LogicalErzor

Nice to meet you! I'm interested in Linux, FOSS, and FOSH, but I'm interested in other topics as well :D

Boosts != endorsements. Opinions own. I treat Mastodon like an interactive RSS feed.

Working on mainlining Linux on the Samsung Expressatt, Pinephone Pro, and the Surface RT.

postmarketOS Trusted Contributor

#fedi22 #Tech #Foss #Linux

Website: https://logicalerzor.codeberg.page/

Public Key

npub1fq87kg24ntac2fesrqmqlk2ka03ere786x4d8whxw48kdf57ferqj2ftuk

NIP-05 Address

logical_error_at_fosstodon.org@momostr.pink

Profile Code

nprofile1qqsysrlty92e47u9yucpsds0m9twhcu3ulrar2knhtn82nmx560yu3sprpmhxue69uhhyetvv9ujumt0d4hhxarj9ecxjmntzg0qth

Publishing to

relay.momostr.pink

Author Public Key

npub1fq87kg24ntac2fesrqmqlk2ka03ere786x4d8whxw48kdf57ferqj2ftuk

Show more details

Last Notes

2026-05-05 08:05:31 UTC

- reply

mind explaining? no idea what the background is here

2026-03-06 08:23:26 UTC

- reply

u might hate this fact then: people switching to linux from windows are primarily using ai to understand what to do in the terminal :P

2026-03-06 08:16:20 UTC

- reply

haha thats fair. u can review the cmds it runs beforehand to audit it. prob useful for those who arent as tech savvy or if u just want it to work
tbh i would never do this on my personal computer as well

2026-03-06 08:05:15 UTC

- reply

also slightly related, but im listening to a lot more audio books now :). its nice to multitask (eg. reading a book and working out). previous tts software just grated my ears and now local cpu generated audio sounds so nice
handy, a local speech to text ai, is useful to quickly type up stuff
anyways, i find it faster to review&edit than to code in one go, so looking forward to local models getting better. useful understanding unknown codebases and to parallelize some work

2026-03-06 07:56:34 UTC

- reply

i found ai does generate better code on FOSS software over proprietary (prob since it is trained on it)
its also useful to understand/prototype on the codebase as well. eg i ask what xyz ticket requires, and it points out good spots (proprietary codebase so its massive and everyone is busy)
another useful thing is letting ai do stuff on ur computer (eg. install xyz program) but i wouldnt trust this too much bc of injection attacks
translation to other langs is good too

2026-03-06 07:49:43 UTC

- reply

it really depends on the model. best rn is opus 4.6 (i will never pay out of my own pocket to scummy ai companies so i understand not wanting to try it)
they get 95% of the way there, and the last tuning part is the hardest
using subagents and letting it iterate itself (especially if u have testcases that it can use to verify itself) is handy
for me, its useful because work and my hobby (which is linux mobile) are diff. i can parallelize my work and enjoy other life too

2026-02-20 15:56:35 UTC

- reply

i wonder if fluxer will consider becoming part of the fediverse :D
i believe they can then piggyback on the e2ee work thats already being done here
https://soatok.blog/category/technology/open-source/fediverse-e2ee-project/
also https://github.com/shazow/ssh-chat is a nice proof of concept of using ssh to chat with others. e2ee encrypted because of ssh :)

2026-02-09 10:51:36 UTC

- reply

expressatt android phone progress - road to mainline linux
got linuxboot working! :)
https://codeberg.org/LogicalErzor/LinuxBoot
tbh using linux as a bootloader makes a lot of sense. uboot and lk2nd is just duplicated effort. a linux distro with kexec tools can work too
benefits: linuxboot now shows the memory holes to mainline linux, something that lk2nd was lacking. can now boot from a sd card :). no more wearing down the emmc
thanks @npub103e…p05x for showing me this :D

2026-02-08 09:31:38 UTC

- reply

expressatt android phone progress - road to mainline linux
someone has gotten wifi/bt working, but i havent been able to test it because lk2nd is giving me trouble
after reading https://www.anthropic.com/engineering/building-c-compiler i wondered if claude could easily get downstream stuff working on mainline linux
after a couple days, i got this commit in: https://codeberg.org/LogicalErzor/linux/commit/6878d1cc1670fc63ba3cba52d4eb08c7a889db17
most of downstream should be ported i think. also vibecoded uboot and linuxboot support
will see if it works when i get the time

2026-02-04 00:50:24 UTC

- reply

expressatt downstream reading
u know how u can connect the pinephone pro to an external monitor and it just works as an external screen? apparently before usb c display alt mode (which is the functionality that makes the feature work) there used to be MHL cables going from micro to hdmi with usb a as an external power source
cursed but somehow i love it lol

2026-01-13 06:51:30 UTC

- reply

#PinePhonePro update: sent wifi, bluetooth, and modem patch upstream
https://lore.kernel.org/all/20260112-ppp_connectivity-v1-0-284d6e83b212@gmail.com/
at this point if you would like to test it, please reply to the patches with a Tested-by: Name <email> as shown in this link
https://www.kernel.org/doc/html/v5.7/process/submitting-patches.html
i think i will get battery/charger next and then switch the pmOS kernel to use mainline Linux. if u would like a feature of the ppp to be included, please let me know and be available to test my changes

2026-01-09 09:18:26 UTC

- reply

is anyone able to test the modem patch for the #Pine64 #PinePhonePro?
i dont have a sim card for the device as its unfortunately not my daily driver. will give u pmbootstrap commands and the kernel tree so u can build it locally to test. can also field any questions if u have any
bonus: u can get credit in the linux kernel for testing the patch too! would need name and email for that
i think ill also try to get wifi and bluetooth working too. phosh seems to recognize the modem with my patch

2026-01-09 03:26:18 UTC

small pinephone pro update
magnetometer, light, proximity, accelerometer, and fixing the volume button patch has been merged upstream
i will be attempting the modem next as it seems to just be dts changes

2025-12-23 05:48:18 UTC

- reply

how i started with linux
with a heavy focus on expressatt :)
https://logicalerzor.codeberg.page/blog/how-i-started-using-linux/

2025-12-22 10:34:42 UTC

- reply

expressatt LLVM issue
turns out lk2nd cant boot anything larger than 21 MB…
welp time to focus on the bootloader now (or make linux-next configs smaller). i want to do linuxboot (using linux as the bootloader) but might go to uboot if thats easier

2025-12-17 10:35:26 UTC

- reply

expressatt mainlining
starting to look like an clang/llvm issue for armv7 devices as i can replicate it on qemu-arm
https://gitlab.postmarketos.org/postmarketOS/pmbootstrap/-/issues/2635

2025-12-14 06:57:19 UTC

- reply

been a couple of months, but finally got around to reporting that rust doesnt work on armv7 if u only specify LLVM=1 on Linux. if u specify LLVM=1 and CC+=gcc, only then it will work
figured this out ~1 month ago because adrianyyy@chaos.social fixed linux-next to only use LLVM (whereas before it was using LLVM and gcc). this seemed to break my setup, and confirmed what was happening when i was externally building the kernel
https://gitlab.postmarketos.org/postmarketOS/pmbootstrap/-/issues/2635#note_520266

2025-12-11 11:18:39 UTC

- reply

expressatt mainlining
and here's the short vibrator blog!
https://logicalerzor.codeberg.page/blog/vibrating-vibrators/

2025-12-08 09:36:48 UTC

- reply

expressatt mainline journey
recently got vibrator up and running and i just have to clean it up and send it upstream. that leaves the following harder stuff left:
display/gpu
speakers/microphone
cameras/flash
wifi/bluetooth/gps (from modem?)
battery/charging
any good readings for the qcom platform? some of this is upstream (i think charging, wifi/bluetooth, gpio expander (camera), and iommu (display) are upstream)
im assuming ill need some amt of background context to get this implemented

2025-12-05 11:42:09 UTC

- reply

expressatt mainlining journey: touchkeys are working now!
https://logicalerzor.codeberg.page/blog/touching-touchkeys/

2025-11-20 09:38:30 UTC

- reply

here's my blogpost on using AI to create linux kernel patches!
https://logicalerzor.codeberg.page/blog/sensing-sensors/

2025-11-17 20:21:08 UTC

- reply

holy moly just got nfc working as well
what literally was taking me months to fail at solving took mere minutes for the ai to solve
this upcoming patch series is gonna be a big boi. really excited to write up a blog once i send the patch series
most importantly, im having so much fun! im not getting frustrated at being stuck and then being ghosted when i ask for help, since there is always something to do and something to look into :D

2025-11-17 07:04:35 UTC

- reply

ai is a huge help ngl. its turning what used to take me weeks to something that i can churn out in a couple of days
once the next stable release comes out, i will blog about all the patches i will send upstream and the usefulness of ai

2025-11-17 06:46:03 UTC

- reply

so multiple updates:
- i think iommu works now! afaik this is mem management for the gpu and a neccessary prereq. try to get adreno 225 working but proving to be a lot harder. luckily (or unluckily) iommu is broken for apq8064 so i can test if my patch works
- got the prox/light sensor working. prox is giving bogus values but light works
- linux-next with just llvm might be an issue? apparently the change i made was using both llvm and gcc, so thats why it was working https://gitlab.postmarketos.org/postmarketOS/pmaports/-/merge_requests/7335/diffs

2025-11-09 07:46:09 UTC

- reply

expressatt: possible compiler bug?
so a while ago i was having trouble cross compiling rust+linux on my computer. the kernel was working when i built it in pmbootstrap but when i built rust+linux outside pmbootstrap and simply told pmbootstrap to use the result, it wouldnt boot
i just realized now that it could be a cross compilation bug within llvm. pmbootstrap uses a chroot (which emulates a arm32 env) and cross compilation is x86
how do i report this? how do i narrow this down?

2025-11-01 16:09:14 UTC

- reply

expressatt accelerometer
easily got the accelerometer working after the magnetometer. was on the same gsbi bus so it was as simple as adding the device and specifying certain things
ngl ai also listed other things that were also already upstreamed in mainline linux, so hopefully i can easily get nfc working with its help now
mount matrix is likely wrong so the orientation of the device will be off

2025-10-27 10:54:03 UTC

- reply

the waiting game works and uart is working again :D
——
i also decided to try to use ai to get a peripheral working. so far i got the magnetometer working but not sure if the orientation is right
gonna write a blogpost about my experience. tldr, ai has a great breadth of knowledge to pick at, but sometimes its confidently wrong which is annoying
i hope to also get the accelerometer working. as its on the same i2c bus. maybe other things like the vibra might be easy to get working too?

2025-10-25 09:56:20 UTC

- reply

once there is any sort of personal connection to any of the parties involved, 3rd parties/external help should immediately be involved imo. the fact that this wasnt done is informative and disappointing
also i hope u are noting all this down and sending it to ur lawyer for litigation

2025-10-11 04:12:57 UTC

- reply

the strangest part is that other msm8960(t), and even msm8226 device owners arent encountering any problems like this

2025-10-11 03:53:55 UTC

- reply

so now apparently expressatt isn't outputting anything over uart in v6.17, but it is for v6.16
bisecting it leads to @npub164g…k0lj's patch https://github.com/torvalds/linux/commit/9f2cbfcda595b6d741adf71058c60476b8d21028
but it doesn't even make sense as this doesnt even remotely touch msm8960 or uart
not sure what to do :/. was gonna test my vibra patch on top of the latest patches... but its seems like im encountering something else entirely
maybe reserved memory or something? i have no idea how to debug this at all

2025-09-24 05:26:54 UTC

- reply

related: clams monitoring water quality in poland
https://desis.osu.edu/seniorthesis/index.php/2024/09/03/in-poznan-poland-eight-clams-get-to-decide-if-people-in-the-city-get-water-or-not/

2025-07-28 07:23:18 UTC

- reply

i wanted to continue my first linux & rust driver but a weird bug keeps biting my butt. essentially:
- generated pmOS img (with rust included in linux-next) works: https://pkgs.postmarketos.org/package/master/postmarketos/armv7/device-samsung-expressatt
- locally built kernel without rust works - locally built kernel with rust doesnt work
local cmds
make LLVM=1 O=.output ARCH=arm qcom_defconfig pmos.config devices.config all
pmb build --envkernel linux-next
pmb build --force device-samsung-expressatt
dunno how to debug this :/

2025-07-25 08:10:39 UTC

whoever thought of the name Evergreen [Linux] (would prefer if it was called EvergreenOS to come up in search engines) is a fucking genius
multiple meanings (atleast for me):
- evergreen trees usually grow on mountains (referring to alpine)
- green: envokes the idea of recycling and a nod to pmOS colors and roots
- evergreen trees are green all yr round… meaning we dont want to give up on old forgotten devices
*chefs kiss*
i really hope this is the new name

2025-07-01 10:33:36 UTC

if distrobox manages to run windows apps in a container, it would be a great stepping stone towards migrating windows users over to linux
issue: https://github.com/89luca89/distrobox/issues/1279

2025-06-27 04:15:29 UTC

seems like i can only go back to about 3 days in a list? was out for a couple days and wouldve liked to see updates from before that
this is using the fosstodon website as well. unless im hitting a limit of some kind?
cc @npub1z3w…c039

2025-05-19 04:43:52 UTC

- reply

oh yeah, i also hit this bug when trying to build against an Alpine Linux pkg:
https://gitlab.postmarketos.org/postmarketOS/pmbootstrap/-/merge_requests/2601
big thanks to @npub1jdw…mdn3 for solving this!
it seems like it's currently unknown why find() is returning none when it shouldnt. mypy isn't complaining about it either
#rustlang probably would have prevented this bug (sorry i couldnt help it :P)

2025-05-19 04:25:45 UTC

- reply

ill prob be working on my first driver soon. it will be based on this:
https://codeberg.org/LogicalErzor/Android_Kernel_Samsung_D2/src/branch/cm-14.1/drivers/motor/tspdrv.c
if anyone has any suggestions on how i can convert this into an upstream driver, im all ears! noobie advice is welcome :)
most of my mainlining time is reserved for weekends, so i would really appreciate tips on figure out how to be faster as well :D

2025-05-19 04:16:40 UTC

- reply

if you have a spare phone, you can check https://www.gsmarena.com/ to see what the SoC is
if it's one of the following...
APQ8064
APQ8060A
MSM8660A
MSM8260A
APQ8030
MSM8930
MSM8630
MSM8230
MSM8627
MSM8227
MSM8625
MSM8225
IPQ8064
MSM8960
...consider joining us!
we have a matrix chat: https://matrix.to/#/#linux-apq8064:matrix.org
doesnt matter if ur totally new too :)
if it's not on the list, u can check https://wiki.postmarketos.org/ to see if there are others with the same SoC and join them

2025-05-19 04:07:47 UTC

- reply

a lot of cleanup has happened since this time. some notable things
https://gitlab.postmarketos.org/postmarketOS/pmaports/-/tree/master/device/testing/device-samsung-expressatt -> useful b/c i can point other msm8960 devices owners to get their device up and running and it publicly documents a working kernel (one has already mainlined with this!)
https://wiki.postmarketos.org/wiki/Samsung_Galaxy_Express_SGH-I437_(samsung-expressatt) -> revamped my page to help others mainline and document tests
https://wiki.postmarketos.org/wiki/Qualcomm_S4/Krait_200_mainlining -> wctrl and i decided to merge the mainlining guide for the entire S4 family to encourage others to join

2025-05-12 05:32:03 UTC

- reply

the tech market is kinda bad rn (layoffs.fyi shows the extent of it). if i were u, i wouldnt wait to get some sort of stable income and would start searching asap
hopefully someone on #FediHire can work something out for u. im sure there are a lot of folks willing to hire someone of ur caliber
maybe u can do vtubing on weekends? hopefully it doesnt mean the end of something u love to do

2025-04-28 15:26:21 UTC

- reply

as one of the users who wanted to see the human, thanks kev. i appreciate what you’ve done to help fosstodon and the entire fedi grow
if there are needs for modding, where can i apply? it seems like the person linked in ur blog left for hachyderm, but in the contrary im happy to stay and help if needed

2025-03-19 06:21:43 UTC

- reply

checking ur dts on msm8960 platforms should now yield no errors!
https://lore.kernel.org/all/20250318-expressatt-solve-dts-errors-v1-1-14012a4bc315@gmail.com/
edit: forgot to thank @npub1shv…wgvr for showing me the patches for apq8064 (msm8960 and apq8064 are very similar, enough so that i can copy some patches from there and it will “just work”)

2025-03-13 13:18:02 UTC

- reply

seems like most people dont mind going matrix-only (me included)
might be worth having a wiki page on how to setup ur own bridge/list of community bridges irc users can connect to
not sure if thats possible because ive barely used irc tbh

2025-03-03 00:43:00 UTC

- reply

was getting a bit annoyed that when sending patches upstream, i would usually go through the same commands when submitting patches upstream. thought it would be better to make an interactive session so introduced `b4 flow`:
https://lore.kernel.org/tools/20250302-flow-v1-1-3247c022bbac@gmail.com/T/#u
Based on:
- https://www.youtube.com/watch?v=QBR06ml2YLQ
- https://www.youtube.com/watch?v=-Vld9aoFPwU

2025-02-27 06:16:11 UTC

- reply

man, some linux maintainers’ comments rival the squiggles on doctor’s notes
u dont understand what they try to get at

2025-02-25 11:04:16 UTC

- reply

thanks to wctrl’s help, ive gotten tsens working for the msm8960!
https://logicalerzor.codeberg.page/blog/increasing-the-temperature/
nfc is super annoying. dunno why even tho its matching downstream it refuses to work

2025-02-10 06:16:16 UTC

- reply

Blog!
https://logicalerzor.codeberg.page/blog/lk2nd-and-shells/

2025-02-08 10:25:02 UTC

- reply

oh my fucking gosh it actually boots into a shell!!!!!
im so happy!!!
i was considering abandoning this project entirely just a day ago and i cant believe this is working now. holy shit
HUGE thanks to Sam… wow without him this would be dead in the water
i will probably make a short blog about this!

2025-02-05 17:17:55 UTC

- reply

main problem here is https://pastebin.com/qGbfFGUt
small rw’s are fine, but big ones hang the system: https://pastebin.com/Tkx7Bw0q
the person trying to help me couldn’t figure it out either
at this point ill try for a bit more before just giving up on this unfortunately

2025-02-04 08:45:18 UTC

- reply

Hi! Thanks for making dowhile0. It’s come in use so many times while i am trying to mainline my phone
i have a problem thats been confusing me for a while: my emmc and sd card can rw, but it seems like if there are too many ops at once it hangs. do u know a way to debug this? there is downstream but it uses board files
as u can see above, ive been thinking this is a memory problem for the longest time. any information for this newbie is much appreciated. thanks!

2025-02-01 12:33:19 UTC

- reply

wow, the pmos community has changed, and for the better too
huge thanks to sam:samcday.com on matrix. he basically sat and helped me work thru lk2nd (bootloader) bugs, until i actually got it working! prob spent 8 hours at least just helping me
unfortunately i havent solved the above issue yet, but people in the chat seem to think its not a memory problem but maybe a config one. i need to play around with some of the suggestions and see if any work for me

2025-01-31 09:03:58 UTC

- reply

decided to give mainlining a go once again, but hitting memory issues. decided to make a pmbootstrap issue about it: https://gitlab.postmarketos.org/postmarketOS/pmaports/-/issues/3512
not too sure how to proceed. asking online gets me no answers. i don't know anyone who can help me. just frustrating being in a dead end...

2025-01-27 08:24:34 UTC

@npub13fk…z5hw for some reason i cant follow [@npr](https://flipboard.com/@npr ). keep getting 404 record not found when trying to follow and then put them in a list. refreshing the page shows that i seemingly unfollowed them

2025-01-25 10:36:58 UTC

- reply

I do want to try and give this one last shot
Let’s meetup at Barebottle Beer Garden at Salesforce Park. It’s also close to the Embarcadero Station in SF. Would people like to meet here at 2 PM on Feb 1st?
Other places/times/topics are welcome. Just wanted to hangout with people interested in FOSS 😊
#MobileLinux #Linux #FOSS

2025-01-18 01:43:09 UTC

- reply

Unfortunately, it seems like there is only one person interested so I’ll go ahead and cancel it. Was worth a shot!
#MobileLinux

2025-01-15 06:33:52 UTC

- reply

We can meetup at Barebottle Beer Garden at Salesforce Park. It’s also close to the Embarcadero Station in SF. Would people like to meet here at 1 PM on Jan 20th (MLK day)?
Please only star/like this if you intend to come (I won’t track boosts so feel free to go ham lol)
Other places/times/topics are welcome
#MobileLinux

2025-01-14 11:43:56 UTC

Would anyone be interested in meeting up in SF/Bay Area about #MobileLinux? I’m trying to gauge if there is interest like the person doing the meetups in New Zealand
I’m not entirely sure what we would do, so suggestions are welcome. I was maybe thinking to hangout near a BART station for accessibility purposes

2024-10-03 16:29:00 UTC

- reply

Hm, I do agree with summarizing the outcome for others to catch up on, however it might be better to have software that groups chat up in a certain way? Like Zulip or Codeberg issues
Problem with Matrix/Discord is that you get repeated questions because 1) search isn’t good 2) there are no “threads” to limit a conversation to one topic

2024-09-25 07:06:58 UTC

- reply

He’s dead now. The death penalty should be banned.
https://edition.cnn.com/2024/09/24/us/marcellus-williams-scheduled-execution-date/index.html

2024-09-19 17:18:23 UTC

- reply

This is cool! Would you know how this setup is different from https://mailinabox.email/?

2024-09-19 17:04:48 UTC

- reply

I can’t wait until the day when office buildings are converted into homes because of the economics of WFH

2024-08-25 21:13:45 UTC

- reply

Ended up making a mainline builder script that just has Linux and busybox: https://codeberg.org/LogicalErzor/mainline_builder/src/branch/master
No root needed (for now), and this is all built locally. I'll probably expand on this to include more tools to help me mainline
This should hopefully help me use i2cdetect and other tools. I can easily kexec into another kernel now too to save my eMMC from being written to oblivion and also hopefully have more space to enable configs

2024-07-01 06:42:03 UTC

- reply

A bit irritated here… Here’s what I have to do to figure out why my i2c gpio devices aren’t working:
- Add i2c-gpio devices in dts (Doesn’t work)
- i2cdetect (Doesn’t work in debug shell)
- Enable configs to boot Linux (Doesn’t fit in partition)
- u-boot (Stuck trying to get UART output)
- Use Kexec instead of u-boot (pmOS now thinks my phone is ARM64)
Anyone know how to write an ARMv7 UART program in assembly? UART works in Linux, but frustrating not knowing what I’m doing wrong

2024-06-25 03:56:34 UTC

- reply

So I kept enabling more and more linux configs, and now it doesn’t fit in my partition anymore :/
I knew I would have to do this eventually, but I guess nows the time to add armv7 support to u-boot. Luckily it doesn’t look like a lot of work since I’ll be building on top of Caleb’s work, but I also don’t know what I’m doing lol
I could modify lk2nd to boot from another partition… but meh, seems like it’ll be smarter to get u-boot instead long term

2024-06-16 09:20:02 UTC

- reply

So I got telnet working in debug shell (postmarketOS now uses NCM rather than RNDIS and seems like there’s an upstream USB regression as well) but still struggling to get telnet/ssh after debug-shell
I think I’m supposed to wait till #OpenRC boots up all the way for ssh access? Got through some problems (enabled UNIX config and increased ramdisk offset)
The current problem I’m facing is OpenRC hanging on “SVCNAME=dmesg”. /etc/init.d/dmesg is a simple script. Maybe missing kernel configs? Dunno

2024-03-18 05:33:29 UTC

- reply

Hm, have a weird ssh/telnet issue going on.
telnet works in debug-shell, but neither telnet nor ssh work after booting past that
Doing https://wiki.postmarketos.org/wiki/USB_Network#SSH_issues makes telnet stop working in debug-shell
I have no other way to get a console. openrc throws errors regarding udev, logging, and clock skew, but i don’t think any of these should affect ssh, right?
Banging my head against the wall trying to understand and fix the problem here

2024-03-12 07:09:11 UTC

- reply

Now trying to test my i2c stuff. I have two nodes in the dts: NFC and touchkeys
Touchkeys doesn’t want to work, even though the dts looks the same as the boardfile
Trying to get NFC working, but neard doesn’t run in debug shell. Having trouble figuring out how to add the neard executable into debug shell or sshing/telneting into my phone without a debug shell
If my phone can detect NFC that’d be great. I’d then know that the device node was bad. Otherwise back to the drawing board

2024-02-18 01:38:13 UTC

- reply

The downstream boot.img stored it in a location close to the kernel. So when I added more configs to the kernel, it decompressed beyond the kernel’s size and overwrote the ramdisk that was in memory
I increased where the ramdisk is stored in memory, and I no longer get the initramfs error
Sucks that it took me this long to figure out. I wish postmarketOS had a forum so that a question can be searchable via a search engine. Matrix search sucks

2024-02-18 01:35:40 UTC

- reply

Randomly figured out what was causing the initramfs problem. The bootloader that’s built into the phone can only boot Android boot.img. Contained within it is a kernel and a ramdisk
I never paid much attention to the ramdisk, since I’ve only modified the kernel, but turns out that the ramdisk is the initramfs. When running mkbootimg (this is to create a boot.img that my phone can boot), you specify where the ramdisk is stored in memory

2024-02-13 09:05:08 UTC

- reply

There are around 600 remaining configs that I have to go through in the downstream config. I can get an accurate understanding on what’s mainlined already (a lot of what I’ve done so far is just enable mainlined functionality, which is nice) and what feature I need to create drivers for. It’ll also help me understand what features there are in my phone as well

2024-02-13 09:01:30 UTC

- reply

Hm, it seems I can boot now, though I don’t know why. I saw that a bunch of TV configs were =m so I removed them. It started working after that. I also modularized a bunch of configs. If I were smart, I would’ve saved my config progress and tried to find the config that causes this issue. Would most likely be a false positive though
Now that I got it booting again, I can test both NFC and touchkeys. Seems like I’m still having trouble with i2c bus :(

2024-02-06 04:25:53 UTC

- reply

Argh, I’m hitting the initrd error once again. I switched over to try and get nfc working on my device, but reenabling the networking stack seems to make my device panic :(
Seems like I have to spend a lot more time on figuring this out, at least to get it booting again
The problem: https://wiki.postmarketos.org/wiki/Qualcomm_Snapdragon_S4_Plus_(MSM8960)

2024-02-03 07:18:31 UTC

- reply

NFC support is also in mainline, but it’s also on its own I2C bus. So I have to figure out this issue first before I can add these two peripherals to my device

2024-02-03 07:16:47 UTC

- reply

Hm, struggling with getting an i2c-bus up. Seems like touchkeys isn’t on GSBI10. Looks like Linux wants to add 512 to the GPIO for some reason?
Eg. when I specify that I want to have an i2c-bus on GPIO 71(sda) and 72(scl), the logs says “using lines 583 (SDA) and 584 (SCL)”
Don’t know why it’s doing this or how to solve this :(

2024-02-01 04:39:23 UTC

- reply

My touchscreen blog is up!
https://logicalerzor.codeberg.page/blog/mxt224s-touchscreen-kernel-dev/

2024-01-29 23:08:18 UTC

- reply

Woo! Got touchscreen working :)
Gotta make another blog (first gotta refactor my blog to use Hugo) and send patches upstream
Probably will try to attempt the touchkeys again while attempting the vibrator and tsens as well
Some things missing:
- display
- speakers/mic
- cameras/flash
- sensors (proximity, light, etc.)
- gyroscope, accelerometers
- wifi/bluetooth/gps (functionality provided by modem?)
- nfc
- battery/charging
This is what I know of, but there might be others that are missing

2024-01-27 08:30:31 UTC

- reply

I almost got the Atmel touchscreen working today. Seems to randomly work and not work, but I don’t know why. Or maybe it’s fooling me into thinking that it’s working, I don’t know
I think I figured out for certain that the touchkeys don’t work because they aren’t on the correct i2c bus. It seems to be on gsbi10 by the looks of it, but there’s no data sheet to be found unfortunately
Might try to get the vibrator and tsens working. Hopefully low hanging fruit

2024-01-26 08:32:50 UTC

- reply

Hm, I think I might have to hunker down and just read John Madieu “Linux Device Drivers Development” cover to cover. This is a bigger uphill climb than I realized since it’s not just driver development, but also figuring out what features are missing in the upstream SoC itself and implementing/enabling that. Searching downstream code is hard
I’ll use this time to get a setup a better neovim development environment and also get my blog sorted out too. Hopefully I can get through this quickly

2024-01-25 05:14:42 UTC

- reply

Currently a bit stuck of touchkeys. There’s a mainline driver thats extremely similar to what’s already been done, but for some reason it’s just not working. My hunch is that the sda/scl (the communication lines) are messed up or maybe the dts is, but everything looks right to me.
I’m just going to switch over to trying to enable the touchscreen. Hopefully that is easier?

2024-01-20 12:11:21 UTC

- reply

And here's my post on gpio-keys! I'm thinking about switching to markdown cause manually editing HTML was not fun
https://logicalerzor.codeberg.page/content/gpio-keys-an-intro-to-kernel-dev.html

2024-01-20 08:57:31 UTC

- reply

Kinda was dreading sending emails upstream, but I found a very friendly b4 video: https://www.youtube.com/watch?v=QBR06ml2YLQ
This is a very well done video

2024-01-20 08:56:59 UTC

- reply

Got gpio-keys working! 🥳 Going to send patches upstream, make a blog post on how I approached the problem, and then work on cypress touchkeys next
Huge thanks to @npub1a02…4rre for helping me! It was great to see how he figured out the problem

2024-01-19 09:23:46 UTC

- reply

Quick updates:
- Pivoted to GPIO keys to understand the DTS more. Very similar to touchkeys. evtest doesn't work so DTS is wrong :(
- Couple hours wasted only to find out $builddir was set wrong
- Need to setup LSPs for neovim. It's slowing me down trying to grep for things. Using kickstart.nvim
- In downstream kernel, converted all configs that were in defconfig to true. Saves time just looking at the source code and not referring to the defconfig
- Building kernel on PPP is slow :(

2024-01-17 08:26:31 UTC

- reply

There seems to be support for firmware upgrades, touchkey calibration, and raw data support in the downstream driver, but I’m unsure if that will be necessary in the mainline kernel. IIRC the tm2 driver doesn’t have that support
Since I’m essentially the only user of this phone, I think I’ll put that in the backlog for now

2024-01-17 08:23:04 UTC

- reply

Awesome, after a nice holiday break and a workflow upgrade detour, I finally got back into kernel development
Looked at mainline linux and downstream for a while and figured out a few things
- touchkeys are a simple way to start
- cypress,tm2-touchkey is a generic mainline driver that I should put my touchkey support in
- board-<machine> is essentially the dts file. need to manually parse it to convert it into a modern dts
After enabling the touchkey and led regulator, the led turns on!

2023-12-19 04:42:03 UTC

- reply

Cool, got a minimum kernel where all the .{c,s,S,h} files are all the important ones. Initally I didn't want to make a parser like I originally stated, so I attempted to brute force it. Tried various algorithms and doing full rebuilds (partial rebuilds didn't recompile everything), but it would take a couple days
Finally got a way to output all dependecies:
- remove `rm -f $(depfile);` (*.d files contain dependecies)
- run `make` with `-d` (I ran it on --debug=implicit)

2023-12-14 06:59:50 UTC

- reply

While compiling and doing other things on my computer, it started to thrash. Guess it’s starting to show its age!
I like old computers like this though. Even though it takes time to compile, I feel like using older devices is nice for the reasons below (no particular order)
1) Sympathizing with those with minimal computing resources
2) Environmentally friendly
3) Decrease consumerism
4) Reduce impact in slavery (eg. DRC)
5) Emphasizes lighter software
That’s why I like postmarketOS so much!

2023-12-14 06:50:10 UTC

- reply

Before this I was trying to get the framebuffer working, but my device has its framebuffer disabled. I’ll have to setup a display driver, but searching through the code was a pain, so that’s why I’m trying to reduce the downstream kernel to the bare minimum
Bringing up i2c and then working on touchkeys is a good first step. I don’t know the i2c subsystem well and touchkeys seems simple enough to understand. Lots of other peripherals depend on i2c, so this will be good to test and learn things

2023-12-14 06:43:47 UTC

- reply

To do that I’m going to add KBUILD_VERBOSE=1 and then build it. It should output all .c files it needs, and I should be able to just delete the other .c files
Then I’ll probably go through the .c files and get all the .h files needed by it. Those .h files might be dependent on other .h files, so I’ll prob have to recursively iterate that. I’ll prob ignore <> since that’ll be included with -I? idk
In the end I should have a minimal kernel? Need to test this theory
I’m having a lot of fun :D

2023-12-14 06:37:32 UTC

- reply

Thought I would build my downstream kernel for fun. Downloaded 16.06 Ubuntu, loaded up gcc and gnueabihf version 4.9 on it + other necessary build tools, and passed the appropriate ARCH, CC, HOSTCC, and CROSS_COMPILE variables to make. Recent versions of Ubuntu couldn’t get gcc4
I think it’ll be useful, since there are a lot of files that I don’t understand in the downstream kernel. I’m going to remove files that aren’t necessary to builds and then run ctags

2023-12-13 04:35:06 UTC

- reply

I thought I could get around flashing constantly by doing the USB and UART combo. I could then use “fastboot boot” instead to directly load it in memory
While rigging up the USB part above, it was interesting to note that I had to twist the D+/- wires! Without it, my computer couldn’t read my phone’s output. It was cool to see electromagnetic fields cause corruption on USB
Another strange bug I encountered was that turning off my laptop’s WiFi helped the Pi connect. Unsure why

2023-12-13 02:04:31 UTC

- reply

Lol, Raspberry Pi Bookworm’s networking doesn’t work out of the box on the Zero W. Downloading the 2020-02-13-raspbian-buster-lite.img seems to work so I’ll just stick with that. Finally got my UART cable working
Tried to be smart and hook up a USB cable and a UART cable to my phone, but it doesn’t work. I think for debugging, I’ll have to flash the emmc. Hopefully I can use lk2nd soon so i don’t wear out the emmc
Now time to debug why enabling a lot of configs crashes the kernel

2023-12-12 06:06:20 UTC

- reply

Hm, anyone have a way to setup a headless/USB gadget on the #pi zero w? For some reason my pi won’t connect to WiFi even though I have the SSID and password correct. Tried USB gadget, but no matter what recent guide I follow, I get “connection refused” even though I enable ssh on my pi through the installer
In the past, we could modify WiFi config via a file on the sd card, but seems like they removed that functionality
If this doesn’t work out then I’ll just make this: https://wiki.postmarketos.org/index.php?title=Serial_debugging:Cable_schematics#microUSB/Carkit_debug_cable

2023-12-07 11:45:18 UTC

- reply

Gah, trying to spend a couple of days on this and cant figure it out
I reduced my kernel config just in case I had to get around this for now, and it looks like I do. Weird thing is that my device seems to get hotter, but oh well. I at least know the configs to disable now if I need space
I tried making /proc/iomem look like downstream, but it’s still failing to boot
At this point I’ll have to setup my UART cable again and debug via KGDB
I just want to start driver dev tho :(

2023-12-01 08:47:41 UTC

- reply

Hm, adding too many configs causes the device to not boot because the kernel takes too much space? My theory is that while the kernel is decompressing, it overwrites itself causing memory corruption. Not sure how to validate this theory or how to solve it
This happens on lk2nd and the stock bootloader, so I’m guess it’s something with Linux. If I can’t figure out how to debug this, I’ll first strip my current config to it’s bare minimum and then add in downstream configs that exist in mainline

2023-11-30 07:03:25 UTC

- reply

android_kernel_samsung_d2$ cat ./arch/arm/configs/cyanogen_expressatt_defconfig | grep -v "#" | sort | uniq
Configs enabled in downstream kernel. Todo: see what the differences are and upstream

2023-11-30 07:00:15 UTC

Modify the kernel and deploy it (for future me)
cd mainline-linux
. pmbootstrap/helpers/envkernel.sh
cp defconfig arch/arm/configs/expressatt_defconfig
make expressatt_defconfig
make menuconfig
make -j8
deactivate
pmbootstrap build --envkernel linux-postmarketos-qcom-msm8960
pmbootstrap initfs hook_add debug-shell
pmbootstrap export
fastboot boot /tmp/postmarketOS-export/boot.img
sleep 40
telnet 172.16.42.1
lk2nd should be installed in BOOT
lmk if there's a better workflow #postmarketos