elly

*oven ding* it's done Took me a while to finish this because of life-related reasons, but finally sat down and fixed (most) remaining issues during the weekend. Pretty much everything* is working - including S3 suspend/resume, automatic fan control and so on (plus it draws less power than stock, I have no idea what AMI is doing). Feels kinda surreal to have a fully-functional computer If you have an "MSI PRO B760-P WIFI DDR4" motherboard (or "MSI B760 GAMING PLUS WIFI DDR4", pretty sure they're the same just with different UISetup skin) you can now flash #coreboot on it. DDR5 versions can be easily added if someone is willing to work/test it, just a matter of telling FSP you're using a different memory type. Will document stuff (like making flashing cable that won't void warranty) later. Enjoy https://review.coreboot.org/c/coreboot/+/93127

Ended up hacking on Lynx R1 last weekend... and at this point I'm wondering if I'll make it work before Steam Frame hits the market

We now have:

1. U-Boot (Tauchgang) 2026.04
- Pretty much everything working except for UFS
- According to logs I need to write SM8250 interconnect driver for that

2. Linux 7.0.1
- sm8250-mainline tree was an unmaintainable disaster and it's abandoned
- Straight from kernel.org with not-yet-upstreamed patches for pm8150b-charger, devicetree for this device and initial panel driver
- Working WiFi/BT, UFS, MicroSD, USB3.1 Gen... uuuh, whatever SoC supports, charging (but it's flaky)

Some stuff that I need to figure out:
1. Possibly mark boot as successful with qbootctl, currently we get 6 boots and get dropped into fastboot where we need to re-flash u-boot.
2. How to make fan work, charging currently stops after ~15 minutes because SoC reaches 75*C. It also doesn't negotiate higher PD profile (5V/3A max) so that needs more work.
3. Decode downstream DSI init sequence, as DSC is kicking my butt despite having display datasheets.
4. How the heck volume buttons are connected, they're not controlled by GPIO nor PMIC :blobcat_bonk_googly_large_trout:

Pretty sure we can get audio and cameras working as well, need to pull files for hexagonrpcd to serve from Android, as currently CDSP/SLPI are crashing (though not sure if we can use hand-tracking model running on Compute DSP in mainline, probably not).

Me: "Okay, time to put this into drawer and revisit it when I'll have more time" My brain: "So... you're just going to let the device win?" Me: "...hold my tea"

@whynothugo Assisted-By: Metylphenidate (CR, 20mg)

@pomagarnet@chaos.social nah, @famfo@frogs.lgbt is likely dealing with “unable to change state from D3cold to D0: device inaccessible”, followed by controller reset. Usually it means either PCIe mess or (very rarely) Linux regressions. In the worst case it means that your drive is dying… I just checked my work drive and damn, I should give it a raise or something, it’s pulling overtime: Data Units Read : 30213998 (15.47 TB) Data Units Written : 334273399 (171.15 TB) power_on_hours : 17428

@radex@social.hackerspace.pl Not fair, that’s a personal attack

@Edent If they're afraid of "AI hacking", they should get rid of legacy hardware/software first. Like, seriously. In 2023 NHS was still running Solaris 8 in production on Sun Blades 150 (workstations dumped into racks) and Ministry of Justice was running Tru64 UNIX on (emulated) DEC Alphas :neocat_facepalm:

@Umbreon that’s a pretty good challenge though :akko_thonk: Rescue 30 computers (or more!) from ending up in a landfill Probably have something close to this amount in my apartment/basement (most of them either gifted to me, or found dirt-cheap with broken screens to mainline/use in CI) so… check? Upcycle 30 phones with a free mobile operating system That’s a real challenge, I only ported Linux to ~4 phones so far and don’t think I have enough time to do so before I turn 30 (just 2 months after KDE… yes, I’m old) Clean up 30 hectares of woodland Actually good idea, I live in the forest and local community is organizing to clean it up next week Convert 30 people to a free operating system Only managed ~16 people so far, gotta step up my game Take 30 techbros to court so they stop building AI datacenters Uhh, I don’t think we have enough money to counter techbros unfortunately

@jn @domi @q3k pretty sure if I would go "screw this, I'm out", I would go into dressmaking. That was my childhood dream (shout-out to my mom's soviet-era sewing machine, that thing was a tank) until capitalism made me realize it wasn't a profitable business model

@ariac@donotsta.re doesn't seem like the same phone though

@whynothugo I believe so, Concerta never quite worked for me, neither did the regular (ER)

@whynothugo Constant Release

@mxk@hachyderm.io > Japan has a Hello Kitty Shinkansen Not anymore :sadcat: https://www.jr-hellokittyshinkansen.jp/

I have to say, this new MacBook Neo is really pretty

@domi @aelin %domain%

@Remiberry@donotsta.re especially the palmrest, I hate the feeling of cold aluminum

Bundenspolizei dillema: Is it a #postmarketos developer, or terrorist with a bomb? Wires sticking out of the phoneWall of text scrolling by after pressing the power buttonPhone making weird noises (In case anyone would want to play with this device at #gpn24, please do not squish the right side of the case, while it should be safe, I cannot guarantee you wouldn’t puncture the battery if you would go “full ham” on it)

I really need to know where folks are getting those smol neocat plushies, they're way too cute

@Curve25519@infosec.exchange we have pretty good toolkit for reverse-engineering the hardware, often you can also find leaked schematics (especially if it's a laptop). Worst case you arm yourself with oscilloscope/logic analyzer, multimeter, soldering iron and some free time. It's (usually) fun

This might be a hot take, but I will never understand why people new to hacking on mobile/embedded devices insist on trying to use vendor-provided sources.

Even if you get a kernel source from $vendor (which is not a given, welcome to the magical land of GPL violation) you will run into numerous issues even trying to compile it, only to end up with a pile of legacy crap that will never* work correctly with upstream userspace (unless you resort to using Ubuntu Touch or Droidian, which AFAIK use halium and can use Android HALs).

Seriously, if you're new to this and want to learn how to do it *properly* on a completely unsupported SoC, you should:

1. Dump downstream devicetree from running device. It's much easier to understand than vendor sources once you understand how it works: `dtc -I dtb /sys/firmware/fdt -O dts > /storage/emulated/0/Downloads/dumped.dts` (and pull it using adb).

2. Locate UART pins on your device (beware that most ARM SoCs use 1.8V or under) and what's the address for that (for example on MT6789/MT8781 it's `0x11002000`). Using simple-framebuffer is a last resort but you really want functional UART for early bringup.

3. Write initial upstream devicetree and minimal driver set. You more-or-less need the following to get into initramfs:
- Clocks
- Core cluster definition (in DT)
- Possibly reserved-memory (memory regions that shouldn't be used, often used by modem/wifi, gpu, remoteprocs etc)
- Interrupt Controller (usually ARM GICv3 unless it's Apple, then AIC)
- Pinctrl
- UART
*(Hopefully not forgetting about anything?)*

4. Go from there (drivers for interconnect, i2c/spi etc).

Using kernel trees for the same SoC (even if it's a different vendor) is a protip, it's all $soc_vendor's BSP after all. Or if you're in really deep, throwing kernel modules at ghidra is also an option

First time in Netherlands

One of the first impressions:

• Look, this guy is overtaking tram on a bike
• Something something weakest Dutch cyclist or something

@nicolas17 well :neocat_laugh_sweat:

@famfo@frogs.lgbt @pomagarnet@chaos.social same, I'm also a filesystem "boomer". XFS has zero advantages if you're not using it for databases, BTRFS is just unstable, ZFS is out-of-tree and so on... EXT4 is still fastest/most reliable Also, is it a Smasnug drive by any chance?

Welp, after spending 3 days and nights debugging my workstation I can say with 100% certainty that all issues I've been having recently are caused by silicon degradation.

First it was PCIe, then random (sporadic) lockups every ~month or so, then issues with USB controller, then iGPU and then lockups became so frequent that it became unusable (2 - 3 per day). AMDGPU issues might even be related to this.

Cleaned the system and put it together, but I expect it to crash while I will be sleeping... so I guess it's time to look for a normal motherboard and CPU (thankfully I can get used MSI DDR4 board which doesn't have BootGuard and 14Gen i5 for roughly ~300EUR and it's going to be 20% faster than the current setup as well).

This concludes a 3 year-old experiment, if you're thinking of buying cheap pre-production hardware that fell from the back of the truck in China: Don't. It's funny at first, but save yourself the headache.

EDIT: It already crashed, it lasted 2 hours and 44 minutes

Just got another funny/silly idea... which one would be more cursed?

@IngaLovinde craziest thing I've seen people transporting on U-Bahn was a fridge, but it did fit without any issues :meowgiggle:

Another month, another shitty behavior from Bambu Lab... and at this point they're just *asking* to get hacked (I guess they haven't heard of the "don't piss off the hackers" rule).

Apparently they sent a DMCA to someone working on cloud integration in OrcaSlicer. Not something I would use, but it just... pisses me off (especially because I'm waiting for A1 to arrive myself).

Started looking into it, based on pictures I see:
Mainboard:
- Spintrol SPC2168 (Dual-core Cortex-M4)
- ESP32-S3
- GigaDevice 25Q128ESIG
- 4x stepper motor controller under the heatsink?

TH board:
- Spintrol 50KB (Cortex-M4)
- 2x ZHONGKEWEI AT8236

Looks to be similar to other models made by them, as documented under:
https://github.com/opensourcemanufacturing/OpenBL/

While ARM MCUs have OTP, worst-case they can easily be replaced with STM32 equivalent thanks to LQFP packages.

Firmware on bambu's website looks to be encrypted (lol, lmao), but dumping it from device in hand (when I'll receive my printer) should solve that problem.
Those files also are self-describing if we look at the board and amount of flash space available:

- AMS: Didn't look into it (and didn't buy one), but it's under 100KB so AMS likely also has a SPC1168APE MCU in it
- AP-ES3: Juicy, easy target. It's 4.9MB in size, so it must live in the SPI flash and store firmware that ESP32-S3 is running.
- AP-HMS: Based on image files in the folder and 700KB size, this is likely firmware that SPC2168 is running and controls the display in front of the device.
- MC: Purely a speculation, maybe for the built-in camera?
- N3F/N3S/N2S: No idea, possibly an NFC chip?
- TH: 50KB, running on SPC1168APE

Pinout can be beeped-out, all you need are GPIOs for motor controllers anyway. Display would be a bit trickier, but oscilloscope solves that problem just fine. Besides, the MCU doesn't have MIPI or anything of sort so it's likely just an SPI for display and I2C/SPI for touchscreen

OH:
- Does alpine not have nested virtualization for it's test runners?
- What now runners?
- Does alpine not have testing infra?
- They do (it is @fossdd@chaos.social laptop) /j

@nicolas17 @yama @codemonkeymike @paulywill this, most modern machines use NVRAM for variable store. You can't reset it by just yoinking the power. Not sure how it's done on T2-based x86 (assuming T2 acts as ROT), x86 itself isn't fused so firmware isn't tamper-protected but it could be done by T2 (from what I remember, T2 emulates SPI to the x86 host and actual x86 UEFI lives in dedicated portion of an "SSD". T2 should be vulnerable to checkra1n though, so it should be possible to fool the ROT and at least modify NVRAM variables to change security policy but it would require some research.

@moses_izumi @domi I've been procrastinating making a blog for 5 years :throwing-hands-up-crying: Deploying nginx, getting familiar with hugo (and picking a theme, so many options...) or writing something by hand, protecting against LLM scraping these days just sounds like a lot of work

@Curve25519@infosec.exchange shameless self-promo media.ccc.de/v/38c3-corebooting-intel-based-systems

@mimimitsumii@donotsta.re @Remiberry@donotsta.re it's even weirder because board layout and datasheets for mosfets suggest that center pin is a GND, but according to wiring and input it's +20V :acat_huh: (And yes, PSU is center-positive, PCGA-AC19V)

Does anyone know how to match thermopads? Of course GPU vendors never provide this information...

Back in 2024 I bought the fastest, most expensive card I ever had (~500EUR), it started crashing last summer and memory was idling at 70*C. It got better during the winter, but recently (before my motherboard died) it got worse.

Never had to replace thermopads before (even on 10 year-old cards they usually were fine), so I'd appreciate any tips.

I'm currently assembling a new workstation, thanks to friends ( @BluRaf@donotsta.re, @TadeusTaD@donotsta.re, @Remiberry@donotsta.re, @dragoonaethis@mstdn.social ) I was able to afford an LGA1700 DDR4 motherboard (there's no way I'm buying DDR5 in current market conditions even if it costs a fair bit of performance) and i9-14900K which will also be ~50% faster than Erying with ES i9-11980HK that... disintegrated.

Over the past 2 weeks I tested all components from my (dead) workstation, with following results:
- Heatsink: Bent in shipping (thanks LaPoste), will use a publicly-accessible vice in OBI to strengthen it. It should be able to handle this CPU.
- Seasonic 620W: Works perfectly, dusted it off and swapped a new fan in
- 32GB of DDR4 I bought back in 2023: Works without XMP enabled, but instability at 3200MHz might be caused by either bad board design or heatwave.
- SSDs/HDDs: Perfectly fine, I got all of my data uncorrupted.
- GPU (Sapphire RX7800XT Pulse): Overheats even at idle (bad fan profile), fan control doesn't work (even though it was supposed to). Opened it up to find completely dry thermal paste and thermopads crumbling if touched.

Measuring current pads (obviously squished) with calipers gave me following results:
- Backplate -> PCB (above VRAM): 3.3mm
- VRAM -> heatsink: 1.8mm
- VRM: 2mm

Should I just order the same thickness, or do they have to be slightly thicker (since heatsink should compress them)?

Thanks! :neocat_heart:

When in doubt, add more printf's

@Lili@donotsta.re I don't think that's possible, unibody was completely re-designed. I've worked on 2008 before, battery was removable from the bottom where you removed the bracket to get to ram slots and the drive, pretty sure layout had changed

"Yes we accept LLM contributions, but our code standards are high"

@tranquillity@mastodon.minionflo.net well, I know from 4 separate sources that AMD measures employee performance in token usage... I just did not expect that was the case over a year ago

@famfo@frogs.lgbt @pomagarnet@chaos.social ooof… 02:00.0 Non-Volatile memory controller [0108]: Kingston Technology Company, Inc. KC3000/FURY Renegade NVMe SSD [E18] [2646:5013] (rev 01) (prog-if 02 [NVM Express]) Subsystem: Kingston Technology Company, Inc. KC3000/FURY Renegade NVMe SSD [E18] [2646:5013] I’m still on kernel 7.0.10 (really should reboot my system…), but I think you just got unlucky unfortunately. Try pcie_aspm=off but I guess it’s gonna be a warranty replacement

I can't believe someone from @netbsd@mastodon.sdf.org managed to nerdsnipe me into bringing back AMD F15h support in coreboot

@portaloffreedom@social.linux.pizza that's not really... feasible to be honest. Support for ARM ACPI in Linux is pretty much non-existent outside of Ampere (at least that was the case last time I checked)

@kayla@kittyy.mom @tranquillity@mastodon.minionflo.net I'm sure some people will hate me for this, but after working on OpenSIL I can confidently say that I won't buy an AMD system again

@j-g00da@donotsta.re I also find it ironic how useless RCB is. Every other country I know deployed proper EAS to warn people about a possible danger or even threat to their life. In Poland you get a text message which you'll likely miss if something really *does* happen (I regularly find about it hours after receiving it because my phone is permanently in silent mode and is only used for banking and accepting calls as that's not something my main phone can do at the moment (xiaomi-pyxis w/postmarketOS)

It lives!

@moses_izumi I assume it's related to PMIC or detection. I don't own any SDM845 devices, but on SDM710 it works... but requires manual switching and powered device. If you patch USB node in devicetree to be dual-mode and have role-switch property, then issue manual switch it should work (after all it's just a designware usb controller): https://wiki.postmarketos.org/wiki/Qualcomm_Snapdragon_710_(SDM710)#USB It's possible that PMI8998 would even enable VBUS, but it likely would need a manual GPIO toggle to do so or something but again, I haven't worked on that PMIC/SoC so I don't want to mislead anyone

ok, I'm pretty sure AMD's GOP driver is vibe-coded