Alba 🌸 :v_pat:

hi i'm a poly lesbian catgirl who loves hummus, kernels and gfs 🏳️‍⚧️

I'm pretty strict about only boosting stuff with alt text. i try not to post/boost unCWed politics

pronouns: she/her
website: https://alba.sh
age: 29

Public Key

npub1eeqfjq6r2pa8cj0pvlgxg33pxqeaq8dwv7cwlm0s730swgykzqlqw6fehf

NIP-05 Address

[email protected]

Profile Code

nprofile1qqsvusyeqdp4q7nuf8sk05rygcsnqv7srkhx0v80ahc0ghc8yztpq0spz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsg5gw0e

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1eeqfjq6r2pa8cj0pvlgxg33pxqeaq8dwv7cwlm0s730swgykzqlqw6fehf

Show more details

Last Notes

2025-11-29 18:17:37 GMT

je m'alloc (i manage my own memory, for those who don't speak french)

2025-11-24 09:53:42 GMT

- reply

inb4 i get doxxed too

2025-11-24 09:41:38 GMT

god what the FUCK is wrong with this guy? the level of gaslighting is insane https://chaos.social/@gsuberland/115599251535775251

2025-11-18 12:23:30 GMT

- reply

https://media.tech.lgbt/media_attachments/files/115/570/676/872/463/464/original/b84414cd372aff6c.png

2025-11-18 12:17:24 GMT

has someone made the xkcd yet

2025-11-03 18:11:47 GMT

- reply

it's been long since i did RISC-V, what does "invalid virtual address" mean in this context? unaligned? unmapped? mapped with invalid permissions?

2025-09-18 13:38:32 GMT

i can't believe this is how i found out oracle is buying tiktok
https://media.tech.lgbt/media_attachments/files/115/225/567/087/773/348/original/8ab5b1c18960e5b1.png

2025-09-09 21:42:16 GMT

lil bro just sent me this without comment
https://media.tech.lgbt/media_attachments/files/115/176/510/163/682/339/original/ef558a5c55905f94.png

2025-09-02 03:28:22 GMT

- reply

thanks :neocat_heart: that sounds promising

2025-09-02 03:22:34 GMT

- reply

yeah, actually most of them criminalize it when it's for commercial uses... but i was even more wrong, the US does criminalize it in all cases

2025-09-01 21:22:26 GMT

- reply

not really? again, in many countries the distribution of pirated material is actually illegal, and as a matter of fact that's why my gf is uncomfortable with us torrenting (the seeding)

2025-09-01 21:20:05 GMT

- reply

IIUC the aspect of piracy was brought up in the lawsuit, and rejected by the judge because Meta took appropriate precautions to avoid seeding the files (which would actually be a violation)

2025-09-01 21:14:21 GMT

- reply

sure, it feels wrong that a 2 billion dollar company can just pirate materials *en masse* from loads of small authors. but is it illegal? most countries do not criminalize the *download* part of piracy, only the *upload* part. if you criminalize download that wouldn't just be for Meta, it'd also apply to you when you pirate Disney

2025-09-01 21:11:10 GMT

Alba's hot take for today is: I don't necessarily disagree with the judge who ruled Meta's pirating of materials for training to be fair use. the problem here is that intellectual property, like any form of private property, is a flawed premise from the start

2025-08-26 04:41:37 GMT

- reply

...but you're pushing apps to do their OS integrity checks using the standard attestation API instead of Play Integrity. if that's what you prefer, wouldn't it be good to act consequently? is it at least planned for the future?
(also I'd argue that even the original uses of attestation, where you only get to check that the key is in a genuine keystore, are DRM. as a user I should be able to control how I store my keys. that a service can refuse to operate unless I store my session key in a place that prevents me from migrating it to another device is an antifeature)

2025-08-26 04:25:35 GMT

- reply

you seem to be very focused on making apps work over making them DRM free... DRM is still DRM even if the service provider has allowlisted your OS and the app works there. if an app is only working because I'm using your particular signed builds of your particular OS, that's IMO just as important to know as when the app doesn't work. it's why Firefox notifies/asks even when the website's DRM is compatible with it

2025-08-26 04:16:42 GMT

- reply

nice. but what about the standard attestation API? do I get to know when applications fetch the attestation certificate chain? I imagine the root is in the project's control?

2025-08-26 04:00:27 GMT

- reply

yes, i know blocking access to DRM features will make many apps cease to work. still as a user I want to at least know when an app is attempting to use it and be able to control access. you know, [like Firefox does with EME by default](https://geekrewind.com/wp-content/uploads/2021/02/netflix-ubuntu-watch.webp )

2025-08-26 03:32:46 GMT

- reply

does grapheneos allow users to disable apps access to root-based attestation?

2025-08-26 03:27:42 GMT

- reply

good to know. that's certainly not the impression I got when someone from grapheneos reacted to the EU wallet prototype using Play Integrity by basically asking them to instead use the standard Android API, though... my memory isn't the best but I don't think the message was pushing back against attestation at all. but i can chalk that as a one off

2025-08-25 20:36:34 GMT

- reply

this was one of the options that came to mind while I was writing the second paragraph. nice that google is already doing that! they could've done... more of that, instead of 100% blocking users from sideloading things
(but also I'd be surprised if pegasus installed itself in a legal way, so it's not like this check will affect that kind of malware in any way?)

2025-08-25 20:30:58 GMT

- reply

i could be convinced that their interests are aligned with users when it comes to protecting people from viruses and such. but I wonder how much that played into the decision to introduce these checks, and how much was an excuse to get around anticompetence and monetize users more heavily
especially considering there's many intermediate forms of this that would've still allowed advanced users to sideload their stuff

2025-08-25 20:10:46 GMT

- reply

you'll also have to agree to some terms of service that are not yet public AFAIK, give them an email + phone number, verify them, then give them your public key and sign a challenge using the key (to prove you have the private part) and only then you may be given ownership of a package name.
you don't have to give google<li>your private key</li><li>the APK you'll be distributing</li>

2025-08-25 20:01:49 GMT

in case you want to skip all the corporate bullshit, this explains the process you'll have to go through in order to be able to sideload an app on your own fucking Android device
https://developer.android.com/developer-verification/assets/pdfs/introducing-the-android-developer-console.pdf
you have to provide legal ID that Google will verify, and PAY $25 unless you're a "hobbyist or student"

2025-07-29 10:12:33 GMT

- reply

nuuu (>'__')> if you ever wanna overcome it, 3b1b has some videos about the Fourier transform that i find very nice to build some intuition https://youtu.be/spUNpyF58BY

2025-07-29 10:05:35 GMT

- reply

ahh that's another thing, and sadly more complicated imo. I mean you can look at the definition, it's not that complicated, but building an intuition for what it does takes time
I'd recommend first learning about the (continuous) Fourier Transform and then you can look at what happens when you discretize it (which you need to do in order to, y'know, be able to compute it)

2025-07-29 10:01:45 GMT

- reply

oh, the Discrete Fourier Transform is the mathematical transform itself and the Fast Fourier Transform is the algorithm you end up implementing to calculate a DFT efficiently

2025-07-29 09:57:09 GMT

- reply

oh, for me it was pretty easy to understand in the case of a 2 divisor, you can get the DFT of the entire sequence from the DFT of the even samples and the DFT of the odd samples

2025-07-29 09:48:49 GMT

- reply

speaking of furry transfems, today I'll try implementing<li>an FFT</li><li>for the first time</li><li>with WebGL</li>

2025-07-29 09:07:52 GMT

she's so me
https://media.tech.lgbt/media_attachments/files/114/935/720/249/131/299/original/d2b5a77ef05a3436.png

2025-06-04 09:30:43 GMT

that reminded me of this meme which I'll put here for easy future access
https://media.tech.lgbt/media_attachments/files/114/624/392/333/973/088/original/01430eebc9514c4b.png

2025-03-13 21:17:36 GMT

- reply

yeah I'm skeptical of both things

2025-03-13 21:06:06 GMT

- reply

I'm also curious... i think they said stuff like C# or Kotlin was too slow

2025-03-09 19:31:07 GMT

https://media.tech.lgbt/media_attachments/files/114/134/130/935/890/206/original/6a7c39b892944e86.png

2025-01-19 01:01:16 GMT

i am once again stealing a thing to add alt text to it
https://media.tech.lgbt/media_attachments/files/113/852/304/297/865/539/original/a9b9a2de97389a9f.png

2025-01-13 10:32:52 GMT

stolen to add alt text
https://media.tech.lgbt/media_attachments/files/113/820/582/235/635/977/original/844c78937a58284a.jpeg

2024-12-26 11:12:56 GMT

https://media.tech.lgbt/media_attachments/files/113/718/823/503/307/976/original/94fbc1f30b0b45de.jpg

2024-11-17 23:05:20 GMT

- reply

what the fuck does libublk require serde_json for??

2024-11-17 22:39:31 GMT

- reply

now, I'd love to be able to "mount" the EBS snapshot as a userspace block device in my machine... but it looks like no one has done that yet :(

2024-11-17 22:37:02 GMT

- reply

oh hey, [lightsail snapshots can be exported into an AMI on your account](https://docs.aws.amazon.com/lightsail/latest/userguide/amazon-lightsail-exporting-snapshots-to-amazon-ec2.html ). all I had to do was do "keep snapshot" to create a manual snapshot from the automatic snapshot, and then "export to ec2". both these processes take an excruciatingly long time, why? a lightsail snapshot is already an AMI, you should just need to transfer it around??

2024-11-17 21:53:52 GMT

- reply

I could recover them from yesterday's backup... but will AWS let me access lightsail snapshots directly using the EBS APIs? we all know the response is "no"

2024-11-17 21:46:55 GMT

i just lost 4 months of personal server logs... definitely not a good sysadmin :neofox__w:

2024-11-17 19:23:56 GMT

I'm a simple girl. give me hummus and a solution to P=NP and I'm happy

2024-10-18 15:15:31 GMT

- reply

this is actually in std: [ManuallyDrop](https://doc.rust-lang.org/std/mem/struct.ManuallyDrop.html )

2024-09-29 08:52:36 GMT

- reply

oh, but where do you share your patches then? :o (edit: no worries I'll figure it out)

2024-09-29 02:31:41 GMT

- reply

I think Lina told me someone had already booted M3 and developed some drivers, but if that's true, either they didn't end up opening a PR or I'm looking at the wrong place, because I checked PRs and branches

2024-09-29 02:24:58 GMT

- reply

I think I have a basic DT ready. cpus, memory, soc, aic, reserved-mem and the uart. maybe I'm lacking a timer, but linux should know how to use the arm one i think?
now i just have to port the little changes to aic3 into linux, but that'll have to wait for tomorrow...

2024-09-29 01:24:53 GMT

- reply

okay, yes, that's it. apparently the AIC has a way to generate interrupts from software? and this is how we mimic the same IRQ that would come from the real thing

2024-09-29 01:20:35 GMT

- reply

I don't understand how the guest gets informed of the virtual UART. I was expecting it to be through the ADT, but I see no such thing. maybe it mimics the interface of the real debug UART and is mapped at exactly the same addresses? is that it?

2024-09-29 00:55:18 GMT

- reply

tip: if you're starting out at Asahi and don't know what a certain acronym means, I recommend to<li>check out <a href="https://github.com/AsahiLinux/docs/wiki/Introduction-to-Apple-Silicon#firmware-overview" target="_blank" rel="nofollow noopener noreferrer">this table</a></li><li>search with <code>site:asahilinux.org</code> (maybe they talked about it in the blog)</li><li>grep the linux and m1n1 repos</li>
some other random acronyms, mostly peripherals of the SoC:
AIC → Apple Interrupt Controller
AP → Application Processor (where the OS runs)
DCP → Display Controller Processor (video output)
MTP → Multitouch Processor (trackpad)
PMGR → Power Manager (probably the most important peripheral. it controls power, voltage, clock, .. to the rest of the peripherals)
RTKit → RTOS made by Apple that runs in several coprocessors (DCP, SMC, ANE, MTP, ...)
boot related:
ADT → Apple Device Tree (this is how Asahi refers to the device tree given to us by iBoot)
KC → Kernel cache
FuOS → Fully Untrusted OS
SFR → System Firmware and Recovery (thing in NOR flash plus the System Recovery partition)
1TR → One True Recovery
BP → Boot Policy

2024-09-28 20:48:34 GMT

- reply

what is the reg in a cpu FDT used for? the values used in the existing FDTs seem not to match those of the ADTs...

2024-09-28 20:40:58 GMT

- reply

I'm making the DT and it looks like Apple made a small typo in CPU8?
https://media.tech.lgbt/media_attachments/files/113/217/112/734/164/619/original/e9ce8a0c9411d03d.png

2024-09-28 14:10:43 GMT

- reply

thinking on which defconfig to use, the obvious choice seems to be arch/arm64/configs/defconfig but the last change to it (a few months ago) was marcan saying "this defconfig is broken" soooo

2024-09-28 14:08:27 GMT

- reply

hmm, marcan sure likes to force push to the repos

2024-09-28 13:53:12 GMT

- reply

okay, let's get building the kernel...

2024-09-28 02:08:22 GMT

- reply

actually before going to sleep i tried running a bare m1n1 inside its HV and it went well, other than a few "Dangerous write" to the PMGR that look to me more like preemptive warnings than anything else.
the child m1n1 booted fully and I can interact with it through the second uart, which is nice, while the child also has access to the other usb ports? nice :o

2024-09-28 01:56:21 GMT

- reply

leaving here a link to the [aarch64 linux boot protocol](https://www.kernel.org/doc/html/latest/arch/arm64/booting.html ) just in case I need it tomorrow

2024-09-28 01:41:54 GMT

- reply

I've signed a new stage1 with the build at this point. tomorrow I'll build a dumb kernel and begin playing with the HV, hopefully it just works

2024-09-28 01:14:49 GMT

- reply

anyway, all secondary cores work including IPIs. I'm not sure why IPIs work but they do... I'll move on, I hope this doesn't come back to bite me later

2024-09-28 01:10:21 GMT

- reply

ah wait, it was only CPU 0, and it was indeed one of these assumptions that boot_cpu == 0
https://media.tech.lgbt/media_attachments/files/113/212/505/735/286/260/original/487ffcdce9d76a6e.png

2024-09-28 01:00:54 GMT

- reply

I can't get the secondary cores to talk to me. they are clearly alive, as they respond to changes in the spin_table, and yet...

2024-09-28 00:38:51 GMT

- reply

I've also applied the mcc changes, so far so good...

2024-09-28 00:32:14 GMT

- reply

yep, that was it. mmu works

2024-09-28 00:31:17 GMT

- reply

it's also possible I finally found why the MMU was failing to start... it's because one of the memory sizes reported is now zero, leading to an empty mapping registered https://github.com/AsahiLinux/m1n1/pull/363/files

2024-09-28 00:26:28 GMT

- reply

looking at the [changes](https://github.com/AsahiLinux/m1n1/pull/402/files ) the only things I'm missing are<li>mostly obvious HV changes (due to boot_cpu not being 0 in this model)</li><li>AIC3 support (but I'll use <a href="https://github.com/AsahiLinux/m1n1/pull/364/files" target="_blank" rel="nofollow noopener noreferrer">this one instead</a> for it)</li><li>very small changes in MCC</li>
it also looks like [cpufreq hasn't changed much in T6031](https://github.com/AsahiLinux/m1n1/pull/367 ) so it could be the same as in T6030, but I think I'll not pull that yet

2024-09-28 00:12:55 GMT

- reply

waitwaitwait, it looks like someone already did most of the bringup for M3 and then some couple weeks ago someone did it for the M3 Pro... it's just that these are unmerged.
they seem to have most of my changes so far

2024-09-27 23:33:19 GMT

- reply

but then again, due to the whole "clusters have a max of 4 cores" thing which is no longer true, the last two cores should not be receiving IRQs... let me try to send them an IPI and see if they get it

2024-09-27 23:31:09 GMT

- reply

ah, I had added a new case to the switch but forgot the break :_) typical Alba. the cores do start up!

2024-09-27 23:16:36 GMT

- reply

it does not, I still get a fault in that offset

2024-09-27 22:55:18 GMT

- reply

...anyway, the offset for M3 Pro is 0x34000 according to my findings. let's see if it works

2024-09-27 21:54:52 GMT

- reply

and also, good thing I found (read: Lina pointed me to) this policy before doing anything more serious :blobcatnotlikethisgoogly:

2024-09-27 21:28:32 GMT

- reply

(or at least, none that I could find)
the other good news is that this doesn't ban me from contributing to other areas

2024-09-27 21:23:14 GMT

- reply

however it looks like it would be safe for me to write documentation about what I've found... does that include documentation contributed to the [wiki](https://github.com/AsahiLinux/docs/wiki ), or are there better channels to contribute it to? what worries me is that there is no review process for including things in the wiki...

2024-09-27 21:13:39 GMT

- reply

I was thinking on contributing this to m1n1, but I'm reading [the reverse engineering policy](https://asahilinux.org/copyright/#reverse-engineering-policy ) and it doesn't look like the contribution would be allowed. maybe I should contact marcan, but I lack the self-esteem to do so

2024-09-27 20:28:16 GMT

@npub1cxq…6wtz I try not to ask questions to avoid bothering you all, but this felt important... is adding this comment to m1n1 acceptable, or do we prefer not going into that much detail about Apple's codebase?
https://media.tech.lgbt/media_attachments/files/113/211/396/631/597/642/original/db6200b01e6e94ba.png

2024-09-27 20:14:08 GMT

- reply

so now, theoretically, I should be able to start the rest of the cores.
I'm a bit skeptical of that since m1n1 currently assumes a maximum of 4 cores per cluster, which seemed to be upheld until the M3...

2024-09-27 20:11:58 GMT

- reply

okay, I have found the "CPU start" offset for M3 Pro (which Apple calls the "MISC_CORES reg group").
in most device trees this offset is found in the misc-cores-offset ADT attribute, but for M3 Pro *in particular* they forgot to add it since they are also hardcoding it in the kext

2024-09-25 02:22:25 GMT

- reply

I feel like I'm starting to get my bearings in everything, slowly

2024-09-25 00:26:10 GMT

- reply

in fact almost 50% of the lines are spent on a single property of the PMGR (the devices list). ofc that'd probably change if we spent more time parsing more of the fields in other nodes, which right now appear as single-line hex blobs

2024-09-25 00:04:43 GMT

- reply

in other matters, I personally find it amazing how 60%+ of the lines of a (formatted) device tree is spent describing a single node (the power manager)

2024-09-24 22:24:08 GMT

- reply

I am a bit confused. the dSYM file is appropriately marked with the 0xA type (which per Wikipedia means "Companion file with only debug sections")
but objdump seems to indicate it has the actual sections too?

2024-09-24 22:13:45 GMT

- reply

it looks like (thanks @npub1dee…mxpr) odjdump does support mach-o, but instead of there being an option to enable it, it is [included only if the target is darwin](https://github.com/aixoss/binutils/blob/5a0ca69650b790b1ea7b7b5835e91a304d9cdc23/binutils/configure.ac#L420-L422 )

2024-09-24 21:41:19 GMT

- reply

looking at how m1n1 does it... they [wrote their own mach-o parser](https://github.com/AsahiLinux/m1n1/blob/main/proxyclient/m1n1/macho.py )

2024-09-24 21:37:41 GMT

- reply

loading the binary in lldb makes it crash. it tells me to upload the diagnostics when opening a bug report.
the diagnostics: an empty file

2024-09-24 20:53:16 GMT

- reply

are you telling me that binutils can manipulate PEs but not mach-o? disappointing. rizin/r2 seem to understand it

2024-09-24 20:45:05 GMT

- reply

chat how do I manipulate Mach-O binaries. is there a set of binutils that works for Mach-O

2024-09-24 20:43:34 GMT

- reply

so basically, for each target there are three kernels:<li>release</li><li>development</li><li>kasan</li>
"target" being either "no suffix" (x64) or the name of the SoC (arm64)

2024-09-24 16:36:52 GMT

- reply

now, how do I find a darwin kernel with debug symbols...

2024-09-24 16:35:51 GMT

- reply

looking through the device trees, it looks like most PMU types specify leg_scrpad as a single-item list, but there are two that use an int:<li>simetra (non-primary PMU in the early M1s)</li><li>chapel (primary PMU in M3 Pro and M3 Max)</li>
so it looks like a silly mistake?

2024-09-22 21:38:28 GMT

- reply

eh, i'm guessing the device tree changed from one chip to the other... tomorrow I'll compare device trees, for now I skip that step.
and I can correctly chainload to itself :) it's so much nicer

2024-09-22 21:06:33 GMT

- reply

USB appears to be working! at least the console part. the proxy's shell then fails with this error:
https://media.tech.lgbt/media_attachments/files/113/183/239/905/566/042/original/40ec0f4762da54df.png

2024-09-22 19:10:14 GMT

- reply

aight so let's let it continue with the initialization and see how far we get, I'm curious

2024-09-22 19:04:59 GMT

- reply

it worked! I only had to slightly modify the console code to allow for early initialization
https://media.tech.lgbt/media_attachments/files/113/182/761/333/315/069/original/ccfb4fae58558f68.jpeg

2024-09-22 18:51:51 GMT

- reply

ah no, it seems to pass it in both places. cool :)

2024-09-22 18:34:51 GMT

- reply

right, so apparently<li>the boot args have a slot specifically for the framebuffer</li><li>but iBoot doesn't pass us a framebuffer there</li><li>instead it populates the /vram node in the ADT</li><li>the node is populated dynamically (addr/size is zero in the disk ADT)</li>

2024-09-22 15:50:05 GMT

- reply

(pause because I have to shower)

2024-09-22 15:18:08 GMT

- reply

let me first make sure I know how to display things in the screen... Lina made it sound easy but keeping in mind how dumb I am, I want to check

2024-09-22 15:16:28 GMT

M3 Pro bringup, hopefully for real this time :neofox:
my [previous attempt](https://tech.lgbt/@mildsunrise/113081388169064850 ) got me sidetracked for different reasons

2024-09-21 23:57:53 GMT

- reply

this would've maybe prevented the panic and brick, but most of the risk is still there because at some point I'll need to bless my stub again. nyeh

2024-09-21 23:56:44 GMT

- reply

additional precautions in case i ever want to attempt the dd again (which i very much doubt):<li>HAVE A SECOND MAC. not just to run AC but also the debugging UART (though the latter also requires a fancy USB cable)</li><li>after booting into the main OS, shrinking its container and creating an unformatted partition, bless the main OS as default boot volume and reboot</li><li>dd the source partition into an image, check its size, disconnect origin disk and reboot (again)</li><li>dd image file into target partition, delete image file and reboot </li>
basically do nothing from recovery