untold basin Jan 23, 2025, 11:50 AM

#

wtf

#

i dont even know anymore

#

im gonna blame qemu pl011 emulation

#

this doesnt break on 8250 uart

untold basin Jan 23, 2025, 4:31 PM

#

okay

#

ill charge the m1

#

and fix the linux boot header

untold basin Jan 23, 2025, 6:01 PM

#

okay

#

it hard reset really quickly

#

fuck

#

no logs

#

oh wait

#

im not running it on hv

#

damn!

#

ok good that i have a usb reset

#

in hv i get this error

#

cool

#

i guess

#

damn

#

ah

#

caches!

#

i need to turn those on

#

yeah its an atomic to log::STATE

#

https://esr.arm64.dev/#0x96000035

#

IMPLEMENTATION DEFINED fault (Unsupported Exclusive or Atomic access).

#

wtf

glad bough Jan 23, 2025, 7:32 PM

#

https://tenor.com/view/cat-despair-meme-atone-gif-18083281511005463831

Tenor

untold basin Jan 23, 2025, 7:56 PM

#

tbh

#

i think im gonna switch to libfdt instead of my homebrew fdt code

#

later

trim seal Jan 23, 2025, 7:57 PM

#

why?

untold basin Jan 23, 2025, 7:57 PM

#

trim seal why?

because my code sucks ass lol

trim seal Jan 23, 2025, 7:58 PM

#

make a uFDT

untold basin Jan 23, 2025, 7:58 PM

#

that would involve writing C code

#

libfdt is already pretty good too

trim seal Jan 23, 2025, 7:58 PM

#

well not necessarily

#

u can make C bindings for a rust library

untold basin Jan 23, 2025, 7:58 PM

#

that would suck ass

#

and i would never recommend anyone use that

trim seal Jan 23, 2025, 7:59 PM

#

lol

#

why does your code suck tho, arent fdts simple

untold basin Jan 23, 2025, 8:00 PM

#

yes

#

but it is one of my core skills to write bad code for simple things

trim seal Jan 23, 2025, 8:02 PM

#

lol

untold basin Jan 23, 2025, 8:02 PM

#

in this case, the issue is rust ownership

#

the dtb is global

#

actually that doesnt matter i think

#

does it

sacred spire Jan 23, 2025, 8:13 PM

#

just throw an rc around everything bro

#

troll2

untold basin Jan 23, 2025, 8:13 PM

#

lmao

glad bough Jan 23, 2025, 9:10 PM

#

untold basin the dtb is global

'static

untold basin Jan 23, 2025, 9:10 PM

#

glad bough 'static

yeah thats what i do right now

#

but theres a lot of big ugly there

untold basin Jan 24, 2025, 10:51 AM

#

okay time to replace every single place that does any DMA with a proper dma buffer allocation

#

and proper cache flushes

#

i should add backtracing from exceptions on arm64

#

okay first crash

#

fdt parsing is fucked

rose sonnet Jan 24, 2025, 11:57 AM

#

there's a nice library by r4

untold basin Jan 24, 2025, 12:03 PM

#

hmm

#

C though

#

also it needs allocation

#

also i need write

open osprey Jan 24, 2025, 12:32 PM

#

untold basin also i need write

why write?

untold basin Jan 24, 2025, 12:33 PM

#

open osprey why write?

i need to edit the FDT to adjust the /chosen node

#

plus, limine requires me to remove memory nodes

untold basin Jan 24, 2025, 3:13 PM

#

okay progress update

#

im on fwcfg now

untold basin Jan 24, 2025, 5:55 PM

#

okay, now pci

#

and now im getting a weird L1 translation fault

#

wtf

#

ah its oob

#

oookay so on qemu the pci window is outside of the low 128 gigs

#

fucking hell

#

okay

#

new plan

#

i'll construct a page table with a single PTE

#

which maps just the loader

#

then i parse the FDT

#

i'll assume it lives within a single 1 gig contingous area

#

so it can just get its own PTE

#

actually im overcomplicating it, i'll just parse the FDT in a prekernel, clean up the memory map, then map my loader and be done with it

#

and then build the prekernel with -mstrict-align

#

i'll call it "cmodule" instead of "prekernel" though

#

and reuse it for a bunch of other stuff

#

e.g. SMP spinning

#

hmm or maybe not

limber abyss Jan 24, 2025, 8:45 PM

#

I sometimes look in this thread and wonder what the hell is happening

untold basin Jan 24, 2025, 8:52 PM

#

lmao

untold basin Jan 24, 2025, 9:17 PM

#

limber abyss I sometimes look in this thread and wonder what the hell is happening

me too, me too

lunar rock Jan 24, 2025, 11:23 PM

#

lmao

agile nexus Jan 25, 2025, 6:56 AM

#

untold basin also i need write

fwiw it does do write, its opt-in behind a define and can operate with a static buffer if you're using it pre-allocator (its a little gross but it works).

#

afaik the only thing its missing is being able to set the reserved memory entries, which I guess you probably do care about

#

ah and ~~good~~ docs

#

but yeah, cant help you with the 'being written in c' part 😛

untold basin Jan 26, 2025, 12:08 PM

#

agile nexus fwiw it does do write, its opt-in behind a define and can operate with a static ...

oh cool

#

i might use it then idk

#

since using C FFI is a famous strategy of escaping rust UB

#

have rust UB? just put the UB in C code and its not UB anymore

limber abyss Jan 26, 2025, 1:01 PM

#

have rust? just put the rust code in C and it's not rust anymore

untold basin Jan 26, 2025, 1:34 PM

#

lol

untold basin Jan 27, 2025, 12:12 PM

#

okay i think i'll do a two stage bootloader

#

to maintain my sanity

untold basin Jan 27, 2025, 12:12 PM

#

agile nexus fwiw it does do write, its opt-in behind a define and can operate with a static ...

actually, can it turn the changed DTB back into bytes?

agile nexus Jan 27, 2025, 12:13 PM

#

It can make a new fdt if that's what you mean

untold basin Jan 27, 2025, 12:13 PM

#

ah yeah

#

thats nice

#

i probably still want rsvmem support though

agile nexus Jan 27, 2025, 12:13 PM

#

Mm I think that requires an allocator though, for the buffer to write into

agile nexus Jan 27, 2025, 12:14 PM

#

untold basin i probably still want rsvmem support though

Fair

untold basin Jan 27, 2025, 12:14 PM

#

unless i just binpatch your FDT to add a rsvmem region

#

wait do you not expose rsvmem at all???

agile nexus Jan 27, 2025, 12:14 PM

#

Lol or that

untold basin Jan 27, 2025, 12:14 PM

#

you need to consider it to construct a valid memory map

#

always

agile nexus Jan 27, 2025, 12:14 PM

#

untold basin wait do you not expose rsvmem at all???

I haven't needed to personally, so I haven't bothered

untold basin Jan 27, 2025, 12:15 PM

#

ah you just use it with limine?

agile nexus Jan 27, 2025, 12:15 PM

#

Its a long story, in the cpp version I did, but since making smoldtb I'm yet to need it

untold basin Jan 27, 2025, 12:15 PM

#

fair enough

agile nexus Jan 27, 2025, 12:15 PM

#

Pretty much

#

The CPP version was before limine had rv support

#

So it had those features

untold basin Jan 27, 2025, 12:16 PM

#

ahhh

agile nexus Jan 27, 2025, 12:16 PM

#

Honestly, if you're willing to give me a weekend I'll add them

untold basin Jan 27, 2025, 12:16 PM

#

apart from the fact that it requires allocations, it looks quite nice

#

also why uintmax_t

#

i'd prefer to be given raw bytes

#

there's like a bunch of places where i need the raw cells too

agile nexus Jan 27, 2025, 12:19 PM

#

untold basin also why uintmax_t

its just what I wanted at the time. I was using size_t's for everything but got bitten by that somewhere (I forget where now). uintmax_t is hilariously overkill I know but solved my issue.

agile nexus Jan 27, 2025, 12:20 PM

#

untold basin i'd prefer to be given raw bytes

yeah fair, could be easily added. ~~You could get that via read_string(0), which willl return a char* to the beginning of the property data.~~ actually dtb_stat_prop() returns the base + length of the cells.

#

I should implement be an api for that though, maybe also returning the number of cells. thanks.

untold basin Jan 27, 2025, 12:27 PM

#

agile nexus yeah fair, could be easily added. ~~You could get that via `read_string(0)`, whi...

i'd want to get the length

#

aah stat_prop works thanks

agile nexus Jan 27, 2025, 12:27 PM

#

I will eventually finish the docs, sorry

untold basin Jan 27, 2025, 12:28 PM

#

no worries

#

okay the new new plan is to bring back a preloader

#

i'll map it with two 2mb pages, then give it 2 more 2mb pages for mapping the FDT and a 4k page for mapping whatever physical memory

#

running off of the stack that i'll later reuse as the normal shkwve stack

untold basin Jan 27, 2025, 12:38 PM

#

untold basin okay the new new plan is to bring back a preloader

and the preloader will map shkwve into virtual memory

#

and also map all of physical RAM

#

and then device memory (e.g. PCI BARs) need to be mapped later by device drivers, on demand

lunar rock Jan 27, 2025, 1:54 PM

#

agile nexus its just what I wanted at the time. I was using size_t's for everything but got ...

About that...

/home/blanham/projects/rooster/build/_deps/smoldtb-src/smoldtb.c:350:34: error: passing argument 3 of 'dtb_read_prop_1' from incompatible pointer type [-Wincompatible-pointer-types]
  350 |         dtb_read_prop_1(prop, 1, &handle);
      |                                  ^~~~~~~
      |                                  |
      |                                  size_t * {aka unsigned int *}
In file included from /home/blanham/projects/rooster/build/_deps/smoldtb-src/smoldtb.c:1:
/home/blanham/projects/rooster/build/_deps/smoldtb-src/smoldtb.h:90:70: note: expected 'uintmax_t *' {aka 'long long unsigned int *'} but argument is of type 'size_t *' {aka 'unsigned int *'}
   90 | size_t dtb_read_prop_1(dtb_prop* prop, size_t cell_count, uintmax_t* vals);
      |                                                           ~~~~~~~~~~~^~~~
/home/blanham/projects/rooster/build/_deps/smoldtb-src/smoldtb.c:360:34: error: passing argument 3 of 'dtb_read_prop_1' from incompatible pointer type [-Wincompatible-pointer-types]
  360 |         dtb_read_prop_1(prop, 1, &handle);
      |                                  ^~~~~~~
      |                                  |
      |                                  size_t * {aka unsigned int *}
/home/blanham/projects/rooster/build/_deps/smoldtb-src/smoldtb.h:90:70: note: expected 'uintmax_t *' {aka 'long long unsigned int *'} but argument is of type 'size_t *' {aka 'unsigned int *'}
   90 | size_t dtb_read_prop_1(dtb_prop* prop, size_t cell_count, uintmax_t* vals);
      |                                                           ~~~~~~~~~~~^~~~```

untold basin Jan 27, 2025, 1:55 PM

#

lmfao

lunar rock Jan 27, 2025, 1:55 PM

#

meme

glad bough Jan 27, 2025, 1:55 PM

#

lol

agile nexus Jan 27, 2025, 1:56 PM

#

rip

agile nexus Jan 27, 2025, 11:56 PM

#

lunar rock About that... ```/home/blanham/projects/rooster/build/_deps/smoldtb-src/smoldtb...

Btw this was fixed last night (I also exposed the reserved memory stuff). What target was that for?

lunar rock Jan 27, 2025, 11:58 PM

#

agile nexus Btw this was fixed last night (I also exposed the reserved memory stuff). What t...

I got the same error on eight different 32-bit targets meme

agile nexus Jan 27, 2025, 11:58 PM

#

That's a bit embaressing lol

lunar rock Jan 28, 2025, 12:02 AM

#

agile nexus That's a bit embaressing lol

tbh, some of those don't even use DTB, but I haven't done any fine level conditional dependencies in my build system yet so it got pulled into all of them meme

rose sonnet Jan 28, 2025, 12:02 AM

#

lunar rock I got the same error on eight different 32-bit targets <:meme:575445094589661215...

wait a second...

agile nexus Jan 28, 2025, 12:03 AM

#

lunar rock tbh, some of those don't even use DTB, but I haven't done any fine level conditi...

I mean it's nice that it was tested anyway

agile nexus Jan 28, 2025, 12:03 AM

#

rose sonnet wait a second...

There's been a handful of updates since you integrated it into your kernel

rose sonnet Jan 28, 2025, 12:03 AM

#

yeah i need to update it

untold basin Jan 29, 2025, 1:38 PM

#

okay now i'm working on the shkwve first stage loader(tm)

#

so far i have an FDT parser

#

and a naming scheme that is truly horrifying

#

oh and a print function straight out of c++

#

i've heard that C doesn't have function overloading

#

but clearly anyone who says that has never used clang(::overloadable)

#

why yes i would like to have C++ name mangling in my C program

glad bough Jan 29, 2025, 1:44 PM

#

untold basin i've heard that C doesn't have function overloading

it kinda does

#

if you have a very broad definition of overloading

#

_Generic() meme

untold basin Jan 29, 2025, 1:46 PM

#

glad bough _Generic() <:meme:575445094589661215>

yeah but that sucks

#

you can't even define an overload for void* and have it catch every kind of pointer

glad bough Jan 29, 2025, 1:48 PM

#

you can afaik?

#

@untold basin https://godbolt.org/z/8YnY81cfW

Compiler Explorer - C (x86-64 gcc 14.2)

struct bar {
int a, b;
};

void foo_struct(struct bar* argument);
void foo_void(void* argument);

#define foo(X) _Generic((X),
struct bar*: foo_struct,
default: foo_void
)(X)

int main() {
struct bar arg;
void* arg2;
foo(&arg);
foo(arg2);
}

untold basin Jan 29, 2025, 1:54 PM

#

glad bough <@696410219663851552> https://godbolt.org/z/8YnY81cfW

yeah but that's a stupid catch all

#

which is boring

#

overloadable is way better

glad bough Jan 29, 2025, 1:54 PM

#

overloads are boring

untold basin Jan 29, 2025, 1:54 PM

#

its also decentralized so including more headers can add more things

#

so for example including my endianness support header makes my print function know how to print le32 and be32

#

(my endian support header is so hilariously overengineered too lol https://paste.sr.ht/~pitust/2f27640c30bae08e335ca24cd5c0ad93c78e5201)

#

hmm i wonder

#

should i respect /chosen/stdout-path

#

or should any use every serial i find as the stdout

#

probably the former tbh

rose sonnet Jan 29, 2025, 2:10 PM

#

untold basin but clearly anyone who says that has never used clang(::overloadable)

this is terrifying

untold basin Jan 29, 2025, 2:12 PM

#

rose sonnet this is terrifying

terrifyingly good you mean?

rose sonnet Jan 29, 2025, 2:13 PM

#

terrifyingly neutral

untold basin Jan 29, 2025, 2:13 PM

#

it is actually quite nice

#

if you dont want all of C++ but you want just one thing or whatever

rose sonnet Jan 29, 2025, 2:13 PM

#

clang is not all C

untold basin Jan 29, 2025, 2:13 PM

#

rose sonnet clang is not all C

yes it is

#

name one other compiler

rose sonnet Jan 29, 2025, 2:13 PM

#

idk if you want C++ features why not just use C++

untold basin Jan 29, 2025, 2:14 PM

#

rose sonnet idk if you want C++ features why not just use C++

that would require me to write clang++ in my compiler invocation

#

and exposes me to C++ naysayers

#

also it would probably make me do a lot of c++ crap i dont want to have

rose sonnet Jan 29, 2025, 2:14 PM

#

untold basin name one other compiler

icc

untold basin Jan 29, 2025, 2:14 PM

#

idk

untold basin Jan 29, 2025, 2:14 PM

#

rose sonnet icc

thats just clang

rose sonnet Jan 29, 2025, 2:14 PM

#

gcc is not clang

untold basin Jan 29, 2025, 2:15 PM

#

you said icc

#

icc is clang

rose sonnet Jan 29, 2025, 2:15 PM

#

ok clang is almost all C

untold basin Jan 29, 2025, 2:15 PM

#

ok fine

untold basin Jan 29, 2025, 2:15 PM

#

rose sonnet gcc is not clang

gcc is a shitty compiler though

#

so irrelevant

#

meme

rose sonnet Jan 29, 2025, 2:16 PM

#

I think this server need to write its own C compiler

#

There's a shortage of them

untold basin Jan 29, 2025, 2:16 PM

#

not really

#

we have clang

#

clang is already the best C compiler

rose sonnet Jan 29, 2025, 2:17 PM

#

but only having 1 compiler is boring

untold basin Jan 29, 2025, 2:17 PM

#

true

glad bough Jan 29, 2025, 2:24 PM

#

rose sonnet I think this server need to write its own C compiler

yea

untold basin Jan 29, 2025, 2:51 PM

#

#

so thats more progress ig

#

okay next steps

#

uh

#

memory map

untold basin Jan 29, 2025, 4:15 PM

#

perfect

#

now time to clean it up

untold basin Jan 29, 2025, 4:43 PM

#

memory map cleanup!

#

and now i have 4k alignment

#

tbh in rust the binary for boot0 would be like 50x bigger

#

and not 44k

#

which is mostly bss

zenith lion Jan 29, 2025, 5:03 PM

#

you're rewriting in c?

untold basin Jan 29, 2025, 5:04 PM

#

no

#

i'm adding a C first-stage loader

#

because i need it for bootstrap

#

hmm maybe i should rewrite in C

zenith lion Jan 29, 2025, 5:04 PM

#

ah

#

M-mode?

untold basin Jan 29, 2025, 5:04 PM

#

ill think about it later

untold basin Jan 29, 2025, 5:04 PM

#

zenith lion M-mode?

EL1

#

so no

zenith lion Jan 29, 2025, 5:05 PM

#

ah arm true

untold basin Jan 29, 2025, 5:05 PM

#

yeah arm

#

for now

#

ill be booting with it on riscv too

#

actually now that you said it

#

ill defo use it in M mode if im gonna be supporting split S/M

#

and if im gonna be supporting returning to EL2 etc etc

zenith lion Jan 29, 2025, 5:06 PM

#

if it's gonna hold the function of an SPL, do M mode

#

if it's not do it in S mode

untold basin Jan 29, 2025, 5:06 PM

#

it wont do load

#

no

zenith lion Jan 29, 2025, 5:06 PM

#

because otherwise you're gonna lose SBI :3

untold basin Jan 29, 2025, 5:06 PM

#

so it will be hard tied to the rest of the logic

untold basin Jan 29, 2025, 5:06 PM

#

zenith lion because otherwise you're gonna lose SBI :3

well no i can implement my own SBI stuff

#

would be a lot of effort but its totally doable

zenith lion Jan 29, 2025, 5:07 PM

#

there's a rust sbi implementation i think it's by xvanc?

untold basin Jan 29, 2025, 5:07 PM

#

yeah but why

#

:^)

#

also now that you said it maybe i should rewrite in C

#

idk

#

ill think about it

zenith lion Jan 29, 2025, 5:07 PM

#

rust is fine

#

this loader might be constrained by size tho

#

so C for this and rust for the rest makes perfect sense

untold basin Jan 29, 2025, 5:08 PM

#

zenith lion this loader might be constrained by size tho

its more that i dont want to add too much to code size

#

the entire thing is gonna be one big binary

zenith lion Jan 29, 2025, 5:08 PM

#

rust for the rest:^) rhymes

untold basin Jan 29, 2025, 5:08 PM

#

baked together by some build type process

zenith lion Jan 29, 2025, 5:08 PM

#

untold basin the entire thing is gonna be one big binary

oh(

untold basin Jan 29, 2025, 5:09 PM

#

lol i dont want to write two disk drivers

zenith lion Jan 29, 2025, 5:09 PM

#

i thought this was gonna be an SPL essentially so stored on rom

untold basin Jan 29, 2025, 5:09 PM

#

nah

#

its to set up page tables and shit

#

and standardize the environment and such

zenith lion Jan 29, 2025, 5:09 PM

#

ah

untold basin Jan 29, 2025, 5:09 PM

#

because i dont want to compile everything without aligned access support

#

and also rust without paging on arm64 isnt really feasible

zenith lion Jan 29, 2025, 5:10 PM

#

ah ok i understand

untold basin Jan 29, 2025, 5:10 PM

#

untold basin and also rust without paging on arm64 isnt really feasible

because you need paging to do atomics

zenith lion Jan 29, 2025, 5:11 PM

#

btw i really hate the fact that u-boot is basically the only SPL option in risc-v

untold basin Jan 29, 2025, 5:11 PM

#

i see but also im not gonna write stupid amounts of disk driver code

#

it would easily double the code size and require me to write lots of careful handover code

zenith lion Jan 29, 2025, 5:12 PM

#

zenith lion btw i really hate the fact that u-boot is basically the only SPL option in risc-...

one day I'll work on some hardware and I'll write an in-house firmware for it because i have an issue or two with everything out there (uboot, edk2, whatever)

#

maybe not coreboot, coreboot's cool

untold basin Jan 31, 2025, 9:41 AM

#

okay the new version of the shkwve boot0 can now set up paging too

#

now i'll do handover ig

untold basin Jan 31, 2025, 10:43 AM

#

hmm i saw how linux does iommu stuff

#

ill probably steal it tbh

untold basin Jan 31, 2025, 12:01 PM

#

hmm now im wondering if i should do like

#

rust for drivers and C for everything else or something idk

trim seal Jan 31, 2025, 12:04 PM

#

why is that

glad bough Jan 31, 2025, 12:04 PM

#

untold basin rust for drivers and C for everything else or something idk

nah

#

rust for everything

trim seal Jan 31, 2025, 12:04 PM

#

i have a better idea

#

C for everything

untold basin Jan 31, 2025, 12:05 PM

#

glad bough rust for everything

rust binaries are so fucking big though

trim seal Jan 31, 2025, 12:05 PM

#

meme

untold basin Jan 31, 2025, 12:05 PM

#

lmfao

trim seal Jan 31, 2025, 12:05 PM

#

they also consume terabytes of stack

untold basin Jan 31, 2025, 12:05 PM

#

and their ub rules are insane

glad bough Jan 31, 2025, 12:05 PM

#

nooo

untold basin Jan 31, 2025, 12:05 PM

#

trim seal they also consume terabytes of stack

only 32k its fineeeeee (lol)

#

im seriously considering using more C

#

especially because C codegen is a lot better lul

devout swift Jan 31, 2025, 12:05 PM

#

untold basin rust binaries are so fucking big though

eh, with lto + strip its not that bad

untold basin Jan 31, 2025, 12:06 PM

#

devout swift eh, with lto + strip its not that bad

eeeh

#

i can lto and strip C code too lol

glad bough Jan 31, 2025, 12:06 PM

#

it's fine tbh because you have nostd anyways

#

with std i would agree

untold basin Jan 31, 2025, 12:06 PM

#

im using C already for the first stage boot logic

#

but yeah idk maybe not

untold basin Jan 31, 2025, 12:07 PM

#

untold basin im using C already for the first stage boot logic

because i dont want to have too big of a binary

glad bough Jan 31, 2025, 12:07 PM

#

then doing it in rust was a mistake to begin with KEKW

untold basin Jan 31, 2025, 12:07 PM

#

fair lmao

trim seal Jan 31, 2025, 12:07 PM

#

why do u have multiple stages now?

untold basin Jan 31, 2025, 12:08 PM

#

trim seal why do u have multiple stages now?

i'll be packing them into one binary

#

i need to have some paging setup

#

and the paging setup is nontrivial because it needs to allocate and parse FDT and shit

#

and i need it to not be rust

#

and i need it to be a separate TU

untold basin Jan 31, 2025, 12:11 PM

#

glad bough then doing it in rust was a mistake to begin with <:KEKW:913708372329644042>

also, im fine with having one copy of the rust stdlib

#

two copies are a bit excessive

glad bough Jan 31, 2025, 12:11 PM

#

dynamic driver modules!

#

have a dynsymtab the size of the whole binary!

zenith lion Jan 31, 2025, 12:22 PM

#

trim seal they also consume terabytes of stack

want to fix it? implement new-in-place in rustc :^3

#

e.g. Box::new() and stuff that doesn't first allocate on stack

#

it's been an open issue for decades

trim seal Jan 31, 2025, 12:24 PM

#

nah id rather use a language like zig or c

zenith lion Jan 31, 2025, 12:24 PM

#

:(

#

i would be so happy to have new-in-place

#

and a syntax for new-in-place too (like C++'s new-in-place

trim seal Jan 31, 2025, 12:25 PM

#

be the change u want to see meme

zenith lion Jan 31, 2025, 12:25 PM

#

sure but there's lots of bikeshedding

#

i wish rust wasn't as big as it is and we could do changes faster

#

at some point I'll return to langdev

#

I would really enjoy a language that only I can use

#

that would allow me to not care about other people's opinions

#

lzcunt software license

Definitions

Licensor means lzcunt and lzcunt only, if you're not her you're not licensed to use this software

open osprey Jan 31, 2025, 1:04 PM

#

untold basin rust for drivers and C for everything else or something idk

I am of this opinion that rust is a reasonable languagr to write drivers in for a system written in another language

untold basin Jan 31, 2025, 1:05 PM

#

zenith lion # lzcunt software license ## Definitions Licensor means lzcunt and lzcunt only,...

you dont need a license for that?

open osprey Jan 31, 2025, 1:05 PM

#

Because a good driver API is an abstract one. So the drivers don't have to deal so much in the problems rust can't answer

zenith lion Jan 31, 2025, 1:07 PM

#

untold basin you dont need a license for that?

it was supposed to be funny

#

that's why it's a license

untold basin Jan 31, 2025, 1:43 PM

#

anyway

untold basin Jan 31, 2025, 1:43 PM

#

open osprey I am of this opinion that rust is a reasonable languagr to write drivers in for ...

maybe

#

maybe i'll do C++

rose sonnet Jan 31, 2025, 2:03 PM

#

Why not Go

untold basin Jan 31, 2025, 3:19 PM

#

no.

untold basin Jan 31, 2025, 4:56 PM

#

okay so how should i implement my new shitty driver model

#

hmm maybe i should use message based async

#

what if i did C++

#

and used C++ coroutines

#

because C++ has implicitly shit perf instead of explicitly shit perf

#

and i have some C code for boot0 already

untold basin Jan 31, 2025, 7:11 PM

#

holy shit who came up with this crap

trim seal Jan 31, 2025, 7:29 PM

#

untold basin and used C++ coroutines

Who needs oom handling amirit

untold basin Jan 31, 2025, 7:31 PM

#

trim seal Who needs oom handling amirit

it's a bootloader

#

OOM handling isn't a thing there anyway

sacred spire Jan 31, 2025, 9:02 PM

#

untold basin and used C++ coroutines

C++ coroutines are a headache compared to rust async

untold basin Jan 31, 2025, 9:03 PM

#

sacred spire C++ coroutines are a headache compared to rust async

nah

#

well

#

maybe

#

but consider this

#

its not rust

sacred spire Jan 31, 2025, 9:05 PM

#

true

rose sonnet Jan 31, 2025, 10:46 PM

#

sacred spire C++ coroutines are a headache compared to rust async

c++ coroutines are fun (to debug)

untold basin Jan 31, 2025, 10:49 PM

#

lol

tawdry knoll Feb 1, 2025, 1:19 AM

#

What are you doing to make your Rust binaries so large? I see this brought up a lot, but never experienced it myself, so I am wondering whats causing it?

tawdry knoll Feb 1, 2025, 1:22 AM

#

zenith lion e.g. Box::new() and stuff that doesn't first allocate on stack

I am pretty sure this has been fixed: https://github.com/rust-lang/rust/pull/82806

GitHub

Enable MemorySSA in MemCpyOpt by nikic · Pull Request #82806 · rust...

LLVM 12 ships with an implementation of MemCpyOpt which is based on MSSA instead of MDA. This implementation can eliminate memcpys across blocks, and as such fixes many (but not all) failures to el...

#

Also really cool project btw, I like the way your register dump looks lol

zenith lion Feb 1, 2025, 6:38 AM

#

tawdry knoll I am pretty sure this has been fixed: https://github.com/rust-lang/rust/pull/828...

I don't think this is what I mean

#

what I mean is like guaranteed new in place

#

e.g. in debug mode

untold basin Feb 1, 2025, 10:16 AM

#

tawdry knoll What are you doing to make your Rust binaries so large? I see this brought up a ...

i don't really know, tbh

zenith lion Feb 1, 2025, 10:36 AM

#

zenith lion e.g. in debug mode

yep I asked in the rust server

#

this is just enabling an LLVM optimization for this

#

there's no rust guarantee that the stack allocation will be removed

#

so your max allocation is still limited by your stack size kinda

#

you can overflow the stack when creating huge strings or similar

sacred spire Feb 1, 2025, 11:55 AM

#

zenith lion you can overflow the stack when creating huge strings or similar

How the fuck does this happen

#

Does it just not put string contents on the heap?

zenith lion Feb 1, 2025, 11:56 AM

#

it does but heap allocations are first on the stack

#

e.g. take Box as an example

#

Box::new(123) will stack allocate 123 then memcpy it to the heap

#

LLVM can and will optimise that out but in debug builds and stuff it might not

sacred spire Feb 1, 2025, 11:57 AM

#

zenith lion Box::new(123) will stack allocate 123 then memcpy it to the heap

Borrow checker meme?

zenith lion Feb 1, 2025, 11:57 AM

#

so if you have a huge boxed array for example

zenith lion Feb 1, 2025, 11:57 AM

#

sacred spire Borrow checker meme?

nah

#

new() is a function so it's arguments are created as they normally would

#

so 123 is created as it normally would be

#

then passed

#

tho 123 will likely be assigned a register so imagine something more complicated

sacred spire Feb 1, 2025, 11:58 AM

#

zenith lion new() is a function so it's arguments are created as they normally would

Oh and new doesn't take references?

zenith lion Feb 1, 2025, 11:58 AM

#

nope it takes ownership of the value it can't take references

#

the solution to this problem is to guarantee that temporaries passed in this way are initialised in place

#

so guaranteeing the optimisation

#

and sometimes you want to do it manually as well which is new-in-place in C++

#

new (ptr) Foo();

#

rust doesn't have an equivalent

#

it used to have box expressions as an equivalent but it was removed

sacred spire Feb 1, 2025, 12:00 PM

#

That's a shame

#

Is there an RFC or smth tracking this?

zenith lion Feb 1, 2025, 12:00 PM

#

the rust "equivalent" is *rawptr = whatever;

zenith lion Feb 1, 2025, 12:01 PM

#

sacred spire Is there an RFC or smth tracking this?

i remember there is but i couldn't find it

#

and there's lots of bikeshedding

sacred spire Feb 1, 2025, 12:01 PM

#

zenith lion and there's lots of bikeshedding

ofc there is

zenith lion Feb 1, 2025, 12:02 PM

#

zenith lion so guaranteeing the optimisation

this is awesome because it requires no extra syntax

#

it's basically copy elision but you're eliding a stack allocation

#

getting rid of the temporary

#

bc the temporary can sometimes not fit in the stack

#

e.g. it's hard to create huge boxed arrays

#

not impossible mind you

untold basin Feb 1, 2025, 2:08 PM

#

zenith lion it used to have box expressions as an equivalent but it was removed

box expressions are the same as Box::new

#

Box::new is a language item

#

and the impl is literally fn new(x: T) -> Box<T> { Box::new(x) }

zenith lion Feb 1, 2025, 2:10 PM

#

I thought the impl was // compiler implemented or something and it was #[rustc_box]ed

untold basin Feb 1, 2025, 2:49 PM

#

oh yeah maybe

#

nvm

#

but its the same thing

#

but its inline(always) anyway

tawdry knoll Feb 1, 2025, 3:59 PM

#

Ahh I see. I guess you always have the fallback option of allocating a zeroed box and then repeatedly moving small amounts of data into it after it's already constructed. I don't remember ever encouraging a stack issue however, I do agree that it should just behave like that by default.

tawdry knoll Feb 1, 2025, 4:19 PM

#

The thing in my eyes is that, yes -- Rust has a lot of growing pains (especially around osdev) and is kinda difficult to use when navigating around them. However, when it works it works really well, and for most things (that I care about at least) are actively getting worked on. In the grand scheme of things, Rust is still a baby language. I mean yeah obviously you have Zig and others that might not have some of the same annoyances, but I find that Rust's rules are much more complex than Zig. I cannot go back to not having proc_macros and not having fmt etc... lol

tawdry knoll Feb 1, 2025, 4:31 PM

#

untold basin i don't really know, tbh

Something you could look into is using the defmt crate. Its a much lighter version of fmt which is quite large for low level stuff. I've seen fmt consume ~30k of executable size alone, and without it you should notice much smaller executables. I think its like this because dyn has quite a bit of executable complexity, and the default formatter is optimized for zero branching. Oh, and you should disable stack unwinding. You can still get stack traces, just that Rust cannot recover from a panic and unwind its stack like C++'s exception can. Its not impossible to get executable sizes really tiny, just a bit of work.

untold basin Feb 1, 2025, 4:31 PM

#

tawdry knoll Something you could look into is using the `defmt` crate. Its a much lighter ver...

i obviously have unwind off

#

i dont have panic-unwind anyway

tawdry knoll Feb 1, 2025, 4:31 PM

#

Ahh okay good

untold basin Feb 1, 2025, 4:31 PM

#

i have my own unwinding logic though, for stacktraces

#

and i build with frame pointers on

tawdry knoll Feb 1, 2025, 4:32 PM

#

yeah thats fine, shouldn't contribute much to large executables

untold basin Feb 1, 2025, 4:32 PM

#

the tables are actually pretty big tbf

#

but its not THAT much

#

i might look at defmt yeah

tawdry knoll Feb 1, 2025, 4:33 PM

#

like are you including your debug symbols in your executable?

untold basin Feb 1, 2025, 4:33 PM

#

no

#

i am taking dwarf tables

#

doing custom magic

#

then including the results of custom magic in the final executable

#

the tables are a lot smaller than dwarf

#

still quite big but not ruinously big

#

isn't defmt serialization-only?

#

which kinda sucks

tawdry knoll Feb 1, 2025, 4:34 PM

#

ahh I was about to say that Rust's symbols are really large lol, I have seen ELFs >2Mib for a <100Kib final executable

#

Anyway, hopefully you end up figuring out something that works. Good luck on the project!

#

Oh and one last thing, if you haven't already, you can make custom compile profiles in your Cargo.toml file to enable some optimizations even in debug mode. Because in debug mode Rust will stub out every BinOp to a function to ensure it doesn't overflow, this gets optimized out in Release mode. You can disable this check without enabling other optimizations, so it's still easier to debug

untold basin Feb 1, 2025, 4:45 PM

#

yeah i know

#

i have it at opt-level=1

tawdry knoll Feb 1, 2025, 4:46 PM

#

But do you have overflow-checks=false

untold basin Feb 1, 2025, 4:46 PM

#

no

#

i want overflow checks though

#

they are useful for debugging

tawdry knoll Feb 1, 2025, 4:48 PM

#

Yeah unfortunately they are pretty big, and I also leave them on were I can... Idk I wouldn't stress about debug's executable size too much unless it's causing tons of issues.

untold basin Feb 1, 2025, 4:54 PM

#

even the release ones are massive though

#

which kinda is a problem

zenith lion Feb 1, 2025, 5:00 PM

#

untold basin even the release ones are massive though

https://github.com/johnthagen/min-sized-rust

GitHub

GitHub - johnthagen/min-sized-rust: 🦀 How to minimize Rust binary s...

🦀 How to minimize Rust binary size 📦. Contribute to johnthagen/min-sized-rust development by creating an account on GitHub.

#

idk if you already do all these

untold basin Feb 1, 2025, 5:01 PM

#

i dont have location details off

tawdry knoll Feb 1, 2025, 5:01 PM

#

untold basin even the release ones are massive though

Release mode with opt-level="z" galaxybrain

untold basin Feb 1, 2025, 5:02 PM

#

and fmt::Debug

#

because they both break shit quite aggressively

zenith lion Feb 1, 2025, 5:04 PM

#

true but

#

why rely on debug by default??

untold basin Feb 1, 2025, 5:05 PM

#

i dont rely on it

#

other than like

#

actually maybe i dont at all

zenith lion Feb 1, 2025, 5:09 PM

#

for release build where you need no debug

#

it makes perfect sense

untold basin Feb 1, 2025, 5:17 PM

#

ehh

#

not really

untold basin Feb 7, 2025, 5:31 PM

#

@vapid echo please tell me more

vapid echo Feb 7, 2025, 5:31 PM

#

idk if you remember the library with rust-like functions

untold basin Feb 7, 2025, 5:31 PM

#

sounds interesting

vapid echo Feb 7, 2025, 5:31 PM

#

the one with the hamt and stuff

#

https://github.com/nyx-os/atlas

GitHub

GitHub - nyx-os/atlas: Rich freestanding C++ library

Rich freestanding C++ library. Contribute to nyx-os/atlas development by creating an account on GitHub.

#

that

untold basin Feb 7, 2025, 5:31 PM

#

i see

#

.hpp tho

vapid echo Feb 7, 2025, 5:31 PM

#

meme

untold basin Feb 7, 2025, 5:31 PM

#

oh also you need the std header crap

vapid echo Feb 7, 2025, 5:32 PM

#

yeah

untold basin Feb 7, 2025, 5:32 PM

#

i dont have it

vapid echo Feb 7, 2025, 5:32 PM

#

so add it meme

untold basin Feb 7, 2025, 5:32 PM

#

why?

#

i have c++20 vomit instead

vapid echo Feb 7, 2025, 5:32 PM

#

freestanding headers good

untold basin Feb 7, 2025, 5:32 PM

#

including these thingies

untold basin Feb 7, 2025, 5:32 PM

#

vapid echo freestanding headers good

what does it really give me tho

vapid echo Feb 7, 2025, 5:33 PM

#

concepts thingies

#

and standard types

untold basin Feb 7, 2025, 5:33 PM

#

vapid echo concepts thingies

they are trivial to write

vapid echo Feb 7, 2025, 5:33 PM

#

yes

untold basin Feb 7, 2025, 5:33 PM

#

vapid echo and standard types

i have types that are standard for me :^)

#

vapid echo Feb 7, 2025, 5:33 PM

#

the only thing you really need is like

#

#

I think

untold basin Feb 7, 2025, 5:33 PM

#

i have my own thing

#

that i use

#

idk

vapid echo Feb 7, 2025, 5:34 PM

#

whatever idrc I just thought it might be cool to use

#


    auto v = arr.iter().collect<Vec<int>>();

#

literally rust

#

meme

untold basin Feb 7, 2025, 5:34 PM

#

vapid echo ```c++ auto v = arr.iter().collect<Vec<int>>(); ```

yeah true

#

that looks cool

#

ill think about it

vapid echo Feb 7, 2025, 5:34 PM

#

idk if u were thinking about using frigg

#

but it's basically it but better meme

untold basin Feb 7, 2025, 5:35 PM

#

vapid echo idk if u were thinking about using frigg

no

#

i was not

#

your thing definitely looks better than frigg

#

idk about the names but like

#

the rest is fine

#

also i dont like std

vapid echo Feb 7, 2025, 5:36 PM

#

untold basin idk about the names but like

wdym

untold basin Feb 7, 2025, 5:36 PM

#

like why have StringView

#

etc

#

tbh personally i prefer just using Span<const u8> for strings if possible

vapid echo Feb 7, 2025, 5:36 PM

#

String is an allocated string and StringView is well, a string view

#

lol

untold basin Feb 7, 2025, 5:36 PM

#

but yeah whatever it doesnt matter

vapid echo Feb 7, 2025, 5:37 PM

#

untold basin tbh personally i prefer just using Span<const u8> for strings if possible

u can just use StringView::as_slice()

untold basin Feb 7, 2025, 5:37 PM

#

true

#

idk

#

ill think about it™️

#

no bounds check cringe https://github.com/nyx-os/atlas/blob/master/include/atlas/string_view.hpp#L21

vapid echo Feb 7, 2025, 5:37 PM

#

ah I may have forgotten

#

I have it in slices though

#

but yea whether you end up using it or not, I'll probably look into contributing to your thing cuz it would be cool as fuck to boot my OS using a custom firmware

untold basin Feb 7, 2025, 5:39 PM

#

trueee

untold basin Feb 7, 2025, 6:23 PM

#

okay i think im gonna have to do bootloader stuff for shkwve now

zenith lion Feb 7, 2025, 6:45 PM

#

untold basin including these thingies

good stuff

zenith lion Feb 7, 2025, 6:45 PM

#

untold basin

based

#

make it rust FFI-able

#

it would probably need a fork of the rust stdlib tho

#

because rust does not have a standard struct layout in memory

#

you'd need to make rust stuff #[repr(C)]

#

but it would be cool to be able to pass any rust stdlib type (that your thing also defines in C++) over an ffi boundary

untold basin Feb 7, 2025, 6:49 PM

#

zenith lion make it rust FFI-able

but why

#

im doing it all in C++ now

#

fuck rust :^)

zenith lion Feb 7, 2025, 6:50 PM

#

DON'T FUCK RUST

#

YOU MIGHT GET TETHANOSIS

untold basin Feb 8, 2025, 12:31 PM

#

fun!

#

looks like im targetting c++2c then

#

ugh "const unsigned char" is not """similar""" to "const char"

#

i have -funsigned-char too lol

#

so its like. literally the same type

untold basin Feb 8, 2025, 1:08 PM

#

okay so lets get shkwve++ booting

lofty hearth Feb 8, 2025, 1:09 PM

#

untold basin so its like. literally the same type

except it isn't because lol

untold basin Feb 8, 2025, 1:09 PM

#

it has the same size, alignment, signdness, layout

lofty hearth Feb 8, 2025, 1:10 PM

#

std::is_same<unsigned char, char> || std::is_same<signed char, char> is always false

untold basin Feb 8, 2025, 1:10 PM

#

yeah i know i know

untold basin Feb 8, 2025, 1:10 PM

#

lofty hearth `std::is_same<unsigned char, char> || std::is_same<signed char, char>` is always...

or in custom c++ goop: is::Same<unsigned char, char> is false :^)

untold basin Feb 8, 2025, 3:42 PM

#

okay now i have a custom "span" helper type in my C++ template goop

#

fun!

#

okay wtf

#

how do i make my initializer_list not weird

devout swift Feb 8, 2025, 4:17 PM

#

untold basin how do i make my initializer_list not weird

does it only have begin/end ptrs or alternatively begin ptr + size?

#

those are the only variants of it that are supported

untold basin Feb 8, 2025, 4:17 PM

#

yeah i figured

#

and the begin pointer must be const T*

#

lol

#

now i have a Vec<T>

vapid echo Feb 8, 2025, 4:28 PM

#

atlas::Vec<T> though 😭

#

nah its ok I forgive you

untold basin Feb 8, 2025, 4:29 PM

#

vapid echo atlas::Vec<T> though 😭

doesn't integrate with my print() ~~function~~ macro

vapid echo Feb 8, 2025, 4:30 PM

#

what I did for my fmt thing is I have a Formatter<T> class that types can extend

#

so you can do like

#

struct Formatter<Vec<T>> {}

#

then do a display function

untold basin Feb 8, 2025, 4:30 PM

#

thats cool but consider this

#

i don't have one

#

:^)

#

i probably could add one tbh

vapid echo Feb 8, 2025, 4:31 PM

#

https://github.com/nyx-os/atlas/blob/master/include/atlas/fmt.hpp#L94

#

like that

untold basin Feb 8, 2025, 4:31 PM

#

ah thats ugly too

#

i just have global state

#

for printing

vapid echo Feb 8, 2025, 4:31 PM

#

cringe

untold basin Feb 8, 2025, 4:32 PM

#

its a bootloader

#

i do not need nonglobal printing anyway

untold basin Feb 8, 2025, 4:32 PM

#

untold basin for printing

because that integrates well with my ~~evil~~ wonderful print macro

vapid echo Feb 8, 2025, 4:32 PM

#

well in my kernel

#

I just have a global sink

untold basin Feb 8, 2025, 4:32 PM

#

ah yeah that makes sense

#

tbh

#

passing around a sink doesnt produce good codegen tho

vapid echo Feb 8, 2025, 4:33 PM

#

I checked

#

and it wasnt that bad

#

a hello world was roughly the same size as frigg (or smaller? I forgor)

untold basin Feb 8, 2025, 4:40 PM

#

ok makes sense

#

okay i have SmallVec

#

okay now im gonna add a refcounted pointer i think

#

okay now i have Box and Ref

#

now i will create a String type

#

a totally unique idea i did not steal from anyone else!

vapid echo Feb 8, 2025, 5:23 PM

#

I called my Ref<> Rc<> meme

untold basin Feb 8, 2025, 6:01 PM

#

okay im gonna use a Formatter type

#

its definitely a better approach

#

tho it has some disadvantages, especially around overload resolution

#

but its probably a better approach

vapid echo Feb 8, 2025, 6:08 PM

#

literal plagiarism

untold basin Feb 8, 2025, 6:12 PM

#

noooo

#

im not writing to a format buffer

#

also my impl puts __thunk on every function which is why its totally unique!

#

#define __thunk [[gnu::always_inline]] [[gnu::artificial]] inline

vapid echo Feb 8, 2025, 6:13 PM

#

I just use inline

#

I let the compiler inline stuff if it deems it fit

untold basin Feb 8, 2025, 6:15 PM

#

yeah probably a good idea

#

i should stop putting it on like everything lol

#

i now have this distinction too

#

which prints as expected

glad bough Feb 8, 2025, 6:24 PM

#

hi

#

what are you doing

untold basin Feb 8, 2025, 6:24 PM

#

oh and other things like this

untold basin Feb 8, 2025, 6:24 PM

#

glad bough what are you doing

implementing random C++ data structure crap for my new rewrite

glad bough Feb 8, 2025, 6:24 PM

#

what rewrite

untold basin Feb 8, 2025, 6:25 PM

#

yeah im rewriting shkwve!

vapid echo Feb 8, 2025, 6:25 PM

#

untold basin oh and other things like this

bro

untold basin Feb 8, 2025, 6:25 PM

#

in c++

glad bough Feb 8, 2025, 6:25 PM

#

💀

#

bro

#

at least don't use go for building then

untold basin Feb 8, 2025, 6:25 PM

#

vapid echo bro

hey. 100% from https://paste.sr.ht/~pitust/631cc4f2c6cb9decaf26fe28bf9d6fdbcb5d0d67

untold basin Feb 8, 2025, 6:25 PM

#

glad bough at least don't use go for building then

why?

#

oh yeah good point

#

ill still keep the go stuff

#

ill just make it generate ninja for depfile handling

#

because having symbolized stacktraces is like. a cool thing

vapid echo Feb 8, 2025, 6:26 PM

#

untold basin hey. 100% from https://paste.sr.ht/~pitust/631cc4f2c6cb9decaf26fe28bf9d6fdbcb5d0...

mhm

#

I think my way is cleaner

#

its very strange tho

untold basin Feb 8, 2025, 6:26 PM

#

what is?

#

my old shit?

vapid echo Feb 8, 2025, 6:26 PM

#

my way

untold basin Feb 8, 2025, 6:26 PM

#

ah yeah i see

#

nah i like my impl

vapid echo Feb 8, 2025, 6:26 PM

#

bunch of decltype stuff

#

https://github.com/nyx-os/atlas/blob/master/include/atlas/iter.hpp#L165

untold basin Feb 8, 2025, 6:27 PM

#

oh you mean for collect?

#

or like in general

#

tbh i like the iterator vs cursor distinction

#

but yeah anyway maybe i should make it run on bare metal

vapid echo Feb 8, 2025, 6:28 PM

#

in general

#

 return Iterator<decltype(next_func)>(next_func);

#

this kind of stuff

untold basin Feb 8, 2025, 6:28 PM

#

untold basin but yeah anyway maybe i should make it run on bare metal

since unless i want to go wild with implementing literally every DS possible

#

im kinda out of things to implement

zenith lion Feb 8, 2025, 6:41 PM

#

untold basin now i have a Vec<T>

THIS IS JUST RUST

#

does your formatter support {:x} and stuff

#

i did that for mine in C++

#

it was tough

untold basin Feb 8, 2025, 6:42 PM

#

zenith lion does your formatter support {:x} and stuff

no

#

also i produce better codegen than rust does

zenith lion Feb 8, 2025, 6:43 PM

#

because i was using very recent C++ features I was just learning

zenith lion Feb 8, 2025, 6:43 PM

#

untold basin also i produce better codegen than rust does

in what way?

#

example?

#

for the formatter?

untold basin Feb 8, 2025, 6:44 PM

#

zenith lion for the formatter?

yeah

#

i dont do loop stuff

#

or like slice stuff

#

uh

#

rust produces really bad codegen for println! lol

#

https://godbolt.org/z/EqxshbPTa

Compiler Explorer - Rust (rustc 1.84.0)

#[no_mangle]
pub fn square(num: i32) -> i32 {
println!("foo: {}", 3);
num * num
}

zenith lion Feb 8, 2025, 6:48 PM

#

Screenshot_2025-02-08-21-48-12-796-edit_com.android.chrome.jpg

#

since when is my tabs :D

untold basin Feb 8, 2025, 6:48 PM

#

it means >100

#

lol

#

my codegen looks more like this

zenith lion Feb 8, 2025, 6:49 PM

#

it was ♾️ before

untold basin Feb 8, 2025, 6:49 PM

#

the Formatter<Span<int32_t>>::debug stuff has a bunch of logic etc

#

(putchar is temporary its gonna turn into real debug io stuff later)

#

but its all fully monomorphised

#

and i parse format strings at compile time

untold basin Feb 9, 2025, 9:49 AM

#

@vapid echo lmao atlas doesnt even pass its own test suite

#

oh and now i managed to induce a crash

#

  TEST_CASE("collide") {
    struct FakeHash {
      uint64_t operator()(uint64_t a, size_t gen = 0) const {
        (void)a;
        return gen;
      }
    };

    Hamt<uint64_t, uint64_t, TracingAllocator, FakeHash> h;
    h.insert(1, 1);
    h.insert(2, 2);
    h.insert(3, 3);
    h.insert(4, 4);
    h.remove(1);
    h.dump();
    if (Option<uint64_t> v = h.get(4)) {
      CHECK(v.unwrap() == 4);
    }
  }

#

you should prooobably handle this kind of endless hash collision

#

    struct FakeHash {
      uint64_t operator()(uint64_t a, size_t gen = 0) const {
        return gen ? a : 0;
      }
    };

    Hamt<uint64_t, uint64_t, TracingAllocator, FakeHash> h;
    h.insert(1, 1);
    h.insert(2, 2);
    h.insert(3, 3);
    h.insert(4, 4);
    h.remove(1);
    if (Option<uint64_t> v = h.get(4)) {
      CHECK(v.unwrap() == 4);
    } else {
      CHECK(false);
    }

#

and this fails in the CHECK(false)

#

i wanted to see how you dealt with these kinds of collisions and the answer is that you do not KEKW

#

tbh i might externally allocate entries

#

its less efficient

#

but its easier too so

#

actually fuck it i dont need a hashmap for now

#

or any map

#

i might port frigg's rbtree

#

or implement an AA tree or something

#

idk

#

okay it is time to make it do os stuff

vapid echo Feb 9, 2025, 1:44 PM

#

untold basin <@379432087620550658> lmao atlas doesnt even pass its own test suite

Huh?

#

All passes for me

#

You probably need to run it in the proper directory

#

I read an elf file

untold basin Feb 9, 2025, 1:46 PM

#

sounds like a broken test suite

#

but anyway i foud two different breakages with your HAMT too so

#

lol

vapid echo Feb 9, 2025, 1:46 PM

#

well

#

Those issues are the hash's fault

#

Not mine

#

trl

untold basin Feb 9, 2025, 1:47 PM

#

the first one yes

#

the second one no

#

the second issue is your 100% fault

vapid echo Feb 9, 2025, 1:48 PM

#

untold basin ```c++ struct FakeHash { uint64_t operator()(uint64_t a, size_t gen = ...

This one?

untold basin Feb 9, 2025, 1:48 PM

#

yes

#

its a valid hash

#

this is the case if you get a first-level hash collision

#

which can happen

vapid echo Feb 9, 2025, 1:48 PM

#

I did extensively test collisions

untold basin Feb 9, 2025, 1:49 PM

#

your testing was clearly crap

#

since it took me 5 minutes to break it

vapid echo Feb 9, 2025, 1:49 PM

#

bro I tried like a 4mb word dictionary

#

There were collisions

#

And it handled them

untold basin Feb 9, 2025, 1:49 PM

#

idk man

#

my tests fails

#

and it very much shouldnt

vapid echo Feb 9, 2025, 1:50 PM

#

I'll see what I can do

#

But it's weird considering I did test collisions like I said

untold basin Feb 9, 2025, 1:51 PM

#

its called testing using brain instead of using brute force meme

#

also i almost got my shit build!

#

:^)

#

(almost)

#

progress

#

it links

vapid echo Feb 9, 2025, 2:04 PM

#

untold basin its called testing using brain instead of using brute force <:meme:5754450945896...

I distinctively remember searching for collisions for testing them

untold basin Feb 9, 2025, 2:06 PM

#

vapid echo I distinctively remember searching for collisions for testing them

idk sounds like you just missed an edge case

vapid echo Feb 9, 2025, 2:10 PM

#

yea so to run tests you need to call ./build/tests directly

#

ninja -C build test wont work

untold basin Feb 9, 2025, 2:10 PM

#

i did ./tests

vapid echo Feb 9, 2025, 2:10 PM

#

because of the elf thing I told you

untold basin Feb 9, 2025, 2:10 PM

#

lol

#

okay yeah whatever

#

same thing

#

i commented out that test and it worked

#

anyway time to fix arm bullshit

#

hmm weird it should work fine

#

wtf

vapid echo Feb 9, 2025, 2:24 PM

#

ok i fixed it

#

https://github.com/nyx-os/atlas/blob/master/include/atlas/hamt.hpp#L409-L418

#

this seemed to be the problem

#

just removing that condition fixes it

untold basin Feb 9, 2025, 2:32 PM

#

esr suggests L3 translation fault

#

the ptes have bits 0x403 except for the last level which is 0x401

vapid echo Feb 9, 2025, 2:36 PM

#

vapid echo https://github.com/nyx-os/atlas/blob/master/include/atlas/hamt.hpp#L409-L418

yeah that was definitely a wrong codepath

untold basin Feb 9, 2025, 2:36 PM

#

403 is ACCESS|VALID|TABLE

vapid echo Feb 9, 2025, 2:36 PM

#

I tested with the huge word dictionary

untold basin Feb 9, 2025, 2:36 PM

#

401 is ACCESS|VALID

vapid echo Feb 9, 2025, 2:36 PM

#

and it doesnt get hit once

untold basin Feb 9, 2025, 2:36 PM

#

does my test pass now?

vapid echo Feb 9, 2025, 2:36 PM

#

yeah

untold basin Feb 9, 2025, 2:37 PM

#

cool

lofty hearth Feb 9, 2025, 2:37 PM

#

untold basin 401 is ACCESS|VALID

i think the page also needs bit 0 set?

untold basin Feb 9, 2025, 2:37 PM

#

lofty hearth i think the page also needs bit 0 set?

isnt that the bit 0?

#

VALID is 0

lofty hearth Feb 9, 2025, 2:37 PM

#

or bit 1

#

inline constexpr uint64_t kPageValid = 1;
inline constexpr uint64_t kPageTable = (1 << 1);
inline constexpr uint64_t kPageL3Page = (1 << 1);

#

from mammogram

untold basin Feb 9, 2025, 2:37 PM

#

huhhh

lofty hearth Feb 9, 2025, 2:38 PM

#

i think the only case bit 1 is unset in a valid entry is 2M/1G/512G blocks

untold basin Feb 9, 2025, 2:38 PM

#

okay thanks!

#

yeah that would make sense i guess

vapid echo Feb 9, 2025, 2:38 PM

#

I should probably make the elf test suite better

#

so it doesnt not work when calling ninja -C build test

lofty hearth Feb 9, 2025, 2:38 PM

#

also page tables don't need ACCESS set

#

iirc those bits are just ignored for non-leaf entries

untold basin Feb 9, 2025, 2:39 PM

#

i see

#

still nope tho

#

okay yeah now its crashing with FAR=0x18

#

and an L1 translation abort

vapid echo Feb 9, 2025, 2:44 PM

#

thanks for the bug report btw i wouldnt have caught it

untold basin Feb 9, 2025, 3:33 PM

#

okay next problem

#

i need to map peripherals

#

okay i have boot0 working (tho with a massive hack)

#

okay now the hack is gone too

untold basin Feb 9, 2025, 4:02 PM

#

boot1!

#

the bad news is that weird things are happening

#

like it shouldnt reach MISC_abort_slow

#

ever

untold basin Feb 9, 2025, 4:20 PM

#

yeah it works at -O0

#

and i get a garbage elf at -O1

#

wtf

#

oh

#

oh god

#

i hate LLVM's "lets just optimize out random """empty""" loops" pass

#

it optimized out my panic code

#

oh!

#

its a clang bug

#

fixed in 19

#

fedora only packages 18

#

grumble

#

and then i get to an allocator failure which is expected

#

cool

quartz storm Feb 9, 2025, 4:27 PM

#

untold basin its a clang bug

that one time when the UB is actually a compiler bug lol

untold basin Feb 9, 2025, 4:27 PM

#

quartz storm that one time when the UB is actually a compiler bug lol

i dont actually know if it is tbh

#

the standard is stupid here

#

newer clang fixes it

#

C mode fixes it too

quartz storm Feb 9, 2025, 4:29 PM

#

afaik for(;;) { /* some code with no 'break' */ } cannot be removed

#

fear the Big Bad Optimizing Compiler

untold basin Feb 9, 2025, 4:30 PM

#

quartz storm afaik `for(;;) { /* some code with no 'break' */ }` cannot be removed

by a standard compiler, no

#

by clang++ versions 18 and older, yes

#

also thats not true

#

for (;;) {} cannot be removed

#

for (;;) { /* any statement */ } can

quartz storm Feb 9, 2025, 4:32 PM

#

💀

lofty hearth Feb 9, 2025, 4:33 PM

#

untold basin `for (;;) {}` cannot be removed

was that a DR to a previous standard or changed in a new standard

#

not that i expect the compiler to actually change optimizations based on -std=c++XY but i'm curious

devout swift Feb 9, 2025, 4:45 PM

#

lofty hearth was that a DR to a previous standard or changed in a new standard

pretty sure it was a dr

#

and clang was basically the only compiler that optimized it so aggressively

untold basin Feb 9, 2025, 4:51 PM

#

lmao

untold basin Feb 9, 2025, 4:52 PM

#

lofty hearth not that i expect the compiler to actually change optimizations based on `-std=c...

im on -std=gnu++2c too so

lofty hearth Feb 9, 2025, 4:53 PM

#

fair enough

#

can't wait for c++2d

untold basin Feb 9, 2025, 4:54 PM

#

lol

lofty hearth Feb 9, 2025, 4:58 PM

#

hmm there is no way to fit more than 4 standards into a decade right?

#

hmm no ... 20 23 26 29 32 35 38 41 44 47 50 ...

untold basin Feb 9, 2025, 4:59 PM

#

lofty hearth hmm there is no way to fit more than 4 standards into a decade right?

10/3 lol

lofty hearth Feb 9, 2025, 4:59 PM

#

shhhh

untold basin Feb 9, 2025, 5:00 PM

#

the max you can do is obviously 00, 03, 06, 09

#

although

#

09 12 15 18 21 would work if they were really slow for 09 and really fast for 21

lofty hearth Feb 9, 2025, 5:02 PM

#

i was more interested in calendar decades, so that you get c++X{a,b,c,d,e} or whatever

#

but that obviously can't happen

untold basin Feb 9, 2025, 5:06 PM

#

ahh ahh

untold basin Feb 10, 2025, 11:51 AM

#

#

handoff

#

memory map from C++ (boot1)

#

and from C (boot0)

#

the 0x980 LoaderImage is zeros for padding

#

okay now it even runs with TCR_EL1.EPD0=1

#

https://git.sr.ht/~pitust/shkwve-v2

#

it crashes on no malloc atm though

#

and its ugly as hell

untold basin Feb 10, 2025, 12:38 PM

#

okay i think its time to write a shitty malloc

untold basin Feb 10, 2025, 1:24 PM

#

and by "write" i mean "steal"

#

and ill probably have to write a new one because im stupid

#

but its fine

#

okay so now that i have a ~~stolen~~ completely novel memory allocator, what's next

#

@agile nexus i hit that incompatible uintmax_t type thing too btw

#

also. you should stop vomiting global state into the public symbol namespace with the state global

agile nexus Feb 10, 2025, 11:06 PM

#

untold basin <@337185227719966721> i hit that incompatible uintmax_t type thing too btw

Bloody hell, thanks for letting me know. Just checking you're on the latest commit? Should be 72d3f85.
Ahah yeah that's fair about the global, I'll change that.

untold basin Feb 11, 2025, 7:38 AM

#

agile nexus Bloody hell, thanks for letting me know. Just checking you're on the latest comm...

you pass a uintmax_t* to a size_t*

#

tbh i forgot where because i just typedefed them together lol

#

also the global is the only way of getting the fdt root

#

other than path stuff ofc

agile nexus Feb 12, 2025, 11:43 AM

#

untold basin tbh i forgot where because i just typedefed them together lol

lol, I'll have a look at this soon hopefully

untold basin Feb 12, 2025, 7:06 PM

#

okay heres my new device subsystem tree thingy

quartz storm Feb 12, 2025, 9:00 PM

#

untold basin okay heres my new device subsystem tree thingy

does this have any use besides being nice to look at? and what is nextLetter?

untold basin Feb 12, 2025, 9:00 PM

#

quartz storm does this have any use besides being nice to look at? and what is nextLetter?

its a link

#

its test stuff

#

yes it will be used for things

#

when i do more impl shit

#

laterrrr

quartz storm Feb 12, 2025, 9:04 PM

#

untold basin its a link

but like... what is the point of nextLetter? if this is part of the filesystem hierarchy (like devfs or sysfs), code can just iterate over readdir

untold basin Feb 12, 2025, 9:04 PM

#

quartz storm but like... what is the point of nextLetter? if this is part of the filesystem ...

its a test entry

#

lol

#

like its a symlink type of thing

quartz storm Feb 12, 2025, 9:05 PM

#

yeah

#

but what does it permit software to do that it cannot without it?

#

also, assuming a usual VFS, is it not better to have the symlink contents be ../b instead of /b? the latter would lookup b in the filesystem root

#

or maybe it is implemented as a 'magic symlink' and it does not matter, idk.

untold basin Feb 12, 2025, 9:10 PM

#

quartz storm also, assuming a usual VFS, is it not better to have the symlink contents be `.....

thats not how its implemented no

untold basin Feb 12, 2025, 9:10 PM

#

quartz storm but what does it permit software to do that it cannot without it?

its for introspection stuff

#

like for modelling relations between devices

#

or whatever

quartz storm Feb 12, 2025, 9:11 PM

#

but is there any meaningful relationship between two "adjacent" devices on a bus, other than belonging to the same bus?

untold basin Feb 12, 2025, 9:12 PM

#

quartz storm but is there any meaningful relationship between two "adjacent" devices on a bus...

no

#

the "nextLetter" entry is just a test thingy

#

to make sure that this mechanism works correctly

quartz storm Feb 12, 2025, 9:13 PM

#

ah

#

ok

untold basin Feb 24, 2025, 5:47 PM

#

okay im back from implementing a shitty raytracer

#

and i decided to redesign the device system yet again

#

// idl, defining a kind of device
class Foo
    poke(u8 thingy) -> void

// a thing that is this kind of device
struct AFoo : Device(Foo) {
    void poke(u8 thingy) { print("poked: {}", thingy); }
};

// user API
Dev<AFoo> root = Dev<AFoo>::make(new AFoo{});
root->poke(0x41_u8);

Dev<Foo> lower = root.downcast<Foo>();
lower->poke(0x42_u8);

#

basically its rust traits except in C++

open osprey Feb 24, 2025, 6:06 PM

#

don't forget to insert a STREAMS implementation as well

untold basin Feb 24, 2025, 6:16 PM

#

open osprey don't forget to insert a STREAMS implementation as well

that sounds like an interesting, although not necessarily good, idea

#

i'm definitely planning to add a network stack

untold basin Feb 24, 2025, 9:24 PM

#

now i have an event thingy as well: ```c++
// subscribtions
static void someHandler(evt::FooRelatedEvent* relev) {
print("handle a foo-related event");
}
static void someHandler2(evt::FooLocation* loc) {
print("handle a foo-located event");
}
EventHandler(someHandler);
EventHandler(someHandler2);

// creation
evt::FooLocation::dispatch(true, false);
evt::FooRelatedEvent::dispatch(true);

// idl
event FooRelatedEvent
bool isRealEvent

event FooLocation : FooRelatedEvent
bool requestCreation

#

with some cheap type casting logic which handles the simple single-inheritance case

#

(i form a tree of the events, then do a dfs over all of the events and assign IDs as i go, so typecheck is just min <= id && id < max)

#

my plan is to allow for events to be "claimed" by device drivers as my mechanism for creating new devices

vapid echo Feb 25, 2025, 2:07 AM

#

untold basin basically its rust traits except in C++

so

#

concepts

untold basin Feb 25, 2025, 7:37 AM

#

no not concepts

untold basin Feb 25, 2025, 7:43 AM

#

vapid echo concepts

concepts are in no way equivalent to rust traits

vapid echo Feb 25, 2025, 11:25 AM

#

untold basin concepts are in no way equivalent to rust traits

They are kinda similar

#

Traits are better tho

untold basin Feb 25, 2025, 2:27 PM

#

okay i think i'll implement STREAMS now

#

maybe

#

for uart

#

less priorities, because i do not need them

#

ig

#

and also other things probably, because looking things up is hard too

untold basin Feb 25, 2025, 5:19 PM

#

okay i have probing going again

#

still need to do clk and pll torture

#

and implement a gic + pl011 driver

sacred spire Feb 25, 2025, 5:32 PM

#

untold basin now i have an event thingy as well: ```c++ // subscribtions static void someHand...

What language is this

untold basin Feb 25, 2025, 5:33 PM

#

sacred spire What language is this

first two C++, last is custom idl thing

#

#include <fdt.hh>
#include <irq.hh>
#include <option.hh>

// generic protocols
class InterruptController
    resolveFDTInterrupt(fdt::Cells cells) -> Option<u32>
    bindIRQ(u32 id, InterruptSlot* slot) -> void

// arm64-specific devices
#if arm64
class ARM64PlatformIC
    handleIRQ()
    handleFIQ()
#endif

// devices
event GICv2Found
    fdt::Node fdtNode

event FDTNode
    fdt::Node fdtNode
    // TODO: also add vmspace when i add that as an abstraction

``` it has a bunch of new maximally cursed features now

#

like #if and #include support

#

which get lowered to C++ as needed

zenith lion Feb 25, 2025, 5:54 PM

#

untold basin like #if and #include support

is that implemented with the c preprocessor?

untold basin Feb 25, 2025, 5:54 PM

#

zenith lion is that implemented with the c preprocessor?

ofc not lol

zenith lion Feb 25, 2025, 5:54 PM

#

well it could be

untold basin Feb 25, 2025, 5:54 PM

#

first of all, not really

#

second of all, why would i do that

zenith lion Feb 25, 2025, 5:54 PM

#

if you make your idl lowered to headers you could include idl and headers

untold basin Feb 25, 2025, 5:54 PM

#

zenith lion if you make your idl lowered to headers you could include idl and headers

uh, the headers define types

#

those definitions arent compatible with my idl parser

zenith lion Feb 25, 2025, 5:55 PM

#

ah ok

untold basin Feb 25, 2025, 7:18 PM

#

okay i implemented a very shitty pl011 "driver"

untold basin Feb 25, 2025, 8:26 PM

#

okay now i need to implement mapping pages

#

and a better allocator

untold basin Feb 26, 2025, 3:00 PM

#

okay i have my own totally unique memcpy! https://paste.sr.ht/~pitust/b9ac15faff533b44ae7d7073ebce626250aad48f

#

totally different from mangarm's

#

[[gnu::artificial]] [[gnu::always_inline]] static inline void memstore(u8*& dst, word value) {
    __builtin_memcpy_inline(dst, &value, wordSize);
    dst += wordSize;
}
[[gnu::artificial]] [[gnu::always_inline]] static inline word memload(const u8*& src) {
    word dst;
    __builtin_memcpy_inline(&dst, src, wordSize);
    src += wordSize;
    return dst;
}

but hey i have this impl which is a lot nicer than the weird enum tricks managarm uses

#

also i mess around with the impl to induce ldp/stp on arm

glad bough Feb 26, 2025, 3:01 PM

#

real

#

what is that white line on that sourcehut link

#

a ruler?

untold basin Feb 26, 2025, 3:01 PM

#

glad bough what is that white line on that sourcehut link

80 columns

#

aka stupid line

#

yes its super distracting

glad bough Feb 26, 2025, 3:02 PM

#

cant you turn it off?

untold basin Feb 26, 2025, 3:02 PM

#

glad bough cant you turn it off?

you can using stylus :^)

untold basin Feb 26, 2025, 3:03 PM

#

untold basin okay i have my own totally unique memcpy! https://paste.sr.ht/~pitust/b9ac15faff...

its the same approach as managarm, except that i induce stp/ldp on arm + i use __builtin_memcpy_inline

glad bough Feb 26, 2025, 3:04 PM

#

lol

untold basin Mar 2, 2025, 9:36 AM

#

okay i think its time to acutally write a reasonable memory allocator

untold basin Mar 3, 2025, 8:58 PM

#

okay i have my shitty malloc now

#

and i have page table logic working i think

#

kinda

#

still need to improve it a bit

untold basin Mar 3, 2025, 9:45 PM

#

okay i have mapping pages down i think

untold basin Mar 4, 2025, 1:50 PM

#

i think i have a bit too many debug assertions

#

also wtf is clang's optimizer doing

#

ahh yes, now that is true clang

#

if (x) { if (x) { do stuff(); call a noreturn function() } if (x) { call a noreturn function() } call a noreturn function() }

#

does clang just not do noreturn analysis at -O1?

untold basin Mar 4, 2025, 2:04 PM

#

untold basin if (x) { if (x) { do stuff(); call a noreturn function() } if (x) { call a nore...

and immediatly after that: if (x) { call a noreturn function() }

untold basin Mar 4, 2025, 2:24 PM

#

tbh it would be cool if c++ had something that worked like #[track_caller]

#

and on one hand i could write a clang plugin to add like, [[track_caller]] or w/e

#

but on the other hand

#

that sounds really fucking hard

#

so i wont

#

~~clangir would make this way easier tho~~

#

or maybe not idk

#

actually it probably would yeah

#

unfortunately its also really fucking broken

untold basin Mar 4, 2025, 2:57 PM

#

okay now i init my pl011+gic driver more better

#

and map pages

#

next step

#

interrupts

#

ugh

#

in newer qemu i get alignment faults

#

in the pre-vm logic

untold basin Mar 4, 2025, 3:52 PM

#

basically assertboot meme

#

okay i routed an irq to the pl011 driver!

#

okay now im handling the dpc

untold basin Mar 4, 2025, 5:00 PM

#

okay now i have serial

#

with interrupts!

#

time for shitty STREAMS

#

or maybe i wont

#

hm

#

okay lets start by entering the hell that is C++20 coroutines

untold basin Mar 4, 2025, 5:27 PM

#

one thing i have to give to rust's async system props for

#

it is way easier to understand

untold basin Mar 4, 2025, 5:51 PM

#

hmm i think i have a race condition somewhere

#

and now i'm reading from a different coroutine via an MPMC

untold basin Mar 4, 2025, 6:21 PM

#

my new set of async primitives is basically just spamming out semaphores tbh

#

i should switch to a proper round robin style scheduler

trim seal Mar 4, 2025, 6:24 PM

#

untold basin with interrupts!

Another successful case of IDT init... LETSFUCKINGGOOOOOO

untold basin Mar 4, 2025, 6:25 PM

#

trim seal Another successful case of IDT init... <:LETSFUCKINGGOOOOOO:1003833035688513606>

actually its VBAR_EL1 init

#

meme

#

and also theres a bunch of other things that go into my interrupt handling because (a) arm moment and (b) pitust shitcode

#

also also i initialize the idt as the second instruction in the proper loader ("boot1")

trim seal Mar 4, 2025, 6:26 PM

#

boring, they should've called it IDT_EL1

untold basin Mar 4, 2025, 6:26 PM

#

ok sorry 5th

#

the first thing is setting up the stack

#

and handoff info

#

the third thing is vbar_el1

#

so things i have to do:

a real round robin scheduler
actual drivers
idk all the other thinsg

#

okay i think ill do more drivers

#

hmm. maybe i can do clk and pll torture

untold basin Mar 4, 2025, 6:58 PM

#

or device selection or something idk

#

okay lets actually do something useful

#

like an fwcfg driver

random elbow Mar 4, 2025, 7:03 PM

#

I like the fact it's written in Go that's pretty cool!

untold basin Mar 4, 2025, 7:05 PM

#

its not written in go

#

lol

#

the go is only the build system

#

the build entrypoint is basically ```sh
#!/bin/sh

set -e

mkdir -p obj

fence_rev=r22
if ! [ -e obj/x ] || ! [ -e obj/fence-buildsrc-$fence_rev ]; then
rm -rf obj
mkdir -p obj

go get ./build-src
go build -o obj/x ./build-src
touch obj/fence-buildsrc-$fence_rev
fi

obj/x "$@"```

#

obj/x is just the main builder program

rose sonnet Mar 4, 2025, 8:39 PM

#

untold basin i think i have a bit too many debug assertions

There is no such thing as having too many assertions assertos

untold basin Mar 4, 2025, 8:39 PM

#

rose sonnet There is no such thing as having too many assertions <:assertos:1082452358732202...

actually thats not true

#

"too many" just means that i cant show off my cool status code system

#

:^)

untold basin Mar 4, 2025, 9:46 PM

#

okay that's fwcfg file listing working

#

huh i wonder what the tpm log is

#

sounds interesting but also really boring

untold basin Mar 4, 2025, 10:17 PM

#

okay i could do ramfb

#

or i could write a shitty virtio driver

#

actually ig pci first

#shkwve: an incomplete kernel/maybe bootloader/UEFI thing/maybe something else if i get bored again

lzcunt software license

Definitions