pmOS - microkernel OS for RISC-V, x86, and LoongArch | Operating System Development | Page 10

carmine nacelle Mar 10, 2026, 4:08 PM

#

I feel like I have to improve lapic/tsc calibration first

#

probably AMD?

#

like they had the whole cpu cores unlocking thing

twilit talon Mar 10, 2026, 4:08 PM

#

CPUs can be software-disabled for many reasons

carmine nacelle Mar 10, 2026, 4:09 PM

#

intel with avx...

#

Can there be repeated entries?

twilit talon Mar 10, 2026, 4:11 PM

#

yes

carmine nacelle Mar 10, 2026, 4:11 PM

#

In MADT

twilit talon Mar 10, 2026, 4:11 PM

#

ofc

carmine nacelle Mar 10, 2026, 4:11 PM

#

but I mean two x2apic entries with the same id

twilit talon Mar 10, 2026, 4:11 PM

#

yes

carmine nacelle Mar 10, 2026, 4:11 PM

#

bruh

twilit talon Mar 10, 2026, 4:11 PM

#

u have to be able to deduplicate them

carmine nacelle Mar 10, 2026, 4:12 PM

#

and what if there's only x2apic entry for id < 255

#

or is this too insane?

twilit talon Mar 10, 2026, 4:12 PM

#

why does it matter

#

u pretend they dont exist at all

carmine nacelle Mar 10, 2026, 4:12 PM

#

fun

twilit talon Mar 10, 2026, 4:13 PM

#

(if lapic entries also exist)

carmine nacelle Mar 10, 2026, 4:13 PM

#

you get less cpu cores sadmeme

twilit talon Mar 10, 2026, 4:13 PM

#

u dont

#

u dont ignore them if there arent any lapic entries

carmine nacelle Mar 10, 2026, 4:30 PM

#

CPU usable is flags & ACPI_PIC_ENABLED or flags & ACPI_PIC_ONLINE_CAPABLE, right?

twilit talon Mar 10, 2026, 4:30 PM

#

static bool cpu_usable(u32 flags)
{
    if (flags & ACPI_PIC_ENABLED)
        return true;

    if (!s_online_capable_bit_usable)
        return false;

    return flags & ACPI_PIC_ONLINE_CAPABLE;
}

carmine nacelle Mar 10, 2026, 4:31 PM

#

oh no

twilit talon Mar 10, 2026, 4:31 PM

#

nah its simple

#

static void setup_online_capable_bit(void)
{
    struct acpi_fadt *fadt;
    uacpi_status st;

    st = uacpi_table_fadt(&fadt);
    BUG_ON(st != UACPI_STATUS_OK);

    /*
     * Online capable bit was introduced in ACPI 6.3, prior revisions might set
     * it erroneously so don't look at it for those versions.
     */
    s_online_capable_bit_usable =
        (fadt->hdr.revision > 6) ||
        (fadt->hdr.revision == 6 && fadt->fadt_minor_verison >= 3);
}

carmine nacelle Mar 10, 2026, 4:33 PM

#

How many helpers are there in uACPI? nooo

twilit talon Mar 10, 2026, 4:33 PM

#

lol

carmine nacelle Mar 10, 2026, 4:34 PM

#

twilit talon ```c static void setup_online_capable_bit(void) { struct acpi_fadt *fadt; ...

So fadt only gets unreferenced once I do the unitialization thing, right?

#

(I forgot what was the function)

twilit talon Mar 10, 2026, 4:35 PM

#

uacpi_table_fadt returns an internal version that doesnt need ref/unref

carmine nacelle Mar 10, 2026, 4:35 PM

#

So uacpi just stores it in its buffer

twilit talon Mar 10, 2026, 4:35 PM

#

yeah, since the raw copy requires a lot of sanitization

carmine nacelle Mar 10, 2026, 4:44 PM

#

So I guess this basically is as correct as it gets (except the smp_wake_everyone_else_up function trl ) https://gitlab.com/mishakov/pmos/-/blob/dev/kernel/arch/x86/cpus/cpus.cc?ref_type=heads#L261

#

And not checking if I'm on bsp

twilit talon Mar 10, 2026, 4:45 PM

#

looks good

hexed acorn Mar 10, 2026, 4:54 PM

#

twilit talon ```c static void setup_online_capable_bit(void) { struct acpi_fadt *fadt; ...

is there no acpi 6.4?

twilit talon Mar 10, 2026, 4:54 PM

#

Wdym?

carmine nacelle Mar 10, 2026, 4:54 PM

#

hexed acorn is there no acpi 6.4?

wdym

hexed acorn Mar 10, 2026, 4:54 PM

#

ah no i misread the check

#

i thought the latter part was == 3

twilit talon Mar 10, 2026, 4:55 PM

#

The current version is 6.6 I think

carmine nacelle Mar 10, 2026, 4:55 PM

#

I have 6.5 in my osdev folder

twilit talon Mar 10, 2026, 4:55 PM

#

It came out in like December

carmine nacelle Mar 10, 2026, 9:29 PM

#

Small update: I've started publishing PCI devices to pmbus as well, but it's currently broken because of broken puts() or whatever Rust is using nooo

#

Maybe I should just finally switch to mlibc? thonk

#

Though I wanted to implement futex first coolmeme

#

I think I'm going insane, why is my todo list so large nooo

carmine nacelle Mar 10, 2026, 11:08 PM

#

This is cool

#

I think this can all fall nicely into place

#

I don't have energy to do it right now, but I think I can add handles to the interrupts in kernel (which can use iommu as well), which can act as a permission system

#

Then also do this, with a different handle

carmine nacelle Mar 10, 2026, 11:11 PM

#

carmine nacelle This is cool

(anyway, now I can match AHCI devices in a better form, I think I'm going to make my ahci driver use this, then do the disk thing, then hopefully I can have disk?)

carmine nacelle Mar 11, 2026, 4:42 PM

#

Wtf

#

It just loaded, and died

carmine nacelle Mar 11, 2026, 5:01 PM

#

I think the kernel OOMed

carmine nacelle Mar 11, 2026, 5:29 PM

#

Ok, this looks extremely promising (it's failing because I haven't implemented that syscall yet), and it found a mouse and a keyboard, and tried to start the service twice (even though i8042 should have only started once), so I can probably make it restart servers as well with very little work

#

I can probably even forgo dependencies for servers for now, and make it start servers as the things get published in pmbus

#

Because the idea is to have devicesd launch and publish pci devices -> this will start ahci server which will publish the disk -> this will start the disks server, which will parse the partition tables (iirc I had code for that already) -> which will start the filesystem drivers, which can finally start the vfs or whatever

#

(I need to decide how to handle the mountpoints)

carmine nacelle Mar 11, 2026, 5:35 PM

#

carmine nacelle Because the idea is to have devicesd launch and publish pci devices -> this will...

And all of this is just passing capabilities between everything

#

Idk, I have so many ideas for this

#

(I really have to study though, so I don't have time for this now)

#

(famous last words)

carmine nacelle Mar 11, 2026, 9:46 PM

#

carmine nacelle Ok, this looks extremely promising (it's failing because I haven't implemented t...

This is fully working now
(aparently, I had everything already in the kernel, just needed to make it a syscall...)

carmine nacelle Mar 12, 2026, 8:54 PM

#

I am going insane

#

My yaml parser is trolling me

twilit talon Mar 12, 2026, 9:15 PM

#

lol

carmine nacelle Mar 12, 2026, 9:21 PM

#

It's working, but I'm exhausted

#

Though I'm planning to make 64 bit port bootable with Hyper as well

#

After this

dense carbon Mar 12, 2026, 9:27 PM

#

carmine nacelle My yaml parser is trolling me

what's the yaml for?

carmine nacelle Mar 12, 2026, 9:29 PM

#

dense carbon what's the yaml for?

Init server

#

(I've used libyaml, but I think I'm doing it wrong, since the code I'm writing is horrible)

dense carbon Mar 12, 2026, 9:32 PM

#

yaml is famously complicated

carmine nacelle Mar 12, 2026, 11:07 PM

#

yeah, it seemed really simple though

#

Anyway, this is not big, but it's working

twilit talon Mar 12, 2026, 11:07 PM

#

cool

carmine nacelle Mar 12, 2026, 11:08 PM

#

And this is my limine file, which can hopefully make Hyper work as well

twilit talon Mar 12, 2026, 11:08 PM

#

yeah thats the thing about yaml, if u actually use advanced features its ridiculously complicated

twilit talon Mar 12, 2026, 11:08 PM

#

carmine nacelle And this is my limine file, which can hopefully make Hyper work as well

dont see why not

carmine nacelle Mar 12, 2026, 11:08 PM

#

My i686 port might be broken

#

I haven't tested it in like a year

#

And I've changed a bunch of x86 stuff recently

#

(I mean I unified more of x86_64 and i686)

#

My config trl https://gitlab.com/mishakov/pmos/-/blob/dev/limine/init.yaml?ref_type=heads

GitLab

limine/init.yaml · dev · Mikhail Kovalev / pmOS · GitLab

GitLab.com

twilit talon Mar 12, 2026, 11:12 PM

#

yeah that definitely looks better than shoving everything into name

left dew Mar 12, 2026, 11:14 PM

#

does the init server do device matching to launch the servers?

carmine nacelle Mar 12, 2026, 11:14 PM

#

left dew does the init server do device matching to launch the servers?

yes

#

with the pmbus

left dew Mar 12, 2026, 11:14 PM

#

yeah i figured that much

carmine nacelle Mar 12, 2026, 11:14 PM

#

and then it just starts them and gives them a handle to the device

twilit talon Mar 12, 2026, 11:15 PM

#

Btw what if I have multiple ps2 keyboards

carmine nacelle Mar 12, 2026, 11:16 PM

#

twilit talon Btw what if I have multiple ps2 keyboards

It launches i8042 just once

#

Then that is supposed to publish two ps2 port object (which I haven't implemented yet)

twilit talon Mar 12, 2026, 11:16 PM

#

IMO ALWAYS_ONCE should be implied by the lack of a match specifier

carmine nacelle Mar 12, 2026, 11:16 PM

#

Then that launches the PS/2 server

carmine nacelle Mar 12, 2026, 11:17 PM

#

twilit talon IMO ALWAYS_ONCE should be implied by the lack of a match specifier

I guess, but other servers run without match

twilit talon Mar 12, 2026, 11:17 PM

#

Wdym

carmine nacelle Mar 12, 2026, 11:17 PM

#

I also though about it, and the stuff like POSIX server can interact with the bootstrapd directly

carmine nacelle Mar 12, 2026, 11:17 PM

#

carmine nacelle I guess, but other servers run without match

Like devicesd and pmbus itself just always run

#

(I also wanted to add a mechanism to restart them, and so on)

twilit talon Mar 12, 2026, 11:18 PM

#

Right

carmine nacelle Mar 12, 2026, 11:20 PM

#

Also, I need to "urgently" fix my build system trl

#

(recompiling everything for a different arch is a mess)

twilit talon Mar 12, 2026, 11:22 PM

#

How do u not recompile everything if its a different arch

carmine nacelle Mar 12, 2026, 11:22 PM

#

I mean I do

#

But I have to recompile libc, then reinstall llvm's libs, and then recompile the rest

#

But like it's not automatic

#

Cool, the i686 kernel doesn't compile

twilit talon Mar 12, 2026, 11:23 PM

#

KEKW

carmine nacelle Mar 12, 2026, 11:23 PM

#

For a dumb reason actually

#

Missing includes...

twilit talon Mar 12, 2026, 11:24 PM

#

💀

carmine nacelle Mar 12, 2026, 11:24 PM

#

and missing using namespace ultrameme

#

What even is this error? 💀

#

Cool, it gets to userspace, and crashes after that...

#

(In i686 port)

#

(I mean the userspace crashes)

#

I think I'll fix it tomorrow

#

Which probably implies the kernel fully works

twilit talon Mar 12, 2026, 11:44 PM

#

Thats cool

#

Is that the hyper version

carmine nacelle Mar 12, 2026, 11:44 PM

#

Yeah, ofc

twilit talon Mar 12, 2026, 11:44 PM

#

Cool

carmine nacelle Mar 12, 2026, 11:45 PM

#

Just qemu-system-i386 -hdd pmos-hyper.img -serial stdio

twilit talon Mar 12, 2026, 11:45 PM

#

Thats like 128 megs of ram

carmine nacelle Mar 12, 2026, 11:46 PM

#

who cares trl

#

Maybe I should fix it now?

#

I'm too sleepy

twilit talon Mar 12, 2026, 11:46 PM

#

1 am is rookie time

carmine nacelle Mar 12, 2026, 11:46 PM

#

16 is syscall_create_right

#

which shouldn't be too hard to fix, whatever it is

#

And -ESRCH means that it can't find the task group?

twilit talon Mar 12, 2026, 11:47 PM

#

I hope 15 is create_left trl

carmine nacelle Mar 12, 2026, 11:47 PM

#

SYSCALL_SET_INTERRUPT trl

twilit talon Mar 12, 2026, 11:47 PM

#

Nooo

carmine nacelle Mar 12, 2026, 11:48 PM

#

carmine nacelle SYSCALL_SET_INTERRUPT <:trl:829316641191886888>

the liberalism

carmine nacelle Mar 12, 2026, 11:49 PM

#

carmine nacelle who cares <:trl:829316641191886888>

Though iirc I'm not reclaiming bootloader reclaimable memory with hyper

#

Which might be suboptimal

#

(I mean I removed it from limine as well, because of SMP considerations, but since I do it myself, the other cores shouldn't be touching it anyway, so there shouldn't be race conditions with it)

#

Yeah, the fact it's doing this makes no sense...

#

I'll do it tomorrow

twilit talon Mar 12, 2026, 11:56 PM

#

carmine nacelle Yeah, the fact it's doing this makes no sense...

Doing what?

carmine nacelle Mar 12, 2026, 11:56 PM

#

Failing to create a right

#

It's failing because the rights namespace (a.k.a. the current task group) doesn't exist, which makes no sense

#

At least from my recollection of code that I wrote like a year ago

twilit talon Mar 12, 2026, 11:58 PM

#

Interesting

carmine nacelle Mar 13, 2026, 8:04 AM

#

What actually pagefaulted was syscall_return???

#

And the stack was unmapped for some reason

carmine nacelle Mar 13, 2026, 9:14 AM

#

I've no idea what's going on

#

Or rather, how did 64 bit port work?

#

Huh?

#

I think I just "forgot" to set it on pthread entry

#

Because for some reason I did it in assembly??

#

This looks more promissing...

carmine nacelle Mar 13, 2026, 11:55 AM

#

Ugh

#

Since when are uint64_t s are 8 bytes aligned on x86??

#

Or wait

#

Wait, it's not it

#

bruh

#

Also, since when is movq valid on ia32???

#

Cool, the stack is off...

#

I think the kernel is ignoring the fixed flag

twilit talon Mar 13, 2026, 12:04 PM

#

thats because by default clang assumes a higher baseline

#

so it uses sse unconditionally even for 32 bit

carmine nacelle Mar 13, 2026, 12:05 PM

#

It's not the problem with sse, but that my stack is not where it's supposed to be bruh

twilit talon Mar 13, 2026, 12:05 PM

#

yeah im just saying movq and sse is because of that

carmine nacelle Mar 13, 2026, 12:05 PM

#

Ofc it's available with i686 target

#

I think movq should be fine with misaligned uint64_t though

twilit talon Mar 13, 2026, 12:10 PM

#

yes

carmine nacelle Mar 13, 2026, 12:39 PM

#

I think I'm getting somewhere

carmine nacelle Mar 13, 2026, 12:40 PM

#

carmine nacelle Ugh

This was exploding because of a dumb error in the auxvec serialization logic

#

Which I've changed for 64 bit pointers, but forgot for 32 bit ones

carmine nacelle Mar 13, 2026, 12:41 PM

#

carmine nacelle I think I'm getting somewhere

All of the errors are in Rust...

#

Though why didn't it print the rest of the stuff?

#

Now it just decided to break for no reason...

#

It must be something stupid

carmine nacelle Mar 13, 2026, 4:01 PM

#

The log server isn't starting

#

I found the error trl

#

Ok, i686 is kinda working

#

(what's broken is syscall return...)

#

I think I want to port x86_64 to hyper as well

twilit talon Mar 13, 2026, 4:25 PM

#

I will add a cmdline field to modules soon but I guess u don't need it anymore

carmine nacelle Mar 13, 2026, 4:40 PM

#

Yeah...

twilit talon Mar 13, 2026, 4:41 PM

#

I mean a config is better anyway

carmine nacelle Mar 13, 2026, 5:12 PM

#

carmine nacelle I think I want to port x86_64 to hyper as well

I give up

twilit talon Mar 13, 2026, 5:15 PM

#

Lol why

toxic torrent Mar 13, 2026, 5:15 PM

#

lol

carmine nacelle Mar 13, 2026, 5:16 PM

#

I'm tired and want to fix i8042 driver first

twilit talon Mar 13, 2026, 5:16 PM

#

How did you break it

carmine nacelle Mar 13, 2026, 5:16 PM

#

carmine nacelle I'm tired and want to fix i8042 driver first

I've looked at my hyper code, and there were too many uint32_ts in there

carmine nacelle Mar 13, 2026, 5:17 PM

#

twilit talon How did you break it

I've changed everything around it, and it was kinda left abandoned

#

Like the logic probably works, but I want it to use pmbus, etc.

twilit talon Mar 13, 2026, 5:17 PM

#

Ah

carmine nacelle Mar 13, 2026, 5:19 PM

#

carmine nacelle I've looked at my hyper code, and there were too many uint32_ts in there

I need to unify my memory initialization logic

#

Because it's 2 1k loc files, that are different for 2 protocols I support...

carmine nacelle Mar 13, 2026, 6:08 PM

#

I think I need a new C helper for pmbus...

#

(why am I writing C?)

carmine nacelle Mar 13, 2026, 6:32 PM

#

Also, maybe I should finally implement futex??

carmine nacelle Mar 14, 2026, 1:26 PM

#

i8042 is trolling me (in QEMU!)

carmine nacelle Mar 14, 2026, 3:10 PM

#

This is concerning...

carmine nacelle Mar 14, 2026, 6:01 PM

#

wtf

#

Or maybe it's actually fine

#

But my CPU cooler is like at 100% for some reason

#

Which tells me it's not trl

#

thonk

#

I'm dumb

carmine nacelle Mar 15, 2026, 8:58 AM

#

I think my rbtree impl is trolling me nooo

carmine nacelle Mar 15, 2026, 1:27 PM

#

carmine nacelle I think my rbtree impl is trolling me <:nooo:750466870993223750>

Not anymore KEKW

carmine nacelle Mar 15, 2026, 4:02 PM

#

Now I kinda want to properly do interrupts stuff, like it's done on managarm in kdb server, but I kinda don't have energy for it

carmine nacelle Mar 15, 2026, 4:41 PM

#

Should I just finally do proper interrupt rights? thonk

#

(a.k.a. handles)

carmine nacelle Mar 15, 2026, 11:19 PM

#

Talking of which, I have decided that I will make my rights also be messages in the kernel, so that they can be pushed onto the ports

#

(since then it lets the kernel send messages without having to allocate memory for them)

carmine nacelle Mar 16, 2026, 11:11 AM

#

I know this is basically just a bunch of userspace stuff, but why is it so cool to see it do this? (devicesd -> i8042 -> ps2d, all launching new drivers instances through pmbus)

#

(the ordering here is from async IPC and the fact that I think I don't have SMP enabled in qemu...)

carmine nacelle Mar 16, 2026, 4:06 PM

#

What's concerning now is that my ns16550 driver breaks half the times, for some reason

#

(Which I think is because of broken interrupts)

#

Cool, the keyboard is fixed

#

(totally not overengineered)

#

Time to figure out evdev?

twilit talon Mar 16, 2026, 4:15 PM

#

why do u need evdev even

#

i thought u werent doing linux compat

carmine nacelle Mar 16, 2026, 4:16 PM

#

how do I represent keyboard and mouse data then?

twilit talon Mar 16, 2026, 4:16 PM

#

thats up to you to design

carmine nacelle Mar 16, 2026, 4:16 PM

#

Like I though it was just some standard structs + char devices at /dev/input

twilit talon Mar 16, 2026, 4:16 PM

#

standard for linux yes

carmine nacelle Mar 16, 2026, 4:16 PM

#

Since Managarm and freebsd have it, why not do it as well

twilit talon Mar 16, 2026, 4:17 PM

#

managarm has it because its aiming for linux compatibility

#

although at api not abi level

#

and freebsd literally has linuxlator so yeah

carmine nacelle Mar 16, 2026, 4:17 PM

#

what do I use then??

#

to not reinvent the wheel

#

nooo

twilit talon Mar 16, 2026, 4:18 PM

#

depends on your goals

#

if u dont want to reinvent the wheel do linux compat

carmine nacelle Mar 16, 2026, 4:18 PM

#

I don't want to deal with input tbh

#

All I wanted was drivers

#

Also, I have no idea how human input and terminals and GUIs work

carmine nacelle Mar 16, 2026, 7:36 PM

#

I see Wikipedia mentions STREAMS when talking about line discipline trl

carmine nacelle Mar 16, 2026, 11:27 PM

#

I have decided, I'll do my tty server in C++ trl

#

I like coroutines, and people have said that the Rust's ones are too complicated

plain jungle Mar 17, 2026, 6:56 AM

#

who said that?

#

Rust async/await is fine

left dew Mar 17, 2026, 7:02 AM

#

yeah, no, it’s fine

#

probably easier and better integrated than c++ coroutines

#

tho that comes with a few drawbacks, mainly lifetimes that are wack

plain jungle Mar 17, 2026, 7:44 AM

#

yeah, if you just compare the mechanism, C++ coroutines are probably a bit easier than rust async/await

#

but with rust you can use a lot of existing libraries

left dew Mar 17, 2026, 7:56 AM

#

tbh i have never even taken a look at c++ coroutines, so that's why i think rust futures are easier :^)

carmine nacelle Mar 17, 2026, 10:57 AM

#

Another detail is that I already have stuff which uses C++ coroutines, I just haven't bothered yet to make them handle rights (I mean this kinda also was the reason to implement them in the first place)

carmine nacelle Mar 17, 2026, 11:24 AM

#

Wtf is this

#

My IPC in a nutshell

carmine nacelle Mar 18, 2026, 12:15 PM

#

Ok, my C++ insanity has almost worked first time

#

And is now trolling me nooo

carmine nacelle Mar 18, 2026, 12:46 PM

#

what the fuck

#

Ok, this was actually not a C++ bug

#

This is very promissing (timers are broken again nooo )

carmine nacelle Mar 18, 2026, 1:08 PM

#

Yeah, it's actually kinda mostly working...

carmine nacelle Mar 18, 2026, 4:16 PM

#

Oof

carmine nacelle Mar 18, 2026, 5:21 PM

#

My AHCI is hardcore trolling me

#

It's just calling exit() somewhere, with no error messages

#

Also, I'm fed up with my timers

#

It just breaks half of the time

#

I'm done

carmine nacelle Mar 18, 2026, 5:42 PM

#

It was a very dumb double free

carmine nacelle Mar 18, 2026, 6:36 PM

#

thonk why does limine not load with 128mb of ram?

twilit talon Mar 18, 2026, 6:37 PM

#

Does hyper work

carmine nacelle Mar 18, 2026, 6:37 PM

#

it was last time I checked

twilit talon Mar 18, 2026, 6:38 PM

#

Big

carmine nacelle Mar 18, 2026, 6:38 PM

#

though hyper is still i386 only

#

huge todo list nooo

lunar island Mar 18, 2026, 6:40 PM

#

carmine nacelle <:thonk:575445129863888896> why does limine not load with 128mb of ram?

large kernel modules?

carmine nacelle Mar 18, 2026, 6:40 PM

#

#

(I think I finally need to implement shared libraries though)

lunar island Mar 18, 2026, 6:42 PM

#

this is still smaller than a rust hello world trl

carmine nacelle Mar 18, 2026, 6:43 PM

#

pmbus is in rust KEKW

lunar island Mar 18, 2026, 6:43 PM

#

damn

carmine nacelle Mar 18, 2026, 8:21 PM

#

I think the time has come to fix the timers...

#

The only issue is that I haven't decided what interface I want to use

carmine nacelle Mar 18, 2026, 8:51 PM

#

I hope I don't get AI psychosis trl
But I've talked to a clanker, and I think I have a very good solution trl

carmine nacelle Mar 18, 2026, 9:13 PM

#

This is insightful

twilit talon Mar 18, 2026, 9:17 PM

#

sycophancy final boss

#

llm are trained to do that so

carmine nacelle Mar 18, 2026, 9:18 PM

#

I need to think about what it said

#

But I think I'll add timer capabilities, which send messages as programmed (so I can just tread them as recieve rights for messages on the other side), and it's like what I have already, but better. And I was worried about the concurrency, but it suggested me that I can just IPI other cores to invalidate the timers on them (since my timers are per-CPU, but multiple threads have access to capabilities).

(It also said to look at QNX, Mach and L4, which is like reasonable...)

#

KEKW

carmine nacelle Mar 18, 2026, 11:50 PM

#

After some (passive) thinking, I think it just repeated what I said to it

#

So this doesn't count

#

Why do I even bother with asking AI

carmine nacelle Mar 19, 2026, 11:40 AM

#

I think I'm in the mood for fixing timers...

#

Though I won't redo the timers interface for now...

#

Just make it absolute, and universal

carmine nacelle Mar 19, 2026, 11:57 AM

#

Though while I'm at it, what's stopping me from making the kernel fully preemptible? trl

#

I just won't reschedule if I'm preempted...

carmine nacelle Mar 20, 2026, 12:52 AM

#

I think I'm going insane trl

#

What if I just make it use HPET for scheduling? thonk thonk thonk

manic girder Mar 20, 2026, 7:34 AM

#

I mean could be worse, iirc lemonOS uses (and only has support for) the PIT.

twilit talon Mar 20, 2026, 7:58 AM

#

Damn lemonos

#

nooo

carmine nacelle Mar 20, 2026, 9:54 PM

#

My new timer infra is having the time of its life

#

(It looks like it overflows?)

carmine nacelle Mar 20, 2026, 10:42 PM

#

bruh

carmine nacelle Mar 20, 2026, 11:06 PM

#

I think the ACPI PM time source is working now...
The only issue is that it seems that it's a bit faster than it should be in QEMU (but not by too much...)

twilit talon Mar 20, 2026, 11:06 PM

#

how did u hook it up into the kernel

carmine nacelle Mar 20, 2026, 11:07 PM

#

carmine nacelle I think the ACPI PM time source is working now... The only issue is that it seem...

Like 60 seconds passes there, while it's 56 in reality, and I don't think it can be broken code since its frequency is fixed

carmine nacelle Mar 20, 2026, 11:07 PM

#

twilit talon how did u hook it up into the kernel

It's just used as a time base, and then I use lapic in one shot mode to get to deadlines

twilit talon Mar 20, 2026, 11:07 PM

#

try on real hw

carmine nacelle Mar 20, 2026, 11:07 PM

#

I would need find the said real hw

#

And the com modem cable

twilit talon Mar 20, 2026, 11:08 PM

#

dont u have display output

carmine nacelle Mar 20, 2026, 11:08 PM

#

thonk

twilit talon Mar 20, 2026, 11:08 PM

#

like fbcon

carmine nacelle Mar 20, 2026, 11:13 PM

#

It's in a different branch trl

twilit talon Mar 20, 2026, 11:13 PM

#

bruh

#

how old is your project again KEKW

carmine nacelle Mar 20, 2026, 11:13 PM

#

since 2022

twilit talon Mar 20, 2026, 11:14 PM

#

its crazy that u have so many features yet a black screen lmao

carmine nacelle Mar 20, 2026, 11:14 PM

#

No, I can just rebase on it

#

It should probably work

#

no, the driver is in userspace

#

it's not black, it's just that if the kernel doesn't boot, there is no screen

#

I think it should shough

carmine nacelle Mar 20, 2026, 11:15 PM

#

twilit talon its crazy that u have so many features yet a black screen lmao

(I was actually planning on implementing tty soon, and bash or something, to the screen)

twilit talon Mar 20, 2026, 11:16 PM

#

yeah u should

carmine nacelle Mar 20, 2026, 11:16 PM

#

I didn't boot on my test machine trl

#

But I would imagine it just doesn't have acpi pm

twilit talon Mar 20, 2026, 11:18 PM

#

i dont think ive seen an x86 box without a pm timer

carmine nacelle Mar 20, 2026, 11:19 PM

#

It's a mac

twilit talon Mar 20, 2026, 11:19 PM

#

anyway its easy to check via acpidump

carmine nacelle Mar 20, 2026, 11:20 PM

#

Idr if I ever sent you its dump...

twilit talon Mar 20, 2026, 11:20 PM

#

i have like 4 dumps from u

#

not sure which one is which tbh

carmine nacelle Mar 20, 2026, 11:20 PM

#

I have like 10 machines

#

It's an intel mac

twilit talon Mar 20, 2026, 11:20 PM

#

lmao

#

i mean acpi dumps dont exactly tell u any hardware information

carmine nacelle Mar 20, 2026, 11:21 PM

#

My kernel is trolling me

#

I'm building the kernel-fb branch...

#

Can you get the dump from OS X?

twilit talon Mar 20, 2026, 11:22 PM

#

most likely yes

#

but idk if it can be done via acpidump

carmine nacelle Mar 20, 2026, 11:25 PM

#

if not, I can just boot ubuntu on it, or something

twilit talon Mar 20, 2026, 11:26 PM

#

i mean even windows supports dumping it

#

i think osx should have a way too

carmine nacelle Mar 20, 2026, 11:27 PM

#

tbh I should make a flantem in kernel a define instead of a separate branch

twilit talon Mar 20, 2026, 11:27 PM

#

yes

#

do kconfig

carmine nacelle Mar 20, 2026, 11:27 PM

#

Or like just have it at all time, since it useful for panics

#

Since now I just print it to com

#

trl

#

Yeah, it's a broadcast IPI dip skill issue

#

I can already see it's very accurate

#

(the timer)

#

I mean, why wouldn't it be

#

And shutdown still works

#

The keyboard had a skill issue though, for some reason

#

Though knowing it's a mac, it probably has a usb or i2c keyboard, or some weirdness like that

#

AHCI is busted though

#

Ok, too large of a todo list

#

I'll do HPET, then TSC (deadline), then the KVM thing probably

#

twilit talon Mar 20, 2026, 11:56 PM

#

carmine nacelle <:trl:829316641191886888>

yeah see it also reports garbage ioapic id

#

its a common theme

#

because back in the day ids had to be unique

#

but it doesnt actually program that id in

#

btw are those found cpu lapic logs only for APs?

carmine nacelle Mar 21, 2026, 12:10 AM

#

twilit talon btw are those found cpu lapic logs only for APs?

Yeah

#

But it prints it when it finds them in madt

twilit talon Mar 21, 2026, 12:11 AM

#

why isnt there a print for BSP

carmine nacelle Mar 21, 2026, 12:17 AM

#

IIt prints it when it initializes data for an AP in prepare_cpu function. And it doesn't get called for the BSP
https://gitlab.com/mishakov/pmos/-/blob/dev/kernel/arch/x86/cpus/cpus.cc?ref_type=heads#L286

GitLab

kernel/arch/x86/cpus/cpus.cc · dev · Mikhail Kovalev / pmOS · Gi...

GitLab.com

#

(Need to fix the typo)

#

It's all very simple trl

twilit talon Mar 21, 2026, 12:18 AM

#

Coildn't

#

☠️

#

i like how u have for_each and then a handrolled iteration loop

carmine nacelle Mar 21, 2026, 12:19 AM

#

Yeah, because you said that there was for_each after I did that loop, and I didn't bother with rewriting it

twilit talon Mar 21, 2026, 12:20 AM

#

lol

carmine nacelle Mar 21, 2026, 12:20 AM

#

I mean it works

carmine nacelle Mar 21, 2026, 8:23 AM

#

thonk do I want to bother with hpet interrupts?

carmine nacelle Mar 21, 2026, 9:42 AM

#

(No)

#

HPET is refusing to cooperate nooo

#

Which I think is a skill issue with units trl

#

Yeah, HPET in qemu also runs a little bit faster than it should

#

(though I'm using qemu in wsl2, so it's double virt, so it's kinda not out of the question=

#

Also, my keyboard and AHCI drivers now work consistently (with either of them)

#

I want to do TSC...

carmine nacelle Mar 21, 2026, 11:16 AM

#

Why is APIC timer code on oasdev wiki in assembly??

plain jungle Mar 21, 2026, 11:23 AM

#

carmine nacelle Yeah, HPET in qemu also runs a little bit faster than it should

i doubt that it does

#

how much faster?

carmine nacelle Mar 21, 2026, 11:25 AM

#

Like the same as ACPI timer

plain jungle Mar 21, 2026, 11:26 AM

#

they should all run at the proper frequency

carmine nacelle Mar 21, 2026, 11:26 AM

#

carmine nacelle Like the same as ACPI timer

My phone counts about 56-58 seconds while it counts 60

#

And I can see it's a bit faster

#

And it was seemengly very accurate on the phys hardware

manic girder Mar 21, 2026, 11:27 AM

#

carmine nacelle Why is APIC timer code on oasdev wiki in assembly??

would you rather it be in c? trl

carmine nacelle Mar 21, 2026, 11:27 AM

#

carmine nacelle And it was seemengly very accurate on the phys hardware

(I haven't tested HPET, but ACPI PM was)

plain jungle Mar 21, 2026, 11:27 AM

#

that's almost certainly an issue on your side

carmine nacelle Mar 21, 2026, 11:27 AM

#

Yeah, I think it's an issue with WSL2

plain jungle Mar 21, 2026, 11:28 AM

#

56/60 deviation is huge

carmine nacelle Mar 21, 2026, 11:28 AM

#

Because ACPI PM's frequency is fixed, and my kernel uses deadlines based on it, I don't think the code can be bad

#

I want to implement kvm clock and see if it improves

carmine nacelle Mar 21, 2026, 11:31 AM

#

carmine nacelle Because ACPI PM's frequency is fixed, and my kernel uses deadlines based on it, ...

Like I take PMTMR/HPET as a reference, and lapic oneshot deadlines, and I've set it up to print something in kernel every second (after the previous deadline), so it should have been precisely at 1 Hz (even if I don't get an interrupt on time)

#

(so even if my lapic calibration is bad, it shouldn't matter)

carmine nacelle Mar 21, 2026, 12:04 PM

#

Cool, it's tripple faulting if I use tsc deadline on wrmsr nooo

#

Wtf

#

Wtf

#

I think I'll just trust that it's working and not test it trl

#

it found a ps/2 keyboard -> timers are not fucked

carmine nacelle Mar 21, 2026, 1:58 PM

#

wait, the pvclock is like really nice actually?

#

it gives me the tsc frequency and time in nanoseconds, and just have to lapic dedline it and that's it??

#

No calibration or weirdness needed

left dew Mar 21, 2026, 1:59 PM

#

that's kind of the whole point

twilit talon Mar 21, 2026, 2:48 PM

#

What did you expect lol

carmine nacelle Mar 21, 2026, 3:03 PM

#

idk, the typical x86 jank

carmine nacelle Mar 21, 2026, 5:28 PM

#

I take my words back

#

I don't understand it

carmine nacelle Mar 21, 2026, 10:53 PM

#

Oh no, loongarch actually has really bad timer nooo

#

Or actually not?

carmine nacelle Mar 21, 2026, 11:26 PM

#

nevermind

#

loongarch is like tsc + lapic (without tsl deadline), but it's like architectural, with a fixed frequency that's given to you in cpuid

carmine nacelle Mar 22, 2026, 9:55 AM

#

Apparently the timers weren't working there because I never actually called the initialization function trl

carmine nacelle Mar 22, 2026, 11:44 AM

#

thonk coild the limine smp init code for risc-v be broken?

carmine nacelle Mar 22, 2026, 7:03 PM

#

RISC-V is trolling me nooo

plain jungle Mar 22, 2026, 7:29 PM

#

why are you switching between some many different tasks all the time lol

carmine nacelle Mar 22, 2026, 7:30 PM

#

I'm just fixing the broken stuff

#

since I switched away from using limine mp request (on x86), I had to reimplement it on risc-v

#

and I kinda wanted to have smp

#

(I mean maybe I'll implement smp loongarch as well)

#

cool, it's working

#

the only broken thing now is acpi power button interrupt, for some reason

#

(also, smp trampoline on risc-v is like super easy)

carmine nacelle Mar 22, 2026, 10:30 PM

#

I have to stop, since I have exams

#

(I'm disappearing till friday)

carmine nacelle Mar 23, 2026, 4:35 PM

#

I think I'll add clock source, calibration source and interrupt source to kernel as an abstraction on x86

#

This way I can support hpet and kvm clock elegantly...

#

(as an interrupt source, with the kvm clock handling, etc.)

carmine nacelle Mar 23, 2026, 9:16 PM

#

I think I either won't support PIT in the kernel, or make it a cmdline flag

#

Fuck me

carmine nacelle Mar 23, 2026, 10:59 PM

#

Cool, I think my HPET calibration produces more or less the same frequency as ACPI PM

carmine nacelle Mar 24, 2026, 12:58 AM

#

(anyway, my timer infra is proper now, and I don't use PIT anymore at all)

#

It's ACPI PM, HPET and CPUID 0x15 TSC

#

(or like the first two for calibration)

#

I think I can live without KVM timer for now...

#

(I'll maybe do it this weekend)

#

But I'm happy and this feels complete

carmine nacelle Mar 24, 2026, 7:00 PM

#

Huge? https://gitlab.com/mishakov/pmos/-/merge_requests/4/diffs

GitLab

Add stuff from dev (!4) · Merge requests · Mikhail Kovalev / pmOS...

Add all the changes done since the June 2025, which include: Services in the bootstrap/init server A bunch of changes related to rights...

#

carmine nacelle Mar 25, 2026, 1:41 AM

#

I need to look at how people do POSIX servers. because what I have feels weird

carmine nacelle Mar 26, 2026, 12:14 AM

#

thonk should I seek contributors?

twilit talon Mar 26, 2026, 12:18 AM

#

How would that even work

#

Where would u look

carmine nacelle Mar 26, 2026, 12:18 AM

#

idk

#

but like now all of the code is mine

#

(except the libraries and stuff like that)

#

idk how to say what I want to say

#

I can't keep doing it alone forever

light basin Mar 26, 2026, 12:33 AM

#

I would love to contribute when I get to a point of not being utterly incompetent at C/C++ pepeflower

carmine nacelle Mar 26, 2026, 12:52 AM

#

I mean I have Rust in userspace as well coolmeme

toxic torrent Mar 26, 2026, 7:01 AM

#

carmine nacelle I mean I have Rust in userspace as well <:coolmeme:808387874268905504>

oh right

#

on that note, how do you build your std?

#

my problem is that I don't know what functions i need to add to the libc crate that aren't needed for the unix group

#

actually

#

where tf IS your rust std

#

nvm

#

hm

#

did you just add random headers to the libc crate until it worked?

carmine nacelle Mar 26, 2026, 8:35 AM

#

Kinda, yeah

#

I think I've just defined my os as unix, then added headers by looking at other operating systems, then implemented everything that was needed for it to link

toxic torrent Mar 26, 2026, 8:39 AM

#

k

carmine nacelle Mar 26, 2026, 10:10 AM

#

bruh I've tried to implement smap and smep, but qemu is trolling me

#

This doesn't detect it for some reason

#

thonk

#

the kernel works, but my userspace breaks with smap 💀💀

toxic torrent Mar 26, 2026, 11:24 AM

#

carmine nacelle the kernel works, but my userspace breaks with smap 💀💀

how

#

do you fault on something?

carmine nacelle Mar 26, 2026, 11:24 AM

#

My mmap syscall fails 💀

#

(or like not the actual mmap, but mapping memory objects)

hexed knot Mar 26, 2026, 11:24 AM

#

so basically your code is broken?

carmine nacelle Mar 26, 2026, 11:24 AM

#

#

yeah

hexed knot Mar 26, 2026, 11:25 AM

#

where does it fail

carmine nacelle Mar 26, 2026, 11:25 AM

#

i've no idea

#

it just fails

hexed knot Mar 26, 2026, 11:25 AM

#

see this is why i put tracing probes all over my kernel so you can enable them to get debug prints all over the place 🙃

#

(and yes they are not terribly fast, but they are very cool)

carmine nacelle Mar 26, 2026, 11:25 AM

#

I though it was a copy from user thing, which didn't have smap, but apparently it's not?

#

anyway, I have an exam in a few hours and I'm just procrastinating

#

I though it would be a quick thing since I already had it on risc-v?

#

I think that my copy_from_user catches the pagefault incorrectly or something?

hexed knot Mar 26, 2026, 11:27 AM

#

hexed knot (and yes they are not terribly fast, but they are very cool)

and the overhead isnt even that crazy if they are inactive

carmine nacelle Mar 26, 2026, 11:32 AM

#

Yeah, I am pretty sure it's the copy_from_user stuff, because why would it fail otherwise

carmine nacelle Mar 26, 2026, 4:36 PM

#

Cool, I have smap and smep on amd64

#

it was some assembly skill issue

#

in my copy to user...

carmine nacelle Mar 26, 2026, 5:23 PM

#

wtf

carmine nacelle Mar 26, 2026, 6:26 PM

#

My i686 port is trolling me nooo

#

Not anymore trl

#

Do I fix my build system?

carmine nacelle Mar 26, 2026, 6:52 PM

#

Do I go with xbstrap or jinx thonk

left dew Mar 26, 2026, 6:53 PM

#

totally unbiased answer: jinx

carmine nacelle Mar 26, 2026, 6:56 PM

#

So I can't eventually self-host it? thonk

left dew Mar 26, 2026, 7:01 PM

#

unless you want to implement linux features, no

carmine nacelle Mar 26, 2026, 7:02 PM

#

Xbstrap also talks about containers...

toxic torrent Mar 26, 2026, 7:57 PM

#

carmine nacelle So I can't eventually self-host it? <:thonk:575445129863888896>

i have requested this before

#

I don't remember the outcome of the discussion

#

being able to self host the distro is something I'd like to do at some point

plain jungle Mar 26, 2026, 8:01 PM

#

xbstrap can run without containers although I wouldn't recommend that

carmine nacelle Mar 26, 2026, 8:01 PM

#

though I think I'm still quite far from self hosting

carmine nacelle Mar 26, 2026, 10:09 PM

#

why are all the build systems so confusing?

#

#

bruh

#

this is cool

carmine nacelle Mar 26, 2026, 10:42 PM

#

what the fuck

carmine nacelle Mar 26, 2026, 11:04 PM

#

carmine nacelle Mar 26, 2026, 11:20 PM

#

Cool, it's compiling clang

toxic torrent Mar 27, 2026, 7:37 AM

#

carmine nacelle what the fuck

hm?

carmine nacelle Mar 27, 2026, 7:47 AM

#

The image didn't send

#

The first result for make on Google is some ai slop?

toxic torrent Mar 27, 2026, 7:51 AM

#

cursed

carmine nacelle Mar 27, 2026, 10:35 AM

#

What am I supposed to do?

twilit talon Mar 27, 2026, 10:36 AM

#

sudo !!

carmine nacelle Mar 27, 2026, 10:36 AM

#

I know

#

But I'm trying to make jinx compile it

modest thistle Mar 27, 2026, 10:36 AM

#

patch it maybe?

#

or change install dir

carmine nacelle Mar 27, 2026, 10:38 AM

#

no

#

it's a host recipe

#

I did this

#!/bin/sh

name=clang
revision=1
from_source=clang-src
# hostdeps="cmake"

configure() {
    cmake -S ${source_dir}/llvm  -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_PROJECTS="clang;lld;lldb" -DLLVM_TARGETS_TO_BUILD="X86;RISCV;LoongArch" -DCMAKE_INSTALL_PREFIX=${prefix} -DLLVM_TARGET_TRIPLE=x86_64-pmos -DCLANG_DEFAULT_CXX_STDLIB=libc++ -DCLANG_DEFAULT_LINKER=lld
}

build() {
    make -j ${parallelism}
}

package() {
    make install
}

#

Cool, I think it worked

#

didn't know llvm also took DESTDIR

modest thistle Mar 27, 2026, 10:47 AM

#

carmine nacelle didn't know llvm also took DESTDIR

yeah i use it to make it install into a subfolder in it's project XD

carmine nacelle Mar 27, 2026, 10:55 AM

#

for some reason, it's not finding the directory?

carmine nacelle Mar 27, 2026, 11:17 AM

#

I'm too dumb for this

#

How do I make it find clang?

#

If it's not in the path

carmine nacelle Mar 27, 2026, 11:36 AM

#

Why

#!/bin/sh

name=libc
version=0.0.1
revision=1

source_dir=lib/libc
deps="libc-headers"
hostdeps="clang"

configure() {
    cmake -S ${source_dir} -DSYSROOT=${sysroot_dir} -DTARGET_ARCH=${JINX_ARCH} -DINSTALL_DIR="${dest_dir}/usr/lib"
}

build() {
    make -j ${parallelism}
}

package() {
    INSTALL_DIR="${dest_dir}/usr/lib" make install
}

left dew Mar 27, 2026, 11:55 AM

#

DESTDIR

#

for make install

carmine nacelle Mar 27, 2026, 11:55 AM

#

This is my CMake

cmake_minimum_required(VERSION 3.22)
set(CMAKE_C_STANDARD 17)

set(TOOLCHAIN_PREFIX "${TARGET_ARCH}-pmos")

set(CMAKE_C_COMPILER "clang")
set(CMAKE_ASM_COMPILER "clang")
set(CMAKE_AR "llvm-ar")

SET(CMAKE_C_COMPILER_TARGET ${TOOLCHAIN_PREFIX})
SET(CMAKE_ASM_COMPILER_TARGET ${TOOLCHAIN_PREFIX})

set(CMAKE_TRY_COMPILE_TARGET_TYPE "STATIC_LIBRARY")
set(CMAKE_ASM_OUTPUT_EXTENSION_REPLACE 1)
set(CMAKE_C_OUTPUT_EXTENSION_REPLACE 1)

set(CMAKE_C_FLAGS "-Wall -Wextra -O3 -g -nostdlib -fPIC --sysroot=${SYSROOT} -fno-omit-frame-pointer -mno-omit-leaf-frame-pointer -flto -ffat-lto-objects")

if (${TARGET_ARCH} STREQUAL "riscv64")
    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=rva20u64 -mabi=lp64d")
endif()

set(CMAKE_ASM_FLAGS ${CMAKE_C_FLAGS})

project(libc VERSION 0.1 LANGUAGES C ASM)

set(SRC_FOLDER "src")

file(GLOB_RECURSE GENERIC_SRC "${SRC_FOLDER}/generic/*.c" "${SRC_FOLDER}/generic/*.S")
file(GLOB_RECURSE ARCH_SRC "${SRC_FOLDER}/arch/${TARGET_ARCH}/*.c" "${SRC_FOLDER}/arch/${TARGET_ARCH}/*.S")

list(FILTER ARCH_SRC EXCLUDE REGEX ".*/crti\.S$")
list(FILTER ARCH_SRC EXCLUDE REGEX ".*/crtn\.S$")
list(FILTER ARCH_SRC EXCLUDE REGEX ".*/crt0\.S$")

set(SRC ${GENERIC_SRC} ${ARCH_SRC})

# GCC shenanigans
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67618
# https://github.com/zephyrproject-rtos/zephyr/issues/64941
# https://github.com/zephyrproject-rtos/zephyr/pull/65082
# https://gcc.gnu.org/onlinedocs/gcc/Other-Builtins.html
#
# Apply -fno-inline-malloc so GCC doesn't optimize calloc to itself, creating infinite recursion
set_source_files_properties(${SRC_FOLDER}/generic/malloc/malloc.c PROPERTIES COMPILE_FLAGS "-fno-builtin-malloc -fno-builtin-free -fno-builtin-calloc -fno-builtin-realloc")
set_source_files_properties(${SRC_FOLDER}/generic/string.c PROPERTIES COMPILE_FLAGS "-fno-builtin")

add_library(csrc OBJECT ${SRC})
add_library(c STATIC $<TARGET_OBJECTS:csrc>)
# add_library(c_static SHARED $<TARGET_OBJECTS:csrc>)

# set_target_properties(c_static PROPERTIES OUTPUT_NAME c)

# crti and crtn
add_library(crtstuff OBJECT ${SRC_FOLDER}/arch/${TARGET_ARCH}/crti.S ${SRC_FOLDER}/arch/${TARGET_ARCH}/crtn.S ${SRC_FOLDER}/arch/${TARGET_ARCH}/crt0.S)

# set(INSTALL_DIR "${CMAKE_CURRENT_SOURCE_DIR}/../../sysroot/usr/lib" CACHE PATH "Installation directory")
install(TARGETS c ARCHIVE DESTINATION ${INSTALL_DIR})
# install(TARGETS c_static DESTINATION ${INSTALL_DIR})

install(FILES $<TARGET_OBJECTS:crtstuff> DESTINATION ${INSTALL_DIR})

#

I don't get it

#

It worked the second time...

left dew Mar 27, 2026, 11:57 AM

#

i'm pretty sure cmake just handles DESTDIR for you out of the box

#

no need for all that INSTALL_DIR stuff

carmine nacelle Mar 27, 2026, 11:57 AM

#

Or not

#

#

I had to do this??

#!/bin/sh

name=libc
version=0.0.1
revision=1

source_dir=lib/libc
deps="libc-headers"
hostdeps="clang"

configure() {
    cmake -S ${source_dir} -DSYSROOT=${sysroot_dir} -DTARGET_ARCH=${JINX_ARCH} -DDESTDIR="${dest_dir}/usr/lib"
}

build() {
    make -j ${parallelism}
}

package() {
    DESTDIR="${dest_dir}" make install
}

#

Ok, I've made it work...

left dew Mar 27, 2026, 12:01 PM

#

you don't need DESTDIR when configuting

carmine nacelle Mar 27, 2026, 12:01 PM

#

yeah

carmine nacelle Mar 27, 2026, 5:37 PM

#

Jinx's trolling me again nooo

#

#!/bin/sh

name=kernel
version=0.0.1
revision=1

source_dir=kernel
deps="libc-headers libcxx"
hostdeps="clang"

configure() {
    cmake -S ${source_dir} -DTARGET_ARCH=${JINX_ARCH} -DSYSROOT=${sysroot_dir}
}

build() {
    make -j ${parallelism}
}

left dew Mar 27, 2026, 5:38 PM

#

yeah? you don't have a package step bro

#

you're trolling yourself

carmine nacelle Mar 27, 2026, 5:38 PM

#

#

If I add it it still does it

#

#!/bin/sh

name=kernel
version=0.0.1
revision=1

source_dir=kernel
deps="libc-headers libcxx"
hostdeps="clang"

configure() {
    cmake -S ${source_dir} -DTARGET_ARCH=${JINX_ARCH} -DSYSROOT=${sysroot_dir}
}

build() {
    make -j ${parallelism}
}

package() {
    DESTDIR="${dest_dir}" make install
}

left dew Mar 27, 2026, 5:39 PM

#

well i didn't know what the error was

#

clearly it has nothing to do with the step missing

#

try to rebuild libcxx

carmine nacelle Mar 27, 2026, 5:40 PM

#

I tried twice

#

And renamed it from libc++ to libcxx

left dew Mar 27, 2026, 5:40 PM

#

idk i dont have the brain capcity to even think of a possible solution

#

maybe it's the rc in the name

#

who knows

#

rindex is failing for some reason

#

it was added libcxx to the pool

carmine nacelle Mar 27, 2026, 5:41 PM

#

It doesn't have version

#

thonk

#

Wouldn't it infer it from the source?

left dew Mar 27, 2026, 5:41 PM

#

well that'd do it

#

uh yes

#

it would

#

show recipe pls

carmine nacelle Mar 27, 2026, 5:41 PM

#

#!/bin/sh

name=libcxx
revision=1
from_source=clang-src
deps="libc-headers libc"
hostdeps="clang"

triplet="${JINX_ARCH}-pmos"

configure() {
    cmake -S ${source_dir}/runtimes \
        -DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi;libunwind;compiler-rt" \
        -DCMAKE_BUILD_TYPE=Release \
        -DCMAKE_INSTALL_PREFIX=/usr/ \
        -DLLVM_TARGET_TRIPLE=loongarch64-pmos \
        -DCMAKE_CXX_COMPILER_TARGET="${triplet}" \
        -DCMAKE_C_COMPILER="clang" \
        -DCMAKE_CXX_COMPILER="clang++" \
        -DCMAKE_C_COMPILER_TARGET="${triplet}" \
        -DCMAKE_ASM_COMPILER_TARGET="${triplet}" \
        -DCOMPILER_RT_BUILD_BUILTINS=ON \
        -DCMAKE_C_FLAGS="--sysroot=${sysroot_dir} -nostartfiles" \
        -DCMAKE_CXX_FLAGS="--sysroot=${sysroot_dir} -nostdlib" \
        -DLIBCXXABI_USE_LLVM_UNWINDER=true \
        -DLIBCXX_ENABLE_SHARED=OFF \
        -DLIBUNWIND_ENABLE_SHARED=OFF \
        -DLIBCXXABI_ENABLE_SHARED=OFF \
        -DCOMPILER_RT_BUILD_BUILTINS=ON \
        -DCOMPILER_RT_BUILD_ORC=OFF \
        -DCOMPILER_RT_BUILD_SANITIZERS=OFF \
        -DCOMPILER_RT_BUILD_MEMPROF=OFF \
        -DCOMPILER_RT_BUILD_XRAY=OFF \
        -DCOMPILER_RT_BUILD_LIBFUZZER=OFF \
        -DCOMPILER_RT_BUILD_PROFILE=OFF \
        -DCOMPILER_RT_INSTALL_LIBRARY_DIR="/usr/lib" \
        -DCOMPILER_RT_BUILD_CTX_PROFILE=OFF
}

build() {
    make -j ${parallelism}
}

package() {
    DESTDIR="${dest_dir}" make install
}

left dew Mar 27, 2026, 5:41 PM

#

should be good

carmine nacelle Mar 27, 2026, 5:44 PM

#

Idk

#

I've added version and revision to libcxx and rebuilt it

#

If I remove libcxx, it starts to build it (but obviously can't because of the missing headers)

#

And like idk, there shouldn't be anything wrong

#

And like also it had built clang fine from the same source

#

yeah, I gues it really doesn't like the -

#

I've changed clang to 19.1.0_rc3 and got this

#

Ok, the kernel is compiling...

#

I guess the userspace shoudn't be too difficult

#

I just need to figure out how to package it into a disk file, and generate bootloader configs

carmine nacelle Mar 27, 2026, 6:31 PM

#

Is there something to do conditional building?

#

I think I'll make a clanker generate a script trl

#

I think that it inherits the version, but not revision, from the source recipe?

carmine nacelle Mar 28, 2026, 10:32 AM

#

Time to figure out how to compile rust

carmine nacelle Mar 28, 2026, 11:00 AM

#

But tbh it's much nicer than my make insanity so far

#

(Also, I've switched everything to CMake trl )

#

I'm too flat brained to figure out out of tree gnumake compilation

carmine nacelle Mar 28, 2026, 4:36 PM

#

bruh

carmine nacelle Mar 28, 2026, 11:20 PM

#

I'm too dumb for this nooo

#

(does anyone know how to fix it?)

#

This is interesting...
https://github.com/Andy-Python-Programmer/aero/blob/master/host-recipes/rust

GitHub

aero/host-recipes/rust at master · Andy-Python-Programmer/aero

Aero is a new modern, experimental, UNIX-like operating system following the monolithic kernel design. Supporting modern PC features such as long mode, 5-level paging, and SMP (multicore), to name ...

#

I think it still doesn't fix my issue gdtdoesnotwork

left dew Mar 29, 2026, 12:08 AM

#

it would be nice if you shared the recipe

#

but iirc you need to do this https://github.com/48cf/jinix/blob/82e89b58dfa5faa4508f9d4d7e83dae91ecaa955/source-recipes/rust#L9-L11

carmine nacelle Mar 29, 2026, 10:07 AM

#

https://gitlab.com/mishakov/pmos/-/blob/jinx/host-recipes/rust?ref_type=heads

GitLab

host-recipes/rust · jinx · Mikhail Kovalev / pmOS · GitLab

GitLab.com

#

but also this?

left dew Mar 29, 2026, 10:18 AM

#

lmao what

#

just rm -rf sources/rust-src*

#

and see if that fixes it

carmine nacelle Mar 29, 2026, 10:19 AM

#

I think it really doesn't like - in names...

#

and _

left dew Mar 29, 2026, 10:19 AM

#

it should be fine

#

example: any project using jinx

carmine nacelle Mar 29, 2026, 10:37 AM

#

left dew just rm -rf sources/rust-src*

I removed the builds directory and didn't work

#

I've changed the recipe to just rust and now it seems to be compiling?

#

Though my another complaint is that it fails easily on a flaky wifi connection

left dew Mar 29, 2026, 10:45 AM

#

well that doesn’t sound like a jinx issue?

carmine nacelle Mar 29, 2026, 3:17 PM

#

Everything else retries

#

anyway

#

running: /build_dir/host-builds/rust/build/x86_64-unknown-linux-gnu/stage0/bin/cargo build --jobs=default --manifest-path /base_dir/sources/rust/src/bootstrap/Cargo.toml -Zroot-dir=/base_dir/sources/rust -Zwarnings
warning: unused config key `build.docs` in `/build_dir/.cargo/config.toml`
error: failed to run `rustc` to learn about target-specific information

Caused by:
  process didn't exit successfully: `/build_dir/host-builds/rust/build/x86_64-unknown-linux-gnu/stage0/bin/rustc - --crate-name ___ --print=file-names -Zallow-features= -Wrust_2018_idioms -Wunused_lifetimes --target x86_64-unknown-pmos --crate-type bin --crate-type rlib --crate-type dylib --crate-type cdylib --crate-type staticlib --crate-type proc-macro --print=sysroot --print=split-debuginfo --print=crate-name --print=cfg -Wwarnings` (exit status: 1)
  --- stderr
  error: error loading target specification: could not find specification for target "x86_64-unknown-pmos"
    |
    = help: run `rustc --print target-list` for a list of built-in targets

Traceback (most recent call last):
  File "/base_dir/sources/rust/x.py", line 53, in <module>
    bootstrap.main()
    ~~~~~~~~~~~~~~^^
  File "/base_dir/sources/rust/src/bootstrap/bootstrap.py", line 1413, in main
    bootstrap(args)
    ~~~~~~~~~^^^^^^
  File "/base_dir/sources/rust/src/bootstrap/bootstrap.py", line 1363, in bootstrap
    build.build_bootstrap()
    ~~~~~~~~~~~~~~~~~~~~~^^
  File "/base_dir/sources/rust/src/bootstrap/bootstrap.py", line 1030, in build_bootstrap
    run(args, env=env, verbose=self.verbose, cwd=self.rust_root)
    ~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/base_dir/sources/rust/src/bootstrap/bootstrap.py", line 237, in run
    raise RuntimeError(err)
RuntimeError: failed to run: /build_dir/host-builds/rust/build/x86_64-unknown-linux-gnu/stage0/bin/cargo build --jobs=default --manifest-path /base_dir/sources/rust/src/bootstrap/Cargo.toml -Zroot-dir=/base_dir/sources/rust -Zwarnings

#

gdtdoesnotwork

carmine nacelle Mar 29, 2026, 5:13 PM

#

No, this is weird

#

there is no sources??

#

wait

#

It's weird

carmine nacelle Mar 29, 2026, 5:19 PM

#

left dew just rm -rf sources/rust-src*

I've deleted rust and now it's this??

#

#

#!/bin/sh

name=rust
revision=1
from_source=rust
allow_network=yes

configure() {
    cat > config.toml <<EOL
[llvm]
download-ci-llvm = false
targets = "X86"

[build]
target = ["x86_64-unknown-pmos", "x86_64-unknown-linux-gnu"]
build-dir = "${base_dir}/host-builds/rust"
docs = false

[install]
prefix = "${prefix}"
sysconfdir = "${prefix}/etc"
EOL
}

build() {
    CARGO_HOME="${build_dir}/.cargo" ${source_dir}/x.py build --stage 2 -j${parallelism} --verbose 
}

package() {
    CARGO_HOME="${build_dir}/.cargo" DESTDIR="${dest_dir}" ${source_dir}/x.py install -j${parallelism}

    
    find ${dest_dir} -name "*.old" -delete

    strip_command=strip
    post_package_strip
}

#!/bin/sh

name=rust
version=1.0.0
revision=1
git_url=https://gitlab.com/mishakov/rust-pmos.git
commit="39dedc845fe6715741cf9a7d425cd389f9e26f0f"

early_prepare() {
    git submodule update --init --recursive --depth=1
}

#

ok, nevermind

carmine nacelle Mar 29, 2026, 6:14 PM

#

I think it's ignoring my bootstrap.toml?

#

I think I've figured it out...

toxic torrent Mar 29, 2026, 7:38 PM

#

carmine nacelle ``` #!/bin/sh name=rust revision=1 from_source=rust allow_network=yes configur...

oh wait you don't need to specify build.host?

#

TIL

carmine nacelle Mar 29, 2026, 7:46 PM

#

I think it compiled?

#

what's post package strip anyway?

toxic torrent Mar 29, 2026, 7:47 PM

#

carmine nacelle what's post package strip anyway?

what do you think it is KEKW

carmine nacelle Mar 29, 2026, 7:47 PM

#

no, i mean where is it coming from?

toxic torrent Mar 29, 2026, 7:47 PM

#

Jinxfile

#

did you yoink this recipe from somewhere

carmine nacelle Mar 29, 2026, 7:48 PM

#

maybe trl

#

I mean I yoinked it from 5 places

toxic torrent Mar 29, 2026, 7:49 PM

#

well you have this thing called a Jinxfile

carmine nacelle Mar 29, 2026, 7:49 PM

#

i know

toxic torrent Mar 29, 2026, 7:49 PM

#

post_package_strip() {
    if [ -z "$strip_command" ]; then
        strip_command="${OS_TRIPLET}-strip"
    fi

    for f in $(find "${dest_dir}"); do
        if file "$f" | grep 'not stripped' >/dev/null; then
            echo "* stripping '$f'..."
            stripped_file="$(mktemp)"
            ${strip_command} "$f" -o "$stripped_file"
            chmod --reference="$f" "$stripped_file"
            mv -f "$stripped_file" "$f"
        fi
    done
}

carmine nacelle Mar 29, 2026, 7:49 PM

#

it's just a bunch of shell scripts

toxic torrent Mar 29, 2026, 7:49 PM

#

this is what i have in there

carmine nacelle Mar 29, 2026, 10:46 PM

#

I'm too dumb for this

#

it didn't build cargo??

#

toxic torrent Mar 30, 2026, 10:19 AM

#

carmine nacelle it didn't build cargo??

wdym

#

isn't cargo a separate repo

carmine nacelle Mar 30, 2026, 10:41 AM

#

Idk

#

Clanker said it isn't

#

https://github.com/rust-lang/cargo#compiling-from-source

GitHub

GitHub - rust-lang/cargo: The Rust package manager

The Rust package manager. Contribute to rust-lang/cargo development by creating an account on GitHub.

#

But I don't get it

#

The recipes that I've seen just run cargo install on cargo

#

But like where is it coming from?

toxic torrent Mar 30, 2026, 10:52 AM

#

carmine nacelle The recipes that I've seen just run cargo install on cargo

steal my recipes

#

i have cargo and rustc

carmine nacelle Mar 30, 2026, 10:53 AM

#

yeah, I've seen it

#

But you just call cargo install in here https://github.com/zinnia-os/bootstrap/blob/main/host-recipes/cargo

#

Imagedeps???

#

bruh

#

Also, is there a way to get the host tripple?

#

I wanted to run the build server on an arm board

modest thistle Mar 30, 2026, 12:00 PM

#

carmine nacelle Also, is there a way to get the host tripple?

doesn't it get it automatically?

carmine nacelle Mar 30, 2026, 1:32 PM

#

idk

#

maybe?

carmine nacelle Mar 30, 2026, 8:55 PM

#

I don't get it

#

I yoinked the cargo recipe, and it should be working fine, but it still can't find cargo

toxic torrent Mar 30, 2026, 10:13 PM

#

carmine nacelle I yoinked the cargo recipe, and it should be working fine, but it still can't fi...

how

#

what's the error

toxic torrent Mar 30, 2026, 10:13 PM

#

carmine nacelle Also, is there a way to get the host tripple?

uname -m meme

carmine nacelle Mar 30, 2026, 10:13 PM

#

toxic torrent Mar 30, 2026, 10:14 PM

#

did you add the cargo to hostdeps

carmine nacelle Mar 30, 2026, 10:14 PM

#

yes

toxic torrent Mar 30, 2026, 10:14 PM

#

show

carmine nacelle Mar 30, 2026, 10:14 PM

#

it's just ignoring it

#

https://gitlab.com/mishakov/pmos/-/blob/jinx/recipes/pmbus?ref_type=heads

toxic torrent Mar 30, 2026, 10:15 PM

#

my cargo recipe is different

carmine nacelle Mar 30, 2026, 10:16 PM

#

find install thing doesn't fail, so it should be fine

toxic torrent Mar 30, 2026, 10:16 PM

#

cargo install?

carmine nacelle Mar 30, 2026, 10:17 PM

#

no

#

I can't find where I stole it from trl

#

https://github.com/48cf/jinix/blob/trunk/host-recipes/cargo-cbuild

toxic torrent Mar 30, 2026, 10:18 PM

#

#! /bin/sh

name=cargo
from_source=cargo
revision=1
imagedeps="build-essential libssl-dev pkg-config cargo"
hostrundeps="gcc rust"
allow_network=yes

build() {
        CARGO_HOME="${build_dir}/.cargo" \
    cargo install \
        --target x86_64-unknown-linux-gnu \
        --target-dir "${build_dir}" \
        --path "${source_dir}" \
        -j${parallelism} \
        --root "${build_dir}/out"
}

package() {
        CARGO_HOME="${build_dir}/.cargo" \
    cargo install \
        --target x86_64-unknown-linux-gnu \
        --target-dir "${build_dir}" \
        --path "${source_dir}" \
        -j${parallelism} \
        --root "${dest_dir}${prefix}"

    # Strip binaries
    strip_command=strip \
    post_package_strip
}

#

this is mine

carmine nacelle Mar 30, 2026, 10:20 PM

#

I'm rebuilding it

carmine nacelle Mar 30, 2026, 10:22 PM

#

toxic torrent ``` #! /bin/sh name=cargo from_source=cargo revision=1 imagedeps="build-essenti...

Still cargo not found

#

this has to be something dumb

#

If I add ls /usr/local, it's empty

#

Even though it does install it

carmine nacelle Mar 31, 2026, 7:14 PM

#

Yeah, it's ignoring hostdeps for some reason...

carmine nacelle Mar 31, 2026, 10:16 PM

#

It works if I comment source_dir out?????

carmine nacelle Apr 1, 2026, 8:58 AM

#

Yeah, it breaks as soon as I add it...

carmine nacelle Apr 1, 2026, 9:18 AM

#

Just if I remove .jinx-cache and sources and build directory, and run ../jinx build pmbus with this recipe, it just gets the debian image, and doesn't even try to compile the compilers and anything

#!/bin/sh

name=pmbus
version=1.0.0
revision=1

source_dir=userspace/ps2d
deps="libc pmos-rust-lib"
hostdeps="cargo rust"
allow_network=yes

prepare() {
    CARGO_HOME="${build_dir}/.cargo" \
    cargo fetch --target ${JINX_ARCH}-unknown-pmos
}

build() {
    CARGO_HOME="${build_dir}/.cargo" \
    cargo build --target ${JINX_ARCH}-unknown-pmos --jobs ${parallelism}
}

package() {
    CARGO_HOME="${build_dir}/.cargo" \
    DESTDIR="${dest_dir}" \
    RUSTFLAGS="-C link-arg=--sysroot=${SYSROOT} -C linker=clang++ -C link-arg=--target=${JINX_ARCH}-unknown-pmos -C link-arg=-lunwind -C link-arg=-static" \
    cargo install --target ${JINX_ARCH}-unknown-pmos --root ${dest_dir} --path ${source_dir}
}

#

#

https://gitlab.com/mishakov/pmos/-/tree/jinx?ref_type=heads

GitLab

Files · jinx · Mikhail Kovalev / pmOS · GitLab

GitLab.com

left dew Apr 1, 2026, 1:42 PM

#

because the jinx cache isnt where the built tools are

#

make sure the cargo host tool is built correctly

#

i think the directory is called host-pkgs

#

there should be cargo in there

#

and it should be installed under $prefix, so /ust/local for tools

#

so the final path for the executable should be host-pkgs/cargo/usr/local/bin/cargo

carmine nacelle Apr 1, 2026, 1:45 PM

#

left dew make sure the cargo host tool is built correctly

But it's not even trying to build it

left dew Apr 1, 2026, 1:46 PM

#

ls host-pkgs/cargo

#

it probably built but is incomplete somehow

#

maybe your recipe is messed up

carmine nacelle Apr 1, 2026, 1:47 PM

#

#

carmine nacelle Apr 1, 2026, 1:48 PM

#

left dew maybe your recipe is messed up

but I don't see anything wrong with it

left dew Apr 1, 2026, 1:50 PM

#

carmine nacelle

what the fuck

#

where is ANYTHING

carmine nacelle Apr 1, 2026, 1:50 PM

#

if I comment source_dir out then it starts doing some stuff

left dew Apr 1, 2026, 1:50 PM

#

where is binutils even

carmine nacelle Apr 1, 2026, 1:50 PM

#

carmine nacelle if I comment source_dir out then it starts doing some stuff

and source dir isn't even important here

#

like what it's set to

#

and shell scripts are too complicated for me idk

left dew Apr 1, 2026, 1:52 PM

#

how is shell complicated 💀

#

anyway let me clone and try building

#

or just take a look but on pc this time

#

which jinx version are you using?

carmine nacelle Apr 1, 2026, 1:55 PM

#

the latest from main

#

tbh I need to make a makefile for it

left dew Apr 1, 2026, 1:55 PM

#

i see you didn't check in the jinx script and you don't have a makefile target to download it

#

yeah

#

that'd be nice

carmine nacelle Apr 1, 2026, 1:55 PM

#

0.8 I think

left dew Apr 1, 2026, 1:57 PM

#

ok time to build this thing

carmine nacelle Apr 1, 2026, 1:57 PM

#

do you have a fast PC?

#

it gets to fail quickly on my 5900X

left dew Apr 1, 2026, 1:58 PM

#

yeah

left dew Apr 1, 2026, 2:23 PM

#

ok it took me way too long to find out

#

but

#

you are running cargo in prepare

#

that's part of the source recipe, not the recipe itself

#

so you need source_hostdeps="cargo rust"

#

after changing source_dir to pmbus:```
[iretq@raptor build]$ ../jinx build pmbus

building package: pmbus
error: failed to get pmos as a dependency of package pmbus v0.1.0 (/base_dir/userspace/pmbus)

Caused by:
failed to load source for dependency pmos

Caused by:
Unable to update /base_dir/userspace/rust/pmos

Caused by:
failed to read /base_dir/userspace/rust/pmos/Cargo.toml

Caused by:
No such file or directory (os error 2)```

carmine nacelle Apr 1, 2026, 2:25 PM

#

bruh

#

Thanks a lot

toxic torrent Apr 1, 2026, 2:56 PM

#

galaxybrain

carmine nacelle Apr 1, 2026, 7:17 PM

#

More trolling gdtdoesnotwork

carmine nacelle Apr 1, 2026, 11:57 PM

#

(I'll probably figure it out tomorrow)

carmine nacelle Apr 2, 2026, 10:18 PM

#

gdtdoesnotwork

#

Maybe I should just retire Rust trl

carmine nacelle Apr 3, 2026, 10:50 AM

#

https://rust-lang.github.io/rustup/cross-compilation.html ?

Cross-compilation - The rustup book

carmine nacelle Apr 3, 2026, 2:26 PM

#

I think it's finally building it

toxic torrent Apr 3, 2026, 2:26 PM

#

carmine nacelle Maybe I should just retire Rust <:trl:829316641191886888>

real

carmine nacelle Apr 3, 2026, 2:26 PM

#

I've no idea what I'm doing

toxic torrent Apr 3, 2026, 2:27 PM

#

carmine nacelle I think it's finally building it

why is it unsupported

#

https://gitlab.com/mishakov/rust-pmos/-/blob/master/compiler/rustc_target/src/spec/base/pmos.rs?ref_type=heads

GitLab

compiler/rustc_target/src/spec/base/pmos.rs · master · Mikhail Ko...

GitLab.com

#

you're missing a dynamic_linking: true

carmine nacelle Apr 3, 2026, 2:29 PM

#

I don't have dynamic linker

#

I'm not using mlibc and my libc is too dumb

toxic torrent Apr 3, 2026, 2:30 PM

#

bruh

#

💀

carmine nacelle Apr 3, 2026, 2:41 PM

#

Bruh, I think it's working

#

just need to figure out cargo install

toxic torrent Apr 3, 2026, 3:12 PM

#

holy

carmine nacelle Apr 3, 2026, 3:14 PM

#

Time to recompile rustc for the millionth time...

#

Also, I need a script to assemble all of this into an image

carmine nacelle Apr 3, 2026, 3:26 PM

#

carmine nacelle Time to recompile rustc for the millionth time...

The only saving thing is my 5900X

#

It takes forever on my laptop...

twilit talon Apr 3, 2026, 4:30 PM

#

Thats why ill never do rust shit

#

This toolchain situation is miserable

toxic torrent Apr 3, 2026, 4:30 PM

#

twilit talon This toolchain situation is miserable

yeah this tbh

#

i like rust as a language

#

but jfc tooling is horrible

carmine nacelle Apr 3, 2026, 4:31 PM

#

Nah, I think I've gotten it to work

#

It's just because nobody's using rust with Jinx, and it's not very well documented?

carmine nacelle Apr 3, 2026, 4:32 PM

#

toxic torrent but jfc tooling is horrible

btw, I think your rust recipe might be broken¿

toxic torrent Apr 3, 2026, 4:32 PM

#

why

carmine nacelle Apr 3, 2026, 4:33 PM

#

config.toml didn't work for me

#

idk, I've made mine use bootstrap.toml

carmine nacelle Apr 3, 2026, 6:14 PM

#

bruh

carmine nacelle Apr 3, 2026, 6:32 PM

#

Jinx is trolling again (totally not because shit build scripts done by me)

carmine nacelle Apr 3, 2026, 7:21 PM

#

What am I looking at?

#

Why does everything build limine as a normal recipe and not host recipe

#

bruh

#

gdtdoesnotwork

#

Time to downgrade limine trl

twilit talon Apr 3, 2026, 7:40 PM

#

Why do u need such an old revision

carmine nacelle Apr 3, 2026, 7:41 PM

#

Or upgrade it

#

I yoinked the 10.6.0 recipe

#

Cool, I've managed to build the system image...

#

The only issue is that userspace doesn't work for some reason...

carmine nacelle Apr 3, 2026, 8:14 PM

#

Cool, it fully works

toxic torrent Apr 3, 2026, 8:31 PM

#

carmine nacelle Why does everything build limine as a normal recipe and not host recipe

wdym

carmine nacelle Apr 3, 2026, 8:32 PM

#

toxic torrent wdym

https://github.com/zinnia-os/bootstrap/blob/main/recipes/limine

GitHub

bootstrap/recipes/limine at main · zinnia-os/bootstrap

A Distribution for the Zinnia kernel. Contribute to zinnia-os/bootstrap development by creating an account on GitHub.

#

everyone, not everything

toxic torrent Apr 3, 2026, 8:32 PM

#

only i build it as a normal recipe

carmine nacelle Apr 3, 2026, 8:32 PM

#

I can't speak today

carmine nacelle Apr 3, 2026, 8:32 PM

#

toxic torrent only i build it as a normal recipe

jinix does the same

toxic torrent Apr 3, 2026, 8:32 PM

#

astral for example doesn't

#

i do it because i have no need for limine's host tool

#

and it's more like a normal package for my os

carmine nacelle Apr 3, 2026, 9:11 PM

#

This is very nice of flanterm...

plain jungle Apr 3, 2026, 9:52 PM

#

toxic torrent i do it because i have no need for limine's host tool

how do you install limine w/o the host tool?

toxic torrent Apr 3, 2026, 9:52 PM

#

cp

#

i don't support bios boot anyways

#

i just copy the BOOTX64.EFI to the esp

#

that's it

mellow sleet Apr 4, 2026, 2:13 AM

#

toxic torrent astral for example doesn't

I build it as both host and recipe

carmine nacelle Apr 4, 2026, 9:13 AM

#

bruh I need to figure out risc-v and other arches now...

#

Since zinnia does this? https://github.com/zinnia-os/bootstrap/blob/main/GNUmakefile

GitHub

bootstrap/GNUmakefile at main · zinnia-os/bootstrap

A Distribution for the Zinnia kernel. Contribute to zinnia-os/bootstrap development by creating an account on GitHub.

#

But I haven't seen $ARCH be used anywhere?

#

Huh?

#

Is it normal that there are 2 variables?

carmine nacelle Apr 4, 2026, 11:15 AM

#

#

bruh Ubuntu's repos strike again

carmine nacelle Apr 4, 2026, 12:16 PM

#

This is fun

#

cool, the latest SBI doesn't even boot

vale hinge Apr 4, 2026, 12:29 PM

#

carmine nacelle

use a newer qemu, that's the only solution sorry

carmine nacelle Apr 4, 2026, 12:33 PM

#

vale hinge use a newer qemu, that's the only solution sorry

it is the latest version

vale hinge Apr 4, 2026, 12:51 PM

#

really?

twilit talon Apr 4, 2026, 12:58 PM

#

carmine nacelle This is fun

I like how it says ANIC

toxic torrent Apr 4, 2026, 1:34 PM

#

twilit talon I like how it says ANIC

wild P

carmine nacelle Apr 4, 2026, 4:30 PM

#

Time to set up the build server...

carmine nacelle Apr 4, 2026, 5:14 PM

#

lol, who even uses docker (coudflare) on weekends...

carmine nacelle Apr 4, 2026, 5:40 PM

#

(it's blocked by my isp again)

modest thistle Apr 4, 2026, 5:57 PM

#

carmine nacelle (it's blocked by my isp again)

vpn time?

carmine nacelle Apr 4, 2026, 6:01 PM

#

I guess I'll just wait

carmine nacelle Apr 4, 2026, 9:03 PM

#

I think I'll merge it to main today

#

Btw, I think the project structure here is outdated https://codeberg.org/Mintsuki/jinx/src/branch/trunk/DOCUMENTATION.md#basic-project-structure

#

(I can send patches)

carmine nacelle Apr 4, 2026, 9:16 PM

#

carmine nacelle I think I'll merge it to main today

I've merged it

#

You can now actually build everything with make qemu-x86

#

(if someone wants to try it for some reason)

carmine nacelle Apr 5, 2026, 12:20 PM

#

Bruh, I want to do disk...

carmine nacelle Apr 6, 2026, 8:29 AM

#

Aparently, this is a skill issue somewhere inside Rust?#1385970208631427173 message

carmine nacelle Apr 6, 2026, 10:24 AM

#

I think it's a printf skill issue trl

carmine nacelle Apr 6, 2026, 5:25 PM

#

I think I should finally add rights to memory objects?
But I don't really know how to handle them (like do I do weak/strong references??)

#

GNU/Hurs is very well documented, as usual https://www.gnu.org/software/hurd/microkernel/mach/memory_object.html

carmine nacelle Apr 7, 2026, 12:06 PM

#

carmine nacelle I think I should finally add rights to memory objects? But I don't really know ...

Surprisingly, it was very easy...

carmine nacelle Apr 7, 2026, 12:23 PM

#

This is nice...

#

I guess it's time to figure out Rust's coroutines nooo
And integrate https://discord.com/channels/440442961147199490/1479184639145938974 ultrameme

carmine nacelle Apr 7, 2026, 1:03 PM

#

Also, I think I figured out how I should do multithreaded servers?

#

I'll just make programs (e.g. libc) explicitly request rights that correspond to the right cores

#

The kernel doesn't have to do anything

carmine nacelle Apr 9, 2026, 9:03 AM

#

carmine nacelle I guess it's time to figure out Rust's coroutines <:nooo:750466870993223750> An...

The Rust's futures are weird af

#

I think I'd need my own runtime for my IPC?

plain jungle Apr 9, 2026, 2:03 PM

#

yes

#

a runtime is really small in rust though

carmine nacelle Apr 9, 2026, 8:39 PM

#

I wonder how fucked my userspace would be if I introduce blocking send() thonk trl

#

Which I think I kinda have to do for Rust...

plain jungle Apr 9, 2026, 8:46 PM

#

Hm?

#

Which I think I kinda have to do for Rust...
^ this sounds weird

#

Rust doesn't force you into a particular IPC model?

carmine nacelle Apr 9, 2026, 8:48 PM

#

My kernel only lets me get the message that's at the front of the queue, and I think I'll have to have a future per right, so I'll have to get messages eagerly and push those the the future's queue (so there aren't deadlocks), so someone can easily DOS my rust servers by sending a lot of messages and making it OOM

#

So the solution is limiting how many messages can be sent

#

Idk, I still don't quite get what I have to do

plain jungle Apr 9, 2026, 8:52 PM

#

how do you avoid this in C?

#

there should be no need to have a separate queue that buffers stuff

#

Rust's future design certainly doesn't require it

carmine nacelle Apr 9, 2026, 8:59 PM

#

plain jungle how do you avoid this in C?

I don't

#

I mean maybe it doesn't matter

#

It's only an issue if you don't immediately await on the future

#

I mean unbounded queues are problematic in general?

plain jungle Apr 9, 2026, 9:01 PM

#

Rust doesn't require you to add an unbounded queue

carmine nacelle Apr 9, 2026, 9:02 PM

#

they are unbounded in the kernel

plain jungle Apr 9, 2026, 9:03 PM

#

that sounds like a problem indeed

#

but anyway, I don't see how Rust futures make it more (or less) problematic

carmine nacelle Apr 9, 2026, 9:08 PM

#

Yeah, but I might as well fix it

carmine nacelle Apr 10, 2026, 11:31 AM

#

plain jungle how do you avoid this in C?

I mean C uses callbacks

plain jungle Apr 10, 2026, 11:31 AM

#

why does that matter?

#

you can convert callback based code to Rust futures or vice versa

carmine nacelle Apr 10, 2026, 11:41 AM

#

carmine nacelle I mean C uses callbacks

Then it's kinda your fault (in my eyes) if you don't register the callbacks in the right order

plain jungle Apr 10, 2026, 11:47 AM

#

i don't get it. you can have a rust future that registers a callback and completes the future on completion

#

you just have some kind of control block that stores the future's current waker

#

or similar

carmine nacelle Apr 10, 2026, 11:50 AM

#

I want send to make new futures

#

It's not a problem with the callbacks/futures, but with the order of doing things being important if I carry on with unbound queues

#

in kernel

#

Since I'd have to buffer everything in userspace, and it's bad

#

(infinitely)

#

So I think I'll make it bounded

#

This would also solve the problem of not needing to allocate on every send

#

Like if I make it bounded, then the buffers can be preallocated, and it can't fail on OOM or weirdness like that, which is kinda important for drivers

#

(Since I can't just block them on oom)

plain jungle Apr 10, 2026, 11:52 AM

#

carmine nacelle I want send to make new futures

you can just do that

#

there is no need to buffer anything in user space

#

https://github.com/managarm/managarm/blob/master/rust/hel/src/submission.rs#L68

GitHub

managarm/rust/hel/src/submission.rs at master · managarm/managarm

Pragmatic microkernel-based OS with fully asynchronous I/O - managarm/managarm

carmine nacelle Apr 10, 2026, 11:53 AM

#

plain jungle https://github.com/managarm/managarm/blob/master/rust/hel/src/submission.rs#L68

Yeah, I wanted to do that

#

But still, it's problematic

#

Like you create 2 futures, then if await on only one of them, the second one can deadlock everything

plain jungle Apr 10, 2026, 11:54 AM

#

why?

carmine nacelle Apr 10, 2026, 11:54 AM

#

Also, I think my IPC is a bit different from Managarm's?

carmine nacelle Apr 10, 2026, 11:55 AM

#

carmine nacelle Like you create 2 futures, then if await on only one of them, the second one can...

Because you can only pop_front from my IPC

plain jungle Apr 10, 2026, 11:55 AM

#

you want the future to do the actual work at await time, not at future creation time

carmine nacelle Apr 10, 2026, 11:55 AM

#

Idk if I'm not explaining it well

#

But I have (recieve) ports, which then have many recieve rights (bound to a port)

#

You can only get messages from the front of the port

#

I want to make futures wait on rights

#

So if there is a message for the given right, and I can't move it to the future (it's not being polled), I have to block

#

So it creates a deadlock

#

Or I'll have to throw messages away

#

Or I have to buffer them indefinitely in userspace

#

Or you have to wait on all of the futures immediately

plain jungle Apr 10, 2026, 11:59 AM

#

plain jungle you want the future to do the actual work at await time, not at future creation ...

^

carmine nacelle Apr 10, 2026, 12:00 PM

#

plain jungle you want the future to do the actual work at await time, not at future creation ...

No, it does the work at await time

#

The Stream/Waker would be polling the message queue

#

I think managarm just lets you wait on any stream, no?

plain jungle Apr 10, 2026, 12:05 PM

#

wdym by "on any stream"? Managarm allows you to submit multiple IPC ops concurrently

carmine nacelle Apr 10, 2026, 12:05 PM

#

My kernel doesn't

plain jungle Apr 10, 2026, 12:06 PM

#

and wait for their completion

#

But messages on each stream are strictly ordered

carmine nacelle Apr 10, 2026, 12:08 PM

#

Yes, but can you wait on several ops from one thread at the same time?

plain jungle Apr 10, 2026, 12:08 PM

#

yes. but if you can't do that, why are you trying to use an async runtime anyway

carmine nacelle Apr 10, 2026, 12:09 PM

#

I don't know how to explain it

#

It works if you buffer it

#

This is my IPC https://gitlab.com/mishakov/pmos/-/blob/main/kernel/generic/messaging/messaging.hh?ref_type=heads#L157
https://gitlab.com/mishakov/pmos/-/blob/main/libs/pmoscxx/src/helpers.cc?ref_type=heads#L128

#

My AHCI driver is fully async in C++

left dew Apr 10, 2026, 12:18 PM

#

why does it matter which operation completes in what order?

#

i don't quite understand what your problem with rust's async is

#

ah i think i see your problem

carmine nacelle Apr 10, 2026, 12:20 PM

#

Let's say you have rights A and B, both can recieve multiple messages
You receive a a b in this order (which is buffered in kernel)
You would get a, move that to the future, then you would get a again. You either have to throw it away, or buffer it, or block on it. If you block on a, but you're awaiting on B before a, you can't make progress, since you can't recieve b because a is at the front

#

I mean I had an argument with a clanker and it really insisted that I don't do that

#

If you buffer a a in stream, then you can recieve b and there is no issue. But since my queues are unbounded, and there is no way to stop new messages, someone can just send a lot of messages, and it might DoS my server

left dew Apr 10, 2026, 12:22 PM

#

i would have to look at your ipc model more closely but it really doesn't sound too bad, do you have a kernel api that lets you receive something? you're saying "recv A, recv A, recv B" - what syscalls are involved in this process? what are A and B in this case?

#

actually lemme take a look at the ipc link you posted

carmine nacelle Apr 10, 2026, 12:23 PM

#

carmine nacelle This is my IPC https://gitlab.com/mishakov/pmos/-/blob/main/kernel/generic/messa...

The second link is userspace C++ wrapper over syscalls

#

I think my async C++ thing is also somewhere in there

#

Like I can just ignore it and keep the order in mind and it will be fine, but it won't be "fearless concurrency"

left dew Apr 10, 2026, 12:25 PM

#

sounds to me like you need a select-like syscall

carmine nacelle Apr 10, 2026, 12:25 PM

#

yes

#

or that

left dew Apr 10, 2026, 12:26 PM

#

or recv-many or whatever

carmine nacelle Apr 10, 2026, 12:26 PM

#

But my queues are just a list/deque of messages

left dew Apr 10, 2026, 12:26 PM

#

same idea i guess

carmine nacelle Apr 10, 2026, 12:26 PM

#

So I would need to redo it

#

Or, if you make it bounded, then you can preallocate the buffers in userspace, and just do it in there

carmine nacelle Apr 10, 2026, 12:28 PM

#

carmine nacelle I think my async C++ thing is also somewhere in there

It's this https://gitlab.com/mishakov/pmos/-/blob/main/libs/pmoscxx/src/helpers.cc?ref_type=heads#L96

plain jungle Apr 10, 2026, 12:29 PM

#

carmine nacelle Let's say you have rights A and B, both can recieve multiple messages You receiv...

But with any reasonable executor, all futures can make progress at the same time

#

You get a, process it, you create another future that gets the next a, you process it, you get b

left dew Apr 10, 2026, 12:30 PM

#

that's the "i can just keep the order right myself" thing

#

seems like that's also how it works in c++

carmine nacelle Apr 10, 2026, 12:31 PM

#

It's not really a future, but a Stream in Rust

carmine nacelle Apr 10, 2026, 12:31 PM

#

left dew seems like that's also how it works in c++

Yeah, it's also problematic, but I've only noticed it now

plain jungle Apr 10, 2026, 12:31 PM

#

I think your issue stems from confusion about push and pull

carmine nacelle Apr 10, 2026, 12:32 PM

#

It would be both I guess?

plain jungle Apr 10, 2026, 12:32 PM

#

Rust futures/streams pull things from an object

#

It seems that you want to make it push based

carmine nacelle Apr 10, 2026, 12:33 PM

#

I guess?

plain jungle Apr 10, 2026, 12:33 PM

#

which doesn't work

#

So the fix is to not make it push based

carmine nacelle Apr 10, 2026, 12:33 PM

#

But my kernel doesn't have the API to make it pull based

#

And I would need to move the message queues to the rights

#

If I want to be able to pull on rights instead of ports

carmine nacelle Apr 10, 2026, 12:34 PM

#

plain jungle So the fix is to not make it push based

But like also why not do this

#

I wanted to add a field that says that it's not the last message in the queue

#

So you can push several messages into the futures

#

Also, it doesn't matter, it's an abstraction over IPC

#

If something builds on top of it, it would be pull based

plain jungle Apr 10, 2026, 12:36 PM

#

Ah i think i get your problem now

#

You want to distribute messages from a common receiver to multiple consumers but you don't know the consumers in advance

carmine nacelle Apr 10, 2026, 12:37 PM

#

yes

plain jungle Apr 10, 2026, 12:37 PM

#

Yeah that sounds like a design issue indeed

carmine nacelle Apr 10, 2026, 12:37 PM

#

Or if I know them in advance, then I have to buffer

plain jungle Apr 10, 2026, 12:37 PM

#

But one that you'd also have with callbacks

carmine nacelle Apr 10, 2026, 12:38 PM

#

yeah

#

that's why I said C++ is also problematic

#

But with C/C++ you can just say "it's on you to do the things in the right order"

plain jungle Apr 10, 2026, 12:39 PM

#

Well one thing that you can do is have a stream that returns (message, target) pairs

#

And have higher level logic deal with this issue

carmine nacelle Apr 10, 2026, 12:40 PM

#

I don't want to do that

#

Then why would you want async in the first place

#

If you can only wait on one port

carmine nacelle Apr 10, 2026, 12:41 PM

#

plain jungle And have higher level logic deal with this issue

This is just moving the issue to that higher level logic

plain jungle Apr 10, 2026, 12:41 PM

#

That's true

carmine nacelle Apr 10, 2026, 12:41 PM

#

I mean that higher level logic is what I'm trying to write

plain jungle Apr 10, 2026, 12:42 PM

#

But higher level logic may be better equipped to solve it (maybe. Depends on your design)

carmine nacelle Apr 10, 2026, 12:42 PM

#

But unbounded queues are also problematic on other abstractions/stuff that builds on top of it, e.g. Unix pipes

#

Which have to block if they're full

#

Or not...

#

thonk

#

No, with pipes, you'll just wait for the reply

#

Idr, there were a bunch of issues with having no way to set a boundary on how many messages can be in the queue

#

But with my ports/rights thing, it should be somewhat straightforward to be able to add limits on rights

#

(with the biggest issue being the scheduler/blocking, which needs improving)

#

Which I also need for futex

carmine nacelle Apr 10, 2026, 5:04 PM

#

thonk What if I do VFS in Rust?

#

which would be a test for my async thing...

carmine nacelle Apr 10, 2026, 7:10 PM

#

Though I think my issue is that I don't quite understand VFS

#

Namely how the directory tree should be, what to do with mountpoints and hardlinks...

carmine nacelle Apr 10, 2026, 8:16 PM

#

thonk

carmine nacelle Apr 11, 2026, 8:36 PM

#

Small progress update -- I ended up doing it in Rust

#

And I'm currently fighting the borrow checker and stuff like that

carmine nacelle Apr 11, 2026, 9:08 PM

#

This insanity works??
(create_named_stream does a request to bootstrapd)

#

(also, I think mut isn't needed here...)

carmine nacelle Apr 12, 2026, 4:30 PM

#

Anyway, since I can't be bothered with writing a proper posix_spawn, I think I'll be launching filesystem drivers through bootstrapd

#

Namely, I think I'll first be launching drivers to probe the partitions, and then doing it again on mount

#

I need to look into the other operating systems thonk

carmine nacelle Apr 12, 2026, 5:36 PM

#

The clanker just repeated what I said to it

plain jungle Apr 12, 2026, 5:43 PM

#

whenever somebody says that AI will fully replace software engineers in the short term I am reminded of stuff like this

#

carmine nacelle Apr 12, 2026, 5:50 PM

#

Wait, are all filesystems all inside of libblockfs and POSIX server in managarm?

hexed acorn Apr 12, 2026, 5:50 PM

#

on-disk fses are handled by libblockfs yes

#

but its a dso that is linked into storage drivers, not a central server

carmine nacelle Apr 12, 2026, 5:51 PM

#

thonk

#

I wanted to make my disk drivers and filesystems be separate servers...

carmine nacelle Apr 12, 2026, 5:53 PM

#

carmine nacelle Anyway, since I can't be bothered with writing a proper posix_spawn, I think I'l...

But this seems to be very complicated

plain jungle Apr 12, 2026, 6:14 PM

#

I don't think separate servers are substantially more complicated

#pmOS - microkernel OS for RISC-V, x86, and LoongArch