SnowOS | Operating System Development | Page 5

narrow storm Mar 25, 2026, 3:11 PM

#

and most things by default target it for non MSVC compilers etc

hushed spire Mar 25, 2026, 3:11 PM

#

fair

#

anyhow Imma go write a todo list for today :3

narrow storm Mar 25, 2026, 3:12 PM

#

#

on windows ABI you cannot pass a 128 bit return value

#

slongeslab

#

and only 4 registers are used for arguments

#

otherwise its not as bad as i rememeber?

void elm Mar 25, 2026, 3:15 PM

#

it doesnt allow to pass struct fields as registers

#

so e.g. string_view or similar gets a lot more expensive than on sysv

narrow storm Mar 25, 2026, 3:15 PM

#

what does it do in that case?

#

dump it onto the stack?

void elm Mar 25, 2026, 3:15 PM

#

pointer to the struct on the caller's stack i think

narrow storm Mar 25, 2026, 3:16 PM

#

also sysv is generaly just nice to work with in assembly imo

#

#

like look at this stack crap

hushed spire Mar 25, 2026, 3:18 PM

#

KEKW

#

lol

narrow storm Mar 25, 2026, 3:18 PM

#

On sysv I never really have to touch the stack for like most things

#

Esp for tiny leaf functions im making

#

Like this

hushed spire Mar 25, 2026, 3:21 PM

#

uhhh, is there a flag to tell clang to use ths SysV ABI? I'd rather not spam __attribute__((sysv_abi)) everywhere

narrow storm Mar 25, 2026, 3:21 PM

#

its the default in 99% of cases
its a compile flag

#

hushed spire Mar 25, 2026, 3:22 PM

#

ah okay

narrow storm Mar 25, 2026, 3:22 PM

#

You also want to disable the redzone

#

Because it’s a fancy part of the ABI

#

And for kernels interrupt handlers will fuck up the redzone

#

It’s basically a part of the stack for leaf functions to use as scratch space and not need to allocate from the shack

#

And IRQ handlers may push shit inside of it

#

🫠

#

Well

hushed spire Mar 25, 2026, 3:23 PM

#

oh

#

uh 💀

narrow storm Mar 25, 2026, 3:23 PM

#

Rather the CPU will

narrow storm Mar 25, 2026, 3:23 PM

#

hushed spire oh

Your using a MSVC target lol

#

That won’t support sysv

hushed spire Mar 25, 2026, 3:24 PM

#

I just told clang "make PE" pepekek

narrow storm Mar 25, 2026, 3:24 PM

#

sysv is a lot more of a Unix thing

#

PSP1G_pspTrollar

hushed spire Mar 25, 2026, 3:24 PM

#

narrow storm That won’t support sysv

figured

narrow storm Mar 25, 2026, 3:24 PM

#

narrow storm sysv is a lot more of a Unix thing

But also something that you possibly want to use in any case

hushed spire Mar 25, 2026, 3:29 PM

#

uhhh, I think that's it? Not particularly interesting but eh pain

[] - Tweak Spinlock Implementation
[] - Clean up code some
[] - Timer Interrupts

narrow storm Mar 25, 2026, 3:30 PM

#

use LAPIC for timer IRQs

#

:3

#

its garunteed on all x86-64 chips

#

just needs clibration from another timer

hushed spire Mar 25, 2026, 3:30 PM

#

narrow storm use LAPIC for timer IRQs

Ye that's what I'm planning one :3

hushed spire Mar 25, 2026, 3:30 PM

#

narrow storm just needs clibration from another timer

was gonna use the HPET for that

narrow storm Mar 25, 2026, 3:30 PM

#

waow based based based based based

narrow storm Mar 25, 2026, 3:31 PM

#

hushed spire was gonna use the HPET for that

my kernel picks from ACPI Timer -> HPET -> PIT

#

then if the TSC is constant calibrate that

#

then use that for clibrstion of LAPIC

hushed spire Mar 25, 2026, 3:31 PM

#

oh ye, I keep forgetting about the ACPI timer 😆

hushed spire Mar 25, 2026, 4:02 PM

#

Oh! I also need to update to Limine 11

hushed spire Mar 25, 2026, 4:29 PM

#

I think this should work?

static inline VOID SpinlockAcquire(PSPINLOCK Lock)
    {
        // Get the state of the IF before disabling interrupts
        Lock->IntsEnabled = HalInterruptsEnabled();

        __asm__ volatile ("cli" ::: "memory");

        while (true)
        {
            if (!__atomic_exchange_n(&Lock->Flag, 1, __ATOMIC_ACQUIRE))
            {
                return;
            }

            while (__atomic_load_n(&Lock->Flag, __ATOMIC_RELAXED))
            {
                __asm__ volatile ("pause" ::: "memory");
            }
        }
    }

    static inline VOID SpinlockRelease(PSPINLOCK Lock)
    {
        __atomic_store_n(&Lock->Flag, 0, __ATOMIC_RELEASE);
        if (Lock->IntsEnabled)
        {
            __asm__ volatile ("sti" ::: "memory");
        }

        else
        {
            __asm__ volatile ("cli" ::: "memory");
        }
    }

narrow storm Mar 25, 2026, 4:31 PM

#

i wouldnt store it in the lock

#

it should be a return value

#

and you have the atomics reversed to be proper

#

and the return

#

🥀

#

also i hatr this code style but trl

hushed spire Mar 25, 2026, 4:33 PM

#

narrow storm it should be a return value

how come GuraThink

narrow storm Mar 25, 2026, 4:34 PM

#

somthing takes the locks with IRQs disabled

#

then somthing takes it with them enabled

#

then when the first one returns

#

they get enabled

#

💥

#

using the [[nodiscard]] attrbute makes it so you cannot ignore the return value

hushed spire Mar 25, 2026, 4:35 PM

#

ah

#

okie then

narrow storm Mar 25, 2026, 4:35 PM

#

making you pass it to the unlock funcion or similar

hushed spire Mar 25, 2026, 4:45 PM

#

narrow storm making you pass it to the unlock funcion or similar

this better?

[[nodiscard]]
static inline BOOL SpinlockAcquire(PSPINLOCK Lock)
{
  // Get the state of the IF before disabling interrupts
  BOOL IntsEnabled = HalInterruptsEnabled();

  __asm__ volatile ("cli" ::: "memory");

  while (true)
  {
    if (!__atomic_exchange_n(&Lock->Flag, 1, __ATOMIC_ACQUIRE))
    {
      break;
    }

    while (__atomic_load_n(&Lock->Flag, __ATOMIC_RELAXED))
    {
      __asm__ volatile ("pause" ::: "memory");
    }
  }

  return IntsEnabled;
}

static inline VOID SpinlockRelease(PSPINLOCK Lock, BOOL IntsEnabled)
{
  __atomic_store_n(&Lock->Flag, 0, __ATOMIC_RELEASE);
  if (IntsEnabled)
  {
    __asm__ volatile ("sti" ::: "memory");
  }

  else
  {
    __asm__ volatile ("cli" ::: "memory");
  }
}

narrow storm Mar 25, 2026, 4:49 PM

#

yes i think so

hushed spire Mar 25, 2026, 4:50 PM

#

party

narrow storm Mar 25, 2026, 4:57 PM

#

But please

#

Reconsider this coding style

#

🥀

hushed spire Mar 25, 2026, 4:57 PM

#

no trl

narrow storm Mar 25, 2026, 4:57 PM

#

You don’t even need the brackets

hushed spire Mar 25, 2026, 4:57 PM

#

what about it?

narrow storm Mar 25, 2026, 4:57 PM

#

You can do an if and else

hushed spire Mar 25, 2026, 4:57 PM

#

which brackets?

narrow storm Mar 25, 2026, 4:58 PM

#

static inline VOID SpinlockRelease(PSPINLOCK Lock, BOOL IntsEnabled) {
  __atomic_store_n(&Lock->Flag, 0, __ATOMIC_RELEASE);
  if (IntsEnabled)
    __asm__ volatile ("sti" ::: "memory");
  else
    __asm__ volatile ("cli" ::: "memory");
}

#

thats valid

hushed spire Mar 25, 2026, 4:58 PM

#

I personally don't really like how that looks, hence why I didn't do it KEKW

narrow storm Mar 25, 2026, 4:59 PM

#

But why do you make it take like 3x likes of code

#

is it just to get your stats up

#

heh_face

hushed spire Mar 25, 2026, 4:59 PM

#

narrow storm But why do you make it take like 3x likes of code

does it really matter 😭

narrow storm Mar 25, 2026, 4:59 PM

#

No

#

I just can’t read it that well

#

wahhgone

hushed spire Mar 25, 2026, 5:00 PM

#

lol

narrow storm Mar 25, 2026, 5:00 PM

#

I hate gnu styles

hushed spire Mar 25, 2026, 5:00 PM

#

I mean ig I can change it zerotwoshrug

#

party

#

I mean there's nothing for the other CPUs to do but hey KEKW

narrow storm Mar 25, 2026, 5:09 PM

#

What’s the bootstrap allocator for?

#

For a freelist PMM and a buddy you can store the freelist nodes inside of the free pages themselves ;3

hushed spire Mar 25, 2026, 5:10 PM

#

allocating pages for the PFNdb and mapping said PFNdb

narrow storm Mar 25, 2026, 5:10 PM

#

Ahh so not a bump allocator etc

#

Just a PMM

hushed spire Mar 25, 2026, 5:10 PM

#

no it is a bump allocator 🤣

#

also If I decide to go down the route of "scheduler inits before everything else" I'm pretty sure I need a bootstrap allocator as well

narrow storm Mar 25, 2026, 5:15 PM

#

I mean you could just delay it until you have a heap setup

#

Do GDT IDT (to catch faults and panic) PMM/PFNDB VMM Heap -> scheduler

hushed spire Mar 25, 2026, 5:17 PM

#

ik but liam said it was a good idea to setup the scheduler first before pretty much anything else :3 (ofc I'm doing it the easier way first because my brain can't handle that order of operations right now pain )

narrow storm Mar 25, 2026, 5:18 PM

#

Why would you want a scheduler before you can do basic memory management?

hushed spire Mar 25, 2026, 5:18 PM

#

lots of things depend on the scheduler, or so I've been told zerotwoshrug

narrow storm Mar 25, 2026, 5:19 PM

#

My kernel inits the schedular very late lol

#

You also possibly want timers before the scheduler

hushed spire Mar 25, 2026, 5:23 PM

#

I found it

#

I really should just take a screenshot of it at this point

#

so I don't have to go searching for it every time 🤣

narrow storm Mar 25, 2026, 5:27 PM

#

How would you setup the scheduler if you cannot allocate memory properly?

hushed spire Mar 25, 2026, 5:28 PM

#

Ask liam not me KEKW but from my understanding, the scheduler doesn't really rely upon process creation to work, all it does is look a queue and say "yup, that thread gets to run next" pepekek

hushed spire Mar 26, 2026, 2:12 AM

#

flobsh

#

yikes

#

okay both Limine and Flanterm should be updated now :3 lemme push the commit

#

there! party

hushed spire Mar 26, 2026, 2:34 AM

#

hushed spire uhhh, I think that's it? Not particularly interesting but eh <:pain:679047090474...

also I got distracted lol

#

I'll uh, work on timer interrupts now ^w^

#

was debating on whether or not to support the PIT. On one hand, I would think any modern PC would have the HPET/ACPI Timer/TSC, on the other hand, I presume the PIT's fairly simple so it's kind of seems silly to not support it zerotwoshrug

#

would ACPI be a Hal thing or a Ke thing, or maybe something else? GuraThink

#

idk I'll put it in Hal for now and I can move it somewhere else later if I need to

hushed spire Mar 26, 2026, 3:07 AM

#

siiiiigh

hushed spire Mar 26, 2026, 3:24 AM

#

huh? 😭

#

sadcrydepression

#

nothing is ever simple

#

what if I just copy the header to freestnd-c-hdrs trl

#

no

#

nooo

#

once again, clang posing as MSVC is causing problems (at least I think that's the issue, intrin.h seems to be a MSVC thing)

#

okay wait hol' up

#

haha! I shrimply commented out the include in compiler.h KEKW

hushed spire Mar 26, 2026, 4:02 AM

#

hushed spire idk I'll put it in Hal for now and I can move it somewhere else later if I need ...

don't ask why I said this, I don't think this makes any sense pepekek

#

hmmm GuraThink

#

I'm guessing it's mapped with the HHDM?

#

seems like it

hushed spire Mar 26, 2026, 4:35 AM

#

FeelsBadMan

#

I really gotta implement proper VA allocation

#

in fact

#

I'll probably start on that right after timer interrupts

#

I'm just now seeing an issue with this code pepekek

#

big oof

#

I'll read the vmem paper tmrw

hushed spire Mar 26, 2026, 4:50 AM

#

hushed spire <:FeelsBadMan:575445132535398422>

this entire function is broken in fact (I'm fairly sure) what the frick was I thinking when I wrote this 🤣

hushed spire Mar 26, 2026, 5:12 AM

#

hushed spire I'll probably start on that right after timer interrupts

Y'know what? on second thought maybe we'll go ahead and implement vmem tmrw lol

#

I can't wait to work on scheduling :3

#

fricking context-switching is what thwarted me the last time I tried to write SnowOS 😭

#

as evidenced by my crashouts trl

visual walrus Mar 26, 2026, 7:49 AM

#

you probably don't want to auto enable interrupts, lol

hushed spire Mar 26, 2026, 4:54 PM

#

visual walrus you probably don't want to auto enable interrupts, lol

I'm aware, it's been fixed KEKW

hushed spire Mar 27, 2026, 4:42 AM

#

I just didn't really have the energy to work on this today

#

pepepray tmrw for sure

hushed spire Mar 27, 2026, 7:40 PM

#

I should implement a ring buffer for logging GuraThink

narrow storm Mar 27, 2026, 7:51 PM

#

hushed spire I should implement a ring buffer for logging <:GuraThink:1387252570510327808>

It’s not so hard

#

I’d probably have a staging page or a few on BSS to log before your allocator is up

#

Then allocate a larger buffer

#

And copy it and reclaim the old buffer

#

Then you can let userspace read it

#

Ala dmesg

hushed spire Mar 27, 2026, 7:54 PM

#

ahhh, okie

#

seems simple enough :3

narrow storm Mar 27, 2026, 7:57 PM

#

Make sure that it’s page aligned on the linker script

#

And that it’s padded properly etc

#

Though you could maybe use attributes

hushed spire Mar 27, 2026, 7:59 PM

#

narrow storm Make sure that it’s page aligned on the linker script

I don't have a linker script meme

narrow storm Mar 27, 2026, 7:59 PM

#

pain

hushed spire Mar 27, 2026, 7:59 PM

#

I dunno how to write one for PEs 😭

narrow storm Mar 27, 2026, 8:00 PM

#

Why do you need to use a PE for the kernel

hushed spire Mar 27, 2026, 8:03 PM

#

narrow storm Why do you need to use a PE for the kernel

well I don't need to but I wanna

vivid helm Mar 27, 2026, 10:49 PM

#

hushed spire I don't have a linker script <:meme:575445094589661215>

You don't need a linker script for making a section for any purpose in PE. Just look into samples of windows drivers for example. pragmas.
resource heaven:

https://learn.microsoft.com/en-us/cpp/preprocessor/section?view=msvc-170

section pragma

Learn more about the section pragma directive in Microsoft C/C++

vivid helm Mar 27, 2026, 10:50 PM

#

narrow storm Why do you need to use a PE for the kernel

It's incomparably better than elf.

hushed spire Mar 28, 2026, 3:37 PM

#

I'm going on a road trip party

#

Which means SnowOS time!!!! letsgo (or at least til my laptop dies in the car KEKW )

hushed spire Mar 28, 2026, 7:16 PM

#

I should probably consider writing some unit tests

hushed spire Mar 28, 2026, 8:09 PM

#

Okay wait, my old Slab impl doesn't really sound like the same thing the paper describes. It seems like I pretty much wrote a Slab but just without object caching, which... seems like it kind of defeats the point of a Slab 😅

hushed spire Mar 28, 2026, 8:45 PM

#

I should also take a looksie at how Windows does this thinkong

hushed spire Mar 29, 2026, 12:05 AM

#

Okay Vmem doesn't seem too dreadfully hard (I'm still too dumb to completely understand it but oh well trl )

hushed spire Mar 29, 2026, 1:16 AM

#

I should also read up on UVM :3

narrow storm Mar 29, 2026, 1:45 AM

#

hushed spire Okay Vmem doesn't seem too dreadfully hard (I'm still too dumb to completely und...

If you manage to figure it out

#

Let me know

#

PSP1G_pspTrollar

hushed spire Mar 29, 2026, 1:46 AM

#

togescared

narrow storm Mar 29, 2026, 1:48 AM

#

Don’t be scared

#

I only skimmed it while I was half asleep

#

💀

#

That’s why I don’t get it to be clear

hushed spire Mar 29, 2026, 1:48 AM

#

oh lol

#

it's still confuzzling me tho KEKW

narrow storm Mar 29, 2026, 1:48 AM

#

It dosnt look too bad

#

I think it’s like for virtual allocations?

hushed spire Mar 29, 2026, 1:49 AM

#

ye (resource allocation in general it seems, but virtual allocations seems to be the big thing)

narrow storm Mar 29, 2026, 1:50 AM

#

Virtual allocations will be heavily wanted by Userspace iirc?

#

For the kernel I plan to use my buddy allocater for large allocations for kernel stacks etc

hushed spire Mar 29, 2026, 2:10 AM

#

well, while I try to wrap my head around vmem, imma implment page unmapping functions, and maybe try to figure out how I can make the hal a dll thinkong

narrow storm Mar 29, 2026, 2:11 AM

#

Btw page unmap can be the same as mapping a page except setting its PTE to zero and doing a tlb invalidation

#

And maybe checking if that level is empty etc

#

(If you want to free the tables as you go but that can be seperate)

hushed spire Mar 29, 2026, 2:13 AM

#

narrow storm Btw page unmap can be the same as mapping a page except setting its PTE to zero ...

yup, that's exactly what I'm doing :3

narrow storm Mar 29, 2026, 2:13 AM

#

Nice :3

hushed spire Mar 29, 2026, 3:04 AM

#

okay so an arena is just a range of integers, simple enough

#

but like this I don't understand
The slab allocator can provide object caching for any vmem arena

#

doesn't the slab depend upon vmem?

#

how is it providing object caching then? 😭

lunar vortex Mar 29, 2026, 3:56 AM

#

hushed spire how is it providing object caching then? 😭

They both depend on each other

#

vmem has this quantum caching thing that uses the slab allocator

#

I don't think it's worth it tho

#

https://github.com/rdmsr/tinyvmem

GitHub

GitHub - rdmsr/tinyvmem: Implementation of the VMem Resource Allocator

Implementation of the VMem Resource Allocator. Contribute to rdmsr/tinyvmem development by creating an account on GitHub.

#

Here's my implementation from a while ago

hushed spire Mar 29, 2026, 3:57 AM

#

thx! I'll take a look at it :3

lunar vortex Mar 29, 2026, 3:58 AM

#

It kinda sucks tho

#

Well it's not that good

#

it's fine

hushed spire Mar 29, 2026, 3:59 AM

#

can't suck anymore than whatever I'll end up writing pepekek

narrow lake Mar 29, 2026, 9:31 AM

#

lunar vortex I don't think it's worth it tho

It’s pretty easy to implement anyways

hushed spire Mar 29, 2026, 8:25 PM

#

okie, back to trying to implement vmem :3

hushed spire Mar 29, 2026, 8:43 PM

#

is there some reaons this needs to return a pointer confusedshiki

#

cuz like, doesn't that mean I need some way to allocate a vmem arena

#

oh wait

#

ah

At compile time we statically declare a few vmem arena structures and boundary tags to get us through boot.

#

well I guess there's my answer

lunar vortex Mar 29, 2026, 8:51 PM

#

hushed spire cuz like, doesn't that mean I need some way to *allocate* a vmem arena

solaris has a convention to do that for some reason, but I think it is better to let the caller pass a pointer

hushed spire Mar 29, 2026, 8:51 PM

#

I was thinking that too GuraThink

#

okie dokie :3

hushed spire Mar 29, 2026, 10:57 PM

#

well I have a function that chucks stuff into a vmem arena struct

#

not much

#

but better than nothing KEKW

hushed spire Mar 30, 2026, 3:22 PM

#

make to the osdev grind

#

maybe I'll actually do something useful today KEKW

hushed spire Mar 31, 2026, 2:57 PM

#

hushed spire maybe I'll actually do something useful today <:KEKW:913708372329644042>

no

#

pepekek

#

I'm on vacation gimme a break lol

#

I will actually try today tho

hushed spire Mar 31, 2026, 3:26 PM

#

gonna write a generic linked list impl

#

should class names be the same style as struct names or in PascalCase like I usually do? GuraThink

hushed spire Mar 31, 2026, 3:48 PM

#

function overloading! Everyone's favorite KEKW

#

having data in the second overload is actually kind of redundant now that I think of it

#

actually overloading might be annoying for debugging now that I think about it

#

ofc I can't debug my kernel rn anyhow so pain

hushed spire Mar 31, 2026, 4:00 PM

#

hushed spire actually overloading might be annoying for debugging now that I think about it

I say as if I don't use namespaces, which cause namemangling anyhow pepekek

visual walrus Mar 31, 2026, 4:11 PM

#

hushed spire gonna write a generic linked list impl

want mine in case?

#

it works somewhat well

hushed spire Mar 31, 2026, 4:11 PM

#

sure zerotwoshrug

visual walrus Mar 31, 2026, 4:12 PM

#

https://github.com/Diamantino-Op/HorizonOS/blob/main/src/libs/hos-base/src/LinkedList.hpp

hushed spire Mar 31, 2026, 4:26 PM

#

oh ye, I forgor I can overload new thinkong

hushed spire Mar 31, 2026, 5:22 PM

#

also seems like the segment list needs to be ordered

hushed spire Mar 31, 2026, 6:25 PM

#

range to use for the kernel heap``` okay so clearly this is the arena for VAs used for the kernel heap

#

Once we have the heap arena, we can create new boundary tags dynamically but then I can suddenly allocate things now that I have it? Even though it doesn't actually allocate pages

#

am I supposed to start the slab after creating the heap_arena?

#

I would think?

hushed spire Mar 31, 2026, 7:47 PM

#

also pretty sure I need to use a hashmap for this

#

so I need to learn about those

#

ah, they don't seem that complicated

hushed spire Mar 31, 2026, 11:30 PM

#

I shouldn't be having this much trouble trying to implement this sadge

#

what if I just try to do a more naive implementation of a VA allocator and come back to this later pepekek

#

just so that I can get to do other things

#

eh screw it I think I'll do that zerotwoshrug

mellow current Apr 1, 2026, 1:31 AM

#

sub title should be winter will come

hushed spire Apr 1, 2026, 2:12 PM

#

I'm officially announcing that SnowOS will be completely rewritten into Rust! party

hushed spire Apr 1, 2026, 3:58 PM

#

SnowOS - A modern OS written entirely in Rust

hushed spire Apr 2, 2026, 3:02 PM

#

SnowOS

#

I'm pulling y'all's leg pepekek

#

I don't think you could pay me enough to use a different language for this lol

#

anyhow

narrow storm Apr 2, 2026, 3:03 PM

#

hushed spire I don't think you could pay me enough to use a different language for this lol

10 grand?

hushed spire Apr 2, 2026, 3:03 PM

#

I really would like to get VA allocation setup to day 😔

hushed spire Apr 2, 2026, 3:04 PM

#

narrow storm 10 grand?

no trl

narrow storm Apr 2, 2026, 3:04 PM

#

25 grand

#

PSP1G_pspTrollar

#

When I get a job

hushed spire Apr 2, 2026, 3:04 PM

#

C++ foevaaaaa!!! >:3

narrow storm Apr 2, 2026, 3:05 PM

#

You mark a hard bargain

#

200 grand when I get a load and take 3 mortgages on my house

#

PSP1G_pspTrollar

#

That’s my final offer

hushed spire Apr 2, 2026, 3:11 PM

#

https://tenor.com/view/soggycereal-gif-18454991

Tenor

#

no :3

hushed spire Apr 3, 2026, 12:48 AM

#

okie I'm finally getting somewhere party

#

userspace impl of vmem is going well

#

I can't imagine actually implementing it in the kernel will be much harder :3

hushed spire Apr 3, 2026, 1:44 AM

#

also might make the code style slightly less NT-pilled, for personality™

hushed spire Apr 3, 2026, 3:31 AM

#

oh ye

#

still wanna make Hal into a DLL

#

think I might get started on that

hushed spire Apr 3, 2026, 4:36 AM

#

okay I think I'll focus on vmem for now, this seems slightly harder than I initially thought 😅

narrow storm Apr 3, 2026, 12:21 PM

#

hushed spire also might make the code style *slightly* less NT-pilled, for *personality*™

You need to make it more NT pilled than NT

hushed spire Apr 4, 2026, 3:22 PM

#

been a kind of slow few days lol

#

hopefully I'll get something working today

hushed spire Apr 5, 2026, 2:09 PM

#

hm

#

guess my insertion code isn't working meme

#

ah hah! Fixed it :3

hushed spire Apr 6, 2026, 2:56 PM

#

gonna be heading back home today

#

maybe I'll actually get some work done :p

#

currently working on the freelists for an arena

#

I think I'll make my goal for this month to be to launch a userspace process

#

Surely I can get the scheduler done by the end of the month trl

visual walrus Apr 6, 2026, 3:19 PM

#

hushed spire I think I'll make my goal for this month to be to launch a userspace process

it was a bit of a pain for me XD

#

but now i can do that!

hushed spire Apr 6, 2026, 3:19 PM

#

Mine didn't even work the last time I tried KEKW

visual walrus Apr 6, 2026, 3:20 PM

#

the big problem i had was to compile llvm's libc++ and mlibc

#

for my OS

hushed spire Apr 6, 2026, 3:21 PM

#

ahhh

#

I still wanna try and look into porting mlibc GuraThink

visual walrus Apr 6, 2026, 3:21 PM

#

it's not that bad

hushed spire Apr 6, 2026, 3:22 PM

#

if your POSIX-compliant maybe not pepekek

#

poggers

#

we're getting somewhere!

hushed spire Apr 6, 2026, 9:03 PM

#

oh hey

#

I've apparently reached my first 1000 loc party

hushed spire Apr 6, 2026, 9:45 PM

#

Might clean the code up some before I continue

hushed spire Apr 7, 2026, 5:42 PM

#

hushed spire also might make the code style *slightly* less NT-pilled, for *personality*™

I forgot how boring this was KEKW

hushed spire Apr 7, 2026, 8:14 PM

#

if I ever try to change to code-style again, please yell at me pepekek

#

I do think I like this style better though

#

extern "C" void keMain(void *snowbootInfo)
{
    keRunConstructors();

    hal::initialize();
    hal::printString("Snow Operating System (c) 2025, 2026 UtsumiFuyuki\r\n");
    ke::print("Yuki Kernel Version %d.%d.%d\r\n", YUKI_VERSION_MAJOR, YUKI_VERSION_MINOR, YUKI_VERSION_PATCH);
    ke::print("Booted by: ");
    
    if (snowbootInfo == nullptr)
    {
        ke::print("Limine %s\r\n\r\n", hal::blVersion());
    }
    else
    {
        ke::print("SnowBoot\r\n");
    }

    hal::initCpu();
    mm::earlyInit();
    hal::initializePaging();
    mm::initialize();
    mm::initializeVmm();

    // We're done, just hang...
    hal::haltCpu();
}

narrow storm Apr 7, 2026, 8:59 PM

#

hushed spire if I ever try to change to code-style again, please yell at me <:pepekek:1255025...

this code style is still ass because the
whatever
{
PSP1G_pspTrollar
}

hushed spire Apr 7, 2026, 8:59 PM

#

Why is that an issue? 😭

narrow storm Apr 7, 2026, 8:59 PM

#

nothing

#

i just dislike it

#

trl

hushed spire Apr 7, 2026, 8:59 PM

#

pepekek

narrow storm Apr 7, 2026, 8:59 PM

#

its too gnu

#

also why are you putting \r\n in every line

#

when your printf can convert \n to \r\n

hushed spire Apr 7, 2026, 9:00 PM

#

I don't think I've ever seen a line of gnu code so I wouldn't know zerotwoshrug

hushed spire Apr 7, 2026, 9:00 PM

#

narrow storm when your printf can convert \n to \r\n

I can't be bothered rn, mostly

#

more pressing matters

narrow storm Apr 7, 2026, 9:02 PM

#

narrow storm Apr 7, 2026, 9:02 PM

#

hushed spire I can't be bothered rn, mostly

its 4 lines of code

#

🥀

hushed spire Apr 7, 2026, 9:04 PM

#

narrow storm

Fine I'll change it 😛 for what it's worth I flip between the two anyhow trl

narrow storm Apr 7, 2026, 9:05 PM

#

#

That’s straight up nasty 😭

#

At least get a clang format file

hushed spire Apr 7, 2026, 9:06 PM

#

narrow storm That’s straight up nasty 😭

I'm not a very experienced programmer gimme a break nooo

#

I'm trying to find mah style sigmapepe

narrow storm Apr 7, 2026, 9:07 PM

#

Fair

#

trl

#

I’ve been doing TS over like 6 years now prolly?

hushed spire Apr 7, 2026, 9:07 PM

#

coolio :3

visual walrus Apr 7, 2026, 9:46 PM

#

narrow storm

the fact that Allman comes standard with C#...

#

i hate it XD

hushed spire Apr 8, 2026, 3:13 AM

#

bmlflllaldjks

#

okay

#

I think I'm done refactoring for now

#

pushed the commit

serene lake Apr 8, 2026, 3:14 AM

#

could have sworn i was following this already

#

i am now 😄

hushed spire Apr 8, 2026, 3:14 AM

#

party

#

yay

#

And this thread has reached 20 stars! party2

#

~~I definitely don't deserve it but eh~~

serene lake Apr 8, 2026, 3:25 AM

#

the font makes your name look like O++ Shill

#

#

barely any black separating space, because of the glow

hushed spire Apr 8, 2026, 3:28 AM

#

Oh it does KEKW

serene lake Apr 8, 2026, 3:28 AM

#

maybe its a special secret blood type

#

i am O positive positive

#

😄

hushed spire Apr 8, 2026, 3:28 AM

#

pepekek

hushed spire Apr 8, 2026, 3:15 PM

#

when'd I start trying to implment vmem?

#

like a week ago pain

#

anyhow

#

think I hafta implement vmemFree now

#

just trying to get something simple setup :3

#

also need to figure out what the heck I do about sources 😭

hushed spire Apr 8, 2026, 5:19 PM

#

freeing is a bit more complicated than allocating lol

lunar vortex Apr 8, 2026, 6:30 PM

#

freeing is simpler imo

#

You just have one logic

#

Allocation has three policies

lunar vortex Apr 8, 2026, 6:31 PM

#

lunar vortex You just have one logic

You do have to keep coalescing in mind tho

hushed spire Apr 8, 2026, 6:46 PM

#

I only have instantfit implemented at the moment lol. Though bestfit and nextfit don't seem that complicated

hushed spire Apr 8, 2026, 9:05 PM

#

WE GOT SOMETHING WORKING!!! letsgo

hushed spire Apr 9, 2026, 12:15 AM

#

took a break :3

#

anyhow

#

now I'm gonna make it actually coalesce segments now

hushed spire Apr 9, 2026, 12:45 AM

#

seems to work! ^w^

#

I guess I'll find out later if it's broken trl

#

Anyhow now I have to make it actually put the newly freed segment back onto the freelist, because currently I'm only manipulating the segment list

hushed spire Apr 9, 2026, 1:26 AM

#

this is so messy KEKW

#

granted this is my first time writing a coalescing, ordered freelist, so I guess it's bound to be a little messy

lunar vortex Apr 9, 2026, 1:28 AM

#

what are you having trouble with

hushed spire Apr 9, 2026, 1:29 AM

#

nothing in particular I just think the way I'm doing it is really messy pepekek

#

lunar vortex Apr 9, 2026, 1:31 AM

#

#

this is what I do

lunar vortex Apr 9, 2026, 1:33 AM

#

hushed spire

ahhh nooo

#

why are you using pow

#

you can very quickly get one size's freelist using log2

#

and log2 is pretty damn fast

hushed spire Apr 9, 2026, 1:33 AM

#

uh, is there an issue with pow?

#

wut's log2?

lunar vortex Apr 9, 2026, 1:33 AM

#

yes

#

dont do that

lunar vortex Apr 9, 2026, 1:34 AM

#

hushed spire wut's log2?

base 2 logarithm

hushed spire Apr 9, 2026, 1:34 AM

#

ah

lunar vortex Apr 9, 2026, 1:34 AM

#

you can implement it very cheaply using ctlzl

hushed spire Apr 9, 2026, 1:34 AM

#

yeah I dunno how to do logarithms so that probably explains why I didn't think of it pepekek now that you say it though that does make more sense

lunar vortex Apr 9, 2026, 1:34 AM

#

you dont ever need to iterate over 64 freelists

lunar vortex Apr 9, 2026, 1:38 AM

#

lunar vortex

the rotor handling is wrong i am an idiot

hushed spire Apr 9, 2026, 1:39 AM

#

I don't even know what "rotor" means 😭

lunar vortex Apr 9, 2026, 1:39 AM

#

it's for next fit

#

it's the last allocated segment

hushed spire Apr 9, 2026, 1:39 AM

#

ah

#

okie :3

#

okay I went ahead and pushed the commit, I'll work on the freelist manipulation tmrw :p

hushed spire Apr 9, 2026, 10:25 PM

#

aight, Imma go work on my math some, and then by like, 7 or 8 I'll do some more OSDev >:3

hushed spire Apr 10, 2026, 3:46 PM

#

Aight, I need to write a log2 function, uh, however I'm supposed to do that zerotwoshrug

serene lake Apr 10, 2026, 5:04 PM

#

feel free to look at mine, be aware it isn't my work. it's a combination of freebsd code and questioning an LLM to kind of understand the concepts as a non mathematician

#

but I know it works, and didn't didn't want to spend time making maths stuff

narrow storm Apr 10, 2026, 5:06 PM

#

According to stackoverflow if you just need it to work for integers

#

Or looping methods

hushed spire Apr 10, 2026, 5:08 PM

#

ah okie! neat :3

serene lake Apr 10, 2026, 5:09 PM

#

it's more work for floating point

hushed spire Apr 10, 2026, 5:10 PM

#

I can't be bothered to do floating point rn KEKW

serene lake Apr 10, 2026, 5:10 PM

#

check log and log2f

https://github.com/brainboxdotcc/retro-rocket/blob/master/src/maths.c#L158

hushed spire Apr 10, 2026, 5:10 PM

#

hushed spire I can't be bothered to do floating point rn <:KEKW:913708372329644042>

(I don't think I need it rn anyhow)

serene lake Apr 10, 2026, 5:10 PM

#

I needed it to work with double, and didn't know my naive approach was wrong until I tried to make an xm player

lunar vortex Apr 10, 2026, 5:18 PM

#

lunar vortex you can implement it very cheaply using ctlzl

This

serene lake Apr 10, 2026, 5:18 PM

#

narrow storm According to stackoverflow if you just need it to work for integers

this is undefined for zero and only works for positive integers

hushed spire Apr 10, 2026, 5:19 PM

#

lunar vortex This

what's ctlzl? 😅

serene lake Apr 10, 2026, 5:19 PM

#

the 2nd one never increments index so is an infinite loop

narrow storm Apr 10, 2026, 5:19 PM

#

serene lake this is undefined for zero and only works for positive integers

Me when you can add a few if statments

lunar vortex Apr 10, 2026, 5:19 PM

#

Sorry clzl

serene lake Apr 10, 2026, 5:19 PM

#

isn't there an sse2 log2?

hushed spire Apr 10, 2026, 5:19 PM

#

lunar vortex Sorry clzl

okay that makes more sense lol

lunar vortex Apr 10, 2026, 5:20 PM

#

Integer log2 is really cheap idk why you'd use sse or something more complex

narrow storm Apr 10, 2026, 5:20 PM

#

https://www.felixcloutier.com/x86/lzcnt

#

Crap

#

Is that the right one?

#

I cant find the ones your mentioning

lunar vortex Apr 10, 2026, 5:21 PM

#

Yeah but use gcc intrinsics

serene lake Apr 10, 2026, 5:22 PM

#

lunar vortex Integer log2 is really cheap idk why you'd use sse or something more complex

for when you want real maths

hushed spire Apr 10, 2026, 5:22 PM

#

just to ask, is there a reason I shouldn't use std::countl_zero()?

lunar vortex Apr 10, 2026, 5:22 PM

#

serene lake for when you want real maths

You don't need that here

serene lake Apr 10, 2026, 5:22 PM

#

if you're implementing libm you'd need it

lunar vortex Apr 10, 2026, 5:22 PM

#

He's not implementing libm

serene lake Apr 10, 2026, 5:22 PM

#

well they haven't said what the use case of log2 is

hushed spire Apr 10, 2026, 5:22 PM

#

what's libm? confusedshiki

serene lake Apr 10, 2026, 5:22 PM

#

generic maths lib

hushed spire Apr 10, 2026, 5:23 PM

#

serene lake well they haven't said what the use case of log2 is

finding which freelist a segment belongs to (vmem)

#

that is, freelist[n] contains all free segments
whose sizes are in the range [2n, 2n+1). To allocate a
segment we search the appropriate freelist for a
segment large enough to satisfy the allocation.```

serene lake Apr 10, 2026, 5:23 PM

#

isn't this just bit shifts

#

it's <<

#

and yeah lzcnt

#

I use the bit count intrinsics in my FS driver

lunar vortex Apr 10, 2026, 5:25 PM

#

What you wanna do is 64 - clzl(num) - 1

hushed spire Apr 10, 2026, 5:26 PM

#

okay yeah, figured :3

#

it works! party

#

uhhhh, how do I actually coalesce the block though pain because it's getting chucked onto the freelist the block it was allocated from isn't on

#

actually do I have to change the freelist a block is on upon allocation? Since it's size has changed?

lunar vortex Apr 10, 2026, 5:33 PM

#

hmm?

hushed spire Apr 10, 2026, 5:38 PM

#

Like, my initial free segment has a size of 0xF0000000 and get's put on freelist 31. And say eventually after some allocations it's size goes down to 0x70000000, which should now be on freelist 30, even though it should be on freelist 30 now, it remains on freelist 31, which seems like it would kind of defeat the point of having a size segregated freelist GuraThink

lunar vortex Apr 10, 2026, 5:38 PM

#

You would split it

hushed spire Apr 10, 2026, 5:39 PM

#

confusedshiki

lunar vortex Apr 10, 2026, 5:42 PM

#

hushed spire Like, my initial free segment has a size of `0xF0000000` and get's put on freeli...

You need to remove the segment and put it back on its appropriate freelist

hushed spire Apr 10, 2026, 5:42 PM

#

okay figured 👍

#

segment we search the appropriate freelist for a
segment large enough to satisfy the allocation.```
Slight question, I would *presume* that I also check freelists above n, in the case that freelist[n] is empty. I don't get the feeling I'm supposed to just fail the allocation if freelist n is empty, wouldn't make much sense ![KEKW](https://cdn.discordapp.com/emojis/913708372329644042.webp?size=128 "KEKW")

lunar vortex Apr 10, 2026, 5:56 PM

#

Yes

hushed spire Apr 11, 2026, 12:47 AM

#

okay I think I just have to fix this?

#

And then I should have an, albeit janky, technically working vmem implementation

#

Emphasis on the "technically" KEKW

#

it'll probably break eventually, but I'll cross that road when I get there

hushed spire Apr 11, 2026, 3:50 PM

#

I have awoken!

#

Back to implementing vmem :3

narrow storm Apr 11, 2026, 3:50 PM

#

No FRED implementing trl

#

I mean it would be a decent break

#

It’s actually not too bad to implement imo

#

Though it does look like your vmem is going well

hushed spire Apr 11, 2026, 3:52 PM

#

FRED I'll either try to implement after this or after I work on scheduling some GuraThink

#

I really wanna work on scheduling >:3

narrow storm Apr 11, 2026, 3:52 PM

#

FRED does swap up how you may do some of that with entry into userspace :3

#

(It’s not bad tho)

hushed spire Apr 11, 2026, 3:56 PM

#

uhh, as incomplete as my vmem implementation is I know that shouldn't be happening pepekek

#

time to spam printfs trl

narrow storm Apr 11, 2026, 4:02 PM

#

hushed spire uhh, as incomplete as my vmem implementation is I know that shouldn't be happeni...

lmfaoooo

hushed spire Apr 11, 2026, 4:03 PM

#

ah yes, vmemAllocate is broken, idk why I did it the way I did KEKW

#

for (auto *currentNode = arena->segmentList.getHead(); currentNode != nullptr; currentNode = currentNode->next) {
        // Find corresponding node in segment list
        if (currentNode->data.segmentBase == head->data.segmentBase && currentNode->data.type != VMEM_SEGMENT_SPAN) {
            if (currentNode->data.segmentSize == size) {
                currentNode->data.type = VMEM_SEGMENT_ALLOCATED;
                return currentNode->data.segmentBase;
            }
            auto *allocatedNode = &initNodes[nextFreeNode];
            allocatedNode->data.segmentSize = size;
            allocatedNode->data.segmentBase = currentNode->data.segmentBase;
            allocatedNode->data.type = VMEM_SEGMENT_ALLOCATED;

            currentNode->data.segmentBase += size;

            arena->segmentList.insert(currentNode, allocatedNode);
            nextFreeNode++;
            return allocatedNode->data.segmentBase;
        }
    }
``` for some reason I do this, which would *usually* work, except because the freelist doesn't get coalesced properly (working on figuring out how to do that) there's no longer a "corresponding node" and it just returns 0 lol

#

pretty simple fix tho :3

hushed spire Apr 11, 2026, 5:05 PM

#

freeing must also be broken I suppose pain

#

how it's broken idk kongoudisgust

#

but evidently uACPI is not happy

#

eh, I'll figure it out later :p

#

I honestly may work on scheduling some rn thinkong

#

since I don't actually need a timer

hushed spire Apr 11, 2026, 6:52 PM

#

well that ain't very helpful =/

hushed spire Apr 11, 2026, 10:08 PM

#

boy I sure wish LLDB would work lol

lunar vortex Apr 11, 2026, 10:08 PM

#

hushed spire well that ain't very helpful =/

try llvm-addr2line or just objdump it

hushed spire Apr 11, 2026, 10:12 PM

#

ah well that worked! :3

#

weird

#

I shouldn't be calling any uACPI functions atm =/

#

I'm guessing my context switch is screwed up in some way then pepekek

lunar vortex Apr 11, 2026, 10:14 PM

#

hushed spire I'm guessing my context switch is screwed up in some way then <:pepekek:12550253...

you may have rebuilt the executable and now that address is wrong

hushed spire Apr 11, 2026, 10:15 PM

#

ah yeah

#

oh god I'm dumb

#

KEKW

#

mov rsp, [rcx+128]

mov rax, [rcx+8]
mov rbx, [rcx+16]
mov rcx, [rcx+24]
mov rdx, [rcx+32]
``` gee, I wonder what the issue here could possibly be ![pepekek](https://cdn.discordapp.com/emojis/1255025354276470846.webp?size=128 "pepekek") (does it matter if I move in values from a thread struct into the gp regs, or should I put them on the stack?)

#

hmm, yeah seems like they should be on the stack GuraThink

narrow storm Apr 11, 2026, 10:30 PM

#

hushed spire ```x86asm mov rsp, [rcx+128] mov rax, [rcx+8] mov rbx, [rcx+16] mov rcx, [rcx+2...

What’s this for?

#

/ what’s the context for this

hushed spire Apr 11, 2026, 10:30 PM

#

context switch

narrow storm Apr 11, 2026, 10:30 PM

#

And Intel or Nasm

hushed spire Apr 11, 2026, 10:30 PM

#

intel

#

or well

#

nasm

#

what's the difference? confusedshiki

#

doesn't nasm use intel syntax?

narrow storm Apr 11, 2026, 10:31 PM

#

Wait fuck

#

I meant Intel or att

hushed spire Apr 11, 2026, 10:31 PM

#

lol

narrow storm Apr 11, 2026, 10:31 PM

#

trl

narrow storm Apr 11, 2026, 10:31 PM

#

hushed spire ```x86asm mov rsp, [rcx+128] mov rax, [rcx+8] mov rbx, [rcx+16] mov rcx, [rcx+2...

this should be all you need

#

extern void thread_switch(spinlock_t* spinlock, int irqs, uint64_t* old_sp, uint64_t new_sp);

#

then after calculating the thread its a

    uint64_t* old_rsp = &previous_thread->krsp;
    uint64_t  new_rsp = next_thread->krsp;
    thread_switch(&scheduler_spinlock, lock1r, old_rsp, new_rsp);

hushed spire Apr 11, 2026, 10:31 PM

#

aren't you supposed to iret?

narrow storm Apr 11, 2026, 10:32 PM

#

Not for a thread switch

hushed spire Apr 11, 2026, 10:32 PM

#

ah

narrow storm Apr 11, 2026, 10:32 PM

#

That’s for when you need a privelege change

narrow storm Apr 11, 2026, 10:32 PM

#

narrow storm this should be all you need

This does assume that you setup a threads stack properly though

#

And pre push things initially

#

eg

hushed spire Apr 11, 2026, 10:33 PM

#

uhhh, what does ret pop off the stack again?

narrow storm Apr 11, 2026, 10:33 PM

#

The instruction pointer

#

Call pushes the IP to return to

hushed spire Apr 11, 2026, 10:33 PM

#

okie

narrow storm Apr 11, 2026, 10:35 PM

#

And I push many function pointers there

#

So when one returns it goes to the next etc

narrow storm Apr 11, 2026, 10:35 PM

#

narrow storm this should be all you need

The first return here pops the first one off etc

#

And this will also break if you try and switch to the same exact thread iirc?

hushed spire Apr 11, 2026, 10:36 PM

#

so wut happens when there's only one thread in the queue?

narrow storm Apr 11, 2026, 10:36 PM

#

My schedular checks if the next thread is the same as the previous

#

And if so it won’t run switch

#

But just return

hushed spire Apr 11, 2026, 10:36 PM

#

ah

narrow storm Apr 11, 2026, 10:36 PM

#

this is the trampoline i use to make it properly follow sysv aswell

#

and this is to jump to CPL3 just to be complete

#

Oh btw sorry @hushed spire for dumping code

#

😅

#

I should have asked

hushed spire Apr 11, 2026, 10:43 PM

#

all good lol

#

coolio sigmapepe

narrow storm Apr 11, 2026, 10:49 PM

#

Yooooo

#

Make one print A and the other B

#

mmLol

#

or even better

#

make them take arguments proper etc

#

to control the behavior at thread start

hushed spire Apr 11, 2026, 10:49 PM

#

now uhh, lemme implement the rest of the switch, because currently it only does half the work KEKW

narrow storm Apr 11, 2026, 10:49 PM

#

(use malloc if you do a string)

hushed spire Apr 11, 2026, 10:49 PM

#

narrow storm to control the behavior at thread start

will do :3

hushed spire Apr 11, 2026, 10:50 PM

#

narrow storm (use malloc if you do a string)

I don't have malloc yet meme

narrow storm Apr 11, 2026, 10:50 PM

#

Or any kind of memory allocation

#

Because you can’t give it a local reference to a new thread

hushed spire Apr 11, 2026, 10:50 PM

#

true

narrow storm Apr 11, 2026, 10:50 PM

#

and you can only pass word sized values

#

Btw what ABI did you end up with?

hushed spire Apr 11, 2026, 10:59 PM

#

The Microsoft one I'm pretty sure

narrow storm Apr 11, 2026, 11:08 PM

#

My code is for sysv ABI

#

Because it’s actualy sane

#

mmLol

#

Your gonna need diff code then

hushed spire Apr 11, 2026, 11:09 PM

#

ye figured

#

I couldn't figure out how to compile a PE with the SysV ABI 😭

#

x64 ABI hasn't gotten in my way that much yet tho so zerotwoshrug

narrow storm Apr 11, 2026, 11:11 PM

#

Fingers crossed

#

Just tweak the code to save and restore and swap the right registers

hushed spire Apr 12, 2026, 4:38 PM

#

Okay, I'm gonna draw some, read OSTEP some, then I'll work more on thread-switching :3

#

after that I think I'm gonna go through the code and see if there's anything I can clean up

hushed spire Apr 12, 2026, 8:22 PM

#

Aight letsgo

narrow storm Apr 12, 2026, 8:23 PM

#

What did OSTEP tell ya?

hushed spire Apr 12, 2026, 8:24 PM

#

narrow storm What did OSTEP tell ya?

I haven't read it yet, it's too flipping windy outside nooo

#

I should just start reading it before I leave at this point KEKW

narrow storm Apr 12, 2026, 8:24 PM

#

FWIW it probably won’t effect your task switching code

#

Maybe your thread structs

hushed spire Apr 12, 2026, 8:24 PM

#

I feel like it's never not windy

narrow storm Apr 12, 2026, 8:24 PM

#

But not the low level code that’s required as per ABI and architecture

hushed spire Apr 12, 2026, 8:24 PM

#

true

narrow storm Apr 12, 2026, 8:25 PM

#

narrow storm Maybe your thread structs

Which is generally easy enough to change early on

#

And esp if you make static asserts for the position if things

#

So your assembly won’t magically fuck up

#

mmLol

hushed spire Apr 12, 2026, 8:27 PM

#

anyhow I think it should maybe work rn GuraThink in theory anyways. Current conundrum is that setupStack well, pushes things onto the stack, and when I try to preform the switch the old thread will push things onto the stack (because it doesn't know it's already been setup) and then 💥

narrow storm Apr 12, 2026, 8:28 PM

#

No

#

When you switch to the stack for the first time

#

It pops them off

hushed spire Apr 12, 2026, 8:28 PM

#

I'm thinking maybe just write a small trampoline zerotwoshrug

narrow storm Apr 12, 2026, 8:29 PM

#

That’s why you push them on the first time

#

And then when the initial thread gets preempted

#

It pushes things into the empty stack

#

So it just kinda works

#

it saves things on the current stack. swaps stacks. then returns

#

When you first setup scheduling

#

And enable it

#

The thing that gets preempted

#

Gets saved first

#

So it’s a bit funky but it works out for me

#

You can control that more specialy by making a dummy thread to get populated when it switches from it

narrow storm Apr 12, 2026, 8:32 PM

#

narrow storm it saves things on the current stack. swaps stacks. then returns

And this either returns back into the previous thread from when after it called the task switchi

#

Or it returns into the dummy things you put into the stack

hushed spire Apr 12, 2026, 8:32 PM

#

thinkong

narrow storm Apr 12, 2026, 8:33 PM

#

anything can call the scheudle function

#

not just an interupt

#

thread A -> calls task switch
the return value is pushed onto the stack
the asm pushes a bunch of things onto the stack
it swaps stacks
restores off of thread B's stack
Returns

#

thread B would have had the return value on its stack from when it called task switch

#

or if you had just made it

#

it wiuld have the things you pushed

#

for return to pop from

#

and jump to

hushed spire Apr 12, 2026, 8:37 PM

#

okay, so what happens if I had just made thread A and thread B for the first time, and then tried to switch to one of them? confusedshiki

narrow storm Apr 12, 2026, 8:37 PM

#

how i have it here is not the best for SMP as shouldSchedule should be CPU local but

#

the idle thrread

#

thats the first thread the schedular will try and use

#

the kmain function then gets assigned

#

to that bit

#

because it does a save state

#

so the stuff you allocated and pushed to it dosnt matter

#

it just gets reused into the idle loop

#

in the rewrite id make it more intentinal but

#

mmLol

#

it just sorta plays out fine

#

though it caused me confusion

#

and some issues now knowing it did that

#

@hushed spire because the task switch function says. okay im gonna save the currently running codes state on this stack. swap stacks. and then restore as if i had saved state before

hushed spire Apr 12, 2026, 8:42 PM

#

hmmm, okay Imma just guess I wrote schedule() wrong or something then zerotwoshrug

narrow storm Apr 12, 2026, 8:42 PM

#

when you make a thread. you push things onto the stack so it can "restore" state when its the first time running

#

so you can do it blindly

narrow storm Apr 12, 2026, 8:44 PM

#

hushed spire hmmm, okay Imma just guess I wrote schedule() wrong or something then <:zerotwos...

show me

hushed spire Apr 12, 2026, 8:44 PM

#

void ke::schedule() {
    if (queue.empty()) {
        ke::log(__FILE__, "Can't schedule anything, sadge =(\r\n");
    }
    else {
        LL_NODE<THREAD_CONTROL_BLOCK> *oldThread = currentThread;
        currentThread = currentThread->next;
        contextSwitch(&oldThread->data, &currentThread->data);
    }
}

narrow storm Apr 12, 2026, 8:45 PM

#

and context switch?

#

you prolly want a bit more than that tho

hushed spire Apr 12, 2026, 8:46 PM

#

contextSwitch:
    mov rsp, [rdx+128]

    ; Save current context
    push r15
    push r14
    push r13
    push r12
    push r11
    push r10
    push r9
    push r8
    push rbp
    push rsi
    push rdi
    push rdx
    push rcx
    push rbx
    push rax

    mov [rdx+128], rsp
    mov rsp, [rcx+128]

    pop rax
    pop rbx
    pop rcx
    pop rdx
    pop rdi
    pop rsi
    pop rbp
    pop r8
    pop r9
    pop r10
    pop r11
    pop r12
    pop r13
    pop r14
    pop r15

    ret
``` I don't think I have to push *every* gp registers, but not sure if that matters for correctness

narrow storm Apr 12, 2026, 8:48 PM

#

it would be alot easier if you did a

    uint64_t* old_rsp = &previous_thread->krsp;
    uint64_t  new_rsp = next_thread->krsp;
    thread_switch(&scheduler_spinlock, lock1r, old_rsp, new_rsp);

then its just

thread_switch:
    ; push crap
    ; swap stacks
    mov [rdx], rsp
    mov rsp, rcx
    ; unlock schedular when its safe and enable preemption if lock does so
    call spinlock_unlock_nil
    ; pop same crap
    ret

#

it puts the RSP into the thread strcuture

#

then it just needs to load the new one

#

doing it like this also lets the C/++ code do more of the work

#

for the compiler to do its job and optmise

hushed spire Apr 12, 2026, 8:49 PM

#

fair

narrow storm Apr 12, 2026, 8:50 PM

#

also makes it harder toi fuck up

#

if you do need magic offsets in ASM make sure to annotate them and add static asserts in the high level defintions too

#

so if you change things

#

you get reminded to fix the assemblt

#

trl

#

else you will have bad things happen™

hushed spire Apr 12, 2026, 8:51 PM

#

speaking of which I need to write assert() meme

narrow storm Apr 12, 2026, 8:51 PM

#

thats part of the C standard btw

#

mmLol

#

atleast for static assert

hushed spire Apr 12, 2026, 9:51 PM

#

well I did kind of have to do a hack to get it to work, but poggers

#

now I'm cleaning my code up before doing anything else pepekek

narrow storm Apr 12, 2026, 9:53 PM

#

hushed spire well I did kind of have to do a hack to get it to work, but <:poggers:1337923490...

Did my magic words help?

#

https://tenor.com/view/jfk-clone-high-i-like-your-funny-words-magic-man-jack-black-gif-18659433

Tenor

hushed spire Apr 12, 2026, 9:56 PM

#

narrow storm Did my magic words help?

A little bit? I still don't really know what I'm supposed to do tbh zerotwoshrug

#

ig maybe I should have a main kernel thread?

narrow storm Apr 12, 2026, 10:07 PM

#

not really

#

I don’t use one

#

I spawn the other threads

#

And then they call into kernel code where needed

#

Either via syscalls or if it’s a kernel thread just by calling it

#

You don’t need a “game loop”

#

You just spawn a thread for that sub system

#

Like for an NVME driver that may need to wake up and do work when it gets an interrupt

#

And some drivers are simple enough to just in on an interrupt context without its own thread for its own things

#

(Serial PS/2 somtimes)

hushed spire Apr 12, 2026, 10:09 PM

#

How would I get the ball rolling then? GuraThink because trying to preform a context switch when there's no context to save it doesn't really like that 😅

#

Anyhow, I'll clean the codebase up some and then probably get started on FRED support if I haven't decided I've coded enough for one day thinkong

narrow storm Apr 12, 2026, 10:20 PM

#

hushed spire Anyhow, I'll clean the codebase up some and then probably get started on FRED su...

FRED is life

#

FRED is god

#

FRED is the savior

lunar vortex Apr 12, 2026, 10:44 PM

#

hushed spire How would I get the ball rolling then? <:GuraThink:1387252570510327808> because ...

you should have a context_load function

hushed spire Apr 12, 2026, 10:55 PM

#

lunar vortex you should have a context_load function

Oh, yeah wait I think that's what I'm doing

narrow storm Apr 13, 2026, 12:28 AM

#

lunar vortex you should have a context_load function

that would be pushing things on a new threads stack yeah?

lunar vortex Apr 13, 2026, 12:28 AM

#

no

#

it would just load the register state

narrow storm Apr 13, 2026, 12:29 AM

#

huh

#

a context switch without the other half?

lunar vortex Apr 13, 2026, 12:29 AM

#

yes

narrow storm Apr 13, 2026, 12:30 AM

#

when would you want that?

lunar vortex Apr 13, 2026, 12:31 AM

#

you might

#

for example, my CPUs load into their idle thread

hushed spire Apr 13, 2026, 2:13 AM

#

was playing kirby :p

#

now I'll actually clean up the codebase some

#

anyone who has the time to take a looksie through the codebase and suggest improvments would be appreciated :3

hushed spire Apr 13, 2026, 2:58 PM

#

I'm awake :3

hushed spire Apr 13, 2026, 5:58 PM

#

The agenda for today will probably mainly be FRED

#

I'll also probably work on vmen some more if I have the time

#

Oh, I also think my PFNDB is mapping more pages than it's actually using, so I should fix that.

narrow storm Apr 13, 2026, 6:14 PM

#

hushed spire The agenda for today will probably mainly be FRED

https://tenor.com/view/family-guy-joe-joe-taken-away-gif-17520860855595872089

Tenor

serene lake Apr 13, 2026, 6:16 PM

#

hushed spire The agenda for today will probably mainly be FRED

https://tenor.com/view/homer-simpson-bart-simpson-join-us-sit-here-gif-15631014

Tenor

hushed spire Apr 13, 2026, 6:44 PM

#

pepekek

narrow storm Apr 13, 2026, 10:09 PM

#

@hushed spire

#

So how’s that FRED going

hushed spire Apr 13, 2026, 10:11 PM

#

narrow storm So how’s that FRED going

I was playing smash with my brother pepekek and I have to run by Barnes & Noble rq

#

but when I get back... FRED TIME!!! letsgo

narrow storm Apr 13, 2026, 10:12 PM

#

Yoooooo

#

Remember if you need help

#

Press the triangle key on your keyboard for help

hushed spire Apr 13, 2026, 10:12 PM

#

I'll read the FRED spec in the car trl

narrow storm Apr 13, 2026, 10:12 PM

#

56 pages or so yeah

serene lake Apr 13, 2026, 10:24 PM

#

hushed spire but when I get back... FRED TIME!!! <:letsgo:1282155513156468747>

https://tenor.com/view/fred-flintstones-shock-horror-fred-flintstone-gif-14196352199004985926

Tenor

narrow storm Apr 13, 2026, 10:28 PM

#

serene lake https://tenor.com/view/fred-flintstones-shock-horror-fred-flintstone-gif-1419635...

https://tenor.com/view/assimilate-star-trek-picard-gif-14160230

Tenor

serene lake Apr 13, 2026, 10:36 PM

#

narrow storm https://tenor.com/view/assimilate-star-trek-picard-gif-14160230

narrow storm Apr 13, 2026, 10:36 PM

#

All kernels must FRED

#

https://tenor.com/view/asdfmovie-dotheflop-everybody-do-the-flop-gif-13722398

Tenor

#

(Let’s lay off a bit tho lmfaooo)

hushed spire Apr 14, 2026, 12:44 AM

#

FRED event delivery is always to ring 0 and ERETU returns only to ring 3, it is not 
possible to enter ring 1 or ring 2 while FRED transitions are enabled.``` Lmao, even Intel knows rings 1 and 2 are useless ![pepekek](https://cdn.discordapp.com/emojis/1255025354276470846.webp?size=128 "pepekek")

narrow storm Apr 14, 2026, 12:48 AM

#

Whaaaaat nooooo

#

Honestly they would be less useless if they had more protection things you can apply :x

#

But no they can still access all ram

#

And they barely lock anything off

hushed spire Apr 14, 2026, 1:01 AM

#

heck is a shadow stack? confusedshiki

narrow storm Apr 14, 2026, 1:03 AM

#

hushed spire heck is a shadow stack? <:confusedshiki:990486394294112286>

its a stack that prevents ROP chains

#

its seperate and hidden

#

all call and ret functions also push there

#

and it verifies that and the normal stack match

hushed spire Apr 14, 2026, 1:04 AM

#

narrow storm its a stack that prevents ROP chains

ROP chains?

narrow storm Apr 14, 2026, 1:06 AM

#

https://en.wikipedia.org/wiki/Return-oriented_programming

Return-oriented programming

Return-oriented programming (ROP) is a computer security exploit technique that allows an attacker to execute code in the presence of security defenses such as executable-space protection and code signing.
In this technique, an attacker gains control of the call stack to hijack program control flow and then executes carefully chosen machine inst...

hushed spire Apr 14, 2026, 1:07 AM

#

Ah

hushed spire Apr 14, 2026, 1:53 AM

#

@narrow storm where were those QEMU patches again?

narrow storm Apr 14, 2026, 1:54 AM

#

You will never find them

#

You must use SIMICS

hushed spire Apr 14, 2026, 1:54 AM

#

togescared

narrow storm Apr 14, 2026, 1:56 AM

#

#resources message

#

@hushed spire have fun

hushed spire Apr 14, 2026, 1:57 AM

#

thinkong

#

lets see here...

#

ah wait

At least 16 GB of RAM

#

I might want my beefier laptop

narrow storm Apr 14, 2026, 2:03 AM

#

Yeah it by default dedicated 8gb to the VM

#

lol

#

Also I hope that laptop has an Intel CPU

hushed spire Apr 14, 2026, 2:04 AM

#

both my latitude and MSI have an intel yes lol

narrow storm Apr 14, 2026, 2:04 AM

#

Then you should install thier custom kernel module

#

Which makes it less slow

#

mmLol

hushed spire Apr 14, 2026, 2:05 AM

#

ah, okie

narrow storm Apr 14, 2026, 2:06 AM

#

Me on AMD

#

(It’s 3 times slower than TCG)

#

wahhgone

narrow storm Apr 14, 2026, 2:19 AM

#

narrow storm (It’s 3 times slower than TCG)

Also I’m sure it’s even slower but that’s just what I can eyeball from a few things

hushed spire Apr 14, 2026, 2:29 AM

#

party

narrow storm Apr 14, 2026, 2:30 AM

#

Now how much do you hate SIMICS

#

trl

hushed spire Apr 14, 2026, 2:30 AM

#

I mean it actually wasn't that bad zerotwoshrug

narrow storm Apr 14, 2026, 2:30 AM

#

It’s mostly the docs that are annoying tbf

#

Oh also you know what I picked for my bootstrap allocator?

#

Lazy free list

#

And tracking how much I’ve allocated into it

#

So then later when I do the later stage PMM

#

I can know what memory to not use from the mem map (as it’s bootstrap stuff)

#

And I can reclaim what’s left on the freelist after going through the memory map

hushed spire Apr 14, 2026, 2:32 AM

#

oh cool, I just use a bump allocator tbh pain

narrow storm Apr 14, 2026, 2:32 AM

#

Mines basically that but it supports free so you can use it as primary if desired. Since you shouldn’t be freeing on bootstrap stuff it’s “the same”

#

But it also handles cross memory map regions etc

hushed spire Apr 14, 2026, 2:33 AM

#

I really got to clean up my code pepepray I'm sure there's a bunch of small stuff I haven't implemented pepekek

narrow storm Apr 14, 2026, 2:34 AM

#

We’ll get ready for FREZ

#

FRED

#

Do you wana do it in a hacky way

#

Or a non hacky way

#

mmLol

#

Since the interrupt frames differ on the stack

hushed spire Apr 14, 2026, 2:34 AM

#

non hacky way preferably. I think I do things in a hacky way enough as is lmao

narrow storm Apr 14, 2026, 2:35 AM

#

You may wana use some magic 🪄

#

Which is chopping up the frame

#

Refer to how I did it because it’s nice

#

But basically I split the vector the registers saved and the frame pushed by the CPU

#

And then I pass the arguments through assembly for the IDT seperatly

#

Then in FRED it goes to C code which does the decoding and then sends it off with pointers to the right parts

#

    push r12
    push r13
    push r14
    push r15

    mov rdi, rsp         ; irq_saved_regs_t ; also same as lea rdi, [rsp + 0]
    lea rsi, [rsp + 128] ; irq_cpu_frame_t
    mov rdx, [rsp + 120] ; uint64_t vector

    cld
    sti
    call dispatch_interrupt
    cli

    pop r15
    pop r14
    pop r13

#

For example this for IDT

#

Which id rework first to use this system

#

And to edit your C side structures where possible

#

void dispatch_interrupt(irq_saved_regs_t* regs, irq_cpu_frame_t* frame, uint64_t vector)

hushed spire Apr 14, 2026, 2:39 AM

#

okie :3

narrow storm Apr 14, 2026, 2:40 AM

#

Afaik this is the best way for performance and keeping it sane on both sides so

#

Nobody has corrected me

#

mmLol

#

(Please correct me and show me benchmarks or proof)

hushed spire Apr 14, 2026, 3:15 AM

#

fixed my cpuid function :3

narrow storm Apr 14, 2026, 3:27 AM

#

hushed spire fixed my cpuid function :3

When checking for FRED make sure that the subleaf is supported too :3

#

You need to make sure the leaf and subleaf

serene lake Apr 14, 2026, 6:45 AM

#

simics isnt that slow on my AMD pc

#

not fast, but bearable for testing

#

wouldn't want to do rendering on it

hushed spire Apr 14, 2026, 6:49 AM

#

fair

#

imma head to bed

#

more fred in da morning

#

https://tenor.com/view/kirby-gif-16890639496203212917

Tenor

serene lake Apr 14, 2026, 7:03 AM

#

better bed than fred

#

true

hushed spire Apr 14, 2026, 2:58 PM

#

I have awoken

#

ultrameme

#

After I finish with my school stuff I might work on this some more :3

narrow storm Apr 14, 2026, 4:07 PM

#

serene lake wouldn't want to do rendering on it

Bad Apple runs at like 10FPS or so

serene lake Apr 14, 2026, 5:23 PM

#

narrow storm Bad Apple runs at like 10FPS or so

hmmm, mine is a lot faster than that on simics

#

ok not a lot

narrow storm Apr 14, 2026, 5:24 PM

#

whats your CPU?

serene lake Apr 14, 2026, 5:25 PM

#

maybe 15fps? but that is gif streaming and gzip decompression

narrow storm Apr 14, 2026, 5:25 PM

#

also my bad apple demo runs in CPL3 and uses SSE and other things. and its RLE compressON

#

so context switch overhead etc

#

esp for the mass of syscalls it spams for audio

serene lake Apr 14, 2026, 5:25 PM

#

its not fully utilising it

#

this is my fastest demo on simics

#

this is bad apple

#

bear in mind, its running on an interpterer, too

narrow storm Apr 14, 2026, 6:09 PM

#

Maybe on windows they use hyperv instead of a custom driver

serene lake Apr 14, 2026, 6:18 PM

#

if they did, fred wouldn't work?

narrow storm Apr 14, 2026, 6:19 PM

#

They can still decide to trap and emulate some stuff ig?

#

Idk how hyperv works

serene lake Apr 14, 2026, 6:22 PM

#

badly lol

#

I don't think it's hyperv, qemu runs under hyperv for me and that's blazingly fast

#

but a pain to get working at all

serene lake Apr 14, 2026, 6:30 PM

#

narrow storm esp for the mass of syscalls it spams for audio

does audio work for you in simics?

narrow storm Apr 14, 2026, 6:30 PM

#

Nope

#

But I didn’t try and enable it

#

I only do PC speaker

serene lake Apr 14, 2026, 6:31 PM

#

ah yeah that will cause syscall spam

#

you're sending each note

narrow storm Apr 14, 2026, 6:52 PM

#

and checking for the time etc

hushed spire Apr 14, 2026, 11:58 PM

#

I'm finished with my school stuff, gonna give my other projects some love first tho :3

hushed spire Apr 15, 2026, 2:49 AM

#

okie dokie

#

time for more fred ^w^

#

I uh, need to figure out what I'm actually doing next tho lol

narrow storm Apr 15, 2026, 3:00 AM

#

You can always ask

hushed spire Apr 15, 2026, 3:01 AM

#

okie, what should I do next evalyn KEKW

narrow storm Apr 15, 2026, 3:01 AM

#

FRED

hushed spire Apr 15, 2026, 3:01 AM

#

oh wait

narrow storm Apr 15, 2026, 3:01 AM

#

After that

#

I need a list of your current doings

hushed spire Apr 15, 2026, 3:02 AM

#

I might need rdmsr and wrmsr functions KEKW

narrow storm Apr 15, 2026, 3:02 AM

#

Steal mine

#

trl

#

Not like il break your knees

hushed spire Apr 15, 2026, 3:02 AM

#

narrow storm I need a list of your current doings

I mostly meant about FRED in specific lol, I have a rough idea of what I'm doing after this

narrow storm Apr 15, 2026, 3:02 AM

#

Ahh

hushed spire Apr 15, 2026, 3:03 AM

#

although I will take suggestions lol

narrow storm Apr 15, 2026, 3:03 AM

#

Well first you want to tweak the IDT path to be more normalized

#

Like how I have it

#

So it can work with FRED easier

hushed spire Apr 15, 2026, 3:03 AM

#

normalized? confusedshiki

narrow storm Apr 15, 2026, 3:03 AM

#

Since you want them to both end up calling the same dispatch function

hushed spire Apr 15, 2026, 3:03 AM

#

ah

#

fair

narrow storm Apr 15, 2026, 3:03 AM

#

hushed spire normalized? <:confusedshiki:990486394294112286>

The formats differ

#

But are similar enough

#

So you can get away with splitting it up into 3 things

#

The things you push. The vector and what the CPU pushes

hushed spire Apr 15, 2026, 3:04 AM

#

okie dokie :3

hushed spire Apr 15, 2026, 3:42 AM

#

So a FRED stack frame would look something like this right? GuraThink

narrow storm Apr 15, 2026, 3:53 AM

#

hushed spire So a FRED stack frame would look something like this right? <:GuraThink:13872525...

(i have a C struct if you wana make sure)
but uhh it looks about right yeah. the parts not of event data and the reserved are the same as IDT btw except it always pushes it

hushed spire Apr 15, 2026, 3:54 AM

#

narrow storm (i have a C struct if you wana make sure) but uhh it looks about right yeah. the...

okie figured sigmapepe

narrow storm Apr 15, 2026, 3:54 AM

#

My new kernel is actually a good reference for FRED this time too (well it will be perfect once I get userspace up soonish)

hushed spire Apr 15, 2026, 3:57 AM

#

how come you clear the direction flag here btw?

    cld
    sti
    call dispatch_interrupt
    cli

narrow storm Apr 15, 2026, 3:57 AM

#

It’s required by the systemv ABI

#

And things can be really bad if they aren’t compared to most other flags

hushed spire Apr 15, 2026, 3:58 AM

#

ah

narrow storm Apr 15, 2026, 3:58 AM

#

Eg making string operations run in the wrong direction

hushed spire Apr 15, 2026, 4:30 AM

#

okay I think the interrupt dispatch code should be good now ^w^

#

Kind of late now so I think I'll start on actually implementing fred tomorrow

narrow storm Apr 15, 2026, 4:30 AM

#

hushed spire okay I think the interrupt dispatch code should be good now ^w^

Show the class

hushed spire Apr 15, 2026, 4:31 AM

#

...the class? confusedshiki

narrow storm Apr 15, 2026, 4:31 AM

#

This channel

#

It’s a saying

#

I wana see lol

hushed spire Apr 15, 2026, 4:31 AM

#

ohhhhh

#

sorry I'm tired lol

narrow storm Apr 15, 2026, 4:31 AM

#

Real

#

It’s almost 1am

hushed spire Apr 15, 2026, 4:33 AM

#

I mean it's pretty much the same as yours except for some minor naming differences zerotwoshrug

keInterruptDispatch:
    push r15
    push r14
    push r13
    push r12
    push r11
    push r10
    push r9
    push r8
    push rsi
    push rdi
    push rbp
    push rdx
    push rcx
    push rbx
    push rax

    mov rcx, rsp
    lea rdx, [rsp + 128]
    mov r8, [rsp + 120]

    cld
    sti
    call keInterruptHandler
    cli

    pop rax
    pop rbx
    pop rcx
    pop rdx
    pop rbp
    pop rdi
    pop rsi
    pop r8
    pop r9
    pop r10
    pop r11
    pop r12
    pop r13
    pop r14
    pop r15

    add rsp, 16
    iretq

narrow storm Apr 15, 2026, 4:33 AM

#

hushed spire I mean it's pretty much the same as yours except for some minor naming differenc...

Did you make sure to add static asserts to the offsets in the C structure?

#

So if you ever change it it causes a compile time error

#

(Because then the assembly will break if you don’t fix it)

hushed spire Apr 15, 2026, 4:34 AM

#

ah no, lemme do that :p

narrow storm Apr 15, 2026, 4:34 AM

#

patpat

#

Static asserts are very nice for stuff esp for things that may be compiler/platform specific

#

(And ofc ASM interop)

hushed spire Apr 15, 2026, 4:42 AM

#

aight, there we go :3

narrow storm Apr 15, 2026, 4:43 AM

#

Now go to sleep

#

https://tenor.com/view/cat-gif-3018175339177676319

Tenor

hushed spire Apr 15, 2026, 4:45 AM

#

https://tenor.com/view/kirby-gif-16890639496203212917

Tenor

hushed spire Apr 15, 2026, 5:47 PM

#

#

i'm up :3

#

fredrick time letsgo

narrow storm Apr 15, 2026, 6:39 PM

#

@hushed spire what’s your plan for rdmsr and cpuid abstractions?

#

(You need both)

hushed spire Apr 15, 2026, 6:40 PM

#

narrow storm <@888666019709722664> what’s your plan for rdmsr and cpuid abstractions?

you mean like, arch abstractions?

narrow storm Apr 15, 2026, 6:40 PM

#

Well rdmsr and cpuid is arch specific

#

But you want to abstract it to make it not annoying to use

hushed spire Apr 15, 2026, 6:41 PM

#

ah, uhm

#

I just kind of wrote wrapper functions

narrow storm Apr 15, 2026, 6:41 PM

#

tis what i do. instead of seperate functions for it all

#

then its a cpuid_check(CPUID_HAS_FRED) etc

hushed spire Apr 15, 2026, 6:42 PM

#

ah, yeah mine is slightly less abstracted lmao

narrow storm Apr 15, 2026, 6:42 PM

#

may i see?

hushed spire Apr 15, 2026, 6:44 PM

#

namespace hal {
    namespace x64 {
        static inline CPUID_REGISTERS getCpuid(uint64_t rax, uint64_t rcx) {
            CPUID_REGISTERS cpuidRegs{};

            __asm__ volatile (
                "cpuid" :
                "=a"(cpuidRegs.rax),
                "=b"(cpuidRegs.rbx),
                "=c"(cpuidRegs.rcx),
                "=d"(cpuidRegs.rdx) :
                "a"(rax),
                "c"(rcx)
            );
            return cpuidRegs;
        }
    }
}
``` I also need to add a check for whether or not CPUID is actually supported :p

narrow storm Apr 15, 2026, 6:44 PM

#

All x86-64 CPUs have it

hushed spire Apr 15, 2026, 6:45 PM

#

I mean I would think I don't really need a check in that case GuraThink

narrow storm Apr 15, 2026, 6:45 PM

#

hushed spire ```c++ namespace hal { namespace x64 { static inline CPUID_REGISTERS...

yeah you may wana abstract it extra becase you need alot of extra checks to ensure you dont get a garbage result

narrow storm Apr 15, 2026, 6:45 PM

#

hushed spire I mean I would think I don't really need a check in that case <:GuraThink:138725...

Yes

#

CPUID is on late 486s and all 586s

#

So like

#

Your fine lol

#

CPUID can also tell you your inside a virtual machine and what kind it is

#

And the CPU name and vendor

hushed spire Apr 15, 2026, 6:46 PM

#

coolio sigmapepe

lunar vortex Apr 15, 2026, 7:07 PM

#

hushed spire ```c++ namespace hal { namespace x64 { static inline CPUID_REGISTERS...

you need to check in cpuid if cpuid is supported

#

trl

lunar vortex Apr 15, 2026, 7:09 PM

#

narrow storm tis what i do. instead of seperate functions for it all

what I do is I only do a few cpuid on boot and cache it

narrow storm Apr 15, 2026, 7:10 PM

#

lunar vortex what I do is I only do a few cpuid on boot and cache it

true, but some things you only need to check once in place

lunar vortex Apr 15, 2026, 7:10 PM

#

pub const CpuFeatures = struct {
    x2apic: bool,
    five_level_paging: bool,
    gib_pages: bool,
    tsc_deadline: bool,
    fxsave: bool,
    xsave: bool,
    invariant_tsc: bool,
    nx: bool,
    pcid: bool,
    smap: bool,
    smep: bool,
    pge: bool,
    vendor: CpuVendor,
    family: u8,
    brand_string: [48]u8,
};

#

I fill this

hushed spire Apr 15, 2026, 7:12 PM

#

I was thinking of doing something like that thinkong

void elm Apr 15, 2026, 7:28 PM

#

Same, I just pre parse all features I care about into bitmaps

hushed spire Apr 15, 2026, 7:59 PM

#

@narrow storm what's up with the OR with 3 here? confusedshiki

uint64_t star = ((uint64_t)(0x18 | 3) << 48) | ((uint64_t)0x08 << 32);

narrow storm Apr 15, 2026, 7:59 PM

#

That sets it as a ring3 selector

#

As this is related to the GDT

hushed spire Apr 15, 2026, 8:00 PM

#

ah

narrow storm Apr 15, 2026, 8:00 PM

#

The bottom 2 bits of the segment selector/register set the CPL

#

And if that’s incorrect there

#

Your kernel uhhhh

#

https://tenor.com/view/explosion-mushroom-cloud-atomic-bomb-bomb-boom-gif-4464831

Tenor

Atomic bomb - Explosion

▶ Play video

#

And have fun tracking it down to that

#

And it also depends on your CPU brand if it will or won’t do that too

hushed spire Apr 15, 2026, 8:37 PM

#

oh how come I can't use an or here? :p

"bts $32, %%rax\n"

#

actually wait I may have just made a typo lol

#

but anyhow

#

@narrow storm poggers

narrow storm Apr 15, 2026, 8:45 PM

#

hushed spire <@618537432002068491> <:poggers:1337923490329792552>

You did it

#

Pog

#

You may want a bit more boot logs lol

hushed spire Apr 15, 2026, 8:46 PM

#

I'm sure the code isn't perfect and I'll have to tweak it some, but it works! party

hushed spire Apr 15, 2026, 8:46 PM

#

narrow storm You may want a bit more boot logs lol

pretty sure I do, but they get printed to serial so they don't clog up the screen :p

narrow storm Apr 15, 2026, 8:46 PM

#

Ah

#

I like big logs scrolling on my screen lol

#

But fair

#

Does your serial driver work on real hardware?

#

goodidea

visual walrus Apr 15, 2026, 8:47 PM

#

narrow storm Does your serial driver work on real hardware?

mine prob does

#

never checked XD

hushed spire Apr 15, 2026, 8:47 PM

#

I have no way of testing it at the moment I'm afraid 😔

narrow storm Apr 15, 2026, 8:47 PM

#

hushed spire I have no way of testing it at the moment I'm afraid 😔

(My desktop has a working serial port and I have a cable)

#

(Next time I go and use it want me to test yours too)

hushed spire Apr 15, 2026, 8:48 PM

#

narrow storm (Next time I go and use it want me to test yours too)

sure :3

narrow storm Apr 15, 2026, 8:53 PM

#

Also @hushed spire you may wana trigger a page fault for testing this

#

As you get a non zero error and vector

#

Also use format spesifiers to make your printfs look consistent

#

wahhgone

hushed spire Apr 15, 2026, 8:54 PM

#

narrow storm Also <@888666019709722664> you may wana trigger a page fault for testing this

literally just did as you said that lol works fine :3

narrow storm Apr 15, 2026, 8:54 PM

#

Sick :3

hushed spire Apr 15, 2026, 9:13 PM

#

So, now that I've done that side-quest, I think I'll go back to finishing up my vmem implementation (for the time being), then move on to writing the Slab allocator, and then after that... working on the Object Manager maybe? GuraThink

narrow storm Apr 15, 2026, 9:14 PM

#

hushed spire So, now that I've done that side-quest, I think I'll go back to finishing up my ...

Perhaps

hushed spire Apr 16, 2026, 12:30 AM

#

aight, pretty sure my vmem implementation is complete enough that I can start on the slab allocator :3

#

Think I may just axe object caching and only cache object size for now GuraThink

lunar vortex Apr 16, 2026, 12:44 AM

#

hushed spire aight, pretty sure my vmem implementation is complete *enough* that I can start ...

send vmem implementation

fading talon Apr 16, 2026, 12:44 AM

#

ho sorry by my fault this passed on 4950

lunar vortex Apr 16, 2026, 12:45 AM

#

fading talon ho sorry by my fault this passed on 4950

un français qui fait du zig

#

insane

fading talon Apr 16, 2026, 12:45 AM

#

lunar vortex un français qui fait du zig

https://giphy.com/gifs/george-sigma-shorch-mewing-georg-Nl3TNvZJb6t23J172M

Giphy

ojotas blancas

#

pas le seul et pas l'unique en personne

hushed spire Apr 16, 2026, 12:46 AM

#

lunar vortex send vmem implementation

main thing I gotta figure out is proper coalescing thinkong https://github.com/UtsumiFuyuki/SnowOS/blob/main/yuki/source/mm/vmem.cpp

hushed spire Apr 16, 2026, 12:46 AM

#

fading talon

...is there something I'm not getting here? confusedshiki

lunar vortex Apr 16, 2026, 12:46 AM

#

fading talon pas le seul et pas l'unique en personne

ya un serveur osdev français en passant

lunar vortex Apr 16, 2026, 12:46 AM

#

hushed spire main thing I gotta figure out is proper coalescing <:thinkong:815589724424241152...

I'll check it out and roast you

hushed spire Apr 16, 2026, 12:47 AM

#

pepesad

fading talon Apr 16, 2026, 12:47 AM

#

hushed spire ...is there something I'm not getting here? <:confusedshiki:990486394294112286>

this is a double number, a perfect, like 1111, 2222, 3333, i love view this

hushed spire Apr 16, 2026, 12:47 AM

#

fair lol

fading talon Apr 16, 2026, 12:47 AM

#

lunar vortex ya un serveur osdev français en passant

j'arrive 👀

lunar vortex Apr 16, 2026, 12:47 AM

#

hushed spire <:pepesad:1387252492743737437>

um

#

well

#

there's a lot of stuff

fading talon Apr 16, 2026, 12:48 AM

#

fading talon j'arrive 👀

il est pas ref dans la barre de recherche des serveurs pour ca je l'ai pas trouvé

lunar vortex Apr 16, 2026, 12:48 AM

#

fading talon il est pas ref dans la barre de recherche des serveurs pour ca je l'ai pas trouv...

dm

hushed spire Apr 16, 2026, 12:49 AM

#

lunar vortex there's a lot of stuff

I'm sure there is, I don't have very high faith in my code KEKW

lunar vortex Apr 16, 2026, 12:49 AM

#

also there seems to be NT larp but the code style is inconsistent

hushed spire Apr 16, 2026, 12:51 AM

#

hushed spire if I ever try to change to code-style again, please yell at me <:pepekek:1255025...

#

I did this all manually 💀 so yeah there's bound to be some screw-ups in places lol

lunar vortex Apr 16, 2026, 12:51 AM

#

idk why youd larp the code style tho

#

it's like the worst part of the thing

hushed spire Apr 16, 2026, 12:52 AM

#

lunar vortex idk why youd larp the code style tho

idk, ask me from like, 6 months ago zerotwoshrug

hushed spire Apr 16, 2026, 5:28 AM

#

procrastinated a tad bit sigmapepe

#

I'll work on slab for a lil bit before heading to bed

hushed spire Apr 16, 2026, 9:40 PM

#

I presume the slab can allocate from itself for the bufctls for large slabs

#

also interestingly

#

of all its slabs. The slab list is partially sorted, in
that the empty slabs (all buffers allocated) come
first, followed by the partial slabs (some buffers
allocated, some free), and finally the complete slabs
(all buffers free, refcnt == 0).``` It seems my old slab impl used three seperate freelist instead of one partially sorted one

#

surely having three separate lists would be better though no? GuraThink

hushed spire Apr 17, 2026, 2:58 AM

#

honestly I think imma just clean this up some first

hushed spire Apr 18, 2026, 2:48 PM

#

wasn't really in an osdevving mood yesterday

#

I'll try to get some work done today tho :3

#

Now if only I had enough braincells to figure out what to fix kongoudisgust

hushed spire Apr 19, 2026, 12:18 AM

#

I don't really know where to start so.... cleaning up my page frame allocator it is I suppose?

#

I should also make it SMP-friendly

#

Surely just using a spinlock should be good enough for this shoddy kernel trl

#

I also have 0 clue how to benchmark any of this stuff :p

hushed spire Apr 19, 2026, 2:34 AM

#

gosh fricking dang it

#

here I am wondering "hmmm, why isn't this booting on real hw"

#

I haven't updated Limine on my machine 😭

#SnowOS