saalty's unnamed OS

[still pebble]

Hello, I have recently been putting a lot of work into my OS, so much so that I wanted to do progress reports here. For context, it is 64-bit, was microkernel-based but now is different, has software context switching, virtual memory allocation with de-allocation, userspace, processes (I call each background helper process a service/module), Linux-like logging, queue-based PS/2 (soon to be USB) input system, works fully on real hardware with it's own bootloader, a full graphics library for the OS, and is public domain

[bitter echo]

Always cool to see stuff that runs on real hardware, shame about the closed source nature though.

[still pebble]

Yeah, I will probably change my mind about the closed source decision.

It's not mature enough for me to confidently to say what it will be

tbh, I mainly decided on it because I am afraid my code is pretty ugly

https://github.com/waternine9/untitled-os

alright it's public domain now

normally I wouldn't change my mind so fast, but I was unsure of it being proprietary myself

anyways, lets get to my OS's goals

## Goals for my OS:
# 1. USB
I plan on adding EHCI support soon.
# 2. Games
I am planning on porting some 3d games to it, with my half-assed OpenGL 3.3 implementation.
# 3. Compatibility
I will add PE-64 (is this how you call it?) file loading and ELF64. 32-bit programs will not be able to run in this OS.
# 4. Window manager
I will replicate how Android looks.
# 5. File system
I will try making this OS have no file system :)

So right now I am working on adding a nice temporary WM.

I could also add a mouse service while I am at it.

also, here is a random screenshot of it on real hardware

[still pebble]

Now I have to decide whether I want a terminal or not.

[still pebble]

just experienced this weird bug when attempting to print keyboard input

Jesus fucking christ, this bug is scary debugging-wise

The actual text rendering seems fine

What in the fuck

How does it mess up that beautifully

Also I have to fix logging it seems

[soft solar]

Is your OS using long mode?

or protected mode

[still pebble]

Yup, long mode

[soft solar]

Interesting project, will check out the Github

[still pebble]

Alright!

Do note that some parts of the code may be slightly off.

Just confirmed that the OS still works on real hardware, awesome

Also sorry for my blurry lens

[soft solar]

seems cool

[proud pecan]

nice

[still pebble]

thanks

I will be taking a 1 day break from this project, so see you till then!

[still pebble]

Is it worth it to implement AHCI?

I am thinking of implementing it over using ATA PIO to read

[unborn forge]

ATA PIO doesn't exist anymore on real hardware and even when it did, it was piss slow

i would suggest NVMe as an alternative to AHCI

it's very common now and writing a driver is simpler than for AHCI

[daring citrus]

ahci is still very common, so i would say its worth supporting if you want to cover a wider range of today's hardware

obviously nvme is superior, mainly because of the speeds, but hard drives are still the king for price/size ratio and i dont think sata will go anywhere anytime soon

[unborn forge]

they will

[still pebble]

Alright, I don't have an NVMe drive anywhere, so i guess AHCI is my best option then.

I only have SATA drives

[daring citrus]

i mean nvme drives are really affordable but i still can get an hdd twice the capacity for the same/slightly lower price, ~300 usd for 4tb nvme and ~250 usd for 8tb hdd

gen 3 nvmes are slightly cheaper at around ~250 usd for 4tb

but im still not sure how i feel about solid state drives for archiving purposes

[still pebble]

I personally wouldn't call over 75 dollar drives affordable lol

[daring citrus]

it all comes down to price/gb imo

no point in buying 4 $75 drives when i can buy a $250 one and have 4 time the capacity in a single drive

and the $250 one has a better price/gb ratio, so you get more for less, technically

its just that you pay more to get more, obviously

[still pebble]

I'm just saying, I never needed more than 500 gigs 😄

[daring citrus]

😬

[still pebble]

I am not hoarding data at all, so I have used around 200 gb so far lol

But, I am getting an NVMe drive in a few days

Because it looks like AHCI is quite similar to EHCI, and I do not like EHCI

One quick look at the NVMe osdev wiki page shows that yeah, I will have a lot more fun with NVMe

[daring citrus]

AHCI is pretty simple, idk how EHCI works at all but given its usb i wouldnt expect it to work anything like that

interesting

qemu also knows how to emulate NVMe so that should help 😄

[still pebble]

Yeah, thank god qemu has NVMe emulation

wait, what am I doing, I can start writing the driver right now then 😄

[still pebble]

The structures and registers seem very similar

[still pebble]

https://github.com/codepilot/NVMe-how-to/tree/master/JavaScript

this is pretty funny

Javascript is the last language I would use for osdev

I also sadly caught a cold so development will be slow for a few days

[still pebble]

alright, now I have an NVMe drive

[shut shoal]

Someone somewhere is probably developing an electron based javascript operating system

50% memory usage at idle

[still pebble]

Good idea

Lmao

NVMe fully working on my machine

It's driver time

[still pebble]

https://tenor.com/view/tgwtg-that-guy-with-the-glasses-doug-walker-point-computer-gif-17022655

Thanks, Windows! Now I can make a GPU driver for an RTX card 😎

also, havent done any progress on the OS so far because the cold I caught requires that I take a lot of rest

In a few days I'll start with making an NVMe driver

[still pebble]

ah fuck it, I'll start making it right now. I'll document my progress.

Right now, I am making a brute force PCI scanner for NVMe devices because I am lazy

Alright, NVMe device successfully found on PCI

Now to test if it finds my actual NVMe on real hardware

Phew, I think it found it

Alright, next step is uh

This probably is going to be the least fun step

Busmastering DMA enabled I think

god I am such a noob

But I believe in my ability to learn

[still pebble]

alright, really liking my progress so far, I have calculated the max page size for an NVMe controller

this is not too bad

It always feels nice when specifications start to make sense 😄

Now the controller is reset

[still pebble]

Alright, here is the exciting part, I am at the point where I can enable the controller

Getting really close to having a somewhat capable NVMe driver

well, somewhat close

[still pebble]

Phew, moment of truth, time to send a read sectors command. My I/O queues are set up

[still pebble]

This really sucks, I have to apparently use IOAPIC as well

And apparently I could have to parse AML

This is so sketchy

I am actually angry

But looks like NVMe drives don't even need IOAPIC to be initialized

Thank god

[bitter echo]

You shouldn't need pin based interrupts for nvme

I remember the spec strongly recommends msix (and msi as a fallback)

[still pebble]

Yeah I now use msi-x

But I don't see the opcode for the read command anywhere in the specs

Oh, a little ctrl-f shows it's 0x2

[bitter echo]

ah yes, so for the IO queues you'll want to reference the NVM command set spec

[still pebble]

yup

I cant see where I select the namespace to use for the read

Maybe I need to set it when creating my I/O queues?

[bitter echo]

There's a field called nsid, it's used for most commands so I think it's listed under the generic command structure

it was dword 1 or 2 from memory

[still pebble]

oh yeah, dword 1, alright

[still pebble]

ah crap, I implemented 2.0 spec instead of 1.4

and qemu is 1.4

[still pebble]

also - am I calculating my 64-bit bar0 correctly?

uint64_t bar0 = ((uint64_t)pci_header.BAR0 & 0xFFFFFFF0) | ((uint64_t)pci_header.BAR1 << 32);

I have a feeling that I am not even communicating with the NVMe controller

because read command is not doing anything

[still pebble]

Finally, NVMe Identify command is working!

Only took 17 continuous hours!

I am really proud of myself

Now I have to fix the read command and then I can take a well-deserved break

[still pebble]

How the hell does my OS work on real hardware but not on emulator

The emulator in question is Virtualbox

[misty token]

that can happen quite often

[proud pecan]

yeah mine triple faults

or restarts

I don't know yet

[still pebble]

cli
hlt

is your best friend

[proud pecan]

that's a good point

it's practically a breakpoint

[still pebble]

yup

[still pebble]

I really done just spilled water on my computer

I would've done osdev if now I didn't have to wait for it to dry

And my USB stick that I copy my OS to might be fucked

And my drives... might be fucked too

[livid summit]

i prefer break

[still pebble]

Is it an assembly instruction?

[livid summit]

no its a gdb command

[still pebble]

break is pretty nice, but in cases where you cannot connect to gdb you might have no other choice than cli hlt

[livid summit]

why would you not be able to connect to gdb

[still pebble]

Real hardware I guess

I am not sure whether VirtualBox has gdb support too

[still pebble]

Computer is running fine it seems

[still pebble]

Installing Gentoo, so I will not be osdeving for a day or two

It's been a long break but I'll be back soon

I don't know why I bother to even say that, no one cares lol

[unborn forge]

my condolences

[still pebble]

Alright, gentoo installed, time to migrate my OS to this gentoo installation

[still pebble]

But first some Rocket League

ADHD moment

[still pebble]

I have not given up on this OS, I will start work on the nvme driver again today

Good thing with Linux I can just mount my drive where my OS is stored and just copy it over to my Gentoo installation

Okay now I have it on Gentoo

Great

Now I only have to install Qemu then off to debugging

[still pebble]

I'm back, so the nvme driver still doesn't work

After weeks of debugging

It'll be easy they said

At least I can send admin commands successfully, not IO though

[still pebble]

FINALLY IT WORKS

[worthy jasper]

https://tenor.com/view/lets-go-lets-goo-lest-gooooooooooooooooo-gif-19416648

WOOOOOOOO

[still pebble]

FUCK YEAH

Now making my own file system, which is the logical next step

[sharp glade]

OR a driver for an existing fs design

[still pebble]

No, I want to be unique

It's great fun designing a fs

[soft solar]

how does one even go about designing a filesystem?

[sharp glade]

by knowing what the goals are

and having a toolbox of knowledge about data structures and which ones best apply to disks and stuff

[livid summit]

90% of the time someones "own file system" ends up being fat with a couple minor changes

[still pebble]

Idk, just fuck around I guess

[still pebble]

You're completely right, because implementing things exactly to spec is boring when you can just not

[sharp glade]

yeah my own is just FAT with an inode table

[still pebble]

Alright guys, should I port DWM directly or Xorg first

After the filesystem

[livid summit]

do you have a shell yet

[still pebble]

Yes

[worthy jasper]

just dwm

why adhere to the xorg spec when you can just hack together your own implementation of whatever the fuck you want B)

as long as you understand it, it's all good

[soft solar]

true lmao

[still pebble]

Yeah I'll just port DWM directly

But right now, I have a filesystem which supports directories of any size and any depth

Adding files soon

My brother actually made me a background which will be the default background for this OS

[unborn forge]

that's actually pretty

[still pebble]

Yeah 😄

[unborn forge]

reminds me of windows vista

[still pebble]

Yup, I think instead of embedding the background image in the OS, I will just write some file manager tool for Linux that can add files to my filesystem, like this background

[worthy jasper]

wow, that's really nice!

[still pebble]

😄

[still pebble]

Now I cannot test my OS on real hardware because I don't want to overwrite my very important nvme drive

[still pebble]

File read/writing almost done in my filesystem driver, soon I will make an ls, cat, and a little makeshift nano

Oh, and touch and rm

Mkdir

Wow, already have a ton of possibilities

[still pebble]

Latest progress, getting closer to linux

There is also a recursive remove command which I am proud of

[worthy jasper]

nice work 😎 👍

[still pebble]

Thanks!

Another example

[worthy jasper]

ahh, nice!

just curious, how's the work on your version of nano / a text editor going?

[still pebble]

I haven't done work on it yet, however I plan for it to be finished tomorrow when I wake up 😄

[worthy jasper]

o7

[still pebble]

https://github.com/waternine9/untitled-os

Github updated

[proud pecan]

I have a name: SaltyOS

[still pebble]

Sorry, I've already settled on TheOS (it's already the name in my default background)

[still pebble]

found here

[proud pecan]

https://en.wikipedia.org/wiki/THEOS

TheOS is already taken

[still pebble]

Alright, then The OS

[proud pecan]

Ok then

[still pebble]

I am soon going to make a text editor now which will be pretty exciting, but first cat command

[still pebble]

Okay guys, time to test my OS on bare metal.

I moved my important files out of my NVMe drive.

Page fault!

Uh oh

And when I remove the NVMe initialization it works

Well, not disk io of course, but everything else works

This will be a long night

Of debugging

[worthy jasper]

good luck! o7

happy hacking :)

[still pebble]

😄👍 This definitely will be fun

[still pebble]

Okay, fixed page fault!

Now I am close to getting it working on real hardware

[soft solar]

the window manager is a tui?

[still pebble]

It's GUI using VBE as a backend, it's just the debug console

[soft solar]

ah

[still pebble]

Ah! I think it's not working because I forgot to align my data structures to the NVME controller's page size

Nope

I will create a debug kernel log function for convenience

[still pebble]

It's been like 3 hours of debugging so far, but making amazing progress

[still pebble]

My BAR calculation or finding is not correct it seems

[still pebble]

4 hours

[still pebble]

Got admin commands working on real hardware so far

[worthy jasper]

testing on real hardware is based

[still pebble]

Yeah, I will never be satisfied if my OS only works in emulator

Once my NVMe driver works on real hardware, I will be very, very well set for a ton of features

[worthy jasper]

yeah, i feel the same way. what's the point if you OS doesn't work on your hardware? i understand the usefulness of virtual machines for quick development

[still pebble]

Yeah, virtual machines are nice for when you have the critical stuff out of the way like drivers and are just making some UI or something

[lean dagger]

I just found this thread and it's really cool that you got this far by yourself! Also are you planning to add something so that you can make apps for your os?

[still pebble]

Thanks! I am planning to add shell script support to my OS that I can save to my OS filesystem and run them, and after that I will put my OpenGL 3.3 implementation into the OS to allow 3d apps

[lean dagger]

nice

[still pebble]

The OpenGL implementation I made even has its own shader compiler which I am proud of lol

[lean dagger]

wow

[still pebble]

My nvme driver now works on real hardware!

Ez4saalty

The NVMe driver on real hardware

[worthy jasper]

https://tenor.com/view/lets-go-lets-goo-lest-gooooooooooooooooo-gif-19416648

[soft solar]

nice colors

[still pebble]

Thanks!

[livid summit]

since when was ls recursive

[still pebble]

What do you mean?

[still pebble]

now that I am back, I just now realized you meant the "touch bruh?hello". It would not create a file inside bruh, there was just a bug that made slashes always be question marks

better example

[still pebble]

It feels so nice how my computer flashes lights on every one of my disk reads, it recognizes it fully 😄

Now it automatically adds home, bin, and etc directories

[worthy jasper]

good stuff 😎 👍

[worthy jasper]

ahh, that's a nice touch

[still pebble]

Yup, next step: I made an interpreter for my programming language that should be extremely easy to port

I will make my first programs in my OS!

[worthy jasper]

that's so based

[still pebble]

Thanks, it also comes with a compiler that luckily compiles to 64bit nasm assembly, so maybe I could port an assembler later

[livid summit]

dont you already have some

like ls

etc

[still pebble]

They're built/embedded into the OS, so I would not say they are programs

[soft solar]

I'm so jealous lolz, that's some great progress

[still pebble]

thanks!

I finished porting my language by the way

just doing some finishing touches

And whats even better: I HAVE SPINNING DONUT CODE FOR IT!

With lighting and all that stuff

[worthy jasper]

TORUS GANG

[still pebble]

Man, this port is taking way longer than expected, sorry guys

[still pebble]

Got the interpreter running code, but has a few bugs to fix before I show you guys

[still pebble]

Successfully ported!

the output is the numbers after "Running BF file..."

[worthy jasper]

oh that's neat!

[still pebble]

Spinning donut coming soon!

Also, thanks so much guys for the support of this OS

5.6k loc so far

[bitter echo]

Do you have a github with your current progress in it?

[still pebble]

https://github.com/waternine9/untitled-os here you go, its public domain

[bitter echo]

Ty

[unborn forge]

https://github.com/waternine9/untitled-os/blob/255372fb997792e4ade0a489c5613f40703034a6/boot/boot32.s#L21

[still pebble]

👍

[unborn forge]

i'll let you find the bug

:p

[still pebble]

Oh shit

you mean I should set the sp before ss?

[unborn forge]

no

it's a bit more subtle

[still pebble]

well im lost

😅

havent touched the bootloader in a while

oh bruh

this

global bootentry
bootentry: jmp entry0

db "UNTITLED_OS"

entry0:
    jmp 0:entry

needs to be

global bootentry
bootentry: jmp 0:entry0

db "UNTITLED_OS"

entry0:
    jmp entry

[still pebble]

also this is taken straight out of my local repo as it is right now, so its not too polished

[unborn forge]

the mov to sp needs to be right after the mov to ss

because a mov to ss disables interrupts for the next instruction

either that, or wrap it in cli/sti

but that wastes more valuable bytes

[still pebble]

oh i never knew that, thanks

ill patch it up rq

[still pebble]

Spinning donut working!

It's poorly formatted but whatever

[still pebble]

Can't believe this is my own programming language, OS, kernel, and bootloader. I feel proud of myself

https://github.com/waternine9/untitled-os

also updated the github

Next I have to port dwm and make some terminal emulator

[livid summit]

this looks like C

[worthy jasper]

i think you should be! it's quite the achievement

[still pebble]

Thanks!

[still pebble]

What do you mean?

[livid summit]

you said its your own language

[still pebble]

Yeah, it looks like C but it has differences https://github.com/waternine9/bflang

(porting the C++ was a pain)

Main thing to take away is the difference in casting:

byte:someByte = 120;
int:final = cast(int:someByte);

[still pebble]

However what made it easier is I removed the compiler, as I only want to interpret the code, not compile it to assembly

[livid summit]

https://github.com/waternine9/untitled-os/blob/main/kernel/64/vmalloc.c this is C

[worthy jasper]

i don't think saalty meant to imply that the OS was written in their language, just that they had written an interpreter for their own language

(to be fair, i haven't had a look at the repo)

[still pebble]

I meant that the spinning donut is in my language

[willow path]

oh haha that was one of the first things i did too

[still pebble]

Nice

[still pebble]

Oh I mean I did my spinning donut in my own programming language in the video

[willow path]

is it running in userspace btw?

[still pebble]

Yeah 64bit userspace

[willow path]

nice

i think mine wasn't back then

i don't remember lol, haven't worked on my kernel in a while

well i have written a few lines here and there but i've had other projects

[still pebble]

Yeah, it's the same with me, I've only recently been putting in work on it

[willow path]

like my fantasy computer for example

i don't think i've posted about it here on osdev yet, might do later on when there is some actual software for it

[still pebble]

Nice

[willow path]

currently the only software is a test suite lmao

i wrote an emulator in C++ and i also have a implementation in verilog

verilog one is rather simple, it isn't even pipelined

[still pebble]

What is Verilog?

[willow path]

a hardware description language, you design hardware with it

[still pebble]

Oh alright.

[willow path]

it's different from something like C

althrough verilog does have C-like syntax

i'm actually surprised i managed to even pull this off within a month. I never did anything in verilog before

[still pebble]

Yeah, I've heard of people using it for various hardware projects, I wish I knew about Verilog before designing my breadboard CPU

[willow path]

i mean for a breadboard CPU it wouldn't be that useful i think?

i don't think you can get any easy to build logic out of it

well unless you work with it at the logic gate level

[willow path]

but that takes away basically all the convenience

[still pebble]

It was at logic gate level, just used some ICs for memory and that's it

[willow path]

i mean

i guess verilog could have been useful

still easier than putting logic gates together in some software and on top of that you can run your CPU on any FPGA

[still pebble]

I wouldn't want to use FPGAs as I was, and still am, designing the PCB for it

I don't intend it to be practical, more of a learning experience

[willow path]

i mean nowadays designing a CPU that is practical is generally hard

it's gonna have to be complicated otherwise there will always just be one with better performance and lower power consumption

and if you make a simple one like this it's gonna be incredibly far from anything useful for anything other than having fun

mine is also just for fun

[still pebble]

Welp I don't have to worry about that, I program to attract girls, not practicality

😎

[willow path]

lmao

[still pebble]

Progress update so far is I implemented arrays in the interpreter I ported, and working now on debugging APIC and making the NVMe driver interrupt based

[still pebble]

How hard would it be to port xorg?

[unborn forge]

it's hard but not that hard all things considered

the hardest part is probably sockets

[still pebble]

Hmm, yeah I would need some sort of networking for it

[unborn forge]

i wouldn't call UNIX sockets networking strictly

[still pebble]

Oh you mean that type of sockets

Yeah, I think it'll be alright to implement

First got to get my NVMe driver and language to a stable point though

[livid summit]

i think you want a good working system before you even think about porting something like xorg

you dont have a proper shell yet afaict

[still pebble]

Yeah my system is still premature and the shell was lazily done, but it should be easy to polish things up and get everything stable, the main problem is the NVMe driver right now I would say

[livid summit]

well do you plan on porting bash or creating your own shell

[still pebble]

My own shell

[livid summit]

that will take some time then

assuming it isnt literally just looking up a program name then running it

[still pebble]

Yeah, I don't want my shell to be nearly as complex as bash right now, maybe later I will port bash

[livid summit]

why not as complex as bash?

or maybe not quite as much but typing the name of a program is hardly a shell

you need things like pipes etc

[still pebble]

I will add parameter handling and stuff, I am not too sure yet 😅

[still pebble]

NVMe driver works fully interrupt based now with MSI-X and APIC, previously I was waiting a constant time by reading from port 0x80

I feel really, really good now that I got that out the way

I knew that was insanity 😅

[bitter echo]

Lol, that's a bit cursed

[still pebble]

yeah lmao

[still pebble]

AML is batshit crazy

[still pebble]

Alright, I detected the IOAPIC on real hardware

[still pebble]

taking a 1 day break from the OS

will debug the IOAPIC tomorrow

[still pebble]

I will take another day break as I am going to be dropping out of high school, and I have to get things sorted out.

[sharp glade]

you've probably made up your mind but everyone I know who dropped out of high school regretted it immensely

In almost every case it either stunts you for a few years (which is if you get a GED or get lucky) or dooms you to mediocrity for life

If there's any way you can graduate even a little late you should do that instead even if it swallows up a whole summer or something

[soft solar]

^

listen to hyenasky

[sharp glade]

I almost dropped out at one point and I'm extremely glad that I didn't

[livid summit]

my dad keeps telling me to drop out

[sharp glade]

I incorrectly thought all was lost because even if I graduated my gpa would be so bad that I wouldn't get into even a state university so I thought it didn't matter

I graduated by the skin of my teeth and then found out your HS gpa doesn't even matter because you should probably do your gen eds and lower division major courses at a community college anyway (because it's absurdly cheaper, and they accept anybody, IF you have a high school diploma or equivalent!) and that's all they'll look at when you transfer

[sharp glade]

Where do you live

[livid summit]

scotland

[sharp glade]

Your dad is silly

[livid summit]

he seems to think since i got an internship that i no longer require education

[still pebble]

Yeah, I am aware that it can screw a lot of things up, but sadly I don't have options school-wise, as it has been getting harder and harder to do school at all due to severe OCD and ADHD, and I have reached my breaking point. I don't want to go fully into the details, but all I can say is I wish things wouldn't have turned out this way. Today my dad was actually quite excited, because he was going to buy me a new processor because I was going to do my schoolwork after a year and a half. Few hours later I mentally broke down because of that, so I told him to just not buy it. It's quite sad

[sharp glade]

Well like I said if you live in the US you'll get a second chance at community college no matter how bad you did in HS but only if you get your diploma

College is easier than HS at least in my experience

[still pebble]

I live in Canada

[sharp glade]

It's probably similar

[soft solar]

usually its the opposite

[still pebble]

hey guys, I will most probably not drop out first of all, second of all I will now be accepting any contributions with open hands while I take a few days to work on a side project

https://github.com/waternine9/untitled-os

github is here, I will accept PRs that are beneficial to the project 😄

after the break I will continue debugging my interrupt handling

(note, running it wont show much)

[still pebble]

What resources does the innovative GPU man use for developing lil igpu library

I have decided that breaks are for losers, and will start making an iGPU driver

[granite remnant]

Very nice

[shut pewter]

I highly recommend integrating lil to save you the torture, if you’re sure you want to do it yourself, it’s a tiny bit of intel docs, sprinkled with some Linux code, based on a metric truckload of reverse engineering GOP drivers, fucking around with passthrough for testing, comparing register dumps against Linux, and a bunch of heavy lifting from potentially already existing drm code in managarm and Mesa llvmpipe for the software rendering (eventually mesa hardware acceleration but we’re not there yet)

[still pebble]

alright, I will save myself and port lil instead which I hope will be easier

I will have to port it because there is no haswell option last time I checked

[shut pewter]

Not yet

It’s on the todo list

[still pebble]

alright, I think its most wise for me to just wait

Holy shit, I was using flags or'ed together which were secretly

#define MALLOC_AVAILABLE_BIT 52
#define MALLOC_NO_EXECUTE_BIT 63

This is for page allocation

This might be why APIC is failing

It might not actually disable cache

Nope

[still pebble]

BRUH

Now it works

The redirection table entries, I thought the first entry was for ISA IRQ 0

Turns out its second

god I hope this works on real hw

[still pebble]

ACPI is utter dogshit, for this next step I need to parse AML. I am not going to

Oh wait, AML is not needed

PHEW

At least my OS boots now

Finally keyboard works now too

Alright, this shit is done, next step is debugging my language interpreter

shit is not done, NVMe driver hangs in userspace

gotta love osdev

this is confusing, NVMe driver handles multiple I/O commands just fine at kernel space

On a side note, the OS is now 7k loc, including kernel and bootloader

Close to being my first ever 10k loc project

[sharp glade]

lmk when u get to 100k club...

100k....................

6 figya.....

[still pebble]

That would take too long for some kid who develops at 500 loc per week

I focus on increasing the functional density of code rather than lines

[still pebble]

Going to spend more time now on polishing and modernizing aspects of the OS

[still pebble]

sometimes I feel like this stuff doesnt even interest me and that I am just a masochist

[worthy jasper]

any updates?

[still pebble]

Yeah, sorry I took a 2 week break, I am debugging the NVMe driver right now

My motivation is a sine wave

[soft solar]

same lol

[snow cloud]

if presumably 1 means you're motivated, and 0 means you're not motivated, then what would -1 mean?

you motivate others?

[soft solar]

motivated to not do anything related probably

[still pebble]

this

I just had -1 motivation again, now should be all motivated though

The problem right now is I am using my NVMe driver in my syscall handler, and I cannot wait for the NVMe interrupt in another interrupt

[loud bear]

just suspend the process until its done?

[still pebble]

No, that would suck

And what do you even mean by process? It is done in the kernel

[loud bear]

you were talking abou syscalls, so I assumed youz have a scheduler sorry

Im just trolling

[still pebble]

I need to wait for the NVMe interrupt inside my syscall handler

[loud bear]

1,suspend the currently running process

fuck

[still pebble]

😐

[loud bear]

I accidentally did ctrl+enter instead of shift+enter

anyways

[loud bear]

1, suspend the currently running process
2, go to the next process, if its the only process, go to a dummy waiting thing
3, when the interrupt arrives, do the thing
4, unsuspend the process so it will run next time

[still pebble]

You are completely missing my point, I am waiting for the interrupt synchronously. I am calling my NVMe driver many times in my syscall handler for filesystem tasks.

[loud bear]

what am I missing?

highlight the word

[still pebble]

"process" there is nothing process related in what I am doing

I also call my NVMe driver before any scheduler even starts (and it works in this case), but when I call it in my syscall handler called by processes, it is nested interrupts

[loud bear]

my point is to exit the handler and do smth else in the meantime

[still pebble]

thats basically impossible, as filesystem tasks are not done by a process, its done by the kernel syscall handler

[loud bear]

then look at the Windows source code or smt

h

(Im trolling again)

[still pebble]

I am so bored

[soft solar]

how about creating a separate thread for the nvme driver?

nah that probably wouldn't work

[still pebble]

yeah I dont think theres a good way to do that

[loud bear]

then maybe do AHCI or ATA instead

who needs NVME in 1923 anyways?

see what I did?

warped back 100 years

[soft solar]

i'd probably look at how OpenBSD's nvme driver works, I heard it's (it as in OpenBSD, not the driver itself) is written very well

[still pebble]

I dont have words for how stupid what you just said is, I am not giving up on NVMe. If any other OS can do it, I can too

I will find a way

[loud bear]

how does NVME work on architectures with no interrupt support?

like interrupts has to be an Intel thing right?

[bitter echo]

... What?

[soft solar]

what

[still pebble]

man, if you dont have anything helpful to say, dont say it

[bitter echo]

btw its useful to have some kind of 'wait for event' function, for things like this. So you can wait on the nvme driver to do its thing.
Also you dont need to have interrupts disabled inside your system call handler the whole time, really just any critical stuff you do at the entry - then it can continue on like any other thread. Meaning you can receive other interrupts, like from devices.

welcome back btw

[still pebble]

Oh alright, so I guess I should be able to do a simple sti while waiting for interrupt and then restore the interrupt flag after

[bitter echo]

yeah that'd work, you'd benefit from making it more generic though

a system call is just an entry to the kernel doing something for a thread

[daring citrus]

you should be able to switch from the syscall handler, it should be still executing in the context of a process, it’s just that the stack, cs, as and rip aren’t user ones

[still pebble]

Yeah, I'd really have to redesign things for that though. I will redesign it later though, I just want to do it quick and dirty for now

[bitter echo]

why ?

[still pebble]

so I guess I should make the syscall add itself as a process, and suspend the process that called it until the syscall is complete.
HOLY SHIT! Thank you!

[daring citrus]

i said the same thing right after that :p

[still pebble]

That is the solution

[bitter echo]

oh right 😆

[daring citrus]

nonono

[bitter echo]

gotta be quick around here 😛

[daring citrus]

basically when you enter the syscall you don’t do anything special

you just do kernel stuff instead i guess

ugh i am terrible at explaining

[bitter echo]

even if the hardware does special things like disabling interrupts, you want to return to some kind of normal state - so you can be pre-empted like normal, etc

[daring citrus]

and i’m not sure what you currently do for syscalls

[still pebble]

my syscall is done the old fashioned way, through int 0x80

[bitter echo]

so ideally it would look like:

• user code performs a syscall
• kernel code starts running with whatever the cpu setup for you
• some critical syscall setup happens
• kernel drops back to normal operating state and continues on with syscall
  [ ... normal operation happens ... ]
• at some point later the syscall does what it's thing, and returns to the user code.

how you actually trigger a syscall is an implementation detail

and ideally the critical section at the start (time with interrupts disabled) is very small

[still pebble]

hmm, I think what I will end up doing is the syscall will do everything purely inside it only if it doesnt require interrupts itself. If it does need interrupts, like in this case, then an OS worker process will be added that will do the things it needs to (the process that called this syscall would be suspended during this time)

so the syscall handler itself will exit right after the worker process is added

[bitter echo]

one of the other reasons for wanting to keep interrupts enabled (at least imho) is for timekeeping

I mean that as a general comment, not a critique of your idea 🙂

[still pebble]

yeah, because timers need interrupts to function most of the time

Thanks for all the help man, I will try to get my idea working soon

[bitter echo]

all good, and good luck

[still pebble]

thanks

[livid summit]

why?

[still pebble]

initial FS setup

if thats what you meant

[livid summit]

why not just do it in a thread context

[still pebble]

Why would I do it in a thread context

[still pebble]

I will be accepting suggestions for the name of this OS

I will give some of my ideas, and will give upvote/downvote reactions to them so you can vote

Sadly I dont have any ideas

[still pebble]

I have decided that syscalls will now be processed in their own thread and not inside an interrupt context, to avoid interference with other processes and its safer than nested interrupts

This should be quick to implement

I also think that this is what Windows does

[livid summit]

not all syscalls really need that

something like exit basically consists of moving the process out of the queue and then letting waiting threads know it has exited

[still pebble]

right, I will only put the handler in a thread if it requires interrupts

like for nvme

[livid summit]

yes that makes more sense

[still pebble]

yup

[daring citrus]

why do that

when u can just enable interrupts and let the syscall block

:p

[worthy jasper]

SAALTY'S BACK BAYBEEE LET'S FUCKING GO

[worthy jasper]

here's an easy one, (though not necessarily a good one)

S.O.S Saalty's Operating System

[soft solar]

ILS - i love salt

[still pebble]

wdym?

[still pebble]

hey, doubles as a call for help! Good name!

[worthy jasper]

L.O.S.S. Operating System Schematics

hmgmd not happy with this one, but loss lmfao

[still pebble]

I think I will just do SOS

SOS sounds cool

[worthy jasper]

L.O.S.S. Operating System, Saalty('s)

lmfao

[daring citrus]

why overcomplicate by spawning new threads, blocking on other threads when you can just do all the work in the syscall handler? if you need to wait for a device (like an nvme drive, though that should be abstracted away obviously) you can simply wait, maybe reschedule, if you don't have an event system you can busy loop (bad) and let the process get preempted, and when an nvme interrupt arrives and the interrupt handler does its job you can break out of the busy loop (BAD BAD BAD) in the syscall handler and proceed

though you really shouldn't have to loop and wait for that stuff in the syscall handler... as i said it should be neatly abstracted away behind some device interface, done through the vfs

so all you should do in the syscall handler is call vfs_read or whatever you're trying to do, and that might block - but that's okay, it will wait for an event that will be raised by the nvme completion handler or whatever, and the thread will be unblocked, vfs_read will return back to the syscall handler and return back to userspace

[still pebble]

I dont have a VFS, and I think I will just busy loop as this event system stuff doesnt really interest me. I dont see whats so wrong with a busy loop

[daring citrus]

because that way you don't give up the cpu time to other threads, instead you just waste it waiting for something

of course instead of busy looping you can just yield

idk why i didnt say that in the example

[still pebble]

i have round robin single-processor scheduler, it shouldnt matter

[daring citrus]

ok let me put it that way, every thread gets its own time slice, and if one thread spends its time slice doing nothing, that's bad

of course you can't control what user programs do exactly, but you should avoid doing that in the kernel

that gives other threads more time to run while you wait for an interrupt to occur

[still pebble]

ah whatever, I dont care too much about the performance of this OS, just functionality

[daring citrus]

if a few threads try to read/write to the disk at once it will also become unusable... so

its nothing crazy, you literally reschedule to another process instead of waiting pointlessly...

its not even an optimization it's something you should do 😄

[still pebble]

I will not have multiple disk IO at the same time, it will be purely synchronous

if another process wants to do IO while another process is doing IO, it will be forced to wait through mutex because anything more advanced I am not getting paid enough for lol

[still pebble]

IT WORKS YOU FUCKERS, IT WORKS!

😂

Finally!

Now syscalls get cleanly handled in their own thread and I can finally do disk I/O again

Now I have to get back to debugging the OS's script interpreter

[still pebble]

and disk I/O is much safer now

We should celebrate this somehow, it took me months to get this working

[soft solar]

wine

[still pebble]

I cant lol

but the OS is 69 source code files now as well

and a total of 7k loc

[still pebble]

If I put my OpenGL implementation in the OS, it will turn into a mean 10k loc

[worthy jasper]

LETS GOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

WOOOOOOOOOOOOOOOO

🎉 🥳 🎉

[still pebble]

Now thinking of whether its worth it to implement SMP... I have it implemented in my old OS, but adapting it to my current OS would be a pain...

I am just REALLY glad I got interrupts for NVMe fully working

[still pebble]

The rust crab irritates the shit out of me, just saying

[bitter echo]

if you're going to do SMP, do it from the start

adding it later on can be painful

[hollow lintel]

^^^ I agree with that

when implementing SMP you will have to rewrite lots of code just because either it was thread unsafe, either it wouldn't apply to a smp system

designing the os from ground up with smp in mind will result in more reliable and solid code

trust me, I've gone through a no-smp to a smp transition and was a pain to do so

[still pebble]

yeah, I guess its staying single core, I am way past the point of the code being flexible enough to allow SMP

[hollow lintel]

I did that transition when I had a full userspace with drivers working

[still pebble]

its just too much of a pain

[still pebble]

I will get this OS on real hardware again, only thing I need to fix is the APIC/IOAPIC initialization

Everything works on real hardware other than APIC/IOAPIC initialization

[still pebble]

I am going to reverse engineer my motherboard's firmware's GOP implementation to hopefully get started on iGPU driver for this OS

[livid summit]

seems a bit strange to me to do that now but 👍

[hollow lintel]

isn't the bios encrypted in some way or at least obsfucated? ik that apple does that

firmware**

[still pebble]

my motherboard's one isnt encrypted or obfuscated

I am getting logical and reasonable assembly code when disassembling it

[hollow lintel]

then I wish you good luck!

[still pebble]

thanks!

I will get to it in 1 hour

if I dont then someone please remind me with ping

[still pebble]

AAA instruction perfectly describes how I am feeling right now

AAAAAAAAAAAAAAAAAAAA

[still pebble]

im going to take break from re and write a POSIX compat layer

[still pebble]

will take 1 week break from this OS to write an x86_64 PC emulator

[still pebble]

I think I am burnt out heavily, so I will take a huge break until I feel motivated enough again to work on this OS. Actually in december, I will gain a development team member (who is my friend), so will probably come back in december

Really sorry for all these interruptions in development, I fully intend to keep developing this OS, I just need to get out of burnout

[worthy jasper]

that's okay bud, i'm here for you (probably won't be able to help much cuz i'm very inexperienced), at least as moral support. enjoy your break :)

[still pebble]

Thanks! I will try to enjoy it lol

[loud bear]

burnt out? mee too lmao

well

no need to repeat what mangomeep said

[still pebble]

burnout gang

also nice nickname

[loud bear]

now its even better

btw ur a walking salt shaker that ovenengineers everything

[still pebble]

Alpha Male Certified Nickname

yeah

[still pebble]

I feel motivated!

I am going to make a GLSL compiler for this OS

[worthy jasper]

:0!

[still pebble]

Making good progress, and this is actually enjoyable to develop

#ifndef FREESL_H
#define FREESL_H

#ifdef FREESL_FREESTANDING

#ifdef FREESL_MALLOC_IMPL
#define malloc(_Size) FREESL_MALLOC_IMPL(_Size)
#else
#error "freesl: Please provide a `malloc` implementation defined as FREESL_MALLOC_IMPL"
#endif

#ifdef FREESL_FREE_IMPL
#define free(_Block) FREESL_FREE_IMPL(_Block)
#else
#error "freesl: Please provide a `free` implementation defined as FREESL_FREE_IMPL"
#endif


#else

#include <stdlib.h>

#endif // FREESL_FREESTANDING

#include <stdbool.h>
#include <stdint.h>
#include <stddef.h>

typedef enum 
{
    FSL_TOK_ASSIGN,
    FSL_TOK_ADD,
    FSL_TOK_SUB,
    FSL_TOK_MUL,
    FSL_TOK_DIV,
    FSL_TOK_VAR,
    FSL_TOK_CONST,
    FSL_TOK_SWIZZLE,
    FSL_TOK_CAST,
    FSL_TOK_CALL,
    FSL_TOK_MEMBER_ACCESS
} FreeSL_TokenType;

typedef enum 
{
    FSL_TYPE_INT,
    FSL_TYPE_FLOAT,
    FSL_TYPE_VEC2,
    FSL_TYPE_VEC3,
    FSL_TYPE_VEC4,
    FSL_TYPE_MAT2,
    FSL_TYPE_MAT3,
    FSL_TYPE_MAT4,
    FSL_TYPE_INT,
    FSL_TYPE_CUSTOM
} FreeSL_TypeOrigin;

typedef struct _FreeSL_Type
{
    FreeSL_TypeOrigin origin;
    size_t size;

    // Only if origin is custom
    struct _FreeSL_Type* members;
    size_t member_count;
} FreeSL_Type;

typedef struct
{
    char* data;
    size_t size;
} FreeSL_String;

typedef struct
{
    FreeSL_String name;
    FreeSL_Type type;
    void* data;
    bool is_array;

    // Only if is array
    size_t count;
} FreeSL_Var;

typedef struct
{
    FreeSL_TypeOrigin type; // Not FreeSL_Type because constants can only be builtin types
    void* data;
} FreeSL_Const;

typedef struct
{
    FreeSL_TokenType type;
    
    // Only if is var
    FreeSL_Var* var;

    // Only if is const
    FreeSL_Const constant;
} FreeSL_Token;

#endif // FREESL_H

what I have so far

It will support structs too

version 330 will hopefully be mostly supported

At first, it will just directly execute the AST tree for simplicity, but then I will make it compile into my own bytecode

[still pebble]

its going good so far

[still pebble]

about 50% done the core C interpreter for it that I will build off of

working blazing fast after burnout lol

proof of progress

[still pebble]

so I can generate AST for C code now

maybe could make two branches of it, one for GLSL and the other one for plain C

people be porting some C compiler while I make a custom one 😎

I will share progress soon

[still pebble]

Boom!

[worthy jasper]

well, if you've written the code already...

[still pebble]

GLSL parser is done! Very very close to having modern OpenGL support in my OS!

[still pebble]

what do you mean?

[worthy jasper]

for this ^

if you've already written some of the code for a c compiler, you might as well write a custom one

[still pebble]

even if I had no code for it yet, I would do it because it hurts to port such massive compilers, plus its extra cool kid points to do it all myself

[worthy jasper]

fuck yeah B)

[loud bear]

opengl without accelarated graphics? are you on crack?

[still pebble]

what do you mean "are you on crack", the performance is not modern but the feature set is

I dont plan to fucking run cyberpunk lmao

Maybe like 10-poly quick demos

[waxen zenith]

u really wrote ur own compiler damn

are the lexers and parsers done?

[unborn forge]

no, you're just ignorant

[still pebble]

the lexers/parsers are fully done, except one tiny bug which I will fix today

best thing about it, is its for an existing language (glsl) so I can run existing shaders

[waxen zenith]

have you made any other programming languages?

[still pebble]

yes I made 5 others

in increasing complexity

this is by far my most complex parser I have ever made

[waxen zenith]

im workingg on a language on the side myself parsers are not easy lol

[still pebble]

yeah, its one of the harder things I learned to do, but the key is, is abstracting away the operations on the raw source code as much as possible

[misty token]

I mean, why not? Software implementations do exist.

Ofc it won't be able to run the latest games in real time, but often that's not the only reason one would do that.

[shut pewter]

Managarm does it too. Your point?

[shut pewter]

Llvm to the rescue

[still pebble]

too massive

[soft solar]

You are just straight up calling him ignorant, that's just rude,especially without saying why you did that

[soft solar]

You just said "why not?" And then brought in the facts. No "you are ignorant" or "your point?"

699th message

[soft solar]

That "your point?" is a bit rude

Don't be assholes

Look at his BIO, he is very depressed right now

My point? It is that you should be like lukflug

[still pebble]

Considering that what lolguy said was rude too, I dont get your point here

Depression is not an excuse to be rude as well

plus I dont think calling someone ignorant is very rude, just informs them to look more deeply into what they are saying I guess

Enough of this though, back to my compiler

[bitter echo]

https://tenor.com/view/spider-man-we-one-gif-18212100

What's the current progress of your opengl implementation?

[still pebble]

Parsed this huge ShaderToy shader into an AST tree (no codegen yet).
I got a preprocessor, structs, and much more. (this is the GLSL compiler for it)

no actual openGL functions implemented yet, just decided to start with hardest thing first

also just implemented function overloading

[misty token]

Stop referring to yourself in the third person, you're not fooling anyone.
And people were rude because you yourself were rude with your comment.

[hollow lintel]

the fact that opengl is usually implemented in hardware doesn't make it hardware-only

[still pebble]

thats the beauty of it

[hollow lintel]

yeah

plus it can be efficient in software if avx or any kinf of simd is used

basically opengl works with lots of number and matrix operations and those can be accelerated using said simd

I don't find software opengl that bad

I really want to see your implementation run half life 1

[exotic lintel]

I wouldn't have treated the "are you on crack?" as an insult

[soft solar]

tbh I never knew OpenGL wasn't hardware-only

[misty token]

then "your point?" isn't either

[daring citrus]

hm i never knew x86 cpu's weren't hardware only

:)

what about arm cpus? they surely must be hardware only right?

enough mocking around, anything can be implemented in software

[soft solar]

didn't think of it this way :p

[misty token]

except touching grass

[soft solar]

touching grass is only hardware sadly

[still pebble]

Finally I am back

I got the OS working fully on real hardware again just recently, this time with the NVMe driver being interrupt based

Also happy new year guys!

[bitter echo]

Welcome back

Happy new year 🎉

[misty token]

welcome back!

[worthy jasper]

merry new year & welcome back saalty!

[worthy jasper]

i had a look at the github and it doesn't seem like it's been updated in a while, but i'm interesting in having a look at the latest version. i'd like to see if i could get it running on my hardware lol

[still pebble]

Thanks guys!

[still pebble]

I pushed the latest version just now for you 😄

It will use the first NVMe drive it finds as the filesystem, so make sure you have at least 1 NVMe drive in your computer and make sure you are not using any of them for anything else

You can use this disk image I compiled

[proud pecan]

What if said drive had multiple partitions?

[still pebble]

partition table just gets overwritten probably, it doesnt care about partitions

[proud pecan]

That sounds like a future issue but you do you

[still pebble]

probably not going to be an issue as I am not using my NVMe drive for anything else

[still pebble]

also the kernel now uses APIC and IOAPIC for interrupts rather than the old PIC

[worthy jasper]

well... that's gonna be an issue for me, seeing as i've only got a laptop with a builtin nvme drive... with all of my stuff on it

i'll check it out with qemu tho :3

[worthy jasper]

well i cloned the repo and ran make

/usr/bin/ld: /tmp/ccnTGF6I.o: in function `InitTable':
vmalloc32.c:(.text+0x3ba): undefined reference to `__stack_chk_fail'

[still pebble]

yeah, you shouldnt run it then

(on your laptop)

[still pebble]

that is really weird, it works for me

[unborn forge]

-fno-stack-protector -fno-stack-check

those flags exist in the c-template for a reason...

alternatively, you'll have to implement the stack protector runtime, which evidently it's not in this case

[proud pecan]

The stack protector runtime is only like a function and a variable

Pretty useful sometimes

[still pebble]

didnt know about these flags, should be fixed now

[still pebble]

every OS needs some entertainment, so I am adding a gambling system

it will be a CS-like case opener

[worthy jasper]

well

it draws a white line at the top of the screen, then it quits qemu

[still pebble]

qemu-system-x86_64 -hda bin/boot.img -drive file=bin/storage.disk,format=raw,if=none,id=nvm -device nvme,serial=deadbeef,id=nvme-ctrl-0 -device nvme-ns,drive=nvm,nsid=1 -no-reboot -m 4G -smp 1
make sure you are doing this command

[soft solar]

lmao

[granite remnant]

lmao what

if you have bad luck it deletes itself

[soft solar]

suicide linux vibes

[still pebble]

if you get bankrupt then you cannot play the game ever again unless you cheat and put a new balance inside the /sys/gamble.conf file

[still pebble]

Its gambling time

Next I will make an art program for it

[proud pecan]

Nice!

[worthy jasper]

how big does the storage.disk file have to be btw? i just used dd to fill up 5GB

[still pebble]

any size

but 128 MiB is recommended

[worthy jasper]

oh okay lol

[worthy jasper]

well the command in the makefile or the .bat file doesn't have "-smp 1", but it still crashes the same way. it gets to draw a white line up the top, then it closes qemu

i don't know much about qemu but i believe there's some debugging built into it? again, i don't know much, but if somebody taught me i'd be happy to provide whatever info is needed

[bitter echo]

so have you decided on a name for this os yet? 😛

[worthy jasper]

well it's either TheOS or SOS, afaik

[still pebble]

hmm, I will try to find whats causing this later, for now I am going to start on the network stack

[still pebble]

The name should be uhhhhhhhhhh

god I am indecisive

[bitter echo]

uhhhhhhhh

[still pebble]

Wait thats perfect

Alright, uhhhhhhhh is the name of the OS

Its official

saalty's named OS (uhhhhhhhhOS)

[worthy jasper]

u8h

[still pebble]

u8h doesnt seem as funi

[misty token]

yay! finally got a name

[still pebble]

yup!

[still pebble]

Actually instead of a network stack, I'll implement xHCI for USB mouse and keyboard

More urgent

xHCI docs downloaded, time to grab some popcorn.

[soft solar]

grab some tissues, you'll need them when you break down while reading the docs

[still pebble]

😢

[still pebble]

xHCI driver going good, getting nice progress.

Some of it is a lot like NVMe

Which is good

[still pebble]

XHCI driver is going amazing so far, 90% of the way to having the xHCI controller fully initialized!

[bitter echo]

Wow cool

[still pebble]

typedef struct
{
    uint32_t RunStop : 1;
    uint32_t HCReset : 1;
    uint32_t IntEnable : 1;
    uint32_t HSErrorEnable : 1;
    uint32_t Reserved0 : 3;
    uint32_t LHCReset : 1;
    uint32_t CSS : 1;
    uint32_t CRS : 1;
    uint32_t EWE : 1;
    uint32_t EU3Stop : 1;
    uint32_t Reserved1 : 1;
    uint32_t CMEnable : 1;
    uint32_t ETEnable : 1;
    uint32_t TSCEnable : 1;
    uint32_t VTIOEnable : 1;
    uint32_t Reserved2 : 15;
} USBCMD;

Thank god for this syntax man, this has saved me so much time, I love it, I really love it

I think they are called bit fields

[worthy jasper]

what is that syntax? / what does it do? i've never seen it before

[still pebble]

Its where you can specify the size of an integer struct member in bits. So when you access that member, the value of it will be zeroed out after the size in bits (but the next member will still have its value intact). Note that the sign bit is also at bit (size - 1), if the struct member is signed.

[bitter echo]

do note that bitfields can have weird behaviour

[soft solar]

why? and what exactly?

[bitter echo]

You'll still be loading and storing bytes/dwords/etc as thats how cpus work. If you're dealing with mmio this could lead to side effects.

[granite remnant]

wouldnt you need to do this anyway?

[still pebble]

Wanted to take a moment to thank everyone here who has supported me so far, I am truly grateful. OS development is changing my life in good ways, its just a field like no other I have ever done, and I credit my progress to you guys who have been sticking with me in this journey. Thank you a thousand times ❤️

[still pebble]

hey guys, theres not any progress update yet because I am being hit by a wave of procrastination, will try to work more on this now though

sorry about this 😢

[versed pendant]

No problem!

[worthy jasper]

i don't know why but i assumed this was just C, is it C or C++? (or something else entirely?)

[still pebble]

its C 😄 everything in the OS is C or assembly

[still pebble]

I love C, C is great
Loving C is my fate
I love that char size is always eight
I would go to a different state
Just so that me and C could go on a date

As you can tell, I am tired
I am restless, my mood has expired
I dont know why I am rhyming
At least its for something that is loving

[still pebble]

i need an emotional support buddy

[worthy jasper]

aw buddy

[still pebble]

My friend is also now working on this OS, and the XHCI driver is now in the debugging stage

[still pebble]

https://github.com/waternine9/untitled-os

This is the project link by the way

https://github.com/waternine9/untitled-os/blob/master/kernel/64/drivers/usb/xhci.c

XHCI driver is here

or, at least the progress on it 😄

[worthy jasper]

it was never joever,

we're joe back

[still pebble]

Indeed... 😅?

[worthy jasper]

i'm sorry i'm suffer from extreme mental degradation due to being online

a small translation: welcome back :)

[still pebble]

thanks!
and me too, i am just good at keeping my degradation away 😅

[nocturne shale]

we got xhci reset working! we didn't do aligned reads, and the only way we found out that was the issue is by reading source code from qemu

[still pebble]

^ ske is the one contributing as well

we did a lot of progress today

The next thing we will likely work on is getting HPET working for a precise sleep function for drivers

(because we need sleep function for xhci)

[nocturne shale]

the ultimate goal is to implement the mouse

[still pebble]

yeah

we could stick to ps/2 keyboard, but its probably best to do usb keyboard too

[nocturne shale]

the question is

what is the embedded keyboard on real hardware

is it also usb?

[still pebble]

i have never heard of an embedded keyboard before

[slim edge]

laptop keyboards?

[nocturne shale]

i mean like, laptop keybard

[slim edge]

most are ps/2

[still pebble]

ok then we can probably use ps/2 as a fallback if there is no usb keyboard

[stray latch]

does the bootloader use UEFI or BIOS loading since on all intel processors from 2020 and above do not support BIOS anymore.

[unborn forge]

that is incorrect btw

intel processors after 2020 can totally support BIOS with an external GPU that supports the legacy video modes needed

[still pebble]

the bootloader is solely BIOS-based, but this is not really a concern as my CPU will probably not be upgraded for years

[nocturne shale]

EFI will probably come.. at some point

while it is more fun to develop the BIOS version we'll do that

i think at least that part of osdev is having that enjoyment of doing it yourself

[unborn forge]

sure whatever you say

[nocturne shale]

i mean of course if you done it many times it's not that interesting, probably will be same for me eventually

it's a hobby because it's interesting

[unborn forge]

bootloader dev isn't really osdev and it easily detracts from the actual osdev experience if you get distracted

it's also not a super trivial thing to get right, especially if running on real hardware is a concern

[still pebble]

Hello everyone, its been long since we last posted but right now I am finally working on this OS again, so right now I am making a driver structure system

I am making a generalized storage driver structure right now, so that we can abstract away the type of storage devices and just have, y'know, storage devices

There will be commands to find what driver a storage device is using, for example

| storage0 - nvme
| storage1 - usb
| storage2 - usb
| storage3 - sata

On the XHCI side of things, we'll probably have to stop using bitfields which will take some time to rewrite all the structure members

[bitter echo]

Cool to see you're still going

[still pebble]

yeah, this must be my only long-term project lol

[still pebble]

the goals have definitely changed lol

[worthy jasper]

well what are they now?

[still pebble]

## Goals for my OS:
# 1. USB
Currently implementing xHCI.
# 2. Games
I am planning on porting some 3d games to it, with my half-assed OpenGL 3.3 implementation. And will add my GLSL parser/soon-to-be-interpreter to it too
# 3. Compatibility
Not sure about being compatible with existing applications, will probably just make my own format
# 4. Window manager
I will replicate how Android looks, but will add features from PC operating systems, like Windows
# 5. File system
My own custom file system will be implemented (done)
# 6. Driver management
Create a way to generalize/abstract drivers to make the OS cleaner and more modern

@worthy jasper

[still pebble]

Oh, and I will make drivers get their IRQ dynamically allocated!

After this will also have to finally make a kernel panic function...

[still pebble]

https://github.com/waternine9/untitled-os

Pushed my latest progress

Coming together very nicely

[still pebble]

Added panic

Finally

[still pebble]

It was thought to be impossible

[still pebble]

Huge things are coming to this OS!

[still pebble]

Major update!!!!
The changelog:

- Added a driver manager which can abstract storage devices from their drivers and automatically get their format detected with any of the available Filesystem drivers
    - The driver manager accepts StorageDrivers and FilesystemDrivers for now, but once XHCI/USB is finished there will be mouse and keyboard drivers
- Deprecated the IDE/ATA driver
- The codebase got a big cleanup!
    - Notably, idt.c is much much cleaner now and doesnt contain the kernel's drawing functions anymore
    - The kernel's drawing functions got moved to draw.c/draw.h in the same directory
- The drivers now no longer have their designated IRQ line hardcoded; the driver manager passes the next available IRQ line to a driver if the driver requests IRQ support, which the driver can then hook up to its own IRQ handler
- Added a panic function with its own string formatter, also displays the file which caused the panic

https://github.com/waternine9/untitled-os/commit/3ac5510bf9ee7956b90d07a0e04669bd9c76dc5a

@nocturne shale helped a lot too, notably the string formatter (which I suck at making lol)

Next thing to do right now is rewrite the XHCI driver to not use bitfields, and make commands to list devices and their respective drivers and more file ops

[still pebble]

https://codeberg.org/thesalt/UnnamedOS

The OS is moved to codeberg because I lost my github account due to its mandatory 2FA

[worthy jasper]

o7

[still pebble]

saalty's unnamed OS

Decided to also scrap that name and keep this OS unnamed, until I get a great name idea which might be never lmao

[worthy jasper]

definitely fair tbh

[still pebble]

yeah i think there is beauty in anonymity, so its probably going to stay this way

[kind oyster]

Yeah I think so too! beauty in anonymity

[exotic lintel]

LOL

why not just... get 2FA?

contrary to popular belief you don't need a phone number for 2FA; you can use an authenticator app/device

I don't exactly care that you aren't using github anymore but that is a stupid reason to give it up

[still pebble]

thing is, i tried using an authenticator app. And that backfired dramatically when I uninstalled windows and switched to linux, since windows had the original installation with its secret code that i must have in order to login to other devices.

I just want to be able to login painlessly

[atomic juniper]

that's why you keep the backup codes somewhere

[worthy jasper]

eh github stinki and owned by microsoft

[still pebble]

Yeah I should've stored them in the cloud, but its whatever. Github is starting to suck anyway, and 2fa is annoying in general.

[still pebble]

Been procrastinating recently, like a lot, but this project is far from abandoned for anyone interested in its development.

I still have plans for major features (like OpenGL 3.3 support), and while I cannot predict how long it will take, I can say its coming.

But, first thing to finish right now is USB/xHCI support, which will provide much better support for HID than the PS/2 emulation from the firmware that the OS is reliant on right now.

For example, PS/2 mouse doesnt work on my hardware, which could be because my PS/2 driver is flawed, or because the firmware is flawed. I dont want to spend a lot of time debugging just to find out its the firmware's issue.

Figure 1, graph of my motivation. (x axis being time, y axis being motivation)

[versed pendant]

Yeah, I feel this.

[atomic juniper]

Same here, but this would be more accurate

[still pebble]

oh man, april fools even reaching this server's "serious" announcements

Limine will now be a GRUB fork, in order to benefit from a lightweight I knew it was a joke after reading "lightweight"

but, april fools aside, me and ske will be developing a windows compatibility layer for this OS. (meaning xhci will have to wait for now)

The goal is to support windows 10 notepad

[pastel onyx]

Umm

Have you ever seen || ?

all these if statements make me want to delete myself

[still pebble]

where in particular are you talking about?

[still pebble]

drawman.c is probably what you are talking about
I was quite indeed aware of logical or, but I personally think seperating each of those conditions looks more readable

[pastel onyx]

All files

[still pebble]

there are only 2 other files that have if statements than can reduce to ||, and there is not much cases of them

for the other files i might reduce them, but the cases in drawman.c are intentional

[stray latch]

A windows 10 compatibility layer well all I can say is best of luck man hope you can somehow survive that programming hell since that means you have to support ... windows.h dun dun dun no but for real I feel bad for you for having to remake windows.h.

[exotic lintel]

remaking the early versions of the API is not that bad

COM and OLE are where it's at

[tight prawn]

He doesn't have to remake windows.h

Programs for windows are already compiled

[livid summit]

i think windows.h is the least of your problems if you want compatibility with windows anyway

[tight prawn]

yeah

writing ntdll is your problem

[stray latch]

Well they said they wanted windows 10 compatibility which means they need to be able to reverse every single bit of windows that notepad relays on. And how would they even do that like would they make apps run on the OS itself or on a wine like app.

[tight prawn]

Notepad probably encompasses a lot of stuff

[exotic lintel]

What else do you think

[misty token]

port Wine

[livid summit]

actually writing all of those functions

[exotic lintel]

[still pebble]

isnt it just a bunch of syscalls, dll support, and making some system dlls and making a pe file loader

and I wont have to write windows.h because I will not support compilation for windows programs

[snow cloud]

To be honest I have no idea about how stuff like that works but it seems like if it were that simple we would see more people implementing Windows apps' support. I could be wrong though

[tight prawn]

It is

Its just that... Windows has quite a lot

https://hfiref0x.github.io/NT10_syscalls.html

[stray latch]

Oh I thought you wanted to be able to like build and run windows apps

[snow cloud]

I really don't want to sound discouraging and pessimistic and I'm sorry from the bottom of my heart if I do but I mean there's the ReactOS project which has lots of contributors and they've been trying to get Windows stuff to run for years (edit: it does run but it's far from perfect). It's a very ambitious goal what you're talking about

But if you're up for a challenge and don't necessarily aim for something perfect, then that's a fun project indeed, best of luck!

[exotic lintel]

Basic Windows stuff probably isn't that difficult. You could run most 1985-1993 Windows apps within a few months

[snow cloud]

Yeah, sounds fair enough

[exotic lintel]

But after win95 things take a complication turn

[still pebble]

thing is, reactos aims to support much much more than notepad

like browsers

[snow cloud]

yeah I get what you're saying

[still pebble]

there is a guy who made windows xp notepad work in 30k loc of just assembly

http://board.kolibrios.org/viewtopic.php?t=5490

so this shows me that its not toooo ambitious

but I am attempting to do win10 which will introduce many complications

but then again, you dont do osdev to do things the easy way.

[sturdy swallow]

the thing is that the windows usermode dlls have a ton of functions

[still pebble]

I'm thinking I can look at the symbol table of notepad's PE file and just implement the symbols it uses

[sturdy swallow]

yeah you can do that, but its still likely a good amount

also one recommendation: don't put nulls into the import address table entries because that makes it very painful to see which function it called, I did that mistake

[tight prawn]

its definitely doable dont listen to what anyone says

just go for it

[stray latch]

I mean you could target windows 1.0 notepad instead

it would be faster but not as good

faster to make

and run

[woeful rover]

i think not

[stray latch]

I forgot /s

[woeful rover]

i hope so

it would be quite tedious to support running win16 apps

[stray latch]

True xp on the other hand

[proud pecan]

You'd probably have to support both 32-bit and 64-bit DLLs to support modern windows programs

As iirc, some applications use both 32-bit and 64-bit DLLs

Nvm

I forgot that that would be practically impossible

[slim edge]

a good chunk of windows programs are 32-bit still though

even ones shipped with windows afaik

[tight prawn]

Steam

[still pebble]

got it!

[still pebble]

exactly

[still pebble]

no

This is what I think is the symbol table of Notepad, which doesnt seem like tooooo many things

well, its quite a lot actually

[proud pecan]

every heard of: objdump

[still pebble]

i dont have it

[proud pecan]

No WSL?

[still pebble]

i have wsl but i am lazy

[proud pecan]

You'd really rather see that garbled mess than type two commands

[still pebble]

inspecting notepad in notepad is what god intended

[proud pecan]

Where even is notepad

located

(filepath)

[still pebble]

windows/system32/notepad

[proud pecan]

objdump -tT /mnt/c/Windows/System32/notepad.exe```

wait it doesn't work

[slim edge]

-x should work iirc

[proud pecan]

Is notepad stripped?

[still pebble]

its divine intervention preventing you from doing such acts, you must use Notepad

[proud pecan]

llvm-readobj --coff-imports /mnt/c/Windows/System32/notepad.exe```

[slim edge]

$ objdump -x  ./.var/app/com.valvesoftware.Steam/.local/share/Steam/steamapps/compatdata/1493710/pfx/drive_c/windows/system32/notepad.exe

...
There is an import table in .idata at 0x14000c000

The Import Tables (interpreted .idata section contents)
 vma:            Hint    Time      Forward  DLL       First
                 Table   Stamp     Chain    Name      Thunk
 0000c000    0000c0c8 00000000 00000000 0000d208 0000c510

    DLL Name: advapi32.dll
    vma:  Hint/Ord Member-Name Bound-To
    c958     253  IsTextUnicode
    c968     391  RegCloseKey
    c978     398  RegCreateKeyExW
    c98c     436  RegOpenKeyW
    c99c     446  RegQueryValueExW
    c9b0     463  RegSetValueExW

...

[still pebble]

inspecting steam proton notepad i see

[woeful rover]

i think they rewrote notepad

might be less painful to go with the one from before the rewrite