OBOS (not vibecoded) | Operating System Development | Page 26

flint idol Nov 20, 2024, 1:54 PM

#

git clone -b userspace-work https://github.com/OBOS-dev/obos

#

@limber wave

#

Which include build prerequisites

limber wave Nov 20, 2024, 1:56 PM

#

in command promot#

#

?

flint idol Nov 20, 2024, 1:56 PM

#

Yes

#

Luckily for you, if you have all the tools required, building on windows works

limber wave Nov 20, 2024, 1:58 PM

#

flint idol `git clone -b userspace-work https://github.com/OBOS-dev/obos`

and what does this command do?

flint idol Nov 20, 2024, 1:58 PM

#

Clones the obos repository

limber wave Nov 20, 2024, 1:59 PM

#

now what?

flint idol Nov 20, 2024, 2:00 PM

#

Look at instructions in readme

limber wave Nov 20, 2024, 2:00 PM

#

alr

#

C:\Users\Xavier>cmake -GNinja -DCMAKE_BUILD_TYPE=Debug --toolchain=src/build/x86_64/toolchain.cmake -B build/x86_64 .chmod +x dependencies/hyper/hyper_install-linux-x86_64cmake --build build/x86_64scripts/generate_initrd-x86_64.shcmake --build build/x86_64
CMake Warning:
Ignoring extra path from command line:

"build/x86_64scripts/generate_initrd-x86_64.shcmake"

CMake Error: Unknown argument --build
CMake Error: Run 'cmake --help' for all supported options.

C:\Users\Xavier>

#

the 2 step doesnt work

flint idol Nov 20, 2024, 2:04 PM

#

Copy one command at a time

limber wave Nov 20, 2024, 2:04 PM

#

also

#

wheres the cloned repo

flint idol Nov 20, 2024, 2:05 PM

#

In the directory you ran git clone in

limber wave Nov 20, 2024, 2:06 PM

#

but it wont make it defualt system right?

flint idol Nov 20, 2024, 2:06 PM

#

No

limber wave Nov 20, 2024, 2:06 PM

#

when you type in the final step

#

alr

flint idol Nov 20, 2024, 2:07 PM

#

Ofc not

#

Final step launches a vm

#

Using qemu

limber wave Nov 20, 2024, 2:07 PM

#

ah alr

flint idol Nov 20, 2024, 2:09 PM

#

Btw if any build errors happen, it is probably because you are missing a prerequisite tool

limber wave Nov 20, 2024, 2:09 PM

#

also

#

is hyper v recommanded

#

as vm

flint idol Nov 20, 2024, 2:09 PM

#

limber wave is hyper v recommanded

My script to launch qemu doesn't enable it

limber wave Nov 20, 2024, 2:10 PM

#

so no/

flint idol Nov 20, 2024, 2:10 PM

#

But yeah I'd recommend using hw accelerated vms

#

Because it's a lot faster

limber wave Nov 20, 2024, 2:10 PM

#

Like vb

flint idol Nov 20, 2024, 2:10 PM

#

Especially on debug mode

limber wave Nov 20, 2024, 2:10 PM

#

alr

flint idol Nov 20, 2024, 2:10 PM

#

Which is what those commands compile by default

limber wave Nov 20, 2024, 2:18 PM

#

which file do i use

flint idol Nov 20, 2024, 2:18 PM

#

Wdym?

#

@limber wave

limber wave Nov 20, 2024, 2:19 PM

#

there are 4 files

flint idol Nov 20, 2024, 2:19 PM

#

Yes and what are you tryna do

limber wave Nov 20, 2024, 2:19 PM

#

which do i use to put in the vm

flint idol Nov 20, 2024, 2:19 PM

#

ISO is stored under obos/out/obos.iso

#

Assuming the build was a success

limber wave Nov 20, 2024, 2:20 PM

#

wait imm install again

flint idol Nov 20, 2024, 2:22 PM

#

kk

#

Not any further than I was a couple days ago

#

Yup

#

No comment

limber wave Nov 20, 2024, 2:27 PM

#

wen i slect add in virtual box they ask me the file name what is it

flint idol Nov 20, 2024, 2:28 PM

#

limber wave wen i slect add in virtual box they ask me the file name what is it

out/obos.iso

#

If that doesn't exist your build failed

limber wave Nov 20, 2024, 2:28 PM

#

yep

#

then imma reinstall it

flint idol Nov 20, 2024, 2:28 PM

#

Make sure you have all prerequisites

flint idol Nov 20, 2024, 2:28 PM

#

flint idol Make sure you have all prerequisites

Otherwise it shouldn't fail

limber wave Nov 20, 2024, 2:29 PM

#

brb

flint idol Nov 20, 2024, 2:29 PM

#

Yeah I gtg too

#

Be back in 30 minutes

#

At the earliest

limber wave Nov 20, 2024, 2:30 PM

#

i will

limber wave Nov 20, 2024, 2:48 PM

#

w

limber wave Nov 20, 2024, 3:12 PM

#

bruh

#

how do you install cmake

#

@flint idol

#

@flint idol

flint idol Nov 20, 2024, 3:15 PM

#

Will assist you in 15 minutes

#

Currently in a crowded area

#

Aka masjid al haram

limber wave Nov 20, 2024, 3:16 PM

#

blud is riding a camal

#

jk

flint idol Nov 20, 2024, 3:17 PM

#

bru

#

It's far too crowded to ride a camel

#

Also I saw what you deleted

limber wave Nov 20, 2024, 3:19 PM

#

...

#

forgive me

weak kestrel Nov 20, 2024, 3:21 PM

#

limber wave how do you install cmake

sudo <package manager> install cmake

#

or if you are on a sane distro just type cmake and it'll ask you if you want to install it

flint idol Nov 20, 2024, 3:26 PM

#

weak kestrel sudo <package manager> install cmake

He's on windows

#

Basically

#

Search cmake

#

Find installer on website

#

Use installer

limber wave Nov 20, 2024, 3:33 PM

#

weak kestrel sudo <package manager> install cmake

also just corious what distro you use

flint idol Nov 20, 2024, 3:33 PM

#

I use debian

#

(Insert debian my beloved here)

#

L

#

Debian sid my beloved

limber wave Nov 20, 2024, 3:35 PM

#

arach vs debian

flint idol Nov 20, 2024, 3:35 PM

#

Boo

limber wave Nov 20, 2024, 3:35 PM

#

i heard debian had a bug which uninstalled your graphics driver'

flint idol Nov 20, 2024, 3:35 PM

#

Anyway gtg

limber wave Nov 20, 2024, 3:35 PM

#

randomly

flint idol Nov 20, 2024, 3:35 PM

#

limber wave i heard debian had a bug which uninstalled your graphics driver'

Wut

limber wave Nov 20, 2024, 3:35 PM

#

alr

flint idol Nov 20, 2024, 3:35 PM

#

I haven't had any major problems with debian

#

That were caused by debian itself

weak kestrel Nov 20, 2024, 4:07 PM

#

limber wave also just corious what distro you use

fedora

flint idol Nov 21, 2024, 7:28 AM

#

Yello guys

#

Kbye

weak kestrel Nov 21, 2024, 7:44 AM

#

flint idol Yello guys

Yellow guys

elfin wolf Nov 21, 2024, 4:14 PM

#

flint idol Aka masjid al haram

bro gave tech support at hajj

flint idol Nov 21, 2024, 5:03 PM

#

elfin wolf bro gave tech support at hajj

No hajj is during dhul hijjah

#

I was at umrah a day prior

#

Wait for file upload

#

@real pecan

real pecan Nov 21, 2024, 5:12 PM

#

flint idol <@353593948523986945>

This is legacy stuff

#

Acpi tables are built by qemu and not seabios

weak kestrel Nov 21, 2024, 5:18 PM

#

limber wave forgive me

https://tenor.com/view/virtual-sin-forgiveness-dan-hentschel-absolved-sins-sin-forgiveness-gif-5529119985214096225

Tenor

torpid wigeon Nov 21, 2024, 7:41 PM

#

flint idol Yello guys

bello

#

https://tenor.com/view/minions-minion-gif-18245938

Tenor

flint idol Nov 21, 2024, 8:58 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

torpid wigeon Nov 21, 2024, 9:06 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

weak kestrel Nov 21, 2024, 9:21 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

flint idol Nov 21, 2024, 9:27 PM

#

Just start a chain of that until I get back to obos work

torpid wigeon Nov 21, 2024, 9:29 PM

#

dont care

#

write better code

flint idol Nov 21, 2024, 9:29 PM

#

Huh

torpid wigeon Nov 21, 2024, 9:30 PM

#

oh wait

#

i thought you said "until i get obos back to work"

flint idol Nov 21, 2024, 9:31 PM

#

torpid wigeon oh wait

How does one

torpid wigeon Nov 21, 2024, 9:31 PM

#

reading fail

white mulch Nov 21, 2024, 9:31 PM

#

make a multisyscall bytecode in obos troll

flint idol Nov 21, 2024, 9:32 PM

#

white mulch make a multisyscall bytecode in obos <:troll:796858892985434152>

Why not ultrameme

white mulch Nov 21, 2024, 9:32 PM

#

I though it would be kinda funny, basically a syscall that you give a bytecode that can do more syscalls and move arguments/return values around

#

then you could write a hello world using one real syscall that combines print + exit (but with no additional context switch overhead)

#

or for a less cursed alternative just without the bytecode part like you though at some point KEKW

flint idol Nov 21, 2024, 9:36 PM

#

white mulch or for a less cursed alternative just without the bytecode part like you though ...

Just use AML

torpid wigeon Nov 21, 2024, 9:37 PM

#

aml syscall

flint idol Nov 21, 2024, 9:37 PM

#

Would be amazing if it weren't for the direct system access

white mulch Nov 21, 2024, 9:38 PM

#

you could do security checks in the uacpi kernel api fns

#

to make sure that userspace supplied aml can't do reads/writes to anywhere in mmio/io/pci

torpid wigeon Nov 21, 2024, 9:39 PM

#

just allow it

short mortar Nov 21, 2024, 9:56 PM

#

Who cares about security

torpid wigeon Nov 21, 2024, 11:04 PM

#

only nerds

flint idol Nov 22, 2024, 1:52 AM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

weak kestrel Nov 22, 2024, 4:05 AM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

flint idol Nov 22, 2024, 4:08 AM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

honest citrus Nov 22, 2024, 2:08 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

weak kestrel Nov 22, 2024, 3:29 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

flint idol Nov 22, 2024, 9:13 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

empty kernel Nov 22, 2024, 9:55 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

flint idol Nov 23, 2024, 6:37 AM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

sterile sparrow Nov 23, 2024, 8:17 AM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

flint idol Nov 23, 2024, 7:37 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

weak kestrel Nov 23, 2024, 8:20 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

flint idol Nov 23, 2024, 10:24 PM

#

Don't think my pc should be doing that

rn_image_picker_lib_temp_e5590c04-677e-44a2-8084-ebce6994ceee.jpg

#

Huh

#

It worked when I clicked a key

#

oh god

#

idk if it's because I haven't typed at this pace for a long time or that I somehow became faster at typing on my vacation

#

but my typing feels a lot faster

#

also I think osdev does make me go insane

#

I swear I felt a lot different before I left for vacation

#

probably just osdev and school driving me nuts

flint idol Nov 23, 2024, 11:09 PM

#

time to get back on obos grind

flint idol Nov 23, 2024, 11:25 PM

#

@white mulch I need you to test a seabios image

#

pls

#

📎 bios.bin

#

make sure ur stuff works with it since it does the mmconfig

#

on suspend

white mulch Nov 23, 2024, 11:30 PM

#

flint idol make sure ur stuff works with it since it does the mmconfig

it doesn't, the writes to the pci bars don't work

flint idol Nov 23, 2024, 11:30 PM

#

kthx

#

hmm

#

it seems to use ioconfig

#

on resume for me

white mulch Nov 23, 2024, 11:35 PM

#

actually yeah nvm lol, I were accidentally using the wrong run target that didn't use the bios

flint idol Nov 23, 2024, 11:36 PM

#

white mulch actually yeah nvm lol, I were accidentally using the wrong run target that didn'...

so it worked

#

?

white mulch Nov 23, 2024, 11:36 PM

#

yes

flint idol Nov 23, 2024, 11:36 PM

#

kthx

flint idol Nov 24, 2024, 12:23 PM

#

@real pecan my cousin decided to test obos on a laptop of his

#

and we'll see how uacpi works there

#

although currently it is refusing to go into obos

real pecan Nov 24, 2024, 12:24 PM

#

nice

#

im going to merge the first batch of thread safety soon

flint idol Nov 24, 2024, 12:24 PM

#

I told him to change the boot priority of the usb

real pecan Nov 24, 2024, 12:24 PM

#

it also adds an absolute ton of convenience functions

#

and reworks a lot of api

flint idol Nov 24, 2024, 12:25 PM

#

in case it was a bug with his bios' boot selector

flint idol Nov 24, 2024, 12:25 PM

#

real pecan and reworks a lot of api

cool

#

will be waiting

real pecan Nov 24, 2024, 12:25 PM

#

inb4 obos dies too fast

#

are u back from vacation?

flint idol Nov 24, 2024, 12:25 PM

#

yup

real pecan Nov 24, 2024, 12:25 PM

#

did u enjoy it

flint idol Nov 24, 2024, 12:25 PM

#

very

real pecan Nov 24, 2024, 12:25 PM

#

nice

#

oh also the new version will have successfully loaded & executed 1 AML blob in 2ms (49 opcodes, avg 19483 ops/s) etc

flint idol Nov 24, 2024, 12:26 PM

#

cool

#

more unneeded verbosity to my logs

#

(not complaining btw)

weak kestrel Nov 24, 2024, 12:26 PM

#

flint idol very

how was the sand halfmemeright
(sorry)

flint idol Nov 24, 2024, 12:26 PM

#

[  LOG  ] Arch_KernelMainBootstrap: Done early boot.
[  LOG  ] Currently at 8892 KiB of committed memory (1332 KiB pageable), 0 KiB paged out, 8756 KiB non-paged, and 0 KiB uncommitted. 16228 KiB of physical memory in use. Page faulted 933 times (0 hard, 933 soft).```

real pecan Nov 24, 2024, 12:26 PM

#

im not sure if thats the final look

flint idol Nov 24, 2024, 12:26 PM

#

weak kestrel how was the sand <:halfmemeright:715831702164996096> (sorry)

quite nice

real pecan Nov 24, 2024, 12:27 PM

#

i hate winter so some sand could be nice right now

#

@flint idol is there any way u could test the new pr before i merge it

#

i mean on real hw and qemu etc

flint idol Nov 24, 2024, 12:33 PM

#

probably

#

does it change any namespace node api?

real pecan Nov 24, 2024, 12:33 PM

#

yes

flint idol Nov 24, 2024, 12:33 PM

#

how much?

real pecan Nov 24, 2024, 12:33 PM

#

not a lot

#

one extra parameter, and u will have to call a different function

flint idol Nov 24, 2024, 12:33 PM

#

I'll try tn

#

*rn

real pecan Nov 24, 2024, 12:34 PM

#

thanks

flint idol Nov 24, 2024, 12:34 PM

#

real pecan not a lot

what about uacpi_object api

real pecan Nov 24, 2024, 12:34 PM

#

that hasnt changed

flint idol Nov 24, 2024, 12:35 PM

#

ok

#

I should be relatively good

real pecan Nov 24, 2024, 12:36 PM

#

yeah

#

this is final api btw

#

i wont change it further probably

flint idol Nov 24, 2024, 12:36 PM

#

big W

real pecan Nov 24, 2024, 12:38 PM

#

if i stop being lazy af 1.0 could be a few weeks away

flint idol Nov 24, 2024, 12:49 PM

#

cool

#

nooo

#

#

obos hangs on his laptop

real pecan Nov 24, 2024, 12:52 PM

#

unlucky

#

do u know where it hangs approximately?

flint idol Nov 24, 2024, 12:52 PM

#

not at all

#

only thing I know is it hangs after scheduler timer init

real pecan Nov 24, 2024, 12:52 PM

#

more debugging to do i guess

flint idol Nov 24, 2024, 12:53 PM

#

yup

#

but now I wanna test uacpi

#

's thread safety

real pecan Nov 24, 2024, 12:53 PM

#

do it

#

now u should be able to enable events right away

#

it shouldnt race with init

#

things that are still racey: opregion install/uninstall, event handler install/uninstall

#

but only if done at runtime

#

the interpreter and namespace stuff should be fully thread safe now

flint idol Nov 24, 2024, 12:54 PM

#

nice

real pecan Nov 24, 2024, 12:55 PM

#

flint idol

at least hyper worked so thats nice

#

can try uefi mode too

flint idol Nov 24, 2024, 12:55 PM

#

legacy didn't work

real pecan Nov 24, 2024, 12:55 PM

#

ah

flint idol Nov 24, 2024, 12:55 PM

#

uefi mode hung in obos

#

legacy didn't boot in the first place, just went onto booting into linux

real pecan Nov 24, 2024, 12:56 PM

#

cursed laptop

flint idol Nov 24, 2024, 12:56 PM

#

real

real pecan Nov 24, 2024, 12:56 PM

#

very good RetarBIOS implementation

barren merlin Nov 24, 2024, 12:57 PM

#

flint idol

nice T420

flint idol Nov 24, 2024, 1:00 PM

#

real pecan very good RetarBIOS implementation

I can ignore node_depth, right?

real pecan Nov 24, 2024, 1:00 PM

#

ye

flint idol Nov 24, 2024, 1:00 PM

#

kthx

real pecan Nov 24, 2024, 1:01 PM

#

uacpi_namespace_for_each_child is the full one, u can use uacpi_namespace_for_each_child_simple, which is more like the old api

flint idol Nov 24, 2024, 1:01 PM

#

k

real pecan Nov 24, 2024, 1:01 PM

#

but the former can filter by type etc which is convenient

#

also eval has a billion shorthands now

flint idol Nov 24, 2024, 1:02 PM

#

since I'm only testing to see if it will work rn, I'll just use the one more like the old api

real pecan Nov 24, 2024, 1:02 PM

#

ye

flint idol Nov 24, 2024, 1:08 PM

#

@real pecan

#

#

actually

#

more so
shareable.c:(.text+0x236): undefined reference to uacpi_atomic_inc32' /home/oberrow/x86_64-elf-tools/bin/../lib/gcc/x86_64-elf/13.2.0/../../../../x86_64-elf/bin/ld: ../uACPI/libuacpi.a(shareable.c.obj): in function uacpi_shareable_unref':
shareable.c:(.text+0x2e0): undefined reference to uacpi_atomic_dec32' /home/oberrow/x86_64-elf-tools/bin/../lib/gcc/x86_64-elf/13.2.0/../../../../x86_64-elf/bin/ld: ../uACPI/libuacpi.a(shareable.c.obj): in function uacpi_shareable_refcount':
shareable.c:(.text+0x3de): undefined reference to `uacpi_atomic_load32'\

#

those ones

#

the others I know are my fault

real pecan Nov 24, 2024, 1:09 PM

#

they're in platform/atomic.h

#

did u override it?

flint idol Nov 24, 2024, 1:09 PM

#

probably not

real pecan Nov 24, 2024, 1:09 PM

#

if not then it should be there

flint idol Nov 24, 2024, 1:09 PM

#

real pecan Nov 24, 2024, 1:10 PM

#

#define uacpi_atomic_inc32(ptr) __atomic_add_fetch(ptr, 1, __ATOMIC_ACQ_REL)

#

does ur atomic.h header have that?

#

its this commit: https://github.com/UltraOS/uACPI/pull/93/commits/17cc35203ca608e21290618c54ee0d72dce82632

#

ah nvm

#

it's because u override arch helpers

#

and arch helpers include atomic

#

and other files expected to get atomic through it

#

ill fix those but now u can just add that include

real pecan Nov 24, 2024, 1:13 PM

#

flint idol more so shareable.c:(.text+0x236): undefined reference to `uacpi_atomic_inc32' /...

or well i guess its just the shareable.c file

real pecan Nov 24, 2024, 1:15 PM

#

real pecan and other files expected to get atomic through it

thats wrong ofc, so good thing we found this problem

flint idol Nov 24, 2024, 1:28 PM

#

yup

flint idol Nov 24, 2024, 1:35 PM

#

real pecan ill fix those but now u can just add that include

what do I name it?

#

uacpi_atomic.h

#

?

real pecan Nov 24, 2024, 1:35 PM

#

in your arch helpers just add #include <uacpi/platform/atomic.h>

flint idol Nov 24, 2024, 1:36 PM

#

kk

real pecan Nov 24, 2024, 1:37 PM

#

https://github.com/OBOS-dev/obos/blob/userspace-work/src/uACPI/uacpi_arch_helpers.h

GitHub

obos/src/uACPI/uacpi_arch_helpers.h at userspace-work · OBOS-dev/obos

I was bored... Contribute to OBOS-dev/obos development by creating an account on GitHub.

#

u actually use uacpi's atomics here

#

so would be a good idea to include it anyway

flint idol Nov 24, 2024, 1:38 PM

#

oh yea

#

I do

short mortar Nov 24, 2024, 1:44 PM

#

real pecan im going to merge the first batch of thread safety soon

I have no time for anything rn, but I can update my OS to test it

real pecan Nov 24, 2024, 1:44 PM

#

would be cool if u have time, but keep in mind its just a pr right now and not merged

flint idol Nov 24, 2024, 1:47 PM

#

@real pecan It works

real pecan Nov 24, 2024, 1:47 PM

#

real hw too?

flint idol Nov 24, 2024, 1:47 PM

#

on qemu

#

real hw I will test shortly

real pecan Nov 24, 2024, 1:47 PM

#

well thats nice anyway

flint idol Nov 24, 2024, 1:50 PM

#

k rebooting now to test it

#

It appears to hang

#

Idk if that's my fault or yours

flint idol Nov 24, 2024, 1:52 PM

#

flint idol It appears to hang

Specifically on the event stuff

#

@real pecan

real pecan Nov 24, 2024, 1:52 PM

#

i mean idk either

#

can u add printfs or something

flint idol Nov 24, 2024, 1:52 PM

#

I'll see if it hangs when I use uacpi master

real pecan Nov 24, 2024, 1:53 PM

#

that wouldnt mean much

#

because it now uses mutex api and event api and stuff and maybe u handle some of that wrong

#

do u have deadlock detection in your mutexes?

flint idol Nov 24, 2024, 1:54 PM

#

real pecan do u have deadlock detection in your mutexes?

think so

real pecan Nov 24, 2024, 1:54 PM

#

then it's probably your skill issue that it hangs

#

only way it would hang in uacpi is mutex deadlock if some logic is wrong somewhere

flint idol Nov 24, 2024, 1:55 PM

#

It does check for recursive locks

#

And asserts it

real pecan Nov 24, 2024, 1:55 PM

#

u should add like

#

mutex acquire time outs

flint idol Nov 24, 2024, 1:55 PM

#

Eventually

real pecan Nov 24, 2024, 1:55 PM

#

so that it panics and dumps the waiters

real pecan Nov 24, 2024, 1:56 PM

#

flint idol And asserts it

it could be like

T0:          T1:
acquire_a()  acquire_b()
acquire_b()  acquire_a()

#

anyway more printffing is needed

#

yeah im looking at event code rn and dont see anything wrong instantly

flint idol Nov 24, 2024, 2:04 PM

#

I decided to log every mutex acquire and release

#

This is gonna take a long time

real pecan Nov 24, 2024, 2:04 PM

#

indeed

flint idol Nov 24, 2024, 2:06 PM

#

It's been 4 minutes

#

Still no sign of it being close to that code

real pecan Nov 24, 2024, 2:08 PM

#

lol

flint idol Nov 24, 2024, 2:09 PM

#

Nearly 10 minutes

#

Still eating away at logs

#

Bruh

#

It hung outside of mutex code

real pecan Nov 24, 2024, 2:11 PM

#

damn

#

well now u know its unrelated to mutexes

flint idol Nov 24, 2024, 2:12 PM

#

Yup

real pecan Nov 24, 2024, 2:13 PM

#

didnt u already have this bug before where events would hang or something?

flint idol Nov 24, 2024, 2:14 PM

#

btw, what my kernel's idea of "event" is different than uacpi's idea, iirc, so I made some half-assed implementation

#

that I never tested

real pecan Nov 24, 2024, 2:14 PM

#

events are only used for global firmware lock

#

well and aml Event objects

#

that code hasnt changed in a while

real pecan Nov 24, 2024, 3:01 PM

#

did u figure it out?

flint idol Nov 24, 2024, 3:01 PM

#

helping my cousin run an obos vm

real pecan Nov 24, 2024, 3:02 PM

#

ah

inland radish Nov 24, 2024, 3:16 PM

#

real pecan it could be like ``` T0: T1: acquire_a() acquire_b() acquire_b() acqu...

do you do this

flint idol Nov 24, 2024, 3:16 PM

#

no

inland radish Nov 24, 2024, 3:16 PM

#

im talking about infy

flint idol Nov 24, 2024, 3:16 PM

#

unless you were talking to infy

#

o

inland radish Nov 24, 2024, 3:17 PM

#

does UACPI acquire mutexes while having acquired mutexes

real pecan Nov 24, 2024, 3:20 PM

#

inland radish does UACPI acquire mutexes while having acquired mutexes

Yes, but without the abba

#

In limited cases

#

Certainly not where it currently hangs in obos

inland radish Nov 24, 2024, 3:22 PM

#

real pecan Yes, but without the abba

Does ACPI code acquire mutexes recursively and sometimes in reverse order?

#

Does Windows handle that case by deadlocking?

real pecan Nov 24, 2024, 3:26 PM

#

inland radish Does ACPI code acquire mutexes recursively and sometimes in reverse order?

This is not related to aml, and yes aml mutexes are recursive and yes they work in nt

#

And yes abba aml acquires would hang in nt

#

But thats not what happens here

inland radish Nov 24, 2024, 3:27 PM

#

ok, cool

real pecan Nov 24, 2024, 3:29 PM

#

Well sane aml doesnt do that generally

flint idol Nov 24, 2024, 4:13 PM

#

@real pecan this bug was there before I updated uacpi

#

The hanging

real pecan Nov 24, 2024, 4:20 PM

#

flint idol <@353593948523986945> this bug was there before I updated uacpi

Ah

#

So where does it hang exactly

flint idol Nov 24, 2024, 4:20 PM

#

Still don't know

real pecan Nov 24, 2024, 4:21 PM

#

Why did it work before

flint idol Nov 24, 2024, 4:21 PM

#

flint idol Still don't know

Only idea I had was because of my gpe bug

#

On x86_64

#

But I fixed that

real pecan Nov 24, 2024, 4:23 PM

#

Maybe thats why it also dies on your cousins laptop

flint idol Nov 24, 2024, 4:23 PM

#

real pecan Maybe thats why it also dies on your cousins laptop

Nah it died earlier than uacpi init

real pecan Nov 24, 2024, 4:23 PM

#

So like this worked before because u had events disabled?

flint idol Nov 24, 2024, 4:24 PM

#

real pecan So like this worked before because u had events disabled?

Still trying to figure out why it worked before

real pecan Nov 24, 2024, 4:25 PM

#

Lol

#

Obos curse from the past

flint idol Nov 24, 2024, 4:26 PM

#

Wtf

#

It did work before

#

But checking out that commit

#

Only reveals the opposite

#

Bruh wtf

#

I swear it worked before

real pecan Nov 24, 2024, 4:40 PM

#

Inb4 obos never even booted once and it was all a dream

flint idol Nov 24, 2024, 4:45 PM

#

Btw, a seabios maintainer asked if they could add it to the beginning of pci_resume, to which I replied "yes," after asking qwinci to test it

#

God damn it discord

#

Why didn't discord reply to the god damn message

#

Meant to reply to this

real pecan Nov 24, 2024, 4:46 PM

#

flint idol Btw, a seabios maintainer asked if they could add it to the beginning of pci_res...

They asked you to add it and then resend the patch

flint idol Nov 24, 2024, 4:47 PM

#

Yea the bug was fixed

thick jolt Nov 24, 2024, 7:09 PM

#

hihi

weak kestrel Nov 24, 2024, 7:56 PM

#

thick jolt hihi

https://tenor.com/view/haha-mickael-jackson-queendugif-ahah-hehe-gif-13642339

Tenor

flint idol Nov 24, 2024, 11:53 PM

#

@spark phoenix this thing is obos

#

anyway now that you're here

#

a star would be appreciated

spark phoenix Nov 24, 2024, 11:54 PM

#

how on mobile

flint idol Nov 26, 2024, 11:55 AM

#

broooooooooooo

#

I need to fix the hanging bug

#

but I don't wannaaaaaaaa

#

ah well

#

how sad

#

guess I have to

real pecan Nov 26, 2024, 12:06 PM

#

fix it 😡

#

so i can be sure its not a uacpi bug

flint idol Nov 26, 2024, 12:20 PM

#

real pecan so i can be sure its not a uacpi bug

It isn't

#

Since it persists with uacpi master

real pecan Nov 26, 2024, 12:21 PM

#

oh well

#

still debug

flint idol Nov 26, 2024, 12:21 PM

#

Yeah

#

Thank God it isn't with gpes

#

rn_image_picker_lib_temp_683ee589-ca66-46aa-b910-9f2c54b8d14c.jpg

#

@real pecan watdat mean

real pecan Nov 26, 2024, 12:25 PM

#

u dont have a notify handler installed for this device

#

it doesnt really mean anything

flint idol Nov 26, 2024, 12:25 PM

#

Oki

real pecan Nov 26, 2024, 12:25 PM

#

flint idol Thank God it isn't with gpes

wdym

flint idol Nov 26, 2024, 12:26 PM

#

Remember the gpe bug I fixed a couple weeks back

real pecan Nov 26, 2024, 12:26 PM

#

oh yeah

flint idol Nov 26, 2024, 12:26 PM

#

I was scared that it manifested itself again

#

Which was why I added the logs in bootstrap irq handler

real pecan Nov 30, 2024, 9:55 PM

#

@flint idol did u figure out the bug?

flint idol Nov 30, 2024, 10:26 PM

#

Haven't worked on obos in days

#

So nope

real pecan Nov 30, 2024, 11:04 PM

#

damn

#

what are u up to these days

flint idol Nov 30, 2024, 11:04 PM

#

School

#

nooo

real pecan Nov 30, 2024, 11:04 PM

#

f

torpid wigeon Nov 30, 2024, 11:08 PM

#

just don't go

flint idol Nov 30, 2024, 11:09 PM

#

I would never

torpid wigeon Nov 30, 2024, 11:09 PM

#

how much school u got

flint idol Nov 30, 2024, 11:09 PM

#

obos is love

torpid wigeon Nov 30, 2024, 11:09 PM

#

i have a full time job and more time than you

flint idol Nov 30, 2024, 11:09 PM

#

obos is life

#

Well some time is school

#

Some time is being tired after school

#

Some time is ignoring this bug

torpid wigeon Nov 30, 2024, 11:09 PM

#

no time for tired

#

you can be tired when you're sleeping

flint idol Nov 30, 2024, 11:10 PM

#

Most time is ignoring this bug

real pecan Nov 30, 2024, 11:10 PM

#

torpid wigeon i have a full time job and more time than you

likewise lol

torpid wigeon Nov 30, 2024, 11:10 PM

#

whenever i encounter a bug i can't figure out i rewrite the system that's around it and that usually fixes it

#

its effectiveness is debatable

flint idol Nov 30, 2024, 11:11 PM

#

For you, infy, I will debug this bug

torpid wigeon Nov 30, 2024, 11:11 PM

#

inshallah 🙏

real pecan Nov 30, 2024, 11:11 PM

#

this is a 5th obos rewrite

flint idol Nov 30, 2024, 11:12 PM

#

Now if only

#

I could get KDE to work properly

#

It broken

torpid wigeon Nov 30, 2024, 11:12 PM

#

kde user spotted

flint idol Nov 30, 2024, 11:12 PM

#

After update

flint idol Nov 30, 2024, 11:12 PM

#

real pecan this is a 5th obos rewrite

Yes.

real pecan Nov 30, 2024, 11:12 PM

#

flint idol For you, infy, I will debug this bug

do it

flint idol Nov 30, 2024, 11:13 PM

#

So I hang after events are registered

#

Seemingly

#

It's not with gpe handling

#

What

#

It works on bios

#

And on uefi

#

@real pecan the bug has decided to disappear

real pecan Nov 30, 2024, 11:22 PM

#

obos is stable once again

flint idol Nov 30, 2024, 11:23 PM

#

rn_image_picker_lib_temp_ae6cef18-2310-44c8-987b-879d316e9180.jpg

#

Although my time keeping is broke

#

Whatever happened to this bug, idk

#

Maybe you accidentally broke something in uacpi and unknowingly fixed it

real pecan Nov 30, 2024, 11:25 PM

#

you're using an outdated versin too

#

nah i doubt it's a uacpi problem

flint idol Nov 30, 2024, 11:25 PM

#

Well bugs don't just fix themselves, do they

real pecan Nov 30, 2024, 11:25 PM

#

hard reset the pc

flint idol Nov 30, 2024, 11:26 PM

#

real pecan hard reset the pc

The bug must have fixed itself then

#

This is the commit where you added the ops/s

real pecan Nov 30, 2024, 11:27 PM

#

wrong commit because i can see outdated formatting

real pecan Nov 30, 2024, 11:27 PM

#

flint idol The bug must have fixed itself then

probably a race condition in your code that is sometimes triggered idk

flint idol Nov 30, 2024, 11:27 PM

#

I will try it a couple more times

#

And will remove my printf

real pecan Nov 30, 2024, 11:28 PM

#

didnt u say even old uacpi was broken

flint idol Nov 30, 2024, 11:28 PM

#

And update my uacpi

flint idol Nov 30, 2024, 11:28 PM

#

real pecan didnt u say even old uacpi was broken

Yes

#

Before you merged this branch

real pecan Nov 30, 2024, 11:28 PM

#

yeah idk

flint idol Nov 30, 2024, 11:28 PM

#

But at one point it also worked

#

Must be magic or smth

real pecan Nov 30, 2024, 11:29 PM

#

curse

flint idol Nov 30, 2024, 11:30 PM

#

Maybe I need to do an exorcism on obos

real pecan Nov 30, 2024, 11:35 PM

#

well if it works it works

flint idol Nov 30, 2024, 11:35 PM

#

Until it doesn't

#

Which is why I'm making sure it does

#

For now

real pecan Nov 30, 2024, 11:36 PM

#

is this bug reproducable or gone completely?

flint idol Nov 30, 2024, 11:36 PM

#

Gine

#

*gone

real pecan Nov 30, 2024, 11:37 PM

#

flint idol

3800 ops is only 2x diff from qemu

#

pretty small aml

#

as a comparison, mint's latest pc mb is 16k ops

flint idol Nov 30, 2024, 11:37 PM

#

Also I think I might know why obos' ops/s is so slow

#

I have ubsan and kasan enabled

real pecan Nov 30, 2024, 11:37 PM

#

probably

flint idol Nov 30, 2024, 11:38 PM

#

I turned that off and it's 9k ops a second

#

Which still sucks

#

But I can also blame that on my.. questionable timekeeping method for uacpi

#

Or in general

real pecan Nov 30, 2024, 11:39 PM

#

true

#

a bit hehind keyronex at 750k

flint idol Nov 30, 2024, 11:40 PM

#

flint idol But I can also blame that on my.. questionable timekeeping method for uacpi

It's a 500hz tick count converted to ms

#

So not very precise

#

3800 ops in 200ms

#

On real hw

#

Which is a big number when converted to ops/s

#

Although that's probably a broken number

real pecan Nov 30, 2024, 11:42 PM

#

3800 * 5

flint idol Nov 30, 2024, 11:43 PM

#

rn_image_picker_lib_temp_d406c081-db35-4378-adf9-a97dd83b4275.jpg

#

Damn camera

#

rn_image_picker_lib_temp_42e4e05c-35cf-4190-9524-9a834b11bc70.jpg

real pecan Nov 30, 2024, 11:44 PM

#

yea...

flint idol Nov 30, 2024, 11:44 PM

#

Idk what's up with ur calculation

#

rn_image_picker_lib_temp_74371c8e-50f1-4ef8-8dac-7bd10341a88c.jpg

real pecan Nov 30, 2024, 11:45 PM

#

        uacpi_u64 ops = g_uacpi_rt_ctx.opcodes_executed;
        uacpi_u64 ops_per_sec = ops * UACPI_NANOSECONDS_PER_SEC;

        if (uacpi_likely(end_ts > begin_ts))
            ops_per_sec /= end_ts - begin_ts;

#

its literally this

flint idol Nov 30, 2024, 11:45 PM

#

Well

#

Idk why the number is like that atm

real pecan Nov 30, 2024, 11:46 PM

#

static uacpi_u64 elapsed_ms(uacpi_u64 begin_ns, uacpi_u64 end_ns)
{
    return (end_ns - begin_ns) / (1000ull * 1000ull);
}

#

maybe end ends up being less than begin in your case

flint idol Nov 30, 2024, 11:46 PM

#

Bru

real pecan Nov 30, 2024, 11:46 PM

#

and so it wraps around

flint idol Nov 30, 2024, 11:47 PM

#

I hate how obos uses so much memory, and I have no idea where

#

Like how is 17M physical memory in use

#

Where is it being used???

#

Symbol table is one thing

#

Drivers are another

#

Albeit not that significant

#

In terms of memory usage

#

Maybe it's pagecache

#

Along with uacpi

#

Since that uses many memory

real pecan Nov 30, 2024, 11:50 PM

#

allocate with tags

#

so u know

flint idol Nov 30, 2024, 11:50 PM

#

Good idea

#

If I want to port obos to esp32, I need to get that count down

#

By a lot

lean glen Nov 30, 2024, 11:51 PM

#

real pecan allocate with tags

I did that and I reduced allocations by a lot

empty kernel Nov 30, 2024, 11:56 PM

#

flint idol Along with uacpi

if you are allocating a new region for every uacpi_kernel_map, then that might be part of the issue

flint idol Nov 30, 2024, 11:56 PM

#

I am not

#

I just throw it in hhdm and call it a day

#

(I shouldn't)

flint idol Dec 1, 2024, 12:01 AM

#

real pecan yea...

I rebooted and it was a saner number

short mortar Dec 1, 2024, 12:03 AM

#

empty kernel if you are allocating a new region for every `uacpi_kernel_map`, then that might...

what else can be done?

empty kernel Dec 1, 2024, 12:04 AM

#

consolidate any physical regions in the same page into one virtual allocation

flint idol Dec 1, 2024, 2:00 AM

#

Ok so around 2M is page cache

#

110 kib is uacpi

#

158K by the kernel symbol table

flint idol Dec 1, 2024, 2:25 AM

#

Todo add an option to the kernel's early bump allocator to configure the size of the region

short mortar Dec 1, 2024, 7:58 AM

#

I'm just taking the largest region from the memory map, before I initialize my PMM

flint idol Dec 1, 2024, 3:55 PM

#

I'm gonna try to tag allocations now

#

so I'll start with the new api

#

void* (*TagAllocate)(struct allocator_info *This, size_t nBytes, allocator_tag tag, obos_status* status);```

#

typedef enum allocator_tag
{
    ALLOCATOR_TAG_MISC,
    ALLOCATOR_TAG_ACPI,
    ALLOCATOR_TAG_VFS,
    ALLOCATOR_TAG_SYMBOL,
    ALLOCATOR_TAG_DRIVER_INTERFACE,
    ALLOCATOR_TAG_HANDLE,
    ALLOCATOR_TAG_HANDLE_TABLE = ALLOCATOR_TAG_HANDLE,
    ALLOCATOR_TAG_SCHEDULER,
    ALLOCATOR_TAG_IRQ, // include dpcs, timers, irq objects
    ALLOCATOR_TAG_ARCH,
    ALLOCATOR_TAG_DRIVER,
    ALLOCATOR_TAG_SYNC_PRIMITIVE,
} allocator_tag;```

#

there are my allocator tags

#

stuff like strings go into misc

#

the rest is self explanatory

real pecan Dec 1, 2024, 4:14 PM

#

Could've just done 4 or 8 byte strings

#

Like nt

flint idol Dec 1, 2024, 4:14 PM

#

tru

#

but I was also maybe thinking of a global variable

#

real pecan Dec 1, 2024, 4:15 PM

#

U can compute this at runtime

flint idol Dec 1, 2024, 4:15 PM

#

true

real pecan Dec 1, 2024, 4:15 PM

#

By walking over every allocated range

flint idol Dec 1, 2024, 4:15 PM

#

mhm

lean glen Dec 1, 2024, 4:29 PM

#

flint idol

cherno

#

crigne

flint idol Dec 1, 2024, 4:30 PM

#

for reference:

#

that's my channel list on cherno server

#

the other channels are 💩

#

hotgarbage on top

#

the people who can take a god damn joke are in hotgarbage

#

and the rest are in the other channels

#

fun fact: that channel is the most active

flint idol Dec 1, 2024, 7:20 PM

#

oberrow@AcerAIO:~/Downloads/obos-uacpi-test$ grep -n '>ZeroAllocate' `find src/ -name '*.c'` | wc -l
68```
Oh god

#

68 more allocations I need to tag

real pecan Dec 1, 2024, 7:31 PM

#

acpica does it so no

flint idol Dec 1, 2024, 7:32 PM

#

uhhh

real pecan Dec 1, 2024, 7:32 PM

#

lol

flint idol Dec 1, 2024, 7:32 PM

#

that's a near two month old message

real pecan Dec 1, 2024, 7:32 PM

#

for some reason that was the message that was open on my pc

flint idol Dec 1, 2024, 7:32 PM

#

and you already responded

real pecan Dec 1, 2024, 7:32 PM

#

but enjoy ur answer

#

lmao

flint idol Dec 1, 2024, 7:32 PM

#

#1141057599584878645 message

real pecan Dec 1, 2024, 7:32 PM

#

oh well

#

i wonder how i ended up there

flint idol Dec 1, 2024, 7:34 PM

#

I did link a message around that point

#

here

#

#1302525344095408171 message

#

maybe you clicked on it

#

and discord did discord stuffs

real pecan Dec 1, 2024, 7:34 PM

#

yeah probably

flint idol Dec 1, 2024, 7:34 PM

#

for future reference, tagging allocations is boring

#

timer* Core_TimerObjectAllocate(obos_status* status)
{
    if (!OBOS_NonPagedPoolAllocator)
        return OBOS_KernelAllocator->ZeroAllocate(OBOS_KernelAllocator, 1, sizeof(process), status);
    return OBOS_NonPagedPoolAllocator->ZeroAllocate(OBOS_NonPagedPoolAllocator, 1, sizeof(process), status);
}```

#

wtf

#

as we can see, a struct timer is quite obviously the same thing as a struct process

real pecan Dec 1, 2024, 7:45 PM

#

did u expect it to be fun

flint idol Dec 1, 2024, 7:45 PM

#

well no

real pecan Dec 1, 2024, 7:46 PM

#

flint idol ```c timer* Core_TimerObjectAllocate(obos_status* status) { if (!OBOS_NonPag...

lol

#

part of why obos is so stable

flint idol Dec 1, 2024, 7:46 PM

#

copypasta

#

20 more to go

real pecan Dec 1, 2024, 7:49 PM

#

i hope process is larger than timer

flint idol Dec 1, 2024, 7:49 PM

#

it is

#

I think

#

lol

#

a FAT dent

#

oberrow@AcerAIO:~/Downloads/obos-uacpi-test$ grep -n '>ZeroAllocate' `find src/ -name '*.c' ! -path "*mbr.c" ! -path "*disk_swap.c" ! -path "*gpt.c" ! -path "*slowfat/io.c"` | wc -l 0

#

finally

#

no more allocations to tag

real pecan Dec 1, 2024, 7:57 PM

#

nice

flint idol Dec 1, 2024, 7:59 PM

#

ok it built

#

it still boots

#

which is a good sign

#

now time to dump the tags

flint idol Dec 1, 2024, 8:29 PM

#

Tag memory usage:
  UNKNOWN: 0 B
  PAGE: 247 KiB
  WS_NODE: 14 KiB
  PAGEINFO: 64 B
  PG_RANGE: 13 KiB
  WS_ENTRY: 44 KiB
  PCREGION: 608 B
  CONTEXT: 592 B
  CMDPARAM: 224 B
  PROC: 480 B
  THR_NODE: 288 B
  x86_64: 240 B
  CPULOCAL: 3 KiB
  THREAD: 1 KiB
  IRQ_NODE: 224 B
  IRQ: 272 B
  INITSWP: 14 KiB
  CLINEBMP: 16 B
  SYNC: 1 KiB
  ACPI: 113 KiB
  SYMNAME: 41 KiB
  SYMBOL: 98 KiB
  DRV_ID: 3 KiB
  DVINITRD: 33 KiB
  DRV_HDR: 2 KiB
  UART_STR: 16 B
  TMP: 80 B
  FD: 48 B
  FATFNODE: 64 B
  FATBPB: 704 B
  FATCACHE: 432 B
  FATDENT: 1 KiB
  DPC: 560 B
  SIGHDR: 3 KiB
  VFS: 6 KiB```

#

vmm context

#

I immediately see 80b leaked here

#

In TMP

#

[  LOG  ] Processing tagged memory allocations...
Tag memory usage:
  UNKNOWN: 1 KiB
  PAGE: 247 KiB
  WS_NODE: 14 KiB
  PAGEINFO: 64 B
  PG_RANGE: 13 KiB
  WS_ENTRY: 44 KiB
  PCREGION: 576 B
  CONTEXT: 240 B
  CMDPARAM: 224 B
  PROC: 416 B
  THR_NODE: 288 B
  x86_64: 240 B
  CPULOCAL: 3 KiB
  THREAD: 1 KiB
  IRQ_NODE: 224 B
  IRQ: 272 B
  INITSWP: 14 KiB
  CLINEBMP: 16 B
  SYNC: 1 KiB
  ACPI: 113 KiB
  SYMNAME: 41 KiB
  SYMBOL: 98 KiB
  DRV_ID: 3 KiB
  DVINITRD: 33 KiB
  DRV_HDR: 2 KiB
  UART_STR: 16 B
  TMP: 80 B
  FD: 48 B
  FATFNODE: 64 B
  FATBPB: 512 B
  FATCACHE: 432 B
  FATDENT: 1 KiB
  DPC: 48 B
  SIGHDR: 3 KiB
  VFS: 6 KiB
[  LOG  ] Mm_Allocator: 691 KiB in-use
[  LOG  ] OBOS_KernelAllocator: 1020 KiB in-use
[  LOG  ] OBOS_NonPagedPoolAllocator: 16 KiB in-use
[  LOG  ] Vfs_Allocator: 16 KiB in-use```

#

this is not making snese

#

all those tagged allocations combined are less than the amount used by the allocators

#

that were processed

#

and if I missed tagging some, it would be in UNKNOWN

#

well since this actually goes through each allocated block and accumulated the sizes that way

#

I'm going to assume those other numbers are buggy

flint idol Dec 1, 2024, 9:26 PM

#

I should make a slab allocator for obos

#

I start now

#

@real pecan when uAllocator

weak kestrel Dec 1, 2024, 9:32 PM

#

when obos can run managarm inside qemu/kvm

flint idol Dec 1, 2024, 9:32 PM

#

tomorrow

weak kestrel Dec 1, 2024, 9:32 PM

#

so in about 2-3 days

#

yes

flint idol Dec 1, 2024, 9:39 PM

#

/*
 * include/slab.h
 *
 * Copyright (c) 2024 Omar Berrow
 */

#pragma once

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

// To be defined by user.

// Read-Write No-Execute
extern void *slb_map(size_t size);
extern void  slb_unmap(void* base, size_t size);

typedef struct slab_allocator slab_allocator_t;

int   slb_init(slab_allocator_t* state);
void* slb_alloc(slab_allocator_t* state, size_t size);
void* slb_calloc(slab_allocator_t* state, size_t objs, size_t szObjs);
void* slb_realloc(slab_allocator_t* state, void* blk, size_t newsize);
void  slb_free(slab_allocator_t* state, void* blk);
void  slb_getsize(slab_allocator_t* state, void* blk, size_t* sz);```

real pecan Dec 1, 2024, 9:47 PM

#

flint idol <@353593948523986945> when uAllocator

im absolutely terrible at algorithms

flint idol Dec 1, 2024, 11:47 PM

#

this has changed heavily

#

so far

lean glen Dec 1, 2024, 11:57 PM

#

good

#

because that sucks

flint idol Dec 1, 2024, 11:58 PM

#

now the actual allocation api is the same because idk how I would change it

weak kestrel Dec 1, 2024, 11:58 PM

#

can you overload the new operator to take an extra argument

flint idol Dec 1, 2024, 11:58 PM

#

in C, no

weak kestrel Dec 1, 2024, 11:59 PM

#

like for what oberrow is doing

flint idol Dec 1, 2024, 11:59 PM

#

in C++, no

weak kestrel Dec 1, 2024, 11:59 PM

#

in C++

#

new(tag) type { };

lean glen Dec 1, 2024, 11:59 PM

#

weak kestrel can you overload the new operator to take an extra argument

yes

flint idol Dec 1, 2024, 11:59 PM

#

doubt

lean glen Dec 1, 2024, 11:59 PM

#

weak kestrel new(tag) type { };

yea

#

I do that

flint idol Dec 1, 2024, 11:59 PM

#

oh wait nvm

weak kestrel Dec 1, 2024, 11:59 PM

#

lean glen I do that

nice

lean glen Dec 1, 2024, 11:59 PM

#

https://github.com/nyx-os/gaia/blob/master/src/vm/anon.cpp#L46

#

example

flint idol Dec 2, 2024, 12:00 AM

#

#ifndef CACHE_SIZE_DEFAULT
#   define CACHE_SIZE_DEFAULT 4096
#endif

#ifndef slb_popcount
#   error Define slb_popcount. With GCC C extentions, this looks like __builtin_popcountll.
#endif

#ifndef slb_bsf
#   error Define slb_bsf. With GCC C extensions, this looks like __builtin_clzll.
#endif

#ifndef slb_bsr
#   error Define slb_bsr. With GCC C extensions, this looks like __builtin_ctzll.
#endif


#ifndef SLB_ALIGN
#   error Define SLB_ALIGN
#   define SLB_ALIGN(struct_, align_to)
#endif

// To be defined by user.
// Read-Write No-Execute
extern void *slb_map(size_t size);
extern void  slb_unmap(void* base, size_t size);

typedef struct slab_allocator *slab_allocator_t;

// Returns sizeof(struct slab_allocator)
size_t slb_allocator_size();

// NOTE: This is not thread-safe!
// Thread-safety must be added on top of these APIs.

// Returns zero on success, one on failure.
// maxSlabSize: The maximum slab size; must be a power of two greater than 32, and less than cacheSize. Hard limit is 0x40000.
//   cacheSize: The cache size; must be a power of two greater than maxSlabSize. If zero, CACHE_SIZE_DEFAULT is assumed. Hard limit is 0x40000
int slb_init(slab_allocator_t state, size_t maxSlabSize, size_t cacheSize);

void* slb_alloc(slab_allocator_t state, size_t size);
void* slb_calloc(slab_allocator_t state, size_t objs, size_t szObjs);
void* slb_realloc(slab_allocator_t state, void* blk, size_t newsize);
void  slb_free(slab_allocator_t state, void* blk);
void  slb_getsize(slab_allocator_t state, void* blk, size_t* sz);```

#

now the header is that

real pecan Dec 2, 2024, 12:00 AM

#

lean glen I do that

doesnt that conflict with placement new syntax

#

idk how that works

lean glen Dec 2, 2024, 12:00 AM

#

no

#

operator new(Tag) or whatever

real pecan Dec 2, 2024, 12:01 AM

#

ah ok nice that u can overload that

lean glen Dec 2, 2024, 12:01 AM

#

flint idol ```c #ifndef CACHE_SIZE_DEFAULT # define CACHE_SIZE_DEFAULT 4096 #endif #ifnd...

bro read the slab paper

flint idol Dec 2, 2024, 12:01 AM

#

no

lean glen Dec 2, 2024, 12:01 AM

#

that is not a slab 😭

flint idol Dec 2, 2024, 12:01 AM

#

it's something

real pecan Dec 2, 2024, 12:01 AM

#

SLOBOS

weak kestrel Dec 2, 2024, 12:02 AM

#

https://tenor.com/view/courage-the-cowardly-dog-gif-25234702

Tenor

#

I have no idea what this means

flint idol Dec 2, 2024, 12:02 AM

#

real pecan SLOBOS

good idea

#

because I ain't readin allat (yet)

real pecan Dec 2, 2024, 12:09 AM

#

I think SLOB is the slab implementation Linux uses for embedded

flint idol Dec 2, 2024, 12:10 AM

#

yes

#

but SLOBOS is some allocator for obos

real pecan Dec 2, 2024, 12:16 AM

#

Yup

flint idol Dec 2, 2024, 12:16 AM

#

which I'm hoping will be faster than whatever shit I have rn

vale nymph Dec 2, 2024, 12:23 AM

#

call it slop

flint idol Dec 2, 2024, 12:23 AM

#

not even a bad idea

#

because it's probably still gonna be shit

#

I'll call it slop

real pecan Dec 5, 2024, 11:25 PM

#

How is it going

flint idol Dec 5, 2024, 11:30 PM

#

real pecan How is it going

Obos progress died for a couple days

#

Today I need to finish some school work

#

And prove my friend wrong

#

He thinks I can't write a Morse code translator in one day

real pecan Dec 5, 2024, 11:32 PM

#

its literally a lookup table to ascii char?

flint idol Dec 5, 2024, 11:32 PM

#

Exactly

#

But he doesn't know that

flint idol Dec 5, 2024, 11:32 PM

#

real pecan its literally a lookup table to ascii char?

From English to Morse, so other way around

#

But yeah

vale nymph Dec 5, 2024, 11:32 PM

#

flint idol He thinks I can't write a Morse code translator in one day

I dare you to work on obos more I bet you couldn't do it troll

flint idol Dec 5, 2024, 11:33 PM

#

I will after making that stupidly simple thing

real pecan Dec 5, 2024, 11:33 PM

#

i bet u cant make the next year the year of obos desktop

flint idol Dec 5, 2024, 11:33 PM

#

Ez

#

I just copy Windows syscalls

#

Vam

#

*bam

flint idol Dec 6, 2024, 1:46 AM

#

flint idol He thinks I can't write a Morse code translator in one day

he challenged me specifically to get it to work on an arduino

#

void print_morse(const char* morse)
{
  for (int i = 0; morse[i]; i++)
  {
    digitalWrite(PULSE_PIN, HIGH);
    delay(pls);
    digitalWrite(PULSE_PIN, LOW);
    switch(morse[i])
    {
      case '.':
        delay(dit);
        break;
      case '-':
        delay(dah);
        break;
      case '/':
        delay(spc);
        break;
      default: break;
    }
  }
}

void serialEvent()
{
    int recv = Serial.read(); 
    if (recv == -1) 
      return;
    if (recv >= 127)
      return;
    const char* ent = table[recv];
    Serial.write(ent);
    Serial.write(" ");
    print_morse(ent);
}```

#

which was easy-ish

#

I also decided to mess around with power states so I could get an equivalent of x86's hlt instruction

#

table is just an ascii -> morse lookup table

#

const char* table[128] = {
    0,0,0,0,0,0,0,0,                                                 // 0-7
    0,0,0,0,0,0,0,0,                                                 // 8-15
    0,0,0,0,0,0,0,0,                                                 // 16-23
    0,0,0,0,0,0,0,0,                                                 // 24-31
    "/", 0,0,0,0,0,0,0,                                              // 32-39
    0,0,0,0,0,0,0,0,                                                 // 40-47
    "-----",".----","..---","...--","....-",".....","-....","--...", // 48-55
    "---..","----.",0,0,0,0,0,0,                                     // 56-63
    0,".-","-...","-.-.","-..",".","..-.","--.",                     // 64-71
    "....","..",".---","-.-",".-..","--","-.","---",                 // 72-79
    ".--.","--.-",".-.","...","-","..-","...-",".--",                // 80-87
    "-..-","-.--","--..",0,0,0,0,0,                                  // 88-95
    0,".-","-...","-.-.","-..",".","..-.","--.",                     // 96-103
    "....","..",".---","-.-",".-..","--","-.","---",                 // 104-111
    ".--.","--.-",".-.","...","-","..-","...-",".--",                // 112-119
    "-..-","-.--","--..",0,0,0,0                                     // 120-127
};```
now this is the actual table, beautiful, isn't it

flint idol Dec 6, 2024, 11:20 PM

#

ok I was working on a new allocator last time I was here

#

so now I'll work on that more

#

ok I fixed some bugs with it

#

now for the real test case, uacpi

#

(I couldn't be bothered to write a proper test case)

#

((but this is probably good enough))

real pecan Dec 6, 2024, 11:37 PM

#

why are u rewriting it again

flint idol Dec 6, 2024, 11:37 PM

#

speed

#

this one is slow as balls

#

as in the old one

#

that is in use for obos

#

it's like O(n) to allocate, with n being the amount of allocated regions in the allocator

real pecan Dec 6, 2024, 11:40 PM

#

i thought it was a normal slab allocator

flint idol Dec 6, 2024, 11:40 PM

#

what is?

#

the one currently being used in obos?

#

no that one is a shitallocator

real pecan Dec 6, 2024, 11:41 PM

#

yes

flint idol Dec 6, 2024, 11:41 PM

#

I made it with the goal of it not crashing

#

which is going well
but as for its speed...

#

this allocator crashes after a bit of uacpi

flint idol Dec 7, 2024, 12:13 AM

#

struct slobos* slobos = (void*)((uintptr_t)blk & ~(state->cacheSize-1));```

#

I have this line

#

to get a struct slobos from an allocation

#

now, if state->cacheSize were the page size, it should work

#

but it isn't

#

and since the slobos isn't guaranteed to be aligned to that margin, that line breaks

#

and idk how to fix it

torpid wigeon Dec 7, 2024, 12:32 AM

#

flint idol ```c struct slobos* slobos = (void*)((uintptr_t)blk & ~(state->cacheSize-1));```

why are they called slobos

flint idol Dec 7, 2024, 12:34 AM

#

torpid wigeon why are they called slobos

Because I didn't want to read the slab paper, and wanted to make a slab allocator from what I thought was a slab allocator

torpid wigeon Dec 7, 2024, 12:34 AM

#

slopos

flint idol Dec 7, 2024, 12:34 AM

#

Then abbix comes in and says not a slab allocator

lean glen Dec 7, 2024, 12:36 AM

#

yes

real pecan Dec 7, 2024, 12:37 AM

#

flint idol Then abbix comes in and says not a slab allocator

Wasn't that me

lean glen Dec 7, 2024, 12:37 AM

#

idk I did too

real pecan Dec 7, 2024, 12:37 AM

#

Slobos is my idea anyway

lean glen Dec 7, 2024, 12:37 AM

#

slobos is such a bad name

#

sloberrow when

real pecan Dec 7, 2024, 12:38 AM

#

Very much in line with the impl meme

flint idol Dec 7, 2024, 12:39 AM

#

real pecan Wasn't that me

You suggested the name

#

Abbix said it wasn't a slab allocator

flint idol Dec 7, 2024, 12:40 AM

#

lean glen slobos is such a bad name

Maybe I should call it slup

lean glen Dec 7, 2024, 12:40 AM

#

or just

#

read the paper

#

😭

flint idol Dec 7, 2024, 12:40 AM

#

No.

vale nymph Dec 7, 2024, 12:49 AM

#

real pecan Slobos is my idea anyway

slop os

torpid wigeon Dec 7, 2024, 7:17 AM

#

lean glen slobos is such a bad name

slob on my oberrow

thick jolt Dec 7, 2024, 2:22 PM

#

boo

flint idol Dec 7, 2024, 2:23 PM

#

ahhhh

#

scary

weak kestrel Dec 7, 2024, 2:29 PM

#

https://tenor.com/view/cat-jumpscare-box-gif-12399387153107465639

Tenor

flint idol Dec 7, 2024, 2:34 PM

#

after having fixed the bug with freeing

#

I think it passes uacpi test

#

the allocator

#

there was one error, but that was because I didn't load any ssdts

#

loading ssdts fixed that

#

now to port this to obos

#

now unfortunately, this allocator will probably never support tagging

#

as the only metadata it has on an allocation is that it exists in a struct slobos somewhere

#

but because of that it (probably) uses a lot less memory than it actually needs

flint idol Dec 7, 2024, 2:45 PM

#

flint idol I think it passes uacpi test

idk if it's gonna pass obos

#

we'll see

flint idol Dec 7, 2024, 3:05 PM

#

https://github.com/OBOS-dev/slobos/
I pushed that allocator

GitHub

GitHub - OBOS-dev/slobos

Contribute to OBOS-dev/slobos development by creating an account on GitHub.

#

to github

flint idol Dec 7, 2024, 9:41 PM

#

I ported the allocator to obos and wrote the neccessary glue code

#

now I just gotta use it somewhere

#

aaand it page faults almost immediately

#

I am able to get it to run a bit further

#

until it page faults again

#

meme

#

but I am 75% sure the page faults are a result of how I configured the allocator in this case

#

and not an allocator bug

#

ok now it got to uacpi init

#

where it hangs randomly

real pecan Dec 7, 2024, 9:58 PM

#

id blame uacpi tbh

coarse elk Dec 7, 2024, 11:13 PM

#

flint idol https://github.com/OBOS-dev/slobos/ I pushed that allocator

As a romanian I suggest you change that repo's name, its too close to the word sloboz

lean glen Dec 7, 2024, 11:13 PM

#

lmao

#

change it to sloberrow

flint idol Dec 7, 2024, 11:14 PM

#

coarse elk As a romanian I suggest you change that repo's name, its too close to the word s...

Which means...

coarse elk Dec 7, 2024, 11:16 PM

#

flint idol Which means...

a not so good way of saying "c*m"

real pecan Dec 7, 2024, 11:18 PM

#

bruh

flint idol Dec 8, 2024, 1:07 AM

#

oh fun fact: basically all work I've done on obos for the past two weeks has been done using nano

empty kernel Dec 8, 2024, 1:11 AM

#

why would you do that to yourself

flint idol Dec 8, 2024, 1:12 AM

#

because I couldn't bother configuring clangd for the two other copies of obos I have

empty kernel Dec 8, 2024, 1:13 AM

#

you use cmake right?

flint idol Dec 8, 2024, 1:13 AM

#

yes

empty kernel Dec 8, 2024, 1:13 AM

#

then clangd is easy

flint idol Dec 8, 2024, 1:13 AM

#

and I also couldn't bother setting up my ide

empty kernel Dec 8, 2024, 1:13 AM

#

what ide?

flint idol Dec 8, 2024, 1:13 AM

#

well less of an ide

#

and more of an advanced text editor

#

Kate

empty kernel Dec 8, 2024, 1:14 AM

#

ah

flint idol Dec 8, 2024, 1:14 AM

#

the KDE editor thing

empty kernel Dec 8, 2024, 1:14 AM

#

I know what it is, I've used KDE before

flint idol Dec 8, 2024, 1:14 AM

#

it has LSP support, but with how I have it, it's pretty dodgy

empty kernel Dec 8, 2024, 1:14 AM

#

why not use vscode or similar

flint idol Dec 8, 2024, 1:14 AM

#

wasn't a fan of vscode

#

and I don't know of any other free IDEs for linux

empty kernel Dec 8, 2024, 1:16 AM

#

you can get all the jetbrains ones using an open source license or a student license. They can be a bit buggy with clangformat though

weary hound Dec 8, 2024, 1:17 AM

#

emacs :^)

flint idol Dec 8, 2024, 1:17 AM

#

too much work

#

for both messages

empty kernel Dec 8, 2024, 1:17 AM

#

there is always good old vim (or neovim)

flint idol Dec 8, 2024, 1:18 AM

#

remove the 'good' part /j

lean glen Dec 8, 2024, 1:34 AM

#

nano 💀

slow dove Dec 8, 2024, 1:34 AM

#

nano is too good

#

how about echo "" >> "" trol

flint idol Dec 8, 2024, 1:35 AM

#

lean glen nano 💀

needed a text editor for "slight changes"

lean glen Dec 8, 2024, 1:35 AM

#

vim

#

or

#

emacs

#

meme

flint idol Dec 8, 2024, 1:35 AM

#

I wrote that allocator using nano

flint idol Dec 8, 2024, 1:36 AM

#

flint idol needed a text editor for "slight changes"

53 files changed, 356 insertions(+), 180 deletions(-)```

coarse elk Dec 8, 2024, 1:36 AM

#

nah bro ed is the shit

#

the standard Unix text editor

white mulch Dec 8, 2024, 12:22 PM

#

empty kernel you can get all the jetbrains ones using an open source license or a student lic...

they don't offer them for open source projects anymore (or well they do but only for bigger projects)

thick jolt Dec 8, 2024, 1:37 PM

#

white mulch they don't offer them for open source projects anymore (or well they do but only...

qwinci is alive????

thick jolt Dec 8, 2024, 1:38 PM

#

flint idol ahhhh

isnae!!

white mulch Dec 8, 2024, 2:08 PM

#

thick jolt qwinci is alive????

when have I not been alive lol

flint idol Dec 8, 2024, 2:09 PM

#

tomorrow /j

torpid wigeon Dec 8, 2024, 2:27 PM

#

🤨

weak kestrel Dec 8, 2024, 2:27 PM

#

oberrow what do you have planned

torpid wigeon Dec 8, 2024, 2:28 PM

#

do not the qwinci

flint idol Dec 8, 2024, 5:19 PM

#

my debian install seems to be completely fucked

#

so I am reinstalling it

#

be back in a bit

#

Uh oh

#

deb boot strap decided to die on me

#

In the middle of install

#

Luckily debian still booted

flint idol Dec 8, 2024, 6:24 PM

#

just reinstalled debian

empty kernel Dec 8, 2024, 8:46 PM

#

white mulch they don't offer them for open source projects anymore (or well they do but only...

Oh yeah, that starts next year

flint idol Dec 9, 2024, 5:22 PM

#

thick jolt Dec 9, 2024, 6:06 PM

#

white mulch when have I not been alive lol

do not the qwinci

flint idol Dec 11, 2024, 11:28 PM

#

kk I'm back

#

time to get back to pci interface

#

how fun, amiright

#

fuck it

#

I am refactoring the kernel entry

#

and init

real pecan Dec 11, 2024, 11:54 PM

#

refactoring how?

flint idol Dec 11, 2024, 11:54 PM

#

well for one

#

all timekeeping shit is in entry.c for some stupid reason

#

I moved that out of there

#

I said many moons ago that with the PCI interface redesign, I also make the ahci support multiple controllers

#

which has now been put off as a TODO

#

(aka never will do)

flint idol Dec 13, 2024, 11:27 PM

#

ok time to get back to doing this nooo

flint idol Dec 13, 2024, 11:46 PM

#

ok ahci driver ported

#

to new pci interface

#

except it's not fully done

#

so nothing will uhh

#

actually work

real pecan Dec 13, 2024, 11:46 PM

#

does it support suspend now

flint idol Dec 13, 2024, 11:46 PM

#

in the ahci driver

flint idol Dec 13, 2024, 11:46 PM

#

real pecan does it support suspend now

you see, everything is stubbed

#

in the new pci thing

#

oh wait it isn;t

#

*isn't

#

most stuff is

#

not all

#

I can update a resource in the pci config space

#

but I can't actually

#

identify any pci devices

#

and populate the trees n'stuff

#

4-5 months ago, I said I would be in userspace by the end of the month

#

nooo

real pecan Dec 13, 2024, 11:58 PM

#

Lol

flint idol Dec 14, 2024, 12:02 AM

#

flint idol 4-5 months ago, I said I would be in userspace by the end of the month

as in, bash or mlibc port

real pecan Dec 14, 2024, 12:02 AM

#

Well if you actually keep working

flint idol Dec 14, 2024, 12:02 AM

#

if even astral has came back to life

real pecan Dec 14, 2024, 12:05 AM

#

Astral is about to have full self hosting once gcc links

flint idol Dec 14, 2024, 12:05 AM

#

cool sauce

#

obos will be there within the next 9 years, trust

real pecan Dec 14, 2024, 12:06 AM

#

Chrome next

real pecan Dec 14, 2024, 12:06 AM

#

flint idol obos will be there within the next 9 years, trust

Not bad

flint idol Dec 14, 2024, 12:07 AM

#

typedef struct pci_bus {
    uint8_t busNumber;
    pci_device* devices;
    size_t nDevices;
    struct uacpi_namespace_node* acpiNode;
} pci_bus;```
when I'm initializing this, I am wondering whether I should probe the bus twice, once to see the device count, and then initialize the actual devices, or to just realloc(devices) for each new device

real pecan Dec 14, 2024, 12:07 AM

#

Why not have a linked list

#

Those should be fairly large structs

flint idol Dec 14, 2024, 12:08 AM

#

real pecan Those should be fairly large structs

it's 56 bytes

#

so not too big

flint idol Dec 14, 2024, 12:08 AM

#

real pecan Why not have a linked list

forgor why

#

probably because I won't be adding to the list after the bus is initialized

real pecan Dec 14, 2024, 12:08 AM

#

Id do a linked list unless performance critical

flint idol Dec 14, 2024, 12:08 AM

#

so insert and deletion isn't really important

lean glen Dec 14, 2024, 12:09 AM

#

why do you need this

flint idol Dec 14, 2024, 12:09 AM

#

abstraction

lean glen Dec 14, 2024, 12:09 AM

#

why

#

just do driver probing at bus iteration time

flint idol Dec 14, 2024, 12:10 AM

#

this is bus iteration time

#

I previously had no sane pci interface

#

just some irq stuff

lean glen Dec 14, 2024, 12:10 AM

#

ok

#

doesnt answer my question

flint idol Dec 14, 2024, 12:10 AM

#

and I'm not probing drivers either

lean glen Dec 14, 2024, 12:10 AM

#

why do you need to keep a list of devices

flint idol Dec 14, 2024, 12:10 AM

#

rn

flint idol Dec 14, 2024, 12:10 AM

#

lean glen why do you need to keep a list of devices

so I can probe drivers later

real pecan Dec 14, 2024, 12:11 AM

#

lean glen why do you need to keep a list of devices

Restoring state and bars after resume from suspend for one

lean glen Dec 14, 2024, 12:11 AM

#

flint idol so I can probe drivers later

do probing when iterating the bus?

flint idol Dec 14, 2024, 12:11 AM

#

lean glen do probing when iterating the bus?

boooring

lean glen Dec 14, 2024, 12:11 AM

#

if a driver gets added

#

reiterate it

lean glen Dec 14, 2024, 12:11 AM

#

real pecan Restoring state and bars after resume from suspend for one

same thing here

flint idol Dec 14, 2024, 12:11 AM

#

real pecan Restoring state and bars after resume from suspend for one

also this

lean glen Dec 14, 2024, 12:11 AM

#

just do reiteration

flint idol Dec 14, 2024, 12:12 AM

#

can't

#

the pci config space is ded

#

after resume from suspend

flint idol Dec 14, 2024, 12:12 AM

#

flint idol the pci config space is ded

all bars are reset

#

all everything is

lean glen Dec 14, 2024, 12:12 AM

#

ah

#

makes sense then

flint idol Dec 14, 2024, 12:12 AM

#

I needed to keep all pci devices' config in memory, otherwise I wouldn't be able to use them after resume

real pecan Dec 14, 2024, 12:13 AM

#

Why would the config be dead lol

flint idol Dec 14, 2024, 12:13 AM

#

well like

#

the pci devices

#

idk what it's called

real pecan Dec 14, 2024, 12:13 AM

#

Config is alive but bars are zeroed

flint idol Dec 14, 2024, 12:13 AM

#

yeah

#

and iirc irq stuff

#

msi stuff

real pecan Dec 14, 2024, 12:14 AM

#

So u can still enumerate them

flint idol Dec 14, 2024, 12:14 AM

#

not msi-x

#

well maybe partially

#

msi-x

lean glen Dec 14, 2024, 12:14 AM

#

ok so do that instead

#

reallocate bars

#

done

flint idol Dec 14, 2024, 12:15 AM

#

vale nymph Dec 14, 2024, 12:27 AM

#

real pecan Astral is about to have full self hosting once gcc links

gcc links, its just that libgcc configure (or build, I forgot) is fucked atm because of the test -x "" thing

flint idol Dec 14, 2024, 12:27 AM

#

what about binutils

vale nymph Dec 14, 2024, 12:27 AM

#

flint idol what about binutils

works, I compiled it once iirc

flint idol Dec 14, 2024, 12:27 AM

#

whoa

#

and how long does this all take?

vale nymph Dec 14, 2024, 12:28 AM

#

NaN

flint idol Dec 14, 2024, 12:46 AM

#

bruhhhhhhhh

#

I have to relearn everything I figured out

#

about bars

#

and detecting their existance

#

actually

#

no

#

ignore that

flint idol Dec 14, 2024, 4:49 PM

#

So I wanna do osdev

#

But I have to do homework

#

Now the question is what do I do

#

The correct answer is homework

#

But instead I will be flipping a coin

#

Tails is homework

#

Heads

#

Aw shucks

#

Guess I gotta do osdev now.

#

if (!module)
{
    OBOS_Warning("Could not load driver %s. Status: %d\n", module->name, OBOS_STATUS_NOT_FOUND);
    if (namelen != len)
        namelen++;
    iter += namelen;
    continue;
}```
oops

flint idol Dec 14, 2024, 5:21 PM

#

oops

#

I accidentally broke the framebuffer in my PCI initialization code

flint idol Dec 14, 2024, 5:27 PM

#

flint idol I accidentally broke the framebuffer in my PCI initialization code

well more like everything PCI

#

but anyway

flint idol Dec 14, 2024, 5:43 PM

#

fixed

#

I have BAR info now

#

in the pci resource list thingy

#

[ DEBUG ] Arch_KernelMainBootstrap: Initializing PCI bus 0
[ DEBUG ] 00:01:00: Initialized 32-bit, prefetchable BAR (BAR 0). BAR ranges from 0x00000000fd000000-0x00000000fdfffff0
[ DEBUG ] 00:01:00: Initialized 32-bit BAR (BAR 2). BAR ranges from 0x00000000febd4000-0x00000000febd4ff0
[ DEBUG ] 00:02:00: Initialized 32-bit BAR (BAR 0). BAR ranges from 0x00000000feb80000-0x00000000feb9fff0
[ DEBUG ] 00:02:00: Initialized 32-bit BAR (BAR 1). BAR ranges from 0x00000000feba0000-0x00000000febbfff0
[ DEBUG ] 00:02:00: Initialized I/O BAR (BAR 2). BAR ranges from 0x000000000000c040-0x000000000000c050
[ DEBUG ] 00:02:00: Initialized 32-bit BAR (BAR 3). BAR ranges from 0x00000000febd0000-0x00000000febd3ff0
[ DEBUG ] 00:1f:02: Initialized I/O BAR (BAR 4). BAR ranges from 0x000000000000c060-0x000000000000c070
[ DEBUG ] 00:1f:02: Initialized 32-bit BAR (BAR 5). BAR ranges from 0x00000000febd5000-0x00000000febd5ff0
[ DEBUG ] 00:1f:03: Initialized I/O BAR (BAR 4). BAR ranges from 0x0000000000000700-0x0000000000000730```

real pecan Dec 14, 2024, 5:57 PM

#

Nice

flint idol Dec 14, 2024, 5:57 PM

#

static void initialize_capability_resources(pci_device* dev)
{

}```
now all that's left is this

#

which simply formulates a list of PCI capabilities

#

and puts them in the pci resource list

#

of the device

#

then I need to implement the more advanced pci initialize thing

#

Drv_PCIInitialize()

#

which initializes other buses

#

then I need to implement Drv_PCIUpdateResource

#

which simply reads the actual value of a resource in the config space, and updates the resource to that value

#

then I want to add PCIe support

#

then I will be changing my "pnp" thing to support the new PCI interface

flint idol Dec 14, 2024, 7:30 PM

#

flint idol ```c static void initialize_capability_resources(pci_device* dev) { }``` now al...

done this

#

and I also added some more logs

#

becuz why not

#

for one device:

[ DEBUG ] PCI: 00:1f:02: Device HID: 01:06:01, Vendor ID: 0x2922, Device ID: 0x8086
[ DEBUG ] PCI: 00:1f:02: Initialized I/O BAR (BAR 4). BAR ranges from 0x000000000000c060-0x000000000000c070
[ DEBUG ] PCI: 00:1f:02: Initialized 32-bit BAR (BAR 5). BAR ranges from 0x00000000febd5000-0x00000000febd5ff0
[ DEBUG ] PCI: 00:1f:02: Found and registered MSI Capability.
[ DEBUG ] PCI: 00:1f:02: Found and registered SATA Data/Index Configuration Capability.```

#

imma change the capability log message

#

[ DEBUG ] PCI: 00:1f:02: Device HID: 01:06:01, Vendor ID: 0x2922, Device ID: 0x8086
[ DEBUG ] PCI: 00:1f:02: Initialized I/O BAR (BAR 4). BAR ranges from 0x000000000000c060-0x000000000000c070
[ DEBUG ] PCI: 00:1f:02: Initialized 32-bit BAR (BAR 5). BAR ranges from 0x00000000febd5000-0x00000000febd5ff0
[ DEBUG ] PCI: 00:1f:02: Found MSI Capability.
[ DEBUG ] PCI: 00:1f:02: Found SATA Data/Index Configuration Capability.
[ DEBUG ] PCI: 00:1f:02: Device has IRQ capabilities.```

#

#

Frick

#

It hangs on real hw

#

it hangs somewhere in the BAR initalization code

#

it could be a bug with how I handle 64-bit bars

#

since it says it initialized a 64-bit bar

#

then it simply hangs

#

BAR0 of the device at 00:02:00 was a 64-bit bar

#

therefore it should be skipping BAR1 since there isn't one

#

then it goes on to BAR2

#

I'm gonna add some printf

flint idol Dec 14, 2024, 7:54 PM

#

flint idol then it goes on to BAR2

Which indeed happens

#

bruhhhh

#

why does it hang

#

unless I accidentally mess something up causing the framebuffer to stop working properly

torpid wigeon Dec 14, 2024, 7:57 PM

#

check serial vro

flint idol Dec 14, 2024, 7:57 PM

#

torpid wigeon check serial vro

ah yes

real pecan Dec 14, 2024, 7:57 PM

#

flint idol unless I accidentally mess something up causing the framebuffer to stop working ...

Are you overwriting the fb bar to check size?

#OBOS (not vibecoded)