EvalynOS

[hazy saddle]

[rotund furnace]

I have no c compiler

ugh

[hazy saddle]

@viral bison that CPUID is barely supported it seems

[hollow crater]

10th gen is too old now
insert "feel old yet?" meme

[hazy saddle]

https://tenor.com/view/deltarune-gerson-boom-gerson-i'm-old-im-old-gif-13265713427338879500

[rotund furnace]

it also seems to be fully unsupported on amd

[hazy saddle]

Thanks Intel

[rotund furnace]

nvm

[hazy saddle]

And other hoo rah for intel right @rotund furnace

[rotund furnace]

limited amd support

eax: 0x3
ebx: 0xab
ecx: 0x124f800
edx: 0x0

for leaf 0x15

[rotund furnace]

linux lmao

[hazy saddle]

Uhhh so what’s the math for thoes

I’m on mobile rn

[hazy saddle]

[rotund furnace]

if I'm not wrong (might be)

598830 hz

[hazy saddle]

It should like be in HGz

GHz

[rotund furnace]

uhhhhhh

[hazy saddle]

Because the TSC runs at like GHz

[rotund furnace]

I added the calculation to my little test program

ecx * (ebx / eax)

and got

[hazy saddle]

I think it’s just fucking busted

[rotund furnace]

1094400000 hz

[hazy saddle]

That’s like 1ghz

[rotund furnace]

which is 1ghz

[hazy saddle]

I think

[rotund furnace]

which SEEMS accurate

[hazy saddle]

Yeah that sounds right fit that laptop

What does Linux say when it setups

I think it logs TSC calibration

[rotund furnace]

yep

tsc: detected 1094.400 mhz processor

and tsc deadline support

damn

intel does not play around with it's budget laptop chips

[hazy saddle]

So then what the fuck is up with the other laptop

[rotund furnace]

"kfence: initialized - using ... bytes for ... objects at 0x(____ptrval____)-0x(____ptrval____)"

[hazy saddle]

Wha

[rotund furnace]

that's in dmesg

lmao

[hazy saddle]

Looks like it’s censored

[rotund furnace]

oh no root will know where the kernel mode fence objects are

better censor it

[hazy saddle]

I mean by default it does that for all pointers

[rotund furnace]

no?

yeah lmao

[hollow crater]

amd just tells u "lol ACPI"

[rotund furnace]

the acpi table virtual addresses are in dmesg

[rotund furnace]

Aaaaaaaaaaaaaaaa

[hollow crater]

its also kinda more convenient?

[hazy saddle]

[hollow crater]

kekw

[hazy saddle]

What’s p0 frequency

[hollow crater]

whats so funny is that the cpu knows the fucking ratio 💔

[hazy saddle]

Honestly just calibrate the fucking TSC with HPET or PIT

They aren’t going anywhere

[hazy saddle]

This still needs a timer to wait

[hollow crater]

it seems to be stored an acpi structure but its basically the nominal frequency on a p0 p state
i might be wrong here

[hollow crater]

[hazy saddle]

Dosnt Linux just calibrate the TSC

[hollow crater]

yea lmao

[marble ridge]

they are

you cannot rely on either of them existing

[hollow crater]

ig it does first calibrate it with pit
THEN attempts to use hpet to do it again or something

[hazy saddle]

Replaced by what?

The PIT and HPET?

[marble ridge]

mostly the TSC and APIC timer and ACPI timer

[hollow crater]

ooo

[hazy saddle]

but they need calibration to work from other times

[marble ridge]

the ACPI timer does not need calibration and the TSC/APIC may not need calibration either depending on the CPU

[hazy saddle]

And trying to get the TSC frequency was being a little messy

[mild glacier]

evalyn do I make bad apple for my distro 🔥

[hazy saddle]

Kind of off topic rn but sure

[mild glacier]

i was waiting for this convo to end but then suddenly sped up bad apple appears xD

[hazy saddle]

Interesting. So I should just be able to grab ACPI timer along with PIC or HPET to be safe?

[hazy saddle]

[marble ridge]

frankly i would exclusively use the ACPI timer for calibration if i were to make a new OS right now

unless you plan to support ACPI-less x86 machines (you don't)

[hazy saddle]

Huh interesting

What is the ACPI timer?

[hazy saddle]

I mean I that that’s pre x86-64 anyways? So I should be fine

[rotund furnace]

kinda cooked when you need the apic timer before you have aml parsing in userspace working 🙃

[soft snow]

ehh i think you dont rly need to do anything when making bad apple for ur distro

[marble ridge]

well that sounds like a not-my-issue lol

i don't think you need AML parsing for the ACPI timer though?

[rotund furnace]

do you?

[hazy saddle]

You only need table parsing

[rotund furnace]

I have no idea

[hazy saddle]

Right?

[rotund furnace]

oh shit

sick

🔥

[hazy saddle]

It should be a table

[marble ridge]

https://wiki.osdev.org/ACPI_Timer

[hollow crater]

ig thats why u would just get some preliminary calibration
also i dont think u need timing for it
u just parse the tables

[hazy saddle]

I guess it has a chance of non existing?

[rotund furnace]

yeah doesn't seem like it

[marble ridge]

it will exist on literally every x86-64 machine afaik

[hazy saddle]

I could do a fallback if ACPI to HPET to PIT to your fucked lol

[rotund furnace]

the hardware/firmware people love this one simple trick to save a tiny bit of money 🔥

the software people can go and fuck themselves

[marble ridge]

i mean, sure, if you want to be 100% safe, but i don't think i know of an x86-64 machine without an ACPI timer, meanwhile i know of and own machines without the PIT and where the HPET is unreliably present (it may be disabled in the BIOS and sometimes it is so by default)

[hazy saddle]

I mean my current system lets me chain them really easy

So it’s no biggie

If one fails it attempts the next and replaces function pointers for relavent services

[marble ridge]

like i am not saying there is a reason why it is bad to have the fallbacks - fallbacks are never bad, beyond the fact that it could be annoying to maintain

[hazy saddle]

Have you ever found one without an IOAPIC btw?

[marble ridge]

nope

[hazy saddle]

I just won’t maintain it beyond compiling if I don’t notice it

[marble ridge]

troll

[rotund furnace]

"yeah the timers can automatically fall back if one isn't supported"

"oh neat do those timers work"

"idfk"

[hazy saddle]

“They did when I coded them”

[rotund furnace]

yep!

tbf I just calibrate the apic timer with the pit chip rn

[marble ridge]

i mean thankfully hardware doesn't really change, if anything

[rotund furnace]

could I do better? sure

am I lazy as fuck? yeah

[hazy saddle]

@rotund furnace Once I finish my suite of timer drivers steal mine

[rotund furnace]

bet 🔥

but it's mit :(

[hazy saddle]

Il give you a special exception as a friend that I won’t break your legs

[rotund furnace]

sick

[hazy saddle]

What is your licence?

[rotund furnace]

3bsd

[hazy saddle]

That’s compatible

[rotund furnace]

it is?????

[hazy saddle]

BSD0 may not be tho

It’s like the same thing

[rotund furnace]

oh it is

sick

nothing can stop me from stealing the code 🔥

[hazy saddle]

Just like include the licence in a comment on the header

🔥

Else this will happen

https://youtu.be/ZomwVcGt0LE

@marble ridge thanks for telling me about the ACPI timer. That should make it bullet proof I think in terms of timing devices along with PIT APIC HPET and TSC

[hollow crater]

told ya u will have a whole lot of timers at ur disposal

now drop pit

[marble ridge]

np!

[hazy saddle]

NEVER

[rotund furnace]

add acpi timer to the list of shit to procrastinate on

[hollow crater]

😔

[hazy saddle]

You don’t know the work it took for me to make it work

https://tenor.com/view/rice-cakes-family-guy-data-moshing-glitch-peter-griffin-gif-6554193917992872111

[rotund furnace]

as well as usermode server discovery and dropping uacpi in kernel mode (no hard feelings just pretty overkill for just table parsing)

[hazy saddle]

That will be part of my timer suite

And the ACPI parsing

[rotund furnace]

I wanna write my own acpi shit

because like why the fuck not

[hollow crater]

like its really slow
veeeery bad precision
and like ig every other one has some advantage over it in some way or another 💀

[hollow crater]

oh fair

[hazy saddle]

But the effort 😭

[hollow crater]

.name = "PIT, its just setting there for nostalgic reasons"

[hazy saddle]

Wdym?

[hollow crater]

o well i thought i can abstract timers through some common interface here
so i describe them with structures and then the backing clock source and so on
not sure if thats actually a good idea anyway but i was joking about some name field describing pit which would have that string

since
in ur case u are keeping it for the effort it took u to get it up and stuff xP

[hazy saddle]

I do

Function pointers

Whenever a timer or abstracted device is setup

It replaces the function pointers for it

I could do a struct

But I just have them plainly there

Because it’s the same shit

[hollow crater]

i thought id like
have the functions for setting, changing modes, reading current value, eoi and so on as function pointer for each struct
and i store irq line, divisor, freq, and whether its active as members
but then never implemented a lot of that and got distracted away to do memory management stuff
but i thought it'd be cool to abstract them like that
tho its just not necessary since one mostly wont end up using most of them anyway
so like it solves no problem

while ur approach is just having global pointers and set them to what u will be using
which kinda makes more sense ig

:P

[hollow crater]

yea makes sense

[viral bison]

that is from WSL tho

idk if WSL is causing issues

[hazy saddle]

Store the members inside the driver file

And then have functions to read or write them

Then change function pointers

[hollow crater]

i still have no driver layer yet
well i basically have nothing yet in that regard
just pmm/vmm and some other scattered pieces atp lolz

[hazy saddle]

I don’t either

It’s just exported in the header file

And a varrible

Ezpz

If I get dynamic loading I may freshen it up a bit though

With some tips from others

[hazy saddle]

Legacy APIC driver complete

Also why does this laptop have two IOAPICs

Why does this cheap ass shit ass laptop have TWO

What could it even be using them all for

@rotund furnace @hollow crater

[hollow crater]

cool

[hollow crater]

um is that even possible 💀

[hazy saddle]

That’s what my ACPI parsing says 💀

And my desktop also has two

[hollow crater]

also have u made bad apple play with a sane framerate yet

[hazy saddle]

Which is unusual I guess because it’s also cheap and not a server

[hazy saddle]

Always have been

That was just a fuckup with the timer calibration

[hollow crater]

yea thats weird..
sad I cant try shit now cuz im
supposedly
studying for an exam tmrw

[hollow crater]

wooooooooo

well thats cool
im sure my shitty thing would break under 2 out of these 3 machines

[hazy saddle]

Though it does have RGB so I guess they slapped an extra one on there for good measure?

[hazy saddle]

My kernel is pretty stable tbh

Just hollow with a few things

[hollow crater]

well if thats for the PC then does the same reasoning apply to the laptop.. 💀

[hollow crater]

still cool to watch it running on multiple machines
satisfying xP

[hazy saddle]

This laptop came with 4gb ddr3

💀

No RGB

720p screen

Entirely made of plastic

Also came with a hard drive

My grandma gave it 16gb ram and an ssd though in 2015 🙏

[hollow crater]

then how 😭 😭

[hollow crater]

lmao well now im wondering how is it even alive

[hazy saddle]

The CMOS battery is dead

But it’s my childhood laptop

And I took good care of it

I played so much heavily modded Minecraft on that shit 🙏

[hollow crater]

😭 😭
well sure it does appreciate that care
u shall get pardoned at the 2067 agi genocide against humans

[hazy saddle]

I treat all my computers and devices with respect

And they respect me

@hollow crater when taking the USB stick out of the laptop I pushed it down and the thing froze for a second 💀

But like this is not a bad condition for a laptop from 2015 tbh

The battery was removed because it died

And I can’t find replacements online

[mild glacier]

damn that's in like 5th grade now 💀

[hollow crater]

other newer devices were thrown away just bc they dont support win 11 🥀

shit runs evalynOS wdym unsupported device

[hazy saddle]

My goal is to support every x86-64 computer ever

And any other 64bit platforms later

Also

I’m cooking up some ideas for a micro kernel

Once I get IPC and other stuff working in this one

I may build a microkernel off of the EvalynOS kernel

I’m gonna call it pwease

With

As the logo

Because it’s a micro kernel

[hollow crater]

based (good luck with that kekw)
tbh I wanna have a riscv64 pory if I ever had a solid x86-64 one
but an aarch64 seems tough
at elast thats how it thought of it

[rotund furnace]

:O

we can compete on who has the most mid microkernel

[hollow crater]

Im doing monolithic yet I name it lil kernel
now that u r saying it ig it couldve been much cooler if I went microkernel just for the name

[hazy saddle]

I’m gonna try and have dynamic memory allocation in userspace

Same for scheduling

I would have a fixed set of slots for core ųKernel threads

For core services

So like the file server would call up the memory server for some memory

@rotund furnace This is where my
Commit is at right now

[mild glacier]

-# :3

[mild glacier]

struct window 🔥
idk why, but still

struct window {
  short xres;
  short yres;
  char layer; // when to get rendered
  unsigned char *buf;
}

prolly not guud

[rotund furnace]

okay so is this like in kernel mode

or user mode

[rotund furnace]

you know you don't have to add EVERYTHING in one commit

[hazy saddle]

A lot of this shit is interconnected and as I was going I was reworking all the systems

I start on one thing and I do many

[opaque marten]

some hp laptops do for some reason, and linux will confirm it

[hazy saddle]

Yeah that is an HP laptop

Another HP laptop only has one

Though the one with two is AMD

My desktop also has AMD

[hazy saddle]

and all the legacy crap still works and gets all fallen back to

(tested by disabling ACPI parsing)

@hollow crater this also has the missing IOAPIC message

i also handle legacy PIC EOIs by just sending them whenever a legacy device IRQ is ran

regardless because the lesser speed isnt a biggie

[hazy saddle]

@rotund furnace you got a moment to test my iso

[rotund furnace]

what am I testing

[hazy saddle]

that the timers come up properly

and bad apple runs at proper speed

[rotund furnace]

what timers should come up?

[hazy saddle]

and a photo of the init text

[hazy saddle]

it auto picks based on your system

[rotund furnace]

and it #pfed on boot

[hazy saddle]

wha

i was having issues with ISOs not updating right

leme try this

@rotund furnace even the fuckin checkups diddnt match 💀

[rotund furnace]

wait

do you compile

with debug symbs on?

[hazy saddle]

yes

[rotund furnace]

it crashed on strncmp

[hazy saddle]

huh

try this one

[rotund furnace]

cool kpanic btw

[hazy saddle]

what CR2 does it fault on

[hazy saddle]

ty

[rotund furnace]

hm

strncmp

the first param

HERE

acpi_parse_fadt

[hazy saddle]

same with new iso?

[rotund furnace]

I tracked the stack trace

no

testing it now

read from non present

with lines up

same

[hazy saddle]

gimme IP

[rotund furnace]

kpanic on boot

[hazy saddle]

this code should be fine?

whats where i strncmp

firmware giving bad data?

send a photo

[rotund furnace]

rip -> ffffffff80007880
ffffffff80005eb3
ffffffff80005e2d
ffffffff8000e0ef
ffffffff800012f4
ffffffff80001828
ffffffff80001025
ffffffff800062c6```

that's the whole trace

cr2 = rdi = 0xffff800075063000

still fadt parsing

[hazy saddle]

seems the firmware is giving bad data?

[rotund furnace]

hm

are your changes on codeberg rn

[hazy saddle]

no

[rotund furnace]

damn

[hazy saddle]

it is parsing the madt too

[rotund furnace]

could you put them on codeberg

even on a branch

or as a patch set

cuz I'm happy to debug this rn

I aint got shit else to do

[hazy saddle]

il make a patch whilst i cleanup some shit

[rotund furnace]

ty

[hazy saddle]

do you have other PCs to test?

[rotund furnace]

nope

other then my macbook

I don't have a monitor or powercable for my ewaste pc yet

sadly

[hazy saddle]

prolly add some print statments to the this thing

see what it gives

[rotund furnace]

again I'm happy to do so

[hazy saddle]

tyty

leme make the ascii logo for my micro kernel rq

[rotund furnace]

ugh

git thinks this patch doesn't apply

for some reason

[hazy saddle]

did you goto dev branch

its based off dev

[rotund furnace]

ye

[hazy saddle]

awe

but i did it and il clean it up later

[rotund furnace]

sync dev to master is the latest right?

[hazy saddle]

no

wait i never pushed to berg

[rotund furnace]

😭

[hazy saddle]

https://git.evalyngoemer.com/evalynOS/evalynOS/src/branch/dev

its pushed there

[hazy saddle]

what even kind of fucking filter even nammed this?

[rotund furnace]

how the fuck do I apply this

git apply huh.patch doesn't work

[hazy saddle]

huh

leme remake it?

[rotund furnace]

BRO

please

[hazy saddle]

what

[hazy saddle]

💀

wtf

maybe this will help

[rancid viper]

lol

how are u creating the patch?

[rotund furnace]

okay

yeah

that one worked

[hazy saddle]

git diff --staged > huh.patch

[rotund furnace]

so I have bad news

uh

it's crashing instantly

on like

the first acpi_find_sdt

[hazy saddle]

in qemu?

[rotund furnace]

real hw only

firmware must just be aids

uacpi handles it fine

[hazy saddle]

wait

is it acpi 1.0?

[rotund furnace]

nope

[hazy saddle]

no i do check for that too

leme find my USB

why tf do i use it as a fidget toy

btw what happens if you manualy craft a USB @rotund furnace

with the old method

[rotund furnace]

idk

cba tbh

[hazy saddle]

Yeah @rotund furnace The ISO works fine

The firmware is just shit it seems

@rotund furnace I wonder what UACPI does differently for table parsing?

[rotund furnace]

why are you assuming the xsdt's pointers are aligned

[hazy saddle]

@rotund furnace What happens if you skip the table parsing?

[rotund furnace]

they aren't

[hazy saddle]

Wait what?

[rotund furnace]

they are at byte 36 they can't be aligned

[hazy saddle]

Where

[rotund furnace]

vs uacpi

[hazy saddle]

it very much is fine here

[rotund furnace]

hm

wtf

[hazy saddle]

And if that was bad QEMU would fuck up

[rotund furnace]

yeah wasn't that

hm

wtf

[hazy saddle]

i do not check the checksums tho

[rotund furnace]

well I found ya problem

the first entry is at 75063000 but limine only marks 75049000-7504a000 and 7504b000-7505a000 as acpi tables

the next entry is at 75068000

sadly I cannot debug further as I must slumber

[hazy saddle]

Rare limine bug?

I assume ųACPI maps it its self

[rotund furnace]

Yes

That’s why you give it vmm alloc and free

[rotund furnace]

Not sure

[hazy saddle]

I could probably change it to check the length of tables and map them its self?

[rotund furnace]

ų ≠ μ

[rotund furnace]

Idk

[hazy saddle]

My phone didn’t have the right symbol 😭

[hazy saddle]

I mean no that would work

But like it feels like an issue if limine isn’t detecting it right?

Guaranteed

I guess the protocol was violated

unless this is bad

[rotund furnace]

Hm idk

This is far above me

[hazy saddle]

And me

It’s a firmware or some other fuckup

Limine does have some more recent commits about ACPI it seems?

[gentle void]

why are you mapping the memory regardless of its type

you do not want to map acpi tables as RW

[hazy saddle]

I can change that and should

But it does at least work technically

[gentle void]

also, have you tried booting any other os with limine on that?

[hazy saddle]

It is @rotund furnace S laptop

Thier kernel boots but uses uacpi for tables

Which I assume traverses memory differently

Eg mapping the tables its self vs just using it like this

With HHDM

[gentle void]

that's only meant to access static tables

[hazy saddle]

I am only touching static tables

I don’t do any AML

Just FADT HPET and whatever the one is for APICs

And it works on every other device

[gentle void]

is this an old device

[hazy saddle]

I think so

That’s why I suspect shit firmware or some such

[hazy saddle]

Being my 5700x desktop a steamdeck a laptop with a 12gen i5 and a pre zen amd APU laptop

[marble ridge]

what's the full memory map?

[hazy saddle]

I think water fell asleep. I have a MMAP command in my kernel “shell” that can print it if they disable the ACPI parsing or just pull up another kernel that can provide the limine one

[mild glacier]

iso file :>
yipeee

and i got a friendly welcoming panic 💀

[mild glacier]

cute lil beggar :>

[mild glacier]

kernel shell?

[hollow crater]

lol glad u backed down from the super mean one

[hazy saddle]

Can you take a picture and show me the crash log?

No picture no crash

[hazy saddle]

Shut up 😭 . I can make a userspace one in a day or less but I have way bigger structural fish to fry

[rotund furnace]

mb I fell asleep but it only crashes on real hw so it's gonna be annoying as fuck to copy

give me like

a bit

[hazy saddle]

Comment out the ACPI line

Run MMAP

Take a photo

Il copy it for you

I have nothing better to do rn

[rotund furnace]

lemme try the limine test kernel rq

cuz I wonder

oh

it doesn't really parse shit

[hazy saddle]

It would atleast print out the memory map right

[rotund furnace]

figured out the problem

you're code is cooked

also the limine memory map on this system is cursed as fuck

but that shouldn't be an issue

some how this causes it to miss the page

[hazy saddle]

Damn

[hazy saddle]

How cursed we talking

[hazy saddle]

Why does it align up

[rotund furnace]

it prints so much 😭

[hazy saddle]

Shouldn’t it be align down for the base?

[rotund furnace]

RESERVED
ACPI_TABLES
RESERVED
ACPI_TABLES
RESERVED
ACPI_TABLES
RESERVED
ACPI_TABLES
RESERVED
ACPI_TABLES
RESERVED
ACPI_TABLES
RESERVED
ACPI_TABLES
[... repeat ALOT...]

[hazy saddle]

Does that file have an align down macro

[hazy saddle]

🥀

That’s fucking corded

Cursed

[rotund furnace]

here is what I do

[hazy saddle]

ngl i blame the firmware here

" I have a firm belief that most firmware developers are not actually humans, but are instead caged rodents fed a solid diet of crack cocaine. Because that would explain a lot."

whatever the fuck that shit is crazy

[hazy saddle]

whats your align down macro?

should just be masking the 0xFFF bits nvm

[rotund furnace]

I'll send a patch in a bit

wait

lemme test it first

[hazy saddle]

@rotund furnace try this

[rotund furnace]

still dies

hmmm

[hazy saddle]

that iso or your patch?

[rotund furnace]

that iso

[hazy saddle]

it is being properly made now tho

@rotund furnace that memorymap is still so cursed as fuck

leme test sold older limine versions

version 10.0.0 fully locked up?

1.3.0 boots

leme make an iso with it

wait idk if that works for that

[rotund furnace]

what the fuck

every time

I write this iso to a usb

[hazy saddle]

huh

[rotund furnace]

it happens at 875mb/s

and doesn't do shit

linux!!!

[hazy saddle]

💀

run sync

sync

[rotund furnace]

I always use sync

[hazy saddle]

tf

make a manual UEFI usb with this

[rotund furnace]

okay

so I just wrote your os

plug in the drive

JUST KIDDING

it booted mine

what the fuck

[hazy saddle]

https://tenor.com/view/skull-skull-skull-skull-skull-skull-skull-skull-gif-14574828933741070167

[hazy saddle]

this has an older limine version

btw

[hazy saddle]

i wonder if limine should condence shit like this

depending on how large they are

[rotund furnace]

ehhh

maybe

that's for mint to decide

anyway

[rotund furnace]

how the fuck do I build this myself

[hazy saddle]

it has pros and cons at the very least

[rotund furnace]

cuz I gotta test something

[hazy saddle]

fornat usb as fat32

mark as esp/boot

unzip it on there

so it looks like this

[rotund furnace]

HUH????

[hazy saddle]

wuh

[hazy saddle]

use gparted

make new parition table

gpt

its formated as a single iso458oo3u8o parition or whatever

[rotund furnace]

I don't know how do use gparted

nor do I have it installed

[hazy saddle]

its easy af

[rotund furnace]

istg

linux keeps fucking being stupid with usb drives

every fucking time

[reboot]

[hazy saddle]

when i get USB il make sure it works well 🙏

i also plan a hypervisor at some point

With IOMMU support of I can help it

So even if my kernel is fucking shit you can atleast game on it with GPU passthrough VMs

[rotund furnace]

bro

istg

[hazy saddle]

What

[rotund furnace]

WHAT?

the partition table just came back from the dead

[hazy saddle]

💀

[rotund furnace]

okay

now fdisk worked

[hazy saddle]

Your firmware may just be shit

[rotund furnace]

what the fuck

doesn't happen on windows btw

and now dd works

sometimes

linux just stops syncing to usb sticks

IT'S ALIVE

[hollow crater]

did u have an updated kernel in place waiting for reboot o_O

[rotund furnace]

nope

[hazy saddle]

https://tenor.com/view/its-alive-its-moving-mr-burns-simpsons-frankenstein-gif-19716003

[hollow crater]

weird

[rotund furnace]

lemme make a patch rq

[hazy saddle]

Did it actually work though

[rotund furnace]

yes

[hazy saddle]

Damn

Was that my ISO?

[rotund furnace]

I gotta disable the idisplay and capture card it via hdmi

[hazy saddle]

What other changes did you make

[rotund furnace]

no

lemme remove my debugging shit and I'll send a patch

[hazy saddle]

Is it a patch over my patch?

[rotund furnace]

yes

[hazy saddle]

Record a video of the MMAP

I wana see how cursed the memory map is

[rotund furnace]

i lwk just kinda ported my vmm init code

and it worked

BRO

WHY IS THE SCREEN

SO DAMN GLOSSY

ts pmo

hold on

[hazy saddle]

I wonder if I can get a minimal sub set of it going

For just the bug

Otherwise it may be better for me to commit

And then for you to make a PR with the improvements

[rotund furnace]

there

[hazy saddle]

@rotund furnace okay try this

sorry for doubting you mint

[rotund furnace]

yep

& badapple works

[hazy saddle]

Zamm

[rotund furnace]

okay

[hazy saddle]

How is doom

[rotund furnace]

do not run doom and badapple at the same

[hazy saddle]

Wdym

[rotund furnace]

I had both doom

and badapple

running at the same time

on top of eachother

but yeah doom did work

[hazy saddle]

Get me that memory map

[rotund furnace]

ugh

[hazy saddle]

I wana see that cursed shit in all its glory

[rotund furnace]

LMAO

your limine memmap to string function doesn't work

for ACPI_TABLES

[hazy saddle]

Damn

Let me fix that

[rotund furnace]

not shown are the entries that go off the screen 😭

[hazy saddle]

how many go off?

[rotund furnace]

only like

4-5

[hazy saddle]

qemu memory map for reference

[rotund furnace]

that are uninteresting

lenovo sucks at firmware

let's just shove all acpi tables

1 page apart

BUT

also kinda nice

[hazy saddle]

that firmware looks like they emebeded the crack inside the the fucking bios chip

[rotund furnace]

because in theory you can guard page in between

¯_(ツ)_/¯

[hazy saddle]

this one has proper printing

[hazy saddle]

Still just make it normal memory

It’s reclaimable anyways

Just gave it be zeroed memory but not show it in that fuck ass map

[rotund furnace]

no point

it's all gonna be acpi tables anyway

[rotund furnace]

yeah

oh well

[hazy saddle]

That way if somthing goes wrong you get null pointers

Instead of whatever the fuck that shit is

[rotund furnace]

if mint wants to do what all os devs eventually have to do

which is clean up after the firmware guys

then sure

[hazy saddle]

i mean this isnt too bad to clean up after to merge repeated sections like that

also i was refering to the firmware devs

not mint

[rotund furnace]

oh my god

there are gaps in the memory map

[hazy saddle]

what

[rotund furnace]

of like 4 bytes

[hazy saddle]

😭

[rotund furnace]

it goes

START, LEN OF TABLE
START + LEN OF TABLE, SOME ADDRESS ABOVE - SOME VALUE
SOME ADDRESS ABOVE - SOME VALUE BUT PAGE ALIGNED, LEN OF TABLE

[hazy saddle]

https://youtu.be/L3G4NyelU2Y

[rotund furnace]

the lenovo firmware devs doing lines before writing code istg

[hazy saddle]

Maybe if memory map regions were garunteed to be page aligned on limine?

Just to stop that fucking shit

[rotund furnace]

nope

[hazy saddle]

Wdym?

[rotund furnace]

because what if two memory map regions cross a page boundary

[hazy saddle]

Have a mixed type

[rotund furnace]

shit would break so quick

[rotund furnace]

NAH

PLEASE NO

WHAT

[hazy saddle]

And then for that type there is a pointer to get the sub types

Sorry I’m thinking like a Lenovo firmware dev

[rotund furnace]

ohhh

that makes alot of sense

[hazy saddle]

Yeah yeah

That way for HHDM you can just go broadly

But if you need to be precise you have the sub mappings

And you don’t have stupid page issues

And you have less messes of that in the main memory map

[rotund furnace]

eh

maybe

[hazy saddle]

Or have it be an option on the request for page aligned?

That way you can ask for it if it dosnt matter much

And give priorities to certain types

[rotund furnace]

who knows bro

[hazy saddle]

Nobody

[rotund furnace]

that's for smarter people to decide

[hazy saddle]

Certainly not the people who made this though

[mild glacier]

, hold on

is this a "500 quadrillion bytes off" error again 😭

[hazy saddle]

try this ISO

also in QEMU??

fuck kinda dlags are you using

i dont even use QEMU with the iso

[mild glacier]

only -m 4G
i think 4GB RAM should be perfectly fine

[rotund furnace]

qemu with iso works fine for me

[mild glacier]

vbox's .deb files dont install coz something is missing, too old or too new :/

[hazy saddle]

    qemu-system-x86_64 \
        -machine q35,accel=kvm,smm=on \
        -cpu host,+x2apic,+invtsc \
        -m 512M \
        -cdrom ./evalynOS.iso \
        -boot d \
        -audiodev pa,id=speaker -machine pcspk-audiodev=speaker \
        -serial stdio

use that

[mild glacier]

perplexing

[hazy saddle]

it works for me

[mild glacier]

🤦‍♂️

[hazy saddle]

though for some reason that breaks acpi in qemu

@rotund furnace

[mild glacier]

holy shit backspace is so close to enter 😭

[rotund furnace]

WHAT?

it works for me

OJKHASWFHJKLO AFHNJLK AFSKHJ

[hazy saddle]

i have to use OVMF

otherwise no ACPI

[rotund furnace]

revision 0?

hm

[hazy saddle]

do you have ovmf on?

[rotund furnace]

yeah

[hazy saddle]

    qemu-system-x86_64 \
        -machine q35,accel=kvm,smm=on \
        -cpu host,+x2apic,+invtsc \
        -m 512M \
        -drive if=pflash,format=raw,readonly=on,file=./OVMF_CODE.4m.fd \
        -drive if=pflash,format=raw,readonly=on,file=./OVMF_VARS.4m.fd \
        -cdrom ./evalynOS.iso \
        -boot d \
        -audiodev pa,id=speaker -machine pcspk-audiodev=speaker \
        -serial stdio

yeah i diddnt before

[hazy saddle]

[rotund furnace]

oh lmao

[hazy saddle]

so no shit theres no ACPI when there is no firmware

[rotund furnace]

seabios my guy?

[hazy saddle]

🤷‍♀️

[rotund furnace]

yk

I have no idea where the acpi tables are on a real pc

or in qemu

are they in the bios rom?

I assume so

[hazy saddle]

ig seabios yeah

[rotund furnace]

in that case it's qemu's job

[hazy saddle]

ig they dont provide acpi?

this is a lil ??

[rotund furnace]

wdym

[hazy saddle]

?????

what revison do you get

[rotund furnace]

uacpi info RSDP 0x00000000000F6500 00000014 v00 (BOCHS )

[hazy saddle]

maybe its a qemu thing?

    qemu-system-x86_64 \
        -machine q35,accel=kvm,smm=on \
        -cpu host,+x2apic,+invtsc \
        -m 512M \
        -cdrom ./evalynOS.iso \
        -boot d \
        -audiodev pa,id=speaker -machine pcspk-audiodev=speaker \
        -serial stdio

try this with my iso?

[rotund furnace]

no

you are just being stupid

acpi 1.0 is rev 0

and you just ignore it

[hazy saddle]

i dont have acpi 1.0 support

[rotund furnace]

exactly

seabios only gives you acpi 1.0

[hazy saddle]

yeah im dumb

[rotund furnace]

no rsdt support?

damn

[hazy saddle]

isnt like every x86-64 computer acpi2 anyways

[rotund furnace]

just support it tho

it's like

[hazy saddle]

the spec came out before x86-64

[rotund furnace]

no extra work tbh

[hazy saddle]

true

just some fucking extra if elses

but like

Advanced Configuration and Power Interface Specification Version 2.0 – Published Jul. 27, 2000 (1.5 MB)

CMON

@rotund furnace if sombody finds a machine thats x86-64. with only acpi 1.0. in real life

i will bake them a fucking pie

and ship it to them

and il fix it myself

else talk cheap send patches

[rotund furnace]

fuck you

because now you know I gotta fix it

[hazy saddle]

XD

[rotund furnace]

how the fuck did you even boot non uefi

[hazy saddle]

I removed the OVMF lines

Everyone has ACPI 2.0+ dipshit, it came free with your fucking x86-64 computer.

I didn't get it, I have the oldest x86-64 known to man.

[rotund furnace]

yeah but removing them makes the cdrom not boot

[hazy saddle]

What

It works for me

Do you have limine installed

[rotund furnace]

tcg:
    ./src/build-scripts/get-deps
    ./src/build-scripts/apply-patches.sh
    make clean
    ./src/build-scripts/generate-all.sh
    make all -j${nproc}
    ./src/build-scripts/undo-patches.sh
    cp ./bin-x86_64/kernel.elf ./src/generated/iso/kernel.elf
    ./src/build-scripts/generate-iso.sh
    qemu-system-x86_64 \
        -machine q35,accel=tcg \
        -m 512M \
        -drive format=raw,file=fat:rw:./src/generated/iso \
        -boot d \
        -audiodev pa,id=speaker -machine pcspk-audiodev=speaker \
        -serial stdio```

[hazy saddle]

So it can do the BIOS pass on the iso

[hazy saddle]

Your not including the CDrom

[hazy saddle]

This

That method turns a folder into a virtual partition

Which is fancy

[rotund furnace]

fuck you

talk is very cheap

[hazy saddle]

Thank you kindly

[rotund furnace]

I'm putting open source contributor on my cv now

[hazy saddle]

If you need my phone number I can be your reference

[rotund furnace]

sick

[marble ridge]

did you people figure out the issue?

[mild glacier]

prolly less than a calorie is wasted when you say "hello world"
so i guess talk is very cheap

[rotund furnace]

yeah

they were mapping the hhdm wrong by ALIGN_UP the base and ALIGN_DOWN the size, instead of ALIGN_DOWN the base, ALIGN_UP the size

[rotund furnace]

,

[marble ridge]

welp

i guess it wasn't a rare Limine bug after all :p

[hazy saddle]

The memory map was fucking cursed though

[marble ridge]

well that's "intended" behaviour

that firmware must be storing ACPI tables in a reserved entry, so when Limine converts it to ACPI tables entry, it splits it up like that, if the tables aren't contiguous

[hazy saddle]

Oh no that’s fair. Just a lil cursed so no wonder it broke a few things.

Sorry for wasting your time

[marble ridge]

no worries lol

i mean technically-correct-but-cursed edge cases are good for sorting bugs out

[hazy saddle]

I wonder if there is a better way to handle that in the bootloader to make it look less like that.

But yeah they are really good for that

https://tenor.com/view/correct-futurama-the-best-kind-of-correct-yes-yep-gif-5787390

[marble ridge]

there are options

the main issue is that the firmware decided to store ACPI tables data in a reserved region

most modern UEFI firmware will store those in an ACPI reclaimable region

where then this does not happen as the bootloader won't need to convert any types

[hazy saddle]

What is a reserved region? I assumed it was invalid region

[rotund furnace]

doesn't have to be I assume

[marble ridge]

it depends, it can literally be anything

[rotund furnace]

there is LIMINE_MEMMAP_BAD_MEMORY

which I assume is for "invalid" memory

(nothing is gonna stop you from reading/writing to it tho)

[marble ridge]

from the kernel's perspective one might as well treat reserved and bad memory the same

[rotund furnace]

reserved memory also contains shit like mmio

sometimes

[marble ridge]

that being the same as memory holes not covered by any entry altogether

[marble ridge]

yeah

but not always

often mmio is not covered by any entries at all

[rotund furnace]

yeah

which I guess is reserved

I treat it all the same (aka don't map it)

[hazy saddle]

So basically it’s just old firmware jank there isn’t a clean way to work arround

[marble ridge]

for now yes

you can assume that

[rotund furnace]

and mmio is mapped itself

[hazy saddle]

So basically what happened is there is a large reserved region the firmware has. But also puts the ACPI tables there

I wonder if overlapping memory map regions could help?

[marble ridge]

Limine kind of avoids creating overlapping regions at all, they are only passed if the firmware itself creates them, and only sometimes

another solution would be to pad out ACPI tables regions to 4k for base and length and then hope that the regions will coalesce

another option is to use base revision 3 with doesn't have this problem at all but then you need to map ACPI tables yourself :p

[hazy saddle]

Are ACPI table memory mappings made by firmware or limine its self

[marble ridge]

ACPI table entries are a base revision 4 meme

made by Limine if the firmware doesn't already cover the tables in the other 2 ACPI regions

[hazy saddle]

Wouldn’t it make more sense for them to still be HHDMd but their memory entitled to be part of the RDSP request and allow them to overlap with the other part potentially but have them be page aligned!

[marble ridge]

idk i'll think about a solution, if anything, and let you know

[rotund furnace]

🔥

[marble ridge]

feel free to ping me with ideas if you get any

[hazy saddle]

Based as always

[hazy saddle]

Will do. There is probably a clean solution somewhere. Maybe an opt in to overlapping regions for certain things or some other? Or having ACPI table regions being a special case of overlapping and having them always page aligned

[mild glacier]

add a :3 emoji
pls i need this

[rotund furnace]

wait

how the fuck does limine map the rsdt entries in the hhdm

does it also ident map

[mild glacier]

https://tenor.com/view/blinking-eyes-white-guy-blinking-meme-what-huh-gif-26334322

i only understood "limine" 😭

[rotund furnace]

because they are uint32_t's

oh I'm stupid and misread the protocol

whoops

[hazy saddle]

[gentle void]

wdym how

[rotund furnace]

no I was tripping

I misread the spec

[hazy saddle]

@mild glacier this ISO should work

@rotund furnace Still seems to be working on deck

Although the first attempt it almost gave me a seizure because the OLED panel freaked the fuck out

[rotund furnace]

😭

[hazy saddle]

It did not happen after that

I think it’s a thing the steamdeck can just kinda do 🥀

With like slightly buggy OS stuff even on Linux

If it happens again il try and take a video

[mild glacier]

what are the commands 🙏

help no work ;-;

[hazy saddle]

maybe read the documentation

and in this world there is no help

bad apple is BADAPPLE and doom is DOOM

[mild glacier]

the BADAPPLE extra low tones literally hurt my ears

noted: stacking 3x BADAPPLE causes a page fault

[rotund furnace]

run badapple and doom at the same time

🔥

[hazy saddle]

I am aware of this lmao

It’s because it renders off the frame buffer

Which is?

[mild glacier]

yea...

[hazy saddle]

Invalid memory

It’s a page fault

I just don’t handle it yet

[mild glacier]

bum 💥

[hazy saddle]

@rotund furnace Shit I need to commit soon

I think the code is all good clean and ready tho right?

Also @mild glacier if you get a bigger monitor three will work just fine as on this laptop

[mild glacier]

what res?

[hazy saddle]

The iso will default you to 720p

[mild glacier]

oh

well that's some full terms and conditions (in hindi) :/

to be clear

[hazy saddle]

There is a reason it does that

Because I use 1080p monitors

And I don’t want my QEMU taking up the entire screen

So it defaults to 720p

[mild glacier]

ohh ok got it

[hazy saddle]

looks good to merge right @rotund furnace

[hazy saddle]

i think im going to do a beta1-rc1 release once i redo the PMM Heap and VMM stuff and get a user shell without janky systemcalls

may not be a CoW though]

it may or may not have the disk drivers though

that may come on beta2

and rc1 will be removed depending on bugs found

[hazy saddle]

timers are all up and pushed

ISO for this commit

May publish this as a release on my git server but it’s not worth it

Plus I have nitro so I can abuse discord as a file server

[hazy saddle]

better idea i can just archive them on my file server
https://files.evalyngoemer.com/share/evalynOS Builds/

dosnt make sense to have them on git since its not "nightlys" but just archives

[mild glacier]

how do you make the sound tho
i neeed to take notes

[hazy saddle]

wdym

[mild glacier]

how does EvalynOS create audio
the beeps and boops of Bad Apple

[hazy saddle]

read source code

[mild glacier]

🤦‍♂️

agh

[hazy saddle]

https://tenor.com/view/hannibal-buress-right-booing-gif-13861878

I use the PC speaker

The driver is really simple

[soft snow]

it is.

[hazy saddle]

And you can see how the things gets called

The source code is really simple and self documenting for the most part

[mild glacier]

not booing tho 🔥

[hazy saddle]

You are making a mistake 🔥

What command are you using

Wait miss read

[mild glacier]

uhh im lost in the /src directory 😭

nvm

nvm im actually still lost, this time in /src/kernel/drivers

[hazy saddle]

/src/kernel/drivers/x86_64/pcspkr.c

https://tenor.com/view/madagascar-skipper-can-you-read-gif-22646390

[mild glacier]

❌

well uhh you have pcskpr.h and pcspkr.c

[hazy saddle]

Yes

[mild glacier]

skeaper

[hazy saddle]

Those are the functions use to drive the pc speakers

[mild glacier]

but uh reread that .h one

[hazy saddle]

It dosnt mater

Who cares

[mild glacier]

🤷

[hazy saddle]

Il do a pass of spelling mistakes later

But it’s just internal code names

It dosnt matter

There are way more spelling mistakes elsewhere e