pmOS - microkernel OS for RISC-V, x86, and LoongArch | Operating System Development | Page 4

carmine nacelle Dec 23, 2024, 4:08 PM

#

In what way

#

I'm using old uACPI

hoary moat Dec 23, 2024, 4:09 PM

#

try to boot into it

#

nothing gets printed

carmine nacelle Dec 23, 2024, 4:10 PM

#

Thonk

#

That may be kernel panic

hoary moat Dec 23, 2024, 4:10 PM

#

I'd expect logs

carmine nacelle Dec 23, 2024, 4:10 PM

#

Or broken uACPI

hoary moat Dec 23, 2024, 4:10 PM

#

But alas

carmine nacelle Dec 23, 2024, 4:10 PM

#

My kernel doesn't have framebuffer driver

#

Try this

📎 pmOS.iso

hoary moat Dec 23, 2024, 4:11 PM

#

In a bit

carmine nacelle Dec 23, 2024, 4:11 PM

#

It's master uACPI

hoary moat Dec 23, 2024, 4:12 PM

#

carmine nacelle It's master uACPI

What is the expected behavior when it boot successful

carmine nacelle Dec 23, 2024, 4:13 PM

#

You see a bunch of stuff on screen

hoary moat Dec 23, 2024, 4:13 PM

#

Ok

carmine nacelle Dec 23, 2024, 4:13 PM

#

Since that's done from userspace

#

Does your PC have PIC?

hoary moat Dec 23, 2024, 4:18 PM

#

probably emulated

#

pic

carmine nacelle Dec 23, 2024, 4:19 PM

#

Emulated should be fine

#

Wait

#

PIT*

hoary moat Dec 23, 2024, 4:19 PM

#

emulated probably

#

still don't see nothin

carmine nacelle Dec 23, 2024, 4:20 PM

#

Ok

carmine nacelle Dec 23, 2024, 10:20 PM

#

How sad https://h30434.www3.hp.com/t5/Notebook-Hardware-and-Upgrade-Questions/HP-SPECTRE-X360-13-AE013DX-won-t-turn-on-or-charge/td-p/8646626

community.hp.com

HP SPECTRE X360 13-AE013DX won't turn on or charge

Hello. I purchased the $1500 HP SPECTRE X360 13-AE013DX in April 2018 for college. It is less than four years later, and the laptop will not turn on or charge. The power light flashes for a second, then immediately goes off. When I plug a charger into the USB C ports, the charging light illuminates....

#

I have the exact same problem

#

carmine nacelle Dec 24, 2024, 1:20 PM

#

Ok I think I'm inspired to work on memory objects now...

carmine nacelle Dec 25, 2024, 8:59 PM

#

What a fun discovery: my pthread_mutex_unblock yielded at all function calls

carmine nacelle Dec 25, 2024, 10:00 PM

#

And apparently munmap was doing nothing galaxybrain

carmine nacelle Dec 25, 2024, 10:49 PM

#

A few bug fixes later

#

Everything is an object now

#

And I think I should be able to implement CoW on fork of everything quite trivially

#

But I'm feeling too tired so that's probably for tomorrow

#

(also, uACPI score increased galaxybrain )

twilit talon Dec 25, 2024, 11:01 PM

#

by how much

left dew Dec 25, 2024, 11:04 PM

#

should spinlock_lock disable irqs?

#

oh wow ☠️ wrong thread

twilit talon Dec 25, 2024, 11:05 PM

#

lol

left dew Dec 25, 2024, 11:05 PM

#

i was reading random threads and coding at the same time and didnt realize im in the pmos thread my bad

carmine nacelle Dec 25, 2024, 11:07 PM

#

twilit talon by how much

It's gotten to 800k+, after I ran it once

#

I want to try it on my destop though

twilit talon Dec 25, 2024, 11:09 PM

#

cool cool

carmine nacelle Dec 25, 2024, 11:12 PM

#

I want to have AHCI reading though

carmine nacelle Dec 26, 2024, 4:22 PM

#

After three days of searching, I've found the charger for the asus laptop

#

And it doesn't post

#

Nice

#

Events are working

#

@twilit talon Why would uACPI eval of an even fail with not found?

twilit talon Dec 26, 2024, 4:30 PM

#

no such event

carmine nacelle Dec 26, 2024, 4:30 PM

#

Of _Q<some shit>

twilit talon Dec 26, 2024, 4:30 PM

#

Yeah

carmine nacelle Dec 26, 2024, 4:30 PM

#

Broken firmware?

twilit talon Dec 26, 2024, 4:30 PM

#

grep your blob for _Q<IDK>

twilit talon Dec 26, 2024, 4:31 PM

#

carmine nacelle Broken firmware?

probably outdated bios generating garbage

carmine nacelle Dec 26, 2024, 4:31 PM

#

There's no new bios

twilit talon Dec 26, 2024, 4:32 PM

#

all hope is lost then PoroSad

carmine nacelle Dec 26, 2024, 4:32 PM

#

twilit talon Dec 26, 2024, 4:32 PM

#

waybackmachine

#

or fire up your favorite patched linux lmao

#

to see if its some sort of a skill issue

carmine nacelle Dec 26, 2024, 4:33 PM

#

It reboots when uACPI tries to turn it off btw

twilit talon Dec 26, 2024, 4:33 PM

#

it or your os meme

#

managarm test ig

carmine nacelle Dec 26, 2024, 4:34 PM

#

Hmm, it just turned off

twilit talon Dec 26, 2024, 4:34 PM

#

probably triple faults in some cases

carmine nacelle Dec 26, 2024, 4:34 PM

#

Why would it

twilit talon Dec 26, 2024, 4:34 PM

#

managarm test needed

twilit talon Dec 26, 2024, 4:34 PM

#

carmine nacelle Why would it

bug

#

in your kapi perhaps

#

or some race

carmine nacelle Dec 26, 2024, 4:35 PM

#

I have double fault handler

#

And uACPI runs in userspace

twilit talon Dec 26, 2024, 4:35 PM

#

well we'll never know until u debug it further

carmine nacelle Dec 26, 2024, 4:36 PM

#

pmOS doesn't tripple fault ultrameme

twilit talon Dec 26, 2024, 4:36 PM

#

got a dump btw

#

also figure out the event idx

#

and test managarm

carmine nacelle Dec 26, 2024, 4:36 PM

#

It's refusing to post after turning off 💀

twilit talon Dec 26, 2024, 4:36 PM

#

lol

#

maybe its just not feeling well

carmine nacelle Dec 26, 2024, 4:37 PM

#

Could be

twilit talon Dec 26, 2024, 4:37 PM

#

is it possible the battery is drained faster than its recharged

carmine nacelle Dec 26, 2024, 4:37 PM

#

Its power button is broken

#

Meybe it turned back on because it got pressed

twilit talon Dec 26, 2024, 4:37 PM

#

lol

carmine nacelle Dec 26, 2024, 4:38 PM

#

This laptop has terrible design flaw

left dew Dec 26, 2024, 4:38 PM

#

it's firmware was written by proper idiots?

carmine nacelle Dec 26, 2024, 4:38 PM

#

I mean Windows and Linux couldn't turn it off in the past

twilit talon Dec 26, 2024, 4:39 PM

#

EC generating non existant events a new one

carmine nacelle Dec 26, 2024, 4:39 PM

#

It would just hang

twilit talon Dec 26, 2024, 4:39 PM

#

every new laptop uncovers more retardness in firmware

carmine nacelle Dec 26, 2024, 4:39 PM

#

I'll try to boot it to fedora

#

It has gentoo I think

twilit talon Dec 26, 2024, 4:39 PM

#

carmine nacelle I mean Windows and Linux couldn't turn it off in the past

uacpi W

carmine nacelle Dec 26, 2024, 4:39 PM

#

But it would take a few days to update with its 1GHz CPU

#

AMD C-60

twilit talon Dec 26, 2024, 4:40 PM

#

u would think they'd all have similar blobs

carmine nacelle Dec 26, 2024, 4:40 PM

#

Retarded laptop with retarded cpu

twilit talon Dec 26, 2024, 4:40 PM

#

but every time i uncover new shit

#

like while with 50 or statements

carmine nacelle Dec 26, 2024, 4:40 PM

#

carmine nacelle I mean Windows and Linux couldn't turn it off in the past

When booting with UEFI

#

With BIOS it worked fine

twilit talon Dec 26, 2024, 4:41 PM

#

try uefi boot and shutdown with uacpi

#

wonder what the error is

carmine nacelle Dec 26, 2024, 4:41 PM

#

carmine nacelle Dec 26, 2024, 4:41 PM

#

twilit talon try uefi boot and shutdown with uacpi

No it did shutdown

#

And linux also started to work

#

And I'm not gonna test windows since it would take several hours to install

twilit talon Dec 26, 2024, 4:41 PM

#

maybe it just needed new enough updates

carmine nacelle Dec 26, 2024, 4:42 PM

#

Windows 10 that is

twilit talon Dec 26, 2024, 4:42 PM

#

maybe it requires ec to shutdown

carmine nacelle Dec 26, 2024, 4:42 PM

#

It's refusing to post again

left dew Dec 26, 2024, 4:42 PM

#

twilit talon maybe it requires ec to shutdown

.

#

maybe lack of an ec driver makes it "shut down"

#

but it's not actually shut down

carmine nacelle Dec 26, 2024, 4:43 PM

#

I do have EC driver

left dew Dec 26, 2024, 4:43 PM

#

oh

twilit talon Dec 26, 2024, 4:43 PM

#

and it does shutdown in pmos

left dew Dec 26, 2024, 4:43 PM

#

but it refuses to post

twilit talon Dec 26, 2024, 4:43 PM

#

but not on old linuxes apparently?

carmine nacelle Dec 26, 2024, 4:43 PM

#

And I'm only hobby OS with AMD GPIO driver sunglas

left dew Dec 26, 2024, 4:43 PM

#

also, when you press the power button after shut down does it light up or is like dead lol

twilit talon Dec 26, 2024, 4:43 PM

#

left dew but it refuses to post

apparently thats the power button being broken?

carmine nacelle Dec 26, 2024, 4:44 PM

#

There's a second button to boot it into the quick boot proprietary linux shit memes

#

Which works

twilit talon Dec 26, 2024, 4:44 PM

#

wtf is that button

carmine nacelle Dec 26, 2024, 4:44 PM

#

And it also turns on when I plug the charger in

twilit talon Dec 26, 2024, 4:45 PM

#

least insane laptop

carmine nacelle Dec 26, 2024, 4:45 PM

#

carmine nacelle There's a second button to boot it into the quick boot proprietary linux shit me...

It's very typical for 2010s laptops

left dew Dec 26, 2024, 4:45 PM

#

i mean it has a 1ghz amd processor nobody has ever heard of

carmine nacelle Dec 26, 2024, 4:45 PM

#

And 2000s

twilit talon Dec 26, 2024, 4:45 PM

#

whats the purpose of that proprietary linux

carmine nacelle Dec 26, 2024, 4:45 PM

#

quick boot

left dew Dec 26, 2024, 4:46 PM

#

Wha

twilit talon Dec 26, 2024, 4:46 PM

#

actual insanity

carmine nacelle Dec 26, 2024, 4:46 PM

#

There was a series by some yt guy about it

#

I have 2 or 3 laptops which have that

twilit talon Dec 26, 2024, 4:47 PM

#

i think i remember watching something like that

#

was it michael

carmine nacelle Dec 26, 2024, 4:47 PM

#

Yes

twilit talon Dec 26, 2024, 4:47 PM

#

but i dont remember exactly what it was about

carmine nacelle Dec 26, 2024, 4:47 PM

#

There was some insanity where some laptop manufacturers would even put DVD drive SoCs into laptops

#

And DELL put a second ARM CPU

#

For fast boot

twilit talon Dec 26, 2024, 4:48 PM

#

wild early 2000s

carmine nacelle Dec 26, 2024, 4:48 PM

#

But it's usually just some Linux distro

hoary moat Dec 26, 2024, 4:48 PM

#

cursed

carmine nacelle Dec 26, 2024, 4:48 PM

#

And a separate button

carmine nacelle Dec 26, 2024, 4:48 PM

#

twilit talon was it michael

what's his chanel name?

twilit talon Dec 26, 2024, 4:49 PM

#

Michael MJD

carmine nacelle Dec 26, 2024, 4:49 PM

#

https://www.youtube.com/watch?v=ssob-7sGVWs

YouTube

Cathode Ray Dude - CRD

Quick Start, Episode 7: Crimes... And Felonies.

HP took a look at all the other Quick Start products and said "oh, we can do worse." They did.

Support me on Patreon: https://www.patreon.com/cathoderaydude
Tip me: https://ko-fi.com/cathoderaydude

Chapters:
00:00 Quickweb recap
04:02 Feature 1: QuickLook
09:15 QuickLook under the hood
15:30 What's an EFI application?
29:15 QuickLook in summa...

▶ Play video

#

Not Michael apparently

twilit talon Dec 26, 2024, 4:49 PM

#

wait i have this exact laptop

#

does it have qb

#

waait it can do that lol??

carmine nacelle Dec 26, 2024, 4:51 PM

#

He has a whole series on it

#

Different laptop manufacturers did different wild shit

#

Wild and insane

twilit talon Dec 26, 2024, 4:51 PM

#

ive tested very early uacpi on this laptop

#

worked fine iirc

carmine nacelle Dec 26, 2024, 4:52 PM

#

No but that's different laptop from mine

twilit talon Dec 26, 2024, 4:52 PM

#

yeah ik

#

im just surprised i have it

#

this laptop has a very bad uefi impl

#

without a GOP

carmine nacelle Dec 26, 2024, 4:54 PM

#

Bruh

#

My new laptop has no CMS btw

twilit talon Dec 26, 2024, 4:55 PM

#

or rather, it has gop but it only supports Blit() and u dont get a framebuffer

carmine nacelle Dec 26, 2024, 4:55 PM

#

End of an era

twilit talon Dec 26, 2024, 4:55 PM

#

yeah modern laptops L

carmine nacelle Dec 26, 2024, 4:55 PM

#

And no S3

twilit talon Dec 26, 2024, 4:55 PM

#

L

carmine nacelle Dec 26, 2024, 4:57 PM

#

Ok I'm booting fedora before it completely dies

#

Fedora is def liking it

twilit talon Dec 26, 2024, 5:01 PM

#

oh well

carmine nacelle Dec 26, 2024, 5:01 PM

#

It booted

#

ok how do I dump acpi from linux

twilit talon Dec 26, 2024, 5:02 PM

#

acpidump > dump.txt

carmine nacelle Dec 26, 2024, 5:03 PM

#

Ok it's trying to install acpica tools

#

And struggling

#

Idk how I managed to install gentoo on it....

#

📎 dump.txt

#

@twilit talon

twilit talon Dec 26, 2024, 5:10 PM

#

lol

#

now lets find out the query number

carmine nacelle Dec 26, 2024, 5:11 PM

#

What is this

hexed acorn Dec 26, 2024, 5:11 PM

#

twilit talon lol

is MCTH a custom string comparison function? meme

carmine nacelle Dec 26, 2024, 5:12 PM

#

I should take dmesg now that I think of it

twilit talon Dec 26, 2024, 5:12 PM

#

hexed acorn is MCTH a custom string comparison function? <:meme:575445094589661215>

match() yeah

hexed acorn Dec 26, 2024, 5:12 PM

#

doesn't aml let you compare strings natively

twilit talon Dec 26, 2024, 5:12 PM

#

    Method (MCTH, 2, NotSerialized)
    {
        If ((SizeOf (Arg0) < SizeOf (Arg1)))
        {
            Return (Zero)
        }

        Local0 = (SizeOf (Arg0) + One)
        Name (BUF0, Buffer (Local0){})
        Name (BUF1, Buffer (Local0){})
        BUF0 = Arg0
        BUF1 = Arg1
        While (Local0)
        {
            Local0--
            If ((DerefOf (BUF0 [Local0]) != DerefOf (BUF1 [Local0]
                )))
            {
                Return (Zero)
            }
        }

        Return (One)
    }

twilit talon Dec 26, 2024, 5:13 PM

#

hexed acorn doesn't aml let you compare strings natively

well NT fails if lhs type != rhs type, maybe they wanted to compare strings against buffers or something

#

or maybe its firmware dev brain rot

hexed acorn Dec 26, 2024, 5:13 PM

#

and ig Buffer (BUF) {} works?

#

where BUF is a buffer already

#

oh wait no

#

nvm

carmine nacelle Dec 26, 2024, 5:13 PM

#

What if I put OpenVMS (whatever that is) on it? ultrameme

twilit talon Dec 26, 2024, 5:13 PM

#

yeah

hexed acorn Dec 26, 2024, 5:13 PM

#

Local0 is the size

twilit talon Dec 26, 2024, 5:13 PM

#

yup

#

string length + null terminator

#

then memcpy the strings

hexed acorn Dec 26, 2024, 5:14 PM

#

BUF0 = Arg0 is a memcpy of the contents and not a reassignment of BUF0 to the arg?

#

wack

twilit talon Dec 26, 2024, 5:14 PM

#

in acpica it is a reassignment

#

on windows its a memcpy

carmine nacelle Dec 26, 2024, 5:14 PM

#

📎 dmesg.txt

twilit talon Dec 26, 2024, 5:14 PM

#

but they clearly do it for windows because Buffer (Local0){} preallocates the buffer with zeroes with correct size

hoary moat Dec 26, 2024, 5:15 PM

#

technically if AML is insane then doesn't that mean that uACPI is an insane interpreter

#

since it can interpret insane AML

hexed acorn Dec 26, 2024, 5:16 PM

#

hexed acorn wack

i guess this behavior would make sense in a strongly typed language

carmine nacelle Dec 26, 2024, 5:17 PM

#

[    0.032063] ACPI: SSDT 0x00000000A7CB0A98 00168E (v02 AMD    ALIB     00000001 MSFT 04000000)```?

twilit talon Dec 26, 2024, 5:17 PM

#

carmine nacelle

@hoary moat btw this has tons of NVS regions

twilit talon Dec 26, 2024, 5:17 PM

#

carmine nacelle ```[ 0.032039] ACPI: SSDT 0x00000000A7CB0788 00030C (v01 AMD POWERNOW 0000...

what

carmine nacelle Dec 26, 2024, 5:18 PM

#

carmine nacelle ```[ 0.032039] ACPI: SSDT 0x00000000A7CB0788 00030C (v01 AMD POWERNOW 0000...

Two vendors?

hexed acorn Dec 26, 2024, 5:18 PM

#

ALIB = ACPI LIB? meme

carmine nacelle Dec 26, 2024, 5:18 PM

#

twilit talon <@841762704766140486> btw this has tons of NVS regions

What's that

twilit talon Dec 26, 2024, 5:18 PM

#

carmine nacelle Two vendors?

well some are shipped by bios, some are shipped by firmware, some are gpu, etc

#

aml is a patchwork of all components

twilit talon Dec 26, 2024, 5:18 PM

#

hexed acorn ALIB = ACPI LIB? <:meme:575445094589661215>

yeah probably

hoary moat Dec 26, 2024, 5:18 PM

#

carmine nacelle What's that

ACPI non volatile regions

hoary moat Dec 26, 2024, 5:18 PM

#

hoary moat ACPI non volatile regions

basically they should be conserved when going to S4

#

and linux also saves it on S3 suspend

twilit talon Dec 26, 2024, 5:20 PM

#

carmine nacelle What's that

memory that must be preserved during suspend for firmware state

carmine nacelle Dec 26, 2024, 5:24 PM

#

carmine nacelle

what's your verdict? risc

twilit talon Dec 26, 2024, 5:25 PM

#

looks normal

carmine nacelle Dec 26, 2024, 5:26 PM

#

Lol

[   64.734853] watchdog: BUG: soft lockup - CPU#0 stuck for 26s! [(udev-worker):601]
[   64.734980] CPU#0 Utilization every 4s during lockup:
[   64.734987]     #1: 100% system,      0% softirq,      0% hardirq,      0% idle
[   64.734998]     #2: 101% system,      0% softirq,      0% hardirq,      0% idle
[   64.735007]     #3: 100% system,      0% softirq,      0% hardirq,      0% idle
[   64.735015]     #4: 101% system,      0% softirq,      0% hardirq,      0% idle
[   64.735023]     #5: 100% system,      0% softirq,      0% hardirq,      0% idle

twilit talon Dec 26, 2024, 5:26 PM

#

it has an ECDT even damn

carmine nacelle Dec 26, 2024, 5:29 PM

#

What's RTCF?

twilit talon Dec 26, 2024, 5:29 PM

#

read the ... fucking idk?

#

acpica doesnt know

#

it has an oem id of RTCONFIG

#

so some sort of proprietary config

carmine nacelle Dec 26, 2024, 5:35 PM

#

Wth is this discussion? https://forum.osdev.org/viewtopic.php?t=57109

#

I think I should shill pmOS there ultrameme

twilit talon Dec 26, 2024, 5:36 PM

#

yeah

#

some clueless people there

left dew Dec 26, 2024, 5:38 PM

#

I don't think you can run ACPI in user space. It requires interrupts.

#

what

twilit talon Dec 26, 2024, 5:41 PM

#

least clueless osdev forums user

left dew Dec 26, 2024, 5:41 PM

#

oh i just realized you were talking in that thread

twilit talon Dec 26, 2024, 5:41 PM

#

yeah

left dew Dec 26, 2024, 5:41 PM

#

but...

8infy wrote:

backlight

fans, temperatures, etc
Most of these are documented in chipsets.

#

WHAT

twilit talon Dec 26, 2024, 5:41 PM

#

lmao

left dew Dec 26, 2024, 5:41 PM

#

this guy might actually be clinically dumb

twilit talon Dec 26, 2024, 5:42 PM

#

just hack every manufacturer and steal chipset specs

dense carbon Dec 26, 2024, 5:46 PM

#

he is notoriously opinionated

#

he invented his own dubious "neanderthal theory" which you can read at https://rdos.net/eng/

RDOS operating system and the Neanderthal theory

RDOS operating system, homo sapiens and Neanderthal hybridization cause of Asperger, ADHD and autism, x86 emulator, anti psychiatry

twilit talon Dec 26, 2024, 5:47 PM

#

I don't think you can run ACPI in user space. It requires interrupts.

This kinda invalidates his credibility

dense carbon Dec 26, 2024, 5:48 PM

#

twilit talon >I don't think you can run ACPI in user space. It requires interrupts. This kin...

his reply is odd

Maybe possible, but you don't want to do it. For one, the interrupt might not even happen in the process where ACPI is mapped.

#

generally you can't count on interrupts happening in the process that you want to handle them

twilit talon Dec 26, 2024, 5:49 PM

#

why does this matter?

dense carbon Dec 26, 2024, 5:49 PM

#

it doesn't

#

i don't know what he thinks you should be doing

twilit talon Dec 26, 2024, 5:49 PM

#

does he think ill have idt in userspace or something

dense carbon Dec 26, 2024, 5:50 PM

#

just letting a hpyothetical acpi server install a userland address into the vector table apparently

left dew Dec 26, 2024, 5:50 PM

#

i think what he thinks

#

is that you'd set the idt entry to like

twilit talon Dec 26, 2024, 5:50 PM

#

dense carbon just letting a hpyothetical acpi server install a userland address into the vect...

lol

left dew Dec 26, 2024, 5:50 PM

#

have a cs with cpl=3

dense carbon Dec 26, 2024, 5:50 PM

#

which no one would ever do

left dew Dec 26, 2024, 5:50 PM

#

and have it jump to userspace code

#

but i dont think bro takes into consideration that youcan just signal an event

twilit talon Dec 26, 2024, 5:50 PM

#

like u have to have a very limited understanding of how any of it works to say something like that

#

or to think that

left dew Dec 26, 2024, 5:51 PM

#

and dispach something to a userspace process outside the irq

twilit talon Dec 26, 2024, 5:51 PM

#

just literaly set a pending bit on an eventfd

left dew Dec 26, 2024, 5:51 PM

#

his brain is stuck in the protected mode single address space era

twilit talon Dec 26, 2024, 5:51 PM

#

looks like it

left dew Dec 26, 2024, 5:51 PM

#

i should try and make a microkernel

#

it sounds like fun

dense carbon Dec 26, 2024, 5:52 PM

#

left dew but i dont think bro takes into consideration that youcan just signal an event

good god, solaris is already doing routing of interrupts into foreign threads in 1991 - an interrupt handler on solaris, if it tries to block, the state will be saved and automagically copied over to a thread waiting to take over

left dew Dec 26, 2024, 5:53 PM

#

even better

#

my approach was kind of naive and dpc-like

#

but you can definitely do something smart and more optimized for a microkernel

twilit talon Dec 26, 2024, 5:55 PM

#

dense carbon good god, solaris is already doing routing of interrupts into foreign threads in...

Lmao

#

Solaris engineers should have clearly consulted him

#

Then they would know its impossible

dense carbon Dec 26, 2024, 5:55 PM

#

what i might imagine if you need to do something very fast, like a multiqueue NIC or disk adapter driver, is to have interrupt threads of your driver server be affine to a CPU, and of a very high priority (so they preempt anything else) and when an interrupt comes in, directly switch into their context

twilit talon Dec 26, 2024, 5:56 PM

#

Yeah

dense carbon Dec 26, 2024, 5:56 PM

#

i think in practice people will be making the threads that handle interrupts very high priority anyway, so this may not even be a special case

twilit talon Dec 26, 2024, 5:57 PM

#

True

left dew Dec 26, 2024, 5:57 PM

#

the cpu pinning + switching directly into the thread context is smart though

hoary moat Dec 26, 2024, 5:57 PM

#

this guy is fucking stupid

dense carbon Dec 26, 2024, 5:58 PM

#

USB resides on PCI, so no idea why ACPI would benefit USB.
what is striking is how he just confidently speaks without the remotest curiosity

hoary moat Dec 26, 2024, 5:58 PM

#

left dew Dec 26, 2024, 5:58 PM

#

least based infy response

hoary moat Dec 26, 2024, 5:59 PM

#

And ACPI cannot tell you if there is a functional PS/2 keyboard or mouse.
bs

left dew Dec 26, 2024, 5:59 PM

#

can it tell you if there's a functional ps/2 keyboard/mouse? lol

hoary moat Dec 26, 2024, 5:59 PM

#

it can tell you if there is one

left dew Dec 26, 2024, 5:59 PM

#

i think you'd ask the controller about that

dense carbon Dec 26, 2024, 5:59 PM

#

hoary moat > And ACPI cannot tell you if there is a functional PS/2 keyboard or mouse. bs

here i was thinking that that's why it has the bloody nodes for both

left dew Dec 26, 2024, 5:59 PM

#

it can tell you if there's an i8042 controller and where it is located, no?

#

but that's probably the most acpi will tell you

#

or does it actually tell you more?

hexed acorn Dec 26, 2024, 6:00 PM

#

dense carbon he invented his own dubious "neanderthal theory" which you can read at https://r...

[...] cause of Asperger, ADHD and autism, x86 emulator [...]
i wonder what's the cause of x86 emulator

left dew Dec 26, 2024, 6:01 PM

#

great website

#

it just refuses to load the blog iframe

#

mind stuck in the 90s, website designed for equally old browsers i guess

hoary moat Dec 26, 2024, 6:03 PM

#

my brain literally started hurting reading him

left dew Dec 26, 2024, 6:04 PM

#

http://svn.rdos.se/rdos/trunk/

apps
aspie-quiz
...
kernel

#

what

#

first operating system with a built in asperger's test 🔥

dense carbon Dec 26, 2024, 6:05 PM

#

left dew http://svn.rdos.se/rdos/trunk/ - apps - aspie-quiz - ... - kernel

"Based on these evidences, a study in Aspie Quiz asked participants if they could detect somebody's presence without seeing them, which many more neurodiverse people answered 'yes' to. If neurodiverse people can detect presence without using the normal sensory channels, this means they must have some other physical sensory channel. Given the evidences of sensing electromagnetic fields and having psychic abilities, it's assumed that neurodiverse people can use the electromagnetic medium to communicate, and that this might work over large distances."

"The use of Ephedra as an stimulant by Neanderthal might explain the need for Ritalina or Amphetamine in todays individuals with ADHD."

left dew Dec 26, 2024, 6:05 PM

#

"The use of Ephedra as an stimulant by Neanderthal might explain the need for Ritalina or Amphetamine in todays individuals with ADHD."
that is so dumb

#

who comes up with these

carmine nacelle Dec 26, 2024, 6:06 PM

#

Lol

carmine nacelle Dec 26, 2024, 6:08 PM

#

left dew i should try and make a microkernel

Maybe pmOS needs contributors ultrameme

left dew Dec 26, 2024, 6:08 PM

#

maybe...

#

i'll take a look but i don't promise anything

#

but at this point i'll probably do anything to not be bored

#

as long as ur project doesn't make me want to off myself just by working on it

#

i.e insane build system or lack of proper maintainer tools (i.e clang-tidy/clang-format)

carmine nacelle Dec 26, 2024, 6:09 PM

#

Idk I haven't really though about accepting contributions and it's a bit of a mess

carmine nacelle Dec 26, 2024, 6:10 PM

#

left dew i.e insane build system or lack of proper maintainer tools (i.e clang-tidy/clang...

I have clang-format, but it's half broken

left dew Dec 26, 2024, 6:10 PM

#

send repo pls i don't want to look for first message meme

carmine nacelle Dec 26, 2024, 6:10 PM

#

In vscode

#

https://gitlab.com/mishakov/pmos

GitLab

Mikhail Kovalev / pmOS · GitLab

GitLab.com

left dew Dec 26, 2024, 6:10 PM

#

wow gitlab

dense carbon Dec 26, 2024, 6:10 PM

#

microkernel is an interesting problem space

carmine nacelle Dec 26, 2024, 6:10 PM

#

It's mirrored on github

#

microkernel is an interesting problem space
and mine is weird

dense carbon Dec 26, 2024, 6:11 PM

#

to name one thing, i wonder what microkernel would like to do when it comes to zero-copy or as close to zero-copy as possible network i/o?

left dew Dec 26, 2024, 6:12 PM

#

probably some sort of kernel api to allocate dma memory?

carmine nacelle Dec 26, 2024, 6:12 PM

#

dense carbon to name one thing, i wonder what microkernel would like to do when it comes to z...

Just make network cards write to memory objects?

#

That was my idea of implementing disk IO

left dew Dec 26, 2024, 6:12 PM

#

that's mapped into the process + you get to pass it to the device because you get a physical address

carmine nacelle Dec 26, 2024, 6:12 PM

#

And then you move pages around

#

And it's zero copy disk IO

left dew Dec 26, 2024, 6:12 PM

#

that works nicely with iommu too because the kernel can handle that for you and you just get something you pass to the device anyway

carmine nacelle Dec 26, 2024, 6:12 PM

#

That's why I want working set before AHCI

carmine nacelle Dec 26, 2024, 6:13 PM

#

left dew probably some sort of kernel api to allocate dma memory?

Yeah my memory objects take DMA flag, which allocates contiguous memory from pmm

#

Which would not be swapped out to disk, once I have that

dense carbon Dec 26, 2024, 6:13 PM

#

carmine nacelle Just make network cards write to memory objects?

there's all kinds of scattery gathery stuff that the nic driver will want to deal in, i had something in mind which seemed challenging but i've completely forgotten it

#

i'll let you know if i remember it again

#

but i suppose for implementing (say) sendfile, you can map in chunks of the file VM object and wire those, have some API to get the physical addresses, as to the packet headers you could have a pool of mbuf type structs in wired memory, and deal in those

carmine nacelle Dec 26, 2024, 6:17 PM

#

I don't know the unix history/right terms 🥶

dense carbon Dec 26, 2024, 6:17 PM

#

carmine nacelle I don't know the unix history/right terms 🥶

they're described well at https://man.freebsd.org/cgi/man.cgi?mbuf(9)

carmine nacelle Dec 26, 2024, 6:18 PM

#

Hmm

#

This doesn't feel very microkernely to me (who doesn't know how kernels usually work)

#

Although I guess it could be a good kernel IPC mechanism

#

Yeah

dense carbon Dec 26, 2024, 6:20 PM

#

carmine nacelle This doesn't feel very microkernely to me (who doesn't know how kernels usually ...

well i would be expecting pool of mbufs would be kept in the networking stack server

#

in wired memory

carmine nacelle Dec 26, 2024, 6:20 PM

#

But then how do you do zero copy

#

You're gonna be eating a lot of TLB misses if you move pages around

dense carbon Dec 26, 2024, 6:21 PM

#

the server would send a scatter-gather list of the physical addresses of the mbufs + the body of the data to the NIC server

carmine nacelle Dec 26, 2024, 6:22 PM

#

That's fine, but how do you give data to applications

dense carbon Dec 26, 2024, 6:22 PM

#

that's always been the harder part

#

zero-copy rx is complicated

carmine nacelle Dec 26, 2024, 6:23 PM

#

dense carbon the server would send a scatter-gather list of the physical addresses of the mbu...

Alternatively you can just use shared memory

dense carbon Dec 26, 2024, 6:24 PM

#

carmine nacelle Alternatively you can just use shared memory

yes, some kind of ring buffer with a doorbell probably ideal way to communicate to the NIC driver what you want to send

carmine nacelle Dec 26, 2024, 6:25 PM

#

You could probably try to avoid the network server all together, and do some sorcery to make applications talk to NIC driver processes directly and write straight to their buffers with scatter-gather, with some kernel mechanism making sure userspace doesn't pull the rug out of it

#

That sounds like fun

#

Although the server middle man being a library in the address space of the NIC driver is an implementation detail I guess and is probably not relevant...

hexed acorn Dec 26, 2024, 7:18 PM

#

dense carbon to name one thing, i wonder what microkernel would like to do when it comes to z...

ig in managarm you could have a memory object with the data to send, send the handle to it over ipc, then have the nic dma straight to it via scatter-gather

carmine nacelle Dec 28, 2024, 4:45 PM

#

nice c++ skill issue

carmine nacelle Dec 28, 2024, 11:29 PM

#

I'm going insane

#

I've added the new CoW thingie, which remaps the R/W pages are read only

#

I've double checked that invlpg gets called

#

info tlb shows it doesn't have write flag

#

and yet it doesn't page fault on call, even though the stack doesn't have the read flag

#

(I'm cloning stack in fork, among other things)

#

But sometimes it seems like it actually does

carmine nacelle Dec 28, 2024, 11:38 PM

#

carmine nacelle I've double checked that invlpg gets called

Thonk Thonk Thonk Thonk

Copying page 7FFFFFFFF000 to 7FFFFFFFF000 tables 9 13
invlpg 7FFF801FF000

#

iwannadie

#

What the hell, how does that even happen

#

I'm blaming copilot

carmine nacelle Dec 29, 2024, 10:35 AM

#

Switching to task 13
syscall_asign_page_table 23 0 3 task 13
Copying page 1000 to 2000 tables 9 12
Copying page 2000 to 4000 tables 9 12
Copying page 3000 to 6000 tables 9 12
Copying page 4000 to 8000 tables 9 12
Copying page 5000 to 9000 tables 9 12
Copying page 13000 to 17000 tables 9 12
Copying page 14000 to 28000 tables 9 12
Copying page 267000 to 2CE000 tables 9 12
Copying page 268000 to 2D0000 tables 9 12
Copying page 26A000 to 2D2000 tables 9 12
Copying page 46A000 to 46A000 tables 9 12
Copying page 7FFFFFFFF000 to 7FFFFFFFF000 tables 9 12
clone page table returned to 13
Switching to task 20
Switching to task 13
Page already allocated table 9
Anonymous page
Page refcount 3
Page 7FFFFFFFFE78 resolved
Switching to task 20
Switching to task 1

Yep it's a nasty tlb skill issue

carmine nacelle Dec 29, 2024, 11:33 AM

#

Omg

#

I found the bug

#

And it's the dumbest bug I've ever had

#

6 hours of debugging?

#

It's because of return 0

#

Judging by the logs CoW is working perfectly

noble mason Dec 29, 2024, 11:36 AM

#

carmine nacelle It's because of `return 0`

skull

carmine nacelle Dec 29, 2024, 11:36 AM

#

carmine nacelle It's because of `return 0`

Which is right

#

but that casted to ReturnStr<bool>, which is like broken std::optional

#

So it returned false, which means "page isn't available, block the task"

#

Average C++ skill issue

pearl narwhal Dec 29, 2024, 11:40 AM

#

~~mandatory reminder that this wouldn't happen in rust~~

carmine nacelle Dec 29, 2024, 11:41 AM

#

Nah I think I have an explicit copy constructor from the result

#

For "convenience"

#

(extra debugging time)

carmine nacelle Dec 29, 2024, 11:42 AM

#

pearl narwhal ~~mandatory reminder that this wouldn't happen in rust~~

Also I have been thinking about using Rust for drivers and whatnot

pearl narwhal Dec 29, 2024, 11:43 AM

#

carmine nacelle Also I have been thinking about using Rust for drivers and whatnot

would be cool to have a higher level driver framework for drivers in rust

carmine nacelle Dec 29, 2024, 11:43 AM

#

Nah

#

Drivers in Rust with it's standard lib

pearl narwhal Dec 29, 2024, 11:44 AM

#

sure but surely drivers have extra stuff they need

#

like requesting DMA memory

carmine nacelle Dec 29, 2024, 11:45 AM

#

I have a couple of functions for that

#

Hmm

#

There must be an existing framework for that

#

uAPI rust bindings?

noble mason Dec 29, 2024, 11:45 AM

#

carmine nacelle 6 hours of debugging?

a month ago i had segfault when i was writing a jvm in pure c, after 2 hours of debugging the problem was in ONE symbol (i dont remember exactly, * or +), i understood that when i did a valgrind by tip of chatgpt, and it says what program is "corrupted" pointer, not linked with osdev but yes

carmine nacelle Dec 29, 2024, 11:47 AM

#

The issue is that nothing was crashing

#

The kernel was just doing its job and quietly blocking the task

#

Like when something crashes at least you know where to look

noble mason Dec 29, 2024, 11:48 AM

#

carmine nacelle The kernel was just doing its job and quietly blocking the task

~~if it quietly blocked the task then try to add the scream into speakers when its blocks the task~~

carmine nacelle Dec 29, 2024, 11:49 AM

#

No, but my IPC blocks the tasks

noble mason Dec 29, 2024, 11:49 AM

#

oh

carmine nacelle Dec 29, 2024, 11:49 AM

#

So it's done a lot

noble mason Dec 29, 2024, 11:49 AM

#

a sick of microkernels

#

many blocks of tasks

carmine nacelle Dec 29, 2024, 11:50 AM

#

I mean if you do read on linux from pipe, it also blocks

noble mason Dec 29, 2024, 11:50 AM

#

i know that

pearl narwhal Dec 29, 2024, 11:51 AM

#

carmine nacelle No, but my IPC blocks the tasks

is your IPC super complicated or?

carmine nacelle Dec 29, 2024, 11:51 AM

#

asynchronous message queues

#

So like send() and recieve(), the first never blocks and the second blocks if there are no messages

pearl narwhal Dec 29, 2024, 11:52 AM

#

is it zero copy

carmine nacelle Dec 29, 2024, 11:52 AM

#

No

#

It's 2 copies and 3 mallocs KEKW

pearl narwhal Dec 29, 2024, 11:53 AM

#

i really want to get my IPC right

#

i have 0 experience tho

carmine nacelle Dec 29, 2024, 11:53 AM

#

It's super dumb but I wrote it a year ago and it just always works, and I don't want to touch it

pearl narwhal Dec 29, 2024, 11:54 AM

#

so it will not be right

pearl narwhal Dec 29, 2024, 11:54 AM

#

carmine nacelle It's super dumb but I wrote it a year ago and it just always works, and I don't ...

well my first buddy allocator used a VecDeque as a freelist

#

i think that's even more dumb

carmine nacelle Dec 29, 2024, 1:50 PM

#

Fuck

#

Not broken locks again

#

I think I'll just steal x86 implementation

carmine nacelle Dec 29, 2024, 3:07 PM

#

Cool, RISC-V port is broken for some reason (I'm blaming timers)

carmine nacelle Dec 29, 2024, 4:40 PM

#

I'm kinda half depressed because I don't know when it broke

#

And what's broken

carmine nacelle Dec 29, 2024, 6:47 PM

#

Ok I've just realized I've been writing in uACPI thread

#

This is weird

#

I think it's 2 bugs

#

And fork is def broken on risc-v

#

But through exit()debugging ( galaxybrain ) I've found that the issue is not where I expected it to be

carmine nacelle Dec 29, 2024, 7:28 PM

#

This is insanity

#

Something is calling exit()

#

But there is a tiny issue

#

There is nothing that could be doing that

#

I need GDB

#

But idk if it's gonna be of help

#

Since it reorders execution of threads

#

The only explanation that I can come up with it tham the thread stacks are somehow mapped with MAP_SHARED

#

and are not being cloned

carmine nacelle Dec 29, 2024, 10:32 PM

#

After a lot more debugging

#

I think it's the same bug which breaks the framebuffer on x86 on some machines galaxybrain

#

syscall_get_message() turns into exit()

carmine nacelle Dec 31, 2024, 2:42 AM

#

I pushed the patches to main, I now have anonymous pages I suppose, which do CoW on fork (and now fork doesn't copy any pages)

#

And a random thought at 3:40 am: why have one kernel, when I can have two Thonk
Maybe pmOS needs a sister hybrid or monolithic kernel, which would run the same userspace

carmine nacelle Dec 31, 2024, 1:18 PM

#

@cyan elk how do I stop lto from assuming if (LIMINE_BASE_REVISION_SUPPORTED == false) is always true?

dense carbon Dec 31, 2024, 1:25 PM

#

carmine nacelle <@379163868364275712> how do I stop lto from assuming `if (LIMINE_BASE_REVISION_...

does it doing this even with volatile?

cyan elk Dec 31, 2024, 1:34 PM

#

carmine nacelle <@379163868364275712> how do I stop lto from assuming `if (LIMINE_BASE_REVISION_...

this does not happen to me

#

with base revision declared as such: c __attribute__((used, section(".limine_requests"))) static volatile LIMINE_BASE_REVISION(3);

#

i made sure that the Limine boot protocol tolerates LTO well

carmine nacelle Dec 31, 2024, 1:43 PM

#

__attribute__((used)) fixed it

#

Wow, -O3 + lto assembly is unreadable

#

Not unreadable, but verry different from C++

#

My vector is pagefaulting in a very weird place

#

What the fuck, how does adding assert(tag) on line 646 fix everything?

#

What soer of UB is that?

carmine nacelle Dec 31, 2024, 2:32 PM

#

This is insanity

#

I have no clue

carmine nacelle Dec 31, 2024, 9:32 PM

#

Anyway, pmOS kernel now works with -O3 + lto

mellow sleet Dec 31, 2024, 9:53 PM

#

What did you even do to fix that

carmine nacelle Dec 31, 2024, 9:54 PM

#

I was overwriting random shit on stack with memcpy KEKW

#

galaxybrain

twilit talon Dec 31, 2024, 9:57 PM

#

carmine nacelle Anyway, pmOS kernel now works with -O3 + lto

LETSFUCKINGGOOOOOO

carmine nacelle Jan 1, 2025, 12:26 PM

#

I think the time has come to resurrect some old code (PS/2 ultrameme )

#

The implementation was only half-cursed, and I have improved the interrupts memes system, so I'm hoping for the best

carmine nacelle Jan 1, 2025, 1:45 PM

#

Can anyone explain why my keyboard driver needs threads?

twilit talon Jan 1, 2025, 1:46 PM

#

carmine nacelle I think the time has come to resurrect some old code (PS/2 <:ultrameme:729026898...

you will be using acpi to discover ps/2 devices right

carmine nacelle Jan 1, 2025, 1:46 PM

#

I'm not yet, but I certainly can

#

I'm trying to resurrect some old code

#

But having a device have 2 interrupts kinda makes it "difficult"

#

Or I can just poll them for now, to make sure the impl has survived

#

Like that driver is from when I didn't have ACPI

carmine nacelle Jan 1, 2025, 1:50 PM

#

twilit talon you will be using acpi to discover ps/2 devices right

I just check the bit in FADT, right?

#

(or I can use PNP I guess)

twilit talon Jan 1, 2025, 1:50 PM

#

carmine nacelle I just check the bit in FADT, right?

no

carmine nacelle Jan 1, 2025, 1:50 PM

#

How does windows do this (this is a question to myself)

twilit talon Jan 1, 2025, 1:50 PM

#

windows enumerates all devices in the namespace

#

and binds drivers to them

carmine nacelle Jan 1, 2025, 1:51 PM

#

I have no such system yet

#

(that's a separate thing)

twilit talon Jan 1, 2025, 1:51 PM

#

u can do ad-hoc find_devices() rn

carmine nacelle Jan 1, 2025, 1:51 PM

#

I have to think how to make that work with microkernel

#

And ideally making drivers hot-pluggable

twilit talon Jan 1, 2025, 1:51 PM

#

managarm publishes discovered stuff to the bus

carmine nacelle Jan 1, 2025, 1:52 PM

#

Hmm

#

But that has to have all the drivers running

#

Or I can have some sort of discovery server

twilit talon Jan 1, 2025, 1:52 PM

#

it has a script to check if a driver should be started i think

#

yeah

dense carbon Jan 1, 2025, 1:52 PM

#

twilit talon it has a script to check if a driver should be started i think

uDevD maybe

carmine nacelle Jan 1, 2025, 1:52 PM

#

carmine nacelle Or I can have some sort of discovery server

(Or I can integrate that into the devices server)

#

And I've been thinking of making some sort of directory which will have driver descriptors or something like that

#

Which can tell which servers correspond to which driver

twilit talon Jan 1, 2025, 1:53 PM

#

u can just do

uacpi_find_devices("<idk>", spawn_ps2_driver, ...)

for now i guess

carmine nacelle Jan 1, 2025, 1:54 PM

#

carmine nacelle Which can tell which servers correspond to which driver

And with microkernel, as I do with AHCI driver which fork()s itself on all controller it finds, I can just launch server as needed (and kill them)

#

And monitor that they don't crash/etc.

#

I think my design is starting to piece itself together

carmine nacelle Jan 1, 2025, 1:55 PM

#

dense carbon uDevD maybe

I have "devicesd"

#

Which has uACPI

#

And also FDT parsing on RISC-V

dense carbon Jan 1, 2025, 1:56 PM

#

carmine nacelle And also FDT parsing on RISC-V

that must be fun

carmine nacelle Jan 1, 2025, 1:56 PM

#

Which is responsible for device discovery and interrupts management

carmine nacelle Jan 1, 2025, 1:56 PM

#

dense carbon that must be fun

@manic girder wrote a great library for it, which makes it easy and relatively nice

dense carbon Jan 1, 2025, 1:57 PM

#

carmine nacelle <@337185227719966721> wrote a great library for it, which makes it easy and rela...

i mean in general dealing with ACPI and an FDT in userland not kerneland

carmine nacelle Jan 1, 2025, 1:57 PM

#

Oh

#

Idk

#

Why not

#

Besides making it difficult to put stuff in kernel

#

But I guess it's working out with my commitment of having "everything" in userspace

twilit talon Jan 1, 2025, 1:58 PM

#

how are u going to be dealing with resume from suspend?

#

since that happens in real mode

carmine nacelle Jan 1, 2025, 1:58 PM

#

carmine nacelle Besides making it difficult to put stuff in kernel

Like HPET

carmine nacelle Jan 1, 2025, 1:58 PM

#

twilit talon since that happens in real mode

?

#

Don't I just need to jump to long mode, at which point I have userspace?

twilit talon Jan 1, 2025, 1:59 PM

#

yes but the wake up trampoline is programmed via acpi

#

and then u need acpi to wake up the firmware

#

so ud have to have some tight coupling with devicesd

carmine nacelle Jan 1, 2025, 1:59 PM

#

twilit talon yes but the wake up trampoline is programmed via acpi

How is that done?

#

I can make kernel return the wakeup vector

twilit talon Jan 1, 2025, 2:00 PM

#

facs->wake_address = ...

carmine nacelle Jan 1, 2025, 2:00 PM

#

I already have functions for that

#

So it just wakes up and jumps to a trampoline (?)

#

And then the kernel can just wake other cores

#

And I have userspace

#

The kernel only needs lapic

#

Everything else's userspace responsibility KEKW

twilit talon Jan 1, 2025, 2:01 PM

#

ud have to somehow tell your drivers that all of the devices are about to suspend and then resume

carmine nacelle Jan 1, 2025, 2:01 PM

#

carmine nacelle Everything else's userspace responsibility <:KEKW:913708372329644042>

Idk userspace has to figure that out

carmine nacelle Jan 1, 2025, 2:01 PM

#

twilit talon ud have to somehow tell your drivers that all of the devices are about to suspen...

IPC?

#

And coupling to device server

twilit talon Jan 1, 2025, 2:01 PM

#

anyway that will be a whole load of fun to design

carmine nacelle Jan 1, 2025, 2:01 PM

#

And kernel events

left dew Jan 1, 2025, 2:02 PM

#

you probably need to tell every server to suspend its devices and wait for an ack

hoary moat Jan 1, 2025, 2:02 PM

#

never underestimate the amount of shit you need to do on wake from suspend

left dew Jan 1, 2025, 2:02 PM

#

suspend purely from userspace sounds like pain

hoary moat Jan 1, 2025, 2:02 PM

#

carmine nacelle The kernel only needs lapic

how about I/O APIC

#

?

#

MTTRs?

#

PAT?

#

those stuff would be dead

carmine nacelle Jan 1, 2025, 2:03 PM

#

hoary moat how about I/O APIC

I manage that in userspace

left dew Jan 1, 2025, 2:03 PM

#

MTRR* meme

carmine nacelle Jan 1, 2025, 2:03 PM

#

And kernel doesn't even touch it

hoary moat Jan 1, 2025, 2:03 PM

#

wake from suspend (for the cpu, if you ignore all devices) is the equivalent of starting a new CPU

carmine nacelle Jan 1, 2025, 2:03 PM

#

hoary moat MTTRs?

I could also do that from userspace

carmine nacelle Jan 1, 2025, 2:03 PM

#

hoary moat wake from suspend (for the cpu, if you ignore all devices) is the equivalent of ...

Yeah

#

I've been using multiboot in the past

#

I can bring that stuff up

left dew Jan 1, 2025, 2:04 PM

#

suspend preserves ram content right?

twilit talon Jan 1, 2025, 2:04 PM

#

S3 suspend does yeah

hoary moat Jan 1, 2025, 2:04 PM

#

left dew suspend preserves ram content right?

S3 yes

twilit talon Jan 1, 2025, 2:04 PM

#

but literally al hw is reset

carmine nacelle Jan 1, 2025, 2:04 PM

#

I can also bring back the cursedness where the bootloader program loaded by multiboot2 would start in ring 0, load the kernel, and continue in ring 3

hoary moat Jan 1, 2025, 2:04 PM

#

except ACPI_NVRAM physical memory might or might not be gone

left dew Jan 1, 2025, 2:04 PM

#

yeah as long as u communicate that to ur servers u should be fine

carmine nacelle Jan 1, 2025, 2:05 PM

#

hoary moat except ACPI_NVRAM physical memory might or might not be gone

I can also preserve that from userspace

hoary moat Jan 1, 2025, 2:05 PM

#

make sure to have an option to disable it

#

apparently some hardware it's not needed

#

because it's not actually said anywhere in the acpi spec

#

but linux says that windows does it

#

so linux does it

carmine nacelle Jan 1, 2025, 2:06 PM

#

I have exams, so I will probably put everything on hold for a few weeks after finishing the PS/2 meme

#

But I can try implementing S3 suspend

hoary moat Jan 1, 2025, 2:06 PM

#

https://github.com/OBOS-dev/obos/blob/master/src/oboskrnl/arch/x86_64/wake.c

carmine nacelle Jan 1, 2025, 2:06 PM

#

After that

#

To see how cursed it gets

hoary moat Jan 1, 2025, 2:06 PM

#

hoary moat https://github.com/OBOS-dev/obos/blob/master/src/oboskrnl/arch/x86_64/wake.c

this is my architecture specific wake code

carmine nacelle Jan 1, 2025, 2:06 PM

#

Are there microkernels which support S3?

twilit talon Jan 1, 2025, 2:07 PM

#

restoring PCI BARs is also fun

carmine nacelle Jan 1, 2025, 2:07 PM

#

That are not Xnu which isn't a microkernel

hoary moat Jan 1, 2025, 2:07 PM

#

https://github.com/OBOS-dev/obos/blob/master/src/oboskrnl/power/suspend.c#L45-L107
this is my platform independent wake code

carmine nacelle Jan 1, 2025, 2:07 PM

#

twilit talon restoring PCI BARs is also fun

Can also be done from userspace

#

I already enumerate PCI there

#

*in userspace

#

Kernel doesn't know and doesn't care about PCI

hoary moat Jan 1, 2025, 2:08 PM

#

make sure PCI would get restored before restoring devices though

#

you need to have a guarantee on that

carmine nacelle Jan 1, 2025, 2:08 PM

#

Not kernel's problem chad

#

carmine nacelle Jan 1, 2025, 2:24 PM

#

Oh wow PS/2 driver crashed the kernel risc

#

userspace crash kernel bad

left dew Jan 1, 2025, 2:26 PM

#

kernel debugger is a very nice thing to have

carmine nacelle Jan 1, 2025, 2:27 PM

#

I had it but removed it

#

It was bzt's meme

#

My kernel doesn't need a debugger

#

(except when it does)

#

Didn't support risc-v

mellow sleet Jan 1, 2025, 2:30 PM

#

carmine nacelle Oh wow PS/2 driver crashed the kernel <:risc:815143445764243476>

Astral curse

carmine nacelle Jan 1, 2025, 2:34 PM

#

Btw

#

Are there good examples of resolving symbols in stack traces?

#

I've tried to do that in the past with libunwind

#

But it didn't work

left dew Jan 1, 2025, 2:35 PM

#

you can the simple way by getting a list of symbols and if addr >= start && addr < end then print that symbol's name

carmine nacelle Jan 1, 2025, 2:35 PM

#

I've seen that Nyaux had it...

left dew Jan 1, 2025, 2:35 PM

#

that is what nyaux does

#

just parse the kernel's symbol table

manic girder Jan 1, 2025, 2:35 PM

#

I don't know about good, but I made a copy of any function symbols and you can do a search on them later

left dew Jan 1, 2025, 2:36 PM

#

of course that won't handle inlined call sites and such but it's good enough for debug builds

manic girder Jan 1, 2025, 2:36 PM

#

I found you can get rid of about half of the symbol table in a debut build, which saved some memory, fwiw

carmine nacelle Jan 1, 2025, 2:38 PM

#

Thonk I think this is getting out of hand

left dew Jan 1, 2025, 2:39 PM

#

why all of those are getting passed "init" cmdline

carmine nacelle Jan 1, 2025, 2:39 PM

#

So they are loaded by the bootstrap server

#

on boot

manic girder Jan 1, 2025, 2:39 PM

#

I mean, afaik there's no limit on the number of modules

carmine nacelle Jan 1, 2025, 2:40 PM

#

Yeah, but I might need to write some script which generates limine configs

#

Or finally write a disk driver

#

I mean I have AHCI

left dew Jan 1, 2025, 2:40 PM

#

maybe you should have an init server that loads all the servers from disk? :^)

carmine nacelle Jan 1, 2025, 2:40 PM

#

Maybe I need to give up on working set

#

Before disk

left dew Jan 1, 2025, 2:40 PM

#

then you can list them in a text file instead of the limine config

carmine nacelle Jan 1, 2025, 2:40 PM

#

Yeah

manic girder Jan 1, 2025, 2:41 PM

#

Is there any reason you don't roll evrrything into an initrd and pass that to the boostrap program?

carmine nacelle Jan 1, 2025, 2:41 PM

#

I haven't implemented it

#

I can do that

manic girder Jan 1, 2025, 2:42 PM

#

Something like tar is as simple as it gets if youre concerned about complexity

carmine nacelle Jan 1, 2025, 2:42 PM

#

I have ustar parser thingy which I planned to use as ramfs

manic girder Jan 1, 2025, 2:42 PM

#

Yeah nice

carmine nacelle Jan 1, 2025, 2:42 PM

#

But everything's unfinished

#

It parses archives fine

#

But I need to work on the filesystem aspect of it

#

Also my VFS is horrendous

#

I think I need to rewrite it

#

In C++

#

Or Go

#

Because why not have VFS in Go

#

(though I need to get it working first)

#

(it was only half-working)

#

Like I would create 10 goroutines with printf and only five of them would print something and it would stop working KEKW

#

Too many things to do nooo

#

Too ambitious of a project

#

How with 50k LoC I still don't have a shell or disk driver?

carmine nacelle Jan 1, 2025, 2:48 PM

#

carmine nacelle Also my VFS is horrendous

Some very carcinogenic C

#

Written with chatgpt KEKW

carmine nacelle Jan 1, 2025, 2:58 PM

#

manic girder Is there any reason you don't roll evrrything into an initrd and pass that to th...

The reason why I didn't implement initrd yet is because I'm not very clear on the dynamics of remapping /

#

Like if I map /, then /dev, then /proc and then remap root, what happens to those

#

(which would get handled by the devices server and POSIX process server)

left dew Jan 1, 2025, 3:28 PM

#

linux "layers" the mounts tmpfs on /home/hacker/test1 type tmpfs (rw,relatime) devtmpfs on /home/hacker/test1/test2 type devtmpfs (rw,relatime,size=16360856k,nr_inodes=4090214,mode=755) sysfs on /home/hacker/test1 type sysfs (rw,relatime)

that's me mounting tmpfs on ~/test1, creating a directory called test2 inside and mounting devtmpfs there and then mounting sysfs on ~/test1

#

the two previous mounts are still there but you can't access them because sysfs is mounted on top of it

#

hacker@raptor:~$ sudo umount test1
hacker@raptor:~$ ls test1
hi  test2```

#

unmounting the sysfs mounted there allows me to access them again

#

i was not aware that this is how it works actually, i might want to think hard about how i want to write my vfs to support that 😅

carmine nacelle Jan 1, 2025, 3:43 PM

#

Fun, I think O3 + lto might have surfaced another bug

#

Or not

#

I have no idea what's going on

#

Wth, how does it get to %rax=0xFFFF800000019170 (which is right) and %r15=0xFFFFFFFF (wtf) and doesn't fail the assert?

#

#

cpu is u32 and cpus.size() is size_t, so u64

left dew Jan 1, 2025, 3:50 PM

#

what if cpu_affinity is 1

#

so you do cpu = cpu_affnity - 1

#

and then cpu - 1

carmine nacelle Jan 1, 2025, 3:51 PM

#

bruh

#

I'm dumb

left dew Jan 1, 2025, 3:51 PM

#

what is it?

carmine nacelle Jan 1, 2025, 3:51 PM

#

cpu - 1

left dew Jan 1, 2025, 3:51 PM

#

oh

#

lmaoo

carmine nacelle Jan 1, 2025, 3:52 PM

#

should be cpu

#

My cpu_affinity is if (cpu_affinity == 0) then any cpu else cpu with id = affinity - 1

left dew Jan 1, 2025, 3:53 PM

#

interesting

#

i would expect affnity to be a bitmask

#

of cpus a thread can be scheduled on

carmine nacelle Jan 1, 2025, 3:54 PM

#

Yeah

#

The name of the variable is wrong

#

It's the CPU that the task is bound to

#

Which decides if it goes to global run queues or cpu-local ones

#

(My scheduler is a bit of a mess)

left dew Jan 1, 2025, 3:55 PM

#

affinity is kind of the right name for that

#

it absolutely makes sense

carmine nacelle Jan 1, 2025, 3:55 PM

#

I want to remove the global queues eventually

#

"It probably creates a lot of lock contentions"

#

(because I haven't measured if it does, and everything is slow anyway so it doesn't matter)

#

Also bitmasks are kinda somewhat problematic because of what if you boot on machine with > 64 cores

#

There needs to be some mechanism for allocating them dynamically

left dew Jan 1, 2025, 3:56 PM

#

just have enough to store the affinity for all cpus

#

yeah

carmine nacelle Jan 1, 2025, 3:57 PM

#

And I haven't bothered to implement that

#

And with global run queues bitmasks kinda make no sense

#

so

#

a mess

left dew Jan 1, 2025, 3:58 PM

#

with global run queue(s) bitmasks make a lot of sense

carmine nacelle Jan 1, 2025, 3:58 PM

#

Why is it failing? nooo (talking to myself)

left dew Jan 1, 2025, 3:58 PM

#

cpus need to know whether they can take the task or not

carmine nacelle Jan 1, 2025, 3:58 PM

#

But what if you have a lot of tasks and a lot of cpus

left dew Jan 1, 2025, 3:58 PM

#

unless you meant per-cpu queues

#

uh yeah i am not the best person to reason about this :^)

carmine nacelle Jan 1, 2025, 3:59 PM

#

maybe I think too much about everything

carmine nacelle Jan 1, 2025, 4:01 PM

#

carmine nacelle Why is it failing? <:nooo:750466870993223750> (talking to myself)

Ugh

#

I thought I commented that out

#

Cool, Makefiles skill issue

#

How do I make QEMU issue SPCR? roundnthonk

left dew Jan 1, 2025, 4:06 PM

#

what the heck is an SPCR

carmine nacelle Jan 1, 2025, 4:07 PM

#

SPCR ACPI table

#

Which tells operating systems to shit the logs to serial controller

#

https://www.qemu.org/docs/master/specs/pci-serial.html Thonk

left dew Jan 1, 2025, 4:10 PM

#

interesting

carmine nacelle Jan 1, 2025, 4:11 PM

#

It's a thing on RISC-V

#

But doing grep, it seems that it's only for ARM and RISC-V

twilit talon Jan 1, 2025, 4:12 PM

#

u can make it append any table using -acpitable

carmine nacelle Jan 1, 2025, 4:14 PM

#

I want some standard mechanism for my OS to decide that it can send logs to serial

#

I can just discover COM ports I guess...

twilit talon Jan 1, 2025, 4:15 PM

#

com ports should be exposed via both fdt and acpi

carmine nacelle Jan 1, 2025, 4:15 PM

#

But what if that has a COM mouse connected to it

#

It's not gonna print logs KEKW

#

risc-v, arch and loongarch have SPCR to tell OS that com is usable for logs

#

there must be something for servers on x86

mellow sleet Jan 1, 2025, 4:17 PM

#

twilit talon com ports should be exposed via both fdt and acpi

Fdt?

carmine nacelle Jan 1, 2025, 4:19 PM

#

carmine nacelle risc-v, arch and loongarch have SPCR to tell OS that com is usable for logs

Although maybe they did that because there wasn't other standard mechanism for discovering serial ports with ACPI on RISC-V risc

#

(probably besides running AML)

twilit talon Jan 1, 2025, 4:20 PM

#

mellow sleet Fdt?

flat device trees

mellow sleet Jan 1, 2025, 4:21 PM

#

twilit talon flat device trees

O

carmine nacelle Jan 1, 2025, 4:23 PM

#

https://stackoverflow.com/questions/9226082/device-misdetected-as-serial-mouse

Stack Overflow

Device misdetected as serial mouse

I'm working on a device which communicates with a PC through a (virtual) serial port. The problem is that the data we are sending occasionally gets incorrectly identified by Windows as a bus mouse,...

#

There should be some probing mechanism...

#

What am I getting myself into... https://en.wikibooks.org/wiki/Serial_Programming

Serial Programming

This book explains different aspects of serial data communication. Serial data communications is the foundation for most forms of data communications used with modern computing devices. The focus of the articles in this book will be around the implementation of RS-232 (aka RS-232C, aka V.24, aka EIA-232D, etc.) based serial data communication an...

#

I'll try asking ChatGPT more nicely

#

Do mice still use the same protocol as serial ones???

#

https://web.archive.org/web/20090302000300/https://freedos-32.sourceforge.net/showdoc.php?page=sermouse

left dew Jan 1, 2025, 4:33 PM

#

on x86 you've got ps2 and usb

carmine nacelle Jan 1, 2025, 4:33 PM

#

Do I even need serial mouse?

left dew Jan 1, 2025, 4:33 PM

#

probably not lol

carmine nacelle Jan 1, 2025, 4:34 PM

#

My (B550) PC has a serial port ultrameme

#

Ouch, no

left dew Jan 1, 2025, 4:35 PM

#

lmaoo

#

u can get a usb mouse for 10th of that price

carmine nacelle Jan 1, 2025, 4:36 PM

#

But I'm sure some people still have DOS boxes so they should still be made

#

Lol

#

Anyway

#

I almost have PS/2

#

Just need to sprinkle pthread_mutex_lock around the driver

#

letsgo

#

And it doesn't work

hoary moat Jan 1, 2025, 4:56 PM

#

astral curse

mellow sleet Jan 1, 2025, 4:57 PM

#

Astral curse

carmine nacelle Jan 1, 2025, 4:57 PM

#

I was missing fflush

#

Now time to add interrupts

#

Surprisingly, the driver has survived almost all of the changes

#

That I did around it

hoary moat Jan 1, 2025, 4:59 PM

#

ps2 driver
fflush

carmine nacelle Jan 1, 2025, 5:00 PM

#

Why not

#

ps2 driver
pthreads

twilit talon Jan 1, 2025, 5:03 PM

#

carmine nacelle <:letsgo:1282155513156468747>

How do u discover it?

carmine nacelle Jan 1, 2025, 5:03 PM

#

hoary moat >ps2 driver >fflush

only a tiny bit overengineered

carmine nacelle Jan 1, 2025, 5:03 PM

#

twilit talon How do u discover it?

I assume it exists unconditionally trl

#

(I'll add ACPI after interrupts)

#

Does ACPI tell me IO ports, interrupts and so on?

hoary moat Jan 1, 2025, 5:04 PM

#

yes

#

in resources

carmine nacelle Jan 1, 2025, 5:04 PM

#

Are there PCI PS/2 controllers?

hoary moat Jan 1, 2025, 5:04 PM

#

no

hoary moat Jan 1, 2025, 5:04 PM

#

carmine nacelle Are there PCI PS/2 controllers?

*there

#

https://tenor.com/view/minor-spelling-mistake-gif-21179057

Tenor

carmine nacelle Jan 1, 2025, 5:07 PM

#

Can there be more than 1 controller?

twilit talon Jan 1, 2025, 5:08 PM

#

Yes

carmine nacelle Jan 1, 2025, 5:08 PM

#

i8042 that is

twilit talon Jan 1, 2025, 5:08 PM

#

Its an unaswerable question

#

I can make a vm that does have 10 of them

#

Are there real world examples? Who knows

carmine nacelle Jan 1, 2025, 5:09 PM

#

Ok, then I'll support multiple

carmine nacelle Jan 1, 2025, 5:09 PM

#

twilit talon I can make a vm that does have 10 of them

How do I do that? KEKW

#

I don't want to assume anything

twilit talon Jan 1, 2025, 5:10 PM

#

Add support for multiple i8042 devices to qemu

#

Ez

#

Can be done in 30 minutes

carmine nacelle Jan 1, 2025, 5:11 PM

#

https://bbs.archlinux.org/viewtopic.php?id=291569

#

Which PNP is PS/2?

twilit talon Jan 1, 2025, 5:12 PM

#

Idr, but you can take a look at managarm

#

It does ps2 via acpi

carmine nacelle Jan 1, 2025, 5:17 PM

#

Fun

#

They're using mbus

#

Input event codes?

#

I need to implement that

carmine nacelle Jan 1, 2025, 5:47 PM

#

Fun, it's not AMD GPIO bug

carmine nacelle Jan 1, 2025, 6:11 PM

#

Astral curse nooo

#

My interrupts are refusing to work

twilit talon Jan 1, 2025, 6:18 PM

#

carmine nacelle Fun, it's not AMD GPIO bug

what is it

twilit talon Jan 1, 2025, 6:18 PM

#

carmine nacelle Astral curse <:nooo:750466870993223750>

astral curse is ps2 never working

carmine nacelle Jan 1, 2025, 6:20 PM

#

Yes my PS/2 driver just doesn't work

#

For no reason

#

I'm suspecting it could be the same bug that's haunting RISC-V port

#

Like how does this happen

#

My IPC can't be broken

#

I know what's happeningn

#

I think

twilit talon Jan 1, 2025, 6:29 PM

#

inb4 u have some stupid memcpy bug again

carmine nacelle Jan 1, 2025, 6:30 PM

#

I don't think so

#

Although everything is possible

#

I think I have an explanation how this could happen

carmine nacelle Jan 1, 2025, 6:34 PM

#

carmine nacelle I'm suspecting it could be the same bug that's haunting RISC-V port

It probably is

#

pause_task cancells the blocking syscalls with -EINTR and that's probably why it can't get message

#

(if someone thought my fork was overengineered, my interrupts management is even better) https://gitlab.com/mishakov/pmos/-/blob/main/devicesd/generic/interrupts.c?ref_type=heads#L8

#

Another great question is, how does AHCId work?

#

Galaxybrain kernel interface

#

And fork is broken because I'm not casting resut to int???

#

Anyway

#

I don't get it

#

How does it return 0 yet fail?

#

Something's trolling me

#

I have another theory

#

Nope

#

But it is the same bug

#

On RISC-V, the return value which is the same register as syscall number gets set to 0 and it calls syscall_exit after fork

#

On x86 it returns success

#

(which is 0)

#

They must be related

#

fork uses syscall_pause and syscall_resume

#

interrupts use the same mechanism

#

Yep, it never executes the syscall and returns success

#

pmOS curse : something messing with saved registers

carmine nacelle Jan 1, 2025, 7:32 PM

#

What the hell

#

Here's the bug

#

Now, what's setting %rax to 0?

#

I thonk I found it

#

(line 1604)

#

Now it just pagefaults

carmine nacelle Jan 1, 2025, 8:59 PM

#

Well, the driver has been resurrected I guess

#

If someone is brave to test trl

📎 pmOS.iso

left dew Jan 1, 2025, 9:00 PM

#

on real hw?

carmine nacelle Jan 1, 2025, 9:01 PM

#

I guess

#

It doesn't use ACPI though

#

And only reads scancodes

left dew Jan 1, 2025, 9:01 PM

#

i can try on my older machine which im unsure it does ps2 emulation

carmine nacelle Jan 1, 2025, 9:01 PM

#

Like I don't have a shell

left dew Jan 1, 2025, 9:01 PM

#

will it print stuff tho?

carmine nacelle Jan 1, 2025, 9:01 PM

#

Yeah

left dew Jan 1, 2025, 9:03 PM

#

does it do uefi?

#

the iso

#

ah fuck it will test it

carmine nacelle Jan 1, 2025, 9:03 PM

#

it's limine

carmine nacelle Jan 1, 2025, 9:04 PM

#

carmine nacelle If someone is brave to test <:trl:829316641191886888>

I'm testing on my laptop

left dew Jan 1, 2025, 9:04 PM

#

#

not much interesting stuff happened

#

ill comment out ahci

#

carmine nacelle Jan 1, 2025, 9:05 PM

#

I think it's not recieving interrupts

#

yeah...

carmine nacelle Jan 1, 2025, 9:08 PM

#

left dew

Did you manage to compile it?

left dew Jan 1, 2025, 9:10 PM

#

nope i just edited limine config

carmine nacelle Jan 1, 2025, 9:10 PM

#

is there a way to mask interrupts in lapic?

#

I guess if the kernel complains about no handler for interrupt, it is being lost

#

I think I never enable interrupts...

vale hazel Jan 1, 2025, 11:25 PM

#

eeeeetick

manic girder Jan 1, 2025, 11:42 PM

#

carmine nacelle Also bitmasks are kinda somewhat problematic because of what if you boot on mach...

I solved this by grouping cores together into 'clusters', 4 cores in my case. Each cluster has a queue that replaces what the global queue might do.

#

And yeah that way you can use a bitmask and not worry about the number of cpus

hoary moat Jan 1, 2025, 11:43 PM

#

carmine nacelle Also bitmasks are kinda somewhat problematic because of what if you boot on mach...

I thought about that so I decided to use a __uint128_t

#

because realistically, who is booting obos, let alone booting obos on a cpu with >128 cpus

manic girder Jan 1, 2025, 11:45 PM

#

128 cores should be enough for anyone

carmine nacelle Jan 1, 2025, 11:46 PM

#

4 cores was enough 10 years ago

#

Idk, of course it's impractical considering the other parts of the kernel, but still

#

"why limit myself"

carmine nacelle Jan 2, 2025, 2:26 AM

#

this is something new...

#

It booted once just fine

#

And this became consistent

#

(and this laptop was refusing to turn on)

#

Insane firmware?

#

@twilit talon it crashed in uacpi_namespace_node_resolve in while (node) { if (node->name.id == name.id) something something

#

 222490: 44 09 c6                         orl    %r8d, %esi
  222493: 40 0f b6 ff                      movzbl    %dil, %edi
  222497: 01 fe                            addl    %edi, %esi
  222499: 81 c6 00 00 00 5f                addl    $0x5f000000, %esi       # imm = 0x5F000000
;         if (bytes_left != 0 && single_nameseg)
  22249f: a8 01                            testb    $0x1, %al
  2224a1: 0f b6 c0                         movzbl    %al, %eax
  2224a4: 0f 45 c1                         cmovnel    %ecx, %eax
;     if (parent == UACPI_NULL)
  2224a7: 4d 85 e4                         testq    %r12, %r12
  2224aa: 4c 0f 44 e2                      cmoveq    %rdx, %r12
  2224ae: 4d 8b 64 24 20                   movq    0x20(%r12), %r12
;     while (node) {
  2224b3: 4d 85 e4                         testq    %r12, %r12
  2224b6: 74 63                            je    0x22251b <uacpi_namespace_node_resolve+0x2ab>
  2224b8: 0f 1f 84 00 00 00 00 00          nopl    (%rax,%rax)
;         if (node->name.id == name.id)
  2224c0: 41 39 74 24 04                   cmpl    %esi, 0x4(%r12)
  2224c5: 74 19                            je    0x2224e0 <uacpi_namespace_node_resolve+0x270>
  2224c7: 4d 8b 64 24 28                   movq    0x28(%r12), %r12
;     while (node) {
  2224cc: 4d 85 e4                         testq    %r12, %r12
  2224cf: 75 ef                            jne    0x2224c0 <uacpi_namespace_node_resolve+0x250>
  2224d1: eb 48                            jmp    0x22251b <uacpi_namespace_node_resolve+0x2ab>
  2224d3: 66 66 66 66 2e 0f 1f 84 00 00 00 00 00           nopw    %cs:(%rax,%rax)
;     while (bytes_left != 0) {
  2224e0: 4d 85 ff                         testq    %r15, %r15
  2224e3: 0f 85 dc fe ff ff                jne    0x2223c5 <uacpi_namespace_node_resolve+0x155>
  2224e9: e9 75 fe ff ff                   jmp    0x222363 <uacpi_namespace_node_resolve+0xf3>

#

📎 devicesd.elf

hoary moat Jan 2, 2025, 2:34 AM

#

don't forget to send the asl blob

#

*aml

carmine nacelle Jan 2, 2025, 2:34 AM

#

(leaving this here since I want to finish with PS/2 first)

#

I have sent it

#

before

#

I give up

#

I don't know who's galaxybrain but my laptop replies with 0xfa and does not report the layout that is used

#

I think I've sunk too much time into it, I'll just commit everything and fix that later

#

#

Something fishy's going on with interrupts though

#

It worked on 2/3 of my laptops

twilit talon Jan 2, 2025, 9:42 AM

#

carmine nacelle <@353593948523986945> it crashed in `uacpi_namespace_node_resolve` in `while (no...

Needs a dump, this doesnt tell me much tbh

twilit talon Jan 2, 2025, 10:05 AM

#

carmine nacelle <@353593948523986945> it crashed in `uacpi_namespace_node_resolve` in `while (no...

Also u say it booted once, that kinda gives me a hint that it might be a bug on your side with uninitialized memory somewhere perhaps?

#

Anyways send a dump and a dmesg that Linux produces if you have that

carmine nacelle Jan 2, 2025, 11:28 AM

#

I don't

carmine nacelle Jan 2, 2025, 11:32 AM

#

twilit talon Needs a dump, this doesnt tell me much tbh

You mean AML dump?

twilit talon Jan 2, 2025, 11:32 AM

#

yeah

carmine nacelle Jan 2, 2025, 11:36 AM

#

I think it's this

📎 dump.txt

twilit talon Jan 2, 2025, 11:37 AM

#

@carmine nacelle btw any way u can enable the debug log level at least

#

to see where exactly this happens

carmine nacelle Jan 2, 2025, 11:38 AM

#

uacpi_context_set_log_level?

twilit talon Jan 2, 2025, 11:39 AM

#

yeah

twilit talon Jan 2, 2025, 11:40 AM

#

carmine nacelle I think it's this

❯ /mnt/d/uacpi/tests/runner/build-linux-64bits/test-runner dsdt.dat -x ssdt1.dat ssdt2.dat --log-level info
[uACPI][INFO] RSDP 0x00007FFF3EFBC000 00000024 v02 (uOEMID)
[uACPI][INFO] XSDT 0x0000606000000320 0000003C v02 (_ASUS_ uTESTTBL)
[uACPI][INFO] DSDT 0x0000630000000400 0000F116 v02 (_ASUS_ Notebook)
[uACPI][INFO] FACP 0x00006120000004C0 00000114 v06 (uOEMID uTESTTBL)
[uACPI][INFO] SSDT 0x0000622000000100 0000168E v02 (AMD    ALIB    )
[uACPI][INFO] SSDT 0x0000618000000080 0000030C v01 (AMD    POWERNOW)
[uACPI][INFO] FACS 0x0000606000000380 00000040
WARN: remapping physical 0x00000000BEBEBEBE with size 4096
[uACPI][INFO] successfully loaded 3 AML blobs, 2738 ops in 23ms (avg 118951/s)
got a GPE
WARN: remapping physical 0x00000000BEBEBEBE with size 128
[uACPI][INFO] namespace initialization done in 158ms: 90 devices, 1 thermal zones
[uACPI][ERROR] unable to lookup named object 'ECFL' within (or above) scope '\_SB_.PCI0.SBRG.WEC_._REG': not found
[uACPI][ERROR] aborting method invocation due to previous error: AML referenced an undefined object
[uACPI][ERROR]     #0 in \_SB_.PCI0.SBRG.WEC_._REG()
[uACPI][ERROR] error during _REG execution for (EmbeddedControl) operation region \_SB_.PCI0.SBRG.WEC_.ECOR: AML referenced an undefined o...

works in userspace as always

#

but this doesnt mean much ofc

carmine nacelle Jan 2, 2025, 11:40 AM

#

Idk maybe it's my bug

twilit talon Jan 2, 2025, 11:40 AM

#

not the first bug that can only manifest in a real kernel

#

lets see debug logs

carmine nacelle Jan 2, 2025, 11:40 AM

#

But this laptop wasn't turning on, then it did turn on, booted once fine, and then had this error

twilit talon Jan 2, 2025, 11:41 AM

#

2738 ops its pretty tiny too

carmine nacelle Jan 2, 2025, 11:41 AM

#

Judging by the logs it has crashed here

twilit talon Jan 2, 2025, 11:42 AM

#

which line?

carmine nacelle Jan 2, 2025, 11:43 AM

#

83 I guess?

#

Or later

#

Anyway

#

I'm trying it

twilit talon Jan 2, 2025, 11:44 AM

#

if its 83 should be easy to debug then

carmine nacelle Jan 2, 2025, 11:44 AM

#

Also, it's unsing my sprintf implementation

#

The laptop is refusing to turn on again 😦

twilit talon Jan 2, 2025, 11:53 AM

#

carmine nacelle Judging by the logs it has crashed here

ive tested line 83 with your blob and it behaves correctly

carmine nacelle Jan 2, 2025, 11:57 AM

#

It could be anywhere tbh

#

I think it's in line 573 of namespace.c

#

btw

twilit talon Jan 2, 2025, 12:20 PM

#

carmine nacelle I think it's in line 573 of namespace.c

yeah this is the line i tested and it works fine

#

its probably line 83 in your init_from_ecdt right

carmine nacelle Jan 2, 2025, 12:21 PM

#

could be

#

or could be after that

twilit talon Jan 2, 2025, 12:22 PM

#

can u test?

carmine nacelle Jan 2, 2025, 12:22 PM

#

the laptop is not turning on

twilit talon Jan 2, 2025, 12:22 PM

#

lmao

twilit talon Jan 2, 2025, 12:43 PM

#

Yeah no I dont see how this could be a uacpi bug, I've looked pretty much everywhere I could

#

So im going to wait for you to dump more logs if you can turn it on

twilit talon Jan 2, 2025, 1:00 PM

#

carmine nacelle Judging by the logs it has crashed here

btw u never unref tables so u leak all the mappings

carmine nacelle Jan 2, 2025, 1:09 PM

#

uacpi_table_find_next_with_same_signature?

carmine nacelle Jan 2, 2025, 1:09 PM

#

twilit talon btw u never unref tables so u leak all the mappings

I though that should unref it

twilit talon Jan 2, 2025, 1:10 PM

#

carmine nacelle I though that should unref it

yes but u do early returns

carmine nacelle Jan 2, 2025, 1:11 PM

#

Oh

#

bruh

#

yeah

#

Also uacpi_namespace_node_find might leak?

twilit talon Jan 2, 2025, 1:12 PM

#

nodes are permanent

carmine nacelle Jan 2, 2025, 1:45 PM

#

lol

twilit talon Jan 2, 2025, 1:48 PM

#

reminds me of the ultra ps2 driver

#

unless thats where u stole that from

#

https://github.com/UltraOS/UltraLegacy/blob/master/Kernel/Drivers/PS2/PS2Controller.cpp

#

look at repeat_until loops with for (;;);

#

this shit was bulletproof

#

and it worked on a whole bunch of laptops with multiple ps2 mice

#

aka the touchpad & the joystick thing

carmine nacelle Jan 2, 2025, 1:52 PM

#

carmine nacelle lol

linux

carmine nacelle Jan 2, 2025, 1:58 PM

#

twilit talon https://github.com/UltraOS/UltraLegacy/blob/master/Kernel/Drivers/PS2/PS2Control...

Are there keyboards which don't support querying current scancode?
My laptop, for some reason responds with two acks (for 2 bytes) of 0xF0 0x00 but doesn't follow that up with the current scancode
Do you know anything about it?

twilit talon Jan 2, 2025, 1:59 PM

#

no idea

carmine nacelle Jan 2, 2025, 2:17 PM

#

It looks like you were never doing that in your code

#

Idk maybe I should remove it as well

#

Synaptics touchpad driver is cool

twilit talon Jan 2, 2025, 2:24 PM

#

carmine nacelle Synaptics touchpad driver is cool

yeah it was pretty interesting to write

carmine nacelle Jan 2, 2025, 2:28 PM

#

Maybe I need usb...

#

Do I just need XHCI and OHCI?

left dew Jan 2, 2025, 2:33 PM

#

xhci is enough for anything modern-ish

#

xhci does both usb 3 and 2

carmine nacelle Jan 2, 2025, 2:33 PM

#

I have laptops without USB 3

left dew Jan 2, 2025, 2:33 PM

#

then you will need ehci

#

ochi is usb 1

#

or actually idk

carmine nacelle Jan 2, 2025, 2:33 PM

#

but ohci also handles that?

left dew Jan 2, 2025, 2:33 PM

#

there are 4 controllers

#

ohci, uhci, ehci and xhci

#

i think ohci/uhci are usb 1

#

i know ehci is usb 2

#

also, just because you don't see a blue usb 3 port that might not mean there isn't an xhci controller, but it would be kinda weird

#

still a good idea to lspci and see if there's an xhci controller

carmine nacelle Jan 2, 2025, 2:36 PM

#

It's a low-end netbook from 2012

left dew Jan 2, 2025, 2:36 PM

#

are u 100% positive it doesn't do xhci?

#

in that case, you want to implement ehci too

carmine nacelle Jan 2, 2025, 2:37 PM

#

It's refusing to boot so kinda difficult to check sadmeme

left dew Jan 2, 2025, 2:37 PM

#

ah that one meme

#

is that the only machine without xhci?

carmine nacelle Jan 2, 2025, 2:38 PM

#

ok it booted

left dew Jan 2, 2025, 2:38 PM

#

ah you said laptops, nvm

carmine nacelle Jan 2, 2025, 2:38 PM

#

I have another laptop

#

And my old PC, which needed drivers for USB 3

#

It was a high-end motherboard

left dew Jan 2, 2025, 2:39 PM

#

carmine nacelle And my old PC, which needed drivers for USB 3

lspci of that would be nice

carmine nacelle Jan 2, 2025, 2:40 PM

#

uACPI debug build first meme

#

#

@twilit talon last thing it printed

#

Considering that the keyboard doesn't work, it should have died

#

It's so slow Gnome is struggling with it

twilit talon Jan 2, 2025, 2:52 PM

#

carmine nacelle <@353593948523986945> last thing it printed

wheres the gpf

#

also what are these random numbers in log

carmine nacelle Jan 2, 2025, 2:52 PM

#

This is what it has @left dew

carmine nacelle Jan 2, 2025, 2:52 PM

#

twilit talon wheres the gpf

Idk

#

It's from uACPI trl

carmine nacelle Jan 2, 2025, 2:53 PM

#

carmine nacelle This is what it has <@796462670995390524>

I see OHCI and EHCI...

carmine nacelle Jan 2, 2025, 2:54 PM

#

twilit talon wheres the gpf

It's also possible that the kernel had run out of memory KEKW

carmine nacelle Jan 2, 2025, 2:56 PM

#

carmine nacelle

I have a theory that my sprintf doesn't terminate strings

left dew Jan 2, 2025, 3:03 PM

#

carmine nacelle This is what it has <@796462670995390524>

damn that's shit

#

but yeah do ehci, it far less insane than ohci

#

im not sure if these two interact somehow like xhci does with ehci tho

#

you'll have to read up on that

carmine nacelle Jan 2, 2025, 3:04 PM

#

left dew but yeah do ehci, it far less insane than ohci

What's insane about it?

left dew Jan 2, 2025, 3:05 PM

#

it's not a very nice experience to write a driver for that iirc

#

idk the details, i have never tried obviously lol

hoary moat Jan 2, 2025, 3:05 PM

#

carmine nacelle I have a theory that my sprintf doesn't terminate strings

make sure if it's snprintf, you add one to len to include the nul-terminator

#

at least if your printf is pre-built

left dew Jan 2, 2025, 3:05 PM

#

but ehci is easier iirc

#

and of course xhci is the easiest, it's kinda similar to nvme

carmine nacelle Jan 2, 2025, 3:06 PM

#

hoary moat at least if your printf is pre-built

I'm using my own printf galaxybrain

#

And my write functions don't terminate, so I think my sprintf also doesn't

carmine nacelle Jan 2, 2025, 3:08 PM

#

carmine nacelle I'm using my own printf <:galaxybrain:701707619651420251>

I like to live dangerously

#

Nice

twilit talon Jan 2, 2025, 3:13 PM

#

carmine nacelle It's also possible that the kernel had run out of memory <:KEKW:9137083723296440...

Perhaps

carmine nacelle Jan 2, 2025, 6:04 PM

#

Ok I'm seriously thinking about rewriting my VFS in C++

#

With coroutines ultrameme

dense carbon Jan 2, 2025, 7:43 PM

#

carmine nacelle With coroutines <:ultrameme:729026898666979428>

pray what for

carmine nacelle Jan 2, 2025, 7:43 PM

#

?

dense carbon Jan 2, 2025, 7:43 PM

#

pray what would the coroutines be for

carmine nacelle Jan 2, 2025, 7:44 PM

#

To resolve the tree

dense carbon Jan 2, 2025, 7:44 PM

#

why?

carmine nacelle Jan 2, 2025, 7:44 PM

#

It's a "vfs cache"

#

I have a central server which is in charge of mountpoints and whatnot

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