Nyaux

[molten grotto]

you don't have to do anything on the router for that to work

[flat nymph]

My router already uses dnsmasq

[surreal path]

okay its the first statement

[flat nymph]

And I have a linux PC which is always on for tftp server

[surreal path]

okay i blame the way im doing the method

well

same way obos does it

[flat nymph]

print what you're giving it?

[surreal path]

kk

[flat nymph]

Ok I'm in mood of setting that up...

[flat nymph]

@surreal path

[surreal path]

thats right right?

how do i get uacpi to print all methods of a node

i will also strlen

[molten grotto]

uacpi_namespace_for_each_child with depth 1 (and type mask of UACPI_OBJECT_METHOD_BIT if you want only actual methods ig, or just UACPI_OBJECT_ANY_BIT if you want all child regardless if its a property or a method)

[surreal path]

maybe theres an invalid character?

okay thanks qwinci

[flat nymph]

#1236769772805554206 message

[flat nymph]

I guess

[flat nymph]

https://www.iventoy.com/en/index.html

[surreal path]

bro

uacpi

i swear to god

[flat nymph]

It's UB if you don't return anything from a function

[surreal path]

yea i will

still tho

[molten grotto]

uacpi_namespace_node_name

[surreal path]

why is it a incomplete type chat

[molten grotto]

there are accessor functions you are supposed to use

[surreal path]

oohhhh

[kind root]

Generate_absolute_path

[surreal path]

oh i should use that instead?

okay

[kind root]

Yes

[surreal path]

cool

new day new iso

[molten grotto]

that's a leak of the string returned from generate path but ig it doesn't matter as its not permanent (hopefully)

[surreal path]

wait what

how do i fix that

wdym leak of the string

[molten grotto]

uacpi_free_absolute_path

[surreal path]

ohhhh

so it allocates a string

MY BAD

sorry chat

😭

[molten grotto]

its a string allocated in the heap by uacpi, the full path isn't stored within the node

yeah

not that it matters for this temporary code tho

[finite summit]

isn't it just uacpi_kernel_free

?

[molten grotto]

yes but might as well use the function made specifically for paths

[surreal path]

@flat nymph this iso

[edgy pilot]

taking a nap I see

[flat nymph]

I'm reading about iPXE

And how can I make it boot ISOs from a dir on my server

But I think I'll need to write a custom script or something for it to do that

I'm tired of having to rewrite my USB drive

[molten grotto]

I haven't bothered booting isos with pxe, I just put the raw files to a folder and as basically all of them use limine I just have a limine conf and some folders with files for different os's

[surreal path]

is there no methods on ec node??

[flat nymph]

My build script generates ISO

[flat nymph]

Maybe I flashed the wrong file

[surreal path]

okay

reflash

[molten grotto]

you pass 0 to type_mask

which means it probably doesn't match anything

[flat nymph]

Since I'm also passing a bunch of modules

To load userspace

servers

[surreal path]

what

are there no methods

????

[molten grotto]

did you change the type mask

[flat nymph]

People on random reddit post say this is doable

[surreal path]

wdym

the type mask is 0

[molten grotto]

yes which means it doesn't match anything lol

[surreal path]

oh

how do i make it match everything related to methods

[molten grotto]

you have to pass it what UACPI_OBJECT_*_BIT you want

there is one for actual methods for an example

or you can just pass UACPI_OBJECT_ANY_BIT to get everything

[surreal path]

i found it

try

[flat nymph]

Ok this is the last time I'm trying it

I have to use my laptop

[surreal path]

okay

[flat nymph]

there's like a bunch of stuff

Ask @kind root he has AML dump of my laptop

[surreal path]

@kind root whats going on

theres no _Q09 method

[surreal path]

which is what the ec told me to execute

[flat nymph]

Maybe I should hold the development of my AHCI driver, and instead implement the network subsystem and iSCSI or something

Which implies USB

[thorn bramble]

usb is nice

or, well, at least xhci is

[flat nymph]

Since my laptop doesn't have built in ethernet

And my PC has 2.5G NIC

[thorn bramble]

usb networking is piss easy

[flat nymph]

But I need

1. USB
2. Networking
3. USB network driver

[thorn bramble]

the arguably hard part here would be the network stack

and that probably isnt ev entoo hard

but it='s literally yeeting ethernet frames over a usb endpoint

so as i said it's REALLY easy

[flat nymph]

Is it universal?

[thorn bramble]

and xhcci is like 500 lines of code on its own

yeah pretty much

as long as your usb nic is somewhat standard

[flat nymph]

Because why do Android boxes only support some of the nics

[thorn bramble]

aka works under linux without any special setup

[molten grotto]

except with rndis you have additional headers but its still pretty easy

[flat nymph]

Like I have two USB NICs, one works on Nintendo Switch but doesn't on Xiaomi Android box, and another one is the opposite

I don't think it's just Linux stupidity

https://www.olimex.com/Products/USB-Modules/USB-GIGABIT/resources/rtl8153.pdf

The RTL8153 supports two networking configurations; ECM (Ethernet Control Model) configuration and
in-house configuration. The ECM configuration complies with CDC-ECM, and is a general Ethernet
networking model that enables network communication without installing additional vendor specific
drivers. The in-house configuration requires a vendor specific driver to support enhanced features and
optimized performance.

Nice

So I just implement USB CDC and automatically support all USB NICs?

[molten grotto]

well there is rndis which for an example a lot of slightly older phones use for usb tethering (which is basically cdc with extra headers on top)

[flat nymph]

Wait, XHCI is both USB 2.0 and 3.0? (according to chatgpt)

[surreal path]

eee

[tawdry mirage]

yes xhci handles all usb

[flat nymph]

It printed a bunch of text

Your terminal is not fullscreen

[tawdry mirage]

well, excluding usb 4 ig but that's thunderbolt in disguise

[flat nymph]

My laptop has that

And I have a thunderbolt 3 laptop from 2017

[tawdry mirage]

for usb 4 ig the usb3 part of it is wired to xhci and the pcie part goes somewhere else?

[flat nymph]

And a thunderbolt 2 laptop

[surreal path]

ill make it

no margin

[flat nymph]

Not for public release = it gets downloaded and stashed into my osdev PDFs folder

[surreal path]

okkk

[flat nymph]

Give it more ram

[thorn bramble]

that is the easy one

cdc ncm is easy too

[surreal path]

what

[flat nymph]

I guess I'll just support that then

[flat nymph]

I assume you're not giving flanterm malloc

[tawdry mirage]

you're really going to ignore johnrussell on that?

[molten grotto]

lol

[flat nymph]

my pc is my property

so not public

[kind root]

Idk, disassemble the dump maybe its there, or maybe its your skill issue

I dont have my pc rn

[surreal path]

yea im not

[tawdry mirage]

inb4 its missing because some kernel api is unimplemented

[surreal path]

bump allocator

[surreal path]

nooo not rlly, only functions im missing are things related to scheduling

[tawdry mirage]

and the evaluation for it's definition fails

[surreal path]

can u give the dump @flat nymph

[flat nymph]

It has my Windows key

Which was expensive

[surreal path]

bru

💀

im not gonna touch ur windows key

i jsut want the fuckin

dump about the ec

[molten grotto]

imagine buying windows keys

just transfer it from some garbage laptop

[surreal path]

mr @kind root can u ask him very nicely for the dump

[kind root]

You can ask him nicely to just strip the msdm table

[flat nymph]

I had to because I wanted touchscreen

[surreal path]

yea mishkavo do that

[flat nymph]

Idk how

[kind root]

Its an ASCII dump, so u can open it in e.g. Visual code and remove it

[flat nymph]

Only that?

[molten grotto]

and maybe its not even there to begin with if you bought it separately

[kind root]

Ye

[kind root]

Yeah only if it was OEM

[surreal path]

inb4 @tawdry mirage i am a cookie

[flat nymph]

No, I had to buy laptop with OS for it to have touchscreen

[surreal path]

https://tenor.com/view/lesbian-fortune-cookie-coming-out-yassification-sorry-gif-24558091

[flat nymph]

Like you can't have it and no OS

[surreal path]

(real)

[molten grotto]

ah

[surreal path]

/j

[surreal path]

thanksss

ill run this through iasl

[kind root]

I have a pin

On how to use that dump

[surreal path]

where

[kind root]

In uacpi

[surreal path]

okok

[plush hearth]

How is EC stuff going

[surreal path]

we are getting closer

[plush hearth]

Have you managed to get events yet

[surreal path]

yes

[plush hearth]

Nice

I can test on my laptop if you want

[flat nymph]

You can also try this

[surreal path]

which table file can i find the ec methods

[kind root]

Well the handler gets called but he gets a bogus event number or something

[plush hearth]

Ah

That is an improvement

[flat nymph]

I mean _Q09 is plausible

[surreal path]

which one chat

[finite summit]

dsdt

[surreal path]

kk

[flat nymph]

use grep

[kind root]

Yeah

[surreal path]

not on linux

[flat nymph]

Use WSL

[surreal path]

doesnt exist

[flat nymph]

👀

[surreal path]

very interesting i see

[flat nymph]

"pmOS works fine on my laptop"

[plush hearth]

is there multiple dsdts?

[kind root]

Dsdt is always one

[surreal path]

nope

[kind root]

There are multiple ssdts tho

[plush hearth]

oh

[kind root]

U can look at those

[plush hearth]

I know my laptop has 12 AML blobs

[kind root]

Yeah 1 dsdt and 11 ssdts

Ls /sys/firmware/acpi/tables

[surreal path]

soooo

is _Q09 supposed to be in the dsdt

[kind root]

Check sstds

[surreal path]

kkk

[kind root]

No one knows

[plush hearth]

even if it supposed to be, doesn't mean the firmware dev put it there

[flat nymph]

@plush hearth if you're in mood for testing stuff, could you also try pmOS?

[plush hearth]

sure

Can you send an ISO

oh nvm

[surreal path]

can u test nyaux first

@kind root ive checked all ssdts

no _Q09

[kind root]

Yeah bogus number then

[surreal path]

lets just have frosty test it

if its also bogus

problem with my code

if not

thats an issue to investigate

[plush hearth]

a lot happens on a lid event

ending with a panic

[surreal path]

what happens

whats the panic

tell me

[plush hearth]

"failed to evaulate aml because of not found" (that is not how you spell evaluate)

[flat nymph]

💀

[surreal path]

💀

what was the method

[plush hearth]

A lot got printed, but the last was \_SB_.PCI0.LPCB.EC0_.HWNP

[surreal path]

no i mean

[plush hearth]

panic doesn't say

[surreal path]

method im trying to give bro

[plush hearth]

I didn't see it

must of scrolled it off the screen

[surreal path]

i see

try this

[plush hearth]

maybe if you disable the printing of node found in ec_node, name is ...

then I should be able to see it

[surreal path]

i js did

[plush hearth]

ok

got ec event
in eccoolnes: ec status reg: 101000
in ec_querytime: ec status reg: 11000
attempting to handle ec eml
method im trying to give bro: _Q25, length is 4
failed to evaulate aml because of not found
NYAUX Panic! Reason: failed

[surreal path]

i see

[kind root]

Nyaux always gets non existent numbers 💀

[plush hearth]

oh, there was some unimplemented errors from uacpi

[surreal path]

😭

[surreal path]

show

[plush hearth]

uacpi called kernel schedule work
uacpi: unable to schedule GPE(0F) for execution: unimplemented

[surreal path]

when did i call schedule work

huh

[kind root]

Thats a different type of event

[plush hearth]

attempting to execute uacpi irq method

[kind root]

Exactly why u should implement it

[surreal path]

ill js

run the executed thing

that it wants to schedule

[kind root]

Yeah

[surreal path]

not in my scheduler

ill js straight up run it

yess

[flat nymph]

Does pmOS work?

[plush hearth]

haven't tried it yet

[surreal path]

run this after pmos

[finite summit]

I told you I'd send you an obos iso 15 hours ago, here it is:

[plush hearth]

thanks

[finite summit]

it's 9M now because it has mlibc 😎

[plush hearth]

it just appears to freeze without printing to the screen

[finite summit]

try power button

[plush hearth]

nice

[finite summit]

in pmos

[plush hearth]

power button causes the laptop to turn off instantly, which means that it is not in acpi mode

[finite summit]

yeah he should have a three-second timeout

iirc

[kind root]

Yeah

[finite summit]

he said that "there should be a three second timeout then it shuts down"

[kind root]

How is pmos so skill issued with framebuffers

[finite summit]

for every hobby os, there is a curse

[plush hearth]

my laptop is weird with its framebuffer

[kind root]

Nah its a theme with pmos

[finite summit]

^

[flat nymph]

It has been broken on 3 or 4 laptops

All of which are not mine

I'm hoping it will get fixed when uDRM drops

[plush hearth]

OBOS gets stuck after printing about running as uniprocessor

[flat nymph]

What's your GPU?

Also, I had an older image which had framebuffer in kernel

[finite summit]

shit

[plush hearth]

Intel UHD something from 2017

does print the correct CPU model name though

[flat nymph]

My intel UHD laptop died 😦

[finite summit]

that's done through cpuid

[plush hearth]

power button is instant shutdown

which means no ACPI mode

[flat nymph]

Can you try this?

[plush hearth]

sure

[flat nymph]

It doesn't have userspace framebuffer at all

But does have the kernel one

[surreal path]

when ur able to run this

[flat nymph]

For some reason, my fork isn't working on RISC-V with recent commits, because the stuff is calling exit with random error codes

Maybe it's the same problem

[plush hearth]

freezes after printing that TSC invariant bit is set

[flat nymph]

Does you PC have PIC?

[plush hearth]

idk

[flat nymph]

I'm just assuming it

[surreal path]

??

[flat nymph]

[plush hearth]

I'll do that now

[surreal path]

kk

^^

[flat nymph]

It probably doesn't

That's fun

Time has come to implement other calibration mechanisms

[plush hearth]

do you mean PIT?

because I think it doesn't

not properly anyway

[flat nymph]

Yeah

That

[plush hearth]

maybe use HPET

or invariant TSC

[flat nymph]

I'm only using it for LAPIC calibration

[plush hearth]

didn't even close the lid, and it got an event

[flat nymph]

It's something that I wrote two years ago and didn't touch since then

[plush hearth]

probably battery related

[surreal path]

did it panic

[plush hearth]

yes

[surreal path]

logs

what itd panic about same thing? no more unimpl'd stuff tho right?

[plush hearth]

same as before, except no uacpi complaints

[flat nymph]

Since it's a newer laptop, I think I can use LAPIC frequency MSR...

[surreal path]

okok thats better

ill try mishkavos way of extracting the query method

[flat nymph]

Just stolen from managram

[surreal path]

[flat nymph]

Though it should be the same as sprintf

[plush hearth]

you completely misspelt his name...

[flat nymph]

Unless the implementation is broken

[surreal path]

[plush hearth]

same result

[surreal path]

damn

[flat nymph]

How do I calibrate against TSC?

[surreal path]

try this

[plush hearth]

here is the ACPI dump if it helps (already removed MSDM table):

[surreal path]

its good dw

just try the iso

[plush hearth]

same

[surreal path]

damn

that is so fucking weird

[plush hearth]

it prints _Q25, which I have confirmed is not in the dsdt or ssdts

[tawdry mirage]

what's the current iso

[plush hearth]

.

[surreal path]

this

show these 2 logs pls

[tawdry mirage]

need a little break from procrastinating on writing php

[surreal path]

fair

[tawdry mirage]

so i'll test nyaux

[surreal path]

😭

[plush hearth]

idx 37, method num 36 from ec

[surreal path]

interesting

[flat nymph]

Bruh Linux TSC calibration stuff is horrible

[plush hearth]

37 is 0x25, so that makes sense

[surreal path]

seems like something happened

between that

will add another print

let me know of that log

[tawdry mirage]

_Q03 does not exist on my laptop

[plush hearth]

0 in eccoolness, 83 from ec, idx 37

[surreal path]

0???

oh wait 0 makes sense

[plush hearth]

yep

[surreal path]

i just realized why

I PUT THE FUCKING GPE IDX

IN THE EVULATE METHOD

try now should work

[plush hearth]

that is expected, his code is failing to encode the method name

[flat nymph]

But linux only does this on atoms

[tawdry mirage]

_Q33 is a thing though, and it does some rtc stuff

[surreal path]

good thing it was a simple mistake

[plush hearth]

36 from ec, idx 36, _Q24, but still fails to evaluate aml

[surreal path]

oh damn

[plush hearth]

I'll see if my laptops AML has _Q24

it does not

[surreal path]

that should be correct

[tawdry mirage]

seems to work, i pressed the brightness up key combo and it triggered the notification for it

but it works only once, no notifies happen after that

[surreal path]

okay thats cool

i see.

should i do a renable after i handle a ec event

i think i need to

yea that makes sense

try this qookie

[surreal path]

as for ur laptop

wtf

reflash and try

[tawdry mirage]

still stuck after one query

[surreal path]

[finite summit]

no

[surreal path]

why?

[tawdry mirage]

lets see what is _Q8A

[finite summit]

it's handled automatically

[surreal path]

oh

[plush hearth]

brightness button results in a notify

[surreal path]

oh

thats good

but its stuck after one query right

[plush hearth]

yeah

[surreal path]

i see

[flat nymph]

Do I trust Linux?

[surreal path]

now

if i was uacpi

why would i get stuck

i will enable trace

[finite summit]

because you forgot to send an eoi

try that

[surreal path]

wait ur supposed to send an eoi to the ec???

[finite summit]

no

but if you don't send it to the lapic

you won't get interrupts after that one

(you knew that, right?)

[surreal path]

oh yeha

[edgy pilot]

trust ilobilix
https://github.com/ilobilo/ilobilix/blob/34e3ea9f6bf5a8d76371086c9dce4c11392d7c1f/kernel/source/arch/x86_64/drivers/timers/tsc.cpp#L53-L80

[surreal path]

my bad

try now

now i send an eoi

[flat nymph]

Oh god, C++

[surreal path]

i forgot i was routing this through the ioapic into my lapic

[tawdry mirage]

okay _Q8A is some lid event, and it happened when i closed the lid

[surreal path]

💀

[tawdry mirage]

but for some reason it did not fire the notification

[flat nymph]

(As if my kernel is not written in it )

(\s)

[surreal path]

try the iso\

[plush hearth]

it is weird now

[surreal path]

it SHOULD fire now

[surreal path]

wdym

[plush hearth]

I'll try take a photo

[surreal path]

okok

thats crazy

[tawdry mirage]

btw @kind root you were wondering what Notify(\_TZ_.THRM, 0x80) was, and apparently it's the temperature change notification for thermal zones

[surreal path]

all at the same time

[molten grotto]

@tawdry mirage btw do you mind testing my kernel at some point too? mostly to just see if the ec + power button works on qacpi, didn't it have some issues on uacpi or am I misremembering infy saying something about that?

[tawdry mirage]

Thermal zone-wide temperature sensor hardware that supports asynchronous temperature change notifications does so using an SCI. The AML code that responds to this SCI must execute a Notify(thermal_zone, 0x80) statement to inform OSPM that a meaningful change in temperature has occurred. Alternatively, devices with embedded temperature sensors may signal their associated device drivers and the drivers may use an OS-specific interface to signal OSPM’s thermal policy driver. A device driver may also invoke a device specific control method that executes a Notify(thermal_zone, 0x80) statement.

[tawdry mirage]

power button wont work unless you wrote a driver for the amd gpio controller

[surreal path]

test first

[molten grotto]

ah, that's kinda cursed

[surreal path]

WDYM AML HANDLER GONE??????????

bro i didnt fucking disappear

[flat nymph]

[tawdry mirage]

same thing on @flat nymph s laptop

he has a driver for it in pmos

[surreal path]

CHAT DID I DISAPPEAR?

[tawdry mirage]

i'd guess its the same on most new amd laptops?

[surreal path]

can u test ur laptop qookie

[plush hearth]

no more events show up after that mess

[flat nymph]

Idk I only have one

[flat nymph]

I think they do that because the GPIO controller can do debouncing

[surreal path]

@kind root how can i disappear

how did i ghost the EC

https://tenor.com/view/ghost-spooky-flying-scary-ghost-images-gif-23894972

left the EC on read

[flat nymph]

qacpi?

[molten grotto]

my aml interpreter written in c++

[plush hearth]

somehow nyaux gets further than obos and pmOS on my laptop

[flat nymph]

Wow, that's cool I guess

[flat nymph]

I just need to fix timer calibration shenanigans

[tawdry mirage]

multiple events work now, but it's still getting stuck in the lid event handler

i'd guess on a write to ec ram

[surreal path]

explain

whats exactly happening

[plush hearth]

I think it might have got a brightness and a battery event at a similar time

[flat nymph]

But yeah nyaux is becoming stable

[tawdry mirage]

your code prints that it's invoking _Q8A but the notification that the aml method does never happens, and the aml looks like this ```
OperationRegion (ERAM, EmbeddedControl, Zero, 0xFF)
Field (ERAM, ByteAcc, Lock, Preserve)
{
...
Offset (0x4E),
LIDE, 1,
Offset (0x52),
LIDS, 1,
...
}
...
Method (_Q8A, 0, NotSerialized) // _Qxx: EC Query, xx=0x00-0xFF
{
LIDE = Zero
Notify (LID, 0x80) // Status Change
}

so i am assuming it's stuck on LIDE = Zero

so on an ec write (or read)

the notification never happens and no events happen after it

[flat nymph]

Do you not support LAPIC timer?

[edgy pilot]

haven't needed it yet

I'll write that single function whenever I get around to making a scheduler

[flat nymph]

What do you do in QEMU with no tsc invariant bit?

[edgy pilot]

just no tsc

kvm timer, hpet or pit

or acpi pmtimer

[flat nymph]

But what do you use for scheduling?

[edgy pilot]

I don't have a scheduler in this rewrite

[flat nymph]

Oh

[surreal path]

why is it stuck on LIDE = zero

???

[surreal path]

yea it is

runs on a lot of hw now

[plush hearth]

can't read/write to EC

[surreal path]

im trying my best to make nyaux stable

[surreal path]

what, wdym

ive installed the address space handler

[edgy pilot]

is lapic timer frequency the same as invariant tsc one

[surreal path]

it should be able

[plush hearth]

idk

[flat nymph]

I think it's not

[plush hearth]

maybe

[edgy pilot]

could be tbh

[flat nymph]

At least in QEMU

[plush hearth]

it could be, but might be something else

[surreal path]

does anyone have the slightest idea why its stuck at lide 0

[flat nymph]

put printfs at read

[surreal path]

@tawdry mirage after it gets "stuck" with the lid close event, does it still print birghtness change events

[flat nymph]

and write

[surreal path]

kk

[tawdry mirage]

it doesn't

[surreal path]

deadlock?

[tawdry mirage]

it only prints your message, nothing from uacpi

[tawdry mirage]

are you asking me?

[surreal path]

no i mean it could be possible

hold on ill put prints

try this

SORRY QWINCI ACCIDENTAL PING

@tawdry mirage

[kind root]

Makes sense

[kind root]

Very old uACPI bug which has been long fixed

[surreal path]

im on the latest uacpi tho

??

[kind root]

I guess u aren't

[surreal path]

no?

[silver yarrow]

lol

[kind root]

Show your submodule git log

[surreal path]

kk

(how do i do that)

[kind root]

CD there and type git log or show

[surreal path]

deadlock it seems

[kind root]

Dec 22 is far from latest

[surreal path]

whats the latest

[kind root]

Git pull

[surreal path]

k

done

will run it in a few

currentl;y playing roblox lol

[kind root]

I mean I would expect it to die tbh, because he evaluates callbacks right away in an irq handler without scheduling them

If they do anything non trivial he will go up in flames

[surreal path]

what do u want me to use my scheduler or smt

[silver yarrow]

[kind root]

What does git log say now

[surreal path]

bru

[tawdry mirage]

interesting that the non-trivial thing that took it down was a write to a field in an ec opregion

[surreal path]

make clean solves nothing

[tawdry mirage]

oh yeah i should update uacpi in managarm

[kind root]

He gets it from another irq handler then tries to go into ec code again and lock it again

So it hangs obviously

[thorn bramble]

no shit read the error bruh

[silver yarrow]

lol

[thorn bramble]

reading with comprehension left the chat

[tawdry mirage]

i more meant that it's surprising that other events dont do anything non-trivial enough to hang

[surreal path]

hold on im playing roblox

[tawdry mirage]

like brightness up/down

[kind root]

Yeah I guess firmware devs decided to go easy on nyaux

[tawdry mirage]

hm now i wonder if LIDE = Zero is what turns off the backlight or something

[kind root]

Nah I think this just updates the status so that when ospm re reads it, it will report closed lid

Just a guess

[tawdry mirage]

LIDE is a register in the ec ram

[kind root]

Where does lid read its current status from?

If u look at the lid object

[tawdry mirage]

LIDS

which is a few bits after LIDE

[kind root]

Interesting

[tawdry mirage]

LIDS is also from ec ram and is never written to in aml

[kind root]

Yeah who knows what that black box does lmao

[tawdry mirage]

wonder if i could extract the ec firmware from the bios update

and disassemble it

[kind root]

Is your ecam ram literal bit values?

Usually they do bytes

[tawdry mirage]

there are a lot of fields with , 1,

[kind root]

Never seen that, cool

[tawdry mirage]

there are bytes too

and one 256 bit field

[kind root]

Is it preserve id hope?

[tawdry mirage]

yeah the ec ram field is Preserve

[kind root]

Makes sense

[kind root]

Damn

Whats it called

[tawdry mirage]

SMD0

[kind root]

System management data?

[surreal path]

does uacpi scheduling require things like

thread blocking

??

[tawdry mirage]

lets see where it is accessed

SMRD and SMWR methods

SMRD is called from ```
Method (GBIF, 1, NotSerialized)
{
Debug = "HP WMI Command 0x7 (BIOS Read)"

SMWR is never called

[kind root]

Oh yeah looks like its indeed system management stuff

[flat nymph]

Implement kernel threads, it should be <100 LoC

[kind root]

What else would bios read mean

[surreal path]

i already have kernel threads

but i mean

thread blocking

how

do i need it

for uacp

upaci

uacpi*

[kind root]

Either system management or shadow memory

[tawdry mirage]

hm SMRD is also called from a method that's called from _BIX

which is battery information extended

[kind root]

Ye

[flat nymph]

It's very easy to do

[kind root]

Interesting

[tawdry mirage]

and if it doesn't go into the method that talks to the EC it returns static data instead: ```

        Method (IVBX, 0, NotSerialized)
        {
            PBIX [0x02] = Ones
            PBIX [0x03] = Ones
            PBIX [0x05] = Ones
            PBIX [0x06] = 0x19B4
            PBIX [0x07] = 0x1086
            PBIX [0x08] = 0x01A7
            PBIX [0x09] = 0x75
            PBIX [0x0A] = Ones
            PBIX [0x0C] = Ones
            PBIX [0x0D] = Ones
            PBIX [0x10] = "Bad"
            PBIX [0x11] = "      "
            PBIX [0x12] = "Bad"
            PBIX [0x13] = "Bad"
        }

[surreal path]

do i need to cast device

[surreal path]

ive never impl'd that

im scared

[surreal path]

or what

[flat nymph]

Bruh

Have a queue

And a counter

[surreal path]

i have a queue

[flat nymph]

Slap a spinlock on it

[tawdry mirage]

yeah it calls SMRD then based on the data from that it fills in the fields

[kind root]

Bad bad bad bad

Thanks firmware devs

[surreal path]

counter where

where the queue is?

[flat nymph]

(if you're feeling adventurous, you can do lock-free stuff, but this will do)

Yes

[surreal path]

like

[flat nymph]

That is your mutex

[surreal path]

a number

oh

[flat nymph]

Yes

[surreal path]

u mean a spinlock

[flat nymph]

A number

No

[tawdry mirage]

hmm i still have all the crap the bios updater dumped in the efi system partition

[flat nymph]

A number protected by spinlock

[surreal path]

okay

[kind root]

Does it perhaps ask the firmware to memcpy from shadow memory into that ec ram or something idk

[surreal path]

and what do i do

if i want to block a thread

[flat nymph]

If number++ == 0 you take the lock

If it's not, you push the thread onto the lock and schedule another one

[tawdry mirage]

the bios updater is pretty nice too, don't need windows to actually update, the windows program just dumps shit on the ESP and reboots into it

[kind root]

Nice

[surreal path]

wdym push thread onto lock

[flat nymph]

With what you do in your timer function

[flat nymph]

Have a scheduling queue in pthread_mutex

[surreal path]

i dont have pthread_mutex

[tawdry mirage]

unfortunately it's all signed so i can't just flash custom bios images

[surreal path]

what is that

[flat nymph]

Or not pthread

Just kernel mutex idk

[surreal path]

???

like what is device

why is it a handle now

[elder shoal]

its a handle

you decide what it is

in my impl for pci open I literally just malloc some memory and memcpy the address there (too lazy to do anything else)

[surreal path]

so now i cant even get the bus device or function

[flat nymph]

I haven't gotten to implementing futexes in my kernel, but it should be something like this

[elder shoal]

pci open function has the address

[surreal path]

oh

[flat nymph]

If you're going to RISC-V, then you have to support ecam

It's the same as on x86

Like exactly

[surreal path]

please dont through things on me i dont understand yet

and are far off

im just trying to get by with fixing this ec driver already

😭

[flat nymph]

Alternatively, call sched yield in your spinlocks

[surreal path]

i dont understand

just

let me fix my uacpi kernel api first

[flat nymph]

Like idk have kernel threads that run kernel code instead of user code

[surreal path]

i already have that

[flat nymph]

That's giving it a different return_to_user function

Oh

[surreal path]

ill ask about how to impl blocking in #schedulers later

[flat nymph]

Then use that to service interrupts?

[surreal path]

for now let me hyper focus for a bit

lol

[flat nymph]

This?

[surreal path]

yea i still dont understand this very well

but ill ask about it later

dw

[tawdry mirage]

lets see if i can easily find the ec blob here

[flat nymph]

struct mutex {
    spinlock_t lock;
    bool is_locked;
    struct thread_t* blocked_queue;
};

probably something like this

[kind root]

Is this your firmware?

[tawdry mirage]

yes

[kind root]

Nice

[tawdry mirage]

graciously left behind by the bios updater in \EFI\HP\BIOS\Current

on the esp

[flat nymph]

(my impl is lock-free but dw about it I guess)

[kind root]

Steam deck has an open source ec controller but I tried reading the source and understood absolutely nothing

[flat nymph]

Where can I find it?

[kind root]

Hmm it may be a fork on valves github

[surreal path]

WHAT

INVALID STAUTS??

i return

[elder shoal]

did you return something bad

[surreal path]

status ok

no

[surreal path]

that is lit what i do

[kind root]

Attach is open()

Not read and write

[surreal path]

oh

that makes sense

hold ojn

[thorn bramble]

did bro not implement pci_open_device ☠️

[surreal path]

owrks now

[surreal path]

bro was sad all day 💀 like me too so

this isnt even a comeback

i was also sad all day

lol

[thorn bramble]

show open_device

[tawdry mirage]

also this image has 2 copies of everything apart from nvram

[tawdry mirage]

A/B partitioning but for bios updates?

[thorn bramble]

how is that supposed to work

[surreal path]

wdym

[kind root]

What the fucj

[thorn bramble]

😭

[kind root]

Bro like

[surreal path]

what am i supposed to do

explain to me

[thorn bramble]

like

use the address

[kind root]

The address variable

[thorn bramble]

uacpi asks u to open...

[kind root]

Is given to you for a reason

[surreal path]

oh wait

💀

[thorn bramble]

guys why my ec driver no work

[surreal path]

okay chill chill

sorry

ill try to complain less about smal lthings

i didnt notice

[kind root]

How do u write code that u miss stuff like that 💀

[surreal path]

okay

[kind root]

Enable unused parameter warnings at least

[surreal path]

chill guys chill please

it is enabled

i must have missed the warning

[kind root]

Make warnings generate errors

[surreal path]

im sorry

okay

okay

[kind root]

-Werror

[surreal path]

-Werror right?

[thorn bramble]

why is bro apologizing

yes

[kind root]

Yeah

[broken depot]

Not having gotten quite to this point yet myself, I assume this is supposed to be something along the lines of getting a handle to the PCI configuration space?

[surreal path]

because i am sorry

[kind root]

Host specific handle that the host can then use to access the devices config space when asked

[surreal path]

like i feel yall are mad at me yk

[thorn bramble]

not me at least

this is fun

[surreal path]

fun? wdym

[kind root]

Nah why would I be mad lmao

[surreal path]

okay

[broken depot]

I thought as much

[surreal path]

okay this makes sense

[thorn bramble]

noooooooooooooooooooooothrgyfn

wtf lmao

[broken depot]

I don't know what the segment number in the PCI address is supposed to be though.

[kind root]

U cant be serious

[thorn bramble]

it's the pci segment number

literally

[surreal path]

wait

[thorn bramble]

it's how you address pci devices

[surreal path]

im stupid sorry

please dont roast me

okay

[kind root]

Pci express introduced one more element to the bdf hierarchy called a segment

[surreal path]

i just realized what i did wrong

[broken depot]

Wait is segment number something that doesn't apply to PCIe?

[tawdry mirage]

ran it through strings just to see, i wonder what this is ```
Widevine - Device unique - ECDSA Key

[thorn bramble]

oh NO

[kind root]

It does

[surreal path]

there

much much better

[kind root]

Its part of the pcie spec

[thorn bramble]

ye that's better

[kind root]

Thank you

[surreal path]

thanks

[surreal path]

no problem

[supple robin]

fent

[thorn bramble]

the ec will now personally thank you

[surreal path]

real

[broken depot]

I'm going to have to look up what the segment number is, then.

[tawdry mirage]

lmao ```
system cannot recover from this point, brick it

[surreal path]

ran fine

[thorn bramble]

?>?????

[supple robin]

what exactly is the point of this

[kind root]

Yet another level to the pci address space, almost always 0 in practice

[surreal path]

just to see if everything looks fine

[thorn bramble]

you need to use ecam to address segment != 0 tho, right

[surreal path]

now

i need to learn

[elder shoal]

pls check for an allocation failure

[surreal path]

whats ecam

[broken depot]

According to StackOverflow it's supposed to distinguish between PCIe controllers?

[thorn bramble]

no way to address it using the funny 0xcf? io ports

[kind root]

EC? Used for event routing on modern hw, reading fan/battery state on older hw

[broken depot]

A memory-mapped configuration space for PCIe

[thorn bramble]

enhanced configuration access mechanism

[kind root]

That could be a reason to use segments, yes

[surreal path]

how do i enable this

[supple robin]

no i mean allocating memory to store the pci address

[thorn bramble]

you get this shit from the madt

you dont enable it

it's pci config space but memory mapped

[finite summit]

this is one of those moments where I mute the thread

[broken depot]

Ah okay. BadgerOS doesn't support multiple PCIe segments for the time being then

[finite summit]

until he fixes the bug

[kind root]

Nah nyaux won't figure out how to store that stuff in a pointer

[supple robin]

[plush hearth]

mcfg not madt

[surreal path]

nonono ill make suer

[thorn bramble]

yeha probably

[surreal path]

dw

[thorn bramble]

my bad

[plush hearth]

madt is APIC stuff

[kind root]

Aml also has the _CBA method to get ecam base

Which qemu ships

[surreal path]

theres no bug lol

[thorn bramble]

is that better to use than getting it from the mcfg?

[kind root]

A lot of kernels dont, but its not exactly hard to support

[thorn bramble]

also what is that aml method located on

[broken depot]

I'll just put it on the ever-growing TODO list :P

[thorn bramble]

on a pci bus? or a controller?

or smth

[kind root]

Tbh I don't have the quirk table from Linux to tell u, id bet some hw incorrectly reports in mcfg and other in aml

[kind root]

Pci bus id guess

[thorn bramble]

fair

[tawdry mirage]

ig if you're finding root buses via aml you might as well run _CBA on it

[elder shoal]

yeah

[kind root]

AML only reports the root busses, which is basically the same thing as a separate controller

[tawdry mirage]

or you could just say you CBA to do it

[elder shoal]

once I get around to doing my device stuff I will just enumerate everything from acpi downwards

[thorn bramble]

lmao

[thorn bramble]

good idea

i wanna do that but

[kind root]

Id be interested to see which one is preferred by Linux

[thorn bramble]

kernel not advanced enough

[flat nymph]

people wouldn't be helping you

[surreal path]

understandable

[plush hearth]

this thread has ~41800 messages

[tawdry mirage]

sad, no one took note of the can't be arsed joke

[surreal path]

@plush hearth

[kind root]

Id trust CBA more because it’s more likely to be retrieved dynamically whereas mcfg is a hardcoded table

[thorn bramble]

i did

[surreal path]

i need u to test this iso

[plush hearth]

alright

[tawdry mirage]

oh

[thorn bramble]

troll face emoji

[kind root]

Where

[thorn bramble]

u replied to it

[tawdry mirage]

.

[thorn bramble]

lmao

[kind root]

Ah lmao

Didnt read into it

[kind root]

Fair

[flat nymph]

But you need ecam for RISC-V anyway (?)
At least I had to implement it

[kind root]

It will still be segment 0 so

[broken depot]

ECAM is not a RISC-V specific thing.

It's a PCIe thing.

[plush hearth]

pressing brightness button results in a notify, but then a ec event appears imediately after, and can't be handled

[broken depot]

BadgerOS will boot fine without PCIe but it won't be able to detect any PCIe devices .

[surreal path]

wdym cant be handled

[plush hearth]

I'll take a photo

[thorn bramble]

ECAM is just a way to access the PCI configuration space

apparently present on other platforms than x86

[kind root]

Yeah I dont think its even pcie specific

[flat nymph]

Yes, but if you're supporting it then adding multiple segments should be very easy anyway

[thorn bramble]

i guess you've got the system io thing and ECAM on x86

[broken depot]

Yeah it works the same on RISC-V

[thorn bramble]

that's nice

[flat nymph]

And RISC-V doesn't have system IO and you've got to use ECAM

[thorn bramble]

i wonder what you need to do on aarch64

[surreal path]

@kind root do i run uacpi under a seperate thread?

[flat nymph]

And aarch64 has some cursed thing?

[thorn bramble]

if you know what you're doing

(don't)

[broken depot]

I know but there's more important things to add to the kernel than multiple segments. Like a SATA driver. Or TTYs and pipes. Or displays. Or USB and USB device drivers.

[surreal path]

oh do i run it in kentry

[surreal path]

its stuck at attempting to execute a uacpi irq method?

hold on

[kind root]

Only the work queues

[surreal path]

hold on

[surreal path]

okay

[flat nymph]

Is that spectre x360?

[plush hearth]

no

[surreal path]

ffirst lets add ECAM

to pci

[flat nymph]

Doesn't matter

[surreal path]

wdym doesnt matter

[kind root]

Depends on the controller, some do ecam, some do other mmio mechanisms

[thorn bramble]

because system io works just fine

[surreal path]

so i dont need to do ecam for the moment at least? \

[kind root]

U dont need it at all

[flat nymph]

If what you have works no

[thorn bramble]

yeah you never needed it

unless you find a computer that addresses segment != 0

[plush hearth]

hp pavilion x360 convertible 14 inch

[thorn bramble]

then you will need to add it

[surreal path]

yea

[flat nymph]

Or port to other architectures

[flat nymph]

So almost the same thing

[thorn bramble]

but i like having ecam support just because it's mmio

[flat nymph]

I had spectre x360 but it died 😦

In a beautiful black/gold color

Head turning

[plush hearth]

my laptop probably won't last for much longer

the battery is almost gone

[flat nymph]

After 6 years

[surreal path]

this is so smart

#warning

[tawdry mirage]

aarch64 has either just ecam, or some custom method depending on the pcie controller driver, or some firmware interface similar to psci where you just give it the pci device address and register and it does the io for you

[plush hearth]

and the touchscreen is being weird a lot

[flat nymph]

I replaced mine, and I think it killed the charger controller

[plush hearth]

I'll need to replace mine sometime in the new year

[flat nymph]

With "hp original battery" from ebay, which was definetely not original

Sad

It was a nice laptop

[kind root]

What are your thoughts on psci

[thorn bramble]

ah so either ecam or some proprietary shit you need to write drivers for, nice

[tawdry mirage]

it's nice

[kind root]

Thats aarch64 for u

[surreal path]

okay then so first lets bring uacpi into the kentry kthread

[kind root]

Its skewed towards the latter tho

[thorn bramble]

unlucky

[tawdry mirage]

depends on the class of hardware you look at ig

[flat nymph]

Are you supposed to run interrupts in irq threads?

I don't know how monolithic kernels do that typically

[tawdry mirage]

higher end workstation/server stuff has mostly compliant stuff (mainly ig because they want to support windows), but the sbcs are kinda cancerous

not all but some at least

[kind root]

They either do DPCs or bottom halves

No one runs any work in the actual handler

[elder shoal]

I think theres a kernel that does a lot of irq thread stuff

[kind root]

Thank god windows refuses to support crappy hardware without acpi or compliant stuff

[broken depot]

SiFive, in their inifinite wisdom, has opted to use a proprietary PCIe controller for their unmatched board, which means that I have to have a whole second driver that does nothing other than parse the DTB slightly differently

[tawdry mirage]

technically you can hack around the nonstandard pcie controllers with acpi to make it work with mcfg

[kind root]

Dtb is stored in pcie somehow?

[broken depot]

No

The DTB is differently formatted

Because there is more nonstandard extra shit in it

[kind root]

Why did they do that

[broken depot]

Good question

[kind root]

Like