Nyaux | Operating System Development | Page 8

worldly condor May 4, 2024, 9:31 PM

#

lemme check the message count

#

#

on the same order of magnitude

finite summit May 4, 2024, 9:32 PM

#

except for him it's over 16 days

#

me it's over a couple of months

#

I mean tbf I do kind of spam my thread

#

A lot more while debugging the allocator bugs

#

also I'm the same as him in terms of development of my OS, I do it usually near-immediately when I get back from school

#

although about two weeks ago I got terraria, and I'd played it for every day instead of osdev

#

anyway enough literring this thread

worldly condor May 4, 2024, 9:37 PM

#

but half of them are after April 1st 2024

finite summit May 4, 2024, 9:37 PM

#

true

surreal path May 4, 2024, 9:57 PM

#

how big should the bump allocator be

#

in pages

#

gonna allocate one big vmm region for it

thorn bramble May 4, 2024, 9:58 PM

#

idk however big you want it to be

surreal path May 4, 2024, 9:58 PM

#

idk how much pages uacpi will need

thorn bramble May 4, 2024, 9:58 PM

#

me neither

#

just give it 16 pages

#

if its not enough

#

bump it up

#

lol

kind root May 4, 2024, 9:58 PM

#

yeah

thorn bramble May 4, 2024, 9:59 PM

#

just make sure that alloc checks if there is enough room left

#

keep that around somewhere so you can check and not return bad pointers

surreal path May 4, 2024, 10:01 PM

#

something like this i guess

#

then i return result

thorn bramble May 4, 2024, 10:02 PM

#

you can do size = align_up(size, 8)

#

at the start

kind root May 4, 2024, 10:02 PM

#

return null at the end

thorn bramble May 4, 2024, 10:02 PM

#

and use size elsewhere

kind root May 4, 2024, 10:02 PM

#

if it fails

thorn bramble May 4, 2024, 10:02 PM

#

oh yeah that too

#

and if the check fails make sure to panic

kind root May 4, 2024, 10:02 PM

#

nah its fine if it fails

#

he will see a UACPI_STATUS_OUT_OF_MEMORY somewhere

thorn bramble May 4, 2024, 10:02 PM

#

but his kernel also uses it

kind root May 4, 2024, 10:02 PM

#

oh lol

thorn bramble May 4, 2024, 10:03 PM

#

i mean presumably

surreal path May 4, 2024, 10:03 PM

#

yea it does

#

well im not replacing the allocator for the kernel

#

just for the

#

uacpi

thorn bramble May 4, 2024, 10:03 PM

#

well you should also replace it for kernel if it doesnt solve the issue

surreal path May 4, 2024, 10:03 PM

#

yea okay i wil

#

in that case

#

man that was quick

#

10 pages is too little

#

lol

#

wow

#

i screwed up a bump allocator

#

first time it tries to allocate it fails!

thorn bramble May 4, 2024, 10:09 PM

#

impressive

kind root May 4, 2024, 10:09 PM

#

surreal path i screwed up a bump allocator

bro corrupted his bump allocator

surreal path May 4, 2024, 10:09 PM

#

kind root bro corrupted his bump allocator

i screwed up a fkn bump allocator 💀

kind root May 4, 2024, 10:09 PM

#

now thats next level oberrow instability

surreal path May 4, 2024, 10:10 PM

#

what

#

bro

#

4096 * 10 is not 32

kind root May 4, 2024, 10:10 PM

#

bro

#

can you focus

surreal path May 4, 2024, 10:11 PM

#

im trying to

#

oh i know what i did wrong

#

yea 10 pages isnt enough

thorn bramble May 4, 2024, 10:13 PM

#

fair, 40k isnt much

kind root May 4, 2024, 10:13 PM

#

aml execution context alone is 2 pages

surreal path May 4, 2024, 10:14 PM

#

giving it 400mb

#

cries

kind root May 4, 2024, 10:14 PM

#

this is not 400 mb

surreal path May 4, 2024, 10:14 PM

#

what is that then

kind root May 4, 2024, 10:14 PM

#

0.4 mb

surreal path May 4, 2024, 10:14 PM

#

oh

kind root May 4, 2024, 10:15 PM

#

it says right there..

surreal path May 4, 2024, 10:15 PM

#

runs with 1000 pages

kind root May 4, 2024, 10:15 PM

#

doesnt die anymore?

surreal path May 4, 2024, 10:15 PM

#

no

#

on qemu fine

#

time to test on vbox

kind root May 4, 2024, 10:15 PM

#

try vmware

#

vbox i mean

#

yea

finite summit May 4, 2024, 10:15 PM

#

inb4 out of memory

surreal path May 4, 2024, 10:17 PM

#

still faults

#

not an allocator issue

kind root May 4, 2024, 10:17 PM

#

thats nice

thorn bramble May 4, 2024, 10:17 PM

#

replace the allocator for the entire kernel with a bump

surreal path May 4, 2024, 10:17 PM

#

surreal path May 4, 2024, 10:17 PM

#

thorn bramble replace the allocator for the entire kernel with a bump

okay

#

yea same thing

thorn bramble May 4, 2024, 10:19 PM

#

then you are doing something very wrong somewhere else

#

good luck!

surreal path May 4, 2024, 10:19 PM

#

at least it isnt the allocator

kind root May 4, 2024, 10:19 PM

#

seeing as u fucked up even the bump allocator thats very well possible

surreal path May 4, 2024, 10:19 PM

#

kind root seeing as u fucked up even the bump allocator thats very well possible

this is legit all the code

#include "lib/kpanic.h"
#include "main.h"
#include "vmm.h"
#include "drivers/serial.h"

uint64_t base = 0;
uint64_t max_size = 4096 * 10000;
char meow[64];
void bump_init()
{
    base = vmm_region_alloc(max_size, NYA_OS_VMM_PRESENT | NYA_OS_VMM_RW);
    max_size = base + max_size;
}
void *bump_alloc(size_t size)
{
    
    if (base < max_size && base + align_up(size, 8) < max_size)
    {
        
        size_t aligned = align_up(size, 8);
        void *result = base;
        base += aligned;
        return result;
    }
    else
    {
        
        kpanic("Out of Memory!\n", NULL);
        return NULL;
    }
}

this is the entire file

kind root May 4, 2024, 10:20 PM

#

can u try giving qemu 8 gigs of ram

surreal path May 4, 2024, 10:20 PM

#

ok

kind root May 4, 2024, 10:20 PM

#

then setting the base to 4294967296

#

hhdm

#

bypass your region alloc shit

thorn bramble May 4, 2024, 10:20 PM

#

you need to keep base and offset from the base

#

currently

#

you are checking if the address allocated and bumped is < the max size

surreal path May 4, 2024, 10:21 PM

#

surreal path ok

doesnt fault

thorn bramble May 4, 2024, 10:21 PM

#

that is so wrong

#

not the reason it doesnt work

#

but its just broken lol

kind root May 4, 2024, 10:21 PM

#

surreal path doesnt fault

what did you change?

surreal path May 4, 2024, 10:21 PM

#

kind root what did you change?

on qemu, gave 8G

#

doesnt fault

kind root May 4, 2024, 10:21 PM

#

yeah thats not even what i aksed

#

set the base to 4GiB + hhdm

surreal path May 4, 2024, 10:21 PM

#

ok

kind root May 4, 2024, 10:22 PM

#

bypass your region managament

#

and set vbox ram to 8 gigs as well

surreal path May 4, 2024, 10:22 PM

#

kind root then setting the base to 4294967296

this number is in hex?

kind root May 4, 2024, 10:22 PM

#

0x100000000

#

this is where the higher memory starts

surreal path May 4, 2024, 10:22 PM

#

ok

surreal path May 4, 2024, 10:23 PM

#

thorn bramble you need to keep base and offset from the base

?

surreal path May 4, 2024, 10:23 PM

#

kind root 0x100000000

trying on vbox

#

sec

#

ok gave vbox 8gb and did the hhdm thing

thorn bramble May 4, 2024, 10:25 PM

#

surreal path ?

you are setting base to the address of the region you allocate for the heap and then you compare it to the max size you allocate

surreal path May 4, 2024, 10:25 PM

#

trying out now

thorn bramble May 4, 2024, 10:25 PM

#

doesnt that sound stupid?

#

address < size

kind root May 4, 2024, 10:25 PM

#

bruh

thorn bramble May 4, 2024, 10:25 PM

#

not even the right units to compare

#

obvious logic error

kind root May 4, 2024, 10:25 PM

#

these bugs are unfathomable

surreal path May 4, 2024, 10:25 PM

#

kind root these bugs are unfathomable

i get im a bad programmer

#

im sorry

thorn bramble May 4, 2024, 10:26 PM

#

why are you sorry

#

☠️

kind root May 4, 2024, 10:26 PM

#

focus

thorn bramble May 4, 2024, 10:26 PM

#

exactly

#

fix bug

#

dont apologize

surreal path May 4, 2024, 10:26 PM

#

ok

kind root May 4, 2024, 10:26 PM

#

surreal path trying out now

so

surreal path May 4, 2024, 10:27 PM

#

same thing

#

same fault

#

same exact fault

kind root May 4, 2024, 10:27 PM

#

ok yeah so definitely not related to the allocator

surreal path May 4, 2024, 10:27 PM

#

yea

kind root May 4, 2024, 10:28 PM

#

so probably you're corrupting some uacpi memory somewhere

surreal path May 4, 2024, 10:28 PM

#

kind root so probably you're corrupting some uacpi memory somewhere

i replaced every allocation call with the bump allocator too

#

in the kernel

kind root May 4, 2024, 10:29 PM

#

how would that prevent use-after-free or out-of-bounds

finite summit May 4, 2024, 10:29 PM

#

Implement KASAN

surreal path May 4, 2024, 10:30 PM

#

kind root how would that prevent use-after-free or out-of-bounds

i removed every free, i just dont free with the bump. as for out of bounds i make very little allocation calls in the kernel and i've looked over how im using each bit of memory and seems fine

surreal path May 4, 2024, 10:30 PM

#

kind root so probably you're corrupting some uacpi memory somewhere

im not doing anything else while uacpi starts

kind root May 4, 2024, 10:30 PM

#

also like

surreal path May 4, 2024, 10:30 PM

#

finite summit Implement KASAN

kasan?

kind root May 4, 2024, 10:30 PM

#

u didnt even put a single printf

#

in uacpi code

finite summit May 4, 2024, 10:31 PM

#

surreal path kasan?

kernel address sanitizer

kind root May 4, 2024, 10:31 PM

#

to trace like what each variable contains

#

in that function where it dies etc

#

like learn printf debugging

#

or u even have gdb u can inspect it yourself

surreal path May 4, 2024, 10:31 PM

#

kind root to trace like what each variable contains

i will put printfs everywhere again

kind root May 4, 2024, 10:31 PM

#

enable UACPI_LOG_DEBUG too

surreal path May 4, 2024, 10:31 PM

#

okay

kind root May 4, 2024, 10:32 PM

#

or whatever thats called

#

the log level

surreal path May 4, 2024, 10:32 PM

#

UACPI_LOG_TRACE

kind root May 4, 2024, 10:32 PM

#

DEBUG

#

not trace

surreal path May 4, 2024, 10:32 PM

#

ok

#

this might take a while

#

since im printing every single fkn call

#

opening in vbox

#

#

same thing

#

addresses being returned look valid

#

same fault

kind root May 4, 2024, 10:41 PM

#

like

#

just put afucking printf in the function that prints out absolute path

#

see what branches its taking

#

how its ending up with a null

#

seriously like

surreal path May 4, 2024, 10:42 PM

#

ok

kind root May 4, 2024, 10:42 PM

#

how is it not the first thing that you do

surreal path May 4, 2024, 10:43 PM

#

because i dont know what you want me to do, im sorry for making you feel frustrated. ill put the printfs into the absolute path now

kind root May 4, 2024, 10:43 PM

#

i want you to visualize the code flow with printfs

#

you have inspected the abort function before

#

you see that its ending up with null but you dont know how

surreal path May 4, 2024, 10:43 PM

#

yea

kind root May 4, 2024, 10:43 PM

#

now to figure that out you put printfs

#

to figure out the code flow that leads up to that

surreal path May 4, 2024, 10:44 PM

#

okay i will

kind root May 4, 2024, 10:44 PM

#

do that

surreal path May 4, 2024, 10:48 PM

#

weird

#

#

this makes zero sense

kind root May 4, 2024, 10:49 PM

#

you're not printing out the absolute path pointer

surreal path May 4, 2024, 10:49 PM

#

oh whoops

kind root May 4, 2024, 10:49 PM

#

also there's a uacpi_trace shorthand

surreal path May 4, 2024, 10:51 PM

#

huh

kind root May 4, 2024, 10:51 PM

#

so your nanoprintf port is broken somehow

surreal path May 4, 2024, 10:52 PM

#

so its not rlly my fault then?

kind root May 4, 2024, 10:52 PM

#

it is

surreal path May 4, 2024, 10:52 PM

#

in which way

kind root May 4, 2024, 10:52 PM

#

lets see how u implement it

#

i know nothing about nanoprintf

surreal path May 4, 2024, 10:52 PM

#

ok

#

i can screenshare the code on my screen in #voice-1 if you'd like, or ill provide the pastebin or screenshots, or i can push everything to github for u to see

#

whatever works best for u

kind root May 4, 2024, 10:55 PM

#

what line is it faulting at in nanoprintf

surreal path May 4, 2024, 10:55 PM

#

this line

#

second

kind root May 4, 2024, 10:55 PM

#

surreal path huh

also why am i not seeing %zu formatted here?

surreal path May 4, 2024, 10:55 PM

#

kind root also why am i not seeing %zu formatted here?

i do not know

#

here

#

these are the defaults nanoprintf uses

thorn bramble May 4, 2024, 10:58 PM

#

large format specifiers

#

enable them

surreal path May 4, 2024, 10:58 PM

#

okay

thorn bramble May 4, 2024, 10:58 PM

#

#define NANOPRINTF_IMPLEMENTATION
#define NANOPRINTF_USE_FIELD_WIDTH_FORMAT_SPECIFIERS 1
#define NANOPRINTF_USE_PRECISION_FORMAT_SPECIFIERS 0
#define NANOPRINTF_USE_FLOAT_FORMAT_SPECIFIERS 0
#define NANOPRINTF_USE_LARGE_FORMAT_SPECIFIERS 1
#define NANOPRINTF_USE_BINARY_FORMAT_SPECIFIERS 0
#define NANOPRINTF_USE_WRITEBACK_FORMAT_SPECIFIERS 0
#define NANOPRINTF_VISIBILITY_STATIC```

#

i do this

kind root May 4, 2024, 10:59 PM

#

precision is also needed, uacpi uses that a lot

thorn bramble May 4, 2024, 10:59 PM

#

precision is for floats

#

i think

kind root May 4, 2024, 10:59 PM

#

no, like %.4s

thorn bramble May 4, 2024, 10:59 PM

#

oh nvm

kind root May 4, 2024, 10:59 PM

#

string with up to 4 chars

thorn bramble May 4, 2024, 10:59 PM

#

yeah i just checked the npf code again

#

they have checks that you dont enable conflicting options

#

or rather disable options without disabling everything it relies on

#

but they make sure that if you enable float you enable specifier

#

so i was misremembering

surreal path May 4, 2024, 11:01 PM

#

oh frick u

kind root May 4, 2024, 11:01 PM

#

@surreal path so i think i know what happened

#

u disabled %zu

#

and it tries to print #0 in \

#

but its end up printing #zu in %s where s is 0

#

because zu never got printed

thorn bramble May 4, 2024, 11:02 PM

#

LMAO

surreal path May 4, 2024, 11:02 PM

#

bro

thorn bramble May 4, 2024, 11:02 PM

#

no way it was that

surreal path May 4, 2024, 11:02 PM

#

is that my fucking issue

#

IS THAT

#

MY MOTHER

#

FUCKING

#

ISSUE

#

bro.

kind root May 4, 2024, 11:02 PM

#

you would have found this out had you actually put printfs in that function first thing

#

idk how u didnt do that

#

and instead did all the other shit

thorn bramble May 4, 2024, 11:03 PM

#

this is so funny

#

holy fuck

surreal path May 4, 2024, 11:03 PM

#

next time dont instantly blame my allocator halfmemeright

kind root May 4, 2024, 11:03 PM

#

i assumed u did the bare minimum debugging

#

which is fucking printfs in the faulting function

surreal path May 4, 2024, 11:03 PM

#

surreal path oh frick u

now it wont compile

#

bruh

thorn bramble May 4, 2024, 11:04 PM

#

hmm i wonder why it works for me

#

maybe i dont pass -nostdinc

surreal path May 4, 2024, 11:04 PM

#

kind root May 4, 2024, 11:04 PM

#

just replace it

#

with your types

thorn bramble May 4, 2024, 11:05 PM

#

czapek@raptor-wsl:~/source/kernel$ make all 
INFO: autodetecting backend as ninja
INFO: calculating backend command to run: /usr/bin/ninja -C /home/czapek/source/kernel/build
ninja: Entering directory `/home/czapek/source/kernel/build'
[3/3] Linking target platform/amd64/kernel
czapek@raptor-wsl:~/source/kernel$ git diff src/main.cpp
diff --git a/src/main.cpp b/src/main.cpp
index a18ce69..db1c388 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -1,6 +1,7 @@
 #include <lib/print.h>
 #include <limine.h>
 #include <memory/pmem.h>
+#include <sys/types.h>
 #include <system/cpu.h>
 
 using namespace kernel;```

#

fun

surreal path May 4, 2024, 11:05 PM

#

lol urs dont work!

kind root May 4, 2024, 11:05 PM

#

what

thorn bramble May 4, 2024, 11:05 PM

#

pmem: registering physical region 0x4d000-0x9f000, 328 KiB, 82 pages, 4 KiB for PFDB
pmem: registering physical region 0x100000-0x7fdfe000, 2094072 KiB, 523518 pages, 12272 KiB for PFDB
pmem: skipping physical region 0x7ff1d000-0x7ff1e000, too small
pmem: registering physical region 0x100000000-0x180000000, 2097152 KiB, 524288 pages, 12292 KiB for PFDB
phys1: 0x4e000
phys2: 0x4f000
phys3: 0x50000
reached end of _start, halting```

#

very much works

surreal path May 4, 2024, 11:05 PM

#

oh

#

meme

thorn bramble May 4, 2024, 11:06 PM

#

using %zx for all addresses in my kernel

#

seems to work fine

surreal path May 4, 2024, 11:06 PM

#

kind root with your types

you mean the PRIXx thingys

thorn bramble May 4, 2024, 11:06 PM

#

its just a types header

#

oh just copy them from your system headers

kind root May 4, 2024, 11:06 PM

#

with whatever header in your os has size_t

surreal path May 4, 2024, 11:06 PM

#

i use stddef

thorn bramble May 4, 2024, 11:06 PM

#

# ifndef __suseconds_t_defined
typedef __suseconds_t suseconds_t;
#  define __suseconds_t_defined
# endif```

#

i love sys/types.h

surreal path May 4, 2024, 11:07 PM

#

who names size_t

#

ssize_t

kind root May 4, 2024, 11:07 PM

#

signed size_t

surreal path May 4, 2024, 11:07 PM

#

kind root with whatever header in your os has size_t

its legit just stddef

kind root May 4, 2024, 11:07 PM

#

its a completely different type

surreal path May 4, 2024, 11:08 PM

#

well it wants it

#

i dont have

kind root May 4, 2024, 11:08 PM

#

long

surreal path May 4, 2024, 11:08 PM

#

this is the most big brained

#

galaxybrain

kind root May 4, 2024, 11:09 PM

#

typedef long ssize_t

leaden hinge May 4, 2024, 11:09 PM

#

surreal path this is the most big brained

my eyes

surreal path May 4, 2024, 11:10 PM

#

works but why does it look like that

#

%.4s

#

oh wait

#

percision

kind root May 4, 2024, 11:11 PM

#

u disabled precision

surreal path May 4, 2024, 11:11 PM

#

sec

#

works

kind root May 4, 2024, 11:12 PM

#

bro wasted so much time

#

now try hw

surreal path May 4, 2024, 11:12 PM

#

💀

thorn bramble May 4, 2024, 11:12 PM

#

that is insane

surreal path May 4, 2024, 11:12 PM

#

ok

thorn bramble May 4, 2024, 11:12 PM

#

lmfaoo

surreal path May 4, 2024, 11:12 PM

#

sec

#

lemme try bring BACK my allocators

#

sec

kind root May 4, 2024, 11:12 PM

#

i hope u learned some debugging lessons from this

surreal path May 4, 2024, 11:13 PM

#

i did

#

i learned a lot from this experience

kind root May 4, 2024, 11:13 PM

#

then its not in vein

surreal path May 4, 2024, 11:13 PM

#

yep

#

works even with my allocators on vbox

#

will flash iso to usb and reboot

#

sec

#

rebooting

#

No page fault yet but it’s taking a while

kind root May 4, 2024, 11:19 PM

#

surreal path works

also implement the io helpers lol, its like 4 lines of code

#

u already have the raw io r/w helpers

surreal path May 4, 2024, 11:20 PM

#

what’s the difference between them

kind root May 4, 2024, 11:20 PM

#

read the header

thorn bramble May 4, 2024, 11:20 PM

#

raw io is memory, io is port io?

#

i think

#

yea read the header

surreal path May 4, 2024, 11:20 PM

#

it’s loading it !!!!

#

things are happening

#

It’s a lot of stuff tho

#

Works !!!

#

Works in real hardware!!!!!!

thorn bramble May 4, 2024, 11:21 PM

#

good job

kind root May 4, 2024, 11:21 PM

#

nice

#

only 5 hours of debugging

thorn bramble May 4, 2024, 11:21 PM

#

trl

surreal path May 4, 2024, 11:21 PM

#

😭

surreal path May 4, 2024, 11:33 PM

#

kind root also implement the io helpers lol, its like 4 lines of code

yea i see what it means now after reading obos's implmentation for it, its simply just an abstraction thingy

kind root May 4, 2024, 11:34 PM

#

yup

surreal path May 4, 2024, 11:34 PM

#

i implemented now

kind root May 4, 2024, 11:34 PM

#

i think u also dont call uacpi_initialize_namespace

surreal path May 4, 2024, 11:34 PM

#

no i dont

#

call it?

kind root May 4, 2024, 11:35 PM

#

read https://wiki.osdev.org/UACPI

#

specifically the initialization section

surreal path May 4, 2024, 11:37 PM

#

i dont have events or whatever tf so it doesnt rlly work

kind root May 4, 2024, 11:37 PM

#

who cares

#

thats a warning, not an error

surreal path May 4, 2024, 11:37 PM

#

ok

#

works

kind root May 4, 2024, 11:39 PM

#

lol

surreal path May 4, 2024, 11:39 PM

#

also what is a gpe

kind root May 4, 2024, 11:39 PM

#

thats probably because those are guarded by a mutex

#

and u dont implement that

finite summit May 4, 2024, 11:40 PM

#

surreal path also what is a gpe

general purpose event

thorn bramble May 4, 2024, 11:40 PM

#

namespace initialization done
Amazing!
meow

surreal path May 4, 2024, 11:40 PM

#

finite summit general purpose event

so like, this could mean when you press the power button or smthin thats a gpe?

finite summit May 4, 2024, 11:40 PM

#

indeed

surreal path May 4, 2024, 11:40 PM

#

ah okay

kind root May 4, 2024, 11:40 PM

#

power button is a fixed event, not a general purpose one

finite summit May 4, 2024, 11:41 PM

#

meh same thing

#

potato potato

surreal path May 4, 2024, 11:41 PM

#

i barely know much about acpi that much to the level yall know 😭

kind root May 4, 2024, 11:41 PM

#

its like 4 lines of code to hook up power button https://wiki.osdev.org/UACPI#Hooking_Up_the_Power_Button

#

but u must implement register_irq_handler

finite summit May 4, 2024, 11:42 PM

#

surreal path i barely know much about acpi that much to the level yall know 😭

bruh all I know is "ACPI makes power management go brrr" and "ACPI gives random ah devices that you can't find on PCI"

surreal path May 4, 2024, 11:42 PM

#

kind root but u must implement register_irq_handler

oh fudge, uhhh

#

so heres the thing

#

yea...

#

i dont have a common stub

finite summit May 4, 2024, 11:42 PM

#

then make one

surreal path May 4, 2024, 11:42 PM

#

shoot

#

okay

kind root May 4, 2024, 11:42 PM

#

well thats what you do next then

#

how did u even hook up timer ticks

finite summit May 4, 2024, 11:43 PM

#

through a not-common stub

#

probably isr_stub_32

surreal path May 4, 2024, 11:43 PM

#

finite summit through a not-common stub

ye

#

ok so for a common stub

kind root May 4, 2024, 11:43 PM

#

just redirect gsi 9 to that stub as well

#

via ioapic

surreal path May 4, 2024, 11:43 PM

#

why gsi 9

finite summit May 4, 2024, 11:44 PM

#

yeah anyway I got a page table manager to debug

surreal path May 4, 2024, 11:44 PM

#

whats on gsi 9

kind root May 4, 2024, 11:44 PM

#

thats where acpi interrupt lives

#

the SCI interrupt

surreal path May 4, 2024, 11:44 PM

#

do u have a spec sheet i can read so i can see what gsis go where

#

for future use

kind root May 4, 2024, 11:44 PM

#

u have the madt table..?

surreal path May 4, 2024, 11:44 PM

#

oh right

kind root May 4, 2024, 11:45 PM

#

technically its not fixed to 9

#

but all computers do it that way for some reason

surreal path May 4, 2024, 11:45 PM

#

kind root just redirect gsi 9 to that stub as well

the common stub?

kind root May 4, 2024, 11:45 PM

#

any idt index u want

surreal path May 4, 2024, 11:45 PM

#

ill make a common stub

#

well i dont actually NEED to. i can use this stub i already have as a common stub in fact!

#

actually no

#

i cant

#

i need the int number

kind root May 4, 2024, 11:47 PM

#

ye

surreal path May 4, 2024, 11:47 PM

#

how tf am i gonna get that

#

cpu doesnt push that on the stack

thorn bramble May 4, 2024, 11:47 PM

#

push it urself

surreal path May 4, 2024, 11:48 PM

#

how am i gonna "push it myself". this is a common stub, i dont know what interrupt i am in this common stub

kind root May 4, 2024, 11:48 PM

#

use brain

thorn bramble May 4, 2024, 11:49 PM

#

i was about to spoonfeed :^)

kind root May 4, 2024, 11:49 PM

#

lol

finite summit May 4, 2024, 11:49 PM

#

bruh

surreal path May 4, 2024, 11:50 PM

#

thorn bramble i was about to spoonfeed :^)

dont

thorn bramble May 4, 2024, 11:50 PM

#

i am not

#

dw

surreal path May 4, 2024, 11:50 PM

#

i need to learn not copy code

thorn bramble May 4, 2024, 11:50 PM

#

but really there is only one way to do it

#

macros and the push instruction

#

go crazy

#

you can figure it out

surreal path May 4, 2024, 11:51 PM

#

i dont need a common stub as i already have one

#

i have a register interurpt function

#

am i fucking dumb

#

just create the isr stubs

#

wait

#

reading ur code @finite summit and why do u push %1-0x20

finite summit May 5, 2024, 12:03 AM

#

that's for my abstract irq interface

#

you can ignore it

surreal path May 5, 2024, 12:03 AM

#

okay

kind root May 5, 2024, 12:03 AM

#

surreal path reading ur code <@841762704766140486> and why do u push %1-0x20

first 32 numbers are reserved

#

for the legacy exceptions

finite summit May 5, 2024, 12:04 AM

#

yeah

surreal path May 5, 2024, 12:04 AM

#

?

finite summit May 5, 2024, 12:04 AM

#

you can ignore it

surreal path May 5, 2024, 12:04 AM

#

alr

kind root May 5, 2024, 12:04 AM

#

0xE - pagefault

#

0x0 - div by zero

#

etc

#

there's 32 of these

surreal path May 5, 2024, 12:04 AM

#

yea i know

kind root May 5, 2024, 12:04 AM

#

so u get it then?

#

it pushes the irq number

#

barring the reserved range of 0..31

surreal path May 5, 2024, 12:05 AM

#

kind root it pushes the irq number

oh i think i know why he does that

#

yea thats smart

#

then u can like have a common irq handler

#

and have that handler the irqs

#

😎

#

correct?

kind root May 5, 2024, 12:06 AM

#

yes

surreal path May 5, 2024, 12:07 AM

#

alr epic

#

im gonna implment this rn!!!

finite summit May 5, 2024, 12:07 AM

#

I say look into my code

#

and try to see what it does

#

abstract irq interface is under irq/irq.cpp

#

if you have any questions you can ping me

surreal path May 5, 2024, 12:09 AM

#

alr

surreal path May 5, 2024, 12:48 AM

#

bro

#

nooo

#

line 33 causes a general protecxtion fault now

#

cause i wanted to add irq to my interrupt frame

#

@finite summit pain

#

#

should work RIGHT?

#

wrong !

#

push everything correctly ! yet it still has the stack get fucked or smthin

surreal path May 5, 2024, 12:59 AM

#

surreal path line 33 causes a general protecxtion fault now

wait

thorn bramble May 5, 2024, 1:00 AM

#

why push it twice

thorn bramble May 5, 2024, 1:02 AM

#

surreal path bro

you can just call [idt_handlers + %1 * 8] btw

#

no need to store it in rax first

surreal path May 5, 2024, 1:18 AM

#

@finite summitexplain (uintptr_t)(&__b_isr_handler + (i * 32)

#

i know u aligned all the stubs

#

but like

#

u dont know the size of each stub

#

and im tryna do something similar but the addresses im getting are all wacky

#

finite summit May 5, 2024, 1:23 AM

#

Well all stubs here are pretty much guarenteed to be < than 32 bytes

#

so with the alignment, that makes all stubs be offset by 32 bytes

thorn bramble May 5, 2024, 1:25 AM

#

i'd recommend you build an array of pointers in assembly

#

just so you don't have to rely on the fact that each handler is 32 bytes with padding

#

instead the assembler builds the array for you with all the pointers

surreal path May 5, 2024, 1:26 AM

#

thorn bramble instead the assembler builds the array for you with all the pointers

hm?

thorn bramble May 5, 2024, 1:27 AM

#

stubs:
  dq isr_stub_0
  dq isr_stub_1
  dq isr_stub_2
  ; ...
  dq isr_stub_255```

#

except you can do it with some preprocessor magic

surreal path May 5, 2024, 1:27 AM

#

?

thorn bramble May 5, 2024, 1:27 AM

#

im sorry what are you confused about?

surreal path May 5, 2024, 1:27 AM

#

im thinking second

#

so ur dq'ing a label

#

so if u dq a label its just gonna be a pointer to that label?

#

correct?

thorn bramble May 5, 2024, 1:28 AM

#

the address of that label, yes

surreal path May 5, 2024, 1:28 AM

#

oh this is actually much smarter

thorn bramble May 5, 2024, 1:30 AM

#

i mean both ways are valid, but i prefer that one because i rely less on the code padding

surreal path May 5, 2024, 1:32 AM

#

ummm

thorn bramble May 5, 2024, 1:34 AM

#

good question

#

hard to say

#

without seeing the entire file

#

also what assembler is that

#

gas?

finite summit May 5, 2024, 1:34 AM

#

nasm

surreal path May 5, 2024, 1:34 AM

#

nasm

kind root May 5, 2024, 1:34 AM

#

thorn bramble hard to say

Look closer

thorn bramble May 5, 2024, 1:34 AM

#

.dq

#

or whatever

#

.quad

#

idk nasm

kind root May 5, 2024, 1:35 AM

#

He's not typing out the name of the label

#

But the macro name

thorn bramble May 5, 2024, 1:35 AM

#

is he not?

#

i thought these were the names of the labels

surreal path May 5, 2024, 1:35 AM

#

oh

#

💀

kind root May 5, 2024, 1:35 AM

#

isr_stub_9

#

Not isr_stub 9

thorn bramble May 5, 2024, 1:35 AM

#

ugh

#

oh dear lord

#

ur right

surreal path May 5, 2024, 1:36 AM

#

i mean its 2am

#

my brain is 💥

kind root May 5, 2024, 1:36 AM

#

U mean 4:30

thorn bramble May 5, 2024, 1:36 AM

#

3:36 here

finite summit May 5, 2024, 1:36 AM

#

it's 21:36 idk what yall mean

thorn bramble May 5, 2024, 1:36 AM

#

we all over the place

surreal path May 5, 2024, 1:36 AM

#

no its 02:36am 😠

#

i control the time

#

😤

finite summit May 5, 2024, 1:36 AM

#

is anyone here upside down

#

?

thorn bramble May 5, 2024, 1:37 AM

#

in this chat i dont think so

finite summit May 5, 2024, 1:37 AM

#

other than @plush hearth

thorn bramble May 5, 2024, 1:37 AM

#

oh i didnt know

finite summit May 5, 2024, 1:37 AM

#

he's upside down

thorn bramble May 5, 2024, 1:37 AM

#

i know that andy is an aussie

plush hearth May 5, 2024, 1:37 AM

#

finite summit he's upside down

haha. yes, I'm in Australia

finite summit May 5, 2024, 1:37 AM

#

what if they accidentally take off the things on them that make them not fall into space

surreal path May 5, 2024, 1:37 AM

#

oh frick u i cant use macros cause nasm hates that

#

time to write all these out manually

finite summit May 5, 2024, 1:38 AM

#

you mistyped

thorn bramble May 5, 2024, 1:38 AM

#

isr_stub_%q maybe

surreal path May 5, 2024, 1:38 AM

#

tried

plush hearth May 5, 2024, 1:38 AM

#

finite summit what if they accidentally take off the things on them that make them not fall in...

I have spoken to people on discord that actually believe this is true

thorn bramble May 5, 2024, 1:38 AM

#

yeah good luck idk nasm

surreal path May 5, 2024, 1:38 AM

#

finite summit May 5, 2024, 1:38 AM

#

plush hearth I have spoken to people on discord that actually believe this is true

lol

plush hearth May 5, 2024, 1:38 AM

#

surreal path no its 02:36am 😠

no, its 11:38 am

finite summit May 5, 2024, 1:38 AM

#

There can be some... delusional people on discord to say the least

#

well not just on discord

#

but yeah

surreal path May 5, 2024, 1:39 AM

#

surreal path

why cant i use %q angr

#

guess i have no other option but to write all these manually!!!

surreal path May 5, 2024, 2:00 AM

#

IT WORKS

#

IRETQ YOUR SO PRO

#

😎

#

i wrote a python script to write the dq's for me

#

it all works!!!

thorn bramble May 5, 2024, 2:05 AM

#

you are very much welcome

#

enjoy your objectively better way of making an isr table

cinder plinth May 5, 2024, 2:13 AM

#

username checks out

surreal path May 5, 2024, 2:14 AM

#

cinder plinth username checks out

😭

surreal path May 5, 2024, 2:48 AM

#

thorn bramble enjoy your objectively better way of making an isr table

yes i will!

#

mister master of interrupts !!!

desert haven May 5, 2024, 6:31 AM

#

kind root u disabled %zu

lmao, I hit the same issue. Although I caught it when I had a few %zus popping up in the logs.

surreal path May 5, 2024, 2:00 PM

#

halfmemeright

#

bro thats an unsigned int wtf

#

#

like bro.

kind root May 5, 2024, 2:03 PM

#

how are u printing it

surreal path May 5, 2024, 2:03 PM

#

its a uint64

kind root May 5, 2024, 2:03 PM

#

bro doesnt know how 2s complement works

surreal path May 5, 2024, 2:04 PM

#

the heck is that

kind root May 5, 2024, 2:04 PM

#

kind root how are u printing it

.

surreal path May 5, 2024, 2:04 PM

#

kind root how are u printing it

%d

#

printf

kind root May 5, 2024, 2:04 PM

#

jesus

#

%d is signed int

surreal path May 5, 2024, 2:05 PM

#

oh

#

💀

kind root May 5, 2024, 2:05 PM

#

%u is unsigned int

#

%lu is unsigned long

#

%ld is signed long

#

etc

#

u cant just put a %d there

tawdry mirage May 5, 2024, 2:05 PM

#

surreal path bro thats an unsigned int wtf

inb4 using %d

surreal path May 5, 2024, 2:05 PM

#

so in this case i need to do lu

tawdry mirage May 5, 2024, 2:05 PM

#

oh i was scrolled up

surreal path May 5, 2024, 2:05 PM

#

😭

kind root May 5, 2024, 2:06 PM

#

doesnt nanoprintf have __attribute__((format))

surreal path May 5, 2024, 2:06 PM

#

idk

tawdry mirage May 5, 2024, 2:06 PM

#

nanoprintf may do but the custom wrapper for writing to the screen/serial most likely doesn't

kind root May 5, 2024, 2:06 PM

#

surreal path idk

look at UACPI_PRINTF_DECL

#

reuse it for your own helper

surreal path May 5, 2024, 2:06 PM

#

troll

kind root May 5, 2024, 2:06 PM

#

then the compiler will scream at u if u do it wrong

tawdry mirage May 5, 2024, 2:06 PM

#

and the compiler can't peek into valist to determine whether format args make sense

surreal path May 5, 2024, 2:16 PM

#

#

the fact that i ignore pretty much every compiler warning

surreal path May 5, 2024, 2:17 PM

#

surreal path <:troll:796858892985434152>

anyways i dont know why the irq number is this giant mess

#

ima figure this out

tawdry mirage May 5, 2024, 2:22 PM

#

surreal path <:troll:796858892985434152>

>>> hex(18446744073709551598)
'0xffffffffffffffee'

#

sign extended byte 0xee?

#

do your asm irq stubs do push byte N by any chance

surreal path May 5, 2024, 2:23 PM

#

#

%1-0x20

ionic jetty May 5, 2024, 2:35 PM

#

surreal path

you dont always wanna push the error code

surreal path May 5, 2024, 2:36 PM

#

ionic jetty you dont always wanna push the error code

always push a dummy error code in that case

ionic jetty May 5, 2024, 2:37 PM

#

i dont see your whole macro but does it use the exact same one for every exception

surreal path May 5, 2024, 2:42 PM

#

ionic jetty i dont see your whole macro but does it use the exact same one for every excepti...

no

#

i have two types of stubs

tawdry mirage May 5, 2024, 2:43 PM

#

surreal path

wait why -0x20

surreal path May 5, 2024, 2:43 PM

#

for the ones that push an error code i dont push an error code

hasty sedge May 5, 2024, 2:43 PM

#

        push $0
    .endif```
you need something like this

tawdry mirage May 5, 2024, 2:43 PM

#

0xee+0x20 = 0x10e => 0xe => page fault

surreal path May 5, 2024, 2:43 PM

#

tawdry mirage wait why -0x20

cause exceptions 0-32 are reserved

ionic jetty May 5, 2024, 2:43 PM

#

surreal path for the ones that push an error code i dont push an error code

yeah thats correct i didnt see it fully

ionic jetty May 5, 2024, 2:43 PM

#

surreal path cause exceptions 0-32 are reserved

31

tawdry mirage May 5, 2024, 2:44 PM

#

surreal path cause exceptions 0-32 are reserved

and?

cinder plinth May 5, 2024, 2:44 PM

#

surreal path cause exceptions 0-32 are reserved

doesnt make sense

#

just push the interrupt number

#

no need for -0x20

#

for whatever reason

finite summit May 5, 2024, 2:44 PM

#

He took that from my kernel

surreal path May 5, 2024, 2:45 PM

#

cause this is for the like, irq number. i wanna implment something similar to omars implmentation of a irq dispatcher

finite summit May 5, 2024, 2:45 PM

#

and I said it was for my irq interface

ionic jetty May 5, 2024, 2:45 PM

#

exposed

surreal path May 5, 2024, 2:45 PM

#

yea, i wanna do something similar

thorn bramble May 5, 2024, 2:45 PM

#

cinder plinth no need for -0x20

thats what i told him lol

cinder plinth May 5, 2024, 2:45 PM

#

finite summit and I said it was for my irq interface

and why cant you do that calculation in C?

thorn bramble May 5, 2024, 2:45 PM

#

he has like two separate ones

finite summit May 5, 2024, 2:45 PM

#

cinder plinth and why cant you do that calculation in C?

Because reasons

cinder plinth May 5, 2024, 2:45 PM

#

literally just

#

irq_dispatch(int-0x20)

ionic jetty May 5, 2024, 2:46 PM

#

finite summit Because reasons

what reasons

cinder plinth May 5, 2024, 2:46 PM

#

lol

finite summit May 5, 2024, 2:46 PM

#

well

#

the irq dispatcher is hooked directly to the isr stub thing

thorn bramble May 5, 2024, 2:46 PM

#

bro cannot run sub rax,0x20 because reasons

finite summit May 5, 2024, 2:46 PM

#

and it's supposed to be completely arch-dependant

surreal path May 5, 2024, 2:46 PM

#

cinder plinth irq_dispatch(int-0x20)

actually ur right

#

i can just do that

finite summit May 5, 2024, 2:46 PM

#

It takes an interrupt_frame

#

which is arch-dependant

#

but has a field that should exist on all archs

#

intNumber

ionic jetty May 5, 2024, 2:47 PM

#

but you can still do intNumber - 0x20 in your isr can you

finite summit May 5, 2024, 2:47 PM

#

ionic jetty but you can still do intNumber - 0x20 in your isr can you

No

cinder plinth May 5, 2024, 2:47 PM

#

intNumber not irqNumber

finite summit May 5, 2024, 2:47 PM

#

because then it depends on the first 32 vectors being reserved

cinder plinth May 5, 2024, 2:47 PM

#

thats stupid

thorn bramble May 5, 2024, 2:47 PM

#

sounds silly to me

finite summit May 5, 2024, 2:47 PM

#

which is x86-specific

thorn bramble May 5, 2024, 2:47 PM

#

anyway

surreal path May 5, 2024, 2:47 PM

#

yea im changing it

#

if i want the irq number i just, intnum-0x20

#

in c

ionic jetty May 5, 2024, 2:48 PM

#

I had a look at your code a few weeks ago does this interrupt dispatcher create these dpc threads you have there?

finite summit May 5, 2024, 2:48 PM

#

indeed

ionic jetty May 5, 2024, 2:48 PM

#

thats neat

finite summit May 5, 2024, 2:49 PM

#

Except I never got a chance to use them

#

Because nothing except uACPI could use them

#

But uACPI needed them to be used in a way that's different than what my kernel does

thorn bramble May 5, 2024, 2:49 PM

#

btw you said you ignore all warnings anyway a while ago, you really shouldnt do that, warnings are emitted for good reasons and you should just take your time to read and fix them if you can

finite summit May 5, 2024, 2:49 PM

#

finite summit intNumber

wait it's called vector

surreal path May 5, 2024, 2:50 PM

#

thorn bramble btw you said you ignore all warnings anyway a while ago, you really shouldnt do ...

i will actually

finite summit May 5, 2024, 2:50 PM

#

thorn bramble btw you said you ignore all warnings anyway a while ago, you really shouldnt do ...

did I?

surreal path May 5, 2024, 2:50 PM

#

fix them warnings

thorn bramble May 5, 2024, 2:50 PM

#

please do

#

-Werror

#

:^)

surreal path May 5, 2024, 2:50 PM

#

hes talking about me lmao 😭

ionic jetty May 5, 2024, 2:50 PM

#

thorn bramble -Werror

I had that typed out already but you were faster 🤣

thorn bramble May 5, 2024, 2:50 PM

#

finite summit did I?

its nyaux thread not obos lol

#

but you should too if you dont :^)

finite summit May 5, 2024, 2:51 PM

#

I have all warnings enabled, just not -Werror to keep compatibility with future GCC versions

#

In case some future GCC version defines some new warning that my code triggers

surreal path May 5, 2024, 2:52 PM

#

ok i do the calculation in C now

thorn bramble May 5, 2024, 2:53 PM

#

learn from this guy :^)

thorn bramble May 5, 2024, 2:53 PM

#

surreal path ok i do the calculation in C now

0xe-0x20

surreal path May 5, 2024, 2:53 PM

#

yea

thorn bramble May 5, 2024, 2:53 PM

#

you know what happens if you subtract them?

surreal path May 5, 2024, 2:54 PM

#

yep

thorn bramble May 5, 2024, 2:54 PM

#

you get a negative result

cinder plinth May 5, 2024, 2:54 PM

#

frame->intnum >= 0x20 ? frame->intnum - 0x20 : idk something

surreal path May 5, 2024, 2:54 PM

#

thorn bramble you get a negative result

yea i know

thorn bramble May 5, 2024, 2:54 PM

#

just print intnum

#

no subtraction

cinder plinth May 5, 2024, 2:54 PM

#

yes

#

that

surreal path May 5, 2024, 2:56 PM

#

since i use exception 32 for my lapic timer and scheduler

#

im gonna do 0x21

kind root May 5, 2024, 2:57 PM

#

Why

#

Lapic timer is just normal irq

surreal path May 5, 2024, 2:58 PM

#

yea it is, wait sorry im stupid 💀

#

give me a second let me think this through

#

idk why im so slow today

#

i think its cause i didnt sleep for much

leaden hinge May 5, 2024, 2:58 PM

#

exception 32?

surreal path May 5, 2024, 2:58 PM

#

thats not an exception right

leaden hinge May 5, 2024, 2:58 PM

#

its not one of the cpu exception vectors no

surreal path May 5, 2024, 2:59 PM

#

sorry im doing a lot of things that are stupid

#

just let me

#

think about this

#

okay heres what im gonna do

thorn bramble May 5, 2024, 2:59 PM

#

you’re gonna print intnum

surreal path May 5, 2024, 2:59 PM

#

no

thorn bramble May 5, 2024, 3:00 PM

#

welp

surreal path May 5, 2024, 3:00 PM

#

why

#

why would i do that?

thorn bramble May 5, 2024, 3:00 PM

#

because thats all you need

#

why subtract anything

surreal path May 5, 2024, 3:00 PM

#

ok wait

#

i need to think about this very hard

#

i need to activate my two brain cells

#

galaxybrain

#

ok

#

so

#

we have a global irq dispatcher, right.

void irq_dispatcher(int irq)
{
  ...
}

okay and this irq dispatcher has an array of function pointers of MAX_IRQ_HANDLERS or smthin

#

if i want to register an irq

#

RegisterInterruptHandler(vector, global_irq_dispatacher);

#

void global_irq_dispatacher(struct interrupt_frame *frame)
{
  irq_dispatacher(frame->intnum - 0x20);
}

thorn bramble May 5, 2024, 3:05 PM

#

wha

#

i mean whatever works for you

#

but sure

surreal path May 5, 2024, 3:05 PM

#

if theres a better design for this

#

tell me

#

i want to learn more and know

thorn bramble May 5, 2024, 3:06 PM

#

well im on phone but ill get on pc soon

surreal path May 5, 2024, 3:06 PM

#

oka

#

okay i get it now

#

thanks iretq

surreal path May 5, 2024, 4:51 PM

#

thanks to huge help from iretq i made the ability to route interrupts and stuff from the ioapic

#

working on ps2 driver when i get back from store to get bar of choclate

surreal path May 5, 2024, 8:09 PM

#

ps2 driver done

#

again thanks to huge help from iretq for helping me understand a lot of concepts

#

well um....

#

vfs now???

#

trl nyaux panics if no one consumes the keyboard event buffer and its full trl trl

#

probs should make it clear out the buffer or smthin

kind root May 5, 2024, 8:14 PM

#

surreal path ps2 driver done

pls tell me you dont hardcode ps/2 irqs and io ports

#

u have literal uacpi to figure that out

thorn bramble May 5, 2024, 8:14 PM

#

oh yeah i forgot about that

#

@surreal path have fun

#

implementing the interrupt stuff

#

for uacpi

#

:^)

kind root May 5, 2024, 8:15 PM

#

u dont even need that

wicked loom May 5, 2024, 8:15 PM

#

it's fine to hardcode ISA devices IRQs, MMIOs, and ports

#

on standard PCs

kind root May 5, 2024, 8:15 PM

#

to find out whether ps/2 exists and where it lives

wicked loom May 5, 2024, 8:15 PM

#

i mean of course if you have a ACPI impl it's better to use that

kind root May 5, 2024, 8:15 PM

#

wicked loom on standard PCs

qemu micro-pc or whatever doesnt have that i think

wicked loom May 5, 2024, 8:15 PM

#

it's the proper™️ way of detecting ISA devices

thorn bramble May 5, 2024, 8:15 PM

#

oh i thought that there was a way to ask uacpi to route the irq for you

surreal path May 5, 2024, 8:15 PM

#

kind root to find out whether ps/2 exists and where it lives

how could i use uacpi to do that?

kind root May 5, 2024, 8:16 PM

#

surreal path how could i use uacpi to do that?

wiki finding devices section

wicked loom May 5, 2024, 8:16 PM

#

kind root qemu micro-pc or whatever doesnt have that i think

that's why i said standard PCs. for example Vinix checks if your hardware is ACPI reduced and refuses to boot

kind root May 5, 2024, 8:16 PM

#

lol

wicked loom May 5, 2024, 8:16 PM

#

because it hardcodes the resources of ISA devices

kind root May 5, 2024, 8:16 PM

#

theres a AT_PC compliant flag in fadt

#

if u really wanna be pedantic

wicked loom May 5, 2024, 8:16 PM

#

is that better?

kind root May 5, 2024, 8:16 PM

#

yeah

wicked loom May 5, 2024, 8:16 PM

#

i did not know about any better method of detecting standard PCs back then

kind root May 5, 2024, 8:17 PM

#

you cant assume isa exists without checking that

wicked loom May 5, 2024, 8:17 PM

#

wicked loom i did not know about any better method of detecting standard PCs back then

as in, better than checking for ACPI reduced hw

kind root May 5, 2024, 8:17 PM

#

wicked loom as in, better than checking for ACPI reduced hw

wicked loom May 5, 2024, 8:17 PM

#

these are separate flags though

kind root May 5, 2024, 8:18 PM

#

ye i thought it was just the one

#

guess not

#

bit 0 is enough probably

wicked loom May 5, 2024, 8:18 PM

#

so it's not AT_PC, it's IA_PC

#

hell it's probably wise to check both that, and !reduced

kind root May 5, 2024, 8:19 PM

#

maybe there's some other flag im forgetting idk

wicked loom May 5, 2024, 8:19 PM

#

after that check it's fine to assume ISA devices are where they are

#

that's kinda the point of backwards compat luckly

#

and yeah of course if you have uACPI or something else you should detect them dynamically via ACPI

#

i am talking for Vinix or other OSes that do not into ACPI

thorn bramble May 5, 2024, 8:20 PM

#

i wonder if there is an acpi way to find out which irq the keyboard uses?

kind root May 5, 2024, 8:20 PM

#

yes!

thorn bramble May 5, 2024, 8:20 PM

#

ofc its unlikely to be != 1

kind root May 5, 2024, 8:20 PM

#

uacpi_get_current_resources()

kind root May 5, 2024, 8:20 PM

#

thorn bramble i wonder if there is an acpi way to find out which irq the keyboard uses?

here:

Name (_CRS, ResourceTemplate ()  // _CRS: Current Resource Settings
{
    IO (Decode16,
        0x0060,             // Range Minimum
        0x0060,             // Range Maximum
        0x01,               // Alignment
        0x01,               // Length
        )
    IO (Decode16,
        0x0064,             // Range Minimum
        0x0064,             // Range Maximum
        0x01,               // Alignment
        0x01,               // Length
        )
    IRQ (Edge, ActiveHigh, Exclusive, )
        {1}
})

#

this is a ps/2 _CRS dump from some AML blob

thorn bramble May 5, 2024, 8:21 PM

#

oh that is amazing

kind root May 5, 2024, 8:21 PM

#

this means IO port 0x60, 0x64, and irq 1

#

uacpi provides a way to retrieve these

thorn bramble May 5, 2024, 8:21 PM

#

oh and i use uacpi_get_current_resources and find the irq resource?

#

that is so cool

kind root May 5, 2024, 8:22 PM

#

yeah

#

theres a uacpi_for_each_resource

surreal path May 5, 2024, 8:22 PM

#

is PNP0303 the id for the ps2 keyboard

thorn bramble May 5, 2024, 8:22 PM

#

acpi is such a bitch

#

but when it works

#

its beautiful

kind root May 5, 2024, 8:22 PM

#

surreal path is PNP0303 the id for the ps2 keyboard

yup

#

lol

#

here's the entire definition

Device (PS2K)
{
    Name (_HID, EisaId ("PNP0303") /* IBM Enhanced Keyboard (101/102-key, PS/2 Mouse) */)  // _HID: Hardware ID
    Method (_STA, 0, NotSerialized)  // _STA: Status
    {
        If (((PKBD == 0xFF) & (PKBC == 0xFF)))
        {
            Return (Zero)
        }

        Return (0x0F)
    }

    Name (_CRS, ResourceTemplate ()  // _CRS: Current Resource Settings
    {
        IO (Decode16,
            0x0060,             // Range Minimum
            0x0060,             // Range Maximum
            0x01,               // Alignment
            0x01,               // Length
            )
        IO (Decode16,
            0x0064,             // Range Minimum
            0x0064,             // Range Maximum
            0x01,               // Alignment
            0x01,               // Length
            )
        IRQ (Edge, ActiveHigh, Exclusive, )
            {1}
    })

surreal path May 5, 2024, 8:25 PM

#

kind root theres a uacpi_for_each_resource

how do i use that

#

wiki doesnt say much about it

kind root May 5, 2024, 8:25 PM

#

read code?

#

its trivial

#

u give it a callback

surreal path May 5, 2024, 8:25 PM

#

and what is this callback for

kind root May 5, 2024, 8:26 PM

#

if the function is called

#

for each resource

#

what do you think the callback receives

surreal path May 5, 2024, 8:26 PM

#

i dont know, i just want the irq number and port for the keyboard so i can route it

kind root May 5, 2024, 8:26 PM

#

right

#

u want me to spoonfeed?

surreal path May 5, 2024, 8:26 PM

#

no

#

i want to understand

kind root May 5, 2024, 8:27 PM

#

the keyboard has 3 resources

#

io port 0

#

io port 1

#

irq

#

you iterate the resources uacpi gives u

#

look at the resource type

#

if its an io resource u record the io port

#

if its an irq resource u record the irq

#

done

surreal path May 5, 2024, 8:28 PM

#

oh okay

kind root May 5, 2024, 8:28 PM

#

same for the mouse btw

#

Device (PS2M)
{
    Name (_HID, EisaId ("PNP0F13") /* PS/2 Mouse */)  // _HID: Hardware ID
    Name (_CRS, ResourceTemplate ()  // _CRS: Current Resource Settings
    {
        IRQ (Edge, ActiveHigh, Exclusive, )
            {12}
    })

kind root May 5, 2024, 8:30 PM

#

surreal path how do i use that

u dont have to use that btw

#

u can just iterate them manually

#

its just a bit more annoying because they're variable length

surreal path May 5, 2024, 8:30 PM

#

ok how do i get the resource

kind root May 5, 2024, 8:30 PM

#

its provided to u in the callback

surreal path May 5, 2024, 8:30 PM

#

oh

#

thats node?

kind root May 5, 2024, 8:31 PM

#

uacpi_resource_iteration_decision my_callback(void *user, uacpi_resource *resource);

surreal path May 5, 2024, 8:31 PM

#

oh

thorn bramble May 5, 2024, 8:32 PM

#

kind root here's the entire definition ``` Device (PS2K) { Name (_HID, EisaId ("PNP030...

also it's pretty clear what the irq is for, as there is only one and it serves one purpose, but how do you make use of the port io resources? they dont seem to have a description or type or flags or whatever

#

sorry for a weird question but i dont know how to ask it properly

kind root May 5, 2024, 8:32 PM

#

thorn bramble also it's pretty clear what the irq is for, as there is only one and it serves o...

well you look for a specific device right

#

aka a ps/2 keyboard

thorn bramble May 5, 2024, 8:32 PM

#

yup

kind root May 5, 2024, 8:32 PM

#

u know how to use the ports

#

and how many to expect

thorn bramble May 5, 2024, 8:32 PM

#

but is it like, i know there are two ports

#

and i know the first port is the data port?

kind root May 5, 2024, 8:33 PM

#

and for the general kernel resource collision detection u dont care about what they're used for

kind root May 5, 2024, 8:33 PM

#

thorn bramble and i know the first port is the data port?

yeah

#

the order and types are generally fixed

thorn bramble May 5, 2024, 8:33 PM

#

so i just assume port_io_resources[0] is a data port, and instead of inb(0x60) i do inb(port_io_res[0].port)

tawdry mirage May 5, 2024, 8:33 PM

#

kind root here's the entire definition ``` Device (PS2K) { Name (_HID, EisaId ("PNP030...

you just need to maintain a list of special _HID names for some devices in case the aml does something odd, like say, renaming the device on windows 8 and above meme

kind root May 5, 2024, 8:33 PM

#

agreed upon so to speak

kind root May 5, 2024, 8:33 PM

#

thorn bramble so i just assume `port_io_resources[0]` is a data port, and instead of `inb(0x60...

yup

kind root May 5, 2024, 8:34 PM

#

tawdry mirage you just need to maintain a list of special _HID names for some devices in case ...

lmao yeah

surreal path May 5, 2024, 8:34 PM

#

bruh

thorn bramble May 5, 2024, 8:34 PM

#

is there any benefit to doing that opposed to just hardcoding the isa resources?

#

is there, realistically, hardware out there that deviates?

#

does windows or linux do that? lol

kind root May 5, 2024, 8:35 PM

#

idk if there's hardware that does that honestly

#

but technically nothing prevents me from making an emulator that does that

#

and windows/linux will work there

tawdry mirage May 5, 2024, 8:35 PM

#

the technically-correct pc emulator

kind root May 5, 2024, 8:35 PM

#

because they dont hardcode the values

kind root May 5, 2024, 8:35 PM

#

thorn bramble is there, realistically, hardware out there that deviates?

also devices have a _SRS method

#

sometimes

#

which allows u to override these

#

so u could tell the hardware to use a different irq

#

in that case _CRS will return a different irq every time

surreal path May 5, 2024, 8:36 PM

#

page faulting on this instruction

#

#

thorn bramble May 5, 2024, 8:40 PM

#

kind root in that case _CRS will return a different irq every time

is _CRS like a method you call on a resource to get the value of it or whatever

kind root May 5, 2024, 8:40 PM

#

thorn bramble is _CRS like a method you call on a resource to get the value of it or whatever

it returns an encoded list of resources consumed/produced by the device at the moment

surreal path May 5, 2024, 8:40 PM

#

surreal path

it faults cause of my callback?

kind root May 5, 2024, 8:41 PM

#

no idea

surreal path May 5, 2024, 8:41 PM

#

kind root May 5, 2024, 8:43 PM

#

i dont think it dies in your callback

#

look at the trace you posted

#

    if (spec->extra_size_for_native)
        final_size += spec->extra_size_for_native(spec, data, size);

#

it dies here

#

some big memory corruption as usual

surreal path May 5, 2024, 8:43 PM

#

offff course

kind root May 5, 2024, 8:44 PM

#

it specifically checks that the function pointer is not null

#

but then its null somehow

#

when called

#

remember the printf technique i taught u

surreal path May 5, 2024, 8:47 PM

#

spec seems to be a valid address

#

data seems to not be null

#

size isnt null

kind root May 5, 2024, 8:47 PM

#

what about the callback

#

pointer

surreal path May 5, 2024, 8:49 PM

#

valid

#

kind root May 5, 2024, 8:49 PM

#

how is it null when called lol

kind root May 5, 2024, 8:49 PM

#

kind root ``` if (spec->extra_size_for_native) final_size += spec->extra_size_...

im talking about this place @surreal path

#

not your code

surreal path May 5, 2024, 8:49 PM

#

no this isnt my code

#

this is in uacpI_find_devices

kind root May 5, 2024, 8:50 PM

#

not that place

surreal path May 5, 2024, 8:50 PM

#

i dont know what the callback is in that function ur refering to

kind root May 5, 2024, 8:50 PM

#

final_size += spec->extra_size_for_native(spec, data, size);

surreal path May 5, 2024, 8:51 PM

#

is final_size the callback or what

#

sorry im very confused

kind root May 5, 2024, 8:51 PM

#

bro

#

do u see the circular brackets

surreal path May 5, 2024, 8:51 PM

#

yes

kind root May 5, 2024, 8:51 PM

#

what does that mean

surreal path May 5, 2024, 8:51 PM

#

thats the callback

kind root May 5, 2024, 8:51 PM

#

you guessed it!

surreal path May 5, 2024, 8:52 PM

#

valid

#

kind root May 5, 2024, 8:52 PM

#

yeah looks like further debugging is required

surreal path May 5, 2024, 8:53 PM

#

seems like getting the resource

kind root May 5, 2024, 8:54 PM

#

didnt we already figure that out by looking at the stack trace

#

it dies when calling that callback

#

that you just inspected

kind root May 5, 2024, 8:54 PM

#

surreal path

here

#

but the callback is valid

#

so something very funky is going on

surreal path May 5, 2024, 8:55 PM

#

i dont know but in the stack trace for some reason the rip is 0

#

like straight up 0

#

same with the rax

#

and cr2 is 0

kind root May 5, 2024, 8:55 PM

#

yeah

surreal path May 5, 2024, 8:55 PM

#

stack fuckery?

kind root May 5, 2024, 8:55 PM

#

perhaps

#

u disable red zone right?

surreal path May 5, 2024, 8:55 PM

#

red zone?

kind root May 5, 2024, 8:55 PM

#

-mno-redzone

surreal path May 5, 2024, 8:55 PM

#

yea

#

kind root May 5, 2024, 8:56 PM

#

could be a stack overflow or something

surreal path May 5, 2024, 8:56 PM

#

why

kind root May 5, 2024, 8:56 PM

#

no idea

#

im just guessing

surreal path May 5, 2024, 8:57 PM

#

0x6?

#

what

#

here

#

thats the "invalid opcode"

kind root May 5, 2024, 8:59 PM

#

your qemu doesn't have popcount i guess

surreal path May 5, 2024, 8:59 PM

#

why

kind root May 5, 2024, 8:59 PM

#

-cpu max

surreal path May 5, 2024, 8:59 PM

#

ok

kind root May 5, 2024, 9:00 PM

#

or -cpu host

surreal path May 5, 2024, 9:00 PM

#

no fault

kind root May 5, 2024, 9:00 PM

#

surreal path 0x6?

how did u even get here

surreal path May 5, 2024, 9:00 PM

#

kind root how did u even get here

-d int

kind root May 5, 2024, 9:00 PM

#

why didnt your kernel catch the invalid opcode

surreal path May 5, 2024, 9:01 PM

#

i dont have a handler for it

kind root May 5, 2024, 9:01 PM

#

its a trap so it should've just been an infinite recursion

surreal path May 5, 2024, 9:02 PM

#

its spamming zero on the screen

#

oh

#

when i try to get the irq

kind root May 5, 2024, 9:02 PM

#

these functions all have a return value you know

#

but who cares about compiler warnings

surreal path May 5, 2024, 9:03 PM

#

why is the irq 0

#

when its clearly 1

kind root May 5, 2024, 9:03 PM

#

no idea

#

first fix ub

surreal path May 5, 2024, 9:04 PM

#

ub?

kind root May 5, 2024, 9:04 PM

#

you dont return anything

surreal path May 5, 2024, 9:04 PM

#

0 is 2

kind root May 5, 2024, 9:05 PM

#

and

#

thats not what that irq number is

surreal path May 5, 2024, 9:05 PM

#

ok

kind root May 5, 2024, 9:05 PM

#

fix code

#

return the value that the function expects

surreal path May 5, 2024, 9:06 PM

#

ok

kind root May 5, 2024, 9:06 PM

#

its not supposed to go on forever

surreal path May 5, 2024, 9:08 PM

#

yea returning it nothing prints

kind root May 5, 2024, 9:08 PM

#

thats not the correct enum value

#

for the first function

surreal path May 5, 2024, 9:08 PM

#

what is it

kind root May 5, 2024, 9:08 PM

#

literally look

surreal path May 5, 2024, 9:09 PM

#

ok found it

kind root May 5, 2024, 9:09 PM

#

the return value types are different

surreal path May 5, 2024, 9:09 PM

#

now its spamming 0

kind root May 5, 2024, 9:09 PM

#

what did u do

surreal path May 5, 2024, 9:09 PM

#

kind root May 5, 2024, 9:10 PM

#

lol

surreal path May 5, 2024, 9:10 PM

#

it goes on forever

kind root May 5, 2024, 9:11 PM

#

like

#

enable fucking compiler warnings

surreal path May 5, 2024, 9:11 PM

#

ok

kind root May 5, 2024, 9:11 PM

#

i see the bug right there

surreal path May 5, 2024, 9:12 PM

#

oh

kind root May 5, 2024, 9:12 PM

#

yup

surreal path May 5, 2024, 9:12 PM

#

gotta cast it

#

ig

kind root May 5, 2024, 9:12 PM

#

what

surreal path May 5, 2024, 9:12 PM

#

what no

#

ok fixed

kind root May 5, 2024, 9:12 PM

#

think harder

surreal path May 5, 2024, 9:13 PM

#

#

resource port 257

#

thats the irq?

kind root May 5, 2024, 9:13 PM

#

port?

#

what

surreal path May 5, 2024, 9:13 PM

#

kind root May 5, 2024, 9:13 PM

#

can you show code

surreal path May 5, 2024, 9:14 PM

#

ok

#

static uacpi_resource_iteration_decision ps2_res(void *user, uacpi_resource *resource)
{
    if (resource->type == UACPI_RESOURCE_TYPE_IRQ)
    {
        kprintf("resource irq %d\n", resource->irq);
    }
    return UACPI_RESOURCE_ITERATION_CONTINUE;
}
static uacpi_ns_iteration_decision found_ps2(void *user, uacpi_namespace_node *node)
{
    uacpi_resources *kb_res;
    kprintf("Bruh\n");
    uacpi_status ret = uacpi_get_current_resources(node, &kb_res);
    kprintf("Um\n");
    if (uacpi_unlikely_error(ret)) {
        kpanic("unable to retrieve Get PS2 resources", NULL);
        return UACPI_NS_ITERATION_DECISION_NEXT_PEER;
    }
    uacpi_for_each_resource(kb_res, ps2_res, NULL);
    return UACPI_NS_ITERATION_DECISION_CONTINUE;
 
}
void ps2_init()
{
    uacpi_find_devices(PS2K_PNP_ID, found_ps2, NULL);
    for (;;)
    {
        asm ("hlt");
    }
}

#Nyaux