and then idk
#Nyaux
kind root
Yeah we'll work that out
kind root
will there be an OS to complement Hyper and uACPI?
Well I have the UltraOS
Those were supposed to be part of that
wicked loom
if it works, i'll track upstream and bump the commit whenever new commits come o...
i already do this for several things so one more isn't a big deal
unkempt relic
Well I have the UltraOS
oh yeah, i remember it now
kind root
We could make a limine->hyper Abi for providing a predefined config
unkempt relic
https://github.com/UltraOS/UltraLegacy/blob/master/Kernel/ACPI/ACPI.cpp uACPI has certainly grown
kind root
Lmao
wicked loom
as i mentioned above, i was thinking of just having hyper scan the efi vars for a predefined config
and Limine would construct the config and put it into an efivar on the fly based on the Limine config options
kind root
as i mentioned above, i was thinking of just having hyper scan the efi vars for ...
Oh yeah
kind root
and Limine would construct the config and put it into an efivar on the fly based...
Yup
Makes sense
wicked loom
i do wonder though
kind root
also idk if i should do `protocol: hyper` or `ultra` (or allow both)
Well the protocol is ultra
So idk
You could just boot into hyper directly for hyper protocol
wicked loom
it would be nice to also pass the kernel and modules or whatever directly (as in, have Limine itself load them into memory)
so they can also be loaded off of ext4 or whatever else Limine supports, using Limine style paths
unkempt relic
yeah
wicked loom
i readded it a while back
unkempt relic
i thought you quit ext
elder shoal
It was brought back but experimental™️
wicked loom
^
unkempt relic
zfs next in line
kind root
i readded it a while back
Why is that?
wicked loom
the puppy linux creator bitched enough to make me wanna add it back
kind root
not sure about BIOS lol
Same thing but something else instead of efi vars
wicked loom
then we could use the same method on both
kind root
True
molten grotto
I don't think efi vars are that great for potentially large temporary one-boot data anyway if its stored to some flash (unless there is a way to have non-persistent variables)
kind root
Anyways we'll work out the details but the idea overall sounds pretty interesting
unkempt relic
a missed opportunity with EFI's design in general:
amiga's rigid disk block allowed you to embed filesystem drivers with your partitions
no one would hate and everyone would love pluggable FS drivers for EFI
wicked loom
I don't think efi vars are that great for potentially large temporary one-boot d...
that's true
wicked loom
Anyways we'll work out the details but the idea overall sounds pretty interestin...
yeah! i like this
kind root
amiga's rigid disk block allowed you to embed filesystem drivers with your parti...
Interesting
unkempt relic
Interesting
amiga was just insanely based in many respects in general
a toast to object oriented design: you can install a DataType module for WebP into your Amiga running AmigaOS from the early 90s and basically all amiga programs that deal with graphics can instantly work with WebPs
kind root
Njce, too ahead of its time ig
steady flume
i was able to find a terminal emulator on an iphone i was gifted, unfortunately however that was a long time ago and i forgor 💀
with that, you might be able to set up sshd on your home computer then ssh in on your phone when you need/want to
don't forget flanterm :^)
thorn bramble
i was able to find a terminal emulator on an iphone i was gifted, unfortunately ...
ish?
surreal path
good mornin
to all
i still havent figured out the nasm crate yet
but im tryinggggg
surreal path
just use global_asm!
I don’t know how macros work with gas syntax and it looks really weird and confusing
also on the plane will be home to Ireland in like 3-4 hours
steady flume
ish?
actually, yeah, it was,
https://ish.app
thorn bramble
yea its kinda ass
frigid cliff
I don’t know how macros work with gas syntax and it looks really weird and confu...
Just use rust macros™️
shut laurel
yea its kinda ass
i mean its ok if you just want to have access to a normal ssh with no features paywalled like the ssh clients on the appstore have
flat nymph
iOS is annoying for SSH because of Apple's background apps' limitations
flat nymph
Android doesn't have those
(Although I think they might've also started doing that)
Like you could have an ssh session going for days
Whereas on iOS it would get closed if the app is running in background
marble surge
iOS is annoying for SSH because of Apple's background apps' limitations
iirc the ish thingie let you access the location in background (something like cat /dev/location > /dev/null&) and if you did that it ran in the background with no problem at all
Though idk I've never used ish myself nor iPhone
flat nymph
You can just ssh into your machine at home
flat nymph
You can just ssh into your machine at home
I think you even get easy to configure tunnels through VS Code
Or you can use darknets
(if you can't open ports)
marble surge
What's the issue at hand?
plush hearth
Keep in mind that the error code only gets pushed for some exceptions
surreal path
Keep in mind that the error code only gets pushed for some exceptions
i know
cool
now i need to just make an array of function pointers that global or whatevs
or maybe
i can just store the interrupt number
and have a global function that dispatches the
crap towards which function
yea maybe that'd work
flat nymph
I have some GAS assembly macros in my OS which create an array of functions that push interrupt number
surreal path
I have some GAS assembly macros in my OS which create an array of functions that...
hm?
thorn bramble
they use assembler macros to generate isr stubs that set up an interrupt frame with a consistent layout so a generic interrupt handler can be invoked
exactly what you wanted to do
surreal path
exactly what you wanted to do
im working on it dw
surreal path
having trouble getting a variable using gas syntax
or even using my macros
.extern function_handlers
.macro interrupterror num
.endm
.global stubs
stubs:
.long interrupterror
this doesnt work i do not know why
keeps saying undefined symbol
finite summit
so define the symbol
surreal path
and you marked it with pub and extern "C" + #[no_mangle]?
Now I did
But now I’m getting macro problems
Fun
flat nymph
(plus this https://gitlab.com/mishakov/pmos/-/blob/main/kernel/arch/x86_64/interrupts/asm_helpers.S?ref_type=heads)
(plus some more macros)
frigid cliff
You don’t need and to do interrupts with rust, or really anything as long as it has inline asm and naked functions
I can’t remove the embed because I’m on mobile 
The x86-interrupt ABI is unstable, but with rust it can be helpful for this stuff
surreal path
You don’t need and to do interrupts with rust, or really anything as long as it ...
what about returning
where do u
pop everything out
in your inline asm
i do not see
nvm im blind
frigid cliff
It’s somewhere in x86_64/mod.rs but I’m kinda bad about dealing with registers so I’m not sure if I do 😅
surreal path
u have a function called syscall, what does sym mean? what are u trying to do in your syscall handler
what is a virtual pointer ive never seen something like that
(i come from C so beware)
frigid cliff
syscall is the function I pass into the 0x80 interrupt (it’s kinda garbage tho so it can be ignored if you want) sym passes a function to an asm block as a pointer iirc
molten grotto
u have a function called syscall, what does sym mean? what are u trying to do in...
sym gives the symbol name of the thing you give to it (meaning the mangled name if its a mangled symbol and the plain name otherwise)
surreal path
i see
frigid cliff
what is a virtual pointer ive never seen something like that
VirtualPtr is a *mut _ wrapper that lets me know if it’s a pointer that’s mapped or not
frigid cliff
Yes, pretty much
surreal path
also what are you doing here
frigid cliff
Ah yes, that cluster fuck of a macro
I’m generally using it to give an error code to my exception handler (should be just above that), I should probably rework that so it calls panic or something but it’s a little bit of a mess
Rdi is the register that rust uses for the first argument so when I push the number onto the stack, I’m taking the stack pointer which now points to that number and moving it into rdi since it now points to the interrupt number I passed to the macro and so the exception function knows what exception happened
frigid cliff
I think I stole it from Felix lol
surreal path
im sorry i dont do things the rust way 💀
i should maybe learn rust macros more
but anyways
Nyaux needs to be worked on
frigid cliff
It’s changed since I stole it but it’s pretty similar to their implementation
Macros are annoying, but you get used to them once you are eventually forced to use them because rust doesn’t have a stable way to pass variadic arguments to a function without a macro
molten grotto
rust macros are pretty nice in comparison to c's lol
frigid cliff
Oh yes 100000%
surreal path
i would know
molten grotto
also that push_gprs function seems a bit weird, as like the compiler can save some stuff to the stack on the entry to there and then its supposed to restore it at the end but then you pushed some random stuff to there and told the compiler that you didn't use the stack
frigid cliff
Yeah, my handling of registers is kinda terrible and I knew that xd
molten grotto
and push_gprs might be called too late after the compiler already used some of the registers for something
surreal path
and push_gprs might be called too late after the compiler already used some of t...
is there a way to make the compiler not do anything
molten grotto
make the handler entry fn naked and then instead of push_gprs being a function either just inline the asm there manually or make it a macro
frigid cliff
is there a way to make the compiler not do anything
#[naked]
frigid cliff
Can I see the whole macro?
surreal path
yep
macro_rules! exception_function {
($code:expr, $handler:ident) => {
#[inline(always)]
#[naked]
extern "C" fn $handler() {
unsafe {
core::arch::asm!(
"push rax",
"push rbx",
"push rcx",
"push rdx",
"push rsi",
"push rdi",
"push rbp",
"push r8",
"push r9",
"push r10",
"push r11",
"push r12",
"push r13",
"push r14",
"push r15",
"push {0:r}",
"mov rdi, rsp",
"call {1}",
"pop {0:r}",
"mov rsp, rdi",
"pop r15",
"pop r14",
"pop r13",
"pop r12",
"pop r11",
"pop r10",
"pop r9",
"pop r8",
"pop rbp",
"pop rdi",
"pop rsi",
"pop rdx",
"pop rcx",
"pop rbx",
"pop rax",
in(reg) $code,
sym exception_handler,
options(noreturn)
);
};
loop {
}
}
};
}
And what’s the error it says??
surreal path
only const and sym operands are supported in naked functions
frigid cliff
Oh yeah
surreal path
wait i solved it by using const
frigid cliff
In main.rs/lib.rs?
surreal path
yep
surreal path
yep
surreal path
Can I see perchance?
shut laurel
I can’t remove the embed because I’m on mobile <:nooo:750466870993223750>
you can use <> around links I think
surreal path
double fault sadly :c
frigid cliff
frigid cliff
you can use <> around links I think
This is good to know, thanks
steady flume
I can’t remove the embed because I’m on mobile <:nooo:750466870993223750>
i'm pretty sure if you put angled brackets around a link, it removes the embed :)
<https://www.youtube.com/watch?v=dQw4w9WgXcQ>
oh 💀
frigid cliff
not the rick roll link xd
steady flume
i mean, it's the perfect use case
alpine mulch
get well soon man dw about it
surreal path
get well soon man dw about it
actually i will get some work done, just a bit sick so not a lot but ill try my best to get something done
alpine mulch
Don't pressure yourself, that's the important part
Do only what you feel like doing if you feel like doing it
surreal path
general protection fault over here
surreal path
Do only what you feel like doing if you feel like doing it
yea i agree
but i wanna work on it
surreal path
general protection fault over here
on this line for some odd reason
not when initing the idt
but when using it
manually interrupting the cpu
even tho
the IDTR
looks correct
this is very odd
it looks valid
further debugging doesnt show much
other then the fact that its general protection faults when an interrupt ever occurs
and the backtrace shows that its happening when an interrupt occurs
general protection error code is 0 so nothing segment related
not fun
size is 4096
which is expected, 16 bytes * 256 is 4096
more or less a single page
very odd
banging my head againist the wall trying to figure this out lol
surreal path
what does qemu think of your idt?
idt looks very odd
surreal path
thats not valid
kind root
Do only what you feel like doing if you feel like doing it
I should say this to my boss
desert haven
but the base address is totally wrong
surreal path
also you limit is 1 byte higher than it should be
?
desert haven
limit is the highest usable byte of the table (same goes for gdt)
surreal path
right
desert haven
so 0 -> 0xFFF is 0x1000 bytes
desert haven
did you fix the base address?
surreal path
the bass address looks correct in the IDTR
desert haven
what about in qemu?
surreal path
still completely stupid
desert haven
are you packing the idtr struct?
desert haven
I dont know rust, but I imagine its inserting some padding after the limit field
ah yeah, it would be then
desert haven
oh lol
surreal path
😎
also ordered some kfc
😎 😎
everythings works nicely
even the saving of the registers
nice
now we have an IDT
whats nextttt uhhhh
uacpi integration now??? hmmm
surreal path
no
i have a heap allocator, pmm, vmm as well
i can allocate virtual addresses and shi
same implmentation from old nyaux
surreal path
got the lapic address
desert haven
Almost! You forgot to mask out some flag bits in the msr
median goblet
@desert haven out of curiosity
can kernel software simply assume the lapic lives at 0xfee00000
desert haven
I mean I wouldnt advise it, but I have have seen kernels that assert that and carry on assuming it to be true
median goblet
I see
desert haven
I mean I wouldnt advise it, but I have have seen kernels that assert that and ca...
but then again its really not that much more work to do it properly and get the address from the msr - and that covers you in case you end up on a system where they've been moved.
median goblet
but then again its really not that much more work to do it properly and get the ...
i was about to cite performance problems but i realise you can cache the msr's value somewhere and use it
surreal path
now is correct
edgy pilot
osdev on speedrun.com when
shut laurel
bash any%
finite summit
for a second I thought he had bash
and I was gonna be like
"how tf does he have bash in < 2 weeks, and I've been at osdev for a year and still haven't ported mlibc once in 5 revisions of the kernel"
thorn bramble
i had a vfs and a few syscalls the first night of osdev in zig
safe to say it was shit lol
but was it quick? yeah
flat nymph
porting (m)libc is boring
finite summit
ik, as I have tried to once
flat nymph
(as in I'm rolling my own libc)
edgy pilot
porting (m)libc is boring
Blasphemy!
surreal path
"how tf does he have bash in < 2 weeks, and I've been at osdev for a year and st...
Thanks to sickness not speedrunning this
I’m too sick
😭😭
help me
surreal path
hi i saw u type fadanoid
hii
@kind root does uacpi have rust bindings i kinda need to use it
to get a bunch of tables
i need
finite summit
Thanks to sickness not speedrunning this
Ok take it easy then
Unless you like pain
Then go on
surreal path
trying to get uacpi integeration
using the uacpi-rs bindings by @silk wind
few problems tho
silk wind
What problems do you get?
Also I took over the development after the previous owner stopped maintaining it and I haven't found time to rework it
surreal path
What problems do you get?
Also I took over the development after the previous ow...
error[E0432]: unresolved import `alloc::boxed`
--> src/acpi/mod.rs:3:12
|
3 | use alloc::boxed::Box;
| ^^^^^ could not find `boxed` in `alloc`
the trait asks for boxed
i dont have
box
for some reason alloc doesnt define it
even tho i have a global allocator
that works
finite summit
you could use C :^)
surreal path
and marked it with #[global_allocator]
surreal path
you could use C :^)
lol
finite summit
never too late to rewrite
surreal path
3rd rewrite when
molten grotto
for some reason alloc doesnt define it
the compiler tells you what you need to do
extern crate alloc;
molten grotto
surreal path
now it works
silk wind
The rust analyzer in vsc doesn't give you the same level of help as rustc does
surreal path
i know
surreal path
is the address in rsdp request limine gives virtual or physical
molten grotto
is the address in rsdp request limine gives virtual or physical
virtual
surreal path
and im guessing uacpi expects a physical address so
i gotta unadd the hhdm offset from it
still page faultsss
:c
page faults here not matter what i do
:ccc
even tho i add the hhdm offset to it
edgy pilot
add or subtract
surreal path
add when using the map function
i give uacpi the physical rsdp
i let it call the kernel api map
which adds the hhdm offset to it
and returns it
edgy pilot
What's the page fault error code
surreal path
0
surreal path
should be tho i went over the entire memory map and mapped everything
i dont think its my map function cause its the same implmentation from old nyaux
edgy pilot
bit 0 not being set means page isn't present
surreal path
there is something i dont do i do in old nyaux
for (uint64_t i = 0; i < 0x100000000; i += 0x1000)
{
map((uint64_t*)((uint64_t)kernel_pagemap.pml4 + hhdm_request.response->offset), (hhdm_request.response->offset + i), i, NYA_OS_VMM_PRESENT | NYA_OS_VMM_RW);
hhdm_pages += 0x1000;
}
this line
i dont understand why i do this
in old nyaux
edgy pilot
that seems to map first 4 gb to hhdm
edgy pilot
in your rust rewrite where do you map the memmap entries
surreal path
in your rust rewrite where do you map the memmap entries
to their hhdm virtual addresses
to map the hhdm
edgy pilot
no I mean in which file and line
edgy pilot
try actually mapping in your uacpi map function
edgy pilot
wdm?
edgy pilot
I'm confucius
kind root
page faults here not matter what i do
The fault address here is way below
surreal path
The fault address here is way below
how are u so sure
kind root
Cr2?
surreal path
you don't have a virtual address to map it to?
yea i know
im gonna solve that
cause vmm_region_alloc gives memory
edgy pilot
just get a random one away from the ram and map it to there
surreal path
allocated and mapped contingous memory
edgy pilot
you can get top of the memory from memmap entries
kind root
```C
for (uint64_t i = 0; i < 0x100000000; i += 0x1000)
{
map((uint6...
I mean clearly this is broken
Because memory below 1mb is not mapped
edgy pilot
you can get top of the memory from memmap entries
add a few gigs and done
surreal path
I mean clearly this is broken
its finene i mean it worked
kind root
What does Info mem say
kind root
Yeah u don't map any of reserved memory
kind root
What worked
surreal path
nvm
seems like uacpi over here is giving me a status error
out of memory
lovely
so im not gonna map the reversed regions then
cause it causes that
edgy pilot
how much ram are you giving to qemu
surreal path
2g
edgy pilot
you shouldn't run out of memory by mapping reserved regions
surreal path
i know
surreal path
doesnt matter if i do the same thing
while i != 0x100000000
{
q.map(
HDDM_OFFSET.get_response().unwrap().offset() + i,
i,
VMMFlags::KTPRESENT.bits() | VMMFlags::KTWRITEALLOWED.bits()
).unwrap();
i += 0x1000;
hhdm_pages += 0x1000;
}
as in old nyaux
i have no idea whats going on ughh
finite summit
and im guessing uacpi expects a physical address so
wrong
I think
maybe
lemme check
nvm it's phys
surreal path
okay i may take a break for an hour my brain hurts debugging this
maybe ill have an idea on why when i come back from the break
flat nymph
3rd rewrite when <:memedown:1267962938132533248>
Zig
finite summit
No.
flat nymph
But like what else
finite summit
C
flat nymph
Ada?
finite summit
shhhh, you'll wake streaksu
flat nymph
C is obnoxious
flat nymph
A language without templates is painful
finite summit
skill issue ngl
elder shoal
Just macro so hard you reinvent c++
finite summit
I've never had much problems with macros
flat nymph
Also no RAII
finite summit
skill issue
just free your memory
I rather use 22M of memory in my kernel, or leak that much
but who cares, they invented swap for a reason
flat nymph
It's not a skill issue, it's an old language
finite summit
it's a skill issue
flat nymph
My kernel doesn't count how much is used so I'm gonna assume it's very little 
finite summit
flat nymph
Also no RAII
Also no SFINAE
flat nymph
C is obnoxious
Although somehow the majority of my OS is still written in it so...
surreal path
fine ill go back to C and stop rewriting
but im DEF adding rust interop
finite summit
rust is cancer
don't use rust
or associate yourself with it
I will be met with large backlash saying that
which is why I will have this thread muted for a day
surreal path
no im gonna take the existing project
and im gonna have both rust components and some componets written in C
and that decesion you must respect
wine about it like a baba
or actually no
im gonna have it written in rust and im just gonna go fix my allocator bug
okay
thats it
i already made progress
im not stopping
@kind root do you know anything about when uacpi throws a out of memory status?
when you just init it
kind root
<@353593948523986945> do you know anything about when uacpi throws a out of memo...
when u return a null
from anything that allocates
flat nymph
Idk, I'm almost finished with reading the Rust book, and the language just feels like C++ without exceptions and with fancier names for its features
surreal path
when u return a null
O
OHHHH
OHHHHHH
wait
its my slab allocator
so my slab allocator is returning 0?
hmm
kind root
u tell me
surreal path
wtf
edgy pilot
rewrite in c#
surreal path
NO
edgy pilot
yes
surreal path
ill check dealloc then
kind root
check e.g. mutex_alloc
surreal path
okay
it worked!
literarly ONE Number
ONE FUCKING NUMBER
😭
well now that works its time to actually
create the namespace
init level mismatch?
what does that mean
huh???
does anyone know what that means orrr
i have no clue
what kind of error that it
kind root
does anyone know what that means orrr
yes
incorrect order of initialization or api usage
surreal path
okay i used it the right way now
memory error
of course
wait wrong screenshot
this one
and i know vmm_region_alloc panics BEFORE it returns the NONE value so
this is something else
why is the current pagemap none
kind root
btw if u just need the tables u dont have to load the namespace
just uacpi_initialize is enough
surreal path
btw if u just need the tables u dont have to load the namespace
i got so far to this
wrong image sorry
kind root
yeah thats the beginning of namespace_load
kind root
💀
finite summit
obos curse
surreal path
so hard to think
finite summit
so sleep
surreal path
i want to get this out of the way first thooo
finite summit
if (on_sleep_tablet)
Sleep();```writing code while tired has caused me the weirdest bugs
that were because I added schizophrenic code
finite summit
I don't think I've taken much caffiene in my life
surreal path
i never tried a redbull
elder shoal
enlightened: never program before a bottle of beer
surreal path
what does it taste like
kind root
i never tried a redbull
thats why u have bugs 💀
surreal path
thats why u have bugs 💀
maybe i should try one
finite summit
infy you're a fucking genius
kind root
lol
surreal path
PAGE FAULT OF COURSE
page fault non present bit
now lets see the rip
also why tf is the rip at 0
/home/rayan/.cargo/git/checkouts/uacpi-rs-859d27ced179e937/4960769/uacpi-sys/vendor/source/interpreter.c:292 (discriminator 1)
[rayan@linux NyauxKT]$```lovellllly
edgy pilot
nyauxmaster 5.0 name leaked
finite summit
I already knew it
surreal path
on this line
kind root
wdym his name is not nyauxmaster?
surreal path
wdym his name is not nyauxmaster?
no it def it
surreal path
RAHHHHh
kind root
on this line
i had a crash on this line in Astral but that was because his memcpy was bugged
surreal path
i had a crash on this line in Astral but that was because his memcpy was bugged
could it be
that the memcpy this uacpi-rs
surreal path
i had a crash on this line in Astral but that was because his memcpy was bugged
i dont feel like its my allocator
because if it couldnt allocate
i would get unwrap or result errors
kind root
it could be anything else
surreal path
because i do the rust error checking
surreal path
it could be anything else
i have no idea then
i have no idea how im gonna debug this
flat nymph
I start staring at the disassembled code in these situations
kind root
i have no idea how im gonna debug this
expand the macro in that code
see which actual line it dies at
surreal path
expand the macro in that code
it wont let me expand it
only the non impl
CR2=ffff80010008203d
this is what cr2 looks like
does that look valid
finite summit
no
surreal path
what does that look like
finite summit
it's after the hhdm
but not too far off from the hhdm
surreal path
ummmm so why is it wanting that address
surreal path
wha?
finite summit
make sure you don't go over the hhdm limit
when mapping physical pages
or well
putting phys. pages in the hhdm
or getting their address rather
surreal path
make sure you don't go over the hhdm limit
i dont tho
kind root
i doubt its anything super complex
just u doing a rarted thing somewhere
is your calloc ok?
kind root
maybe the rust bindings are fucked somewhere
they're untested for hte most part
and also outdated
surreal path
unsafe fn alloc(&self, layout: core::alloc::Layout) -> *mut u8 {
MemoryManager.alloc(layout)
}
unsafe fn dealloc(&self, ptr: *mut u8, layout: core::alloc::Layout) {
MemoryManager.dealloc(ptr, layout);
}
kind root
inspect them
surreal path
#[no_mangle]
pub(crate) unsafe extern "C" fn uacpi_kernel_calloc(count: usize, size: usize) -> *mut c_void {
get_kernel_api()
.alloc(Layout::from_size_align(count * size, 8).unwrap())
.cast()
}
it does this
for calloc
kind root
no idea if that zeroes anything
kind root
zero it then lol
surreal path
okay lmao ill just zero all my memory then
kind root
tbh
debian linux has that config option on
absolutely any memory allocated anywhere is always zeroed
surreal path
thats actually really cool
oh for the big memory allocations
i forgot to memzero them
but no i do
it faults when it trys to vmm region allocate
nvm
now we are page faults
with an error code of 2
[rayan@linux NyauxKT]$ addr2line -e kernel/kernel ffffffff8004f75d --demangle
/rust/deps/compiler_builtins-0.1.121/src/mem/x86_64.rs:117
[rayan@linux NyauxKT]$
what???
why here
nvm
back to the same issue
@kind root is namespace init needed for getting tables
or no
cause this is clearly some weird issue with some old version of uacpi
im not wanting to debug
probs something that happened before
kind root
cause this is clearly some weird issue with some old version of uacpi
nah its not
its ur skill issue pretty confident 
kind root
<@353593948523986945> is namespace init needed for getting tables
no
only uacpi_initialize
surreal path
its ur skill issue pretty confident <:meme:575445094589661215>
nuh uh
kind root
oh wait
if its using old uacpi
it probably depends on those libc helpers that these bindings ship
those probably suck
as theyre untested
so idk
kind root
just update uacpi lol
surreal path
im not making rust bindings for it
kind root
F
surreal path
i have an OS to work on to be fair
kind root
yeah thats why i mentioned earlier just forget about namespace for now
uacpi_initialize is enough
surreal path
yea
surreal path
back to work
i think i shall use a different acpi aml parser
cause the rust bindings for uacpi are incomplete
silk wind
There is a rust aml parser but idk how feature complete it is
surreal path
There is a rust aml parser but idk how feature complete it is
let me see
ill use it for NOW
kind root
nonono
it doesnt support half the opcodes, its recursive, and other other half of opcodes are implemented incorrectly
u can try but i just wouldnt use any
haughty kite
kind root
lol yes
thats toally easier than unfucking the current rust bindings
maybe there are acpica binding for rust
but probably not
flat nymph
Rewrite in C
finite summit
~~Rewrite in C~~
remove the ~~
kind root
yeah idk man rust for nyaux wasnt a very good choice i think
flat nymph
Or have it as a userapace server (written in C)
kind root
yup
flat nymph
It basically just needs a bit of POSIX and mapping of physical memory
surreal path
yeah idk man rust for nyaux wasnt a very good choice i think
look i dont have good rust bindings yet for uacpi
ill use the shitty rust version
and then switch to the new uacpi rust bindings when thats ready
that sound good?
i need to get a move on with the project
i plan to make everything modular
so
i will be able to rewrite the acpi part of the kernel just fine
kind root
that sound good?
i mean its your project lol
surreal path
yea lol
kind root
im simply warning you that the rust ACPI crate is not very good
surreal path
im simply warning you that the rust ACPI crate is not very good
might just use the lai bindings at this point
yea i saw its shit
kind root
might just use the lai bindings at this point
thats a thing?
i mean sure lol
lai doesnt do table parsing for u tho
kind root
lol
finite summit
fucking hell what do i use
C
surreal path
i cant do this bro notihing works it keeps page faulting
edgy pilot
surreal path
debugging the memory manager
seems like theres an issue
with the slab allocator
:c
let obj_amount = (4096 - size_of::<slab_header>()) / size;
i feel as if this is wrong
for i in 1..obj_amount
{
let mut new = (start as u64 + (i as u64 * size as u64)) as *mut KTNode;
println!("node: {:#?}", new);
(*prev).next = Some(new);
prev = new;
}
(*prev).next = None;
i cannot know whats causing some of the nodes to point to 0x1
i cant figure it out
😭
kind root
wtf is this formatting
finite summit
copy pasting from vscode is funky
surreal path
copy pasting from vscode is funky
true
kind root
At least its not pointing to 0x2
finite summit
At least its not pointing to 0x2
^
surreal path
At least its not pointing to 0x2
sometimes it points to 0x0
@molten grotto i need your help im dying i cannot figure out the bug in my slab allocator code
kind root
Didn't you have one that was relatively good
kind root
Maybe u fucked up because of rust semantics
surreal path
i dont think so i did everything correctly
i read the documentation
write_bytes does count * size of T
maybe theres something i dont know
kind root
I mean I dont see any other possibilities
surreal path
yea
surreal path
where is that code?
nyauxkt/kernel/src/mem/phys/mod.rs
thorn bramble
this code
is giving me
multiple aneurysms
i can feel blood vessels popping in my brain
surreal path
ik but it should work
surreal path
😭
surreal path
lol
surreal path
i know im writing like its C
thorn bramble
jesus man idk how you read this
no
you are writing it
then compiling
compiler tells you something
and you do the complete opposite
(probably)
idk how you actually come up with this code
surreal path
😭
thorn bramble
but comparing booleans to booleans is criminal
idk why people do it in ANY language
like a bool is already a bool
why compare it to another bool
'outer: while h.is_none() == false
like here
surreal path
i know
surreal path
yea ikik
thorn bramble
or even better!!!
while let Some(h) = h {}
rust has structured bindings and pattern matching!!!
you dont need to unwrap everything!!!
you can just do that Lol
but yeah this is hard to read sorry
im not too sure why stuff is borrken
surreal path
i understand
surreal path
im not too sure why stuff is borrken
yea me too
surreal path
im trying too
flat nymph
and you do the complete opposite
That's an average C++ experience
i know im writing like its C
thorn bramble
not really because the c++ compiler does not give you actual tips on how to fix errors
while the rust compiler does :^)
at least for the easy stuff
flat nymph
clang does
thorn bramble
since when lol
flat nymph
I was surprised
thorn bramble
maybe VERY simple stuff like "oh you forgot a deref"
or "there is a similarly named type/function/whatever"
but other than that i just splurges out a million template instantiations and tells u to fuck yourself
flat nymph
"Can't find the variable. Did you mean this: X?"
Or something
It's better than GCC though
thorn bramble
at the cost of compile times, i guess
at least thats my experience with llvm in general
flat nymph
Can you make a binding to proven C version?
surreal path
Can you make a binding to proven C version?
rather not
kill me
wtf
why is it some 0x0
i said
None
(*new).next = None;
its supposed to be none
wtf
@thorn bramble u know anything about this
flat nymph
(*new).next = None;
Isn't something like this UB?
surreal path
Isn't something like this UB?
how
molten grotto
if new doesn't contain an object previously written to there then yes
surreal path
if new doesn't contain an object previously written to there then yes
but i wrote to it
@molten grotto
for i in 1..obj_amount
{
let mut new = (start as u64 + (i as u64 * size as u64)) as *mut KTNode;
new.write_bytes(0, 1);
(*new).next = None;
(*prev).next = Some(new);
println!("info about this new node new: {:?} prev: {:?}", (*new), (*prev));
let mut e = 0;
for c in 0..800000
{
e += 1;
}
prev = new;
}
here
new.write_bytes(0,1);
^^^
molten grotto
yeah that should be fine then assuming that all zeros is a valid representation of KTNode
(which it most likely is)
surreal path
yeah that should be fine then assuming that all zeros is a valid representation ...
do you know?
im really stuck
😭
im banging my head againist ze wall
molten grotto
do you know?
I mean you can try to .write to there instead of write_bytes but I doubt that makes a difference
surreal path
I mean you can try to .write to there instead of write_bytes but I doubt that ma...
didnt solve it
nothings working qwinci 😭
flat nymph
Try a different compiler
surreal path
no
surreal path
tried setting up rust santizer didnt even help
wtf
no errors???
bro
yall know this shit?
@molten grotto do you know about rust santiziers? i did -Z santizer=kcfi
still no errors
molten grotto
that's for control flow integrity anyway, not for ub
surreal path
that's for control flow integrity anyway, not for ub
error: undefined symbol: __asan_register_globals when with -Z santizer=kernel-address
what does that mean
molten grotto
asan needs quite a bit of work to get working
surreal path
asan needs quite a bit of work to get working
how do i get it working
it would be very helpful to have
molten grotto
well you need a shadow space where those functions that it asks for mark specific regions of memory as accessible or poisoned
surreal path
well you need a shadow space where those functions that it asks for mark specifi...
oh
molten grotto
like I said its not the easiest thing ever to implement
surreal path
like I said its not the easiest thing ever to implement
im not doing that then
i guess im stuck for the time being with this issue of Some(0x0)
😃
molten grotto
if you push it I can take a look again
surreal path
okay second
surreal path
if you push it I can take a look again
pushed
molten grotto
the problem is that the size of KTNode is 16 because Option must be able to hold both Some(nullptr) and None as separate values and your slab size is 8
so either make the min slab size 16 or use Option<NonNull<KTNode>> instead of Option<*mut KTNode>
(NonNull is just a wrapper type over a pointer that can't be nullptr so then Option<NonNull<T>> can have the same size as the pointer itself as null can represent None)
surreal path
the problem is that the size of KTNode is 16 because Option must be able to hold...
why is my slab size 8
oh right
im idiot
surreal path
so either make the min slab size 16 or use Option<NonNull<KTNode>> instead of Op...
NonNull?
surreal path
(NonNull is just a wrapper type over a pointer that can't be nullptr so then Opt...
So i dont need an Option?
molten grotto
well it depends on how you want to do it
if you want KTNode to have a size of 8 then you either use the raw pointer without option checking for null instead of None or use Option<NonNull<T>>
surreal path
if you want KTNode to have a size of 8 then you either use the raw pointer witho...
Ill just use nonnull then
nonnull is of size 8 right?
@molten grotto ?
molten grotto
yes it has the same size as the pointer itself as its just a wrapper over a pointer
and Option<NonNull<T>> has the same size as a pointer too
surreal path
right
or you know what
ill just not have sizes of 8
it works now thanks qwinci!
surreal path
nvm @molten grotto having a page fault when using the acpi crate in here
/rustc/9c01301c52df5d2d7b6fe337707a74e011d68d6f/library/core/src/array/equality.rs:150
:c
molten grotto
nvm <@379250891838193667> having a page fault when using the acpi crate in here
...
the problem is that you don't map the whole hhdm, you only map the usable regions there
surreal path
the problem is that you don't map the whole hhdm, you only map the usable region...
so i need to map the reversed regions too?
molten grotto
if you use the hhdm to access them then yes
also do note that they might not be page aligned
surreal path
it works now tho
molten grotto
also do note that they might not be page aligned
so you want to first align down base to page size and then use that (and add base % page size) to the length before you align the length up)
surreal path
so you want to first align down base to page size and then use that (and add bas...
okay done
it works
sadly
the interrupt handler keeps showing rip at 0x0
i cannot figure it out
😔
hmmm
ill just use gdb and see
surreal path
so you want to first align down base to page size and then use that (and add bas...
is there a way to trigger an interrupt with gdb
or run an asm instruction
manually
molten grotto
idk
surreal path
yea somethings corrupting the stack
okay
reading the disassmebly
something funking is happening
ffffffff80004140 <NyauxKT::idt::page_fault>:
push %rax
push %rbx
ffffffff80004142: push %rcx
ffffffff80004143: push %rdx
ffffffff80004144: push %rsi
ffffffff80004145: push %rdi
ffffffff80004146: push %rbp
ffffffff80004147: push %r8
ffffffff80004149: push %r9
ffffffff8000414b: push %r10
ffffffff8000414d: push %r11
ffffffff8000414f: push %r12
ffffffff80004151: push %r13
ffffffff80004153: push %r14
ffffffff80004155: push %r15
ffffffff80004157: push $0xe
ffffffff80004159: mov %rsp,%rdi
ffffffff8000415c: call ffffffff80003c70 <NyauxKT::idt::exception_handler>
ffffffff80004161: add $0x8,%rsp
ffffffff80004165: mov %rdi,%rsp
ffffffff80004168: pop %r15
ffffffff8000416a: pop %r14
ffffffff8000416c: pop %r13
ffffffff8000416e: pop %r12
ffffffff80004170: pop %r11
ffffffff80004172: pop %r10
ffffffff80004174: pop %r9
ffffffff80004176: pop %r8
ffffffff80004178: pop %rbp
ffffffff80004179: pop %rdi
ffffffff8000417a: pop %rsi
ffffffff8000417b: pop %rdx
ffffffff8000417c: pop %rcx
ffffffff8000417d: pop %rbx
ffffffff8000417e: pop %rax
ffffffff8000417f: ud2
ffffffff80004181: int3
ffffffff80004182: int3
ffffffff80004183: int3
ffffffff80004184: int3
ffffffff80004185: int3
ffffffff80004186: int3
ffffffff80004187: int3
ffffffff80004188: int3
ffffffff80004189: int3
ffffffff8000418a: int3
ffffffff8000418b: int3
ffffffff8000418c: int3
ffffffff8000418d: int3
ffffffff8000418e: int3
ffffffff8000418f: int3
wtf
surreal path
```rust
ffffffff80004140 <NyauxKT::idt::page_fault>:
pus...
wheres my iretq
fuck i didnt push the rip
my bad
wait no
the sdm states
that the rip should be on the stack wtf
molten grotto
btw here is a slightly unfucked uacpi-rs https://github.com/Qwinci/uacpi-rs
well I didn't add all the functions, mostly it just has eval and reboot/sleep functions
and documented the kernel api https://qwinci.github.io/uacpi-rs/uacpi/kernel_api/trait.KernelApi.html (though its ofc documented within uacpi itself too)
surreal path
well I didn't add all the functions, mostly it just has eval and reboot/sleep fu...
finally
but still
im trying to figure out my
interrupt handler
why its broken lol
why is rip 0x0
am i doing everything right?
like huh
molten grotto
where does it get to rip 0
surreal path
where does it get to rip 0
when it calls the handler
molten grotto
the page fault handler?
surreal path
the page fault handler?
no the exception handler
i dont understasnd why rip is 0
wtf
running it in gdb it solves the problem
it works now?
wtf
molten grotto
how do you trigger the exception
and what specific interrupt number if with the int instruction
surreal path
and what specific interrupt number if with the int instruction
int 0xe
asm!("int 0xe");
molten grotto
asm!("int 0xe");
then your int stub expects an error code to be on the stack no?
surreal path
then your int stub expects an error code to be on the stack no?
ohhhhhhhhhh
molten grotto
so then because int 0xe doesn't push it (because its not a cpu triggered pagefault) the registers aren't saved in the same places
try triggering a real pagefault
eg. with (8usize as *mut i32).write_volatile(0);
molten grotto
well actually looks like you don't take the error code into account in your registers struct or the idt register pushing code at all, it would be between rax and rip if its there (and if its not then you likely want to push a dumb zero to that place)
surreal path
well actually looks like you don't take the error code into account in your regi...
yea i should
kind root
btw here is a slightly unfucked uacpi-rs https://github.com/Qwinci/uacpi-rs
Damn nice, any way u can pr those into the official ™️ repo
molten grotto
Damn nice, any way u can pr those into the official ™️ repo
yeah ig I can, its missing a lot of the functions rn tho
like the event stuff, namespace node discovery stuff etc
molten grotto
I mean yeah at least it compiles with the latest uacpi and has some basic eval stuff but to use that you'd have to know the path ofc
kind root
Nice
