where
#Nyaux
surreal path
finite summit
you just make them yourself
surreal path
bru
spice yarrow
Making closures in c is weird
finite summit
for (size_t i = 0; i < arr.size; i++)
// do shit
i hate dealing with "strings" in C
spice yarrow
iterators are way nicer than for loops
finite summit
iterators are way nicer than for loops
lies
surreal path
one of my biggest reasons why i struggle
cinder plinth
Making closures in c is weird
you dont do closures in C
spice yarrow
You can... you just gotta pass around both a function pointer and a data pointer
But it is kinda cursed
surreal path
look im just not gonna switch language and keep banging my head againist the wall until everything works
even if it takes years
because it wont matter what language i use
im just a horrible programmer in general
plus i havent used C/C++ in a long time
finite summit
I have
surreal path
look im just not gonna switch language and keep banging my head againist the wal...
maybe itll work if i bang my head hard enough looking at the disassembly
idfk
finite summit
I used c++ for the longest time
cinder plinth
depends where
but would probs be something like
iter::rev() -> Slice::iter() -> main
finite summit
anyway, I use C and I will not fall into these language features
spice yarrow
there's not a lot of ways stuff can crash in an iterator unless you explicitly use unwrap or something
cinder plinth
yeah
they cant rlly fail
the only thing I could think of is with fixed arrays
like
well
smallvec
which is a vector with a fixed limit
finite summit
I remain content with my for loop.
spice yarrow
I mean smallvec ran out of space is simple enough to figure out
Also in rust you don't even need backtraces much, which is why they are disabled by default
cinder plinth
Array<int, 10> array;
array.iter().collect<SmallVec<int, 5>>();
this'll just panic
I could maybe make my Container trait require push_back returns a Result<> but that would break stl containers
kind root
wdym lol
thorn bramble
he said that right as he said he hates rust, so i think he expects a bit more than an operator++ wrapper
edgy pilot
C++ has ranges
thorn bramble
i mean sure, you can probably write some crazy shit with the new views stuff, i haven’t looked into that exactly
or ranges..?
thorn bramble
i winder how the codegen is compared to rust iterators :^)
iirc they are optimized pretty heavily in rust
edgy pilot
i winder how the codegen is compared to rust iterators :^)
godbolt time
thorn bramble
eh i cba writing anything more complex in rust and rewriting it in c++ with apis that i dont know lmao
especially on phone :’)
edgy pilot
write a rust one and I'll provide the C++ version
thorn bramble
actually i could just chatgpt that shit
edgy pilot
thorn bramble
but yeah ill see once i get on my pc lol
spice yarrow
write a rust one and I'll provide the C++ version
I really wanna see if you can do this in c++ https://docs.rs/uom/latest/uom/
cinder plinth
I don't see why not
in fact, it'd probs be easier to do in C++
just a bunch of operator overloads
flat nymph
Why are they stupid?
cinder plinth
they're not but you shouldnt use them most times
edgy pilot
I really wanna see if you can do this in c++ https://docs.rs/uom/latest/uom/
Implement a library like that?
yeah just operator overloads
something like
Velocity operator/(Distance, Time)
Distance operator*(Time, Velocity)
there's already std::chrono for time
velocity and distance can be simple classes
cinder plinth
yea
spice yarrow
It's not that simple, it's generic for every possible si unit
edgy pilot
It's not that simple, it's generic for every possible si unit
you don't need generics or templates
spice yarrow
mul<L1, M1, T1, L2, M2, T2>(x: Unit<L1, M1, T1> y<L2, M2, T2>) = Unit<L1 + L2, M1 + M2, T1 + T
2>
that's overcomplicated imo
you would do the same thing chrono does
just have different constructors for si units in Distance and Velocity classes
and use metres and m/s as internal storage
then do operations on them
and have operator overloads for conversions
ezpz
spice yarrow
Right, but what if I need to use something like m^3 * kg^-1 * s^-2
In my example above, Distance would be an alias for Unit<1, 0, 0> and Velocity would be Unit<1, -1, 0>
spice yarrow
okay apparently it's actually really easy
template <int L1, int M1, int T1>
template <int L2, int M2, int T2>
Unit<L1 + L2, M1 + M2, T1 + T2>
Unit<L1, M1, T1>::operator*(Unit<L2, M2, T2> const &other) {
return Unit<L1 + L2, M1 + M2, T1 + T2>(this->value * other.value);
}
just add operators for integer multiplication and division?
I don't quite understand what you're doing
spice yarrow
I can't store the result of that without either hardcoding a ton of units or creating a generic unit type
edgy pilot
oh yeah then have another class Unit
spice yarrow
Right, then Distance can just be an alias for Unit<1, 0, 0>
edgy pilot
yep
spice yarrow
And you can have a distance_from_feet and other functions
edgy pilot
or just have every operator overload return numerical values
no more need for templates 
spice yarrow
Yeah but then you don't get the automatic dimensional analysis
edgy pilot
simple wrapper over a number that stores that extra information
I need to try to implement it first to know if it will work or not
surreal path
silly fella aren’t u
i am
:3
anyways ill just debug this problem after i uh
finish eating
which is gonna be like
a while lmao
we are outside at kfc!!!
imagine
surreal path
ummm might not be able to get any work done today :c, dad taking away all my devices cause of me skipping school for osdev
😔
edgy pilot
you shouldn't be skipping school for osdev
you shouldn't be skipping school at all, unless your ill
or depressed
surreal path
i got closer, used .mask instead of .mask_with()
seems like a garbage addr
.mask does &
.mask_with allows u to do whatever operation u want
its good we are closer
i still have to figure this out tho
i got my devices back (yippie!)
but my dad is putting a time limit on my laptop for the time being
😔
surreal path
Personally, I wouldn't be able to work on an os in that case
yea but yk
i am determined to make bash run on nyaux one day !!
surreal path
but my dad is putting a time limit on my laptop for the time being
ima keep debugging
edgy pilot
don't skip school
rigid fable
i got closer, used .mask instead of .mask_with()
reading the kernel name in my mind sounded very alike to meaw kitty
tender gorge
skip school -> 24 hr osdev mute
surreal path
reading the kernel name in my mind sounded very alike to meaw kitty
check the very first post
lol
okay seems valid here
found it
okay so this bitmask is just wrong, its not
thorn bramble
whatever .map_addr is
ah smart!
let me take the virtual address of the pte
i don't think that's how this works man...
surreal path
let me take the virtual address of the pte
so the bitmask is wrong
thorn bramble
no?
surreal path
okay so whats the issue???
surreal path
i cast pt
thorn bramble
pt = entry.map_addr(|a| a & ...)
surreal path
into a *mut u8
surreal path
yes
thorn bramble
what that mask is?
surreal path
yes
surreal path
i want to get the next table
elder shoal
what that mask is?
cuz thats what the point of the mask is
thorn bramble
what is it masking here
surreal path
to get the physical address to the next table
elder shoal
:dream:
thorn bramble
READ THE ENTRY
surreal path
I AM READING THE ENTRY
thorn bramble
DONT MASK THE FUCKING PTE POINTER
DO YOU EVEN KNOW WHAT MAP_ADDR DOES?
RTFM
UR CREATING INVALID POINTERS
elder shoal
iretq is going joker mode
surreal path
OKAY THEN WHAT AM I SUPPOSED TO USE
thorn bramble
jkoefaKLPFGMWERsjkpgBRKRsKR
'GKRSRS
adfs
sf
sorry
i told u
already
Read the PTE.
THEN mask.
thorn bramble
.
surreal path
fucking idiot i am
jaunty hingeBOT
elder shoal
.!unmute @thorn bramble
jaunty hingeBOT
thorn bramble
I FORGOT THIS DAMN BOT EXISTED
surreal path
😭
elder shoal
I feel like yorue gonna have a fucking breakdown at this rate 😭
surreal path
okay ill read the docs next time before even using any rust function
not just glancing at the vscode
surreal path
😭 😭
thorn bramble
who would have thought...
surreal path
lets check
if the uacpi gdb bug is still there
@molten grotto ITS STILL FUCKING
DOING THIS
surreal path
I DID
thorn bramble
i dont see a reason why this would happen only under gdb
surreal path
surreal path
i dont see a reason why this would happen only under gdb
qwinci said it has to be my allocator or something
and through gdb i found out it was my vmm
so i fixed that
and its STILL doing this
thorn bramble
does it happen without gdb?
surreal path
no
thorn bramble
lol
rigid dust
wait chat active
surreal path
lit perfectly fine
surreal path
wait chat active
yea and
thorn bramble
objdump -d kernel | grep '800002c0:'
?
u can addr2line that too i guess
and figure out what it is
surreal path
objdump -d kernel | grep '800002c0:'
theres no ub
here
wtf
and addr2lines rip
thorn bramble
like u know that
surreal path
doesnt taking me anywhere
surreal path
lovely
thorn bramble
i mean i cant help much so thats all i can say
surreal path
i found where its
doing that
@kind root
do you know whats going on
im reading the object dump and like
idfk know whats going on
ud2 is called a bunch in uacpis own code too
the rip points to nothing
its my PMM
i knew this day would come
edgy pilot
don't ghost ping
surreal path
sorry
SORRY WHAT
it ud2s here???
fun
okay on release
it page faults on vmm_region_alloc
and on
debug
it ud2s
kill me
genuinelly
how does this code cause ```rs
let mut store: Option<&mut VMMRegion> = None;
for (idx, i) in self.head.iter_mut().enumerate() {
if store.is_none() {
store = Some(i);
continue;
}
let temp = store.unwrap();
if i.base.wrapping_sub(temp.base + temp.length)
>= align_up(size as usize, 4096) as usize + 0x1000
{
let new_guy = VMMRegion {
base: temp.base + temp.length,
length: align_up(size, 4096),
flags,
iskernel: false,
};
let amou = align_up(size as usize, 4096) / 4096;
for i in 0..amou {
let data = {
let o = pmm_alloc().unwrap() as *mut u8;
unsafe {
o.add(HHDM.get_response().unwrap().offset() as usize)
.write_bytes(0, 4096);
}
o
};
self.map(
data.addr(),
new_guy.base + (i * 0x1000),
new_guy.flags.bits(),
);
}
let h = core::ptr::with_exposed_provenance_mut::<u8>(new_guy.base);
unsafe { h.write_bytes(0, new_guy.length) };
self.head.push(new_guy);
return Some(h);
} else {
store = Some(i);
continue;
}
}
UB
page fault non present of course
im losing my mind
i wanna die
chat
flat nymph
Probably just stop sprinkling unsafe everywhere
spice yarrow
If you're gonna write unsafe everywhere just write c
flat nymph
You're kinda supposed to abstract/encapsulate it
silk wind
The channel said one ping the inbox says no ping :/
kind root
because the message was deleted probably
surreal path
Avoid unsafe where possible
im trying to
also sorry chat
no work for today
i feel awful
im sick :c
desert haven
agreed, rest up and take care of yourself first 🙂
plush hearth
I FORGOT THIS DAMN BOT EXISTED
This made me notice that you aren't mod anymore
wicked loom
he didn't wanna be
plush hearth
That makes sense
surreal path
still sick
ill be on that nyaux grind when we no longer sick
also theres like a storm going on here
lmao
surreal path
OKAY WE MIGHT BE ABLE TO GET SOME WORK DONE TMRW
IM STARTING TO RECOVER CHAT
IM SORRY ITS BEEN NEARLY A WEEK AT THIS POINT
ACTUALLY ITS BEEN A FEW DAYS BUT WHATEVS
cinder plinth
Nice
surreal path
IM SORRY ITS BEEN NEARLY A WEEK AT THIS POINT
FEELS LIKE A WEEK TO ME 😭
frigid cliff
surreal path
sorry chat been sleeping all day will be working after i take another nap
my sleep schedule is fgucked
😭 🙏
surreal path
IM BACKKK
LES GOOOO
WOKR TIME
WHOOHOOOOHOOH
NYAUX GRIND
!!!!
okay
i have found two bugs
one related to
pmm _init
clearly some UB
and one related to
something else
bru
kind root
Least buggy nyaux snippet
surreal path
Least buggy nyaux snippet
im fixing it
like wait
WHY IS THE CR2
FUCKING
THIS
WTF
kind root
Lmao
surreal path
oen sec im gonna print the hhdm
OKAY THATS SO WEIRd
PRINTING THE HDDM MAKES IT FINE
BUT WHEN NOT PRINTING
IT JUST
CRIES?
AND DOES SOME GOOFY AH SHIT
WTF
?????
edgy pilot
why not use a simple bitmap
surreal path
@kind root all the bugs im going through is most likely coming from the PMM itself
i need some kinda of equivent to std::launder
kind root
Its called C
surreal path
💀
surreal path
okay even after
that
still getting the HDDM causes a opcode exception
fucking hell
ffffffff800045d0 <NyauxKT::mem::pmm::pmm_init::{{closure}}>:
ffffffff800045d0: 55 push %rbp
ffffffff800045d1: 48 89 e5 mov %rsp,%rbp
ffffffff800045d4: 48 83 ec 10 sub $0x10,%rsp
ffffffff800045d8: 48 89 7d f0 mov %rdi,-0x10(%rbp)
ffffffff800045dc: 48 89 75 f8 mov %rsi,-0x8(%rbp)
ffffffff800045e0: 48 8b 06 mov (%rsi),%rax
ffffffff800045e3: 48 8b 38 mov (%rax),%rdi
ffffffff800045e6: 48 83 c7 10 add $0x10,%rdi
ffffffff800045ea: 48 c7 c6 00 18 06 80 mov $0xffffffff80061800,%rsi
ffffffff800045f1: e8 7a 00 00 00 call ffffffff80004670 <<limine::memory_map::EntryType as core::cmp::PartialEq>::eq>
ffffffff800045f6: 24 01 and $0x1,%al
ffffffff800045f8: 48 83 c4 10 add $0x10,%rsp
ffffffff800045fc: 5d pop %rbp
ffffffff800045fd: c3 ret
ffffffff800045fe: cc int3
disassembly looks like tis
this
surreal path
okay im using a debugger for this
rust people say the stack might be getting fucked or smthin
OKAY WTF
IT WORKS FINE ON GDB
yea okay
we are switching to C
im done
@finite summit u happy!!!!
nyaux rewrite #13
im not joking
rust aint for osdev
skill issue or not
idfk im switching to C, rust is fine as a userspace language but i love C for kernel development
😃
finite summit
Yes
finite summit
idfk im switching to C, rust is fine as a userspace language but i love C for ke...
Based opinion
surreal path
now im gonna be straight honest its been a while for me with C but
im sure this wont be too big of an issue since i kept writing rust C like
💀
btw what is a cc runtime
????
surreal path
and libgcc provides what?
finite summit
and libgcc provides what?
Some compiler stuff
surreal path
what happened to it
edgy pilot
LMAO
surreal path
BRO
haughty kite
exception handling???
for C++ lol
surreal path
okay we are so back
im gonna experiment with va lists rqrq in userspace C
brb
:)
spice yarrow
idfk im switching to C, rust is fine as a userspace language but i love C for ke...
Probably a good idea, especially for the way you code. I'm planning on sticking with rust, but I'm gonna be very careful with unsafe.
surreal path
Probably a good idea, especially for the way you code. I'm planning on sticking ...
fair
surreal path
okay ive been experimenting with userspace C and ive got to say
UNIONS ARE SO FUCKING COOL
OMGOMGOGMOMG
I LOVE UNIONS
@finite summit do u love unions
i love unions
finite summit
useless
surreal path
ill give u one use for them
typedef enum {
FAIL,
STRING,
PTR
CHAR
FLOAT,
} Type;
typedef struct {
Type type;
union {
char *str;
void *ptr;
char cha;
float floa;
}
} result;
result higuys() {
result hi = {.type = CHAR, .cha = 'H'}
return hi;
}
DONT TELL ME THATS NOT OCOL OBERROW
and the size would be like
super small cause as u know
unions only have one piece of memory and whatever variable u modify in a union changes that piece of memory
@finite summit if u dont write code u like ur missing out
rigid fable
im gonna experiment with va lists rqrq in userspace C
what exactly about va lists?
surreal path
what exactly about va lists?
idk just mess around with them lol
they are funny funny hehehahah
rigid fable
they are funny funny hehehahah
I abandoned the idea of creating a kernel in Pascal because it did not have va args
surreal path
@kind root are u happy about this change
surreal path
I abandoned the idea of creating a kernel in Pascal because it did not have va a...
😭
rigid fable
(and base 1 array indexing)
rigid fable
imagine doing vga_buffer[0] := 65; (or whatever) and it does not work
only to figure out it does
mov bx, 0xb7fff
mov al, 0x41
mov [bx], al
something like this
then I realized it was base 1 indexing
surreal path
only to figure out it does
mov bx, 0xb7fff
mov al, 0x41
mov [bx], al
isnt that just making bx a ptr to 0xb7fff and putting the value of whatever is in al into bx derefed?
rigid fable
isnt that just making bx a ptr to 0xb7fff and putting the value of whatever is i...
yes
that's what that assembly does
but the thing is, the vga_buffer variable was 0xb8000, for the text mode video memory
so index 1 is 0 so 0xb8000 and index 0 is -1 so 0xb7fff
surreal path
yo how do i make clangd stop with this goofy ah errors
wheres my lsp completeion
rahhh
finite summit
yo how do i make clangd stop with this goofy ah errors
you have configured it right, right?
and given it a compile_commands.json
surreal path
you have configured it right, right?
no
finite summit
and given it a `compile_commands.json`
.
finite summit
there is some tool
called bear
idk how good it is
but it can make one for you
or if you're using cmake
surreal path
oh i see
finite summit
I think that clangd assumes it is in $(project_root)/build/
finite summit
in any case
you can tell it what directory it is in
and do not put this on the git repo btw
--compile-commands-dir=
surreal path
`--compile-commands-dir=`
where do i put this option
finite summit
to clangd
wicked loom
bear is honestly one of the best tools
surreal path
ia gree
i am green
bro
oh right its not in src
cool got flanterm and nanoprintf working
welp gdt time ez fr
elder shoal
nyaux c rewrite?
surreal path
yep
elder shoal
damn
surreal path
got tired of rust
the fucking UB shit
too much
we have been stuck for months with this hsit
bug after bug
its annoying
so like
time to actually get a real working operating system
wicked loom
based
surreal path
exactly
ok gdt done
idt next lol
now on github
idt time ig
lol
wait no actually the first thing im gonna do is extract my symbols for stack traces
just gonna translate the code from rust to C
should be easy considering i write rust C style
wait nvm
i need memory
so uhh idt first
then pmm -> page tables and vmm -> yk the drill
kind root
Lol
silver yarrow
literally no use for them
thats a wild take
edgy pilot
I have only used them in two cases:
classes where you can have either stack or dynamic buffer
structs where you want to access either bitfields or entire value
kind root
Oberrow L take
flat nymph
Why not C++? (instead of C)
surreal path
ok chat im gonna sleep some more
then we go on nyaux grind
chat i may have an addiction to sleeping
also my gf is staying over wit me rn and she doesnt want me out of bed rn
kind root
also my gf is staying over wit me rn and she doesnt want me out of bed rn
Didnt she dump u
surreal path
Didnt she dump u
she did but we got back together
kind root
Also ur like 13
kind root
surreal path
very epic truly
surreal path
Also ur like 13
infy after forgetting my age for the 2nd time
😡
arent u like 20 somethin
idfk
elder shoal
Didnt she dump u
Never has been real, low level devs dont get bitches
kind root
infy after forgetting my age for the 2nd time
No I just saw ur pic
green elbow
damn F's in the chat for sanana
kind root
i miss her
you are better than this
surreal path
does anyone have an assert macro i can use in C
like how do i get asserts going
in C
kind root
#define assert(x) do { if (!(x)) die(#x); } while(0)
rigid fable
it's literally the Limine template 💀
and why is the gdt ptr struct not in the .h file 
surreal path
and why is the gdt ptr struct not in the .h file <:thonk:575445129863888896>
because its not gonna be used by anyone
only gdt.c
tawdry mirage
why are you rewriting it again
kind root
why are you rewriting it again
Rust ub
surreal path
why are you rewriting it again
drives me insane man
tawdry mirage
now it's gonna be c ub instead
surreal path
now it's gonna be c ub instead
no
kind root
now it's gonna be c ub instead
Harder to invoke but probably
surreal path
Harder to invoke but probably
i swear to god
if yall say im gonna make UB
thats not even fucking funny
kind root
Every c kernel has some sort of ub
surreal path
yea but im not gonna make UB thats gonna make it not work
finite summit
use ubsan
I copy pasted some random kernels implementation
(I should probably add in the license now that I think of it)
((it was mit iirc))
surreal path
Every c kernel has some sort of ub
including linux?
or astral
or mangarm
wait managarm c++
elder shoal
managarm asserts the ub away 
surreal path
managarm asserts the ub away <:halfmemeleft:715831625233203250>
i wont have UB
elder shoal
assert(ub == no)
surreal path
i will assert the ub way too
nyaux == managarm 2
@tawdry mirage what is ur opinion on this information
surreal path
anyone know of a spinlock library or somthin
i know barely anything about atomic shit and like
lol
cinder plinth
spinlocks are trivial
you basically just pause the cpu and you check for an atomic bool
spice yarrow
just lock.compare_exchange(FREE, HELD) in a loop until it is successful
There's more stuff that you can do but that's all that's needed
finite summit
use atomic flag
spice yarrow
Isn't that a c++ thing?
finite summit
note that you should have preemption disabled with syscalls held
elder shoal
note that you should have preemption disabled with syscalls held
syscalls
finite summit
*spinlocks
lol
idk why but when I use lock cmpxchg my syscalls go to shit
*spinlocks
goddamnit
surreal path
just `lock.compare_exchange(FREE, HELD)` in a loop until it is successful
what
finite summit
since I'm cool and use irqls I can raise the irql to IRQL_DISPATCH to stop scheduler preemption
but keep other irqs
😎
spice yarrow
what
Here's what spin does https://docs.rs/spin/latest/src/spin/mutex/spin.rs.html#177-195
spice yarrow
im not using rust anymore tho
I know, but that doesn't mean you can't read rust code
finite summit
Here's what spin does https://docs.rs/spin/latest/src/spin/mutex/spin.rs.html#17...
Be banished, rustian
surreal path
Be banished, rustian
nah hes chill
surreal path
I know, but that doesn't mean you can't read rust code
fair
cinder plinth
https://github.com/OBOS-dev/obos/blob/master/src/oboskrnl/locks/spinlock.c#L30
you managed to make spinlock code unreadable
congrats
finite summit
meh
there is some ifdef
irql newIrql = minIrql == IRQL_INVALID ? IRQL_INVALID : Core_GetIrql() < minIrql ? irqlNthrVariant ? Core_RaiseIrqlNoThread(minIrql) : Core_RaiseIrql(minIrql) : IRQL_INVALID;```uh oh
surreal path
wtf is this
finite summit
dw about it lmao
surreal path
this shit makesm e wanna commit rewrite nyaux in assembly
cinder plinth
#define SPINLOCK_INITIALIZER 0
typedef int spinlock_t;
static inline void spinlock_lock(spinlock_t *lock)
{
while (!__sync_bool_compare_and_swap(lock, 0, 1)) {
#if defined(__x86_64__)
__asm__ volatile("pause");
#endif
}
}
static inline void spinlock_unlock(spinlock_t *lock)
{
__sync_bool_compare_and_swap(lock, 1, 0);
}
@surreal path
for free
spice yarrow
this shit makesm e wanna commit rewrite nyaux in assembly
Go for it, become truly based
surreal path
```c
#define SPINLOCK_INITIALIZER 0
typedef int spinlock_t;
static inline void...
NAH DONT BE WRITING CODE FOR ME 😭
cinder plinth
i didnt write it for you
surreal path
ill look up the functions being called here tho
cinder plinth
__sync_bool_compare_and_swap(lock, 0, 1) means if spinlock == 0, then spinlock = 1 and return 1
spice yarrow
```c
#define SPINLOCK_INITIALIZER 0
typedef int spinlock_t;
static inline void...
I'd assert the unlock swap, cause if that fails something has gone very wrong
cinder plinth
how can it fail
cinder plinth
its literally setting a bool
spice yarrow
If you try to unlock a lock that's isn't held
spice yarrow
Yeah but if you ever try to do that something has gone very wrong
Maybe it's just me but I like to assert anything that's easy to check and shouldn't happen
spice yarrow
Every c kernel has some sort of ub
is that a challenge?
spice yarrow
Also, I feel like there should be kernels that run with both ubsan and kasan set to fatal
molten grotto
Also, I feel like there should be kernels that run with both ubsan and kasan set...
tbh ubsan doesn't really catch all kinds of ub either so its not that hard to pass it
it does catch some things like if your stack is not properly aligned
spice yarrow
what stuff doesn't it catch?
kind root
Aka most scary ub
molten grotto
same even tho its unlikely to cause any issues in practice I definitely do stuff that violates it and it wouldn't be fun to debug some random issues caused by the compiler taking advantage of it
surreal path
OMG what happened?
suicide attempt
kind root
Wtf why
kind root
Long story
We have the time
tender gorge
suicide attempt
🫂
Hopefully you get the help you need, we're here for you if you want to talk
surreal path
look its fine, this isnt the place to talk about such stuff
lets just
get back on nyaux grind
should the pmm have a spinlock
its a basic freelist pmm as usual
surreal path
okay
cinder plinth
rule of thumb is if it's going to be accessed by someone else then you lock it
surreal path
rule of thumb is if it's going to be accessed by someone else then you lock it
even when allocating and deallocing
use the lock
okay
surreal path
okay
finite summit
int 3```
upbeat turret
as simply running asm("int $0":::); triple faults
finite summit
as simply running `asm("int $0":::);` triple faults
well maybe
just maybe
your IDT is broken
upbeat turret
it's the handler
finite summit
so fix the handler
finite summit
and you know that because you used a debugger to pinpoint that
upbeat turret
when i use lidt
finite summit
that means nothing
the cpu only looks at an idt descriptor when its specific ISR should be called
upbeat turret
ik
anyway can't fix it rn
focused on paging
then i'll re-do interrupts
oh shit
this is no where near related to nyaux
spice yarrow
as simply running `asm("int $0":::);` triple faults
Why does it triple fault? You can use -d int to log every interrupt
surreal path
as simply running `asm("int $0":::);` triple faults
ur interrupt handler is bad
do -d int
when running qemu
upbeat turret
ur interrupt handler is bad
ik
surreal path
lost it :D
what
upbeat turret
registers rax-r15
surreal path
spice yarrow
are you not looking at the code right now?
upbeat turret
are you not looking at the code right now?
nah
finite summit
miracle debugging
upbeat turret
i lost it
finite summit
waiting until the code fixes itself
upbeat turret
stupid linux mint didnt save
spice yarrow
wdym it didn't save?
upbeat turret
i can explain
spice yarrow
if you executed it then it had to have been saved
upbeat turret
sure ig
finite summit
linux mint more like linux sour
spice yarrow
I mean the first one was broken, so maybe rewriting will fix it
upbeat turret
anyway
finite summit
wait why is my uptime 11 days lol
upbeat turret
off to rewriting that idt from memory
surreal path
stuck to window cause acer gaming laptop support sucks
upbeat turret
- help from the intel sdm and osdev wiki
surreal path
surreal path
+ help from the intel sdm and osdev wiki
these 2 only
finite summit
yeah my friend has a gaming laptop
surreal path
save ur registers
spice yarrow
wait wtf?
upbeat turret
his laptop starting smoking when he ran linux
huh?!?!
upbeat turret
my pc is beefy
finite summit
after spitting some acpi errors
upbeat turret
had no problems
finite summit
and bluetooth errors
then starting smoking (according to him)
and takes minutes to boot because of the bluetooth errors
after a timeout it boots
spice yarrow
I was about to say maybe we could help you with it, but it seems like you've been cursed by the hardware gods
finite summit
surreal path
these 2 only
struct regs {
uint64_t rax, r10, r9, ...
} __attribute__((packed))
void handler(struct regs *frame) {
kprintf("weee");
}
asm stub handler:
push r9
push r10
push rax
call handler
pop rax
pop r10
pop r9
iretq
its lit this easy
also
there
perfect
if u fail this up ur just like me
nyaux try not to fail basic ah task
frfr
spice yarrow
```C
struct regs {
uint64_t rax, r10, r9, ...
} __attribute__((packed))
void h...
Not quite, you gotta handle the fact that some interrupts push error codes
surreal path
Not quite, you gotta handle the fact that some interrupts push error codes
yea true
upbeat turret
```C
struct regs {
uint64_t rax, r10, r9, ...
} __attribute__((packed))
void h...
doesnt the error code usually get pushed
upbeat turret
i feel like that's something to capture
surreal path
what i do is push a dummy error code
upbeat turret
what i do is push a dummy error code
i do the same
(only with isrs that don't push error codes)
surreal path
```C
struct regs {
uint64_t rax, r10, r9, ...
} __attribute__((packed))
void h...
struct regs {
uint64_t error_code, rax, r10, r9, ...
} __attribute__((packed))
asm exception stub handler
push r9
push r10
push rax
call handler
add rsp, 8
pop rax
pop r10
pop r9
iretq
asm non exception stub handler
push 0
push r9
push r10
push rax
call handler
add rsp, 8
pop rax
pop r10
pop r9
iretq
something like this ^
if u use nasm
use some nasm macro shanngains
surreal path
?
upbeat turret
first i save it to rbx
upbeat turret
or do i not need to align
finite summit
some abis require aligning the stack
surreal path
some abis require aligning the stack
not for my abi
finite summit
not for my abi
nope
surreal path
u dont need to
upbeat turret
i'll try something out
finite summit
x86_64 systemv abi
surreal path
I OWN SYSV ABI
finite summit
requires stack to be aligned to 16-bytes at call
upbeat turret
requires stack to be aligned to 16-bytes at call
yeah
surreal path
requires stack to be aligned to 16-bytes at call
usually is
upbeat turret
i aligned the stack before calling the int handler
surreal path
so theres like no need to check
finite summit
but if you do it it's no harm ig
finite summit
as long as you don't round it down
surreal path
upbeat turret
<@919283329587626045>
got it
finite summit
and rsp, ~16```lmao
finite summit
I love xor
surreal path
me too
finite summit
surreal path
0 -> 0
1 -> 0
0 -> 1
0 -> 1
1 bit stack
get boom'd
😱
@spice yarrow
unwrap in C
crazy
ik
i also have asserts :)
finite summit
le very cool
surreal path
yes
finite summit
rust=🤮
finite summit
(before I sleep, I want to say: don't attempt suicide again though, we'd genuinely be sad if you died)
surreal path
(before I sleep, I want to say: don't attempt suicide again though, we'd genuine...
okay
surreal path
gn
upbeat turret
so far lgtm
though in this context, a null IDT would be loaded, the gates need to be set
upbeat turret
quick question
can IRQs be just an ISR that pushes a dummy error code and calls some other handler other than the isr default
@surreal path (please beat me up if you don't want to be pinged)
hollow herald
```x86asm
and rsp, ~16```
15
upbeat turret
show code
i am da triple fault master
it happened
%macro pushr 0
push rax
push rcx
push rdx
push rbx
push rsi
push rdi
push rsp
push r8
push r9
push r10
push r11
push r12
push r13
push r14
push r15
%endmacro
%macro popr 0
pop r15
pop r14
pop r13
pop r12
pop r11
pop r10
pop r9
pop r8
pop rsp
pop rdi
pop rsi
pop rbx
pop rdx
pop rcx
pop rax
%endmacro
extern handleInterrupt
extern handleIrq
%macro isr 1
global int%1_stub
int%1_stub:
push 0
push %1
pushr
mov rdi, rsp
call handleInterrupt
popr
add rsp, 16
iretq
%endmacro
%macro isr_err 1
global int%1_stub
int%1_stub:
push %1
pushr
mov rdi, rsp
call handleInterrupt
popr
add rsp, 16
iretq
%endmacro
%macro irq 1
global int%1_stub
int%1_stub:
push 0
push %1
pushr
mov rdi, rsp
call handleIrq
popr
add rsp, 16
iretq
%endmacro
isr 0
i've only defined isr 0 for testing purposes
i'll define the rest later
spice yarrow
i am da triple fault master
what's the output when you use -d int (post it on pastebin)
tender gorge
(just some examples if you don't believe the last part)
You forgot the part where I said I was gonna cry myself to sleep because I was imagining my friends dying
finite summit
Oh yeah
upbeat turret
what's the output when you use `-d int` (post it on pastebin)
this was the last log before it triple faulted
hollow goblet
anyone know of a spinlock library or somthin
just use spin or some other crate until you know exactly what you need for your locking primitives
and for some stuff use Atomic* types from core
like i store my HHDM in a AtomicUsize
which technically is also UB because that doesn't keep the provenance of the HHDM pointer but it's really fine
when you're writing a kernel in rust, you are really fine doing stuff that would otherwise be not fine
just you keep overdoing it, then expect nothing to do what you want it to do (e.g. memory corruptions
hollow goblet
just use `spin` or some other crate until you know exactly what you need for you...
this really applies to everything that you can take from std or crates
if you can take something from an external dependency, do so and replace it when it isn't exactly what you want