OBOS (not vibecoded) | Operating System Development | Page 25

white mulch Nov 3, 2024, 11:50 AM

#

why wouldn't it speed it up lol

flint idol Nov 3, 2024, 11:51 AM

#

although it speeds up by so much

white mulch Nov 3, 2024, 11:51 AM

#

without optimizations there are a lot of redundant memory accesses for an example

flint idol Nov 3, 2024, 11:51 AM

#

unlike any other time I've used optimizations

weak kestrel Nov 3, 2024, 11:54 AM

#

flint idol unlike any other time I've used optimizations

larger the code base, more optimisations compiler can do

#

and performance difference becomes more apparent

flint idol Nov 3, 2024, 12:10 PM

#

well LTO is being a pain\

#

it's removing my ubsan symbols

#

I'll move ubsan/kasan/stack checker to a separate lib

weak kestrel Nov 3, 2024, 12:10 PM

#

flint idol it's removing my ubsan symbols

attribute used?

flint idol Nov 3, 2024, 12:11 PM

#

undefined reference

flint idol Nov 3, 2024, 12:11 PM

#

weak kestrel attribute used?

no, like

weak kestrel Nov 3, 2024, 12:11 PM

#

weird

#

works fine for me

#

ubsan is for debug mode anyway
and you don't use lto with debug mode cool_meme

flint idol Nov 3, 2024, 12:13 PM

#

and if there is a kernel module in debug mode

#

being used on a release kernel

#

it doesn't find ubsan

#

so it starts crying

flint idol Nov 3, 2024, 12:58 PM

#

the linker is rearted

#

ok

#

I define the symbol OBOS_InitrdBinary

#

export it

#

reference it in 5 different places

#

WHERE THE FUCK

#

is my symbol

#

looks like it needed to be volatile

#

fuck binutils

#

fuck LTO

#

OBOS_EXPORT const void *hashmap_get(struct hashmap *map, const void *item);```

#

makes the visibility of

#

hashmap_get

#

'default'

#

[ DEBUG ] Could not resolve symbol 'hashmap_get' (symbol is hidden) referenced within a driver.```

white mulch Nov 3, 2024, 1:12 PM

#

what if you add -Wl,--export-dynamic?

flint idol Nov 3, 2024, 1:12 PM

#

that fixes that

#

now I just need to fix the nullptr access in the COM driver

#

bruh

#

inlining is ruining me stack traces

white mulch Nov 3, 2024, 1:25 PM

#

implement dwarf cfi interpreter meme

flint idol Nov 3, 2024, 1:26 PM

#

ok wtf is up with LTO

#

how is it making

#

Symbol->name become nullptr

#

JUST HOW

#

yeah goodbye LTO

flint idol Nov 3, 2024, 4:43 PM

#

time to stop rotting away behind my phone scrolling

#

and instead

#

rot away behind my computer

#

writing the new PCI interface

#

and actually

#

I will be using early table-access

#

in uacpi

#

don't ask how cpu_local.h somehow ends up including pci.h

#

meme

real pecan Nov 3, 2024, 4:56 PM

#

Wtf

flint idol Nov 3, 2024, 4:56 PM

#

well it includes cpu_local_arch.h

#

which includes arch/x86_64/gdbstub/connection.h

#

which includes driver_interface/header.h

flint idol Nov 3, 2024, 4:56 PM

#

flint idol which includes `arch/x86_64/gdbstub/connection.h`

for like, cpu-local gdb stub stuff

flint idol Nov 3, 2024, 4:57 PM

#

flint idol which includes `driver_interface/header.h`

which then includes driver_interface/pci.h

#

for the struct pci_hid

#

I'm also going to be refactoring the driver interface init

flint idol Nov 3, 2024, 10:37 PM

#

fun fact: the obos master branch still doesn't have dirty/standby lists

#

I need to merge userspace-work soon....

#

while I am on the topic of the vmm

#

I think I want to merge struct working_set_node and struct page

#

someway

#

because there might be a bug related to that

#

(I haven't observed the bug yet, since it has to do with shared pages between two vmm contexts)

#

basically, the amount of working sets a page is in is stored in the working set node

#

but each time a page is put into a working-set

#

a new one is made

#

potentially causing fnuy stuff to happen
such as:

a page being removed from a working-set (therefore being paged out), while it's still in another process' working-set

#

the way to solve that is really simple

#

put a working_set_node in the struct page

#

like a pointer to it

#

I just realized that locking virtual pages really just means "locking it into the working-set"

#

so pseudocode for that would be:

lock(&pages->ctx)
prefault(pages)
foreach page in pages:
  working_set_node n = insert_page_into_workingset(&pages->ctx->ws, page)
  n->locked = true
unlock(&pages->ctx)```

#

and the opposite would be the same, but without the prefaulting, and instead of inserting the page into the working-set, it gets the page's ws entry, then sets n->locked = false

#

I wonder how prefault will work

#

I could just call Mm_HandlePageFault for each page in there

#

Mm_HandlePageFault(ctx, page->virt, page->present ? PF_EC_PRESENT : 0);

flint idol Nov 3, 2024, 11:12 PM

#

in like a week

#

I'm off on vacation

#

so no osdev there

#

I'm sure yall will miss me

#

I'll bring my laptop :^)

real pecan Nov 3, 2024, 11:14 PM

#

u cant osdev from laptop?

flint idol Nov 3, 2024, 11:14 PM

#

huh

#

what prevents me from osdeving on a laptop?

real pecan Nov 3, 2024, 11:14 PM

#

it was a legit question

#

since u said no osdev

flint idol Nov 3, 2024, 11:14 PM

#

well idk if I'll be bringing my laptop

#

if I can

#

(I don't think my family would be fine with me bringing my laptop on a 2-week vacation)

#

((just to do the exact thing I do everyday))

real pecan Nov 3, 2024, 11:15 PM

#

two weeks is enough to get extremely bored at a vacation

flint idol Nov 3, 2024, 11:16 PM

#

last couple days are doing the pilgrimage in makkah, saudi arabia

#

anyway

#

we'll see

flint idol Nov 4, 2024, 10:34 PM

#

can someone pin this?

white mulch Nov 4, 2024, 10:34 PM

#

flint idol Nov 4, 2024, 10:34 PM

#

thanks

thick jolt Nov 4, 2024, 10:34 PM

#

flint idol can someone pin this?

fr

flint idol Nov 5, 2024, 12:38 AM

#

lmao

dire belfry Nov 5, 2024, 2:42 AM

#

♫ 137 little bugs in the code ♫
♫ 137 little bugs ♫
♫ take one down, patch it around ♫
♫ 93,428,471,930,587,234,682,318,095,723,456,932,108,653,294,867,394,583,495,715,902,847,523,905,

rapid pulsar Nov 5, 2024, 2:43 AM

#

♫ 20,730,897,752,653,674,714,720,723,427,459,000,889,249,915,529,600,981,530,433,365,335,188,589,206,775,527,139,229,601,995,115,836,043,535,786,629,580,792,144,773,078,426,776,834,268,298,332,187,236,818,042,380,226,849,443,884,860,369,688,406,406,660,118,018,751,041,554,543,095,806,670,784,324,017,067,690,408,791,078,125,463,951,378,030,499,174,512,211,021,762,053,899,146,321,811,754,326,676,200,848,317,668,615,746,032,791,477,810,189,370,765,796,872,015,454,208,726,535,891,065,392,365,497,780,680,934,294,765,890,614,462,433,230,434,642,019,850,210,194,574,941,837,149,904,506,668,092,623,531,566,731,463,976,957,855,738,022,148,581,771,099,684,223,664,263,637,368,004,862,059,648,627,935,705,330,725,186,314,285,961,032,283,586,277,966,145,966,280,929,918,887,266,909,890,820,024,385,375,444,923,126,652,775,589,102,263,540,978,186,901,034,212,346,494,403,140,899,551,856,996,521,257,241,285,133,877,814,522,994,967,358,822,224,685,926,377,660,467,871,022,924,275,654,082,181,950,373,765,807,211,473,562,917,453,950,351,601 little bugs in the code ♫

flint idol Nov 5, 2024, 2:43 AM

#

they're only little bugs

#

WIP

#

my bf interpreter is cursed

#

dire belfry Nov 5, 2024, 2:58 AM

#

flint idol

��

flint idol Nov 5, 2024, 3:08 AM

#

better

#

although it isn't perfect

#

the interpreter

#

it kinda works

#

and it also can run a bf interpreter written in bf in it

#

lean glen Nov 5, 2024, 3:25 AM

#

Why the fuck are you making a brainfuck interpreter

thick jolt Nov 5, 2024, 3:26 AM

#

lean glen Why the fuck are you making a brainfuck interpreter

because he said so

#

😎

rapid pulsar Nov 5, 2024, 3:57 AM

#

@flint idol Is this the bf-in-bf you're using? https://raw.githubusercontent.com/canoon/bfbf/refs/heads/master/bf.bf

$ time echo '++++++++[>++++[>++>+++>+++>+<<<<-]>+>+>->>+[<]<-]>>.>---.+++++++..+++.>>.<-.<.+++.------.--------.>>+.>++.!' | kuroko bf.krk "$(cat bf.bf)"
Hello World!

real    2m29.956s
user    2m29.879s
sys    0m0.068s

flint idol Nov 5, 2024, 10:51 AM

#

rapid pulsar <@841762704766140486> Is this the bf-in-bf you're using? <https://raw.githubuser...

nah this is https://brainfuck.org/dbfi.b

flint idol Nov 5, 2024, 10:51 AM

#

lean glen Why the fuck are you making a brainfuck interpreter

was bored

#

didn't feel like osdev

rapid pulsar Nov 5, 2024, 10:53 AM

#

oh that's a lot faster than the one I linked

flint idol Nov 5, 2024, 10:54 AM

#

flint idol and it also can run a bf interpreter written in bf in it

that was a lie

#

I ran the wrong command

rapid pulsar Nov 5, 2024, 10:54 AM

#

$ time echo '++++++++[>++++[>++>+++>+++>+<<<<-]>+>+>->>+[<]<-]>>.>---.+++++++..+++.>>.<-.<.+++.------.--------.>>+.>++.!' | ./kuroko bf.krk "$(cat bf.bf)"
Hello World!

real    0m0.362s
user    0m0.354s
sys    0m0.011s

lol

inland radish Nov 5, 2024, 2:33 PM

#

rapid pulsar <@841762704766140486> Is this the bf-in-bf you're using? <https://raw.githubuser...

2 minutes?!

#

wow

thick jolt Nov 5, 2024, 7:15 PM

#

flint idol didn't feel like osdev

dude respond to dms 😭

flint idol Nov 5, 2024, 10:05 PM

#

back to osdev lesgo

torpid wigeon Nov 5, 2024, 10:05 PM

#

lesgo

flint idol Nov 5, 2024, 10:15 PM

#

been a couple months

#

still barely any userspace

vale nymph Nov 5, 2024, 10:16 PM

#

just say soon™️ and then never do it

thick jolt Nov 5, 2024, 10:16 PM

#

im unsure

#

how to impl this

flint idol Nov 5, 2024, 10:17 PM

#

https://github.com/OBOS-dev/obos/blob/4a6e9a9830b312f6a8994a74d2952205eca19926/src/sanitizers/asan.c

GitHub

obos/src/sanitizers/asan.c at 4a6e9a9830b312f6a8994a74d2952205eca19...

I was bored... Contribute to OBOS-dev/obos development by creating an account on GitHub.

thick jolt Nov 5, 2024, 10:18 PM

#

i still dont understand what im looking at

#

like at all

flint idol Nov 5, 2024, 10:18 PM

#

u have functions

#

called __asan_load1

#

to 8

#

but only powers of two

thick jolt Nov 5, 2024, 10:19 PM

#

to 8?

#

what

flint idol Nov 5, 2024, 10:19 PM

#

__asan_load1
then that last number is all the powers of two up to and including 8

thick jolt Nov 5, 2024, 10:19 PM

#

what

flint idol Nov 5, 2024, 10:20 PM

#

bru

white mulch Nov 5, 2024, 10:20 PM

#

__asan_load1, __asan_load2, __asan_load4, __asan_load8

thick jolt Nov 5, 2024, 10:20 PM

#

oh

#

ok

flint idol Nov 5, 2024, 10:20 PM

#

and then the same thing, but with the word store instead of load

thick jolt Nov 5, 2024, 10:20 PM

#

ok

flint idol Nov 5, 2024, 10:20 PM

#

then in that, you call some other function, which I call asan_verify

#

which verifies the memory isn't poisoned

#

and I explain what "poison" is in the messages here

#

and in the case that the memory is poisoned

#

you panic

#

you can poison just about anything

#

freed memory, out-of-bounds memory

#

non zeroed memory

thick jolt Nov 5, 2024, 10:22 PM

#

i think im just gonna take a break from osdev for a week i dont feel motivated anymore. i feel like every piece of code i write is utter shit so i would probs also screw up the kasan impl as well

#

to be honest

flint idol Nov 5, 2024, 10:23 PM

#

ok

#

cya

thick jolt Nov 5, 2024, 10:23 PM

#

bro is saying ok cya

#

look fine ill impl kasan

#

but ur gonna see how shit

#

it will be

#

look

#

ill probs

#

triple fault or some shit

flint idol Nov 5, 2024, 10:24 PM

#

you probably will if you forget to mark the asan helper functions as __attribute__((no_sanitize("address")))

#

you'd stack overflow if you forget to do that

torpid wigeon Nov 5, 2024, 10:25 PM

#

thick jolt i think im just gonna take a break from osdev for a week i dont feel motivated a...

lol

#

rule 1 of osdev

#

don't expect to write linux 2, but don't let your code reflect that

flint idol Nov 5, 2024, 10:26 PM

#

https://osdev.wiki/wiki/Rules_of_OSDev

OSDev.wiki

Rules of OSDev

Rule of osdev #1469: never assume memory is zero'd - moondeck
Rule of osdev #1468: never assume indices are sequential - mniip
Rule of osdev #1470: don't divide when you can shift - glauxosdever
Rule of osdev #fbf89521-21f1-4a52-b5cb-a0bf0f761bb4: probably avoid collisions with globally unique identifiers - sortie
Rule of osdev #1985: Never try ...

#

white mulch Nov 5, 2024, 10:27 PM

#

I tried asan once but idk how do you actually make it work because you have to have initialized enough stuff to make it work in the first place

flint idol Nov 5, 2024, 10:27 PM

#

wut

#

torpid wigeon Nov 5, 2024, 10:28 PM

#

just don't write bad code

thick jolt Nov 5, 2024, 10:28 PM

#

okay i did this now what

#

who is even

white mulch Nov 5, 2024, 10:28 PM

#

flint idol wut

well like you have to have an allocator for an example to be able to use asan so you'd have to mark all functions needed to bring that up with no sanitize address

thick jolt Nov 5, 2024, 10:28 PM

#

calling these functions

#

idfk

flint idol Nov 5, 2024, 10:29 PM

#

the compiler

#

generates them for each memory access

weak kestrel Nov 5, 2024, 10:29 PM

#

don't you need a prekernel

#

with clang at least

flint idol Nov 5, 2024, 10:29 PM

#

huh

thick jolt Nov 5, 2024, 10:29 PM

#

okay and what do i need to do in __asan_load_n?

white mulch Nov 5, 2024, 10:29 PM

#

white mulch well like you have to have an allocator for an example to be able to use asan so...

or ig with out of line instrumentation you could just skip any asan checks when its not initialized

flint idol Nov 5, 2024, 10:29 PM

#

it calls asan_verify

#

or well

#

it calls something

#

that verifies the memory is unpoisoned

thick jolt Nov 5, 2024, 10:30 PM

#

what am i supposed to put posion values in the memory or smthin

#

what

#

im lost now

flint idol Nov 5, 2024, 10:30 PM

#

thick jolt what am i supposed to put posion values in the memory or smthin

yes

thick jolt Nov 5, 2024, 10:30 PM

#

flint idol yes

in load and store?

flint idol Nov 5, 2024, 10:30 PM

#

no

thick jolt Nov 5, 2024, 10:30 PM

#

where

flint idol Nov 5, 2024, 10:30 PM

#

you have a region after your slab

#

or in your slab

#

which you memset to the poison

thick jolt Nov 5, 2024, 10:31 PM

#

this wont verify my vmm region tho

flint idol Nov 5, 2024, 10:31 PM

#

y

thick jolt Nov 5, 2024, 10:31 PM

#

because

weak kestrel Nov 5, 2024, 10:31 PM

#

flint idol huh

how would you poison/unpoison kernel without it

flint idol Nov 5, 2024, 10:31 PM

#

weak kestrel how would you poison/unpoison kernel without it

just don't bother with globals

thick jolt Nov 5, 2024, 10:31 PM

#

thats virtually memory mapped memory, and like

#

idfk

#

im lost

weak kestrel Nov 5, 2024, 10:31 PM

#

flint idol just don't bother with globals

❓

thick jolt Nov 5, 2024, 10:31 PM

#

flint idol which you memset to the poison

oh

flint idol Nov 5, 2024, 10:31 PM

#

if you wanted to poison globals

thick jolt Nov 5, 2024, 10:31 PM

#

but then

#

uacpi will get garbage

#

data

weak kestrel Nov 5, 2024, 10:32 PM

#

flint idol if you wanted to poison globals

global variables?

flint idol Nov 5, 2024, 10:32 PM

#

no it won't?????

thick jolt Nov 5, 2024, 10:32 PM

#

why

white mulch Nov 5, 2024, 10:32 PM

#

weak kestrel with clang at least

you can make clang emit out of line function calls everywhere so then you could probably skip the asan checks before you have initialized everything

flint idol Nov 5, 2024, 10:32 PM

#

thick jolt why

because why would it?

thick jolt Nov 5, 2024, 10:32 PM

#

because uacpi expects

#

zero'd memory

flint idol Nov 5, 2024, 10:32 PM

#

no it doesn't

thick jolt Nov 5, 2024, 10:32 PM

#

in one of its

#

calloc

#

functions

#

and if i posion the memory

flint idol Nov 5, 2024, 10:32 PM

#

IT'S AFTER THE ALLOCATED REGION

#

oops caps lock\

thick jolt Nov 5, 2024, 10:33 PM

#

so like

[page]posion value

#

like

#

after the page

flint idol Nov 5, 2024, 10:33 PM

#

it's like

#

no

#

it's after an allocated piece of memory

#

what I do is in my allocator

#

if kasan is enabled

#

I add like 256 bytes I think

#

to the size

thick jolt Nov 5, 2024, 10:34 PM

#

what

flint idol Nov 5, 2024, 10:34 PM

#

before allocating anything

#

and then after allocation

#

I do memset(addr+size-256, poison, 256);

thick jolt Nov 5, 2024, 10:34 PM

#

what why 256

flint idol Nov 5, 2024, 10:34 PM

#

nice number

thick jolt Nov 5, 2024, 10:34 PM

#

what if that goes over that page

flint idol Nov 5, 2024, 10:34 PM

#

then it corrupts memory over that page

#

:)

thick jolt Nov 5, 2024, 10:34 PM

#

wow

#

so u cant use 256

#

cause like

#

yea

flint idol Nov 5, 2024, 10:35 PM

#

no

thick jolt Nov 5, 2024, 10:35 PM

#

because what if

#

ur on the last allocated region in the slab

#

ur just gonna

#

overwrite another page

#

fun

flint idol Nov 5, 2024, 10:35 PM

#

then just

thick jolt Nov 5, 2024, 10:35 PM

#

memory corruption

flint idol Nov 5, 2024, 10:35 PM

#

what

#

https://github.com/OBOS-dev/obos/blob/4a6e9a9830b312f6a8994a74d2952205eca19926/src/oboskrnl/allocators/basic_allocator.c#L210

GitHub

obos/src/oboskrnl/allocators/basic_allocator.c at 4a6e9a9830b312f6a...

I was bored... Contribute to OBOS-dev/obos development by creating an account on GitHub.

#

this is what I meant

thick jolt Nov 5, 2024, 10:35 PM

#

oh

flint idol Nov 5, 2024, 10:35 PM

#

https://github.com/OBOS-dev/obos/blob/4a6e9a9830b312f6a8994a74d2952205eca19926/src/oboskrnl/allocators/basic_allocator.c#L325-L327

GitHub

obos/src/oboskrnl/allocators/basic_allocator.c at 4a6e9a9830b312f6a...

I was bored... Contribute to OBOS-dev/obos development by creating an account on GitHub.

thick jolt Nov 5, 2024, 10:36 PM

#

so you add 256 bytes to the slab

#

size

flint idol Nov 5, 2024, 10:36 PM

#

no

thick jolt Nov 5, 2024, 10:36 PM

#

oh

flint idol Nov 5, 2024, 10:36 PM

#

that's the malloc function

thick jolt Nov 5, 2024, 10:36 PM

#

what

flint idol Nov 5, 2024, 10:36 PM

#

it adds 256 to the size parameter

thick jolt Nov 5, 2024, 10:36 PM

#

now ur just making me more confused

#

what

#

what size function

flint idol Nov 5, 2024, 10:36 PM

#

use that thing you have in your head

#

wassit called

#

brain

#

and think of what I could've meant

#

when I said "size function"

thick jolt Nov 5, 2024, 10:37 PM

#

size parameter

#

so like

#

in this calculation

#

u remove -256 bytes

#

giving the page 256 more bytes

flint idol Nov 5, 2024, 10:38 PM

#

no I don't

thick jolt Nov 5, 2024, 10:38 PM

#

oh

flint idol Nov 5, 2024, 10:38 PM

#

for one I don't use a slab allocator

thick jolt Nov 5, 2024, 10:38 PM

#

oh but for a slab allocator would u do this

#

yes or no

flint idol Nov 5, 2024, 10:38 PM

#

probs not

thick jolt Nov 5, 2024, 10:38 PM

#

oh

flint idol Nov 5, 2024, 10:38 PM

#

what I mean is

#

here

#

you would add 256 to amount

#

at the beginning of the function

#

(if kasan is enabled)

thick jolt Nov 5, 2024, 10:39 PM

#

oh

#

so if ksan

#

malloc(amount + 256)

#

?

flint idol Nov 5, 2024, 10:39 PM

#

yea

thick jolt Nov 5, 2024, 10:39 PM

#

ok

#

how do i check if ksan is enabled

flint idol Nov 5, 2024, 10:40 PM

#

ifdef __SANITIZE_ADDRESS__

thick jolt Nov 5, 2024, 10:40 PM

#

ok

#

@obe#

#

like this?

flint idol Nov 5, 2024, 10:42 PM

#

yeah

#

but

#

after you allocate the memory

#

memset(him + amount, poison, 256)

thick jolt Nov 5, 2024, 10:42 PM

#

ok can my posion be whatever i want?

flint idol Nov 5, 2024, 10:42 PM

#

sure

#

as long as it's not:

thick jolt Nov 5, 2024, 10:42 PM

#

ok

flint idol Nov 5, 2024, 10:42 PM

#

zero

#

and it must be consistent with what your asan impl. expects

thick jolt Nov 5, 2024, 10:43 PM

#

there

#

whats next

#

and yes i removed the duplicate ifdef

flint idol Nov 5, 2024, 10:43 PM

#

go to asan_verify

weak kestrel Nov 5, 2024, 10:43 PM

#

thick jolt okay i did this now what

couldn't you rename it lol

thick jolt Nov 5, 2024, 10:43 PM

#

im confused looking at this

thick jolt Nov 5, 2024, 10:44 PM

#

weak kestrel couldn't you rename it lol

rename what

weak kestrel Nov 5, 2024, 10:44 PM

#

OBOS_AsanPoisonValues

flint idol Nov 5, 2024, 10:44 PM

#

thick jolt im confused looking at this

it checks the memory against three different poison values

thick jolt Nov 5, 2024, 10:44 PM

#

ok

thick jolt Nov 5, 2024, 10:44 PM

#

flint idol it checks the memory against three different poison values

why does it need size again

flint idol Nov 5, 2024, 10:44 PM

#

size is the size of the access

thick jolt Nov 5, 2024, 10:44 PM

#

and what is this function

flint idol Nov 5, 2024, 10:45 PM

#

it checks if it's accessing a poisoned region

#

and if so panics

thick jolt Nov 5, 2024, 10:45 PM

#

do i have to impl it

flint idol Nov 5, 2024, 10:45 PM

#

yes?

thick jolt Nov 5, 2024, 10:45 PM

#

ok

flint idol Nov 5, 2024, 10:45 PM

#

you can rename it

#

read the comments of my implementation

#

but not the actual code

thick jolt Nov 5, 2024, 10:46 PM

#

yea i dont know what this bitmask is checking for

#

shhifting at by 47

#

and inverting it

flint idol Nov 5, 2024, 10:46 PM

#

checking if the memory is canonical

thick jolt Nov 5, 2024, 10:46 PM

#

and oring it with

#

ok ill just copy that if statement

#

:)

#

what is memcmp_b whats the difference between memcmp and memcmp_b

flint idol Nov 5, 2024, 10:48 PM

#

memcmp_b checks the memory against an integer

#

as opposed to another memory block

thick jolt Nov 5, 2024, 10:49 PM

#

is there a impl of this function i can yoink

flint idol Nov 5, 2024, 10:49 PM

#

make it urself

thick jolt Nov 5, 2024, 10:49 PM

#

is memcmp_b doing#

#

if (address + size) == value:
return 1
else
return 0

#

?

flint idol Nov 5, 2024, 10:50 PM

#

no????

thick jolt Nov 5, 2024, 10:50 PM

#

oh

flint idol Nov 5, 2024, 10:50 PM

#

it returns a bool for one

#

if the memory is equal

#

to val

#

return true

#

else return false

thick jolt Nov 5, 2024, 10:50 PM

#

okay so bool instead of number

#

okay

#

but the if statement is correct yers?

flint idol Nov 5, 2024, 10:51 PM

#

not at all

thick jolt Nov 5, 2024, 10:51 PM

#

oh

#

fun when im always stupid and never understand things

flint idol Nov 5, 2024, 10:52 PM

#

I wonder who's been cloning obos thinkong

thick jolt Nov 5, 2024, 10:52 PM

#

me

#

i cloned it

#

like a bit

#

ago

flint idol Nov 5, 2024, 10:52 PM

#

november 2nd?

thick jolt Nov 5, 2024, 10:52 PM

#

yea

flint idol Nov 5, 2024, 10:52 PM

#

someone cloned it

thick jolt Nov 5, 2024, 10:52 PM

#

flint idol not at all

so what is it then

flint idol Nov 5, 2024, 10:52 PM

#

take memcmp

#

instead of if (blk1[i] == blk2[i]
do if (blk[i] == val)

thick jolt Nov 5, 2024, 10:53 PM

#

and what is i

#

size?

flint idol Nov 5, 2024, 10:53 PM

#

look at memcmp

thick jolt Nov 5, 2024, 10:54 PM

#

yea im looking at memcmp

#

what about it

#

oh

#

so u loop over

#

if (p1[i] == val

#

do whatever

#

okay

#

there

flint idol Nov 5, 2024, 10:55 PM

#

incorrect

thick jolt Nov 5, 2024, 10:55 PM

#

oh

flint idol Nov 5, 2024, 10:55 PM

#

think about that if statement

#

it should be if p1[i] != val

#

return false

thick jolt Nov 5, 2024, 10:56 PM

#

okay

#

solved

#

tf is short circuit#

#

what is this

#

ltt

#

linus tech tips

flint idol Nov 5, 2024, 10:58 PM

#

that's why I said read the comments

#

and not the code for that function

thick jolt Nov 5, 2024, 10:58 PM

#

yea im reading

#

your checking if the ptr

#

16 bytes before is posioned

#

if one of the sides is posioned

#

yk

weak kestrel Nov 5, 2024, 10:59 PM

#

who wants to implement asan in my kernel halfmemeright

thick jolt Nov 5, 2024, 11:01 PM

#

@flint idol why at + size

#

oh wait

#

nvm

#

sorry

flint idol Nov 5, 2024, 11:02 PM

#

it's all good

thick jolt Nov 5, 2024, 11:07 PM

#

is this correct oberrow

#

@flint idol ?

#

actually wait no is this correct

flint idol Nov 5, 2024, 11:14 PM

#

thick jolt actually wait no is this correct

if (isposright || isposleft) panic

thick jolt Nov 5, 2024, 11:14 PM

#

flint idol `if (isposright || isposleft) panic`

okay

flint idol Nov 5, 2024, 11:15 PM

#

btw

#

this is wrong

thick jolt Nov 5, 2024, 11:15 PM

#

oh

#

fun

flint idol Nov 5, 2024, 11:15 PM

#

you index into nyaux_poison

#

when you check if it's poisoned

thick jolt Nov 5, 2024, 11:16 PM

#

mhm

flint idol Nov 5, 2024, 11:16 PM

#

but you don't index into nyaux_poison when you poison the memory

#

you had used some other value

#

which I stated before

thick jolt Nov 5, 2024, 11:16 PM

#

what

#

https://tenor.com/view/squidward-throw-away-brain-spongebob-gif-12540895558420688599

Tenor

#

????

#

i dont understand

#

what does it mean when u say "when checking if its posioned but not when posioned you had used some other value"

#

like

#

i dont understand what you meant

thick jolt Nov 5, 2024, 11:19 PM

#

flint idol but you don't index into nyaux_poison when you poison the memory

oh yeah#

#

but i remembered

#

#

to do this

#

so

#

its right

#

😎

flint idol Nov 5, 2024, 11:20 PM

#

0x1a is for non-zeroed memory

thick jolt Nov 5, 2024, 11:20 PM

#

okay

#

okay i did this

#

now what

flint idol Nov 5, 2024, 11:33 PM

#

thick jolt now what

profit

thick jolt Nov 5, 2024, 11:33 PM

#

but i didnt impl __asan_load_n

#

or whatever

#

what am i supposed to do

flint idol Nov 5, 2024, 11:33 PM

#

white mulch `__asan_load1`, `__asan_load2`, `__asan_load4`, `__asan_load8`

but don't forget to define these

flint idol Nov 5, 2024, 11:33 PM

#

thick jolt what am i supposed to do

call into asan_verify

thick jolt Nov 5, 2024, 11:34 PM

#

flint idol but don't forget to define these

i do __asan_load_n

flint idol Nov 5, 2024, 11:34 PM

#

and?

thick jolt Nov 5, 2024, 11:34 PM

#

why do i need to define __asan_load1

flint idol Nov 5, 2024, 11:34 PM

#

because

thick jolt Nov 5, 2024, 11:34 PM

#

etc

flint idol Nov 5, 2024, 11:34 PM

#

that's what the compiler expects

thick jolt Nov 5, 2024, 11:35 PM

#

okay

#

ill just

#

manually make the functions

#

wait

#

what is OBOS_EXPORT

flint idol Nov 5, 2024, 11:36 PM

#

dw about it

thick jolt Nov 5, 2024, 11:36 PM

#

ok

#

functions done now what

#

am i done

flint idol Nov 5, 2024, 11:39 PM

#

thick jolt am i done

nearly

#

add -fsanitize=kernel-address

#

to your CC_FLAGS

#

and also asan_store doesn't take a size paramter

#

and also, asan_verify needs kasan disabled

thick jolt Nov 5, 2024, 11:42 PM

#

flint idol and also, asan_verify needs kasan disabled

i do that with NYAUX_NO_KSAN

thick jolt Nov 5, 2024, 11:42 PM

#

flint idol add `-fsanitize=kernel-address`

ok

flint idol Nov 5, 2024, 11:42 PM

#

ok

#

maybe to the linker too

thick jolt Nov 5, 2024, 11:43 PM

#

now it wants

flint idol Nov 5, 2024, 11:43 PM

#

idk

thick jolt Nov 5, 2024, 11:43 PM

#

no abourt

#

abort

flint idol Nov 5, 2024, 11:43 PM

#

just add _noabort

#

to the end

#

of all those functions

thick jolt Nov 5, 2024, 11:43 PM

#

okay

flint idol Nov 5, 2024, 11:43 PM

#

__asan_load1 becomes __asan_load1_noabort

#

etc

#

btw, no underscore

#

between the number

#

and load/store

#

so instead of __asan_load_1

#

you do __asan_load1

thick jolt Nov 5, 2024, 11:45 PM

#

now it wants load16????

flint idol Nov 5, 2024, 11:46 PM

#

thick jolt now it wants load16????

so add one for 16

thick jolt Nov 5, 2024, 11:46 PM

#

now its talking about load_n?

flint idol Nov 5, 2024, 11:46 PM

#

instead of __asan_load_n

#

do __asan_loadN_noabort

thick jolt Nov 5, 2024, 11:47 PM

#

nvm i did that

#

compiles

flint idol Nov 5, 2024, 11:48 PM

#

wooo

#

now does it run?

thick jolt Nov 5, 2024, 11:48 PM

#

nope

#

triple fault

#

called it

flint idol Nov 5, 2024, 11:48 PM

#

make sure you

#

disable kasan on all kasan functions

thick jolt Nov 5, 2024, 11:49 PM

#

#

16th line

thick jolt Nov 5, 2024, 11:49 PM

#

flint idol disable kasan on all kasan functions

i did

#

cr2 is

#

ffff800006672ff8

#

then from 0xe

#

it 0xd's

#

then 0x8

#

then 0xd

flint idol Nov 5, 2024, 11:50 PM

#

you know how I check if I cross a page boundary

thick jolt Nov 5, 2024, 11:50 PM

#

yea

flint idol Nov 5, 2024, 11:50 PM

#

do that

#

you're probably accessing far too memory

thick jolt Nov 5, 2024, 11:50 PM

#

how do i check that

flint idol Nov 5, 2024, 11:51 PM

#

math

thick jolt Nov 5, 2024, 11:51 PM

#

#define OBOS_CROSSES_PAGE_BOUNDARY(base, size) (round_down_to_page(base) == round_down_to_page((uintptr_t)(base) + (size)))

#

mine

#

now

#

:)

torpid wigeon Nov 5, 2024, 11:51 PM

#

chad

#

rob him if his code

flint idol Nov 5, 2024, 11:51 PM

#

I use MIT license for a reason

flint idol Nov 5, 2024, 11:52 PM

#

thick jolt #define OBOS_CROSSES_PAGE_BOUNDARY(base, size) (round_down_to_page(base) == roun...

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

thick jolt Nov 5, 2024, 11:52 PM

#

okay and what do i do if i did cross a page boundary

flint idol Nov 5, 2024, 11:52 PM

#

go to the next comparison

thick jolt Nov 5, 2024, 11:53 PM

#

and what if thats false or true

#

what happens

#

?

#

???

rapid pulsar Nov 5, 2024, 11:55 PM

#

?????????

thick jolt Nov 5, 2024, 11:55 PM

#

like

#

if i cross a page boundary and either check passes

#

do i panic saying a page boundary was crossed or what

#

is that what i do?

#

i dont know if that is correct so im asking to verify if thats correct or not

flint idol Nov 5, 2024, 11:56 PM

#

thick jolt do i panic saying a page boundary was crossed or what

no

thick jolt Nov 5, 2024, 11:56 PM

#

what do i do then

flint idol Nov 5, 2024, 11:56 PM

#

go to the next comparison

thick jolt Nov 5, 2024, 11:56 PM

#

i said either

#

so both page boundary checks

flint idol Nov 5, 2024, 11:57 PM

#

return;

thick jolt Nov 5, 2024, 11:57 PM

#

just return?

#

isnt it a bad thing if u cross a page boundary

#

anyways i did the checks

#

still the same triple fault

#

#

still at memcmp_b

#

nooo

torpid wigeon Nov 6, 2024, 12:03 AM

#

who the hell uses 2 space indents unironically

thick jolt Nov 6, 2024, 12:04 AM

#

im having vscode format it

#

not my fault

#

but anyways i dont know whats going on

#

im going sleep

#

tmrw we figure this out

#

changes are on git

torpid wigeon Nov 6, 2024, 12:04 AM

#

get a .clang-format

thick jolt Nov 6, 2024, 12:04 AM

#

no thanks

#

goodnight

flint idol Nov 6, 2024, 12:22 AM

#

static obos_status writeback_bar(pci_device* device, pci_bar* bar)
{
    obos_status status = OBOS_STATUS_SUCCESS;
    switch (bar->type) {
        case PCI_BAR32:
            status = DrvS_WritePCIRegister(device->location, 0x10+bar->idx*4, 0x4, bar->phys & 0xffffffff);
            break;
        case PCI_BAR64:
            status = DrvS_WritePCIRegister(device->location, 0x10+bar->idx*4, 0x4, bar->phys & 0xffffffff);
            status = DrvS_WritePCIRegister(device->location, 0x10+bar->idx*4+4, 0x4, bar->phys >> 32);
            break;
        case PCI_BARIO:
            status = DrvS_WritePCIRegister(device->location, 0x10+bar->idx*4, 4, bar->phys & 0xffffffff);
            break;
        default:
            return OBOS_STATUS_INVALID_ARGUMENT;
    }
    return status;
}```

thick jolt Nov 6, 2024, 12:32 AM

#

niceeeeee

#

oh wait shit

#

i should be asleep

#

😭

#

my dad gonna kill me

flint idol Nov 6, 2024, 12:32 AM

#

go to sleep.

thick jolt Nov 6, 2024, 12:32 AM

#

okiiii

#

goodnight for real this time

flint idol Nov 6, 2024, 12:33 AM

#

gn\

flint idol Nov 6, 2024, 1:18 AM

#

other than the fact that I need to essentially rewrite the pci part of pnp

#

and the ahci driver doesn't compile

#

a large part of the new pci interface is implemented

#

well not large part

#

but a part

#

writing pci resources

#

and capabilities

#

now of course I don't initialize any buses yet

#

so it's useless

#

now my head hurts so I'm going to sleep early

#

but tomorrow I will probably continue on this

flint idol Nov 6, 2024, 11:33 AM

#

when I'm done my new pci interface, the ahci driver will have support for multiple ahci controllers

#

hopefully

#

how would I name the drives though? sd1a

#

for ahci controller 1, drive a

#

sd2a
for ahci controller 2, drive a

short mortar Nov 6, 2024, 12:28 PM

#

flint idol when I'm done my new pci interface, the ahci driver will have support for multip...

Can't you just link the same module in multiple times?

flint idol Nov 6, 2024, 12:29 PM

#

No

white mulch Nov 6, 2024, 9:39 PM

#

btw uHDA now has the locking stuff, fixed path configuration (as the old one was broken on real hw) and some doc on the concepts if you want to try it with your kernel (it would be good to have some other kernels to see if the kernel api is good enough though it may change with the possible unified api and also would be good to see if it works in general on different real hardware cards)

real pecan Nov 6, 2024, 9:42 PM

#

also oberrow has some old hw so would be interesting to test

white mulch Nov 6, 2024, 9:48 PM

#

one thing I am slightly worried about is cards that have broken stream position reporting meme though I don't think its that hard to fix as there is an other mechanism to get it too so you'd just switch to that if the position didn't change at all between irqs

flint idol Nov 8, 2024, 2:42 PM

#

I can't wait to use my new panic screen

#

when I merge userspace-work with master

#

sharp pike Nov 8, 2024, 2:44 PM

#

i5-4590T

flint idol Nov 8, 2024, 2:46 PM

#

sharp pike i5-4590T

Indeed

#

My computer is from 2013

sharp pike Nov 8, 2024, 2:47 PM

#

which one?

torpid wigeon Nov 8, 2024, 2:54 PM

#

flint idol

unpresent sounds weird

flint idol Nov 8, 2024, 2:57 PM

#

sharp pike which one?

a one

flint idol Nov 8, 2024, 2:57 PM

#

torpid wigeon unpresent sounds weird

sounded right to me

#

https://en.wiktionary.org/wiki/unpresent#:~:text=present (adjective).-,Adjective,Not present.

Wiktionary

unpresent

torpid wigeon Nov 8, 2024, 2:57 PM

#

just say not present idk

flint idol Nov 8, 2024, 2:57 PM

#

that wastes two bytes of memory

#

so no

torpid wigeon Nov 8, 2024, 2:58 PM

#

printing out your long git hash wastes memory too

flint idol Nov 8, 2024, 2:58 PM

#

torpid wigeon printing out your long git hash wastes memory too

yeah but it's useful for bug reports

torpid wigeon Nov 8, 2024, 2:58 PM

#

(not mentioning the ascii art)

flint idol Nov 8, 2024, 2:58 PM

#

ascii art just looks good

#

so I obviously need that

vale nymph Nov 8, 2024, 2:59 PM

#

Oopsie! The cpu did a fucky wucky!!!!

flint idol Nov 8, 2024, 2:59 PM

#

vale nymph Oopsie! The cpu did a fucky wucky!!!!

isn't that the astral panic screen

torpid wigeon Nov 8, 2024, 2:59 PM

#

i'm writing a qr code generator for the drm panic handler

flint idol Nov 8, 2024, 3:00 PM

#

(I may or may not have thought about doing that once)

vale nymph Nov 8, 2024, 3:00 PM

#

flint idol isn't that the astral panic screen

Cpu0: oops.

torpid wigeon Nov 8, 2024, 3:00 PM

#

i'll make it like uACPI

#

independent from everything else, just plug and play

flint idol Nov 8, 2024, 3:00 PM

#

vale nymph Cpu0: oops.

what happened to cpu1

#

does it just not crash

vale nymph Nov 8, 2024, 3:01 PM

#

Cpu1 watched cpu0 crash and decided to retire

flint idol Nov 8, 2024, 3:01 PM

#

ah

#

sometimes I'm scared of what auto-merge can do

#

like when git merge decides it thinks it knows how to fix a merge conflict

#

uh oh
why is the current obos branch so much slower than master

#

vale nymph Nov 8, 2024, 3:18 PM

#

Did you forget -O0

flint idol Nov 8, 2024, 3:18 PM

#

they're both on -O2

#

I have a feeling it's my slowfat driver

torpid wigeon Nov 8, 2024, 3:21 PM

#

fat + slow

flint idol Nov 8, 2024, 3:21 PM

#

torpid wigeon fat + slow

the name of the fat driver is a parody of NT's fastfat driver

#

because like

#

my fat driver is slow af

#

I'm just going to blame it on uacpi

#

meme

lean glen Nov 8, 2024, 3:24 PM

#

flint idol

you're printing a bunch more stuff

#

tho

flint idol Nov 8, 2024, 3:25 PM

#

true

#

I did also make the log level higher on userspace-work

#

and the speed seems matched

#

and it uses 9M less memory

flint idol Nov 8, 2024, 5:44 PM

#

since I'm going to the UAE I might be able to test obos on some relatives devices (with consent first, of course)

#

cc @real pecan

#

apparently obos doesn't like CPUs reporting as AuthenticAMD

vale nymph Nov 8, 2024, 5:59 PM

#

Have you never tested with tcg? IIRC tcg returns AuthenticAMD

flint idol Nov 8, 2024, 6:02 PM

#

vale nymph Have you never tested with tcg? IIRC tcg returns AuthenticAMD

I have with -cpu Haswell

#

but never with -cpu qemu64

#

or rather, no -cpu argument

#

it literally just triple faults on initializing the page tables

#

maybe I'm missing some sorta check

#

huh?

#

it triple faults in stb sprintf

#

__builtin_popcount crashes

#

probably due to #UD

#

yup

real pecan Nov 8, 2024, 6:16 PM

#

flint idol since I'm going to the UAE I might be able to test obos on some relatives device...

Nice

flint idol Nov 8, 2024, 6:22 PM

#

flint idol `__builtin_popcount` crashes

I told the compiler to compile for x86_64-v1

#

and it works

#

since __builtin_popcount calls a libgcc function

#

I just added a config option to obos

#

to configure the value passed to march

#

OBOS_X86-64_VERSION

#

v1 is x86-64
v2 is x86-64-v2
v3 is x86_64-v2
v4 is x86_64-v4

torpid wigeon Nov 8, 2024, 6:40 PM

#

is it feasible to have opcode emulation for ud exceptions

#

or is that unusable

flint idol Nov 8, 2024, 6:46 PM

#

sounds delusional

#

~~might as well do it~~

flint idol Nov 8, 2024, 10:07 PM

#

I'm gonna see if obos works on qemu on my phone

#

I got termius installed

#

*termux

#

And will be cloning obos

#

Oh wait

#

I have no obos iso

#

Oh cool termux uses apt

#

Cannot link executable cmake

#

Libcrypto not found

#

#

And apt file

#

Doesn't work

haughty abyss Nov 8, 2024, 10:42 PM

#

loool

inland radish Nov 10, 2024, 2:14 PM

#

vale nymph Have you never tested with tcg? IIRC tcg returns AuthenticAMD

TCG reports GenuineIntel on windows for me, iirc

flint idol Nov 10, 2024, 2:15 PM

#

inland radish TCG reports GenuineIntel on windows for me, iirc

The qemu64 cpu makes it use authenticamd

inland radish Nov 10, 2024, 2:15 PM

#

ah, I must have been testing on qemu-system-i386 only

flint idol Nov 10, 2024, 2:16 PM

#

Ok

thick jolt Nov 12, 2024, 4:54 PM

#

obos is now my os muhahaha

#

ill make a commit today obos will be ported to mlibc! yes muhahahahaha and it will run bash !!!

#

muheheheh

flint idol Nov 12, 2024, 6:53 PM

#

thick jolt ill make a commit today obos will be ported to mlibc! yes muhahahahaha and it wi...

Then obos becomes more cursed nooo

thick jolt Nov 12, 2024, 7:34 PM

#

flint idol Then obos becomes more cursed <:nooo:750466870993223750>

trl

flint idol Nov 13, 2024, 6:00 PM

#

Just got me hands on an esp32

rn_image_picker_lib_temp_81463426-7f70-4994-9505-b453350b0bc4.jpg

#

As soon as I come back and am done writing the new pci interface

#

I will port obos

#

To xtensa

#

Or really after userspace-work is merged

short mortar Nov 13, 2024, 6:21 PM

#

Some ESP32s have RISC-V cores...

torpid wigeon Nov 13, 2024, 6:24 PM

#

like this one :P

flint idol Nov 13, 2024, 6:26 PM

#

short mortar Some ESP32s have RISC-V cores...

Then I'd be porting to that

torpid wigeon Nov 13, 2024, 6:26 PM

#

obos riscv port based

short mortar Nov 13, 2024, 6:26 PM

#

I have some ESP32-C3s (I think) spread around home doing IoT stuff

#

I'm wondering how much can be implemented with their MMUs...

flint idol Nov 13, 2024, 8:32 PM

#

Just realized obos uses waaaay too much memory

#

At least for esp32

#

This model specifically seems to have 4.5M total

#

And obos uses 10M memory

#

*virtual memory

#

And a lot more physical

#

memory

#

Although a couple MB of physical memory is framebuffer

#

And my physical memory count doesn't even include any VMM structures (including PTs)

#

I'll just make it start swapping memory aggressively

thick jolt Nov 13, 2024, 9:18 PM

#

flint idol Just got me hands on an esp32

damn cooollll

real pecan Nov 13, 2024, 10:19 PM

#

flint idol Just got me hands on an esp32

OBOS proprietary edition when

thick jolt Nov 13, 2024, 10:19 PM

#

real pecan OBOS proprietary edition when

nyaux goes closed source and buys out obos

#

😱

flint idol Nov 13, 2024, 11:41 PM

#

real pecan OBOS proprietary edition when

I don't get it

thick jolt Nov 13, 2024, 11:41 PM

#

flint idol I don't get it

im gonna buy out obos

#

and ur gonna

#

nyaobos

#

no

#

nyabos

real pecan Nov 13, 2024, 11:54 PM

#

flint idol I don't get it

Nothing I just hate embedded

flint idol Nov 14, 2024, 12:35 AM

#

I wonder if esp32 has acpi

torpid wigeon Nov 14, 2024, 12:44 AM

#

no

short mortar Nov 14, 2024, 6:53 AM

#

It also kinda doesn't have proper virtual memory

flint idol Nov 14, 2024, 6:57 AM

#

short mortar It also kinda doesn't have proper virtual memory

How come?

short mortar Nov 14, 2024, 6:57 AM

#

It's a microcontroller

#

They don't have niceties of big computers

flint idol Nov 14, 2024, 6:58 AM

#

Yeah but what makes it non proper

short mortar Nov 14, 2024, 7:04 AM

#

I don't remember how it works, but it is very limited

flint idol Nov 14, 2024, 7:05 AM

#

I'll just figure that out on the wag

#

*way

short mortar Nov 14, 2024, 7:07 AM

#

Like you can only have 32 or so memory regions and it's meant for memory protection, not to have virtual memory

#

And only select ESP32s have user mode

flint idol Nov 14, 2024, 7:11 AM

#

Challenge accepted

#

And user mode isn't really needed

short mortar Nov 14, 2024, 7:58 AM

#

Which ESP32 did you get?

flint idol Nov 14, 2024, 7:59 AM

#

short mortar Which ESP32 did you get?

Got it from my uncle, but it said something like ai thinker

#

On the chip

short mortar Nov 14, 2024, 8:03 AM

#

I think it's ESP32S

#

Which has Xtensa cores

flint idol Nov 14, 2024, 8:07 AM

#

short mortar Which ESP32 did you get?

rn_image_picker_lib_temp_a3340b10-7d38-4e67-956d-a20b6400816c.jpg

#

This is it closer up

dire belfry Nov 15, 2024, 7:03 AM

#

rapid pulsar ♫ 20,730,897,752,653,674,714,720,723,427,459,000,889,249,915,529,600,981,530,433...

♫ 20.73 Sesquinquagintaducentillion little bugs in the code ♫

flint idol Nov 15, 2024, 5:47 PM

#

Bruh

short mortar Nov 15, 2024, 8:02 PM

#

Division Error

flint idol Nov 16, 2024, 3:44 PM

#

short mortar Division Error

Page fault

torpid wigeon Nov 16, 2024, 3:45 PM

#

flint idol Page fault

Invalid Opcode

#

(tried to execute sse)

flint idol Nov 16, 2024, 3:46 PM

#

General protection fault (loaded 0x58 to cs when gdtr.limit was 0x38)

#

Because int 0x3 is for SIGTRAP

torpid wigeon Nov 16, 2024, 3:49 PM

#

because it's privileged 🤓

flint idol Nov 16, 2024, 3:49 PM

#

Used in debuggers

#

And other interrupts shouldn't be called manually otherwise it can cause cpu funniness

#

Btw that is only for interrupts through the int instruction

#

For example the page fault exception handler expects an error code

#

If it's not there it probably crashes

#

(It is not pushed when you do the int instruction to invoke int 0x14)

#

Type attribute in the idt entry

#

idtEntry.typeAttributes = DEFAULT_TYPE_ATTRIBUTE | (canUsermodeCall ? TYPE_ATTRIBUTE_USER_MODE : 0);

#

Debug trap

torpid wigeon Nov 16, 2024, 3:58 PM

#

@flint idol ```c
typedef struct ATTR(packed)
{
u16 base_0_15;
u16 selector;
#if CONFIG_bits >= 64
Bits ist:2;
Bits reserved:6;
#else
bits reserved:8;
#endif
u8 type;
u16 base_16_31;
#if CONFIG_bits >= 64
u32 base_32_63;
u32 reserved2;
#endif
} IdtDesc;

which bit would that be here

flint idol Nov 16, 2024, 3:59 PM

#

torpid wigeon <@841762704766140486> ```c typedef struct ATTR(packed) { u16 base_0_15; ...

U see that type field

torpid wigeon Nov 16, 2024, 3:59 PM

#

yea

flint idol Nov 16, 2024, 3:59 PM

#

That's it

torpid wigeon Nov 16, 2024, 3:59 PM

#

which bit

flint idol Nov 16, 2024, 3:59 PM

#

The entire thing

#

Bte

#

*btw

#

Don't use bitfields here

torpid wigeon Nov 16, 2024, 4:00 PM

#

why

flint idol Nov 16, 2024, 4:00 PM

#

Because that's a general rule for structs like th8s

#

*this

#

Compiler can do wacky stuff to it

torpid wigeon Nov 16, 2024, 4:01 PM

#

how

flint idol Nov 16, 2024, 4:01 PM

#

Search bitfield using discord search thing

#

Has been explained too many times

white mulch Nov 16, 2024, 4:02 PM

#

well for idt it doesn't really matter unless you enable something that generates that irq in the middle when you are modifying the entry but in general its a good idea to avoid yeah

flint idol Nov 16, 2024, 4:02 PM

#

torpid wigeon how

#osdev-misc-0 message

torpid wigeon Nov 16, 2024, 4:03 PM

#

lol i might have made a huge meme then :)))

#

flint idol Nov 16, 2024, 4:03 PM

#

torpid wigeon

Oh god

#

Gl

torpid wigeon Nov 16, 2024, 4:04 PM

#

the code is so clean though 😢

flint idol Nov 16, 2024, 4:04 PM

#

Sorry

white mulch Nov 16, 2024, 4:04 PM

#

you can make pretty clean abstraction that's like bitfields but with structs (at least in c++, idk about c but I can't think it would be too bad there either)

flint idol Nov 16, 2024, 4:04 PM

#

white mulch you can make pretty clean abstraction that's like bitfields but with structs (at...

Example?

torpid wigeon Nov 16, 2024, 4:04 PM

#

yea example pls

flint idol Nov 16, 2024, 4:05 PM

#

Bitfields are fine if you use them for internal structs

#

Or smth

#

That aren't from any sorta spec

#

E.g., https://github.com/OBOS-dev/obos/blob/userspace-work/src%2Foboskrnl%2Fmm%2Fpage.h#L14-L24

torpid wigeon Nov 16, 2024, 4:09 PM

#

why isn't that an enum

white mulch Nov 16, 2024, 4:09 PM

#

namespace op_regs {
   constexpr BitRegister<u32> USBCMD {0x0};
}

namespace usbcmd {
    constexpr BitField<u32, bool> RS {0, 1};
    constexpr BitField<u32, bool> HCRST {1, 1};
}

void test() {
    IoSpace space {0xsomeaddressofxhci};

    auto cmd = op_space.load(op_regs::USBCMD);
    cmd &= ~usbcmd::RS;
    op_space.store(op_regs::USBCMD, cmd);
    cmd |= usbcmd::HCRST(true);
    op_space.store(op_regs::USBCMD, cmd);
}
``` this is basically what I do in my kernel and I borrowed the idea from managarm, basically the BitRegister contains the offset of the register within the space and then the space's load method just does a load from that offset returning a BitValue you can get those individual bitfields out of (or set them)

flint idol Nov 16, 2024, 4:09 PM

#

torpid wigeon why isn't that an enum

Used internally

torpid wigeon Nov 16, 2024, 4:10 PM

#

but why not do this

flint idol Nov 16, 2024, 4:10 PM

#

torpid wigeon but why not do this

That's not its purpose

torpid wigeon Nov 16, 2024, 4:10 PM

#

what is it used for

flint idol Nov 16, 2024, 4:10 PM

#

torpid wigeon what is it used for

For page ranges

#

The virtual alloc thing has its own flags

torpid wigeon Nov 16, 2024, 4:11 PM

#

so you would use it like prot.present = true; prot.user = true;

flint idol Nov 16, 2024, 4:11 PM

#

Yes

torpid wigeon Nov 16, 2024, 4:11 PM

#

instead of prot = present | user;

flint idol Nov 16, 2024, 4:11 PM

#

torpid wigeon instead of `prot = present | user;`

Yes

torpid wigeon Nov 16, 2024, 4:12 PM

#

meh, i won't criticize that, imo the latter is cleaner

flint idol Nov 16, 2024, 4:12 PM

#

Btw my vmm is made to be platform independent

torpid wigeon Nov 16, 2024, 4:12 PM

#

so is mine

#

:3

flint idol Nov 16, 2024, 4:12 PM

#

So I don't write anything like it's about to go in a pte

torpid wigeon Nov 16, 2024, 4:12 PM

#

i don't either

flint idol Nov 16, 2024, 4:12 PM

#

Lies

torpid wigeon Nov 16, 2024, 4:13 PM

#

#

see line 305

flint idol Nov 16, 2024, 4:13 PM

#

_x86

#

lvl=4

#

Seems pretty x86 to me

#

Unless vm map is different for each arch

torpid wigeon Nov 16, 2024, 4:14 PM

#

it is

#

that's just the impl

flint idol Nov 16, 2024, 4:14 PM

#

So it's not cross platform?

torpid wigeon Nov 16, 2024, 4:14 PM

#

wdym

flint idol Nov 16, 2024, 4:15 PM

#

If it's different for each arch it's not really cross platform

torpid wigeon Nov 16, 2024, 4:15 PM

#

the header/usage is, the implementation is obviously architecture dependent

flint idol Nov 16, 2024, 4:15 PM

#

Sorry what the hell is acquire_force

torpid wigeon Nov 16, 2024, 4:15 PM

#

how the hell would that work otherwise

flint idol Nov 16, 2024, 4:16 PM

#

torpid wigeon how the hell would that work otherwise

If you really care I can send a link to my code

torpid wigeon Nov 16, 2024, 4:16 PM

#

i do care

flint idol Nov 16, 2024, 4:16 PM

#

torpid wigeon i do care

https://github.com/OBOS-dev/obos/blob/userspace-work/src%2Foboskrnl%2Fmm%2Falloc.c#L103

GitHub

obos/src/oboskrnl/mm/alloc.c at userspace-work · OBOS-dev/obos

I was bored... Contribute to OBOS-dev/obos development by creating an account on GitHub.

torpid wigeon Nov 16, 2024, 4:17 PM

#

flint idol Sorry what the hell is acquire_force

attempts to acquire the spinlock until it succeeds

#

spin_acquire can fail

flint idol Nov 16, 2024, 4:17 PM

#

torpid wigeon attempts to acquire the spinlock until it succeeds

I mean if I were you, I'd make this acquire

real pecan Nov 16, 2024, 4:17 PM

#

torpid wigeon spin_acquire can fail

💀

flint idol Nov 16, 2024, 4:17 PM

#

And the fallible one is try_acquire

torpid wigeon Nov 16, 2024, 4:17 PM

#

ig spin_try_acquire is a better name

#

yea

white mulch Nov 16, 2024, 4:18 PM

#

white mulch ```cpp namespace op_regs { constexpr BitRegister<u32> USBCMD {0x0}; } namesp...

though actually idk doing something like this in c might be painful because of no generics or methods, ig the best you can do is macros to extract/set specific fields which is kinda a c skill issue ngl

flint idol Nov 16, 2024, 4:19 PM

#

flint idol https://github.com/OBOS-dev/obos/blob/userspace-work/src%2Foboskrnl%2Fmm%2Falloc...

Imagine having to repeat this kinda code for each arch

torpid wigeon Nov 16, 2024, 4:19 PM

#

😔

white mulch Nov 16, 2024, 4:19 PM

#

I don't think they are repeating that kind of code for each arch? it was just the code to map one page

#

and that obviously needs to be arch dependent

flint idol Nov 16, 2024, 4:20 PM

#

It only has some helpers for mapping a page

#

And query page info

torpid wigeon Nov 16, 2024, 4:20 PM

#

#

this is my map interface, nothing arch dependent here

flint idol Nov 16, 2024, 4:20 PM

#

torpid wigeon this is my map interface, nothing arch dependent here

I'm talking about the implementation itself

torpid wigeon Nov 16, 2024, 4:21 PM

#

yea but how can that possibly be arch independent

white mulch Nov 16, 2024, 4:21 PM

#

vm_map maps one page tho so the impl has to be arch dependent

#

though the desc is kinda undescriptive lol

flint idol Nov 16, 2024, 4:22 PM

#

torpid wigeon yea but how can that possibly be arch independent

Oh I thought that was like the mmap

torpid wigeon Nov 16, 2024, 4:22 PM

#

white mulch though the desc is kinda undescriptive lol

true

white mulch Nov 16, 2024, 4:22 PM

#

maybe you should rename it to something like vm_map_page

torpid wigeon Nov 16, 2024, 4:22 PM

#

but what else would i be mapping

flint idol Nov 16, 2024, 4:23 PM

#

torpid wigeon but what else would i be mapping

Files

white mulch Nov 16, 2024, 4:23 PM

#

or a bigger region

torpid wigeon Nov 16, 2024, 4:23 PM

#

hm

#

i don't have that at all lol

flint idol Nov 16, 2024, 4:23 PM

#

Skill issue

torpid wigeon Nov 16, 2024, 4:23 PM

#

yep

#

btw if you mean the mmap syscall impl that is actually arch independent

#

https://github.com/menix-os/menix/blob/main/kernel/syscall/memory.c

GitHub

menix/kernel/syscall/memory.c at main · menix-os/menix

A minimal and expandable Unix-like kernel. Contribute to menix-os/menix development by creating an account on GitHub.

#

probably the worst code you'll see today

#

but it does its thing

real pecan Nov 16, 2024, 4:27 PM

#

torpid wigeon https://github.com/menix-os/menix/blob/main/kernel/syscall/memory.c

https://github.com/menix-os/menix/blob/f7fdf5a9d74381e65c923618b7e78c1be7ac7f01/kernel/syscall/memory.c#L73

GitHub

menix/kernel/syscall/memory.c at f7fdf5a9d74381e65c923618b7e78c1be7...

A minimal and expandable Unix-like kernel. Contribute to menix-os/menix development by creating an account on GitHub.

#

doesnt this leak half of these mappings

#

or like depending on where it OOMs

torpid wigeon Nov 16, 2024, 4:29 PM

#

yes it does

#

it also doesn't unmap the addresses

real pecan Nov 16, 2024, 4:30 PM

#

isnt that kinda

#

not good

flint idol Nov 16, 2024, 4:30 PM

#

lol

torpid wigeon Nov 16, 2024, 4:30 PM

#

yes

#

that's what i was talking about in #polls

#

this shit is completely broken because i didn't know wtf i was doing

real pecan Nov 16, 2024, 4:30 PM

#

lol

torpid wigeon Nov 16, 2024, 4:31 PM

#

like

#

check out my genius isatty impl

#

flint idol Nov 16, 2024, 4:33 PM

#

torpid wigeon like

Like

real pecan Nov 16, 2024, 4:33 PM

#

if it works it works

torpid wigeon Nov 16, 2024, 4:33 PM

#

ultrameme

flint idol Nov 16, 2024, 4:34 PM

#

torpid wigeon this shit is completely broken because i didn't know wtf i was doing

Smhsmh no file mapping

vale nymph Nov 16, 2024, 5:22 PM

#

torpid wigeon

pipe()
dup()
broken halfmemeright

torpid wigeon Nov 16, 2024, 5:23 PM

#

who needs pipes

#

shared memory + doorbell meme

real pecan Nov 16, 2024, 5:26 PM

#

vale nymph pipe() dup() broken <:halfmemeright:715831702164996096>

or just anything launched not from a terminal

vale nymph Nov 16, 2024, 5:27 PM

#

real pecan or just anything launched not from a terminal

the solution here is clear: have everything be associated with a pty

#

troll

lean glen Nov 16, 2024, 5:30 PM

#

always return true

torpid wigeon Nov 16, 2024, 5:31 PM

#

#define } __builtin_trap(); }

short mortar Nov 17, 2024, 10:49 AM

#

real pecan 💀

Currently, my pthread_mutex_unlock can fail with ENOMEM ultrameme

flint idol Nov 17, 2024, 11:20 AM

#

short mortar Currently, my `pthread_mutex_unlock` can fail with ENOMEM <:ultrameme:7290268986...

W h a t

#

Microkernel skill issue

short mortar Nov 17, 2024, 11:24 AM

#

I'm thinking of reworking it and moving it to the kernel (instead of relying on IPC)

#

(also because of the deadlock issues related to scheduler priorities)

flint idol Nov 17, 2024, 11:38 AM

#

Microkernel skill issue

shy forge Nov 17, 2024, 12:41 PM

#

Imagine ever reporting no memory

#

Post made by the dynamic paging gang

short mortar Nov 17, 2024, 12:53 PM

#

post made by managram assertos

flint idol Nov 17, 2024, 1:16 PM

#

shy forge Imagine ever reporting no memory

Post made by the vmm with paging

real pecan Nov 17, 2024, 1:19 PM

#

shy forge Post made by the dynamic paging gang

but dynamic paging doesnt imply overcommit

flint idol Nov 17, 2024, 1:20 PM

#

flint idol Post made by the vmm with paging

*swap

#

I have ahci

#

Nvme is planned™️

#

Pci interface rewrite

#

Of course I'm on vacation so I'm not working on it now

#

Lmao

#

I'm gonna support PCIe

#

Currently the pci interface is very shit™️

short mortar Nov 17, 2024, 1:30 PM

#

What's the difference?

#

Besides ECAM

flint idol Nov 17, 2024, 1:31 PM

#

short mortar Besides ECAM

Not much

#

Except MSI is guaranteed to be supported

#

On the software side ofc

short mortar Nov 17, 2024, 1:31 PM

#

I don't support MSI (yet)

flint idol Nov 17, 2024, 1:31 PM

#

Hw side there is more difference s

#

A way to get interrupts

short mortar Nov 17, 2024, 1:32 PM

#

Message Signaled Interrupts

#

Basically a "better" way to do them

flint idol Nov 17, 2024, 1:32 PM

#

I support MSI and MSI-X

short mortar Nov 17, 2024, 1:32 PM

#

Except that it doesn't work on RISC-V

flint idol Nov 17, 2024, 1:32 PM

#

And legacy irqs

#

Through int pin

#

It's paid

#

Just use osdev wiki, should be good enough

short mortar Nov 17, 2024, 1:34 PM

#

I don't have it

flint idol Nov 17, 2024, 1:35 PM

#

I do have a spec that explains the relationship between pci and acpi

short mortar Nov 17, 2024, 1:35 PM

#

I found parts of it on a random website with hw (homework) solutions stuff

#

And seen enough of it to implement ACPI

flint idol Nov 17, 2024, 1:35 PM

#

I have the osdev wiki page

#

meme

short mortar Nov 17, 2024, 1:36 PM

#

Wdym

flint idol Nov 17, 2024, 1:37 PM

#

Did you buy it or sum?

white mulch Nov 17, 2024, 1:41 PM

#

you can also find them literally everywhere lol, even on intel's site

flint idol Nov 17, 2024, 1:41 PM

#

white mulch you can also find them literally everywhere lol, even on intel's site

Proof?

white mulch Nov 17, 2024, 1:42 PM

#

just google "pci specification pdf" for yourself, ofc its likely not intended to be public and its some old spec (iirc like 2.0)

flint idol Nov 17, 2024, 1:46 PM

#

this cat says hi

short mortar Nov 17, 2024, 2:05 PM

#

white mulch you can also find them literally everywhere lol, even on intel's site

Not PCIe spec though

white mulch Nov 17, 2024, 2:06 PM

#

short mortar Not PCIe spec though

it is pcie

short mortar Nov 17, 2024, 2:06 PM

#

PCI Firmware*

white mulch Nov 17, 2024, 2:07 PM

#

yeah, for that you have to use archive.org lol

short mortar Nov 17, 2024, 2:08 PM

#

Huh

real pecan Nov 17, 2024, 2:08 PM

#

there was a gh repo where someone leaked all of the most recent pcie and pci specs

#

probably down now tho

white mulch Nov 17, 2024, 2:08 PM

#

short mortar Nov 17, 2024, 2:09 PM

#

I've just found 2021 verison

white mulch Nov 17, 2024, 2:09 PM

#

white mulch

though its not terribly useful imo, the acpi spec already contains most of the stuff iirc

short mortar Nov 17, 2024, 2:14 PM

#

It said "refer to PCI spec"

#

I think

thick jolt Nov 17, 2024, 5:28 PM

#

flint idol It's paid

i buy it :)

white mulch Nov 17, 2024, 5:52 PM

#

thick jolt i buy it :)

for 2k? troll

thick jolt Nov 17, 2024, 5:52 PM

#

white mulch for 2k? <:troll:796858892985434152>

nooo

white mulch Nov 17, 2024, 5:53 PM

#

2k for a little pdf

weak kestrel Nov 17, 2024, 5:57 PM

#

I think drafts are free

flint idol Nov 18, 2024, 7:41 PM

#

Hi guys

#

Kbye

weak kestrel Nov 18, 2024, 8:16 PM

#

kthxbye

vale nymph Nov 18, 2024, 8:44 PM

#

flint idol Kbye

KiBye

flint idol Nov 19, 2024, 3:57 PM

#

Wassup

#

Kbyenow

weak kestrel Nov 19, 2024, 3:58 PM

#

https://tenor.com/view/delay-cat-gif-21402724

Tenor

flint idol Nov 20, 2024, 1:47 PM

#

weak kestrel https://tenor.com/view/delay-cat-gif-21402724

https://tenor.com/view/delay-cat-gif-21402724

Tenor

#

someone forked obos

#

Have no idea who

#

Well I know the github account

#

Yes yes definitely

#

100%

#

Is ur os not on github?

#

Then why isn't it on that github account

#

https://github.com/codingisapaintodo

GitHub

codingisapaintodo - Overview

codingisapaintodo has 6 repositories available. Follow their code on GitHub.

#

That's the account that forked obos

#

Oh yeah it is

#

I saw yeah

#

Btw master is missing many features from userspace-work

limber wave Nov 20, 2024, 1:52 PM

#

wait i did

#

sorry

flint idol Nov 20, 2024, 1:52 PM

#

Np

#

It doesn't really mean anything

limber wave Nov 20, 2024, 1:53 PM

#

wait

#

actualy

#

how do i try it out

#

without making it my default

flint idol Nov 20, 2024, 1:54 PM

#

clone userspace-work

#

And build instructions are in readme

limber wave Nov 20, 2024, 1:54 PM

#

ah alr imma download it

#OBOS (not vibecoded)