Nyaux

[surreal path]

nope

sys_exit

[elder shoal]

Is your bash compiled with readline

[surreal path]

idk

i think it is

wha

[surreal path]

whats wrong with that

??

[elder shoal]

Bash with readline needs a more working system lol

Whats your jinx recipe for bash

[surreal path]

name=bash
version=5.2.26
revision=1
tarball_url="https://ftp.gnu.org/gnu/bash/bash-5.2.21.tar.gz"
tarball_blake2b="6789c9a0d9eb1ad167d4199bf1438d77934a7bbeae9f9fdd7167cae006b17b3894852440248db1bb6e9cf6d930e8a18b6448a3bb4db8831b2e6d1445b56a2065"
source_allow_network=yes
source_imagedeps="curl"
source_hostdeps="autoconf automake libtool pkg-config"
imagedeps="gcc"
hostdeps="gcc autoconf automake libtool pkg-config"
deps="core-libs ncurses readline"

regenerate() {
    curl -Lo bash52-022 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-022
    curl -Lo bash52-023 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-023
    curl -Lo bash52-024 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-024
    curl -Lo bash52-025 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-025
    curl -Lo bash52-026 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-026

    for f in bash52-02*; do
        patch -p0 < $f
    done

        AUTOHEADER=true \
    autotools_recursive_regen
}

build() {
    autotools_configure \
        --with-curses \
        --enable-readline \
        --without-bash-malloc \
        --with-installed-readline="${sysroot}/usr/lib"

    make -j${parallelism}
}

package() {
    make install DESTDIR="${dest_dir}"
    ln -s bash "${dest_dir}${prefix}"/bin/sh

    # post_package_strip
}

copied from vinix

[green elbow]

you have --enable-readline so im guessing it is using readline :^)

you could try remove it and recompile

[elder shoal]

Yeah remove the readline flags

[tawdry mirage]

might explicitly need --disable-readline?

if it looks for it on it's own by default

[surreal path]

umm?

using this flag causes these errors

works without enable readline tho

clearly still dying

fuck

does my bash recipe look wrong?

https://github.com/rayanmargham/Nyaux/blob/master/recipes/bash

[green elbow]

remove the --with-installed-readline too

[surreal path]

oka

bash is STILL Dying

bruh#

[green elbow]

do you actually print stderr now?

[surreal path]

i tried

its printing nothing

[green elbow]

you should just always print that

[surreal path]

void syscall_write(struct syscall_frame *frame, struct per_thread_cpu_info_t *ptr)
{
    int fd = frame->rsi;
    void *buf = (void*)frame->rdx;
    size_t count = frame->r10;
    if (fd == 1)
    {
        // SINCE WE DONT HAVE DEVICES WE JUST WRITE DIRECTLY TO FLANTERM
        flanterm_write(ctx, buf, count);
        frame->rdx = 0;
        frame->rax = count;
        return;
    }
    else if (fd == 2)
    {
        flanterm_write(ctx, buf, count);
        frame->rdx = 0;
        frame->rax = count;
    }
    else {
        kprintf("syscall_write: failure at fd: %d\n", fd);
        frame->rdx = -1;
        return;
    }
    
}

i clearly do

[green elbow]

solid

why does it still say failure tho

[surreal path]

wait

im stupid

it did get to

bash

but then it exited

lemme remove all the debug messages

unknown error code?

its looking for bashrc

like what

[green elbow]

that shouldn't be fatal

[surreal path]

okay this is what cleaner output looks like

[elder shoal]

I think the reason why it exits is orobably related to your read

[surreal path]

why?

[elder shoal]

What does your read return

Or do you even handle stdin

[surreal path]

-1 if err

void syscall_read(struct syscall_frame *frame, struct per_thread_cpu_info_t *ptr)
{
    int fd = frame->rsi;
    void *buf = (void*)frame->rdx;
    size_t size_of_buf = frame->r10;
    struct process_info *pro = get_cur_process_info();
    struct FileDescriptor *d = &pro->Descriptors[fd];
    if (d)
    {
        if (d->ptr)
        {
            kprintf("Buf: %p Offset: %d, fd: %d\n", buf, d->offset, fd);
            kprintf("Addr of vnode: %p\n", d->ptr);
            int hm = d->ptr->ops->v_rdwr(d->ptr, size_of_buf, d->offset, buf, 0);
            if (hm != -1)
            {
                kprintf("old offset is %d ", d->offset);
                d->offset += hm;
                kprintf("new offset is %d\n", d->offset);
                frame->rdx = 0;
                return;
            }
            frame->rdx = -1;
            return;
        }
        else {
            kprintf("Clearly exists\n");
            frame->rdx = -1;
            return;
        }
    }
    frame->rdx = -1;
    return;
}

legit the entire function

[elder shoal]

Do you have a stdin

[surreal path]

no

[elder shoal]

Thats probably it then

[green elbow]

yah could be that it's just reading EOF

and exiting

[surreal path]

wha

now ur making me confused

[green elbow]

add a log to syscall_read and log the fd

[surreal path]

oka

it just reads fd 0 a bunch

[green elbow]

yes that is stdin

that should be reading from the keyboard buffer

[surreal path]

now its not reading fd 0

as im making sure that remains allocated

[green elbow]

allocated?

[surreal path]

as in

used

i have a bitmap that says which fds are used

and which fds are not used

[green elbow]

do you know what stdin is

[surreal path]

standard input

[green elbow]

right

where does that come from

in your kernel right now

[surreal path]

i dont have it rn

at all

[green elbow]

okay that is probably your issue

[surreal path]

fair

[green elbow]

to test it you can even make it infinitely return "a" or smth

[surreal path]

on read?

[green elbow]

yea

in syscall_read add a special case for fd 0 that always returns "a"

in principle that means that your bash prompt will just be filled with "a"

[surreal path]

ok

[unkempt relic]

or even \n

should make bash reprint her prompt

[surreal path]

something like this i suppose

[tawdry mirage]

you need to return the amount of data writteen

[green elbow]

^

[surreal path]

right

mlibc still panicking

[tawdry mirage]

presumably because exit is unimplemented?

[surreal path]

im not trying to exit tho

why is bash wanting to exit

[green elbow]

what does your sys_read look like?

[surreal path]

void syscall_read(struct syscall_frame *frame, struct per_thread_cpu_info_t *ptr)
{
    int fd = frame->rsi;
    void *buf = (void*)frame->rdx;
    size_t size_of_buf = frame->r10;
    if (fd == 0)
    {
        // stdin
        if (size_of_buf > sizeof(char))
        {
            ((char*)buf)[0] = '\n';
            frame->rdx = 0;
            return;
        }
    }
    struct process_info *pro = get_cur_process_info();
    struct FileDescriptor *d = &pro->Descriptors[fd];
    //kprintf("syscall_read: reading fd %d\n", fd);
    if (d)
    {
        if (d->ptr)
        {
            // kprintf("Buf: %p Offset: %d, fd: %d\n", buf, d->offset, fd);
            // kprintf("Addr of vnode: %p\n", d->ptr);
            int hm = d->ptr->ops->v_rdwr(d->ptr, size_of_buf, d->offset, buf, 0);
            if (hm != -1)
            {
                // kprintf("old offset is %d ", d->offset);
                d->offset += hm;
                // kprintf("new offset is %d\n", d->offset);
                frame->rdx = 0;
                return;
            }
            frame->rdx = -1;
            return;
        }
        else {
            // kprintf("Clearly exists\n");
            frame->rdx = -1;
            return;
        }
    }
    frame->rdx = -1;
    return;
}

[elder shoal]

I dont see you returning the amount of bytes read?

[surreal path]

its because i do this in my syscall

*bytes_read = count

returning bytes_read properly doesnt solve it tho

[green elbow]

[surreal path]

wdym concern

[green elbow]

you should properly return how many bytes are read lol

how did you fix it?

[surreal path]

i return the amount of bytes read

properly

as asked

still dont know why bash is quitting tho

its like fuck this shit im out

[wide nexus]

bash is actually running but the little gremlins inside your computer are typing exit every time

just to mess with u

[unkempt relic]

bash instantly exits if it thinks it's not getting any more input, something like that

[surreal path]

but im giving input

'\n' remember

[unkempt relic]

does it actually get as far as calling sys_read()?

[tawdry mirage]

maybe it bails because of poll failing?

[unkempt relic]

i had this problem when i first ported bash, it instantly exited, i forgot the details of what was going on

[wicked loom]

likewise

[surreal path]

im back after 4 days

isnt that crazy

yall thought i left again

[finite summit]

didn't you say somewhere you would take a break from osdev because school

[surreal path]

i did

for studying

i have the exams in a few days but

i've done my studies

i can breath now

anyways last thing we were trying to do on nyaux

hmm

that was bash unexecpectingly quitting

ah yes

[wicked loom]

i did ngl

[surreal path]

why u think that

thats a lil mean

[wicked loom]

because it happened before like twice

[surreal path]

but i gotten so far

in this iteration

its not gonna happen again

[wicked loom]

that's true

[plush hearth]

Are you sure about that?

[surreal path]

yes

😎

[ionic jetty]

Yes

Mindblowing

[surreal path]

absolutely

[ionic jetty]

Indeed

[surreal path]

just wanna let yall know that nyaux is not dead im just not feeling up mentally

to be honest

so its difficult to do anything really atm

im sorry

ill try to resolve these mental issues as soon as possible so i can continue to work on the project

[green elbow]

take it easy bro no pressure

[surreal path]

i will be back on monday tomorrow

to work on nyaux

i had an appointment with my therapist and i feel a little better

[surreal path]

thanks bro means a lot

[shut laurel]

Do a little bit every day, dont work for hours on end

[surreal path]

fair

anyways okay so what we were doing

right

bash

so we loaded bash

but uhhhhh

no stdin

so bash goes like fuck u

right

and ofc

weird ass page fault

happening on my system

cause im using wsl this time

great

which is like wtf

and ofc make run-uefi isnt working on wsl cause fuck me

hmmmmmmmmmmmmmmmmmmmm

how tf am i gonna debug this

lovely

its a general protection fault

when using xsave????

bru

[elder shoal]

Check alignment

Of the address you pass to xsave

It has to be 16 bytes aligned iirc

[surreal path]

wha

its happening at line 22

in this function

on the asm instruction

what i do

[surreal path]

its not happening on task switch

[surreal path]

its happening here for some reason

cause the cpu hates me

like usual

[elder shoal]

https://www.felixcloutier.com/x86/xsetbv

[surreal path]

yea i dont see an issue?

[elder shoal]

Protected Mode Exceptions

[surreal path]

okay so its def not cause of cpl

not xcr either

could be the third exception

umm

hmmm

ughhh what is the issue

i lit set xcr 0 to 7

if the cpu has avx

i cant remember how the layout of the xcr registers are but

im sure thats valid

as it worked before

why now is there an issue

[elder shoal]

Does your host cpu have avx? Idk if qemu software emulation emulates it

[surreal path]

yes it does

on linux is worked

it *

[elder shoal]

Try running with kvm and see what it does

[surreal path]

solves issue on kvm?

tf???

I CHECK FOR AVX????

HUH???

[elder shoal]

Then the checking code is bad or smth or its a qemu issue

[surreal path]

i believe this is correct

yea

so like wtf???

?????

maybe this is not worth fixing, ill just make a linux vm or smthin

fuck wsl honestly

[elder shoal]

[surreal path]

whats wrong with that

wait

oh

yea

but wait

no wait

shouldnt it like

wait

my brain is confused

shouldnt it work fine cause like its a

if statement-ish

[elder shoal]

Read that statement slowly

[surreal path]

no

sorry

my brain is actually tard

poop

i dont know

like im ANDing the bit

so i should get just that one bit right???

why is this statement an issue

as u can tell its been quite a while

i forgot a lot of things

and maybe minus 20iq points

[elder shoal]

Youre checking the edx register for something in the ecx register......

[surreal path]

oh my god

am i

that stupid

wtf

☠️

😭

anyways more issues yay

setting xcr 0 to value 2 is an issue for some reason

fuck where was that diagram of the xcr registers in the sdm again

[elder shoal]

Currently, only XCR0 is supported. Thus, all other values of ECX are reserved and will cause a #GP(0). Note that bit 0 of XCR0 (corresponding to x87 state) must be set to 1; the instruction will cause a #GP(0) if an attempt is made to clear this bit

[surreal path]

the value of 2 does not clear bit 0

wait no

it does

☠️

it should be 3

okay now thats fixed

now we have vfs issues

yay !!!!

fuckin hell

truly a

moment

oh

bash is not compiling???

wheres my mlibc???

WHERES MY SHIT

😠

YO JINX

WHERES MY USERSPACE

yea where tf is my userspace huh

why didnt jinx build mlibc or

bash

or anything

[sleek gazelle]

you made an os and a somewhat usable but still shitty userspace in a span of a month yet my dumb autistic possibly adhd brain still can't figure out how to do smp and apic

[surreal path]

its not shitty userspace

its pro userspace

😎

[sleek gazelle]

my mistake

[surreal path]

u sound like me a bit ago

i dont even have smp in my kernel

and apic is actually kinda easy

its just

[sleek gazelle]

lol

[surreal path]

a bunch of registers u write to

[sleek gazelle]

i do have smp though its still somewhat shit

[surreal path]

can we eat jinx

im actually quite mad

wheres my damn userspace

compile it jinx

rawr

[elder shoal]

The kernel can have a little instabilty as a treat, it adds risk and reward to the computing experience

[sleek gazelle]

i somewhat have problems with doing apic since theres no comprehensive code and docs for apic that i can reference

[surreal path]

maybe i need to do dist clean?

maybe something fucked up with the build or something

and jinx didnt compile bash and mlibc for some whatever reason

this compile may take like

30-40 minutes

[sleek gazelle]

and some apic code written in zig doesn't really fit in with (how i setup) my code

[surreal path]

(my kernel )

anyways i will try to run

make dist-clean

and see if jinx will be epic gamer and compile mlibc and shit for me

[surreal path]

bro

why isnt jinx

compiling bash

i did dist clean

STILL

i believe it was my makefile

should be solved now hopefully

🤞

ok

now we are back to the same point we were before

where bash poopies for some reaosn

reason*

uhhhhh

okay so why does bash do this hmm

[elder shoal]

Likely because it thinks it hit an end of file

[surreal path]

huh??? why????

with what??

????

[elder shoal]

See what read returns

[surreal path]

this doesnt get printed

nor this

so clearly something else is happening???

huh???

is it cause i error out when someone asks to duplicate a file descriptor?

[elder shoal]

You dont need dup for a basic bash promot

Whats your whiole read function both mlibc and kernel side

[surreal path]

void syscall_read(struct syscall_frame *frame, struct per_thread_cpu_info_t *ptr)
{
    int fd = frame->rsi;
    void *buf = (void*)frame->rdx;
    size_t size_of_buf = frame->r10;
    if (fd == 0)
    {
        // stdin
        if (size_of_buf > sizeof(char))
        {
            ((char*)buf)[0] = '\n';
            frame->rdx = 0;
            frame->rax = sizeof(char);
            kprintf("read(): stdin was read\n");
            return;
        }
    }
    struct process_info *pro = get_cur_process_info();
    struct FileDescriptor *d = &pro->Descriptors[fd];
    //kprintf("syscall_read: reading fd %d\n", fd);
    if (d)
    {
        if (d->ptr)
        {
            // kprintf("Buf: %p Offset: %d, fd: %d\n", buf, d->offset, fd);
            // kprintf("Addr of vnode: %p\n", d->ptr);
            int hm = d->ptr->ops->v_rdwr(d->ptr, size_of_buf, d->offset, buf, 0);
            if (hm != -1)
            {
                // kprintf("old offset is %d ", d->offset);
                d->offset += hm;
                // kprintf("new offset is %d\n", d->offset);
                frame->rdx = 0;
                frame->rax = hm;
                return;
            }
            frame->rdx = -1;
            return;
        }
        else {
            kprintf("read(): descriptor of value %d was not found\n", fd);
            frame->rdx = -1;
            return;
        }
    }
    frame->rdx = -1;
    return;
}

mlibc:

int sys_read(int fd, void *buf, size_t count, ssize_t *bytes_read) 
{
    __syscall_ret ret = __syscall(3, fd, buf, count);
    if (ret.errno != 0)
    {
        return ret.errno;
    }
    *bytes_read = count;
    return 0;
}

[elder shoal]

You should add a system call log

[surreal path]

i have

i have a syscall log

???

[elder shoal]

Also fix the bytes read to be what you return from the read syscall

[surreal path]

yea changed that

to which it solved nothing

[elder shoal]

What fd is it trying to read from?

[surreal path]

lemme see

fd 2

[elder shoal]

Is mlibc complaining about any missing functions? How is your bash recipe

[surreal path]

i dont think so?

doesnt seem like it?

this is my bash recipe

name=bash
version=5.2.26
revision=1
tarball_url="https://ftp.gnu.org/gnu/bash/bash-5.2.21.tar.gz"
tarball_blake2b="6789c9a0d9eb1ad167d4199bf1438d77934a7bbeae9f9fdd7167cae006b17b3894852440248db1bb6e9cf6d930e8a18b6448a3bb4db8831b2e6d1445b56a2065"
source_allow_network=yes
source_imagedeps="curl"
source_hostdeps="autoconf automake libtool pkg-config"
imagedeps="gcc"
hostdeps="gcc autoconf automake libtool pkg-config"
deps="core-libs ncurses readline"

regenerate() {
    curl -Lo bash52-022 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-022
    curl -Lo bash52-023 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-023
    curl -Lo bash52-024 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-024
    curl -Lo bash52-025 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-025
    curl -Lo bash52-026 https://ftp.gnu.org/gnu/bash/bash-5.2-patches/bash52-026

    for f in bash52-02*; do
        patch -p0 < $f
    done

        AUTOHEADER=true \
    autotools_recursive_regen
}

build() {
    autotools_configure \
        --with-curses \
        --without-bash-malloc

    make -j${parallelism}
}

package() {
    make install DESTDIR="${dest_dir}"
    ln -s bash "${dest_dir}${prefix}"/bin/sh

    # post_package_strip
}

[elder shoal]

Well id seems like youre building with readline. Did you stub poll

[surreal path]

yes

[elder shoal]

Readline needs poll

Yeah thats why its breaking lol

[surreal path]

idk what readline is so

ima remove the dep

[elder shoal]

What lets you line edit in bash

[surreal path]

huh

ok building without readline

still same issue

so sad

what is ncursers

do i remove that

is that causing the issue

??

[elder shoal]

Could be, try removing it too

[surreal path]

okay

[elder shoal]

And remove --with-curses

From the configure

[surreal path]

okay

so

okay

yea

linker errors

for some reason

[elder shoal]

I think its trying to build with its internal readline

Do --without-readline

[surreal path]

okay

still

even with

[elder shoal]

Wait its

--disable-readline

https://github.com/Mathewnd/Astral/commit/145df46088b978421691de5d4c502a7cb8eefcfe

This one worked without poll

[surreal path]

compiles now

OKAY IT WORKS

LETS GO

AND MY FUNNY \n STDIN SHANNGAINS

WORKS TOO

WOHO!

okay do i hook up my ps2 driver to stdin

or what

[green elbow]

that would be a solid next step

[elder shoal]

You make it a device

[surreal path]

i dont have devices

[elder shoal]

You create them

[surreal path]

i think thats bash

is that normal

i think thats normal right

[surreal path]

how

i have zero idea how to implment devices

[elder shoal]

Theyre vnodes just like any other file

[surreal path]

uh huh yea

i get that

but what extra functionality do dev devices need or whatever

[elder shoal]

They have major and minor numbers you use to see what device the vnode points to

[green elbow]

not much except they have different implementations of read() and write()

and sometimes seek is meaningless on them

[surreal path]

wha

my brain has just exploded

💥

[green elbow]

you don't need that

device files are just files which are not backed by a filesystem but instead by some mechanism

for example /dev/zero is just a file with a read() function that looks like

int read(char *data, size_t amount) {
  memset(data, '\0', amount);
  return amount;
}```

[surreal path]

so for my case

i could have a file like

/dev/keyboard

and its read could look like

it could return the thing in the keyboard buffer and stuff

and its write??? will do nothing i suppose

is this a good way of implmenting this?

[elder shoal]

For terminal io it gets a bit more complex

[surreal path]

oh

[elder shoal]

But yeah for a keyboard file its p much that

[surreal path]

please explain

i want to implment this correctly

[elder shoal]

You have to do some processing on the data you receive

Look up tty demistified

I think thats the name of that article

And ig you can also look at astral shitty™️ code for some implementation

[surreal path]

reading the article is just making everything even more confusing

💥

[green elbow]

read it again

[surreal path]

i am trying

my brain is still cooked trying to understand any of this

[green elbow]

dont read about cooked mode then

[surreal path]

bru

[elder shoal]

Is that when you put the terminal in the oven

[wicked loom]

nyaux

[green elbow]

yah and then add it to a stir fry

[surreal path]

nyaux

yes what about nyaux

[surreal path]

1 whole egg

2 flour

1 cool bean

mix together

anyways i may take a nap

be back in 2 hours

[leaden hinge]

and all the other things probably

like mmap

[surreal path]

ok back

[surreal path]

so

hm

what was i doing again

right

[kind root]

procrastinating

[surreal path]

im not

😠

[kind root]

i know i am

[surreal path]

i have a tty to implment

what about you hm??

exactly

😎

[kind root]

just opened a pr

[surreal path]

fair

[kind root]

but im still not happy about it

[surreal path]

i keep rereading the tty demystified article but nothing of what it says makes sense to me at all

maybe im stupid

[finite summit]

I doubt I'd understand ttys

they seem like a dumb concept

[surreal path]

yea

[kind root]

it wont work unless u write code in userspace to experiment

e.g. i wrote code to hijack a working tty's stdin and it helped me reinforce my understanding

[surreal path]

the thing is i dont understand a single thing like at all

[kind root]

read again and again until u do

what is it u dont understand

[finite summit]

inb4 everything

[surreal path]

like this whole line displince thing

yep

[kind root]

often times u need multiple source to grasp something

google line discipline

see different explanations

look on youtube

different sources

if you dont put in effort you dont get the knowledge

[surreal path]

i am

[kind root]

then you'll get it

[surreal path]

looked through a bunch of resources and i still dont, i think im clearly braindead

maybe the 3 weeks of me being away from nyaux has made my iq go down by 20 points

i think i played too much valorant during that time

[elder shoal]

it processes a stream of bytes that come from, for example, a serial port

if you want an implementation to look at you can take a look at astral io/tty.c, io/pty.c (for the pseudo terminal backend) or io/console.c (for the /dev/console backend)

[steady flume]

nya?

[edgy pilot]

ux

[kind root]

master

[plush hearth]

4.0

[ionic jetty]

OP

[marble surge]

Jun 8, 2024

[kind root]

looked through a bunch of resources and i still dont, i think im clearly braindead

[marble surge]

I think that message might have been meant for another thread unless I'm misunderstanding something

[plush hearth]

You are misunderstanding. Have a read of this message: #1230349543623757845 message

[marble surge]

Ohh okay 😅

[plush hearth]

maybe the 3 weeks of me being away from nyaux has made my iq go down by 20 points

[median goblet]

No it hasn't

[plush hearth]

you don't understand what is happening here

[kind root]

i think i played too much valorant during that time

[median goblet]

I'm pretty sure it's physically impossible to lose 20 points of IQ in just 2 weeks

You need advanced stage brainrot to do that

[plush hearth]

yes, but by you replying to me like that, it makes me think you don't understand what is happening

[kind root]

Because they dont

[thorn bramble]

did linuxmaster take over your account

[kind root]

lol

[kind root]

Looks like nyaux died

[finite summit]

RIP

[finite summit]

@surreal path You good?

[ionic jetty]

They promised us that they‘d return, right?

Last seen 28 days ago

[plush hearth]

There was a lot of progress fairly quickly at the start of this, and then it just seemed to die

[ionic jetty]

I hope the thing that I tend to do is not happening, being incredibly interested in something for a while and then suddenly not really caring about it anymore

[finite summit]

linuxmaster seems to have disappeared along with his OS

[plush hearth]

.

It happened again

[thorn bramble]

rip

[wicked loom]

i called it

[kind root]

It was obvious tbh

Tty ended up being too hard to understand

And people weren't spoon-feeding as much

[thorn bramble]

i've been wanting to give up on my osdev journey for like 6 months at this point

writing code that doesn't suck and isn't fundamentally flawed too hard :D

[finite summit]

I'd give up on osdev but I got nothing else to do

except game dev

but osdev > game dev

[finite summit]

https://github.com/rayanmargham/Nyaux/commit/7f35e6475165f3e0d7663cc1a149179dab274c4f

just found this commit from nyaux master

[elder shoal]

holy shit

where is nyauxmaster...

[kind root]

Its also completely wrong lol

But looks like hes alive at least which is nice

[finite summit]

so I'm not the only one who thought he died or something

good

[ionic jetty]

@surreal path

[surreal path]

HI

im back

for the

fuckin trillionth time

@kind root guess whos still alive

[surreal path]

im good

[surreal path]

im fine yea

just want to say sorry to everyone for being gone again

for the billionth time

my motivation to do anything is like nearly 0

😭

so i just stopped doing any kind of anything

for a while

[surreal path]

you did lmao

i dunno what to do now, continue the project. scrap and rewrite? not rewrite and just continue? idfk

im still alive (sadly)

which is the worst thing

time to check if nyaux still compiles

[kind root]

Where the f have you been

Its been like months lmao

[surreal path]

mental illness

!!

[kind root]

[surreal path]

i could not escape the big depression

anyways im gonna verify if nyaux still compiles and go from there

[magic kernel]

https://tenor.com/view/the-return-of-the-king-lord-of-the-rings-ro-tk-title-card-gif-22191928

welcome back

you were making pretty good progress, so looking forward to see what's next

[surreal path]

thank you !!!

time to wait forever to jinx do setup the container

or whatevs

lovely

clearly something bad

is jinx outdated

probs

okay got the newer jinx

oh yeah btw @kind root forgot to ask, hows life going with you? hows the uacpi project going?

[leaden hinge]

welcome back

[surreal path]

hi solar

[leaden hinge]

howve you been

[surreal path]

not great which is why ive been gone for a while

but im back now

[leaden hinge]

ah i see

well i hope things are better

and nice to see you back

[surreal path]

nice to see you too

[leaden hinge]

gonna continue with nyaux?

[surreal path]

yep

[leaden hinge]

it was going pretty well last i recall

[surreal path]

right back where we left off

i believe it was

[leaden hinge]

super

usually this would be the point where i rewrite everything for no reason

[surreal path]

something with implmententing device files

[leaden hinge]

this is where the fun begins!

[surreal path]

yep !

[leaden hinge]

i just finished doing that the other day actually

[surreal path]

im just recompiling nyaux

see if everything still compiles well

[surreal path]

oh great !

[elder shoal]

Holy shit nyauxmaster 4.0 is back

[plush hearth]

welcome back

[ionic jetty]

no way

They have returned

what‘s on the agenda

[kind root]

ive been on a bit of a hiatus as well but slowly getting back and wrapping uacpi up for a 1.0 release hopefully some time soon

nice to see that ur back

[unkempt relic]

we can post it on hacker news and reddit

ought to be fun

[leaden hinge]

"whats aml" incoming

[unkempt relic]

it will be full of people denouncing ACPI as the devil's daughter

[leaden hinge]

or that yes

[kind root]

that would be cool lmao

[surreal path]

yep

[surreal path]

thank youu

[surreal path]

okay so gotta fix some of these

[kind root]

yeah this is new stuff

[surreal path]

yea ik

okay got it to compile

huh weird

it takes forever

oh im running out of ram

lmao

huh it still stays stuck here

something weird

is happening

not cool

fun

yea clearly this is a nonsense address

[desert haven]

Welcome back 😁

[surreal path]

hi r4

[desert haven]

Oh someone already posted it haha

[surreal path]

difficult to debug when its one of those kind of issues

😭

[leaden hinge]

anything useful in the call stack?

[surreal path]

good question

[surreal path]

lovely

[leaden hinge]

my favorite kind of error

[surreal path]

truly

no interrupt happened as well

and no cpl change

so somehow the cpu is jumping to some weirdo address and just doing whatever tf

[desert haven]

Maybe take a look at what the other codes are doing

[leaden hinge]

did you iret?

or sysret or whatever

[surreal path]

im unsure

this shouldnt happen tho

cause all interrupts are handled with an iret

[leaden hinge]

you might be messing the stack up before iret

[surreal path]

possible

somewhere here

there seems to be an issue

oh

could be cause uacpi changed a fuck ton while i was gone

[desert haven]

Yes it certainly has

[surreal path]

yea i suppose

lovely that ossdev wiki is down for the billionth time

so i cant see how the api changed for uacpi

[leaden hinge]

osdev.wiki is up for me

[surreal path]

its hella slow for me

wait its coming back

i dont see a problem with how im trying to get the madt table

hmm

man i love maintaining really old code

ie code i wrote 3 months ago

okay

thats correct

also correct

[kind root]

why is table a pointer here

[surreal path]

good question

yea still not working

the urge to rewrite everything

[kind root]

madt = table is wrong as well

do u not have warnings or what

[surreal path]

of course i do but i havent touched this code in 3 months

[finite summit]

Nah it's fine. Hope you're okay now

[finite summit]

There is a fork of the osdev wiki at https://osdev.wiki

[thorn bramble]

holy fucking shit

i thought this day would never come

welcome back king

[elder shoal]

The nyauxmaster has come to save us all

[finite summit]

https://tenor.com/view/the-return-of-the-king-lord-of-the-rings-ro-tk-title-card-gif-22191928

[surreal path]

😎

okay so

the code in nyaux is WAY too crusty

difficult to wrap my head around what the fuck i was trying to do a few months ago, nearly unreadable code

i think the best option is to do a rewrite

NOW BEFORE anything hear me out

[thorn bramble]

please no rust

[surreal path]

😭

[thorn bramble]

or, actually, go have fun :^)

i should try rust osdev one day

[finite summit]

no rust

but I'm serious /j

[leaden hinge]

i would personally advise against it

it takes longer to rewrite than to get used to your old code

[surreal path]

ill see what i will do

[surreal path]

on a trip on italy btw

[elder shoal]

nyauxmaster 4.0 travel arc

[frigid cliff]

Have fun

[surreal path]

okay got my laptop pluggedi n

brung it with me for italy

lets

fuckin

do this

rewriting nyaux day 1

[kind root]

Brung?

[surreal path]

i cannot spell

😭

im working on nyaux in vc0

[elder shoal]

nyaux in rust real

[surreal path]

real

yep i got macros wokring

[steady flume]

isn't that "bring" but in past tense?

[kind root]

Brought

[steady flume]

🤦 of course

[steady flume]

LETS FUCKING GO

[kind root]

Is this a rewrite

[steady flume]

probably

main.rs

[surreal path]

yep

btw i got some nice pics of Italy, u wanna see @kind root ?

[kind root]

Yes

[surreal path]

wait

@kind root thats me btw face reveal crazy ik

[kind root]

lol nice

u look like ur pfp

[surreal path]

i do indeed

i get that a lot

anywayssss time to get some memory management goinggg

maybe ill do a bitmap allocator this time around

nah

[wicked loom]

bruh you look like a child

[surreal path]

bro im 17 mint 😭

why does everyone say that

[wicked loom]

wtf

off of that pic i'd say you were 14 or smth lol

[surreal path]

😭 😭

[wicked loom]

maybe even less

[surreal path]

idk why i look that young

[wicked loom]

i mean it's not necessarily a bad thing

to most people at least

[surreal path]

yea i guess

cool got memory entries

lets heckin

throw a bunch of freelist nodes into each free page 🌞

😎

and link em all

then boom done finished freelist allocator

kkkk nice

[alpine mulch]

isnt that bracelet one of those that they give you in hospitals/mental institutions

are you on the run

[surreal path]

no 😭

[alpine mulch]

we can hide you

theyll never get more of us osdevers alive

[surreal path]

!!!

rust is a mistake

[finite summit]

common rust L

[surreal path]

truly

[surreal path]

hi guys so still working on the PMM

im sorry im not at the speed i should be at

difficult when on holiday and doing it in a completely different language

but its a fun learning experience

[surreal path]

😭

i was expecting to use unsafe a lot more actually

only rlly needed it for dereferencing raw pointers

thats great btw it works

my memory allocation works !!!!

what abouttttttt

500 pages

it did it very nice

and its not even running under kvm

thank you @thorn bramble for the suggestion of using this kind of allocator

[finite summit]

Don't worry about it

[surreal path]

thanks for being really supportive

[finite summit]

Always know that no one here is going to pressure you to work on this

[surreal path]

thanks again

im def just gonna work on it whenever i have free time

problem now tho

[cinder plinth]

when did you learn rust

[surreal path]

like a month ago

[finite summit]

You're the rust compiler you should know

[surreal path]

exactly

actually the rust compiler has been really nice to me

i dont fight with it too much

only for raw pointers

anyways slab allocator is next

man i love

making the terminal look pretty

[finite summit]

who doesn't

[surreal path]

ikr

[thorn bramble]

i got past the "must look pretty" phase

i like informative logs

i like to pack as much info as possible without making the line too long

[finite summit]

my logs have mainly been:

bonjour``` for the past few days

[thorn bramble]

ew frenchie

[finite summit]

#1141057599584878645 message

[finite summit]

there's this sticker on another discord I'm on

and I was thinking of something I could log in my irq handlers so I would know they got hit

so I chose 'bonjour'

and also 'hola'

don't worry I'm not french

[thorn bramble]

thank god

[elder shoal]

my logs are
modulename: slop

[plush hearth]

I was just thinking that...

[surreal path]

apologies i dont get much time

difficult to work

but im trying my best

anyways

we need a PMM working do we have that yes we do

but deallocation

oh hell nah

gonna be difficult

lets do this

[surreal path]

okay

dealloc works i think

[finite summit]

if it's a freelist you just make a node add it to the list

[surreal path]

no its a bit different

[finite summit]

with a page count?

[surreal path]

yep

[finite summit]

same thing

just add the node to the end

and set the page count

[surreal path]

right

[finite summit]

https://github.com/OBOS-dev/obos/blob/master/src/oboskrnl/mm/pmm.c#L189

you can ignore all from size_t nMemoryLeft = nPages * OBOS_PAGE_SIZE; to the end of the function

that's just so the kernel can wait for physical pages to be freed

[surreal path]

where the structs

like freelist node

[finite summit]

top of file

[surreal path]

right im stupid

sorry

[finite summit]

npnp

[surreal path]

wait

i think i have a better design

im thinking wait

[thorn bramble]

yea i kinda trolled u with the pmm

but as oberrow said you can add single pages back to the list as new regions

but in general what i do in my kernel is i have N lists of free blocks of order N, what that means is that each block is of size 2^(N+12) and is aligned to that boundary

i still have that "memory region" thing just to make sure whenever im freeing memory it does belong to a valid memory region + i reserve some memory out of each usable region for the page info storage

[median goblet]

i just dont deal in more than 1 page at a time.

has served me well enough

[surreal path]

bru

im just gonna do a freelist

im not getting trolled

yea inserting it in EVERY page

is stupid

but idc

okay freelist is done actually

its the same simple one page allocation freelist

but it doesnt take forever to boot

thankfully

[surreal path]

hi all just implmented the gdt into nyaux, cause rust is a bitch when it comes to asm files i have to fuckin do this janky ah system with using build.rs to compile my asm code

and

it uses

gas syntax

🤮

so i got a tool online to convert nasm syntax to gas

cause i dont know how to write gas synta

call it skill issue idc

ull never find me writing gax syntax

[elder shoal]

Liquid syntax when

[surreal path]

tf is a liquid syntax???

[molten grotto]

global_asm!(include_str!("myfile.asm"))

[surreal path]

fucking thank you 😭

[molten grotto]

and it uses intel syntax because its parsed by rust the same way as inline asm (well it is inline asm just loaded from a file)

[surreal path]

wait nvm its still gas syntax

[molten grotto]

no

you have to use gas directives but the syntax is intel

meaning eg. instead of global you use .global

[surreal path]

how do i do a relative jump in gas

its easy with nasm

but im unsure how to do it in nasm

gas*

[molten grotto]

lea rax, [rip + whatever]

[surreal path]

works

lets gooo

🌞

😎

triple fault!

lovely

time to see the issueeee

WHERES MY DAMN LOG QEMU 😭

it should give me some

debug

where is the debug output

i see nothing in my terminal

yea no matter what flags i put

no debug output

amazing !

[ionic jetty]

normally it‘s because of kvm

but you don’t seem to have it enabled

try logging to a file instead

[surreal path]

how to do it again

sorry i havent used qemu in a while 😭

[ionic jetty]

-d int -D qemulog -serial file:seriallog

[finite summit]

you could use .intel_syntax noprefix

but I do agree that at&t "GAS" syntax is dogshit

[finite summit]

based

[sleek gazelle]

if you thought gas syntax was dogshit, you haven't seen plan9 syntax

they use untypeable unicode symbols

that you need to either copypaste or type with the compose key if you have that

[surreal path]

Will be back soon me and my family going to the beach

[finite summit]

have fun

[surreal path]

thanks Omar

[shut laurel]

:

[surreal path]

nekoneko nokonoko dostantan

[shut laurel]

didn't intend to send this lol??

[finite summit]

Lies.

[surreal path]

u cannot escape the brainrot

[surreal path]

they changed the qemu logging 😭

nvm

page fault

fun

me when rust

how tf

am i supposed to debug this

help

so its happening at reload cs

how do i make this possible to debug with rust

[thorn bramble]

look at the disassembly

[surreal path]

oh im idiot

i forgot i dont have a tss

done

now we have gdt

😎

uhhhh whats next

right

page tables

fun

lets open the sdm and get to work!!!

[surreal path]

okay i did the things for the flags and stuff

love bitflags great crate

now im uhhhshit

my brain is gone

[kind root]

There's a crate for... flags?

[surreal path]

yk like uhhhh what u call it

#define FLAG_w (1 << 2)

things like this right

this crate gives me that functionality

[kind root]

Why is a crate needed for that?

[surreal path]

i dont fucking know

to be honest

but its whatever

[edgy pilot]

when rust:

[surreal path]

literarly

okay im gonna go read back all the things from last month

from everyone explaining paging to me

and be back in like 20 minutes

will have a page table manager done then after like 25 minutes hopefully

[edgy pilot]

1 << 1 should be wallowed

[surreal path]

walloed?

[edgy pilot]

write allowed

it's always readable when page is present

[surreal path]

okay i renamed it

one thing im guessing is i need to map the kernel binary

if im going to create a new page map

as well as the hhdm?

wait

no looking back at my old code

now im completely confused

my brain has exploded fuck

okay thank god old rayan made a diagarm for me to go off by

i need to copy the hhdm and the kernel virtual mappings into the new pagemap

right in order to map the kernel

i need the kernel size in bytes

which i do NOT know how to get in rust

C its easy with the linker script