Astral

[prime juniper]

sshd needs tcp server side

ssh had some funnies with resolving which port to connect to

so doesn't work (yet)

[grand shadow]

I have tcp server side so I should see if it works once I make my tcp stack less bad

[split gazelle]

minecraft bedrock is written in c++ and has microsoft code in it

[lean kettle]

minecraft bedrock is basically a shared library

[grand shadow]

for all the haters who doubted me

tcp can now download at a miraculous rate of ~300 kb/s

technology is awesome

time to build https support into this thing

[prime juniper]

Nice job

[lean kettle]

now im interested in finding out how fast managarms TCP/IP stack is

i dont think its actually that slow

[grand shadow]

I'm going to port openssl and try to download some big file and see how fast this can get

[robust geyser]

according to the video dennis recorded, he managed to download gnu hello tarball at 160KB/s

common monolithic kernel W

(for astral obviously)

[robust geyser]

is the bottleneck now your connection? the fs? or still the netstack?

[lean kettle]

i tried seeing if i could install speedtest-cli with pip on managarm but i realised that

1. we dont ship pip
2. venv doesnt work lmao

[robust geyser]

i kinda want to port my xhci driver+usb stack to astral, it could be a fun challenge

[lean kettle]

i just got 550 kb/s

common managarm W

[robust geyser]

maybe it was related to the level of logging dennis had configured

i think he was logging quite a lot

[lean kettle]

probably more related to dennis iirc still running his vm with a single core and without kvm because reasons

[robust geyser]

nah hes running multicore with kvm

[lean kettle]

although i am probably also logging a bit less than he is

i am logging all posix requests tho

without logging requests i get between 700kb and 1000kb/s

[grand shadow]

no idea, I still gotta get openssl to test on a large file

[analog berry]

Whats the bottleneck?

[lean kettle]

currently trying to find out

as much as i know we have bigger problems, i kinda want to optimize the net stack now

[grand shadow]

wtf is this build system agh

[lean kettle]

currently trying to make managarm system compile with -Ofast instead of -O2

[robust geyser]

yeah openssl is terrible lmao

[analog berry]

Just to flex on astral lmao

[lean kettle]

correct

we need a victory

[robust geyser]

let me build astral

[lean kettle]

although astral is also super based and impressive

[grand shadow]

I will rename astral to tcpstral and focus all my effort on the netstack

[robust geyser]

ill get that going at a gigabit soon

:^)

[grand shadow]

I don't even have tcp upstreamed yet

[lean kettle]

all hobby os's that can get shit working are automatically based (unless they are written in rust)

[grand shadow]

and the code is bad tbh

[robust geyser]

:^(

[analog berry]

Netbsdstral

[robust geyser]

let me try and build astral anyway

[grand shadow]

like its messy and I really need to do a proper cleanup on it

do it

[robust geyser]

jinx, nice

[grand shadow]

honestly I will just take heavy inspiration™️ from the managarm recipes for this I'm not gonna spend the next few hours figuring this out

mammogram wins again

[analog berry]

Nyaux had a working tarball fs

Does that make it based

[robust geyser]

vinix recipes are jinx

i recommend those instead ;^)

[grand shadow]

does vinix have openssl

[robust geyser]

of course

or at least

it should

maybe not...

[grand shadow]

it does

[robust geyser]

damn

i wonder what it needs it for

[grand shadow]

apparently taisei and libzip

[robust geyser]

hmm fs/tmpfs.c: In function ‘tmpfs_link’: fs/tmpfs.c:248:23: warning: comparison between pointer and integer 248 | if (node->vfs == V_TYPE_DIR) | ^~ [CC] io/block/vioblk.o [CC] io/pci.o io/block/nvme.c:10:10: fatal error: event.h: No such file or directory 10 | #include <event.h> | ^~~~~~~~~ compilation terminated.

[analog berry]

Wait why is vfs pointer storing the file type

[grand shadow]

wait wtf

how the fuck did I not notice that

aaaaaa I thought I nuked all includes of event.h

that had a nasty lost wakeup problem

@robust geyser pull now

[robust geyser]

yeah that seems to have worked, thanks

[grand shadow]

I do have like 12381292183129 unstaged changes rn so once I get openssl ported I will probably upstream them

[robust geyser]

now building the rest of the distro

funnn

[grand shadow]

well https is here

[robust geyser]

based

[grand shadow]

gonna see that certificate stuff

[wild marsh]

the recipes build with -Og by default?

that's what debugoptimized is

[robust geyser]

i notice that sometimes astral is eating my keystrokes

when typing fast in the terminal sometimes it skips a key or two

[grand shadow]

weird, that doesn't happen here

is it on fvwm or the console

[robust geyser]

fvwm

i just boot using the default option

[grand shadow]

what is your qemu version

[robust geyser]

uhhh

8.2.4

[grand shadow]

hm then that weird bug isn't supposed to happen

maybe its because of the mouse focus thing

that never happened to me

[robust geyser]

doesnt happen anymore, weird i agree

also nvme boot not working, is that expected?

[lean kettle]

managarm-system was compiling with -O2

i checked by adding an #error to a file and checking the command

also compiling as release + removing a couple logs gives me a consistent 1MB/s

http is also slower than https for some reason, lmao

[robust geyser]

__assert(devfs_getbyname(root, &backing) == 0); in kernel_entry fails when using any of the nvme boot options

im assuming the nvme device just isnt there

[wild marsh]

huh i misremembered debugoptimized is -O2 -g

[robust geyser]

which would make sense, i dont see anything on the qemu cmdline regarding that

[wild marsh]

not -Og -g

my bad

[lean kettle]

im going to try -Ofast now

[wild marsh]

Ofast won't help for managarm

[grand shadow]

I don't actually ship the script to generate an nvme image lmao

[lean kettle]

[grand shadow]

there isn't even an nvme drive attached to the vm

[wild marsh]

or well, it won't be any better than O3

[grand shadow]

you can use vioblk as well (and change the root in limine.cfg)

[robust geyser]

also, have you thought about shipping a clang-format file?

[prime juniper]

Mind you, slowdowns due to recording were observed

[wild marsh]

or maybe only slightly better, due to -fno-semantic-interposition

[grand shadow]

what does that do

[lean kettle]

from 1mb/s -> 160kb, lol

[prime juniper]

Used to. And always with KVM I’m not that sadistic

[robust geyser]

https://clang.llvm.org/docs/ClangFormat.html

basically a tool to format your code

you can configure it using a .clang-format file

[prime juniper]

Yeah it sucks. Took me a while and a tiny mis configure lead me to a two week debugging of the ssl state machine. That sucked

[robust geyser]

a lot of people like to format code to keep it consistent, and a clang-format file would be nice for other people to have a way to follow the code style of astral

[prime juniper]

Traitor

[robust geyser]

with all due respect

jinx > xbstrap

and you know it :^)

[grand shadow]

hmmmmm

[prime juniper]

You really think I take 8 mins to configure gun hello in my unrecorded test runs? Lol

[lean kettle]

no

[grand shadow]

I should get lynx ported so I don't have to manually type in the download link of stuff every time

[prime juniper]

Still all based on my work so I’ll be taking that W thank you very much

[grand shadow]

its not webkitgtk but its something

[prime juniper]

Doable (or something along those lines). I forgot if I ported lynx or links

But I have a text mode webbrowser in managarm

[robust geyser]

i mean i dont want to force you to do that, it would be nice but it may take some time to get everything tweaked so it formats the code exactly the way you want it to

and even then it wont be perfect

for that you just need to // clang-format: off + // clang-format: on

or learn how to tame it with strategic // at the end of line :^)

[prime juniper]

It was links

[grand shadow]

do I go for lynx or links

hmmm

[robust geyser]

managarm does links

[grand shadow]

I mean links is graphical so I might as well do that

[prime juniper]

Also supports non graphical if you run it without -g

if I recall correctly

[grand shadow]

oh perfect then

[robust geyser]

also... include/x86-64/arch/ist.h

why do you call the tss "ist"

[prime juniper]

There must be a reason why I didn’t do lynx. Maybe shit build system or some weirdness

https://github.com/managarm/bootstrap-managarm/blob/master/bootstrap.d/www-client.yml contains links

[grand shadow]

its called the interrupt stack table in x8664 iirc

doesn't really matter

[robust geyser]

i thought it was the "task state segment"

[grand shadow]

oh wait it is tss

oh well I'll rename it later

[robust geyser]

it contains the ist, but the entire structure is called the tss :^)

no worries

im just snooping around

this just caught my eye in a particular way

im trying to pick up the coding conventions, see how things work

so much code that could be simplified

but i must resist the urge

[prime juniper]

Local man gonna work on astral?

What happened to your own OS iteration 69420

[robust geyser]

yea, maybe an xhci driver or something

i dont have motivation

and i dont know how to do shit

[prime juniper]

Classic

[prime juniper]

Same

[prime juniper]

Go to userspace. Run wayland

You can do it

[robust geyser]

yeah i can

i mean i know i can

but i just have to do so much shit

i want to do stuff properly

[prime juniper]

As one should

[robust geyser]

yea but idk how 😭

[prime juniper]

From the start

[robust geyser]

lol ur so right

[prime juniper]

Get bash and coreutils and friends

Then probably /dev and /sys

[robust geyser]

yea let me port them to kernel space real quick

[prime juniper]

I hear udev

[prime juniper]

No no from userspace duh

[robust geyser]

yea but i need so much shit for that

i dont want to speedrun a shit scheduler + vfs again

[prime juniper]

Then get to work

[winged basin]

keyronex networking is drastically slowed by printfs in the code

that's one to watch out for

[robust geyser]

i know i know

its just too much stuff to handle

without a clear path

[winged basin]

ordinarily networking is prioritised over everything else because the stack below the sockets themselves run in DPCs

[prime juniper]

Speaking of keyronex related items, I should upstream the InitWare changes

[winged basin]

but with printfs in the way the DPCs get hung

i have no idea if astral prints during networking but it's a low hanging fruit

[prime juniper]

and tell @winged basin to continue work on it, when new udev and tmpfiles

[robust geyser]

i feel like im never gonna get my shit together and write anything good

[prime juniper]

Relatable

Also incorrect

[robust geyser]

every time i sit down and try to do something useful i just get overwhelmed by how much shit i have to do

[prime juniper]

Take it slow. One item at a time

I fell and fall for that trap too often still

[winged basin]

some day...

[prime juniper]

Just in a different scope

[robust geyser]

tbh its easier said than done, i wish i could plan things out but i dont have enough knowledge on what to actually do

[prime juniper]

Feared you’d say that. Maybe I can entice you with Managarm support PRs for InitWare (for real tho, I need to get this old ass udev to behave while I’m waiting on new udev)

[robust geyser]

i want to have a nice vmm, robust interrupt system, i have no idea how everything is tied together just yet and that annoys and discourages me

[prime juniper]

That I can’t help with. I’m awful with planning and I never did the steps to get to userspace

[prime juniper]

Understandable

[lean kettle]

i can feel the awful planning part atm

[grand shadow]

oops

think its get sock opt

either way I'm probably gonna at least upstream all of these changes I have here jesus christ

[grand shadow]

couldn't get the graphical version to work

nevermid it was google that didn't work in specific

epic

[grand shadow]

I mean hey its not too bad to use

the tcp ip stack doesn't immediatelly combust

the only thing I ran into was an oom but I was also running at 256 mb of ram

hm the issue was that it was default to ipv6

or not

idk I'm tired today was productive

[prime juniper]

Nice work

[neon crane]

it's kiloBYTES/s (kB/s)

dont confuse!

i mean it's about 2.4mbps which is still slow as hell but at least it's not 56kbps 😄

[neon crane]

agree

[grand shadow]

From now on I will measure everything in kilo nibbles to prevent further confusion

[grand shadow]

I wanna do something fun today and get an ssh server on astral

[grand shadow]

The sys futex wait signal interruption stuff still hasnt been upstreamed into mlibc damn its like an 8 line diff

[craggy tide]

PR?

[grand shadow]

#1068

[craggy tide]

why did the CI fail?

also can i trust you that this is correct?

[grand shadow]

Apparently building the managarm sysdeps failed because it failed to get libdrm or something

[grand shadow]

Ive been running with it on astral ever since I opened the pr, I guess you can wait until someone who wrote the code confirm its good but all it does is just retry to take the locks on an eintr and go back to sleep if not

[wild marsh]

i restarted the ci jobs

freedesktop seems back up? at least the gitlab instance loaded in my browser

as for the code i looked at it and it seemed fine

could you run a rebase to get rid of the merge commit though?

[grand shadow]

I will once Im back home

In a few hours

[wild marsh]

we could also just squash right now ig

when merging

[grand shadow]

Yeah I dont think squashing would be a problem since its just a single commit anyways

[wild marsh]

okay ci passed now

[grand shadow]

Im surprised you didnt run into this when doing mlibc on linux since FUTEX_WAIT is interruptible there

[wild marsh]

inb4 linux returns EAGAIN if futex_wait is interrupted by a signal :^)

hm no it doesn't ```
EINTR A FUTEX_WAIT or FUTEX_WAIT_BITSET operation was interrupted by a signal (see signal(7)). Before Linux 2.6.22, this error could also be returned for a spurious wakeup; since Linux 2.6.22, this no longer happens.

[grand shadow]

What kept dying because of this was fvwm iirc and fvwm does start new processes once in a while

So I think it was a sigchld interrupting it

[wild marsh]

i guess geert already merged it

without squashing

[grand shadow]

Yeah lmao

[grand shadow]

hmmmmm

I will need socketpair for sshd

yeah localsocket socketpair doesn't sound so bad to implement

[grand shadow]

damn sshd is making me implement all of the socket system calls I was too lazy to implement

oop

[robust geyser]

chroot is pretty simple

[grand shadow]

yeah I literally just copied my chdir and changed a single thing

I already had the root stuff implemented just never the chroot syscall itself

[robust geyser]

ahh nice :^)

did you try it after that change?

[grand shadow]

yeah I'm stubbing some more stuff the mlibc sysdeps now

hm

[robust geyser]

wtf

[grand shadow]

I don't handle uid/gid stuff so its probably complaining that it can change the gid because I just stubbed those

I can probably just hack something together lmao

or actually

disable that in openssh

yeah its just setting a flag in configure.ac

WHAT DOES IT WANT FROM ME I DON'T HAVE PASSWORD SUPPORT

wait there isn't even a mathewnd user

oh

[analog berry]

lol

Does this mean I can p4wn astral now

[grand shadow]

if you can connect to it because not even I can lmao

I need to figure out how to make this work

I either implement crypt(3) or stub it to return the bullshit I put in /etc/passwd

random bullshit go

[analog berry]

ssh speedrun

[grand shadow]

I mean its an mlibc panic because of a function I don't even know how to implement being stubbed

I will just make it so that like if the first input is some specific string it will return another string

and otherwise do the normal unimplemented

[wild marsh]

libxcrypt

[grand shadow]

does that overwrite mlibc's stub

[wild marsh]

yeah if you have libxcrypt you can just disable the mlibc crypt option

likewise with libiconv

[grand shadow]

ohhhh

I should do that then

[wild marsh]

most things should pick the dedicated libraries

libxcrypt should definitely work because not even glibc provides crypt on it's own nowadays

(or at least distros use libxcrypt)

[grand shadow]

well ty for the pointer

[bright kernel]

ur trying to make https work?

or is it some other internet stuff

[robust geyser]

sshd

[grand shadow]

https works fine, I have a web browser ported

[bright kernel]

ah I see, insane!

[grand shadow]

I'm just going after sshd now

[bright kernel]

oh dang

yeah getting https ported is one of my OS goals, but I'm still a long long way from it

and then other stuff it may need which just adds to the complexity

great stuff on ur OS tho!

[craggy tide]

tbh you can (and perhaps should) import most changes from the vinix repo

when it comes to ports

i could even do it for you like i do it for Gloire

i don't mind

[grand shadow]

you don't have to, I'm probably gonna have to take a day to do port maintenance sooner or later anyways

[craggy tide]

sure

just ask if you want tho

[grand shadow]

I ported firefox to astral

[robust geyser]

what a madman

[robust geyser]

great work

[wild marsh]

does X forwarding work? :^)

[grand shadow]

I should try

[robust geyser]

yeah that would be amazing

[analog berry]

That was quick

Good job

[analog berry]

Astral is a windows 11 fork confirmed

[grand shadow]

I just got libxcrypt ported and then remembered I could let ssh use empty passwords

[grand shadow]

true....

[robust geyser]

man

every day i need to do more and more to catch up

:^)

i should start doing something

[obsidian pivot]

I'm on my VMM and haven't written a single line of code for it since the day I started

but we both gotta catch up

I'll start catching up

uhhh

the 22nd

[robust geyser]

i probably wont lmao

[obsidian pivot]

my arm with a cast is itchy

but the itch

is bad

like

it needs to be itched

[grand shadow]

idk if x forwarding doesn't work with wsl2, if it doesn't work with astral or if I'm just too dumb to set it up

whatever I have openssh thats what matters :3

[robust geyser]

should work on wsl, at least it worked last time i tried it

but great work nonetheless :3

[grand shadow]

seems pretty stable on ssh too, was able to run links and stuff

I should do the ssh from one astral vm to the other

[robust geyser]

u should host a public astral vm :D

[grand shadow]

@robust geyser

[grand shadow]

that's a fun idea and I would need to add permssion stuff cuz like I know in the first 10 seconds someone will rm -rf /usr/bin/* lmao

[bright kernel]

better than windows™️

[robust geyser]

that is so god damn based

[analog berry]

you should really implement absolute ps/2 mouse extension

but nice thats really cool

[grand shadow]

I should add credentials tbh

And unstub all those functions from mlibc

[analog berry]

no no dont do that

mount your host fs into astral and expose over ssh

and host on a public ip

[errant zephyr]

perfect

[errant zephyr]

then put up a public ssh

and run printf "flag{%s}" <(head -c 12 /dev/urandom | xxd -ps) > /flag

and see how long it takes for someone to recover that

[grand shadow]

Implementing nfs just to get my host fs messed with 🙏

[grand shadow]

Honestly sounds like a fun idea

[analog berry]

qemu has fat: thing\

u can mount any directory as a fat disk

[errant zephyr]

make sure to instance qemu though

[warped plinth]

awwww that's awesome

[winged basin]

i found the easiest thing is to implement virtio-9p

[errant zephyr]

i mean

[winged basin]

and if you are prudent when you implement it, you can then implement networked 9p without changing the actual 9p driver (just the transport driver, which should be separate)

[errant zephyr]

its plan9 so

[winged basin]

qemu implements 9p2000.L which is 9p, minus 80% of the plan9 stuff, plus serialised linux vfs ops

so it's much less pain

[errant zephyr]

already found one thing that might be exploitable

syscall_hostname assumes a terminator and then performs a strcpy into a buffer that is too small

oh and

PLEASE enable SMAP and SMEP

hostname also has a kernel read which is cool

oh and write

[grand shadow]

I dont really have a read from user implemented yet lmao

Old astral had it but Im too lazy to backport that

[errant zephyr]

https://github.com/Mathewnd/Astral/blob/rewrite/kernel-src/sys/syscalls/write.c#L30 nuff said

[errant zephyr]

do it

its a major security feature

[grand shadow]

Yeah

Guess I will add it together with credentials

Since everythijg still runs as root

[errant zephyr]

https://github.com/Mathewnd/Astral/blob/rewrite/kernel-src/sys/syscalls/sigaltstack.c this is racey if you have cpu migration

[grand shadow]

I dont lmao

My scheduler works with smp but its bad

Actually wait

[errant zephyr]

ah i see

[grand shadow]

Yeah that can be vary bad oop

I should probably add a getrunningthread func or something

[analog berry]

what the fuck

this is serenity os level stuff

why cant vfs_write lock user pages until its written?

you dont need this memcpy bs

also u have an interesting code style

no underscores nor camel

itsjustonewordcase

veryhardtoreadbutwhatever

[neon crane]

lol

[wild marsh]

slower than a memcpy that just bails out on a page fault

[neon crane]

was about to say that

just bail on an invalid page fault

[analog berry]

wdym

thats not even what im talking about

im saying he straight up allocates a new kernel buffer

so it copies the entire range from user, then does DMA from that range, then frees the range

thats extremely expensive

[neon crane]

my guess is he doesn't want it to crash WITHIN vfs_write

[analog berry]

copy_from_user is for like tiny structs

not disk writes

[wild marsh]

i mean just copy_from_user into the page cache

[neon crane]

does he have such a thing

[analog berry]

if he has one then yeah

[neon crane]

but what if you pass an invalid memory region to the function

and it bails in the middle of writing

[analog berry]

u return the page back to cache

[winged basin]

an exception is thrown

[neon crane]

and then you've half-written the file already

[wild marsh]

undefined

return EFAULT and carry on

[winged basin]

half the file has been written then

[analog berry]

NT verifies the entire range and locks it before doing anything

for example

u cant half write

[winged basin]

expensive and impossible

[grand shadow]

I never said the code was good

[analog berry]

how is it impossible?

for page in pages
if page.lock() < 0: return EFAULT

[neon crane]

that's a problem isn't it

[winged basin]

it is legal to have, let's say, 1 tib of data in virtual memory, and to write that whole lot to a file

[neon crane]

you could pull that off with an MDL

[winged basin]

with only, say, 32 gib of ram

[analog berry]

sooo how does that make it impossible to lock pages

so that unmap hangs until the data is copied

[wild marsh]

you can't lock all 1TB into memory at once

[analog berry]

u lock page by page

[wild marsh]

locking implies it can't be paged out

[winged basin]

it can't fit

[analog berry]

if any fails u unlock all of them

by lock i mean the virutal address cannot be unmaped by user

it doesnt have to be present

u just say this virtual address cannot be unmapped for the duration of the operation

[winged basin]

in that case possible but requires a little work

[analog berry]

u have to have a similar mechanism for paging out stuff anyway, while in a page fault handler you shouldnt allow a different thread to unmap said page

[winged basin]

you need to inhibit files for which some mapping exists from being truncated, for example

it also breaks a bit on some networked filesystems where mmap() can be risky in general (because the file may be remotely truncated and then what should you do? provide zero-filled pages or fault?)

[analog berry]

probably doesnt really matter because this is implementation defined behavior

[winged basin]

my feeling is that if someone tries to write data part of which is virtually unmapped then it is fair to punish them by only writing the virtually mapped part of it

[neon crane]

isn't that just locking the address space

[analog berry]

fair enough but what i described above is not that difficult to implement

[winged basin]

for write() style interface anyway, for asynchronous writes not a virtual but a physical wiring of the pages is necessary

[grand shadow]

When I wrote that I didnt even have a page cache but I might do that once I have a copy from user anyways

[analog berry]

but thats all u need

if its present it will be present

[analog berry]

yeah, but you will write private data if that page is freed by the user before the write is finished

which is why we lock stuff

or you just bump the page reference count if u have struct page

[winged basin]

if it's physically wired it's not subject to reuse

[analog berry]

what does physically wired mean

[winged basin]

in keyronex this is the use v.s. reference distinction

[analog berry]

so like ref count?

[winged basin]

a page use can be e.g. file shared, anonymous, pml3, or deleted; a page is freed when its use is "deleted" at the same time as its reference count is 0

[analog berry]

i mean yeah that works

[winged basin]

and a reference count > 0 = physical wiring because the page is not subject to either freeing or reuse

[grand shadow]

Hm another thing that phases me more than the performance aspect of it does is that it needs to allocate the memory for writing all of that especially if its like a huge file which is dumb yeah

Whatever Ill deal with it later Id need user memcpy for that anyways

It was just the easiest solution at the time

[analog berry]

Well since u have no cache see my solution above

Just dma from those pages

[grand shadow]

I do have a cache?

I will just user copy into it like I already do in vfs_write

[analog berry]

oh then im even more confused

why not memcpy into the cache directly

[grand shadow]

Because I wrote that before I had the page cache

I didnt implement it with my vmm I did it much later because I had no idea how to

I did this because it was the simplest thing to do at the time

[analog berry]

I see

[grand shadow]

Anyhow for today I will probably go around the kernel doing a few todos here and there while seeing how I will be implementing the credential stuff

Maybe even the user copy stuff too

Since thats part of the todos lmao

[winged basin]

kauth

[grand shadow]

it seems like an interesting idea

I got a friend to ssh into astral all the way from poland so serverside stuff defo works over the internet too

[robust geyser]

can i try aswell

im polish too :D

[wild marsh]

POLAND MOUNTAIN 🇵🇱

[robust geyser]

its pride month

i can confidently say that im polish now

🇵🇱

[grand shadow]

I already undid the port forwarding and stuff 💔

[grand shadow]

my family is trying to get italian citizenship and the hearing is in the 24th I really hope we get it and get it this month so I can come out as italian during pride month lmao

its probably not happening because bureocracy is slooow but one can hope

[prime juniper]

Fuck now Managarm really need that server stack asap

Nice work

[craggy tide]

i'm so sorry

[neon crane]

why not portugal? they speak portuguese so you won't need to learn another language, just the different accent

[grand shadow]

I dont think we can trace anyone on our side of the family tree thats from portugal

[neon crane]

I didn't say it because of that

I said it because since you live in Brazil, you most likely speak Portuguese

[grand shadow]

I mean isnt portugal an eu country

[craggy tide]

it is

[serene parcel]

Great, I'm polish as well 😉

[spark sentinel]

and italy is too?

[grand shadow]

I have been very productive these past days yes I love productivity

I ported shadow (still not 100%, passwd works with the putpwent pr and login will work once I get to add the handling of a few wtranscode errors in more functions)

that way once I start working on stuff related to different users I can quickly test stuff out

and today I didn't do a lot other than add a commit to that pr and try to build gcc inside astral for fun (which during a mmap test in the configure the makefile calls ran into a page refcount issue which I still have to find the cause of, neat)

this weekend I really should get going again

[warped plinth]

u got this

[robust geyser]

also, if you start working on filesystem permissions you might notice jinx conveniently makes everything owned by $yourhostuseruid, which probably doesn't map cleanly into the astral vm :3

that's because of the unshare+uid/gid mapping trick it uses to gain fake root in the containers without running as root outside the container

which also sucks because shit inside the container (install scripts or whatever) can't set the uid/gid of certain files/directories, which might be needed for some things

[errant zephyr]

you can just fakeroot tar tho

[robust geyser]

but the files on the filesystem wont be owned by root

?

thats the problem

you are root inside the container

but on the outside you aren't

[errant zephyr]

fakeroot is like the entire userns meme, it remaps the uids

so tar thinks that its all root

[robust geyser]

yeah but its not actually root

so the files on the outside are owned by your user, not root

thats the problem

[errant zephyr]

but that doesnt matter, what matters is what is written in fs metadata

[robust geyser]

but linux maps uid 0 inside the container to your user's uid on the outside

so what's written in the fs metadata is actually different depending if you look at it from inside or outside the container

at least if you look at it through linux apis

the actual values written there are what you'd see outside the container

[craggy tide]

if it didn't work like that it would be a security issue

what you can do is just adjust the ownership of the files when you pack the initrd

and no it's not just a Jinx issue, it would be the same with xbstrap

unless you run it as root

[prime juniper]

Yes that is correct

[errant zephyr]

but you pack the initrd inside

and then no problem

or with fakeroot which is equivalent

[craggy tide]

^

[grand shadow]

I ran into an issue with sshd because of that, it expected the chroot dir it used to be owned by root which made me have to go and chown it every time I first ran a disk image

[craggy tide]

Gloire manually runs chown -R root:root on the sysroot as root for similar reasons

but yeah you can also pack the initrd in a fakeroot as pitust said, assuming you use an initrd

[grand shadow]

I was going to do that as I barely use the initrd to test in my machine since its faster to just boot from disk if Im doing some random testing

[craggy tide]

fair

[grand shadow]

That as in what gloire does

[craggy tide]

i mean there is no other way

other than running as root

unless you manually edit the filesystem inodes from userland to be all root

:^)

maybe guestfish can do something? idk

but guestfish is an abomination

[errant zephyr]

its also possible that whatever tool you used can do that

mksquashfs can, idk about mke2fs or tar

[warped plinth]

https://www.php.net/manual/en/class.gender.php#gender-gender.constants.italy

good for you

;)

[supple hatch]

congrats on mod staging btw mathew

also keep up the good work with astral

[grand shadow]

Ty

[grand shadow]

You can now ssh into it from the other side of the world (and do anything because theres no separation between users)

[supple hatch]

woah

thats actually rlly cool

must have taken a lot of work to get working

[grand shadow]

It was mostly having a working tcp impoementation

Also theres a web browser.now that works for some pages

[supple hatch]

woah

niceeee

[proud vine]

bro is the definition of "diy"

[neon crane]

I just hope that the OS doesn't turn out messy out the other end like (iirc) SCAL/UX was

[grand shadow]

its already messy

eithre way

boom thats one bug fixed, no longer dies because of the page refcount

lets see how far into gcc compilation this goes

waitpid deadlock memes at the end of intl configure

[obsidian pivot]

does astral say, boot on real hardware?

[grand shadow]

I tried once like last year and it got to a bash prompt but the keyboard didn't work

I should try again someday

I think I already know whats happening and it should be an easy fix

just some lock ordering oops I did in my signal code

I'm gonna leave it building gcc while I go out for lunch

it began

anyways hopefully it doesn't crash and burn when I come back

oh lol

[lean kettle]

ill take this as a challenge on managarm

[prime juniper]

Configure won’t work without at least one hack I have locally and something something fcntl.h check

[lean kettle]

tar -xvf also fails with weird EBADF errors?

[craggy tide]

that's honestly really impressive matt

[obsidian pivot]

I, for one, could only dream of getting to that point

so to speed that up

I'll stop talking for now

I'll be back

[grand shadow]

testing user copy stuff going on well lmao

I think I fucked up really badly

[warped plinth]

i'd love to run astral on my laptop

[grand shadow]

still not able to install it to disk sadly, I need fat32 so limine can be installed

[grand shadow]

works now

time to go through all of the 70 system calls and switch what they use

[grand shadow]

I'm curious, is xhci super difficult to implement?

[lean kettle]

cc @wild marsh

[wild marsh]

conceptually not really, but i've been having trouble getting stuff working on real hardware

it's just a bunch of ring buffers for submitting commands and transfers and receiving events (incl completions), and some context data structures that hold the device and endpoint states

[wild marsh]

most recent struggle, after fixing some other stuff, is usb hubs

[robust geyser]

xhci itself is pretty simple, i was considering implementing it in astral for a while but i couldnt really find any motivation to read and learn the codebase lol

i think the hard part is making it work with different controllers on real hardware

but that might just be because my scheduler is absolute dog poop, the xhci driver did seem to work fine for the most part

usb is pretty simple too, it just builds on top of xhci and it gives you so much for free basically

[wild marsh]

i mean linux defines like 46 quirks for xhci :^)

[wild marsh]

well you probably want to build a nice abstraction if you want to add ehci/uhci/ohci support

[robust geyser]

yeah, s/xhci/hcd basically

dont they work pretty similarly? i have not looked into anything else at all so i wouldnt know

[wild marsh]

well for the older ones you need to build the schedule for packets to send within a microframe yourself

and for ehci you need to manage generating split transactions for high speed usb hubs

thankfully xhci manages that on its own

[obsidian pivot]

run curl http://ascii.live/can-you-hear-me

[grand shadow]

gimme a sec I'm making execve work with copy from user

[grand shadow]

bro

😭

mfw ported curl just to get rick rolled

[obsidian pivot]

tbh I thought you already had it

[grand shadow]

I had wget

oh well now I have curl AND wget

[obsidian pivot]

HTTP request sent, awaiting response... 417 Expectation Failed
2024-06-09 18:33:06 ERROR 417: Expectation Failed.```

I get that with wget

[grand shadow]

On astral?

[obsidian pivot]

no

wsl ubuntu

I couldn't be bothered to compile astral

[grand shadow]

Oh lmao

[obsidian pivot]

last time I tried

it needed to build gcc

I ain't goin' through all that

[craggy tide]

as if that's the biggest bother

wait til you see LLVM

[grand shadow]

I really should get around to seeing if I could unassert a few wtranscode errors like someone else did in another function

it gets really close to being done but it doesn't actually get done

[prime juniper]

based

doing it without passwords and passing the user on the command line does work (with some hacks)

debug logging is elogind for session management

[grand shadow]

with the sysconf pr you're now able to actually pass the login user in the prompt

[prime juniper]

only if we implement cooked mode I think

[grand shadow]

oh lmao

and hopefully with me handling the illegal_input mbrlen error thing it should be able to log in with a password

[prime juniper]

based

I patched pam to just ignore password and return ok

but that's only because input is fucked

[grand shadow]

based tbh you can't be hacked if the system is public anyways

[prime juniper]

I want to fix it but I have negative braincells so idk how well that goes lol

[prime juniper]

[grand shadow]

hm

idk how this wchar stuff works

well thats fun

uhhh

making that return an illegal_input instead of an assert failure kinda worked

and with kinda I mean

wtf is this bash prompt

I'm still root uid though since I don't have a setuid or anything like that

maybe its that

this is really weird though

it seems like its not being treated as a login shell

[craggy tide]

i suggest reading the patches/recipes/extra files from Vinix

[grand shadow]

lemme see

[prime juniper]

Seems like it indeed

[grand shadow]

still doesn't explain the weird bash name

in xterm it works

maybe that wchar stuff gets called from bash and not from login

yeah bash is the one that calls mbrlen etc

gonna compile bash with the vinix configure arguments and see if that helps

ah well whatever

I'll look into this tmr

[grand shadow]

Im a pussy and I really dont feel like debugging bash so Ill drop this login thing for now and instead go fix some things I have on my backlog and work on fat32

[robust geyser]

i remember having a similar issue on linux but i dont exactly remember how it was fixed

does it ever segfault for you when using login?

[grand shadow]

Nope, all that happened was a wtranscode error

That I patched out

I genuinelt have no idea wtf bash wants from me, I think its thinking its own name isnt just ascii or smth

[bright kernel]

well atleast you know it works x)

[grand shadow]

I'm thinking of buying a domain and a vps or something to host the packages for astral once you can install it to disk

tho I have no idea how that stuff works since I've never looked into it too much

other than buying a domain a few years ago because I was going around a bit and wanted to ssh to my computer at home

[bright kernel]

why dont u just set up a basic https server that reads off package names on a request, and then returns it's address link where then the pac man would just download off of

[grand shadow]

the issue mostly is the hosting part of things, I could self host it but I really don't feel like exposing my home network that much nor do I have any spare machines to leave on 24/7
with that out of the way its just a matter of doing it

[bright kernel]

you could buy a simple server, the other problem would be the pricing, but for a simple server I think that it wouldn't be that expensive

[grand shadow]

which provider though

I searched in this server and one called "tornado vps" caught my attention

idrk atp I'll probably leave that to future me

[grand shadow]

@fervent fjord are you still using this and is it worth

I can barely find anything about it online

[bright kernel]

future you will remember u...

[bright kernel]

then that's not a good sign, usually something more known will be more reliable

[bright kernel]

yo @grand shadow did you need to recompile the WSL kernel for KVM to work?

[obsidian pivot]

I didn't

if that helps

[grand shadow]

nope

[bright kernel]

huh

how did yall get kvm working then

I tried adding myself to the group

and staring up the virt manager (and what it needs)

but it still didnt work

and checking with kvm-ok it said my CPU didnt support it (it ran fine on linux)

[neon crane]

yeah why does it need to do that

[obsidian pivot]

probably for the astral gcc toolchain

[bright kernel]

oh wait, u run windows 11 right?

[grand shadow]

yeah my laptop uses it

[bright kernel]

oh that's why

windows 10 doesnt support nested virtualization

dang

I might need to reinstall windows 11 or just dualboot

(I don't have much space for dualboot)

😭

[fervent fjord]

its neat yeah

[neon crane]

?

how come

im pretty sure it does

[bright kernel]

I tried enabling nestedvirtualization on .wslconfig and when WSL started it said "This machine does NOT support nested virtualization"

and running kvm-ok it said "this CPU does not support KVM run with sudo for more info.."

[winged basin]

it blamed your machine, not windows

[bright kernel]

the problem being that my machine supports it

[winged basin]

oh, very bizarre

[bright kernel]

and other people facing the same issue on gh reported it

and most answers were that windows 10 doesnt support it

apparently you could enable it using windbg

but it's not safe

[grand shadow]

I saw you were running ironclad there and that caught my interest, I will be checking them out later

[bright kernel]

yes, the problem was windows

[grand shadow]

I did it

I'm gonna see if I can download a minimal chroot with bash and dependencies

[analog berry]

What is links exactly?

[wild marsh]

a web browser

[analog berry]

Does it load youtube

[grand shadow]

don't think so

[analog berry]

What can it load

[grand shadow]

it could load github

it doesn't load a lot of things and what it does load is us quesitonable

also rip

got close

[analog berry]

Allocator bug?

[grand shadow]

hmm

double free

might be some vnode refcounting memes

well the download part of it worked lmao

averaged like between 600 and 700 kb/s

not that good but better than 30 kb/s

[bright kernel]

install fortnite next

[grand shadow]

well I added an assert ( reference) to my VOP_HOLD to see if I can catch that the first time it happens

[analog berry]

Lol

[grand shadow]

Time to get my ass back on track again today

I got a list of todos from the source and I will try to do at least half today

My friend is also making a simple landing page for astral-os.org because I know nothing about web dev

[warped plinth]

hell yeah

ngl i love the bare html

[neon crane]

@grand shadow question, why aren't you popping FS and GS here: https://github.com/Mathewnd/Astral/blob/rewrite/kernel-src/arch/x86-64/isr.asm#L54

Also perhaps you should have your ISR handlers be like:

isr_X:
    if x != 8 && (x < 10 || x > 14)
    push qword 0
    endif
    push qword X
    jmp  isr_common

isr_common:...

[grand shadow]

My gs and fs are always 0

[grand shadow]

Yeah true

It was one of those things that has been on my todo list for so long but it has never mattered enough for me to do it

Ill add it to the list of stuff I wanna do today tho

Im gonna work on a bunch of stuff I left behind before I do fat32

[neon crane]

its so easy tho

and it saves quite a bit in binary size

[neon crane]

So then why preserve them

[grand shadow]

In case something corrupts it and completeness sake I guess

[neon crane]

I guess I'd also make sure that ds and es are the kernel data segment before you enter the interrupt ISR

[grand shadow]

Oops yeah I think I forgot that

I guess it never really mattered that much because all my checks are done on the cs

[neon crane]

It's not about that

It's about using the CPU in an inconsistent state I think

Maybe it doesn't matter if DS and ES have a CPL of 3 while CS and SS, a CPL of 0

but for completeness

[grand shadow]

yeah

my system call handler also suffers from that from what I can see

[obsidian pivot]

Is Astral cursed on the inside

i.e., bugs that are pretty big, but a worked around instead of fixed

[grand shadow]

I hope not lmao

I usually try to fix bugs when I find them

well those 2 fixes (the segment thing and interrupt code size) took like less than 10 minutes

goddamn if I werent so lazy

[analog berry]

did you figure out the tar corruption?

[obsidian pivot]

and you guys say my kernel is cursed ||(it is)||

is the tar corruption

for the ramdisk?

[analog berry]

.

[obsidian pivot]

cpu0: oops

best error message ever

[grand shadow]

still haven't tried to reproduce it

I probably should soon

I am just lazy (and work has been taking up quite a bit of my energy)

[grand shadow]

dead project xd

I think I did burn out a bit from working way too much on this

[lean kettle]

You don't have to constantly work on the same project, that's fine

[grand shadow]

I should port an irc client

[warped plinth]

https://c9x.me/irc/

:^)

[grand shadow]

I ported xneko

best port ever

[robust geyser]

unironically

[warped plinth]

oh my god yes that's awesome

i genuinely look forward to the day i can use astral on my hardware

[analog berry]

like, today then

[warped plinth]

smp would be nice, but beggars can' be choosers

[analog berry]

it does have smp

[warped plinth]

well, bugless smp, i suppose

iirc there's still some problems with the current impl

[grand shadow]

you can pass nosmp to the kernel commandline in limine and it will boot without smp

the biggest smp issue rn is that my tlb shootdown is painfully slow

and some things still arent 100% smp safe and racey

[warped plinth]

oh fuck yeah, well i'd love to check it out then

[prime juniper]

Port hexchat. Port GTK2

[grand shadow]

perhaps...

[grand shadow]

fixed the certificate memes finally

[grand shadow]

I do wonder if it could post in the osdev forums, it can log in and navigate around just fine

[tawny oak]

haha wow, very nice

[prime juniper]

Never tried it from there

Works fine from WebKitGTK tho

[grand shadow]

is there anything in specific you needed kernel side for webkitgtk

well sysdep side since managarm is a microkernel

[obsidian pivot]

when smp fix?

even obos runs with SMP

except not with 128 cores

ok fixed

[grand shadow]

when I stop being lazy and fix a race that affects like 90% of the files in the kernel and make the tlb shootdown not horrible

[obsidian pivot]

that reminds me

I need tlb shootdown

[grand shadow]

funnily enough some of the stuff I had thought were smp bugs were just my scheduler doing an oopsie with the stack

that also happened out of smp but were more common with smp

[obsidian pivot]

lol

OBOS is better than Astral until SMP on Astral

[grand shadow]

soon™️

smp on astral has always been "almost there but never really quite there"

[obsidian pivot]

SMP is surprisingly the most stable thing on OBOS

[grand shadow]

it just adds thrill to the computing experience

[obsidian pivot]

guess the least stable

(difficulty: impossible)

[grand shadow]

allocator....

[obsidian pivot]

no that was rewrite 4

now it's the timer interface

[obsidian pivot]

actually

hmm

[grand shadow]

I want astral to be self hosting enough to be able to:
open osdev wiki on browser
follow barebones from compiling an i686-elf gcc cross compiler cuz why not
run qemu and the barebones and post in #osdev-misc-0 asking how do paging

[lean kettle]

if you manage all of this im donating 10€ to a (good) charity of your choice

[bright kernel]

he just has to remember the (good) part

[lean kettle]

by good i mean morally decent if that makes sense

[grand shadow]

Bet

Before I add new things I still gotta finish my todo list though

Tomorrow Im gonna go after 2 of them where a thread does nonstop yield waiting for I/O because my old event thing was bad, so Im gonna write another one that doesnt suffer from that nasty lost wakeup

Its gonna be more general than the poll system thing

So just an attach etc

[grand shadow]

More specifically when its waiting for a page to be ready in the page cache and waiting for an arp response lol

[prime juniper]

Well, considering I always lie to the programs and say that (if needed) just use Linux code paths, probably some Linux stuff if you follow that route. I know we had to do some netlink shit cuz glib. I know there is a BSD backend which uses a different interface. I know I told GTK just assume we’re Linux, which means you may or may not run into some memes there too. My advice? Just try it. If you want to be Linux like, there is a good chance you’ll need some of the following (list is in random order and not guaranteed you need anything from it, list might be incomplete): eventfd, timerfd, signalfd, inotify, netlink, DRM. Most, if not all, can probably be substituted for something else, if you claim that you aren’t Linux but more like a BSD for example. Best way to find out is to try

[prime juniper]

Does that also count if Managarm can do it?

Cuz the only thing in this list we can’t do yet is compile GCC

Qookie wrote a bootsector on Managarm and ran it in Qemu, I have had discord conversations from Managarm

[prime juniper]

Free tip, GTK and friends (especially GTK3 and 4) really like their mime database (shared-mime-info, run update-mime-database /usr/share/mime). Oh and GTK kinda wants to refuse to run as root? I patched that check, might want to borrow that patch

[lean kettle]

Sure

[grand shadow]

shadow really doesn't like it when you don't have a valid shell

[grand shadow]

seems like the weird bash name isn't in bash but rather in login, makes things easier to debug I think

I wont have to read bash code

so it could be that exec

what the fuck?

this can't be right

am I going insane or

I think this was supposed to be argv[0] = argv0?

idk

actually wait

lemme try thios

this was taken from musl so it must be right

gonna see whtats in that argv[0]

yeah no this is wrong

from musl source

fucking hell

another mlibc bug it seems

yeah

that was why bash was acting all weird

execle is broken in mlibc

[eager copper]

be the change you want to see™️

[grand shadow]

Has been merged for like 10h already smh

[robust geyser]

lapfed slow as usual

[grand shadow]

My list of things to fix is getting smaller, soon I will actually be able to progress again

[warped plinth]

woohoo!

[fleet osprey]

soon™️ or actually soon

[grand shadow]

if I stop being lazy soon otherwise soon®️

[fleet osprey]

skill issue tbh
I've been working on the same project daily for the past 2 months or more, I don't even remember anymore
legitimately not 1 day that I haven't worked on it

I don't think this is healthy

[grand shadow]

I did that too for like 3 months until I ported openssh then I kinda burnt out a bit

now I'm just slowly grinding away at a todo list of random fixes and improvements

[grand shadow]

did a thing

I should try to compile it

well the kernel

jinx Is Not Going To Work

oops yeah its missing flanterm and stuff

but hey git works

[craggy tide]

damn that's fucking crazy

not even managarm has working git iirc

[warped plinth]

oh fuck yeah, nice

[grand shadow]

damn

still another step towards self hosting, which would be pushing its own code into the internet

I don't know if pushing works and I'm too lazy to configure the credentials and stuff

I will only consider astral truly self hosting once I can compile gcc though

because what good is for that it can compile itself if it cant compile the tool to compile itself?

[prime juniper]

I worked on it years ago. Back then, netstack was the biggest issue. So no, technically we don’t, but I’m pretty sure it’s a recompile away. Minor fixes at most (at least for git clone). If I dig I can probably find screenshots of git doing stuff

Yep. Got a screenshot from the 15th of April 2021 running git log

#678608576151027734 message

Still, getting git clone working is massive for self hosting. Amazing job

[craggy tide]

ah nice

[obsidian pivot]

cool

but does it run doom

?

[grand shadow]

Quake and doom on x11

[obsidian pivot]

Based

[charred sinew]

What syscalls do you need to support X11?

[regal stone]

is astral based on a kernel?

cause if not

holy shit this is impressive

[prime mulch]

it is based on a kernel. but its a custom kernel

and yea it is

impresive

[regal stone]

which kernel

linux?

[prime mulch]

the astral kernel

[regal stone]

so it has its own

[prime mulch]

yep

[regal stone]

wow

[winged basin]

it is the traditional fashion here to make our own kernels

though maybe someone could legitimately do something on top of Se.L4

or w/e

[neon crane]

@grand shadow i suggest you consider porting sm64 to Astral

would be cool

[obsidian pivot]

port astral to the 68k

[grand shadow]

the core of it is giving it somewhere to draw like using a xf86-video-fbdev driver and a way for applications to talk between them and the server like unix sockets, so a good enough posix support w/ the socket stuff gets you a long way

[grand shadow]

what did you use for nanoshell

[grand shadow]

or you can just not even have the X server be on your os and do it thru the network like god intended

[regal stone]

they would need an n64 emulator

[wild marsh]

sm64 has been decompiled and that has been ported to various other platforms

i think it just needs sdl2 and opengl?

or something thereabout anyway

[grand shadow]

I don't have mesa and I'd need to port llvm for that and I wanna fix up some other stuff first

[fleet osprey]

can it run QEMU?

[grand shadow]

Idk never tried it

[wild marsh]

qemu should be simple enough i think

on the outside you might want sdl2 or gtk for the ui, on the inside it wants pthreads

and it does some signal trickery (sigaltstack + sigsetjmp & siglongjmp) to switch stacks for coroutines but that's also not that bad

[winged basin]

why llvm?

[fervent fjord]

llvm-pipe

[winged basin]

screw llvmpipe

use softpipe

[fervent fjord]

bit slow

but its usable

[grand shadow]

Ah I should have everything then I think

[neon crane]

The software renderer made by "fgsfdsfgs" for his DOS port

[fervent fjord]

great name

[grand shadow]

I do wonder something
eventually I will need a better init system than just running a shell script at boot and all
how easy would it be to port openrc, hm

its not something I really need atm (or for the close future anyways) but I am curious

[errant zephyr]

why not s6 or upstart?

chromeos uses upstart fwiw

[warped plinth]

runit, perhaps?

[sacred gyro]

nah, port systemd

[grand shadow]

Cuz Ive always used openrc and it really doesnt matter enough

[neon crane]

What's an init system do other than launch the shell

[grand shadow]

Make init a symlink to bash smh

[errant zephyr]

reap orphans

[neon crane]

I assume orphans that exited right?

[winged basin]

Depends how one defines it

Will discuss when home

[errant zephyr]

well, yeah

usually it also deals with service managment

because some services always orphan themselves

and so you have to be pid1 to get a notification

[prime juniper]

Pretty sure mlibc can (almost) run it

I should try that under LFS

[fleet osprey]

I wonder if it can run my "debugging init"

give me a sec to clean it up so it has a chance, and I'll drop it here

this is much more a test of how much of a generic userspace you have and less of a C library test (it's literally a script), but still

made this to get some info as to why somebody's OS wasn't booting for (non-osdev, it's Linux) project I maintain, just to gather some basic system info

cleaned it up a little for here, since I threw it together in like 60s over there

and it was meant for recording the screen with a phone camera, hence reading stuff out slowly

curious to see what they all give, since I know neofetch was being pretty funky in regards to RAM, and didn't even list a CPU on Astral last I saw it

[grand shadow]

gonna try it

actually wait

free -h doesnt work, I don't have /proc, I don't have dmesg, I don't have an /etc/os-release

[robust geyser]

i don't think /etc/os-release is necessary

but procfs is definitely something you need if you want neofetch to pick up your cpu/memory info

[grand shadow]

yeah I will have to look into it sometime

[grand shadow]

Dead project xd

Argh I need to finish fixing stuff so I can work on fat32 already

[analog berry]

Dont make the mistake of only supporting fat32

Support all types