Astral | Operating System Development | Page 11

carmine swallow Sep 6, 2024, 9:32 PM

#

it just waits on a semaphore

grand shadow Sep 6, 2024, 9:32 PM

#

You can pass an interruptible flag to a semaphore

#

And check for an error

carmine swallow Sep 6, 2024, 9:33 PM

#

does it return like EINTR or something on error?

grand shadow Sep 6, 2024, 9:35 PM

#

It returns the wakeup reason which can be SCHED_WAKEUP_REASON_INTERRUPTED or something like that

analog berry Sep 6, 2024, 9:35 PM

#

any reason why its an int and not an enum?

carmine swallow Sep 6, 2024, 9:36 PM

#

grand shadow It returns the wakeup reason which can be SCHED_WAKEUP_REASON_INTERRUPTED or som...

ah, ig I can add that (though idk what is actually the sensible thing to do there, should it just stop the whole audio stream?)

#

ig that would work

grand shadow Sep 6, 2024, 9:36 PM

#

analog berry any reason why its an `int` and not an enum?

I completely forgot enums existed in c tbh it sounds like such a c++ thing

#

I just use a define

analog berry Sep 6, 2024, 9:36 PM

#

i see

#

i use enums extensively because its impossible to tell what that int might end up storing

grand shadow Sep 6, 2024, 9:37 PM

#

carmine swallow ah, ig I can add that (though idk what is actually the sensible thing to do ther...

Idk I see you loop there I think just clean up and leave?

grand shadow Sep 6, 2024, 9:37 PM

#

analog berry i use enums extensively because its impossible to tell what that `int` might end...

Makes sense

#

if you do implement interruptible sleep uh send a new iso cuz its gonna take a little eternity to test all dsps

#

I'm still at dps1

carmine swallow Sep 6, 2024, 9:39 PM

#

I could also add the name thing so it would be a little easier to know which ones you want to test

#

eg. just printing them to the log for now

grand shadow Sep 6, 2024, 9:40 PM

#

carmine swallow I could also add the name thing so it would be a little easier to know which one...

yeah true

#

would save some time

#

also @fleet osprey merged ty

fleet osprey Sep 6, 2024, 9:42 PM

#

Np, anytime

carmine swallow Sep 6, 2024, 9:42 PM

#

grand shadow Idk I see you loop there I think just clean up and leave?

well actually ig I don't really need to do anything in there then? its just a write loop and close() waits for the data currently in the buffer to finish playing anyway

grand shadow Sep 6, 2024, 9:43 PM

#

question

carmine swallow Sep 6, 2024, 9:43 PM

#

Ill try just returning EINTR and see how that goes

grand shadow Sep 6, 2024, 9:43 PM

#

does oss ever like

#

want non blocking writes

carmine swallow Sep 6, 2024, 9:43 PM

#

it depends on the app, possibly

#

rn I don't have that implemented tho

#

nor mmap for that matter

grand shadow Sep 6, 2024, 9:43 PM

#

ah

#

well I'll leave my desktop off and wait for the name and eintr stuff

carmine swallow Sep 6, 2024, 9:45 PM

#

yeah that shouldn't take too long to implement

carmine swallow Sep 6, 2024, 10:06 PM

#

it doesn't look like I ever get SCHED_WAKEUP_REASON_INTERRUPTED tho

#

no matter how long I press ctrl c

#

also one thing which I forgot about was that sometimes its actually possible to route output to eg. mic inputs (this is from the controller that I have on my pc passed through to qemu)

flat copper Sep 6, 2024, 10:13 PM

#

analog berry i use enums extensively because its impossible to tell what that `int` might end...

and also it makes it far easier to see in debugger

carmine swallow Sep 6, 2024, 10:16 PM

#

@grand shadow here is a new iso with the names, I don't know how to make the intr stuff work

📎 astral.iso

grand shadow Sep 6, 2024, 10:16 PM

#

carmine swallow it doesn't look like I ever get SCHED_WAKEUP_REASON_INTERRUPTED tho

does it ever return from the sleep and if so what does it return

#

maybe it doesn't work with semaphores?

#

but I think waitpid uses a semaphore

grand shadow Sep 6, 2024, 10:17 PM

#

carmine swallow <@215823874615541761> here is a new iso with the names, I don't know how to make...

I'll try it out

carmine swallow Sep 6, 2024, 10:18 PM

#

grand shadow does it ever return from the sleep and if so what does it return

it does return after the sched_yield() and it always returns zero

grand shadow Sep 6, 2024, 10:19 PM

#

maybe my signals are borked which is weird because it works everywhere else

#

lemme flash it to my usb and try it out

carmine swallow Sep 6, 2024, 10:21 PM

#

now I realized that a shorter file could have been a good idea lol

#

then it would return earlier

#

I mean ig you can find some random text file and cat that to there, though idk what they sound like

grand shadow Sep 6, 2024, 10:22 PM

#

true

#

@carmine swallow I have a headphone connected to the back so I assume its one of the rear speakers?

#

Or rear line out?

carmine swallow Sep 6, 2024, 10:25 PM

#

could be line out yeah

#

also don't do it with text files, they don't seem to work very well

#

it just gets stuck with no sound at all

grand shadow Sep 6, 2024, 10:26 PM

#

my phone ran out of battery shit damn

#

I'll try line out

#

I also have 2 hda controllers it seems like

carmine swallow Sep 6, 2024, 10:26 PM

#

could be for like gpu or whatever

grand shadow Sep 6, 2024, 10:27 PM

#

JESUS CHRIST

#

it outputs sounds to my mic

#

but its all broken

carmine swallow Sep 6, 2024, 10:27 PM

#

lol

grand shadow Sep 6, 2024, 10:27 PM

#

not mic

#

headphone

analog berry Sep 6, 2024, 10:27 PM

#

Sound to mic?

#

Oh

grand shadow Sep 6, 2024, 10:27 PM

#

I actually jumped

carmine swallow Sep 6, 2024, 10:27 PM

#

what does it sound like

grand shadow Sep 6, 2024, 10:27 PM

#

noise

#

really loud noise

obsidian pivot Sep 6, 2024, 10:27 PM

#

https://tenor.com/view/xqc-linux-aplay-urandom-random-gif-26003711

Tenor

grand shadow Sep 6, 2024, 10:27 PM

#

random noise

#

on one hand it does play sound
on the other I don't think this is what should come out of it

carmine swallow Sep 6, 2024, 10:29 PM

#

did you cat the wav into it

flat copper Sep 6, 2024, 10:29 PM

#

wrong format likely

grand shadow Sep 6, 2024, 10:29 PM

#

carmine swallow did you cat the wav into it

I did

#

cat whatevername.wav > /dev/dsp5

flat copper Sep 6, 2024, 10:29 PM

#

would be quite cursed if it didnt support 48khz 16bit

obsidian pivot Sep 6, 2024, 10:29 PM

#

wait is .wav just raw audio data???

carmine swallow Sep 6, 2024, 10:30 PM

#

pretty much, though there are some headers at the start that cause some audible clicks if you play them as the data

obsidian pivot Sep 6, 2024, 10:30 PM

#

ah

carmine swallow Sep 6, 2024, 10:35 PM

#

grand shadow on one hand it does play sound on the other I don't think this is what should co...

yeah that's weird, can you do cat /proc/asound/card*/codec* > dump and send that?

grand shadow Sep 6, 2024, 10:35 PM

#

carmine swallow yeah that's weird, can you do `cat /proc/asound/card*/codec* > dump` and send th...

I don't have linux on that card

#

not card

#

computer

#

.

#

I am tired

fleet osprey Sep 6, 2024, 10:36 PM

#

I noticed when you said it output sound to your mic 😅

carmine swallow Sep 6, 2024, 10:37 PM

#

grand shadow I don't have linux on that card

ah

grand shadow Sep 6, 2024, 10:40 PM

#

I think I should've output it into another dsp maybe?

#

idk

#

windows says my headphones are on a "speaker" and not on the "digital out"

#

but the again its windows

carmine swallow Sep 6, 2024, 10:41 PM

#

yeah ig its possible

#

also I should probably do something about the default volume

grand shadow Sep 6, 2024, 10:42 PM

#

its on the green thing on the back

#

I think I should connect it to the one on the front

fleet osprey Sep 6, 2024, 10:45 PM

#

yeah, if I'm not mistaken:
- green - speaker / line out
- blue - line in
- pink - mic in

#

green should be good for heaphones, but iirc will be horrifyingly loud by default, unless set by software

grand shadow Sep 6, 2024, 10:47 PM

#

yeah front headphone out still gives out the damn fucked up noise

carmine swallow Sep 6, 2024, 10:57 PM

#

I have no idea why that would be, unless its like playing some garbage that the allocated pages would initially have (assuming your page alloc doesn't zero them, I am not sure if it does)

grand shadow Sep 6, 2024, 10:58 PM

#

the page alloc doesnt

carmine swallow Sep 6, 2024, 10:58 PM

#

and as for why that would happen no idea, it does have buffer underrun debug messages but they are only printed to e9

analog berry Sep 6, 2024, 10:58 PM

#

carmine swallow I have no idea why that would be, unless its like playing some garbage that the ...

Does catting work on your laptop?

carmine swallow Sep 6, 2024, 10:59 PM

#

at least on the passed through desktop card it does, Ill test it on the laptop too

grand shadow Sep 6, 2024, 10:59 PM

#

you can have it cat automatically on /etc/rc if the keyboard doesnt work

analog berry Sep 6, 2024, 11:05 PM

#

carmine swallow at least on the passed through desktop card it does, Ill test it on the laptop t...

Btw how do you pass it through? Do u have a script for that?

carmine swallow Sep 6, 2024, 11:07 PM

#

yes I do, I bind all the devices that are in the same iommu group as the card to vfio and then use -device vfio-pci,host=00:1f.3

#

but the script isn't that portable because the devices/ids vary between systems

analog berry Sep 6, 2024, 11:08 PM

#

I see

#

How do u make it so its not acquired by host driver

carmine swallow Sep 6, 2024, 11:09 PM

#

well originally it is but when you manually do ```bash
echo "0000:00:1f.3" > /sys/bus/pci/devices/0000:00:1f.3/driver/unbind

analog berry Sep 6, 2024, 11:09 PM

#

Oh I see

carmine swallow Sep 6, 2024, 11:20 PM

#

carmine swallow at least on the passed through desktop card it does, Ill test it on the laptop t...

looks like it doesn't for some reason which is weird

#

Ill have to take a look at the widget dump for it again, maybe the first path (and the only one that I use in the astral driver) is some weird one

analog berry Sep 6, 2024, 11:21 PM

#

What is that tigerlake thing?

#

Like some hda extensions?

carmine swallow Sep 6, 2024, 11:24 PM

#

analog berry Like some hda extensions?

basically, it can do some post-processing and something related to voice recognition (ig it has some kind of more advanced processor)

#

though I think there are some limitations with the hda on that to eg. the sample rates it can support without uploading firmware to the card

analog berry Sep 6, 2024, 11:25 PM

#

So it is hda but with extensions?

carmine swallow Sep 6, 2024, 11:25 PM

#

yes

analog berry Sep 6, 2024, 11:25 PM

#

And they're fully backwards compatible?

carmine swallow Sep 6, 2024, 11:26 PM

#

at least as far as I have tested yes, though like I said afaik it needs firmware upload for some sample rates and formats to work

analog berry Sep 6, 2024, 11:26 PM

#

Thats interesting

#

Didn't know that was a thing

grand shadow Sep 6, 2024, 11:43 PM

#

carmine swallow looks like it doesn't for some reason which is weird

did it have the same issue as me?

carmine swallow Sep 6, 2024, 11:44 PM

#

I just didn't get any sound at all out of any of the dsp nodes which is weird considering that I have def got sound both from its internal speakers and headphone jack when playing sound to all possible paths

#

so yeah ig there are multiple issues to fix with it now

grand shadow Sep 6, 2024, 11:50 PM

#

damn

#

hopefully theyre both symptoms of the same issue

fleet osprey Sep 7, 2024, 2:59 AM

#

oh yeah I still need to try the new ISO on my Optiplex

#

I forgot about that

#

yeah I think mint was right the first time about it being an issue with it being an ISO
even with the latest limine (8.0.9), the ISO that's generated, directly flashed to a USB stick, does not boot on BIOS

#

welp, time to hack in a make img to generate a bootable USB/HDD image

fleet osprey Sep 7, 2024, 3:26 AM

#

progress has been made but my build env is kinda screwed somehow (git weirdness?), so doing a clean rebuild

#

gonna head out for a bit and hope it's done soon™️, then go try out the image

#

most of the image gen borrowed from the limine-c-template makefile, adjusted to be in a script, similar to your gendisk.sh, and called via the makefile

#

or ofc can be called directly

fleet osprey Sep 7, 2024, 3:44 AM

#

it works

#

ok that's actually kinda a lie

#

I've made the plumbing to
- make it gen the image
- be able to run the image, irrespective of the ISO

#

now I just need to fix it (doesn't seem to copy files "correctly", limine can't load the initrd

#

but that's a problem for future me (i.e. in 1 hour, after I take a shower)

#

but this is still huge

#

nvm, just realized my mistake and fixed it in 5s

#

real hardware test coming after I'm back though, USBs are slow

craggy tide Sep 7, 2024, 4:09 AM

#

awesome!

fleet osprey Sep 7, 2024, 4:24 AM

#

hmmm, the optiplex does not like this USB for some reason

#

it at least.... tries to do something with it now, but it just ends up resetting back to the BIOS

#

I'm not entirely sure if this is a case of
- it's just too old (doubt, it's a Core 2 Duo)
- old USB stick crapping itself
- Limine bug
- something else?

#

I've booted Limine on this PC before, but not recently

#

going all the way back to Limine 5.x in fact

#

but at least I know it can't be an image problem, since it boots fine in QEMU

#

I can probably rule the USB stick out too just by swapping the image file in QEMU to the bdev of my USB stick, hang on

#

hmmm, it actually doesn't work there, interesting

#

rufus probably botched it then, I'm just gonna directly dd it from my laptop

#

🎉

#

Rip no PS2
ACPI works tho

#

@grand shadow so this is another machine that gives that error, despite the fact that I know it has PS2 emulation
it may not be good PS2 emulation, but it does exist and is enabled
also note that after the kernel started (and only after the kernel started, worked fine in the BIOS and Limine), the USB keyboard actually turned off
since I have fancy RGB on it, I can actually see that it's just off now, lol

#

also I'll chuck the bootable USB/HDD image creation code I slapped together to get this thing booting at all, in a PR, should be useful

fleet osprey Sep 7, 2024, 5:13 AM

#

#1061407633745125397 message
also, just tried these to no avail on the optiplex (verrified they work in QEMU)
one sec, will try on the thinkpad

#

Here's what it does repeatedly mashing (more than 4) keys

#

welp, that's enough debugging and programming for me for today, gonna call it

craggy tide Sep 7, 2024, 5:23 AM

#

so what is the report wrt Limine working or not on some hardware (with up to date version)

fleet osprey Sep 7, 2024, 5:28 AM

#

- does not work if Rufus is used to flash an ISO to USB
- does not work if an ISO is dd'd to the USB, v7 or v8.
- does not work but differently, if Rufus is used to flash an HDD image to USB
- partially works if a v7 ISO is booted with Ventoy - breaks Limine loading the initrd?
- maybe works if a v8 ISO is booted with Ventoy (untested)
- maybe works with v7 if an HDD image is dd'd to USB (untested)
- does work with the latest version if an HDD version is dd'd

#

all are for BIOS
UEFI mostly just works no matter what

craggy tide Sep 7, 2024, 5:31 AM

#

is that specific to that one machine or like, in general?

fleet osprey Sep 7, 2024, 5:32 AM

#

in general on all of my real hw I've tested

#

not sure if any of those would still apply in a VM

#

would be interesting to test, but not today

#

I wish i still had that Dell Latitude 5491, it was the one machine I was able to get to boot to actually boot the ISO dd'd to USB over BIOS

#

haven't had luck with any other machine, and I sold that one a bit ago

craggy tide Sep 7, 2024, 5:35 AM

#

it would be interesting to know as well if you are generating the iso/hdd file using the C template's recipe or some other potentially broken system

fleet osprey Sep 7, 2024, 5:36 AM

#

I also haven't put the C template's ISO through the same tests

#

that would also be another good thing to try tomorrow

craggy tide Sep 7, 2024, 5:36 AM

#

yeah

#

like those things you listed are a surprising amount of failures for what it should be

#

certainly the BIOS port of Limine has been somewhat neglected recently, but i doubt it was neglected to the point that it is half broken consistently on most systems

#

that would be pretty bad

#

unless it's an issue withh isohybrid and xorriso commands

fleet osprey Sep 7, 2024, 5:37 AM

#

nah it's not
the issue is primarily with the ISO

#

I've had nonstop issues with that

#

the HDD image seems mostly fine

#

the only HDD image failure I encountered is Rufus, and I bet is trying to inject it's own MBR there, and it's just exploding

#

given that it even fails on QEMU

craggy tide Sep 7, 2024, 5:40 AM

#

rufus in non-dd mode is always annoying

#

not just rufus, but pretty much all these "boot image burners" or whatever that actually try to inject their own changes into the mix

#

instead of just copying the stream of bytes as-is

fleet osprey Sep 7, 2024, 5:40 AM

#

lovely

#

well yeah, I'm positive that Rufus is the issue, the exact cause of why it's the issue is unknown to me, and I don't have the energy to go investigate it

#

I just know that it works perfectly in every test case I've thrown at it so far if I just dd the HDD image across, and dies dramatically (it might actually be triple faulting, it fully halts QEMU) if I use Rufus

craggy tide Sep 7, 2024, 5:42 AM

#

yeah alright

#

i'll be waiting for tomorrow to run the tests off of the vanilla C template, and in the meantime i'll go double check what xorriso commands grub2-mkrescue uses, because isohybrids are so damn annoying to deal with - or more in general, ISO is really annoying to deal with

fleet osprey Sep 7, 2024, 5:44 AM

#

fair
I've had nothing but issues with USB booting ISOs on BIOS, it seems to just suck in general, regardless of what is actually booting (Limine, GRUB, ISOLinux, hell even Windows Boot Manager has crapped out on me before)

craggy tide Sep 7, 2024, 5:44 AM

#

it's really fragile

#

it is tempting to just say "screw it" and not support ISO altogether anymore

#

really reminds me of the Puppy Linux guy's post bashing the ISO format (for good reasons)

#

no one burns optical media anymore

#

https://easyos.org/about/why-the-iso-format-has-to-die.html

fleet osprey Sep 7, 2024, 5:46 AM

#

although to be fair to modern devices, UEFI has pretty much "just worked" in every case, ISO or HDD

#

this seems to be all specific to booting ISOs, from non-optical media, specifically on legacy BIOS

craggy tide Sep 7, 2024, 5:46 AM

#

i mean yeah UEFI is much less finnicky

#

that said ISO failed to work on the snapdragon laptop a friend of mine tried this stuff on

#

while HDD image worked fine

fleet osprey Sep 7, 2024, 5:48 AM

#

interesting
most stuff is just EDK2, which should support it no problem
the manufacturer must've either been using a REALLY old version, or somehow, intentionally or not, broken whatever code is responsible for loading in that manner

#

would love to know how that happened, but we'll probably never know

craggy tide Sep 7, 2024, 5:49 AM

#

yeah, i am not sure about the details because i do not have such laptop myself

#

i should probably get one to be able to test an arm port of a future project of mine

#

but eh, it's expensive

fleet osprey Sep 7, 2024, 5:50 AM

#

a lot of old (and even some new) phones (unofficially) can run UEFI as well

#

you can go pick up a Google Pixel 3a or the like for cheap, and have a tiny ARM device with a built in screen and storage that can run EDK2

#

anyways, I really need to actually go to sleep now lol, I've put it off for long enough, and if I sit here for too much longer I'll wake up in my chair in 8 hours :P

grand shadow Sep 7, 2024, 12:56 PM

#

fleet osprey Here's what it does repeatedly mashing (more than 4) keys

the fuck

#

so it gets some interrupts but not all?

supple hatch Sep 7, 2024, 1:17 PM

#

grand shadow the fuck

hello im the keyboard and i have a very important message

obsidian pivot Sep 7, 2024, 1:19 PM

#

E

grand shadow Sep 7, 2024, 2:59 PM

#

wonder what I will do today

#

@carmine swallow @robust geyser @analog berry anything yall need implemented/ported?

#

other than real hardware ps2 I cant even begin to debug that

carmine swallow Sep 7, 2024, 3:08 PM

#

well the oneshot event would be nice, though I think it could just be a variant of the signal method on a semaphore?

#

basically it would only increment the count if the count is zero to begin with so then multiple calls to it don't increment the count more

grand shadow Sep 7, 2024, 3:10 PM

#

would my event.h stuff suffice?
its interruptible and actually returns EINTR (as its based on my poll stuff)

#

though it wakes up all threads waiting on the event

#

instead of the semaphore 1

#

even has a timeout if you need that

carmine swallow Sep 7, 2024, 3:14 PM

#

grand shadow would my event.h stuff suffice? its interruptible and actually returns EINTR (as...

ah there is event already KEKW I didn't see that, it probably would assuming it basically does what I said (so triggering it multiple times would only allow either the currently waiting threads to wake up or if there are no waiting threads then the next thread that tries to wait for it)

grand shadow Sep 7, 2024, 3:15 PM

#

oh it doesnt do the next thread part

#

hm

#

I suppose I could add a semaphore_signal type of thing that can have a limit passed to it?

carmine swallow Sep 7, 2024, 3:19 PM

#

yeah that would probably work, the reason why it would be good for it to allow the next thread to proceed is that its triggered within the hda irq handler if the ringbuffer space is getting low so then even if there are no threads immediately waiting for it the next thread can proceed to write to there (though now that I think about it it should do that anyway because there is a check for how much space there is within the buffer)

#

so idk if its really needed

#

also btw the code that I have so far is in https://github.com/Qwinci/Astral if you are interested

grand shadow Sep 7, 2024, 3:24 PM

#

yeah seems like it wouldn't be needed as you already check for the available space anyways

#

though you'd probably have to make sure the semaphore is only signaled if the buffer changed from enough to not enough?

#

idk how it works but you might get what I mean

analog berry Sep 7, 2024, 3:26 PM

#

grand shadow <@379250891838193667> <@796462670995390524> <@353593948523986945> anything yall ...

Hmm no idea

#

I guess a port of wine would be nice

grand shadow Sep 7, 2024, 3:27 PM

#

grand shadow though you'd probably have to make sure the semaphore is only signaled if the bu...

seems like you do

#

        size_t avail;
        while (true) {
            bool irqstate = spinlock_acquireirqclear(&strm->bufferlock);
            avail = RINGBUFFER_DATACOUNT(&strm->buffer);
            spinlock_releaseirqrestore(&strm->bufferlock, irqstate);
            if (avail <= path->oss.fragmentsize) {
                break;
            }
            else {
                int err = semaphore_wait(&strm->buffersem, true);
                if (err == SCHED_WAKEUP_REASON_INTERRUPTED) {
                    return EINTR;
                }
            }
        }

in hdaqueue
if two threads are running here at the same time it looks like they can both leave the loop even though there wouldn't be space for both of them

grand shadow Sep 7, 2024, 3:31 PM

#

analog berry I guess a port of wine would be nice

I'll look into that

charred sinew Sep 7, 2024, 3:32 PM

#

add --with-insults to the ./configure of sudo build to get insulted when getting the password wrong

analog berry Sep 7, 2024, 3:40 PM

#

grand shadow I'll look into that

Nice

#

Iirc its easy for 64 bit, for 32 u need ldt syscalls

carmine swallow Sep 7, 2024, 3:46 PM

#

yeah for 64-bit you basically only need the syscall to set gsbase (and then signals and whatever else)

grand shadow Sep 7, 2024, 3:46 PM

#

seems like I would need the better sigaction

#

that returns siginfo and a ucontext

carmine swallow Sep 7, 2024, 3:46 PM

#

hopefully it runs better on astral than on managarm

analog berry Sep 7, 2024, 3:46 PM

#

grand shadow that returns siginfo and a ucontext

Thats useful to have anyway

carmine swallow Sep 7, 2024, 3:46 PM

#

ah yeah it needs ucontext

analog berry Sep 7, 2024, 3:47 PM

#

carmine swallow hopefully it runs better on astral than on managarm

I mean it will by default because no ipc

obsidian pivot Sep 7, 2024, 3:47 PM

#

common microkernel L

analog berry Sep 7, 2024, 3:50 PM

#

Even ls on astral feels an order of magnitude faster

carmine swallow Sep 7, 2024, 3:50 PM

#

also another fun thing would be openjdk

analog berry Sep 7, 2024, 3:50 PM

#

carmine swallow also another fun thing would be openjdk

Oh yeah

#

Thats probably quite a bit harder than wine

carmine swallow Sep 7, 2024, 3:50 PM

#

its not that hard either

#

at least in terms of patches

analog berry Sep 7, 2024, 3:50 PM

#

Can I finally play Minecraft on astral

carmine swallow Sep 7, 2024, 3:50 PM

#

in terms of stuff it uses idk, at least it works on managarm

analog berry Sep 7, 2024, 3:51 PM

#

Well u still cant play Minecraft there

#

Only the server

carmine swallow Sep 7, 2024, 3:51 PM

#

and for the server it would need the proper tcp server stuff that I didn't figure out

grand shadow Sep 7, 2024, 3:51 PM

#

wouldnt for the minecraft launcher you need a web browser anyways

carmine swallow Sep 7, 2024, 3:51 PM

#

you can launch minecraft without the launcher

grand shadow Sep 7, 2024, 3:51 PM

#

carmine swallow and for the server it would need the proper tcp server stuff that I didn't figur...

astral has serverside tcp meme

carmine swallow Sep 7, 2024, 3:51 PM

#

nice

grand shadow Sep 7, 2024, 3:52 PM

#

(tcp tx is not really optimised tho)

obsidian pivot Sep 7, 2024, 3:52 PM

#

just use wine

#

to play mc on openjdk

analog berry Sep 7, 2024, 3:52 PM

#

grand shadow wouldnt for the minecraft launcher you need a web browser anyways

Nah

grand shadow Sep 7, 2024, 3:52 PM

#

carmine swallow you can launch minecraft without the launcher

how does one even do that

obsidian pivot Sep 7, 2024, 3:52 PM

#

java -jar

grand shadow Sep 7, 2024, 3:52 PM

#

yeah but what arguments

carmine swallow Sep 7, 2024, 3:52 PM

#

yeah that and then a lot of arguments

obsidian pivot Sep 7, 2024, 3:52 PM

#

java -jar <jar file> -Xmx2G -XX:+UnlockExperimentalVMOptions -XX:+UseG1GC -XX:G1NewSizePercent=20 -XX:G1ReservePercent=20 -XX:MaxGCPauseMillis=50 -XX:G1HeapRegionSize=32M

carmine swallow Sep 7, 2024, 3:53 PM

#

they are listed in the version json (if you want auth stuff)

analog berry Sep 7, 2024, 3:53 PM

#

Even just single player would be huge

carmine swallow Sep 7, 2024, 3:54 PM

#

carmine swallow at least in terms of patches

obsidian pivot Sep 7, 2024, 3:54 PM

#

port mc for windows 10 meme

#

just run that on wine

#

done

#

(doesn't actually work)

grand shadow Sep 7, 2024, 3:58 PM

#

I have zero willpower to touch the signal code but also openjdk and wine need siginfo and ucontext and it'd be very funny.....

analog berry Sep 7, 2024, 3:58 PM

#

Do it

grand shadow Sep 7, 2024, 3:58 PM

#

hmm

#

siginfo + ucontext -> possible qemu port -> wine -> openjdk

carmine swallow Sep 7, 2024, 4:00 PM

#

tbh openjdk's template interpreter might work without ucontext, I am not completely sure tho (and even if it does work its a lot slower afaik)

#

because its an interpreter like the name says

wild marsh Sep 7, 2024, 4:00 PM

#

grand shadow siginfo + ucontext -> possible qemu port -> wine -> openjdk

qemu needs pthread

grand shadow Sep 7, 2024, 4:00 PM

#

wild marsh qemu needs pthread

I have thread stuff since xorg needed it anyways

wild marsh Sep 7, 2024, 4:00 PM

#

ah

grand shadow Sep 7, 2024, 4:00 PM

#

unless theres some more things it needs from the kernel?

wild marsh Sep 7, 2024, 4:01 PM

#

it needs some signal stuff (sigsuspend, sigaltstack)

grand shadow Sep 7, 2024, 4:01 PM

#

hmm

#

I dont have sigsuspend but I do have sigaltstack

#

sigsuspend should be fine

wild marsh Sep 7, 2024, 4:02 PM

#

qemu uses sigaltstack and siglongjmp to spawn a coroutine on a separate stack

obsidian pivot Sep 7, 2024, 4:02 PM

#

is siglongjmp in the kernel?

#

or is it a libc thing

wild marsh Sep 7, 2024, 4:03 PM

#

libc

#

it's a longjmp that also restores the signal mask

obsidian pivot Sep 7, 2024, 4:03 PM

#

I see

wild marsh Sep 7, 2024, 4:04 PM

#

wild marsh qemu uses sigaltstack and siglongjmp to spawn a coroutine on a separate stack

https://github.com/qemu/qemu/blob/master/util/coroutine-sigaltstack.c

#

it sigsetjmps in the signal handler that's on an alternative stack, then returns, then siglongjmps back into it from outside the handler

grand shadow Sep 7, 2024, 4:07 PM

#

seems like qemu only uses sig info in userspace emulation so I should be fine just implementing sigsuspend and trying qemu?

#

sigsuspend -> qemu -> siginfo + ucontext -> wine -> openjdk

wild marsh Sep 7, 2024, 4:08 PM

#

yeah that should work i think

grand shadow Sep 7, 2024, 4:08 PM

#

sigsuspend is a stupid easy system call anyways

#

its an atomic sigprocmask with a pause without timeout

wild marsh Sep 7, 2024, 4:09 PM

#

iirc ucontext+siginfo was properly implemented in managarm after we had qemu portd

#

oh you'll also need sigwait

#

or signalfd

#

https://github.com/qemu/qemu/blob/master/util/compatfd.c

carmine swallow Sep 7, 2024, 4:10 PM

#

wild marsh iirc ucontext+siginfo was properly implemented in managarm after we had qemu por...

yeah ucontext was just passed as null before I did the wine stuff iirc

wild marsh Sep 7, 2024, 4:12 PM

#

wild marsh oh you'll also need `sigwait`

for mlibc you'll need sigtimedwait since sigwait is implemented using sigwaitinfo and that's implemented using sigtimedwait

grand shadow Sep 7, 2024, 4:14 PM

#

does it need thread signaling stuff?

#

like tgkill or whatever the call was

wild marsh Sep 7, 2024, 4:14 PM

#

don't think so

#

it's just using signalfd or it's own emulation to tie signals into the event loop ig

obsidian pivot Sep 7, 2024, 4:15 PM

#

doesn't managarm just make tgkill just call kill

#

in its sysdep

wild marsh Sep 7, 2024, 4:15 PM

#

obsidian pivot doesn't managarm just make tgkill just call kill

yes because we don't have thread groups

#

so each thread has it's own pid

winged basin Sep 7, 2024, 4:16 PM

#

signalfd/epoll interaction frightens me

#

they say when you read a signalfd, it always reads the current thread's signals

fleet osprey Sep 7, 2024, 5:50 PM

#

grand shadow so it gets some interrupts but not all?

nope
It never actually gets any real keys it seems
I kept mashing for like, 5m after that

#

and nothing showed up

#

those (2x press and release, of the same key, if I remember my scancodes correctly?) appeared within the first few seonds of boot

#

when I had barely pressed any

#

then nothing at all after

grand shadow Sep 7, 2024, 7:04 PM

#

conclusion: ps2 is haunted

#

guess I'll just make enough NIC drivers so that you'll only use astral through ssh on real hardware meme

fleet osprey Sep 7, 2024, 7:09 PM

#

lol

#

I may try to help unfuck PS2
@analog berry did you have any luck w/ your attempt?

analog berry Sep 7, 2024, 7:15 PM

#

Nah I was working on uACPI so go ahead if you have the time

fleet osprey Sep 7, 2024, 7:15 PM

#

alright

fleet osprey Sep 7, 2024, 7:31 PM

#

alright so here's my plan to try to get the best of both worlds

#

PS2 Init Process
- try to test if it works by default
- yes? finish setup and hope it works (similar to my driver that works)
- no? go through the rest of the init process that normally goes on in the current driver

analog berry Sep 7, 2024, 7:39 PM

#

Why, just remove the old one, its bad

fleet osprey Sep 7, 2024, 7:40 PM

#

because it works fine for Mathewnd's machines
would rather not break that

#

because I have no idea if my old setup of "just pray it's good by default", that seems to be the only thing that works on my machines (maybe it's a Dell thing?), will work on his

analog berry Sep 7, 2024, 7:41 PM

#

Literally everything works on his machines tbh

grand shadow Sep 7, 2024, 7:41 PM

#

Not hda meme

#

Well it works 1/2

#

And yeah because theyre the only places I have to test so ofc theyll work 😭

#

The other old amd64 machine I have is broken

carmine swallow Sep 7, 2024, 7:43 PM

#

I didn't yet figure out the hda issue with my laptop either, it should work ™️ and I also tried to test astral on my desktop but it hang in the hda code when enumerating codecs which isn't good

analog berry Sep 7, 2024, 7:44 PM

#

Damn

fleet osprey Sep 7, 2024, 7:44 PM

#

I think this perfectly describes how awful this will be

grand shadow Sep 7, 2024, 7:53 PM

#

carmine swallow I didn't yet figure out the hda issue with my laptop either, it should work ™️ a...

I never tested the hda code on my laptop I should do it rn

carmine swallow Sep 7, 2024, 7:56 PM

#

yeah it would be nice, though based on all the current issues I kinda doubt that it will work

grand shadow Sep 7, 2024, 7:56 PM

#

Out the headphones its the same weird noise

carmine swallow Sep 7, 2024, 7:56 PM

#

likely its the same issue then

grand shadow Sep 7, 2024, 7:57 PM

#

Ou the main speaker too

#

A different type of noise but still just loud noise too

robust geyser Sep 7, 2024, 7:58 PM

#

wish i could try hda 😔

carmine swallow Sep 7, 2024, 7:58 PM

#

the bad thing is that I have no idea what can these be caused by, except for the issue that I have on my desktop I suspect that the irqs are not triggering at all so then it gets stuck waiting for some command

#

I mean I already found one bug related to selecting the connection between widgets but that didn't really affect anything (and idk how common that code path even is, it might be pretty rare)

#

and all that it could cause is that there would not be any sound at all

grand shadow Sep 7, 2024, 8:00 PM

#

Through the headphones its a loud noise that sounds like something made to kill your ears but thru the main speaker it sounds just like static

fleet osprey Sep 7, 2024, 8:01 PM

#

my hacky PS2 init method works fine on QEMU :D
gonna try it out on my Optiplex and see if it's gonna like me now

oh also made a minor fix the Makefile, you may want to add -f to the rm command in the initrd target in the Makefile. It blows up if a previous build was interrupted for me

carmine swallow Sep 7, 2024, 8:03 PM

#

grand shadow Through the headphones its a loud noise that sounds like something made to kill ...

do you have linux on that machine? or is it windows too

#

though I kinda doubt that I could get anything out of the widgets anyway tbh

grand shadow Sep 7, 2024, 8:04 PM

#

carmine swallow do you have linux on that machine? or is it windows too

Both are windows

#

Id have linux in my laptop but my mom uses it from time to time so cant

#

though

#

I do think theres a linux install in one of the ssds in my main machine

#

actually lemme see this I might've forgotten that I had linux installed at all

#

I think windows killed it lol

#

I could boot up a live cd or something if itd be helpful @carmine swallow

#

Wait bruh this is fucking cursed lol

#

Windows boot manager is in the hdd but the install in the ssd

fleet osprey Sep 7, 2024, 8:13 PM

#

fleet osprey my hacky PS2 init method works fine on QEMU :D gonna try it out on my Optiplex a...

alright, testing now
I swear if this works... man PS/2 is weird, lol

carmine swallow Sep 7, 2024, 8:14 PM

#

grand shadow I could boot up a live cd or something if itd be helpful <@379250891838193667>

there might also be a windows tool, Ill test if it works first

grand shadow Sep 7, 2024, 8:14 PM

#

Yeah no for some reason only the hdd windows boot manager works and everywehre else drops into a grub rescue shell

grand shadow Sep 7, 2024, 8:14 PM

#

carmine swallow there might also be a windows tool, Ill test if it works first

Alrovht

#

Aleight

#

Bleh

#

Alright

sacred gyro Sep 7, 2024, 8:18 PM

#

so I tried booting Astral on another one of my machines

#

and it appears as if as soon as Limine executes the kernel, it triplefaults

grand shadow Sep 7, 2024, 8:18 PM

#

wtf

fleet osprey Sep 7, 2024, 8:22 PM

#

wtf.... so on my optiplex, either the self test, or whatever ps2kbd_init() does, is causing the BIOS to just completely give up on trying to do USB -> PS2 emulation

grand shadow Sep 7, 2024, 8:23 PM

#

all ps2kbd init does is

#

    isr_t *isr = interrupt_allocate(kbdisr, arch_apic_eoi, IPL_KEYBOARD);
    __assert(isr);
    arch_ioapic_setirq(KEYBOARDIRQ, isr->id & 0xff, _cpu()->id, false);
    kb = keyboard_new();
    __assert(kb);
    printf("ps2kbd: irq enabled with vector %u\n", isr->id & 0xff);

fleet osprey Sep 7, 2024, 8:23 PM

#

¯_(ツ)_/¯

#

well, one of the things in my shitty codepath is causing it to just entirely give up, and drop all power to the USB port

// XXX this is a terrible idea that really shouldn't need to be done,
// but a lot of stuff, especially USB-based PS/2 emulation via the BIOS, craps out
// if you try to do the full init process!  Try to selftest immediately.
// If it works, immediately finalize and get out of here.
// If it doesn't, actually try to set it up.

printf("ps2kbd: trying method 1\n");
ps2_write_command(PS2_CTLR_CMD_READCFG);
uint8_t control = ps2_read_data();
int workingflag = 0;
bool dualport = control & PS2_CONFIG_CLOCKP2;


ps2_write_command(PS2_CTLR_CMD_P1SELFTEST);
uint8_t result = ps2_read_data();
if (result == 0)
    workingflag |= 1;


if (dualport) {
    ps2_write_command(PS2_CTLR_CMD_P2SELFTEST);
    result = ps2_read_data();
    if (result == 0)
        workingflag |= 2;
}

// looking good!
if (workingflag != 0) {
    ps2kbd_init();
    printf("ps2kbd: method 1 success\n");
    goto out;
}

grand shadow Sep 7, 2024, 8:25 PM

#

does it have real ps2

fleet osprey Sep 7, 2024, 8:25 PM

#

so somehow even this minimal of an init kills it

#

no

grand shadow Sep 7, 2024, 8:25 PM

#

hm

fleet osprey Sep 7, 2024, 8:25 PM

#

tbf, my init is even less than this, it doesn't even test it, or read the config

#

so maybe just self testing nukes it?

carmine swallow Sep 7, 2024, 8:26 PM

#

grand shadow Alrovht

so there is https://people.canonical.com/~hwang4/rtdump/ but this utility is absolute garbage from realtek™️ and it only works like one time in a boot and if you run it many times and close it while its "busy" it causes a bsod (and you should tick the all pci devices option if you do try it)

grand shadow Sep 7, 2024, 8:29 PM

#

it created an empty .txt bruh

carmine swallow Sep 7, 2024, 8:29 PM

#

it should give a prompt when its done

#

it took like 5s or smth

grand shadow Sep 7, 2024, 8:29 PM

#

didnt give anything 💀

#

oop

#

running it again worked

#

ok so wtf am I looking for here

carmine swallow Sep 7, 2024, 8:35 PM

#

I don't really know, it doesn't even look like it dumped the connections at all which isn't very useful

#

if you'd know the connection that's used on a working system and all the widget parameters from that path then they could be compared to the ones that they are configured to in astral

#

(though not with the image that I sent to you as it doesn't dump them)

carmine swallow Sep 7, 2024, 8:40 PM

#

carmine swallow I don't really know, it doesn't even look like it dumped the connections at all ...

actually no it did, it just didn't print the ones that the driver uses but the default configs

#

the lines with Wid=XX Codec=XXXXXXXX

grand shadow Sep 7, 2024, 8:42 PM

#

if it helps (I think this is my headphone one, it says jack is plugged)


********** Wid=[0x19] **********
<VREF Hi-Z> <Input Disable><Output Enabled><H-Phn Enabled><Jack is plugged>
Output MUX select to index 0x00 <WID=0x0C>
Amplifier Gain :
    Output Amplifier Gain :
        <Gain Number of step=0, 0dB Offset=0, Step Size=0.00 dB,From 0.00dB to 0.00dB>
        => Index 00 <WID=0x0C> Mute L:0 R:0 , Gain L:0.00dB(0x00) R:0.00dB(0x00)
    Input Amplifier Gain :
        <Gain Number of step=3, 0dB Offset=0, Step Size=10.00 dB, From 0.00dB to 30.00dB>
        => Gain L:0.00dB(0x00) R:0.00dB(0x00)
Power State : [PS-Act : D0], [PS-Set : D0]

Wid=19  Codec=02A19050  Drv=02A19C50  Loc=00030200

#


********** Wid=[0x22] **********
Amplifier Gain :
    Input Amplifier Gain :
        <Gain Number of step=0, 0dB Offset=0, Step Size=0.00 dB,From 0.00dB to 0.00dB>
        => Index 00 <WID=0x18> Mute L:1 R:1 , Gain L:0.00dB(0x00) R:0.00dB(0x00)
        => Index 01 <WID=0x19> Mute L:1 R:1 , Gain L:0.00dB(0x00) R:0.00dB(0x00)

#

idk

fleet osprey Sep 7, 2024, 9:12 PM

#

I'm gonna try just infinite looping after each thing that PS2 init does meme

#

I see no better way to see where it breaks

grand shadow Sep 7, 2024, 9:26 PM

#

print "doing thing"
do thing
sleep idk 5 seconds
repeat for next thing

fleet osprey Sep 7, 2024, 9:26 PM

#

that's.... actually a much more sensible plan than this, wow

#

there's a reason I don't design things :P

carmine swallow Sep 7, 2024, 9:49 PM

#

grand shadow idk

I found at least one bug, I were subtracting the channel count two times when setting the converter format so there was a mismatch between the converter and the stream format (idk how much would that affect the audio tho)

📎 astral.iso

#

(also again be careful because of the volume if you test this, I should really do something about that eg. by dividing it by two or smth)

wild marsh Sep 7, 2024, 9:55 PM

#

btw @grand shadow i just remembered i added a debug allocator into mlibc

#

a while ago

#

if you want to try debugging the gcc issue

#

it uses mmap for each allocation, adds guard pages, places the allocation at the end of the mapped range so an overflow segfaults, and it never frees any addresses (it remaps as PROT_NONE so no actual memory should be used) so use-after-free should just immediately crash

#

you can enable it with -Ddebug_allocator=true when configuring mlibc

robust geyser Sep 7, 2024, 9:58 PM

#

ill try using that and building gcc

#

../../../base_dir/sources/mlibc/options/glibc/generic/malloc.cpp: In function ‘size_t malloc_usable_size(void*)’:
../../../base_dir/sources/mlibc/options/glibc/generic/malloc.cpp:5:31: error: ‘struct MemoryAllocator’ has no member named ‘get_size’
    5 |         return getAllocator().get_size(p);
      |                               ^~~~~~~~```

#

hmm

wild marsh Sep 7, 2024, 9:59 PM

#

oh it's missing that

#

should be easy enough to implement :^)

robust geyser Sep 7, 2024, 10:00 PM

#

me no know how nooo

#

i thought this would be easy sadge

wild marsh Sep 7, 2024, 10:01 PM

#

size_t MemoryAllocator::get_size(void *ptr) {
    if constexpr (logAllocations)
        mlibc::infoLogger() << "MemoryAllocator::get_size(" << ptr << ")" << frg::endlog;

    uintptr_t page_addr = reinterpret_cast<uintptr_t>(ptr) & ~size_t{pageSize - 1};
    AllocatorMeta *meta = reinterpret_cast<AllocatorMeta *>(page_addr - pageSize);

    if (meta->magic != allocatorMagic)
        mlibc::panicLogger() << "Invalid allocator metadata magic in MemoryAllocator::get_size" << frg::endlog;

    return meta->allocatedSize;
}

robust geyser Sep 7, 2024, 10:01 PM

#

holy

#

qookie the goat

#

okay it compiles, ill try running it now, thanks!

wild marsh Sep 7, 2024, 10:05 PM

#

actually, for it to never release addresses you need to change a variable

#

https://github.com/managarm/mlibc/blob/master/options/internal/generic/allocator.cpp#L61

robust geyser Sep 7, 2024, 10:05 PM

#

does that matter?

wild marsh Sep 7, 2024, 10:06 PM

#

well with it off an use-after-free doesn't crash

grand shadow Sep 7, 2024, 10:06 PM

#

carmine swallow I found at least one bug, I were subtracting the channel count two times when se...

I'll try it

robust geyser Sep 7, 2024, 10:06 PM

#

ok ill run with the option enabled

grand shadow Sep 7, 2024, 10:06 PM

#

wild marsh btw <@215823874615541761> i just remembered i added a debug allocator into mlibc

I tried that and grep was bugged for some reason

robust geyser Sep 7, 2024, 10:06 PM

#

constexpr bool neverReleaseVa = true;

#

need to remake image nooo

grand shadow Sep 7, 2024, 10:06 PM

#

you only need to rebuild mlibc thankfully

#

the joy of dynamic linking

wild marsh Sep 7, 2024, 10:07 PM

#

wtf are you in my head

robust geyser Sep 7, 2024, 10:07 PM

#

yeah but mkfs.ext2 is slow as balls

wild marsh Sep 7, 2024, 10:07 PM

#

grand shadow the joy of dynamic linking

i was about to say exactly this

wild marsh Sep 7, 2024, 10:08 PM

#

grand shadow I tried that and grep was bugged for some reason

hmm strange

#

well on managarm grep is a bit bugged even without it so

robust geyser Sep 7, 2024, 10:10 PM

#

lmao what the fuck

grand shadow Sep 7, 2024, 10:11 PM

#

yeah

#

it wasnt just grep

#

sed too

obsidian pivot Sep 7, 2024, 10:11 PM

#

the memory is tired

#

let it sleep

grand shadow Sep 7, 2024, 10:11 PM

#

carmine swallow I found at least one bug, I were subtracting the channel count two times when se...

still happens on my laptop

#

so I assume on the desktop it'll be fucked too

wild marsh Sep 7, 2024, 10:11 PM

#

robust geyser lmao what the fuck

are PROT_NONE anon mappings still backed by pages?

grand shadow Sep 7, 2024, 10:11 PM

#

I'll run it again and show the output

grand shadow Sep 7, 2024, 10:12 PM

#

wild marsh are PROT_NONE anon mappings still backed by pages?

shit I think they are

#

I mean it makes sense

wild marsh Sep 7, 2024, 10:12 PM

#

wild marsh are PROT_NONE anon mappings still backed by pages?

also it's not just remap, it's mmap that replaces the existing mapping with a new one

grand shadow Sep 7, 2024, 10:12 PM

#

youre not unmapping just changing the prot

grand shadow Sep 7, 2024, 10:12 PM

#

wild marsh also it's not just remap, it's mmap that replaces the existing mapping with a ne...

that one destroys the old pages it just does an atomic unmap() map()

wild marsh Sep 7, 2024, 10:12 PM

#

    if constexpr (neverReleaseVa) {
        void *unused;
        if (int e = mlibc::sys_vm_map(meta, meta->pagesSize + pageSize * 2, PROT_NONE, MAP_ANONYMOUS | MAP_PRIVATE | MAP_FIXED, -1, 0, &unused))
            mlibc::panicLogger() << "sys_vm_map failed in MemoryAllocator::deallocate (errno " << e << ")" << frg::endlog;
    } else {
        if (int e = mlibc::sys_vm_unmap(meta, meta->pagesSize + pageSize * 2))
            mlibc::panicLogger() << "sys_vm_unmap failed in MemoryAllocator::deallocate (errno " << e << ")" << frg::endlog;
    }

wild marsh Sep 7, 2024, 10:13 PM

#

grand shadow that one destroys the old pages it just does an atomic unmap() map()

i'd expect that the initial PROT_NONE mapping doesn't cause any pages to get allocated

#

until you mprotect as accessible and access them

prime juniper Sep 7, 2024, 10:15 PM

#

wild marsh well on managarm grep is a bit bugged even without it so

I believe grep works fine on managarm? It triggered a memory leak a few years ago but we fixed that

wild marsh Sep 7, 2024, 10:15 PM

#

i had it hang or crash once yesterday or so? when trying to use -i iirc

prime juniper Sep 7, 2024, 10:16 PM

#

Oh

#

Damn

grand shadow Sep 7, 2024, 10:16 PM

#

wild marsh i'd expect that the initial PROT_NONE mapping doesn't cause any pages to get all...

it doesnt yeah

wild marsh Sep 7, 2024, 10:16 PM

#

so that should work then hmm

prime juniper Sep 7, 2024, 10:16 PM

#

If you can repro, you should open an issue against bootstrap (or mlibc / Managarm if it can be traced to either of them)

wild marsh Sep 7, 2024, 10:16 PM

#

i haven't investigated yet since i ran into it when investigating the egl issue

grand shadow Sep 7, 2024, 10:17 PM

#

@carmine swallow

wild marsh Sep 7, 2024, 10:17 PM

#

i was doing something | grep -i egl or so

prime juniper Sep 7, 2024, 10:17 PM

#

Interesting

wild marsh Sep 7, 2024, 10:17 PM

#

wild marsh i was doing `something | grep -i egl` or so

i think readelf -aW /usr/lib/something? but i don't remember

prime juniper Sep 7, 2024, 10:18 PM

#

If I don’t forget I might try it tomorrow (just a simple grep -i on some meme in usr/include)

grand shadow Sep 7, 2024, 10:20 PM

#

robust geyser lmao what the fuck

if you run it with system call logging it'll show the number of pages free

carmine swallow Sep 7, 2024, 10:20 PM

#

grand shadow <@379250891838193667>

💀

grand shadow Sep 7, 2024, 10:20 PM

#

plus, the vmm would be screaming and sending sigbuses everywhere lol

grand shadow Sep 7, 2024, 10:20 PM

#

carmine swallow 💀

is anything wrong there

carmine swallow Sep 7, 2024, 10:21 PM

#

well I don't really know, it just shows the path and looking at that its a dac connected directly to the output

#

these kinds of things are hard to debug

robust geyser Sep 7, 2024, 10:26 PM

#

why does typing in bash make free page count go down by 2 LMAO @grand shadow

#

syscall: pid 35 tid 36: mmap: hint 0000000002591000 length 8192 prot 3 flags 50 fd -1 offset 0 (468 cached pages, 1007775 free pages)
syscall return: pid 35 tid 36: 39391232 0 (468 cached pages, 1007775 free pages)
syscall: pid 35 tid 36: mmap: hint 0000000000000000 length 12288 prot 0 flags 34 fd -1 offset 0 (468 cached pages, 1007773 free pages)
syscall return: pid 35 tid 36: 39403520 0 (468 cached pages, 1007773 free pages)
syscall: pid 35 tid 36: mmap: hint 0000000002594000 length 8192 prot 3 flags 50 fd -1 offset 0 (468 cached pages, 1007773 free pages)
syscall return: pid 35 tid 36: 39403520 0 (468 cached pages, 1007773 free pages)
syscall: pid 35 tid 36: write: fd 2 buffer 00000000011C3FF8 size 4 (468 cached pages, 1007771 free pages)
syscall return: pid 35 tid 36: 4 0 (468 cached pages, 1007771 free pages)
syscall: pid 35 tid 36: mmap: hint 0000000000000000 length 12288 prot 0 flags 34 fd -1 offset 0 (468 cached pages, 1007771 free pages)
syscall return: pid 35 tid 36: 39415808 0 (468 cached pages, 1007771 free pages)
syscall: pid 35 tid 36: mmap: hint 0000000002597000 length 8192 prot 3 flags 50 fd -1 offset 0 (468 cached pages, 1007771 free pages)
syscall return: pid 35 tid 36: 39415808 0 (468 cached pages, 1007771 free pages)
syscall: pid 35 tid 36: ppoll: fds 0000000002598FF0 nfds 1 timeout 0000000000000000 sigmask 0000000010306020 (468 cached pages, 1007769 free pages)
syscall return: pid 35 tid 36: 1 0 (468 cached pages, 1007769 free pages)
syscall: pid 35 tid 36: mmap: hint 0000000002597000 length 12288 prot 0 flags 50 fd -1 offset 0 (468 cached pages, 1007769 free pages)
syscall return: pid 35 tid 36: 39415808 0 (468 cached pages, 1007771 free pages)
syscall: pid 35 tid 36: read: fd 0 buffer 00007FFFFFFFEBEF size 1 (468 cached pages, 1007771 free pages)
syscall return: pid 35 tid 36: 1 0 (468 cached pages, 1007771 free pages)
syscall: pid 35 tid 36: mmap: hint 0000000000000000 length 12288 prot 0 flags 34 fd -1 offset 0 (468 cached pages, 1007771 free pages)
syscall return: pid 35 tid 36: 39428096 0 (468 cached pages, 1007771 free pages)
syscall: pid 35 tid 36: mmap: hint 000000000259A000 length 8192 prot 3 flags 50 fd -1 offset 0 (468 cached pages, 1007771 free pages)
syscall return: pid 35 tid 36: 39428096 0 (468 cached pages, 1007771 free pages)
syscall: pid 35 tid 36: mmap: hint 0000000000000000 length 12288 prot 0 flags 34 fd -1 offset 0 (468 cached pages, 1007769 free pages)
syscall return: pid 35 tid 36: 39440384 0 (468 cached pages, 1007769 free pages)
syscall: pid 35 tid 36: mmap: hint 000000000259D000 length 8192 prot 3 flags 50 fd -1 offset 0 (468 cached pages, 1007769 free pages)
syscall return: pid 35 tid 36: 39440384 0 (468 cached pages, 1007769 free pages)
syscall: pid 35 tid 36: write: fd 2 buffer 00000000011C3FF8 size 4 (468 cached pages, 1007767 free pages)
syscall return: pid 35 tid 36: 4 0 (468 cached pages, 1007767 free pages)
syscall: pid 35 tid 36: mmap: hint 0000000000000000 length 12288 prot 0 flags 34 fd -1 offset 0 (468 cached pages, 1007767 free pages)
syscall return: pid 35 tid 36: 39452672 0 (468 cached pages, 1007767 free pages)
syscall: pid 35 tid 36: mmap: hint 00000000025A0000 length 8192 prot 3 flags 50 fd -1 offset 0 (468 cached pages, 1007767 free pages)
syscall return: pid 35 tid 36: 39452672 0 (468 cached pages, 1007767 free pages)
syscall: pid 35 tid 36: ppoll: fds 00000000025A1FF0 nfds 1 timeout 0000000000000000 sigmask 0000000010306020 (468 cached pages, 1007765 free pages)```

wild marsh Sep 7, 2024, 10:28 PM

#

the minimum allocation is 2 pages

robust geyser Sep 7, 2024, 10:29 PM

#

yeah but why would it allocate every time i type 😭

#

well its gonna take a while until it runs out of memory, theres still 998k pages remaining

#

but somehow, even with that much memory, sed says "memory exhausted"

#

maybe it just runs out of virtual memory?

wild marsh Sep 7, 2024, 10:36 PM

#

there are 34359738368 pages in the lower half of the address space

robust geyser Sep 7, 2024, 10:37 PM

#

well, thats a lot of virtual address space

#

idk why else would it give that error tho

grand shadow Sep 7, 2024, 10:42 PM

#

I wouldnt think its running out of memory

#

since my vmm would start sending a sigbus to whoever tries to do page in

craggy tide Sep 7, 2024, 10:43 PM

#

@fleet osprey did you get around to doing those tests?

fleet osprey Sep 7, 2024, 10:43 PM

#

I asked a question in #lounge-0 that you never answered

#

#lounge-0 message

craggy tide Sep 7, 2024, 10:44 PM

#

well, i was sleeping

fleet osprey Sep 7, 2024, 10:44 PM

#

ah, timezone differences ig
3PM over here

craggy tide Sep 7, 2024, 10:45 PM

#

i sleep regardless of the position of the sun anyways

fleet osprey Sep 7, 2024, 10:46 PM

#

btw @grand shadow you shouldn't have sched_sleep in the scheduler header if it doesn't really exist, lol

#

it threw me off, and got a linker error

grand shadow Sep 7, 2024, 10:47 PM

#

oh god

#

I think thats there from before I realized a sched_sleep() the way I was gonna do would be a lost wakeup

#

use sched_sleepus

fleet osprey Sep 7, 2024, 10:47 PM

#

yeah I noticed and am adjusting my code

#

just wanted to make note of that

#

perfect, that works exactly like I need it to, time to try it

#

PS2 dies reading the config meme

#

what.....

#

I wonder what happens if I just blindly assume it's a dual port

#

sending selftest also kills it.... ???

#

does sending any data to it at all just kill it?

#

I think it must

#

I wish I knew what DOS was doing to it behind the scenes

#

although actually I guess FreeDOS also works with it

#

I could go look at the source there

#

this is awful, nevermind
I give up trying to read all of that

#

@grand shadow so yeah that's the issue, it dies as soon as you send a single command to it

#

and that's why mine worked, because I didn't send anything to it, just blindly assume it's on IRQ1

grand shadow Sep 7, 2024, 11:09 PM

#

bruh

#

wtf

fleet osprey Sep 7, 2024, 11:10 PM

#

yeah, dell memes ig

#

I guess the weird buffer issue isn't the only bug with it

#

I have no idea how many Dell machines that applies to, I only have 2 of them, an Optiplex 755 from 2007, and a Precision T1650 from 2012

#

if anyone else has a Dell desktop, I can chuck you an ISO or USB image that you can test, that way we can determine if it's actually worth fixing

#

if it's specific to old Dell desktops, it's probably not worth fixing tbh

#

but if it's an issue with any Dell desktop (or maybe even laptops?), then that may actually be worth looking into a workaround

analog berry Sep 7, 2024, 11:14 PM

#

fleet osprey yeah, dell memes ig

i doubt tbh

#

just wait for me to figure this out ig

fleet osprey Sep 7, 2024, 11:15 PM

#

really? I don't see any other reason for it

#

why would it explode the second that you try to read the controller config

#

or do a self test

analog berry Sep 7, 2024, 11:15 PM

#

one possible reason is blindly assuming ps2 ports

#

u need acpi to get the actual port number

#

it could be different

fleet osprey Sep 7, 2024, 11:16 PM

#

oh? interesting

#

because the keyboard entirely turns off with just these 2 lines:

ps2_write_command(PS2_CTLR_CMD_READCFG);
uint8_t control = ps2_read_data();

analog berry Sep 7, 2024, 11:17 PM

#

turns off how?

fleet osprey Sep 7, 2024, 11:17 PM

#

as in, the entire USB keyboard just drops power

analog berry Sep 7, 2024, 11:17 PM

#

ah

fleet osprey Sep 7, 2024, 11:17 PM

#

it's an RGB keyboard, RGB is always on if it's getting 5V

#

everything turns off

analog berry Sep 7, 2024, 11:17 PM

#

well also like

#

linux and windows probably never even hit this codepath in firmware

#

because they force disable emulation

#

so its possible that its bugged

#

but it doesnt work on your laptop either right

fleet osprey Sep 7, 2024, 11:18 PM

#

it works fine with FreeDOS though

fleet osprey Sep 7, 2024, 11:18 PM

#

analog berry but it doesnt work on your laptop either right

PS2 init works on my laptop

analog berry Sep 7, 2024, 11:18 PM

#

freedos is also pretty advanced lol

#

it does have ehci right

fleet osprey Sep 7, 2024, 11:18 PM

#

yes but not by default

#

default is PS2

#

and not even ehci iirc

#

uhci only

analog berry Sep 7, 2024, 11:19 PM

#

its also not 64 bits right

fleet osprey Sep 7, 2024, 11:19 PM

#

and my machine isn't supported by it

fleet osprey Sep 7, 2024, 11:19 PM

#

analog berry its also not 64 bits right

FreeDOS, or the optiplex?

analog berry Sep 7, 2024, 11:19 PM

#

anyway its possible that astral is doing something that makes the firmware expect a modern os

fleet osprey Sep 7, 2024, 11:19 PM

#

the Optiplex is a Core 2 Duo E8400, it's 64-bit

analog berry Sep 7, 2024, 11:19 PM

#

and then uses this codepath

#

freedos ofc

fleet osprey Sep 7, 2024, 11:20 PM

#

FreeDOS is a DOS, lol
it's 16-bit real mode

analog berry Sep 7, 2024, 11:20 PM

#

also

#

https://elixir.bootlin.com/linux/v6.10.8/source/drivers/input/serio/i8042.c#L976

i8042.c - drivers/input/serio/i8042.c - Linux source...

Elixir Cross Referencer - source file of Linux (version v6.10.8). Browsed file: /drivers/input/serio/i8042.c

fleet osprey Sep 7, 2024, 11:21 PM

#

yeah but it still sends a selftest

grand shadow Sep 7, 2024, 11:21 PM

#

analog berry https://elixir.bootlin.com/linux/v6.10.8/source/drivers/input/serio/i8042.c#L976

is that 32 bit only?

fleet osprey Sep 7, 2024, 11:21 PM

#

which is enough to knock the entire thing offline here

analog berry Sep 7, 2024, 11:21 PM

#

grand shadow is that 32 bit only?

no

fleet osprey Sep 7, 2024, 11:21 PM

#

I haven't tried Linux w/o USB support to force PS/2 on here yet

#

entirely possible that not even Linux works on it meme

analog berry Sep 7, 2024, 11:23 PM

#

https://elixir.bootlin.com/linux/v6.10.8/source/drivers/input/serio/i8042-acpipnpio.h

i8042-acpipnpio.h - drivers/input/serio/i8042-acpipnp...

Elixir Cross Referencer - source file of Linux (version v6.10.8). Browsed file: /drivers/input/serio/i8042-acpipnpio.h

#

here's the quirk table

fleet osprey Sep 7, 2024, 11:24 PM

#

my Optiplex isn't there

#

so yeah will try booting some Linux distro while blacklisting all USB

#

try to force PS/2

#

will report if it works, I doubt it

analog berry Sep 7, 2024, 11:25 PM

#

i wonder if ultra legacy ps/2 driver would work there

#

📎 Ultra64HDD-flat.vmdk

#

can u try booting this there

fleet osprey Sep 7, 2024, 11:26 PM

#

yea, sec

#

ok testing now

analog berry Sep 7, 2024, 11:35 PM

#

im expecting it to not work

#

hopefully

fleet osprey Sep 7, 2024, 11:36 PM

#

it actually dies with video stuff

analog berry Sep 7, 2024, 11:36 PM

#

what did it say?

fleet osprey Sep 7, 2024, 11:37 PM

#

Reading kernel from disk...
Preparing kernel environment...
Failed to set the video mode!
Press any key to reboot...

analog berry Sep 7, 2024, 11:37 PM

#

F

#

i guess will never know

fleet osprey Sep 7, 2024, 11:37 PM

#

GPU is an RX 580

#

I can try a few others

#

I have an HD 5450 and a GT 210

analog berry Sep 7, 2024, 11:37 PM

#

yeah the bootloader relies on finding some specific video mode or something idk, its a pretty crappy one

#

if u can thatd be great

fleet osprey Sep 7, 2024, 11:38 PM

#

k, one sec

analog berry Sep 7, 2024, 11:39 PM

#

it wants at least 1024x768@32bpp

#

if it cant find it it will fail

fleet osprey Sep 7, 2024, 11:39 PM

#

yeah hang on

#

I need to wait for it to cool

#

mf is at like, 100C

analog berry Sep 7, 2024, 11:39 PM

#

lol wtf

fleet osprey Sep 7, 2024, 11:40 PM

#

I tried to unplug it and am in excrutiating pain

analog berry Sep 7, 2024, 11:40 PM

#

why is it so hot?

fleet osprey Sep 7, 2024, 11:40 PM

#

no idea

#

it was literally just driving a monitor

#

not even accelerated

#

was just idling at a terminal in Astral

#

lol

analog berry Sep 7, 2024, 11:41 PM

#

lol

#

well astral obviously doesnt do any thermal zone management

#

so it probably just kept working at that fan speed

fleet osprey Sep 7, 2024, 11:41 PM

#

yeah

#

the fan wasn't spinning meme

analog berry Sep 7, 2024, 11:41 PM

#

lmao

#

like the gpu fan or cpu?

fleet osprey Sep 7, 2024, 11:42 PM

#

GPU

#

CPU fan was

analog berry Sep 7, 2024, 11:43 PM

#

ah

#

well the gpu fan lives on its own so idk

fleet osprey Sep 7, 2024, 11:43 PM

#

it's a Dell

#

CPU fan control sucks

analog berry Sep 7, 2024, 11:43 PM

#

yea

fleet osprey Sep 7, 2024, 11:43 PM

#

it's controlled by the firmware unless you do horrific hacks to talk to SMM

analog berry Sep 7, 2024, 11:43 PM

#

which astral does btw via aml

#

and it can even explicitly via control methods, but obv doesnt for now

fleet osprey Sep 7, 2024, 11:44 PM

#

it's not in ACPI iirc

#

it's on an embedded controller

#

undocumented

analog berry Sep 7, 2024, 11:44 PM

#

acpi abstracts it away, u dont need to know the EC layout

#

u just execute a method of the FAN object

fleet osprey Sep 7, 2024, 11:45 PM

#

yeah you can read the RPM from ACPI iirc

#

but can't control it

#

it's i8k

#

https://github.com/TomFreudenberg/dell-bios-fan-control/blob/master/dell-bios-fan-control.c

GitHub

dell-bios-fan-control/dell-bios-fan-control.c at master · TomFreude...

Simple tool to enable or disable the SMBIOS fan (auto) fan control on Dell 9560 - TomFreudenberg/dell-bios-fan-control

analog berry Sep 7, 2024, 11:45 PM

#

https://uefi.org/specs/ACPI/6.5/11_Thermal_Management.html#fsl-fan-set-level

fleet osprey Sep 7, 2024, 11:45 PM

#

Dell doesn't expose that feature

#

you just can't lol, trust me, I've tried

#

the only way is the SMM jank

wild marsh Sep 7, 2024, 11:46 PM

#

ssdt that adds the control methods :^)

analog berry Sep 7, 2024, 11:46 PM

#

^

#

what is i8k?

fleet osprey Sep 7, 2024, 11:46 PM

#

I don't know what it actually stands for

analog berry Sep 7, 2024, 11:47 PM

#

oh its dell smm interface

fleet osprey Sep 7, 2024, 11:47 PM

#

yes

#

that's the only fan control method for Dell machines

#

Desktop, Laptop, anything

analog berry Sep 7, 2024, 11:47 PM

#

bruh

fleet osprey Sep 7, 2024, 11:47 PM

#

it's the only way

analog berry Sep 7, 2024, 11:50 PM

#

a vendor without weird proprietary crap is a bad vendor

fleet osprey Sep 7, 2024, 11:50 PM

#

yeah

#

my friend actually knows the guy who wrote the code

#

in the firmware

analog berry Sep 7, 2024, 11:50 PM

#

lol

#

he should really congratulate the guy on inventing such a beautiful new interface

fleet osprey Sep 7, 2024, 11:53 PM

#

the magic hex bytes in decimal are 13731 and 13475, and he says that the guy wanted the machine to boot in 13.475 seconds

#

so that just became an inside joke because it's stupidly precise goal

#

and now they use 13475 (and various variations of it in hex) whenever they're exposing something that they think is stupid

analog berry Sep 7, 2024, 11:54 PM

#

lol

fleet osprey Sep 7, 2024, 11:54 PM

#

anyways I've stalled for long enough, I bet the GPU has cooled down enough to pull it out now

analog berry Sep 7, 2024, 11:55 PM

#

btw 2007 is probably earliest AML uacpi has been tested with which is interesting

#

its from the more crazy days

#

does it have any errors or anything in the log?

#

in astral i mean

fleet osprey Sep 7, 2024, 11:56 PM

#

1 warning

#

checksum mismatch

#

that's it

analog berry Sep 7, 2024, 11:57 PM

#

nice

#

your other pc had duplicate methods i think

#

for handling firmware events

fleet osprey Sep 7, 2024, 11:57 PM

#

"my other PC" could refer to many things

#

was it the laptop I should here recently?

analog berry Sep 7, 2024, 11:58 PM

#

oh yeah probably

fleet osprey Sep 7, 2024, 11:58 PM

#

that's a ThinkPad T14 Gen 1 AMD

#

pretty recent

#

shocking it has issues

#

it's from at least 2020 iirc

#

your image just hangs

analog berry Sep 7, 2024, 11:59 PM

#

fleet osprey Here's what it does repeatedly mashing (more than 4) keys

yeah this

fleet osprey Sep 7, 2024, 11:59 PM

#

black screen, everything locks up

fleet osprey Sep 8, 2024, 12:00 AM

#

analog berry yeah this

yeah that's the T14

analog berry Sep 8, 2024, 12:00 AM

#

i wonder if its stuck in ps2 init 💀

fleet osprey Sep 8, 2024, 12:00 AM

#

well the kbd didn't turn off

analog berry Sep 8, 2024, 12:01 AM

#

fleet osprey shocking it has issues

meh its not really that big of an issue

#

any way u can do acpidump > dump.txt on it?

fleet osprey Sep 8, 2024, 12:02 AM

#

📎 dump.txt

#

anything useful in there?

analog berry Sep 8, 2024, 12:05 AM

#

yeah

fleet osprey Sep 8, 2024, 12:06 AM

#

cool, I'm not an ACPI nerd, so I don't know what any of that means, lol
I recognize some of the table names from very early attempts at poking at it in my OS

#

and Hackintoshing as well, I had to poke at them

analog berry Sep 8, 2024, 12:07 AM

#

im mostly looking at the aml disassembly, u can see yourself

acpixtract -a dump.txt
iasl dsdt.dat

obsidian pivot Sep 8, 2024, 12:07 AM

#

unfortunately no MSDM table

fleet osprey Sep 8, 2024, 12:07 AM

#

want me to chuck the output here?

#

oh wait

#

nvm just re-read ur msg

analog berry Sep 8, 2024, 12:08 AM

#

yeah just in case u were curious

obsidian pivot Sep 8, 2024, 12:09 AM

#

whats dmar table

#

in acpi\

#

seems to have something to do with the iommu

#

or dma

fleet osprey Sep 8, 2024, 12:10 AM

#

the generated dsl file all just looks like gibberish to me, but I trust that it's actually useful somehow
never got far enough into my OS to really try doing ACPI stuff

#

would probably get more meaning out of it if I did

analog berry Sep 8, 2024, 12:11 AM

#

i think its a bug in uacpi

obsidian pivot Sep 8, 2024, 12:11 AM

#

impossible

analog berry Sep 8, 2024, 12:11 AM

#

its harmless but yeah

obsidian pivot Sep 8, 2024, 12:12 AM

#

Name (_S5, Package (0x04)  // _S5_: S5 System State
    {
        0x07, 
        Zero, 
        Zero, 
        Zero
    })```

#

how do I interpret this

#

(I'm doing fnuy things with ports)

analog berry Sep 8, 2024, 12:12 AM

#

0x07 is the value that should be written to SLP_TYPa

#

for the system to enter S5

wild marsh Sep 8, 2024, 12:12 AM

#

what are the other 3 values for?

obsidian pivot Sep 8, 2024, 12:13 AM

#

analog berry 0x07 is the value that should be written to SLP_TYPa

so this:

[040h 0064   4]   PM1A Control Block Address : 00001804```

analog berry Sep 8, 2024, 12:13 AM

#

wild marsh what are the other 3 values for?

value two is SLP_TYPb if it exists

#

the rest are unused i think, they used to mean something

wild marsh Sep 8, 2024, 12:14 AM

#

huh

analog berry Sep 8, 2024, 12:15 AM

#

wild marsh Sep 8, 2024, 12:15 AM

#

interesting

analog berry Sep 8, 2024, 12:17 AM

#

obsidian pivot so this: ``` [040h 0064 4] PM1A Control Block Address : 00001804```

yeah

obsidian pivot Sep 8, 2024, 12:17 AM

#

ok

analog berry Sep 8, 2024, 12:18 AM

#

anyway there's a bug here that makes the iteration escape the scope in some case

#

ill debug tomorrow

obsidian pivot Sep 8, 2024, 12:19 AM

#

analog berry yeah

it's a four byte write, right?

analog berry Sep 8, 2024, 12:21 AM

#

whatever the GAS structure says

analog berry Sep 8, 2024, 12:23 AM

#

obsidian pivot so this: ``` [040h 0064 4] PM1A Control Block Address : 00001804```

Fadt also has a length field for it I think, which if u dont have gas basically is the access size

obsidian pivot Sep 8, 2024, 12:24 AM

#

ok

analog berry Sep 8, 2024, 12:24 AM

#

Just scroll below

#

Anyway will have to get back to ps2 later

#

I dont see it in the quirks table so im assuming astral skill issue for now

analog berry Sep 8, 2024, 12:27 AM

#

fleet osprey the generated dsl file all just looks like gibberish to me, but I trust that it'...

Its basically the device tree of your pc, plus methods to interact with said devices or detect resources they're consuming and/or configure them

#

Btw if u can run Linux without USB drivers on it that would be nice

fleet osprey Sep 8, 2024, 12:43 AM

#

oh on the optiplex?

#

ye

#

hang on

obsidian pivot Sep 8, 2024, 12:54 AM

#

obsidian pivot (I'm doing fnuy things with ports)

BITS 64
DEFAULT REL
extern iopl
global main
main:
        mov rdi, 3
        call iopl
        mov dx, 0x1804
        mov ax, (7<<10)|(1<<13)
        out dx,ax
        xor rax,rax
        ret```

fleet osprey Sep 8, 2024, 1:06 AM

#

analog berry Btw if u can run Linux without USB drivers on it that would be nice

Ubuntu server doesn't seem to like me

#

great

#

let me try on Arch ig, hang on

#

hell, may need to just go compile a new kernel, I bet Arch's also has USB built in

#

yeah, Ubuntu's config just sucks meme

CONFIG_USB_ARCH_HAS_HCD=y
CONFIG_USB_XHCI_HCD=y
CONFIG_USB_EHCI_HCD=y
CONFIG_USB_EHCI_HCD_PLATFORM=y
CONFIG_USB_OHCI_HCD=y
CONFIG_USB_OHCI_HCD_PCI=y
CONFIG_USB_OHCI_HCD_PLATFORM=y
CONFIG_USB_UHCI_HCD=y
CONFIG_USB_SL811_HCD_ISO=y
# CONFIG_USB_HCD_TEST_MODE is not set
CONFIG_USB_ISP1760_HCD=y
# CONFIG_USB_DUMMY_HCD is not set

#

damn even Arch does that

#

welp, time to go recompile Linux, so I can test if the Linux PS2 drivers works, so I can see if the issue is Dell's PS2 implementation, or Astral's driver being iffy

grand shadow Sep 8, 2024, 1:12 AM

#

likely the latter

fleet osprey Sep 8, 2024, 1:12 AM

#

well Dell's implementation is known dogshit

#

I sent the description of the weird buffer issue before

obsidian pivot Sep 8, 2024, 1:12 AM

#

fleet osprey well Dell's implementation is known dogshit

iirc my dell computer duplicates chars on ps/2 emulation

fleet osprey Sep 8, 2024, 1:12 AM

#

mine is the opposite, lol

obsidian pivot Sep 8, 2024, 1:13 AM

#

well it's more like

#

I click a key

#

and it thinks I'm holding it

grand shadow Sep 8, 2024, 1:13 AM

#

my desktop seems to do ps2 emulation perfectly, even emulates the mouse

#

guess I was lucky

#

or it just emulates it in a way astral's driver expects meme

obsidian pivot Sep 8, 2024, 1:14 AM

#

u could just replace all sort of display stuff with serial

#

instead of keyboard

#

use serial

#

instead of Xorg/bash

#

ssh over serial

grand shadow Sep 8, 2024, 1:15 AM

#

atp write a nic driver and have ssh and all the other benefits of a working internet connection

fleet osprey Sep 8, 2024, 1:17 AM

#

oh lol, mine is not at all that
mine has the unfortunate issue of
- press key #1
- puts it in key buffer, and fires an IRQ
- OS pops char from buffer, buffer now empty
- press key #2
- puts it in buffer, does not fire an IRQ
- press key #3
- puts it in buffer, does fire an IRQ
- OS pops key #2 from buffer, buffer now has 1 dead char
- press key #4
- puts it in buffer, does not fire an IRQ
- press key #5
- puts it in buffer, does fire an IRQ
- OS pops key #3 from buffer, buffer now has 2 dead chars

#

and this repeats into in infinity

obsidian pivot Sep 8, 2024, 1:18 AM

#

obvious solution is to just

#

have a thread polling the ps/2 statys thing

#

for data /j

fleet osprey Sep 8, 2024, 1:18 AM

#

it's bad enough that it breaks DOS

robust geyser Sep 8, 2024, 1:18 AM

#

no, the obvious solution is to drain the buffer on each irq

#

k

#

fair enough

fleet osprey Sep 8, 2024, 1:19 AM

#

yeah

#

so I really wouldn't be shocked if their driver has other bugs

grand shadow Sep 8, 2024, 1:26 AM

#

how does it even get past qa

robust geyser Sep 8, 2024, 1:26 AM

#

works on linux™️

fleet osprey Sep 8, 2024, 1:30 AM

#

grand shadow how does it even get past qa

you think they have QA for legacy PS/2 emulation for USB keyboards on a 2007 Dell office PC?

#

lol

#

if they tested it at all (they probably didn't), they probably keysmashed and checked if anything at all appeared

grand shadow Sep 8, 2024, 1:40 AM

#

gotta make sure it works for that one guy still running an ancient windows version

fleet osprey Sep 8, 2024, 1:53 AM

#

it's an OptiPlex

#

it's an office PC

#

it's meant to be used as a glorified typewritter

#

or to run excel

#

it's meant to have the drive imaged with the company's Windows XP install and then left to sit for 8 years until it's decomissioned

fleet osprey Sep 8, 2024, 2:09 AM

#

analog berry Btw if u can run Linux without USB drivers on it that would be nice

can't really confirm because no video, but the keyboard turned off

#

I bet it died

grand shadow Sep 8, 2024, 2:10 AM

#

💀

#

So it wasnt astral

fleet osprey Sep 8, 2024, 2:14 AM

#

yeah

#

it just eats hit and dies if you ever send a command to the PS2 controller

grand shadow Sep 8, 2024, 2:15 AM

#

Does it have any nic

#

And if so which

fleet osprey Sep 8, 2024, 2:16 AM

#

e1000

grand shadow Sep 8, 2024, 2:16 AM

#

Hmmmm

#

I do need to write an e1000 driver

fleet osprey Sep 8, 2024, 2:16 AM

#

I also have another one that's also e1000 but different™️

#

iirc lspci shows different model, but still e1000 driver

grand shadow Sep 8, 2024, 2:17 AM

#

Maybe it has some fucked up quirk

fleet osprey Sep 8, 2024, 2:22 AM

#

yeah the Precision has an 82579LM "Lewisville"
Optiplex has 82566DM-2

#

oh yeah I also have 3x R8169 devices

#

my laptop, primary desktop, and mini PC, are all Realtek 8169

analog berry Sep 8, 2024, 8:32 AM

#

fleet osprey can't really confirm because no video, but the keyboard turned off

What happened to video?

fleet osprey Sep 8, 2024, 8:33 AM

#

nvidia GPU

#

need I say more?

analog berry Sep 8, 2024, 8:34 AM

#

But you have more GPUs right

fleet osprey Sep 8, 2024, 8:34 AM

#

I probably could test it out, but I'm busy

#

and definetly not rn

#

it's 1:30AM

analog berry Sep 8, 2024, 8:39 AM

#

Yeah sure

supple hatch Sep 8, 2024, 9:27 AM

#

why do i always get the weird issues mathew 😭

#

like does god hate me or smthin

#

no matter what language, C or rust i always get weird ah issues that take forever to deebug

#

its so annoying

#

also ps2 is weird

#

usually usb emulation is shitty

#

so if u dont get that shit working dont beat yourself up about it, not your fault rlly

analog berry Sep 8, 2024, 9:30 AM

#

Im unironically tempted to do uPS2

#

But maybe later at some point

supple hatch Sep 8, 2024, 9:31 AM

#

analog berry Im unironically tempted to do uPS2

u should to be fair

#

!!

analog berry Sep 8, 2024, 9:31 AM

#

The main point would be quirk tables for broken ps2 emulation

#

So that init is skipped in that case

supple hatch Sep 8, 2024, 9:31 AM

#

yea honestly that would be cool

#

i think it would be really cool if u did somthin like that

#

but whatever works best for u, i think it'd help hobby oses a lot

#

usb is difficult

analog berry Sep 8, 2024, 9:32 AM

#

I would/could ship touchpad and multiplexed ps2 drivers

supple hatch Sep 8, 2024, 9:32 AM

#

that would be cool also

analog berry Sep 8, 2024, 9:33 AM

#

Thanks, ill think about how I would design it

supple hatch Sep 8, 2024, 9:33 AM

#

def! i'd be glad to support a project like that

#

might even contribute if i can

analog berry Sep 8, 2024, 9:33 AM

#

Well you most likely wouldn't be able to use it because rust KEKW

supple hatch Sep 8, 2024, 9:34 AM

#

analog berry Well you most likely wouldn't be able to use it because rust <:KEKW:913708372329...

nuh uh!

#

i can still write C

#

im not a total rust shill

#

im only writing in rust for nyaux cause build system lol

obsidian pivot Sep 8, 2024, 10:42 AM

#

yet

grand shadow Sep 8, 2024, 12:54 PM

#

The astral curse is ps2 never working

analog berry Sep 8, 2024, 12:57 PM

#

Id like to see the acpi dump of that ps2 emulation pc, whether it mentions ps2 devices in the dsdt

#

If it doesn't I can easily work around that in astral and skip init in that case

grand shadow Sep 8, 2024, 1:56 PM

#

anyways

#

for today uh

#

I will implement sigtimedwait and port qemu

#

damn 90 syscalls already

grand shadow Sep 8, 2024, 4:24 PM

#

wonder which web server would be easier to port

carmine swallow Sep 8, 2024, 4:41 PM

#

civetweb was pretty easy

#

if you just want something to test stuff with

grand shadow Sep 8, 2024, 4:53 PM

#

hmmmm

carmine swallow Sep 8, 2024, 4:54 PM

#

iirc it didn't even need any patches

errant zephyr Sep 8, 2024, 5:29 PM

#

grand shadow wonder which web server would be easier to port

pitust® lolttpd™

#

it performs like shit despite using epoll and whatnot

grand shadow Sep 8, 2024, 5:30 PM

#

epoll

#

__ensure(!"unimplemented"); meme

errant zephyr Sep 8, 2024, 5:35 PM

#

grand shadow __ensure(!"unimplemented"); <:meme:575445094589661215>

well dw you can provide kqueue instead

#

i support both

grand shadow Sep 8, 2024, 6:42 PM

#

qemu time

supple hatch Sep 8, 2024, 6:44 PM

#

hehehe

earnest zinc Sep 8, 2024, 6:45 PM

#

Tail wags Hello!

supple hatch Sep 8, 2024, 6:46 PM

#

hey

obsidian pivot Sep 8, 2024, 7:53 PM

#

supple hatch hey

who hast thou brought upon us?

#

"Max The Cat"

fleet osprey Sep 8, 2024, 9:08 PM

#

new astral issue, btw
Dell Precision T1650, Xeon E3-1240 V2, only happens with UEFI boot
Possibly specific to Ventoy? Need to test it more

#

Oh, and my primary desktop in UEFI mode is also affected by this.

grand shadow Sep 8, 2024, 9:15 PM

#

fleet osprey new astral issue, btw Dell Precision T1650, Xeon E3-1240 V2, only happens with U...

what iso is this I thought I had fixed this bug

grand shadow Sep 8, 2024, 9:15 PM

#

fleet osprey Oh, and my primary desktop in UEFI mode is also affected by this.

and this is probably aml wanting to allocate a bigger memory size than the alloc() supports

analog berry Sep 8, 2024, 9:16 PM

#

qwinci had a fix for this

#

iirc

grand shadow Sep 8, 2024, 9:16 PM

#

yeah

fleet osprey Sep 8, 2024, 9:16 PM

#

grand shadow what iso is this I thought I had fixed this bug

I built it myself at most 2 days ago
I can pull and rebuild just in case if you want

obsidian pivot Sep 8, 2024, 9:16 PM

#

ok I think I got it now:
say "m68k" to summon fadanoid
say "aml" to summon infy

analog berry Sep 8, 2024, 9:16 PM

#

yes

grand shadow Sep 8, 2024, 9:16 PM

#

fleet osprey I built it myself at most 2 days ago I can pull and rebuild just in case if you ...

bruh

#

what commit are oyu at

fleet osprey Sep 8, 2024, 9:17 PM

#

a0a902e968ab8dcd6c6c8cc87f5c31fb27be7c12

grand shadow Sep 8, 2024, 9:18 PM

#

hm

fleet osprey Sep 8, 2024, 9:18 PM

#

it looks like I'm missing your merge commit for my PR, and one other

grand shadow Sep 8, 2024, 9:18 PM

#

something is overlapping in the same page when allocating space for the hhdm

#

if (e->type != LIMINE_MEMMAP_USABLE && e->type != LIMINE_MEMMAP_BOOTLOADER_RECLAIMABLE && e->type != LIMINE_MEMMAP_KERNEL_AND_MODULES && e->type != LIMINE_MEMMAP_FRAMEBUFFER)

#

its one of these

#

USABLE and RECLAIMABLE are guaranteed to be aligned iirc

fleet osprey Sep 8, 2024, 9:20 PM

#

just in case, rebuilding the ISO fully up to date with your gh

grand shadow Sep 8, 2024, 9:21 PM

#

can you add a print before the vmm_map to see each thing it tries to map

#

here

obsidian pivot Sep 8, 2024, 9:21 PM

#

btw this probably counts as oberrow curse

grand shadow Sep 8, 2024, 9:21 PM

#

its not a vmm corruption or anything its just real hardware being stupid meme

obsidian pivot Sep 8, 2024, 9:22 PM

#

I think oberrow curse is just memory bugs

#

in general

#

sooooo

fleet osprey Sep 8, 2024, 9:23 PM

#

so what am I printing here, the info from entry (e)? I haven't looked at the Astral vmm yet (and never actually got far enough to make one in my OS meme ) so I'm not necessarily an expert here

grand shadow Sep 8, 2024, 9:24 PM

#

printf("base %p length %x type %d\n", e->base, e->length, e->type);

fleet osprey Sep 8, 2024, 9:24 PM

#

👍
will retest on the t1650, one sec

#

grand shadow Sep 8, 2024, 9:37 PM

#

fleet osprey

wtf

#

that doesnt seem page aligned?

analog berry Sep 8, 2024, 9:40 PM

#

limine bug ig

fleet osprey Sep 8, 2024, 9:40 PM

#

or ventoy bug

#

ventoy is fucky

#

as mentioned, I need to test w/o it

analog berry Sep 8, 2024, 9:41 PM

#

limine is the one doing allocations

fleet osprey Sep 8, 2024, 9:41 PM

#

true, I wonder why it's not page aligned then, since the spec says it's garuanteed to be

analog berry Sep 8, 2024, 9:42 PM

#

@craggy tide any ideas?

grand shadow Sep 8, 2024, 10:06 PM

#

welp I have to do a clean build

#

and once I'm done yall will have too

#

switching out mlibc's iconv and intl options to libiconv and libintl (from gettext)

fleet osprey Sep 8, 2024, 10:07 PM

#

eh, that's not that bad with DISTROTYPE=minimal

#

only takes ~15-20m on my machine, depending on outside system load

craggy tide Sep 8, 2024, 10:11 PM

#

analog berry <@379163868364275712> any ideas?

what limine version?

fleet osprey Sep 8, 2024, 10:11 PM

#

~~unless it was modified from my PR that bumped the version,~~ (just checked, it wasn't) it's 8.0.9

craggy tide Sep 8, 2024, 10:18 PM

#

it seems like it is carving the framebuffer entry out of bootloader reclaimable memory

#

but that is weird unless the firmware originally marked the framebuffer as usable memory

#

which was then allocated as bootloader reclaimable

fleet osprey Sep 8, 2024, 10:19 PM

#

it may very well be a Ventoy bug, since it does set the video mode
give me 2m here and I can test directly w/o ventoy

craggy tide Sep 8, 2024, 10:20 PM

#

okay

fleet osprey Sep 8, 2024, 10:21 PM

#

yeah it totally doesn't do that when booting direct

#

only happens when booting from ventoy, specifically in UEFI mode on this machine

#

it may affect others, but none that I personally owen

craggy tide Sep 8, 2024, 10:22 PM

#

can you show the memory map

#

when not booting with ventoy

flat copper Sep 8, 2024, 10:22 PM

#

you should specifically check if you're booting with ventoy and show a big fat warning that it's not supported

analog berry Sep 8, 2024, 10:23 PM

#

i mean unless its not a ventoy bug

craggy tide Sep 8, 2024, 10:23 PM

#

analog berry limine is the one doing allocations

yeah but the firmware is the one passing nonsensical memory maps

analog berry Sep 8, 2024, 10:23 PM

#

we havent seen it yet but yeah possible

craggy tide Sep 8, 2024, 10:23 PM

#

like the only way this can happen really is if the framebuffer is marked as usable memory

#

which would be weird

analog berry Sep 8, 2024, 10:24 PM

#

how would that be possible, its usualy just one of the bars addresses

#

of the gpu

craggy tide Sep 8, 2024, 10:24 PM

#

i don't know

#

i am trying to think of another mechanism you could end up with that memory map

analog berry Sep 8, 2024, 10:24 PM

#

maybe u could add printf to limine to dump the memory map given to u

#

and we could run that

craggy tide Sep 8, 2024, 10:25 PM

#

i mean yeah we can do tests

analog berry Sep 8, 2024, 10:25 PM

#

then we can at least submit a bug report

fleet osprey Sep 8, 2024, 10:25 PM

#

ah, that works
I was about to go write a tiny kernel that dumps the memory map to COM1

#

lol

craggy tide Sep 8, 2024, 10:25 PM

#

if confirmed i would still like to work around the issue

analog berry Sep 8, 2024, 10:26 PM

#

the workaround would be to scan the bars manually and see if they fall into any usable entries?

fleet osprey Sep 8, 2024, 10:28 PM

#

fleet osprey ah, that works I was about to go write a tiny kernel that dumps the memory map t...

I actually have code for this in my OS, I would just need to port it up to Limine v8, it's so old that it's still stuck on v5

#

if yall want I can spend a few minutes doing that and get the memory map

analog berry Sep 8, 2024, 10:28 PM

#

nono we need the original firmware map, not the limine map

fleet osprey Sep 8, 2024, 10:28 PM

#

oh 🤦

analog berry Sep 8, 2024, 10:28 PM

#

the limine map we already saw in astral

fleet osprey Sep 8, 2024, 10:29 PM

#

does Limine have some builtin way to dump that? I thought there used to be a console you could open or something, but don't see it in v8

craggy tide Sep 8, 2024, 10:29 PM

#

yes

#

there is a function to dump the memory map

#

i am trying to figure out a good way to print the fb address, give me a sec

fleet osprey Sep 8, 2024, 10:30 PM

#

👍 will continue working on that compatibility spreasheet then

analog berry Sep 8, 2024, 10:30 PM

#

the what?

fleet osprey Sep 8, 2024, 10:31 PM

#

#lounge-0 message

fleet osprey Sep 8, 2024, 10:31 PM

#

fleet osprey \- does **not** work if Rufus is used to flash an ISO to USB \- does **not** wor...

expanded version of ^^^

analog berry Sep 8, 2024, 10:32 PM

#

im honestly astonished theres a difference

#

like these are tools to do the same thing

#

lol

fleet osprey Sep 8, 2024, 10:32 PM

#

yup

#

current status:

#

with a 4KB text file of notes about various failures meme

craggy tide Sep 8, 2024, 10:33 PM

#

📎 BOOTX64.EFI

#

try this

fleet osprey Sep 8, 2024, 10:33 PM

#

one sec

craggy tide Sep 8, 2024, 10:33 PM

#

it prints the memory map (as raw as possible) followed by the framebuffer address

gritty shore Sep 8, 2024, 10:34 PM

#

fleet osprey does Limine have some builtin way to dump that? I thought there used to be a co...

bring back command line

#

I mean

#

limine shell

#

whatever

craggy tide Sep 8, 2024, 10:35 PM

#

no

analog berry Sep 8, 2024, 10:35 PM

#

bring back the ntfs driver

craggy tide Sep 8, 2024, 10:35 PM

#

i will if you maintain it, deal?

analog berry Sep 8, 2024, 10:35 PM

#

i will maintain it if its complete meme

craggy tide Sep 8, 2024, 10:35 PM

#

it's not complete

#

lol

analog berry Sep 8, 2024, 10:35 PM

#

once*

craggy tide Sep 8, 2024, 10:35 PM

#

never will be

#

unless someone steps up and does it

analog berry Sep 8, 2024, 10:36 PM

#

true

craggy tide Sep 8, 2024, 10:36 PM

#

i have 0 interest in doing it myself

fleet osprey Sep 8, 2024, 10:36 PM

#

Without ventoy

analog berry Sep 8, 2024, 10:37 PM

#

wheres the fb address?

craggy tide Sep 8, 2024, 10:37 PM

#

that's not the full memmap

#

you have to press a key to continue

#

(it shows the rest)

fleet osprey Sep 8, 2024, 10:37 PM

#

oh 🤦 I thought that meant like, to continue to the menu

#

sec

craggy tide Sep 8, 2024, 10:37 PM

#

nope

analog berry Sep 8, 2024, 10:37 PM

#

wait why did u do it like that lol

craggy tide Sep 8, 2024, 10:37 PM

#

like what?

analog berry Sep 8, 2024, 10:38 PM

#

having to press the key to see the whole thing

craggy tide Sep 8, 2024, 10:38 PM

#

because efi console

#

in qemu it's really tiny

analog berry Sep 8, 2024, 10:38 PM

#

ohh

craggy tide Sep 8, 2024, 10:38 PM

#

i did not want it to scroll away too fast

analog berry Sep 8, 2024, 10:38 PM

#

its usually giant on real hw which is nice

craggy tide Sep 8, 2024, 10:38 PM

#

and the memmap can be quite big

#

yeah

fleet osprey Sep 8, 2024, 10:38 PM

#

holy shit
it's like, 20 screens, and the console doesn't fill the whole screen

#

hang on

analog berry Sep 8, 2024, 10:39 PM

#

now i see why meme

fleet osprey Sep 8, 2024, 10:40 PM

#

Thankfully it fits 2 chunks at a time

#

#

each of those are 2 keypresses (1 screen) apart

#

that's one hell of a memory map

#

wtf

analog berry Sep 8, 2024, 10:43 PM

#

thats normal, u can compress it quite well

#Astral