Managarm and related projects | Operating System Development | Page 3

upbeat smelt Mar 17, 2024, 9:15 PM

#

send the whole machine to this address

worn rampart Mar 17, 2024, 9:15 PM

#

and the 3 digits in the back of your credit card

brittle mauve Mar 17, 2024, 9:15 PM

#

https://cdn.discordapp.com/attachments/829403502349647952/1072985432066887731/acpi.tar?ex=6604890e&is=65f2140e&hm=c2f43a9d06f9e881ddc4b7fd4195c10c6b9d6c47313d7ada1735890301c81030& should have all the acpi tables, I don't think I have updated the bios or anything since I did that dump

native prairie Mar 17, 2024, 9:16 PM

#

Thanks

#

Yeah your bios is fucked in the head

#

But its good because I can test against that

worn rampart Mar 17, 2024, 9:16 PM

#

if windows boots its not fucked enough

native prairie Mar 17, 2024, 9:17 PM

#

They dont even look at the windows debugger shitting itself

#

They just assume if it boots its good enough

worn rampart Mar 17, 2024, 9:17 PM

#

I mean, why wouldn't it be fine if it boots :^)

native prairie Mar 17, 2024, 9:17 PM

#

Lmao

brittle mauve Mar 17, 2024, 9:18 PM

#

Also desktop

rn_image_picker_lib_temp_c91b536b-d138-4c87-b43e-181c935c1172.jpg

native prairie Mar 17, 2024, 9:18 PM

#

Oh interesting

native prairie Mar 17, 2024, 9:18 PM

#

brittle mauve Also desktop

Can you also dump those as well

storm plover Mar 17, 2024, 9:18 PM

#

brittle mauve Also desktop

same thing happens on my desktop

native prairie Mar 17, 2024, 9:19 PM

#

I wonder if it has no _PRT

#

I dont see any errors in the logs above

native prairie Mar 17, 2024, 9:19 PM

#

storm plover same thing happens on my desktop

Were there errors above that message?

storm plover Mar 17, 2024, 9:19 PM

#

no

native prairie Mar 17, 2024, 9:19 PM

#

Dump tables pls if you dont mind

brittle mauve Mar 17, 2024, 9:22 PM

#

📎 acpitables.tar.gz

safe glen Mar 17, 2024, 9:25 PM

#

unlikely but you can just try running this iso

📎 template.iso

native prairie Mar 17, 2024, 9:25 PM

#

@brittle mauve ill add additional logging for _PRT to see which one fails, that dump has like 50 different ones

worn rampart Mar 17, 2024, 9:26 PM

#

safe glen unlikely but you can just try running this iso

yeah I am just gonna assume its a managarm issue

#

given limine itself shows up correctly

#

and I am pretty sure I have used limine on that laptop before

safe glen Mar 17, 2024, 9:26 PM

#

it's not the first report i hear of managarm presenting a blank screen when attempting to boot it

#

i would still test that iso, which admittedly doesn't necessarily 100% rule out potential issues in the multiboot2 implementation of Limine, but still

native prairie Mar 17, 2024, 9:31 PM

#

brittle mauve Also desktop

might be this thing that fails

    obj = uacpi_namespace_node_get_object(parent);
    if (uacpi_unlikely(obj == UACPI_NULL || obj->type != UACPI_OBJECT_DEVICE))
        return UACPI_STATUS_INVALID_ARGUMENT;

#

maybe the bus device is not a device but something else

#

we're about to see

#

I can remove that check if thats the case

forest bear Mar 17, 2024, 9:32 PM

#

first time i see someone use unlikely() apart from linux

native prairie Mar 17, 2024, 9:32 PM

#

well its all over uacpi LULW

worn rampart Mar 17, 2024, 9:32 PM

#

my CHECK macro usually has unlikely spiced into it

native prairie Mar 17, 2024, 9:32 PM

#

it does change code gen quite a lot

brittle mauve Mar 17, 2024, 9:38 PM

#

also if the shutdown would work on my desktop it would be interesting to see if wake by usb works after it because as far as I remember when I perform a shutdown from windows then it doesn't for some reason but if from linux then it does work

native prairie Mar 17, 2024, 9:38 PM

#

wake requires quite a lot of code on managarm's side

brittle mauve Mar 17, 2024, 9:39 PM

#

not from sleep but from s5

#

as in turning on the pc when I press a key on the keyboard

native prairie Mar 17, 2024, 9:39 PM

#

brittle mauve not from sleep but from s5

well thats a full shutdown

brittle mauve Mar 17, 2024, 9:39 PM

#

yes

native prairie Mar 17, 2024, 9:41 PM

#

📎 image.tar.xz

#

can you guys please try this @storm plover @brittle mauve

#

this has more logging if prt fails

storm plover Mar 17, 2024, 9:52 PM

#

output is still the same

native prairie Mar 17, 2024, 9:52 PM

#

this adds extra logs

#

can u show what it says?

storm plover Mar 17, 2024, 9:52 PM

#

I don’t see any extra logs

native prairie Mar 17, 2024, 9:52 PM

#

what do u see

brittle mauve Mar 17, 2024, 9:54 PM

#

native prairie can u show what it says?

it doesn't show anything more than with the old image

#

it just shows the fault

native prairie Mar 17, 2024, 9:54 PM

#

and the line above the fault?

#

I added this

#

wait im retarded

#

i didnt regen the image

storm plover Mar 17, 2024, 9:55 PM

#

native prairie Mar 17, 2024, 9:55 PM

#

for fucks sake

brittle mauve Mar 17, 2024, 9:55 PM

#

can you just send the changed kernel

native prairie Mar 17, 2024, 9:55 PM

#

yup

brittle mauve Mar 17, 2024, 9:55 PM

#

it would be faster

native prairie Mar 17, 2024, 9:56 PM

#

📎 thor

brittle mauve Mar 17, 2024, 9:59 PM

#

\_SB_.PC00.PEG1.HPEV

native prairie Mar 17, 2024, 9:59 PM

#

and the error code?

brittle mauve Mar 17, 2024, 10:00 PM

#

invalid argument

native prairie Mar 17, 2024, 10:01 PM

#

uhhh lol

#

thats a method

#

why does it try to _PRT on that hmm

#

_PRT exists a level above, in PEG1

#

oh fuck i probably know why

#

📎 thor

#

try this

#

or actually no

#

dont

#

ill make a better version

#

📎 thor

#

here

brittle mauve Mar 17, 2024, 10:09 PM

#

nice, now it works (other than unhandled minor usb revision but thats unrelated)

#

shutdown worked and also it powered up when I pressed a key so thats nice

native prairie Mar 17, 2024, 10:10 PM

#

yeah it was me being a bit on the retarded side when looking for a device matching an address

#

great to hear

#

did it get to desktop or no because usb?

brittle mauve Mar 17, 2024, 10:10 PM

#

no because it died inside posix

upbeat smelt Mar 17, 2024, 10:10 PM

#

usb 3.2 too?

native prairie Mar 17, 2024, 10:10 PM

#

i see

brittle mauve Mar 17, 2024, 10:10 PM

#

yeah I guess its usb 3.2

storm plover Mar 17, 2024, 10:10 PM

#

usb 3.2 fails, shutdown works tho

native prairie Mar 17, 2024, 10:11 PM

#

okay dammn

#

@upbeat smelt imagine on how many computers this could work if xhci worked

brittle mauve Mar 17, 2024, 10:11 PM

#

shouldn't be that hard to add the desc tho like I already did for 3.1

native prairie Mar 17, 2024, 10:11 PM

#

so we got like at least 6(?) I think different computers tested

native prairie Mar 17, 2024, 10:12 PM

#

brittle mauve shouldn't be that hard to add the desc tho like I already did for 3.1

dont worry it doesnt work for other revisions as well

#

@brittle mauve and on these computers lai died for some reason?

brittle mauve Mar 17, 2024, 10:13 PM

#

on the desktop I haven't tried lai but on the laptop it didn't shut it down when pressing the power button (and uacpi didn't yet work because of the duplicated node thing)

native prairie Mar 17, 2024, 10:13 PM

#

right

#

lai supports duplicated nodes?

#

thats interesting

#

it supports that yet doesnt support some core aml

upbeat smelt Mar 17, 2024, 10:14 PM

#

brittle mauve shouldn't be that hard to add the desc tho like I already did for 3.1

yeah if you could add the 3.2 descriptor that'd be greatly appreciated

#

i think that one should be the last one

native prairie Mar 17, 2024, 10:15 PM

#

native prairie so we got like at least 6(?) I think different computers tested

two from mint where on one of them lai would die

trim eagle Mar 17, 2024, 10:28 PM

#

@native prairie if you provide me the latest and greatest test image after all fixes are done (if you have any) I’ll test on a few laptops tomorrow

#

Maybe fire up the old desktop too

brittle mauve Mar 17, 2024, 10:30 PM

#

@native prairie ping me once you have something for the duplicated node thing, I am going to go sleep now but I can test again tomorrow

native prairie Mar 17, 2024, 10:31 PM

#

trim eagle <@353593948523986945> if you provide me the latest and greatest test image after...

Ill send one tomorrow morning

trim eagle Mar 17, 2024, 10:31 PM

#

Sounds good

native prairie Mar 17, 2024, 10:31 PM

#

brittle mauve <@353593948523986945> ping me once you have something for the duplicated node th...

Sure, we'll see what I can come up with

native prairie Mar 17, 2024, 10:32 PM

#

trim eagle Sounds good

Thanks

brittle mauve Mar 17, 2024, 10:34 PM

#

also I made a pr that should fix the usb thing so it should be possible to boot it to the desktop on pcs with 3.2 too (though ig there isn't really anything related to the acpi part that you could do there)

native prairie Mar 17, 2024, 10:34 PM

#

Ill poll them in if they get merged

#

Although we weren't able to get xhci working on any tested computer today

#

Even those without 3.2

#

Really hoping qookie figures it out

trim eagle Mar 17, 2024, 10:40 PM

#

Quirks. Quirks everywhere

#

We really should focus on stability and bugfixes for a bit

#

And stuff like make our implementation more compliant

native prairie Mar 17, 2024, 10:41 PM

#

Yup

trim eagle Mar 17, 2024, 10:42 PM

#

Which I guess we’re doing somewhat (readlinkat, inotify and recently some socket memes from me, acpi from you, usb 3.2 from qwinci) but for some reason I feel like KDE isn’t it meme (but I will continue debugging that and InitWare tomorrow)

native prairie Mar 17, 2024, 10:42 PM

#

Isn't it in what sense?

native prairie Mar 18, 2024, 6:12 AM

#

📎 image.tar.xz

#

@trim eagle this should have all the latest fixes

trim eagle Mar 18, 2024, 8:31 AM

#

native prairie Isn't it in what sense?

Isn’t focussing on stability and bugfixes but on porting new shit

native prairie Mar 18, 2024, 8:33 AM

#

I see

brittle mauve Mar 18, 2024, 8:46 AM

#

also disk io performance should be improved somehow KEKW because then wine would probably timeout less often and ig it would be nice otherwise too

trim eagle Mar 18, 2024, 8:52 AM

#

Yep. Performance in general tho

#

Not just disk

brittle mauve Mar 18, 2024, 8:52 AM

#

yeah

trim eagle Mar 18, 2024, 8:53 AM

#

Mind you, probably doesn’t help that a bunch of packages are compiled in debug mode with no optimisations

#

Including the likes of GTK and QT

brittle mauve Mar 18, 2024, 8:53 AM

#

aren't most of them built with relwithdebinfo?

trim eagle Mar 18, 2024, 8:53 AM

#

All my ports default to debug

brittle mauve Mar 18, 2024, 8:54 AM

#

ah I guess its just some of them then

trim eagle Mar 18, 2024, 8:54 AM

#

Autotools does what autotools does

#

Meson and cmake I set to debug

vestal sapphire Mar 18, 2024, 10:00 AM

#

Maybe we should do debugoptimized instead

fervent zodiac Mar 18, 2024, 10:01 AM

#

compile everything with -Ofast -flto -s -fomit-frame-pointer to prevent assertions

safe glen Mar 18, 2024, 11:57 AM

#

-funroll-loops

native prairie Mar 18, 2024, 12:50 PM

#

@brittle mauve will you be able to retest with duplicate node support today?

brittle mauve Mar 18, 2024, 12:50 PM

#

sure

#

did you already make a kernel with the fixes?

native prairie Mar 18, 2024, 12:53 PM

#

no but i can do that real quick if you're up for testing rn

brittle mauve Mar 18, 2024, 12:54 PM

#

yeah I could test it if you send it

native prairie Mar 18, 2024, 12:54 PM

#

ok great

#

📎 thor

brittle mauve Mar 18, 2024, 1:07 PM

#

now it boots, but the power button doesn't work (and I don't really know what the log says because the kernel log is invisible after it boots into kmscon/weston

native prairie Mar 18, 2024, 1:08 PM

#

Probably because its an aml power button object

#

Does it work with lai?

brittle mauve Mar 18, 2024, 1:09 PM

#

no

native prairie Mar 18, 2024, 1:09 PM

#

Actually I have your aml dump I can just check

native prairie Mar 18, 2024, 1:09 PM

#

brittle mauve no

No as in boots fine but no power button?

brittle mauve Mar 18, 2024, 1:09 PM

#

yes

native prairie Mar 18, 2024, 1:09 PM

#

I see

brittle mauve Mar 18, 2024, 1:10 PM

#

well I don't really know if the presses work but at least it doesn't shut it down

native prairie Mar 18, 2024, 1:10 PM

#

And on your pc it literally boots into weston?

nocturne tide Mar 18, 2024, 1:10 PM

#

day 128731273 of waiting for someone to implement kernel log buffer

#

meme

native prairie Mar 18, 2024, 1:10 PM

#

Ikr

#

If it does you can Ctrl alt delete to shutdown I think

native prairie Mar 18, 2024, 1:12 PM

#

brittle mauve well I don't really know if the presses work but at least it doesn't shut it dow...

Yeah I made it so it dies if shutdown call fails so you would've known

brittle mauve Mar 18, 2024, 1:18 PM

#

native prairie If it does you can Ctrl alt delete to shutdown I think

it can boot to kmscon but to weston it doesn't because theres some weird resolution issue where it somehow ends up creating a gfx buffer with resolution 1920x1088 when the actual fb res is 1920x1080)

native prairie Mar 18, 2024, 1:18 PM

#

brittle mauve it can boot to kmscon but to weston it doesn't because theres some weird resolut...

            Control Method Power Button (V1) : 1
            Control Method Sleep Button (V1) : 1

#

yeah your PC doesnt have a fixed event power button

brittle mauve Mar 18, 2024, 1:19 PM

#

hmm

native prairie Mar 18, 2024, 1:19 PM

#

u have this tho

Scope (_SB)
{
    Device (PWRB)
    {
        Name (_HID, EisaId ("PNP0C0C") /* Power Button Device */)  // _HID: Hardware ID

#

I could try hacking together a handler for this

#

so we can enable the corresponding GPE and handle it

#

unfortunately your power button is routed in a very modern way

#

Method (_Q33, 0, NotSerialized)  // _Qxx: EC Query, xx=0x00-0xFF
{
    Debug = "=====QUERY_33====="
    P80H = 0x33
    ADBG ("EC PB press")
    PWPR ()
}

Method (_Q34, 0, NotSerialized)  // _Qxx: EC Query, xx=0x00-0xFF
{
    Debug = "=====QUERY_34====="
    P80H = 0x34
    ADBG ("EC PB release")
    PWRR ()
}

#

you have to have an embedded controller driver, then read the query, then schedule the handler to run

#

and then that notifies the system that power button was pressed

#

or maybe it works without an EC as well but we have no kernel logs so no way to check if a GPE gets generated

brittle mauve Mar 18, 2024, 1:32 PM

#

I think a headless launch should keep the kernel log available

#

Ill try

native prairie Mar 18, 2024, 1:33 PM

#

Device (EC0)
{
    Name (_HID, EisaId ("PNP0C09") /* Embedded Controller Device */)  // _HID: Hardware ID
    Name (_GPE, 0x6E)  // _GPE: General Purpose Events

most likely GPE 0x6E gets generated

#

which uacpi doesnt enable because it has no handler

#

i should really write an EC driver

#

Also your PC recommends PM Profile : 02 [Mobile] mobile power profile for some reason lmao

brittle mauve Mar 18, 2024, 1:37 PM

#

well its a laptop

native prairie Mar 18, 2024, 1:37 PM

#

oh

brittle mauve Mar 18, 2024, 1:40 PM

#

brittle mauve I think a headless launch should keep the kernel log available

apparently not, well I guess that makes sense as udev is launched before the launch argument is checked so gfx claims the device and the output is disabled

native prairie Mar 18, 2024, 1:41 PM

#

the solution is to buy a standalone xhci controller, plug that in, boot off of that and let xhci driver crash

#

then u see logs

#

LULW

wary wing Mar 18, 2024, 1:45 PM

#

sounds like a plan

brittle mauve Mar 18, 2024, 1:49 PM

#

or remove the udev rules for vga, that looks to have worked

#

well theres still nothing to be seen when the button is pressed but at least the log is showed

native prairie Mar 18, 2024, 1:50 PM

#

yeah so no fallback routing for the button if no ec support

#

oh well

brittle mauve Mar 18, 2024, 1:51 PM

#

I wonder if the lid open/close event is also using the ec

native prairie Mar 18, 2024, 1:51 PM

#

lets see

#

sure enough

Method (_Q15, 0, NotSerialized)  // _Qxx: EC Query, xx=0x00-0xFF
{
    Debug = "=====QUERY_15====="
    P80H = 0x15
    If (IGDS)
    {
        Local0 = LIDF /* \_SB_.PC00.LPCB.EC0_.LIDF */
        Local0 = ~Local0
        Local0 &= One
        Local0 += 0x02
        If (^^^GFX0.GLID (Local0))
        {
            ^^^GFX0.CLID |= 0x80000000 /* External reference */
        }
    }

    Notify (LID0, 0x80) // Status Change
}

brittle mauve Mar 18, 2024, 1:58 PM

#

what actually calls that method?

native prairie Mar 18, 2024, 1:58 PM

#

Also just tested on my main pc, also boots and shuts down just fine, but dies because USB revision

native prairie Mar 18, 2024, 2:02 PM

#

brittle mauve what actually calls that method?

The kernel enumerates the namespace, sees an EC device, calls its _GPE method to find the index of the event that it generates what it has outstanding queries, the kernel then enables that event and installs a handler. Then once that event occurs, it reads a register in EC space called EC_QUERY, that returns the index of the event handler in AML to execute, e.g 0x15 for your lid, then the kernel schedules EC. _Q15 for execution, then that handler notifies the kernel of a status change of lid device by doing Notify(Lid, 0x80)

brittle mauve Mar 18, 2024, 2:04 PM

#

native prairie The kernel enumerates the namespace, sees an EC device, calls its _GPE method to...

interesting

#

makes sense tho

native prairie Mar 18, 2024, 2:04 PM

#

brittle mauve interesting

yeah its quite involved as you can see

shell timber Mar 18, 2024, 6:08 PM

#

safe glen `-funroll-loops`

fun roll loops!!!!

woven escarp Mar 19, 2024, 3:19 AM

#

native prairie

Tried it on my laptop and it seems to work fine!

#

Though it defaults to using /dev/sda as the root

#

I wonder if there's some kind of way to change it to /dev/sdc since I'm booting off USB

native prairie Mar 19, 2024, 5:34 AM

#

woven escarp Tried it on my laptop and it seems to work fine!

Cool, did it work with lai before?

woven escarp Mar 19, 2024, 5:41 AM

#

Haven't tested it yet

native prairie Mar 19, 2024, 5:47 AM

#

By work fine you mean it boots into desktop?

#

That would be a first

#

Usually it dies in some driver

nocturne tide Mar 19, 2024, 7:44 AM

#

he said it doesn’t fully boot though

#

“waiting for /dev/sda” means the usb drive was not fully detected

#

it’s not actually looking for /dev/sda to my knowledge, just for a managarm root partition on any drive (that includes usb drives if they are detected)

#

but don’t quote me on that, check the source code yourself if you want to know

#

nevermind, it actually does specifically look for /dev/sda0... interesting

#

i could have sworn i've seen it look for a partition with a specific guid at some point

#

https://github.com/managarm/managarm/blob/f9515870a734758944a9dc4699c77a6cbbaa8eb9/drivers/libblockfs/src/libblockfs.cpp#L822-L829 i wasn't going crazy

trim eagle Mar 19, 2024, 7:51 AM

#

Pretty sure we do look for GUID

trim eagle Mar 19, 2024, 7:51 AM

#

nocturne tide it’s not actually looking for /dev/sda to my knowledge, just for a managarm root...

I saw that. Mana game partition

#

meme

nocturne tide Mar 19, 2024, 7:52 AM

#

ios autocorrect fucking me over

trim eagle Mar 19, 2024, 7:52 AM

#

Relatable

nocturne tide Mar 19, 2024, 7:52 AM

#

like how do they not know who mana game is

#

managram*

#

managarm*

#

:)

trim eagle Mar 19, 2024, 7:52 AM

#

Tho by now I’ve forced it do to it correctly. Managarm Managarm Managarm

#

Capitalisation it sometimes forgets tho

fervent zodiac Mar 19, 2024, 7:52 AM

#

mangranam

nocturne tide Mar 19, 2024, 7:53 AM

#

pain

trim eagle Mar 19, 2024, 7:53 AM

#

.!t managarm

wintry pivotBOT Mar 19, 2024, 7:53 AM

#

It's Managarm, not Managram

trim eagle Mar 19, 2024, 7:53 AM

#

*Or any other spellings

fervent zodiac Mar 19, 2024, 7:53 AM

#

shouldve put that in the tag ig

trim eagle Mar 19, 2024, 7:53 AM

#

Ig time to update the tag

fervent zodiac Mar 19, 2024, 7:54 AM

#

but honestly you cant even tell

native prairie Mar 19, 2024, 8:09 AM

#

nocturne tide “waiting for /dev/sda” means the usb drive was not fully detected

i like how you said "not fully detected" caret

#

aka the driver fucking died

nocturne tide Mar 19, 2024, 8:10 AM

#

very much possible, hard to tell without a full log

native prairie Mar 19, 2024, 8:10 AM

#

@woven escarp can you show us how it died

woven escarp Mar 19, 2024, 8:11 AM

#

nocturne tide “waiting for /dev/sda” means the usb drive was not fully detected

I believe its actually detected

#

I saw along the lines of "mounted /dev/sdc"

#

Which most likely refer to the USB drive

nocturne tide Mar 19, 2024, 8:11 AM

#

the only message similar to that is "mounted /dev/sda0"

native prairie Mar 19, 2024, 8:12 AM

#

does your laptop have xhci?

woven escarp Mar 19, 2024, 8:12 AM

#

Yes

nocturne tide Mar 19, 2024, 8:12 AM

#

in which case it's even more interesting as to how it failed

native prairie Mar 19, 2024, 8:12 AM

#

can you actually take the photo of the logs

nocturne tide Mar 19, 2024, 8:12 AM

#

usb is one of the main failure points in managarm to my knowledge

native prairie Mar 19, 2024, 8:12 AM

#

yup

woven escarp Mar 19, 2024, 8:12 AM

#

My laptop has two drives. One is an SSD and another is HDD

#

They're probably mounted as /dev/sda and /dev/sdb in Managarm

#

So the USB would be /dev/sdc

trim eagle Mar 19, 2024, 8:13 AM

#

No. Unless you have a managarm@root partition there I doubt it. We only mount a partition with the managarm GUID

#

Which was randomly generated by us

woven escarp Mar 19, 2024, 8:14 AM

#

So any partition in all available drives with the Managarm GUID?

nocturne tide Mar 19, 2024, 8:14 AM

#

yes

trim eagle Mar 19, 2024, 8:14 AM

#

The first one it finds I think

nocturne tide Mar 19, 2024, 8:14 AM

#

https://github.com/managarm/managarm/blob/f9515870a734758944a9dc4699c77a6cbbaa8eb9/drivers/libblockfs/src/libblockfs.cpp#L827-L829

native prairie Mar 19, 2024, 8:14 AM

#

woven escarp They're probably mounted as /dev/sda and /dev/sdb in Managarm

...or you could take a photo of logs

nocturne tide Mar 19, 2024, 8:15 AM

#

well yes, the first one it finds

#

that is right

trim eagle Mar 19, 2024, 8:15 AM

#

Mounting the root is quite hardcoded. We want to just pass the partition uuid on the cmdline like Linux but that’s currently not the case

nocturne tide Mar 19, 2024, 8:15 AM

#

if you have one on your internal drive and it gets iterated before the usb drive does, then that's the one it will find

native prairie Mar 19, 2024, 8:15 AM

#

@brittle mauve i'll try to hack together an experimental EC driver for managarm today we'll see if it picks up your power button presses

trim eagle Mar 19, 2024, 8:15 AM

#

The block stack needs some improvements in general (just like the usb stack)

fervent zodiac Mar 19, 2024, 8:16 AM

#

wait

#

does managarm handle oom?

native prairie Mar 19, 2024, 8:16 AM

#

no

nocturne tide Mar 19, 2024, 8:16 AM

#

define handle

#

meme

fervent zodiac Mar 19, 2024, 8:16 AM

#

or is it just like "oops well no ram :/ HEL_CHECK(false)"

native prairie Mar 19, 2024, 8:16 AM

#

yes

#

it just dies

fervent zodiac Mar 19, 2024, 8:16 AM

#

lmaoo

nocturne tide Mar 19, 2024, 8:17 AM

#

that is one way to handle an oom

woven escarp Mar 19, 2024, 8:17 AM

#

nocturne tide if you have one on your internal drive and it gets iterated before the usb drive...

Since it's iterated before the USB, do I need to copy the Managarm partition from the image to my hard drive?

native prairie Mar 19, 2024, 8:17 AM

#

fervent zodiac lmaoo

I was surprised as well

fervent zodiac Mar 19, 2024, 8:17 AM

#

nocturne tide that is one way to handle an oom

i wouldnt call it handling

#

more like "running away from the problem"

native prairie Mar 19, 2024, 8:17 AM

#

thats like a userspace shitty command line app oom handling

nocturne tide Mar 19, 2024, 8:17 AM

#

woven escarp Since it's iterated before the USB, do I need to copy the Managarm partition fro...

does your internal drive have a managarm root partition?

woven escarp Mar 19, 2024, 8:17 AM

#

no

fervent zodiac Mar 19, 2024, 8:17 AM

#

native prairie thats like a userspace shitty command line app oom handling

in the block device stack

native prairie Mar 19, 2024, 8:18 AM

#

worse, in the kernel

#

block device stack is in usersapce

fervent zodiac Mar 19, 2024, 8:18 AM

#

oh the kernel dies on oom?

#

oh jesus christ

native prairie Mar 19, 2024, 8:18 AM

#

...yes

woven escarp Mar 19, 2024, 8:18 AM

#

woven escarp no

i want to partition my drive to have the managarm root partition

fervent zodiac Mar 19, 2024, 8:18 AM

#

what are you even doing that allocates

nocturne tide Mar 19, 2024, 8:18 AM

#

why would you do that

native prairie Mar 19, 2024, 8:18 AM

#

the kernel has all sorts of dynamic arrays @fervent zodiac

#

including storing allocated memory

fervent zodiac Mar 19, 2024, 8:18 AM

#

native prairie the kernel has all sorts of dynamic arrays <@696410219663851552>

theres your problem

native prairie Mar 19, 2024, 8:19 AM

#

not my problem LULW

fervent zodiac Mar 19, 2024, 8:19 AM

#

either handle errors on that

#

or dont use them :^)

native prairie Mar 19, 2024, 8:19 AM

#

im just here to provide a stable acpi implemetation

fervent zodiac Mar 19, 2024, 8:19 AM

#

yeah ik

native prairie Mar 19, 2024, 8:19 AM

#

that handles oom caret

fervent zodiac Mar 19, 2024, 8:19 AM

#

does it really handle it?

native prairie Mar 19, 2024, 8:19 AM

#

yep, every possible allocation is checked and propagated

fervent zodiac Mar 19, 2024, 8:19 AM

#

what happens if you randomly fail some allocations

native prairie Mar 19, 2024, 8:20 AM

#

if it happens in the interpreter it cleans up anything that might've been allocated and returns UACPI_STATUS_OUT_OF_MEMORY

fervent zodiac Mar 19, 2024, 8:20 AM

#

also wait, does this mean that aml can run halfway through

native prairie Mar 19, 2024, 8:20 AM

#

yup

nocturne tide Mar 19, 2024, 8:20 AM

#

is uacpi on managarm running in userspace or kernel?

trim eagle Mar 19, 2024, 8:20 AM

#

Kernel

native prairie Mar 19, 2024, 8:20 AM

#

kernel ofc, just like lai

fervent zodiac Mar 19, 2024, 8:20 AM

#

native prairie yup

what happens if it dies while holding locks

native prairie Mar 19, 2024, 8:20 AM

#

the interpreter unlocks them when exiting

trim eagle Mar 19, 2024, 8:21 AM

#

But yeah OOM handling is also something we need to propagate

fervent zodiac Mar 19, 2024, 8:21 AM

#

okay at least thats good

nocturne tide Mar 19, 2024, 8:21 AM

#

not that obvious, lil can run in userspace :^)

native prairie Mar 19, 2024, 8:21 AM

#

i've specifically written it with oom handling in mind so every single allocation has a check and a way to gracefully handle failure

trim eagle Mar 19, 2024, 8:21 AM

#

Again, the list of things to do is massive. And my time unfortunately not so. But we’re making some nice improvements in the past week or so. Quite happy with development atm

fervent zodiac Mar 19, 2024, 8:22 AM

#

native prairie i've specifically written it with oom handling in mind so every single allocatio...

you might want to block on oom

#

instead of dying

#

like go to sleep and maybe hopefully that makes things better

native prairie Mar 19, 2024, 8:22 AM

#

i dont control that, the kernel is free to do it in the allocation callback

nocturne tide Mar 19, 2024, 8:22 AM

#

also @trim eagle do you happen to know what kills performance in managarm?

#

the whole thing is awfully slow at times, even without debug prints

native prairie Mar 19, 2024, 8:22 AM

#

on linux GFP_KERNEL is basically almost infalliable

nocturne tide Mar 19, 2024, 8:23 AM

#

ls -la /proc/self/fd takes like >1 second to complete

trim eagle Mar 19, 2024, 8:23 AM

#

The entire thing being unoptimized in most places, and several key userspace components running with debug mode

trim eagle Mar 19, 2024, 8:23 AM

#

nocturne tide ls -la /proc/self/fd takes like >1 second to complete

Hmm. That’s just a regular fs operation. 1 second is very excessive even for us I think

#

@carmine current might know or have ideas what the big points here are

nocturne tide Mar 19, 2024, 8:24 AM

#

fs operations that never touch the physical fs

#

meme

#

does managarm not have a block cache or something

trim eagle Mar 19, 2024, 8:26 AM

#

I sure hope we do. I’d like to think we do. But you’d need to check source to know

nocturne tide Mar 19, 2024, 8:26 AM

#

yeah i cant be bothered to read through this code anymore

#

god damn phd code

trim eagle Mar 19, 2024, 8:26 AM

#

Not now at least

#

We all know what happens at the next managarm development session

#

Which might well be tonight lol

nocturne tide Mar 19, 2024, 8:27 AM

#

initware

native prairie Mar 19, 2024, 8:27 AM

#

Kernel log ring?

nocturne tide Mar 19, 2024, 8:27 AM

#

!!

trim eagle Mar 19, 2024, 8:27 AM

#

native prairie Kernel log ring?

That’s Leo’s thing and he’s waiting on the mbus refractor last I heard

trim eagle Mar 19, 2024, 8:27 AM

#

nocturne tide initware

That said, have you seen the screenshot in managarm general?

native prairie Mar 19, 2024, 8:27 AM

#

ok next coolest thing: KDE

nocturne tide Mar 19, 2024, 8:28 AM

#

managarm general no..

trim eagle Mar 19, 2024, 8:28 AM

#

Related to that yes

nocturne tide Mar 19, 2024, 8:28 AM

#

let me take a look

trim eagle Mar 19, 2024, 8:28 AM

#

native prairie ok next coolest thing: KDE

Current path is InitWare for well, init, get SDDM working, boot KDE

native prairie Mar 19, 2024, 8:28 AM

#

yeah i saw the thread, cool work so far

nocturne tide Mar 19, 2024, 8:28 AM

#

oh the same thing that you posted in voice yesterday

#

nice work :^)

trim eagle Mar 19, 2024, 8:28 AM

#

If only it worked consistently (or at all meme )

nocturne tide Mar 19, 2024, 8:29 AM

#

any idea whats up with the input?

trim eagle Mar 19, 2024, 8:30 AM

#

My best guess is eudev being old and not a fan of booting via systemd / InitWare as a bunch of shit was nuked from it. Udev usually reports via sd-notify that it’s ready to do shit but eudev doesn’t have that anymore

#

So I’ll probably look at replacing eudev

#

It’s unmaintained anyway

carmine current Mar 19, 2024, 8:31 AM

#

trim eagle Hmm. That’s just a regular fs operation. 1 second is very excessive even for us ...

Yeah one second sounds really excessive

#

Maybe look at the posix trace and check if there's anything strange

#

This should be just a few calls to posix, right? Like a dozen or so

trim eagle Mar 19, 2024, 8:32 AM

#

Probably

carmine current Mar 19, 2024, 8:32 AM

#

And posix can handle many more than that in a sec

trim eagle Mar 19, 2024, 8:32 AM

#

Yep

nocturne tide Mar 19, 2024, 8:33 AM

#

#

TE_Shrug

#

how to do the trace thingy

trim eagle Mar 19, 2024, 8:33 AM

#

Debug-options.hpp enable logRequests and rebuild system

#

Warning, will@spam your log

#

Might wanna redirect to file and inspect after the run

nocturne tide Mar 19, 2024, 8:33 AM

#

i expect nothing less than that

#

i got a big scollback buffer

#

no worries

#

meme

#

im only interested in one part of the log anyway which shouldnt be too big

trim eagle Mar 19, 2024, 8:34 AM

#

Aight. Enjoy the GET_TID supercall spam

trim eagle Mar 19, 2024, 8:34 AM

#

nocturne tide

Even on / one second. Weird

nocturne tide Mar 19, 2024, 8:37 AM

#

fuck i ls -la'd /proc/self/fd

#

i forgot im on the broken tree i was working on

trim eagle Mar 19, 2024, 8:37 AM

#

meme

#

That shouldn’t be fatal anymore

nocturne tide Mar 19, 2024, 8:37 AM

#

ik ik i just havent updated since then

#

we need to update the pr today lol

trim eagle Mar 19, 2024, 8:38 AM

#

What is wrong?

nocturne tide Mar 19, 2024, 8:39 AM

#

have you not seen qookie's review?

#

posix log for ls -la / https://gist.github.com/48cf/eabc3ee8428ddd1675310ab46b234dff

#

i have no idea what im looking at

#

it doesnt seem to stop at anything for a significant amount of time, so its just a lot of requests and it takes time?

#

huhh... it might be related to this as well: xbstrap: Running ['@SOURCE_ROOT@/scripts/vm-util.py', 'qemu', '--arch', '@OPTION:arch@'] (tools: []) No hardware virtualization available!

#

czapek@raptor-wsl:~/source/managarm/build$ stat /dev/kvm
  File: /dev/kvm
  Size: 0               Blocks: 0          IO Block: 4096   character special file
Device: 0,5     Inode: 81          Links: 1     Device type: 10,232
Access: (0660/crw-rw----)  Uid: (    0/    root)   Gid: (  994/     kvm)
Access: 2024-03-19 09:32:11.769564111 +0100
Modify: 2024-03-19 09:32:11.769564111 +0100
Change: 2024-03-19 09:32:11.769564111 +0100
 Birth: -```

#

¯_(ツ)_/¯

#

the qemu i built seems to be less than capable

#

probably missing some stupid dependency

#

im just gonna reconfigure and install everything i can lol

#

nope i cant get it to work no matter what

#

so annoying

trim eagle Mar 19, 2024, 8:53 AM

#

Grep the qemu source for the error and trace what it wants

trim eagle Mar 19, 2024, 8:53 AM

#

nocturne tide have you not seen qookie's review?

Oh yeah that

nocturne tide Mar 19, 2024, 8:54 AM

#

czapek@raptor-wsl:~/source/managarm/src/ports/qemu$ rg "No hardware virtualization available"
czapek@raptor-wsl:~/source/managarm/src/ports/qemu$```

#

hmm

#

yeah its whatever

median wharf Mar 19, 2024, 8:59 AM

#

nocturne tide does managarm not have a block cache or something

we do

trim eagle Mar 19, 2024, 8:59 AM

#

nocturne tide ``` czapek@raptor-wsl:~/source/managarm/src/ports/qemu$ rg "No hardware virtuali...

Weird

nocturne tide Mar 19, 2024, 9:00 AM

#

maybe its ripgrep picking up parent .gitignore and seeing ports is ignored?

trim eagle Mar 19, 2024, 9:00 AM

#

Oh it might be vm-util.py printing it?

trim eagle Mar 19, 2024, 9:01 AM

#

nocturne tide maybe its ripgrep picking up parent .gitignore and seeing ports is ignored?

Nah I can rg through ports

nocturne tide Mar 19, 2024, 9:01 AM

#

czapek@raptor-wsl:~$ source/managarm/build/tools/host-qemu/bin/qemu-system-x86_64 -enable-kvm
Could not access KVM kernel module: Permission denied
qemu-system-x86_64: failed to initialize kvm: Permission denied```

#

hey now that's something

#

oh im not part of the kvm group

trim eagle Mar 19, 2024, 9:01 AM

#

Are you in the kvm usergroup?

nocturne tide Mar 19, 2024, 9:01 AM

#

:)

trim eagle Mar 19, 2024, 9:01 AM

#

meme

#

Fix that then it should be quicker

nocturne tide Mar 19, 2024, 9:02 AM

#

yep it works now

#

lmao

#

one could say it is much faster...

#

0.192s with posix logs to ls -la /

trim eagle Mar 19, 2024, 9:03 AM

#

I sure hope it’s faster

median wharf Mar 19, 2024, 9:03 AM

#

thousands of lines of phd code just to take 200ms for ls

#

galaxybrain

trim eagle Mar 19, 2024, 9:04 AM

#

ultrameme

nocturne tide Mar 19, 2024, 9:04 AM

#

hey its still pretty good with all the vmexits for the helout

trim eagle Mar 19, 2024, 9:05 AM

#

yep

carmine current Mar 19, 2024, 9:44 AM

#

Well, w/o hardware virt, the perf is obviously going to be bad

#

We don't optimize for software virt

cedar cosmos Mar 19, 2024, 9:46 AM

#

that sucks

trim eagle Mar 19, 2024, 9:47 AM

#

yeah not surprised there

safe glen Mar 19, 2024, 9:47 AM

#

for the record ls on Lyre with TCG is pretty much instant still :^p

#

with 100% less PhD code

#

~~and asserts~~

trim eagle Mar 19, 2024, 9:48 AM

#

there's probably always room for improvement meme

safe glen Mar 19, 2024, 9:48 AM

#

i mean yeah

#

Lyre is still crazy slow compared to Linux

cedar cosmos Mar 19, 2024, 9:50 AM

#

safe glen for the record `ls` on Lyre with TCG is pretty much instant still :^p

But did you do it from a disk image or a RAM disk?

safe glen Mar 19, 2024, 9:50 AM

#

ramdisk

#

but tbh managarm is still slower even off a ramdisk

#

i don't think the bottleneck is storage

native prairie Mar 19, 2024, 10:08 AM

#

safe glen i don't think the bottleneck is storage

IPC halfmemeright

trim eagle Mar 19, 2024, 10:23 AM

#

the IPC should be pretty ok. @carmine current benchmarked that a while back no?

fervent zodiac Mar 19, 2024, 10:30 AM

#

have you tried profiling

#

managarm ipc has fixed requets/reply pairings, right?

#

so copy every ipc request and reply you send and receive to a buffer with a tsc readout

#

and then you can see where its spending time

native prairie Mar 19, 2024, 2:32 PM

#

#1217009725711847465 message

#

@brittle mauve we can now test on your laptop

#

where power button didnt work

#

it should now send notifications

brittle mauve Mar 19, 2024, 2:32 PM

#

nice

native prairie Mar 19, 2024, 2:32 PM

#

also charger, lid, and fn+whatever keys as well

trim eagle Mar 19, 2024, 3:07 PM

#

Based

brittle mauve Mar 19, 2024, 5:13 PM

#

@native prairie could you send the kernel with the ec stuff implemented?

native prairie Mar 19, 2024, 5:15 PM

#

Yup, once I boot into my shitbuntu

#

10 mins

nocturne tide Mar 19, 2024, 5:27 PM

#

are you dualbooting windows and linux for development?

#

if so, i'd recommend you look into wsl

#

wsl v2 is basically a lightweight linux vm that runs a fully functional kernel

#

supports kvm, and you can use a fairly recent kernel (6.1 i believe) if you compile it yourself

#

on windows 11 it also supports wayland applications and runs an xwayland server out of the box

#

i think you could even do pci and usb passthrough with some powershell memes

#

and i know for a fact you can passthrough raw block devices, so it's really got everything you might ever need :^)

native prairie Mar 19, 2024, 5:34 PM

#

i develop in wsl almost exclusively yeah

#

for some heavy stuff like managarm i use ubuntu because wsl io is slow as shit

nocturne tide Mar 19, 2024, 5:34 PM

#

for me its just as if it was running natively

native prairie Mar 19, 2024, 5:34 PM

#

not when accesing files on windows drive

#

through /mnt/X

nocturne tide Mar 19, 2024, 5:35 PM

#

oh yeah that one sucks

native prairie Mar 19, 2024, 5:35 PM

#

ye

nocturne tide Mar 19, 2024, 5:35 PM

#

just use the root partition lol

native prairie Mar 19, 2024, 5:35 PM

#

i use native clion and native git fork

nocturne tide Mar 19, 2024, 5:35 PM

#

oof

native prairie Mar 19, 2024, 5:35 PM

#

so i must use it that way

nocturne tide Mar 19, 2024, 5:35 PM

#

clion supports remote workspaces

#

so you could edit via ssh

#

i use it with vscode like that, except there is an official extension that runs a vscode server on the wsl distro

#

but there is also a vscode extension to do the same over ssh

native prairie Mar 19, 2024, 5:37 PM

#

i see

brittle mauve Mar 19, 2024, 5:37 PM

#

I have used the vscode open remote extension and it mostly worked fine

native prairie Mar 19, 2024, 5:37 PM

#

📎 thor

brittle mauve Mar 19, 2024, 5:38 PM

#

except that things like saving are a little slower ofc

native prairie Mar 19, 2024, 5:38 PM

#

idk im kinda used to have things on my windows drive but i know its not good

#

also my nvme is almost out of space

brittle mauve Mar 19, 2024, 5:38 PM

#

native prairie

thanks, Ill try it

native prairie Mar 19, 2024, 5:38 PM

#

this should spam power button notifications

#

but also try charger and lid ofc

#

and maybe some fn keys if u want

#

then its just a matter of installing a notify handler on the button object from managarm

#

if we want to hook that notify

nocturne tide Mar 19, 2024, 5:39 PM

#

is there a standardized way to install a "power button" notifier?

native prairie Mar 19, 2024, 5:40 PM

#

there are two standard ways

nocturne tide Mar 19, 2024, 5:40 PM

#

are these objects named consistently across acpi implementations?

native prairie Mar 19, 2024, 5:40 PM

#

u never look at the object names, u look at the id returned from _HID (hardware id) or _CID (compatible ids)

#

and yes, they're always the same

#

object names are internal to the aml bytecode

#

they are basically reference handles

nocturne tide Mar 19, 2024, 5:41 PM

#

oh so the _HID and _CID ids are standard and you can basically use those to listen for specific things?

native prairie Mar 19, 2024, 5:41 PM

#

usually u enumerate the entire namespace once its loaded and you have some sort of "bus" where all drivers that support ACPI list their IDs

#

this is how I find an EC device for example uacpi_find_devices("PNP0C09", ...

native prairie Mar 19, 2024, 5:42 PM

#

nocturne tide oh so the _HID and _CID ids are standard and you can basically use those to list...

they're like PCI vendor_id/device_id and other fields combined

#

so basically yes

nocturne tide Mar 19, 2024, 5:43 PM

#

so technically all vendors use the same _HID/_CID for their power buttons and lids?

#

im sorry if im being dense

native prairie Mar 19, 2024, 5:44 PM

#

they can use their own extensions and make a custom device and make a new _HID, but still list the classic power button id in the _CID (aka compatible)

nocturne tide Mar 19, 2024, 5:44 PM

#

but i dont know much about acpi as you can tell

#

ah so _CID is what you want to look for

native prairie Mar 19, 2024, 5:44 PM

#

yeah, but its not present for all devices, so u inspect both and look for a match

#

some devices have just _HID, some have a _CID with like 10 extra entries

#

depends on how simple or complex a device is

brittle mauve Mar 19, 2024, 5:45 PM

#

native prairie and maybe some fn keys if u want

looks like they all work, brightness/sleep/wlan fn keys, power press/release, lid open/close and ac connect/disconnect

native prairie Mar 19, 2024, 5:45 PM

#

daaamn

nocturne tide Mar 19, 2024, 5:45 PM

#

damn that is poggers

#

can we get a pogchamp in the chat

native prairie Mar 19, 2024, 5:45 PM

#

u see it trying to Notify?

nocturne tide Mar 19, 2024, 5:45 PM

#

fr though, good job on this acpi thingy

native prairie Mar 19, 2024, 5:46 PM

#

thank you

nocturne tide Mar 19, 2024, 5:46 PM

#

mr portmaster is going to have a lot of fun porting new stuff

brittle mauve Mar 19, 2024, 5:46 PM

#

native prairie u see it trying to Notify?

yeah, it just says that they have no listeners and thor: EC fired but no SCI event

native prairie Mar 19, 2024, 5:46 PM

#

yeah thats because i disable burst mode

#

and that sends an extra irq

native prairie Mar 19, 2024, 5:46 PM

#

brittle mauve yeah, it just says that they have no listeners and `thor: EC fired but no SCI ev...

ye makes sense

nocturne tide Mar 19, 2024, 5:47 PM

#

holy shit does acpi just work™️?

native prairie Mar 19, 2024, 5:47 PM

#

if we wanted to hook those we would install a notify handler

nocturne tide Mar 19, 2024, 5:47 PM

#

if you have a capable aml parser

native prairie Mar 19, 2024, 5:47 PM

#

yup

native prairie Mar 19, 2024, 5:47 PM

#

nocturne tide if you have a capable aml parser

thats the hard part

nocturne tide Mar 19, 2024, 5:47 PM

#

well of course meme

brittle mauve Mar 19, 2024, 5:48 PM

#

also interestingly it looked like long pressing the power button also generates some kind of different event before it powered off

#

didn't really have time to see what it was

native prairie Mar 19, 2024, 5:48 PM

#

interesting

#

probably "you're about to die"

nocturne tide Mar 19, 2024, 5:48 PM

#

that is fun

brittle mauve Mar 19, 2024, 5:49 PM

#

yeah there's not really anything the os could do about it

#

though ig you could try syncing data to disk really quickly troll

native prairie Mar 19, 2024, 5:49 PM

#

lol

nocturne tide Mar 19, 2024, 5:49 PM

#

eh you can probably try to do some cleanup in the few milliseconds before you get killed

native prairie Mar 19, 2024, 5:49 PM

#

i wonder if os can NAK that

nocturne tide Mar 19, 2024, 5:49 PM

#

hey that would be pretty smart

#

"nah not yet pls"

brittle mauve Mar 19, 2024, 5:50 PM

#

it would also mean that the os could prevent force shutdown completely

native prairie Mar 19, 2024, 5:50 PM

#

yea

nocturne tide Mar 19, 2024, 5:50 PM

#

yeah but for a good reason

#

you can always just cut power completely

#

by taking out the battery

#

:^)

#

nothing they can do about it

brittle mauve Mar 19, 2024, 5:50 PM

#

except when its not removable :p

nocturne tide Mar 19, 2024, 5:51 PM

#

im sure that if you hold the power button long enough it just shuts down with no acpi event

brittle mauve Mar 19, 2024, 5:51 PM

#

though my laptop still has a hole in the bottom that you can push with some screwdriver or whatever to force disconnect the battery

nocturne tide Mar 19, 2024, 5:51 PM

#

if there is something you can cancel of course

native prairie Mar 19, 2024, 5:51 PM

#

anyway pogger result

#

the driver is only 300 loc

nocturne tide Mar 19, 2024, 5:52 PM

#

i wonder how would one control the backlight?

#

since you get notify events when pressing brightness control buttons

native prairie Mar 19, 2024, 5:52 PM

#

acpi has methods for that

#

the kernel can call them

nocturne tide Mar 19, 2024, 5:52 PM

#

like a method defined by aml?

native prairie Mar 19, 2024, 5:52 PM

#

some are non standard and require a separate driver (aka for thinkpads i think)

#

yeah an AML method

nocturne tide Mar 19, 2024, 5:53 PM

#

interesting

#

that seems really easy

native prairie Mar 19, 2024, 5:53 PM

#

nocturne tide Mar 19, 2024, 5:53 PM

#

thats all the standard shit there is?

native prairie Mar 19, 2024, 5:53 PM

#

nocturne tide Mar 19, 2024, 5:53 PM

#

where do you get that from

native prairie Mar 19, 2024, 5:54 PM

#

nocturne tide thats all the standard shit there is?

thats a small portion of standard methods

native prairie Mar 19, 2024, 5:54 PM

#

nocturne tide where do you get that from

https://uefi.org/specs/ACPI/6.5/Apx_B_Video_Extensions.html#bcm-set-the-brightness-level

nocturne tide Mar 19, 2024, 5:54 PM

#

oh thats so cool, thanks!

native prairie Mar 19, 2024, 5:54 PM

#

yeah

#

having a working interpreter allows u to do an absolute shit ton of things

#

many people dont realize it

trim eagle Mar 19, 2024, 5:58 PM

#

nocturne tide mr portmaster is going to have a lot of fun porting new stuff

just acpid and upower I think

nocturne tide Mar 19, 2024, 5:59 PM

#

i wonder how wiring it all up to userspace will look like

#

would thor just expose it all over mbus?

native prairie Mar 19, 2024, 6:00 PM

#

I would think so

nocturne tide Mar 19, 2024, 6:00 PM

#

and let posix just create sysfs entries for all the shit

trim eagle Mar 19, 2024, 6:00 PM

#

Yep. Likely

nocturne tide Mar 19, 2024, 6:01 PM

#

how would one handle the funny "set backlight brightness" with mbus? does it support rpc?

native prairie Mar 19, 2024, 6:01 PM

#

@nocturne tide for example this is a power button device from qwinci's laptop

Device (PWRB)
{
    Name (_HID, EisaId ("PNP0C0C") /* Power Button Device */)  // _HID: Hardware ID

#

in uacpi u would do install_notify_handler(PWRB, handle_button)

upbeat smelt Mar 19, 2024, 6:02 PM

#

nocturne tide how would one handle the funny "set backlight brightness" with mbus? does it sup...

? how do you think any other hardware device is handled

nocturne tide Mar 19, 2024, 6:02 PM

#

i have absolutely no idea

#

also it was a genuine question, i do not know how managarm ipc works at all

#

so i was wondering if you guys have rpc

upbeat smelt Mar 19, 2024, 6:03 PM

#

mbus is only a list of objects which have properties, and, when bound, give back a handle to an ipc lane

nocturne tide Mar 19, 2024, 6:04 PM

#

ah i thought mbus also handled ipc

native prairie Mar 19, 2024, 6:04 PM

#

tbh just starting with ac plug/unplug and battery charge level would be good

upbeat smelt Mar 19, 2024, 6:04 PM

#

then you just construct the request message, send it over the lane, get the response

nocturne tide Mar 19, 2024, 6:05 PM

#

so you could just have a "set backlight brightness" message then i guess

#

that makes sense

upbeat smelt Mar 19, 2024, 6:07 PM

#

we don't do automatic rpc where you only see a function and it automagically generates the code to marshall all the in and out parameters into request and response messages, mainly because korona is not a fan of that idea, and it'd make it potentially harder to add support for other languages

nocturne tide Mar 19, 2024, 6:07 PM

#

yeah i get it now, i was missing some part of how all of that works

#

so im assuming a server would "announce" an object to mbus and a client can then find it and that allows you to create a connection?

trim eagle Mar 19, 2024, 6:09 PM

#

You’d publish it if I recall correctly yeah

upbeat smelt Mar 19, 2024, 6:09 PM

#

yeah you publish an object on mbus, and if the properties match it notifies all pending observers

nocturne tide Mar 19, 2024, 6:10 PM

#

ok yeah i totally dont get how that works so i'll have to read up on that

#

im just being extra dense so dw

trim eagle Mar 19, 2024, 6:10 PM

#

nocturne tide im just being extra dense so dw

No worries. Same

nocturne tide Mar 19, 2024, 6:10 PM

#

when home

#

we need to fix the pr

#

meme

native prairie Mar 19, 2024, 6:11 PM

#

Now we only need for xhci to not suck and managarm will boot on most hardware we tested so far

upbeat smelt Mar 19, 2024, 6:11 PM

#

currently it uses a push model, where the server sends messages to clients which have active observers, but on my mbus refactoring branch i've reworked it to a pull model, where you send a request to the mbus server, and it will respond once there are matching objects (either immediately, or only after new ones are added, controlled by sequence numbers)

#

ignoring the fact you could just not respond to the server messages and get it stuck, the pull model also integrates into async code in a significantly nicer way

nocturne tide Mar 19, 2024, 6:14 PM

#

that sounds like a good change

#

are there any docs on how mbus works?

#

or do i have to read the source code

#

i'll read the source code anyway but i was wondering if there was documentation

native prairie Mar 19, 2024, 6:14 PM

#

theres even a talk

nocturne tide Mar 19, 2024, 6:14 PM

#

https://docs.managarm.org/handbook/sys-arch/mbus/index.html

#

that doesn't say much

#

meme

trim eagle Mar 19, 2024, 6:14 PM

#

nocturne tide we need to fix the pr

Yes. But beer

trim eagle Mar 19, 2024, 6:15 PM

#

native prairie theres even a talk

Multiple. But not on mbus per se

upbeat smelt Mar 19, 2024, 6:15 PM

#

yeah mbus itself is rather undocumented

#

as is the case with most of managarm :^)

#

some day it'd be nice to sit down and document stuff, but there's more pressing issues to work on, and i'm not that good at writing meme

nocturne tide Mar 19, 2024, 6:16 PM

#

how would one create a new server? what kind of an object do i publish? does the server know which object was used to create an ipc lane?

#

im assuming creating a new server is just spawning a new process that publishes objects on mbus somehow

upbeat smelt Mar 19, 2024, 6:18 PM

#

every thread is given the lane to talk to mbus over via ManagarmProcessData (obtained via a supercall)

#

the object you publish is just a set of properties (name-value pairs), and (in the current model) after publishing an object the mbus server gives you a handle to a management lane, on which it sends bind requests, to which you respond by preparing a new stream and sending the handle to a lane back

native prairie Mar 19, 2024, 6:22 PM

#

yeah this is impossible to understand without code examples LULW

nocturne tide Mar 19, 2024, 6:22 PM

#

yeah ngl it is kinda confusing

#

i'll just take a look at posix in a while, maybe i can understand it after staring at it way too long

upbeat smelt Mar 19, 2024, 6:23 PM

#

okay lets take usb as an example

nocturne tide Mar 19, 2024, 6:23 PM

#

mbus definitely does need some docs

upbeat smelt Mar 19, 2024, 6:23 PM

#

when the xhci driver is done setting up a new usb device, it publishes an object for it, so the other drivers can find it

native prairie Mar 19, 2024, 6:23 PM

#

upbeat smelt okay lets take usb as an example

assert(0)

#

LULW

nocturne tide Mar 19, 2024, 6:24 PM

#

so like xhci.device with properties like vid=..., pid=..., subclass=..., if=... whatever

upbeat smelt Mar 19, 2024, 6:24 PM

#

here's the code for that in xhci: https://github.com/managarm/managarm/blob/master/drivers/usb/hcds/xhci/src/main.cpp#L436-L461

native prairie Mar 19, 2024, 6:25 PM

#

and thats supposed to be picked up by a hid driver for example?

nocturne tide Mar 19, 2024, 6:25 PM

#

hm i was close enough

#

im guessing the hid driver then looks for objects that match usb.class/subclass to look for hid devices

upbeat smelt Mar 19, 2024, 6:25 PM

#

native prairie and thats supposed to be picked up by a hid driver for example?

yeah, https://github.com/managarm/managarm/blob/master/drivers/usb/devices/hid/src/main.cpp#L728-L743

native prairie Mar 19, 2024, 6:25 PM

#

okay makes sense

nocturne tide Mar 19, 2024, 6:25 PM

#

somehow that feels really inefficient

upbeat smelt Mar 19, 2024, 6:26 PM

#

nocturne tide im guessing the hid driver then looks for objects that match usb.class/subclass ...

well for usb class/subclass often are 0 because they are defined by the interface

native prairie Mar 19, 2024, 6:26 PM

#

nocturne tide somehow that feels really inefficient

hey at least the kernel doesnt go down along with the driver

upbeat smelt Mar 19, 2024, 6:27 PM

#

nocturne tide somehow that feels really inefficient

how so, the mbus server is only there to pass the lane of the stream between the two threads initially

#

after that they talk directly

nocturne tide Mar 19, 2024, 6:27 PM

#

that filter thing i mean

#

to check whether the object is what you're looking for

#

i can see it getting slow with all those string compares when there is a lot of objects

upbeat smelt Mar 19, 2024, 6:28 PM

#

well it's more efficient than serializing all objects and having the client filter them on it's own

nocturne tide Mar 19, 2024, 6:28 PM

#

unless you're not just mindlessly iterating them all and comparing them

upbeat smelt Mar 19, 2024, 6:28 PM

#

and besides, you usually filter for like 3 properties at most, and i'd guess mbus objects only have <10 properties on average

#

and you can bail out early if one property doesn't match

#

since in the server's view of the object they are in a map

fervent zodiac Mar 19, 2024, 6:29 PM

#

device hotplug cant be that common anyway

native prairie Mar 19, 2024, 6:30 PM

#

also its not the bottleneck of managarm lol

#

nor its core issue atm

nocturne tide Mar 19, 2024, 6:30 PM

#

oh definitely not

#

i just hate slow code in general

#

im wondering how would you approach making acpi available to userspace

#

since it's handled in kernel are you going to have to have an acpi server built into the kernel?

upbeat smelt Mar 19, 2024, 6:32 PM

#

the kernel already exposes an mbus object for rebooting, and dennis was working on extending that to include shutdown as well

native prairie Mar 19, 2024, 6:33 PM

#

you could eventually move the interpreter into userspace

nocturne tide Mar 19, 2024, 6:33 PM

#

so it is sort of a server built into thor

#

i wonder though how would one handle acl so not everyone can reboot the computer

#

with posix you have uid/gid

#

but you probably can't utilize it effectively for that purpose

#

does managarm have some system wide credentials?

upbeat smelt Mar 19, 2024, 6:35 PM

#

threads have a set of kernel-controlled credentials that they can present in ipc messages

#

via ImbueCredentials/ExtractCredentials

native prairie Mar 19, 2024, 6:36 PM

#

lol imbue

upbeat smelt Mar 19, 2024, 6:36 PM

#

nocturne tide i wonder though how would one handle acl so not everyone can reboot the computer

well we'd limit the access to that so only posix can access the mbus object directly

native prairie Mar 19, 2024, 6:36 PM

#

the word i found out about because some c++ helper used it

upbeat smelt Mar 19, 2024, 6:37 PM

#

upbeat smelt well we'd limit the access to that so only posix can access the mbus object dire...

and you have to send a request to it to shut down the machine or whatever

nocturne tide Mar 19, 2024, 6:37 PM

#

ahh that is smart

upbeat smelt Mar 19, 2024, 6:37 PM

#

currently mbus doesn't have access control though, but i'm working on that on my mbus refactoring branch

nocturne tide Mar 19, 2024, 6:37 PM

#

all that phd code finally pays off

upbeat smelt Mar 19, 2024, 6:38 PM

#

to help with that i added token credential objects, which just represent credentials you can send via ImbueCredentials

#

in place of thread credentials

#

the idea is that after publishing an object, you add credentials to the acl for it, and then when someone wants to bind to it they have to send the right credential in the request

#

and to help servers have common credentials, the kernel can have some named ones that it gives to servers, based off of their svrctl configuration

#

so posix-subsystem would list that it needs the power credentials, and then it can include them when it binds to the mbus object

trim eagle Mar 19, 2024, 6:51 PM

#

nocturne tide does managarm have some system wide credentials?

Just thread UID/GID cuz reboot / shutdown from userspace should be (and will be) handled via posix

nocturne tide Mar 19, 2024, 6:51 PM

#

shutdown from userspace can very well be initiated by a random program that just talks to the kernel

trim eagle Mar 19, 2024, 6:51 PM

#

If it runs as root it’s fine meme

nocturne tide Mar 19, 2024, 6:51 PM

#

or at least thats my assumption with how mbus works

#

a random process could very well say "hey i want to bind to reboot"

trim eagle Mar 19, 2024, 6:52 PM

#

Hmm

nocturne tide Mar 19, 2024, 6:52 PM

#

unless there are some checks that i am not aware of

trim eagle Mar 19, 2024, 6:52 PM

#

Yes that would work in the current situation I think

upbeat smelt Mar 19, 2024, 6:52 PM

#

well currently yes, but as i've just said i am working on it

trim eagle Mar 19, 2024, 6:52 PM

#

As mbus doesn’t do ACL or the like yet

#

Yeah

nocturne tide Mar 19, 2024, 6:53 PM

#

yes yes i understand

#

i was just objecting to dennis' reply :^)

trim eagle Mar 19, 2024, 6:53 PM

#

Yeah fair

nocturne tide Mar 19, 2024, 6:53 PM

#

i like the mbus rework acl approach though

#

managarm kernel sounds like a lot of fun

trim eagle Mar 19, 2024, 6:55 PM

#

It is. Same for server

#

Until you go insane on PhD code. Like I went almost 4 years ago

nocturne tide Mar 19, 2024, 6:56 PM

#

yeah, phd code is the not so fun part of it

upbeat smelt Mar 19, 2024, 6:56 PM

#

it's really not that bad

#

:^)

nocturne tide Mar 19, 2024, 6:56 PM

#

but im sure once you spend enough time around it you learn how to read it

#

just like with any project

trim eagle Mar 19, 2024, 6:57 PM

#

Supposedly I should be able to handle it. If I go by the PRs I sometimes can too. Idk how the fuck I’m doing it meme

native prairie Mar 19, 2024, 7:02 PM

#

Whats the phd code meme lmao

#

what code is phd?

trim eagle Mar 19, 2024, 7:04 PM

#

Korona started it. Korona has a PhD. Korona code is PhD code. It’s usually good, but not always the easiest to understand (but that’s also cuz it’s quite advanced sometimes)

native prairie Mar 19, 2024, 7:05 PM

#

lol

nocturne tide Mar 19, 2024, 7:05 PM

#

ill try and fix one managarm issue today on stream

#

assuming dennis is going to come home meme

native prairie Mar 19, 2024, 7:05 PM

#

can you stream on twitch and not discord

#

or both

nocturne tide Mar 19, 2024, 7:05 PM

#

why twitch

native prairie Mar 19, 2024, 7:05 PM

#

i can watch on tv and ipad

#

and other people as well

trim eagle Mar 19, 2024, 7:05 PM

#

Discord stream better stream

nocturne tide Mar 19, 2024, 7:06 PM

#

true

native prairie Mar 19, 2024, 7:06 PM

#

but why

trim eagle Mar 19, 2024, 7:06 PM

#

I’m too shy for twitch stream

trim eagle Mar 19, 2024, 7:06 PM

#

nocturne tide assuming dennis is going to come home <:meme:575445094589661215>

It is planned

#

But

nocturne tide Mar 19, 2024, 7:06 PM

#

i could stream on yt but its not gonna be very entertaining

trim eagle Mar 19, 2024, 7:06 PM

#

native prairie Mar 19, 2024, 7:06 PM

#

well thats kind of said

#

sad

trim eagle Mar 19, 2024, 7:06 PM

#

I might try twitch one day but I like the two way voice comms

nocturne tide Mar 19, 2024, 7:06 PM

#

same

trim eagle Mar 19, 2024, 7:06 PM

#

And on twitch I won’t have those

native prairie Mar 19, 2024, 7:06 PM

#

i mean u can still be in discord voice

nocturne tide Mar 19, 2024, 7:06 PM

#

anyway you can drink beer at home

#

meme

trim eagle Mar 19, 2024, 7:07 PM

#

native prairie i mean u can still be in discord voice

I’ll consider it once again

#

Maybe it is time to try it

native prairie Mar 19, 2024, 7:08 PM

#

that way your vod also stays up

#

it could be educational for a lot of people

trim eagle Mar 19, 2024, 7:08 PM

#

Most of my stream is spamming swear words in logs lol

nocturne tide Mar 19, 2024, 7:08 PM

#

oh i dont think hours of dennis saying stupid shit and typing "kanker" all over the place would be educational

#

meme

native prairie Mar 19, 2024, 7:08 PM

#

lol

trim eagle Mar 19, 2024, 7:08 PM

#

meme

nocturne tide Mar 19, 2024, 7:09 PM

#

if dennis did more coding it could be educational though

trim eagle Mar 19, 2024, 7:09 PM

#

That requires me to have more than one braincells

nocturne tide Mar 19, 2024, 7:09 PM

#

same

#

i tend to do good but my adhd is bad and i cant focus for long periods of time

#

and sometimes i just dont have the braincells laying around

upbeat smelt Mar 19, 2024, 7:10 PM

#

a long while ago i did some streams on twitch of me working on managarm

trim eagle Mar 19, 2024, 7:11 PM

#

I Just don’t have them

upbeat smelt Mar 19, 2024, 7:11 PM

#

nocturne tide i tend to do good but my adhd is bad and i cant focus for long periods of time

well it's nice in that it forces you to focus on something at least

#

so it doesn't end up with you just watching youtube videos on stream or something :^)

trim eagle Mar 19, 2024, 7:12 PM

#

upbeat smelt so it doesn't end up with you just watching youtube videos on stream or somethin...

meme

nocturne tide Mar 19, 2024, 7:12 PM

#

the problem is i often just cant focus at all

#

so i end up forcing myself to do something rather than watch youtube

trim eagle Mar 19, 2024, 7:13 PM

#

nocturne tide the problem is i often just cant focus at all

I’m just too damn stupid

nocturne tide Mar 19, 2024, 7:13 PM

#

sometimes listening to music even distracts me

trim eagle Mar 19, 2024, 7:13 PM

#

Literally negative iq here

nocturne tide Mar 19, 2024, 7:13 PM

#

i really need to start working so i can either get insurance through my employer or afford it myself so i can go to a psychiatrist and get a diagnosis lol

upbeat smelt Mar 19, 2024, 7:14 PM

#

just microdose regular meth instead of prescribed meth

trim eagle Mar 19, 2024, 7:15 PM

#

It doesn’t always get better with a diagnosis. But yeah

nocturne tide Mar 19, 2024, 7:15 PM

#

at least i get the meth legally that way

#

halfmemeright

trim eagle Mar 19, 2024, 7:15 PM

#

True

#

Same

median wharf Mar 19, 2024, 7:54 PM

#

trim eagle Discord stream better stream

wrong

#

twitch stream better

trim eagle Mar 19, 2024, 7:55 PM

#

I’ll consider it

native prairie Mar 20, 2024, 8:35 PM

#

@brittle mauve do u wanna try the final version on your laptop? I just submitted the PR to managarm

#

I hooked up a notify handler to the power button as well

#

so that should work now

brittle mauve Mar 20, 2024, 9:04 PM

#

native prairie <@379250891838193667> do u wanna try the final version on your laptop? I just su...

sure

native prairie Mar 20, 2024, 9:05 PM

#

....and you said that right after i booted into windows

brittle mauve Mar 20, 2024, 9:05 PM

#

well I can just build it myself

native prairie Mar 20, 2024, 9:05 PM

#

yeah actually

#

just grab the PR

brittle mauve Mar 20, 2024, 9:06 PM

#

yeah, Ill have to make a new clone of managarm tho as I only have an aarch64 copy locally and then apply the pr

native prairie Mar 20, 2024, 9:07 PM

#

cant u just clone my fork directly

#

or just u know git remote add uacpi-fork <url>

brittle mauve Mar 20, 2024, 9:09 PM

#

yeah that would work or just git fetch origin pull/642/head:uacpi-rewrite

native prairie Mar 20, 2024, 9:09 PM

#

oh yeah u could do that in github

#

if forgor

brittle mauve Mar 20, 2024, 9:14 PM

#

the pr lacks uacpi itself btw (a wrap or submodule or whatever)

native prairie Mar 20, 2024, 9:15 PM

#

no, its just in a different repo

#

its linked to that pr

trim eagle Mar 20, 2024, 9:16 PM

#

@brittle mauve see the most recent bootstrap PR

#

That should solve your issues

brittle mauve Mar 20, 2024, 9:16 PM

#

oh yeah

#

well I just cloned it there

#

/var/lib/managarm-buildenv/build/tools/system-gcc/lib/gcc/x86_64-managarm/13.2.0/../../../../x86_64-managarm/include/c++/13.2.0/optional:477: _Tp &std::_Optional_base_impl<helix::UniqueDescriptor, std::_Optional_base<helix::UniqueDescriptor, false, false>
>::_M_get() [_Tp = helix::UniqueDescriptor, _Dp = std::_Optional_base<helix::UniqueDescriptor, false, false>]: Assertion 'this->_M_is_engaged()' failed.
posix: Unsupported flag in recvMsg
thor: Illegal POSIX request type 38
HEL_CHECK failed: recv_resp.error()
    In file ../../../src/mlibc/sysdeps/managarm/generic/fork-exec.cpp on line 459

End of lane


thor: Panic in server /usr/bin/netserver
``` I wonder what is this about (this is unrelated to the pr tho)

#

also it seems to prevent it from going to the desktop like it did on aarch64 too

brittle mauve Mar 20, 2024, 9:36 PM

#

native prairie so that should work now

it looks like the power button doesn't generate the ec notify's anymore thonk

#

or at least it doesn't print/do anything

native prairie Mar 20, 2024, 9:36 PM

#

hmm

#

do fn keys work?

brittle mauve Mar 20, 2024, 9:36 PM

#

fn keys and lid open/close do work yeah

native prairie Mar 20, 2024, 9:36 PM

#

lol really strange

#

maybe try rebooting?

brittle mauve Mar 20, 2024, 9:38 PM

#

still the same

native prairie Mar 20, 2024, 9:38 PM

#

like literally nothing gets logged?

brittle mauve Mar 20, 2024, 9:38 PM

#

yeah it doesn't log anything

native prairie Mar 20, 2024, 9:38 PM

#

hmmm

#

i wonder what changed

#

its not like u somehow enable separate events...

#

we can try experimenting tomorrow a bit

#

if u dont mind

#

spam a bit more logs

brittle mauve Mar 20, 2024, 9:40 PM

#

yeah

native prairie Mar 20, 2024, 9:49 PM

#

@brittle mauve I have one simple idea actually, lets try rn

#

I'll send thor

#

📎 thor

brittle mauve Mar 20, 2024, 9:51 PM

#

Ill try that

native prairie Mar 20, 2024, 9:52 PM

#

i moved ec init to post namespace load

#

idk it shouldnt matter but who knows

#

also enabled trace logs

brittle mauve Mar 20, 2024, 9:54 PM

#

it appears to work now

native prairie Mar 20, 2024, 9:54 PM

#

holy shit

#

does it shut down?

brittle mauve Mar 20, 2024, 9:55 PM

#

yes

native prairie Mar 20, 2024, 9:55 PM

#

damn

#

i see your embedded controller is doing tons of shit in the _REG code

#

maybe it relies on that being run after init

#

@brittle mauve and this laptop is the one with duplicated objects right?

brittle mauve Mar 20, 2024, 9:56 PM

#

yeah

native prairie Mar 20, 2024, 9:56 PM

#

okay so you have no ECDT table

#

this means the firmware is not asking for early access to the EC

#

this is probably why

#

ill make it so its only regged early if firmware asked for it

#

Method (_REG, 2, NotSerialized)  // _REG: Region Availability
{
    If ((Arg0 == 0x03))
    {
        OKEC = Arg1
        OSTP ()
        ECFG (One)
        MSFG = (S0ID & One)
        PBOV = (PB1E & One)
    }
}

#

usually this is implemented as a write to one global variable

#

but your aml is doing tons of shit

#

@brittle mauve okay i implemented that properly i think

#

can you verify that it works from the PR head that you built locally?

#

so that we rule out other stuff

#

dont forget to fetch tho

brittle mauve Mar 20, 2024, 10:18 PM

#

yeah it does work

native prairie Mar 20, 2024, 10:18 PM

#

YEsss

#

Amazing

#

thank you

#

lucky guess on my part

#

@brittle mauve btw how old is this laptop?

brittle mauve Mar 20, 2024, 10:24 PM

#

iirc it was released in 2020

native prairie Mar 20, 2024, 10:25 PM

#

interesting

#

do u know the model?

#

also i think i traced why it didnt work, in the _REG method it sets feature bits via SMI and it detects those based on the OS type global variable. That global variable is not yet initialized at that point if we run _REG too early

#

so it sets the barest feature set possible

brittle mauve Mar 20, 2024, 10:29 PM

#

native prairie do u know the model?

acer aspire 5 a515-56g

brittle mauve Mar 20, 2024, 10:29 PM

#

native prairie so it sets the barest feature set possible

interesting

native prairie Mar 20, 2024, 10:29 PM

#

brittle mauve acer aspire 5 a515-56g

cool, thanks

native prairie Mar 23, 2024, 5:11 PM

#

PR to replace LAI with uACPI just got merged!

#

It's now official

quaint fern Mar 23, 2024, 5:22 PM

#

native prairie PR to replace LAI with uACPI just got merged!

incredible...

native prairie Mar 23, 2024, 5:22 PM

#

quaint fern incredible...

lol

#

have you made your decision yet

quaint fern Mar 23, 2024, 5:29 PM

#

native prairie have you made your decision yet

i believe in supporting our community

#

so uacpi is as natural a fit as limine and mlibc for me

#

i will go figure out what all i do with lai (i have been playing with the m68k port of keyronex almost exclusively lately and i need to refamiliarise myself with the amd64 port) and i'm sure uacpi will already provide for it all

native prairie Mar 23, 2024, 5:30 PM

#

this is the way

safe glen Mar 23, 2024, 7:46 PM

#

congratulations on getting uacpi merged

#

i'd probably be rushing to replace lai with it in my OSes if i used any ACPI library at all :p

native prairie Mar 23, 2024, 7:51 PM

#

safe glen congratulations on getting uacpi merged

well thank you!

mighty gust Mar 23, 2024, 9:31 PM

#

safe glen i'd probably be rushing to replace lai with it in my OSes if i used any ACPI lib...

do you have your own ACPI stuff, or do you just not touch ACPI?

safe glen Mar 23, 2024, 9:32 PM

#

latter

mighty gust Mar 23, 2024, 9:34 PM

#

omg based!!!

trim eagle Mar 28, 2024, 1:04 AM

#

Managarm, now booted by InitWare (based on SystemD)

#

I'd like to thank @quaint fern for the help (and @upbeat smelt and @carmine current and @nocturne tide for the debugging today, good session)

nocturne tide Mar 28, 2024, 1:17 AM

#

amazing work 🎉

native prairie Mar 28, 2024, 6:49 AM

#

Nice!

#

KDE when

molten meteor Mar 28, 2024, 3:55 PM

#

interesting

jaunty charm Mar 29, 2024, 8:56 AM

#

Is InitWare still maintained

quaint fern Apr 1, 2024, 8:06 PM

#

I am pleased to announce that an agreement has been made to merge Keyronex and Managarm into a new OS, to be called Keyrogarm.

The tech industry in general and software engineering in particular has a rich history of successes borne of mergers that promote the interests of all parties. Who can forget the golden age for IBM technologies in operating systems after their dealing with Bill Gates? Where would modern UNIX be without the acclaimed Project Monterey, the unification of HP-UX, AIX, and DYNIX/PTX?

Managarm is recognised as the greatest profoundly-asynchronous pragmatic microkernel on the OSDev server. Keyronex is recognised as the greatest operating system, targetting Motorola 68k-series CPUs, that was created within the 2020s.

By leveraging common technologies, not least the asynchronous I/O frameworks, the new Keyrogarm can break ground as the greatest profoundly-asynchronous pragmatic microkernel on the OSDev server targetting Motorola 68k-series CPUs that was created within the 2020s.

Immediate directions for the new kernel include a comprehensive rewrite of C++ components into traditional Objective-C, the immediate prioritisation of support for 68k-family machines by repurposing manpower previously allocated towards porting KDE and driving Intel GPUs, and the replacement of kmscon by systemd-consoled.

trim eagle Apr 1, 2024, 8:07 PM

#

I look forward to this partnership!

native prairie Apr 1, 2024, 8:09 PM

#

quaint fern I am pleased to announce that an agreement has been made to merge Keyronex and M...

Would unironically be cool

upbeat smelt Apr 1, 2024, 8:18 PM

#

looking forward to learning objective-c

fervent zodiac Apr 1, 2024, 9:48 PM

#

quaint fern I am pleased to announce that an agreement has been made to merge Keyronex and M...

most intriguing

#

looking forward to future updates!

#

I assume you will be making the transition through Objective-C++

#

Are you planning to stick to bragi? Or are you going to replace it with a more Objective-C-flavoured interface?

trim eagle Apr 1, 2024, 9:50 PM

#

Obviously it needs replacing

#

We’re also considering rewriting uACPI in rust so we have blazingly fast AML execution

quaint fern Apr 1, 2024, 9:51 PM

#

fervent zodiac Are you planning to stick to bragi? Or are you going to replace it with a more O...

we are reviving NeXTSTEP PortableDistributedObjects

fervent zodiac Apr 1, 2024, 9:51 PM

#

Does it though? I mean it might be possible to adapt even Protocol Buffers and gRPC to work natively with Objective-C

fervent zodiac Apr 1, 2024, 9:51 PM

#

quaint fern we are reviving NeXTSTEP PortableDistributedObjects

I feel like a modern reimplementation of NSDistributedObject's might not be too difficult

quaint fern Apr 1, 2024, 9:52 PM

#

fervent zodiac I feel like a modern reimplementation of NSDistributedObject's might not be too ...

i already have the code to pluck out data and pointers from the parameter & return passing stack/registers on amd64 as well as m68k and dynamically serialise the whole lot

fervent zodiac Apr 1, 2024, 9:52 PM

#

Also, what are your plans wrt security on those objects? I believe that one of the reasons they aren't used in macOS much anymore is because of security issues

fervent zodiac Apr 1, 2024, 9:53 PM

#

quaint fern i already have the code to pluck out data and pointers from the parameter & retu...

Is that fast? I would have thought a more traditional DSL would perform a lot better

quaint fern Apr 1, 2024, 9:53 PM

#

fervent zodiac Also, what are your plans wrt security on those objects? I believe that one of t...

i think the bigger problem is that it's slow as fuck!

#

to briefly explain how it works

fervent zodiac Apr 1, 2024, 9:53 PM

#

yeah i get it

#

you could use an optimized JIT compiler to avoid most of the performance bottlenecks though

quaint fern Apr 1, 2024, 9:55 PM

#

message sent to a proxy object -> method not found handler contacts the remote object to ask for the method signature -> NSInvocation built by consulting stack/registers and interpreting according to the method signature -> serialised and sent to remote -> remote constructs an NSInvocation and invokes it -> remote serialises return value and sends it back -> return deserialised on the other side and stuffed into registers/stack

fervent zodiac Apr 1, 2024, 9:56 PM

#

that sounds ultra slow

quaint fern Apr 1, 2024, 9:56 PM

#

it's evil to have to make a roundtrip just to decide how you're going to serialise

fervent zodiac Apr 1, 2024, 9:56 PM

#

actually, you could do a funny meme with a JIT here

quaint fern Apr 1, 2024, 9:57 PM

#

NSXPC forces you to designate protocols of proxy objects and permits proxy objects to be given only when in a position where the protocol that the proxy object should understand, so it eliminates that roundtrip, and it also goes asynchronous, and instead of return values, you can pass a block as a parameter, and remote can evaluate the block

fervent zodiac Apr 1, 2024, 9:57 PM

#

a method not found handler saves all the registers, does the lookup roundtrip, builds a proper JIT thunk, restores registers and runs the thunk

#

and you bind the JIT thunk retroactively into the class

#

so that you only need to do this once

fervent zodiac Apr 1, 2024, 9:58 PM

#

quaint fern NSXPC forces you to designate protocols of proxy objects and permits proxy objec...

oh so it's basically apple wayland

quaint fern Apr 1, 2024, 10:00 PM

#

fervent zodiac and you bind the JIT thunk retroactively into the class

this is quite dangerously based actually

fervent zodiac Apr 1, 2024, 10:01 PM

#

in fact on all of {arm, x86, riscv} enough codegen for marshalling doesn't even need any fancy anything

#

i mean not even fancy regalloc

#

because ~all registers are ~equal

native prairie Apr 1, 2024, 10:07 PM

#

trim eagle We’re also considering rewriting uACPI in rust so we have blazingly fast AML exe...

Fast and safe

fervent zodiac Apr 1, 2024, 10:09 PM

#

quaint fern this is quite dangerously based actually

yes, its actually what i would do if i had to do this and optimize aggresively for speed

trim eagle Apr 1, 2024, 10:22 PM

#

native prairie Fast _and_ safe

Ah yes how could I forget

#

Oh we were also thinking to compress the IPC data with xz for that amazing security

median wharf Apr 2, 2024, 8:52 AM

#

based

#

i say we link xz into every binary by default on managarm

#

everyone will need compression at some point right?

trim eagle Apr 2, 2024, 8:52 AM

#

Yeah

fervent zodiac Apr 2, 2024, 8:53 AM

#

i propose you link every managarm binary with libsupplychain

#

(which dlopens all dynamic libraries available)

onyx cloud Apr 3, 2024, 3:30 AM

#

fervent zodiac (which dlopens all dynamic libraries available)

ah yea, good idea

#

just in case u wanna use another lib

median wharf Apr 3, 2024, 12:16 PM

#

hey bro I heard you liked shared libraries

onyx cloud Apr 4, 2024, 4:15 AM

#

sharing is caring

#

😊

#

i wanna share my liblzma.so with u

#

😍

storm plover Apr 5, 2024, 3:50 AM

#

onyx cloud i wanna share my liblzma.so with u

Omg, thank you so much, thats what i needed all this time

quaint fern Apr 9, 2024, 4:34 PM

#

let me just say i think it's an outrage that despite a PR open for over a year, managarm has still yet to be added to the http://www.microkernel.info website

median wharf Apr 9, 2024, 6:22 PM

#

there are even 2 prs open lol

trim eagle Apr 9, 2024, 7:14 PM

#

Lol

zenith phoenix Apr 9, 2024, 7:26 PM

#

Looks like maintainer is inactive

#

They don’t have any gh activity this year

vestal sapphire Apr 9, 2024, 8:31 PM

#

~~fork it~~

trim eagle Apr 9, 2024, 8:47 PM

#

Cringe

quaint fern Apr 19, 2024, 6:13 PM

#

vestal sapphire ~~fork it~~

not the worst idea actually

#

you managarmers should establish a secretive organisation cloaked in mystery as to what dark money funds it, to be called the Microkernels Group or something like that

cedar cosmos Apr 24, 2024, 5:06 AM

#

managamers

cedar cosmos Apr 24, 2024, 5:07 AM

#

quaint fern you managarmers should establish a secretive organisation cloaked in mystery as ...

Portable Systems Group

upbeat smelt Jun 3, 2024, 1:19 AM

#

after losing my mind only slightly i got mammogram booting on my laptop off of an usb stick with the usb keyboard working at the same time

#

#

still need to debug why usb hubs don't work, and fix the usb hid driver dying when i give it my mouse (although this i can test in qemu)

#

the laptop touchpad is out of reach for now, i don't feel like writing an i2c driver just yet

native prairie Jun 3, 2024, 8:31 AM

#

upbeat smelt

me when 15 AML blobs, and this is not even including the potential dynamic blobs that arent loaded by managarm yet because no cpu power states

trim eagle Jun 3, 2024, 8:45 AM

#

AML moment

wild stone Jun 4, 2024, 4:03 PM

#

reviving the managarm thread to show how our network stack is actually reasonably fast if you compile everything as release with -O3 halfmemeright

#

its a bit random but i dont know if thats my internet or managarm, the highest i got was 65 mbit/s

#

this is using the minibrowser

trim eagle Jun 4, 2024, 4:04 PM

#

Nice

wild stone Jun 4, 2024, 4:04 PM

#

also, normally i can only test it once or twice before OOMing

#

halfmemeleft

trim eagle Jun 4, 2024, 4:04 PM

#

That’s why I run with 8 gigs of RAM

#

But we should fix that shit yeah

fervent zodiac Jun 4, 2024, 4:23 PM

#

wild stone reviving the managarm thread to show how our network stack is actually reasonabl...

what is it on linux?

#

also

#

how well does it perform if you use comcast --device=whateverbridge --latency=250 --target-bw=1000000 --packet-loss=10%

wild stone Jun 4, 2024, 4:25 PM

#

fervent zodiac what is it on linux?

atm 60 mbit

fervent zodiac Jun 4, 2024, 4:27 PM

#

so you are losing about 20 megabits not bad

fervent zodiac Jun 4, 2024, 4:27 PM

#

fervent zodiac how well does it perform if you use [comcast](<https://github.com/tylertreat/com...

what about that though

wild stone Jun 4, 2024, 4:27 PM

#

looking at how id set that up atm

upbeat smelt Jun 4, 2024, 4:32 PM

#

lol comcast

fervent zodiac Jun 4, 2024, 5:50 PM

#

wild stone looking at how id set that up atm

the docs sound easy enough

fervent zodiac Jun 4, 2024, 5:51 PM

#

upbeat smelt lol comcast

lmfao yeah the only reason i remember the name of the tool

abstract coral Jun 7, 2024, 7:26 PM

#

upbeat smelt after losing my mind only slightly i got mammogram booting on my laptop off of a...

I think managarm wants you to run

upbeat smelt Jun 7, 2024, 7:37 PM

#

abstract coral I think managarm wants you to run

don't worry about it

wild stone Jun 8, 2024, 11:31 PM

#

i wonder if its less omegakanker to get thor booting without eir rather then relocating thor or eir to a different virtual address, since thor want to be loaded where limine would load eir halfmemeright

nocturne tide Jun 9, 2024, 7:11 AM

#

you can ask limine to load eir at a different address btw

wild stone Jun 9, 2024, 8:02 AM

#

nocturne tide you can ask limine to load eir at a different address btw

Not into lower half lol

nocturne tide Jun 9, 2024, 8:03 AM

#

no but you can relocate it above the kernel

vestal sapphire Jun 9, 2024, 8:27 AM

#

I managed to get phone tethering over USB working on managarm

carmine current Jun 9, 2024, 8:32 AM

#

very nice

native prairie Jun 9, 2024, 8:56 AM

#

very nice

shrewd grove Jun 9, 2024, 8:56 AM

#

very nice

nocturne tide Jun 9, 2024, 8:58 AM

#

very nice

odd schooner Jun 9, 2024, 8:58 AM

#

very nice

wild stone Jun 9, 2024, 9:16 AM

#

very nice

dusk crystal Jun 9, 2024, 9:36 AM

#

very nice

median wharf Jun 9, 2024, 10:27 AM

#

very nice

dense umbra Jun 9, 2024, 12:54 PM

#

very nice

forest bear Jun 9, 2024, 1:16 PM

#

nery vice

wary wing Jun 9, 2024, 2:53 PM

#

vicy nere

vagrant thunder Jun 12, 2024, 12:11 AM

#

very nice

civic vortex Jun 12, 2024, 4:28 PM

#

very nice

short gazelle Jun 12, 2024, 6:31 PM

#

very nice

lament grove Jun 12, 2024, 9:58 PM

#

very nice

abstract coral Jun 13, 2024, 10:37 AM

#

very nice

indigo urchin Jun 13, 2024, 10:37 AM

#

very nice

graceful temple Jun 13, 2024, 8:32 PM

#

nery vice

shy minnow Jun 13, 2024, 8:53 PM

#

yere nivc

woven escarp Jun 14, 2024, 12:55 AM

#

very nice

silent oak Jun 14, 2024, 9:20 PM

#

very nice

mossy swallow Jun 14, 2024, 9:42 PM

#

vairie naïsse

wary wing Jun 14, 2024, 10:08 PM

#

ベリーナイス

trim eagle Jun 27, 2024, 6:27 AM

#

Managarm now shows a notification if a network interface gets an address. May not look like much considering it’s just a window, but it’s a complex operation with multiple moving parts (just like it is on Linux). Of course, this goes over netlink, so we need to broadcast that a new address has been assigned to an interface. On Linux, NetworkManager listens for this and does things with it, here it’s netlink-notify which just notices it and sends away a call with libnotify to the notification-daemon over dbus. We chose for lxqt-notificationd, as we were having some issues with notification-daemon (gnome and kde ship their own in the shell). The daemon then uses Qt6 (in debug mode, so it’s quite slow atm) to display the notification. This process goes over the dbus session bus, not the system wide one, so some dbus memes were needed to not hard rely on elogind. All in all a fruitful endeavor as this work can also be used with mail clients notifying you of new mail for example

fervent zodiac Jun 27, 2024, 10:15 AM

#

does the notification itself go over netlink?

#

as in the text?

#

or i guess dbus?

trim eagle Jun 27, 2024, 10:16 AM

#

pretty sure that's all handled over dbus

fervent zodiac Jun 27, 2024, 10:16 AM

#

i see i see

trim eagle Jun 27, 2024, 1:37 PM

#

new irc client just dropped (this is konversation)

wary wing Jun 27, 2024, 1:53 PM

#

yooo

trim eagle Jun 27, 2024, 2:20 PM

#

when paired with the notification daemon work, I can get notified of incoming messages

#

happy with that

lament grove Jun 27, 2024, 5:35 PM

#

Have you ever compiled Managarm on a machine running Managarm?

wild stone Jun 27, 2024, 5:36 PM

#

self hosting does not work on managarm

#

yet meme

lament grove Jun 27, 2024, 5:36 PM

#

meme

lament grove Jun 27, 2024, 6:31 PM

#

BTW: I notice it's always root@, have users not been implemented or is it just configured this way by default?

wild stone Jun 27, 2024, 6:35 PM

#

lament grove BTW: I notice it's always root@, have users not been implemented or is it just c...

iirc you can change to a different user but no checking of any kind of privs exists
havent actually tried changing user now thinking about it

trim eagle Jun 27, 2024, 6:47 PM

#

You can’t (unless you write a program yourself and call setuid). I can (shadow is marked as broken and hacky, but I can do it). If configured correctly (and by that I mean workarounds) it’ll auto login as managarm, but only in kmscon

#

Permission checking is very limited, but I’m 75% sure we check (e)uid and friends on calls to setuid meme

fervent zodiac Jun 27, 2024, 8:53 PM

#

the managarm way

#

write the most half-assed internals possible

#

but we have ports

#

where we just carbon copy linux

trim eagle Jun 27, 2024, 8:54 PM

#

Surprisingly, it works decently

#

That said, there’s a pretty decent base to work on

fervent zodiac Jun 27, 2024, 8:57 PM

#

did yall go through with the strace """port"""

trim eagle Jun 27, 2024, 8:59 PM

#

Nah. But we’ll get one eventually. Needs ptrace tho

vestal sapphire Jun 27, 2024, 9:03 PM

#

porting strace itself it probably ez (if you disregard autotools pain for a sec)

#

ptrace(2) is pain tho

#Managarm and related projects