uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 5

north holly May 12, 2024, 9:18 PM

#

as plain text

zenith crescent May 12, 2024, 9:18 PM

#

the MSDM table contains it

dusky glade May 12, 2024, 9:18 PM

#

what in the actual

north holly May 12, 2024, 9:18 PM

#

in the acpi tables

dusky glade May 12, 2024, 9:18 PM

#

fuck

#

how to make millions:

get thousands of acpi table dumps somehow
extract all ms keys
sell on cdkeys or smthin
profit ???

north holly May 12, 2024, 9:19 PM

#

possibly not legal

dusky glade May 12, 2024, 9:19 PM

#

bru

#

😔

north holly May 12, 2024, 9:20 PM

#

dusky glade how to make millions: 1. get thousands of acpi table dumps somehow 2. extract a...

also some linux repo has a bunch of acpi table dumps

#

from 100s of computers

zenith crescent May 12, 2024, 9:20 PM

#

I don't think it works for keys that are linked to microsoft accounts

north holly May 12, 2024, 9:20 PM

#

mine is

#

and it worked anyway

zenith crescent May 12, 2024, 9:20 PM

#

mine is also, but I meant that I don't think it can be used if its linked

fiery turtle May 12, 2024, 9:20 PM

#

@zenith crescent whats the acpica error for that pc?

zenith crescent May 12, 2024, 9:21 PM

#

fiery turtle <@983999369848193044> whats the acpica error for that pc?

[    0.621519] ACPI BIOS Error (bug): Could not resolve symbol [\_SB.CHRG], AE_NOT_FOUND (20230628/psargs-330)
[    0.621545] ACPI Error: Aborting method \PNOT due to previous error (AE_NOT_FOUND) (20230628/psparse-529)
[    0.621563] ACPI Error: Aborting method \_SB.PCI0.LPCB.EC0._REG due to previous error (AE_NOT_FOUND) (20230628/psparse-529)

fiery turtle May 12, 2024, 9:22 PM

#

I see

#

Brain damage

dusky glade May 12, 2024, 9:22 PM

#

i also get acpi errors

#

with my laptop

#

using acpica

#

idk if its related but

#

is smthing

hollow elm May 12, 2024, 9:23 PM

#

gobless ```
[ 1.086734] ACPI Error: AE_NOT_FOUND, While resolving a named reference package element - _PR_.P000 (20230628/dspkginit-438)

#

there is a suspiciously named \_SB_.PLTF.P000 though :^)

left orbit May 12, 2024, 9:27 PM

#

i think acpica throws an error on my hardware too?

#

but i cant remember

#

and i cant get into my debian install because i have mismatched kernel and nvidia modules halfmemeright

fiery turtle May 12, 2024, 9:33 PM

#

Yeah undefined references in some of these is a classic

#

Of firmware dev brain damage

kindred beacon May 13, 2024, 10:19 AM

#

I'll do that once I test managarm on my test PCs

fiery turtle May 13, 2024, 10:20 AM

#

Thanks!

stable topaz May 13, 2024, 10:59 AM

#

This is a zen 4 system

📎 res.txt

fiery turtle May 13, 2024, 11:02 AM

#

Thanks

tired crater May 13, 2024, 1:55 PM

#

Why does so much stuff around acpi during system initialization fail on some pcs? For example, on linux according to dmesg, it complains about symbols that aren’t resolvable (note: acpi bios error (bug)), complains about invalid values like thermal threshold (here it notes: Firmware Bug), sometimes it fails to parse AML opcodes because they seem to be invalid, duplicate or non existent, …

#

Is this really just to blame on awful firmware (devs)

left orbit May 13, 2024, 2:09 PM

#

i'd assume yes mostly

#

but there are for sure some bugs in ACPICA

fiery turtle May 13, 2024, 2:12 PM

#

tired crater Is this really just to blame on awful firmware (devs)

Yeah

tired crater May 13, 2024, 2:12 PM

#

halfmemeright

fiery turtle May 13, 2024, 2:39 PM

#

tired crater Why does so much stuff around acpi during system initialization fail on some pcs...

Those are basically untested code paths for the most part and none of them are critical

#

Since aml is just a loosely typed language you can compile completely broken code

#

Similar to python

#

And the error won't uncover itself until that code runs

mortal yoke May 13, 2024, 2:40 PM

#

its kinda weird if there are errors in the top level namespace tho, like in one of the test blobs there is an out of bounds createqwordfield and then it just stops the namespace load there

#

but ig because its very much at the end it doesn't really cause that much problems

tired crater May 13, 2024, 2:41 PM

#

fiery turtle Those are basically untested code paths for the most part and none of them are ...

yeah that makes sense
||just put that burden on kernel devs||

fiery turtle May 13, 2024, 2:41 PM

#

mortal yoke its kinda weird if there are errors in the top level namespace tho, like in one ...

Which blob?

#

Thats because NT just ignores the bounds until the field is accessed

#

So on nt thats not an error

loud ice May 13, 2024, 2:42 PM

#

wait what

#

wtf

fiery turtle May 13, 2024, 2:42 PM

#

Yeah

loud ice May 13, 2024, 2:43 PM

#

can aml ask to load more aml? or is that not how it works

mortal yoke May 13, 2024, 2:43 PM

#

fiery turtle Which blob?

notebook_avell_high_performance_c75_c75_rtx_muv-g1750_rtx_muv_b4cc55e338b5.aml line 54359

fiery turtle May 13, 2024, 2:43 PM

#

loud ice can aml ask to load more aml? or is that not how it works

It can load arbitary buffers as aml

#

So yes

loud ice May 13, 2024, 2:43 PM

#

fun!

fiery turtle May 13, 2024, 2:43 PM

#

Thats how advanced features work

#

Like advanced cpu power management

loud ice May 13, 2024, 2:43 PM

#

ah

#

fair

fiery turtle May 13, 2024, 2:44 PM

#

If kernel says it supports it aml just loads extra crap dynamically

fiery turtle May 13, 2024, 2:44 PM

#

mortal yoke notebook_avell_high_performance_c75_c75_rtx_muv-g1750_rtx_muv_b4cc55e338b5.aml l...

Probably not a problem in uACPI since load aborts dont abort initialization entirely, but I check bounds at creation time as well

#

Which is technically not nt compatible

stable topaz May 13, 2024, 2:59 PM

#

is it possible that uacpi is trying to free a null address, or is it more a vmm problem?

loud ice May 13, 2024, 3:00 PM

#

free(null) is legal c

stable topaz May 13, 2024, 3:01 PM

#

Should i just ignore that? Because my vmm is shitting itself when uacpi frees that

loud ice May 13, 2024, 3:02 PM

#

yes

fiery turtle May 13, 2024, 3:07 PM

#

stable topaz is it possible that uacpi is trying to free a null address, or is it more a vmm ...

Yes, all the time

hollow crescent May 13, 2024, 3:08 PM

#

loud ice free(null) is legal c

angr

fiery turtle May 13, 2024, 3:10 PM

#

I thought I had a comment there

#

Maybe not

#

Btw @stable topaz I saw u have some beginnings of rust bindings, would be really nice to upstream that eventually if you get that done

left orbit May 13, 2024, 3:22 PM

#

looking at their crate, it's probably a better idea to use something like bindgen to automatically generate these bindings instead of manually writing them

#

eliminates the chance of you accidentally getting something wrong and is much faster to update, then you could write a proper rust wrapper around that

stable topaz May 13, 2024, 3:24 PM

#

i wanted to do it as a poc for my os, but i can do it the right way so i can upstream it

left orbit May 13, 2024, 3:26 PM

#

@fiery turtle btw since you mentioned that you could add zig support, would you like to have a zig wrapper in-tree? or do you want something similar to uacpi.cmake/meson.build to collect source files so something else can do the wrapping?

#

either way i'll look into wrapping the C api so it's more ziggy

#

i can maintain it out-of-tree with the addition of an in-tree build.zig that will do the source file meme so you don't have to deal with maintaining zig code

fiery turtle May 13, 2024, 3:31 PM

#

left orbit <@353593948523986945> btw since you mentioned that you could add zig support, wo...

In tree is probably better so that we can verify that it doesn't go out of date?

#

If u mean like the bindings

left orbit May 13, 2024, 3:33 PM

#

well so there is basically 2 ways to consume uacpi from zig right now:

translate-c: compiler generates raw api bindings, not very nice to use from zig code
a proper, hand written wrapper: it uses the compiler generated bindings to provide a more user friendy interface, so wrapping raw pointers and making it all feel like proper zig code

#

the first one you won't really have to do anything, as it's done by the compiler, ideally you would only provide a build.zig file similar to uacpi.cmake/meson.build so users can easily add the source files to the compilation unit, ideally in tree so you can update it so downstream users dont have to worry about updating the source list manually

#

the second one is a bit more work, in addition to maintaining the source list you would also need to worry about the standard library changes/compiler changes (keep in mind zig isn't 1.0 lol)

loud ice May 13, 2024, 3:37 PM

#

hollow crescent <:angr:991436925116350494>

it actually is kinda useful

fiery turtle May 13, 2024, 3:38 PM

#

left orbit the second one is a bit more work, in addition to maintaining the source list yo...

Well that one definitely sounds nicer, but yeah I dont know who would maintain it in case of compiler changes etc, because I know nothing about zig

left orbit May 13, 2024, 3:38 PM

#

yeah thats why i suggested out of tree since then i could maintain it

#

but if you want it in tree we can arrange that

fiery turtle May 13, 2024, 3:42 PM

#

left orbit yeah thats why i suggested out of tree since then i could maintain it

You could maintain in tree as well tho but if out of tree works better than sure thats fine as well!

left orbit May 13, 2024, 3:43 PM

#

i do think in tree would be better, just because of the added benefit of you being able to sync small changes between build systems

fiery turtle May 13, 2024, 3:45 PM

#

Yeah, and we could also add a GitHub actions ci check that it builds and stuff

left orbit May 13, 2024, 3:45 PM

#

yep absolutely

left orbit May 13, 2024, 4:18 PM

#

i'm thinking of somehow providing a ziggy way of handling kernel api

loud ice May 13, 2024, 4:19 PM

#

left orbit i'm thinking of somehow providing a ziggy way of handling kernel api

that actually sounds rather hard because of the callouts

left orbit May 13, 2024, 4:19 PM

#

my idea currently is to provide default kernel api that returns everything as unimplemented and having the user export an instance of an API struct from the root package

loud ice May 13, 2024, 4:20 PM

#

left orbit my idea currently is to provide default kernel api that returns everything as un...

you can export symbols comptime, can't you

#

so all you need is @import("root").uacpi_hooks and if comptime (has feature) export()

loud ice May 13, 2024, 4:21 PM

#

loud ice you can export symbols comptime, can't you

yeah of course you can, just need a bit of asm :^)

#

or import memes

left orbit May 13, 2024, 4:21 PM

#

i was thinking of providing the kernel api exports myself and just calling the user provided functions from them

#

makes for a better compile time error when a new kernel api gets added

#

instead of a linker error you get a missing field or something

loud ice May 13, 2024, 4:22 PM

#

left orbit i was thinking of providing the kernel api exports myself and just calling the u...

well yeah

#

wait does uacpi use weak symbols?

left orbit May 13, 2024, 4:22 PM

#

nope

loud ice May 13, 2024, 4:22 PM

#

ah

#

yeah then just do that

left orbit May 13, 2024, 4:23 PM

#

also i want the callbacks to accept my zigified wrappers where needed

#

i dont think there is any api that actually accepts a node or whatever

#

but i do want to wrap around that as well

#

im currently trying to figure out how the module args work so i dont need to export two modules, one regular uacpi and one reduced hw

#

ideally the user can specify that somehow

#

also printf and some string.h functions... idk what to do about those either lol

loud ice May 13, 2024, 4:24 PM

#

left orbit also printf and some string.h functions... idk what to do about those either lol

wait uacpi needs printf?

left orbit May 13, 2024, 4:25 PM

#

yeah

loud ice May 13, 2024, 4:25 PM

#

what the actual fuck

left orbit May 13, 2024, 4:25 PM

#

also memset and friends

#

but i think maybe i can get zig to link in compiler-rt for that?

#

idk

loud ice May 13, 2024, 4:25 PM

#

left orbit also memset and friends

they have to be provided anyway

left orbit May 13, 2024, 4:25 PM

#

nope

loud ice May 13, 2024, 4:25 PM

#

for zig code to work

#

wdym nope

#

llvm requires them

left orbit May 13, 2024, 4:25 PM

#

oh?

#

hmm

loud ice May 13, 2024, 4:25 PM

#

afaik you cant turn that off

#

some of them

#

not memmove

left orbit May 13, 2024, 4:26 PM

#

export fn strlen(s: [*]const c_char) callconv(.C) usize {
export fn strcmp(s1: [*]const c_char, s2: [*]const c_char) callconv(.C) i32 {
export fn strnlen(s: [*]const c_char, maxlen: usize) callconv(.C) usize {
export fn strncmp(s1: [*]const c_char, s2: [*]const c_char, n: usize) callconv(.C) i32 {

#

yeah those no

loud ice May 13, 2024, 4:26 PM

#

imo uacpi should just not depend on these things like what are they even used for

#

like memcpy/set/cmp is fine

#

but the other ones why

left orbit May 13, 2024, 4:27 PM

#

because most kernels should have them anyway, idk

loud ice May 13, 2024, 4:27 PM

#

not really

left orbit May 13, 2024, 4:27 PM

#

slightly annoying for ~C(++)

loud ice May 13, 2024, 4:27 PM

#

zig ones won't

#

and C++ ones might not either

#

idk i like writing c++ that uses as few null terminated strings as possible

#

also lol strcmp and strncmp have one use each

#

memmove has more tbf

#

all in one function

#

and strlen and strnlen have two references each

loud ice May 13, 2024, 4:30 PM

#

left orbit because most kernels should have them anyway, idk

also not snprintf

left orbit May 13, 2024, 4:30 PM

#

i mean every time i implemented printf i also made an snprintf

#

and also vsnprintf

#

all the other ones are just built on top of that one

loud ice May 13, 2024, 4:31 PM

#

left orbit i mean every time i implemented printf i also made an snprintf

that of course assumes you implement printf

#

which again isn't necessarily the case

#

even in C kernels

#

and honestly for this level of use just implement your own "append int to buffer" func

#

its like 5 lines of code

left orbit May 13, 2024, 4:33 PM

#

yea im not the one to decide any of that

#

its a bit of a pita but it is what it is

fiery turtle May 13, 2024, 4:48 PM

#

loud ice what the actual fuck

Logging is built around printf, tons of aml conversion stuff needs snprintf

#

It wasn't built with zig in mind

#

Its a c library

#

But yeah shipping own printf would be nice

#

If its replaceable especially

left orbit May 13, 2024, 4:49 PM

#

nanoprintf is nice, and it's 0bsd

#

actually i realized i can make zig fetch nanoprintf for me and it doesnt look like it'll be too bad

fiery turtle May 13, 2024, 4:50 PM

#

Yeah thats the first thing I recommend to people that complain

loud ice May 13, 2024, 5:05 PM

#

fiery turtle Logging is built around printf, tons of aml conversion stuff needs snprintf

for logging, fair enough

#

but all the aml conversion stuff only needs numbers afaict

#

also it's not about that, it's about like

#

it's a pretty weird and rather annoying assumption to make

fiery turtle May 13, 2024, 5:09 PM

#

source/interpreter.c:    repr_len = uacpi_snprintf(
source/interpreter.c-        int_buf, sizeof(int_buf),
source/interpreter.c-        is_hex ? "%"UACPI_PRIX64 : "%"UACPI_PRIu64,
--
source/interpreter.c:        repr_len = uacpi_snprintf(
source/interpreter.c-            int_buf, sizeof(int_buf),
source/interpreter.c-            is_hex ? "0x%02X" : "%d",
--
source/interpreter.c:            size = uacpi_snprintf(int_buf, sizeof(int_buf), "%"UACPI_PRIx64,
source/interpreter.c-                                  arg1->integer);
source/interpreter.c-            if (size < 0)
--
source/utilities.c:    uacpi_snprintf(
source/utilities.c-        id_string->value, CLS_REPR_SIZE, "%02X%02X%02X",
source/utilities.c-        class_codes[0], class_codes[1], class_codes[2]
--
source/utilities.c:        size = uacpi_snprintf(UACPI_NULL, 0, "%"UACPI_PRIu64, obj->integer) + 1;
source/utilities.c-    }
source/utilities.c-
--
source/utilities.c:        uacpi_snprintf(
source/utilities.c-            id_string->value, id_string->size, "%"UACPI_PRIu64, obj->integer
source/utilities.c-        );

fiery turtle May 13, 2024, 5:09 PM

#

loud ice it's a pretty weird and rather annoying assumption to make

what assumption

#

no one assumes it exists

#

its a requirement that you implement it if you dont have one

#

at least for now

#

since its a C library (or at least was built as a C library initially) its pretty fair to assume a printf would already exist

#

since there are people using other languages it makes sense to try an require as little as possible

#

thats why going forward id like to ship own printf by default with an option to override it

left orbit May 13, 2024, 5:13 PM

#

that would be amazing if uacpi could just ship with nanoprintf or something by default, and if you have your own printf then you can define something that makes uacpi not compile nanoprintf

fiery turtle May 13, 2024, 5:13 PM

#

yeah

#

id like to do that for most not-so-common helpers

#

e.g. strncmp

left orbit May 13, 2024, 5:14 PM

#

oh yeah those too

fiery turtle May 13, 2024, 5:14 PM

#

ideally the stdlib.h header goes away with more atomic overrides if the client wishes to do so

#

and then you would do like #define uacpi_override_strncmp etc

loud ice May 13, 2024, 5:15 PM

#

fiery turtle what assumption

in that you don't assume a compiler or like standard header

#

but then you also pull in an osdep for strcmp which you use literally once

fiery turtle May 13, 2024, 5:16 PM

#

right, i dont see anything contradictory here

loud ice May 13, 2024, 5:16 PM

#

it's weird

fiery turtle May 13, 2024, 5:16 PM

#

what would be the not weird thing?

loud ice May 13, 2024, 5:16 PM

#

just implement your own strcmp for the literal one call site?

fiery turtle May 13, 2024, 5:17 PM

#

and then ill add usage 2

#

then usage 3

#

then move it to osdep

#

?

#

or keep using my impl?

loud ice May 13, 2024, 5:17 PM

#

or just don't move it to an osdep?

#

it's not like it's something that talks to hardware

#

or is os-specific in any way

fiery turtle May 13, 2024, 5:17 PM

#

its signficant enough to have an optimized implementation in something like linux

#

why would i force my crappy implementation

#

its 3 lines of code, just write it yourself

#

same goes for snprintf and other helpers

loud ice May 13, 2024, 5:18 PM

#

also snprintf is like

fiery turtle May 13, 2024, 5:18 PM

#

thats why i said ideally i ship my own but add an option to override it

loud ice May 13, 2024, 5:18 PM

#

you want precise outputs

fiery turtle May 13, 2024, 5:19 PM

#

it just wasnt the focus of the library at the early stages of development

loud ice May 13, 2024, 5:19 PM

#

fair enough yeah

fiery turtle May 13, 2024, 5:19 PM

#

loud ice you want precise outputs

I agree

#

relying on that to produce consistent output between kernels is probably not ideal

#

i already did that for strtoull btw, initially it was a requirement as well

#

right now there's an internal more cut down version

#

also what i didnt like about shipping own printf is like logging helpers get transformed into either

write one char at a time
write a string

#

to do the latter you have to have the buffer somewhere

#

so that either has to be preallocated (so must be lock guarded), stored on the stack (so waste stack space), or dynamically allocated (slow af)

loud ice May 13, 2024, 5:23 PM

#

fiery turtle also what i didnt like about shipping own printf is like logging helpers get tra...

yeah that's more understandable

fiery turtle May 13, 2024, 5:23 PM

#

that will have to be tunable as well i guess

#

so that kernels with builtin printf get the printf log callbacks

#

others just get one of those two

left orbit May 13, 2024, 5:24 PM

#

i allocate a buffer on the stack

#

sadmeme

fiery turtle May 13, 2024, 5:24 PM

#

that works for single threaded stuff

#

or wait

#

stack

#

well

left orbit May 13, 2024, 5:24 PM

#

yup stack

#

wtf ""

loud ice May 13, 2024, 5:24 PM

#

i have an on-stack buffer too

left orbit May 13, 2024, 5:24 PM

#

wtf is that lmao

fiery turtle May 13, 2024, 5:25 PM

#

the on-stack is the most common approach

left orbit May 13, 2024, 5:25 PM

#

how did i type a 0x7f

fiery turtle May 13, 2024, 5:25 PM

#

u could do like char[128] or [256]

#

i doubt it would try to log more than that

loud ice May 13, 2024, 5:25 PM

#

you could also go char[64] and flush in segments

fiery turtle May 13, 2024, 5:25 PM

#

oh yeah

loud ice May 13, 2024, 5:25 PM

#

oh god i have cursed idea

fiery turtle May 13, 2024, 5:25 PM

#

but segments kinda suck because no atomicity

#

linux still suffers from pr_cont shit

loud ice May 13, 2024, 5:26 PM

#

using recursion to allocate multiple stack segments

fiery turtle May 13, 2024, 5:26 PM

#

lmao

#

at that point char buf[snprintf(...));

loud ice May 13, 2024, 5:26 PM

#

fiery turtle at that point `char buf[snprintf(...));`

not required to be supported

#

so

fiery turtle May 13, 2024, 5:27 PM

#

ok ok _alloca(snprintf) for msvc

left orbit May 13, 2024, 5:27 PM

#

fiery turtle at that point `char buf[snprintf(...));`

what the fuck

loud ice May 13, 2024, 5:27 PM

#

fiery turtle ok ok `_alloca(snprintf)` for msvc

thats

#

still not standard

#

and also that has a much slower slow path

fiery turtle May 13, 2024, 5:27 PM

#

oh well

#

since we're discussing cursed ideas anyway

loud ice May 13, 2024, 5:27 PM

#

the recursive approach is at least standard lmao

fiery turtle May 13, 2024, 5:28 PM

#

anyway fixed size buffer should do the trick probably

#

128 bytes is unlikely to cause a stack overflow

#

and if your kernel doesnt have a printf it probably uses something other than C and at that point already consumes tons of stack space anyway

loud ice May 13, 2024, 5:29 PM

#

you can just guarantee that you cant overflow

loud ice May 13, 2024, 5:29 PM

#

fiery turtle and if your kernel doesnt have a printf it probably uses something other than C ...

nah

#

other languages can be efficient too

fiery turtle May 13, 2024, 5:29 PM

#

that was mostly a meme but yeah

#

i know rust can hog the stack

fiery turtle May 13, 2024, 5:29 PM

#

loud ice you can just guarantee that you cant overflow

how?

loud ice May 13, 2024, 5:29 PM

#

yeah thats just because its a shitty language :^)

fiery turtle May 13, 2024, 5:30 PM

#

uacpi_kernel_stack_space_left() halfmemeright

loud ice May 13, 2024, 5:30 PM

#

fiery turtle how?

void append(char* bufstart, int* buflen, char c) {
  if (*buflen == 128) (*buflen)--;
  bufstart[(*buflen)++] = c;
}

#

or alternatively

#

if (bufidx + segment_len < buflen) memcpy() /* fast path */ else /* slow path */

fiery turtle May 13, 2024, 5:31 PM

#

uhhh

#

what is that doing?

loud ice May 13, 2024, 5:31 PM

#

well so like if you are doing append not one character at a time

#

but in larger segments

#

you can take the fast path if it cant overflow

#

and otherwise take the (probably not inlined if you really care about that) slow path

fiery turtle May 13, 2024, 5:32 PM

#

i see

#

yeah i guess

kindred beacon May 13, 2024, 6:33 PM

#

fiery turtle But yeah shipping own printf would be nice

frigg meme

#

(I'm trolling at this point)

hollow elm May 13, 2024, 6:37 PM

#

lai has a small printf impl (written by yours truly) you could maybe take :^)

fiery turtle May 13, 2024, 6:55 PM

#

This is what I call peak NT compatibility:

--- a/include/uacpi/types.h
+++ b/include/uacpi/types.h
@@ -96,6 +96,7 @@ typedef struct uacpi_buffer_field {
     uacpi_size bit_index;
     uacpi_u32 bit_length;
     uacpi_bool force_buffer;
+    uacpi_bool out_of_bounds;
 } uacpi_buffer_field;

 typedef struct uacpi_buffer_index {
diff --git a/source/interpreter.c b/source/interpreter.c
index 7b08ca6..f532ade 100644
--- a/source/interpreter.c
+++ b/source/interpreter.c
@@ -3817,12 +3817,12 @@ static uacpi_status handle_create_buffer_field(struct execution_context *ctx)

     if (uacpi_unlikely((field->bit_index + field->bit_length) >
                        src_buf->size * 8)) {
-        uacpi_error(
+        uacpi_warn(
             "Invalid buffer field: bits [%zu..%zu], buffer size is %zu bytes\n",
             field->bit_index, field->bit_index + field->bit_length,
             src_buf->size
         );
-        return UACPI_STATUS_AML_OUT_OF_BOUNDS_INDEX;
+        field->out_of_bounds = UACPI_TRUE;
     }

     field->backing = src_buf;
@@ -4211,6 +4211,8 @@ static uacpi_status handle_inc_dec(struct execution_context *ctx)
             if (uacpi_unlikely_error(ret))
                 return ret;
         } else {
+            if (uacpi_unlikely(src->buffer_field.out_of_bounds))
+                return UACPI_STATUS_AML_OUT_OF_BOUNDS_INDEX;
             uacpi_read_buffer_field(&src->buffer_field, &dst->integer);
         }
         break;

fiery turtle May 13, 2024, 7:27 PM

#

wait

#

gcc and clang have __SIZEOF_POINTER__

verbal herald May 13, 2024, 7:34 PM

#

left orbit well so there is basically 2 ways to consume uacpi from zig right now: - transla...

Well I currently use uACPI directly using Zig's @cImport, though a proper binding would be really nice

left orbit May 13, 2024, 7:38 PM

#

yep thats part of my plan for an in tree wrapper

#

it’ll be a lot of wrapping though

#

im also figuring out the proper way to do everything

fiery turtle May 13, 2024, 8:33 PM

#

I merged all of the fixes, and API constness + got rid of bitfields in public api

#

also pointer detection so riscv should work now

fiery turtle May 13, 2024, 8:34 PM

#

fiery turtle I merged all of the fixes, and API constness + got rid of bitfields in public ap...

@left orbit you should probably stop using that PR branch now

#

i havent merge the global lock stuff yet because there's a lot of controversial choices there that i will need to think about

left orbit May 13, 2024, 8:49 PM

#

fiery turtle <@796462670995390524> you should probably stop using that PR branch now

i mainly wanted it for the bitfield stuff so thank you for that :^)

verbal herald May 13, 2024, 9:01 PM

#

left orbit yep thats part of my plan for an in tree wrapper

i'm actually thinking of writing my own zig wrapper too, though i haven't wrote any code for it yet

left orbit May 13, 2024, 9:02 PM

#

well i’m just starting out so it’ll be a while until there is something to use, but i hope it won’t be too long

fiery turtle May 13, 2024, 9:22 PM

#

maybe you could join efforts?

rustic compass May 14, 2024, 5:55 AM

#

Maybe we should create a list of who works on which API wrapper to prevent multiple redundant projects

fiery turtle May 14, 2024, 8:30 AM

#

im all for it if people are willing to maintain those, I can fix it up myself even as long as its not insanely complicated to compile

#

(if its stored in the main tree somehow)

left orbit May 14, 2024, 10:12 AM

#

/home/czapek/source/uacpi-zig/example/zig-cache/o/5d8157b26ae4e1098fb65e53e66d4d88/cimport.h:1:10: error: 'string.h' file not found
#include <string.h>
         ^```

#

uhhhhh

#

what the Heck zig

loud ice May 14, 2024, 10:13 AM

#

lmao

#

zig moment

left orbit May 14, 2024, 10:13 AM

#

i'm not even building freestanding or anything? lmao

#

just a "standard" target

#

oh i need to tell zig to link in the C library if i want the headers...

#

on one hand that is smart, but on the other hand fuck zig

#

i guess i'm gonna have to provide some standard headers for uacpi

verbal herald May 14, 2024, 10:19 AM

#

well i did that while using uacpi on my zig os

left orbit May 14, 2024, 10:19 AM

#

i think im gonna yoink https://github.com/mintsuki/freestanding-headers

GitHub

GitHub - mintsuki/freestanding-headers: Collection of freestanding ...

Collection of freestanding C headers for GCC and Clang. - mintsuki/freestanding-headers

#

it doesn't have inttypes.h though

left orbit May 14, 2024, 10:37 AM

#

ok i got the missing headers, when im back from the store i'll implement the missing libc functions, provide a dummy kernel api so it links and put it on github so we can start working on this (if someone wants to contribute)

fiery turtle May 14, 2024, 11:01 AM

#

Nice

left orbit May 14, 2024, 5:33 PM

#

hmm uacpi is calling uacpi_kernel_alloc with size=11 but freeing with size=12 thinkong

#

[    0.011000] [CPU0] debug(acpi): allocation 11 bytes at 0xFFFFA0000021DAD0
[    0.013000] [CPU0] debug(acpi): allocation 88 bytes at 0xFFFFA0000040B600
[    0.015000] [CPU0] debug(acpi): freeing 12 bytes at 0xFFFFA0000021DAD0
[    0.020000] [CPU0] err(gpa): Allocation size 11 bytes does not match free size 12. Allocation:  Free: 
[    0.022000] [CPU0] debug(acpi): allocation 16 bytes at 0xFFFFA0000021DAE0```

#

[    0.028000] [CPU0] err(debug):   0xFFFFFFFF800C21B3: uacpi_kernel_free at /usr/local/zig/lib/std/mem/Allocator.zig:98
[    0.028000] [CPU0] err(debug):   0xFFFFFFFF8001D02D: uacpi_free_dynamic_string at kernel/./uacpi/source/utilities.c:1208
[    0.028000] [CPU0] err(debug):   0xFFFFFFFF8005B615: uacpi_free_absolute_path at kernel/./uacpi/source/namespace.c:550```

#

oh my god its my shitty str(n)len i think

#

YUP!

#

my bad

#

works perfectly fine now if i don't return the wrong length...

fiery turtle May 14, 2024, 5:43 PM

#

Yeah I have tests for verifying sizes

stable topaz May 15, 2024, 11:31 AM

#

is it normal on qemu that uacpi sometimes gives me this warning and error:

warn unsupported GAS access size 128
error unable to create FADT GPE block 0: unimplemented

it sometimes doesn't happen and i don't know why

fiery turtle May 15, 2024, 11:41 AM

#

nope not normal

#

you're probably returning an incorrectly mapped pointer

#

or corrupting the memory

fiery turtle May 15, 2024, 8:45 PM

#

This is how ACPICA solves the need for PRIu64 macros:

/*
 * printf() format helper. This macro is a workaround for the difficulties
 * with emitting 64-bit integers and 64-bit pointers with the same code
 * for both 32-bit and 64-bit hosts.
 */
#define ACPI_FORMAT_UINT64(i)           ACPI_HIDWORD(i), ACPI_LODWORD(i)

#

idk man lol

#

whatever I guess, this simplifies my printf implementation

#

ill do this too

#

Also I'm deleting the platform/stdlib.h header in favor of platform/libc.h

#

#

this is all its going to contain

#

no more nanoprintf or random header imports needed

#

ill try to solve the popcount issue as well

fiery turtle May 15, 2024, 8:50 PM

#

fiery turtle

@loud ice caret and also @left orbit

loud ice May 15, 2024, 8:52 PM

#

neat

fiery turtle May 15, 2024, 8:53 PM

#

now to implement all of those..

fiery turtle May 15, 2024, 9:14 PM

#

okay did everything but printf

#

noticably slower but whatever

#

now printf

#

i'll also make the printf-style-kernel-log optional

#

something like UACPI_PLAIN_LOGGING

#

if thats enabled it will just pass a string to the log calback

#

and vlog callback will disappear

#

or actually maybe the default should be reversed

#

#ifndef UACPI_FORMATTED_LOGGING
void uacpi_kernel_log(enum uacpi_log_level, const uacpi_char*);
#else
UACPI_PRINTF_DECL(2, 3)
void uacpi_kernel_log(enum uacpi_log_level, const uacpi_char*, ...);
void uacpi_kernel_vlog(enum uacpi_log_level, const uacpi_char*, uacpi_va_list);
#endif

fiery turtle May 15, 2024, 9:42 PM

#

how the fuck do u emulate PRIu64 with two integers...

#

i dont think u can even

#

oh wait, i think precision of 0 works

#

yay printf("%.0X%.0X\n", (unsigned)(val >> 32), (unsigned)val); does what i want

loud ice May 15, 2024, 10:05 PM

#

fiery turtle noticably slower but whatever

importing mem{cpy,set,cmp} should probably be the default

#

but the rest no

kind mantle May 15, 2024, 10:06 PM

#

loud ice importing mem{cpy,set,cmp} should probably be the default

100%

#

I have complete implementations of all of them chilling here if you need them: https://github.com/badgeteam/BadgerOS/blob/main/kernel/src/badgelib/badge_strings.c

GitHub

BadgerOS/kernel/src/badgelib/badge_strings.c at main · badgeteam/Ba...

A new operating system designed to bring portable native apps to electronic event badges. - badgeteam/BadgerOS

#

It even does multi-word copy optimizations

fiery turtle May 15, 2024, 10:21 PM

#

loud ice importing mem{cpy,set,cmp} should probably be the default

Yeah probably, ill let the user decide tho

#

I cant make it default because I have to assume a specific header, at which point you have to make an override file so it’s even more annoying

#

Or I can assume default signature but that's just dirty

fiery turtle May 15, 2024, 10:24 PM

#

kind mantle I have complete implementations of all of them chilling here if you need them: h...

Is your os riscv only?

#

Thats interesting

kind mantle May 15, 2024, 10:25 PM

#

fiery turtle Is your os riscv only?

Currently yes

#

I made an effort to still separare CPU-specific code into it's own folder so it may be ported at a later date.

#

But the target chip is RISC-V and of course so will be the OS.

fiery turtle May 15, 2024, 10:26 PM

#

Why that chip specifically?

kind mantle May 15, 2024, 10:27 PM

#

TL;DR: Espressif sponsors us (as in Badge.team) and they give us the new ESP chips in exchange for us making them part of the official electronic event badge for a camp with ~3000 attendees.

#

https://why2025.org/ is the next camp

fiery turtle May 15, 2024, 10:28 PM

#

Oh cool

kind mantle May 15, 2024, 10:28 PM

#

Very

fiery turtle May 15, 2024, 10:28 PM

#

You actually get paid for your os

kind mantle May 15, 2024, 10:28 PM

#

Nope

#

Volunteer

#

The entire camp is volunteers

fiery turtle May 15, 2024, 10:28 PM

#

Oh

loud ice May 15, 2024, 10:29 PM

#

oh wow

kind mantle May 15, 2024, 10:29 PM

#

So yeah if you have spare time for OSDEV and want something valuable on your CV, hit me up :P

fiery turtle May 15, 2024, 10:30 PM

#

kind mantle https://why2025.org/ is the next camp

Im sorry I tried reading this but I dont understand what this event is about

#

What are electronic event badges

loud ice May 15, 2024, 10:31 PM

#

dont ask stupid questions like "why"

kind mantle May 15, 2024, 10:33 PM

#

This camp is the next one in a series of quadrennial hacker camps in the Netherlands. Since 2017, the event has official "badges" made by, fittingly, Badge.team. These badges can show your name, but also run apps.
The apps can be games, they can be a real GPS map, they can be a talks schedule, whatever you want really.

fiery turtle May 15, 2024, 10:34 PM

#

Ohh

#

What do they look like?

kind mantle May 15, 2024, 10:35 PM

#

let me get you a picture, hang on

fiery turtle May 16, 2024, 8:18 PM

#

Ported printf from hyper and adapted for uACPI

#

#

Seems to work great

#

now uACPI no longer depends on a single function from libc besides kernel_api.h by default

#

should greately simplify interop with other languages

#

and just overall make porting less annoying

#

This should also greately simplify language interop

#

im also considering making log level a static prefix behind a define

#

so that there isnt any extra annoying work to append it

left orbit May 16, 2024, 8:35 PM

#

i actually like that it's passed in because i can forward it to zig's standard logging

fiery turtle May 16, 2024, 8:36 PM

#

im curious why does zig standard logging need uacpi log level?

north holly May 16, 2024, 8:39 PM

#

fiery turtle im curious why does zig standard logging need uacpi log level?

the integer values can probably be cast directly to those of the zig standard library and mean the same thing

fiery turtle May 16, 2024, 9:31 PM

#

Yay

#

not merged yet but PR up

#

will merge after some testing and a bit of review

left orbit May 16, 2024, 9:41 PM

#

fiery turtle im curious why does zig standard logging need uacpi log level?

what oberrow said, except i do the mapping explicitly since i map UACPI_LOG_TRACE (or whatever the log level is called) to the debug log level, but i basically just call the equivalent log function for each uacpi log level

#

const message = buffer[0..length];
switch (level) {
    C.UACPI_LOG_DEBUG, C.UACPI_LOG_TRACE => logger.debug("{s}", .{message}),
    C.UACPI_LOG_INFO => logger.info("{s}", .{message}),
    C.UACPI_LOG_WARN => logger.warn("{s}", .{message}),
    C.UACPI_LOG_ERROR => logger.err("{s}", .{message}),
    else => unreachable,
}```

#

basically what i do

fiery turtle May 16, 2024, 9:42 PM

#

oh i see

#

but u could also do for example #define UACPI_LOG_TRACE_PREFIX "\x1

#

and then switch (buffer[0])

#

(had i actually implemented that)

left orbit May 16, 2024, 9:43 PM

#

that's kinda nasty lmaoo

#

that would work yeah

fiery turtle May 16, 2024, 9:44 PM

#

thats how linux does log levels lmao

#

that kinda reduces reg usage and stuff ya know

left orbit May 16, 2024, 9:44 PM

#

i suppose yeah

fiery turtle May 16, 2024, 9:44 PM

#

but yeah idk

mortal yoke May 16, 2024, 9:45 PM

#

fiery turtle that kinda reduces reg usage and stuff ya know

how though? you still need to preserve the arg registers if you have something else in them

#

even if the function doesn't take that many args

left orbit May 16, 2024, 9:45 PM

#

not really critical though imho, you are calling into a kernel api that you have no control over so i don't think that micro-optimization will make a difference

fiery turtle May 16, 2024, 9:45 PM

#

yeah

left orbit May 16, 2024, 9:45 PM

#

it makes sense in a more tightly integrated system like linux where they control everything

fiery turtle May 16, 2024, 9:46 PM

#

another thing is this makes the string sort of atomic, as in it already contains the level prefix, u can just memcpy it into the log buffer for example

#

or log it in one shot

#

e.g i've seen people do

print(log_level_str);
print(msg)

#

like i know this is terrible if you have SMP

mortal yoke May 16, 2024, 9:50 PM

#

just have the print function accept multiple things lol

#

or explicitly take a print lock at the start and then release it after you have done printing everything

fiery turtle May 16, 2024, 9:50 PM

#

anyway ill keep the log_level thing for now, we'll see if the need arises later on

loud ice May 16, 2024, 9:51 PM

#

mortal yoke or explicitly take a print lock at the start and then release it after you have ...

or buffer up all the things and only take the lock for the print path

left orbit May 16, 2024, 9:52 PM

#

fiery turtle e.g i've seen people do ``` print(log_level_str); print(msg) ```

i used to just print("%s %s", log_level_str, msg); lol

#

you could always make it a compile time switch :^)

#

but i feel like that'd be silly

fiery turtle May 16, 2024, 9:52 PM

#

left orbit i used to just `print("%s %s", log_level_str, msg);` lol

see this is what im talking about lol

#

what it was already preformatted with that

fiery turtle May 16, 2024, 9:52 PM

#

left orbit but i feel like that'd be silly

maybe

left orbit May 16, 2024, 9:53 PM

#

i mean on one hand it's nice to let the user pick what they really want

#

but i can see how that could get annoying one you start adding more things

#

more and more things to configure

#

and more and more code to maintain all those alternate code paths in uacpi itself

#

even if they are trivial like in the log level case

#

#ifdef UACPI_LOG_WITH_LEVEL
...
#else
...
#endif```

#

you need another CI variant to make sure nothing breaks and all that

#

not that good of a solution

mortal yoke May 16, 2024, 9:55 PM

#

left orbit you need another CI variant to make sure nothing breaks and all that

I mean it could likely just be made a macro that is switched at compile-time, then the other code wouldn't have to care about it

#

but yeah its probably not a good idea, idk

left orbit May 16, 2024, 9:55 PM

#

oh that is true actually

fiery turtle May 16, 2024, 10:06 PM

#

mortal yoke I mean it could likely just be made a macro that is switched at compile-time, th...

wdym

fiery turtle May 16, 2024, 10:06 PM

#

left orbit you need another CI variant to make sure nothing breaks and all that

yeah i can see this leading to a combinatorial explosion

#

which im slowly approaching

fiery turtle May 17, 2024, 9:55 PM

#

Ran uACPI against a giant AML blob with --namespace-enumerate with MSVC printf vs my printf

#

looks like exactly the same output which is amazing

#

pointer addresses are different but thats kinda expected lol

#

anyway this should be good to merge now

hollow elm May 17, 2024, 9:56 PM

#

there's some sysctl you could set to turn off aslr :^)

fiery turtle May 17, 2024, 9:56 PM

#

not on windows but yeah

hollow elm May 17, 2024, 9:57 PM

#

ah fair

#

gdb also disables aslr somehow for it's child process

#

at least on linux

fiery turtle May 17, 2024, 9:57 PM

#

gdb freaks out if you have asan

#

so yeah

#

but idk how it does it

hollow elm May 17, 2024, 9:58 PM

#

ime it doesn't freak out but it doesn't always break into the debugger when asan trips on something

fiery turtle May 17, 2024, 9:58 PM

#

oh yeah? for me it segfaulted

#

maybe its fixed now

hollow elm May 17, 2024, 9:58 PM

#

mainly when asan in the child exits with an abort instead of a segfault

fiery turtle May 17, 2024, 9:58 PM

#

but maybe that was an ubuntu 22 bug where asan entropy caused the app to segfault every now and again

#

i wonder if I should delete all the helpers i added to mangarm that only uacpi uses

#

it has a bunch of useless libc crap now

hollow elm May 17, 2024, 10:00 PM

#

might be a good idea

fiery turtle May 17, 2024, 10:00 PM

#

the libc headers need a general cleanup as well

#

because they export symbols that dont exist

#

for some of them

fiery turtle May 17, 2024, 10:01 PM

#

hollow elm might be a good idea

ill do it on next upstream sync then

hollow elm May 17, 2024, 10:01 PM

#

👍

fiery turtle May 17, 2024, 10:02 PM

#

my hacked together crappy printf will also go away finally lmao

#

the one with frigg callbacks

left orbit May 17, 2024, 10:11 PM

#

fiery turtle not on windows but yeah

link with /DYNAMICBASE:NO but i dont know if these pointers will help since those are heap addresses

#

that will just disable load base randomization

fiery turtle May 17, 2024, 10:11 PM

#

yeah

#

its not a problem honetsly, its just these 4 lines

fiery turtle May 17, 2024, 11:14 PM

#

merged!

#

no more dependencies besides kernel_api.h

#

should be very very easy to port now

hollow elm May 17, 2024, 11:15 PM

#

based

fiery turtle May 17, 2024, 11:16 PM

#

Thank you!

hasty plinth May 18, 2024, 7:39 AM

#

very nice

#

and I see the platform-specific stuff in platform/compiler.h got replaced too 😄

fiery turtle May 18, 2024, 9:01 AM

#

hasty plinth and I see the platform-specific stuff in platform/compiler.h got replaced too 😄

The pointer size thing? Yeah

fiery turtle May 19, 2024, 12:15 PM

#

@hasty plinth uhhh

#

https://github.com/DeanoBurrito/northport/blob/e8f478002bd34246f8a422e52416636eaa447f5a/drivers/uacpi/BindsMisc.cpp#L25

GitHub

northport/drivers/uacpi/BindsMisc.cpp at e8f478002bd34246f8a422e524...

Monolithic kernel and support libraries for riscv64 and x86_64. - DeanoBurrito/northport

#

i think this might be leaking memory

hasty plinth May 19, 2024, 12:16 PM

#

Uh yeah, it certainly is

#

Thanks for the ping

fiery turtle May 19, 2024, 12:17 PM

#

np!

hasty plinth May 19, 2024, 12:19 PM

#

Tbh I just assumed uacpi needed lots of memory when I saw the memory usage go up 😅

fiery turtle May 19, 2024, 12:19 PM

#

lol

hollow crescent May 19, 2024, 1:06 PM

#

hasty plinth Uh yeah, it certainly is

What's up with that?

#

BTW, typo: const npk_monotomic_time uptime = npk_get_monotomic_time(); it's supposed to be monotonic

#

it's one tone

#

not one tome

hasty plinth May 19, 2024, 1:12 PM

#

hollow crescent BTW, typo: `const npk_monotomic_time uptime = npk_get_monotomic_time();` it's su...

Haha cheers, that's probably one spelling mistake that's been propagated everywhere via lsp. Woops.

hasty plinth May 19, 2024, 1:14 PM

#

hollow crescent What's up with that?

Probably just writing code too late at night

hollow crescent May 19, 2024, 1:16 PM

#

hasty plinth Haha cheers, that's probably one spelling mistake that's been propagated everywh...

what's lsp?

hollow crescent May 19, 2024, 1:17 PM

#

hasty plinth Probably just writing code too late at night

i mean like

#

is it uacpi leaking? or northport

#

what was going on

hasty plinth May 19, 2024, 1:17 PM

#

Nah it's my kernel

hollow crescent May 19, 2024, 1:17 PM

#

ah

hasty plinth May 19, 2024, 1:17 PM

#

I allocated once, and if that succeeded I zeroed the memory. Then returned memory from a new allocation.

hollow crescent May 19, 2024, 1:18 PM

#

oh LOL

hasty plinth May 19, 2024, 1:18 PM

#

Yeah lol

hollow crescent May 19, 2024, 1:18 PM

#

tbh i wish i had more motivation because i'm stuck on some seemingly trivial boilerplate

hasty plinth May 19, 2024, 1:18 PM

#

hollow crescent what's lsp?

Language server protocol, I quite heavily use the completion suggestions

hollow crescent May 19, 2024, 1:19 PM

#

I wanna do a lot of stuff just that I'm not at that stage yet

hasty plinth May 19, 2024, 1:20 PM

#

Thats frustrating

thorny orbit May 19, 2024, 1:28 PM

#

hollow crescent tbh i wish i had more motivation because i'm stuck on some seemingly trivial boi...

What are you stuck with right now

hollow crescent May 19, 2024, 1:30 PM

#

thorny orbit What are you stuck with right now

seemingly trivial boilerplate, right now I need to add a system owned object handle table to my process struct, and then I need to create the IO manager

#

I don't want to create something like unix, I want to create something akin to NT, basically a fundamentally asynchronous I/O system

#

I understand a lot of what makes it tick but I also don't understand a lot

tired crater May 19, 2024, 7:22 PM

#

(kinda) acpi related question: is there any acpi power state that can put the cpu in a lower power mode than the cpus min mhz?

#

iirc S2 puts the cpu in idle state, so on a cpu with min mhz = 800 it puts it to 800 mhz

north holly May 19, 2024, 7:23 PM

#

S5 turns off the CPU

#

0 mhz

#

so does S4

tired crater May 19, 2024, 7:23 PM

#

well ones that dont turn it off completely

#

now i'm having a problem where my whole laptop becomes INSANELY unresponsive when its plugged in

#

and its fine if not plugged in

mortal yoke May 19, 2024, 7:24 PM

#

tired crater iirc S2 puts the cpu in idle state, so on a cpu with min mhz = 800 it puts it to...

S2 can also completely power off the cpu

north holly May 19, 2024, 7:24 PM

#

tired crater iirc S2 puts the cpu in idle state, so on a cpu with min mhz = 800 it puts it to...

Don't all sleep states but S0 make it so the cpu doesn't run

tired crater May 19, 2024, 7:24 PM

#

when not plugged in it idles at 800mhz and tops at however many ghz my cpu can run at and when plugged in it idles at 200mhz and tops at 800

north holly May 19, 2024, 7:24 PM

#

anything

tired crater May 19, 2024, 7:26 PM

#

is frequency scaling affected by acpi

#

well it's controlled by acpi

#

i think my ac is broken

#

ok it is, when plugged in linux reports charging state at a rate of 0?

fiery turtle May 19, 2024, 7:28 PM

#

tired crater (kinda) acpi related question: is there any acpi power state that can put the cp...

that can probably be done via advanced p/cstate management

#

aka not directly related to acpi but somewhat

tired crater May 19, 2024, 7:29 PM

#

so it could be that linux is going ahead and saying: well you're plugged in, but not getting any decent voltage, so fuck yourself and live with an 800 mhz cpu?

fiery turtle May 19, 2024, 7:29 PM

#

iirc linux calls _OSC on the cpu object, which tells aml it supports advanced power mangament

#

which loads new objects

fiery turtle May 19, 2024, 7:29 PM

#

tired crater so it could be that linux is going ahead and saying: well you're plugged in, but...

yeah

tired crater May 19, 2024, 7:29 PM

#

could it also be my motherboard

#

or some other hardware component

fiery turtle May 19, 2024, 7:30 PM

#

yes

tired crater May 19, 2024, 7:30 PM

#

fun

fiery turtle May 19, 2024, 7:30 PM

#

check if windows also does this

tired crater May 19, 2024, 7:30 PM

#

thanks tho

#

yes it does

fiery turtle May 19, 2024, 7:30 PM

#

yeah im afraid hw issue then

#

if you have proprietary chipset drivers installed as well

tired crater May 19, 2024, 7:31 PM

#

ive never had that problem so i'm quite convinced it's to blame on my power adapter or something similar

fiery turtle May 19, 2024, 7:31 PM

#

yeah that could be the issue

tired crater May 19, 2024, 7:32 PM

#

anyways, laptop gonna be gone in a few minutes since I have pretty much no battery charge left

fiery turtle May 19, 2024, 8:06 PM

#

im so close to "finishing" uacpi (at least a stable 1.0), but have been a bit lazy to do anything lately

#

spent the weekend playing minecraft

#

caret

jaunty fox May 19, 2024, 8:07 PM

#

fiery turtle im so close to "finishing" uacpi (at least a stable 1.0), but have been a bit la...

will the APIs also become stable after 1.0?

fiery turtle May 19, 2024, 8:07 PM

#

yup

jaunty fox May 19, 2024, 8:08 PM

#

very helpful, i will have to upgrade and see what i've missed out on

fiery turtle May 19, 2024, 8:08 PM

#

im a fan of stable api myself so i hate to break it even now

#

even though the readme explicitly mentions that its unstable right now

north holly May 19, 2024, 8:09 PM

#

jaunty fox May 19, 2024, 8:09 PM

#

it's still young and at least there actually is a 1.0 in sight

fiery turtle May 19, 2024, 8:09 PM

#

but sometimes its better to break than to live with crappy api

fiery turtle May 19, 2024, 8:10 PM

#

jaunty fox it's still young and at least there actually is a 1.0 in sight

yeah definitely

#

im aiming for june/july 1.0

#

then i can finally go back to other projects after a year long detour

#

i wonder if its worth wasting my time to make everything numa aware from the start

left orbit May 19, 2024, 9:04 PM

#

no longer CopyObject Abuser

#

rip

#

definitely looking forward to 1.0 though :^)

fiery turtle May 19, 2024, 9:49 PM

#

Thanks

fiery turtle May 19, 2024, 9:49 PM

#

left orbit no longer CopyObject Abuser

People were unironically referring to me by that handle so I thought I should stop there

left orbit May 19, 2024, 9:50 PM

#

cant wait to be called "thread 210425 panic"

dawn pine May 19, 2024, 9:50 PM

#

hello thread 210425 panic

dusky glade May 19, 2024, 9:52 PM

#

left orbit cant wait to be called "thread 210425 panic"

tkill 210425

#

trl

left orbit May 19, 2024, 9:52 PM

#

dusky glade tkill 210425

Operation not permitted

dusky glade May 19, 2024, 9:52 PM

#

left orbit Operation not permitted

nooo

dawn pine May 19, 2024, 9:52 PM

#

sudo tkill 210425

left orbit May 19, 2024, 9:53 PM

#

trollvanish

lapis hawk May 21, 2024, 5:24 AM

#

apology for bad english

where were u wen thread 210425 die

fiery turtle May 22, 2024, 9:20 PM

#

okay so... where was i

#

cleaning up my global lock PR

#

by default uACPI will use void* as thread id

#

this should hopefully cover 99.9% of users, if not override it ofc

#

this is the general idea

fiery turtle May 22, 2024, 9:53 PM

#

Aaand look at that!

#

Reading from a Lock'ed AML field

#

The PR will need a bit more cleanup and some real hw tests ofc

#

but looks like it works

fiery turtle May 22, 2024, 10:13 PM

#

made a simple test case as well

#

this acquires the global lock a ton of times, one of them being by a locked field write

fiery turtle May 22, 2024, 10:38 PM

#

#

well looks like it works in managarm as well

#

at least on qemu

#

for whatever reason reading from hpet needs the global lock

#

also pretty convenient

vagrant hull May 23, 2024, 6:59 AM

#

i should add uACPI to Vinix like next month

#

sounds like something fun to do on the side

fiery turtle May 23, 2024, 7:11 AM

#

That would be nice

#

Does V have a c bindings generator?

vagrant hull May 23, 2024, 7:14 AM

#

well it has a c2v thing but it's not quite a bindings generator

#

at least i think

#

i would have to see what @ancient sable has to say about that

ancient sable May 23, 2024, 7:20 AM

#

vagrant hull well it has a c2v thing but it's not quite a bindings generator

it does have a bindings generator

#

i'm actually working on integrating into v

#

so that you can just #include "foo.h" and call C.fns(3,4,5) right away

#

without declaring the functions

#

for now you do have to declare them, but you can use v translate -wrapper library.h

vagrant hull May 23, 2024, 7:22 AM

#

ancient sable it does have a bindings generator

ah that's good to know

#

i think i actually used that before, just completely forgot about it

vagrant hull May 23, 2024, 7:22 AM

#

ancient sable so that you can just #include "foo.h" and call C.fns(3,4,5) right away

that sounds pretty awesome

ancient sable May 23, 2024, 7:22 AM

#

yeah it's super useful and it's been improved a lot

#

it can handle entire sokol library for example

vagrant hull May 23, 2024, 7:23 AM

#

amazing

#

i will try throwing uacpi at it then

ancient sable May 23, 2024, 7:23 AM

#

well sokol is clean and modern

#

older libs with macros may be trickier

vagrant hull May 23, 2024, 7:23 AM

#

well uacpi is quite modern too

#

idk how many macros it uses though

fiery turtle May 23, 2024, 7:24 AM

#

User facing pretty much none

#

Or none that are mandatory to use

vagrant hull May 23, 2024, 7:24 AM

#

nice then

fiery turtle May 23, 2024, 7:25 AM

#

I specifically simplified some headers too

#

So that bindings are easier to make

#

Like there are no user facing bitfields for example

#

Zig just doesn't support that

fiery turtle May 23, 2024, 12:18 PM

#

okay well global lock seems to work on real hardware

#

although i didnt see any contention on it unfortunately, at least from briefly looking at the logs (no way to view them after as managarm dies after boot because driver bugs)

trim siren May 23, 2024, 7:06 PM

#

fiery turtle Zig just doesn't support that

doesnt zig support arbitrary bit-width ints?

loud ice May 23, 2024, 7:07 PM

#

they do, but probably @cInclude is just broken

left orbit May 23, 2024, 7:10 PM

#

translate-c doesn't understand bitfields

verbal herald May 23, 2024, 7:26 PM

#

trim siren doesnt zig support arbitrary bit-width ints?

yes, but not bitfields

#

variable bit-width ints != bitfields

trim siren May 23, 2024, 7:36 PM

#

verbal herald yes, but not bitfields

wouldn't packed + bit-width int make a bitfield basically?

loud ice May 23, 2024, 7:40 PM

#

translate-c cant produce that

#

so

fiery turtle May 23, 2024, 7:43 PM

#

bitfields are also very unstable between platforms so

#

i wouldnt rely on zig's ability to translate them correctly

left orbit May 23, 2024, 7:45 PM

#

i dont think packed structs are compatible with extern structs

#

so yeah, translate-c just cant produce it

verbal herald May 23, 2024, 7:52 PM

#

trim siren wouldn't packed + bit-width int make a bitfield basically?

no

fiery turtle May 24, 2024, 1:01 PM

#

Designing the API for advanced _OSI configuration

#

this is mostly for very advanced kernels but should be implemented nonetheless

left orbit May 24, 2024, 2:42 PM

#

wouldn't it make sense for the uacpi_{un,}register_interface to take a uacpi_interface instead of a const char *? it seems that they are somewhat standardized and you already have an enum of them

thorny orbit May 24, 2024, 2:52 PM

#

left orbit wouldn't it make sense for the `uacpi_{un,}register_interface` to take a `uacpi_...

aren’t these functions used to register new interfaces? I don’t see how it would make sense to register already existing interfaces

fiery turtle May 24, 2024, 2:57 PM

#

thorny orbit aren’t these functions used to register new interfaces? I don’t see how it would...

Yup

fiery turtle May 24, 2024, 3:05 PM

#

left orbit wouldn't it make sense for the `uacpi_{un,}register_interface` to take a `uacpi_...

These are basically for registering new custom interfaces, e.g Darwin

#

The enum above is standard interfaces like Windows versions

fiery turtle May 24, 2024, 3:06 PM

#

fiery turtle These are basically for registering new custom interfaces, e.g Darwin

This can be an arbitrary string

left orbit May 24, 2024, 3:08 PM

#

ahh i see

fiery turtle May 24, 2024, 3:31 PM

#

anyway this is a draft and the name uacpi_interface was misleading

#

i renamed to uacpi_vendor_interface, which is what it is

#

the idea is certain quirk handling code needs to know the latest windows version AML queried support for

#

because different windows versions need different workarounds

#

example:

fiery turtle May 24, 2024, 4:03 PM

#

also in hindsight uacpi_kernel_acquire_mutex returning a bool is the most retarded decision ive made

#

ill change that to uacpi_status

rain nimbus May 24, 2024, 5:32 PM

#

fiery turtle also in hindsight `uacpi_kernel_acquire_mutex` returning a bool is the most reta...

At what point should it be allowed to fail tho?

fiery turtle May 24, 2024, 5:33 PM

#

it might oom or fail for other reasons

#

otherwise im forced to do this shit everywhere

if (uacpi_unlikely(!uacpi_kernel_acquire_mutex(mtx, 0xFFFF)))
    return UACPI_STATUS_INTERNAL_ERROR;

#

so i lose the chance to propagate callee errors

#

also this allows to propagate timeout in a uniform fashion

#

UACPI_STATUS_TIMEOUT is a thing

rain nimbus May 24, 2024, 5:57 PM

#

Hmm I guess

fiery turtle May 24, 2024, 7:22 PM

#

...aaand this PR merge marks the end of missing ACPI features

#

everything else thats left is either utility api or locking

rain nimbus May 25, 2024, 4:47 AM

#

I'm so gonna have fun integrating uACPI in a few months, once I finish my other two projects

fiery turtle May 25, 2024, 7:59 AM

#

rain nimbus I'm so gonna have fun integrating uACPI in a few months, *once I finish my other...

The long awaited Marca return to osdev

rain nimbus May 25, 2024, 8:01 AM

#

Yes, currently "researching" different methods of rendering multiple swapchains through vulkan in a way that results in good performance at all times.
Which is kinda difficult when DWM is like "Oh I see you want to present this window, well lemme just spend 600 us doing nothing for no reason"

#

And then once I'm done with that I'll spend time writing a HTTP socket system for at least 1.1 and 2, can't really be bothered to implement QUIC xD
And once I'm done with that I can complete the secret manager, and once that's done I'll move back to osdev :>

#

So I'm expecting I'll get back to osdev around august, september ish

fiery turtle May 25, 2024, 1:36 PM

#

Made a test case for _OSI

#

For whatever reason even added compatibility for ACPICA

#

the first command in the terminal is this test fed to ACPICA

#

the second is uACPI test runner

jaunty fox May 25, 2024, 1:36 PM

#

acpica is identified by AnotherTestString?

fiery turtle May 25, 2024, 1:37 PM

#

uACPI test runner:

#

ACPICA:

fiery turtle May 25, 2024, 1:37 PM

#

jaunty fox acpica is identified by AnotherTestString?

funny enough yes

#

lmao

fiery turtle May 25, 2024, 8:52 PM

#

Just merged the interfaces PR + simplified the initialization structure

#

❯ cloc --vcs=git .
     134 text files.
     134 unique files.
      60 files ignored.

github.com/AlDanial/cloc v 1.90  T=2.19 s (34.7 files/s, 13586.8 lines/s)
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
C                               19           3207            810          14741
C/C++ Header                    44            831           1074           4536
C++                              4            379            493           2421
Python                           3            207             30            561
Markdown                         1             59              0            184
CMake                            3             23             14            109
YAML                             1             10              5             56
Meson                            1              2              0             23
-------------------------------------------------------------------------------
SUM:                            76           4718           2426          22631
-------------------------------------------------------------------------------

#

This is where it currently stands

#

I dont expect it to grow much more

#

Pretty much 95% of the features & helper API you could possible need is done

#

What's left is general thread safety, maybe a few misc api features im still considering

fiery turtle May 25, 2024, 11:37 PM

#

I have just realized I dont have any call stack depth limits... Should probably limit it to like 256

fiery turtle May 26, 2024, 9:06 AM

#

fiery turtle I have just realized I dont have any call stack depth limits... Should probably ...

#

#

Done!

hasty plinth May 26, 2024, 9:07 AM

#

silly question, is the depth configurable?

rain nimbus May 26, 2024, 9:07 AM

#

Would you make the limit configurable?

#

kekw

hasty plinth May 26, 2024, 9:07 AM

#

first

rain nimbus May 26, 2024, 9:07 AM

#

"Great minds think alike"

fiery turtle May 26, 2024, 9:07 AM

#

hasty plinth silly question, is the depth configurable?

absolutely!

hasty plinth May 26, 2024, 9:07 AM

#

and idiots never differ 😉

rain nimbus May 26, 2024, 9:07 AM

#

lmao

fiery turtle May 26, 2024, 9:08 AM

#

i know there would have to be this one aml blob that does some polling recursively

#

so it must be configurable

rain nimbus May 26, 2024, 9:09 AM

#

fiery turtle i know there would have to be this one aml blob that does some polling recursive...

void delay(size_t millis)
{
  if (millis == 0)
    return;
  Sleep(1);
  delay(millis - 1);
}

kekw

fiery turtle May 26, 2024, 9:09 AM

#

lol

rain nimbus May 26, 2024, 9:09 AM

#

cplusplus users in 2013 at its finest

fiery turtle May 26, 2024, 9:10 AM

#

lol ACPICA doesnt pass this test

#

uACPI can resume blob execution even after a fatal failure by unwinding back

#

acpica cannot here

rain nimbus May 26, 2024, 9:11 AM

#

What about microsoft's one?

fiery turtle May 26, 2024, 9:12 AM

#

i would have to pull up my kernel debugger for this...

rain nimbus May 26, 2024, 9:12 AM

#

I would almost assume they probably can handle it

fiery turtle May 26, 2024, 9:12 AM

#

but my guess is it has no recursion limit at all and would hang

fiery turtle May 26, 2024, 9:12 AM

#

rain nimbus I would almost assume they probably can handle it

yeah probably

#

fuck, lets test i guess

#

its going...

#

the debugger is not feeling well

fiery turtle May 26, 2024, 9:21 AM

#

fiery turtle but my guess is it has no recursion limit at all and would hang

looks like my theory was correct

#

but lets see

#

what should I call the status code?

#

i dont like this too much

#

but maybe its fine

rain nimbus May 26, 2024, 9:25 AM

#

fiery turtle looks like my theory was correct

Rather funky...

#

Maybe they just have a very large depth limit?

fiery turtle May 26, 2024, 9:25 AM

#

let me try without the debugger

#

because it just takes 10 yers

#

what i've learned with NT interpreter is that all they do is very very simple

#

so i wouldnt be surprised

#

(if theres no limit at all)

hollow crescent May 26, 2024, 9:27 AM

#

why would firmware have an infinite unbounded recursion wtf

hasty plinth May 26, 2024, 9:27 AM

#

"not our problem" meme

#

but good question

fiery turtle May 26, 2024, 9:27 AM

#

they already have infinite unbounded loops all over the place

#

so this wouldnt be new

fiery turtle May 26, 2024, 9:27 AM

#

hollow crescent why would firmware have an infinite unbounded recursion wtf

waiting on some register to change for example

hollow crescent May 26, 2024, 9:28 AM

#

with RECURSION??

fiery turtle May 26, 2024, 9:28 AM

#

you would use a while loop ofc but its impossible to predict what comes in their corrupted minds

hollow crescent May 26, 2024, 9:28 AM

#

rather than a while loop

fiery turtle May 26, 2024, 9:28 AM

#

also

#

aml is patched by bios at boot all the time

#

if it patches aml incorrectly it could produce an infinite recursion loop

#

its just too easy to fuck this up

rain nimbus May 26, 2024, 9:29 AM

#

rain nimbus ```cpp void delay(size_t millis) { if (millis == 0) return; Sleep(1); ...

With people advocating code like this, obviously it can very well happen

fiery turtle May 26, 2024, 9:29 AM

#

fiery turtle let me try without the debugger

still spinning...

rain nimbus May 26, 2024, 9:29 AM

#

Sounds like only uACPI handles it then lmao

fiery turtle May 26, 2024, 9:30 AM

#

qemu memory usage growing by 200kb every few seconds

#

i gave it 4g so this will take years

fiery turtle May 26, 2024, 9:30 AM

#

rain nimbus Sounds like only uACPI handles it then <:lmao:1166278910871359488>

acpica also handles it, but as a fatal abort

rain nimbus May 26, 2024, 9:30 AM

#

So the nt one just grows the stack when it reaches the stack limit?

fiery turtle May 26, 2024, 9:30 AM

#

i mean u have to be retarded to run aml recursively

#

it grows the internal call stack "array" or list or whatever

#

the actual interpreter runs on a normal fixed kernel stack

rain nimbus May 26, 2024, 9:31 AM

#

fiery turtle i mean u have to be retarded to run aml recursively

Very inefficient so yeah can agree

fiery turtle May 26, 2024, 9:31 AM

#

recursion in kernel is generally not allowed at all

#

same as VLAs

rain nimbus May 26, 2024, 9:32 AM

#

Makes sense

fiery turtle May 26, 2024, 9:32 AM

#

ACPICA is generally shit at this

rain nimbus May 26, 2024, 9:32 AM

#

Things just have to work

hollow crescent May 26, 2024, 9:32 AM

#

fiery turtle same as VLAs

VLA?

fiery turtle May 26, 2024, 9:32 AM

#

NT interpreter is pretty good i think, no recursions there

rain nimbus May 26, 2024, 9:32 AM

#

hollow crescent VLA?

C Variable Length Arrays

hollow crescent May 26, 2024, 9:32 AM

#

oh like alloca

rain nimbus May 26, 2024, 9:32 AM

#

Ye

fiery turtle May 26, 2024, 9:32 AM

#

yeah

rain nimbus May 26, 2024, 9:32 AM

#

Just without alloca

fiery turtle May 26, 2024, 9:32 AM

#

consuming an unpredictable amont of stack space

rain nimbus May 26, 2024, 9:32 AM

#

The compiler can generate whatever it wants to do the sub rsp

fiery turtle May 26, 2024, 9:33 AM

#

uACPI does it via a "vector"-like thing

#

with inline fixed storage

#

so like first 4 function calls don't allocate

#

after that each new frame is an allocation

rain nimbus May 26, 2024, 9:34 AM

#

Hmm, do you immediately free it when you exit the frame too?
Cuz that'd be kinda inefficient

fiery turtle May 26, 2024, 9:35 AM

#

nope

#

it does a pop

#

which just decrements size

#

its freed at the very end

rain nimbus May 26, 2024, 9:36 AM

#

Ah ok, so like if you enter a hundred frames, it will keep the memory forever?

fiery turtle May 26, 2024, 9:36 AM

#

yeah, until the very last function exits

#

its all stored in the "execution_context" struct

#

which is a few pages big

#

for most aml no dynamic allocations besides inline storage in execution context is needed

#

for more insane aml some dynamic memory is allocated on demand

#

its like a vector with two parts, inline storage and dynamic storage

#

fiery turtle May 27, 2024, 7:08 PM

#

Fixing some of the retarded uacpi_warn usages where it should clearly be uacpi_error, e.g.

fiery turtle May 29, 2024, 9:26 AM

#

Going to rewrite everything in tables.c

#

Its the sloppiest garbage i have in the entire code base

#

i did refactor it a bit in the past but its not on par with the rest of the project

#

no tracking of duplicate tables, no api to hook table installs, no override

#

I'm also not sure whether its worth to have an overcomplicated refcount per table and have it unmaped when the last reference is removed

#

like ACPICA does

#

tables are relatively tiny and the biggest ones like DSDTs must be mapped at all times

fiery turtle May 29, 2024, 10:09 AM

#

One thing uACPI is way better than ACPICA at is project structure

#

ACPICA has literally all public api dumped into one header, its impossible to tell whats what

#

and it has publically exported global variables that you're supposed to update directly

#

its also impossible to tell which api u need to implement, which is mandatory, which is not

#

like this is hell to port

#

You can really tell it's custom made for linux, and everything else is an afterthought

jaunty fox May 29, 2024, 12:47 PM

#

fiery turtle You can really tell it's custom made for linux, and everything else is an aftert...

it's a wonder it still exists as a formally independent project

#

echoes of DRM here

#

that was an independent project maintained in its own repository, then one day linux unilaterally decided to import the upstream code into the linux kernel monorepo and discontinue contributing back

fiery turtle May 29, 2024, 12:48 PM

#

Pretty much

loud ice May 29, 2024, 12:48 PM

#

i mean acpica is also inlined into linuxx

jaunty fox May 29, 2024, 12:48 PM

#

that was the death of portable DRM

fiery turtle May 29, 2024, 12:48 PM

#

Its maintained by one person that doesn't know much about the project too

jaunty fox May 29, 2024, 12:48 PM

#

loud ice i mean acpica is also inlined into linuxx

they have a script that automatically puts the Linux code style on it, reports say

loud ice May 29, 2024, 12:48 PM

#

how the fuck that does work

#

and also

#

what the fuck

fiery turtle May 29, 2024, 12:49 PM

#

Its a giant script

#

And they have a json describing every acpica struct

fiery turtle May 29, 2024, 12:49 PM

#

jaunty fox they have a script that automatically puts the Linux code style on it, reports s...

Its used for upstream sync, but nowadays most changes go directly into Linux bypassing acpica

jaunty fox May 29, 2024, 12:49 PM

#

https://github.com/acpica/acpica/tree/master/source/tools/acpisrc i believe this is the conversion tool

fiery turtle May 29, 2024, 12:50 PM

#

jaunty fox https://github.com/acpica/acpica/tree/master/source/tools/acpisrc i believe this...

https://github.com/acpica/acpica/blob/master/generate/linux/linuxize.sh

#

And it calls that yeah

jaunty fox May 29, 2024, 12:52 PM

#

i wonder why they can't just use the ACPICA style

fiery turtle May 29, 2024, 12:52 PM

#

Linus hates it probably

#

And it is pretty shit, they try to look like Microsoft

jaunty fox May 29, 2024, 12:52 PM

#

the loud TYPE_NAMES!! is offputting but it's not unbearable

fiery turtle May 29, 2024, 12:53 PM

#

Function names are all Pascal case

#

So is everything else

vast kestrel May 29, 2024, 1:58 PM

#

I mean, Linux MM has a ton of pascal case macros

#

And you can find inconsistent code styles everywhere in vendor code in the kernel

#

Most likely they decided to be nice and make it match

fiery turtle May 29, 2024, 2:13 PM

#

vast kestrel I mean, Linux MM has a ton of pascal case macros

Yeah I found that pretty weird lol

jaunty fox May 29, 2024, 3:30 PM

#

fiery turtle Yeah I found that pretty weird lol

wait till you struct task_struct

#

just in case you were in any doubt what kind of type that is

median crest May 29, 2024, 3:49 PM

#

I think its a struct but I'm not sure, maybe we should rename it to struct structure_task_struct

fiery turtle May 29, 2024, 5:08 PM

#

jaunty fox wait till you `struct task_struct`

i think it might be a struct

#

but maybe its also a task

north holly May 29, 2024, 5:10 PM

#

what if it were a union

vast kestrel May 29, 2024, 5:44 PM

#

nah its probably a union

fiery turtle May 29, 2024, 6:44 PM

#

my github action runs randomly get canceled??

#

anyone else seen that?

#

and that only happens on ubuntu for whatever reason

vagrant hull May 29, 2024, 6:48 PM

#

yeah it can happen i think

fiery turtle May 29, 2024, 7:00 PM

#

like canceled by whom lmao

loud ice May 29, 2024, 7:01 PM

#

github

#

probably because it's running on a preemptible instance

fiery turtle May 29, 2024, 7:03 PM

#

interesting, but I've never seen that before

#

and i did like at least a few hundred runs

#

maybe just luck

torpid ferry May 29, 2024, 7:05 PM

#

fiery turtle like canceled by whom lmao

afaik it's qos stuff, ci instances for free accs run on idle cpu time

#

so if there's a lot of paid gh work the thing can get preempted indefinitvely

vast kestrel May 29, 2024, 7:05 PM

#

pay microsoft moneies

fiery turtle May 29, 2024, 7:05 PM

#

that makes sense

torpid ferry May 29, 2024, 7:06 PM

#

vast kestrel pay microsoft moneies

inb4 it's all for bing ai queries :^)

fiery turtle May 30, 2024, 5:52 PM

#

#

Taking better naming suggestions but i cant come up with any

left orbit May 30, 2024, 6:17 PM

#

BAN -> DISALLOW?

fiery turtle May 30, 2024, 6:21 PM

#

is that better?

torpid ferry May 30, 2024, 6:21 PM

#

deny?

fiery turtle May 30, 2024, 6:21 PM

#

deny sounds cool

torpid ferry May 30, 2024, 6:21 PM

#

allow/deny makes it clearer imo

fiery turtle May 30, 2024, 6:22 PM

#

true

loud ice May 30, 2024, 8:27 PM

#

what if accept/drop/replace_(phys/virt)

fiery turtle May 30, 2024, 8:33 PM

#

interesting

hollow crescent May 31, 2024, 5:05 AM

#

I think drop would fit better

tired crater May 31, 2024, 5:08 AM

#

I don’t know how this disposition table thing works but drop kinda implies that some attribute or whatever gets, well, dropped

#

But depending on how this thingy works there is nothing actually being dropped

#

So deny would make more sense

vast kestrel May 31, 2024, 8:56 AM

#

I associate drop with networking too much, so in my mind ignore might sound better

loud ice May 31, 2024, 12:22 PM

#

accept/reject?

fiery turtle May 31, 2024, 9:52 PM

#

Okay i finally rewrote all of it, it's now much more advanced and thread safe

#

Added an ability to override tables dynamically + install/uninstall/load public api

#

The table subsystem was the weakest link, notw it no longer is

#

i need to split this up into commits and do some testing ofc because a ton of things have been changed

#

Don't u just love it when clion randomly decides that an entire function is unreachable code

#

its both unreachable and within a recursive call chain

vast kestrel May 31, 2024, 10:25 PM

#

Clion can do really weird things sometimes

#

Which I don't understand

#

Like I have had cases where it was omega smart and actually did find some weird recursive chains or checks that always fail

#

But sometimes it's stupid and just thinks "ah yes, this check is totally never going to pass"

fiery turtle May 31, 2024, 10:27 PM

#

Yup can confirm I've seen both

#

Mostly the latter

#

But as long as it sometimes finds bugs im fine with it

jaunty fox May 31, 2024, 10:30 PM

#

i should try this thing out

fiery turtle May 31, 2024, 10:36 PM

#

jaunty fox i should try this thing out

Yes definitely do, its very good

#

They give out free licenses for open source projects

#

You can apply and get one

vast kestrel May 31, 2024, 10:54 PM

#

I just bought a license with student discount smugFace

fiery turtle May 31, 2024, 10:55 PM

#

vast kestrel I just bought a license with student discount <:smugFace:356732242246107138>

Could've got one for free for tomatos probably

vast kestrel May 31, 2024, 10:56 PM

#

I don't have it worked on enough to have it consistent

fiery turtle May 31, 2024, 11:04 PM

#

Get back to work then 😡

fiery turtle Jun 1, 2024, 10:35 AM

#

Added a test for table overrides

#

vast kestrel Jun 1, 2024, 11:41 AM

#

Tbh this kinda makes me write a kernel that uses acpi for all the drivers

#

All the ability to load arbitrary buffers gives a lot of power

fiery turtle Jun 1, 2024, 11:46 AM

#

yeah it is pretty cool stuff

kindred beacon Jun 1, 2024, 11:55 AM

#

vast kestrel Tbh this kinda makes me write a kernel that uses acpi for all the drivers

finally, portable drivers!

#

ultrameme

vast kestrel Jun 1, 2024, 11:56 AM

#

Y e s

#

Tbh for legacy drivers with bank registers and index registers it's even a pretty decent

#

But with modern drivers where everything is just mmio it's kinda easier to just use native code

#

Oh yeah that's one problem, can acpi even get a pointer to a buffer

fiery turtle Jun 1, 2024, 11:58 AM

#

vast kestrel Oh yeah that's one problem, can acpi even get a pointer to a buffer

yeah

vast kestrel Jun 1, 2024, 11:58 AM

#

How

fiery turtle Jun 1, 2024, 11:58 AM

#

Depending what you mean exactly

vast kestrel Jun 1, 2024, 11:58 AM

#

You need to give hardware a pointer to a buffer

#

Now excluding pre allocated buffers that the firmware can allocate, does acpi on its own has the ability to allocate a buffer and get a pointer to it

fiery turtle Jun 1, 2024, 11:59 AM

#

hmm let me think

#

technically there isn't any way to extract an address of an object, but you can totally design an interface to do that

#

with a native call for example

#

or even like a Notify(obj, ADDRESS_OF)

#

if you agree on a value for ADDRESS_OF

#

or just make a method that will resolve to a native call that will return the address of a buffer passed into the function

#

there's already _OSI for calling into the os

#

you can totally reuse a similar thing

#

so yeah, totally possible

#

provided you add a few more os-defined functions

#

similar to NodeJS

vast kestrel Jun 1, 2024, 2:27 PM

#

yeah, it would need some extensions

torpid root Jun 7, 2024, 6:23 PM

#

@fiery turtle shouldn't these be info->flags & UACPI_NS_NODE_INFO_HAS_HID and info->flags & UACPI_NS_NODE_INFO_HAS_CID

fiery turtle Jun 7, 2024, 7:25 PM

#

torpid root <@353593948523986945> shouldn't these be ``info->flags & UACPI_NS_NODE_INFO_HAS_...

Yeah this is pre bitflags API

#

Outdated stuff

#

Thanks for pointing out

fiery turtle Jun 7, 2024, 10:16 PM

#

torpid root <@353593948523986945> shouldn't these be ``info->flags & UACPI_NS_NODE_INFO_HAS_...

I saw that u rebased to latest master, very nice

#

always pleasant to see commits with more code removed than added

fiery turtle Jun 7, 2024, 10:52 PM

#

added optional compile-time mutex tracing

#

i think i'll also add tracing for every single method returning a uacpi_status

fiery turtle Jun 8, 2024, 9:13 AM

#

so close

left orbit Jun 8, 2024, 9:14 AM

#

here, 99 :^)

#

let me log into my alt /s

fiery turtle Jun 8, 2024, 9:17 AM

#

lol

#

thank you

vagrant hull Jun 8, 2024, 9:18 AM

#

i would star it but i already have long ago :p

sterile egret Jun 8, 2024, 9:19 AM

#

Same^

fiery turtle Jun 8, 2024, 9:21 AM

#

much appreciated ❤️

sterile egret Jun 8, 2024, 9:23 AM

#

Now...when uAPM? trollfacePolite

fiery turtle Jun 8, 2024, 9:24 AM

#

lol

#

i think apm is realtively tiny(?) compared to acpi

#

the only problem is u need a bios call helper to drop down to real mode

sterile egret Jun 8, 2024, 9:26 AM

#

I might fiddle with it someday if I ever try running anything on one of my old laptops

fiery turtle Jun 8, 2024, 9:27 AM

#

i have an apm enable switch in bios for my 2019 pc motherboard

sterile egret Jun 8, 2024, 9:27 AM

#

lmao

fiery turtle Jun 8, 2024, 9:27 AM

#

its off by defautl but i mean lol

kind mantle Jun 8, 2024, 11:36 AM

#

You're welcome :P

fiery turtle Jun 8, 2024, 11:39 AM

#

kind mantle You're welcome :P

yay we did it

#

thank you robot

woven citrus Jun 9, 2024, 8:36 AM

#

fiery turtle thank you

Ima star

fiery turtle Jun 11, 2024, 8:22 PM

#

the TODO list grows along with my laziness

#

although i'm almost done with p.3

#

I constantly produce huge PRs then proceed to stall

#

because they appear to generally work but still need review and additional testing

kindred beacon Jun 12, 2024, 8:05 AM

#

Story of my life

fiery turtle Jun 12, 2024, 9:22 AM

#

Lol

fiery turtle Jun 12, 2024, 1:53 PM

#

I wonder if I should make uacpi_object an opaque struct like i did for uacpi_namespace_node

#

a lot of internal shit is leaked via the public types.h header which is something i dont like too much

kindred beacon Jun 12, 2024, 3:25 PM

#

fiery turtle I wonder if I should make `uacpi_object` an opaque struct like i did for `uacpi_...

I consider such structuring to be a massive W

#

the amount of people who rely on internals no matter what you say to prevent them from doing it is surprising

kind mantle Jun 12, 2024, 3:26 PM

#

kindred beacon the amount of people who rely on internals no matter what you say to prevent the...

Add a deprecation warning to all internal types so their builds will be spammed?

kindred beacon Jun 12, 2024, 3:26 PM

#

just make the structs opaque instead meme

kind mantle Jun 12, 2024, 3:28 PM

#

But yeah I know what you're talking about - I have a graphics library and wish I hadn't exposed internal types :P

#

Because it happens slightly too often that people try to use those (let alone change them) and get stuff they weren't expecting

fiery turtle Jun 12, 2024, 3:43 PM

#

kindred beacon the amount of people who rely on internals no matter what you say to prevent the...

yeah i know its the way forward, kind of scared of the amount of refactoring this would introduce tho

fiery turtle Jun 14, 2024, 8:59 PM

#

uACPI test runner would've been 10 times more advanced if C++ had anything like python argparse

#

its so fucking annoying that c++ has nothing to assist in argument parsing

jaunty fox Jun 14, 2024, 8:59 PM

#

fiery turtle its so fucking annoying that c++ has nothing to assist in argument parsing

getopt

fiery turtle Jun 14, 2024, 8:59 PM

#

its way way less advanced than argparse but yeah

hollow crescent Jun 14, 2024, 9:00 PM

#

fiery turtle its so fucking annoying that c++ has nothing to assist in argument parsing

maybe not in freestanding C++ but how about std::stringstream

fiery turtle Jun 14, 2024, 9:00 PM

#

oh im not talking about freestanding

#

its a normal userland program

jaunty fox Jun 14, 2024, 9:00 PM

#

hollow crescent maybe not in freestanding C++ but how about `std::stringstream`

that's just a string stream though, isn't it? no args parsing

#

i was being slightly facetious with getopt

hollow crescent Jun 14, 2024, 9:01 PM

#

jaunty fox that's just a string stream though, isn't it? no args parsing

it parses a string

jaunty fox Jun 14, 2024, 9:01 PM

#

but what's it that pytson's argsparse does so well?

hollow crescent Jun 14, 2024, 9:01 PM

#

what more would you want

jaunty fox Jun 14, 2024, 9:01 PM

#

hollow crescent it parses a string

i thought it just streamed one

fiery turtle Jun 14, 2024, 9:01 PM

#

in what way does it parse strings?

hollow crescent Jun 14, 2024, 9:01 PM

#

you can read from it as if you read from iostream

jaunty fox Jun 14, 2024, 9:01 PM

#

that's just streaming

hollow crescent Jun 14, 2024, 9:01 PM

#

if you have formatted data separated by spaces you can read it directly into its type

fiery turtle Jun 14, 2024, 9:01 PM

#

jaunty fox but what's it that pytson's argsparse does so well?

example:

def main() -> int:
    parser = argparse.ArgumentParser(description="Run uACPI tests")
    parser.add_argument("--asl-compiler",
                        help="Compiler to use to build test cases",
                        default="iasl")
    parser.add_argument("--acpi-extractor",
                        help="ACPI extractor utility to use for ACPI dumps",
                        default="acpixtract")
    parser.add_argument("--test-dir",
                        default=test_relpath("test-cases"),
                        help="The directory to run tests from, defaults to "
                             "'test-cases' in the same directory")
    parser.add_argument("--test-runner",
                        help="The test runner binary to invoke")
    parser.add_argument("--binary-directory",
                        default=test_relpath("bin"),
                        help="The directory to store intermediate files in, "
                             "created & deleted automatically. Defaults to "
                             "'bin' in the same directory")
    parser.add_argument("--bitness", default=64, choices=[32, 64], type=int,
                        help="uACPI build bitness")
    parser.add_argument("--large", action="store_true",
                        help="Run the large test suite as well")

#

default, choices, type

#

help

hollow crescent Jun 14, 2024, 9:02 PM

#

like cpp stringstream ss("42 something 4.0"); int x; string y; float z; ss >>x >>y >>z;

fiery turtle Jun 14, 2024, 9:02 PM

#

i can refer to the options as args.<name> etc

hollow crescent Jun 14, 2024, 9:02 PM

#

fiery turtle example: ```python def main() -> int: parser = argparse.ArgumentParser(descr...

wow that's pretty cool

#

while getopt is C I guess you can just use that if nothing else

jaunty fox Jun 14, 2024, 9:03 PM

#

fiery turtle example: ```python def main() -> int: parser = argparse.ArgumentParser(descr...

getopt_long a bit lacking then in conveniences compared to this

fiery turtle Jun 14, 2024, 9:03 PM

#

yea..

jaunty fox Jun 14, 2024, 9:03 PM

#

i'm surprised they havne't added something to c++ template library for this

fiery turtle Jun 14, 2024, 9:03 PM

#

and this also supports unnamed arguments, short arguments, all that stuff

#

just done for you automagically

jaunty fox Jun 14, 2024, 9:03 PM

#

they've sure as hell added everything else in the world to it

fiery turtle Jun 14, 2024, 9:03 PM

#

i know right

hollow crescent Jun 14, 2024, 9:03 PM

#

jaunty fox i'm surprised they havne't added something to c++ template library for this

Why would they

fiery turtle Jun 14, 2024, 9:03 PM

#

hence why im mad

hollow crescent Jun 14, 2024, 9:04 PM

#

Ok they added regexes, fair enough

fiery turtle Jun 14, 2024, 9:04 PM

#

that stack overflows and is a piece of crap

#

like msvc regex is known for just obliterating the stack

#

its unusable

#

i doubt glibc one is different

hollow crescent Jun 14, 2024, 9:05 PM

#

Fair point

#

I actually had to use boost regex at one point but it was due to performance, and not stack problems

fiery turtle Jun 14, 2024, 9:05 PM

#

performance too yeah

#

boost as bloated as it is at least has those implemented decently

#

it has a proper arg parser too iirc

hollow crescent Jun 14, 2024, 9:06 PM

#

fiery turtle boost as bloated as it is at least has those implemented decently

Oh no no I don't use the entirety of boost

#

Just the regex and base64 parts

#

I deleted the rest because they weren't necessary and I wasn't using them

fiery turtle Jun 14, 2024, 9:07 PM

#

i doubt anyone is using the entirety of boost

#

they would end up with a few gb worth of code

left orbit Jun 14, 2024, 9:19 PM

#

https://github.com/p-ranav/argparse maybe?

#

it's single header so it's easy to embed

#

also MIT

fiery turtle Jun 14, 2024, 9:19 PM

#

yeah this is pretty close to python argparse

#

somehow python stuff always has the best api in my experience

#

it just works TM

fiery turtle Jun 15, 2024, 11:10 PM

#

Finally got over my laziness and made a proper argument parser, and with that...

#

uACPI test runer now has the ability to load any number of extra tables

#

this is it loading all of 25 tables from my thinkpad T14S

#uACPI - a portable and easy-to-integrate ACPI implementation