uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 13

fiery turtle Dec 13, 2024, 9:37 PM

#

indeed

#

i actually found two bugs because of your blobs, and one random one by browsing code while looking for your bug

#

all fixes pushed

#

the rest were minor but still

calm latch Dec 13, 2024, 9:42 PM

#

(I'm trying to target Eee PC 700 now)

#

Wonder how that thing's ACPI is

fiery turtle Dec 13, 2024, 9:42 PM

#

have u tried on there?

calm latch Dec 13, 2024, 9:42 PM

#

It's 32 bit only

#

So no

fiery turtle Dec 13, 2024, 9:42 PM

#

that would be interesting

#

but its old so probably not that big

#

but quirky af most likely

#

so good test

#

can u get a quick dump from it?

calm latch Dec 13, 2024, 9:44 PM

#

No

#

It's my friend's laptop

fiery turtle Dec 13, 2024, 9:45 PM

#

unlucky

calm latch Dec 13, 2024, 9:45 PM

#

Ok I can't get TSC to cooperate

#

I'm too dumb for timers

fiery turtle Dec 13, 2024, 9:45 PM

#

get managarm gods or mathewnd to help u

calm latch Dec 13, 2024, 9:46 PM

#

I've stolen their code

#

And it doesn't work

fiery turtle Dec 13, 2024, 9:46 PM

#

steal it again until it works

calm latch Dec 13, 2024, 9:48 PM

#

I have frequency in nHz

#

And period in nHz^-1 (aka ns)

#

Why multiplying tsc * period gives me garbage

hollow elm Dec 13, 2024, 9:49 PM

#

nanohenry :^)

calm latch Dec 13, 2024, 9:49 PM

#

Too much electronics

#

Is 32949014999 a sane frequency of TSC?

frank canopy Dec 13, 2024, 9:52 PM

#

calm latch Why multiplying tsc * period gives me garbage

recommend doing it as a fraction, if youre using an int for the period thatll be losing precision

fiery turtle Dec 13, 2024, 9:52 PM

#

nah

#

it should be like 1000000000 or something i think?

calm latch Dec 13, 2024, 9:53 PM

#

frank canopy recommend doing it as a fraction, if youre using an int for the period thatll be...

I'm using fractions

frank canopy Dec 13, 2024, 9:53 PM

#

3600042kHz is my emu's tsc freq

fiery turtle Dec 13, 2024, 9:53 PM

#

fiery turtle it should be like 1000000000 or something i think?

or maybe thats lapic

calm latch Dec 13, 2024, 9:53 PM

#

Lapic is 625699999

#

According to my measurements against PIC

calm latch Dec 13, 2024, 9:55 PM

#

frank canopy 3600042kHz is my emu's tsc freq

So 32949014999 should be close

#

Relatively

median crest Dec 13, 2024, 9:55 PM

#

fiery turtle it should be like 1000000000 or something i think?

kvm timer is 1 ns ticks

frank canopy Dec 13, 2024, 9:56 PM

#

calm latch So 32949014999 should be close

about 10 times faster than what qemu is giving me if im counting digits right

fiery turtle Dec 13, 2024, 9:56 PM

#

median crest kvm timer is 1 ns ticks

ah

calm latch Dec 13, 2024, 9:56 PM

#

thinkong

#

I can't count

fiery turtle Dec 13, 2024, 9:56 PM

#

can u help pmos get a working timer mathewnd

median crest Dec 13, 2024, 9:57 PM

#

I mean my timers arent that good either halfmemeright

fiery turtle Dec 13, 2024, 9:57 PM

#

last time i checked astral time keeping is super accurate

calm latch Dec 13, 2024, 10:03 PM

#

Is there something obviously wrong here?https://gitlab.com/mishakov/pmos/-/blob/ia32/kernel/arch/x86_64/interrupts/apic.cc?ref_type=heads#L95

#

(I kinda don't want to deal with timers right now)

calm latch Dec 13, 2024, 10:05 PM

#

calm latch So 32949014999 should be close

This is 9.15 times faster than the theoretical value

#

So 5,644,145/9.15 = 616'846 ops/s

vast kestrel Dec 13, 2024, 10:09 PM

#

what you cooking

slim panther Dec 13, 2024, 10:09 PM

#

#1316457641353936956

vast kestrel Dec 13, 2024, 10:10 PM

#

damn

fiery turtle Dec 13, 2024, 10:11 PM

#

yeah we're thinking of a common kernel api structure

#

not sure yet how its going to look

vast kestrel Dec 13, 2024, 10:12 PM

#

I mean, the uacpi kernel api is pretty good

fiery turtle Dec 13, 2024, 10:12 PM

#

problem is, some drivers need more, some need less

vast kestrel Dec 13, 2024, 10:12 PM

#

and it provides most things that a driver would need (ignoring actual subsystem abstractions)

fiery turtle Dec 13, 2024, 10:12 PM

#

e.g. udrm will need a ton more stuff

#

phys page allocation, bar mapping, msi, cpu affinity, threads, work queues etc

vast kestrel Dec 13, 2024, 10:13 PM

#

isn't work queues already abstracted?

fiery turtle Dec 13, 2024, 10:13 PM

#

sorta

vast kestrel Dec 13, 2024, 10:13 PM

#

you have the schedule work or something alike

fiery turtle Dec 13, 2024, 10:13 PM

#

yeah

#

but i dont control when or how they're created

vast kestrel Dec 13, 2024, 10:13 PM

#

bar mapping is basically the normal map function

fiery turtle Dec 13, 2024, 10:13 PM

#

not for arm stuff

vast kestrel Dec 13, 2024, 10:14 PM

#

true, the arm pcie controllers are garbage

fiery turtle Dec 13, 2024, 10:14 PM

#

yea

vast kestrel Dec 13, 2024, 10:14 PM

#

ecam was created for a freaking reason

#

but nooooo

#

you need to be special and have cfg0/cfg1 memes

#

and the weird offsets

#

thing is that defining too much and you might as well define how the kernel should behave

calm latch Dec 13, 2024, 10:18 PM

#

vast kestrel you need to be special and have cfg0/cfg1 memes

I think that's one of the things RISC-V did right

fiery turtle Dec 15, 2024, 1:26 PM

#

alright im on vacation so time to hardcore push uACPI release 1.0

#

first things is ironing out final thread safety stuff

#

So far fixed this one

slim panther Dec 15, 2024, 1:36 PM

#

fiery turtle alright im on vacation so time to hardcore push uACPI release 1.0

https://media.discordapp.net/attachments/393177629886251018/797999295885869106/image0-45.gif

fiery turtle Dec 15, 2024, 1:56 PM

#

the unfortunate news is some problems are unsolvable without a recursive mutex

slim panther Dec 15, 2024, 1:57 PM

#

ew

mortal yoke Dec 15, 2024, 1:57 PM

#

can't you just internally keep the recursion count?

fiery turtle Dec 15, 2024, 1:58 PM

#

because there's handler install code that needs the ns write lock, then releases it for the duration of calling a user's handler because that might also call into the interpreter, but the problem is when it does it the installation is in half-finished state

fiery turtle Dec 15, 2024, 1:58 PM

#

mortal yoke can't you just internally keep the recursion count?

yeah ofc and i already do for aml mutexes

#

its just unfortunate overall

#

i have all the kernel api to implement one

#

normal mutex + thread_id getter

fiery turtle Dec 15, 2024, 2:00 PM

#

mortal yoke can't you just internally keep the recursion count?

uacpi_status uacpi_acquire_aml_mutex(uacpi_mutex *mutex, uacpi_u16 timeout)
{
    uacpi_thread_id this_id;
    uacpi_status ret = UACPI_STATUS_OK;

    this_id = uacpi_kernel_get_thread_id();
    if (UACPI_ATOMIC_LOAD_THREAD_ID(&mutex->owner) == this_id) {
        if (uacpi_unlikely(mutex->depth == 0xFFFF)) {
            uacpi_warn(
                "failing an attempt to acquire mutex @%p, too many recursive "
                "acquires\n", mutex
            );
            return UACPI_STATUS_DENIED;
        }

        mutex->depth++;
        return ret;
    }

    uacpi_namespace_write_unlock();
    ret = uacpi_acquire_native_mutex_with_timeout(mutex->handle, timeout);
    if (ret != UACPI_STATUS_OK)
        goto out;

    if (mutex->handle == g_uacpi_rt_ctx.global_lock_mutex->handle) {
        ret = uacpi_acquire_global_lock_from_firmware();
        if (uacpi_unlikely_error(ret)) {
            uacpi_release_native_mutex(mutex->handle);
            goto out;
        }
    }

    UACPI_ATOMIC_STORE_THREAD_ID(&mutex->owner, this_id);
    mutex->depth = 1;

out:
    uacpi_namespace_write_lock();
    return ret;
}

#

this is already existing code, but looks like ill have to add similar code for native mutexes

fiery turtle Dec 15, 2024, 2:44 PM

#

void u()
{
    while (1) {

        uacpi_install_address_space_handler(
            uacpi_namespace_root(), UACPI_ADDRESS_SPACE_PCC,
            handle_ec, nullptr
        );
    }
}

void uu()
{
    while (1) {

        uacpi_uninstall_address_space_handler(
            uacpi_namespace_root(), UACPI_ADDRESS_SPACE_PCC
        );
    }
}

std::thread t0(u);
std::thread t1(uu);
t0.join();

#

least insane test

#

/mnt/d/uACPI/source/io.c:231:35: runtime error: member access within null pointer of type 'struct uacpi_address_space_handler'
AddressSanitizer:DEADLYSIGNAL
=================================================================
==2566==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000010 (pc 0x564416b1ad16 bp 0x7f5918afe310 sp 0x7f5918afe1e0 T1)
==2566==The signal is caused by a READ memory access.
==2566==Hint: address points to the zero page.
    #0 0x564416b1ad16 in dispatch_field_io /mnt/d/uACPI/source/io.c:231
    #1 0x564416b1bb92 in access_field_unit /mnt/d/uACPI/source/io.c:312
    #2 0x564416b1d35b in uacpi_write_field_unit /mnt/d/uACPI/source/io.c:461
    #3 0x564416acf837 in object_assign_with_implicit_cast /mnt/d/uACPI/source/interpreter.c:978
    #4 0x564416adde91 in store_to_reference /mnt/d/uACPI/source/interpreter.c:2332
    #5 0x564416af65ee in store_to_target /mnt/d/uACPI/source/interpreter.c:4194
    #6 0x564416af6acf in handle_copy_object_or_store /mnt/d/uACPI/source/interpreter.c:4227
    #7 0x564416affa20 in exec_op /mnt/d/uACPI/source/interpreter.c:5403
    #8 0x564416b02b7d in uacpi_execute_control_method /mnt/d/uACPI/source/interpreter.c:5755
    #9 0x564416b0ffa9 in region_run_reg /mnt/d/uACPI/source/opregion.c:143
    #10 0x564416b12b51 in do_run_reg /mnt/d/uACPI/source/opregion.c:451
    #11 0x564416b0897a in uacpi_namespace_do_for_each_child /mnt/d/uACPI/source/namespace.c:876
    #12 0x564416b13249 in reg_or_unreg_all_opregions /mnt/d/uACPI/source/opregion.c:506
    #13 0x564416b13ef3 in uacpi_install_address_space_handler /mnt/d/uACPI/source/opregion.c:642

#

this is what im trying to protect against

#

this is my AML blob:

Device (X) {
    OperationRegion (Y, PCC, 0, 100)

    Device (X) { OperationRegion (Y, PCC, 0, 100)
        Device (Z) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (A) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (C) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (D) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (B) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (F) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (G) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (H) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (L) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}

    }
    Device (Z) { OperationRegion (Y, PCC, 0, 100) }
    Device (A) { OperationRegion (Y, PCC, 0, 100) }
    Device (C) { OperationRegion (Y, PCC, 0, 100) }
    Device (D) { OperationRegion (Y, PCC, 0, 100) }
    Device (B) { OperationRegion (Y, PCC, 0, 100) }
    Device (F) { OperationRegion (Y, PCC, 0, 100)
        Device (Z) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (A) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (C) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (D) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (B) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (F) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (G) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (H) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
        Device (L) { OperationRegion (Y, PCC, 0, 100) Method (_REG, 2) { Field (Y, AnyAcc, NoLock, WriteAsZeros) { Z, 1 } Z = 1 }}
    }
    Device (G) { OperationRegion (Y, PCC, 0, 100) }
    Device (H) { OperationRegion (Y, PCC, 0, 100) }
    Device (L) { OperationRegion (Y, PCC, 0, 100) }
}

#

    uacpi_namespace_write_unlock();
    ret = handler->callback(UACPI_REGION_OP_ATTACH, &attach_data);
    uacpi_namespace_write_lock();

    if (region->handler == UACPI_NULL) {
        uacpi_error("Lol handler is gone while we were attaching it\n");
    }

#

this is what im talking about

#

why i need recursive mutexes

#

[uACPI][ERROR] Lol handler is gone while we were attaching it

#

at least some mutex must be held while we're calling into a user-provided handler

#

    uacpi_object_ref(obj);
    uacpi_namespace_write_unlock();
    ret = handler->callback(UACPI_REGION_OP_ATTACH, &attach_data);
    uacpi_namespace_write_lock();

    // Handler might've disappeared while we were attaching it
    if (uacpi_unlikely(region->handler == UACPI_NULL))
        return UACPI_STATUS_NO_HANDLER;

Hmm this check technically fixes the segfault and i can cleanly abort in this case

#

i wonder if this could be the way out

rustic compass Dec 15, 2024, 3:18 PM

#

you are removing the only excuse i had to procrastinate and not work on the uacpi rust bindings nooo

fiery turtle Dec 15, 2024, 3:19 PM

#

99% of the api is final btw

#

so u have no excuse meme

rustic compass Dec 15, 2024, 3:19 PM

#

death

fiery turtle Dec 15, 2024, 3:34 PM

#

[uACPI][TRACE] deactivated all 'PCC' opregions controlled by '\ ', 29 _REG() calls (18 errors)

#

now thats what i call a nice race condition

#

18 races out of 29 calls

#

but seems to not crash at least

#

my _REG serialization solution was bogus and causes deadlock tho so thats nice

#

whatever time to go workout and get some food

deft canopy Dec 15, 2024, 3:40 PM

#

U can do it infy, i believe in u

fiery turtle Dec 15, 2024, 3:41 PM

#

thanks

#

this is an unsolved problem in acpica so i get brag rights

#

if i do it that is

deft canopy Dec 15, 2024, 3:43 PM

#

You've already got quite a few bragging rights over ACPICA

fiery turtle Dec 15, 2024, 3:54 PM

#

lol

frank canopy Dec 15, 2024, 4:13 PM

#

fiery turtle this is already existing code, but looks like ill have to add similar code for n...

if you do need reentrant kernel mutexes, can you provide an option to use ones the kernel already has instead of forcing uacpi to implement them?

#

my imaginarium ones already have that option (though I think it's actually bugged right now lmao)

fiery turtle Dec 15, 2024, 4:52 PM

#

frank canopy if you do need reentrant kernel mutexes, can you provide an option to use ones t...

possibly, but I'm not sure that's useful given i have all the primitives needed for them?

frank canopy Dec 15, 2024, 4:53 PM

#

if you want to do it yourself then sure lol

fiery turtle Dec 15, 2024, 4:53 PM

#

i already had to implement a special-cased version for AML mutex objects so kinda

#

nice, my acpica fix for linux was added to a bunch of linux stable branches

vagrant hull Dec 15, 2024, 5:04 PM

#

big W

fiery turtle Dec 15, 2024, 5:04 PM

#

thanks

#

this was a type of bug that would make keyronex panic, but I guess linux had a workaround for it

#

so it never manifested itself in prod

fiery turtle Dec 15, 2024, 11:31 PM

#

okay i think i got it finally

#

    std::thread t0([] {
        while (1) {
            uacpi_install_address_space_handler(
                uacpi_namespace_root(), UACPI_ADDRESS_SPACE_PCC,
                handle_ec, nullptr
            );
        }
    });
    std::thread t1([] {
        while (1) {
            uacpi_uninstall_address_space_handler(
                uacpi_namespace_root(), UACPI_ADDRESS_SPACE_PCC
            );
        }
    });
    t0.join();

#

this is now fully thread safe and serialized in uacpi

#

just so u know the amount of work this does on install:

attach a new address space handler to the device object
scan every opregion node under this device and attach the handler to it
run _REG for every new opregion found
if this opregion is accessed, dispatch to user handler first with OP_ATTACH, then with OP_READ/WRITE, both release the interpreter lock for the duration

#

and uninstall does the same, but in reverse, while also calling into the user handler with OP_DETACH

#

so its insane to serialize properly

#

static uacpi_status upgrade_to_opregion_lock(void)
{
    uacpi_status ret;

    /*
     * Drop the namespace lock, and reacquire it after the opregion lock
     * so we keep the ordering with user API.
     */
    uacpi_namespace_write_unlock();

    ret = uacpi_recursive_lock_acquire(&g_opregion_lock);
    uacpi_namespace_write_lock();
    return ret;
}

I have this atrocity where the interpreter might call into the opregion subsystem tho

#

attaching EC 17028
[uACPI][TRACE] write to [\X___.X___.Z___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17029
[uACPI][TRACE] write to [\X___.X___.A___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17030
[uACPI][TRACE] write to [\X___.X___.C___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17031
[uACPI][TRACE] write to [\X___.X___.D___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17032
[uACPI][TRACE] write to [\X___.X___.B___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17033
[uACPI][TRACE] write to [\X___.X___.F___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17034
[uACPI][TRACE] write to [\X___.X___.G___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17035
[uACPI][TRACE] write to [\X___.X___.H___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17036
[uACPI][TRACE] write to [\X___.X___.L___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17037
[uACPI][TRACE] write to [\X___.F___.Z___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17038
[uACPI][TRACE] write to [\X___.F___.A___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17039
[uACPI][TRACE] write to [\X___.F___.C___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17040
[uACPI][TRACE] write to [\X___.F___.D___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17041
[uACPI][TRACE] write to [\X___.F___.B___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17042
[uACPI][TRACE] write to [\X___.F___.F___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17043
[uACPI][TRACE] write to [\X___.F___.G___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17044
[uACPI][TRACE] write to [\X___.F___.H___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
attaching EC 17045
[uACPI][TRACE] write to [\X___.F___.L___.Y___] (1 bytes) PCC[0x0000000000000000] = 0x1
[uACPI][TRACE] activated all 'PCC' opregions controlled by '\   ', 29 _REG() calls (0 errors)
detaching EC 17046
detaching EC 17047
detaching EC 17048
detaching EC 17049
detaching EC 17050
detaching EC 17051
detaching EC 17052
detaching EC 17053
detaching EC 17054
detaching EC 17055
detaching EC 17056
detaching EC 17057
detaching EC 17058
detaching EC 17059
detaching EC 17060
detaching EC 17061
detaching EC 17062
detaching EC 17063

At least all logs are like this now, no races

flat badge Dec 15, 2024, 11:53 PM

#

fiery turtle ```cpp static uacpi_status upgrade_to_opregion_lock(void) { uacpi_status ret...

how is this safe?

#

this looks like a weird mechanism

#

you drop the namespace_write lock, so other threads can take it now

#

but if it's fine for other threads to take it, why do you have it in the first place?

fiery turtle Dec 16, 2024, 12:03 AM

#

flat badge but if it's fine for other threads to take it, why do you have it in the first p...

lock not held means the namespace state is unknown, temporary nodes referenced might've been destroyed, objects held might've been deleted since the owner node got copyobject'ed into

it's safe to drop the lock as long as any objects needed for the duration of the operation have an extra reference keeping them alive and when we don't intend to modify the namespace

#

in this case i dont do either of those, and i verify that the object that i have with namespace locked is still what i want, e.g. that it didnt get overwritten by other aml during lock upgrade

#

so basically without the ns lock held its unsafe to:

scan the namespace
access objects at permanent namespace nodes
access temporary namespace nodes

flat badge Dec 16, 2024, 12:09 AM

#

i see

fiery turtle Dec 16, 2024, 12:11 AM

#

flat badge i see

the problem is there's user api to manipulate opregion handlers, and it calls into the interpreter, which may call back into opregion code, which may in turn need to run the user IO handler (for the duration of which i have to release the namespace lock so user can run aml inside handlers), during this window anyone can call uninstall_handler, which will uninstall the handler that is currently being used for AML IO

#

so there has to be some mutex ensuring serialized access to opregion handlers

#

which the ns lock is not good enough for because user handlers require releasing it

#

this is further made more difficult by the fact that a device might have any number of opregions underneath, and all must be regged and unregged in a seralized manner as well

#

idk if what im saying makes sense, its kinda hard to understand without the full context

fiery turtle Dec 16, 2024, 12:16 AM

#

fiery turtle the problem is there's user api to manipulate opregion handlers, and it calls in...

ACPICA also drops the ns lock for user handler IO, but it doesn't care to protect against anyone calling handler_uninstall() from a different thread during this window, so u can blow it up quite easily, nor does it serialize reg/unreg

fiery turtle Dec 16, 2024, 12:37 AM

#

obligatory astral test

#

while i wait for CI

#

@north holly does obos currently boot on real hw

north holly Dec 16, 2024, 12:38 AM

#

fiery turtle <@841762704766140486> does obos currently boot on real hw

Yes

fiery turtle Dec 16, 2024, 12:38 AM

#

do u want to test new thread safety stuff meme

north holly Dec 16, 2024, 12:38 AM

#

Sure

fiery turtle Dec 16, 2024, 12:38 AM

#

thanks

#

git fetch && git checkout thread-safety-p2

north holly Dec 16, 2024, 12:39 AM

#

Mhm

fiery turtle Dec 16, 2024, 12:39 AM

#

no api changes

north holly Dec 16, 2024, 12:48 AM

#

K

north holly Dec 16, 2024, 12:49 AM

#

fiery turtle do u want to test new thread safety stuff <:meme:575445094589661215>

It worked on real hw

fiery turtle Dec 16, 2024, 12:49 AM

#

nice

north holly Dec 16, 2024, 12:49 AM

#

Unless

#

Why isn't it loading my com driver

#

I will debug that

fiery turtle Dec 16, 2024, 12:50 AM

#

nooo my com collection

#

Anyways thanks for testing

#

All tests passed as well

#

Alright thats one huge roadblock before 1.0 solved

#

The last non-thread-safe thing in uACPI is events

#

I will fix that tomorrow

#

Only took like 6 hours today

#

Figuring out what code paths are possible

#

Im not a giant fan of the solution but i cant currently even theoretically think of a different one

#

frank canopy Dec 16, 2024, 1:22 AM

#

nice

median crest Dec 16, 2024, 1:31 AM

#

fiery turtle obligatory astral test

I really need to figure out whats making it so slow

fiery turtle Dec 16, 2024, 10:50 AM

#

Ok so

#

Event thread safety

#

I will have to change the meaning of uacpi_kernel_wait_for_work_completion

#

It will have to be also wait for in-flight ACPI irqs to finish

#

basically similar to synchronize_hardirq(); in linux

#

otherwise i cant guarantee that the interrupt handler won't touch event handlers that are being deleted

#

/*
 * Waits for two types of work to finish:
 * - All work scheduled via uacpi_kernel_schedule_work
 * - All in-flight interrupts installed via uacpi_kernel_install_interrupt_handler
 */
uacpi_status uacpi_kernel_wait_for_work_completion(void);

#

basically like this

#

so event handler deinstallation will do:

mask the event in the GPE register block
wait for in-flight interrupts to complete
unregister the handler
restore the old native handler if one existed
unmask

calm latch Dec 16, 2024, 12:50 PM

#

Is EC driver basically just functions to read and write to/from its registers, passed by uACPI?

fiery turtle Dec 16, 2024, 2:09 PM

#

calm latch Is EC driver basically just functions to read and write to/from its registers, p...

yup

calm latch Dec 16, 2024, 2:09 PM

#

Can there be more than one controller??

fiery turtle Dec 16, 2024, 2:10 PM

#

technically nothing prevents it

#

but almost impossible

#

managarm supports only one i think

calm latch Dec 16, 2024, 2:11 PM

#

The ACPI standard supports multiple embedded controllers in a system, each with its own resources

fiery turtle Dec 16, 2024, 2:12 PM

#

sure why wouldnt the standard support it

calm latch Dec 16, 2024, 2:16 PM

#

I guess I'll support multiple because why not

fiery turtle Dec 16, 2024, 2:17 PM

#

im not sure linux supports multiple tbh

#

no it does

calm latch Dec 16, 2024, 2:22 PM

#

It doesn't look difficult to do

calm latch Dec 16, 2024, 2:55 PM

#

Does it not pass ECDT?

fiery turtle Dec 16, 2024, 3:00 PM

#

your laptop doesnt have ECDT

calm latch Dec 16, 2024, 5:14 PM

#

What does global lock refer to in https://github.com/managarm/managarm/blob/275c0704a05cb904ae10d8964c6a300ac30d2dc7/kernel/thor/system/acpi/ec.cpp#L297 ?

GitHub

managarm/kernel/thor/system/acpi/ec.cpp at 275c0704a05cb904ae10d896...

Pragmatic microkernel-based OS with fully asynchronous I/O - managarm/managarm

fiery turtle Dec 16, 2024, 5:20 PM

#

global ACPI firmware lock

fiery turtle Dec 16, 2024, 5:22 PM

#

calm latch What does global lock refer to in https://github.com/managarm/managarm/blob/275c...

uacpi_acquire_global_lock

calm latch Dec 16, 2024, 5:23 PM

#

What do I need it for?

#

(where am I supposed to acquire it)

#

Did anyone notice that managram doesn't initialize gpeNode?

#

Is that intended?

fiery turtle Dec 16, 2024, 5:44 PM

#

calm latch What do I need it for?

You dont

#

The firmware might ask you to lock it

#

When u access EC

fiery turtle Dec 16, 2024, 5:45 PM

#

calm latch Is that intended?

Yes, the null node is allowed, take a look at uacpi event header

#

Its resolved into _GPE automatically

calm latch Dec 16, 2024, 5:49 PM

#

I've already looked at it

#

I guess they have clearing allocator

#

Can EC be in memory? Thonk

fiery turtle Dec 16, 2024, 5:50 PM

#

calm latch I guess they have clearing allocator

The what

fiery turtle Dec 16, 2024, 5:50 PM

#

calm latch Can EC be in memory? <:Thonk:1271231807945441402>

Who knows

calm latch Dec 16, 2024, 5:52 PM

#

fiery turtle The what

calloc

#

Since managram only takes UACPI_RESOURCE_TYPE_IO and UACPI_RESOURCE_TYPE_FIXED_IO

#

assertos

fiery turtle Dec 16, 2024, 5:53 PM

#

Yeah I doubt other exist but u can take a look at which ones Linux expects

fiery turtle Dec 16, 2024, 5:55 PM

#

calm latch Since managram only takes UACPI_RESOURCE_TYPE_IO and UACPI_RESOURCE_TYPE_FIXED_I...

https://elixir.bootlin.com/linux/v6.12.5/source/drivers/acpi/ec.c#L1775

ec.c - drivers/acpi/ec.c - Linux source code v6.12.5...

Elixir Cross Referencer - source code of Linux v6.12.5: drivers/acpi/ec.c

vast kestrel Dec 16, 2024, 6:01 PM

#

Btw why doesn't the aml just implement the EC driver? Is there an actual reason or just yet another acpi weirdness?

fiery turtle Dec 16, 2024, 6:02 PM

#

vast kestrel Btw why doesn't the aml just implement the EC driver? Is there an actual reason ...

it could I guess, fundamentally it's an asynchronous device, which might take an unpredicatble amount of time to respond so I guess they decided to offload that to the kernel

vast kestrel Dec 16, 2024, 6:03 PM

#

But acpi has everything it needs to handle async code

#

It has the event mechanism and the gpe mechanism

fiery turtle Dec 16, 2024, 6:03 PM

#

yeah

#

it would require non-trivial aml with events and stuff, I guess firmware devs are not at that level yet

vast kestrel Dec 16, 2024, 6:04 PM

#

The only actual reason I can think of is if maybe the OEM wants its own driver to do some EC stuff, but they could have just used the aml driver for both

#

Shrug

fiery turtle Dec 16, 2024, 6:05 PM

#

one other reason i can think of is EC is used directly by the host to request hw specific information about thermals and firmware settings etc

#

maybe they did it to avoid races with host

loud ice Dec 16, 2024, 6:05 PM

#

i think the real reason is because aml is a shitty programming language

vast kestrel Dec 16, 2024, 6:05 PM

#

Is there a reason uacpi doesn't provide it's own EC driver?

loud ice Dec 16, 2024, 6:05 PM

#

and vendors dont want to write even more aml

fiery turtle Dec 16, 2024, 6:05 PM

#

already answered before

vast kestrel Dec 16, 2024, 6:05 PM

#

Lemme search then

fiery turtle Dec 16, 2024, 6:06 PM

#

#1217009725711847465 message

vast kestrel Dec 16, 2024, 6:06 PM

#

How the fuck do you search discord on mobile

#

Bruh I asked that

fiery turtle Dec 16, 2024, 6:06 PM

#

lmao

north holly Dec 16, 2024, 6:06 PM

#

vast kestrel How the fuck do you search discord on mobile

It doesn't work in threada

loud ice Dec 16, 2024, 6:07 PM

#

like in general

vast kestrel Dec 16, 2024, 6:07 PM

#

Tbh I can't really agree with the quirk thing because if you won't handle it no one would, but the coupling with the kernel I can understand

fiery turtle Dec 16, 2024, 6:07 PM

#

handling it requires uacpi to have SMBIOS hooks in kernel api

#

so basically just too many extra kernel api functions needed for that

vast kestrel Dec 16, 2024, 6:08 PM

#

Yeah that sucks

#

Could be a nice uDriver maybe :^)

fiery turtle Dec 16, 2024, 6:08 PM

#

yeah definitely

#

cool idea

vast kestrel Dec 16, 2024, 6:08 PM

#

Given for that you will def need smbios hooks

#

Like uapi

fiery turtle Dec 16, 2024, 6:09 PM

#

uapi will definitely need platform info yeah

vast kestrel Dec 16, 2024, 6:09 PM

#

Its funny because I am not sure if I will even use any of these stuff in my C# kernel giving the whole goal is to have managed drivers 3sGudaEhehe

fiery turtle Dec 16, 2024, 6:09 PM

#

C# bindings for uAPI meme

vast kestrel Dec 16, 2024, 6:10 PM

#

I am actually kinda waiting to see the API to see if I can take inspiration from it lol

fiery turtle Dec 16, 2024, 6:10 PM

#

lol

#

nice

vast kestrel Dec 16, 2024, 6:10 PM

#

Like I need "safe" abstractions over interrupts and hardware resources

#

And I wasn't that impressed with the rust Linux bindings for inspiration

fiery turtle Dec 16, 2024, 6:10 PM

#

lol

#

i can imagine that

vast kestrel Dec 16, 2024, 6:11 PM

#

The closest to a nice api inspiration is WDK from all things

fiery turtle Dec 16, 2024, 6:11 PM

#

yup

#

storport meme

vast kestrel Dec 16, 2024, 6:11 PM

#

Ah yes I too like a switch case over every kernel function because ?????

fiery turtle Dec 16, 2024, 6:13 PM

#

ikr

#

not a fan of that giant switch

calm latch Dec 16, 2024, 6:13 PM

#

fiery turtle The firmware might ask you to lock it

Where would it be asking?

fiery turtle Dec 16, 2024, 6:14 PM

#

_GLK evaluating to 1

#

https://elixir.bootlin.com/linux/v6.12.5/source/drivers/acpi/ec.c#L1483

ec.c - drivers/acpi/ec.c - Linux source code v6.12.5...

Elixir Cross Referencer - source code of Linux v6.12.5: drivers/acpi/ec.c

calm latch Dec 16, 2024, 6:15 PM

#

Oh so basically when doing reads and writes?

fiery turtle Dec 16, 2024, 6:15 PM

#

ye

calm latch Dec 16, 2024, 6:17 PM

#

But not in GPE handler

fiery turtle Dec 17, 2024, 12:34 PM

#

#

uACPI will be fully thread safe after I merge this PR!

#

it's basically done, just need to test it now

#

to test this properly ill need managarm because no one else uses events properly

calm latch Dec 17, 2024, 12:40 PM

#

Are they difficult to implement?

fiery turtle Dec 17, 2024, 12:41 PM

#

well the managarms ec driver

#

did u implement one?

calm latch Dec 17, 2024, 12:41 PM

#

No

#

I'll finish it today

north holly Dec 17, 2024, 12:41 PM

#

I should implement ec driver

calm latch Dec 17, 2024, 12:41 PM

#

I've started writing it

fiery turtle Dec 17, 2024, 12:42 PM

#

ok nice

fiery turtle Dec 17, 2024, 12:42 PM

#

north holly I should implement ec driver

yes

calm latch Dec 17, 2024, 12:42 PM

#

calm latch I'll finish it today

But I have an exam now...

north holly Dec 17, 2024, 12:42 PM

#

Like maybe before adding the new syscalls

fiery turtle Dec 17, 2024, 12:44 PM

#

i can test obos on like 5 laptops if u do it

#

ill have to pull them out and find chargers tho

calm latch Dec 17, 2024, 12:45 PM

#

I have a bunch of PCs

#

8

fiery turtle Dec 17, 2024, 12:45 PM

#

damn

calm latch Dec 17, 2024, 12:45 PM

#

I think

fiery turtle Dec 17, 2024, 12:45 PM

#

does your os boot on all?

calm latch Dec 17, 2024, 12:45 PM

#

Yes

fiery turtle Dec 17, 2024, 12:45 PM

#

damn

#

i guess that includes uacpi?

north holly Dec 17, 2024, 12:45 PM

#

fiery turtle i can test obos on like 5 laptops if u do it

Oh cool

#

Will be doing that soon

calm latch Dec 17, 2024, 12:45 PM

#

Yes

#

Although I haven't tested it for a long time

fiery turtle Dec 17, 2024, 12:46 PM

#

nice, if u add an ec driver that should be a very large coverage

#

if u do fn+various keys it should spam EC events

calm latch Dec 17, 2024, 12:46 PM

#

How would I see them?

fiery turtle Dec 17, 2024, 12:46 PM

#

same for AC charger/lid and stuff like that if u have laptops

calm latch Dec 17, 2024, 12:46 PM

#

calm latch 8

One of them broke limine lmao

fiery turtle Dec 17, 2024, 12:47 PM

#

https://github.com/managarm/managarm/blob/cdc7ca0c631b222502a44ac151f43cfe66178bcd/kernel/thor/system/acpi/ec.cpp#L168

GitHub

managarm/kernel/thor/system/acpi/ec.cpp at cdc7ca0c631b222502a44ac1...

Pragmatic microkernel-based OS with fully asynchronous I/O - managarm/managarm

#

the EC makes u run AML

#

that it wants

calm latch Dec 17, 2024, 12:47 PM

#

Yeah my code is very similar ~~(copy-pasted)~~

#

Except in C

calm latch Dec 17, 2024, 12:49 PM

#

calm latch 8

Although I'll only be able to test it on the weekend

#

And potentially not all of them are working

#

Like my old laptop died a month ago sadmeme

#

But it's a battery problem so I can disassemble it and disconnect it I'd suppose

#

And it's a weird HP 2 in 1, which struggles with sound in Linux

#

Anyway, I think I'll try to finish it today's evening, and at least try it on my Lenovo laptop

fiery turtle Dec 17, 2024, 12:56 PM

#

okay astral revealed a slight skill issue but seems to be fixed now

#

#

seems to work

fiery turtle Dec 17, 2024, 12:57 PM

#

calm latch Like my old laptop died a month ago <:sadmeme:659848285510565907>

F

fiery turtle Dec 17, 2024, 12:57 PM

#

calm latch Anyway, I think I'll try to finish it today's evening, and at least try it on my...

that would be nice

#

Event stuff was much easier to make thread safe and ended up cleaner and safer than ACPICA

#

ACPICA has a complex mutex + spinlock design

#

And still subject to race windows

#

Whereas i just use a mutex for public API + mask the corresponding event when modifying its handlers so it can't even occur and thus race against handler modification

median crest Dec 17, 2024, 1:00 PM

#

fiery turtle okay astral revealed a slight skill issue but seems to be fixed now

What was it

fiery turtle Dec 17, 2024, 1:01 PM

#

median crest What was it

I moved some logic into a separate helper and it was page faulting

static uacpi_status for_all_gpes_locked(struct do_for_all_gpes_ctx *ctx)
{
    uacpi_status ret;

    UACPI_ENSURE_INIT_LEVEL_AT_LEAST(UACPI_INIT_LEVEL_SUBSYSTEM_INITIALIZED);

    ret = uacpi_recursive_lock_acquire(&g_event_lock);
    if (uacpi_unlikely_error(ret))
        return ret;

    for_each_gpe_block(do_for_all_gpes, &ctx);

    uacpi_recursive_lock_release(&g_event_lock);
    return ctx->ret;
}

#

can u spot the bug?

calm latch Dec 17, 2024, 1:02 PM

#

ctx?

median crest Dec 17, 2024, 1:02 PM

#

&ctx

fiery turtle Dec 17, 2024, 1:02 PM

#

yup

#

also I noticed astral doesnt even initialize events lol

#

i forgor to add it 💀

#

--- a/kernel-src/io/acpi/acpi.c
+++ b/kernel-src/io/acpi/acpi.c
@@ -35,6 +35,8 @@ void acpi_init(void) {
        __assert(ret == UACPI_STATUS_OK);

        uacpi_install_fixed_event_handler(UACPI_FIXED_EVENT_POWER_BUTTON, handle_pwrbtn, UACPI_NULL);
+
+       uacpi_finalize_gpe_initialization();
 }

#

I decided to use recursive locks but it's mostly to prevent the user from deadlocking themselves if they want to do something about the event from an event handler

#

which technically u shouldn't be doing but whatever

calm latch Dec 17, 2024, 1:13 PM

#

Does it use recursive mutex?

#

From kernel api

fiery turtle Dec 17, 2024, 1:13 PM

#

no such kernel api

calm latch Dec 17, 2024, 1:13 PM

#

So it's spinlocks

fiery turtle Dec 17, 2024, 1:13 PM

#

its normal mutexes

#

just with depth count

calm latch Dec 17, 2024, 1:14 PM

#

thonk

fiery turtle Dec 17, 2024, 1:14 PM

#

uacpi_status uacpi_recursive_lock_acquire(struct uacpi_recursive_lock *lock)
{
    uacpi_thread_id this_id;
    uacpi_status ret = UACPI_STATUS_OK;

    this_id = uacpi_kernel_get_thread_id();
    if (UACPI_ATOMIC_LOAD_THREAD_ID(&lock->owner) == this_id) {
        lock->depth++;
        return ret;
    }

    ret = uacpi_acquire_native_mutex(lock->mutex);
    if (uacpi_unlikely_error(ret))
        return ret;

    UACPI_ATOMIC_STORE_THREAD_ID(&lock->owner, this_id);
    lock->depth = 1;
    return ret;
}

calm latch Dec 17, 2024, 1:14 PM

#

Cool

#

My mutex implements recursive locks, but I don't know if it can be trusted...

fiery turtle Dec 17, 2024, 1:15 PM

#

lol

#

yeah less work for kernel api is better

#

less space for user bugs

north holly Dec 17, 2024, 1:16 PM

#

Does recursive lock include recursive unlock?

fiery turtle Dec 17, 2024, 1:16 PM

#

yes

north holly Dec 17, 2024, 1:17 PM

#

Like if I lock twice, the first time I unlock doesn't release it, but the 2nd time it does

fiery turtle Dec 17, 2024, 1:17 PM

#

north holly Like if I lock twice, the first time I unlock doesn't release it, but the 2nd ti...

uacpi_status uacpi_recursive_lock_release(struct uacpi_recursive_lock *lock)
{
    if (lock->depth-- > 1)
        return UACPI_STATUS_OK;

    UACPI_ATOMIC_STORE_THREAD_ID(&lock->owner, UACPI_THREAD_ID_NONE);
    return uacpi_release_native_mutex(lock->mutex);
}

hollow elm Dec 17, 2024, 1:22 PM

#

fiery turtle ```cpp uacpi_status uacpi_recursive_lock_release(struct uacpi_recursive_lock *lo...

maybe check that the owner is the current thread before decrementing the recursion depth?

fiery turtle Dec 17, 2024, 1:23 PM

#

hollow elm maybe check that the owner is the current thread before decrementing the recursi...

maybe, but then again its kinda internal api so something has gone horribly wrong if that check fails

#

i could add that to just log BUG: ...

hollow elm Dec 17, 2024, 1:25 PM

#

i was thinking just assert that thats the case

north holly Dec 17, 2024, 1:25 PM

#

hollow elm i was thinking just assert that thats the case

Uacpi has no assert

fiery turtle Dec 17, 2024, 1:26 PM

#

^

#

uACPI's ideology is that it has no right to decide what is worthy of panic

#

so all errros are propagated back to the host

#

it does reset its own state if some critical init fails tho

median crest Dec 17, 2024, 1:30 PM

#

uAssert when

calm latch Dec 17, 2024, 1:31 PM

#

uKernel

rustic compass Dec 17, 2024, 1:31 PM

#

wouldnt it be better if all manageable error are handled and the rest is an Abort?

fiery turtle Dec 17, 2024, 1:32 PM

#

whats an example of an unmanagable error?

rustic compass Dec 17, 2024, 1:33 PM

#

fiery turtle maybe, but then again its kinda internal api so something has gone horribly wron...

.

fiery turtle Dec 17, 2024, 1:34 PM

#

thats just an internal error IMO

#

like it could signify a memory corruption for example ofc

#

but idk if its worth to abort just because it might be one

loud ice Dec 17, 2024, 1:36 PM

#

debug assertions are good

#

add them

dusky glade Dec 17, 2024, 1:36 PM

#

guys just do it like managarm

loud ice Dec 17, 2024, 1:36 PM

#

lol

dusky glade Dec 17, 2024, 1:36 PM

#

add asserts everywhere

fiery turtle Dec 17, 2024, 1:36 PM

#

yeah as an optional "fortified" uacpi build i could definitely add that tbh

loud ice Dec 17, 2024, 1:37 PM

#

DEBUG build lol

dusky glade Dec 17, 2024, 1:37 PM

#

assert(if cpu exists)

fiery turtle Dec 17, 2024, 1:37 PM

#

for debugging shitty kernel api

loud ice Dec 17, 2024, 1:37 PM

#

yeah

slim panther Dec 17, 2024, 1:37 PM

#

fiery turtle uACPI's ideology is that it has no right to decide what is worthy of panic

which is based

dusky glade Dec 17, 2024, 1:37 PM

#

have u asserted if the cpu actually exists

loud ice Dec 17, 2024, 1:37 PM

#

and possibly internal errors

fiery turtle Dec 17, 2024, 1:38 PM

#

#ifdef UACPI_MAYBE_BROKEN_KERNEL_API
void uacpi_kernel_abort(const uacpi_char *why);
#endif

calm latch Dec 17, 2024, 1:38 PM

#

(100% of operating systems that use uACPI)

loud ice Dec 17, 2024, 1:38 PM

#

fiery turtle ```cpp #ifdef UACPI_MAYBE_BROKEN_KERNEL_API void uacpi_kernel_abort(const uacpi_...

nah i think ```c
#ifdef UACPI_DEBUG_ASSERTIONS
#define UACPI_CHECK(cond) do { if (cond) { while (1) {} } } while (0)
#else
#define UACPI_CHECK(cond) do { } while(0)
#endif

fiery turtle Dec 17, 2024, 1:39 PM

#

I was memeing but yeah

#

thats what I would do

dusky glade Dec 17, 2024, 1:39 PM

#

fiery turtle ```cpp #ifdef UACPI_MAYBE_BROKEN_KERNEL_API void uacpi_kernel_abort(const uacpi_...

// please add this to uacpi
void uacpi_init() {
  uacpi_allocator_init();
  uacpi_arch_init();
  uacpi_scheduler_init();
  uacpi_vfs_init();
  uacpi_own_entry_init();
}

loud ice Dec 17, 2024, 1:39 PM

#

loud ice nah i think ```c #ifdef UACPI_DEBUG_ASSERTIONS #define UACPI_CHECK(cond) do { if...

and ud2 in there or something maybe

fiery turtle Dec 17, 2024, 1:39 PM

#

but i would still add a kernel api hook

#

just hanging is not that great

loud ice Dec 17, 2024, 1:39 PM

#

true

#

but if internal consistency checks are failing, the kernel hooks are more than a bit suspect

dusky glade Dec 17, 2024, 1:39 PM

#

dusky glade ```C // please add this to uacpi void uacpi_init() { uacpi_allocator_init(); ...

infy add this galaxybrain

calm latch Dec 17, 2024, 1:40 PM

#

Have you considered adding heap to uACPI?

gentle peak Dec 17, 2024, 1:40 PM

#

loud ice nah i think ```c #ifdef UACPI_DEBUG_ASSERTIONS #define UACPI_CHECK(cond) do { if...

something like this except it's wrapped in #ifndef UACPI_CHECK so the kernel can provide an assert that's a bit more visible

fiery turtle Dec 17, 2024, 1:40 PM

#

loud ice but if internal consistency checks are failing, the kernel hooks are more than a...

or it would be a UACPI_ARCH_UD2 but at that point might as well have a hook instead

dusky glade Dec 17, 2024, 1:40 PM

#

calm latch Have you considered adding heap to uACPI?

yes

calm latch Dec 17, 2024, 1:40 PM

#

calm latch Have you considered adding heap to uACPI?

So it just calls mmap

loud ice Dec 17, 2024, 1:40 PM

#

fiery turtle or it would be a `UACPI_ARCH_UD2` but at that point might as well have a hook in...

if its gcc or clang you use __builtin_trap()

#

else just hang

fiery turtle Dec 17, 2024, 1:40 PM

#

msvc also has this

#

i think

dusky glade Dec 17, 2024, 1:40 PM

#

who compiles a kernel with msvc

fiery turtle Dec 17, 2024, 1:40 PM

#

__debugbreak or something

calm latch Dec 17, 2024, 1:40 PM

#

fiery turtle or it would be a `UACPI_ARCH_UD2` but at that point might as well have a hook in...

I think llvm already does that too well

fiery turtle Dec 17, 2024, 1:41 PM

#

dusky glade who compiles a kernel with msvc

I do for testing

rustic compass Dec 17, 2024, 1:41 PM

#

nullptr and its value is an index into a error table

dusky glade Dec 17, 2024, 1:41 PM

#

fiery turtle I do for testing

insane

loud ice Dec 17, 2024, 1:41 PM

#

loud ice if its gcc or clang you use `__builtin_trap()`

and in optimized builds you can __builtin_assume that it doesnt happen too

fiery turtle Dec 17, 2024, 1:41 PM

#

calm latch Have you considered adding heap to uACPI?

would be nice to have such option

calm latch Dec 17, 2024, 1:41 PM

#

Does msvc output to elf?

loud ice Dec 17, 2024, 1:41 PM

#

no

loud ice Dec 17, 2024, 1:41 PM

#

fiery turtle would be nice to have such option

uHeap meme

fiery turtle Dec 17, 2024, 1:41 PM

#

lol

calm latch Dec 17, 2024, 1:42 PM

#

calm latch Does msvc output to elf?

I could add PE linker Thonk

fiery turtle Dec 17, 2024, 1:42 PM

#

I already had to sacrafice a lot of make uacpi compilable under msvc so at this point its sunk cost fallacy

loud ice Dec 17, 2024, 1:43 PM

#

what about uACPITableBuilder

fiery turtle Dec 17, 2024, 1:43 PM

#

like an asl compiler?

dusky glade Dec 17, 2024, 1:43 PM

#

fiery turtle I already had to sacrafice a lot of make uacpi compilable under msvc so at this ...

why

#

why support a compiler

#

no one is gonna use

fiery turtle Dec 17, 2024, 1:44 PM

#

i work on windows

dusky glade Dec 17, 2024, 1:44 PM

#

i havent seen one osdev compile their kernel with msvc

fiery turtle Dec 17, 2024, 1:44 PM

#

plus its an extra platform for testing

#

plus kernels that go for nt compat

#

since uacpi is nt compatible

loud ice Dec 17, 2024, 1:44 PM

#

fiery turtle plus kernels that go for nt compat

NT compat doesnt mean you have to use the cringe compiler

#

uacpi.sys when

dusky glade Dec 17, 2024, 1:44 PM

#

^

fiery turtle Dec 17, 2024, 1:44 PM

#

yeah still tho

fiery turtle Dec 17, 2024, 1:44 PM

#

loud ice uacpi.sys when

already exists

loud ice Dec 17, 2024, 1:44 PM

#

oh

fiery turtle Dec 17, 2024, 1:44 PM

#

wip in reactos

loud ice Dec 17, 2024, 1:44 PM

#

damn

#

what about uacpi on NT?

calm latch Dec 17, 2024, 1:45 PM

#

fiery turtle i work on windows

there's wsl

fiery turtle Dec 17, 2024, 1:45 PM

#

this

dusky glade Dec 17, 2024, 1:45 PM

#

loud ice what about uacpi on NT?

could u somehow replace the aml interrepter with uacpi

#

good question

fiery turtle Dec 17, 2024, 1:46 PM

#

loud ice what about uacpi on NT?

would be based for testing

dusky glade Dec 17, 2024, 1:46 PM

#

insane

loud ice Dec 17, 2024, 1:46 PM

#

fiery turtle would be based for testing

yeah

fiery turtle Dec 17, 2024, 1:47 PM

#

apparently uacpi fixed touchscreen detection in reactos on that tablet

#

idk how exactly but thats cool

#

they were previously using acpica

calm latch Dec 17, 2024, 1:47 PM

#

Did reactos switch to uacpi?

fiery turtle Dec 17, 2024, 1:48 PM

#

darkfire is working hard on it but not yet

#

a lot of tech debt to solve before that

calm latch Dec 17, 2024, 1:48 PM

#

That's cool

fiery turtle Dec 17, 2024, 1:48 PM

#

yeah that would be a huge milestone for uacpi as well

calm latch Dec 17, 2024, 1:49 PM

#

Imagine Linux switching one day

fiery turtle Dec 17, 2024, 1:49 PM

#

yeah its doable, just a lot of files to go 1 by 1

#

but uacpi is very close to feature parity with acpica

#

there's like a few thousand acpica api calls

#

each one has to be carefully changed

#

although a lot of them are almost identical

#

also uACPI's repo is a bit over 2 years old now

#

which is crazy

#

i finished the interpreter last december

#

and its been all the rest of the features since then

#

and fixes etc

north holly Dec 17, 2024, 1:53 PM

#

First uacpi port in Feburary

#

I think

fiery turtle Dec 17, 2024, 1:53 PM

#

yeah it couldnt do shit back then

north holly Dec 17, 2024, 1:53 PM

#

Or earlier

#

Made by some guy named oberrow

#

On the cherno server

fiery turtle Dec 17, 2024, 1:54 PM

#

yeah lol

north holly Dec 17, 2024, 1:54 PM

#

I can link his kernel wait

#

https://github.com/OBOS-dev/obos

GitHub

GitHub - OBOS-dev/obos: I was bored...

I was bored... Contribute to OBOS-dev/obos development by creating an account on GitHub.

#

Daily obos shilling complete

fiery turtle Dec 17, 2024, 1:54 PM

#

never seen it before

north holly Dec 17, 2024, 1:54 PM

#

Yeah it is a pretty cool kernel

#

Although since it first ported uacpi, it has had some rewrites

fiery turtle Dec 17, 2024, 1:56 PM

#

lol

dusky glade Dec 17, 2024, 2:04 PM

#

i think we can talk about the second kernel that ported uacpi

#

its pretty cool

#

it was made by a guy i can link his kernel wait

#

https://github.com/rayanmargham/NyauxKC

GitHub

GitHub - rayanmargham/NyauxKC

Contribute to rayanmargham/NyauxKC development by creating an account on GitHub.

#

its had some rewrites tho

#

idk

#

trl

#

Daily nyaux shilling complete

torpid root Dec 17, 2024, 2:07 PM

#

guys what about kernel rewrite by some guy named ilobilo
I hear it throws a coprocessor segment overrun exception if you press a key in vmware

#

that dumass probably messed up the pic again

fiery turtle Dec 17, 2024, 2:08 PM

#

torpid root guys what about kernel rewrite by some guy named ilobilo I hear it throws a copr...

Lmao

#

Actually I dont know which kernel was second

#

I wanna say managarm

torpid root Dec 17, 2024, 2:08 PM

#

I guess I should only enable interrupts when the pic is remapped lol

torpid root Dec 17, 2024, 2:08 PM

#

fiery turtle Actually I dont know which kernel was second

mine was definitely after managarm

calm latch Dec 17, 2024, 2:11 PM

#

https://gitlab.com/mishakov/pmos/-/commit/7fcb3e82a509838be194ddaead94288f01054779

north holly Dec 17, 2024, 2:11 PM

#

dusky glade i think we can talk about the second kernel that ported uacpi

But it kinda wasn't

#

Managarm was 2nd

#

I believe

dusky glade Dec 17, 2024, 2:12 PM

#

damn

#

nooo

calm latch Dec 17, 2024, 2:12 PM

#

Was I third?

north holly Dec 17, 2024, 2:13 PM

#

On the 28th of January 2024, I ported uacpi to obos

#

#560042023638269955 message

fiery turtle Dec 17, 2024, 2:13 PM

#

calm latch Was I third?

Perhaps

median crest Dec 17, 2024, 2:14 PM

#

Astral was first because all others arent real operating systems

#

Astral <------- real operating system
Linux windows etc etc <------- delusional cosplayers

calm latch Dec 17, 2024, 2:15 PM

#

Although I did that on May 13

#

It's kinda late

#

Maybe not

north holly Dec 17, 2024, 2:16 PM

#

I am also the first one using uacpi to implement suspend functionality

#

And the first one to implement device detection via uacpi (sometime in February)

calm latch Dec 17, 2024, 2:17 PM

#

I'm racing you to be the second OS to have EC

north holly Dec 17, 2024, 2:17 PM

#

Well I can't work on obos for another couple hours

fiery turtle Dec 17, 2024, 2:17 PM

#

median crest Astral was first because all others arent real operating systems

Lol

#

True true

calm latch Dec 17, 2024, 2:17 PM

#

I can't either

loud ice Dec 17, 2024, 2:17 PM

#

calm latch I'm racing you to be the second OS to have EC

time to speedrun a cros_ec driver

fiery turtle Dec 17, 2024, 2:18 PM

#

north holly I am also the first one using uacpi to implement suspend functionality

Also true

north holly Dec 17, 2024, 2:18 PM

#

Time to copy managarm code

#

I meant read the spec

fiery turtle Dec 17, 2024, 2:18 PM

#

loud ice time to speedrun a cros_ec driver

Managarm ec driver port speedrun

calm latch Dec 17, 2024, 2:18 PM

#

ctrl+c ctrl+v speed contest

north holly Dec 17, 2024, 2:19 PM

#

found ec spec

loud ice Dec 17, 2024, 2:19 PM

#

what kind of stuff does an EC do anyway?

north holly Dec 17, 2024, 2:19 PM

#

magical stuff

fiery turtle Dec 17, 2024, 2:20 PM

#

loud ice what kind of stuff does an EC do anyway?

Its two registers

calm latch Dec 17, 2024, 2:20 PM

#

Was I the second to do PCI interrupts routing with uACPI?

loud ice Dec 17, 2024, 2:20 PM

#

oh lol

#

what does it do tho

fiery turtle Dec 17, 2024, 2:20 PM

#

Abstraction over proprietary firmware state basically

north holly Dec 17, 2024, 2:20 PM

#

loud ice what does it do tho

what aml wants it to do

loud ice Dec 17, 2024, 2:20 PM

#

ah so nothing useful

#

got it

north holly Dec 17, 2024, 2:20 PM

#

well no, you need it on modern devices

#

for power button

#

n'shit

loud ice Dec 17, 2024, 2:21 PM

#

"need"

fiery turtle Dec 17, 2024, 2:21 PM

#

It can toggle fans, features, thermals, runtime options etc

#

Read temperatures

loud ice Dec 17, 2024, 2:21 PM

#

north holly well no, you need it on modern devices

i can get an m1

calm latch Dec 17, 2024, 2:21 PM

#

north holly well no, you need it on modern devices

On broken modern devices

fiery turtle Dec 17, 2024, 2:21 PM

#

Basically all firmware stuff

loud ice Dec 17, 2024, 2:21 PM

#

and then i get the power button from something else

north holly Dec 17, 2024, 2:21 PM

#

calm latch On broken modern devices

yeah aka laptops

calm latch Dec 17, 2024, 2:21 PM

#

No

#

MacBooks

#

And my 2023 lenovo laptop for no reason

north holly Dec 17, 2024, 2:21 PM

#

$120 school laptops

#

or sorry

#

$60

#

/j

loud ice Dec 17, 2024, 2:23 PM

#

i think chromebooks have their own linux driver

fiery turtle Dec 17, 2024, 2:23 PM

#

Basically all x86 hw after circa 2010 has one

fiery turtle Dec 17, 2024, 2:23 PM

#

loud ice i think chromebooks have their own linux driver

Yeah some have a fucked up quirky ec

#

So they get a special driver or quirk

loud ice Dec 17, 2024, 2:23 PM

#

fiery turtle Yeah some have a fucked up quirky ec

i wouldnt call it fucked up and quirky

#

they have their own entire driver lol

#

called cros_ec

#

its also in frameworks btw

fiery turtle Dec 17, 2024, 2:24 PM

#

The thing is ec is abstract away it shouldn't need special drivers lol

calm latch Dec 17, 2024, 2:24 PM

#

But how does Windows deal with it

loud ice Dec 17, 2024, 2:24 PM

#

fiery turtle The thing is ec is abstract away it shouldn't need special drivers lol

why would they support this?

#

it would imply writing aml

#

and they are vendoring their own kernels anyway so might as well

fiery turtle Dec 17, 2024, 2:25 PM

#

Dont they expose it via aml as well

#

Chromebooks have acpi

loud ice Dec 17, 2024, 2:25 PM

#

no fucking idea

#

also its open source so you'll see it more and more i guess

fiery turtle Dec 17, 2024, 2:25 PM

#

calm latch But how does Windows deal with it

Special chipset driver

loud ice Dec 17, 2024, 2:25 PM

#

frameworks also have a cros_ec

#

so

fiery turtle Dec 17, 2024, 2:25 PM

#

I've had someone boot uacpi on a Chromebook so yeah

loud ice Dec 17, 2024, 2:26 PM

#

oh cool

fiery turtle Dec 17, 2024, 2:26 PM

#

It does have aml albeit tiny

loud ice Dec 17, 2024, 2:26 PM

#

i think i sent you my aml dump

fiery turtle Dec 17, 2024, 2:26 PM

#

Compared to other hw

loud ice Dec 17, 2024, 2:26 PM

#

from my chromebook?

fiery turtle Dec 17, 2024, 2:26 PM

#

Think so yeah

#

A while ago

loud ice Dec 17, 2024, 2:26 PM

#

which is dead now, rip charging controller

fiery turtle Dec 17, 2024, 2:27 PM

#

Oof

torpid root Dec 17, 2024, 2:34 PM

#

my uacpi commit was on June 1

#

I'll take the fourth place

loud ice Dec 17, 2024, 2:36 PM

#

maybe i should add uacpi support to my bootloader/firmware thing

calm latch Dec 17, 2024, 2:36 PM

#

you def need userspace /s

loud ice Dec 17, 2024, 2:36 PM

#

userspace?

calm latch Dec 17, 2024, 2:36 PM

#

to run stuff in it

#

in bootloader

#

(unrelated to uACPI)

loud ice Dec 17, 2024, 2:37 PM

#

why would i do userspace

#

much better to write my own scripting lang

calm latch Dec 17, 2024, 2:37 PM

#

because it's cursed

loud ice Dec 17, 2024, 2:37 PM

#

an AOT bytecoded one so i dont ship the compiler

#

my binaries are already way too big

fiery turtle Dec 17, 2024, 2:40 PM

#

loud ice maybe i should add uacpi support to my bootloader/firmware thing

Absolutely (not biased opinion in any way KEKW )

loud ice Dec 17, 2024, 2:41 PM

#

fiery turtle Absolutely (not biased opinion in any way <:KEKW:913708372329644042> )

uacpi cant make new acpi tables though

#

so idk maybe not

fiery turtle Dec 17, 2024, 2:41 PM

#

Aml ones u just propagate as is, the rest is just hardcoded struct + data so

loud ice Dec 17, 2024, 2:42 PM

#

fiery turtle Aml ones u just propagate as is, the rest is just hardcoded struct + data so

well no i cant

#

i need to build them dynamically

#

thats like the point

fiery turtle Dec 17, 2024, 2:45 PM

#

loud ice i need to build them dynamically

Like from fdt?

#

I thought u dropped that idea

#

But yeah if thats your only goal for acpi then yes, unless u want to use aml to discover available non pci devices in the bootloader directly

loud ice Dec 17, 2024, 2:46 PM

#

fiery turtle I thought u dropped that idea

i still kinda want to

#

its on the list of things i may want to do at some point

fiery turtle Dec 17, 2024, 2:46 PM

#

That would be a truly unique thing

loud ice Dec 17, 2024, 2:47 PM

#

yes but it would involve dealing with aml shit

#

id probably have to write an aml compiler

fiery turtle Dec 17, 2024, 2:47 PM

#

I would just use a DSL like qemu

#

Aml is very good for that

loud ice Dec 17, 2024, 2:47 PM

#

fiery turtle I would just use a DSL like qemu

DSL?

fiery turtle Dec 17, 2024, 2:48 PM

#

I can show u an example of how qemu does it if u want

fiery turtle Dec 17, 2024, 2:48 PM

#

loud ice DSL?

Yeah

loud ice Dec 17, 2024, 2:48 PM

#

fiery turtle I can show u an example of how qemu does it if u want

sure

fiery turtle Dec 17, 2024, 2:50 PM

#

loud ice sure

https://github.com/qemu/qemu/blob/8032c78e556cd0baec111740a6c636863f9bd7c8/hw/acpi/tpm.c#L23

GitHub

qemu/hw/acpi/tpm.c at 8032c78e556cd0baec111740a6c636863f9bd7c8 · qe...

Official QEMU mirror. Please see https://www.qemu.org/contribute/ for how to submit changes to QEMU. Pull Requests are ignored. Please only use release tarballs from the QEMU website. - qemu/qemu

#

example for a TPM device

loud ice Dec 17, 2024, 2:51 PM

#

ah

fiery turtle Dec 17, 2024, 2:51 PM

#

ifctx = aml_if(aml_lgreater_equal(op, aml_int(0x100)));

loud ice Dec 17, 2024, 2:51 PM

#

ugly

fiery turtle Dec 17, 2024, 2:51 PM

#

literally trivial

#

no compiler needed

loud ice Dec 17, 2024, 2:51 PM

#

id much rather make a good compiler

#

or a good dsl at lesat

fiery turtle Dec 17, 2024, 2:51 PM

#

you would produce ascii text then compile it?

loud ice Dec 17, 2024, 2:51 PM

#

actually ig i could abuse rust

loud ice Dec 17, 2024, 2:51 PM

#

fiery turtle you would produce ascii text then compile it?

of course not lol

fiery turtle Dec 17, 2024, 2:52 PM

#

rust macros could be abused to make a very powerful dsl here

loud ice Dec 17, 2024, 2:52 PM

#

true

dusky glade Dec 17, 2024, 4:11 PM

#

@fiery turtle this u?

fiery turtle Dec 17, 2024, 4:52 PM

#

Wut

dusky glade Dec 17, 2024, 5:43 PM

#

fiery turtle Wut

the osdev irc channel thinks this is your username

#

on the irc channel

fiery turtle Dec 17, 2024, 5:46 PM

#

Never used irc before

dusky glade Dec 17, 2024, 5:46 PM

#

told them that now

#

cant believe im using irc in 2024

#

smh

fiery turtle Dec 17, 2024, 5:47 PM

#

Is it active?

dusky glade Dec 17, 2024, 5:47 PM

#

yea

fiery turtle Dec 17, 2024, 5:47 PM

#

dusky glade the osdev irc channel thinks this is your username

How would it even know

dusky glade Dec 17, 2024, 5:48 PM

#

fiery turtle How would it even know

i dont even know they kept hinting at that 😭

#

they are weird

#

but its an active irc channel so

#

🤷‍♂️

fiery turtle Dec 17, 2024, 5:56 PM

#

Didnt understand any of that but w/e

dusky glade Dec 17, 2024, 6:02 PM

#

fiery turtle Didnt understand any of that but w/e

w/e?

#

wwe?

#

john cena? did u say john cina infy

deft canopy Dec 17, 2024, 6:19 PM

#

dusky glade w/e?

"whatever"

calm latch Dec 17, 2024, 6:20 PM

#

Ok, I'm back

#

Time to finish EC

winter orbit Dec 17, 2024, 6:20 PM

#

fiery turtle yeah that would be a huge milestone for uacpi as well

Well hopefully this will allow some pretty useful stuff to be sent back to uACPI

#

That’s the goal

winter orbit Dec 17, 2024, 6:21 PM

#

loud ice what about uacpi on NT?

This is the plan

#

Im not sure if Infy wants a vs solution sent back into a PR or something that remains to be figured out

calm latch Dec 17, 2024, 6:22 PM

#

Btw, what should interrupt handlers return on failure?

winter orbit Dec 17, 2024, 6:22 PM

#

But uACPI will be able to be plugged into vista from reactos

dusky glade Dec 17, 2024, 6:22 PM

#

deft canopy "whatever"

https://tenor.com/view/john-cena-wwe-urstrulyshailesh-gif-20805283

Tenor

#

wwe

#

real

#

infy is john cena

#

real

north holly Dec 17, 2024, 6:53 PM

#

calm latch Time to finish EC

Noooo

#

I'm still at school

#

rn_image_picker_lib_temp_b99177cb-3167-415c-afbf-372da5994669.jpg

fiery turtle Dec 17, 2024, 6:54 PM

#

winter orbit But uACPI will be able to be plugged into vista from reactos

Wait really? You can just port the reactos driver and it will work?

calm latch Dec 17, 2024, 6:55 PM

#

north holly I'm still at school

How do you have the school that late?

north holly Dec 17, 2024, 6:55 PM

#

It's 2 pm

calm latch Dec 17, 2024, 6:55 PM

#

Thonk

#

I confused you and linuxmaster2.0

fiery turtle Dec 17, 2024, 6:56 PM

#

winter orbit Im not sure if Infy wants a vs solution sent back into a PR or something that re...

Yeah I mean if there's a way to build a driver somehow for regression testing id be interested to hear

north holly Dec 17, 2024, 6:58 PM

#

I only have a small base for an EC driver

#

Basically nothing

#

I could ssh into my 0x

#

Pc

#

From school

calm latch Dec 17, 2024, 6:59 PM

#

Thonk

north holly Dec 17, 2024, 7:00 PM

#

If only I had that port forwarded

loud ice Dec 17, 2024, 7:00 PM

#

north holly If only I had that port forwarded

this is why you should use tailscale

north holly Dec 17, 2024, 7:01 PM

#

Wait

#

I will compile a cross compiler

#

For x86 64 elf

#

Then compile obos with it

#

Termux can't run static executables, unfortunately

#

So that won't work

#

calm latch Dec 17, 2024, 7:11 PM

#

I'm eating, so you have time bruh

calm latch Dec 17, 2024, 7:32 PM

#

Do I have to use burst?

#

Managram does it, but since I'm running in userspace, I'm afraid this stuff could be preempted

#

(though I have priorities in scheduler, I don't want to use that)

winter orbit Dec 17, 2024, 7:33 PM

#

fiery turtle Wait really? You can just port the reactos driver and it will work?

hmm? sorry i didn't see this till now.

yes that's what im working on

calm latch Dec 17, 2024, 7:33 PM

#

And it would disengage

north holly Dec 17, 2024, 7:34 PM

#

Infy don't answer his questions

#

Until I get back home

calm latch Dec 17, 2024, 7:36 PM

#

bruh

#

Does QEMU support EC?

#

I'm gonna try it on my laptop...

#

Looks like QEMU doesn't have it..?

#

@fiery turtle it's failing what am I doing wrong

#

But I did get an even pressing f keys...

fiery turtle Dec 17, 2024, 7:52 PM

#

winter orbit hmm? sorry i didn't see this till now. yes that's what im working on

Nice

fiery turtle Dec 17, 2024, 7:53 PM

#

calm latch <@353593948523986945> it's failing what am I doing wrong

Do u initialize EC before namespace_initialize?

calm latch Dec 17, 2024, 7:53 PM

#

omega bruh

fiery turtle Dec 17, 2024, 7:54 PM

#

calm latch omega bruh

What did you do 💀

calm latch Dec 17, 2024, 7:54 PM

#

Idk it doesn't boot

fiery turtle Dec 17, 2024, 7:54 PM

#

Hard power off

calm latch Dec 17, 2024, 7:54 PM

#

Ok it booted

fiery turtle Dec 17, 2024, 7:54 PM

#

Least broken ec driver

calm latch Dec 17, 2024, 7:54 PM

#

But it booted to Windows, turned off and then I had this

fiery turtle Dec 17, 2024, 7:56 PM

#

calm latch <@353593948523986945> it's failing what am I doing wrong

Basically installing ec handler must happen before namespace initialize

calm latch Dec 17, 2024, 7:56 PM

#

I'm gonna push my code to gitlab

#

I'm doing this https://gitlab.com/mishakov/pmos/-/blob/dev/devicesd/generic/acpi/acpi.c?ref_type=heads#L212

fiery turtle Dec 17, 2024, 8:01 PM

#

Basically you should early reg for your laptop

#

This error doesnt happen on Linux right? Idk what logic it uses to decide if you should or not

calm latch Dec 17, 2024, 8:03 PM

#

Idk

#

I think it's secure boot

#

Which I have to turn on because of bitlocker

fiery turtle Dec 17, 2024, 8:03 PM

#

Your laptop is strange in that it doesnt give an ECDT but expects _REG to be called before _STA

#

I have a laptop which wants the opposite

calm latch Dec 17, 2024, 8:04 PM

#

What am I supposed to do?

#

Managram broken?

calm latch Dec 17, 2024, 8:05 PM

#

fiery turtle Your laptop is strange in that it doesnt give an ECDT but expects _REG to be cal...

I've never tried Linux on it

#

Which part of spec should I look at?

fiery turtle Dec 17, 2024, 8:05 PM

#

Look at how Linux decides when to install the handler I guess, but for your hw the managarm logic is wrong

fiery turtle Dec 17, 2024, 8:06 PM

#

calm latch Which part of spec should I look at?

None, its not specified there

calm latch Dec 17, 2024, 8:06 PM

#

I find linux source to be intimidating

fiery turtle Dec 17, 2024, 8:06 PM

#

BSDs

#

I can take a look later

#

Btw try to boot live Linux on it

calm latch Dec 17, 2024, 8:07 PM

#

#ifdef CONFIG_ACPI_EC
    /*
     * ACPI 2.0 requires the EC driver to be loaded and work before
     * the EC device is found in the namespace (i.e. before acpi_initialize_objects()
     * is called).
     *
     * This is accomplished by looking for the ECDT table, and getting 
     * the EC parameters out of that.
     */
    status = acpi_ec_ecdt_probe();
    /* Ignore result. Not having an ECDT is not fatal. */
#endif
```?

#

What's the website with linux source

#

drivers/acpi/bus.c:689

fiery turtle Dec 17, 2024, 8:07 PM

#

Bootlin

calm latch Dec 17, 2024, 8:08 PM

#

ret = acpi_ec_get_real_ecdt();
/* Try to make a fake ECDT */
if (ret && acpi_fake_ecdt_enabled) {
  ret = acpi_ec_fake_ecdt();
}
```![KEKW](https://cdn.discordapp.com/emojis/913708372329644042.webp?size=128 "KEKW")

fiery turtle Dec 17, 2024, 8:09 PM

#

Lol

#

What decides if there should be a fake ECDT

calm latch Dec 17, 2024, 8:09 PM

#

I'm gonna try Linux

fiery turtle Dec 17, 2024, 8:09 PM

#

Yeah try it

calm latch Dec 17, 2024, 8:09 PM

#

I think it's a kernel option

#

But I would imagine someone using Linux on it

#

It's 1 year old laptop

fiery turtle Dec 17, 2024, 8:10 PM

#

Yeah

calm latch Dec 17, 2024, 8:10 PM

#

So it must work.......

#

Ok, it's 5 minutes to try

fiery turtle Dec 17, 2024, 8:11 PM

#

Yeah try and we'll try to decide from there

#

Its not a critical error or anything but still

calm latch Dec 17, 2024, 8:11 PM

#

What fedora ships ppc?

#

Bruh and it's 2 gb ISO

#

I'm gonna download ubuntu as well...

calm latch Dec 17, 2024, 8:12 PM

#

calm latch But I did get an even pressing f keys...

But still I've gotten en event here

#

But power button didn't work

fiery turtle Dec 17, 2024, 8:13 PM

#

Dont see u running the query

calm latch Dec 17, 2024, 8:13 PM

#

Which query

fiery turtle Dec 17, 2024, 8:13 PM

#

... In response to the event

#

_QXX

calm latch Dec 17, 2024, 8:14 PM

#

Oof

#

Maybe it locked up

#

I need more printfs

fiery turtle Dec 17, 2024, 8:14 PM

#

Yup

calm latch Dec 17, 2024, 8:15 PM

#

I was kinda hoping it would work first try

fiery turtle Dec 17, 2024, 8:15 PM

#

I had to debug hangs as well, but it was a skill issue

#

When making the managarm ec driver

calm latch Dec 17, 2024, 8:17 PM

#

calm latch But I did get an even pressing f keys...

What's event 1 btw

calm latch Dec 17, 2024, 8:17 PM

#

calm latch But I did get an even pressing f keys...

Wth

#

I've basically copied managram and so it should be firing uacpi_kernel_schedule_work, which should be printing stuff without taking locks

#

Maybe I do have a deadlock in kernel after all..

fiery turtle Dec 17, 2024, 8:19 PM

#

calm latch What's event 1 btw

Its platform specific

#

See your aml _Q01

calm latch Dec 17, 2024, 8:20 PM

#

I'm feeling dumb

#

There's nowhere to lock up between 2 printfs

#

Ok, something fishy is going on

#

(maybe I do need kernel panic to framebuffer after all)

slim panther Dec 17, 2024, 8:22 PM

#

serial meme

calm latch Dec 17, 2024, 8:23 PM

#

No that's the only output I have from kernel

#

The issue is that I only have it in VMs and on couple of my PCs

#

And not on laptops

#

So sometimes I don't know if the kernel has panicked

#

Ok anyways I've been looking at the wrong part of code again galaxybrain

#

Somehow the kernel never crashes

frank canopy Dec 17, 2024, 8:30 PM

#

dang updating to the new recursive-locks version broke my kernel

calm latch Dec 17, 2024, 8:31 PM

#

Btw can EC be tested in VMs?

north holly Dec 17, 2024, 8:31 PM

#

I still have time to make a more stable ec driver than you

calm latch Dec 17, 2024, 8:34 PM

#

EC on PCI??

#

I'm trying Linux

torpid root Dec 17, 2024, 8:34 PM

#

calm latch EC on PCI??

say hello from me too to the pthreads

calm latch Dec 17, 2024, 8:39 PM

#

https://pastebin.com/W9zrQntR

Pastebin

Lenovo dmesg - Pastebin.com

Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.

#

Phoenix bios?????

#

I thought they were out of business

#

What is [ 1.057783] Unstable clock detected, switching default tracing clock to "global"?

#

https://www.linuxquestions.org/questions/slackware-14/tsc-found-unstable-after-boot-4175711883/

TSC found unstable after boot

#

So TSC is broken on this laptop?

#

Lol

#

And it's using PIT for calibration

#

What snd_hda_intel 0000:c3:00.1: enabling device (0000 -> 0002)

#

This is AMD only laptop

#

Will uhda work on it? thinkong

#

[   12.512529] snd_hda_intel 0000:c3:00.1: enabling device (0000 -> 0002)
[   12.512944] snd_hda_intel 0000:c3:00.1: Handle vga_switcheroo audio client
[   12.513058] snd_hda_intel 0000:c3:00.6: enabling device (0000 -> 0002)
[   12.531287] amd_atl: AMD Address Translation Library initialized
[   12.531624] intel_rapl_common: Found RAPL domain package
[   12.531627] intel_rapl_common: Found RAPL domain core
[   12.532028] snd_hda_intel 0000:c3:00.1: bound 0000:c3:00.0 (ops amdgpu_dm_audio_component_bind_ops [amdgpu])
[   12.534327] input: HD-Audio Generic HDMI/DP,pcm=3 as /devices/pci0000:00/0000:00:08.1/0000:c3:00.1/sound/card0/input16
[   12.534388] input: HD-Audio Generic HDMI/DP,pcm=7 as /devices/pci0000:00/0000:00:08.1/0000:c3:00.1/sound/card0/input17
[   12.534454] input: HD-Audio Generic HDMI/DP,pcm=8 as /devices/pci0000:00/0000:00:08.1/0000:c3:00.1/sound/card0/input18

fiery turtle Dec 17, 2024, 8:54 PM

#

calm latch Btw can EC be tested in VMs?

unfortunately no

fiery turtle Dec 17, 2024, 8:55 PM

#

calm latch EC on PCI??

under the LPC0 device

frank canopy Dec 17, 2024, 8:58 PM

#

oh thats why the new recursive mutex thing broke my kernel, my get thread id crashes before my processor-local control blocks get initialized which i dont do until after i init uacpi so i can have the MADT

#

does early table access use mutexes?

#

(by crashes i mean that before processor-local control blocks are initialized the gs-base is 0 and so the gs-relative pointer page faults accessing virt address 0 when i try to check the current thread id)

fiery turtle Dec 17, 2024, 9:00 PM

#

frank canopy does early table access use mutexes?

definitely not the recursive ones

#

or actually no it doesnt use any

#

it doesnt use heap whatsoever

frank canopy Dec 17, 2024, 9:00 PM

#

cool, then i can use early table access to ensure i dont need thread ids until after i get processor local stuff figured out

fiery turtle Dec 17, 2024, 9:00 PM

#

yeah definitely

#

early table access uses logging and map/unmap

frank canopy Dec 17, 2024, 9:01 PM

#

will need to figure out memory management for the early table access buffer (which i assume can be freed after the normal initialize) but ill need memory management anyway by that point for the map/unmap so thats fine

fiery turtle Dec 17, 2024, 9:01 PM

#

yeah

#

linux just puts it in an __init section

#

which is automatically freed after init

frank canopy Dec 17, 2024, 9:02 PM

#

ill have full kernel-mode alloc ready to use well before that point most likely anyway

fiery turtle Dec 17, 2024, 9:08 PM

#

ah ok then u can just kmalloc(4096)

#

then free after uacpi_initialize

#

there are kernel initialization hooks u may reuse for this

frank canopy Dec 17, 2024, 9:09 PM

#

also a question for those who know windows synchronization stuff, is the uacpi_kernel_xxx_event stuff expected to be closer to a KSEMAPHORE or to a SynchronizationEvent?

#

cool

calm latch Dec 17, 2024, 9:49 PM

#

calm latch What is `[ 1.057783] Unstable clock detected, switching default tracing clock...

@fiery turtle could TSC be broken?

mortal yoke Dec 17, 2024, 9:52 PM

#

calm latch Will uhda work on it?<:thinkong:815589724424241152>

only one way to find out :p

fiery turtle Dec 17, 2024, 9:52 PM

#

calm latch <@353593948523986945> could TSC be broken?

maybe

calm latch Dec 17, 2024, 9:52 PM

#

I think Linux has gotten the same frequency

mortal yoke Dec 17, 2024, 9:53 PM

#

mortal yoke only one way to find out :p

I should make a test image but I still haven't figured out the issue with astral and on my own kernel switching the output would be painful because the xhci is broken (though ig I could add a kernel param but that's not ideal)

hollow elm Dec 17, 2024, 9:54 PM

#

calm latch <@353593948523986945> could TSC be broken?

i remember getting the same message on my laptop and it went away after a bios update

calm latch Dec 17, 2024, 9:54 PM

#

I think I have the latest bios though

calm latch Dec 17, 2024, 9:55 PM

#

mortal yoke I should make a test image but I still haven't figured out the issue with astral...

I don't have a shell, but I guess I could try and port it...

mortal yoke Dec 17, 2024, 9:59 PM

#

shouldn't be that hard

calm latch Dec 17, 2024, 9:59 PM

#

But I want shell first...

#

And for that I need AHCI

#

And for that I need working set

north holly Dec 17, 2024, 10:00 PM

#

@fiery turtle does ec init have to happen before namespace load or init?

fiery turtle Dec 17, 2024, 10:01 PM

#

it cant happen before load

#

either before init or after

#

depending on something

#

idk what specifically

north holly Dec 17, 2024, 10:01 PM

#

does the spec know?

slim panther Dec 17, 2024, 10:02 PM

#

https://tenor.com/view/riddler-paul-dano-new-rockstars-thumbnail-does-he-know-he-doesnt-know-gif-27552850

Tenor

fiery turtle Dec 17, 2024, 10:02 PM

#

no

#

only linux knows

#

managarm does if ECDT exists then after load, otherwise after init

torpid root Dec 17, 2024, 10:02 PM

#

linux knows all

median crest Dec 17, 2024, 10:03 PM

#

who is this linux person

slim panther Dec 17, 2024, 10:03 PM

#

linux torvald

calm latch Dec 17, 2024, 10:04 PM

#

Wait can I unconditionally call it after init?

fiery turtle Dec 17, 2024, 10:04 PM

#

that will break some init code like your battery

calm latch Dec 17, 2024, 10:04 PM

#

Like I'm iterating through it, where in it does it request early init?

torpid root Dec 17, 2024, 10:04 PM

#

slim panther linux torvald

linux sebastian

north holly Dec 17, 2024, 10:05 PM

#

where is linux' ec driver

#

found it

#

https://elixir.bootlin.com/linux/v6.12.5/source/drivers/acpi/ec.c

ec.c - drivers/acpi/ec.c - Linux source code v6.12.5...

Elixir Cross Referencer - source code of Linux v6.12.5: drivers/acpi/ec.c

calm latch Dec 17, 2024, 10:07 PM

#

Ok I'm having a skill issue with my pthreads

#

I think I've found it 👍

north holly Dec 17, 2024, 10:08 PM

#

https://elixir.bootlin.com/linux/v6.12.5/source/drivers/acpi/ec.c#L1808

ec.c - drivers/acpi/ec.c - Linux source code v6.12.5...

Elixir Cross Referencer - source code of Linux v6.12.5: drivers/acpi/ec.c

calm latch Dec 17, 2024, 10:09 PM

#

(for some reason, pinning threads to other CPUs seems to be broken...)

#

Wait, no

#

Do GPE handlers want CPU 0?

#

bootstrap CPU that is

north holly Dec 17, 2024, 10:10 PM

#

I believe so

#

what linux does:

before namespace load, it probes the ecdt table, if it exists
if that exists, all is fine
otherwise, sometime after namespace init, it probes the acpi namespace for an EC

calm latch Dec 17, 2024, 10:11 PM

#

north holly what linux does: - before namespace load, it probes the ecdt table, if it exists...

I did that and somehow it doesn't work on my laptop

north holly Dec 17, 2024, 10:11 PM

#

well does linux work?

calm latch Dec 17, 2024, 10:12 PM

#

It looks like it

north holly Dec 17, 2024, 10:12 PM

#

https://elixir.bootlin.com/linux/v6.12.5/source/drivers/acpi/bus.c#L1359

bus.c - drivers/acpi/bus.c - Linux source code v6.12...

Elixir Cross Referencer - source code of Linux v6.12.5: drivers/acpi/bus.c

#

linux init sequence there

north holly Dec 17, 2024, 10:12 PM

#

calm latch It looks like it

check journalctl or smth

#

or dmesh

#

*dmesg

#

to make sure nothing errored out

calm latch Dec 17, 2024, 10:12 PM

#

calm latch https://pastebin.com/W9zrQntR

.

fiery turtle Dec 17, 2024, 10:12 PM

#

north holly https://elixir.bootlin.com/linux/v6.12.5/source/drivers/acpi/bus.c#L1359

yup

#

it installs it to root if there's an ECDT

#

https://elixir.bootlin.com/linux/v6.12.5/source/drivers/acpi/ec.c#L1530

ec.c - drivers/acpi/ec.c - Linux source code v6.12.5...

Elixir Cross Referencer - source code of Linux v6.12.5: drivers/acpi/ec.c

calm latch Dec 17, 2024, 10:14 PM

#

ACPI BIOS Error (bug): Failure creating named object [\_SB.MHSP], AE_ALREADY_EXISTS (20240322/dswload2-326)

fiery turtle Dec 17, 2024, 10:14 PM

#

yeah this one uacpi produces as well

#

its unrelated

calm latch Dec 17, 2024, 10:14 PM

#

What is this ACPI: [Firmware Bug]: BIOS _OSI(Linux) query ignored?

fiery turtle Dec 17, 2024, 10:15 PM

#

also unrelated

#

long story

#

just to make sure, there's definitely no ECDT right?

calm latch Dec 17, 2024, 10:16 PM

#

Could I need to be calling initialization code before finalize_gpe_initialization?

fiery turtle Dec 17, 2024, 10:16 PM

#

nah thats unrelated

#

_REG happens when installing the handler

calm latch Dec 17, 2024, 10:17 PM

#

calm latch https://pastebin.com/W9zrQntR

This also doesn't seem to be mentioning it anywhere

fiery turtle Dec 17, 2024, 10:17 PM

#

so u should be installing it earlier

#

can u paste the full dmesg

calm latch Dec 17, 2024, 10:17 PM

#

calm latch https://pastebin.com/W9zrQntR

This is full dmesg

#

If I'm not missing anything

fiery turtle Dec 17, 2024, 10:18 PM

#

calm latch Phoenix bios?????

the person porting uacpi to reactos literally works there lol

calm latch Dec 17, 2024, 10:18 PM

#

Cool

#

I though everyone was using AMI

#

Anyway

fiery turtle Dec 17, 2024, 10:19 PM

#

[    0.249958] ACPI: EC: EC started
[    0.249959] ACPI: EC: interrupt blocked
[    0.587395] ACPI: EC: EC_CMD/EC_SC=0x66, EC_DATA=0x62
[    0.587399] ACPI: \_SB_.PCI0.LPC0.EC0_: Boot DSDT EC used to handle transactions

calm latch Dec 17, 2024, 10:20 PM

#

DSDT EC?

#

https://wiki.archlinux.org/title/DSDT

DSDT

fiery turtle Dec 17, 2024, 10:21 PM

#

https://elixir.bootlin.com/linux/v6.12.5/source/drivers/acpi/bus.c#L1394

bus.c - drivers/acpi/bus.c - Linux source code v6.12...

Elixir Cross Referencer - source code of Linux v6.12.5: drivers/acpi/bus.c

#

this call here

fiery turtle Dec 17, 2024, 10:21 PM

#

calm latch DSDT EC?

just means it was found in the AML namespace

#

so yeah in your case linux also calls _STA and _INI before

#

but i guess maybe they initialize locals to 0 by default and no Uninitialized

#

anyway that shouldnt matter for now

#

its not that big of an issue

calm latch Dec 17, 2024, 10:22 PM

#

But my power button doesn't work

#

(and my threads are broken)

fiery turtle Dec 17, 2024, 10:22 PM

#

thats your driver's skill issue

calm latch Dec 17, 2024, 10:22 PM

#

(maybe that is the reason)

fiery turtle Dec 17, 2024, 10:22 PM

#

its unrelated to this

calm latch Dec 17, 2024, 10:23 PM

#

That's my kernel's skill issue because why doesn't it work

fiery turtle Dec 17, 2024, 10:23 PM

#

could try managarm to see if its a uacpi bug

#

but do u even run the query?

calm latch Dec 17, 2024, 10:23 PM

#

I run uacpi_kernel_schedule_work

#

And it creates a thread, which pins itself to bootstrap CPU

#

And after that syscall, it disappears

#

Looking at printfs

fiery turtle Dec 17, 2024, 10:24 PM

#

Method (_Q32, 0, NotSerialized)  // _Qxx: EC Query, xx=0x00-0xFF
{
    P80H = 0x32
    Notify (PWRB, 0x80) // Status Change
}

calm latch Dec 17, 2024, 10:24 PM

#

Or the kernel hangs

fiery turtle Dec 17, 2024, 10:24 PM

#

this is the event u should be getting

#

0x32

#

so your reading of 0x01 is bogus also

calm latch Dec 17, 2024, 10:24 PM

#

From EC?

fiery turtle Dec 17, 2024, 10:24 PM

#

yes

calm latch Dec 17, 2024, 10:24 PM

#

Bruh

#

Anyway

#

I'll fix bogus reading

#

I need to know what's going on with that thread

#

Maybe the kernel is panicking

fiery turtle Dec 17, 2024, 10:25 PM

#

yeah

#

Hmm maybe uacpi should also zero-initialize locals, i'll see what NT does

calm latch Dec 17, 2024, 10:26 PM

#

How do you know what it does?

calm latch Dec 17, 2024, 10:27 PM

#

calm latch Maybe the kernel is panicking

I think I'll add IPI panick so all the cores stop and at least I see that it's stuck sunflush

#

Or I could test it in VM

#

Anyway, I think this is irrelevant to uACPI

fiery turtle Dec 17, 2024, 10:28 PM

#

calm latch How do you know what it does?

one sec

calm latch Dec 17, 2024, 10:29 PM

#

(I urgently need a keyboard driver)

fiery turtle Dec 17, 2024, 10:29 PM

#

calm latch How do you know what it does?

#1217009725711847465 message

torpid root Dec 17, 2024, 10:30 PM

#

infy I have a request to make

#

✨ uacpi benchmark leaderboard ✨

fiery turtle Dec 17, 2024, 10:30 PM

#

lmao

#

true true

#

ill maybe add that to the README

calm latch Dec 17, 2024, 10:32 PM

#

lol

fiery turtle Dec 17, 2024, 10:32 PM

#

Please give me your submissions so I can record them, like project link: x ops/s

north holly Dec 17, 2024, 10:33 PM

#

@fiery turtle I think I found a bug in managarm's ec code
https://github.com/managarm/managarm/blob/cdc7ca0c631b222502a44ac151f43cfe66178bcd/kernel/thor/system/acpi/ec.cpp#L302

GitHub

managarm/kernel/thor/system/acpi/ec.cpp at cdc7ca0c631b222502a44ac1...

Pragmatic microkernel-based OS with fully asynchronous I/O - managarm/managarm

#

_GPE can resolve to a package

#

but you only evaluate it as an integer

#

or they do

#

idk who wrote that driver

fiery turtle Dec 17, 2024, 10:34 PM

#

can it?

mortal yoke Dec 17, 2024, 10:34 PM

#

north holly _GPE can resolve to a package

in practice it doesn't

north holly Dec 17, 2024, 10:34 PM

#

north holly Dec 17, 2024, 10:34 PM

#

mortal yoke in practice it doesn't

better safe than sorry

calm latch Dec 17, 2024, 10:34 PM

#

calm latch I need to know what's going on with that thread

Ok I'm able to trigger it in a vm

fiery turtle Dec 17, 2024, 10:35 PM

#

north holly

ah, GPE block devices were never added

#

they dont exist and arent supported

north holly Dec 17, 2024, 10:35 PM

#

wtf

#

acpi spec stoobid

fiery turtle Dec 17, 2024, 10:35 PM

#

ye

north holly Dec 17, 2024, 10:37 PM

#

my ec driver is nearly complete

#

I just need

#

to implement initializing it from the acpi namespace

#

my pc does indeed have an ec

#

wut

#

why is there no _GPE

fiery turtle Dec 17, 2024, 10:39 PM

#

might be there just added later

#

with Scope and stuff

north holly Dec 17, 2024, 10:40 PM

#

only two mentions of _GPE in the entire dsdt

#

so I doubt it

#

both of which in the namespace root

#

first is a Scope(_GPE) in the root

#

and the 2nd is Scope(\_GPE)

#

or smth

#

the other way around

fiery turtle Dec 17, 2024, 10:41 PM

#

do u have any _QXX methods?

north holly Dec 17, 2024, 10:41 PM

#

nope

fiery turtle Dec 17, 2024, 10:42 PM

#

yeah your ec just doesnt have events

#

it doesnt have to

north holly Dec 17, 2024, 10:42 PM

#

so what's really the point of it now then?

fiery turtle Dec 17, 2024, 10:42 PM

#

AML can still access it

#

it stores firmware state

north holly Dec 17, 2024, 10:43 PM

#

guess so

#

I wonder what would happen if I wrote random shit to it

#

what could possibly go wrong

torpid root Dec 17, 2024, 10:44 PM

#

fiery turtle Please give me your submissions so I can record them, like project link: x ops/s

https://github.com/ilobilo/ilobilix meme
it's still very wip though, not even on the master branch

fiery turtle Dec 17, 2024, 10:44 PM

#

north holly *what could possibly go wrong*

well pmos tried and it couldnt boot after a reboot soo

fiery turtle Dec 17, 2024, 10:44 PM

#

torpid root <https://github.com/ilobilo/ilobilix> <:meme:575445094589661215> it's still ***...

thanks, recorded

north holly Dec 17, 2024, 10:44 PM

#

wtf

#

there is no _CRS

#

in the ec device

#

guess it just doesn't exist

fiery turtle Dec 17, 2024, 10:45 PM

#

yeah just a stub ig

calm latch Dec 17, 2024, 10:46 PM

#

My scheduler has a giant skill issue

#

Wtf

#

Wtf

pine leaf Dec 17, 2024, 10:48 PM

#

man now I'm half tempted to revive techflash-os just so that I can compete on the leaderboard meme

fiery turtle Dec 17, 2024, 10:48 PM

#

north holly guess it just doesn't exist

any EmbeddedControl opregions?

calm latch Dec 17, 2024, 10:48 PM

#

calm latch Wtf

This code is two years old

north holly Dec 17, 2024, 10:48 PM

#

nope

calm latch Dec 17, 2024, 10:48 PM

#

And it had a huge bug all along

#

And I was wondering why threads were disappearing

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