uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 30

hollow elm Jan 5, 2025, 9:54 PM

#

i could just merge it ig

fiery turtle Jan 5, 2025, 9:55 PM

#

📎 message.txt

#

yeah LAI didnt survive the test very well

#

although some did pass

#

SUMMARY: AddressSanitizer: 5573531 byte(s) leaked in 30885 allocation(s). ☠️

mortal yoke Jan 5, 2025, 9:56 PM

#

ah I guess that means that the dataregion missing shouldn't affect it

#

weird

fiery turtle Jan 5, 2025, 9:56 PM

#

i commented data region out

#

and recompiled the test

mortal yoke Jan 5, 2025, 9:57 PM

#

can you send that, maybe it works better on qacpi too?

fiery turtle Jan 5, 2025, 9:57 PM

#

📎 bfield.aml

mortal yoke Jan 5, 2025, 9:57 PM

#

thanks

fiery turtle Jan 5, 2025, 9:58 PM

#

hollow elm i could just merge it ig

yeah

fiery turtle Jan 5, 2025, 9:58 PM

#

mortal yoke thanks

how did qacpi do

mortal yoke Jan 5, 2025, 10:01 PM

#

aml debug: 64-bit mode
aml debug: Release of parent ACPICA code 0x2
aml debug: Release of ASLTS test suite  0x16
aml debug: Settings of ASLTS test suite 0x5
aml debug: TEST (functional), Buffer Field
aml debug: m211 test started
aml debug: TEST: m211, Create 1-Bit Buffer Field:
aml debug: :CTST:functional:bfield:m211:PASS:
aml debug: m213 test started
aml debug: TEST: m213, Create 8-Bit Buffer Field:
aml debug: :CTST:functional:bfield:m213:PASS:
aml debug: m215 test started
aml debug: TEST: m215, Create 16-Bit Buffer Field:
0
aml debug: 64-bit mode
aml debug: Release of parent ACPICA code 0x2
aml debug: Release of ASLTS test suite  0x16
aml debug: Settings of ASLTS test suite 0x5
aml debug: TEST (functional), Buffer Field
aml debug: :CTST:functional:bfield:215:PASS:
aml debug: m211 test started
aml debug: TEST: m211, Create 1-Bit Buffer Field:
aml debug: :CTST:functional:bfield:m211:PASS:
aml debug: m213 test started
aml debug: TEST: m213, Create 8-Bit Buffer Field:
aml debug: :CTST:functional:bfield:m213:PASS:
aml debug: m215 test started
aml debug: TEST: m215, Create 16-Bit Buffer Field:
aml debug: :CTST:functional:bfield:m215:PASS:
aml debug: Run time (in seconds): 0x
aml debug: The total number of exceptions handled: 0x
aml debug: ========= ROOT METHODS SUMMARY (max 600):
aml debug: :STST:functional:bfield:m211:PASS:
aml debug: :STST:functional:bfield:m213:PASS:
aml debug: :STST:functional:bfield:215:PASS:
aml debug: :STST:functional:bfield:m211:PASS:
aml debug: :STST:functional:bfield:m213:PASS:
aml debug: :STST:functional:bfield:m215:PASS:
aml debug: ========= END.
aml debug: TEST ACPICA: 64-bit : PASS

fiery turtle Jan 5, 2025, 10:01 PM

#

fiery turtle

back when i was just working on field read code (like a year ago) i was having some skill issues with edge cases in this case so tdb helped me with this bit_copy function

fiery turtle Jan 5, 2025, 10:01 PM

#

mortal yoke ``` aml debug: 64-bit mode aml debug: Release of parent ACPICA code 0x2 aml debu...

nice!

fiery turtle Jan 5, 2025, 10:02 PM

#

mortal yoke ``` aml debug: 64-bit mode aml debug: Release of parent ACPICA code 0x2 aml debu...

lol u log it the same way as lai

#

just with a space

fiery turtle Jan 5, 2025, 10:03 PM

#

mortal yoke ``` aml debug: 64-bit mode aml debug: Release of parent ACPICA code 0x2 aml debu...

also:

[uACPI][TRACE] [AML DEBUG] String => "========= ROOT METHODS SUMMARY (max 600):"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m211:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m213:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m215:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m217:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m219:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m21d:PASS:"
[uACPI][TRACE] [AML DEBUG] String => "========= END."

U lost the m somewhere

#

and the last thing should be 21d i think

fiery turtle Jan 5, 2025, 10:04 PM

#

mortal yoke ``` aml debug: 64-bit mode aml debug: Release of parent ACPICA code 0x2 aml debu...

actually i might know why

#

did u run MAIN?

#

it already does it while loading the table to work in LAI

mortal yoke Jan 5, 2025, 10:04 PM

#

yes

fiery turtle Jan 5, 2025, 10:04 PM

#

if i run the body + MAIN i get this

[uACPI][TRACE] [AML DEBUG] String => "========= ROOT METHODS SUMMARY (max 600):"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m211:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m213:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m215:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m217:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m219:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m21d:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m211:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m213:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m215:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m217:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m219:PASS:"
[uACPI][TRACE] [AML DEBUG] String => ":STST:functional:bfield:m21d:PASS:"
[uACPI][TRACE] [AML DEBUG] String => "========= END."

#

somehow your array gets truncated or something

mortal yoke Jan 5, 2025, 10:07 PM

#

yeah idk

fiery turtle Jan 5, 2025, 10:08 PM

#

It prints these from a huge package

hollow elm Jan 5, 2025, 10:08 PM

#

fiery turtle lol u log it the same way as lai

well qacpi doesn't log the type and lai logs it as a sexp of (type value)

fiery turtle Jan 5, 2025, 10:09 PM

#

a sexp? flush

hollow elm Jan 5, 2025, 10:09 PM

#

s expression

fiery turtle Jan 5, 2025, 10:10 PM

#

what is that?

#

ah nvm

#

like (string <value>)

mortal yoke Jan 5, 2025, 10:10 PM

#

also about the arithmetic test I am not sure if its because of the data region or do I actually have that many broken things lol

📎 message.txt

vast kestrel Jan 5, 2025, 10:10 PM

#

did I hear lai

hollow elm Jan 5, 2025, 10:11 PM

#

mortal yoke also about the arithmetic test I am not sure if its because of the data region o...

interesting integers ```
aml debug: **** Expected Result: 0xp004, (p004)
aml debug: **** Actual Result : 0x0x7, (7)

fiery turtle Jan 5, 2025, 10:11 PM

#

mortal yoke also about the arithmetic test I am not sure if its because of the data region o...

I can patch it for u

hollow elm Jan 5, 2025, 10:11 PM

#

0xp004 and 0x0x7

mortal yoke Jan 5, 2025, 10:11 PM

#

hollow elm interesting integers ``` aml debug: **** Expected Result: 0xp004, (p004) aml deb...

yes idk what is that lol

#

or well the second one I can kinda know why

#

there is probably 0x as a part of the string and then it prints the integer after which itself adds 0x

fiery turtle Jan 5, 2025, 10:12 PM

#

📎 arithmetic.aml

#

📎 message.txt

#

LAI passess all but SHL/SHR

#

this is what I get

📎 message.txt

hollow elm Jan 5, 2025, 10:15 PM

#

fiery turtle

strange that we leak stuff

#

wonder if it's just due to Debug = ...

fiery turtle Jan 5, 2025, 10:15 PM

#

it creates a lot of temporary nodes

#

i think those are leaked

#

And some strings are leaked as well yeah

fiery turtle Jan 5, 2025, 10:16 PM

#

hollow elm 0xp004 and 0x0x7

☠️

mortal yoke Jan 5, 2025, 10:18 PM

#

I found two out of bounds shifts that it did and after adding checks for those I get aml debug: TEST ACPICA: 64-bit : FAIL : Errors # 0xJ, Failed tests # 0x1 and the failing one is FindSetLeftBit

fiery turtle Jan 5, 2025, 10:19 PM

#

Ah yes 0xJ

#

the new hex number

#

Concatenate (Local0, "FAIL:Errors # ", Local2) @mortal yoke your concatenate might be broken i guess?

#

or wait

#

Concatenate (Local0, "FAIL:Errors # ", Local2)
B000 = Local7
Concatenate (Local2, B000, Local0)
Concatenate (Local0, ":", Local1)
ERR6++

#

yeah

#

i guess your concatenate is not feeling welll

mortal yoke Jan 5, 2025, 10:21 PM

#

maybe

#

Ill check that after I fix the fsl

fiery turtle Jan 5, 2025, 10:21 PM

#

i remember the bit finding ops have slightly weird semantics

#

but should be simple to fix

#

uacpi_u8 uacpi_bit_scan_forward(uacpi_u64 value)
{
#ifdef _MSC_VER
    unsigned char ret;
    unsigned long index;

#ifdef _WIN64
    ret = _BitScanForward64(&index, value);
    if (ret == 0)
        return 0;

    return (uacpi_u8)index + 1;
#else
    ret = _BitScanForward(&index, value);
    if (ret == 0) {
        ret = _BitScanForward(&index, value >> 32);
        if (ret == 0)
            return 0;

        return (uacpi_u8)index + 33;
    }

    return (uacpi_u8)index + 1;
#endif

#else
    return __builtin_ffsll(value);
#endif
}

#

i think i return this value raw

hollow elm Jan 5, 2025, 10:23 PM

#

gnu on top their builtin maps onto aml directly meme

fiery turtle Jan 5, 2025, 10:23 PM

#

lmao

#

gnu builtins designed specifically for aml

#

the backward one is slightly more ugly:

if (value == 0)
    return 0;

return 64 - __builtin_clzll(value);

north holly Jan 5, 2025, 10:29 PM

#

__builtin_ctzll?

fiery turtle Jan 5, 2025, 10:30 PM

#

perhaps, i dont remember what any of those do

north holly Jan 5, 2025, 10:30 PM

#

iirc that compiles to bsf

#

https://godbolt.org/z/d11M9WT7T

Compiler Explorer - C++ (x86-64 gcc 14.2)

int ctz(int a)
{
return __builtin_ctz(a);
}
int clz(int a)
{
return __builtin_clz(a);
}

fiery turtle Jan 5, 2025, 10:32 PM

#

why the hell is there a rep there lmao

left orbit Jan 5, 2025, 10:33 PM

#

don't think it does anything

north holly Jan 5, 2025, 10:33 PM

#

probably just padding

#

https://godbolt.org/z/a9ncvWYPK
at -O1 it goes away

Compiler Explorer - C++ (x86-64 gcc 14.2)

int ctz(int a)
{
return __builtin_ctz(a);
}
int clz(int a)
{
return __builtin_clz(a);
}

fiery turtle Jan 5, 2025, 10:33 PM

#

ah

north holly Jan 5, 2025, 10:34 PM

#

even at -Og it goes away

#

so probably nothing important

fiery turtle Jan 5, 2025, 10:35 PM

#

oh this one is actually interesting, I think yeah it actually hangs because the SMI handler gets some bogus data from lai

#

idk if a spurious read can cause that tho

hollow elm Jan 5, 2025, 10:36 PM

#

hmm i thought it was just due to the extra read from the smi register

fiery turtle Jan 5, 2025, 10:36 PM

#

usually the way this works is theres some NVS region that stores the SMM opcode + args

#

and then it just invokes SMI

#

but who knows how this specific one works

mortal yoke Jan 5, 2025, 10:38 PM

#

fiery turtle the backward one is slightly more ugly: ```c if (value == 0) return 0; retu...

I don't understand how is that different from what I do which is basically that done manually (but somehow the result is different and idk why) ```c
result = 0;
for (int i = int_size * 8; i > 0; --i) {
if (int_value & uint64_t {1} << (i - 1)) {
result = (int_size * 8) - i + 1;
break;
}
}

fiery turtle Jan 5, 2025, 10:39 PM

#

mortal yoke I don't understand how is that different from what I do which is basically that ...

maybe its because u dont do parens

#

like i have no idea what the order here is

mortal yoke Jan 5, 2025, 10:39 PM

#

& has less precedence than shifts

fiery turtle Jan 5, 2025, 10:39 PM

#

i guess if ur confident about that

fiery turtle Jan 5, 2025, 10:40 PM

#

mortal yoke I don't understand how is that different from what I do which is basically that ...

dont see anything immediately wrong ig

mortal yoke Jan 5, 2025, 10:58 PM

#

oh right its the bit location, I though it was supposed to be the other way around (so if the msb would be set it would return 1 and if lsb then 64)

#

also the concat thing is because of ascii detection

fiery turtle Jan 5, 2025, 11:42 PM

#

mortal yoke also the concat thing is because of ascii detection

wdym?

mortal yoke Jan 5, 2025, 11:44 PM

#

it tries to convert the rhs integer to a string as the lhs is a string using the same semantics as Store basically

fiery turtle Jan 5, 2025, 11:44 PM

#

ah u dont do integer_to_string

#

uacpi 1.0 is going well 💀

mortal yoke Jan 5, 2025, 11:47 PM

#

fiery turtle ah u dont do integer_to_string

the concat does have its own special to string impl but I only use it if the one with the store semantics fails, maybe I should just always use it

mortal yoke Jan 5, 2025, 11:48 PM

#

fiery turtle uacpi 1.0 is going well 💀

💀 what are you trying to change?

fiery turtle Jan 5, 2025, 11:49 PM

#

nvm it was the dumbest thing

fiery turtle Jan 5, 2025, 11:49 PM

#

mortal yoke 💀 what are you trying to change?

i have like 99999 changes everywhere to support these special fields

mortal yoke Jan 5, 2025, 11:50 PM

#

ah

fiery turtle Jan 5, 2025, 11:50 PM

#

some of them also support direct read besides wtr

#

so i have to support propagating the response buffer from a normal read as well

#

anyway its just hoisting crap from one place to another everywhere

#

and also new structs for op attach and r/w

frank canopy Jan 5, 2025, 11:55 PM

#

what are these special fields, of curiosity?

fiery turtle Jan 5, 2025, 11:59 PM

#

IPMI, SMBus, GenericSerialBus, FFixedHW, FCC, GeneralPurposeIO (less annoying to support), PCC (Even less annoying)

#

GenericSerialBus is the most annoying of all

#

not only do u have to propagate the smbus protocol, field attribute, i2c controller connection, but u also have to allocate correctly sized buffers for the response and then propagate back to aml

#

What's funny is SMBus is the address space to use for SMBus, but there isnt a single usage in any AML blobs nor does linux support it at all

#

They use GenericSerialBus for SMBus instead for whatever reason

#

❯ grep -rni ', GenericSerialBus' -l | wc -l
16

#

only 16 out of 500 blobs have it

#

thats why its generally a very niche thing

#

only needed for tablets and SoC

#

e.g. microsoft surface

#

for IPMI i have only one blob

#

❯ grep -rni ', GeneralPurposeIO' -l | wc -l
11

#

even less for gpio

#

❯ grep -rni ', SMBus' -l | wc -l
0
❯ grep -rni ', FCC' -l | wc -l
4
❯ grep -rni ', FFixedHW' -l | wc -l
0
❯ grep -rni ', IPMI' -l | wc -l
1

#

not much to work with

#

nvm, for FCC there's also 0, my expresion was just wrong

frank canopy Jan 6, 2025, 12:06 AM

#

mhm

fiery turtle Jan 6, 2025, 12:08 AM

#

Waaaait

#

❯ grep -rni ', PlatformRtMechanism,'
Server/Intel/ArcherCity/ArcherCity/356135E81941:1842143:            OperationRegion (PRMR, PlatformRtMechanism, 0x00, 0x01)

#

thats what its called apparently lmao

#

Device (PRMT)
{
    Name (_HID, "80860223")  // _HID: Hardware ID
    Name (_CID, "80860223")  // _CID: Compatible ID
    Name (_DDN, "PRM Test Device")  // _DDN: DOS Device Name
    OperationRegion (PRMR, PlatformRtMechanism, 0x00, 0x01)
    Field (PRMR, BufferAcc, NoLock, Preserve)
    {
        PRMF,   8
    }

    Method (RUNS, 1, NotSerialized)
    {
        Local0 = Buffer (0x1A) {}
        CreateByteField (Local0, 0x00, PSTA)
        CreateQWordField (Local0, 0x01, USTA)
        CreateByteField (Local0, 0x09, CMD)
        CreateField (Local0, 0x50, 0x80, GUID)
        CMD = 0x00
        GUID = Arg0
        Local0 = PRMF = Local0
        Return (PSTA) /* \_SB_.PRMT.RUNS.PSTA */
    }

    Method (LCKH, 1, NotSerialized)
    {
        Local0 = Buffer (0x1A) {}
        CreateByteField (Local0, 0x00, STAT)
        CreateByteField (Local0, 0x09, CMD)
        CreateField (Local0, 0x50, 0x80, GUID)
        CMD = 0x01
        GUID = Arg0
        Local0 = PRMF = Local0
        Return (STAT) /* \_SB_.PRMT.LCKH.STAT */
    }

    Method (ULCK, 1, NotSerialized)
    {
        Local0 = Buffer (0x1A) {}
        CreateByteField (Local0, 0x00, STAT)
        CreateByteField (Local0, 0x09, CMD)
        CreateField (Local0, 0x50, 0x80, GUID)
        CMD = 0x02
        GUID = Arg0
        Local0 = PRMF = Local0
        Return (STAT) /* \_SB_.PRMT.ULCK.STAT */
    }

    Method (_STA, 0, NotSerialized)  // _STA: Status
    {
        Return (0x0B)
    }
}

#

thats pretty cool

fiery turtle Jan 6, 2025, 12:15 AM

#

fiery turtle ```c Device (PRMT) { Name (_HID, "80860223") // _HID: Hardware ID Name ...

@mortal yoke see what it does with that PRMF field?

#

thats basically how you're supposed to do it from aml

#

and that field bit width u just ignore completely because why not

#

they dont even have to specify anything sane there

#

just 8 usually

mortal yoke Jan 6, 2025, 12:18 AM

#

if there are multiple fields do they all work the same too?

fiery turtle Jan 6, 2025, 12:18 AM

#

some protocols care about the absolute offset of the field relative to opregion base as that indicates some sort of command value

#

some protocols dont because they have a AccessAs + Connection before every field

#

like GPIO

#

OperationRegion (GPOP, GeneralPurposeIo, Zero, 0x0C)
Field (GPOP, ByteAcc, NoLock, Preserve)
{
    Connection (
        GpioIo (Exclusive, PullDefault, 0x0000, 0x0000, IoRestrictionOutputOnly,
            "\\_SB.GPO1", 0x00, ResourceConsumer, ,
            )
            {   // Pin list
                0x000F
            }
    ), 
    BST5,   1, 
    Connection (
        GpioIo (Exclusive, PullDefault, 0x0000, 0x0000, IoRestrictionOutputOnly,
            "\\_SB.GPO1", 0x00, ResourceConsumer, ,
            )
            {   // Pin list
                0x001A
            }
    ), 
    TCD3,   1
}

example

mortal yoke Jan 6, 2025, 12:21 AM

#

interesting

fiery turtle Jan 6, 2025, 12:21 AM

#

here the offset of the field means absolutely nothing

#

because a Connection resets it

mortal yoke Jan 6, 2025, 12:21 AM

#

do you also expose the connection somehow? or is it just used internally

fiery turtle Jan 6, 2025, 12:21 AM

#

yup u do absolutely

#

the handler must be able to parse the resources from it

#

to know which i2c to talk to

#

typedef struct uacpi_region_gpio_rw_data
{
    void *handler_context;
    void *region_context;
    uacpi_data_view connection;
    uacpi_u32 pin_offset;
    uacpi_u32 num_pins;
    uacpi_u64 value;
} uacpi_region_gpio_rw_data;

#

this is what uACPI is going to provide for GPIO

#

pin offset is basically offset since previous Connection

#

num_pins is the bit width of the field

#

then u would use uacpi_get_resources_from_buffer to parse the connnection attribute

#

from that u know which i2c and which pin from that i2c

#

or which gpio in this case

#

GenericSerialBus is similar idea to this but way more complex

#

since there are a lot of SMBus protocols with AccessAs

#

Name (XPMC, ResourceTemplate ()
{
    I2cSerialBusV2 (0x0034, ControllerInitiated, 0x00061A80,
        AddressingMode7Bit, "\\_SB.I2C5",
        0x00, ResourceConsumer, , Exclusive,
        )
})
OperationRegion (XSCG, GenericSerialBus, Zero, 0x0100)
Field (XSCG, BufferAcc, NoLock, Preserve)
{
    Connection (XPMC), 
    AccessAs (BufferAcc, AttribByte), 
    XP00,   8, 
    XP01,   8
}

Field (XSCG, BufferAcc, NoLock, Preserve)
{
    Connection (XPMC), 
    Offset (0xB9), 
    AccessAs (BufferAcc, AttribByte), 
    XPB9,   8
}

Example for SMBus

#

AttribByte means this

#

there's like 10 types u must handle

winter orbit Jan 6, 2025, 12:50 AM

#

fiery turtle this is what uACPI is going to provide for GPIO

lol

#

This is basically all windows provides too

#

Plus some extra windows specific stuff

fiery turtle Jan 6, 2025, 12:51 AM

#

nice

mortal yoke Jan 6, 2025, 1:00 AM

#

I think I finally fixed Load leaks in qacpi, it was mostly just me being dumb and doing ```cpp
if (method_frame.node_link) {
method_frame.node_link->link = context->all_nodes;
context->all_nodes = method_frame.node_link;
}

#

(method_frame.node_link starts a singly linked list and its link might not be null if its part of a larger chain so the other nodes would just be lost in the void)

fiery turtle Jan 6, 2025, 1:02 AM

#

what does this do?

frank canopy Jan 6, 2025, 1:02 AM

#

fiery turtle IPMI, SMBus, GenericSerialBus, FFixedHW, FCC, GeneralPurposeIO (less annoying to...

oh those things

hollow elm Jan 6, 2025, 1:03 AM

#

what's PlatformRtMechanism?

fiery turtle Jan 6, 2025, 1:03 AM

#

https://elixir.bootlin.com/linux/v6.13-rc3/source/drivers/acpi/prmt.c

prmt.c - drivers/acpi/prmt.c - Linux source code v6....

Elixir Cross Referencer - source code of Linux v6.13-rc3: drivers/acpi/prmt.c

#

this thing

#

AML talking to EFI firmware or something idk

hollow elm Jan 6, 2025, 1:04 AM

#

lol

mortal yoke Jan 6, 2025, 1:04 AM

#

fiery turtle what does this do?

adds namespace nodes created inside Load to the context's node list that is used to keep track of all global nodes that are supposed to be deleted when the context is destroyed (originally they are added to the method frame's list and destroyed from there if the Load fails)

hollow elm Jan 6, 2025, 1:05 AM

#

2 firmware teams couldn't talk to eachother so instead they invented a new mechanism for the firmware to talk to itself

fiery turtle Jan 6, 2025, 1:05 AM

#

hollow elm 2 firmware teams couldn't talk to eachother so instead they invented a new mecha...

lmao

#

basically

fiery turtle Jan 6, 2025, 1:05 AM

#

mortal yoke adds namespace nodes created inside Load to the context's node list that is used...

ohhh

#

so u support Load and LoadTable now?

mortal yoke Jan 6, 2025, 1:06 AM

#

Load yes but LoadTable no because I still haven't implemented the table api

#

though as LoadTable is basically exactly the same as Load but with data from the table api it should just work

fiery turtle Jan 6, 2025, 1:06 AM

#

yeah

#

it has annoying semantics with scopes tho

frank canopy Jan 6, 2025, 1:06 AM

#

yall made me curious and turns out my laptop only uses systemio, systemmemory, and a single instance of PCI_Config opregions

fiery turtle Jan 6, 2025, 1:07 AM

#

yeah i mean

#

16 out of 500 use these special opregions

frank canopy Jan 6, 2025, 1:07 AM

#

this is special niche stuff for a reason ig lol

fiery turtle Jan 6, 2025, 1:07 AM

#

mostly tablets/mini pcs/chinese laptops

#

for IPMI and prmt stuff its servers

frank canopy Jan 6, 2025, 1:09 AM

#

ah ok

#

anything actually use the pci bar opregion?

#

thats the one im most unsure how id manage to implement so im curious

fiery turtle Jan 6, 2025, 1:12 AM

#

frank canopy anything actually use the pci bar opregion?

funny enough no

#

its not supported by linux nor used by anyone

frank canopy Jan 6, 2025, 1:12 AM

#

(i should REALLY get to finishing off my new memory manager so i can do more device drivers and work more on uacpi zig bindings tbh)

#

oh interesting

fiery turtle Jan 6, 2025, 1:13 AM

#

frank canopy thats the one im most unsure how id manage to implement so im curious

wdym?

#

like isnt it just map(read_bar(), ...)

frank canopy Jan 6, 2025, 1:13 AM

#

you can have io port bars too i think and overall im not sure how i plan to handle BAR management in general yet

#

unless i can rely on firmware setting up the BARs for me in which case yeah its probably just that

#

but "rely on firmware" is the sort of thing that sounds like a bad idea imo lol

fiery turtle Jan 6, 2025, 1:15 AM

#

yeah u have to have an abstraction layer for this

#

so that bar type isnt exposed to each user

#

since its not relevant

#

u just want like io_map io_read io_write etc

frank canopy Jan 6, 2025, 1:16 AM

#

yeah

fiery turtle Jan 6, 2025, 1:16 AM

#

and io_read would dispatch to in/out or mmio

frank canopy Jan 6, 2025, 1:16 AM

#

ive got that for pci config already and it works great. even lets me easily add some possible efficiency gains with the map_pci things

fiery turtle Jan 6, 2025, 1:17 AM

#

nice

frank canopy Jan 6, 2025, 1:18 AM

#

i literally just allocate the structure that im already using for holding pci config access info for my drivers and my pci config read/write functions take that struct and the register offset its nice. im debating giving that a dedicated pool once i get allocation working more too

fiery turtle Jan 6, 2025, 1:18 AM

#

yeah thats pretty cool

frank canopy Jan 6, 2025, 1:19 AM

#

im assuming issue 75 is the to-do for 1.0 btw?

fiery turtle Jan 6, 2025, 1:21 AM

#

its outdated but yeah

#

See if we need special handling for SMBus and similar address spaces basically im doing this

#

the other two are not important

winter orbit Jan 6, 2025, 1:22 AM

#

frank canopy but "rely on firmware" is the sort of thing that sounds like a bad idea imo lol

Can confirm

#

meme

fiery turtle Jan 6, 2025, 1:22 AM

#

ultrameme

frank canopy Jan 6, 2025, 1:24 AM

#

lol

fiery turtle Jan 6, 2025, 1:25 AM

#

i think linux tries to claim the existing bar address (aka sanity check it and allocate internally)

#

if that fails it assigns a new address

frank canopy Jan 6, 2025, 1:25 AM

#

mhm

fiery turtle Jan 6, 2025, 1:26 AM

#

but most firmware should be sane enough to do bar allocation correctly

frank canopy Jan 6, 2025, 1:26 AM

#

cool

fiery turtle Jan 6, 2025, 1:26 AM

#

problems start appearing when u have devices with huge bars

#

not all firmware is ready for those

calm latch Jan 6, 2025, 1:43 AM

#

fiery turtle its not supported by linux nor used by anyone

Does Windows implement all ACPI features?

#

(out of curiosity)

fiery turtle Jan 6, 2025, 1:49 AM

#

calm latch Does Windows implement all ACPI features?

No idea, this isnt used by any firmware so I imagine its because of nt not supporting it

#

Maybe darkfire knows if windows supports PciBarTarget

winter orbit Jan 6, 2025, 2:24 AM

#

I don’t think it is

#

Windows pci hasn’t gotten a lot of love in recent years

#

I haven’t investigated win11 pci though

rustic compass Jan 6, 2025, 7:25 AM

#

fiery turtle but most firmware should be sane enough to do bar allocation correctly

How should an os do bar allocations? I haven't seen a mention about that in the pcie 5.0 specs (I searched for something else so I could have missed it).

vast kestrel Jan 6, 2025, 8:03 AM

#

hollow elm what's `PlatformRtMechanism`?

So you know how currently random drivers are implemented in smm and acpi calls them? Well to avoid that now instead the firmware will give you PE files you will map yourself and run yourself (along side the MMIO regions they need), and a way for them to communicate with acpi

#

Funnily enough one of the features here is to allow line updates of these modules

fiery turtle Jan 6, 2025, 10:03 AM

#

rustic compass How should an os do bar allocations? I haven't seen a mention about that in the ...

Just by allocating arbitrary addresses outside of the reserved e820 ranges as far as I understand

#

Or rather, addresses outside of all ranges mentioned in e820

vast kestrel Jan 6, 2025, 10:10 AM

#

Depends a bit on controller, in x86 it can be pretty much anywhere not in ram

fiery turtle Jan 6, 2025, 10:11 AM

#

Ye

vast kestrel Jan 6, 2025, 10:11 AM

#

On alot of arm boards they have a specific decode range you can allocate from

#

Linux does reallocate bars on x86 for sriov

flat badge Jan 6, 2025, 11:17 AM

#

That's not true (that any address outside the e820/uefi map works on x86)

#

You should look at _CRS and allocate within the address regions

fiery turtle Jan 6, 2025, 11:20 AM

#

at crs of what?

flat badge Jan 6, 2025, 11:20 AM

#

The root bridge

fiery turtle Jan 6, 2025, 11:20 AM

#

really?

flat badge Jan 6, 2025, 11:20 AM

#

Yes, of course. You can't just allocate random addresses to the root bridge

#

And if the BARs is not directly behind the root bridge, you need to take the address windows of intermediate bridges into account as well

#

But these are in PCI config space, not in ACPI

fiery turtle Jan 6, 2025, 11:23 AM

#

ill double check that

#

Name (_CRS, ResourceTemplate ()  // _CRS: Current Resource Settings
{
    WordBusNumber (ResourceProducer, MinFixed, MaxFixed, PosDecode,
        0x0000,             // Granularity
        0x0000,             // Range Minimum
        0x00FF,             // Range Maximum
        0x0000,             // Translation Offset
        0x0100,             // Length
        ,, )
    IO (Decode16,
        0x0CF8,             // Range Minimum
        0x0CF8,             // Range Maximum
        0x01,               // Alignment
        0x08,               // Length
        )
    WordIO (ResourceProducer, MinFixed, MaxFixed, PosDecode, EntireRange,
        0x0000,             // Granularity
        0x0000,             // Range Minimum
        0x0CF7,             // Range Maximum
        0x0000,             // Translation Offset
        0x0CF8,             // Length
        ,, , TypeStatic, DenseTranslation)
    WordIO (ResourceProducer, MinFixed, MaxFixed, PosDecode, EntireRange,
        0x0000,             // Granularity
        0x0D00,             // Range Minimum
        0xFFFF,             // Range Maximum
        0x0000,             // Translation Offset
        0xF300,             // Length
        ,, , TypeStatic, DenseTranslation)
    DWordMemory (ResourceProducer, PosDecode, MinFixed, MaxFixed, Cacheable, ReadWrite,
        0x00000000,         // Granularity
        0x000A0000,         // Range Minimum
        0x000BFFFF,         // Range Maximum
        0x00000000,         // Translation Offset
        0x00020000,         // Length
        ,, , AddressRangeMemory, TypeStatic)
    DWordMemory (ResourceProducer, PosDecode, MinFixed, MaxFixed, NonCacheable, ReadWrite,
        0x00000000,         // Granularity
        0x08000000,         // Range Minimum
        0xAFFFFFFF,         // Range Maximum
        0x00000000,         // Translation Offset
        0xA8000000,         // Length
        ,, , AddressRangeMemory, TypeStatic)
    DWordMemory (ResourceProducer, PosDecode, MinFixed, MaxFixed, NonCacheable, ReadWrite,
        0x00000000,         // Granularity
        0xC0000000,         // Range Minimum
        0xFEBFFFFF,         // Range Maximum
        0x00000000,         // Translation Offset
        0x3EC00000,         // Length
        ,, , AddressRangeMemory, TypeStatic)
    QWordMemory (ResourceProducer, PosDecode, MinFixed, MaxFixed, Cacheable, ReadWrite,
        0x0000000000000000, // Granularity
        0x0000000100000000, // Range Minimum
        0x00000008FFFFFFFF, // Range Maximum
        0x0000000000000000, // Translation Offset
        0x0000000800000000, // Length
        ,, , AddressRangeMemory, TypeStatic)
})

#

this is what they have on qemu

#

@flat badge does this mean u can place BARs in any of these?

#

It even specifies the allowed bus number range

#

pretty interesting

#

TIL

flat badge Jan 6, 2025, 11:30 AM

#

yes

fiery turtle Jan 6, 2025, 11:32 AM

#

flat badge yes

so if u dont have an aml interpreter u cant even know where to allocate bars?

flat badge Jan 6, 2025, 11:34 AM

#

Yes. Unless you're the firmware I guess

#

or if you have a chipset driver that knows the allowed ranges

calm latch Jan 6, 2025, 11:34 AM

#

You need to have a bar allocator if you want to support USB 4/Thunderbolt, right?

fiery turtle Jan 6, 2025, 11:37 AM

#

for any sort of hotplugging

#

well not normal usb devices

flat badge Jan 6, 2025, 11:43 AM

#

There are situations where not BARs are allocated

#

Iirc omvf for example only initializes the first segment

#

And for arm/riscv you also need BAR allocation

fiery turtle Jan 6, 2025, 11:45 AM

#

wtf are sufrace laptops checking for LULW

#

microsoft had to add quirk handling to their own laptop

#

ADBG ("Let\'s bring TBT RPs out of D3Cold")

#

ADBG ("Drop TG0N due to it is already exit D3 cold")

#

indian outsource firmware

calm latch Jan 6, 2025, 11:53 AM

#

flat badge And for arm/riscv you also need BAR allocation

In QEMU ovmf allocated bards for me on RISC-V

fiery turtle Jan 6, 2025, 11:55 AM

#

@winter orbit btw I just realized, by default uACPI will respond true to _OSI queries up to Windows 11 22H2

#

u probably want to disable all that ReactOS doesnt support

winter orbit Jan 6, 2025, 11:56 AM

#

yolo go strong go hard

#

no im kidding good idea lol

fiery turtle Jan 6, 2025, 11:56 AM

#

lol

#

fyi its in the osi.h header

#

it should probably work as is, but idk if react os has an ec driver for example

#

which firmware might assume unconditional prescense of

#

etc

winter orbit Jan 6, 2025, 11:57 AM

#

uacpi_latest_queried_vendor_interface would probably be funny to make try and just read the windows version from ntoskrnl properly

#

some kind of overload

winter orbit Jan 6, 2025, 11:58 AM

#

fiery turtle it should probably work as is, but idk if react os has an ec driver for example

i had to make one to boot win10

#

it's shoved inside acpi.sys

fiery turtle Jan 6, 2025, 11:58 AM

#

winter orbit `uacpi_latest_queried_vendor_interface` would probably be funny to make try and ...

yeah

fiery turtle Jan 6, 2025, 11:58 AM

#

winter orbit i had to make one to boot win10

lmao

kindred beacon Jan 6, 2025, 1:52 PM

#

uACPI, the new standard implementation™️

flat badge Jan 6, 2025, 1:55 PM

#

calm latch In QEMU ovmf allocated bards for me on RISC-V

yes, but uefi is not usable on all (most?) risc-v boards

calm latch Jan 6, 2025, 1:55 PM

#

btw, does anyone know of good cheap risc-v sbcs?

flat badge Jan 6, 2025, 1:56 PM

#

i have a bpi-f3

#

and you can look at the milk-v ones

#

they should also be decent

calm latch Jan 6, 2025, 1:57 PM

#

Which hardware does it have?

#

Like USB is XHCI and not prorietary weirdness like raspberry pi

#

?

#

oof

calm latch Jan 6, 2025, 2:20 PM

#

Looks like normal EHCI KEKW

flat badge Jan 6, 2025, 2:25 PM

#

yeah i think it's xhci in general but it's wrapped into some priorietary device (the dwc3-spacemit thing)

#

the linux kernel patches to make this work are open source, but i haven't looked at them

#

for the NIC it doesn't look bad, no idea how it is for the usb

fiery turtle Jan 6, 2025, 6:33 PM

#

groundwork for esoteric field support done

#

A few more regions left, but they reuse the bulk of logic from IPMI

#

well and also the bulk from GPIO

loud ice Jan 6, 2025, 6:37 PM

#

IPMI?

fiery turtle Jan 6, 2025, 6:37 PM

#

https://en.wikipedia.org/wiki/Intelligent_Platform_Management_Interface

Intelligent Platform Management Interface

The Intelligent Platform Management Interface (IPMI) is a set of computer interface specifications for an autonomous computer subsystem that provides management and monitoring capabilities independently of the host system's CPU, firmware (BIOS or UEFI) and operating system. IPMI defines a set of interfaces used by system administrators for out-o...

#

power management for server equipment

loud ice Jan 6, 2025, 6:45 PM

#

what does uacpi have to do with IPMI???

fiery turtle Jan 6, 2025, 6:46 PM

#

.

#

TLDR AML can access IPMI for various configuration & counters

#

in this case power meter device

left orbit Jan 6, 2025, 6:48 PM

#

does acpica do those?

fiery turtle Jan 6, 2025, 6:49 PM

#

yes: https://elixir.bootlin.com/linux/v6.13-rc3/source/drivers/acpi/acpi_ipmi.c#L524

acpi_ipmi.c - drivers/acpi/acpi_ipmi.c - Linux sourc...

Elixir Cross Referencer - source code of Linux v6.13-rc3: drivers/acpi/acpi_ipmi.c

loud ice Jan 6, 2025, 6:55 PM

#

that actually sounds quite useful

fiery turtle Jan 6, 2025, 6:57 PM

#

its pretty cool yeah

thorny orbit Jan 6, 2025, 7:20 PM

#

What do I need to implement in the kernel api for hooking power buttons ? (on systems which don't have ECDT)

fiery turtle Jan 6, 2025, 7:27 PM

#

Attaching IPMI
IPMI command deadbeef
[uACPI][TRACE] write-then-read from [\MAIN.VARM] IPMI[0x00000000DEADBEEF] = <buffer of 66 bytes>
[uACPI][TRACE] [AML DEBUG] Buffer @0000603000001cc0 (0000603000001d20) (66 bytes)

Yayy

#

OperationRegion (VARM, IPMI, 0xDEADBEEF, 0x08)
Field (VARM, BufferAcc, NoLock, Preserve) {
    AccessAs (BufferAcc, 0x01),
    FIL0, 32,
}

Debug = FIL0 = 123

fiery turtle Jan 6, 2025, 7:29 PM

#

thorny orbit What do I need to implement in the kernel api for hooking power buttons ? (on sy...

all locking related stuff

#

and irq handler

thorny orbit Jan 6, 2025, 7:29 PM

#

fiery turtle all locking related stuff

and I can init it how it showed in wiki page ?

fiery turtle Jan 6, 2025, 7:30 PM

#

try and find out

#

if u want to be able to shutdown also u must implement every kernel api

#

well on qemu u can get away with less

#

but i cant tell u the exact subset

thorny orbit Jan 6, 2025, 7:31 PM

#

thorny orbit and I can init it how it showed in wiki page ?

like this

Screenshot_2025-01-06-22-31-03-101_com.android.chrome-edit.jpg

thorny orbit Jan 6, 2025, 7:31 PM

#

fiery turtle if u want to be able to shutdown also u must implement every kernel api

uacpi 😭😭

loud ice Jan 6, 2025, 7:32 PM

#

on qemu you only need rawio

fiery turtle Jan 6, 2025, 7:32 PM

#

where did u get this fucked formatting from

loud ice Jan 6, 2025, 7:32 PM

#

i think

thorny orbit Jan 6, 2025, 7:32 PM

#

fiery turtle where did u get this fucked formatting from

mobile

fiery turtle Jan 6, 2025, 7:32 PM

#

yeah on qemu u only need io & irq handling

loud ice Jan 6, 2025, 7:32 PM

#

fiery turtle where did u get this fucked formatting from

word wrap

fiery turtle Jan 6, 2025, 7:32 PM

#

loud ice word wrap

not much to wrap here

loud ice Jan 6, 2025, 7:32 PM

#

fiery turtle not much to wrap here

phone screen tho

fiery turtle Jan 6, 2025, 7:32 PM

#

bruh

#

qemu should add a _PTS method to make it more annoying

#

so that people dont get away with half assing uacpi kernel api

loud ice Jan 6, 2025, 7:33 PM

#

lol

fiery turtle Jan 6, 2025, 7:33 PM

#

and also require an ec on top of that

#

trl

loud ice Jan 6, 2025, 7:34 PM

#

perfect

fiery turtle Jan 6, 2025, 7:34 PM

#

thorny orbit like this

there's a difference between just receiving the event and actually handling it

loud ice Jan 6, 2025, 7:34 PM

#

i kinda want to write a system emulator and if i do i will definitely run a riscv core

#

so that i can put a cros_ec there

#

to force you to go through the full uacpi experience

fiery turtle Jan 6, 2025, 7:35 PM

#

im gonna take a look at that cros_ec, im curious how its out of spec

loud ice Jan 6, 2025, 7:35 PM

#

no clue

loud ice Jan 6, 2025, 7:35 PM

#

loud ice i kinda want to write a system emulator and if i do i will definitely run a risc...

like a second core on top of the main cpu

#

ofc

thorny orbit Jan 6, 2025, 7:35 PM

#

fiery turtle there's a difference between just receiving the event and actually handling it

uacpi is so cruel

loud ice Jan 6, 2025, 7:36 PM

#

and include an ACPI spec test in _PTS

#

go through every feature i can find in the spec before i allow you to sleep

#

genius

calm latch Jan 6, 2025, 7:41 PM

#

thorny orbit What do I need to implement in the kernel api for hooking power buttons ? (on sy...

Not having ECDT does not mean you don't have EC trl

fiery turtle Jan 6, 2025, 7:44 PM

#

loud ice genius

rm -rf home dir if the user doesnt do everything correctly

loud ice Jan 6, 2025, 7:44 PM

#

fiery turtle `rm -rf` home dir if the user doesnt do everything correctly

we don't need to be that dramatic

#

we can just make the user post a screenshot to discord

calm latch Jan 6, 2025, 7:44 PM

#

(ACPI cult)

loud ice Jan 6, 2025, 7:45 PM

#

showing their noncompliant impl

#

and shaming it

#

how about posting "L L L imagine using a BROKEN ACPI impl and thinking you could SLEEP!!! 🤣 🤣 🤣"

fiery turtle Jan 6, 2025, 7:46 PM

#

https://www.kernel.org/doc/Documentation/devicetree/bindings/mfd/cros-ec.txt

#

nvm cros_ec is not even your standard ec

#

its some weirdness

loud ice Jan 6, 2025, 7:47 PM

#

lol

#

yeah probably

thorny orbit Jan 6, 2025, 7:47 PM

#

calm latch (ACPI cult)

we need a sacrifice for the cult of ACPI

calm latch Jan 6, 2025, 7:48 PM

#

loud ice how about posting "L L L imagine using a *BROKEN* ACPI impl and thinking you cou...

uACPI has the power to shut the computer down

#

Just no operating system for you if you're not compiant

#

(I can't type)

fiery turtle Jan 6, 2025, 7:49 PM

#

loud ice yeah probably

although your chromebook dump seems to have a standard embedded controller device

#

maybe cros ec is like an extension idk

loud ice Jan 6, 2025, 7:49 PM

#

fiery turtle although your chromebook dump seems to have a standard embedded controller devic...

so there are TWO ECs?

fiery turtle Jan 6, 2025, 7:50 PM

#

if i had to guess

#

there's a standard two-port interface for reading ec ram

#

well and writing it

loud ice Jan 6, 2025, 7:50 PM

#

its probably not EC ram

fiery turtle Jan 6, 2025, 7:50 PM

#

and then the extension on top

#

which is chromebook specific stuff

loud ice Jan 6, 2025, 7:50 PM

#

cros_ec does security stuff

fiery turtle Jan 6, 2025, 7:50 PM

#

loud ice its probably not EC ram

whatever, abstract r/w interface over that address space

#

OperationRegion (ERAM, EmbeddedControl, Zero, 0x20)
Field (ERAM, ByteAcc, Lock, Preserve)
{
    RAMV,   8, 
    TSTB,   8, 
    TSTC,   8, 
    KBLV,   8, 
    FAND,   8, 
    PATI,   8, 
    PATT,   8, 
    PATC,   8, 
    CHGL,   8, 
    TBMD,   1, 
    DDPN,   3, 
    Offset (0x0A), 
    DFUD,   1, 
    FLSH,   1, 
    PFAN,   1, 
    KBLE,   1, 
    LTBR,   1, 
    LEDC,   1, 
    MTNS,   1, 
    KEYB,   1, 
    PSTR,   1, 
    P80P,   1, 
    THRM,   1, 
    SBKL,   1, 
    WIFI,   1, 
    HOST,   1, 
    GPIO,   1, 
    I2CB,   1, 
    CHRG,   1, 
    BATT,   1, 
    SBAT,   1, 
    HANG,   1, 
    PMUI,   1, 
    DSEC,   1, 
    UPDC,   1, 
    UMUX,   1, 
    MSFF,   1, 
    TVST,   1, 
    TCMV,   1, 
    RTCD,   1, 
    FPRD,   1, 
    TPAD,   1, 
    RWSG,   1, 
    DEVE,   1, 
    Offset (0x12), 
    BTID,   8, 
    USPP,   8
}

thorny orbit Jan 6, 2025, 7:51 PM

#

calm latch uACPI has the power to shut the computer down

https://tenor.com/view/thanos-smile-smiling-relax-chilling-gif-16321442

Tenor

fiery turtle Jan 6, 2025, 7:51 PM

#

fiery turtle ``` OperationRegion (ERAM, EmbeddedControl, Zero, 0x20) Field (ERAM, ByteAcc, Lo...

these are the fields it seems to have

#

i wonder what the HANG field does lmao

#

its also funny that chromebook logs EC events to aml stdout

#

not even their internal debug ports, literal Debug =

loud ice Jan 6, 2025, 7:56 PM

#

lol

fiery turtle Jan 6, 2025, 7:56 PM

#

#

also why the f does it have device tree properties in aml objects

loud ice Jan 6, 2025, 7:56 PM

#

lmfao

#

"compatible" "google,cr50"

fiery turtle Jan 6, 2025, 7:57 PM

#

#

linux,probed

#

huh 💀

loud ice Jan 6, 2025, 7:57 PM

#

i dont even

fiery turtle Jan 6, 2025, 7:58 PM

#

#

lmfao i cant

#

google merged aml and device trees ☠️

loud ice Jan 6, 2025, 7:58 PM

#

this isnt actually DT

#

also its a normal linux thing

#

you can search up this uuid in the kernel code

fiery turtle Jan 6, 2025, 7:59 PM

#

include/acpi/acuuid.h:#define UUID_DEVICE_PROPERTIES "daffd814-6eba-4d8c-8a91-bc9bbf4aa301"

loud ice Jan 6, 2025, 7:59 PM

#

grep for daffd814

calm latch Jan 6, 2025, 7:59 PM

#

Infy what do I need to do to support i2c devices? (like touch screens)

dusky glade Jan 6, 2025, 8:00 PM

#

fiery turtle `linux,probed`

probed?

#

sus

fiery turtle Jan 6, 2025, 8:00 PM

#

loud ice grep for daffd814

message above

fiery turtle Jan 6, 2025, 8:00 PM

#

calm latch Infy what do I need to do to support i2c devices? (like touch screens)

smbus driver

loud ice Jan 6, 2025, 8:00 PM

#

the funny thing is that the kernel driver only cares about the acpi id ```c
#ifdef CONFIG_ACPI
static const struct acpi_device_id cr50_i2c_acpi_id[] = {
{ "GOOG0005", 0 },
{}
};
MODULE_DEVICE_TABLE(acpi, cr50_i2c_acpi_id);
#endif

fiery turtle Jan 6, 2025, 8:00 PM

#

loud ice grep for daffd814

Documentation/arch/arm64/arm-acpi.rst explains this

loud ice Jan 6, 2025, 8:01 PM

#

and not the fake ass compatible

loud ice Jan 6, 2025, 8:01 PM

#

fiery turtle `Documentation/arch/arm64/arm-acpi.rst` explains this

ye i know

fiery turtle Jan 6, 2025, 8:01 PM

#

loud ice the funny thing is that the kernel driver only cares about the acpi id ```c #ifd...

this one is not even in the dump

loud ice Jan 6, 2025, 8:01 PM

#

wtf

fiery turtle Jan 6, 2025, 8:01 PM

#

theres GOOG{3,4,6,A,9999}

loud ice Jan 6, 2025, 8:02 PM

#

okay i give up

#

chromeos sucks

#

GOOG9999 is for ramoops

fiery turtle Jan 6, 2025, 8:02 PM

#

Device (PS2K)
{
    Name (_UID, Zero)  // _UID: Unique ID
    Name (_ADR, Zero)  // _ADR: Address
    Name (_HID, "GOOG000A")  // _HID: Hardware ID
    Name (_CID, Package (0x02)  // _CID: Compatible ID
    {
        EisaId ("PNP0303") /* IBM Enhanced Keyboard (101/102-key, PS/2 Mouse) */, 
        EisaId ("PNP030B")
    })

#

ps2 keyboard is GOOG000A lol

loud ice Jan 6, 2025, 8:02 PM

#

GOOG0006 is for tablet mode

#

4 is for power delivery

#

oh wait o

#

4 is EC

fiery turtle Jan 6, 2025, 8:03 PM

#

loud ice GOOG9999 is for ramoops

the what

dusky glade Jan 6, 2025, 8:03 PM

#

infy

loud ice Jan 6, 2025, 8:03 PM

#

fiery turtle the what

its where you write your crash log when you panic

#

its peristent ram

dusky glade Jan 6, 2025, 8:03 PM

#

why is it called prob

#

probe*

fiery turtle Jan 6, 2025, 8:03 PM

#

loud ice its where you write your crash log when you panic

ahhh

dusky glade Jan 6, 2025, 8:03 PM

#

why is probe mentioned

loud ice Jan 6, 2025, 8:03 PM

#

and 3 is USB-PD

dusky glade Jan 6, 2025, 8:03 PM

#

what firmware dev thought it was a good idea to mention probe

fiery turtle Jan 6, 2025, 8:03 PM

#

Device (CREC)
{
    Name (_HID, "GOOG0004")  // _HID: Hardware ID
    Name (_UID, One)  // _UID: Unique ID
    Name (_DDN, "EC Command Device")  // _DDN: DOS Device Name
    Name (_PRW, Package (0x02)  // _PRW: Power Resources for Wake
    {
        0x41, 
        0x05
    })
    Name (_CRS, ResourceTemplate ()  // _CRS: Current Resource Settings
    {
        Interrupt (ResourceConsumer, Level, ActiveLow, Exclusive, ,, )
        {
            0x0000005D,
        }
    })
    Method (_STA, 0, NotSerialized)  // _STA: Status
    {
        Return (0x0B)
    }
}

Okay that makes sense now

#

they have a custom ec device stub for this

loud ice Jan 6, 2025, 8:04 PM

#

how does that work?

#

i dont know anything about EC

fiery turtle Jan 6, 2025, 8:05 PM

#

loud ice how does that work?

they have a "normal" EC interface that would work with e.g. pmos ec driver

#

and then this thing on top

#

that is picked up by cros_ec driver

loud ice Jan 6, 2025, 8:05 PM

#

ah yeah

#

i do not know where the normal stuff is implemented though

fiery turtle Jan 6, 2025, 8:05 PM

#

drivers/platform/chrome/cros_ec_lpc.c this is their ec

#

normal ec is drivers/acpi/ec.c

calm latch Jan 6, 2025, 8:06 PM

#

Why do you even need EC extensions?

loud ice Jan 6, 2025, 8:06 PM

#

because google

fiery turtle Jan 6, 2025, 8:06 PM

#

// This driver uses the ChromeOS EC byte-level message-based protocol for
// communicating the keyboard state (which keys are pressed) from a keyboard EC
// to the AP over some bus (such as i2c, lpc, spi).  The EC does debouncing,
// but everything else (including deghosting) is done here.  The main
// motivation for this is to keep the EC firmware as simple as possible, since
// it cannot be easily upgraded and EC flash/IRAM space is relatively
// expensive.

loud ice Jan 6, 2025, 8:06 PM

#

also it does a ton of stuff

#

like security shit iirc

fiery turtle Jan 6, 2025, 8:06 PM

#

this just says "because its cheap to do in software"

calm latch Jan 6, 2025, 8:07 PM

#

So they're doing keyboard through EC?

fiery turtle Jan 6, 2025, 8:07 PM

#

yeah

loud ice Jan 6, 2025, 8:07 PM

#

ah the acpi shit is implemented here: https://source.chromium.org/chromiumos/chromiumos/codesearch/+/main:src/platform/ec/common/acpi.c;l=137

calm latch Jan 6, 2025, 8:07 PM

#

fiery turtle this just says "because its cheap to do in software"

I mean AMD laptops do debouncing through the GPIO meme

fiery turtle Jan 6, 2025, 8:07 PM

#

for the power button yeah

fiery turtle Jan 6, 2025, 8:08 PM

#

loud ice ah the acpi shit is implemented here: https://source.chromium.org/chromiumos/chr...

oh this is like their ec firmware?

loud ice Jan 6, 2025, 8:08 PM

#

yes

#

cros_ec is a software thing

dusky glade Jan 6, 2025, 8:08 PM

#

fiery turtle

infy if u were a firmware dev would u remove the word probe like a sane person

#

or are u an alien whos gonna probe

fiery turtle Jan 6, 2025, 8:09 PM

#

why are u so triggered by this word

dusky glade Jan 6, 2025, 8:09 PM

#

fiery turtle why are u so triggered by this word

probe

#

sus

loud ice Jan 6, 2025, 8:09 PM

#

what

#

its what you call device detection

dusky glade Jan 6, 2025, 8:09 PM

#

its what aliens do

kind mantle Jan 6, 2025, 8:10 PM

#

https://tenor.com/bQD4H.gif

Tenor

fiery turtle Jan 6, 2025, 8:10 PM

#

loud ice cros_ec is a software thing

oh yeah

loud ice Jan 6, 2025, 8:10 PM

#

they also do PS/2 emulation on the EC afaict

fiery turtle Jan 6, 2025, 8:10 PM

#

so they do read from the ec ram

dusky glade Jan 6, 2025, 8:10 PM

#

kind mantle https://tenor.com/bQD4H.gif

shut up im a teenager

#

😡

fiery turtle Jan 6, 2025, 8:10 PM

#

its literally some mmaped range they have

loud ice Jan 6, 2025, 8:10 PM

#

yeah idk how it works tbh

kind mantle Jan 6, 2025, 8:10 PM

#

dusky glade shut up im a teenager

That makes the meme even more appropriate because it was created when I was a teenager

fiery turtle Jan 6, 2025, 8:11 PM

#

apparently its so expensive to read they cache it

loud ice Jan 6, 2025, 8:11 PM

#

or it could be expensive

fiery turtle Jan 6, 2025, 8:11 PM

#

but yeah pretty interesting to see the software side of EC

#

or rather, the firmware side

loud ice Jan 6, 2025, 8:12 PM

#

yeah thats also cool

fiery turtle Jan 6, 2025, 8:12 PM

#

its close to what i would expect i guess

#

so they do this acpi compat layer and then whatever their platform needs, makes sense

fiery turtle Jan 6, 2025, 8:46 PM

#

#

ACPICA maintainer lore update

hollow elm Jan 6, 2025, 8:50 PM

#

lmao

#

switching companies from intel to arm

fiery turtle Jan 6, 2025, 8:51 PM

#

he also starred the device tree compiler yesterday

#

soo

#

LULW

mortal yoke Jan 6, 2025, 9:13 PM

#

btw how does uacpi pass some of the blob tests like the dell xps one, from what I can see it declares a field in ssdt6 but then it uses it in the top level of ssdt1?

fiery turtle Jan 6, 2025, 9:14 PM

#

can u give some specific example?

#

i can double check

mortal yoke Jan 6, 2025, 9:14 PM

#

OperationRegion (OPG0, SystemMemory, (XBAS + 0x8000), 0x1000)

fiery turtle Jan 6, 2025, 9:14 PM

#

no like

#

the blob

mortal yoke Jan 6, 2025, 9:15 PM

#

https://github.com/UltraOS/ACPIDumps/blob/master/Desktop/Dell/XPS/XPS 8940/9B79A7439313

#

and there that op region line it uses XBAS which is defined in a later ssdt from what I can see

fiery turtle Jan 6, 2025, 9:18 PM

#

[uACPI][ERROR] unable to lookup named object '\_SB_.CPID': not found
[uACPI][ERROR] aborting table load due to previous error: AML referenced an undefined object
[uACPI][ERROR]     #0 in \()
[uACPI][WARN] failed to load 'SSDT' (OEM ID 'Intel ' OEM Table ID 'PegSsdt '): AML referenced an undefined object
[uACPI][WARN] unusable/empty operation region IO_D
[uACPI][WARN] unusable/empty operation region IO_P
[uACPI][INFO] loaded 16 AML blobs in 121ms (1 error)
[uACPI][INFO] namespace initialization done in 9ms: 256 devices, 1 thermal zones

#

this is what i get

fiery turtle Jan 6, 2025, 9:20 PM

#

mortal yoke and there that op region line it uses XBAS which is defined in a later ssdt from...

yeah looks like a firmware bug

#

do u abort like literally everything if one load fails?

mortal yoke Jan 6, 2025, 9:22 PM

#

yes, maybe that's why

fiery turtle Jan 6, 2025, 9:22 PM

#

i just dont do that yeah (nor does acpica)

#

iirc

mortal yoke Jan 6, 2025, 9:24 PM

#

ig Ill just make the load method ignore the errors too

fiery turtle Jan 6, 2025, 9:24 PM

#

maybe for dsdt it matters more

#

but i dont abort even for dsdts

mortal yoke Jan 6, 2025, 9:30 PM

#

ig the remaining things are like 10 blobs that are somehow getting timed out even after the timeout stuff that I added, Ill have to take a look at those at some point

fiery turtle Jan 6, 2025, 9:31 PM

#

nice

#

do u handle nested whiles correctly?

#

like is the timeout value saved in this case

mortal yoke Jan 6, 2025, 9:33 PM

#

I am pretty sure they do work, ig I can test again to make sure

#

though tbh its likely not that important as on real hw there probably aren't that many hanging while loops hopefully

#

more important is the fact that you currently need to disable interrupts when installing a gpe handler to avoid races KEKW

#

also I think I can implement loadtable + overrides + load prevention tomorrow as I got some kind of table api done

fiery turtle Jan 6, 2025, 9:35 PM

#

mortal yoke though tbh its likely not that important as on real hw there probably aren't tha...

who knows LULW

fiery turtle Jan 6, 2025, 9:35 PM

#

mortal yoke more important is the fact that you currently need to disable interrupts when in...

wait why

mortal yoke Jan 6, 2025, 9:37 PM

#

fiery turtle wait why

well my theory is that you can get a sci in the middle of the enable function when it modifies the handlers array and then it gets screwed up

#

I haven't verified that but considering it works with an irq guard I think its something like that

fiery turtle Jan 6, 2025, 9:38 PM

#

mortal yoke well my theory is that you can get a sci in the middle of the enable function wh...

sure u can but why cant u just do

handlers[gpe_idx] = handler;
enable_gpe(gpe_idx);

mortal yoke Jan 6, 2025, 9:39 PM

#

yes but handlers is not a constant size array so it may get resized

fiery turtle Jan 6, 2025, 9:39 PM

#

also your protection only works if SCI is configured to fire at the current cpu where the code is being executed

winter orbit Jan 6, 2025, 9:39 PM

#

fiery turtle ACPICA maintainer lore update

pretty sure Intel just fired a bunch more people like 8 hours ago

fiery turtle Jan 6, 2025, 9:40 PM

#

mortal yoke yes but handlers is not a constant size array so it may get resized

💀

fiery turtle Jan 6, 2025, 9:40 PM

#

winter orbit pretty sure Intel just fired a bunch more people like 8 hours ago

ah damn

#

i guess acpica is unmaintained now

#

if thats true

north holly Jan 6, 2025, 9:40 PM

#

Another one bites the dust

#

Linux will start using uACPI

winter orbit Jan 6, 2025, 9:40 PM

#

fiery turtle i guess acpica is unmaintained now

We’ll see I guess

north holly Jan 6, 2025, 9:40 PM

#

north holly Linux will start using uACPI

100% yes

winter orbit Jan 6, 2025, 9:41 PM

#

north holly Linux will start using uACPI

no because that would “just work” and Linux doesn’t like that

#

trl

calm latch Jan 6, 2025, 9:41 PM

#

I need to upload my dumps...

#

My laptop is also slighly fucked

north holly Jan 6, 2025, 9:41 PM

#

Also doesn't Linux enjoy taking in the entire source of the thing into the Linux repo

#

Instead of a submodule

winter orbit Jan 6, 2025, 9:42 PM

#

ROS does that too

#

We have 15 years of like scripts that break with sub modules

calm latch Jan 6, 2025, 9:42 PM

#

mortal yoke though tbh its likely not that important as on real hw there probably aren't tha...

some galaxybrain dev using mutex timeouts as sleep(): trl

#

Oh I misread the message

#

anyway

loud ice Jan 6, 2025, 9:50 PM

#

winter orbit We have 15 years of like scripts that break with sub modules

you can rewrite them though lol

winter orbit Jan 6, 2025, 9:50 PM

#

i think so

#

Probably won’t happen anytime soon though

fiery turtle Jan 6, 2025, 9:59 PM

#

north holly Also doesn't Linux enjoy taking in the entire source of the thing into the Linux...

thats because they do private patches

#

and also convert acpica source style to linux style

north holly Jan 6, 2025, 10:00 PM

#

Would they do the same to uacpi

fiery turtle Jan 6, 2025, 10:01 PM

#

uacpi is already linux style

loud ice Jan 6, 2025, 10:08 PM

#

fiery turtle and also convert acpica source style to linux style

how the fuck does that work lol

fiery turtle Jan 6, 2025, 10:10 PM

#

loud ice how the fuck does that work lol

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

GitHub

acpica/generate/linux/libacpica.sh at master · acpica/acpica

The ACPI Component Architecture (ACPICA) project provides an open-source operating system-independent implementation of the Advanced Configuration and Power Interface specification (ACPI). For det...

loud ice Jan 6, 2025, 10:10 PM

#

just no

fiery turtle Jan 6, 2025, 10:11 PM

#

wait

#

they have an entire C program for that

#

https://github.com/acpica/acpica/tree/master/source/tools/acpisrc

GitHub

acpica/source/tools/acpisrc at master · acpica/acpica

The ACPI Component Architecture (ACPICA) project provides an open-source operating system-independent implementation of the Advanced Configuration and Power Interface specification (ACPI). For det...

calm latch Jan 6, 2025, 10:11 PM

#

that's sad

loud ice Jan 6, 2025, 10:12 PM

#

yeah tbh

fiery turtle Jan 6, 2025, 10:12 PM

#

its like 5k loc total 💀

winter orbit Jan 6, 2025, 10:12 PM

#

Incredible

#

that’s the most Linux thing I’ve seen all week

calm latch Jan 6, 2025, 10:14 PM

#

no wonder ACPICA has no maintainers

#

trl

loud ice Jan 6, 2025, 10:14 PM

#

yeah lol

fiery turtle Jan 6, 2025, 10:14 PM

#

do they literally have an entire handrolled c parser there bruh

winter orbit Jan 6, 2025, 10:14 PM

#

Linux would rather write a program to change source style than fix actual problems

fiery turtle Jan 6, 2025, 10:14 PM

#

its intels money sooo who cares LULW

loud ice Jan 6, 2025, 10:15 PM

#

lmao fair

calm latch Jan 6, 2025, 10:17 PM

#

did you choose Linux style so uACPI doesn't suffer the same faith?

fiery turtle Jan 6, 2025, 10:18 PM

#

nah its just what im used to lol

winter orbit Jan 6, 2025, 10:18 PM

#

uACPI designed for support in every OS from the get go

calm latch Jan 6, 2025, 10:18 PM

#

wasn't that the gist of ACPICA?

fiery turtle Jan 6, 2025, 10:18 PM

#

script to convert to reactos style when ultrameme

winter orbit Jan 6, 2025, 10:18 PM

#

ACPI spec is just replaced by uACPI source code in 2026

loud ice Jan 6, 2025, 10:19 PM

#

i mean

winter orbit Jan 6, 2025, 10:19 PM

#

fiery turtle script to convert to reactos style when <:ultrameme:729026898666979428>

You can’t write a script to convert to a buggy mess

loud ice Jan 6, 2025, 10:19 PM

#

right now the acpi spec is whatever the fuck microsoft developers did

fiery turtle Jan 6, 2025, 10:19 PM

#

i wish it was that

loud ice Jan 6, 2025, 10:19 PM

#

winter orbit You can’t write a script to convert to a buggy mess

yes you can, you can put __builtin_unreachable(); in a random selection of 30% of functions

fiery turtle Jan 6, 2025, 10:19 PM

#

right now the spec is full of lies

loud ice Jan 6, 2025, 10:19 PM

#

fiery turtle right now the spec is full of lies

thats not the spec

#

it calls itself a spec

winter orbit Jan 6, 2025, 10:19 PM

#

yeah ACPI.sys ignores 30% of the ACPI spec

loud ice Jan 6, 2025, 10:19 PM

#

but the spec is ACPI.sys

calm latch Jan 6, 2025, 10:19 PM

#

"hardware reduced platforms" KEKW

fiery turtle Jan 6, 2025, 10:20 PM

#

true

winter orbit Jan 6, 2025, 10:20 PM

#

which is funny because MS didn’t write ACPI.sys

#

Technically

loud ice Jan 6, 2025, 10:20 PM

#

wait really?

#

wtf

winter orbit Jan 6, 2025, 10:20 PM

#

It’s contracted work

loud ice Jan 6, 2025, 10:20 PM

#

oh god

fiery turtle Jan 6, 2025, 10:20 PM

#

is that why its so bad

loud ice Jan 6, 2025, 10:20 PM

#

that might explain it

#

the NT people are usually pretty good about implementation

winter orbit Jan 6, 2025, 10:20 PM

#

I mean it’s why there’s 90 interfaces

calm latch Jan 6, 2025, 10:20 PM

#

acpi spec = subcontracted developers, probably in India

winter orbit Jan 6, 2025, 10:20 PM

#

uh no

#

It’s people from Phoenix, INSYDE or AMI

#

Alternating

fiery turtle Jan 6, 2025, 10:21 PM

#

acpi spec is uefi board

#

which is all major tech companies

hollow elm Jan 6, 2025, 10:21 PM

#

winter orbit It’s people from Phoenix, INSYDE or AMI

so brain damaged people meme

loud ice Jan 6, 2025, 10:21 PM

#

winter orbit It’s people from Phoenix, INSYDE or AMI

those are not known for good code

winter orbit Jan 6, 2025, 10:21 PM

#

Yeah

loud ice Jan 6, 2025, 10:21 PM

#

lmfao

calm latch Jan 6, 2025, 10:21 PM

#

firmware people writing windows code is even worse

winter orbit Jan 6, 2025, 10:21 PM

#

loud ice those are not known for good code

hey buddy

#

you’re right

#

But that’s mean

#

KEKW

#

I write code that’s like

#

Close to the spec 90% of the time

#

chad

fiery turtle Jan 6, 2025, 10:22 PM

#

@winter orbit do u fetch the aml blobs from some rom, or do u construct them at runtime?

winter orbit Jan 6, 2025, 10:22 PM

#

I fetch them from ROMs

#

I just know how to dump them out of the raw SPI partitions in most hardware

fiery turtle Jan 6, 2025, 10:23 PM

#

so they're mostly hardcoded by the mobo?

#

i guess bioses do patch them too depepnding on config

winter orbit Jan 6, 2025, 10:23 PM

#

pretty much every device has some kind of device specific change in ACPI

#

Like two devices on the assembly line won’t have an identical ACPI

fiery turtle Jan 6, 2025, 10:24 PM

#

why is that?

winter orbit Jan 6, 2025, 10:25 PM

#

so I can tell you why for thinkpads specifically

#

there’s a tool phoenix provides that is basically an all in one debugging / compiler service

It’s supports source debugging firmware etc super cool

Okay so when you’re compiling the entire bios with this tool you also flash it with the same tool

#

do you know how you generate project tags at runtime per commit sometimes

fiery turtle Jan 6, 2025, 10:26 PM

#

ye

winter orbit Jan 6, 2025, 10:26 PM

#

it’s like that, you press play on visual studios, and it will write the entire thing to device

#

Take that concept and scale it to automated deployment on an assembly line

loud ice Jan 6, 2025, 10:27 PM

#

winter orbit Close to the spec 90% of the time

thats not everyone

#

and the 10% is the problem

winter orbit Jan 6, 2025, 10:27 PM

#

You’re hoping every line uses the same code base 1:1 and even if that’s true there’s just going to be some autogenerated Stuff

#

That’s the tldr

winter orbit Jan 6, 2025, 10:28 PM

#

loud ice and the 10% is the problem

the real problem is that what’s shipped to vendors as a basis from Intel is already flawed

#

Each chipset has a package that gets integrated into edk2

#

It includes ACPI stuff too as a basis

loud ice Jan 6, 2025, 10:28 PM

#

tbh its every single step of the way that did a half assed job

winter orbit Jan 6, 2025, 10:28 PM

#

yup

loud ice Jan 6, 2025, 10:29 PM

#

ms did a half assed job and outsourced acpi.sys to firmware vendors

#

inte is doing a half assed job and wrote dodgy aml

#

then firmware devs do weird shit 10% of the time and making it even worse

winter orbit Jan 6, 2025, 10:30 PM

#

mhmm. But you can’t really fix it at any spot

#

because what happens when we try to make our bios conform

calm latch Jan 6, 2025, 10:30 PM

#

Put uACPI in Windows KEKW

winter orbit Jan 6, 2025, 10:30 PM

#

Okay then automated Linux and windows testing fails

#

I did

#

You missed the lore

#

chad

fiery turtle Jan 6, 2025, 10:30 PM

#

calm latch Put uACPI in Windows <:KEKW:913708372329644042>

u must have missed a video of windows 10 booting with uacpi

#

or whichever one that was

loud ice Jan 6, 2025, 10:30 PM

#

holy shit

calm latch Jan 6, 2025, 10:30 PM

#

I did see

kind mantle Jan 6, 2025, 10:31 PM

#

fiery turtle u must have missed a video of windows 10 booting with uacpi

how

calm latch Jan 6, 2025, 10:31 PM

#

calm latch Put uACPI in Windows <:KEKW:913708372329644042>

But as the solution I mean

loud ice Jan 6, 2025, 10:31 PM

#

its not complete though right?

winter orbit Jan 6, 2025, 10:31 PM

#

not even close

#

Rewriting it now to be clean

loud ice Jan 6, 2025, 10:31 PM

#

makes sense

winter orbit Jan 6, 2025, 10:31 PM

#

but it’ll boot on win11 when it’s done

loud ice Jan 6, 2025, 10:31 PM

#

winter orbit Okay then automated Linux and windows testing fails

why?

fiery turtle Jan 6, 2025, 10:31 PM

#

kind mantle how

https://www.youtube.com/shorts/gUEzkglsXn4

YouTube

The_DarkFire_

ReactOS ACPI / uACPI windows driver wakes up windows 10 for the fir...

▶ Play video

loud ice Jan 6, 2025, 10:31 PM

#

a conforming bios should work

winter orbit Jan 6, 2025, 10:31 PM

#

Because windows already does completely stupid crap

loud ice Jan 6, 2025, 10:31 PM

#

fucking hell

winter orbit Jan 6, 2025, 10:32 PM

#

ofc now ReactOS does too

#

technically

fiery turtle Jan 6, 2025, 10:32 PM

#

winter orbit Because windows already does completely stupid crap

like what

kind mantle Jan 6, 2025, 10:32 PM

#

fiery turtle https://www.youtube.com/shorts/gUEzkglsXn4

How does that work? Is this the ReactOS kernel or is there some shenanigans going on here?

fiery turtle Jan 6, 2025, 10:33 PM

#

just a uacpi-based driver exporting vtables expected by nt from acpi.sys

kind mantle Jan 6, 2025, 10:33 PM

#

Oh interesting

#

ACPI is not built directly into the windows kernel?

loud ice Jan 6, 2025, 10:33 PM

#

no its a driver

winter orbit Jan 6, 2025, 10:34 PM

#

I actually don’t know anything off hand for win10/11 pepethink now that I think about it

#

maybe I’ll find something

loud ice Jan 6, 2025, 10:34 PM

#

what about linux?

fiery turtle Jan 6, 2025, 10:34 PM

#

would be fun to know the details

winter orbit Jan 6, 2025, 10:34 PM

#

Linux is a disaster we’ve tried fighting with before

#

lying about the windows versions is a mess

fiery turtle Jan 6, 2025, 10:35 PM

#

lmfao

#

but firmware devs also did bogus shit when _OSI(Linux)

winter orbit Jan 6, 2025, 10:35 PM

#

actually I can go on about Linux for awhile but they’re refused to fix anything we’ve suggested

fiery turtle Jan 6, 2025, 10:35 PM

#

so it was disabled for a reason

winter orbit Jan 6, 2025, 10:35 PM

#

fiery turtle but firmware devs also did bogus shit when _OSI(Linux)

like just make it unbootable

#

KEKW

fiery turtle Jan 6, 2025, 10:35 PM

#

basically

loud ice Jan 6, 2025, 10:35 PM

#

genius

winter orbit Jan 6, 2025, 10:36 PM

#

yeah but I know Ami and phoenix have both sat down and tried to get Linux to establish dome basic rules

#

Or just try and work with us to fix OSI(Linux)

loud ice Jan 6, 2025, 10:36 PM

#

behind the scenes, or?

fiery turtle Jan 6, 2025, 10:36 PM

#

like what?

winter orbit Jan 6, 2025, 10:36 PM

#

yeah

#

It’s been basically a eff off

#

Though

#

AFAIK

fiery turtle Jan 6, 2025, 10:36 PM

#

yeah like

#

why not have a proper OSI (Linux 5) for example

#

with actual versions

winter orbit Jan 6, 2025, 10:37 PM

#

we want to

#

Linux doesn’t want to

fiery turtle Jan 6, 2025, 10:37 PM

#

brain damage

calm latch Jan 6, 2025, 10:37 PM

#

gnuslashtroll

winter orbit Jan 6, 2025, 10:37 PM

#

most of the UEFI board has been willing to dedicate effort to implementing Linux specific logic

#

But even now Linux sleep doesn’t even work

#

we wanted it too

loud ice Jan 6, 2025, 10:37 PM

#

what if you implement linux specific logic

calm latch Jan 6, 2025, 10:37 PM

#

my laptop doesn't sleep with linux

loud ice Jan 6, 2025, 10:38 PM

#

that performs an acpi.sys spec test

#

and if it fails then you crash the system

fiery turtle Jan 6, 2025, 10:38 PM

#

what was their argument exactly

loud ice Jan 6, 2025, 10:38 PM

#

that will make linux deal with ACPI correctly!

winter orbit Jan 6, 2025, 10:38 PM

#

“Linux is an open environment and aligning with a close interpretation of the spec is against our values!”

fiery turtle Jan 6, 2025, 10:38 PM

#

loud ice that performs an acpi.sys spec test

i know a trick or two for fellow firmware devs to detect acpica aml abi trl

loud ice Jan 6, 2025, 10:38 PM

#

fucking what

loud ice Jan 6, 2025, 10:39 PM

#

fiery turtle i know a trick or two for fellow firmware devs to detect acpica aml abi <:trl:82...

no thats not the point

winter orbit Jan 6, 2025, 10:39 PM

#

but this conversation has been happening for 5+ years

loud ice Jan 6, 2025, 10:39 PM

#

the point is you do a compliance test

#

if you fail the test, you break boot

winter orbit Jan 6, 2025, 10:39 PM

#

Way before I joined phoenix

loud ice Jan 6, 2025, 10:39 PM

#

this means to get booting you need to pass the test

fiery turtle Jan 6, 2025, 10:39 PM

#

btw I have suggested this in the past https://github.com/acpica/acpica/discussions/956

winter orbit Jan 6, 2025, 10:39 PM

#

I was in high school when Linux and phoenix/ami started beefing

fiery turtle Jan 6, 2025, 10:40 PM

#

well their loss

#

even worse linux desktop experience

winter orbit Jan 6, 2025, 10:40 PM

#

yeah but the issue is we’re getting paid to do automated testing against several distros

#

I actually don’t think I can tell you which ones iirc

#

pepethink

#

Now that I think about it

fiery turtle Jan 6, 2025, 10:41 PM

#

ubuntu

winter orbit Jan 6, 2025, 10:41 PM

#

but you can probably figure it out

fiery turtle Jan 6, 2025, 10:41 PM

#

yeah lmao

winter orbit Jan 6, 2025, 10:41 PM

#

anyway we’re doing this, and testing against latest windows

fiery turtle Jan 6, 2025, 10:41 PM

#

winter orbit yeah but the issue is we’re getting paid to do automated testing against several...

by the distros themselves?

loud ice Jan 6, 2025, 10:41 PM

#

ubuntu, redhat, suse

#

is my guess

#

i.e. the commercial ones

winter orbit Jan 6, 2025, 10:41 PM

#

fiery turtle by the distros themselves?

I don’t know

calm latch Jan 6, 2025, 10:41 PM

#

who cares about suse

fiery turtle Jan 6, 2025, 10:41 PM

#

@slim panther

winter orbit Jan 6, 2025, 10:42 PM

#

majority of the enterprise Linux community

#

but yeah

#

Who cares

#

trl

slim panther Jan 6, 2025, 10:42 PM

#

calm latch who cares about suse

😢

winter orbit Jan 6, 2025, 10:42 PM

#

Just the people with the money

slim panther Jan 6, 2025, 10:42 PM

#

i do

fiery turtle Jan 6, 2025, 10:42 PM

#

winter orbit anyway we’re doing this, and testing against latest windows

so u dont even test against older windows?

winter orbit Jan 6, 2025, 10:42 PM

#

lol no

fiery turtle Jan 6, 2025, 10:42 PM

#

bruh

calm latch Jan 6, 2025, 10:42 PM

#

But idk everyone releases things for RHEL and Ubuntu

loud ice Jan 6, 2025, 10:42 PM

#

fun

slim panther Jan 6, 2025, 10:42 PM

#

calm latch But idk everyone releases things for RHEL and Ubuntu

we have big companies and governments as clients

fiery turtle Jan 6, 2025, 10:42 PM

#

Older windows aml interpreters couldnt even handle basic shit

winter orbit Jan 6, 2025, 10:43 PM

#

Looking at you vista

slim panther Jan 6, 2025, 10:43 PM

#

rhel is dead

calm latch Jan 6, 2025, 10:43 PM

#

The university I study at uses suse

loud ice Jan 6, 2025, 10:43 PM

#

slim panther rhel is dead

is it

slim panther Jan 6, 2025, 10:43 PM

#

well, that's what our stats say at least

loud ice Jan 6, 2025, 10:43 PM

#

i see

#

i see

slim panther Jan 6, 2025, 10:43 PM

#

we took over support for some eol distro

#

i think it was centos

#

because rh refuses to provide service for it

#

that means i gotta deal with ancient software though

winter orbit Jan 6, 2025, 10:44 PM

#

the only thing I hate more than Linux is firmware devs

slim panther Jan 6, 2025, 10:44 PM

#

is windows firmware devs

winter orbit Jan 6, 2025, 10:44 PM

#

windows doesn’t have firmware devs

slim panther Jan 6, 2025, 10:45 PM

#

I'm retarded

winter orbit Jan 6, 2025, 10:45 PM

#

Even the surface bios was just made via contracting with phoenix and AMI

slim panther Jan 6, 2025, 10:45 PM

#

i read "linux firmware devs"

winter orbit Jan 6, 2025, 10:46 PM

#

it’s funny too because there’s not that many firmware devs in the world

#

We can all fit in a building

#

So Intel, Ami, phoenix, insyde, Samsung all trade people often

calm latch Jan 6, 2025, 10:46 PM

#

samsung makes their firmware?

winter orbit Jan 6, 2025, 10:47 PM

#

So does dell

#

Also dell is the most conformant since they control the whole process on their own devices

#

Go figure

#

dontforgive

fiery turtle Jan 6, 2025, 10:48 PM

#

winter orbit Even the surface bios was just made via contracting with phoenix and AMI

btw, since u mentioned surface If (_OSI ("Intel-Ice-Lake-Low-Power-S0-Idle-Always-Set-XTAL-Bypass"))

#

know anything about this? LULW

winter orbit Jan 6, 2025, 10:48 PM

#

oof

#

yeah

fiery turtle Jan 6, 2025, 10:48 PM

#

lmfao

#

what is this thing

winter orbit Jan 6, 2025, 10:49 PM

#

I can’t really tell you

fiery turtle Jan 6, 2025, 10:49 PM

#

sad

frank canopy Jan 6, 2025, 10:49 PM

#

wow

winter orbit Jan 6, 2025, 10:50 PM

#

Microsoft meme

fiery turtle Jan 6, 2025, 10:51 PM

#

i see XTAL is a field meme

#

could probably reverse what this guards

winter orbit Jan 6, 2025, 10:51 PM

#

You can

#

You’ll be very disappointed though

fiery turtle Jan 6, 2025, 10:52 PM

#

Damn

loud ice Jan 6, 2025, 10:55 PM

#

what is XTAL

kind mantle Jan 6, 2025, 10:55 PM

#

Crystal

fiery turtle Jan 6, 2025, 10:56 PM

#

loud ice what is XTAL

some global aml state

kind mantle Jan 6, 2025, 10:56 PM

#

Don't ask why that's abbreviated the way it is

fiery turtle Jan 6, 2025, 10:56 PM

#

a byte in ram

kind mantle Jan 6, 2025, 10:57 PM

#

In the hardware world, XTAL means crystal, which of course has to do with clocks.

#

Since we're talking about sleep states here, I think that's the more likely explanation: XTAL is an MMIO register.

fiery turtle Jan 6, 2025, 10:58 PM

#

its a boolean

#

Method (XCFG, 2, NotSerialized)
{
    If ((Arg1 == Zero))
    {
        Local0 = 0x016E3600
        Local1 = 0x005B8D80
        Local2 = 0x7D
        Local3 = 0x02
    }
    ElseIf ((Arg1 == One))
    {
        Local0 = 0x0249F000
        Local1 = 0x00493E00
        Local2 = 0x32
        Local3 = 0x04
    }

Depending on if its 1 or 0 some shit is configured with these values

kind mantle Jan 6, 2025, 10:59 PM

#

Ok, not literally an MMIO register.

#

Still looks like it affects some MMIO stuff though

fiery turtle Jan 6, 2025, 11:00 PM

#

Package (0x02)
{
    "intel-quirk-mask", 
    Zero
},

#

idk

loud ice Jan 6, 2025, 11:02 PM

#

can you send the whole acpi table

north holly Jan 6, 2025, 11:02 PM

#

winter orbit So does dell

Is Dell's firmware sane?

winter orbit Jan 6, 2025, 11:03 PM

#

everything after 2008 iirc is the date

Yeah it’s better than most

fiery turtle Jan 6, 2025, 11:03 PM

#

loud ice can you send the whole acpi table

https://github.com/UltraOS/ACPIDumps/blob/master/Tablet/Microsoft/Surface/Surface Laptop 3/4E426AB8062D

fiery turtle Jan 6, 2025, 11:03 PM

#

winter orbit everything after 2008 iirc is the date Yeah it’s better than most

why 2008

winter orbit Jan 6, 2025, 11:03 PM

#

AMI->in-house bios

hollow elm Jan 6, 2025, 11:03 PM

#

i assume it's just configuring some clock to remain on when going into a lower power state

kind mantle Jan 6, 2025, 11:04 PM

#

Nah. Too many bits changed for that. I think it's setting a frequency.

#

Local0 and Local1 are clearly decimal numbers

loud ice Jan 6, 2025, 11:05 PM

#

oh

kind mantle Jan 6, 2025, 11:05 PM

#

That looks like something related to clock speed to me.

loud ice Jan 6, 2025, 11:06 PM

#

    If ((Arg1 == Zero))
    {
        Local0 = 24M
        Local1 = 6M
        Local2 = 125
        Local3 = 2
    }
    ElseIf ((Arg1 == One))
    {
        Local0 = 38.4 M
        Local1 = 4.8 M
        Local2 = 50
        Local3 = 4
    }```

fiery turtle Jan 6, 2025, 11:06 PM

#

then these are stored in some package

#

it does describe clocks

loud ice Jan 6, 2025, 11:07 PM

#

XCFG is probably XTAL configure

fiery turtle Jan 6, 2025, 11:07 PM

#

loud ice Jan 6, 2025, 11:08 PM

#

oh yeah its our favorite "definitely not FDT i swear" aml meme

fiery turtle Jan 6, 2025, 11:08 PM

#

yeah

winter orbit Jan 6, 2025, 11:08 PM

#

I wonder if this is how MS feels when they watch me reverse their entire gpu architecture

loud ice Jan 6, 2025, 11:08 PM

#

so 38.4 M is the sdw ip clock

fiery turtle Jan 6, 2025, 11:09 PM

#

apparently

loud ice Jan 6, 2025, 11:09 PM

#

which is the clock for the Intel Soundwire Manager

fiery turtle Jan 6, 2025, 11:09 PM

#

winter orbit I wonder if this is how MS feels when they watch me reverse their entire gpu arc...

lol

fiery turtle Jan 6, 2025, 11:09 PM

#

loud ice which is the clock for the Intel Soundwire Manager

maybe that bypass thing makes firmware ignore this or smth

loud ice Jan 6, 2025, 11:10 PM

#

fiery turtle maybe that bypass thing makes firmware ignore this or smth

im assuming its windows only

#

which code path is the bypass one?

#

4.8 M is the clock for MIPI discovery and configuration

fiery turtle Jan 6, 2025, 11:11 PM

#

If (_OSI ("Intel-Ice-Lake-Low-Power-S0-Idle-Always-Set-XTAL-Bypass"))
{
    S0I3 = Zero
}

loud ice Jan 6, 2025, 11:11 PM

#

MIPI is display stuff

fiery turtle Jan 6, 2025, 11:11 PM

#

S0I3 stuff

loud ice Jan 6, 2025, 11:12 PM

#

SoundWire is a new interface ratified in 2015 by the MIPI Alliance. SoundWire is used for transporting data typically related to audio functions. SoundWire interface is optimized to integrate audio devices in mobile or mobile inspired systems.

#

ah okay no its not display stuff

#

its audio stuff???

kind mantle Jan 6, 2025, 11:12 PM

#

MIPI does all kinds of wacky standards

#

There's also stuff like I³C, which IIRC is also a MIPI standard

winter orbit Jan 6, 2025, 11:13 PM

#

kind mantle MIPI does all kinds of wacky standards

lol yeah

#

KEKW

loud ice Jan 6, 2025, 11:13 PM

#

the linux code is in drivers/soundwire

#

    /*
     * The hardware relies on an internal counter, typically 4kHz,
     * to generate the SoundWire SSP - which defines a 'safe'
     * synchronization point between commands and audio transport
     * and allows for multi link synchronization. The SYNCPRD value
     * is only dependent on the oscillator clock provided to
     * the IP, so adjust based on _DSD properties reported in DSDT
     * tables. The values reported are based on either 24MHz
     * (CNL/CML) or 38.4 MHz (ICL/TGL+). On MeteorLake additional
     * frequencies are available with the MLCS clock source selection.
     */

#

24M is the clock when icelake xtal bypass is on

north holly Jan 6, 2025, 11:14 PM

#

kind mantle There's also stuff like I³C, which IIRC is also a MIPI standard

i cubed c?

kind mantle Jan 6, 2025, 11:14 PM

#

Yup

#

It's supposed to be the I²C successor

calm latch Jan 6, 2025, 11:14 PM

#

bruh

kind mantle Jan 6, 2025, 11:15 PM

#

But way more proprietary 💀

fiery turtle Jan 6, 2025, 11:15 PM

#

loud ice 24M is the clock when icelake xtal bypass is on

interesting

kind mantle Jan 6, 2025, 11:15 PM

#

Just for the record, I don't have the MIPI I³C standards in a folder on my computer

winter orbit Jan 6, 2025, 11:16 PM

#

kind mantle But way more proprietary 💀

Not really

#

Only proprietary because it hasn’t been used that much yet

kind mantle Jan 6, 2025, 11:16 PM

#

The standard is not fully open

#

Whereas I²C basically is

winter orbit Jan 6, 2025, 11:16 PM

#

I guess

loud ice Jan 6, 2025, 11:17 PM

#

so my guess is that when the bypass is on, they save power by downclocking the soundwire controller under the spec

#

or something

fiery turtle Jan 6, 2025, 11:17 PM

#

@winter orbit did he reverse it correctly? LULW

winter orbit Jan 6, 2025, 11:18 PM

#

Idk did he

#

meme

calm latch Jan 6, 2025, 11:18 PM

#

sound chip not sleeping?

#

(I have no idea but it's fun to witness)

loud ice Jan 6, 2025, 11:21 PM

#

loud ice so my guess is that when the bypass is on, they save power by downclocking the s...

or alternatively, that the supposed crystal isnt a real crystal

#

and is derived by a clock divider or some such

#

and the clock its derived from slows down when in s0 idle

winter orbit Jan 6, 2025, 11:24 PM

#

Waiting for @fiery turtle firmware dev arc

#

drink

fiery turtle Jan 6, 2025, 11:25 PM

#

💀

winter orbit Jan 6, 2025, 11:25 PM

#

if you ever apply somewhere I’ll be your reference

#

chad

#

Oh Infy I remember him, back from the days before Reactos took over the world

fiery turtle Jan 6, 2025, 11:28 PM

#

lol

vast kestrel Jan 7, 2025, 8:08 AM

#

Join the dark side of breaking the spec

winter orbit Jan 7, 2025, 8:18 AM

#

yeah

#

there you go

fiery turtle Jan 7, 2025, 10:42 AM

#

the test zig file is lost behind all aml stuff LULW

#

ill get back to it hopefully lol

winter orbit Jan 7, 2025, 10:45 AM

#

I just got a score of 65k ops/s in ROS

#

I think it’s because of our mutex implementation

#

It seems our implementation is 650 times slower than windows

#

kongoudisgust

fiery turtle Jan 7, 2025, 10:46 AM

#

winter orbit I just got a score of 65k ops/s in ROS

💀

#

nyaux is faster than ros confirmed

winter orbit Jan 7, 2025, 10:47 AM

#

confirmed

fiery turtle Jan 7, 2025, 10:47 AM

#

is this tcg?

winter orbit Jan 7, 2025, 10:47 AM

#

no

fiery turtle Jan 7, 2025, 10:47 AM

#

bruh

winter orbit Jan 7, 2025, 10:47 AM

#

Just fuzzing around a little I think it’s mutexs

#

So I think uACPI found a bug in ROS we willsee

fiery turtle Jan 7, 2025, 10:48 AM

#

u can try stubbing them to a no-op and seeing if that changes anytrhing

winter orbit Jan 7, 2025, 10:48 AM

#

yeah it goes to like 1.1 million

fiery turtle Jan 7, 2025, 10:48 AM

#

ah

fiery turtle Jan 7, 2025, 10:48 AM

#

winter orbit So I think uACPI found a bug in ROS we willsee

what sort of bug?

winter orbit Jan 7, 2025, 10:48 AM

#

mutexs being 650 times slower than windows is a bug

fiery turtle Jan 7, 2025, 10:49 AM

#

yeah doesnt seem right

#

it should be one atomic op in case of uacpi init, there shouldnt be any contention

winter orbit Jan 7, 2025, 10:49 AM

#

coooooool KEKW

fiery turtle Jan 7, 2025, 10:49 AM

#

lmao

#

i assume you're confident that the timekeeping is accurate?

winter orbit Jan 7, 2025, 10:50 AM

#

Pretty confident

#

Here I posted my os layer I’ve been using without the PCI logic

#

Since that needs like.. 10k lines of changes

fiery turtle Jan 7, 2025, 10:51 AM

#

where?

winter orbit Jan 7, 2025, 10:51 AM

#

https://github.com/DarkFire01/reactos/tree/PR-uACPI-Source

GitHub

GitHub - DarkFire01/reactos at PR-uACPI-Source

A free Windows-compatible Operating System. Contribute to DarkFire01/reactos development by creating an account on GitHub.

#

mind you the rsdp, PCI stuff all need the private dispatch tables

#

Which need to be separate PRs I guess

fiery turtle Jan 7, 2025, 10:56 AM

#

Looks like it should work for the most part

#

is you mapiospace smart enough to know which is a pio address and which is a phys address?

winter orbit Jan 7, 2025, 10:57 AM

#

yeah I know it says Io but that’s not really what it does

fiery turtle Jan 7, 2025, 10:57 AM

#

does it do like if <= 0xFFFF: map_pio()?

winter orbit Jan 7, 2025, 10:57 AM

#

yeah

fiery turtle Jan 7, 2025, 10:57 AM

#

ah ok cool

#

I think req is a pointer not an integer lol

winter orbit Jan 7, 2025, 10:58 AM

#

OH WAIT

#

Good point

#

KEKW

#

know what who needs break points

#

Failure aren’t real

#

Firmware devs are known for good code

fiery turtle Jan 7, 2025, 10:59 AM

#

btw, this is how i do nanoseseconds_since_boot on windows (for my userspace test runner)

uacpi_u64 uacpi_kernel_get_nanoseconds_since_boot(void)
{
#ifdef _WIN32
    static LARGE_INTEGER frequency;
    LARGE_INTEGER counter;

    if (frequency.QuadPart == 0) {
        if (!QueryPerformanceFrequency(&frequency)) {
            puts("QueryPerformanceFrequency() returned an error");
            std::abort();
        }
    }

    if (!QueryPerformanceCounter(&counter)) {
        puts("QueryPerformanceCounter() returned an error");
        std::abort();
    }

    counter.QuadPart *= nanoseconds_per_second;
    return counter.QuadPart / frequency.QuadPart;

#

is this wrong?

#

beacuse i noticed u do it differently

winter orbit Jan 7, 2025, 10:59 AM

#

It’s actually the same exact thing

#

I just use kernel APIs

#

queryPeformanceCounter eventually just calls what I did there :)

fiery turtle Jan 7, 2025, 11:00 AM

#

ah ok

#

/* Thread ID must be non-zero */ is questionable, maybe that was an acpica constraint of some kind

winter orbit Jan 7, 2025, 11:01 AM

#

is it not the same constraints there

#

ACPICA takes another burn

fiery turtle Jan 7, 2025, 11:01 AM

#

uacpi would be mad only if u returned this: https://github.com/UltraOS/uACPI/blob/master/include/uacpi/platform/arch_helpers.h#L35

winter orbit Jan 7, 2025, 11:02 AM

#

https://github.com/reactos/reactos/blob/master/drivers/bus/acpi/osl.c#L217

GitHub

reactos/drivers/bus/acpi/osl.c at master · reactos/reactos

A free Windows-compatible Operating System. Contribute to reactos/reactos development by creating an account on GitHub.

winter orbit Jan 7, 2025, 11:02 AM

#

fiery turtle uacpi would be mad only if u returned this: https://github.com/UltraOS/uACPI/blo...

oh yeah

#

Cool

#

That won’t happen

#

KEKW

fiery turtle Jan 7, 2025, 11:03 AM

#

its configurable in acpica as well i think, but maybe they also check for 0 as an error idk

fiery turtle Jan 7, 2025, 11:04 AM

#

winter orbit https://github.com/DarkFire01/reactos/tree/PR-uACPI-Source

i like how it calls it Fast mutexes LULW

winter orbit Jan 7, 2025, 11:04 AM

#

They aren’t fast

#

nooo

fiery turtle Jan 7, 2025, 11:04 AM

#

petition to rename to ExceptionallySlowMutex LULW

winter orbit Jan 7, 2025, 11:04 AM

#

You see the Unsafe call

fiery turtle Jan 7, 2025, 11:04 AM

#

yeah

winter orbit Jan 7, 2025, 11:05 AM

#

You spend so much time doing stuff between mutex locks windows was switching threads

#

And it loses the IRQL

loud ice Jan 7, 2025, 11:05 AM

#

wha

#

how does that work

fiery turtle Jan 7, 2025, 11:05 AM

#

💀

winter orbit Jan 7, 2025, 11:05 AM

#

I shouldn’t actually be using a mutex for this

#

Basically

loud ice Jan 7, 2025, 11:05 AM

#

how can you even lose an irql?

winter orbit Jan 7, 2025, 11:05 AM

#

windows mutates need to be at APC

#

you don’t you just get dispatched

#

And that’s a higher priority

fiery turtle Jan 7, 2025, 11:06 AM

#

Im confused ngl lol

#

what does that mean

winter orbit Jan 7, 2025, 11:07 AM

#

I should be making it so the thread running uACPI doesn’t go back into the scheduler on stall

#

But because of what I wrote there it will hit a delay and then get set to idle

fiery turtle Jan 7, 2025, 11:07 AM

#

wdym by stall exactly

winter orbit Jan 7, 2025, 11:07 AM

#

any kind of wait

fiery turtle Jan 7, 2025, 11:08 AM

#

you arent allowed to sleep with a mutex held in ros?

winter orbit Jan 7, 2025, 11:08 AM

#

no you can

#

Sorry I’m explaining it bad

#

It’s 3am

fiery turtle Jan 7, 2025, 11:08 AM

#

lol np

#

uacpi does release the namespace write lock on SleepOp and AcquireOp etc btw

winter orbit Jan 7, 2025, 11:09 AM

#

yeah it’s written write for windows drivers

#

I just need to enter a critical section tldr

fiery turtle Jan 7, 2025, 11:10 AM

#

written write 💀

fiery turtle Jan 7, 2025, 11:10 AM

#

winter orbit I just need to enter a critical section tldr

ah

winter orbit Jan 7, 2025, 11:10 AM

#

So if actually gets back to an IRQL of APC or below for release

#

the unsafe call is a way to make it ignore this fact

#

If you wanna SPEEEED

fiery turtle Jan 7, 2025, 11:10 AM

#

Wellll about that LULW

winter orbit Jan 7, 2025, 11:10 AM

#

and don’t care if another processor takes the job

#

Windows scheduler is fully active by the time ACPI.sys is loaded

fiery turtle Jan 7, 2025, 11:11 AM

#

tbh maybe reactos loads this driver after everything is up and there's contention for the cpu, other osses init it befor smp even

#

that could explain the poor score

winter orbit Jan 7, 2025, 11:12 AM

#

Maybe but windows was much faster

fiery turtle Jan 7, 2025, 11:12 AM

#

i mean id hope so lol

#

but yeah not 20 times

winter orbit Jan 7, 2025, 11:13 AM

#

fiery turtle tbh maybe reactos loads this driver after everything is up and there's contentio...

To be fair I don’t think this is a bad thing tbh

fiery turtle Jan 7, 2025, 11:13 AM

#

yeah

winter orbit Jan 7, 2025, 11:14 AM

#

NTs main hybrid kernel trait is that it waits before it’s ready to load user mode before it loads any drivers

#

like the whole kernel starts up and is ready

#

Early Reactos was just a shel and the kernel

#

No drivers

fiery turtle Jan 7, 2025, 11:14 AM

#

or maybe i lied, i think i mixed up early uacpi init (aka just parsing the tables) and loading the namespace

#

i hope people do namespace_load after their scheduler is up

#

because that needs scheduler for aml

fiery turtle Jan 7, 2025, 11:15 AM

#

winter orbit NTs main hybrid kernel trait is that it waits before it’s ready to load user mod...

does that mean it has some builtin early acpi parser for smp startup etc

winter orbit Jan 7, 2025, 11:15 AM

#

yeah

fiery turtle Jan 7, 2025, 11:15 AM

#

and numa stuff

winter orbit Jan 7, 2025, 11:15 AM

#

yup

fiery turtle Jan 7, 2025, 11:16 AM

#

makes sense

winter orbit Jan 7, 2025, 11:16 AM

#

Actually the bootloader even has some ACPI parsing

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