LibACPI: Add enough stuff to load the Qemu DSDT (3/N) by cqundefine · Pull Request #27134 · SerenityOS/serenity

cqundefine · 2026-10-29T11:39:06Z

This adds enough features to allow our AML interpreter to load the standard DSDT provided but Qemu.

Some opcodes encode how many bytes a list of something takes, this will be needed very soon.

Sometimes reading an operand doesn't produce an operand, but a node or something else.

When an opcode would get filtered at root it would try to get the opcode value to print the error message, which is absent on the root frame.

The comment noted these should be added, but I accidentally left them out when refactoring. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

Those two always come in pairs so lets do it in one commit. This doesn't yet support evaluating fields as it's not yet needed and would require setting up callbacks from the interpreter. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

It was mistakenly added in the previous PR even though it never has any other argument than nullptr.

For now just create a node, but don't actually do anything with the body as it's not needed for now. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

Do this similarly to repeating over a byte count.

Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

While this opcode is deprecated, it is still widely used and we need to support it. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

Now that we have all the primitives we need, let's add the missing objects. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

Just as the previous test, the AML table was compiled with using IASL 20260930 without any flags.

spholz

Thanks!

spholz · 2026-10-30T10:55:22Z

+    PCC = 10,
+    PRM = 11,


nit: the spec also has these as full words. We usually prefer full words over abbreviations unless where the abbreviation is easier to understand and more common.

(I personally don't know what these stand for.)

spholz · 2026-10-30T10:59:37Z

+    PCC = 10,
+    PRM = 11,


very nit: Why didn't you keep these as hex? The spec has them as hex.

spholz · 2026-10-30T11:02:22Z

    // opcodes here, I'm adding it in here.
-    // DeviceOp is also missing. Adding it here too.
-    // MutexOp likewise missing. It's named and creates an object, so here it be.
+    // DeviceOp, MethodOp and MutexOp are also missing. Adding it here too beacuse they also create an object.


nit: s/it/them, s/beacuse/because

spholz · 2026-10-30T11:03:55Z

+    OperandType::ByteData,
+    OperandType::TermArg,
+    OperandType::TermArg,
+    OperandType::End


nit: please add trailing commas for every multi-line initializer like this.

spholz · 2026-10-30T11:12:21Z

+
+protected:
+    u8 m_space_raw;
+    i64 m_offset;


You are using i64 here now, but u64 everywhere else for ACPI integers. Are ACPI integers signed or unsigned? Or does that depend on the usage of them?

spholz · 2026-10-30T14:29:57Z

        auto value = MUST(m_data.as<String>());
        TRY(builder.try_appendff("('{}')"sv, StringView(value.buffer())));
+    } else if (m_data.is<Buffer>()) {
+        TRY(builder.try_append("Buffer"sv));


Should we print the buffer contents (with :hex-dump maybe)?

spholz · 2026-10-30T14:39:18Z

    } else if (m_data.is<Buffer>()) {
        TRY(builder.try_append("Buffer"sv));
+    } else if (m_data.is<Package>()) {
+        TRY(builder.try_append("Package"sv));


Might be nice to print the contents here. Easiest way to do so would be to add a formatter for Object. Then the Vector formatter should just work.

spholz · 2026-10-30T14:47:39Z

+        DEBUG_LOG("Reading package element...");
+
+        auto initial = TRY(stream.peek_byte());
+        if (is_namestring_start(initial)) {


Shouldn't we print an interpreter error for types that we don't support yet here? Or do I not understand how this works?

I'm also confused by the "// NOTE: Reading NameString will be handled a different way." comment. In what different way?

spholz · 2026-10-30T14:51:41Z

+    StringView kind() const override { return "Processor"sv; }
+
+protected:
+    u32 m_address { 0 };


The ACPI 6.3 spec seems to call this the address of the "processor register block". And m_block_length the length of it.

spholz · 2026-10-30T14:57:19Z

+{
+    auto interpreter = TRY_OR_FAIL(load_test_table("data_objects_and_named_objects.aml"sv));
+
+    auto string = TRY_OR_FAIL(interpreter->evaluate(TRY_OR_FAIL(ACPI::AML::NameString::from_string_view("\\STR0"sv))));


Seeing these ACPI::AML::NameString::from_string_view everywhere makes me feel like we should add a StringView overload to evaluate maybe?

cqundefine and others added 18 commits October 29, 2026 08:14

LibACPI: Allow repeating an operand over a byte count

4d167f7

Some opcodes encode how many bytes a list of something takes, this will be needed very soon.

LibACPI: Add missing GenericAddressStructure spec links

6ccdd7c

LibACPI: Add missing AddressSpace enum values

8df40a8

LibACPI: Add a finish_operand variant without an actual operand

723dcfa

Sometimes reading an operand doesn't produce an operand, but a node or something else.

LibACPI: Fix a crash on filtered opcode at root

9af18fc

When an opcode would get filtered at root it would try to get the opcode value to print the error message, which is absent on the root frame.

LibACPI: Add missing opcodes to named_object_filter

ac54964

The comment noted these should be added, but I accidentally left them out when refactoring. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

LibACPI: Remove the Node constructor

b0ba1dc

It was mistakenly added in the previous PR even though it never has any other argument than nullptr.

LibACPI: Add support for reading methods

bdfb286

For now just create a node, but don't actually do anything with the body as it's not needed for now. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

LibACPI: Add support for reading DeviceOp

5c6548d

Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

LibACPI: Add support for strings

38340b3

Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

LibACPI: Add support for buffers

99ca413

Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

LibACPI: Add support for reading MutexOp

ed1d3d0

Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

LibACPI: Allow repeating an operand over a count

04b3775

Do this similarly to repeating over a byte count.

LibACPI: Add support for packages

1975fb6

Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

LibACPI: Add support for reading ProcessorOp

248c8b4

While this opcode is deprecated, it is still widely used and we need to support it. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

LibACPI: Add missing predefined namespace objects

7a6d373

Now that we have all the primitives we need, let's add the missing objects. Co-authored-by: RatcheT2497 <ratchetnumbers@proton.me>

Tests: Add some tests for new LibACPI features

3a8ef00

Just as the previous test, the AML table was compiled with using IASL 20260930 without any flags.

cqundefine changed the title ~~LibACPI: Add enough stuff to load the Qemu DSDT~~ LibACPI: Add enough stuff to load the Qemu DSDT (3/N) Oct 29, 2026

github-actions Bot added the 👀 pr-needs-review PR needs review from a maintainer or community member label Oct 29, 2026

spholz reviewed Oct 30, 2026

View reviewed changes

Navigation Menu

Repository navigation

LibACPI: Add enough stuff to load the Qemu DSDT (3/N) - #27134

LibACPI: Add enough stuff to load the Qemu DSDT (3/N)#27134
cqundefine wants to merge 18 commits into
SerenityOS:masterfrom
cqundefine:acpi3

cqundefine commented Oct 29, 2026

Uh oh!

spholz left a comment

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

spholz Oct 30, 2026

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

cqundefine commented Oct 29, 2026

Uh oh!

spholz left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants