Kernel ABI: Difference between revisions

From 3dbrew
Jump to navigation Jump to search
← Older edit
Neobrain (talk | contribs)
549 edits
No edit summary
m typo
 
(11 intermediate revisions by 5 users not shown)
Line 1: Line 1:
Inputs are read from registers starting from r0 and outputs are written back to the same registers (also starting with r0).
= Calling Convention =
Seems to be [https://github.com/ARM-software/abi-aa/blob/main/aapcs32/aapcs32.rst#the-base-procedure-call-standard AAPCS]-based (with modifications)


== Overview ==
=== Inputs ===
* <code>r0–r3</code> : Argument / Scratch registers (caller-saved), inherited from AAPCS. ''If an input is to be placed on the stack, it will instead use the next free register starting from <code>r0</code>.''
=== Outputs ===
* <code>r0-r1</code> : Result, inherited from AAPCS. ''If multiple outputs are returned (e.g., <code>ControlMemory</code>), they are placed in consecutive registers starting from <code>r0</code>.''
== Example ==
<code>Result ControlMemory(uintptr_t* out, uintptr_t addr0, uintptr_t addr1, size_t size, MemoryOperation operation, MemoryPermission permissions)</code>
=== Inputs ===
Following standard [https://github.com/ARM-software/abi-aa/blob/main/aapcs32/aapcs32.rst#the-base-procedure-call-standard AAPCS] register selection:
; <code>r1</code> -> addr0
; <code>r2</code> -> addr1
; <code>r3</code> -> size
As there are still more arguments, the next free registers are selected starting from <code>r0</code>
; <code>r0</code> -> operation
; <code>r4</code> -> permissions
=== Outputs ===
; <code>r0</code> -> Result
; <code>r1</code> -> uintptr_t out
== System calls ==
{| class="wikitable" border="1"
{| class="wikitable" border="1"
|-
|-
!  Id
!  ID
!  Name
!  Name
!  Inputs
!  Inputs
!  Outputs
!  Outputs
|-
| 0x01
| ControlMemory
|
<code>r0</code>: <code>[[Memory Management#enum_MemoryOperation|MemoryOperation]] operation</code><br>
<code>r1</code>: <code>u32/void* addr0</code><br>
<code>r2</code>: <code>u32/void* addr1</code><br>
<code>r3</code>: <code>u32 size</code><br>
<code>r4</code>: <code>[[Memory Management#enum_MemoryPermission|MemoryPermission]] permissions</code>
|
<code>r0</code>: <code>Result</code><br>
<code>r1</code>: <code>u32/void* addr_out</code>
|-
| 0x02
| QueryMemory
|
<code>r2</code>?: <code>u32/void* addr</code>
|
<code>r0</code>: <code>Result</code><br>
<code>r1</code>: <code>u32 base_process_virtual_address</code><br>
<code>r2</code>: <code>u32 size</code><br>
<code>r3</code>: <code>[[Memory Management#enum_MemoryPermission|MemoryPermission]] permission</code><br>
<code>r4</code>: <code>[[Memory_Management#enum_MemoryState|MemoryState]] state</code><br>
<code>r5</code>: <code>[[Memory_Management#enum_PageFlags|PageFlags]] page_flags</code>
|-
| 0x03
| ExitProcess
| None
| None, doesn't return
|-
| 0x04
| GetProcessAffinityMask
|
<code>r0</code>: <code>[out] u8* affinitymask</code><br>
<code>r1</code>: <code>Handle<[[KProcess]]> process</code><br>
<code>r2</code>: <code>s32 processorcount</code>
|
<code>r0</code>: <code>Result</code>
|-
| 0x05
| SetProcessAffinityMask
|
<code>r0</code>: <code>Handle<[[KProcess]]> process</code><br>
<code>r1</code>: <code>[in] const u8* affinitymask</code><br>
<code>r2</code>: <code>s32 processorcount</code>
|
<code>r0</code>: <code>Result</code>
|-
| 0x06
| GetProcessIdealProcessor
|
<code>r1</code>?: <code>Handle<[[KProcess]]> process</code>
|
<code>r0</code>: <code>Result</code><br>
<code>r1</code>: <code>s32 processorid</code><br>
<code>r2</code>: Clobbered?
|-
| 0x07
| SetProcessIdealProcessor
|
<code>r0</code>: <code>Handle<[[KProcess]]> process</code><br>
<code>r1</code>: <code>s32 processorid</code>
|
<code>r0</code>: <code>Result</code><br>
|-
| 0x08
| CreateThread
|
<code>r0</code>: <code>s32 thread_priority</code><br>
<code>r1</code>: <code>ThreadFunc entrypoint</code><br>
<code>r2</code>: <code>u32 arg</code><br>
<code>r3</code>: <code>u32/void* stack_top</code><br>
<code>r4</code>: <code>s32 processor_id</code>
|
<code>r0</code>: <code>Result</code><br>
<code>r1</code>: <code>Handle<[[KThread]]> thread_handle</code>
|-
| 0x09
| ExitThread
| None
| None, doesn't return
|-
| 0x0A
| SleepThread
|
<code>r0-r1</code>: <code>s64 nanoseconds</code>
| None
|-
| 0x0B
| GetThreadPriority
|
<code>r0</code>: Ignored?<br>
<code>r1</code>: <code>Handle thread_handle</code>
|
<code>r0</code>: <code>Result</code><br>
<code>r1</code>: <code>s32 thread_priority</code><br>
<code>r2</code>: Clobbered?
|-
| 0x0C
| SetThreadPriority
|
<code>r0</code>: <code>Handle<[[KThread]]> thread_handle</code><br>
<code>r1</code>: <code>s32 thread_priority</code>
|
<code>r0</code>: <code>Result</code>
|-
| 0x0D
| GetThreadAffinityMask
|
<code>r0</code>: <code>[out] u8* affinitymask</code><br>
<code>r1</code>: <code>Handle<[[KThread]]> thread_handle</code><br>
<code>r2</code>: <code>s32 processorcount</code>
|
<code>r0</code>: <code>Result</code>
|-
| 0x0E
| SetThreadAffinityMask
|
<code>r0</code>: <code>Handle<[[KThread]]> thread_handle</code><br>
<code>r1</code>: <code>[in] const u8* affinitymask</code><br>
<code>r2</code>: <code>s32 processorcount</code>
|
<code>r0</code>: <code>Result</code>
|-
| 0x0F
| GetThreadIdealProcessor
|
<code>r0</code>: Ignored?<br>
<code>r1</code>: <code>Handle<[[KThread]]> thread_handle</code>
|
<code>r0</code>: <code>Result</code><br>
<code>r1</code>: <code>s32 processorid</code>
|-
| 0x10
| SetThreadIdealProcessor
|
<code>r0</code>: <code>Handle<[[KThread]]> thread_handle</code><br>
<code>r1</code>: <code>s32 processorid</code>
|
<code>r0</code>: <code>Result</code>
|-
| 0x11
| GetProcessorID
|
None
|
<code>r0</code>: <code>s32 processorid</code>
|-
| 0x24
| WaitSynchronization1
|
<code>r0</code>: <code>Handle<[[KSynchronizationObject]]> handle</code><br>
<code>r2-r3</code>: <code>s64 timeout</code>
|
<code>r0</code>: <code>Result</code>
|-
|-
| 0x1B
| 0x1B
Line 12: Line 194:
| Timer handle, initial_low, interval_low, initial_high, interval_high
| Timer handle, initial_low, interval_low, initial_high, interval_high
| Result
| Result
|-
| 0x28
| GetSystemTick
| None
|
<code>r0</code>: <code>Low 32 bits of the tick count</code>
<br><code>r1</code>: <code>High 32 bits of the tick count</code>
|-
|-
| 0x2D
| 0x2D
| ConnectToPort
| ConnectToPort
| ??, pointer to port name
| ??<br><code>r1</code>: <code>pointer to port name</code>
| Result, handle to [[KClientSession]]
| <code>r0</code>: <code>Result</code><br><code>r1</code>: <code>handle to [[KClientSession]]</code>
|-
|-
| 0x32
| 0x32
| SendSyncRequest
| SendSyncRequest
| handle to [[KClientSession]]
| <code>r0</code>: <code>handle to [[KClientSession]]</code>
| Result
| <code>r0</code>: <code>Result<code>
|}
|}

Latest revision as of 21:08, 29 August 2025

Calling Convention

Seems to be AAPCS-based (with modifications)

Overview

Inputs

  • r0–r3 : Argument / Scratch registers (caller-saved), inherited from AAPCS. If an input is to be placed on the stack, it will instead use the next free register starting from r0.

Outputs

  • r0-r1 : Result, inherited from AAPCS. If multiple outputs are returned (e.g., ControlMemory), they are placed in consecutive registers starting from r0.

Example

Result ControlMemory(uintptr_t* out, uintptr_t addr0, uintptr_t addr1, size_t size, MemoryOperation operation, MemoryPermission permissions)

Inputs

Following standard AAPCS register selection:

r1 -> addr0
r2 -> addr1
r3 -> size

As there are still more arguments, the next free registers are selected starting from r0

r0 -> operation
r4 -> permissions

Outputs

r0 -> Result
r1 -> uintptr_t out

System calls

ID Name Inputs Outputs
0x01 ControlMemory

r0: MemoryOperation operation
r1: u32/void* addr0
r2: u32/void* addr1
r3: u32 size
r4: MemoryPermission permissions

r0: Result
r1: u32/void* addr_out

0x02 QueryMemory

r2?: u32/void* addr

r0: Result
r1: u32 base_process_virtual_address
r2: u32 size
r3: MemoryPermission permission
r4: MemoryState state
r5: PageFlags page_flags

0x03 ExitProcess None None, doesn't return
0x04 GetProcessAffinityMask

r0: [out] u8* affinitymask
r1: Handle<KProcess> process
r2: s32 processorcount

r0: Result

0x05 SetProcessAffinityMask

r0: Handle<KProcess> process
r1: [in] const u8* affinitymask
r2: s32 processorcount

r0: Result

0x06 GetProcessIdealProcessor

r1?: Handle<KProcess> process

r0: Result
r1: s32 processorid
r2: Clobbered?

0x07 SetProcessIdealProcessor

r0: Handle<KProcess> process
r1: s32 processorid

r0: Result

0x08 CreateThread

r0: s32 thread_priority
r1: ThreadFunc entrypoint
r2: u32 arg
r3: u32/void* stack_top
r4: s32 processor_id

r0: Result
r1: Handle<KThread> thread_handle

0x09 ExitThread None None, doesn't return
0x0A SleepThread

r0-r1: s64 nanoseconds

None
0x0B GetThreadPriority

r0: Ignored?
r1: Handle thread_handle

r0: Result
r1: s32 thread_priority
r2: Clobbered?

0x0C SetThreadPriority

r0: Handle<KThread> thread_handle
r1: s32 thread_priority

r0: Result

0x0D GetThreadAffinityMask

r0: [out] u8* affinitymask
r1: Handle<KThread> thread_handle
r2: s32 processorcount

r0: Result

0x0E SetThreadAffinityMask

r0: Handle<KThread> thread_handle
r1: [in] const u8* affinitymask
r2: s32 processorcount

r0: Result

0x0F GetThreadIdealProcessor

r0: Ignored?
r1: Handle<KThread> thread_handle

r0: Result
r1: s32 processorid

0x10 SetThreadIdealProcessor

r0: Handle<KThread> thread_handle
r1: s32 processorid

r0: Result

0x11 GetProcessorID

None

r0: s32 processorid

0x24 WaitSynchronization1

r0: Handle<KSynchronizationObject> handle
r2-r3: s64 timeout

r0: Result

0x1B SetTimer Timer handle, initial_low, interval_low, initial_high, interval_high Result
0x28 GetSystemTick None

r0: Low 32 bits of the tick count
r1: High 32 bits of the tick count

0x2D ConnectToPort ??
r1: pointer to port name 		r0: Result
r1: handle to KClientSession
0x32 SendSyncRequest r0: handle to KClientSession r0: Result
Retrieved from "https://www.3dbrew.org/w/index.php?title=Kernel_ABI&oldid=23691"