kernel: Implement code address offset for Skyline compatibility

Add support for the Skyline 32-bit modding framework by implementing
a code address offset and improving memory state verification.

Changes:
- Add CodeStartOffset constant (0x500000) applied to 32-bit code
  address space types (Is32Bit and Is32BitNoMap) in KProcess::LoadFromMetadata
- Add debug logging for 32-bit syscall argument tracking in
  SetProcessMemoryPermission, MapProcessCodeMemory, and
  UnmapProcessCodeMemory wrapper functions
- Replace CheckMemoryStateContiguous with CheckMemoryState in
  KPageTableBase::UnmapCodeMemory for more flexible memory state
  verification

This implementation enables compatibility with:
- Skyline 32-bit modding framework
- CTGP-DX (Mario Kart 8 Deluxe mod)
- Other homebrew using 32-bit Skyline

Based on similar fixes in Ryujinx (commit 5e9678c8fe) and Eden emulator.

Co-authored-by: JPikachu <jpikachu.eden@gmail.com>
Co-authored-by: JPikachu <jpikachu@eden-emu.dev>
Signed-off-by: Zephyron <zephyron@citron-emu.org>

src/core/hle/kernel/k_page_table_base.cpp

@@ -1,4 +1,5 @@
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 citron Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include "common/scope_exit.h"
@@ -1274,7 +1275,7 @@ Result KPageTableBase::UnmapCodeMemory(KProcessAddress dst_address, KProcessAddr
 
     // Verify that the destination memory is aliasable code.
     size_t num_dst_allocator_blocks;
-    R_TRY(this->CheckMemoryStateContiguous(
+    R_TRY(this->CheckMemoryState(
         std::addressof(num_dst_allocator_blocks), dst_address, size, KMemoryState::FlagCanCodeAlias,
         KMemoryState::FlagCanCodeAlias, KMemoryPermission::None, KMemoryPermission::None,
         KMemoryAttribute::All & ~KMemoryAttribute::PermissionLocked, KMemoryAttribute::None));

src/core/hle/kernel/k_process.cpp

@@ -1,4 +1,5 @@
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 citron Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <random>
@@ -25,6 +26,12 @@ namespace Kernel {
 
 namespace {
 
+// Code offset for 32-bit processes to ensure compatibility with Skyline modding framework.
+// Skyline assumes memory exists before the entry point. This matches the approach used for
+// 39-bit processes (which load at 0x8000'0000 instead of 0x800'0000 for the same reason).
+// This can only be removed if Skyline is updated to not depend on pre-entry-point memory.
+constexpr u64 CodeStartOffset = 0x500000UL;
+
 Result TerminateChildren(KernelCore& kernel, KProcess* process,
                          const KThread* thread_to_not_terminate) {
     // Request that all children threads terminate.
@@ -1195,11 +1202,11 @@ Result KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata, std:
         break;
     case FileSys::ProgramAddressSpaceType::Is32Bit:
         flag |= Svc::CreateProcessFlag::AddressSpace32Bit;
-        code_address = 0x20'0000;
+        code_address = 0x20'0000 + CodeStartOffset;
         break;
     case FileSys::ProgramAddressSpaceType::Is32BitNoMap:
         flag |= Svc::CreateProcessFlag::AddressSpace32BitWithoutAlias;
-        code_address = 0x20'0000;
+        code_address = 0x20'0000 + CodeStartOffset;
         break;
     }
 

src/core/hle/kernel/svc.cpp

@@ -1,4 +1,5 @@
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 citron Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 // This file is automatically generated using svc_generator.py.
@@ -1825,6 +1826,10 @@ static void SvcWrap_SetProcessMemoryPermission64From32(Core::System& system, std
     uint64_t size{};
     MemoryPermission perm{};
 
+    LOG_DEBUG(Kernel_SVC, "Raw args, [0]={:#x} [1]={:#x} [2]={:#x} [3]={:#x} [4]={:#x} [5]={:#x}",
+              GetArg32(args, 0), GetArg32(args, 1), GetArg32(args, 2),
+              GetArg32(args, 3), GetArg32(args, 4), GetArg32(args, 5));
+
     process_handle = Convert<Handle>(GetArg32(args, 0));
     std::array<uint32_t, 2> address_gather{};
     address_gather[0] = GetArg32(args, 2);
@@ -1912,6 +1917,10 @@ static void SvcWrap_MapProcessCodeMemory64From32(Core::System& system, std::span
     uint64_t src_address{};
     uint64_t size{};
 
+    LOG_DEBUG(Kernel_SVC, "Raw args, [0]={:#x} [1]={:#x} [2]={:#x} [3]={:#x} [4]={:#x} [5]={:#x} [6]={:#x}",
+              GetArg32(args, 0), GetArg32(args, 1), GetArg32(args, 2),
+              GetArg32(args, 3), GetArg32(args, 4), GetArg32(args, 5), GetArg32(args, 6));
+
     process_handle = Convert<Handle>(GetArg32(args, 0));
     std::array<uint32_t, 2> dst_address_gather{};
     dst_address_gather[0] = GetArg32(args, 2);
@@ -1939,6 +1948,10 @@ static void SvcWrap_UnmapProcessCodeMemory64From32(Core::System& system, std::sp
     uint64_t src_address{};
     uint64_t size{};
 
+    LOG_DEBUG(Kernel_SVC, "Raw args, [0]={:#x} [1]={:#x} [2]={:#x} [3]={:#x} [4]={:#x} [5]={:#x} [6]={:#x}",
+              GetArg32(args, 0), GetArg32(args, 1), GetArg32(args, 2),
+              GetArg32(args, 3), GetArg32(args, 4), GetArg32(args, 5), GetArg32(args, 6));
+
     process_handle = Convert<Handle>(GetArg32(args, 0));
     std::array<uint32_t, 2> dst_address_gather{};
     dst_address_gather[0] = GetArg32(args, 2);
@@ -1953,6 +1966,9 @@ static void SvcWrap_UnmapProcessCodeMemory64From32(Core::System& system, std::sp
     size_gather[1] = GetArg32(args, 6);
     size = Convert<uint64_t>(size_gather);
 
+    LOG_DEBUG(Kernel_SVC, "Reconstructed, handle={:#x} dst={:#x} src={:#x} size={:#x}",
+              process_handle, dst_address, src_address, size);
+
     ret = UnmapProcessCodeMemory64From32(system, process_handle, dst_address, src_address, size);
 
     SetArg32(args, 0, Convert<uint32_t>(ret));