diff --git a/src/xenia/base/string.cc b/src/xenia/base/string.cc
index 068c1b6ac..acd00ce76 100644
--- a/src/xenia/base/string.cc
+++ b/src/xenia/base/string.cc
@@ -68,6 +68,20 @@ std::vector<std::string> split_path(const std::string& path) {
   return parts;
 }
 
+std::string join_paths(const std::string& left, const std::string& right,
+                       char sep) {
+  if (!left.size()) {
+    return right;
+  } else if (!right.size()) {
+    return left;
+  }
+  if (left[left.size() - 1] == sep) {
+    return left + right;
+  } else {
+    return left + sep + right;
+  }
+}
+
 std::wstring join_paths(const std::wstring& left, const std::wstring& right,
                         wchar_t sep) {
   if (!left.size()) {
@@ -166,4 +180,20 @@ std::wstring find_name_from_path(const std::wstring& path) {
   return name;
 }
 
+std::string find_base_path(const std::string& path) {
+  auto last_slash = path.find_last_of('\\');
+  if (last_slash == std::string::npos) {
+    return path;
+  } else if (last_slash == path.length() - 1) {
+    auto prev_slash = path.find_last_of('\\', last_slash - 1);
+    if (prev_slash == std::string::npos) {
+      return "";
+    } else {
+      return path.substr(0, prev_slash + 1);
+    }
+  } else {
+    return path.substr(0, last_slash + 1);
+  }
+}
+
 }  // namespace xe
diff --git a/src/xenia/base/string.h b/src/xenia/base/string.h
index 9ed7a1603..79f159e35 100644
--- a/src/xenia/base/string.h
+++ b/src/xenia/base/string.h
@@ -32,6 +32,8 @@ std::wstring to_absolute_path(const std::wstring& path);
 std::vector<std::string> split_path(const std::string& path);
 
 // Joins two path segments with the given separator.
+std::string join_paths(const std::string& left, const std::string& right,
+                       char sep = xe::path_separator);
 std::wstring join_paths(const std::wstring& left, const std::wstring& right,
                         wchar_t sep = xe::path_separator);
 
@@ -42,10 +44,13 @@ std::wstring fix_path_separators(const std::wstring& source,
 std::string fix_path_separators(const std::string& source,
                                 char new_sep = xe::path_separator);
 
-// Find the top directory name or filename from a path
+// Find the top directory name or filename from a path.
 std::string find_name_from_path(const std::string& path);
 std::wstring find_name_from_path(const std::wstring& path);
 
+// Get parent path of the given directory or filename.
+std::string find_base_path(const std::string& path);
+
 }  // namespace xe
 
 #endif  // XENIA_BASE_STRING_H_
diff --git a/src/xenia/cpu/processor.cc b/src/xenia/cpu/processor.cc
index ffb884a7f..6a7e0ef0a 100644
--- a/src/xenia/cpu/processor.cc
+++ b/src/xenia/cpu/processor.cc
@@ -325,15 +325,18 @@ bool Processor::Execute(ThreadState* thread_state, uint32_t address) {
 
   PPCContext* context = thread_state->context();
 
+  // Setup registers.
+  uint64_t previous_lr = context->lr;
   // This could be set to anything to give us a unique identifier to track
   // re-entrancy/etc.
-  uint32_t lr = 0xBEBEBEBE;
-
-  // Setup registers.
-  context->lr = lr;
+  context->lr = 0xBEBEBEBE;
 
   // Execute the function.
-  return fn->Call(thread_state, lr);
+  auto result = fn->Call(thread_state, uint32_t(context->lr));
+
+  context->lr = previous_lr;
+
+  return result;
 }
 
 uint64_t Processor::Execute(ThreadState* thread_state, uint32_t address,
diff --git a/src/xenia/kernel/kernel_state.cc b/src/xenia/kernel/kernel_state.cc
index 151f862e7..453eb2dcb 100644
--- a/src/xenia/kernel/kernel_state.cc
+++ b/src/xenia/kernel/kernel_state.cc
@@ -13,6 +13,7 @@
 
 #include "xenia/base/assert.h"
 #include "xenia/base/string.h"
+#include "xenia/cpu/processor.h"
 #include "xenia/emulator.h"
 #include "xenia/kernel/dispatcher.h"
 #include "xenia/kernel/xam_module.h"
@@ -113,8 +114,11 @@ XModule* KernelState::GetModule(const char* name) {
   } else {
     std::lock_guard<std::mutex> lock(object_mutex_);
 
-    for (XUserModule *module : user_modules_) {
-      if (module->name() == name) {
+    for (XUserModule* module : user_modules_) {
+      if ((strcasecmp(xe::find_name_from_path(module->path()).c_str(), name) ==
+           0) ||
+          (strcasecmp(module->name().c_str(), name) == 0) ||
+          (strcasecmp(module->path().c_str(), name) == 0)) {
         module->Retain();
         return module;
       }
@@ -147,27 +151,54 @@ void KernelState::SetExecutableModule(XUserModule* module) {
   }
 }
 
-XUserModule* KernelState::LoadUserModule(const char *name) {
-  std::lock_guard<std::mutex> lock(object_mutex_);
+XUserModule* KernelState::LoadUserModule(const char* raw_name) {
+  // Some games try to load relative to launch module, others specify full path.
+  std::string name = xe::find_name_from_path(raw_name);
+  std::string path(raw_name);
+  if (name == raw_name) {
+    path = xe::join_paths(xe::find_base_path(executable_module_->path()), name);
+  }
 
-  // See if we've already loaded it
-  for (XUserModule *module : user_modules_) {
-    if (module->name() == name) {
-      module->Retain();
-      return module;
+  XUserModule* module = nullptr;
+  {
+    std::lock_guard<std::mutex> lock(object_mutex_);
+
+    // See if we've already loaded it
+    for (XUserModule* existing_module : user_modules_) {
+      if (existing_module->path() == path) {
+        existing_module->Retain();
+        return existing_module;
+      }
     }
+
+    // Module wasn't loaded, so load it.
+    module = new XUserModule(this, path.c_str());
+    X_STATUS status = module->LoadFromFile(path);
+    if (XFAILED(status)) {
+      module->Release();
+      return nullptr;
+    }
+
+    user_modules_.push_back(module);
+    module->Retain();
   }
 
-  // Module wasn't loaded, so load it.
-  XUserModule *module = new XUserModule(this, name);
-  X_STATUS status = module->LoadFromFile(name);
-  if (XFAILED(status)) {
-    module->Release();
-    return nullptr;
+  module->Dump();
+
+  auto xex_header = module->xex_header();
+  if (xex_header->exe_entry_point) {
+    // Call DllMain(DLL_PROCESS_ATTACH):
+    // https://msdn.microsoft.com/en-us/library/windows/desktop/ms682583%28v=vs.85%29.aspx
+    uint64_t args[] = {
+        module->handle(),
+        1,  // DLL_PROCESS_ATTACH
+        0,  // 0 because always dynamic
+    };
+    auto thread_state = XThread::GetCurrentThread()->thread_state();
+    processor()->Execute(thread_state, xex_header->exe_entry_point, args,
+                         xe::countof(args));
   }
 
-  user_modules_.push_back(module);
-  module->Retain();
   return module;
 }
 
@@ -184,6 +215,52 @@ void KernelState::UnregisterThread(XThread* thread) {
   }
 }
 
+void KernelState::OnThreadExecute(XThread* thread) {
+  std::lock_guard<std::mutex> lock(object_mutex_);
+
+  // Must be called on executing thread.
+  assert_true(XThread::GetCurrentThread() == thread);
+
+  // Call DllMain(DLL_THREAD_ATTACH) for each user module:
+  // https://msdn.microsoft.com/en-us/library/windows/desktop/ms682583%28v=vs.85%29.aspx
+  auto thread_state = thread->thread_state();
+  for (auto user_module : user_modules_) {
+    auto xex_header = user_module->xex_header();
+    if (xex_header->exe_entry_point) {
+      uint64_t args[] = {
+          user_module->handle(),
+          2,  // DLL_THREAD_ATTACH
+          0,  // 0 because always dynamic
+      };
+      processor()->Execute(thread_state, xex_header->exe_entry_point, args,
+                           xe::countof(args));
+    }
+  }
+}
+
+void KernelState::OnThreadExit(XThread* thread) {
+  std::lock_guard<std::mutex> lock(object_mutex_);
+
+  // Must be called on executing thread.
+  assert_true(XThread::GetCurrentThread() == thread);
+
+  // Call DllMain(DLL_THREAD_DETACH) for each user module:
+  // https://msdn.microsoft.com/en-us/library/windows/desktop/ms682583%28v=vs.85%29.aspx
+  auto thread_state = thread->thread_state();
+  for (auto user_module : user_modules_) {
+    auto xex_header = user_module->xex_header();
+    if (xex_header->exe_entry_point) {
+      uint64_t args[] = {
+          user_module->handle(),
+          3,  // DLL_THREAD_DETACH
+          0,  // 0 because always dynamic
+      };
+      processor()->Execute(thread_state, xex_header->exe_entry_point, args,
+                           xe::countof(args));
+    }
+  }
+}
+
 XThread* KernelState::GetThreadByID(uint32_t thread_id) {
   std::lock_guard<std::mutex> lock(object_mutex_);
   XThread* thread = nullptr;
diff --git a/src/xenia/kernel/kernel_state.h b/src/xenia/kernel/kernel_state.h
index b2a413607..18bcdc3e5 100644
--- a/src/xenia/kernel/kernel_state.h
+++ b/src/xenia/kernel/kernel_state.h
@@ -73,6 +73,8 @@ class KernelState {
 
   void RegisterThread(XThread* thread);
   void UnregisterThread(XThread* thread);
+  void OnThreadExecute(XThread* thread);
+  void OnThreadExit(XThread* thread);
   XThread* GetThreadByID(uint32_t thread_id);
 
   void RegisterNotifyListener(XNotifyListener* listener);
diff --git a/src/xenia/kernel/objects/xthread.cc b/src/xenia/kernel/objects/xthread.cc
index 1de9bb376..7d293f426 100644
--- a/src/xenia/kernel/objects/xthread.cc
+++ b/src/xenia/kernel/objects/xthread.cc
@@ -109,10 +109,6 @@ uint32_t XThread::GetCurrentThreadId(const uint8_t* thread_state_block) {
   return xe::load_and_swap<uint32_t>(thread_state_block + 0x14C);
 }
 
-uint32_t XThread::thread_state() { return thread_state_address_; }
-
-uint32_t XThread::thread_id() { return thread_id_; }
-
 uint32_t XThread::last_error() {
   uint8_t* p = memory()->TranslateVirtual(thread_state_address_);
   return xe::load_and_swap<uint32_t>(p + 0x160);
@@ -326,6 +322,9 @@ void XThread::Execute() {
               thread_id_, handle(), name_.c_str(),
               xe::threading::current_thread_id());
 
+  // Let the kernel know we are starting.
+  kernel_state()->OnThreadExecute(this);
+
   // All threads get a mandatory sleep. This is to deal with some buggy
   // games that are assuming the 360 is so slow to create threads that they
   // have time to initialize shared structures AFTER CreateThread (RR).
@@ -349,6 +348,9 @@ void XThread::Execute() {
     // Treat the return code as an implicit exit code.
     Exit(exit_code);
   }
+
+  // Let the kernel know we are exiting.
+  kernel_state()->OnThreadExit(this);
 }
 
 void XThread::EnterCriticalRegion() {
diff --git a/src/xenia/kernel/objects/xthread.h b/src/xenia/kernel/objects/xthread.h
index 90aaa8a5c..d210d7edd 100644
--- a/src/xenia/kernel/objects/xthread.h
+++ b/src/xenia/kernel/objects/xthread.h
@@ -35,8 +35,9 @@ class XThread : public XObject {
   static uint32_t GetCurrentThreadHandle();
   static uint32_t GetCurrentThreadId(const uint8_t* thread_state_block);
 
-  uint32_t thread_state();
-  uint32_t thread_id();
+  uint32_t thread_state_ptr() const { return thread_state_address_; }
+  cpu::ThreadState* thread_state() const { return thread_state_; }
+  uint32_t thread_id() const { return thread_id_; }
   uint32_t last_error();
   void set_last_error(uint32_t error_code);
   const std::string& name() const { return name_; }
diff --git a/src/xenia/kernel/objects/xuser_module.cc b/src/xenia/kernel/objects/xuser_module.cc
index 5ea4d268b..265898daa 100644
--- a/src/xenia/kernel/objects/xuser_module.cc
+++ b/src/xenia/kernel/objects/xuser_module.cc
@@ -34,7 +34,7 @@ const xe_xex2_header_t* XUserModule::xex_header() {
   return xe_xex2_get_header(xex_);
 }
 
-X_STATUS XUserModule::LoadFromFile(const char* path) {
+X_STATUS XUserModule::LoadFromFile(std::string path) {
   X_STATUS result = X_STATUS_UNSUCCESSFUL;
   XFile* file = NULL;
 
@@ -42,7 +42,7 @@ X_STATUS XUserModule::LoadFromFile(const char* path) {
   // TODO(benvanik): make this code shared?
   auto fs_entry = kernel_state()->file_system()->ResolvePath(path);
   if (!fs_entry) {
-    XELOGE("File not found: %s", path);
+    XELOGE("File not found: %s", path.c_str());
     return X_STATUS_NO_SUCH_FILE;
   }
 
diff --git a/src/xenia/kernel/objects/xuser_module.h b/src/xenia/kernel/objects/xuser_module.h
index e97165e49..3ea84b315 100644
--- a/src/xenia/kernel/objects/xuser_module.h
+++ b/src/xenia/kernel/objects/xuser_module.h
@@ -10,9 +10,10 @@
 #ifndef XENIA_KERNEL_XBOXKRNL_XUSER_MODULE_H_
 #define XENIA_KERNEL_XBOXKRNL_XUSER_MODULE_H_
 
-#include "xenia/kernel/objects/xmodule.h"
+#include <string>
 
 #include "xenia/cpu/export_resolver.h"
+#include "xenia/kernel/objects/xmodule.h"
 #include "xenia/kernel/util/xex2.h"
 #include "xenia/xbox.h"
 
@@ -29,7 +30,7 @@ class XUserModule : public XModule {
 
   uint32_t execution_info_ptr() const { return execution_info_ptr_; }
 
-  X_STATUS LoadFromFile(const char* path);
+  X_STATUS LoadFromFile(std::string path);
   X_STATUS LoadFromMemory(const void* addr, const size_t length);
 
   uint32_t GetProcAddressByOrdinal(uint16_t ordinal) override;
diff --git a/src/xenia/kernel/xboxkrnl_modules.cc b/src/xenia/kernel/xboxkrnl_modules.cc
index 3afae32ca..f105a8436 100644
--- a/src/xenia/kernel/xboxkrnl_modules.cc
+++ b/src/xenia/kernel/xboxkrnl_modules.cc
@@ -8,6 +8,7 @@
  */
 
 #include "xenia/base/logging.h"
+#include "xenia/cpu/processor.h"
 #include "xenia/kernel/kernel_state.h"
 #include "xenia/kernel/objects/xuser_module.h"
 #include "xenia/kernel/util/shim_utils.h"
@@ -15,8 +16,6 @@
 #include "xenia/kernel/xboxkrnl_private.h"
 #include "xenia/xbox.h"
 
-#include "xenia/cpu/processor.h"
-
 namespace xe {
 namespace kernel {
 
@@ -218,51 +217,33 @@ SHIM_CALL XexLoadImage_shim(PPCContext* ppc_state, KernelState* state) {
 
   X_STATUS result = X_STATUS_NO_SUCH_FILE;
 
+  X_HANDLE module_handle = X_INVALID_HANDLE_VALUE;
   XModule* module = state->GetModule(module_name);
   if (module) {
-    module->RetainHandle();
-    SHIM_SET_MEM_32(handle_ptr, module->handle());
-    module->Release();
-
-    result = X_STATUS_SUCCESS;
+    // Existing module found, just add a reference and obtain a handle.
+    result = state->object_table()->AddHandle(module, &module_handle);
   } else {
-    XUserModule* usermod = state->LoadUserModule(module_name);
-    if (usermod) {
-      // If the module has an entry point function, we have to call it.
-      const xe_xex2_header_t* header = usermod->xex_header();
-      if (header->exe_entry_point) {
-        // Return address
-        uint32_t lr = ppc_state->thread_state->context()->lr;
-
-        // TODO: What are these args for?
-        // param 2: val 1 seems to make CRT initialize
-        uint64_t args[] = { 0, 1, 0 };
-        state->processor()->Execute(ppc_state->thread_state,
-                                    header->exe_entry_point,
-                                    args, xe::countof(args));
-
-        ppc_state->thread_state->context()->lr = lr;
-      }
-
+    // Not found; attempt to load as a user module.
+    module = state->LoadUserModule(module_name);
+    if (module) {
+      module->RetainHandle();
+      module_handle = module->handle();
       result = X_STATUS_SUCCESS;
-
-      usermod->RetainHandle();
-      SHIM_SET_MEM_32(handle_ptr, usermod->handle());
-      usermod->Release();
     }
   }
+  SHIM_SET_MEM_32(handle_ptr, module_handle);
 
   SHIM_SET_RETURN_32(result);
 }
 
 SHIM_CALL XexUnloadImage_shim(PPCContext* ppc_state, KernelState* state) {
-  uint32_t handle = SHIM_GET_ARG_32(0);
+  uint32_t module_handle = SHIM_GET_ARG_32(0);
 
-  XELOGD("XexUnloadImage(%.8X)", handle);
+  XELOGD("XexUnloadImage(%.8X)", module_handle);
 
   X_STATUS result = X_STATUS_INVALID_HANDLE;
 
-  result = state->object_table()->RemoveHandle(handle);
+  result = state->object_table()->RemoveHandle(module_handle);
 
   SHIM_SET_RETURN_32(result);
 }
diff --git a/src/xenia/kernel/xboxkrnl_ob.cc b/src/xenia/kernel/xboxkrnl_ob.cc
index 538652ff0..9aaaf1e09 100644
--- a/src/xenia/kernel/xboxkrnl_ob.cc
+++ b/src/xenia/kernel/xboxkrnl_ob.cc
@@ -69,7 +69,7 @@ SHIM_CALL ObReferenceObjectByHandle_shim(PPCContext* ppc_state,
       case 0xD01BBEEF:  // ExThreadObjectType
       {
         XThread* thread = (XThread*)object;
-        native_ptr = thread->thread_state();
+        native_ptr = thread->thread_state_ptr();
       } break;
       default:
         assert_unhandled_case(object_type_ptr);