Kernel: Make sure we free the thread stack on thread exit

This adds two new arguments to the thread_exit system call which let
a thread unmap an arbitrary VM range on thread exit. LibPthread
uses this functionality to unmap the thread stack.

Fixes #7267.

Kernel/Process.h

@@ -379,7 +379,7 @@ public:
     KResultOr<int> sys$sched_setparam(pid_t pid, Userspace<const struct sched_param*>);
     KResultOr<int> sys$sched_getparam(pid_t pid, Userspace<struct sched_param*>);
     KResultOr<int> sys$create_thread(void* (*)(void*), Userspace<const Syscall::SC_create_thread_params*>);
-    [[noreturn]] void sys$exit_thread(Userspace<void*>);
+    [[noreturn]] void sys$exit_thread(Userspace<void*>, Userspace<void*>, size_t);
     KResultOr<int> sys$join_thread(pid_t tid, Userspace<void**> exit_value);
     KResultOr<int> sys$detach_thread(pid_t tid);
     KResultOr<int> sys$set_thread_name(pid_t tid, Userspace<const char*> buffer, size_t buffer_size);

Kernel/Syscall.cpp

@@ -97,7 +97,7 @@ KResultOr<FlatPtr> handle(RegisterState& regs, FlatPtr function, FlatPtr arg1, F
             process.sys$exit(arg1);
             break;
         case SC_exit_thread:
-            process.sys$exit_thread(arg1);
+            process.sys$exit_thread(arg1, arg2, arg3);
             break;
         default:
             VERIFY_NOT_REACHED();

Kernel/Syscalls/thread.cpp

@@ -78,7 +78,7 @@ KResultOr<int> Process::sys$create_thread(void* (*entry)(void*), Userspace<const
     return thread->tid().value();
 }
 
-void Process::sys$exit_thread(Userspace<void*> exit_value)
+void Process::sys$exit_thread(Userspace<void*> exit_value, Userspace<void*> stack_location, size_t stack_size)
 {
     REQUIRE_PROMISE(thread);
 
@@ -90,6 +90,12 @@ void Process::sys$exit_thread(Userspace<void*> exit_value)
     auto current_thread = Thread::current();
     PerformanceManager::add_thread_exit_event(*current_thread);
 
+    if (stack_location) {
+        auto unmap_result = space().unmap_mmap_range(VirtualAddress { stack_location }, stack_size);
+        if (unmap_result.is_error())
+            dbgln("Failed to unmap thread stack, terminating thread anyway. Error code: {}", unmap_result.error());
+    }
+
     current_thread->exit(reinterpret_cast<void*>(exit_value.ptr()));
     VERIFY_NOT_REACHED();
 }

Userland/Libraries/LibPthread/pthread.cpp

@@ -33,13 +33,18 @@ static constexpr size_t required_stack_alignment = 4 * MiB;
 static constexpr size_t highest_reasonable_guard_size = 32 * PAGE_SIZE;
 static constexpr size_t highest_reasonable_stack_size = 8 * MiB; // That's the default in Ubuntu?
 
+__thread void* s_stack_location;
+__thread size_t s_stack_size;
+
 #define __RETURN_PTHREAD_ERROR(rc) \
     return ((rc) < 0 ? -(rc) : 0)
 
 extern "C" {
 
-static void* pthread_create_helper(void* (*routine)(void*), void* argument)
+static void* pthread_create_helper(void* (*routine)(void*), void* argument, void* stack_location, size_t stack_size)
 {
+    s_stack_location = stack_location;
+    s_stack_size = stack_size;
     void* ret_val = routine(argument);
     pthread_exit(ret_val);
     return nullptr;
@@ -57,10 +62,12 @@ static int create_thread(pthread_t* thread, void* (*entry)(void*), void* argumen
 
     // We set up the stack for pthread_create_helper.
     // Note that we need to align the stack to 16B, accounting for
-    // the fact that we also push 8 bytes.
-    while (((uintptr_t)stack - 8) % 16 != 0)
+    // the fact that we also push 16 bytes.
+    while (((uintptr_t)stack - 16) % 16 != 0)
         push_on_stack(nullptr);
 
+    push_on_stack((void*)thread_params->m_stack_size);
+    push_on_stack(thread_params->m_stack_location);
     push_on_stack(argument);
     push_on_stack((void*)entry);
     VERIFY((uintptr_t)stack % 16 == 0);
@@ -74,10 +81,10 @@ static int create_thread(pthread_t* thread, void* (*entry)(void*), void* argumen
     __RETURN_PTHREAD_ERROR(rc);
 }
 
-[[noreturn]] static void exit_thread(void* code)
+[[noreturn]] static void exit_thread(void* code, void* stack_location, size_t stack_size)
 {
     __pthread_key_destroy_for_current_thread();
-    syscall(SC_exit_thread, code);
+    syscall(SC_exit_thread, code, stack_location, stack_size);
     VERIFY_NOT_REACHED();
 }
 
@@ -119,7 +126,7 @@ int pthread_create(pthread_t* thread, pthread_attr_t* attributes, void* (*start_
 
 void pthread_exit(void* value_ptr)
 {
-    exit_thread(value_ptr);
+    exit_thread(value_ptr, s_stack_location, s_stack_size);
 }
 
 void pthread_cleanup_push([[maybe_unused]] void (*routine)(void*), [[maybe_unused]] void* arg)