That function is a sort of paraphrase of and glued-together-mess of XNU's `thread_deallocate_complete` and `thread_terminate_queue_invoke`, and I guess I accidentally deleted the `task_unlock` part.
Major changes:
* We now use BSD `proc` and `uthread` structures
* We now use XNU's own pthread interface
* This is done to increase compatibility with XNU code and make it easier to include XNU code that depends on it
* This is pretty easy due to Apple having modularized the pthread implementation into a kext: all we have to do is pretend to be that kext
* We compile the Blocks runtime (libclosure)
* GCC doesn't support blocks, so the LKM only compiles with Clang at the moment
* Unfortunately, Apple has decided to use them in the Mach code of XNU now.
I didn't look into just how much they use them and how easy/difficult it would be to remove them,
but that's the current situation.
* We compile and use(?) IPC importance code
Other than that, there were many more small changes that needed to be made (which was to be expected; we jumped from XNU-3000-ish to XNU-6153).
This code has *not* been tested to run yet, and probably won't run successfully without further modifications.
All the files that were building with the previous version of XNU are now building again (with Clang), however, the module does not link yet (due to missing symbols; *lots* of them). A lot of them are probably new additions to XNU, but some of them are probably from me not knowing the `ifdef` certain things that should be.
Also, I completely overhauled the Makefile to make it simpler to manage settings and flags across files, folders, modules, and Linux/KBuild versions. I didn't add this feature (because I don't need it yet), but it can be easily extended to easily allow per-compiler flags
Full list of all missing symbols reported by `MODPOST`:
task_is_driver
thread_get_state_to_user
machine_exception
catch_mach_exception_raise_state_identity
turnstile_has_waiters
mach_vm_allocate_kernel
processor_start_from_user
catch_mach_exception_raise_state
task_violated_guard
hw_atomic_test_and_set32
task_is_importance_donor
catch_mach_exception_raise
ipc_importance_task_reference
work_interval_port_notify
random_bool_init
os_ref_retain_try_internal
mach_zone_info_for_zone
processor_exit_from_user
pqueue_pair_meld
mach_vm_page_range_query
turnstile_complete
mach_vm_wire_external
mach_vm_allocate_external
vm_allocate_kernel
ux_handler_init
uext_server
turnstile_waitq_add_thread_priority_queue
thread_depress_abort_from_user
turnstile_deallocate
_Block_object_assign
turnstile_cleanup
turnstile_kernel_update_inheritor_on_wake_locked
_pthread_priority_normalize_for_ipc
filt_ipc_kqueue_turnstile
lck_spin_assert
IOTaskHasEntitlement
thread_get_requested_qos
task_watchport_elem_deallocate
os_ref_init_count_internal
lck_mtx_lock_spin_always
turnstile_update_inheritor_complete
task_inspect
thread_bootstrap_return
thread_setstatus_from_user
turnstile_recompute_priority_locked
mach_vm_remap_external
zone_require
thread_getstatus_to_user
turnstile_hash_bucket_unlock
_NSConcreteGlobalBlock
filt_machport_kqueue_has_turnstile
mach_continuous_time
ipc_importance_init
turnstile_update_inheritor_locked
turnstile_alloc
ipc_importance_send
ipc_importance_thread_call_init
bank_get_bank_ledger_thread_group_and_persona
turnstile_hash_bucket_lock
filt_machport_turnstile_prepare_lazily
turnstile_deallocate_safe
os_ref_release_barrier_internal
turnstile_reference
vm_map_wire_kernel
thread_deallocate_safe
turnstile_stats_update
thread_set_pending_block_hint
task_info_from_user
thread_inspect_deallocate
catch_exc_subsystem
ux_handler_stop
lck_mtx_assert
mach_vm_map_external
filt_machport_stash_port
ipc_importance_task_hold_internal_assertion
lck_spin_lock_grp
_Block_object_dispose
ipc_importance_check_circularity
task_restartable_subsystem
ipc_importance_task_drop_internal_assertion
random_bool_gen_bits
turnstile_update_inheritor
lck_spin_try_lock_grp
ipc_importance_task_release
kdp_lck_spin_is_acquired
ipc_importance_receive
os_ref_retain_internal
task_inspect_deallocate
_NSConcreteStackBlock
task_get_exc_guard_behavior
pid_from_task
sched_thread_unpromote_reason
memory_entry_subsystem
turnstile_prepare
act_get_state_to_user
sched_thread_promote_reason
task_set_exc_guard_behavior
ipc_importance_task_is_any_receiver_type
knote_vanish
thread_user_promotion_qos_for_pri
I'm not entirely sure if Darling *should* have this, but it can't hurt to build it, right? Besides, if we end up not needing it, we can just not add it with `add_subdirectory` in Darling's `src/CMakeLists.txt`
When a message is being sent from inside the kernel to a kernel thread,
mach_kernelrpc.c expects that re-sending the message will not be interrupted again.
Without this change, it would loop forever.
