end-to-end impl of global loop w/ high-level ref counting.. needs work
- starting/stoping the loop based on client work is functioning, correctly - the issue appears to be that, when the process is about to exit, the signal to let weak tasks know that they need to exit isn't getting fired.
This commit is contained in:
parent
e6f6a8ced4
commit
31ba223c26
@ -6,27 +6,135 @@
|
||||
libuv functionality.
|
||||
"];
|
||||
|
||||
export high_level_loop;
|
||||
export run_high_level_loop, interact, ref_handle, unref_handle;
|
||||
// this will eventually move into its own, unexported (from std) module
|
||||
export get_global_loop;
|
||||
|
||||
import ll = uv_ll;
|
||||
|
||||
export high_level_loop;
|
||||
export interact, prepare_loop;
|
||||
native mod rustrt {
|
||||
fn rust_uv_get_kernel_global_chan_ptr() -> *libc::uintptr_t;
|
||||
fn rust_uv_get_kernel_global_async_handle() -> **libc::c_void;
|
||||
fn rust_uv_set_kernel_global_async_handle(handle: *ll::uv_async_t);
|
||||
fn rust_uv_free_kernel_global_async_handle();
|
||||
}
|
||||
|
||||
#[doc = "
|
||||
Used to abstract-away direct interaction with a libuv loop.
|
||||
|
||||
# Arguments
|
||||
|
||||
* async_handle - a pointer to a uv_async_t struct used to 'poke'
|
||||
* async_handle - a pointer to a pointer to a uv_async_t struct used to 'poke'
|
||||
the C uv loop to process any pending callbacks
|
||||
|
||||
* op_chan - a channel used to send function callbacks to be processed
|
||||
by the C uv loop
|
||||
"]
|
||||
type high_level_loop = {
|
||||
async_handle: *ll::uv_async_t,
|
||||
op_chan: comm::chan<fn~(*libc::c_void)>
|
||||
async_handle: **ll::uv_async_t,
|
||||
op_chan: comm::chan<high_level_msg>
|
||||
};
|
||||
|
||||
#[doc = "
|
||||
Race-free helper to get access to a global task where a libuv
|
||||
loop is running.
|
||||
|
||||
# Return
|
||||
|
||||
* A `high_level_loop` that encapsulates communication with the global loop.
|
||||
"]
|
||||
fn get_global_loop() -> high_level_loop unsafe {
|
||||
let global_loop_chan_ptr = rustrt::rust_uv_get_kernel_global_chan_ptr();
|
||||
log(debug, #fmt("ENTERING get_global_loop() loop chan: %?",
|
||||
global_loop_chan_ptr));
|
||||
log(debug, #fmt("ENTERING get_global_loop() loop chan: %?",
|
||||
global_loop_chan_ptr));
|
||||
log(debug,#fmt("value of loop ptr: %?", *global_loop_chan_ptr));
|
||||
|
||||
let builder_fn = {||
|
||||
let builder = task::builder();
|
||||
let opts = {
|
||||
supervise: true,
|
||||
notify_chan: none,
|
||||
sched:
|
||||
some({mode: task::manual_threads(1u),
|
||||
native_stack_size: none })
|
||||
};
|
||||
task::set_opts(builder, opts);
|
||||
builder
|
||||
};
|
||||
unsafe {
|
||||
log(debug, "before priv::chan_from_global_ptr");
|
||||
let chan = priv::chan_from_global_ptr::<high_level_msg>(
|
||||
global_loop_chan_ptr,
|
||||
builder_fn) {|port|
|
||||
|
||||
// the actual body of our global loop lives here
|
||||
log(debug, "initialized global port task!");
|
||||
log(debug, "GLOBAL!!!! initialized global port task!");
|
||||
outer_global_loop_body(port);
|
||||
};
|
||||
log(debug, "after priv::chan_from_global_ptr");
|
||||
let handle = get_global_async_handle();
|
||||
ret { async_handle: handle, op_chan: chan };
|
||||
}
|
||||
}
|
||||
|
||||
#[doc = "
|
||||
Takes a vanilla libuv `uv_loop_t*` ptr, performs some setup and then calls
|
||||
`uv_run()`. Users will be able to access this loop via a provided
|
||||
`async_handle` and `msg_ptr_po`. On top of libuv's internal handle refcount,
|
||||
the high_level_loop manages its own lifetime with a similar refcount scheme.
|
||||
|
||||
This call blocks for the lifetime of the libuv loop.
|
||||
|
||||
# Arguments
|
||||
|
||||
* loop_ptr - a pointer to a currently unused libuv loop. Its `data` field
|
||||
will be overwritten before the loop begins
|
||||
* async_handle - a pointer to a _fresh_ `ll::uv_async_t` record that _has
|
||||
not_ been initialized via `uv_async_init`, `ll::uv::async_init`, etc. It
|
||||
must be a pointer to a clean rust `uv_async_t` record
|
||||
* before_run - a unique closure that is invoked just before the call to
|
||||
`uv_run`
|
||||
* before_msg_drain - a unique closure that is invoked every time the loop is
|
||||
awoken, but before the port pointed to in the `msg_po` argument is drained.
|
||||
"]
|
||||
unsafe fn run_high_level_loop(loop_ptr: *libc::c_void,
|
||||
msg_po: comm::port<high_level_msg>,
|
||||
before_run: fn~(*global_loop_data),
|
||||
before_msg_drain: fn~() -> bool,
|
||||
before_tear_down: fn~()) {
|
||||
// set up the special async handle we'll use to allow multi-task
|
||||
// communication with this loop
|
||||
let async = ll::async_t();
|
||||
let async_handle = ptr::addr_of(async);
|
||||
// associate the async handle with the loop
|
||||
ll::async_init(loop_ptr, async_handle, high_level_wake_up_cb);
|
||||
|
||||
// initialize our loop data and store it in the loop
|
||||
let data: global_loop_data = {
|
||||
async_handle: async_handle,
|
||||
mut active: true,
|
||||
before_msg_drain: before_msg_drain,
|
||||
before_tear_down: before_tear_down,
|
||||
msg_po_ptr: ptr::addr_of(msg_po),
|
||||
mut refd_handles: [mut],
|
||||
mut unrefd_handles: [mut]
|
||||
};
|
||||
let data_ptr = ptr::addr_of(data);
|
||||
ll::set_data_for_uv_handle(async_handle, data_ptr);
|
||||
|
||||
// call before_run
|
||||
before_run(data_ptr);
|
||||
|
||||
log(debug, "about to run high level loop");
|
||||
// enter the loop... this blocks until the loop is done..
|
||||
ll::run(loop_ptr);
|
||||
log(debug, "high-level loop ended");
|
||||
}
|
||||
|
||||
#[doc = "
|
||||
Pass in a callback to be processed on the running libuv loop's thread
|
||||
|
||||
@ -35,7 +143,6 @@
|
||||
* a_loop - a high_level_loop record that represents a channel of
|
||||
communication with an active libuv loop running on a thread
|
||||
somwhere in the current process
|
||||
|
||||
* cb - a function callback to be processed on the running loop's
|
||||
thread. The only parameter is an opaque pointer to the running
|
||||
uv_loop_t. You can use this pointer to initiate or continue any
|
||||
@ -43,61 +150,343 @@
|
||||
"]
|
||||
unsafe fn interact(a_loop: high_level_loop,
|
||||
-cb: fn~(*libc::c_void)) {
|
||||
comm::send(a_loop.op_chan, cb);
|
||||
ll::async_send(a_loop.async_handle);
|
||||
send_high_level_msg(a_loop, interaction(cb));
|
||||
}
|
||||
|
||||
#[doc = "
|
||||
Prepares a clean, inactive uv_loop_t* to be used with any of the
|
||||
functions in the `uv::hl` module.
|
||||
iface uv_handle_manager<T> {
|
||||
fn init() -> T;
|
||||
}
|
||||
|
||||
Library developers can use this function to prepare a given
|
||||
`uv_loop_t*`, whose lifecycle they manage, to be used, ran
|
||||
and controlled with the tools in this module.
|
||||
resource uv_safe_handle<T>(handle_val: uv_handle_manager<T>) {
|
||||
}
|
||||
|
||||
After this is ran against a loop, a library developer can run
|
||||
the loop in its own thread and then use the returned
|
||||
`high_level_loop` to interact with it.
|
||||
|
||||
# Arguments
|
||||
|
||||
* loop_ptr - a pointer to a newly created `uv_loop_t*` with no
|
||||
handles registered (this will interfere with the internal lifecycle
|
||||
management this module provides). Ideally, this should be called
|
||||
immediately after using `uv::ll::loop_new()`
|
||||
|
||||
# Return
|
||||
|
||||
A `high_level_loop` record that can be used to interact with the
|
||||
loop (after you use `uv::ll::run()` on the `uv_loop_t*`, of course
|
||||
#[doc="
|
||||
"]
|
||||
unsafe fn prepare_loop(loop_ptr: *libc::c_void)
|
||||
-> high_level_loop {
|
||||
// will probably need to stake out a data record
|
||||
// here, as well, to keep whatever state we want to
|
||||
// use with the loop
|
||||
|
||||
// move this into a malloc
|
||||
let async = ll::async_t();
|
||||
let async_ptr = ptr::addr_of(async);
|
||||
let op_port = comm::port::<fn~(*libc::c_void)>();
|
||||
let async_result = ll::async_init(loop_ptr,
|
||||
async_ptr,
|
||||
interact_poke);
|
||||
if (async_result != 0i32) {
|
||||
fail ll::get_last_err_info(loop_ptr);
|
||||
}
|
||||
// need to store the port and async_ptr in the top-level
|
||||
// of the provided loop ..
|
||||
ret { async_handle: async_ptr,
|
||||
op_chan: comm::chan::<fn~(*libc::c_void)>(op_port)
|
||||
};
|
||||
fn ref_handle<T>(hl_loop: high_level_loop, handle: *T) unsafe {
|
||||
send_high_level_msg(hl_loop, auto_ref_handle(handle as *libc::c_void));
|
||||
}
|
||||
#[doc="
|
||||
"]
|
||||
fn unref_handle<T>(hl_loop: high_level_loop, handle: *T) unsafe {
|
||||
send_high_level_msg(hl_loop, auto_unref_handle(handle as *libc::c_void));
|
||||
}
|
||||
|
||||
// this will be invoked by a called to uv::hl::interact(), so
|
||||
// we'll drain the port of pending callbacks, processing each
|
||||
crust fn interact_poke(async_handle: *libc::c_void) {
|
||||
/////////////////////
|
||||
// INTERNAL API
|
||||
/////////////////////
|
||||
|
||||
unsafe fn send_high_level_msg(hl_loop: high_level_loop,
|
||||
-msg: high_level_msg) unsafe {
|
||||
comm::send(hl_loop.op_chan, msg);
|
||||
|
||||
// if the global async handle == 0, then that means
|
||||
// the loop isn't active, so we don't need to wake it up,
|
||||
// (the loop's enclosing task should be blocking on a message
|
||||
// receive on this port)
|
||||
if (*(hl_loop.async_handle) != 0 as *ll::uv_async_t) {
|
||||
log(debug,"global async handle != 0, waking up loop..");
|
||||
ll::async_send(*(hl_loop.async_handle));
|
||||
}
|
||||
else {
|
||||
log(debug,"GLOBAL ASYNC handle == 0");
|
||||
}
|
||||
}
|
||||
|
||||
// this will be invoked by a call to uv::hl::interact() with
|
||||
// the high_level_loop corresponding to this async_handle. We
|
||||
// simply check if the loop is active and, if so, invoke the
|
||||
// user-supplied on_wake callback that is stored in the loop's
|
||||
// data member
|
||||
crust fn high_level_wake_up_cb(async_handle: *libc::c_void,
|
||||
status: int) unsafe {
|
||||
// nothing here, yet.
|
||||
log(debug, #fmt("interact_poke crust.. handle: %?",
|
||||
log(debug, #fmt("high_level_wake_up_cb crust.. handle: %?",
|
||||
async_handle));
|
||||
let loop_ptr = ll::get_loop_for_uv_handle(async_handle);
|
||||
let data = ll::get_data_for_uv_handle(async_handle) as *global_loop_data;
|
||||
// we check to see if the loop is "active" (the loop is set to
|
||||
// active = false the first time we realize we need to 'tear down',
|
||||
// set subsequent calls to the global async handle may be triggered
|
||||
// before all of the uv_close() calls are processed and loop exits
|
||||
// on its own. So if the loop isn't active, we won't run the user's
|
||||
// on_wake callback (and, consequently, let messages pile up, probably
|
||||
// in the loops msg_po)
|
||||
if (*data).active {
|
||||
log(debug, "before on_wake");
|
||||
let mut do_msg_drain = (*data).before_msg_drain();
|
||||
let mut continue = true;
|
||||
if do_msg_drain {
|
||||
let msg_po = *((*data).msg_po_ptr);
|
||||
if comm::peek(msg_po) {
|
||||
// if this is true, we'll iterate over the
|
||||
// msgs waiting in msg_po until there's no more
|
||||
log(debug,"got msg_po");
|
||||
while(continue) {
|
||||
log(debug,"before alt'ing on high_level_msg");
|
||||
alt comm::recv(msg_po) {
|
||||
interaction(cb) {
|
||||
log(debug,"got interaction, before cb..");
|
||||
// call it..
|
||||
cb(loop_ptr);
|
||||
log(debug,"after calling cb");
|
||||
}
|
||||
auto_ref_handle(handle) {
|
||||
high_level_ref(data, handle);
|
||||
}
|
||||
auto_unref_handle(handle) {
|
||||
high_level_unref(data, handle, false);
|
||||
}
|
||||
tear_down {
|
||||
log(debug,"incoming hl_msg: got tear_down");
|
||||
}
|
||||
}
|
||||
continue = comm::peek(msg_po);
|
||||
}
|
||||
}
|
||||
}
|
||||
log(debug, #fmt("after on_wake, continue? %?", continue));
|
||||
if !do_msg_drain {
|
||||
high_level_tear_down(data);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
crust fn tear_down_close_cb(handle: *ll::uv_async_t) unsafe {
|
||||
log(debug, #fmt("tear_down_close_cb called, closing handle at %?",
|
||||
handle));
|
||||
// TODO: iterate through open handles on the loop and uv_close()
|
||||
// them all
|
||||
let data = ll::get_data_for_uv_handle(handle) as *global_loop_data;
|
||||
}
|
||||
|
||||
fn high_level_tear_down(data: *global_loop_data) unsafe {
|
||||
log(debug, "high_level_tear_down() called, close async_handle");
|
||||
// call user-suppled before_tear_down cb
|
||||
(*data).before_tear_down();
|
||||
let async_handle = (*data).async_handle;
|
||||
ll::close(async_handle as *libc::c_void, tear_down_close_cb);
|
||||
}
|
||||
|
||||
unsafe fn high_level_ref(data: *global_loop_data, handle: *libc::c_void) {
|
||||
log(debug,"incoming hl_msg: got auto_ref_handle");
|
||||
let mut refd_handles = (*data).refd_handles;
|
||||
let handle_already_refd = refd_handles.contains(handle);
|
||||
if handle_already_refd {
|
||||
fail "attempt to do a high-level ref an already ref'd handle";
|
||||
}
|
||||
refd_handles += [handle];
|
||||
(*data).refd_handles = refd_handles;
|
||||
}
|
||||
|
||||
crust fn auto_unref_close_cb(handle: *libc::c_void) {
|
||||
log(debug, "closing handle via high_level_unref");
|
||||
}
|
||||
|
||||
unsafe fn high_level_unref(data: *global_loop_data, handle: *libc::c_void,
|
||||
manual_unref: bool) {
|
||||
log(debug,"incoming hl_msg: got auto_unref_handle");
|
||||
let mut refd_handles = (*data).refd_handles;
|
||||
let mut unrefd_handles = (*data).unrefd_handles;
|
||||
let handle_already_refd = refd_handles.contains(handle);
|
||||
if !handle_already_refd {
|
||||
fail "attempting to high-level unref an untracked handle";
|
||||
}
|
||||
let double_unref = unrefd_handles.contains(handle);
|
||||
if double_unref {
|
||||
if manual_unref {
|
||||
// will allow a user to manual unref, but only signal
|
||||
// a fail when a double-unref is caused by a user
|
||||
fail "attempting to high-level unref an unrefd handle";
|
||||
}
|
||||
}
|
||||
else {
|
||||
ll::close(handle, auto_unref_close_cb);
|
||||
let last_idx = vec::len(refd_handles) - 1u;
|
||||
let handle_idx = vec::position_elem(refd_handles, handle);
|
||||
alt handle_idx {
|
||||
none {
|
||||
fail "trying to remove handle that isn't in refd_handles";
|
||||
}
|
||||
some(idx) {
|
||||
refd_handles[idx] <-> refd_handles[last_idx];
|
||||
vec::pop(refd_handles);
|
||||
}
|
||||
}
|
||||
(*data).refd_handles = refd_handles;
|
||||
unrefd_handles += [handle];
|
||||
(*data).unrefd_handles = unrefd_handles;
|
||||
if vec::len(refd_handles) == 0u {
|
||||
log(debug, "0 referenced handles, start loop teardown");
|
||||
high_level_tear_down(data);
|
||||
}
|
||||
else {
|
||||
log(debug, "more than 0 referenced handles");
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
enum high_level_msg {
|
||||
interaction (fn~(*libc::c_void)),
|
||||
auto_ref_handle (*libc::c_void),
|
||||
auto_unref_handle (*libc::c_void),
|
||||
tear_down
|
||||
}
|
||||
|
||||
fn get_global_async_handle() -> **ll::uv_async_t {
|
||||
ret rustrt::rust_uv_get_kernel_global_async_handle() as **ll::uv_async_t;
|
||||
}
|
||||
|
||||
fn set_global_async_handle(handle: *ll::uv_async_t) {
|
||||
rustrt::rust_uv_set_kernel_global_async_handle(handle);
|
||||
}
|
||||
|
||||
type global_loop_data = {
|
||||
async_handle: *ll::uv_async_t,
|
||||
mut active: bool,
|
||||
before_msg_drain: fn~() -> bool,
|
||||
before_tear_down: fn~(),
|
||||
msg_po_ptr: *comm::port<high_level_msg>,
|
||||
mut refd_handles: [mut *libc::c_void],
|
||||
mut unrefd_handles: [mut *libc::c_void]
|
||||
};
|
||||
|
||||
unsafe fn outer_global_loop_body(msg_po: comm::port<high_level_msg>) {
|
||||
// we're going to use a single libuv-generated loop ptr
|
||||
// for the duration of the process
|
||||
let loop_ptr = ll::loop_new();
|
||||
|
||||
// data structure for loop goes here..
|
||||
|
||||
// immediately weaken the task this is running in.
|
||||
priv::weaken_task() {|weak_exit_po|
|
||||
// when we first enter this loop, we're going
|
||||
// to wait on stand-by to receive a request to
|
||||
// fire-up the libuv loop
|
||||
let mut continue = true;
|
||||
while continue {
|
||||
log(debug, "in outer_loop...");
|
||||
continue = either::either(
|
||||
{|left_val|
|
||||
// bail out..
|
||||
// if we catch this msg at this point,
|
||||
// we should just be able to exit because
|
||||
// the loop isn't active
|
||||
log(debug, "got msg on weak_exit_po in outer loop");
|
||||
false
|
||||
}, {|right_val|
|
||||
log(debug, "about to enter inner loop");
|
||||
inner_global_loop_body(weak_exit_po, msg_po, loop_ptr,
|
||||
copy(right_val))
|
||||
}, comm::select2(weak_exit_po, msg_po));
|
||||
log(debug,#fmt("GLOBAL LOOP EXITED, WAITING TO RESTART? %?",
|
||||
continue));
|
||||
}
|
||||
};
|
||||
|
||||
ll::loop_delete(loop_ptr);
|
||||
// once we get here, show's over.
|
||||
rustrt::rust_uv_free_kernel_global_async_handle();
|
||||
}
|
||||
|
||||
unsafe fn inner_global_loop_body(weak_exit_po_in: comm::port<()>,
|
||||
msg_po_in: comm::port<high_level_msg>,
|
||||
loop_ptr: *libc::c_void,
|
||||
-first_interaction: high_level_msg) -> bool {
|
||||
// resend the msg
|
||||
comm::send(comm::chan(msg_po_in), first_interaction);
|
||||
|
||||
// black magic
|
||||
let weak_exit_po_ptr = ptr::addr_of(weak_exit_po_in);
|
||||
run_high_level_loop(
|
||||
loop_ptr,
|
||||
msg_po_in,
|
||||
// before_run
|
||||
{|data|
|
||||
// set the handle as the global
|
||||
set_global_async_handle((*data).async_handle);
|
||||
// when this is ran, our async_handle is set up, so let's
|
||||
// do an async_send with it
|
||||
ll::async_send((*data).async_handle);
|
||||
},
|
||||
// before_msg_drain
|
||||
{||
|
||||
log(debug,"entering before_msg_drain for the global loop");
|
||||
let weak_exit_po = *weak_exit_po_ptr;
|
||||
if(comm::peek(weak_exit_po)) {
|
||||
// if this is true, immediately bail and return false, causing
|
||||
// the libuv loop to start tearing down
|
||||
log(debug,"got weak_exit meg inside libuv loop");
|
||||
comm::recv(weak_exit_po);
|
||||
false
|
||||
}
|
||||
// if no weak_exit_po msg is received, then we'll let the
|
||||
// loop continue
|
||||
else {
|
||||
true
|
||||
}
|
||||
},
|
||||
// before_tear_down
|
||||
{||
|
||||
set_global_async_handle(0 as *ll::uv_async_t);
|
||||
});
|
||||
// supposed to return a bool to indicate to the enclosing loop whether
|
||||
// it should continue or not..
|
||||
ret true;
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod test {
|
||||
crust fn simple_timer_close_cb(timer_ptr: *ll::uv_timer_t) unsafe {
|
||||
log(debug, "UNUSED...");
|
||||
}
|
||||
crust fn simple_timer_cb(timer_ptr: *ll::uv_timer_t,
|
||||
status: libc::c_int) unsafe {
|
||||
log(debug, "in simple timer cb");
|
||||
let exit_ch_ptr = ll::get_data_for_uv_handle(
|
||||
timer_ptr as *libc::c_void) as *comm::chan<bool>;
|
||||
ll::timer_stop(timer_ptr);
|
||||
let hl_loop = get_global_loop();
|
||||
interact(hl_loop) {|loop_ptr|
|
||||
log(debug, "closing timer");
|
||||
//ll::close(timer_ptr as *libc::c_void, simple_timer_close_cb);
|
||||
unref_handle(hl_loop, timer_ptr);
|
||||
log(debug, "about to deref exit_ch_ptr");
|
||||
let exit_ch = *exit_ch_ptr;
|
||||
comm::send(exit_ch, true);
|
||||
log(debug, "after msg sent on deref'd exit_ch");
|
||||
};
|
||||
log(debug, "exiting simple timer cb");
|
||||
}
|
||||
#[test]
|
||||
fn test_uv_hl_simple_timer() unsafe {
|
||||
let exit_po = comm::port::<bool>();
|
||||
let exit_ch = comm::chan(exit_po);
|
||||
let exit_ch_ptr = ptr::addr_of(exit_ch);
|
||||
let hl_loop = get_global_loop();
|
||||
let timer_handle = ll::timer_t();
|
||||
let timer_ptr = ptr::addr_of(timer_handle);
|
||||
interact(hl_loop) {|loop_ptr|
|
||||
log(debug, "user code inside interact loop!!!");
|
||||
let init_status = ll::timer_init(loop_ptr, timer_ptr);
|
||||
if(init_status == 0i32) {
|
||||
ref_handle(hl_loop, timer_ptr);
|
||||
ll::set_data_for_uv_handle(
|
||||
timer_ptr as *libc::c_void,
|
||||
exit_ch_ptr as *libc::c_void);
|
||||
let start_status = ll::timer_start(timer_ptr, simple_timer_cb,
|
||||
1u, 0u);
|
||||
if(start_status == 0i32) {
|
||||
}
|
||||
else {
|
||||
fail "failure on ll::timer_start()";
|
||||
}
|
||||
}
|
||||
else {
|
||||
fail "failure on ll::timer_init()";
|
||||
}
|
||||
};
|
||||
comm::recv(exit_po);
|
||||
log(debug, "test_uv_hl_simple_timer: msg recv on exit_po, done..");
|
||||
}
|
||||
}
|
Loading…
Reference in New Issue
Block a user