use std::os::fd::{AsRawFd, RawFd}; use std::sync::atomic::{AtomicU64, Ordering}; use std::sync::{Arc, Mutex}; use std::time::{Duration, Instant}; use self::fd_monitor::{c_void, new_fd_event_signaller, FdEventSignaller, ItemWakeReason}; use crate::fd_readable_set::FdReadableSet; use crate::fds::AutoCloseFd; use crate::ffi::void_ptr; use crate::flog::FLOG; use crate::wutil::perror; use cxx::SharedPtr; #[cxx::bridge] mod fd_monitor { /// Reason for waking an item #[repr(u8)] #[cxx_name = "item_wake_reason_t"] enum ItemWakeReason { /// The fd became readable (or was HUP'd) Readable, /// The requested timeout was hit Timeout, /// The item was "poked" (woken up explicitly) Poke, } // Defines and exports a type shared between C++ and rust struct c_void { _unused: u8, } unsafe extern "C++" { include!("fds.h"); /// An event signaller implemented using a file descriptor, so it can plug into /// [`select()`](libc::select). /// /// This is like a binary semaphore. A call to [`post()`](FdEventSignaller::post) will /// signal an event, making the fd readable. Multiple calls to `post()` may be coalesced. /// On Linux this uses [`eventfd()`](libc::eventfd), on other systems this uses a pipe. /// [`try_consume()`](FdEventSignaller::try_consume) may be used to consume the event. /// Importantly this is async signal safe. Of course it is `CLO_EXEC` as well. #[rust_name = "FdEventSignaller"] type fd_event_signaller_t = crate::ffi::fd_event_signaller_t; #[rust_name = "new_fd_event_signaller"] fn ffi_new_fd_event_signaller_t() -> SharedPtr; } extern "Rust" { #[cxx_name = "fd_monitor_item_id_t"] type FdMonitorItemId; } extern "Rust" { #[cxx_name = "fd_monitor_item_t"] type FdMonitorItem; #[cxx_name = "make_fd_monitor_item_t"] fn new_fd_monitor_item_ffi( fd: i32, timeout_usecs: u64, callback: *const c_void, param: *const c_void, ) -> Box; } extern "Rust" { #[cxx_name = "fd_monitor_t"] type FdMonitor; #[cxx_name = "make_fd_monitor_t"] fn new_fd_monitor_ffi() -> Box; #[cxx_name = "add_item"] fn add_item_ffi( &mut self, fd: i32, timeout_usecs: u64, callback: *const c_void, param: *const c_void, ) -> u64; #[cxx_name = "poke_item"] fn poke_item_ffi(&self, item_id: u64); #[cxx_name = "add"] pub fn add_ffi(&mut self, item: Box) -> u64; } } // TODO: Remove once we're no longer using the FFI variant of FdEventSignaller unsafe impl Sync for FdEventSignaller {} unsafe impl Send for FdEventSignaller {} /// Each item added to fd_monitor_t is assigned a unique ID, which is not recycled. Items may have /// their callback triggered immediately by passing the ID. Zero is a sentinel. #[derive(Clone, Copy, Debug, PartialEq, Eq, PartialOrd, Ord)] pub struct FdMonitorItemId(u64); type FfiCallback = extern "C" fn(*mut AutoCloseFd, u8, void_ptr); /// The callback type used by [`FdMonitorItem`]. It is passed a mutable reference to the /// `FdMonitorItem`'s [`FdMonitorItem::fd`] and [the reason](ItemWakeupReason) for the wakeup. The /// callback may close the fd, in which case the `FdMonitorItem` is removed from [`FdMonitor`]'s /// set. /// /// As capturing C++ closures can't be safely used via ffi interop and cxx bridge doesn't support /// passing typed `fn(...)` pointers from C++ to rust, we have a separate variant of the type that /// uses the C abi to invoke a callback. This will be removed when the dependent C++ code (currently /// only `src/io.cpp`) is ported to rust enum FdMonitorCallback { None, Native(Box), Ffi(FfiCallback /* fn ptr */, void_ptr /* param */), } /// An item containing an fd and callback, which can be monitored to watch when it becomes readable /// and invoke the callback. pub struct FdMonitorItem { /// The fd to monitor fd: AutoCloseFd, /// A callback to be invoked when the fd is readable, or when we are timed out. If we time out, /// then timed_out will be true. If the fd is invalid on return from the function, then the item /// is removed from the [`FdMonitor`] set. callback: FdMonitorCallback, /// The timeout associated with waiting on this item or `None` to wait indefinitely. A timeout /// of `0` is not supported. timeout: Option, /// The last time we were called or the time of initialization. last_time: Option, /// The id for this item, assigned by [`FdMonitor`]. item_id: FdMonitorItemId, } /// Unlike C++, rust's `Vec` has `Vec::retain()` instead of `std::remove_if(...)` with the inverse /// logic. It's hard to keep track of which bool means what across the different layers, so be more /// explicit. #[derive(PartialEq, Eq)] enum ItemAction { Remove, Retain, } impl FdMonitorItem { /// Return the duration until the timeout should trigger or `None`. A return of `0` means we are /// at or past the timeout. fn remaining_time(&self, now: &Instant) -> Option { let last_time = self.last_time.expect("Should always have a last_time!"); let timeout = self.timeout?; assert!(now >= &last_time, "Steady clock went backwards or bug!"); let since = *now - last_time; Some(if since >= timeout { Duration::ZERO } else { timeout - since }) } /// Invoke this item's callback if its value (when its value is set in the fd or has timed out). /// Returns `true` if the item should be retained or `false` if it should be removed from the /// set. fn service_item(&mut self, fds: &FdReadableSet, now: &Instant) -> ItemAction { let mut result = ItemAction::Retain; let readable = fds.test(self.fd.as_raw_fd()); let timed_out = !readable && self.remaining_time(now) == Some(Duration::ZERO); if readable || timed_out { self.last_time = Some(*now); let reason = if readable { ItemWakeReason::Readable } else { ItemWakeReason::Timeout }; match &self.callback { FdMonitorCallback::None => panic!("Callback not assigned!"), FdMonitorCallback::Native(callback) => (callback)(&mut self.fd, reason), FdMonitorCallback::Ffi(callback, param) => { // Safety: identical objects are generated on both sides by cxx bridge as // integers of the same size (minimum size to fit the enum). let reason = unsafe { std::mem::transmute(reason) }; (callback)(&mut self.fd as *mut _, reason, *param) } } if !self.fd.is_valid() { result = ItemAction::Remove; } } return result; } /// Invoke this item's callback with a poke, if its id is present in the sorted poke list. fn maybe_poke_item(&mut self, pokelist: &[FdMonitorItemId]) -> ItemAction { if self.item_id.0 == 0 || pokelist.binary_search(&self.item_id).is_err() { // Not pokeable or not in the poke list. return ItemAction::Retain; } match &self.callback { FdMonitorCallback::None => panic!("Callback not assigned!"), FdMonitorCallback::Native(callback) => (callback)(&mut self.fd, ItemWakeReason::Poke), FdMonitorCallback::Ffi(callback, param) => { // Safety: identical objects are generated on both sides by cxx bridge as // integers of the same size (minimum size to fit the enum). let reason = unsafe { std::mem::transmute(ItemWakeReason::Poke) }; (callback)(&mut self.fd as *mut _, reason, *param) } } // Return `ItemAction::Remove` if the callback closed the fd match self.fd.is_valid() { true => ItemAction::Retain, false => ItemAction::Remove, } } fn new() -> Self { Self { callback: FdMonitorCallback::None, fd: AutoCloseFd::empty(), timeout: None, last_time: None, item_id: FdMonitorItemId(0), } } fn set_callback_ffi(&mut self, callback: *const c_void, param: *const c_void) { // Safety: we are just marshalling our function pointers with identical definitions on both // sides of the ffi bridge as void pointers to keep cxx bridge happy. Whether we invoke the // raw function as a void pointer or as a typed fn that helps us keep track of what we're // doing is unsafe in all cases, so might as well make the best of it. let callback = unsafe { std::mem::transmute(callback) }; self.callback = FdMonitorCallback::Ffi(callback, void_ptr(param as _)); } } // cxx bridge does not support "static member functions" in C++ or rust, so we need a top-level fn. fn new_fd_monitor_ffi() -> Box { Box::new(FdMonitor::new()) } // cxx bridge does not support "static member functions" in C++ or rust, so we need a top-level fn. fn new_fd_monitor_item_ffi( fd: RawFd, timeout_usecs: u64, callback: *const c_void, param: *const c_void, ) -> Box { // Safety: we are just marshalling our function pointers with identical definitions on both // sides of the ffi bridge as void pointers to keep cxx bridge happy. Whether we invoke the // raw function as a void pointer or as a typed fn that helps us keep track of what we're // doing is unsafe in all cases, so might as well make the best of it. let callback = unsafe { std::mem::transmute(callback) }; let mut item = FdMonitorItem::new(); item.fd.reset(fd); item.callback = FdMonitorCallback::Ffi(callback, void_ptr(param as _)); if timeout_usecs != FdReadableSet::kNoTimeout { item.timeout = Some(Duration::from_micros(timeout_usecs)); } return Box::new(item); } /// A thread-safe class which can monitor a set of fds, invoking a callback when any becomes /// readable (or has been HUP'd) or when per-item-configurable timeouts are reached. pub struct FdMonitor { /// Our self-signaller. When this is written to, it means there are new items pending, new items /// in the poke list, or terminate has been set. change_signaller: SharedPtr, /// The data shared between the background thread and the `FdMonitor` instance. data: Arc>, /// The last ID assigned or `0` if none. last_id: AtomicU64, } // We don't want to manually implement `Sync` for `FdMonitor` but we do want to make sure that it's // always using interior mutability correctly and therefore automatically `Sync`. const _: () = { // It is sufficient to declare the generic function pointers; calling them too would require // using `const fn` with Send/Sync constraints which wasn't stabilized until rustc 1.61.0 fn assert_sync() {} let _ = assert_sync::; }; /// Data shared between the `FdMonitor` instance and its associated `BackgroundFdMonitor`. struct SharedData { /// Pending items. This is set by the main thread with the mutex locked, then the background /// thread grabs them. pending: Vec, /// List of IDs for items that need to be poked (explicitly woken up). pokelist: Vec, /// Whether the background thread is running. running: bool, /// Used to signal that the background thread should terminate. terminate: bool, } /// The background half of the fd monitor, running on its own thread. struct BackgroundFdMonitor { /// The list of items to monitor. This is only accessed from the background thread. /// This doesn't need to be in any particular order. items: Vec, /// Our self-signaller. When this is written to, it means there are new items pending, new items /// in the poke list, or terminate has been set. change_signaller: SharedPtr, /// The data shared between the background thread and the `FdMonitor` instance. data: Arc>, } impl FdMonitor { pub fn add_ffi(&self, item: Box) -> u64 { self.add(*item).0 } /// Add an item to the monitor. Returns the [`FdMonitorItemId`] assigned to the item. pub fn add(&self, mut item: FdMonitorItem) -> FdMonitorItemId { assert!(item.fd.is_valid()); assert!(item.timeout != Some(Duration::ZERO), "Invalid timeout!"); assert!( item.item_id == FdMonitorItemId(0), "Item should not already have an id!" ); let item_id = self.last_id.fetch_add(1, Ordering::Relaxed) + 1; let item_id = FdMonitorItemId(item_id); let start_thread = { // Lock around a local region let mut data = self.data.lock().expect("Mutex poisoned!"); // Assign an id and add the item to pending item.item_id = item_id; data.pending.push(item); // Start the thread if it hasn't already been started let already_started = data.running; data.running = true; !already_started }; if start_thread { FLOG!(fd_monitor, "Thread starting"); let background_monitor = BackgroundFdMonitor { data: Arc::clone(&self.data), change_signaller: SharedPtr::clone(&self.change_signaller), items: Vec::new(), }; crate::threads::spawn(move || { background_monitor.run(); }); } item_id } /// Avoid requiring a separate UniquePtr for each item C++ wants to add to the set by giving an /// all-in-one entry point that can initialize the item on our end and insert it to the set. fn add_item_ffi( &mut self, fd: RawFd, timeout_usecs: u64, callback: *const c_void, param: *const c_void, ) -> u64 { // Safety: we are just marshalling our function pointers with identical definitions on both // sides of the ffi bridge as void pointers to keep cxx bridge happy. Whether we invoke the // raw function as a void pointer or as a typed fn that helps us keep track of what we're // doing is unsafe in all cases, so might as well make the best of it. let callback = unsafe { std::mem::transmute(callback) }; let mut item = FdMonitorItem::new(); item.fd.reset(fd); item.callback = FdMonitorCallback::Ffi(callback, void_ptr(param as _)); if timeout_usecs != FdReadableSet::kNoTimeout { item.timeout = Some(Duration::from_micros(timeout_usecs)); } let item_id = self.add(item).0; item_id } /// Mark that the item with the given ID needs to be woken up explicitly. pub fn poke_item(&self, item_id: FdMonitorItemId) { assert!(item_id.0 > 0, "Invalid item id!"); let needs_notification = { let mut data = self.data.lock().expect("Mutex poisoned!"); let needs_notification = data.pokelist.is_empty(); // Insert it, sorted. But not if it already exists. if let Err(pos) = data.pokelist.binary_search(&item_id) { data.pokelist.insert(pos, item_id); }; needs_notification }; if needs_notification { self.change_signaller.post(); } } fn poke_item_ffi(&self, item_id: u64) { self.poke_item(FdMonitorItemId(item_id)) } pub fn new() -> Self { Self { data: Arc::new(Mutex::new(SharedData { pending: Vec::new(), pokelist: Vec::new(), running: false, terminate: false, })), change_signaller: new_fd_event_signaller(), last_id: AtomicU64::new(0), } } } impl BackgroundFdMonitor { /// Starts monitoring the fd set and listening for new fds to add to the set. Takes ownership /// over its instance so that this method cannot be called again. fn run(mut self) { let mut pokelist: Vec = Vec::new(); let mut fds = FdReadableSet::new(); loop { // Poke any items that need it if !pokelist.is_empty() { self.poke(&mut pokelist); pokelist.clear(); } fds.clear(); // Our change_signaller is special-cased let change_signal_fd = self.change_signaller.read_fd().into(); fds.add(change_signal_fd); let mut now = Instant::now(); // Use Duration::MAX to represent no timeout for comparison purposes. let mut timeout = Duration::MAX; for item in &mut self.items { fds.add(item.fd.as_raw_fd()); if !item.last_time.is_some() { item.last_time = Some(now); } timeout = timeout.min(item.timeout.unwrap_or(Duration::MAX)); } // If we have no items, then we wish to allow the thread to exit, but after a time, so // we aren't spinning up and tearing down the thread repeatedly. Set a timeout of 256 // msec; if nothing becomes readable by then we will exit. We refer to this as the // wait-lap. let is_wait_lap = self.items.is_empty(); if is_wait_lap { assert!( timeout == Duration::MAX, "Should not have a timeout on wait lap!" ); timeout = Duration::from_millis(256); } // Don't leave Duration::MAX as an actual timeout value let timeout = match timeout { Duration::MAX => None, timeout => Some(timeout), }; // Call select() let ret = fds.check_readable( timeout .map(|duration| duration.as_micros() as u64) .unwrap_or(FdReadableSet::kNoTimeout), ); if ret < 0 && errno::errno().0 != libc::EINTR { // Surprising error perror("select"); } // Update the value of `now` after waiting on `fds.check_readable()`; it's used in the // servicer closure. now = Instant::now(); // A predicate which services each item in turn, returning true if it should be removed let servicer = |item: &mut FdMonitorItem| { let fd = item.fd.as_raw_fd(); if item.service_item(&fds, &now) == ItemAction::Remove { FLOG!(fd_monitor, "Removing fd", fd); return ItemAction::Remove; } return ItemAction::Retain; }; // Service all items that are either readable or have timed out, and remove any which // say to do so. self.items .retain_mut(|item| servicer(item) == ItemAction::Retain); // Handle any changes if the change signaller was set. Alternatively, this may be the // wait lap, in which case we might want to commit to exiting. let change_signalled = fds.test(change_signal_fd); if change_signalled || is_wait_lap { // Clear the change signaller before processing incoming changes self.change_signaller.try_consume(); let mut data = self.data.lock().expect("Mutex poisoned!"); // Move from `pending` to the end of `items` self.items.extend(&mut data.pending.drain(..)); // Grab any poke list assert!( pokelist.is_empty(), "poke list should be empty or else we're dropping pokes!" ); std::mem::swap(&mut pokelist, &mut data.pokelist); if data.terminate || (is_wait_lap && self.items.is_empty() && pokelist.is_empty() && !change_signalled) { // Maybe terminate is set. Alternatively, maybe we had no items, waited a bit, // and still have no items. It's important to do this while holding the lock, // otherwise we race with new items being added. assert!( data.running, "Thread should be running because we're that thread" ); FLOG!(fd_monitor, "Thread exiting"); data.running = false; break; } } } } /// Poke items in the poke list, removing any items that close their fd in their callback. The /// poke list is consumed after this. This is only called from the background thread. fn poke(&mut self, pokelist: &[FdMonitorItemId]) { self.items.retain_mut(|item| { let action = item.maybe_poke_item(&*pokelist); if action == ItemAction::Remove { FLOG!(fd_monitor, "Removing fd", item.fd.as_raw_fd()); } return action == ItemAction::Retain; }); } } /// In ordinary usage, we never invoke the destructor. This is used in the tests to not leave stale /// fds arounds; this is why it's very hacky! impl Drop for FdMonitor { fn drop(&mut self) { // Safety: this is a port of the C++ code and we are running in the destructor. The C++ code // had no way to bubble back any errors encountered here, and the pthread mutex the C++ code // uses does not have a concept of mutex poisoning. self.data.lock().expect("Mutex poisoned!").terminate = true; self.change_signaller.post(); // Safety: see note above. while self.data.lock().expect("Mutex poisoned!").running { std::thread::sleep(Duration::from_millis(5)); } } }