1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
//! The core event handling primitives of Sunrise. //! //! The Sunrise kernel supports a couple sources of events, such as IRQs, timers, //! userspace-triggered events, and other. It must be possible to await for //! one or multiple events at the same time. //! //! In order to do this, we have the Event trait. It works by (in theory) //! registering an interest in an event, and putting the current process to //! sleep (deregistering it from the scheduler). When the event is triggered, //! the scheduler will wake the process up, allowing it work. use core::sync::atomic::{AtomicUsize, AtomicBool, Ordering}; use core::fmt::Debug; use alloc::sync::Arc; use crate::sync::{SpinLock, SpinLockIRQ}; use alloc::vec::Vec; use crate::error::{KernelError, UserspaceError}; use crate::process::ThreadStruct; use crate::scheduler; use failure::Backtrace; /// A waitable item. /// /// There are essentially two kinds of Waitables: user-signaled and IRQ-backed. /// IRQ-backed waitables are implemented by [IRQEvent], while user-signaled /// events are implemented by [ReadableEvent]. /// /// It is possible that a raw waitable is not flexible enough though. For /// instance, if we want to wait for 1 second, it might be necessary to wait on /// the timer event multiple times. To do this, it is possible to implement our /// own Waitable, that defers register to the underlying IRQEvent, but adds /// additional logic to is_signaled. For example: /// /// ``` /// use sunrise_kernel::event::{IRQEvent, Waitable}; /// use core::sync::atomic::{AtomicUsize, Ordering}; /// struct WaitFor5Ticks(IRQEvent, AtomicUsize); /// impl Waitable for WaitFor5Ticks { /// fn is_signaled(&self) -> bool { /// self.1.compare_and_swap(0, 5, Ordering::SeqCst); /// if self.0.is_signaled() { /// if self.1.fetch_sub(1) == 0 { /// return true; /// } else { /// return false; /// } /// } else { /// return false; /// } /// } /// fn register(&self) { /// self.0.register() /// } /// } /// ``` pub trait Waitable: Debug + Send + Sync { /// Checks whether the Waitable was signalled. /// /// If it returns false, the register function will be called again, in order /// to get notified of the next wakeup. /// /// This will likely require to change state - and yet it takes self by value. /// the reason for this is that it's possible for multiple threads, and /// potentially multiple CPUs, to wait on the same Waitable. Think of servers: /// you might want to wait for multiple threads for the arrival of a new socket. /// When this happens, **only a single thread should return true**. Make extra /// sure your Atomic operations are written properly! /// /// You'll probably want to check out AtomicUsize::fetch_update to make sure your /// atomic update loops are correct. fn is_signaled(&self) -> bool; /// Register the waitable with the scheduler. /// /// This should ensure that when the event is (or is likely to be) triggered, /// the scheduler puts the Process back in the running Vec. Most implementors /// will want to defer this to an IRQEvent. For instance: /// /// ``` /// #use sunrise_kernel::event::{IRQEvent, Waitable}; /// #struct Wait(IRQEvent); /// #impl Waitable for WaitFor5Ticks { /// #fn is_signaled(&self) -> bool { /// #self.0.is_signaled() /// #} /// fn register(&self) { /// self.0.register() /// } /// #} /// ``` fn register(&self); } /// Waits for an event to occur on one of the given Waitable objects. pub fn wait<'wait, INTOITER>(waitable_intoiter: INTOITER) -> Result<&'wait dyn Waitable, UserspaceError> where INTOITER: IntoIterator<Item=&'wait dyn Waitable>, <INTOITER as IntoIterator>::IntoIter: Clone { let _thread = scheduler::get_current_thread(); let waitable = waitable_intoiter.into_iter(); let interrupt_manager = SpinLockIRQ::new(()); loop { // Early-check for events that have already been signaled. for item in waitable.clone() { if item.is_signaled() { return Ok(item); } } // Disable interrupts between registration and unschedule. let lock = interrupt_manager.lock(); // Register the process for wakeup on all the possible events for item in waitable.clone() { item.register(); } // TODO: check that the current process is registered for an event, // bug otherwise. // Schedule scheduler::unschedule(&interrupt_manager, lock)?; } } /// The underlying shared object of a [ReadableEvent]/[WritableEvent]. #[derive(Debug)] struct Event { /// The state determines whether the event is signaled or not. When it is true, /// the event is signaled, and calls to WaitSynchronization with this event /// will immediately return. state: AtomicBool, /// List of processes waiting on this IRQ. When this IRQ is triggered, all /// those processes will be rescheduled. waiting_processes: SpinLock<Vec<Arc<ThreadStruct>>> } /// Create a new pair of [WritableEvent]/[ReadableEvent]. pub fn new_pair() -> (WritableEvent, ReadableEvent) { let event = Arc::new(Event { state: AtomicBool::new(false), waiting_processes: SpinLock::new(Vec::new()) }); (WritableEvent { parent: event.clone() }, ReadableEvent { parent: event }) } /// The readable part of an event. The user shall use this end to verify if the /// event is signaled, and wait for the signaling through wait_synchronization. /// The user can also use this handle to clear the signaled state through /// [ReadableEvent::clear_signal()]. #[derive(Debug, Clone)] pub struct ReadableEvent { /// Pointer to the shared event representation. parent: Arc<Event> } impl ReadableEvent { /// Clears the signaled state. /// /// # Errors /// /// - `InvalidState` /// - The event wasn't signaled. pub fn clear_signal(&self) -> Result<(), KernelError> { let oldstate = self.parent.state.swap(false, Ordering::SeqCst); if !oldstate { return Err(KernelError::InvalidState { backtrace: Backtrace::new() }) } Ok(()) } } impl Waitable for ReadableEvent { fn is_signaled(&self) -> bool { self.parent.state.load(Ordering::SeqCst) } fn register(&self) { self.parent.waiting_processes.lock().push(scheduler::get_current_thread()); } } /// The writable part of an event. The user shall use this end to signal (and /// wake up threads waiting on the event). #[derive(Debug, Clone)] pub struct WritableEvent { /// Pointer to the shared event representation. parent: Arc<Event> } impl WritableEvent { /// Signals the event, setting its state to signaled and waking up any /// thread waiting on its value. pub fn signal(&self) { self.parent.state.store(true, Ordering::SeqCst); let mut processes = self.parent.waiting_processes.lock(); while let Some(process) = processes.pop() { scheduler::add_to_schedule_queue(process); } } /// Clears the signaled state. /// /// # Errors /// /// - `InvalidState` /// - The event wasn't signaled. pub fn clear_signal(&self) -> Result<(), KernelError> { let oldstate = self.parent.state.swap(false, Ordering::SeqCst); if !oldstate { return Err(KernelError::InvalidState { backtrace: Backtrace::new() }) } Ok(()) } } /// An event waiting for an IRQ. /// /// When created, is_signaled is called and the IRQ was triggered, it will /// increment the ACK count by 1. This means that if multiple IRQs happened /// between wait calls, it will immediately return true. // TODO: Allow configuring edge vs level triggering. #[derive(Debug)] pub struct IRQEvent { /// The global state of the IRQ this event is listening on. /// Contains the IRQ trigger count. state: &'static IRQState, /// Acknowledgement counter for this IRQEvent instance. Each time we get /// signaled, this counter is incremented until it matches the counter in /// state. ack: AtomicUsize, } impl Waitable for IRQEvent { fn is_signaled(&self) -> bool { self.ack.fetch_update(|x| { if x < self.state.counter.load(Ordering::SeqCst) { // TODO: If level-triggered, set this to the counter. Some(x + 1) } else { None } }, Ordering::SeqCst, Ordering::SeqCst) .is_ok() } fn register(&self) { let curproc = scheduler::get_current_thread(); let mut veclock = self.state.waiting_processes.lock(); debug!("Registering {:010x} for irq {}", &*curproc as *const _ as usize, self.state.irqnum); if veclock.iter().find(|v| Arc::ptr_eq(&curproc, v)).is_none() { veclock.push(scheduler::get_current_thread()); } } } /// Signal the scheduler and waiters that an IRQ has been triggered. /// /// Usually, the IRQ handling code calls this. But it may be used to generate /// synthetic IRQs. pub fn dispatch_event(irq: usize) { IRQ_STATES[irq].counter.fetch_add(1, Ordering::SeqCst); let mut processes = IRQ_STATES[irq].waiting_processes.lock(); while let Some(process) = processes.pop() { scheduler::add_to_schedule_queue(process); } } /// Creates an IRQEvent waiting for the given IRQ number. pub fn wait_event(irq: u8) -> IRQEvent { debug!("Waiting for {}", irq); crate::i386::interrupt::unmask(irq); IRQEvent { state: &IRQ_STATES[irq as usize], ack: AtomicUsize::new(IRQ_STATES[irq as usize].counter.load(Ordering::SeqCst)) } } /// Global state of an IRQ. /// /// Counts the number of times this IRQ was triggered from kernel boot. #[derive(Debug)] struct IRQState { /// The irq number this state represents. Only used for debug logs. irqnum: usize, /// The number of time this IRQ was triggered from kernel boot. counter: AtomicUsize, /// List of processes waiting on this IRQ. When this IRQ is triggered, all /// those processes will be rescheduled. waiting_processes: SpinLockIRQ<Vec<Arc<ThreadStruct>>> } impl IRQState { /// Create a new IRQState for the given IRQ number, with the counter set to /// 0. pub const fn new(irqnum: usize) -> IRQState { IRQState { irqnum, counter: AtomicUsize::new(0), waiting_processes: SpinLockIRQ::new(Vec::new()) } } } /// Global state for all the IRQ handled by the IOAPIC. static IRQ_STATES: [IRQState; 17] = [ IRQState::new(0x20), IRQState::new(0x21), IRQState::new(0x22), IRQState::new(0x23), IRQState::new(0x24), IRQState::new(0x25), IRQState::new(0x26), IRQState::new(0x27), IRQState::new(0x28), IRQState::new(0x29), IRQState::new(0x2A), IRQState::new(0x2B), IRQState::new(0x2C), IRQState::new(0x2D), IRQState::new(0x2E), IRQState::new(0x2F), IRQState::new(0x30), ];