include/boost/corosio/native/detail/epoll/epoll_scheduler.hpp

82.7% Lines (124/150) 90.0% List of functions (9/10)

[ ] prev/next file n p prev/next uncovered line

epoll_scheduler.hpp

f(x) Functions (10)

Function Calls Lines Blocks

Line	Hits	Source Code
1		//
2		// Copyright (c) 2026 Steve Gerbino
3		//
4		// Distributed under the Boost Software License, Version 1.0. (See accompanying
5		// file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
6		//
7		// Official repository: https://github.com/cppalliance/corosio
8		//
9
10		#ifndef BOOST_COROSIO_NATIVE_DETAIL_EPOLL_EPOLL_SCHEDULER_HPP
11		#define BOOST_COROSIO_NATIVE_DETAIL_EPOLL_EPOLL_SCHEDULER_HPP
12
13		#include <boost/corosio/detail/platform.hpp>
14
15		#if BOOST_COROSIO_HAS_EPOLL
16
17		#include <boost/corosio/detail/config.hpp>
18		#include <boost/capy/ex/execution_context.hpp>
19
20		#include <boost/corosio/native/detail/reactor/reactor_scheduler.hpp>
21
22		#include <boost/corosio/native/detail/epoll/epoll_traits.hpp>
23		#include <boost/corosio/detail/timer_service.hpp>
24		#include <boost/corosio/native/detail/make_err.hpp>
25		#include <boost/corosio/native/detail/posix/posix_resolver_service.hpp>
26		#include <boost/corosio/native/detail/posix/posix_signal_service.hpp>
27		#include <boost/corosio/native/detail/posix/posix_stream_file_service.hpp>
28		#include <boost/corosio/native/detail/posix/posix_random_access_file_service.hpp>
29
30		#include <boost/corosio/detail/except.hpp>
31
32		#include <atomic>
33		#include <chrono>
34		#include <cstdint>
35		#include <mutex>
36		#include <vector>
37
38		#include <errno.h>
39		#include <sys/epoll.h>
40		#include <sys/eventfd.h>
41		#include <sys/timerfd.h>
42		#include <unistd.h>
43
44		namespace boost::corosio::detail {
45
46		/** Linux scheduler using epoll for I/O multiplexing.
47
48		This scheduler implements the scheduler interface using Linux epoll
49		for efficient I/O event notification. It uses a single reactor model
50		where one thread runs epoll_wait while other threads
51		wait on a condition variable for handler work. This design provides:
52
53		- Handler parallelism: N posted handlers can execute on N threads
54		- No thundering herd: condition_variable wakes exactly one thread
55		- IOCP parity: Behavior matches Windows I/O completion port semantics
56
57		When threads call run(), they first try to execute queued handlers.
58		If the queue is empty and no reactor is running, one thread becomes
59		the reactor and runs epoll_wait. Other threads wait on a condition
60		variable until handlers are available.
61
62		@par Thread Safety
63		All public member functions are thread-safe.
64		*/
65		class BOOST_COROSIO_DECL epoll_scheduler final : public reactor_scheduler
66		{
67		public:
68		/** Construct the scheduler.
69
70		Creates an epoll instance, eventfd for reactor interruption,
71		and timerfd for kernel-managed timer expiry.
72
73		@param ctx Reference to the owning execution_context.
74		@param concurrency_hint Hint for expected thread count (unused).
75		*/
76		epoll_scheduler(capy::execution_context& ctx, int concurrency_hint = -1);
77
78		/// Destroy the scheduler.
79		~epoll_scheduler() override;
80
81		epoll_scheduler(epoll_scheduler const&) = delete;
82		epoll_scheduler& operator=(epoll_scheduler const&) = delete;
83
84		/// Shut down the scheduler, draining pending operations.
85		void shutdown() override;
86
87		/// Apply runtime configuration, resizing the event buffer.
88		void configure_reactor(
89		unsigned max_events,
90		unsigned budget_init,
91		unsigned budget_max,
92		unsigned unassisted) override;
93
94		/** Return the epoll file descriptor.
95
96		Used by socket services to register file descriptors
97		for I/O event notification.
98
99		@return The epoll file descriptor.
100		*/
101		int epoll_fd() const noexcept
102		{
103		return epoll_fd_;
104		}
105
106		/** Register a descriptor for persistent monitoring.
107
108		The fd is registered once and stays registered until explicitly
109		deregistered. Events are dispatched via reactor_descriptor_state which
110		tracks pending read/write/connect operations.
111
112		@param fd The file descriptor to register.
113		@param desc Pointer to descriptor data (stored in epoll_event.data.ptr).
114		*/
115		void register_descriptor(int fd, reactor_descriptor_state* desc) const;
116
117		/** Deregister a persistently registered descriptor.
118
119		@param fd The file descriptor to deregister.
120		*/
121		void deregister_descriptor(int fd) const;
122
123		private:
124		void
125		run_task(lock_type& lock, context_type* ctx,
126		long timeout_us) override;
127		void interrupt_reactor() const override;
128		void update_timerfd() const;
129
130		int epoll_fd_;
131		int event_fd_;
132		int timer_fd_;
133
134		// Edge-triggered eventfd state
135		mutable std::atomic<bool> eventfd_armed_{false};
136
137		// Set when the earliest timer changes; flushed before epoll_wait
138		mutable std::atomic<bool> timerfd_stale_{false};
139
140		// Event buffer sized from max_events_per_poll_ (set at construction,
141		// resized by configure_reactor via io_context_options).
142		std::vector<epoll_event> event_buffer_;
143		};
144
145	366x	inline epoll_scheduler::epoll_scheduler(capy::execution_context& ctx, int)
146	366x	: epoll_fd_(-1)
147	366x	, event_fd_(-1)
148	366x	, timer_fd_(-1)
149	732x	, event_buffer_(max_events_per_poll_)
150		{
151	366x	epoll_fd_ = ::epoll_create1(EPOLL_CLOEXEC);
152	366x	if (epoll_fd_ < 0)
153	✗	detail::throw_system_error(make_err(errno), "epoll_create1");
154
155	366x	event_fd_ = ::eventfd(0, EFD_NONBLOCK \| EFD_CLOEXEC);
156	366x	if (event_fd_ < 0)
157		{
158	✗	int errn = errno;
159	✗	::close(epoll_fd_);
160	✗	detail::throw_system_error(make_err(errn), "eventfd");
161		}
162
163	366x	timer_fd_ = ::timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK \| TFD_CLOEXEC);
164	366x	if (timer_fd_ < 0)
165		{
166	✗	int errn = errno;
167	✗	::close(event_fd_);
168	✗	::close(epoll_fd_);
169	✗	detail::throw_system_error(make_err(errn), "timerfd_create");
170		}
171
172	366x	epoll_event ev{};
173	366x	ev.events = EPOLLIN \| EPOLLET;
174	366x	ev.data.ptr = nullptr;
175	366x	if (::epoll_ctl(epoll_fd_, EPOLL_CTL_ADD, event_fd_, &ev) < 0)
176		{
177	✗	int errn = errno;
178	✗	::close(timer_fd_);
179	✗	::close(event_fd_);
180	✗	::close(epoll_fd_);
181	✗	detail::throw_system_error(make_err(errn), "epoll_ctl");
182		}
183
184	366x	epoll_event timer_ev{};
185	366x	timer_ev.events = EPOLLIN \| EPOLLERR;
186	366x	timer_ev.data.ptr = &timer_fd_;
187	366x	if (::epoll_ctl(epoll_fd_, EPOLL_CTL_ADD, timer_fd_, &timer_ev) < 0)
188		{
189	✗	int errn = errno;
190	✗	::close(timer_fd_);
191	✗	::close(event_fd_);
192	✗	::close(epoll_fd_);
193	✗	detail::throw_system_error(make_err(errn), "epoll_ctl (timerfd)");
194		}
195
196	366x	timer_svc_ = &get_timer_service(ctx, *this);
197	366x	timer_svc_->set_on_earliest_changed(
198	5569x	timer_service::callback(this, [](void* p) {
199	5203x	auto* self = static_cast<epoll_scheduler*>(p);
200	5203x	self->timerfd_stale_.store(true, std::memory_order_release);
201	5203x	self->interrupt_reactor();
202	5203x	}));
203
204	366x	get_resolver_service(ctx, *this);
205	366x	get_signal_service(ctx, *this);
206	366x	get_stream_file_service(ctx, *this);
207	366x	get_random_access_file_service(ctx, *this);
208
209	366x	completed_ops_.push(&task_op_);
210	366x	}
211
212	732x	inline epoll_scheduler::~epoll_scheduler()
213		{
214	366x	if (timer_fd_ >= 0)
215	366x	::close(timer_fd_);
216	366x	if (event_fd_ >= 0)
217	366x	::close(event_fd_);
218	366x	if (epoll_fd_ >= 0)
219	366x	::close(epoll_fd_);
220	732x	}
221
222		inline void
223	366x	epoll_scheduler::shutdown()
224		{
225	366x	shutdown_drain();
226
227	366x	if (event_fd_ >= 0)
228	366x	interrupt_reactor();
229	366x	}
230
231		inline void
232	✗	epoll_scheduler::configure_reactor(
233		unsigned max_events,
234		unsigned budget_init,
235		unsigned budget_max,
236		unsigned unassisted)
237		{
238	✗	reactor_scheduler::configure_reactor(
239		max_events, budget_init, budget_max, unassisted);
240	✗	event_buffer_.resize(max_events_per_poll_);
241	✗	}
242
243		inline void
244	10128x	epoll_scheduler::register_descriptor(int fd, reactor_descriptor_state* desc) const
245		{
246	10128x	epoll_event ev{};
247	10128x	ev.events = EPOLLIN \| EPOLLOUT \| EPOLLET \| EPOLLERR \| EPOLLHUP;
248	10128x	ev.data.ptr = desc;
249
250	10128x	if (::epoll_ctl(epoll_fd_, EPOLL_CTL_ADD, fd, &ev) < 0)
251	✗	detail::throw_system_error(make_err(errno), "epoll_ctl (register)");
252
253	10128x	desc->registered_events = ev.events;
254	10128x	desc->fd = fd;
255	10128x	desc->scheduler_ = this;
256	10128x	desc->mutex.set_enabled(!single_threaded_);
257	10128x	desc->ready_events_.store(0, std::memory_order_relaxed);
258
259	10128x	conditionally_enabled_mutex::scoped_lock lock(desc->mutex);
260	10128x	desc->impl_ref_.reset();
261	10128x	desc->read_ready = false;
262	10128x	desc->write_ready = false;
263	10128x	}
264
265		inline void
266	10128x	epoll_scheduler::deregister_descriptor(int fd) const
267		{
268	10128x	::epoll_ctl(epoll_fd_, EPOLL_CTL_DEL, fd, nullptr);
269	10128x	}
270
271		inline void
272	5878x	epoll_scheduler::interrupt_reactor() const
273		{
274	5878x	bool expected = false;
275	5878x	if (eventfd_armed_.compare_exchange_strong(
276		expected, true, std::memory_order_release,
277		std::memory_order_relaxed))
278		{
279	5646x	std::uint64_t val = 1;
280	5646x	[[maybe_unused]] auto r = ::write(event_fd_, &val, sizeof(val));
281		}
282	5878x	}
283
284		inline void
285	10364x	epoll_scheduler::update_timerfd() const
286		{
287	10364x	auto nearest = timer_svc_->nearest_expiry();
288
289	10364x	itimerspec ts{};
290	10364x	int flags = 0;
291
292	10364x	if (nearest == timer_service::time_point::max())
293		{
294		// No timers — disarm by setting to 0 (relative)
295		}
296		else
297		{
298	10308x	auto now = std::chrono::steady_clock::now();
299	10308x	if (nearest <= now)
300		{
301		// Use 1ns instead of 0 — zero disarms the timerfd
302	206x	ts.it_value.tv_nsec = 1;
303		}
304		else
305		{
306	10102x	auto nsec = std::chrono::duration_cast<std::chrono::nanoseconds>(
307	10102x	nearest - now)
308	10102x	.count();
309	10102x	ts.it_value.tv_sec = nsec / 1000000000;
310	10102x	ts.it_value.tv_nsec = nsec % 1000000000;
311	10102x	if (ts.it_value.tv_sec == 0 && ts.it_value.tv_nsec == 0)
312	✗	ts.it_value.tv_nsec = 1;
313		}
314		}
315
316	10364x	if (::timerfd_settime(timer_fd_, flags, &ts, nullptr) < 0)
317	✗	detail::throw_system_error(make_err(errno), "timerfd_settime");
318	10364x	}
319
320		inline void
321	39890x	epoll_scheduler::run_task(
322		lock_type& lock, context_type* ctx, long timeout_us)
323		{
324		int timeout_ms;
325	39890x	if (task_interrupted_)
326	24970x	timeout_ms = 0;
327	14920x	else if (timeout_us < 0)
328	14912x	timeout_ms = -1;
329		else
330	8x	timeout_ms = static_cast<int>((timeout_us + 999) / 1000);
331
332	39890x	if (lock.owns_lock())
333	14920x	lock.unlock();
334
335	39890x	task_cleanup on_exit{this, &lock, ctx};
336
337		// Flush deferred timerfd programming before blocking
338	39890x	if (timerfd_stale_.exchange(false, std::memory_order_acquire))
339	5181x	update_timerfd();
340
341	39890x	int nfds = ::epoll_wait(
342		epoll_fd_, event_buffer_.data(),
343	39890x	static_cast<int>(event_buffer_.size()), timeout_ms);
344
345	39890x	if (nfds < 0 && errno != EINTR)
346	✗	detail::throw_system_error(make_err(errno), "epoll_wait");
347
348	39890x	bool check_timers = false;
349	39890x	op_queue local_ops;
350
351	88903x	for (int i = 0; i < nfds; ++i)
352		{
353	49013x	if (event_buffer_[i].data.ptr == nullptr)
354		{
355		std::uint64_t val;
356		// NOLINTNEXTLINE(clang-analyzer-unix.BlockInCriticalSection)
357	5280x	[[maybe_unused]] auto r = ::read(event_fd_, &val, sizeof(val));
358	5280x	eventfd_armed_.store(false, std::memory_order_relaxed);
359	5280x	continue;
360	5280x	}
361
362	43733x	if (event_buffer_[i].data.ptr == &timer_fd_)
363		{
364		std::uint64_t expirations;
365		// NOLINTNEXTLINE(clang-analyzer-unix.BlockInCriticalSection)
366		[[maybe_unused]] auto r =
367	5183x	::read(timer_fd_, &expirations, sizeof(expirations));
368	5183x	check_timers = true;
369	5183x	continue;
370	5183x	}
371
372		auto* desc =
373	38550x	static_cast<reactor_descriptor_state*>(event_buffer_[i].data.ptr);
374	38550x	desc->add_ready_events(event_buffer_[i].events);
375
376	38550x	bool expected = false;
377	38550x	if (desc->is_enqueued_.compare_exchange_strong(
378		expected, true, std::memory_order_release,
379		std::memory_order_relaxed))
380		{
381	38550x	local_ops.push(desc);
382		}
383		}
384
385	39890x	if (check_timers)
386		{
387	5183x	timer_svc_->process_expired();
388	5183x	update_timerfd();
389		}
390
391	39890x	lock.lock();
392
393	39890x	if (!local_ops.empty())
394	24426x	completed_ops_.splice(local_ops);
395	39890x	}
396
397		} // namespace boost::corosio::detail
398
399		#endif // BOOST_COROSIO_HAS_EPOLL
400
401		#endif // BOOST_COROSIO_NATIVE_DETAIL_EPOLL_EPOLL_SCHEDULER_HPP
402