libs/corosio/src/corosio/src/detail/epoll/op.hpp

84.8% Lines (95/112) 81.0% Functions (17/21) 77.8% Branches (21/27)
libs/corosio/src/corosio/src/detail/epoll/op.hpp
Line Branch 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_DETAIL_EPOLL_OP_HPP
11 #define BOOST_COROSIO_DETAIL_EPOLL_OP_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/corosio/io_object.hpp>
19 #include <boost/corosio/endpoint.hpp>
20 #include <boost/capy/ex/executor_ref.hpp>
21 #include <boost/capy/coro.hpp>
22 #include <boost/capy/error.hpp>
23 #include <system_error>
24
25 #include "src/detail/make_err.hpp"
26 #include "src/detail/resume_coro.hpp"
27 #include "src/detail/scheduler_op.hpp"
28 #include "src/detail/endpoint_convert.hpp"
29
30 #include <unistd.h>
31 #include <errno.h>
32
33 #include <atomic>
34 #include <cstddef>
35 #include <memory>
36 #include <mutex>
37 #include <optional>
38 #include <stop_token>
39
40 #include <netinet/in.h>
41 #include <sys/socket.h>
42 #include <sys/uio.h>
43
44 /*
45 epoll Operation State
46 =====================
47
48 Each async I/O operation has a corresponding epoll_op-derived struct that
49 holds the operation's state while it's in flight. The socket impl owns
50 fixed slots for each operation type (conn_, rd_, wr_), so only one
51 operation of each type can be pending per socket at a time.
52
53 Persistent Registration
54 -----------------------
55 File descriptors are registered with epoll once (via descriptor_state) and
56 stay registered until closed. The descriptor_state tracks which operations
57 are pending (read_op, write_op, connect_op). When an event arrives, the
58 reactor dispatches to the appropriate pending operation.
59
60 Impl Lifetime Management
61 ------------------------
62 When cancel() posts an op to the scheduler's ready queue, the socket impl
63 might be destroyed before the scheduler processes the op. The `impl_ptr`
64 member holds a shared_ptr to the impl, keeping it alive until the op
65 completes. This is set by cancel() and cleared in operator() after the
66 coroutine is resumed.
67
68 EOF Detection
69 -------------
70 For reads, 0 bytes with no error means EOF. But an empty user buffer also
71 returns 0 bytes. The `empty_buffer_read` flag distinguishes these cases.
72
73 SIGPIPE Prevention
74 ------------------
75 Writes use sendmsg() with MSG_NOSIGNAL instead of writev() to prevent
76 SIGPIPE when the peer has closed.
77 */
78
79 namespace boost::corosio::detail {
80
81 // Forward declarations
82 class epoll_socket_impl;
83 class epoll_acceptor_impl;
84 struct epoll_op;
85
86 // Forward declaration
87 class epoll_scheduler;
88
89 /** Per-descriptor state for persistent epoll registration.
90
91 Tracks pending operations for a file descriptor. The fd is registered
92 once with epoll and stays registered until closed.
93
94 This struct extends scheduler_op to support deferred I/O processing.
95 When epoll events arrive, the reactor sets ready_events and queues
96 this descriptor for processing. When popped from the scheduler queue,
97 operator() performs the actual I/O and queues completion handlers.
98
99 @par Deferred I/O Model
100 The reactor no longer performs I/O directly. Instead:
101 1. Reactor sets ready_events and queues descriptor_state
102 2. Scheduler pops descriptor_state and calls operator()
103 3. operator() performs I/O under mutex and queues completions
104
105 This eliminates per-descriptor mutex locking from the reactor hot path.
106
107 @par Thread Safety
108 The mutex protects operation pointers and ready flags during I/O.
109 ready_events_ and is_enqueued_ are atomic for lock-free reactor access.
110 */
111 struct descriptor_state : scheduler_op
112 {
113 std::mutex mutex;
114
115 // Protected by mutex
116 epoll_op* read_op = nullptr;
117 epoll_op* write_op = nullptr;
118 epoll_op* connect_op = nullptr;
119
120 // Caches edge events that arrived before an op was registered
121 bool read_ready = false;
122 bool write_ready = false;
123
124 // Set during registration only (no mutex needed)
125 std::uint32_t registered_events = 0;
126 int fd = -1;
127
128 // For deferred I/O - set by reactor, read by scheduler
129 std::atomic<std::uint32_t> ready_events_{0};
130 std::atomic<bool> is_enqueued_{false};
131 epoll_scheduler const* scheduler_ = nullptr;
132
133 // Prevents impl destruction while this descriptor_state is queued.
134 // Set by close_socket() when is_enqueued_ is true, cleared by operator().
135 std::shared_ptr<void> impl_ref_;
136
137 /// Add ready events atomically.
138 125169 void add_ready_events(std::uint32_t ev) noexcept
139 {
140 125169 ready_events_.fetch_or(ev, std::memory_order_relaxed);
141 125169 }
142
143 /// Perform deferred I/O and queue completions.
144 void operator()() override;
145
146 /// Destroy without invoking.
147 void destroy() override {}
148 };
149
150 struct epoll_op : scheduler_op
151 {
152 struct canceller
153 {
154 epoll_op* op;
155 void operator()() const noexcept;
156 };
157
158 capy::coro h;
159 capy::executor_ref ex;
160 std::error_code* ec_out = nullptr;
161 std::size_t* bytes_out = nullptr;
162
163 int fd = -1;
164 int errn = 0;
165 std::size_t bytes_transferred = 0;
166
167 std::atomic<bool> cancelled{false};
168 std::optional<std::stop_callback<canceller>> stop_cb;
169
170 // Prevents use-after-free when socket is closed with pending ops.
171 // See "Impl Lifetime Management" in file header.
172 std::shared_ptr<void> impl_ptr;
173
174 // For stop_token cancellation - pointer to owning socket/acceptor impl.
175 // When stop is requested, we call back to the impl to perform actual I/O cancellation.
176 epoll_socket_impl* socket_impl_ = nullptr;
177 epoll_acceptor_impl* acceptor_impl_ = nullptr;
178
179 30210 epoll_op() = default;
180
181 240281 void reset() noexcept
182 {
183 240281 fd = -1;
184 240281 errn = 0;
185 240281 bytes_transferred = 0;
186 240281 cancelled.store(false, std::memory_order_relaxed);
187 240281 impl_ptr.reset();
188 240281 socket_impl_ = nullptr;
189 240281 acceptor_impl_ = nullptr;
190 240281 }
191
192 230237 void operator()() override
193 {
194 230237 stop_cb.reset();
195
196
1/2
✓ Branch 0 taken 230237 times.
✗ Branch 1 not taken.
230237 if (ec_out)
197 {
198
2/2
✓ Branch 1 taken 206 times.
✓ Branch 2 taken 230031 times.
230237 if (cancelled.load(std::memory_order_acquire))
199 206 *ec_out = capy::error::canceled;
200
2/2
✓ Branch 0 taken 1 time.
✓ Branch 1 taken 230030 times.
230031 else if (errn != 0)
201 1 *ec_out = make_err(errn);
202
6/6
✓ Branch 1 taken 114977 times.
✓ Branch 2 taken 115053 times.
✓ Branch 3 taken 5 times.
✓ Branch 4 taken 114972 times.
✓ Branch 5 taken 5 times.
✓ Branch 6 taken 230025 times.
230030 else if (is_read_operation() && bytes_transferred == 0)
203 5 *ec_out = capy::error::eof;
204 else
205 230025 *ec_out = {};
206 }
207
208
1/2
✓ Branch 0 taken 230237 times.
✗ Branch 1 not taken.
230237 if (bytes_out)
209 230237 *bytes_out = bytes_transferred;
210
211 // Move to stack before resuming coroutine. The coroutine might close
212 // the socket, releasing the last wrapper ref. If impl_ptr were the
213 // last ref and we destroyed it while still in operator(), we'd have
214 // use-after-free. Moving to local ensures destruction happens at
215 // function exit, after all member accesses are complete.
216 230237 capy::executor_ref saved_ex( std::move( ex ) );
217 230237 capy::coro saved_h( std::move( h ) );
218 230237 auto prevent_premature_destruction = std::move(impl_ptr);
219
1/1
✓ Branch 1 taken 230237 times.
230237 resume_coro(saved_ex, saved_h);
220 230237 }
221
222 115052 virtual bool is_read_operation() const noexcept { return false; }
223 virtual void cancel() noexcept = 0;
224
225 void destroy() override
226 {
227 stop_cb.reset();
228 impl_ptr.reset();
229 }
230
231 45769 void request_cancel() noexcept
232 {
233 45769 cancelled.store(true, std::memory_order_release);
234 45769 }
235
236 235255 void start(std::stop_token token, epoll_socket_impl* impl)
237 {
238 235255 cancelled.store(false, std::memory_order_release);
239 235255 stop_cb.reset();
240 235255 socket_impl_ = impl;
241 235255 acceptor_impl_ = nullptr;
242
243
2/2
✓ Branch 1 taken 106 times.
✓ Branch 2 taken 235149 times.
235255 if (token.stop_possible())
244 106 stop_cb.emplace(token, canceller{this});
245 235255 }
246
247 5026 void start(std::stop_token token, epoll_acceptor_impl* impl)
248 {
249 5026 cancelled.store(false, std::memory_order_release);
250 5026 stop_cb.reset();
251 5026 socket_impl_ = nullptr;
252 5026 acceptor_impl_ = impl;
253
254
2/2
✓ Branch 1 taken 9 times.
✓ Branch 2 taken 5017 times.
5026 if (token.stop_possible())
255 9 stop_cb.emplace(token, canceller{this});
256 5026 }
257
258 240199 void complete(int err, std::size_t bytes) noexcept
259 {
260 240199 errn = err;
261 240199 bytes_transferred = bytes;
262 240199 }
263
264 virtual void perform_io() noexcept {}
265 };
266
267
268 struct epoll_connect_op : epoll_op
269 {
270 endpoint target_endpoint;
271
272 5018 void reset() noexcept
273 {
274 5018 epoll_op::reset();
275 5018 target_endpoint = endpoint{};
276 5018 }
277
278 5018 void perform_io() noexcept override
279 {
280 // connect() completion status is retrieved via SO_ERROR, not return value
281 5018 int err = 0;
282 5018 socklen_t len = sizeof(err);
283
1/2
✗ Branch 1 not taken.
✓ Branch 2 taken 5018 times.
5018 if (::getsockopt(fd, SOL_SOCKET, SO_ERROR, &err, &len) < 0)
284 err = errno;
285 5018 complete(err, 0);
286 5018 }
287
288 // Defined in sockets.cpp where epoll_socket_impl is complete
289 void operator()() override;
290 void cancel() noexcept override;
291 };
292
293
294 struct epoll_read_op : epoll_op
295 {
296 static constexpr std::size_t max_buffers = 16;
297 iovec iovecs[max_buffers];
298 int iovec_count = 0;
299 bool empty_buffer_read = false;
300
301 114978 bool is_read_operation() const noexcept override
302 {
303 114978 return !empty_buffer_read;
304 }
305
306 115180 void reset() noexcept
307 {
308 115180 epoll_op::reset();
309 115180 iovec_count = 0;
310 115180 empty_buffer_read = false;
311 115180 }
312
313 98 void perform_io() noexcept override
314 {
315 98 ssize_t n = ::readv(fd, iovecs, iovec_count);
316
2/2
✓ Branch 0 taken 52 times.
✓ Branch 1 taken 46 times.
98 if (n >= 0)
317 52 complete(0, static_cast<std::size_t>(n));
318 else
319 46 complete(errno, 0);
320 98 }
321
322 void cancel() noexcept override;
323 };
324
325
326 struct epoll_write_op : epoll_op
327 {
328 static constexpr std::size_t max_buffers = 16;
329 iovec iovecs[max_buffers];
330 int iovec_count = 0;
331
332 115057 void reset() noexcept
333 {
334 115057 epoll_op::reset();
335 115057 iovec_count = 0;
336 115057 }
337
338 void perform_io() noexcept override
339 {
340 msghdr msg{};
341 msg.msg_iov = iovecs;
342 msg.msg_iovlen = static_cast<std::size_t>(iovec_count);
343
344 ssize_t n = ::sendmsg(fd, &msg, MSG_NOSIGNAL);
345 if (n >= 0)
346 complete(0, static_cast<std::size_t>(n));
347 else
348 complete(errno, 0);
349 }
350
351 void cancel() noexcept override;
352 };
353
354
355 struct epoll_accept_op : epoll_op
356 {
357 int accepted_fd = -1;
358 io_object::io_object_impl* peer_impl = nullptr;
359 io_object::io_object_impl** impl_out = nullptr;
360
361 5026 void reset() noexcept
362 {
363 5026 epoll_op::reset();
364 5026 accepted_fd = -1;
365 5026 peer_impl = nullptr;
366 5026 impl_out = nullptr;
367 5026 }
368
369 5015 void perform_io() noexcept override
370 {
371 5015 sockaddr_in addr{};
372 5015 socklen_t addrlen = sizeof(addr);
373 5015 int new_fd = ::accept4(fd, reinterpret_cast<sockaddr*>(&addr),
374 &addrlen, SOCK_NONBLOCK | SOCK_CLOEXEC);
375
376
1/2
✓ Branch 0 taken 5015 times.
✗ Branch 1 not taken.
5015 if (new_fd >= 0)
377 {
378 5015 accepted_fd = new_fd;
379 5015 complete(0, 0);
380 }
381 else
382 {
383 complete(errno, 0);
384 }
385 5015 }
386
387 // Defined in acceptors.cpp where epoll_acceptor_impl is complete
388 void operator()() override;
389 void cancel() noexcept override;
390 };
391
392 } // namespace boost::corosio::detail
393
394 #endif // BOOST_COROSIO_HAS_EPOLL
395
396 #endif // BOOST_COROSIO_DETAIL_EPOLL_OP_HPP
397