]>
Commit | Line | Data |
---|---|---|
11fdf7f2 TL |
1 | /* |
2 | * This file is open source software, licensed to you under the terms | |
3 | * of the Apache License, Version 2.0 (the "License"). See the NOTICE file | |
4 | * distributed with this work for additional information regarding copyright | |
5 | * ownership. You may not use this file except in compliance with the License. | |
6 | * | |
7 | * You may obtain a copy of the License at | |
8 | * | |
9 | * http://www.apache.org/licenses/LICENSE-2.0 | |
10 | * | |
11 | * Unless required by applicable law or agreed to in writing, | |
12 | * software distributed under the License is distributed on an | |
13 | * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY | |
14 | * KIND, either express or implied. See the License for the | |
15 | * specific language governing permissions and limitations | |
16 | * under the License. | |
17 | */ | |
18 | /* | |
19 | * Copyright (C) 2014 Cloudius Systems, Ltd. | |
20 | */ | |
21 | ||
22 | #pragma once | |
23 | ||
24 | #include <memory> | |
25 | #include <vector> | |
26 | #include <cstring> | |
27 | #include <seastar/core/future.hh> | |
28 | #include <seastar/net/byteorder.hh> | |
29 | #include <seastar/net/socket_defs.hh> | |
30 | #include <seastar/net/packet.hh> | |
31 | #include <seastar/core/print.hh> | |
32 | #include <seastar/core/temporary_buffer.hh> | |
33 | #include <seastar/core/iostream.hh> | |
34 | #include <seastar/util/std-compat.hh> | |
35 | #include <sys/types.h> | |
36 | ||
37 | namespace seastar { | |
38 | ||
39 | static inline | |
40 | bool is_ip_unspecified(ipv4_addr &addr) { | |
41 | return addr.ip == 0; | |
42 | } | |
43 | ||
44 | static inline | |
45 | bool is_port_unspecified(ipv4_addr &addr) { | |
46 | return addr.port == 0; | |
47 | } | |
48 | ||
49 | static inline | |
50 | std::ostream& operator<<(std::ostream &os, ipv4_addr addr) { | |
51 | fmt_print(os, "{:d}.{:d}.{:d}.{:d}", | |
52 | (addr.ip >> 24) & 0xff, | |
53 | (addr.ip >> 16) & 0xff, | |
54 | (addr.ip >> 8) & 0xff, | |
55 | (addr.ip) & 0xff); | |
56 | return os << ":" << addr.port; | |
57 | } | |
58 | ||
59 | static inline | |
60 | socket_address make_ipv4_address(ipv4_addr addr) { | |
61 | socket_address sa; | |
62 | sa.u.in.sin_family = AF_INET; | |
63 | sa.u.in.sin_port = htons(addr.port); | |
64 | sa.u.in.sin_addr.s_addr = htonl(addr.ip); | |
65 | return sa; | |
66 | } | |
67 | ||
68 | inline | |
69 | socket_address make_ipv4_address(uint32_t ip, uint16_t port) { | |
70 | socket_address sa; | |
71 | sa.u.in.sin_family = AF_INET; | |
72 | sa.u.in.sin_port = htons(port); | |
73 | sa.u.in.sin_addr.s_addr = htonl(ip); | |
74 | return sa; | |
75 | } | |
76 | ||
77 | namespace net { | |
78 | ||
79 | // see linux tcp(7) for parameter explanation | |
80 | struct tcp_keepalive_params { | |
81 | std::chrono::seconds idle; // TCP_KEEPIDLE | |
82 | std::chrono::seconds interval; // TCP_KEEPINTVL | |
83 | unsigned count; // TCP_KEEPCNT | |
84 | }; | |
85 | ||
86 | // see linux sctp(7) for parameter explanation | |
87 | struct sctp_keepalive_params { | |
88 | std::chrono::seconds interval; // spp_hbinterval | |
89 | unsigned count; // spp_pathmaxrt | |
90 | }; | |
91 | ||
92 | using keepalive_params = compat::variant<tcp_keepalive_params, sctp_keepalive_params>; | |
93 | ||
94 | /// \cond internal | |
95 | class connected_socket_impl; | |
96 | class socket_impl; | |
97 | class server_socket_impl; | |
98 | class udp_channel_impl; | |
99 | class get_impl; | |
100 | /// \endcond | |
101 | ||
102 | class udp_datagram_impl { | |
103 | public: | |
104 | virtual ~udp_datagram_impl() {}; | |
105 | virtual ipv4_addr get_src() = 0; | |
106 | virtual ipv4_addr get_dst() = 0; | |
107 | virtual uint16_t get_dst_port() = 0; | |
108 | virtual packet& get_data() = 0; | |
109 | }; | |
110 | ||
111 | class udp_datagram final { | |
112 | private: | |
113 | std::unique_ptr<udp_datagram_impl> _impl; | |
114 | public: | |
115 | udp_datagram(std::unique_ptr<udp_datagram_impl>&& impl) : _impl(std::move(impl)) {}; | |
116 | ipv4_addr get_src() { return _impl->get_src(); } | |
117 | ipv4_addr get_dst() { return _impl->get_dst(); } | |
118 | uint16_t get_dst_port() { return _impl->get_dst_port(); } | |
119 | packet& get_data() { return _impl->get_data(); } | |
120 | }; | |
121 | ||
122 | class udp_channel { | |
123 | private: | |
124 | std::unique_ptr<udp_channel_impl> _impl; | |
125 | public: | |
126 | udp_channel(); | |
127 | udp_channel(std::unique_ptr<udp_channel_impl>); | |
128 | ~udp_channel(); | |
129 | ||
130 | udp_channel(udp_channel&&); | |
131 | udp_channel& operator=(udp_channel&&); | |
132 | ||
133 | future<udp_datagram> receive(); | |
134 | future<> send(ipv4_addr dst, const char* msg); | |
135 | future<> send(ipv4_addr dst, packet p); | |
136 | bool is_closed() const; | |
137 | /// Causes a pending receive() to complete (possibly with an exception) | |
138 | void shutdown_input(); | |
139 | /// Causes a pending send() to complete (possibly with an exception) | |
140 | void shutdown_output(); | |
141 | /// Close the channel and releases all resources. | |
142 | /// | |
143 | /// Must be called only when there are no unfinished send() or receive() calls. You | |
144 | /// can force pending calls to complete soon by calling shutdown_input() and | |
145 | /// shutdown_output(). | |
146 | void close(); | |
147 | }; | |
148 | ||
149 | } /* namespace net */ | |
150 | ||
151 | /// \addtogroup networking-module | |
152 | /// @{ | |
153 | ||
154 | /// A TCP (or other stream-based protocol) connection. | |
155 | /// | |
156 | /// A \c connected_socket represents a full-duplex stream between | |
157 | /// two endpoints, a local endpoint and a remote endpoint. | |
158 | class connected_socket { | |
159 | friend class net::get_impl; | |
160 | std::unique_ptr<net::connected_socket_impl> _csi; | |
161 | public: | |
162 | /// Constructs a \c connected_socket not corresponding to a connection | |
163 | connected_socket(); | |
164 | ~connected_socket(); | |
165 | ||
166 | /// \cond internal | |
167 | explicit connected_socket(std::unique_ptr<net::connected_socket_impl> csi); | |
168 | /// \endcond | |
169 | /// Moves a \c connected_socket object. | |
170 | connected_socket(connected_socket&& cs) noexcept; | |
171 | /// Move-assigns a \c connected_socket object. | |
172 | connected_socket& operator=(connected_socket&& cs) noexcept; | |
173 | /// Gets the input stream. | |
174 | /// | |
175 | /// Gets an object returning data sent from the remote endpoint. | |
176 | input_stream<char> input(); | |
177 | /// Gets the output stream. | |
178 | /// | |
179 | /// Gets an object that sends data to the remote endpoint. | |
180 | /// \param buffer_size how much data to buffer | |
181 | output_stream<char> output(size_t buffer_size = 8192); | |
182 | /// Sets the TCP_NODELAY option (disabling Nagle's algorithm) | |
183 | void set_nodelay(bool nodelay); | |
184 | /// Gets the TCP_NODELAY option (Nagle's algorithm) | |
185 | /// | |
186 | /// \return whether the nodelay option is enabled or not | |
187 | bool get_nodelay() const; | |
188 | /// Sets SO_KEEPALIVE option (enable keepalive timer on a socket) | |
189 | void set_keepalive(bool keepalive); | |
190 | /// Gets O_KEEPALIVE option | |
191 | /// \return whether the keepalive option is enabled or not | |
192 | bool get_keepalive() const; | |
193 | /// Sets TCP keepalive parameters | |
194 | void set_keepalive_parameters(const net::keepalive_params& p); | |
195 | /// Get TCP keepalive parameters | |
196 | net::keepalive_params get_keepalive_parameters() const; | |
197 | ||
198 | /// Disables output to the socket. | |
199 | /// | |
200 | /// Current or future writes that have not been successfully flushed | |
201 | /// will immediately fail with an error. This is useful to abort | |
202 | /// operations on a socket that is not making progress due to a | |
203 | /// peer failure. | |
204 | void shutdown_output(); | |
205 | /// Disables input from the socket. | |
206 | /// | |
207 | /// Current or future reads will immediately fail with an error. | |
208 | /// This is useful to abort operations on a socket that is not making | |
209 | /// progress due to a peer failure. | |
210 | void shutdown_input(); | |
211 | }; | |
212 | /// @} | |
213 | ||
214 | /// \addtogroup networking-module | |
215 | /// @{ | |
216 | ||
217 | /// The seastar socket. | |
218 | /// | |
219 | /// A \c socket that allows a connection to be established between | |
220 | /// two endpoints. | |
221 | class socket { | |
222 | std::unique_ptr<net::socket_impl> _si; | |
223 | public: | |
224 | ~socket(); | |
225 | ||
226 | /// \cond internal | |
227 | explicit socket(std::unique_ptr<net::socket_impl> si); | |
228 | /// \endcond | |
229 | /// Moves a \c seastar::socket object. | |
230 | socket(socket&&) noexcept; | |
231 | /// Move-assigns a \c seastar::socket object. | |
232 | socket& operator=(socket&&) noexcept; | |
233 | ||
234 | /// Attempts to establish the connection. | |
235 | /// | |
236 | /// \return a \ref connected_socket representing the connection. | |
237 | future<connected_socket> connect(socket_address sa, socket_address local = socket_address(::sockaddr_in{AF_INET, INADDR_ANY, {0}}), transport proto = transport::TCP); | |
238 | /// Stops any in-flight connection attempt. | |
239 | /// | |
240 | /// Cancels the connection attempt if it's still in progress, and | |
241 | /// terminates the connection if it has already been established. | |
242 | void shutdown(); | |
243 | }; | |
244 | ||
245 | /// @} | |
246 | ||
247 | /// \addtogroup networking-module | |
248 | /// @{ | |
249 | ||
250 | /// A listening socket, waiting to accept incoming network connections. | |
251 | class server_socket { | |
252 | std::unique_ptr<net::server_socket_impl> _ssi; | |
253 | bool _aborted = false; | |
254 | public: | |
255 | enum class load_balancing_algorithm { | |
256 | // This algorithm tries to distribute all connections equally between all shards. | |
257 | // It does this by sending new connections to a shard with smallest amount of connections. | |
258 | connection_distribution, | |
259 | // This algorithm distributes new connection based on peer's tcp port. Destination shard | |
260 | // is calculated as a port number modulo number of shards. This allows a client to connect | |
261 | // to a specific shard in a server given it knows how many shards server has by choosing | |
262 | // src port number accordingly. | |
263 | port, | |
264 | default_ = connection_distribution | |
265 | }; | |
266 | /// Constructs a \c server_socket not corresponding to a connection | |
267 | server_socket(); | |
268 | /// \cond internal | |
269 | explicit server_socket(std::unique_ptr<net::server_socket_impl> ssi); | |
270 | /// \endcond | |
271 | /// Moves a \c server_socket object. | |
272 | server_socket(server_socket&& ss) noexcept; | |
273 | ~server_socket(); | |
274 | /// Move-assigns a \c server_socket object. | |
275 | server_socket& operator=(server_socket&& cs) noexcept; | |
276 | ||
277 | /// Accepts the next connection to successfully connect to this socket. | |
278 | /// | |
279 | /// \return a \ref connected_socket representing the connection, and | |
280 | /// a \ref socket_address describing the remote endpoint. | |
281 | /// | |
282 | /// \see listen(socket_address sa) | |
283 | /// \see listen(socket_address sa, listen_options opts) | |
284 | future<connected_socket, socket_address> accept(); | |
285 | ||
286 | /// Stops any \ref accept() in progress. | |
287 | /// | |
288 | /// Current and future \ref accept() calls will terminate immediately | |
289 | /// with an error. | |
290 | void abort_accept(); | |
291 | }; | |
292 | /// @} | |
293 | ||
294 | struct listen_options { | |
295 | bool reuse_address = false; | |
296 | server_socket::load_balancing_algorithm lba = server_socket::load_balancing_algorithm::default_; | |
297 | transport proto = transport::TCP; | |
298 | }; | |
299 | ||
300 | class network_stack { | |
301 | public: | |
302 | virtual ~network_stack() {} | |
303 | virtual server_socket listen(socket_address sa, listen_options opts) = 0; | |
304 | // FIXME: local parameter assumes ipv4 for now, fix when adding other AF | |
305 | future<connected_socket> connect(socket_address sa, socket_address local = socket_address(::sockaddr_in{AF_INET, INADDR_ANY, {0}}), transport proto = transport::TCP) { | |
306 | return do_with(socket(), [sa, local, proto](::seastar::socket& s) { | |
307 | return s.connect(sa, local, proto); | |
308 | }); | |
309 | } | |
310 | virtual ::seastar::socket socket() = 0; | |
311 | virtual net::udp_channel make_udp_channel(ipv4_addr addr = {}) = 0; | |
312 | virtual future<> initialize() { | |
313 | return make_ready_future(); | |
314 | } | |
315 | virtual bool has_per_core_namespace() = 0; | |
316 | }; | |
317 | ||
318 | } |