1NAL_CONNECTION_NEW(2) distcache NAL_CONNECTION_NEW(2)
2
3
4
6 NAL_CONNECTION_new, NAL_CONNECTION_free, NAL_CONNECTION_create,
7 NAL_CONNECTION_create_pair, NAL_CONNECTION_create_dummy, NAL_CONNEC‐
8 TION_set_size, NAL_CONNECTION_get_read, NAL_CONNECTION_get_send,
9 NAL_CONNECTION_io, NAL_CONNECTION_io_cap, NAL_CONNECTION_is_estab‐
10 lished, NAL_CONNECTION_add_to_selector, NAL_CONNECTION_del_from_selec‐
11 tor - libnal connection functions
12
14 #include <libnal/nal.h>
15
16 #define NAL_SELECT_FLAG_READ (unsigned int)0x0001
17 #define NAL_SELECT_FLAG_SEND (unsigned int)0x0002
18 #define NAL_SELECT_FLAG_RW (NAL_SELECT_FLAG_READ ⎪ NAL_SELECT_FLAG_SEND)
19
20 NAL_CONNECTION *NAL_CONNECTION_new(void);
21 void NAL_CONNECTION_free(NAL_CONNECTION *conn);
22 void NAL_CONNECTION_reset(NAL_CONNECTION *conn);
23 int NAL_CONNECTION_create(NAL_CONNECTION *conn, const NAL_ADDRESS *addr);
24 int NAL_CONNECTION_accept(NAL_CONNECTION *conn, NAL_LISTENER *list,
25 NAL_SELECTOR *sel);
26 int NAL_CONNECTION_create_pair(NAL_CONNECTION *conn1, NAL_CONNECTION *conn2,
27 unsigned int def_buffer_size);
28 #if 0
29 int NAL_CONNECTION_create_dummy(NAL_CONNECTION *conn,
30 unsigned int def_buffer_size);
31 #endif
32 int NAL_CONNECTION_set_size(NAL_CONNECTION *conn, unsigned int size);
33 NAL_BUFFER *NAL_CONNECTION_get_read(NAL_CONNECTION *conn);
34 NAL_BUFFER *NAL_CONNECTION_get_send(NAL_CONNECTION *conn);
35 const NAL_BUFFER *NAL_CONNECTION_get_read_c(const NAL_CONNECTION *conn);
36 const NAL_BUFFER *NAL_CONNECTION_get_send_c(const NAL_CONNECTION *conn);
37 int NAL_CONNECTION_io(NAL_CONNECTION *conn, NAL_SELECTOR *sel);
38 int NAL_CONNECTION_io_cap(NAL_CONNECTION *conn, NAL_SELECTOR *sel,
39 unsigned int max_read, unsigned int max_send);
40 int NAL_CONNECTION_is_established(const NAL_CONNECTION *conn);
41 void NAL_CONNECTION_add_to_selector(const NAL_CONNECTION *conn,
42 NAL_SELECTOR *sel);
43 void NAL_CONNECTION_add_to_selector_ex(const NAL_CONNECTION *conn,
44 NAL_SELECTOR *sel,
45 unsigned int flags);
46 void NAL_CONNECTION_del_from_selector(const NAL_CONNECTION *conn,
47 NAL_SELECTOR *sel);
48
50 NAL_CONNECTION_new() allocates and initialises a new NAL_CONNECTION
51 object.
52
53 NAL_CONNECTION_free() destroys a NAL_CONNECTION object.
54
55 NAL_CONNECTION_reset() will, if necessary, cleanup any prior state in
56 conn so that it can be reused in NAL_CONNECTION_create(). Internally,
57 there are other optimisations and benefits to using NAL_CONNEC‐
58 TION_reset() instead of NAL_CONNECTION_free() and NAL_CONNECTION_new()
59 - the implementation can try to avoid repeated reallocation and reini‐
60 tialisation of state, only doing full cleanup and reinitialisation when
61 necessary.
62
63 NAL_CONNECTION_create() will attempt to connect to the address repre‐
64 sented by addr. If this succeeds, it means either that the underlying
65 connection of conn is established, or that a non-blocking connect was
66 successfully initiated but has not yet completed (it may still be
67 rejected by the peer eventually). Typically, unix domain sockets con‐
68 nect or fail immediately, and usually TCP/IPv4 connect non-blocking,
69 though this may not be true for some interfaces such as `localhost'.
70 NAL_CONNECTION_is_established() can be used to distinguish the differ‐
71 ence. The size of the connection's underlying read and send NAL_BUFFERs
72 is initialised to the default that was created in addr. See the
73 "NOTES" section for more discussion of connection semantics.
74
75 NAL_CONNECTION_accept() will not block waiting for incoming connection
76 requests on list, but will accept any pending connection request that
77 had already been identified by a previous call to NAL_SELEC‐
78 TOR_select(2) on sel. See "NOTES".
79
80 NAL_CONNECTION_create_pair() will initialise conn1 and conn2 to be end-
81 points of a single connection. This is typically implemented using the
82 socketpair(2) function, and is designed to allow for an IPC mechanism
83 that integrates with libnal. def_buffer_size will control the size of
84 the read and send buffers of both connections if the functions succeed.
85 See the EXAMPLES section for some uses of ``pairs''.
86
87 NAL_CONNECTION_create_dummy() will implement a virtual FIFO that has no
88 underlying network resource associated with it. Writing data to the
89 connection amounts to pushing data onto the front of the FIFO, and
90 reading data from the connection amounts to popping data off the end of
91 the FIFO. The size of the FIFO is specified by def_buffer_size. See the
92 "BUGS" section for a note on using these connection types with
93 NAL_SELECTOR.
94
95 NAL_CONNECTION_set_size() will resize the read and send buffers of conn
96 to size. The default size of those buffers is inherited from the set‐
97 ting created in the NAL_ADDRESS that initialised conn, or if conn was
98 accepted from a NAL_LISTENER object, then from the address that created
99 the listener. The individual buffers can be resized independantly by
100 using the following two functions to obtain the buffesr and using
101 NAL_BUFFER functions directly.
102
103 NAL_CONNECTION_get_read() and NAL_CONNECTION_get_send() return the read
104 and send buffers of conn. This is how reading and writing is performed
105 on conn, as NAL_BUFFER functions may be used on these buffers directly.
106 NAL_CONNECTION_get_read_c() and NAL_CONNECTION_get_send_c() perform the
107 same function but on a constant conn parameter and returning constant
108 pointers to the corresponding buffers.
109
110 NAL_CONNECTION_io() will perform any network input/output that is pos‐
111 sible given the state in sel. Unless conn had been added to sel via
112 NAL_SELECTOR_add_conn() (or its `_ex' variant) and a resulting call to
113 NAL_SELECTOR_select() had revealed readability and/or writability on
114 conn, this function will silently succeed. Otherwise it will attempt to
115 perform whatever reading or writing was required. If this function
116 fails, that indicates that the connection is no longer valid - this
117 represents a disconnection by the peer, the result of a non-blocking
118 connect that had been initiated but was unable to connect, or some net‐
119 work error that makes conn unusable. See the "NOTES" section.
120
121 NAL_CONNECTION_io_cap() is a version of NAL_CONNECTION_io() that allows
122 the caller to specify a limit on the maximum amount conn should read
123 from, or send to, the network. Whether this amount is read or sent (or
124 even whether reading or sending takes place at all) depends on; the
125 data (and space) available is in the connection's buffers, what the
126 results of the last select on sel were, and how much data the host sys‐
127 tem's networking support will accept or provide to conn.
128
129 NAL_CONNECTION_is_established() is useful for determining when a non-
130 blocking connect has completed. See the "NOTES" section.
131
132 NAL_CONNECTION_add_to_selector() registers conn with the selector sel
133 for any events relevant to it. NAL_CONNECTION_del_from_selector() can
134 be used to reverse this if called before any subsequent call to
135 NAL_SELECTOR_select(). NAL_CONNECTION_add_to_selector_ex() extends
136 NAL_CONNECTION_add_to_selector() by allowing a bit-mask to be supplied
137 to control what events the connection can be selected on, these flags
138 are indicated above prefixed with NAL_SELECT_FLAG_.
139
141 NAL_CONNECTION_new() returns a valid NAL_CONNECTION object on success,
142 NULL otherwise.
143
144 NAL_CONNECTION_free(), NAL_CONNECTION_reset(), NAL_CONNEC‐
145 TION_add_to_selector(), NAL_CONNECTION_add_to_selector_ex(), and
146 NAL_CONNECTION_del_from_selector() have no return value.
147
148 NAL_CONNECTION_get_read(), NAL_CONNECTION_get_send(), NAL_CONNEC‐
149 TION_get_read_c(), and NAL_CONNECTION_get_send_c() return pointers to
150 the connection's buffer objects or NULL for failure.
151
152 NAL_CONNECTION_accept() returns non-zero if a connection was accepted
153 and is represented by the provided NAL_CONNECTION object, or zero if no
154 connection attempt was pending (or if there was but an error prevented
155 the accept operation).
156
157 All other NAL_CONNECTION functions return zero for failure or false,
158 and non-zero for success or true.
159
161 A NAL_CONNECTION object encapsulates two NAL_BUFFER objects and a non-
162 blocking socket. Any data that has been read from the socket is placed
163 in the read buffer, and applications write data into the send buffer
164 for it to be (eventually) written out to the socket. The NAL_SELECTOR
165 type provides the ability to poll for any requested network events and
166 then allow connections and listeners to perform their network
167 input/output based on the results.
168
169 NAL_CONNECTION_add_to_selector() uses the following logic; the connec‐
170 tion is always selected for exception events, and will be selected for
171 readability if its read buffer is not full and writability if its send
172 buffer is not empty.
173
174 NAL_CONNECTION_io() is used after calling NAL_CONNECTION_add_to_selec‐
175 tor() and a subsequent call to NAL_SELECTOR_select(). It observes the
176 following logic; if an exception event has occured it returns failure,
177 if readability is indicated it will read incoming data up to the limit
178 of the available space in the read buffer, and if writability is indi‐
179 cated it will send as much of the send buffer's data as possible. If
180 NAL_CONNECTION_io() returns failure, the connection is considered bro‐
181 ken for some reason and no further I/O operations should be attempted
182 (the behaviour is undefined). NB: The connection object is not automat‐
183 ically cleaned up so as to allow the caller to continue reading any
184 data in the read buffer and/or examine any unsent data in the send buf‐
185 fer.
186
187 The above is almost true, BTW :-) The special case is that of non-
188 blocking connects. If NAL_CONNECTION_create() cannot immediately con‐
189 nect without blocking, it will return success but subsequent calls to
190 NAL_CONNECTION_is_established() will reveal that the connection is not
191 yet complete. Any connection that is not complete will request selec‐
192 tion for sendability inside NAL_CONNECTION_add_to_selector(), whether
193 the application has provided data to send or not. The completion (or
194 failure) of the non-blocking connect will thus cause any subsequent
195 NAL_SELECTOR_select() operation to break. As with all other semantics,
196 it is the follow up call to NAL_CONNECTION_io() that changes the state
197 of the connection object - if it returns failure, the non-blocking con‐
198 nect failed. If it returns success, you should still call NAL_CONNEC‐
199 TION_is_established() to determine if the connection is complete, as
200 the selector could have broken because of signals or network events on
201 other objects.
202
203 NAL_CONNECTION_accept() will return immediately, and will only succeed
204 if the NAL_LISTENER object had already been added to the selector using
205 NAL_LISTENER_add_to_select(), the selector had been subsequently
206 selected using NAL_SELECTOR_select(2), and this indicated an incoming
207 connection request waiting on the listener.
208
209 It should be noted that the actual transport in use is virtualised to
210 allow for multiple transports and, because of this, multiple semantics
211 for how the network functionality behaves. TCP/IPv4 and unix domain
212 socket based connections, as well as connection pairs from NAL_CONNEC‐
213 TION_create_pair(), operate very much as described here. The FIFO con‐
214 nection type, created by NAL_CONNECTION_create_dummy() is not yet con‐
215 sistent with this and is described in the "BUGS" section.
216
218 Dummy FIFO connections created using NAL_CONNECTION_create_dummy()
219 should be trivially selectable if anyone's daft enough to try. Ie. if
220 you add a dummy connection to a selector, the NAL_SELECTOR_select()
221 should break instantly if the FIFO is non-empty otherwise the FIFO
222 should have no influence at all on the real select(2). Right now,
223 NAL_CONNECTION_add_to_selector() silently ignores dummy connections
224 completely.
225
227 A typical state-machine implementation using a single connection is
228 illustrated here (without error-checking);
229
230 NAL_BUFFER *c_read, *c_send;
231 NAL_SELECTOR *sel = NAL_SELECTOR_new();
232 NAL_CONNECTION *conn = NAL_CONNECTION_new();
233 NAL_ADDRESS *addr = retrieve_the_desired_address();
234
235 /* Setup */
236 NAL_CONNECTION_create(conn, addr);
237 c_read = NAL_CONNECTION_get_read(conn);
238 c_send = NAL_CONNECTION_get_send(conn);
239
240 /* Loop */
241 do {
242 /* This is where the state-machine code should process as much data as
243 * possible from 'c_read' and/or produce as much output to 'c_send' as
244 * it can. */
245 ...
246 ... user code
247 ...
248 /* block on (relevant) network events for 'conn' */
249 NAL_CONNECTION_add_to_selector(conn, sel);
250 NAL_SELECTOR_select(sel, 0, 0);
251 /* Do network I/O after the above blocking select and continue looping
252 * only if the connection is still alive. */
253 } while(NAL_CONNECTION_io(conn, sel));
254
255 An example of using a connection pair (with 2 Kb read and send buffers
256 for each connection) to create IPC between a parent process and its
257 child (again, no error checking);
258
259 NAL_CONNECTION *ipc_to_parent = NAL_CONNECTION_new();
260 NAL_CONNECTION *ipc_to_child = NAL_CONNECTION_new();
261
262 /* Setup */
263 NAL_CONNECTION_create_pair(ipc_to_parent, ipc_to_child, 2048);
264
265 /* Create child process */
266 switch(fork()) {
267 case 0:
268 /* Inside the child process, close our copy of the parent's side */
269 NAL_CONNECTION_free(ipc_to_child);
270 /* Do child process things, and use 'ipc_to_parent' to communicate
271 * with the parent. */
272 do_child_logic(ipc_to_parent);
273 exit(0);
274 default:
275 /* Inside the parent process, close our copy of the child's side */
276 NAL_CONNECTION_free(ipc_to_parent);
277 break;
278 }
279 /* Continue in the parent process, and use 'ipc_to_child' to communicate
280 * with the child. */
281 do_parent_logic(ipc_to_child);
282
283 Note that these connection pairs can also be a useful way of handling
284 process termination that allow you to bypass signal handling alto‐
285 gether. If a child process terminates, the connection between the pair
286 will be broken and so this will be noticed in the parent process by any
287 selector selecting on the ipc_to_child connection - the subsequent
288 NAL_CONNECTION_io() operation will fail indicating that the child
289 process is dead (or in the process of dying) and so the parent could
290 immediately call wait(2) or waitpid(2). Whether the SIGCHLD signal
291 arrives before the NAL_CONNECTION_io() call or not is not too impor‐
292 tant, at worst it might prematurely interrupt NAL_SELECTOR_select()
293 (causing it to return zero) so that a redundant loop of the state-
294 machine runs before the next select operation will notice the discon‐
295 nection. If you already need IPC between the parent and child for
296 exchange of data anyway, this mechanism could be useful in avoiding
297 global variables, signal handlers, and the associated difficulties.
298
300 NAL_CONNECTION_new(2) - Functions for the NAL_CONNECTION type.
301
302 NAL_LISTENER_new(2) - Functions for the NAL_LISTENER type.
303
304 NAL_SELECTOR_new(2) - Functions for the NAL_SELECTOR type.
305
306 NAL_BUFFER_new(2) - Functions for the NAL_BUFFER type.
307
308 distcache(8) - Overview of the distcache architecture.
309
310 http://www.distcache.org/ - Distcache home page.
311
313 This toolkit was designed and implemented by Geoff Thorpe for Crypto‐
314 graphic Appliances Incorporated. Since the project was released into
315 open source, it has a home page and a project environment where devel‐
316 opment, mailing lists, and releases are organised. For problems with
317 the software or this man page please check for new releases at the
318 project web-site below, mail the users mailing list described there, or
319 contact the author at geoff@geoffthorpe.net.
320
321 Home Page: http://www.distcache.org
322
323
324
3251.4.5 2004.03.23 NAL_CONNECTION_NEW(2)