6 ASYNC_init_thread, ASYNC_cleanup_thread, ASYNC_start_job, ASYNC_pause_job,
7 ASYNC_get_current_job, ASYNC_block_pause, ASYNC_unblock_pause, ASYNC_is_capable,
8 ASYNC_stack_alloc_fn, ASYNC_stack_free_fn, ASYNC_set_mem_functions, ASYNC_get_mem_functions
9 - asynchronous job management functions
13 #include <openssl/async.h>
15 int ASYNC_init_thread(size_t max_size, size_t init_size);
16 void ASYNC_cleanup_thread(void);
18 int ASYNC_start_job(ASYNC_JOB **job, ASYNC_WAIT_CTX *ctx, int *ret,
19 int (*func)(void *), void *args, size_t size);
20 int ASYNC_pause_job(void);
22 ASYNC_JOB *ASYNC_get_current_job(void);
23 ASYNC_WAIT_CTX *ASYNC_get_wait_ctx(ASYNC_JOB *job);
24 void ASYNC_block_pause(void);
25 void ASYNC_unblock_pause(void);
27 int ASYNC_is_capable(void);
29 typedef void *(*ASYNC_stack_alloc_fn)(size_t *num);
30 typedef void (*ASYNC_stack_free_fn)(void *addr);
31 int ASYNC_set_mem_functions(ASYNC_stack_alloc_fn alloc_fn,
32 ASYNC_stack_free_fn free_fn);
33 void ASYNC_get_mem_functions(ASYNC_stack_alloc_fn *alloc_fn,
34 ASYNC_stack_free_fn *free_fn);
38 OpenSSL implements asynchronous capabilities through an B<ASYNC_JOB>. This
39 represents code that can be started and executes until some event occurs. At
40 that point the code can be paused and control returns to user code until some
41 subsequent event indicates that the job can be resumed.
43 The creation of an B<ASYNC_JOB> is a relatively expensive operation. Therefore,
44 for efficiency reasons, jobs can be created up front and reused many times. They
45 are held in a pool until they are needed, at which point they are removed from
46 the pool, used, and then returned to the pool when the job completes. If the
47 user application is multi-threaded, then ASYNC_init_thread() may be called for
48 each thread that will initiate asynchronous jobs. Before
49 user code exits per-thread resources need to be cleaned up. This will normally
50 occur automatically (see L<OPENSSL_init_crypto(3)>) but may be explicitly
51 initiated by using ASYNC_cleanup_thread(). No asynchronous jobs must be
52 outstanding for the thread when ASYNC_cleanup_thread() is called. Failing to
53 ensure this will result in memory leaks.
55 The I<max_size> argument limits the number of B<ASYNC_JOB>s that will be held in
56 the pool. If I<max_size> is set to 0 then no upper limit is set. When an
57 B<ASYNC_JOB> is needed but there are none available in the pool already then one
58 will be automatically created, as long as the total of B<ASYNC_JOB>s managed by
59 the pool does not exceed I<max_size>. When the pool is first initialised
60 I<init_size> B<ASYNC_JOB>s will be created immediately. If ASYNC_init_thread()
61 is not called before the pool is first used then it will be called automatically
62 with a I<max_size> of 0 (no upper limit) and an I<init_size> of 0 (no
63 B<ASYNC_JOB>s created up front).
65 An asynchronous job is started by calling the ASYNC_start_job() function.
66 Initially I<*job> should be NULL. I<ctx> should point to an B<ASYNC_WAIT_CTX>
67 object created through the L<ASYNC_WAIT_CTX_new(3)> function. I<ret> should
68 point to a location where the return value of the asynchronous function should
69 be stored on completion of the job. I<func> represents the function that should
70 be started asynchronously. The data pointed to by I<args> and of size I<size>
71 will be copied and then passed as an argument to I<func> when the job starts.
72 ASYNC_start_job will return one of the following values:
78 An error occurred trying to start the job. Check the OpenSSL error queue (e.g.
79 see L<ERR_print_errors(3)>) for more details.
81 =item B<ASYNC_NO_JOBS>
83 There are no jobs currently available in the pool. This call can be retried
84 again at a later time.
88 The job was successfully started but was "paused" before it completed (see
89 ASYNC_pause_job() below). A handle to the job is placed in I<*job>. Other work
90 can be performed (if desired) and the job restarted at a later time. To restart
91 a job call ASYNC_start_job() again passing the job handle in I<*job>. The
92 I<func>, I<args> and I<size> parameters will be ignored when restarting a job.
93 When restarting a job ASYNC_start_job() B<must> be called from the same thread
94 that the job was originally started from.
98 The job completed. I<*job> will be NULL and the return value from I<func> will
103 At any one time there can be a maximum of one job actively running per thread
104 (you can have many that are paused). ASYNC_get_current_job() can be used to get
105 a pointer to the currently executing B<ASYNC_JOB>. If no job is currently
106 executing then this will return NULL.
108 If executing within the context of a job (i.e. having been called directly or
109 indirectly by the function "func" passed as an argument to ASYNC_start_job())
110 then ASYNC_pause_job() will immediately return control to the calling
111 application with B<ASYNC_PAUSE> returned from the ASYNC_start_job() call. A
112 subsequent call to ASYNC_start_job passing in the relevant B<ASYNC_JOB> in the
113 I<*job> parameter will resume execution from the ASYNC_pause_job() call. If
114 ASYNC_pause_job() is called whilst not within the context of a job then no
115 action is taken and ASYNC_pause_job() returns immediately.
117 ASYNC_get_wait_ctx() can be used to get a pointer to the B<ASYNC_WAIT_CTX>
118 for the I<job>. B<ASYNC_WAIT_CTX>s contain two different ways to notify
119 applications that a job is ready to be resumed. One is a "wait" file
120 descriptor, and the other is a "callback" mechanism.
122 The "wait" file descriptor associated with B<ASYNC_WAIT_CTX> is used for
123 applications to wait for the file descriptor to be ready for "read" using a
124 system function call such as select or poll (being ready for "read" indicates
125 that the job should be resumed). If no file descriptor is made available then
126 an application will have to periodically "poll" the job by attempting to restart
127 it to see if it is ready to continue.
129 B<ASYNC_WAIT_CTX>s also have a "callback" mechanism to notify applications. The
130 callback is set by an application, and it will be automatically called when an
131 engine completes a cryptography operation, so that the application can resume
132 the paused work flow without polling. An engine could be written to look whether
133 the callback has been set. If it has then it would use the callback mechanism
134 in preference to the file descriptor notifications. If a callback is not set
135 then the engine may use file descriptor based notifications. Please note that
136 not all engines may support the callback mechanism, so the callback may not be
137 used even if it has been set. See ASYNC_WAIT_CTX_new() for more details.
139 The ASYNC_block_pause() function will prevent the currently active job from
140 pausing. The block will remain in place until a subsequent call to
141 ASYNC_unblock_pause(). These functions can be nested, e.g. if you call
142 ASYNC_block_pause() twice then you must call ASYNC_unblock_pause() twice in
143 order to re-enable pausing. If these functions are called while there is no
144 currently active job then they have no effect. This functionality can be useful
145 to avoid deadlock scenarios. For example during the execution of an B<ASYNC_JOB>
146 an application acquires a lock. It then calls some cryptographic function which
147 invokes ASYNC_pause_job(). This returns control back to the code that created
148 the B<ASYNC_JOB>. If that code then attempts to acquire the same lock before
149 resuming the original job then a deadlock can occur. By calling
150 ASYNC_block_pause() immediately after acquiring the lock and
151 ASYNC_unblock_pause() immediately before releasing it then this situation cannot
154 Some platforms cannot support async operations. The ASYNC_is_capable() function
155 can be used to detect whether the current platform is async capable or not.
157 Custom memory allocation functions are supported for the POSIX platform.
158 Custom memory allocation functions allow alternative methods of allocating
159 stack memory such as mmap, or using stack memory from the current thread.
160 Using an ASYNC_stack_alloc_fn callback also allows manipulation of the stack
161 size, which defaults to 32k.
162 The stack size can be altered by allocating a stack of a size different to
163 the requested size, and passing back the new stack size in the callback's I<*num>
168 ASYNC_init_thread returns 1 on success or 0 otherwise.
170 ASYNC_start_job returns one of B<ASYNC_ERR>, B<ASYNC_NO_JOBS>, B<ASYNC_PAUSE> or
171 B<ASYNC_FINISH> as described above.
173 ASYNC_pause_job returns 0 if an error occurred or 1 on success. If called when
174 not within the context of an B<ASYNC_JOB> then this is counted as success so 1
177 ASYNC_get_current_job returns a pointer to the currently executing B<ASYNC_JOB>
178 or NULL if not within the context of a job.
180 ASYNC_get_wait_ctx() returns a pointer to the B<ASYNC_WAIT_CTX> for the job.
182 ASYNC_is_capable() returns 1 if the current platform is async capable or 0
185 ASYNC_set_mem_functions returns 1 if custom stack allocators are supported by
186 the current platform and no allocations have already occurred or 0 otherwise.
190 On Windows platforms the F<< <openssl/async.h> >> header is dependent on some
191 of the types customarily made available by including F<< <windows.h> >>. The
192 application developer is likely to require control over when the latter
193 is included, commonly as one of the first included headers. Therefore,
194 it is defined as an application developer's responsibility to include
195 F<< <windows.h> >> prior to F<< <openssl/async.h> >>.
199 The following example demonstrates how to use most of the core async APIs:
202 # include <windows.h>
206 #include <openssl/async.h>
207 #include <openssl/crypto.h>
211 void cleanup(ASYNC_WAIT_CTX *ctx, const void *key, OSSL_ASYNC_FD r, void *vw)
213 OSSL_ASYNC_FD *w = (OSSL_ASYNC_FD *)vw;
220 int jobfunc(void *arg)
224 int pipefds[2] = {0, 0};
228 currjob = ASYNC_get_current_job();
229 if (currjob != NULL) {
230 printf("Executing within a job\n");
232 printf("Not executing within a job - should not happen\n");
236 msg = (unsigned char *)arg;
237 printf("Passed in message is: %s\n", msg);
239 if (pipe(pipefds) != 0) {
240 printf("Failed to create pipe\n");
243 wptr = OPENSSL_malloc(sizeof(OSSL_ASYNC_FD));
245 printf("Failed to malloc\n");
249 ASYNC_WAIT_CTX_set_wait_fd(ASYNC_get_wait_ctx(currjob), &unique,
250 pipefds[0], wptr, cleanup);
253 * Normally some external event would cause this to happen at some
254 * later point - but we do it here for demo purposes, i.e.
255 * immediately signalling that the job is ready to be woken up after
256 * we return to main via ASYNC_pause_job().
258 write(pipefds[1], &buf, 1);
260 /* Return control back to main */
263 /* Clear the wake signal */
264 read(pipefds[0], &buf, 1);
266 printf ("Resumed the job after a pause\n");
273 ASYNC_JOB *job = NULL;
274 ASYNC_WAIT_CTX *ctx = NULL;
276 OSSL_ASYNC_FD waitfd;
279 unsigned char msg[13] = "Hello world!";
281 printf("Starting...\n");
283 ctx = ASYNC_WAIT_CTX_new();
285 printf("Failed to create ASYNC_WAIT_CTX\n");
290 switch (ASYNC_start_job(&job, ctx, &ret, jobfunc, msg, sizeof(msg))) {
293 printf("An error occurred\n");
296 printf("Job was paused\n");
299 printf("Job finished with return value %d\n", ret);
303 /* Wait for the job to be woken */
304 printf("Waiting for the job to be woken up\n");
306 if (!ASYNC_WAIT_CTX_get_all_fds(ctx, NULL, &numfds)
308 printf("Unexpected number of fds\n");
311 ASYNC_WAIT_CTX_get_all_fds(ctx, &waitfd, &numfds);
313 FD_SET(waitfd, &waitfdset);
314 select(waitfd + 1, &waitfdset, NULL, NULL, NULL);
318 ASYNC_WAIT_CTX_free(ctx);
319 printf("Finishing\n");
324 The expected output from executing the above example program is:
327 Executing within a job
328 Passed in message is: Hello world!
330 Waiting for the job to be woken up
331 Resumed the job after a pause
332 Job finished with return value 1
337 L<crypto(7)>, L<ERR_print_errors(3)>
341 ASYNC_init_thread, ASYNC_cleanup_thread,
342 ASYNC_start_job, ASYNC_pause_job, ASYNC_get_current_job, ASYNC_get_wait_ctx(),
343 ASYNC_block_pause(), ASYNC_unblock_pause() and ASYNC_is_capable() were first
344 added in OpenSSL 1.1.0.
348 Copyright 2015-2022 The OpenSSL Project Authors. All Rights Reserved.
350 Licensed under the Apache License 2.0 (the "License"). You may not use
351 this file except in compliance with the License. You can obtain a copy
352 in the file LICENSE in the source distribution or at
353 L<https://www.openssl.org/source/license.html>.