1 /***
2 This file is part of PulseAudio.
3
4 Copyright 2014 David Henningsson, Canonical Ltd.
5
6 PulseAudio is free software; you can redistribute it and/or modify
7 it under the terms of the GNU Lesser General Public License as
8 published by the Free Software Foundation; either version 2.1 of the
9 License, or (at your option) any later version.
10
11 PulseAudio is distributed in the hope that it will be useful, but
12 WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public
17 License along with PulseAudio; if not, see <http://www.gnu.org/licenses/>.
18 ***/
19
20 #ifdef HAVE_CONFIG_H
21 #include <config.h>
22 #endif
23
24 #include "srbchannel.h"
25
26 #include <pulsecore/atomic.h>
27 #include <pulse/xmalloc.h>
28
29 /* #define DEBUG_SRBCHANNEL */
30
31 /* This ringbuffer might be useful in other contexts too, but
32 * right now it's only used inside the srbchannel, so let's keep it here
33 * for the time being. */
34 typedef struct pa_ringbuffer pa_ringbuffer;
35
36 struct pa_ringbuffer {
37 pa_atomic_t *count; /* amount of data in the buffer */
38 int capacity;
39 uint8_t *memory;
40 int readindex, writeindex;
41 };
42
pa_ringbuffer_peek(pa_ringbuffer * r,int * count)43 static void *pa_ringbuffer_peek(pa_ringbuffer *r, int *count) {
44 int c = pa_atomic_load(r->count);
45
46 if (r->readindex + c > r->capacity)
47 *count = r->capacity - r->readindex;
48 else
49 *count = c;
50
51 return r->memory + r->readindex;
52 }
53
54 /* Returns true only if the buffer was completely full before the drop. */
pa_ringbuffer_drop(pa_ringbuffer * r,int count)55 static bool pa_ringbuffer_drop(pa_ringbuffer *r, int count) {
56 bool b = pa_atomic_sub(r->count, count) >= r->capacity;
57
58 r->readindex += count;
59 r->readindex %= r->capacity;
60
61 return b;
62 }
63
pa_ringbuffer_begin_write(pa_ringbuffer * r,int * count)64 static void *pa_ringbuffer_begin_write(pa_ringbuffer *r, int *count) {
65 int c = pa_atomic_load(r->count);
66
67 *count = PA_MIN(r->capacity - r->writeindex, r->capacity - c);
68
69 return r->memory + r->writeindex;
70 }
71
pa_ringbuffer_end_write(pa_ringbuffer * r,int count)72 static void pa_ringbuffer_end_write(pa_ringbuffer *r, int count) {
73 pa_atomic_add(r->count, count);
74 r->writeindex += count;
75 r->writeindex %= r->capacity;
76 }
77
78 struct pa_srbchannel {
79 pa_ringbuffer rb_read, rb_write;
80 pa_fdsem *sem_read, *sem_write;
81 pa_memblock *memblock;
82
83 void *cb_userdata;
84 pa_srbchannel_cb_t callback;
85
86 pa_io_event *read_event;
87 pa_defer_event *defer_event;
88 pa_mainloop_api *mainloop;
89 };
90
91 /* We always listen to sem_read, and always signal on sem_write.
92 *
93 * This means we signal the same semaphore for two scenarios:
94 * 1) We have written something to our send buffer, and want the other
95 * side to read it
96 * 2) We have read something from our receive buffer that was previously
97 * completely full, and want the other side to continue writing
98 */
99
pa_srbchannel_write(pa_srbchannel * sr,const void * data,size_t l)100 size_t pa_srbchannel_write(pa_srbchannel *sr, const void *data, size_t l) {
101 size_t written = 0;
102
103 while (l > 0) {
104 int towrite;
105 void *ptr = pa_ringbuffer_begin_write(&sr->rb_write, &towrite);
106
107 if ((size_t) towrite > l)
108 towrite = l;
109
110 if (towrite == 0) {
111 #ifdef DEBUG_SRBCHANNEL
112 pa_log("srbchannel output buffer full");
113 #endif
114 break;
115 }
116
117 memcpy(ptr, data, towrite);
118 pa_ringbuffer_end_write(&sr->rb_write, towrite);
119 written += towrite;
120 data = (uint8_t*) data + towrite;
121 l -= towrite;
122 }
123 #ifdef DEBUG_SRBCHANNEL
124 pa_log("Wrote %d bytes to srbchannel, signalling fdsem", (int) written);
125 #endif
126
127 pa_fdsem_post(sr->sem_write);
128 return written;
129 }
130
pa_srbchannel_read(pa_srbchannel * sr,void * data,size_t l)131 size_t pa_srbchannel_read(pa_srbchannel *sr, void *data, size_t l) {
132 size_t isread = 0;
133
134 while (l > 0) {
135 int toread;
136 void *ptr = pa_ringbuffer_peek(&sr->rb_read, &toread);
137
138 if ((size_t) toread > l)
139 toread = l;
140
141 if (toread == 0)
142 break;
143
144 memcpy(data, ptr, toread);
145
146 if (pa_ringbuffer_drop(&sr->rb_read, toread)) {
147 #ifdef DEBUG_SRBCHANNEL
148 pa_log("Read from full output buffer, signalling fdsem");
149 #endif
150 pa_fdsem_post(sr->sem_write);
151 }
152
153 isread += toread;
154 data = (uint8_t*) data + toread;
155 l -= toread;
156 }
157
158 #ifdef DEBUG_SRBCHANNEL
159 pa_log("Read %d bytes from srbchannel", (int) isread);
160 #endif
161
162 return isread;
163 }
164
165 /* This is the memory layout of the ringbuffer shm block. It is followed by
166 read and write ringbuffer memory. */
167 struct srbheader {
168 pa_atomic_t read_count;
169 pa_atomic_t write_count;
170
171 pa_fdsem_data read_semdata;
172 pa_fdsem_data write_semdata;
173
174 int capacity;
175 int readbuf_offset;
176 int writebuf_offset;
177
178 /* TODO: Maybe a marker here to make sure we talk to a server with equally sized struct */
179 };
180
srbchannel_rwloop(pa_srbchannel * sr)181 static void srbchannel_rwloop(pa_srbchannel* sr) {
182 do {
183 #ifdef DEBUG_SRBCHANNEL
184 int q;
185 pa_ringbuffer_peek(&sr->rb_read, &q);
186 pa_log("In rw loop from srbchannel, before callback, count = %d", q);
187 #endif
188
189 if (sr->callback) {
190 if (!sr->callback(sr, sr->cb_userdata)) {
191 #ifdef DEBUG_SRBCHANNEL
192 pa_log("Aborting read loop from srbchannel");
193 #endif
194 return;
195 }
196 }
197
198 #ifdef DEBUG_SRBCHANNEL
199 pa_ringbuffer_peek(&sr->rb_read, &q);
200 pa_log("In rw loop from srbchannel, after callback, count = %d", q);
201 #endif
202
203 } while (pa_fdsem_before_poll(sr->sem_read) < 0);
204 }
205
semread_cb(pa_mainloop_api * m,pa_io_event * e,int fd,pa_io_event_flags_t events,void * userdata)206 static void semread_cb(pa_mainloop_api *m, pa_io_event *e, int fd, pa_io_event_flags_t events, void *userdata) {
207 pa_srbchannel* sr = userdata;
208
209 pa_fdsem_after_poll(sr->sem_read);
210 srbchannel_rwloop(sr);
211 }
212
defer_cb(pa_mainloop_api * m,pa_defer_event * e,void * userdata)213 static void defer_cb(pa_mainloop_api *m, pa_defer_event *e, void *userdata) {
214 pa_srbchannel* sr = userdata;
215
216 #ifdef DEBUG_SRBCHANNEL
217 pa_log("Calling rw loop from deferred event");
218 #endif
219
220 m->defer_enable(e, 0);
221 srbchannel_rwloop(sr);
222 }
223
pa_srbchannel_new(pa_mainloop_api * m,pa_mempool * p)224 pa_srbchannel* pa_srbchannel_new(pa_mainloop_api *m, pa_mempool *p) {
225 int capacity;
226 int readfd;
227 struct srbheader *srh;
228
229 pa_srbchannel* sr = pa_xmalloc0(sizeof(pa_srbchannel));
230 sr->mainloop = m;
231 sr->memblock = pa_memblock_new_pool(p, -1);
232 if (!sr->memblock)
233 goto fail;
234
235 srh = pa_memblock_acquire(sr->memblock);
236 pa_zero(*srh);
237
238 sr->rb_read.memory = (uint8_t*) srh + PA_ALIGN(sizeof(*srh));
239 srh->readbuf_offset = sr->rb_read.memory - (uint8_t*) srh;
240
241 capacity = (pa_memblock_get_length(sr->memblock) - srh->readbuf_offset) / 2;
242
243 sr->rb_write.memory = PA_ALIGN_PTR(sr->rb_read.memory + capacity);
244 srh->writebuf_offset = sr->rb_write.memory - (uint8_t*) srh;
245
246 capacity = PA_MIN(capacity, srh->writebuf_offset - srh->readbuf_offset);
247
248 pa_log_debug("SHM block is %d bytes, ringbuffer capacity is 2 * %d bytes",
249 (int) pa_memblock_get_length(sr->memblock), capacity);
250
251 srh->capacity = sr->rb_read.capacity = sr->rb_write.capacity = capacity;
252
253 sr->rb_read.count = &srh->read_count;
254 sr->rb_write.count = &srh->write_count;
255
256 sr->sem_read = pa_fdsem_new_shm(&srh->read_semdata);
257 if (!sr->sem_read)
258 goto fail;
259
260 sr->sem_write = pa_fdsem_new_shm(&srh->write_semdata);
261 if (!sr->sem_write)
262 goto fail;
263
264 readfd = pa_fdsem_get(sr->sem_read);
265
266 #ifdef DEBUG_SRBCHANNEL
267 pa_log("Enabling io event on fd %d", readfd);
268 #endif
269
270 sr->read_event = m->io_new(m, readfd, PA_IO_EVENT_INPUT, semread_cb, sr);
271 m->io_enable(sr->read_event, PA_IO_EVENT_INPUT);
272
273 return sr;
274
275 fail:
276 pa_srbchannel_free(sr);
277
278 return NULL;
279 }
280
pa_srbchannel_swap(pa_srbchannel * sr)281 static void pa_srbchannel_swap(pa_srbchannel *sr) {
282 pa_srbchannel temp = *sr;
283
284 sr->sem_read = temp.sem_write;
285 sr->sem_write = temp.sem_read;
286 sr->rb_read = temp.rb_write;
287 sr->rb_write = temp.rb_read;
288 }
289
pa_srbchannel_new_from_template(pa_mainloop_api * m,pa_srbchannel_template * t)290 pa_srbchannel* pa_srbchannel_new_from_template(pa_mainloop_api *m, pa_srbchannel_template *t)
291 {
292 int temp;
293 struct srbheader *srh;
294 pa_srbchannel* sr = pa_xmalloc0(sizeof(pa_srbchannel));
295
296 sr->mainloop = m;
297 sr->memblock = t->memblock;
298 pa_memblock_ref(sr->memblock);
299 srh = pa_memblock_acquire(sr->memblock);
300
301 sr->rb_read.capacity = sr->rb_write.capacity = srh->capacity;
302 sr->rb_read.count = &srh->read_count;
303 sr->rb_write.count = &srh->write_count;
304
305 sr->rb_read.memory = (uint8_t*) srh + srh->readbuf_offset;
306 sr->rb_write.memory = (uint8_t*) srh + srh->writebuf_offset;
307
308 sr->sem_read = pa_fdsem_open_shm(&srh->read_semdata, t->readfd);
309 if (!sr->sem_read)
310 goto fail;
311
312 sr->sem_write = pa_fdsem_open_shm(&srh->write_semdata, t->writefd);
313 if (!sr->sem_write)
314 goto fail;
315
316 pa_srbchannel_swap(sr);
317 temp = t->readfd; t->readfd = t->writefd; t->writefd = temp;
318
319 #ifdef DEBUG_SRBCHANNEL
320 pa_log("Enabling io event on fd %d", t->readfd);
321 #endif
322
323 sr->read_event = m->io_new(m, t->readfd, PA_IO_EVENT_INPUT, semread_cb, sr);
324 m->io_enable(sr->read_event, PA_IO_EVENT_INPUT);
325
326 return sr;
327
328 fail:
329 pa_srbchannel_free(sr);
330
331 return NULL;
332 }
333
pa_srbchannel_export(pa_srbchannel * sr,pa_srbchannel_template * t)334 void pa_srbchannel_export(pa_srbchannel *sr, pa_srbchannel_template *t) {
335 t->memblock = sr->memblock;
336 t->readfd = pa_fdsem_get(sr->sem_read);
337 t->writefd = pa_fdsem_get(sr->sem_write);
338 }
339
pa_srbchannel_set_callback(pa_srbchannel * sr,pa_srbchannel_cb_t callback,void * userdata)340 void pa_srbchannel_set_callback(pa_srbchannel *sr, pa_srbchannel_cb_t callback, void *userdata) {
341 if (sr->callback)
342 pa_fdsem_after_poll(sr->sem_read);
343
344 sr->callback = callback;
345 sr->cb_userdata = userdata;
346
347 if (sr->callback) {
348 /* If there are events to be read already in the ringbuffer, we will not get any IO event for that,
349 because that's how pa_fdsem works. Therefore check the ringbuffer in a defer event instead. */
350 if (!sr->defer_event)
351 sr->defer_event = sr->mainloop->defer_new(sr->mainloop, defer_cb, sr);
352 sr->mainloop->defer_enable(sr->defer_event, 1);
353 }
354 }
355
pa_srbchannel_free(pa_srbchannel * sr)356 void pa_srbchannel_free(pa_srbchannel *sr)
357 {
358 #ifdef DEBUG_SRBCHANNEL
359 pa_log("Freeing srbchannel");
360 #endif
361 pa_assert(sr);
362
363 if (sr->defer_event)
364 sr->mainloop->defer_free(sr->defer_event);
365 if (sr->read_event)
366 sr->mainloop->io_free(sr->read_event);
367
368 if (sr->sem_read)
369 pa_fdsem_free(sr->sem_read);
370 if (sr->sem_write)
371 pa_fdsem_free(sr->sem_write);
372
373 if (sr->memblock) {
374 pa_memblock_release(sr->memblock);
375 pa_memblock_unref(sr->memblock);
376 }
377
378 pa_xfree(sr);
379 }
380