]>
Commit | Line | Data |
---|---|---|
7c673cae FG |
1 | .. BSD LICENSE |
2 | Copyright(c) 2010-2014 Intel Corporation. All rights reserved. | |
3 | All rights reserved. | |
4 | ||
5 | Redistribution and use in source and binary forms, with or without | |
6 | modification, are permitted provided that the following conditions | |
7 | are met: | |
8 | ||
9 | * Redistributions of source code must retain the above copyright | |
10 | notice, this list of conditions and the following disclaimer. | |
11 | * Redistributions in binary form must reproduce the above copyright | |
12 | notice, this list of conditions and the following disclaimer in | |
13 | the documentation and/or other materials provided with the | |
14 | distribution. | |
15 | * Neither the name of Intel Corporation nor the names of its | |
16 | contributors may be used to endorse or promote products derived | |
17 | from this software without specific prior written permission. | |
18 | ||
19 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | |
20 | "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
21 | LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | |
22 | A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | |
23 | OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | |
24 | SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | |
25 | LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | |
26 | DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | |
27 | THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | |
28 | (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | |
29 | OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
30 | ||
31 | .. _Ring_Library: | |
32 | ||
33 | Ring Library | |
34 | ============ | |
35 | ||
36 | The ring allows the management of queues. | |
37 | Instead of having a linked list of infinite size, the rte_ring has the following properties: | |
38 | ||
39 | * FIFO | |
40 | ||
41 | * Maximum size is fixed, the pointers are stored in a table | |
42 | ||
43 | * Lockless implementation | |
44 | ||
45 | * Multi-consumer or single-consumer dequeue | |
46 | ||
47 | * Multi-producer or single-producer enqueue | |
48 | ||
49 | * Bulk dequeue - Dequeues the specified count of objects if successful; otherwise fails | |
50 | ||
51 | * Bulk enqueue - Enqueues the specified count of objects if successful; otherwise fails | |
52 | ||
53 | * Burst dequeue - Dequeue the maximum available objects if the specified count cannot be fulfilled | |
54 | ||
55 | * Burst enqueue - Enqueue the maximum available objects if the specified count cannot be fulfilled | |
56 | ||
57 | The advantages of this data structure over a linked list queue are as follows: | |
58 | ||
59 | * Faster; only requires a single Compare-And-Swap instruction of sizeof(void \*) instead of several double-Compare-And-Swap instructions. | |
60 | ||
61 | * Simpler than a full lockless queue. | |
62 | ||
63 | * Adapted to bulk enqueue/dequeue operations. | |
64 | As pointers are stored in a table, a dequeue of several objects will not produce as many cache misses as in a linked queue. | |
65 | Also, a bulk dequeue of many objects does not cost more than a dequeue of a simple object. | |
66 | ||
67 | The disadvantages: | |
68 | ||
69 | * Size is fixed | |
70 | ||
71 | * Having many rings costs more in terms of memory than a linked list queue. An empty ring contains at least N pointers. | |
72 | ||
73 | A simplified representation of a Ring is shown in with consumer and producer head and tail pointers to objects stored in the data structure. | |
74 | ||
75 | .. _figure_ring1: | |
76 | ||
77 | .. figure:: img/ring1.* | |
78 | ||
79 | Ring Structure | |
80 | ||
81 | ||
82 | References for Ring Implementation in FreeBSD* | |
83 | ---------------------------------------------- | |
84 | ||
85 | The following code was added in FreeBSD 8.0, and is used in some network device drivers (at least in Intel drivers): | |
86 | ||
87 | * `bufring.h in FreeBSD <http://svn.freebsd.org/viewvc/base/release/8.0.0/sys/sys/buf_ring.h?revision=199625&view=markup>`_ | |
88 | ||
89 | * `bufring.c in FreeBSD <http://svn.freebsd.org/viewvc/base/release/8.0.0/sys/kern/subr_bufring.c?revision=199625&view=markup>`_ | |
90 | ||
91 | Lockless Ring Buffer in Linux* | |
92 | ------------------------------ | |
93 | ||
94 | The following is a link describing the `Linux Lockless Ring Buffer Design <http://lwn.net/Articles/340400/>`_. | |
95 | ||
96 | Additional Features | |
97 | ------------------- | |
98 | ||
99 | Name | |
100 | ~~~~ | |
101 | ||
102 | A ring is identified by a unique name. | |
103 | It is not possible to create two rings with the same name (rte_ring_create() returns NULL if this is attempted). | |
104 | ||
105 | Water Marking | |
106 | ~~~~~~~~~~~~~ | |
107 | ||
108 | The ring can have a high water mark (threshold). | |
109 | Once an enqueue operation reaches the high water mark, the producer is notified, if the water mark is configured. | |
110 | ||
111 | This mechanism can be used, for example, to exert a back pressure on I/O to inform the LAN to PAUSE. | |
112 | ||
113 | Debug | |
114 | ~~~~~ | |
115 | ||
116 | When debug is enabled (CONFIG_RTE_LIBRTE_RING_DEBUG is set), | |
117 | the library stores some per-ring statistic counters about the number of enqueues/dequeues. | |
118 | These statistics are per-core to avoid concurrent accesses or atomic operations. | |
119 | ||
120 | Use Cases | |
121 | --------- | |
122 | ||
123 | Use cases for the Ring library include: | |
124 | ||
125 | * Communication between applications in the DPDK | |
126 | ||
127 | * Used by memory pool allocator | |
128 | ||
129 | Anatomy of a Ring Buffer | |
130 | ------------------------ | |
131 | ||
132 | This section explains how a ring buffer operates. | |
133 | The ring structure is composed of two head and tail couples; one is used by producers and one is used by the consumers. | |
134 | The figures of the following sections refer to them as prod_head, prod_tail, cons_head and cons_tail. | |
135 | ||
136 | Each figure represents a simplified state of the ring, which is a circular buffer. | |
137 | The content of the function local variables is represented on the top of the figure, | |
138 | and the content of ring structure is represented on the bottom of the figure. | |
139 | ||
140 | Single Producer Enqueue | |
141 | ~~~~~~~~~~~~~~~~~~~~~~~ | |
142 | ||
143 | This section explains what occurs when a producer adds an object to the ring. | |
144 | In this example, only the producer head and tail (prod_head and prod_tail) are modified, | |
145 | and there is only one producer. | |
146 | ||
147 | The initial state is to have a prod_head and prod_tail pointing at the same location. | |
148 | ||
149 | Enqueue First Step | |
150 | ^^^^^^^^^^^^^^^^^^ | |
151 | ||
152 | First, *ring->prod_head* and ring->cons_tail are copied in local variables. | |
153 | The prod_next local variable points to the next element of the table, or several elements after in case of bulk enqueue. | |
154 | ||
155 | If there is not enough room in the ring (this is detected by checking cons_tail), it returns an error. | |
156 | ||
157 | ||
158 | .. _figure_ring-enqueue1: | |
159 | ||
160 | .. figure:: img/ring-enqueue1.* | |
161 | ||
162 | Enqueue first step | |
163 | ||
164 | ||
165 | Enqueue Second Step | |
166 | ^^^^^^^^^^^^^^^^^^^ | |
167 | ||
168 | The second step is to modify *ring->prod_head* in ring structure to point to the same location as prod_next. | |
169 | ||
170 | A pointer to the added object is copied in the ring (obj4). | |
171 | ||
172 | ||
173 | .. _figure_ring-enqueue2: | |
174 | ||
175 | .. figure:: img/ring-enqueue2.* | |
176 | ||
177 | Enqueue second step | |
178 | ||
179 | ||
180 | Enqueue Last Step | |
181 | ^^^^^^^^^^^^^^^^^ | |
182 | ||
183 | Once the object is added in the ring, ring->prod_tail in the ring structure is modified to point to the same location as *ring->prod_head*. | |
184 | The enqueue operation is finished. | |
185 | ||
186 | ||
187 | .. _figure_ring-enqueue3: | |
188 | ||
189 | .. figure:: img/ring-enqueue3.* | |
190 | ||
191 | Enqueue last step | |
192 | ||
193 | ||
194 | Single Consumer Dequeue | |
195 | ~~~~~~~~~~~~~~~~~~~~~~~ | |
196 | ||
197 | This section explains what occurs when a consumer dequeues an object from the ring. | |
198 | In this example, only the consumer head and tail (cons_head and cons_tail) are modified and there is only one consumer. | |
199 | ||
200 | The initial state is to have a cons_head and cons_tail pointing at the same location. | |
201 | ||
202 | Dequeue First Step | |
203 | ^^^^^^^^^^^^^^^^^^ | |
204 | ||
205 | First, ring->cons_head and ring->prod_tail are copied in local variables. | |
206 | The cons_next local variable points to the next element of the table, or several elements after in the case of bulk dequeue. | |
207 | ||
208 | If there are not enough objects in the ring (this is detected by checking prod_tail), it returns an error. | |
209 | ||
210 | ||
211 | .. _figure_ring-dequeue1: | |
212 | ||
213 | .. figure:: img/ring-dequeue1.* | |
214 | ||
215 | Dequeue last step | |
216 | ||
217 | ||
218 | Dequeue Second Step | |
219 | ^^^^^^^^^^^^^^^^^^^ | |
220 | ||
221 | The second step is to modify ring->cons_head in the ring structure to point to the same location as cons_next. | |
222 | ||
223 | The pointer to the dequeued object (obj1) is copied in the pointer given by the user. | |
224 | ||
225 | ||
226 | .. _figure_ring-dequeue2: | |
227 | ||
228 | .. figure:: img/ring-dequeue2.* | |
229 | ||
230 | Dequeue second step | |
231 | ||
232 | ||
233 | Dequeue Last Step | |
234 | ^^^^^^^^^^^^^^^^^ | |
235 | ||
236 | Finally, ring->cons_tail in the ring structure is modified to point to the same location as ring->cons_head. | |
237 | The dequeue operation is finished. | |
238 | ||
239 | ||
240 | .. _figure_ring-dequeue3: | |
241 | ||
242 | .. figure:: img/ring-dequeue3.* | |
243 | ||
244 | Dequeue last step | |
245 | ||
246 | ||
247 | Multiple Producers Enqueue | |
248 | ~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
249 | ||
250 | This section explains what occurs when two producers concurrently add an object to the ring. | |
251 | In this example, only the producer head and tail (prod_head and prod_tail) are modified. | |
252 | ||
253 | The initial state is to have a prod_head and prod_tail pointing at the same location. | |
254 | ||
255 | Multiple Producers Enqueue First Step | |
256 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
257 | ||
258 | On both cores, *ring->prod_head* and ring->cons_tail are copied in local variables. | |
259 | The prod_next local variable points to the next element of the table, | |
260 | or several elements after in the case of bulk enqueue. | |
261 | ||
262 | If there is not enough room in the ring (this is detected by checking cons_tail), it returns an error. | |
263 | ||
264 | ||
265 | .. _figure_ring-mp-enqueue1: | |
266 | ||
267 | .. figure:: img/ring-mp-enqueue1.* | |
268 | ||
269 | Multiple producer enqueue first step | |
270 | ||
271 | ||
272 | Multiple Producers Enqueue Second Step | |
273 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
274 | ||
275 | The second step is to modify ring->prod_head in the ring structure to point to the same location as prod_next. | |
276 | This operation is done using a Compare And Swap (CAS) instruction, which does the following operations atomically: | |
277 | ||
278 | * If ring->prod_head is different to local variable prod_head, | |
279 | the CAS operation fails, and the code restarts at first step. | |
280 | ||
281 | * Otherwise, ring->prod_head is set to local prod_next, | |
282 | the CAS operation is successful, and processing continues. | |
283 | ||
284 | In the figure, the operation succeeded on core 1, and step one restarted on core 2. | |
285 | ||
286 | ||
287 | .. _figure_ring-mp-enqueue2: | |
288 | ||
289 | .. figure:: img/ring-mp-enqueue2.* | |
290 | ||
291 | Multiple producer enqueue second step | |
292 | ||
293 | ||
294 | Multiple Producers Enqueue Third Step | |
295 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
296 | ||
297 | The CAS operation is retried on core 2 with success. | |
298 | ||
299 | The core 1 updates one element of the ring(obj4), and the core 2 updates another one (obj5). | |
300 | ||
301 | ||
302 | .. _figure_ring-mp-enqueue3: | |
303 | ||
304 | .. figure:: img/ring-mp-enqueue3.* | |
305 | ||
306 | Multiple producer enqueue third step | |
307 | ||
308 | ||
309 | Multiple Producers Enqueue Fourth Step | |
310 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
311 | ||
312 | Each core now wants to update ring->prod_tail. | |
313 | A core can only update it if ring->prod_tail is equal to the prod_head local variable. | |
314 | This is only true on core 1. The operation is finished on core 1. | |
315 | ||
316 | ||
317 | .. _figure_ring-mp-enqueue4: | |
318 | ||
319 | .. figure:: img/ring-mp-enqueue4.* | |
320 | ||
321 | Multiple producer enqueue fourth step | |
322 | ||
323 | ||
324 | Multiple Producers Enqueue Last Step | |
325 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
326 | ||
327 | Once ring->prod_tail is updated by core 1, core 2 is allowed to update it too. | |
328 | The operation is also finished on core 2. | |
329 | ||
330 | ||
331 | .. _figure_ring-mp-enqueue5: | |
332 | ||
333 | .. figure:: img/ring-mp-enqueue5.* | |
334 | ||
335 | Multiple producer enqueue last step | |
336 | ||
337 | ||
338 | Modulo 32-bit Indexes | |
339 | ~~~~~~~~~~~~~~~~~~~~~ | |
340 | ||
341 | In the preceding figures, the prod_head, prod_tail, cons_head and cons_tail indexes are represented by arrows. | |
342 | In the actual implementation, these values are not between 0 and size(ring)-1 as would be assumed. | |
343 | The indexes are between 0 and 2^32 -1, and we mask their value when we access the pointer table (the ring itself). | |
344 | 32-bit modulo also implies that operations on indexes (such as, add/subtract) will automatically do 2^32 modulo | |
345 | if the result overflows the 32-bit number range. | |
346 | ||
347 | The following are two examples that help to explain how indexes are used in a ring. | |
348 | ||
349 | .. note:: | |
350 | ||
351 | To simplify the explanation, operations with modulo 16-bit are used instead of modulo 32-bit. | |
352 | In addition, the four indexes are defined as unsigned 16-bit integers, | |
353 | as opposed to unsigned 32-bit integers in the more realistic case. | |
354 | ||
355 | ||
356 | .. _figure_ring-modulo1: | |
357 | ||
358 | .. figure:: img/ring-modulo1.* | |
359 | ||
360 | Modulo 32-bit indexes - Example 1 | |
361 | ||
362 | ||
363 | This ring contains 11000 entries. | |
364 | ||
365 | ||
366 | .. _figure_ring-modulo2: | |
367 | ||
368 | .. figure:: img/ring-modulo2.* | |
369 | ||
370 | Modulo 32-bit indexes - Example 2 | |
371 | ||
372 | ||
373 | This ring contains 12536 entries. | |
374 | ||
375 | .. note:: | |
376 | ||
377 | For ease of understanding, we use modulo 65536 operations in the above examples. | |
378 | In real execution cases, this is redundant for low efficiency, but is done automatically when the result overflows. | |
379 | ||
380 | The code always maintains a distance between producer and consumer between 0 and size(ring)-1. | |
381 | Thanks to this property, we can do subtractions between 2 index values in a modulo-32bit base: | |
382 | that's why the overflow of the indexes is not a problem. | |
383 | ||
384 | At any time, entries and free_entries are between 0 and size(ring)-1, | |
385 | even if only the first term of subtraction has overflowed: | |
386 | ||
387 | .. code-block:: c | |
388 | ||
389 | uint32_t entries = (prod_tail - cons_head); | |
390 | uint32_t free_entries = (mask + cons_tail -prod_head); | |
391 | ||
392 | References | |
393 | ---------- | |
394 | ||
395 | * `bufring.h in FreeBSD <http://svn.freebsd.org/viewvc/base/release/8.0.0/sys/sys/buf_ring.h?revision=199625&view=markup>`_ (version 8) | |
396 | ||
397 | * `bufring.c in FreeBSD <http://svn.freebsd.org/viewvc/base/release/8.0.0/sys/kern/subr_bufring.c?revision=199625&view=markup>`_ (version 8) | |
398 | ||
399 | * `Linux Lockless Ring Buffer Design <http://lwn.net/Articles/340400/>`_ |