Rev 48 | Rev 62 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
9 | bondari | 1 | <?xml version="1.0" encoding="UTF-8"?> |
41 | jermar | 2 | <chapter id="sync"> |
3 | <?dbhtml filename="sync.html"?> |
||
9 | bondari | 4 | |
41 | jermar | 5 | <title>Mutual exclusion and synchronization</title> |
9 | bondari | 6 | |
41 | jermar | 7 | <section> |
8 | <title>Introduction</title> |
||
9 | bondari | 9 | |
45 | jermar | 10 | <para>The HelenOS operating system is designed to make use of the |
11 | parallelism offered by the hardware and to exploit concurrency of both the |
||
12 | kernel and userspace tasks. This is achieved through multiprocessor |
||
13 | support and several levels of multiprogramming such as multitasking, |
||
14 | multithreading and also through userspace pseudo threads. However, such a |
||
15 | highly concurrent environment needs safe and efficient ways to handle |
||
16 | mutual exclusion and synchronization of many execution flows.</para> |
||
41 | jermar | 17 | </section> |
18 | |||
19 | <section> |
||
20 | <title>Active kernel primitives</title> |
||
21 | |||
9 | bondari | 22 | <section> |
41 | jermar | 23 | <title>Spinlocks</title> |
9 | bondari | 24 | |
45 | jermar | 25 | <para>The basic mutual exclusion primitive is the spinlock. The spinlock |
26 | implements active waiting for the availability of a memory lock (i.e. |
||
41 | jermar | 27 | simple variable) in a multiprocessor-safe manner. This safety is |
28 | achieved through the use of a specialized, architecture-dependent, |
||
29 | atomic test-and-set operation which either locks the spinlock (i.e. sets |
||
30 | the variable) or, provided that it is already locked, leaves it |
||
31 | unaltered. In any case, the test-and-set operation returns a value, thus |
||
32 | signalling either success (i.e. zero return value) or failure (i.e. |
||
45 | jermar | 33 | non-zero value) in acquiring the lock. Note that this makes a |
41 | jermar | 34 | fundamental difference between the naive algorithm that doesn't use the |
35 | atomic operation and the spinlock algortihm. While the naive algorithm |
||
45 | jermar | 36 | is prone to race conditions on SMP configurations and thus is completely |
41 | jermar | 37 | SMP-unsafe, the spinlock algorithm eliminates the possibility of race |
38 | conditions and is suitable for mutual exclusion use.</para> |
||
9 | bondari | 39 | |
41 | jermar | 40 | <para>The semantics of the test-and-set operation is that the spinlock |
41 | remains unavailable until this operation called on the respective |
||
45 | jermar | 42 | spinlock returns zero. HelenOS builds two functions on top of the |
43 | test-and-set operation. The first function is the unconditional attempt |
||
57 | jermar | 44 | to acquire the spinlock and is called <code>spinlock_lock</code>. It |
45 | simply loops until the test-and-set returns a zero value. The other |
||
46 | function, <code>spinlock_trylock</code>, is the conditional lock |
||
47 | operation and calls the test-and-set only once to find out whether it |
||
48 | managed to acquire the spinlock or not. The conditional operation is |
||
49 | useful in situations in which an algorithm cannot acquire more spinlocks |
||
50 | in the proper order and a deadlock cannot be avoided. In such a case, |
||
51 | the algorithm would detect the danger and instead of possibly |
||
52 | deadlocking the system it would simply release some spinlocks it already |
||
53 | holds and retry the whole operation with the hope that it will succeed |
||
54 | next time. The unlock function, <code>spinlock_unlock</code>, is quite |
||
55 | easy - it merely clears the spinlock variable.</para> |
||
9 | bondari | 56 | |
41 | jermar | 57 | <para>Nevertheless, there is a special issue related to hardware |
45 | jermar | 58 | optimizations that modern processors implement. Particularly problematic |
59 | is the out-of-order execution of instructions within the critical |
||
60 | section protected by a spinlock. The processors are always |
||
41 | jermar | 61 | self-consistent so that they can carry out speculatively executed |
62 | instructions in the right order with regard to dependencies among those |
||
63 | instructions. However, the dependency between instructions inside the |
||
64 | critical section and those that implement locking and unlocking of the |
||
45 | jermar | 65 | respective spinlock is not implicit on some processor architectures. As |
66 | a result, the processor needs to be explicitly told about each |
||
67 | occurrence of such a dependency. Therefore, HelenOS adds |
||
57 | jermar | 68 | architecture-specific hooks to all <code>spinlock_lock</code>, |
69 | <code>spinlock_trylock</code> and <code>spinlock_unlock</code> functions |
||
70 | to prevent the instructions inside the critical section from permeating |
||
71 | out. On some architectures, these hooks can be void because the |
||
72 | dependencies are implicitly there because of the special properties of |
||
73 | locking and unlocking instructions. However, other architectures need to |
||
74 | instrument these hooks with different memory barriers, depending on what |
||
75 | operations could permeate out.</para> |
||
9 | bondari | 76 | |
41 | jermar | 77 | <para>Spinlocks have one significant drawback: when held for longer time |
45 | jermar | 78 | periods, they harm both parallelism and concurrency. The processor |
57 | jermar | 79 | executing <code>spinlock_lock</code> does not do any fruitful work and |
80 | is effectively halted until it can grab the lock and proceed. |
||
45 | jermar | 81 | Similarily, other execution flows cannot execute on the processor that |
82 | holds the spinlock because the kernel disables preemption on that |
||
83 | processor when a spinlock is held. The reason behind disabling |
||
84 | preemption is priority inversion problem avoidance. For the same reason, |
||
85 | threads are strongly discouraged from sleeping when they hold a |
||
86 | spinlock.</para> |
||
9 | bondari | 87 | |
41 | jermar | 88 | <para>To summarize, spinlocks represent very simple and essential mutual |
89 | exclusion primitive for SMP systems. On the other hand, spinlocks scale |
||
90 | poorly because of the active loop they are based on. Therefore, |
||
45 | jermar | 91 | spinlocks are used in HelenOS only for short-time mutual exclusion and |
41 | jermar | 92 | in cases where the mutual exclusion is required out of thread context. |
93 | Lastly, spinlocks are used in the construction of passive |
||
94 | synchronization primitives.</para> |
||
95 | </section> |
||
96 | </section> |
||
9 | bondari | 97 | |
41 | jermar | 98 | <section> |
99 | <title>Passive kernel synchronization</title> |
||
9 | bondari | 100 | |
41 | jermar | 101 | <section> |
43 | jermar | 102 | <title>Wait queues</title> |
9 | bondari | 103 | |
43 | jermar | 104 | <para>A wait queue is the basic passive synchronization primitive on |
45 | jermar | 105 | which all other passive synchronization primitives are built. Simply |
106 | put, it allows a thread to sleep until an event associated with the |
||
107 | particular wait queue occurs. Multiple threads are notified about |
||
108 | incoming events in a first come, first served fashion. Moreover, should |
||
109 | the event come before any thread waits for it, it is recorded in the |
||
110 | wait queue as a missed wakeup and later forwarded to the first thread |
||
111 | that decides to wait in the queue. The inner structures of the wait |
||
112 | queue are protected by a spinlock.</para> |
||
43 | jermar | 113 | |
114 | <para>The thread that wants to wait for a wait queue event uses the |
||
57 | jermar | 115 | <code>waitq_sleep_timeout</code> function. The algorithm then checks the |
116 | wait queue's counter of missed wakeups and if there are any missed |
||
117 | wakeups, the call returns immediately. The call also returns immediately |
||
118 | if only a conditional wait was requested. Otherwise the thread is |
||
119 | enqueued in the wait queue's list of sleeping threads and its state is |
||
120 | changed to <constant>Sleeping</constant>. It then sleeps until one of |
||
121 | the following events happens:</para> |
||
43 | jermar | 122 | |
123 | <orderedlist> |
||
124 | <listitem> |
||
57 | jermar | 125 | <para>another thread calls <code>waitq_wakeup</code> and the thread |
126 | is the first thread in the wait queue's list of sleeping |
||
45 | jermar | 127 | threads;</para> |
43 | jermar | 128 | </listitem> |
129 | |||
130 | <listitem> |
||
57 | jermar | 131 | <para>another thread calls <code>waitq_interrupt_sleep</code> on the |
132 | sleeping thread;</para> |
||
43 | jermar | 133 | </listitem> |
134 | |||
135 | <listitem> |
||
45 | jermar | 136 | <para>the sleep times out provided that none of the previous |
137 | occurred within a specified time limit; the limit can be |
||
138 | infinity.</para> |
||
43 | jermar | 139 | </listitem> |
140 | </orderedlist> |
||
141 | |||
142 | <para>All five possibilities (immediate return on success, immediate |
||
143 | return on failure, wakeup after sleep, interruption and timeout) are |
||
57 | jermar | 144 | distinguishable by the return value of <code>waitq_sleep_timeout</code>. |
145 | Being able to interrupt a sleeping thread is essential for externally |
||
146 | initiated thread termination. The ability to wait only for a certain |
||
147 | amount of time is used, for instance, to passively delay thread |
||
148 | execution by several microseconds or even seconds in |
||
149 | <code>thread_sleep</code> function. Due to the fact that all other |
||
150 | passive kernel synchronization primitives are based on wait queues, they |
||
151 | also have the option of being interrutped and, more importantly, can |
||
152 | timeout. All of them also implement the conditional operation. |
||
153 | Furthemore, this very fundamental interface reaches up to the |
||
154 | implementation of futexes - userspace synchronization primitive, which |
||
155 | makes it possible for a userspace thread to request a synchronization |
||
156 | operation with a timeout or a conditional operation.</para> |
||
43 | jermar | 157 | |
158 | <para>From the description above, it should be apparent, that when a |
||
57 | jermar | 159 | sleeping thread is woken by <code>waitq_wakeup</code> or when |
160 | <code>waitq_sleep_timeout</code> succeeds immediately, the thread can be |
||
161 | sure that the event has occurred. The thread need not and should not |
||
162 | verify this fact. This approach is called direct hand-off and is |
||
163 | characteristic for all passive HelenOS synchronization primitives, with |
||
164 | the exception as described below.</para> |
||
41 | jermar | 165 | </section> |
9 | bondari | 166 | |
41 | jermar | 167 | <section> |
168 | <title>Semaphores</title> |
||
9 | bondari | 169 | |
43 | jermar | 170 | <para>The interesting point about wait queues is that the number of |
171 | missed wakeups is equal to the number of threads that will not block in |
||
57 | jermar | 172 | <code>watiq_sleep_timeout</code> and would immediately succeed instead. |
173 | On the other hand, semaphores are synchronization primitives that will |
||
174 | let predefined amount of threads into their critical section and block |
||
175 | any other threads above this count. However, these two cases are exactly |
||
176 | the same. Semaphores in HelenOS are therefore implemented as wait queues |
||
177 | with a single semantic change: their wait queue is initialized to have |
||
178 | so many missed wakeups as is the number of threads that the semphore |
||
179 | intends to let into its critical section simultaneously.</para> |
||
43 | jermar | 180 | |
181 | <para>In the semaphore language, the wait queue operation |
||
57 | jermar | 182 | <code>waitq_sleep_timeout</code> corresponds to semaphore |
183 | <code>down</code> operation, represented by the function |
||
184 | <code>semaphore_down_timeout</code> and by way of similitude the wait |
||
185 | queue operation waitq_wakeup corresponds to semaphore <code>up</code> |
||
186 | operation, represented by the function <code>sempafore_up</code>. The |
||
187 | conditional down operation is called |
||
188 | <code>semaphore_trydown</code>.</para> |
||
41 | jermar | 189 | </section> |
9 | bondari | 190 | |
41 | jermar | 191 | <section> |
43 | jermar | 192 | <title>Mutexes</title> |
9 | bondari | 193 | |
45 | jermar | 194 | <para>Mutexes are sometimes referred to as binary sempahores. That means |
195 | that mutexes are like semaphores that allow only one thread in its |
||
43 | jermar | 196 | critical section. Indeed, mutexes in HelenOS are implemented exactly in |
45 | jermar | 197 | this way: they are built on top of semaphores. From another point of |
198 | view, they can be viewed as spinlocks without busy waiting. Their |
||
199 | semaphore heritage provides good basics for both conditional operation |
||
200 | and operation with timeout. The locking operation is called |
||
57 | jermar | 201 | <code>mutex_lock</code>, the conditional locking operation is called |
202 | <code>mutex_trylock</code> and the unlocking operation is called |
||
203 | <code>mutex_unlock</code>.</para> |
||
41 | jermar | 204 | </section> |
9 | bondari | 205 | |
41 | jermar | 206 | <section> |
43 | jermar | 207 | <title>Reader/writer locks</title> |
9 | bondari | 208 | |
43 | jermar | 209 | <para>Reader/writer locks, or rwlocks, are by far the most complicated |
210 | synchronization primitive within the kernel. The goal of these locks is |
||
45 | jermar | 211 | to improve concurrency of applications, in which threads need to |
212 | synchronize access to a shared resource, and that access can be |
||
43 | jermar | 213 | partitioned into a read-only mode and a write mode. Reader/writer locks |
214 | should make it possible for several, possibly many, readers to enter the |
||
215 | critical section, provided that no writer is currently in the critical |
||
216 | section, or to be in the critical section contemporarily. Writers are |
||
217 | allowed to enter the critical section only individually, provided that |
||
45 | jermar | 218 | no reader is in the critical section already. Applications, in which the |
219 | majority of operations can be done in the read-only mode, can benefit |
||
43 | jermar | 220 | from increased concurrency created by reader/writer locks.</para> |
221 | |||
45 | jermar | 222 | <para>During reader/writer lock construction, a decision should be made |
43 | jermar | 223 | whether readers will be prefered over writers or whether writers will be |
224 | prefered over readers in cases when the lock is not currently held and |
||
225 | both a reader and a writer want to gain the lock. Some operating systems |
||
226 | prefer one group over the other, creating thus a possibility for |
||
227 | starving the unprefered group. In the HelenOS operating system, none of |
||
45 | jermar | 228 | the two groups is prefered. The lock is granted on a first come, first |
43 | jermar | 229 | served basis with the additional note that readers are granted the lock |
45 | jermar | 230 | in the biggest possible batch.</para> |
43 | jermar | 231 | |
232 | <para>With this policy and the timeout modes of operation, the direct |
||
233 | hand-off becomes much more complicated. For instance, a writer leaving |
||
234 | the critical section must wake up all leading readers in the rwlock's |
||
235 | wait queue or one leading writer or no-one if no thread is waiting. |
||
236 | Similarily, the last reader leaving the critical section must wakeup the |
||
45 | jermar | 237 | sleeping writer if there are any sleeping threads left at all. As |
238 | another example, if a writer at the beginning of the rwlock's wait queue |
||
239 | times out and the lock is held by at least one reader, the writer which |
||
240 | has timed out must first wake up all readers that follow him in the |
||
241 | queue prior to signalling the timeout itself and giving up.</para> |
||
43 | jermar | 242 | |
45 | jermar | 243 | <para>Due to the issues mentioned in the previous paragraph, the |
244 | reader/writer lock imlpementation needs to walk the rwlock wait queue's |
||
245 | list of sleeping threads directly, in order to find out the type of |
||
43 | jermar | 246 | access that the queueing threads demand. This makes the code difficult |
247 | to understand and dependent on the internal implementation of the wait |
||
248 | queue. Nevertheless, it remains unclear to the authors of HelenOS |
||
249 | whether a simpler but equivalently fair solution exists.</para> |
||
250 | |||
251 | <para>The implementation of rwlocks as it has been already put, makes |
||
252 | use of one single wait queue for both readers and writers, thus avoiding |
||
253 | any possibility of starvation. In fact, rwlocks use a mutex rather than |
||
57 | jermar | 254 | a bare wait queue. This mutex is called <code>exclusive</code> and is |
255 | used to synchronize writers. The writer's lock operation, |
||
256 | <code>rwlock_write_lock_timeout</code>, simply tries to acquire the |
||
257 | exclusive mutex. If it succeeds, the writer is granted the rwlock. |
||
44 | jermar | 258 | However, if the operation fails (e.g. times out), the writer must check |
259 | for potential readers at the head of the list of sleeping threads |
||
45 | jermar | 260 | associated with the mutex's wait queue and then proceed according to the |
44 | jermar | 261 | procedure outlined above.</para> |
43 | jermar | 262 | |
263 | <para>The exclusive mutex plays an important role in reader |
||
264 | synchronization as well. However, a reader doing the reader's lock |
||
57 | jermar | 265 | operation, <code>rwlock_read_lock_timeout</code>, may bypass this mutex |
266 | when it detects that:</para> |
||
43 | jermar | 267 | |
268 | <orderedlist> |
||
269 | <listitem> |
||
45 | jermar | 270 | <para>there are other readers in the critical section and</para> |
43 | jermar | 271 | </listitem> |
272 | |||
273 | <listitem> |
||
274 | <para>there are no sleeping threads waiting for the exclusive |
||
45 | jermar | 275 | mutex.</para> |
43 | jermar | 276 | </listitem> |
277 | </orderedlist> |
||
278 | |||
279 | <para>If both conditions are true, the reader will bypass the mutex, |
||
45 | jermar | 280 | increment the number of readers in the critical section and then enter |
281 | the critical section. Note that if there are any sleeping threads at the |
||
282 | beginning of the wait queue, the first must be a writer. If the |
||
43 | jermar | 283 | conditions are not fulfilled, the reader normally waits until the |
284 | exclusive mutex is granted to it.</para> |
||
41 | jermar | 285 | </section> |
9 | bondari | 286 | |
287 | <section> |
||
41 | jermar | 288 | <title>Condition variables</title> |
9 | bondari | 289 | |
48 | jermar | 290 | <para>Condition variables can be used for waiting until a condition |
291 | becomes true. In this respect, they are similar to wait queues. But |
||
292 | contrary to wait queues, condition variables have different semantics |
||
293 | that allows events to be lost when there is no thread waiting for them. |
||
294 | In order to support this, condition variables don't use direct hand-off |
||
295 | and operate in a way similar to the example below. A thread waiting for |
||
296 | the condition becoming true does the following:</para> |
||
297 | |||
298 | <para><programlisting language="C"><function>mutex_lock</function>(<varname>mtx</varname>); |
||
299 | while (!<varname>condition</varname>) |
||
300 | <function>condvar_wait_timeout</function>(<varname>cv</varname>, <varname>mtx</varname>); |
||
301 | /* <remark>the condition is true, do something</remark> */ |
||
302 | <function>mutex_unlock</function>(<varname>mtx</varname>);</programlisting></para> |
||
303 | |||
304 | <para>A thread that causes the condition become true signals this event |
||
305 | like this:</para> |
||
306 | |||
307 | <para><programlisting><function>mutex_lock</function>(<varname>mtx</varname>); |
||
308 | <varname>condition</varname> = <constant>true</constant>; |
||
309 | <function>condvar_signal</function>(<varname>cv</varname>); /* <remark>condvar_broadcast(cv);</remark> */ |
||
310 | <function>mutex_unlock</function>(<varname>mtx</varname>);</programlisting></para> |
||
311 | |||
57 | jermar | 312 | <para>The wait operation, <code>condvar_wait_timeout</code>, always puts |
313 | the calling thread to sleep. The thread then sleeps until another thread |
||
314 | invokes <code>condvar_broadcast</code> on the same condition variable or |
||
315 | until it is woken up by <code>condvar_signal</code>. The |
||
316 | <code>condvar_signal</code> operation unblocks the first thread blocking |
||
317 | on the condition variable while the <code>condvar_broadcast</code> |
||
318 | operation unblocks all threads blocking there. If there are no blocking |
||
319 | threads, these two operations have no efect.</para> |
||
48 | jermar | 320 | |
321 | <para>Note that the threads must synchronize over a dedicated mutex. To |
||
57 | jermar | 322 | prevent race condition between <code>condvar_wait_timeout</code> and |
323 | <code>condvar_signal</code> or <code>condvar_broadcast</code>, the mutex |
||
324 | is passed to <code>condvar_wait_timeout</code> which then atomically |
||
325 | puts the calling thread asleep and unlocks the mutex. When the thread |
||
326 | eventually wakes up, <code>condvar_wait</code> regains the mutex and |
||
48 | jermar | 327 | returns.</para> |
328 | |||
329 | <para>Also note, that there is no conditional operation for condition |
||
330 | variables. Such an operation would make no sence since condition |
||
331 | variables are defined to forget events for which there is no waiting |
||
57 | jermar | 332 | thread and because <code>condvar_wait</code> must always go to sleep. |
333 | The operation with timeout is supported as usually.</para> |
||
48 | jermar | 334 | |
335 | <para>In HelenOS, condition variables are based on wait queues. As it is |
||
336 | already mentioned above, wait queues remember missed events while |
||
337 | condition variables must not do so. This is reasoned by the fact that |
||
338 | condition variables are designed for scenarios in which an event might |
||
339 | occur very many times without being picked up by any waiting thread. On |
||
340 | the other hand, wait queues would remember any event that had not been |
||
57 | jermar | 341 | picked up by a call to <code>waitq_sleep_timeout</code>. Therefore, if |
342 | wait queues were used directly and without any changes to implement |
||
343 | condition variables, the missed_wakeup counter would hurt performance of |
||
344 | the implementation: the <code>while</code> loop in |
||
345 | <code>condvar_wait_timeout</code> would effectively do busy waiting |
||
346 | until all missed wakeups were discarded.</para> |
||
48 | jermar | 347 | |
348 | <para>The requirement on the wait operation to atomically put the caller |
||
349 | to sleep and release the mutex poses an interesting problem on |
||
57 | jermar | 350 | <code>condvar_wait_timeout</code>. More precisely, the thread should |
351 | sleep in the condvar's wait queue prior to releasing the mutex, but it |
||
352 | must not hold the mutex when it is sleeping.</para> |
||
48 | jermar | 353 | |
354 | <para>Problems described in the two previous paragraphs are addressed in |
||
57 | jermar | 355 | HelenOS by dividing the <code>waitq_sleep_timeout</code> function into |
356 | three pieces:</para> |
||
48 | jermar | 357 | |
358 | <orderedlist> |
||
359 | <listitem> |
||
57 | jermar | 360 | <para><code>waitq_sleep_prepare</code> prepares the thread to go to |
361 | sleep by, among other things, locking the wait queue;</para> |
||
48 | jermar | 362 | </listitem> |
363 | |||
364 | <listitem> |
||
57 | jermar | 365 | <para><code>waitq_sleep_timeout_unsafe</code> implements the core |
366 | blocking logic;</para> |
||
48 | jermar | 367 | </listitem> |
368 | |||
369 | <listitem> |
||
57 | jermar | 370 | <para><code>waitq_sleep_finish</code> performs cleanup after |
371 | <code>waitq_sleep_timeout_unsafe</code>; after this call, the wait |
||
372 | queue spinlock is guaranteed to be unlocked by the caller</para> |
||
48 | jermar | 373 | </listitem> |
374 | </orderedlist> |
||
375 | |||
57 | jermar | 376 | <para>The stock <code>waitq_sleep_timeout</code> is then a mere wrapper |
377 | that calls these three functions. It is provided for convenience in |
||
378 | cases where the caller doesn't require such a low level control. |
||
379 | However, the implementation of <code>condvar_wait_timeout</code> does |
||
380 | need this finer-grained control because it has to interleave calls to |
||
381 | these functions by other actions. It carries its operations out in the |
||
382 | following order:</para> |
||
48 | jermar | 383 | |
384 | <orderedlist> |
||
385 | <listitem> |
||
57 | jermar | 386 | <para>calls <code>waitq_sleep_prepare</code> in order to lock the |
387 | condition variable's wait queue,</para> |
||
48 | jermar | 388 | </listitem> |
389 | |||
390 | <listitem> |
||
391 | <para>releases the mutex,</para> |
||
392 | </listitem> |
||
393 | |||
394 | <listitem> |
||
395 | <para>clears the counter of missed wakeups,</para> |
||
396 | </listitem> |
||
397 | |||
398 | <listitem> |
||
57 | jermar | 399 | <para>calls <code>waitq_sleep_timeout_unsafe</code>,</para> |
48 | jermar | 400 | </listitem> |
401 | |||
402 | <listitem> |
||
403 | <para>retakes the mutex,</para> |
||
404 | </listitem> |
||
405 | |||
406 | <listitem> |
||
57 | jermar | 407 | <para>calls <code>waitq_sleep_finish</code>.</para> |
48 | jermar | 408 | </listitem> |
409 | </orderedlist> |
||
9 | bondari | 410 | </section> |
41 | jermar | 411 | </section> |
9 | bondari | 412 | |
41 | jermar | 413 | <section> |
414 | <title>Userspace synchronization</title> |
||
9 | bondari | 415 | |
41 | jermar | 416 | <section> |
417 | <title>Futexes</title> |
||
418 | |||
419 | <para></para> |
||
420 | </section> |
||
421 | </section> |
||
422 | </chapter> |