rculfhash: document max_nr_buckets = 0
[userspace-rcu.git] / README
1 Userspace RCU Implementation
2 by Mathieu Desnoyers and Paul E. McKenney
3
4 BUILDING
5 --------
6
7 ./bootstrap (skip if using tarball)
8 ./configure
9 make
10 make install
11 ldconfig
12
13 Hints: Forcing 32-bit build:
14 * CFLAGS="-m32 -g -O2" ./configure
15
16 Forcing 64-bit build:
17 * CFLAGS="-m64 -g -O2" ./configure
18
19 Forcing a 32-bit build with 386 backward compatibility:
20 * CFLAGS="-m32 -g -O2" ./configure --host=i386-pc-linux-gnu
21
22 Forcing a 32-bit build for Sparcv9 (typical for Sparc v9)
23 * CFLAGS="-m32 -Wa,-Av9a -g -O2" ./configure
24
25
26 ARCHITECTURES SUPPORTED
27 -----------------------
28
29 Currently, Linux x86 (i386, i486, i586, i686), x86 64-bit, PowerPC 32/64,
30 S390, S390x, ARM, MIPS, Alpha, ia64 and Sparcv9 32/64 are supported.
31 Tested on Linux, FreeBSD 8.2/8.3/9.0/9.1/10.0 i386/amd64, and Cygwin.
32 Should also work on: Android, NetBSD 5, OpenBSD, Darwin (more testing
33 needed before claiming support for these OS).
34
35 Linux ARM depends on running a Linux kernel 2.6.15 or better, GCC 4.4 or
36 better.
37
38 The gcc compiler versions 3.3, 3.4, 4.0, 4.1, 4.2, 4.3, 4.4 and 4.5 are
39 supported, with the following exceptions:
40
41 - gcc 3.3 and 3.4 have a bug that prevents them from generating volatile
42 accesses to offsets in a TLS structure on 32-bit x86. These versions are
43 therefore not compatible with liburcu on x86 32-bit (i386, i486, i586, i686).
44 The problem has been reported to the gcc community:
45 http://www.mail-archive.com/gcc-bugs@gcc.gnu.org/msg281255.html
46 - gcc 3.3 cannot match the "xchg" instruction on 32-bit x86 build.
47 See: http://kerneltrap.org/node/7507
48 - Alpha, ia64 and ARM architectures depend on gcc 4.x with atomic builtins
49 support. For ARM this was introduced with gcc 4.4:
50 http://gcc.gnu.org/gcc-4.4/changes.html
51
52 Clang version 3.0 (based on LLVM 3.0) is supported.
53
54 Building on MacOS X (Darwin) requires a work-around for processor
55 detection:
56 # 32-bit
57 ./configure --build=i686-apple-darwin11
58 # 64-bit
59 ./configure --build=x86_64-apple-darwin11
60
61 For developers using the git tree:
62
63 This source tree is based on the autotools suite from GNU to simplify
64 portability. Here are some things you should have on your system in order to
65 compile the git repository tree :
66
67 - GNU autotools (automake >=1.10, autoconf >=2.50, autoheader >=2.50)
68 (make sure your system wide "automake" points to a recent version!)
69 - GNU Libtool >=2.2
70 (for more information, go to http://www.gnu.org/software/autoconf/)
71
72 If you get the tree from the repository, you will need to use the "bootstrap"
73 script in the root of the tree. It calls all the GNU tools needed to prepare the
74 tree configuration.
75
76 Test scripts provided in the tests/ directory of the source tree depend
77 on "bash" and the "seq" program.
78
79
80 API
81 ---
82
83 See the relevant API documentation files in doc/. The APIs provided by
84 Userspace RCU are, by prefix:
85
86 - rcu_ : Read-Copy Update (see doc/rcu-api.txt)
87 - cmm_ : Concurrent Memory Model
88 - caa_ : Concurrent Architecture Abstraction
89 - cds_ : Concurrent Data Structures (see doc/cds-api.txt)
90 - uatomic_: Userspace Atomic (see doc/uatomic-api.txt)
91
92
93 QUICK START GUIDE
94 -----------------
95
96 Usage of all urcu libraries
97
98 * Define _LGPL_SOURCE (only) if your code is LGPL or GPL compatible
99 before including the urcu.h or urcu-qsbr.h header. If your application
100 is distributed under another license, function calls will be generated
101 instead of inlines, so your application can link with the library.
102 * Linking with one of the libraries below is always necessary even for
103 LGPL and GPL applications.
104 * Define URCU_INLINE_SMALL_FUNCTIONS before including Userspace RCU
105 headers if you want Userspace RCU to inline small functions (10
106 lines or less) into the application. It can be used by applications
107 distributed under any kind of license, and does *not* make the
108 application a derived work of Userspace RCU.
109
110 Those small inlined functions are guaranteed to match the library
111 content as long as the library major version is unchanged.
112 Therefore, the application *must* be compiled with headers matching
113 the library major version number. Applications using
114 URCU_INLINE_SMALL_FUNCTIONS may be unable to use debugging
115 features of Userspace RCU without being recompiled.
116
117
118 Usage of liburcu
119
120 * #include <urcu.h>
121 * Link the application with "-lurcu".
122 * This is the preferred version of the library, in terms of
123 grace-period detection speed, read-side speed and flexibility.
124 Dynamically detects kernel support for sys_membarrier(). Falls back
125 on urcu-mb scheme if support is not present, which has slower
126 read-side.
127
128 Usage of liburcu-qsbr
129
130 * #include <urcu-qsbr.h>
131 * Link with "-lurcu-qsbr".
132 * The QSBR flavor of RCU needs to have each reader thread executing
133 rcu_quiescent_state() periodically to progress. rcu_thread_online()
134 and rcu_thread_offline() can be used to mark long periods for which
135 the threads are not active. It provides the fastest read-side at the
136 expense of more intrusiveness in the application code.
137
138 Usage of liburcu-mb
139
140 * #include <urcu.h>
141 * Compile any _LGPL_SOURCE code using this library with "-DRCU_MB".
142 * Link with "-lurcu-mb".
143 * This version of the urcu library uses memory barriers on the writer
144 and reader sides. This results in faster grace-period detection, but
145 results in slower reads.
146
147 Usage of liburcu-signal
148
149 * #include <urcu.h>
150 * Compile any _LGPL_SOURCE code using this library with "-DRCU_SIGNAL".
151 * Link the application with "-lurcu-signal".
152 * Version of the library that requires a signal, typically SIGUSR1. Can
153 be overridden with -DSIGRCU by modifying Makefile.build.inc.
154
155 Usage of liburcu-bp
156
157 * #include <urcu-bp.h>
158 * Link with "-lurcu-bp".
159 * The BP library flavor stands for "bulletproof". It is specifically
160 designed to help tracing library to hook on applications without
161 requiring to modify these applications. rcu_init(),
162 rcu_register_thread() and rcu_unregister_thread() all become nops.
163 The state is dealt with by the library internally at the expense of
164 read-side and write-side performance.
165
166 Initialization
167
168 Each thread that has reader critical sections (that uses
169 rcu_read_lock()/rcu_read_unlock() must first register to the URCU
170 library. This is done by calling rcu_register_thread(). Unregistration
171 must be performed before exiting the thread by using
172 rcu_unregister_thread().
173
174 Reading
175
176 Reader critical sections must be protected by locating them between
177 calls to rcu_read_lock() and rcu_read_unlock(). Inside that lock,
178 rcu_dereference() may be called to read an RCU protected pointer.
179
180 Writing
181
182 rcu_assign_pointer() and rcu_xchg_pointer() may be called anywhere.
183 After, synchronize_rcu() must be called. When it returns, the old
184 values are not in usage anymore.
185
186 Usage of liburcu-defer
187
188 * Follow instructions for either liburcu, liburcu-qsbr,
189 liburcu-mb, liburcu-signal, or liburcu-bp above.
190 The liburcu-defer functionality is pulled into each of
191 those library modules.
192 * Provides defer_rcu() primitive to enqueue delayed callbacks. Queued
193 callbacks are executed in batch periodically after a grace period.
194 Do _not_ use defer_rcu() within a read-side critical section, because
195 it may call synchronize_rcu() if the thread queue is full.
196 This can lead to deadlock or worse.
197 * Requires that rcu_defer_barrier() must be called in library destructor
198 if a library queues callbacks and is expected to be unloaded with
199 dlclose().
200 * Its API is currently experimental. It may change in future library
201 releases.
202
203 Usage of urcu-call-rcu
204
205 * Follow instructions for either liburcu, liburcu-qsbr,
206 liburcu-mb, liburcu-signal, or liburcu-bp above.
207 The urcu-call-rcu functionality is provided for each of
208 these library modules.
209 * Provides the call_rcu() primitive to enqueue delayed callbacks
210 in a manner similar to defer_rcu(), but without ever delaying
211 for a grace period. On the other hand, call_rcu()'s best-case
212 overhead is not quite as good as that of defer_rcu().
213 * Provides call_rcu() to allow asynchronous handling of RCU
214 grace periods. A number of additional functions are provided
215 to manage the helper threads used by call_rcu(), but reasonable
216 defaults are used if these additional functions are not invoked.
217 See rcu-api.txt in userspace-rcu documentation for more details.
218
219 Being careful with signals
220
221 The liburcu library uses signals internally. The signal handler is
222 registered with the SA_RESTART flag. However, these signals may cause
223 some non-restartable system calls to fail with errno = EINTR. Care
224 should be taken to restart system calls manually if they fail with this
225 error. A list of non-restartable system calls may be found in
226 signal(7). The liburcu-mb and liburcu-qsbr versions of the Userspace RCU
227 library do not require any signal.
228
229 Read-side critical sections are allowed in a signal handler,
230 except those setup with sigaltstack(2), with liburcu and
231 liburcu-mb. Be careful, however, to disable these signals
232 between thread creation and calls to rcu_register_thread(), because a
233 signal handler nesting on an unregistered thread would not be
234 allowed to call rcu_read_lock().
235
236 Read-side critical sections are _not_ allowed in a signal handler with
237 liburcu-qsbr, unless signals are disabled explicitly around each
238 rcu_quiescent_state() calls, when threads are put offline and around
239 calls to synchronize_rcu(). Even then, we do not recommend it.
240
241 Interaction with mutexes
242
243 One must be careful to do not cause deadlocks due to interaction of
244 synchronize_rcu() and RCU read-side with mutexes. If synchronize_rcu()
245 is called with a mutex held, this mutex (or any mutex which has this
246 mutex in its dependency chain) should not be acquired from within a RCU
247 read-side critical section.
248
249 This is especially important to understand in the context of the
250 QSBR flavor: a registered reader thread being "online" by
251 default should be considered as within a RCU read-side critical
252 section unless explicitly put "offline". Therefore, if
253 synchronize_rcu() is called with a mutex held, this mutex, as
254 well as any mutex which has this mutex in its dependency chain
255 should only be taken when the RCU reader thread is "offline"
256 (this can be performed by calling rcu_thread_offline()).
257
258 Interaction with fork()
259
260 Special care must be taken for applications performing fork() without
261 any following exec(). This is caused by the fact that Linux only clones
262 the thread calling fork(), and thus never replicates any of the other
263 parent thread into the child process. Most liburcu implementations
264 require that all registrations (as reader, defer_rcu and call_rcu
265 threads) should be released before a fork() is performed, except for the
266 rather common scenario where fork() is immediately followed by exec() in
267 the child process. The only implementation not subject to that rule is
268 liburcu-bp, which is designed to handle fork() by calling
269 rcu_bp_before_fork, rcu_bp_after_fork_parent and
270 rcu_bp_after_fork_child.
271
272 Applications that use call_rcu() and that fork() without
273 doing an immediate exec() must take special action. The parent
274 must invoke call_rcu_before_fork() before the fork() and
275 call_rcu_after_fork_parent() after the fork(). The child
276 process must invoke call_rcu_after_fork_child().
277 Even though these three APIs are suitable for passing to
278 pthread_atfork(), use of pthread_atfork() is *STRONGLY
279 DISCOURAGED* for programs calling the glibc memory allocator
280 (malloc(), calloc(), free(), ...) within call_rcu callbacks.
281 This is due to limitations in the way glibc memory allocator
282 handles calls to the memory allocator from concurrent threads
283 while the pthread_atfork() handlers are executing.
284 Combining e.g.:
285 * call to free() from callbacks executed within call_rcu worker
286 threads,
287 * executing call_rcu atfork handlers within the glibc pthread
288 atfork mechanism,
289 will sometimes trigger interesting process hangs. This usually
290 hangs on a memory allocator lock within glibc.
291
292 Thread Local Storage (TLS)
293
294 Userspace RCU can fall back on pthread_getspecific() to emulate
295 TLS variables on systems where it is not available. This behavior
296 can be forced by specifying --disable-compiler-tls as configure
297 argument.
298
299 Usage of DEBUG_RCU
300
301 DEBUG_RCU is used to add internal debugging self-checks to the
302 RCU library. This define adds a performance penalty when enabled.
303 Can be enabled by uncommenting the corresponding line in
304 Makefile.build.inc.
305
306 Usage of DEBUG_YIELD
307
308 DEBUG_YIELD is used to add random delays in the code for testing
309 purposes.
310
311 SMP support
312
313 By default the library is configured to use synchronization primitives
314 adequate for SMP systems. On uniprocessor systems, support for SMP
315 systems can be disabled with:
316
317 ./configure --disable-smp-support
318
319 theoretically yielding slightly better performance.
320
321 MAKE TARGETS
322 ------------
323
324 In addition to the usual "make check" target, Userspace RCU features
325 "make regtest" and "make bench" targets.
326
327 make check: Short tests, meant to be run when rebuilding or porting
328 Userspace RCU.
329
330 make regtest: Long (many hours) test, meant to be run when modifying
331 Userspace RCU or porting it to a new architecture or
332 operating system.
333
334 make bench: Long (many hours) benchmarks.
This page took 0.035709 seconds and 4 git commands to generate.