README.md

   1 Userspace RCU Implementation
   2 ============================
   3
   4 by Mathieu Desnoyers and Paul E. McKenney
   5
   6
   7 Building
   8 --------
   9
  10     ./bootstrap # skip if using tarball
  11     ./configure
  12     make
  13     make install
  14     ldconfig
  15
  16 Hints:
  17
  18   - Forcing 32-bit build:
  19
  20         CFLAGS="-m32 -g -O2" ./configure
  21
  22   - Forcing 64-bit build:
  23
  24         CFLAGS="-m64 -g -O2" ./configure
  25
  26   - Forcing a 32-bit build with 386 backward compatibility:
  27
  28         CFLAGS="-m32 -g -O2" ./configure --host=i386-pc-linux-gnu
  29
  30   - Forcing a 32-bit build for Sparcv9 (typical for Sparc v9)
  31
  32         CFLAGS="-m32 -Wa,-Av9a -g -O2" ./configure
  33
  34
  35 Architectures supported
  36 -----------------------
  37
  38 Currently, the following architectures are supported:
  39
  40   - Linux x86 (i386, i486, i586, i686)
  41   - x86 64-bit
  42   - PowerPC 32/64
  43   - S390, S390x
  44   - ARM 32/64
  45   - MIPS
  46   - NIOS2
  47   - Alpha
  48   - ia64
  49   - Sparcv9 32/64
  50   - Tilera
  51   - hppa/PA-RISC
  52
  53 Tested on Linux, FreeBSD 8.2/8.3/9.0/9.1/10.0 i386/amd64, and Cygwin.
  54 Should also work on:
  55
  56   - Android
  57   - NetBSD 5
  58   - OpenBSD
  59   - Darwin
  60
  61 (more testing needed before claiming support for these OS).
  62
  63 Linux ARM depends on running a Linux kernel 2.6.15 or better, GCC 4.4 or
  64 better.
  65
  66 The GCC compiler versions 3.3, 3.4, 4.0, 4.1, 4.2, 4.3, 4.4 and 4.5 are
  67 supported, with the following exceptions:
  68
  69   - GCC 3.3 and 3.4 have a bug that prevents them from generating volatile
  70     accesses to offsets in a TLS structure on 32-bit x86. These versions are
  71     therefore not compatible with `liburcu` on x86 32-bit
  72     (i386, i486, i586, i686).
  73     The problem has been reported to the GCC community:
  74     http://www.mail-archive.com/gcc-bugs@gcc.gnu.org/msg281255.html
  75   - GCC 3.3 cannot match the "xchg" instruction on 32-bit x86 build.
  76     See http://kerneltrap.org/node/7507
  77   - Alpha, ia64 and ARM architectures depend on GCC 4.x with atomic builtins
  78     support. For ARM this was introduced with GCC 4.4:
  79     http://gcc.gnu.org/gcc-4.4/changes.html.
  80
  81 Clang version 3.0 (based on LLVM 3.0) is supported.
  82
  83 Building on MacOS X (Darwin) requires a work-around for processor
  84 detection:
  85
  86   - 32-bit:
  87
  88         ./configure --build=i686-apple-darwin11
  89
  90   - 64-bit:
  91
  92         ./configure --build=x86_64-apple-darwin11
  93
  94 For developers using the Git tree:
  95
  96 This source tree is based on the autotools suite from GNU to simplify
  97 portability. Here are some things you should have on your system in order to
  98 compile the git repository tree :
  99
 100   - GNU autotools (automake >=1.10, autoconf >=2.50, autoheader >=2.50)
 101     (make sure your system wide `automake` points to a recent version!)
 102   - GNU Libtool >=2.2
 103     (for more information, go to http://www.gnu.org/software/autoconf/)
 104
 105 If you get the tree from the repository, you will need to use the `bootstrap`
 106 script in the root of the tree. It calls all the GNU tools needed to prepare
 107 the tree configuration.
 108
 109 Test scripts provided in the `tests/` directory of the source tree depend
 110 on `bash` and the `seq` program.
 111
 112
 113 API
 114 ---
 115
 116 See the relevant API documentation files in `doc/`. The APIs provided by
 117 Userspace RCU are, by prefix:
 118
 119   - `rcu_`: Read-Copy Update (see [`doc/rcu-api.md`](doc/rcu-api.md))
 120   - `cmm_`: Concurrent Memory Model
 121   - `caa_`: Concurrent Architecture Abstraction
 122   - `cds_`: Concurrent Data Structures
 123     (see [`doc/cds-api.md`](doc/cds-api.md))
 124   - `uatomic_`: Userspace Atomic
 125     (see [`doc/uatomic-api.md`](doc/uatomic-api.md))
 126
 127
 128 Quick start guide
 129 -----------------
 130
 131 ### Usage of all urcu libraries:
 132
 133   - Define `_LGPL_SOURCE` (only) if your code is LGPL or GPL compatible
 134     before including the `urcu.h` or `urcu-qsbr.h` header. If your application
 135     is distributed under another license, function calls will be generated
 136     instead of inlines, so your application can link with the library.
 137   - Linking with one of the libraries below is always necessary even for
 138     LGPL and GPL applications.
 139   - Define `URCU_INLINE_SMALL_FUNCTIONS` before including Userspace RCU
 140     headers if you want Userspace RCU to inline small functions (10
 141     lines or less) into the application. It can be used by applications
 142     distributed under any kind of license, and does *not* make the
 143     application a derived work of Userspace RCU.
 144
 145 Those small inlined functions are guaranteed to match the library
 146 content as long as the library major version is unchanged.
 147 Therefore, the application *must* be compiled with headers matching
 148 the library major version number. Applications using
 149 `URCU_INLINE_SMALL_FUNCTIONS` may be unable to use debugging
 150 features of Userspace RCU without being recompiled.
 151
 152
 153 ### Usage of `liburcu`
 154
 155   1. `#include <urcu.h>`
 156   2. Link the application with `-lurcu`
 157
 158 This is the preferred version of the library, in terms of
 159 grace-period detection speed, read-side speed and flexibility.
 160 Dynamically detects kernel support for `sys_membarrier()`. Falls back
 161 on `urcu-mb` scheme if support is not present, which has slower
 162 read-side. Use the --disable-sys-membarrier-fallback configure option
 163 to disable the fall back, thus requiring `sys_membarrier()` to be
 164 available. This gives a small speedup when `sys_membarrier()` is
 165 supported by the kernel, and aborts in the library constructor if not
 166 supported.
 167
 168
 169 ### Usage of `liburcu-qsbr`
 170
 171   1. `#include <urcu-qsbr.h>`
 172   2. Link with `-lurcu-qsbr`
 173
 174 The QSBR flavor of RCU needs to have each reader thread executing
 175 `rcu_quiescent_state()` periodically to progress. `rcu_thread_online()`
 176 and `rcu_thread_offline()` can be used to mark long periods for which
 177 the threads are not active. It provides the fastest read-side at the
 178 expense of more intrusiveness in the application code.
 179
 180
 181 ### Usage of `liburcu-mb`
 182
 183   1. `#include <urcu.h>`
 184   2. Compile any `_LGPL_SOURCE` code using this library with `-DRCU_MB`
 185   3. Link with `-lurcu-mb`
 186
 187 This version of the urcu library uses memory barriers on the writer
 188 and reader sides. This results in faster grace-period detection, but
 189 results in slower reads.
 190
 191
 192 ### Usage of `liburcu-signal`
 193
 194   1. `#include <urcu.h>`
 195   2. Compile any `_LGPL_SOURCE` code using this library with `-DRCU_SIGNAL`
 196   3. Link the application with `-lurcu-signal`
 197
 198 Version of the library that requires a signal, typically `SIGUSR1`. Can
 199 be overridden with `-DSIGRCU` by modifying `Makefile.build.inc`.
 200
 201
 202 ### Usage of `liburcu-bp`
 203
 204   1. `#include <urcu-bp.h>`
 205   2. Link with `-lurcu-bp`
 206
 207 The BP library flavor stands for "bulletproof". It is specifically
 208 designed to help tracing library to hook on applications without
 209 requiring to modify these applications. `rcu_init()`,
 210 `rcu_register_thread()` and `rcu_unregister_thread()` all become nops.
 211 The state is dealt with by the library internally at the expense of
 212 read-side and write-side performance.
 213
 214
 215 ### Initialization
 216
 217 Each thread that has reader critical sections (that uses
 218 `rcu_read_lock()`/`rcu_read_unlock()` must first register to the URCU
 219 library. This is done by calling `rcu_register_thread()`. Unregistration
 220 must be performed before exiting the thread by using
 221 `rcu_unregister_thread()`.
 222
 223
 224 ### Reading
 225
 226 Reader critical sections must be protected by locating them between
 227 calls to `rcu_read_lock()` and `rcu_read_unlock()`. Inside that lock,
 228 `rcu_dereference()` may be called to read an RCU protected pointer.
 229
 230
 231 ### Writing
 232
 233 `rcu_assign_pointer()` and `rcu_xchg_pointer()` may be called anywhere.
 234 After, `synchronize_rcu()` must be called. When it returns, the old
 235 values are not in usage anymore.
 236
 237
 238 ### Usage of `liburcu-defer`
 239
 240   - Follow instructions for either `liburcu`, `liburcu-qsbr`,
 241     `liburcu-mb`, `liburcu-signal`, or `liburcu-bp` above.
 242     The `liburcu-defer` functionality is pulled into each of
 243     those library modules.
 244   - Provides `defer_rcu()` primitive to enqueue delayed callbacks. Queued
 245     callbacks are executed in batch periodically after a grace period.
 246     Do _not_ use `defer_rcu()` within a read-side critical section, because
 247     it may call `synchronize_rcu()` if the thread queue is full.
 248     This can lead to deadlock or worse.
 249   - Requires that `rcu_defer_barrier()` must be called in library destructor
 250     if a library queues callbacks and is expected to be unloaded with
 251     `dlclose()`.
 252
 253 Its API is currently experimental. It may change in future library releases.
 254
 255
 256 ### Usage of `urcu-call-rcu`
 257
 258   - Follow instructions for either `liburcu`, `liburcu-qsbr`,
 259     `liburcu-mb`, `liburcu-signal`, or `liburcu-bp` above.
 260     The `urcu-call-rcu` functionality is pulled into each of
 261     those library modules.
 262   - Provides the `call_rcu()` primitive to enqueue delayed callbacks
 263     in a manner similar to `defer_rcu()`, but without ever delaying
 264     for a grace period.  On the other hand, `call_rcu()`'s best-case
 265     overhead is not quite as good as that of `defer_rcu()`.
 266   - Provides `call_rcu()` to allow asynchronous handling of RCU
 267     grace periods.  A number of additional functions are provided
 268     to manage the helper threads used by `call_rcu()`, but reasonable
 269     defaults are used if these additional functions are not invoked.
 270     See [`doc/rcu-api.md`](doc/rcu-api.md) in userspace-rcu documentation
 271     for more details.
 272
 273
 274 ### Being careful with signals
 275
 276 The `liburcu` library uses signals internally. The signal handler is
 277 registered with the `SA_RESTART` flag. However, these signals may cause
 278 some non-restartable system calls to fail with `errno = EINTR`. Care
 279 should be taken to restart system calls manually if they fail with this
 280 error. A list of non-restartable system calls may be found in
 281 `signal(7)`. The `liburcu-mb` and `liburcu-qsbr` versions of the Userspace RCU
 282 library do not require any signal.
 283
 284 Read-side critical sections are allowed in a signal handler,
 285 except those setup with `sigaltstack(2)`, with `liburcu` and
 286 `liburcu-mb`. Be careful, however, to disable these signals
 287 between thread creation and calls to `rcu_register_thread()`, because a
 288 signal handler nesting on an unregistered thread would not be
 289 allowed to call `rcu_read_lock()`.
 290
 291 Read-side critical sections are _not_ allowed in a signal handler with
 292 `liburcu-qsbr`, unless signals are disabled explicitly around each
 293 `rcu_quiescent_state()` calls, when threads are put offline and around
 294 calls to `synchronize_rcu()`. Even then, we do not recommend it.
 295
 296
 297 ### Interaction with mutexes
 298
 299 One must be careful to do not cause deadlocks due to interaction of
 300 `synchronize_rcu()` and RCU read-side with mutexes. If `synchronize_rcu()`
 301 is called with a mutex held, this mutex (or any mutex which has this
 302 mutex in its dependency chain) should not be acquired from within a RCU
 303 read-side critical section.
 304
 305 This is especially important to understand in the context of the
 306 QSBR flavor: a registered reader thread being "online" by
 307 default should be considered as within a RCU read-side critical
 308 section unless explicitly put "offline". Therefore, if
 309 `synchronize_rcu()` is called with a mutex held, this mutex, as
 310 well as any mutex which has this mutex in its dependency chain
 311 should only be taken when the RCU reader thread is "offline"
 312 (this can be performed by calling `rcu_thread_offline()`).
 313
 314
 315 ### Interaction with `fork()`
 316
 317 Special care must be taken for applications performing `fork()` without
 318 any following `exec()`. This is caused by the fact that Linux only clones
 319 the thread calling `fork()`, and thus never replicates any of the other
 320 parent thread into the child process. Most `liburcu` implementations
 321 require that all registrations (as reader, `defer_rcu` and `call_rcu`
 322 threads) should be released before a `fork()` is performed, except for the
 323 rather common scenario where `fork()` is immediately followed by `exec()` in
 324 the child process. The only implementation not subject to that rule is
 325 `liburcu-bp`, which is designed to handle `fork()` by calling
 326 `rcu_bp_before_fork`, `rcu_bp_after_fork_parent` and
 327 `rcu_bp_after_fork_child`.
 328
 329 Applications that use `call_rcu()` and that `fork()` without
 330 doing an immediate `exec()` must take special action.  The parent
 331 must invoke `call_rcu_before_fork()` before the `fork()` and
 332 `call_rcu_after_fork_parent()` after the `fork()`. The child
 333 process must invoke `call_rcu_after_fork_child()`.
 334 Even though these three APIs are suitable for passing to
 335 `pthread_atfork()`, use of `pthread_atfork()` is **STRONGLY
 336 DISCOURAGED** for programs calling the glibc memory allocator
 337 (`malloc()`, `calloc()`, `free()`, ...) within `call_rcu` callbacks.
 338 This is due to limitations in the way glibc memory allocator
 339 handles calls to the memory allocator from concurrent threads
 340 while the `pthread_atfork()` handlers are executing.
 341
 342 Combining e.g.:
 343
 344   - call to `free()` from callbacks executed within `call_rcu` worker
 345     threads,
 346   - executing `call_rcu` atfork handlers within the glibc pthread
 347     atfork mechanism,
 348
 349 will sometimes trigger interesting process hangs. This usually
 350 hangs on a memory allocator lock within glibc.
 351
 352
 353 ### Thread Local Storage (TLS)
 354
 355 Userspace RCU can fall back on `pthread_getspecific()` to emulate
 356 TLS variables on systems where it is not available. This behavior
 357 can be forced by specifying `--disable-compiler-tls` as configure
 358 argument.
 359
 360
 361 ### Usage of `DEBUG_RCU` & `--enable-rcu-debug`
 362
 363 By default the library is configured with internal debugging
 364 self-checks disabled.
 365
 366 For always-on debugging self-checks:
 367         ./configure --enable-rcu-debug
 368
 369 For fine grained enabling of debugging self-checks, build
 370 urserspace-rcu with DEBUG_RCU defined and compile dependent
 371 applications with DEBUG_RCU defined when necessary.
 372
 373 Warning: Enabling this feature result in a performance penalty.
 374
 375
 376 ### Usage of `DEBUG_YIELD`
 377
 378 `DEBUG_YIELD` is used to add random delays in the code for testing
 379 purposes.
 380
 381
 382 ### SMP support
 383
 384 By default the library is configured to use synchronization primitives
 385 adequate for SMP systems. On uniprocessor systems, support for SMP
 386 systems can be disabled with:
 387
 388     ./configure --disable-smp-support
 389
 390 theoretically yielding slightly better performance.
 391
 392
 393 Make targets
 394 ------------
 395
 396 In addition to the usual `make check` target, Userspace RCU features
 397 `make regtest` and `make bench` targets:
 398
 399   - `make check`: short tests, meant to be run when rebuilding or
 400     porting Userspace RCU.
 401   - `make regtest`: long (many hours) test, meant to be run when
 402     modifying Userspace RCU or porting it to a new architecture or
 403     operating system.
 404   - `make bench`: long (many hours) benchmarks.
 405
 406
 407 Contacts
 408 --------
 409
 410 You can contact the maintainers on the following mailing list:
 411 `lttng-dev@lists.lttng.org`.