README.md

   1 <!--
   2 SPDX-FileCopyrightText: 2023 EfficiOS Inc.
   3
   4 SPDX-License-Identifier: CC-BY-4.0
   5 -->
   6
   7 Userspace RCU Implementation
   8 ============================
   9
  10 by Mathieu Desnoyers and Paul E. McKenney
  11
  12
  13 Building
  14 --------
  15
  16     ./bootstrap # skip if using tarball
  17     ./configure
  18     make
  19     make install
  20     ldconfig
  21
  22 Hints:
  23
  24   - Forcing 32-bit build:
  25
  26         CFLAGS="-m32 -g -O2" ./configure
  27
  28   - Forcing 64-bit build:
  29
  30         CFLAGS="-m64 -g -O2" ./configure
  31
  32   - Forcing a 32-bit build with 386 backward compatibility:
  33
  34         CFLAGS="-m32 -g -O2" ./configure --host=i386-pc-linux-gnu
  35
  36   - Forcing a 32-bit build for Sparcv9 (typical for Sparc v9)
  37
  38         CFLAGS="-m32 -Wa,-Av9a -g -O2" ./configure
  39
  40
  41 Architectures supported
  42 -----------------------
  43
  44 Currently, the following architectures are supported:
  45
  46   - x86 (i386, i486, i586, i686)
  47   - amd64 / x86\_64
  48   - PowerPC 32/64
  49   - S390, S390x
  50   - ARM 32/64
  51   - MIPS
  52   - NIOS2
  53   - Alpha
  54   - ia64
  55   - Sparcv9 32/64
  56   - Tilera
  57   - hppa/PA-RISC
  58   - m68k
  59   - RISC-V
  60   - LoongArch
  61
  62 Tested on:
  63
  64   - Linux all architectures
  65   - FreeBSD 13 i386/amd64
  66   - Cygwin i386/amd64
  67   - MacOS amd64/arm64
  68
  69 Should also work on:
  70
  71   - Android
  72   - NetBSD 5
  73   - OpenBSD
  74   - Solaris
  75
  76 (more testing needed before claiming support for these OS).
  77
  78
  79 Toolchain support
  80 -----------------
  81
  82 The C compiler used needs to support at least C99. The C++ compiler used needs
  83 to support at least C++11. The oldest GCC version officialy supported and
  84 tested is 4.8.
  85
  86 Older GCC versions might still work with the following exceptions:
  87
  88   - GCC 3.3 and 3.4 have a bug that prevents them from generating volatile
  89     accesses to offsets in a TLS structure on 32-bit x86. These versions are
  90     therefore not compatible with `liburcu` on x86 32-bit
  91     (i386, i486, i586, i686).
  92     The problem has been reported to the GCC community:
  93     <http://www.mail-archive.com/gcc-bugs@gcc.gnu.org/msg281255.html>
  94   - GCC 3.3 cannot match the "xchg" instruction on 32-bit x86 build.
  95     See <http://kerneltrap.org/node/7507>
  96   - Alpha, ia64 and ARM architectures depend on GCC 4.x with atomic builtins
  97     support. For ARM this was introduced with GCC 4.4:
  98     <http://gcc.gnu.org/gcc-4.4/changes.html>.
  99   - Linux aarch64 depends on GCC 5.1 or better because prior versions
 100     perform unsafe access to deallocated stack.
 101
 102 Clang version 3.0 (based on LLVM 3.0) is supported.
 103
 104 Glibc >= 2.4 should work but the older version we test against is
 105 currently 2.17.
 106
 107
 108 Build system
 109 ------------
 110
 111 For developers using the Git tree:
 112
 113 This source tree is based on the autotools suite from GNU to simplify
 114 portability. Here are some things you should have on your system in order to
 115 compile the git repository tree :
 116
 117   - GNU autotools (automake >=1.12, autoconf >=2.69)
 118     (make sure your system wide `automake` points to a recent version!)
 119   - GNU Libtool >=2.2
 120     (for more information, go to <http://www.gnu.org/software/autoconf/>)
 121
 122 If you get the tree from the repository, you will need to use the `bootstrap`
 123 script in the root of the tree. It calls all the GNU tools needed to prepare
 124 the tree configuration.
 125
 126 Test scripts provided in the `tests/` directory of the source tree depend
 127 on `bash` and the `seq` program.
 128
 129
 130 API
 131 ---
 132
 133 See the relevant API documentation files in `doc/`. The APIs provided by
 134 Userspace RCU are, by prefix:
 135
 136   - `rcu_`: Read-Copy Update (see [`doc/rcu-api.md`](doc/rcu-api.md))
 137   - `cmm_`: Concurrent Memory Model
 138   - `caa_`: Concurrent Architecture Abstraction
 139   - `cds_`: Concurrent Data Structures
 140     (see [`doc/cds-api.md`](doc/cds-api.md))
 141   - `uatomic_`: Userspace Atomic
 142     (see [`doc/uatomic-api.md`](doc/uatomic-api.md))
 143
 144
 145 Quick start guide
 146 -----------------
 147
 148 ### Usage of all urcu libraries:
 149
 150   - Define `_LGPL_SOURCE` (only) if your code is LGPL or GPL compatible
 151     before including the `urcu.h` or `urcu-qsbr.h` header. If your application
 152     is distributed under another license, function calls will be generated
 153     instead of inlines, so your application can link with the library.
 154   - Linking with one of the libraries below is always necessary even for
 155     LGPL and GPL applications.
 156   - Define `URCU_INLINE_SMALL_FUNCTIONS` before including Userspace RCU
 157     headers if you want Userspace RCU to inline small functions (10
 158     lines or less) into the application. It can be used by applications
 159     distributed under any kind of license, and does *not* make the
 160     application a derived work of Userspace RCU.
 161
 162 Those small inlined functions are guaranteed to match the library
 163 content as long as the library major version is unchanged.
 164 Therefore, the application *must* be compiled with headers matching
 165 the library major version number. Applications using
 166 `URCU_INLINE_SMALL_FUNCTIONS` may be unable to use debugging
 167 features of Userspace RCU without being recompiled.
 168
 169 There are multiple flavors of liburcu available:
 170
 171   - `memb`,
 172   - `qsbr`,
 173   - `mb`,
 174   - `bp`.
 175
 176 The API members start with the prefix `urcu_<flavor>_`, where
 177 `<flavor>` is the chosen flavor name.
 178
 179
 180 ### Usage of `liburcu-memb`
 181
 182   1. `#include <urcu/urcu-memb.h>`
 183   2. Link the application with `-lurcu-memb`
 184
 185 This is the preferred version of the library, in terms of
 186 grace-period detection speed, read-side speed and flexibility.
 187 Dynamically detects kernel support for `sys_membarrier()`. Falls back
 188 on `urcu-mb` scheme if support is not present, which has slower
 189 read-side. Use the `--disable-sys-membarrier-fallback` configure option
 190 to disable the fall back, thus requiring `sys_membarrier()` to be
 191 available. This gives a small speedup when `sys_membarrier()` is
 192 supported by the kernel, and aborts in the library constructor if not
 193 supported.
 194
 195
 196 ### Usage of `liburcu-qsbr`
 197
 198   1. `#include <urcu/urcu-qsbr.h>`
 199   2. Link with `-lurcu-qsbr`
 200
 201 The QSBR flavor of RCU needs to have each reader thread executing
 202 `rcu_quiescent_state()` periodically to progress. `rcu_thread_online()`
 203 and `rcu_thread_offline()` can be used to mark long periods for which
 204 the threads are not active. It provides the fastest read-side at the
 205 expense of more intrusiveness in the application code.
 206
 207
 208 ### Usage of `liburcu-mb`
 209
 210   1. `#include <urcu/urcu-mb.h>`
 211   2. Link with `-lurcu-mb`
 212
 213 This version of the urcu library uses memory barriers on the writer
 214 and reader sides. This results in faster grace-period detection, but
 215 results in slower reads.
 216
 217
 218 ### Usage of `liburcu-bp`
 219
 220   1. `#include <urcu/urcu-bp.h>`
 221   2. Link with `-lurcu-bp`
 222
 223 The BP library flavor stands for "bulletproof". It is specifically
 224 designed to help tracing library to hook on applications without
 225 requiring to modify these applications. `urcu_bp_init()`, and
 226 `urcu_bp_unregister_thread()` all become nops, whereas calling
 227 `urcu_bp_register_thread()` becomes optional. The state is dealt with by
 228 the library internally at the expense of read-side and write-side
 229 performance.
 230
 231
 232 ### Initialization
 233
 234 Each thread that has reader critical sections (that uses
 235 `urcu_<flavor>_read_lock()`/`urcu_<flavor>_read_unlock()` must first
 236 register to the URCU library. This is done by calling
 237 `urcu_<flavor>_register_thread()`. Unregistration must be performed
 238 before exiting the thread by using `urcu_<flavor>_unregister_thread()`.
 239
 240
 241 ### Reading
 242
 243 Reader critical sections must be protected by locating them between
 244 calls to `urcu_<flavor>_read_lock()` and `urcu_<flavor>_read_unlock()`.
 245 Inside that lock, `rcu_dereference()` may be called to read an RCU
 246 protected pointer.
 247
 248
 249 ### Writing
 250
 251 `rcu_assign_pointer()` and `rcu_xchg_pointer()` may be called anywhere.
 252 After, `urcu_<flavor>_synchronize_rcu()` must be called. When it
 253 returns, the old values are not in usage anymore.
 254
 255 As an alternative to `urcu_<flavor>_synchronize_rcu()`,
 256 it is also possible to use the urcu polling mechanism to wait for a
 257 grace period to elapse. This can be done by using
 258 `urcu_<flavor>_start_poll_synchronize_rcu()`
 259 to start the grace period polling, and then invoke
 260 `urcu_<flavor>_poll_state_synchronize_rcu()`, which returns true if
 261 the grace period has completed, false otherwise.
 262
 263
 264 ### Usage of `liburcu-defer`
 265
 266   - Follow instructions for either `liburcu-memb`, `liburcu-qsbr`,
 267     `liburcu-mb`, or `liburcu-bp` above.
 268     The `liburcu-defer` functionality is pulled into each of
 269     those library modules.
 270   - Provides `urcu_<flavor>_defer_rcu()` primitive to enqueue delayed
 271     callbacks. Queued callbacks are executed in batch periodically after
 272     a grace period.  Do _not_ use `urcu_<flavor>_defer_rcu()` within a
 273     read-side critical section, because it may call
 274     `urcu_<flavor>_synchronize_rcu()` if the thread queue is full.  This
 275     can lead to deadlock or worse.
 276   - Requires that `urcu_<flavor>_defer_barrier()` must be called in
 277     library destructor if a library queues callbacks and is expected to
 278     be unloaded with `dlclose()`.
 279
 280 Its API is currently experimental. It may change in future library releases.
 281
 282
 283 ### Usage of `urcu-call-rcu`
 284
 285   - Follow instructions for either `liburcu-memb`, `liburcu-qsbr`,
 286     `liburcu-mb`, or `liburcu-bp` above.
 287     The `urcu-call-rcu` functionality is pulled into each of
 288     those library modules.
 289   - Provides the `urcu_<flavor>_call_rcu()` primitive to enqueue delayed
 290     callbacks in a manner similar to `urcu_<flavor>_defer_rcu()`, but
 291     without ever delaying for a grace period.  On the other hand,
 292     `urcu_<flavor>_call_rcu()`'s best-case overhead is not quite as good
 293     as that of `urcu_<flavor>_defer_rcu()`.
 294   - Provides `urcu_<flavor>_call_rcu()` to allow asynchronous handling
 295     of RCU grace periods.  A number of additional functions are provided
 296     to manage the helper threads used by `urcu_<flavor>_call_rcu()`, but
 297     reasonable defaults are used if these additional functions are not
 298     invoked. See [`doc/rcu-api.md`](doc/rcu-api.md) in userspace-rcu
 299     documentation for more details.
 300
 301
 302 ### Being careful with signals
 303
 304 Read-side critical sections are allowed in a signal handler,
 305 except those setup with `sigaltstack(2)`, with `liburcu-memb` and
 306 `liburcu-mb`. Be careful, however, to disable these signals
 307 between thread creation and calls to `urcu_<flavor>_register_thread()`,
 308 because a signal handler nesting on an unregistered thread would not be
 309 allowed to call `urcu_<flavor>_read_lock()`.
 310
 311 Read-side critical sections are _not_ allowed in a signal handler with
 312 `liburcu-qsbr`, unless signals are disabled explicitly around each
 313 `urcu_qsbr_quiescent_state()` calls, when threads are put offline and around
 314 calls to `urcu_qsbr_synchronize_rcu()`. Even then, we do not recommend it.
 315
 316
 317 ### Interaction with mutexes
 318
 319 One must be careful to do not cause deadlocks due to interaction of
 320 `urcu_<flavor>_synchronize_rcu()` and RCU read-side with mutexes. If
 321 `urcu_<flavor>_synchronize_rcu()` is called with a mutex held, this
 322 mutex (or any mutex which has this mutex in its dependency chain) should
 323 not be acquired from within a RCU read-side critical section.
 324
 325 This is especially important to understand in the context of the
 326 QSBR flavor: a registered reader thread being "online" by
 327 default should be considered as within a RCU read-side critical
 328 section unless explicitly put "offline". Therefore, if
 329 `urcu_qsbr_synchronize_rcu()` is called with a mutex held, this mutex,
 330 as well as any mutex which has this mutex in its dependency chain should
 331 only be taken when the RCU reader thread is "offline" (this can be
 332 performed by calling `urcu_qsbr_thread_offline()`).
 333
 334
 335 ### Interaction with `fork()`
 336
 337 Special care must be taken for applications performing `fork()` without
 338 any following `exec()`. This is caused by the fact that Linux only clones
 339 the thread calling `fork()`, and thus never replicates any of the other
 340 parent thread into the child process. Most `liburcu` implementations
 341 require that all registrations (as reader, `defer_rcu` and `call_rcu`
 342 threads) should be released before a `fork()` is performed, except for the
 343 rather common scenario where `fork()` is immediately followed by `exec()` in
 344 the child process. The only implementation not subject to that rule is
 345 `liburcu-bp`, which is designed to handle `fork()` by calling
 346 `urcu_bp_before_fork`, `urcu_bp_after_fork_parent` and
 347 `urcu_bp_after_fork_child`.
 348
 349 Applications that use `urcu_<flavor>_call_rcu()` and that `fork()`
 350 without doing an immediate `exec()` must take special action.  The
 351 parent must invoke `urcu_<flavor>_call_rcu_before_fork()` before the
 352 `fork()` and `urcu_<flavor>_call_rcu_after_fork_parent()` after the
 353 `fork()`. The child process must invoke
 354 `urcu_<flavor>_call_rcu_after_fork_child()`.  Even though these three
 355 APIs are suitable for passing to `pthread_atfork()`, use of
 356 `pthread_atfork()` is **STRONGLY DISCOURAGED** for programs calling the
 357 glibc memory allocator (`malloc()`, `calloc()`, `free()`, ...) within
 358 `urcu_<flavor>_call_rcu` callbacks.  This is due to limitations in the
 359 way glibc memory allocator handles calls to the memory allocator from
 360 concurrent threads while the `pthread_atfork()` handlers are executing.
 361
 362 Combining e.g.:
 363
 364   - call to `free()` from callbacks executed within
 365     `urcu_<flavor>_call_rcu` worker threads,
 366   - executing `urcu_<flavor>_call_rcu` atfork handlers within the glibc
 367     pthread atfork mechanism,
 368
 369 will sometimes trigger interesting process hangs. This usually
 370 hangs on a memory allocator lock within glibc.
 371
 372
 373 ### Thread Local Storage (TLS)
 374
 375 Userspace RCU can fall back on `pthread_getspecific()` to emulate
 376 TLS variables on systems where it is not available. This behavior
 377 can be forced by specifying `--disable-compiler-tls` as configure
 378 argument.
 379
 380
 381 ### Usage of `DEBUG_RCU` & `--enable-rcu-debug`
 382
 383 By default the library is configured with internal debugging
 384 self-checks disabled.
 385
 386 For always-on debugging self-checks:
 387
 388     ./configure --enable-rcu-debug
 389
 390 For fine grained enabling of debugging self-checks, build
 391 userspace-rcu with `DEBUG_RCU` defined and compile dependent
 392 applications with `DEBUG_RCU` defined when necessary.
 393
 394 Warning: Enabling this feature result in a performance penalty.
 395
 396
 397 ### Usage of `DEBUG_YIELD`
 398
 399 `DEBUG_YIELD` is used to add random delays in the code for testing
 400 purposes.
 401
 402
 403 ### SMP support
 404
 405 By default the library is configured to use synchronization primitives
 406 adequate for SMP systems. On uniprocessor systems, support for SMP
 407 systems can be disabled with:
 408
 409     ./configure --disable-smp-support
 410
 411 theoretically yielding slightly better performance.
 412
 413
 414 ### Usage of `--enable-cds-lfht-iter-debug`
 415
 416 By default the library is configured with extra debugging checks for
 417 lock-free hash table iterator traversal disabled.
 418
 419 Building liburcu with `--enable-cds-lfht-iter-debug` and rebuilding
 420 application to match the ABI change allows finding cases where the hash
 421 table iterator is re-purposed to be used on a different hash table while
 422 still being used to iterate on a hash table.
 423
 424 This option alters the rculfhash ABI. Make sure to compile both library
 425 and application with matching configuration.
 426
 427 ### Usage of `--enable-compiler-atomic-builtins`
 428
 429 Building liburcu with `--enable-compiler-atomic-builtins` implements the uatomic
 430 API with the compiler atomic builtins if supported.
 431
 432 Make targets
 433 ------------
 434
 435 In addition to the usual `make check` target, Userspace RCU features
 436 `make regtest`, `make short_bench` and `make long_bench` targets:
 437
 438   - `make check`: short tests, meant to be run when rebuilding or
 439     porting Userspace RCU.
 440   - `make regtest`: long (many hours) test, meant to be run when
 441     modifying Userspace RCU or porting it to a new architecture or
 442     operating system.
 443   - `make short_bench`: short benchmarks, 3 seconds per test.
 444   - `make long_bench`: long (many hours) benchmarks, 30 seconds per test.
 445
 446
 447 Known issues
 448 ------------
 449
 450 There is an application vs library compatibility issue between
 451 applications built using Userspace RCU 0.10 headers linked against
 452 Userspace RCU 0.11 or 0.12 shared objects. The problem occurs as
 453 follows:
 454
 455   - An application executable is built with `_LGPL_SOURCE` defined, includes
 456     any of the Userspace RCU 0.10 urcu flavor headers, and is built
 457     without the `-fpic` compiler option.
 458
 459   - The Userspace RCU 0.10 library shared objects are updated to 0.11
 460     or 0.12 without rebuilding the application.
 461
 462   - The application will hang, typically when RCU grace period
 463     (synchronize_rcu) is invoked.
 464
 465 Some possible work-arounds for this are:
 466
 467   - Rebuild the application against Userspace RCU 0.11+.
 468
 469   - Rebuild the application with `-fpic`.
 470
 471   - Upgrade Userspace RCU to 0.13+ without installing 0.11 nor 0.12.
 472
 473
 474 Contacts
 475 --------
 476
 477 You can contact the maintainers on the following mailing list:
 478 `lttng-dev@lists.lttng.org`.
 479
 480 IRC channel: [#lttng](irc://irc.oftc.net/lttng) on the OFTC network
 481
 482 Bug tracker: [Userspace RCU bug tracker](https://bugs.lttng.org/projects/urcu)
 483
 484 Code review: [_userspace-rcu_ project](https://review.lttng.org/q/project:userspace-rcu) on LTTng Review
 485
 486 Continuous integration: [Userspace RCU](https://ci.lttng.org/view/Liburcu/) on LTTng's CI
 487
 488 GitHub mirror: [urcu/userspace-rcu](https://github.com/urcu/userspace-rcu)
 489
 490 Patches are principally submitted and reviewed on [LTTng Review](https://review.lttng.org),
 491 but may also be submitted to the [mailing list](mailto:lttng-dev@lists.lttng.org)
 492 with the subject prefix `PATCH urcu` or by pull request on the
 493 [GitHub mirror](https://github.com/urcu/userspace-rcu).