Userspace RCU Implementation by Mathieu Desnoyers and Paul E. McKenney BUILDING -------- ./bootstrap (skip if using tarball) ./configure make make install Hints: Forcing 32-bit build: * CFLAGS="-m32 -g -O2" ./configure Forcing 64-bit build: * CFLAGS="-m64 -g -O2" ./configure Forcing a 32-bit build with 386 backward compatibility: * CFLAGS="-m32 -g -O2" ./configure --host=i386-pc-linux-gnu Forcing a 32-bit build for Sparcv9 (typical for Sparc v9) * CFLAGS="-m32 -Wa,-Av9a -g -O2" ./configure ARCHITECTURES SUPPORTED ----------------------- Currently, x86 (i386, i486, i586, i686), x86 64-bit, PowerPC 32/64, S390, S390x ARMv7l, Alpha, ia64 and Sparcv9 32/64 are supported. Only tested on Linux so far, but should theoretically work on other operating systems. ARMv7l depends on running a Linux kernel 2.6.15 or better. The gcc compiler versions 3.3, 3.4, 4.0, 4.1, 4.2, 4.3, 4.4 and 4.5 are supported, with the following exceptions: - gcc 3.3 and 3.4 have a bug that prevents them from generating volatile accesses to offsets in a TLS structure on 32-bit x86. These versions are therefore not compatible with liburcu on x86 32-bit (i386, i486, i586, i686). The problem has been reported to the gcc community: http://www.mail-archive.com/gcc-bugs@gcc.gnu.org/msg281255.html - Alpha, ia64 and ARMv7l architectures depend on 4.x gcc with atomic builtins support. QUICK START GUIDE ----------------- Usage of all urcu libraries * Define _LGPL_SOURCE (only) if your code is LGPL or GPL compatible before including the urcu.h or urcu-qsbr.h header. If your application is distributed under another license, function calls will be generated instead of inlines, so your application can link with the library. * Linking with one of the libraries below is always necessary even for LGPL and GPL applications. Usage of liburcu * #include * Link the application with "-lurcu". * This is the preferred version of the library, in terms of grace-period detection speed, read-side speed and flexibility. Dynamically detects kernel support for sys_membarrier(). Falls back on urcu-mb scheme if support is not present, which has slower read-side. Usage of liburcu-qsbr * #include * Link with "-lurcu-qsbr". * The QSBR flavor of RCU needs to have each reader thread executing rcu_quiescent_state() periodically to progress. rcu_thread_online() and rcu_thread_offline() can be used to mark long periods for which the threads are not active. It provides the fastest read-side at the expense of more intrusiveness in the application code. Usage of liburcu-mb * #include * Compile any _LGPL_SOURCE code using this library with "-DRCU_MB". * Link with "-lurcu-mb". * This version of the urcu library uses memory barriers on the writer and reader sides. This results in faster grace-period detection, but results in slower reads. Usage of liburcu-signal * #include * Compile any _LGPL_SOURCE code using this library with "-DRCU_SIGNAL". * Link the application with "-lurcu-signal". * Version of the library that requires a signal, typically SIGUSR1. Can be overridden with -DSIGRCU by modifying Makefile.build.inc. Usage of liburcu-bp * #include * Link with "-lurcu-bp". * The BP library flavor stands for "bulletproof". It is specifically designed to help tracing library to hook on applications without requiring to modify these applications. rcu_init(), rcu_register_thread() and rcu_unregister_thread() all become nops. The state is dealt with by the library internally at the expense of read-side and write-side performance. Initialization Each thread that has reader critical sections (that uses rcu_read_lock()/rcu_read_unlock() must first register to the URCU library. This is done by calling rcu_register_thread(). Unregistration must be performed before exiting the thread by using rcu_unregister_thread(). Reading Reader critical sections must be protected by locating them between calls to rcu_read_lock() and rcu_read_unlock(). Inside that lock, rcu_dereference() may be called to read an RCU protected pointer. Writing rcu_assign_pointer() and rcu_xchg_pointer() may be called anywhere. After, synchronize_rcu() must be called. When it returns, the old values are not in usage anymore. Usage of liburcu-defer * #include * Link with "-lurcu-defer", and also with one of the urcu library (either urcu, urcu-bp, urcu-mb or urcu-qsbr). * Provides defer_rcu() primitive to enqueue delayed callbacks. Queued callbacks are executed in batch periodically after a grace period. Do _not_ use defer_rcu() within a read-side critical section, because it may call synchronize_rcu() if the thread queue is full. * Provides defer_rcu_ratelimit() primitive, which acts just like defer_rcu(), but takes an additional rate limiter callback forcing synchronized callback execution of the limiter returns non-zero. * Requires that rcu_defer_barrier() must be called in library destructor if a library queues callbacks and is expected to be unloaded with dlclose(). * Its API is currently experimental. It may change in future library releases. Being careful with signals The liburcu library uses signals internally. The signal handler is registered with the SA_RESTART flag. However, these signals may cause some non-restartable system calls to fail with errno = EINTR. Care should be taken to restart system calls manually if they fail with this error. A list of non-restartable system calls may be found in signal(7). The liburcu-mb and liburcu-qsbr versions of the Userspace RCU library do not require any signal. Read-side critical sections are allowed in a signal handler with liburcu and liburcu-mb. Be careful, however, to disable these signals between thread creation and calls to rcu_register_thread(), because a signal handler nesting on an unregistered thread would not be allowed to call rcu_read_lock(). Read-side critical sections are _not_ allowed in a signal handler with liburcu-qsbr, unless signals are disabled explicitly around each rcu_quiescent_state() calls, when threads are put offline and around calls to synchronize_rcu(). Even then, we do not recommend it. Interaction with mutexes One must be careful to do not cause deadlocks due to interaction of synchronize_rcu() and RCU read-side with mutexes. If synchronize_rcu() is called with a mutex held, this mutex (or any mutex which has this mutex in its dependency chain) should not be acquired from within a RCU read-side critical section. Usage of DEBUG_RCU DEBUG_RCU is used to add internal debugging self-checks to the RCU library. This define adds a performance penalty when enabled. Can be enabled by uncommenting the corresponding line in Makefile.build.inc. Usage of DEBUG_YIELD DEBUG_YIELD is used to add random delays in the code for testing purposes. SMP support By default the library is configured to use synchronization primitives adequate for SMP systems. On uniprocessor systems, support for SMP systems can be disabled with: ./configure --disable-smp-support theoretically yielding slightly better performance.