util/rcu.c

   1 /*
   2  * urcu-mb.c
   3  *
   4  * Userspace RCU library with explicit memory barriers
   5  *
   6  * Copyright (c) 2009 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
   7  * Copyright (c) 2009 Paul E. McKenney, IBM Corporation.
   8  * Copyright 2015 Red Hat, Inc.
   9  *
  10  * Ported to QEMU by Paolo Bonzini  <pbonzini@redhat.com>
  11  *
  12  * This library is free software; you can redistribute it and/or
  13  * modify it under the terms of the GNU Lesser General Public
  14  * License as published by the Free Software Foundation; either
  15  * version 2.1 of the License, or (at your option) any later version.
  16  *
  17  * This library is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  20  * Lesser General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU Lesser General Public
  23  * License along with this library; if not, write to the Free Software
  24  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  25  *
  26  * IBM's contributions to this file may be relicensed under LGPLv2 or later.
  27  */
  28
  29 #include "qemu/osdep.h"
  30 #include "qemu/rcu.h"
  31 #include "qemu/atomic.h"
  32 #include "qemu/thread.h"
  33 #include "qemu/main-loop.h"
  34 #include "qemu/lockable.h"
  35 #if defined(CONFIG_MALLOC_TRIM)
  36 #include <malloc.h>
  37 #endif
  38
  39 /*
  40  * Global grace period counter.  Bit 0 is always one in rcu_gp_ctr.
  41  * Bits 1 and above are defined in synchronize_rcu.
  42  */
  43 #define RCU_GP_LOCKED           (1UL << 0)
  44 #define RCU_GP_CTR              (1UL << 1)
  45
  46 unsigned long rcu_gp_ctr = RCU_GP_LOCKED;
  47
  48 QemuEvent rcu_gp_event;
  49 static int in_drain_call_rcu;
  50 static QemuMutex rcu_registry_lock;
  51 static QemuMutex rcu_sync_lock;
  52
  53 /*
  54  * Check whether a quiescent state was crossed between the beginning of
  55  * update_counter_and_wait and now.
  56  */
  57 static inline int rcu_gp_ongoing(unsigned long *ctr)
  58 {
  59     unsigned long v;
  60
  61     v = qatomic_read(ctr);
  62     return v && (v != rcu_gp_ctr);
  63 }
  64
  65 /* Written to only by each individual reader. Read by both the reader and the
  66  * writers.
  67  */
  68 QEMU_DEFINE_CO_TLS(struct rcu_reader_data, rcu_reader)
  69
  70 /* Protected by rcu_registry_lock.  */
  71 typedef QLIST_HEAD(, rcu_reader_data) ThreadList;
  72 static ThreadList registry = QLIST_HEAD_INITIALIZER(registry);
  73
  74 /* Wait for previous parity/grace period to be empty of readers.  */
  75 static void wait_for_readers(void)
  76 {
  77     ThreadList qsreaders = QLIST_HEAD_INITIALIZER(qsreaders);
  78     struct rcu_reader_data *index, *tmp;
  79
  80     for (;;) {
  81         /* We want to be notified of changes made to rcu_gp_ongoing
  82          * while we walk the list.
  83          */
  84         qemu_event_reset(&rcu_gp_event);
  85
  86         /* Instead of using qatomic_mb_set for index->waiting, and
  87          * qatomic_mb_read for index->ctr, memory barriers are placed
  88          * manually since writes to different threads are independent.
  89          * qemu_event_reset has acquire semantics, so no memory barrier
  90          * is needed here.
  91          */
  92         QLIST_FOREACH(index, &registry, node) {
  93             qatomic_set(&index->waiting, true);
  94         }
  95
  96         /* Here, order the stores to index->waiting before the loads of
  97          * index->ctr.  Pairs with smp_mb_placeholder() in rcu_read_unlock(),
  98          * ensuring that the loads of index->ctr are sequentially consistent.
  99          */
 100         smp_mb_global();
 101
 102         QLIST_FOREACH_SAFE(index, &registry, node, tmp) {
 103             if (!rcu_gp_ongoing(&index->ctr)) {
 104                 QLIST_REMOVE(index, node);
 105                 QLIST_INSERT_HEAD(&qsreaders, index, node);
 106
 107                 /* No need for mb_set here, worst of all we
 108                  * get some extra futex wakeups.
 109                  */
 110                 qatomic_set(&index->waiting, false);
 111             } else if (qatomic_read(&in_drain_call_rcu)) {
 112                 notifier_list_notify(&index->force_rcu, NULL);
 113             }
 114         }
 115
 116         if (QLIST_EMPTY(&registry)) {
 117             break;
 118         }
 119
 120         /* Wait for one thread to report a quiescent state and try again.
 121          * Release rcu_registry_lock, so rcu_(un)register_thread() doesn't
 122          * wait too much time.
 123          *
 124          * rcu_register_thread() may add nodes to &registry; it will not
 125          * wake up synchronize_rcu, but that is okay because at least another
 126          * thread must exit its RCU read-side critical section before
 127          * synchronize_rcu is done.  The next iteration of the loop will
 128          * move the new thread's rcu_reader from &registry to &qsreaders,
 129          * because rcu_gp_ongoing() will return false.
 130          *
 131          * rcu_unregister_thread() may remove nodes from &qsreaders instead
 132          * of &registry if it runs during qemu_event_wait.  That's okay;
 133          * the node then will not be added back to &registry by QLIST_SWAP
 134          * below.  The invariant is that the node is part of one list when
 135          * rcu_registry_lock is released.
 136          */
 137         qemu_mutex_unlock(&rcu_registry_lock);
 138         qemu_event_wait(&rcu_gp_event);
 139         qemu_mutex_lock(&rcu_registry_lock);
 140     }
 141
 142     /* put back the reader list in the registry */
 143     QLIST_SWAP(&registry, &qsreaders, node);
 144 }
 145
 146 void synchronize_rcu(void)
 147 {
 148     QEMU_LOCK_GUARD(&rcu_sync_lock);
 149
 150     /* Write RCU-protected pointers before reading p_rcu_reader->ctr.
 151      * Pairs with smp_mb_placeholder() in rcu_read_lock().
 152      */
 153     smp_mb_global();
 154
 155     QEMU_LOCK_GUARD(&rcu_registry_lock);
 156     if (!QLIST_EMPTY(&registry)) {
 157         /* In either case, the qatomic_mb_set below blocks stores that free
 158          * old RCU-protected pointers.
 159          */
 160         if (sizeof(rcu_gp_ctr) < 8) {
 161             /* For architectures with 32-bit longs, a two-subphases algorithm
 162              * ensures we do not encounter overflow bugs.
 163              *
 164              * Switch parity: 0 -> 1, 1 -> 0.
 165              */
 166             qatomic_mb_set(&rcu_gp_ctr, rcu_gp_ctr ^ RCU_GP_CTR);
 167             wait_for_readers();
 168             qatomic_mb_set(&rcu_gp_ctr, rcu_gp_ctr ^ RCU_GP_CTR);
 169         } else {
 170             /* Increment current grace period.  */
 171             qatomic_mb_set(&rcu_gp_ctr, rcu_gp_ctr + RCU_GP_CTR);
 172         }
 173
 174         wait_for_readers();
 175     }
 176 }
 177
 178
 179 #define RCU_CALL_MIN_SIZE        30
 180
 181 /* Multi-producer, single-consumer queue based on urcu/static/wfqueue.h
 182  * from liburcu.  Note that head is only used by the consumer.
 183  */
 184 static struct rcu_head dummy;
 185 static struct rcu_head *head = &dummy, **tail = &dummy.next;
 186 static int rcu_call_count;
 187 static QemuEvent rcu_call_ready_event;
 188
 189 static void enqueue(struct rcu_head *node)
 190 {
 191     struct rcu_head **old_tail;
 192
 193     node->next = NULL;
 194     old_tail = qatomic_xchg(&tail, &node->next);
 195     qatomic_mb_set(old_tail, node);
 196 }
 197
 198 static struct rcu_head *try_dequeue(void)
 199 {
 200     struct rcu_head *node, *next;
 201
 202 retry:
 203     /* Test for an empty list, which we do not expect.  Note that for
 204      * the consumer head and tail are always consistent.  The head
 205      * is consistent because only the consumer reads/writes it.
 206      * The tail, because it is the first step in the enqueuing.
 207      * It is only the next pointers that might be inconsistent.
 208      */
 209     if (head == &dummy && qatomic_mb_read(&tail) == &dummy.next) {
 210         abort();
 211     }
 212
 213     /* If the head node has NULL in its next pointer, the value is
 214      * wrong and we need to wait until its enqueuer finishes the update.
 215      */
 216     node = head;
 217     next = qatomic_mb_read(&head->next);
 218     if (!next) {
 219         return NULL;
 220     }
 221
 222     /* Since we are the sole consumer, and we excluded the empty case
 223      * above, the queue will always have at least two nodes: the
 224      * dummy node, and the one being removed.  So we do not need to update
 225      * the tail pointer.
 226      */
 227     head = next;
 228
 229     /* If we dequeued the dummy node, add it back at the end and retry.  */
 230     if (node == &dummy) {
 231         enqueue(node);
 232         goto retry;
 233     }
 234
 235     return node;
 236 }
 237
 238 static void *call_rcu_thread(void *opaque)
 239 {
 240     struct rcu_head *node;
 241
 242     rcu_register_thread();
 243
 244     for (;;) {
 245         int tries = 0;
 246         int n = qatomic_read(&rcu_call_count);
 247
 248         /* Heuristically wait for a decent number of callbacks to pile up.
 249          * Fetch rcu_call_count now, we only must process elements that were
 250          * added before synchronize_rcu() starts.
 251          */
 252         while (n == 0 || (n < RCU_CALL_MIN_SIZE && ++tries <= 5)) {
 253             g_usleep(10000);
 254             if (n == 0) {
 255                 qemu_event_reset(&rcu_call_ready_event);
 256                 n = qatomic_read(&rcu_call_count);
 257                 if (n == 0) {
 258 #if defined(CONFIG_MALLOC_TRIM)
 259                     malloc_trim(4 * 1024 * 1024);
 260 #endif
 261                     qemu_event_wait(&rcu_call_ready_event);
 262                 }
 263             }
 264             n = qatomic_read(&rcu_call_count);
 265         }
 266
 267         qatomic_sub(&rcu_call_count, n);
 268         synchronize_rcu();
 269         qemu_mutex_lock_iothread();
 270         while (n > 0) {
 271             node = try_dequeue();
 272             while (!node) {
 273                 qemu_mutex_unlock_iothread();
 274                 qemu_event_reset(&rcu_call_ready_event);
 275                 node = try_dequeue();
 276                 if (!node) {
 277                     qemu_event_wait(&rcu_call_ready_event);
 278                     node = try_dequeue();
 279                 }
 280                 qemu_mutex_lock_iothread();
 281             }
 282
 283             n--;
 284             node->func(node);
 285         }
 286         qemu_mutex_unlock_iothread();
 287     }
 288     abort();
 289 }
 290
 291 void call_rcu1(struct rcu_head *node, void (*func)(struct rcu_head *node))
 292 {
 293     node->func = func;
 294     enqueue(node);
 295     qatomic_inc(&rcu_call_count);
 296     qemu_event_set(&rcu_call_ready_event);
 297 }
 298
 299
 300 struct rcu_drain {
 301     struct rcu_head rcu;
 302     QemuEvent drain_complete_event;
 303 };
 304
 305 static void drain_rcu_callback(struct rcu_head *node)
 306 {
 307     struct rcu_drain *event = (struct rcu_drain *)node;
 308     qemu_event_set(&event->drain_complete_event);
 309 }
 310
 311 /*
 312  * This function ensures that all pending RCU callbacks
 313  * on the current thread are done executing
 314
 315  * drops big qemu lock during the wait to allow RCU thread
 316  * to process the callbacks
 317  *
 318  */
 319
 320 void drain_call_rcu(void)
 321 {
 322     struct rcu_drain rcu_drain;
 323     bool locked = qemu_mutex_iothread_locked();
 324
 325     memset(&rcu_drain, 0, sizeof(struct rcu_drain));
 326     qemu_event_init(&rcu_drain.drain_complete_event, false);
 327
 328     if (locked) {
 329         qemu_mutex_unlock_iothread();
 330     }
 331
 332
 333     /*
 334      * RCU callbacks are invoked in the same order as in which they
 335      * are registered, thus we can be sure that when 'drain_rcu_callback'
 336      * is called, all RCU callbacks that were registered on this thread
 337      * prior to calling this function are completed.
 338      *
 339      * Note that since we have only one global queue of the RCU callbacks,
 340      * we also end up waiting for most of RCU callbacks that were registered
 341      * on the other threads, but this is a side effect that shoudn't be
 342      * assumed.
 343      */
 344
 345     qatomic_inc(&in_drain_call_rcu);
 346     call_rcu1(&rcu_drain.rcu, drain_rcu_callback);
 347     qemu_event_wait(&rcu_drain.drain_complete_event);
 348     qatomic_dec(&in_drain_call_rcu);
 349
 350     if (locked) {
 351         qemu_mutex_lock_iothread();
 352     }
 353
 354 }
 355
 356 void rcu_register_thread(void)
 357 {
 358     assert(get_ptr_rcu_reader()->ctr == 0);
 359     qemu_mutex_lock(&rcu_registry_lock);
 360     QLIST_INSERT_HEAD(&registry, get_ptr_rcu_reader(), node);
 361     qemu_mutex_unlock(&rcu_registry_lock);
 362 }
 363
 364 void rcu_unregister_thread(void)
 365 {
 366     qemu_mutex_lock(&rcu_registry_lock);
 367     QLIST_REMOVE(get_ptr_rcu_reader(), node);
 368     qemu_mutex_unlock(&rcu_registry_lock);
 369 }
 370
 371 void rcu_add_force_rcu_notifier(Notifier *n)
 372 {
 373     qemu_mutex_lock(&rcu_registry_lock);
 374     notifier_list_add(&get_ptr_rcu_reader()->force_rcu, n);
 375     qemu_mutex_unlock(&rcu_registry_lock);
 376 }
 377
 378 void rcu_remove_force_rcu_notifier(Notifier *n)
 379 {
 380     qemu_mutex_lock(&rcu_registry_lock);
 381     notifier_remove(n);
 382     qemu_mutex_unlock(&rcu_registry_lock);
 383 }
 384
 385 static void rcu_init_complete(void)
 386 {
 387     QemuThread thread;
 388
 389     qemu_mutex_init(&rcu_registry_lock);
 390     qemu_mutex_init(&rcu_sync_lock);
 391     qemu_event_init(&rcu_gp_event, true);
 392
 393     qemu_event_init(&rcu_call_ready_event, false);
 394
 395     /* The caller is assumed to have iothread lock, so the call_rcu thread
 396      * must have been quiescent even after forking, just recreate it.
 397      */
 398     qemu_thread_create(&thread, "call_rcu", call_rcu_thread,
 399                        NULL, QEMU_THREAD_DETACHED);
 400
 401     rcu_register_thread();
 402 }
 403
 404 static int atfork_depth = 1;
 405
 406 void rcu_enable_atfork(void)
 407 {
 408     atfork_depth++;
 409 }
 410
 411 void rcu_disable_atfork(void)
 412 {
 413     atfork_depth--;
 414 }
 415
 416 #ifdef CONFIG_POSIX
 417 static void rcu_init_lock(void)
 418 {
 419     if (atfork_depth < 1) {
 420         return;
 421     }
 422
 423     qemu_mutex_lock(&rcu_sync_lock);
 424     qemu_mutex_lock(&rcu_registry_lock);
 425 }
 426
 427 static void rcu_init_unlock(void)
 428 {
 429     if (atfork_depth < 1) {
 430         return;
 431     }
 432
 433     qemu_mutex_unlock(&rcu_registry_lock);
 434     qemu_mutex_unlock(&rcu_sync_lock);
 435 }
 436
 437 static void rcu_init_child(void)
 438 {
 439     if (atfork_depth < 1) {
 440         return;
 441     }
 442
 443     memset(&registry, 0, sizeof(registry));
 444     rcu_init_complete();
 445 }
 446 #endif
 447
 448 static void __attribute__((__constructor__)) rcu_init(void)
 449 {
 450     smp_mb_global_init();
 451 #ifdef CONFIG_POSIX
 452     pthread_atfork(rcu_init_lock, rcu_init_unlock, rcu_init_child);
 453 #endif
 454     rcu_init_complete();
 455 }