summaryrefslogtreecommitdiff
path: root/util/rcu.c
blob: 9adc5e4a368ab600d1c53f7076840b51093d9624 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
/*
 * urcu-mb.c
 *
 * Userspace RCU library with explicit memory barriers
 *
 * Copyright (c) 2009 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 * Copyright (c) 2009 Paul E. McKenney, IBM Corporation.
 * Copyright 2015 Red Hat, Inc.
 *
 * Ported to QEMU by Paolo Bonzini  <pbonzini@redhat.com>
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 *
 * IBM's contributions to this file may be relicensed under LGPLv2 or later.
 */

#include "qemu/osdep.h"
#include "qemu-common.h"
#include "qemu/rcu.h"
#include "qemu/atomic.h"
#include "qemu/thread.h"
#include "qemu/main-loop.h"

/*
 * Global grace period counter.  Bit 0 is always one in rcu_gp_ctr.
 * Bits 1 and above are defined in synchronize_rcu.
 */
#define RCU_GP_LOCKED           (1UL << 0)
#define RCU_GP_CTR              (1UL << 1)

unsigned long rcu_gp_ctr = RCU_GP_LOCKED;

QemuEvent rcu_gp_event;
static QemuMutex rcu_registry_lock;
static QemuMutex rcu_sync_lock;

/*
 * Check whether a quiescent state was crossed between the beginning of
 * update_counter_and_wait and now.
 */
static inline int rcu_gp_ongoing(unsigned long *ctr)
{
    unsigned long v;

    v = atomic_read(ctr);
    return v && (v != rcu_gp_ctr);
}

/* Written to only by each individual reader. Read by both the reader and the
 * writers.
 */
__thread struct rcu_reader_data rcu_reader;

/* Protected by rcu_registry_lock.  */
typedef QLIST_HEAD(, rcu_reader_data) ThreadList;
static ThreadList registry = QLIST_HEAD_INITIALIZER(registry);

/* Wait for previous parity/grace period to be empty of readers.  */
static void wait_for_readers(void)
{
    ThreadList qsreaders = QLIST_HEAD_INITIALIZER(qsreaders);
    struct rcu_reader_data *index, *tmp;

    for (;;) {
        /* We want to be notified of changes made to rcu_gp_ongoing
         * while we walk the list.
         */
        qemu_event_reset(&rcu_gp_event);

        /* Instead of using atomic_mb_set for index->waiting, and
         * atomic_mb_read for index->ctr, memory barriers are placed
         * manually since writes to different threads are independent.
         * qemu_event_reset has acquire semantics, so no memory barrier
         * is needed here.
         */
        QLIST_FOREACH(index, &registry, node) {
            atomic_set(&index->waiting, true);
        }

        /* Here, order the stores to index->waiting before the
         * loads of index->ctr.
         */
        smp_mb();

        QLIST_FOREACH_SAFE(index, &registry, node, tmp) {
            if (!rcu_gp_ongoing(&index->ctr)) {
                QLIST_REMOVE(index, node);
                QLIST_INSERT_HEAD(&qsreaders, index, node);

                /* No need for mb_set here, worst of all we
                 * get some extra futex wakeups.
                 */
                atomic_set(&index->waiting, false);
            }
        }

        if (QLIST_EMPTY(&registry)) {
            break;
        }

        /* Wait for one thread to report a quiescent state and try again.
         * Release rcu_registry_lock, so rcu_(un)register_thread() doesn't
         * wait too much time.
         *
         * rcu_register_thread() may add nodes to &registry; it will not
         * wake up synchronize_rcu, but that is okay because at least another
         * thread must exit its RCU read-side critical section before
         * synchronize_rcu is done.  The next iteration of the loop will
         * move the new thread's rcu_reader from &registry to &qsreaders,
         * because rcu_gp_ongoing() will return false.
         *
         * rcu_unregister_thread() may remove nodes from &qsreaders instead
         * of &registry if it runs during qemu_event_wait.  That's okay;
         * the node then will not be added back to &registry by QLIST_SWAP
         * below.  The invariant is that the node is part of one list when
         * rcu_registry_lock is released.
         */
        qemu_mutex_unlock(&rcu_registry_lock);
        qemu_event_wait(&rcu_gp_event);
        qemu_mutex_lock(&rcu_registry_lock);
    }

    /* put back the reader list in the registry */
    QLIST_SWAP(&registry, &qsreaders, node);
}

void synchronize_rcu(void)
{
    qemu_mutex_lock(&rcu_sync_lock);
    qemu_mutex_lock(&rcu_registry_lock);

    if (!QLIST_EMPTY(&registry)) {
        /* In either case, the atomic_mb_set below blocks stores that free
         * old RCU-protected pointers.
         */
        if (sizeof(rcu_gp_ctr) < 8) {
            /* For architectures with 32-bit longs, a two-subphases algorithm
             * ensures we do not encounter overflow bugs.
             *
             * Switch parity: 0 -> 1, 1 -> 0.
             */
            atomic_mb_set(&rcu_gp_ctr, rcu_gp_ctr ^ RCU_GP_CTR);
            wait_for_readers();
            atomic_mb_set(&rcu_gp_ctr, rcu_gp_ctr ^ RCU_GP_CTR);
        } else {
            /* Increment current grace period.  */
            atomic_mb_set(&rcu_gp_ctr, rcu_gp_ctr + RCU_GP_CTR);
        }

        wait_for_readers();
    }

    qemu_mutex_unlock(&rcu_registry_lock);
    qemu_mutex_unlock(&rcu_sync_lock);
}


#define RCU_CALL_MIN_SIZE        30

/* Multi-producer, single-consumer queue based on urcu/static/wfqueue.h
 * from liburcu.  Note that head is only used by the consumer.
 */
static struct rcu_head dummy;
static struct rcu_head *head = &dummy, **tail = &dummy.next;
static int rcu_call_count;
static QemuEvent rcu_call_ready_event;

static void enqueue(struct rcu_head *node)
{
    struct rcu_head **old_tail;

    node->next = NULL;
    old_tail = atomic_xchg(&tail, &node->next);
    atomic_mb_set(old_tail, node);
}

static struct rcu_head *try_dequeue(void)
{
    struct rcu_head *node, *next;

retry:
    /* Test for an empty list, which we do not expect.  Note that for
     * the consumer head and tail are always consistent.  The head
     * is consistent because only the consumer reads/writes it.
     * The tail, because it is the first step in the enqueuing.
     * It is only the next pointers that might be inconsistent.
     */
    if (head == &dummy && atomic_mb_read(&tail) == &dummy.next) {
        abort();
    }

    /* If the head node has NULL in its next pointer, the value is
     * wrong and we need to wait until its enqueuer finishes the update.
     */
    node = head;
    next = atomic_mb_read(&head->next);
    if (!next) {
        return NULL;
    }

    /* Since we are the sole consumer, and we excluded the empty case
     * above, the queue will always have at least two nodes: the
     * dummy node, and the one being removed.  So we do not need to update
     * the tail pointer.
     */
    head = next;

    /* If we dequeued the dummy node, add it back at the end and retry.  */
    if (node == &dummy) {
        enqueue(node);
        goto retry;
    }

    return node;
}

static void *call_rcu_thread(void *opaque)
{
    struct rcu_head *node;

    rcu_register_thread();

    for (;;) {
        int tries = 0;
        int n = atomic_read(&rcu_call_count);

        /* Heuristically wait for a decent number of callbacks to pile up.
         * Fetch rcu_call_count now, we only must process elements that were
         * added before synchronize_rcu() starts.
         */
        while (n == 0 || (n < RCU_CALL_MIN_SIZE && ++tries <= 5)) {
            g_usleep(10000);
            if (n == 0) {
                qemu_event_reset(&rcu_call_ready_event);
                n = atomic_read(&rcu_call_count);
                if (n == 0) {
                    qemu_event_wait(&rcu_call_ready_event);
                }
            }
            n = atomic_read(&rcu_call_count);
        }

        atomic_sub(&rcu_call_count, n);
        synchronize_rcu();
        qemu_mutex_lock_iothread();
        while (n > 0) {
            node = try_dequeue();
            while (!node) {
                qemu_mutex_unlock_iothread();
                qemu_event_reset(&rcu_call_ready_event);
                node = try_dequeue();
                if (!node) {
                    qemu_event_wait(&rcu_call_ready_event);
                    node = try_dequeue();
                }
                qemu_mutex_lock_iothread();
            }

            n--;
            node->func(node);
        }
        qemu_mutex_unlock_iothread();
    }
    abort();
}

void call_rcu1(struct rcu_head *node, void (*func)(struct rcu_head *node))
{
    node->func = func;
    enqueue(node);
    atomic_inc(&rcu_call_count);
    qemu_event_set(&rcu_call_ready_event);
}

void rcu_register_thread(void)
{
    assert(rcu_reader.ctr == 0);
    qemu_mutex_lock(&rcu_registry_lock);
    QLIST_INSERT_HEAD(&registry, &rcu_reader, node);
    qemu_mutex_unlock(&rcu_registry_lock);
}

void rcu_unregister_thread(void)
{
    qemu_mutex_lock(&rcu_registry_lock);
    QLIST_REMOVE(&rcu_reader, node);
    qemu_mutex_unlock(&rcu_registry_lock);
}

static void rcu_init_complete(void)
{
    QemuThread thread;

    qemu_mutex_init(&rcu_registry_lock);
    qemu_mutex_init(&rcu_sync_lock);
    qemu_event_init(&rcu_gp_event, true);

    qemu_event_init(&rcu_call_ready_event, false);

    /* The caller is assumed to have iothread lock, so the call_rcu thread
     * must have been quiescent even after forking, just recreate it.
     */
    qemu_thread_create(&thread, "call_rcu", call_rcu_thread,
                       NULL, QEMU_THREAD_DETACHED);

    rcu_register_thread();
}

#ifdef CONFIG_POSIX
static void rcu_init_lock(void)
{
    qemu_mutex_lock(&rcu_sync_lock);
    qemu_mutex_lock(&rcu_registry_lock);
}

static void rcu_init_unlock(void)
{
    qemu_mutex_unlock(&rcu_registry_lock);
    qemu_mutex_unlock(&rcu_sync_lock);
}
#endif

void rcu_after_fork(void)
{
    memset(&registry, 0, sizeof(registry));
    rcu_init_complete();
}

static void __attribute__((__constructor__)) rcu_init(void)
{
#ifdef CONFIG_POSIX
    pthread_atfork(rcu_init_lock, rcu_init_unlock, rcu_init_unlock);
#endif
    rcu_init_complete();
}