aio-posix.c 6.81 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
/*
 * QEMU aio implementation
 *
 * Copyright IBM, Corp. 2008
 *
 * Authors:
 *  Anthony Liguori   <aliguori@us.ibm.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
12 13
 * Contributions after 2012-01-13 are licensed under the terms of the
 * GNU GPL, version 2 or (at your option) any later version.
14 15 16
 */

#include "qemu-common.h"
17
#include "block/block.h"
18 19
#include "qemu/queue.h"
#include "qemu/sockets.h"
20 21 22

struct AioHandler
{
23
    GPollFD pfd;
24 25 26
    IOHandler *io_read;
    IOHandler *io_write;
    int deleted;
27
    int pollfds_idx;
28
    void *opaque;
29
    QLIST_ENTRY(AioHandler) node;
30 31
};

32
static AioHandler *find_aio_handler(AioContext *ctx, int fd)
33 34 35
{
    AioHandler *node;

36
    QLIST_FOREACH(node, &ctx->aio_handlers, node) {
37
        if (node->pfd.fd == fd)
Alexander Graf's avatar
Alexander Graf committed
38 39
            if (!node->deleted)
                return node;
40 41 42 43 44
    }

    return NULL;
}

45 46 47 48 49
void aio_set_fd_handler(AioContext *ctx,
                        int fd,
                        IOHandler *io_read,
                        IOHandler *io_write,
                        void *opaque)
50 51 52
{
    AioHandler *node;

53
    node = find_aio_handler(ctx, fd);
54 55 56 57

    /* Are we deleting the fd handler? */
    if (!io_read && !io_write) {
        if (node) {
58 59
            g_source_remove_poll(&ctx->source, &node->pfd);

60
            /* If the lock is held, just mark the node as deleted */
61
            if (ctx->walking_handlers) {
62
                node->deleted = 1;
63 64
                node->pfd.revents = 0;
            } else {
65 66 67 68
                /* Otherwise, delete it for real.  We can't just mark it as
                 * deleted because deleted nodes are only cleaned up after
                 * releasing the walking_handlers lock.
                 */
69
                QLIST_REMOVE(node, node);
70
                g_free(node);
71 72 73 74 75
            }
        }
    } else {
        if (node == NULL) {
            /* Alloc and insert if it's not already there */
76
            node = g_malloc0(sizeof(AioHandler));
77
            node->pfd.fd = fd;
78
            QLIST_INSERT_HEAD(&ctx->aio_handlers, node, node);
79 80

            g_source_add_poll(&ctx->source, &node->pfd);
81 82 83 84 85
        }
        /* Update handler with latest information */
        node->io_read = io_read;
        node->io_write = io_write;
        node->opaque = opaque;
86
        node->pollfds_idx = -1;
87

88 89
        node->pfd.events = (io_read ? G_IO_IN | G_IO_HUP | G_IO_ERR : 0);
        node->pfd.events |= (io_write ? G_IO_OUT | G_IO_ERR : 0);
90
    }
91 92

    aio_notify(ctx);
93 94
}

95 96
void aio_set_event_notifier(AioContext *ctx,
                            EventNotifier *notifier,
97
                            EventNotifierHandler *io_read)
98
{
99
    aio_set_fd_handler(ctx, event_notifier_get_fd(notifier),
100
                       (IOHandler *)io_read, NULL, notifier);
101 102
}

103 104 105 106 107
bool aio_prepare(AioContext *ctx)
{
    return false;
}

108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126
bool aio_pending(AioContext *ctx)
{
    AioHandler *node;

    QLIST_FOREACH(node, &ctx->aio_handlers, node) {
        int revents;

        revents = node->pfd.revents & node->pfd.events;
        if (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR) && node->io_read) {
            return true;
        }
        if (revents & (G_IO_OUT | G_IO_ERR) && node->io_write) {
            return true;
        }
    }

    return false;
}

127
bool aio_dispatch(AioContext *ctx)
128
{
129
    AioHandler *node;
130
    bool progress = false;
131

132 133 134 135 136 137 138 139 140
    /*
     * If there are callbacks left that have been queued, we need to call them.
     * Do not call select in this case, because it is possible that the caller
     * does not need a complete flush (as is the case for aio_poll loops).
     */
    if (aio_bh_poll(ctx)) {
        progress = true;
    }

141
    /*
142
     * We have to walk very carefully in case aio_set_fd_handler is
143 144 145 146 147 148 149 150 151 152 153 154
     * called while we're walking.
     */
    node = QLIST_FIRST(&ctx->aio_handlers);
    while (node) {
        AioHandler *tmp;
        int revents;

        ctx->walking_handlers++;

        revents = node->pfd.revents & node->pfd.events;
        node->pfd.revents = 0;

155 156 157
        if (!node->deleted &&
            (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR)) &&
            node->io_read) {
158
            node->io_read(node->opaque);
159 160 161 162 163

            /* aio_notify() does not count as progress */
            if (node->opaque != &ctx->notifier) {
                progress = true;
            }
164
        }
165 166 167
        if (!node->deleted &&
            (revents & (G_IO_OUT | G_IO_ERR)) &&
            node->io_write) {
168 169 170 171 172 173 174 175 176 177 178 179 180 181
            node->io_write(node->opaque);
            progress = true;
        }

        tmp = node;
        node = QLIST_NEXT(node, node);

        ctx->walking_handlers--;

        if (!ctx->walking_handlers && tmp->deleted) {
            QLIST_REMOVE(tmp, node);
            g_free(tmp);
        }
    }
182 183 184 185

    /* Run our timers */
    progress |= timerlistgroup_run_timers(&ctx->tlg);

186 187 188 189 190 191
    return progress;
}

bool aio_poll(AioContext *ctx, bool blocking)
{
    AioHandler *node;
192
    bool was_dispatching;
193
    int ret;
194
    bool progress;
195

196
    was_dispatching = ctx->dispatching;
197 198
    progress = false;

199 200
    /* aio_notify can avoid the expensive event_notifier_set if
     * everything (file descriptors, bottom halves, timers) will
201 202 203
     * be re-evaluated before the next blocking poll().  This is
     * already true when aio_poll is called with blocking == false;
     * if blocking == true, it is only true after poll() returns.
204 205 206 207 208 209 210
     *
     * If we're in a nested event loop, ctx->dispatching might be true.
     * In that case we can restore it just before returning, but we
     * have to clear it now.
     */
    aio_set_dispatching(ctx, !blocking);

211
    ctx->walking_handlers++;
212

213
    g_array_set_size(ctx->pollfds, 0);
214

215
    /* fill pollfds */
216
    QLIST_FOREACH(node, &ctx->aio_handlers, node) {
217 218 219 220 221 222 223 224
        node->pollfds_idx = -1;
        if (!node->deleted && node->pfd.events) {
            GPollFD pfd = {
                .fd = node->pfd.fd,
                .events = node->pfd.events,
            };
            node->pollfds_idx = ctx->pollfds->len;
            g_array_append_val(ctx->pollfds, pfd);
225 226
        }
    }
227

228
    ctx->walking_handlers--;
229

230
    /* wait until next event */
231 232
    ret = qemu_poll_ns((GPollFD *)ctx->pollfds->data,
                         ctx->pollfds->len,
233
                         blocking ? aio_compute_timeout(ctx) : 0);
234 235 236

    /* if we have any readable fds, dispatch event */
    if (ret > 0) {
237 238 239 240 241
        QLIST_FOREACH(node, &ctx->aio_handlers, node) {
            if (node->pollfds_idx != -1) {
                GPollFD *pfd = &g_array_index(ctx->pollfds, GPollFD,
                                              node->pollfds_idx);
                node->pfd.revents = pfd->revents;
242
            }
243
        }
244 245 246
    }

    /* Run dispatch even if there were no readable fds to run timers */
247
    aio_set_dispatching(ctx, true);
248 249
    if (aio_dispatch(ctx)) {
        progress = true;
250
    }
251

252
    aio_set_dispatching(ctx, was_dispatching);
253
    return progress;
254
}