summaryrefslogtreecommitdiffstats
path: root/block/linux-aio.c
blob: 3c0527c2bf8970960bd66077d6ba81da6a7cc85c (plain) (blame)
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
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
/*
 * Linux native AIO support.
 *
 * Copyright (C) 2009 IBM, Corp.
 * Copyright (C) 2009 Red Hat, Inc.
 *
 * This work is licensed under the terms of the GNU GPL, version 2 or later.
 * See the COPYING file in the top-level directory.
 */
#include "qemu/osdep.h"
#include "block/aio.h"
#include "qemu/queue.h"
#include "block/block.h"
#include "block/raw-aio.h"
#include "qemu/event_notifier.h"
#include "qemu/coroutine.h"
#include "qapi/error.h"

#include <libaio.h>

/*
 * Queue size (per-device).
 *
 * XXX: eventually we need to communicate this to the guest and/or make it
 *      tunable by the guest.  If we get more outstanding requests at a time
 *      than this we will get EAGAIN from io_submit which is communicated to
 *      the guest as an I/O error.
 */
#define MAX_EVENTS 1024

struct qemu_laiocb {
    Coroutine *co;
    LinuxAioState *ctx;
    struct iocb iocb;
    ssize_t ret;
    size_t nbytes;
    QEMUIOVector *qiov;
    bool is_read;
    QSIMPLEQ_ENTRY(qemu_laiocb) next;
};

typedef struct {
    int plugged;
    unsigned int in_queue;
    unsigned int in_flight;
    bool blocked;
    QSIMPLEQ_HEAD(, qemu_laiocb) pending;
} LaioQueue;

struct LinuxAioState {
    AioContext *aio_context;

    io_context_t ctx;
    EventNotifier e;

    /* io queue for submit at batch.  Protected by AioContext lock. */
    LaioQueue io_q;

    /* I/O completion processing.  Only runs in I/O thread.  */
    QEMUBH *completion_bh;
    int event_idx;
    int event_max;
};

static void ioq_submit(LinuxAioState *s);

static inline ssize_t io_event_ret(struct io_event *ev)
{
    return (ssize_t)(((uint64_t)ev->res2 << 32) | ev->res);
}

/*
 * Completes an AIO request.
 */
static void qemu_laio_process_completion(struct qemu_laiocb *laiocb)
{
    int ret;

    ret = laiocb->ret;
    if (ret != -ECANCELED) {
        if (ret == laiocb->nbytes) {
            ret = 0;
        } else if (ret >= 0) {
            /* Short reads mean EOF, pad with zeros. */
            if (laiocb->is_read) {
                qemu_iovec_memset(laiocb->qiov, ret, 0,
                    laiocb->qiov->size - ret);
            } else {
                ret = -ENOSPC;
            }
        }
    }

    laiocb->ret = ret;

    /*
     * If the coroutine is already entered it must be in ioq_submit() and
     * will notice laio->ret has been filled in when it eventually runs
     * later.  Coroutines cannot be entered recursively so avoid doing
     * that!
     */
    if (!qemu_coroutine_entered(laiocb->co)) {
        aio_co_wake(laiocb->co);
    }
}

/**
 * aio_ring buffer which is shared between userspace and kernel.
 *
 * This copied from linux/fs/aio.c, common header does not exist
 * but AIO exists for ages so we assume ABI is stable.
 */
struct aio_ring {
    unsigned    id;    /* kernel internal index number */
    unsigned    nr;    /* number of io_events */
    unsigned    head;  /* Written to by userland or by kernel. */
    unsigned    tail;

    unsigned    magic;
    unsigned    compat_features;
    unsigned    incompat_features;
    unsigned    header_length;  /* size of aio_ring */

    struct io_event io_events[];
};

/**
 * io_getevents_peek:
 * @ctx: AIO context
 * @events: pointer on events array, output value

 * Returns the number of completed events and sets a pointer
 * on events array.  This function does not update the internal
 * ring buffer, only reads head and tail.  When @events has been
 * processed io_getevents_commit() must be called.
 */
static inline unsigned int io_getevents_peek(io_context_t ctx,
                                             struct io_event **events)
{
    struct aio_ring *ring = (struct aio_ring *)ctx;
    unsigned int head = ring->head, tail = ring->tail;
    unsigned int nr;

    nr = tail >= head ? tail - head : ring->nr - head;
    *events = ring->io_events + head;
    /* To avoid speculative loads of s->events[i] before observing tail.
       Paired with smp_wmb() inside linux/fs/aio.c: aio_complete(). */
    smp_rmb();

    return nr;
}

/**
 * io_getevents_commit:
 * @ctx: AIO context
 * @nr: the number of events on which head should be advanced
 *
 * Advances head of a ring buffer.
 */
static inline void io_getevents_commit(io_context_t ctx, unsigned int nr)
{
    struct aio_ring *ring = (struct aio_ring *)ctx;

    if (nr) {
        ring->head = (ring->head + nr) % ring->nr;
    }
}

/**
 * io_getevents_advance_and_peek:
 * @ctx: AIO context
 * @events: pointer on events array, output value
 * @nr: the number of events on which head should be advanced
 *
 * Advances head of a ring buffer and returns number of elements left.
 */
static inline unsigned int
io_getevents_advance_and_peek(io_context_t ctx,
                              struct io_event **events,
                              unsigned int nr)
{
    io_getevents_commit(ctx, nr);
    return io_getevents_peek(ctx, events);
}

/**
 * qemu_laio_process_completions:
 * @s: AIO state
 *
 * Fetches completed I/O requests and invokes their callbacks.
 *
 * The function is somewhat tricky because it supports nested event loops, for
 * example when a request callback invokes aio_poll().  In order to do this,
 * indices are kept in LinuxAioState.  Function schedules BH completion so it
 * can be called again in a nested event loop.  When there are no events left
 * to complete the BH is being canceled.
 */
static void qemu_laio_process_completions(LinuxAioState *s)
{
    struct io_event *events;

    /* Reschedule so nested event loops see currently pending completions */
    qemu_bh_schedule(s->completion_bh);

    while ((s->event_max = io_getevents_advance_and_peek(s->ctx, &events,
                                                         s->event_idx))) {
        for (s->event_idx = 0; s->event_idx < s->event_max; ) {
            struct iocb *iocb = events[s->event_idx].obj;
            struct qemu_laiocb *laiocb =
                container_of(iocb, struct qemu_laiocb, iocb);

            laiocb->ret = io_event_ret(&events[s->event_idx]);

            /* Change counters one-by-one because we can be nested. */
            s->io_q.in_flight--;
            s->event_idx++;
            qemu_laio_process_completion(laiocb);
        }
    }

    qemu_bh_cancel(s->completion_bh);

    /* If we are nested we have to notify the level above that we are done
     * by setting event_max to zero, upper level will then jump out of it's
     * own `for` loop.  If we are the last all counters droped to zero. */
    s->event_max = 0;
    s->event_idx = 0;
}

static void qemu_laio_process_completions_and_submit(LinuxAioState *s)
{
    aio_context_acquire(s->aio_context);
    qemu_laio_process_completions(s);

    if (!s->io_q.plugged && !QSIMPLEQ_EMPTY(&s->io_q.pending)) {
        ioq_submit(s);
    }
    aio_context_release(s->aio_context);
}

static void qemu_laio_completion_bh(void *opaque)
{
    LinuxAioState *s = opaque;

    qemu_laio_process_completions_and_submit(s);
}

static void qemu_laio_completion_cb(EventNotifier *e)
{
    LinuxAioState *s = container_of(e, LinuxAioState, e);

    if (event_notifier_test_and_clear(&s->e)) {
        qemu_laio_process_completions_and_submit(s);
    }
}

static bool qemu_laio_poll_cb(void *opaque)
{
    EventNotifier *e = opaque;
    LinuxAioState *s = container_of(e, LinuxAioState, e);
    struct io_event *events;

    if (!io_getevents_peek(s->ctx, &events)) {
        return false;
    }

    qemu_laio_process_completions_and_submit(s);
    return true;
}

static void ioq_init(LaioQueue *io_q)
{
    QSIMPLEQ_INIT(&io_q->pending);
    io_q->plugged = 0;
    io_q->in_queue = 0;
    io_q->in_flight = 0;
    io_q->blocked = false;
}

static void ioq_submit(LinuxAioState *s)
{
    int ret, len;
    struct qemu_laiocb *aiocb;
    struct iocb *iocbs[MAX_EVENTS];
    QSIMPLEQ_HEAD(, qemu_laiocb) completed;

    do {
        if (s->io_q.in_flight >= MAX_EVENTS) {
            break;
        }
        len = 0;
        QSIMPLEQ_FOREACH(aiocb, &s->io_q.pending, next) {
            iocbs[len++] = &aiocb->iocb;
            if (s->io_q.in_flight + len >= MAX_EVENTS) {
                break;
            }
        }

        ret = io_submit(s->ctx, len, iocbs);
        if (ret == -EAGAIN) {
            break;
        }
        if (ret < 0) {
            /* Fail the first request, retry the rest */
            aiocb = QSIMPLEQ_FIRST(&s->io_q.pending);
            QSIMPLEQ_REMOVE_HEAD(&s->io_q.pending, next);
            s->io_q.in_queue--;
            aiocb->ret = ret;
            qemu_laio_process_completion(aiocb);
            continue;
        }

        s->io_q.in_flight += ret;
        s->io_q.in_queue  -= ret;
        aiocb = container_of(iocbs[ret - 1], struct qemu_laiocb, iocb);
        QSIMPLEQ_SPLIT_AFTER(&s->io_q.pending, aiocb, next, &completed);
    } while (ret == len && !QSIMPLEQ_EMPTY(&s->io_q.pending));
    s->io_q.blocked = (s->io_q.in_queue > 0);

    if (s->io_q.in_flight) {
        /* We can try to complete something just right away if there are
         * still requests in-flight. */
        qemu_laio_process_completions(s);
        /*
         * Even we have completed everything (in_flight == 0), the queue can
         * have still pended requests (in_queue > 0).  We do not attempt to
         * repeat submission to avoid IO hang.  The reason is simple: s->e is
         * still set and completion callback will be called shortly and all
         * pended requests will be submitted from there.
         */
    }
}

void laio_io_plug(BlockDriverState *bs, LinuxAioState *s)
{
    s->io_q.plugged++;
}

void laio_io_unplug(BlockDriverState *bs, LinuxAioState *s)
{
    assert(s->io_q.plugged);
    if (--s->io_q.plugged == 0 &&
        !s->io_q.blocked && !QSIMPLEQ_EMPTY(&s->io_q.pending)) {
        ioq_submit(s);
    }
}

static int laio_do_submit(int fd, struct qemu_laiocb *laiocb, off_t offset,
                          int type)
{
    LinuxAioState *s = laiocb->ctx;
    struct iocb *iocbs = &laiocb->iocb;
    QEMUIOVector *qiov = laiocb->qiov;

    switch (type) {
    case QEMU_AIO_WRITE:
        io_prep_pwritev(iocbs, fd, qiov->iov, qiov->niov, offset);
        break;
    case QEMU_AIO_READ:
        io_prep_preadv(iocbs, fd, qiov->iov, qiov->niov, offset);
        break;
    /* Currently Linux kernel does not support other operations */
    default:
        fprintf(stderr, "%s: invalid AIO request type 0x%x.\n",
                        __func__, type);
        return -EIO;
    }
    io_set_eventfd(&laiocb->iocb, event_notifier_get_fd(&s->e));

    QSIMPLEQ_INSERT_TAIL(&s->io_q.pending, laiocb, next);
    s->io_q.in_queue++;
    if (!s->io_q.blocked &&
        (!s->io_q.plugged ||
         s->io_q.in_flight + s->io_q.in_queue >= MAX_EVENTS)) {
        ioq_submit(s);
    }

    return 0;
}

int coroutine_fn laio_co_submit(BlockDriverState *bs, LinuxAioState *s, int fd,
                                uint64_t offset, QEMUIOVector *qiov, int type)
{
    int ret;
    struct qemu_laiocb laiocb = {
        .co         = qemu_coroutine_self(),
        .nbytes     = qiov->size,
        .ctx        = s,
        .ret        = -EINPROGRESS,
        .is_read    = (type == QEMU_AIO_READ),
        .qiov       = qiov,
    };

    ret = laio_do_submit(fd, &laiocb, offset, type);
    if (ret < 0) {
        return ret;
    }

    if (laiocb.ret == -EINPROGRESS) {
        qemu_coroutine_yield();
    }
    return laiocb.ret;
}

void laio_detach_aio_context(LinuxAioState *s, AioContext *old_context)
{
    aio_set_event_notifier(old_context, &s->e, false, NULL, NULL);
    qemu_bh_delete(s->completion_bh);
    s->aio_context = NULL;
}

void laio_attach_aio_context(LinuxAioState *s, AioContext *new_context)
{
    s->aio_context = new_context;
    s->completion_bh = aio_bh_new(new_context, qemu_laio_completion_bh, s);
    aio_set_event_notifier(new_context, &s->e, false,
                           qemu_laio_completion_cb,
                           qemu_laio_poll_cb);
}

LinuxAioState *laio_init(Error **errp)
{
    int rc;
    LinuxAioState *s;

    s = g_malloc0(sizeof(*s));
    rc = event_notifier_init(&s->e, false);
    if (rc < 0) {
        error_setg_errno(errp, -rc, "failed to to initialize event notifier");
        goto out_free_state;
    }

    rc = io_setup(MAX_EVENTS, &s->ctx);
    if (rc < 0) {
        error_setg_errno(errp, -rc, "failed to create linux AIO context");
        goto out_close_efd;
    }

    ioq_init(&s->io_q);

    return s;

out_close_efd:
    event_notifier_cleanup(&s->e);
out_free_state:
    g_free(s);
    return NULL;
}

void laio_cleanup(LinuxAioState *s)
{
    event_notifier_cleanup(&s->e);

    if (io_destroy(s->ctx) != 0) {
        fprintf(stderr, "%s: destroy AIO context %p failed\n",
                        __func__, &s->ctx);
    }
    g_free(s);
}