[qemu.git] / posix-aio-compat.c

/*
 * QEMU posix-aio emulation
 *
 * 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.
 *
 */

#include <sys/ioctl.h>
#include <pthread.h>
#include <unistd.h>
#include <errno.h>
#include <time.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>
#include "osdep.h"
#include "qemu-common.h"

#include "posix-aio-compat.h"

static pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
static pthread_cond_t cond = PTHREAD_COND_INITIALIZER;
static pthread_t thread_id;
static pthread_attr_t attr;
static int max_threads = 64;
static int cur_threads = 0;
static int idle_threads = 0;
static TAILQ_HEAD(, qemu_paiocb) request_list;

#ifdef CONFIG_PREADV
static int preadv_present = 1;
#else
static int preadv_present = 0;
#endif

static void die2(int err, const char *what)
{
    fprintf(stderr, "%s failed: %s\n", what, strerror(err));
    abort();
}

static void die(const char *what)
{
    die2(errno, what);
}

static void mutex_lock(pthread_mutex_t *mutex)
{
    int ret = pthread_mutex_lock(mutex);
    if (ret) die2(ret, "pthread_mutex_lock");
}

static void mutex_unlock(pthread_mutex_t *mutex)
{
    int ret = pthread_mutex_unlock(mutex);
    if (ret) die2(ret, "pthread_mutex_unlock");
}

static int cond_timedwait(pthread_cond_t *cond, pthread_mutex_t *mutex,
                           struct timespec *ts)
{
    int ret = pthread_cond_timedwait(cond, mutex, ts);
    if (ret && ret != ETIMEDOUT) die2(ret, "pthread_cond_timedwait");
    return ret;
}

static void cond_signal(pthread_cond_t *cond)
{
    int ret = pthread_cond_signal(cond);
    if (ret) die2(ret, "pthread_cond_signal");
}

static void thread_create(pthread_t *thread, pthread_attr_t *attr,
                          void *(*start_routine)(void*), void *arg)
{
    int ret = pthread_create(thread, attr, start_routine, arg);
    if (ret) die2(ret, "pthread_create");
}

static size_t handle_aiocb_ioctl(struct qemu_paiocb *aiocb)
{
	int ret;

	ret = ioctl(aiocb->aio_fildes, aiocb->aio_ioctl_cmd, aiocb->aio_ioctl_buf);
	if (ret == -1)
		return -errno;

	/*
	 * This looks weird, but the aio code only consideres a request
	 * successfull if it has written the number full number of bytes.
	 *
	 * Now we overload aio_nbytes as aio_ioctl_cmd for the ioctl command,
	 * so in fact we return the ioctl command here to make posix_aio_read()
	 * happy..
	 */
	return aiocb->aio_nbytes;
}

#ifdef CONFIG_PREADV

static ssize_t
qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset)
{
    return preadv(fd, iov, nr_iov, offset);
}

static ssize_t
qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset)
{
    return pwritev(fd, iov, nr_iov, offset);
}

#else

static ssize_t
qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset)
{
    return -ENOSYS;
}

static ssize_t
qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset)
{
    return -ENOSYS;
}

#endif

/*
 * Check if we need to copy the data in the aiocb into a new
 * properly aligned buffer.
 */
static int aiocb_needs_copy(struct qemu_paiocb *aiocb)
{
    if (aiocb->aio_flags & QEMU_AIO_SECTOR_ALIGNED) {
        int i;

        for (i = 0; i < aiocb->aio_niov; i++)
            if ((uintptr_t) aiocb->aio_iov[i].iov_base % 512)
                return 1;
    }

    return 0;
}

static size_t handle_aiocb_rw_vector(struct qemu_paiocb *aiocb)
{
    size_t offset = 0;
    ssize_t len;

    do {
        if (aiocb->aio_type == QEMU_PAIO_WRITE)
            len = qemu_pwritev(aiocb->aio_fildes,
                               aiocb->aio_iov,
                               aiocb->aio_niov,
                               aiocb->aio_offset + offset);
         else
            len = qemu_preadv(aiocb->aio_fildes,
                              aiocb->aio_iov,
                              aiocb->aio_niov,
                              aiocb->aio_offset + offset);
    } while (len == -1 && errno == EINTR);

    if (len == -1)
        return -errno;
    return len;
}

static size_t handle_aiocb_rw_linear(struct qemu_paiocb *aiocb, char *buf)
{
    size_t offset = 0;
    size_t len;

    while (offset < aiocb->aio_nbytes) {
         if (aiocb->aio_type == QEMU_PAIO_WRITE)
             len = pwrite(aiocb->aio_fildes,
                          (const char *)buf + offset,
                          aiocb->aio_nbytes - offset,
                          aiocb->aio_offset + offset);
         else
             len = pread(aiocb->aio_fildes,
                         buf + offset,
                         aiocb->aio_nbytes - offset,
                         aiocb->aio_offset + offset);

         if (len == -1 && errno == EINTR)
             continue;
         else if (len == -1) {
             offset = -errno;
             break;
         } else if (len == 0)
             break;

         offset += len;
    }

    return offset;
}

static size_t handle_aiocb_rw(struct qemu_paiocb *aiocb)
{
    size_t nbytes;
    char *buf;

    if (!aiocb_needs_copy(aiocb)) {
        /*
         * If there is just a single buffer, and it is properly aligned
         * we can just use plain pread/pwrite without any problems.
         */
        if (aiocb->aio_niov == 1)
             return handle_aiocb_rw_linear(aiocb, aiocb->aio_iov->iov_base);

        /*
         * We have more than one iovec, and all are properly aligned.
         *
         * Try preadv/pwritev first and fall back to linearizing the
         * buffer if it's not supported.
         */
	if (preadv_present) {
            nbytes = handle_aiocb_rw_vector(aiocb);
            if (nbytes == aiocb->aio_nbytes)
	        return nbytes;
            if (nbytes < 0 && nbytes != -ENOSYS)
                return nbytes;
            preadv_present = 0;
        }

        /*
         * XXX(hch): short read/write.  no easy way to handle the reminder
         * using these interfaces.  For now retry using plain
         * pread/pwrite?
         */
    }

    /*
     * Ok, we have to do it the hard way, copy all segments into
     * a single aligned buffer.
     */
    buf = qemu_memalign(512, aiocb->aio_nbytes);
    if (aiocb->aio_type == QEMU_PAIO_WRITE) {
        char *p = buf;
        int i;

        for (i = 0; i < aiocb->aio_niov; ++i) {
            memcpy(p, aiocb->aio_iov[i].iov_base, aiocb->aio_iov[i].iov_len);
            p += aiocb->aio_iov[i].iov_len;
        }
    }

    nbytes = handle_aiocb_rw_linear(aiocb, buf);
    if (aiocb->aio_type != QEMU_PAIO_WRITE) {
        char *p = buf;
        size_t count = aiocb->aio_nbytes, copy;
        int i;

        for (i = 0; i < aiocb->aio_niov && count; ++i) {
            copy = count;
            if (copy > aiocb->aio_iov[i].iov_len)
                copy = aiocb->aio_iov[i].iov_len;
            memcpy(aiocb->aio_iov[i].iov_base, p, copy);
            p     += copy;
            count -= copy;
        }
    }
    qemu_vfree(buf);

    return nbytes;
}

static void *aio_thread(void *unused)
{
    pid_t pid;
    sigset_t set;

    pid = getpid();

    /* block all signals */
    if (sigfillset(&set)) die("sigfillset");
    if (sigprocmask(SIG_BLOCK, &set, NULL)) die("sigprocmask");

    while (1) {
        struct qemu_paiocb *aiocb;
        size_t ret = 0;
        qemu_timeval tv;
        struct timespec ts;

        qemu_gettimeofday(&tv);
        ts.tv_sec = tv.tv_sec + 10;
        ts.tv_nsec = 0;

        mutex_lock(&lock);

        while (TAILQ_EMPTY(&request_list) &&
               !(ret == ETIMEDOUT)) {
            ret = cond_timedwait(&cond, &lock, &ts);
        }

        if (TAILQ_EMPTY(&request_list))
            break;

        aiocb = TAILQ_FIRST(&request_list);
        TAILQ_REMOVE(&request_list, aiocb, node);
        aiocb->active = 1;
        idle_threads--;
        mutex_unlock(&lock);

        switch (aiocb->aio_type) {
        case QEMU_PAIO_READ:
        case QEMU_PAIO_WRITE:
		ret = handle_aiocb_rw(aiocb);
		break;
        case QEMU_PAIO_IOCTL:
		ret = handle_aiocb_ioctl(aiocb);
		break;
	default:
		fprintf(stderr, "invalid aio request (0x%x)\n", aiocb->aio_type);
		ret = -EINVAL;
		break;
	}

        mutex_lock(&lock);
        aiocb->ret = ret;
        idle_threads++;
        mutex_unlock(&lock);

        if (kill(pid, aiocb->ev_signo)) die("kill failed");
    }

    idle_threads--;
    cur_threads--;
    mutex_unlock(&lock);

    return NULL;
}

static void spawn_thread(void)
{
    cur_threads++;
    idle_threads++;
    thread_create(&thread_id, &attr, aio_thread, NULL);
}

int qemu_paio_init(struct qemu_paioinit *aioinit)
{
    int ret;

    ret = pthread_attr_init(&attr);
    if (ret) die2(ret, "pthread_attr_init");

    ret = pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
    if (ret) die2(ret, "pthread_attr_setdetachstate");

    TAILQ_INIT(&request_list);

    return 0;
}

static int qemu_paio_submit(struct qemu_paiocb *aiocb, int type)
{
    aiocb->aio_type = type;
    aiocb->ret = -EINPROGRESS;
    aiocb->active = 0;
    mutex_lock(&lock);
    if (idle_threads == 0 && cur_threads < max_threads)
        spawn_thread();
    TAILQ_INSERT_TAIL(&request_list, aiocb, node);
    mutex_unlock(&lock);
    cond_signal(&cond);

    return 0;
}

int qemu_paio_read(struct qemu_paiocb *aiocb)
{
    return qemu_paio_submit(aiocb, QEMU_PAIO_READ);
}

int qemu_paio_write(struct qemu_paiocb *aiocb)
{
    return qemu_paio_submit(aiocb, QEMU_PAIO_WRITE);
}

int qemu_paio_ioctl(struct qemu_paiocb *aiocb)
{
    return qemu_paio_submit(aiocb, QEMU_PAIO_IOCTL);
}

ssize_t qemu_paio_return(struct qemu_paiocb *aiocb)
{
    ssize_t ret;

    mutex_lock(&lock);
    ret = aiocb->ret;
    mutex_unlock(&lock);

    return ret;
}

int qemu_paio_error(struct qemu_paiocb *aiocb)
{
    ssize_t ret = qemu_paio_return(aiocb);

    if (ret < 0)
        ret = -ret;
    else
        ret = 0;

    return ret;
}

int qemu_paio_cancel(int fd, struct qemu_paiocb *aiocb)
{
    int ret;

    mutex_lock(&lock);
    if (!aiocb->active) {
        TAILQ_REMOVE(&request_list, aiocb, node);
        aiocb->ret = -ECANCELED;
        ret = QEMU_PAIO_CANCELED;
    } else if (aiocb->ret == -EINPROGRESS)
        ret = QEMU_PAIO_NOTCANCELED;
    else
        ret = QEMU_PAIO_ALLDONE;
    mutex_unlock(&lock);

    return ret;
}
Commit	Line	Data
3c529d93 AL	1	/*
	2	* QEMU posix-aio emulation
	3	*
	4	* Copyright IBM, Corp. 2008
	5	*
	6	* Authors:
	7	* Anthony Liguori <aliguori@us.ibm.com>
	8	*
	9	* This work is licensed under the terms of the GNU GPL, version 2. See
	10	* the COPYING file in the top-level directory.
	11	*
	12	*/
	13
221f715d	14	#include <sys/ioctl.h>
3c529d93 AL	15	#include <pthread.h>
	16	#include <unistd.h>
	17	#include <errno.h>
30525aff	18	#include <time.h>
8653c015	19	#include <string.h>
	20	#include <stdlib.h>
	21	#include <stdio.h>
3c529d93	22	#include "osdep.h"
f141eafe	23	#include "qemu-common.h"
3c529d93 AL	24
	25	#include "posix-aio-compat.h"
	26
	27	static pthread_mutex_t lock = PTHREAD_MUTEX_INITIALIZER;
	28	static pthread_cond_t cond = PTHREAD_COND_INITIALIZER;
	29	static pthread_t thread_id;
a8227a5a	30	static pthread_attr_t attr;
3c529d93 AL	31	static int max_threads = 64;
	32	static int cur_threads = 0;
	33	static int idle_threads = 0;
	34	static TAILQ_HEAD(, qemu_paiocb) request_list;
	35
2341f9a1	36	#ifdef CONFIG_PREADV
ceb42de8 AL	37	static int preadv_present = 1;
	38	#else
	39	static int preadv_present = 0;
	40	#endif
	41
8653c015	42	static void die2(int err, const char *what)
	43	{
	44	fprintf(stderr, "%s failed: %s\n", what, strerror(err));
	45	abort();
	46	}
	47
	48	static void die(const char *what)
	49	{
	50	die2(errno, what);
	51	}
	52
	53	static void mutex_lock(pthread_mutex_t *mutex)
	54	{
	55	int ret = pthread_mutex_lock(mutex);
	56	if (ret) die2(ret, "pthread_mutex_lock");
	57	}
	58
	59	static void mutex_unlock(pthread_mutex_t *mutex)
	60	{
	61	int ret = pthread_mutex_unlock(mutex);
	62	if (ret) die2(ret, "pthread_mutex_unlock");
	63	}
	64
	65	static int cond_timedwait(pthread_cond_t cond, pthread_mutex_t mutex,
	66	struct timespec *ts)
	67	{
	68	int ret = pthread_cond_timedwait(cond, mutex, ts);
	69	if (ret && ret != ETIMEDOUT) die2(ret, "pthread_cond_timedwait");
	70	return ret;
	71	}
	72
5d47e372	73	static void cond_signal(pthread_cond_t *cond)
8653c015	74	{
5d47e372	75	int ret = pthread_cond_signal(cond);
5d47e372	76	if (ret) die2(ret, "pthread_cond_signal");
8653c015	77	}
	78
	79	static void thread_create(pthread_t thread, pthread_attr_t attr,
	80	void (start_routine)(void), void arg)
	81	{
	82	int ret = pthread_create(thread, attr, start_routine, arg);
	83	if (ret) die2(ret, "pthread_create");
	84	}
	85
f141eafe AL	86	static size_t handle_aiocb_ioctl(struct qemu_paiocb *aiocb)
	87	{
	88	int ret;
	89
	90	ret = ioctl(aiocb->aio_fildes, aiocb->aio_ioctl_cmd, aiocb->aio_ioctl_buf);
	91	if (ret == -1)
	92	return -errno;
e7d54ae8 CH	93
	94	/*
	95	* This looks weird, but the aio code only consideres a request
	96	* successfull if it has written the number full number of bytes.
	97	*
	98	* Now we overload aio_nbytes as aio_ioctl_cmd for the ioctl command,
	99	* so in fact we return the ioctl command here to make posix_aio_read()
	100	* happy..
	101	*/
	102	return aiocb->aio_nbytes;
f141eafe AL	103	}
f141eafe AL	104
2341f9a1	105	#ifdef CONFIG_PREADV
ceb42de8 AL	106
	107	static ssize_t
	108	qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset)
	109	{
	110	return preadv(fd, iov, nr_iov, offset);
	111	}
	112
	113	static ssize_t
	114	qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset)
	115	{
	116	return pwritev(fd, iov, nr_iov, offset);
	117	}
	118
	119	#else
	120
	121	static ssize_t
	122	qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset)
	123	{
	124	return -ENOSYS;
	125	}
	126
	127	static ssize_t
	128	qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset)
	129	{
	130	return -ENOSYS;
	131	}
	132
	133	#endif
	134
f141eafe AL	135	/*
	136	* Check if we need to copy the data in the aiocb into a new
	137	* properly aligned buffer.
	138	*/
	139	static int aiocb_needs_copy(struct qemu_paiocb *aiocb)
	140	{
	141	if (aiocb->aio_flags & QEMU_AIO_SECTOR_ALIGNED) {
	142	int i;
	143
	144	for (i = 0; i < aiocb->aio_niov; i++)
	145	if ((uintptr_t) aiocb->aio_iov[i].iov_base % 512)
	146	return 1;
	147	}
	148
	149	return 0;
	150	}
	151
ceb42de8 AL	152	static size_t handle_aiocb_rw_vector(struct qemu_paiocb *aiocb)
	153	{
	154	size_t offset = 0;
	155	ssize_t len;
	156
	157	do {
	158	if (aiocb->aio_type == QEMU_PAIO_WRITE)
	159	len = qemu_pwritev(aiocb->aio_fildes,
	160	aiocb->aio_iov,
	161	aiocb->aio_niov,
	162	aiocb->aio_offset + offset);
	163	else
	164	len = qemu_preadv(aiocb->aio_fildes,
	165	aiocb->aio_iov,
	166	aiocb->aio_niov,
	167	aiocb->aio_offset + offset);
	168	} while (len == -1 && errno == EINTR);
	169
	170	if (len == -1)
	171	return -errno;
	172	return len;
	173	}
	174
f141eafe	175	static size_t handle_aiocb_rw_linear(struct qemu_paiocb aiocb, char buf)
221f715d AL	176	{
221f715d AL	177	size_t offset = 0;
f141eafe	178	size_t len;
221f715d AL	179
221f715d AL	180	while (offset < aiocb->aio_nbytes) {
f141eafe AL	181	if (aiocb->aio_type == QEMU_PAIO_WRITE)
	182	len = pwrite(aiocb->aio_fildes,
	183	(const char *)buf + offset,
	184	aiocb->aio_nbytes - offset,
	185	aiocb->aio_offset + offset);
	186	else
	187	len = pread(aiocb->aio_fildes,
	188	buf + offset,
221f715d AL	189	aiocb->aio_nbytes - offset,
221f715d AL	190	aiocb->aio_offset + offset);
221f715d	191
f141eafe AL	192	if (len == -1 && errno == EINTR)
	193	continue;
	194	else if (len == -1) {
	195	offset = -errno;
	196	break;
	197	} else if (len == 0)
	198	break;
	199
	200	offset += len;
221f715d AL	201	}
	202
	203	return offset;
	204	}
	205
f141eafe	206	static size_t handle_aiocb_rw(struct qemu_paiocb *aiocb)
221f715d	207	{
f141eafe AL	208	size_t nbytes;
	209	char *buf;
	210
ceb42de8	211	if (!aiocb_needs_copy(aiocb)) {
f141eafe AL	212	/*
	213	* If there is just a single buffer, and it is properly aligned
	214	* we can just use plain pread/pwrite without any problems.
	215	*/
ceb42de8 AL	216	if (aiocb->aio_niov == 1)
	217	return handle_aiocb_rw_linear(aiocb, aiocb->aio_iov->iov_base);
	218
	219	/*
	220	* We have more than one iovec, and all are properly aligned.
	221	*
	222	* Try preadv/pwritev first and fall back to linearizing the
	223	* buffer if it's not supported.
	224	*/
	225	if (preadv_present) {
	226	nbytes = handle_aiocb_rw_vector(aiocb);
	227	if (nbytes == aiocb->aio_nbytes)
	228	return nbytes;
	229	if (nbytes < 0 && nbytes != -ENOSYS)
	230	return nbytes;
	231	preadv_present = 0;
	232	}
	233
	234	/*
	235	* XXX(hch): short read/write. no easy way to handle the reminder
	236	* using these interfaces. For now retry using plain
	237	* pread/pwrite?
	238	*/
f141eafe	239	}
221f715d	240
f141eafe AL	241	/*
	242	* Ok, we have to do it the hard way, copy all segments into
	243	* a single aligned buffer.
	244	*/
	245	buf = qemu_memalign(512, aiocb->aio_nbytes);
	246	if (aiocb->aio_type == QEMU_PAIO_WRITE) {
	247	char *p = buf;
	248	int i;
	249
	250	for (i = 0; i < aiocb->aio_niov; ++i) {
	251	memcpy(p, aiocb->aio_iov[i].iov_base, aiocb->aio_iov[i].iov_len);
	252	p += aiocb->aio_iov[i].iov_len;
	253	}
	254	}
	255
	256	nbytes = handle_aiocb_rw_linear(aiocb, buf);
	257	if (aiocb->aio_type != QEMU_PAIO_WRITE) {
	258	char *p = buf;
	259	size_t count = aiocb->aio_nbytes, copy;
	260	int i;
	261
	262	for (i = 0; i < aiocb->aio_niov && count; ++i) {
	263	copy = count;
	264	if (copy > aiocb->aio_iov[i].iov_len)
	265	copy = aiocb->aio_iov[i].iov_len;
	266	memcpy(aiocb->aio_iov[i].iov_base, p, copy);
	267	p += copy;
	268	count -= copy;
	269	}
	270	}
	271	qemu_vfree(buf);
	272
	273	return nbytes;
221f715d AL	274	}
221f715d AL	275
3c529d93 AL	276	static void aio_thread(void unused)
3c529d93 AL	277	{
a8227a5a	278	pid_t pid;
3c529d93 AL	279	sigset_t set;
3c529d93 AL	280
a8227a5a	281	pid = getpid();
a8227a5a	282
3c529d93	283	/* block all signals */
8653c015	284	if (sigfillset(&set)) die("sigfillset");
8653c015	285	if (sigprocmask(SIG_BLOCK, &set, NULL)) die("sigprocmask");
3c529d93 AL	286
	287	while (1) {
	288	struct qemu_paiocb *aiocb;
221f715d	289	size_t ret = 0;
30525aff	290	qemu_timeval tv;
	291	struct timespec ts;
	292
	293	qemu_gettimeofday(&tv);
	294	ts.tv_sec = tv.tv_sec + 10;
	295	ts.tv_nsec = 0;
3c529d93	296
8653c015	297	mutex_lock(&lock);
3c529d93 AL	298
	299	while (TAILQ_EMPTY(&request_list) &&
	300	!(ret == ETIMEDOUT)) {
8653c015	301	ret = cond_timedwait(&cond, &lock, &ts);
3c529d93 AL	302	}
3c529d93 AL	303
514f7a27	304	if (TAILQ_EMPTY(&request_list))
3c529d93 AL	305	break;
	306
	307	aiocb = TAILQ_FIRST(&request_list);
	308	TAILQ_REMOVE(&request_list, aiocb, node);
3c529d93	309	aiocb->active = 1;
3c529d93	310	idle_threads--;
8653c015	311	mutex_unlock(&lock);
3c529d93	312
221f715d AL	313	switch (aiocb->aio_type) {
	314	case QEMU_PAIO_READ:
	315	case QEMU_PAIO_WRITE:
f141eafe	316	ret = handle_aiocb_rw(aiocb);
221f715d AL	317	break;
	318	case QEMU_PAIO_IOCTL:
	319	ret = handle_aiocb_ioctl(aiocb);
	320	break;
	321	default:
	322	fprintf(stderr, "invalid aio request (0x%x)\n", aiocb->aio_type);
	323	ret = -EINVAL;
	324	break;
	325	}
3c529d93	326
8653c015	327	mutex_lock(&lock);
221f715d	328	aiocb->ret = ret;
3c529d93	329	idle_threads++;
8653c015	330	mutex_unlock(&lock);
3c529d93	331
a8227a5a	332	if (kill(pid, aiocb->ev_signo)) die("kill failed");
3c529d93 AL	333	}
	334
	335	idle_threads--;
	336	cur_threads--;
8653c015	337	mutex_unlock(&lock);
3c529d93 AL	338
	339	return NULL;
	340	}
	341
8653c015	342	static void spawn_thread(void)
3c529d93	343	{
3c529d93 AL	344	cur_threads++;
3c529d93 AL	345	idle_threads++;
8653c015	346	thread_create(&thread_id, &attr, aio_thread, NULL);
3c529d93 AL	347	}
	348
	349	int qemu_paio_init(struct qemu_paioinit *aioinit)
	350	{
a8227a5a	351	int ret;
	352
	353	ret = pthread_attr_init(&attr);
	354	if (ret) die2(ret, "pthread_attr_init");
	355
	356	ret = pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
	357	if (ret) die2(ret, "pthread_attr_setdetachstate");
	358
3c529d93 AL	359	TAILQ_INIT(&request_list);
	360
	361	return 0;
	362	}
	363
221f715d	364	static int qemu_paio_submit(struct qemu_paiocb *aiocb, int type)
3c529d93	365	{
221f715d	366	aiocb->aio_type = type;
3c529d93 AL	367	aiocb->ret = -EINPROGRESS;
3c529d93 AL	368	aiocb->active = 0;
8653c015	369	mutex_lock(&lock);
3c529d93 AL	370	if (idle_threads == 0 && cur_threads < max_threads)
	371	spawn_thread();
	372	TAILQ_INSERT_TAIL(&request_list, aiocb, node);
8653c015	373	mutex_unlock(&lock);
5d47e372	374	cond_signal(&cond);
3c529d93 AL	375
	376	return 0;
	377	}
	378
	379	int qemu_paio_read(struct qemu_paiocb *aiocb)
	380	{
221f715d	381	return qemu_paio_submit(aiocb, QEMU_PAIO_READ);
3c529d93 AL	382	}
	383
	384	int qemu_paio_write(struct qemu_paiocb *aiocb)
	385	{
221f715d AL	386	return qemu_paio_submit(aiocb, QEMU_PAIO_WRITE);
	387	}
	388
	389	int qemu_paio_ioctl(struct qemu_paiocb *aiocb)
	390	{
	391	return qemu_paio_submit(aiocb, QEMU_PAIO_IOCTL);
3c529d93 AL	392	}
	393
	394	ssize_t qemu_paio_return(struct qemu_paiocb *aiocb)
	395	{
	396	ssize_t ret;
	397
8653c015	398	mutex_lock(&lock);
3c529d93	399	ret = aiocb->ret;
8653c015	400	mutex_unlock(&lock);
3c529d93 AL	401
	402	return ret;
	403	}
	404
	405	int qemu_paio_error(struct qemu_paiocb *aiocb)
	406	{
	407	ssize_t ret = qemu_paio_return(aiocb);
	408
	409	if (ret < 0)
	410	ret = -ret;
	411	else
	412	ret = 0;
	413
	414	return ret;
	415	}
	416
	417	int qemu_paio_cancel(int fd, struct qemu_paiocb *aiocb)
	418	{
	419	int ret;
	420
8653c015	421	mutex_lock(&lock);
3c529d93 AL	422	if (!aiocb->active) {
	423	TAILQ_REMOVE(&request_list, aiocb, node);
	424	aiocb->ret = -ECANCELED;
	425	ret = QEMU_PAIO_CANCELED;
	426	} else if (aiocb->ret == -EINPROGRESS)
	427	ret = QEMU_PAIO_NOTCANCELED;
	428	else
	429	ret = QEMU_PAIO_ALLDONE;
8653c015	430	mutex_unlock(&lock);
3c529d93 AL	431
	432	return ret;
	433	}