[mirror_qemu.git] / block / backup.c

/*
 * QEMU backup
 *
 * Copyright (C) 2013 Proxmox Server Solutions
 *
 * Authors:
 *  Dietmar Maurer (dietmar@proxmox.com)
 *
 * 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 <stdio.h>
#include <errno.h>
#include <unistd.h>

#include "trace.h"
#include "block/block.h"
#include "block/block_int.h"
#include "block/blockjob.h"
#include "qemu/ratelimit.h"

#define BACKUP_CLUSTER_BITS 16
#define BACKUP_CLUSTER_SIZE (1 << BACKUP_CLUSTER_BITS)
#define BACKUP_SECTORS_PER_CLUSTER (BACKUP_CLUSTER_SIZE / BDRV_SECTOR_SIZE)

#define SLICE_TIME 100000000ULL /* ns */

typedef struct CowRequest {
    int64_t start;
    int64_t end;
    QLIST_ENTRY(CowRequest) list;
    CoQueue wait_queue; /* coroutines blocked on this request */
} CowRequest;

typedef struct BackupBlockJob {
    BlockJob common;
    BlockDriverState *target;
    MirrorSyncMode sync_mode;
    RateLimit limit;
    BlockdevOnError on_source_error;
    BlockdevOnError on_target_error;
    CoRwlock flush_rwlock;
    uint64_t sectors_read;
    HBitmap *bitmap;
    QLIST_HEAD(, CowRequest) inflight_reqs;
} BackupBlockJob;

/* See if in-flight requests overlap and wait for them to complete */
static void coroutine_fn wait_for_overlapping_requests(BackupBlockJob *job,
                                                       int64_t start,
                                                       int64_t end)
{
    CowRequest *req;
    bool retry;

    do {
        retry = false;
        QLIST_FOREACH(req, &job->inflight_reqs, list) {
            if (end > req->start && start < req->end) {
                qemu_co_queue_wait(&req->wait_queue);
                retry = true;
                break;
            }
        }
    } while (retry);
}

/* Keep track of an in-flight request */
static void cow_request_begin(CowRequest *req, BackupBlockJob *job,
                                     int64_t start, int64_t end)
{
    req->start = start;
    req->end = end;
    qemu_co_queue_init(&req->wait_queue);
    QLIST_INSERT_HEAD(&job->inflight_reqs, req, list);
}

/* Forget about a completed request */
static void cow_request_end(CowRequest *req)
{
    QLIST_REMOVE(req, list);
    qemu_co_queue_restart_all(&req->wait_queue);
}

static int coroutine_fn backup_do_cow(BlockDriverState *bs,
                                      int64_t sector_num, int nb_sectors,
                                      bool *error_is_read)
{
    BackupBlockJob *job = (BackupBlockJob *)bs->job;
    CowRequest cow_request;
    struct iovec iov;
    QEMUIOVector bounce_qiov;
    void *bounce_buffer = NULL;
    int ret = 0;
    int64_t start, end;
    int n;

    qemu_co_rwlock_rdlock(&job->flush_rwlock);

    start = sector_num / BACKUP_SECTORS_PER_CLUSTER;
    end = DIV_ROUND_UP(sector_num + nb_sectors, BACKUP_SECTORS_PER_CLUSTER);

    trace_backup_do_cow_enter(job, start, sector_num, nb_sectors);

    wait_for_overlapping_requests(job, start, end);
    cow_request_begin(&cow_request, job, start, end);

    for (; start < end; start++) {
        if (hbitmap_get(job->bitmap, start)) {
            trace_backup_do_cow_skip(job, start);
            continue; /* already copied */
        }

        trace_backup_do_cow_process(job, start);

        n = MIN(BACKUP_SECTORS_PER_CLUSTER,
                job->common.len / BDRV_SECTOR_SIZE -
                start * BACKUP_SECTORS_PER_CLUSTER);

        if (!bounce_buffer) {
            bounce_buffer = qemu_blockalign(bs, BACKUP_CLUSTER_SIZE);
        }
        iov.iov_base = bounce_buffer;
        iov.iov_len = n * BDRV_SECTOR_SIZE;
        qemu_iovec_init_external(&bounce_qiov, &iov, 1);

        ret = bdrv_co_readv(bs, start * BACKUP_SECTORS_PER_CLUSTER, n,
                            &bounce_qiov);
        if (ret < 0) {
            trace_backup_do_cow_read_fail(job, start, ret);
            if (error_is_read) {
                *error_is_read = true;
            }
            goto out;
        }

        if (buffer_is_zero(iov.iov_base, iov.iov_len)) {
            ret = bdrv_co_write_zeroes(job->target,
                                       start * BACKUP_SECTORS_PER_CLUSTER,
                                       n, BDRV_REQ_MAY_UNMAP);
        } else {
            ret = bdrv_co_writev(job->target,
                                 start * BACKUP_SECTORS_PER_CLUSTER, n,
                                 &bounce_qiov);
        }
        if (ret < 0) {
            trace_backup_do_cow_write_fail(job, start, ret);
            if (error_is_read) {
                *error_is_read = false;
            }
            goto out;
        }

        hbitmap_set(job->bitmap, start, 1);

        /* Publish progress, guest I/O counts as progress too.  Note that the
         * offset field is an opaque progress value, it is not a disk offset.
         */
        job->sectors_read += n;
        job->common.offset += n * BDRV_SECTOR_SIZE;
    }

out:
    if (bounce_buffer) {
        qemu_vfree(bounce_buffer);
    }

    cow_request_end(&cow_request);

    trace_backup_do_cow_return(job, sector_num, nb_sectors, ret);

    qemu_co_rwlock_unlock(&job->flush_rwlock);

    return ret;
}

static int coroutine_fn backup_before_write_notify(
        NotifierWithReturn *notifier,
        void *opaque)
{
    BdrvTrackedRequest *req = opaque;
    int64_t sector_num = req->offset >> BDRV_SECTOR_BITS;
    int nb_sectors = req->bytes >> BDRV_SECTOR_BITS;

    assert((req->offset & (BDRV_SECTOR_SIZE - 1)) == 0);
    assert((req->bytes & (BDRV_SECTOR_SIZE - 1)) == 0);

    return backup_do_cow(req->bs, sector_num, nb_sectors, NULL);
}

static void backup_set_speed(BlockJob *job, int64_t speed, Error **errp)
{
    BackupBlockJob *s = container_of(job, BackupBlockJob, common);

    if (speed < 0) {
        error_set(errp, QERR_INVALID_PARAMETER, "speed");
        return;
    }
    ratelimit_set_speed(&s->limit, speed / BDRV_SECTOR_SIZE, SLICE_TIME);
}

static void backup_iostatus_reset(BlockJob *job)
{
    BackupBlockJob *s = container_of(job, BackupBlockJob, common);

    bdrv_iostatus_reset(s->target);
}

static const BlockJobDriver backup_job_driver = {
    .instance_size  = sizeof(BackupBlockJob),
    .job_type       = BLOCK_JOB_TYPE_BACKUP,
    .set_speed      = backup_set_speed,
    .iostatus_reset = backup_iostatus_reset,
};

static BlockErrorAction backup_error_action(BackupBlockJob *job,
                                            bool read, int error)
{
    if (read) {
        return block_job_error_action(&job->common, job->common.bs,
                                      job->on_source_error, true, error);
    } else {
        return block_job_error_action(&job->common, job->target,
                                      job->on_target_error, false, error);
    }
}

typedef struct {
    int ret;
} BackupCompleteData;

static void backup_complete(BlockJob *job, void *opaque)
{
    BackupBlockJob *s = container_of(job, BackupBlockJob, common);
    BackupCompleteData *data = opaque;

    bdrv_unref(s->target);

    block_job_completed(job, data->ret);
    g_free(data);
}

static void coroutine_fn backup_run(void *opaque)
{
    BackupBlockJob *job = opaque;
    BackupCompleteData *data;
    BlockDriverState *bs = job->common.bs;
    BlockDriverState *target = job->target;
    BlockdevOnError on_target_error = job->on_target_error;
    NotifierWithReturn before_write = {
        .notify = backup_before_write_notify,
    };
    int64_t start, end;
    int ret = 0;

    QLIST_INIT(&job->inflight_reqs);
    qemu_co_rwlock_init(&job->flush_rwlock);

    start = 0;
    end = DIV_ROUND_UP(job->common.len / BDRV_SECTOR_SIZE,
                       BACKUP_SECTORS_PER_CLUSTER);

    job->bitmap = hbitmap_alloc(end, 0);

    bdrv_set_enable_write_cache(target, true);
    bdrv_set_on_error(target, on_target_error, on_target_error);
    bdrv_iostatus_enable(target);

    bdrv_add_before_write_notifier(bs, &before_write);

    if (job->sync_mode == MIRROR_SYNC_MODE_NONE) {
        while (!block_job_is_cancelled(&job->common)) {
            /* Yield until the job is cancelled.  We just let our before_write
             * notify callback service CoW requests. */
            job->common.busy = false;
            qemu_coroutine_yield();
            job->common.busy = true;
        }
    } else {
        /* Both FULL and TOP SYNC_MODE's require copying.. */
        for (; start < end; start++) {
            bool error_is_read;

            if (block_job_is_cancelled(&job->common)) {
                break;
            }

            /* we need to yield so that bdrv_drain_all() returns.
             * (without, VM does not reboot)
             */
            if (job->common.speed) {
                uint64_t delay_ns = ratelimit_calculate_delay(
                        &job->limit, job->sectors_read);
                job->sectors_read = 0;
                block_job_sleep_ns(&job->common, QEMU_CLOCK_REALTIME, delay_ns);
            } else {
                block_job_sleep_ns(&job->common, QEMU_CLOCK_REALTIME, 0);
            }

            if (block_job_is_cancelled(&job->common)) {
                break;
            }

            if (job->sync_mode == MIRROR_SYNC_MODE_TOP) {
                int i, n;
                int alloced = 0;

                /* Check to see if these blocks are already in the
                 * backing file. */

                for (i = 0; i < BACKUP_SECTORS_PER_CLUSTER;) {
                    /* bdrv_is_allocated() only returns true/false based
                     * on the first set of sectors it comes across that
                     * are are all in the same state.
                     * For that reason we must verify each sector in the
                     * backup cluster length.  We end up copying more than
                     * needed but at some point that is always the case. */
                    alloced =
                        bdrv_is_allocated(bs,
                                start * BACKUP_SECTORS_PER_CLUSTER + i,
                                BACKUP_SECTORS_PER_CLUSTER - i, &n);
                    i += n;

                    if (alloced == 1 || n == 0) {
                        break;
                    }
                }

                /* If the above loop never found any sectors that are in
                 * the topmost image, skip this backup. */
                if (alloced == 0) {
                    continue;
                }
            }
            /* FULL sync mode we copy the whole drive. */
            ret = backup_do_cow(bs, start * BACKUP_SECTORS_PER_CLUSTER,
                    BACKUP_SECTORS_PER_CLUSTER, &error_is_read);
            if (ret < 0) {
                /* Depending on error action, fail now or retry cluster */
                BlockErrorAction action =
                    backup_error_action(job, error_is_read, -ret);
                if (action == BLOCK_ERROR_ACTION_REPORT) {
                    break;
                } else {
                    start--;
                    continue;
                }
            }
        }
    }

    notifier_with_return_remove(&before_write);

    /* wait until pending backup_do_cow() calls have completed */
    qemu_co_rwlock_wrlock(&job->flush_rwlock);
    qemu_co_rwlock_unlock(&job->flush_rwlock);

    hbitmap_free(job->bitmap);

    bdrv_iostatus_disable(target);
    bdrv_op_unblock_all(target, job->common.blocker);

    data = g_malloc(sizeof(*data));
    data->ret = ret;
    block_job_defer_to_main_loop(&job->common, backup_complete, data);
}

void backup_start(BlockDriverState *bs, BlockDriverState *target,
                  int64_t speed, MirrorSyncMode sync_mode,
                  BlockdevOnError on_source_error,
                  BlockdevOnError on_target_error,
                  BlockCompletionFunc *cb, void *opaque,
                  Error **errp)
{
    int64_t len;

    assert(bs);
    assert(target);
    assert(cb);

    if (bs == target) {
        error_setg(errp, "Source and target cannot be the same");
        return;
    }

    if ((on_source_error == BLOCKDEV_ON_ERROR_STOP ||
         on_source_error == BLOCKDEV_ON_ERROR_ENOSPC) &&
        !bdrv_iostatus_is_enabled(bs)) {
        error_set(errp, QERR_INVALID_PARAMETER, "on-source-error");
        return;
    }

    if (!bdrv_is_inserted(bs)) {
        error_setg(errp, "Device is not inserted: %s",
                   bdrv_get_device_name(bs));
        return;
    }

    if (!bdrv_is_inserted(target)) {
        error_setg(errp, "Device is not inserted: %s",
                   bdrv_get_device_name(target));
        return;
    }

    if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
        return;
    }

    if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_BACKUP_TARGET, errp)) {
        return;
    }

    len = bdrv_getlength(bs);
    if (len < 0) {
        error_setg_errno(errp, -len, "unable to get length for '%s'",
                         bdrv_get_device_name(bs));
        return;
    }

    BackupBlockJob *job = block_job_create(&backup_job_driver, bs, speed,
                                           cb, opaque, errp);
    if (!job) {
        return;
    }

    bdrv_op_block_all(target, job->common.blocker);

    job->on_source_error = on_source_error;
    job->on_target_error = on_target_error;
    job->target = target;
    job->sync_mode = sync_mode;
    job->common.len = len;
    job->common.co = qemu_coroutine_create(backup_run);
    qemu_coroutine_enter(job->common.co, job);
}
Commit	Line	Data
98d2c6f2 DM	1	/*
	2	* QEMU backup
	3	*
	4	* Copyright (C) 2013 Proxmox Server Solutions
	5	*
	6	* Authors:
	7	* Dietmar Maurer (dietmar@proxmox.com)
	8	*
	9	* This work is licensed under the terms of the GNU GPL, version 2 or later.
	10	* See the COPYING file in the top-level directory.
	11	*
	12	*/
	13
	14	#include <stdio.h>
	15	#include <errno.h>
	16	#include <unistd.h>
	17
	18	#include "trace.h"
	19	#include "block/block.h"
	20	#include "block/block_int.h"
	21	#include "block/blockjob.h"
	22	#include "qemu/ratelimit.h"
	23
	24	#define BACKUP_CLUSTER_BITS 16
	25	#define BACKUP_CLUSTER_SIZE (1 << BACKUP_CLUSTER_BITS)
	26	#define BACKUP_SECTORS_PER_CLUSTER (BACKUP_CLUSTER_SIZE / BDRV_SECTOR_SIZE)
	27
	28	#define SLICE_TIME 100000000ULL /* ns */
	29
	30	typedef struct CowRequest {
	31	int64_t start;
	32	int64_t end;
	33	QLIST_ENTRY(CowRequest) list;
	34	CoQueue wait_queue; /* coroutines blocked on this request */
	35	} CowRequest;
	36
	37	typedef struct BackupBlockJob {
	38	BlockJob common;
	39	BlockDriverState *target;
fc5d3f84	40	MirrorSyncMode sync_mode;
98d2c6f2 DM	41	RateLimit limit;
	42	BlockdevOnError on_source_error;
	43	BlockdevOnError on_target_error;
	44	CoRwlock flush_rwlock;
	45	uint64_t sectors_read;
	46	HBitmap *bitmap;
	47	QLIST_HEAD(, CowRequest) inflight_reqs;
	48	} BackupBlockJob;
	49
	50	/* See if in-flight requests overlap and wait for them to complete */
	51	static void coroutine_fn wait_for_overlapping_requests(BackupBlockJob *job,
	52	int64_t start,
	53	int64_t end)
	54	{
	55	CowRequest *req;
	56	bool retry;
	57
	58	do {
	59	retry = false;
	60	QLIST_FOREACH(req, &job->inflight_reqs, list) {
	61	if (end > req->start && start < req->end) {
	62	qemu_co_queue_wait(&req->wait_queue);
	63	retry = true;
	64	break;
	65	}
	66	}
	67	} while (retry);
	68	}
	69
	70	/* Keep track of an in-flight request */
	71	static void cow_request_begin(CowRequest req, BackupBlockJob job,
	72	int64_t start, int64_t end)
	73	{
	74	req->start = start;
	75	req->end = end;
	76	qemu_co_queue_init(&req->wait_queue);
	77	QLIST_INSERT_HEAD(&job->inflight_reqs, req, list);
	78	}
	79
	80	/* Forget about a completed request */
	81	static void cow_request_end(CowRequest *req)
	82	{
	83	QLIST_REMOVE(req, list);
	84	qemu_co_queue_restart_all(&req->wait_queue);
	85	}
	86
	87	static int coroutine_fn backup_do_cow(BlockDriverState *bs,
	88	int64_t sector_num, int nb_sectors,
	89	bool *error_is_read)
	90	{
	91	BackupBlockJob job = (BackupBlockJob )bs->job;
	92	CowRequest cow_request;
	93	struct iovec iov;
	94	QEMUIOVector bounce_qiov;
	95	void *bounce_buffer = NULL;
	96	int ret = 0;
	97	int64_t start, end;
	98	int n;
	99
	100	qemu_co_rwlock_rdlock(&job->flush_rwlock);
	101
	102	start = sector_num / BACKUP_SECTORS_PER_CLUSTER;
	103	end = DIV_ROUND_UP(sector_num + nb_sectors, BACKUP_SECTORS_PER_CLUSTER);
	104
105	trace_backup_do_cow_enter(job, start, sector_num, nb_sectors);
106
107	wait_for_overlapping_requests(job, start, end);
108	cow_request_begin(&cow_request, job, start, end);
109
110	for (; start < end; start++) {
111	if (hbitmap_get(job->bitmap, start)) {
112	trace_backup_do_cow_skip(job, start);
113	continue; /* already copied */
114	}
115
116	trace_backup_do_cow_process(job, start);
117
118	n = MIN(BACKUP_SECTORS_PER_CLUSTER,
119	job->common.len / BDRV_SECTOR_SIZE -
120	start * BACKUP_SECTORS_PER_CLUSTER);
121
122	if (!bounce_buffer) {
123	bounce_buffer = qemu_blockalign(bs, BACKUP_CLUSTER_SIZE);
124	}
125	iov.iov_base = bounce_buffer;
126	iov.iov_len = n * BDRV_SECTOR_SIZE;
127	qemu_iovec_init_external(&bounce_qiov, &iov, 1);
128
129	ret = bdrv_co_readv(bs, start * BACKUP_SECTORS_PER_CLUSTER, n,
130	&bounce_qiov);
131	if (ret < 0) {
132	trace_backup_do_cow_read_fail(job, start, ret);
133	if (error_is_read) {
134	*error_is_read = true;
135	}
136	goto out;
137	}
138
139	if (buffer_is_zero(iov.iov_base, iov.iov_len)) {
140	ret = bdrv_co_write_zeroes(job->target,
aa7bfbff	141	start * BACKUP_SECTORS_PER_CLUSTER,
d32f35cb	142	n, BDRV_REQ_MAY_UNMAP);
98d2c6f2 DM	143	} else {
	144	ret = bdrv_co_writev(job->target,
	145	start * BACKUP_SECTORS_PER_CLUSTER, n,
	146	&bounce_qiov);
	147	}
	148	if (ret < 0) {
	149	trace_backup_do_cow_write_fail(job, start, ret);
	150	if (error_is_read) {
	151	*error_is_read = false;
	152	}
	153	goto out;
	154	}
	155
	156	hbitmap_set(job->bitmap, start, 1);
	157
	158	/* Publish progress, guest I/O counts as progress too. Note that the
	159	* offset field is an opaque progress value, it is not a disk offset.
	160	*/
	161	job->sectors_read += n;
	162	job->common.offset += n * BDRV_SECTOR_SIZE;
	163	}
	164
	165	out:
	166	if (bounce_buffer) {
	167	qemu_vfree(bounce_buffer);
	168	}
	169
	170	cow_request_end(&cow_request);
	171
	172	trace_backup_do_cow_return(job, sector_num, nb_sectors, ret);
	173
	174	qemu_co_rwlock_unlock(&job->flush_rwlock);
	175
	176	return ret;
	177	}
	178
	179	static int coroutine_fn backup_before_write_notify(
	180	NotifierWithReturn *notifier,
	181	void *opaque)
	182	{
	183	BdrvTrackedRequest *req = opaque;
793ed47a KW	184	int64_t sector_num = req->offset >> BDRV_SECTOR_BITS;
793ed47a KW	185	int nb_sectors = req->bytes >> BDRV_SECTOR_BITS;
98d2c6f2	186
793ed47a KW	187	assert((req->offset & (BDRV_SECTOR_SIZE - 1)) == 0);
	188	assert((req->bytes & (BDRV_SECTOR_SIZE - 1)) == 0);
	189
	190	return backup_do_cow(req->bs, sector_num, nb_sectors, NULL);
98d2c6f2 DM	191	}
	192
	193	static void backup_set_speed(BlockJob job, int64_t speed, Error *errp)
	194	{
	195	BackupBlockJob *s = container_of(job, BackupBlockJob, common);
	196
	197	if (speed < 0) {
	198	error_set(errp, QERR_INVALID_PARAMETER, "speed");
	199	return;
	200	}
	201	ratelimit_set_speed(&s->limit, speed / BDRV_SECTOR_SIZE, SLICE_TIME);
	202	}
	203
	204	static void backup_iostatus_reset(BlockJob *job)
	205	{
	206	BackupBlockJob *s = container_of(job, BackupBlockJob, common);
	207
	208	bdrv_iostatus_reset(s->target);
	209	}
	210
3fc4b10a	211	static const BlockJobDriver backup_job_driver = {
98d2c6f2	212	.instance_size = sizeof(BackupBlockJob),
79e14bf7	213	.job_type = BLOCK_JOB_TYPE_BACKUP,
98d2c6f2 DM	214	.set_speed = backup_set_speed,
	215	.iostatus_reset = backup_iostatus_reset,
	216	};
	217
	218	static BlockErrorAction backup_error_action(BackupBlockJob *job,
	219	bool read, int error)
	220	{
	221	if (read) {
	222	return block_job_error_action(&job->common, job->common.bs,
	223	job->on_source_error, true, error);
	224	} else {
	225	return block_job_error_action(&job->common, job->target,
	226	job->on_target_error, false, error);
	227	}
	228	}
	229
761731b1 SH	230	typedef struct {
	231	int ret;
	232	} BackupCompleteData;
	233
	234	static void backup_complete(BlockJob job, void opaque)
	235	{
	236	BackupBlockJob *s = container_of(job, BackupBlockJob, common);
	237	BackupCompleteData *data = opaque;
	238
	239	bdrv_unref(s->target);
	240
	241	block_job_completed(job, data->ret);
	242	g_free(data);
	243	}
	244
98d2c6f2 DM	245	static void coroutine_fn backup_run(void *opaque)
	246	{
	247	BackupBlockJob *job = opaque;
761731b1	248	BackupCompleteData *data;
98d2c6f2 DM	249	BlockDriverState *bs = job->common.bs;
	250	BlockDriverState *target = job->target;
	251	BlockdevOnError on_target_error = job->on_target_error;
	252	NotifierWithReturn before_write = {
	253	.notify = backup_before_write_notify,
	254	};
	255	int64_t start, end;
	256	int ret = 0;
	257
	258	QLIST_INIT(&job->inflight_reqs);
	259	qemu_co_rwlock_init(&job->flush_rwlock);
	260
	261	start = 0;
	262	end = DIV_ROUND_UP(job->common.len / BDRV_SECTOR_SIZE,
	263	BACKUP_SECTORS_PER_CLUSTER);
	264
	265	job->bitmap = hbitmap_alloc(end, 0);
	266
	267	bdrv_set_enable_write_cache(target, true);
	268	bdrv_set_on_error(target, on_target_error, on_target_error);
	269	bdrv_iostatus_enable(target);
	270
	271	bdrv_add_before_write_notifier(bs, &before_write);
	272
fc5d3f84 IM	273	if (job->sync_mode == MIRROR_SYNC_MODE_NONE) {
	274	while (!block_job_is_cancelled(&job->common)) {
	275	/* Yield until the job is cancelled. We just let our before_write
	276	* notify callback service CoW requests. */
	277	job->common.busy = false;
	278	qemu_coroutine_yield();
	279	job->common.busy = true;
98d2c6f2	280	}
fc5d3f84 IM	281	} else {
	282	/* Both FULL and TOP SYNC_MODE's require copying.. */
	283	for (; start < end; start++) {
	284	bool error_is_read;
98d2c6f2	285
fc5d3f84 IM	286	if (block_job_is_cancelled(&job->common)) {
	287	break;
	288	}
98d2c6f2	289
a7282330	290	/* we need to yield so that bdrv_drain_all() returns.
fc5d3f84 IM	291	* (without, VM does not reboot)
	292	*/
	293	if (job->common.speed) {
	294	uint64_t delay_ns = ratelimit_calculate_delay(
	295	&job->limit, job->sectors_read);
	296	job->sectors_read = 0;
7483d1e5	297	block_job_sleep_ns(&job->common, QEMU_CLOCK_REALTIME, delay_ns);
fc5d3f84	298	} else {
7483d1e5	299	block_job_sleep_ns(&job->common, QEMU_CLOCK_REALTIME, 0);
fc5d3f84	300	}
98d2c6f2	301
fc5d3f84	302	if (block_job_is_cancelled(&job->common)) {
98d2c6f2	303	break;
fc5d3f84 IM	304	}
	305
	306	if (job->sync_mode == MIRROR_SYNC_MODE_TOP) {
	307	int i, n;
	308	int alloced = 0;
	309
	310	/* Check to see if these blocks are already in the
	311	* backing file. */
	312
	313	for (i = 0; i < BACKUP_SECTORS_PER_CLUSTER;) {
bdad13b9	314	/* bdrv_is_allocated() only returns true/false based
4c293dc6	315	* on the first set of sectors it comes across that
fc5d3f84 IM	316	* are are all in the same state.
	317	* For that reason we must verify each sector in the
	318	* backup cluster length. We end up copying more than
	319	* needed but at some point that is always the case. */
	320	alloced =
bdad13b9	321	bdrv_is_allocated(bs,
fc5d3f84 IM	322	start * BACKUP_SECTORS_PER_CLUSTER + i,
	323	BACKUP_SECTORS_PER_CLUSTER - i, &n);
	324	i += n;
	325
d40593dd	326	if (alloced == 1 \|\| n == 0) {
fc5d3f84 IM	327	break;
	328	}
	329	}
	330
	331	/* If the above loop never found any sectors that are in
	332	* the topmost image, skip this backup. */
	333	if (alloced == 0) {
	334	continue;
	335	}
	336	}
	337	/* FULL sync mode we copy the whole drive. */
	338	ret = backup_do_cow(bs, start * BACKUP_SECTORS_PER_CLUSTER,
	339	BACKUP_SECTORS_PER_CLUSTER, &error_is_read);
	340	if (ret < 0) {
	341	/* Depending on error action, fail now or retry cluster */
	342	BlockErrorAction action =
	343	backup_error_action(job, error_is_read, -ret);
a589569f	344	if (action == BLOCK_ERROR_ACTION_REPORT) {
fc5d3f84 IM	345	break;
	346	} else {
	347	start--;
	348	continue;
	349	}
98d2c6f2 DM	350	}
	351	}
	352	}
	353
	354	notifier_with_return_remove(&before_write);
	355
	356	/* wait until pending backup_do_cow() calls have completed */
	357	qemu_co_rwlock_wrlock(&job->flush_rwlock);
	358	qemu_co_rwlock_unlock(&job->flush_rwlock);
	359
	360	hbitmap_free(job->bitmap);
	361
	362	bdrv_iostatus_disable(target);
c29c1dd3	363	bdrv_op_unblock_all(target, job->common.blocker);
98d2c6f2	364
761731b1 SH	365	data = g_malloc(sizeof(*data));
	366	data->ret = ret;
	367	block_job_defer_to_main_loop(&job->common, backup_complete, data);
98d2c6f2 DM	368	}
	369
	370	void backup_start(BlockDriverState bs, BlockDriverState target,
fc5d3f84	371	int64_t speed, MirrorSyncMode sync_mode,
98d2c6f2 DM	372	BlockdevOnError on_source_error,
98d2c6f2 DM	373	BlockdevOnError on_target_error,
097310b5	374	BlockCompletionFunc cb, void opaque,
98d2c6f2 DM	375	Error **errp)
	376	{
	377	int64_t len;
	378
	379	assert(bs);
	380	assert(target);
	381	assert(cb);
	382
c29c1dd3 FZ	383	if (bs == target) {
	384	error_setg(errp, "Source and target cannot be the same");
	385	return;
	386	}
	387
98d2c6f2 DM	388	if ((on_source_error == BLOCKDEV_ON_ERROR_STOP \|\|
	389	on_source_error == BLOCKDEV_ON_ERROR_ENOSPC) &&
	390	!bdrv_iostatus_is_enabled(bs)) {
	391	error_set(errp, QERR_INVALID_PARAMETER, "on-source-error");
	392	return;
	393	}
	394
c29c1dd3 FZ	395	if (!bdrv_is_inserted(bs)) {
	396	error_setg(errp, "Device is not inserted: %s",
	397	bdrv_get_device_name(bs));
	398	return;
	399	}
	400
	401	if (!bdrv_is_inserted(target)) {
	402	error_setg(errp, "Device is not inserted: %s",
	403	bdrv_get_device_name(target));
	404	return;
	405	}
	406
	407	if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
	408	return;
	409	}
	410
	411	if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_BACKUP_TARGET, errp)) {
	412	return;
	413	}
	414
98d2c6f2 DM	415	len = bdrv_getlength(bs);
	416	if (len < 0) {
	417	error_setg_errno(errp, -len, "unable to get length for '%s'",
	418	bdrv_get_device_name(bs));
	419	return;
	420	}
	421
3fc4b10a	422	BackupBlockJob *job = block_job_create(&backup_job_driver, bs, speed,
98d2c6f2 DM	423	cb, opaque, errp);
	424	if (!job) {
	425	return;
	426	}
	427
c29c1dd3 FZ	428	bdrv_op_block_all(target, job->common.blocker);
c29c1dd3 FZ	429
98d2c6f2 DM	430	job->on_source_error = on_source_error;
	431	job->on_target_error = on_target_error;
	432	job->target = target;
fc5d3f84	433	job->sync_mode = sync_mode;
98d2c6f2 DM	434	job->common.len = len;
	435	job->common.co = qemu_coroutine_create(backup_run);
	436	qemu_coroutine_enter(job->common.co, job);
	437	}