[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 "qapi/qmp/qerror.h"
#include "qemu/ratelimit.h"
#include "sysemu/block-backend.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;
    /* bitmap for sync=incremental */
    BdrvDirtyBitmap *sync_bitmap;
    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,
                                      bool is_write_notifier)
{
    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);

        if (is_write_notifier) {
            ret = bdrv_co_no_copy_on_readv(bs,
                                           start * BACKUP_SECTORS_PER_CLUSTER,
                                           n, &bounce_qiov);
        } else {
            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, true);
}

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

    if (speed < 0) {
        error_setg(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);

    if (s->target->blk) {
        blk_iostatus_reset(s->target->blk);
    }
}

static void backup_cleanup_sync_bitmap(BackupBlockJob *job, int ret)
{
    BdrvDirtyBitmap *bm;
    BlockDriverState *bs = job->common.bs;

    if (ret < 0 || block_job_is_cancelled(&job->common)) {
        /* Merge the successor back into the parent, delete nothing. */
        bm = bdrv_reclaim_dirty_bitmap(bs, job->sync_bitmap, NULL);
        assert(bm);
    } else {
        /* Everything is fine, delete this bitmap and install the backup. */
        bm = bdrv_dirty_bitmap_abdicate(bs, job->sync_bitmap, NULL);
        assert(bm);
    }
}

static void backup_commit(BlockJob *job)
{
    BackupBlockJob *s = container_of(job, BackupBlockJob, common);
    if (s->sync_bitmap) {
        backup_cleanup_sync_bitmap(s, 0);
    }
}

static void backup_abort(BlockJob *job)
{
    BackupBlockJob *s = container_of(job, BackupBlockJob, common);
    if (s->sync_bitmap) {
        backup_cleanup_sync_bitmap(s, -1);
    }
}

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,
    .commit         = backup_commit,
    .abort          = backup_abort,
};

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 bool coroutine_fn yield_and_check(BackupBlockJob *job)
{
    if (block_job_is_cancelled(&job->common)) {
        return true;
    }

    /* 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)) {
        return true;
    }

    return false;
}

static int coroutine_fn backup_run_incremental(BackupBlockJob *job)
{
    bool error_is_read;
    int ret = 0;
    int clusters_per_iter;
    uint32_t granularity;
    int64_t sector;
    int64_t cluster;
    int64_t end;
    int64_t last_cluster = -1;
    BlockDriverState *bs = job->common.bs;
    HBitmapIter hbi;

    granularity = bdrv_dirty_bitmap_granularity(job->sync_bitmap);
    clusters_per_iter = MAX((granularity / BACKUP_CLUSTER_SIZE), 1);
    bdrv_dirty_iter_init(job->sync_bitmap, &hbi);

    /* Find the next dirty sector(s) */
    while ((sector = hbitmap_iter_next(&hbi)) != -1) {
        cluster = sector / BACKUP_SECTORS_PER_CLUSTER;

        /* Fake progress updates for any clusters we skipped */
        if (cluster != last_cluster + 1) {
            job->common.offset += ((cluster - last_cluster - 1) *
                                   BACKUP_CLUSTER_SIZE);
        }

        for (end = cluster + clusters_per_iter; cluster < end; cluster++) {
            do {
                if (yield_and_check(job)) {
                    return ret;
                }
                ret = backup_do_cow(bs, cluster * BACKUP_SECTORS_PER_CLUSTER,
                                    BACKUP_SECTORS_PER_CLUSTER, &error_is_read,
                                    false);
                if ((ret < 0) &&
                    backup_error_action(job, error_is_read, -ret) ==
                    BLOCK_ERROR_ACTION_REPORT) {
                    return ret;
                }
            } while (ret < 0);
        }

        /* If the bitmap granularity is smaller than the backup granularity,
         * we need to advance the iterator pointer to the next cluster. */
        if (granularity < BACKUP_CLUSTER_SIZE) {
            bdrv_set_dirty_iter(&hbi, cluster * BACKUP_SECTORS_PER_CLUSTER);
        }

        last_cluster = cluster - 1;
    }

    /* Play some final catchup with the progress meter */
    end = DIV_ROUND_UP(job->common.len, BACKUP_CLUSTER_SIZE);
    if (last_cluster + 1 < end) {
        job->common.offset += ((end - last_cluster - 1) * BACKUP_CLUSTER_SIZE);
    }

    return ret;
}

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, BACKUP_CLUSTER_SIZE);

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

    bdrv_set_enable_write_cache(target, true);
    if (target->blk) {
        blk_set_on_error(target->blk, on_target_error, on_target_error);
        blk_iostatus_enable(target->blk);
    }

    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 if (job->sync_mode == MIRROR_SYNC_MODE_INCREMENTAL) {
        ret = backup_run_incremental(job);
    } else {
        /* Both FULL and TOP SYNC_MODE's require copying.. */
        for (; start < end; start++) {
            bool error_is_read;
            if (yield_and_check(job)) {
                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, false);
            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);

    if (target->blk) {
        blk_iostatus_disable(target->blk);
    }
    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,
                  BdrvDirtyBitmap *sync_bitmap,
                  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) &&
        (!bs->blk || !blk_iostatus_is_enabled(bs->blk))) {
        error_setg(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;
    }

    if (sync_mode == MIRROR_SYNC_MODE_INCREMENTAL) {
        if (!sync_bitmap) {
            error_setg(errp, "must provide a valid bitmap name for "
                             "\"incremental\" sync mode");
            return;
        }

        /* Create a new bitmap, and freeze/disable this one. */
        if (bdrv_dirty_bitmap_create_successor(bs, sync_bitmap, errp) < 0) {
            return;
        }
    } else if (sync_bitmap) {
        error_setg(errp,
                   "a sync_bitmap was provided to backup_run, "
                   "but received an incompatible sync_mode (%s)",
                   MirrorSyncMode_lookup[sync_mode]);
        return;
    }

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

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

    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->sync_bitmap = sync_mode == MIRROR_SYNC_MODE_INCREMENTAL ?
                       sync_bitmap : NULL;
    job->common.len = len;
    job->common.co = qemu_coroutine_create(backup_run);
    qemu_coroutine_enter(job->common.co, job);
    return;

 error:
    if (sync_bitmap) {
        bdrv_reclaim_dirty_bitmap(bs, sync_bitmap, NULL);
    }
}
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"
cc7a8ea7	22	#include "qapi/qmp/qerror.h"
98d2c6f2	23	#include "qemu/ratelimit.h"
373340b2	24	#include "sysemu/block-backend.h"
98d2c6f2 DM	25
	26	#define BACKUP_CLUSTER_BITS 16
	27	#define BACKUP_CLUSTER_SIZE (1 << BACKUP_CLUSTER_BITS)
	28	#define BACKUP_SECTORS_PER_CLUSTER (BACKUP_CLUSTER_SIZE / BDRV_SECTOR_SIZE)
	29
	30	#define SLICE_TIME 100000000ULL /* ns */
	31
	32	typedef struct CowRequest {
	33	int64_t start;
	34	int64_t end;
	35	QLIST_ENTRY(CowRequest) list;
	36	CoQueue wait_queue; /* coroutines blocked on this request */
	37	} CowRequest;
	38
	39	typedef struct BackupBlockJob {
	40	BlockJob common;
	41	BlockDriverState *target;
4b80ab2b	42	/* bitmap for sync=incremental */
d58d8453	43	BdrvDirtyBitmap *sync_bitmap;
fc5d3f84	44	MirrorSyncMode sync_mode;
98d2c6f2 DM	45	RateLimit limit;
	46	BlockdevOnError on_source_error;
	47	BlockdevOnError on_target_error;
	48	CoRwlock flush_rwlock;
	49	uint64_t sectors_read;
	50	HBitmap *bitmap;
	51	QLIST_HEAD(, CowRequest) inflight_reqs;
	52	} BackupBlockJob;
	53
	54	/* See if in-flight requests overlap and wait for them to complete */
	55	static void coroutine_fn wait_for_overlapping_requests(BackupBlockJob *job,
	56	int64_t start,
	57	int64_t end)
	58	{
	59	CowRequest *req;
	60	bool retry;
	61
	62	do {
	63	retry = false;
	64	QLIST_FOREACH(req, &job->inflight_reqs, list) {
	65	if (end > req->start && start < req->end) {
	66	qemu_co_queue_wait(&req->wait_queue);
	67	retry = true;
	68	break;
	69	}
	70	}
	71	} while (retry);
	72	}
	73
	74	/* Keep track of an in-flight request */
	75	static void cow_request_begin(CowRequest req, BackupBlockJob job,
	76	int64_t start, int64_t end)
	77	{
	78	req->start = start;
	79	req->end = end;
	80	qemu_co_queue_init(&req->wait_queue);
	81	QLIST_INSERT_HEAD(&job->inflight_reqs, req, list);
	82	}
	83
	84	/* Forget about a completed request */
	85	static void cow_request_end(CowRequest *req)
	86	{
	87	QLIST_REMOVE(req, list);
	88	qemu_co_queue_restart_all(&req->wait_queue);
	89	}
	90
	91	static int coroutine_fn backup_do_cow(BlockDriverState *bs,
	92	int64_t sector_num, int nb_sectors,
06c3916b WC	93	bool *error_is_read,
06c3916b WC	94	bool is_write_notifier)
98d2c6f2 DM	95	{
	96	BackupBlockJob job = (BackupBlockJob )bs->job;
	97	CowRequest cow_request;
	98	struct iovec iov;
	99	QEMUIOVector bounce_qiov;
	100	void *bounce_buffer = NULL;
	101	int ret = 0;
	102	int64_t start, end;
	103	int n;
	104
	105	qemu_co_rwlock_rdlock(&job->flush_rwlock);
	106
	107	start = sector_num / BACKUP_SECTORS_PER_CLUSTER;
	108	end = DIV_ROUND_UP(sector_num + nb_sectors, BACKUP_SECTORS_PER_CLUSTER);
	109
	110	trace_backup_do_cow_enter(job, start, sector_num, nb_sectors);
	111
	112	wait_for_overlapping_requests(job, start, end);
	113	cow_request_begin(&cow_request, job, start, end);
	114
	115	for (; start < end; start++) {
	116	if (hbitmap_get(job->bitmap, start)) {
	117	trace_backup_do_cow_skip(job, start);
	118	continue; /* already copied */
	119	}
	120
	121	trace_backup_do_cow_process(job, start);
	122
	123	n = MIN(BACKUP_SECTORS_PER_CLUSTER,
	124	job->common.len / BDRV_SECTOR_SIZE -
	125	start * BACKUP_SECTORS_PER_CLUSTER);
	126
	127	if (!bounce_buffer) {
	128	bounce_buffer = qemu_blockalign(bs, BACKUP_CLUSTER_SIZE);
	129	}
	130	iov.iov_base = bounce_buffer;
	131	iov.iov_len = n * BDRV_SECTOR_SIZE;
	132	qemu_iovec_init_external(&bounce_qiov, &iov, 1);
	133
06c3916b WC	134	if (is_write_notifier) {
	135	ret = bdrv_co_no_copy_on_readv(bs,
	136	start * BACKUP_SECTORS_PER_CLUSTER,
	137	n, &bounce_qiov);
	138	} else {
	139	ret = bdrv_co_readv(bs, start * BACKUP_SECTORS_PER_CLUSTER, n,
	140	&bounce_qiov);
	141	}
98d2c6f2 DM	142	if (ret < 0) {
	143	trace_backup_do_cow_read_fail(job, start, ret);
	144	if (error_is_read) {
	145	*error_is_read = true;
	146	}
	147	goto out;
	148	}
	149
	150	if (buffer_is_zero(iov.iov_base, iov.iov_len)) {
	151	ret = bdrv_co_write_zeroes(job->target,
aa7bfbff	152	start * BACKUP_SECTORS_PER_CLUSTER,
d32f35cb	153	n, BDRV_REQ_MAY_UNMAP);
98d2c6f2 DM	154	} else {
	155	ret = bdrv_co_writev(job->target,
	156	start * BACKUP_SECTORS_PER_CLUSTER, n,
	157	&bounce_qiov);
	158	}
	159	if (ret < 0) {
	160	trace_backup_do_cow_write_fail(job, start, ret);
	161	if (error_is_read) {
	162	*error_is_read = false;
	163	}
	164	goto out;
	165	}
	166
	167	hbitmap_set(job->bitmap, start, 1);
	168
	169	/* Publish progress, guest I/O counts as progress too. Note that the
	170	* offset field is an opaque progress value, it is not a disk offset.
	171	*/
	172	job->sectors_read += n;
	173	job->common.offset += n * BDRV_SECTOR_SIZE;
	174	}
	175
	176	out:
	177	if (bounce_buffer) {
	178	qemu_vfree(bounce_buffer);
	179	}
	180
	181	cow_request_end(&cow_request);
	182
	183	trace_backup_do_cow_return(job, sector_num, nb_sectors, ret);
	184
	185	qemu_co_rwlock_unlock(&job->flush_rwlock);
	186
	187	return ret;
	188	}
	189
	190	static int coroutine_fn backup_before_write_notify(
	191	NotifierWithReturn *notifier,
	192	void *opaque)
	193	{
	194	BdrvTrackedRequest *req = opaque;
793ed47a KW	195	int64_t sector_num = req->offset >> BDRV_SECTOR_BITS;
793ed47a KW	196	int nb_sectors = req->bytes >> BDRV_SECTOR_BITS;
98d2c6f2	197
793ed47a KW	198	assert((req->offset & (BDRV_SECTOR_SIZE - 1)) == 0);
	199	assert((req->bytes & (BDRV_SECTOR_SIZE - 1)) == 0);
	200
06c3916b	201	return backup_do_cow(req->bs, sector_num, nb_sectors, NULL, true);
98d2c6f2 DM	202	}
	203
	204	static void backup_set_speed(BlockJob job, int64_t speed, Error *errp)
	205	{
	206	BackupBlockJob *s = container_of(job, BackupBlockJob, common);
	207
	208	if (speed < 0) {
c6bd8c70	209	error_setg(errp, QERR_INVALID_PARAMETER, "speed");
98d2c6f2 DM	210	return;
	211	}
	212	ratelimit_set_speed(&s->limit, speed / BDRV_SECTOR_SIZE, SLICE_TIME);
	213	}
	214
	215	static void backup_iostatus_reset(BlockJob *job)
	216	{
	217	BackupBlockJob *s = container_of(job, BackupBlockJob, common);
	218
373340b2 HR	219	if (s->target->blk) {
	220	blk_iostatus_reset(s->target->blk);
	221	}
98d2c6f2 DM	222	}
98d2c6f2 DM	223
b976ea3c FZ	224	static void backup_cleanup_sync_bitmap(BackupBlockJob *job, int ret)
	225	{
	226	BdrvDirtyBitmap *bm;
	227	BlockDriverState *bs = job->common.bs;
	228
	229	if (ret < 0 \|\| block_job_is_cancelled(&job->common)) {
	230	/* Merge the successor back into the parent, delete nothing. */
	231	bm = bdrv_reclaim_dirty_bitmap(bs, job->sync_bitmap, NULL);
	232	assert(bm);
	233	} else {
	234	/* Everything is fine, delete this bitmap and install the backup. */
	235	bm = bdrv_dirty_bitmap_abdicate(bs, job->sync_bitmap, NULL);
	236	assert(bm);
	237	}
	238	}
	239
c347b2c6 JS	240	static void backup_commit(BlockJob *job)
	241	{
	242	BackupBlockJob *s = container_of(job, BackupBlockJob, common);
	243	if (s->sync_bitmap) {
	244	backup_cleanup_sync_bitmap(s, 0);
	245	}
	246	}
	247
	248	static void backup_abort(BlockJob *job)
	249	{
	250	BackupBlockJob *s = container_of(job, BackupBlockJob, common);
	251	if (s->sync_bitmap) {
	252	backup_cleanup_sync_bitmap(s, -1);
	253	}
	254	}
	255
3fc4b10a	256	static const BlockJobDriver backup_job_driver = {
98d2c6f2	257	.instance_size = sizeof(BackupBlockJob),
79e14bf7	258	.job_type = BLOCK_JOB_TYPE_BACKUP,
98d2c6f2 DM	259	.set_speed = backup_set_speed,
98d2c6f2 DM	260	.iostatus_reset = backup_iostatus_reset,
c347b2c6 JS	261	.commit = backup_commit,
c347b2c6 JS	262	.abort = backup_abort,
98d2c6f2 DM	263	};
	264
	265	static BlockErrorAction backup_error_action(BackupBlockJob *job,
	266	bool read, int error)
	267	{
	268	if (read) {
	269	return block_job_error_action(&job->common, job->common.bs,
	270	job->on_source_error, true, error);
	271	} else {
	272	return block_job_error_action(&job->common, job->target,
	273	job->on_target_error, false, error);
	274	}
	275	}
	276
761731b1 SH	277	typedef struct {
	278	int ret;
	279	} BackupCompleteData;
	280
	281	static void backup_complete(BlockJob job, void opaque)
	282	{
	283	BackupBlockJob *s = container_of(job, BackupBlockJob, common);
	284	BackupCompleteData *data = opaque;
	285
	286	bdrv_unref(s->target);
	287
	288	block_job_completed(job, data->ret);
	289	g_free(data);
	290	}
	291
d58d8453 JS	292	static bool coroutine_fn yield_and_check(BackupBlockJob *job)
	293	{
	294	if (block_job_is_cancelled(&job->common)) {
	295	return true;
	296	}
	297
	298	/* we need to yield so that bdrv_drain_all() returns.
	299	* (without, VM does not reboot)
	300	*/
	301	if (job->common.speed) {
	302	uint64_t delay_ns = ratelimit_calculate_delay(&job->limit,
	303	job->sectors_read);
	304	job->sectors_read = 0;
	305	block_job_sleep_ns(&job->common, QEMU_CLOCK_REALTIME, delay_ns);
	306	} else {
	307	block_job_sleep_ns(&job->common, QEMU_CLOCK_REALTIME, 0);
	308	}
	309
	310	if (block_job_is_cancelled(&job->common)) {
	311	return true;
	312	}
	313
	314	return false;
	315	}
	316
	317	static int coroutine_fn backup_run_incremental(BackupBlockJob *job)
	318	{
	319	bool error_is_read;
	320	int ret = 0;
	321	int clusters_per_iter;
	322	uint32_t granularity;
	323	int64_t sector;
	324	int64_t cluster;
	325	int64_t end;
	326	int64_t last_cluster = -1;
	327	BlockDriverState *bs = job->common.bs;
	328	HBitmapIter hbi;
	329
	330	granularity = bdrv_dirty_bitmap_granularity(job->sync_bitmap);
	331	clusters_per_iter = MAX((granularity / BACKUP_CLUSTER_SIZE), 1);
20dca810	332	bdrv_dirty_iter_init(job->sync_bitmap, &hbi);
d58d8453 JS	333
	334	/* Find the next dirty sector(s) */
	335	while ((sector = hbitmap_iter_next(&hbi)) != -1) {
	336	cluster = sector / BACKUP_SECTORS_PER_CLUSTER;
	337
	338	/* Fake progress updates for any clusters we skipped */
	339	if (cluster != last_cluster + 1) {
	340	job->common.offset += ((cluster - last_cluster - 1) *
	341	BACKUP_CLUSTER_SIZE);
	342	}
	343
	344	for (end = cluster + clusters_per_iter; cluster < end; cluster++) {
	345	do {
	346	if (yield_and_check(job)) {
	347	return ret;
	348	}
	349	ret = backup_do_cow(bs, cluster * BACKUP_SECTORS_PER_CLUSTER,
06c3916b WC	350	BACKUP_SECTORS_PER_CLUSTER, &error_is_read,
06c3916b WC	351	false);
d58d8453 JS	352	if ((ret < 0) &&
	353	backup_error_action(job, error_is_read, -ret) ==
	354	BLOCK_ERROR_ACTION_REPORT) {
	355	return ret;
	356	}
	357	} while (ret < 0);
	358	}
	359
	360	/* If the bitmap granularity is smaller than the backup granularity,
	361	* we need to advance the iterator pointer to the next cluster. */
	362	if (granularity < BACKUP_CLUSTER_SIZE) {
	363	bdrv_set_dirty_iter(&hbi, cluster * BACKUP_SECTORS_PER_CLUSTER);
	364	}
	365
	366	last_cluster = cluster - 1;
	367	}
	368
	369	/* Play some final catchup with the progress meter */
	370	end = DIV_ROUND_UP(job->common.len, BACKUP_CLUSTER_SIZE);
	371	if (last_cluster + 1 < end) {
	372	job->common.offset += ((end - last_cluster - 1) * BACKUP_CLUSTER_SIZE);
	373	}
	374
	375	return ret;
	376	}
	377
98d2c6f2 DM	378	static void coroutine_fn backup_run(void *opaque)
	379	{
	380	BackupBlockJob *job = opaque;
761731b1	381	BackupCompleteData *data;
98d2c6f2 DM	382	BlockDriverState *bs = job->common.bs;
	383	BlockDriverState *target = job->target;
	384	BlockdevOnError on_target_error = job->on_target_error;
	385	NotifierWithReturn before_write = {
	386	.notify = backup_before_write_notify,
	387	};
	388	int64_t start, end;
	389	int ret = 0;
	390
	391	QLIST_INIT(&job->inflight_reqs);
	392	qemu_co_rwlock_init(&job->flush_rwlock);
	393
	394	start = 0;
d58d8453	395	end = DIV_ROUND_UP(job->common.len, BACKUP_CLUSTER_SIZE);
98d2c6f2 DM	396
	397	job->bitmap = hbitmap_alloc(end, 0);
	398
	399	bdrv_set_enable_write_cache(target, true);
373340b2 HR	400	if (target->blk) {
	401	blk_set_on_error(target->blk, on_target_error, on_target_error);
	402	blk_iostatus_enable(target->blk);
	403	}
98d2c6f2 DM	404
	405	bdrv_add_before_write_notifier(bs, &before_write);
	406
fc5d3f84 IM	407	if (job->sync_mode == MIRROR_SYNC_MODE_NONE) {
	408	while (!block_job_is_cancelled(&job->common)) {
	409	/* Yield until the job is cancelled. We just let our before_write
	410	* notify callback service CoW requests. */
	411	job->common.busy = false;
	412	qemu_coroutine_yield();
	413	job->common.busy = true;
98d2c6f2	414	}
4b80ab2b	415	} else if (job->sync_mode == MIRROR_SYNC_MODE_INCREMENTAL) {
d58d8453	416	ret = backup_run_incremental(job);
fc5d3f84 IM	417	} else {
	418	/* Both FULL and TOP SYNC_MODE's require copying.. */
	419	for (; start < end; start++) {
	420	bool error_is_read;
d58d8453	421	if (yield_and_check(job)) {
98d2c6f2	422	break;
fc5d3f84 IM	423	}
	424
	425	if (job->sync_mode == MIRROR_SYNC_MODE_TOP) {
	426	int i, n;
	427	int alloced = 0;
	428
	429	/* Check to see if these blocks are already in the
	430	* backing file. */
	431
	432	for (i = 0; i < BACKUP_SECTORS_PER_CLUSTER;) {
bdad13b9	433	/* bdrv_is_allocated() only returns true/false based
4c293dc6	434	* on the first set of sectors it comes across that
fc5d3f84 IM	435	* are are all in the same state.
	436	* For that reason we must verify each sector in the
	437	* backup cluster length. We end up copying more than
	438	* needed but at some point that is always the case. */
	439	alloced =
bdad13b9	440	bdrv_is_allocated(bs,
fc5d3f84 IM	441	start * BACKUP_SECTORS_PER_CLUSTER + i,
	442	BACKUP_SECTORS_PER_CLUSTER - i, &n);
	443	i += n;
	444
d40593dd	445	if (alloced == 1 \|\| n == 0) {
fc5d3f84 IM	446	break;
	447	}
	448	}
	449
	450	/* If the above loop never found any sectors that are in
	451	* the topmost image, skip this backup. */
	452	if (alloced == 0) {
	453	continue;
	454	}
	455	}
	456	/* FULL sync mode we copy the whole drive. */
	457	ret = backup_do_cow(bs, start * BACKUP_SECTORS_PER_CLUSTER,
06c3916b	458	BACKUP_SECTORS_PER_CLUSTER, &error_is_read, false);
fc5d3f84 IM	459	if (ret < 0) {
	460	/* Depending on error action, fail now or retry cluster */
	461	BlockErrorAction action =
	462	backup_error_action(job, error_is_read, -ret);
a589569f	463	if (action == BLOCK_ERROR_ACTION_REPORT) {
fc5d3f84 IM	464	break;
	465	} else {
	466	start--;
	467	continue;
	468	}
98d2c6f2 DM	469	}
	470	}
	471	}
	472
	473	notifier_with_return_remove(&before_write);
	474
	475	/* wait until pending backup_do_cow() calls have completed */
	476	qemu_co_rwlock_wrlock(&job->flush_rwlock);
	477	qemu_co_rwlock_unlock(&job->flush_rwlock);
98d2c6f2 DM	478	hbitmap_free(job->bitmap);
98d2c6f2 DM	479
373340b2 HR	480	if (target->blk) {
	481	blk_iostatus_disable(target->blk);
	482	}
c29c1dd3	483	bdrv_op_unblock_all(target, job->common.blocker);
98d2c6f2	484
761731b1 SH	485	data = g_malloc(sizeof(*data));
	486	data->ret = ret;
	487	block_job_defer_to_main_loop(&job->common, backup_complete, data);
98d2c6f2 DM	488	}
	489
	490	void backup_start(BlockDriverState bs, BlockDriverState target,
fc5d3f84	491	int64_t speed, MirrorSyncMode sync_mode,
d58d8453	492	BdrvDirtyBitmap *sync_bitmap,
98d2c6f2 DM	493	BlockdevOnError on_source_error,
98d2c6f2 DM	494	BlockdevOnError on_target_error,
097310b5	495	BlockCompletionFunc cb, void opaque,
98d2c6f2 DM	496	Error **errp)
	497	{
	498	int64_t len;
	499
	500	assert(bs);
	501	assert(target);
	502	assert(cb);
	503
c29c1dd3 FZ	504	if (bs == target) {
	505	error_setg(errp, "Source and target cannot be the same");
	506	return;
	507	}
	508
98d2c6f2 DM	509	if ((on_source_error == BLOCKDEV_ON_ERROR_STOP \|\|
98d2c6f2 DM	510	on_source_error == BLOCKDEV_ON_ERROR_ENOSPC) &&
373340b2	511	(!bs->blk \|\| !blk_iostatus_is_enabled(bs->blk))) {
c6bd8c70	512	error_setg(errp, QERR_INVALID_PARAMETER, "on-source-error");
98d2c6f2 DM	513	return;
	514	}
	515
c29c1dd3 FZ	516	if (!bdrv_is_inserted(bs)) {
	517	error_setg(errp, "Device is not inserted: %s",
	518	bdrv_get_device_name(bs));
	519	return;
	520	}
	521
	522	if (!bdrv_is_inserted(target)) {
	523	error_setg(errp, "Device is not inserted: %s",
	524	bdrv_get_device_name(target));
	525	return;
	526	}
	527
	528	if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
	529	return;
	530	}
	531
	532	if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_BACKUP_TARGET, errp)) {
	533	return;
	534	}
	535
4b80ab2b	536	if (sync_mode == MIRROR_SYNC_MODE_INCREMENTAL) {
d58d8453 JS	537	if (!sync_bitmap) {
d58d8453 JS	538	error_setg(errp, "must provide a valid bitmap name for "
4b80ab2b	539	"\"incremental\" sync mode");
d58d8453 JS	540	return;
	541	}
	542
	543	/* Create a new bitmap, and freeze/disable this one. */
	544	if (bdrv_dirty_bitmap_create_successor(bs, sync_bitmap, errp) < 0) {
	545	return;
	546	}
	547	} else if (sync_bitmap) {
	548	error_setg(errp,
	549	"a sync_bitmap was provided to backup_run, "
	550	"but received an incompatible sync_mode (%s)",
	551	MirrorSyncMode_lookup[sync_mode]);
	552	return;
	553	}
	554
98d2c6f2 DM	555	len = bdrv_getlength(bs);
	556	if (len < 0) {
	557	error_setg_errno(errp, -len, "unable to get length for '%s'",
	558	bdrv_get_device_name(bs));
d58d8453	559	goto error;
98d2c6f2 DM	560	}
98d2c6f2 DM	561
3fc4b10a	562	BackupBlockJob *job = block_job_create(&backup_job_driver, bs, speed,
98d2c6f2 DM	563	cb, opaque, errp);
98d2c6f2 DM	564	if (!job) {
d58d8453	565	goto error;
98d2c6f2 DM	566	}
98d2c6f2 DM	567
c29c1dd3 FZ	568	bdrv_op_block_all(target, job->common.blocker);
c29c1dd3 FZ	569
98d2c6f2 DM	570	job->on_source_error = on_source_error;
	571	job->on_target_error = on_target_error;
	572	job->target = target;
fc5d3f84	573	job->sync_mode = sync_mode;
4b80ab2b	574	job->sync_bitmap = sync_mode == MIRROR_SYNC_MODE_INCREMENTAL ?
d58d8453	575	sync_bitmap : NULL;
98d2c6f2 DM	576	job->common.len = len;
	577	job->common.co = qemu_coroutine_create(backup_run);
	578	qemu_coroutine_enter(job->common.co, job);
d58d8453 JS	579	return;
	580
	581	error:
	582	if (sync_bitmap) {
	583	bdrv_reclaim_dirty_bitmap(bs, sync_bitmap, NULL);
	584	}
98d2c6f2	585	}