diff mbox

[6/7] QEMU live block copy

Message ID 20110523213411.003695437@amt.cnet
State New
Headers show

Commit Message

Marcelo Tosatti May 23, 2011, 9:31 p.m. UTC
Support live image copy + switch. That is, copy an image backing
a guest hard disk to a destination image (destination image must
be created separately), and switch to this copy.

Command syntax:

block_copy device filename [-i] -- live block copy device to image
             -i for incremental copy (base image shared between src and destination)

Please refer to qmp-commands diff for more details.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Comments

Blue Swirl May 24, 2011, 7:15 p.m. UTC | #1
On Tue, May 24, 2011 at 12:31 AM, Marcelo Tosatti <mtosatti@redhat.com> wrote:
> Support live image copy + switch. That is, copy an image backing
> a guest hard disk to a destination image (destination image must
> be created separately), and switch to this copy.
>
> Command syntax:
>
> block_copy device filename [-i] -- live block copy device to image
>             -i for incremental copy (base image shared between src and destination)
>
> Please refer to qmp-commands diff for more details.
>
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
>
> Index: qemu-block-copy/block-copy.c
> ===================================================================
> --- /dev/null
> +++ qemu-block-copy/block-copy.c
> @@ -0,0 +1,754 @@
> +/*
> + * QEMU live block copy
> + *
> + * Copyright (C) 2010 Red Hat Inc.
> + *
> + * Authors: Marcelo Tosatti <mtosatti@redhat.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2.  See
> + * the COPYING file in the top-level directory.
> + *
> + */
> +
> +#include "qemu-common.h"
> +#include "block_int.h"
> +#include "blockdev.h"
> +#include "qemu-queue.h"
> +#include "qemu-timer.h"
> +#include "monitor.h"
> +#include "block-copy.h"
> +#include "migration.h"
> +#include "sysemu.h"
> +#include "qjson.h"
> +#include <assert.h>
> +
> +#define BLOCK_SIZE (BDRV_SECTORS_PER_DIRTY_CHUNK << BDRV_SECTOR_BITS)
> +#define MAX_IS_ALLOCATED_SEARCH 65536
> +
> +/*
> + * Stages:
> + *
> + * STAGE_BULK: bulk reads/writes in progress
> + * STAGE_BULK_FINISHED: bulk reads finished, bulk writes in progress
> + * STAGE_DIRTY: bulk writes finished, dirty reads/writes in progress
> + * STAGE_MIRROR_WRITES: copy finished, writes mirrored to both images.
> + * STAGE_SWITCH_FINISHED: switched to new image.
> + */
> +
> +enum BdrvCopyStage {
> +    STAGE_BULK,
> +    STAGE_BULK_FINISHED,
> +    STAGE_DIRTY,
> +    STAGE_MIRROR_WRITES,
> +    STAGE_SWITCH_FINISHED,
> +};
> +
> +typedef struct BdrvCopyState {
> +    BlockDriverState *src;
> +    BlockDriverState *dst;
> +    bool shared_base;
> +
> +    int64_t curr_sector;
> +    int64_t completed_sectors;
> +    int64_t nr_sectors;
> +
> +    enum BdrvCopyStage stage;
> +    int inflight_reads;
> +    int error;
> +    int failed;
> +    int cancelled;
> +    QLIST_HEAD(, BdrvCopyBlock) io_list;
> +    unsigned long *aio_bitmap;
> +    QEMUTimer *aio_timer;
> +    QLIST_ENTRY(BdrvCopyState) list;
> +
> +    int64_t blocks;
> +    int64_t total_time;
> +
> +    char src_device_name[32];
> +    char src_filename[1024];

A #define for the buffer size would be nice.

> +} BdrvCopyState;
> +
> +typedef struct BdrvCopyBlock {
> +    BdrvCopyState *state;
> +    uint8_t *buf;
> +    int64_t sector;
> +    int64_t nr_sectors;
> +    struct iovec iov;
> +    QEMUIOVector qiov;
> +    BlockDriverAIOCB *aiocb;
> +    int64_t time;
> +    QLIST_ENTRY(BdrvCopyBlock) list;
> +} BdrvCopyBlock;
> +
> +static QLIST_HEAD(, BdrvCopyState) block_copy_list =
> +    QLIST_HEAD_INITIALIZER(block_copy_list);
> +
> +static void alloc_aio_bitmap(BdrvCopyState *s)
> +{
> +    BlockDriverState *bs = s->src;
> +    int64_t bitmap_size;
> +
> +    bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
> +            BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
> +    bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
> +
> +    s->aio_bitmap = qemu_mallocz(bitmap_size);
> +}
> +
> +static bool aio_inflight(BdrvCopyState *s, int64_t sector)
> +{
> +    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
> +
> +    if (s->aio_bitmap &&
> +        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(s->src)) {
> +        return !!(s->aio_bitmap[chunk / (sizeof(unsigned long) * 8)] &
> +            (1UL << (chunk % (sizeof(unsigned long) * 8))));

Please use the bitmap functions in bitmap.h, also in the next function.

> +    } else {
> +        return 0;
> +    }
> +}
> +
> +static void set_aio_inflight(BdrvCopyState *s, int64_t sector_num,
> +                             int nb_sectors, int set)
> +{
> +    int64_t start, end;
> +    unsigned long val, idx, bit;
> +
> +    start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
> +    end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
> +
> +    for (; start <= end; start++) {
> +        idx = start / (sizeof(unsigned long) * 8);
> +        bit = start % (sizeof(unsigned long) * 8);
> +        val = s->aio_bitmap[idx];
> +        if (set) {
> +            if (!(val & (1UL << bit))) {
> +                val |= 1UL << bit;
> +            }
> +        } else {
> +            if (val & (1UL << bit)) {
> +                val &= ~(1UL << bit);
> +            }
> +        }
> +        s->aio_bitmap[idx] = val;
> +    }
> +}
> +
> +static void blkcopy_set_stage(BdrvCopyState *s, enum BdrvCopyStage stage)
> +{
> +    s->stage = stage;
> +
> +    switch (stage) {
> +    case STAGE_BULK:
> +        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_BULK);
> +        break;
> +    case STAGE_BULK_FINISHED:
> +        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_BULK_FINISHED);
> +        break;
> +    case STAGE_DIRTY:
> +        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_DIRTY);
> +        break;
> +    case STAGE_MIRROR_WRITES:
> +        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_MIRROR_WRITES);
> +        break;
> +    case STAGE_SWITCH_FINISHED:
> +        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_SWITCH_FINISHED);
> +        break;
> +    default:
> +        break;
> +    }
> +}
> +
> +static void blk_copy_handle_cb_error(BdrvCopyState *s, int ret)
> +{
> +    s->error = ret;
> +    qemu_mod_timer(s->aio_timer, qemu_get_clock_ms(rt_clock));
> +}
> +
> +static inline void add_avg_transfer_time(BdrvCopyState *s, int64_t time)
> +{
> +    s->blocks++;
> +    s->total_time += time;
> +}
> +
> +static void blk_copy_write_cb(void *opaque, int ret)
> +{
> +    BdrvCopyBlock *blk = opaque;
> +    BdrvCopyState *s = blk->state;
> +
> +    if (ret < 0) {
> +        QLIST_REMOVE(blk, list);
> +        qemu_free(blk->buf);
> +        qemu_free(blk);
> +        blk_copy_handle_cb_error(s, ret);
> +        return;
> +    }
> +
> +    QLIST_REMOVE(blk, list);
> +    add_avg_transfer_time(s, qemu_get_clock_ns(rt_clock) - blk->time);
> +
> +    /* schedule switch to STAGE_DIRTY on last bulk write completion */
> +    if (blk->state->stage == STAGE_BULK_FINISHED) {
> +        qemu_mod_timer(s->aio_timer, qemu_get_clock_ms(rt_clock));
> +    }
> +
> +    if (blk->state->stage > STAGE_BULK_FINISHED) {
> +        set_aio_inflight(blk->state, blk->sector, blk->nr_sectors, 0);
> +    }
> +
> +    qemu_free(blk->buf);
> +    qemu_free(blk);
> +}
> +
> +static void blk_copy_issue_write(BdrvCopyState *s, BdrvCopyBlock *read_blk)
> +{
> +    BdrvCopyBlock *blk = qemu_mallocz(sizeof(BdrvCopyBlock));
> +    blk->state = s;
> +    blk->sector = read_blk->sector;
> +    blk->nr_sectors = read_blk->nr_sectors;
> +    blk->time = read_blk->time;
> +    blk->buf = read_blk->buf;
> +    QLIST_INSERT_HEAD(&s->io_list, blk, list);
> +
> +    blk->iov.iov_base = read_blk->buf;
> +    blk->iov.iov_len = read_blk->iov.iov_len;
> +    qemu_iovec_init_external(&blk->qiov, &blk->iov, 1);
> +
> +    BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_AIO_WRITE);
> +    blk->aiocb = bdrv_aio_writev(s->dst, blk->sector, &blk->qiov,
> +                                 blk->iov.iov_len / BDRV_SECTOR_SIZE,
> +                                 blk_copy_write_cb, blk);
> +    if (!blk->aiocb) {
> +        s->error = 1;
> +        goto error;
> +    }
> +
> +    return;
> +
> +error:
> +    QLIST_REMOVE(blk, list);
> +    qemu_free(read_blk->buf);
> +    qemu_free(blk);
> +}
> +
> +static void blk_copy_read_cb(void *opaque, int ret)
> +{
> +    BdrvCopyBlock *blk = opaque;
> +    BdrvCopyState *s = blk->state;
> +
> +    s->inflight_reads--;
> +    if (ret < 0) {
> +        QLIST_REMOVE(blk, list);
> +        qemu_free(blk->buf);
> +        qemu_free(blk);
> +        blk_copy_handle_cb_error(s, ret);
> +        return;
> +    }
> +    blk_copy_issue_write(s, blk);
> +    QLIST_REMOVE(blk, list);
> +    qemu_free(blk);
> +    qemu_mod_timer(s->aio_timer, qemu_get_clock_ms(rt_clock));
> +}
> +
> +static void blk_copy_issue_read(BdrvCopyState *s, int64_t sector,
> +                                int nr_sectors)
> +{
> +    BdrvCopyBlock *blk = qemu_mallocz(sizeof(BdrvCopyBlock));
> +    blk->buf = qemu_mallocz(BLOCK_SIZE);
> +    blk->state = s;
> +    blk->sector = sector;
> +    blk->nr_sectors = nr_sectors;
> +    QLIST_INSERT_HEAD(&s->io_list, blk, list);
> +
> +    blk->iov.iov_base = blk->buf;
> +    blk->iov.iov_len = nr_sectors * BDRV_SECTOR_SIZE;
> +    qemu_iovec_init_external(&blk->qiov, &blk->iov, 1);
> +
> +    s->inflight_reads++;
> +    blk->time = qemu_get_clock_ns(rt_clock);
> +    blk->aiocb = bdrv_aio_readv(s->src, sector, &blk->qiov, nr_sectors,
> +                                blk_copy_read_cb, blk);
> +    if (!blk->aiocb) {
> +        s->error = 1;
> +        goto error;
> +    }
> +
> +    return;
> +
> +error:
> +    s->inflight_reads--;
> +    QLIST_REMOVE(blk, list);
> +    qemu_free(blk->buf);
> +    qemu_free(blk);
> +}
> +
> +static bool blkcopy_can_switch(BdrvCopyState *s)
> +{
> +    int64_t remaining_dirty;
> +    int64_t avg_transfer_time;
> +
> +    remaining_dirty = bdrv_get_dirty_count(s->src);
> +    if (remaining_dirty == 0 || s->blocks == 0) {
> +        return true;
> +    }
> +
> +    avg_transfer_time = s->total_time / s->blocks;
> +    if ((remaining_dirty * avg_transfer_time) <= migrate_max_downtime()) {
> +        return true;
> +    }
> +    return false;
> +}
> +
> +static int blk_issue_reads_dirty(BdrvCopyState *s)
> +{
> +    int64_t sector;
> +
> +    for (sector = s->curr_sector; sector < s->nr_sectors;) {
> +        if (bdrv_get_dirty(s->src, sector) && !aio_inflight(s, sector)) {
> +            int nr_sectors = MIN(s->nr_sectors - s->curr_sector,
> +                                 BDRV_SECTORS_PER_DIRTY_CHUNK);
> +
> +            blk_copy_issue_read(s, sector, nr_sectors);
> +            bdrv_reset_dirty(s->src, sector, nr_sectors);
> +            set_aio_inflight(s, sector, nr_sectors, 1);
> +            break;
> +        }
> +
> +        sector += BDRV_SECTORS_PER_DIRTY_CHUNK;
> +        s->curr_sector = sector;
> +    }
> +
> +    if (sector >= s->nr_sectors) {
> +        s->curr_sector = 0;
> +    }
> +    return 0;
> +}
> +
> +static int blk_issue_reads_bulk(BdrvCopyState *s)
> +{
> +    int nr_sectors;
> +    int64_t curr_sector = s->curr_sector;
> +
> +    if (s->shared_base) {
> +        while (curr_sector < s->nr_sectors &&
> +                !bdrv_is_allocated(s->src, curr_sector,
> +                                   MAX_IS_ALLOCATED_SEARCH, &nr_sectors)) {
> +                curr_sector += nr_sectors;
> +        }
> +    }
> +
> +    if (curr_sector >= s->nr_sectors) {
> +        s->curr_sector = 0;
> +        return 1;
> +    }
> +
> +    curr_sector &= ~((int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK - 1);
> +    nr_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;
> +
> +    blk_copy_issue_read(s, s->curr_sector, nr_sectors);
> +    s->curr_sector += nr_sectors;
> +    s->completed_sectors = curr_sector;
> +    return 0;
> +}
> +
> +static void blkcopy_finish(BdrvCopyState *s)
> +{
> +    int64_t sector;
> +    uint8_t *buf;
> +
> +    buf = qemu_malloc(BLOCK_SIZE);
> +
> +    /* FIXME: speed up loop, get_next_dirty_block? */
> +    for (sector = 0; sector < s->nr_sectors;
> +         sector += BDRV_SECTORS_PER_DIRTY_CHUNK) {
> +        if (bdrv_get_dirty(s->src, sector)) {
> +            int nr_sectors = MIN(s->nr_sectors - sector,
> +                                 BDRV_SECTORS_PER_DIRTY_CHUNK);
> +
> +            memset(buf, 0, BLOCK_SIZE);
> +            if (bdrv_read(s->src, sector, buf, nr_sectors) < 0) {
> +                goto error;
> +            }
> +            if (bdrv_write(s->dst, sector, buf, nr_sectors) < 0) {
> +                goto error;
> +            }
> +            bdrv_reset_dirty(s->src, sector, nr_sectors);
> +        }
> +
> +        if (bdrv_get_dirty_count(s->src) == 0)
> +            break;

Braces, please use checkpatch.pl.

> +    }
> +    qemu_free(buf);
> +    return;
> +
> +error:
> +    qemu_free(buf);
> +    s->error = 1;
> +}
> +
> +static void blkcopy_cleanup(BdrvCopyState *s)
> +{
> +    assert(s->inflight_reads == 0);
> +    assert(QLIST_EMPTY(&s->io_list));
> +    bdrv_set_dirty_tracking(s->src, 0);
> +    drive_put_ref(drive_get_by_blockdev(s->src));
> +    bdrv_set_in_use(s->src, 0);
> +    if (s->stage >= STAGE_DIRTY)
> +        qemu_free(s->aio_bitmap);
> +    qemu_del_timer(s->aio_timer);
> +}
> +
> +static void blkcopy_free(BdrvCopyState *s)
> +{
> +    QLIST_REMOVE(s, list);
> +    qemu_free(s);
> +}
> +
> +static void handle_error(BdrvCopyState *s)
> +{
> +    if (!QLIST_EMPTY(&s->io_list))
> +        return;
> +    s->failed = 1;
> +    blkcopy_cleanup(s);
> +}
> +
> +static void blkcopy_switch(BdrvCopyState *s)
> +{
> +    char src_filename[1024], mirror_name[2048];

#defines instead of magic constants, please.

> +    int open_flags, ret;
> +
> +    strncpy(src_filename, s->src->filename, sizeof(src_filename));

Are the lengths checked somewhere?

> +    open_flags = s->src->open_flags;
> +
> +    assert(s->stage == STAGE_DIRTY);
> +
> +    vm_stop(VMSTOP_BLOCKCOPY);
> +    /* flush any guest writes, dirty bitmap uptodate after this.
> +     * copy AIO also finished.
> +     */
> +    qemu_aio_flush();
> +    assert(QLIST_EMPTY(&s->io_list));
> +    if (s->error) {
> +        handle_error(s);
> +        goto vm_start;
> +    }
> +    blkcopy_finish(s);
> +    if (s->error) {
> +        handle_error(s);
> +        goto vm_start;
> +    }
> +    assert(bdrv_get_dirty_count(s->src) == 0);
> +    /* turn dirty bitmap off */
> +    bdrv_set_dirty_tracking(s->src, 0);
> +    /* switch to double writes */
> +    bdrv_flush_all();
> +    bdrv_close(s->src);
> +    bdrv_close(s->dst);
> +
> +    snprintf(mirror_name, sizeof(mirror_name)-1,
> +             "blkmirror:%s:%s", s->dst->filename, s->src->filename);
> +
> +    ret = bdrv_open(s->src, mirror_name, s->src->open_flags, NULL);
> +    if (ret < 0) {
> +        error_report("%s: cannot open blkmirror device, err %d",
> +                      mirror_name, ret);
> +        s->failed = 1;
> +        goto err;
> +    }
> +
> +    blkcopy_set_stage(s, STAGE_MIRROR_WRITES);
> +    qemu_del_timer(s->aio_timer);
> +
> +vm_start:
> +    vm_start();
> +    return;
> +
> +err:
> +    if (bdrv_open(s->src, src_filename, open_flags, NULL) < 0) {
> +        error_report("%s: %s: cannot fallback to source image\n", __func__,
> +                     s->src_filename);
> +        abort();
> +    }
> +    blkcopy_cleanup(s);
> +    goto vm_start;
> +}
> +
> +#define BLKCOPY_INFLIGHT 2
> +
> +/*
> + * To simplify the implementation, the IO completion callbacks do not
> + * handle stage control or submit IO for further blocks. A timer is used
> + * for such purpose.
> + */
> +
> +static void aio_timer(void *opaque)
> +{
> +    BdrvCopyState *s = opaque;
> +
> +    assert(s->cancelled == 0);
> +    assert(s->stage < STAGE_MIRROR_WRITES);
> +
> +    if (s->error) {
> +        handle_error(s);
> +        return;
> +    }
> +
> +    while (s->stage == STAGE_BULK) {
> +        if (s->inflight_reads >= BLKCOPY_INFLIGHT) {
> +            break;
> +        }
> +        if (blk_issue_reads_bulk(s)) {
> +            blkcopy_set_stage(s, STAGE_BULK_FINISHED);
> +        }
> +    }
> +
> +    if (s->stage == STAGE_BULK_FINISHED) {
> +        if (QLIST_EMPTY(&s->io_list)) {
> +            blkcopy_set_stage(s, STAGE_DIRTY);
> +            alloc_aio_bitmap(s);
> +        }
> +    }
> +
> +    while (s->stage == STAGE_DIRTY) {
> +        if (s->inflight_reads >= BLKCOPY_INFLIGHT) {
> +            break;
> +        }
> +        blk_issue_reads_dirty(s);
> +        if (blkcopy_can_switch(s)) {
> +            BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_SWITCH_START);
> +            blkcopy_switch(s);
> +            return;
> +        }
> +    }
> +}
> +
> +
> +int do_bdrv_copy_switch(Monitor *mon, const QDict *qdict, QObject **ret_data)
> +{
> +    const char *device = qdict_get_str(qdict, "device");
> +    BdrvCopyState *s = NULL;
> +    int open_flags;
> +
> +    QLIST_FOREACH(s, &block_copy_list, list) {
> +        if (!strcmp(s->src_device_name, device)) {
> +            if (s->stage != STAGE_MIRROR_WRITES) {
> +                qerror_report(QERR_IN_PROGRESS, "block copy");
> +                return -1;
> +            }
> +            break;
> +        }
> +    }
> +
> +    if (!s) {
> +        qerror_report(QERR_DEVICE_NOT_FOUND, device);
> +        return -1;
> +    }
> +
> +    open_flags = s->src->open_flags;
> +
> +    /* switch from mirrored writes to destination only */
> +    bdrv_flush_all();
> +    bdrv_close(s->src);
> +    if (bdrv_open(s->src, s->dst->filename, s->src->open_flags, NULL) < 0) {
> +        s->failed = 1;
> +        goto err;
> +    }
> +
> +    blkcopy_set_stage(s, STAGE_SWITCH_FINISHED);
> +    blkcopy_cleanup(s);
> +    return 0;
> +
> +err:
> +    if (bdrv_open(s->src, s->src_filename, open_flags, NULL) < 0) {
> +        error_report("%s: %s: cannot fallback to source image\n", __func__,
> +                     s->src_filename);

Below qerror_report() is used.

> +        abort();
> +    }
> +    return -1;
> +}
> +
> +static int bdrv_copy(Monitor *mon, const char * device, BlockDriverState *src,
> +                     BlockDriverState *dst, bool shared_base)
> +{
> +    int64_t sectors;
> +    BdrvCopyState *blkcopy, *safe;
> +
> +    QLIST_FOREACH_SAFE(blkcopy, &block_copy_list, list, safe) {
> +        if (!strcmp(blkcopy->src_device_name, src->device_name)) {
> +            if (blkcopy->stage == STAGE_SWITCH_FINISHED || blkcopy->failed) {
> +                blkcopy_free(blkcopy);
> +            } else {
> +                qerror_report(QERR_IN_PROGRESS, "block copy");
> +                return -1;
> +            }
> +        }
> +    }
> +
> +    sectors = bdrv_getlength(src) >> BDRV_SECTOR_BITS;
> +    if (sectors != bdrv_getlength(dst) >> BDRV_SECTOR_BITS) {
> +        qerror_report(QERR_BLOCKCOPY_IMAGE_SIZE_DIFFERS);
> +        return -1;
> +    }
> +
> +    blkcopy = qemu_mallocz(sizeof(BdrvCopyState));
> +    blkcopy->src = src;
> +    blkcopy->dst = dst;
> +    blkcopy->curr_sector = 0;
> +    blkcopy->nr_sectors = sectors;
> +    blkcopy_set_stage(blkcopy, STAGE_BULK);
> +    blkcopy->aio_timer = qemu_new_timer_ms(rt_clock, aio_timer, blkcopy);
> +    blkcopy->shared_base = shared_base;
> +    strncpy(blkcopy->src_device_name, blkcopy->src->device_name,
> +            sizeof(blkcopy->src_device_name) - 1);
> +    strncpy(blkcopy->src_filename, blkcopy->src->filename,
> +            sizeof(blkcopy->src_filename) - 1);
> +
> +    bdrv_set_dirty_tracking(src, 1);
> +    qemu_mod_timer(blkcopy->aio_timer, qemu_get_clock_ms(rt_clock));
> +    drive_get_ref(drive_get_by_blockdev(src));
> +    bdrv_set_in_use(src, 1);
> +
> +    QLIST_INSERT_HEAD(&block_copy_list, blkcopy, list);
> +    return 0;
> +}
> +
> +int do_bdrv_copy(Monitor *mon, const QDict *qdict, QObject **ret_data)
> +{
> +    const char *device = qdict_get_str(qdict, "device");
> +    const char *filename = qdict_get_str(qdict, "filename");
> +    bool shared_base = qdict_get_try_bool(qdict, "incremental", 0);
> +    BlockDriverState *new_bs, *bs;
> +    int ret;
> +
> +    if (migration_active()) {
> +        qerror_report(QERR_IN_PROGRESS, "migration");
> +        return -1;
> +    }
> +
> +    bs = bdrv_find(device);
> +    if (!bs) {
> +        qerror_report(QERR_DEVICE_NOT_FOUND, device);
> +        return -1;
> +    }
> +
> +    new_bs = bdrv_new("");
> +    if (bdrv_open(new_bs, filename, bs->open_flags, NULL) < 0) {
> +        bdrv_delete(new_bs);
> +        qerror_report(QERR_OPEN_FILE_FAILED, filename);
> +        return -1;
> +    }
> +
> +    ret = bdrv_copy(mon, device, bs, new_bs, shared_base);
> +    return ret;
> +}
> +
> +int do_bdrv_copy_cancel(Monitor *mon, const QDict *qdict, QObject **ret_data)
> +{
> +    BdrvCopyState *blkcopy, *s = NULL;
> +    const char *device = qdict_get_str(qdict, "device");
> +
> +    QLIST_FOREACH(blkcopy, &block_copy_list, list) {
> +        if (!strcmp(blkcopy->src_device_name, device)) {
> +            s = blkcopy;
> +            break;
> +        }
> +    }
> +
> +    if (!s || s->stage == STAGE_SWITCH_FINISHED || s->failed) {
> +        qerror_report(QERR_DEVICE_NOT_FOUND, device);
> +        return -1;
> +    }
> +
> +    s->cancelled = 1;
> +    do {
> +        qemu_aio_flush();
> +    } while (!QLIST_EMPTY(&s->io_list));
> +    blkcopy_cleanup(s);
> +    blkcopy_free(s);
> +
> +    return 0;
> +}
> +
> +static void blockcopy_print_dict(QObject *obj, void *opaque)
> +{
> +    QDict *c_dict;
> +    Monitor *mon = opaque;
> +
> +    c_dict = qobject_to_qdict(obj);
> +
> +    monitor_printf(mon, "%s: status=%s ",
> +                        qdict_get_str(c_dict, "device"),
> +                        qdict_get_str(c_dict, "status"));
> +
> +    if (qdict_haskey(c_dict, "info")) {
> +        QDict *qdict = qobject_to_qdict(qdict_get(c_dict, "info"));
> +
> +        monitor_printf(mon, "percentage=%ld %%",
> +                       qdict_get_int(qdict, "percentage"));
> +    }
> +
> +    monitor_printf(mon, "\n");
> +}
> +
> +void do_info_blockcopy_print(Monitor *mon, const QObject *data)
> +{
> +    qlist_iter(qobject_to_qlist(data), blockcopy_print_dict, mon);
> +}
> +
> +void do_info_blockcopy(Monitor *mon, QObject **ret_data)
> +{
> +    QList *c_list;
> +    BdrvCopyState *s;
> +
> +    c_list = qlist_new();
> +
> +    QLIST_FOREACH(s, &block_copy_list, list) {
> +        QObject *c_obj;
> +        static const char *status[] = { "failed", "active", "mirrored", "completed" };
> +        int i;
> +
> +        if (s->failed) {
> +            i = 0;
> +        } else if (s->stage < STAGE_MIRROR_WRITES) {
> +            i = 1;
> +        } else if (s->stage < STAGE_SWITCH_FINISHED) {
> +            i = 2;
> +        } else {
> +            i = 3;
> +        }
> +
> +        c_obj = qobject_from_jsonf("{ 'device': %s, 'status': %s }",
> +                                    s->src_device_name, status[i]);
> +
> +        if (i == 1) {
> +            QDict *dict = qobject_to_qdict(c_obj);
> +            QObject *obj;
> +
> +            /* FIXME: add dirty stage progress? */
> +            obj = qobject_from_jsonf("{ 'percentage': %" PRId64 "}",
> +                                     s->completed_sectors * 100 / s->nr_sectors);
> +            qdict_put_obj(dict, "info", obj);
> +        }
> +        qlist_append_obj(c_list, c_obj);
> +    }
> +
> +    *ret_data = QOBJECT(c_list);
> +}
> +
> +bool block_copy_active(void)
> +{
> +    BdrvCopyState *s;
> +
> +    QLIST_FOREACH(s, &block_copy_list, list) {
> +        if (s->failed) {
> +            continue;
> +        }
> +        if (s->stage < STAGE_SWITCH_FINISHED) {
> +            return true;
> +        }
> +    }
> +
> +    return false;
> +}
> +
> Index: qemu-block-copy/block-copy.h
> ===================================================================
> --- /dev/null
> +++ qemu-block-copy/block-copy.h
> @@ -0,0 +1,26 @@
> +/*
> + * QEMU live block copy
> + *
> + * Copyright (C) 2010 Red Hat Inc.
> + *
> + * Authors: Marcelo Tosatti <mtosatti@redhat.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2.  See
> + * the COPYING file in the top-level directory.
> + *
> + */
> +
> +#ifndef BLOCK_COPY_H
> +#define BLOCK_COPY_H
> +
> +int do_bdrv_copy(Monitor *mon, const QDict *qdict, QObject **ret_data);
> +int do_bdrv_copy_cancel(Monitor *mon, const QDict *qdict, QObject **ret_data);
> +int do_bdrv_copy_switch(Monitor *mon, const QDict *qdict, QObject **ret_data);
> +
> +void do_info_blockcopy_print(Monitor *mon, const QObject *data);
> +void do_info_blockcopy(Monitor *mon, QObject **ret_data);
> +
> +bool block_copy_active(void);
> +
> +#endif /* BLOCK_COPY_H */
> +
> Index: qemu-block-copy/hmp-commands.hx
> ===================================================================
> --- qemu-block-copy.orig/hmp-commands.hx
> +++ qemu-block-copy/hmp-commands.hx
> @@ -806,6 +806,63 @@ Set maximum speed to @var{value} (in byt
>  ETEXI
>
>     {
> +        .name       = "block_copy",
> +        .args_type  = "device:s,filename:s,incremental:-i",
> +        .params     = "device filename [-i]",
> +        .help       = "live block copy device to image"
> +                      "\n\t\t\t -i for incremental copy "
> +                      "(base image shared between original and destination)",
> +        .user_print = monitor_user_noop,
> +        .mhandler.cmd_new = do_bdrv_copy,
> +    },
> +
> +STEXI
> +@item block_copy @var{device} @var{filename} [-i]
> +@findex block_copy
> +Live copy block device @var{device} to image @var{filename}.
> +        -i for incremental copy (base image is shared)
> +
> +Destination image @var{filename} must be created with qemu-img prior
> +to execution of this command, with image size equal to the original
> +image size.
> +
> +Incremental copy allows the destination image @var{filename} to share
> +a common base image with the original image. This option skips copying
> +blocks which are not allocated in the original image.
> +ETEXI
> +
> +    {
> +        .name       = "block_copy_cancel",
> +        .args_type  = "device:s",
> +        .params     = "device",
> +        .help       = "cancel live block copy",
> +        .user_print = monitor_user_noop,
> +        .mhandler.cmd_new = do_bdrv_copy_cancel,
> +    },
> +
> +STEXI
> +@item block_copy_cancel @var{device}
> +@findex block_copy_cancel
> +Cancel live block copy on @var{device}.
> +ETEXI
> +
> +    {
> +        .name       = "block_copy_switch",
> +        .args_type  = "device:s",
> +        .params     = "device",
> +        .help       = "finish live block copy",
> +        .user_print = monitor_user_noop,
> +        .mhandler.cmd_new = do_bdrv_copy_switch,
> +    },
> +
> +STEXI
> +@item block_copy_switch @var{device}
> +@findex block_copy_switch
> +Finish live block copy on @var{device} by switching
> +to destination image.
> +ETEXI
> +
> +    {
>         .name       = "migrate_set_downtime",
>         .args_type  = "value:T",
>         .params     = "value",
> @@ -1352,6 +1409,8 @@ show device tree
>  show qdev device model list
>  @item info roms
>  show roms
> +@item info block-copy
> +show block copy status
>  @end table
>  ETEXI
>
> Index: qemu-block-copy/monitor.c
> ===================================================================
> --- qemu-block-copy.orig/monitor.c
> +++ qemu-block-copy/monitor.c
> @@ -45,6 +45,7 @@
>  #include "balloon.h"
>  #include "qemu-timer.h"
>  #include "migration.h"
> +#include "block-copy.h"
>  #include "kvm.h"
>  #include "acl.h"
>  #include "qint.h"
> @@ -3101,6 +3102,14 @@ static const mon_cmd_t info_cmds[] = {
>     },
>  #endif
>     {
> +        .name       = "block-copy",
> +        .args_type  = "",
> +        .params     = "",
> +        .help       = "show block copy status",
> +        .user_print = do_info_blockcopy_print,
> +        .mhandler.info_new = do_info_blockcopy,
> +    },
> +    {
>         .name       = NULL,
>     },
>  };
> @@ -3242,6 +3251,14 @@ static const mon_cmd_t qmp_query_cmds[]
>         .mhandler.info_async = do_info_balloon,
>         .flags      = MONITOR_CMD_ASYNC,
>     },
> +    {
> +        .name       = "block-copy",
> +        .args_type  = "",
> +        .params     = "",
> +        .help       = "show block copy status",
> +        .user_print = do_info_blockcopy_print,
> +        .mhandler.info_new = do_info_blockcopy,
> +    },
>     { /* NULL */ },
>  };
>
> Index: qemu-block-copy/qmp-commands.hx
> ===================================================================
> --- qemu-block-copy.orig/qmp-commands.hx
> +++ qemu-block-copy/qmp-commands.hx
> @@ -581,6 +581,98 @@ Example:
>  EQMP
>
>     {
> +        .name       = "block_copy",
> +        .args_type  = "device:s,filename:s,inc:-i",
> +        .params     = "device filename [-i]",
> +        .help       = "live block copy device to image"
> +                      "\n\t\t\t -i for incremental copy "
> +                      "(base image shared between src and destination)",
> +        .user_print = monitor_user_noop,
> +        .mhandler.cmd_new = do_bdrv_copy,
> +    },
> +
> +SQMP
> +block-copy
> +-------
> +
> +Live block copy.
> +
> +Arguments:
> +
> +- "device": device name (json-string)
> +- "filename": target image filename (json-string)
> +- "incremental": incremental disk copy (json-bool, optional)
> +
> +Example:
> +
> +-> { "execute": "block_copy",
> +                            "arguments": { "device": "ide0-hd1",
> +                               "filename": "/mnt/new-disk.img",
> +                             } }
> +
> +<- { "return": {} }
> +
> +Notes:
> +
> +(1) The 'query-block-copy' command should be used to check block copy progress
> +    and final result (this information is provided by the 'status' member)
> +(2) Boolean argument "incremental" defaults to false
> +
> +EQMP
> +
> +    {
> +        .name       = "block_copy_cancel",
> +        .args_type  = "device:s",
> +        .params     = "device",
> +        .help       = "cancel live block copy",
> +        .user_print = monitor_user_noop,
> +        .mhandler.cmd_new = do_bdrv_copy_cancel,
> +    },
> +
> +SQMP
> +block_copy_cancel
> +--------------
> +
> +Cancel live block copy.
> +
> +Arguments:
> +
> +- device: device name (json-string)
> +
> +Example:
> +
> +-> { "execute": "block_copy_cancel", "arguments": { "device": "ide0-hd1" } }
> +<- { "return": {} }
> +
> +EQMP
> +
> +    {
> +        .name       = "block_copy_switch",
> +        .args_type  = "device:s",
> +        .params     = "device",
> +        .help       = "finish live block copy",
> +        .user_print = monitor_user_noop,
> +        .mhandler.cmd_new = do_bdrv_copy_switch,
> +    },
> +
> +SQMP
> +block_copy_switch
> +--------------
> +
> +Finish live block copy, switching device to destination image.
> +
> +Arguments:
> +
> +- device: device name (json-string)
> +
> +Example:
> +
> +-> { "execute": "block_copy_switch", "arguments": { "device": "ide0-hd1" } }
> +<- { "return": {} }
> +
> +EQMP
> +
> +    {
>         .name       = "netdev_add",
>         .args_type  = "netdev:O",
>         .params     = "[user|tap|socket],id=str[,prop=value][,...]",
> @@ -1744,6 +1836,44 @@ Examples:
>  EQMP
>
>  SQMP
> +query-block-copy
> +-------------
> +
> +Live block copy status.
> +
> +Each block copy instance information is stored in a json-object and the returned
> +value is a json-array of all instances.
> +
> +Each json-object contains the following:
> +
> +- "device": device name (json-string)
> +- "status": block copy status (json-string)
> +    - Possible values: "active", "failed", "mirrored", "completed"
> +- "info": A json-object with the statistics information, if status is "active":
> +    - "percentage": percentage completed (json-int)
> +
> +Example:
> +
> +Block copy for "ide1-hd0" active and block copy for "ide1-hd1" failed:
> +
> +-> { "execute": "query-block-copy" }
> +<- {
> +      "return":[
> +        {"device":"ide1-hd0",
> +            "status":"active",
> +            "info":{
> +               "percentage":23,
> +            }
> +        },
> +        {"device":"ide1-hd1",
> +         "status":"failed"
> +        }
> +      ]
> +   }
> +
> +EQMP
> +
> +SQMP
>  query-balloon
>  -------------
>
> Index: qemu-block-copy/Makefile.objs
> ===================================================================
> --- qemu-block-copy.orig/Makefile.objs
> +++ qemu-block-copy/Makefile.objs
> @@ -98,7 +98,7 @@ common-obj-y += buffered_file.o migratio
>  common-obj-y += qemu-char.o savevm.o #aio.o
>  common-obj-y += msmouse.o ps2.o
>  common-obj-y += qdev.o qdev-properties.o
> -common-obj-y += block-migration.o iohandler.o
> +common-obj-y += block-migration.o iohandler.o block-copy.o
>  common-obj-y += pflib.o
>  common-obj-y += bitmap.o bitops.o
>
>
>
>
>
Avi Kivity May 29, 2011, 8:54 a.m. UTC | #2
On 05/24/2011 12:31 AM, Marcelo Tosatti wrote:
> Support live image copy + switch. That is, copy an image backing
> a guest hard disk to a destination image (destination image must
> be created separately), and switch to this copy.
>
> Command syntax:
>
> block_copy device filename [-i] -- live block copy device to image
>               -i for incremental copy (base image shared between src and destination)
>
> Please refer to qmp-commands diff for more details.

IMO it would have been nicer to use the mirror driver for all copying; 
there would be no STAGE_DIRTY; simply a background process that copies 
over all blocks, taking care not to conflict with ongoing writes.  It 
would also remove the conflict with migration.

But that's an implementation detail and can be changed later.
Marcelo Tosatti May 31, 2011, 4:06 p.m. UTC | #3
On Sun, May 29, 2011 at 11:54:25AM +0300, Avi Kivity wrote:
> On 05/24/2011 12:31 AM, Marcelo Tosatti wrote:
> >Support live image copy + switch. That is, copy an image backing
> >a guest hard disk to a destination image (destination image must
> >be created separately), and switch to this copy.
> >
> >Command syntax:
> >
> >block_copy device filename [-i] -- live block copy device to image
> >              -i for incremental copy (base image shared between src and destination)
> >
> >Please refer to qmp-commands diff for more details.
> 
> IMO it would have been nicer to use the mirror driver for all
> copying; there would be no STAGE_DIRTY; simply a background process
> that copies over all blocks, taking care not to conflict with
> ongoing writes.  

Don't see the advantage of doing that. Disadvantages:

- Guest write performance is affected during copying (guest writes
  compete with stream of writes from copy).
- Complexity to handle copy write conflict (there is no need to handle
  that with current solution).
- Unability to have the mirror functionality in a separate driver.

> It would also remove the conflict with migration.

There is no fundamental problem with migration, its simply unsupported
now.

> But that's an implementation detail and can be changed later.
Avi Kivity May 31, 2011, 4:14 p.m. UTC | #4
On 05/31/2011 07:06 PM, Marcelo Tosatti wrote:
> On Sun, May 29, 2011 at 11:54:25AM +0300, Avi Kivity wrote:
> >  On 05/24/2011 12:31 AM, Marcelo Tosatti wrote:
> >  >Support live image copy + switch. That is, copy an image backing
> >  >a guest hard disk to a destination image (destination image must
> >  >be created separately), and switch to this copy.
> >  >
> >  >Command syntax:
> >  >
> >  >block_copy device filename [-i] -- live block copy device to image
> >  >               -i for incremental copy (base image shared between src and destination)
> >  >
> >  >Please refer to qmp-commands diff for more details.
> >
> >  IMO it would have been nicer to use the mirror driver for all
> >  copying; there would be no STAGE_DIRTY; simply a background process
> >  that copies over all blocks, taking care not to conflict with
> >  ongoing writes.
>
> Don't see the advantage of doing that.

No dirty logging
No conflict with migration
Simpler conceptually (IMO)

> Disadvantages:
>
> - Guest write performance is affected during copying (guest writes
>    compete with stream of writes from copy).

Competes anyway with your background task?

> - Complexity to handle copy write conflict (there is no need to handle
>    that with current solution).

If new write from source A overlaps an in-flight write from source B, 
hold it in a list off the B request, and re-issue it on B completion.

> - Unability to have the mirror functionality in a separate driver.

Why?

> >  It would also remove the conflict with migration.
>
> There is no fundamental problem with migration, its simply unsupported
> now.

Why?

> >  But that's an implementation detail and can be changed later.
>
Marcelo Tosatti May 31, 2011, 4:38 p.m. UTC | #5
On Tue, May 31, 2011 at 07:14:57PM +0300, Avi Kivity wrote:
> On 05/31/2011 07:06 PM, Marcelo Tosatti wrote:
> >On Sun, May 29, 2011 at 11:54:25AM +0300, Avi Kivity wrote:
> >>  On 05/24/2011 12:31 AM, Marcelo Tosatti wrote:
> >>  >Support live image copy + switch. That is, copy an image backing
> >>  >a guest hard disk to a destination image (destination image must
> >>  >be created separately), and switch to this copy.
> >>  >
> >>  >Command syntax:
> >>  >
> >>  >block_copy device filename [-i] -- live block copy device to image
> >>  >               -i for incremental copy (base image shared between src and destination)
> >>  >
> >>  >Please refer to qmp-commands diff for more details.
> >>
> >>  IMO it would have been nicer to use the mirror driver for all
> >>  copying; there would be no STAGE_DIRTY; simply a background process
> >>  that copies over all blocks, taking care not to conflict with
> >>  ongoing writes.
> >
> >Don't see the advantage of doing that.
> 
> No dirty logging
> No conflict with migration
> Simpler conceptually (IMO)
> 
> >Disadvantages:
> >
> >- Guest write performance is affected during copying (guest writes
> >   compete with stream of writes from copy).
> 
> Competes anyway with your background task?

No because guest writes are to the source and copy writes are to
destination (which are potentially different disks or set of disks).

> >- Complexity to handle copy write conflict (there is no need to handle
> >   that with current solution).
> 
> If new write from source A overlaps an in-flight write from source
> B, hold it in a list off the B request, and re-issue it on B
> completion.
> 
> >- Unability to have the mirror functionality in a separate driver.
> 
> Why?

Because you need to handle the interaction between guest writes and 
copy writes someplace which can be aware of both.

For example, in-flight copy writes must be invalidated in the guest
write path.

> >>  It would also remove the conflict with migration.
> >
> >There is no fundamental problem with migration, its simply unsupported
> >now.
> 
> Why?

Don't see the need for that, management can simply wait for livecopy to
finish or cancel livecopy and restart on destination after migration.

But yeah, it should be implemented sometime...
Avi Kivity May 31, 2011, 4:53 p.m. UTC | #6
On 05/31/2011 07:38 PM, Marcelo Tosatti wrote:
> >
> >  >Disadvantages:
> >  >
> >  >- Guest write performance is affected during copying (guest writes
> >  >    compete with stream of writes from copy).
> >
> >  Competes anyway with your background task?
>
> No because guest writes are to the source and copy writes are to
> destination (which are potentially different disks or set of disks).

Your copy task reads from the source disk and writes to the destination 
disk.  A guest write doesn't touch the destination disk, but schedules a 
later write due to dirty tracking.  In all, the total number of writes 
should be similar.  Background copy is a more efficient since multiple 
writes to the same location are coalesced, but that's not very common 
(except perhaps for transaction logs).

> >
> >  >- Unability to have the mirror functionality in a separate driver.
> >
> >  Why?
>
> Because you need to handle the interaction between guest writes and
> copy writes someplace which can be aware of both.
>
> For example, in-flight copy writes must be invalidated in the guest
> write path.

guest -> device -> mirror driver -> [ source disk, destination disk ]

> >  >>   It would also remove the conflict with migration.
> >  >
> >  >There is no fundamental problem with migration, its simply unsupported
> >  >now.
> >
> >  Why?
>
> Don't see the need for that, management can simply wait for livecopy to
> finish or cancel livecopy and restart on destination after migration.

They can do it, but I imagine it's pretty hard for them.

> But yeah, it should be implemented sometime...
>


We should make an effort not to introduce interdependencies which make 
management's life harder.
Marcelo Tosatti June 3, 2011, 3:59 p.m. UTC | #7
On Tue, May 24, 2011 at 10:15:09PM +0300, Blue Swirl wrote:
> > +static bool aio_inflight(BdrvCopyState *s, int64_t sector)
> > +{
> > +    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
> > +
> > +    if (s->aio_bitmap &&
> > +        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(s->src)) {
> > +        return !!(s->aio_bitmap[chunk / (sizeof(unsigned long) * 8)] &
> > +            (1UL << (chunk % (sizeof(unsigned long) * 8))));
> 
> Please use the bitmap functions in bitmap.h, also in the next function.

Can't do, this bitmap can be larger than 2^32, while bitmap.h uses 'int'
for index.

> > +    if (bdrv_open(s->src, s->src_filename, open_flags, NULL) < 0) {
> > +        error_report("%s: %s: cannot fallback to source image\n", __func__,
> > +                     s->src_filename);
> 
> Below qerror_report() is used.

What is the problem? Don't get the point.

Fixed other comments, thanks for reviewing.
Marcelo Tosatti June 3, 2011, 4:20 p.m. UTC | #8
On Tue, May 31, 2011 at 07:53:44PM +0300, Avi Kivity wrote:
> >Don't see the need for that, management can simply wait for livecopy to
> >finish or cancel livecopy and restart on destination after migration.
> 
> They can do it, but I imagine it's pretty hard for them.

Well, they have to handle migration failures anyway.

> 
> >But yeah, it should be implemented sometime...
> >
> 
> 
> We should make an effort not to introduce interdependencies which
> make management's life harder.

While this is a good principle there is no way around complexity for
management here. If migration is supported, the block copy target image
filenames must be specified on the destination.
diff mbox

Patch

Index: qemu-block-copy/block-copy.c
===================================================================
--- /dev/null
+++ qemu-block-copy/block-copy.c
@@ -0,0 +1,754 @@ 
+/*
+ * QEMU live block copy
+ *
+ * Copyright (C) 2010 Red Hat Inc.
+ *
+ * Authors: Marcelo Tosatti <mtosatti@redhat.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu-common.h"
+#include "block_int.h"
+#include "blockdev.h"
+#include "qemu-queue.h"
+#include "qemu-timer.h"
+#include "monitor.h"
+#include "block-copy.h"
+#include "migration.h"
+#include "sysemu.h"
+#include "qjson.h"
+#include <assert.h>
+
+#define BLOCK_SIZE (BDRV_SECTORS_PER_DIRTY_CHUNK << BDRV_SECTOR_BITS)
+#define MAX_IS_ALLOCATED_SEARCH 65536
+
+/*
+ * Stages:
+ *
+ * STAGE_BULK: bulk reads/writes in progress
+ * STAGE_BULK_FINISHED: bulk reads finished, bulk writes in progress
+ * STAGE_DIRTY: bulk writes finished, dirty reads/writes in progress
+ * STAGE_MIRROR_WRITES: copy finished, writes mirrored to both images.
+ * STAGE_SWITCH_FINISHED: switched to new image.
+ */
+
+enum BdrvCopyStage {
+    STAGE_BULK,
+    STAGE_BULK_FINISHED,
+    STAGE_DIRTY,
+    STAGE_MIRROR_WRITES,
+    STAGE_SWITCH_FINISHED,
+};
+
+typedef struct BdrvCopyState {
+    BlockDriverState *src;
+    BlockDriverState *dst;
+    bool shared_base;
+
+    int64_t curr_sector;
+    int64_t completed_sectors;
+    int64_t nr_sectors;
+
+    enum BdrvCopyStage stage;
+    int inflight_reads;
+    int error;
+    int failed;
+    int cancelled;
+    QLIST_HEAD(, BdrvCopyBlock) io_list;
+    unsigned long *aio_bitmap;
+    QEMUTimer *aio_timer;
+    QLIST_ENTRY(BdrvCopyState) list;
+
+    int64_t blocks;
+    int64_t total_time;
+
+    char src_device_name[32];
+    char src_filename[1024];
+} BdrvCopyState;
+
+typedef struct BdrvCopyBlock {
+    BdrvCopyState *state;
+    uint8_t *buf;
+    int64_t sector;
+    int64_t nr_sectors;
+    struct iovec iov;
+    QEMUIOVector qiov;
+    BlockDriverAIOCB *aiocb;
+    int64_t time;
+    QLIST_ENTRY(BdrvCopyBlock) list;
+} BdrvCopyBlock;
+
+static QLIST_HEAD(, BdrvCopyState) block_copy_list =
+    QLIST_HEAD_INITIALIZER(block_copy_list);
+
+static void alloc_aio_bitmap(BdrvCopyState *s)
+{
+    BlockDriverState *bs = s->src;
+    int64_t bitmap_size;
+
+    bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
+            BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
+    bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
+
+    s->aio_bitmap = qemu_mallocz(bitmap_size);
+}
+
+static bool aio_inflight(BdrvCopyState *s, int64_t sector)
+{
+    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
+
+    if (s->aio_bitmap &&
+        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(s->src)) {
+        return !!(s->aio_bitmap[chunk / (sizeof(unsigned long) * 8)] &
+            (1UL << (chunk % (sizeof(unsigned long) * 8))));
+    } else {
+        return 0;
+    }
+}
+
+static void set_aio_inflight(BdrvCopyState *s, int64_t sector_num,
+                             int nb_sectors, int set)
+{
+    int64_t start, end;
+    unsigned long val, idx, bit;
+
+    start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
+    end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
+
+    for (; start <= end; start++) {
+        idx = start / (sizeof(unsigned long) * 8);
+        bit = start % (sizeof(unsigned long) * 8);
+        val = s->aio_bitmap[idx];
+        if (set) {
+            if (!(val & (1UL << bit))) {
+                val |= 1UL << bit;
+            }
+        } else {
+            if (val & (1UL << bit)) {
+                val &= ~(1UL << bit);
+            }
+        }
+        s->aio_bitmap[idx] = val;
+    }
+}
+
+static void blkcopy_set_stage(BdrvCopyState *s, enum BdrvCopyStage stage)
+{
+    s->stage = stage;
+
+    switch (stage) {
+    case STAGE_BULK:
+        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_BULK);
+        break;
+    case STAGE_BULK_FINISHED:
+        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_BULK_FINISHED);
+        break;
+    case STAGE_DIRTY:
+        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_DIRTY);
+        break;
+    case STAGE_MIRROR_WRITES:
+        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_MIRROR_WRITES);
+        break;
+    case STAGE_SWITCH_FINISHED:
+        BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_STAGE_SWITCH_FINISHED);
+        break;
+    default:
+        break;
+    }
+}
+
+static void blk_copy_handle_cb_error(BdrvCopyState *s, int ret)
+{
+    s->error = ret;
+    qemu_mod_timer(s->aio_timer, qemu_get_clock_ms(rt_clock));
+}
+
+static inline void add_avg_transfer_time(BdrvCopyState *s, int64_t time)
+{
+    s->blocks++;
+    s->total_time += time;
+}
+
+static void blk_copy_write_cb(void *opaque, int ret)
+{
+    BdrvCopyBlock *blk = opaque;
+    BdrvCopyState *s = blk->state;
+
+    if (ret < 0) {
+        QLIST_REMOVE(blk, list);
+        qemu_free(blk->buf);
+        qemu_free(blk);
+        blk_copy_handle_cb_error(s, ret);
+        return;
+    }
+
+    QLIST_REMOVE(blk, list);
+    add_avg_transfer_time(s, qemu_get_clock_ns(rt_clock) - blk->time);
+
+    /* schedule switch to STAGE_DIRTY on last bulk write completion */
+    if (blk->state->stage == STAGE_BULK_FINISHED) {
+        qemu_mod_timer(s->aio_timer, qemu_get_clock_ms(rt_clock));
+    }
+
+    if (blk->state->stage > STAGE_BULK_FINISHED) {
+        set_aio_inflight(blk->state, blk->sector, blk->nr_sectors, 0);
+    }
+
+    qemu_free(blk->buf);
+    qemu_free(blk);
+}
+
+static void blk_copy_issue_write(BdrvCopyState *s, BdrvCopyBlock *read_blk)
+{
+    BdrvCopyBlock *blk = qemu_mallocz(sizeof(BdrvCopyBlock));
+    blk->state = s;
+    blk->sector = read_blk->sector;
+    blk->nr_sectors = read_blk->nr_sectors;
+    blk->time = read_blk->time;
+    blk->buf = read_blk->buf;
+    QLIST_INSERT_HEAD(&s->io_list, blk, list);
+
+    blk->iov.iov_base = read_blk->buf;
+    blk->iov.iov_len = read_blk->iov.iov_len;
+    qemu_iovec_init_external(&blk->qiov, &blk->iov, 1);
+
+    BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_AIO_WRITE);
+    blk->aiocb = bdrv_aio_writev(s->dst, blk->sector, &blk->qiov,
+                                 blk->iov.iov_len / BDRV_SECTOR_SIZE,
+                                 blk_copy_write_cb, blk);
+    if (!blk->aiocb) {
+        s->error = 1;
+        goto error;
+    }
+
+    return;
+
+error:
+    QLIST_REMOVE(blk, list);
+    qemu_free(read_blk->buf);
+    qemu_free(blk);
+}
+
+static void blk_copy_read_cb(void *opaque, int ret)
+{
+    BdrvCopyBlock *blk = opaque;
+    BdrvCopyState *s = blk->state;
+
+    s->inflight_reads--;
+    if (ret < 0) {
+        QLIST_REMOVE(blk, list);
+        qemu_free(blk->buf);
+        qemu_free(blk);
+        blk_copy_handle_cb_error(s, ret);
+        return;
+    }
+    blk_copy_issue_write(s, blk);
+    QLIST_REMOVE(blk, list);
+    qemu_free(blk);
+    qemu_mod_timer(s->aio_timer, qemu_get_clock_ms(rt_clock));
+}
+
+static void blk_copy_issue_read(BdrvCopyState *s, int64_t sector,
+                                int nr_sectors)
+{
+    BdrvCopyBlock *blk = qemu_mallocz(sizeof(BdrvCopyBlock));
+    blk->buf = qemu_mallocz(BLOCK_SIZE);
+    blk->state = s;
+    blk->sector = sector;
+    blk->nr_sectors = nr_sectors;
+    QLIST_INSERT_HEAD(&s->io_list, blk, list);
+
+    blk->iov.iov_base = blk->buf;
+    blk->iov.iov_len = nr_sectors * BDRV_SECTOR_SIZE;
+    qemu_iovec_init_external(&blk->qiov, &blk->iov, 1);
+
+    s->inflight_reads++;
+    blk->time = qemu_get_clock_ns(rt_clock);
+    blk->aiocb = bdrv_aio_readv(s->src, sector, &blk->qiov, nr_sectors,
+                                blk_copy_read_cb, blk);
+    if (!blk->aiocb) {
+        s->error = 1;
+        goto error;
+    }
+
+    return;
+
+error:
+    s->inflight_reads--;
+    QLIST_REMOVE(blk, list);
+    qemu_free(blk->buf);
+    qemu_free(blk);
+}
+
+static bool blkcopy_can_switch(BdrvCopyState *s)
+{
+    int64_t remaining_dirty;
+    int64_t avg_transfer_time;
+
+    remaining_dirty = bdrv_get_dirty_count(s->src);
+    if (remaining_dirty == 0 || s->blocks == 0) {
+        return true;
+    }
+
+    avg_transfer_time = s->total_time / s->blocks;
+    if ((remaining_dirty * avg_transfer_time) <= migrate_max_downtime()) {
+        return true;
+    }
+    return false;
+}
+
+static int blk_issue_reads_dirty(BdrvCopyState *s)
+{
+    int64_t sector;
+
+    for (sector = s->curr_sector; sector < s->nr_sectors;) {
+        if (bdrv_get_dirty(s->src, sector) && !aio_inflight(s, sector)) {
+            int nr_sectors = MIN(s->nr_sectors - s->curr_sector,
+                                 BDRV_SECTORS_PER_DIRTY_CHUNK);
+
+            blk_copy_issue_read(s, sector, nr_sectors);
+            bdrv_reset_dirty(s->src, sector, nr_sectors);
+            set_aio_inflight(s, sector, nr_sectors, 1);
+            break;
+        }
+
+        sector += BDRV_SECTORS_PER_DIRTY_CHUNK;
+        s->curr_sector = sector;
+    }
+
+    if (sector >= s->nr_sectors) {
+        s->curr_sector = 0;
+    }
+    return 0;
+}
+
+static int blk_issue_reads_bulk(BdrvCopyState *s)
+{
+    int nr_sectors;
+    int64_t curr_sector = s->curr_sector;
+
+    if (s->shared_base) {
+        while (curr_sector < s->nr_sectors &&
+                !bdrv_is_allocated(s->src, curr_sector,
+                                   MAX_IS_ALLOCATED_SEARCH, &nr_sectors)) {
+                curr_sector += nr_sectors;
+        }
+    }
+
+    if (curr_sector >= s->nr_sectors) {
+        s->curr_sector = 0;
+        return 1;
+    }
+
+    curr_sector &= ~((int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK - 1);
+    nr_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;
+
+    blk_copy_issue_read(s, s->curr_sector, nr_sectors);
+    s->curr_sector += nr_sectors;
+    s->completed_sectors = curr_sector;
+    return 0;
+}
+
+static void blkcopy_finish(BdrvCopyState *s)
+{
+    int64_t sector;
+    uint8_t *buf;
+
+    buf = qemu_malloc(BLOCK_SIZE);
+
+    /* FIXME: speed up loop, get_next_dirty_block? */
+    for (sector = 0; sector < s->nr_sectors;
+         sector += BDRV_SECTORS_PER_DIRTY_CHUNK) {
+        if (bdrv_get_dirty(s->src, sector)) {
+            int nr_sectors = MIN(s->nr_sectors - sector,
+                                 BDRV_SECTORS_PER_DIRTY_CHUNK);
+
+            memset(buf, 0, BLOCK_SIZE);
+            if (bdrv_read(s->src, sector, buf, nr_sectors) < 0) {
+                goto error;
+            }
+            if (bdrv_write(s->dst, sector, buf, nr_sectors) < 0) {
+                goto error;
+            }
+            bdrv_reset_dirty(s->src, sector, nr_sectors);
+        }
+
+        if (bdrv_get_dirty_count(s->src) == 0)
+            break;
+    }
+    qemu_free(buf);
+    return;
+
+error:
+    qemu_free(buf);
+    s->error = 1;
+}
+
+static void blkcopy_cleanup(BdrvCopyState *s)
+{
+    assert(s->inflight_reads == 0);
+    assert(QLIST_EMPTY(&s->io_list));
+    bdrv_set_dirty_tracking(s->src, 0);
+    drive_put_ref(drive_get_by_blockdev(s->src));
+    bdrv_set_in_use(s->src, 0);
+    if (s->stage >= STAGE_DIRTY)
+        qemu_free(s->aio_bitmap);
+    qemu_del_timer(s->aio_timer);
+}
+
+static void blkcopy_free(BdrvCopyState *s)
+{
+    QLIST_REMOVE(s, list);
+    qemu_free(s);
+}
+
+static void handle_error(BdrvCopyState *s)
+{
+    if (!QLIST_EMPTY(&s->io_list))
+        return;
+    s->failed = 1;
+    blkcopy_cleanup(s);
+}
+
+static void blkcopy_switch(BdrvCopyState *s)
+{
+    char src_filename[1024], mirror_name[2048];
+    int open_flags, ret;
+
+    strncpy(src_filename, s->src->filename, sizeof(src_filename));
+    open_flags = s->src->open_flags;
+
+    assert(s->stage == STAGE_DIRTY);
+
+    vm_stop(VMSTOP_BLOCKCOPY);
+    /* flush any guest writes, dirty bitmap uptodate after this.
+     * copy AIO also finished.
+     */
+    qemu_aio_flush();
+    assert(QLIST_EMPTY(&s->io_list));
+    if (s->error) {
+        handle_error(s);
+        goto vm_start;
+    }
+    blkcopy_finish(s);
+    if (s->error) {
+        handle_error(s);
+        goto vm_start;
+    }
+    assert(bdrv_get_dirty_count(s->src) == 0);
+    /* turn dirty bitmap off */
+    bdrv_set_dirty_tracking(s->src, 0);
+    /* switch to double writes */
+    bdrv_flush_all();
+    bdrv_close(s->src);
+    bdrv_close(s->dst);
+
+    snprintf(mirror_name, sizeof(mirror_name)-1,
+             "blkmirror:%s:%s", s->dst->filename, s->src->filename);
+
+    ret = bdrv_open(s->src, mirror_name, s->src->open_flags, NULL);
+    if (ret < 0) {
+        error_report("%s: cannot open blkmirror device, err %d",
+                      mirror_name, ret);
+        s->failed = 1;
+        goto err;
+    }
+
+    blkcopy_set_stage(s, STAGE_MIRROR_WRITES);
+    qemu_del_timer(s->aio_timer);
+
+vm_start:
+    vm_start();
+    return;
+
+err:
+    if (bdrv_open(s->src, src_filename, open_flags, NULL) < 0) {
+        error_report("%s: %s: cannot fallback to source image\n", __func__,
+                     s->src_filename);
+        abort();
+    }
+    blkcopy_cleanup(s);
+    goto vm_start;
+}
+
+#define BLKCOPY_INFLIGHT 2
+
+/*
+ * To simplify the implementation, the IO completion callbacks do not
+ * handle stage control or submit IO for further blocks. A timer is used
+ * for such purpose.
+ */
+
+static void aio_timer(void *opaque)
+{
+    BdrvCopyState *s = opaque;
+
+    assert(s->cancelled == 0);
+    assert(s->stage < STAGE_MIRROR_WRITES);
+
+    if (s->error) {
+        handle_error(s);
+        return;
+    }
+
+    while (s->stage == STAGE_BULK) {
+        if (s->inflight_reads >= BLKCOPY_INFLIGHT) {
+            break;
+        }
+        if (blk_issue_reads_bulk(s)) {
+            blkcopy_set_stage(s, STAGE_BULK_FINISHED);
+        }
+    }
+
+    if (s->stage == STAGE_BULK_FINISHED) {
+        if (QLIST_EMPTY(&s->io_list)) {
+            blkcopy_set_stage(s, STAGE_DIRTY);
+            alloc_aio_bitmap(s);
+        }
+    }
+
+    while (s->stage == STAGE_DIRTY) {
+        if (s->inflight_reads >= BLKCOPY_INFLIGHT) {
+            break;
+        }
+        blk_issue_reads_dirty(s);
+        if (blkcopy_can_switch(s)) {
+            BLKDBG_EVENT(s->dst->file, BLKDBG_BLKCOPY_SWITCH_START);
+            blkcopy_switch(s);
+            return;
+        }
+    }
+}
+
+
+int do_bdrv_copy_switch(Monitor *mon, const QDict *qdict, QObject **ret_data)
+{
+    const char *device = qdict_get_str(qdict, "device");
+    BdrvCopyState *s = NULL;
+    int open_flags;
+
+    QLIST_FOREACH(s, &block_copy_list, list) {
+        if (!strcmp(s->src_device_name, device)) {
+            if (s->stage != STAGE_MIRROR_WRITES) {
+                qerror_report(QERR_IN_PROGRESS, "block copy");
+                return -1;
+            }
+            break;
+        }
+    }
+
+    if (!s) {
+        qerror_report(QERR_DEVICE_NOT_FOUND, device);
+        return -1;
+    }
+
+    open_flags = s->src->open_flags;
+
+    /* switch from mirrored writes to destination only */
+    bdrv_flush_all();
+    bdrv_close(s->src);
+    if (bdrv_open(s->src, s->dst->filename, s->src->open_flags, NULL) < 0) {
+        s->failed = 1;
+        goto err;
+    }
+
+    blkcopy_set_stage(s, STAGE_SWITCH_FINISHED);
+    blkcopy_cleanup(s);
+    return 0;
+
+err:
+    if (bdrv_open(s->src, s->src_filename, open_flags, NULL) < 0) {
+        error_report("%s: %s: cannot fallback to source image\n", __func__,
+                     s->src_filename);
+        abort();
+    }
+    return -1;
+}
+
+static int bdrv_copy(Monitor *mon, const char * device, BlockDriverState *src,
+                     BlockDriverState *dst, bool shared_base)
+{
+    int64_t sectors;
+    BdrvCopyState *blkcopy, *safe;
+
+    QLIST_FOREACH_SAFE(blkcopy, &block_copy_list, list, safe) {
+        if (!strcmp(blkcopy->src_device_name, src->device_name)) {
+            if (blkcopy->stage == STAGE_SWITCH_FINISHED || blkcopy->failed) {
+                blkcopy_free(blkcopy);
+            } else {
+                qerror_report(QERR_IN_PROGRESS, "block copy");
+                return -1;
+            }
+        }
+    }
+
+    sectors = bdrv_getlength(src) >> BDRV_SECTOR_BITS;
+    if (sectors != bdrv_getlength(dst) >> BDRV_SECTOR_BITS) {
+        qerror_report(QERR_BLOCKCOPY_IMAGE_SIZE_DIFFERS);
+        return -1;
+    }
+
+    blkcopy = qemu_mallocz(sizeof(BdrvCopyState));
+    blkcopy->src = src;
+    blkcopy->dst = dst;
+    blkcopy->curr_sector = 0;
+    blkcopy->nr_sectors = sectors;
+    blkcopy_set_stage(blkcopy, STAGE_BULK);
+    blkcopy->aio_timer = qemu_new_timer_ms(rt_clock, aio_timer, blkcopy);
+    blkcopy->shared_base = shared_base;
+    strncpy(blkcopy->src_device_name, blkcopy->src->device_name,
+            sizeof(blkcopy->src_device_name) - 1);
+    strncpy(blkcopy->src_filename, blkcopy->src->filename,
+            sizeof(blkcopy->src_filename) - 1);
+
+    bdrv_set_dirty_tracking(src, 1);
+    qemu_mod_timer(blkcopy->aio_timer, qemu_get_clock_ms(rt_clock));
+    drive_get_ref(drive_get_by_blockdev(src));
+    bdrv_set_in_use(src, 1);
+
+    QLIST_INSERT_HEAD(&block_copy_list, blkcopy, list);
+    return 0;
+}
+
+int do_bdrv_copy(Monitor *mon, const QDict *qdict, QObject **ret_data)
+{
+    const char *device = qdict_get_str(qdict, "device");
+    const char *filename = qdict_get_str(qdict, "filename");
+    bool shared_base = qdict_get_try_bool(qdict, "incremental", 0);
+    BlockDriverState *new_bs, *bs;
+    int ret;
+
+    if (migration_active()) {
+        qerror_report(QERR_IN_PROGRESS, "migration");
+        return -1;
+    }
+
+    bs = bdrv_find(device);
+    if (!bs) {
+        qerror_report(QERR_DEVICE_NOT_FOUND, device);
+        return -1;
+    }
+
+    new_bs = bdrv_new("");
+    if (bdrv_open(new_bs, filename, bs->open_flags, NULL) < 0) {
+        bdrv_delete(new_bs);
+        qerror_report(QERR_OPEN_FILE_FAILED, filename);
+        return -1;
+    }
+
+    ret = bdrv_copy(mon, device, bs, new_bs, shared_base);
+    return ret;
+}
+
+int do_bdrv_copy_cancel(Monitor *mon, const QDict *qdict, QObject **ret_data)
+{
+    BdrvCopyState *blkcopy, *s = NULL;
+    const char *device = qdict_get_str(qdict, "device");
+
+    QLIST_FOREACH(blkcopy, &block_copy_list, list) {
+        if (!strcmp(blkcopy->src_device_name, device)) {
+            s = blkcopy;
+            break;
+        }
+    }
+
+    if (!s || s->stage == STAGE_SWITCH_FINISHED || s->failed) {
+        qerror_report(QERR_DEVICE_NOT_FOUND, device);
+        return -1;
+    }
+
+    s->cancelled = 1;
+    do {
+        qemu_aio_flush();
+    } while (!QLIST_EMPTY(&s->io_list));
+    blkcopy_cleanup(s);
+    blkcopy_free(s);
+
+    return 0;
+}
+
+static void blockcopy_print_dict(QObject *obj, void *opaque)
+{
+    QDict *c_dict;
+    Monitor *mon = opaque;
+
+    c_dict = qobject_to_qdict(obj);
+
+    monitor_printf(mon, "%s: status=%s ",
+                        qdict_get_str(c_dict, "device"),
+                        qdict_get_str(c_dict, "status"));
+
+    if (qdict_haskey(c_dict, "info")) {
+        QDict *qdict = qobject_to_qdict(qdict_get(c_dict, "info"));
+
+        monitor_printf(mon, "percentage=%ld %%",
+                       qdict_get_int(qdict, "percentage"));
+    }
+
+    monitor_printf(mon, "\n");
+}
+
+void do_info_blockcopy_print(Monitor *mon, const QObject *data)
+{
+    qlist_iter(qobject_to_qlist(data), blockcopy_print_dict, mon);
+}
+
+void do_info_blockcopy(Monitor *mon, QObject **ret_data)
+{
+    QList *c_list;
+    BdrvCopyState *s;
+
+    c_list = qlist_new();
+
+    QLIST_FOREACH(s, &block_copy_list, list) {
+        QObject *c_obj;
+        static const char *status[] = { "failed", "active", "mirrored", "completed" };
+        int i;
+
+        if (s->failed) {
+            i = 0;
+        } else if (s->stage < STAGE_MIRROR_WRITES) {
+            i = 1;
+        } else if (s->stage < STAGE_SWITCH_FINISHED) {
+            i = 2;
+        } else {
+            i = 3;
+        }
+
+        c_obj = qobject_from_jsonf("{ 'device': %s, 'status': %s }",
+                                    s->src_device_name, status[i]);
+
+        if (i == 1) {
+            QDict *dict = qobject_to_qdict(c_obj);
+            QObject *obj;
+
+            /* FIXME: add dirty stage progress? */
+            obj = qobject_from_jsonf("{ 'percentage': %" PRId64 "}",
+                                     s->completed_sectors * 100 / s->nr_sectors);
+            qdict_put_obj(dict, "info", obj);
+        }
+        qlist_append_obj(c_list, c_obj);
+    }
+
+    *ret_data = QOBJECT(c_list);
+}
+
+bool block_copy_active(void)
+{
+    BdrvCopyState *s;
+
+    QLIST_FOREACH(s, &block_copy_list, list) {
+        if (s->failed) {
+            continue;
+        }
+        if (s->stage < STAGE_SWITCH_FINISHED) {
+            return true;
+        }
+    }
+
+    return false;
+}
+
Index: qemu-block-copy/block-copy.h
===================================================================
--- /dev/null
+++ qemu-block-copy/block-copy.h
@@ -0,0 +1,26 @@ 
+/*
+ * QEMU live block copy
+ *
+ * Copyright (C) 2010 Red Hat Inc.
+ *
+ * Authors: Marcelo Tosatti <mtosatti@redhat.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef BLOCK_COPY_H
+#define BLOCK_COPY_H
+
+int do_bdrv_copy(Monitor *mon, const QDict *qdict, QObject **ret_data);
+int do_bdrv_copy_cancel(Monitor *mon, const QDict *qdict, QObject **ret_data);
+int do_bdrv_copy_switch(Monitor *mon, const QDict *qdict, QObject **ret_data);
+
+void do_info_blockcopy_print(Monitor *mon, const QObject *data);
+void do_info_blockcopy(Monitor *mon, QObject **ret_data);
+
+bool block_copy_active(void);
+
+#endif /* BLOCK_COPY_H */
+
Index: qemu-block-copy/hmp-commands.hx
===================================================================
--- qemu-block-copy.orig/hmp-commands.hx
+++ qemu-block-copy/hmp-commands.hx
@@ -806,6 +806,63 @@  Set maximum speed to @var{value} (in byt
 ETEXI
 
     {
+        .name       = "block_copy",
+        .args_type  = "device:s,filename:s,incremental:-i",
+        .params     = "device filename [-i]",
+        .help       = "live block copy device to image"
+                      "\n\t\t\t -i for incremental copy "
+                      "(base image shared between original and destination)",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_bdrv_copy,
+    },
+
+STEXI
+@item block_copy @var{device} @var{filename} [-i]
+@findex block_copy
+Live copy block device @var{device} to image @var{filename}.
+        -i for incremental copy (base image is shared)
+
+Destination image @var{filename} must be created with qemu-img prior
+to execution of this command, with image size equal to the original
+image size.
+
+Incremental copy allows the destination image @var{filename} to share
+a common base image with the original image. This option skips copying
+blocks which are not allocated in the original image.
+ETEXI
+
+    {
+        .name       = "block_copy_cancel",
+        .args_type  = "device:s",
+        .params     = "device",
+        .help       = "cancel live block copy",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_bdrv_copy_cancel,
+    },
+
+STEXI
+@item block_copy_cancel @var{device}
+@findex block_copy_cancel
+Cancel live block copy on @var{device}.
+ETEXI
+
+    {
+        .name       = "block_copy_switch",
+        .args_type  = "device:s",
+        .params     = "device",
+        .help       = "finish live block copy",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_bdrv_copy_switch,
+    },
+
+STEXI
+@item block_copy_switch @var{device}
+@findex block_copy_switch
+Finish live block copy on @var{device} by switching
+to destination image.
+ETEXI
+
+    {
         .name       = "migrate_set_downtime",
         .args_type  = "value:T",
         .params     = "value",
@@ -1352,6 +1409,8 @@  show device tree
 show qdev device model list
 @item info roms
 show roms
+@item info block-copy
+show block copy status
 @end table
 ETEXI
 
Index: qemu-block-copy/monitor.c
===================================================================
--- qemu-block-copy.orig/monitor.c
+++ qemu-block-copy/monitor.c
@@ -45,6 +45,7 @@ 
 #include "balloon.h"
 #include "qemu-timer.h"
 #include "migration.h"
+#include "block-copy.h"
 #include "kvm.h"
 #include "acl.h"
 #include "qint.h"
@@ -3101,6 +3102,14 @@  static const mon_cmd_t info_cmds[] = {
     },
 #endif
     {
+        .name       = "block-copy",
+        .args_type  = "",
+        .params     = "",
+        .help       = "show block copy status",
+        .user_print = do_info_blockcopy_print,
+        .mhandler.info_new = do_info_blockcopy,
+    },
+    {
         .name       = NULL,
     },
 };
@@ -3242,6 +3251,14 @@  static const mon_cmd_t qmp_query_cmds[] 
         .mhandler.info_async = do_info_balloon,
         .flags      = MONITOR_CMD_ASYNC,
     },
+    {
+        .name       = "block-copy",
+        .args_type  = "",
+        .params     = "",
+        .help       = "show block copy status",
+        .user_print = do_info_blockcopy_print,
+        .mhandler.info_new = do_info_blockcopy,
+    },
     { /* NULL */ },
 };
 
Index: qemu-block-copy/qmp-commands.hx
===================================================================
--- qemu-block-copy.orig/qmp-commands.hx
+++ qemu-block-copy/qmp-commands.hx
@@ -581,6 +581,98 @@  Example:
 EQMP
 
     {
+        .name       = "block_copy",
+        .args_type  = "device:s,filename:s,inc:-i",
+        .params     = "device filename [-i]",
+        .help       = "live block copy device to image"
+                      "\n\t\t\t -i for incremental copy "
+                      "(base image shared between src and destination)",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_bdrv_copy,
+    },
+
+SQMP
+block-copy
+-------
+
+Live block copy.
+
+Arguments:
+
+- "device": device name (json-string)
+- "filename": target image filename (json-string)
+- "incremental": incremental disk copy (json-bool, optional)
+
+Example:
+
+-> { "execute": "block_copy",
+                            "arguments": { "device": "ide0-hd1",
+                               "filename": "/mnt/new-disk.img",
+                             } }
+
+<- { "return": {} }
+
+Notes:
+
+(1) The 'query-block-copy' command should be used to check block copy progress
+    and final result (this information is provided by the 'status' member)
+(2) Boolean argument "incremental" defaults to false
+
+EQMP
+
+    {
+        .name       = "block_copy_cancel",
+        .args_type  = "device:s",
+        .params     = "device",
+        .help       = "cancel live block copy",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_bdrv_copy_cancel,
+    },
+
+SQMP
+block_copy_cancel
+--------------
+
+Cancel live block copy.
+
+Arguments:
+
+- device: device name (json-string)
+
+Example:
+
+-> { "execute": "block_copy_cancel", "arguments": { "device": "ide0-hd1" } }
+<- { "return": {} }
+
+EQMP
+
+    {
+        .name       = "block_copy_switch",
+        .args_type  = "device:s",
+        .params     = "device",
+        .help       = "finish live block copy",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_bdrv_copy_switch,
+    },
+
+SQMP
+block_copy_switch
+--------------
+
+Finish live block copy, switching device to destination image.
+
+Arguments:
+
+- device: device name (json-string)
+
+Example:
+
+-> { "execute": "block_copy_switch", "arguments": { "device": "ide0-hd1" } }
+<- { "return": {} }
+
+EQMP
+
+    {
         .name       = "netdev_add",
         .args_type  = "netdev:O",
         .params     = "[user|tap|socket],id=str[,prop=value][,...]",
@@ -1744,6 +1836,44 @@  Examples:
 EQMP
 
 SQMP
+query-block-copy
+-------------
+
+Live block copy status.
+
+Each block copy instance information is stored in a json-object and the returned
+value is a json-array of all instances.
+
+Each json-object contains the following:
+
+- "device": device name (json-string)
+- "status": block copy status (json-string)
+    - Possible values: "active", "failed", "mirrored", "completed"
+- "info": A json-object with the statistics information, if status is "active":
+    - "percentage": percentage completed (json-int)
+
+Example:
+
+Block copy for "ide1-hd0" active and block copy for "ide1-hd1" failed:
+
+-> { "execute": "query-block-copy" }
+<- {
+      "return":[
+        {"device":"ide1-hd0",
+            "status":"active",
+            "info":{
+               "percentage":23,
+            }
+        },
+        {"device":"ide1-hd1",
+         "status":"failed"
+        }
+      ]
+   }
+
+EQMP
+
+SQMP
 query-balloon
 -------------
 
Index: qemu-block-copy/Makefile.objs
===================================================================
--- qemu-block-copy.orig/Makefile.objs
+++ qemu-block-copy/Makefile.objs
@@ -98,7 +98,7 @@  common-obj-y += buffered_file.o migratio
 common-obj-y += qemu-char.o savevm.o #aio.o
 common-obj-y += msmouse.o ps2.o
 common-obj-y += qdev.o qdev-properties.o
-common-obj-y += block-migration.o iohandler.o
+common-obj-y += block-migration.o iohandler.o block-copy.o
 common-obj-y += pflib.o
 common-obj-y += bitmap.o bitops.o