[qemu.git] / block-migration.c

/*
 * QEMU live block migration
 *
 * Copyright IBM, Corp. 2009
 *
 * Authors:
 *  Liran Schour   <lirans@il.ibm.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 */

#include "qemu-common.h"
#include "block_int.h"
#include "hw/hw.h"
#include "block-migration.h"
#include <assert.h>

#define BLOCK_SIZE (BDRV_SECTORS_PER_DIRTY_CHUNK << BDRV_SECTOR_BITS)

#define BLK_MIG_FLAG_DEVICE_BLOCK       0x01
#define BLK_MIG_FLAG_EOS                0x02

#define MAX_IS_ALLOCATED_SEARCH 65536
#define MAX_BLOCKS_READ 10000
#define BLOCKS_READ_CHANGE 100
#define INITIAL_BLOCKS_READ 100

//#define DEBUG_BLK_MIGRATION

#ifdef DEBUG_BLK_MIGRATION
#define dprintf(fmt, ...) \
    do { printf("blk_migration: " fmt, ## __VA_ARGS__); } while (0)
#else
#define dprintf(fmt, ...) \
    do { } while (0)
#endif

typedef struct BlkMigDevState {
    BlockDriverState *bs;
    int bulk_completed;
    int shared_base;
    struct BlkMigDevState *next;
    int64_t cur_sector;
    int64_t total_sectors;
    int64_t dirty;
} BlkMigDevState;

typedef struct BlkMigBlock {
    uint8_t *buf;
    BlkMigDevState *bmds;
    int64_t sector;
    struct iovec iov;
    QEMUIOVector qiov;
    BlockDriverAIOCB *aiocb;
    int ret;
    struct BlkMigBlock *next;
} BlkMigBlock;

typedef struct BlkMigState {
    int blk_enable;
    int shared_base;
    BlkMigDevState *bmds_first;
    BlkMigBlock *first_blk;
    BlkMigBlock *last_blk;
    int submitted;
    int read_done;
    int transferred;
    int64_t print_completion;
} BlkMigState;

static BlkMigState block_mig_state;

static void blk_mig_read_cb(void *opaque, int ret)
{
    BlkMigBlock *blk = opaque;

    blk->ret = ret;

    /* insert at the end */
    if (block_mig_state.last_blk == NULL) {
        block_mig_state.first_blk = blk;
        block_mig_state.last_blk = blk;
    } else {
        block_mig_state.last_blk->next = blk;
        block_mig_state.last_blk = blk;
    }

    block_mig_state.submitted--;
    block_mig_state.read_done++;
    assert(block_mig_state.submitted >= 0);
}

static int mig_read_device_bulk(QEMUFile *f, BlkMigDevState *bms)
{
    int nr_sectors;
    int64_t total_sectors, cur_sector = 0;
    BlockDriverState *bs = bms->bs;
    BlkMigBlock *blk;

    blk = qemu_malloc(sizeof(BlkMigBlock));
    blk->buf = qemu_malloc(BLOCK_SIZE);

    cur_sector = bms->cur_sector;
    total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;

    if (bms->shared_base) {
        while (cur_sector < bms->total_sectors &&
               !bdrv_is_allocated(bms->bs, cur_sector,
                                  MAX_IS_ALLOCATED_SEARCH, &nr_sectors)) {
            cur_sector += nr_sectors;
        }
    }

    if (cur_sector >= total_sectors) {
        bms->cur_sector = total_sectors;
        qemu_free(blk->buf);
        qemu_free(blk);
        return 1;
    }

    if (cur_sector >= block_mig_state.print_completion) {
        printf("Completed %" PRId64 " %%\r", cur_sector * 100 / total_sectors);
        fflush(stdout);
        block_mig_state.print_completion +=
            (BDRV_SECTORS_PER_DIRTY_CHUNK * 10000);
    }

    /* we are going to transfer a full block even if it is not allocated */
    nr_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;

    cur_sector &= ~((int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK - 1);

    if (total_sectors - cur_sector < BDRV_SECTORS_PER_DIRTY_CHUNK) {
        nr_sectors = (total_sectors - cur_sector);
    }

    bms->cur_sector = cur_sector + nr_sectors;
    blk->sector = cur_sector;
    blk->bmds = bms;
    blk->next = NULL;

    blk->iov.iov_base = blk->buf;
    blk->iov.iov_len = nr_sectors * BDRV_SECTOR_SIZE;
    qemu_iovec_init_external(&blk->qiov, &blk->iov, 1);

    blk->aiocb = bdrv_aio_readv(bs, cur_sector, &blk->qiov,
                                nr_sectors, blk_mig_read_cb, blk);

    if (!blk->aiocb) {
        printf("Error reading sector %" PRId64 "\n", cur_sector);
        qemu_free(blk->buf);
        qemu_free(blk);
        return 0;
    }

    bdrv_reset_dirty(bms->bs, cur_sector, nr_sectors);
    block_mig_state.submitted++;

    return (bms->cur_sector >= total_sectors);
}

static int mig_save_device_bulk(QEMUFile *f, BlkMigDevState *bmds)
{
    int len, nr_sectors;
    int64_t total_sectors = bmds->total_sectors, cur_sector = 0;
    uint8_t *tmp_buf = NULL;
    BlockDriverState *bs = bmds->bs;

    tmp_buf = qemu_malloc(BLOCK_SIZE);

    cur_sector = bmds->cur_sector;

    if (bmds->shared_base) {
        while (cur_sector < bmds->total_sectors &&
               !bdrv_is_allocated(bmds->bs, cur_sector,
                                  MAX_IS_ALLOCATED_SEARCH, &nr_sectors)) {
            cur_sector += nr_sectors;
        }
    }

    if (cur_sector >= total_sectors) {
        bmds->cur_sector = total_sectors;
        qemu_free(tmp_buf);
        return 1;
    }

    if (cur_sector >= block_mig_state.print_completion) {
        printf("Completed %" PRId64 " %%\r", cur_sector * 100 / total_sectors);
        fflush(stdout);
        block_mig_state.print_completion +=
            (BDRV_SECTORS_PER_DIRTY_CHUNK * 10000);
    }

    cur_sector &= ~((int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK - 1);

    /* we are going to transfer a full block even if it is not allocated */
    nr_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;

    if (total_sectors - cur_sector < BDRV_SECTORS_PER_DIRTY_CHUNK) {
        nr_sectors = (total_sectors - cur_sector);
    }

    if (bdrv_read(bs, cur_sector, tmp_buf, nr_sectors) < 0) {
        printf("Error reading sector %" PRId64 "\n", cur_sector);
    }

    bdrv_reset_dirty(bs, cur_sector, nr_sectors);

    /* sector number and flags */
    qemu_put_be64(f, (cur_sector << BDRV_SECTOR_BITS)
                     | BLK_MIG_FLAG_DEVICE_BLOCK);

    /* device name */
    len = strlen(bs->device_name);
    qemu_put_byte(f, len);
    qemu_put_buffer(f, (uint8_t *)bs->device_name, len);

    qemu_put_buffer(f, tmp_buf, BLOCK_SIZE);

    bmds->cur_sector = cur_sector + BDRV_SECTORS_PER_DIRTY_CHUNK;

    qemu_free(tmp_buf);

    return (bmds->cur_sector >= total_sectors);
}

static void send_blk(QEMUFile *f, BlkMigBlock * blk)
{
    int len;

    /* sector number and flags */
    qemu_put_be64(f, (blk->sector << BDRV_SECTOR_BITS)
                     | BLK_MIG_FLAG_DEVICE_BLOCK);

    /* device name */
    len = strlen(blk->bmds->bs->device_name);
    qemu_put_byte(f, len);
    qemu_put_buffer(f, (uint8_t *)blk->bmds->bs->device_name, len);

    qemu_put_buffer(f, blk->buf, BLOCK_SIZE);
}

static void set_dirty_tracking(int enable)
{
    BlkMigDevState *bmds;
    for (bmds = block_mig_state.bmds_first; bmds != NULL; bmds = bmds->next) {
        bdrv_set_dirty_tracking(bmds->bs, enable);
    }
}

static void init_blk_migration(QEMUFile *f)
{
    BlkMigDevState **pbmds, *bmds;
    BlockDriverState *bs;

    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
        if (bs->type == BDRV_TYPE_HD) {
            bmds = qemu_mallocz(sizeof(BlkMigDevState));
            bmds->bs = bs;
            bmds->bulk_completed = 0;
            bmds->total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
            bmds->shared_base = block_mig_state.shared_base;

            if (bmds->shared_base) {
                printf("Start migration for %s with shared base image\n",
                       bs->device_name);
            } else {
                printf("Start full migration for %s\n", bs->device_name);
            }

            /* insert at the end */
            pbmds = &block_mig_state.bmds_first;
            while (*pbmds != NULL) {
                pbmds = &(*pbmds)->next;
            }
            *pbmds = bmds;
        }
    }
}

static int blk_mig_save_bulked_block(QEMUFile *f, int is_async)
{
    BlkMigDevState *bmds;

    for (bmds = block_mig_state.bmds_first; bmds != NULL; bmds = bmds->next) {
        if (bmds->bulk_completed == 0) {
            if (is_async) {
                if (mig_read_device_bulk(f, bmds) == 1) {
                    /* completed bulk section for this device */
                    bmds->bulk_completed = 1;
                }
            } else {
                if (mig_save_device_bulk(f, bmds) == 1) {
                    /* completed bulk section for this device */
                    bmds->bulk_completed = 1;
                }
            }
            return 1;
        }
    }

    /* we reached here means bulk is completed */
    return 0;
}

#define MAX_NUM_BLOCKS 4

static void blk_mig_save_dirty_blocks(QEMUFile *f)
{
    BlkMigDevState *bmds;
    uint8_t *buf;
    int64_t sector;
    int len;

    buf = qemu_malloc(BLOCK_SIZE);

    for (bmds = block_mig_state.bmds_first; bmds != NULL; bmds = bmds->next) {
        for (sector = 0; sector < bmds->cur_sector;) {
            if (bdrv_get_dirty(bmds->bs, sector)) {
                if (bdrv_read(bmds->bs, sector, buf,
                              BDRV_SECTORS_PER_DIRTY_CHUNK) < 0) {
                    /* FIXME: add error handling */
                }

                /* sector number and flags */
                qemu_put_be64(f, (sector << BDRV_SECTOR_BITS)
                                 | BLK_MIG_FLAG_DEVICE_BLOCK);

                /* device name */
                len = strlen(bmds->bs->device_name);
                qemu_put_byte(f, len);
                qemu_put_buffer(f, (uint8_t *)bmds->bs->device_name, len);

                qemu_put_buffer(f, buf, BLOCK_SIZE);

                bdrv_reset_dirty(bmds->bs, sector,
                                 BDRV_SECTORS_PER_DIRTY_CHUNK);
            }
            sector += BDRV_SECTORS_PER_DIRTY_CHUNK;
        }
    }

    qemu_free(buf);
}

static void flush_blks(QEMUFile* f)
{
    BlkMigBlock *blk, *next;

    dprintf("%s Enter submitted %d read_done %d transferred %d\n",
            __FUNCTION__, block_mig_state.submitted, block_mig_state.read_done,
            block_mig_state.transferred);

    for (blk = block_mig_state.first_blk;
         blk != NULL && !qemu_file_rate_limit(f);
         blk = next) {
        send_blk(f, blk);

        next = blk->next;
        qemu_free(blk->buf);
        qemu_free(blk);

        block_mig_state.read_done--;
        block_mig_state.transferred++;
        assert(block_mig_state.read_done >= 0);
    }
    block_mig_state.first_blk = blk;

    if (block_mig_state.first_blk == NULL) {
        block_mig_state.last_blk = NULL;
    }

    dprintf("%s Exit submitted %d read_done %d transferred %d\n", __FUNCTION__,
            block_mig_state.submitted, block_mig_state.read_done,
            block_mig_state.transferred);
}

static int is_stage2_completed(void)
{
    BlkMigDevState *bmds;

    if (block_mig_state.submitted > 0) {
        return 0;
    }

    for (bmds = block_mig_state.bmds_first; bmds != NULL; bmds = bmds->next) {
        if (bmds->bulk_completed == 0) {
            return 0;
        }
    }

    return 1;
}

static int block_save_live(QEMUFile *f, int stage, void *opaque)
{
    dprintf("Enter save live stage %d submitted %d transferred %d\n",
            stage, block_mig_state.submitted, block_mig_state.transferred);

    if (block_mig_state.blk_enable != 1) {
        /* no need to migrate storage */
        qemu_put_be64(f, BLK_MIG_FLAG_EOS);
        return 1;
    }

    if (stage == 1) {
        init_blk_migration(f);

        /* start track dirty blocks */
        set_dirty_tracking(1);
    }

    flush_blks(f);

    /* control the rate of transfer */
    while ((block_mig_state.submitted +
            block_mig_state.read_done) * BLOCK_SIZE <
           qemu_file_get_rate_limit(f)) {
        if (blk_mig_save_bulked_block(f, 1) == 0) {
            /* no more bulk blocks for now */
            break;
        }
    }

    flush_blks(f);

    if (stage == 3) {
        while (blk_mig_save_bulked_block(f, 0) != 0) {
            /* empty */
        }

        blk_mig_save_dirty_blocks(f);

        /* stop track dirty blocks */
        set_dirty_tracking(0);

        printf("\nBlock migration completed\n");
    }

    qemu_put_be64(f, BLK_MIG_FLAG_EOS);

    return ((stage == 2) && is_stage2_completed());
}

static int block_load(QEMUFile *f, void *opaque, int version_id)
{
    int len, flags;
    char device_name[256];
    int64_t addr;
    BlockDriverState *bs;
    uint8_t *buf;

    do {
        addr = qemu_get_be64(f);

        flags = addr & ~BDRV_SECTOR_MASK;
        addr >>= BDRV_SECTOR_BITS;

        if (flags & BLK_MIG_FLAG_DEVICE_BLOCK) {
            /* get device name */
            len = qemu_get_byte(f);

            qemu_get_buffer(f, (uint8_t *)device_name, len);
            device_name[len] = '\0';

            bs = bdrv_find(device_name);

            buf = qemu_malloc(BLOCK_SIZE);

            qemu_get_buffer(f, buf, BLOCK_SIZE);
            if (bs != NULL) {
                bdrv_write(bs, addr, buf, BDRV_SECTORS_PER_DIRTY_CHUNK);
            } else {
                printf("Error unknown block device %s\n", device_name);
                /* FIXME: add error handling */
            }

            qemu_free(buf);
        } else if (!(flags & BLK_MIG_FLAG_EOS)) {
            printf("Unknown flags\n");
            /* FIXME: add error handling */
        }
    } while (!(flags & BLK_MIG_FLAG_EOS));

    return 0;
}

static void block_set_params(int blk_enable, int shared_base, void *opaque)
{
    block_mig_state.blk_enable = blk_enable;
    block_mig_state.shared_base = shared_base;

    /* shared base means that blk_enable = 1 */
    block_mig_state.blk_enable |= shared_base;
}

void blk_mig_init(void)
{
    register_savevm_live("block", 0, 1, block_set_params, block_save_live,
                         NULL, block_load, &block_mig_state);
}
Commit	Line	Data
c163b5ca	1	/*
	2	* QEMU live block migration
	3	*
	4	* Copyright IBM, Corp. 2009
	5	*
	6	* Authors:
	7	* Liran Schour <lirans@il.ibm.com>
	8	*
	9	* This work is licensed under the terms of the GNU GPL, version 2. See
	10	* the COPYING file in the top-level directory.
	11	*
	12	*/
	13
	14	#include "qemu-common.h"
	15	#include "block_int.h"
	16	#include "hw/hw.h"
	17	#include "block-migration.h"
	18	#include <assert.h>
c163b5ca	19
6ea44308	20	#define BLOCK_SIZE (BDRV_SECTORS_PER_DIRTY_CHUNK << BDRV_SECTOR_BITS)
c163b5ca	21
	22	#define BLK_MIG_FLAG_DEVICE_BLOCK 0x01
	23	#define BLK_MIG_FLAG_EOS 0x02
	24
	25	#define MAX_IS_ALLOCATED_SEARCH 65536
	26	#define MAX_BLOCKS_READ 10000
	27	#define BLOCKS_READ_CHANGE 100
	28	#define INITIAL_BLOCKS_READ 100
	29
	30	//#define DEBUG_BLK_MIGRATION
	31
	32	#ifdef DEBUG_BLK_MIGRATION
a55eb92c	33	#define dprintf(fmt, ...) \
c163b5ca	34	do { printf("blk_migration: " fmt, ## __VA_ARGS__); } while (0)
c163b5ca	35	#else
a55eb92c	36	#define dprintf(fmt, ...) \
c163b5ca	37	do { } while (0)
	38	#endif
	39
a55eb92c JK	40	typedef struct BlkMigDevState {
	41	BlockDriverState *bs;
	42	int bulk_completed;
	43	int shared_base;
	44	struct BlkMigDevState *next;
	45	int64_t cur_sector;
	46	int64_t total_sectors;
	47	int64_t dirty;
	48	} BlkMigDevState;
	49
c163b5ca	50	typedef struct BlkMigBlock {
	51	uint8_t *buf;
	52	BlkMigDevState *bmds;
	53	int64_t sector;
	54	struct iovec iov;
	55	QEMUIOVector qiov;
	56	BlockDriverAIOCB *aiocb;
	57	int ret;
	58	struct BlkMigBlock *next;
	59	} BlkMigBlock;
	60
	61	typedef struct BlkMigState {
c163b5ca	62	int blk_enable;
c163b5ca	63	int shared_base;
c163b5ca	64	BlkMigDevState *bmds_first;
c163b5ca	65	BlkMigBlock *first_blk;
	66	BlkMigBlock *last_blk;
	67	int submitted;
	68	int read_done;
	69	int transferred;
	70	int64_t print_completion;
	71	} BlkMigState;
	72
d11ecd3d	73	static BlkMigState block_mig_state;
c163b5ca	74
	75	static void blk_mig_read_cb(void *opaque, int ret)
	76	{
	77	BlkMigBlock *blk = opaque;
a55eb92c	78
c163b5ca	79	blk->ret = ret;
a55eb92c	80
c163b5ca	81	/* insert at the end */
d11ecd3d JK	82	if (block_mig_state.last_blk == NULL) {
	83	block_mig_state.first_blk = blk;
	84	block_mig_state.last_blk = blk;
c163b5ca	85	} else {
d11ecd3d JK	86	block_mig_state.last_blk->next = blk;
d11ecd3d JK	87	block_mig_state.last_blk = blk;
c163b5ca	88	}
a55eb92c	89
d11ecd3d JK	90	block_mig_state.submitted--;
	91	block_mig_state.read_done++;
	92	assert(block_mig_state.submitted >= 0);
c163b5ca	93	}
	94
	95	static int mig_read_device_bulk(QEMUFile f, BlkMigDevState bms)
a55eb92c	96	{
c163b5ca	97	int nr_sectors;
	98	int64_t total_sectors, cur_sector = 0;
	99	BlockDriverState *bs = bms->bs;
	100	BlkMigBlock *blk;
a55eb92c	101
c163b5ca	102	blk = qemu_malloc(sizeof(BlkMigBlock));
c163b5ca	103	blk->buf = qemu_malloc(BLOCK_SIZE);
a55eb92c	104
c163b5ca	105	cur_sector = bms->cur_sector;
6ea44308	106	total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
a55eb92c JK	107
	108	if (bms->shared_base) {
	109	while (cur_sector < bms->total_sectors &&
	110	!bdrv_is_allocated(bms->bs, cur_sector,
	111	MAX_IS_ALLOCATED_SEARCH, &nr_sectors)) {
c163b5ca	112	cur_sector += nr_sectors;
	113	}
	114	}
a55eb92c JK	115
a55eb92c JK	116	if (cur_sector >= total_sectors) {
c163b5ca	117	bms->cur_sector = total_sectors;
	118	qemu_free(blk->buf);
	119	qemu_free(blk);
	120	return 1;
	121	}
a55eb92c	122
d11ecd3d	123	if (cur_sector >= block_mig_state.print_completion) {
c163b5ca	124	printf("Completed %" PRId64 " %%\r", cur_sector * 100 / total_sectors);
c163b5ca	125	fflush(stdout);
d11ecd3d	126	block_mig_state.print_completion +=
6ea44308	127	(BDRV_SECTORS_PER_DIRTY_CHUNK * 10000);
c163b5ca	128	}
a55eb92c	129
6ea44308 JK	130	/* we are going to transfer a full block even if it is not allocated */
6ea44308 JK	131	nr_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;
c163b5ca	132
6ea44308	133	cur_sector &= ~((int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK - 1);
a55eb92c	134
6ea44308	135	if (total_sectors - cur_sector < BDRV_SECTORS_PER_DIRTY_CHUNK) {
c163b5ca	136	nr_sectors = (total_sectors - cur_sector);
c163b5ca	137	}
a55eb92c	138
c163b5ca	139	bms->cur_sector = cur_sector + nr_sectors;
	140	blk->sector = cur_sector;
	141	blk->bmds = bms;
	142	blk->next = NULL;
a55eb92c	143
c163b5ca	144	blk->iov.iov_base = blk->buf;
6ea44308	145	blk->iov.iov_len = nr_sectors * BDRV_SECTOR_SIZE;
c163b5ca	146	qemu_iovec_init_external(&blk->qiov, &blk->iov, 1);
a55eb92c	147
c163b5ca	148	blk->aiocb = bdrv_aio_readv(bs, cur_sector, &blk->qiov,
c163b5ca	149	nr_sectors, blk_mig_read_cb, blk);
a55eb92c JK	150
a55eb92c JK	151	if (!blk->aiocb) {
c163b5ca	152	printf("Error reading sector %" PRId64 "\n", cur_sector);
	153	qemu_free(blk->buf);
	154	qemu_free(blk);
	155	return 0;
	156	}
	157
	158	bdrv_reset_dirty(bms->bs, cur_sector, nr_sectors);
d11ecd3d	159	block_mig_state.submitted++;
a55eb92c	160
c163b5ca	161	return (bms->cur_sector >= total_sectors);
	162	}
	163
	164	static int mig_save_device_bulk(QEMUFile f, BlkMigDevState bmds)
a55eb92c	165	{
c163b5ca	166	int len, nr_sectors;
	167	int64_t total_sectors = bmds->total_sectors, cur_sector = 0;
	168	uint8_t *tmp_buf = NULL;
	169	BlockDriverState *bs = bmds->bs;
	170
	171	tmp_buf = qemu_malloc(BLOCK_SIZE);
a55eb92c	172
c163b5ca	173	cur_sector = bmds->cur_sector;
a55eb92c JK	174
	175	if (bmds->shared_base) {
	176	while (cur_sector < bmds->total_sectors &&
	177	!bdrv_is_allocated(bmds->bs, cur_sector,
	178	MAX_IS_ALLOCATED_SEARCH, &nr_sectors)) {
c163b5ca	179	cur_sector += nr_sectors;
	180	}
	181	}
a55eb92c JK	182
a55eb92c JK	183	if (cur_sector >= total_sectors) {
c163b5ca	184	bmds->cur_sector = total_sectors;
	185	qemu_free(tmp_buf);
	186	return 1;
	187	}
a55eb92c	188
d11ecd3d	189	if (cur_sector >= block_mig_state.print_completion) {
c163b5ca	190	printf("Completed %" PRId64 " %%\r", cur_sector * 100 / total_sectors);
c163b5ca	191	fflush(stdout);
d11ecd3d	192	block_mig_state.print_completion +=
6ea44308	193	(BDRV_SECTORS_PER_DIRTY_CHUNK * 10000);
c163b5ca	194	}
a55eb92c	195
6ea44308	196	cur_sector &= ~((int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK - 1);
a55eb92c	197
6ea44308 JK	198	/* we are going to transfer a full block even if it is not allocated */
6ea44308 JK	199	nr_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;
a55eb92c	200
6ea44308	201	if (total_sectors - cur_sector < BDRV_SECTORS_PER_DIRTY_CHUNK) {
c163b5ca	202	nr_sectors = (total_sectors - cur_sector);
c163b5ca	203	}
a55eb92c JK	204
a55eb92c JK	205	if (bdrv_read(bs, cur_sector, tmp_buf, nr_sectors) < 0) {
c163b5ca	206	printf("Error reading sector %" PRId64 "\n", cur_sector);
	207	}
	208
	209	bdrv_reset_dirty(bs, cur_sector, nr_sectors);
a55eb92c JK	210
a55eb92c JK	211	/* sector number and flags */
6ea44308 JK	212	qemu_put_be64(f, (cur_sector << BDRV_SECTOR_BITS)
6ea44308 JK	213	\| BLK_MIG_FLAG_DEVICE_BLOCK);
a55eb92c JK	214
a55eb92c JK	215	/* device name */
c163b5ca	216	len = strlen(bs->device_name);
	217	qemu_put_byte(f, len);
	218	qemu_put_buffer(f, (uint8_t *)bs->device_name, len);
a55eb92c JK	219
	220	qemu_put_buffer(f, tmp_buf, BLOCK_SIZE);
	221
6ea44308	222	bmds->cur_sector = cur_sector + BDRV_SECTORS_PER_DIRTY_CHUNK;
a55eb92c	223
c163b5ca	224	qemu_free(tmp_buf);
a55eb92c	225
c163b5ca	226	return (bmds->cur_sector >= total_sectors);
	227	}
	228
	229	static void send_blk(QEMUFile f, BlkMigBlock blk)
	230	{
	231	int len;
a55eb92c JK	232
a55eb92c JK	233	/* sector number and flags */
6ea44308 JK	234	qemu_put_be64(f, (blk->sector << BDRV_SECTOR_BITS)
6ea44308 JK	235	\| BLK_MIG_FLAG_DEVICE_BLOCK);
a55eb92c JK	236
a55eb92c JK	237	/* device name */
c163b5ca	238	len = strlen(blk->bmds->bs->device_name);
	239	qemu_put_byte(f, len);
	240	qemu_put_buffer(f, (uint8_t *)blk->bmds->bs->device_name, len);
a55eb92c JK	241
a55eb92c JK	242	qemu_put_buffer(f, blk->buf, BLOCK_SIZE);
c163b5ca	243	}
c163b5ca	244
c163b5ca	245	static void set_dirty_tracking(int enable)
	246	{
	247	BlkMigDevState *bmds;
d11ecd3d	248	for (bmds = block_mig_state.bmds_first; bmds != NULL; bmds = bmds->next) {
a55eb92c	249	bdrv_set_dirty_tracking(bmds->bs, enable);
c163b5ca	250	}
c163b5ca	251	}
	252
	253	static void init_blk_migration(QEMUFile *f)
	254	{
	255	BlkMigDevState *pbmds, bmds;
	256	BlockDriverState *bs;
a55eb92c	257
c163b5ca	258	for (bs = bdrv_first; bs != NULL; bs = bs->next) {
a55eb92c	259	if (bs->type == BDRV_TYPE_HD) {
c163b5ca	260	bmds = qemu_mallocz(sizeof(BlkMigDevState));
	261	bmds->bs = bs;
	262	bmds->bulk_completed = 0;
6ea44308	263	bmds->total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
d11ecd3d	264	bmds->shared_base = block_mig_state.shared_base;
a55eb92c JK	265
	266	if (bmds->shared_base) {
	267	printf("Start migration for %s with shared base image\n",
c163b5ca	268	bs->device_name);
	269	} else {
	270	printf("Start full migration for %s\n", bs->device_name);
	271	}
a55eb92c	272
c163b5ca	273	/* insert at the end */
d11ecd3d	274	pbmds = &block_mig_state.bmds_first;
a55eb92c	275	while (*pbmds != NULL) {
c163b5ca	276	pbmds = &(*pbmds)->next;
a55eb92c	277	}
c163b5ca	278	*pbmds = bmds;
c163b5ca	279	}
a55eb92c	280	}
c163b5ca	281	}
	282
	283	static int blk_mig_save_bulked_block(QEMUFile *f, int is_async)
	284	{
	285	BlkMigDevState *bmds;
	286
d11ecd3d	287	for (bmds = block_mig_state.bmds_first; bmds != NULL; bmds = bmds->next) {
a55eb92c JK	288	if (bmds->bulk_completed == 0) {
	289	if (is_async) {
	290	if (mig_read_device_bulk(f, bmds) == 1) {
c163b5ca	291	/* completed bulk section for this device */
	292	bmds->bulk_completed = 1;
	293	}
	294	} else {
a55eb92c	295	if (mig_save_device_bulk(f, bmds) == 1) {
c163b5ca	296	/* completed bulk section for this device */
	297	bmds->bulk_completed = 1;
	298	}
	299	}
	300	return 1;
	301	}
	302	}
a55eb92c	303
c163b5ca	304	/* we reached here means bulk is completed */
c163b5ca	305	return 0;
c163b5ca	306	}
	307
	308	#define MAX_NUM_BLOCKS 4
	309
	310	static void blk_mig_save_dirty_blocks(QEMUFile *f)
	311	{
	312	BlkMigDevState *bmds;
575a58d7	313	uint8_t *buf;
c163b5ca	314	int64_t sector;
c163b5ca	315	int len;
a55eb92c	316
575a58d7 JK	317	buf = qemu_malloc(BLOCK_SIZE);
575a58d7 JK	318
d11ecd3d	319	for (bmds = block_mig_state.bmds_first; bmds != NULL; bmds = bmds->next) {
a55eb92c JK	320	for (sector = 0; sector < bmds->cur_sector;) {
	321	if (bdrv_get_dirty(bmds->bs, sector)) {
	322	if (bdrv_read(bmds->bs, sector, buf,
6ea44308	323	BDRV_SECTORS_PER_DIRTY_CHUNK) < 0) {
a55eb92c	324	/* FIXME: add error handling */
c163b5ca	325	}
a55eb92c JK	326
a55eb92c JK	327	/* sector number and flags */
6ea44308	328	qemu_put_be64(f, (sector << BDRV_SECTOR_BITS)
a55eb92c JK	329	\| BLK_MIG_FLAG_DEVICE_BLOCK);
a55eb92c JK	330
c163b5ca	331	/* device name */
c163b5ca	332	len = strlen(bmds->bs->device_name);
c163b5ca	333	qemu_put_byte(f, len);
c163b5ca	334	qemu_put_buffer(f, (uint8_t *)bmds->bs->device_name, len);
a55eb92c	335
6ea44308	336	qemu_put_buffer(f, buf, BLOCK_SIZE);
a55eb92c JK	337
a55eb92c JK	338	bdrv_reset_dirty(bmds->bs, sector,
6ea44308	339	BDRV_SECTORS_PER_DIRTY_CHUNK);
a55eb92c	340	}
6ea44308	341	sector += BDRV_SECTORS_PER_DIRTY_CHUNK;
c163b5ca	342	}
c163b5ca	343	}
575a58d7 JK	344
575a58d7 JK	345	qemu_free(buf);
c163b5ca	346	}
	347
	348	static void flush_blks(QEMUFile* f)
	349	{
a55eb92c JK	350	BlkMigBlock blk, next;
a55eb92c JK	351
d11ecd3d JK	352	dprintf("%s Enter submitted %d read_done %d transferred %d\n",
	353	__FUNCTION__, block_mig_state.submitted, block_mig_state.read_done,
	354	block_mig_state.transferred);
a55eb92c	355
d11ecd3d	356	for (blk = block_mig_state.first_blk;
a55eb92c JK	357	blk != NULL && !qemu_file_rate_limit(f);
a55eb92c JK	358	blk = next) {
c163b5ca	359	send_blk(f, blk);
a55eb92c JK	360
a55eb92c JK	361	next = blk->next;
c163b5ca	362	qemu_free(blk->buf);
c163b5ca	363	qemu_free(blk);
a55eb92c	364
d11ecd3d JK	365	block_mig_state.read_done--;
	366	block_mig_state.transferred++;
	367	assert(block_mig_state.read_done >= 0);
c163b5ca	368	}
d11ecd3d	369	block_mig_state.first_blk = blk;
a55eb92c	370
d11ecd3d JK	371	if (block_mig_state.first_blk == NULL) {
d11ecd3d JK	372	block_mig_state.last_blk = NULL;
c163b5ca	373	}
c163b5ca	374
d11ecd3d JK	375	dprintf("%s Exit submitted %d read_done %d transferred %d\n", __FUNCTION__,
	376	block_mig_state.submitted, block_mig_state.read_done,
	377	block_mig_state.transferred);
c163b5ca	378	}
	379
	380	static int is_stage2_completed(void)
	381	{
	382	BlkMigDevState *bmds;
a55eb92c	383
d11ecd3d	384	if (block_mig_state.submitted > 0) {
c163b5ca	385	return 0;
c163b5ca	386	}
a55eb92c	387
d11ecd3d	388	for (bmds = block_mig_state.bmds_first; bmds != NULL; bmds = bmds->next) {
a55eb92c	389	if (bmds->bulk_completed == 0) {
c163b5ca	390	return 0;
	391	}
	392	}
a55eb92c	393
c163b5ca	394	return 1;
	395	}
	396
	397	static int block_save_live(QEMUFile f, int stage, void opaque)
	398	{
d11ecd3d JK	399	dprintf("Enter save live stage %d submitted %d transferred %d\n",
d11ecd3d JK	400	stage, block_mig_state.submitted, block_mig_state.transferred);
a55eb92c	401
d11ecd3d	402	if (block_mig_state.blk_enable != 1) {
c163b5ca	403	/* no need to migrate storage */
a55eb92c	404	qemu_put_be64(f, BLK_MIG_FLAG_EOS);
c163b5ca	405	return 1;
c163b5ca	406	}
a55eb92c JK	407
a55eb92c JK	408	if (stage == 1) {
c163b5ca	409	init_blk_migration(f);
a55eb92c	410
c163b5ca	411	/* start track dirty blocks */
c163b5ca	412	set_dirty_tracking(1);
c163b5ca	413	}
	414
	415	flush_blks(f);
a55eb92c	416
c163b5ca	417	/* control the rate of transfer */
d11ecd3d JK	418	while ((block_mig_state.submitted +
d11ecd3d JK	419	block_mig_state.read_done) * BLOCK_SIZE <
a55eb92c JK	420	qemu_file_get_rate_limit(f)) {
	421	if (blk_mig_save_bulked_block(f, 1) == 0) {
	422	/* no more bulk blocks for now */
c163b5ca	423	break;
a55eb92c	424	}
c163b5ca	425	}
a55eb92c	426
c163b5ca	427	flush_blks(f);
a55eb92c JK	428
	429	if (stage == 3) {
	430	while (blk_mig_save_bulked_block(f, 0) != 0) {
	431	/* empty */
	432	}
	433
c163b5ca	434	blk_mig_save_dirty_blocks(f);
a55eb92c	435
c163b5ca	436	/* stop track dirty blocks */
a55eb92c JK	437	set_dirty_tracking(0);
	438
	439	printf("\nBlock migration completed\n");
c163b5ca	440	}
a55eb92c JK	441
	442	qemu_put_be64(f, BLK_MIG_FLAG_EOS);
	443
c163b5ca	444	return ((stage == 2) && is_stage2_completed());
	445	}
	446
	447	static int block_load(QEMUFile f, void opaque, int version_id)
	448	{
	449	int len, flags;
	450	char device_name[256];
	451	int64_t addr;
	452	BlockDriverState *bs;
	453	uint8_t *buf;
a55eb92c	454
c163b5ca	455	do {
c163b5ca	456	addr = qemu_get_be64(f);
a55eb92c	457
6ea44308 JK	458	flags = addr & ~BDRV_SECTOR_MASK;
6ea44308 JK	459	addr >>= BDRV_SECTOR_BITS;
a55eb92c JK	460
a55eb92c JK	461	if (flags & BLK_MIG_FLAG_DEVICE_BLOCK) {
c163b5ca	462	/* get device name */
c163b5ca	463	len = qemu_get_byte(f);
a55eb92c	464
c163b5ca	465	qemu_get_buffer(f, (uint8_t *)device_name, len);
c163b5ca	466	device_name[len] = '\0';
a55eb92c	467
c163b5ca	468	bs = bdrv_find(device_name);
a55eb92c	469
575a58d7 JK	470	buf = qemu_malloc(BLOCK_SIZE);
575a58d7 JK	471
a55eb92c JK	472	qemu_get_buffer(f, buf, BLOCK_SIZE);
a55eb92c JK	473	if (bs != NULL) {
6ea44308	474	bdrv_write(bs, addr, buf, BDRV_SECTORS_PER_DIRTY_CHUNK);
c163b5ca	475	} else {
c163b5ca	476	printf("Error unknown block device %s\n", device_name);
a55eb92c	477	/* FIXME: add error handling */
c163b5ca	478	}
575a58d7 JK	479
575a58d7 JK	480	qemu_free(buf);
a55eb92c	481	} else if (!(flags & BLK_MIG_FLAG_EOS)) {
c163b5ca	482	printf("Unknown flags\n");
a55eb92c	483	/* FIXME: add error handling */
c163b5ca	484	}
a55eb92c JK	485	} while (!(flags & BLK_MIG_FLAG_EOS));
a55eb92c JK	486
c163b5ca	487	return 0;
	488	}
	489
	490	static void block_set_params(int blk_enable, int shared_base, void *opaque)
	491	{
d11ecd3d JK	492	block_mig_state.blk_enable = blk_enable;
d11ecd3d JK	493	block_mig_state.shared_base = shared_base;
a55eb92c	494
c163b5ca	495	/* shared base means that blk_enable = 1 */
d11ecd3d	496	block_mig_state.blk_enable \|= shared_base;
c163b5ca	497	}
c163b5ca	498
c163b5ca	499	void blk_mig_init(void)
a55eb92c	500	{
a55eb92c	501	register_savevm_live("block", 0, 1, block_set_params, block_save_live,
d11ecd3d	502	NULL, block_load, &block_mig_state);
c163b5ca	503	}