Re: [Qemu-devel] [PATCH 2/3 v4] Block live migration

[Anthony Liguori <anthony@codemonkey.ws>]

lirans@il.ibm.com wrote:
> This patch introduces block migration called during live migration. Block 
> are being copied to the destination in an async way. First the code will 
> transfer the whole disk and then transfer all dirty blocks accumulted during 
> the migration.
> Still need to improve transition from the iterative phase of migration to the 
> end phase. For now transition will take place when all blocks transfered once,
> all the dirty blocks will be transfered during the end phase (guest is 
> suspended).
>
>
> diff --git a/Makefile b/Makefile
> index 7d4d75c..3f34459 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -125,6 +125,7 @@ obj-y += qemu-char.o aio.o net-checksum.o savevm.o
>  obj-y += msmouse.o ps2.o
>  obj-y += qdev.o qdev-properties.o
>  obj-y += qint.o qstring.o qdict.o qlist.o qemu-config.o
> +obj-y += block-migration.o
>  
>  obj-$(CONFIG_BRLAPI) += baum.o
>  obj-$(CONFIG_WIN32) += tap-win32.o
> diff --git a/block-migration.c b/block-migration.c
> new file mode 100644
> index 0000000..7ca9d85
> --- /dev/null
> +++ b/block-migration.c
> @@ -0,0 +1,624 @@
> +/*
> + * 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 "qemu-timer.h"
> +#include "block-migration.h"
> +#include <assert.h>
> +#include <pthread.h>
> +
> +#define SECTOR_BITS 9
> +#define SECTOR_SIZE (1 << SECTOR_BITS)
> +#define SECTOR_MASK ~(SECTOR_SIZE - 1);
> +
> +#define SECTORS_PER_BLOCK 8
> +#define BLOCK_SIZE (SECTORS_PER_BLOCK << 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 BlkMigBlock {
> +    uint8_t buf[BLOCK_SIZE];
> +    BlkMigDevState *bmds;
> +    int64_t sector;
> +    struct iovec iov;
> +    QEMUIOVector qiov;
> +    BlockDriverAIOCB *aiocb;
> +    int ret;
> +    struct BlkMigBlock *next;
> +} BlkMigBlock;
> +
> +typedef struct BlkMigState {
> +    int bulk_completed;
> +    int blk_enable;
> +    int shared_base;
> +    int no_dirty;
> +    QEMUFile *load_file;
> +    BlkMigDevState *bmds_first;
> +    QEMUTimer *timer;
> +} BlkMigState;
> +
> +static BlkMigState block_mig_state;
> +
> +static BlkMigBlock *first_blk = NULL;
> +static BlkMigBlock *last_blk = NULL;
> +
> +static int submitted = 0;
> +static int read_done = 0;
> +static int transferred = 0;
>   

It seems to me that this could all get moved to a global state (that's 
dynamically allocated).

> +
> +static int64_t print_completion = 0;
> +static void mark_clean(BlkMigDevState *bmds, int64_t sector, 
> +		       int sector_num);
> +static int is_dirty(BlkMigDevState *bmds, int64_t sector);
> +
> +static void blk_mig_read_cb(void *opaque, int ret)
> +{
> +    BlkMigBlock *blk = opaque;
> +  
> +    blk->ret = ret;
> +  
> +    /* insert at the end */
> +    if(last_blk == NULL) {
> +	first_blk = last_blk = blk;
> +    } else {
> +	last_blk->next = blk;
> +	last_blk = blk;
> +    }
>   

CodingStyle is off.

> +    /* Device name */
> +    qemu_put_be64(f,(cur_sector << SECTOR_BITS) | BLK_MIG_FLAG_DEVICE_BLOCK);
> +  
> +    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);
>   

Would be good to do a simple check for an all zero block.  That would 
help the initial transfer.


Regards,

Anthony Liguori