[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-block] [PATCH v10 10/12] migration: add postcopy migration of
|
From: |
Dr. David Alan Gilbert |
|
Subject: |
Re: [Qemu-block] [PATCH v10 10/12] migration: add postcopy migration of dirty bitmaps |
|
Date: |
Mon, 12 Mar 2018 16:09:02 +0000 |
|
User-agent: |
Mutt/1.9.2 (2017-12-15) |
* Vladimir Sementsov-Ogievskiy (address@hidden) wrote:
> Postcopy migration of dirty bitmaps. Only named dirty bitmaps are migrated.
>
> If destination qemu is already containing a dirty bitmap with the same name
> as a migrated bitmap (for the same node), then, if their granularities are
> the same the migration will be done, otherwise the error will be generated.
>
> If destination qemu doesn't contain such bitmap it will be created.
>
> Signed-off-by: Vladimir Sementsov-Ogievskiy <address@hidden>
> ---
> include/migration/misc.h | 3 +
> migration/migration.h | 3 +
> migration/block-dirty-bitmap.c | 737
> +++++++++++++++++++++++++++++++++++++++++
> migration/migration.c | 5 +
> migration/savevm.c | 2 +
> vl.c | 1 +
> migration/Makefile.objs | 1 +
> migration/trace-events | 14 +
> 8 files changed, 766 insertions(+)
> create mode 100644 migration/block-dirty-bitmap.c
>
> diff --git a/include/migration/misc.h b/include/migration/misc.h
> index 77fd4f587c..4ebf24c6c2 100644
> --- a/include/migration/misc.h
> +++ b/include/migration/misc.h
> @@ -56,4 +56,7 @@ bool migration_has_failed(MigrationState *);
> bool migration_in_postcopy_after_devices(MigrationState *);
> void migration_global_dump(Monitor *mon);
>
> +/* migration/block-dirty-bitmap.c */
> +void dirty_bitmap_mig_init(void);
> +
> #endif
> diff --git a/migration/migration.h b/migration/migration.h
> index 861cdfaa96..79f72b7e50 100644
> --- a/migration/migration.h
> +++ b/migration/migration.h
> @@ -233,4 +233,7 @@ void migrate_send_rp_pong(MigrationIncomingState *mis,
> void migrate_send_rp_req_pages(MigrationIncomingState *mis, const char*
> rbname,
> ram_addr_t start, size_t len);
>
> +void dirty_bitmap_mig_before_vm_start(void);
> +void init_dirty_bitmap_incoming_migration(void);
> +
> #endif
> diff --git a/migration/block-dirty-bitmap.c b/migration/block-dirty-bitmap.c
> new file mode 100644
> index 0000000000..5b41f7140d
> --- /dev/null
> +++ b/migration/block-dirty-bitmap.c
> @@ -0,0 +1,737 @@
> +/*
> + * Block dirty bitmap postcopy migration
> + *
> + * Copyright IBM, Corp. 2009
> + * Copyright (c) 2016-2017 Virtuozzo International GmbH. All rights reserved.
> + *
> + * Authors:
> + * Liran Schour <address@hidden>
> + * Vladimir Sementsov-Ogievskiy <address@hidden>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2. See
> + * the COPYING file in the top-level directory.
> + * This file is derived from migration/block.c, so it's author and IBM
> copyright
> + * are here, although content is quite different.
> + *
> + * Contributions after 2012-01-13 are licensed under the terms of the
> + * GNU GPL, version 2 or (at your option) any later version.
> + *
> + * ***
> + *
> + * Here postcopy migration of dirty bitmaps is realized. Only QMP-addressable
> + * bitmaps are migrated.
> + *
> + * Bitmap migration implies creating bitmap with the same name and
> granularity
> + * in destination QEMU. If the bitmap with the same name (for the same node)
> + * already exists on destination an error will be generated.
> + *
> + * format of migration:
> + *
> + * # Header (shared for different chunk types)
> + * 1, 2 or 4 bytes: flags (see qemu_{put,put}_flags)
> + * [ 1 byte: node name size ] \ flags & DEVICE_NAME
> + * [ n bytes: node name ] /
> + * [ 1 byte: bitmap name size ] \ flags & BITMAP_NAME
> + * [ n bytes: bitmap name ] /
> + *
> + * # Start of bitmap migration (flags & START)
> + * header
> + * be64: granularity
> + * 1 byte: bitmap flags (corresponds to BdrvDirtyBitmap)
> + * bit 0 - bitmap is enabled
> + * bit 1 - bitmap is persistent
> + * bit 2 - bitmap is autoloading
> + * bits 3-7 - reserved, must be zero
> + *
> + * # Complete of bitmap migration (flags & COMPLETE)
> + * header
> + *
> + * # Data chunk of bitmap migration
> + * header
> + * be64: start sector
> + * be32: number of sectors
> + * [ be64: buffer size ] \ ! (flags & ZEROES)
> + * [ n bytes: buffer ] /
> + *
> + * The last chunk in stream should contain flags & EOS. The chunk may skip
> + * device and/or bitmap names, assuming them to be the same with the previous
> + * chunk.
> + */
> +
> +#include "qemu/osdep.h"
> +#include "block/block.h"
> +#include "block/block_int.h"
> +#include "sysemu/block-backend.h"
> +#include "qemu/main-loop.h"
> +#include "qemu/error-report.h"
> +#include "migration/misc.h"
> +#include "migration/migration.h"
> +#include "migration/qemu-file.h"
> +#include "migration/vmstate.h"
> +#include "migration/register.h"
> +#include "qemu/hbitmap.h"
> +#include "sysemu/sysemu.h"
> +#include "qemu/cutils.h"
> +#include "qapi/error.h"
> +#include "trace.h"
> +
> +#define CHUNK_SIZE (1 << 10)
> +
> +/* Flags occupy one, two or four bytes (Big Endian). The size is determined
> as
> + * follows:
> + * in first (most significant) byte bit 8 is clear --> one byte
> + * in first byte bit 8 is set --> two or four bytes, depending on second
> + * byte:
> + * | in second byte bit 8 is clear --> two bytes
> + * | in second byte bit 8 is set --> four bytes
> + */
> +#define DIRTY_BITMAP_MIG_FLAG_EOS 0x01
> +#define DIRTY_BITMAP_MIG_FLAG_ZEROES 0x02
> +#define DIRTY_BITMAP_MIG_FLAG_BITMAP_NAME 0x04
> +#define DIRTY_BITMAP_MIG_FLAG_DEVICE_NAME 0x08
> +#define DIRTY_BITMAP_MIG_FLAG_START 0x10
> +#define DIRTY_BITMAP_MIG_FLAG_COMPLETE 0x20
> +#define DIRTY_BITMAP_MIG_FLAG_BITS 0x40
> +
> +#define DIRTY_BITMAP_MIG_EXTRA_FLAGS 0x80
> +
> +#define DIRTY_BITMAP_MIG_START_FLAG_ENABLED 0x01
> +#define DIRTY_BITMAP_MIG_START_FLAG_PERSISTENT 0x02
> +/* 0x04 was "AUTOLOAD" flags on elder versions, no it is ignored */
> +#define DIRTY_BITMAP_MIG_START_FLAG_RESERVED_MASK 0xf8
> +
> +typedef struct DirtyBitmapMigBitmapState {
> + /* Written during setup phase. */
> + BlockDriverState *bs;
> + const char *node_name;
> + BdrvDirtyBitmap *bitmap;
> + uint64_t total_sectors;
> + uint64_t sectors_per_chunk;
> + QSIMPLEQ_ENTRY(DirtyBitmapMigBitmapState) entry;
> + uint8_t flags;
> +
> + /* For bulk phase. */
> + bool bulk_completed;
> + uint64_t cur_sector;
> +} DirtyBitmapMigBitmapState;
> +
> +typedef struct DirtyBitmapMigState {
> + QSIMPLEQ_HEAD(dbms_list, DirtyBitmapMigBitmapState) dbms_list;
> +
> + bool bulk_completed;
> + bool no_bitmaps;
> +
> + /* for send_bitmap_bits() */
> + BlockDriverState *prev_bs;
> + BdrvDirtyBitmap *prev_bitmap;
> +} DirtyBitmapMigState;
> +
> +typedef struct DirtyBitmapLoadState {
> + uint32_t flags;
> + char node_name[256];
> + char bitmap_name[256];
> + BlockDriverState *bs;
> + BdrvDirtyBitmap *bitmap;
> +} DirtyBitmapLoadState;
> +
> +static DirtyBitmapMigState dirty_bitmap_mig_state;
> +
> +typedef struct DirtyBitmapLoadBitmapState {
> + BlockDriverState *bs;
> + BdrvDirtyBitmap *bitmap;
> + bool migrated;
> +} DirtyBitmapLoadBitmapState;
> +static GSList *enabled_bitmaps;
> +QemuMutex finish_lock;
> +
> +void init_dirty_bitmap_incoming_migration(void)
> +{
> + qemu_mutex_init(&finish_lock);
> +}
> +
> +static uint32_t qemu_get_bitmap_flags(QEMUFile *f)
> +{
> + uint8_t flags = qemu_get_byte(f);
> + if (flags & DIRTY_BITMAP_MIG_EXTRA_FLAGS) {
> + flags = flags << 8 | qemu_get_byte(f);
> + if (flags & DIRTY_BITMAP_MIG_EXTRA_FLAGS) {
> + flags = flags << 16 | qemu_get_be16(f);
> + }
> + }
> +
> + return flags;
> +}
> +
> +static void qemu_put_bitmap_flags(QEMUFile *f, uint32_t flags)
> +{
> + /* The code currently do not send flags more than one byte */
> + assert(!(flags & (0xffffff00 | DIRTY_BITMAP_MIG_EXTRA_FLAGS)));
> +
> + qemu_put_byte(f, flags);
> +}
> +
> +static void send_bitmap_header(QEMUFile *f, DirtyBitmapMigBitmapState *dbms,
> + uint32_t additional_flags)
> +{
> + BlockDriverState *bs = dbms->bs;
> + BdrvDirtyBitmap *bitmap = dbms->bitmap;
> + uint32_t flags = additional_flags;
> + trace_send_bitmap_header_enter();
> +
> + if (bs != dirty_bitmap_mig_state.prev_bs) {
> + dirty_bitmap_mig_state.prev_bs = bs;
> + flags |= DIRTY_BITMAP_MIG_FLAG_DEVICE_NAME;
> + }
> +
> + if (bitmap != dirty_bitmap_mig_state.prev_bitmap) {
> + dirty_bitmap_mig_state.prev_bitmap = bitmap;
> + flags |= DIRTY_BITMAP_MIG_FLAG_BITMAP_NAME;
> + }
> +
> + qemu_put_bitmap_flags(f, flags);
> +
> + if (flags & DIRTY_BITMAP_MIG_FLAG_DEVICE_NAME) {
> + qemu_put_counted_string(f, dbms->node_name);
> + }
> +
> + if (flags & DIRTY_BITMAP_MIG_FLAG_BITMAP_NAME) {
> + qemu_put_counted_string(f, bdrv_dirty_bitmap_name(bitmap));
> + }
> +}
> +
> +static void send_bitmap_start(QEMUFile *f, DirtyBitmapMigBitmapState *dbms)
> +{
> + send_bitmap_header(f, dbms, DIRTY_BITMAP_MIG_FLAG_START);
> + qemu_put_be32(f, bdrv_dirty_bitmap_granularity(dbms->bitmap));
> + qemu_put_byte(f, dbms->flags);
> +}
> +
> +static void send_bitmap_complete(QEMUFile *f, DirtyBitmapMigBitmapState
> *dbms)
> +{
> + send_bitmap_header(f, dbms, DIRTY_BITMAP_MIG_FLAG_COMPLETE);
> +}
> +
> +static void send_bitmap_bits(QEMUFile *f, DirtyBitmapMigBitmapState *dbms,
> + uint64_t start_sector, uint32_t nr_sectors)
> +{
> + /* align for buffer_is_zero() */
> + uint64_t align = 4 * sizeof(long);
> + uint64_t unaligned_size =
> + bdrv_dirty_bitmap_serialization_size(
> + dbms->bitmap, start_sector << BDRV_SECTOR_BITS,
> + (uint64_t)nr_sectors << BDRV_SECTOR_BITS);
> + uint64_t buf_size = QEMU_ALIGN_UP(unaligned_size, align);
> + uint8_t *buf = g_malloc0(buf_size);
> + uint32_t flags = DIRTY_BITMAP_MIG_FLAG_BITS;
> +
> + bdrv_dirty_bitmap_serialize_part(
> + dbms->bitmap, buf, start_sector << BDRV_SECTOR_BITS,
> + (uint64_t)nr_sectors << BDRV_SECTOR_BITS);
> +
> + if (buffer_is_zero(buf, buf_size)) {
> + g_free(buf);
> + buf = NULL;
> + flags |= DIRTY_BITMAP_MIG_FLAG_ZEROES;
> + }
> +
> + trace_send_bitmap_bits(flags, start_sector, nr_sectors, buf_size);
> +
> + send_bitmap_header(f, dbms, flags);
> +
> + qemu_put_be64(f, start_sector);
> + qemu_put_be32(f, nr_sectors);
> +
> + /* if a block is zero we need to flush here since the network
> + * bandwidth is now a lot higher than the storage device bandwidth.
> + * thus if we queue zero blocks we slow down the migration. */
> + if (flags & DIRTY_BITMAP_MIG_FLAG_ZEROES) {
> + qemu_fflush(f);
> + } else {
> + qemu_put_be64(f, buf_size);
> + qemu_put_buffer(f, buf, buf_size);
> + }
> +
> + g_free(buf);
> +}
> +
> +/* Called with iothread lock taken. */
> +static void dirty_bitmap_mig_cleanup(void)
> +{
> + DirtyBitmapMigBitmapState *dbms;
> +
> + while ((dbms = QSIMPLEQ_FIRST(&dirty_bitmap_mig_state.dbms_list)) !=
> NULL) {
> + QSIMPLEQ_REMOVE_HEAD(&dirty_bitmap_mig_state.dbms_list, entry);
> + bdrv_dirty_bitmap_set_qmp_locked(dbms->bitmap, false);
> + bdrv_unref(dbms->bs);
> + g_free(dbms);
> + }
> +}
> +
> +/* Called with iothread lock taken. */
> +static int init_dirty_bitmap_migration(void)
> +{
> + BlockDriverState *bs;
> + BdrvDirtyBitmap *bitmap;
> + DirtyBitmapMigBitmapState *dbms;
> + BdrvNextIterator it;
> +
> + dirty_bitmap_mig_state.bulk_completed = false;
> + dirty_bitmap_mig_state.prev_bs = NULL;
> + dirty_bitmap_mig_state.prev_bitmap = NULL;
> + dirty_bitmap_mig_state.no_bitmaps = false;
> +
> + for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
> + const char *drive_name = bdrv_get_device_or_node_name(bs);
> +
> + /* skip automatically inserted nodes */
> + while (bs && bs->drv && bs->implicit) {
> + bs = backing_bs(bs);
> + }
> +
> + for (bitmap = bdrv_dirty_bitmap_next(bs, NULL); bitmap;
> + bitmap = bdrv_dirty_bitmap_next(bs, bitmap))
> + {
> + if (!bdrv_dirty_bitmap_name(bitmap)) {
> + continue;
> + }
> +
> + if (drive_name == NULL) {
> + error_report("Found bitmap '%s' in unnamed node %p. It can't
> "
> + "be migrated", bdrv_dirty_bitmap_name(bitmap),
> bs);
> + goto fail;
> + }
> +
> + if (bdrv_dirty_bitmap_frozen(bitmap)) {
> + error_report("Can't migrate frozen dirty bitmap: '%s",
> + bdrv_dirty_bitmap_name(bitmap));
> + goto fail;
> + }
> +
> + if (bdrv_dirty_bitmap_qmp_locked(bitmap)) {
> + error_report("Can't migrate locked dirty bitmap: '%s",
> + bdrv_dirty_bitmap_name(bitmap));
> + goto fail;
> + }
> +
> + bdrv_ref(bs);
> + bdrv_dirty_bitmap_set_qmp_locked(bitmap, true);
> +
> + dbms = g_new0(DirtyBitmapMigBitmapState, 1);
> + dbms->bs = bs;
> + dbms->node_name = drive_name;
> + dbms->bitmap = bitmap;
> + dbms->total_sectors = bdrv_nb_sectors(bs);
> + dbms->sectors_per_chunk = CHUNK_SIZE * 8 *
> + bdrv_dirty_bitmap_granularity(bitmap) >> BDRV_SECTOR_BITS;
> + if (bdrv_dirty_bitmap_enabled(bitmap)) {
> + dbms->flags |= DIRTY_BITMAP_MIG_START_FLAG_ENABLED;
> + }
> + if (bdrv_dirty_bitmap_get_persistance(bitmap)) {
> + dbms->flags |= DIRTY_BITMAP_MIG_START_FLAG_PERSISTENT;
> + }
> +
> + QSIMPLEQ_INSERT_TAIL(&dirty_bitmap_mig_state.dbms_list,
> + dbms, entry);
> + }
> + }
> +
> + /* unset persistance here, to not roll back it */
> + QSIMPLEQ_FOREACH(dbms, &dirty_bitmap_mig_state.dbms_list, entry) {
> + bdrv_dirty_bitmap_set_persistance(dbms->bitmap, false);
> + }
> +
> + if (QSIMPLEQ_EMPTY(&dirty_bitmap_mig_state.dbms_list)) {
> + dirty_bitmap_mig_state.no_bitmaps = true;
> + }
> +
> + return 0;
> +
> +fail:
> + dirty_bitmap_mig_cleanup();
> +
> + return -1;
> +}
> +
> +/* Called with no lock taken. */
> +static void bulk_phase_send_chunk(QEMUFile *f, DirtyBitmapMigBitmapState
> *dbms)
> +{
> + uint32_t nr_sectors = MIN(dbms->total_sectors - dbms->cur_sector,
> + dbms->sectors_per_chunk);
> +
> + send_bitmap_bits(f, dbms, dbms->cur_sector, nr_sectors);
> +
> + dbms->cur_sector += nr_sectors;
> + if (dbms->cur_sector >= dbms->total_sectors) {
> + dbms->bulk_completed = true;
> + }
> +}
> +
> +/* Called with no lock taken. */
> +static void bulk_phase(QEMUFile *f, bool limit)
> +{
> + DirtyBitmapMigBitmapState *dbms;
> +
> + QSIMPLEQ_FOREACH(dbms, &dirty_bitmap_mig_state.dbms_list, entry) {
> + while (!dbms->bulk_completed) {
> + bulk_phase_send_chunk(f, dbms);
> + if (limit && qemu_file_rate_limit(f)) {
> + return;
> + }
> + }
> + }
> +
> + dirty_bitmap_mig_state.bulk_completed = true;
> +}
> +
> +/* for SaveVMHandlers */
> +static void dirty_bitmap_save_cleanup(void *opaque)
> +{
> + dirty_bitmap_mig_cleanup();
> +}
> +
> +static int dirty_bitmap_save_iterate(QEMUFile *f, void *opaque)
> +{
> + trace_dirty_bitmap_save_iterate(migration_in_postcopy());
> +
> + if (migration_in_postcopy() && !dirty_bitmap_mig_state.bulk_completed) {
> + bulk_phase(f, true);
> + }
> +
> + qemu_put_bitmap_flags(f, DIRTY_BITMAP_MIG_FLAG_EOS);
> +
> + return dirty_bitmap_mig_state.bulk_completed;
> +}
> +
> +/* Called with iothread lock taken. */
> +
> +static int dirty_bitmap_save_complete(QEMUFile *f, void *opaque)
> +{
> + DirtyBitmapMigBitmapState *dbms;
> + trace_dirty_bitmap_save_complete_enter();
> +
> + if (!dirty_bitmap_mig_state.bulk_completed) {
> + bulk_phase(f, false);
> + }
> +
> + QSIMPLEQ_FOREACH(dbms, &dirty_bitmap_mig_state.dbms_list, entry) {
> + send_bitmap_complete(f, dbms);
> + }
> +
> + qemu_put_bitmap_flags(f, DIRTY_BITMAP_MIG_FLAG_EOS);
> +
> + trace_dirty_bitmap_save_complete_finish();
> +
> + dirty_bitmap_mig_cleanup();
> + return 0;
> +}
> +
> +static void dirty_bitmap_save_pending(QEMUFile *f, void *opaque,
> + uint64_t max_size,
> + uint64_t *res_precopy_only,
> + uint64_t *res_compatible,
> + uint64_t *res_postcopy_only)
> +{
> + DirtyBitmapMigBitmapState *dbms;
> + uint64_t pending = 0;
> +
> + qemu_mutex_lock_iothread();
> +
> + QSIMPLEQ_FOREACH(dbms, &dirty_bitmap_mig_state.dbms_list, entry) {
> + uint64_t gran = bdrv_dirty_bitmap_granularity(dbms->bitmap);
> + uint64_t sectors = dbms->bulk_completed ? 0 :
> + dbms->total_sectors - dbms->cur_sector;
> +
> + pending += DIV_ROUND_UP(sectors * BDRV_SECTOR_SIZE, gran);
> + }
> +
> + qemu_mutex_unlock_iothread();
> +
> + trace_dirty_bitmap_save_pending(pending, max_size);
> +
> + *res_postcopy_only += pending;
> +}
> +
> +/* First occurrence of this bitmap. It should be created if doesn't exist */
> +static int dirty_bitmap_load_start(QEMUFile *f, DirtyBitmapLoadState *s)
> +{
> + Error *local_err = NULL;
> + uint32_t granularity = qemu_get_be32(f);
> + uint8_t flags = qemu_get_byte(f);
> +
> + if (s->bitmap) {
> + error_report("Bitmap with the same name ('%s') already exists on "
> + "destination", bdrv_dirty_bitmap_name(s->bitmap));
> + return -EINVAL;
> + } else {
> + s->bitmap = bdrv_create_dirty_bitmap(s->bs, granularity,
> + s->bitmap_name, &local_err);
> + if (!s->bitmap) {
> + error_report_err(local_err);
> + return -EINVAL;
> + }
> + }
> +
> + if (flags & DIRTY_BITMAP_MIG_START_FLAG_RESERVED_MASK) {
> + error_report("Unknown flags in migrated dirty bitmap header: %x",
> + flags);
> + return -EINVAL;
> + }
> +
> + if (flags & DIRTY_BITMAP_MIG_START_FLAG_PERSISTENT) {
> + bdrv_dirty_bitmap_set_persistance(s->bitmap, true);
> + }
> +
> + bdrv_disable_dirty_bitmap(s->bitmap);
> + if (flags & DIRTY_BITMAP_MIG_START_FLAG_ENABLED) {
> + DirtyBitmapLoadBitmapState *b;
> +
> + bdrv_dirty_bitmap_create_successor(s->bs, s->bitmap, &local_err);
> + if (local_err) {
> + error_report_err(local_err);
> + return -EINVAL;
> + }
> +
> + b = g_new(DirtyBitmapLoadBitmapState, 1);
> + b->bs = s->bs;
> + b->bitmap = s->bitmap;
> + b->migrated = false;
> + enabled_bitmaps = g_slist_prepend(enabled_bitmaps, b);
> + }
> +
> + return 0;
> +}
> +
> +void dirty_bitmap_mig_before_vm_start(void)
> +{
> + GSList *item;
> +
> + qemu_mutex_lock(&finish_lock);
> +
> + for (item = enabled_bitmaps; item; item = g_slist_next(item)) {
> + DirtyBitmapLoadBitmapState *b = item->data;
> +
> + if (b->migrated) {
> + bdrv_enable_dirty_bitmap(b->bitmap);
> + } else {
> + bdrv_dirty_bitmap_enable_successor(b->bitmap);
> + }
> +
> + g_free(b);
> + }
> +
> + g_slist_free(enabled_bitmaps);
> + enabled_bitmaps = NULL;
> +
> + qemu_mutex_unlock(&finish_lock);
> +}
> +
> +static void dirty_bitmap_load_complete(QEMUFile *f, DirtyBitmapLoadState *s)
> +{
> + GSList *item;
> + trace_dirty_bitmap_load_complete();
> + bdrv_dirty_bitmap_deserialize_finish(s->bitmap);
> +
> + qemu_mutex_lock(&finish_lock);
> +
> + for (item = enabled_bitmaps; item; item = g_slist_next(item)) {
> + DirtyBitmapLoadBitmapState *b = item->data;
> +
> + if (b->bitmap == s->bitmap) {
> + b->migrated = true;
> + break;
> + }
> + }
> +
> + if (bdrv_dirty_bitmap_frozen(s->bitmap)) {
> + bdrv_dirty_bitmap_lock(s->bitmap);
> + if (enabled_bitmaps == NULL) {
> + /* in postcopy */
> + bdrv_reclaim_dirty_bitmap_locked(s->bs, s->bitmap, &error_abort);
> + bdrv_enable_dirty_bitmap(s->bitmap);
> + } else {
> + /* target not started, successor must be empty */
> + int64_t count = bdrv_get_dirty_count(s->bitmap);
> + BdrvDirtyBitmap *ret = bdrv_reclaim_dirty_bitmap_locked(s->bs,
> +
> s->bitmap,
> + NULL);
> + /* bdrv_reclaim_dirty_bitmap can fail only on no successor (it
> + * must be) or on merge fail, but merge can't fail when second
> + * bitmap is empty
> + */
> + assert(ret == s->bitmap &&
> + count == bdrv_get_dirty_count(s->bitmap));
> + }
> + bdrv_dirty_bitmap_unlock(s->bitmap);
> + }
> +
> + qemu_mutex_unlock(&finish_lock);
> +}
> +
> +static int dirty_bitmap_load_bits(QEMUFile *f, DirtyBitmapLoadState *s)
> +{
> + uint64_t first_byte = qemu_get_be64(f) << BDRV_SECTOR_BITS;
> + uint64_t nr_bytes = (uint64_t)qemu_get_be32(f) << BDRV_SECTOR_BITS;
> + trace_dirty_bitmap_load_bits_enter(first_byte >> BDRV_SECTOR_BITS,
> + nr_bytes >> BDRV_SECTOR_BITS);
> +
> + if (s->flags & DIRTY_BITMAP_MIG_FLAG_ZEROES) {
> + trace_dirty_bitmap_load_bits_zeroes();
> + bdrv_dirty_bitmap_deserialize_zeroes(s->bitmap, first_byte, nr_bytes,
> + false);
> + } else {
> + uint8_t *buf;
> + uint64_t buf_size = qemu_get_be64(f);
> + uint64_t needed_size =
> + bdrv_dirty_bitmap_serialization_size(s->bitmap,
> + first_byte, nr_bytes);
> +
> + if (needed_size > buf_size) {
> + error_report("Error: Migrated bitmap granularity doesn't "
> + "match the destination bitmap '%s' granularity",
> + bdrv_dirty_bitmap_name(s->bitmap));
> + return -EINVAL;
> + }
> +
> + buf = g_malloc(buf_size);
> + qemu_get_buffer(f, buf, buf_size);
Two things to be careful of there:
a) buf_size is read from the stream, so be careful about whether you
trust it could be a crazy value
b) You don't check the return value of qemu_get_buffer; which actually
we're pretty bad at in lots of places, but it's probably a good idea
before trying to deserialise it.
> + bdrv_dirty_bitmap_deserialize_part(s->bitmap, buf, first_byte,
> nr_bytes,
> + false);
> + g_free(buf);
> + }
> +
> + return 0;
> +}
> +
> +static int dirty_bitmap_load_header(QEMUFile *f, DirtyBitmapLoadState *s)
> +{
> + Error *local_err = NULL;
> + bool nothing;
> + s->flags = qemu_get_bitmap_flags(f);
> + trace_dirty_bitmap_load_header(s->flags);
> +
> + nothing = s->flags == (s->flags & DIRTY_BITMAP_MIG_FLAG_EOS);
> +
> + if (s->flags & DIRTY_BITMAP_MIG_FLAG_DEVICE_NAME) {
> + if (!qemu_get_counted_string(f, s->node_name)) {
> + error_report("Unable to read node name string");
> + return -EINVAL;
> + }
> + s->bs = bdrv_lookup_bs(s->node_name, s->node_name, &local_err);
> + if (!s->bs) {
> + error_report_err(local_err);
> + return -EINVAL;
> + }
> + } else if (!s->bs && !nothing) {
> + error_report("Error: block device name is not set");
> + return -EINVAL;
> + }
> +
> + if (s->flags & DIRTY_BITMAP_MIG_FLAG_BITMAP_NAME) {
> + if (!qemu_get_counted_string(f, s->bitmap_name)) {
> + error_report("Unable to read bitmap name string");
> + return -EINVAL;
> + }
> + s->bitmap = bdrv_find_dirty_bitmap(s->bs, s->bitmap_name);
> +
> + /* bitmap may be NULL here, it wouldn't be an error if it is the
> + * first occurrence of the bitmap */
> + if (!s->bitmap && !(s->flags & DIRTY_BITMAP_MIG_FLAG_START)) {
> + error_report("Error: unknown dirty bitmap "
> + "'%s' for block device '%s'",
> + s->bitmap_name, s->node_name);
> + return -EINVAL;
> + }
> + } else if (!s->bitmap && !nothing) {
> + error_report("Error: block device name is not set");
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +static int dirty_bitmap_load(QEMUFile *f, void *opaque, int version_id)
> +{
> + static DirtyBitmapLoadState s;
> + int ret = 0;
> +
> + trace_dirty_bitmap_load_enter();
> +
> + if (version_id != 1) {
> + return -EINVAL;
> + }
> +
> + do {
> + ret = dirty_bitmap_load_header(f, &s);
> +
> + if (s.flags & DIRTY_BITMAP_MIG_FLAG_START) {
> + ret = dirty_bitmap_load_start(f, &s);
> + } else if (s.flags & DIRTY_BITMAP_MIG_FLAG_COMPLETE) {
> + dirty_bitmap_load_complete(f, &s);
> + } else if (s.flags & DIRTY_BITMAP_MIG_FLAG_BITS) {
> + ret = dirty_bitmap_load_bits(f, &s);
> + }
> +
> + if (!ret) {
> + ret = qemu_file_get_error(f);
> + }
> +
> + if (ret) {
> + return ret;
> + }
> + } while (!(s.flags & DIRTY_BITMAP_MIG_FLAG_EOS));
> +
> + trace_dirty_bitmap_load_success();
> + return 0;
> +}
> +
> +static int dirty_bitmap_save_setup(QEMUFile *f, void *opaque)
> +{
> + DirtyBitmapMigBitmapState *dbms = NULL;
> + if (init_dirty_bitmap_migration() < 0) {
> + return -1;
> + }
> +
> + QSIMPLEQ_FOREACH(dbms, &dirty_bitmap_mig_state.dbms_list, entry) {
> + send_bitmap_start(f, dbms);
> + }
> + qemu_put_bitmap_flags(f, DIRTY_BITMAP_MIG_FLAG_EOS);
> +
> + return 0;
> +}
> +
> +static bool dirty_bitmap_is_active(void *opaque)
> +{
> + return migrate_dirty_bitmaps() && !dirty_bitmap_mig_state.no_bitmaps;
> +}
> +
> +static bool dirty_bitmap_is_active_iterate(void *opaque)
> +{
> + return dirty_bitmap_is_active(opaque) && !runstate_is_running();
> +}
> +
> +static bool dirty_bitmap_has_postcopy(void *opaque)
> +{
> + return true;
> +}
> +
> +static SaveVMHandlers savevm_dirty_bitmap_handlers = {
> + .save_setup = dirty_bitmap_save_setup,
> + .save_live_complete_postcopy = dirty_bitmap_save_complete,
> + .save_live_complete_precopy = dirty_bitmap_save_complete,
> + .has_postcopy = dirty_bitmap_has_postcopy,
> + .save_live_pending = dirty_bitmap_save_pending,
> + .save_live_iterate = dirty_bitmap_save_iterate,
> + .is_active_iterate = dirty_bitmap_is_active_iterate,
> + .load_state = dirty_bitmap_load,
> + .save_cleanup = dirty_bitmap_save_cleanup,
> + .is_active = dirty_bitmap_is_active,
> +};
> +
> +void dirty_bitmap_mig_init(void)
> +{
> + QSIMPLEQ_INIT(&dirty_bitmap_mig_state.dbms_list);
> +
> + register_savevm_live(NULL, "dirty-bitmap", 0, 1,
> + &savevm_dirty_bitmap_handlers,
> + &dirty_bitmap_mig_state);
> +}
> diff --git a/migration/migration.c b/migration/migration.c
> index d09f53d6c3..4128e05d7c 100644
> --- a/migration/migration.c
> +++ b/migration/migration.c
> @@ -155,6 +155,9 @@ MigrationIncomingState
> *migration_incoming_get_current(void)
> memset(&mis_current, 0, sizeof(MigrationIncomingState));
> qemu_mutex_init(&mis_current.rp_mutex);
> qemu_event_init(&mis_current.main_thread_load_event, false);
> +
> + init_dirty_bitmap_incoming_migration();
> +
You might want to consider if that's better in vl.c near where
ram_mig_init() is, OR whether there should be a call in
migratation_incoming_state_destroy to clean it up.
(Although I doubt the cases where the destroy happens are interesting
for postcopy bitmaps).
> once = true;
> }
> return &mis_current;
> @@ -297,6 +300,8 @@ static void process_incoming_migration_bh(void *opaque)
> state, we need to obey autostart. Any other state is set with
> runstate_set. */
>
> + dirty_bitmap_mig_before_vm_start();
> +
> if (!global_state_received() ||
> global_state_get_runstate() == RUN_STATE_RUNNING) {
> if (autostart) {
> diff --git a/migration/savevm.c b/migration/savevm.c
> index e5d557458e..93b339646b 100644
> --- a/migration/savevm.c
> +++ b/migration/savevm.c
> @@ -1673,6 +1673,8 @@ static void loadvm_postcopy_handle_run_bh(void *opaque)
>
> trace_loadvm_postcopy_handle_run_vmstart();
>
> + dirty_bitmap_mig_before_vm_start();
> +
> if (autostart) {
> /* Hold onto your hats, starting the CPU */
> vm_start();
> diff --git a/vl.c b/vl.c
> index e517a8d995..0ef3f2b5a2 100644
> --- a/vl.c
> +++ b/vl.c
> @@ -4514,6 +4514,7 @@ int main(int argc, char **argv, char **envp)
>
> blk_mig_init();
> ram_mig_init();
> + dirty_bitmap_mig_init();
>
> /* If the currently selected machine wishes to override the units-per-bus
> * property of its default HBA interface type, do so now. */
> diff --git a/migration/Makefile.objs b/migration/Makefile.objs
> index 99e038024d..c83ec47ba8 100644
> --- a/migration/Makefile.objs
> +++ b/migration/Makefile.objs
> @@ -6,6 +6,7 @@ common-obj-y += qemu-file.o global_state.o
> common-obj-y += qemu-file-channel.o
> common-obj-y += xbzrle.o postcopy-ram.o
> common-obj-y += qjson.o
> +common-obj-y += block-dirty-bitmap.o
>
> common-obj-$(CONFIG_RDMA) += rdma.o
>
> diff --git a/migration/trace-events b/migration/trace-events
> index a04fffb877..e9eb8078d4 100644
> --- a/migration/trace-events
> +++ b/migration/trace-events
> @@ -227,3 +227,17 @@ colo_vm_state_change(const char *old, const char *new)
> "Change '%s' => '%s'"
> colo_send_message(const char *msg) "Send '%s' message"
> colo_receive_message(const char *msg) "Receive '%s' message"
> colo_failover_set_state(const char *new_state) "new state %s"
> +
> +# migration/block-dirty-bitmap.c
> +send_bitmap_header_enter(void) ""
> +send_bitmap_bits(uint32_t flags, uint64_t start_sector, uint32_t nr_sectors,
> uint64_t data_size) "\n flags: 0x%x\n start_sector: %" PRIu64 "\n
> nr_sectors: %" PRIu32 "\n data_size: %" PRIu64 "\n"
Tracing doesn't have \n's in
> +dirty_bitmap_save_iterate(int in_postcopy) "in postcopy: %d"
> +dirty_bitmap_save_complete_enter(void) ""
> +dirty_bitmap_save_complete_finish(void) ""
> +dirty_bitmap_save_pending(uint64_t pending, uint64_t max_size) "pending %"
> PRIu64 " max: %" PRIu64
> +dirty_bitmap_load_complete(void) ""
> +dirty_bitmap_load_bits_enter(uint64_t first_sector, uint32_t nr_sectors)
> "chunk: %" PRIu64 " %" PRIu32
> +dirty_bitmap_load_bits_zeroes(void) ""
> +dirty_bitmap_load_header(uint32_t flags) "flags 0x%x"
> +dirty_bitmap_load_enter(void) ""
> +dirty_bitmap_load_success(void) ""
So other than minor bits, this one looks OK from a migration side; I
can't say I've followed the block side of the patch though.
Dave
> --
> 2.11.1
>
--
Dr. David Alan Gilbert / address@hidden / Manchester, UK
- Re: [Qemu-block] [PATCH v10 10/12] migration: add postcopy migration of dirty bitmaps,
Dr. David Alan Gilbert <=