2015-05-07 19:33:31 +02:00
|
|
|
/*
|
|
|
|
* QEMU System Emulator
|
|
|
|
*
|
|
|
|
* Copyright (c) 2003-2008 Fabrice Bellard
|
2015-05-08 13:20:21 +02:00
|
|
|
* Copyright (c) 2011-2015 Red Hat Inc
|
|
|
|
*
|
|
|
|
* Authors:
|
|
|
|
* Juan Quintela <quintela@redhat.com>
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
|
|
* of this software and associated documentation files (the "Software"), to deal
|
|
|
|
* in the Software without restriction, including without limitation the rights
|
|
|
|
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
|
|
* copies of the Software, and to permit persons to whom the Software is
|
|
|
|
* furnished to do so, subject to the following conditions:
|
|
|
|
*
|
|
|
|
* The above copyright notice and this permission notice shall be included in
|
|
|
|
* all copies or substantial portions of the Software.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
|
|
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
|
|
|
* THE SOFTWARE.
|
|
|
|
*/
|
2016-01-26 19:16:54 +01:00
|
|
|
#include "qemu/osdep.h"
|
2016-03-15 16:58:45 +01:00
|
|
|
#include "qemu-common.h"
|
|
|
|
#include "cpu.h"
|
2015-05-07 19:33:31 +02:00
|
|
|
#include <zlib.h>
|
2015-12-16 12:47:36 +01:00
|
|
|
#include "qapi-event.h"
|
2016-03-20 18:16:19 +01:00
|
|
|
#include "qemu/cutils.h"
|
2015-05-07 19:33:31 +02:00
|
|
|
#include "qemu/bitops.h"
|
|
|
|
#include "qemu/bitmap.h"
|
2015-05-08 13:54:36 +02:00
|
|
|
#include "qemu/timer.h"
|
|
|
|
#include "qemu/main-loop.h"
|
2015-05-07 19:33:31 +02:00
|
|
|
#include "migration/migration.h"
|
2015-11-05 19:11:02 +01:00
|
|
|
#include "migration/postcopy-ram.h"
|
2015-05-07 19:33:31 +02:00
|
|
|
#include "exec/address-spaces.h"
|
|
|
|
#include "migration/page_cache.h"
|
|
|
|
#include "qemu/error-report.h"
|
|
|
|
#include "trace.h"
|
|
|
|
#include "exec/ram_addr.h"
|
|
|
|
#include "qemu/rcu_queue.h"
|
2016-10-27 08:42:59 +02:00
|
|
|
#include "migration/colo.h"
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
/***********************************************************/
|
|
|
|
/* ram save/restore */
|
|
|
|
|
|
|
|
#define RAM_SAVE_FLAG_FULL 0x01 /* Obsolete, not used anymore */
|
|
|
|
#define RAM_SAVE_FLAG_COMPRESS 0x02
|
|
|
|
#define RAM_SAVE_FLAG_MEM_SIZE 0x04
|
|
|
|
#define RAM_SAVE_FLAG_PAGE 0x08
|
|
|
|
#define RAM_SAVE_FLAG_EOS 0x10
|
|
|
|
#define RAM_SAVE_FLAG_CONTINUE 0x20
|
|
|
|
#define RAM_SAVE_FLAG_XBZRLE 0x40
|
|
|
|
/* 0x80 is reserved in migration.h start with 0x100 next */
|
|
|
|
#define RAM_SAVE_FLAG_COMPRESS_PAGE 0x100
|
|
|
|
|
2016-10-24 17:26:49 +02:00
|
|
|
static uint8_t *ZERO_TARGET_PAGE;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
static inline bool is_zero_range(uint8_t *p, uint64_t size)
|
|
|
|
{
|
2016-08-29 20:46:14 +02:00
|
|
|
return buffer_is_zero(p, size);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* struct contains XBZRLE cache and a static page
|
|
|
|
used by the compression */
|
|
|
|
static struct {
|
|
|
|
/* buffer used for XBZRLE encoding */
|
|
|
|
uint8_t *encoded_buf;
|
|
|
|
/* buffer for storing page content */
|
|
|
|
uint8_t *current_buf;
|
|
|
|
/* Cache for XBZRLE, Protected by lock. */
|
|
|
|
PageCache *cache;
|
|
|
|
QemuMutex lock;
|
|
|
|
} XBZRLE;
|
|
|
|
|
|
|
|
/* buffer used for XBZRLE decoding */
|
|
|
|
static uint8_t *xbzrle_decoded_buf;
|
|
|
|
|
|
|
|
static void XBZRLE_cache_lock(void)
|
|
|
|
{
|
|
|
|
if (migrate_use_xbzrle())
|
|
|
|
qemu_mutex_lock(&XBZRLE.lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void XBZRLE_cache_unlock(void)
|
|
|
|
{
|
|
|
|
if (migrate_use_xbzrle())
|
|
|
|
qemu_mutex_unlock(&XBZRLE.lock);
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* xbzrle_cache_resize: resize the xbzrle cache
|
|
|
|
*
|
|
|
|
* This function is called from qmp_migrate_set_cache_size in main
|
|
|
|
* thread, possibly while a migration is in progress. A running
|
|
|
|
* migration may be using the cache and might finish during this call,
|
|
|
|
* hence changes to the cache are protected by XBZRLE.lock().
|
|
|
|
*
|
|
|
|
* Returns the new_size or negative in case of error.
|
|
|
|
*
|
|
|
|
* @new_size: new cache size
|
2015-05-07 19:33:31 +02:00
|
|
|
*/
|
|
|
|
int64_t xbzrle_cache_resize(int64_t new_size)
|
|
|
|
{
|
|
|
|
PageCache *new_cache;
|
|
|
|
int64_t ret;
|
|
|
|
|
|
|
|
if (new_size < TARGET_PAGE_SIZE) {
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
XBZRLE_cache_lock();
|
|
|
|
|
|
|
|
if (XBZRLE.cache != NULL) {
|
|
|
|
if (pow2floor(new_size) == migrate_xbzrle_cache_size()) {
|
|
|
|
goto out_new_size;
|
|
|
|
}
|
|
|
|
new_cache = cache_init(new_size / TARGET_PAGE_SIZE,
|
|
|
|
TARGET_PAGE_SIZE);
|
|
|
|
if (!new_cache) {
|
|
|
|
error_report("Error creating cache");
|
|
|
|
ret = -1;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
cache_fini(XBZRLE.cache);
|
|
|
|
XBZRLE.cache = new_cache;
|
|
|
|
}
|
|
|
|
|
|
|
|
out_new_size:
|
|
|
|
ret = pow2floor(new_size);
|
|
|
|
out:
|
|
|
|
XBZRLE_cache_unlock();
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
/* State of RAM for migration */
|
|
|
|
struct RAMState {
|
|
|
|
/* Last block that we have visited searching for dirty pages */
|
|
|
|
RAMBlock *last_seen_block;
|
|
|
|
/* Last block from where we have sent data */
|
|
|
|
RAMBlock *last_sent_block;
|
|
|
|
/* Last offset we have sent data from */
|
|
|
|
ram_addr_t last_offset;
|
|
|
|
/* last ram version we have seen */
|
|
|
|
uint32_t last_version;
|
|
|
|
/* We are in the first round */
|
|
|
|
bool ram_bulk_stage;
|
2017-03-13 19:35:50 +01:00
|
|
|
/* How many times we have dirty too many pages */
|
|
|
|
int dirty_rate_high_cnt;
|
2017-03-13 19:39:02 +01:00
|
|
|
/* How many times we have synchronized the bitmap */
|
|
|
|
uint64_t bitmap_sync_count;
|
2017-03-13 19:44:57 +01:00
|
|
|
/* these variables are used for bitmap sync */
|
|
|
|
/* last time we did a full bitmap_sync */
|
|
|
|
int64_t time_last_bitmap_sync;
|
2017-03-28 14:59:01 +02:00
|
|
|
/* bytes transferred at start_time */
|
2017-03-28 14:59:54 +02:00
|
|
|
uint64_t bytes_xfer_prev;
|
2017-03-28 15:02:43 +02:00
|
|
|
/* number of dirty pages since start_time */
|
2017-03-28 15:05:53 +02:00
|
|
|
uint64_t num_dirty_pages_period;
|
2017-03-13 19:49:19 +01:00
|
|
|
/* xbzrle misses since the beginning of the period */
|
|
|
|
uint64_t xbzrle_cache_miss_prev;
|
2017-03-13 19:51:13 +01:00
|
|
|
/* number of iterations at the beginning of period */
|
|
|
|
uint64_t iterations_prev;
|
2017-03-13 20:30:21 +01:00
|
|
|
/* Accounting fields */
|
|
|
|
/* number of zero pages. It used to be pages filled by the same char. */
|
|
|
|
uint64_t zero_pages;
|
2017-03-13 20:40:53 +01:00
|
|
|
/* number of normal transferred pages */
|
|
|
|
uint64_t norm_pages;
|
2017-03-13 20:51:34 +01:00
|
|
|
/* Iterations since start */
|
|
|
|
uint64_t iterations;
|
2017-03-13 19:26:29 +01:00
|
|
|
};
|
|
|
|
typedef struct RAMState RAMState;
|
|
|
|
|
|
|
|
static RAMState ram_state;
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
/* accounting for migration statistics */
|
|
|
|
typedef struct AccountingInfo {
|
|
|
|
uint64_t xbzrle_bytes;
|
|
|
|
uint64_t xbzrle_pages;
|
|
|
|
uint64_t xbzrle_cache_miss;
|
|
|
|
double xbzrle_cache_miss_rate;
|
|
|
|
uint64_t xbzrle_overflows;
|
|
|
|
} AccountingInfo;
|
|
|
|
|
|
|
|
static AccountingInfo acct_info;
|
|
|
|
|
|
|
|
static void acct_clear(void)
|
|
|
|
{
|
|
|
|
memset(&acct_info, 0, sizeof(acct_info));
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t dup_mig_pages_transferred(void)
|
|
|
|
{
|
2017-03-13 20:30:21 +01:00
|
|
|
return ram_state.zero_pages;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t norm_mig_pages_transferred(void)
|
|
|
|
{
|
2017-03-13 20:40:53 +01:00
|
|
|
return ram_state.norm_pages;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t xbzrle_mig_bytes_transferred(void)
|
|
|
|
{
|
|
|
|
return acct_info.xbzrle_bytes;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t xbzrle_mig_pages_transferred(void)
|
|
|
|
{
|
|
|
|
return acct_info.xbzrle_pages;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t xbzrle_mig_pages_cache_miss(void)
|
|
|
|
{
|
|
|
|
return acct_info.xbzrle_cache_miss;
|
|
|
|
}
|
|
|
|
|
|
|
|
double xbzrle_mig_cache_miss_rate(void)
|
|
|
|
{
|
|
|
|
return acct_info.xbzrle_cache_miss_rate;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t xbzrle_mig_pages_overflow(void)
|
|
|
|
{
|
|
|
|
return acct_info.xbzrle_overflows;
|
|
|
|
}
|
|
|
|
|
2015-07-02 14:18:06 +02:00
|
|
|
static QemuMutex migration_bitmap_mutex;
|
2015-05-07 19:33:31 +02:00
|
|
|
static uint64_t migration_dirty_pages;
|
|
|
|
|
2015-09-23 16:27:10 +02:00
|
|
|
/* used by the search for pages to send */
|
|
|
|
struct PageSearchStatus {
|
|
|
|
/* Current block being searched */
|
|
|
|
RAMBlock *block;
|
|
|
|
/* Current offset to search from */
|
|
|
|
ram_addr_t offset;
|
|
|
|
/* Set once we wrap around */
|
|
|
|
bool complete_round;
|
|
|
|
};
|
|
|
|
typedef struct PageSearchStatus PageSearchStatus;
|
|
|
|
|
2015-09-28 13:41:58 +02:00
|
|
|
static struct BitmapRcu {
|
|
|
|
struct rcu_head rcu;
|
2015-11-05 19:11:01 +01:00
|
|
|
/* Main migration bitmap */
|
2015-09-28 13:41:58 +02:00
|
|
|
unsigned long *bmap;
|
2015-11-05 19:11:01 +01:00
|
|
|
/* bitmap of pages that haven't been sent even once
|
|
|
|
* only maintained and used in postcopy at the moment
|
|
|
|
* where it's used to send the dirtymap at the start
|
|
|
|
* of the postcopy phase
|
|
|
|
*/
|
|
|
|
unsigned long *unsentmap;
|
2015-09-28 13:41:58 +02:00
|
|
|
} *migration_bitmap_rcu;
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
struct CompressParam {
|
|
|
|
bool done;
|
2016-05-05 09:32:56 +02:00
|
|
|
bool quit;
|
2015-05-07 19:33:31 +02:00
|
|
|
QEMUFile *file;
|
|
|
|
QemuMutex mutex;
|
|
|
|
QemuCond cond;
|
|
|
|
RAMBlock *block;
|
|
|
|
ram_addr_t offset;
|
|
|
|
};
|
|
|
|
typedef struct CompressParam CompressParam;
|
|
|
|
|
|
|
|
struct DecompressParam {
|
2016-05-05 09:32:51 +02:00
|
|
|
bool done;
|
2016-05-05 09:32:56 +02:00
|
|
|
bool quit;
|
2015-05-07 19:33:31 +02:00
|
|
|
QemuMutex mutex;
|
|
|
|
QemuCond cond;
|
|
|
|
void *des;
|
2016-01-22 16:09:21 +01:00
|
|
|
uint8_t *compbuf;
|
2015-05-07 19:33:31 +02:00
|
|
|
int len;
|
|
|
|
};
|
|
|
|
typedef struct DecompressParam DecompressParam;
|
|
|
|
|
|
|
|
static CompressParam *comp_param;
|
|
|
|
static QemuThread *compress_threads;
|
|
|
|
/* comp_done_cond is used to wake up the migration thread when
|
|
|
|
* one of the compression threads has finished the compression.
|
|
|
|
* comp_done_lock is used to co-work with comp_done_cond.
|
|
|
|
*/
|
2016-05-05 09:32:59 +02:00
|
|
|
static QemuMutex comp_done_lock;
|
|
|
|
static QemuCond comp_done_cond;
|
2015-05-07 19:33:31 +02:00
|
|
|
/* The empty QEMUFileOps will be used by file in CompressParam */
|
|
|
|
static const QEMUFileOps empty_ops = { };
|
|
|
|
|
|
|
|
static bool compression_switch;
|
|
|
|
static DecompressParam *decomp_param;
|
|
|
|
static QemuThread *decompress_threads;
|
2016-05-05 09:32:51 +02:00
|
|
|
static QemuMutex decomp_done_lock;
|
|
|
|
static QemuCond decomp_done_cond;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2016-05-05 09:32:57 +02:00
|
|
|
static int do_compress_ram_page(QEMUFile *f, RAMBlock *block,
|
|
|
|
ram_addr_t offset);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
static void *do_data_compress(void *opaque)
|
|
|
|
{
|
|
|
|
CompressParam *param = opaque;
|
2016-05-05 09:32:57 +02:00
|
|
|
RAMBlock *block;
|
|
|
|
ram_addr_t offset;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2016-05-05 09:32:57 +02:00
|
|
|
qemu_mutex_lock(¶m->mutex);
|
2016-05-05 09:32:56 +02:00
|
|
|
while (!param->quit) {
|
2016-05-05 09:32:57 +02:00
|
|
|
if (param->block) {
|
|
|
|
block = param->block;
|
|
|
|
offset = param->offset;
|
|
|
|
param->block = NULL;
|
|
|
|
qemu_mutex_unlock(¶m->mutex);
|
|
|
|
|
|
|
|
do_compress_ram_page(param->file, block, offset);
|
|
|
|
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_mutex_lock(&comp_done_lock);
|
2016-05-05 09:32:57 +02:00
|
|
|
param->done = true;
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_cond_signal(&comp_done_cond);
|
|
|
|
qemu_mutex_unlock(&comp_done_lock);
|
2016-05-05 09:32:57 +02:00
|
|
|
|
|
|
|
qemu_mutex_lock(¶m->mutex);
|
|
|
|
} else {
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_cond_wait(¶m->cond, ¶m->mutex);
|
|
|
|
}
|
|
|
|
}
|
2016-05-05 09:32:57 +02:00
|
|
|
qemu_mutex_unlock(¶m->mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void terminate_compression_threads(void)
|
|
|
|
{
|
|
|
|
int idx, thread_count;
|
|
|
|
|
|
|
|
thread_count = migrate_compress_threads();
|
2017-03-23 15:06:39 +01:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
for (idx = 0; idx < thread_count; idx++) {
|
|
|
|
qemu_mutex_lock(&comp_param[idx].mutex);
|
2016-05-05 09:32:56 +02:00
|
|
|
comp_param[idx].quit = true;
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_cond_signal(&comp_param[idx].cond);
|
|
|
|
qemu_mutex_unlock(&comp_param[idx].mutex);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void migrate_compress_threads_join(void)
|
|
|
|
{
|
|
|
|
int i, thread_count;
|
|
|
|
|
|
|
|
if (!migrate_use_compression()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
terminate_compression_threads();
|
|
|
|
thread_count = migrate_compress_threads();
|
|
|
|
for (i = 0; i < thread_count; i++) {
|
|
|
|
qemu_thread_join(compress_threads + i);
|
|
|
|
qemu_fclose(comp_param[i].file);
|
|
|
|
qemu_mutex_destroy(&comp_param[i].mutex);
|
|
|
|
qemu_cond_destroy(&comp_param[i].cond);
|
|
|
|
}
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_mutex_destroy(&comp_done_lock);
|
|
|
|
qemu_cond_destroy(&comp_done_cond);
|
2015-05-07 19:33:31 +02:00
|
|
|
g_free(compress_threads);
|
|
|
|
g_free(comp_param);
|
|
|
|
compress_threads = NULL;
|
|
|
|
comp_param = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
void migrate_compress_threads_create(void)
|
|
|
|
{
|
|
|
|
int i, thread_count;
|
|
|
|
|
|
|
|
if (!migrate_use_compression()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
compression_switch = true;
|
|
|
|
thread_count = migrate_compress_threads();
|
|
|
|
compress_threads = g_new0(QemuThread, thread_count);
|
|
|
|
comp_param = g_new0(CompressParam, thread_count);
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_cond_init(&comp_done_cond);
|
|
|
|
qemu_mutex_init(&comp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
for (i = 0; i < thread_count; i++) {
|
2016-07-29 09:10:31 +02:00
|
|
|
/* comp_param[i].file is just used as a dummy buffer to save data,
|
|
|
|
* set its ops to empty.
|
2015-05-07 19:33:31 +02:00
|
|
|
*/
|
|
|
|
comp_param[i].file = qemu_fopen_ops(NULL, &empty_ops);
|
|
|
|
comp_param[i].done = true;
|
2016-05-05 09:32:56 +02:00
|
|
|
comp_param[i].quit = false;
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_mutex_init(&comp_param[i].mutex);
|
|
|
|
qemu_cond_init(&comp_param[i].cond);
|
|
|
|
qemu_thread_create(compress_threads + i, "compress",
|
|
|
|
do_data_compress, comp_param + i,
|
|
|
|
QEMU_THREAD_JOINABLE);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2017-03-23 15:06:39 +01:00
|
|
|
* save_page_header: write page header to wire
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
|
|
|
* If this is the 1st block, it also writes the block identification
|
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns the number of bytes written
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @block: block that contains the page we want to send
|
|
|
|
* @offset: offset inside the block for the page
|
|
|
|
* in the lower bits, it contains flags
|
|
|
|
*/
|
|
|
|
static size_t save_page_header(QEMUFile *f, RAMBlock *block, ram_addr_t offset)
|
|
|
|
{
|
2015-07-13 11:34:10 +02:00
|
|
|
size_t size, len;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
qemu_put_be64(f, offset);
|
|
|
|
size = 8;
|
|
|
|
|
|
|
|
if (!(offset & RAM_SAVE_FLAG_CONTINUE)) {
|
2015-07-13 11:34:10 +02:00
|
|
|
len = strlen(block->idstr);
|
|
|
|
qemu_put_byte(f, len);
|
|
|
|
qemu_put_buffer(f, (uint8_t *)block->idstr, len);
|
|
|
|
size += 1 + len;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
return size;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* mig_throttle_guest_down: throotle down the guest
|
|
|
|
*
|
|
|
|
* Reduce amount of guest cpu execution to hopefully slow down memory
|
|
|
|
* writes. If guest dirty memory rate is reduced below the rate at
|
|
|
|
* which we can transfer pages to the destination then we should be
|
|
|
|
* able to complete migration. Some workloads dirty memory way too
|
|
|
|
* fast and will not effectively converge, even with auto-converge.
|
2015-09-08 19:12:35 +02:00
|
|
|
*/
|
|
|
|
static void mig_throttle_guest_down(void)
|
|
|
|
{
|
|
|
|
MigrationState *s = migrate_get_current();
|
2016-04-27 12:05:14 +02:00
|
|
|
uint64_t pct_initial = s->parameters.cpu_throttle_initial;
|
|
|
|
uint64_t pct_icrement = s->parameters.cpu_throttle_increment;
|
2015-09-08 19:12:35 +02:00
|
|
|
|
|
|
|
/* We have not started throttling yet. Let's start it. */
|
|
|
|
if (!cpu_throttle_active()) {
|
|
|
|
cpu_throttle_set(pct_initial);
|
|
|
|
} else {
|
|
|
|
/* Throttling already on, just increase the rate */
|
|
|
|
cpu_throttle_set(cpu_throttle_get_percentage() + pct_icrement);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* xbzrle_cache_zero_page: insert a zero page in the XBZRLE cache
|
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2017-03-23 15:06:39 +01:00
|
|
|
* @current_addr: address for the zero page
|
|
|
|
*
|
|
|
|
* Update the xbzrle cache to reflect a page that's been sent as all 0.
|
2015-05-07 19:33:31 +02:00
|
|
|
* The important thing is that a stale (not-yet-0'd) page be replaced
|
|
|
|
* by the new data.
|
|
|
|
* As a bonus, if the page wasn't in the cache it gets added so that
|
2017-03-23 15:06:39 +01:00
|
|
|
* when a small write is made into the 0'd page it gets XBZRLE sent.
|
2015-05-07 19:33:31 +02:00
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
static void xbzrle_cache_zero_page(RAMState *rs, ram_addr_t current_addr)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
2017-03-13 19:26:29 +01:00
|
|
|
if (rs->ram_bulk_stage || !migrate_use_xbzrle()) {
|
2015-05-07 19:33:31 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* We don't care if this fails to allocate a new cache page
|
|
|
|
* as long as it updated an old one */
|
|
|
|
cache_insert(XBZRLE.cache, current_addr, ZERO_TARGET_PAGE,
|
2017-03-13 19:39:02 +01:00
|
|
|
rs->bitmap_sync_count);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
#define ENCODING_FLAG_XBZRLE 0x1
|
|
|
|
|
|
|
|
/**
|
|
|
|
* save_xbzrle_page: compress and send current page
|
|
|
|
*
|
|
|
|
* Returns: 1 means that we wrote the page
|
|
|
|
* 0 means that page is identical to the one already sent
|
|
|
|
* -1 means that xbzrle would be longer than normal
|
|
|
|
*
|
2017-03-13 19:39:02 +01:00
|
|
|
* @rs: current RAM state
|
2015-05-07 19:33:31 +02:00
|
|
|
* @f: QEMUFile where to send the data
|
2017-03-23 15:06:39 +01:00
|
|
|
* @current_data: pointer to the address of the page contents
|
|
|
|
* @current_addr: addr of the page
|
2015-05-07 19:33:31 +02:00
|
|
|
* @block: block that contains the page we want to send
|
|
|
|
* @offset: offset inside the block for the page
|
|
|
|
* @last_stage: if we are at the completion stage
|
|
|
|
* @bytes_transferred: increase it with the number of transferred bytes
|
|
|
|
*/
|
2017-03-13 19:39:02 +01:00
|
|
|
static int save_xbzrle_page(RAMState *rs, QEMUFile *f, uint8_t **current_data,
|
2015-05-07 19:33:31 +02:00
|
|
|
ram_addr_t current_addr, RAMBlock *block,
|
|
|
|
ram_addr_t offset, bool last_stage,
|
|
|
|
uint64_t *bytes_transferred)
|
|
|
|
{
|
|
|
|
int encoded_len = 0, bytes_xbzrle;
|
|
|
|
uint8_t *prev_cached_page;
|
|
|
|
|
2017-03-13 19:39:02 +01:00
|
|
|
if (!cache_is_cached(XBZRLE.cache, current_addr, rs->bitmap_sync_count)) {
|
2015-05-07 19:33:31 +02:00
|
|
|
acct_info.xbzrle_cache_miss++;
|
|
|
|
if (!last_stage) {
|
|
|
|
if (cache_insert(XBZRLE.cache, current_addr, *current_data,
|
2017-03-13 19:39:02 +01:00
|
|
|
rs->bitmap_sync_count) == -1) {
|
2015-05-07 19:33:31 +02:00
|
|
|
return -1;
|
|
|
|
} else {
|
|
|
|
/* update *current_data when the page has been
|
|
|
|
inserted into cache */
|
|
|
|
*current_data = get_cached_data(XBZRLE.cache, current_addr);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
prev_cached_page = get_cached_data(XBZRLE.cache, current_addr);
|
|
|
|
|
|
|
|
/* save current buffer into memory */
|
|
|
|
memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
|
|
|
|
|
|
|
|
/* XBZRLE encoding (if there is no overflow) */
|
|
|
|
encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
|
|
|
|
TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
|
|
|
|
TARGET_PAGE_SIZE);
|
|
|
|
if (encoded_len == 0) {
|
2017-01-23 22:32:05 +01:00
|
|
|
trace_save_xbzrle_page_skipping();
|
2015-05-07 19:33:31 +02:00
|
|
|
return 0;
|
|
|
|
} else if (encoded_len == -1) {
|
2017-01-23 22:32:05 +01:00
|
|
|
trace_save_xbzrle_page_overflow();
|
2015-05-07 19:33:31 +02:00
|
|
|
acct_info.xbzrle_overflows++;
|
|
|
|
/* update data in the cache */
|
|
|
|
if (!last_stage) {
|
|
|
|
memcpy(prev_cached_page, *current_data, TARGET_PAGE_SIZE);
|
|
|
|
*current_data = prev_cached_page;
|
|
|
|
}
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* we need to update the data in the cache, in order to get the same data */
|
|
|
|
if (!last_stage) {
|
|
|
|
memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Send XBZRLE based compressed page */
|
|
|
|
bytes_xbzrle = save_page_header(f, block, offset | RAM_SAVE_FLAG_XBZRLE);
|
|
|
|
qemu_put_byte(f, ENCODING_FLAG_XBZRLE);
|
|
|
|
qemu_put_be16(f, encoded_len);
|
|
|
|
qemu_put_buffer(f, XBZRLE.encoded_buf, encoded_len);
|
|
|
|
bytes_xbzrle += encoded_len + 1 + 2;
|
|
|
|
acct_info.xbzrle_pages++;
|
|
|
|
acct_info.xbzrle_bytes += bytes_xbzrle;
|
|
|
|
*bytes_transferred += bytes_xbzrle;
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* migration_bitmap_find_dirty: find the next dirty page from start
|
2015-11-05 19:11:01 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Called with rcu_read_lock() to protect migration_bitmap
|
|
|
|
*
|
|
|
|
* Returns the byte offset within memory region of the start of a dirty page
|
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2017-03-23 15:06:39 +01:00
|
|
|
* @rb: RAMBlock where to search for dirty pages
|
|
|
|
* @start: starting address (typically so we can continue from previous page)
|
|
|
|
* @ram_addr_abs: pointer into which to store the address of the dirty page
|
|
|
|
* within the global ram_addr space
|
2015-11-05 19:11:01 +01:00
|
|
|
*/
|
2015-05-07 19:33:31 +02:00
|
|
|
static inline
|
2017-03-13 19:26:29 +01:00
|
|
|
ram_addr_t migration_bitmap_find_dirty(RAMState *rs, RAMBlock *rb,
|
2015-11-05 19:11:09 +01:00
|
|
|
ram_addr_t start,
|
|
|
|
ram_addr_t *ram_addr_abs)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
2015-08-13 12:51:30 +02:00
|
|
|
unsigned long base = rb->offset >> TARGET_PAGE_BITS;
|
2015-05-07 19:33:31 +02:00
|
|
|
unsigned long nr = base + (start >> TARGET_PAGE_BITS);
|
2015-08-13 12:51:30 +02:00
|
|
|
uint64_t rb_size = rb->used_length;
|
|
|
|
unsigned long size = base + (rb_size >> TARGET_PAGE_BITS);
|
2015-07-02 14:18:05 +02:00
|
|
|
unsigned long *bitmap;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
unsigned long next;
|
|
|
|
|
2015-09-28 13:41:58 +02:00
|
|
|
bitmap = atomic_rcu_read(&migration_bitmap_rcu)->bmap;
|
2017-03-13 19:26:29 +01:00
|
|
|
if (rs->ram_bulk_stage && nr > base) {
|
2015-05-07 19:33:31 +02:00
|
|
|
next = nr + 1;
|
|
|
|
} else {
|
2015-07-02 14:18:05 +02:00
|
|
|
next = find_next_bit(bitmap, size, nr);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:01 +01:00
|
|
|
*ram_addr_abs = next << TARGET_PAGE_BITS;
|
2015-05-07 19:33:31 +02:00
|
|
|
return (next - base) << TARGET_PAGE_BITS;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:09 +01:00
|
|
|
static inline bool migration_bitmap_clear_dirty(ram_addr_t addr)
|
|
|
|
{
|
|
|
|
bool ret;
|
|
|
|
int nr = addr >> TARGET_PAGE_BITS;
|
|
|
|
unsigned long *bitmap = atomic_rcu_read(&migration_bitmap_rcu)->bmap;
|
|
|
|
|
|
|
|
ret = test_and_clear_bit(nr, bitmap);
|
|
|
|
|
|
|
|
if (ret) {
|
|
|
|
migration_dirty_pages--;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2017-03-28 15:02:43 +02:00
|
|
|
static void migration_bitmap_sync_range(RAMState *rs, ram_addr_t start,
|
|
|
|
ram_addr_t length)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
2015-07-02 14:18:05 +02:00
|
|
|
unsigned long *bitmap;
|
2015-09-28 13:41:58 +02:00
|
|
|
bitmap = atomic_rcu_read(&migration_bitmap_rcu)->bmap;
|
2017-03-14 02:55:07 +01:00
|
|
|
migration_dirty_pages += cpu_physical_memory_sync_dirty_bitmap(bitmap,
|
2017-03-28 15:02:43 +02:00
|
|
|
start, length, &rs->num_dirty_pages_period);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
2017-03-13 19:44:57 +01:00
|
|
|
static void migration_bitmap_sync_init(RAMState *rs)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
2017-03-13 19:44:57 +01:00
|
|
|
rs->time_last_bitmap_sync = 0;
|
2017-03-28 14:59:01 +02:00
|
|
|
rs->bytes_xfer_prev = 0;
|
2017-03-28 15:02:43 +02:00
|
|
|
rs->num_dirty_pages_period = 0;
|
2017-03-13 19:49:19 +01:00
|
|
|
rs->xbzrle_cache_miss_prev = 0;
|
2017-03-13 19:51:13 +01:00
|
|
|
rs->iterations_prev = 0;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_pagesize_summary: calculate all the pagesizes of a VM
|
|
|
|
*
|
|
|
|
* Returns a summary bitmap of the page sizes of all RAMBlocks
|
|
|
|
*
|
|
|
|
* For VMs with just normal pages this is equivalent to the host page
|
|
|
|
* size. If it's got some huge pages then it's the OR of all the
|
|
|
|
* different page sizes.
|
2017-02-24 19:28:29 +01:00
|
|
|
*/
|
|
|
|
uint64_t ram_pagesize_summary(void)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
uint64_t summary = 0;
|
|
|
|
|
|
|
|
QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
|
|
|
|
summary |= block->page_size;
|
|
|
|
}
|
|
|
|
|
|
|
|
return summary;
|
|
|
|
}
|
|
|
|
|
2017-03-13 19:35:50 +01:00
|
|
|
static void migration_bitmap_sync(RAMState *rs)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
MigrationState *s = migrate_get_current();
|
|
|
|
int64_t end_time;
|
2017-03-28 14:59:54 +02:00
|
|
|
uint64_t bytes_xfer_now;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2017-03-13 19:39:02 +01:00
|
|
|
rs->bitmap_sync_count++;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2017-03-28 14:59:01 +02:00
|
|
|
if (!rs->bytes_xfer_prev) {
|
|
|
|
rs->bytes_xfer_prev = ram_bytes_transferred();
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
2017-03-13 19:44:57 +01:00
|
|
|
if (!rs->time_last_bitmap_sync) {
|
|
|
|
rs->time_last_bitmap_sync = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
trace_migration_bitmap_sync_start();
|
2016-09-22 16:08:31 +02:00
|
|
|
memory_global_dirty_log_sync();
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2015-07-02 14:18:06 +02:00
|
|
|
qemu_mutex_lock(&migration_bitmap_mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
rcu_read_lock();
|
|
|
|
QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
|
2017-03-28 15:02:43 +02:00
|
|
|
migration_bitmap_sync_range(rs, block->offset, block->used_length);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
2015-07-02 14:18:06 +02:00
|
|
|
qemu_mutex_unlock(&migration_bitmap_mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2017-03-28 15:02:43 +02:00
|
|
|
trace_migration_bitmap_sync_end(rs->num_dirty_pages_period);
|
2017-03-14 02:55:07 +01:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
end_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
|
|
|
|
|
|
|
|
/* more than 1 second = 1000 millisecons */
|
2017-03-13 19:44:57 +01:00
|
|
|
if (end_time > rs->time_last_bitmap_sync + 1000) {
|
2015-05-07 19:33:31 +02:00
|
|
|
if (migrate_auto_converge()) {
|
|
|
|
/* The following detection logic can be refined later. For now:
|
|
|
|
Check to see if the dirtied bytes is 50% more than the approx.
|
|
|
|
amount of bytes that just got transferred since the last time we
|
2015-09-08 19:12:35 +02:00
|
|
|
were in this routine. If that happens twice, start or increase
|
|
|
|
throttling */
|
2015-05-07 19:33:31 +02:00
|
|
|
bytes_xfer_now = ram_bytes_transferred();
|
2015-09-08 19:12:35 +02:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
if (s->dirty_pages_rate &&
|
2017-03-28 15:02:43 +02:00
|
|
|
(rs->num_dirty_pages_period * TARGET_PAGE_SIZE >
|
2017-03-28 14:59:01 +02:00
|
|
|
(bytes_xfer_now - rs->bytes_xfer_prev) / 2) &&
|
2017-03-13 19:35:50 +01:00
|
|
|
(rs->dirty_rate_high_cnt++ >= 2)) {
|
2015-05-07 19:33:31 +02:00
|
|
|
trace_migration_throttle();
|
2017-03-13 19:35:50 +01:00
|
|
|
rs->dirty_rate_high_cnt = 0;
|
2015-09-08 19:12:35 +02:00
|
|
|
mig_throttle_guest_down();
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2017-03-28 14:59:01 +02:00
|
|
|
rs->bytes_xfer_prev = bytes_xfer_now;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2015-09-08 19:12:35 +02:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
if (migrate_use_xbzrle()) {
|
2017-03-13 20:51:34 +01:00
|
|
|
if (rs->iterations_prev != rs->iterations) {
|
2015-05-07 19:33:31 +02:00
|
|
|
acct_info.xbzrle_cache_miss_rate =
|
|
|
|
(double)(acct_info.xbzrle_cache_miss -
|
2017-03-13 19:49:19 +01:00
|
|
|
rs->xbzrle_cache_miss_prev) /
|
2017-03-13 20:51:34 +01:00
|
|
|
(rs->iterations - rs->iterations_prev);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2017-03-13 20:51:34 +01:00
|
|
|
rs->iterations_prev = rs->iterations;
|
2017-03-13 19:49:19 +01:00
|
|
|
rs->xbzrle_cache_miss_prev = acct_info.xbzrle_cache_miss;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2017-03-28 15:02:43 +02:00
|
|
|
s->dirty_pages_rate = rs->num_dirty_pages_period * 1000
|
2017-03-13 19:44:57 +01:00
|
|
|
/ (end_time - rs->time_last_bitmap_sync);
|
2015-05-07 19:33:31 +02:00
|
|
|
s->dirty_bytes_rate = s->dirty_pages_rate * TARGET_PAGE_SIZE;
|
2017-03-13 19:44:57 +01:00
|
|
|
rs->time_last_bitmap_sync = end_time;
|
2017-03-28 15:02:43 +02:00
|
|
|
rs->num_dirty_pages_period = 0;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2017-03-13 19:39:02 +01:00
|
|
|
s->dirty_sync_count = rs->bitmap_sync_count;
|
2015-12-16 12:47:36 +01:00
|
|
|
if (migrate_use_events()) {
|
2017-03-13 19:39:02 +01:00
|
|
|
qapi_event_send_migration_pass(rs->bitmap_sync_count, NULL);
|
2015-12-16 12:47:36 +01:00
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2017-03-23 15:06:39 +01:00
|
|
|
* save_zero_page: send the zero page to the stream
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns the number of pages written.
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
2017-03-13 20:30:21 +01:00
|
|
|
* @rs: current RAM state
|
2015-05-07 19:33:31 +02:00
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @block: block that contains the page we want to send
|
|
|
|
* @offset: offset inside the block for the page
|
|
|
|
* @p: pointer to the page
|
|
|
|
* @bytes_transferred: increase it with the number of transferred bytes
|
|
|
|
*/
|
2017-03-13 20:30:21 +01:00
|
|
|
static int save_zero_page(RAMState *rs, QEMUFile *f, RAMBlock *block,
|
|
|
|
ram_addr_t offset,
|
2015-05-07 19:33:31 +02:00
|
|
|
uint8_t *p, uint64_t *bytes_transferred)
|
|
|
|
{
|
|
|
|
int pages = -1;
|
|
|
|
|
|
|
|
if (is_zero_range(p, TARGET_PAGE_SIZE)) {
|
2017-03-13 20:30:21 +01:00
|
|
|
rs->zero_pages++;
|
2015-05-07 19:33:31 +02:00
|
|
|
*bytes_transferred += save_page_header(f, block,
|
|
|
|
offset | RAM_SAVE_FLAG_COMPRESS);
|
|
|
|
qemu_put_byte(f, 0);
|
|
|
|
*bytes_transferred += 1;
|
|
|
|
pages = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return pages;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:11:59 +01:00
|
|
|
static void ram_release_pages(MigrationState *ms, const char *rbname,
|
2017-02-03 16:23:20 +01:00
|
|
|
uint64_t offset, int pages)
|
|
|
|
{
|
|
|
|
if (!migrate_release_ram() || !migration_in_postcopy(ms)) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:11:59 +01:00
|
|
|
ram_discard_range(NULL, rbname, offset, pages << TARGET_PAGE_BITS);
|
2017-02-03 16:23:20 +01:00
|
|
|
}
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
/**
|
2017-03-23 15:06:39 +01:00
|
|
|
* ram_save_page: send the given page to the stream
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns the number of pages written.
|
2015-12-10 17:31:46 +01:00
|
|
|
* < 0 - error
|
|
|
|
* >=0 - Number of pages written - this might legally be 0
|
|
|
|
* if xbzrle noticed the page was the same.
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2017-03-23 15:06:39 +01:00
|
|
|
* @ms: current migration state
|
2015-05-07 19:33:31 +02:00
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @block: block that contains the page we want to send
|
|
|
|
* @offset: offset inside the block for the page
|
|
|
|
* @last_stage: if we are at the completion stage
|
|
|
|
* @bytes_transferred: increase it with the number of transferred bytes
|
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
static int ram_save_page(RAMState *rs, MigrationState *ms, QEMUFile *f,
|
|
|
|
PageSearchStatus *pss, bool last_stage,
|
|
|
|
uint64_t *bytes_transferred)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
|
|
|
int pages = -1;
|
|
|
|
uint64_t bytes_xmit;
|
|
|
|
ram_addr_t current_addr;
|
|
|
|
uint8_t *p;
|
|
|
|
int ret;
|
|
|
|
bool send_async = true;
|
2016-01-15 04:37:44 +01:00
|
|
|
RAMBlock *block = pss->block;
|
|
|
|
ram_addr_t offset = pss->offset;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2015-08-13 12:51:30 +02:00
|
|
|
p = block->host + offset;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
/* In doubt sent page as normal */
|
|
|
|
bytes_xmit = 0;
|
|
|
|
ret = ram_control_save_page(f, block->offset,
|
|
|
|
offset, TARGET_PAGE_SIZE, &bytes_xmit);
|
|
|
|
if (bytes_xmit) {
|
|
|
|
*bytes_transferred += bytes_xmit;
|
|
|
|
pages = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
XBZRLE_cache_lock();
|
|
|
|
|
|
|
|
current_addr = block->offset + offset;
|
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
if (block == rs->last_sent_block) {
|
2015-05-07 19:33:31 +02:00
|
|
|
offset |= RAM_SAVE_FLAG_CONTINUE;
|
|
|
|
}
|
|
|
|
if (ret != RAM_SAVE_CONTROL_NOT_SUPP) {
|
|
|
|
if (ret != RAM_SAVE_CONTROL_DELAYED) {
|
|
|
|
if (bytes_xmit > 0) {
|
2017-03-13 20:40:53 +01:00
|
|
|
rs->norm_pages++;
|
2015-05-07 19:33:31 +02:00
|
|
|
} else if (bytes_xmit == 0) {
|
2017-03-13 20:30:21 +01:00
|
|
|
rs->zero_pages++;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
2017-03-13 20:30:21 +01:00
|
|
|
pages = save_zero_page(rs, f, block, offset, p, bytes_transferred);
|
2015-05-07 19:33:31 +02:00
|
|
|
if (pages > 0) {
|
|
|
|
/* Must let xbzrle know, otherwise a previous (now 0'd) cached
|
|
|
|
* page would be stale
|
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
xbzrle_cache_zero_page(rs, current_addr);
|
2017-02-03 16:23:20 +01:00
|
|
|
ram_release_pages(ms, block->idstr, pss->offset, pages);
|
2017-03-13 19:26:29 +01:00
|
|
|
} else if (!rs->ram_bulk_stage &&
|
2017-02-03 16:23:19 +01:00
|
|
|
!migration_in_postcopy(ms) && migrate_use_xbzrle()) {
|
2017-03-13 19:39:02 +01:00
|
|
|
pages = save_xbzrle_page(rs, f, &p, current_addr, block,
|
2015-05-07 19:33:31 +02:00
|
|
|
offset, last_stage, bytes_transferred);
|
|
|
|
if (!last_stage) {
|
|
|
|
/* Can't send this cached data async, since the cache page
|
|
|
|
* might get updated before it gets to the wire
|
|
|
|
*/
|
|
|
|
send_async = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* XBZRLE overflow or normal page */
|
|
|
|
if (pages == -1) {
|
|
|
|
*bytes_transferred += save_page_header(f, block,
|
|
|
|
offset | RAM_SAVE_FLAG_PAGE);
|
|
|
|
if (send_async) {
|
2017-02-03 16:23:20 +01:00
|
|
|
qemu_put_buffer_async(f, p, TARGET_PAGE_SIZE,
|
|
|
|
migrate_release_ram() &
|
|
|
|
migration_in_postcopy(ms));
|
2015-05-07 19:33:31 +02:00
|
|
|
} else {
|
|
|
|
qemu_put_buffer(f, p, TARGET_PAGE_SIZE);
|
|
|
|
}
|
|
|
|
*bytes_transferred += TARGET_PAGE_SIZE;
|
|
|
|
pages = 1;
|
2017-03-13 20:40:53 +01:00
|
|
|
rs->norm_pages++;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
XBZRLE_cache_unlock();
|
|
|
|
|
|
|
|
return pages;
|
|
|
|
}
|
|
|
|
|
2016-05-05 09:32:57 +02:00
|
|
|
static int do_compress_ram_page(QEMUFile *f, RAMBlock *block,
|
|
|
|
ram_addr_t offset)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
|
|
|
int bytes_sent, blen;
|
2016-05-05 09:32:57 +02:00
|
|
|
uint8_t *p = block->host + (offset & TARGET_PAGE_MASK);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2016-05-05 09:32:57 +02:00
|
|
|
bytes_sent = save_page_header(f, block, offset |
|
2015-05-07 19:33:31 +02:00
|
|
|
RAM_SAVE_FLAG_COMPRESS_PAGE);
|
2016-05-05 09:32:57 +02:00
|
|
|
blen = qemu_put_compression_data(f, p, TARGET_PAGE_SIZE,
|
2015-05-07 19:33:31 +02:00
|
|
|
migrate_compress_level());
|
2016-05-05 09:32:54 +02:00
|
|
|
if (blen < 0) {
|
|
|
|
bytes_sent = 0;
|
|
|
|
qemu_file_set_error(migrate_get_current()->to_dst_file, blen);
|
|
|
|
error_report("compressed data failed!");
|
|
|
|
} else {
|
|
|
|
bytes_sent += blen;
|
2017-02-03 16:23:20 +01:00
|
|
|
ram_release_pages(migrate_get_current(), block->idstr,
|
|
|
|
offset & TARGET_PAGE_MASK, 1);
|
2016-05-05 09:32:54 +02:00
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
return bytes_sent;
|
|
|
|
}
|
|
|
|
|
|
|
|
static uint64_t bytes_transferred;
|
|
|
|
|
|
|
|
static void flush_compressed_data(QEMUFile *f)
|
|
|
|
{
|
|
|
|
int idx, len, thread_count;
|
|
|
|
|
|
|
|
if (!migrate_use_compression()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
thread_count = migrate_compress_threads();
|
2016-05-05 09:32:57 +02:00
|
|
|
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_mutex_lock(&comp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
for (idx = 0; idx < thread_count; idx++) {
|
2016-05-05 09:32:57 +02:00
|
|
|
while (!comp_param[idx].done) {
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_cond_wait(&comp_done_cond, &comp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2016-05-05 09:32:57 +02:00
|
|
|
}
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_mutex_unlock(&comp_done_lock);
|
2016-05-05 09:32:57 +02:00
|
|
|
|
|
|
|
for (idx = 0; idx < thread_count; idx++) {
|
|
|
|
qemu_mutex_lock(&comp_param[idx].mutex);
|
2016-05-05 09:32:56 +02:00
|
|
|
if (!comp_param[idx].quit) {
|
2015-05-07 19:33:31 +02:00
|
|
|
len = qemu_put_qemu_file(f, comp_param[idx].file);
|
|
|
|
bytes_transferred += len;
|
|
|
|
}
|
2016-05-05 09:32:57 +02:00
|
|
|
qemu_mutex_unlock(&comp_param[idx].mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void set_compress_params(CompressParam *param, RAMBlock *block,
|
|
|
|
ram_addr_t offset)
|
|
|
|
{
|
|
|
|
param->block = block;
|
|
|
|
param->offset = offset;
|
|
|
|
}
|
|
|
|
|
2017-03-13 20:40:53 +01:00
|
|
|
static int compress_page_with_multi_thread(RAMState *rs, QEMUFile *f,
|
|
|
|
RAMBlock *block, ram_addr_t offset,
|
2015-05-07 19:33:31 +02:00
|
|
|
uint64_t *bytes_transferred)
|
|
|
|
{
|
|
|
|
int idx, thread_count, bytes_xmit = -1, pages = -1;
|
|
|
|
|
|
|
|
thread_count = migrate_compress_threads();
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_mutex_lock(&comp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
while (true) {
|
|
|
|
for (idx = 0; idx < thread_count; idx++) {
|
|
|
|
if (comp_param[idx].done) {
|
2016-05-05 09:32:57 +02:00
|
|
|
comp_param[idx].done = false;
|
2015-05-07 19:33:31 +02:00
|
|
|
bytes_xmit = qemu_put_qemu_file(f, comp_param[idx].file);
|
2016-05-05 09:32:57 +02:00
|
|
|
qemu_mutex_lock(&comp_param[idx].mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
set_compress_params(&comp_param[idx], block, offset);
|
2016-05-05 09:32:57 +02:00
|
|
|
qemu_cond_signal(&comp_param[idx].cond);
|
|
|
|
qemu_mutex_unlock(&comp_param[idx].mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
pages = 1;
|
2017-03-13 20:40:53 +01:00
|
|
|
rs->norm_pages++;
|
2015-05-07 19:33:31 +02:00
|
|
|
*bytes_transferred += bytes_xmit;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (pages > 0) {
|
|
|
|
break;
|
|
|
|
} else {
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_cond_wait(&comp_done_cond, &comp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
}
|
2016-05-05 09:32:59 +02:00
|
|
|
qemu_mutex_unlock(&comp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
return pages;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* ram_save_compressed_page: compress the given page and send it to the stream
|
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns the number of pages written.
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2017-03-23 15:06:39 +01:00
|
|
|
* @ms: current migration state
|
2015-05-07 19:33:31 +02:00
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @block: block that contains the page we want to send
|
|
|
|
* @offset: offset inside the block for the page
|
|
|
|
* @last_stage: if we are at the completion stage
|
|
|
|
* @bytes_transferred: increase it with the number of transferred bytes
|
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
static int ram_save_compressed_page(RAMState *rs, MigrationState *ms,
|
|
|
|
QEMUFile *f,
|
2017-02-03 16:23:19 +01:00
|
|
|
PageSearchStatus *pss, bool last_stage,
|
2015-05-07 19:33:31 +02:00
|
|
|
uint64_t *bytes_transferred)
|
|
|
|
{
|
|
|
|
int pages = -1;
|
2016-05-05 09:32:55 +02:00
|
|
|
uint64_t bytes_xmit = 0;
|
2015-05-07 19:33:31 +02:00
|
|
|
uint8_t *p;
|
2016-05-05 09:32:55 +02:00
|
|
|
int ret, blen;
|
2016-01-15 04:37:44 +01:00
|
|
|
RAMBlock *block = pss->block;
|
|
|
|
ram_addr_t offset = pss->offset;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2015-08-13 12:51:30 +02:00
|
|
|
p = block->host + offset;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
ret = ram_control_save_page(f, block->offset,
|
|
|
|
offset, TARGET_PAGE_SIZE, &bytes_xmit);
|
|
|
|
if (bytes_xmit) {
|
|
|
|
*bytes_transferred += bytes_xmit;
|
|
|
|
pages = 1;
|
|
|
|
}
|
|
|
|
if (ret != RAM_SAVE_CONTROL_NOT_SUPP) {
|
|
|
|
if (ret != RAM_SAVE_CONTROL_DELAYED) {
|
|
|
|
if (bytes_xmit > 0) {
|
2017-03-13 20:40:53 +01:00
|
|
|
rs->norm_pages++;
|
2015-05-07 19:33:31 +02:00
|
|
|
} else if (bytes_xmit == 0) {
|
2017-03-13 20:30:21 +01:00
|
|
|
rs->zero_pages++;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/* When starting the process of a new block, the first page of
|
|
|
|
* the block should be sent out before other pages in the same
|
|
|
|
* block, and all the pages in last block should have been sent
|
|
|
|
* out, keeping this order is important, because the 'cont' flag
|
|
|
|
* is used to avoid resending the block name.
|
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
if (block != rs->last_sent_block) {
|
2015-05-07 19:33:31 +02:00
|
|
|
flush_compressed_data(f);
|
2017-03-13 20:30:21 +01:00
|
|
|
pages = save_zero_page(rs, f, block, offset, p, bytes_transferred);
|
2015-05-07 19:33:31 +02:00
|
|
|
if (pages == -1) {
|
2016-05-05 09:32:55 +02:00
|
|
|
/* Make sure the first page is sent out before other pages */
|
|
|
|
bytes_xmit = save_page_header(f, block, offset |
|
|
|
|
RAM_SAVE_FLAG_COMPRESS_PAGE);
|
|
|
|
blen = qemu_put_compression_data(f, p, TARGET_PAGE_SIZE,
|
|
|
|
migrate_compress_level());
|
|
|
|
if (blen > 0) {
|
|
|
|
*bytes_transferred += bytes_xmit + blen;
|
2017-03-13 20:40:53 +01:00
|
|
|
rs->norm_pages++;
|
2016-05-05 09:32:54 +02:00
|
|
|
pages = 1;
|
2016-05-05 09:32:55 +02:00
|
|
|
} else {
|
|
|
|
qemu_file_set_error(f, blen);
|
|
|
|
error_report("compressed data failed!");
|
2016-05-05 09:32:54 +02:00
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2017-02-03 16:23:20 +01:00
|
|
|
if (pages > 0) {
|
|
|
|
ram_release_pages(ms, block->idstr, pss->offset, pages);
|
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
} else {
|
2016-05-05 09:32:55 +02:00
|
|
|
offset |= RAM_SAVE_FLAG_CONTINUE;
|
2017-03-13 20:30:21 +01:00
|
|
|
pages = save_zero_page(rs, f, block, offset, p, bytes_transferred);
|
2015-05-07 19:33:31 +02:00
|
|
|
if (pages == -1) {
|
2017-03-13 20:40:53 +01:00
|
|
|
pages = compress_page_with_multi_thread(rs, f, block, offset,
|
2015-05-07 19:33:31 +02:00
|
|
|
bytes_transferred);
|
2017-02-03 16:23:20 +01:00
|
|
|
} else {
|
|
|
|
ram_release_pages(ms, block->idstr, pss->offset, pages);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return pages;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* find_dirty_block: find the next dirty page and update any state
|
|
|
|
* associated with the search process.
|
2015-09-23 16:27:11 +02:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns if a page is found
|
2015-09-23 16:27:11 +02:00
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2017-03-23 15:06:39 +01:00
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @pss: data about the state of the current dirty page scan
|
|
|
|
* @again: set to false if the search has scanned the whole of RAM
|
|
|
|
* @ram_addr_abs: pointer into which to store the address of the dirty page
|
|
|
|
* within the global ram_addr space
|
2015-09-23 16:27:11 +02:00
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
static bool find_dirty_block(RAMState *rs, QEMUFile *f, PageSearchStatus *pss,
|
2015-11-05 19:11:01 +01:00
|
|
|
bool *again, ram_addr_t *ram_addr_abs)
|
2015-09-23 16:27:11 +02:00
|
|
|
{
|
2017-03-13 19:26:29 +01:00
|
|
|
pss->offset = migration_bitmap_find_dirty(rs, pss->block, pss->offset,
|
2015-11-05 19:11:09 +01:00
|
|
|
ram_addr_abs);
|
2017-03-13 19:26:29 +01:00
|
|
|
if (pss->complete_round && pss->block == rs->last_seen_block &&
|
|
|
|
pss->offset >= rs->last_offset) {
|
2015-09-23 16:27:11 +02:00
|
|
|
/*
|
|
|
|
* We've been once around the RAM and haven't found anything.
|
|
|
|
* Give up.
|
|
|
|
*/
|
|
|
|
*again = false;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (pss->offset >= pss->block->used_length) {
|
|
|
|
/* Didn't find anything in this RAM Block */
|
|
|
|
pss->offset = 0;
|
|
|
|
pss->block = QLIST_NEXT_RCU(pss->block, next);
|
|
|
|
if (!pss->block) {
|
|
|
|
/* Hit the end of the list */
|
|
|
|
pss->block = QLIST_FIRST_RCU(&ram_list.blocks);
|
|
|
|
/* Flag that we've looped */
|
|
|
|
pss->complete_round = true;
|
2017-03-13 19:26:29 +01:00
|
|
|
rs->ram_bulk_stage = false;
|
2015-09-23 16:27:11 +02:00
|
|
|
if (migrate_use_xbzrle()) {
|
|
|
|
/* If xbzrle is on, stop using the data compression at this
|
|
|
|
* point. In theory, xbzrle can do better than compression.
|
|
|
|
*/
|
|
|
|
flush_compressed_data(f);
|
|
|
|
compression_switch = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
/* Didn't find anything this time, but try again on the new block */
|
|
|
|
*again = true;
|
|
|
|
return false;
|
|
|
|
} else {
|
|
|
|
/* Can go around again, but... */
|
|
|
|
*again = true;
|
|
|
|
/* We've found something so probably don't need to */
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* unqueue_page: gets a page of the queue
|
|
|
|
*
|
2015-11-05 19:11:09 +01:00
|
|
|
* Helper for 'get_queued_page' - gets a page off the queue
|
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns the block of the page (or NULL if none available)
|
|
|
|
*
|
|
|
|
* @ms: current migration state
|
|
|
|
* @offset: used to return the offset within the RAMBlock
|
|
|
|
* @ram_addr_abs: pointer into which to store the address of the dirty page
|
|
|
|
* within the global ram_addr space
|
2015-11-05 19:11:09 +01:00
|
|
|
*/
|
|
|
|
static RAMBlock *unqueue_page(MigrationState *ms, ram_addr_t *offset,
|
|
|
|
ram_addr_t *ram_addr_abs)
|
|
|
|
{
|
|
|
|
RAMBlock *block = NULL;
|
|
|
|
|
|
|
|
qemu_mutex_lock(&ms->src_page_req_mutex);
|
|
|
|
if (!QSIMPLEQ_EMPTY(&ms->src_page_requests)) {
|
|
|
|
struct MigrationSrcPageRequest *entry =
|
|
|
|
QSIMPLEQ_FIRST(&ms->src_page_requests);
|
|
|
|
block = entry->rb;
|
|
|
|
*offset = entry->offset;
|
|
|
|
*ram_addr_abs = (entry->offset + entry->rb->offset) &
|
|
|
|
TARGET_PAGE_MASK;
|
|
|
|
|
|
|
|
if (entry->len > TARGET_PAGE_SIZE) {
|
|
|
|
entry->len -= TARGET_PAGE_SIZE;
|
|
|
|
entry->offset += TARGET_PAGE_SIZE;
|
|
|
|
} else {
|
|
|
|
memory_region_unref(block->mr);
|
|
|
|
QSIMPLEQ_REMOVE_HEAD(&ms->src_page_requests, next_req);
|
|
|
|
g_free(entry);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
qemu_mutex_unlock(&ms->src_page_req_mutex);
|
|
|
|
|
|
|
|
return block;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* get_queued_page: unqueue a page from the postocpy requests
|
|
|
|
*
|
|
|
|
* Skips pages that are already sent (!dirty)
|
2015-11-05 19:11:09 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns if a queued page is found
|
2015-11-05 19:11:09 +01:00
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2017-03-23 15:06:39 +01:00
|
|
|
* @ms: current migration state
|
|
|
|
* @pss: data about the state of the current dirty page scan
|
|
|
|
* @ram_addr_abs: pointer into which to store the address of the dirty page
|
|
|
|
* within the global ram_addr space
|
2015-11-05 19:11:09 +01:00
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
static bool get_queued_page(RAMState *rs, MigrationState *ms,
|
|
|
|
PageSearchStatus *pss,
|
2015-11-05 19:11:09 +01:00
|
|
|
ram_addr_t *ram_addr_abs)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
ram_addr_t offset;
|
|
|
|
bool dirty;
|
|
|
|
|
|
|
|
do {
|
|
|
|
block = unqueue_page(ms, &offset, ram_addr_abs);
|
|
|
|
/*
|
|
|
|
* We're sending this page, and since it's postcopy nothing else
|
|
|
|
* will dirty it, and we must make sure it doesn't get sent again
|
|
|
|
* even if this queue request was received after the background
|
|
|
|
* search already sent it.
|
|
|
|
*/
|
|
|
|
if (block) {
|
|
|
|
unsigned long *bitmap;
|
|
|
|
bitmap = atomic_rcu_read(&migration_bitmap_rcu)->bmap;
|
|
|
|
dirty = test_bit(*ram_addr_abs >> TARGET_PAGE_BITS, bitmap);
|
|
|
|
if (!dirty) {
|
|
|
|
trace_get_queued_page_not_dirty(
|
|
|
|
block->idstr, (uint64_t)offset,
|
|
|
|
(uint64_t)*ram_addr_abs,
|
|
|
|
test_bit(*ram_addr_abs >> TARGET_PAGE_BITS,
|
|
|
|
atomic_rcu_read(&migration_bitmap_rcu)->unsentmap));
|
|
|
|
} else {
|
|
|
|
trace_get_queued_page(block->idstr,
|
|
|
|
(uint64_t)offset,
|
|
|
|
(uint64_t)*ram_addr_abs);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
} while (block && !dirty);
|
|
|
|
|
|
|
|
if (block) {
|
|
|
|
/*
|
|
|
|
* As soon as we start servicing pages out of order, then we have
|
|
|
|
* to kill the bulk stage, since the bulk stage assumes
|
|
|
|
* in (migration_bitmap_find_and_reset_dirty) that every page is
|
|
|
|
* dirty, that's no longer true.
|
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
rs->ram_bulk_stage = false;
|
2015-11-05 19:11:09 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* We want the background search to continue from the queued page
|
|
|
|
* since the guest is likely to want other pages near to the page
|
|
|
|
* it just requested.
|
|
|
|
*/
|
|
|
|
pss->block = block;
|
|
|
|
pss->offset = offset;
|
|
|
|
}
|
|
|
|
|
|
|
|
return !!block;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:08 +01:00
|
|
|
/**
|
2017-04-03 22:06:54 +02:00
|
|
|
* migration_page_queue_free: drop any remaining pages in the ram
|
|
|
|
* request queue
|
2015-11-05 19:11:08 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* It should be empty at the end anyway, but in error cases there may
|
|
|
|
* be some left. in case that there is any page left, we drop it.
|
|
|
|
*
|
|
|
|
* @ms: current migration state
|
2015-11-05 19:11:08 +01:00
|
|
|
*/
|
2017-04-03 22:06:54 +02:00
|
|
|
void migration_page_queue_free(MigrationState *ms)
|
2015-11-05 19:11:08 +01:00
|
|
|
{
|
|
|
|
struct MigrationSrcPageRequest *mspr, *next_mspr;
|
|
|
|
/* This queue generally should be empty - but in the case of a failed
|
|
|
|
* migration might have some droppings in.
|
|
|
|
*/
|
|
|
|
rcu_read_lock();
|
|
|
|
QSIMPLEQ_FOREACH_SAFE(mspr, &ms->src_page_requests, next_req, next_mspr) {
|
|
|
|
memory_region_unref(mspr->rb->mr);
|
|
|
|
QSIMPLEQ_REMOVE_HEAD(&ms->src_page_requests, next_req);
|
|
|
|
g_free(mspr);
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2017-03-23 15:06:39 +01:00
|
|
|
* ram_save_queue_pages: queue the page for transmission
|
|
|
|
*
|
|
|
|
* A request from postcopy destination for example.
|
|
|
|
*
|
|
|
|
* Returns zero on success or negative on error
|
|
|
|
*
|
|
|
|
* @ms: current migration state
|
|
|
|
* @rbname: Name of the RAMBLock of the request. NULL means the
|
|
|
|
* same that last one.
|
|
|
|
* @start: starting address from the start of the RAMBlock
|
|
|
|
* @len: length (in bytes) to send
|
2015-11-05 19:11:08 +01:00
|
|
|
*/
|
|
|
|
int ram_save_queue_pages(MigrationState *ms, const char *rbname,
|
|
|
|
ram_addr_t start, ram_addr_t len)
|
|
|
|
{
|
|
|
|
RAMBlock *ramblock;
|
|
|
|
|
2016-06-13 13:16:42 +02:00
|
|
|
ms->postcopy_requests++;
|
2015-11-05 19:11:08 +01:00
|
|
|
rcu_read_lock();
|
|
|
|
if (!rbname) {
|
|
|
|
/* Reuse last RAMBlock */
|
|
|
|
ramblock = ms->last_req_rb;
|
|
|
|
|
|
|
|
if (!ramblock) {
|
|
|
|
/*
|
|
|
|
* Shouldn't happen, we can't reuse the last RAMBlock if
|
|
|
|
* it's the 1st request.
|
|
|
|
*/
|
|
|
|
error_report("ram_save_queue_pages no previous block");
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
ramblock = qemu_ram_block_by_name(rbname);
|
|
|
|
|
|
|
|
if (!ramblock) {
|
|
|
|
/* We shouldn't be asked for a non-existent RAMBlock */
|
|
|
|
error_report("ram_save_queue_pages no block '%s'", rbname);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
ms->last_req_rb = ramblock;
|
|
|
|
}
|
|
|
|
trace_ram_save_queue_pages(ramblock->idstr, start, len);
|
|
|
|
if (start+len > ramblock->used_length) {
|
2015-11-10 17:42:05 +01:00
|
|
|
error_report("%s request overrun start=" RAM_ADDR_FMT " len="
|
|
|
|
RAM_ADDR_FMT " blocklen=" RAM_ADDR_FMT,
|
2015-11-05 19:11:08 +01:00
|
|
|
__func__, start, len, ramblock->used_length);
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct MigrationSrcPageRequest *new_entry =
|
|
|
|
g_malloc0(sizeof(struct MigrationSrcPageRequest));
|
|
|
|
new_entry->rb = ramblock;
|
|
|
|
new_entry->offset = start;
|
|
|
|
new_entry->len = len;
|
|
|
|
|
|
|
|
memory_region_ref(ramblock->mr);
|
|
|
|
qemu_mutex_lock(&ms->src_page_req_mutex);
|
|
|
|
QSIMPLEQ_INSERT_TAIL(&ms->src_page_requests, new_entry, next_req);
|
|
|
|
qemu_mutex_unlock(&ms->src_page_req_mutex);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
err:
|
|
|
|
rcu_read_unlock();
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:09 +01:00
|
|
|
/**
|
2017-03-23 15:06:39 +01:00
|
|
|
* ram_save_target_page: save one target page
|
2015-11-05 19:11:09 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns the number of pages written
|
2015-11-05 19:11:09 +01:00
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2017-03-23 15:06:39 +01:00
|
|
|
* @ms: current migration state
|
2015-11-05 19:11:09 +01:00
|
|
|
* @f: QEMUFile where to send the data
|
2017-03-23 15:06:39 +01:00
|
|
|
* @pss: data about the page we want to send
|
2015-11-05 19:11:09 +01:00
|
|
|
* @last_stage: if we are at the completion stage
|
|
|
|
* @bytes_transferred: increase it with the number of transferred bytes
|
2017-03-23 15:06:39 +01:00
|
|
|
* @dirty_ram_abs: address of the start of the dirty page in ram_addr_t space
|
2015-11-05 19:11:09 +01:00
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
static int ram_save_target_page(RAMState *rs, MigrationState *ms, QEMUFile *f,
|
2016-01-15 04:37:44 +01:00
|
|
|
PageSearchStatus *pss,
|
2015-11-05 19:11:09 +01:00
|
|
|
bool last_stage,
|
|
|
|
uint64_t *bytes_transferred,
|
|
|
|
ram_addr_t dirty_ram_abs)
|
|
|
|
{
|
|
|
|
int res = 0;
|
|
|
|
|
|
|
|
/* Check the pages is dirty and if it is send it */
|
|
|
|
if (migration_bitmap_clear_dirty(dirty_ram_abs)) {
|
|
|
|
unsigned long *unsentmap;
|
|
|
|
if (compression_switch && migrate_use_compression()) {
|
2017-03-13 19:26:29 +01:00
|
|
|
res = ram_save_compressed_page(rs, ms, f, pss,
|
2015-11-05 19:11:09 +01:00
|
|
|
last_stage,
|
|
|
|
bytes_transferred);
|
|
|
|
} else {
|
2017-03-13 19:26:29 +01:00
|
|
|
res = ram_save_page(rs, ms, f, pss, last_stage,
|
2015-11-05 19:11:09 +01:00
|
|
|
bytes_transferred);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (res < 0) {
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
unsentmap = atomic_rcu_read(&migration_bitmap_rcu)->unsentmap;
|
|
|
|
if (unsentmap) {
|
|
|
|
clear_bit(dirty_ram_abs >> TARGET_PAGE_BITS, unsentmap);
|
|
|
|
}
|
2015-12-10 17:31:46 +01:00
|
|
|
/* Only update last_sent_block if a block was actually sent; xbzrle
|
|
|
|
* might have decided the page was identical so didn't bother writing
|
|
|
|
* to the stream.
|
|
|
|
*/
|
|
|
|
if (res > 0) {
|
2017-03-13 19:26:29 +01:00
|
|
|
rs->last_sent_block = pss->block;
|
2015-12-10 17:31:46 +01:00
|
|
|
}
|
2015-11-05 19:11:09 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2017-03-23 15:06:39 +01:00
|
|
|
* ram_save_host_page: save a whole host page
|
2015-11-05 19:11:09 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Starting at *offset send pages up to the end of the current host
|
|
|
|
* page. It's valid for the initial offset to point into the middle of
|
|
|
|
* a host page in which case the remainder of the hostpage is sent.
|
|
|
|
* Only dirty target pages are sent. Note that the host page size may
|
|
|
|
* be a huge page for this block.
|
2015-11-05 19:11:09 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns the number of pages written or negative on error
|
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2017-03-23 15:06:39 +01:00
|
|
|
* @ms: current migration state
|
2015-11-05 19:11:09 +01:00
|
|
|
* @f: QEMUFile where to send the data
|
2017-03-23 15:06:39 +01:00
|
|
|
* @pss: data about the page we want to send
|
2015-11-05 19:11:09 +01:00
|
|
|
* @last_stage: if we are at the completion stage
|
|
|
|
* @bytes_transferred: increase it with the number of transferred bytes
|
|
|
|
* @dirty_ram_abs: Address of the start of the dirty page in ram_addr_t space
|
|
|
|
*/
|
2017-03-13 19:26:29 +01:00
|
|
|
static int ram_save_host_page(RAMState *rs, MigrationState *ms, QEMUFile *f,
|
2016-01-15 04:37:44 +01:00
|
|
|
PageSearchStatus *pss,
|
|
|
|
bool last_stage,
|
2015-11-05 19:11:09 +01:00
|
|
|
uint64_t *bytes_transferred,
|
|
|
|
ram_addr_t dirty_ram_abs)
|
|
|
|
{
|
|
|
|
int tmppages, pages = 0;
|
2017-02-24 19:28:39 +01:00
|
|
|
size_t pagesize = qemu_ram_pagesize(pss->block);
|
|
|
|
|
2015-11-05 19:11:09 +01:00
|
|
|
do {
|
2017-03-13 19:26:29 +01:00
|
|
|
tmppages = ram_save_target_page(rs, ms, f, pss, last_stage,
|
2015-11-05 19:11:09 +01:00
|
|
|
bytes_transferred, dirty_ram_abs);
|
|
|
|
if (tmppages < 0) {
|
|
|
|
return tmppages;
|
|
|
|
}
|
|
|
|
|
|
|
|
pages += tmppages;
|
2016-01-15 04:37:44 +01:00
|
|
|
pss->offset += TARGET_PAGE_SIZE;
|
2015-11-05 19:11:09 +01:00
|
|
|
dirty_ram_abs += TARGET_PAGE_SIZE;
|
2017-02-24 19:28:39 +01:00
|
|
|
} while (pss->offset & (pagesize - 1));
|
2015-11-05 19:11:09 +01:00
|
|
|
|
|
|
|
/* The offset we leave with is the last one we looked at */
|
2016-01-15 04:37:44 +01:00
|
|
|
pss->offset -= TARGET_PAGE_SIZE;
|
2015-11-05 19:11:09 +01:00
|
|
|
return pages;
|
|
|
|
}
|
2015-11-05 19:11:08 +01:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
/**
|
2017-03-23 15:06:39 +01:00
|
|
|
* ram_find_and_save_block: finds a dirty page and sends it to f
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
|
|
|
* Called within an RCU critical section.
|
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns the number of pages written where zero means no dirty pages
|
2015-05-07 19:33:31 +02:00
|
|
|
*
|
2017-03-13 19:26:29 +01:00
|
|
|
* @rs: current RAM state
|
2015-05-07 19:33:31 +02:00
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @last_stage: if we are at the completion stage
|
|
|
|
* @bytes_transferred: increase it with the number of transferred bytes
|
2015-11-05 19:11:09 +01:00
|
|
|
*
|
|
|
|
* On systems where host-page-size > target-page-size it will send all the
|
|
|
|
* pages in a host page that are dirty.
|
2015-05-07 19:33:31 +02:00
|
|
|
*/
|
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
static int ram_find_and_save_block(RAMState *rs, QEMUFile *f, bool last_stage,
|
2015-05-07 19:33:31 +02:00
|
|
|
uint64_t *bytes_transferred)
|
|
|
|
{
|
2015-09-23 16:27:10 +02:00
|
|
|
PageSearchStatus pss;
|
2015-11-05 19:11:09 +01:00
|
|
|
MigrationState *ms = migrate_get_current();
|
2015-05-07 19:33:31 +02:00
|
|
|
int pages = 0;
|
2015-09-23 16:27:11 +02:00
|
|
|
bool again, found;
|
2015-11-05 19:11:01 +01:00
|
|
|
ram_addr_t dirty_ram_abs; /* Address of the start of the dirty page in
|
|
|
|
ram_addr_t space */
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2017-02-08 15:28:45 +01:00
|
|
|
/* No dirty page as there is zero RAM */
|
|
|
|
if (!ram_bytes_total()) {
|
|
|
|
return pages;
|
|
|
|
}
|
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
pss.block = rs->last_seen_block;
|
|
|
|
pss.offset = rs->last_offset;
|
2015-09-23 16:27:10 +02:00
|
|
|
pss.complete_round = false;
|
|
|
|
|
|
|
|
if (!pss.block) {
|
|
|
|
pss.block = QLIST_FIRST_RCU(&ram_list.blocks);
|
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2015-09-23 16:27:11 +02:00
|
|
|
do {
|
2015-11-05 19:11:09 +01:00
|
|
|
again = true;
|
2017-03-13 19:26:29 +01:00
|
|
|
found = get_queued_page(rs, ms, &pss, &dirty_ram_abs);
|
2015-09-23 16:27:11 +02:00
|
|
|
|
2015-11-05 19:11:09 +01:00
|
|
|
if (!found) {
|
|
|
|
/* priority queue empty, so just search for something dirty */
|
2017-03-13 19:26:29 +01:00
|
|
|
found = find_dirty_block(rs, f, &pss, &again, &dirty_ram_abs);
|
2015-11-05 19:11:09 +01:00
|
|
|
}
|
2015-11-05 19:11:01 +01:00
|
|
|
|
2015-11-05 19:11:09 +01:00
|
|
|
if (found) {
|
2017-03-13 19:26:29 +01:00
|
|
|
pages = ram_save_host_page(rs, ms, f, &pss,
|
2015-11-05 19:11:09 +01:00
|
|
|
last_stage, bytes_transferred,
|
|
|
|
dirty_ram_abs);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2015-09-23 16:27:11 +02:00
|
|
|
} while (!pages && again);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
rs->last_seen_block = pss.block;
|
|
|
|
rs->last_offset = pss.offset;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
return pages;
|
|
|
|
}
|
|
|
|
|
|
|
|
void acct_update_position(QEMUFile *f, size_t size, bool zero)
|
|
|
|
{
|
|
|
|
uint64_t pages = size / TARGET_PAGE_SIZE;
|
2017-03-13 20:30:21 +01:00
|
|
|
RAMState *rs = &ram_state;
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
if (zero) {
|
2017-03-13 20:30:21 +01:00
|
|
|
rs->zero_pages += pages;
|
2015-05-07 19:33:31 +02:00
|
|
|
} else {
|
2017-03-13 20:40:53 +01:00
|
|
|
rs->norm_pages += pages;
|
2015-05-07 19:33:31 +02:00
|
|
|
bytes_transferred += size;
|
|
|
|
qemu_update_position(f, size);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static ram_addr_t ram_save_remaining(void)
|
|
|
|
{
|
|
|
|
return migration_dirty_pages;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t ram_bytes_remaining(void)
|
|
|
|
{
|
|
|
|
return ram_save_remaining() * TARGET_PAGE_SIZE;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t ram_bytes_transferred(void)
|
|
|
|
{
|
|
|
|
return bytes_transferred;
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t ram_bytes_total(void)
|
|
|
|
{
|
|
|
|
RAMBlock *block;
|
|
|
|
uint64_t total = 0;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
QLIST_FOREACH_RCU(block, &ram_list.blocks, next)
|
|
|
|
total += block->used_length;
|
|
|
|
rcu_read_unlock();
|
|
|
|
return total;
|
|
|
|
}
|
|
|
|
|
|
|
|
void free_xbzrle_decoded_buf(void)
|
|
|
|
{
|
|
|
|
g_free(xbzrle_decoded_buf);
|
|
|
|
xbzrle_decoded_buf = NULL;
|
|
|
|
}
|
|
|
|
|
2015-09-28 13:41:58 +02:00
|
|
|
static void migration_bitmap_free(struct BitmapRcu *bmap)
|
|
|
|
{
|
|
|
|
g_free(bmap->bmap);
|
2015-11-05 19:11:01 +01:00
|
|
|
g_free(bmap->unsentmap);
|
2015-09-28 13:41:58 +02:00
|
|
|
g_free(bmap);
|
|
|
|
}
|
|
|
|
|
2015-11-02 08:37:03 +01:00
|
|
|
static void ram_migration_cleanup(void *opaque)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
2015-07-02 14:18:05 +02:00
|
|
|
/* caller have hold iothread lock or is in a bh, so there is
|
|
|
|
* no writing race against this migration_bitmap
|
|
|
|
*/
|
2015-09-28 13:41:58 +02:00
|
|
|
struct BitmapRcu *bitmap = migration_bitmap_rcu;
|
|
|
|
atomic_rcu_set(&migration_bitmap_rcu, NULL);
|
2015-07-02 14:18:05 +02:00
|
|
|
if (bitmap) {
|
2015-05-07 19:33:31 +02:00
|
|
|
memory_global_dirty_log_stop();
|
2015-09-28 13:41:58 +02:00
|
|
|
call_rcu(bitmap, migration_bitmap_free, rcu);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
XBZRLE_cache_lock();
|
|
|
|
if (XBZRLE.cache) {
|
|
|
|
cache_fini(XBZRLE.cache);
|
|
|
|
g_free(XBZRLE.encoded_buf);
|
|
|
|
g_free(XBZRLE.current_buf);
|
2016-10-24 17:26:49 +02:00
|
|
|
g_free(ZERO_TARGET_PAGE);
|
2015-05-07 19:33:31 +02:00
|
|
|
XBZRLE.cache = NULL;
|
|
|
|
XBZRLE.encoded_buf = NULL;
|
|
|
|
XBZRLE.current_buf = NULL;
|
|
|
|
}
|
|
|
|
XBZRLE_cache_unlock();
|
|
|
|
}
|
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
static void ram_state_reset(RAMState *rs)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
2017-03-13 19:26:29 +01:00
|
|
|
rs->last_seen_block = NULL;
|
|
|
|
rs->last_sent_block = NULL;
|
|
|
|
rs->last_offset = 0;
|
|
|
|
rs->last_version = ram_list.version;
|
|
|
|
rs->ram_bulk_stage = true;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
#define MAX_WAIT 50 /* ms, half buffered_file limit */
|
|
|
|
|
2015-07-02 14:18:06 +02:00
|
|
|
void migration_bitmap_extend(ram_addr_t old, ram_addr_t new)
|
|
|
|
{
|
|
|
|
/* called in qemu main thread, so there is
|
|
|
|
* no writing race against this migration_bitmap
|
|
|
|
*/
|
2015-09-28 13:41:58 +02:00
|
|
|
if (migration_bitmap_rcu) {
|
|
|
|
struct BitmapRcu *old_bitmap = migration_bitmap_rcu, *bitmap;
|
|
|
|
bitmap = g_new(struct BitmapRcu, 1);
|
|
|
|
bitmap->bmap = bitmap_new(new);
|
2015-07-02 14:18:06 +02:00
|
|
|
|
|
|
|
/* prevent migration_bitmap content from being set bit
|
|
|
|
* by migration_bitmap_sync_range() at the same time.
|
|
|
|
* it is safe to migration if migration_bitmap is cleared bit
|
|
|
|
* at the same time.
|
|
|
|
*/
|
|
|
|
qemu_mutex_lock(&migration_bitmap_mutex);
|
2015-09-28 13:41:58 +02:00
|
|
|
bitmap_copy(bitmap->bmap, old_bitmap->bmap, old);
|
|
|
|
bitmap_set(bitmap->bmap, old, new - old);
|
2015-11-05 19:11:01 +01:00
|
|
|
|
|
|
|
/* We don't have a way to safely extend the sentmap
|
|
|
|
* with RCU; so mark it as missing, entry to postcopy
|
|
|
|
* will fail.
|
|
|
|
*/
|
|
|
|
bitmap->unsentmap = NULL;
|
|
|
|
|
2015-09-28 13:41:58 +02:00
|
|
|
atomic_rcu_set(&migration_bitmap_rcu, bitmap);
|
2015-07-02 14:18:06 +02:00
|
|
|
qemu_mutex_unlock(&migration_bitmap_mutex);
|
|
|
|
migration_dirty_pages += new - old;
|
2015-09-28 13:41:58 +02:00
|
|
|
call_rcu(old_bitmap, migration_bitmap_free, rcu);
|
2015-07-02 14:18:06 +02:00
|
|
|
}
|
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2015-11-05 19:10:38 +01:00
|
|
|
/*
|
|
|
|
* 'expected' is the value you expect the bitmap mostly to be full
|
|
|
|
* of; it won't bother printing lines that are all this value.
|
|
|
|
* If 'todump' is null the migration bitmap is dumped.
|
|
|
|
*/
|
|
|
|
void ram_debug_dump_bitmap(unsigned long *todump, bool expected)
|
|
|
|
{
|
|
|
|
int64_t ram_pages = last_ram_offset() >> TARGET_PAGE_BITS;
|
|
|
|
|
|
|
|
int64_t cur;
|
|
|
|
int64_t linelen = 128;
|
|
|
|
char linebuf[129];
|
|
|
|
|
|
|
|
if (!todump) {
|
|
|
|
todump = atomic_rcu_read(&migration_bitmap_rcu)->bmap;
|
|
|
|
}
|
|
|
|
|
|
|
|
for (cur = 0; cur < ram_pages; cur += linelen) {
|
|
|
|
int64_t curb;
|
|
|
|
bool found = false;
|
|
|
|
/*
|
|
|
|
* Last line; catch the case where the line length
|
|
|
|
* is longer than remaining ram
|
|
|
|
*/
|
|
|
|
if (cur + linelen > ram_pages) {
|
|
|
|
linelen = ram_pages - cur;
|
|
|
|
}
|
|
|
|
for (curb = 0; curb < linelen; curb++) {
|
|
|
|
bool thisbit = test_bit(cur + curb, todump);
|
|
|
|
linebuf[curb] = thisbit ? '1' : '.';
|
|
|
|
found = found || (thisbit != expected);
|
|
|
|
}
|
|
|
|
if (found) {
|
|
|
|
linebuf[curb] = '\0';
|
|
|
|
fprintf(stderr, "0x%08" PRIx64 " : %s\n", cur, linebuf);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:02 +01:00
|
|
|
/* **** functions for postcopy ***** */
|
|
|
|
|
2017-02-03 16:23:21 +01:00
|
|
|
void ram_postcopy_migrated_memory_release(MigrationState *ms)
|
|
|
|
{
|
|
|
|
struct RAMBlock *block;
|
|
|
|
unsigned long *bitmap = atomic_rcu_read(&migration_bitmap_rcu)->bmap;
|
|
|
|
|
|
|
|
QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
|
|
|
|
unsigned long first = block->offset >> TARGET_PAGE_BITS;
|
|
|
|
unsigned long range = first + (block->used_length >> TARGET_PAGE_BITS);
|
|
|
|
unsigned long run_start = find_next_zero_bit(bitmap, range, first);
|
|
|
|
|
|
|
|
while (run_start < range) {
|
|
|
|
unsigned long run_end = find_next_bit(bitmap, range, run_start + 1);
|
|
|
|
ram_discard_range(NULL, block->idstr, run_start << TARGET_PAGE_BITS,
|
|
|
|
(run_end - run_start) << TARGET_PAGE_BITS);
|
|
|
|
run_start = find_next_zero_bit(bitmap, range, run_end + 1);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* postcopy_send_discard_bm_ram: discard a RAMBlock
|
|
|
|
*
|
|
|
|
* Returns zero on success
|
|
|
|
*
|
2015-11-05 19:11:02 +01:00
|
|
|
* Callback from postcopy_each_ram_send_discard for each RAMBlock
|
|
|
|
* Note: At this point the 'unsentmap' is the processed bitmap combined
|
|
|
|
* with the dirtymap; so a '1' means it's either dirty or unsent.
|
2017-03-23 15:06:39 +01:00
|
|
|
*
|
|
|
|
* @ms: current migration state
|
|
|
|
* @pds: state for postcopy
|
|
|
|
* @start: RAMBlock starting page
|
|
|
|
* @length: RAMBlock size
|
2015-11-05 19:11:02 +01:00
|
|
|
*/
|
|
|
|
static int postcopy_send_discard_bm_ram(MigrationState *ms,
|
|
|
|
PostcopyDiscardState *pds,
|
|
|
|
unsigned long start,
|
|
|
|
unsigned long length)
|
|
|
|
{
|
|
|
|
unsigned long end = start + length; /* one after the end */
|
|
|
|
unsigned long current;
|
|
|
|
unsigned long *unsentmap;
|
|
|
|
|
|
|
|
unsentmap = atomic_rcu_read(&migration_bitmap_rcu)->unsentmap;
|
|
|
|
for (current = start; current < end; ) {
|
|
|
|
unsigned long one = find_next_bit(unsentmap, end, current);
|
|
|
|
|
|
|
|
if (one <= end) {
|
|
|
|
unsigned long zero = find_next_zero_bit(unsentmap, end, one + 1);
|
|
|
|
unsigned long discard_length;
|
|
|
|
|
|
|
|
if (zero >= end) {
|
|
|
|
discard_length = end - one;
|
|
|
|
} else {
|
|
|
|
discard_length = zero - one;
|
|
|
|
}
|
2016-06-13 13:16:40 +02:00
|
|
|
if (discard_length) {
|
|
|
|
postcopy_discard_send_range(ms, pds, one, discard_length);
|
|
|
|
}
|
2015-11-05 19:11:02 +01:00
|
|
|
current = one + discard_length;
|
|
|
|
} else {
|
|
|
|
current = one;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* postcopy_each_ram_send_discard: discard all RAMBlocks
|
|
|
|
*
|
|
|
|
* Returns 0 for success or negative for error
|
|
|
|
*
|
2015-11-05 19:11:02 +01:00
|
|
|
* Utility for the outgoing postcopy code.
|
|
|
|
* Calls postcopy_send_discard_bm_ram for each RAMBlock
|
|
|
|
* passing it bitmap indexes and name.
|
|
|
|
* (qemu_ram_foreach_block ends up passing unscaled lengths
|
|
|
|
* which would mean postcopy code would have to deal with target page)
|
2017-03-23 15:06:39 +01:00
|
|
|
*
|
|
|
|
* @ms: current migration state
|
2015-11-05 19:11:02 +01:00
|
|
|
*/
|
|
|
|
static int postcopy_each_ram_send_discard(MigrationState *ms)
|
|
|
|
{
|
|
|
|
struct RAMBlock *block;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
|
|
|
|
unsigned long first = block->offset >> TARGET_PAGE_BITS;
|
|
|
|
PostcopyDiscardState *pds = postcopy_discard_send_init(ms,
|
|
|
|
first,
|
|
|
|
block->idstr);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Postcopy sends chunks of bitmap over the wire, but it
|
|
|
|
* just needs indexes at this point, avoids it having
|
|
|
|
* target page specific code.
|
|
|
|
*/
|
|
|
|
ret = postcopy_send_discard_bm_ram(ms, pds, first,
|
|
|
|
block->used_length >> TARGET_PAGE_BITS);
|
|
|
|
postcopy_discard_send_finish(ms, pds);
|
|
|
|
if (ret) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* postcopy_chunk_hostpages_pass: canocalize bitmap in hostpages
|
|
|
|
*
|
|
|
|
* Helper for postcopy_chunk_hostpages; it's called twice to
|
|
|
|
* canonicalize the two bitmaps, that are similar, but one is
|
|
|
|
* inverted.
|
2015-11-05 19:11:15 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Postcopy requires that all target pages in a hostpage are dirty or
|
|
|
|
* clean, not a mix. This function canonicalizes the bitmaps.
|
2015-11-05 19:11:15 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* @ms: current migration state
|
|
|
|
* @unsent_pass: if true we need to canonicalize partially unsent host pages
|
|
|
|
* otherwise we need to canonicalize partially dirty host pages
|
|
|
|
* @block: block that contains the page we want to canonicalize
|
|
|
|
* @pds: state for postcopy
|
2015-11-05 19:11:15 +01:00
|
|
|
*/
|
|
|
|
static void postcopy_chunk_hostpages_pass(MigrationState *ms, bool unsent_pass,
|
|
|
|
RAMBlock *block,
|
|
|
|
PostcopyDiscardState *pds)
|
|
|
|
{
|
|
|
|
unsigned long *bitmap;
|
|
|
|
unsigned long *unsentmap;
|
2017-02-24 19:28:31 +01:00
|
|
|
unsigned int host_ratio = block->page_size / TARGET_PAGE_SIZE;
|
2015-11-05 19:11:15 +01:00
|
|
|
unsigned long first = block->offset >> TARGET_PAGE_BITS;
|
|
|
|
unsigned long len = block->used_length >> TARGET_PAGE_BITS;
|
|
|
|
unsigned long last = first + (len - 1);
|
|
|
|
unsigned long run_start;
|
|
|
|
|
2017-02-24 19:28:31 +01:00
|
|
|
if (block->page_size == TARGET_PAGE_SIZE) {
|
|
|
|
/* Easy case - TPS==HPS for a non-huge page RAMBlock */
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:15 +01:00
|
|
|
bitmap = atomic_rcu_read(&migration_bitmap_rcu)->bmap;
|
|
|
|
unsentmap = atomic_rcu_read(&migration_bitmap_rcu)->unsentmap;
|
|
|
|
|
|
|
|
if (unsent_pass) {
|
|
|
|
/* Find a sent page */
|
|
|
|
run_start = find_next_zero_bit(unsentmap, last + 1, first);
|
|
|
|
} else {
|
|
|
|
/* Find a dirty page */
|
|
|
|
run_start = find_next_bit(bitmap, last + 1, first);
|
|
|
|
}
|
|
|
|
|
|
|
|
while (run_start <= last) {
|
|
|
|
bool do_fixup = false;
|
|
|
|
unsigned long fixup_start_addr;
|
|
|
|
unsigned long host_offset;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the start of this run of pages is in the middle of a host
|
|
|
|
* page, then we need to fixup this host page.
|
|
|
|
*/
|
|
|
|
host_offset = run_start % host_ratio;
|
|
|
|
if (host_offset) {
|
|
|
|
do_fixup = true;
|
|
|
|
run_start -= host_offset;
|
|
|
|
fixup_start_addr = run_start;
|
|
|
|
/* For the next pass */
|
|
|
|
run_start = run_start + host_ratio;
|
|
|
|
} else {
|
|
|
|
/* Find the end of this run */
|
|
|
|
unsigned long run_end;
|
|
|
|
if (unsent_pass) {
|
|
|
|
run_end = find_next_bit(unsentmap, last + 1, run_start + 1);
|
|
|
|
} else {
|
|
|
|
run_end = find_next_zero_bit(bitmap, last + 1, run_start + 1);
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* If the end isn't at the start of a host page, then the
|
|
|
|
* run doesn't finish at the end of a host page
|
|
|
|
* and we need to discard.
|
|
|
|
*/
|
|
|
|
host_offset = run_end % host_ratio;
|
|
|
|
if (host_offset) {
|
|
|
|
do_fixup = true;
|
|
|
|
fixup_start_addr = run_end - host_offset;
|
|
|
|
/*
|
|
|
|
* This host page has gone, the next loop iteration starts
|
|
|
|
* from after the fixup
|
|
|
|
*/
|
|
|
|
run_start = fixup_start_addr + host_ratio;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* No discards on this iteration, next loop starts from
|
|
|
|
* next sent/dirty page
|
|
|
|
*/
|
|
|
|
run_start = run_end + 1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (do_fixup) {
|
|
|
|
unsigned long page;
|
|
|
|
|
|
|
|
/* Tell the destination to discard this page */
|
|
|
|
if (unsent_pass || !test_bit(fixup_start_addr, unsentmap)) {
|
|
|
|
/* For the unsent_pass we:
|
|
|
|
* discard partially sent pages
|
|
|
|
* For the !unsent_pass (dirty) we:
|
|
|
|
* discard partially dirty pages that were sent
|
|
|
|
* (any partially sent pages were already discarded
|
|
|
|
* by the previous unsent_pass)
|
|
|
|
*/
|
|
|
|
postcopy_discard_send_range(ms, pds, fixup_start_addr,
|
|
|
|
host_ratio);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Clean up the bitmap */
|
|
|
|
for (page = fixup_start_addr;
|
|
|
|
page < fixup_start_addr + host_ratio; page++) {
|
|
|
|
/* All pages in this host page are now not sent */
|
|
|
|
set_bit(page, unsentmap);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remark them as dirty, updating the count for any pages
|
|
|
|
* that weren't previously dirty.
|
|
|
|
*/
|
|
|
|
migration_dirty_pages += !test_and_set_bit(page, bitmap);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (unsent_pass) {
|
|
|
|
/* Find the next sent page for the next iteration */
|
|
|
|
run_start = find_next_zero_bit(unsentmap, last + 1,
|
|
|
|
run_start);
|
|
|
|
} else {
|
|
|
|
/* Find the next dirty page for the next iteration */
|
|
|
|
run_start = find_next_bit(bitmap, last + 1, run_start);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* postcopy_chuck_hostpages: discrad any partially sent host page
|
|
|
|
*
|
2015-11-05 19:11:15 +01:00
|
|
|
* Utility for the outgoing postcopy code.
|
|
|
|
*
|
|
|
|
* Discard any partially sent host-page size chunks, mark any partially
|
2017-02-24 19:28:31 +01:00
|
|
|
* dirty host-page size chunks as all dirty. In this case the host-page
|
|
|
|
* is the host-page for the particular RAMBlock, i.e. it might be a huge page
|
2015-11-05 19:11:15 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns zero on success
|
|
|
|
*
|
|
|
|
* @ms: current migration state
|
2015-11-05 19:11:15 +01:00
|
|
|
*/
|
|
|
|
static int postcopy_chunk_hostpages(MigrationState *ms)
|
|
|
|
{
|
2017-03-13 19:26:29 +01:00
|
|
|
RAMState *rs = &ram_state;
|
2015-11-05 19:11:15 +01:00
|
|
|
struct RAMBlock *block;
|
|
|
|
|
|
|
|
/* Easiest way to make sure we don't resume in the middle of a host-page */
|
2017-03-13 19:26:29 +01:00
|
|
|
rs->last_seen_block = NULL;
|
|
|
|
rs->last_sent_block = NULL;
|
|
|
|
rs->last_offset = 0;
|
2015-11-05 19:11:15 +01:00
|
|
|
|
|
|
|
QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
|
|
|
|
unsigned long first = block->offset >> TARGET_PAGE_BITS;
|
|
|
|
|
|
|
|
PostcopyDiscardState *pds =
|
|
|
|
postcopy_discard_send_init(ms, first, block->idstr);
|
|
|
|
|
|
|
|
/* First pass: Discard all partially sent host pages */
|
|
|
|
postcopy_chunk_hostpages_pass(ms, true, block, pds);
|
|
|
|
/*
|
|
|
|
* Second pass: Ensure that all partially dirty host pages are made
|
|
|
|
* fully dirty.
|
|
|
|
*/
|
|
|
|
postcopy_chunk_hostpages_pass(ms, false, block, pds);
|
|
|
|
|
|
|
|
postcopy_discard_send_finish(ms, pds);
|
|
|
|
} /* ram_list loop */
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_postcopy_send_discard_bitmap: transmit the discard bitmap
|
|
|
|
*
|
|
|
|
* Returns zero on success
|
|
|
|
*
|
2015-11-05 19:11:02 +01:00
|
|
|
* Transmit the set of pages to be discarded after precopy to the target
|
|
|
|
* these are pages that:
|
|
|
|
* a) Have been previously transmitted but are now dirty again
|
|
|
|
* b) Pages that have never been transmitted, this ensures that
|
|
|
|
* any pages on the destination that have been mapped by background
|
|
|
|
* tasks get discarded (transparent huge pages is the specific concern)
|
|
|
|
* Hopefully this is pretty sparse
|
2017-03-23 15:06:39 +01:00
|
|
|
*
|
|
|
|
* @ms: current migration state
|
2015-11-05 19:11:02 +01:00
|
|
|
*/
|
|
|
|
int ram_postcopy_send_discard_bitmap(MigrationState *ms)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
unsigned long *bitmap, *unsentmap;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
|
|
|
|
/* This should be our last sync, the src is now paused */
|
2017-03-13 19:35:50 +01:00
|
|
|
migration_bitmap_sync(&ram_state);
|
2015-11-05 19:11:02 +01:00
|
|
|
|
|
|
|
unsentmap = atomic_rcu_read(&migration_bitmap_rcu)->unsentmap;
|
|
|
|
if (!unsentmap) {
|
|
|
|
/* We don't have a safe way to resize the sentmap, so
|
|
|
|
* if the bitmap was resized it will be NULL at this
|
|
|
|
* point.
|
|
|
|
*/
|
|
|
|
error_report("migration ram resized during precopy phase");
|
|
|
|
rcu_read_unlock();
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2017-02-24 19:28:31 +01:00
|
|
|
/* Deal with TPS != HPS and huge pages */
|
2015-11-05 19:11:15 +01:00
|
|
|
ret = postcopy_chunk_hostpages(ms);
|
|
|
|
if (ret) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:02 +01:00
|
|
|
/*
|
|
|
|
* Update the unsentmap to be unsentmap = unsentmap | dirty
|
|
|
|
*/
|
|
|
|
bitmap = atomic_rcu_read(&migration_bitmap_rcu)->bmap;
|
|
|
|
bitmap_or(unsentmap, unsentmap, bitmap,
|
|
|
|
last_ram_offset() >> TARGET_PAGE_BITS);
|
|
|
|
|
|
|
|
|
|
|
|
trace_ram_postcopy_send_discard_bitmap();
|
|
|
|
#ifdef DEBUG_POSTCOPY
|
|
|
|
ram_debug_dump_bitmap(unsentmap, true);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
ret = postcopy_each_ram_send_discard(ms);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_discard_range: discard dirtied pages at the beginning of postcopy
|
2015-11-05 19:11:02 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* Returns zero on success
|
2015-11-05 19:11:02 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* @mis: current migration incoming state
|
2017-03-23 15:11:59 +01:00
|
|
|
* @rbname: name of the RAMBlock of the request. NULL means the
|
|
|
|
* same that last one.
|
2017-03-23 15:06:39 +01:00
|
|
|
* @start: RAMBlock starting page
|
|
|
|
* @length: RAMBlock size
|
2015-11-05 19:11:02 +01:00
|
|
|
*/
|
|
|
|
int ram_discard_range(MigrationIncomingState *mis,
|
2017-03-23 15:11:59 +01:00
|
|
|
const char *rbname,
|
2015-11-05 19:11:02 +01:00
|
|
|
uint64_t start, size_t length)
|
|
|
|
{
|
|
|
|
int ret = -1;
|
|
|
|
|
2017-03-23 15:11:59 +01:00
|
|
|
trace_ram_discard_range(rbname, start, length);
|
2017-02-24 19:28:32 +01:00
|
|
|
|
2015-11-05 19:11:02 +01:00
|
|
|
rcu_read_lock();
|
2017-03-23 15:11:59 +01:00
|
|
|
RAMBlock *rb = qemu_ram_block_by_name(rbname);
|
2015-11-05 19:11:02 +01:00
|
|
|
|
|
|
|
if (!rb) {
|
2017-03-23 15:11:59 +01:00
|
|
|
error_report("ram_discard_range: Failed to find block '%s'", rbname);
|
2015-11-05 19:11:02 +01:00
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
|
2017-02-24 19:28:32 +01:00
|
|
|
ret = ram_block_discard_range(rb, start, length);
|
2015-11-05 19:11:02 +01:00
|
|
|
|
|
|
|
err:
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
static int ram_save_init_globals(RAMState *rs)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
|
|
|
int64_t ram_bitmap_pages; /* Size of bitmap in pages, including gaps */
|
|
|
|
|
2017-03-13 19:35:50 +01:00
|
|
|
rs->dirty_rate_high_cnt = 0;
|
2017-03-13 19:39:02 +01:00
|
|
|
rs->bitmap_sync_count = 0;
|
2017-03-13 20:30:21 +01:00
|
|
|
rs->zero_pages = 0;
|
2017-03-13 20:40:53 +01:00
|
|
|
rs->norm_pages = 0;
|
2017-03-13 20:51:34 +01:00
|
|
|
rs->iterations = 0;
|
2017-03-13 19:44:57 +01:00
|
|
|
migration_bitmap_sync_init(rs);
|
2015-07-02 14:18:06 +02:00
|
|
|
qemu_mutex_init(&migration_bitmap_mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
if (migrate_use_xbzrle()) {
|
|
|
|
XBZRLE_cache_lock();
|
2016-10-24 17:26:49 +02:00
|
|
|
ZERO_TARGET_PAGE = g_malloc0(TARGET_PAGE_SIZE);
|
2015-05-07 19:33:31 +02:00
|
|
|
XBZRLE.cache = cache_init(migrate_xbzrle_cache_size() /
|
|
|
|
TARGET_PAGE_SIZE,
|
|
|
|
TARGET_PAGE_SIZE);
|
|
|
|
if (!XBZRLE.cache) {
|
|
|
|
XBZRLE_cache_unlock();
|
|
|
|
error_report("Error creating cache");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
XBZRLE_cache_unlock();
|
|
|
|
|
|
|
|
/* We prefer not to abort if there is no memory */
|
|
|
|
XBZRLE.encoded_buf = g_try_malloc0(TARGET_PAGE_SIZE);
|
|
|
|
if (!XBZRLE.encoded_buf) {
|
|
|
|
error_report("Error allocating encoded_buf");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
XBZRLE.current_buf = g_try_malloc(TARGET_PAGE_SIZE);
|
|
|
|
if (!XBZRLE.current_buf) {
|
|
|
|
error_report("Error allocating current_buf");
|
|
|
|
g_free(XBZRLE.encoded_buf);
|
|
|
|
XBZRLE.encoded_buf = NULL;
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
acct_clear();
|
|
|
|
}
|
|
|
|
|
2016-02-15 19:57:57 +01:00
|
|
|
/* For memory_global_dirty_log_start below. */
|
|
|
|
qemu_mutex_lock_iothread();
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_mutex_lock_ramlist();
|
|
|
|
rcu_read_lock();
|
|
|
|
bytes_transferred = 0;
|
2017-03-13 19:26:29 +01:00
|
|
|
ram_state_reset(rs);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2015-11-05 19:11:01 +01:00
|
|
|
migration_bitmap_rcu = g_new0(struct BitmapRcu, 1);
|
2017-02-08 15:28:45 +01:00
|
|
|
/* Skip setting bitmap if there is no RAM */
|
|
|
|
if (ram_bytes_total()) {
|
|
|
|
ram_bitmap_pages = last_ram_offset() >> TARGET_PAGE_BITS;
|
|
|
|
migration_bitmap_rcu->bmap = bitmap_new(ram_bitmap_pages);
|
|
|
|
bitmap_set(migration_bitmap_rcu->bmap, 0, ram_bitmap_pages);
|
|
|
|
|
|
|
|
if (migrate_postcopy_ram()) {
|
|
|
|
migration_bitmap_rcu->unsentmap = bitmap_new(ram_bitmap_pages);
|
|
|
|
bitmap_set(migration_bitmap_rcu->unsentmap, 0, ram_bitmap_pages);
|
|
|
|
}
|
2015-11-05 19:11:01 +01:00
|
|
|
}
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
/*
|
|
|
|
* Count the total number of pages used by ram blocks not including any
|
|
|
|
* gaps due to alignment or unplugs.
|
|
|
|
*/
|
|
|
|
migration_dirty_pages = ram_bytes_total() >> TARGET_PAGE_BITS;
|
|
|
|
|
|
|
|
memory_global_dirty_log_start();
|
2017-03-13 19:35:50 +01:00
|
|
|
migration_bitmap_sync(rs);
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_mutex_unlock_ramlist();
|
2016-02-15 19:57:57 +01:00
|
|
|
qemu_mutex_unlock_iothread();
|
2016-10-27 08:42:59 +02:00
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/*
|
|
|
|
* Each of ram_save_setup, ram_save_iterate and ram_save_complete has
|
2016-10-27 08:42:59 +02:00
|
|
|
* long-running RCU critical section. When rcu-reclaims in the code
|
|
|
|
* start to become numerous it will be necessary to reduce the
|
|
|
|
* granularity of these critical sections.
|
|
|
|
*/
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_save_setup: Setup RAM for migration
|
|
|
|
*
|
|
|
|
* Returns zero to indicate success and negative for error
|
|
|
|
*
|
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @opaque: RAMState pointer
|
|
|
|
*/
|
2016-10-27 08:42:59 +02:00
|
|
|
static int ram_save_setup(QEMUFile *f, void *opaque)
|
|
|
|
{
|
2017-03-13 19:26:29 +01:00
|
|
|
RAMState *rs = opaque;
|
2016-10-27 08:42:59 +02:00
|
|
|
RAMBlock *block;
|
|
|
|
|
|
|
|
/* migration has already setup the bitmap, reuse it. */
|
|
|
|
if (!migration_in_colo_state()) {
|
2017-03-13 19:26:29 +01:00
|
|
|
if (ram_save_init_globals(rs) < 0) {
|
2016-10-27 08:42:59 +02:00
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
rcu_read_lock();
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
qemu_put_be64(f, ram_bytes_total() | RAM_SAVE_FLAG_MEM_SIZE);
|
|
|
|
|
|
|
|
QLIST_FOREACH_RCU(block, &ram_list.blocks, next) {
|
|
|
|
qemu_put_byte(f, strlen(block->idstr));
|
|
|
|
qemu_put_buffer(f, (uint8_t *)block->idstr, strlen(block->idstr));
|
|
|
|
qemu_put_be64(f, block->used_length);
|
2017-02-24 19:28:30 +01:00
|
|
|
if (migrate_postcopy_ram() && block->page_size != qemu_host_page_size) {
|
|
|
|
qemu_put_be64(f, block->page_size);
|
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
ram_control_before_iterate(f, RAM_CONTROL_SETUP);
|
|
|
|
ram_control_after_iterate(f, RAM_CONTROL_SETUP);
|
|
|
|
|
|
|
|
qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_save_iterate: iterative stage for migration
|
|
|
|
*
|
|
|
|
* Returns zero to indicate success and negative for error
|
|
|
|
*
|
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @opaque: RAMState pointer
|
|
|
|
*/
|
2015-05-07 19:33:31 +02:00
|
|
|
static int ram_save_iterate(QEMUFile *f, void *opaque)
|
|
|
|
{
|
2017-03-13 19:26:29 +01:00
|
|
|
RAMState *rs = opaque;
|
2015-05-07 19:33:31 +02:00
|
|
|
int ret;
|
|
|
|
int i;
|
|
|
|
int64_t t0;
|
2016-11-04 14:10:17 +01:00
|
|
|
int done = 0;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
rcu_read_lock();
|
2017-03-13 19:26:29 +01:00
|
|
|
if (ram_list.version != rs->last_version) {
|
|
|
|
ram_state_reset(rs);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Read version before ram_list.blocks */
|
|
|
|
smp_rmb();
|
|
|
|
|
|
|
|
ram_control_before_iterate(f, RAM_CONTROL_ROUND);
|
|
|
|
|
|
|
|
t0 = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
|
|
|
|
i = 0;
|
|
|
|
while ((ret = qemu_file_rate_limit(f)) == 0) {
|
|
|
|
int pages;
|
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
pages = ram_find_and_save_block(rs, f, false, &bytes_transferred);
|
2015-05-07 19:33:31 +02:00
|
|
|
/* no more pages to sent */
|
|
|
|
if (pages == 0) {
|
2016-11-04 14:10:17 +01:00
|
|
|
done = 1;
|
2015-05-07 19:33:31 +02:00
|
|
|
break;
|
|
|
|
}
|
2017-03-13 20:51:34 +01:00
|
|
|
rs->iterations++;
|
2015-09-08 19:12:35 +02:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
/* we want to check in the 1st loop, just in case it was the 1st time
|
|
|
|
and we had to sync the dirty bitmap.
|
|
|
|
qemu_get_clock_ns() is a bit expensive, so we only check each some
|
|
|
|
iterations
|
|
|
|
*/
|
|
|
|
if ((i & 63) == 0) {
|
|
|
|
uint64_t t1 = (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - t0) / 1000000;
|
|
|
|
if (t1 > MAX_WAIT) {
|
2017-01-23 22:32:05 +01:00
|
|
|
trace_ram_save_iterate_big_wait(t1, i);
|
2015-05-07 19:33:31 +02:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
flush_compressed_data(f);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Must occur before EOS (or any QEMUFile operation)
|
|
|
|
* because of RDMA protocol.
|
|
|
|
*/
|
|
|
|
ram_control_after_iterate(f, RAM_CONTROL_ROUND);
|
|
|
|
|
|
|
|
qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
|
|
|
|
bytes_transferred += 8;
|
|
|
|
|
|
|
|
ret = qemu_file_get_error(f);
|
|
|
|
if (ret < 0) {
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2016-11-04 14:10:17 +01:00
|
|
|
return done;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_save_complete: function called to send the remaining amount of ram
|
|
|
|
*
|
|
|
|
* Returns zero to indicate success
|
|
|
|
*
|
|
|
|
* Called with iothread lock
|
|
|
|
*
|
|
|
|
* @f: QEMUFile where to send the data
|
|
|
|
* @opaque: RAMState pointer
|
|
|
|
*/
|
2015-05-07 19:33:31 +02:00
|
|
|
static int ram_save_complete(QEMUFile *f, void *opaque)
|
|
|
|
{
|
2017-03-13 19:26:29 +01:00
|
|
|
RAMState *rs = opaque;
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
rcu_read_lock();
|
|
|
|
|
2015-11-05 19:11:13 +01:00
|
|
|
if (!migration_in_postcopy(migrate_get_current())) {
|
2017-03-13 19:35:50 +01:00
|
|
|
migration_bitmap_sync(rs);
|
2015-11-05 19:11:13 +01:00
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
ram_control_before_iterate(f, RAM_CONTROL_FINISH);
|
|
|
|
|
|
|
|
/* try transferring iterative blocks of memory */
|
|
|
|
|
|
|
|
/* flush all remaining blocks regardless of rate limiting */
|
|
|
|
while (true) {
|
|
|
|
int pages;
|
|
|
|
|
2017-03-13 19:26:29 +01:00
|
|
|
pages = ram_find_and_save_block(rs, f, !migration_in_colo_state(),
|
2016-10-27 08:42:59 +02:00
|
|
|
&bytes_transferred);
|
2015-05-07 19:33:31 +02:00
|
|
|
/* no more blocks to sent */
|
|
|
|
if (pages == 0) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
flush_compressed_data(f);
|
|
|
|
ram_control_after_iterate(f, RAM_CONTROL_FINISH);
|
|
|
|
|
|
|
|
rcu_read_unlock();
|
2015-07-09 08:47:58 +02:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2015-11-05 19:10:54 +01:00
|
|
|
static void ram_save_pending(QEMUFile *f, void *opaque, uint64_t max_size,
|
|
|
|
uint64_t *non_postcopiable_pending,
|
|
|
|
uint64_t *postcopiable_pending)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
2017-03-13 19:35:50 +01:00
|
|
|
RAMState *rs = opaque;
|
2015-05-07 19:33:31 +02:00
|
|
|
uint64_t remaining_size;
|
|
|
|
|
|
|
|
remaining_size = ram_save_remaining() * TARGET_PAGE_SIZE;
|
|
|
|
|
2015-11-05 19:11:13 +01:00
|
|
|
if (!migration_in_postcopy(migrate_get_current()) &&
|
|
|
|
remaining_size < max_size) {
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_mutex_lock_iothread();
|
|
|
|
rcu_read_lock();
|
2017-03-13 19:35:50 +01:00
|
|
|
migration_bitmap_sync(rs);
|
2015-05-07 19:33:31 +02:00
|
|
|
rcu_read_unlock();
|
|
|
|
qemu_mutex_unlock_iothread();
|
|
|
|
remaining_size = ram_save_remaining() * TARGET_PAGE_SIZE;
|
|
|
|
}
|
2015-11-05 19:10:54 +01:00
|
|
|
|
|
|
|
/* We can do postcopy, and all the data is postcopiable */
|
|
|
|
*postcopiable_pending += remaining_size;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int load_xbzrle(QEMUFile *f, ram_addr_t addr, void *host)
|
|
|
|
{
|
|
|
|
unsigned int xh_len;
|
|
|
|
int xh_flags;
|
2015-12-16 12:47:37 +01:00
|
|
|
uint8_t *loaded_data;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
if (!xbzrle_decoded_buf) {
|
|
|
|
xbzrle_decoded_buf = g_malloc(TARGET_PAGE_SIZE);
|
|
|
|
}
|
2015-12-16 12:47:37 +01:00
|
|
|
loaded_data = xbzrle_decoded_buf;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
/* extract RLE header */
|
|
|
|
xh_flags = qemu_get_byte(f);
|
|
|
|
xh_len = qemu_get_be16(f);
|
|
|
|
|
|
|
|
if (xh_flags != ENCODING_FLAG_XBZRLE) {
|
|
|
|
error_report("Failed to load XBZRLE page - wrong compression!");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (xh_len > TARGET_PAGE_SIZE) {
|
|
|
|
error_report("Failed to load XBZRLE page - len overflow!");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
/* load data and decode */
|
2015-12-16 12:47:37 +01:00
|
|
|
qemu_get_buffer_in_place(f, &loaded_data, xh_len);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
/* decode RLE */
|
2015-12-16 12:47:37 +01:00
|
|
|
if (xbzrle_decode_buffer(loaded_data, xh_len, host,
|
2015-05-07 19:33:31 +02:00
|
|
|
TARGET_PAGE_SIZE) == -1) {
|
|
|
|
error_report("Failed to load XBZRLE page - decode error!");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_block_from_stream: read a RAMBlock id from the migration stream
|
|
|
|
*
|
|
|
|
* Must be called from within a rcu critical section.
|
|
|
|
*
|
2015-05-07 19:33:31 +02:00
|
|
|
* Returns a pointer from within the RCU-protected ram_list.
|
2015-11-05 19:11:11 +01:00
|
|
|
*
|
2017-03-23 15:06:39 +01:00
|
|
|
* @f: QEMUFile where to read the data from
|
|
|
|
* @flags: Page flags (mostly to see if it's a continuation of previous block)
|
2015-11-05 19:11:11 +01:00
|
|
|
*/
|
2017-03-23 15:06:39 +01:00
|
|
|
static inline RAMBlock *ram_block_from_stream(QEMUFile *f, int flags)
|
2015-05-07 19:33:31 +02:00
|
|
|
{
|
|
|
|
static RAMBlock *block = NULL;
|
|
|
|
char id[256];
|
|
|
|
uint8_t len;
|
|
|
|
|
|
|
|
if (flags & RAM_SAVE_FLAG_CONTINUE) {
|
2016-01-15 04:37:41 +01:00
|
|
|
if (!block) {
|
2015-05-07 19:33:31 +02:00
|
|
|
error_report("Ack, bad migration stream!");
|
|
|
|
return NULL;
|
|
|
|
}
|
2016-01-15 04:37:41 +01:00
|
|
|
return block;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
len = qemu_get_byte(f);
|
|
|
|
qemu_get_buffer(f, (uint8_t *)id, len);
|
|
|
|
id[len] = 0;
|
|
|
|
|
2015-11-05 19:10:33 +01:00
|
|
|
block = qemu_ram_block_by_name(id);
|
2016-01-15 04:37:41 +01:00
|
|
|
if (!block) {
|
|
|
|
error_report("Can't find block %s", id);
|
|
|
|
return NULL;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
2016-01-15 04:37:41 +01:00
|
|
|
return block;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void *host_from_ram_block_offset(RAMBlock *block,
|
|
|
|
ram_addr_t offset)
|
|
|
|
{
|
|
|
|
if (!offset_in_ramblock(block, offset)) {
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
return block->host + offset;
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_handle_compressed: handle the zero page case
|
|
|
|
*
|
2015-05-07 19:33:31 +02:00
|
|
|
* If a page (or a whole RDMA chunk) has been
|
|
|
|
* determined to be zero, then zap it.
|
2017-03-23 15:06:39 +01:00
|
|
|
*
|
|
|
|
* @host: host address for the zero page
|
|
|
|
* @ch: what the page is filled from. We only support zero
|
|
|
|
* @size: size of the zero page
|
2015-05-07 19:33:31 +02:00
|
|
|
*/
|
|
|
|
void ram_handle_compressed(void *host, uint8_t ch, uint64_t size)
|
|
|
|
{
|
|
|
|
if (ch != 0 || !is_zero_range(host, size)) {
|
|
|
|
memset(host, ch, size);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void *do_data_decompress(void *opaque)
|
|
|
|
{
|
|
|
|
DecompressParam *param = opaque;
|
|
|
|
unsigned long pagesize;
|
2016-05-05 09:32:58 +02:00
|
|
|
uint8_t *des;
|
|
|
|
int len;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2016-05-05 09:32:58 +02:00
|
|
|
qemu_mutex_lock(¶m->mutex);
|
2016-05-05 09:32:56 +02:00
|
|
|
while (!param->quit) {
|
2016-05-05 09:32:58 +02:00
|
|
|
if (param->des) {
|
|
|
|
des = param->des;
|
|
|
|
len = param->len;
|
|
|
|
param->des = 0;
|
|
|
|
qemu_mutex_unlock(¶m->mutex);
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
pagesize = TARGET_PAGE_SIZE;
|
2016-05-05 09:32:51 +02:00
|
|
|
/* uncompress() will return failed in some case, especially
|
|
|
|
* when the page is dirted when doing the compression, it's
|
|
|
|
* not a problem because the dirty page will be retransferred
|
|
|
|
* and uncompress() won't break the data in other pages.
|
|
|
|
*/
|
2016-05-05 09:32:58 +02:00
|
|
|
uncompress((Bytef *)des, &pagesize,
|
|
|
|
(const Bytef *)param->compbuf, len);
|
2016-05-05 09:32:51 +02:00
|
|
|
|
2016-05-05 09:32:58 +02:00
|
|
|
qemu_mutex_lock(&decomp_done_lock);
|
|
|
|
param->done = true;
|
|
|
|
qemu_cond_signal(&decomp_done_cond);
|
|
|
|
qemu_mutex_unlock(&decomp_done_lock);
|
|
|
|
|
|
|
|
qemu_mutex_lock(¶m->mutex);
|
|
|
|
} else {
|
|
|
|
qemu_cond_wait(¶m->cond, ¶m->mutex);
|
|
|
|
}
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
2016-05-05 09:32:58 +02:00
|
|
|
qemu_mutex_unlock(¶m->mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2016-05-05 09:32:52 +02:00
|
|
|
static void wait_for_decompress_done(void)
|
|
|
|
{
|
|
|
|
int idx, thread_count;
|
|
|
|
|
|
|
|
if (!migrate_use_compression()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
thread_count = migrate_decompress_threads();
|
|
|
|
qemu_mutex_lock(&decomp_done_lock);
|
|
|
|
for (idx = 0; idx < thread_count; idx++) {
|
|
|
|
while (!decomp_param[idx].done) {
|
|
|
|
qemu_cond_wait(&decomp_done_cond, &decomp_done_lock);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
qemu_mutex_unlock(&decomp_done_lock);
|
|
|
|
}
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
void migrate_decompress_threads_create(void)
|
|
|
|
{
|
|
|
|
int i, thread_count;
|
|
|
|
|
|
|
|
thread_count = migrate_decompress_threads();
|
|
|
|
decompress_threads = g_new0(QemuThread, thread_count);
|
|
|
|
decomp_param = g_new0(DecompressParam, thread_count);
|
2016-05-05 09:32:51 +02:00
|
|
|
qemu_mutex_init(&decomp_done_lock);
|
|
|
|
qemu_cond_init(&decomp_done_cond);
|
2015-05-07 19:33:31 +02:00
|
|
|
for (i = 0; i < thread_count; i++) {
|
|
|
|
qemu_mutex_init(&decomp_param[i].mutex);
|
|
|
|
qemu_cond_init(&decomp_param[i].cond);
|
|
|
|
decomp_param[i].compbuf = g_malloc0(compressBound(TARGET_PAGE_SIZE));
|
2016-05-05 09:32:51 +02:00
|
|
|
decomp_param[i].done = true;
|
2016-05-05 09:32:56 +02:00
|
|
|
decomp_param[i].quit = false;
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_thread_create(decompress_threads + i, "decompress",
|
|
|
|
do_data_decompress, decomp_param + i,
|
|
|
|
QEMU_THREAD_JOINABLE);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void migrate_decompress_threads_join(void)
|
|
|
|
{
|
|
|
|
int i, thread_count;
|
|
|
|
|
|
|
|
thread_count = migrate_decompress_threads();
|
|
|
|
for (i = 0; i < thread_count; i++) {
|
|
|
|
qemu_mutex_lock(&decomp_param[i].mutex);
|
2016-05-05 09:32:56 +02:00
|
|
|
decomp_param[i].quit = true;
|
2015-05-07 19:33:31 +02:00
|
|
|
qemu_cond_signal(&decomp_param[i].cond);
|
|
|
|
qemu_mutex_unlock(&decomp_param[i].mutex);
|
|
|
|
}
|
|
|
|
for (i = 0; i < thread_count; i++) {
|
|
|
|
qemu_thread_join(decompress_threads + i);
|
|
|
|
qemu_mutex_destroy(&decomp_param[i].mutex);
|
|
|
|
qemu_cond_destroy(&decomp_param[i].cond);
|
|
|
|
g_free(decomp_param[i].compbuf);
|
|
|
|
}
|
|
|
|
g_free(decompress_threads);
|
|
|
|
g_free(decomp_param);
|
|
|
|
decompress_threads = NULL;
|
|
|
|
decomp_param = NULL;
|
|
|
|
}
|
|
|
|
|
2015-12-16 12:47:38 +01:00
|
|
|
static void decompress_data_with_multi_threads(QEMUFile *f,
|
2015-05-07 19:33:31 +02:00
|
|
|
void *host, int len)
|
|
|
|
{
|
|
|
|
int idx, thread_count;
|
|
|
|
|
|
|
|
thread_count = migrate_decompress_threads();
|
2016-05-05 09:32:51 +02:00
|
|
|
qemu_mutex_lock(&decomp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
while (true) {
|
|
|
|
for (idx = 0; idx < thread_count; idx++) {
|
2016-05-05 09:32:51 +02:00
|
|
|
if (decomp_param[idx].done) {
|
2016-05-05 09:32:58 +02:00
|
|
|
decomp_param[idx].done = false;
|
|
|
|
qemu_mutex_lock(&decomp_param[idx].mutex);
|
2015-12-16 12:47:38 +01:00
|
|
|
qemu_get_buffer(f, decomp_param[idx].compbuf, len);
|
2015-05-07 19:33:31 +02:00
|
|
|
decomp_param[idx].des = host;
|
|
|
|
decomp_param[idx].len = len;
|
2016-05-05 09:32:58 +02:00
|
|
|
qemu_cond_signal(&decomp_param[idx].cond);
|
|
|
|
qemu_mutex_unlock(&decomp_param[idx].mutex);
|
2015-05-07 19:33:31 +02:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (idx < thread_count) {
|
|
|
|
break;
|
2016-05-05 09:32:51 +02:00
|
|
|
} else {
|
|
|
|
qemu_cond_wait(&decomp_done_cond, &decomp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
}
|
2016-05-05 09:32:51 +02:00
|
|
|
qemu_mutex_unlock(&decomp_done_lock);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_postcopy_incoming_init: allocate postcopy data structures
|
|
|
|
*
|
|
|
|
* Returns 0 for success and negative if there was one error
|
|
|
|
*
|
|
|
|
* @mis: current migration incoming state
|
|
|
|
*
|
|
|
|
* Allocate data structures etc needed by incoming migration with
|
|
|
|
* postcopy-ram. postcopy-ram's similarly names
|
|
|
|
* postcopy_ram_incoming_init does the work.
|
2015-11-05 19:11:03 +01:00
|
|
|
*/
|
|
|
|
int ram_postcopy_incoming_init(MigrationIncomingState *mis)
|
|
|
|
{
|
|
|
|
size_t ram_pages = last_ram_offset() >> TARGET_PAGE_BITS;
|
|
|
|
|
|
|
|
return postcopy_ram_incoming_init(mis, ram_pages);
|
|
|
|
}
|
|
|
|
|
2017-03-23 15:06:39 +01:00
|
|
|
/**
|
|
|
|
* ram_load_postcopy: load a page in postcopy case
|
|
|
|
*
|
|
|
|
* Returns 0 for success or -errno in case of error
|
|
|
|
*
|
2015-11-05 19:11:11 +01:00
|
|
|
* Called in postcopy mode by ram_load().
|
|
|
|
* rcu_read_lock is taken prior to this being called.
|
2017-03-23 15:06:39 +01:00
|
|
|
*
|
|
|
|
* @f: QEMUFile where to send the data
|
2015-11-05 19:11:11 +01:00
|
|
|
*/
|
|
|
|
static int ram_load_postcopy(QEMUFile *f)
|
|
|
|
{
|
|
|
|
int flags = 0, ret = 0;
|
|
|
|
bool place_needed = false;
|
2017-02-24 19:28:37 +01:00
|
|
|
bool matching_page_sizes = false;
|
2015-11-05 19:11:11 +01:00
|
|
|
MigrationIncomingState *mis = migration_incoming_get_current();
|
|
|
|
/* Temporary page that is later 'placed' */
|
|
|
|
void *postcopy_host_page = postcopy_get_tmp_page(mis);
|
2015-11-05 19:11:12 +01:00
|
|
|
void *last_host = NULL;
|
2015-11-11 15:02:28 +01:00
|
|
|
bool all_zero = false;
|
2015-11-05 19:11:11 +01:00
|
|
|
|
|
|
|
while (!ret && !(flags & RAM_SAVE_FLAG_EOS)) {
|
|
|
|
ram_addr_t addr;
|
|
|
|
void *host = NULL;
|
|
|
|
void *page_buffer = NULL;
|
|
|
|
void *place_source = NULL;
|
2017-02-24 19:28:35 +01:00
|
|
|
RAMBlock *block = NULL;
|
2015-11-05 19:11:11 +01:00
|
|
|
uint8_t ch;
|
|
|
|
|
|
|
|
addr = qemu_get_be64(f);
|
|
|
|
flags = addr & ~TARGET_PAGE_MASK;
|
|
|
|
addr &= TARGET_PAGE_MASK;
|
|
|
|
|
|
|
|
trace_ram_load_postcopy_loop((uint64_t)addr, flags);
|
|
|
|
place_needed = false;
|
|
|
|
if (flags & (RAM_SAVE_FLAG_COMPRESS | RAM_SAVE_FLAG_PAGE)) {
|
2017-02-24 19:28:35 +01:00
|
|
|
block = ram_block_from_stream(f, flags);
|
2016-01-15 04:37:41 +01:00
|
|
|
|
|
|
|
host = host_from_ram_block_offset(block, addr);
|
2015-11-05 19:11:11 +01:00
|
|
|
if (!host) {
|
|
|
|
error_report("Illegal RAM offset " RAM_ADDR_FMT, addr);
|
|
|
|
ret = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
2017-02-24 19:28:37 +01:00
|
|
|
matching_page_sizes = block->page_size == TARGET_PAGE_SIZE;
|
2015-11-05 19:11:11 +01:00
|
|
|
/*
|
2017-02-24 19:28:37 +01:00
|
|
|
* Postcopy requires that we place whole host pages atomically;
|
|
|
|
* these may be huge pages for RAMBlocks that are backed by
|
|
|
|
* hugetlbfs.
|
2015-11-05 19:11:11 +01:00
|
|
|
* To make it atomic, the data is read into a temporary page
|
|
|
|
* that's moved into place later.
|
|
|
|
* The migration protocol uses, possibly smaller, target-pages
|
|
|
|
* however the source ensures it always sends all the components
|
|
|
|
* of a host page in order.
|
|
|
|
*/
|
|
|
|
page_buffer = postcopy_host_page +
|
2017-02-24 19:28:37 +01:00
|
|
|
((uintptr_t)host & (block->page_size - 1));
|
2015-11-05 19:11:11 +01:00
|
|
|
/* If all TP are zero then we can optimise the place */
|
2017-02-24 19:28:37 +01:00
|
|
|
if (!((uintptr_t)host & (block->page_size - 1))) {
|
2015-11-05 19:11:11 +01:00
|
|
|
all_zero = true;
|
2015-11-05 19:11:12 +01:00
|
|
|
} else {
|
|
|
|
/* not the 1st TP within the HP */
|
|
|
|
if (host != (last_host + TARGET_PAGE_SIZE)) {
|
2015-12-18 16:35:19 +01:00
|
|
|
error_report("Non-sequential target page %p/%p",
|
2015-11-05 19:11:12 +01:00
|
|
|
host, last_host);
|
|
|
|
ret = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
2015-11-05 19:11:11 +01:00
|
|
|
}
|
|
|
|
|
2015-11-05 19:11:12 +01:00
|
|
|
|
2015-11-05 19:11:11 +01:00
|
|
|
/*
|
|
|
|
* If it's the last part of a host page then we place the host
|
|
|
|
* page
|
|
|
|
*/
|
|
|
|
place_needed = (((uintptr_t)host + TARGET_PAGE_SIZE) &
|
2017-02-24 19:28:37 +01:00
|
|
|
(block->page_size - 1)) == 0;
|
2015-11-05 19:11:11 +01:00
|
|
|
place_source = postcopy_host_page;
|
|
|
|
}
|
2015-11-05 19:11:12 +01:00
|
|
|
last_host = host;
|
2015-11-05 19:11:11 +01:00
|
|
|
|
|
|
|
switch (flags & ~RAM_SAVE_FLAG_CONTINUE) {
|
|
|
|
case RAM_SAVE_FLAG_COMPRESS:
|
|
|
|
ch = qemu_get_byte(f);
|
|
|
|
memset(page_buffer, ch, TARGET_PAGE_SIZE);
|
|
|
|
if (ch) {
|
|
|
|
all_zero = false;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case RAM_SAVE_FLAG_PAGE:
|
|
|
|
all_zero = false;
|
|
|
|
if (!place_needed || !matching_page_sizes) {
|
|
|
|
qemu_get_buffer(f, page_buffer, TARGET_PAGE_SIZE);
|
|
|
|
} else {
|
|
|
|
/* Avoids the qemu_file copy during postcopy, which is
|
|
|
|
* going to do a copy later; can only do it when we
|
|
|
|
* do this read in one go (matching page sizes)
|
|
|
|
*/
|
|
|
|
qemu_get_buffer_in_place(f, (uint8_t **)&place_source,
|
|
|
|
TARGET_PAGE_SIZE);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case RAM_SAVE_FLAG_EOS:
|
|
|
|
/* normal exit */
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
error_report("Unknown combination of migration flags: %#x"
|
|
|
|
" (postcopy mode)", flags);
|
|
|
|
ret = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (place_needed) {
|
|
|
|
/* This gets called at the last target page in the host page */
|
2017-02-24 19:28:35 +01:00
|
|
|
void *place_dest = host + TARGET_PAGE_SIZE - block->page_size;
|
|
|
|
|
2015-11-05 19:11:11 +01:00
|
|
|
if (all_zero) {
|
2017-02-24 19:28:35 +01:00
|
|
|
ret = postcopy_place_page_zero(mis, place_dest,
|
|
|
|
block->page_size);
|
2015-11-05 19:11:11 +01:00
|
|
|
} else {
|
2017-02-24 19:28:35 +01:00
|
|
|
ret = postcopy_place_page(mis, place_dest,
|
|
|
|
place_source, block->page_size);
|
2015-11-05 19:11:11 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
if (!ret) {
|
|
|
|
ret = qemu_file_get_error(f);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
static int ram_load(QEMUFile *f, void *opaque, int version_id)
|
|
|
|
{
|
|
|
|
int flags = 0, ret = 0;
|
|
|
|
static uint64_t seq_iter;
|
|
|
|
int len = 0;
|
2015-11-05 19:11:11 +01:00
|
|
|
/*
|
|
|
|
* If system is running in postcopy mode, page inserts to host memory must
|
|
|
|
* be atomic
|
|
|
|
*/
|
|
|
|
bool postcopy_running = postcopy_state_get() >= POSTCOPY_INCOMING_LISTENING;
|
2017-02-24 19:28:30 +01:00
|
|
|
/* ADVISE is earlier, it shows the source has the postcopy capability on */
|
|
|
|
bool postcopy_advised = postcopy_state_get() >= POSTCOPY_INCOMING_ADVISE;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
|
|
|
seq_iter++;
|
|
|
|
|
|
|
|
if (version_id != 4) {
|
|
|
|
ret = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* This RCU critical section can be very long running.
|
|
|
|
* When RCU reclaims in the code start to become numerous,
|
|
|
|
* it will be necessary to reduce the granularity of this
|
|
|
|
* critical section.
|
|
|
|
*/
|
|
|
|
rcu_read_lock();
|
2015-11-05 19:11:11 +01:00
|
|
|
|
|
|
|
if (postcopy_running) {
|
|
|
|
ret = ram_load_postcopy(f);
|
|
|
|
}
|
|
|
|
|
|
|
|
while (!postcopy_running && !ret && !(flags & RAM_SAVE_FLAG_EOS)) {
|
2015-05-07 19:33:31 +02:00
|
|
|
ram_addr_t addr, total_ram_bytes;
|
2015-11-05 19:10:39 +01:00
|
|
|
void *host = NULL;
|
2015-05-07 19:33:31 +02:00
|
|
|
uint8_t ch;
|
|
|
|
|
|
|
|
addr = qemu_get_be64(f);
|
|
|
|
flags = addr & ~TARGET_PAGE_MASK;
|
|
|
|
addr &= TARGET_PAGE_MASK;
|
|
|
|
|
2015-11-05 19:10:39 +01:00
|
|
|
if (flags & (RAM_SAVE_FLAG_COMPRESS | RAM_SAVE_FLAG_PAGE |
|
|
|
|
RAM_SAVE_FLAG_COMPRESS_PAGE | RAM_SAVE_FLAG_XBZRLE)) {
|
2016-01-15 04:37:41 +01:00
|
|
|
RAMBlock *block = ram_block_from_stream(f, flags);
|
|
|
|
|
|
|
|
host = host_from_ram_block_offset(block, addr);
|
2015-11-05 19:10:39 +01:00
|
|
|
if (!host) {
|
|
|
|
error_report("Illegal RAM offset " RAM_ADDR_FMT, addr);
|
|
|
|
ret = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
switch (flags & ~RAM_SAVE_FLAG_CONTINUE) {
|
|
|
|
case RAM_SAVE_FLAG_MEM_SIZE:
|
|
|
|
/* Synchronize RAM block list */
|
|
|
|
total_ram_bytes = addr;
|
|
|
|
while (!ret && total_ram_bytes) {
|
|
|
|
RAMBlock *block;
|
|
|
|
char id[256];
|
|
|
|
ram_addr_t length;
|
|
|
|
|
|
|
|
len = qemu_get_byte(f);
|
|
|
|
qemu_get_buffer(f, (uint8_t *)id, len);
|
|
|
|
id[len] = 0;
|
|
|
|
length = qemu_get_be64(f);
|
|
|
|
|
2015-11-05 19:10:33 +01:00
|
|
|
block = qemu_ram_block_by_name(id);
|
|
|
|
if (block) {
|
|
|
|
if (length != block->used_length) {
|
|
|
|
Error *local_err = NULL;
|
2015-05-07 19:33:31 +02:00
|
|
|
|
2016-05-10 04:04:59 +02:00
|
|
|
ret = qemu_ram_resize(block, length,
|
2015-11-05 19:10:33 +01:00
|
|
|
&local_err);
|
|
|
|
if (local_err) {
|
|
|
|
error_report_err(local_err);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|
|
|
|
}
|
2017-02-24 19:28:30 +01:00
|
|
|
/* For postcopy we need to check hugepage sizes match */
|
|
|
|
if (postcopy_advised &&
|
|
|
|
block->page_size != qemu_host_page_size) {
|
|
|
|
uint64_t remote_page_size = qemu_get_be64(f);
|
|
|
|
if (remote_page_size != block->page_size) {
|
|
|
|
error_report("Mismatched RAM page size %s "
|
|
|
|
"(local) %zd != %" PRId64,
|
|
|
|
id, block->page_size,
|
|
|
|
remote_page_size);
|
|
|
|
ret = -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
2015-11-05 19:10:33 +01:00
|
|
|
ram_control_load_hook(f, RAM_CONTROL_BLOCK_REG,
|
|
|
|
block->idstr);
|
|
|
|
} else {
|
2015-05-07 19:33:31 +02:00
|
|
|
error_report("Unknown ramblock \"%s\", cannot "
|
|
|
|
"accept migration", id);
|
|
|
|
ret = -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
total_ram_bytes -= length;
|
|
|
|
}
|
|
|
|
break;
|
2015-11-05 19:10:39 +01:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
case RAM_SAVE_FLAG_COMPRESS:
|
|
|
|
ch = qemu_get_byte(f);
|
|
|
|
ram_handle_compressed(host, ch, TARGET_PAGE_SIZE);
|
|
|
|
break;
|
2015-11-05 19:10:39 +01:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
case RAM_SAVE_FLAG_PAGE:
|
|
|
|
qemu_get_buffer(f, host, TARGET_PAGE_SIZE);
|
|
|
|
break;
|
|
|
|
|
2015-11-05 19:10:39 +01:00
|
|
|
case RAM_SAVE_FLAG_COMPRESS_PAGE:
|
2015-05-07 19:33:31 +02:00
|
|
|
len = qemu_get_be32(f);
|
|
|
|
if (len < 0 || len > compressBound(TARGET_PAGE_SIZE)) {
|
|
|
|
error_report("Invalid compressed data length: %d", len);
|
|
|
|
ret = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
2015-12-16 12:47:38 +01:00
|
|
|
decompress_data_with_multi_threads(f, host, len);
|
2015-05-07 19:33:31 +02:00
|
|
|
break;
|
2015-11-05 19:10:39 +01:00
|
|
|
|
2015-05-07 19:33:31 +02:00
|
|
|
case RAM_SAVE_FLAG_XBZRLE:
|
|
|
|
if (load_xbzrle(f, addr, host) < 0) {
|
|
|
|
error_report("Failed to decompress XBZRLE page at "
|
|
|
|
RAM_ADDR_FMT, addr);
|
|
|
|
ret = -EINVAL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case RAM_SAVE_FLAG_EOS:
|
|
|
|
/* normal exit */
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
if (flags & RAM_SAVE_FLAG_HOOK) {
|
2015-06-11 19:17:23 +02:00
|
|
|
ram_control_load_hook(f, RAM_CONTROL_HOOK, NULL);
|
2015-05-07 19:33:31 +02:00
|
|
|
} else {
|
|
|
|
error_report("Unknown combination of migration flags: %#x",
|
|
|
|
flags);
|
|
|
|
ret = -EINVAL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (!ret) {
|
|
|
|
ret = qemu_file_get_error(f);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-05-05 09:32:52 +02:00
|
|
|
wait_for_decompress_done();
|
2015-05-07 19:33:31 +02:00
|
|
|
rcu_read_unlock();
|
2017-01-23 22:32:05 +01:00
|
|
|
trace_ram_load_complete(ret, seq_iter);
|
2015-05-07 19:33:31 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static SaveVMHandlers savevm_ram_handlers = {
|
|
|
|
.save_live_setup = ram_save_setup,
|
|
|
|
.save_live_iterate = ram_save_iterate,
|
2015-11-05 19:11:00 +01:00
|
|
|
.save_live_complete_postcopy = ram_save_complete,
|
2015-11-05 19:10:41 +01:00
|
|
|
.save_live_complete_precopy = ram_save_complete,
|
2015-05-07 19:33:31 +02:00
|
|
|
.save_live_pending = ram_save_pending,
|
|
|
|
.load_state = ram_load,
|
2015-11-02 08:37:03 +01:00
|
|
|
.cleanup = ram_migration_cleanup,
|
2015-05-07 19:33:31 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
void ram_mig_init(void)
|
|
|
|
{
|
|
|
|
qemu_mutex_init(&XBZRLE.lock);
|
2017-03-13 19:26:29 +01:00
|
|
|
register_savevm_live(NULL, "ram", 0, 4, &savevm_ram_handlers, &ram_state);
|
2015-05-07 19:33:31 +02:00
|
|
|
}
|