arch_init.c 37.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
/*
 * QEMU System Emulator
 *
 * Copyright (c) 2003-2008 Fabrice Bellard
 *
 * 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.
 */
#include <stdint.h>
#include <stdarg.h>
26
#include <stdlib.h>
27
#ifndef _WIN32
28
#include <sys/types.h>
29 30 31
#include <sys/mman.h>
#endif
#include "config.h"
32
#include "monitor/monitor.h"
33
#include "sysemu/sysemu.h"
34 35
#include "qemu/bitops.h"
#include "qemu/bitmap.h"
36
#include "sysemu/arch_init.h"
37
#include "audio/audio.h"
Paolo Bonzini's avatar
Paolo Bonzini committed
38
#include "hw/i386/pc.h"
39
#include "hw/pci/pci.h"
Paolo Bonzini's avatar
Paolo Bonzini committed
40
#include "hw/audio/audio.h"
41
#include "sysemu/kvm.h"
42
#include "migration/migration.h"
Paolo Bonzini's avatar
Paolo Bonzini committed
43
#include "hw/i386/smbios.h"
44
#include "exec/address-spaces.h"
Paolo Bonzini's avatar
Paolo Bonzini committed
45
#include "hw/audio/pcspk.h"
46
#include "migration/page_cache.h"
47
#include "qemu/config-file.h"
48
#include "qemu/error-report.h"
49
#include "qmp-commands.h"
50
#include "trace.h"
51
#include "exec/cpu-all.h"
52
#include "exec/ram_addr.h"
53
#include "hw/acpi/acpi.h"
54
#include "qemu/host-utils.h"
55

Orit Wasserman's avatar
Orit Wasserman committed
56 57 58 59 60 61 62 63
#ifdef DEBUG_ARCH_INIT
#define DPRINTF(fmt, ...) \
    do { fprintf(stdout, "arch_init: " fmt, ## __VA_ARGS__); } while (0)
#else
#define DPRINTF(fmt, ...) \
    do { } while (0)
#endif

64 65 66 67 68 69 70
#ifdef TARGET_SPARC
int graphic_width = 1024;
int graphic_height = 768;
int graphic_depth = 8;
#else
int graphic_width = 800;
int graphic_height = 600;
71
int graphic_depth = 32;
72 73 74 75 76 77 78 79 80 81 82 83 84
#endif


#if defined(TARGET_ALPHA)
#define QEMU_ARCH QEMU_ARCH_ALPHA
#elif defined(TARGET_ARM)
#define QEMU_ARCH QEMU_ARCH_ARM
#elif defined(TARGET_CRIS)
#define QEMU_ARCH QEMU_ARCH_CRIS
#elif defined(TARGET_I386)
#define QEMU_ARCH QEMU_ARCH_I386
#elif defined(TARGET_M68K)
#define QEMU_ARCH QEMU_ARCH_M68K
Michael Walle's avatar
Michael Walle committed
85 86
#elif defined(TARGET_LM32)
#define QEMU_ARCH QEMU_ARCH_LM32
87 88 89 90
#elif defined(TARGET_MICROBLAZE)
#define QEMU_ARCH QEMU_ARCH_MICROBLAZE
#elif defined(TARGET_MIPS)
#define QEMU_ARCH QEMU_ARCH_MIPS
91 92
#elif defined(TARGET_MOXIE)
#define QEMU_ARCH QEMU_ARCH_MOXIE
93 94
#elif defined(TARGET_OPENRISC)
#define QEMU_ARCH QEMU_ARCH_OPENRISC
95 96 97 98 99 100 101 102
#elif defined(TARGET_PPC)
#define QEMU_ARCH QEMU_ARCH_PPC
#elif defined(TARGET_S390X)
#define QEMU_ARCH QEMU_ARCH_S390X
#elif defined(TARGET_SH4)
#define QEMU_ARCH QEMU_ARCH_SH4
#elif defined(TARGET_SPARC)
#define QEMU_ARCH QEMU_ARCH_SPARC
Max Filippov's avatar
Max Filippov committed
103 104
#elif defined(TARGET_XTENSA)
#define QEMU_ARCH QEMU_ARCH_XTENSA
105 106
#elif defined(TARGET_UNICORE32)
#define QEMU_ARCH QEMU_ARCH_UNICORE32
107 108 109
#endif

const uint32_t arch_type = QEMU_ARCH;
110 111 112
static bool mig_throttle_on;
static int dirty_rate_high_cnt;
static void check_guest_throttling(void);
113

114 115
static uint64_t bitmap_sync_count;

116 117 118
/***********************************************************/
/* ram save/restore */

119 120 121 122 123 124
#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
125
#define RAM_SAVE_FLAG_XBZRLE   0x40
Michael R. Hines's avatar
Michael R. Hines committed
126
/* 0x80 is reserved in migration.h start with 0x100 next */
127

128 129
static struct defconfig_file {
    const char *filename;
130 131
    /* Indicates it is an user config file (disabled by -no-user-config) */
    bool userconfig;
132
} default_config_files[] = {
133
    { CONFIG_QEMU_CONFDIR "/qemu.conf",                   true },
134
    { CONFIG_QEMU_CONFDIR "/target-" TARGET_NAME ".conf", true },
135 136 137
    { NULL }, /* end of list */
};

138
static const uint8_t ZERO_TARGET_PAGE[TARGET_PAGE_SIZE];
139

140
int qemu_read_default_config_files(bool userconfig)
141 142
{
    int ret;
143
    struct defconfig_file *f;
144

145
    for (f = default_config_files; f->filename; f++) {
146 147 148
        if (!userconfig && f->userconfig) {
            continue;
        }
149 150 151 152
        ret = qemu_read_config_file(f->filename);
        if (ret < 0 && ret != -ENOENT) {
            return ret;
        }
153
    }
Laszlo Ersek's avatar
Laszlo Ersek committed
154

155 156 157
    return 0;
}

158
static inline bool is_zero_range(uint8_t *p, uint64_t size)
159
{
160
    return buffer_find_nonzero_offset(p, size) == size;
161 162
}

163 164 165 166 167 168 169
/* 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;
170
    /* Cache for XBZRLE, Protected by lock. */
171
    PageCache *cache;
172
    QemuMutex lock;
173 174
} XBZRLE;

175 176
/* buffer used for XBZRLE decoding */
static uint8_t *xbzrle_decoded_buf;
177

178 179 180 181 182 183 184 185 186 187 188 189
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);
}

190 191 192 193 194 195
/*
 * called from qmp_migrate_set_cache_size in main thread, possibly while
 * a migration is in progress.
 * A running migration maybe using the cache and might finish during this
 * call, hence changes to the cache are protected by XBZRLE.lock().
 */
196 197
int64_t xbzrle_cache_resize(int64_t new_size)
{
198 199
    PageCache *new_cache;
    int64_t ret;
200

201 202 203 204
    if (new_size < TARGET_PAGE_SIZE) {
        return -1;
    }

205 206
    XBZRLE_cache_lock();

207
    if (XBZRLE.cache != NULL) {
208
        if (pow2floor(new_size) == migrate_xbzrle_cache_size()) {
209
            goto out_new_size;
210 211 212 213
        }
        new_cache = cache_init(new_size / TARGET_PAGE_SIZE,
                                        TARGET_PAGE_SIZE);
        if (!new_cache) {
214 215 216
            error_report("Error creating cache");
            ret = -1;
            goto out;
217 218
        }

219 220
        cache_fini(XBZRLE.cache);
        XBZRLE.cache = new_cache;
221
    }
222

223 224 225 226 227
out_new_size:
    ret = pow2floor(new_size);
out:
    XBZRLE_cache_unlock();
    return ret;
228 229
}

230 231 232
/* accounting for migration statistics */
typedef struct AccountingInfo {
    uint64_t dup_pages;
233
    uint64_t skipped_pages;
234 235
    uint64_t norm_pages;
    uint64_t iterations;
Orit Wasserman's avatar
Orit Wasserman committed
236 237 238 239
    uint64_t xbzrle_bytes;
    uint64_t xbzrle_pages;
    uint64_t xbzrle_cache_miss;
    uint64_t xbzrle_overflows;
240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258
} AccountingInfo;

static AccountingInfo acct_info;

static void acct_clear(void)
{
    memset(&acct_info, 0, sizeof(acct_info));
}

uint64_t dup_mig_bytes_transferred(void)
{
    return acct_info.dup_pages * TARGET_PAGE_SIZE;
}

uint64_t dup_mig_pages_transferred(void)
{
    return acct_info.dup_pages;
}

259 260 261 262 263 264 265 266 267 268
uint64_t skipped_mig_bytes_transferred(void)
{
    return acct_info.skipped_pages * TARGET_PAGE_SIZE;
}

uint64_t skipped_mig_pages_transferred(void)
{
    return acct_info.skipped_pages;
}

269 270 271 272 273 274 275 276 277 278
uint64_t norm_mig_bytes_transferred(void)
{
    return acct_info.norm_pages * TARGET_PAGE_SIZE;
}

uint64_t norm_mig_pages_transferred(void)
{
    return acct_info.norm_pages;
}

Orit Wasserman's avatar
Orit Wasserman committed
279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298
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;
}

uint64_t xbzrle_mig_pages_overflow(void)
{
    return acct_info.xbzrle_overflows;
}

299 300
static size_t save_block_hdr(QEMUFile *f, RAMBlock *block, ram_addr_t offset,
                             int cont, int flag)
Orit Wasserman's avatar
Orit Wasserman committed
301
{
302 303 304 305
    size_t size;

    qemu_put_be64(f, offset | cont | flag);
    size = 8;
Orit Wasserman's avatar
Orit Wasserman committed
306

307 308 309 310 311 312 313
    if (!cont) {
        qemu_put_byte(f, strlen(block->idstr));
        qemu_put_buffer(f, (uint8_t *)block->idstr,
                        strlen(block->idstr));
        size += 1 + strlen(block->idstr);
    }
    return size;
Orit Wasserman's avatar
Orit Wasserman committed
314 315
}

316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343
/* This is the last block that we have visited serching for dirty pages
 */
static RAMBlock *last_seen_block;
/* This is the last block from where we have sent data */
static RAMBlock *last_sent_block;
static ram_addr_t last_offset;
static unsigned long *migration_bitmap;
static uint64_t migration_dirty_pages;
static uint32_t last_version;
static bool ram_bulk_stage;

/* Update the xbzrle cache to reflect a page that's been sent as all 0.
 * 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
 * when a small write is made into the 0'd page it gets XBZRLE sent
 */
static void xbzrle_cache_zero_page(ram_addr_t current_addr)
{
    if (ram_bulk_stage || !migrate_use_xbzrle()) {
        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);
}

344 345
#define ENCODING_FLAG_XBZRLE 0x1

346
static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
347
                            ram_addr_t current_addr, RAMBlock *block,
348
                            ram_addr_t offset, int cont, bool last_stage)
349 350 351 352 353
{
    int encoded_len = 0, bytes_sent = -1;
    uint8_t *prev_cached_page;

    if (!cache_is_cached(XBZRLE.cache, current_addr)) {
354
        acct_info.xbzrle_cache_miss++;
355
        if (!last_stage) {
356
            if (cache_insert(XBZRLE.cache, current_addr, *current_data) == -1) {
357
                return -1;
358 359 360 361
            } else {
                /* update *current_data when the page has been
                   inserted into cache */
                *current_data = get_cached_data(XBZRLE.cache, current_addr);
362
            }
363
        }
364 365 366 367 368 369
        return -1;
    }

    prev_cached_page = get_cached_data(XBZRLE.cache, current_addr);

    /* save current buffer into memory */
370
    memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
371 372 373 374 375 376 377 378 379 380

    /* 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) {
        DPRINTF("Skipping unmodified page\n");
        return 0;
    } else if (encoded_len == -1) {
        DPRINTF("Overflow\n");
Orit Wasserman's avatar
Orit Wasserman committed
381
        acct_info.xbzrle_overflows++;
382
        /* update data in the cache */
383 384 385 386
        if (!last_stage) {
            memcpy(prev_cached_page, *current_data, TARGET_PAGE_SIZE);
            *current_data = prev_cached_page;
        }
387 388 389 390
        return -1;
    }

    /* we need to update the data in the cache, in order to get the same data */
391 392 393
    if (!last_stage) {
        memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
    }
394 395

    /* Send XBZRLE based compressed page */
396
    bytes_sent = save_block_hdr(f, block, offset, cont, RAM_SAVE_FLAG_XBZRLE);
397 398 399
    qemu_put_byte(f, ENCODING_FLAG_XBZRLE);
    qemu_put_be16(f, encoded_len);
    qemu_put_buffer(f, XBZRLE.encoded_buf, encoded_len);
400
    bytes_sent += encoded_len + 1 + 2;
Orit Wasserman's avatar
Orit Wasserman committed
401 402
    acct_info.xbzrle_pages++;
    acct_info.xbzrle_bytes += bytes_sent;
403 404 405 406

    return bytes_sent;
}

407 408 409
static inline
ram_addr_t migration_bitmap_find_and_reset_dirty(MemoryRegion *mr,
                                                 ram_addr_t start)
410
{
411 412
    unsigned long base = mr->ram_addr >> TARGET_PAGE_BITS;
    unsigned long nr = base + (start >> TARGET_PAGE_BITS);
413 414
    uint64_t mr_size = TARGET_PAGE_ALIGN(memory_region_size(mr));
    unsigned long size = base + (mr_size >> TARGET_PAGE_BITS);
Juan Quintela's avatar
Juan Quintela committed
415

416 417 418 419 420 421 422
    unsigned long next;

    if (ram_bulk_stage && nr > base) {
        next = nr + 1;
    } else {
        next = find_next_bit(migration_bitmap, size, nr);
    }
423

424 425
    if (next < size) {
        clear_bit(next, migration_bitmap);
Juan Quintela's avatar
Juan Quintela committed
426
        migration_dirty_pages--;
427
    }
428
    return (next - base) << TARGET_PAGE_BITS;
429 430
}

431
static inline bool migration_bitmap_set_dirty(ram_addr_t addr)
432
{
Juan Quintela's avatar
Juan Quintela committed
433
    bool ret;
434
    int nr = addr >> TARGET_PAGE_BITS;
435

Juan Quintela's avatar
Juan Quintela committed
436 437 438 439
    ret = test_and_set_bit(nr, migration_bitmap);

    if (!ret) {
        migration_dirty_pages++;
440
    }
Juan Quintela's avatar
Juan Quintela committed
441
    return ret;
442 443
}

444 445 446
static void migration_bitmap_sync_range(ram_addr_t start, ram_addr_t length)
{
    ram_addr_t addr;
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
    unsigned long page = BIT_WORD(start >> TARGET_PAGE_BITS);

    /* start address is aligned at the start of a word? */
    if (((page * BITS_PER_LONG) << TARGET_PAGE_BITS) == start) {
        int k;
        int nr = BITS_TO_LONGS(length >> TARGET_PAGE_BITS);
        unsigned long *src = ram_list.dirty_memory[DIRTY_MEMORY_MIGRATION];

        for (k = page; k < page + nr; k++) {
            if (src[k]) {
                unsigned long new_dirty;
                new_dirty = ~migration_bitmap[k];
                migration_bitmap[k] |= src[k];
                new_dirty &= src[k];
                migration_dirty_pages += ctpopl(new_dirty);
                src[k] = 0;
            }
        }
    } else {
        for (addr = 0; addr < length; addr += TARGET_PAGE_SIZE) {
            if (cpu_physical_memory_get_dirty(start + addr,
                                              TARGET_PAGE_SIZE,
                                              DIRTY_MEMORY_MIGRATION)) {
                cpu_physical_memory_reset_dirty(start + addr,
                                                TARGET_PAGE_SIZE,
                                                DIRTY_MEMORY_MIGRATION);
                migration_bitmap_set_dirty(start + addr);
            }
475 476 477 478 479
        }
    }
}


480 481
/* Needs iothread lock! */

482 483
static void migration_bitmap_sync(void)
{
Juan Quintela's avatar
Juan Quintela committed
484 485
    RAMBlock *block;
    uint64_t num_dirty_pages_init = migration_dirty_pages;
486 487
    MigrationState *s = migrate_get_current();
    static int64_t start_time;
488
    static int64_t bytes_xfer_prev;
489 490
    static int64_t num_dirty_pages_period;
    int64_t end_time;
491 492
    int64_t bytes_xfer_now;

493 494
    bitmap_sync_count++;

495 496 497
    if (!bytes_xfer_prev) {
        bytes_xfer_prev = ram_bytes_transferred();
    }
498 499

    if (!start_time) {
500
        start_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
501
    }
502 503

    trace_migration_bitmap_sync_start();
504
    address_space_sync_dirty_bitmap(&address_space_memory);
Juan Quintela's avatar
Juan Quintela committed
505

506
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
507
        migration_bitmap_sync_range(block->mr->ram_addr, block->length);
Juan Quintela's avatar
Juan Quintela committed
508 509
    }
    trace_migration_bitmap_sync_end(migration_dirty_pages
510
                                    - num_dirty_pages_init);
511
    num_dirty_pages_period += migration_dirty_pages - num_dirty_pages_init;
512
    end_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
513 514 515

    /* more than 1 second = 1000 millisecons */
    if (end_time > start_time + 1000) {
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
        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
               were in this routine. If that happens >N times (for now N==4)
               we turn on the throttle down logic */
            bytes_xfer_now = ram_bytes_transferred();
            if (s->dirty_pages_rate &&
               (num_dirty_pages_period * TARGET_PAGE_SIZE >
                   (bytes_xfer_now - bytes_xfer_prev)/2) &&
               (dirty_rate_high_cnt++ > 4)) {
                    trace_migration_throttle();
                    mig_throttle_on = true;
                    dirty_rate_high_cnt = 0;
             }
             bytes_xfer_prev = bytes_xfer_now;
        } else {
             mig_throttle_on = false;
        }
535 536
        s->dirty_pages_rate = num_dirty_pages_period * 1000
            / (end_time - start_time);
537
        s->dirty_bytes_rate = s->dirty_pages_rate * TARGET_PAGE_SIZE;
538 539
        start_time = end_time;
        num_dirty_pages_period = 0;
540
        s->dirty_sync_count = bitmap_sync_count;
541
    }
542 543
}

544 545 546
/*
 * ram_save_block: Writes a page of memory to the stream f
 *
547 548
 * Returns:  The number of bytes written.
 *           0 means no dirty pages
549 550
 */

551
static int ram_save_block(QEMUFile *f, bool last_stage)
552
{
553
    RAMBlock *block = last_seen_block;
554
    ram_addr_t offset = last_offset;
555
    bool complete_round = false;
556
    int bytes_sent = 0;
Avi Kivity's avatar
Avi Kivity committed
557
    MemoryRegion *mr;
558
    ram_addr_t current_addr;
559

560
    if (!block)
561
        block = QTAILQ_FIRST(&ram_list.blocks);
562

563
    while (true) {
Avi Kivity's avatar
Avi Kivity committed
564
        mr = block->mr;
565 566 567 568 569 570 571 572 573 574 575
        offset = migration_bitmap_find_and_reset_dirty(mr, offset);
        if (complete_round && block == last_seen_block &&
            offset >= last_offset) {
            break;
        }
        if (offset >= block->length) {
            offset = 0;
            block = QTAILQ_NEXT(block, next);
            if (!block) {
                block = QTAILQ_FIRST(&ram_list.blocks);
                complete_round = true;
576
                ram_bulk_stage = false;
577 578
            }
        } else {
Michael R. Hines's avatar
Michael R. Hines committed
579
            int ret;
580
            uint8_t *p;
581
            bool send_async = true;
Juan Quintela's avatar
Juan Quintela committed
582
            int cont = (block == last_sent_block) ?
583
                RAM_SAVE_FLAG_CONTINUE : 0;
584

Avi Kivity's avatar
Avi Kivity committed
585
            p = memory_region_get_ram_ptr(mr) + offset;
586

587 588
            /* In doubt sent page as normal */
            bytes_sent = -1;
Michael R. Hines's avatar
Michael R. Hines committed
589 590 591
            ret = ram_control_save_page(f, block->offset,
                               offset, TARGET_PAGE_SIZE, &bytes_sent);

592 593
            XBZRLE_cache_lock();

594
            current_addr = block->offset + offset;
Michael R. Hines's avatar
Michael R. Hines committed
595 596 597 598 599 600 601 602
            if (ret != RAM_SAVE_CONTROL_NOT_SUPP) {
                if (ret != RAM_SAVE_CONTROL_DELAYED) {
                    if (bytes_sent > 0) {
                        acct_info.norm_pages++;
                    } else if (bytes_sent == 0) {
                        acct_info.dup_pages++;
                    }
                }
603
            } else if (is_zero_range(p, TARGET_PAGE_SIZE)) {
604
                acct_info.dup_pages++;
605 606 607 608
                bytes_sent = save_block_hdr(f, block, offset, cont,
                                            RAM_SAVE_FLAG_COMPRESS);
                qemu_put_byte(f, 0);
                bytes_sent++;
609 610 611 612
                /* Must let xbzrle know, otherwise a previous (now 0'd) cached
                 * page would be stale
                 */
                xbzrle_cache_zero_page(current_addr);
613
            } else if (!ram_bulk_stage && migrate_use_xbzrle()) {
614
                bytes_sent = save_xbzrle_page(f, &p, current_addr, block,
615 616
                                              offset, cont, last_stage);
                if (!last_stage) {
617 618 619 620
                    /* Can't send this cached data async, since the cache page
                     * might get updated before it gets to the wire
                     */
                    send_async = false;
621
                }
622 623
            }

624
            /* XBZRLE overflow or normal page */
625
            if (bytes_sent == -1) {
626
                bytes_sent = save_block_hdr(f, block, offset, cont, RAM_SAVE_FLAG_PAGE);
627 628 629 630 631
                if (send_async) {
                    qemu_put_buffer_async(f, p, TARGET_PAGE_SIZE);
                } else {
                    qemu_put_buffer(f, p, TARGET_PAGE_SIZE);
                }
632
                bytes_sent += TARGET_PAGE_SIZE;
633
                acct_info.norm_pages++;
634 635
            }

636
            XBZRLE_cache_unlock();
637
            /* if page is unmodified, continue to the next */
638
            if (bytes_sent > 0) {
Juan Quintela's avatar
Juan Quintela committed
639
                last_sent_block = block;
640 641
                break;
            }
642
        }
643
    }
644
    last_seen_block = block;
645
    last_offset = offset;
646

647
    return bytes_sent;
648 649 650 651
}

static uint64_t bytes_transferred;

652 653 654 655 656 657 658 659 660 661 662 663
void acct_update_position(QEMUFile *f, size_t size, bool zero)
{
    uint64_t pages = size / TARGET_PAGE_SIZE;
    if (zero) {
        acct_info.dup_pages += pages;
    } else {
        acct_info.norm_pages += pages;
        bytes_transferred += size;
        qemu_update_position(f, size);
    }
}

664 665
static ram_addr_t ram_save_remaining(void)
{
Juan Quintela's avatar
Juan Quintela committed
666
    return migration_dirty_pages;
667 668 669 670 671 672 673 674 675 676 677 678 679 680
}

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)
{
681 682 683
    RAMBlock *block;
    uint64_t total = 0;

684
    QTAILQ_FOREACH(block, &ram_list.blocks, next)
685 686 687
        total += block->length;

    return total;
688 689
}

690 691 692 693 694 695
void free_xbzrle_decoded_buf(void)
{
    g_free(xbzrle_decoded_buf);
    xbzrle_decoded_buf = NULL;
}

Orit Wasserman's avatar
Orit Wasserman committed
696 697
static void migration_end(void)
{
698 699 700 701 702
    if (migration_bitmap) {
        memory_global_dirty_log_stop();
        g_free(migration_bitmap);
        migration_bitmap = NULL;
    }
703

704
    XBZRLE_cache_lock();
705
    if (XBZRLE.cache) {
706 707 708 709 710
        cache_fini(XBZRLE.cache);
        g_free(XBZRLE.cache);
        g_free(XBZRLE.encoded_buf);
        g_free(XBZRLE.current_buf);
        XBZRLE.cache = NULL;
711 712
        XBZRLE.encoded_buf = NULL;
        XBZRLE.current_buf = NULL;
713
    }
714
    XBZRLE_cache_unlock();
Orit Wasserman's avatar
Orit Wasserman committed
715 716
}

717 718 719 720 721
static void ram_migration_cancel(void *opaque)
{
    migration_end();
}

722 723
static void reset_ram_globals(void)
{
724
    last_seen_block = NULL;
Juan Quintela's avatar
Juan Quintela committed
725
    last_sent_block = NULL;
726
    last_offset = 0;
727
    last_version = ram_list.version;
728
    ram_bulk_stage = true;
729 730
}

731 732
#define MAX_WAIT 50 /* ms, half buffered_file limit */

733
static int ram_save_setup(QEMUFile *f, void *opaque)
734
{
735
    RAMBlock *block;
736
    int64_t ram_bitmap_pages; /* Size of bitmap in pages, including gaps */
Juan Quintela's avatar
Juan Quintela committed
737

738 739
    mig_throttle_on = false;
    dirty_rate_high_cnt = 0;
740
    bitmap_sync_count = 0;
741

742
    if (migrate_use_xbzrle()) {
743
        XBZRLE_cache_lock();
744 745 746 747
        XBZRLE.cache = cache_init(migrate_xbzrle_cache_size() /
                                  TARGET_PAGE_SIZE,
                                  TARGET_PAGE_SIZE);
        if (!XBZRLE.cache) {
748 749
            XBZRLE_cache_unlock();
            error_report("Error creating cache");
750 751
            return -1;
        }
752
        XBZRLE_cache_unlock();
753 754 755 756

        /* We prefer not to abort if there is no memory */
        XBZRLE.encoded_buf = g_try_malloc0(TARGET_PAGE_SIZE);
        if (!XBZRLE.encoded_buf) {
757
            error_report("Error allocating encoded_buf");
758 759 760 761 762
            return -1;
        }

        XBZRLE.current_buf = g_try_malloc(TARGET_PAGE_SIZE);
        if (!XBZRLE.current_buf) {
763
            error_report("Error allocating current_buf");
764 765 766 767 768
            g_free(XBZRLE.encoded_buf);
            XBZRLE.encoded_buf = NULL;
            return -1;
        }

769
        acct_clear();
770 771
    }

772 773 774 775 776
    qemu_mutex_lock_iothread();
    qemu_mutex_lock_ramlist();
    bytes_transferred = 0;
    reset_ram_globals();

777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792
    ram_bitmap_pages = last_ram_offset() >> TARGET_PAGE_BITS;
    migration_bitmap = bitmap_new(ram_bitmap_pages);
    bitmap_set(migration_bitmap, 0, ram_bitmap_pages);

    /*
     * Count the total number of pages used by ram blocks not including any
     * gaps due to alignment or unplugs.
     */
    migration_dirty_pages = 0;
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
        uint64_t block_pages;

        block_pages = block->length >> TARGET_PAGE_BITS;
        migration_dirty_pages += block_pages;
    }

793
    memory_global_dirty_log_start();
Juan Quintela's avatar
Juan Quintela committed
794
    migration_bitmap_sync();
795
    qemu_mutex_unlock_iothread();
796

797
    qemu_put_be64(f, ram_bytes_total() | RAM_SAVE_FLAG_MEM_SIZE);
798

799
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
800 801 802
        qemu_put_byte(f, strlen(block->idstr));
        qemu_put_buffer(f, (uint8_t *)block->idstr, strlen(block->idstr));
        qemu_put_be64(f, block->length);
803 804
    }

805
    qemu_mutex_unlock_ramlist();
Michael R. Hines's avatar
Michael R. Hines committed
806 807 808 809

    ram_control_before_iterate(f, RAM_CONTROL_SETUP);
    ram_control_after_iterate(f, RAM_CONTROL_SETUP);

810 811 812 813 814
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);

    return 0;
}

815
static int ram_save_iterate(QEMUFile *f, void *opaque)
816 817 818
{
    int ret;
    int i;
819
    int64_t t0;
820
    int total_sent = 0;
821

822 823
    qemu_mutex_lock_ramlist();

824 825 826 827
    if (ram_list.version != last_version) {
        reset_ram_globals();
    }

Michael R. Hines's avatar
Michael R. Hines committed
828 829
    ram_control_before_iterate(f, RAM_CONTROL_ROUND);

830
    t0 = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
831
    i = 0;
832
    while ((ret = qemu_file_rate_limit(f)) == 0) {
833
        int bytes_sent;
834

835
        bytes_sent = ram_save_block(f, false);
836
        /* no more blocks to sent */
837
        if (bytes_sent == 0) {
838 839
            break;
        }
840
        total_sent += bytes_sent;
841
        acct_info.iterations++;
842
        check_guest_throttling();
843 844 845 846 847 848
        /* 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) {
849
            uint64_t t1 = (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - t0) / 1000000;
850
            if (t1 > MAX_WAIT) {
851
                DPRINTF("big wait: %" PRIu64 " milliseconds, %d iterations\n",
852 853 854 855 856
                        t1, i);
                break;
            }
        }
        i++;
857 858
    }

859 860
    qemu_mutex_unlock_ramlist();

Michael R. Hines's avatar
Michael R. Hines committed
861 862 863 864 865 866
    /*
     * Must occur before EOS (or any QEMUFile operation)
     * because of RDMA protocol.
     */
    ram_control_after_iterate(f, RAM_CONTROL_ROUND);

867 868 869 870 871 872 873 874 875 876
    bytes_transferred += total_sent;

    /*
     * Do not count these 8 bytes into total_sent, so that we can
     * return 0 if no page had been dirtied.
     */
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
    bytes_transferred += 8;

    ret = qemu_file_get_error(f);
877 878 879 880
    if (ret < 0) {
        return ret;
    }

881
    return total_sent;
882 883 884 885
}

static int ram_save_complete(QEMUFile *f, void *opaque)
{
886
    qemu_mutex_lock_ramlist();
887
    migration_bitmap_sync();
888

Michael R. Hines's avatar
Michael R. Hines committed
889 890
    ram_control_before_iterate(f, RAM_CONTROL_FINISH);

891
    /* try transferring iterative blocks of memory */
Orit Wasserman's avatar
Orit Wasserman committed
892

893
    /* flush all remaining blocks regardless of rate limiting */
894
    while (true) {
895 896
        int bytes_sent;

897
        bytes_sent = ram_save_block(f, true);
898
        /* no more blocks to sent */
899
        if (bytes_sent == 0) {
900
            break;
901
        }
902
        bytes_transferred += bytes_sent;
903
    }
Michael R. Hines's avatar
Michael R. Hines committed
904 905

    ram_control_after_iterate(f, RAM_CONTROL_FINISH);
906
    migration_end();
907

908
    qemu_mutex_unlock_ramlist();
909 910
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);

911
    return 0;
912 913
}

914 915 916 917 918 919 920
static uint64_t ram_save_pending(QEMUFile *f, void *opaque, uint64_t max_size)
{
    uint64_t remaining_size;

    remaining_size = ram_save_remaining() * TARGET_PAGE_SIZE;

    if (remaining_size < max_size) {
921
        qemu_mutex_lock_iothread();
922
        migration_bitmap_sync();
923
        qemu_mutex_unlock_iothread();
924 925 926 927 928
        remaining_size = ram_save_remaining() * TARGET_PAGE_SIZE;
    }
    return remaining_size;
}

929 930 931 932 933 934
static int load_xbzrle(QEMUFile *f, ram_addr_t addr, void *host)
{
    int ret, rc = 0;
    unsigned int xh_len;
    int xh_flags;

935 936
    if (!xbzrle_decoded_buf) {
        xbzrle_decoded_buf = g_malloc(TARGET_PAGE_SIZE);
937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
    }

    /* extract RLE header */
    xh_flags = qemu_get_byte(f);
    xh_len = qemu_get_be16(f);

    if (xh_flags != ENCODING_FLAG_XBZRLE) {
        fprintf(stderr, "Failed to load XBZRLE page - wrong compression!\n");
        return -1;
    }

    if (xh_len > TARGET_PAGE_SIZE) {
        fprintf(stderr, "Failed to load XBZRLE page - len overflow!\n");
        return -1;
    }
    /* load data and decode */
953
    qemu_get_buffer(f, xbzrle_decoded_buf, xh_len);
954 955

    /* decode RLE */
956
    ret = xbzrle_decode_buffer(xbzrle_decoded_buf, xh_len, host,
957 958 959 960 961 962 963 964 965 966 967 968 969
                               TARGET_PAGE_SIZE);
    if (ret == -1) {
        fprintf(stderr, "Failed to load XBZRLE page - decode error!\n");
        rc = -1;
    } else  if (ret > TARGET_PAGE_SIZE) {
        fprintf(stderr, "Failed to load XBZRLE page - size %d exceeds %d!\n",
                ret, TARGET_PAGE_SIZE);
        abort();
    }

    return rc;
}

970 971 972 973 974 975 976 977 978 979 980 981 982 983
static inline void *host_from_stream_offset(QEMUFile *f,
                                            ram_addr_t offset,
                                            int flags)
{
    static RAMBlock *block = NULL;
    char id[256];
    uint8_t len;

    if (flags & RAM_SAVE_FLAG_CONTINUE) {
        if (!block) {
            fprintf(stderr, "Ack, bad migration stream!\n");
            return NULL;
        }

984
        return memory_region_get_ram_ptr(block->mr) + offset;
985 986 987 988 989 990
    }

    len = qemu_get_byte(f);
    qemu_get_buffer(f, (uint8_t *)id, len);
    id[len] = 0;

991
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
992
        if (!strncmp(id, block->idstr, sizeof(id)))
993
            return memory_region_get_ram_ptr(block->mr) + offset;
994 995 996 997 998 999
    }

    fprintf(stderr, "Can't find block %s!\n", id);
    return NULL;
}

1000 1001 1002 1003 1004 1005
/*
 * If a page (or a whole RDMA chunk) has been
 * determined to be zero, then zap it.
 */
void ram_handle_compressed(void *host, uint8_t ch, uint64_t size)
{
1006
    if (ch != 0 || !is_zero_range(host, size)) {
1007 1008 1009 1010
        memset(host, ch, size);
    }
}

1011
static int ram_load(QEMUFile *f, void *opaque, int version_id)
1012 1013
{
    ram_addr_t addr;
Orit Wasserman's avatar
Orit Wasserman committed
1014
    int flags, ret = 0;
1015
    int error;
Orit Wasserman's avatar
Orit Wasserman committed
1016 1017 1018
    static uint64_t seq_iter;

    seq_iter++;
1019

ChenLiang's avatar
ChenLiang committed
1020
    if (version_id != 4) {
1021 1022 1023 1024 1025 1026 1027 1028 1029 1030
        return -EINVAL;
    }

    do {
        addr = qemu_get_be64(f);

        flags = addr & ~TARGET_PAGE_MASK;
        addr &= TARGET_PAGE_MASK;

        if (flags & RAM_SAVE_FLAG_MEM_SIZE) {
ChenLiang's avatar
ChenLiang committed
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
            /* Synchronize RAM block list */
            char id[256];
            ram_addr_t length;
            ram_addr_t total_ram_bytes = addr;

            while (total_ram_bytes) {
                RAMBlock *block;
                uint8_t len;

                len = qemu_get_byte(f);
                qemu_get_buffer(f, (uint8_t *)id, len);
                id[len] = 0;
                length = qemu_get_be64(f);

                QTAILQ_FOREACH(block, &ram_list.blocks, next) {
                    if (!strncmp(id, block->idstr, sizeof(id))) {
                        if (block->length != length) {
                            fprintf(stderr,
                                    "Length mismatch: %s: " RAM_ADDR_FMT
                                    " in != " RAM_ADDR_FMT "\n", id, length,
                                    block->length);
                            ret =  -EINVAL;
                            goto done;
1054
                        }
ChenLiang's avatar
ChenLiang committed
1055
                        break;
1056
                    }