arch_init.c 38.1 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
    uint64_t xbzrle_bytes;
    uint64_t xbzrle_pages;
    uint64_t xbzrle_cache_miss;
239
    double xbzrle_cache_miss_rate;
Orit Wasserman's avatar
Orit Wasserman committed
240
    uint64_t xbzrle_overflows;
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259
} 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;
}

260 261 262 263 264 265 266 267 268 269
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;
}

270 271 272 273 274 275 276 277 278 279
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
280 281 282 283 284 285 286 287 288 289 290 291 292 293 294
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;
}

295 296 297 298 299
double xbzrle_mig_cache_miss_rate(void)
{
    return acct_info.xbzrle_cache_miss_rate;
}

Orit Wasserman's avatar
Orit Wasserman committed
300 301 302 303 304
uint64_t xbzrle_mig_pages_overflow(void)
{
    return acct_info.xbzrle_overflows;
}

305 306
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
307
{
308 309 310 311
    size_t size;

    qemu_put_be64(f, offset | cont | flag);
    size = 8;
Orit Wasserman's avatar
Orit Wasserman committed
312

313 314 315 316 317 318 319
    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
320 321
}

322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349
/* 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);
}

350 351
#define ENCODING_FLAG_XBZRLE 0x1

352
static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
353
                            ram_addr_t current_addr, RAMBlock *block,
354
                            ram_addr_t offset, int cont, bool last_stage)
355 356 357 358 359
{
    int encoded_len = 0, bytes_sent = -1;
    uint8_t *prev_cached_page;

    if (!cache_is_cached(XBZRLE.cache, current_addr)) {
360
        acct_info.xbzrle_cache_miss++;
361
        if (!last_stage) {
362
            if (cache_insert(XBZRLE.cache, current_addr, *current_data) == -1) {
363
                return -1;
364 365 366 367
            } else {
                /* update *current_data when the page has been
                   inserted into cache */
                *current_data = get_cached_data(XBZRLE.cache, current_addr);
368
            }
369
        }
370 371 372 373 374 375
        return -1;
    }

    prev_cached_page = get_cached_data(XBZRLE.cache, current_addr);

    /* save current buffer into memory */
376
    memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
377 378 379 380 381 382 383 384 385 386

    /* 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
387
        acct_info.xbzrle_overflows++;
388
        /* update data in the cache */
389 390 391 392
        if (!last_stage) {
            memcpy(prev_cached_page, *current_data, TARGET_PAGE_SIZE);
            *current_data = prev_cached_page;
        }
393 394 395 396
        return -1;
    }

    /* we need to update the data in the cache, in order to get the same data */
397 398 399
    if (!last_stage) {
        memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
    }
400 401

    /* Send XBZRLE based compressed page */
402
    bytes_sent = save_block_hdr(f, block, offset, cont, RAM_SAVE_FLAG_XBZRLE);
403 404 405
    qemu_put_byte(f, ENCODING_FLAG_XBZRLE);
    qemu_put_be16(f, encoded_len);
    qemu_put_buffer(f, XBZRLE.encoded_buf, encoded_len);
406
    bytes_sent += encoded_len + 1 + 2;
Orit Wasserman's avatar
Orit Wasserman committed
407 408
    acct_info.xbzrle_pages++;
    acct_info.xbzrle_bytes += bytes_sent;
409 410 411 412

    return bytes_sent;
}

413 414 415
static inline
ram_addr_t migration_bitmap_find_and_reset_dirty(MemoryRegion *mr,
                                                 ram_addr_t start)
416
{
417 418
    unsigned long base = mr->ram_addr >> TARGET_PAGE_BITS;
    unsigned long nr = base + (start >> TARGET_PAGE_BITS);
419 420
    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
421

422 423 424 425 426 427 428
    unsigned long next;

    if (ram_bulk_stage && nr > base) {
        next = nr + 1;
    } else {
        next = find_next_bit(migration_bitmap, size, nr);
    }
429

430 431
    if (next < size) {
        clear_bit(next, migration_bitmap);
Juan Quintela's avatar
Juan Quintela committed
432
        migration_dirty_pages--;
433
    }
434
    return (next - base) << TARGET_PAGE_BITS;
435 436
}

437
static inline bool migration_bitmap_set_dirty(ram_addr_t addr)
438
{
Juan Quintela's avatar
Juan Quintela committed
439
    bool ret;
440
    int nr = addr >> TARGET_PAGE_BITS;
441

Juan Quintela's avatar
Juan Quintela committed
442 443 444 445
    ret = test_and_set_bit(nr, migration_bitmap);

    if (!ret) {
        migration_dirty_pages++;
446
    }
Juan Quintela's avatar
Juan Quintela committed
447
    return ret;
448 449
}

450 451 452
static void migration_bitmap_sync_range(ram_addr_t start, ram_addr_t length)
{
    ram_addr_t addr;
453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480
    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);
            }
481 482 483 484 485
        }
    }
}


486 487
/* Needs iothread lock! */

488 489
static void migration_bitmap_sync(void)
{
Juan Quintela's avatar
Juan Quintela committed
490 491
    RAMBlock *block;
    uint64_t num_dirty_pages_init = migration_dirty_pages;
492 493
    MigrationState *s = migrate_get_current();
    static int64_t start_time;
494
    static int64_t bytes_xfer_prev;
495 496
    static int64_t num_dirty_pages_period;
    int64_t end_time;
497
    int64_t bytes_xfer_now;
498 499
    static uint64_t xbzrle_cache_miss_prev;
    static uint64_t iterations_prev;
500

501 502
    bitmap_sync_count++;

503 504 505
    if (!bytes_xfer_prev) {
        bytes_xfer_prev = ram_bytes_transferred();
    }
506 507

    if (!start_time) {
508
        start_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
509
    }
510 511

    trace_migration_bitmap_sync_start();
512
    address_space_sync_dirty_bitmap(&address_space_memory);
Juan Quintela's avatar
Juan Quintela committed
513

514
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
515
        migration_bitmap_sync_range(block->mr->ram_addr, block->length);
Juan Quintela's avatar
Juan Quintela committed
516 517
    }
    trace_migration_bitmap_sync_end(migration_dirty_pages
518
                                    - num_dirty_pages_init);
519
    num_dirty_pages_period += migration_dirty_pages - num_dirty_pages_init;
520
    end_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
521 522 523

    /* more than 1 second = 1000 millisecons */
    if (end_time > start_time + 1000) {
524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542
        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;
        }
543 544 545 546 547 548 549 550 551 552
        if (migrate_use_xbzrle()) {
            if (iterations_prev != 0) {
                acct_info.xbzrle_cache_miss_rate =
                   (double)(acct_info.xbzrle_cache_miss -
                            xbzrle_cache_miss_prev) /
                   (acct_info.iterations - iterations_prev);
            }
            iterations_prev = acct_info.iterations;
            xbzrle_cache_miss_prev = acct_info.xbzrle_cache_miss;
        }
553 554
        s->dirty_pages_rate = num_dirty_pages_period * 1000
            / (end_time - start_time);
555
        s->dirty_bytes_rate = s->dirty_pages_rate * TARGET_PAGE_SIZE;
556 557
        start_time = end_time;
        num_dirty_pages_period = 0;
558
        s->dirty_sync_count = bitmap_sync_count;
559
    }
560 561
}

562 563 564
/*
 * ram_save_block: Writes a page of memory to the stream f
 *
565 566
 * Returns:  The number of bytes written.
 *           0 means no dirty pages
567 568
 */

569
static int ram_save_block(QEMUFile *f, bool last_stage)
570
{
571
    RAMBlock *block = last_seen_block;
572
    ram_addr_t offset = last_offset;
573
    bool complete_round = false;
574
    int bytes_sent = 0;
Avi Kivity's avatar
Avi Kivity committed
575
    MemoryRegion *mr;
576
    ram_addr_t current_addr;
577

578
    if (!block)
579
        block = QTAILQ_FIRST(&ram_list.blocks);
580

581
    while (true) {
Avi Kivity's avatar
Avi Kivity committed
582
        mr = block->mr;
583 584 585 586 587 588 589 590 591 592 593
        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;
594
                ram_bulk_stage = false;
595 596
            }
        } else {
Michael R. Hines's avatar
Michael R. Hines committed
597
            int ret;
598
            uint8_t *p;
599
            bool send_async = true;
Juan Quintela's avatar
Juan Quintela committed
600
            int cont = (block == last_sent_block) ?
601
                RAM_SAVE_FLAG_CONTINUE : 0;
602

Avi Kivity's avatar
Avi Kivity committed
603
            p = memory_region_get_ram_ptr(mr) + offset;
604

605 606
            /* In doubt sent page as normal */
            bytes_sent = -1;
Michael R. Hines's avatar
Michael R. Hines committed
607 608 609
            ret = ram_control_save_page(f, block->offset,
                               offset, TARGET_PAGE_SIZE, &bytes_sent);

610 611
            XBZRLE_cache_lock();

612
            current_addr = block->offset + offset;
Michael R. Hines's avatar
Michael R. Hines committed
613 614 615 616 617 618 619 620
            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++;
                    }
                }
621
            } else if (is_zero_range(p, TARGET_PAGE_SIZE)) {
622
                acct_info.dup_pages++;
623 624 625 626
                bytes_sent = save_block_hdr(f, block, offset, cont,
                                            RAM_SAVE_FLAG_COMPRESS);
                qemu_put_byte(f, 0);
                bytes_sent++;
627 628 629 630
                /* Must let xbzrle know, otherwise a previous (now 0'd) cached
                 * page would be stale
                 */
                xbzrle_cache_zero_page(current_addr);
631
            } else if (!ram_bulk_stage && migrate_use_xbzrle()) {
632
                bytes_sent = save_xbzrle_page(f, &p, current_addr, block,
633 634
                                              offset, cont, last_stage);
                if (!last_stage) {
635 636 637 638
                    /* Can't send this cached data async, since the cache page
                     * might get updated before it gets to the wire
                     */
                    send_async = false;
639
                }
640 641
            }

642
            /* XBZRLE overflow or normal page */
643
            if (bytes_sent == -1) {
644
                bytes_sent = save_block_hdr(f, block, offset, cont, RAM_SAVE_FLAG_PAGE);
645 646 647 648 649
                if (send_async) {
                    qemu_put_buffer_async(f, p, TARGET_PAGE_SIZE);
                } else {
                    qemu_put_buffer(f, p, TARGET_PAGE_SIZE);
                }
650
                bytes_sent += TARGET_PAGE_SIZE;
651
                acct_info.norm_pages++;
652 653
            }

654
            XBZRLE_cache_unlock();
655
            /* if page is unmodified, continue to the next */
656
            if (bytes_sent > 0) {
Juan Quintela's avatar
Juan Quintela committed
657
                last_sent_block = block;
658 659
                break;
            }
660
        }
661
    }
662
    last_seen_block = block;
663
    last_offset = offset;
664

665
    return bytes_sent;
666 667 668 669
}

static uint64_t bytes_transferred;

670 671 672 673 674 675 676 677 678 679 680 681
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);
    }
}

682 683
static ram_addr_t ram_save_remaining(void)
{
Juan Quintela's avatar
Juan Quintela committed
684
    return migration_dirty_pages;
685 686 687 688 689 690 691 692 693 694 695 696 697 698
}

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)
{
699 700 701
    RAMBlock *block;
    uint64_t total = 0;

702
    QTAILQ_FOREACH(block, &ram_list.blocks, next)
703 704 705
        total += block->length;

    return total;
706 707
}

708 709 710 711 712 713
void free_xbzrle_decoded_buf(void)
{
    g_free(xbzrle_decoded_buf);
    xbzrle_decoded_buf = NULL;
}

Orit Wasserman's avatar
Orit Wasserman committed
714 715
static void migration_end(void)
{
716 717 718 719 720
    if (migration_bitmap) {
        memory_global_dirty_log_stop();
        g_free(migration_bitmap);
        migration_bitmap = NULL;
    }
721

722
    XBZRLE_cache_lock();
723
    if (XBZRLE.cache) {
724 725 726 727 728
        cache_fini(XBZRLE.cache);
        g_free(XBZRLE.cache);
        g_free(XBZRLE.encoded_buf);
        g_free(XBZRLE.current_buf);
        XBZRLE.cache = NULL;
729 730
        XBZRLE.encoded_buf = NULL;
        XBZRLE.current_buf = NULL;
731
    }
732
    XBZRLE_cache_unlock();
Orit Wasserman's avatar
Orit Wasserman committed
733 734
}

735 736 737 738 739
static void ram_migration_cancel(void *opaque)
{
    migration_end();
}

740 741
static void reset_ram_globals(void)
{
742
    last_seen_block = NULL;
Juan Quintela's avatar
Juan Quintela committed
743
    last_sent_block = NULL;
744
    last_offset = 0;
745
    last_version = ram_list.version;
746
    ram_bulk_stage = true;
747 748
}

749 750
#define MAX_WAIT 50 /* ms, half buffered_file limit */

751
static int ram_save_setup(QEMUFile *f, void *opaque)
752
{
753
    RAMBlock *block;
754
    int64_t ram_bitmap_pages; /* Size of bitmap in pages, including gaps */
Juan Quintela's avatar
Juan Quintela committed
755

756 757
    mig_throttle_on = false;
    dirty_rate_high_cnt = 0;
758
    bitmap_sync_count = 0;
759

760
    if (migrate_use_xbzrle()) {
761
        XBZRLE_cache_lock();
762 763 764 765
        XBZRLE.cache = cache_init(migrate_xbzrle_cache_size() /
                                  TARGET_PAGE_SIZE,
                                  TARGET_PAGE_SIZE);
        if (!XBZRLE.cache) {
766 767
            XBZRLE_cache_unlock();
            error_report("Error creating cache");
768 769
            return -1;
        }
770
        XBZRLE_cache_unlock();
771 772 773 774

        /* We prefer not to abort if there is no memory */
        XBZRLE.encoded_buf = g_try_malloc0(TARGET_PAGE_SIZE);
        if (!XBZRLE.encoded_buf) {
775
            error_report("Error allocating encoded_buf");
776 777 778 779 780
            return -1;
        }

        XBZRLE.current_buf = g_try_malloc(TARGET_PAGE_SIZE);
        if (!XBZRLE.current_buf) {
781
            error_report("Error allocating current_buf");
782 783 784 785 786
            g_free(XBZRLE.encoded_buf);
            XBZRLE.encoded_buf = NULL;
            return -1;
        }

787
        acct_clear();
788 789
    }

790 791 792 793 794
    qemu_mutex_lock_iothread();
    qemu_mutex_lock_ramlist();
    bytes_transferred = 0;
    reset_ram_globals();

795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810
    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;
    }

811
    memory_global_dirty_log_start();
Juan Quintela's avatar
Juan Quintela committed
812
    migration_bitmap_sync();
813
    qemu_mutex_unlock_iothread();
814

815
    qemu_put_be64(f, ram_bytes_total() | RAM_SAVE_FLAG_MEM_SIZE);
816

817
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
818 819 820
        qemu_put_byte(f, strlen(block->idstr));
        qemu_put_buffer(f, (uint8_t *)block->idstr, strlen(block->idstr));
        qemu_put_be64(f, block->length);
821 822
    }

823
    qemu_mutex_unlock_ramlist();
Michael R. Hines's avatar
Michael R. Hines committed
824 825 826 827

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

828 829 830 831 832
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);

    return 0;
}

833
static int ram_save_iterate(QEMUFile *f, void *opaque)
834 835 836
{
    int ret;
    int i;
837
    int64_t t0;
838
    int total_sent = 0;
839

840 841
    qemu_mutex_lock_ramlist();

842 843 844 845
    if (ram_list.version != last_version) {
        reset_ram_globals();
    }

Michael R. Hines's avatar
Michael R. Hines committed
846 847
    ram_control_before_iterate(f, RAM_CONTROL_ROUND);

848
    t0 = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
849
    i = 0;
850
    while ((ret = qemu_file_rate_limit(f)) == 0) {
851
        int bytes_sent;
852

853
        bytes_sent = ram_save_block(f, false);
854
        /* no more blocks to sent */
855
        if (bytes_sent == 0) {
856 857
            break;
        }
858
        total_sent += bytes_sent;
859
        acct_info.iterations++;
860
        check_guest_throttling();
861 862 863 864 865 866
        /* 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) {
867
            uint64_t t1 = (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - t0) / 1000000;
868
            if (t1 > MAX_WAIT) {
869
                DPRINTF("big wait: %" PRIu64 " milliseconds, %d iterations\n",
870 871 872 873 874
                        t1, i);
                break;
            }
        }
        i++;
875 876
    }

877 878
    qemu_mutex_unlock_ramlist();

Michael R. Hines's avatar
Michael R. Hines committed
879 880 881 882 883 884
    /*
     * Must occur before EOS (or any QEMUFile operation)
     * because of RDMA protocol.
     */
    ram_control_after_iterate(f, RAM_CONTROL_ROUND);

885 886 887 888 889 890 891 892 893 894
    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);
895 896 897 898
    if (ret < 0) {
        return ret;
    }

899
    return total_sent;
900 901 902 903
}

static int ram_save_complete(QEMUFile *f, void *opaque)
{
904
    qemu_mutex_lock_ramlist();
905
    migration_bitmap_sync();
906

Michael R. Hines's avatar
Michael R. Hines committed
907 908
    ram_control_before_iterate(f, RAM_CONTROL_FINISH);

909
    /* try transferring iterative blocks of memory */
Orit Wasserman's avatar
Orit Wasserman committed
910

911
    /* flush all remaining blocks regardless of rate limiting */
912
    while (true) {
913 914
        int bytes_sent;

915
        bytes_sent = ram_save_block(f, true);
916
        /* no more blocks to sent */
917
        if (bytes_sent == 0) {
918
            break;
919
        }
920
        bytes_transferred += bytes_sent;
921
    }
Michael R. Hines's avatar
Michael R. Hines committed
922 923

    ram_control_after_iterate(f, RAM_CONTROL_FINISH);
924
    migration_end();
925

926
    qemu_mutex_unlock_ramlist();
927 928
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);

929
    return 0;
930 931
}

932 933 934 935 936 937 938
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) {
939
        qemu_mutex_lock_iothread();
940
        migration_bitmap_sync();
941
        qemu_mutex_unlock_iothread();
942 943 944 945 946
        remaining_size = ram_save_remaining() * TARGET_PAGE_SIZE;
    }
    return remaining_size;
}

947 948 949 950 951 952
static int load_xbzrle(QEMUFile *f, ram_addr_t addr, void *host)
{
    int ret, rc = 0;
    unsigned int xh_len;
    int xh_flags;

953 954
    if (!xbzrle_decoded_buf) {
        xbzrle_decoded_buf = g_malloc(TARGET_PAGE_SIZE);
955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970
    }

    /* 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 */
971
    qemu_get_buffer(f, xbzrle_decoded_buf, xh_len);
972 973

    /* decode RLE */
974
    ret = xbzrle_decode_buffer(xbzrle_decoded_buf, xh_len, host,
975 976 977 978 979 980 981 982 983 984 985 986 987
                               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;
}

988 989 990 991 992 993 994 995 996 997 998 999 1000 1001
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;
        }

1002
        return memory_region_get_ram_ptr(block->mr) + offset;
1003 1004 1005 1006 1007 1008
    }

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

1009
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
1010
        if (!strncmp(id, block->idstr, sizeof(id)))
1011
            return memory_region_get_ram_ptr(block->mr) + offset;
1012 1013 1014 1015 1016 1017
    }

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

1018 1019 1020 1021 1022 1023
/*
 * 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)
{
1024
    if (ch != 0 || !is_zero_range(host, size)) {
1025 1026 1027 1028
        memset(host, ch, size);
    }
}

1029
static int ram_load(QEMUFile *f, void *opaque, int version_id)
1030 1031
{
    ram_addr_t addr;
Orit Wasserman's avatar
Orit Wasserman committed
1032
    int flags, ret = 0;
1033
    int error;
Orit Wasserman's avatar
Orit Wasserman committed
1034 1035 1036
    static uint64_t seq_iter;

    seq_iter++;
1037

ChenLiang's avatar
ChenLiang committed
1038
    if (version_id != 4) {
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
        return -EINVAL;
    }

    do {
        addr = qemu_get_be64(f);

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

        if (flags & RAM_SAVE_FLAG_MEM_SIZE) {