arch_init.c 38.2 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
#elif defined(TARGET_TRICORE)
#define QEMU_ARCH QEMU_ARCH_TRICORE
109 110 111
#endif

const uint32_t arch_type = QEMU_ARCH;
112 113 114
static bool mig_throttle_on;
static int dirty_rate_high_cnt;
static void check_guest_throttling(void);
115

116 117
static uint64_t bitmap_sync_count;

118 119 120
/***********************************************************/
/* ram save/restore */

121 122 123 124 125 126
#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
127
#define RAM_SAVE_FLAG_XBZRLE   0x40
Michael R. Hines's avatar
Michael R. Hines committed
128
/* 0x80 is reserved in migration.h start with 0x100 next */
129

130 131
static struct defconfig_file {
    const char *filename;
132 133
    /* Indicates it is an user config file (disabled by -no-user-config) */
    bool userconfig;
134
} default_config_files[] = {
135
    { CONFIG_QEMU_CONFDIR "/qemu.conf",                   true },
136
    { CONFIG_QEMU_CONFDIR "/target-" TARGET_NAME ".conf", true },
137 138 139
    { NULL }, /* end of list */
};

140
static const uint8_t ZERO_TARGET_PAGE[TARGET_PAGE_SIZE];
141

142
int qemu_read_default_config_files(bool userconfig)
143 144
{
    int ret;
145
    struct defconfig_file *f;
146

147
    for (f = default_config_files; f->filename; f++) {
148 149 150
        if (!userconfig && f->userconfig) {
            continue;
        }
151 152 153 154
        ret = qemu_read_config_file(f->filename);
        if (ret < 0 && ret != -ENOENT) {
            return ret;
        }
155
    }
Laszlo Ersek's avatar
Laszlo Ersek committed
156

157 158 159
    return 0;
}

160
static inline bool is_zero_range(uint8_t *p, uint64_t size)
161
{
162
    return buffer_find_nonzero_offset(p, size) == size;
163 164
}

165 166 167 168 169 170 171
/* 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;
172
    /* Cache for XBZRLE, Protected by lock. */
173
    PageCache *cache;
174
    QemuMutex lock;
175 176
} XBZRLE;

177 178
/* buffer used for XBZRLE decoding */
static uint8_t *xbzrle_decoded_buf;
179

180 181 182 183 184 185 186 187 188 189 190 191
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);
}

192 193 194 195 196 197
/*
 * 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().
 */
198 199
int64_t xbzrle_cache_resize(int64_t new_size)
{
200 201
    PageCache *new_cache;
    int64_t ret;
202

203 204 205 206
    if (new_size < TARGET_PAGE_SIZE) {
        return -1;
    }

207 208
    XBZRLE_cache_lock();

209
    if (XBZRLE.cache != NULL) {
210
        if (pow2floor(new_size) == migrate_xbzrle_cache_size()) {
211
            goto out_new_size;
212 213 214 215
        }
        new_cache = cache_init(new_size / TARGET_PAGE_SIZE,
                                        TARGET_PAGE_SIZE);
        if (!new_cache) {
216 217 218
            error_report("Error creating cache");
            ret = -1;
            goto out;
219 220
        }

221 222
        cache_fini(XBZRLE.cache);
        XBZRLE.cache = new_cache;
223
    }
224

225 226 227 228 229
out_new_size:
    ret = pow2floor(new_size);
out:
    XBZRLE_cache_unlock();
    return ret;
230 231
}

232 233 234
/* accounting for migration statistics */
typedef struct AccountingInfo {
    uint64_t dup_pages;
235
    uint64_t skipped_pages;
236 237
    uint64_t norm_pages;
    uint64_t iterations;
Orit Wasserman's avatar
Orit Wasserman committed
238 239 240
    uint64_t xbzrle_bytes;
    uint64_t xbzrle_pages;
    uint64_t xbzrle_cache_miss;
241
    double xbzrle_cache_miss_rate;
Orit Wasserman's avatar
Orit Wasserman committed
242
    uint64_t xbzrle_overflows;
243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
} 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;
}

262 263 264 265 266 267 268 269 270 271
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;
}

272 273 274 275 276 277 278 279 280 281
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
282 283 284 285 286 287 288 289 290 291 292 293 294 295 296
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;
}

297 298 299 300 301
double xbzrle_mig_cache_miss_rate(void)
{
    return acct_info.xbzrle_cache_miss_rate;
}

Orit Wasserman's avatar
Orit Wasserman committed
302 303 304 305 306
uint64_t xbzrle_mig_pages_overflow(void)
{
    return acct_info.xbzrle_overflows;
}

307 308
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
309
{
310 311 312 313
    size_t size;

    qemu_put_be64(f, offset | cont | flag);
    size = 8;
Orit Wasserman's avatar
Orit Wasserman committed
314

315 316 317 318 319 320 321
    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
322 323
}

324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
/* 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);
}

352 353
#define ENCODING_FLAG_XBZRLE 0x1

354
static int save_xbzrle_page(QEMUFile *f, uint8_t **current_data,
355
                            ram_addr_t current_addr, RAMBlock *block,
356
                            ram_addr_t offset, int cont, bool last_stage)
357 358 359 360 361
{
    int encoded_len = 0, bytes_sent = -1;
    uint8_t *prev_cached_page;

    if (!cache_is_cached(XBZRLE.cache, current_addr)) {
362
        acct_info.xbzrle_cache_miss++;
363
        if (!last_stage) {
364
            if (cache_insert(XBZRLE.cache, current_addr, *current_data) == -1) {
365
                return -1;
366 367 368 369
            } else {
                /* update *current_data when the page has been
                   inserted into cache */
                *current_data = get_cached_data(XBZRLE.cache, current_addr);
370
            }
371
        }
372 373 374 375 376 377
        return -1;
    }

    prev_cached_page = get_cached_data(XBZRLE.cache, current_addr);

    /* save current buffer into memory */
378
    memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
379 380 381 382 383 384 385 386 387 388

    /* 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
389
        acct_info.xbzrle_overflows++;
390
        /* update data in the cache */
391 392 393 394
        if (!last_stage) {
            memcpy(prev_cached_page, *current_data, TARGET_PAGE_SIZE);
            *current_data = prev_cached_page;
        }
395 396 397 398
        return -1;
    }

    /* we need to update the data in the cache, in order to get the same data */
399 400 401
    if (!last_stage) {
        memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
    }
402 403

    /* Send XBZRLE based compressed page */
404
    bytes_sent = save_block_hdr(f, block, offset, cont, RAM_SAVE_FLAG_XBZRLE);
405 406 407
    qemu_put_byte(f, ENCODING_FLAG_XBZRLE);
    qemu_put_be16(f, encoded_len);
    qemu_put_buffer(f, XBZRLE.encoded_buf, encoded_len);
408
    bytes_sent += encoded_len + 1 + 2;
Orit Wasserman's avatar
Orit Wasserman committed
409 410
    acct_info.xbzrle_pages++;
    acct_info.xbzrle_bytes += bytes_sent;
411 412 413 414

    return bytes_sent;
}

415 416 417
static inline
ram_addr_t migration_bitmap_find_and_reset_dirty(MemoryRegion *mr,
                                                 ram_addr_t start)
418
{
419 420
    unsigned long base = mr->ram_addr >> TARGET_PAGE_BITS;
    unsigned long nr = base + (start >> TARGET_PAGE_BITS);
421 422
    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
423

424 425 426 427 428 429 430
    unsigned long next;

    if (ram_bulk_stage && nr > base) {
        next = nr + 1;
    } else {
        next = find_next_bit(migration_bitmap, size, nr);
    }
431

432 433
    if (next < size) {
        clear_bit(next, migration_bitmap);
Juan Quintela's avatar
Juan Quintela committed
434
        migration_dirty_pages--;
435
    }
436
    return (next - base) << TARGET_PAGE_BITS;
437 438
}

439
static inline bool migration_bitmap_set_dirty(ram_addr_t addr)
440
{
Juan Quintela's avatar
Juan Quintela committed
441
    bool ret;
442
    int nr = addr >> TARGET_PAGE_BITS;
443

Juan Quintela's avatar
Juan Quintela committed
444 445 446 447
    ret = test_and_set_bit(nr, migration_bitmap);

    if (!ret) {
        migration_dirty_pages++;
448
    }
Juan Quintela's avatar
Juan Quintela committed
449
    return ret;
450 451
}

452 453 454
static void migration_bitmap_sync_range(ram_addr_t start, ram_addr_t length)
{
    ram_addr_t addr;
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482
    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);
            }
483 484 485 486 487
        }
    }
}


488 489
/* Needs iothread lock! */

490 491
static void migration_bitmap_sync(void)
{
Juan Quintela's avatar
Juan Quintela committed
492 493
    RAMBlock *block;
    uint64_t num_dirty_pages_init = migration_dirty_pages;
494 495
    MigrationState *s = migrate_get_current();
    static int64_t start_time;
496
    static int64_t bytes_xfer_prev;
497 498
    static int64_t num_dirty_pages_period;
    int64_t end_time;
499
    int64_t bytes_xfer_now;
500 501
    static uint64_t xbzrle_cache_miss_prev;
    static uint64_t iterations_prev;
502

503 504
    bitmap_sync_count++;

505 506 507
    if (!bytes_xfer_prev) {
        bytes_xfer_prev = ram_bytes_transferred();
    }
508 509

    if (!start_time) {
510
        start_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
511
    }
512 513

    trace_migration_bitmap_sync_start();
514
    address_space_sync_dirty_bitmap(&address_space_memory);
Juan Quintela's avatar
Juan Quintela committed
515

516
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
517
        migration_bitmap_sync_range(block->mr->ram_addr, block->length);
Juan Quintela's avatar
Juan Quintela committed
518 519
    }
    trace_migration_bitmap_sync_end(migration_dirty_pages
520
                                    - num_dirty_pages_init);
521
    num_dirty_pages_period += migration_dirty_pages - num_dirty_pages_init;
522
    end_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
523 524 525

    /* more than 1 second = 1000 millisecons */
    if (end_time > start_time + 1000) {
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544
        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;
        }
545 546 547 548 549 550 551 552 553 554
        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;
        }
555 556
        s->dirty_pages_rate = num_dirty_pages_period * 1000
            / (end_time - start_time);
557
        s->dirty_bytes_rate = s->dirty_pages_rate * TARGET_PAGE_SIZE;
558 559
        start_time = end_time;
        num_dirty_pages_period = 0;
560
        s->dirty_sync_count = bitmap_sync_count;
561
    }
562 563
}

564
/*
Dr. David Alan Gilbert's avatar
Dr. David Alan Gilbert committed
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639
 * ram_save_page: Send the given page to the stream
 *
 * Returns: Number of bytes written.
 */
static int ram_save_page(QEMUFile *f, RAMBlock* block, ram_addr_t offset,
                         bool last_stage)
{
    int bytes_sent;
    int cont;
    ram_addr_t current_addr;
    MemoryRegion *mr = block->mr;
    uint8_t *p;
    int ret;
    bool send_async = true;

    cont = (block == last_sent_block) ? RAM_SAVE_FLAG_CONTINUE : 0;

    p = memory_region_get_ram_ptr(mr) + offset;

    /* In doubt sent page as normal */
    bytes_sent = -1;
    ret = ram_control_save_page(f, block->offset,
                           offset, TARGET_PAGE_SIZE, &bytes_sent);

    XBZRLE_cache_lock();

    current_addr = block->offset + offset;
    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++;
            }
        }
    } else if (is_zero_range(p, TARGET_PAGE_SIZE)) {
        acct_info.dup_pages++;
        bytes_sent = save_block_hdr(f, block, offset, cont,
                                    RAM_SAVE_FLAG_COMPRESS);
        qemu_put_byte(f, 0);
        bytes_sent++;
        /* Must let xbzrle know, otherwise a previous (now 0'd) cached
         * page would be stale
         */
        xbzrle_cache_zero_page(current_addr);
    } else if (!ram_bulk_stage && migrate_use_xbzrle()) {
        bytes_sent = save_xbzrle_page(f, &p, current_addr, block,
                                      offset, cont, last_stage);
        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 (bytes_sent == -1) {
        bytes_sent = save_block_hdr(f, block, offset, cont, RAM_SAVE_FLAG_PAGE);
        if (send_async) {
            qemu_put_buffer_async(f, p, TARGET_PAGE_SIZE);
        } else {
            qemu_put_buffer(f, p, TARGET_PAGE_SIZE);
        }
        bytes_sent += TARGET_PAGE_SIZE;
        acct_info.norm_pages++;
    }

    XBZRLE_cache_unlock();

    return bytes_sent;
}

/*
 * ram_find_and_save_block: Finds a page to send and sends it to f
640
 *
641 642
 * Returns:  The number of bytes written.
 *           0 means no dirty pages
643 644
 */

Dr. David Alan Gilbert's avatar
Dr. David Alan Gilbert committed
645
static int ram_find_and_save_block(QEMUFile *f, bool last_stage)
646
{
647
    RAMBlock *block = last_seen_block;
648
    ram_addr_t offset = last_offset;
649
    bool complete_round = false;
650
    int bytes_sent = 0;
Avi Kivity's avatar
Avi Kivity committed
651
    MemoryRegion *mr;
652

653
    if (!block)
654
        block = QTAILQ_FIRST(&ram_list.blocks);
655

656
    while (true) {
Avi Kivity's avatar
Avi Kivity committed
657
        mr = block->mr;
658 659 660 661 662 663 664 665 666 667 668
        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;
669
                ram_bulk_stage = false;
670 671
            }
        } else {
Dr. David Alan Gilbert's avatar
Dr. David Alan Gilbert committed
672
            bytes_sent = ram_save_page(f, block, offset, last_stage);
673 674

            /* if page is unmodified, continue to the next */
675
            if (bytes_sent > 0) {
Juan Quintela's avatar
Juan Quintela committed
676
                last_sent_block = block;
677 678
                break;
            }
679
        }
680
    }
681
    last_seen_block = block;
682
    last_offset = offset;
683

684
    return bytes_sent;
685 686 687 688
}

static uint64_t bytes_transferred;

689 690 691 692 693 694 695 696 697 698 699 700
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);
    }
}

701 702
static ram_addr_t ram_save_remaining(void)
{
Juan Quintela's avatar
Juan Quintela committed
703
    return migration_dirty_pages;
704 705 706 707 708 709 710 711 712 713 714 715 716 717
}

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)
{
718 719 720
    RAMBlock *block;
    uint64_t total = 0;

721
    QTAILQ_FOREACH(block, &ram_list.blocks, next)
722 723 724
        total += block->length;

    return total;
725 726
}

727 728 729 730 731 732
void free_xbzrle_decoded_buf(void)
{
    g_free(xbzrle_decoded_buf);
    xbzrle_decoded_buf = NULL;
}

Orit Wasserman's avatar
Orit Wasserman committed
733 734
static void migration_end(void)
{
735 736 737 738 739
    if (migration_bitmap) {
        memory_global_dirty_log_stop();
        g_free(migration_bitmap);
        migration_bitmap = NULL;
    }
740

741
    XBZRLE_cache_lock();
742
    if (XBZRLE.cache) {
743 744 745 746
        cache_fini(XBZRLE.cache);
        g_free(XBZRLE.encoded_buf);
        g_free(XBZRLE.current_buf);
        XBZRLE.cache = NULL;
747 748
        XBZRLE.encoded_buf = NULL;
        XBZRLE.current_buf = NULL;
749
    }
750
    XBZRLE_cache_unlock();
Orit Wasserman's avatar
Orit Wasserman committed
751 752
}

753 754 755 756 757
static void ram_migration_cancel(void *opaque)
{
    migration_end();
}

758 759
static void reset_ram_globals(void)
{
760
    last_seen_block = NULL;
Juan Quintela's avatar
Juan Quintela committed
761
    last_sent_block = NULL;
762
    last_offset = 0;
763
    last_version = ram_list.version;
764
    ram_bulk_stage = true;
765 766
}

767 768
#define MAX_WAIT 50 /* ms, half buffered_file limit */

769
static int ram_save_setup(QEMUFile *f, void *opaque)
770
{
771
    RAMBlock *block;
772
    int64_t ram_bitmap_pages; /* Size of bitmap in pages, including gaps */
Juan Quintela's avatar
Juan Quintela committed
773

774 775
    mig_throttle_on = false;
    dirty_rate_high_cnt = 0;
776
    bitmap_sync_count = 0;
777

778
    if (migrate_use_xbzrle()) {
779
        XBZRLE_cache_lock();
780 781 782 783
        XBZRLE.cache = cache_init(migrate_xbzrle_cache_size() /
                                  TARGET_PAGE_SIZE,
                                  TARGET_PAGE_SIZE);
        if (!XBZRLE.cache) {
784 785
            XBZRLE_cache_unlock();
            error_report("Error creating cache");
786 787
            return -1;
        }
788
        XBZRLE_cache_unlock();
789 790 791 792

        /* We prefer not to abort if there is no memory */
        XBZRLE.encoded_buf = g_try_malloc0(TARGET_PAGE_SIZE);
        if (!XBZRLE.encoded_buf) {
793
            error_report("Error allocating encoded_buf");
794 795 796 797 798
            return -1;
        }

        XBZRLE.current_buf = g_try_malloc(TARGET_PAGE_SIZE);
        if (!XBZRLE.current_buf) {
799
            error_report("Error allocating current_buf");
800 801 802 803 804
            g_free(XBZRLE.encoded_buf);
            XBZRLE.encoded_buf = NULL;
            return -1;
        }

805
        acct_clear();
806 807
    }

808 809 810 811 812
    qemu_mutex_lock_iothread();
    qemu_mutex_lock_ramlist();
    bytes_transferred = 0;
    reset_ram_globals();

813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828
    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;
    }

829
    memory_global_dirty_log_start();
Juan Quintela's avatar
Juan Quintela committed
830
    migration_bitmap_sync();
831
    qemu_mutex_unlock_iothread();
832

833
    qemu_put_be64(f, ram_bytes_total() | RAM_SAVE_FLAG_MEM_SIZE);
834

835
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
836 837 838
        qemu_put_byte(f, strlen(block->idstr));
        qemu_put_buffer(f, (uint8_t *)block->idstr, strlen(block->idstr));
        qemu_put_be64(f, block->length);
839 840
    }

841
    qemu_mutex_unlock_ramlist();
Michael R. Hines's avatar
Michael R. Hines committed
842 843 844 845

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

846 847 848 849 850
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);

    return 0;
}

851
static int ram_save_iterate(QEMUFile *f, void *opaque)
852 853 854
{
    int ret;
    int i;
855
    int64_t t0;
856
    int total_sent = 0;
857

858 859
    qemu_mutex_lock_ramlist();

860 861 862 863
    if (ram_list.version != last_version) {
        reset_ram_globals();
    }

Michael R. Hines's avatar
Michael R. Hines committed
864 865
    ram_control_before_iterate(f, RAM_CONTROL_ROUND);

866
    t0 = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
867
    i = 0;
868
    while ((ret = qemu_file_rate_limit(f)) == 0) {
869
        int bytes_sent;
870

Dr. David Alan Gilbert's avatar
Dr. David Alan Gilbert committed
871
        bytes_sent = ram_find_and_save_block(f, false);
872
        /* no more blocks to sent */
873
        if (bytes_sent == 0) {
874 875
            break;
        }
876
        total_sent += bytes_sent;
877
        acct_info.iterations++;
878
        check_guest_throttling();
879 880 881 882 883 884
        /* 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) {
885
            uint64_t t1 = (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - t0) / 1000000;
886
            if (t1 > MAX_WAIT) {
887
                DPRINTF("big wait: %" PRIu64 " milliseconds, %d iterations\n",
888 889 890 891 892
                        t1, i);
                break;
            }
        }
        i++;
893 894
    }

895 896
    qemu_mutex_unlock_ramlist();

Michael R. Hines's avatar
Michael R. Hines committed
897 898 899 900 901 902
    /*
     * Must occur before EOS (or any QEMUFile operation)
     * because of RDMA protocol.
     */
    ram_control_after_iterate(f, RAM_CONTROL_ROUND);

903 904 905 906 907 908 909 910 911 912
    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);
913 914 915 916
    if (ret < 0) {
        return ret;
    }

917
    return total_sent;
918 919 920 921
}

static int ram_save_complete(QEMUFile *f, void *opaque)
{
922
    qemu_mutex_lock_ramlist();
923
    migration_bitmap_sync();
924

Michael R. Hines's avatar
Michael R. Hines committed
925 926
    ram_control_before_iterate(f, RAM_CONTROL_FINISH);

927
    /* try transferring iterative blocks of memory */
Orit Wasserman's avatar
Orit Wasserman committed
928

929
    /* flush all remaining blocks regardless of rate limiting */
930
    while (true) {
931 932
        int bytes_sent;

Dr. David Alan Gilbert's avatar
Dr. David Alan Gilbert committed
933
        bytes_sent = ram_find_and_save_block(f, true);
934
        /* no more blocks to sent */
935
        if (bytes_sent == 0) {
936
            break;
937
        }
938
        bytes_transferred += bytes_sent;
939
    }
Michael R. Hines's avatar
Michael R. Hines committed
940 941

    ram_control_after_iterate(f, RAM_CONTROL_FINISH);
942
    migration_end();
943

944
    qemu_mutex_unlock_ramlist();
945 946
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);

947
    return 0;
948 949
}

950 951 952 953 954 955 956
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) {
957
        qemu_mutex_lock_iothread();
958
        migration_bitmap_sync();
959
        qemu_mutex_unlock_iothread();
960 961 962 963 964
        remaining_size = ram_save_remaining() * TARGET_PAGE_SIZE;
    }
    return remaining_size;
}

965 966 967 968 969
static int load_xbzrle(QEMUFile *f, ram_addr_t addr, void *host)
{
    unsigned int xh_len;
    int xh_flags;

970 971
    if (!xbzrle_decoded_buf) {
        xbzrle_decoded_buf = g_malloc(TARGET_PAGE_SIZE);
972 973 974 975 976 977 978
    }

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

    if (xh_flags != ENCODING_FLAG_XBZRLE) {
979
        error_report("Failed to load XBZRLE page - wrong compression!");
980 981 982 983
        return -1;
    }

    if (xh_len > TARGET_PAGE_SIZE) {
984
        error_report("Failed to load XBZRLE page - len overflow!");
985 986 987
        return -1;
    }
    /* load data and decode */
988
    qemu_get_buffer(f, xbzrle_decoded_buf, xh_len);
989 990

    /* decode RLE */
991 992
    if (xbzrle_decode_buffer(xbzrle_decoded_buf, xh_len, host,
                             TARGET_PAGE_SIZE) == -1) {
993
        error_report("Failed to load XBZRLE page - decode error!");
994
        return -1;
995 996
    }

997
    return 0;
998 999
}

1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
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) {
1010
            error_report("Ack, bad migration stream!");
1011 1012 1013
            return NULL;
        }

1014
        return memory_region_get_ram_ptr(block->mr) + offset;
1015 1016 1017 1018 1019 1020
    }

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

1021
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
1022
        if (!strncmp(id, block->idstr, sizeof(id)))
1023
            return memory_region_get_ram_ptr(block->mr) + offset;
1024 1025
    }

1026
    error_report("Can't find block %s!", id);
1027 1028 1029
    return NULL;
}

1030 1031 1032 1033 1034 1035
/*
 * 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)
{
1036
    if (ch != 0 || !is_zero_range(host, size)) {
1037 1038 1039 1040
        memset(host, ch, size);
    }
}

1041
static int ram_load(QEMUFile *f, void *opaque, int version_id)
1042 1043
{
    ram_addr_t addr;
Orit Wasserman's avatar
Orit Wasserman committed
1044 1045 1046 1047
    int flags, ret = 0;
    static uint64_t seq_iter;

    seq_iter++;
1048

ChenLiang's avatar
ChenLiang committed
1049
    if (version_id != 4) {
1050
        ret = -EINVAL;
1051 1052
    }

1053
    while (!ret) {
1054 1055 1056 1057 1058 1059
        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
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076
            /* 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) {
Alex Bligh's avatar
Alex Bligh committed
1077 1078
                            error_report("Length mismatch: %s: 0x" RAM_ADDR_FMT
                                         " in != 0x" RAM_ADDR_FMT, id, length,
1079
                                         block->length);
ChenLiang's avatar
ChenLiang committed
1080
                            ret =  -EINVAL;
1081
                        }
ChenLiang's avatar
ChenLiang committed
1082
                        break;