arch_init.c 37.6 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 "qmp-commands.h"
49
#include "trace.h"
50
#include "exec/cpu-all.h"
51
#include "exec/ram_addr.h"
52
#include "hw/acpi/acpi.h"
53
#include "qemu/host-utils.h"
54

Orit Wasserman's avatar
Orit Wasserman committed
55 56 57 58 59 60 61 62
#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

63 64 65 66 67 68 69
#ifdef TARGET_SPARC
int graphic_width = 1024;
int graphic_height = 768;
int graphic_depth = 8;
#else
int graphic_width = 800;
int graphic_height = 600;
70
int graphic_depth = 32;
71 72 73 74 75 76 77 78 79 80 81 82 83
#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
84 85
#elif defined(TARGET_LM32)
#define QEMU_ARCH QEMU_ARCH_LM32
86 87 88 89
#elif defined(TARGET_MICROBLAZE)
#define QEMU_ARCH QEMU_ARCH_MICROBLAZE
#elif defined(TARGET_MIPS)
#define QEMU_ARCH QEMU_ARCH_MIPS
90 91
#elif defined(TARGET_MOXIE)
#define QEMU_ARCH QEMU_ARCH_MOXIE
92 93
#elif defined(TARGET_OPENRISC)
#define QEMU_ARCH QEMU_ARCH_OPENRISC
94 95 96 97 98 99 100 101
#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
102 103
#elif defined(TARGET_XTENSA)
#define QEMU_ARCH QEMU_ARCH_XTENSA
104 105
#elif defined(TARGET_UNICORE32)
#define QEMU_ARCH QEMU_ARCH_UNICORE32
106 107 108
#endif

const uint32_t arch_type = QEMU_ARCH;
109 110 111
static bool mig_throttle_on;
static int dirty_rate_high_cnt;
static void check_guest_throttling(void);
112 113 114 115

/***********************************************************/
/* ram save/restore */

116 117 118 119 120 121
#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
122
#define RAM_SAVE_FLAG_XBZRLE   0x40
Michael R. Hines's avatar
Michael R. Hines committed
123
/* 0x80 is reserved in migration.h start with 0x100 next */
124

125 126
static struct defconfig_file {
    const char *filename;
127 128
    /* Indicates it is an user config file (disabled by -no-user-config) */
    bool userconfig;
129
} default_config_files[] = {
130
    { CONFIG_QEMU_CONFDIR "/qemu.conf",                   true },
131
    { CONFIG_QEMU_CONFDIR "/target-" TARGET_NAME ".conf", true },
132 133 134
    { NULL }, /* end of list */
};

135
static const uint8_t ZERO_TARGET_PAGE[TARGET_PAGE_SIZE];
136

137
int qemu_read_default_config_files(bool userconfig)
138 139
{
    int ret;
140
    struct defconfig_file *f;
141

142
    for (f = default_config_files; f->filename; f++) {
143 144 145
        if (!userconfig && f->userconfig) {
            continue;
        }
146 147 148 149
        ret = qemu_read_config_file(f->filename);
        if (ret < 0 && ret != -ENOENT) {
            return ret;
        }
150
    }
Laszlo Ersek's avatar
Laszlo Ersek committed
151

152 153 154
    return 0;
}

155
static inline bool is_zero_range(uint8_t *p, uint64_t size)
156
{
157
    return buffer_find_nonzero_offset(p, size) == size;
158 159
}

160 161 162 163 164 165 166
/* 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;
167
    /* Cache for XBZRLE, Protected by lock. */
168
    PageCache *cache;
169
    QemuMutex lock;
170 171 172 173 174
} XBZRLE = {
    .encoded_buf = NULL,
    .current_buf = NULL,
    .cache = NULL,
};
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
int64_t xbzrle_cache_resize(int64_t new_size)
{
192 193
    PageCache *new_cache, *cache_to_free;

194 195 196 197
    if (new_size < TARGET_PAGE_SIZE) {
        return -1;
    }

198
    /* no need to lock, the current thread holds qemu big lock */
199
    if (XBZRLE.cache != NULL) {
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
        /* check XBZRLE.cache again later */
        if (pow2floor(new_size) == migrate_xbzrle_cache_size()) {
            return pow2floor(new_size);
        }
        new_cache = cache_init(new_size / TARGET_PAGE_SIZE,
                                        TARGET_PAGE_SIZE);
        if (!new_cache) {
            DPRINTF("Error creating cache\n");
            return -1;
        }

        XBZRLE_cache_lock();
        /* the XBZRLE.cache may have be destroyed, check it again */
        if (XBZRLE.cache != NULL) {
            cache_to_free = XBZRLE.cache;
            XBZRLE.cache = new_cache;
        } else {
            cache_to_free = new_cache;
        }
        XBZRLE_cache_unlock();

        cache_fini(cache_to_free);
222
    }
223

224 225 226
    return pow2floor(new_size);
}

227 228 229
/* accounting for migration statistics */
typedef struct AccountingInfo {
    uint64_t dup_pages;
230
    uint64_t skipped_pages;
231 232
    uint64_t norm_pages;
    uint64_t iterations;
Orit Wasserman's avatar
Orit Wasserman committed
233 234 235 236
    uint64_t xbzrle_bytes;
    uint64_t xbzrle_pages;
    uint64_t xbzrle_cache_miss;
    uint64_t xbzrle_overflows;
237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
} 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;
}

256 257 258 259 260 261 262 263 264 265
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;
}

266 267 268 269 270 271 272 273 274 275
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
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295
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;
}

296 297
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
298
{
299 300 301 302
    size_t size;

    qemu_put_be64(f, offset | cont | flag);
    size = 8;
Orit Wasserman's avatar
Orit Wasserman committed
303

304 305 306 307 308 309 310
    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
311 312
}

313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340
/* 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);
}

341 342 343 344
#define ENCODING_FLAG_XBZRLE 0x1

static int save_xbzrle_page(QEMUFile *f, uint8_t *current_data,
                            ram_addr_t current_addr, RAMBlock *block,
345
                            ram_addr_t offset, int cont, bool last_stage)
346 347 348 349 350
{
    int encoded_len = 0, bytes_sent = -1;
    uint8_t *prev_cached_page;

    if (!cache_is_cached(XBZRLE.cache, current_addr)) {
351
        if (!last_stage) {
352 353 354
            if (cache_insert(XBZRLE.cache, current_addr, current_data) == -1) {
                return -1;
            }
355
        }
Orit Wasserman's avatar
Orit Wasserman committed
356
        acct_info.xbzrle_cache_miss++;
357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373
        return -1;
    }

    prev_cached_page = get_cached_data(XBZRLE.cache, current_addr);

    /* save current buffer into memory */
    memcpy(XBZRLE.current_buf, current_data, TARGET_PAGE_SIZE);

    /* XBZRLE encoding (if there is no overflow) */
    encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
                                       TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
                                       TARGET_PAGE_SIZE);
    if (encoded_len == 0) {
        DPRINTF("Skipping unmodified page\n");
        return 0;
    } else if (encoded_len == -1) {
        DPRINTF("Overflow\n");
Orit Wasserman's avatar
Orit Wasserman committed
374
        acct_info.xbzrle_overflows++;
375 376 377 378 379 380
        /* update data in the cache */
        memcpy(prev_cached_page, current_data, TARGET_PAGE_SIZE);
        return -1;
    }

    /* we need to update the data in the cache, in order to get the same data */
381 382 383
    if (!last_stage) {
        memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
    }
384 385

    /* Send XBZRLE based compressed page */
386
    bytes_sent = save_block_hdr(f, block, offset, cont, RAM_SAVE_FLAG_XBZRLE);
387 388 389
    qemu_put_byte(f, ENCODING_FLAG_XBZRLE);
    qemu_put_be16(f, encoded_len);
    qemu_put_buffer(f, XBZRLE.encoded_buf, encoded_len);
390
    bytes_sent += encoded_len + 1 + 2;
Orit Wasserman's avatar
Orit Wasserman committed
391 392
    acct_info.xbzrle_pages++;
    acct_info.xbzrle_bytes += bytes_sent;
393 394 395 396

    return bytes_sent;
}

397 398 399
static inline
ram_addr_t migration_bitmap_find_and_reset_dirty(MemoryRegion *mr,
                                                 ram_addr_t start)
400
{
401 402
    unsigned long base = mr->ram_addr >> TARGET_PAGE_BITS;
    unsigned long nr = base + (start >> TARGET_PAGE_BITS);
403 404
    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
405

406 407 408 409 410 411 412
    unsigned long next;

    if (ram_bulk_stage && nr > base) {
        next = nr + 1;
    } else {
        next = find_next_bit(migration_bitmap, size, nr);
    }
413

414 415
    if (next < size) {
        clear_bit(next, migration_bitmap);
Juan Quintela's avatar
Juan Quintela committed
416
        migration_dirty_pages--;
417
    }
418
    return (next - base) << TARGET_PAGE_BITS;
419 420
}

421
static inline bool migration_bitmap_set_dirty(ram_addr_t addr)
422
{
Juan Quintela's avatar
Juan Quintela committed
423
    bool ret;
424
    int nr = addr >> TARGET_PAGE_BITS;
425

Juan Quintela's avatar
Juan Quintela committed
426 427 428 429
    ret = test_and_set_bit(nr, migration_bitmap);

    if (!ret) {
        migration_dirty_pages++;
430
    }
Juan Quintela's avatar
Juan Quintela committed
431
    return ret;
432 433
}

434 435 436
static void migration_bitmap_sync_range(ram_addr_t start, ram_addr_t length)
{
    ram_addr_t addr;
437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464
    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);
            }
465 466 467 468 469
        }
    }
}


470 471
/* Needs iothread lock! */

472 473
static void migration_bitmap_sync(void)
{
Juan Quintela's avatar
Juan Quintela committed
474 475
    RAMBlock *block;
    uint64_t num_dirty_pages_init = migration_dirty_pages;
476 477
    MigrationState *s = migrate_get_current();
    static int64_t start_time;
478
    static int64_t bytes_xfer_prev;
479 480
    static int64_t num_dirty_pages_period;
    int64_t end_time;
481 482 483 484 485
    int64_t bytes_xfer_now;

    if (!bytes_xfer_prev) {
        bytes_xfer_prev = ram_bytes_transferred();
    }
486 487

    if (!start_time) {
488
        start_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
489
    }
490 491

    trace_migration_bitmap_sync_start();
492
    address_space_sync_dirty_bitmap(&address_space_memory);
Juan Quintela's avatar
Juan Quintela committed
493

494
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
495
        migration_bitmap_sync_range(block->mr->ram_addr, block->length);
Juan Quintela's avatar
Juan Quintela committed
496 497
    }
    trace_migration_bitmap_sync_end(migration_dirty_pages
498
                                    - num_dirty_pages_init);
499
    num_dirty_pages_period += migration_dirty_pages - num_dirty_pages_init;
500
    end_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
501 502 503

    /* more than 1 second = 1000 millisecons */
    if (end_time > start_time + 1000) {
504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
        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;
        }
523 524
        s->dirty_pages_rate = num_dirty_pages_period * 1000
            / (end_time - start_time);
525
        s->dirty_bytes_rate = s->dirty_pages_rate * TARGET_PAGE_SIZE;
526 527 528
        start_time = end_time;
        num_dirty_pages_period = 0;
    }
529 530
}

531 532 533
/*
 * ram_save_block: Writes a page of memory to the stream f
 *
534 535
 * Returns:  The number of bytes written.
 *           0 means no dirty pages
536 537
 */

538
static int ram_save_block(QEMUFile *f, bool last_stage)
539
{
540
    RAMBlock *block = last_seen_block;
541
    ram_addr_t offset = last_offset;
542
    bool complete_round = false;
543
    int bytes_sent = 0;
Avi Kivity's avatar
Avi Kivity committed
544
    MemoryRegion *mr;
545
    ram_addr_t current_addr;
546

547
    if (!block)
548
        block = QTAILQ_FIRST(&ram_list.blocks);
549

550
    while (true) {
Avi Kivity's avatar
Avi Kivity committed
551
        mr = block->mr;
552 553 554 555 556 557 558 559 560 561 562
        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;
563
                ram_bulk_stage = false;
564 565
            }
        } else {
Michael R. Hines's avatar
Michael R. Hines committed
566
            int ret;
567
            uint8_t *p;
568
            bool send_async = true;
Juan Quintela's avatar
Juan Quintela committed
569
            int cont = (block == last_sent_block) ?
570
                RAM_SAVE_FLAG_CONTINUE : 0;
571

Avi Kivity's avatar
Avi Kivity committed
572
            p = memory_region_get_ram_ptr(mr) + offset;
573

574 575
            /* In doubt sent page as normal */
            bytes_sent = -1;
Michael R. Hines's avatar
Michael R. Hines committed
576 577 578
            ret = ram_control_save_page(f, block->offset,
                               offset, TARGET_PAGE_SIZE, &bytes_sent);

579 580
            XBZRLE_cache_lock();

581
            current_addr = block->offset + offset;
Michael R. Hines's avatar
Michael R. Hines committed
582 583 584 585 586 587 588 589
            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++;
                    }
                }
590
            } else if (is_zero_range(p, TARGET_PAGE_SIZE)) {
591
                acct_info.dup_pages++;
592 593 594 595
                bytes_sent = save_block_hdr(f, block, offset, cont,
                                            RAM_SAVE_FLAG_COMPRESS);
                qemu_put_byte(f, 0);
                bytes_sent++;
596 597 598 599
                /* Must let xbzrle know, otherwise a previous (now 0'd) cached
                 * page would be stale
                 */
                xbzrle_cache_zero_page(current_addr);
600
            } else if (!ram_bulk_stage && migrate_use_xbzrle()) {
601
                bytes_sent = save_xbzrle_page(f, p, current_addr, block,
602 603
                                              offset, cont, last_stage);
                if (!last_stage) {
604 605 606 607
                    /* We must send exactly what's in the xbzrle cache
                     * even if the page wasn't xbzrle compressed, so that
                     * it's right next time.
                     */
608
                    p = get_cached_data(XBZRLE.cache, current_addr);
609 610 611 612 613

                    /* Can't send this cached data async, since the cache page
                     * might get updated before it gets to the wire
                     */
                    send_async = false;
614
                }
615 616
            }

617
            /* XBZRLE overflow or normal page */
618
            if (bytes_sent == -1) {
619
                bytes_sent = save_block_hdr(f, block, offset, cont, RAM_SAVE_FLAG_PAGE);
620 621 622 623 624
                if (send_async) {
                    qemu_put_buffer_async(f, p, TARGET_PAGE_SIZE);
                } else {
                    qemu_put_buffer(f, p, TARGET_PAGE_SIZE);
                }
625
                bytes_sent += TARGET_PAGE_SIZE;
626
                acct_info.norm_pages++;
627 628
            }

629
            XBZRLE_cache_unlock();
630
            /* if page is unmodified, continue to the next */
631
            if (bytes_sent > 0) {
Juan Quintela's avatar
Juan Quintela committed
632
                last_sent_block = block;
633 634
                break;
            }
635
        }
636
    }
637
    last_seen_block = block;
638
    last_offset = offset;
639

640
    return bytes_sent;
641 642 643 644
}

static uint64_t bytes_transferred;

645 646 647 648 649 650 651 652 653 654 655 656
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);
    }
}

657 658
static ram_addr_t ram_save_remaining(void)
{
Juan Quintela's avatar
Juan Quintela committed
659
    return migration_dirty_pages;
660 661 662 663 664 665 666 667 668 669 670 671 672 673
}

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)
{
674 675 676
    RAMBlock *block;
    uint64_t total = 0;

677
    QTAILQ_FOREACH(block, &ram_list.blocks, next)
678 679 680
        total += block->length;

    return total;
681 682
}

683 684 685 686 687 688
void free_xbzrle_decoded_buf(void)
{
    g_free(xbzrle_decoded_buf);
    xbzrle_decoded_buf = NULL;
}

Orit Wasserman's avatar
Orit Wasserman committed
689 690
static void migration_end(void)
{
691 692 693 694 695
    if (migration_bitmap) {
        memory_global_dirty_log_stop();
        g_free(migration_bitmap);
        migration_bitmap = NULL;
    }
696

697
    XBZRLE_cache_lock();
698
    if (XBZRLE.cache) {
699 700 701 702 703
        cache_fini(XBZRLE.cache);
        g_free(XBZRLE.cache);
        g_free(XBZRLE.encoded_buf);
        g_free(XBZRLE.current_buf);
        XBZRLE.cache = NULL;
704 705
        XBZRLE.encoded_buf = NULL;
        XBZRLE.current_buf = NULL;
706
    }
707
    XBZRLE_cache_unlock();
Orit Wasserman's avatar
Orit Wasserman committed
708 709
}

710 711 712 713 714
static void ram_migration_cancel(void *opaque)
{
    migration_end();
}

715 716
static void reset_ram_globals(void)
{
717
    last_seen_block = NULL;
Juan Quintela's avatar
Juan Quintela committed
718
    last_sent_block = NULL;
719
    last_offset = 0;
720
    last_version = ram_list.version;
721
    ram_bulk_stage = true;
722 723
}

724 725
#define MAX_WAIT 50 /* ms, half buffered_file limit */

726
static int ram_save_setup(QEMUFile *f, void *opaque)
727
{
728
    RAMBlock *block;
729
    int64_t ram_bitmap_pages; /* Size of bitmap in pages, including gaps */
Juan Quintela's avatar
Juan Quintela committed
730

731 732
    mig_throttle_on = false;
    dirty_rate_high_cnt = 0;
733

734
    if (migrate_use_xbzrle()) {
735
        qemu_mutex_lock_iothread();
736 737 738 739
        XBZRLE.cache = cache_init(migrate_xbzrle_cache_size() /
                                  TARGET_PAGE_SIZE,
                                  TARGET_PAGE_SIZE);
        if (!XBZRLE.cache) {
740
            qemu_mutex_unlock_iothread();
741 742 743
            DPRINTF("Error creating cache\n");
            return -1;
        }
744 745
        qemu_mutex_init(&XBZRLE.lock);
        qemu_mutex_unlock_iothread();
746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761

        /* We prefer not to abort if there is no memory */
        XBZRLE.encoded_buf = g_try_malloc0(TARGET_PAGE_SIZE);
        if (!XBZRLE.encoded_buf) {
            DPRINTF("Error allocating encoded_buf\n");
            return -1;
        }

        XBZRLE.current_buf = g_try_malloc(TARGET_PAGE_SIZE);
        if (!XBZRLE.current_buf) {
            DPRINTF("Error allocating current_buf\n");
            g_free(XBZRLE.encoded_buf);
            XBZRLE.encoded_buf = NULL;
            return -1;
        }

762
        acct_clear();
763 764
    }

765 766 767 768 769
    qemu_mutex_lock_iothread();
    qemu_mutex_lock_ramlist();
    bytes_transferred = 0;
    reset_ram_globals();

770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785
    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;
    }

786
    memory_global_dirty_log_start();
Juan Quintela's avatar
Juan Quintela committed
787
    migration_bitmap_sync();
788
    qemu_mutex_unlock_iothread();
789

790
    qemu_put_be64(f, ram_bytes_total() | RAM_SAVE_FLAG_MEM_SIZE);
791

792
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
793 794 795
        qemu_put_byte(f, strlen(block->idstr));
        qemu_put_buffer(f, (uint8_t *)block->idstr, strlen(block->idstr));
        qemu_put_be64(f, block->length);
796 797
    }

798
    qemu_mutex_unlock_ramlist();
Michael R. Hines's avatar
Michael R. Hines committed
799 800 801 802

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

803 804 805 806 807
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);

    return 0;
}

808
static int ram_save_iterate(QEMUFile *f, void *opaque)
809 810 811
{
    int ret;
    int i;
812
    int64_t t0;
813
    int total_sent = 0;
814

815 816
    qemu_mutex_lock_ramlist();

817 818 819 820
    if (ram_list.version != last_version) {
        reset_ram_globals();
    }

Michael R. Hines's avatar
Michael R. Hines committed
821 822
    ram_control_before_iterate(f, RAM_CONTROL_ROUND);

823
    t0 = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
824
    i = 0;
825
    while ((ret = qemu_file_rate_limit(f)) == 0) {
826
        int bytes_sent;
827

828
        bytes_sent = ram_save_block(f, false);
829
        /* no more blocks to sent */
830
        if (bytes_sent == 0) {
831 832
            break;
        }
833
        total_sent += bytes_sent;
834
        acct_info.iterations++;
835
        check_guest_throttling();
836 837 838 839 840 841
        /* 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) {
842
            uint64_t t1 = (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - t0) / 1000000;
843
            if (t1 > MAX_WAIT) {
844
                DPRINTF("big wait: %" PRIu64 " milliseconds, %d iterations\n",
845 846 847 848 849
                        t1, i);
                break;
            }
        }
        i++;
850 851
    }

852 853
    qemu_mutex_unlock_ramlist();

Michael R. Hines's avatar
Michael R. Hines committed
854 855 856 857 858 859
    /*
     * Must occur before EOS (or any QEMUFile operation)
     * because of RDMA protocol.
     */
    ram_control_after_iterate(f, RAM_CONTROL_ROUND);

860 861 862 863 864 865 866 867 868 869
    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);
870 871 872 873
    if (ret < 0) {
        return ret;
    }

874
    return total_sent;
875 876 877 878
}

static int ram_save_complete(QEMUFile *f, void *opaque)
{
879
    qemu_mutex_lock_ramlist();
880
    migration_bitmap_sync();
881

Michael R. Hines's avatar
Michael R. Hines committed
882 883
    ram_control_before_iterate(f, RAM_CONTROL_FINISH);

884
    /* try transferring iterative blocks of memory */
Orit Wasserman's avatar
Orit Wasserman committed
885

886
    /* flush all remaining blocks regardless of rate limiting */
887
    while (true) {
888 889
        int bytes_sent;

890
        bytes_sent = ram_save_block(f, true);
891
        /* no more blocks to sent */
892
        if (bytes_sent == 0) {
893
            break;
894
        }
895
        bytes_transferred += bytes_sent;
896
    }
Michael R. Hines's avatar
Michael R. Hines committed
897 898

    ram_control_after_iterate(f, RAM_CONTROL_FINISH);
899
    migration_end();
900

901
    qemu_mutex_unlock_ramlist();
902 903
    qemu_put_be64(f, RAM_SAVE_FLAG_EOS);

904
    return 0;
905 906
}

907 908 909 910 911 912 913
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) {
914
        qemu_mutex_lock_iothread();
915
        migration_bitmap_sync();
916
        qemu_mutex_unlock_iothread();
917 918 919 920 921
        remaining_size = ram_save_remaining() * TARGET_PAGE_SIZE;
    }
    return remaining_size;
}

922 923 924 925 926 927
static int load_xbzrle(QEMUFile *f, ram_addr_t addr, void *host)
{
    int ret, rc = 0;
    unsigned int xh_len;
    int xh_flags;

928 929
    if (!xbzrle_decoded_buf) {
        xbzrle_decoded_buf = g_malloc(TARGET_PAGE_SIZE);
930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945
    }

    /* 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 */
946
    qemu_get_buffer(f, xbzrle_decoded_buf, xh_len);
947 948

    /* decode RLE */
949
    ret = xbzrle_decode_buffer(xbzrle_decoded_buf, xh_len, host,
950 951 952 953 954 955 956 957 958 959 960 961 962
                               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;
}

963 964 965 966 967 968 969 970 971 972 973 974 975 976
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;
        }

977
        return memory_region_get_ram_ptr(block->mr) + offset;
978 979 980 981 982 983
    }

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

984
    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
985
        if (!strncmp(id, block->idstr, sizeof(id)))
986
            return memory_region_get_ram_ptr(block->mr) + offset;
987 988 989 990 991 992
    }

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

993 994 995 996 997 998
/*
 * 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)
{
999
    if (ch != 0 || !is_zero_range(host, size)) {
1000 1001 1002 1003
        memset(host, ch, size);
    }
}

1004
static int ram_load(QEMUFile *f, void *opaque, int version_id)
1005 1006
{
    ram_addr_t addr;
Orit Wasserman's avatar
Orit Wasserman committed
1007
    int flags, ret = 0;
1008
    int error;
Orit Wasserman's avatar
Orit Wasserman committed
1009 1010 1011
    static uint64_t seq_iter;

    seq_iter++;
1012

1013
    if (version_id < 4 || version_id > 4) {
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023
        return -EINVAL;
    }

    do {
        addr = qemu_get_be64(f);

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

        if (flags & RAM_SAVE_FLAG_MEM_SIZE) {
1024
            if (version_id == 4) {
1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
                /* 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);

1039
                    QTAILQ_FOREACH(block, &ram_list.blocks, next) {
1040
                        if (!strncmp(id, block->idstr, sizeof(id))) {
Orit Wasserman's avatar
Orit Wasserman committed
1041
                            if (block->length != length) {
1042 1043 1044
                                fprintf(stderr,
                                        "Length mismatch: %s: " RAM_ADDR_FMT
                                        " in != " RAM_ADDR_FMT "\n", id, length,
1045
                                        block->length);
Orit Wasserman's avatar
Orit Wasserman committed
1046 1047 1048
                                ret =  -EINVAL;
                                goto done;
                            }
1049 1050 1051 1052 1053
                            break;
                        }
                    }

                    if (!block) {
1054 1055
                        fprintf(stderr, "Unknown ramblock \"%s\", cannot "
                                "accept migration\n", id);
Orit Wasserman's avatar
Orit Wasserman committed
1056 1057
                        ret = -EINVAL;
                        goto done;
1058 1059 1060 1061
                    }

                    total_ram_bytes -= length;
                }
1062 1063 1064 1065
            }
        }

        if (flags & RAM_SAVE_FLAG_COMPRESS) {
1066 1067 1068
            void *host;
            uint8_t ch;

1069
            host = host_from_stream_offset(f, addr, flags);
1070 1071 1072
            if (!host) {
                return -EINVAL;
            }
1073 1074

            ch = qemu_get_byte(f);
1075
            ram_handle_compressed(host, ch, TARGET_PAGE_SIZE);
1076
        } else if (flags & RAM_SAVE_FLAG_PAGE) {
1077 1078
            void *host;

1079
            host = host_from_stream_offset(f, addr, flags);
1080 1081 1082
            if (!host) {
                return -EINVAL;
            }
1083 1084

            qemu_get_buffer(f, host, TARGET_PAGE_SIZE);
Orit Wasserman's avatar