1 /* 2 * Image mirroring 3 * 4 * Copyright Red Hat, Inc. 2012 5 * 6 * Authors: 7 * Paolo Bonzini <pbonzini@redhat.com> 8 * 9 * This work is licensed under the terms of the GNU LGPL, version 2 or later. 10 * See the COPYING.LIB file in the top-level directory. 11 * 12 */ 13 14 #include "qemu/osdep.h" 15 #include "qemu/cutils.h" 16 #include "qemu/coroutine.h" 17 #include "qemu/range.h" 18 #include "trace.h" 19 #include "block/blockjob_int.h" 20 #include "block/block_int.h" 21 #include "block/dirty-bitmap.h" 22 #include "sysemu/block-backend.h" 23 #include "qapi/error.h" 24 #include "qapi/qmp/qerror.h" 25 #include "qemu/ratelimit.h" 26 #include "qemu/bitmap.h" 27 #include "qemu/memalign.h" 28 29 #define MAX_IN_FLIGHT 16 30 #define MAX_IO_BYTES (1 << 20) /* 1 Mb */ 31 #define DEFAULT_MIRROR_BUF_SIZE (MAX_IN_FLIGHT * MAX_IO_BYTES) 32 33 /* The mirroring buffer is a list of granularity-sized chunks. 34 * Free chunks are organized in a list. 35 */ 36 typedef struct MirrorBuffer { 37 QSIMPLEQ_ENTRY(MirrorBuffer) next; 38 } MirrorBuffer; 39 40 typedef struct MirrorOp MirrorOp; 41 42 typedef struct MirrorBlockJob { 43 BlockJob common; 44 BlockBackend *target; 45 BlockDriverState *mirror_top_bs; 46 BlockDriverState *base; 47 BlockDriverState *base_overlay; 48 49 /* The name of the graph node to replace */ 50 char *replaces; 51 /* The BDS to replace */ 52 BlockDriverState *to_replace; 53 /* Used to block operations on the drive-mirror-replace target */ 54 Error *replace_blocker; 55 bool is_none_mode; 56 BlockMirrorBackingMode backing_mode; 57 /* Whether the target image requires explicit zero-initialization */ 58 bool zero_target; 59 MirrorCopyMode copy_mode; 60 BlockdevOnError on_source_error, on_target_error; 61 /* Set when the target is synced (dirty bitmap is clean, nothing 62 * in flight) and the job is running in active mode */ 63 bool actively_synced; 64 bool should_complete; 65 int64_t granularity; 66 size_t buf_size; 67 int64_t bdev_length; 68 unsigned long *cow_bitmap; 69 BdrvDirtyBitmap *dirty_bitmap; 70 BdrvDirtyBitmapIter *dbi; 71 uint8_t *buf; 72 QSIMPLEQ_HEAD(, MirrorBuffer) buf_free; 73 int buf_free_count; 74 75 uint64_t last_pause_ns; 76 unsigned long *in_flight_bitmap; 77 unsigned in_flight; 78 int64_t bytes_in_flight; 79 QTAILQ_HEAD(, MirrorOp) ops_in_flight; 80 int ret; 81 bool unmap; 82 int target_cluster_size; 83 int max_iov; 84 bool initial_zeroing_ongoing; 85 int in_active_write_counter; 86 int64_t active_write_bytes_in_flight; 87 bool prepared; 88 bool in_drain; 89 } MirrorBlockJob; 90 91 typedef struct MirrorBDSOpaque { 92 MirrorBlockJob *job; 93 bool stop; 94 bool is_commit; 95 } MirrorBDSOpaque; 96 97 struct MirrorOp { 98 MirrorBlockJob *s; 99 QEMUIOVector qiov; 100 int64_t offset; 101 uint64_t bytes; 102 103 /* The pointee is set by mirror_co_read(), mirror_co_zero(), and 104 * mirror_co_discard() before yielding for the first time */ 105 int64_t *bytes_handled; 106 107 bool is_pseudo_op; 108 bool is_active_write; 109 bool is_in_flight; 110 CoQueue waiting_requests; 111 Coroutine *co; 112 MirrorOp *waiting_for_op; 113 114 QTAILQ_ENTRY(MirrorOp) next; 115 }; 116 117 typedef enum MirrorMethod { 118 MIRROR_METHOD_COPY, 119 MIRROR_METHOD_ZERO, 120 MIRROR_METHOD_DISCARD, 121 } MirrorMethod; 122 123 static BlockErrorAction mirror_error_action(MirrorBlockJob *s, bool read, 124 int error) 125 { 126 s->actively_synced = false; 127 if (read) { 128 return block_job_error_action(&s->common, s->on_source_error, 129 true, error); 130 } else { 131 return block_job_error_action(&s->common, s->on_target_error, 132 false, error); 133 } 134 } 135 136 static void coroutine_fn mirror_wait_on_conflicts(MirrorOp *self, 137 MirrorBlockJob *s, 138 uint64_t offset, 139 uint64_t bytes) 140 { 141 uint64_t self_start_chunk = offset / s->granularity; 142 uint64_t self_end_chunk = DIV_ROUND_UP(offset + bytes, s->granularity); 143 uint64_t self_nb_chunks = self_end_chunk - self_start_chunk; 144 145 while (find_next_bit(s->in_flight_bitmap, self_end_chunk, 146 self_start_chunk) < self_end_chunk && 147 s->ret >= 0) 148 { 149 MirrorOp *op; 150 151 QTAILQ_FOREACH(op, &s->ops_in_flight, next) { 152 uint64_t op_start_chunk = op->offset / s->granularity; 153 uint64_t op_nb_chunks = DIV_ROUND_UP(op->offset + op->bytes, 154 s->granularity) - 155 op_start_chunk; 156 157 if (op == self) { 158 continue; 159 } 160 161 if (ranges_overlap(self_start_chunk, self_nb_chunks, 162 op_start_chunk, op_nb_chunks)) 163 { 164 if (self) { 165 /* 166 * If the operation is already (indirectly) waiting for us, 167 * or will wait for us as soon as it wakes up, then just go 168 * on (instead of producing a deadlock in the former case). 169 */ 170 if (op->waiting_for_op) { 171 continue; 172 } 173 174 self->waiting_for_op = op; 175 } 176 177 qemu_co_queue_wait(&op->waiting_requests, NULL); 178 179 if (self) { 180 self->waiting_for_op = NULL; 181 } 182 183 break; 184 } 185 } 186 } 187 } 188 189 static void coroutine_fn mirror_iteration_done(MirrorOp *op, int ret) 190 { 191 MirrorBlockJob *s = op->s; 192 struct iovec *iov; 193 int64_t chunk_num; 194 int i, nb_chunks; 195 196 trace_mirror_iteration_done(s, op->offset, op->bytes, ret); 197 198 s->in_flight--; 199 s->bytes_in_flight -= op->bytes; 200 iov = op->qiov.iov; 201 for (i = 0; i < op->qiov.niov; i++) { 202 MirrorBuffer *buf = (MirrorBuffer *) iov[i].iov_base; 203 QSIMPLEQ_INSERT_TAIL(&s->buf_free, buf, next); 204 s->buf_free_count++; 205 } 206 207 chunk_num = op->offset / s->granularity; 208 nb_chunks = DIV_ROUND_UP(op->bytes, s->granularity); 209 210 bitmap_clear(s->in_flight_bitmap, chunk_num, nb_chunks); 211 QTAILQ_REMOVE(&s->ops_in_flight, op, next); 212 if (ret >= 0) { 213 if (s->cow_bitmap) { 214 bitmap_set(s->cow_bitmap, chunk_num, nb_chunks); 215 } 216 if (!s->initial_zeroing_ongoing) { 217 job_progress_update(&s->common.job, op->bytes); 218 } 219 } 220 qemu_iovec_destroy(&op->qiov); 221 222 qemu_co_queue_restart_all(&op->waiting_requests); 223 g_free(op); 224 } 225 226 static void coroutine_fn mirror_write_complete(MirrorOp *op, int ret) 227 { 228 MirrorBlockJob *s = op->s; 229 230 if (ret < 0) { 231 BlockErrorAction action; 232 233 bdrv_set_dirty_bitmap(s->dirty_bitmap, op->offset, op->bytes); 234 action = mirror_error_action(s, false, -ret); 235 if (action == BLOCK_ERROR_ACTION_REPORT && s->ret >= 0) { 236 s->ret = ret; 237 } 238 } 239 240 mirror_iteration_done(op, ret); 241 } 242 243 static void coroutine_fn mirror_read_complete(MirrorOp *op, int ret) 244 { 245 MirrorBlockJob *s = op->s; 246 247 if (ret < 0) { 248 BlockErrorAction action; 249 250 bdrv_set_dirty_bitmap(s->dirty_bitmap, op->offset, op->bytes); 251 action = mirror_error_action(s, true, -ret); 252 if (action == BLOCK_ERROR_ACTION_REPORT && s->ret >= 0) { 253 s->ret = ret; 254 } 255 256 mirror_iteration_done(op, ret); 257 return; 258 } 259 260 ret = blk_co_pwritev(s->target, op->offset, op->qiov.size, &op->qiov, 0); 261 mirror_write_complete(op, ret); 262 } 263 264 /* Clip bytes relative to offset to not exceed end-of-file */ 265 static inline int64_t mirror_clip_bytes(MirrorBlockJob *s, 266 int64_t offset, 267 int64_t bytes) 268 { 269 return MIN(bytes, s->bdev_length - offset); 270 } 271 272 /* Round offset and/or bytes to target cluster if COW is needed, and 273 * return the offset of the adjusted tail against original. */ 274 static int mirror_cow_align(MirrorBlockJob *s, int64_t *offset, 275 uint64_t *bytes) 276 { 277 bool need_cow; 278 int ret = 0; 279 int64_t align_offset = *offset; 280 int64_t align_bytes = *bytes; 281 int max_bytes = s->granularity * s->max_iov; 282 283 need_cow = !test_bit(*offset / s->granularity, s->cow_bitmap); 284 need_cow |= !test_bit((*offset + *bytes - 1) / s->granularity, 285 s->cow_bitmap); 286 if (need_cow) { 287 bdrv_round_to_clusters(blk_bs(s->target), *offset, *bytes, 288 &align_offset, &align_bytes); 289 } 290 291 if (align_bytes > max_bytes) { 292 align_bytes = max_bytes; 293 if (need_cow) { 294 align_bytes = QEMU_ALIGN_DOWN(align_bytes, s->target_cluster_size); 295 } 296 } 297 /* Clipping may result in align_bytes unaligned to chunk boundary, but 298 * that doesn't matter because it's already the end of source image. */ 299 align_bytes = mirror_clip_bytes(s, align_offset, align_bytes); 300 301 ret = align_offset + align_bytes - (*offset + *bytes); 302 *offset = align_offset; 303 *bytes = align_bytes; 304 assert(ret >= 0); 305 return ret; 306 } 307 308 static inline void coroutine_fn 309 mirror_wait_for_free_in_flight_slot(MirrorBlockJob *s) 310 { 311 MirrorOp *op; 312 313 QTAILQ_FOREACH(op, &s->ops_in_flight, next) { 314 /* 315 * Do not wait on pseudo ops, because it may in turn wait on 316 * some other operation to start, which may in fact be the 317 * caller of this function. Since there is only one pseudo op 318 * at any given time, we will always find some real operation 319 * to wait on. 320 * Also, do not wait on active operations, because they do not 321 * use up in-flight slots. 322 */ 323 if (!op->is_pseudo_op && op->is_in_flight && !op->is_active_write) { 324 qemu_co_queue_wait(&op->waiting_requests, NULL); 325 return; 326 } 327 } 328 abort(); 329 } 330 331 /* Perform a mirror copy operation. 332 * 333 * *op->bytes_handled is set to the number of bytes copied after and 334 * including offset, excluding any bytes copied prior to offset due 335 * to alignment. This will be op->bytes if no alignment is necessary, 336 * or (new_end - op->offset) if the tail is rounded up or down due to 337 * alignment or buffer limit. 338 */ 339 static void coroutine_fn mirror_co_read(void *opaque) 340 { 341 MirrorOp *op = opaque; 342 MirrorBlockJob *s = op->s; 343 int nb_chunks; 344 uint64_t ret; 345 uint64_t max_bytes; 346 347 max_bytes = s->granularity * s->max_iov; 348 349 /* We can only handle as much as buf_size at a time. */ 350 op->bytes = MIN(s->buf_size, MIN(max_bytes, op->bytes)); 351 assert(op->bytes); 352 assert(op->bytes < BDRV_REQUEST_MAX_BYTES); 353 *op->bytes_handled = op->bytes; 354 355 if (s->cow_bitmap) { 356 *op->bytes_handled += mirror_cow_align(s, &op->offset, &op->bytes); 357 } 358 /* Cannot exceed BDRV_REQUEST_MAX_BYTES + INT_MAX */ 359 assert(*op->bytes_handled <= UINT_MAX); 360 assert(op->bytes <= s->buf_size); 361 /* The offset is granularity-aligned because: 362 * 1) Caller passes in aligned values; 363 * 2) mirror_cow_align is used only when target cluster is larger. */ 364 assert(QEMU_IS_ALIGNED(op->offset, s->granularity)); 365 /* The range is sector-aligned, since bdrv_getlength() rounds up. */ 366 assert(QEMU_IS_ALIGNED(op->bytes, BDRV_SECTOR_SIZE)); 367 nb_chunks = DIV_ROUND_UP(op->bytes, s->granularity); 368 369 while (s->buf_free_count < nb_chunks) { 370 trace_mirror_yield_in_flight(s, op->offset, s->in_flight); 371 mirror_wait_for_free_in_flight_slot(s); 372 } 373 374 /* Now make a QEMUIOVector taking enough granularity-sized chunks 375 * from s->buf_free. 376 */ 377 qemu_iovec_init(&op->qiov, nb_chunks); 378 while (nb_chunks-- > 0) { 379 MirrorBuffer *buf = QSIMPLEQ_FIRST(&s->buf_free); 380 size_t remaining = op->bytes - op->qiov.size; 381 382 QSIMPLEQ_REMOVE_HEAD(&s->buf_free, next); 383 s->buf_free_count--; 384 qemu_iovec_add(&op->qiov, buf, MIN(s->granularity, remaining)); 385 } 386 387 /* Copy the dirty cluster. */ 388 s->in_flight++; 389 s->bytes_in_flight += op->bytes; 390 op->is_in_flight = true; 391 trace_mirror_one_iteration(s, op->offset, op->bytes); 392 393 WITH_GRAPH_RDLOCK_GUARD() { 394 ret = bdrv_co_preadv(s->mirror_top_bs->backing, op->offset, op->bytes, 395 &op->qiov, 0); 396 } 397 mirror_read_complete(op, ret); 398 } 399 400 static void coroutine_fn mirror_co_zero(void *opaque) 401 { 402 MirrorOp *op = opaque; 403 int ret; 404 405 op->s->in_flight++; 406 op->s->bytes_in_flight += op->bytes; 407 *op->bytes_handled = op->bytes; 408 op->is_in_flight = true; 409 410 ret = blk_co_pwrite_zeroes(op->s->target, op->offset, op->bytes, 411 op->s->unmap ? BDRV_REQ_MAY_UNMAP : 0); 412 mirror_write_complete(op, ret); 413 } 414 415 static void coroutine_fn mirror_co_discard(void *opaque) 416 { 417 MirrorOp *op = opaque; 418 int ret; 419 420 op->s->in_flight++; 421 op->s->bytes_in_flight += op->bytes; 422 *op->bytes_handled = op->bytes; 423 op->is_in_flight = true; 424 425 ret = blk_co_pdiscard(op->s->target, op->offset, op->bytes); 426 mirror_write_complete(op, ret); 427 } 428 429 static unsigned mirror_perform(MirrorBlockJob *s, int64_t offset, 430 unsigned bytes, MirrorMethod mirror_method) 431 { 432 MirrorOp *op; 433 Coroutine *co; 434 int64_t bytes_handled = -1; 435 436 op = g_new(MirrorOp, 1); 437 *op = (MirrorOp){ 438 .s = s, 439 .offset = offset, 440 .bytes = bytes, 441 .bytes_handled = &bytes_handled, 442 }; 443 qemu_co_queue_init(&op->waiting_requests); 444 445 switch (mirror_method) { 446 case MIRROR_METHOD_COPY: 447 co = qemu_coroutine_create(mirror_co_read, op); 448 break; 449 case MIRROR_METHOD_ZERO: 450 co = qemu_coroutine_create(mirror_co_zero, op); 451 break; 452 case MIRROR_METHOD_DISCARD: 453 co = qemu_coroutine_create(mirror_co_discard, op); 454 break; 455 default: 456 abort(); 457 } 458 op->co = co; 459 460 QTAILQ_INSERT_TAIL(&s->ops_in_flight, op, next); 461 qemu_coroutine_enter(co); 462 /* At this point, ownership of op has been moved to the coroutine 463 * and the object may already be freed */ 464 465 /* Assert that this value has been set */ 466 assert(bytes_handled >= 0); 467 468 /* Same assertion as in mirror_co_read() (and for mirror_co_read() 469 * and mirror_co_discard(), bytes_handled == op->bytes, which 470 * is the @bytes parameter given to this function) */ 471 assert(bytes_handled <= UINT_MAX); 472 return bytes_handled; 473 } 474 475 static uint64_t coroutine_fn mirror_iteration(MirrorBlockJob *s) 476 { 477 BlockDriverState *source = s->mirror_top_bs->backing->bs; 478 MirrorOp *pseudo_op; 479 int64_t offset; 480 uint64_t delay_ns = 0, ret = 0; 481 /* At least the first dirty chunk is mirrored in one iteration. */ 482 int nb_chunks = 1; 483 bool write_zeroes_ok = bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)); 484 int max_io_bytes = MAX(s->buf_size / MAX_IN_FLIGHT, MAX_IO_BYTES); 485 486 bdrv_dirty_bitmap_lock(s->dirty_bitmap); 487 offset = bdrv_dirty_iter_next(s->dbi); 488 if (offset < 0) { 489 bdrv_set_dirty_iter(s->dbi, 0); 490 offset = bdrv_dirty_iter_next(s->dbi); 491 trace_mirror_restart_iter(s, bdrv_get_dirty_count(s->dirty_bitmap)); 492 assert(offset >= 0); 493 } 494 bdrv_dirty_bitmap_unlock(s->dirty_bitmap); 495 496 /* 497 * Wait for concurrent requests to @offset. The next loop will limit the 498 * copied area based on in_flight_bitmap so we only copy an area that does 499 * not overlap with concurrent in-flight requests. Still, we would like to 500 * copy something, so wait until there are at least no more requests to the 501 * very beginning of the area. 502 */ 503 mirror_wait_on_conflicts(NULL, s, offset, 1); 504 505 job_pause_point(&s->common.job); 506 507 /* Find the number of consective dirty chunks following the first dirty 508 * one, and wait for in flight requests in them. */ 509 bdrv_dirty_bitmap_lock(s->dirty_bitmap); 510 while (nb_chunks * s->granularity < s->buf_size) { 511 int64_t next_dirty; 512 int64_t next_offset = offset + nb_chunks * s->granularity; 513 int64_t next_chunk = next_offset / s->granularity; 514 if (next_offset >= s->bdev_length || 515 !bdrv_dirty_bitmap_get_locked(s->dirty_bitmap, next_offset)) { 516 break; 517 } 518 if (test_bit(next_chunk, s->in_flight_bitmap)) { 519 break; 520 } 521 522 next_dirty = bdrv_dirty_iter_next(s->dbi); 523 if (next_dirty > next_offset || next_dirty < 0) { 524 /* The bitmap iterator's cache is stale, refresh it */ 525 bdrv_set_dirty_iter(s->dbi, next_offset); 526 next_dirty = bdrv_dirty_iter_next(s->dbi); 527 } 528 assert(next_dirty == next_offset); 529 nb_chunks++; 530 } 531 532 /* Clear dirty bits before querying the block status, because 533 * calling bdrv_block_status_above could yield - if some blocks are 534 * marked dirty in this window, we need to know. 535 */ 536 bdrv_reset_dirty_bitmap_locked(s->dirty_bitmap, offset, 537 nb_chunks * s->granularity); 538 bdrv_dirty_bitmap_unlock(s->dirty_bitmap); 539 540 /* Before claiming an area in the in-flight bitmap, we have to 541 * create a MirrorOp for it so that conflicting requests can wait 542 * for it. mirror_perform() will create the real MirrorOps later, 543 * for now we just create a pseudo operation that will wake up all 544 * conflicting requests once all real operations have been 545 * launched. */ 546 pseudo_op = g_new(MirrorOp, 1); 547 *pseudo_op = (MirrorOp){ 548 .offset = offset, 549 .bytes = nb_chunks * s->granularity, 550 .is_pseudo_op = true, 551 }; 552 qemu_co_queue_init(&pseudo_op->waiting_requests); 553 QTAILQ_INSERT_TAIL(&s->ops_in_flight, pseudo_op, next); 554 555 bitmap_set(s->in_flight_bitmap, offset / s->granularity, nb_chunks); 556 while (nb_chunks > 0 && offset < s->bdev_length) { 557 int ret; 558 int64_t io_bytes; 559 int64_t io_bytes_acct; 560 MirrorMethod mirror_method = MIRROR_METHOD_COPY; 561 562 assert(!(offset % s->granularity)); 563 WITH_GRAPH_RDLOCK_GUARD() { 564 ret = bdrv_block_status_above(source, NULL, offset, 565 nb_chunks * s->granularity, 566 &io_bytes, NULL, NULL); 567 } 568 if (ret < 0) { 569 io_bytes = MIN(nb_chunks * s->granularity, max_io_bytes); 570 } else if (ret & BDRV_BLOCK_DATA) { 571 io_bytes = MIN(io_bytes, max_io_bytes); 572 } 573 574 io_bytes -= io_bytes % s->granularity; 575 if (io_bytes < s->granularity) { 576 io_bytes = s->granularity; 577 } else if (ret >= 0 && !(ret & BDRV_BLOCK_DATA)) { 578 int64_t target_offset; 579 int64_t target_bytes; 580 bdrv_round_to_clusters(blk_bs(s->target), offset, io_bytes, 581 &target_offset, &target_bytes); 582 if (target_offset == offset && 583 target_bytes == io_bytes) { 584 mirror_method = ret & BDRV_BLOCK_ZERO ? 585 MIRROR_METHOD_ZERO : 586 MIRROR_METHOD_DISCARD; 587 } 588 } 589 590 while (s->in_flight >= MAX_IN_FLIGHT) { 591 trace_mirror_yield_in_flight(s, offset, s->in_flight); 592 mirror_wait_for_free_in_flight_slot(s); 593 } 594 595 if (s->ret < 0) { 596 ret = 0; 597 goto fail; 598 } 599 600 io_bytes = mirror_clip_bytes(s, offset, io_bytes); 601 io_bytes = mirror_perform(s, offset, io_bytes, mirror_method); 602 if (mirror_method != MIRROR_METHOD_COPY && write_zeroes_ok) { 603 io_bytes_acct = 0; 604 } else { 605 io_bytes_acct = io_bytes; 606 } 607 assert(io_bytes); 608 offset += io_bytes; 609 nb_chunks -= DIV_ROUND_UP(io_bytes, s->granularity); 610 delay_ns = block_job_ratelimit_get_delay(&s->common, io_bytes_acct); 611 } 612 613 ret = delay_ns; 614 fail: 615 QTAILQ_REMOVE(&s->ops_in_flight, pseudo_op, next); 616 qemu_co_queue_restart_all(&pseudo_op->waiting_requests); 617 g_free(pseudo_op); 618 619 return ret; 620 } 621 622 static void mirror_free_init(MirrorBlockJob *s) 623 { 624 int granularity = s->granularity; 625 size_t buf_size = s->buf_size; 626 uint8_t *buf = s->buf; 627 628 assert(s->buf_free_count == 0); 629 QSIMPLEQ_INIT(&s->buf_free); 630 while (buf_size != 0) { 631 MirrorBuffer *cur = (MirrorBuffer *)buf; 632 QSIMPLEQ_INSERT_TAIL(&s->buf_free, cur, next); 633 s->buf_free_count++; 634 buf_size -= granularity; 635 buf += granularity; 636 } 637 } 638 639 /* This is also used for the .pause callback. There is no matching 640 * mirror_resume() because mirror_run() will begin iterating again 641 * when the job is resumed. 642 */ 643 static void coroutine_fn mirror_wait_for_all_io(MirrorBlockJob *s) 644 { 645 while (s->in_flight > 0) { 646 mirror_wait_for_free_in_flight_slot(s); 647 } 648 } 649 650 /** 651 * mirror_exit_common: handle both abort() and prepare() cases. 652 * for .prepare, returns 0 on success and -errno on failure. 653 * for .abort cases, denoted by abort = true, MUST return 0. 654 */ 655 static int mirror_exit_common(Job *job) 656 { 657 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 658 BlockJob *bjob = &s->common; 659 MirrorBDSOpaque *bs_opaque; 660 AioContext *replace_aio_context = NULL; 661 BlockDriverState *src; 662 BlockDriverState *target_bs; 663 BlockDriverState *mirror_top_bs; 664 Error *local_err = NULL; 665 bool abort = job->ret < 0; 666 int ret = 0; 667 668 if (s->prepared) { 669 return 0; 670 } 671 s->prepared = true; 672 673 mirror_top_bs = s->mirror_top_bs; 674 bs_opaque = mirror_top_bs->opaque; 675 src = mirror_top_bs->backing->bs; 676 target_bs = blk_bs(s->target); 677 678 if (bdrv_chain_contains(src, target_bs)) { 679 bdrv_unfreeze_backing_chain(mirror_top_bs, target_bs); 680 } 681 682 bdrv_release_dirty_bitmap(s->dirty_bitmap); 683 684 /* Make sure that the source BDS doesn't go away during bdrv_replace_node, 685 * before we can call bdrv_drained_end */ 686 bdrv_ref(src); 687 bdrv_ref(mirror_top_bs); 688 bdrv_ref(target_bs); 689 690 /* 691 * Remove target parent that still uses BLK_PERM_WRITE/RESIZE before 692 * inserting target_bs at s->to_replace, where we might not be able to get 693 * these permissions. 694 */ 695 blk_unref(s->target); 696 s->target = NULL; 697 698 /* We don't access the source any more. Dropping any WRITE/RESIZE is 699 * required before it could become a backing file of target_bs. Not having 700 * these permissions any more means that we can't allow any new requests on 701 * mirror_top_bs from now on, so keep it drained. */ 702 bdrv_drained_begin(mirror_top_bs); 703 bs_opaque->stop = true; 704 bdrv_child_refresh_perms(mirror_top_bs, mirror_top_bs->backing, 705 &error_abort); 706 if (!abort && s->backing_mode == MIRROR_SOURCE_BACKING_CHAIN) { 707 BlockDriverState *backing = s->is_none_mode ? src : s->base; 708 BlockDriverState *unfiltered_target = bdrv_skip_filters(target_bs); 709 710 if (bdrv_cow_bs(unfiltered_target) != backing) { 711 bdrv_set_backing_hd(unfiltered_target, backing, &local_err); 712 if (local_err) { 713 error_report_err(local_err); 714 local_err = NULL; 715 ret = -EPERM; 716 } 717 } 718 } else if (!abort && s->backing_mode == MIRROR_OPEN_BACKING_CHAIN) { 719 assert(!bdrv_backing_chain_next(target_bs)); 720 ret = bdrv_open_backing_file(bdrv_skip_filters(target_bs), NULL, 721 "backing", &local_err); 722 if (ret < 0) { 723 error_report_err(local_err); 724 local_err = NULL; 725 } 726 } 727 728 if (s->to_replace) { 729 replace_aio_context = bdrv_get_aio_context(s->to_replace); 730 aio_context_acquire(replace_aio_context); 731 } 732 733 if (s->should_complete && !abort) { 734 BlockDriverState *to_replace = s->to_replace ?: src; 735 bool ro = bdrv_is_read_only(to_replace); 736 737 if (ro != bdrv_is_read_only(target_bs)) { 738 bdrv_reopen_set_read_only(target_bs, ro, NULL); 739 } 740 741 /* The mirror job has no requests in flight any more, but we need to 742 * drain potential other users of the BDS before changing the graph. */ 743 assert(s->in_drain); 744 bdrv_drained_begin(target_bs); 745 /* 746 * Cannot use check_to_replace_node() here, because that would 747 * check for an op blocker on @to_replace, and we have our own 748 * there. 749 */ 750 if (bdrv_recurse_can_replace(src, to_replace)) { 751 bdrv_replace_node(to_replace, target_bs, &local_err); 752 } else { 753 error_setg(&local_err, "Can no longer replace '%s' by '%s', " 754 "because it can no longer be guaranteed that doing so " 755 "would not lead to an abrupt change of visible data", 756 to_replace->node_name, target_bs->node_name); 757 } 758 bdrv_drained_end(target_bs); 759 if (local_err) { 760 error_report_err(local_err); 761 ret = -EPERM; 762 } 763 } 764 if (s->to_replace) { 765 bdrv_op_unblock_all(s->to_replace, s->replace_blocker); 766 error_free(s->replace_blocker); 767 bdrv_unref(s->to_replace); 768 } 769 if (replace_aio_context) { 770 aio_context_release(replace_aio_context); 771 } 772 g_free(s->replaces); 773 bdrv_unref(target_bs); 774 775 /* 776 * Remove the mirror filter driver from the graph. Before this, get rid of 777 * the blockers on the intermediate nodes so that the resulting state is 778 * valid. 779 */ 780 block_job_remove_all_bdrv(bjob); 781 bdrv_replace_node(mirror_top_bs, mirror_top_bs->backing->bs, &error_abort); 782 783 bs_opaque->job = NULL; 784 785 bdrv_drained_end(src); 786 bdrv_drained_end(mirror_top_bs); 787 s->in_drain = false; 788 bdrv_unref(mirror_top_bs); 789 bdrv_unref(src); 790 791 return ret; 792 } 793 794 static int mirror_prepare(Job *job) 795 { 796 return mirror_exit_common(job); 797 } 798 799 static void mirror_abort(Job *job) 800 { 801 int ret = mirror_exit_common(job); 802 assert(ret == 0); 803 } 804 805 static void coroutine_fn mirror_throttle(MirrorBlockJob *s) 806 { 807 int64_t now = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); 808 809 if (now - s->last_pause_ns > BLOCK_JOB_SLICE_TIME) { 810 s->last_pause_ns = now; 811 job_sleep_ns(&s->common.job, 0); 812 } else { 813 job_pause_point(&s->common.job); 814 } 815 } 816 817 static int coroutine_fn mirror_dirty_init(MirrorBlockJob *s) 818 { 819 int64_t offset; 820 BlockDriverState *bs = s->mirror_top_bs->backing->bs; 821 BlockDriverState *target_bs = blk_bs(s->target); 822 int ret; 823 int64_t count; 824 825 if (s->zero_target) { 826 if (!bdrv_can_write_zeroes_with_unmap(target_bs)) { 827 bdrv_set_dirty_bitmap(s->dirty_bitmap, 0, s->bdev_length); 828 return 0; 829 } 830 831 s->initial_zeroing_ongoing = true; 832 for (offset = 0; offset < s->bdev_length; ) { 833 int bytes = MIN(s->bdev_length - offset, 834 QEMU_ALIGN_DOWN(INT_MAX, s->granularity)); 835 836 mirror_throttle(s); 837 838 if (job_is_cancelled(&s->common.job)) { 839 s->initial_zeroing_ongoing = false; 840 return 0; 841 } 842 843 if (s->in_flight >= MAX_IN_FLIGHT) { 844 trace_mirror_yield(s, UINT64_MAX, s->buf_free_count, 845 s->in_flight); 846 mirror_wait_for_free_in_flight_slot(s); 847 continue; 848 } 849 850 mirror_perform(s, offset, bytes, MIRROR_METHOD_ZERO); 851 offset += bytes; 852 } 853 854 mirror_wait_for_all_io(s); 855 s->initial_zeroing_ongoing = false; 856 } 857 858 /* First part, loop on the sectors and initialize the dirty bitmap. */ 859 for (offset = 0; offset < s->bdev_length; ) { 860 /* Just to make sure we are not exceeding int limit. */ 861 int bytes = MIN(s->bdev_length - offset, 862 QEMU_ALIGN_DOWN(INT_MAX, s->granularity)); 863 864 mirror_throttle(s); 865 866 if (job_is_cancelled(&s->common.job)) { 867 return 0; 868 } 869 870 WITH_GRAPH_RDLOCK_GUARD() { 871 ret = bdrv_is_allocated_above(bs, s->base_overlay, true, offset, 872 bytes, &count); 873 } 874 if (ret < 0) { 875 return ret; 876 } 877 878 assert(count); 879 if (ret > 0) { 880 bdrv_set_dirty_bitmap(s->dirty_bitmap, offset, count); 881 } 882 offset += count; 883 } 884 return 0; 885 } 886 887 /* Called when going out of the streaming phase to flush the bulk of the 888 * data to the medium, or just before completing. 889 */ 890 static int mirror_flush(MirrorBlockJob *s) 891 { 892 int ret = blk_flush(s->target); 893 if (ret < 0) { 894 if (mirror_error_action(s, false, -ret) == BLOCK_ERROR_ACTION_REPORT) { 895 s->ret = ret; 896 } 897 } 898 return ret; 899 } 900 901 static int coroutine_fn mirror_run(Job *job, Error **errp) 902 { 903 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 904 BlockDriverState *bs = s->mirror_top_bs->backing->bs; 905 MirrorBDSOpaque *mirror_top_opaque = s->mirror_top_bs->opaque; 906 BlockDriverState *target_bs = blk_bs(s->target); 907 bool need_drain = true; 908 BlockDeviceIoStatus iostatus; 909 int64_t length; 910 int64_t target_length; 911 BlockDriverInfo bdi; 912 char backing_filename[2]; /* we only need 2 characters because we are only 913 checking for a NULL string */ 914 int ret = 0; 915 916 if (job_is_cancelled(&s->common.job)) { 917 goto immediate_exit; 918 } 919 920 bdrv_graph_co_rdlock(); 921 s->bdev_length = bdrv_co_getlength(bs); 922 bdrv_graph_co_rdunlock(); 923 924 if (s->bdev_length < 0) { 925 ret = s->bdev_length; 926 goto immediate_exit; 927 } 928 929 target_length = blk_co_getlength(s->target); 930 if (target_length < 0) { 931 ret = target_length; 932 goto immediate_exit; 933 } 934 935 /* Active commit must resize the base image if its size differs from the 936 * active layer. */ 937 if (s->base == blk_bs(s->target)) { 938 if (s->bdev_length > target_length) { 939 ret = blk_co_truncate(s->target, s->bdev_length, false, 940 PREALLOC_MODE_OFF, 0, NULL); 941 if (ret < 0) { 942 goto immediate_exit; 943 } 944 } 945 } else if (s->bdev_length != target_length) { 946 error_setg(errp, "Source and target image have different sizes"); 947 ret = -EINVAL; 948 goto immediate_exit; 949 } 950 951 if (s->bdev_length == 0) { 952 /* Transition to the READY state and wait for complete. */ 953 job_transition_to_ready(&s->common.job); 954 s->actively_synced = true; 955 while (!job_cancel_requested(&s->common.job) && !s->should_complete) { 956 job_yield(&s->common.job); 957 } 958 goto immediate_exit; 959 } 960 961 length = DIV_ROUND_UP(s->bdev_length, s->granularity); 962 s->in_flight_bitmap = bitmap_new(length); 963 964 /* If we have no backing file yet in the destination, we cannot let 965 * the destination do COW. Instead, we copy sectors around the 966 * dirty data if needed. We need a bitmap to do that. 967 */ 968 bdrv_get_backing_filename(target_bs, backing_filename, 969 sizeof(backing_filename)); 970 if (!bdrv_co_get_info(target_bs, &bdi) && bdi.cluster_size) { 971 s->target_cluster_size = bdi.cluster_size; 972 } else { 973 s->target_cluster_size = BDRV_SECTOR_SIZE; 974 } 975 if (backing_filename[0] && !bdrv_backing_chain_next(target_bs) && 976 s->granularity < s->target_cluster_size) { 977 s->buf_size = MAX(s->buf_size, s->target_cluster_size); 978 s->cow_bitmap = bitmap_new(length); 979 } 980 s->max_iov = MIN(bs->bl.max_iov, target_bs->bl.max_iov); 981 982 s->buf = qemu_try_blockalign(bs, s->buf_size); 983 if (s->buf == NULL) { 984 ret = -ENOMEM; 985 goto immediate_exit; 986 } 987 988 mirror_free_init(s); 989 990 s->last_pause_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); 991 if (!s->is_none_mode) { 992 ret = mirror_dirty_init(s); 993 if (ret < 0 || job_is_cancelled(&s->common.job)) { 994 goto immediate_exit; 995 } 996 } 997 998 /* 999 * Only now the job is fully initialised and mirror_top_bs should start 1000 * accessing it. 1001 */ 1002 mirror_top_opaque->job = s; 1003 1004 assert(!s->dbi); 1005 s->dbi = bdrv_dirty_iter_new(s->dirty_bitmap); 1006 for (;;) { 1007 uint64_t delay_ns = 0; 1008 int64_t cnt, delta; 1009 bool should_complete; 1010 1011 if (s->ret < 0) { 1012 ret = s->ret; 1013 goto immediate_exit; 1014 } 1015 1016 job_pause_point(&s->common.job); 1017 1018 if (job_is_cancelled(&s->common.job)) { 1019 ret = 0; 1020 goto immediate_exit; 1021 } 1022 1023 cnt = bdrv_get_dirty_count(s->dirty_bitmap); 1024 /* cnt is the number of dirty bytes remaining and s->bytes_in_flight is 1025 * the number of bytes currently being processed; together those are 1026 * the current remaining operation length */ 1027 job_progress_set_remaining(&s->common.job, 1028 s->bytes_in_flight + cnt + 1029 s->active_write_bytes_in_flight); 1030 1031 /* Note that even when no rate limit is applied we need to yield 1032 * periodically with no pending I/O so that bdrv_drain_all() returns. 1033 * We do so every BLKOCK_JOB_SLICE_TIME nanoseconds, or when there is 1034 * an error, or when the source is clean, whichever comes first. */ 1035 delta = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - s->last_pause_ns; 1036 WITH_JOB_LOCK_GUARD() { 1037 iostatus = s->common.iostatus; 1038 } 1039 if (delta < BLOCK_JOB_SLICE_TIME && 1040 iostatus == BLOCK_DEVICE_IO_STATUS_OK) { 1041 if (s->in_flight >= MAX_IN_FLIGHT || s->buf_free_count == 0 || 1042 (cnt == 0 && s->in_flight > 0)) { 1043 trace_mirror_yield(s, cnt, s->buf_free_count, s->in_flight); 1044 mirror_wait_for_free_in_flight_slot(s); 1045 continue; 1046 } else if (cnt != 0) { 1047 delay_ns = mirror_iteration(s); 1048 } 1049 } 1050 1051 should_complete = false; 1052 if (s->in_flight == 0 && cnt == 0) { 1053 trace_mirror_before_flush(s); 1054 if (!job_is_ready(&s->common.job)) { 1055 if (mirror_flush(s) < 0) { 1056 /* Go check s->ret. */ 1057 continue; 1058 } 1059 /* We're out of the streaming phase. From now on, if the job 1060 * is cancelled we will actually complete all pending I/O and 1061 * report completion. This way, block-job-cancel will leave 1062 * the target in a consistent state. 1063 */ 1064 job_transition_to_ready(&s->common.job); 1065 if (s->copy_mode != MIRROR_COPY_MODE_BACKGROUND) { 1066 s->actively_synced = true; 1067 } 1068 } 1069 1070 should_complete = s->should_complete || 1071 job_cancel_requested(&s->common.job); 1072 cnt = bdrv_get_dirty_count(s->dirty_bitmap); 1073 } 1074 1075 if (cnt == 0 && should_complete) { 1076 /* The dirty bitmap is not updated while operations are pending. 1077 * If we're about to exit, wait for pending operations before 1078 * calling bdrv_get_dirty_count(bs), or we may exit while the 1079 * source has dirty data to copy! 1080 * 1081 * Note that I/O can be submitted by the guest while 1082 * mirror_populate runs, so pause it now. Before deciding 1083 * whether to switch to target check one last time if I/O has 1084 * come in the meanwhile, and if not flush the data to disk. 1085 */ 1086 trace_mirror_before_drain(s, cnt); 1087 1088 s->in_drain = true; 1089 bdrv_drained_begin(bs); 1090 1091 /* Must be zero because we are drained */ 1092 assert(s->in_active_write_counter == 0); 1093 1094 cnt = bdrv_get_dirty_count(s->dirty_bitmap); 1095 if (cnt > 0 || mirror_flush(s) < 0) { 1096 bdrv_drained_end(bs); 1097 s->in_drain = false; 1098 continue; 1099 } 1100 1101 /* The two disks are in sync. Exit and report successful 1102 * completion. 1103 */ 1104 assert(QLIST_EMPTY(&bs->tracked_requests)); 1105 need_drain = false; 1106 break; 1107 } 1108 1109 if (job_is_ready(&s->common.job) && !should_complete) { 1110 delay_ns = (s->in_flight == 0 && 1111 cnt == 0 ? BLOCK_JOB_SLICE_TIME : 0); 1112 } 1113 trace_mirror_before_sleep(s, cnt, job_is_ready(&s->common.job), 1114 delay_ns); 1115 job_sleep_ns(&s->common.job, delay_ns); 1116 s->last_pause_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); 1117 } 1118 1119 immediate_exit: 1120 if (s->in_flight > 0) { 1121 /* We get here only if something went wrong. Either the job failed, 1122 * or it was cancelled prematurely so that we do not guarantee that 1123 * the target is a copy of the source. 1124 */ 1125 assert(ret < 0 || job_is_cancelled(&s->common.job)); 1126 assert(need_drain); 1127 mirror_wait_for_all_io(s); 1128 } 1129 1130 assert(s->in_flight == 0); 1131 qemu_vfree(s->buf); 1132 g_free(s->cow_bitmap); 1133 g_free(s->in_flight_bitmap); 1134 bdrv_dirty_iter_free(s->dbi); 1135 1136 if (need_drain) { 1137 s->in_drain = true; 1138 bdrv_drained_begin(bs); 1139 } 1140 1141 return ret; 1142 } 1143 1144 static void mirror_complete(Job *job, Error **errp) 1145 { 1146 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 1147 1148 if (!job_is_ready(job)) { 1149 error_setg(errp, "The active block job '%s' cannot be completed", 1150 job->id); 1151 return; 1152 } 1153 1154 /* block all operations on to_replace bs */ 1155 if (s->replaces) { 1156 AioContext *replace_aio_context; 1157 1158 s->to_replace = bdrv_find_node(s->replaces); 1159 if (!s->to_replace) { 1160 error_setg(errp, "Node name '%s' not found", s->replaces); 1161 return; 1162 } 1163 1164 replace_aio_context = bdrv_get_aio_context(s->to_replace); 1165 aio_context_acquire(replace_aio_context); 1166 1167 /* TODO Translate this into child freeze system. */ 1168 error_setg(&s->replace_blocker, 1169 "block device is in use by block-job-complete"); 1170 bdrv_op_block_all(s->to_replace, s->replace_blocker); 1171 bdrv_ref(s->to_replace); 1172 1173 aio_context_release(replace_aio_context); 1174 } 1175 1176 s->should_complete = true; 1177 1178 /* If the job is paused, it will be re-entered when it is resumed */ 1179 WITH_JOB_LOCK_GUARD() { 1180 if (!job->paused) { 1181 job_enter_cond_locked(job, NULL); 1182 } 1183 } 1184 } 1185 1186 static void coroutine_fn mirror_pause(Job *job) 1187 { 1188 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 1189 1190 mirror_wait_for_all_io(s); 1191 } 1192 1193 static bool mirror_drained_poll(BlockJob *job) 1194 { 1195 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common); 1196 1197 /* If the job isn't paused nor cancelled, we can't be sure that it won't 1198 * issue more requests. We make an exception if we've reached this point 1199 * from one of our own drain sections, to avoid a deadlock waiting for 1200 * ourselves. 1201 */ 1202 WITH_JOB_LOCK_GUARD() { 1203 if (!s->common.job.paused && !job_is_cancelled_locked(&job->job) 1204 && !s->in_drain) { 1205 return true; 1206 } 1207 } 1208 1209 return !!s->in_flight; 1210 } 1211 1212 static bool mirror_cancel(Job *job, bool force) 1213 { 1214 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 1215 BlockDriverState *target = blk_bs(s->target); 1216 1217 /* 1218 * Before the job is READY, we treat any cancellation like a 1219 * force-cancellation. 1220 */ 1221 force = force || !job_is_ready(job); 1222 1223 if (force) { 1224 bdrv_cancel_in_flight(target); 1225 } 1226 return force; 1227 } 1228 1229 static bool commit_active_cancel(Job *job, bool force) 1230 { 1231 /* Same as above in mirror_cancel() */ 1232 return force || !job_is_ready(job); 1233 } 1234 1235 static const BlockJobDriver mirror_job_driver = { 1236 .job_driver = { 1237 .instance_size = sizeof(MirrorBlockJob), 1238 .job_type = JOB_TYPE_MIRROR, 1239 .free = block_job_free, 1240 .user_resume = block_job_user_resume, 1241 .run = mirror_run, 1242 .prepare = mirror_prepare, 1243 .abort = mirror_abort, 1244 .pause = mirror_pause, 1245 .complete = mirror_complete, 1246 .cancel = mirror_cancel, 1247 }, 1248 .drained_poll = mirror_drained_poll, 1249 }; 1250 1251 static const BlockJobDriver commit_active_job_driver = { 1252 .job_driver = { 1253 .instance_size = sizeof(MirrorBlockJob), 1254 .job_type = JOB_TYPE_COMMIT, 1255 .free = block_job_free, 1256 .user_resume = block_job_user_resume, 1257 .run = mirror_run, 1258 .prepare = mirror_prepare, 1259 .abort = mirror_abort, 1260 .pause = mirror_pause, 1261 .complete = mirror_complete, 1262 .cancel = commit_active_cancel, 1263 }, 1264 .drained_poll = mirror_drained_poll, 1265 }; 1266 1267 static void coroutine_fn 1268 do_sync_target_write(MirrorBlockJob *job, MirrorMethod method, 1269 uint64_t offset, uint64_t bytes, 1270 QEMUIOVector *qiov, int flags) 1271 { 1272 int ret; 1273 size_t qiov_offset = 0; 1274 int64_t bitmap_offset, bitmap_end; 1275 1276 if (!QEMU_IS_ALIGNED(offset, job->granularity) && 1277 bdrv_dirty_bitmap_get(job->dirty_bitmap, offset)) 1278 { 1279 /* 1280 * Dirty unaligned padding: ignore it. 1281 * 1282 * Reasoning: 1283 * 1. If we copy it, we can't reset corresponding bit in 1284 * dirty_bitmap as there may be some "dirty" bytes still not 1285 * copied. 1286 * 2. It's already dirty, so skipping it we don't diverge mirror 1287 * progress. 1288 * 1289 * Note, that because of this, guest write may have no contribution 1290 * into mirror converge, but that's not bad, as we have background 1291 * process of mirroring. If under some bad circumstances (high guest 1292 * IO load) background process starve, we will not converge anyway, 1293 * even if each write will contribute, as guest is not guaranteed to 1294 * rewrite the whole disk. 1295 */ 1296 qiov_offset = QEMU_ALIGN_UP(offset, job->granularity) - offset; 1297 if (bytes <= qiov_offset) { 1298 /* nothing to do after shrink */ 1299 return; 1300 } 1301 offset += qiov_offset; 1302 bytes -= qiov_offset; 1303 } 1304 1305 if (!QEMU_IS_ALIGNED(offset + bytes, job->granularity) && 1306 bdrv_dirty_bitmap_get(job->dirty_bitmap, offset + bytes - 1)) 1307 { 1308 uint64_t tail = (offset + bytes) % job->granularity; 1309 1310 if (bytes <= tail) { 1311 /* nothing to do after shrink */ 1312 return; 1313 } 1314 bytes -= tail; 1315 } 1316 1317 /* 1318 * Tails are either clean or shrunk, so for bitmap resetting 1319 * we safely align the range down. 1320 */ 1321 bitmap_offset = QEMU_ALIGN_UP(offset, job->granularity); 1322 bitmap_end = QEMU_ALIGN_DOWN(offset + bytes, job->granularity); 1323 if (bitmap_offset < bitmap_end) { 1324 bdrv_reset_dirty_bitmap(job->dirty_bitmap, bitmap_offset, 1325 bitmap_end - bitmap_offset); 1326 } 1327 1328 job_progress_increase_remaining(&job->common.job, bytes); 1329 job->active_write_bytes_in_flight += bytes; 1330 1331 switch (method) { 1332 case MIRROR_METHOD_COPY: 1333 ret = blk_co_pwritev_part(job->target, offset, bytes, 1334 qiov, qiov_offset, flags); 1335 break; 1336 1337 case MIRROR_METHOD_ZERO: 1338 assert(!qiov); 1339 ret = blk_co_pwrite_zeroes(job->target, offset, bytes, flags); 1340 break; 1341 1342 case MIRROR_METHOD_DISCARD: 1343 assert(!qiov); 1344 ret = blk_co_pdiscard(job->target, offset, bytes); 1345 break; 1346 1347 default: 1348 abort(); 1349 } 1350 1351 job->active_write_bytes_in_flight -= bytes; 1352 if (ret >= 0) { 1353 job_progress_update(&job->common.job, bytes); 1354 } else { 1355 BlockErrorAction action; 1356 1357 /* 1358 * We failed, so we should mark dirty the whole area, aligned up. 1359 * Note that we don't care about shrunk tails if any: they were dirty 1360 * at function start, and they must be still dirty, as we've locked 1361 * the region for in-flight op. 1362 */ 1363 bitmap_offset = QEMU_ALIGN_DOWN(offset, job->granularity); 1364 bitmap_end = QEMU_ALIGN_UP(offset + bytes, job->granularity); 1365 bdrv_set_dirty_bitmap(job->dirty_bitmap, bitmap_offset, 1366 bitmap_end - bitmap_offset); 1367 job->actively_synced = false; 1368 1369 action = mirror_error_action(job, false, -ret); 1370 if (action == BLOCK_ERROR_ACTION_REPORT) { 1371 if (!job->ret) { 1372 job->ret = ret; 1373 } 1374 } 1375 } 1376 } 1377 1378 static MirrorOp *coroutine_fn active_write_prepare(MirrorBlockJob *s, 1379 uint64_t offset, 1380 uint64_t bytes) 1381 { 1382 MirrorOp *op; 1383 uint64_t start_chunk = offset / s->granularity; 1384 uint64_t end_chunk = DIV_ROUND_UP(offset + bytes, s->granularity); 1385 1386 op = g_new(MirrorOp, 1); 1387 *op = (MirrorOp){ 1388 .s = s, 1389 .offset = offset, 1390 .bytes = bytes, 1391 .is_active_write = true, 1392 .is_in_flight = true, 1393 .co = qemu_coroutine_self(), 1394 }; 1395 qemu_co_queue_init(&op->waiting_requests); 1396 QTAILQ_INSERT_TAIL(&s->ops_in_flight, op, next); 1397 1398 s->in_active_write_counter++; 1399 1400 /* 1401 * Wait for concurrent requests affecting the area. If there are already 1402 * running requests that are copying off now-to-be stale data in the area, 1403 * we must wait for them to finish before we begin writing fresh data to the 1404 * target so that the write operations appear in the correct order. 1405 * Note that background requests (see mirror_iteration()) in contrast only 1406 * wait for conflicting requests at the start of the dirty area, and then 1407 * (based on the in_flight_bitmap) truncate the area to copy so it will not 1408 * conflict with any requests beyond that. For active writes, however, we 1409 * cannot truncate that area. The request from our parent must be blocked 1410 * until the area is copied in full. Therefore, we must wait for the whole 1411 * area to become free of concurrent requests. 1412 */ 1413 mirror_wait_on_conflicts(op, s, offset, bytes); 1414 1415 bitmap_set(s->in_flight_bitmap, start_chunk, end_chunk - start_chunk); 1416 1417 return op; 1418 } 1419 1420 static void coroutine_fn active_write_settle(MirrorOp *op) 1421 { 1422 uint64_t start_chunk = op->offset / op->s->granularity; 1423 uint64_t end_chunk = DIV_ROUND_UP(op->offset + op->bytes, 1424 op->s->granularity); 1425 1426 if (!--op->s->in_active_write_counter && op->s->actively_synced) { 1427 BdrvChild *source = op->s->mirror_top_bs->backing; 1428 1429 if (QLIST_FIRST(&source->bs->parents) == source && 1430 QLIST_NEXT(source, next_parent) == NULL) 1431 { 1432 /* Assert that we are back in sync once all active write 1433 * operations are settled. 1434 * Note that we can only assert this if the mirror node 1435 * is the source node's only parent. */ 1436 assert(!bdrv_get_dirty_count(op->s->dirty_bitmap)); 1437 } 1438 } 1439 bitmap_clear(op->s->in_flight_bitmap, start_chunk, end_chunk - start_chunk); 1440 QTAILQ_REMOVE(&op->s->ops_in_flight, op, next); 1441 qemu_co_queue_restart_all(&op->waiting_requests); 1442 g_free(op); 1443 } 1444 1445 static int coroutine_fn GRAPH_RDLOCK 1446 bdrv_mirror_top_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes, 1447 QEMUIOVector *qiov, BdrvRequestFlags flags) 1448 { 1449 return bdrv_co_preadv(bs->backing, offset, bytes, qiov, flags); 1450 } 1451 1452 static int coroutine_fn GRAPH_RDLOCK 1453 bdrv_mirror_top_do_write(BlockDriverState *bs, MirrorMethod method, 1454 uint64_t offset, uint64_t bytes, QEMUIOVector *qiov, 1455 int flags) 1456 { 1457 MirrorOp *op = NULL; 1458 MirrorBDSOpaque *s = bs->opaque; 1459 int ret = 0; 1460 bool copy_to_target = false; 1461 1462 if (s->job) { 1463 copy_to_target = s->job->ret >= 0 && 1464 !job_is_cancelled(&s->job->common.job) && 1465 s->job->copy_mode == MIRROR_COPY_MODE_WRITE_BLOCKING; 1466 } 1467 1468 if (copy_to_target) { 1469 op = active_write_prepare(s->job, offset, bytes); 1470 } 1471 1472 switch (method) { 1473 case MIRROR_METHOD_COPY: 1474 ret = bdrv_co_pwritev(bs->backing, offset, bytes, qiov, flags); 1475 break; 1476 1477 case MIRROR_METHOD_ZERO: 1478 ret = bdrv_co_pwrite_zeroes(bs->backing, offset, bytes, flags); 1479 break; 1480 1481 case MIRROR_METHOD_DISCARD: 1482 ret = bdrv_co_pdiscard(bs->backing, offset, bytes); 1483 break; 1484 1485 default: 1486 abort(); 1487 } 1488 1489 if (ret < 0) { 1490 goto out; 1491 } 1492 1493 if (copy_to_target) { 1494 do_sync_target_write(s->job, method, offset, bytes, qiov, flags); 1495 } 1496 1497 out: 1498 if (copy_to_target) { 1499 active_write_settle(op); 1500 } 1501 return ret; 1502 } 1503 1504 static int coroutine_fn GRAPH_RDLOCK 1505 bdrv_mirror_top_pwritev(BlockDriverState *bs, int64_t offset, int64_t bytes, 1506 QEMUIOVector *qiov, BdrvRequestFlags flags) 1507 { 1508 MirrorBDSOpaque *s = bs->opaque; 1509 QEMUIOVector bounce_qiov; 1510 void *bounce_buf; 1511 int ret = 0; 1512 bool copy_to_target = false; 1513 1514 if (s->job) { 1515 copy_to_target = s->job->ret >= 0 && 1516 !job_is_cancelled(&s->job->common.job) && 1517 s->job->copy_mode == MIRROR_COPY_MODE_WRITE_BLOCKING; 1518 } 1519 1520 if (copy_to_target) { 1521 /* The guest might concurrently modify the data to write; but 1522 * the data on source and destination must match, so we have 1523 * to use a bounce buffer if we are going to write to the 1524 * target now. */ 1525 bounce_buf = qemu_blockalign(bs, bytes); 1526 iov_to_buf_full(qiov->iov, qiov->niov, 0, bounce_buf, bytes); 1527 1528 qemu_iovec_init(&bounce_qiov, 1); 1529 qemu_iovec_add(&bounce_qiov, bounce_buf, bytes); 1530 qiov = &bounce_qiov; 1531 1532 flags &= ~BDRV_REQ_REGISTERED_BUF; 1533 } 1534 1535 ret = bdrv_mirror_top_do_write(bs, MIRROR_METHOD_COPY, offset, bytes, qiov, 1536 flags); 1537 1538 if (copy_to_target) { 1539 qemu_iovec_destroy(&bounce_qiov); 1540 qemu_vfree(bounce_buf); 1541 } 1542 1543 return ret; 1544 } 1545 1546 static int coroutine_fn GRAPH_RDLOCK bdrv_mirror_top_flush(BlockDriverState *bs) 1547 { 1548 if (bs->backing == NULL) { 1549 /* we can be here after failed bdrv_append in mirror_start_job */ 1550 return 0; 1551 } 1552 return bdrv_co_flush(bs->backing->bs); 1553 } 1554 1555 static int coroutine_fn GRAPH_RDLOCK 1556 bdrv_mirror_top_pwrite_zeroes(BlockDriverState *bs, int64_t offset, 1557 int64_t bytes, BdrvRequestFlags flags) 1558 { 1559 return bdrv_mirror_top_do_write(bs, MIRROR_METHOD_ZERO, offset, bytes, NULL, 1560 flags); 1561 } 1562 1563 static int coroutine_fn GRAPH_RDLOCK 1564 bdrv_mirror_top_pdiscard(BlockDriverState *bs, int64_t offset, int64_t bytes) 1565 { 1566 return bdrv_mirror_top_do_write(bs, MIRROR_METHOD_DISCARD, offset, bytes, 1567 NULL, 0); 1568 } 1569 1570 static void bdrv_mirror_top_refresh_filename(BlockDriverState *bs) 1571 { 1572 if (bs->backing == NULL) { 1573 /* we can be here after failed bdrv_attach_child in 1574 * bdrv_set_backing_hd */ 1575 return; 1576 } 1577 pstrcpy(bs->exact_filename, sizeof(bs->exact_filename), 1578 bs->backing->bs->filename); 1579 } 1580 1581 static void bdrv_mirror_top_child_perm(BlockDriverState *bs, BdrvChild *c, 1582 BdrvChildRole role, 1583 BlockReopenQueue *reopen_queue, 1584 uint64_t perm, uint64_t shared, 1585 uint64_t *nperm, uint64_t *nshared) 1586 { 1587 MirrorBDSOpaque *s = bs->opaque; 1588 1589 if (s->stop) { 1590 /* 1591 * If the job is to be stopped, we do not need to forward 1592 * anything to the real image. 1593 */ 1594 *nperm = 0; 1595 *nshared = BLK_PERM_ALL; 1596 return; 1597 } 1598 1599 bdrv_default_perms(bs, c, role, reopen_queue, 1600 perm, shared, nperm, nshared); 1601 1602 if (s->is_commit) { 1603 /* 1604 * For commit jobs, we cannot take CONSISTENT_READ, because 1605 * that permission is unshared for everything above the base 1606 * node (except for filters on the base node). 1607 * We also have to force-share the WRITE permission, or 1608 * otherwise we would block ourselves at the base node (if 1609 * writes are blocked for a node, they are also blocked for 1610 * its backing file). 1611 * (We could also share RESIZE, because it may be needed for 1612 * the target if its size is less than the top node's; but 1613 * bdrv_default_perms_for_cow() automatically shares RESIZE 1614 * for backing nodes if WRITE is shared, so there is no need 1615 * to do it here.) 1616 */ 1617 *nperm &= ~BLK_PERM_CONSISTENT_READ; 1618 *nshared |= BLK_PERM_WRITE; 1619 } 1620 } 1621 1622 /* Dummy node that provides consistent read to its users without requiring it 1623 * from its backing file and that allows writes on the backing file chain. */ 1624 static BlockDriver bdrv_mirror_top = { 1625 .format_name = "mirror_top", 1626 .bdrv_co_preadv = bdrv_mirror_top_preadv, 1627 .bdrv_co_pwritev = bdrv_mirror_top_pwritev, 1628 .bdrv_co_pwrite_zeroes = bdrv_mirror_top_pwrite_zeroes, 1629 .bdrv_co_pdiscard = bdrv_mirror_top_pdiscard, 1630 .bdrv_co_flush = bdrv_mirror_top_flush, 1631 .bdrv_refresh_filename = bdrv_mirror_top_refresh_filename, 1632 .bdrv_child_perm = bdrv_mirror_top_child_perm, 1633 1634 .is_filter = true, 1635 .filtered_child_is_backing = true, 1636 }; 1637 1638 static BlockJob *mirror_start_job( 1639 const char *job_id, BlockDriverState *bs, 1640 int creation_flags, BlockDriverState *target, 1641 const char *replaces, int64_t speed, 1642 uint32_t granularity, int64_t buf_size, 1643 BlockMirrorBackingMode backing_mode, 1644 bool zero_target, 1645 BlockdevOnError on_source_error, 1646 BlockdevOnError on_target_error, 1647 bool unmap, 1648 BlockCompletionFunc *cb, 1649 void *opaque, 1650 const BlockJobDriver *driver, 1651 bool is_none_mode, BlockDriverState *base, 1652 bool auto_complete, const char *filter_node_name, 1653 bool is_mirror, MirrorCopyMode copy_mode, 1654 Error **errp) 1655 { 1656 MirrorBlockJob *s; 1657 MirrorBDSOpaque *bs_opaque; 1658 BlockDriverState *mirror_top_bs; 1659 bool target_is_backing; 1660 uint64_t target_perms, target_shared_perms; 1661 int ret; 1662 1663 if (granularity == 0) { 1664 granularity = bdrv_get_default_bitmap_granularity(target); 1665 } 1666 1667 assert(is_power_of_2(granularity)); 1668 1669 if (buf_size < 0) { 1670 error_setg(errp, "Invalid parameter 'buf-size'"); 1671 return NULL; 1672 } 1673 1674 if (buf_size == 0) { 1675 buf_size = DEFAULT_MIRROR_BUF_SIZE; 1676 } 1677 1678 if (bdrv_skip_filters(bs) == bdrv_skip_filters(target)) { 1679 error_setg(errp, "Can't mirror node into itself"); 1680 return NULL; 1681 } 1682 1683 target_is_backing = bdrv_chain_contains(bs, target); 1684 1685 /* In the case of active commit, add dummy driver to provide consistent 1686 * reads on the top, while disabling it in the intermediate nodes, and make 1687 * the backing chain writable. */ 1688 mirror_top_bs = bdrv_new_open_driver(&bdrv_mirror_top, filter_node_name, 1689 BDRV_O_RDWR, errp); 1690 if (mirror_top_bs == NULL) { 1691 return NULL; 1692 } 1693 if (!filter_node_name) { 1694 mirror_top_bs->implicit = true; 1695 } 1696 1697 /* So that we can always drop this node */ 1698 mirror_top_bs->never_freeze = true; 1699 1700 mirror_top_bs->total_sectors = bs->total_sectors; 1701 mirror_top_bs->supported_write_flags = BDRV_REQ_WRITE_UNCHANGED; 1702 mirror_top_bs->supported_zero_flags = BDRV_REQ_WRITE_UNCHANGED | 1703 BDRV_REQ_NO_FALLBACK; 1704 bs_opaque = g_new0(MirrorBDSOpaque, 1); 1705 mirror_top_bs->opaque = bs_opaque; 1706 1707 bs_opaque->is_commit = target_is_backing; 1708 1709 bdrv_drained_begin(bs); 1710 ret = bdrv_append(mirror_top_bs, bs, errp); 1711 bdrv_drained_end(bs); 1712 1713 if (ret < 0) { 1714 bdrv_unref(mirror_top_bs); 1715 return NULL; 1716 } 1717 1718 /* Make sure that the source is not resized while the job is running */ 1719 s = block_job_create(job_id, driver, NULL, mirror_top_bs, 1720 BLK_PERM_CONSISTENT_READ, 1721 BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED | 1722 BLK_PERM_WRITE, speed, 1723 creation_flags, cb, opaque, errp); 1724 if (!s) { 1725 goto fail; 1726 } 1727 1728 /* The block job now has a reference to this node */ 1729 bdrv_unref(mirror_top_bs); 1730 1731 s->mirror_top_bs = mirror_top_bs; 1732 1733 /* No resize for the target either; while the mirror is still running, a 1734 * consistent read isn't necessarily possible. We could possibly allow 1735 * writes and graph modifications, though it would likely defeat the 1736 * purpose of a mirror, so leave them blocked for now. 1737 * 1738 * In the case of active commit, things look a bit different, though, 1739 * because the target is an already populated backing file in active use. 1740 * We can allow anything except resize there.*/ 1741 1742 target_perms = BLK_PERM_WRITE; 1743 target_shared_perms = BLK_PERM_WRITE_UNCHANGED; 1744 1745 if (target_is_backing) { 1746 int64_t bs_size, target_size; 1747 bs_size = bdrv_getlength(bs); 1748 if (bs_size < 0) { 1749 error_setg_errno(errp, -bs_size, 1750 "Could not inquire top image size"); 1751 goto fail; 1752 } 1753 1754 target_size = bdrv_getlength(target); 1755 if (target_size < 0) { 1756 error_setg_errno(errp, -target_size, 1757 "Could not inquire base image size"); 1758 goto fail; 1759 } 1760 1761 if (target_size < bs_size) { 1762 target_perms |= BLK_PERM_RESIZE; 1763 } 1764 1765 target_shared_perms |= BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE; 1766 } else if (bdrv_chain_contains(bs, bdrv_skip_filters(target))) { 1767 /* 1768 * We may want to allow this in the future, but it would 1769 * require taking some extra care. 1770 */ 1771 error_setg(errp, "Cannot mirror to a filter on top of a node in the " 1772 "source's backing chain"); 1773 goto fail; 1774 } 1775 1776 s->target = blk_new(s->common.job.aio_context, 1777 target_perms, target_shared_perms); 1778 ret = blk_insert_bs(s->target, target, errp); 1779 if (ret < 0) { 1780 goto fail; 1781 } 1782 if (is_mirror) { 1783 /* XXX: Mirror target could be a NBD server of target QEMU in the case 1784 * of non-shared block migration. To allow migration completion, we 1785 * have to allow "inactivate" of the target BB. When that happens, we 1786 * know the job is drained, and the vcpus are stopped, so no write 1787 * operation will be performed. Block layer already has assertions to 1788 * ensure that. */ 1789 blk_set_force_allow_inactivate(s->target); 1790 } 1791 blk_set_allow_aio_context_change(s->target, true); 1792 blk_set_disable_request_queuing(s->target, true); 1793 1794 s->replaces = g_strdup(replaces); 1795 s->on_source_error = on_source_error; 1796 s->on_target_error = on_target_error; 1797 s->is_none_mode = is_none_mode; 1798 s->backing_mode = backing_mode; 1799 s->zero_target = zero_target; 1800 s->copy_mode = copy_mode; 1801 s->base = base; 1802 s->base_overlay = bdrv_find_overlay(bs, base); 1803 s->granularity = granularity; 1804 s->buf_size = ROUND_UP(buf_size, granularity); 1805 s->unmap = unmap; 1806 if (auto_complete) { 1807 s->should_complete = true; 1808 } 1809 1810 s->dirty_bitmap = bdrv_create_dirty_bitmap(bs, granularity, NULL, errp); 1811 if (!s->dirty_bitmap) { 1812 goto fail; 1813 } 1814 if (s->copy_mode == MIRROR_COPY_MODE_WRITE_BLOCKING) { 1815 bdrv_disable_dirty_bitmap(s->dirty_bitmap); 1816 } 1817 1818 ret = block_job_add_bdrv(&s->common, "source", bs, 0, 1819 BLK_PERM_WRITE_UNCHANGED | BLK_PERM_WRITE | 1820 BLK_PERM_CONSISTENT_READ, 1821 errp); 1822 if (ret < 0) { 1823 goto fail; 1824 } 1825 1826 /* Required permissions are already taken with blk_new() */ 1827 block_job_add_bdrv(&s->common, "target", target, 0, BLK_PERM_ALL, 1828 &error_abort); 1829 1830 /* In commit_active_start() all intermediate nodes disappear, so 1831 * any jobs in them must be blocked */ 1832 if (target_is_backing) { 1833 BlockDriverState *iter, *filtered_target; 1834 uint64_t iter_shared_perms; 1835 1836 /* 1837 * The topmost node with 1838 * bdrv_skip_filters(filtered_target) == bdrv_skip_filters(target) 1839 */ 1840 filtered_target = bdrv_cow_bs(bdrv_find_overlay(bs, target)); 1841 1842 assert(bdrv_skip_filters(filtered_target) == 1843 bdrv_skip_filters(target)); 1844 1845 /* 1846 * XXX BLK_PERM_WRITE needs to be allowed so we don't block 1847 * ourselves at s->base (if writes are blocked for a node, they are 1848 * also blocked for its backing file). The other options would be a 1849 * second filter driver above s->base (== target). 1850 */ 1851 iter_shared_perms = BLK_PERM_WRITE_UNCHANGED | BLK_PERM_WRITE; 1852 1853 for (iter = bdrv_filter_or_cow_bs(bs); iter != target; 1854 iter = bdrv_filter_or_cow_bs(iter)) 1855 { 1856 if (iter == filtered_target) { 1857 /* 1858 * From here on, all nodes are filters on the base. 1859 * This allows us to share BLK_PERM_CONSISTENT_READ. 1860 */ 1861 iter_shared_perms |= BLK_PERM_CONSISTENT_READ; 1862 } 1863 1864 ret = block_job_add_bdrv(&s->common, "intermediate node", iter, 0, 1865 iter_shared_perms, errp); 1866 if (ret < 0) { 1867 goto fail; 1868 } 1869 } 1870 1871 if (bdrv_freeze_backing_chain(mirror_top_bs, target, errp) < 0) { 1872 goto fail; 1873 } 1874 } 1875 1876 QTAILQ_INIT(&s->ops_in_flight); 1877 1878 trace_mirror_start(bs, s, opaque); 1879 job_start(&s->common.job); 1880 1881 return &s->common; 1882 1883 fail: 1884 if (s) { 1885 /* Make sure this BDS does not go away until we have completed the graph 1886 * changes below */ 1887 bdrv_ref(mirror_top_bs); 1888 1889 g_free(s->replaces); 1890 blk_unref(s->target); 1891 bs_opaque->job = NULL; 1892 if (s->dirty_bitmap) { 1893 bdrv_release_dirty_bitmap(s->dirty_bitmap); 1894 } 1895 job_early_fail(&s->common.job); 1896 } 1897 1898 bs_opaque->stop = true; 1899 bdrv_child_refresh_perms(mirror_top_bs, mirror_top_bs->backing, 1900 &error_abort); 1901 bdrv_replace_node(mirror_top_bs, mirror_top_bs->backing->bs, &error_abort); 1902 1903 bdrv_unref(mirror_top_bs); 1904 1905 return NULL; 1906 } 1907 1908 void mirror_start(const char *job_id, BlockDriverState *bs, 1909 BlockDriverState *target, const char *replaces, 1910 int creation_flags, int64_t speed, 1911 uint32_t granularity, int64_t buf_size, 1912 MirrorSyncMode mode, BlockMirrorBackingMode backing_mode, 1913 bool zero_target, 1914 BlockdevOnError on_source_error, 1915 BlockdevOnError on_target_error, 1916 bool unmap, const char *filter_node_name, 1917 MirrorCopyMode copy_mode, Error **errp) 1918 { 1919 bool is_none_mode; 1920 BlockDriverState *base; 1921 1922 GLOBAL_STATE_CODE(); 1923 1924 if ((mode == MIRROR_SYNC_MODE_INCREMENTAL) || 1925 (mode == MIRROR_SYNC_MODE_BITMAP)) { 1926 error_setg(errp, "Sync mode '%s' not supported", 1927 MirrorSyncMode_str(mode)); 1928 return; 1929 } 1930 is_none_mode = mode == MIRROR_SYNC_MODE_NONE; 1931 base = mode == MIRROR_SYNC_MODE_TOP ? bdrv_backing_chain_next(bs) : NULL; 1932 mirror_start_job(job_id, bs, creation_flags, target, replaces, 1933 speed, granularity, buf_size, backing_mode, zero_target, 1934 on_source_error, on_target_error, unmap, NULL, NULL, 1935 &mirror_job_driver, is_none_mode, base, false, 1936 filter_node_name, true, copy_mode, errp); 1937 } 1938 1939 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs, 1940 BlockDriverState *base, int creation_flags, 1941 int64_t speed, BlockdevOnError on_error, 1942 const char *filter_node_name, 1943 BlockCompletionFunc *cb, void *opaque, 1944 bool auto_complete, Error **errp) 1945 { 1946 bool base_read_only; 1947 BlockJob *job; 1948 1949 GLOBAL_STATE_CODE(); 1950 1951 base_read_only = bdrv_is_read_only(base); 1952 1953 if (base_read_only) { 1954 if (bdrv_reopen_set_read_only(base, false, errp) < 0) { 1955 return NULL; 1956 } 1957 } 1958 1959 job = mirror_start_job( 1960 job_id, bs, creation_flags, base, NULL, speed, 0, 0, 1961 MIRROR_LEAVE_BACKING_CHAIN, false, 1962 on_error, on_error, true, cb, opaque, 1963 &commit_active_job_driver, false, base, auto_complete, 1964 filter_node_name, false, MIRROR_COPY_MODE_BACKGROUND, 1965 errp); 1966 if (!job) { 1967 goto error_restore_flags; 1968 } 1969 1970 return job; 1971 1972 error_restore_flags: 1973 /* ignore error and errp for bdrv_reopen, because we want to propagate 1974 * the original error */ 1975 if (base_read_only) { 1976 bdrv_reopen_set_read_only(base, true, NULL); 1977 } 1978 return NULL; 1979 } 1980