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 "qemu/ratelimit.h" 25 #include "qemu/bitmap.h" 26 #include "qemu/memalign.h" 27 28 #define MAX_IN_FLIGHT 16 29 #define MAX_IO_BYTES (1 << 20) /* 1 Mb */ 30 #define DEFAULT_MIRROR_BUF_SIZE (MAX_IN_FLIGHT * MAX_IO_BYTES) 31 32 /* The mirroring buffer is a list of granularity-sized chunks. 33 * Free chunks are organized in a list. 34 */ 35 typedef struct MirrorBuffer { 36 QSIMPLEQ_ENTRY(MirrorBuffer) next; 37 } MirrorBuffer; 38 39 typedef struct MirrorOp MirrorOp; 40 41 typedef struct MirrorBlockJob { 42 BlockJob common; 43 BlockBackend *target; 44 BlockDriverState *mirror_top_bs; 45 BlockDriverState *base; 46 BlockDriverState *base_overlay; 47 48 /* The name of the graph node to replace */ 49 char *replaces; 50 /* The BDS to replace */ 51 BlockDriverState *to_replace; 52 /* Used to block operations on the drive-mirror-replace target */ 53 Error *replace_blocker; 54 bool is_none_mode; 55 BlockMirrorBackingMode backing_mode; 56 /* Whether the target image requires explicit zero-initialization */ 57 bool zero_target; 58 /* 59 * To be accesssed with atomics. Written only under the BQL (required by the 60 * current implementation of mirror_change()). 61 */ 62 MirrorCopyMode copy_mode; 63 BlockdevOnError on_source_error, on_target_error; 64 /* 65 * To be accessed with atomics. 66 * 67 * Set when the target is synced (dirty bitmap is clean, nothing in flight) 68 * and the job is running in active mode. 69 */ 70 bool actively_synced; 71 bool should_complete; 72 int64_t granularity; 73 size_t buf_size; 74 int64_t bdev_length; 75 unsigned long *cow_bitmap; 76 BdrvDirtyBitmap *dirty_bitmap; 77 BdrvDirtyBitmapIter *dbi; 78 uint8_t *buf; 79 QSIMPLEQ_HEAD(, MirrorBuffer) buf_free; 80 int buf_free_count; 81 82 uint64_t last_pause_ns; 83 unsigned long *in_flight_bitmap; 84 unsigned in_flight; 85 int64_t bytes_in_flight; 86 QTAILQ_HEAD(, MirrorOp) ops_in_flight; 87 int ret; 88 bool unmap; 89 int target_cluster_size; 90 int max_iov; 91 bool initial_zeroing_ongoing; 92 int in_active_write_counter; 93 int64_t active_write_bytes_in_flight; 94 bool prepared; 95 bool in_drain; 96 } MirrorBlockJob; 97 98 typedef struct MirrorBDSOpaque { 99 MirrorBlockJob *job; 100 bool stop; 101 bool is_commit; 102 } MirrorBDSOpaque; 103 104 struct MirrorOp { 105 MirrorBlockJob *s; 106 QEMUIOVector qiov; 107 int64_t offset; 108 uint64_t bytes; 109 110 /* The pointee is set by mirror_co_read(), mirror_co_zero(), and 111 * mirror_co_discard() before yielding for the first time */ 112 int64_t *bytes_handled; 113 114 bool is_pseudo_op; 115 bool is_active_write; 116 bool is_in_flight; 117 CoQueue waiting_requests; 118 Coroutine *co; 119 MirrorOp *waiting_for_op; 120 121 QTAILQ_ENTRY(MirrorOp) next; 122 }; 123 124 typedef enum MirrorMethod { 125 MIRROR_METHOD_COPY, 126 MIRROR_METHOD_ZERO, 127 MIRROR_METHOD_DISCARD, 128 } MirrorMethod; 129 130 static BlockErrorAction mirror_error_action(MirrorBlockJob *s, bool read, 131 int error) 132 { 133 qatomic_set(&s->actively_synced, false); 134 if (read) { 135 return block_job_error_action(&s->common, s->on_source_error, 136 true, error); 137 } else { 138 return block_job_error_action(&s->common, s->on_target_error, 139 false, error); 140 } 141 } 142 143 static void coroutine_fn mirror_wait_on_conflicts(MirrorOp *self, 144 MirrorBlockJob *s, 145 uint64_t offset, 146 uint64_t bytes) 147 { 148 uint64_t self_start_chunk = offset / s->granularity; 149 uint64_t self_end_chunk = DIV_ROUND_UP(offset + bytes, s->granularity); 150 uint64_t self_nb_chunks = self_end_chunk - self_start_chunk; 151 152 while (find_next_bit(s->in_flight_bitmap, self_end_chunk, 153 self_start_chunk) < self_end_chunk && 154 s->ret >= 0) 155 { 156 MirrorOp *op; 157 158 QTAILQ_FOREACH(op, &s->ops_in_flight, next) { 159 uint64_t op_start_chunk = op->offset / s->granularity; 160 uint64_t op_nb_chunks = DIV_ROUND_UP(op->offset + op->bytes, 161 s->granularity) - 162 op_start_chunk; 163 164 if (op == self) { 165 continue; 166 } 167 168 if (ranges_overlap(self_start_chunk, self_nb_chunks, 169 op_start_chunk, op_nb_chunks)) 170 { 171 if (self) { 172 /* 173 * If the operation is already (indirectly) waiting for us, 174 * or will wait for us as soon as it wakes up, then just go 175 * on (instead of producing a deadlock in the former case). 176 */ 177 if (op->waiting_for_op) { 178 continue; 179 } 180 181 self->waiting_for_op = op; 182 } 183 184 qemu_co_queue_wait(&op->waiting_requests, NULL); 185 186 if (self) { 187 self->waiting_for_op = NULL; 188 } 189 190 break; 191 } 192 } 193 } 194 } 195 196 static void coroutine_fn mirror_iteration_done(MirrorOp *op, int ret) 197 { 198 MirrorBlockJob *s = op->s; 199 struct iovec *iov; 200 int64_t chunk_num; 201 int i, nb_chunks; 202 203 trace_mirror_iteration_done(s, op->offset, op->bytes, ret); 204 205 s->in_flight--; 206 s->bytes_in_flight -= op->bytes; 207 iov = op->qiov.iov; 208 for (i = 0; i < op->qiov.niov; i++) { 209 MirrorBuffer *buf = (MirrorBuffer *) iov[i].iov_base; 210 QSIMPLEQ_INSERT_TAIL(&s->buf_free, buf, next); 211 s->buf_free_count++; 212 } 213 214 chunk_num = op->offset / s->granularity; 215 nb_chunks = DIV_ROUND_UP(op->bytes, s->granularity); 216 217 bitmap_clear(s->in_flight_bitmap, chunk_num, nb_chunks); 218 QTAILQ_REMOVE(&s->ops_in_flight, op, next); 219 if (ret >= 0) { 220 if (s->cow_bitmap) { 221 bitmap_set(s->cow_bitmap, chunk_num, nb_chunks); 222 } 223 if (!s->initial_zeroing_ongoing) { 224 job_progress_update(&s->common.job, op->bytes); 225 } 226 } 227 qemu_iovec_destroy(&op->qiov); 228 229 qemu_co_queue_restart_all(&op->waiting_requests); 230 g_free(op); 231 } 232 233 static void coroutine_fn mirror_write_complete(MirrorOp *op, int ret) 234 { 235 MirrorBlockJob *s = op->s; 236 237 if (ret < 0) { 238 BlockErrorAction action; 239 240 bdrv_set_dirty_bitmap(s->dirty_bitmap, op->offset, op->bytes); 241 action = mirror_error_action(s, false, -ret); 242 if (action == BLOCK_ERROR_ACTION_REPORT && s->ret >= 0) { 243 s->ret = ret; 244 } 245 } 246 247 mirror_iteration_done(op, ret); 248 } 249 250 static void coroutine_fn mirror_read_complete(MirrorOp *op, int ret) 251 { 252 MirrorBlockJob *s = op->s; 253 254 if (ret < 0) { 255 BlockErrorAction action; 256 257 bdrv_set_dirty_bitmap(s->dirty_bitmap, op->offset, op->bytes); 258 action = mirror_error_action(s, true, -ret); 259 if (action == BLOCK_ERROR_ACTION_REPORT && s->ret >= 0) { 260 s->ret = ret; 261 } 262 263 mirror_iteration_done(op, ret); 264 return; 265 } 266 267 ret = blk_co_pwritev(s->target, op->offset, op->qiov.size, &op->qiov, 0); 268 mirror_write_complete(op, ret); 269 } 270 271 /* Clip bytes relative to offset to not exceed end-of-file */ 272 static inline int64_t mirror_clip_bytes(MirrorBlockJob *s, 273 int64_t offset, 274 int64_t bytes) 275 { 276 return MIN(bytes, s->bdev_length - offset); 277 } 278 279 /* Round offset and/or bytes to target cluster if COW is needed, and 280 * return the offset of the adjusted tail against original. */ 281 static int coroutine_fn mirror_cow_align(MirrorBlockJob *s, int64_t *offset, 282 uint64_t *bytes) 283 { 284 bool need_cow; 285 int ret = 0; 286 int64_t align_offset = *offset; 287 int64_t align_bytes = *bytes; 288 int max_bytes = s->granularity * s->max_iov; 289 290 need_cow = !test_bit(*offset / s->granularity, s->cow_bitmap); 291 need_cow |= !test_bit((*offset + *bytes - 1) / s->granularity, 292 s->cow_bitmap); 293 if (need_cow) { 294 bdrv_round_to_subclusters(blk_bs(s->target), *offset, *bytes, 295 &align_offset, &align_bytes); 296 } 297 298 if (align_bytes > max_bytes) { 299 align_bytes = max_bytes; 300 if (need_cow) { 301 align_bytes = QEMU_ALIGN_DOWN(align_bytes, s->target_cluster_size); 302 } 303 } 304 /* Clipping may result in align_bytes unaligned to chunk boundary, but 305 * that doesn't matter because it's already the end of source image. */ 306 align_bytes = mirror_clip_bytes(s, align_offset, align_bytes); 307 308 ret = align_offset + align_bytes - (*offset + *bytes); 309 *offset = align_offset; 310 *bytes = align_bytes; 311 assert(ret >= 0); 312 return ret; 313 } 314 315 static inline void coroutine_fn 316 mirror_wait_for_free_in_flight_slot(MirrorBlockJob *s) 317 { 318 MirrorOp *op; 319 320 QTAILQ_FOREACH(op, &s->ops_in_flight, next) { 321 /* 322 * Do not wait on pseudo ops, because it may in turn wait on 323 * some other operation to start, which may in fact be the 324 * caller of this function. Since there is only one pseudo op 325 * at any given time, we will always find some real operation 326 * to wait on. 327 * Also, do not wait on active operations, because they do not 328 * use up in-flight slots. 329 */ 330 if (!op->is_pseudo_op && op->is_in_flight && !op->is_active_write) { 331 qemu_co_queue_wait(&op->waiting_requests, NULL); 332 return; 333 } 334 } 335 abort(); 336 } 337 338 /* Perform a mirror copy operation. 339 * 340 * *op->bytes_handled is set to the number of bytes copied after and 341 * including offset, excluding any bytes copied prior to offset due 342 * to alignment. This will be op->bytes if no alignment is necessary, 343 * or (new_end - op->offset) if the tail is rounded up or down due to 344 * alignment or buffer limit. 345 */ 346 static void coroutine_fn mirror_co_read(void *opaque) 347 { 348 MirrorOp *op = opaque; 349 MirrorBlockJob *s = op->s; 350 int nb_chunks; 351 uint64_t ret; 352 uint64_t max_bytes; 353 354 max_bytes = s->granularity * s->max_iov; 355 356 /* We can only handle as much as buf_size at a time. */ 357 op->bytes = MIN(s->buf_size, MIN(max_bytes, op->bytes)); 358 assert(op->bytes); 359 assert(op->bytes < BDRV_REQUEST_MAX_BYTES); 360 *op->bytes_handled = op->bytes; 361 362 if (s->cow_bitmap) { 363 *op->bytes_handled += mirror_cow_align(s, &op->offset, &op->bytes); 364 } 365 /* Cannot exceed BDRV_REQUEST_MAX_BYTES + INT_MAX */ 366 assert(*op->bytes_handled <= UINT_MAX); 367 assert(op->bytes <= s->buf_size); 368 /* The offset is granularity-aligned because: 369 * 1) Caller passes in aligned values; 370 * 2) mirror_cow_align is used only when target cluster is larger. */ 371 assert(QEMU_IS_ALIGNED(op->offset, s->granularity)); 372 /* The range is sector-aligned, since bdrv_getlength() rounds up. */ 373 assert(QEMU_IS_ALIGNED(op->bytes, BDRV_SECTOR_SIZE)); 374 nb_chunks = DIV_ROUND_UP(op->bytes, s->granularity); 375 376 while (s->buf_free_count < nb_chunks) { 377 trace_mirror_yield_in_flight(s, op->offset, s->in_flight); 378 mirror_wait_for_free_in_flight_slot(s); 379 } 380 381 /* Now make a QEMUIOVector taking enough granularity-sized chunks 382 * from s->buf_free. 383 */ 384 qemu_iovec_init(&op->qiov, nb_chunks); 385 while (nb_chunks-- > 0) { 386 MirrorBuffer *buf = QSIMPLEQ_FIRST(&s->buf_free); 387 size_t remaining = op->bytes - op->qiov.size; 388 389 QSIMPLEQ_REMOVE_HEAD(&s->buf_free, next); 390 s->buf_free_count--; 391 qemu_iovec_add(&op->qiov, buf, MIN(s->granularity, remaining)); 392 } 393 394 /* Copy the dirty cluster. */ 395 s->in_flight++; 396 s->bytes_in_flight += op->bytes; 397 op->is_in_flight = true; 398 trace_mirror_one_iteration(s, op->offset, op->bytes); 399 400 WITH_GRAPH_RDLOCK_GUARD() { 401 ret = bdrv_co_preadv(s->mirror_top_bs->backing, op->offset, op->bytes, 402 &op->qiov, 0); 403 } 404 mirror_read_complete(op, ret); 405 } 406 407 static void coroutine_fn mirror_co_zero(void *opaque) 408 { 409 MirrorOp *op = opaque; 410 int ret; 411 412 op->s->in_flight++; 413 op->s->bytes_in_flight += op->bytes; 414 *op->bytes_handled = op->bytes; 415 op->is_in_flight = true; 416 417 ret = blk_co_pwrite_zeroes(op->s->target, op->offset, op->bytes, 418 op->s->unmap ? BDRV_REQ_MAY_UNMAP : 0); 419 mirror_write_complete(op, ret); 420 } 421 422 static void coroutine_fn mirror_co_discard(void *opaque) 423 { 424 MirrorOp *op = opaque; 425 int ret; 426 427 op->s->in_flight++; 428 op->s->bytes_in_flight += op->bytes; 429 *op->bytes_handled = op->bytes; 430 op->is_in_flight = true; 431 432 ret = blk_co_pdiscard(op->s->target, op->offset, op->bytes); 433 mirror_write_complete(op, ret); 434 } 435 436 static unsigned mirror_perform(MirrorBlockJob *s, int64_t offset, 437 unsigned bytes, MirrorMethod mirror_method) 438 { 439 MirrorOp *op; 440 Coroutine *co; 441 int64_t bytes_handled = -1; 442 443 op = g_new(MirrorOp, 1); 444 *op = (MirrorOp){ 445 .s = s, 446 .offset = offset, 447 .bytes = bytes, 448 .bytes_handled = &bytes_handled, 449 }; 450 qemu_co_queue_init(&op->waiting_requests); 451 452 switch (mirror_method) { 453 case MIRROR_METHOD_COPY: 454 co = qemu_coroutine_create(mirror_co_read, op); 455 break; 456 case MIRROR_METHOD_ZERO: 457 co = qemu_coroutine_create(mirror_co_zero, op); 458 break; 459 case MIRROR_METHOD_DISCARD: 460 co = qemu_coroutine_create(mirror_co_discard, op); 461 break; 462 default: 463 abort(); 464 } 465 op->co = co; 466 467 QTAILQ_INSERT_TAIL(&s->ops_in_flight, op, next); 468 qemu_coroutine_enter(co); 469 /* At this point, ownership of op has been moved to the coroutine 470 * and the object may already be freed */ 471 472 /* Assert that this value has been set */ 473 assert(bytes_handled >= 0); 474 475 /* Same assertion as in mirror_co_read() (and for mirror_co_read() 476 * and mirror_co_discard(), bytes_handled == op->bytes, which 477 * is the @bytes parameter given to this function) */ 478 assert(bytes_handled <= UINT_MAX); 479 return bytes_handled; 480 } 481 482 static void coroutine_fn GRAPH_UNLOCKED mirror_iteration(MirrorBlockJob *s) 483 { 484 BlockDriverState *source; 485 MirrorOp *pseudo_op; 486 int64_t offset; 487 /* At least the first dirty chunk is mirrored in one iteration. */ 488 int nb_chunks = 1; 489 bool write_zeroes_ok = bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)); 490 int max_io_bytes = MAX(s->buf_size / MAX_IN_FLIGHT, MAX_IO_BYTES); 491 492 bdrv_graph_co_rdlock(); 493 source = s->mirror_top_bs->backing->bs; 494 bdrv_graph_co_rdunlock(); 495 496 bdrv_dirty_bitmap_lock(s->dirty_bitmap); 497 offset = bdrv_dirty_iter_next(s->dbi); 498 if (offset < 0) { 499 bdrv_set_dirty_iter(s->dbi, 0); 500 offset = bdrv_dirty_iter_next(s->dbi); 501 trace_mirror_restart_iter(s, bdrv_get_dirty_count(s->dirty_bitmap)); 502 assert(offset >= 0); 503 } 504 bdrv_dirty_bitmap_unlock(s->dirty_bitmap); 505 506 /* 507 * Wait for concurrent requests to @offset. The next loop will limit the 508 * copied area based on in_flight_bitmap so we only copy an area that does 509 * not overlap with concurrent in-flight requests. Still, we would like to 510 * copy something, so wait until there are at least no more requests to the 511 * very beginning of the area. 512 */ 513 mirror_wait_on_conflicts(NULL, s, offset, 1); 514 515 job_pause_point(&s->common.job); 516 517 /* Find the number of consecutive dirty chunks following the first dirty 518 * one, and wait for in flight requests in them. */ 519 bdrv_dirty_bitmap_lock(s->dirty_bitmap); 520 while (nb_chunks * s->granularity < s->buf_size) { 521 int64_t next_dirty; 522 int64_t next_offset = offset + nb_chunks * s->granularity; 523 int64_t next_chunk = next_offset / s->granularity; 524 if (next_offset >= s->bdev_length || 525 !bdrv_dirty_bitmap_get_locked(s->dirty_bitmap, next_offset)) { 526 break; 527 } 528 if (test_bit(next_chunk, s->in_flight_bitmap)) { 529 break; 530 } 531 532 next_dirty = bdrv_dirty_iter_next(s->dbi); 533 if (next_dirty > next_offset || next_dirty < 0) { 534 /* The bitmap iterator's cache is stale, refresh it */ 535 bdrv_set_dirty_iter(s->dbi, next_offset); 536 next_dirty = bdrv_dirty_iter_next(s->dbi); 537 } 538 assert(next_dirty == next_offset); 539 nb_chunks++; 540 } 541 542 /* Clear dirty bits before querying the block status, because 543 * calling bdrv_block_status_above could yield - if some blocks are 544 * marked dirty in this window, we need to know. 545 */ 546 bdrv_reset_dirty_bitmap_locked(s->dirty_bitmap, offset, 547 nb_chunks * s->granularity); 548 bdrv_dirty_bitmap_unlock(s->dirty_bitmap); 549 550 /* Before claiming an area in the in-flight bitmap, we have to 551 * create a MirrorOp for it so that conflicting requests can wait 552 * for it. mirror_perform() will create the real MirrorOps later, 553 * for now we just create a pseudo operation that will wake up all 554 * conflicting requests once all real operations have been 555 * launched. */ 556 pseudo_op = g_new(MirrorOp, 1); 557 *pseudo_op = (MirrorOp){ 558 .offset = offset, 559 .bytes = nb_chunks * s->granularity, 560 .is_pseudo_op = true, 561 }; 562 qemu_co_queue_init(&pseudo_op->waiting_requests); 563 QTAILQ_INSERT_TAIL(&s->ops_in_flight, pseudo_op, next); 564 565 bitmap_set(s->in_flight_bitmap, offset / s->granularity, nb_chunks); 566 while (nb_chunks > 0 && offset < s->bdev_length) { 567 int ret; 568 int64_t io_bytes; 569 int64_t io_bytes_acct; 570 MirrorMethod mirror_method = MIRROR_METHOD_COPY; 571 572 assert(!(offset % s->granularity)); 573 WITH_GRAPH_RDLOCK_GUARD() { 574 ret = bdrv_co_block_status_above(source, NULL, offset, 575 nb_chunks * s->granularity, 576 &io_bytes, NULL, NULL); 577 } 578 if (ret < 0) { 579 io_bytes = MIN(nb_chunks * s->granularity, max_io_bytes); 580 } else if (ret & BDRV_BLOCK_DATA) { 581 io_bytes = MIN(io_bytes, max_io_bytes); 582 } 583 584 io_bytes -= io_bytes % s->granularity; 585 if (io_bytes < s->granularity) { 586 io_bytes = s->granularity; 587 } else if (ret >= 0 && !(ret & BDRV_BLOCK_DATA)) { 588 int64_t target_offset; 589 int64_t target_bytes; 590 WITH_GRAPH_RDLOCK_GUARD() { 591 bdrv_round_to_subclusters(blk_bs(s->target), offset, io_bytes, 592 &target_offset, &target_bytes); 593 } 594 if (target_offset == offset && 595 target_bytes == io_bytes) { 596 mirror_method = ret & BDRV_BLOCK_ZERO ? 597 MIRROR_METHOD_ZERO : 598 MIRROR_METHOD_DISCARD; 599 } 600 } 601 602 while (s->in_flight >= MAX_IN_FLIGHT) { 603 trace_mirror_yield_in_flight(s, offset, s->in_flight); 604 mirror_wait_for_free_in_flight_slot(s); 605 } 606 607 if (s->ret < 0) { 608 ret = 0; 609 goto fail; 610 } 611 612 io_bytes = mirror_clip_bytes(s, offset, io_bytes); 613 io_bytes = mirror_perform(s, offset, io_bytes, mirror_method); 614 if (mirror_method != MIRROR_METHOD_COPY && write_zeroes_ok) { 615 io_bytes_acct = 0; 616 } else { 617 io_bytes_acct = io_bytes; 618 } 619 assert(io_bytes); 620 offset += io_bytes; 621 nb_chunks -= DIV_ROUND_UP(io_bytes, s->granularity); 622 block_job_ratelimit_processed_bytes(&s->common, io_bytes_acct); 623 } 624 625 fail: 626 QTAILQ_REMOVE(&s->ops_in_flight, pseudo_op, next); 627 qemu_co_queue_restart_all(&pseudo_op->waiting_requests); 628 g_free(pseudo_op); 629 } 630 631 static void mirror_free_init(MirrorBlockJob *s) 632 { 633 int granularity = s->granularity; 634 size_t buf_size = s->buf_size; 635 uint8_t *buf = s->buf; 636 637 assert(s->buf_free_count == 0); 638 QSIMPLEQ_INIT(&s->buf_free); 639 while (buf_size != 0) { 640 MirrorBuffer *cur = (MirrorBuffer *)buf; 641 QSIMPLEQ_INSERT_TAIL(&s->buf_free, cur, next); 642 s->buf_free_count++; 643 buf_size -= granularity; 644 buf += granularity; 645 } 646 } 647 648 /* This is also used for the .pause callback. There is no matching 649 * mirror_resume() because mirror_run() will begin iterating again 650 * when the job is resumed. 651 */ 652 static void coroutine_fn mirror_wait_for_all_io(MirrorBlockJob *s) 653 { 654 while (s->in_flight > 0) { 655 mirror_wait_for_free_in_flight_slot(s); 656 } 657 } 658 659 /** 660 * mirror_exit_common: handle both abort() and prepare() cases. 661 * for .prepare, returns 0 on success and -errno on failure. 662 * for .abort cases, denoted by abort = true, MUST return 0. 663 */ 664 static int mirror_exit_common(Job *job) 665 { 666 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 667 BlockJob *bjob = &s->common; 668 MirrorBDSOpaque *bs_opaque; 669 BlockDriverState *src; 670 BlockDriverState *target_bs; 671 BlockDriverState *mirror_top_bs; 672 Error *local_err = NULL; 673 bool abort = job->ret < 0; 674 int ret = 0; 675 676 GLOBAL_STATE_CODE(); 677 678 if (s->prepared) { 679 return 0; 680 } 681 s->prepared = true; 682 683 bdrv_graph_rdlock_main_loop(); 684 685 mirror_top_bs = s->mirror_top_bs; 686 bs_opaque = mirror_top_bs->opaque; 687 src = mirror_top_bs->backing->bs; 688 target_bs = blk_bs(s->target); 689 690 if (bdrv_chain_contains(src, target_bs)) { 691 bdrv_unfreeze_backing_chain(mirror_top_bs, target_bs); 692 } 693 694 bdrv_release_dirty_bitmap(s->dirty_bitmap); 695 696 /* Make sure that the source BDS doesn't go away during bdrv_replace_node, 697 * before we can call bdrv_drained_end */ 698 bdrv_ref(src); 699 bdrv_ref(mirror_top_bs); 700 bdrv_ref(target_bs); 701 702 bdrv_graph_rdunlock_main_loop(); 703 704 /* 705 * Remove target parent that still uses BLK_PERM_WRITE/RESIZE before 706 * inserting target_bs at s->to_replace, where we might not be able to get 707 * these permissions. 708 */ 709 blk_unref(s->target); 710 s->target = NULL; 711 712 /* We don't access the source any more. Dropping any WRITE/RESIZE is 713 * required before it could become a backing file of target_bs. Not having 714 * these permissions any more means that we can't allow any new requests on 715 * mirror_top_bs from now on, so keep it drained. */ 716 bdrv_drained_begin(mirror_top_bs); 717 bdrv_drained_begin(target_bs); 718 bs_opaque->stop = true; 719 720 bdrv_graph_rdlock_main_loop(); 721 bdrv_child_refresh_perms(mirror_top_bs, mirror_top_bs->backing, 722 &error_abort); 723 724 if (!abort && s->backing_mode == MIRROR_SOURCE_BACKING_CHAIN) { 725 BlockDriverState *backing = s->is_none_mode ? src : s->base; 726 BlockDriverState *unfiltered_target = bdrv_skip_filters(target_bs); 727 728 if (bdrv_cow_bs(unfiltered_target) != backing) { 729 bdrv_set_backing_hd(unfiltered_target, backing, &local_err); 730 if (local_err) { 731 error_report_err(local_err); 732 local_err = NULL; 733 ret = -EPERM; 734 } 735 } 736 } else if (!abort && s->backing_mode == MIRROR_OPEN_BACKING_CHAIN) { 737 assert(!bdrv_backing_chain_next(target_bs)); 738 ret = bdrv_open_backing_file(bdrv_skip_filters(target_bs), NULL, 739 "backing", &local_err); 740 if (ret < 0) { 741 error_report_err(local_err); 742 local_err = NULL; 743 } 744 } 745 bdrv_graph_rdunlock_main_loop(); 746 747 if (s->should_complete && !abort) { 748 BlockDriverState *to_replace = s->to_replace ?: src; 749 bool ro = bdrv_is_read_only(to_replace); 750 751 if (ro != bdrv_is_read_only(target_bs)) { 752 bdrv_reopen_set_read_only(target_bs, ro, NULL); 753 } 754 755 /* The mirror job has no requests in flight any more, but we need to 756 * drain potential other users of the BDS before changing the graph. */ 757 assert(s->in_drain); 758 bdrv_drained_begin(to_replace); 759 /* 760 * Cannot use check_to_replace_node() here, because that would 761 * check for an op blocker on @to_replace, and we have our own 762 * there. 763 */ 764 bdrv_graph_wrlock(); 765 if (bdrv_recurse_can_replace(src, to_replace)) { 766 bdrv_replace_node(to_replace, target_bs, &local_err); 767 } else { 768 error_setg(&local_err, "Can no longer replace '%s' by '%s', " 769 "because it can no longer be guaranteed that doing so " 770 "would not lead to an abrupt change of visible data", 771 to_replace->node_name, target_bs->node_name); 772 } 773 bdrv_graph_wrunlock(); 774 bdrv_drained_end(to_replace); 775 if (local_err) { 776 error_report_err(local_err); 777 ret = -EPERM; 778 } 779 } 780 if (s->to_replace) { 781 bdrv_op_unblock_all(s->to_replace, s->replace_blocker); 782 error_free(s->replace_blocker); 783 bdrv_unref(s->to_replace); 784 } 785 g_free(s->replaces); 786 787 /* 788 * Remove the mirror filter driver from the graph. Before this, get rid of 789 * the blockers on the intermediate nodes so that the resulting state is 790 * valid. 791 */ 792 block_job_remove_all_bdrv(bjob); 793 bdrv_graph_wrlock(); 794 bdrv_replace_node(mirror_top_bs, mirror_top_bs->backing->bs, &error_abort); 795 bdrv_graph_wrunlock(); 796 797 bdrv_drained_end(target_bs); 798 bdrv_unref(target_bs); 799 800 bs_opaque->job = NULL; 801 802 bdrv_drained_end(src); 803 bdrv_drained_end(mirror_top_bs); 804 s->in_drain = false; 805 bdrv_unref(mirror_top_bs); 806 bdrv_unref(src); 807 808 return ret; 809 } 810 811 static int mirror_prepare(Job *job) 812 { 813 return mirror_exit_common(job); 814 } 815 816 static void mirror_abort(Job *job) 817 { 818 int ret = mirror_exit_common(job); 819 assert(ret == 0); 820 } 821 822 static void coroutine_fn mirror_throttle(MirrorBlockJob *s) 823 { 824 int64_t now = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); 825 826 if (now - s->last_pause_ns > BLOCK_JOB_SLICE_TIME) { 827 s->last_pause_ns = now; 828 job_sleep_ns(&s->common.job, 0); 829 } else { 830 job_pause_point(&s->common.job); 831 } 832 } 833 834 static int coroutine_fn GRAPH_UNLOCKED mirror_dirty_init(MirrorBlockJob *s) 835 { 836 int64_t offset; 837 BlockDriverState *bs; 838 BlockDriverState *target_bs = blk_bs(s->target); 839 int ret; 840 int64_t count; 841 842 bdrv_graph_co_rdlock(); 843 bs = s->mirror_top_bs->backing->bs; 844 bdrv_graph_co_rdunlock(); 845 846 if (s->zero_target) { 847 if (!bdrv_can_write_zeroes_with_unmap(target_bs)) { 848 bdrv_set_dirty_bitmap(s->dirty_bitmap, 0, s->bdev_length); 849 return 0; 850 } 851 852 s->initial_zeroing_ongoing = true; 853 for (offset = 0; offset < s->bdev_length; ) { 854 int bytes = MIN(s->bdev_length - offset, 855 QEMU_ALIGN_DOWN(INT_MAX, s->granularity)); 856 857 mirror_throttle(s); 858 859 if (job_is_cancelled(&s->common.job)) { 860 s->initial_zeroing_ongoing = false; 861 return 0; 862 } 863 864 if (s->in_flight >= MAX_IN_FLIGHT) { 865 trace_mirror_yield(s, UINT64_MAX, s->buf_free_count, 866 s->in_flight); 867 mirror_wait_for_free_in_flight_slot(s); 868 continue; 869 } 870 871 mirror_perform(s, offset, bytes, MIRROR_METHOD_ZERO); 872 offset += bytes; 873 } 874 875 mirror_wait_for_all_io(s); 876 s->initial_zeroing_ongoing = false; 877 } 878 879 /* First part, loop on the sectors and initialize the dirty bitmap. */ 880 for (offset = 0; offset < s->bdev_length; ) { 881 /* Just to make sure we are not exceeding int limit. */ 882 int bytes = MIN(s->bdev_length - offset, 883 QEMU_ALIGN_DOWN(INT_MAX, s->granularity)); 884 885 mirror_throttle(s); 886 887 if (job_is_cancelled(&s->common.job)) { 888 return 0; 889 } 890 891 WITH_GRAPH_RDLOCK_GUARD() { 892 ret = bdrv_co_is_allocated_above(bs, s->base_overlay, true, offset, 893 bytes, &count); 894 } 895 if (ret < 0) { 896 return ret; 897 } 898 899 assert(count); 900 if (ret > 0) { 901 bdrv_set_dirty_bitmap(s->dirty_bitmap, offset, count); 902 } 903 offset += count; 904 } 905 return 0; 906 } 907 908 /* Called when going out of the streaming phase to flush the bulk of the 909 * data to the medium, or just before completing. 910 */ 911 static int coroutine_fn mirror_flush(MirrorBlockJob *s) 912 { 913 int ret = blk_co_flush(s->target); 914 if (ret < 0) { 915 if (mirror_error_action(s, false, -ret) == BLOCK_ERROR_ACTION_REPORT) { 916 s->ret = ret; 917 } 918 } 919 return ret; 920 } 921 922 static int coroutine_fn mirror_run(Job *job, Error **errp) 923 { 924 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 925 BlockDriverState *bs; 926 MirrorBDSOpaque *mirror_top_opaque = s->mirror_top_bs->opaque; 927 BlockDriverState *target_bs = blk_bs(s->target); 928 bool need_drain = true; 929 BlockDeviceIoStatus iostatus; 930 int64_t length; 931 int64_t target_length; 932 BlockDriverInfo bdi; 933 char backing_filename[2]; /* we only need 2 characters because we are only 934 checking for a NULL string */ 935 int ret = 0; 936 937 bdrv_graph_co_rdlock(); 938 bs = bdrv_filter_bs(s->mirror_top_bs); 939 bdrv_graph_co_rdunlock(); 940 941 if (job_is_cancelled(&s->common.job)) { 942 goto immediate_exit; 943 } 944 945 bdrv_graph_co_rdlock(); 946 s->bdev_length = bdrv_co_getlength(bs); 947 bdrv_graph_co_rdunlock(); 948 949 if (s->bdev_length < 0) { 950 ret = s->bdev_length; 951 goto immediate_exit; 952 } 953 954 target_length = blk_co_getlength(s->target); 955 if (target_length < 0) { 956 ret = target_length; 957 goto immediate_exit; 958 } 959 960 /* Active commit must resize the base image if its size differs from the 961 * active layer. */ 962 if (s->base == blk_bs(s->target)) { 963 if (s->bdev_length > target_length) { 964 ret = blk_co_truncate(s->target, s->bdev_length, false, 965 PREALLOC_MODE_OFF, 0, NULL); 966 if (ret < 0) { 967 goto immediate_exit; 968 } 969 } 970 } else if (s->bdev_length != target_length) { 971 error_setg(errp, "Source and target image have different sizes"); 972 ret = -EINVAL; 973 goto immediate_exit; 974 } 975 976 if (s->bdev_length == 0) { 977 /* Transition to the READY state and wait for complete. */ 978 job_transition_to_ready(&s->common.job); 979 qatomic_set(&s->actively_synced, true); 980 while (!job_cancel_requested(&s->common.job) && !s->should_complete) { 981 job_yield(&s->common.job); 982 } 983 goto immediate_exit; 984 } 985 986 length = DIV_ROUND_UP(s->bdev_length, s->granularity); 987 s->in_flight_bitmap = bitmap_new(length); 988 989 /* If we have no backing file yet in the destination, we cannot let 990 * the destination do COW. Instead, we copy sectors around the 991 * dirty data if needed. We need a bitmap to do that. 992 */ 993 bdrv_get_backing_filename(target_bs, backing_filename, 994 sizeof(backing_filename)); 995 bdrv_graph_co_rdlock(); 996 if (!bdrv_co_get_info(target_bs, &bdi) && bdi.cluster_size) { 997 s->target_cluster_size = bdi.cluster_size; 998 } else { 999 s->target_cluster_size = BDRV_SECTOR_SIZE; 1000 } 1001 if (backing_filename[0] && !bdrv_backing_chain_next(target_bs) && 1002 s->granularity < s->target_cluster_size) { 1003 s->buf_size = MAX(s->buf_size, s->target_cluster_size); 1004 s->cow_bitmap = bitmap_new(length); 1005 } 1006 s->max_iov = MIN(bs->bl.max_iov, target_bs->bl.max_iov); 1007 bdrv_graph_co_rdunlock(); 1008 1009 s->buf = qemu_try_blockalign(bs, s->buf_size); 1010 if (s->buf == NULL) { 1011 ret = -ENOMEM; 1012 goto immediate_exit; 1013 } 1014 1015 mirror_free_init(s); 1016 1017 s->last_pause_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); 1018 if (!s->is_none_mode) { 1019 ret = mirror_dirty_init(s); 1020 if (ret < 0 || job_is_cancelled(&s->common.job)) { 1021 goto immediate_exit; 1022 } 1023 } 1024 1025 /* 1026 * Only now the job is fully initialised and mirror_top_bs should start 1027 * accessing it. 1028 */ 1029 mirror_top_opaque->job = s; 1030 1031 assert(!s->dbi); 1032 s->dbi = bdrv_dirty_iter_new(s->dirty_bitmap); 1033 for (;;) { 1034 int64_t cnt, delta; 1035 bool should_complete; 1036 1037 if (s->ret < 0) { 1038 ret = s->ret; 1039 goto immediate_exit; 1040 } 1041 1042 job_pause_point(&s->common.job); 1043 1044 if (job_is_cancelled(&s->common.job)) { 1045 ret = 0; 1046 goto immediate_exit; 1047 } 1048 1049 cnt = bdrv_get_dirty_count(s->dirty_bitmap); 1050 /* cnt is the number of dirty bytes remaining and s->bytes_in_flight is 1051 * the number of bytes currently being processed; together those are 1052 * the current remaining operation length */ 1053 job_progress_set_remaining(&s->common.job, 1054 s->bytes_in_flight + cnt + 1055 s->active_write_bytes_in_flight); 1056 1057 /* Note that even when no rate limit is applied we need to yield 1058 * periodically with no pending I/O so that bdrv_drain_all() returns. 1059 * We do so every BLKOCK_JOB_SLICE_TIME nanoseconds, or when there is 1060 * an error, or when the source is clean, whichever comes first. */ 1061 delta = qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - s->last_pause_ns; 1062 WITH_JOB_LOCK_GUARD() { 1063 iostatus = s->common.iostatus; 1064 } 1065 if (delta < BLOCK_JOB_SLICE_TIME && 1066 iostatus == BLOCK_DEVICE_IO_STATUS_OK) { 1067 if (s->in_flight >= MAX_IN_FLIGHT || s->buf_free_count == 0 || 1068 (cnt == 0 && s->in_flight > 0)) { 1069 trace_mirror_yield(s, cnt, s->buf_free_count, s->in_flight); 1070 mirror_wait_for_free_in_flight_slot(s); 1071 continue; 1072 } else if (cnt != 0) { 1073 mirror_iteration(s); 1074 } 1075 } 1076 1077 should_complete = false; 1078 if (s->in_flight == 0 && cnt == 0) { 1079 trace_mirror_before_flush(s); 1080 if (!job_is_ready(&s->common.job)) { 1081 if (mirror_flush(s) < 0) { 1082 /* Go check s->ret. */ 1083 continue; 1084 } 1085 /* We're out of the streaming phase. From now on, if the job 1086 * is cancelled we will actually complete all pending I/O and 1087 * report completion. This way, block-job-cancel will leave 1088 * the target in a consistent state. 1089 */ 1090 job_transition_to_ready(&s->common.job); 1091 } 1092 if (qatomic_read(&s->copy_mode) != MIRROR_COPY_MODE_BACKGROUND) { 1093 qatomic_set(&s->actively_synced, true); 1094 } 1095 1096 should_complete = s->should_complete || 1097 job_cancel_requested(&s->common.job); 1098 cnt = bdrv_get_dirty_count(s->dirty_bitmap); 1099 } 1100 1101 if (cnt == 0 && should_complete) { 1102 /* The dirty bitmap is not updated while operations are pending. 1103 * If we're about to exit, wait for pending operations before 1104 * calling bdrv_get_dirty_count(bs), or we may exit while the 1105 * source has dirty data to copy! 1106 * 1107 * Note that I/O can be submitted by the guest while 1108 * mirror_populate runs, so pause it now. Before deciding 1109 * whether to switch to target check one last time if I/O has 1110 * come in the meanwhile, and if not flush the data to disk. 1111 */ 1112 trace_mirror_before_drain(s, cnt); 1113 1114 s->in_drain = true; 1115 bdrv_drained_begin(bs); 1116 1117 /* Must be zero because we are drained */ 1118 assert(s->in_active_write_counter == 0); 1119 1120 cnt = bdrv_get_dirty_count(s->dirty_bitmap); 1121 if (cnt > 0 || mirror_flush(s) < 0) { 1122 bdrv_drained_end(bs); 1123 s->in_drain = false; 1124 continue; 1125 } 1126 1127 /* The two disks are in sync. Exit and report successful 1128 * completion. 1129 */ 1130 assert(QLIST_EMPTY(&bs->tracked_requests)); 1131 need_drain = false; 1132 break; 1133 } 1134 1135 if (job_is_ready(&s->common.job) && !should_complete) { 1136 if (s->in_flight == 0 && cnt == 0) { 1137 trace_mirror_before_sleep(s, cnt, job_is_ready(&s->common.job), 1138 BLOCK_JOB_SLICE_TIME); 1139 job_sleep_ns(&s->common.job, BLOCK_JOB_SLICE_TIME); 1140 } 1141 } else { 1142 block_job_ratelimit_sleep(&s->common); 1143 } 1144 s->last_pause_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); 1145 } 1146 1147 immediate_exit: 1148 if (s->in_flight > 0) { 1149 /* We get here only if something went wrong. Either the job failed, 1150 * or it was cancelled prematurely so that we do not guarantee that 1151 * the target is a copy of the source. 1152 */ 1153 assert(ret < 0 || job_is_cancelled(&s->common.job)); 1154 assert(need_drain); 1155 mirror_wait_for_all_io(s); 1156 } 1157 1158 assert(s->in_flight == 0); 1159 qemu_vfree(s->buf); 1160 g_free(s->cow_bitmap); 1161 g_free(s->in_flight_bitmap); 1162 bdrv_dirty_iter_free(s->dbi); 1163 1164 if (need_drain) { 1165 s->in_drain = true; 1166 bdrv_drained_begin(bs); 1167 } 1168 1169 return ret; 1170 } 1171 1172 static void mirror_complete(Job *job, Error **errp) 1173 { 1174 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 1175 1176 if (!job_is_ready(job)) { 1177 error_setg(errp, "The active block job '%s' cannot be completed", 1178 job->id); 1179 return; 1180 } 1181 1182 /* block all operations on to_replace bs */ 1183 if (s->replaces) { 1184 s->to_replace = bdrv_find_node(s->replaces); 1185 if (!s->to_replace) { 1186 error_setg(errp, "Node name '%s' not found", s->replaces); 1187 return; 1188 } 1189 1190 /* TODO Translate this into child freeze system. */ 1191 error_setg(&s->replace_blocker, 1192 "block device is in use by block-job-complete"); 1193 bdrv_op_block_all(s->to_replace, s->replace_blocker); 1194 bdrv_ref(s->to_replace); 1195 } 1196 1197 s->should_complete = true; 1198 1199 /* If the job is paused, it will be re-entered when it is resumed */ 1200 WITH_JOB_LOCK_GUARD() { 1201 if (!job->paused) { 1202 job_enter_cond_locked(job, NULL); 1203 } 1204 } 1205 } 1206 1207 static void coroutine_fn mirror_pause(Job *job) 1208 { 1209 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 1210 1211 mirror_wait_for_all_io(s); 1212 } 1213 1214 static bool mirror_drained_poll(BlockJob *job) 1215 { 1216 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common); 1217 1218 /* If the job isn't paused nor cancelled, we can't be sure that it won't 1219 * issue more requests. We make an exception if we've reached this point 1220 * from one of our own drain sections, to avoid a deadlock waiting for 1221 * ourselves. 1222 */ 1223 WITH_JOB_LOCK_GUARD() { 1224 if (!s->common.job.paused && !job_is_cancelled_locked(&job->job) 1225 && !s->in_drain) { 1226 return true; 1227 } 1228 } 1229 1230 return !!s->in_flight; 1231 } 1232 1233 static bool mirror_cancel(Job *job, bool force) 1234 { 1235 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common.job); 1236 BlockDriverState *target = blk_bs(s->target); 1237 1238 /* 1239 * Before the job is READY, we treat any cancellation like a 1240 * force-cancellation. 1241 */ 1242 force = force || !job_is_ready(job); 1243 1244 if (force) { 1245 bdrv_cancel_in_flight(target); 1246 } 1247 return force; 1248 } 1249 1250 static bool commit_active_cancel(Job *job, bool force) 1251 { 1252 /* Same as above in mirror_cancel() */ 1253 return force || !job_is_ready(job); 1254 } 1255 1256 static void mirror_change(BlockJob *job, BlockJobChangeOptions *opts, 1257 Error **errp) 1258 { 1259 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common); 1260 BlockJobChangeOptionsMirror *change_opts = &opts->u.mirror; 1261 MirrorCopyMode current; 1262 1263 /* 1264 * The implementation relies on the fact that copy_mode is only written 1265 * under the BQL. Otherwise, further synchronization would be required. 1266 */ 1267 1268 GLOBAL_STATE_CODE(); 1269 1270 if (qatomic_read(&s->copy_mode) == change_opts->copy_mode) { 1271 return; 1272 } 1273 1274 if (change_opts->copy_mode != MIRROR_COPY_MODE_WRITE_BLOCKING) { 1275 error_setg(errp, "Change to copy mode '%s' is not implemented", 1276 MirrorCopyMode_str(change_opts->copy_mode)); 1277 return; 1278 } 1279 1280 current = qatomic_cmpxchg(&s->copy_mode, MIRROR_COPY_MODE_BACKGROUND, 1281 change_opts->copy_mode); 1282 if (current != MIRROR_COPY_MODE_BACKGROUND) { 1283 error_setg(errp, "Expected current copy mode '%s', got '%s'", 1284 MirrorCopyMode_str(MIRROR_COPY_MODE_BACKGROUND), 1285 MirrorCopyMode_str(current)); 1286 } 1287 } 1288 1289 static void mirror_query(BlockJob *job, BlockJobInfo *info) 1290 { 1291 MirrorBlockJob *s = container_of(job, MirrorBlockJob, common); 1292 1293 info->u.mirror = (BlockJobInfoMirror) { 1294 .actively_synced = qatomic_read(&s->actively_synced), 1295 }; 1296 } 1297 1298 static const BlockJobDriver mirror_job_driver = { 1299 .job_driver = { 1300 .instance_size = sizeof(MirrorBlockJob), 1301 .job_type = JOB_TYPE_MIRROR, 1302 .free = block_job_free, 1303 .user_resume = block_job_user_resume, 1304 .run = mirror_run, 1305 .prepare = mirror_prepare, 1306 .abort = mirror_abort, 1307 .pause = mirror_pause, 1308 .complete = mirror_complete, 1309 .cancel = mirror_cancel, 1310 }, 1311 .drained_poll = mirror_drained_poll, 1312 .change = mirror_change, 1313 .query = mirror_query, 1314 }; 1315 1316 static const BlockJobDriver commit_active_job_driver = { 1317 .job_driver = { 1318 .instance_size = sizeof(MirrorBlockJob), 1319 .job_type = JOB_TYPE_COMMIT, 1320 .free = block_job_free, 1321 .user_resume = block_job_user_resume, 1322 .run = mirror_run, 1323 .prepare = mirror_prepare, 1324 .abort = mirror_abort, 1325 .pause = mirror_pause, 1326 .complete = mirror_complete, 1327 .cancel = commit_active_cancel, 1328 }, 1329 .drained_poll = mirror_drained_poll, 1330 }; 1331 1332 static void coroutine_fn 1333 do_sync_target_write(MirrorBlockJob *job, MirrorMethod method, 1334 uint64_t offset, uint64_t bytes, 1335 QEMUIOVector *qiov, int flags) 1336 { 1337 int ret; 1338 size_t qiov_offset = 0; 1339 int64_t bitmap_offset, bitmap_end; 1340 1341 if (!QEMU_IS_ALIGNED(offset, job->granularity) && 1342 bdrv_dirty_bitmap_get(job->dirty_bitmap, offset)) 1343 { 1344 /* 1345 * Dirty unaligned padding: ignore it. 1346 * 1347 * Reasoning: 1348 * 1. If we copy it, we can't reset corresponding bit in 1349 * dirty_bitmap as there may be some "dirty" bytes still not 1350 * copied. 1351 * 2. It's already dirty, so skipping it we don't diverge mirror 1352 * progress. 1353 * 1354 * Note, that because of this, guest write may have no contribution 1355 * into mirror converge, but that's not bad, as we have background 1356 * process of mirroring. If under some bad circumstances (high guest 1357 * IO load) background process starve, we will not converge anyway, 1358 * even if each write will contribute, as guest is not guaranteed to 1359 * rewrite the whole disk. 1360 */ 1361 qiov_offset = QEMU_ALIGN_UP(offset, job->granularity) - offset; 1362 if (bytes <= qiov_offset) { 1363 /* nothing to do after shrink */ 1364 return; 1365 } 1366 offset += qiov_offset; 1367 bytes -= qiov_offset; 1368 } 1369 1370 if (!QEMU_IS_ALIGNED(offset + bytes, job->granularity) && 1371 bdrv_dirty_bitmap_get(job->dirty_bitmap, offset + bytes - 1)) 1372 { 1373 uint64_t tail = (offset + bytes) % job->granularity; 1374 1375 if (bytes <= tail) { 1376 /* nothing to do after shrink */ 1377 return; 1378 } 1379 bytes -= tail; 1380 } 1381 1382 /* 1383 * Tails are either clean or shrunk, so for bitmap resetting 1384 * we safely align the range down. 1385 */ 1386 bitmap_offset = QEMU_ALIGN_UP(offset, job->granularity); 1387 bitmap_end = QEMU_ALIGN_DOWN(offset + bytes, job->granularity); 1388 if (bitmap_offset < bitmap_end) { 1389 bdrv_reset_dirty_bitmap(job->dirty_bitmap, bitmap_offset, 1390 bitmap_end - bitmap_offset); 1391 } 1392 1393 job_progress_increase_remaining(&job->common.job, bytes); 1394 job->active_write_bytes_in_flight += bytes; 1395 1396 switch (method) { 1397 case MIRROR_METHOD_COPY: 1398 ret = blk_co_pwritev_part(job->target, offset, bytes, 1399 qiov, qiov_offset, flags); 1400 break; 1401 1402 case MIRROR_METHOD_ZERO: 1403 assert(!qiov); 1404 ret = blk_co_pwrite_zeroes(job->target, offset, bytes, flags); 1405 break; 1406 1407 case MIRROR_METHOD_DISCARD: 1408 assert(!qiov); 1409 ret = blk_co_pdiscard(job->target, offset, bytes); 1410 break; 1411 1412 default: 1413 abort(); 1414 } 1415 1416 job->active_write_bytes_in_flight -= bytes; 1417 if (ret >= 0) { 1418 job_progress_update(&job->common.job, bytes); 1419 } else { 1420 BlockErrorAction action; 1421 1422 /* 1423 * We failed, so we should mark dirty the whole area, aligned up. 1424 * Note that we don't care about shrunk tails if any: they were dirty 1425 * at function start, and they must be still dirty, as we've locked 1426 * the region for in-flight op. 1427 */ 1428 bitmap_offset = QEMU_ALIGN_DOWN(offset, job->granularity); 1429 bitmap_end = QEMU_ALIGN_UP(offset + bytes, job->granularity); 1430 bdrv_set_dirty_bitmap(job->dirty_bitmap, bitmap_offset, 1431 bitmap_end - bitmap_offset); 1432 qatomic_set(&job->actively_synced, false); 1433 1434 action = mirror_error_action(job, false, -ret); 1435 if (action == BLOCK_ERROR_ACTION_REPORT) { 1436 if (!job->ret) { 1437 job->ret = ret; 1438 } 1439 } 1440 } 1441 } 1442 1443 static MirrorOp *coroutine_fn active_write_prepare(MirrorBlockJob *s, 1444 uint64_t offset, 1445 uint64_t bytes) 1446 { 1447 MirrorOp *op; 1448 uint64_t start_chunk = offset / s->granularity; 1449 uint64_t end_chunk = DIV_ROUND_UP(offset + bytes, s->granularity); 1450 1451 op = g_new(MirrorOp, 1); 1452 *op = (MirrorOp){ 1453 .s = s, 1454 .offset = offset, 1455 .bytes = bytes, 1456 .is_active_write = true, 1457 .is_in_flight = true, 1458 .co = qemu_coroutine_self(), 1459 }; 1460 qemu_co_queue_init(&op->waiting_requests); 1461 QTAILQ_INSERT_TAIL(&s->ops_in_flight, op, next); 1462 1463 s->in_active_write_counter++; 1464 1465 /* 1466 * Wait for concurrent requests affecting the area. If there are already 1467 * running requests that are copying off now-to-be stale data in the area, 1468 * we must wait for them to finish before we begin writing fresh data to the 1469 * target so that the write operations appear in the correct order. 1470 * Note that background requests (see mirror_iteration()) in contrast only 1471 * wait for conflicting requests at the start of the dirty area, and then 1472 * (based on the in_flight_bitmap) truncate the area to copy so it will not 1473 * conflict with any requests beyond that. For active writes, however, we 1474 * cannot truncate that area. The request from our parent must be blocked 1475 * until the area is copied in full. Therefore, we must wait for the whole 1476 * area to become free of concurrent requests. 1477 */ 1478 mirror_wait_on_conflicts(op, s, offset, bytes); 1479 1480 bitmap_set(s->in_flight_bitmap, start_chunk, end_chunk - start_chunk); 1481 1482 return op; 1483 } 1484 1485 static void coroutine_fn GRAPH_RDLOCK active_write_settle(MirrorOp *op) 1486 { 1487 uint64_t start_chunk = op->offset / op->s->granularity; 1488 uint64_t end_chunk = DIV_ROUND_UP(op->offset + op->bytes, 1489 op->s->granularity); 1490 1491 if (!--op->s->in_active_write_counter && 1492 qatomic_read(&op->s->actively_synced)) { 1493 BdrvChild *source = op->s->mirror_top_bs->backing; 1494 1495 if (QLIST_FIRST(&source->bs->parents) == source && 1496 QLIST_NEXT(source, next_parent) == NULL) 1497 { 1498 /* Assert that we are back in sync once all active write 1499 * operations are settled. 1500 * Note that we can only assert this if the mirror node 1501 * is the source node's only parent. */ 1502 assert(!bdrv_get_dirty_count(op->s->dirty_bitmap)); 1503 } 1504 } 1505 bitmap_clear(op->s->in_flight_bitmap, start_chunk, end_chunk - start_chunk); 1506 QTAILQ_REMOVE(&op->s->ops_in_flight, op, next); 1507 qemu_co_queue_restart_all(&op->waiting_requests); 1508 g_free(op); 1509 } 1510 1511 static int coroutine_fn GRAPH_RDLOCK 1512 bdrv_mirror_top_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes, 1513 QEMUIOVector *qiov, BdrvRequestFlags flags) 1514 { 1515 return bdrv_co_preadv(bs->backing, offset, bytes, qiov, flags); 1516 } 1517 1518 static bool should_copy_to_target(MirrorBDSOpaque *s) 1519 { 1520 return s->job && s->job->ret >= 0 && 1521 !job_is_cancelled(&s->job->common.job) && 1522 qatomic_read(&s->job->copy_mode) == MIRROR_COPY_MODE_WRITE_BLOCKING; 1523 } 1524 1525 static int coroutine_fn GRAPH_RDLOCK 1526 bdrv_mirror_top_do_write(BlockDriverState *bs, MirrorMethod method, 1527 bool copy_to_target, uint64_t offset, uint64_t bytes, 1528 QEMUIOVector *qiov, int flags) 1529 { 1530 MirrorOp *op = NULL; 1531 MirrorBDSOpaque *s = bs->opaque; 1532 int ret = 0; 1533 1534 if (copy_to_target) { 1535 op = active_write_prepare(s->job, offset, bytes); 1536 } 1537 1538 switch (method) { 1539 case MIRROR_METHOD_COPY: 1540 ret = bdrv_co_pwritev(bs->backing, offset, bytes, qiov, flags); 1541 break; 1542 1543 case MIRROR_METHOD_ZERO: 1544 ret = bdrv_co_pwrite_zeroes(bs->backing, offset, bytes, flags); 1545 break; 1546 1547 case MIRROR_METHOD_DISCARD: 1548 ret = bdrv_co_pdiscard(bs->backing, offset, bytes); 1549 break; 1550 1551 default: 1552 abort(); 1553 } 1554 1555 if (!copy_to_target && s->job && s->job->dirty_bitmap) { 1556 qatomic_set(&s->job->actively_synced, false); 1557 bdrv_set_dirty_bitmap(s->job->dirty_bitmap, offset, bytes); 1558 } 1559 1560 if (ret < 0) { 1561 goto out; 1562 } 1563 1564 if (copy_to_target) { 1565 do_sync_target_write(s->job, method, offset, bytes, qiov, flags); 1566 } 1567 1568 out: 1569 if (copy_to_target) { 1570 active_write_settle(op); 1571 } 1572 return ret; 1573 } 1574 1575 static int coroutine_fn GRAPH_RDLOCK 1576 bdrv_mirror_top_pwritev(BlockDriverState *bs, int64_t offset, int64_t bytes, 1577 QEMUIOVector *qiov, BdrvRequestFlags flags) 1578 { 1579 QEMUIOVector bounce_qiov; 1580 void *bounce_buf; 1581 int ret = 0; 1582 bool copy_to_target = should_copy_to_target(bs->opaque); 1583 1584 if (copy_to_target) { 1585 /* The guest might concurrently modify the data to write; but 1586 * the data on source and destination must match, so we have 1587 * to use a bounce buffer if we are going to write to the 1588 * target now. */ 1589 bounce_buf = qemu_blockalign(bs, bytes); 1590 iov_to_buf_full(qiov->iov, qiov->niov, 0, bounce_buf, bytes); 1591 1592 qemu_iovec_init(&bounce_qiov, 1); 1593 qemu_iovec_add(&bounce_qiov, bounce_buf, bytes); 1594 qiov = &bounce_qiov; 1595 1596 flags &= ~BDRV_REQ_REGISTERED_BUF; 1597 } 1598 1599 ret = bdrv_mirror_top_do_write(bs, MIRROR_METHOD_COPY, copy_to_target, 1600 offset, bytes, qiov, flags); 1601 1602 if (copy_to_target) { 1603 qemu_iovec_destroy(&bounce_qiov); 1604 qemu_vfree(bounce_buf); 1605 } 1606 1607 return ret; 1608 } 1609 1610 static int coroutine_fn GRAPH_RDLOCK bdrv_mirror_top_flush(BlockDriverState *bs) 1611 { 1612 if (bs->backing == NULL) { 1613 /* we can be here after failed bdrv_append in mirror_start_job */ 1614 return 0; 1615 } 1616 return bdrv_co_flush(bs->backing->bs); 1617 } 1618 1619 static int coroutine_fn GRAPH_RDLOCK 1620 bdrv_mirror_top_pwrite_zeroes(BlockDriverState *bs, int64_t offset, 1621 int64_t bytes, BdrvRequestFlags flags) 1622 { 1623 bool copy_to_target = should_copy_to_target(bs->opaque); 1624 return bdrv_mirror_top_do_write(bs, MIRROR_METHOD_ZERO, copy_to_target, 1625 offset, bytes, NULL, flags); 1626 } 1627 1628 static int coroutine_fn GRAPH_RDLOCK 1629 bdrv_mirror_top_pdiscard(BlockDriverState *bs, int64_t offset, int64_t bytes) 1630 { 1631 bool copy_to_target = should_copy_to_target(bs->opaque); 1632 return bdrv_mirror_top_do_write(bs, MIRROR_METHOD_DISCARD, copy_to_target, 1633 offset, bytes, NULL, 0); 1634 } 1635 1636 static void GRAPH_RDLOCK bdrv_mirror_top_refresh_filename(BlockDriverState *bs) 1637 { 1638 if (bs->backing == NULL) { 1639 /* we can be here after failed bdrv_attach_child in 1640 * bdrv_set_backing_hd */ 1641 return; 1642 } 1643 pstrcpy(bs->exact_filename, sizeof(bs->exact_filename), 1644 bs->backing->bs->filename); 1645 } 1646 1647 static void bdrv_mirror_top_child_perm(BlockDriverState *bs, BdrvChild *c, 1648 BdrvChildRole role, 1649 BlockReopenQueue *reopen_queue, 1650 uint64_t perm, uint64_t shared, 1651 uint64_t *nperm, uint64_t *nshared) 1652 { 1653 MirrorBDSOpaque *s = bs->opaque; 1654 1655 if (s->stop) { 1656 /* 1657 * If the job is to be stopped, we do not need to forward 1658 * anything to the real image. 1659 */ 1660 *nperm = 0; 1661 *nshared = BLK_PERM_ALL; 1662 return; 1663 } 1664 1665 bdrv_default_perms(bs, c, role, reopen_queue, 1666 perm, shared, nperm, nshared); 1667 1668 if (s->is_commit) { 1669 /* 1670 * For commit jobs, we cannot take CONSISTENT_READ, because 1671 * that permission is unshared for everything above the base 1672 * node (except for filters on the base node). 1673 * We also have to force-share the WRITE permission, or 1674 * otherwise we would block ourselves at the base node (if 1675 * writes are blocked for a node, they are also blocked for 1676 * its backing file). 1677 * (We could also share RESIZE, because it may be needed for 1678 * the target if its size is less than the top node's; but 1679 * bdrv_default_perms_for_cow() automatically shares RESIZE 1680 * for backing nodes if WRITE is shared, so there is no need 1681 * to do it here.) 1682 */ 1683 *nperm &= ~BLK_PERM_CONSISTENT_READ; 1684 *nshared |= BLK_PERM_WRITE; 1685 } 1686 } 1687 1688 /* Dummy node that provides consistent read to its users without requiring it 1689 * from its backing file and that allows writes on the backing file chain. */ 1690 static BlockDriver bdrv_mirror_top = { 1691 .format_name = "mirror_top", 1692 .bdrv_co_preadv = bdrv_mirror_top_preadv, 1693 .bdrv_co_pwritev = bdrv_mirror_top_pwritev, 1694 .bdrv_co_pwrite_zeroes = bdrv_mirror_top_pwrite_zeroes, 1695 .bdrv_co_pdiscard = bdrv_mirror_top_pdiscard, 1696 .bdrv_co_flush = bdrv_mirror_top_flush, 1697 .bdrv_refresh_filename = bdrv_mirror_top_refresh_filename, 1698 .bdrv_child_perm = bdrv_mirror_top_child_perm, 1699 1700 .is_filter = true, 1701 .filtered_child_is_backing = true, 1702 }; 1703 1704 static BlockJob *mirror_start_job( 1705 const char *job_id, BlockDriverState *bs, 1706 int creation_flags, BlockDriverState *target, 1707 const char *replaces, int64_t speed, 1708 uint32_t granularity, int64_t buf_size, 1709 BlockMirrorBackingMode backing_mode, 1710 bool zero_target, 1711 BlockdevOnError on_source_error, 1712 BlockdevOnError on_target_error, 1713 bool unmap, 1714 BlockCompletionFunc *cb, 1715 void *opaque, 1716 const BlockJobDriver *driver, 1717 bool is_none_mode, BlockDriverState *base, 1718 bool auto_complete, const char *filter_node_name, 1719 bool is_mirror, MirrorCopyMode copy_mode, 1720 Error **errp) 1721 { 1722 MirrorBlockJob *s; 1723 MirrorBDSOpaque *bs_opaque; 1724 BlockDriverState *mirror_top_bs; 1725 bool target_is_backing; 1726 uint64_t target_perms, target_shared_perms; 1727 int ret; 1728 1729 GLOBAL_STATE_CODE(); 1730 1731 if (granularity == 0) { 1732 granularity = bdrv_get_default_bitmap_granularity(target); 1733 } 1734 1735 assert(is_power_of_2(granularity)); 1736 1737 if (buf_size < 0) { 1738 error_setg(errp, "Invalid parameter 'buf-size'"); 1739 return NULL; 1740 } 1741 1742 if (buf_size == 0) { 1743 buf_size = DEFAULT_MIRROR_BUF_SIZE; 1744 } 1745 1746 bdrv_graph_rdlock_main_loop(); 1747 if (bdrv_skip_filters(bs) == bdrv_skip_filters(target)) { 1748 error_setg(errp, "Can't mirror node into itself"); 1749 bdrv_graph_rdunlock_main_loop(); 1750 return NULL; 1751 } 1752 1753 target_is_backing = bdrv_chain_contains(bs, target); 1754 bdrv_graph_rdunlock_main_loop(); 1755 1756 /* In the case of active commit, add dummy driver to provide consistent 1757 * reads on the top, while disabling it in the intermediate nodes, and make 1758 * the backing chain writable. */ 1759 mirror_top_bs = bdrv_new_open_driver(&bdrv_mirror_top, filter_node_name, 1760 BDRV_O_RDWR, errp); 1761 if (mirror_top_bs == NULL) { 1762 return NULL; 1763 } 1764 if (!filter_node_name) { 1765 mirror_top_bs->implicit = true; 1766 } 1767 1768 /* So that we can always drop this node */ 1769 mirror_top_bs->never_freeze = true; 1770 1771 mirror_top_bs->total_sectors = bs->total_sectors; 1772 mirror_top_bs->supported_write_flags = BDRV_REQ_WRITE_UNCHANGED; 1773 mirror_top_bs->supported_zero_flags = BDRV_REQ_WRITE_UNCHANGED | 1774 BDRV_REQ_NO_FALLBACK; 1775 bs_opaque = g_new0(MirrorBDSOpaque, 1); 1776 mirror_top_bs->opaque = bs_opaque; 1777 1778 bs_opaque->is_commit = target_is_backing; 1779 1780 bdrv_drained_begin(bs); 1781 ret = bdrv_append(mirror_top_bs, bs, errp); 1782 bdrv_drained_end(bs); 1783 1784 if (ret < 0) { 1785 bdrv_unref(mirror_top_bs); 1786 return NULL; 1787 } 1788 1789 /* Make sure that the source is not resized while the job is running */ 1790 s = block_job_create(job_id, driver, NULL, mirror_top_bs, 1791 BLK_PERM_CONSISTENT_READ, 1792 BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED | 1793 BLK_PERM_WRITE, speed, 1794 creation_flags, cb, opaque, errp); 1795 if (!s) { 1796 goto fail; 1797 } 1798 1799 /* The block job now has a reference to this node */ 1800 bdrv_unref(mirror_top_bs); 1801 1802 s->mirror_top_bs = mirror_top_bs; 1803 1804 /* No resize for the target either; while the mirror is still running, a 1805 * consistent read isn't necessarily possible. We could possibly allow 1806 * writes and graph modifications, though it would likely defeat the 1807 * purpose of a mirror, so leave them blocked for now. 1808 * 1809 * In the case of active commit, things look a bit different, though, 1810 * because the target is an already populated backing file in active use. 1811 * We can allow anything except resize there.*/ 1812 1813 target_perms = BLK_PERM_WRITE; 1814 target_shared_perms = BLK_PERM_WRITE_UNCHANGED; 1815 1816 if (target_is_backing) { 1817 int64_t bs_size, target_size; 1818 bs_size = bdrv_getlength(bs); 1819 if (bs_size < 0) { 1820 error_setg_errno(errp, -bs_size, 1821 "Could not inquire top image size"); 1822 goto fail; 1823 } 1824 1825 target_size = bdrv_getlength(target); 1826 if (target_size < 0) { 1827 error_setg_errno(errp, -target_size, 1828 "Could not inquire base image size"); 1829 goto fail; 1830 } 1831 1832 if (target_size < bs_size) { 1833 target_perms |= BLK_PERM_RESIZE; 1834 } 1835 1836 target_shared_perms |= BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE; 1837 } else { 1838 bdrv_graph_rdlock_main_loop(); 1839 if (bdrv_chain_contains(bs, bdrv_skip_filters(target))) { 1840 /* 1841 * We may want to allow this in the future, but it would 1842 * require taking some extra care. 1843 */ 1844 error_setg(errp, "Cannot mirror to a filter on top of a node in " 1845 "the source's backing chain"); 1846 bdrv_graph_rdunlock_main_loop(); 1847 goto fail; 1848 } 1849 bdrv_graph_rdunlock_main_loop(); 1850 } 1851 1852 s->target = blk_new(s->common.job.aio_context, 1853 target_perms, target_shared_perms); 1854 ret = blk_insert_bs(s->target, target, errp); 1855 if (ret < 0) { 1856 goto fail; 1857 } 1858 if (is_mirror) { 1859 /* XXX: Mirror target could be a NBD server of target QEMU in the case 1860 * of non-shared block migration. To allow migration completion, we 1861 * have to allow "inactivate" of the target BB. When that happens, we 1862 * know the job is drained, and the vcpus are stopped, so no write 1863 * operation will be performed. Block layer already has assertions to 1864 * ensure that. */ 1865 blk_set_force_allow_inactivate(s->target); 1866 } 1867 blk_set_allow_aio_context_change(s->target, true); 1868 blk_set_disable_request_queuing(s->target, true); 1869 1870 bdrv_graph_rdlock_main_loop(); 1871 s->replaces = g_strdup(replaces); 1872 s->on_source_error = on_source_error; 1873 s->on_target_error = on_target_error; 1874 s->is_none_mode = is_none_mode; 1875 s->backing_mode = backing_mode; 1876 s->zero_target = zero_target; 1877 qatomic_set(&s->copy_mode, copy_mode); 1878 s->base = base; 1879 s->base_overlay = bdrv_find_overlay(bs, base); 1880 s->granularity = granularity; 1881 s->buf_size = ROUND_UP(buf_size, granularity); 1882 s->unmap = unmap; 1883 if (auto_complete) { 1884 s->should_complete = true; 1885 } 1886 bdrv_graph_rdunlock_main_loop(); 1887 1888 s->dirty_bitmap = bdrv_create_dirty_bitmap(s->mirror_top_bs, granularity, 1889 NULL, errp); 1890 if (!s->dirty_bitmap) { 1891 goto fail; 1892 } 1893 1894 /* 1895 * The dirty bitmap is set by bdrv_mirror_top_do_write() when not in active 1896 * mode. 1897 */ 1898 bdrv_disable_dirty_bitmap(s->dirty_bitmap); 1899 1900 bdrv_graph_wrlock(); 1901 ret = block_job_add_bdrv(&s->common, "source", bs, 0, 1902 BLK_PERM_WRITE_UNCHANGED | BLK_PERM_WRITE | 1903 BLK_PERM_CONSISTENT_READ, 1904 errp); 1905 if (ret < 0) { 1906 bdrv_graph_wrunlock(); 1907 goto fail; 1908 } 1909 1910 /* Required permissions are already taken with blk_new() */ 1911 block_job_add_bdrv(&s->common, "target", target, 0, BLK_PERM_ALL, 1912 &error_abort); 1913 1914 /* In commit_active_start() all intermediate nodes disappear, so 1915 * any jobs in them must be blocked */ 1916 if (target_is_backing) { 1917 BlockDriverState *iter, *filtered_target; 1918 uint64_t iter_shared_perms; 1919 1920 /* 1921 * The topmost node with 1922 * bdrv_skip_filters(filtered_target) == bdrv_skip_filters(target) 1923 */ 1924 filtered_target = bdrv_cow_bs(bdrv_find_overlay(bs, target)); 1925 1926 assert(bdrv_skip_filters(filtered_target) == 1927 bdrv_skip_filters(target)); 1928 1929 /* 1930 * XXX BLK_PERM_WRITE needs to be allowed so we don't block 1931 * ourselves at s->base (if writes are blocked for a node, they are 1932 * also blocked for its backing file). The other options would be a 1933 * second filter driver above s->base (== target). 1934 */ 1935 iter_shared_perms = BLK_PERM_WRITE_UNCHANGED | BLK_PERM_WRITE; 1936 1937 for (iter = bdrv_filter_or_cow_bs(bs); iter != target; 1938 iter = bdrv_filter_or_cow_bs(iter)) 1939 { 1940 if (iter == filtered_target) { 1941 /* 1942 * From here on, all nodes are filters on the base. 1943 * This allows us to share BLK_PERM_CONSISTENT_READ. 1944 */ 1945 iter_shared_perms |= BLK_PERM_CONSISTENT_READ; 1946 } 1947 1948 ret = block_job_add_bdrv(&s->common, "intermediate node", iter, 0, 1949 iter_shared_perms, errp); 1950 if (ret < 0) { 1951 bdrv_graph_wrunlock(); 1952 goto fail; 1953 } 1954 } 1955 1956 if (bdrv_freeze_backing_chain(mirror_top_bs, target, errp) < 0) { 1957 bdrv_graph_wrunlock(); 1958 goto fail; 1959 } 1960 } 1961 bdrv_graph_wrunlock(); 1962 1963 QTAILQ_INIT(&s->ops_in_flight); 1964 1965 trace_mirror_start(bs, s, opaque); 1966 job_start(&s->common.job); 1967 1968 return &s->common; 1969 1970 fail: 1971 if (s) { 1972 /* Make sure this BDS does not go away until we have completed the graph 1973 * changes below */ 1974 bdrv_ref(mirror_top_bs); 1975 1976 g_free(s->replaces); 1977 blk_unref(s->target); 1978 bs_opaque->job = NULL; 1979 if (s->dirty_bitmap) { 1980 bdrv_release_dirty_bitmap(s->dirty_bitmap); 1981 } 1982 job_early_fail(&s->common.job); 1983 } 1984 1985 bs_opaque->stop = true; 1986 bdrv_drained_begin(bs); 1987 bdrv_graph_wrlock(); 1988 assert(mirror_top_bs->backing->bs == bs); 1989 bdrv_child_refresh_perms(mirror_top_bs, mirror_top_bs->backing, 1990 &error_abort); 1991 bdrv_replace_node(mirror_top_bs, bs, &error_abort); 1992 bdrv_graph_wrunlock(); 1993 bdrv_drained_end(bs); 1994 1995 bdrv_unref(mirror_top_bs); 1996 1997 return NULL; 1998 } 1999 2000 void mirror_start(const char *job_id, BlockDriverState *bs, 2001 BlockDriverState *target, const char *replaces, 2002 int creation_flags, int64_t speed, 2003 uint32_t granularity, int64_t buf_size, 2004 MirrorSyncMode mode, BlockMirrorBackingMode backing_mode, 2005 bool zero_target, 2006 BlockdevOnError on_source_error, 2007 BlockdevOnError on_target_error, 2008 bool unmap, const char *filter_node_name, 2009 MirrorCopyMode copy_mode, Error **errp) 2010 { 2011 bool is_none_mode; 2012 BlockDriverState *base; 2013 2014 GLOBAL_STATE_CODE(); 2015 2016 if ((mode == MIRROR_SYNC_MODE_INCREMENTAL) || 2017 (mode == MIRROR_SYNC_MODE_BITMAP)) { 2018 error_setg(errp, "Sync mode '%s' not supported", 2019 MirrorSyncMode_str(mode)); 2020 return; 2021 } 2022 2023 bdrv_graph_rdlock_main_loop(); 2024 is_none_mode = mode == MIRROR_SYNC_MODE_NONE; 2025 base = mode == MIRROR_SYNC_MODE_TOP ? bdrv_backing_chain_next(bs) : NULL; 2026 bdrv_graph_rdunlock_main_loop(); 2027 2028 mirror_start_job(job_id, bs, creation_flags, target, replaces, 2029 speed, granularity, buf_size, backing_mode, zero_target, 2030 on_source_error, on_target_error, unmap, NULL, NULL, 2031 &mirror_job_driver, is_none_mode, base, false, 2032 filter_node_name, true, copy_mode, errp); 2033 } 2034 2035 BlockJob *commit_active_start(const char *job_id, BlockDriverState *bs, 2036 BlockDriverState *base, int creation_flags, 2037 int64_t speed, BlockdevOnError on_error, 2038 const char *filter_node_name, 2039 BlockCompletionFunc *cb, void *opaque, 2040 bool auto_complete, Error **errp) 2041 { 2042 bool base_read_only; 2043 BlockJob *job; 2044 2045 GLOBAL_STATE_CODE(); 2046 2047 base_read_only = bdrv_is_read_only(base); 2048 2049 if (base_read_only) { 2050 if (bdrv_reopen_set_read_only(base, false, errp) < 0) { 2051 return NULL; 2052 } 2053 } 2054 2055 job = mirror_start_job( 2056 job_id, bs, creation_flags, base, NULL, speed, 0, 0, 2057 MIRROR_LEAVE_BACKING_CHAIN, false, 2058 on_error, on_error, true, cb, opaque, 2059 &commit_active_job_driver, false, base, auto_complete, 2060 filter_node_name, false, MIRROR_COPY_MODE_BACKGROUND, 2061 errp); 2062 if (!job) { 2063 goto error_restore_flags; 2064 } 2065 2066 return job; 2067 2068 error_restore_flags: 2069 /* ignore error and errp for bdrv_reopen, because we want to propagate 2070 * the original error */ 2071 if (base_read_only) { 2072 bdrv_reopen_set_read_only(base, true, NULL); 2073 } 2074 return NULL; 2075 } 2076