xref: /qemu/block/block-copy.c (revision 8b7b9c5c)
1 /*
2  * block_copy API
3  *
4  * Copyright (C) 2013 Proxmox Server Solutions
5  * Copyright (c) 2019 Virtuozzo International GmbH.
6  *
7  * Authors:
8  *  Dietmar Maurer (dietmar@proxmox.com)
9  *  Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
10  *
11  * This work is licensed under the terms of the GNU GPL, version 2 or later.
12  * See the COPYING file in the top-level directory.
13  */
14 
15 #include "qemu/osdep.h"
16 
17 #include "trace.h"
18 #include "qapi/error.h"
19 #include "block/block-copy.h"
20 #include "block/block_int-io.h"
21 #include "block/dirty-bitmap.h"
22 #include "block/reqlist.h"
23 #include "sysemu/block-backend.h"
24 #include "qemu/units.h"
25 #include "qemu/co-shared-resource.h"
26 #include "qemu/coroutine.h"
27 #include "qemu/ratelimit.h"
28 #include "block/aio_task.h"
29 #include "qemu/error-report.h"
30 #include "qemu/memalign.h"
31 
32 #define BLOCK_COPY_MAX_COPY_RANGE (16 * MiB)
33 #define BLOCK_COPY_MAX_BUFFER (1 * MiB)
34 #define BLOCK_COPY_MAX_MEM (128 * MiB)
35 #define BLOCK_COPY_MAX_WORKERS 64
36 #define BLOCK_COPY_SLICE_TIME 100000000ULL /* ns */
37 #define BLOCK_COPY_CLUSTER_SIZE_DEFAULT (1 << 16)
38 
39 typedef enum {
40     COPY_READ_WRITE_CLUSTER,
41     COPY_READ_WRITE,
42     COPY_WRITE_ZEROES,
43     COPY_RANGE_SMALL,
44     COPY_RANGE_FULL
45 } BlockCopyMethod;
46 
47 static coroutine_fn int block_copy_task_entry(AioTask *task);
48 
49 typedef struct BlockCopyCallState {
50     /* Fields initialized in block_copy_async() and never changed. */
51     BlockCopyState *s;
52     int64_t offset;
53     int64_t bytes;
54     int max_workers;
55     int64_t max_chunk;
56     bool ignore_ratelimit;
57     BlockCopyAsyncCallbackFunc cb;
58     void *cb_opaque;
59     /* Coroutine where async block-copy is running */
60     Coroutine *co;
61 
62     /* Fields whose state changes throughout the execution */
63     bool finished; /* atomic */
64     QemuCoSleep sleep; /* TODO: protect API with a lock */
65     bool cancelled; /* atomic */
66     /* To reference all call states from BlockCopyState */
67     QLIST_ENTRY(BlockCopyCallState) list;
68 
69     /*
70      * Fields that report information about return values and errors.
71      * Protected by lock in BlockCopyState.
72      */
73     bool error_is_read;
74     /*
75      * @ret is set concurrently by tasks under mutex. Only set once by first
76      * failed task (and untouched if no task failed).
77      * After finishing (call_state->finished is true), it is not modified
78      * anymore and may be safely read without mutex.
79      */
80     int ret;
81 } BlockCopyCallState;
82 
83 typedef struct BlockCopyTask {
84     AioTask task;
85 
86     /*
87      * Fields initialized in block_copy_task_create()
88      * and never changed.
89      */
90     BlockCopyState *s;
91     BlockCopyCallState *call_state;
92     /*
93      * @method can also be set again in the while loop of
94      * block_copy_dirty_clusters(), but it is never accessed concurrently
95      * because the only other function that reads it is
96      * block_copy_task_entry() and it is invoked afterwards in the same
97      * iteration.
98      */
99     BlockCopyMethod method;
100 
101     /*
102      * Generally, req is protected by lock in BlockCopyState, Still req.offset
103      * is only set on task creation, so may be read concurrently after creation.
104      * req.bytes is changed at most once, and need only protecting the case of
105      * parallel read while updating @bytes value in block_copy_task_shrink().
106      */
107     BlockReq req;
108 } BlockCopyTask;
109 
110 static int64_t task_end(BlockCopyTask *task)
111 {
112     return task->req.offset + task->req.bytes;
113 }
114 
115 typedef struct BlockCopyState {
116     /*
117      * BdrvChild objects are not owned or managed by block-copy. They are
118      * provided by block-copy user and user is responsible for appropriate
119      * permissions on these children.
120      */
121     BdrvChild *source;
122     BdrvChild *target;
123 
124     /*
125      * Fields initialized in block_copy_state_new()
126      * and never changed.
127      */
128     int64_t cluster_size;
129     int64_t max_transfer;
130     uint64_t len;
131     BdrvRequestFlags write_flags;
132 
133     /*
134      * Fields whose state changes throughout the execution
135      * Protected by lock.
136      */
137     CoMutex lock;
138     int64_t in_flight_bytes;
139     BlockCopyMethod method;
140     BlockReqList reqs;
141     QLIST_HEAD(, BlockCopyCallState) calls;
142     /*
143      * skip_unallocated:
144      *
145      * Used by sync=top jobs, which first scan the source node for unallocated
146      * areas and clear them in the copy_bitmap.  During this process, the bitmap
147      * is thus not fully initialized: It may still have bits set for areas that
148      * are unallocated and should actually not be copied.
149      *
150      * This is indicated by skip_unallocated.
151      *
152      * In this case, block_copy() will query the source’s allocation status,
153      * skip unallocated regions, clear them in the copy_bitmap, and invoke
154      * block_copy_reset_unallocated() every time it does.
155      */
156     bool skip_unallocated; /* atomic */
157     /* State fields that use a thread-safe API */
158     BdrvDirtyBitmap *copy_bitmap;
159     ProgressMeter *progress;
160     SharedResource *mem;
161     RateLimit rate_limit;
162 } BlockCopyState;
163 
164 /* Called with lock held */
165 static int64_t block_copy_chunk_size(BlockCopyState *s)
166 {
167     switch (s->method) {
168     case COPY_READ_WRITE_CLUSTER:
169         return s->cluster_size;
170     case COPY_READ_WRITE:
171     case COPY_RANGE_SMALL:
172         return MIN(MAX(s->cluster_size, BLOCK_COPY_MAX_BUFFER),
173                    s->max_transfer);
174     case COPY_RANGE_FULL:
175         return MIN(MAX(s->cluster_size, BLOCK_COPY_MAX_COPY_RANGE),
176                    s->max_transfer);
177     default:
178         /* Cannot have COPY_WRITE_ZEROES here.  */
179         abort();
180     }
181 }
182 
183 /*
184  * Search for the first dirty area in offset/bytes range and create task at
185  * the beginning of it.
186  */
187 static coroutine_fn BlockCopyTask *
188 block_copy_task_create(BlockCopyState *s, BlockCopyCallState *call_state,
189                        int64_t offset, int64_t bytes)
190 {
191     BlockCopyTask *task;
192     int64_t max_chunk;
193 
194     QEMU_LOCK_GUARD(&s->lock);
195     max_chunk = MIN_NON_ZERO(block_copy_chunk_size(s), call_state->max_chunk);
196     if (!bdrv_dirty_bitmap_next_dirty_area(s->copy_bitmap,
197                                            offset, offset + bytes,
198                                            max_chunk, &offset, &bytes))
199     {
200         return NULL;
201     }
202 
203     assert(QEMU_IS_ALIGNED(offset, s->cluster_size));
204     bytes = QEMU_ALIGN_UP(bytes, s->cluster_size);
205 
206     /* region is dirty, so no existent tasks possible in it */
207     assert(!reqlist_find_conflict(&s->reqs, offset, bytes));
208 
209     bdrv_reset_dirty_bitmap(s->copy_bitmap, offset, bytes);
210     s->in_flight_bytes += bytes;
211 
212     task = g_new(BlockCopyTask, 1);
213     *task = (BlockCopyTask) {
214         .task.func = block_copy_task_entry,
215         .s = s,
216         .call_state = call_state,
217         .method = s->method,
218     };
219     reqlist_init_req(&s->reqs, &task->req, offset, bytes);
220 
221     return task;
222 }
223 
224 /*
225  * block_copy_task_shrink
226  *
227  * Drop the tail of the task to be handled later. Set dirty bits back and
228  * wake up all tasks waiting for us (may be some of them are not intersecting
229  * with shrunk task)
230  */
231 static void coroutine_fn block_copy_task_shrink(BlockCopyTask *task,
232                                                 int64_t new_bytes)
233 {
234     QEMU_LOCK_GUARD(&task->s->lock);
235     if (new_bytes == task->req.bytes) {
236         return;
237     }
238 
239     assert(new_bytes > 0 && new_bytes < task->req.bytes);
240 
241     task->s->in_flight_bytes -= task->req.bytes - new_bytes;
242     bdrv_set_dirty_bitmap(task->s->copy_bitmap,
243                           task->req.offset + new_bytes,
244                           task->req.bytes - new_bytes);
245 
246     reqlist_shrink_req(&task->req, new_bytes);
247 }
248 
249 static void coroutine_fn block_copy_task_end(BlockCopyTask *task, int ret)
250 {
251     QEMU_LOCK_GUARD(&task->s->lock);
252     task->s->in_flight_bytes -= task->req.bytes;
253     if (ret < 0) {
254         bdrv_set_dirty_bitmap(task->s->copy_bitmap, task->req.offset,
255                               task->req.bytes);
256     }
257     if (task->s->progress) {
258         progress_set_remaining(task->s->progress,
259                                bdrv_get_dirty_count(task->s->copy_bitmap) +
260                                task->s->in_flight_bytes);
261     }
262     reqlist_remove_req(&task->req);
263 }
264 
265 void block_copy_state_free(BlockCopyState *s)
266 {
267     if (!s) {
268         return;
269     }
270 
271     ratelimit_destroy(&s->rate_limit);
272     bdrv_release_dirty_bitmap(s->copy_bitmap);
273     shres_destroy(s->mem);
274     g_free(s);
275 }
276 
277 static uint32_t block_copy_max_transfer(BdrvChild *source, BdrvChild *target)
278 {
279     return MIN_NON_ZERO(INT_MAX,
280                         MIN_NON_ZERO(source->bs->bl.max_transfer,
281                                      target->bs->bl.max_transfer));
282 }
283 
284 void block_copy_set_copy_opts(BlockCopyState *s, bool use_copy_range,
285                               bool compress)
286 {
287     /* Keep BDRV_REQ_SERIALISING set (or not set) in block_copy_state_new() */
288     s->write_flags = (s->write_flags & BDRV_REQ_SERIALISING) |
289         (compress ? BDRV_REQ_WRITE_COMPRESSED : 0);
290 
291     if (s->max_transfer < s->cluster_size) {
292         /*
293          * copy_range does not respect max_transfer. We don't want to bother
294          * with requests smaller than block-copy cluster size, so fallback to
295          * buffered copying (read and write respect max_transfer on their
296          * behalf).
297          */
298         s->method = COPY_READ_WRITE_CLUSTER;
299     } else if (compress) {
300         /* Compression supports only cluster-size writes and no copy-range. */
301         s->method = COPY_READ_WRITE_CLUSTER;
302     } else {
303         /*
304          * If copy range enabled, start with COPY_RANGE_SMALL, until first
305          * successful copy_range (look at block_copy_do_copy).
306          */
307         s->method = use_copy_range ? COPY_RANGE_SMALL : COPY_READ_WRITE;
308     }
309 }
310 
311 static int64_t block_copy_calculate_cluster_size(BlockDriverState *target,
312                                                  Error **errp)
313 {
314     int ret;
315     BlockDriverInfo bdi;
316     bool target_does_cow = bdrv_backing_chain_next(target);
317 
318     /*
319      * If there is no backing file on the target, we cannot rely on COW if our
320      * backup cluster size is smaller than the target cluster size. Even for
321      * targets with a backing file, try to avoid COW if possible.
322      */
323     ret = bdrv_get_info(target, &bdi);
324     if (ret == -ENOTSUP && !target_does_cow) {
325         /* Cluster size is not defined */
326         warn_report("The target block device doesn't provide "
327                     "information about the block size and it doesn't have a "
328                     "backing file. The default block size of %u bytes is "
329                     "used. If the actual block size of the target exceeds "
330                     "this default, the backup may be unusable",
331                     BLOCK_COPY_CLUSTER_SIZE_DEFAULT);
332         return BLOCK_COPY_CLUSTER_SIZE_DEFAULT;
333     } else if (ret < 0 && !target_does_cow) {
334         error_setg_errno(errp, -ret,
335             "Couldn't determine the cluster size of the target image, "
336             "which has no backing file");
337         error_append_hint(errp,
338             "Aborting, since this may create an unusable destination image\n");
339         return ret;
340     } else if (ret < 0 && target_does_cow) {
341         /* Not fatal; just trudge on ahead. */
342         return BLOCK_COPY_CLUSTER_SIZE_DEFAULT;
343     }
344 
345     return MAX(BLOCK_COPY_CLUSTER_SIZE_DEFAULT, bdi.cluster_size);
346 }
347 
348 BlockCopyState *block_copy_state_new(BdrvChild *source, BdrvChild *target,
349                                      const BdrvDirtyBitmap *bitmap,
350                                      Error **errp)
351 {
352     ERRP_GUARD();
353     BlockCopyState *s;
354     int64_t cluster_size;
355     BdrvDirtyBitmap *copy_bitmap;
356     bool is_fleecing;
357 
358     cluster_size = block_copy_calculate_cluster_size(target->bs, errp);
359     if (cluster_size < 0) {
360         return NULL;
361     }
362 
363     copy_bitmap = bdrv_create_dirty_bitmap(source->bs, cluster_size, NULL,
364                                            errp);
365     if (!copy_bitmap) {
366         return NULL;
367     }
368     bdrv_disable_dirty_bitmap(copy_bitmap);
369     if (bitmap) {
370         if (!bdrv_merge_dirty_bitmap(copy_bitmap, bitmap, NULL, errp)) {
371             error_prepend(errp, "Failed to merge bitmap '%s' to internal "
372                           "copy-bitmap: ", bdrv_dirty_bitmap_name(bitmap));
373             bdrv_release_dirty_bitmap(copy_bitmap);
374             return NULL;
375         }
376     } else {
377         bdrv_set_dirty_bitmap(copy_bitmap, 0,
378                               bdrv_dirty_bitmap_size(copy_bitmap));
379     }
380 
381     /*
382      * If source is in backing chain of target assume that target is going to be
383      * used for "image fleecing", i.e. it should represent a kind of snapshot of
384      * source at backup-start point in time. And target is going to be read by
385      * somebody (for example, used as NBD export) during backup job.
386      *
387      * In this case, we need to add BDRV_REQ_SERIALISING write flag to avoid
388      * intersection of backup writes and third party reads from target,
389      * otherwise reading from target we may occasionally read already updated by
390      * guest data.
391      *
392      * For more information see commit f8d59dfb40bb and test
393      * tests/qemu-iotests/222
394      */
395     is_fleecing = bdrv_chain_contains(target->bs, source->bs);
396 
397     s = g_new(BlockCopyState, 1);
398     *s = (BlockCopyState) {
399         .source = source,
400         .target = target,
401         .copy_bitmap = copy_bitmap,
402         .cluster_size = cluster_size,
403         .len = bdrv_dirty_bitmap_size(copy_bitmap),
404         .write_flags = (is_fleecing ? BDRV_REQ_SERIALISING : 0),
405         .mem = shres_create(BLOCK_COPY_MAX_MEM),
406         .max_transfer = QEMU_ALIGN_DOWN(
407                                     block_copy_max_transfer(source, target),
408                                     cluster_size),
409     };
410 
411     block_copy_set_copy_opts(s, false, false);
412 
413     ratelimit_init(&s->rate_limit);
414     qemu_co_mutex_init(&s->lock);
415     QLIST_INIT(&s->reqs);
416     QLIST_INIT(&s->calls);
417 
418     return s;
419 }
420 
421 /* Only set before running the job, no need for locking. */
422 void block_copy_set_progress_meter(BlockCopyState *s, ProgressMeter *pm)
423 {
424     s->progress = pm;
425 }
426 
427 /*
428  * Takes ownership of @task
429  *
430  * If pool is NULL directly run the task, otherwise schedule it into the pool.
431  *
432  * Returns: task.func return code if pool is NULL
433  *          otherwise -ECANCELED if pool status is bad
434  *          otherwise 0 (successfully scheduled)
435  */
436 static coroutine_fn int block_copy_task_run(AioTaskPool *pool,
437                                             BlockCopyTask *task)
438 {
439     if (!pool) {
440         int ret = task->task.func(&task->task);
441 
442         g_free(task);
443         return ret;
444     }
445 
446     aio_task_pool_wait_slot(pool);
447     if (aio_task_pool_status(pool) < 0) {
448         co_put_to_shres(task->s->mem, task->req.bytes);
449         block_copy_task_end(task, -ECANCELED);
450         g_free(task);
451         return -ECANCELED;
452     }
453 
454     aio_task_pool_start_task(pool, &task->task);
455 
456     return 0;
457 }
458 
459 /*
460  * block_copy_do_copy
461  *
462  * Do copy of cluster-aligned chunk. Requested region is allowed to exceed
463  * s->len only to cover last cluster when s->len is not aligned to clusters.
464  *
465  * No sync here: neither bitmap nor intersecting requests handling, only copy.
466  *
467  * @method is an in-out argument, so that copy_range can be either extended to
468  * a full-size buffer or disabled if the copy_range attempt fails.  The output
469  * value of @method should be used for subsequent tasks.
470  * Returns 0 on success.
471  */
472 static int coroutine_fn GRAPH_RDLOCK
473 block_copy_do_copy(BlockCopyState *s, int64_t offset, int64_t bytes,
474                    BlockCopyMethod *method, bool *error_is_read)
475 {
476     int ret;
477     int64_t nbytes = MIN(offset + bytes, s->len) - offset;
478     void *bounce_buffer = NULL;
479 
480     assert(offset >= 0 && bytes > 0 && INT64_MAX - offset >= bytes);
481     assert(QEMU_IS_ALIGNED(offset, s->cluster_size));
482     assert(QEMU_IS_ALIGNED(bytes, s->cluster_size));
483     assert(offset < s->len);
484     assert(offset + bytes <= s->len ||
485            offset + bytes == QEMU_ALIGN_UP(s->len, s->cluster_size));
486     assert(nbytes < INT_MAX);
487 
488     switch (*method) {
489     case COPY_WRITE_ZEROES:
490         ret = bdrv_co_pwrite_zeroes(s->target, offset, nbytes, s->write_flags &
491                                     ~BDRV_REQ_WRITE_COMPRESSED);
492         if (ret < 0) {
493             trace_block_copy_write_zeroes_fail(s, offset, ret);
494             *error_is_read = false;
495         }
496         return ret;
497 
498     case COPY_RANGE_SMALL:
499     case COPY_RANGE_FULL:
500         ret = bdrv_co_copy_range(s->source, offset, s->target, offset, nbytes,
501                                  0, s->write_flags);
502         if (ret >= 0) {
503             /* Successful copy-range, increase chunk size.  */
504             *method = COPY_RANGE_FULL;
505             return 0;
506         }
507 
508         trace_block_copy_copy_range_fail(s, offset, ret);
509         *method = COPY_READ_WRITE;
510         /* Fall through to read+write with allocated buffer */
511 
512     case COPY_READ_WRITE_CLUSTER:
513     case COPY_READ_WRITE:
514         /*
515          * In case of failed copy_range request above, we may proceed with
516          * buffered request larger than BLOCK_COPY_MAX_BUFFER.
517          * Still, further requests will be properly limited, so don't care too
518          * much. Moreover the most likely case (copy_range is unsupported for
519          * the configuration, so the very first copy_range request fails)
520          * is handled by setting large copy_size only after first successful
521          * copy_range.
522          */
523 
524         bounce_buffer = qemu_blockalign(s->source->bs, nbytes);
525 
526         ret = bdrv_co_pread(s->source, offset, nbytes, bounce_buffer, 0);
527         if (ret < 0) {
528             trace_block_copy_read_fail(s, offset, ret);
529             *error_is_read = true;
530             goto out;
531         }
532 
533         ret = bdrv_co_pwrite(s->target, offset, nbytes, bounce_buffer,
534                              s->write_flags);
535         if (ret < 0) {
536             trace_block_copy_write_fail(s, offset, ret);
537             *error_is_read = false;
538             goto out;
539         }
540 
541     out:
542         qemu_vfree(bounce_buffer);
543         break;
544 
545     default:
546         abort();
547     }
548 
549     return ret;
550 }
551 
552 static coroutine_fn int block_copy_task_entry(AioTask *task)
553 {
554     BlockCopyTask *t = container_of(task, BlockCopyTask, task);
555     BlockCopyState *s = t->s;
556     bool error_is_read = false;
557     BlockCopyMethod method = t->method;
558     int ret;
559 
560     WITH_GRAPH_RDLOCK_GUARD() {
561         ret = block_copy_do_copy(s, t->req.offset, t->req.bytes, &method,
562                                  &error_is_read);
563     }
564 
565     WITH_QEMU_LOCK_GUARD(&s->lock) {
566         if (s->method == t->method) {
567             s->method = method;
568         }
569 
570         if (ret < 0) {
571             if (!t->call_state->ret) {
572                 t->call_state->ret = ret;
573                 t->call_state->error_is_read = error_is_read;
574             }
575         } else if (s->progress) {
576             progress_work_done(s->progress, t->req.bytes);
577         }
578     }
579     co_put_to_shres(s->mem, t->req.bytes);
580     block_copy_task_end(t, ret);
581 
582     return ret;
583 }
584 
585 static coroutine_fn GRAPH_RDLOCK
586 int block_copy_block_status(BlockCopyState *s, int64_t offset, int64_t bytes,
587                             int64_t *pnum)
588 {
589     int64_t num;
590     BlockDriverState *base;
591     int ret;
592 
593     if (qatomic_read(&s->skip_unallocated)) {
594         base = bdrv_backing_chain_next(s->source->bs);
595     } else {
596         base = NULL;
597     }
598 
599     ret = bdrv_co_block_status_above(s->source->bs, base, offset, bytes, &num,
600                                      NULL, NULL);
601     if (ret < 0 || num < s->cluster_size) {
602         /*
603          * On error or if failed to obtain large enough chunk just fallback to
604          * copy one cluster.
605          */
606         num = s->cluster_size;
607         ret = BDRV_BLOCK_ALLOCATED | BDRV_BLOCK_DATA;
608     } else if (offset + num == s->len) {
609         num = QEMU_ALIGN_UP(num, s->cluster_size);
610     } else {
611         num = QEMU_ALIGN_DOWN(num, s->cluster_size);
612     }
613 
614     *pnum = num;
615     return ret;
616 }
617 
618 /*
619  * Check if the cluster starting at offset is allocated or not.
620  * return via pnum the number of contiguous clusters sharing this allocation.
621  */
622 static int coroutine_fn GRAPH_RDLOCK
623 block_copy_is_cluster_allocated(BlockCopyState *s, int64_t offset,
624                                 int64_t *pnum)
625 {
626     BlockDriverState *bs = s->source->bs;
627     int64_t count, total_count = 0;
628     int64_t bytes = s->len - offset;
629     int ret;
630 
631     assert(QEMU_IS_ALIGNED(offset, s->cluster_size));
632 
633     while (true) {
634         /* protected in backup_run() */
635         ret = bdrv_co_is_allocated(bs, offset, bytes, &count);
636         if (ret < 0) {
637             return ret;
638         }
639 
640         total_count += count;
641 
642         if (ret || count == 0) {
643             /*
644              * ret: partial segment(s) are considered allocated.
645              * otherwise: unallocated tail is treated as an entire segment.
646              */
647             *pnum = DIV_ROUND_UP(total_count, s->cluster_size);
648             return ret;
649         }
650 
651         /* Unallocated segment(s) with uncertain following segment(s) */
652         if (total_count >= s->cluster_size) {
653             *pnum = total_count / s->cluster_size;
654             return 0;
655         }
656 
657         offset += count;
658         bytes -= count;
659     }
660 }
661 
662 void block_copy_reset(BlockCopyState *s, int64_t offset, int64_t bytes)
663 {
664     QEMU_LOCK_GUARD(&s->lock);
665 
666     bdrv_reset_dirty_bitmap(s->copy_bitmap, offset, bytes);
667     if (s->progress) {
668         progress_set_remaining(s->progress,
669                                bdrv_get_dirty_count(s->copy_bitmap) +
670                                s->in_flight_bytes);
671     }
672 }
673 
674 /*
675  * Reset bits in copy_bitmap starting at offset if they represent unallocated
676  * data in the image. May reset subsequent contiguous bits.
677  * @return 0 when the cluster at @offset was unallocated,
678  *         1 otherwise, and -ret on error.
679  */
680 int64_t coroutine_fn block_copy_reset_unallocated(BlockCopyState *s,
681                                                   int64_t offset,
682                                                   int64_t *count)
683 {
684     int ret;
685     int64_t clusters, bytes;
686 
687     ret = block_copy_is_cluster_allocated(s, offset, &clusters);
688     if (ret < 0) {
689         return ret;
690     }
691 
692     bytes = clusters * s->cluster_size;
693 
694     if (!ret) {
695         block_copy_reset(s, offset, bytes);
696     }
697 
698     *count = bytes;
699     return ret;
700 }
701 
702 /*
703  * block_copy_dirty_clusters
704  *
705  * Copy dirty clusters in @offset/@bytes range.
706  * Returns 1 if dirty clusters found and successfully copied, 0 if no dirty
707  * clusters found and -errno on failure.
708  */
709 static int coroutine_fn GRAPH_RDLOCK
710 block_copy_dirty_clusters(BlockCopyCallState *call_state)
711 {
712     BlockCopyState *s = call_state->s;
713     int64_t offset = call_state->offset;
714     int64_t bytes = call_state->bytes;
715 
716     int ret = 0;
717     bool found_dirty = false;
718     int64_t end = offset + bytes;
719     AioTaskPool *aio = NULL;
720 
721     /*
722      * block_copy() user is responsible for keeping source and target in same
723      * aio context
724      */
725     assert(bdrv_get_aio_context(s->source->bs) ==
726            bdrv_get_aio_context(s->target->bs));
727 
728     assert(QEMU_IS_ALIGNED(offset, s->cluster_size));
729     assert(QEMU_IS_ALIGNED(bytes, s->cluster_size));
730 
731     while (bytes && aio_task_pool_status(aio) == 0 &&
732            !qatomic_read(&call_state->cancelled)) {
733         BlockCopyTask *task;
734         int64_t status_bytes;
735 
736         task = block_copy_task_create(s, call_state, offset, bytes);
737         if (!task) {
738             /* No more dirty bits in the bitmap */
739             trace_block_copy_skip_range(s, offset, bytes);
740             break;
741         }
742         if (task->req.offset > offset) {
743             trace_block_copy_skip_range(s, offset, task->req.offset - offset);
744         }
745 
746         found_dirty = true;
747 
748         ret = block_copy_block_status(s, task->req.offset, task->req.bytes,
749                                       &status_bytes);
750         assert(ret >= 0); /* never fail */
751         if (status_bytes < task->req.bytes) {
752             block_copy_task_shrink(task, status_bytes);
753         }
754         if (qatomic_read(&s->skip_unallocated) &&
755             !(ret & BDRV_BLOCK_ALLOCATED)) {
756             block_copy_task_end(task, 0);
757             trace_block_copy_skip_range(s, task->req.offset, task->req.bytes);
758             offset = task_end(task);
759             bytes = end - offset;
760             g_free(task);
761             continue;
762         }
763         if (ret & BDRV_BLOCK_ZERO) {
764             task->method = COPY_WRITE_ZEROES;
765         }
766 
767         if (!call_state->ignore_ratelimit) {
768             uint64_t ns = ratelimit_calculate_delay(&s->rate_limit, 0);
769             if (ns > 0) {
770                 block_copy_task_end(task, -EAGAIN);
771                 g_free(task);
772                 qemu_co_sleep_ns_wakeable(&call_state->sleep,
773                                           QEMU_CLOCK_REALTIME, ns);
774                 continue;
775             }
776         }
777 
778         ratelimit_calculate_delay(&s->rate_limit, task->req.bytes);
779 
780         trace_block_copy_process(s, task->req.offset);
781 
782         co_get_from_shres(s->mem, task->req.bytes);
783 
784         offset = task_end(task);
785         bytes = end - offset;
786 
787         if (!aio && bytes) {
788             aio = aio_task_pool_new(call_state->max_workers);
789         }
790 
791         ret = block_copy_task_run(aio, task);
792         if (ret < 0) {
793             goto out;
794         }
795     }
796 
797 out:
798     if (aio) {
799         aio_task_pool_wait_all(aio);
800 
801         /*
802          * We are not really interested in -ECANCELED returned from
803          * block_copy_task_run. If it fails, it means some task already failed
804          * for real reason, let's return first failure.
805          * Still, assert that we don't rewrite failure by success.
806          *
807          * Note: ret may be positive here because of block-status result.
808          */
809         assert(ret >= 0 || aio_task_pool_status(aio) < 0);
810         ret = aio_task_pool_status(aio);
811 
812         aio_task_pool_free(aio);
813     }
814 
815     return ret < 0 ? ret : found_dirty;
816 }
817 
818 void block_copy_kick(BlockCopyCallState *call_state)
819 {
820     qemu_co_sleep_wake(&call_state->sleep);
821 }
822 
823 /*
824  * block_copy_common
825  *
826  * Copy requested region, accordingly to dirty bitmap.
827  * Collaborate with parallel block_copy requests: if they succeed it will help
828  * us. If they fail, we will retry not-copied regions. So, if we return error,
829  * it means that some I/O operation failed in context of _this_ block_copy call,
830  * not some parallel operation.
831  */
832 static int coroutine_fn GRAPH_RDLOCK
833 block_copy_common(BlockCopyCallState *call_state)
834 {
835     int ret;
836     BlockCopyState *s = call_state->s;
837 
838     qemu_co_mutex_lock(&s->lock);
839     QLIST_INSERT_HEAD(&s->calls, call_state, list);
840     qemu_co_mutex_unlock(&s->lock);
841 
842     do {
843         ret = block_copy_dirty_clusters(call_state);
844 
845         if (ret == 0 && !qatomic_read(&call_state->cancelled)) {
846             WITH_QEMU_LOCK_GUARD(&s->lock) {
847                 /*
848                  * Check that there is no task we still need to
849                  * wait to complete
850                  */
851                 ret = reqlist_wait_one(&s->reqs, call_state->offset,
852                                        call_state->bytes, &s->lock);
853                 if (ret == 0) {
854                     /*
855                      * No pending tasks, but check again the bitmap in this
856                      * same critical section, since a task might have failed
857                      * between this and the critical section in
858                      * block_copy_dirty_clusters().
859                      *
860                      * reqlist_wait_one return value 0 also means that it
861                      * didn't release the lock. So, we are still in the same
862                      * critical section, not interrupted by any concurrent
863                      * access to state.
864                      */
865                     ret = bdrv_dirty_bitmap_next_dirty(s->copy_bitmap,
866                                                        call_state->offset,
867                                                        call_state->bytes) >= 0;
868                 }
869             }
870         }
871 
872         /*
873          * We retry in two cases:
874          * 1. Some progress done
875          *    Something was copied, which means that there were yield points
876          *    and some new dirty bits may have appeared (due to failed parallel
877          *    block-copy requests).
878          * 2. We have waited for some intersecting block-copy request
879          *    It may have failed and produced new dirty bits.
880          */
881     } while (ret > 0 && !qatomic_read(&call_state->cancelled));
882 
883     qatomic_store_release(&call_state->finished, true);
884 
885     if (call_state->cb) {
886         call_state->cb(call_state->cb_opaque);
887     }
888 
889     qemu_co_mutex_lock(&s->lock);
890     QLIST_REMOVE(call_state, list);
891     qemu_co_mutex_unlock(&s->lock);
892 
893     return ret;
894 }
895 
896 static void coroutine_fn block_copy_async_co_entry(void *opaque)
897 {
898     GRAPH_RDLOCK_GUARD();
899     block_copy_common(opaque);
900 }
901 
902 int coroutine_fn block_copy(BlockCopyState *s, int64_t start, int64_t bytes,
903                             bool ignore_ratelimit, uint64_t timeout_ns,
904                             BlockCopyAsyncCallbackFunc cb,
905                             void *cb_opaque)
906 {
907     int ret;
908     BlockCopyCallState *call_state = g_new(BlockCopyCallState, 1);
909 
910     *call_state = (BlockCopyCallState) {
911         .s = s,
912         .offset = start,
913         .bytes = bytes,
914         .ignore_ratelimit = ignore_ratelimit,
915         .max_workers = BLOCK_COPY_MAX_WORKERS,
916         .cb = cb,
917         .cb_opaque = cb_opaque,
918     };
919 
920     ret = qemu_co_timeout(block_copy_async_co_entry, call_state, timeout_ns,
921                           g_free);
922     if (ret < 0) {
923         assert(ret == -ETIMEDOUT);
924         block_copy_call_cancel(call_state);
925         /* call_state will be freed by running coroutine. */
926         return ret;
927     }
928 
929     ret = call_state->ret;
930     g_free(call_state);
931 
932     return ret;
933 }
934 
935 BlockCopyCallState *block_copy_async(BlockCopyState *s,
936                                      int64_t offset, int64_t bytes,
937                                      int max_workers, int64_t max_chunk,
938                                      BlockCopyAsyncCallbackFunc cb,
939                                      void *cb_opaque)
940 {
941     BlockCopyCallState *call_state = g_new(BlockCopyCallState, 1);
942 
943     *call_state = (BlockCopyCallState) {
944         .s = s,
945         .offset = offset,
946         .bytes = bytes,
947         .max_workers = max_workers,
948         .max_chunk = max_chunk,
949         .cb = cb,
950         .cb_opaque = cb_opaque,
951 
952         .co = qemu_coroutine_create(block_copy_async_co_entry, call_state),
953     };
954 
955     qemu_coroutine_enter(call_state->co);
956 
957     return call_state;
958 }
959 
960 void block_copy_call_free(BlockCopyCallState *call_state)
961 {
962     if (!call_state) {
963         return;
964     }
965 
966     assert(qatomic_read(&call_state->finished));
967     g_free(call_state);
968 }
969 
970 bool block_copy_call_finished(BlockCopyCallState *call_state)
971 {
972     return qatomic_read(&call_state->finished);
973 }
974 
975 bool block_copy_call_succeeded(BlockCopyCallState *call_state)
976 {
977     return qatomic_load_acquire(&call_state->finished) &&
978            !qatomic_read(&call_state->cancelled) &&
979            call_state->ret == 0;
980 }
981 
982 bool block_copy_call_failed(BlockCopyCallState *call_state)
983 {
984     return qatomic_load_acquire(&call_state->finished) &&
985            !qatomic_read(&call_state->cancelled) &&
986            call_state->ret < 0;
987 }
988 
989 bool block_copy_call_cancelled(BlockCopyCallState *call_state)
990 {
991     return qatomic_read(&call_state->cancelled);
992 }
993 
994 int block_copy_call_status(BlockCopyCallState *call_state, bool *error_is_read)
995 {
996     assert(qatomic_load_acquire(&call_state->finished));
997     if (error_is_read) {
998         *error_is_read = call_state->error_is_read;
999     }
1000     return call_state->ret;
1001 }
1002 
1003 /*
1004  * Note that cancelling and finishing are racy.
1005  * User can cancel a block-copy that is already finished.
1006  */
1007 void block_copy_call_cancel(BlockCopyCallState *call_state)
1008 {
1009     qatomic_set(&call_state->cancelled, true);
1010     block_copy_kick(call_state);
1011 }
1012 
1013 BdrvDirtyBitmap *block_copy_dirty_bitmap(BlockCopyState *s)
1014 {
1015     return s->copy_bitmap;
1016 }
1017 
1018 int64_t block_copy_cluster_size(BlockCopyState *s)
1019 {
1020     return s->cluster_size;
1021 }
1022 
1023 void block_copy_set_skip_unallocated(BlockCopyState *s, bool skip)
1024 {
1025     qatomic_set(&s->skip_unallocated, skip);
1026 }
1027 
1028 void block_copy_set_speed(BlockCopyState *s, uint64_t speed)
1029 {
1030     ratelimit_set_speed(&s->rate_limit, speed, BLOCK_COPY_SLICE_TIME);
1031 
1032     /*
1033      * Note: it's good to kick all call states from here, but it should be done
1034      * only from a coroutine, to not crash if s->calls list changed while
1035      * entering one call. So for now, the only user of this function kicks its
1036      * only one call_state by hand.
1037      */
1038 }
1039