xref: /spdk/lib/blob/blobstore.c (revision 994d4c38ba50ff2f3a8b5a63aef116b26e8ad851)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright (c) Intel Corporation.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Intel Corporation nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include "spdk/stdinc.h"
35 
36 #include "spdk/blob.h"
37 #include "spdk/crc32.h"
38 #include "spdk/env.h"
39 #include "spdk/queue.h"
40 #include "spdk/thread.h"
41 #include "spdk/bit_array.h"
42 #include "spdk/likely.h"
43 #include "spdk/util.h"
44 #include "spdk/string.h"
45 
46 #include "spdk_internal/assert.h"
47 #include "spdk_internal/log.h"
48 
49 #include "blobstore.h"
50 
51 #define BLOB_CRC32C_INITIAL    0xffffffffUL
52 
53 static int spdk_bs_register_md_thread(struct spdk_blob_store *bs);
54 static int spdk_bs_unregister_md_thread(struct spdk_blob_store *bs);
55 static void _spdk_blob_close_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno);
56 static void _spdk_blob_insert_cluster_on_md_thread(struct spdk_blob *blob, uint32_t cluster_num,
57 		uint64_t cluster, spdk_blob_op_complete cb_fn, void *cb_arg);
58 
59 static int _spdk_blob_set_xattr(struct spdk_blob *blob, const char *name, const void *value,
60 				uint16_t value_len, bool internal);
61 static int _spdk_blob_get_xattr_value(struct spdk_blob *blob, const char *name,
62 				      const void **value, size_t *value_len, bool internal);
63 static int _spdk_blob_remove_xattr(struct spdk_blob *blob, const char *name, bool internal);
64 
65 static void
66 _spdk_blob_verify_md_op(struct spdk_blob *blob)
67 {
68 	assert(blob != NULL);
69 	assert(spdk_get_thread() == blob->bs->md_thread);
70 	assert(blob->state != SPDK_BLOB_STATE_LOADING);
71 }
72 
73 static struct spdk_blob_list *
74 _spdk_bs_get_snapshot_entry(struct spdk_blob_store *bs, spdk_blob_id blobid)
75 {
76 	struct spdk_blob_list *snapshot_entry = NULL;
77 
78 	TAILQ_FOREACH(snapshot_entry, &bs->snapshots, link) {
79 		if (snapshot_entry->id == blobid) {
80 			break;
81 		}
82 	}
83 
84 	return snapshot_entry;
85 }
86 
87 static void
88 _spdk_bs_claim_cluster(struct spdk_blob_store *bs, uint32_t cluster_num)
89 {
90 	assert(cluster_num < spdk_bit_array_capacity(bs->used_clusters));
91 	assert(spdk_bit_array_get(bs->used_clusters, cluster_num) == false);
92 	assert(bs->num_free_clusters > 0);
93 
94 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Claiming cluster %u\n", cluster_num);
95 
96 	spdk_bit_array_set(bs->used_clusters, cluster_num);
97 	bs->num_free_clusters--;
98 }
99 
100 static int
101 _spdk_blob_insert_cluster(struct spdk_blob *blob, uint32_t cluster_num, uint64_t cluster)
102 {
103 	uint64_t *cluster_lba = &blob->active.clusters[cluster_num];
104 
105 	_spdk_blob_verify_md_op(blob);
106 
107 	if (*cluster_lba != 0) {
108 		return -EEXIST;
109 	}
110 
111 	*cluster_lba = _spdk_bs_cluster_to_lba(blob->bs, cluster);
112 	return 0;
113 }
114 
115 static int
116 _spdk_bs_allocate_cluster(struct spdk_blob *blob, uint32_t cluster_num,
117 			  uint64_t *lowest_free_cluster, bool update_map)
118 {
119 	pthread_mutex_lock(&blob->bs->used_clusters_mutex);
120 	*lowest_free_cluster = spdk_bit_array_find_first_clear(blob->bs->used_clusters,
121 			       *lowest_free_cluster);
122 	if (*lowest_free_cluster == UINT32_MAX) {
123 		/* No more free clusters. Cannot satisfy the request */
124 		pthread_mutex_unlock(&blob->bs->used_clusters_mutex);
125 		return -ENOSPC;
126 	}
127 
128 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Claiming cluster %lu for blob %lu\n", *lowest_free_cluster, blob->id);
129 	_spdk_bs_claim_cluster(blob->bs, *lowest_free_cluster);
130 	pthread_mutex_unlock(&blob->bs->used_clusters_mutex);
131 
132 	if (update_map) {
133 		_spdk_blob_insert_cluster(blob, cluster_num, *lowest_free_cluster);
134 	}
135 
136 	return 0;
137 }
138 
139 static void
140 _spdk_bs_release_cluster(struct spdk_blob_store *bs, uint32_t cluster_num)
141 {
142 	assert(cluster_num < spdk_bit_array_capacity(bs->used_clusters));
143 	assert(spdk_bit_array_get(bs->used_clusters, cluster_num) == true);
144 	assert(bs->num_free_clusters < bs->total_clusters);
145 
146 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Releasing cluster %u\n", cluster_num);
147 
148 	pthread_mutex_lock(&bs->used_clusters_mutex);
149 	spdk_bit_array_clear(bs->used_clusters, cluster_num);
150 	bs->num_free_clusters++;
151 	pthread_mutex_unlock(&bs->used_clusters_mutex);
152 }
153 
154 static void
155 _spdk_blob_xattrs_init(struct spdk_blob_xattr_opts *xattrs)
156 {
157 	xattrs->count = 0;
158 	xattrs->names = NULL;
159 	xattrs->ctx = NULL;
160 	xattrs->get_value = NULL;
161 }
162 
163 void
164 spdk_blob_opts_init(struct spdk_blob_opts *opts)
165 {
166 	opts->num_clusters = 0;
167 	opts->thin_provision = false;
168 	opts->clear_method = BLOB_CLEAR_WITH_DEFAULT;
169 	_spdk_blob_xattrs_init(&opts->xattrs);
170 }
171 
172 void
173 spdk_blob_open_opts_init(struct spdk_blob_open_opts *opts)
174 {
175 	opts->clear_method = BLOB_CLEAR_WITH_DEFAULT;
176 }
177 
178 static struct spdk_blob *
179 _spdk_blob_alloc(struct spdk_blob_store *bs, spdk_blob_id id)
180 {
181 	struct spdk_blob *blob;
182 
183 	blob = calloc(1, sizeof(*blob));
184 	if (!blob) {
185 		return NULL;
186 	}
187 
188 	blob->id = id;
189 	blob->bs = bs;
190 
191 	blob->parent_id = SPDK_BLOBID_INVALID;
192 
193 	blob->state = SPDK_BLOB_STATE_DIRTY;
194 	blob->active.num_pages = 1;
195 	blob->active.pages = calloc(1, sizeof(*blob->active.pages));
196 	if (!blob->active.pages) {
197 		free(blob);
198 		return NULL;
199 	}
200 
201 	blob->active.pages[0] = _spdk_bs_blobid_to_page(id);
202 
203 	TAILQ_INIT(&blob->xattrs);
204 	TAILQ_INIT(&blob->xattrs_internal);
205 
206 	return blob;
207 }
208 
209 static void
210 _spdk_xattrs_free(struct spdk_xattr_tailq *xattrs)
211 {
212 	struct spdk_xattr	*xattr, *xattr_tmp;
213 
214 	TAILQ_FOREACH_SAFE(xattr, xattrs, link, xattr_tmp) {
215 		TAILQ_REMOVE(xattrs, xattr, link);
216 		free(xattr->name);
217 		free(xattr->value);
218 		free(xattr);
219 	}
220 }
221 
222 static void
223 _spdk_blob_free(struct spdk_blob *blob)
224 {
225 	assert(blob != NULL);
226 
227 	free(blob->active.clusters);
228 	free(blob->clean.clusters);
229 	free(blob->active.pages);
230 	free(blob->clean.pages);
231 
232 	_spdk_xattrs_free(&blob->xattrs);
233 	_spdk_xattrs_free(&blob->xattrs_internal);
234 
235 	if (blob->back_bs_dev) {
236 		blob->back_bs_dev->destroy(blob->back_bs_dev);
237 	}
238 
239 	free(blob);
240 }
241 
242 struct freeze_io_ctx {
243 	struct spdk_bs_cpl cpl;
244 	struct spdk_blob *blob;
245 };
246 
247 static void
248 _spdk_blob_io_sync(struct spdk_io_channel_iter *i)
249 {
250 	spdk_for_each_channel_continue(i, 0);
251 }
252 
253 static void
254 _spdk_blob_execute_queued_io(struct spdk_io_channel_iter *i)
255 {
256 	struct spdk_io_channel *_ch = spdk_io_channel_iter_get_channel(i);
257 	struct spdk_bs_channel *ch = spdk_io_channel_get_ctx(_ch);
258 	struct freeze_io_ctx *ctx = spdk_io_channel_iter_get_ctx(i);
259 	struct spdk_bs_request_set	*set;
260 	struct spdk_bs_user_op_args	*args;
261 	spdk_bs_user_op_t *op, *tmp;
262 
263 	TAILQ_FOREACH_SAFE(op, &ch->queued_io, link, tmp) {
264 		set = (struct spdk_bs_request_set *)op;
265 		args = &set->u.user_op;
266 
267 		if (args->blob == ctx->blob) {
268 			TAILQ_REMOVE(&ch->queued_io, op, link);
269 			spdk_bs_user_op_execute(op);
270 		}
271 	}
272 
273 	spdk_for_each_channel_continue(i, 0);
274 }
275 
276 static void
277 _spdk_blob_io_cpl(struct spdk_io_channel_iter *i, int status)
278 {
279 	struct freeze_io_ctx *ctx = spdk_io_channel_iter_get_ctx(i);
280 
281 	ctx->cpl.u.blob_basic.cb_fn(ctx->cpl.u.blob_basic.cb_arg, 0);
282 
283 	free(ctx);
284 }
285 
286 static void
287 _spdk_blob_freeze_io(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
288 {
289 	struct freeze_io_ctx *ctx;
290 
291 	ctx = calloc(1, sizeof(*ctx));
292 	if (!ctx) {
293 		cb_fn(cb_arg, -ENOMEM);
294 		return;
295 	}
296 
297 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
298 	ctx->cpl.u.blob_basic.cb_fn = cb_fn;
299 	ctx->cpl.u.blob_basic.cb_arg = cb_arg;
300 	ctx->blob = blob;
301 
302 	/* Freeze I/O on blob */
303 	blob->frozen_refcnt++;
304 
305 	if (blob->frozen_refcnt == 1) {
306 		spdk_for_each_channel(blob->bs, _spdk_blob_io_sync, ctx, _spdk_blob_io_cpl);
307 	} else {
308 		cb_fn(cb_arg, 0);
309 		free(ctx);
310 	}
311 }
312 
313 static void
314 _spdk_blob_unfreeze_io(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
315 {
316 	struct freeze_io_ctx *ctx;
317 
318 	ctx = calloc(1, sizeof(*ctx));
319 	if (!ctx) {
320 		cb_fn(cb_arg, -ENOMEM);
321 		return;
322 	}
323 
324 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
325 	ctx->cpl.u.blob_basic.cb_fn = cb_fn;
326 	ctx->cpl.u.blob_basic.cb_arg = cb_arg;
327 	ctx->blob = blob;
328 
329 	assert(blob->frozen_refcnt > 0);
330 
331 	blob->frozen_refcnt--;
332 
333 	if (blob->frozen_refcnt == 0) {
334 		spdk_for_each_channel(blob->bs, _spdk_blob_execute_queued_io, ctx, _spdk_blob_io_cpl);
335 	} else {
336 		cb_fn(cb_arg, 0);
337 		free(ctx);
338 	}
339 }
340 
341 static int
342 _spdk_blob_mark_clean(struct spdk_blob *blob)
343 {
344 	uint64_t *clusters = NULL;
345 	uint32_t *pages = NULL;
346 
347 	assert(blob != NULL);
348 
349 	if (blob->active.num_clusters) {
350 		assert(blob->active.clusters);
351 		clusters = calloc(blob->active.num_clusters, sizeof(*blob->active.clusters));
352 		if (!clusters) {
353 			return -ENOMEM;
354 		}
355 		memcpy(clusters, blob->active.clusters, blob->active.num_clusters * sizeof(*blob->active.clusters));
356 	}
357 
358 	if (blob->active.num_pages) {
359 		assert(blob->active.pages);
360 		pages = calloc(blob->active.num_pages, sizeof(*blob->active.pages));
361 		if (!pages) {
362 			free(clusters);
363 			return -ENOMEM;
364 		}
365 		memcpy(pages, blob->active.pages, blob->active.num_pages * sizeof(*blob->active.pages));
366 	}
367 
368 	free(blob->clean.clusters);
369 	free(blob->clean.pages);
370 
371 	blob->clean.num_clusters = blob->active.num_clusters;
372 	blob->clean.clusters = blob->active.clusters;
373 	blob->clean.num_pages = blob->active.num_pages;
374 	blob->clean.pages = blob->active.pages;
375 
376 	blob->active.clusters = clusters;
377 	blob->active.pages = pages;
378 
379 	/* If the metadata was dirtied again while the metadata was being written to disk,
380 	 *  we do not want to revert the DIRTY state back to CLEAN here.
381 	 */
382 	if (blob->state == SPDK_BLOB_STATE_LOADING) {
383 		blob->state = SPDK_BLOB_STATE_CLEAN;
384 	}
385 
386 	return 0;
387 }
388 
389 static int
390 _spdk_blob_deserialize_xattr(struct spdk_blob *blob,
391 			     struct spdk_blob_md_descriptor_xattr *desc_xattr, bool internal)
392 {
393 	struct spdk_xattr                       *xattr;
394 
395 	if (desc_xattr->length != sizeof(desc_xattr->name_length) +
396 	    sizeof(desc_xattr->value_length) +
397 	    desc_xattr->name_length + desc_xattr->value_length) {
398 		return -EINVAL;
399 	}
400 
401 	xattr = calloc(1, sizeof(*xattr));
402 	if (xattr == NULL) {
403 		return -ENOMEM;
404 	}
405 
406 	xattr->name = malloc(desc_xattr->name_length + 1);
407 	if (xattr->name == NULL) {
408 		free(xattr);
409 		return -ENOMEM;
410 	}
411 	memcpy(xattr->name, desc_xattr->name, desc_xattr->name_length);
412 	xattr->name[desc_xattr->name_length] = '\0';
413 
414 	xattr->value = malloc(desc_xattr->value_length);
415 	if (xattr->value == NULL) {
416 		free(xattr->name);
417 		free(xattr);
418 		return -ENOMEM;
419 	}
420 	xattr->value_len = desc_xattr->value_length;
421 	memcpy(xattr->value,
422 	       (void *)((uintptr_t)desc_xattr->name + desc_xattr->name_length),
423 	       desc_xattr->value_length);
424 
425 	TAILQ_INSERT_TAIL(internal ? &blob->xattrs_internal : &blob->xattrs, xattr, link);
426 
427 	return 0;
428 }
429 
430 
431 static int
432 _spdk_blob_parse_page(const struct spdk_blob_md_page *page, struct spdk_blob *blob)
433 {
434 	struct spdk_blob_md_descriptor *desc;
435 	size_t	cur_desc = 0;
436 	void *tmp;
437 
438 	desc = (struct spdk_blob_md_descriptor *)page->descriptors;
439 	while (cur_desc < sizeof(page->descriptors)) {
440 		if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_PADDING) {
441 			if (desc->length == 0) {
442 				/* If padding and length are 0, this terminates the page */
443 				break;
444 			}
445 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_FLAGS) {
446 			struct spdk_blob_md_descriptor_flags	*desc_flags;
447 
448 			desc_flags = (struct spdk_blob_md_descriptor_flags *)desc;
449 
450 			if (desc_flags->length != sizeof(*desc_flags) - sizeof(*desc)) {
451 				return -EINVAL;
452 			}
453 
454 			if ((desc_flags->invalid_flags | SPDK_BLOB_INVALID_FLAGS_MASK) !=
455 			    SPDK_BLOB_INVALID_FLAGS_MASK) {
456 				return -EINVAL;
457 			}
458 
459 			if ((desc_flags->data_ro_flags | SPDK_BLOB_DATA_RO_FLAGS_MASK) !=
460 			    SPDK_BLOB_DATA_RO_FLAGS_MASK) {
461 				blob->data_ro = true;
462 				blob->md_ro = true;
463 			}
464 
465 			if ((desc_flags->md_ro_flags | SPDK_BLOB_MD_RO_FLAGS_MASK) !=
466 			    SPDK_BLOB_MD_RO_FLAGS_MASK) {
467 				blob->md_ro = true;
468 			}
469 
470 			if ((desc_flags->data_ro_flags & SPDK_BLOB_READ_ONLY)) {
471 				blob->data_ro = true;
472 				blob->md_ro = true;
473 			}
474 
475 			blob->invalid_flags = desc_flags->invalid_flags;
476 			blob->data_ro_flags = desc_flags->data_ro_flags;
477 			blob->md_ro_flags = desc_flags->md_ro_flags;
478 
479 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE) {
480 			struct spdk_blob_md_descriptor_extent_rle	*desc_extent_rle;
481 			unsigned int				i, j;
482 			unsigned int				cluster_count = blob->active.num_clusters;
483 
484 			desc_extent_rle = (struct spdk_blob_md_descriptor_extent_rle *)desc;
485 
486 			if (desc_extent_rle->length == 0 ||
487 			    (desc_extent_rle->length % sizeof(desc_extent_rle->extents[0]) != 0)) {
488 				return -EINVAL;
489 			}
490 
491 			for (i = 0; i < desc_extent_rle->length / sizeof(desc_extent_rle->extents[0]); i++) {
492 				for (j = 0; j < desc_extent_rle->extents[i].length; j++) {
493 					if (desc_extent_rle->extents[i].cluster_idx != 0) {
494 						if (!spdk_bit_array_get(blob->bs->used_clusters,
495 									desc_extent_rle->extents[i].cluster_idx + j)) {
496 							return -EINVAL;
497 						}
498 					}
499 					cluster_count++;
500 				}
501 			}
502 
503 			if (cluster_count == 0) {
504 				return -EINVAL;
505 			}
506 			tmp = realloc(blob->active.clusters, cluster_count * sizeof(*blob->active.clusters));
507 			if (tmp == NULL) {
508 				return -ENOMEM;
509 			}
510 			blob->active.clusters = tmp;
511 			blob->active.cluster_array_size = cluster_count;
512 
513 			for (i = 0; i < desc_extent_rle->length / sizeof(desc_extent_rle->extents[0]); i++) {
514 				for (j = 0; j < desc_extent_rle->extents[i].length; j++) {
515 					if (desc_extent_rle->extents[i].cluster_idx != 0) {
516 						blob->active.clusters[blob->active.num_clusters++] = _spdk_bs_cluster_to_lba(blob->bs,
517 								desc_extent_rle->extents[i].cluster_idx + j);
518 					} else if (spdk_blob_is_thin_provisioned(blob)) {
519 						blob->active.clusters[blob->active.num_clusters++] = 0;
520 					} else {
521 						return -EINVAL;
522 					}
523 				}
524 			}
525 
526 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR) {
527 			int rc;
528 
529 			rc = _spdk_blob_deserialize_xattr(blob,
530 							  (struct spdk_blob_md_descriptor_xattr *) desc, false);
531 			if (rc != 0) {
532 				return rc;
533 			}
534 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL) {
535 			int rc;
536 
537 			rc = _spdk_blob_deserialize_xattr(blob,
538 							  (struct spdk_blob_md_descriptor_xattr *) desc, true);
539 			if (rc != 0) {
540 				return rc;
541 			}
542 		} else {
543 			/* Unrecognized descriptor type.  Do not fail - just continue to the
544 			 *  next descriptor.  If this descriptor is associated with some feature
545 			 *  defined in a newer version of blobstore, that version of blobstore
546 			 *  should create and set an associated feature flag to specify if this
547 			 *  blob can be loaded or not.
548 			 */
549 		}
550 
551 		/* Advance to the next descriptor */
552 		cur_desc += sizeof(*desc) + desc->length;
553 		if (cur_desc + sizeof(*desc) > sizeof(page->descriptors)) {
554 			break;
555 		}
556 		desc = (struct spdk_blob_md_descriptor *)((uintptr_t)page->descriptors + cur_desc);
557 	}
558 
559 	return 0;
560 }
561 
562 static int
563 _spdk_blob_parse(const struct spdk_blob_md_page *pages, uint32_t page_count,
564 		 struct spdk_blob *blob)
565 {
566 	const struct spdk_blob_md_page *page;
567 	uint32_t i;
568 	int rc;
569 
570 	assert(page_count > 0);
571 	assert(pages[0].sequence_num == 0);
572 	assert(blob != NULL);
573 	assert(blob->state == SPDK_BLOB_STATE_LOADING);
574 	assert(blob->active.clusters == NULL);
575 
576 	/* The blobid provided doesn't match what's in the MD, this can
577 	 * happen for example if a bogus blobid is passed in through open.
578 	 */
579 	if (blob->id != pages[0].id) {
580 		SPDK_ERRLOG("Blobid (%lu) doesn't match what's in metadata (%lu)\n",
581 			    blob->id, pages[0].id);
582 		return -ENOENT;
583 	}
584 
585 	for (i = 0; i < page_count; i++) {
586 		page = &pages[i];
587 
588 		assert(page->id == blob->id);
589 		assert(page->sequence_num == i);
590 
591 		rc = _spdk_blob_parse_page(page, blob);
592 		if (rc != 0) {
593 			return rc;
594 		}
595 	}
596 
597 	return 0;
598 }
599 
600 static int
601 _spdk_blob_serialize_add_page(const struct spdk_blob *blob,
602 			      struct spdk_blob_md_page **pages,
603 			      uint32_t *page_count,
604 			      struct spdk_blob_md_page **last_page)
605 {
606 	struct spdk_blob_md_page *page;
607 
608 	assert(pages != NULL);
609 	assert(page_count != NULL);
610 
611 	if (*page_count == 0) {
612 		assert(*pages == NULL);
613 		*page_count = 1;
614 		*pages = spdk_malloc(SPDK_BS_PAGE_SIZE, SPDK_BS_PAGE_SIZE,
615 				     NULL, SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
616 	} else {
617 		assert(*pages != NULL);
618 		(*page_count)++;
619 		*pages = spdk_realloc(*pages,
620 				      SPDK_BS_PAGE_SIZE * (*page_count),
621 				      SPDK_BS_PAGE_SIZE);
622 	}
623 
624 	if (*pages == NULL) {
625 		*page_count = 0;
626 		*last_page = NULL;
627 		return -ENOMEM;
628 	}
629 
630 	page = &(*pages)[*page_count - 1];
631 	memset(page, 0, sizeof(*page));
632 	page->id = blob->id;
633 	page->sequence_num = *page_count - 1;
634 	page->next = SPDK_INVALID_MD_PAGE;
635 	*last_page = page;
636 
637 	return 0;
638 }
639 
640 /* Transform the in-memory representation 'xattr' into an on-disk xattr descriptor.
641  * Update required_sz on both success and failure.
642  *
643  */
644 static int
645 _spdk_blob_serialize_xattr(const struct spdk_xattr *xattr,
646 			   uint8_t *buf, size_t buf_sz,
647 			   size_t *required_sz, bool internal)
648 {
649 	struct spdk_blob_md_descriptor_xattr	*desc;
650 
651 	*required_sz = sizeof(struct spdk_blob_md_descriptor_xattr) +
652 		       strlen(xattr->name) +
653 		       xattr->value_len;
654 
655 	if (buf_sz < *required_sz) {
656 		return -1;
657 	}
658 
659 	desc = (struct spdk_blob_md_descriptor_xattr *)buf;
660 
661 	desc->type = internal ? SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL : SPDK_MD_DESCRIPTOR_TYPE_XATTR;
662 	desc->length = sizeof(desc->name_length) +
663 		       sizeof(desc->value_length) +
664 		       strlen(xattr->name) +
665 		       xattr->value_len;
666 	desc->name_length = strlen(xattr->name);
667 	desc->value_length = xattr->value_len;
668 
669 	memcpy(desc->name, xattr->name, desc->name_length);
670 	memcpy((void *)((uintptr_t)desc->name + desc->name_length),
671 	       xattr->value,
672 	       desc->value_length);
673 
674 	return 0;
675 }
676 
677 static void
678 _spdk_blob_serialize_extent_rle(const struct spdk_blob *blob,
679 				uint64_t start_cluster, uint64_t *next_cluster,
680 				uint8_t **buf, size_t *buf_sz)
681 {
682 	struct spdk_blob_md_descriptor_extent_rle *desc_extent_rle;
683 	size_t cur_sz;
684 	uint64_t i, extent_idx;
685 	uint64_t lba, lba_per_cluster, lba_count;
686 
687 	/* The buffer must have room for at least one extent */
688 	cur_sz = sizeof(struct spdk_blob_md_descriptor) + sizeof(desc_extent_rle->extents[0]);
689 	if (*buf_sz < cur_sz) {
690 		*next_cluster = start_cluster;
691 		return;
692 	}
693 
694 	desc_extent_rle = (struct spdk_blob_md_descriptor_extent_rle *)*buf;
695 	desc_extent_rle->type = SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE;
696 
697 	lba_per_cluster = _spdk_bs_cluster_to_lba(blob->bs, 1);
698 
699 	lba = blob->active.clusters[start_cluster];
700 	lba_count = lba_per_cluster;
701 	extent_idx = 0;
702 	for (i = start_cluster + 1; i < blob->active.num_clusters; i++) {
703 		if ((lba + lba_count) == blob->active.clusters[i] && lba != 0) {
704 			/* Run-length encode sequential non-zero LBA */
705 			lba_count += lba_per_cluster;
706 			continue;
707 		} else if (lba == 0 && blob->active.clusters[i] == 0) {
708 			/* Run-length encode unallocated clusters */
709 			lba_count += lba_per_cluster;
710 			continue;
711 		}
712 		desc_extent_rle->extents[extent_idx].cluster_idx = lba / lba_per_cluster;
713 		desc_extent_rle->extents[extent_idx].length = lba_count / lba_per_cluster;
714 		extent_idx++;
715 
716 		cur_sz += sizeof(desc_extent_rle->extents[extent_idx]);
717 
718 		if (*buf_sz < cur_sz) {
719 			/* If we ran out of buffer space, return */
720 			*next_cluster = i;
721 			goto finish;
722 		}
723 
724 		lba = blob->active.clusters[i];
725 		lba_count = lba_per_cluster;
726 	}
727 
728 	desc_extent_rle->extents[extent_idx].cluster_idx = lba / lba_per_cluster;
729 	desc_extent_rle->extents[extent_idx].length = lba_count / lba_per_cluster;
730 	extent_idx++;
731 
732 	*next_cluster = blob->active.num_clusters;
733 
734 finish:
735 	desc_extent_rle->length = sizeof(desc_extent_rle->extents[0]) * extent_idx;
736 	*buf_sz -= sizeof(struct spdk_blob_md_descriptor) + desc_extent_rle->length;
737 	*buf += sizeof(struct spdk_blob_md_descriptor) + desc_extent_rle->length;
738 
739 	return;
740 }
741 
742 static int
743 _spdk_blob_serialize_extents_rle(const struct spdk_blob *blob,
744 				 struct spdk_blob_md_page **pages,
745 				 struct spdk_blob_md_page *cur_page,
746 				 uint32_t *page_count, uint8_t **buf,
747 				 size_t *remaining_sz)
748 {
749 	uint64_t				last_cluster;
750 	int					rc;
751 
752 	last_cluster = 0;
753 	while (last_cluster < blob->active.num_clusters) {
754 		_spdk_blob_serialize_extent_rle(blob, last_cluster, &last_cluster, buf, remaining_sz);
755 
756 		if (last_cluster == blob->active.num_clusters) {
757 			break;
758 		}
759 
760 		rc = _spdk_blob_serialize_add_page(blob, pages, page_count, &cur_page);
761 		if (rc < 0) {
762 			return rc;
763 		}
764 
765 		*buf = (uint8_t *)cur_page->descriptors;
766 		*remaining_sz = sizeof(cur_page->descriptors);
767 	}
768 
769 	return 0;
770 }
771 
772 static void
773 _spdk_blob_serialize_flags(const struct spdk_blob *blob,
774 			   uint8_t *buf, size_t *buf_sz)
775 {
776 	struct spdk_blob_md_descriptor_flags *desc;
777 
778 	/*
779 	 * Flags get serialized first, so we should always have room for the flags
780 	 *  descriptor.
781 	 */
782 	assert(*buf_sz >= sizeof(*desc));
783 
784 	desc = (struct spdk_blob_md_descriptor_flags *)buf;
785 	desc->type = SPDK_MD_DESCRIPTOR_TYPE_FLAGS;
786 	desc->length = sizeof(*desc) - sizeof(struct spdk_blob_md_descriptor);
787 	desc->invalid_flags = blob->invalid_flags;
788 	desc->data_ro_flags = blob->data_ro_flags;
789 	desc->md_ro_flags = blob->md_ro_flags;
790 
791 	*buf_sz -= sizeof(*desc);
792 }
793 
794 static int
795 _spdk_blob_serialize_xattrs(const struct spdk_blob *blob,
796 			    const struct spdk_xattr_tailq *xattrs, bool internal,
797 			    struct spdk_blob_md_page **pages,
798 			    struct spdk_blob_md_page *cur_page,
799 			    uint32_t *page_count, uint8_t **buf,
800 			    size_t *remaining_sz)
801 {
802 	const struct spdk_xattr	*xattr;
803 	int	rc;
804 
805 	TAILQ_FOREACH(xattr, xattrs, link) {
806 		size_t required_sz = 0;
807 
808 		rc = _spdk_blob_serialize_xattr(xattr,
809 						*buf, *remaining_sz,
810 						&required_sz, internal);
811 		if (rc < 0) {
812 			/* Need to add a new page to the chain */
813 			rc = _spdk_blob_serialize_add_page(blob, pages, page_count,
814 							   &cur_page);
815 			if (rc < 0) {
816 				spdk_free(*pages);
817 				*pages = NULL;
818 				*page_count = 0;
819 				return rc;
820 			}
821 
822 			*buf = (uint8_t *)cur_page->descriptors;
823 			*remaining_sz = sizeof(cur_page->descriptors);
824 
825 			/* Try again */
826 			required_sz = 0;
827 			rc = _spdk_blob_serialize_xattr(xattr,
828 							*buf, *remaining_sz,
829 							&required_sz, internal);
830 
831 			if (rc < 0) {
832 				spdk_free(*pages);
833 				*pages = NULL;
834 				*page_count = 0;
835 				return rc;
836 			}
837 		}
838 
839 		*remaining_sz -= required_sz;
840 		*buf += required_sz;
841 	}
842 
843 	return 0;
844 }
845 
846 static int
847 _spdk_blob_serialize(const struct spdk_blob *blob, struct spdk_blob_md_page **pages,
848 		     uint32_t *page_count)
849 {
850 	struct spdk_blob_md_page		*cur_page;
851 	int					rc;
852 	uint8_t					*buf;
853 	size_t					remaining_sz;
854 
855 	assert(pages != NULL);
856 	assert(page_count != NULL);
857 	assert(blob != NULL);
858 	assert(blob->state == SPDK_BLOB_STATE_DIRTY);
859 
860 	*pages = NULL;
861 	*page_count = 0;
862 
863 	/* A blob always has at least 1 page, even if it has no descriptors */
864 	rc = _spdk_blob_serialize_add_page(blob, pages, page_count, &cur_page);
865 	if (rc < 0) {
866 		return rc;
867 	}
868 
869 	buf = (uint8_t *)cur_page->descriptors;
870 	remaining_sz = sizeof(cur_page->descriptors);
871 
872 	/* Serialize flags */
873 	_spdk_blob_serialize_flags(blob, buf, &remaining_sz);
874 	buf += sizeof(struct spdk_blob_md_descriptor_flags);
875 
876 	/* Serialize xattrs */
877 	rc = _spdk_blob_serialize_xattrs(blob, &blob->xattrs, false,
878 					 pages, cur_page, page_count, &buf, &remaining_sz);
879 	if (rc < 0) {
880 		return rc;
881 	}
882 
883 	/* Serialize internal xattrs */
884 	rc = _spdk_blob_serialize_xattrs(blob, &blob->xattrs_internal, true,
885 					 pages, cur_page, page_count, &buf, &remaining_sz);
886 	if (rc < 0) {
887 		return rc;
888 	}
889 
890 	/* Serialize extents */
891 	rc = _spdk_blob_serialize_extents_rle(blob, pages, cur_page, page_count, &buf, &remaining_sz);
892 
893 	return rc;
894 }
895 
896 struct spdk_blob_load_ctx {
897 	struct spdk_blob		*blob;
898 
899 	struct spdk_blob_md_page	*pages;
900 	uint32_t			num_pages;
901 	spdk_bs_sequence_t	        *seq;
902 
903 	spdk_bs_sequence_cpl		cb_fn;
904 	void				*cb_arg;
905 };
906 
907 static uint32_t
908 _spdk_blob_md_page_calc_crc(void *page)
909 {
910 	uint32_t		crc;
911 
912 	crc = BLOB_CRC32C_INITIAL;
913 	crc = spdk_crc32c_update(page, SPDK_BS_PAGE_SIZE - 4, crc);
914 	crc ^= BLOB_CRC32C_INITIAL;
915 
916 	return crc;
917 
918 }
919 
920 static void
921 _spdk_blob_load_final(void *cb_arg, int bserrno)
922 {
923 	struct spdk_blob_load_ctx	*ctx = cb_arg;
924 	struct spdk_blob		*blob = ctx->blob;
925 
926 	if (bserrno == 0) {
927 		_spdk_blob_mark_clean(blob);
928 	}
929 
930 	ctx->cb_fn(ctx->seq, ctx->cb_arg, bserrno);
931 
932 	/* Free the memory */
933 	spdk_free(ctx->pages);
934 	free(ctx);
935 }
936 
937 static void
938 _spdk_blob_load_snapshot_cpl(void *cb_arg, struct spdk_blob *snapshot, int bserrno)
939 {
940 	struct spdk_blob_load_ctx	*ctx = cb_arg;
941 	struct spdk_blob		*blob = ctx->blob;
942 
943 	if (bserrno == 0) {
944 		blob->back_bs_dev = spdk_bs_create_blob_bs_dev(snapshot);
945 		if (blob->back_bs_dev == NULL) {
946 			bserrno = -ENOMEM;
947 		}
948 	}
949 	if (bserrno != 0) {
950 		SPDK_ERRLOG("Snapshot fail\n");
951 	}
952 
953 	_spdk_blob_load_final(ctx, bserrno);
954 }
955 
956 static void _spdk_blob_update_clear_method(struct spdk_blob *blob);
957 
958 static void
959 _spdk_blob_load_backing_dev(void *cb_arg)
960 {
961 	struct spdk_blob_load_ctx	*ctx = cb_arg;
962 	struct spdk_blob		*blob = ctx->blob;
963 	const void			*value;
964 	size_t				len;
965 	int				rc;
966 
967 	if (spdk_blob_is_thin_provisioned(blob)) {
968 		rc = _spdk_blob_get_xattr_value(blob, BLOB_SNAPSHOT, &value, &len, true);
969 		if (rc == 0) {
970 			if (len != sizeof(spdk_blob_id)) {
971 				_spdk_blob_load_final(ctx, -EINVAL);
972 				return;
973 			}
974 			/* open snapshot blob and continue in the callback function */
975 			blob->parent_id = *(spdk_blob_id *)value;
976 			spdk_bs_open_blob(blob->bs, blob->parent_id,
977 					  _spdk_blob_load_snapshot_cpl, ctx);
978 			return;
979 		} else {
980 			/* add zeroes_dev for thin provisioned blob */
981 			blob->back_bs_dev = spdk_bs_create_zeroes_dev();
982 		}
983 	} else {
984 		/* standard blob */
985 		blob->back_bs_dev = NULL;
986 	}
987 	_spdk_blob_load_final(ctx, 0);
988 }
989 
990 static void
991 _spdk_blob_load_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
992 {
993 	struct spdk_blob_load_ctx	*ctx = cb_arg;
994 	struct spdk_blob		*blob = ctx->blob;
995 	struct spdk_blob_md_page	*page;
996 	int				rc;
997 	uint32_t			crc;
998 
999 	if (bserrno) {
1000 		SPDK_ERRLOG("Metadata page read failed: %d\n", bserrno);
1001 		_spdk_blob_load_final(ctx, bserrno);
1002 		return;
1003 	}
1004 
1005 	page = &ctx->pages[ctx->num_pages - 1];
1006 	crc = _spdk_blob_md_page_calc_crc(page);
1007 	if (crc != page->crc) {
1008 		SPDK_ERRLOG("Metadata page %d crc mismatch\n", ctx->num_pages);
1009 		_spdk_blob_load_final(ctx, -EINVAL);
1010 		return;
1011 	}
1012 
1013 	if (page->next != SPDK_INVALID_MD_PAGE) {
1014 		uint32_t next_page = page->next;
1015 		uint64_t next_lba = _spdk_bs_md_page_to_lba(blob->bs, next_page);
1016 
1017 		/* Read the next page */
1018 		ctx->num_pages++;
1019 		ctx->pages = spdk_realloc(ctx->pages, (sizeof(*page) * ctx->num_pages),
1020 					  sizeof(*page));
1021 		if (ctx->pages == NULL) {
1022 			_spdk_blob_load_final(ctx, -ENOMEM);
1023 			return;
1024 		}
1025 
1026 		spdk_bs_sequence_read_dev(seq, &ctx->pages[ctx->num_pages - 1],
1027 					  next_lba,
1028 					  _spdk_bs_byte_to_lba(blob->bs, sizeof(*page)),
1029 					  _spdk_blob_load_cpl, ctx);
1030 		return;
1031 	}
1032 
1033 	/* Parse the pages */
1034 	rc = _spdk_blob_parse(ctx->pages, ctx->num_pages, blob);
1035 	if (rc) {
1036 		_spdk_blob_load_final(ctx, rc);
1037 		return;
1038 	}
1039 	ctx->seq = seq;
1040 
1041 	/* Check the clear_method stored in metadata vs what may have been passed
1042 	 * via spdk_bs_open_blob_ext() and update accordingly.
1043 	 */
1044 	_spdk_blob_update_clear_method(blob);
1045 
1046 	_spdk_blob_load_backing_dev(ctx);
1047 }
1048 
1049 /* Load a blob from disk given a blobid */
1050 static void
1051 _spdk_blob_load(spdk_bs_sequence_t *seq, struct spdk_blob *blob,
1052 		spdk_bs_sequence_cpl cb_fn, void *cb_arg)
1053 {
1054 	struct spdk_blob_load_ctx *ctx;
1055 	struct spdk_blob_store *bs;
1056 	uint32_t page_num;
1057 	uint64_t lba;
1058 
1059 	_spdk_blob_verify_md_op(blob);
1060 
1061 	bs = blob->bs;
1062 
1063 	ctx = calloc(1, sizeof(*ctx));
1064 	if (!ctx) {
1065 		cb_fn(seq, cb_arg, -ENOMEM);
1066 		return;
1067 	}
1068 
1069 	ctx->blob = blob;
1070 	ctx->pages = spdk_realloc(ctx->pages, SPDK_BS_PAGE_SIZE, SPDK_BS_PAGE_SIZE);
1071 	if (!ctx->pages) {
1072 		free(ctx);
1073 		cb_fn(seq, cb_arg, -ENOMEM);
1074 		return;
1075 	}
1076 	ctx->num_pages = 1;
1077 	ctx->cb_fn = cb_fn;
1078 	ctx->cb_arg = cb_arg;
1079 	ctx->seq = seq;
1080 
1081 	page_num = _spdk_bs_blobid_to_page(blob->id);
1082 	lba = _spdk_bs_md_page_to_lba(blob->bs, page_num);
1083 
1084 	blob->state = SPDK_BLOB_STATE_LOADING;
1085 
1086 	spdk_bs_sequence_read_dev(seq, &ctx->pages[0], lba,
1087 				  _spdk_bs_byte_to_lba(bs, SPDK_BS_PAGE_SIZE),
1088 				  _spdk_blob_load_cpl, ctx);
1089 }
1090 
1091 struct spdk_blob_persist_ctx {
1092 	struct spdk_blob		*blob;
1093 
1094 	struct spdk_bs_super_block	*super;
1095 
1096 	struct spdk_blob_md_page	*pages;
1097 
1098 	spdk_bs_sequence_t		*seq;
1099 	spdk_bs_sequence_cpl		cb_fn;
1100 	void				*cb_arg;
1101 };
1102 
1103 static void
1104 spdk_bs_batch_clear_dev(struct spdk_blob_persist_ctx *ctx, spdk_bs_batch_t *batch, uint64_t lba,
1105 			uint32_t lba_count)
1106 {
1107 	switch (ctx->blob->clear_method) {
1108 	case BLOB_CLEAR_WITH_DEFAULT:
1109 	case BLOB_CLEAR_WITH_UNMAP:
1110 		spdk_bs_batch_unmap_dev(batch, lba, lba_count);
1111 		break;
1112 	case BLOB_CLEAR_WITH_WRITE_ZEROES:
1113 		spdk_bs_batch_write_zeroes_dev(batch, lba, lba_count);
1114 		break;
1115 	case BLOB_CLEAR_WITH_NONE:
1116 	default:
1117 		break;
1118 	}
1119 }
1120 
1121 static void
1122 _spdk_blob_persist_complete(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1123 {
1124 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1125 	struct spdk_blob		*blob = ctx->blob;
1126 
1127 	if (bserrno == 0) {
1128 		_spdk_blob_mark_clean(blob);
1129 	}
1130 
1131 	/* Call user callback */
1132 	ctx->cb_fn(seq, ctx->cb_arg, bserrno);
1133 
1134 	/* Free the memory */
1135 	spdk_free(ctx->pages);
1136 	free(ctx);
1137 }
1138 
1139 static void
1140 _spdk_blob_persist_clear_clusters_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1141 {
1142 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1143 	struct spdk_blob		*blob = ctx->blob;
1144 	struct spdk_blob_store		*bs = blob->bs;
1145 	size_t				i;
1146 
1147 	/* Release all clusters that were truncated */
1148 	for (i = blob->active.num_clusters; i < blob->active.cluster_array_size; i++) {
1149 		uint32_t cluster_num = _spdk_bs_lba_to_cluster(bs, blob->active.clusters[i]);
1150 
1151 		/* Nothing to release if it was not allocated */
1152 		if (blob->active.clusters[i] != 0) {
1153 			_spdk_bs_release_cluster(bs, cluster_num);
1154 		}
1155 	}
1156 
1157 	if (blob->active.num_clusters == 0) {
1158 		free(blob->active.clusters);
1159 		blob->active.clusters = NULL;
1160 		blob->active.cluster_array_size = 0;
1161 	} else if (blob->active.num_clusters != blob->active.cluster_array_size) {
1162 #ifndef __clang_analyzer__
1163 		void *tmp;
1164 
1165 		/* scan-build really can't figure reallocs, workaround it */
1166 		tmp = realloc(blob->active.clusters, sizeof(*blob->active.clusters) * blob->active.num_clusters);
1167 		assert(tmp != NULL);
1168 		blob->active.clusters = tmp;
1169 #endif
1170 		blob->active.cluster_array_size = blob->active.num_clusters;
1171 	}
1172 
1173 	_spdk_blob_persist_complete(seq, ctx, bserrno);
1174 }
1175 
1176 static void
1177 _spdk_blob_persist_clear_clusters(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1178 {
1179 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1180 	struct spdk_blob		*blob = ctx->blob;
1181 	struct spdk_blob_store		*bs = blob->bs;
1182 	spdk_bs_batch_t			*batch;
1183 	size_t				i;
1184 	uint64_t			lba;
1185 	uint32_t			lba_count;
1186 
1187 	/* Clusters don't move around in blobs. The list shrinks or grows
1188 	 * at the end, but no changes ever occur in the middle of the list.
1189 	 */
1190 
1191 	batch = spdk_bs_sequence_to_batch(seq, _spdk_blob_persist_clear_clusters_cpl, ctx);
1192 
1193 	/* Clear all clusters that were truncated */
1194 	lba = 0;
1195 	lba_count = 0;
1196 	for (i = blob->active.num_clusters; i < blob->active.cluster_array_size; i++) {
1197 		uint64_t next_lba = blob->active.clusters[i];
1198 		uint32_t next_lba_count = _spdk_bs_cluster_to_lba(bs, 1);
1199 
1200 		if (next_lba > 0 && (lba + lba_count) == next_lba) {
1201 			/* This cluster is contiguous with the previous one. */
1202 			lba_count += next_lba_count;
1203 			continue;
1204 		}
1205 
1206 		/* This cluster is not contiguous with the previous one. */
1207 
1208 		/* If a run of LBAs previously existing, clear them now */
1209 		if (lba_count > 0) {
1210 			spdk_bs_batch_clear_dev(ctx, batch, lba, lba_count);
1211 		}
1212 
1213 		/* Start building the next batch */
1214 		lba = next_lba;
1215 		if (next_lba > 0) {
1216 			lba_count = next_lba_count;
1217 		} else {
1218 			lba_count = 0;
1219 		}
1220 	}
1221 
1222 	/* If we ended with a contiguous set of LBAs, clear them now */
1223 	if (lba_count > 0) {
1224 		spdk_bs_batch_clear_dev(ctx, batch, lba, lba_count);
1225 	}
1226 
1227 	spdk_bs_batch_close(batch);
1228 }
1229 
1230 static void
1231 _spdk_blob_persist_zero_pages_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1232 {
1233 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1234 	struct spdk_blob		*blob = ctx->blob;
1235 	struct spdk_blob_store		*bs = blob->bs;
1236 	size_t				i;
1237 
1238 	/* This loop starts at 1 because the first page is special and handled
1239 	 * below. The pages (except the first) are never written in place,
1240 	 * so any pages in the clean list must be zeroed.
1241 	 */
1242 	for (i = 1; i < blob->clean.num_pages; i++) {
1243 		spdk_bit_array_clear(bs->used_md_pages, blob->clean.pages[i]);
1244 	}
1245 
1246 	if (blob->active.num_pages == 0) {
1247 		uint32_t page_num;
1248 
1249 		page_num = _spdk_bs_blobid_to_page(blob->id);
1250 		spdk_bit_array_clear(bs->used_md_pages, page_num);
1251 	}
1252 
1253 	/* Move on to clearing clusters */
1254 	_spdk_blob_persist_clear_clusters(seq, ctx, 0);
1255 }
1256 
1257 static void
1258 _spdk_blob_persist_zero_pages(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1259 {
1260 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1261 	struct spdk_blob		*blob = ctx->blob;
1262 	struct spdk_blob_store		*bs = blob->bs;
1263 	uint64_t			lba;
1264 	uint32_t			lba_count;
1265 	spdk_bs_batch_t			*batch;
1266 	size_t				i;
1267 
1268 	batch = spdk_bs_sequence_to_batch(seq, _spdk_blob_persist_zero_pages_cpl, ctx);
1269 
1270 	lba_count = _spdk_bs_byte_to_lba(bs, SPDK_BS_PAGE_SIZE);
1271 
1272 	/* This loop starts at 1 because the first page is special and handled
1273 	 * below. The pages (except the first) are never written in place,
1274 	 * so any pages in the clean list must be zeroed.
1275 	 */
1276 	for (i = 1; i < blob->clean.num_pages; i++) {
1277 		lba = _spdk_bs_md_page_to_lba(bs, blob->clean.pages[i]);
1278 
1279 		spdk_bs_batch_write_zeroes_dev(batch, lba, lba_count);
1280 	}
1281 
1282 	/* The first page will only be zeroed if this is a delete. */
1283 	if (blob->active.num_pages == 0) {
1284 		uint32_t page_num;
1285 
1286 		/* The first page in the metadata goes where the blobid indicates */
1287 		page_num = _spdk_bs_blobid_to_page(blob->id);
1288 		lba = _spdk_bs_md_page_to_lba(bs, page_num);
1289 
1290 		spdk_bs_batch_write_zeroes_dev(batch, lba, lba_count);
1291 	}
1292 
1293 	spdk_bs_batch_close(batch);
1294 }
1295 
1296 static void
1297 _spdk_blob_persist_write_page_root(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1298 {
1299 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1300 	struct spdk_blob		*blob = ctx->blob;
1301 	struct spdk_blob_store		*bs = blob->bs;
1302 	uint64_t			lba;
1303 	uint32_t			lba_count;
1304 	struct spdk_blob_md_page	*page;
1305 
1306 	if (blob->active.num_pages == 0) {
1307 		/* Move on to the next step */
1308 		_spdk_blob_persist_zero_pages(seq, ctx, 0);
1309 		return;
1310 	}
1311 
1312 	lba_count = _spdk_bs_byte_to_lba(bs, sizeof(*page));
1313 
1314 	page = &ctx->pages[0];
1315 	/* The first page in the metadata goes where the blobid indicates */
1316 	lba = _spdk_bs_md_page_to_lba(bs, _spdk_bs_blobid_to_page(blob->id));
1317 
1318 	spdk_bs_sequence_write_dev(seq, page, lba, lba_count,
1319 				   _spdk_blob_persist_zero_pages, ctx);
1320 }
1321 
1322 static void
1323 _spdk_blob_persist_write_page_chain(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1324 {
1325 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1326 	struct spdk_blob		*blob = ctx->blob;
1327 	struct spdk_blob_store		*bs = blob->bs;
1328 	uint64_t			lba;
1329 	uint32_t			lba_count;
1330 	struct spdk_blob_md_page	*page;
1331 	spdk_bs_batch_t			*batch;
1332 	size_t				i;
1333 
1334 	/* Clusters don't move around in blobs. The list shrinks or grows
1335 	 * at the end, but no changes ever occur in the middle of the list.
1336 	 */
1337 
1338 	lba_count = _spdk_bs_byte_to_lba(bs, sizeof(*page));
1339 
1340 	batch = spdk_bs_sequence_to_batch(seq, _spdk_blob_persist_write_page_root, ctx);
1341 
1342 	/* This starts at 1. The root page is not written until
1343 	 * all of the others are finished
1344 	 */
1345 	for (i = 1; i < blob->active.num_pages; i++) {
1346 		page = &ctx->pages[i];
1347 		assert(page->sequence_num == i);
1348 
1349 		lba = _spdk_bs_md_page_to_lba(bs, blob->active.pages[i]);
1350 
1351 		spdk_bs_batch_write_dev(batch, page, lba, lba_count);
1352 	}
1353 
1354 	spdk_bs_batch_close(batch);
1355 }
1356 
1357 static int
1358 _spdk_blob_resize(struct spdk_blob *blob, uint64_t sz)
1359 {
1360 	uint64_t	i;
1361 	uint64_t	*tmp;
1362 	uint64_t	lfc; /* lowest free cluster */
1363 	uint64_t	num_clusters;
1364 	struct spdk_blob_store *bs;
1365 
1366 	bs = blob->bs;
1367 
1368 	_spdk_blob_verify_md_op(blob);
1369 
1370 	if (blob->active.num_clusters == sz) {
1371 		return 0;
1372 	}
1373 
1374 	if (blob->active.num_clusters < blob->active.cluster_array_size) {
1375 		/* If this blob was resized to be larger, then smaller, then
1376 		 * larger without syncing, then the cluster array already
1377 		 * contains spare assigned clusters we can use.
1378 		 */
1379 		num_clusters = spdk_min(blob->active.cluster_array_size,
1380 					sz);
1381 	} else {
1382 		num_clusters = blob->active.num_clusters;
1383 	}
1384 
1385 	/* Do two passes - one to verify that we can obtain enough clusters
1386 	 * and another to actually claim them.
1387 	 */
1388 
1389 	if (spdk_blob_is_thin_provisioned(blob) == false) {
1390 		lfc = 0;
1391 		for (i = num_clusters; i < sz; i++) {
1392 			lfc = spdk_bit_array_find_first_clear(bs->used_clusters, lfc);
1393 			if (lfc == UINT32_MAX) {
1394 				/* No more free clusters. Cannot satisfy the request */
1395 				return -ENOSPC;
1396 			}
1397 			lfc++;
1398 		}
1399 	}
1400 
1401 	if (sz > num_clusters) {
1402 		/* Expand the cluster array if necessary.
1403 		 * We only shrink the array when persisting.
1404 		 */
1405 		tmp = realloc(blob->active.clusters, sizeof(*blob->active.clusters) * sz);
1406 		if (sz > 0 && tmp == NULL) {
1407 			return -ENOMEM;
1408 		}
1409 		memset(tmp + blob->active.cluster_array_size, 0,
1410 		       sizeof(*blob->active.clusters) * (sz - blob->active.cluster_array_size));
1411 		blob->active.clusters = tmp;
1412 		blob->active.cluster_array_size = sz;
1413 	}
1414 
1415 	blob->state = SPDK_BLOB_STATE_DIRTY;
1416 
1417 	if (spdk_blob_is_thin_provisioned(blob) == false) {
1418 		lfc = 0;
1419 		for (i = num_clusters; i < sz; i++) {
1420 			_spdk_bs_allocate_cluster(blob, i, &lfc, true);
1421 			lfc++;
1422 		}
1423 	}
1424 
1425 	blob->active.num_clusters = sz;
1426 
1427 	return 0;
1428 }
1429 
1430 static void
1431 _spdk_blob_persist_generate_new_md(struct spdk_blob_persist_ctx *ctx)
1432 {
1433 	spdk_bs_sequence_t *seq = ctx->seq;
1434 	struct spdk_blob *blob = ctx->blob;
1435 	struct spdk_blob_store *bs = blob->bs;
1436 	uint64_t i;
1437 	uint32_t page_num;
1438 	void *tmp;
1439 	int rc;
1440 
1441 	/* Generate the new metadata */
1442 	rc = _spdk_blob_serialize(blob, &ctx->pages, &blob->active.num_pages);
1443 	if (rc < 0) {
1444 		_spdk_blob_persist_complete(seq, ctx, rc);
1445 		return;
1446 	}
1447 
1448 	assert(blob->active.num_pages >= 1);
1449 
1450 	/* Resize the cache of page indices */
1451 	tmp = realloc(blob->active.pages, blob->active.num_pages * sizeof(*blob->active.pages));
1452 	if (!tmp) {
1453 		_spdk_blob_persist_complete(seq, ctx, -ENOMEM);
1454 		return;
1455 	}
1456 	blob->active.pages = tmp;
1457 
1458 	/* Assign this metadata to pages. This requires two passes -
1459 	 * one to verify that there are enough pages and a second
1460 	 * to actually claim them. */
1461 	page_num = 0;
1462 	/* Note that this loop starts at one. The first page location is fixed by the blobid. */
1463 	for (i = 1; i < blob->active.num_pages; i++) {
1464 		page_num = spdk_bit_array_find_first_clear(bs->used_md_pages, page_num);
1465 		if (page_num == UINT32_MAX) {
1466 			_spdk_blob_persist_complete(seq, ctx, -ENOMEM);
1467 			return;
1468 		}
1469 		page_num++;
1470 	}
1471 
1472 	page_num = 0;
1473 	blob->active.pages[0] = _spdk_bs_blobid_to_page(blob->id);
1474 	for (i = 1; i < blob->active.num_pages; i++) {
1475 		page_num = spdk_bit_array_find_first_clear(bs->used_md_pages, page_num);
1476 		ctx->pages[i - 1].next = page_num;
1477 		/* Now that previous metadata page is complete, calculate the crc for it. */
1478 		ctx->pages[i - 1].crc = _spdk_blob_md_page_calc_crc(&ctx->pages[i - 1]);
1479 		blob->active.pages[i] = page_num;
1480 		spdk_bit_array_set(bs->used_md_pages, page_num);
1481 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Claiming page %u for blob %lu\n", page_num, blob->id);
1482 		page_num++;
1483 	}
1484 	ctx->pages[i - 1].crc = _spdk_blob_md_page_calc_crc(&ctx->pages[i - 1]);
1485 	/* Start writing the metadata from last page to first */
1486 	blob->state = SPDK_BLOB_STATE_CLEAN;
1487 	_spdk_blob_persist_write_page_chain(seq, ctx, 0);
1488 }
1489 
1490 static void
1491 _spdk_blob_persist_start(struct spdk_blob_persist_ctx *ctx)
1492 {
1493 	spdk_bs_sequence_t *seq = ctx->seq;
1494 	struct spdk_blob *blob = ctx->blob;
1495 
1496 	if (blob->active.num_pages == 0) {
1497 		/* This is the signal that the blob should be deleted.
1498 		 * Immediately jump to the clean up routine. */
1499 		assert(blob->clean.num_pages > 0);
1500 		blob->state = SPDK_BLOB_STATE_CLEAN;
1501 		_spdk_blob_persist_zero_pages(seq, ctx, 0);
1502 		return;
1503 
1504 	}
1505 
1506 	_spdk_blob_persist_generate_new_md(ctx);
1507 }
1508 
1509 static void
1510 _spdk_blob_persist_dirty_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1511 {
1512 	struct spdk_blob_persist_ctx *ctx = cb_arg;
1513 
1514 	ctx->blob->bs->clean = 0;
1515 
1516 	spdk_free(ctx->super);
1517 
1518 	_spdk_blob_persist_start(ctx);
1519 }
1520 
1521 static void
1522 _spdk_bs_write_super(spdk_bs_sequence_t *seq, struct spdk_blob_store *bs,
1523 		     struct spdk_bs_super_block *super, spdk_bs_sequence_cpl cb_fn, void *cb_arg);
1524 
1525 
1526 static void
1527 _spdk_blob_persist_dirty(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1528 {
1529 	struct spdk_blob_persist_ctx *ctx = cb_arg;
1530 
1531 	ctx->super->clean = 0;
1532 	if (ctx->super->size == 0) {
1533 		ctx->super->size = ctx->blob->bs->dev->blockcnt * ctx->blob->bs->dev->blocklen;
1534 	}
1535 
1536 	_spdk_bs_write_super(seq, ctx->blob->bs, ctx->super, _spdk_blob_persist_dirty_cpl, ctx);
1537 }
1538 
1539 
1540 /* Write a blob to disk */
1541 static void
1542 _spdk_blob_persist(spdk_bs_sequence_t *seq, struct spdk_blob *blob,
1543 		   spdk_bs_sequence_cpl cb_fn, void *cb_arg)
1544 {
1545 	struct spdk_blob_persist_ctx *ctx;
1546 
1547 	_spdk_blob_verify_md_op(blob);
1548 
1549 	if (blob->state == SPDK_BLOB_STATE_CLEAN) {
1550 		cb_fn(seq, cb_arg, 0);
1551 		return;
1552 	}
1553 
1554 	ctx = calloc(1, sizeof(*ctx));
1555 	if (!ctx) {
1556 		cb_fn(seq, cb_arg, -ENOMEM);
1557 		return;
1558 	}
1559 	ctx->blob = blob;
1560 	ctx->seq = seq;
1561 	ctx->cb_fn = cb_fn;
1562 	ctx->cb_arg = cb_arg;
1563 
1564 	if (blob->bs->clean) {
1565 		ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
1566 					  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
1567 		if (!ctx->super) {
1568 			cb_fn(seq, cb_arg, -ENOMEM);
1569 			free(ctx);
1570 			return;
1571 		}
1572 
1573 		spdk_bs_sequence_read_dev(seq, ctx->super, _spdk_bs_page_to_lba(blob->bs, 0),
1574 					  _spdk_bs_byte_to_lba(blob->bs, sizeof(*ctx->super)),
1575 					  _spdk_blob_persist_dirty, ctx);
1576 	} else {
1577 		_spdk_blob_persist_start(ctx);
1578 	}
1579 }
1580 
1581 struct spdk_blob_copy_cluster_ctx {
1582 	struct spdk_blob *blob;
1583 	uint8_t *buf;
1584 	uint64_t page;
1585 	uint64_t new_cluster;
1586 	spdk_bs_sequence_t *seq;
1587 };
1588 
1589 static void
1590 _spdk_blob_allocate_and_copy_cluster_cpl(void *cb_arg, int bserrno)
1591 {
1592 	struct spdk_blob_copy_cluster_ctx *ctx = cb_arg;
1593 	struct spdk_bs_request_set *set = (struct spdk_bs_request_set *)ctx->seq;
1594 	TAILQ_HEAD(, spdk_bs_request_set) requests;
1595 	spdk_bs_user_op_t *op;
1596 
1597 	TAILQ_INIT(&requests);
1598 	TAILQ_SWAP(&set->channel->need_cluster_alloc, &requests, spdk_bs_request_set, link);
1599 
1600 	while (!TAILQ_EMPTY(&requests)) {
1601 		op = TAILQ_FIRST(&requests);
1602 		TAILQ_REMOVE(&requests, op, link);
1603 		if (bserrno == 0) {
1604 			spdk_bs_user_op_execute(op);
1605 		} else {
1606 			spdk_bs_user_op_abort(op);
1607 		}
1608 	}
1609 
1610 	spdk_free(ctx->buf);
1611 	free(ctx);
1612 }
1613 
1614 static void
1615 _spdk_blob_insert_cluster_cpl(void *cb_arg, int bserrno)
1616 {
1617 	struct spdk_blob_copy_cluster_ctx *ctx = cb_arg;
1618 
1619 	if (bserrno) {
1620 		if (bserrno == -EEXIST) {
1621 			/* The metadata insert failed because another thread
1622 			 * allocated the cluster first. Free our cluster
1623 			 * but continue without error. */
1624 			bserrno = 0;
1625 		}
1626 		_spdk_bs_release_cluster(ctx->blob->bs, ctx->new_cluster);
1627 	}
1628 
1629 	spdk_bs_sequence_finish(ctx->seq, bserrno);
1630 }
1631 
1632 static void
1633 _spdk_blob_write_copy_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1634 {
1635 	struct spdk_blob_copy_cluster_ctx *ctx = cb_arg;
1636 	uint32_t cluster_number;
1637 
1638 	if (bserrno) {
1639 		/* The write failed, so jump to the final completion handler */
1640 		spdk_bs_sequence_finish(seq, bserrno);
1641 		return;
1642 	}
1643 
1644 	cluster_number = _spdk_bs_page_to_cluster(ctx->blob->bs, ctx->page);
1645 
1646 	_spdk_blob_insert_cluster_on_md_thread(ctx->blob, cluster_number, ctx->new_cluster,
1647 					       _spdk_blob_insert_cluster_cpl, ctx);
1648 }
1649 
1650 static void
1651 _spdk_blob_write_copy(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1652 {
1653 	struct spdk_blob_copy_cluster_ctx *ctx = cb_arg;
1654 
1655 	if (bserrno != 0) {
1656 		/* The read failed, so jump to the final completion handler */
1657 		spdk_bs_sequence_finish(seq, bserrno);
1658 		return;
1659 	}
1660 
1661 	/* Write whole cluster */
1662 	spdk_bs_sequence_write_dev(seq, ctx->buf,
1663 				   _spdk_bs_cluster_to_lba(ctx->blob->bs, ctx->new_cluster),
1664 				   _spdk_bs_cluster_to_lba(ctx->blob->bs, 1),
1665 				   _spdk_blob_write_copy_cpl, ctx);
1666 }
1667 
1668 static void
1669 _spdk_bs_allocate_and_copy_cluster(struct spdk_blob *blob,
1670 				   struct spdk_io_channel *_ch,
1671 				   uint64_t io_unit, spdk_bs_user_op_t *op)
1672 {
1673 	struct spdk_bs_cpl cpl;
1674 	struct spdk_bs_channel *ch;
1675 	struct spdk_blob_copy_cluster_ctx *ctx;
1676 	uint32_t cluster_start_page;
1677 	uint32_t cluster_number;
1678 	int rc;
1679 
1680 	ch = spdk_io_channel_get_ctx(_ch);
1681 
1682 	if (!TAILQ_EMPTY(&ch->need_cluster_alloc)) {
1683 		/* There are already operations pending. Queue this user op
1684 		 * and return because it will be re-executed when the outstanding
1685 		 * cluster allocation completes. */
1686 		TAILQ_INSERT_TAIL(&ch->need_cluster_alloc, op, link);
1687 		return;
1688 	}
1689 
1690 	/* Round the io_unit offset down to the first page in the cluster */
1691 	cluster_start_page = _spdk_bs_io_unit_to_cluster_start(blob, io_unit);
1692 
1693 	/* Calculate which index in the metadata cluster array the corresponding
1694 	 * cluster is supposed to be at. */
1695 	cluster_number = _spdk_bs_io_unit_to_cluster_number(blob, io_unit);
1696 
1697 	ctx = calloc(1, sizeof(*ctx));
1698 	if (!ctx) {
1699 		spdk_bs_user_op_abort(op);
1700 		return;
1701 	}
1702 
1703 	assert(blob->bs->cluster_sz % blob->back_bs_dev->blocklen == 0);
1704 
1705 	ctx->blob = blob;
1706 	ctx->page = cluster_start_page;
1707 
1708 	if (blob->parent_id != SPDK_BLOBID_INVALID) {
1709 		ctx->buf = spdk_malloc(blob->bs->cluster_sz, blob->back_bs_dev->blocklen,
1710 				       NULL, SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
1711 		if (!ctx->buf) {
1712 			SPDK_ERRLOG("DMA allocation for cluster of size = %" PRIu32 " failed.\n",
1713 				    blob->bs->cluster_sz);
1714 			free(ctx);
1715 			spdk_bs_user_op_abort(op);
1716 			return;
1717 		}
1718 	}
1719 
1720 	rc = _spdk_bs_allocate_cluster(blob, cluster_number, &ctx->new_cluster, false);
1721 	if (rc != 0) {
1722 		spdk_free(ctx->buf);
1723 		free(ctx);
1724 		spdk_bs_user_op_abort(op);
1725 		return;
1726 	}
1727 
1728 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
1729 	cpl.u.blob_basic.cb_fn = _spdk_blob_allocate_and_copy_cluster_cpl;
1730 	cpl.u.blob_basic.cb_arg = ctx;
1731 
1732 	ctx->seq = spdk_bs_sequence_start(_ch, &cpl);
1733 	if (!ctx->seq) {
1734 		_spdk_bs_release_cluster(blob->bs, ctx->new_cluster);
1735 		spdk_free(ctx->buf);
1736 		free(ctx);
1737 		spdk_bs_user_op_abort(op);
1738 		return;
1739 	}
1740 
1741 	/* Queue the user op to block other incoming operations */
1742 	TAILQ_INSERT_TAIL(&ch->need_cluster_alloc, op, link);
1743 
1744 	if (blob->parent_id != SPDK_BLOBID_INVALID) {
1745 		/* Read cluster from backing device */
1746 		spdk_bs_sequence_read_bs_dev(ctx->seq, blob->back_bs_dev, ctx->buf,
1747 					     _spdk_bs_dev_page_to_lba(blob->back_bs_dev, cluster_start_page),
1748 					     _spdk_bs_dev_byte_to_lba(blob->back_bs_dev, blob->bs->cluster_sz),
1749 					     _spdk_blob_write_copy, ctx);
1750 	} else {
1751 		_spdk_blob_insert_cluster_on_md_thread(ctx->blob, cluster_number, ctx->new_cluster,
1752 						       _spdk_blob_insert_cluster_cpl, ctx);
1753 	}
1754 }
1755 
1756 static void
1757 _spdk_blob_calculate_lba_and_lba_count(struct spdk_blob *blob, uint64_t io_unit, uint64_t length,
1758 				       uint64_t *lba,	uint32_t *lba_count)
1759 {
1760 	*lba_count = length;
1761 
1762 	if (!_spdk_bs_io_unit_is_allocated(blob, io_unit)) {
1763 		assert(blob->back_bs_dev != NULL);
1764 		*lba = _spdk_bs_io_unit_to_back_dev_lba(blob, io_unit);
1765 		*lba_count = _spdk_bs_io_unit_to_back_dev_lba(blob, *lba_count);
1766 	} else {
1767 		*lba = _spdk_bs_blob_io_unit_to_lba(blob, io_unit);
1768 	}
1769 }
1770 
1771 struct op_split_ctx {
1772 	struct spdk_blob *blob;
1773 	struct spdk_io_channel *channel;
1774 	uint64_t io_unit_offset;
1775 	uint64_t io_units_remaining;
1776 	void *curr_payload;
1777 	enum spdk_blob_op_type op_type;
1778 	spdk_bs_sequence_t *seq;
1779 };
1780 
1781 static void
1782 _spdk_blob_request_submit_op_split_next(void *cb_arg, int bserrno)
1783 {
1784 	struct op_split_ctx	*ctx = cb_arg;
1785 	struct spdk_blob	*blob = ctx->blob;
1786 	struct spdk_io_channel	*ch = ctx->channel;
1787 	enum spdk_blob_op_type	op_type = ctx->op_type;
1788 	uint8_t			*buf = ctx->curr_payload;
1789 	uint64_t		offset = ctx->io_unit_offset;
1790 	uint64_t		length = ctx->io_units_remaining;
1791 	uint64_t		op_length;
1792 
1793 	if (bserrno != 0 || ctx->io_units_remaining == 0) {
1794 		spdk_bs_sequence_finish(ctx->seq, bserrno);
1795 		free(ctx);
1796 		return;
1797 	}
1798 
1799 	op_length = spdk_min(length, _spdk_bs_num_io_units_to_cluster_boundary(blob,
1800 			     offset));
1801 
1802 	/* Update length and payload for next operation */
1803 	ctx->io_units_remaining -= op_length;
1804 	ctx->io_unit_offset += op_length;
1805 	if (op_type == SPDK_BLOB_WRITE || op_type == SPDK_BLOB_READ) {
1806 		ctx->curr_payload += op_length * blob->bs->io_unit_size;
1807 	}
1808 
1809 	switch (op_type) {
1810 	case SPDK_BLOB_READ:
1811 		spdk_blob_io_read(blob, ch, buf, offset, op_length,
1812 				  _spdk_blob_request_submit_op_split_next, ctx);
1813 		break;
1814 	case SPDK_BLOB_WRITE:
1815 		spdk_blob_io_write(blob, ch, buf, offset, op_length,
1816 				   _spdk_blob_request_submit_op_split_next, ctx);
1817 		break;
1818 	case SPDK_BLOB_UNMAP:
1819 		spdk_blob_io_unmap(blob, ch, offset, op_length,
1820 				   _spdk_blob_request_submit_op_split_next, ctx);
1821 		break;
1822 	case SPDK_BLOB_WRITE_ZEROES:
1823 		spdk_blob_io_write_zeroes(blob, ch, offset, op_length,
1824 					  _spdk_blob_request_submit_op_split_next, ctx);
1825 		break;
1826 	case SPDK_BLOB_READV:
1827 	case SPDK_BLOB_WRITEV:
1828 		SPDK_ERRLOG("readv/write not valid\n");
1829 		spdk_bs_sequence_finish(ctx->seq, -EINVAL);
1830 		free(ctx);
1831 		break;
1832 	}
1833 }
1834 
1835 static void
1836 _spdk_blob_request_submit_op_split(struct spdk_io_channel *ch, struct spdk_blob *blob,
1837 				   void *payload, uint64_t offset, uint64_t length,
1838 				   spdk_blob_op_complete cb_fn, void *cb_arg, enum spdk_blob_op_type op_type)
1839 {
1840 	struct op_split_ctx *ctx;
1841 	spdk_bs_sequence_t *seq;
1842 	struct spdk_bs_cpl cpl;
1843 
1844 	assert(blob != NULL);
1845 
1846 	ctx = calloc(1, sizeof(struct op_split_ctx));
1847 	if (ctx == NULL) {
1848 		cb_fn(cb_arg, -ENOMEM);
1849 		return;
1850 	}
1851 
1852 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
1853 	cpl.u.blob_basic.cb_fn = cb_fn;
1854 	cpl.u.blob_basic.cb_arg = cb_arg;
1855 
1856 	seq = spdk_bs_sequence_start(ch, &cpl);
1857 	if (!seq) {
1858 		free(ctx);
1859 		cb_fn(cb_arg, -ENOMEM);
1860 		return;
1861 	}
1862 
1863 	ctx->blob = blob;
1864 	ctx->channel = ch;
1865 	ctx->curr_payload = payload;
1866 	ctx->io_unit_offset = offset;
1867 	ctx->io_units_remaining = length;
1868 	ctx->op_type = op_type;
1869 	ctx->seq = seq;
1870 
1871 	_spdk_blob_request_submit_op_split_next(ctx, 0);
1872 }
1873 
1874 static void
1875 _spdk_blob_request_submit_op_single(struct spdk_io_channel *_ch, struct spdk_blob *blob,
1876 				    void *payload, uint64_t offset, uint64_t length,
1877 				    spdk_blob_op_complete cb_fn, void *cb_arg, enum spdk_blob_op_type op_type)
1878 {
1879 	struct spdk_bs_cpl cpl;
1880 	uint64_t lba;
1881 	uint32_t lba_count;
1882 
1883 	assert(blob != NULL);
1884 
1885 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
1886 	cpl.u.blob_basic.cb_fn = cb_fn;
1887 	cpl.u.blob_basic.cb_arg = cb_arg;
1888 
1889 	_spdk_blob_calculate_lba_and_lba_count(blob, offset, length, &lba, &lba_count);
1890 
1891 	if (blob->frozen_refcnt) {
1892 		/* This blob I/O is frozen */
1893 		spdk_bs_user_op_t *op;
1894 		struct spdk_bs_channel *bs_channel = spdk_io_channel_get_ctx(_ch);
1895 
1896 		op = spdk_bs_user_op_alloc(_ch, &cpl, op_type, blob, payload, 0, offset, length);
1897 		if (!op) {
1898 			cb_fn(cb_arg, -ENOMEM);
1899 			return;
1900 		}
1901 
1902 		TAILQ_INSERT_TAIL(&bs_channel->queued_io, op, link);
1903 
1904 		return;
1905 	}
1906 
1907 	switch (op_type) {
1908 	case SPDK_BLOB_READ: {
1909 		spdk_bs_batch_t *batch;
1910 
1911 		batch = spdk_bs_batch_open(_ch, &cpl);
1912 		if (!batch) {
1913 			cb_fn(cb_arg, -ENOMEM);
1914 			return;
1915 		}
1916 
1917 		if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
1918 			/* Read from the blob */
1919 			spdk_bs_batch_read_dev(batch, payload, lba, lba_count);
1920 		} else {
1921 			/* Read from the backing block device */
1922 			spdk_bs_batch_read_bs_dev(batch, blob->back_bs_dev, payload, lba, lba_count);
1923 		}
1924 
1925 		spdk_bs_batch_close(batch);
1926 		break;
1927 	}
1928 	case SPDK_BLOB_WRITE:
1929 	case SPDK_BLOB_WRITE_ZEROES: {
1930 		if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
1931 			/* Write to the blob */
1932 			spdk_bs_batch_t *batch;
1933 
1934 			if (lba_count == 0) {
1935 				cb_fn(cb_arg, 0);
1936 				return;
1937 			}
1938 
1939 			batch = spdk_bs_batch_open(_ch, &cpl);
1940 			if (!batch) {
1941 				cb_fn(cb_arg, -ENOMEM);
1942 				return;
1943 			}
1944 
1945 			if (op_type == SPDK_BLOB_WRITE) {
1946 				spdk_bs_batch_write_dev(batch, payload, lba, lba_count);
1947 			} else {
1948 				spdk_bs_batch_write_zeroes_dev(batch, lba, lba_count);
1949 			}
1950 
1951 			spdk_bs_batch_close(batch);
1952 		} else {
1953 			/* Queue this operation and allocate the cluster */
1954 			spdk_bs_user_op_t *op;
1955 
1956 			op = spdk_bs_user_op_alloc(_ch, &cpl, op_type, blob, payload, 0, offset, length);
1957 			if (!op) {
1958 				cb_fn(cb_arg, -ENOMEM);
1959 				return;
1960 			}
1961 
1962 			_spdk_bs_allocate_and_copy_cluster(blob, _ch, offset, op);
1963 		}
1964 		break;
1965 	}
1966 	case SPDK_BLOB_UNMAP: {
1967 		spdk_bs_batch_t *batch;
1968 
1969 		batch = spdk_bs_batch_open(_ch, &cpl);
1970 		if (!batch) {
1971 			cb_fn(cb_arg, -ENOMEM);
1972 			return;
1973 		}
1974 
1975 		if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
1976 			spdk_bs_batch_unmap_dev(batch, lba, lba_count);
1977 		}
1978 
1979 		spdk_bs_batch_close(batch);
1980 		break;
1981 	}
1982 	case SPDK_BLOB_READV:
1983 	case SPDK_BLOB_WRITEV:
1984 		SPDK_ERRLOG("readv/write not valid\n");
1985 		cb_fn(cb_arg, -EINVAL);
1986 		break;
1987 	}
1988 }
1989 
1990 static void
1991 _spdk_blob_request_submit_op(struct spdk_blob *blob, struct spdk_io_channel *_channel,
1992 			     void *payload, uint64_t offset, uint64_t length,
1993 			     spdk_blob_op_complete cb_fn, void *cb_arg, enum spdk_blob_op_type op_type)
1994 {
1995 	assert(blob != NULL);
1996 
1997 	if (blob->data_ro && op_type != SPDK_BLOB_READ) {
1998 		cb_fn(cb_arg, -EPERM);
1999 		return;
2000 	}
2001 
2002 	if (offset + length > _spdk_bs_cluster_to_lba(blob->bs, blob->active.num_clusters)) {
2003 		cb_fn(cb_arg, -EINVAL);
2004 		return;
2005 	}
2006 	if (length <= _spdk_bs_num_io_units_to_cluster_boundary(blob, offset)) {
2007 		_spdk_blob_request_submit_op_single(_channel, blob, payload, offset, length,
2008 						    cb_fn, cb_arg, op_type);
2009 	} else {
2010 		_spdk_blob_request_submit_op_split(_channel, blob, payload, offset, length,
2011 						   cb_fn, cb_arg, op_type);
2012 	}
2013 }
2014 
2015 struct rw_iov_ctx {
2016 	struct spdk_blob *blob;
2017 	struct spdk_io_channel *channel;
2018 	spdk_blob_op_complete cb_fn;
2019 	void *cb_arg;
2020 	bool read;
2021 	int iovcnt;
2022 	struct iovec *orig_iov;
2023 	uint64_t io_unit_offset;
2024 	uint64_t io_units_remaining;
2025 	uint64_t io_units_done;
2026 	struct iovec iov[0];
2027 };
2028 
2029 static void
2030 _spdk_rw_iov_done(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
2031 {
2032 	assert(cb_arg == NULL);
2033 	spdk_bs_sequence_finish(seq, bserrno);
2034 }
2035 
2036 static void
2037 _spdk_rw_iov_split_next(void *cb_arg, int bserrno)
2038 {
2039 	struct rw_iov_ctx *ctx = cb_arg;
2040 	struct spdk_blob *blob = ctx->blob;
2041 	struct iovec *iov, *orig_iov;
2042 	int iovcnt;
2043 	size_t orig_iovoff;
2044 	uint64_t io_units_count, io_units_to_boundary, io_unit_offset;
2045 	uint64_t byte_count;
2046 
2047 	if (bserrno != 0 || ctx->io_units_remaining == 0) {
2048 		ctx->cb_fn(ctx->cb_arg, bserrno);
2049 		free(ctx);
2050 		return;
2051 	}
2052 
2053 	io_unit_offset = ctx->io_unit_offset;
2054 	io_units_to_boundary = _spdk_bs_num_io_units_to_cluster_boundary(blob, io_unit_offset);
2055 	io_units_count = spdk_min(ctx->io_units_remaining, io_units_to_boundary);
2056 	/*
2057 	 * Get index and offset into the original iov array for our current position in the I/O sequence.
2058 	 *  byte_count will keep track of how many bytes remaining until orig_iov and orig_iovoff will
2059 	 *  point to the current position in the I/O sequence.
2060 	 */
2061 	byte_count = ctx->io_units_done * blob->bs->io_unit_size;
2062 	orig_iov = &ctx->orig_iov[0];
2063 	orig_iovoff = 0;
2064 	while (byte_count > 0) {
2065 		if (byte_count >= orig_iov->iov_len) {
2066 			byte_count -= orig_iov->iov_len;
2067 			orig_iov++;
2068 		} else {
2069 			orig_iovoff = byte_count;
2070 			byte_count = 0;
2071 		}
2072 	}
2073 
2074 	/*
2075 	 * Build an iov array for the next I/O in the sequence.  byte_count will keep track of how many
2076 	 *  bytes of this next I/O remain to be accounted for in the new iov array.
2077 	 */
2078 	byte_count = io_units_count * blob->bs->io_unit_size;
2079 	iov = &ctx->iov[0];
2080 	iovcnt = 0;
2081 	while (byte_count > 0) {
2082 		assert(iovcnt < ctx->iovcnt);
2083 		iov->iov_len = spdk_min(byte_count, orig_iov->iov_len - orig_iovoff);
2084 		iov->iov_base = orig_iov->iov_base + orig_iovoff;
2085 		byte_count -= iov->iov_len;
2086 		orig_iovoff = 0;
2087 		orig_iov++;
2088 		iov++;
2089 		iovcnt++;
2090 	}
2091 
2092 	ctx->io_unit_offset += io_units_count;
2093 	ctx->io_units_remaining -= io_units_count;
2094 	ctx->io_units_done += io_units_count;
2095 	iov = &ctx->iov[0];
2096 
2097 	if (ctx->read) {
2098 		spdk_blob_io_readv(ctx->blob, ctx->channel, iov, iovcnt, io_unit_offset,
2099 				   io_units_count, _spdk_rw_iov_split_next, ctx);
2100 	} else {
2101 		spdk_blob_io_writev(ctx->blob, ctx->channel, iov, iovcnt, io_unit_offset,
2102 				    io_units_count, _spdk_rw_iov_split_next, ctx);
2103 	}
2104 }
2105 
2106 static void
2107 _spdk_blob_request_submit_rw_iov(struct spdk_blob *blob, struct spdk_io_channel *_channel,
2108 				 struct iovec *iov, int iovcnt, uint64_t offset, uint64_t length,
2109 				 spdk_blob_op_complete cb_fn, void *cb_arg, bool read)
2110 {
2111 	struct spdk_bs_cpl	cpl;
2112 
2113 	assert(blob != NULL);
2114 
2115 	if (!read && blob->data_ro) {
2116 		cb_fn(cb_arg, -EPERM);
2117 		return;
2118 	}
2119 
2120 	if (length == 0) {
2121 		cb_fn(cb_arg, 0);
2122 		return;
2123 	}
2124 
2125 	if (offset + length > _spdk_bs_cluster_to_lba(blob->bs, blob->active.num_clusters)) {
2126 		cb_fn(cb_arg, -EINVAL);
2127 		return;
2128 	}
2129 
2130 	/*
2131 	 * For now, we implement readv/writev using a sequence (instead of a batch) to account for having
2132 	 *  to split a request that spans a cluster boundary.  For I/O that do not span a cluster boundary,
2133 	 *  there will be no noticeable difference compared to using a batch.  For I/O that do span a cluster
2134 	 *  boundary, the target LBAs (after blob offset to LBA translation) may not be contiguous, so we need
2135 	 *  to allocate a separate iov array and split the I/O such that none of the resulting
2136 	 *  smaller I/O cross a cluster boundary.  These smaller I/O will be issued in sequence (not in parallel)
2137 	 *  but since this case happens very infrequently, any performance impact will be negligible.
2138 	 *
2139 	 * This could be optimized in the future to allocate a big enough iov array to account for all of the iovs
2140 	 *  for all of the smaller I/Os, pre-build all of the iov arrays for the smaller I/Os, then issue them
2141 	 *  in a batch.  That would also require creating an intermediate spdk_bs_cpl that would get called
2142 	 *  when the batch was completed, to allow for freeing the memory for the iov arrays.
2143 	 */
2144 	if (spdk_likely(length <= _spdk_bs_num_io_units_to_cluster_boundary(blob, offset))) {
2145 		uint32_t lba_count;
2146 		uint64_t lba;
2147 
2148 		cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
2149 		cpl.u.blob_basic.cb_fn = cb_fn;
2150 		cpl.u.blob_basic.cb_arg = cb_arg;
2151 
2152 		if (blob->frozen_refcnt) {
2153 			/* This blob I/O is frozen */
2154 			enum spdk_blob_op_type op_type;
2155 			spdk_bs_user_op_t *op;
2156 			struct spdk_bs_channel *bs_channel = spdk_io_channel_get_ctx(_channel);
2157 
2158 			op_type = read ? SPDK_BLOB_READV : SPDK_BLOB_WRITEV;
2159 			op = spdk_bs_user_op_alloc(_channel, &cpl, op_type, blob, iov, iovcnt, offset, length);
2160 			if (!op) {
2161 				cb_fn(cb_arg, -ENOMEM);
2162 				return;
2163 			}
2164 
2165 			TAILQ_INSERT_TAIL(&bs_channel->queued_io, op, link);
2166 
2167 			return;
2168 		}
2169 
2170 		_spdk_blob_calculate_lba_and_lba_count(blob, offset, length, &lba, &lba_count);
2171 
2172 		if (read) {
2173 			spdk_bs_sequence_t *seq;
2174 
2175 			seq = spdk_bs_sequence_start(_channel, &cpl);
2176 			if (!seq) {
2177 				cb_fn(cb_arg, -ENOMEM);
2178 				return;
2179 			}
2180 
2181 			if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
2182 				spdk_bs_sequence_readv_dev(seq, iov, iovcnt, lba, lba_count, _spdk_rw_iov_done, NULL);
2183 			} else {
2184 				spdk_bs_sequence_readv_bs_dev(seq, blob->back_bs_dev, iov, iovcnt, lba, lba_count,
2185 							      _spdk_rw_iov_done, NULL);
2186 			}
2187 		} else {
2188 			if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
2189 				spdk_bs_sequence_t *seq;
2190 
2191 				seq = spdk_bs_sequence_start(_channel, &cpl);
2192 				if (!seq) {
2193 					cb_fn(cb_arg, -ENOMEM);
2194 					return;
2195 				}
2196 
2197 				spdk_bs_sequence_writev_dev(seq, iov, iovcnt, lba, lba_count, _spdk_rw_iov_done, NULL);
2198 			} else {
2199 				/* Queue this operation and allocate the cluster */
2200 				spdk_bs_user_op_t *op;
2201 
2202 				op = spdk_bs_user_op_alloc(_channel, &cpl, SPDK_BLOB_WRITEV, blob, iov, iovcnt, offset,
2203 							   length);
2204 				if (!op) {
2205 					cb_fn(cb_arg, -ENOMEM);
2206 					return;
2207 				}
2208 
2209 				_spdk_bs_allocate_and_copy_cluster(blob, _channel, offset, op);
2210 			}
2211 		}
2212 	} else {
2213 		struct rw_iov_ctx *ctx;
2214 
2215 		ctx = calloc(1, sizeof(struct rw_iov_ctx) + iovcnt * sizeof(struct iovec));
2216 		if (ctx == NULL) {
2217 			cb_fn(cb_arg, -ENOMEM);
2218 			return;
2219 		}
2220 
2221 		ctx->blob = blob;
2222 		ctx->channel = _channel;
2223 		ctx->cb_fn = cb_fn;
2224 		ctx->cb_arg = cb_arg;
2225 		ctx->read = read;
2226 		ctx->orig_iov = iov;
2227 		ctx->iovcnt = iovcnt;
2228 		ctx->io_unit_offset = offset;
2229 		ctx->io_units_remaining = length;
2230 		ctx->io_units_done = 0;
2231 
2232 		_spdk_rw_iov_split_next(ctx, 0);
2233 	}
2234 }
2235 
2236 static struct spdk_blob *
2237 _spdk_blob_lookup(struct spdk_blob_store *bs, spdk_blob_id blobid)
2238 {
2239 	struct spdk_blob *blob;
2240 
2241 	TAILQ_FOREACH(blob, &bs->blobs, link) {
2242 		if (blob->id == blobid) {
2243 			return blob;
2244 		}
2245 	}
2246 
2247 	return NULL;
2248 }
2249 
2250 static void
2251 _spdk_blob_get_snapshot_and_clone_entries(struct spdk_blob *blob,
2252 		struct spdk_blob_list **snapshot_entry, struct spdk_blob_list **clone_entry)
2253 {
2254 	assert(blob != NULL);
2255 	*snapshot_entry = NULL;
2256 	*clone_entry = NULL;
2257 
2258 	if (blob->parent_id == SPDK_BLOBID_INVALID) {
2259 		return;
2260 	}
2261 
2262 	TAILQ_FOREACH(*snapshot_entry, &blob->bs->snapshots, link) {
2263 		if ((*snapshot_entry)->id == blob->parent_id) {
2264 			break;
2265 		}
2266 	}
2267 
2268 	if (*snapshot_entry != NULL) {
2269 		TAILQ_FOREACH(*clone_entry, &(*snapshot_entry)->clones, link) {
2270 			if ((*clone_entry)->id == blob->id) {
2271 				break;
2272 			}
2273 		}
2274 
2275 		assert(clone_entry != NULL);
2276 	}
2277 }
2278 
2279 static int
2280 _spdk_bs_channel_create(void *io_device, void *ctx_buf)
2281 {
2282 	struct spdk_blob_store		*bs = io_device;
2283 	struct spdk_bs_channel		*channel = ctx_buf;
2284 	struct spdk_bs_dev		*dev;
2285 	uint32_t			max_ops = bs->max_channel_ops;
2286 	uint32_t			i;
2287 
2288 	dev = bs->dev;
2289 
2290 	channel->req_mem = calloc(max_ops, sizeof(struct spdk_bs_request_set));
2291 	if (!channel->req_mem) {
2292 		return -1;
2293 	}
2294 
2295 	TAILQ_INIT(&channel->reqs);
2296 
2297 	for (i = 0; i < max_ops; i++) {
2298 		TAILQ_INSERT_TAIL(&channel->reqs, &channel->req_mem[i], link);
2299 	}
2300 
2301 	channel->bs = bs;
2302 	channel->dev = dev;
2303 	channel->dev_channel = dev->create_channel(dev);
2304 
2305 	if (!channel->dev_channel) {
2306 		SPDK_ERRLOG("Failed to create device channel.\n");
2307 		free(channel->req_mem);
2308 		return -1;
2309 	}
2310 
2311 	TAILQ_INIT(&channel->need_cluster_alloc);
2312 	TAILQ_INIT(&channel->queued_io);
2313 
2314 	return 0;
2315 }
2316 
2317 static void
2318 _spdk_bs_channel_destroy(void *io_device, void *ctx_buf)
2319 {
2320 	struct spdk_bs_channel *channel = ctx_buf;
2321 	spdk_bs_user_op_t *op;
2322 
2323 	while (!TAILQ_EMPTY(&channel->need_cluster_alloc)) {
2324 		op = TAILQ_FIRST(&channel->need_cluster_alloc);
2325 		TAILQ_REMOVE(&channel->need_cluster_alloc, op, link);
2326 		spdk_bs_user_op_abort(op);
2327 	}
2328 
2329 	while (!TAILQ_EMPTY(&channel->queued_io)) {
2330 		op = TAILQ_FIRST(&channel->queued_io);
2331 		TAILQ_REMOVE(&channel->queued_io, op, link);
2332 		spdk_bs_user_op_abort(op);
2333 	}
2334 
2335 	free(channel->req_mem);
2336 	channel->dev->destroy_channel(channel->dev, channel->dev_channel);
2337 }
2338 
2339 static void
2340 _spdk_bs_dev_destroy(void *io_device)
2341 {
2342 	struct spdk_blob_store *bs = io_device;
2343 	struct spdk_blob	*blob, *blob_tmp;
2344 
2345 	bs->dev->destroy(bs->dev);
2346 
2347 	TAILQ_FOREACH_SAFE(blob, &bs->blobs, link, blob_tmp) {
2348 		TAILQ_REMOVE(&bs->blobs, blob, link);
2349 		_spdk_blob_free(blob);
2350 	}
2351 
2352 	pthread_mutex_destroy(&bs->used_clusters_mutex);
2353 
2354 	spdk_bit_array_free(&bs->used_blobids);
2355 	spdk_bit_array_free(&bs->used_md_pages);
2356 	spdk_bit_array_free(&bs->used_clusters);
2357 	/*
2358 	 * If this function is called for any reason except a successful unload,
2359 	 * the unload_cpl type will be NONE and this will be a nop.
2360 	 */
2361 	spdk_bs_call_cpl(&bs->unload_cpl, bs->unload_err);
2362 
2363 	free(bs);
2364 }
2365 
2366 static int
2367 _spdk_bs_blob_list_add(struct spdk_blob *blob)
2368 {
2369 	spdk_blob_id snapshot_id;
2370 	struct spdk_blob_list *snapshot_entry = NULL;
2371 	struct spdk_blob_list *clone_entry = NULL;
2372 
2373 	assert(blob != NULL);
2374 
2375 	snapshot_id = blob->parent_id;
2376 	if (snapshot_id == SPDK_BLOBID_INVALID) {
2377 		return 0;
2378 	}
2379 
2380 	snapshot_entry = _spdk_bs_get_snapshot_entry(blob->bs, snapshot_id);
2381 	if (snapshot_entry == NULL) {
2382 		/* Snapshot not found */
2383 		snapshot_entry = calloc(1, sizeof(struct spdk_blob_list));
2384 		if (snapshot_entry == NULL) {
2385 			return -ENOMEM;
2386 		}
2387 		snapshot_entry->id = snapshot_id;
2388 		TAILQ_INIT(&snapshot_entry->clones);
2389 		TAILQ_INSERT_TAIL(&blob->bs->snapshots, snapshot_entry, link);
2390 	} else {
2391 		TAILQ_FOREACH(clone_entry, &snapshot_entry->clones, link) {
2392 			if (clone_entry->id == blob->id) {
2393 				break;
2394 			}
2395 		}
2396 	}
2397 
2398 	if (clone_entry == NULL) {
2399 		/* Clone not found */
2400 		clone_entry = calloc(1, sizeof(struct spdk_blob_list));
2401 		if (clone_entry == NULL) {
2402 			return -ENOMEM;
2403 		}
2404 		clone_entry->id = blob->id;
2405 		TAILQ_INIT(&clone_entry->clones);
2406 		TAILQ_INSERT_TAIL(&snapshot_entry->clones, clone_entry, link);
2407 		snapshot_entry->clone_count++;
2408 	}
2409 
2410 	return 0;
2411 }
2412 
2413 static void
2414 _spdk_bs_blob_list_remove(struct spdk_blob *blob)
2415 {
2416 	struct spdk_blob_list *snapshot_entry = NULL;
2417 	struct spdk_blob_list *clone_entry = NULL;
2418 
2419 	_spdk_blob_get_snapshot_and_clone_entries(blob, &snapshot_entry, &clone_entry);
2420 
2421 	if (snapshot_entry == NULL) {
2422 		return;
2423 	}
2424 
2425 	blob->parent_id = SPDK_BLOBID_INVALID;
2426 	TAILQ_REMOVE(&snapshot_entry->clones, clone_entry, link);
2427 	free(clone_entry);
2428 
2429 	snapshot_entry->clone_count--;
2430 }
2431 
2432 static int
2433 _spdk_bs_blob_list_free(struct spdk_blob_store *bs)
2434 {
2435 	struct spdk_blob_list *snapshot_entry;
2436 	struct spdk_blob_list *snapshot_entry_tmp;
2437 	struct spdk_blob_list *clone_entry;
2438 	struct spdk_blob_list *clone_entry_tmp;
2439 
2440 	TAILQ_FOREACH_SAFE(snapshot_entry, &bs->snapshots, link, snapshot_entry_tmp) {
2441 		TAILQ_FOREACH_SAFE(clone_entry, &snapshot_entry->clones, link, clone_entry_tmp) {
2442 			TAILQ_REMOVE(&snapshot_entry->clones, clone_entry, link);
2443 			free(clone_entry);
2444 		}
2445 		TAILQ_REMOVE(&bs->snapshots, snapshot_entry, link);
2446 		free(snapshot_entry);
2447 	}
2448 
2449 	return 0;
2450 }
2451 
2452 static void
2453 _spdk_bs_free(struct spdk_blob_store *bs)
2454 {
2455 	_spdk_bs_blob_list_free(bs);
2456 
2457 	spdk_bs_unregister_md_thread(bs);
2458 	spdk_io_device_unregister(bs, _spdk_bs_dev_destroy);
2459 }
2460 
2461 void
2462 spdk_bs_opts_init(struct spdk_bs_opts *opts)
2463 {
2464 	opts->cluster_sz = SPDK_BLOB_OPTS_CLUSTER_SZ;
2465 	opts->num_md_pages = SPDK_BLOB_OPTS_NUM_MD_PAGES;
2466 	opts->max_md_ops = SPDK_BLOB_OPTS_MAX_MD_OPS;
2467 	opts->max_channel_ops = SPDK_BLOB_OPTS_DEFAULT_CHANNEL_OPS;
2468 	opts->clear_method = BS_CLEAR_WITH_UNMAP;
2469 	memset(&opts->bstype, 0, sizeof(opts->bstype));
2470 	opts->iter_cb_fn = NULL;
2471 	opts->iter_cb_arg = NULL;
2472 }
2473 
2474 static int
2475 _spdk_bs_opts_verify(struct spdk_bs_opts *opts)
2476 {
2477 	if (opts->cluster_sz == 0 || opts->num_md_pages == 0 || opts->max_md_ops == 0 ||
2478 	    opts->max_channel_ops == 0) {
2479 		SPDK_ERRLOG("Blobstore options cannot be set to 0\n");
2480 		return -1;
2481 	}
2482 
2483 	return 0;
2484 }
2485 
2486 static int
2487 _spdk_bs_alloc(struct spdk_bs_dev *dev, struct spdk_bs_opts *opts, struct spdk_blob_store **_bs)
2488 {
2489 	struct spdk_blob_store	*bs;
2490 	uint64_t dev_size;
2491 	int rc;
2492 
2493 	dev_size = dev->blocklen * dev->blockcnt;
2494 	if (dev_size < opts->cluster_sz) {
2495 		/* Device size cannot be smaller than cluster size of blobstore */
2496 		SPDK_INFOLOG(SPDK_LOG_BLOB, "Device size %" PRIu64 " is smaller than cluster size %" PRIu32 "\n",
2497 			     dev_size, opts->cluster_sz);
2498 		return -ENOSPC;
2499 	}
2500 	if (opts->cluster_sz < SPDK_BS_PAGE_SIZE) {
2501 		/* Cluster size cannot be smaller than page size */
2502 		SPDK_ERRLOG("Cluster size %" PRIu32 " is smaller than page size %d\n",
2503 			    opts->cluster_sz, SPDK_BS_PAGE_SIZE);
2504 		return -EINVAL;
2505 	}
2506 	bs = calloc(1, sizeof(struct spdk_blob_store));
2507 	if (!bs) {
2508 		return -ENOMEM;
2509 	}
2510 
2511 	TAILQ_INIT(&bs->blobs);
2512 	TAILQ_INIT(&bs->snapshots);
2513 	bs->dev = dev;
2514 	bs->md_thread = spdk_get_thread();
2515 	assert(bs->md_thread != NULL);
2516 
2517 	/*
2518 	 * Do not use _spdk_bs_lba_to_cluster() here since blockcnt may not be an
2519 	 *  even multiple of the cluster size.
2520 	 */
2521 	bs->cluster_sz = opts->cluster_sz;
2522 	bs->total_clusters = dev->blockcnt / (bs->cluster_sz / dev->blocklen);
2523 	bs->pages_per_cluster = bs->cluster_sz / SPDK_BS_PAGE_SIZE;
2524 	bs->num_free_clusters = bs->total_clusters;
2525 	bs->used_clusters = spdk_bit_array_create(bs->total_clusters);
2526 	bs->io_unit_size = dev->blocklen;
2527 	if (bs->used_clusters == NULL) {
2528 		free(bs);
2529 		return -ENOMEM;
2530 	}
2531 
2532 	bs->max_channel_ops = opts->max_channel_ops;
2533 	bs->super_blob = SPDK_BLOBID_INVALID;
2534 	memcpy(&bs->bstype, &opts->bstype, sizeof(opts->bstype));
2535 
2536 	/* The metadata is assumed to be at least 1 page */
2537 	bs->used_md_pages = spdk_bit_array_create(1);
2538 	bs->used_blobids = spdk_bit_array_create(0);
2539 
2540 	pthread_mutex_init(&bs->used_clusters_mutex, NULL);
2541 
2542 	spdk_io_device_register(bs, _spdk_bs_channel_create, _spdk_bs_channel_destroy,
2543 				sizeof(struct spdk_bs_channel), "blobstore");
2544 	rc = spdk_bs_register_md_thread(bs);
2545 	if (rc == -1) {
2546 		spdk_io_device_unregister(bs, NULL);
2547 		pthread_mutex_destroy(&bs->used_clusters_mutex);
2548 		spdk_bit_array_free(&bs->used_blobids);
2549 		spdk_bit_array_free(&bs->used_md_pages);
2550 		spdk_bit_array_free(&bs->used_clusters);
2551 		free(bs);
2552 		/* FIXME: this is a lie but don't know how to get a proper error code here */
2553 		return -ENOMEM;
2554 	}
2555 
2556 	*_bs = bs;
2557 	return 0;
2558 }
2559 
2560 /* START spdk_bs_load, spdk_bs_load_ctx will used for both load and unload. */
2561 
2562 struct spdk_bs_load_ctx {
2563 	struct spdk_blob_store		*bs;
2564 	struct spdk_bs_super_block	*super;
2565 
2566 	struct spdk_bs_md_mask		*mask;
2567 	bool				in_page_chain;
2568 	uint32_t			page_index;
2569 	uint32_t			cur_page;
2570 	struct spdk_blob_md_page	*page;
2571 
2572 	spdk_bs_sequence_t			*seq;
2573 	spdk_blob_op_with_handle_complete	iter_cb_fn;
2574 	void					*iter_cb_arg;
2575 	struct spdk_blob			*blob;
2576 	spdk_blob_id				blobid;
2577 };
2578 
2579 static void
2580 _spdk_bs_load_ctx_fail(spdk_bs_sequence_t *seq, struct spdk_bs_load_ctx *ctx, int bserrno)
2581 {
2582 	assert(bserrno != 0);
2583 
2584 	spdk_free(ctx->super);
2585 	spdk_bs_sequence_finish(seq, bserrno);
2586 	_spdk_bs_free(ctx->bs);
2587 	free(ctx);
2588 }
2589 
2590 static void
2591 _spdk_bs_set_mask(struct spdk_bit_array *array, struct spdk_bs_md_mask *mask)
2592 {
2593 	uint32_t i = 0;
2594 
2595 	while (true) {
2596 		i = spdk_bit_array_find_first_set(array, i);
2597 		if (i >= mask->length) {
2598 			break;
2599 		}
2600 		mask->mask[i / 8] |= 1U << (i % 8);
2601 		i++;
2602 	}
2603 }
2604 
2605 static int
2606 _spdk_bs_load_mask(struct spdk_bit_array **array_ptr, struct spdk_bs_md_mask *mask)
2607 {
2608 	struct spdk_bit_array *array;
2609 	uint32_t i;
2610 
2611 	if (spdk_bit_array_resize(array_ptr, mask->length) < 0) {
2612 		return -ENOMEM;
2613 	}
2614 
2615 	array = *array_ptr;
2616 	for (i = 0; i < mask->length; i++) {
2617 		if (mask->mask[i / 8] & (1U << (i % 8))) {
2618 			spdk_bit_array_set(array, i);
2619 		}
2620 	}
2621 
2622 	return 0;
2623 }
2624 
2625 static void
2626 _spdk_bs_write_super(spdk_bs_sequence_t *seq, struct spdk_blob_store *bs,
2627 		     struct spdk_bs_super_block *super, spdk_bs_sequence_cpl cb_fn, void *cb_arg)
2628 {
2629 	/* Update the values in the super block */
2630 	super->super_blob = bs->super_blob;
2631 	memcpy(&super->bstype, &bs->bstype, sizeof(bs->bstype));
2632 	super->crc = _spdk_blob_md_page_calc_crc(super);
2633 	spdk_bs_sequence_write_dev(seq, super, _spdk_bs_page_to_lba(bs, 0),
2634 				   _spdk_bs_byte_to_lba(bs, sizeof(*super)),
2635 				   cb_fn, cb_arg);
2636 }
2637 
2638 static void
2639 _spdk_bs_write_used_clusters(spdk_bs_sequence_t *seq, void *arg, spdk_bs_sequence_cpl cb_fn)
2640 {
2641 	struct spdk_bs_load_ctx	*ctx = arg;
2642 	uint64_t	mask_size, lba, lba_count;
2643 
2644 	/* Write out the used clusters mask */
2645 	mask_size = ctx->super->used_cluster_mask_len * SPDK_BS_PAGE_SIZE;
2646 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL,
2647 				 SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
2648 	if (!ctx->mask) {
2649 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
2650 		return;
2651 	}
2652 
2653 	ctx->mask->type = SPDK_MD_MASK_TYPE_USED_CLUSTERS;
2654 	ctx->mask->length = ctx->bs->total_clusters;
2655 	assert(ctx->mask->length == spdk_bit_array_capacity(ctx->bs->used_clusters));
2656 
2657 	_spdk_bs_set_mask(ctx->bs->used_clusters, ctx->mask);
2658 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_cluster_mask_start);
2659 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_cluster_mask_len);
2660 	spdk_bs_sequence_write_dev(seq, ctx->mask, lba, lba_count, cb_fn, arg);
2661 }
2662 
2663 static void
2664 _spdk_bs_write_used_md(spdk_bs_sequence_t *seq, void *arg, spdk_bs_sequence_cpl cb_fn)
2665 {
2666 	struct spdk_bs_load_ctx	*ctx = arg;
2667 	uint64_t	mask_size, lba, lba_count;
2668 
2669 	if (seq->bserrno) {
2670 		_spdk_bs_load_ctx_fail(seq, ctx, seq->bserrno);
2671 		return;
2672 	}
2673 
2674 	mask_size = ctx->super->used_page_mask_len * SPDK_BS_PAGE_SIZE;
2675 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL,
2676 				 SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
2677 	if (!ctx->mask) {
2678 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
2679 		return;
2680 	}
2681 
2682 	ctx->mask->type = SPDK_MD_MASK_TYPE_USED_PAGES;
2683 	ctx->mask->length = ctx->super->md_len;
2684 	assert(ctx->mask->length == spdk_bit_array_capacity(ctx->bs->used_md_pages));
2685 
2686 	_spdk_bs_set_mask(ctx->bs->used_md_pages, ctx->mask);
2687 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_page_mask_start);
2688 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_page_mask_len);
2689 	spdk_bs_sequence_write_dev(seq, ctx->mask, lba, lba_count, cb_fn, arg);
2690 }
2691 
2692 static void
2693 _spdk_bs_write_used_blobids(spdk_bs_sequence_t *seq, void *arg, spdk_bs_sequence_cpl cb_fn)
2694 {
2695 	struct spdk_bs_load_ctx	*ctx = arg;
2696 	uint64_t	mask_size, lba, lba_count;
2697 
2698 	if (ctx->super->used_blobid_mask_len == 0) {
2699 		/*
2700 		 * This is a pre-v3 on-disk format where the blobid mask does not get
2701 		 *  written to disk.
2702 		 */
2703 		cb_fn(seq, arg, 0);
2704 		return;
2705 	}
2706 
2707 	mask_size = ctx->super->used_blobid_mask_len * SPDK_BS_PAGE_SIZE;
2708 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL, SPDK_ENV_SOCKET_ID_ANY,
2709 				 SPDK_MALLOC_DMA);
2710 	if (!ctx->mask) {
2711 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
2712 		return;
2713 	}
2714 
2715 	ctx->mask->type = SPDK_MD_MASK_TYPE_USED_BLOBIDS;
2716 	ctx->mask->length = ctx->super->md_len;
2717 	assert(ctx->mask->length == spdk_bit_array_capacity(ctx->bs->used_blobids));
2718 
2719 	_spdk_bs_set_mask(ctx->bs->used_blobids, ctx->mask);
2720 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_blobid_mask_start);
2721 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_blobid_mask_len);
2722 	spdk_bs_sequence_write_dev(seq, ctx->mask, lba, lba_count, cb_fn, arg);
2723 }
2724 
2725 static void
2726 _spdk_blob_set_thin_provision(struct spdk_blob *blob)
2727 {
2728 	_spdk_blob_verify_md_op(blob);
2729 	blob->invalid_flags |= SPDK_BLOB_THIN_PROV;
2730 	blob->state = SPDK_BLOB_STATE_DIRTY;
2731 }
2732 
2733 static void
2734 _spdk_blob_set_clear_method(struct spdk_blob *blob, enum blob_clear_method clear_method)
2735 {
2736 	_spdk_blob_verify_md_op(blob);
2737 	blob->clear_method = clear_method;
2738 	blob->md_ro_flags |= (clear_method << SPDK_BLOB_CLEAR_METHOD_SHIFT);
2739 	blob->state = SPDK_BLOB_STATE_DIRTY;
2740 }
2741 
2742 static void _spdk_bs_load_iter(void *arg, struct spdk_blob *blob, int bserrno);
2743 
2744 static void
2745 _spdk_bs_delete_corrupted_blob_cpl(void *cb_arg, int bserrno)
2746 {
2747 	struct spdk_bs_load_ctx *ctx = cb_arg;
2748 	spdk_blob_id id;
2749 	int64_t page_num;
2750 
2751 	/* Iterate to next blob (we can't use spdk_bs_iter_next function as our
2752 	 * last blob has been removed */
2753 	page_num = _spdk_bs_blobid_to_page(ctx->blobid);
2754 	page_num++;
2755 	page_num = spdk_bit_array_find_first_set(ctx->bs->used_blobids, page_num);
2756 	if (page_num >= spdk_bit_array_capacity(ctx->bs->used_blobids)) {
2757 		_spdk_bs_load_iter(ctx, NULL, -ENOENT);
2758 		return;
2759 	}
2760 
2761 	id = _spdk_bs_page_to_blobid(page_num);
2762 
2763 	spdk_bs_open_blob(ctx->bs, id, _spdk_bs_load_iter, ctx);
2764 }
2765 
2766 static void
2767 _spdk_bs_delete_corrupted_close_cb(void *cb_arg, int bserrno)
2768 {
2769 	struct spdk_bs_load_ctx *ctx = cb_arg;
2770 
2771 	if (bserrno != 0) {
2772 		SPDK_ERRLOG("Failed to close corrupted blob\n");
2773 		spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2774 		return;
2775 	}
2776 
2777 	spdk_bs_delete_blob(ctx->bs, ctx->blobid, _spdk_bs_delete_corrupted_blob_cpl, ctx);
2778 }
2779 
2780 static void
2781 _spdk_bs_delete_corrupted_blob(void *cb_arg, int bserrno)
2782 {
2783 	struct spdk_bs_load_ctx *ctx = cb_arg;
2784 	uint64_t i;
2785 
2786 	if (bserrno != 0) {
2787 		SPDK_ERRLOG("Failed to close clone of a corrupted blob\n");
2788 		spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2789 		return;
2790 	}
2791 
2792 	/* Snapshot and clone have the same copy of cluster map at this point.
2793 	 * Let's clear cluster map for snpashot now so that it won't be cleared
2794 	 * for clone later when we remove snapshot. Also set thin provision to
2795 	 * pass data corruption check */
2796 	for (i = 0; i < ctx->blob->active.num_clusters; i++) {
2797 		ctx->blob->active.clusters[i] = 0;
2798 	}
2799 
2800 	ctx->blob->md_ro = false;
2801 
2802 	_spdk_blob_set_thin_provision(ctx->blob);
2803 
2804 	ctx->blobid = ctx->blob->id;
2805 
2806 	spdk_blob_close(ctx->blob, _spdk_bs_delete_corrupted_close_cb, ctx);
2807 }
2808 
2809 static void
2810 _spdk_bs_update_corrupted_blob(void *cb_arg, int bserrno)
2811 {
2812 	struct spdk_bs_load_ctx *ctx = cb_arg;
2813 
2814 	if (bserrno != 0) {
2815 		SPDK_ERRLOG("Failed to close clone of a corrupted blob\n");
2816 		spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2817 		return;
2818 	}
2819 
2820 	ctx->blob->md_ro = false;
2821 	_spdk_blob_remove_xattr(ctx->blob, SNAPSHOT_PENDING_REMOVAL, true);
2822 	_spdk_blob_remove_xattr(ctx->blob, SNAPSHOT_IN_PROGRESS, true);
2823 	spdk_blob_set_read_only(ctx->blob);
2824 
2825 	if (ctx->iter_cb_fn) {
2826 		ctx->iter_cb_fn(ctx->iter_cb_arg, ctx->blob, 0);
2827 	}
2828 	_spdk_bs_blob_list_add(ctx->blob);
2829 
2830 	spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2831 }
2832 
2833 static void
2834 _spdk_bs_examine_clone(void *cb_arg, struct spdk_blob *blob, int bserrno)
2835 {
2836 	struct spdk_bs_load_ctx *ctx = cb_arg;
2837 
2838 	if (bserrno != 0) {
2839 		SPDK_ERRLOG("Failed to open clone of a corrupted blob\n");
2840 		spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2841 		return;
2842 	}
2843 
2844 	if (blob->parent_id == ctx->blob->id) {
2845 		/* Power failure occured before updating clone (snapshot delete case)
2846 		 * or after updating clone (creating snapshot case) - keep snapshot */
2847 		spdk_blob_close(blob, _spdk_bs_update_corrupted_blob, ctx);
2848 	} else {
2849 		/* Power failure occured after updating clone (snapshot delete case)
2850 		 * or before updating clone (creating snapshot case) - remove snapshot */
2851 		spdk_blob_close(blob, _spdk_bs_delete_corrupted_blob, ctx);
2852 	}
2853 }
2854 
2855 static void
2856 _spdk_bs_load_iter(void *arg, struct spdk_blob *blob, int bserrno)
2857 {
2858 	struct spdk_bs_load_ctx *ctx = arg;
2859 	const void *value;
2860 	size_t len;
2861 	int rc = 0;
2862 
2863 	if (bserrno == 0) {
2864 		/* Examine blob if it is corrupted after power failure. Fix
2865 		 * the ones that can be fixed and remove any other corrupted
2866 		 * ones. If it is not corrupted just process it */
2867 		rc = _spdk_blob_get_xattr_value(blob, SNAPSHOT_PENDING_REMOVAL, &value, &len, true);
2868 		if (rc != 0) {
2869 			rc = _spdk_blob_get_xattr_value(blob, SNAPSHOT_IN_PROGRESS, &value, &len, true);
2870 			if (rc != 0) {
2871 				/* Not corrupted - process it and continue with iterating through blobs */
2872 				if (ctx->iter_cb_fn) {
2873 					ctx->iter_cb_fn(ctx->iter_cb_arg, blob, 0);
2874 				}
2875 				_spdk_bs_blob_list_add(blob);
2876 				spdk_bs_iter_next(ctx->bs, blob, _spdk_bs_load_iter, ctx);
2877 				return;
2878 			}
2879 
2880 		}
2881 
2882 		assert(len == sizeof(spdk_blob_id));
2883 
2884 		ctx->blob = blob;
2885 
2886 		/* Open clone to check if we are able to fix this blob or should we remove it */
2887 		spdk_bs_open_blob(ctx->bs, *(spdk_blob_id *)value, _spdk_bs_examine_clone, ctx);
2888 		return;
2889 	} else if (bserrno == -ENOENT) {
2890 		bserrno = 0;
2891 	} else {
2892 		/*
2893 		 * This case needs to be looked at further.  Same problem
2894 		 *  exists with applications that rely on explicit blob
2895 		 *  iteration.  We should just skip the blob that failed
2896 		 *  to load and continue on to the next one.
2897 		 */
2898 		SPDK_ERRLOG("Error in iterating blobs\n");
2899 	}
2900 
2901 	ctx->iter_cb_fn = NULL;
2902 
2903 	spdk_free(ctx->super);
2904 	spdk_free(ctx->mask);
2905 	spdk_bs_sequence_finish(ctx->seq, bserrno);
2906 	free(ctx);
2907 }
2908 
2909 static void
2910 _spdk_bs_load_complete(spdk_bs_sequence_t *seq, struct spdk_bs_load_ctx *ctx, int bserrno)
2911 {
2912 	ctx->seq = seq;
2913 	spdk_bs_iter_first(ctx->bs, _spdk_bs_load_iter, ctx);
2914 }
2915 
2916 static void
2917 _spdk_bs_load_used_blobids_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
2918 {
2919 	struct spdk_bs_load_ctx *ctx = cb_arg;
2920 	int rc;
2921 
2922 	/* The type must be correct */
2923 	assert(ctx->mask->type == SPDK_MD_MASK_TYPE_USED_BLOBIDS);
2924 
2925 	/* The length of the mask (in bits) must not be greater than
2926 	 * the length of the buffer (converted to bits) */
2927 	assert(ctx->mask->length <= (ctx->super->used_blobid_mask_len * SPDK_BS_PAGE_SIZE * 8));
2928 
2929 	/* The length of the mask must be exactly equal to the size
2930 	 * (in pages) of the metadata region */
2931 	assert(ctx->mask->length == ctx->super->md_len);
2932 
2933 	rc = _spdk_bs_load_mask(&ctx->bs->used_blobids, ctx->mask);
2934 	if (rc < 0) {
2935 		spdk_free(ctx->mask);
2936 		_spdk_bs_load_ctx_fail(seq, ctx, rc);
2937 		return;
2938 	}
2939 
2940 	_spdk_bs_load_complete(seq, ctx, bserrno);
2941 }
2942 
2943 static void
2944 _spdk_bs_load_used_clusters_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
2945 {
2946 	struct spdk_bs_load_ctx *ctx = cb_arg;
2947 	uint64_t		lba, lba_count, mask_size;
2948 	int			rc;
2949 
2950 	/* The type must be correct */
2951 	assert(ctx->mask->type == SPDK_MD_MASK_TYPE_USED_CLUSTERS);
2952 	/* The length of the mask (in bits) must not be greater than the length of the buffer (converted to bits) */
2953 	assert(ctx->mask->length <= (ctx->super->used_cluster_mask_len * sizeof(
2954 					     struct spdk_blob_md_page) * 8));
2955 	/* The length of the mask must be exactly equal to the total number of clusters */
2956 	assert(ctx->mask->length == ctx->bs->total_clusters);
2957 
2958 	rc = _spdk_bs_load_mask(&ctx->bs->used_clusters, ctx->mask);
2959 	if (rc < 0) {
2960 		spdk_free(ctx->mask);
2961 		_spdk_bs_load_ctx_fail(seq, ctx, rc);
2962 		return;
2963 	}
2964 
2965 	ctx->bs->num_free_clusters = spdk_bit_array_count_clear(ctx->bs->used_clusters);
2966 	assert(ctx->bs->num_free_clusters <= ctx->bs->total_clusters);
2967 
2968 	spdk_free(ctx->mask);
2969 
2970 	/* Read the used blobids mask */
2971 	mask_size = ctx->super->used_blobid_mask_len * SPDK_BS_PAGE_SIZE;
2972 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL, SPDK_ENV_SOCKET_ID_ANY,
2973 				 SPDK_MALLOC_DMA);
2974 	if (!ctx->mask) {
2975 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
2976 		return;
2977 	}
2978 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_blobid_mask_start);
2979 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_blobid_mask_len);
2980 	spdk_bs_sequence_read_dev(seq, ctx->mask, lba, lba_count,
2981 				  _spdk_bs_load_used_blobids_cpl, ctx);
2982 }
2983 
2984 static void
2985 _spdk_bs_load_used_pages_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
2986 {
2987 	struct spdk_bs_load_ctx *ctx = cb_arg;
2988 	uint64_t		lba, lba_count, mask_size;
2989 	int			rc;
2990 
2991 	/* The type must be correct */
2992 	assert(ctx->mask->type == SPDK_MD_MASK_TYPE_USED_PAGES);
2993 	/* The length of the mask (in bits) must not be greater than the length of the buffer (converted to bits) */
2994 	assert(ctx->mask->length <= (ctx->super->used_page_mask_len * SPDK_BS_PAGE_SIZE *
2995 				     8));
2996 	/* The length of the mask must be exactly equal to the size (in pages) of the metadata region */
2997 	assert(ctx->mask->length == ctx->super->md_len);
2998 
2999 	rc = _spdk_bs_load_mask(&ctx->bs->used_md_pages, ctx->mask);
3000 	if (rc < 0) {
3001 		spdk_free(ctx->mask);
3002 		_spdk_bs_load_ctx_fail(seq, ctx, rc);
3003 		return;
3004 	}
3005 
3006 	spdk_free(ctx->mask);
3007 
3008 	/* Read the used clusters mask */
3009 	mask_size = ctx->super->used_cluster_mask_len * SPDK_BS_PAGE_SIZE;
3010 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL, SPDK_ENV_SOCKET_ID_ANY,
3011 				 SPDK_MALLOC_DMA);
3012 	if (!ctx->mask) {
3013 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
3014 		return;
3015 	}
3016 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_cluster_mask_start);
3017 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_cluster_mask_len);
3018 	spdk_bs_sequence_read_dev(seq, ctx->mask, lba, lba_count,
3019 				  _spdk_bs_load_used_clusters_cpl, ctx);
3020 }
3021 
3022 static void
3023 _spdk_bs_load_read_used_pages(spdk_bs_sequence_t *seq, void *cb_arg)
3024 {
3025 	struct spdk_bs_load_ctx	*ctx = cb_arg;
3026 	uint64_t lba, lba_count, mask_size;
3027 
3028 	/* Read the used pages mask */
3029 	mask_size = ctx->super->used_page_mask_len * SPDK_BS_PAGE_SIZE;
3030 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL,
3031 				 SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3032 	if (!ctx->mask) {
3033 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
3034 		return;
3035 	}
3036 
3037 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_page_mask_start);
3038 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_page_mask_len);
3039 	spdk_bs_sequence_read_dev(seq, ctx->mask, lba, lba_count,
3040 				  _spdk_bs_load_used_pages_cpl, ctx);
3041 }
3042 
3043 static int
3044 _spdk_bs_load_replay_md_parse_page(const struct spdk_blob_md_page *page, struct spdk_blob_store *bs)
3045 {
3046 	struct spdk_blob_md_descriptor *desc;
3047 	size_t	cur_desc = 0;
3048 
3049 	desc = (struct spdk_blob_md_descriptor *)page->descriptors;
3050 	while (cur_desc < sizeof(page->descriptors)) {
3051 		if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_PADDING) {
3052 			if (desc->length == 0) {
3053 				/* If padding and length are 0, this terminates the page */
3054 				break;
3055 			}
3056 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE) {
3057 			struct spdk_blob_md_descriptor_extent_rle	*desc_extent_rle;
3058 			unsigned int				i, j;
3059 			unsigned int				cluster_count = 0;
3060 			uint32_t				cluster_idx;
3061 
3062 			desc_extent_rle = (struct spdk_blob_md_descriptor_extent_rle *)desc;
3063 
3064 			for (i = 0; i < desc_extent_rle->length / sizeof(desc_extent_rle->extents[0]); i++) {
3065 				for (j = 0; j < desc_extent_rle->extents[i].length; j++) {
3066 					cluster_idx = desc_extent_rle->extents[i].cluster_idx;
3067 					/*
3068 					 * cluster_idx = 0 means an unallocated cluster - don't mark that
3069 					 * in the used cluster map.
3070 					 */
3071 					if (cluster_idx != 0) {
3072 						spdk_bit_array_set(bs->used_clusters, cluster_idx + j);
3073 						if (bs->num_free_clusters == 0) {
3074 							return -ENOSPC;
3075 						}
3076 						bs->num_free_clusters--;
3077 					}
3078 					cluster_count++;
3079 				}
3080 			}
3081 			if (cluster_count == 0) {
3082 				return -EINVAL;
3083 			}
3084 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR) {
3085 			/* Skip this item */
3086 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL) {
3087 			/* Skip this item */
3088 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_FLAGS) {
3089 			/* Skip this item */
3090 		} else {
3091 			/* Error */
3092 			return -EINVAL;
3093 		}
3094 		/* Advance to the next descriptor */
3095 		cur_desc += sizeof(*desc) + desc->length;
3096 		if (cur_desc + sizeof(*desc) > sizeof(page->descriptors)) {
3097 			break;
3098 		}
3099 		desc = (struct spdk_blob_md_descriptor *)((uintptr_t)page->descriptors + cur_desc);
3100 	}
3101 	return 0;
3102 }
3103 
3104 static bool _spdk_bs_load_cur_md_page_valid(struct spdk_bs_load_ctx *ctx)
3105 {
3106 	uint32_t crc;
3107 
3108 	crc = _spdk_blob_md_page_calc_crc(ctx->page);
3109 	if (crc != ctx->page->crc) {
3110 		return false;
3111 	}
3112 
3113 	if (ctx->page->sequence_num == 0 &&
3114 	    _spdk_bs_page_to_blobid(ctx->cur_page) != ctx->page->id) {
3115 		return false;
3116 	}
3117 	return true;
3118 }
3119 
3120 static void
3121 _spdk_bs_load_replay_cur_md_page(spdk_bs_sequence_t *seq, void *cb_arg);
3122 
3123 static void
3124 _spdk_bs_load_write_used_clusters_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3125 {
3126 	struct spdk_bs_load_ctx	*ctx = cb_arg;
3127 
3128 	_spdk_bs_load_complete(seq, ctx, bserrno);
3129 }
3130 
3131 static void
3132 _spdk_bs_load_write_used_blobids_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3133 {
3134 	struct spdk_bs_load_ctx	*ctx = cb_arg;
3135 
3136 	spdk_free(ctx->mask);
3137 	ctx->mask = NULL;
3138 
3139 	_spdk_bs_write_used_clusters(seq, cb_arg, _spdk_bs_load_write_used_clusters_cpl);
3140 }
3141 
3142 static void
3143 _spdk_bs_load_write_used_pages_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3144 {
3145 	struct spdk_bs_load_ctx	*ctx = cb_arg;
3146 
3147 	spdk_free(ctx->mask);
3148 	ctx->mask = NULL;
3149 
3150 	_spdk_bs_write_used_blobids(seq, cb_arg, _spdk_bs_load_write_used_blobids_cpl);
3151 }
3152 
3153 static void
3154 _spdk_bs_load_write_used_md(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3155 {
3156 	_spdk_bs_write_used_md(seq, cb_arg, _spdk_bs_load_write_used_pages_cpl);
3157 }
3158 
3159 static void
3160 _spdk_bs_load_replay_md_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3161 {
3162 	struct spdk_bs_load_ctx *ctx = cb_arg;
3163 	uint64_t num_md_clusters;
3164 	uint64_t i;
3165 	uint32_t page_num;
3166 
3167 	if (bserrno != 0) {
3168 		_spdk_bs_load_ctx_fail(seq, ctx, bserrno);
3169 		return;
3170 	}
3171 
3172 	page_num = ctx->cur_page;
3173 	if (_spdk_bs_load_cur_md_page_valid(ctx) == true) {
3174 		if (ctx->page->sequence_num == 0 || ctx->in_page_chain == true) {
3175 			spdk_bit_array_set(ctx->bs->used_md_pages, page_num);
3176 			if (ctx->page->sequence_num == 0) {
3177 				spdk_bit_array_set(ctx->bs->used_blobids, page_num);
3178 			}
3179 			if (_spdk_bs_load_replay_md_parse_page(ctx->page, ctx->bs)) {
3180 				_spdk_bs_load_ctx_fail(seq, ctx, -EILSEQ);
3181 				return;
3182 			}
3183 			if (ctx->page->next != SPDK_INVALID_MD_PAGE) {
3184 				ctx->in_page_chain = true;
3185 				ctx->cur_page = ctx->page->next;
3186 				_spdk_bs_load_replay_cur_md_page(seq, cb_arg);
3187 				return;
3188 			}
3189 		}
3190 	}
3191 
3192 	ctx->in_page_chain = false;
3193 
3194 	do {
3195 		ctx->page_index++;
3196 	} while (spdk_bit_array_get(ctx->bs->used_md_pages, ctx->page_index) == true);
3197 
3198 	if (ctx->page_index < ctx->super->md_len) {
3199 		ctx->cur_page = ctx->page_index;
3200 		_spdk_bs_load_replay_cur_md_page(seq, cb_arg);
3201 	} else {
3202 		/* Claim all of the clusters used by the metadata */
3203 		num_md_clusters = spdk_divide_round_up(ctx->super->md_len, ctx->bs->pages_per_cluster);
3204 		for (i = 0; i < num_md_clusters; i++) {
3205 			_spdk_bs_claim_cluster(ctx->bs, i);
3206 		}
3207 		spdk_free(ctx->page);
3208 		_spdk_bs_load_write_used_md(seq, ctx, bserrno);
3209 	}
3210 }
3211 
3212 static void
3213 _spdk_bs_load_replay_cur_md_page(spdk_bs_sequence_t *seq, void *cb_arg)
3214 {
3215 	struct spdk_bs_load_ctx *ctx = cb_arg;
3216 	uint64_t lba;
3217 
3218 	assert(ctx->cur_page < ctx->super->md_len);
3219 	lba = _spdk_bs_md_page_to_lba(ctx->bs, ctx->cur_page);
3220 	spdk_bs_sequence_read_dev(seq, ctx->page, lba,
3221 				  _spdk_bs_byte_to_lba(ctx->bs, SPDK_BS_PAGE_SIZE),
3222 				  _spdk_bs_load_replay_md_cpl, ctx);
3223 }
3224 
3225 static void
3226 _spdk_bs_load_replay_md(spdk_bs_sequence_t *seq, void *cb_arg)
3227 {
3228 	struct spdk_bs_load_ctx *ctx = cb_arg;
3229 
3230 	ctx->page_index = 0;
3231 	ctx->cur_page = 0;
3232 	ctx->page = spdk_zmalloc(SPDK_BS_PAGE_SIZE, SPDK_BS_PAGE_SIZE,
3233 				 NULL, SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3234 	if (!ctx->page) {
3235 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
3236 		return;
3237 	}
3238 	_spdk_bs_load_replay_cur_md_page(seq, cb_arg);
3239 }
3240 
3241 static void
3242 _spdk_bs_recover(spdk_bs_sequence_t *seq, void *cb_arg)
3243 {
3244 	struct spdk_bs_load_ctx *ctx = cb_arg;
3245 	int		rc;
3246 
3247 	rc = spdk_bit_array_resize(&ctx->bs->used_md_pages, ctx->super->md_len);
3248 	if (rc < 0) {
3249 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
3250 		return;
3251 	}
3252 
3253 	rc = spdk_bit_array_resize(&ctx->bs->used_blobids, ctx->super->md_len);
3254 	if (rc < 0) {
3255 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
3256 		return;
3257 	}
3258 
3259 	rc = spdk_bit_array_resize(&ctx->bs->used_clusters, ctx->bs->total_clusters);
3260 	if (rc < 0) {
3261 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
3262 		return;
3263 	}
3264 
3265 	ctx->bs->num_free_clusters = ctx->bs->total_clusters;
3266 	_spdk_bs_load_replay_md(seq, cb_arg);
3267 }
3268 
3269 static void
3270 _spdk_bs_load_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3271 {
3272 	struct spdk_bs_load_ctx *ctx = cb_arg;
3273 	uint32_t	crc;
3274 	int		rc;
3275 	static const char zeros[SPDK_BLOBSTORE_TYPE_LENGTH];
3276 
3277 	if (ctx->super->version > SPDK_BS_VERSION ||
3278 	    ctx->super->version < SPDK_BS_INITIAL_VERSION) {
3279 		_spdk_bs_load_ctx_fail(seq, ctx, -EILSEQ);
3280 		return;
3281 	}
3282 
3283 	if (memcmp(ctx->super->signature, SPDK_BS_SUPER_BLOCK_SIG,
3284 		   sizeof(ctx->super->signature)) != 0) {
3285 		_spdk_bs_load_ctx_fail(seq, ctx, -EILSEQ);
3286 		return;
3287 	}
3288 
3289 	crc = _spdk_blob_md_page_calc_crc(ctx->super);
3290 	if (crc != ctx->super->crc) {
3291 		_spdk_bs_load_ctx_fail(seq, ctx, -EILSEQ);
3292 		return;
3293 	}
3294 
3295 	if (memcmp(&ctx->bs->bstype, &ctx->super->bstype, SPDK_BLOBSTORE_TYPE_LENGTH) == 0) {
3296 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Bstype matched - loading blobstore\n");
3297 	} else if (memcmp(&ctx->bs->bstype, zeros, SPDK_BLOBSTORE_TYPE_LENGTH) == 0) {
3298 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Bstype wildcard used - loading blobstore regardless bstype\n");
3299 	} else {
3300 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Unexpected bstype\n");
3301 		SPDK_LOGDUMP(SPDK_LOG_BLOB, "Expected:", ctx->bs->bstype.bstype, SPDK_BLOBSTORE_TYPE_LENGTH);
3302 		SPDK_LOGDUMP(SPDK_LOG_BLOB, "Found:", ctx->super->bstype.bstype, SPDK_BLOBSTORE_TYPE_LENGTH);
3303 		_spdk_bs_load_ctx_fail(seq, ctx, -ENXIO);
3304 		return;
3305 	}
3306 
3307 	if (ctx->super->size > ctx->bs->dev->blockcnt * ctx->bs->dev->blocklen) {
3308 		SPDK_NOTICELOG("Size mismatch, dev size: %lu, blobstore size: %lu\n",
3309 			       ctx->bs->dev->blockcnt * ctx->bs->dev->blocklen, ctx->super->size);
3310 		_spdk_bs_load_ctx_fail(seq, ctx, -EILSEQ);
3311 		return;
3312 	}
3313 
3314 	if (ctx->super->size == 0) {
3315 		ctx->super->size = ctx->bs->dev->blockcnt * ctx->bs->dev->blocklen;
3316 	}
3317 
3318 	if (ctx->super->io_unit_size == 0) {
3319 		ctx->super->io_unit_size = SPDK_BS_PAGE_SIZE;
3320 	}
3321 
3322 	/* Parse the super block */
3323 	ctx->bs->clean = 1;
3324 	ctx->bs->cluster_sz = ctx->super->cluster_size;
3325 	ctx->bs->total_clusters = ctx->super->size / ctx->super->cluster_size;
3326 	ctx->bs->pages_per_cluster = ctx->bs->cluster_sz / SPDK_BS_PAGE_SIZE;
3327 	ctx->bs->io_unit_size = ctx->super->io_unit_size;
3328 	rc = spdk_bit_array_resize(&ctx->bs->used_clusters, ctx->bs->total_clusters);
3329 	if (rc < 0) {
3330 		_spdk_bs_load_ctx_fail(seq, ctx, -ENOMEM);
3331 		return;
3332 	}
3333 	ctx->bs->md_start = ctx->super->md_start;
3334 	ctx->bs->md_len = ctx->super->md_len;
3335 	ctx->bs->total_data_clusters = ctx->bs->total_clusters - spdk_divide_round_up(
3336 					       ctx->bs->md_start + ctx->bs->md_len, ctx->bs->pages_per_cluster);
3337 	ctx->bs->super_blob = ctx->super->super_blob;
3338 	memcpy(&ctx->bs->bstype, &ctx->super->bstype, sizeof(ctx->super->bstype));
3339 
3340 	if (ctx->super->used_blobid_mask_len == 0 || ctx->super->clean == 0) {
3341 		_spdk_bs_recover(seq, ctx);
3342 	} else {
3343 		_spdk_bs_load_read_used_pages(seq, ctx);
3344 	}
3345 }
3346 
3347 void
3348 spdk_bs_load(struct spdk_bs_dev *dev, struct spdk_bs_opts *o,
3349 	     spdk_bs_op_with_handle_complete cb_fn, void *cb_arg)
3350 {
3351 	struct spdk_blob_store	*bs;
3352 	struct spdk_bs_cpl	cpl;
3353 	spdk_bs_sequence_t	*seq;
3354 	struct spdk_bs_load_ctx *ctx;
3355 	struct spdk_bs_opts	opts = {};
3356 	int err;
3357 
3358 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Loading blobstore from dev %p\n", dev);
3359 
3360 	if ((SPDK_BS_PAGE_SIZE % dev->blocklen) != 0) {
3361 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "unsupported dev block length of %d\n", dev->blocklen);
3362 		dev->destroy(dev);
3363 		cb_fn(cb_arg, NULL, -EINVAL);
3364 		return;
3365 	}
3366 
3367 	if (o) {
3368 		opts = *o;
3369 	} else {
3370 		spdk_bs_opts_init(&opts);
3371 	}
3372 
3373 	if (opts.max_md_ops == 0 || opts.max_channel_ops == 0) {
3374 		dev->destroy(dev);
3375 		cb_fn(cb_arg, NULL, -EINVAL);
3376 		return;
3377 	}
3378 
3379 	err = _spdk_bs_alloc(dev, &opts, &bs);
3380 	if (err) {
3381 		dev->destroy(dev);
3382 		cb_fn(cb_arg, NULL, err);
3383 		return;
3384 	}
3385 
3386 	ctx = calloc(1, sizeof(*ctx));
3387 	if (!ctx) {
3388 		_spdk_bs_free(bs);
3389 		cb_fn(cb_arg, NULL, -ENOMEM);
3390 		return;
3391 	}
3392 
3393 	ctx->bs = bs;
3394 	ctx->iter_cb_fn = opts.iter_cb_fn;
3395 	ctx->iter_cb_arg = opts.iter_cb_arg;
3396 
3397 	/* Allocate memory for the super block */
3398 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
3399 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3400 	if (!ctx->super) {
3401 		free(ctx);
3402 		_spdk_bs_free(bs);
3403 		cb_fn(cb_arg, NULL, -ENOMEM);
3404 		return;
3405 	}
3406 
3407 	cpl.type = SPDK_BS_CPL_TYPE_BS_HANDLE;
3408 	cpl.u.bs_handle.cb_fn = cb_fn;
3409 	cpl.u.bs_handle.cb_arg = cb_arg;
3410 	cpl.u.bs_handle.bs = bs;
3411 
3412 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
3413 	if (!seq) {
3414 		spdk_free(ctx->super);
3415 		free(ctx);
3416 		_spdk_bs_free(bs);
3417 		cb_fn(cb_arg, NULL, -ENOMEM);
3418 		return;
3419 	}
3420 
3421 	/* Read the super block */
3422 	spdk_bs_sequence_read_dev(seq, ctx->super, _spdk_bs_page_to_lba(bs, 0),
3423 				  _spdk_bs_byte_to_lba(bs, sizeof(*ctx->super)),
3424 				  _spdk_bs_load_super_cpl, ctx);
3425 }
3426 
3427 /* END spdk_bs_load */
3428 
3429 /* START spdk_bs_dump */
3430 
3431 struct spdk_bs_dump_ctx {
3432 	struct spdk_blob_store		*bs;
3433 	struct spdk_bs_super_block	*super;
3434 	uint32_t			cur_page;
3435 	struct spdk_blob_md_page	*page;
3436 	spdk_bs_sequence_t		*seq;
3437 	FILE				*fp;
3438 	spdk_bs_dump_print_xattr	print_xattr_fn;
3439 	char				xattr_name[4096];
3440 };
3441 
3442 static void
3443 _spdk_bs_dump_finish(spdk_bs_sequence_t *seq, struct spdk_bs_dump_ctx *ctx, int bserrno)
3444 {
3445 	spdk_free(ctx->super);
3446 
3447 	/*
3448 	 * We need to defer calling spdk_bs_call_cpl() until after
3449 	 * dev destruction, so tuck these away for later use.
3450 	 */
3451 	ctx->bs->unload_err = bserrno;
3452 	memcpy(&ctx->bs->unload_cpl, &seq->cpl, sizeof(struct spdk_bs_cpl));
3453 	seq->cpl.type = SPDK_BS_CPL_TYPE_NONE;
3454 
3455 	spdk_bs_sequence_finish(seq, 0);
3456 	_spdk_bs_free(ctx->bs);
3457 	free(ctx);
3458 }
3459 
3460 static void _spdk_bs_dump_read_md_page(spdk_bs_sequence_t *seq, void *cb_arg);
3461 
3462 static void
3463 _spdk_bs_dump_print_md_page(struct spdk_bs_dump_ctx *ctx)
3464 {
3465 	uint32_t page_idx = ctx->cur_page;
3466 	struct spdk_blob_md_page *page = ctx->page;
3467 	struct spdk_blob_md_descriptor *desc;
3468 	size_t cur_desc = 0;
3469 	uint32_t crc;
3470 
3471 	fprintf(ctx->fp, "=========\n");
3472 	fprintf(ctx->fp, "Metadata Page Index: %" PRIu32 " (0x%" PRIx32 ")\n", page_idx, page_idx);
3473 	fprintf(ctx->fp, "Blob ID: 0x%" PRIx64 "\n", page->id);
3474 
3475 	crc = _spdk_blob_md_page_calc_crc(page);
3476 	fprintf(ctx->fp, "CRC: 0x%" PRIx32 " (%s)\n", page->crc, crc == page->crc ? "OK" : "Mismatch");
3477 
3478 	desc = (struct spdk_blob_md_descriptor *)page->descriptors;
3479 	while (cur_desc < sizeof(page->descriptors)) {
3480 		if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_PADDING) {
3481 			if (desc->length == 0) {
3482 				/* If padding and length are 0, this terminates the page */
3483 				break;
3484 			}
3485 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE) {
3486 			struct spdk_blob_md_descriptor_extent_rle	*desc_extent_rle;
3487 			unsigned int				i;
3488 
3489 			desc_extent_rle = (struct spdk_blob_md_descriptor_extent_rle *)desc;
3490 
3491 			for (i = 0; i < desc_extent_rle->length / sizeof(desc_extent_rle->extents[0]); i++) {
3492 				if (desc_extent_rle->extents[i].cluster_idx != 0) {
3493 					fprintf(ctx->fp, "Allocated Extent - Start: %" PRIu32,
3494 						desc_extent_rle->extents[i].cluster_idx);
3495 				} else {
3496 					fprintf(ctx->fp, "Unallocated Extent - ");
3497 				}
3498 				fprintf(ctx->fp, " Length: %" PRIu32, desc_extent_rle->extents[i].length);
3499 				fprintf(ctx->fp, "\n");
3500 			}
3501 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR) {
3502 			struct spdk_blob_md_descriptor_xattr *desc_xattr;
3503 			uint32_t i;
3504 
3505 			desc_xattr = (struct spdk_blob_md_descriptor_xattr *)desc;
3506 
3507 			if (desc_xattr->length !=
3508 			    sizeof(desc_xattr->name_length) + sizeof(desc_xattr->value_length) +
3509 			    desc_xattr->name_length + desc_xattr->value_length) {
3510 			}
3511 
3512 			memcpy(ctx->xattr_name, desc_xattr->name, desc_xattr->name_length);
3513 			ctx->xattr_name[desc_xattr->name_length] = '\0';
3514 			fprintf(ctx->fp, "XATTR: name = \"%s\"\n", ctx->xattr_name);
3515 			fprintf(ctx->fp, "       value = \"");
3516 			ctx->print_xattr_fn(ctx->fp, ctx->super->bstype.bstype, ctx->xattr_name,
3517 					    (void *)((uintptr_t)desc_xattr->name + desc_xattr->name_length),
3518 					    desc_xattr->value_length);
3519 			fprintf(ctx->fp, "\"\n");
3520 			for (i = 0; i < desc_xattr->value_length; i++) {
3521 				if (i % 16 == 0) {
3522 					fprintf(ctx->fp, "               ");
3523 				}
3524 				fprintf(ctx->fp, "%02" PRIx8 " ", *((uint8_t *)desc_xattr->name + desc_xattr->name_length + i));
3525 				if ((i + 1) % 16 == 0) {
3526 					fprintf(ctx->fp, "\n");
3527 				}
3528 			}
3529 			if (i % 16 != 0) {
3530 				fprintf(ctx->fp, "\n");
3531 			}
3532 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL) {
3533 			/* TODO */
3534 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_FLAGS) {
3535 			/* TODO */
3536 		} else {
3537 			/* Error */
3538 		}
3539 		/* Advance to the next descriptor */
3540 		cur_desc += sizeof(*desc) + desc->length;
3541 		if (cur_desc + sizeof(*desc) > sizeof(page->descriptors)) {
3542 			break;
3543 		}
3544 		desc = (struct spdk_blob_md_descriptor *)((uintptr_t)page->descriptors + cur_desc);
3545 	}
3546 }
3547 
3548 static void
3549 _spdk_bs_dump_read_md_page_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3550 {
3551 	struct spdk_bs_dump_ctx *ctx = cb_arg;
3552 
3553 	if (bserrno != 0) {
3554 		_spdk_bs_dump_finish(seq, ctx, bserrno);
3555 		return;
3556 	}
3557 
3558 	if (ctx->page->id != 0) {
3559 		_spdk_bs_dump_print_md_page(ctx);
3560 	}
3561 
3562 	ctx->cur_page++;
3563 
3564 	if (ctx->cur_page < ctx->super->md_len) {
3565 		_spdk_bs_dump_read_md_page(seq, cb_arg);
3566 	} else {
3567 		spdk_free(ctx->page);
3568 		_spdk_bs_dump_finish(seq, ctx, 0);
3569 	}
3570 }
3571 
3572 static void
3573 _spdk_bs_dump_read_md_page(spdk_bs_sequence_t *seq, void *cb_arg)
3574 {
3575 	struct spdk_bs_dump_ctx *ctx = cb_arg;
3576 	uint64_t lba;
3577 
3578 	assert(ctx->cur_page < ctx->super->md_len);
3579 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->md_start + ctx->cur_page);
3580 	spdk_bs_sequence_read_dev(seq, ctx->page, lba,
3581 				  _spdk_bs_byte_to_lba(ctx->bs, SPDK_BS_PAGE_SIZE),
3582 				  _spdk_bs_dump_read_md_page_cpl, ctx);
3583 }
3584 
3585 static void
3586 _spdk_bs_dump_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3587 {
3588 	struct spdk_bs_dump_ctx *ctx = cb_arg;
3589 
3590 	fprintf(ctx->fp, "Signature: \"%.8s\" ", ctx->super->signature);
3591 	if (memcmp(ctx->super->signature, SPDK_BS_SUPER_BLOCK_SIG,
3592 		   sizeof(ctx->super->signature)) != 0) {
3593 		fprintf(ctx->fp, "(Mismatch)\n");
3594 		_spdk_bs_dump_finish(seq, ctx, bserrno);
3595 		return;
3596 	} else {
3597 		fprintf(ctx->fp, "(OK)\n");
3598 	}
3599 	fprintf(ctx->fp, "Version: %" PRIu32 "\n", ctx->super->version);
3600 	fprintf(ctx->fp, "CRC: 0x%x (%s)\n", ctx->super->crc,
3601 		(ctx->super->crc == _spdk_blob_md_page_calc_crc(ctx->super)) ? "OK" : "Mismatch");
3602 	fprintf(ctx->fp, "Blobstore Type: %.*s\n", SPDK_BLOBSTORE_TYPE_LENGTH, ctx->super->bstype.bstype);
3603 	fprintf(ctx->fp, "Cluster Size: %" PRIu32 "\n", ctx->super->cluster_size);
3604 	fprintf(ctx->fp, "Super Blob ID: ");
3605 	if (ctx->super->super_blob == SPDK_BLOBID_INVALID) {
3606 		fprintf(ctx->fp, "(None)\n");
3607 	} else {
3608 		fprintf(ctx->fp, "%" PRIu64 "\n", ctx->super->super_blob);
3609 	}
3610 	fprintf(ctx->fp, "Clean: %" PRIu32 "\n", ctx->super->clean);
3611 	fprintf(ctx->fp, "Used Metadata Page Mask Start: %" PRIu32 "\n", ctx->super->used_page_mask_start);
3612 	fprintf(ctx->fp, "Used Metadata Page Mask Length: %" PRIu32 "\n", ctx->super->used_page_mask_len);
3613 	fprintf(ctx->fp, "Used Cluster Mask Start: %" PRIu32 "\n", ctx->super->used_cluster_mask_start);
3614 	fprintf(ctx->fp, "Used Cluster Mask Length: %" PRIu32 "\n", ctx->super->used_cluster_mask_len);
3615 	fprintf(ctx->fp, "Used Blob ID Mask Start: %" PRIu32 "\n", ctx->super->used_blobid_mask_start);
3616 	fprintf(ctx->fp, "Used Blob ID Mask Length: %" PRIu32 "\n", ctx->super->used_blobid_mask_len);
3617 	fprintf(ctx->fp, "Metadata Start: %" PRIu32 "\n", ctx->super->md_start);
3618 	fprintf(ctx->fp, "Metadata Length: %" PRIu32 "\n", ctx->super->md_len);
3619 
3620 	ctx->cur_page = 0;
3621 	ctx->page = spdk_zmalloc(SPDK_BS_PAGE_SIZE, SPDK_BS_PAGE_SIZE,
3622 				 NULL, SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3623 	if (!ctx->page) {
3624 		_spdk_bs_dump_finish(seq, ctx, -ENOMEM);
3625 		return;
3626 	}
3627 	_spdk_bs_dump_read_md_page(seq, cb_arg);
3628 }
3629 
3630 void
3631 spdk_bs_dump(struct spdk_bs_dev *dev, FILE *fp, spdk_bs_dump_print_xattr print_xattr_fn,
3632 	     spdk_bs_op_complete cb_fn, void *cb_arg)
3633 {
3634 	struct spdk_blob_store	*bs;
3635 	struct spdk_bs_cpl	cpl;
3636 	spdk_bs_sequence_t	*seq;
3637 	struct spdk_bs_dump_ctx *ctx;
3638 	struct spdk_bs_opts	opts = {};
3639 	int err;
3640 
3641 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Dumping blobstore from dev %p\n", dev);
3642 
3643 	spdk_bs_opts_init(&opts);
3644 
3645 	err = _spdk_bs_alloc(dev, &opts, &bs);
3646 	if (err) {
3647 		dev->destroy(dev);
3648 		cb_fn(cb_arg, err);
3649 		return;
3650 	}
3651 
3652 	ctx = calloc(1, sizeof(*ctx));
3653 	if (!ctx) {
3654 		_spdk_bs_free(bs);
3655 		cb_fn(cb_arg, -ENOMEM);
3656 		return;
3657 	}
3658 
3659 	ctx->bs = bs;
3660 	ctx->fp = fp;
3661 	ctx->print_xattr_fn = print_xattr_fn;
3662 
3663 	/* Allocate memory for the super block */
3664 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
3665 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3666 	if (!ctx->super) {
3667 		free(ctx);
3668 		_spdk_bs_free(bs);
3669 		cb_fn(cb_arg, -ENOMEM);
3670 		return;
3671 	}
3672 
3673 	cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
3674 	cpl.u.bs_basic.cb_fn = cb_fn;
3675 	cpl.u.bs_basic.cb_arg = cb_arg;
3676 
3677 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
3678 	if (!seq) {
3679 		spdk_free(ctx->super);
3680 		free(ctx);
3681 		_spdk_bs_free(bs);
3682 		cb_fn(cb_arg, -ENOMEM);
3683 		return;
3684 	}
3685 
3686 	/* Read the super block */
3687 	spdk_bs_sequence_read_dev(seq, ctx->super, _spdk_bs_page_to_lba(bs, 0),
3688 				  _spdk_bs_byte_to_lba(bs, sizeof(*ctx->super)),
3689 				  _spdk_bs_dump_super_cpl, ctx);
3690 }
3691 
3692 /* END spdk_bs_dump */
3693 
3694 /* START spdk_bs_init */
3695 
3696 struct spdk_bs_init_ctx {
3697 	struct spdk_blob_store		*bs;
3698 	struct spdk_bs_super_block	*super;
3699 };
3700 
3701 static void
3702 _spdk_bs_init_persist_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3703 {
3704 	struct spdk_bs_init_ctx *ctx = cb_arg;
3705 
3706 	spdk_free(ctx->super);
3707 	free(ctx);
3708 
3709 	spdk_bs_sequence_finish(seq, bserrno);
3710 }
3711 
3712 static void
3713 _spdk_bs_init_trim_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3714 {
3715 	struct spdk_bs_init_ctx *ctx = cb_arg;
3716 
3717 	/* Write super block */
3718 	spdk_bs_sequence_write_dev(seq, ctx->super, _spdk_bs_page_to_lba(ctx->bs, 0),
3719 				   _spdk_bs_byte_to_lba(ctx->bs, sizeof(*ctx->super)),
3720 				   _spdk_bs_init_persist_super_cpl, ctx);
3721 }
3722 
3723 void
3724 spdk_bs_init(struct spdk_bs_dev *dev, struct spdk_bs_opts *o,
3725 	     spdk_bs_op_with_handle_complete cb_fn, void *cb_arg)
3726 {
3727 	struct spdk_bs_init_ctx *ctx;
3728 	struct spdk_blob_store	*bs;
3729 	struct spdk_bs_cpl	cpl;
3730 	spdk_bs_sequence_t	*seq;
3731 	spdk_bs_batch_t		*batch;
3732 	uint64_t		num_md_lba;
3733 	uint64_t		num_md_pages;
3734 	uint64_t		num_md_clusters;
3735 	uint32_t		i;
3736 	struct spdk_bs_opts	opts = {};
3737 	int			rc;
3738 
3739 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Initializing blobstore on dev %p\n", dev);
3740 
3741 	if ((SPDK_BS_PAGE_SIZE % dev->blocklen) != 0) {
3742 		SPDK_ERRLOG("unsupported dev block length of %d\n",
3743 			    dev->blocklen);
3744 		dev->destroy(dev);
3745 		cb_fn(cb_arg, NULL, -EINVAL);
3746 		return;
3747 	}
3748 
3749 	if (o) {
3750 		opts = *o;
3751 	} else {
3752 		spdk_bs_opts_init(&opts);
3753 	}
3754 
3755 	if (_spdk_bs_opts_verify(&opts) != 0) {
3756 		dev->destroy(dev);
3757 		cb_fn(cb_arg, NULL, -EINVAL);
3758 		return;
3759 	}
3760 
3761 	rc = _spdk_bs_alloc(dev, &opts, &bs);
3762 	if (rc) {
3763 		dev->destroy(dev);
3764 		cb_fn(cb_arg, NULL, rc);
3765 		return;
3766 	}
3767 
3768 	if (opts.num_md_pages == SPDK_BLOB_OPTS_NUM_MD_PAGES) {
3769 		/* By default, allocate 1 page per cluster.
3770 		 * Technically, this over-allocates metadata
3771 		 * because more metadata will reduce the number
3772 		 * of usable clusters. This can be addressed with
3773 		 * more complex math in the future.
3774 		 */
3775 		bs->md_len = bs->total_clusters;
3776 	} else {
3777 		bs->md_len = opts.num_md_pages;
3778 	}
3779 	rc = spdk_bit_array_resize(&bs->used_md_pages, bs->md_len);
3780 	if (rc < 0) {
3781 		_spdk_bs_free(bs);
3782 		cb_fn(cb_arg, NULL, -ENOMEM);
3783 		return;
3784 	}
3785 
3786 	rc = spdk_bit_array_resize(&bs->used_blobids, bs->md_len);
3787 	if (rc < 0) {
3788 		_spdk_bs_free(bs);
3789 		cb_fn(cb_arg, NULL, -ENOMEM);
3790 		return;
3791 	}
3792 
3793 	ctx = calloc(1, sizeof(*ctx));
3794 	if (!ctx) {
3795 		_spdk_bs_free(bs);
3796 		cb_fn(cb_arg, NULL, -ENOMEM);
3797 		return;
3798 	}
3799 
3800 	ctx->bs = bs;
3801 
3802 	/* Allocate memory for the super block */
3803 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
3804 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3805 	if (!ctx->super) {
3806 		free(ctx);
3807 		_spdk_bs_free(bs);
3808 		cb_fn(cb_arg, NULL, -ENOMEM);
3809 		return;
3810 	}
3811 	memcpy(ctx->super->signature, SPDK_BS_SUPER_BLOCK_SIG,
3812 	       sizeof(ctx->super->signature));
3813 	ctx->super->version = SPDK_BS_VERSION;
3814 	ctx->super->length = sizeof(*ctx->super);
3815 	ctx->super->super_blob = bs->super_blob;
3816 	ctx->super->clean = 0;
3817 	ctx->super->cluster_size = bs->cluster_sz;
3818 	ctx->super->io_unit_size = bs->io_unit_size;
3819 	memcpy(&ctx->super->bstype, &bs->bstype, sizeof(bs->bstype));
3820 
3821 	/* Calculate how many pages the metadata consumes at the front
3822 	 * of the disk.
3823 	 */
3824 
3825 	/* The super block uses 1 page */
3826 	num_md_pages = 1;
3827 
3828 	/* The used_md_pages mask requires 1 bit per metadata page, rounded
3829 	 * up to the nearest page, plus a header.
3830 	 */
3831 	ctx->super->used_page_mask_start = num_md_pages;
3832 	ctx->super->used_page_mask_len = spdk_divide_round_up(sizeof(struct spdk_bs_md_mask) +
3833 					 spdk_divide_round_up(bs->md_len, 8),
3834 					 SPDK_BS_PAGE_SIZE);
3835 	num_md_pages += ctx->super->used_page_mask_len;
3836 
3837 	/* The used_clusters mask requires 1 bit per cluster, rounded
3838 	 * up to the nearest page, plus a header.
3839 	 */
3840 	ctx->super->used_cluster_mask_start = num_md_pages;
3841 	ctx->super->used_cluster_mask_len = spdk_divide_round_up(sizeof(struct spdk_bs_md_mask) +
3842 					    spdk_divide_round_up(bs->total_clusters, 8),
3843 					    SPDK_BS_PAGE_SIZE);
3844 	num_md_pages += ctx->super->used_cluster_mask_len;
3845 
3846 	/* The used_blobids mask requires 1 bit per metadata page, rounded
3847 	 * up to the nearest page, plus a header.
3848 	 */
3849 	ctx->super->used_blobid_mask_start = num_md_pages;
3850 	ctx->super->used_blobid_mask_len = spdk_divide_round_up(sizeof(struct spdk_bs_md_mask) +
3851 					   spdk_divide_round_up(bs->md_len, 8),
3852 					   SPDK_BS_PAGE_SIZE);
3853 	num_md_pages += ctx->super->used_blobid_mask_len;
3854 
3855 	/* The metadata region size was chosen above */
3856 	ctx->super->md_start = bs->md_start = num_md_pages;
3857 	ctx->super->md_len = bs->md_len;
3858 	num_md_pages += bs->md_len;
3859 
3860 	num_md_lba = _spdk_bs_page_to_lba(bs, num_md_pages);
3861 
3862 	ctx->super->size = dev->blockcnt * dev->blocklen;
3863 
3864 	ctx->super->crc = _spdk_blob_md_page_calc_crc(ctx->super);
3865 
3866 	num_md_clusters = spdk_divide_round_up(num_md_pages, bs->pages_per_cluster);
3867 	if (num_md_clusters > bs->total_clusters) {
3868 		SPDK_ERRLOG("Blobstore metadata cannot use more clusters than is available, "
3869 			    "please decrease number of pages reserved for metadata "
3870 			    "or increase cluster size.\n");
3871 		spdk_free(ctx->super);
3872 		free(ctx);
3873 		_spdk_bs_free(bs);
3874 		cb_fn(cb_arg, NULL, -ENOMEM);
3875 		return;
3876 	}
3877 	/* Claim all of the clusters used by the metadata */
3878 	for (i = 0; i < num_md_clusters; i++) {
3879 		_spdk_bs_claim_cluster(bs, i);
3880 	}
3881 
3882 	bs->total_data_clusters = bs->num_free_clusters;
3883 
3884 	cpl.type = SPDK_BS_CPL_TYPE_BS_HANDLE;
3885 	cpl.u.bs_handle.cb_fn = cb_fn;
3886 	cpl.u.bs_handle.cb_arg = cb_arg;
3887 	cpl.u.bs_handle.bs = bs;
3888 
3889 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
3890 	if (!seq) {
3891 		spdk_free(ctx->super);
3892 		free(ctx);
3893 		_spdk_bs_free(bs);
3894 		cb_fn(cb_arg, NULL, -ENOMEM);
3895 		return;
3896 	}
3897 
3898 	batch = spdk_bs_sequence_to_batch(seq, _spdk_bs_init_trim_cpl, ctx);
3899 
3900 	/* Clear metadata space */
3901 	spdk_bs_batch_write_zeroes_dev(batch, 0, num_md_lba);
3902 
3903 	switch (opts.clear_method) {
3904 	case BS_CLEAR_WITH_UNMAP:
3905 		/* Trim data clusters */
3906 		spdk_bs_batch_unmap_dev(batch, num_md_lba, ctx->bs->dev->blockcnt - num_md_lba);
3907 		break;
3908 	case BS_CLEAR_WITH_WRITE_ZEROES:
3909 		/* Write_zeroes to data clusters */
3910 		spdk_bs_batch_write_zeroes_dev(batch, num_md_lba, ctx->bs->dev->blockcnt - num_md_lba);
3911 		break;
3912 	case BS_CLEAR_WITH_NONE:
3913 	default:
3914 		break;
3915 	}
3916 
3917 	spdk_bs_batch_close(batch);
3918 }
3919 
3920 /* END spdk_bs_init */
3921 
3922 /* START spdk_bs_destroy */
3923 
3924 static void
3925 _spdk_bs_destroy_trim_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3926 {
3927 	struct spdk_bs_init_ctx *ctx = cb_arg;
3928 	struct spdk_blob_store *bs = ctx->bs;
3929 
3930 	/*
3931 	 * We need to defer calling spdk_bs_call_cpl() until after
3932 	 * dev destruction, so tuck these away for later use.
3933 	 */
3934 	bs->unload_err = bserrno;
3935 	memcpy(&bs->unload_cpl, &seq->cpl, sizeof(struct spdk_bs_cpl));
3936 	seq->cpl.type = SPDK_BS_CPL_TYPE_NONE;
3937 
3938 	spdk_bs_sequence_finish(seq, bserrno);
3939 
3940 	_spdk_bs_free(bs);
3941 	free(ctx);
3942 }
3943 
3944 void
3945 spdk_bs_destroy(struct spdk_blob_store *bs, spdk_bs_op_complete cb_fn,
3946 		void *cb_arg)
3947 {
3948 	struct spdk_bs_cpl	cpl;
3949 	spdk_bs_sequence_t	*seq;
3950 	struct spdk_bs_init_ctx *ctx;
3951 
3952 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Destroying blobstore\n");
3953 
3954 	if (!TAILQ_EMPTY(&bs->blobs)) {
3955 		SPDK_ERRLOG("Blobstore still has open blobs\n");
3956 		cb_fn(cb_arg, -EBUSY);
3957 		return;
3958 	}
3959 
3960 	cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
3961 	cpl.u.bs_basic.cb_fn = cb_fn;
3962 	cpl.u.bs_basic.cb_arg = cb_arg;
3963 
3964 	ctx = calloc(1, sizeof(*ctx));
3965 	if (!ctx) {
3966 		cb_fn(cb_arg, -ENOMEM);
3967 		return;
3968 	}
3969 
3970 	ctx->bs = bs;
3971 
3972 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
3973 	if (!seq) {
3974 		free(ctx);
3975 		cb_fn(cb_arg, -ENOMEM);
3976 		return;
3977 	}
3978 
3979 	/* Write zeroes to the super block */
3980 	spdk_bs_sequence_write_zeroes_dev(seq,
3981 					  _spdk_bs_page_to_lba(bs, 0),
3982 					  _spdk_bs_byte_to_lba(bs, sizeof(struct spdk_bs_super_block)),
3983 					  _spdk_bs_destroy_trim_cpl, ctx);
3984 }
3985 
3986 /* END spdk_bs_destroy */
3987 
3988 /* START spdk_bs_unload */
3989 
3990 static void
3991 _spdk_bs_unload_write_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3992 {
3993 	struct spdk_bs_load_ctx	*ctx = cb_arg;
3994 
3995 	spdk_free(ctx->super);
3996 
3997 	/*
3998 	 * We need to defer calling spdk_bs_call_cpl() until after
3999 	 * dev destruction, so tuck these away for later use.
4000 	 */
4001 	ctx->bs->unload_err = bserrno;
4002 	memcpy(&ctx->bs->unload_cpl, &seq->cpl, sizeof(struct spdk_bs_cpl));
4003 	seq->cpl.type = SPDK_BS_CPL_TYPE_NONE;
4004 
4005 	spdk_bs_sequence_finish(seq, bserrno);
4006 
4007 	_spdk_bs_free(ctx->bs);
4008 	free(ctx);
4009 }
4010 
4011 static void
4012 _spdk_bs_unload_write_used_clusters_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4013 {
4014 	struct spdk_bs_load_ctx	*ctx = cb_arg;
4015 
4016 	spdk_free(ctx->mask);
4017 	ctx->super->clean = 1;
4018 
4019 	_spdk_bs_write_super(seq, ctx->bs, ctx->super, _spdk_bs_unload_write_super_cpl, ctx);
4020 }
4021 
4022 static void
4023 _spdk_bs_unload_write_used_blobids_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4024 {
4025 	struct spdk_bs_load_ctx	*ctx = cb_arg;
4026 
4027 	spdk_free(ctx->mask);
4028 	ctx->mask = NULL;
4029 
4030 	_spdk_bs_write_used_clusters(seq, cb_arg, _spdk_bs_unload_write_used_clusters_cpl);
4031 }
4032 
4033 static void
4034 _spdk_bs_unload_write_used_pages_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4035 {
4036 	struct spdk_bs_load_ctx	*ctx = cb_arg;
4037 
4038 	spdk_free(ctx->mask);
4039 	ctx->mask = NULL;
4040 
4041 	_spdk_bs_write_used_blobids(seq, cb_arg, _spdk_bs_unload_write_used_blobids_cpl);
4042 }
4043 
4044 static void
4045 _spdk_bs_unload_read_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4046 {
4047 	_spdk_bs_write_used_md(seq, cb_arg, _spdk_bs_unload_write_used_pages_cpl);
4048 }
4049 
4050 void
4051 spdk_bs_unload(struct spdk_blob_store *bs, spdk_bs_op_complete cb_fn, void *cb_arg)
4052 {
4053 	struct spdk_bs_cpl	cpl;
4054 	spdk_bs_sequence_t	*seq;
4055 	struct spdk_bs_load_ctx *ctx;
4056 
4057 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Syncing blobstore\n");
4058 
4059 	if (!TAILQ_EMPTY(&bs->blobs)) {
4060 		SPDK_ERRLOG("Blobstore still has open blobs\n");
4061 		cb_fn(cb_arg, -EBUSY);
4062 		return;
4063 	}
4064 
4065 	ctx = calloc(1, sizeof(*ctx));
4066 	if (!ctx) {
4067 		cb_fn(cb_arg, -ENOMEM);
4068 		return;
4069 	}
4070 
4071 	ctx->bs = bs;
4072 
4073 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
4074 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
4075 	if (!ctx->super) {
4076 		free(ctx);
4077 		cb_fn(cb_arg, -ENOMEM);
4078 		return;
4079 	}
4080 
4081 	cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
4082 	cpl.u.bs_basic.cb_fn = cb_fn;
4083 	cpl.u.bs_basic.cb_arg = cb_arg;
4084 
4085 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
4086 	if (!seq) {
4087 		spdk_free(ctx->super);
4088 		free(ctx);
4089 		cb_fn(cb_arg, -ENOMEM);
4090 		return;
4091 	}
4092 
4093 	/* Read super block */
4094 	spdk_bs_sequence_read_dev(seq, ctx->super, _spdk_bs_page_to_lba(bs, 0),
4095 				  _spdk_bs_byte_to_lba(bs, sizeof(*ctx->super)),
4096 				  _spdk_bs_unload_read_super_cpl, ctx);
4097 }
4098 
4099 /* END spdk_bs_unload */
4100 
4101 /* START spdk_bs_set_super */
4102 
4103 struct spdk_bs_set_super_ctx {
4104 	struct spdk_blob_store		*bs;
4105 	struct spdk_bs_super_block	*super;
4106 };
4107 
4108 static void
4109 _spdk_bs_set_super_write_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4110 {
4111 	struct spdk_bs_set_super_ctx	*ctx = cb_arg;
4112 
4113 	if (bserrno != 0) {
4114 		SPDK_ERRLOG("Unable to write to super block of blobstore\n");
4115 	}
4116 
4117 	spdk_free(ctx->super);
4118 
4119 	spdk_bs_sequence_finish(seq, bserrno);
4120 
4121 	free(ctx);
4122 }
4123 
4124 static void
4125 _spdk_bs_set_super_read_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4126 {
4127 	struct spdk_bs_set_super_ctx	*ctx = cb_arg;
4128 
4129 	if (bserrno != 0) {
4130 		SPDK_ERRLOG("Unable to read super block of blobstore\n");
4131 		spdk_free(ctx->super);
4132 		spdk_bs_sequence_finish(seq, bserrno);
4133 		free(ctx);
4134 		return;
4135 	}
4136 
4137 	_spdk_bs_write_super(seq, ctx->bs, ctx->super, _spdk_bs_set_super_write_cpl, ctx);
4138 }
4139 
4140 void
4141 spdk_bs_set_super(struct spdk_blob_store *bs, spdk_blob_id blobid,
4142 		  spdk_bs_op_complete cb_fn, void *cb_arg)
4143 {
4144 	struct spdk_bs_cpl		cpl;
4145 	spdk_bs_sequence_t		*seq;
4146 	struct spdk_bs_set_super_ctx	*ctx;
4147 
4148 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Setting super blob id on blobstore\n");
4149 
4150 	ctx = calloc(1, sizeof(*ctx));
4151 	if (!ctx) {
4152 		cb_fn(cb_arg, -ENOMEM);
4153 		return;
4154 	}
4155 
4156 	ctx->bs = bs;
4157 
4158 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
4159 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
4160 	if (!ctx->super) {
4161 		free(ctx);
4162 		cb_fn(cb_arg, -ENOMEM);
4163 		return;
4164 	}
4165 
4166 	cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
4167 	cpl.u.bs_basic.cb_fn = cb_fn;
4168 	cpl.u.bs_basic.cb_arg = cb_arg;
4169 
4170 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
4171 	if (!seq) {
4172 		spdk_free(ctx->super);
4173 		free(ctx);
4174 		cb_fn(cb_arg, -ENOMEM);
4175 		return;
4176 	}
4177 
4178 	bs->super_blob = blobid;
4179 
4180 	/* Read super block */
4181 	spdk_bs_sequence_read_dev(seq, ctx->super, _spdk_bs_page_to_lba(bs, 0),
4182 				  _spdk_bs_byte_to_lba(bs, sizeof(*ctx->super)),
4183 				  _spdk_bs_set_super_read_cpl, ctx);
4184 }
4185 
4186 /* END spdk_bs_set_super */
4187 
4188 void
4189 spdk_bs_get_super(struct spdk_blob_store *bs,
4190 		  spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4191 {
4192 	if (bs->super_blob == SPDK_BLOBID_INVALID) {
4193 		cb_fn(cb_arg, SPDK_BLOBID_INVALID, -ENOENT);
4194 	} else {
4195 		cb_fn(cb_arg, bs->super_blob, 0);
4196 	}
4197 }
4198 
4199 uint64_t
4200 spdk_bs_get_cluster_size(struct spdk_blob_store *bs)
4201 {
4202 	return bs->cluster_sz;
4203 }
4204 
4205 uint64_t
4206 spdk_bs_get_page_size(struct spdk_blob_store *bs)
4207 {
4208 	return SPDK_BS_PAGE_SIZE;
4209 }
4210 
4211 uint64_t
4212 spdk_bs_get_io_unit_size(struct spdk_blob_store *bs)
4213 {
4214 	return bs->io_unit_size;
4215 }
4216 
4217 uint64_t
4218 spdk_bs_free_cluster_count(struct spdk_blob_store *bs)
4219 {
4220 	return bs->num_free_clusters;
4221 }
4222 
4223 uint64_t
4224 spdk_bs_total_data_cluster_count(struct spdk_blob_store *bs)
4225 {
4226 	return bs->total_data_clusters;
4227 }
4228 
4229 static int
4230 spdk_bs_register_md_thread(struct spdk_blob_store *bs)
4231 {
4232 	bs->md_channel = spdk_get_io_channel(bs);
4233 	if (!bs->md_channel) {
4234 		SPDK_ERRLOG("Failed to get IO channel.\n");
4235 		return -1;
4236 	}
4237 
4238 	return 0;
4239 }
4240 
4241 static int
4242 spdk_bs_unregister_md_thread(struct spdk_blob_store *bs)
4243 {
4244 	spdk_put_io_channel(bs->md_channel);
4245 
4246 	return 0;
4247 }
4248 
4249 spdk_blob_id spdk_blob_get_id(struct spdk_blob *blob)
4250 {
4251 	assert(blob != NULL);
4252 
4253 	return blob->id;
4254 }
4255 
4256 uint64_t spdk_blob_get_num_pages(struct spdk_blob *blob)
4257 {
4258 	assert(blob != NULL);
4259 
4260 	return _spdk_bs_cluster_to_page(blob->bs, blob->active.num_clusters);
4261 }
4262 
4263 uint64_t spdk_blob_get_num_io_units(struct spdk_blob *blob)
4264 {
4265 	assert(blob != NULL);
4266 
4267 	return spdk_blob_get_num_pages(blob) * _spdk_bs_io_unit_per_page(blob->bs);
4268 }
4269 
4270 uint64_t spdk_blob_get_num_clusters(struct spdk_blob *blob)
4271 {
4272 	assert(blob != NULL);
4273 
4274 	return blob->active.num_clusters;
4275 }
4276 
4277 /* START spdk_bs_create_blob */
4278 
4279 static void
4280 _spdk_bs_create_blob_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4281 {
4282 	struct spdk_blob *blob = cb_arg;
4283 
4284 	_spdk_blob_free(blob);
4285 
4286 	spdk_bs_sequence_finish(seq, bserrno);
4287 }
4288 
4289 static int
4290 _spdk_blob_set_xattrs(struct spdk_blob *blob, const struct spdk_blob_xattr_opts *xattrs,
4291 		      bool internal)
4292 {
4293 	uint64_t i;
4294 	size_t value_len = 0;
4295 	int rc;
4296 	const void *value = NULL;
4297 	if (xattrs->count > 0 && xattrs->get_value == NULL) {
4298 		return -EINVAL;
4299 	}
4300 	for (i = 0; i < xattrs->count; i++) {
4301 		xattrs->get_value(xattrs->ctx, xattrs->names[i], &value, &value_len);
4302 		if (value == NULL || value_len == 0) {
4303 			return -EINVAL;
4304 		}
4305 		rc = _spdk_blob_set_xattr(blob, xattrs->names[i], value, value_len, internal);
4306 		if (rc < 0) {
4307 			return rc;
4308 		}
4309 	}
4310 	return 0;
4311 }
4312 
4313 static void
4314 _spdk_bs_create_blob(struct spdk_blob_store *bs,
4315 		     const struct spdk_blob_opts *opts,
4316 		     const struct spdk_blob_xattr_opts *internal_xattrs,
4317 		     spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4318 {
4319 	struct spdk_blob	*blob;
4320 	uint32_t		page_idx;
4321 	struct spdk_bs_cpl	cpl;
4322 	struct spdk_blob_opts	opts_default;
4323 	struct spdk_blob_xattr_opts internal_xattrs_default;
4324 	spdk_bs_sequence_t	*seq;
4325 	spdk_blob_id		id;
4326 	int rc;
4327 
4328 	assert(spdk_get_thread() == bs->md_thread);
4329 
4330 	page_idx = spdk_bit_array_find_first_clear(bs->used_md_pages, 0);
4331 	if (page_idx == UINT32_MAX) {
4332 		cb_fn(cb_arg, 0, -ENOMEM);
4333 		return;
4334 	}
4335 	spdk_bit_array_set(bs->used_blobids, page_idx);
4336 	spdk_bit_array_set(bs->used_md_pages, page_idx);
4337 
4338 	id = _spdk_bs_page_to_blobid(page_idx);
4339 
4340 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Creating blob with id %lu at page %u\n", id, page_idx);
4341 
4342 	blob = _spdk_blob_alloc(bs, id);
4343 	if (!blob) {
4344 		cb_fn(cb_arg, 0, -ENOMEM);
4345 		return;
4346 	}
4347 
4348 	if (!opts) {
4349 		spdk_blob_opts_init(&opts_default);
4350 		opts = &opts_default;
4351 	}
4352 	if (!internal_xattrs) {
4353 		_spdk_blob_xattrs_init(&internal_xattrs_default);
4354 		internal_xattrs = &internal_xattrs_default;
4355 	}
4356 
4357 	rc = _spdk_blob_set_xattrs(blob, &opts->xattrs, false);
4358 	if (rc < 0) {
4359 		_spdk_blob_free(blob);
4360 		cb_fn(cb_arg, 0, rc);
4361 		return;
4362 	}
4363 
4364 	rc = _spdk_blob_set_xattrs(blob, internal_xattrs, true);
4365 	if (rc < 0) {
4366 		_spdk_blob_free(blob);
4367 		cb_fn(cb_arg, 0, rc);
4368 		return;
4369 	}
4370 
4371 	if (opts->thin_provision) {
4372 		_spdk_blob_set_thin_provision(blob);
4373 	}
4374 
4375 	_spdk_blob_set_clear_method(blob, opts->clear_method);
4376 
4377 	rc = _spdk_blob_resize(blob, opts->num_clusters);
4378 	if (rc < 0) {
4379 		_spdk_blob_free(blob);
4380 		cb_fn(cb_arg, 0, rc);
4381 		return;
4382 	}
4383 	cpl.type = SPDK_BS_CPL_TYPE_BLOBID;
4384 	cpl.u.blobid.cb_fn = cb_fn;
4385 	cpl.u.blobid.cb_arg = cb_arg;
4386 	cpl.u.blobid.blobid = blob->id;
4387 
4388 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
4389 	if (!seq) {
4390 		_spdk_blob_free(blob);
4391 		cb_fn(cb_arg, 0, -ENOMEM);
4392 		return;
4393 	}
4394 
4395 	_spdk_blob_persist(seq, blob, _spdk_bs_create_blob_cpl, blob);
4396 }
4397 
4398 void spdk_bs_create_blob(struct spdk_blob_store *bs,
4399 			 spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4400 {
4401 	_spdk_bs_create_blob(bs, NULL, NULL, cb_fn, cb_arg);
4402 }
4403 
4404 void spdk_bs_create_blob_ext(struct spdk_blob_store *bs, const struct spdk_blob_opts *opts,
4405 			     spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4406 {
4407 	_spdk_bs_create_blob(bs, opts, NULL, cb_fn, cb_arg);
4408 }
4409 
4410 /* END spdk_bs_create_blob */
4411 
4412 /* START blob_cleanup */
4413 
4414 struct spdk_clone_snapshot_ctx {
4415 	struct spdk_bs_cpl      cpl;
4416 	int bserrno;
4417 	bool frozen;
4418 
4419 	struct spdk_io_channel *channel;
4420 
4421 	/* Current cluster for inflate operation */
4422 	uint64_t cluster;
4423 
4424 	/* For inflation force allocation of all unallocated clusters and remove
4425 	 * thin-provisioning. Otherwise only decouple parent and keep clone thin. */
4426 	bool allocate_all;
4427 
4428 	struct {
4429 		spdk_blob_id id;
4430 		struct spdk_blob *blob;
4431 	} original;
4432 	struct {
4433 		spdk_blob_id id;
4434 		struct spdk_blob *blob;
4435 	} new;
4436 
4437 	/* xattrs specified for snapshot/clones only. They have no impact on
4438 	 * the original blobs xattrs. */
4439 	const struct spdk_blob_xattr_opts *xattrs;
4440 };
4441 
4442 static void
4443 _spdk_bs_clone_snapshot_cleanup_finish(void *cb_arg, int bserrno)
4444 {
4445 	struct spdk_clone_snapshot_ctx *ctx = cb_arg;
4446 	struct spdk_bs_cpl *cpl = &ctx->cpl;
4447 
4448 	if (bserrno != 0) {
4449 		if (ctx->bserrno != 0) {
4450 			SPDK_ERRLOG("Cleanup error %d\n", bserrno);
4451 		} else {
4452 			ctx->bserrno = bserrno;
4453 		}
4454 	}
4455 
4456 	switch (cpl->type) {
4457 	case SPDK_BS_CPL_TYPE_BLOBID:
4458 		cpl->u.blobid.cb_fn(cpl->u.blobid.cb_arg, cpl->u.blobid.blobid, ctx->bserrno);
4459 		break;
4460 	case SPDK_BS_CPL_TYPE_BLOB_BASIC:
4461 		cpl->u.blob_basic.cb_fn(cpl->u.blob_basic.cb_arg, ctx->bserrno);
4462 		break;
4463 	default:
4464 		SPDK_UNREACHABLE();
4465 		break;
4466 	}
4467 
4468 	free(ctx);
4469 }
4470 
4471 static void
4472 _spdk_bs_snapshot_unfreeze_cpl(void *cb_arg, int bserrno)
4473 {
4474 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4475 	struct spdk_blob *origblob = ctx->original.blob;
4476 
4477 	if (bserrno != 0) {
4478 		if (ctx->bserrno != 0) {
4479 			SPDK_ERRLOG("Unfreeze error %d\n", bserrno);
4480 		} else {
4481 			ctx->bserrno = bserrno;
4482 		}
4483 	}
4484 
4485 	ctx->original.id = origblob->id;
4486 	origblob->locked_operation_in_progress = false;
4487 
4488 	spdk_blob_close(origblob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4489 }
4490 
4491 static void
4492 _spdk_bs_clone_snapshot_origblob_cleanup(void *cb_arg, int bserrno)
4493 {
4494 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4495 	struct spdk_blob *origblob = ctx->original.blob;
4496 
4497 	if (bserrno != 0) {
4498 		if (ctx->bserrno != 0) {
4499 			SPDK_ERRLOG("Cleanup error %d\n", bserrno);
4500 		} else {
4501 			ctx->bserrno = bserrno;
4502 		}
4503 	}
4504 
4505 	if (ctx->frozen) {
4506 		/* Unfreeze any outstanding I/O */
4507 		_spdk_blob_unfreeze_io(origblob, _spdk_bs_snapshot_unfreeze_cpl, ctx);
4508 	} else {
4509 		_spdk_bs_snapshot_unfreeze_cpl(ctx, 0);
4510 	}
4511 
4512 }
4513 
4514 static void
4515 _spdk_bs_clone_snapshot_newblob_cleanup(void *cb_arg, int bserrno)
4516 {
4517 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4518 	struct spdk_blob *newblob = ctx->new.blob;
4519 
4520 	if (bserrno != 0) {
4521 		if (ctx->bserrno != 0) {
4522 			SPDK_ERRLOG("Cleanup error %d\n", bserrno);
4523 		} else {
4524 			ctx->bserrno = bserrno;
4525 		}
4526 	}
4527 
4528 	ctx->new.id = newblob->id;
4529 	spdk_blob_close(newblob, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
4530 }
4531 
4532 /* END blob_cleanup */
4533 
4534 /* START spdk_bs_create_snapshot */
4535 
4536 static void
4537 _spdk_bs_snapshot_swap_cluster_maps(struct spdk_blob *blob1, struct spdk_blob *blob2)
4538 {
4539 	uint64_t *cluster_temp;
4540 
4541 	cluster_temp = blob1->active.clusters;
4542 	blob1->active.clusters = blob2->active.clusters;
4543 	blob2->active.clusters = cluster_temp;
4544 }
4545 
4546 static void
4547 _spdk_bs_snapshot_origblob_sync_cpl(void *cb_arg, int bserrno)
4548 {
4549 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4550 	struct spdk_blob *origblob = ctx->original.blob;
4551 	struct spdk_blob *newblob = ctx->new.blob;
4552 
4553 	if (bserrno != 0) {
4554 		_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4555 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4556 		return;
4557 	}
4558 
4559 	/* Remove metadata descriptor SNAPSHOT_IN_PROGRESS */
4560 	bserrno = _spdk_blob_remove_xattr(newblob, SNAPSHOT_IN_PROGRESS, true);
4561 	if (bserrno != 0) {
4562 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4563 		return;
4564 	}
4565 
4566 	_spdk_bs_blob_list_add(ctx->original.blob);
4567 
4568 	spdk_blob_set_read_only(newblob);
4569 
4570 	/* sync snapshot metadata */
4571 	spdk_blob_sync_md(newblob, _spdk_bs_clone_snapshot_origblob_cleanup, cb_arg);
4572 }
4573 
4574 static void
4575 _spdk_bs_snapshot_newblob_sync_cpl(void *cb_arg, int bserrno)
4576 {
4577 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4578 	struct spdk_blob *origblob = ctx->original.blob;
4579 	struct spdk_blob *newblob = ctx->new.blob;
4580 
4581 	if (bserrno != 0) {
4582 		/* return cluster map back to original */
4583 		_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4584 		_spdk_bs_clone_snapshot_newblob_cleanup(ctx, bserrno);
4585 		return;
4586 	}
4587 
4588 	/* Set internal xattr for snapshot id */
4589 	bserrno = _spdk_blob_set_xattr(origblob, BLOB_SNAPSHOT, &newblob->id, sizeof(spdk_blob_id), true);
4590 	if (bserrno != 0) {
4591 		/* return cluster map back to original */
4592 		_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4593 		_spdk_bs_clone_snapshot_newblob_cleanup(ctx, bserrno);
4594 		return;
4595 	}
4596 
4597 	_spdk_bs_blob_list_remove(origblob);
4598 	origblob->parent_id = newblob->id;
4599 
4600 	/* Create new back_bs_dev for snapshot */
4601 	origblob->back_bs_dev = spdk_bs_create_blob_bs_dev(newblob);
4602 	if (origblob->back_bs_dev == NULL) {
4603 		/* return cluster map back to original */
4604 		_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4605 		_spdk_bs_clone_snapshot_newblob_cleanup(ctx, -EINVAL);
4606 		return;
4607 	}
4608 
4609 	/* set clone blob as thin provisioned */
4610 	_spdk_blob_set_thin_provision(origblob);
4611 
4612 	_spdk_bs_blob_list_add(newblob);
4613 
4614 	/* sync clone metadata */
4615 	spdk_blob_sync_md(origblob, _spdk_bs_snapshot_origblob_sync_cpl, ctx);
4616 }
4617 
4618 static void
4619 _spdk_bs_snapshot_freeze_cpl(void *cb_arg, int rc)
4620 {
4621 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4622 	struct spdk_blob *origblob = ctx->original.blob;
4623 	struct spdk_blob *newblob = ctx->new.blob;
4624 	int bserrno;
4625 
4626 	if (rc != 0) {
4627 		_spdk_bs_clone_snapshot_newblob_cleanup(ctx, rc);
4628 		return;
4629 	}
4630 
4631 	ctx->frozen = true;
4632 
4633 	/* set new back_bs_dev for snapshot */
4634 	newblob->back_bs_dev = origblob->back_bs_dev;
4635 	/* Set invalid flags from origblob */
4636 	newblob->invalid_flags = origblob->invalid_flags;
4637 
4638 	/* inherit parent from original blob if set */
4639 	newblob->parent_id = origblob->parent_id;
4640 	if (origblob->parent_id != SPDK_BLOBID_INVALID) {
4641 		/* Set internal xattr for snapshot id */
4642 		bserrno = _spdk_blob_set_xattr(newblob, BLOB_SNAPSHOT,
4643 					       &origblob->parent_id, sizeof(spdk_blob_id), true);
4644 		if (bserrno != 0) {
4645 			_spdk_bs_clone_snapshot_newblob_cleanup(ctx, bserrno);
4646 			return;
4647 		}
4648 	}
4649 
4650 	/* swap cluster maps */
4651 	_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4652 
4653 	/* Set the clear method on the new blob to match the original. */
4654 	_spdk_blob_set_clear_method(newblob, origblob->clear_method);
4655 
4656 	/* sync snapshot metadata */
4657 	spdk_blob_sync_md(newblob, _spdk_bs_snapshot_newblob_sync_cpl, ctx);
4658 }
4659 
4660 static void
4661 _spdk_bs_snapshot_newblob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
4662 {
4663 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4664 	struct spdk_blob *origblob = ctx->original.blob;
4665 	struct spdk_blob *newblob = _blob;
4666 
4667 	if (bserrno != 0) {
4668 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4669 		return;
4670 	}
4671 
4672 	ctx->new.blob = newblob;
4673 	assert(spdk_blob_is_thin_provisioned(newblob));
4674 	assert(spdk_mem_all_zero(newblob->active.clusters,
4675 				 newblob->active.num_clusters * sizeof(*newblob->active.clusters)));
4676 
4677 	_spdk_blob_freeze_io(origblob, _spdk_bs_snapshot_freeze_cpl, ctx);
4678 }
4679 
4680 static void
4681 _spdk_bs_snapshot_newblob_create_cpl(void *cb_arg, spdk_blob_id blobid, int bserrno)
4682 {
4683 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4684 	struct spdk_blob *origblob = ctx->original.blob;
4685 
4686 	if (bserrno != 0) {
4687 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4688 		return;
4689 	}
4690 
4691 	ctx->new.id = blobid;
4692 	ctx->cpl.u.blobid.blobid = blobid;
4693 
4694 	spdk_bs_open_blob(origblob->bs, ctx->new.id, _spdk_bs_snapshot_newblob_open_cpl, ctx);
4695 }
4696 
4697 
4698 static void
4699 _spdk_bs_xattr_snapshot(void *arg, const char *name,
4700 			const void **value, size_t *value_len)
4701 {
4702 	assert(strncmp(name, SNAPSHOT_IN_PROGRESS, sizeof(SNAPSHOT_IN_PROGRESS)) == 0);
4703 
4704 	struct spdk_blob *blob = (struct spdk_blob *)arg;
4705 	*value = &blob->id;
4706 	*value_len = sizeof(blob->id);
4707 }
4708 
4709 static void
4710 _spdk_bs_snapshot_origblob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
4711 {
4712 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4713 	struct spdk_blob_opts opts;
4714 	struct spdk_blob_xattr_opts internal_xattrs;
4715 	char *xattrs_names[] = { SNAPSHOT_IN_PROGRESS };
4716 
4717 	if (bserrno != 0) {
4718 		_spdk_bs_clone_snapshot_cleanup_finish(ctx, bserrno);
4719 		return;
4720 	}
4721 
4722 	ctx->original.blob = _blob;
4723 
4724 	if (_blob->data_ro || _blob->md_ro) {
4725 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot create snapshot from read only blob with id %lu\n",
4726 			      _blob->id);
4727 		ctx->bserrno = -EINVAL;
4728 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4729 		return;
4730 	}
4731 
4732 	if (_blob->locked_operation_in_progress) {
4733 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot create snapshot - another operation in progress\n");
4734 		ctx->bserrno = -EBUSY;
4735 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4736 		return;
4737 	}
4738 
4739 	_blob->locked_operation_in_progress = true;
4740 
4741 	spdk_blob_opts_init(&opts);
4742 	_spdk_blob_xattrs_init(&internal_xattrs);
4743 
4744 	/* Change the size of new blob to the same as in original blob,
4745 	 * but do not allocate clusters */
4746 	opts.thin_provision = true;
4747 	opts.num_clusters = spdk_blob_get_num_clusters(_blob);
4748 
4749 	/* If there are any xattrs specified for snapshot, set them now */
4750 	if (ctx->xattrs) {
4751 		memcpy(&opts.xattrs, ctx->xattrs, sizeof(*ctx->xattrs));
4752 	}
4753 	/* Set internal xattr SNAPSHOT_IN_PROGRESS */
4754 	internal_xattrs.count = 1;
4755 	internal_xattrs.ctx = _blob;
4756 	internal_xattrs.names = xattrs_names;
4757 	internal_xattrs.get_value = _spdk_bs_xattr_snapshot;
4758 
4759 	_spdk_bs_create_blob(_blob->bs, &opts, &internal_xattrs,
4760 			     _spdk_bs_snapshot_newblob_create_cpl, ctx);
4761 }
4762 
4763 void spdk_bs_create_snapshot(struct spdk_blob_store *bs, spdk_blob_id blobid,
4764 			     const struct spdk_blob_xattr_opts *snapshot_xattrs,
4765 			     spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4766 {
4767 	struct spdk_clone_snapshot_ctx *ctx = calloc(1, sizeof(*ctx));
4768 
4769 	if (!ctx) {
4770 		cb_fn(cb_arg, SPDK_BLOBID_INVALID, -ENOMEM);
4771 		return;
4772 	}
4773 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BLOBID;
4774 	ctx->cpl.u.blobid.cb_fn = cb_fn;
4775 	ctx->cpl.u.blobid.cb_arg = cb_arg;
4776 	ctx->cpl.u.blobid.blobid = SPDK_BLOBID_INVALID;
4777 	ctx->bserrno = 0;
4778 	ctx->frozen = false;
4779 	ctx->original.id = blobid;
4780 	ctx->xattrs = snapshot_xattrs;
4781 
4782 	spdk_bs_open_blob(bs, ctx->original.id, _spdk_bs_snapshot_origblob_open_cpl, ctx);
4783 }
4784 /* END spdk_bs_create_snapshot */
4785 
4786 /* START spdk_bs_create_clone */
4787 
4788 static void
4789 _spdk_bs_xattr_clone(void *arg, const char *name,
4790 		     const void **value, size_t *value_len)
4791 {
4792 	assert(strncmp(name, BLOB_SNAPSHOT, sizeof(BLOB_SNAPSHOT)) == 0);
4793 
4794 	struct spdk_blob *blob = (struct spdk_blob *)arg;
4795 	*value = &blob->id;
4796 	*value_len = sizeof(blob->id);
4797 }
4798 
4799 static void
4800 _spdk_bs_clone_newblob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
4801 {
4802 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4803 	struct spdk_blob *clone = _blob;
4804 
4805 	ctx->new.blob = clone;
4806 	_spdk_bs_blob_list_add(clone);
4807 
4808 	spdk_blob_close(clone, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
4809 }
4810 
4811 static void
4812 _spdk_bs_clone_newblob_create_cpl(void *cb_arg, spdk_blob_id blobid, int bserrno)
4813 {
4814 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4815 
4816 	ctx->cpl.u.blobid.blobid = blobid;
4817 	spdk_bs_open_blob(ctx->original.blob->bs, blobid, _spdk_bs_clone_newblob_open_cpl, ctx);
4818 }
4819 
4820 static void
4821 _spdk_bs_clone_origblob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
4822 {
4823 	struct spdk_clone_snapshot_ctx	*ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4824 	struct spdk_blob_opts		opts;
4825 	struct spdk_blob_xattr_opts internal_xattrs;
4826 	char *xattr_names[] = { BLOB_SNAPSHOT };
4827 
4828 	if (bserrno != 0) {
4829 		_spdk_bs_clone_snapshot_cleanup_finish(ctx, bserrno);
4830 		return;
4831 	}
4832 
4833 	ctx->original.blob = _blob;
4834 
4835 	if (!_blob->data_ro || !_blob->md_ro) {
4836 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Clone not from read-only blob\n");
4837 		ctx->bserrno = -EINVAL;
4838 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4839 		return;
4840 	}
4841 
4842 	if (_blob->locked_operation_in_progress) {
4843 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot create clone - another operation in progress\n");
4844 		ctx->bserrno = -EBUSY;
4845 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4846 		return;
4847 	}
4848 
4849 	_blob->locked_operation_in_progress = true;
4850 
4851 	spdk_blob_opts_init(&opts);
4852 	_spdk_blob_xattrs_init(&internal_xattrs);
4853 
4854 	opts.thin_provision = true;
4855 	opts.num_clusters = spdk_blob_get_num_clusters(_blob);
4856 	if (ctx->xattrs) {
4857 		memcpy(&opts.xattrs, ctx->xattrs, sizeof(*ctx->xattrs));
4858 	}
4859 
4860 	/* Set internal xattr BLOB_SNAPSHOT */
4861 	internal_xattrs.count = 1;
4862 	internal_xattrs.ctx = _blob;
4863 	internal_xattrs.names = xattr_names;
4864 	internal_xattrs.get_value = _spdk_bs_xattr_clone;
4865 
4866 	_spdk_bs_create_blob(_blob->bs, &opts, &internal_xattrs,
4867 			     _spdk_bs_clone_newblob_create_cpl, ctx);
4868 }
4869 
4870 void spdk_bs_create_clone(struct spdk_blob_store *bs, spdk_blob_id blobid,
4871 			  const struct spdk_blob_xattr_opts *clone_xattrs,
4872 			  spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4873 {
4874 	struct spdk_clone_snapshot_ctx	*ctx = calloc(1, sizeof(*ctx));
4875 
4876 	if (!ctx) {
4877 		cb_fn(cb_arg, SPDK_BLOBID_INVALID, -ENOMEM);
4878 		return;
4879 	}
4880 
4881 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BLOBID;
4882 	ctx->cpl.u.blobid.cb_fn = cb_fn;
4883 	ctx->cpl.u.blobid.cb_arg = cb_arg;
4884 	ctx->cpl.u.blobid.blobid = SPDK_BLOBID_INVALID;
4885 	ctx->bserrno = 0;
4886 	ctx->xattrs = clone_xattrs;
4887 	ctx->original.id = blobid;
4888 
4889 	spdk_bs_open_blob(bs, ctx->original.id, _spdk_bs_clone_origblob_open_cpl, ctx);
4890 }
4891 
4892 /* END spdk_bs_create_clone */
4893 
4894 /* START spdk_bs_inflate_blob */
4895 
4896 static void
4897 _spdk_bs_inflate_blob_set_parent_cpl(void *cb_arg, struct spdk_blob *_parent, int bserrno)
4898 {
4899 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4900 	struct spdk_blob *_blob = ctx->original.blob;
4901 
4902 	if (bserrno != 0) {
4903 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4904 		return;
4905 	}
4906 
4907 	assert(_parent != NULL);
4908 
4909 	_spdk_bs_blob_list_remove(_blob);
4910 	_blob->parent_id = _parent->id;
4911 	_spdk_blob_set_xattr(_blob, BLOB_SNAPSHOT, &_blob->parent_id,
4912 			     sizeof(spdk_blob_id), true);
4913 
4914 	_blob->back_bs_dev->destroy(_blob->back_bs_dev);
4915 	_blob->back_bs_dev = spdk_bs_create_blob_bs_dev(_parent);
4916 	_spdk_bs_blob_list_add(_blob);
4917 
4918 	spdk_blob_sync_md(_blob, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
4919 }
4920 
4921 static void
4922 _spdk_bs_inflate_blob_done(void *cb_arg, int bserrno)
4923 {
4924 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4925 	struct spdk_blob *_blob = ctx->original.blob;
4926 	struct spdk_blob *_parent;
4927 
4928 	if (bserrno != 0) {
4929 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4930 		return;
4931 	}
4932 
4933 	if (ctx->allocate_all) {
4934 		/* remove thin provisioning */
4935 		_spdk_bs_blob_list_remove(_blob);
4936 		_spdk_blob_remove_xattr(_blob, BLOB_SNAPSHOT, true);
4937 		_blob->invalid_flags = _blob->invalid_flags & ~SPDK_BLOB_THIN_PROV;
4938 		_blob->back_bs_dev->destroy(_blob->back_bs_dev);
4939 		_blob->back_bs_dev = NULL;
4940 		_blob->parent_id = SPDK_BLOBID_INVALID;
4941 	} else {
4942 		_parent = ((struct spdk_blob_bs_dev *)(_blob->back_bs_dev))->blob;
4943 		if (_parent->parent_id != SPDK_BLOBID_INVALID) {
4944 			/* We must change the parent of the inflated blob */
4945 			spdk_bs_open_blob(_blob->bs, _parent->parent_id,
4946 					  _spdk_bs_inflate_blob_set_parent_cpl, ctx);
4947 			return;
4948 		}
4949 
4950 		_spdk_bs_blob_list_remove(_blob);
4951 		_spdk_blob_remove_xattr(_blob, BLOB_SNAPSHOT, true);
4952 		_blob->parent_id = SPDK_BLOBID_INVALID;
4953 		_blob->back_bs_dev->destroy(_blob->back_bs_dev);
4954 		_blob->back_bs_dev = spdk_bs_create_zeroes_dev();
4955 	}
4956 
4957 	_blob->state = SPDK_BLOB_STATE_DIRTY;
4958 	spdk_blob_sync_md(_blob, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
4959 }
4960 
4961 /* Check if cluster needs allocation */
4962 static inline bool
4963 _spdk_bs_cluster_needs_allocation(struct spdk_blob *blob, uint64_t cluster, bool allocate_all)
4964 {
4965 	struct spdk_blob_bs_dev *b;
4966 
4967 	assert(blob != NULL);
4968 
4969 	if (blob->active.clusters[cluster] != 0) {
4970 		/* Cluster is already allocated */
4971 		return false;
4972 	}
4973 
4974 	if (blob->parent_id == SPDK_BLOBID_INVALID) {
4975 		/* Blob have no parent blob */
4976 		return allocate_all;
4977 	}
4978 
4979 	b = (struct spdk_blob_bs_dev *)blob->back_bs_dev;
4980 	return (allocate_all || b->blob->active.clusters[cluster] != 0);
4981 }
4982 
4983 static void
4984 _spdk_bs_inflate_blob_touch_next(void *cb_arg, int bserrno)
4985 {
4986 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4987 	struct spdk_blob *_blob = ctx->original.blob;
4988 	uint64_t offset;
4989 
4990 	if (bserrno != 0) {
4991 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4992 		return;
4993 	}
4994 
4995 	for (; ctx->cluster < _blob->active.num_clusters; ctx->cluster++) {
4996 		if (_spdk_bs_cluster_needs_allocation(_blob, ctx->cluster, ctx->allocate_all)) {
4997 			break;
4998 		}
4999 	}
5000 
5001 	if (ctx->cluster < _blob->active.num_clusters) {
5002 		offset = _spdk_bs_cluster_to_lba(_blob->bs, ctx->cluster);
5003 
5004 		/* We may safely increment a cluster before write */
5005 		ctx->cluster++;
5006 
5007 		/* Use zero length write to touch a cluster */
5008 		spdk_blob_io_write(_blob, ctx->channel, NULL, offset, 0,
5009 				   _spdk_bs_inflate_blob_touch_next, ctx);
5010 	} else {
5011 		_spdk_bs_inflate_blob_done(cb_arg, bserrno);
5012 	}
5013 }
5014 
5015 static void
5016 _spdk_bs_inflate_blob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
5017 {
5018 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
5019 	uint64_t lfc; /* lowest free cluster */
5020 	uint64_t i;
5021 
5022 	if (bserrno != 0) {
5023 		_spdk_bs_clone_snapshot_cleanup_finish(ctx, bserrno);
5024 		return;
5025 	}
5026 
5027 	ctx->original.blob = _blob;
5028 
5029 	if (_blob->locked_operation_in_progress) {
5030 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot inflate blob - another operation in progress\n");
5031 		ctx->bserrno = -EBUSY;
5032 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
5033 		return;
5034 	}
5035 
5036 	_blob->locked_operation_in_progress = true;
5037 
5038 	if (!ctx->allocate_all && _blob->parent_id == SPDK_BLOBID_INVALID) {
5039 		/* This blob have no parent, so we cannot decouple it. */
5040 		SPDK_ERRLOG("Cannot decouple parent of blob with no parent.\n");
5041 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, -EINVAL);
5042 		return;
5043 	}
5044 
5045 	if (spdk_blob_is_thin_provisioned(_blob) == false) {
5046 		/* This is not thin provisioned blob. No need to inflate. */
5047 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, 0);
5048 		return;
5049 	}
5050 
5051 	/* Do two passes - one to verify that we can obtain enough clusters
5052 	 * and another to actually claim them.
5053 	 */
5054 	lfc = 0;
5055 	for (i = 0; i < _blob->active.num_clusters; i++) {
5056 		if (_spdk_bs_cluster_needs_allocation(_blob, i, ctx->allocate_all)) {
5057 			lfc = spdk_bit_array_find_first_clear(_blob->bs->used_clusters, lfc);
5058 			if (lfc == UINT32_MAX) {
5059 				/* No more free clusters. Cannot satisfy the request */
5060 				_spdk_bs_clone_snapshot_origblob_cleanup(ctx, -ENOSPC);
5061 				return;
5062 			}
5063 			lfc++;
5064 		}
5065 	}
5066 
5067 	ctx->cluster = 0;
5068 	_spdk_bs_inflate_blob_touch_next(ctx, 0);
5069 }
5070 
5071 static void
5072 _spdk_bs_inflate_blob(struct spdk_blob_store *bs, struct spdk_io_channel *channel,
5073 		      spdk_blob_id blobid, bool allocate_all, spdk_blob_op_complete cb_fn, void *cb_arg)
5074 {
5075 	struct spdk_clone_snapshot_ctx *ctx = calloc(1, sizeof(*ctx));
5076 
5077 	if (!ctx) {
5078 		cb_fn(cb_arg, -ENOMEM);
5079 		return;
5080 	}
5081 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
5082 	ctx->cpl.u.bs_basic.cb_fn = cb_fn;
5083 	ctx->cpl.u.bs_basic.cb_arg = cb_arg;
5084 	ctx->bserrno = 0;
5085 	ctx->original.id = blobid;
5086 	ctx->channel = channel;
5087 	ctx->allocate_all = allocate_all;
5088 
5089 	spdk_bs_open_blob(bs, ctx->original.id, _spdk_bs_inflate_blob_open_cpl, ctx);
5090 }
5091 
5092 void
5093 spdk_bs_inflate_blob(struct spdk_blob_store *bs, struct spdk_io_channel *channel,
5094 		     spdk_blob_id blobid, spdk_blob_op_complete cb_fn, void *cb_arg)
5095 {
5096 	_spdk_bs_inflate_blob(bs, channel, blobid, true, cb_fn, cb_arg);
5097 }
5098 
5099 void
5100 spdk_bs_blob_decouple_parent(struct spdk_blob_store *bs, struct spdk_io_channel *channel,
5101 			     spdk_blob_id blobid, spdk_blob_op_complete cb_fn, void *cb_arg)
5102 {
5103 	_spdk_bs_inflate_blob(bs, channel, blobid, false, cb_fn, cb_arg);
5104 }
5105 /* END spdk_bs_inflate_blob */
5106 
5107 /* START spdk_blob_resize */
5108 struct spdk_bs_resize_ctx {
5109 	spdk_blob_op_complete cb_fn;
5110 	void *cb_arg;
5111 	struct spdk_blob *blob;
5112 	uint64_t sz;
5113 	int rc;
5114 };
5115 
5116 static void
5117 _spdk_bs_resize_unfreeze_cpl(void *cb_arg, int rc)
5118 {
5119 	struct spdk_bs_resize_ctx *ctx = (struct spdk_bs_resize_ctx *)cb_arg;
5120 
5121 	if (rc != 0) {
5122 		SPDK_ERRLOG("Unfreeze failed, rc=%d\n", rc);
5123 	}
5124 
5125 	if (ctx->rc != 0) {
5126 		SPDK_ERRLOG("Unfreeze failed, ctx->rc=%d\n", ctx->rc);
5127 		rc = ctx->rc;
5128 	}
5129 
5130 	ctx->blob->locked_operation_in_progress = false;
5131 
5132 	ctx->cb_fn(ctx->cb_arg, rc);
5133 	free(ctx);
5134 }
5135 
5136 static void
5137 _spdk_bs_resize_freeze_cpl(void *cb_arg, int rc)
5138 {
5139 	struct spdk_bs_resize_ctx *ctx = (struct spdk_bs_resize_ctx *)cb_arg;
5140 
5141 	if (rc != 0) {
5142 		ctx->blob->locked_operation_in_progress = false;
5143 		ctx->cb_fn(ctx->cb_arg, rc);
5144 		free(ctx);
5145 		return;
5146 	}
5147 
5148 	ctx->rc = _spdk_blob_resize(ctx->blob, ctx->sz);
5149 
5150 	_spdk_blob_unfreeze_io(ctx->blob, _spdk_bs_resize_unfreeze_cpl, ctx);
5151 }
5152 
5153 void
5154 spdk_blob_resize(struct spdk_blob *blob, uint64_t sz, spdk_blob_op_complete cb_fn, void *cb_arg)
5155 {
5156 	struct spdk_bs_resize_ctx *ctx;
5157 
5158 	_spdk_blob_verify_md_op(blob);
5159 
5160 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Resizing blob %lu to %lu clusters\n", blob->id, sz);
5161 
5162 	if (blob->md_ro) {
5163 		cb_fn(cb_arg, -EPERM);
5164 		return;
5165 	}
5166 
5167 	if (sz == blob->active.num_clusters) {
5168 		cb_fn(cb_arg, 0);
5169 		return;
5170 	}
5171 
5172 	if (blob->locked_operation_in_progress) {
5173 		cb_fn(cb_arg, -EBUSY);
5174 		return;
5175 	}
5176 
5177 	ctx = calloc(1, sizeof(*ctx));
5178 	if (!ctx) {
5179 		cb_fn(cb_arg, -ENOMEM);
5180 		return;
5181 	}
5182 
5183 	blob->locked_operation_in_progress = true;
5184 	ctx->cb_fn = cb_fn;
5185 	ctx->cb_arg = cb_arg;
5186 	ctx->blob = blob;
5187 	ctx->sz = sz;
5188 	_spdk_blob_freeze_io(blob, _spdk_bs_resize_freeze_cpl, ctx);
5189 }
5190 
5191 /* END spdk_blob_resize */
5192 
5193 
5194 /* START spdk_bs_delete_blob */
5195 
5196 static void
5197 _spdk_bs_delete_close_cpl(void *cb_arg, int bserrno)
5198 {
5199 	spdk_bs_sequence_t *seq = cb_arg;
5200 
5201 	spdk_bs_sequence_finish(seq, bserrno);
5202 }
5203 
5204 static void
5205 _spdk_bs_delete_persist_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
5206 {
5207 	struct spdk_blob *blob = cb_arg;
5208 
5209 	if (bserrno != 0) {
5210 		/*
5211 		 * We already removed this blob from the blobstore tailq, so
5212 		 *  we need to free it here since this is the last reference
5213 		 *  to it.
5214 		 */
5215 		_spdk_blob_free(blob);
5216 		_spdk_bs_delete_close_cpl(seq, bserrno);
5217 		return;
5218 	}
5219 
5220 	/*
5221 	 * This will immediately decrement the ref_count and call
5222 	 *  the completion routine since the metadata state is clean.
5223 	 *  By calling spdk_blob_close, we reduce the number of call
5224 	 *  points into code that touches the blob->open_ref count
5225 	 *  and the blobstore's blob list.
5226 	 */
5227 	spdk_blob_close(blob, _spdk_bs_delete_close_cpl, seq);
5228 }
5229 
5230 struct delete_snapshot_ctx {
5231 	struct spdk_blob_list *parent_snapshot_entry;
5232 	struct spdk_blob *snapshot;
5233 	bool snapshot_md_ro;
5234 	struct spdk_blob *clone;
5235 	bool clone_md_ro;
5236 	spdk_blob_op_with_handle_complete cb_fn;
5237 	void *cb_arg;
5238 	int bserrno;
5239 };
5240 
5241 static void
5242 _spdk_delete_blob_cleanup_finish(void *cb_arg, int bserrno)
5243 {
5244 	struct delete_snapshot_ctx *ctx = cb_arg;
5245 
5246 	if (bserrno != 0) {
5247 		SPDK_ERRLOG("Snapshot cleanup error %d\n", bserrno);
5248 	}
5249 
5250 	assert(ctx != NULL);
5251 
5252 	if (bserrno != 0 && ctx->bserrno == 0) {
5253 		ctx->bserrno = bserrno;
5254 	}
5255 
5256 	ctx->cb_fn(ctx->cb_arg, ctx->snapshot, ctx->bserrno);
5257 	free(ctx);
5258 }
5259 
5260 static void
5261 _spdk_delete_snapshot_cleanup_snapshot(void *cb_arg, int bserrno)
5262 {
5263 	struct delete_snapshot_ctx *ctx = cb_arg;
5264 
5265 	if (bserrno != 0) {
5266 		ctx->bserrno = bserrno;
5267 		SPDK_ERRLOG("Clone cleanup error %d\n", bserrno);
5268 	}
5269 
5270 	/* open_ref == 1 menas that only deletion context has opened this snapshot
5271 	 * open_ref == 2 menas that clone has opened this snapshot as well,
5272 	 * so we have to add it back to the blobs list */
5273 	if (ctx->snapshot->open_ref == 2) {
5274 		TAILQ_INSERT_HEAD(&ctx->snapshot->bs->blobs, ctx->snapshot, link);
5275 	}
5276 
5277 	ctx->snapshot->locked_operation_in_progress = false;
5278 	ctx->snapshot->md_ro = ctx->snapshot_md_ro;
5279 
5280 	spdk_blob_close(ctx->snapshot, _spdk_delete_blob_cleanup_finish, ctx);
5281 }
5282 
5283 static void
5284 _spdk_delete_snapshot_cleanup_clone(void *cb_arg, int bserrno)
5285 {
5286 	struct delete_snapshot_ctx *ctx = cb_arg;
5287 
5288 	ctx->clone->locked_operation_in_progress = false;
5289 	ctx->clone->md_ro = ctx->clone_md_ro;
5290 
5291 	spdk_blob_close(ctx->clone, _spdk_delete_snapshot_cleanup_snapshot, ctx);
5292 }
5293 
5294 static void
5295 _spdk_delete_snapshot_unfreeze_cpl(void *cb_arg, int bserrno)
5296 {
5297 	struct delete_snapshot_ctx *ctx = cb_arg;
5298 
5299 	if (bserrno) {
5300 		ctx->bserrno = bserrno;
5301 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5302 		return;
5303 	}
5304 
5305 	ctx->clone->locked_operation_in_progress = false;
5306 	spdk_blob_close(ctx->clone, _spdk_delete_blob_cleanup_finish, ctx);
5307 }
5308 
5309 static void
5310 _spdk_delete_snapshot_sync_snapshot_cpl(void *cb_arg, int bserrno)
5311 {
5312 	struct delete_snapshot_ctx *ctx = cb_arg;
5313 	struct spdk_blob_list *parent_snapshot_entry = NULL;
5314 	struct spdk_blob_list *snapshot_entry = NULL;
5315 	struct spdk_blob_list *clone_entry = NULL;
5316 	struct spdk_blob_list *snapshot_clone_entry = NULL;
5317 
5318 	if (bserrno) {
5319 		SPDK_ERRLOG("Failed to sync MD on blob\n");
5320 		ctx->bserrno = bserrno;
5321 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5322 		return;
5323 	}
5324 
5325 	/* Get snapshot entry for the snapshot we want to remove */
5326 	snapshot_entry = _spdk_bs_get_snapshot_entry(ctx->snapshot->bs, ctx->snapshot->id);
5327 
5328 	assert(snapshot_entry != NULL);
5329 
5330 	/* Remove clone entry in this snapshot (at this point there can be only one clone) */
5331 	clone_entry = TAILQ_FIRST(&snapshot_entry->clones);
5332 	assert(clone_entry != NULL);
5333 	TAILQ_REMOVE(&snapshot_entry->clones, clone_entry, link);
5334 	snapshot_entry->clone_count--;
5335 	assert(TAILQ_EMPTY(&snapshot_entry->clones));
5336 
5337 	if (ctx->snapshot->parent_id != SPDK_BLOBID_INVALID) {
5338 		/* This snapshot is at the same time a clone of another snapshot - we need to
5339 		 * update parent snapshot (remove current clone, add new one inherited from
5340 		 * the snapshot that is being removed) */
5341 
5342 		/* Get snapshot entry for parent snapshot and clone entry within that snapshot for
5343 		 * snapshot that we are removing */
5344 		_spdk_blob_get_snapshot_and_clone_entries(ctx->snapshot, &parent_snapshot_entry,
5345 				&snapshot_clone_entry);
5346 
5347 		/* Switch clone entry in parent snapshot */
5348 		TAILQ_INSERT_TAIL(&parent_snapshot_entry->clones, clone_entry, link);
5349 		TAILQ_REMOVE(&parent_snapshot_entry->clones, snapshot_clone_entry, link);
5350 		free(snapshot_clone_entry);
5351 	} else {
5352 		/* No parent snapshot - just remove clone entry */
5353 		free(clone_entry);
5354 	}
5355 
5356 	/* Restore md_ro flags */
5357 	ctx->clone->md_ro = ctx->clone_md_ro;
5358 	ctx->snapshot->md_ro = ctx->snapshot_md_ro;
5359 
5360 	_spdk_blob_unfreeze_io(ctx->clone, _spdk_delete_snapshot_unfreeze_cpl, ctx);
5361 }
5362 
5363 static void
5364 _spdk_delete_snapshot_sync_clone_cpl(void *cb_arg, int bserrno)
5365 {
5366 	struct delete_snapshot_ctx *ctx = cb_arg;
5367 	uint64_t i;
5368 
5369 	ctx->snapshot->md_ro = false;
5370 
5371 	if (bserrno) {
5372 		SPDK_ERRLOG("Failed to sync MD on clone\n");
5373 		ctx->bserrno = bserrno;
5374 
5375 		/* Restore snapshot to previous state */
5376 		bserrno = _spdk_blob_remove_xattr(ctx->snapshot, SNAPSHOT_PENDING_REMOVAL, true);
5377 		if (bserrno != 0) {
5378 			_spdk_delete_snapshot_cleanup_clone(ctx, bserrno);
5379 			return;
5380 		}
5381 
5382 		spdk_blob_sync_md(ctx->snapshot, _spdk_delete_snapshot_cleanup_clone, ctx);
5383 		return;
5384 	}
5385 
5386 	/* Clear cluster map entries for snapshot */
5387 	for (i = 0; i < ctx->snapshot->active.num_clusters && i < ctx->clone->active.num_clusters; i++) {
5388 		if (ctx->clone->active.clusters[i] == ctx->snapshot->active.clusters[i]) {
5389 			ctx->snapshot->active.clusters[i] = 0;
5390 		}
5391 	}
5392 
5393 	ctx->snapshot->state = SPDK_BLOB_STATE_DIRTY;
5394 
5395 	if (ctx->parent_snapshot_entry != NULL) {
5396 		ctx->snapshot->back_bs_dev = NULL;
5397 	}
5398 
5399 	spdk_blob_sync_md(ctx->snapshot, _spdk_delete_snapshot_sync_snapshot_cpl, ctx);
5400 }
5401 
5402 static void
5403 _spdk_delete_snapshot_sync_snapshot_xattr_cpl(void *cb_arg, int bserrno)
5404 {
5405 	struct delete_snapshot_ctx *ctx = cb_arg;
5406 	uint64_t i;
5407 
5408 	/* Temporarily override md_ro flag for clone for MD modification */
5409 	ctx->clone_md_ro = ctx->clone->md_ro;
5410 	ctx->clone->md_ro = false;
5411 
5412 	if (bserrno) {
5413 		SPDK_ERRLOG("Failed to sync MD with xattr on blob\n");
5414 		ctx->bserrno = bserrno;
5415 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5416 		return;
5417 	}
5418 
5419 	/* Copy snapshot map to clone map (only unallocated clusters in clone) */
5420 	for (i = 0; i < ctx->snapshot->active.num_clusters && i < ctx->clone->active.num_clusters; i++) {
5421 		if (ctx->clone->active.clusters[i] == 0) {
5422 			ctx->clone->active.clusters[i] = ctx->snapshot->active.clusters[i];
5423 		}
5424 	}
5425 
5426 	/* Delete old backing bs_dev from clone (related to snapshot that will be removed) */
5427 	ctx->clone->back_bs_dev->destroy(ctx->clone->back_bs_dev);
5428 
5429 	/* Set/remove snapshot xattr and switch parent ID and backing bs_dev on clone... */
5430 	if (ctx->parent_snapshot_entry != NULL) {
5431 		/* ...to parent snapshot */
5432 		ctx->clone->parent_id = ctx->parent_snapshot_entry->id;
5433 		ctx->clone->back_bs_dev = ctx->snapshot->back_bs_dev;
5434 		_spdk_blob_set_xattr(ctx->clone, BLOB_SNAPSHOT, &ctx->parent_snapshot_entry->id,
5435 				     sizeof(spdk_blob_id),
5436 				     true);
5437 	} else {
5438 		/* ...to blobid invalid and zeroes dev */
5439 		ctx->clone->parent_id = SPDK_BLOBID_INVALID;
5440 		ctx->clone->back_bs_dev = spdk_bs_create_zeroes_dev();
5441 		_spdk_blob_remove_xattr(ctx->clone, BLOB_SNAPSHOT, true);
5442 	}
5443 
5444 	spdk_blob_sync_md(ctx->clone, _spdk_delete_snapshot_sync_clone_cpl, ctx);
5445 }
5446 
5447 static void
5448 _spdk_delete_snapshot_freeze_io_cb(void *cb_arg, int bserrno)
5449 {
5450 	struct delete_snapshot_ctx *ctx = cb_arg;
5451 
5452 	if (bserrno) {
5453 		SPDK_ERRLOG("Failed to freeze I/O on clone\n");
5454 		ctx->bserrno = bserrno;
5455 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5456 		return;
5457 	}
5458 
5459 	/* Temporarily override md_ro flag for snapshot for MD modification */
5460 	ctx->snapshot_md_ro = ctx->snapshot->md_ro;
5461 	ctx->snapshot->md_ro = false;
5462 
5463 	/* Mark blob as pending for removal for power failure safety, use clone id for recovery */
5464 	ctx->bserrno = _spdk_blob_set_xattr(ctx->snapshot, SNAPSHOT_PENDING_REMOVAL, &ctx->clone->id,
5465 					    sizeof(spdk_blob_id), true);
5466 	if (ctx->bserrno != 0) {
5467 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5468 		return;
5469 	}
5470 
5471 	spdk_blob_sync_md(ctx->snapshot, _spdk_delete_snapshot_sync_snapshot_xattr_cpl, ctx);
5472 }
5473 
5474 static void
5475 _spdk_delete_snapshot_open_clone_cb(void *cb_arg, struct spdk_blob *clone, int bserrno)
5476 {
5477 	struct delete_snapshot_ctx *ctx = cb_arg;
5478 
5479 	if (bserrno) {
5480 		SPDK_ERRLOG("Failed to open clone\n");
5481 		ctx->bserrno = bserrno;
5482 		_spdk_delete_snapshot_cleanup_snapshot(ctx, 0);
5483 		return;
5484 	}
5485 
5486 	ctx->clone = clone;
5487 
5488 	if (clone->locked_operation_in_progress) {
5489 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot remove blob - another operation in progress on its clone\n");
5490 		ctx->bserrno = -EBUSY;
5491 		spdk_blob_close(ctx->clone, _spdk_delete_snapshot_cleanup_snapshot, ctx);
5492 		return;
5493 	}
5494 
5495 	clone->locked_operation_in_progress = true;
5496 
5497 	_spdk_blob_freeze_io(clone, _spdk_delete_snapshot_freeze_io_cb, ctx);
5498 }
5499 
5500 static void
5501 _spdk_update_clone_on_snapshot_deletion(struct spdk_blob *snapshot, struct delete_snapshot_ctx *ctx)
5502 {
5503 	struct spdk_blob_list *snapshot_entry = NULL;
5504 	struct spdk_blob_list *clone_entry = NULL;
5505 	struct spdk_blob_list *snapshot_clone_entry = NULL;
5506 
5507 	/* Get snapshot entry for the snapshot we want to remove */
5508 	snapshot_entry = _spdk_bs_get_snapshot_entry(snapshot->bs, snapshot->id);
5509 
5510 	assert(snapshot_entry != NULL);
5511 
5512 	/* Get clone of the snapshot (at this point there can be only one clone) */
5513 	clone_entry = TAILQ_FIRST(&snapshot_entry->clones);
5514 	assert(snapshot_entry->clone_count == 1);
5515 	assert(clone_entry != NULL);
5516 
5517 	/* Get snapshot entry for parent snapshot and clone entry within that snapshot for
5518 	 * snapshot that we are removing */
5519 	_spdk_blob_get_snapshot_and_clone_entries(snapshot, &ctx->parent_snapshot_entry,
5520 			&snapshot_clone_entry);
5521 
5522 	spdk_bs_open_blob(snapshot->bs, clone_entry->id, _spdk_delete_snapshot_open_clone_cb, ctx);
5523 }
5524 
5525 static void
5526 _spdk_bs_delete_blob_finish(void *cb_arg, struct spdk_blob *blob, int bserrno)
5527 {
5528 	spdk_bs_sequence_t *seq = cb_arg;
5529 	struct spdk_blob_list *snapshot_entry = NULL;
5530 	uint32_t page_num;
5531 
5532 	if (bserrno) {
5533 		SPDK_ERRLOG("Failed to remove blob\n");
5534 		spdk_bs_sequence_finish(seq, bserrno);
5535 		return;
5536 	}
5537 
5538 	/* Remove snapshot from the list */
5539 	snapshot_entry = _spdk_bs_get_snapshot_entry(blob->bs, blob->id);
5540 	if (snapshot_entry != NULL) {
5541 		TAILQ_REMOVE(&blob->bs->snapshots, snapshot_entry, link);
5542 		free(snapshot_entry);
5543 	}
5544 
5545 	page_num = _spdk_bs_blobid_to_page(blob->id);
5546 	spdk_bit_array_clear(blob->bs->used_blobids, page_num);
5547 	blob->state = SPDK_BLOB_STATE_DIRTY;
5548 	blob->active.num_pages = 0;
5549 	_spdk_blob_resize(blob, 0);
5550 
5551 	_spdk_blob_persist(seq, blob, _spdk_bs_delete_persist_cpl, blob);
5552 }
5553 
5554 static int
5555 _spdk_bs_is_blob_deletable(struct spdk_blob *blob, bool *update_clone)
5556 {
5557 	struct spdk_blob_list *snapshot_entry = NULL;
5558 	struct spdk_blob_list *clone_entry = NULL;
5559 	struct spdk_blob *clone = NULL;
5560 	bool has_one_clone = false;
5561 
5562 	/* Check if this is a snapshot with clones */
5563 	snapshot_entry = _spdk_bs_get_snapshot_entry(blob->bs, blob->id);
5564 	if (snapshot_entry != NULL) {
5565 		if (snapshot_entry->clone_count > 1) {
5566 			SPDK_ERRLOG("Cannot remove snapshot with more than one clone\n");
5567 			return -EBUSY;
5568 		} else if (snapshot_entry->clone_count == 1) {
5569 			has_one_clone = true;
5570 		}
5571 	}
5572 
5573 	/* Check if someone has this blob open (besides this delete context):
5574 	 * - open_ref = 1 - only this context opened blob, so it is ok to remove it
5575 	 * - open_ref <= 2 && has_one_clone = true - clone is holding snapshot
5576 	 *	and that is ok, because we will update it accordingly */
5577 	if (blob->open_ref <= 2 && has_one_clone) {
5578 		clone_entry = TAILQ_FIRST(&snapshot_entry->clones);
5579 		assert(clone_entry != NULL);
5580 		clone = _spdk_blob_lookup(blob->bs, clone_entry->id);
5581 
5582 		if (blob->open_ref == 2 && clone == NULL) {
5583 			/* Clone is closed and someone else opened this blob */
5584 			SPDK_ERRLOG("Cannot remove snapshot because it is open\n");
5585 			return -EBUSY;
5586 		}
5587 
5588 		*update_clone = true;
5589 		return 0;
5590 	}
5591 
5592 	if (blob->open_ref > 1) {
5593 		SPDK_ERRLOG("Cannot remove snapshot because it is open\n");
5594 		return -EBUSY;
5595 	}
5596 
5597 	assert(has_one_clone == false);
5598 	*update_clone = false;
5599 	return 0;
5600 }
5601 
5602 static void
5603 _spdk_bs_delete_enomem_close_cpl(void *cb_arg, int bserrno)
5604 {
5605 	spdk_bs_sequence_t *seq = cb_arg;
5606 
5607 	spdk_bs_sequence_finish(seq, -ENOMEM);
5608 }
5609 
5610 static void
5611 _spdk_bs_delete_open_cpl(void *cb_arg, struct spdk_blob *blob, int bserrno)
5612 {
5613 	spdk_bs_sequence_t *seq = cb_arg;
5614 	struct delete_snapshot_ctx *ctx;
5615 	bool update_clone = false;
5616 
5617 	if (bserrno != 0) {
5618 		spdk_bs_sequence_finish(seq, bserrno);
5619 		return;
5620 	}
5621 
5622 	_spdk_blob_verify_md_op(blob);
5623 
5624 	ctx = calloc(1, sizeof(*ctx));
5625 	if (ctx == NULL) {
5626 		spdk_blob_close(blob, _spdk_bs_delete_enomem_close_cpl, seq);
5627 		return;
5628 	}
5629 
5630 	ctx->snapshot = blob;
5631 	ctx->cb_fn = _spdk_bs_delete_blob_finish;
5632 	ctx->cb_arg = seq;
5633 
5634 	/* Check if blob can be removed and if it is a snapshot with clone on top of it */
5635 	ctx->bserrno = _spdk_bs_is_blob_deletable(blob, &update_clone);
5636 	if (ctx->bserrno) {
5637 		spdk_blob_close(blob, _spdk_delete_blob_cleanup_finish, ctx);
5638 		return;
5639 	}
5640 
5641 	if (blob->locked_operation_in_progress) {
5642 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot remove blob - another operation in progress\n");
5643 		ctx->bserrno = -EBUSY;
5644 		spdk_blob_close(blob, _spdk_delete_blob_cleanup_finish, ctx);
5645 		return;
5646 	}
5647 
5648 	blob->locked_operation_in_progress = true;
5649 
5650 	/*
5651 	 * Remove the blob from the blob_store list now, to ensure it does not
5652 	 *  get returned after this point by _spdk_blob_lookup().
5653 	 */
5654 	TAILQ_REMOVE(&blob->bs->blobs, blob, link);
5655 
5656 	if (update_clone) {
5657 		/* This blob is a snapshot with active clone - update clone first */
5658 		_spdk_update_clone_on_snapshot_deletion(blob, ctx);
5659 	} else {
5660 		/* This blob does not have any clones - just remove it */
5661 		_spdk_bs_blob_list_remove(blob);
5662 		_spdk_bs_delete_blob_finish(seq, blob, 0);
5663 		free(ctx);
5664 	}
5665 }
5666 
5667 void
5668 spdk_bs_delete_blob(struct spdk_blob_store *bs, spdk_blob_id blobid,
5669 		    spdk_blob_op_complete cb_fn, void *cb_arg)
5670 {
5671 	struct spdk_bs_cpl	cpl;
5672 	spdk_bs_sequence_t	*seq;
5673 
5674 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Deleting blob %lu\n", blobid);
5675 
5676 	assert(spdk_get_thread() == bs->md_thread);
5677 
5678 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
5679 	cpl.u.blob_basic.cb_fn = cb_fn;
5680 	cpl.u.blob_basic.cb_arg = cb_arg;
5681 
5682 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
5683 	if (!seq) {
5684 		cb_fn(cb_arg, -ENOMEM);
5685 		return;
5686 	}
5687 
5688 	spdk_bs_open_blob(bs, blobid, _spdk_bs_delete_open_cpl, seq);
5689 }
5690 
5691 /* END spdk_bs_delete_blob */
5692 
5693 /* START spdk_bs_open_blob */
5694 
5695 static void
5696 _spdk_bs_open_blob_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
5697 {
5698 	struct spdk_blob *blob = cb_arg;
5699 
5700 	if (bserrno != 0) {
5701 		_spdk_blob_free(blob);
5702 		seq->cpl.u.blob_handle.blob = NULL;
5703 		spdk_bs_sequence_finish(seq, bserrno);
5704 		return;
5705 	}
5706 
5707 	blob->open_ref++;
5708 
5709 	TAILQ_INSERT_HEAD(&blob->bs->blobs, blob, link);
5710 
5711 	spdk_bs_sequence_finish(seq, bserrno);
5712 }
5713 
5714 static void _spdk_bs_open_blob(struct spdk_blob_store *bs, spdk_blob_id blobid,
5715 			       struct spdk_blob_open_opts *opts, spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
5716 {
5717 	struct spdk_blob		*blob;
5718 	struct spdk_bs_cpl		cpl;
5719 	struct spdk_blob_open_opts	opts_default;
5720 	spdk_bs_sequence_t		*seq;
5721 	uint32_t			page_num;
5722 
5723 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Opening blob %lu\n", blobid);
5724 	assert(spdk_get_thread() == bs->md_thread);
5725 
5726 	page_num = _spdk_bs_blobid_to_page(blobid);
5727 	if (spdk_bit_array_get(bs->used_blobids, page_num) == false) {
5728 		/* Invalid blobid */
5729 		cb_fn(cb_arg, NULL, -ENOENT);
5730 		return;
5731 	}
5732 
5733 	blob = _spdk_blob_lookup(bs, blobid);
5734 	if (blob) {
5735 		blob->open_ref++;
5736 		cb_fn(cb_arg, blob, 0);
5737 		return;
5738 	}
5739 
5740 	blob = _spdk_blob_alloc(bs, blobid);
5741 	if (!blob) {
5742 		cb_fn(cb_arg, NULL, -ENOMEM);
5743 		return;
5744 	}
5745 
5746 	if (!opts) {
5747 		spdk_blob_open_opts_init(&opts_default);
5748 		opts = &opts_default;
5749 	}
5750 
5751 	blob->clear_method = opts->clear_method;
5752 
5753 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_HANDLE;
5754 	cpl.u.blob_handle.cb_fn = cb_fn;
5755 	cpl.u.blob_handle.cb_arg = cb_arg;
5756 	cpl.u.blob_handle.blob = blob;
5757 
5758 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
5759 	if (!seq) {
5760 		_spdk_blob_free(blob);
5761 		cb_fn(cb_arg, NULL, -ENOMEM);
5762 		return;
5763 	}
5764 
5765 	_spdk_blob_load(seq, blob, _spdk_bs_open_blob_cpl, blob);
5766 }
5767 
5768 void spdk_bs_open_blob(struct spdk_blob_store *bs, spdk_blob_id blobid,
5769 		       spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
5770 {
5771 	_spdk_bs_open_blob(bs, blobid, NULL, cb_fn, cb_arg);
5772 }
5773 
5774 void spdk_bs_open_blob_ext(struct spdk_blob_store *bs, spdk_blob_id blobid,
5775 			   struct spdk_blob_open_opts *opts, spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
5776 {
5777 	_spdk_bs_open_blob(bs, blobid, opts, cb_fn, cb_arg);
5778 }
5779 
5780 /* END spdk_bs_open_blob */
5781 
5782 /* START spdk_blob_set_read_only */
5783 int spdk_blob_set_read_only(struct spdk_blob *blob)
5784 {
5785 	_spdk_blob_verify_md_op(blob);
5786 
5787 	blob->data_ro_flags |= SPDK_BLOB_READ_ONLY;
5788 
5789 	blob->state = SPDK_BLOB_STATE_DIRTY;
5790 	return 0;
5791 }
5792 /* END spdk_blob_set_read_only */
5793 
5794 /* START spdk_blob_sync_md */
5795 
5796 static void
5797 _spdk_blob_sync_md_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
5798 {
5799 	struct spdk_blob *blob = cb_arg;
5800 
5801 	if (bserrno == 0 && (blob->data_ro_flags & SPDK_BLOB_READ_ONLY)) {
5802 		blob->data_ro = true;
5803 		blob->md_ro = true;
5804 	}
5805 
5806 	spdk_bs_sequence_finish(seq, bserrno);
5807 }
5808 
5809 static void
5810 _spdk_blob_sync_md(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
5811 {
5812 	struct spdk_bs_cpl	cpl;
5813 	spdk_bs_sequence_t	*seq;
5814 
5815 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
5816 	cpl.u.blob_basic.cb_fn = cb_fn;
5817 	cpl.u.blob_basic.cb_arg = cb_arg;
5818 
5819 	seq = spdk_bs_sequence_start(blob->bs->md_channel, &cpl);
5820 	if (!seq) {
5821 		cb_fn(cb_arg, -ENOMEM);
5822 		return;
5823 	}
5824 
5825 	_spdk_blob_persist(seq, blob, _spdk_blob_sync_md_cpl, blob);
5826 }
5827 
5828 void
5829 spdk_blob_sync_md(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
5830 {
5831 	_spdk_blob_verify_md_op(blob);
5832 
5833 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Syncing blob %lu\n", blob->id);
5834 
5835 	if (blob->md_ro) {
5836 		assert(blob->state == SPDK_BLOB_STATE_CLEAN);
5837 		cb_fn(cb_arg, 0);
5838 		return;
5839 	}
5840 
5841 	_spdk_blob_sync_md(blob, cb_fn, cb_arg);
5842 }
5843 
5844 /* END spdk_blob_sync_md */
5845 
5846 struct spdk_blob_insert_cluster_ctx {
5847 	struct spdk_thread	*thread;
5848 	struct spdk_blob	*blob;
5849 	uint32_t		cluster_num;	/* cluster index in blob */
5850 	uint32_t		cluster;	/* cluster on disk */
5851 	int			rc;
5852 	spdk_blob_op_complete	cb_fn;
5853 	void			*cb_arg;
5854 };
5855 
5856 static void
5857 _spdk_blob_insert_cluster_msg_cpl(void *arg)
5858 {
5859 	struct spdk_blob_insert_cluster_ctx *ctx = arg;
5860 
5861 	ctx->cb_fn(ctx->cb_arg, ctx->rc);
5862 	free(ctx);
5863 }
5864 
5865 static void
5866 _spdk_blob_insert_cluster_msg_cb(void *arg, int bserrno)
5867 {
5868 	struct spdk_blob_insert_cluster_ctx *ctx = arg;
5869 
5870 	ctx->rc = bserrno;
5871 	spdk_thread_send_msg(ctx->thread, _spdk_blob_insert_cluster_msg_cpl, ctx);
5872 }
5873 
5874 static void
5875 _spdk_blob_insert_cluster_msg(void *arg)
5876 {
5877 	struct spdk_blob_insert_cluster_ctx *ctx = arg;
5878 
5879 	ctx->rc = _spdk_blob_insert_cluster(ctx->blob, ctx->cluster_num, ctx->cluster);
5880 	if (ctx->rc != 0) {
5881 		spdk_thread_send_msg(ctx->thread, _spdk_blob_insert_cluster_msg_cpl, ctx);
5882 		return;
5883 	}
5884 
5885 	ctx->blob->state = SPDK_BLOB_STATE_DIRTY;
5886 	_spdk_blob_sync_md(ctx->blob, _spdk_blob_insert_cluster_msg_cb, ctx);
5887 }
5888 
5889 static void
5890 _spdk_blob_insert_cluster_on_md_thread(struct spdk_blob *blob, uint32_t cluster_num,
5891 				       uint64_t cluster, spdk_blob_op_complete cb_fn, void *cb_arg)
5892 {
5893 	struct spdk_blob_insert_cluster_ctx *ctx;
5894 
5895 	ctx = calloc(1, sizeof(*ctx));
5896 	if (ctx == NULL) {
5897 		cb_fn(cb_arg, -ENOMEM);
5898 		return;
5899 	}
5900 
5901 	ctx->thread = spdk_get_thread();
5902 	ctx->blob = blob;
5903 	ctx->cluster_num = cluster_num;
5904 	ctx->cluster = cluster;
5905 	ctx->cb_fn = cb_fn;
5906 	ctx->cb_arg = cb_arg;
5907 
5908 	spdk_thread_send_msg(blob->bs->md_thread, _spdk_blob_insert_cluster_msg, ctx);
5909 }
5910 
5911 /* START spdk_blob_close */
5912 
5913 static void
5914 _spdk_blob_close_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
5915 {
5916 	struct spdk_blob *blob = cb_arg;
5917 
5918 	if (bserrno == 0) {
5919 		blob->open_ref--;
5920 		if (blob->open_ref == 0) {
5921 			/*
5922 			 * Blobs with active.num_pages == 0 are deleted blobs.
5923 			 *  these blobs are removed from the blob_store list
5924 			 *  when the deletion process starts - so don't try to
5925 			 *  remove them again.
5926 			 */
5927 			if (blob->active.num_pages > 0) {
5928 				TAILQ_REMOVE(&blob->bs->blobs, blob, link);
5929 			}
5930 			_spdk_blob_free(blob);
5931 		}
5932 	}
5933 
5934 	spdk_bs_sequence_finish(seq, bserrno);
5935 }
5936 
5937 void spdk_blob_close(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
5938 {
5939 	struct spdk_bs_cpl	cpl;
5940 	spdk_bs_sequence_t	*seq;
5941 
5942 	_spdk_blob_verify_md_op(blob);
5943 
5944 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Closing blob %lu\n", blob->id);
5945 
5946 	if (blob->open_ref == 0) {
5947 		cb_fn(cb_arg, -EBADF);
5948 		return;
5949 	}
5950 
5951 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
5952 	cpl.u.blob_basic.cb_fn = cb_fn;
5953 	cpl.u.blob_basic.cb_arg = cb_arg;
5954 
5955 	seq = spdk_bs_sequence_start(blob->bs->md_channel, &cpl);
5956 	if (!seq) {
5957 		cb_fn(cb_arg, -ENOMEM);
5958 		return;
5959 	}
5960 
5961 	/* Sync metadata */
5962 	_spdk_blob_persist(seq, blob, _spdk_blob_close_cpl, blob);
5963 }
5964 
5965 /* END spdk_blob_close */
5966 
5967 struct spdk_io_channel *spdk_bs_alloc_io_channel(struct spdk_blob_store *bs)
5968 {
5969 	return spdk_get_io_channel(bs);
5970 }
5971 
5972 void spdk_bs_free_io_channel(struct spdk_io_channel *channel)
5973 {
5974 	spdk_put_io_channel(channel);
5975 }
5976 
5977 void spdk_blob_io_unmap(struct spdk_blob *blob, struct spdk_io_channel *channel,
5978 			uint64_t offset, uint64_t length, spdk_blob_op_complete cb_fn, void *cb_arg)
5979 {
5980 	_spdk_blob_request_submit_op(blob, channel, NULL, offset, length, cb_fn, cb_arg,
5981 				     SPDK_BLOB_UNMAP);
5982 }
5983 
5984 void spdk_blob_io_write_zeroes(struct spdk_blob *blob, struct spdk_io_channel *channel,
5985 			       uint64_t offset, uint64_t length, spdk_blob_op_complete cb_fn, void *cb_arg)
5986 {
5987 	_spdk_blob_request_submit_op(blob, channel, NULL, offset, length, cb_fn, cb_arg,
5988 				     SPDK_BLOB_WRITE_ZEROES);
5989 }
5990 
5991 void spdk_blob_io_write(struct spdk_blob *blob, struct spdk_io_channel *channel,
5992 			void *payload, uint64_t offset, uint64_t length,
5993 			spdk_blob_op_complete cb_fn, void *cb_arg)
5994 {
5995 	_spdk_blob_request_submit_op(blob, channel, payload, offset, length, cb_fn, cb_arg,
5996 				     SPDK_BLOB_WRITE);
5997 }
5998 
5999 void spdk_blob_io_read(struct spdk_blob *blob, struct spdk_io_channel *channel,
6000 		       void *payload, uint64_t offset, uint64_t length,
6001 		       spdk_blob_op_complete cb_fn, void *cb_arg)
6002 {
6003 	_spdk_blob_request_submit_op(blob, channel, payload, offset, length, cb_fn, cb_arg,
6004 				     SPDK_BLOB_READ);
6005 }
6006 
6007 void spdk_blob_io_writev(struct spdk_blob *blob, struct spdk_io_channel *channel,
6008 			 struct iovec *iov, int iovcnt, uint64_t offset, uint64_t length,
6009 			 spdk_blob_op_complete cb_fn, void *cb_arg)
6010 {
6011 	_spdk_blob_request_submit_rw_iov(blob, channel, iov, iovcnt, offset, length, cb_fn, cb_arg, false);
6012 }
6013 
6014 void spdk_blob_io_readv(struct spdk_blob *blob, struct spdk_io_channel *channel,
6015 			struct iovec *iov, int iovcnt, uint64_t offset, uint64_t length,
6016 			spdk_blob_op_complete cb_fn, void *cb_arg)
6017 {
6018 	_spdk_blob_request_submit_rw_iov(blob, channel, iov, iovcnt, offset, length, cb_fn, cb_arg, true);
6019 }
6020 
6021 struct spdk_bs_iter_ctx {
6022 	int64_t page_num;
6023 	struct spdk_blob_store *bs;
6024 
6025 	spdk_blob_op_with_handle_complete cb_fn;
6026 	void *cb_arg;
6027 };
6028 
6029 static void
6030 _spdk_bs_iter_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
6031 {
6032 	struct spdk_bs_iter_ctx *ctx = cb_arg;
6033 	struct spdk_blob_store *bs = ctx->bs;
6034 	spdk_blob_id id;
6035 
6036 	if (bserrno == 0) {
6037 		ctx->cb_fn(ctx->cb_arg, _blob, bserrno);
6038 		free(ctx);
6039 		return;
6040 	}
6041 
6042 	ctx->page_num++;
6043 	ctx->page_num = spdk_bit_array_find_first_set(bs->used_blobids, ctx->page_num);
6044 	if (ctx->page_num >= spdk_bit_array_capacity(bs->used_blobids)) {
6045 		ctx->cb_fn(ctx->cb_arg, NULL, -ENOENT);
6046 		free(ctx);
6047 		return;
6048 	}
6049 
6050 	id = _spdk_bs_page_to_blobid(ctx->page_num);
6051 
6052 	spdk_bs_open_blob(bs, id, _spdk_bs_iter_cpl, ctx);
6053 }
6054 
6055 void
6056 spdk_bs_iter_first(struct spdk_blob_store *bs,
6057 		   spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
6058 {
6059 	struct spdk_bs_iter_ctx *ctx;
6060 
6061 	ctx = calloc(1, sizeof(*ctx));
6062 	if (!ctx) {
6063 		cb_fn(cb_arg, NULL, -ENOMEM);
6064 		return;
6065 	}
6066 
6067 	ctx->page_num = -1;
6068 	ctx->bs = bs;
6069 	ctx->cb_fn = cb_fn;
6070 	ctx->cb_arg = cb_arg;
6071 
6072 	_spdk_bs_iter_cpl(ctx, NULL, -1);
6073 }
6074 
6075 static void
6076 _spdk_bs_iter_close_cpl(void *cb_arg, int bserrno)
6077 {
6078 	struct spdk_bs_iter_ctx *ctx = cb_arg;
6079 
6080 	_spdk_bs_iter_cpl(ctx, NULL, -1);
6081 }
6082 
6083 void
6084 spdk_bs_iter_next(struct spdk_blob_store *bs, struct spdk_blob *blob,
6085 		  spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
6086 {
6087 	struct spdk_bs_iter_ctx *ctx;
6088 
6089 	assert(blob != NULL);
6090 
6091 	ctx = calloc(1, sizeof(*ctx));
6092 	if (!ctx) {
6093 		cb_fn(cb_arg, NULL, -ENOMEM);
6094 		return;
6095 	}
6096 
6097 	ctx->page_num = _spdk_bs_blobid_to_page(blob->id);
6098 	ctx->bs = bs;
6099 	ctx->cb_fn = cb_fn;
6100 	ctx->cb_arg = cb_arg;
6101 
6102 	/* Close the existing blob */
6103 	spdk_blob_close(blob, _spdk_bs_iter_close_cpl, ctx);
6104 }
6105 
6106 static int
6107 _spdk_blob_set_xattr(struct spdk_blob *blob, const char *name, const void *value,
6108 		     uint16_t value_len, bool internal)
6109 {
6110 	struct spdk_xattr_tailq *xattrs;
6111 	struct spdk_xattr	*xattr;
6112 	size_t			desc_size;
6113 
6114 	_spdk_blob_verify_md_op(blob);
6115 
6116 	if (blob->md_ro) {
6117 		return -EPERM;
6118 	}
6119 
6120 	desc_size = sizeof(struct spdk_blob_md_descriptor_xattr) + strlen(name) + value_len;
6121 	if (desc_size > SPDK_BS_MAX_DESC_SIZE) {
6122 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Xattr '%s' of size %ld does not fix into single page %ld\n", name,
6123 			      desc_size, SPDK_BS_MAX_DESC_SIZE);
6124 		return -ENOMEM;
6125 	}
6126 
6127 	if (internal) {
6128 		xattrs = &blob->xattrs_internal;
6129 		blob->invalid_flags |= SPDK_BLOB_INTERNAL_XATTR;
6130 	} else {
6131 		xattrs = &blob->xattrs;
6132 	}
6133 
6134 	TAILQ_FOREACH(xattr, xattrs, link) {
6135 		if (!strcmp(name, xattr->name)) {
6136 			free(xattr->value);
6137 			xattr->value_len = value_len;
6138 			xattr->value = malloc(value_len);
6139 			memcpy(xattr->value, value, value_len);
6140 
6141 			blob->state = SPDK_BLOB_STATE_DIRTY;
6142 
6143 			return 0;
6144 		}
6145 	}
6146 
6147 	xattr = calloc(1, sizeof(*xattr));
6148 	if (!xattr) {
6149 		return -ENOMEM;
6150 	}
6151 	xattr->name = strdup(name);
6152 	xattr->value_len = value_len;
6153 	xattr->value = malloc(value_len);
6154 	memcpy(xattr->value, value, value_len);
6155 	TAILQ_INSERT_TAIL(xattrs, xattr, link);
6156 
6157 	blob->state = SPDK_BLOB_STATE_DIRTY;
6158 
6159 	return 0;
6160 }
6161 
6162 int
6163 spdk_blob_set_xattr(struct spdk_blob *blob, const char *name, const void *value,
6164 		    uint16_t value_len)
6165 {
6166 	return _spdk_blob_set_xattr(blob, name, value, value_len, false);
6167 }
6168 
6169 static int
6170 _spdk_blob_remove_xattr(struct spdk_blob *blob, const char *name, bool internal)
6171 {
6172 	struct spdk_xattr_tailq *xattrs;
6173 	struct spdk_xattr	*xattr;
6174 
6175 	_spdk_blob_verify_md_op(blob);
6176 
6177 	if (blob->md_ro) {
6178 		return -EPERM;
6179 	}
6180 	xattrs = internal ? &blob->xattrs_internal : &blob->xattrs;
6181 
6182 	TAILQ_FOREACH(xattr, xattrs, link) {
6183 		if (!strcmp(name, xattr->name)) {
6184 			TAILQ_REMOVE(xattrs, xattr, link);
6185 			free(xattr->value);
6186 			free(xattr->name);
6187 			free(xattr);
6188 
6189 			if (internal && TAILQ_EMPTY(&blob->xattrs_internal)) {
6190 				blob->invalid_flags &= ~SPDK_BLOB_INTERNAL_XATTR;
6191 			}
6192 			blob->state = SPDK_BLOB_STATE_DIRTY;
6193 
6194 			return 0;
6195 		}
6196 	}
6197 
6198 	return -ENOENT;
6199 }
6200 
6201 int
6202 spdk_blob_remove_xattr(struct spdk_blob *blob, const char *name)
6203 {
6204 	return _spdk_blob_remove_xattr(blob, name, false);
6205 }
6206 
6207 static int
6208 _spdk_blob_get_xattr_value(struct spdk_blob *blob, const char *name,
6209 			   const void **value, size_t *value_len, bool internal)
6210 {
6211 	struct spdk_xattr	*xattr;
6212 	struct spdk_xattr_tailq *xattrs;
6213 
6214 	xattrs = internal ? &blob->xattrs_internal : &blob->xattrs;
6215 
6216 	TAILQ_FOREACH(xattr, xattrs, link) {
6217 		if (!strcmp(name, xattr->name)) {
6218 			*value = xattr->value;
6219 			*value_len = xattr->value_len;
6220 			return 0;
6221 		}
6222 	}
6223 	return -ENOENT;
6224 }
6225 
6226 int
6227 spdk_blob_get_xattr_value(struct spdk_blob *blob, const char *name,
6228 			  const void **value, size_t *value_len)
6229 {
6230 	_spdk_blob_verify_md_op(blob);
6231 
6232 	return _spdk_blob_get_xattr_value(blob, name, value, value_len, false);
6233 }
6234 
6235 struct spdk_xattr_names {
6236 	uint32_t	count;
6237 	const char	*names[0];
6238 };
6239 
6240 static int
6241 _spdk_blob_get_xattr_names(struct spdk_xattr_tailq *xattrs, struct spdk_xattr_names **names)
6242 {
6243 	struct spdk_xattr	*xattr;
6244 	int			count = 0;
6245 
6246 	TAILQ_FOREACH(xattr, xattrs, link) {
6247 		count++;
6248 	}
6249 
6250 	*names = calloc(1, sizeof(struct spdk_xattr_names) + count * sizeof(char *));
6251 	if (*names == NULL) {
6252 		return -ENOMEM;
6253 	}
6254 
6255 	TAILQ_FOREACH(xattr, xattrs, link) {
6256 		(*names)->names[(*names)->count++] = xattr->name;
6257 	}
6258 
6259 	return 0;
6260 }
6261 
6262 int
6263 spdk_blob_get_xattr_names(struct spdk_blob *blob, struct spdk_xattr_names **names)
6264 {
6265 	_spdk_blob_verify_md_op(blob);
6266 
6267 	return _spdk_blob_get_xattr_names(&blob->xattrs, names);
6268 }
6269 
6270 uint32_t
6271 spdk_xattr_names_get_count(struct spdk_xattr_names *names)
6272 {
6273 	assert(names != NULL);
6274 
6275 	return names->count;
6276 }
6277 
6278 const char *
6279 spdk_xattr_names_get_name(struct spdk_xattr_names *names, uint32_t index)
6280 {
6281 	if (index >= names->count) {
6282 		return NULL;
6283 	}
6284 
6285 	return names->names[index];
6286 }
6287 
6288 void
6289 spdk_xattr_names_free(struct spdk_xattr_names *names)
6290 {
6291 	free(names);
6292 }
6293 
6294 struct spdk_bs_type
6295 spdk_bs_get_bstype(struct spdk_blob_store *bs)
6296 {
6297 	return bs->bstype;
6298 }
6299 
6300 void
6301 spdk_bs_set_bstype(struct spdk_blob_store *bs, struct spdk_bs_type bstype)
6302 {
6303 	memcpy(&bs->bstype, &bstype, sizeof(bstype));
6304 }
6305 
6306 bool
6307 spdk_blob_is_read_only(struct spdk_blob *blob)
6308 {
6309 	assert(blob != NULL);
6310 	return (blob->data_ro || blob->md_ro);
6311 }
6312 
6313 bool
6314 spdk_blob_is_snapshot(struct spdk_blob *blob)
6315 {
6316 	struct spdk_blob_list *snapshot_entry;
6317 
6318 	assert(blob != NULL);
6319 
6320 	snapshot_entry = _spdk_bs_get_snapshot_entry(blob->bs, blob->id);
6321 	if (snapshot_entry == NULL) {
6322 		return false;
6323 	}
6324 
6325 	return true;
6326 }
6327 
6328 bool
6329 spdk_blob_is_clone(struct spdk_blob *blob)
6330 {
6331 	assert(blob != NULL);
6332 
6333 	if (blob->parent_id != SPDK_BLOBID_INVALID) {
6334 		assert(spdk_blob_is_thin_provisioned(blob));
6335 		return true;
6336 	}
6337 
6338 	return false;
6339 }
6340 
6341 bool
6342 spdk_blob_is_thin_provisioned(struct spdk_blob *blob)
6343 {
6344 	assert(blob != NULL);
6345 	return !!(blob->invalid_flags & SPDK_BLOB_THIN_PROV);
6346 }
6347 
6348 static void
6349 _spdk_blob_update_clear_method(struct spdk_blob *blob)
6350 {
6351 	enum blob_clear_method stored_cm;
6352 
6353 	assert(blob != NULL);
6354 
6355 	/* If BLOB_CLEAR_WITH_DEFAULT was passed in, use the setting stored
6356 	 * in metadata previously.  If something other than the default was
6357 	 * specified, ignore stored value and used what was passed in.
6358 	 */
6359 	stored_cm = ((blob->md_ro_flags & SPDK_BLOB_CLEAR_METHOD) >> SPDK_BLOB_CLEAR_METHOD_SHIFT);
6360 
6361 	if (blob->clear_method == BLOB_CLEAR_WITH_DEFAULT) {
6362 		blob->clear_method = stored_cm;
6363 	} else if (blob->clear_method != stored_cm) {
6364 		SPDK_WARNLOG("Using passed in clear method 0x%x instead of stored value of 0x%x\n",
6365 			     blob->clear_method, stored_cm);
6366 	}
6367 }
6368 
6369 spdk_blob_id
6370 spdk_blob_get_parent_snapshot(struct spdk_blob_store *bs, spdk_blob_id blob_id)
6371 {
6372 	struct spdk_blob_list *snapshot_entry = NULL;
6373 	struct spdk_blob_list *clone_entry = NULL;
6374 
6375 	TAILQ_FOREACH(snapshot_entry, &bs->snapshots, link) {
6376 		TAILQ_FOREACH(clone_entry, &snapshot_entry->clones, link) {
6377 			if (clone_entry->id == blob_id) {
6378 				return snapshot_entry->id;
6379 			}
6380 		}
6381 	}
6382 
6383 	return SPDK_BLOBID_INVALID;
6384 }
6385 
6386 int
6387 spdk_blob_get_clones(struct spdk_blob_store *bs, spdk_blob_id blobid, spdk_blob_id *ids,
6388 		     size_t *count)
6389 {
6390 	struct spdk_blob_list *snapshot_entry, *clone_entry;
6391 	size_t n;
6392 
6393 	snapshot_entry = _spdk_bs_get_snapshot_entry(bs, blobid);
6394 	if (snapshot_entry == NULL) {
6395 		*count = 0;
6396 		return 0;
6397 	}
6398 
6399 	if (ids == NULL || *count < snapshot_entry->clone_count) {
6400 		*count = snapshot_entry->clone_count;
6401 		return -ENOMEM;
6402 	}
6403 	*count = snapshot_entry->clone_count;
6404 
6405 	n = 0;
6406 	TAILQ_FOREACH(clone_entry, &snapshot_entry->clones, link) {
6407 		ids[n++] = clone_entry->id;
6408 	}
6409 
6410 	return 0;
6411 }
6412 
6413 SPDK_LOG_REGISTER_COMPONENT("blob", SPDK_LOG_BLOB)
6414