xref: /spdk/lib/blob/blobstore.c (revision 2f8bdb3c820c8a65df378cf33fd92054d7b0fd70)
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 	opts->use_extent_table = false;
171 }
172 
173 void
174 spdk_blob_open_opts_init(struct spdk_blob_open_opts *opts)
175 {
176 	opts->clear_method = BLOB_CLEAR_WITH_DEFAULT;
177 }
178 
179 static struct spdk_blob *
180 _spdk_blob_alloc(struct spdk_blob_store *bs, spdk_blob_id id)
181 {
182 	struct spdk_blob *blob;
183 
184 	blob = calloc(1, sizeof(*blob));
185 	if (!blob) {
186 		return NULL;
187 	}
188 
189 	blob->id = id;
190 	blob->bs = bs;
191 
192 	blob->parent_id = SPDK_BLOBID_INVALID;
193 
194 	blob->state = SPDK_BLOB_STATE_DIRTY;
195 	blob->extent_rle_found = false;
196 	blob->extent_table_found = false;
197 	blob->active.num_pages = 1;
198 	blob->active.pages = calloc(1, sizeof(*blob->active.pages));
199 	if (!blob->active.pages) {
200 		free(blob);
201 		return NULL;
202 	}
203 
204 	blob->active.pages[0] = _spdk_bs_blobid_to_page(id);
205 
206 	TAILQ_INIT(&blob->xattrs);
207 	TAILQ_INIT(&blob->xattrs_internal);
208 
209 	return blob;
210 }
211 
212 static void
213 _spdk_xattrs_free(struct spdk_xattr_tailq *xattrs)
214 {
215 	struct spdk_xattr	*xattr, *xattr_tmp;
216 
217 	TAILQ_FOREACH_SAFE(xattr, xattrs, link, xattr_tmp) {
218 		TAILQ_REMOVE(xattrs, xattr, link);
219 		free(xattr->name);
220 		free(xattr->value);
221 		free(xattr);
222 	}
223 }
224 
225 static void
226 _spdk_blob_free(struct spdk_blob *blob)
227 {
228 	assert(blob != NULL);
229 
230 	free(blob->active.clusters);
231 	free(blob->clean.clusters);
232 	free(blob->active.pages);
233 	free(blob->clean.pages);
234 
235 	_spdk_xattrs_free(&blob->xattrs);
236 	_spdk_xattrs_free(&blob->xattrs_internal);
237 
238 	if (blob->back_bs_dev) {
239 		blob->back_bs_dev->destroy(blob->back_bs_dev);
240 	}
241 
242 	free(blob);
243 }
244 
245 struct freeze_io_ctx {
246 	struct spdk_bs_cpl cpl;
247 	struct spdk_blob *blob;
248 };
249 
250 static void
251 _spdk_blob_io_sync(struct spdk_io_channel_iter *i)
252 {
253 	spdk_for_each_channel_continue(i, 0);
254 }
255 
256 static void
257 _spdk_blob_execute_queued_io(struct spdk_io_channel_iter *i)
258 {
259 	struct spdk_io_channel *_ch = spdk_io_channel_iter_get_channel(i);
260 	struct spdk_bs_channel *ch = spdk_io_channel_get_ctx(_ch);
261 	struct freeze_io_ctx *ctx = spdk_io_channel_iter_get_ctx(i);
262 	struct spdk_bs_request_set	*set;
263 	struct spdk_bs_user_op_args	*args;
264 	spdk_bs_user_op_t *op, *tmp;
265 
266 	TAILQ_FOREACH_SAFE(op, &ch->queued_io, link, tmp) {
267 		set = (struct spdk_bs_request_set *)op;
268 		args = &set->u.user_op;
269 
270 		if (args->blob == ctx->blob) {
271 			TAILQ_REMOVE(&ch->queued_io, op, link);
272 			spdk_bs_user_op_execute(op);
273 		}
274 	}
275 
276 	spdk_for_each_channel_continue(i, 0);
277 }
278 
279 static void
280 _spdk_blob_io_cpl(struct spdk_io_channel_iter *i, int status)
281 {
282 	struct freeze_io_ctx *ctx = spdk_io_channel_iter_get_ctx(i);
283 
284 	ctx->cpl.u.blob_basic.cb_fn(ctx->cpl.u.blob_basic.cb_arg, 0);
285 
286 	free(ctx);
287 }
288 
289 static void
290 _spdk_blob_freeze_io(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
291 {
292 	struct freeze_io_ctx *ctx;
293 
294 	ctx = calloc(1, sizeof(*ctx));
295 	if (!ctx) {
296 		cb_fn(cb_arg, -ENOMEM);
297 		return;
298 	}
299 
300 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
301 	ctx->cpl.u.blob_basic.cb_fn = cb_fn;
302 	ctx->cpl.u.blob_basic.cb_arg = cb_arg;
303 	ctx->blob = blob;
304 
305 	/* Freeze I/O on blob */
306 	blob->frozen_refcnt++;
307 
308 	if (blob->frozen_refcnt == 1) {
309 		spdk_for_each_channel(blob->bs, _spdk_blob_io_sync, ctx, _spdk_blob_io_cpl);
310 	} else {
311 		cb_fn(cb_arg, 0);
312 		free(ctx);
313 	}
314 }
315 
316 static void
317 _spdk_blob_unfreeze_io(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
318 {
319 	struct freeze_io_ctx *ctx;
320 
321 	ctx = calloc(1, sizeof(*ctx));
322 	if (!ctx) {
323 		cb_fn(cb_arg, -ENOMEM);
324 		return;
325 	}
326 
327 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
328 	ctx->cpl.u.blob_basic.cb_fn = cb_fn;
329 	ctx->cpl.u.blob_basic.cb_arg = cb_arg;
330 	ctx->blob = blob;
331 
332 	assert(blob->frozen_refcnt > 0);
333 
334 	blob->frozen_refcnt--;
335 
336 	if (blob->frozen_refcnt == 0) {
337 		spdk_for_each_channel(blob->bs, _spdk_blob_execute_queued_io, ctx, _spdk_blob_io_cpl);
338 	} else {
339 		cb_fn(cb_arg, 0);
340 		free(ctx);
341 	}
342 }
343 
344 static int
345 _spdk_blob_mark_clean(struct spdk_blob *blob)
346 {
347 	uint64_t *clusters = NULL;
348 	uint32_t *pages = NULL;
349 
350 	assert(blob != NULL);
351 
352 	if (blob->active.num_clusters) {
353 		assert(blob->active.clusters);
354 		clusters = calloc(blob->active.num_clusters, sizeof(*blob->active.clusters));
355 		if (!clusters) {
356 			return -ENOMEM;
357 		}
358 		memcpy(clusters, blob->active.clusters, blob->active.num_clusters * sizeof(*blob->active.clusters));
359 	}
360 
361 	if (blob->active.num_pages) {
362 		assert(blob->active.pages);
363 		pages = calloc(blob->active.num_pages, sizeof(*blob->active.pages));
364 		if (!pages) {
365 			free(clusters);
366 			return -ENOMEM;
367 		}
368 		memcpy(pages, blob->active.pages, blob->active.num_pages * sizeof(*blob->active.pages));
369 	}
370 
371 	free(blob->clean.clusters);
372 	free(blob->clean.pages);
373 
374 	blob->clean.num_clusters = blob->active.num_clusters;
375 	blob->clean.clusters = blob->active.clusters;
376 	blob->clean.num_pages = blob->active.num_pages;
377 	blob->clean.pages = blob->active.pages;
378 
379 	blob->active.clusters = clusters;
380 	blob->active.pages = pages;
381 
382 	/* If the metadata was dirtied again while the metadata was being written to disk,
383 	 *  we do not want to revert the DIRTY state back to CLEAN here.
384 	 */
385 	if (blob->state == SPDK_BLOB_STATE_LOADING) {
386 		blob->state = SPDK_BLOB_STATE_CLEAN;
387 	}
388 
389 	return 0;
390 }
391 
392 static int
393 _spdk_blob_deserialize_xattr(struct spdk_blob *blob,
394 			     struct spdk_blob_md_descriptor_xattr *desc_xattr, bool internal)
395 {
396 	struct spdk_xattr                       *xattr;
397 
398 	if (desc_xattr->length != sizeof(desc_xattr->name_length) +
399 	    sizeof(desc_xattr->value_length) +
400 	    desc_xattr->name_length + desc_xattr->value_length) {
401 		return -EINVAL;
402 	}
403 
404 	xattr = calloc(1, sizeof(*xattr));
405 	if (xattr == NULL) {
406 		return -ENOMEM;
407 	}
408 
409 	xattr->name = malloc(desc_xattr->name_length + 1);
410 	if (xattr->name == NULL) {
411 		free(xattr);
412 		return -ENOMEM;
413 	}
414 	memcpy(xattr->name, desc_xattr->name, desc_xattr->name_length);
415 	xattr->name[desc_xattr->name_length] = '\0';
416 
417 	xattr->value = malloc(desc_xattr->value_length);
418 	if (xattr->value == NULL) {
419 		free(xattr->name);
420 		free(xattr);
421 		return -ENOMEM;
422 	}
423 	xattr->value_len = desc_xattr->value_length;
424 	memcpy(xattr->value,
425 	       (void *)((uintptr_t)desc_xattr->name + desc_xattr->name_length),
426 	       desc_xattr->value_length);
427 
428 	TAILQ_INSERT_TAIL(internal ? &blob->xattrs_internal : &blob->xattrs, xattr, link);
429 
430 	return 0;
431 }
432 
433 
434 static int
435 _spdk_blob_parse_page(const struct spdk_blob_md_page *page, struct spdk_blob *blob)
436 {
437 	struct spdk_blob_md_descriptor *desc;
438 	size_t	cur_desc = 0;
439 	void *tmp;
440 
441 	desc = (struct spdk_blob_md_descriptor *)page->descriptors;
442 	while (cur_desc < sizeof(page->descriptors)) {
443 		if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_PADDING) {
444 			if (desc->length == 0) {
445 				/* If padding and length are 0, this terminates the page */
446 				break;
447 			}
448 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_FLAGS) {
449 			struct spdk_blob_md_descriptor_flags	*desc_flags;
450 
451 			desc_flags = (struct spdk_blob_md_descriptor_flags *)desc;
452 
453 			if (desc_flags->length != sizeof(*desc_flags) - sizeof(*desc)) {
454 				return -EINVAL;
455 			}
456 
457 			if ((desc_flags->invalid_flags | SPDK_BLOB_INVALID_FLAGS_MASK) !=
458 			    SPDK_BLOB_INVALID_FLAGS_MASK) {
459 				return -EINVAL;
460 			}
461 
462 			if ((desc_flags->data_ro_flags | SPDK_BLOB_DATA_RO_FLAGS_MASK) !=
463 			    SPDK_BLOB_DATA_RO_FLAGS_MASK) {
464 				blob->data_ro = true;
465 				blob->md_ro = true;
466 			}
467 
468 			if ((desc_flags->md_ro_flags | SPDK_BLOB_MD_RO_FLAGS_MASK) !=
469 			    SPDK_BLOB_MD_RO_FLAGS_MASK) {
470 				blob->md_ro = true;
471 			}
472 
473 			if ((desc_flags->data_ro_flags & SPDK_BLOB_READ_ONLY)) {
474 				blob->data_ro = true;
475 				blob->md_ro = true;
476 			}
477 
478 			blob->invalid_flags = desc_flags->invalid_flags;
479 			blob->data_ro_flags = desc_flags->data_ro_flags;
480 			blob->md_ro_flags = desc_flags->md_ro_flags;
481 
482 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE) {
483 			struct spdk_blob_md_descriptor_extent_rle	*desc_extent_rle;
484 			unsigned int				i, j;
485 			unsigned int				cluster_count = blob->active.num_clusters;
486 
487 			if (blob->extent_table_found) {
488 				/* Extent Table already present in the md,
489 				 * both descriptors should never be at the same time. */
490 				return -EINVAL;
491 			}
492 			blob->extent_rle_found = true;
493 
494 			desc_extent_rle = (struct spdk_blob_md_descriptor_extent_rle *)desc;
495 
496 			if (desc_extent_rle->length == 0 ||
497 			    (desc_extent_rle->length % sizeof(desc_extent_rle->extents[0]) != 0)) {
498 				return -EINVAL;
499 			}
500 
501 			for (i = 0; i < desc_extent_rle->length / sizeof(desc_extent_rle->extents[0]); i++) {
502 				for (j = 0; j < desc_extent_rle->extents[i].length; j++) {
503 					if (desc_extent_rle->extents[i].cluster_idx != 0) {
504 						if (!spdk_bit_array_get(blob->bs->used_clusters,
505 									desc_extent_rle->extents[i].cluster_idx + j)) {
506 							return -EINVAL;
507 						}
508 					}
509 					cluster_count++;
510 				}
511 			}
512 
513 			if (cluster_count == 0) {
514 				return -EINVAL;
515 			}
516 			tmp = realloc(blob->active.clusters, cluster_count * sizeof(*blob->active.clusters));
517 			if (tmp == NULL) {
518 				return -ENOMEM;
519 			}
520 			blob->active.clusters = tmp;
521 			blob->active.cluster_array_size = cluster_count;
522 
523 			for (i = 0; i < desc_extent_rle->length / sizeof(desc_extent_rle->extents[0]); i++) {
524 				for (j = 0; j < desc_extent_rle->extents[i].length; j++) {
525 					if (desc_extent_rle->extents[i].cluster_idx != 0) {
526 						blob->active.clusters[blob->active.num_clusters++] = _spdk_bs_cluster_to_lba(blob->bs,
527 								desc_extent_rle->extents[i].cluster_idx + j);
528 					} else if (spdk_blob_is_thin_provisioned(blob)) {
529 						blob->active.clusters[blob->active.num_clusters++] = 0;
530 					} else {
531 						return -EINVAL;
532 					}
533 				}
534 			}
535 
536 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR) {
537 			int rc;
538 
539 			rc = _spdk_blob_deserialize_xattr(blob,
540 							  (struct spdk_blob_md_descriptor_xattr *) desc, false);
541 			if (rc != 0) {
542 				return rc;
543 			}
544 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL) {
545 			int rc;
546 
547 			rc = _spdk_blob_deserialize_xattr(blob,
548 							  (struct spdk_blob_md_descriptor_xattr *) desc, true);
549 			if (rc != 0) {
550 				return rc;
551 			}
552 		} else {
553 			/* Unrecognized descriptor type.  Do not fail - just continue to the
554 			 *  next descriptor.  If this descriptor is associated with some feature
555 			 *  defined in a newer version of blobstore, that version of blobstore
556 			 *  should create and set an associated feature flag to specify if this
557 			 *  blob can be loaded or not.
558 			 */
559 		}
560 
561 		/* Advance to the next descriptor */
562 		cur_desc += sizeof(*desc) + desc->length;
563 		if (cur_desc + sizeof(*desc) > sizeof(page->descriptors)) {
564 			break;
565 		}
566 		desc = (struct spdk_blob_md_descriptor *)((uintptr_t)page->descriptors + cur_desc);
567 	}
568 
569 	return 0;
570 }
571 
572 static int
573 _spdk_blob_parse(const struct spdk_blob_md_page *pages, uint32_t page_count,
574 		 struct spdk_blob *blob)
575 {
576 	const struct spdk_blob_md_page *page;
577 	uint32_t i;
578 	int rc;
579 
580 	assert(page_count > 0);
581 	assert(pages[0].sequence_num == 0);
582 	assert(blob != NULL);
583 	assert(blob->state == SPDK_BLOB_STATE_LOADING);
584 	assert(blob->active.clusters == NULL);
585 
586 	/* The blobid provided doesn't match what's in the MD, this can
587 	 * happen for example if a bogus blobid is passed in through open.
588 	 */
589 	if (blob->id != pages[0].id) {
590 		SPDK_ERRLOG("Blobid (%lu) doesn't match what's in metadata (%lu)\n",
591 			    blob->id, pages[0].id);
592 		return -ENOENT;
593 	}
594 
595 	for (i = 0; i < page_count; i++) {
596 		page = &pages[i];
597 
598 		assert(page->id == blob->id);
599 		assert(page->sequence_num == i);
600 
601 		rc = _spdk_blob_parse_page(page, blob);
602 		if (rc != 0) {
603 			return rc;
604 		}
605 	}
606 
607 	return 0;
608 }
609 
610 static int
611 _spdk_blob_serialize_add_page(const struct spdk_blob *blob,
612 			      struct spdk_blob_md_page **pages,
613 			      uint32_t *page_count,
614 			      struct spdk_blob_md_page **last_page)
615 {
616 	struct spdk_blob_md_page *page;
617 
618 	assert(pages != NULL);
619 	assert(page_count != NULL);
620 
621 	if (*page_count == 0) {
622 		assert(*pages == NULL);
623 		*page_count = 1;
624 		*pages = spdk_malloc(SPDK_BS_PAGE_SIZE, SPDK_BS_PAGE_SIZE,
625 				     NULL, SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
626 	} else {
627 		assert(*pages != NULL);
628 		(*page_count)++;
629 		*pages = spdk_realloc(*pages,
630 				      SPDK_BS_PAGE_SIZE * (*page_count),
631 				      SPDK_BS_PAGE_SIZE);
632 	}
633 
634 	if (*pages == NULL) {
635 		*page_count = 0;
636 		*last_page = NULL;
637 		return -ENOMEM;
638 	}
639 
640 	page = &(*pages)[*page_count - 1];
641 	memset(page, 0, sizeof(*page));
642 	page->id = blob->id;
643 	page->sequence_num = *page_count - 1;
644 	page->next = SPDK_INVALID_MD_PAGE;
645 	*last_page = page;
646 
647 	return 0;
648 }
649 
650 /* Transform the in-memory representation 'xattr' into an on-disk xattr descriptor.
651  * Update required_sz on both success and failure.
652  *
653  */
654 static int
655 _spdk_blob_serialize_xattr(const struct spdk_xattr *xattr,
656 			   uint8_t *buf, size_t buf_sz,
657 			   size_t *required_sz, bool internal)
658 {
659 	struct spdk_blob_md_descriptor_xattr	*desc;
660 
661 	*required_sz = sizeof(struct spdk_blob_md_descriptor_xattr) +
662 		       strlen(xattr->name) +
663 		       xattr->value_len;
664 
665 	if (buf_sz < *required_sz) {
666 		return -1;
667 	}
668 
669 	desc = (struct spdk_blob_md_descriptor_xattr *)buf;
670 
671 	desc->type = internal ? SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL : SPDK_MD_DESCRIPTOR_TYPE_XATTR;
672 	desc->length = sizeof(desc->name_length) +
673 		       sizeof(desc->value_length) +
674 		       strlen(xattr->name) +
675 		       xattr->value_len;
676 	desc->name_length = strlen(xattr->name);
677 	desc->value_length = xattr->value_len;
678 
679 	memcpy(desc->name, xattr->name, desc->name_length);
680 	memcpy((void *)((uintptr_t)desc->name + desc->name_length),
681 	       xattr->value,
682 	       desc->value_length);
683 
684 	return 0;
685 }
686 
687 static void
688 _spdk_blob_serialize_extent_rle(const struct spdk_blob *blob,
689 				uint64_t start_cluster, uint64_t *next_cluster,
690 				uint8_t **buf, size_t *buf_sz)
691 {
692 	struct spdk_blob_md_descriptor_extent_rle *desc_extent_rle;
693 	size_t cur_sz;
694 	uint64_t i, extent_idx;
695 	uint64_t lba, lba_per_cluster, lba_count;
696 
697 	/* The buffer must have room for at least one extent */
698 	cur_sz = sizeof(struct spdk_blob_md_descriptor) + sizeof(desc_extent_rle->extents[0]);
699 	if (*buf_sz < cur_sz) {
700 		*next_cluster = start_cluster;
701 		return;
702 	}
703 
704 	desc_extent_rle = (struct spdk_blob_md_descriptor_extent_rle *)*buf;
705 	desc_extent_rle->type = SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE;
706 
707 	lba_per_cluster = _spdk_bs_cluster_to_lba(blob->bs, 1);
708 
709 	lba = blob->active.clusters[start_cluster];
710 	lba_count = lba_per_cluster;
711 	extent_idx = 0;
712 	for (i = start_cluster + 1; i < blob->active.num_clusters; i++) {
713 		if ((lba + lba_count) == blob->active.clusters[i] && lba != 0) {
714 			/* Run-length encode sequential non-zero LBA */
715 			lba_count += lba_per_cluster;
716 			continue;
717 		} else if (lba == 0 && blob->active.clusters[i] == 0) {
718 			/* Run-length encode unallocated clusters */
719 			lba_count += lba_per_cluster;
720 			continue;
721 		}
722 		desc_extent_rle->extents[extent_idx].cluster_idx = lba / lba_per_cluster;
723 		desc_extent_rle->extents[extent_idx].length = lba_count / lba_per_cluster;
724 		extent_idx++;
725 
726 		cur_sz += sizeof(desc_extent_rle->extents[extent_idx]);
727 
728 		if (*buf_sz < cur_sz) {
729 			/* If we ran out of buffer space, return */
730 			*next_cluster = i;
731 			break;
732 		}
733 
734 		lba = blob->active.clusters[i];
735 		lba_count = lba_per_cluster;
736 	}
737 
738 	if (*buf_sz >= cur_sz) {
739 		desc_extent_rle->extents[extent_idx].cluster_idx = lba / lba_per_cluster;
740 		desc_extent_rle->extents[extent_idx].length = lba_count / lba_per_cluster;
741 		extent_idx++;
742 
743 		*next_cluster = blob->active.num_clusters;
744 	}
745 
746 	desc_extent_rle->length = sizeof(desc_extent_rle->extents[0]) * extent_idx;
747 	*buf_sz -= sizeof(struct spdk_blob_md_descriptor) + desc_extent_rle->length;
748 	*buf += sizeof(struct spdk_blob_md_descriptor) + desc_extent_rle->length;
749 }
750 
751 static int
752 _spdk_blob_serialize_extents_rle(const struct spdk_blob *blob,
753 				 struct spdk_blob_md_page **pages,
754 				 struct spdk_blob_md_page *cur_page,
755 				 uint32_t *page_count, uint8_t **buf,
756 				 size_t *remaining_sz)
757 {
758 	uint64_t				last_cluster;
759 	int					rc;
760 
761 	last_cluster = 0;
762 	while (last_cluster < blob->active.num_clusters) {
763 		_spdk_blob_serialize_extent_rle(blob, last_cluster, &last_cluster, buf, remaining_sz);
764 
765 		if (last_cluster == blob->active.num_clusters) {
766 			break;
767 		}
768 
769 		rc = _spdk_blob_serialize_add_page(blob, pages, page_count, &cur_page);
770 		if (rc < 0) {
771 			return rc;
772 		}
773 
774 		*buf = (uint8_t *)cur_page->descriptors;
775 		*remaining_sz = sizeof(cur_page->descriptors);
776 	}
777 
778 	return 0;
779 }
780 
781 static void
782 _spdk_blob_serialize_flags(const struct spdk_blob *blob,
783 			   uint8_t *buf, size_t *buf_sz)
784 {
785 	struct spdk_blob_md_descriptor_flags *desc;
786 
787 	/*
788 	 * Flags get serialized first, so we should always have room for the flags
789 	 *  descriptor.
790 	 */
791 	assert(*buf_sz >= sizeof(*desc));
792 
793 	desc = (struct spdk_blob_md_descriptor_flags *)buf;
794 	desc->type = SPDK_MD_DESCRIPTOR_TYPE_FLAGS;
795 	desc->length = sizeof(*desc) - sizeof(struct spdk_blob_md_descriptor);
796 	desc->invalid_flags = blob->invalid_flags;
797 	desc->data_ro_flags = blob->data_ro_flags;
798 	desc->md_ro_flags = blob->md_ro_flags;
799 
800 	*buf_sz -= sizeof(*desc);
801 }
802 
803 static int
804 _spdk_blob_serialize_xattrs(const struct spdk_blob *blob,
805 			    const struct spdk_xattr_tailq *xattrs, bool internal,
806 			    struct spdk_blob_md_page **pages,
807 			    struct spdk_blob_md_page *cur_page,
808 			    uint32_t *page_count, uint8_t **buf,
809 			    size_t *remaining_sz)
810 {
811 	const struct spdk_xattr	*xattr;
812 	int	rc;
813 
814 	TAILQ_FOREACH(xattr, xattrs, link) {
815 		size_t required_sz = 0;
816 
817 		rc = _spdk_blob_serialize_xattr(xattr,
818 						*buf, *remaining_sz,
819 						&required_sz, internal);
820 		if (rc < 0) {
821 			/* Need to add a new page to the chain */
822 			rc = _spdk_blob_serialize_add_page(blob, pages, page_count,
823 							   &cur_page);
824 			if (rc < 0) {
825 				spdk_free(*pages);
826 				*pages = NULL;
827 				*page_count = 0;
828 				return rc;
829 			}
830 
831 			*buf = (uint8_t *)cur_page->descriptors;
832 			*remaining_sz = sizeof(cur_page->descriptors);
833 
834 			/* Try again */
835 			required_sz = 0;
836 			rc = _spdk_blob_serialize_xattr(xattr,
837 							*buf, *remaining_sz,
838 							&required_sz, internal);
839 
840 			if (rc < 0) {
841 				spdk_free(*pages);
842 				*pages = NULL;
843 				*page_count = 0;
844 				return rc;
845 			}
846 		}
847 
848 		*remaining_sz -= required_sz;
849 		*buf += required_sz;
850 	}
851 
852 	return 0;
853 }
854 
855 static int
856 _spdk_blob_serialize(const struct spdk_blob *blob, struct spdk_blob_md_page **pages,
857 		     uint32_t *page_count)
858 {
859 	struct spdk_blob_md_page		*cur_page;
860 	int					rc;
861 	uint8_t					*buf;
862 	size_t					remaining_sz;
863 
864 	assert(pages != NULL);
865 	assert(page_count != NULL);
866 	assert(blob != NULL);
867 	assert(blob->state == SPDK_BLOB_STATE_DIRTY);
868 
869 	*pages = NULL;
870 	*page_count = 0;
871 
872 	/* A blob always has at least 1 page, even if it has no descriptors */
873 	rc = _spdk_blob_serialize_add_page(blob, pages, page_count, &cur_page);
874 	if (rc < 0) {
875 		return rc;
876 	}
877 
878 	buf = (uint8_t *)cur_page->descriptors;
879 	remaining_sz = sizeof(cur_page->descriptors);
880 
881 	/* Serialize flags */
882 	_spdk_blob_serialize_flags(blob, buf, &remaining_sz);
883 	buf += sizeof(struct spdk_blob_md_descriptor_flags);
884 
885 	/* Serialize xattrs */
886 	rc = _spdk_blob_serialize_xattrs(blob, &blob->xattrs, false,
887 					 pages, cur_page, page_count, &buf, &remaining_sz);
888 	if (rc < 0) {
889 		return rc;
890 	}
891 
892 	/* Serialize internal xattrs */
893 	rc = _spdk_blob_serialize_xattrs(blob, &blob->xattrs_internal, true,
894 					 pages, cur_page, page_count, &buf, &remaining_sz);
895 	if (rc < 0) {
896 		return rc;
897 	}
898 
899 	if (blob->use_extent_table) {
900 		/* Serialization as extent pages is not yet implemented */
901 		assert(false);
902 		rc = -ENOSYS;
903 	} else {
904 		/* Serialize extents */
905 		rc = _spdk_blob_serialize_extents_rle(blob, pages, cur_page, page_count, &buf, &remaining_sz);
906 	}
907 
908 	return rc;
909 }
910 
911 struct spdk_blob_load_ctx {
912 	struct spdk_blob		*blob;
913 
914 	struct spdk_blob_md_page	*pages;
915 	uint32_t			num_pages;
916 	spdk_bs_sequence_t	        *seq;
917 
918 	spdk_bs_sequence_cpl		cb_fn;
919 	void				*cb_arg;
920 };
921 
922 static uint32_t
923 _spdk_blob_md_page_calc_crc(void *page)
924 {
925 	uint32_t		crc;
926 
927 	crc = BLOB_CRC32C_INITIAL;
928 	crc = spdk_crc32c_update(page, SPDK_BS_PAGE_SIZE - 4, crc);
929 	crc ^= BLOB_CRC32C_INITIAL;
930 
931 	return crc;
932 
933 }
934 
935 static void
936 _spdk_blob_load_final(void *cb_arg, int bserrno)
937 {
938 	struct spdk_blob_load_ctx	*ctx = cb_arg;
939 	struct spdk_blob		*blob = ctx->blob;
940 
941 	if (bserrno == 0) {
942 		_spdk_blob_mark_clean(blob);
943 	}
944 
945 	ctx->cb_fn(ctx->seq, ctx->cb_arg, bserrno);
946 
947 	/* Free the memory */
948 	spdk_free(ctx->pages);
949 	free(ctx);
950 }
951 
952 static void
953 _spdk_blob_load_snapshot_cpl(void *cb_arg, struct spdk_blob *snapshot, int bserrno)
954 {
955 	struct spdk_blob_load_ctx	*ctx = cb_arg;
956 	struct spdk_blob		*blob = ctx->blob;
957 
958 	if (bserrno == 0) {
959 		blob->back_bs_dev = spdk_bs_create_blob_bs_dev(snapshot);
960 		if (blob->back_bs_dev == NULL) {
961 			bserrno = -ENOMEM;
962 		}
963 	}
964 	if (bserrno != 0) {
965 		SPDK_ERRLOG("Snapshot fail\n");
966 	}
967 
968 	_spdk_blob_load_final(ctx, bserrno);
969 }
970 
971 static void _spdk_blob_update_clear_method(struct spdk_blob *blob);
972 
973 static void
974 _spdk_blob_load_backing_dev(void *cb_arg)
975 {
976 	struct spdk_blob_load_ctx	*ctx = cb_arg;
977 	struct spdk_blob		*blob = ctx->blob;
978 	const void			*value;
979 	size_t				len;
980 	int				rc;
981 
982 	if (spdk_blob_is_thin_provisioned(blob)) {
983 		rc = _spdk_blob_get_xattr_value(blob, BLOB_SNAPSHOT, &value, &len, true);
984 		if (rc == 0) {
985 			if (len != sizeof(spdk_blob_id)) {
986 				_spdk_blob_load_final(ctx, -EINVAL);
987 				return;
988 			}
989 			/* open snapshot blob and continue in the callback function */
990 			blob->parent_id = *(spdk_blob_id *)value;
991 			spdk_bs_open_blob(blob->bs, blob->parent_id,
992 					  _spdk_blob_load_snapshot_cpl, ctx);
993 			return;
994 		} else {
995 			/* add zeroes_dev for thin provisioned blob */
996 			blob->back_bs_dev = spdk_bs_create_zeroes_dev();
997 		}
998 	} else {
999 		/* standard blob */
1000 		blob->back_bs_dev = NULL;
1001 	}
1002 	_spdk_blob_load_final(ctx, 0);
1003 }
1004 
1005 static void
1006 _spdk_blob_load_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1007 {
1008 	struct spdk_blob_load_ctx	*ctx = cb_arg;
1009 	struct spdk_blob		*blob = ctx->blob;
1010 	struct spdk_blob_md_page	*page;
1011 	int				rc;
1012 	uint32_t			crc;
1013 
1014 	if (bserrno) {
1015 		SPDK_ERRLOG("Metadata page read failed: %d\n", bserrno);
1016 		_spdk_blob_load_final(ctx, bserrno);
1017 		return;
1018 	}
1019 
1020 	page = &ctx->pages[ctx->num_pages - 1];
1021 	crc = _spdk_blob_md_page_calc_crc(page);
1022 	if (crc != page->crc) {
1023 		SPDK_ERRLOG("Metadata page %d crc mismatch\n", ctx->num_pages);
1024 		_spdk_blob_load_final(ctx, -EINVAL);
1025 		return;
1026 	}
1027 
1028 	if (page->next != SPDK_INVALID_MD_PAGE) {
1029 		uint32_t next_page = page->next;
1030 		uint64_t next_lba = _spdk_bs_md_page_to_lba(blob->bs, next_page);
1031 
1032 		/* Read the next page */
1033 		ctx->num_pages++;
1034 		ctx->pages = spdk_realloc(ctx->pages, (sizeof(*page) * ctx->num_pages),
1035 					  sizeof(*page));
1036 		if (ctx->pages == NULL) {
1037 			_spdk_blob_load_final(ctx, -ENOMEM);
1038 			return;
1039 		}
1040 
1041 		spdk_bs_sequence_read_dev(seq, &ctx->pages[ctx->num_pages - 1],
1042 					  next_lba,
1043 					  _spdk_bs_byte_to_lba(blob->bs, sizeof(*page)),
1044 					  _spdk_blob_load_cpl, ctx);
1045 		return;
1046 	}
1047 
1048 	/* Parse the pages */
1049 	rc = _spdk_blob_parse(ctx->pages, ctx->num_pages, blob);
1050 	if (rc) {
1051 		_spdk_blob_load_final(ctx, rc);
1052 		return;
1053 	}
1054 
1055 	if (blob->extent_table_found == true) {
1056 		/* If EXTENT_TABLE was found, that means support for it should be enabled. */
1057 		assert(blob->extent_rle_found == false);
1058 		blob->use_extent_table = true;
1059 	} else {
1060 		/* If EXTENT_RLE or no extent_* descriptor was found disable support
1061 		 * for extent table. No extent_* descriptors means that blob has length of 0
1062 		 * and no extent_rle descriptors were persisted for it.
1063 		 * EXTENT_TABLE if used, is always present in metadata regardless of length. */
1064 		blob->use_extent_table = false;
1065 	}
1066 
1067 	ctx->seq = seq;
1068 
1069 	/* Check the clear_method stored in metadata vs what may have been passed
1070 	 * via spdk_bs_open_blob_ext() and update accordingly.
1071 	 */
1072 	_spdk_blob_update_clear_method(blob);
1073 
1074 	_spdk_blob_load_backing_dev(ctx);
1075 }
1076 
1077 /* Load a blob from disk given a blobid */
1078 static void
1079 _spdk_blob_load(spdk_bs_sequence_t *seq, struct spdk_blob *blob,
1080 		spdk_bs_sequence_cpl cb_fn, void *cb_arg)
1081 {
1082 	struct spdk_blob_load_ctx *ctx;
1083 	struct spdk_blob_store *bs;
1084 	uint32_t page_num;
1085 	uint64_t lba;
1086 
1087 	_spdk_blob_verify_md_op(blob);
1088 
1089 	bs = blob->bs;
1090 
1091 	ctx = calloc(1, sizeof(*ctx));
1092 	if (!ctx) {
1093 		cb_fn(seq, cb_arg, -ENOMEM);
1094 		return;
1095 	}
1096 
1097 	ctx->blob = blob;
1098 	ctx->pages = spdk_realloc(ctx->pages, SPDK_BS_PAGE_SIZE, SPDK_BS_PAGE_SIZE);
1099 	if (!ctx->pages) {
1100 		free(ctx);
1101 		cb_fn(seq, cb_arg, -ENOMEM);
1102 		return;
1103 	}
1104 	ctx->num_pages = 1;
1105 	ctx->cb_fn = cb_fn;
1106 	ctx->cb_arg = cb_arg;
1107 	ctx->seq = seq;
1108 
1109 	page_num = _spdk_bs_blobid_to_page(blob->id);
1110 	lba = _spdk_bs_md_page_to_lba(blob->bs, page_num);
1111 
1112 	blob->state = SPDK_BLOB_STATE_LOADING;
1113 
1114 	spdk_bs_sequence_read_dev(seq, &ctx->pages[0], lba,
1115 				  _spdk_bs_byte_to_lba(bs, SPDK_BS_PAGE_SIZE),
1116 				  _spdk_blob_load_cpl, ctx);
1117 }
1118 
1119 struct spdk_blob_persist_ctx {
1120 	struct spdk_blob		*blob;
1121 
1122 	struct spdk_bs_super_block	*super;
1123 
1124 	struct spdk_blob_md_page	*pages;
1125 
1126 	spdk_bs_sequence_t		*seq;
1127 	spdk_bs_sequence_cpl		cb_fn;
1128 	void				*cb_arg;
1129 };
1130 
1131 static void
1132 spdk_bs_batch_clear_dev(struct spdk_blob_persist_ctx *ctx, spdk_bs_batch_t *batch, uint64_t lba,
1133 			uint32_t lba_count)
1134 {
1135 	switch (ctx->blob->clear_method) {
1136 	case BLOB_CLEAR_WITH_DEFAULT:
1137 	case BLOB_CLEAR_WITH_UNMAP:
1138 		spdk_bs_batch_unmap_dev(batch, lba, lba_count);
1139 		break;
1140 	case BLOB_CLEAR_WITH_WRITE_ZEROES:
1141 		spdk_bs_batch_write_zeroes_dev(batch, lba, lba_count);
1142 		break;
1143 	case BLOB_CLEAR_WITH_NONE:
1144 	default:
1145 		break;
1146 	}
1147 }
1148 
1149 static void
1150 _spdk_blob_persist_complete(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1151 {
1152 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1153 	struct spdk_blob		*blob = ctx->blob;
1154 
1155 	if (bserrno == 0) {
1156 		_spdk_blob_mark_clean(blob);
1157 	}
1158 
1159 	/* Call user callback */
1160 	ctx->cb_fn(seq, ctx->cb_arg, bserrno);
1161 
1162 	/* Free the memory */
1163 	spdk_free(ctx->pages);
1164 	free(ctx);
1165 }
1166 
1167 static void
1168 _spdk_blob_persist_clear_clusters_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1169 {
1170 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1171 	struct spdk_blob		*blob = ctx->blob;
1172 	struct spdk_blob_store		*bs = blob->bs;
1173 	size_t				i;
1174 
1175 	/* Release all clusters that were truncated */
1176 	for (i = blob->active.num_clusters; i < blob->active.cluster_array_size; i++) {
1177 		uint32_t cluster_num = _spdk_bs_lba_to_cluster(bs, blob->active.clusters[i]);
1178 
1179 		/* Nothing to release if it was not allocated */
1180 		if (blob->active.clusters[i] != 0) {
1181 			_spdk_bs_release_cluster(bs, cluster_num);
1182 		}
1183 	}
1184 
1185 	if (blob->active.num_clusters == 0) {
1186 		free(blob->active.clusters);
1187 		blob->active.clusters = NULL;
1188 		blob->active.cluster_array_size = 0;
1189 	} else if (blob->active.num_clusters != blob->active.cluster_array_size) {
1190 #ifndef __clang_analyzer__
1191 		void *tmp;
1192 
1193 		/* scan-build really can't figure reallocs, workaround it */
1194 		tmp = realloc(blob->active.clusters, sizeof(*blob->active.clusters) * blob->active.num_clusters);
1195 		assert(tmp != NULL);
1196 		blob->active.clusters = tmp;
1197 #endif
1198 		blob->active.cluster_array_size = blob->active.num_clusters;
1199 	}
1200 
1201 	_spdk_blob_persist_complete(seq, ctx, bserrno);
1202 }
1203 
1204 static void
1205 _spdk_blob_persist_clear_clusters(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1206 {
1207 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1208 	struct spdk_blob		*blob = ctx->blob;
1209 	struct spdk_blob_store		*bs = blob->bs;
1210 	spdk_bs_batch_t			*batch;
1211 	size_t				i;
1212 	uint64_t			lba;
1213 	uint32_t			lba_count;
1214 
1215 	/* Clusters don't move around in blobs. The list shrinks or grows
1216 	 * at the end, but no changes ever occur in the middle of the list.
1217 	 */
1218 
1219 	batch = spdk_bs_sequence_to_batch(seq, _spdk_blob_persist_clear_clusters_cpl, ctx);
1220 
1221 	/* Clear all clusters that were truncated */
1222 	lba = 0;
1223 	lba_count = 0;
1224 	for (i = blob->active.num_clusters; i < blob->active.cluster_array_size; i++) {
1225 		uint64_t next_lba = blob->active.clusters[i];
1226 		uint32_t next_lba_count = _spdk_bs_cluster_to_lba(bs, 1);
1227 
1228 		if (next_lba > 0 && (lba + lba_count) == next_lba) {
1229 			/* This cluster is contiguous with the previous one. */
1230 			lba_count += next_lba_count;
1231 			continue;
1232 		}
1233 
1234 		/* This cluster is not contiguous with the previous one. */
1235 
1236 		/* If a run of LBAs previously existing, clear them now */
1237 		if (lba_count > 0) {
1238 			spdk_bs_batch_clear_dev(ctx, batch, lba, lba_count);
1239 		}
1240 
1241 		/* Start building the next batch */
1242 		lba = next_lba;
1243 		if (next_lba > 0) {
1244 			lba_count = next_lba_count;
1245 		} else {
1246 			lba_count = 0;
1247 		}
1248 	}
1249 
1250 	/* If we ended with a contiguous set of LBAs, clear them now */
1251 	if (lba_count > 0) {
1252 		spdk_bs_batch_clear_dev(ctx, batch, lba, lba_count);
1253 	}
1254 
1255 	spdk_bs_batch_close(batch);
1256 }
1257 
1258 static void
1259 _spdk_blob_persist_zero_pages_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1260 {
1261 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1262 	struct spdk_blob		*blob = ctx->blob;
1263 	struct spdk_blob_store		*bs = blob->bs;
1264 	size_t				i;
1265 
1266 	/* This loop starts at 1 because the first page is special and handled
1267 	 * below. The pages (except the first) are never written in place,
1268 	 * so any pages in the clean list must be zeroed.
1269 	 */
1270 	for (i = 1; i < blob->clean.num_pages; i++) {
1271 		spdk_bit_array_clear(bs->used_md_pages, blob->clean.pages[i]);
1272 	}
1273 
1274 	if (blob->active.num_pages == 0) {
1275 		uint32_t page_num;
1276 
1277 		page_num = _spdk_bs_blobid_to_page(blob->id);
1278 		spdk_bit_array_clear(bs->used_md_pages, page_num);
1279 	}
1280 
1281 	/* Move on to clearing clusters */
1282 	_spdk_blob_persist_clear_clusters(seq, ctx, 0);
1283 }
1284 
1285 static void
1286 _spdk_blob_persist_zero_pages(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1287 {
1288 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1289 	struct spdk_blob		*blob = ctx->blob;
1290 	struct spdk_blob_store		*bs = blob->bs;
1291 	uint64_t			lba;
1292 	uint32_t			lba_count;
1293 	spdk_bs_batch_t			*batch;
1294 	size_t				i;
1295 
1296 	batch = spdk_bs_sequence_to_batch(seq, _spdk_blob_persist_zero_pages_cpl, ctx);
1297 
1298 	lba_count = _spdk_bs_byte_to_lba(bs, SPDK_BS_PAGE_SIZE);
1299 
1300 	/* This loop starts at 1 because the first page is special and handled
1301 	 * below. The pages (except the first) are never written in place,
1302 	 * so any pages in the clean list must be zeroed.
1303 	 */
1304 	for (i = 1; i < blob->clean.num_pages; i++) {
1305 		lba = _spdk_bs_md_page_to_lba(bs, blob->clean.pages[i]);
1306 
1307 		spdk_bs_batch_write_zeroes_dev(batch, lba, lba_count);
1308 	}
1309 
1310 	/* The first page will only be zeroed if this is a delete. */
1311 	if (blob->active.num_pages == 0) {
1312 		uint32_t page_num;
1313 
1314 		/* The first page in the metadata goes where the blobid indicates */
1315 		page_num = _spdk_bs_blobid_to_page(blob->id);
1316 		lba = _spdk_bs_md_page_to_lba(bs, page_num);
1317 
1318 		spdk_bs_batch_write_zeroes_dev(batch, lba, lba_count);
1319 	}
1320 
1321 	spdk_bs_batch_close(batch);
1322 }
1323 
1324 static void
1325 _spdk_blob_persist_write_page_root(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1326 {
1327 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1328 	struct spdk_blob		*blob = ctx->blob;
1329 	struct spdk_blob_store		*bs = blob->bs;
1330 	uint64_t			lba;
1331 	uint32_t			lba_count;
1332 	struct spdk_blob_md_page	*page;
1333 
1334 	if (blob->active.num_pages == 0) {
1335 		/* Move on to the next step */
1336 		_spdk_blob_persist_zero_pages(seq, ctx, 0);
1337 		return;
1338 	}
1339 
1340 	lba_count = _spdk_bs_byte_to_lba(bs, sizeof(*page));
1341 
1342 	page = &ctx->pages[0];
1343 	/* The first page in the metadata goes where the blobid indicates */
1344 	lba = _spdk_bs_md_page_to_lba(bs, _spdk_bs_blobid_to_page(blob->id));
1345 
1346 	spdk_bs_sequence_write_dev(seq, page, lba, lba_count,
1347 				   _spdk_blob_persist_zero_pages, ctx);
1348 }
1349 
1350 static void
1351 _spdk_blob_persist_write_page_chain(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1352 {
1353 	struct spdk_blob_persist_ctx	*ctx = cb_arg;
1354 	struct spdk_blob		*blob = ctx->blob;
1355 	struct spdk_blob_store		*bs = blob->bs;
1356 	uint64_t			lba;
1357 	uint32_t			lba_count;
1358 	struct spdk_blob_md_page	*page;
1359 	spdk_bs_batch_t			*batch;
1360 	size_t				i;
1361 
1362 	/* Clusters don't move around in blobs. The list shrinks or grows
1363 	 * at the end, but no changes ever occur in the middle of the list.
1364 	 */
1365 
1366 	lba_count = _spdk_bs_byte_to_lba(bs, sizeof(*page));
1367 
1368 	batch = spdk_bs_sequence_to_batch(seq, _spdk_blob_persist_write_page_root, ctx);
1369 
1370 	/* This starts at 1. The root page is not written until
1371 	 * all of the others are finished
1372 	 */
1373 	for (i = 1; i < blob->active.num_pages; i++) {
1374 		page = &ctx->pages[i];
1375 		assert(page->sequence_num == i);
1376 
1377 		lba = _spdk_bs_md_page_to_lba(bs, blob->active.pages[i]);
1378 
1379 		spdk_bs_batch_write_dev(batch, page, lba, lba_count);
1380 	}
1381 
1382 	spdk_bs_batch_close(batch);
1383 }
1384 
1385 static int
1386 _spdk_blob_resize(struct spdk_blob *blob, uint64_t sz)
1387 {
1388 	uint64_t	i;
1389 	uint64_t	*tmp;
1390 	uint64_t	lfc; /* lowest free cluster */
1391 	uint64_t	num_clusters;
1392 	struct spdk_blob_store *bs;
1393 
1394 	bs = blob->bs;
1395 
1396 	_spdk_blob_verify_md_op(blob);
1397 
1398 	if (blob->active.num_clusters == sz) {
1399 		return 0;
1400 	}
1401 
1402 	if (blob->active.num_clusters < blob->active.cluster_array_size) {
1403 		/* If this blob was resized to be larger, then smaller, then
1404 		 * larger without syncing, then the cluster array already
1405 		 * contains spare assigned clusters we can use.
1406 		 */
1407 		num_clusters = spdk_min(blob->active.cluster_array_size,
1408 					sz);
1409 	} else {
1410 		num_clusters = blob->active.num_clusters;
1411 	}
1412 
1413 	/* Do two passes - one to verify that we can obtain enough clusters
1414 	 * and another to actually claim them.
1415 	 */
1416 
1417 	if (spdk_blob_is_thin_provisioned(blob) == false) {
1418 		lfc = 0;
1419 		for (i = num_clusters; i < sz; i++) {
1420 			lfc = spdk_bit_array_find_first_clear(bs->used_clusters, lfc);
1421 			if (lfc == UINT32_MAX) {
1422 				/* No more free clusters. Cannot satisfy the request */
1423 				return -ENOSPC;
1424 			}
1425 			lfc++;
1426 		}
1427 	}
1428 
1429 	if (sz > num_clusters) {
1430 		/* Expand the cluster array if necessary.
1431 		 * We only shrink the array when persisting.
1432 		 */
1433 		tmp = realloc(blob->active.clusters, sizeof(*blob->active.clusters) * sz);
1434 		if (sz > 0 && tmp == NULL) {
1435 			return -ENOMEM;
1436 		}
1437 		memset(tmp + blob->active.cluster_array_size, 0,
1438 		       sizeof(*blob->active.clusters) * (sz - blob->active.cluster_array_size));
1439 		blob->active.clusters = tmp;
1440 		blob->active.cluster_array_size = sz;
1441 	}
1442 
1443 	blob->state = SPDK_BLOB_STATE_DIRTY;
1444 
1445 	if (spdk_blob_is_thin_provisioned(blob) == false) {
1446 		lfc = 0;
1447 		for (i = num_clusters; i < sz; i++) {
1448 			_spdk_bs_allocate_cluster(blob, i, &lfc, true);
1449 			lfc++;
1450 		}
1451 	}
1452 
1453 	blob->active.num_clusters = sz;
1454 
1455 	return 0;
1456 }
1457 
1458 static void
1459 _spdk_blob_persist_generate_new_md(struct spdk_blob_persist_ctx *ctx)
1460 {
1461 	spdk_bs_sequence_t *seq = ctx->seq;
1462 	struct spdk_blob *blob = ctx->blob;
1463 	struct spdk_blob_store *bs = blob->bs;
1464 	uint64_t i;
1465 	uint32_t page_num;
1466 	void *tmp;
1467 	int rc;
1468 
1469 	/* Generate the new metadata */
1470 	rc = _spdk_blob_serialize(blob, &ctx->pages, &blob->active.num_pages);
1471 	if (rc < 0) {
1472 		_spdk_blob_persist_complete(seq, ctx, rc);
1473 		return;
1474 	}
1475 
1476 	assert(blob->active.num_pages >= 1);
1477 
1478 	/* Resize the cache of page indices */
1479 	tmp = realloc(blob->active.pages, blob->active.num_pages * sizeof(*blob->active.pages));
1480 	if (!tmp) {
1481 		_spdk_blob_persist_complete(seq, ctx, -ENOMEM);
1482 		return;
1483 	}
1484 	blob->active.pages = tmp;
1485 
1486 	/* Assign this metadata to pages. This requires two passes -
1487 	 * one to verify that there are enough pages and a second
1488 	 * to actually claim them. */
1489 	page_num = 0;
1490 	/* Note that this loop starts at one. The first page location is fixed by the blobid. */
1491 	for (i = 1; i < blob->active.num_pages; i++) {
1492 		page_num = spdk_bit_array_find_first_clear(bs->used_md_pages, page_num);
1493 		if (page_num == UINT32_MAX) {
1494 			_spdk_blob_persist_complete(seq, ctx, -ENOMEM);
1495 			return;
1496 		}
1497 		page_num++;
1498 	}
1499 
1500 	page_num = 0;
1501 	blob->active.pages[0] = _spdk_bs_blobid_to_page(blob->id);
1502 	for (i = 1; i < blob->active.num_pages; i++) {
1503 		page_num = spdk_bit_array_find_first_clear(bs->used_md_pages, page_num);
1504 		ctx->pages[i - 1].next = page_num;
1505 		/* Now that previous metadata page is complete, calculate the crc for it. */
1506 		ctx->pages[i - 1].crc = _spdk_blob_md_page_calc_crc(&ctx->pages[i - 1]);
1507 		blob->active.pages[i] = page_num;
1508 		spdk_bit_array_set(bs->used_md_pages, page_num);
1509 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Claiming page %u for blob %lu\n", page_num, blob->id);
1510 		page_num++;
1511 	}
1512 	ctx->pages[i - 1].crc = _spdk_blob_md_page_calc_crc(&ctx->pages[i - 1]);
1513 	/* Start writing the metadata from last page to first */
1514 	blob->state = SPDK_BLOB_STATE_CLEAN;
1515 	_spdk_blob_persist_write_page_chain(seq, ctx, 0);
1516 }
1517 
1518 static void
1519 _spdk_blob_persist_start(struct spdk_blob_persist_ctx *ctx)
1520 {
1521 	spdk_bs_sequence_t *seq = ctx->seq;
1522 	struct spdk_blob *blob = ctx->blob;
1523 
1524 	if (blob->active.num_pages == 0) {
1525 		/* This is the signal that the blob should be deleted.
1526 		 * Immediately jump to the clean up routine. */
1527 		assert(blob->clean.num_pages > 0);
1528 		blob->state = SPDK_BLOB_STATE_CLEAN;
1529 		_spdk_blob_persist_zero_pages(seq, ctx, 0);
1530 		return;
1531 
1532 	}
1533 
1534 	_spdk_blob_persist_generate_new_md(ctx);
1535 }
1536 
1537 static void
1538 _spdk_blob_persist_dirty_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1539 {
1540 	struct spdk_blob_persist_ctx *ctx = cb_arg;
1541 
1542 	ctx->blob->bs->clean = 0;
1543 
1544 	spdk_free(ctx->super);
1545 
1546 	_spdk_blob_persist_start(ctx);
1547 }
1548 
1549 static void
1550 _spdk_bs_write_super(spdk_bs_sequence_t *seq, struct spdk_blob_store *bs,
1551 		     struct spdk_bs_super_block *super, spdk_bs_sequence_cpl cb_fn, void *cb_arg);
1552 
1553 
1554 static void
1555 _spdk_blob_persist_dirty(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1556 {
1557 	struct spdk_blob_persist_ctx *ctx = cb_arg;
1558 
1559 	ctx->super->clean = 0;
1560 	if (ctx->super->size == 0) {
1561 		ctx->super->size = ctx->blob->bs->dev->blockcnt * ctx->blob->bs->dev->blocklen;
1562 	}
1563 
1564 	_spdk_bs_write_super(seq, ctx->blob->bs, ctx->super, _spdk_blob_persist_dirty_cpl, ctx);
1565 }
1566 
1567 
1568 /* Write a blob to disk */
1569 static void
1570 _spdk_blob_persist(spdk_bs_sequence_t *seq, struct spdk_blob *blob,
1571 		   spdk_bs_sequence_cpl cb_fn, void *cb_arg)
1572 {
1573 	struct spdk_blob_persist_ctx *ctx;
1574 
1575 	_spdk_blob_verify_md_op(blob);
1576 
1577 	if (blob->state == SPDK_BLOB_STATE_CLEAN) {
1578 		cb_fn(seq, cb_arg, 0);
1579 		return;
1580 	}
1581 
1582 	ctx = calloc(1, sizeof(*ctx));
1583 	if (!ctx) {
1584 		cb_fn(seq, cb_arg, -ENOMEM);
1585 		return;
1586 	}
1587 	ctx->blob = blob;
1588 	ctx->seq = seq;
1589 	ctx->cb_fn = cb_fn;
1590 	ctx->cb_arg = cb_arg;
1591 
1592 	if (blob->bs->clean) {
1593 		ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
1594 					  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
1595 		if (!ctx->super) {
1596 			cb_fn(seq, cb_arg, -ENOMEM);
1597 			free(ctx);
1598 			return;
1599 		}
1600 
1601 		spdk_bs_sequence_read_dev(seq, ctx->super, _spdk_bs_page_to_lba(blob->bs, 0),
1602 					  _spdk_bs_byte_to_lba(blob->bs, sizeof(*ctx->super)),
1603 					  _spdk_blob_persist_dirty, ctx);
1604 	} else {
1605 		_spdk_blob_persist_start(ctx);
1606 	}
1607 }
1608 
1609 struct spdk_blob_copy_cluster_ctx {
1610 	struct spdk_blob *blob;
1611 	uint8_t *buf;
1612 	uint64_t page;
1613 	uint64_t new_cluster;
1614 	spdk_bs_sequence_t *seq;
1615 };
1616 
1617 static void
1618 _spdk_blob_allocate_and_copy_cluster_cpl(void *cb_arg, int bserrno)
1619 {
1620 	struct spdk_blob_copy_cluster_ctx *ctx = cb_arg;
1621 	struct spdk_bs_request_set *set = (struct spdk_bs_request_set *)ctx->seq;
1622 	TAILQ_HEAD(, spdk_bs_request_set) requests;
1623 	spdk_bs_user_op_t *op;
1624 
1625 	TAILQ_INIT(&requests);
1626 	TAILQ_SWAP(&set->channel->need_cluster_alloc, &requests, spdk_bs_request_set, link);
1627 
1628 	while (!TAILQ_EMPTY(&requests)) {
1629 		op = TAILQ_FIRST(&requests);
1630 		TAILQ_REMOVE(&requests, op, link);
1631 		if (bserrno == 0) {
1632 			spdk_bs_user_op_execute(op);
1633 		} else {
1634 			spdk_bs_user_op_abort(op);
1635 		}
1636 	}
1637 
1638 	spdk_free(ctx->buf);
1639 	free(ctx);
1640 }
1641 
1642 static void
1643 _spdk_blob_insert_cluster_cpl(void *cb_arg, int bserrno)
1644 {
1645 	struct spdk_blob_copy_cluster_ctx *ctx = cb_arg;
1646 
1647 	if (bserrno) {
1648 		if (bserrno == -EEXIST) {
1649 			/* The metadata insert failed because another thread
1650 			 * allocated the cluster first. Free our cluster
1651 			 * but continue without error. */
1652 			bserrno = 0;
1653 		}
1654 		_spdk_bs_release_cluster(ctx->blob->bs, ctx->new_cluster);
1655 	}
1656 
1657 	spdk_bs_sequence_finish(ctx->seq, bserrno);
1658 }
1659 
1660 static void
1661 _spdk_blob_write_copy_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1662 {
1663 	struct spdk_blob_copy_cluster_ctx *ctx = cb_arg;
1664 	uint32_t cluster_number;
1665 
1666 	if (bserrno) {
1667 		/* The write failed, so jump to the final completion handler */
1668 		spdk_bs_sequence_finish(seq, bserrno);
1669 		return;
1670 	}
1671 
1672 	cluster_number = _spdk_bs_page_to_cluster(ctx->blob->bs, ctx->page);
1673 
1674 	_spdk_blob_insert_cluster_on_md_thread(ctx->blob, cluster_number, ctx->new_cluster,
1675 					       _spdk_blob_insert_cluster_cpl, ctx);
1676 }
1677 
1678 static void
1679 _spdk_blob_write_copy(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
1680 {
1681 	struct spdk_blob_copy_cluster_ctx *ctx = cb_arg;
1682 
1683 	if (bserrno != 0) {
1684 		/* The read failed, so jump to the final completion handler */
1685 		spdk_bs_sequence_finish(seq, bserrno);
1686 		return;
1687 	}
1688 
1689 	/* Write whole cluster */
1690 	spdk_bs_sequence_write_dev(seq, ctx->buf,
1691 				   _spdk_bs_cluster_to_lba(ctx->blob->bs, ctx->new_cluster),
1692 				   _spdk_bs_cluster_to_lba(ctx->blob->bs, 1),
1693 				   _spdk_blob_write_copy_cpl, ctx);
1694 }
1695 
1696 static void
1697 _spdk_bs_allocate_and_copy_cluster(struct spdk_blob *blob,
1698 				   struct spdk_io_channel *_ch,
1699 				   uint64_t io_unit, spdk_bs_user_op_t *op)
1700 {
1701 	struct spdk_bs_cpl cpl;
1702 	struct spdk_bs_channel *ch;
1703 	struct spdk_blob_copy_cluster_ctx *ctx;
1704 	uint32_t cluster_start_page;
1705 	uint32_t cluster_number;
1706 	int rc;
1707 
1708 	ch = spdk_io_channel_get_ctx(_ch);
1709 
1710 	if (!TAILQ_EMPTY(&ch->need_cluster_alloc)) {
1711 		/* There are already operations pending. Queue this user op
1712 		 * and return because it will be re-executed when the outstanding
1713 		 * cluster allocation completes. */
1714 		TAILQ_INSERT_TAIL(&ch->need_cluster_alloc, op, link);
1715 		return;
1716 	}
1717 
1718 	/* Round the io_unit offset down to the first page in the cluster */
1719 	cluster_start_page = _spdk_bs_io_unit_to_cluster_start(blob, io_unit);
1720 
1721 	/* Calculate which index in the metadata cluster array the corresponding
1722 	 * cluster is supposed to be at. */
1723 	cluster_number = _spdk_bs_io_unit_to_cluster_number(blob, io_unit);
1724 
1725 	ctx = calloc(1, sizeof(*ctx));
1726 	if (!ctx) {
1727 		spdk_bs_user_op_abort(op);
1728 		return;
1729 	}
1730 
1731 	assert(blob->bs->cluster_sz % blob->back_bs_dev->blocklen == 0);
1732 
1733 	ctx->blob = blob;
1734 	ctx->page = cluster_start_page;
1735 
1736 	if (blob->parent_id != SPDK_BLOBID_INVALID) {
1737 		ctx->buf = spdk_malloc(blob->bs->cluster_sz, blob->back_bs_dev->blocklen,
1738 				       NULL, SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
1739 		if (!ctx->buf) {
1740 			SPDK_ERRLOG("DMA allocation for cluster of size = %" PRIu32 " failed.\n",
1741 				    blob->bs->cluster_sz);
1742 			free(ctx);
1743 			spdk_bs_user_op_abort(op);
1744 			return;
1745 		}
1746 	}
1747 
1748 	rc = _spdk_bs_allocate_cluster(blob, cluster_number, &ctx->new_cluster, false);
1749 	if (rc != 0) {
1750 		spdk_free(ctx->buf);
1751 		free(ctx);
1752 		spdk_bs_user_op_abort(op);
1753 		return;
1754 	}
1755 
1756 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
1757 	cpl.u.blob_basic.cb_fn = _spdk_blob_allocate_and_copy_cluster_cpl;
1758 	cpl.u.blob_basic.cb_arg = ctx;
1759 
1760 	ctx->seq = spdk_bs_sequence_start(_ch, &cpl);
1761 	if (!ctx->seq) {
1762 		_spdk_bs_release_cluster(blob->bs, ctx->new_cluster);
1763 		spdk_free(ctx->buf);
1764 		free(ctx);
1765 		spdk_bs_user_op_abort(op);
1766 		return;
1767 	}
1768 
1769 	/* Queue the user op to block other incoming operations */
1770 	TAILQ_INSERT_TAIL(&ch->need_cluster_alloc, op, link);
1771 
1772 	if (blob->parent_id != SPDK_BLOBID_INVALID) {
1773 		/* Read cluster from backing device */
1774 		spdk_bs_sequence_read_bs_dev(ctx->seq, blob->back_bs_dev, ctx->buf,
1775 					     _spdk_bs_dev_page_to_lba(blob->back_bs_dev, cluster_start_page),
1776 					     _spdk_bs_dev_byte_to_lba(blob->back_bs_dev, blob->bs->cluster_sz),
1777 					     _spdk_blob_write_copy, ctx);
1778 	} else {
1779 		_spdk_blob_insert_cluster_on_md_thread(ctx->blob, cluster_number, ctx->new_cluster,
1780 						       _spdk_blob_insert_cluster_cpl, ctx);
1781 	}
1782 }
1783 
1784 static void
1785 _spdk_blob_calculate_lba_and_lba_count(struct spdk_blob *blob, uint64_t io_unit, uint64_t length,
1786 				       uint64_t *lba,	uint32_t *lba_count)
1787 {
1788 	*lba_count = length;
1789 
1790 	if (!_spdk_bs_io_unit_is_allocated(blob, io_unit)) {
1791 		assert(blob->back_bs_dev != NULL);
1792 		*lba = _spdk_bs_io_unit_to_back_dev_lba(blob, io_unit);
1793 		*lba_count = _spdk_bs_io_unit_to_back_dev_lba(blob, *lba_count);
1794 	} else {
1795 		*lba = _spdk_bs_blob_io_unit_to_lba(blob, io_unit);
1796 	}
1797 }
1798 
1799 struct op_split_ctx {
1800 	struct spdk_blob *blob;
1801 	struct spdk_io_channel *channel;
1802 	uint64_t io_unit_offset;
1803 	uint64_t io_units_remaining;
1804 	void *curr_payload;
1805 	enum spdk_blob_op_type op_type;
1806 	spdk_bs_sequence_t *seq;
1807 };
1808 
1809 static void
1810 _spdk_blob_request_submit_op_split_next(void *cb_arg, int bserrno)
1811 {
1812 	struct op_split_ctx	*ctx = cb_arg;
1813 	struct spdk_blob	*blob = ctx->blob;
1814 	struct spdk_io_channel	*ch = ctx->channel;
1815 	enum spdk_blob_op_type	op_type = ctx->op_type;
1816 	uint8_t			*buf = ctx->curr_payload;
1817 	uint64_t		offset = ctx->io_unit_offset;
1818 	uint64_t		length = ctx->io_units_remaining;
1819 	uint64_t		op_length;
1820 
1821 	if (bserrno != 0 || ctx->io_units_remaining == 0) {
1822 		spdk_bs_sequence_finish(ctx->seq, bserrno);
1823 		free(ctx);
1824 		return;
1825 	}
1826 
1827 	op_length = spdk_min(length, _spdk_bs_num_io_units_to_cluster_boundary(blob,
1828 			     offset));
1829 
1830 	/* Update length and payload for next operation */
1831 	ctx->io_units_remaining -= op_length;
1832 	ctx->io_unit_offset += op_length;
1833 	if (op_type == SPDK_BLOB_WRITE || op_type == SPDK_BLOB_READ) {
1834 		ctx->curr_payload += op_length * blob->bs->io_unit_size;
1835 	}
1836 
1837 	switch (op_type) {
1838 	case SPDK_BLOB_READ:
1839 		spdk_blob_io_read(blob, ch, buf, offset, op_length,
1840 				  _spdk_blob_request_submit_op_split_next, ctx);
1841 		break;
1842 	case SPDK_BLOB_WRITE:
1843 		spdk_blob_io_write(blob, ch, buf, offset, op_length,
1844 				   _spdk_blob_request_submit_op_split_next, ctx);
1845 		break;
1846 	case SPDK_BLOB_UNMAP:
1847 		spdk_blob_io_unmap(blob, ch, offset, op_length,
1848 				   _spdk_blob_request_submit_op_split_next, ctx);
1849 		break;
1850 	case SPDK_BLOB_WRITE_ZEROES:
1851 		spdk_blob_io_write_zeroes(blob, ch, offset, op_length,
1852 					  _spdk_blob_request_submit_op_split_next, ctx);
1853 		break;
1854 	case SPDK_BLOB_READV:
1855 	case SPDK_BLOB_WRITEV:
1856 		SPDK_ERRLOG("readv/write not valid\n");
1857 		spdk_bs_sequence_finish(ctx->seq, -EINVAL);
1858 		free(ctx);
1859 		break;
1860 	}
1861 }
1862 
1863 static void
1864 _spdk_blob_request_submit_op_split(struct spdk_io_channel *ch, struct spdk_blob *blob,
1865 				   void *payload, uint64_t offset, uint64_t length,
1866 				   spdk_blob_op_complete cb_fn, void *cb_arg, enum spdk_blob_op_type op_type)
1867 {
1868 	struct op_split_ctx *ctx;
1869 	spdk_bs_sequence_t *seq;
1870 	struct spdk_bs_cpl cpl;
1871 
1872 	assert(blob != NULL);
1873 
1874 	ctx = calloc(1, sizeof(struct op_split_ctx));
1875 	if (ctx == NULL) {
1876 		cb_fn(cb_arg, -ENOMEM);
1877 		return;
1878 	}
1879 
1880 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
1881 	cpl.u.blob_basic.cb_fn = cb_fn;
1882 	cpl.u.blob_basic.cb_arg = cb_arg;
1883 
1884 	seq = spdk_bs_sequence_start(ch, &cpl);
1885 	if (!seq) {
1886 		free(ctx);
1887 		cb_fn(cb_arg, -ENOMEM);
1888 		return;
1889 	}
1890 
1891 	ctx->blob = blob;
1892 	ctx->channel = ch;
1893 	ctx->curr_payload = payload;
1894 	ctx->io_unit_offset = offset;
1895 	ctx->io_units_remaining = length;
1896 	ctx->op_type = op_type;
1897 	ctx->seq = seq;
1898 
1899 	_spdk_blob_request_submit_op_split_next(ctx, 0);
1900 }
1901 
1902 static void
1903 _spdk_blob_request_submit_op_single(struct spdk_io_channel *_ch, struct spdk_blob *blob,
1904 				    void *payload, uint64_t offset, uint64_t length,
1905 				    spdk_blob_op_complete cb_fn, void *cb_arg, enum spdk_blob_op_type op_type)
1906 {
1907 	struct spdk_bs_cpl cpl;
1908 	uint64_t lba;
1909 	uint32_t lba_count;
1910 
1911 	assert(blob != NULL);
1912 
1913 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
1914 	cpl.u.blob_basic.cb_fn = cb_fn;
1915 	cpl.u.blob_basic.cb_arg = cb_arg;
1916 
1917 	_spdk_blob_calculate_lba_and_lba_count(blob, offset, length, &lba, &lba_count);
1918 
1919 	if (blob->frozen_refcnt) {
1920 		/* This blob I/O is frozen */
1921 		spdk_bs_user_op_t *op;
1922 		struct spdk_bs_channel *bs_channel = spdk_io_channel_get_ctx(_ch);
1923 
1924 		op = spdk_bs_user_op_alloc(_ch, &cpl, op_type, blob, payload, 0, offset, length);
1925 		if (!op) {
1926 			cb_fn(cb_arg, -ENOMEM);
1927 			return;
1928 		}
1929 
1930 		TAILQ_INSERT_TAIL(&bs_channel->queued_io, op, link);
1931 
1932 		return;
1933 	}
1934 
1935 	switch (op_type) {
1936 	case SPDK_BLOB_READ: {
1937 		spdk_bs_batch_t *batch;
1938 
1939 		batch = spdk_bs_batch_open(_ch, &cpl);
1940 		if (!batch) {
1941 			cb_fn(cb_arg, -ENOMEM);
1942 			return;
1943 		}
1944 
1945 		if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
1946 			/* Read from the blob */
1947 			spdk_bs_batch_read_dev(batch, payload, lba, lba_count);
1948 		} else {
1949 			/* Read from the backing block device */
1950 			spdk_bs_batch_read_bs_dev(batch, blob->back_bs_dev, payload, lba, lba_count);
1951 		}
1952 
1953 		spdk_bs_batch_close(batch);
1954 		break;
1955 	}
1956 	case SPDK_BLOB_WRITE:
1957 	case SPDK_BLOB_WRITE_ZEROES: {
1958 		if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
1959 			/* Write to the blob */
1960 			spdk_bs_batch_t *batch;
1961 
1962 			if (lba_count == 0) {
1963 				cb_fn(cb_arg, 0);
1964 				return;
1965 			}
1966 
1967 			batch = spdk_bs_batch_open(_ch, &cpl);
1968 			if (!batch) {
1969 				cb_fn(cb_arg, -ENOMEM);
1970 				return;
1971 			}
1972 
1973 			if (op_type == SPDK_BLOB_WRITE) {
1974 				spdk_bs_batch_write_dev(batch, payload, lba, lba_count);
1975 			} else {
1976 				spdk_bs_batch_write_zeroes_dev(batch, lba, lba_count);
1977 			}
1978 
1979 			spdk_bs_batch_close(batch);
1980 		} else {
1981 			/* Queue this operation and allocate the cluster */
1982 			spdk_bs_user_op_t *op;
1983 
1984 			op = spdk_bs_user_op_alloc(_ch, &cpl, op_type, blob, payload, 0, offset, length);
1985 			if (!op) {
1986 				cb_fn(cb_arg, -ENOMEM);
1987 				return;
1988 			}
1989 
1990 			_spdk_bs_allocate_and_copy_cluster(blob, _ch, offset, op);
1991 		}
1992 		break;
1993 	}
1994 	case SPDK_BLOB_UNMAP: {
1995 		spdk_bs_batch_t *batch;
1996 
1997 		batch = spdk_bs_batch_open(_ch, &cpl);
1998 		if (!batch) {
1999 			cb_fn(cb_arg, -ENOMEM);
2000 			return;
2001 		}
2002 
2003 		if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
2004 			spdk_bs_batch_unmap_dev(batch, lba, lba_count);
2005 		}
2006 
2007 		spdk_bs_batch_close(batch);
2008 		break;
2009 	}
2010 	case SPDK_BLOB_READV:
2011 	case SPDK_BLOB_WRITEV:
2012 		SPDK_ERRLOG("readv/write not valid\n");
2013 		cb_fn(cb_arg, -EINVAL);
2014 		break;
2015 	}
2016 }
2017 
2018 static void
2019 _spdk_blob_request_submit_op(struct spdk_blob *blob, struct spdk_io_channel *_channel,
2020 			     void *payload, uint64_t offset, uint64_t length,
2021 			     spdk_blob_op_complete cb_fn, void *cb_arg, enum spdk_blob_op_type op_type)
2022 {
2023 	assert(blob != NULL);
2024 
2025 	if (blob->data_ro && op_type != SPDK_BLOB_READ) {
2026 		cb_fn(cb_arg, -EPERM);
2027 		return;
2028 	}
2029 
2030 	if (offset + length > _spdk_bs_cluster_to_lba(blob->bs, blob->active.num_clusters)) {
2031 		cb_fn(cb_arg, -EINVAL);
2032 		return;
2033 	}
2034 	if (length <= _spdk_bs_num_io_units_to_cluster_boundary(blob, offset)) {
2035 		_spdk_blob_request_submit_op_single(_channel, blob, payload, offset, length,
2036 						    cb_fn, cb_arg, op_type);
2037 	} else {
2038 		_spdk_blob_request_submit_op_split(_channel, blob, payload, offset, length,
2039 						   cb_fn, cb_arg, op_type);
2040 	}
2041 }
2042 
2043 struct rw_iov_ctx {
2044 	struct spdk_blob *blob;
2045 	struct spdk_io_channel *channel;
2046 	spdk_blob_op_complete cb_fn;
2047 	void *cb_arg;
2048 	bool read;
2049 	int iovcnt;
2050 	struct iovec *orig_iov;
2051 	uint64_t io_unit_offset;
2052 	uint64_t io_units_remaining;
2053 	uint64_t io_units_done;
2054 	struct iovec iov[0];
2055 };
2056 
2057 static void
2058 _spdk_rw_iov_done(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
2059 {
2060 	assert(cb_arg == NULL);
2061 	spdk_bs_sequence_finish(seq, bserrno);
2062 }
2063 
2064 static void
2065 _spdk_rw_iov_split_next(void *cb_arg, int bserrno)
2066 {
2067 	struct rw_iov_ctx *ctx = cb_arg;
2068 	struct spdk_blob *blob = ctx->blob;
2069 	struct iovec *iov, *orig_iov;
2070 	int iovcnt;
2071 	size_t orig_iovoff;
2072 	uint64_t io_units_count, io_units_to_boundary, io_unit_offset;
2073 	uint64_t byte_count;
2074 
2075 	if (bserrno != 0 || ctx->io_units_remaining == 0) {
2076 		ctx->cb_fn(ctx->cb_arg, bserrno);
2077 		free(ctx);
2078 		return;
2079 	}
2080 
2081 	io_unit_offset = ctx->io_unit_offset;
2082 	io_units_to_boundary = _spdk_bs_num_io_units_to_cluster_boundary(blob, io_unit_offset);
2083 	io_units_count = spdk_min(ctx->io_units_remaining, io_units_to_boundary);
2084 	/*
2085 	 * Get index and offset into the original iov array for our current position in the I/O sequence.
2086 	 *  byte_count will keep track of how many bytes remaining until orig_iov and orig_iovoff will
2087 	 *  point to the current position in the I/O sequence.
2088 	 */
2089 	byte_count = ctx->io_units_done * blob->bs->io_unit_size;
2090 	orig_iov = &ctx->orig_iov[0];
2091 	orig_iovoff = 0;
2092 	while (byte_count > 0) {
2093 		if (byte_count >= orig_iov->iov_len) {
2094 			byte_count -= orig_iov->iov_len;
2095 			orig_iov++;
2096 		} else {
2097 			orig_iovoff = byte_count;
2098 			byte_count = 0;
2099 		}
2100 	}
2101 
2102 	/*
2103 	 * Build an iov array for the next I/O in the sequence.  byte_count will keep track of how many
2104 	 *  bytes of this next I/O remain to be accounted for in the new iov array.
2105 	 */
2106 	byte_count = io_units_count * blob->bs->io_unit_size;
2107 	iov = &ctx->iov[0];
2108 	iovcnt = 0;
2109 	while (byte_count > 0) {
2110 		assert(iovcnt < ctx->iovcnt);
2111 		iov->iov_len = spdk_min(byte_count, orig_iov->iov_len - orig_iovoff);
2112 		iov->iov_base = orig_iov->iov_base + orig_iovoff;
2113 		byte_count -= iov->iov_len;
2114 		orig_iovoff = 0;
2115 		orig_iov++;
2116 		iov++;
2117 		iovcnt++;
2118 	}
2119 
2120 	ctx->io_unit_offset += io_units_count;
2121 	ctx->io_units_remaining -= io_units_count;
2122 	ctx->io_units_done += io_units_count;
2123 	iov = &ctx->iov[0];
2124 
2125 	if (ctx->read) {
2126 		spdk_blob_io_readv(ctx->blob, ctx->channel, iov, iovcnt, io_unit_offset,
2127 				   io_units_count, _spdk_rw_iov_split_next, ctx);
2128 	} else {
2129 		spdk_blob_io_writev(ctx->blob, ctx->channel, iov, iovcnt, io_unit_offset,
2130 				    io_units_count, _spdk_rw_iov_split_next, ctx);
2131 	}
2132 }
2133 
2134 static void
2135 _spdk_blob_request_submit_rw_iov(struct spdk_blob *blob, struct spdk_io_channel *_channel,
2136 				 struct iovec *iov, int iovcnt, uint64_t offset, uint64_t length,
2137 				 spdk_blob_op_complete cb_fn, void *cb_arg, bool read)
2138 {
2139 	struct spdk_bs_cpl	cpl;
2140 
2141 	assert(blob != NULL);
2142 
2143 	if (!read && blob->data_ro) {
2144 		cb_fn(cb_arg, -EPERM);
2145 		return;
2146 	}
2147 
2148 	if (length == 0) {
2149 		cb_fn(cb_arg, 0);
2150 		return;
2151 	}
2152 
2153 	if (offset + length > _spdk_bs_cluster_to_lba(blob->bs, blob->active.num_clusters)) {
2154 		cb_fn(cb_arg, -EINVAL);
2155 		return;
2156 	}
2157 
2158 	/*
2159 	 * For now, we implement readv/writev using a sequence (instead of a batch) to account for having
2160 	 *  to split a request that spans a cluster boundary.  For I/O that do not span a cluster boundary,
2161 	 *  there will be no noticeable difference compared to using a batch.  For I/O that do span a cluster
2162 	 *  boundary, the target LBAs (after blob offset to LBA translation) may not be contiguous, so we need
2163 	 *  to allocate a separate iov array and split the I/O such that none of the resulting
2164 	 *  smaller I/O cross a cluster boundary.  These smaller I/O will be issued in sequence (not in parallel)
2165 	 *  but since this case happens very infrequently, any performance impact will be negligible.
2166 	 *
2167 	 * This could be optimized in the future to allocate a big enough iov array to account for all of the iovs
2168 	 *  for all of the smaller I/Os, pre-build all of the iov arrays for the smaller I/Os, then issue them
2169 	 *  in a batch.  That would also require creating an intermediate spdk_bs_cpl that would get called
2170 	 *  when the batch was completed, to allow for freeing the memory for the iov arrays.
2171 	 */
2172 	if (spdk_likely(length <= _spdk_bs_num_io_units_to_cluster_boundary(blob, offset))) {
2173 		uint32_t lba_count;
2174 		uint64_t lba;
2175 
2176 		cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
2177 		cpl.u.blob_basic.cb_fn = cb_fn;
2178 		cpl.u.blob_basic.cb_arg = cb_arg;
2179 
2180 		if (blob->frozen_refcnt) {
2181 			/* This blob I/O is frozen */
2182 			enum spdk_blob_op_type op_type;
2183 			spdk_bs_user_op_t *op;
2184 			struct spdk_bs_channel *bs_channel = spdk_io_channel_get_ctx(_channel);
2185 
2186 			op_type = read ? SPDK_BLOB_READV : SPDK_BLOB_WRITEV;
2187 			op = spdk_bs_user_op_alloc(_channel, &cpl, op_type, blob, iov, iovcnt, offset, length);
2188 			if (!op) {
2189 				cb_fn(cb_arg, -ENOMEM);
2190 				return;
2191 			}
2192 
2193 			TAILQ_INSERT_TAIL(&bs_channel->queued_io, op, link);
2194 
2195 			return;
2196 		}
2197 
2198 		_spdk_blob_calculate_lba_and_lba_count(blob, offset, length, &lba, &lba_count);
2199 
2200 		if (read) {
2201 			spdk_bs_sequence_t *seq;
2202 
2203 			seq = spdk_bs_sequence_start(_channel, &cpl);
2204 			if (!seq) {
2205 				cb_fn(cb_arg, -ENOMEM);
2206 				return;
2207 			}
2208 
2209 			if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
2210 				spdk_bs_sequence_readv_dev(seq, iov, iovcnt, lba, lba_count, _spdk_rw_iov_done, NULL);
2211 			} else {
2212 				spdk_bs_sequence_readv_bs_dev(seq, blob->back_bs_dev, iov, iovcnt, lba, lba_count,
2213 							      _spdk_rw_iov_done, NULL);
2214 			}
2215 		} else {
2216 			if (_spdk_bs_io_unit_is_allocated(blob, offset)) {
2217 				spdk_bs_sequence_t *seq;
2218 
2219 				seq = spdk_bs_sequence_start(_channel, &cpl);
2220 				if (!seq) {
2221 					cb_fn(cb_arg, -ENOMEM);
2222 					return;
2223 				}
2224 
2225 				spdk_bs_sequence_writev_dev(seq, iov, iovcnt, lba, lba_count, _spdk_rw_iov_done, NULL);
2226 			} else {
2227 				/* Queue this operation and allocate the cluster */
2228 				spdk_bs_user_op_t *op;
2229 
2230 				op = spdk_bs_user_op_alloc(_channel, &cpl, SPDK_BLOB_WRITEV, blob, iov, iovcnt, offset,
2231 							   length);
2232 				if (!op) {
2233 					cb_fn(cb_arg, -ENOMEM);
2234 					return;
2235 				}
2236 
2237 				_spdk_bs_allocate_and_copy_cluster(blob, _channel, offset, op);
2238 			}
2239 		}
2240 	} else {
2241 		struct rw_iov_ctx *ctx;
2242 
2243 		ctx = calloc(1, sizeof(struct rw_iov_ctx) + iovcnt * sizeof(struct iovec));
2244 		if (ctx == NULL) {
2245 			cb_fn(cb_arg, -ENOMEM);
2246 			return;
2247 		}
2248 
2249 		ctx->blob = blob;
2250 		ctx->channel = _channel;
2251 		ctx->cb_fn = cb_fn;
2252 		ctx->cb_arg = cb_arg;
2253 		ctx->read = read;
2254 		ctx->orig_iov = iov;
2255 		ctx->iovcnt = iovcnt;
2256 		ctx->io_unit_offset = offset;
2257 		ctx->io_units_remaining = length;
2258 		ctx->io_units_done = 0;
2259 
2260 		_spdk_rw_iov_split_next(ctx, 0);
2261 	}
2262 }
2263 
2264 static struct spdk_blob *
2265 _spdk_blob_lookup(struct spdk_blob_store *bs, spdk_blob_id blobid)
2266 {
2267 	struct spdk_blob *blob;
2268 
2269 	TAILQ_FOREACH(blob, &bs->blobs, link) {
2270 		if (blob->id == blobid) {
2271 			return blob;
2272 		}
2273 	}
2274 
2275 	return NULL;
2276 }
2277 
2278 static void
2279 _spdk_blob_get_snapshot_and_clone_entries(struct spdk_blob *blob,
2280 		struct spdk_blob_list **snapshot_entry, struct spdk_blob_list **clone_entry)
2281 {
2282 	assert(blob != NULL);
2283 	*snapshot_entry = NULL;
2284 	*clone_entry = NULL;
2285 
2286 	if (blob->parent_id == SPDK_BLOBID_INVALID) {
2287 		return;
2288 	}
2289 
2290 	TAILQ_FOREACH(*snapshot_entry, &blob->bs->snapshots, link) {
2291 		if ((*snapshot_entry)->id == blob->parent_id) {
2292 			break;
2293 		}
2294 	}
2295 
2296 	if (*snapshot_entry != NULL) {
2297 		TAILQ_FOREACH(*clone_entry, &(*snapshot_entry)->clones, link) {
2298 			if ((*clone_entry)->id == blob->id) {
2299 				break;
2300 			}
2301 		}
2302 
2303 		assert(clone_entry != NULL);
2304 	}
2305 }
2306 
2307 static int
2308 _spdk_bs_channel_create(void *io_device, void *ctx_buf)
2309 {
2310 	struct spdk_blob_store		*bs = io_device;
2311 	struct spdk_bs_channel		*channel = ctx_buf;
2312 	struct spdk_bs_dev		*dev;
2313 	uint32_t			max_ops = bs->max_channel_ops;
2314 	uint32_t			i;
2315 
2316 	dev = bs->dev;
2317 
2318 	channel->req_mem = calloc(max_ops, sizeof(struct spdk_bs_request_set));
2319 	if (!channel->req_mem) {
2320 		return -1;
2321 	}
2322 
2323 	TAILQ_INIT(&channel->reqs);
2324 
2325 	for (i = 0; i < max_ops; i++) {
2326 		TAILQ_INSERT_TAIL(&channel->reqs, &channel->req_mem[i], link);
2327 	}
2328 
2329 	channel->bs = bs;
2330 	channel->dev = dev;
2331 	channel->dev_channel = dev->create_channel(dev);
2332 
2333 	if (!channel->dev_channel) {
2334 		SPDK_ERRLOG("Failed to create device channel.\n");
2335 		free(channel->req_mem);
2336 		return -1;
2337 	}
2338 
2339 	TAILQ_INIT(&channel->need_cluster_alloc);
2340 	TAILQ_INIT(&channel->queued_io);
2341 
2342 	return 0;
2343 }
2344 
2345 static void
2346 _spdk_bs_channel_destroy(void *io_device, void *ctx_buf)
2347 {
2348 	struct spdk_bs_channel *channel = ctx_buf;
2349 	spdk_bs_user_op_t *op;
2350 
2351 	while (!TAILQ_EMPTY(&channel->need_cluster_alloc)) {
2352 		op = TAILQ_FIRST(&channel->need_cluster_alloc);
2353 		TAILQ_REMOVE(&channel->need_cluster_alloc, op, link);
2354 		spdk_bs_user_op_abort(op);
2355 	}
2356 
2357 	while (!TAILQ_EMPTY(&channel->queued_io)) {
2358 		op = TAILQ_FIRST(&channel->queued_io);
2359 		TAILQ_REMOVE(&channel->queued_io, op, link);
2360 		spdk_bs_user_op_abort(op);
2361 	}
2362 
2363 	free(channel->req_mem);
2364 	channel->dev->destroy_channel(channel->dev, channel->dev_channel);
2365 }
2366 
2367 static void
2368 _spdk_bs_dev_destroy(void *io_device)
2369 {
2370 	struct spdk_blob_store *bs = io_device;
2371 	struct spdk_blob	*blob, *blob_tmp;
2372 
2373 	bs->dev->destroy(bs->dev);
2374 
2375 	TAILQ_FOREACH_SAFE(blob, &bs->blobs, link, blob_tmp) {
2376 		TAILQ_REMOVE(&bs->blobs, blob, link);
2377 		_spdk_blob_free(blob);
2378 	}
2379 
2380 	pthread_mutex_destroy(&bs->used_clusters_mutex);
2381 
2382 	spdk_bit_array_free(&bs->used_blobids);
2383 	spdk_bit_array_free(&bs->used_md_pages);
2384 	spdk_bit_array_free(&bs->used_clusters);
2385 	/*
2386 	 * If this function is called for any reason except a successful unload,
2387 	 * the unload_cpl type will be NONE and this will be a nop.
2388 	 */
2389 	spdk_bs_call_cpl(&bs->unload_cpl, bs->unload_err);
2390 
2391 	free(bs);
2392 }
2393 
2394 static int
2395 _spdk_bs_blob_list_add(struct spdk_blob *blob)
2396 {
2397 	spdk_blob_id snapshot_id;
2398 	struct spdk_blob_list *snapshot_entry = NULL;
2399 	struct spdk_blob_list *clone_entry = NULL;
2400 
2401 	assert(blob != NULL);
2402 
2403 	snapshot_id = blob->parent_id;
2404 	if (snapshot_id == SPDK_BLOBID_INVALID) {
2405 		return 0;
2406 	}
2407 
2408 	snapshot_entry = _spdk_bs_get_snapshot_entry(blob->bs, snapshot_id);
2409 	if (snapshot_entry == NULL) {
2410 		/* Snapshot not found */
2411 		snapshot_entry = calloc(1, sizeof(struct spdk_blob_list));
2412 		if (snapshot_entry == NULL) {
2413 			return -ENOMEM;
2414 		}
2415 		snapshot_entry->id = snapshot_id;
2416 		TAILQ_INIT(&snapshot_entry->clones);
2417 		TAILQ_INSERT_TAIL(&blob->bs->snapshots, snapshot_entry, link);
2418 	} else {
2419 		TAILQ_FOREACH(clone_entry, &snapshot_entry->clones, link) {
2420 			if (clone_entry->id == blob->id) {
2421 				break;
2422 			}
2423 		}
2424 	}
2425 
2426 	if (clone_entry == NULL) {
2427 		/* Clone not found */
2428 		clone_entry = calloc(1, sizeof(struct spdk_blob_list));
2429 		if (clone_entry == NULL) {
2430 			return -ENOMEM;
2431 		}
2432 		clone_entry->id = blob->id;
2433 		TAILQ_INIT(&clone_entry->clones);
2434 		TAILQ_INSERT_TAIL(&snapshot_entry->clones, clone_entry, link);
2435 		snapshot_entry->clone_count++;
2436 	}
2437 
2438 	return 0;
2439 }
2440 
2441 static void
2442 _spdk_bs_blob_list_remove(struct spdk_blob *blob)
2443 {
2444 	struct spdk_blob_list *snapshot_entry = NULL;
2445 	struct spdk_blob_list *clone_entry = NULL;
2446 
2447 	_spdk_blob_get_snapshot_and_clone_entries(blob, &snapshot_entry, &clone_entry);
2448 
2449 	if (snapshot_entry == NULL) {
2450 		return;
2451 	}
2452 
2453 	blob->parent_id = SPDK_BLOBID_INVALID;
2454 	TAILQ_REMOVE(&snapshot_entry->clones, clone_entry, link);
2455 	free(clone_entry);
2456 
2457 	snapshot_entry->clone_count--;
2458 }
2459 
2460 static int
2461 _spdk_bs_blob_list_free(struct spdk_blob_store *bs)
2462 {
2463 	struct spdk_blob_list *snapshot_entry;
2464 	struct spdk_blob_list *snapshot_entry_tmp;
2465 	struct spdk_blob_list *clone_entry;
2466 	struct spdk_blob_list *clone_entry_tmp;
2467 
2468 	TAILQ_FOREACH_SAFE(snapshot_entry, &bs->snapshots, link, snapshot_entry_tmp) {
2469 		TAILQ_FOREACH_SAFE(clone_entry, &snapshot_entry->clones, link, clone_entry_tmp) {
2470 			TAILQ_REMOVE(&snapshot_entry->clones, clone_entry, link);
2471 			free(clone_entry);
2472 		}
2473 		TAILQ_REMOVE(&bs->snapshots, snapshot_entry, link);
2474 		free(snapshot_entry);
2475 	}
2476 
2477 	return 0;
2478 }
2479 
2480 static void
2481 _spdk_bs_free(struct spdk_blob_store *bs)
2482 {
2483 	_spdk_bs_blob_list_free(bs);
2484 
2485 	spdk_bs_unregister_md_thread(bs);
2486 	spdk_io_device_unregister(bs, _spdk_bs_dev_destroy);
2487 }
2488 
2489 void
2490 spdk_bs_opts_init(struct spdk_bs_opts *opts)
2491 {
2492 	opts->cluster_sz = SPDK_BLOB_OPTS_CLUSTER_SZ;
2493 	opts->num_md_pages = SPDK_BLOB_OPTS_NUM_MD_PAGES;
2494 	opts->max_md_ops = SPDK_BLOB_OPTS_MAX_MD_OPS;
2495 	opts->max_channel_ops = SPDK_BLOB_OPTS_DEFAULT_CHANNEL_OPS;
2496 	opts->clear_method = BS_CLEAR_WITH_UNMAP;
2497 	memset(&opts->bstype, 0, sizeof(opts->bstype));
2498 	opts->iter_cb_fn = NULL;
2499 	opts->iter_cb_arg = NULL;
2500 }
2501 
2502 static int
2503 _spdk_bs_opts_verify(struct spdk_bs_opts *opts)
2504 {
2505 	if (opts->cluster_sz == 0 || opts->num_md_pages == 0 || opts->max_md_ops == 0 ||
2506 	    opts->max_channel_ops == 0) {
2507 		SPDK_ERRLOG("Blobstore options cannot be set to 0\n");
2508 		return -1;
2509 	}
2510 
2511 	return 0;
2512 }
2513 
2514 static int
2515 _spdk_bs_alloc(struct spdk_bs_dev *dev, struct spdk_bs_opts *opts, struct spdk_blob_store **_bs)
2516 {
2517 	struct spdk_blob_store	*bs;
2518 	uint64_t dev_size;
2519 	int rc;
2520 
2521 	dev_size = dev->blocklen * dev->blockcnt;
2522 	if (dev_size < opts->cluster_sz) {
2523 		/* Device size cannot be smaller than cluster size of blobstore */
2524 		SPDK_INFOLOG(SPDK_LOG_BLOB, "Device size %" PRIu64 " is smaller than cluster size %" PRIu32 "\n",
2525 			     dev_size, opts->cluster_sz);
2526 		return -ENOSPC;
2527 	}
2528 	if (opts->cluster_sz < SPDK_BS_PAGE_SIZE) {
2529 		/* Cluster size cannot be smaller than page size */
2530 		SPDK_ERRLOG("Cluster size %" PRIu32 " is smaller than page size %d\n",
2531 			    opts->cluster_sz, SPDK_BS_PAGE_SIZE);
2532 		return -EINVAL;
2533 	}
2534 	bs = calloc(1, sizeof(struct spdk_blob_store));
2535 	if (!bs) {
2536 		return -ENOMEM;
2537 	}
2538 
2539 	TAILQ_INIT(&bs->blobs);
2540 	TAILQ_INIT(&bs->snapshots);
2541 	bs->dev = dev;
2542 	bs->md_thread = spdk_get_thread();
2543 	assert(bs->md_thread != NULL);
2544 
2545 	/*
2546 	 * Do not use _spdk_bs_lba_to_cluster() here since blockcnt may not be an
2547 	 *  even multiple of the cluster size.
2548 	 */
2549 	bs->cluster_sz = opts->cluster_sz;
2550 	bs->total_clusters = dev->blockcnt / (bs->cluster_sz / dev->blocklen);
2551 	bs->pages_per_cluster = bs->cluster_sz / SPDK_BS_PAGE_SIZE;
2552 	bs->num_free_clusters = bs->total_clusters;
2553 	bs->used_clusters = spdk_bit_array_create(bs->total_clusters);
2554 	bs->io_unit_size = dev->blocklen;
2555 	if (bs->used_clusters == NULL) {
2556 		free(bs);
2557 		return -ENOMEM;
2558 	}
2559 
2560 	bs->max_channel_ops = opts->max_channel_ops;
2561 	bs->super_blob = SPDK_BLOBID_INVALID;
2562 	memcpy(&bs->bstype, &opts->bstype, sizeof(opts->bstype));
2563 
2564 	/* The metadata is assumed to be at least 1 page */
2565 	bs->used_md_pages = spdk_bit_array_create(1);
2566 	bs->used_blobids = spdk_bit_array_create(0);
2567 
2568 	pthread_mutex_init(&bs->used_clusters_mutex, NULL);
2569 
2570 	spdk_io_device_register(bs, _spdk_bs_channel_create, _spdk_bs_channel_destroy,
2571 				sizeof(struct spdk_bs_channel), "blobstore");
2572 	rc = spdk_bs_register_md_thread(bs);
2573 	if (rc == -1) {
2574 		spdk_io_device_unregister(bs, NULL);
2575 		pthread_mutex_destroy(&bs->used_clusters_mutex);
2576 		spdk_bit_array_free(&bs->used_blobids);
2577 		spdk_bit_array_free(&bs->used_md_pages);
2578 		spdk_bit_array_free(&bs->used_clusters);
2579 		free(bs);
2580 		/* FIXME: this is a lie but don't know how to get a proper error code here */
2581 		return -ENOMEM;
2582 	}
2583 
2584 	*_bs = bs;
2585 	return 0;
2586 }
2587 
2588 /* START spdk_bs_load, spdk_bs_load_ctx will used for both load and unload. */
2589 
2590 struct spdk_bs_load_ctx {
2591 	struct spdk_blob_store		*bs;
2592 	struct spdk_bs_super_block	*super;
2593 
2594 	struct spdk_bs_md_mask		*mask;
2595 	bool				in_page_chain;
2596 	uint32_t			page_index;
2597 	uint32_t			cur_page;
2598 	struct spdk_blob_md_page	*page;
2599 
2600 	spdk_bs_sequence_t			*seq;
2601 	spdk_blob_op_with_handle_complete	iter_cb_fn;
2602 	void					*iter_cb_arg;
2603 	struct spdk_blob			*blob;
2604 	spdk_blob_id				blobid;
2605 };
2606 
2607 static void
2608 _spdk_bs_load_ctx_fail(struct spdk_bs_load_ctx *ctx, int bserrno)
2609 {
2610 	assert(bserrno != 0);
2611 
2612 	spdk_free(ctx->super);
2613 	spdk_bs_sequence_finish(ctx->seq, bserrno);
2614 	_spdk_bs_free(ctx->bs);
2615 	free(ctx);
2616 }
2617 
2618 static void
2619 _spdk_bs_set_mask(struct spdk_bit_array *array, struct spdk_bs_md_mask *mask)
2620 {
2621 	uint32_t i = 0;
2622 
2623 	while (true) {
2624 		i = spdk_bit_array_find_first_set(array, i);
2625 		if (i >= mask->length) {
2626 			break;
2627 		}
2628 		mask->mask[i / 8] |= 1U << (i % 8);
2629 		i++;
2630 	}
2631 }
2632 
2633 static int
2634 _spdk_bs_load_mask(struct spdk_bit_array **array_ptr, struct spdk_bs_md_mask *mask)
2635 {
2636 	struct spdk_bit_array *array;
2637 	uint32_t i;
2638 
2639 	if (spdk_bit_array_resize(array_ptr, mask->length) < 0) {
2640 		return -ENOMEM;
2641 	}
2642 
2643 	array = *array_ptr;
2644 	for (i = 0; i < mask->length; i++) {
2645 		if (mask->mask[i / 8] & (1U << (i % 8))) {
2646 			spdk_bit_array_set(array, i);
2647 		}
2648 	}
2649 
2650 	return 0;
2651 }
2652 
2653 static void
2654 _spdk_bs_write_super(spdk_bs_sequence_t *seq, struct spdk_blob_store *bs,
2655 		     struct spdk_bs_super_block *super, spdk_bs_sequence_cpl cb_fn, void *cb_arg)
2656 {
2657 	/* Update the values in the super block */
2658 	super->super_blob = bs->super_blob;
2659 	memcpy(&super->bstype, &bs->bstype, sizeof(bs->bstype));
2660 	super->crc = _spdk_blob_md_page_calc_crc(super);
2661 	spdk_bs_sequence_write_dev(seq, super, _spdk_bs_page_to_lba(bs, 0),
2662 				   _spdk_bs_byte_to_lba(bs, sizeof(*super)),
2663 				   cb_fn, cb_arg);
2664 }
2665 
2666 static void
2667 _spdk_bs_write_used_clusters(spdk_bs_sequence_t *seq, void *arg, spdk_bs_sequence_cpl cb_fn)
2668 {
2669 	struct spdk_bs_load_ctx	*ctx = arg;
2670 	uint64_t	mask_size, lba, lba_count;
2671 
2672 	/* Write out the used clusters mask */
2673 	mask_size = ctx->super->used_cluster_mask_len * SPDK_BS_PAGE_SIZE;
2674 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL,
2675 				 SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
2676 	if (!ctx->mask) {
2677 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
2678 		return;
2679 	}
2680 
2681 	ctx->mask->type = SPDK_MD_MASK_TYPE_USED_CLUSTERS;
2682 	ctx->mask->length = ctx->bs->total_clusters;
2683 	assert(ctx->mask->length == spdk_bit_array_capacity(ctx->bs->used_clusters));
2684 
2685 	_spdk_bs_set_mask(ctx->bs->used_clusters, ctx->mask);
2686 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_cluster_mask_start);
2687 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_cluster_mask_len);
2688 	spdk_bs_sequence_write_dev(seq, ctx->mask, lba, lba_count, cb_fn, arg);
2689 }
2690 
2691 static void
2692 _spdk_bs_write_used_md(spdk_bs_sequence_t *seq, void *arg, spdk_bs_sequence_cpl cb_fn)
2693 {
2694 	struct spdk_bs_load_ctx	*ctx = arg;
2695 	uint64_t	mask_size, lba, lba_count;
2696 
2697 	mask_size = ctx->super->used_page_mask_len * SPDK_BS_PAGE_SIZE;
2698 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL,
2699 				 SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
2700 	if (!ctx->mask) {
2701 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
2702 		return;
2703 	}
2704 
2705 	ctx->mask->type = SPDK_MD_MASK_TYPE_USED_PAGES;
2706 	ctx->mask->length = ctx->super->md_len;
2707 	assert(ctx->mask->length == spdk_bit_array_capacity(ctx->bs->used_md_pages));
2708 
2709 	_spdk_bs_set_mask(ctx->bs->used_md_pages, ctx->mask);
2710 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_page_mask_start);
2711 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_page_mask_len);
2712 	spdk_bs_sequence_write_dev(seq, ctx->mask, lba, lba_count, cb_fn, arg);
2713 }
2714 
2715 static void
2716 _spdk_bs_write_used_blobids(spdk_bs_sequence_t *seq, void *arg, spdk_bs_sequence_cpl cb_fn)
2717 {
2718 	struct spdk_bs_load_ctx	*ctx = arg;
2719 	uint64_t	mask_size, lba, lba_count;
2720 
2721 	if (ctx->super->used_blobid_mask_len == 0) {
2722 		/*
2723 		 * This is a pre-v3 on-disk format where the blobid mask does not get
2724 		 *  written to disk.
2725 		 */
2726 		cb_fn(seq, arg, 0);
2727 		return;
2728 	}
2729 
2730 	mask_size = ctx->super->used_blobid_mask_len * SPDK_BS_PAGE_SIZE;
2731 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL, SPDK_ENV_SOCKET_ID_ANY,
2732 				 SPDK_MALLOC_DMA);
2733 	if (!ctx->mask) {
2734 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
2735 		return;
2736 	}
2737 
2738 	ctx->mask->type = SPDK_MD_MASK_TYPE_USED_BLOBIDS;
2739 	ctx->mask->length = ctx->super->md_len;
2740 	assert(ctx->mask->length == spdk_bit_array_capacity(ctx->bs->used_blobids));
2741 
2742 	_spdk_bs_set_mask(ctx->bs->used_blobids, ctx->mask);
2743 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_blobid_mask_start);
2744 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_blobid_mask_len);
2745 	spdk_bs_sequence_write_dev(seq, ctx->mask, lba, lba_count, cb_fn, arg);
2746 }
2747 
2748 static void
2749 _spdk_blob_set_thin_provision(struct spdk_blob *blob)
2750 {
2751 	_spdk_blob_verify_md_op(blob);
2752 	blob->invalid_flags |= SPDK_BLOB_THIN_PROV;
2753 	blob->state = SPDK_BLOB_STATE_DIRTY;
2754 }
2755 
2756 static void
2757 _spdk_blob_set_clear_method(struct spdk_blob *blob, enum blob_clear_method clear_method)
2758 {
2759 	_spdk_blob_verify_md_op(blob);
2760 	blob->clear_method = clear_method;
2761 	blob->md_ro_flags |= (clear_method << SPDK_BLOB_CLEAR_METHOD_SHIFT);
2762 	blob->state = SPDK_BLOB_STATE_DIRTY;
2763 }
2764 
2765 static void _spdk_bs_load_iter(void *arg, struct spdk_blob *blob, int bserrno);
2766 
2767 static void
2768 _spdk_bs_delete_corrupted_blob_cpl(void *cb_arg, int bserrno)
2769 {
2770 	struct spdk_bs_load_ctx *ctx = cb_arg;
2771 	spdk_blob_id id;
2772 	int64_t page_num;
2773 
2774 	/* Iterate to next blob (we can't use spdk_bs_iter_next function as our
2775 	 * last blob has been removed */
2776 	page_num = _spdk_bs_blobid_to_page(ctx->blobid);
2777 	page_num++;
2778 	page_num = spdk_bit_array_find_first_set(ctx->bs->used_blobids, page_num);
2779 	if (page_num >= spdk_bit_array_capacity(ctx->bs->used_blobids)) {
2780 		_spdk_bs_load_iter(ctx, NULL, -ENOENT);
2781 		return;
2782 	}
2783 
2784 	id = _spdk_bs_page_to_blobid(page_num);
2785 
2786 	spdk_bs_open_blob(ctx->bs, id, _spdk_bs_load_iter, ctx);
2787 }
2788 
2789 static void
2790 _spdk_bs_delete_corrupted_close_cb(void *cb_arg, int bserrno)
2791 {
2792 	struct spdk_bs_load_ctx *ctx = cb_arg;
2793 
2794 	if (bserrno != 0) {
2795 		SPDK_ERRLOG("Failed to close corrupted blob\n");
2796 		spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2797 		return;
2798 	}
2799 
2800 	spdk_bs_delete_blob(ctx->bs, ctx->blobid, _spdk_bs_delete_corrupted_blob_cpl, ctx);
2801 }
2802 
2803 static void
2804 _spdk_bs_delete_corrupted_blob(void *cb_arg, int bserrno)
2805 {
2806 	struct spdk_bs_load_ctx *ctx = cb_arg;
2807 	uint64_t i;
2808 
2809 	if (bserrno != 0) {
2810 		SPDK_ERRLOG("Failed to close clone of a corrupted blob\n");
2811 		spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2812 		return;
2813 	}
2814 
2815 	/* Snapshot and clone have the same copy of cluster map at this point.
2816 	 * Let's clear cluster map for snpashot now so that it won't be cleared
2817 	 * for clone later when we remove snapshot. Also set thin provision to
2818 	 * pass data corruption check */
2819 	for (i = 0; i < ctx->blob->active.num_clusters; i++) {
2820 		ctx->blob->active.clusters[i] = 0;
2821 	}
2822 
2823 	ctx->blob->md_ro = false;
2824 
2825 	_spdk_blob_set_thin_provision(ctx->blob);
2826 
2827 	ctx->blobid = ctx->blob->id;
2828 
2829 	spdk_blob_close(ctx->blob, _spdk_bs_delete_corrupted_close_cb, ctx);
2830 }
2831 
2832 static void
2833 _spdk_bs_update_corrupted_blob(void *cb_arg, int bserrno)
2834 {
2835 	struct spdk_bs_load_ctx *ctx = cb_arg;
2836 
2837 	if (bserrno != 0) {
2838 		SPDK_ERRLOG("Failed to close clone of a corrupted blob\n");
2839 		spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2840 		return;
2841 	}
2842 
2843 	ctx->blob->md_ro = false;
2844 	_spdk_blob_remove_xattr(ctx->blob, SNAPSHOT_PENDING_REMOVAL, true);
2845 	_spdk_blob_remove_xattr(ctx->blob, SNAPSHOT_IN_PROGRESS, true);
2846 	spdk_blob_set_read_only(ctx->blob);
2847 
2848 	if (ctx->iter_cb_fn) {
2849 		ctx->iter_cb_fn(ctx->iter_cb_arg, ctx->blob, 0);
2850 	}
2851 	_spdk_bs_blob_list_add(ctx->blob);
2852 
2853 	spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2854 }
2855 
2856 static void
2857 _spdk_bs_examine_clone(void *cb_arg, struct spdk_blob *blob, int bserrno)
2858 {
2859 	struct spdk_bs_load_ctx *ctx = cb_arg;
2860 
2861 	if (bserrno != 0) {
2862 		SPDK_ERRLOG("Failed to open clone of a corrupted blob\n");
2863 		spdk_bs_iter_next(ctx->bs, ctx->blob, _spdk_bs_load_iter, ctx);
2864 		return;
2865 	}
2866 
2867 	if (blob->parent_id == ctx->blob->id) {
2868 		/* Power failure occured before updating clone (snapshot delete case)
2869 		 * or after updating clone (creating snapshot case) - keep snapshot */
2870 		spdk_blob_close(blob, _spdk_bs_update_corrupted_blob, ctx);
2871 	} else {
2872 		/* Power failure occured after updating clone (snapshot delete case)
2873 		 * or before updating clone (creating snapshot case) - remove snapshot */
2874 		spdk_blob_close(blob, _spdk_bs_delete_corrupted_blob, ctx);
2875 	}
2876 }
2877 
2878 static void
2879 _spdk_bs_load_iter(void *arg, struct spdk_blob *blob, int bserrno)
2880 {
2881 	struct spdk_bs_load_ctx *ctx = arg;
2882 	const void *value;
2883 	size_t len;
2884 	int rc = 0;
2885 
2886 	if (bserrno == 0) {
2887 		/* Examine blob if it is corrupted after power failure. Fix
2888 		 * the ones that can be fixed and remove any other corrupted
2889 		 * ones. If it is not corrupted just process it */
2890 		rc = _spdk_blob_get_xattr_value(blob, SNAPSHOT_PENDING_REMOVAL, &value, &len, true);
2891 		if (rc != 0) {
2892 			rc = _spdk_blob_get_xattr_value(blob, SNAPSHOT_IN_PROGRESS, &value, &len, true);
2893 			if (rc != 0) {
2894 				/* Not corrupted - process it and continue with iterating through blobs */
2895 				if (ctx->iter_cb_fn) {
2896 					ctx->iter_cb_fn(ctx->iter_cb_arg, blob, 0);
2897 				}
2898 				_spdk_bs_blob_list_add(blob);
2899 				spdk_bs_iter_next(ctx->bs, blob, _spdk_bs_load_iter, ctx);
2900 				return;
2901 			}
2902 
2903 		}
2904 
2905 		assert(len == sizeof(spdk_blob_id));
2906 
2907 		ctx->blob = blob;
2908 
2909 		/* Open clone to check if we are able to fix this blob or should we remove it */
2910 		spdk_bs_open_blob(ctx->bs, *(spdk_blob_id *)value, _spdk_bs_examine_clone, ctx);
2911 		return;
2912 	} else if (bserrno == -ENOENT) {
2913 		bserrno = 0;
2914 	} else {
2915 		/*
2916 		 * This case needs to be looked at further.  Same problem
2917 		 *  exists with applications that rely on explicit blob
2918 		 *  iteration.  We should just skip the blob that failed
2919 		 *  to load and continue on to the next one.
2920 		 */
2921 		SPDK_ERRLOG("Error in iterating blobs\n");
2922 	}
2923 
2924 	ctx->iter_cb_fn = NULL;
2925 
2926 	spdk_free(ctx->super);
2927 	spdk_free(ctx->mask);
2928 	spdk_bs_sequence_finish(ctx->seq, bserrno);
2929 	free(ctx);
2930 }
2931 
2932 static void
2933 _spdk_bs_load_complete(struct spdk_bs_load_ctx *ctx)
2934 {
2935 	spdk_bs_iter_first(ctx->bs, _spdk_bs_load_iter, ctx);
2936 }
2937 
2938 static void
2939 _spdk_bs_load_used_blobids_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
2940 {
2941 	struct spdk_bs_load_ctx *ctx = cb_arg;
2942 	int rc;
2943 
2944 	/* The type must be correct */
2945 	assert(ctx->mask->type == SPDK_MD_MASK_TYPE_USED_BLOBIDS);
2946 
2947 	/* The length of the mask (in bits) must not be greater than
2948 	 * the length of the buffer (converted to bits) */
2949 	assert(ctx->mask->length <= (ctx->super->used_blobid_mask_len * SPDK_BS_PAGE_SIZE * 8));
2950 
2951 	/* The length of the mask must be exactly equal to the size
2952 	 * (in pages) of the metadata region */
2953 	assert(ctx->mask->length == ctx->super->md_len);
2954 
2955 	rc = _spdk_bs_load_mask(&ctx->bs->used_blobids, ctx->mask);
2956 	if (rc < 0) {
2957 		spdk_free(ctx->mask);
2958 		_spdk_bs_load_ctx_fail(ctx, rc);
2959 		return;
2960 	}
2961 
2962 	_spdk_bs_load_complete(ctx);
2963 }
2964 
2965 static void
2966 _spdk_bs_load_used_clusters_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
2967 {
2968 	struct spdk_bs_load_ctx *ctx = cb_arg;
2969 	uint64_t		lba, lba_count, mask_size;
2970 	int			rc;
2971 
2972 	if (bserrno != 0) {
2973 		_spdk_bs_load_ctx_fail(ctx, bserrno);
2974 		return;
2975 	}
2976 
2977 	/* The type must be correct */
2978 	assert(ctx->mask->type == SPDK_MD_MASK_TYPE_USED_CLUSTERS);
2979 	/* The length of the mask (in bits) must not be greater than the length of the buffer (converted to bits) */
2980 	assert(ctx->mask->length <= (ctx->super->used_cluster_mask_len * sizeof(
2981 					     struct spdk_blob_md_page) * 8));
2982 	/* The length of the mask must be exactly equal to the total number of clusters */
2983 	assert(ctx->mask->length == ctx->bs->total_clusters);
2984 
2985 	rc = _spdk_bs_load_mask(&ctx->bs->used_clusters, ctx->mask);
2986 	if (rc < 0) {
2987 		spdk_free(ctx->mask);
2988 		_spdk_bs_load_ctx_fail(ctx, rc);
2989 		return;
2990 	}
2991 
2992 	ctx->bs->num_free_clusters = spdk_bit_array_count_clear(ctx->bs->used_clusters);
2993 	assert(ctx->bs->num_free_clusters <= ctx->bs->total_clusters);
2994 
2995 	spdk_free(ctx->mask);
2996 
2997 	/* Read the used blobids mask */
2998 	mask_size = ctx->super->used_blobid_mask_len * SPDK_BS_PAGE_SIZE;
2999 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL, SPDK_ENV_SOCKET_ID_ANY,
3000 				 SPDK_MALLOC_DMA);
3001 	if (!ctx->mask) {
3002 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
3003 		return;
3004 	}
3005 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_blobid_mask_start);
3006 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_blobid_mask_len);
3007 	spdk_bs_sequence_read_dev(seq, ctx->mask, lba, lba_count,
3008 				  _spdk_bs_load_used_blobids_cpl, ctx);
3009 }
3010 
3011 static void
3012 _spdk_bs_load_used_pages_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3013 {
3014 	struct spdk_bs_load_ctx *ctx = cb_arg;
3015 	uint64_t		lba, lba_count, mask_size;
3016 	int			rc;
3017 
3018 	if (bserrno != 0) {
3019 		_spdk_bs_load_ctx_fail(ctx, bserrno);
3020 		return;
3021 	}
3022 
3023 	/* The type must be correct */
3024 	assert(ctx->mask->type == SPDK_MD_MASK_TYPE_USED_PAGES);
3025 	/* The length of the mask (in bits) must not be greater than the length of the buffer (converted to bits) */
3026 	assert(ctx->mask->length <= (ctx->super->used_page_mask_len * SPDK_BS_PAGE_SIZE *
3027 				     8));
3028 	/* The length of the mask must be exactly equal to the size (in pages) of the metadata region */
3029 	assert(ctx->mask->length == ctx->super->md_len);
3030 
3031 	rc = _spdk_bs_load_mask(&ctx->bs->used_md_pages, ctx->mask);
3032 	if (rc < 0) {
3033 		spdk_free(ctx->mask);
3034 		_spdk_bs_load_ctx_fail(ctx, rc);
3035 		return;
3036 	}
3037 
3038 	spdk_free(ctx->mask);
3039 
3040 	/* Read the used clusters mask */
3041 	mask_size = ctx->super->used_cluster_mask_len * SPDK_BS_PAGE_SIZE;
3042 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL, SPDK_ENV_SOCKET_ID_ANY,
3043 				 SPDK_MALLOC_DMA);
3044 	if (!ctx->mask) {
3045 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
3046 		return;
3047 	}
3048 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_cluster_mask_start);
3049 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_cluster_mask_len);
3050 	spdk_bs_sequence_read_dev(seq, ctx->mask, lba, lba_count,
3051 				  _spdk_bs_load_used_clusters_cpl, ctx);
3052 }
3053 
3054 static void
3055 _spdk_bs_load_read_used_pages(struct spdk_bs_load_ctx *ctx)
3056 {
3057 	uint64_t lba, lba_count, mask_size;
3058 
3059 	/* Read the used pages mask */
3060 	mask_size = ctx->super->used_page_mask_len * SPDK_BS_PAGE_SIZE;
3061 	ctx->mask = spdk_zmalloc(mask_size, 0x1000, NULL,
3062 				 SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3063 	if (!ctx->mask) {
3064 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
3065 		return;
3066 	}
3067 
3068 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_page_mask_start);
3069 	lba_count = _spdk_bs_page_to_lba(ctx->bs, ctx->super->used_page_mask_len);
3070 	spdk_bs_sequence_read_dev(ctx->seq, ctx->mask, lba, lba_count,
3071 				  _spdk_bs_load_used_pages_cpl, ctx);
3072 }
3073 
3074 static int
3075 _spdk_bs_load_replay_md_parse_page(const struct spdk_blob_md_page *page, struct spdk_blob_store *bs)
3076 {
3077 	struct spdk_blob_md_descriptor *desc;
3078 	size_t	cur_desc = 0;
3079 
3080 	desc = (struct spdk_blob_md_descriptor *)page->descriptors;
3081 	while (cur_desc < sizeof(page->descriptors)) {
3082 		if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_PADDING) {
3083 			if (desc->length == 0) {
3084 				/* If padding and length are 0, this terminates the page */
3085 				break;
3086 			}
3087 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE) {
3088 			struct spdk_blob_md_descriptor_extent_rle	*desc_extent_rle;
3089 			unsigned int				i, j;
3090 			unsigned int				cluster_count = 0;
3091 			uint32_t				cluster_idx;
3092 
3093 			desc_extent_rle = (struct spdk_blob_md_descriptor_extent_rle *)desc;
3094 
3095 			for (i = 0; i < desc_extent_rle->length / sizeof(desc_extent_rle->extents[0]); i++) {
3096 				for (j = 0; j < desc_extent_rle->extents[i].length; j++) {
3097 					cluster_idx = desc_extent_rle->extents[i].cluster_idx;
3098 					/*
3099 					 * cluster_idx = 0 means an unallocated cluster - don't mark that
3100 					 * in the used cluster map.
3101 					 */
3102 					if (cluster_idx != 0) {
3103 						spdk_bit_array_set(bs->used_clusters, cluster_idx + j);
3104 						if (bs->num_free_clusters == 0) {
3105 							return -ENOSPC;
3106 						}
3107 						bs->num_free_clusters--;
3108 					}
3109 					cluster_count++;
3110 				}
3111 			}
3112 			if (cluster_count == 0) {
3113 				return -EINVAL;
3114 			}
3115 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR) {
3116 			/* Skip this item */
3117 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL) {
3118 			/* Skip this item */
3119 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_FLAGS) {
3120 			/* Skip this item */
3121 		} else {
3122 			/* Error */
3123 			return -EINVAL;
3124 		}
3125 		/* Advance to the next descriptor */
3126 		cur_desc += sizeof(*desc) + desc->length;
3127 		if (cur_desc + sizeof(*desc) > sizeof(page->descriptors)) {
3128 			break;
3129 		}
3130 		desc = (struct spdk_blob_md_descriptor *)((uintptr_t)page->descriptors + cur_desc);
3131 	}
3132 	return 0;
3133 }
3134 
3135 static bool _spdk_bs_load_cur_md_page_valid(struct spdk_bs_load_ctx *ctx)
3136 {
3137 	uint32_t crc;
3138 
3139 	crc = _spdk_blob_md_page_calc_crc(ctx->page);
3140 	if (crc != ctx->page->crc) {
3141 		return false;
3142 	}
3143 
3144 	if (ctx->page->sequence_num == 0 &&
3145 	    _spdk_bs_page_to_blobid(ctx->cur_page) != ctx->page->id) {
3146 		return false;
3147 	}
3148 	return true;
3149 }
3150 
3151 static void
3152 _spdk_bs_load_replay_cur_md_page(struct spdk_bs_load_ctx *ctx);
3153 
3154 static void
3155 _spdk_bs_load_write_used_clusters_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3156 {
3157 	struct spdk_bs_load_ctx	*ctx = cb_arg;
3158 
3159 	if (bserrno != 0) {
3160 		_spdk_bs_load_ctx_fail(ctx, bserrno);
3161 		return;
3162 	}
3163 
3164 	_spdk_bs_load_complete(ctx);
3165 }
3166 
3167 static void
3168 _spdk_bs_load_write_used_blobids_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3169 {
3170 	struct spdk_bs_load_ctx	*ctx = cb_arg;
3171 
3172 	spdk_free(ctx->mask);
3173 	ctx->mask = NULL;
3174 
3175 	if (bserrno != 0) {
3176 		_spdk_bs_load_ctx_fail(ctx, bserrno);
3177 		return;
3178 	}
3179 
3180 	_spdk_bs_write_used_clusters(seq, ctx, _spdk_bs_load_write_used_clusters_cpl);
3181 }
3182 
3183 static void
3184 _spdk_bs_load_write_used_pages_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3185 {
3186 	struct spdk_bs_load_ctx	*ctx = cb_arg;
3187 
3188 	spdk_free(ctx->mask);
3189 	ctx->mask = NULL;
3190 
3191 	if (bserrno != 0) {
3192 		_spdk_bs_load_ctx_fail(ctx, bserrno);
3193 		return;
3194 	}
3195 
3196 	_spdk_bs_write_used_blobids(seq, ctx, _spdk_bs_load_write_used_blobids_cpl);
3197 }
3198 
3199 static void
3200 _spdk_bs_load_write_used_md(struct spdk_bs_load_ctx *ctx)
3201 {
3202 	_spdk_bs_write_used_md(ctx->seq, ctx, _spdk_bs_load_write_used_pages_cpl);
3203 }
3204 
3205 static void
3206 _spdk_bs_load_replay_md_chain_cpl(struct spdk_bs_load_ctx *ctx)
3207 {
3208 	uint64_t num_md_clusters;
3209 	uint64_t i;
3210 
3211 	ctx->in_page_chain = false;
3212 
3213 	do {
3214 		ctx->page_index++;
3215 	} while (spdk_bit_array_get(ctx->bs->used_md_pages, ctx->page_index) == true);
3216 
3217 	if (ctx->page_index < ctx->super->md_len) {
3218 		ctx->cur_page = ctx->page_index;
3219 		_spdk_bs_load_replay_cur_md_page(ctx);
3220 	} else {
3221 		/* Claim all of the clusters used by the metadata */
3222 		num_md_clusters = spdk_divide_round_up(ctx->super->md_len, ctx->bs->pages_per_cluster);
3223 		for (i = 0; i < num_md_clusters; i++) {
3224 			_spdk_bs_claim_cluster(ctx->bs, i);
3225 		}
3226 		spdk_free(ctx->page);
3227 		_spdk_bs_load_write_used_md(ctx);
3228 	}
3229 }
3230 
3231 static void
3232 _spdk_bs_load_replay_md_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3233 {
3234 	struct spdk_bs_load_ctx *ctx = cb_arg;
3235 	uint32_t page_num;
3236 
3237 	if (bserrno != 0) {
3238 		_spdk_bs_load_ctx_fail(ctx, bserrno);
3239 		return;
3240 	}
3241 
3242 	page_num = ctx->cur_page;
3243 	if (_spdk_bs_load_cur_md_page_valid(ctx) == true) {
3244 		if (ctx->page->sequence_num == 0 || ctx->in_page_chain == true) {
3245 			spdk_bit_array_set(ctx->bs->used_md_pages, page_num);
3246 			if (ctx->page->sequence_num == 0) {
3247 				spdk_bit_array_set(ctx->bs->used_blobids, page_num);
3248 			}
3249 			if (_spdk_bs_load_replay_md_parse_page(ctx->page, ctx->bs)) {
3250 				_spdk_bs_load_ctx_fail(ctx, -EILSEQ);
3251 				return;
3252 			}
3253 			if (ctx->page->next != SPDK_INVALID_MD_PAGE) {
3254 				ctx->in_page_chain = true;
3255 				ctx->cur_page = ctx->page->next;
3256 				_spdk_bs_load_replay_cur_md_page(ctx);
3257 				return;
3258 			}
3259 		}
3260 	}
3261 	_spdk_bs_load_replay_md_chain_cpl(ctx);
3262 }
3263 
3264 static void
3265 _spdk_bs_load_replay_cur_md_page(struct spdk_bs_load_ctx *ctx)
3266 {
3267 	uint64_t lba;
3268 
3269 	assert(ctx->cur_page < ctx->super->md_len);
3270 	lba = _spdk_bs_md_page_to_lba(ctx->bs, ctx->cur_page);
3271 	spdk_bs_sequence_read_dev(ctx->seq, ctx->page, lba,
3272 				  _spdk_bs_byte_to_lba(ctx->bs, SPDK_BS_PAGE_SIZE),
3273 				  _spdk_bs_load_replay_md_cpl, ctx);
3274 }
3275 
3276 static void
3277 _spdk_bs_load_replay_md(struct spdk_bs_load_ctx *ctx)
3278 {
3279 	ctx->page_index = 0;
3280 	ctx->cur_page = 0;
3281 	ctx->page = spdk_zmalloc(SPDK_BS_PAGE_SIZE, SPDK_BS_PAGE_SIZE,
3282 				 NULL, SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3283 	if (!ctx->page) {
3284 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
3285 		return;
3286 	}
3287 	_spdk_bs_load_replay_cur_md_page(ctx);
3288 }
3289 
3290 static void
3291 _spdk_bs_recover(struct spdk_bs_load_ctx *ctx)
3292 {
3293 	int		rc;
3294 
3295 	rc = spdk_bit_array_resize(&ctx->bs->used_md_pages, ctx->super->md_len);
3296 	if (rc < 0) {
3297 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
3298 		return;
3299 	}
3300 
3301 	rc = spdk_bit_array_resize(&ctx->bs->used_blobids, ctx->super->md_len);
3302 	if (rc < 0) {
3303 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
3304 		return;
3305 	}
3306 
3307 	rc = spdk_bit_array_resize(&ctx->bs->used_clusters, ctx->bs->total_clusters);
3308 	if (rc < 0) {
3309 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
3310 		return;
3311 	}
3312 
3313 	ctx->bs->num_free_clusters = ctx->bs->total_clusters;
3314 	_spdk_bs_load_replay_md(ctx);
3315 }
3316 
3317 static void
3318 _spdk_bs_load_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3319 {
3320 	struct spdk_bs_load_ctx *ctx = cb_arg;
3321 	uint32_t	crc;
3322 	int		rc;
3323 	static const char zeros[SPDK_BLOBSTORE_TYPE_LENGTH];
3324 
3325 	if (ctx->super->version > SPDK_BS_VERSION ||
3326 	    ctx->super->version < SPDK_BS_INITIAL_VERSION) {
3327 		_spdk_bs_load_ctx_fail(ctx, -EILSEQ);
3328 		return;
3329 	}
3330 
3331 	if (memcmp(ctx->super->signature, SPDK_BS_SUPER_BLOCK_SIG,
3332 		   sizeof(ctx->super->signature)) != 0) {
3333 		_spdk_bs_load_ctx_fail(ctx, -EILSEQ);
3334 		return;
3335 	}
3336 
3337 	crc = _spdk_blob_md_page_calc_crc(ctx->super);
3338 	if (crc != ctx->super->crc) {
3339 		_spdk_bs_load_ctx_fail(ctx, -EILSEQ);
3340 		return;
3341 	}
3342 
3343 	if (memcmp(&ctx->bs->bstype, &ctx->super->bstype, SPDK_BLOBSTORE_TYPE_LENGTH) == 0) {
3344 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Bstype matched - loading blobstore\n");
3345 	} else if (memcmp(&ctx->bs->bstype, zeros, SPDK_BLOBSTORE_TYPE_LENGTH) == 0) {
3346 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Bstype wildcard used - loading blobstore regardless bstype\n");
3347 	} else {
3348 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Unexpected bstype\n");
3349 		SPDK_LOGDUMP(SPDK_LOG_BLOB, "Expected:", ctx->bs->bstype.bstype, SPDK_BLOBSTORE_TYPE_LENGTH);
3350 		SPDK_LOGDUMP(SPDK_LOG_BLOB, "Found:", ctx->super->bstype.bstype, SPDK_BLOBSTORE_TYPE_LENGTH);
3351 		_spdk_bs_load_ctx_fail(ctx, -ENXIO);
3352 		return;
3353 	}
3354 
3355 	if (ctx->super->size > ctx->bs->dev->blockcnt * ctx->bs->dev->blocklen) {
3356 		SPDK_NOTICELOG("Size mismatch, dev size: %lu, blobstore size: %lu\n",
3357 			       ctx->bs->dev->blockcnt * ctx->bs->dev->blocklen, ctx->super->size);
3358 		_spdk_bs_load_ctx_fail(ctx, -EILSEQ);
3359 		return;
3360 	}
3361 
3362 	if (ctx->super->size == 0) {
3363 		ctx->super->size = ctx->bs->dev->blockcnt * ctx->bs->dev->blocklen;
3364 	}
3365 
3366 	if (ctx->super->io_unit_size == 0) {
3367 		ctx->super->io_unit_size = SPDK_BS_PAGE_SIZE;
3368 	}
3369 
3370 	/* Parse the super block */
3371 	ctx->bs->clean = 1;
3372 	ctx->bs->cluster_sz = ctx->super->cluster_size;
3373 	ctx->bs->total_clusters = ctx->super->size / ctx->super->cluster_size;
3374 	ctx->bs->pages_per_cluster = ctx->bs->cluster_sz / SPDK_BS_PAGE_SIZE;
3375 	ctx->bs->io_unit_size = ctx->super->io_unit_size;
3376 	rc = spdk_bit_array_resize(&ctx->bs->used_clusters, ctx->bs->total_clusters);
3377 	if (rc < 0) {
3378 		_spdk_bs_load_ctx_fail(ctx, -ENOMEM);
3379 		return;
3380 	}
3381 	ctx->bs->md_start = ctx->super->md_start;
3382 	ctx->bs->md_len = ctx->super->md_len;
3383 	ctx->bs->total_data_clusters = ctx->bs->total_clusters - spdk_divide_round_up(
3384 					       ctx->bs->md_start + ctx->bs->md_len, ctx->bs->pages_per_cluster);
3385 	ctx->bs->super_blob = ctx->super->super_blob;
3386 	memcpy(&ctx->bs->bstype, &ctx->super->bstype, sizeof(ctx->super->bstype));
3387 
3388 	if (ctx->super->used_blobid_mask_len == 0 || ctx->super->clean == 0) {
3389 		_spdk_bs_recover(ctx);
3390 	} else {
3391 		_spdk_bs_load_read_used_pages(ctx);
3392 	}
3393 }
3394 
3395 void
3396 spdk_bs_load(struct spdk_bs_dev *dev, struct spdk_bs_opts *o,
3397 	     spdk_bs_op_with_handle_complete cb_fn, void *cb_arg)
3398 {
3399 	struct spdk_blob_store	*bs;
3400 	struct spdk_bs_cpl	cpl;
3401 	struct spdk_bs_load_ctx *ctx;
3402 	struct spdk_bs_opts	opts = {};
3403 	int err;
3404 
3405 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Loading blobstore from dev %p\n", dev);
3406 
3407 	if ((SPDK_BS_PAGE_SIZE % dev->blocklen) != 0) {
3408 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "unsupported dev block length of %d\n", dev->blocklen);
3409 		dev->destroy(dev);
3410 		cb_fn(cb_arg, NULL, -EINVAL);
3411 		return;
3412 	}
3413 
3414 	if (o) {
3415 		opts = *o;
3416 	} else {
3417 		spdk_bs_opts_init(&opts);
3418 	}
3419 
3420 	if (opts.max_md_ops == 0 || opts.max_channel_ops == 0) {
3421 		dev->destroy(dev);
3422 		cb_fn(cb_arg, NULL, -EINVAL);
3423 		return;
3424 	}
3425 
3426 	err = _spdk_bs_alloc(dev, &opts, &bs);
3427 	if (err) {
3428 		dev->destroy(dev);
3429 		cb_fn(cb_arg, NULL, err);
3430 		return;
3431 	}
3432 
3433 	ctx = calloc(1, sizeof(*ctx));
3434 	if (!ctx) {
3435 		_spdk_bs_free(bs);
3436 		cb_fn(cb_arg, NULL, -ENOMEM);
3437 		return;
3438 	}
3439 
3440 	ctx->bs = bs;
3441 	ctx->iter_cb_fn = opts.iter_cb_fn;
3442 	ctx->iter_cb_arg = opts.iter_cb_arg;
3443 
3444 	/* Allocate memory for the super block */
3445 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
3446 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3447 	if (!ctx->super) {
3448 		free(ctx);
3449 		_spdk_bs_free(bs);
3450 		cb_fn(cb_arg, NULL, -ENOMEM);
3451 		return;
3452 	}
3453 
3454 	cpl.type = SPDK_BS_CPL_TYPE_BS_HANDLE;
3455 	cpl.u.bs_handle.cb_fn = cb_fn;
3456 	cpl.u.bs_handle.cb_arg = cb_arg;
3457 	cpl.u.bs_handle.bs = bs;
3458 
3459 	ctx->seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
3460 	if (!ctx->seq) {
3461 		spdk_free(ctx->super);
3462 		free(ctx);
3463 		_spdk_bs_free(bs);
3464 		cb_fn(cb_arg, NULL, -ENOMEM);
3465 		return;
3466 	}
3467 
3468 	/* Read the super block */
3469 	spdk_bs_sequence_read_dev(ctx->seq, ctx->super, _spdk_bs_page_to_lba(bs, 0),
3470 				  _spdk_bs_byte_to_lba(bs, sizeof(*ctx->super)),
3471 				  _spdk_bs_load_super_cpl, ctx);
3472 }
3473 
3474 /* END spdk_bs_load */
3475 
3476 /* START spdk_bs_dump */
3477 
3478 struct spdk_bs_dump_ctx {
3479 	struct spdk_blob_store		*bs;
3480 	struct spdk_bs_super_block	*super;
3481 	uint32_t			cur_page;
3482 	struct spdk_blob_md_page	*page;
3483 	spdk_bs_sequence_t		*seq;
3484 	FILE				*fp;
3485 	spdk_bs_dump_print_xattr	print_xattr_fn;
3486 	char				xattr_name[4096];
3487 };
3488 
3489 static void
3490 _spdk_bs_dump_finish(spdk_bs_sequence_t *seq, struct spdk_bs_dump_ctx *ctx, int bserrno)
3491 {
3492 	spdk_free(ctx->super);
3493 
3494 	/*
3495 	 * We need to defer calling spdk_bs_call_cpl() until after
3496 	 * dev destruction, so tuck these away for later use.
3497 	 */
3498 	ctx->bs->unload_err = bserrno;
3499 	memcpy(&ctx->bs->unload_cpl, &seq->cpl, sizeof(struct spdk_bs_cpl));
3500 	seq->cpl.type = SPDK_BS_CPL_TYPE_NONE;
3501 
3502 	spdk_bs_sequence_finish(seq, 0);
3503 	_spdk_bs_free(ctx->bs);
3504 	free(ctx);
3505 }
3506 
3507 static void _spdk_bs_dump_read_md_page(spdk_bs_sequence_t *seq, void *cb_arg);
3508 
3509 static void
3510 _spdk_bs_dump_print_md_page(struct spdk_bs_dump_ctx *ctx)
3511 {
3512 	uint32_t page_idx = ctx->cur_page;
3513 	struct spdk_blob_md_page *page = ctx->page;
3514 	struct spdk_blob_md_descriptor *desc;
3515 	size_t cur_desc = 0;
3516 	uint32_t crc;
3517 
3518 	fprintf(ctx->fp, "=========\n");
3519 	fprintf(ctx->fp, "Metadata Page Index: %" PRIu32 " (0x%" PRIx32 ")\n", page_idx, page_idx);
3520 	fprintf(ctx->fp, "Blob ID: 0x%" PRIx64 "\n", page->id);
3521 
3522 	crc = _spdk_blob_md_page_calc_crc(page);
3523 	fprintf(ctx->fp, "CRC: 0x%" PRIx32 " (%s)\n", page->crc, crc == page->crc ? "OK" : "Mismatch");
3524 
3525 	desc = (struct spdk_blob_md_descriptor *)page->descriptors;
3526 	while (cur_desc < sizeof(page->descriptors)) {
3527 		if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_PADDING) {
3528 			if (desc->length == 0) {
3529 				/* If padding and length are 0, this terminates the page */
3530 				break;
3531 			}
3532 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE) {
3533 			struct spdk_blob_md_descriptor_extent_rle	*desc_extent_rle;
3534 			unsigned int				i;
3535 
3536 			desc_extent_rle = (struct spdk_blob_md_descriptor_extent_rle *)desc;
3537 
3538 			for (i = 0; i < desc_extent_rle->length / sizeof(desc_extent_rle->extents[0]); i++) {
3539 				if (desc_extent_rle->extents[i].cluster_idx != 0) {
3540 					fprintf(ctx->fp, "Allocated Extent - Start: %" PRIu32,
3541 						desc_extent_rle->extents[i].cluster_idx);
3542 				} else {
3543 					fprintf(ctx->fp, "Unallocated Extent - ");
3544 				}
3545 				fprintf(ctx->fp, " Length: %" PRIu32, desc_extent_rle->extents[i].length);
3546 				fprintf(ctx->fp, "\n");
3547 			}
3548 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR) {
3549 			struct spdk_blob_md_descriptor_xattr *desc_xattr;
3550 			uint32_t i;
3551 
3552 			desc_xattr = (struct spdk_blob_md_descriptor_xattr *)desc;
3553 
3554 			if (desc_xattr->length !=
3555 			    sizeof(desc_xattr->name_length) + sizeof(desc_xattr->value_length) +
3556 			    desc_xattr->name_length + desc_xattr->value_length) {
3557 			}
3558 
3559 			memcpy(ctx->xattr_name, desc_xattr->name, desc_xattr->name_length);
3560 			ctx->xattr_name[desc_xattr->name_length] = '\0';
3561 			fprintf(ctx->fp, "XATTR: name = \"%s\"\n", ctx->xattr_name);
3562 			fprintf(ctx->fp, "       value = \"");
3563 			ctx->print_xattr_fn(ctx->fp, ctx->super->bstype.bstype, ctx->xattr_name,
3564 					    (void *)((uintptr_t)desc_xattr->name + desc_xattr->name_length),
3565 					    desc_xattr->value_length);
3566 			fprintf(ctx->fp, "\"\n");
3567 			for (i = 0; i < desc_xattr->value_length; i++) {
3568 				if (i % 16 == 0) {
3569 					fprintf(ctx->fp, "               ");
3570 				}
3571 				fprintf(ctx->fp, "%02" PRIx8 " ", *((uint8_t *)desc_xattr->name + desc_xattr->name_length + i));
3572 				if ((i + 1) % 16 == 0) {
3573 					fprintf(ctx->fp, "\n");
3574 				}
3575 			}
3576 			if (i % 16 != 0) {
3577 				fprintf(ctx->fp, "\n");
3578 			}
3579 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL) {
3580 			/* TODO */
3581 		} else if (desc->type == SPDK_MD_DESCRIPTOR_TYPE_FLAGS) {
3582 			/* TODO */
3583 		} else {
3584 			/* Error */
3585 		}
3586 		/* Advance to the next descriptor */
3587 		cur_desc += sizeof(*desc) + desc->length;
3588 		if (cur_desc + sizeof(*desc) > sizeof(page->descriptors)) {
3589 			break;
3590 		}
3591 		desc = (struct spdk_blob_md_descriptor *)((uintptr_t)page->descriptors + cur_desc);
3592 	}
3593 }
3594 
3595 static void
3596 _spdk_bs_dump_read_md_page_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3597 {
3598 	struct spdk_bs_dump_ctx *ctx = cb_arg;
3599 
3600 	if (bserrno != 0) {
3601 		_spdk_bs_dump_finish(seq, ctx, bserrno);
3602 		return;
3603 	}
3604 
3605 	if (ctx->page->id != 0) {
3606 		_spdk_bs_dump_print_md_page(ctx);
3607 	}
3608 
3609 	ctx->cur_page++;
3610 
3611 	if (ctx->cur_page < ctx->super->md_len) {
3612 		_spdk_bs_dump_read_md_page(seq, ctx);
3613 	} else {
3614 		spdk_free(ctx->page);
3615 		_spdk_bs_dump_finish(seq, ctx, 0);
3616 	}
3617 }
3618 
3619 static void
3620 _spdk_bs_dump_read_md_page(spdk_bs_sequence_t *seq, void *cb_arg)
3621 {
3622 	struct spdk_bs_dump_ctx *ctx = cb_arg;
3623 	uint64_t lba;
3624 
3625 	assert(ctx->cur_page < ctx->super->md_len);
3626 	lba = _spdk_bs_page_to_lba(ctx->bs, ctx->super->md_start + ctx->cur_page);
3627 	spdk_bs_sequence_read_dev(seq, ctx->page, lba,
3628 				  _spdk_bs_byte_to_lba(ctx->bs, SPDK_BS_PAGE_SIZE),
3629 				  _spdk_bs_dump_read_md_page_cpl, ctx);
3630 }
3631 
3632 static void
3633 _spdk_bs_dump_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3634 {
3635 	struct spdk_bs_dump_ctx *ctx = cb_arg;
3636 
3637 	fprintf(ctx->fp, "Signature: \"%.8s\" ", ctx->super->signature);
3638 	if (memcmp(ctx->super->signature, SPDK_BS_SUPER_BLOCK_SIG,
3639 		   sizeof(ctx->super->signature)) != 0) {
3640 		fprintf(ctx->fp, "(Mismatch)\n");
3641 		_spdk_bs_dump_finish(seq, ctx, bserrno);
3642 		return;
3643 	} else {
3644 		fprintf(ctx->fp, "(OK)\n");
3645 	}
3646 	fprintf(ctx->fp, "Version: %" PRIu32 "\n", ctx->super->version);
3647 	fprintf(ctx->fp, "CRC: 0x%x (%s)\n", ctx->super->crc,
3648 		(ctx->super->crc == _spdk_blob_md_page_calc_crc(ctx->super)) ? "OK" : "Mismatch");
3649 	fprintf(ctx->fp, "Blobstore Type: %.*s\n", SPDK_BLOBSTORE_TYPE_LENGTH, ctx->super->bstype.bstype);
3650 	fprintf(ctx->fp, "Cluster Size: %" PRIu32 "\n", ctx->super->cluster_size);
3651 	fprintf(ctx->fp, "Super Blob ID: ");
3652 	if (ctx->super->super_blob == SPDK_BLOBID_INVALID) {
3653 		fprintf(ctx->fp, "(None)\n");
3654 	} else {
3655 		fprintf(ctx->fp, "%" PRIu64 "\n", ctx->super->super_blob);
3656 	}
3657 	fprintf(ctx->fp, "Clean: %" PRIu32 "\n", ctx->super->clean);
3658 	fprintf(ctx->fp, "Used Metadata Page Mask Start: %" PRIu32 "\n", ctx->super->used_page_mask_start);
3659 	fprintf(ctx->fp, "Used Metadata Page Mask Length: %" PRIu32 "\n", ctx->super->used_page_mask_len);
3660 	fprintf(ctx->fp, "Used Cluster Mask Start: %" PRIu32 "\n", ctx->super->used_cluster_mask_start);
3661 	fprintf(ctx->fp, "Used Cluster Mask Length: %" PRIu32 "\n", ctx->super->used_cluster_mask_len);
3662 	fprintf(ctx->fp, "Used Blob ID Mask Start: %" PRIu32 "\n", ctx->super->used_blobid_mask_start);
3663 	fprintf(ctx->fp, "Used Blob ID Mask Length: %" PRIu32 "\n", ctx->super->used_blobid_mask_len);
3664 	fprintf(ctx->fp, "Metadata Start: %" PRIu32 "\n", ctx->super->md_start);
3665 	fprintf(ctx->fp, "Metadata Length: %" PRIu32 "\n", ctx->super->md_len);
3666 
3667 	ctx->cur_page = 0;
3668 	ctx->page = spdk_zmalloc(SPDK_BS_PAGE_SIZE, SPDK_BS_PAGE_SIZE,
3669 				 NULL, SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3670 	if (!ctx->page) {
3671 		_spdk_bs_dump_finish(seq, ctx, -ENOMEM);
3672 		return;
3673 	}
3674 	_spdk_bs_dump_read_md_page(seq, ctx);
3675 }
3676 
3677 void
3678 spdk_bs_dump(struct spdk_bs_dev *dev, FILE *fp, spdk_bs_dump_print_xattr print_xattr_fn,
3679 	     spdk_bs_op_complete cb_fn, void *cb_arg)
3680 {
3681 	struct spdk_blob_store	*bs;
3682 	struct spdk_bs_cpl	cpl;
3683 	spdk_bs_sequence_t	*seq;
3684 	struct spdk_bs_dump_ctx *ctx;
3685 	struct spdk_bs_opts	opts = {};
3686 	int err;
3687 
3688 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Dumping blobstore from dev %p\n", dev);
3689 
3690 	spdk_bs_opts_init(&opts);
3691 
3692 	err = _spdk_bs_alloc(dev, &opts, &bs);
3693 	if (err) {
3694 		dev->destroy(dev);
3695 		cb_fn(cb_arg, err);
3696 		return;
3697 	}
3698 
3699 	ctx = calloc(1, sizeof(*ctx));
3700 	if (!ctx) {
3701 		_spdk_bs_free(bs);
3702 		cb_fn(cb_arg, -ENOMEM);
3703 		return;
3704 	}
3705 
3706 	ctx->bs = bs;
3707 	ctx->fp = fp;
3708 	ctx->print_xattr_fn = print_xattr_fn;
3709 
3710 	/* Allocate memory for the super block */
3711 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
3712 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3713 	if (!ctx->super) {
3714 		free(ctx);
3715 		_spdk_bs_free(bs);
3716 		cb_fn(cb_arg, -ENOMEM);
3717 		return;
3718 	}
3719 
3720 	cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
3721 	cpl.u.bs_basic.cb_fn = cb_fn;
3722 	cpl.u.bs_basic.cb_arg = cb_arg;
3723 
3724 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
3725 	if (!seq) {
3726 		spdk_free(ctx->super);
3727 		free(ctx);
3728 		_spdk_bs_free(bs);
3729 		cb_fn(cb_arg, -ENOMEM);
3730 		return;
3731 	}
3732 
3733 	/* Read the super block */
3734 	spdk_bs_sequence_read_dev(seq, ctx->super, _spdk_bs_page_to_lba(bs, 0),
3735 				  _spdk_bs_byte_to_lba(bs, sizeof(*ctx->super)),
3736 				  _spdk_bs_dump_super_cpl, ctx);
3737 }
3738 
3739 /* END spdk_bs_dump */
3740 
3741 /* START spdk_bs_init */
3742 
3743 struct spdk_bs_init_ctx {
3744 	struct spdk_blob_store		*bs;
3745 	struct spdk_bs_super_block	*super;
3746 };
3747 
3748 static void
3749 _spdk_bs_init_persist_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3750 {
3751 	struct spdk_bs_init_ctx *ctx = cb_arg;
3752 
3753 	spdk_free(ctx->super);
3754 	free(ctx);
3755 
3756 	spdk_bs_sequence_finish(seq, bserrno);
3757 }
3758 
3759 static void
3760 _spdk_bs_init_trim_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3761 {
3762 	struct spdk_bs_init_ctx *ctx = cb_arg;
3763 
3764 	/* Write super block */
3765 	spdk_bs_sequence_write_dev(seq, ctx->super, _spdk_bs_page_to_lba(ctx->bs, 0),
3766 				   _spdk_bs_byte_to_lba(ctx->bs, sizeof(*ctx->super)),
3767 				   _spdk_bs_init_persist_super_cpl, ctx);
3768 }
3769 
3770 void
3771 spdk_bs_init(struct spdk_bs_dev *dev, struct spdk_bs_opts *o,
3772 	     spdk_bs_op_with_handle_complete cb_fn, void *cb_arg)
3773 {
3774 	struct spdk_bs_init_ctx *ctx;
3775 	struct spdk_blob_store	*bs;
3776 	struct spdk_bs_cpl	cpl;
3777 	spdk_bs_sequence_t	*seq;
3778 	spdk_bs_batch_t		*batch;
3779 	uint64_t		num_md_lba;
3780 	uint64_t		num_md_pages;
3781 	uint64_t		num_md_clusters;
3782 	uint32_t		i;
3783 	struct spdk_bs_opts	opts = {};
3784 	int			rc;
3785 
3786 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Initializing blobstore on dev %p\n", dev);
3787 
3788 	if ((SPDK_BS_PAGE_SIZE % dev->blocklen) != 0) {
3789 		SPDK_ERRLOG("unsupported dev block length of %d\n",
3790 			    dev->blocklen);
3791 		dev->destroy(dev);
3792 		cb_fn(cb_arg, NULL, -EINVAL);
3793 		return;
3794 	}
3795 
3796 	if (o) {
3797 		opts = *o;
3798 	} else {
3799 		spdk_bs_opts_init(&opts);
3800 	}
3801 
3802 	if (_spdk_bs_opts_verify(&opts) != 0) {
3803 		dev->destroy(dev);
3804 		cb_fn(cb_arg, NULL, -EINVAL);
3805 		return;
3806 	}
3807 
3808 	rc = _spdk_bs_alloc(dev, &opts, &bs);
3809 	if (rc) {
3810 		dev->destroy(dev);
3811 		cb_fn(cb_arg, NULL, rc);
3812 		return;
3813 	}
3814 
3815 	if (opts.num_md_pages == SPDK_BLOB_OPTS_NUM_MD_PAGES) {
3816 		/* By default, allocate 1 page per cluster.
3817 		 * Technically, this over-allocates metadata
3818 		 * because more metadata will reduce the number
3819 		 * of usable clusters. This can be addressed with
3820 		 * more complex math in the future.
3821 		 */
3822 		bs->md_len = bs->total_clusters;
3823 	} else {
3824 		bs->md_len = opts.num_md_pages;
3825 	}
3826 	rc = spdk_bit_array_resize(&bs->used_md_pages, bs->md_len);
3827 	if (rc < 0) {
3828 		_spdk_bs_free(bs);
3829 		cb_fn(cb_arg, NULL, -ENOMEM);
3830 		return;
3831 	}
3832 
3833 	rc = spdk_bit_array_resize(&bs->used_blobids, bs->md_len);
3834 	if (rc < 0) {
3835 		_spdk_bs_free(bs);
3836 		cb_fn(cb_arg, NULL, -ENOMEM);
3837 		return;
3838 	}
3839 
3840 	ctx = calloc(1, sizeof(*ctx));
3841 	if (!ctx) {
3842 		_spdk_bs_free(bs);
3843 		cb_fn(cb_arg, NULL, -ENOMEM);
3844 		return;
3845 	}
3846 
3847 	ctx->bs = bs;
3848 
3849 	/* Allocate memory for the super block */
3850 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
3851 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
3852 	if (!ctx->super) {
3853 		free(ctx);
3854 		_spdk_bs_free(bs);
3855 		cb_fn(cb_arg, NULL, -ENOMEM);
3856 		return;
3857 	}
3858 	memcpy(ctx->super->signature, SPDK_BS_SUPER_BLOCK_SIG,
3859 	       sizeof(ctx->super->signature));
3860 	ctx->super->version = SPDK_BS_VERSION;
3861 	ctx->super->length = sizeof(*ctx->super);
3862 	ctx->super->super_blob = bs->super_blob;
3863 	ctx->super->clean = 0;
3864 	ctx->super->cluster_size = bs->cluster_sz;
3865 	ctx->super->io_unit_size = bs->io_unit_size;
3866 	memcpy(&ctx->super->bstype, &bs->bstype, sizeof(bs->bstype));
3867 
3868 	/* Calculate how many pages the metadata consumes at the front
3869 	 * of the disk.
3870 	 */
3871 
3872 	/* The super block uses 1 page */
3873 	num_md_pages = 1;
3874 
3875 	/* The used_md_pages mask requires 1 bit per metadata page, rounded
3876 	 * up to the nearest page, plus a header.
3877 	 */
3878 	ctx->super->used_page_mask_start = num_md_pages;
3879 	ctx->super->used_page_mask_len = spdk_divide_round_up(sizeof(struct spdk_bs_md_mask) +
3880 					 spdk_divide_round_up(bs->md_len, 8),
3881 					 SPDK_BS_PAGE_SIZE);
3882 	num_md_pages += ctx->super->used_page_mask_len;
3883 
3884 	/* The used_clusters mask requires 1 bit per cluster, rounded
3885 	 * up to the nearest page, plus a header.
3886 	 */
3887 	ctx->super->used_cluster_mask_start = num_md_pages;
3888 	ctx->super->used_cluster_mask_len = spdk_divide_round_up(sizeof(struct spdk_bs_md_mask) +
3889 					    spdk_divide_round_up(bs->total_clusters, 8),
3890 					    SPDK_BS_PAGE_SIZE);
3891 	num_md_pages += ctx->super->used_cluster_mask_len;
3892 
3893 	/* The used_blobids mask requires 1 bit per metadata page, rounded
3894 	 * up to the nearest page, plus a header.
3895 	 */
3896 	ctx->super->used_blobid_mask_start = num_md_pages;
3897 	ctx->super->used_blobid_mask_len = spdk_divide_round_up(sizeof(struct spdk_bs_md_mask) +
3898 					   spdk_divide_round_up(bs->md_len, 8),
3899 					   SPDK_BS_PAGE_SIZE);
3900 	num_md_pages += ctx->super->used_blobid_mask_len;
3901 
3902 	/* The metadata region size was chosen above */
3903 	ctx->super->md_start = bs->md_start = num_md_pages;
3904 	ctx->super->md_len = bs->md_len;
3905 	num_md_pages += bs->md_len;
3906 
3907 	num_md_lba = _spdk_bs_page_to_lba(bs, num_md_pages);
3908 
3909 	ctx->super->size = dev->blockcnt * dev->blocklen;
3910 
3911 	ctx->super->crc = _spdk_blob_md_page_calc_crc(ctx->super);
3912 
3913 	num_md_clusters = spdk_divide_round_up(num_md_pages, bs->pages_per_cluster);
3914 	if (num_md_clusters > bs->total_clusters) {
3915 		SPDK_ERRLOG("Blobstore metadata cannot use more clusters than is available, "
3916 			    "please decrease number of pages reserved for metadata "
3917 			    "or increase cluster size.\n");
3918 		spdk_free(ctx->super);
3919 		free(ctx);
3920 		_spdk_bs_free(bs);
3921 		cb_fn(cb_arg, NULL, -ENOMEM);
3922 		return;
3923 	}
3924 	/* Claim all of the clusters used by the metadata */
3925 	for (i = 0; i < num_md_clusters; i++) {
3926 		_spdk_bs_claim_cluster(bs, i);
3927 	}
3928 
3929 	bs->total_data_clusters = bs->num_free_clusters;
3930 
3931 	cpl.type = SPDK_BS_CPL_TYPE_BS_HANDLE;
3932 	cpl.u.bs_handle.cb_fn = cb_fn;
3933 	cpl.u.bs_handle.cb_arg = cb_arg;
3934 	cpl.u.bs_handle.bs = bs;
3935 
3936 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
3937 	if (!seq) {
3938 		spdk_free(ctx->super);
3939 		free(ctx);
3940 		_spdk_bs_free(bs);
3941 		cb_fn(cb_arg, NULL, -ENOMEM);
3942 		return;
3943 	}
3944 
3945 	batch = spdk_bs_sequence_to_batch(seq, _spdk_bs_init_trim_cpl, ctx);
3946 
3947 	/* Clear metadata space */
3948 	spdk_bs_batch_write_zeroes_dev(batch, 0, num_md_lba);
3949 
3950 	switch (opts.clear_method) {
3951 	case BS_CLEAR_WITH_UNMAP:
3952 		/* Trim data clusters */
3953 		spdk_bs_batch_unmap_dev(batch, num_md_lba, ctx->bs->dev->blockcnt - num_md_lba);
3954 		break;
3955 	case BS_CLEAR_WITH_WRITE_ZEROES:
3956 		/* Write_zeroes to data clusters */
3957 		spdk_bs_batch_write_zeroes_dev(batch, num_md_lba, ctx->bs->dev->blockcnt - num_md_lba);
3958 		break;
3959 	case BS_CLEAR_WITH_NONE:
3960 	default:
3961 		break;
3962 	}
3963 
3964 	spdk_bs_batch_close(batch);
3965 }
3966 
3967 /* END spdk_bs_init */
3968 
3969 /* START spdk_bs_destroy */
3970 
3971 static void
3972 _spdk_bs_destroy_trim_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
3973 {
3974 	struct spdk_bs_init_ctx *ctx = cb_arg;
3975 	struct spdk_blob_store *bs = ctx->bs;
3976 
3977 	/*
3978 	 * We need to defer calling spdk_bs_call_cpl() until after
3979 	 * dev destruction, so tuck these away for later use.
3980 	 */
3981 	bs->unload_err = bserrno;
3982 	memcpy(&bs->unload_cpl, &seq->cpl, sizeof(struct spdk_bs_cpl));
3983 	seq->cpl.type = SPDK_BS_CPL_TYPE_NONE;
3984 
3985 	spdk_bs_sequence_finish(seq, bserrno);
3986 
3987 	_spdk_bs_free(bs);
3988 	free(ctx);
3989 }
3990 
3991 void
3992 spdk_bs_destroy(struct spdk_blob_store *bs, spdk_bs_op_complete cb_fn,
3993 		void *cb_arg)
3994 {
3995 	struct spdk_bs_cpl	cpl;
3996 	spdk_bs_sequence_t	*seq;
3997 	struct spdk_bs_init_ctx *ctx;
3998 
3999 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Destroying blobstore\n");
4000 
4001 	if (!TAILQ_EMPTY(&bs->blobs)) {
4002 		SPDK_ERRLOG("Blobstore still has open blobs\n");
4003 		cb_fn(cb_arg, -EBUSY);
4004 		return;
4005 	}
4006 
4007 	cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
4008 	cpl.u.bs_basic.cb_fn = cb_fn;
4009 	cpl.u.bs_basic.cb_arg = cb_arg;
4010 
4011 	ctx = calloc(1, sizeof(*ctx));
4012 	if (!ctx) {
4013 		cb_fn(cb_arg, -ENOMEM);
4014 		return;
4015 	}
4016 
4017 	ctx->bs = bs;
4018 
4019 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
4020 	if (!seq) {
4021 		free(ctx);
4022 		cb_fn(cb_arg, -ENOMEM);
4023 		return;
4024 	}
4025 
4026 	/* Write zeroes to the super block */
4027 	spdk_bs_sequence_write_zeroes_dev(seq,
4028 					  _spdk_bs_page_to_lba(bs, 0),
4029 					  _spdk_bs_byte_to_lba(bs, sizeof(struct spdk_bs_super_block)),
4030 					  _spdk_bs_destroy_trim_cpl, ctx);
4031 }
4032 
4033 /* END spdk_bs_destroy */
4034 
4035 /* START spdk_bs_unload */
4036 
4037 static void
4038 _spdk_bs_unload_finish(struct spdk_bs_load_ctx *ctx, int bserrno)
4039 {
4040 	spdk_bs_sequence_t *seq = ctx->seq;
4041 
4042 	spdk_free(ctx->super);
4043 
4044 	/*
4045 	 * We need to defer calling spdk_bs_call_cpl() until after
4046 	 * dev destruction, so tuck these away for later use.
4047 	 */
4048 	ctx->bs->unload_err = bserrno;
4049 	memcpy(&ctx->bs->unload_cpl, &seq->cpl, sizeof(struct spdk_bs_cpl));
4050 	seq->cpl.type = SPDK_BS_CPL_TYPE_NONE;
4051 
4052 	spdk_bs_sequence_finish(seq, bserrno);
4053 
4054 	_spdk_bs_free(ctx->bs);
4055 	free(ctx);
4056 }
4057 
4058 static void
4059 _spdk_bs_unload_write_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4060 {
4061 	struct spdk_bs_load_ctx	*ctx = cb_arg;
4062 
4063 	_spdk_bs_unload_finish(ctx, bserrno);
4064 }
4065 
4066 static void
4067 _spdk_bs_unload_write_used_clusters_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4068 {
4069 	struct spdk_bs_load_ctx	*ctx = cb_arg;
4070 
4071 	spdk_free(ctx->mask);
4072 
4073 	if (bserrno != 0) {
4074 		_spdk_bs_unload_finish(ctx, bserrno);
4075 		return;
4076 	}
4077 
4078 	ctx->super->clean = 1;
4079 
4080 	_spdk_bs_write_super(seq, ctx->bs, ctx->super, _spdk_bs_unload_write_super_cpl, ctx);
4081 }
4082 
4083 static void
4084 _spdk_bs_unload_write_used_blobids_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4085 {
4086 	struct spdk_bs_load_ctx	*ctx = cb_arg;
4087 
4088 	spdk_free(ctx->mask);
4089 	ctx->mask = NULL;
4090 
4091 	if (bserrno != 0) {
4092 		_spdk_bs_unload_finish(ctx, bserrno);
4093 		return;
4094 	}
4095 
4096 	_spdk_bs_write_used_clusters(seq, ctx, _spdk_bs_unload_write_used_clusters_cpl);
4097 }
4098 
4099 static void
4100 _spdk_bs_unload_write_used_pages_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4101 {
4102 	struct spdk_bs_load_ctx	*ctx = cb_arg;
4103 
4104 	spdk_free(ctx->mask);
4105 	ctx->mask = NULL;
4106 
4107 	if (bserrno != 0) {
4108 		_spdk_bs_unload_finish(ctx, bserrno);
4109 		return;
4110 	}
4111 
4112 	_spdk_bs_write_used_blobids(seq, ctx, _spdk_bs_unload_write_used_blobids_cpl);
4113 }
4114 
4115 static void
4116 _spdk_bs_unload_read_super_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4117 {
4118 	struct spdk_bs_load_ctx	*ctx = cb_arg;
4119 
4120 	if (bserrno != 0) {
4121 		_spdk_bs_unload_finish(ctx, bserrno);
4122 		return;
4123 	}
4124 
4125 	_spdk_bs_write_used_md(seq, cb_arg, _spdk_bs_unload_write_used_pages_cpl);
4126 }
4127 
4128 void
4129 spdk_bs_unload(struct spdk_blob_store *bs, spdk_bs_op_complete cb_fn, void *cb_arg)
4130 {
4131 	struct spdk_bs_cpl	cpl;
4132 	struct spdk_bs_load_ctx *ctx;
4133 
4134 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Syncing blobstore\n");
4135 
4136 	if (!TAILQ_EMPTY(&bs->blobs)) {
4137 		SPDK_ERRLOG("Blobstore still has open blobs\n");
4138 		cb_fn(cb_arg, -EBUSY);
4139 		return;
4140 	}
4141 
4142 	ctx = calloc(1, sizeof(*ctx));
4143 	if (!ctx) {
4144 		cb_fn(cb_arg, -ENOMEM);
4145 		return;
4146 	}
4147 
4148 	ctx->bs = bs;
4149 
4150 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
4151 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
4152 	if (!ctx->super) {
4153 		free(ctx);
4154 		cb_fn(cb_arg, -ENOMEM);
4155 		return;
4156 	}
4157 
4158 	cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
4159 	cpl.u.bs_basic.cb_fn = cb_fn;
4160 	cpl.u.bs_basic.cb_arg = cb_arg;
4161 
4162 	ctx->seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
4163 	if (!ctx->seq) {
4164 		spdk_free(ctx->super);
4165 		free(ctx);
4166 		cb_fn(cb_arg, -ENOMEM);
4167 		return;
4168 	}
4169 
4170 	/* Read super block */
4171 	spdk_bs_sequence_read_dev(ctx->seq, ctx->super, _spdk_bs_page_to_lba(bs, 0),
4172 				  _spdk_bs_byte_to_lba(bs, sizeof(*ctx->super)),
4173 				  _spdk_bs_unload_read_super_cpl, ctx);
4174 }
4175 
4176 /* END spdk_bs_unload */
4177 
4178 /* START spdk_bs_set_super */
4179 
4180 struct spdk_bs_set_super_ctx {
4181 	struct spdk_blob_store		*bs;
4182 	struct spdk_bs_super_block	*super;
4183 };
4184 
4185 static void
4186 _spdk_bs_set_super_write_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4187 {
4188 	struct spdk_bs_set_super_ctx	*ctx = cb_arg;
4189 
4190 	if (bserrno != 0) {
4191 		SPDK_ERRLOG("Unable to write to super block of blobstore\n");
4192 	}
4193 
4194 	spdk_free(ctx->super);
4195 
4196 	spdk_bs_sequence_finish(seq, bserrno);
4197 
4198 	free(ctx);
4199 }
4200 
4201 static void
4202 _spdk_bs_set_super_read_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4203 {
4204 	struct spdk_bs_set_super_ctx	*ctx = cb_arg;
4205 
4206 	if (bserrno != 0) {
4207 		SPDK_ERRLOG("Unable to read super block of blobstore\n");
4208 		spdk_free(ctx->super);
4209 		spdk_bs_sequence_finish(seq, bserrno);
4210 		free(ctx);
4211 		return;
4212 	}
4213 
4214 	_spdk_bs_write_super(seq, ctx->bs, ctx->super, _spdk_bs_set_super_write_cpl, ctx);
4215 }
4216 
4217 void
4218 spdk_bs_set_super(struct spdk_blob_store *bs, spdk_blob_id blobid,
4219 		  spdk_bs_op_complete cb_fn, void *cb_arg)
4220 {
4221 	struct spdk_bs_cpl		cpl;
4222 	spdk_bs_sequence_t		*seq;
4223 	struct spdk_bs_set_super_ctx	*ctx;
4224 
4225 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Setting super blob id on blobstore\n");
4226 
4227 	ctx = calloc(1, sizeof(*ctx));
4228 	if (!ctx) {
4229 		cb_fn(cb_arg, -ENOMEM);
4230 		return;
4231 	}
4232 
4233 	ctx->bs = bs;
4234 
4235 	ctx->super = spdk_zmalloc(sizeof(*ctx->super), 0x1000, NULL,
4236 				  SPDK_ENV_SOCKET_ID_ANY, SPDK_MALLOC_DMA);
4237 	if (!ctx->super) {
4238 		free(ctx);
4239 		cb_fn(cb_arg, -ENOMEM);
4240 		return;
4241 	}
4242 
4243 	cpl.type = SPDK_BS_CPL_TYPE_BS_BASIC;
4244 	cpl.u.bs_basic.cb_fn = cb_fn;
4245 	cpl.u.bs_basic.cb_arg = cb_arg;
4246 
4247 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
4248 	if (!seq) {
4249 		spdk_free(ctx->super);
4250 		free(ctx);
4251 		cb_fn(cb_arg, -ENOMEM);
4252 		return;
4253 	}
4254 
4255 	bs->super_blob = blobid;
4256 
4257 	/* Read super block */
4258 	spdk_bs_sequence_read_dev(seq, ctx->super, _spdk_bs_page_to_lba(bs, 0),
4259 				  _spdk_bs_byte_to_lba(bs, sizeof(*ctx->super)),
4260 				  _spdk_bs_set_super_read_cpl, ctx);
4261 }
4262 
4263 /* END spdk_bs_set_super */
4264 
4265 void
4266 spdk_bs_get_super(struct spdk_blob_store *bs,
4267 		  spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4268 {
4269 	if (bs->super_blob == SPDK_BLOBID_INVALID) {
4270 		cb_fn(cb_arg, SPDK_BLOBID_INVALID, -ENOENT);
4271 	} else {
4272 		cb_fn(cb_arg, bs->super_blob, 0);
4273 	}
4274 }
4275 
4276 uint64_t
4277 spdk_bs_get_cluster_size(struct spdk_blob_store *bs)
4278 {
4279 	return bs->cluster_sz;
4280 }
4281 
4282 uint64_t
4283 spdk_bs_get_page_size(struct spdk_blob_store *bs)
4284 {
4285 	return SPDK_BS_PAGE_SIZE;
4286 }
4287 
4288 uint64_t
4289 spdk_bs_get_io_unit_size(struct spdk_blob_store *bs)
4290 {
4291 	return bs->io_unit_size;
4292 }
4293 
4294 uint64_t
4295 spdk_bs_free_cluster_count(struct spdk_blob_store *bs)
4296 {
4297 	return bs->num_free_clusters;
4298 }
4299 
4300 uint64_t
4301 spdk_bs_total_data_cluster_count(struct spdk_blob_store *bs)
4302 {
4303 	return bs->total_data_clusters;
4304 }
4305 
4306 static int
4307 spdk_bs_register_md_thread(struct spdk_blob_store *bs)
4308 {
4309 	bs->md_channel = spdk_get_io_channel(bs);
4310 	if (!bs->md_channel) {
4311 		SPDK_ERRLOG("Failed to get IO channel.\n");
4312 		return -1;
4313 	}
4314 
4315 	return 0;
4316 }
4317 
4318 static int
4319 spdk_bs_unregister_md_thread(struct spdk_blob_store *bs)
4320 {
4321 	spdk_put_io_channel(bs->md_channel);
4322 
4323 	return 0;
4324 }
4325 
4326 spdk_blob_id spdk_blob_get_id(struct spdk_blob *blob)
4327 {
4328 	assert(blob != NULL);
4329 
4330 	return blob->id;
4331 }
4332 
4333 uint64_t spdk_blob_get_num_pages(struct spdk_blob *blob)
4334 {
4335 	assert(blob != NULL);
4336 
4337 	return _spdk_bs_cluster_to_page(blob->bs, blob->active.num_clusters);
4338 }
4339 
4340 uint64_t spdk_blob_get_num_io_units(struct spdk_blob *blob)
4341 {
4342 	assert(blob != NULL);
4343 
4344 	return spdk_blob_get_num_pages(blob) * _spdk_bs_io_unit_per_page(blob->bs);
4345 }
4346 
4347 uint64_t spdk_blob_get_num_clusters(struct spdk_blob *blob)
4348 {
4349 	assert(blob != NULL);
4350 
4351 	return blob->active.num_clusters;
4352 }
4353 
4354 /* START spdk_bs_create_blob */
4355 
4356 static void
4357 _spdk_bs_create_blob_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
4358 {
4359 	struct spdk_blob *blob = cb_arg;
4360 
4361 	_spdk_blob_free(blob);
4362 
4363 	spdk_bs_sequence_finish(seq, bserrno);
4364 }
4365 
4366 static int
4367 _spdk_blob_set_xattrs(struct spdk_blob *blob, const struct spdk_blob_xattr_opts *xattrs,
4368 		      bool internal)
4369 {
4370 	uint64_t i;
4371 	size_t value_len = 0;
4372 	int rc;
4373 	const void *value = NULL;
4374 	if (xattrs->count > 0 && xattrs->get_value == NULL) {
4375 		return -EINVAL;
4376 	}
4377 	for (i = 0; i < xattrs->count; i++) {
4378 		xattrs->get_value(xattrs->ctx, xattrs->names[i], &value, &value_len);
4379 		if (value == NULL || value_len == 0) {
4380 			return -EINVAL;
4381 		}
4382 		rc = _spdk_blob_set_xattr(blob, xattrs->names[i], value, value_len, internal);
4383 		if (rc < 0) {
4384 			return rc;
4385 		}
4386 	}
4387 	return 0;
4388 }
4389 
4390 static void
4391 _spdk_bs_create_blob(struct spdk_blob_store *bs,
4392 		     const struct spdk_blob_opts *opts,
4393 		     const struct spdk_blob_xattr_opts *internal_xattrs,
4394 		     spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4395 {
4396 	struct spdk_blob	*blob;
4397 	uint32_t		page_idx;
4398 	struct spdk_bs_cpl	cpl;
4399 	struct spdk_blob_opts	opts_default;
4400 	struct spdk_blob_xattr_opts internal_xattrs_default;
4401 	spdk_bs_sequence_t	*seq;
4402 	spdk_blob_id		id;
4403 	int rc;
4404 
4405 	assert(spdk_get_thread() == bs->md_thread);
4406 
4407 	page_idx = spdk_bit_array_find_first_clear(bs->used_md_pages, 0);
4408 	if (page_idx == UINT32_MAX) {
4409 		cb_fn(cb_arg, 0, -ENOMEM);
4410 		return;
4411 	}
4412 	spdk_bit_array_set(bs->used_blobids, page_idx);
4413 	spdk_bit_array_set(bs->used_md_pages, page_idx);
4414 
4415 	id = _spdk_bs_page_to_blobid(page_idx);
4416 
4417 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Creating blob with id %lu at page %u\n", id, page_idx);
4418 
4419 	blob = _spdk_blob_alloc(bs, id);
4420 	if (!blob) {
4421 		cb_fn(cb_arg, 0, -ENOMEM);
4422 		return;
4423 	}
4424 
4425 	if (!opts) {
4426 		spdk_blob_opts_init(&opts_default);
4427 		opts = &opts_default;
4428 	}
4429 
4430 	blob->use_extent_table = opts->use_extent_table;
4431 
4432 	if (!internal_xattrs) {
4433 		_spdk_blob_xattrs_init(&internal_xattrs_default);
4434 		internal_xattrs = &internal_xattrs_default;
4435 	}
4436 
4437 	rc = _spdk_blob_set_xattrs(blob, &opts->xattrs, false);
4438 	if (rc < 0) {
4439 		_spdk_blob_free(blob);
4440 		cb_fn(cb_arg, 0, rc);
4441 		return;
4442 	}
4443 
4444 	rc = _spdk_blob_set_xattrs(blob, internal_xattrs, true);
4445 	if (rc < 0) {
4446 		_spdk_blob_free(blob);
4447 		cb_fn(cb_arg, 0, rc);
4448 		return;
4449 	}
4450 
4451 	if (opts->thin_provision) {
4452 		_spdk_blob_set_thin_provision(blob);
4453 	}
4454 
4455 	_spdk_blob_set_clear_method(blob, opts->clear_method);
4456 
4457 	rc = _spdk_blob_resize(blob, opts->num_clusters);
4458 	if (rc < 0) {
4459 		_spdk_blob_free(blob);
4460 		cb_fn(cb_arg, 0, rc);
4461 		return;
4462 	}
4463 	cpl.type = SPDK_BS_CPL_TYPE_BLOBID;
4464 	cpl.u.blobid.cb_fn = cb_fn;
4465 	cpl.u.blobid.cb_arg = cb_arg;
4466 	cpl.u.blobid.blobid = blob->id;
4467 
4468 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
4469 	if (!seq) {
4470 		_spdk_blob_free(blob);
4471 		cb_fn(cb_arg, 0, -ENOMEM);
4472 		return;
4473 	}
4474 
4475 	_spdk_blob_persist(seq, blob, _spdk_bs_create_blob_cpl, blob);
4476 }
4477 
4478 void spdk_bs_create_blob(struct spdk_blob_store *bs,
4479 			 spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4480 {
4481 	_spdk_bs_create_blob(bs, NULL, NULL, cb_fn, cb_arg);
4482 }
4483 
4484 void spdk_bs_create_blob_ext(struct spdk_blob_store *bs, const struct spdk_blob_opts *opts,
4485 			     spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4486 {
4487 	_spdk_bs_create_blob(bs, opts, NULL, cb_fn, cb_arg);
4488 }
4489 
4490 /* END spdk_bs_create_blob */
4491 
4492 /* START blob_cleanup */
4493 
4494 struct spdk_clone_snapshot_ctx {
4495 	struct spdk_bs_cpl      cpl;
4496 	int bserrno;
4497 	bool frozen;
4498 
4499 	struct spdk_io_channel *channel;
4500 
4501 	/* Current cluster for inflate operation */
4502 	uint64_t cluster;
4503 
4504 	/* For inflation force allocation of all unallocated clusters and remove
4505 	 * thin-provisioning. Otherwise only decouple parent and keep clone thin. */
4506 	bool allocate_all;
4507 
4508 	struct {
4509 		spdk_blob_id id;
4510 		struct spdk_blob *blob;
4511 	} original;
4512 	struct {
4513 		spdk_blob_id id;
4514 		struct spdk_blob *blob;
4515 	} new;
4516 
4517 	/* xattrs specified for snapshot/clones only. They have no impact on
4518 	 * the original blobs xattrs. */
4519 	const struct spdk_blob_xattr_opts *xattrs;
4520 };
4521 
4522 static void
4523 _spdk_bs_clone_snapshot_cleanup_finish(void *cb_arg, int bserrno)
4524 {
4525 	struct spdk_clone_snapshot_ctx *ctx = cb_arg;
4526 	struct spdk_bs_cpl *cpl = &ctx->cpl;
4527 
4528 	if (bserrno != 0) {
4529 		if (ctx->bserrno != 0) {
4530 			SPDK_ERRLOG("Cleanup error %d\n", bserrno);
4531 		} else {
4532 			ctx->bserrno = bserrno;
4533 		}
4534 	}
4535 
4536 	switch (cpl->type) {
4537 	case SPDK_BS_CPL_TYPE_BLOBID:
4538 		cpl->u.blobid.cb_fn(cpl->u.blobid.cb_arg, cpl->u.blobid.blobid, ctx->bserrno);
4539 		break;
4540 	case SPDK_BS_CPL_TYPE_BLOB_BASIC:
4541 		cpl->u.blob_basic.cb_fn(cpl->u.blob_basic.cb_arg, ctx->bserrno);
4542 		break;
4543 	default:
4544 		SPDK_UNREACHABLE();
4545 		break;
4546 	}
4547 
4548 	free(ctx);
4549 }
4550 
4551 static void
4552 _spdk_bs_snapshot_unfreeze_cpl(void *cb_arg, int bserrno)
4553 {
4554 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4555 	struct spdk_blob *origblob = ctx->original.blob;
4556 
4557 	if (bserrno != 0) {
4558 		if (ctx->bserrno != 0) {
4559 			SPDK_ERRLOG("Unfreeze error %d\n", bserrno);
4560 		} else {
4561 			ctx->bserrno = bserrno;
4562 		}
4563 	}
4564 
4565 	ctx->original.id = origblob->id;
4566 	origblob->locked_operation_in_progress = false;
4567 
4568 	spdk_blob_close(origblob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4569 }
4570 
4571 static void
4572 _spdk_bs_clone_snapshot_origblob_cleanup(void *cb_arg, int bserrno)
4573 {
4574 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4575 	struct spdk_blob *origblob = ctx->original.blob;
4576 
4577 	if (bserrno != 0) {
4578 		if (ctx->bserrno != 0) {
4579 			SPDK_ERRLOG("Cleanup error %d\n", bserrno);
4580 		} else {
4581 			ctx->bserrno = bserrno;
4582 		}
4583 	}
4584 
4585 	if (ctx->frozen) {
4586 		/* Unfreeze any outstanding I/O */
4587 		_spdk_blob_unfreeze_io(origblob, _spdk_bs_snapshot_unfreeze_cpl, ctx);
4588 	} else {
4589 		_spdk_bs_snapshot_unfreeze_cpl(ctx, 0);
4590 	}
4591 
4592 }
4593 
4594 static void
4595 _spdk_bs_clone_snapshot_newblob_cleanup(void *cb_arg, int bserrno)
4596 {
4597 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4598 	struct spdk_blob *newblob = ctx->new.blob;
4599 
4600 	if (bserrno != 0) {
4601 		if (ctx->bserrno != 0) {
4602 			SPDK_ERRLOG("Cleanup error %d\n", bserrno);
4603 		} else {
4604 			ctx->bserrno = bserrno;
4605 		}
4606 	}
4607 
4608 	ctx->new.id = newblob->id;
4609 	spdk_blob_close(newblob, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
4610 }
4611 
4612 /* END blob_cleanup */
4613 
4614 /* START spdk_bs_create_snapshot */
4615 
4616 static void
4617 _spdk_bs_snapshot_swap_cluster_maps(struct spdk_blob *blob1, struct spdk_blob *blob2)
4618 {
4619 	uint64_t *cluster_temp;
4620 
4621 	cluster_temp = blob1->active.clusters;
4622 	blob1->active.clusters = blob2->active.clusters;
4623 	blob2->active.clusters = cluster_temp;
4624 }
4625 
4626 static void
4627 _spdk_bs_snapshot_origblob_sync_cpl(void *cb_arg, int bserrno)
4628 {
4629 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4630 	struct spdk_blob *origblob = ctx->original.blob;
4631 	struct spdk_blob *newblob = ctx->new.blob;
4632 
4633 	if (bserrno != 0) {
4634 		_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4635 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4636 		return;
4637 	}
4638 
4639 	/* Remove metadata descriptor SNAPSHOT_IN_PROGRESS */
4640 	bserrno = _spdk_blob_remove_xattr(newblob, SNAPSHOT_IN_PROGRESS, true);
4641 	if (bserrno != 0) {
4642 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4643 		return;
4644 	}
4645 
4646 	_spdk_bs_blob_list_add(ctx->original.blob);
4647 
4648 	spdk_blob_set_read_only(newblob);
4649 
4650 	/* sync snapshot metadata */
4651 	spdk_blob_sync_md(newblob, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
4652 }
4653 
4654 static void
4655 _spdk_bs_snapshot_newblob_sync_cpl(void *cb_arg, int bserrno)
4656 {
4657 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4658 	struct spdk_blob *origblob = ctx->original.blob;
4659 	struct spdk_blob *newblob = ctx->new.blob;
4660 
4661 	if (bserrno != 0) {
4662 		/* return cluster map back to original */
4663 		_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4664 		_spdk_bs_clone_snapshot_newblob_cleanup(ctx, bserrno);
4665 		return;
4666 	}
4667 
4668 	/* Set internal xattr for snapshot id */
4669 	bserrno = _spdk_blob_set_xattr(origblob, BLOB_SNAPSHOT, &newblob->id, sizeof(spdk_blob_id), true);
4670 	if (bserrno != 0) {
4671 		/* return cluster map back to original */
4672 		_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4673 		_spdk_bs_clone_snapshot_newblob_cleanup(ctx, bserrno);
4674 		return;
4675 	}
4676 
4677 	_spdk_bs_blob_list_remove(origblob);
4678 	origblob->parent_id = newblob->id;
4679 
4680 	/* Create new back_bs_dev for snapshot */
4681 	origblob->back_bs_dev = spdk_bs_create_blob_bs_dev(newblob);
4682 	if (origblob->back_bs_dev == NULL) {
4683 		/* return cluster map back to original */
4684 		_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4685 		_spdk_bs_clone_snapshot_newblob_cleanup(ctx, -EINVAL);
4686 		return;
4687 	}
4688 
4689 	/* set clone blob as thin provisioned */
4690 	_spdk_blob_set_thin_provision(origblob);
4691 
4692 	_spdk_bs_blob_list_add(newblob);
4693 
4694 	/* sync clone metadata */
4695 	spdk_blob_sync_md(origblob, _spdk_bs_snapshot_origblob_sync_cpl, ctx);
4696 }
4697 
4698 static void
4699 _spdk_bs_snapshot_freeze_cpl(void *cb_arg, int rc)
4700 {
4701 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4702 	struct spdk_blob *origblob = ctx->original.blob;
4703 	struct spdk_blob *newblob = ctx->new.blob;
4704 	int bserrno;
4705 
4706 	if (rc != 0) {
4707 		_spdk_bs_clone_snapshot_newblob_cleanup(ctx, rc);
4708 		return;
4709 	}
4710 
4711 	ctx->frozen = true;
4712 
4713 	/* set new back_bs_dev for snapshot */
4714 	newblob->back_bs_dev = origblob->back_bs_dev;
4715 	/* Set invalid flags from origblob */
4716 	newblob->invalid_flags = origblob->invalid_flags;
4717 
4718 	/* inherit parent from original blob if set */
4719 	newblob->parent_id = origblob->parent_id;
4720 	if (origblob->parent_id != SPDK_BLOBID_INVALID) {
4721 		/* Set internal xattr for snapshot id */
4722 		bserrno = _spdk_blob_set_xattr(newblob, BLOB_SNAPSHOT,
4723 					       &origblob->parent_id, sizeof(spdk_blob_id), true);
4724 		if (bserrno != 0) {
4725 			_spdk_bs_clone_snapshot_newblob_cleanup(ctx, bserrno);
4726 			return;
4727 		}
4728 	}
4729 
4730 	/* swap cluster maps */
4731 	_spdk_bs_snapshot_swap_cluster_maps(newblob, origblob);
4732 
4733 	/* Set the clear method on the new blob to match the original. */
4734 	_spdk_blob_set_clear_method(newblob, origblob->clear_method);
4735 
4736 	/* sync snapshot metadata */
4737 	spdk_blob_sync_md(newblob, _spdk_bs_snapshot_newblob_sync_cpl, ctx);
4738 }
4739 
4740 static void
4741 _spdk_bs_snapshot_newblob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
4742 {
4743 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4744 	struct spdk_blob *origblob = ctx->original.blob;
4745 	struct spdk_blob *newblob = _blob;
4746 
4747 	if (bserrno != 0) {
4748 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4749 		return;
4750 	}
4751 
4752 	ctx->new.blob = newblob;
4753 	assert(spdk_blob_is_thin_provisioned(newblob));
4754 	assert(spdk_mem_all_zero(newblob->active.clusters,
4755 				 newblob->active.num_clusters * sizeof(*newblob->active.clusters)));
4756 
4757 	_spdk_blob_freeze_io(origblob, _spdk_bs_snapshot_freeze_cpl, ctx);
4758 }
4759 
4760 static void
4761 _spdk_bs_snapshot_newblob_create_cpl(void *cb_arg, spdk_blob_id blobid, int bserrno)
4762 {
4763 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4764 	struct spdk_blob *origblob = ctx->original.blob;
4765 
4766 	if (bserrno != 0) {
4767 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4768 		return;
4769 	}
4770 
4771 	ctx->new.id = blobid;
4772 	ctx->cpl.u.blobid.blobid = blobid;
4773 
4774 	spdk_bs_open_blob(origblob->bs, ctx->new.id, _spdk_bs_snapshot_newblob_open_cpl, ctx);
4775 }
4776 
4777 
4778 static void
4779 _spdk_bs_xattr_snapshot(void *arg, const char *name,
4780 			const void **value, size_t *value_len)
4781 {
4782 	assert(strncmp(name, SNAPSHOT_IN_PROGRESS, sizeof(SNAPSHOT_IN_PROGRESS)) == 0);
4783 
4784 	struct spdk_blob *blob = (struct spdk_blob *)arg;
4785 	*value = &blob->id;
4786 	*value_len = sizeof(blob->id);
4787 }
4788 
4789 static void
4790 _spdk_bs_snapshot_origblob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
4791 {
4792 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4793 	struct spdk_blob_opts opts;
4794 	struct spdk_blob_xattr_opts internal_xattrs;
4795 	char *xattrs_names[] = { SNAPSHOT_IN_PROGRESS };
4796 
4797 	if (bserrno != 0) {
4798 		_spdk_bs_clone_snapshot_cleanup_finish(ctx, bserrno);
4799 		return;
4800 	}
4801 
4802 	ctx->original.blob = _blob;
4803 
4804 	if (_blob->data_ro || _blob->md_ro) {
4805 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot create snapshot from read only blob with id %lu\n",
4806 			      _blob->id);
4807 		ctx->bserrno = -EINVAL;
4808 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4809 		return;
4810 	}
4811 
4812 	if (_blob->locked_operation_in_progress) {
4813 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot create snapshot - another operation in progress\n");
4814 		ctx->bserrno = -EBUSY;
4815 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4816 		return;
4817 	}
4818 
4819 	_blob->locked_operation_in_progress = true;
4820 
4821 	spdk_blob_opts_init(&opts);
4822 	_spdk_blob_xattrs_init(&internal_xattrs);
4823 
4824 	/* Change the size of new blob to the same as in original blob,
4825 	 * but do not allocate clusters */
4826 	opts.thin_provision = true;
4827 	opts.num_clusters = spdk_blob_get_num_clusters(_blob);
4828 	opts.use_extent_table = _blob->use_extent_table;
4829 
4830 	/* If there are any xattrs specified for snapshot, set them now */
4831 	if (ctx->xattrs) {
4832 		memcpy(&opts.xattrs, ctx->xattrs, sizeof(*ctx->xattrs));
4833 	}
4834 	/* Set internal xattr SNAPSHOT_IN_PROGRESS */
4835 	internal_xattrs.count = 1;
4836 	internal_xattrs.ctx = _blob;
4837 	internal_xattrs.names = xattrs_names;
4838 	internal_xattrs.get_value = _spdk_bs_xattr_snapshot;
4839 
4840 	_spdk_bs_create_blob(_blob->bs, &opts, &internal_xattrs,
4841 			     _spdk_bs_snapshot_newblob_create_cpl, ctx);
4842 }
4843 
4844 void spdk_bs_create_snapshot(struct spdk_blob_store *bs, spdk_blob_id blobid,
4845 			     const struct spdk_blob_xattr_opts *snapshot_xattrs,
4846 			     spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4847 {
4848 	struct spdk_clone_snapshot_ctx *ctx = calloc(1, sizeof(*ctx));
4849 
4850 	if (!ctx) {
4851 		cb_fn(cb_arg, SPDK_BLOBID_INVALID, -ENOMEM);
4852 		return;
4853 	}
4854 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BLOBID;
4855 	ctx->cpl.u.blobid.cb_fn = cb_fn;
4856 	ctx->cpl.u.blobid.cb_arg = cb_arg;
4857 	ctx->cpl.u.blobid.blobid = SPDK_BLOBID_INVALID;
4858 	ctx->bserrno = 0;
4859 	ctx->frozen = false;
4860 	ctx->original.id = blobid;
4861 	ctx->xattrs = snapshot_xattrs;
4862 
4863 	spdk_bs_open_blob(bs, ctx->original.id, _spdk_bs_snapshot_origblob_open_cpl, ctx);
4864 }
4865 /* END spdk_bs_create_snapshot */
4866 
4867 /* START spdk_bs_create_clone */
4868 
4869 static void
4870 _spdk_bs_xattr_clone(void *arg, const char *name,
4871 		     const void **value, size_t *value_len)
4872 {
4873 	assert(strncmp(name, BLOB_SNAPSHOT, sizeof(BLOB_SNAPSHOT)) == 0);
4874 
4875 	struct spdk_blob *blob = (struct spdk_blob *)arg;
4876 	*value = &blob->id;
4877 	*value_len = sizeof(blob->id);
4878 }
4879 
4880 static void
4881 _spdk_bs_clone_newblob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
4882 {
4883 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4884 	struct spdk_blob *clone = _blob;
4885 
4886 	ctx->new.blob = clone;
4887 	_spdk_bs_blob_list_add(clone);
4888 
4889 	spdk_blob_close(clone, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
4890 }
4891 
4892 static void
4893 _spdk_bs_clone_newblob_create_cpl(void *cb_arg, spdk_blob_id blobid, int bserrno)
4894 {
4895 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4896 
4897 	ctx->cpl.u.blobid.blobid = blobid;
4898 	spdk_bs_open_blob(ctx->original.blob->bs, blobid, _spdk_bs_clone_newblob_open_cpl, ctx);
4899 }
4900 
4901 static void
4902 _spdk_bs_clone_origblob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
4903 {
4904 	struct spdk_clone_snapshot_ctx	*ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4905 	struct spdk_blob_opts		opts;
4906 	struct spdk_blob_xattr_opts internal_xattrs;
4907 	char *xattr_names[] = { BLOB_SNAPSHOT };
4908 
4909 	if (bserrno != 0) {
4910 		_spdk_bs_clone_snapshot_cleanup_finish(ctx, bserrno);
4911 		return;
4912 	}
4913 
4914 	ctx->original.blob = _blob;
4915 
4916 	if (!_blob->data_ro || !_blob->md_ro) {
4917 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Clone not from read-only blob\n");
4918 		ctx->bserrno = -EINVAL;
4919 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4920 		return;
4921 	}
4922 
4923 	if (_blob->locked_operation_in_progress) {
4924 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot create clone - another operation in progress\n");
4925 		ctx->bserrno = -EBUSY;
4926 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
4927 		return;
4928 	}
4929 
4930 	_blob->locked_operation_in_progress = true;
4931 
4932 	spdk_blob_opts_init(&opts);
4933 	_spdk_blob_xattrs_init(&internal_xattrs);
4934 
4935 	opts.thin_provision = true;
4936 	opts.num_clusters = spdk_blob_get_num_clusters(_blob);
4937 	opts.use_extent_table = _blob->use_extent_table;
4938 	if (ctx->xattrs) {
4939 		memcpy(&opts.xattrs, ctx->xattrs, sizeof(*ctx->xattrs));
4940 	}
4941 
4942 	/* Set internal xattr BLOB_SNAPSHOT */
4943 	internal_xattrs.count = 1;
4944 	internal_xattrs.ctx = _blob;
4945 	internal_xattrs.names = xattr_names;
4946 	internal_xattrs.get_value = _spdk_bs_xattr_clone;
4947 
4948 	_spdk_bs_create_blob(_blob->bs, &opts, &internal_xattrs,
4949 			     _spdk_bs_clone_newblob_create_cpl, ctx);
4950 }
4951 
4952 void spdk_bs_create_clone(struct spdk_blob_store *bs, spdk_blob_id blobid,
4953 			  const struct spdk_blob_xattr_opts *clone_xattrs,
4954 			  spdk_blob_op_with_id_complete cb_fn, void *cb_arg)
4955 {
4956 	struct spdk_clone_snapshot_ctx	*ctx = calloc(1, sizeof(*ctx));
4957 
4958 	if (!ctx) {
4959 		cb_fn(cb_arg, SPDK_BLOBID_INVALID, -ENOMEM);
4960 		return;
4961 	}
4962 
4963 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BLOBID;
4964 	ctx->cpl.u.blobid.cb_fn = cb_fn;
4965 	ctx->cpl.u.blobid.cb_arg = cb_arg;
4966 	ctx->cpl.u.blobid.blobid = SPDK_BLOBID_INVALID;
4967 	ctx->bserrno = 0;
4968 	ctx->xattrs = clone_xattrs;
4969 	ctx->original.id = blobid;
4970 
4971 	spdk_bs_open_blob(bs, ctx->original.id, _spdk_bs_clone_origblob_open_cpl, ctx);
4972 }
4973 
4974 /* END spdk_bs_create_clone */
4975 
4976 /* START spdk_bs_inflate_blob */
4977 
4978 static void
4979 _spdk_bs_inflate_blob_set_parent_cpl(void *cb_arg, struct spdk_blob *_parent, int bserrno)
4980 {
4981 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
4982 	struct spdk_blob *_blob = ctx->original.blob;
4983 
4984 	if (bserrno != 0) {
4985 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
4986 		return;
4987 	}
4988 
4989 	assert(_parent != NULL);
4990 
4991 	_spdk_bs_blob_list_remove(_blob);
4992 	_blob->parent_id = _parent->id;
4993 	_spdk_blob_set_xattr(_blob, BLOB_SNAPSHOT, &_blob->parent_id,
4994 			     sizeof(spdk_blob_id), true);
4995 
4996 	_blob->back_bs_dev->destroy(_blob->back_bs_dev);
4997 	_blob->back_bs_dev = spdk_bs_create_blob_bs_dev(_parent);
4998 	_spdk_bs_blob_list_add(_blob);
4999 
5000 	spdk_blob_sync_md(_blob, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
5001 }
5002 
5003 static void
5004 _spdk_bs_inflate_blob_done(void *cb_arg, int bserrno)
5005 {
5006 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
5007 	struct spdk_blob *_blob = ctx->original.blob;
5008 	struct spdk_blob *_parent;
5009 
5010 	if (bserrno != 0) {
5011 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
5012 		return;
5013 	}
5014 
5015 	if (ctx->allocate_all) {
5016 		/* remove thin provisioning */
5017 		_spdk_bs_blob_list_remove(_blob);
5018 		_spdk_blob_remove_xattr(_blob, BLOB_SNAPSHOT, true);
5019 		_blob->invalid_flags = _blob->invalid_flags & ~SPDK_BLOB_THIN_PROV;
5020 		_blob->back_bs_dev->destroy(_blob->back_bs_dev);
5021 		_blob->back_bs_dev = NULL;
5022 		_blob->parent_id = SPDK_BLOBID_INVALID;
5023 	} else {
5024 		_parent = ((struct spdk_blob_bs_dev *)(_blob->back_bs_dev))->blob;
5025 		if (_parent->parent_id != SPDK_BLOBID_INVALID) {
5026 			/* We must change the parent of the inflated blob */
5027 			spdk_bs_open_blob(_blob->bs, _parent->parent_id,
5028 					  _spdk_bs_inflate_blob_set_parent_cpl, ctx);
5029 			return;
5030 		}
5031 
5032 		_spdk_bs_blob_list_remove(_blob);
5033 		_spdk_blob_remove_xattr(_blob, BLOB_SNAPSHOT, true);
5034 		_blob->parent_id = SPDK_BLOBID_INVALID;
5035 		_blob->back_bs_dev->destroy(_blob->back_bs_dev);
5036 		_blob->back_bs_dev = spdk_bs_create_zeroes_dev();
5037 	}
5038 
5039 	_blob->state = SPDK_BLOB_STATE_DIRTY;
5040 	spdk_blob_sync_md(_blob, _spdk_bs_clone_snapshot_origblob_cleanup, ctx);
5041 }
5042 
5043 /* Check if cluster needs allocation */
5044 static inline bool
5045 _spdk_bs_cluster_needs_allocation(struct spdk_blob *blob, uint64_t cluster, bool allocate_all)
5046 {
5047 	struct spdk_blob_bs_dev *b;
5048 
5049 	assert(blob != NULL);
5050 
5051 	if (blob->active.clusters[cluster] != 0) {
5052 		/* Cluster is already allocated */
5053 		return false;
5054 	}
5055 
5056 	if (blob->parent_id == SPDK_BLOBID_INVALID) {
5057 		/* Blob have no parent blob */
5058 		return allocate_all;
5059 	}
5060 
5061 	b = (struct spdk_blob_bs_dev *)blob->back_bs_dev;
5062 	return (allocate_all || b->blob->active.clusters[cluster] != 0);
5063 }
5064 
5065 static void
5066 _spdk_bs_inflate_blob_touch_next(void *cb_arg, int bserrno)
5067 {
5068 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
5069 	struct spdk_blob *_blob = ctx->original.blob;
5070 	uint64_t offset;
5071 
5072 	if (bserrno != 0) {
5073 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, bserrno);
5074 		return;
5075 	}
5076 
5077 	for (; ctx->cluster < _blob->active.num_clusters; ctx->cluster++) {
5078 		if (_spdk_bs_cluster_needs_allocation(_blob, ctx->cluster, ctx->allocate_all)) {
5079 			break;
5080 		}
5081 	}
5082 
5083 	if (ctx->cluster < _blob->active.num_clusters) {
5084 		offset = _spdk_bs_cluster_to_lba(_blob->bs, ctx->cluster);
5085 
5086 		/* We may safely increment a cluster before write */
5087 		ctx->cluster++;
5088 
5089 		/* Use zero length write to touch a cluster */
5090 		spdk_blob_io_write(_blob, ctx->channel, NULL, offset, 0,
5091 				   _spdk_bs_inflate_blob_touch_next, ctx);
5092 	} else {
5093 		_spdk_bs_inflate_blob_done(cb_arg, bserrno);
5094 	}
5095 }
5096 
5097 static void
5098 _spdk_bs_inflate_blob_open_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
5099 {
5100 	struct spdk_clone_snapshot_ctx *ctx = (struct spdk_clone_snapshot_ctx *)cb_arg;
5101 	uint64_t lfc; /* lowest free cluster */
5102 	uint64_t i;
5103 
5104 	if (bserrno != 0) {
5105 		_spdk_bs_clone_snapshot_cleanup_finish(ctx, bserrno);
5106 		return;
5107 	}
5108 
5109 	ctx->original.blob = _blob;
5110 
5111 	if (_blob->locked_operation_in_progress) {
5112 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot inflate blob - another operation in progress\n");
5113 		ctx->bserrno = -EBUSY;
5114 		spdk_blob_close(_blob, _spdk_bs_clone_snapshot_cleanup_finish, ctx);
5115 		return;
5116 	}
5117 
5118 	_blob->locked_operation_in_progress = true;
5119 
5120 	if (!ctx->allocate_all && _blob->parent_id == SPDK_BLOBID_INVALID) {
5121 		/* This blob have no parent, so we cannot decouple it. */
5122 		SPDK_ERRLOG("Cannot decouple parent of blob with no parent.\n");
5123 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, -EINVAL);
5124 		return;
5125 	}
5126 
5127 	if (spdk_blob_is_thin_provisioned(_blob) == false) {
5128 		/* This is not thin provisioned blob. No need to inflate. */
5129 		_spdk_bs_clone_snapshot_origblob_cleanup(ctx, 0);
5130 		return;
5131 	}
5132 
5133 	/* Do two passes - one to verify that we can obtain enough clusters
5134 	 * and another to actually claim them.
5135 	 */
5136 	lfc = 0;
5137 	for (i = 0; i < _blob->active.num_clusters; i++) {
5138 		if (_spdk_bs_cluster_needs_allocation(_blob, i, ctx->allocate_all)) {
5139 			lfc = spdk_bit_array_find_first_clear(_blob->bs->used_clusters, lfc);
5140 			if (lfc == UINT32_MAX) {
5141 				/* No more free clusters. Cannot satisfy the request */
5142 				_spdk_bs_clone_snapshot_origblob_cleanup(ctx, -ENOSPC);
5143 				return;
5144 			}
5145 			lfc++;
5146 		}
5147 	}
5148 
5149 	ctx->cluster = 0;
5150 	_spdk_bs_inflate_blob_touch_next(ctx, 0);
5151 }
5152 
5153 static void
5154 _spdk_bs_inflate_blob(struct spdk_blob_store *bs, struct spdk_io_channel *channel,
5155 		      spdk_blob_id blobid, bool allocate_all, spdk_blob_op_complete cb_fn, void *cb_arg)
5156 {
5157 	struct spdk_clone_snapshot_ctx *ctx = calloc(1, sizeof(*ctx));
5158 
5159 	if (!ctx) {
5160 		cb_fn(cb_arg, -ENOMEM);
5161 		return;
5162 	}
5163 	ctx->cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
5164 	ctx->cpl.u.bs_basic.cb_fn = cb_fn;
5165 	ctx->cpl.u.bs_basic.cb_arg = cb_arg;
5166 	ctx->bserrno = 0;
5167 	ctx->original.id = blobid;
5168 	ctx->channel = channel;
5169 	ctx->allocate_all = allocate_all;
5170 
5171 	spdk_bs_open_blob(bs, ctx->original.id, _spdk_bs_inflate_blob_open_cpl, ctx);
5172 }
5173 
5174 void
5175 spdk_bs_inflate_blob(struct spdk_blob_store *bs, struct spdk_io_channel *channel,
5176 		     spdk_blob_id blobid, spdk_blob_op_complete cb_fn, void *cb_arg)
5177 {
5178 	_spdk_bs_inflate_blob(bs, channel, blobid, true, cb_fn, cb_arg);
5179 }
5180 
5181 void
5182 spdk_bs_blob_decouple_parent(struct spdk_blob_store *bs, struct spdk_io_channel *channel,
5183 			     spdk_blob_id blobid, spdk_blob_op_complete cb_fn, void *cb_arg)
5184 {
5185 	_spdk_bs_inflate_blob(bs, channel, blobid, false, cb_fn, cb_arg);
5186 }
5187 /* END spdk_bs_inflate_blob */
5188 
5189 /* START spdk_blob_resize */
5190 struct spdk_bs_resize_ctx {
5191 	spdk_blob_op_complete cb_fn;
5192 	void *cb_arg;
5193 	struct spdk_blob *blob;
5194 	uint64_t sz;
5195 	int rc;
5196 };
5197 
5198 static void
5199 _spdk_bs_resize_unfreeze_cpl(void *cb_arg, int rc)
5200 {
5201 	struct spdk_bs_resize_ctx *ctx = (struct spdk_bs_resize_ctx *)cb_arg;
5202 
5203 	if (rc != 0) {
5204 		SPDK_ERRLOG("Unfreeze failed, rc=%d\n", rc);
5205 	}
5206 
5207 	if (ctx->rc != 0) {
5208 		SPDK_ERRLOG("Unfreeze failed, ctx->rc=%d\n", ctx->rc);
5209 		rc = ctx->rc;
5210 	}
5211 
5212 	ctx->blob->locked_operation_in_progress = false;
5213 
5214 	ctx->cb_fn(ctx->cb_arg, rc);
5215 	free(ctx);
5216 }
5217 
5218 static void
5219 _spdk_bs_resize_freeze_cpl(void *cb_arg, int rc)
5220 {
5221 	struct spdk_bs_resize_ctx *ctx = (struct spdk_bs_resize_ctx *)cb_arg;
5222 
5223 	if (rc != 0) {
5224 		ctx->blob->locked_operation_in_progress = false;
5225 		ctx->cb_fn(ctx->cb_arg, rc);
5226 		free(ctx);
5227 		return;
5228 	}
5229 
5230 	ctx->rc = _spdk_blob_resize(ctx->blob, ctx->sz);
5231 
5232 	_spdk_blob_unfreeze_io(ctx->blob, _spdk_bs_resize_unfreeze_cpl, ctx);
5233 }
5234 
5235 void
5236 spdk_blob_resize(struct spdk_blob *blob, uint64_t sz, spdk_blob_op_complete cb_fn, void *cb_arg)
5237 {
5238 	struct spdk_bs_resize_ctx *ctx;
5239 
5240 	_spdk_blob_verify_md_op(blob);
5241 
5242 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Resizing blob %lu to %lu clusters\n", blob->id, sz);
5243 
5244 	if (blob->md_ro) {
5245 		cb_fn(cb_arg, -EPERM);
5246 		return;
5247 	}
5248 
5249 	if (sz == blob->active.num_clusters) {
5250 		cb_fn(cb_arg, 0);
5251 		return;
5252 	}
5253 
5254 	if (blob->locked_operation_in_progress) {
5255 		cb_fn(cb_arg, -EBUSY);
5256 		return;
5257 	}
5258 
5259 	ctx = calloc(1, sizeof(*ctx));
5260 	if (!ctx) {
5261 		cb_fn(cb_arg, -ENOMEM);
5262 		return;
5263 	}
5264 
5265 	blob->locked_operation_in_progress = true;
5266 	ctx->cb_fn = cb_fn;
5267 	ctx->cb_arg = cb_arg;
5268 	ctx->blob = blob;
5269 	ctx->sz = sz;
5270 	_spdk_blob_freeze_io(blob, _spdk_bs_resize_freeze_cpl, ctx);
5271 }
5272 
5273 /* END spdk_blob_resize */
5274 
5275 
5276 /* START spdk_bs_delete_blob */
5277 
5278 static void
5279 _spdk_bs_delete_close_cpl(void *cb_arg, int bserrno)
5280 {
5281 	spdk_bs_sequence_t *seq = cb_arg;
5282 
5283 	spdk_bs_sequence_finish(seq, bserrno);
5284 }
5285 
5286 static void
5287 _spdk_bs_delete_persist_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
5288 {
5289 	struct spdk_blob *blob = cb_arg;
5290 
5291 	if (bserrno != 0) {
5292 		/*
5293 		 * We already removed this blob from the blobstore tailq, so
5294 		 *  we need to free it here since this is the last reference
5295 		 *  to it.
5296 		 */
5297 		_spdk_blob_free(blob);
5298 		_spdk_bs_delete_close_cpl(seq, bserrno);
5299 		return;
5300 	}
5301 
5302 	/*
5303 	 * This will immediately decrement the ref_count and call
5304 	 *  the completion routine since the metadata state is clean.
5305 	 *  By calling spdk_blob_close, we reduce the number of call
5306 	 *  points into code that touches the blob->open_ref count
5307 	 *  and the blobstore's blob list.
5308 	 */
5309 	spdk_blob_close(blob, _spdk_bs_delete_close_cpl, seq);
5310 }
5311 
5312 struct delete_snapshot_ctx {
5313 	struct spdk_blob_list *parent_snapshot_entry;
5314 	struct spdk_blob *snapshot;
5315 	bool snapshot_md_ro;
5316 	struct spdk_blob *clone;
5317 	bool clone_md_ro;
5318 	spdk_blob_op_with_handle_complete cb_fn;
5319 	void *cb_arg;
5320 	int bserrno;
5321 };
5322 
5323 static void
5324 _spdk_delete_blob_cleanup_finish(void *cb_arg, int bserrno)
5325 {
5326 	struct delete_snapshot_ctx *ctx = cb_arg;
5327 
5328 	if (bserrno != 0) {
5329 		SPDK_ERRLOG("Snapshot cleanup error %d\n", bserrno);
5330 	}
5331 
5332 	assert(ctx != NULL);
5333 
5334 	if (bserrno != 0 && ctx->bserrno == 0) {
5335 		ctx->bserrno = bserrno;
5336 	}
5337 
5338 	ctx->cb_fn(ctx->cb_arg, ctx->snapshot, ctx->bserrno);
5339 	free(ctx);
5340 }
5341 
5342 static void
5343 _spdk_delete_snapshot_cleanup_snapshot(void *cb_arg, int bserrno)
5344 {
5345 	struct delete_snapshot_ctx *ctx = cb_arg;
5346 
5347 	if (bserrno != 0) {
5348 		ctx->bserrno = bserrno;
5349 		SPDK_ERRLOG("Clone cleanup error %d\n", bserrno);
5350 	}
5351 
5352 	/* open_ref == 1 menas that only deletion context has opened this snapshot
5353 	 * open_ref == 2 menas that clone has opened this snapshot as well,
5354 	 * so we have to add it back to the blobs list */
5355 	if (ctx->snapshot->open_ref == 2) {
5356 		TAILQ_INSERT_HEAD(&ctx->snapshot->bs->blobs, ctx->snapshot, link);
5357 	}
5358 
5359 	ctx->snapshot->locked_operation_in_progress = false;
5360 	ctx->snapshot->md_ro = ctx->snapshot_md_ro;
5361 
5362 	spdk_blob_close(ctx->snapshot, _spdk_delete_blob_cleanup_finish, ctx);
5363 }
5364 
5365 static void
5366 _spdk_delete_snapshot_cleanup_clone(void *cb_arg, int bserrno)
5367 {
5368 	struct delete_snapshot_ctx *ctx = cb_arg;
5369 
5370 	ctx->clone->locked_operation_in_progress = false;
5371 	ctx->clone->md_ro = ctx->clone_md_ro;
5372 
5373 	spdk_blob_close(ctx->clone, _spdk_delete_snapshot_cleanup_snapshot, ctx);
5374 }
5375 
5376 static void
5377 _spdk_delete_snapshot_unfreeze_cpl(void *cb_arg, int bserrno)
5378 {
5379 	struct delete_snapshot_ctx *ctx = cb_arg;
5380 
5381 	if (bserrno) {
5382 		ctx->bserrno = bserrno;
5383 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5384 		return;
5385 	}
5386 
5387 	ctx->clone->locked_operation_in_progress = false;
5388 	spdk_blob_close(ctx->clone, _spdk_delete_blob_cleanup_finish, ctx);
5389 }
5390 
5391 static void
5392 _spdk_delete_snapshot_sync_snapshot_cpl(void *cb_arg, int bserrno)
5393 {
5394 	struct delete_snapshot_ctx *ctx = cb_arg;
5395 	struct spdk_blob_list *parent_snapshot_entry = NULL;
5396 	struct spdk_blob_list *snapshot_entry = NULL;
5397 	struct spdk_blob_list *clone_entry = NULL;
5398 	struct spdk_blob_list *snapshot_clone_entry = NULL;
5399 
5400 	if (bserrno) {
5401 		SPDK_ERRLOG("Failed to sync MD on blob\n");
5402 		ctx->bserrno = bserrno;
5403 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5404 		return;
5405 	}
5406 
5407 	/* Get snapshot entry for the snapshot we want to remove */
5408 	snapshot_entry = _spdk_bs_get_snapshot_entry(ctx->snapshot->bs, ctx->snapshot->id);
5409 
5410 	assert(snapshot_entry != NULL);
5411 
5412 	/* Remove clone entry in this snapshot (at this point there can be only one clone) */
5413 	clone_entry = TAILQ_FIRST(&snapshot_entry->clones);
5414 	assert(clone_entry != NULL);
5415 	TAILQ_REMOVE(&snapshot_entry->clones, clone_entry, link);
5416 	snapshot_entry->clone_count--;
5417 	assert(TAILQ_EMPTY(&snapshot_entry->clones));
5418 
5419 	if (ctx->snapshot->parent_id != SPDK_BLOBID_INVALID) {
5420 		/* This snapshot is at the same time a clone of another snapshot - we need to
5421 		 * update parent snapshot (remove current clone, add new one inherited from
5422 		 * the snapshot that is being removed) */
5423 
5424 		/* Get snapshot entry for parent snapshot and clone entry within that snapshot for
5425 		 * snapshot that we are removing */
5426 		_spdk_blob_get_snapshot_and_clone_entries(ctx->snapshot, &parent_snapshot_entry,
5427 				&snapshot_clone_entry);
5428 
5429 		/* Switch clone entry in parent snapshot */
5430 		TAILQ_INSERT_TAIL(&parent_snapshot_entry->clones, clone_entry, link);
5431 		TAILQ_REMOVE(&parent_snapshot_entry->clones, snapshot_clone_entry, link);
5432 		free(snapshot_clone_entry);
5433 	} else {
5434 		/* No parent snapshot - just remove clone entry */
5435 		free(clone_entry);
5436 	}
5437 
5438 	/* Restore md_ro flags */
5439 	ctx->clone->md_ro = ctx->clone_md_ro;
5440 	ctx->snapshot->md_ro = ctx->snapshot_md_ro;
5441 
5442 	_spdk_blob_unfreeze_io(ctx->clone, _spdk_delete_snapshot_unfreeze_cpl, ctx);
5443 }
5444 
5445 static void
5446 _spdk_delete_snapshot_sync_clone_cpl(void *cb_arg, int bserrno)
5447 {
5448 	struct delete_snapshot_ctx *ctx = cb_arg;
5449 	uint64_t i;
5450 
5451 	ctx->snapshot->md_ro = false;
5452 
5453 	if (bserrno) {
5454 		SPDK_ERRLOG("Failed to sync MD on clone\n");
5455 		ctx->bserrno = bserrno;
5456 
5457 		/* Restore snapshot to previous state */
5458 		bserrno = _spdk_blob_remove_xattr(ctx->snapshot, SNAPSHOT_PENDING_REMOVAL, true);
5459 		if (bserrno != 0) {
5460 			_spdk_delete_snapshot_cleanup_clone(ctx, bserrno);
5461 			return;
5462 		}
5463 
5464 		spdk_blob_sync_md(ctx->snapshot, _spdk_delete_snapshot_cleanup_clone, ctx);
5465 		return;
5466 	}
5467 
5468 	/* Clear cluster map entries for snapshot */
5469 	for (i = 0; i < ctx->snapshot->active.num_clusters && i < ctx->clone->active.num_clusters; i++) {
5470 		if (ctx->clone->active.clusters[i] == ctx->snapshot->active.clusters[i]) {
5471 			ctx->snapshot->active.clusters[i] = 0;
5472 		}
5473 	}
5474 
5475 	ctx->snapshot->state = SPDK_BLOB_STATE_DIRTY;
5476 
5477 	if (ctx->parent_snapshot_entry != NULL) {
5478 		ctx->snapshot->back_bs_dev = NULL;
5479 	}
5480 
5481 	spdk_blob_sync_md(ctx->snapshot, _spdk_delete_snapshot_sync_snapshot_cpl, ctx);
5482 }
5483 
5484 static void
5485 _spdk_delete_snapshot_sync_snapshot_xattr_cpl(void *cb_arg, int bserrno)
5486 {
5487 	struct delete_snapshot_ctx *ctx = cb_arg;
5488 	uint64_t i;
5489 
5490 	/* Temporarily override md_ro flag for clone for MD modification */
5491 	ctx->clone_md_ro = ctx->clone->md_ro;
5492 	ctx->clone->md_ro = false;
5493 
5494 	if (bserrno) {
5495 		SPDK_ERRLOG("Failed to sync MD with xattr on blob\n");
5496 		ctx->bserrno = bserrno;
5497 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5498 		return;
5499 	}
5500 
5501 	/* Copy snapshot map to clone map (only unallocated clusters in clone) */
5502 	for (i = 0; i < ctx->snapshot->active.num_clusters && i < ctx->clone->active.num_clusters; i++) {
5503 		if (ctx->clone->active.clusters[i] == 0) {
5504 			ctx->clone->active.clusters[i] = ctx->snapshot->active.clusters[i];
5505 		}
5506 	}
5507 
5508 	/* Delete old backing bs_dev from clone (related to snapshot that will be removed) */
5509 	ctx->clone->back_bs_dev->destroy(ctx->clone->back_bs_dev);
5510 
5511 	/* Set/remove snapshot xattr and switch parent ID and backing bs_dev on clone... */
5512 	if (ctx->parent_snapshot_entry != NULL) {
5513 		/* ...to parent snapshot */
5514 		ctx->clone->parent_id = ctx->parent_snapshot_entry->id;
5515 		ctx->clone->back_bs_dev = ctx->snapshot->back_bs_dev;
5516 		_spdk_blob_set_xattr(ctx->clone, BLOB_SNAPSHOT, &ctx->parent_snapshot_entry->id,
5517 				     sizeof(spdk_blob_id),
5518 				     true);
5519 	} else {
5520 		/* ...to blobid invalid and zeroes dev */
5521 		ctx->clone->parent_id = SPDK_BLOBID_INVALID;
5522 		ctx->clone->back_bs_dev = spdk_bs_create_zeroes_dev();
5523 		_spdk_blob_remove_xattr(ctx->clone, BLOB_SNAPSHOT, true);
5524 	}
5525 
5526 	spdk_blob_sync_md(ctx->clone, _spdk_delete_snapshot_sync_clone_cpl, ctx);
5527 }
5528 
5529 static void
5530 _spdk_delete_snapshot_freeze_io_cb(void *cb_arg, int bserrno)
5531 {
5532 	struct delete_snapshot_ctx *ctx = cb_arg;
5533 
5534 	if (bserrno) {
5535 		SPDK_ERRLOG("Failed to freeze I/O on clone\n");
5536 		ctx->bserrno = bserrno;
5537 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5538 		return;
5539 	}
5540 
5541 	/* Temporarily override md_ro flag for snapshot for MD modification */
5542 	ctx->snapshot_md_ro = ctx->snapshot->md_ro;
5543 	ctx->snapshot->md_ro = false;
5544 
5545 	/* Mark blob as pending for removal for power failure safety, use clone id for recovery */
5546 	ctx->bserrno = _spdk_blob_set_xattr(ctx->snapshot, SNAPSHOT_PENDING_REMOVAL, &ctx->clone->id,
5547 					    sizeof(spdk_blob_id), true);
5548 	if (ctx->bserrno != 0) {
5549 		_spdk_delete_snapshot_cleanup_clone(ctx, 0);
5550 		return;
5551 	}
5552 
5553 	spdk_blob_sync_md(ctx->snapshot, _spdk_delete_snapshot_sync_snapshot_xattr_cpl, ctx);
5554 }
5555 
5556 static void
5557 _spdk_delete_snapshot_open_clone_cb(void *cb_arg, struct spdk_blob *clone, int bserrno)
5558 {
5559 	struct delete_snapshot_ctx *ctx = cb_arg;
5560 
5561 	if (bserrno) {
5562 		SPDK_ERRLOG("Failed to open clone\n");
5563 		ctx->bserrno = bserrno;
5564 		_spdk_delete_snapshot_cleanup_snapshot(ctx, 0);
5565 		return;
5566 	}
5567 
5568 	ctx->clone = clone;
5569 
5570 	if (clone->locked_operation_in_progress) {
5571 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot remove blob - another operation in progress on its clone\n");
5572 		ctx->bserrno = -EBUSY;
5573 		spdk_blob_close(ctx->clone, _spdk_delete_snapshot_cleanup_snapshot, ctx);
5574 		return;
5575 	}
5576 
5577 	clone->locked_operation_in_progress = true;
5578 
5579 	_spdk_blob_freeze_io(clone, _spdk_delete_snapshot_freeze_io_cb, ctx);
5580 }
5581 
5582 static void
5583 _spdk_update_clone_on_snapshot_deletion(struct spdk_blob *snapshot, struct delete_snapshot_ctx *ctx)
5584 {
5585 	struct spdk_blob_list *snapshot_entry = NULL;
5586 	struct spdk_blob_list *clone_entry = NULL;
5587 	struct spdk_blob_list *snapshot_clone_entry = NULL;
5588 
5589 	/* Get snapshot entry for the snapshot we want to remove */
5590 	snapshot_entry = _spdk_bs_get_snapshot_entry(snapshot->bs, snapshot->id);
5591 
5592 	assert(snapshot_entry != NULL);
5593 
5594 	/* Get clone of the snapshot (at this point there can be only one clone) */
5595 	clone_entry = TAILQ_FIRST(&snapshot_entry->clones);
5596 	assert(snapshot_entry->clone_count == 1);
5597 	assert(clone_entry != NULL);
5598 
5599 	/* Get snapshot entry for parent snapshot and clone entry within that snapshot for
5600 	 * snapshot that we are removing */
5601 	_spdk_blob_get_snapshot_and_clone_entries(snapshot, &ctx->parent_snapshot_entry,
5602 			&snapshot_clone_entry);
5603 
5604 	spdk_bs_open_blob(snapshot->bs, clone_entry->id, _spdk_delete_snapshot_open_clone_cb, ctx);
5605 }
5606 
5607 static void
5608 _spdk_bs_delete_blob_finish(void *cb_arg, struct spdk_blob *blob, int bserrno)
5609 {
5610 	spdk_bs_sequence_t *seq = cb_arg;
5611 	struct spdk_blob_list *snapshot_entry = NULL;
5612 	uint32_t page_num;
5613 
5614 	if (bserrno) {
5615 		SPDK_ERRLOG("Failed to remove blob\n");
5616 		spdk_bs_sequence_finish(seq, bserrno);
5617 		return;
5618 	}
5619 
5620 	/* Remove snapshot from the list */
5621 	snapshot_entry = _spdk_bs_get_snapshot_entry(blob->bs, blob->id);
5622 	if (snapshot_entry != NULL) {
5623 		TAILQ_REMOVE(&blob->bs->snapshots, snapshot_entry, link);
5624 		free(snapshot_entry);
5625 	}
5626 
5627 	page_num = _spdk_bs_blobid_to_page(blob->id);
5628 	spdk_bit_array_clear(blob->bs->used_blobids, page_num);
5629 	blob->state = SPDK_BLOB_STATE_DIRTY;
5630 	blob->active.num_pages = 0;
5631 	_spdk_blob_resize(blob, 0);
5632 
5633 	_spdk_blob_persist(seq, blob, _spdk_bs_delete_persist_cpl, blob);
5634 }
5635 
5636 static int
5637 _spdk_bs_is_blob_deletable(struct spdk_blob *blob, bool *update_clone)
5638 {
5639 	struct spdk_blob_list *snapshot_entry = NULL;
5640 	struct spdk_blob_list *clone_entry = NULL;
5641 	struct spdk_blob *clone = NULL;
5642 	bool has_one_clone = false;
5643 
5644 	/* Check if this is a snapshot with clones */
5645 	snapshot_entry = _spdk_bs_get_snapshot_entry(blob->bs, blob->id);
5646 	if (snapshot_entry != NULL) {
5647 		if (snapshot_entry->clone_count > 1) {
5648 			SPDK_ERRLOG("Cannot remove snapshot with more than one clone\n");
5649 			return -EBUSY;
5650 		} else if (snapshot_entry->clone_count == 1) {
5651 			has_one_clone = true;
5652 		}
5653 	}
5654 
5655 	/* Check if someone has this blob open (besides this delete context):
5656 	 * - open_ref = 1 - only this context opened blob, so it is ok to remove it
5657 	 * - open_ref <= 2 && has_one_clone = true - clone is holding snapshot
5658 	 *	and that is ok, because we will update it accordingly */
5659 	if (blob->open_ref <= 2 && has_one_clone) {
5660 		clone_entry = TAILQ_FIRST(&snapshot_entry->clones);
5661 		assert(clone_entry != NULL);
5662 		clone = _spdk_blob_lookup(blob->bs, clone_entry->id);
5663 
5664 		if (blob->open_ref == 2 && clone == NULL) {
5665 			/* Clone is closed and someone else opened this blob */
5666 			SPDK_ERRLOG("Cannot remove snapshot because it is open\n");
5667 			return -EBUSY;
5668 		}
5669 
5670 		*update_clone = true;
5671 		return 0;
5672 	}
5673 
5674 	if (blob->open_ref > 1) {
5675 		SPDK_ERRLOG("Cannot remove snapshot because it is open\n");
5676 		return -EBUSY;
5677 	}
5678 
5679 	assert(has_one_clone == false);
5680 	*update_clone = false;
5681 	return 0;
5682 }
5683 
5684 static void
5685 _spdk_bs_delete_enomem_close_cpl(void *cb_arg, int bserrno)
5686 {
5687 	spdk_bs_sequence_t *seq = cb_arg;
5688 
5689 	spdk_bs_sequence_finish(seq, -ENOMEM);
5690 }
5691 
5692 static void
5693 _spdk_bs_delete_open_cpl(void *cb_arg, struct spdk_blob *blob, int bserrno)
5694 {
5695 	spdk_bs_sequence_t *seq = cb_arg;
5696 	struct delete_snapshot_ctx *ctx;
5697 	bool update_clone = false;
5698 
5699 	if (bserrno != 0) {
5700 		spdk_bs_sequence_finish(seq, bserrno);
5701 		return;
5702 	}
5703 
5704 	_spdk_blob_verify_md_op(blob);
5705 
5706 	ctx = calloc(1, sizeof(*ctx));
5707 	if (ctx == NULL) {
5708 		spdk_blob_close(blob, _spdk_bs_delete_enomem_close_cpl, seq);
5709 		return;
5710 	}
5711 
5712 	ctx->snapshot = blob;
5713 	ctx->cb_fn = _spdk_bs_delete_blob_finish;
5714 	ctx->cb_arg = seq;
5715 
5716 	/* Check if blob can be removed and if it is a snapshot with clone on top of it */
5717 	ctx->bserrno = _spdk_bs_is_blob_deletable(blob, &update_clone);
5718 	if (ctx->bserrno) {
5719 		spdk_blob_close(blob, _spdk_delete_blob_cleanup_finish, ctx);
5720 		return;
5721 	}
5722 
5723 	if (blob->locked_operation_in_progress) {
5724 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Cannot remove blob - another operation in progress\n");
5725 		ctx->bserrno = -EBUSY;
5726 		spdk_blob_close(blob, _spdk_delete_blob_cleanup_finish, ctx);
5727 		return;
5728 	}
5729 
5730 	blob->locked_operation_in_progress = true;
5731 
5732 	/*
5733 	 * Remove the blob from the blob_store list now, to ensure it does not
5734 	 *  get returned after this point by _spdk_blob_lookup().
5735 	 */
5736 	TAILQ_REMOVE(&blob->bs->blobs, blob, link);
5737 
5738 	if (update_clone) {
5739 		/* This blob is a snapshot with active clone - update clone first */
5740 		_spdk_update_clone_on_snapshot_deletion(blob, ctx);
5741 	} else {
5742 		/* This blob does not have any clones - just remove it */
5743 		_spdk_bs_blob_list_remove(blob);
5744 		_spdk_bs_delete_blob_finish(seq, blob, 0);
5745 		free(ctx);
5746 	}
5747 }
5748 
5749 void
5750 spdk_bs_delete_blob(struct spdk_blob_store *bs, spdk_blob_id blobid,
5751 		    spdk_blob_op_complete cb_fn, void *cb_arg)
5752 {
5753 	struct spdk_bs_cpl	cpl;
5754 	spdk_bs_sequence_t	*seq;
5755 
5756 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Deleting blob %lu\n", blobid);
5757 
5758 	assert(spdk_get_thread() == bs->md_thread);
5759 
5760 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
5761 	cpl.u.blob_basic.cb_fn = cb_fn;
5762 	cpl.u.blob_basic.cb_arg = cb_arg;
5763 
5764 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
5765 	if (!seq) {
5766 		cb_fn(cb_arg, -ENOMEM);
5767 		return;
5768 	}
5769 
5770 	spdk_bs_open_blob(bs, blobid, _spdk_bs_delete_open_cpl, seq);
5771 }
5772 
5773 /* END spdk_bs_delete_blob */
5774 
5775 /* START spdk_bs_open_blob */
5776 
5777 static void
5778 _spdk_bs_open_blob_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
5779 {
5780 	struct spdk_blob *blob = cb_arg;
5781 
5782 	if (bserrno != 0) {
5783 		_spdk_blob_free(blob);
5784 		seq->cpl.u.blob_handle.blob = NULL;
5785 		spdk_bs_sequence_finish(seq, bserrno);
5786 		return;
5787 	}
5788 
5789 	blob->open_ref++;
5790 
5791 	TAILQ_INSERT_HEAD(&blob->bs->blobs, blob, link);
5792 
5793 	spdk_bs_sequence_finish(seq, bserrno);
5794 }
5795 
5796 static void _spdk_bs_open_blob(struct spdk_blob_store *bs, spdk_blob_id blobid,
5797 			       struct spdk_blob_open_opts *opts, spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
5798 {
5799 	struct spdk_blob		*blob;
5800 	struct spdk_bs_cpl		cpl;
5801 	struct spdk_blob_open_opts	opts_default;
5802 	spdk_bs_sequence_t		*seq;
5803 	uint32_t			page_num;
5804 
5805 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Opening blob %lu\n", blobid);
5806 	assert(spdk_get_thread() == bs->md_thread);
5807 
5808 	page_num = _spdk_bs_blobid_to_page(blobid);
5809 	if (spdk_bit_array_get(bs->used_blobids, page_num) == false) {
5810 		/* Invalid blobid */
5811 		cb_fn(cb_arg, NULL, -ENOENT);
5812 		return;
5813 	}
5814 
5815 	blob = _spdk_blob_lookup(bs, blobid);
5816 	if (blob) {
5817 		blob->open_ref++;
5818 		cb_fn(cb_arg, blob, 0);
5819 		return;
5820 	}
5821 
5822 	blob = _spdk_blob_alloc(bs, blobid);
5823 	if (!blob) {
5824 		cb_fn(cb_arg, NULL, -ENOMEM);
5825 		return;
5826 	}
5827 
5828 	if (!opts) {
5829 		spdk_blob_open_opts_init(&opts_default);
5830 		opts = &opts_default;
5831 	}
5832 
5833 	blob->clear_method = opts->clear_method;
5834 
5835 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_HANDLE;
5836 	cpl.u.blob_handle.cb_fn = cb_fn;
5837 	cpl.u.blob_handle.cb_arg = cb_arg;
5838 	cpl.u.blob_handle.blob = blob;
5839 
5840 	seq = spdk_bs_sequence_start(bs->md_channel, &cpl);
5841 	if (!seq) {
5842 		_spdk_blob_free(blob);
5843 		cb_fn(cb_arg, NULL, -ENOMEM);
5844 		return;
5845 	}
5846 
5847 	_spdk_blob_load(seq, blob, _spdk_bs_open_blob_cpl, blob);
5848 }
5849 
5850 void spdk_bs_open_blob(struct spdk_blob_store *bs, spdk_blob_id blobid,
5851 		       spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
5852 {
5853 	_spdk_bs_open_blob(bs, blobid, NULL, cb_fn, cb_arg);
5854 }
5855 
5856 void spdk_bs_open_blob_ext(struct spdk_blob_store *bs, spdk_blob_id blobid,
5857 			   struct spdk_blob_open_opts *opts, spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
5858 {
5859 	_spdk_bs_open_blob(bs, blobid, opts, cb_fn, cb_arg);
5860 }
5861 
5862 /* END spdk_bs_open_blob */
5863 
5864 /* START spdk_blob_set_read_only */
5865 int spdk_blob_set_read_only(struct spdk_blob *blob)
5866 {
5867 	_spdk_blob_verify_md_op(blob);
5868 
5869 	blob->data_ro_flags |= SPDK_BLOB_READ_ONLY;
5870 
5871 	blob->state = SPDK_BLOB_STATE_DIRTY;
5872 	return 0;
5873 }
5874 /* END spdk_blob_set_read_only */
5875 
5876 /* START spdk_blob_sync_md */
5877 
5878 static void
5879 _spdk_blob_sync_md_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
5880 {
5881 	struct spdk_blob *blob = cb_arg;
5882 
5883 	if (bserrno == 0 && (blob->data_ro_flags & SPDK_BLOB_READ_ONLY)) {
5884 		blob->data_ro = true;
5885 		blob->md_ro = true;
5886 	}
5887 
5888 	spdk_bs_sequence_finish(seq, bserrno);
5889 }
5890 
5891 static void
5892 _spdk_blob_sync_md(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
5893 {
5894 	struct spdk_bs_cpl	cpl;
5895 	spdk_bs_sequence_t	*seq;
5896 
5897 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
5898 	cpl.u.blob_basic.cb_fn = cb_fn;
5899 	cpl.u.blob_basic.cb_arg = cb_arg;
5900 
5901 	seq = spdk_bs_sequence_start(blob->bs->md_channel, &cpl);
5902 	if (!seq) {
5903 		cb_fn(cb_arg, -ENOMEM);
5904 		return;
5905 	}
5906 
5907 	_spdk_blob_persist(seq, blob, _spdk_blob_sync_md_cpl, blob);
5908 }
5909 
5910 void
5911 spdk_blob_sync_md(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
5912 {
5913 	_spdk_blob_verify_md_op(blob);
5914 
5915 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Syncing blob %lu\n", blob->id);
5916 
5917 	if (blob->md_ro) {
5918 		assert(blob->state == SPDK_BLOB_STATE_CLEAN);
5919 		cb_fn(cb_arg, 0);
5920 		return;
5921 	}
5922 
5923 	_spdk_blob_sync_md(blob, cb_fn, cb_arg);
5924 }
5925 
5926 /* END spdk_blob_sync_md */
5927 
5928 struct spdk_blob_insert_cluster_ctx {
5929 	struct spdk_thread	*thread;
5930 	struct spdk_blob	*blob;
5931 	uint32_t		cluster_num;	/* cluster index in blob */
5932 	uint32_t		cluster;	/* cluster on disk */
5933 	int			rc;
5934 	spdk_blob_op_complete	cb_fn;
5935 	void			*cb_arg;
5936 };
5937 
5938 static void
5939 _spdk_blob_insert_cluster_msg_cpl(void *arg)
5940 {
5941 	struct spdk_blob_insert_cluster_ctx *ctx = arg;
5942 
5943 	ctx->cb_fn(ctx->cb_arg, ctx->rc);
5944 	free(ctx);
5945 }
5946 
5947 static void
5948 _spdk_blob_insert_cluster_msg_cb(void *arg, int bserrno)
5949 {
5950 	struct spdk_blob_insert_cluster_ctx *ctx = arg;
5951 
5952 	ctx->rc = bserrno;
5953 	spdk_thread_send_msg(ctx->thread, _spdk_blob_insert_cluster_msg_cpl, ctx);
5954 }
5955 
5956 static void
5957 _spdk_blob_insert_cluster_msg(void *arg)
5958 {
5959 	struct spdk_blob_insert_cluster_ctx *ctx = arg;
5960 
5961 	ctx->rc = _spdk_blob_insert_cluster(ctx->blob, ctx->cluster_num, ctx->cluster);
5962 	if (ctx->rc != 0) {
5963 		spdk_thread_send_msg(ctx->thread, _spdk_blob_insert_cluster_msg_cpl, ctx);
5964 		return;
5965 	}
5966 
5967 	ctx->blob->state = SPDK_BLOB_STATE_DIRTY;
5968 	_spdk_blob_sync_md(ctx->blob, _spdk_blob_insert_cluster_msg_cb, ctx);
5969 }
5970 
5971 static void
5972 _spdk_blob_insert_cluster_on_md_thread(struct spdk_blob *blob, uint32_t cluster_num,
5973 				       uint64_t cluster, spdk_blob_op_complete cb_fn, void *cb_arg)
5974 {
5975 	struct spdk_blob_insert_cluster_ctx *ctx;
5976 
5977 	ctx = calloc(1, sizeof(*ctx));
5978 	if (ctx == NULL) {
5979 		cb_fn(cb_arg, -ENOMEM);
5980 		return;
5981 	}
5982 
5983 	ctx->thread = spdk_get_thread();
5984 	ctx->blob = blob;
5985 	ctx->cluster_num = cluster_num;
5986 	ctx->cluster = cluster;
5987 	ctx->cb_fn = cb_fn;
5988 	ctx->cb_arg = cb_arg;
5989 
5990 	spdk_thread_send_msg(blob->bs->md_thread, _spdk_blob_insert_cluster_msg, ctx);
5991 }
5992 
5993 /* START spdk_blob_close */
5994 
5995 static void
5996 _spdk_blob_close_cpl(spdk_bs_sequence_t *seq, void *cb_arg, int bserrno)
5997 {
5998 	struct spdk_blob *blob = cb_arg;
5999 
6000 	if (bserrno == 0) {
6001 		blob->open_ref--;
6002 		if (blob->open_ref == 0) {
6003 			/*
6004 			 * Blobs with active.num_pages == 0 are deleted blobs.
6005 			 *  these blobs are removed from the blob_store list
6006 			 *  when the deletion process starts - so don't try to
6007 			 *  remove them again.
6008 			 */
6009 			if (blob->active.num_pages > 0) {
6010 				TAILQ_REMOVE(&blob->bs->blobs, blob, link);
6011 			}
6012 			_spdk_blob_free(blob);
6013 		}
6014 	}
6015 
6016 	spdk_bs_sequence_finish(seq, bserrno);
6017 }
6018 
6019 void spdk_blob_close(struct spdk_blob *blob, spdk_blob_op_complete cb_fn, void *cb_arg)
6020 {
6021 	struct spdk_bs_cpl	cpl;
6022 	spdk_bs_sequence_t	*seq;
6023 
6024 	_spdk_blob_verify_md_op(blob);
6025 
6026 	SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Closing blob %lu\n", blob->id);
6027 
6028 	if (blob->open_ref == 0) {
6029 		cb_fn(cb_arg, -EBADF);
6030 		return;
6031 	}
6032 
6033 	cpl.type = SPDK_BS_CPL_TYPE_BLOB_BASIC;
6034 	cpl.u.blob_basic.cb_fn = cb_fn;
6035 	cpl.u.blob_basic.cb_arg = cb_arg;
6036 
6037 	seq = spdk_bs_sequence_start(blob->bs->md_channel, &cpl);
6038 	if (!seq) {
6039 		cb_fn(cb_arg, -ENOMEM);
6040 		return;
6041 	}
6042 
6043 	/* Sync metadata */
6044 	_spdk_blob_persist(seq, blob, _spdk_blob_close_cpl, blob);
6045 }
6046 
6047 /* END spdk_blob_close */
6048 
6049 struct spdk_io_channel *spdk_bs_alloc_io_channel(struct spdk_blob_store *bs)
6050 {
6051 	return spdk_get_io_channel(bs);
6052 }
6053 
6054 void spdk_bs_free_io_channel(struct spdk_io_channel *channel)
6055 {
6056 	spdk_put_io_channel(channel);
6057 }
6058 
6059 void spdk_blob_io_unmap(struct spdk_blob *blob, struct spdk_io_channel *channel,
6060 			uint64_t offset, uint64_t length, spdk_blob_op_complete cb_fn, void *cb_arg)
6061 {
6062 	_spdk_blob_request_submit_op(blob, channel, NULL, offset, length, cb_fn, cb_arg,
6063 				     SPDK_BLOB_UNMAP);
6064 }
6065 
6066 void spdk_blob_io_write_zeroes(struct spdk_blob *blob, struct spdk_io_channel *channel,
6067 			       uint64_t offset, uint64_t length, spdk_blob_op_complete cb_fn, void *cb_arg)
6068 {
6069 	_spdk_blob_request_submit_op(blob, channel, NULL, offset, length, cb_fn, cb_arg,
6070 				     SPDK_BLOB_WRITE_ZEROES);
6071 }
6072 
6073 void spdk_blob_io_write(struct spdk_blob *blob, struct spdk_io_channel *channel,
6074 			void *payload, uint64_t offset, uint64_t length,
6075 			spdk_blob_op_complete cb_fn, void *cb_arg)
6076 {
6077 	_spdk_blob_request_submit_op(blob, channel, payload, offset, length, cb_fn, cb_arg,
6078 				     SPDK_BLOB_WRITE);
6079 }
6080 
6081 void spdk_blob_io_read(struct spdk_blob *blob, struct spdk_io_channel *channel,
6082 		       void *payload, uint64_t offset, uint64_t length,
6083 		       spdk_blob_op_complete cb_fn, void *cb_arg)
6084 {
6085 	_spdk_blob_request_submit_op(blob, channel, payload, offset, length, cb_fn, cb_arg,
6086 				     SPDK_BLOB_READ);
6087 }
6088 
6089 void spdk_blob_io_writev(struct spdk_blob *blob, struct spdk_io_channel *channel,
6090 			 struct iovec *iov, int iovcnt, uint64_t offset, uint64_t length,
6091 			 spdk_blob_op_complete cb_fn, void *cb_arg)
6092 {
6093 	_spdk_blob_request_submit_rw_iov(blob, channel, iov, iovcnt, offset, length, cb_fn, cb_arg, false);
6094 }
6095 
6096 void spdk_blob_io_readv(struct spdk_blob *blob, struct spdk_io_channel *channel,
6097 			struct iovec *iov, int iovcnt, uint64_t offset, uint64_t length,
6098 			spdk_blob_op_complete cb_fn, void *cb_arg)
6099 {
6100 	_spdk_blob_request_submit_rw_iov(blob, channel, iov, iovcnt, offset, length, cb_fn, cb_arg, true);
6101 }
6102 
6103 struct spdk_bs_iter_ctx {
6104 	int64_t page_num;
6105 	struct spdk_blob_store *bs;
6106 
6107 	spdk_blob_op_with_handle_complete cb_fn;
6108 	void *cb_arg;
6109 };
6110 
6111 static void
6112 _spdk_bs_iter_cpl(void *cb_arg, struct spdk_blob *_blob, int bserrno)
6113 {
6114 	struct spdk_bs_iter_ctx *ctx = cb_arg;
6115 	struct spdk_blob_store *bs = ctx->bs;
6116 	spdk_blob_id id;
6117 
6118 	if (bserrno == 0) {
6119 		ctx->cb_fn(ctx->cb_arg, _blob, bserrno);
6120 		free(ctx);
6121 		return;
6122 	}
6123 
6124 	ctx->page_num++;
6125 	ctx->page_num = spdk_bit_array_find_first_set(bs->used_blobids, ctx->page_num);
6126 	if (ctx->page_num >= spdk_bit_array_capacity(bs->used_blobids)) {
6127 		ctx->cb_fn(ctx->cb_arg, NULL, -ENOENT);
6128 		free(ctx);
6129 		return;
6130 	}
6131 
6132 	id = _spdk_bs_page_to_blobid(ctx->page_num);
6133 
6134 	spdk_bs_open_blob(bs, id, _spdk_bs_iter_cpl, ctx);
6135 }
6136 
6137 void
6138 spdk_bs_iter_first(struct spdk_blob_store *bs,
6139 		   spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
6140 {
6141 	struct spdk_bs_iter_ctx *ctx;
6142 
6143 	ctx = calloc(1, sizeof(*ctx));
6144 	if (!ctx) {
6145 		cb_fn(cb_arg, NULL, -ENOMEM);
6146 		return;
6147 	}
6148 
6149 	ctx->page_num = -1;
6150 	ctx->bs = bs;
6151 	ctx->cb_fn = cb_fn;
6152 	ctx->cb_arg = cb_arg;
6153 
6154 	_spdk_bs_iter_cpl(ctx, NULL, -1);
6155 }
6156 
6157 static void
6158 _spdk_bs_iter_close_cpl(void *cb_arg, int bserrno)
6159 {
6160 	struct spdk_bs_iter_ctx *ctx = cb_arg;
6161 
6162 	_spdk_bs_iter_cpl(ctx, NULL, -1);
6163 }
6164 
6165 void
6166 spdk_bs_iter_next(struct spdk_blob_store *bs, struct spdk_blob *blob,
6167 		  spdk_blob_op_with_handle_complete cb_fn, void *cb_arg)
6168 {
6169 	struct spdk_bs_iter_ctx *ctx;
6170 
6171 	assert(blob != NULL);
6172 
6173 	ctx = calloc(1, sizeof(*ctx));
6174 	if (!ctx) {
6175 		cb_fn(cb_arg, NULL, -ENOMEM);
6176 		return;
6177 	}
6178 
6179 	ctx->page_num = _spdk_bs_blobid_to_page(blob->id);
6180 	ctx->bs = bs;
6181 	ctx->cb_fn = cb_fn;
6182 	ctx->cb_arg = cb_arg;
6183 
6184 	/* Close the existing blob */
6185 	spdk_blob_close(blob, _spdk_bs_iter_close_cpl, ctx);
6186 }
6187 
6188 static int
6189 _spdk_blob_set_xattr(struct spdk_blob *blob, const char *name, const void *value,
6190 		     uint16_t value_len, bool internal)
6191 {
6192 	struct spdk_xattr_tailq *xattrs;
6193 	struct spdk_xattr	*xattr;
6194 	size_t			desc_size;
6195 
6196 	_spdk_blob_verify_md_op(blob);
6197 
6198 	if (blob->md_ro) {
6199 		return -EPERM;
6200 	}
6201 
6202 	desc_size = sizeof(struct spdk_blob_md_descriptor_xattr) + strlen(name) + value_len;
6203 	if (desc_size > SPDK_BS_MAX_DESC_SIZE) {
6204 		SPDK_DEBUGLOG(SPDK_LOG_BLOB, "Xattr '%s' of size %ld does not fix into single page %ld\n", name,
6205 			      desc_size, SPDK_BS_MAX_DESC_SIZE);
6206 		return -ENOMEM;
6207 	}
6208 
6209 	if (internal) {
6210 		xattrs = &blob->xattrs_internal;
6211 		blob->invalid_flags |= SPDK_BLOB_INTERNAL_XATTR;
6212 	} else {
6213 		xattrs = &blob->xattrs;
6214 	}
6215 
6216 	TAILQ_FOREACH(xattr, xattrs, link) {
6217 		if (!strcmp(name, xattr->name)) {
6218 			free(xattr->value);
6219 			xattr->value_len = value_len;
6220 			xattr->value = malloc(value_len);
6221 			memcpy(xattr->value, value, value_len);
6222 
6223 			blob->state = SPDK_BLOB_STATE_DIRTY;
6224 
6225 			return 0;
6226 		}
6227 	}
6228 
6229 	xattr = calloc(1, sizeof(*xattr));
6230 	if (!xattr) {
6231 		return -ENOMEM;
6232 	}
6233 	xattr->name = strdup(name);
6234 	xattr->value_len = value_len;
6235 	xattr->value = malloc(value_len);
6236 	memcpy(xattr->value, value, value_len);
6237 	TAILQ_INSERT_TAIL(xattrs, xattr, link);
6238 
6239 	blob->state = SPDK_BLOB_STATE_DIRTY;
6240 
6241 	return 0;
6242 }
6243 
6244 int
6245 spdk_blob_set_xattr(struct spdk_blob *blob, const char *name, const void *value,
6246 		    uint16_t value_len)
6247 {
6248 	return _spdk_blob_set_xattr(blob, name, value, value_len, false);
6249 }
6250 
6251 static int
6252 _spdk_blob_remove_xattr(struct spdk_blob *blob, const char *name, bool internal)
6253 {
6254 	struct spdk_xattr_tailq *xattrs;
6255 	struct spdk_xattr	*xattr;
6256 
6257 	_spdk_blob_verify_md_op(blob);
6258 
6259 	if (blob->md_ro) {
6260 		return -EPERM;
6261 	}
6262 	xattrs = internal ? &blob->xattrs_internal : &blob->xattrs;
6263 
6264 	TAILQ_FOREACH(xattr, xattrs, link) {
6265 		if (!strcmp(name, xattr->name)) {
6266 			TAILQ_REMOVE(xattrs, xattr, link);
6267 			free(xattr->value);
6268 			free(xattr->name);
6269 			free(xattr);
6270 
6271 			if (internal && TAILQ_EMPTY(&blob->xattrs_internal)) {
6272 				blob->invalid_flags &= ~SPDK_BLOB_INTERNAL_XATTR;
6273 			}
6274 			blob->state = SPDK_BLOB_STATE_DIRTY;
6275 
6276 			return 0;
6277 		}
6278 	}
6279 
6280 	return -ENOENT;
6281 }
6282 
6283 int
6284 spdk_blob_remove_xattr(struct spdk_blob *blob, const char *name)
6285 {
6286 	return _spdk_blob_remove_xattr(blob, name, false);
6287 }
6288 
6289 static int
6290 _spdk_blob_get_xattr_value(struct spdk_blob *blob, const char *name,
6291 			   const void **value, size_t *value_len, bool internal)
6292 {
6293 	struct spdk_xattr	*xattr;
6294 	struct spdk_xattr_tailq *xattrs;
6295 
6296 	xattrs = internal ? &blob->xattrs_internal : &blob->xattrs;
6297 
6298 	TAILQ_FOREACH(xattr, xattrs, link) {
6299 		if (!strcmp(name, xattr->name)) {
6300 			*value = xattr->value;
6301 			*value_len = xattr->value_len;
6302 			return 0;
6303 		}
6304 	}
6305 	return -ENOENT;
6306 }
6307 
6308 int
6309 spdk_blob_get_xattr_value(struct spdk_blob *blob, const char *name,
6310 			  const void **value, size_t *value_len)
6311 {
6312 	_spdk_blob_verify_md_op(blob);
6313 
6314 	return _spdk_blob_get_xattr_value(blob, name, value, value_len, false);
6315 }
6316 
6317 struct spdk_xattr_names {
6318 	uint32_t	count;
6319 	const char	*names[0];
6320 };
6321 
6322 static int
6323 _spdk_blob_get_xattr_names(struct spdk_xattr_tailq *xattrs, struct spdk_xattr_names **names)
6324 {
6325 	struct spdk_xattr	*xattr;
6326 	int			count = 0;
6327 
6328 	TAILQ_FOREACH(xattr, xattrs, link) {
6329 		count++;
6330 	}
6331 
6332 	*names = calloc(1, sizeof(struct spdk_xattr_names) + count * sizeof(char *));
6333 	if (*names == NULL) {
6334 		return -ENOMEM;
6335 	}
6336 
6337 	TAILQ_FOREACH(xattr, xattrs, link) {
6338 		(*names)->names[(*names)->count++] = xattr->name;
6339 	}
6340 
6341 	return 0;
6342 }
6343 
6344 int
6345 spdk_blob_get_xattr_names(struct spdk_blob *blob, struct spdk_xattr_names **names)
6346 {
6347 	_spdk_blob_verify_md_op(blob);
6348 
6349 	return _spdk_blob_get_xattr_names(&blob->xattrs, names);
6350 }
6351 
6352 uint32_t
6353 spdk_xattr_names_get_count(struct spdk_xattr_names *names)
6354 {
6355 	assert(names != NULL);
6356 
6357 	return names->count;
6358 }
6359 
6360 const char *
6361 spdk_xattr_names_get_name(struct spdk_xattr_names *names, uint32_t index)
6362 {
6363 	if (index >= names->count) {
6364 		return NULL;
6365 	}
6366 
6367 	return names->names[index];
6368 }
6369 
6370 void
6371 spdk_xattr_names_free(struct spdk_xattr_names *names)
6372 {
6373 	free(names);
6374 }
6375 
6376 struct spdk_bs_type
6377 spdk_bs_get_bstype(struct spdk_blob_store *bs)
6378 {
6379 	return bs->bstype;
6380 }
6381 
6382 void
6383 spdk_bs_set_bstype(struct spdk_blob_store *bs, struct spdk_bs_type bstype)
6384 {
6385 	memcpy(&bs->bstype, &bstype, sizeof(bstype));
6386 }
6387 
6388 bool
6389 spdk_blob_is_read_only(struct spdk_blob *blob)
6390 {
6391 	assert(blob != NULL);
6392 	return (blob->data_ro || blob->md_ro);
6393 }
6394 
6395 bool
6396 spdk_blob_is_snapshot(struct spdk_blob *blob)
6397 {
6398 	struct spdk_blob_list *snapshot_entry;
6399 
6400 	assert(blob != NULL);
6401 
6402 	snapshot_entry = _spdk_bs_get_snapshot_entry(blob->bs, blob->id);
6403 	if (snapshot_entry == NULL) {
6404 		return false;
6405 	}
6406 
6407 	return true;
6408 }
6409 
6410 bool
6411 spdk_blob_is_clone(struct spdk_blob *blob)
6412 {
6413 	assert(blob != NULL);
6414 
6415 	if (blob->parent_id != SPDK_BLOBID_INVALID) {
6416 		assert(spdk_blob_is_thin_provisioned(blob));
6417 		return true;
6418 	}
6419 
6420 	return false;
6421 }
6422 
6423 bool
6424 spdk_blob_is_thin_provisioned(struct spdk_blob *blob)
6425 {
6426 	assert(blob != NULL);
6427 	return !!(blob->invalid_flags & SPDK_BLOB_THIN_PROV);
6428 }
6429 
6430 static void
6431 _spdk_blob_update_clear_method(struct spdk_blob *blob)
6432 {
6433 	enum blob_clear_method stored_cm;
6434 
6435 	assert(blob != NULL);
6436 
6437 	/* If BLOB_CLEAR_WITH_DEFAULT was passed in, use the setting stored
6438 	 * in metadata previously.  If something other than the default was
6439 	 * specified, ignore stored value and used what was passed in.
6440 	 */
6441 	stored_cm = ((blob->md_ro_flags & SPDK_BLOB_CLEAR_METHOD) >> SPDK_BLOB_CLEAR_METHOD_SHIFT);
6442 
6443 	if (blob->clear_method == BLOB_CLEAR_WITH_DEFAULT) {
6444 		blob->clear_method = stored_cm;
6445 	} else if (blob->clear_method != stored_cm) {
6446 		SPDK_WARNLOG("Using passed in clear method 0x%x instead of stored value of 0x%x\n",
6447 			     blob->clear_method, stored_cm);
6448 	}
6449 }
6450 
6451 spdk_blob_id
6452 spdk_blob_get_parent_snapshot(struct spdk_blob_store *bs, spdk_blob_id blob_id)
6453 {
6454 	struct spdk_blob_list *snapshot_entry = NULL;
6455 	struct spdk_blob_list *clone_entry = NULL;
6456 
6457 	TAILQ_FOREACH(snapshot_entry, &bs->snapshots, link) {
6458 		TAILQ_FOREACH(clone_entry, &snapshot_entry->clones, link) {
6459 			if (clone_entry->id == blob_id) {
6460 				return snapshot_entry->id;
6461 			}
6462 		}
6463 	}
6464 
6465 	return SPDK_BLOBID_INVALID;
6466 }
6467 
6468 int
6469 spdk_blob_get_clones(struct spdk_blob_store *bs, spdk_blob_id blobid, spdk_blob_id *ids,
6470 		     size_t *count)
6471 {
6472 	struct spdk_blob_list *snapshot_entry, *clone_entry;
6473 	size_t n;
6474 
6475 	snapshot_entry = _spdk_bs_get_snapshot_entry(bs, blobid);
6476 	if (snapshot_entry == NULL) {
6477 		*count = 0;
6478 		return 0;
6479 	}
6480 
6481 	if (ids == NULL || *count < snapshot_entry->clone_count) {
6482 		*count = snapshot_entry->clone_count;
6483 		return -ENOMEM;
6484 	}
6485 	*count = snapshot_entry->clone_count;
6486 
6487 	n = 0;
6488 	TAILQ_FOREACH(clone_entry, &snapshot_entry->clones, link) {
6489 		ids[n++] = clone_entry->id;
6490 	}
6491 
6492 	return 0;
6493 }
6494 
6495 SPDK_LOG_REGISTER_COMPONENT("blob", SPDK_LOG_BLOB)
6496