Line data Source code
1 : /* SPDX-License-Identifier: BSD-3-Clause
2 : * Copyright (C) 2017 Intel Corporation.
3 : * All rights reserved.
4 : * Copyright (c) 2022-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
5 : */
6 :
7 : #ifndef SPDK_BLOBSTORE_H
8 : #define SPDK_BLOBSTORE_H
9 :
10 : #include "spdk/assert.h"
11 : #include "spdk/blob.h"
12 : #include "spdk/queue.h"
13 : #include "spdk/util.h"
14 : #include "spdk/tree.h"
15 : #include "spdk/thread.h"
16 :
17 : #include "request.h"
18 :
19 : /* In Memory Data Structures
20 : *
21 : * The following data structures exist only in memory.
22 : */
23 :
24 : #define SPDK_BLOB_OPTS_CLUSTER_SZ (1024 * 1024)
25 : #define SPDK_BLOB_OPTS_NUM_MD_PAGES UINT32_MAX
26 : #define SPDK_BLOB_OPTS_MAX_MD_OPS 32
27 : #define SPDK_BLOB_OPTS_DEFAULT_CHANNEL_OPS 512
28 : #define SPDK_BLOB_BLOBID_HIGH_BIT (1ULL << 32)
29 :
30 : struct spdk_xattr {
31 : uint32_t index;
32 : uint16_t value_len;
33 : char *name;
34 : void *value;
35 : TAILQ_ENTRY(spdk_xattr) link;
36 : };
37 :
38 : /* The mutable part of the blob data that is sync'd to
39 : * disk. The data in here is both mutable and persistent.
40 : */
41 : struct spdk_blob_mut_data {
42 : /* Number of data clusters in the blob */
43 : uint64_t num_clusters;
44 :
45 : /* Array LBAs that are the beginning of a cluster, in
46 : * the order they appear in the blob.
47 : */
48 : uint64_t *clusters;
49 :
50 : /* The size of the clusters array. This is greater than or
51 : * equal to 'num_clusters'.
52 : */
53 : size_t cluster_array_size;
54 :
55 : /* The number of allocated clusters in the clusters array */
56 : uint64_t num_allocated_clusters;
57 :
58 : /* Number of extent pages */
59 : uint64_t num_extent_pages;
60 :
61 : /* Array of page offsets into the metadata region,
62 : * containing extents. Can contain entries for not yet
63 : * allocated pages. */
64 : uint32_t *extent_pages;
65 :
66 : /* The size of the extent page array. This is greater than or
67 : * equal to 'num_extent_pages'. */
68 : size_t extent_pages_array_size;
69 :
70 : /* Number of metadata pages */
71 : uint32_t num_pages;
72 :
73 : /* Array of page offsets into the metadata region, in
74 : * the order of the metadata page sequence.
75 : */
76 : uint32_t *pages;
77 : };
78 :
79 : enum spdk_blob_state {
80 : /* The blob in-memory version does not match the on-disk
81 : * version.
82 : */
83 : SPDK_BLOB_STATE_DIRTY,
84 :
85 : /* The blob in memory version of the blob matches the on disk
86 : * version.
87 : */
88 : SPDK_BLOB_STATE_CLEAN,
89 :
90 : /* The in-memory state being synchronized with the on-disk
91 : * blob state. */
92 : SPDK_BLOB_STATE_LOADING,
93 : };
94 :
95 : TAILQ_HEAD(spdk_xattr_tailq, spdk_xattr);
96 :
97 : struct spdk_blob_list {
98 : spdk_blob_id id;
99 : size_t clone_count;
100 : TAILQ_HEAD(, spdk_blob_list) clones;
101 : TAILQ_ENTRY(spdk_blob_list) link;
102 : };
103 :
104 : struct spdk_blob {
105 : struct spdk_blob_store *bs;
106 :
107 : uint32_t open_ref;
108 :
109 : spdk_blob_id id;
110 : spdk_blob_id parent_id;
111 :
112 : enum spdk_blob_state state;
113 :
114 : /* Two copies of the mutable data. One is a version
115 : * that matches the last known data on disk (clean).
116 : * The other (active) is the current data. Syncing
117 : * a blob makes the clean match the active.
118 : */
119 : struct spdk_blob_mut_data clean;
120 : struct spdk_blob_mut_data active;
121 :
122 : bool invalid;
123 : bool data_ro;
124 : bool md_ro;
125 :
126 : uint64_t invalid_flags;
127 : uint64_t data_ro_flags;
128 : uint64_t md_ro_flags;
129 :
130 : struct spdk_bs_dev *back_bs_dev;
131 :
132 : /* TODO: The xattrs are mutable, but we don't want to be
133 : * copying them unnecessarily. Figure this out.
134 : */
135 : struct spdk_xattr_tailq xattrs;
136 : struct spdk_xattr_tailq xattrs_internal;
137 :
138 : RB_ENTRY(spdk_blob) link;
139 :
140 : uint32_t frozen_refcnt;
141 : bool locked_operation_in_progress;
142 : enum blob_clear_method clear_method;
143 : bool extent_rle_found;
144 : bool extent_table_found;
145 : bool use_extent_table;
146 :
147 : /* A list of pending metadata pending_persists */
148 : TAILQ_HEAD(, spdk_blob_persist_ctx) pending_persists;
149 : TAILQ_HEAD(, spdk_blob_persist_ctx) persists_to_complete;
150 :
151 : /* Number of data clusters retrieved from extent table,
152 : * that many have to be read from extent pages. */
153 : uint64_t remaining_clusters_in_et;
154 : };
155 :
156 : struct spdk_blob_store {
157 : uint64_t md_start; /* Offset from beginning of disk, in pages */
158 : uint32_t md_len; /* Count, in pages */
159 :
160 : struct spdk_io_channel *md_channel;
161 : uint32_t max_channel_ops;
162 :
163 : struct spdk_thread *md_thread;
164 :
165 : struct spdk_bs_dev *dev;
166 :
167 : struct spdk_bit_array *used_md_pages; /* Protected by used_lock */
168 : struct spdk_bit_pool *used_clusters; /* Protected by used_lock */
169 : struct spdk_bit_array *used_blobids;
170 : struct spdk_bit_array *open_blobids;
171 :
172 : struct spdk_spinlock used_lock;
173 :
174 : uint32_t cluster_sz;
175 : uint64_t total_clusters;
176 : uint64_t total_data_clusters;
177 : uint64_t num_free_clusters; /* Protected by used_lock */
178 : uint64_t pages_per_cluster;
179 : uint8_t pages_per_cluster_shift;
180 : uint32_t io_unit_size;
181 :
182 : spdk_blob_id super_blob;
183 : struct spdk_bs_type bstype;
184 :
185 : struct spdk_bs_cpl unload_cpl;
186 : int unload_err;
187 :
188 : RB_HEAD(spdk_blob_tree, spdk_blob) open_blobs;
189 : TAILQ_HEAD(, spdk_blob_list) snapshots;
190 :
191 : bool clean;
192 :
193 : spdk_bs_esnap_dev_create esnap_bs_dev_create;
194 : void *esnap_ctx;
195 :
196 : /* If external snapshot channels are being destroyed while
197 : * the blobstore is unloaded, the unload is deferred until
198 : * after the channel destruction completes.
199 : */
200 : uint32_t esnap_channels_unloading;
201 : spdk_bs_op_complete esnap_unload_cb_fn;
202 : void *esnap_unload_cb_arg;
203 : };
204 :
205 : struct spdk_bs_channel {
206 : struct spdk_bs_request_set *req_mem;
207 : TAILQ_HEAD(, spdk_bs_request_set) reqs;
208 :
209 : struct spdk_blob_store *bs;
210 :
211 : struct spdk_bs_dev *dev;
212 : struct spdk_io_channel *dev_channel;
213 :
214 : /* This page is only used during insert of a new cluster. */
215 : struct spdk_blob_md_page *new_cluster_page;
216 :
217 : TAILQ_HEAD(, spdk_bs_request_set) need_cluster_alloc;
218 : TAILQ_HEAD(, spdk_bs_request_set) queued_io;
219 :
220 : RB_HEAD(blob_esnap_channel_tree, blob_esnap_channel) esnap_channels;
221 : };
222 :
223 : /** operation type */
224 : enum spdk_blob_op_type {
225 : SPDK_BLOB_WRITE,
226 : SPDK_BLOB_READ,
227 : SPDK_BLOB_UNMAP,
228 : SPDK_BLOB_WRITE_ZEROES,
229 : SPDK_BLOB_WRITEV,
230 : SPDK_BLOB_READV,
231 : };
232 :
233 : /* back bs_dev */
234 :
235 : #define BLOB_SNAPSHOT "SNAP"
236 : #define SNAPSHOT_IN_PROGRESS "SNAPTMP"
237 : #define SNAPSHOT_PENDING_REMOVAL "SNAPRM"
238 : #define BLOB_EXTERNAL_SNAPSHOT_ID "EXTSNAP"
239 :
240 : struct spdk_blob_bs_dev {
241 : struct spdk_bs_dev bs_dev;
242 : struct spdk_blob *blob;
243 : };
244 :
245 : /* On-Disk Data Structures
246 : *
247 : * The following data structures exist on disk.
248 : */
249 : #define SPDK_BS_INITIAL_VERSION 1
250 : #define SPDK_BS_VERSION 3 /* current version */
251 :
252 : #pragma pack(push, 1)
253 :
254 : #define SPDK_MD_MASK_TYPE_USED_PAGES 0
255 : #define SPDK_MD_MASK_TYPE_USED_CLUSTERS 1
256 : #define SPDK_MD_MASK_TYPE_USED_BLOBIDS 2
257 :
258 : struct spdk_bs_md_mask {
259 : uint8_t type;
260 : uint32_t length; /* In bits */
261 : uint8_t mask[0];
262 : };
263 :
264 : #define SPDK_MD_DESCRIPTOR_TYPE_PADDING 0
265 : #define SPDK_MD_DESCRIPTOR_TYPE_XATTR 2
266 : #define SPDK_MD_DESCRIPTOR_TYPE_FLAGS 3
267 : #define SPDK_MD_DESCRIPTOR_TYPE_XATTR_INTERNAL 4
268 :
269 : /* Following descriptors define cluster layout in a blob.
270 : * EXTENT_RLE cannot be present in blobs metadata,
271 : * at the same time as EXTENT_TABLE and EXTENT_PAGE descriptors. */
272 :
273 : /* EXTENT_RLE descriptor holds an array of LBA that points to
274 : * beginning of allocated clusters. The array is run-length encoded,
275 : * with 0's being unallocated clusters. It is part of serialized
276 : * metadata chain for a blob. */
277 : #define SPDK_MD_DESCRIPTOR_TYPE_EXTENT_RLE 1
278 : /* EXTENT_TABLE descriptor holds array of md page offsets that
279 : * point to pages with EXTENT_PAGE descriptor. The 0's in the array
280 : * are run-length encoded, non-zero values are unallocated pages.
281 : * It is part of serialized metadata chain for a blob. */
282 : #define SPDK_MD_DESCRIPTOR_TYPE_EXTENT_TABLE 5
283 : /* EXTENT_PAGE descriptor holds an array of LBAs that point to
284 : * beginning of allocated clusters. The array is run-length encoded,
285 : * with 0's being unallocated clusters. It is NOT part of
286 : * serialized metadata chain for a blob. */
287 : #define SPDK_MD_DESCRIPTOR_TYPE_EXTENT_PAGE 6
288 :
289 : struct spdk_blob_md_descriptor_xattr {
290 : uint8_t type;
291 : uint32_t length;
292 :
293 : uint16_t name_length;
294 : uint16_t value_length;
295 :
296 : char name[0];
297 : /* String name immediately followed by string value. */
298 : };
299 :
300 : struct spdk_blob_md_descriptor_extent_rle {
301 : uint8_t type;
302 : uint32_t length;
303 :
304 : struct {
305 : uint32_t cluster_idx;
306 : uint32_t length; /* In units of clusters */
307 : } extents[0];
308 : };
309 :
310 : struct spdk_blob_md_descriptor_extent_table {
311 : uint8_t type;
312 : uint32_t length;
313 :
314 : /* Number of data clusters in the blob */
315 : uint64_t num_clusters;
316 :
317 : struct {
318 : uint32_t page_idx;
319 : uint32_t num_pages; /* In units of pages */
320 : } extent_page[0];
321 : };
322 :
323 : struct spdk_blob_md_descriptor_extent_page {
324 : uint8_t type;
325 : uint32_t length;
326 :
327 : /* First cluster index in this extent page */
328 : uint32_t start_cluster_idx;
329 :
330 : uint32_t cluster_idx[0];
331 : };
332 :
333 : #define SPDK_BLOB_THIN_PROV (1ULL << 0)
334 : #define SPDK_BLOB_INTERNAL_XATTR (1ULL << 1)
335 : #define SPDK_BLOB_EXTENT_TABLE (1ULL << 2)
336 : #define SPDK_BLOB_EXTERNAL_SNAPSHOT (1ULL << 3)
337 : #define SPDK_BLOB_INVALID_FLAGS_MASK (SPDK_BLOB_THIN_PROV | SPDK_BLOB_INTERNAL_XATTR | \
338 : SPDK_BLOB_EXTENT_TABLE | SPDK_BLOB_EXTERNAL_SNAPSHOT)
339 :
340 : #define SPDK_BLOB_READ_ONLY (1ULL << 0)
341 : #define SPDK_BLOB_DATA_RO_FLAGS_MASK SPDK_BLOB_READ_ONLY
342 :
343 : #define SPDK_BLOB_CLEAR_METHOD_SHIFT 0
344 : #define SPDK_BLOB_CLEAR_METHOD (3ULL << SPDK_BLOB_CLEAR_METHOD_SHIFT)
345 : #define SPDK_BLOB_MD_RO_FLAGS_MASK SPDK_BLOB_CLEAR_METHOD
346 :
347 : struct spdk_blob_md_descriptor_flags {
348 : uint8_t type;
349 : uint32_t length;
350 :
351 : /*
352 : * If a flag in invalid_flags is set that the application is not aware of,
353 : * it will not allow the blob to be opened.
354 : */
355 : uint64_t invalid_flags;
356 :
357 : /*
358 : * If a flag in data_ro_flags is set that the application is not aware of,
359 : * allow the blob to be opened in data_read_only and md_read_only mode.
360 : */
361 : uint64_t data_ro_flags;
362 :
363 : /*
364 : * If a flag in md_ro_flags is set the application is not aware of,
365 : * allow the blob to be opened in md_read_only mode.
366 : */
367 : uint64_t md_ro_flags;
368 : };
369 :
370 : struct spdk_blob_md_descriptor {
371 : uint8_t type;
372 : uint32_t length;
373 : };
374 :
375 : #define SPDK_INVALID_MD_PAGE UINT32_MAX
376 :
377 : struct spdk_blob_md_page {
378 : spdk_blob_id id;
379 :
380 : uint32_t sequence_num;
381 : uint32_t reserved0;
382 :
383 : /* Descriptors here */
384 : uint8_t descriptors[4072];
385 :
386 : uint32_t next;
387 : uint32_t crc;
388 : };
389 : #define SPDK_BS_PAGE_SIZE 0x1000
390 : SPDK_STATIC_ASSERT(SPDK_BS_PAGE_SIZE == sizeof(struct spdk_blob_md_page), "Invalid md page size");
391 :
392 : #define SPDK_BS_MAX_DESC_SIZE SPDK_SIZEOF_MEMBER(struct spdk_blob_md_page, descriptors)
393 :
394 : /* Maximum number of extents a single Extent Page can fit.
395 : * For an SPDK_BS_PAGE_SIZE of 4K SPDK_EXTENTS_PER_EP would be 512. */
396 : #define SPDK_EXTENTS_PER_EP_MAX ((SPDK_BS_MAX_DESC_SIZE - sizeof(struct spdk_blob_md_descriptor_extent_page)) / sizeof(uint32_t))
397 : #define SPDK_EXTENTS_PER_EP (spdk_align64pow2(SPDK_EXTENTS_PER_EP_MAX + 1) >> 1u)
398 :
399 : #define SPDK_BS_SUPER_BLOCK_SIG "SPDKBLOB"
400 :
401 : struct spdk_bs_super_block {
402 : uint8_t signature[8];
403 : uint32_t version;
404 : uint32_t length;
405 : uint32_t clean; /* If there was a clean shutdown, this is 1. */
406 : spdk_blob_id super_blob;
407 :
408 : uint32_t cluster_size; /* In bytes */
409 :
410 : uint32_t used_page_mask_start; /* Offset from beginning of disk, in pages */
411 : uint32_t used_page_mask_len; /* Count, in pages */
412 :
413 : uint32_t used_cluster_mask_start; /* Offset from beginning of disk, in pages */
414 : uint32_t used_cluster_mask_len; /* Count, in pages */
415 :
416 : uint32_t md_start; /* Offset from beginning of disk, in pages */
417 : uint32_t md_len; /* Count, in pages */
418 :
419 : struct spdk_bs_type bstype; /* blobstore type */
420 :
421 : uint32_t used_blobid_mask_start; /* Offset from beginning of disk, in pages */
422 : uint32_t used_blobid_mask_len; /* Count, in pages */
423 :
424 : uint64_t size; /* size of blobstore in bytes */
425 : uint32_t io_unit_size; /* Size of io unit in bytes */
426 :
427 : uint8_t reserved[4000];
428 : uint32_t crc;
429 : };
430 : SPDK_STATIC_ASSERT(sizeof(struct spdk_bs_super_block) == 0x1000, "Invalid super block size");
431 :
432 : #pragma pack(pop)
433 :
434 : struct spdk_bs_dev *bs_create_zeroes_dev(void);
435 : struct spdk_bs_dev *bs_create_blob_bs_dev(struct spdk_blob *blob);
436 : struct spdk_io_channel *blob_esnap_get_io_channel(struct spdk_io_channel *ch,
437 : struct spdk_blob *blob);
438 : bool blob_backed_with_zeroes_dev(struct spdk_blob *blob);
439 :
440 : /* Unit Conversions
441 : *
442 : * The blobstore works with several different units:
443 : * - Byte: Self explanatory
444 : * - LBA: The logical blocks on the backing storage device.
445 : * - Page: The read/write units of blobs and metadata. This is
446 : * an offset into a blob in units of 4KiB.
447 : * - Cluster Index: The disk is broken into a sequential list of
448 : * clusters. This is the offset from the beginning.
449 : *
450 : * NOTE: These conversions all act on simple magnitudes, not with any sort
451 : * of knowledge about the blobs themselves. For instance, converting
452 : * a page to an lba with the conversion function below simply converts
453 : * a number of pages to an equivalent number of lbas, but that
454 : * lba certainly isn't the right lba that corresponds to a page offset
455 : * for a particular blob.
456 : */
457 : static inline uint64_t
458 32388 : bs_byte_to_lba(struct spdk_blob_store *bs, uint64_t length)
459 : {
460 32388 : assert(length % bs->dev->blocklen == 0);
461 :
462 32388 : return length / bs->dev->blocklen;
463 : }
464 :
465 : static inline uint64_t
466 2660 : bs_dev_byte_to_lba(struct spdk_bs_dev *bs_dev, uint64_t length)
467 : {
468 2660 : assert(length % bs_dev->blocklen == 0);
469 :
470 2660 : return length / bs_dev->blocklen;
471 : }
472 :
473 : static inline uint64_t
474 26946 : bs_page_to_lba(struct spdk_blob_store *bs, uint64_t page)
475 : {
476 26946 : return page * SPDK_BS_PAGE_SIZE / bs->dev->blocklen;
477 : }
478 :
479 : static inline uint64_t
480 18438 : bs_md_page_to_lba(struct spdk_blob_store *bs, uint32_t page)
481 : {
482 18438 : assert(page < bs->md_len);
483 18438 : return bs_page_to_lba(bs, page + bs->md_start);
484 : }
485 :
486 : static inline uint64_t
487 2710 : bs_dev_page_to_lba(struct spdk_bs_dev *bs_dev, uint64_t page)
488 : {
489 2710 : return page * SPDK_BS_PAGE_SIZE / bs_dev->blocklen;
490 : }
491 :
492 : static inline uint64_t
493 161248 : bs_io_unit_per_page(struct spdk_blob_store *bs)
494 : {
495 161248 : return SPDK_BS_PAGE_SIZE / bs->io_unit_size;
496 : }
497 :
498 : static inline uint64_t
499 79830 : bs_io_unit_to_page(struct spdk_blob_store *bs, uint64_t io_unit)
500 : {
501 79830 : return io_unit / bs_io_unit_per_page(bs);
502 : }
503 :
504 : static inline uint64_t
505 28 : bs_cluster_to_page(struct spdk_blob_store *bs, uint32_t cluster)
506 : {
507 28 : return (uint64_t)cluster * bs->pages_per_cluster;
508 : }
509 :
510 : static inline uint32_t
511 412 : bs_page_to_cluster(struct spdk_blob_store *bs, uint64_t page)
512 : {
513 412 : assert(page % bs->pages_per_cluster == 0);
514 :
515 412 : return page / bs->pages_per_cluster;
516 : }
517 :
518 : static inline uint64_t
519 1136028 : bs_cluster_to_lba(struct spdk_blob_store *bs, uint32_t cluster)
520 : {
521 1136028 : assert(bs->cluster_sz / bs->dev->blocklen > 0);
522 :
523 1136028 : return (uint64_t)cluster * (bs->cluster_sz / bs->dev->blocklen);
524 : }
525 :
526 : static inline uint32_t
527 1069563 : bs_lba_to_cluster(struct spdk_blob_store *bs, uint64_t lba)
528 : {
529 1069563 : assert(lba % (bs->cluster_sz / bs->dev->blocklen) == 0);
530 :
531 1069563 : return lba / (bs->cluster_sz / bs->dev->blocklen);
532 : }
533 :
534 : static inline uint64_t
535 6480 : bs_io_unit_to_back_dev_lba(struct spdk_blob *blob, uint64_t io_unit)
536 : {
537 6480 : return io_unit * (blob->bs->io_unit_size / blob->back_bs_dev->blocklen);
538 : }
539 :
540 : static inline uint64_t
541 4722 : bs_cluster_to_extent_table_id(uint64_t cluster_num)
542 : {
543 4722 : return cluster_num / SPDK_EXTENTS_PER_EP;
544 : }
545 :
546 : static inline uint32_t *
547 4722 : bs_cluster_to_extent_page(struct spdk_blob *blob, uint64_t cluster_num)
548 : {
549 4722 : uint64_t extent_table_id = bs_cluster_to_extent_table_id(cluster_num);
550 :
551 4722 : assert(blob->use_extent_table);
552 4722 : assert(extent_table_id < blob->active.extent_pages_array_size);
553 :
554 4722 : return &blob->active.extent_pages[extent_table_id];
555 : }
556 :
557 : static inline uint64_t
558 42464 : bs_io_units_per_cluster(struct spdk_blob *blob)
559 : {
560 : uint64_t io_units_per_cluster;
561 42464 : uint8_t shift = blob->bs->pages_per_cluster_shift;
562 :
563 42464 : if (shift != 0) {
564 42464 : io_units_per_cluster = bs_io_unit_per_page(blob->bs) << shift;
565 : } else {
566 0 : io_units_per_cluster = bs_io_unit_per_page(blob->bs) * blob->bs->pages_per_cluster;
567 : }
568 :
569 42464 : return io_units_per_cluster;
570 : }
571 :
572 : /* End basic conversions */
573 :
574 : static inline uint64_t
575 30645 : bs_blobid_to_page(spdk_blob_id id)
576 : {
577 30645 : return id & 0xFFFFFFFF;
578 : }
579 :
580 : /* The blob id is a 64 bit number. The lower 32 bits are the page_idx. The upper
581 : * 32 bits are not currently used. Stick a 1 there just to catch bugs where the
582 : * code assumes blob id == page_idx.
583 : */
584 : static inline spdk_blob_id
585 2518 : bs_page_to_blobid(uint64_t page_idx)
586 : {
587 2518 : if (page_idx > UINT32_MAX) {
588 0 : return SPDK_BLOBID_INVALID;
589 : }
590 2518 : return SPDK_BLOB_BLOBID_HIGH_BIT | page_idx;
591 : }
592 :
593 : /* Given an io unit offset into a blob, look up the LBA for the
594 : * start of that io unit.
595 : */
596 : static inline uint64_t
597 37362 : bs_blob_io_unit_to_lba(struct spdk_blob *blob, uint64_t io_unit)
598 : {
599 : uint64_t lba;
600 : uint64_t pages_per_cluster;
601 : uint8_t shift;
602 : uint64_t io_units_per_cluster;
603 : uint64_t io_units_per_page;
604 : uint64_t page;
605 :
606 37362 : page = bs_io_unit_to_page(blob->bs, io_unit);
607 :
608 37362 : pages_per_cluster = blob->bs->pages_per_cluster;
609 37362 : shift = blob->bs->pages_per_cluster_shift;
610 37362 : io_units_per_page = bs_io_unit_per_page(blob->bs);
611 :
612 37362 : assert(page < blob->active.num_clusters * pages_per_cluster);
613 :
614 37362 : if (shift != 0) {
615 37362 : io_units_per_cluster = io_units_per_page << shift;
616 37362 : lba = blob->active.clusters[page >> shift];
617 : } else {
618 0 : io_units_per_cluster = io_units_per_page * pages_per_cluster;
619 0 : lba = blob->active.clusters[page / pages_per_cluster];
620 : }
621 37362 : lba += io_unit % io_units_per_cluster;
622 37362 : return lba;
623 : }
624 :
625 : /* Given an io_unit offset into a blob, look up the number of io_units until the
626 : * next cluster boundary.
627 : */
628 : static inline uint32_t
629 42324 : bs_num_io_units_to_cluster_boundary(struct spdk_blob *blob, uint64_t io_unit)
630 : {
631 : uint64_t io_units_per_cluster;
632 :
633 42324 : io_units_per_cluster = bs_io_units_per_cluster(blob);
634 :
635 42324 : return io_units_per_cluster - (io_unit % io_units_per_cluster);
636 : }
637 :
638 : /* Given an io_unit offset into a blob, look up the number of pages into blob to beginning of current cluster */
639 : static inline uint64_t
640 880 : bs_io_unit_to_cluster_start(struct spdk_blob *blob, uint64_t io_unit)
641 : {
642 : uint64_t pages_per_cluster;
643 : uint64_t page;
644 :
645 880 : pages_per_cluster = blob->bs->pages_per_cluster;
646 880 : page = bs_io_unit_to_page(blob->bs, io_unit);
647 :
648 880 : return page - (page % pages_per_cluster);
649 : }
650 :
651 : /* Given an io_unit offset into a blob, look up the number of pages into blob to beginning of current cluster */
652 : static inline uint32_t
653 880 : bs_io_unit_to_cluster_number(struct spdk_blob *blob, uint64_t io_unit)
654 : {
655 880 : uint64_t pages_per_cluster = blob->bs->pages_per_cluster;
656 880 : uint8_t shift = blob->bs->pages_per_cluster_shift;
657 : uint64_t page_offset;
658 :
659 880 : page_offset = io_unit / bs_io_unit_per_page(blob->bs);
660 880 : if (shift != 0) {
661 880 : return page_offset >> shift;
662 : } else {
663 0 : return page_offset / pages_per_cluster;
664 : }
665 : }
666 :
667 : /* Given an io unit offset into a blob, look up if it is from allocated cluster. */
668 : static inline bool
669 41010 : bs_io_unit_is_allocated(struct spdk_blob *blob, uint64_t io_unit)
670 : {
671 : uint64_t lba;
672 : uint64_t page;
673 : uint64_t pages_per_cluster;
674 : uint8_t shift;
675 :
676 41010 : shift = blob->bs->pages_per_cluster_shift;
677 41010 : pages_per_cluster = blob->bs->pages_per_cluster;
678 41010 : page = bs_io_unit_to_page(blob->bs, io_unit);
679 :
680 41010 : assert(page < blob->active.num_clusters * pages_per_cluster);
681 :
682 41010 : if (shift != 0) {
683 41010 : lba = blob->active.clusters[page >> shift];
684 : } else {
685 0 : lba = blob->active.clusters[page / pages_per_cluster];
686 : }
687 :
688 41010 : if (lba == 0) {
689 3364 : assert(spdk_blob_is_thin_provisioned(blob));
690 3364 : return false;
691 : } else {
692 37646 : return true;
693 : }
694 : }
695 :
696 : #endif
|
