xref: /spdk/lib/reduce/reduce.c (revision ae7b5890ef728af40bd233a5011b924c482603bf)
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/reduce.h"
37 #include "spdk/env.h"
38 #include "spdk/string.h"
39 #include "spdk/bit_array.h"
40 #include "spdk/util.h"
41 #include "spdk_internal/log.h"
42 
43 #include "libpmem.h"
44 
45 /* Always round up the size of the PM region to the nearest cacheline. */
46 #define REDUCE_PM_SIZE_ALIGNMENT	64
47 
48 /* Offset into the backing device where the persistent memory file's path is stored. */
49 #define REDUCE_BACKING_DEV_PATH_OFFSET	4096
50 
51 #define REDUCE_EMPTY_MAP_ENTRY	-1ULL
52 
53 #define REDUCE_NUM_VOL_REQUESTS	256
54 
55 /* Structure written to offset 0 of both the pm file and the backing device. */
56 struct spdk_reduce_vol_superblock {
57 	uint8_t				signature[8];
58 	struct spdk_reduce_vol_params	params;
59 	uint8_t				reserved[4048];
60 };
61 SPDK_STATIC_ASSERT(sizeof(struct spdk_reduce_vol_superblock) == 4096, "size incorrect");
62 
63 #define SPDK_REDUCE_SIGNATURE "SPDKREDU"
64 /* null terminator counts one */
65 SPDK_STATIC_ASSERT(sizeof(SPDK_REDUCE_SIGNATURE) - 1 ==
66 		   sizeof(((struct spdk_reduce_vol_superblock *)0)->signature), "size incorrect");
67 
68 #define REDUCE_PATH_MAX 4096
69 
70 /**
71  * Describes a persistent memory file used to hold metadata associated with a
72  *  compressed volume.
73  */
74 struct spdk_reduce_pm_file {
75 	char			path[REDUCE_PATH_MAX];
76 	void			*pm_buf;
77 	int			pm_is_pmem;
78 	uint64_t		size;
79 };
80 
81 #define REDUCE_IO_READV		1
82 #define REDUCE_IO_WRITEV	2
83 
84 struct spdk_reduce_chunk_map {
85 	uint32_t		compressed_size;
86 	uint32_t		reserved;
87 	uint64_t		io_unit_index[0];
88 };
89 
90 struct spdk_reduce_vol_request {
91 	/**
92 	 *  Scratch buffer used for uncompressed chunk.  This is used for:
93 	 *   1) source buffer for compression operations
94 	 *   2) destination buffer for decompression operations
95 	 *   3) data buffer when writing uncompressed chunk to disk
96 	 *   4) data buffer when reading uncompressed chunk from disk
97 	 */
98 	uint8_t					*decomp_buf;
99 	struct iovec				*decomp_buf_iov;
100 
101 	/**
102 	 * These are used to construct the iovecs that are sent to
103 	 *  the decomp engine, they point to a mix of the scratch buffer
104 	 *  and user buffer
105 	 */
106 	struct iovec				decomp_iov[REDUCE_MAX_IOVECS];
107 	int					decomp_iovcnt;
108 
109 	/**
110 	 *  Scratch buffer used for compressed chunk.  This is used for:
111 	 *   1) destination buffer for compression operations
112 	 *   2) source buffer for decompression operations
113 	 *   3) data buffer when writing compressed chunk to disk
114 	 *   4) data buffer when reading compressed chunk from disk
115 	 */
116 	uint8_t					*comp_buf;
117 	struct iovec				*comp_buf_iov;
118 	struct iovec				*iov;
119 	bool					rmw;
120 	struct spdk_reduce_vol			*vol;
121 	int					type;
122 	int					reduce_errno;
123 	int					iovcnt;
124 	int					num_backing_ops;
125 	uint32_t				num_io_units;
126 	bool					chunk_is_compressed;
127 	uint64_t				offset;
128 	uint64_t				logical_map_index;
129 	uint64_t				length;
130 	uint64_t				chunk_map_index;
131 	struct spdk_reduce_chunk_map		*chunk;
132 	spdk_reduce_vol_op_complete		cb_fn;
133 	void					*cb_arg;
134 	TAILQ_ENTRY(spdk_reduce_vol_request)	tailq;
135 	struct spdk_reduce_vol_cb_args		backing_cb_args;
136 };
137 
138 struct spdk_reduce_vol {
139 	struct spdk_reduce_vol_params		params;
140 	uint32_t				backing_io_units_per_chunk;
141 	uint32_t				backing_lba_per_io_unit;
142 	uint32_t				logical_blocks_per_chunk;
143 	struct spdk_reduce_pm_file		pm_file;
144 	struct spdk_reduce_backing_dev		*backing_dev;
145 	struct spdk_reduce_vol_superblock	*backing_super;
146 	struct spdk_reduce_vol_superblock	*pm_super;
147 	uint64_t				*pm_logical_map;
148 	uint64_t				*pm_chunk_maps;
149 
150 	struct spdk_bit_array			*allocated_chunk_maps;
151 	struct spdk_bit_array			*allocated_backing_io_units;
152 
153 	struct spdk_reduce_vol_request		*request_mem;
154 	TAILQ_HEAD(, spdk_reduce_vol_request)	free_requests;
155 	TAILQ_HEAD(, spdk_reduce_vol_request)	executing_requests;
156 	TAILQ_HEAD(, spdk_reduce_vol_request)	queued_requests;
157 
158 	/* Single contiguous buffer used for all request buffers for this volume. */
159 	uint8_t					*buf_mem;
160 	struct iovec				*buf_iov_mem;
161 };
162 
163 static void _start_readv_request(struct spdk_reduce_vol_request *req);
164 static void _start_writev_request(struct spdk_reduce_vol_request *req);
165 static uint8_t *g_zero_buf;
166 static int g_vol_count = 0;
167 
168 /*
169  * Allocate extra metadata chunks and corresponding backing io units to account for
170  *  outstanding IO in worst case scenario where logical map is completely allocated
171  *  and no data can be compressed.  We need extra chunks in this case to handle
172  *  in-flight writes since reduce never writes data in place.
173  */
174 #define REDUCE_NUM_EXTRA_CHUNKS 128
175 
176 static void
177 _reduce_persist(struct spdk_reduce_vol *vol, const void *addr, size_t len)
178 {
179 	if (vol->pm_file.pm_is_pmem) {
180 		pmem_persist(addr, len);
181 	} else {
182 		pmem_msync(addr, len);
183 	}
184 }
185 
186 static uint64_t
187 _get_pm_logical_map_size(uint64_t vol_size, uint64_t chunk_size)
188 {
189 	uint64_t chunks_in_logical_map, logical_map_size;
190 
191 	chunks_in_logical_map = vol_size / chunk_size;
192 	logical_map_size = chunks_in_logical_map * sizeof(uint64_t);
193 
194 	/* Round up to next cacheline. */
195 	return spdk_divide_round_up(logical_map_size, REDUCE_PM_SIZE_ALIGNMENT) *
196 	       REDUCE_PM_SIZE_ALIGNMENT;
197 }
198 
199 static uint64_t
200 _get_total_chunks(uint64_t vol_size, uint64_t chunk_size)
201 {
202 	uint64_t num_chunks;
203 
204 	num_chunks = vol_size / chunk_size;
205 	num_chunks += REDUCE_NUM_EXTRA_CHUNKS;
206 
207 	return num_chunks;
208 }
209 
210 static inline uint32_t
211 _reduce_vol_get_chunk_struct_size(uint64_t backing_io_units_per_chunk)
212 {
213 	return sizeof(struct spdk_reduce_chunk_map) + sizeof(uint64_t) * backing_io_units_per_chunk;
214 }
215 
216 static uint64_t
217 _get_pm_total_chunks_size(uint64_t vol_size, uint64_t chunk_size, uint64_t backing_io_unit_size)
218 {
219 	uint64_t io_units_per_chunk, num_chunks, total_chunks_size;
220 
221 	num_chunks = _get_total_chunks(vol_size, chunk_size);
222 	io_units_per_chunk = chunk_size / backing_io_unit_size;
223 
224 	total_chunks_size = num_chunks * _reduce_vol_get_chunk_struct_size(io_units_per_chunk);
225 
226 	return spdk_divide_round_up(total_chunks_size, REDUCE_PM_SIZE_ALIGNMENT) *
227 	       REDUCE_PM_SIZE_ALIGNMENT;
228 }
229 
230 static struct spdk_reduce_chunk_map *
231 _reduce_vol_get_chunk_map(struct spdk_reduce_vol *vol, uint64_t chunk_map_index)
232 {
233 	uintptr_t chunk_map_addr;
234 
235 	assert(chunk_map_index < _get_total_chunks(vol->params.vol_size, vol->params.chunk_size));
236 
237 	chunk_map_addr = (uintptr_t)vol->pm_chunk_maps;
238 	chunk_map_addr += chunk_map_index *
239 			  _reduce_vol_get_chunk_struct_size(vol->backing_io_units_per_chunk);
240 
241 	return (struct spdk_reduce_chunk_map *)chunk_map_addr;
242 }
243 
244 static int
245 _validate_vol_params(struct spdk_reduce_vol_params *params)
246 {
247 	if (params->vol_size > 0) {
248 		/**
249 		 * User does not pass in the vol size - it gets calculated by libreduce from
250 		 *  values in this structure plus the size of the backing device.
251 		 */
252 		return -EINVAL;
253 	}
254 
255 	if (params->chunk_size == 0 || params->backing_io_unit_size == 0 ||
256 	    params->logical_block_size == 0) {
257 		return -EINVAL;
258 	}
259 
260 	/* Chunk size must be an even multiple of the backing io unit size. */
261 	if ((params->chunk_size % params->backing_io_unit_size) != 0) {
262 		return -EINVAL;
263 	}
264 
265 	/* Chunk size must be an even multiple of the logical block size. */
266 	if ((params->chunk_size % params->logical_block_size) != 0) {
267 		return -1;
268 	}
269 
270 	return 0;
271 }
272 
273 static uint64_t
274 _get_vol_size(uint64_t chunk_size, uint64_t backing_dev_size)
275 {
276 	uint64_t num_chunks;
277 
278 	num_chunks = backing_dev_size / chunk_size;
279 	if (num_chunks <= REDUCE_NUM_EXTRA_CHUNKS) {
280 		return 0;
281 	}
282 
283 	num_chunks -= REDUCE_NUM_EXTRA_CHUNKS;
284 	return num_chunks * chunk_size;
285 }
286 
287 static uint64_t
288 _get_pm_file_size(struct spdk_reduce_vol_params *params)
289 {
290 	uint64_t total_pm_size;
291 
292 	total_pm_size = sizeof(struct spdk_reduce_vol_superblock);
293 	total_pm_size += _get_pm_logical_map_size(params->vol_size, params->chunk_size);
294 	total_pm_size += _get_pm_total_chunks_size(params->vol_size, params->chunk_size,
295 			 params->backing_io_unit_size);
296 	return total_pm_size;
297 }
298 
299 const struct spdk_uuid *
300 spdk_reduce_vol_get_uuid(struct spdk_reduce_vol *vol)
301 {
302 	return &vol->params.uuid;
303 }
304 
305 static void
306 _initialize_vol_pm_pointers(struct spdk_reduce_vol *vol)
307 {
308 	uint64_t logical_map_size;
309 
310 	/* Superblock is at the beginning of the pm file. */
311 	vol->pm_super = (struct spdk_reduce_vol_superblock *)vol->pm_file.pm_buf;
312 
313 	/* Logical map immediately follows the super block. */
314 	vol->pm_logical_map = (uint64_t *)(vol->pm_super + 1);
315 
316 	/* Chunks maps follow the logical map. */
317 	logical_map_size = _get_pm_logical_map_size(vol->params.vol_size, vol->params.chunk_size);
318 	vol->pm_chunk_maps = (uint64_t *)((uint8_t *)vol->pm_logical_map + logical_map_size);
319 }
320 
321 /* We need 2 iovs during load - one for the superblock, another for the path */
322 #define LOAD_IOV_COUNT	2
323 
324 struct reduce_init_load_ctx {
325 	struct spdk_reduce_vol			*vol;
326 	struct spdk_reduce_vol_cb_args		backing_cb_args;
327 	spdk_reduce_vol_op_with_handle_complete	cb_fn;
328 	void					*cb_arg;
329 	struct iovec				iov[LOAD_IOV_COUNT];
330 	void					*path;
331 };
332 
333 static int
334 _allocate_vol_requests(struct spdk_reduce_vol *vol)
335 {
336 	struct spdk_reduce_vol_request *req;
337 	int i;
338 
339 	/* Allocate 2x since we need buffers for both read/write and compress/decompress
340 	 *  intermediate buffers.
341 	 */
342 	vol->buf_mem = spdk_malloc(2 * REDUCE_NUM_VOL_REQUESTS * vol->params.chunk_size,
343 				   64, NULL, SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
344 	if (vol->buf_mem == NULL) {
345 		return -ENOMEM;
346 	}
347 
348 	vol->request_mem = calloc(REDUCE_NUM_VOL_REQUESTS, sizeof(*req));
349 	if (vol->request_mem == NULL) {
350 		spdk_free(vol->buf_mem);
351 		vol->buf_mem = NULL;
352 		return -ENOMEM;
353 	}
354 
355 	/* Allocate 2x since we need iovs for both read/write and compress/decompress intermediate
356 	 *  buffers.
357 	 */
358 	vol->buf_iov_mem = calloc(REDUCE_NUM_VOL_REQUESTS,
359 				  2 * sizeof(struct iovec) * vol->backing_io_units_per_chunk);
360 	if (vol->buf_iov_mem == NULL) {
361 		free(vol->request_mem);
362 		spdk_free(vol->buf_mem);
363 		vol->request_mem = NULL;
364 		vol->buf_mem = NULL;
365 		return -ENOMEM;
366 	}
367 
368 	for (i = 0; i < REDUCE_NUM_VOL_REQUESTS; i++) {
369 		req = &vol->request_mem[i];
370 		TAILQ_INSERT_HEAD(&vol->free_requests, req, tailq);
371 		req->decomp_buf_iov = &vol->buf_iov_mem[(2 * i) * vol->backing_io_units_per_chunk];
372 		req->decomp_buf = vol->buf_mem + (2 * i) * vol->params.chunk_size;
373 		req->comp_buf_iov = &vol->buf_iov_mem[(2 * i + 1) * vol->backing_io_units_per_chunk];
374 		req->comp_buf = vol->buf_mem + (2 * i + 1) * vol->params.chunk_size;
375 	}
376 
377 	return 0;
378 }
379 
380 static void
381 _init_load_cleanup(struct spdk_reduce_vol *vol, struct reduce_init_load_ctx *ctx)
382 {
383 	if (ctx != NULL) {
384 		spdk_free(ctx->path);
385 		free(ctx);
386 	}
387 
388 	if (vol != NULL) {
389 		if (vol->pm_file.pm_buf != NULL) {
390 			pmem_unmap(vol->pm_file.pm_buf, vol->pm_file.size);
391 		}
392 
393 		spdk_free(vol->backing_super);
394 		spdk_bit_array_free(&vol->allocated_chunk_maps);
395 		spdk_bit_array_free(&vol->allocated_backing_io_units);
396 		free(vol->request_mem);
397 		free(vol->buf_iov_mem);
398 		spdk_free(vol->buf_mem);
399 		free(vol);
400 	}
401 }
402 
403 static int
404 _alloc_zero_buff(struct spdk_reduce_vol *vol)
405 {
406 	int rc = 0;
407 
408 	/* The zero buffer is shared between all volumnes and just used
409 	 * for reads so allocate one global instance here if not already
410 	 * allocated when another vol init'd or loaded.
411 	 */
412 	if (g_vol_count++ == 0) {
413 		g_zero_buf = spdk_zmalloc(vol->params.chunk_size,
414 					  64, NULL, SPDK_ENV_LCORE_ID_ANY,
415 					  SPDK_MALLOC_DMA);
416 		if (g_zero_buf == NULL) {
417 			rc = -ENOMEM;
418 		}
419 	}
420 	return rc;
421 }
422 
423 static void
424 _init_write_super_cpl(void *cb_arg, int reduce_errno)
425 {
426 	struct reduce_init_load_ctx *init_ctx = cb_arg;
427 	int rc;
428 
429 	rc = _allocate_vol_requests(init_ctx->vol);
430 	if (rc != 0) {
431 		init_ctx->cb_fn(init_ctx->cb_arg, NULL, rc);
432 		_init_load_cleanup(init_ctx->vol, init_ctx);
433 		return;
434 	}
435 
436 	rc = _alloc_zero_buff(init_ctx->vol);
437 	if (rc != 0) {
438 		init_ctx->cb_fn(init_ctx->cb_arg, NULL, rc);
439 		_init_load_cleanup(init_ctx->vol, init_ctx);
440 		return;
441 	}
442 
443 	init_ctx->cb_fn(init_ctx->cb_arg, init_ctx->vol, reduce_errno);
444 	/* Only clean up the ctx - the vol has been passed to the application
445 	 *  for use now that initialization was successful.
446 	 */
447 	_init_load_cleanup(NULL, init_ctx);
448 }
449 
450 static void
451 _init_write_path_cpl(void *cb_arg, int reduce_errno)
452 {
453 	struct reduce_init_load_ctx *init_ctx = cb_arg;
454 	struct spdk_reduce_vol *vol = init_ctx->vol;
455 
456 	init_ctx->iov[0].iov_base = vol->backing_super;
457 	init_ctx->iov[0].iov_len = sizeof(*vol->backing_super);
458 	init_ctx->backing_cb_args.cb_fn = _init_write_super_cpl;
459 	init_ctx->backing_cb_args.cb_arg = init_ctx;
460 	vol->backing_dev->writev(vol->backing_dev, init_ctx->iov, 1,
461 				 0, sizeof(*vol->backing_super) / vol->backing_dev->blocklen,
462 				 &init_ctx->backing_cb_args);
463 }
464 
465 static int
466 _allocate_bit_arrays(struct spdk_reduce_vol *vol)
467 {
468 	uint64_t total_chunks, total_backing_io_units;
469 	uint32_t i, num_metadata_io_units;
470 
471 	total_chunks = _get_total_chunks(vol->params.vol_size, vol->params.chunk_size);
472 	vol->allocated_chunk_maps = spdk_bit_array_create(total_chunks);
473 	total_backing_io_units = total_chunks * (vol->params.chunk_size / vol->params.backing_io_unit_size);
474 	vol->allocated_backing_io_units = spdk_bit_array_create(total_backing_io_units);
475 
476 	if (vol->allocated_chunk_maps == NULL || vol->allocated_backing_io_units == NULL) {
477 		return -ENOMEM;
478 	}
479 
480 	/* Set backing io unit bits associated with metadata. */
481 	num_metadata_io_units = (sizeof(*vol->backing_super) + REDUCE_PATH_MAX) /
482 				vol->backing_dev->blocklen;
483 	for (i = 0; i < num_metadata_io_units; i++) {
484 		spdk_bit_array_set(vol->allocated_backing_io_units, i);
485 	}
486 
487 	return 0;
488 }
489 
490 void
491 spdk_reduce_vol_init(struct spdk_reduce_vol_params *params,
492 		     struct spdk_reduce_backing_dev *backing_dev,
493 		     const char *pm_file_dir,
494 		     spdk_reduce_vol_op_with_handle_complete cb_fn, void *cb_arg)
495 {
496 	struct spdk_reduce_vol *vol;
497 	struct reduce_init_load_ctx *init_ctx;
498 	uint64_t backing_dev_size;
499 	size_t mapped_len;
500 	int dir_len, max_dir_len, rc;
501 
502 	/* We need to append a path separator and the UUID to the supplied
503 	 * path.
504 	 */
505 	max_dir_len = REDUCE_PATH_MAX - SPDK_UUID_STRING_LEN - 1;
506 	dir_len = strnlen(pm_file_dir, max_dir_len);
507 	/* Strip trailing slash if the user provided one - we will add it back
508 	 * later when appending the filename.
509 	 */
510 	if (pm_file_dir[dir_len - 1] == '/') {
511 		dir_len--;
512 	}
513 	if (dir_len == max_dir_len) {
514 		SPDK_ERRLOG("pm_file_dir (%s) too long\n", pm_file_dir);
515 		cb_fn(cb_arg, NULL, -EINVAL);
516 		return;
517 	}
518 
519 	rc = _validate_vol_params(params);
520 	if (rc != 0) {
521 		SPDK_ERRLOG("invalid vol params\n");
522 		cb_fn(cb_arg, NULL, rc);
523 		return;
524 	}
525 
526 	backing_dev_size = backing_dev->blockcnt * backing_dev->blocklen;
527 	params->vol_size = _get_vol_size(params->chunk_size, backing_dev_size);
528 	if (params->vol_size == 0) {
529 		SPDK_ERRLOG("backing device is too small\n");
530 		cb_fn(cb_arg, NULL, -EINVAL);
531 		return;
532 	}
533 
534 	if (backing_dev->readv == NULL || backing_dev->writev == NULL ||
535 	    backing_dev->unmap == NULL) {
536 		SPDK_ERRLOG("backing_dev function pointer not specified\n");
537 		cb_fn(cb_arg, NULL, -EINVAL);
538 		return;
539 	}
540 
541 	vol = calloc(1, sizeof(*vol));
542 	if (vol == NULL) {
543 		cb_fn(cb_arg, NULL, -ENOMEM);
544 		return;
545 	}
546 
547 	TAILQ_INIT(&vol->free_requests);
548 	TAILQ_INIT(&vol->executing_requests);
549 	TAILQ_INIT(&vol->queued_requests);
550 
551 	vol->backing_super = spdk_zmalloc(sizeof(*vol->backing_super), 0, NULL,
552 					  SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
553 	if (vol->backing_super == NULL) {
554 		cb_fn(cb_arg, NULL, -ENOMEM);
555 		_init_load_cleanup(vol, NULL);
556 		return;
557 	}
558 
559 	init_ctx = calloc(1, sizeof(*init_ctx));
560 	if (init_ctx == NULL) {
561 		cb_fn(cb_arg, NULL, -ENOMEM);
562 		_init_load_cleanup(vol, NULL);
563 		return;
564 	}
565 
566 	init_ctx->path = spdk_zmalloc(REDUCE_PATH_MAX, 0, NULL,
567 				      SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
568 	if (init_ctx->path == NULL) {
569 		cb_fn(cb_arg, NULL, -ENOMEM);
570 		_init_load_cleanup(vol, init_ctx);
571 		return;
572 	}
573 
574 	if (spdk_mem_all_zero(&params->uuid, sizeof(params->uuid))) {
575 		spdk_uuid_generate(&params->uuid);
576 	}
577 
578 	memcpy(vol->pm_file.path, pm_file_dir, dir_len);
579 	vol->pm_file.path[dir_len] = '/';
580 	spdk_uuid_fmt_lower(&vol->pm_file.path[dir_len + 1], SPDK_UUID_STRING_LEN,
581 			    &params->uuid);
582 	vol->pm_file.size = _get_pm_file_size(params);
583 	vol->pm_file.pm_buf = pmem_map_file(vol->pm_file.path, vol->pm_file.size,
584 					    PMEM_FILE_CREATE | PMEM_FILE_EXCL, 0600,
585 					    &mapped_len, &vol->pm_file.pm_is_pmem);
586 	if (vol->pm_file.pm_buf == NULL) {
587 		SPDK_ERRLOG("could not pmem_map_file(%s): %s\n",
588 			    vol->pm_file.path, strerror(errno));
589 		cb_fn(cb_arg, NULL, -errno);
590 		_init_load_cleanup(vol, init_ctx);
591 		return;
592 	}
593 
594 	if (vol->pm_file.size != mapped_len) {
595 		SPDK_ERRLOG("could not map entire pmem file (size=%" PRIu64 " mapped=%" PRIu64 ")\n",
596 			    vol->pm_file.size, mapped_len);
597 		cb_fn(cb_arg, NULL, -ENOMEM);
598 		_init_load_cleanup(vol, init_ctx);
599 		return;
600 	}
601 
602 	vol->backing_io_units_per_chunk = params->chunk_size / params->backing_io_unit_size;
603 	vol->logical_blocks_per_chunk = params->chunk_size / params->logical_block_size;
604 	vol->backing_lba_per_io_unit = params->backing_io_unit_size / backing_dev->blocklen;
605 	memcpy(&vol->params, params, sizeof(*params));
606 
607 	vol->backing_dev = backing_dev;
608 
609 	rc = _allocate_bit_arrays(vol);
610 	if (rc != 0) {
611 		cb_fn(cb_arg, NULL, rc);
612 		_init_load_cleanup(vol, init_ctx);
613 		return;
614 	}
615 
616 	memcpy(vol->backing_super->signature, SPDK_REDUCE_SIGNATURE,
617 	       sizeof(vol->backing_super->signature));
618 	memcpy(&vol->backing_super->params, params, sizeof(*params));
619 
620 	_initialize_vol_pm_pointers(vol);
621 
622 	memcpy(vol->pm_super, vol->backing_super, sizeof(*vol->backing_super));
623 	/* Writing 0xFF's is equivalent of filling it all with SPDK_EMPTY_MAP_ENTRY.
624 	 * Note that this writes 0xFF to not just the logical map but the chunk maps as well.
625 	 */
626 	memset(vol->pm_logical_map, 0xFF, vol->pm_file.size - sizeof(*vol->backing_super));
627 	_reduce_persist(vol, vol->pm_file.pm_buf, vol->pm_file.size);
628 
629 	init_ctx->vol = vol;
630 	init_ctx->cb_fn = cb_fn;
631 	init_ctx->cb_arg = cb_arg;
632 
633 	memcpy(init_ctx->path, vol->pm_file.path, REDUCE_PATH_MAX);
634 	init_ctx->iov[0].iov_base = init_ctx->path;
635 	init_ctx->iov[0].iov_len = REDUCE_PATH_MAX;
636 	init_ctx->backing_cb_args.cb_fn = _init_write_path_cpl;
637 	init_ctx->backing_cb_args.cb_arg = init_ctx;
638 	/* Write path to offset 4K on backing device - just after where the super
639 	 *  block will be written.  We wait until this is committed before writing the
640 	 *  super block to guarantee we don't get the super block written without the
641 	 *  the path if the system crashed in the middle of a write operation.
642 	 */
643 	vol->backing_dev->writev(vol->backing_dev, init_ctx->iov, 1,
644 				 REDUCE_BACKING_DEV_PATH_OFFSET / vol->backing_dev->blocklen,
645 				 REDUCE_PATH_MAX / vol->backing_dev->blocklen,
646 				 &init_ctx->backing_cb_args);
647 }
648 
649 static void
650 _load_read_super_and_path_cpl(void *cb_arg, int reduce_errno)
651 {
652 	struct reduce_init_load_ctx *load_ctx = cb_arg;
653 	struct spdk_reduce_vol *vol = load_ctx->vol;
654 	uint64_t backing_dev_size;
655 	uint64_t i, num_chunks, logical_map_index;
656 	struct spdk_reduce_chunk_map *chunk;
657 	size_t mapped_len;
658 	uint32_t j;
659 	int rc;
660 
661 	if (memcmp(vol->backing_super->signature,
662 		   SPDK_REDUCE_SIGNATURE,
663 		   sizeof(vol->backing_super->signature)) != 0) {
664 		/* This backing device isn't a libreduce backing device. */
665 		rc = -EILSEQ;
666 		goto error;
667 	}
668 
669 	memcpy(&vol->params, &vol->backing_super->params, sizeof(vol->params));
670 	vol->backing_io_units_per_chunk = vol->params.chunk_size / vol->params.backing_io_unit_size;
671 	vol->logical_blocks_per_chunk = vol->params.chunk_size / vol->params.logical_block_size;
672 	vol->backing_lba_per_io_unit = vol->params.backing_io_unit_size / vol->backing_dev->blocklen;
673 
674 	rc = _allocate_bit_arrays(vol);
675 	if (rc != 0) {
676 		goto error;
677 	}
678 
679 	backing_dev_size = vol->backing_dev->blockcnt * vol->backing_dev->blocklen;
680 	if (_get_vol_size(vol->params.chunk_size, backing_dev_size) < vol->params.vol_size) {
681 		SPDK_ERRLOG("backing device size %" PRIi64 " smaller than expected\n",
682 			    backing_dev_size);
683 		rc = -EILSEQ;
684 		goto error;
685 	}
686 
687 	memcpy(vol->pm_file.path, load_ctx->path, sizeof(vol->pm_file.path));
688 	vol->pm_file.size = _get_pm_file_size(&vol->params);
689 	vol->pm_file.pm_buf = pmem_map_file(vol->pm_file.path, 0, 0, 0, &mapped_len,
690 					    &vol->pm_file.pm_is_pmem);
691 	if (vol->pm_file.pm_buf == NULL) {
692 		SPDK_ERRLOG("could not pmem_map_file(%s): %s\n", vol->pm_file.path, strerror(errno));
693 		rc = -errno;
694 		goto error;
695 	}
696 
697 	if (vol->pm_file.size != mapped_len) {
698 		SPDK_ERRLOG("could not map entire pmem file (size=%" PRIu64 " mapped=%" PRIu64 ")\n",
699 			    vol->pm_file.size, mapped_len);
700 		rc = -ENOMEM;
701 		goto error;
702 	}
703 
704 	rc = _allocate_vol_requests(vol);
705 	if (rc != 0) {
706 		goto error;
707 	}
708 
709 	_initialize_vol_pm_pointers(vol);
710 
711 	num_chunks = vol->params.vol_size / vol->params.chunk_size;
712 	for (i = 0; i < num_chunks; i++) {
713 		logical_map_index = vol->pm_logical_map[i];
714 		if (logical_map_index == REDUCE_EMPTY_MAP_ENTRY) {
715 			continue;
716 		}
717 		spdk_bit_array_set(vol->allocated_chunk_maps, logical_map_index);
718 		chunk = _reduce_vol_get_chunk_map(vol, logical_map_index);
719 		for (j = 0; j < vol->backing_io_units_per_chunk; j++) {
720 			if (chunk->io_unit_index[j] != REDUCE_EMPTY_MAP_ENTRY) {
721 				spdk_bit_array_set(vol->allocated_backing_io_units, chunk->io_unit_index[j]);
722 			}
723 		}
724 	}
725 
726 	rc = _alloc_zero_buff(vol);
727 	if (rc) {
728 		goto error;
729 	}
730 
731 	load_ctx->cb_fn(load_ctx->cb_arg, vol, 0);
732 	/* Only clean up the ctx - the vol has been passed to the application
733 	 *  for use now that volume load was successful.
734 	 */
735 	_init_load_cleanup(NULL, load_ctx);
736 	return;
737 
738 error:
739 	load_ctx->cb_fn(load_ctx->cb_arg, NULL, rc);
740 	_init_load_cleanup(vol, load_ctx);
741 }
742 
743 void
744 spdk_reduce_vol_load(struct spdk_reduce_backing_dev *backing_dev,
745 		     spdk_reduce_vol_op_with_handle_complete cb_fn, void *cb_arg)
746 {
747 	struct spdk_reduce_vol *vol;
748 	struct reduce_init_load_ctx *load_ctx;
749 
750 	if (backing_dev->readv == NULL || backing_dev->writev == NULL ||
751 	    backing_dev->unmap == NULL) {
752 		SPDK_ERRLOG("backing_dev function pointer not specified\n");
753 		cb_fn(cb_arg, NULL, -EINVAL);
754 		return;
755 	}
756 
757 	vol = calloc(1, sizeof(*vol));
758 	if (vol == NULL) {
759 		cb_fn(cb_arg, NULL, -ENOMEM);
760 		return;
761 	}
762 
763 	TAILQ_INIT(&vol->free_requests);
764 	TAILQ_INIT(&vol->executing_requests);
765 	TAILQ_INIT(&vol->queued_requests);
766 
767 	vol->backing_super = spdk_zmalloc(sizeof(*vol->backing_super), 64, NULL,
768 					  SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
769 	if (vol->backing_super == NULL) {
770 		_init_load_cleanup(vol, NULL);
771 		cb_fn(cb_arg, NULL, -ENOMEM);
772 		return;
773 	}
774 
775 	vol->backing_dev = backing_dev;
776 
777 	load_ctx = calloc(1, sizeof(*load_ctx));
778 	if (load_ctx == NULL) {
779 		_init_load_cleanup(vol, NULL);
780 		cb_fn(cb_arg, NULL, -ENOMEM);
781 		return;
782 	}
783 
784 	load_ctx->path = spdk_zmalloc(REDUCE_PATH_MAX, 64, NULL,
785 				      SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
786 	if (load_ctx->path == NULL) {
787 		_init_load_cleanup(vol, load_ctx);
788 		cb_fn(cb_arg, NULL, -ENOMEM);
789 		return;
790 	}
791 
792 	load_ctx->vol = vol;
793 	load_ctx->cb_fn = cb_fn;
794 	load_ctx->cb_arg = cb_arg;
795 
796 	load_ctx->iov[0].iov_base = vol->backing_super;
797 	load_ctx->iov[0].iov_len = sizeof(*vol->backing_super);
798 	load_ctx->iov[1].iov_base = load_ctx->path;
799 	load_ctx->iov[1].iov_len = REDUCE_PATH_MAX;
800 	load_ctx->backing_cb_args.cb_fn = _load_read_super_and_path_cpl;
801 	load_ctx->backing_cb_args.cb_arg = load_ctx;
802 	vol->backing_dev->readv(vol->backing_dev, load_ctx->iov, LOAD_IOV_COUNT, 0,
803 				(sizeof(*vol->backing_super) + REDUCE_PATH_MAX) /
804 				vol->backing_dev->blocklen,
805 				&load_ctx->backing_cb_args);
806 }
807 
808 void
809 spdk_reduce_vol_unload(struct spdk_reduce_vol *vol,
810 		       spdk_reduce_vol_op_complete cb_fn, void *cb_arg)
811 {
812 	if (vol == NULL) {
813 		/* This indicates a programming error. */
814 		assert(false);
815 		cb_fn(cb_arg, -EINVAL);
816 		return;
817 	}
818 
819 	if (--g_vol_count == 0) {
820 		spdk_free(g_zero_buf);
821 	}
822 	_init_load_cleanup(vol, NULL);
823 	cb_fn(cb_arg, 0);
824 }
825 
826 struct reduce_destroy_ctx {
827 	spdk_reduce_vol_op_complete		cb_fn;
828 	void					*cb_arg;
829 	struct spdk_reduce_vol			*vol;
830 	struct spdk_reduce_vol_superblock	*super;
831 	struct iovec				iov;
832 	struct spdk_reduce_vol_cb_args		backing_cb_args;
833 	int					reduce_errno;
834 	char					pm_path[REDUCE_PATH_MAX];
835 };
836 
837 static void
838 destroy_unload_cpl(void *cb_arg, int reduce_errno)
839 {
840 	struct reduce_destroy_ctx *destroy_ctx = cb_arg;
841 
842 	if (destroy_ctx->reduce_errno == 0) {
843 		if (unlink(destroy_ctx->pm_path)) {
844 			SPDK_ERRLOG("%s could not be unlinked: %s\n",
845 				    destroy_ctx->pm_path, strerror(errno));
846 		}
847 	}
848 
849 	/* Even if the unload somehow failed, we still pass the destroy_ctx
850 	 * reduce_errno since that indicates whether or not the volume was
851 	 * actually destroyed.
852 	 */
853 	destroy_ctx->cb_fn(destroy_ctx->cb_arg, destroy_ctx->reduce_errno);
854 	spdk_free(destroy_ctx->super);
855 	free(destroy_ctx);
856 }
857 
858 static void
859 _destroy_zero_super_cpl(void *cb_arg, int reduce_errno)
860 {
861 	struct reduce_destroy_ctx *destroy_ctx = cb_arg;
862 	struct spdk_reduce_vol *vol = destroy_ctx->vol;
863 
864 	destroy_ctx->reduce_errno = reduce_errno;
865 	spdk_reduce_vol_unload(vol, destroy_unload_cpl, destroy_ctx);
866 }
867 
868 static void
869 destroy_load_cb(void *cb_arg, struct spdk_reduce_vol *vol, int reduce_errno)
870 {
871 	struct reduce_destroy_ctx *destroy_ctx = cb_arg;
872 
873 	if (reduce_errno != 0) {
874 		destroy_ctx->cb_fn(destroy_ctx->cb_arg, reduce_errno);
875 		spdk_free(destroy_ctx->super);
876 		free(destroy_ctx);
877 		return;
878 	}
879 
880 	destroy_ctx->vol = vol;
881 	memcpy(destroy_ctx->pm_path, vol->pm_file.path, sizeof(destroy_ctx->pm_path));
882 	destroy_ctx->iov.iov_base = destroy_ctx->super;
883 	destroy_ctx->iov.iov_len = sizeof(*destroy_ctx->super);
884 	destroy_ctx->backing_cb_args.cb_fn = _destroy_zero_super_cpl;
885 	destroy_ctx->backing_cb_args.cb_arg = destroy_ctx;
886 	vol->backing_dev->writev(vol->backing_dev, &destroy_ctx->iov, 1, 0,
887 				 sizeof(*destroy_ctx->super) / vol->backing_dev->blocklen,
888 				 &destroy_ctx->backing_cb_args);
889 }
890 
891 void
892 spdk_reduce_vol_destroy(struct spdk_reduce_backing_dev *backing_dev,
893 			spdk_reduce_vol_op_complete cb_fn, void *cb_arg)
894 {
895 	struct reduce_destroy_ctx *destroy_ctx;
896 
897 	destroy_ctx = calloc(1, sizeof(*destroy_ctx));
898 	if (destroy_ctx == NULL) {
899 		cb_fn(cb_arg, -ENOMEM);
900 		return;
901 	}
902 
903 	destroy_ctx->super = spdk_zmalloc(sizeof(*destroy_ctx->super), 64, NULL,
904 					  SPDK_ENV_LCORE_ID_ANY, SPDK_MALLOC_DMA);
905 	if (destroy_ctx->super == NULL) {
906 		free(destroy_ctx);
907 		cb_fn(cb_arg, -ENOMEM);
908 		return;
909 	}
910 	destroy_ctx->cb_fn = cb_fn;
911 	destroy_ctx->cb_arg = cb_arg;
912 	spdk_reduce_vol_load(backing_dev, destroy_load_cb, destroy_ctx);
913 }
914 
915 static bool
916 _request_spans_chunk_boundary(struct spdk_reduce_vol *vol, uint64_t offset, uint64_t length)
917 {
918 	uint64_t start_chunk, end_chunk;
919 
920 	start_chunk = offset / vol->logical_blocks_per_chunk;
921 	end_chunk = (offset + length - 1) / vol->logical_blocks_per_chunk;
922 
923 	return (start_chunk != end_chunk);
924 }
925 
926 typedef void (*reduce_request_fn)(void *_req, int reduce_errno);
927 
928 static void
929 _reduce_vol_complete_req(struct spdk_reduce_vol_request *req, int reduce_errno)
930 {
931 	struct spdk_reduce_vol_request *next_req;
932 	struct spdk_reduce_vol *vol = req->vol;
933 
934 	req->cb_fn(req->cb_arg, reduce_errno);
935 	TAILQ_REMOVE(&vol->executing_requests, req, tailq);
936 
937 	TAILQ_FOREACH(next_req, &vol->queued_requests, tailq) {
938 		if (next_req->logical_map_index == req->logical_map_index) {
939 			TAILQ_REMOVE(&vol->queued_requests, next_req, tailq);
940 			if (next_req->type == REDUCE_IO_READV) {
941 				_start_readv_request(next_req);
942 			} else {
943 				assert(next_req->type == REDUCE_IO_WRITEV);
944 				_start_writev_request(next_req);
945 			}
946 			break;
947 		}
948 	}
949 
950 	TAILQ_INSERT_HEAD(&vol->free_requests, req, tailq);
951 }
952 
953 static void
954 _write_write_done(void *_req, int reduce_errno)
955 {
956 	struct spdk_reduce_vol_request *req = _req;
957 	struct spdk_reduce_vol *vol = req->vol;
958 	uint64_t old_chunk_map_index;
959 	struct spdk_reduce_chunk_map *old_chunk;
960 	uint32_t i;
961 
962 	if (reduce_errno != 0) {
963 		req->reduce_errno = reduce_errno;
964 	}
965 
966 	assert(req->num_backing_ops > 0);
967 	if (--req->num_backing_ops > 0) {
968 		return;
969 	}
970 
971 	if (req->reduce_errno != 0) {
972 		_reduce_vol_complete_req(req, req->reduce_errno);
973 		return;
974 	}
975 
976 	old_chunk_map_index = vol->pm_logical_map[req->logical_map_index];
977 	if (old_chunk_map_index != REDUCE_EMPTY_MAP_ENTRY) {
978 		old_chunk = _reduce_vol_get_chunk_map(vol, old_chunk_map_index);
979 		for (i = 0; i < vol->backing_io_units_per_chunk; i++) {
980 			if (old_chunk->io_unit_index[i] == REDUCE_EMPTY_MAP_ENTRY) {
981 				break;
982 			}
983 			assert(spdk_bit_array_get(vol->allocated_backing_io_units, old_chunk->io_unit_index[i]) == true);
984 			spdk_bit_array_clear(vol->allocated_backing_io_units, old_chunk->io_unit_index[i]);
985 			old_chunk->io_unit_index[i] = REDUCE_EMPTY_MAP_ENTRY;
986 		}
987 		spdk_bit_array_clear(vol->allocated_chunk_maps, old_chunk_map_index);
988 	}
989 
990 	/*
991 	 * We don't need to persist the clearing of the old chunk map here.  The old chunk map
992 	 * becomes invalid after we update the logical map, since the old chunk map will no
993 	 * longer have a reference to it in the logical map.
994 	 */
995 
996 	/* Persist the new chunk map.  This must be persisted before we update the logical map. */
997 	_reduce_persist(vol, req->chunk,
998 			_reduce_vol_get_chunk_struct_size(vol->backing_io_units_per_chunk));
999 
1000 	vol->pm_logical_map[req->logical_map_index] = req->chunk_map_index;
1001 
1002 	_reduce_persist(vol, &vol->pm_logical_map[req->logical_map_index], sizeof(uint64_t));
1003 
1004 	_reduce_vol_complete_req(req, 0);
1005 }
1006 
1007 static void
1008 _issue_backing_ops(struct spdk_reduce_vol_request *req, struct spdk_reduce_vol *vol,
1009 		   reduce_request_fn next_fn, bool is_write)
1010 {
1011 	struct iovec *iov;
1012 	uint8_t *buf;
1013 	uint32_t i;
1014 
1015 	if (req->chunk_is_compressed) {
1016 		iov = req->comp_buf_iov;
1017 		buf = req->comp_buf;
1018 	} else {
1019 		iov = req->decomp_buf_iov;
1020 		buf = req->decomp_buf;
1021 	}
1022 
1023 	req->num_backing_ops = req->num_io_units;
1024 	req->backing_cb_args.cb_fn = next_fn;
1025 	req->backing_cb_args.cb_arg = req;
1026 	for (i = 0; i < req->num_io_units; i++) {
1027 		iov[i].iov_base = buf + i * vol->params.backing_io_unit_size;
1028 		iov[i].iov_len = vol->params.backing_io_unit_size;
1029 		if (is_write) {
1030 			vol->backing_dev->writev(vol->backing_dev, &iov[i], 1,
1031 						 req->chunk->io_unit_index[i] * vol->backing_lba_per_io_unit,
1032 						 vol->backing_lba_per_io_unit, &req->backing_cb_args);
1033 		} else {
1034 			vol->backing_dev->readv(vol->backing_dev, &iov[i], 1,
1035 						req->chunk->io_unit_index[i] * vol->backing_lba_per_io_unit,
1036 						vol->backing_lba_per_io_unit, &req->backing_cb_args);
1037 		}
1038 	}
1039 }
1040 
1041 static void
1042 _reduce_vol_write_chunk(struct spdk_reduce_vol_request *req, reduce_request_fn next_fn,
1043 			uint32_t compressed_size)
1044 {
1045 	struct spdk_reduce_vol *vol = req->vol;
1046 	uint32_t i;
1047 	uint64_t chunk_offset, remainder, total_len = 0;
1048 	uint8_t *buf;
1049 	int j;
1050 
1051 	req->chunk_map_index = spdk_bit_array_find_first_clear(vol->allocated_chunk_maps, 0);
1052 
1053 	/* TODO: fail if no chunk map found - but really this should not happen if we
1054 	 * size the number of requests similarly to number of extra chunk maps
1055 	 */
1056 	assert(req->chunk_map_index != UINT32_MAX);
1057 	spdk_bit_array_set(vol->allocated_chunk_maps, req->chunk_map_index);
1058 
1059 	req->chunk = _reduce_vol_get_chunk_map(vol, req->chunk_map_index);
1060 	req->num_io_units = spdk_divide_round_up(compressed_size,
1061 			    vol->params.backing_io_unit_size);
1062 	req->chunk_is_compressed = (req->num_io_units != vol->backing_io_units_per_chunk);
1063 	req->chunk->compressed_size =
1064 		req->chunk_is_compressed ? compressed_size : vol->params.chunk_size;
1065 
1066 	/* if the chunk is uncompressed we need to copy the data from the host buffers. */
1067 	if (req->chunk_is_compressed == false) {
1068 		chunk_offset = req->offset % vol->logical_blocks_per_chunk;
1069 		buf = req->decomp_buf;
1070 		total_len = chunk_offset * vol->params.logical_block_size;
1071 
1072 		/* zero any offset into chunk */
1073 		if (req->rmw == false && chunk_offset) {
1074 			memset(buf, 0, total_len);
1075 		}
1076 		buf += total_len;
1077 
1078 		/* copy the data */
1079 		for (j = 0; j < req->iovcnt; j++) {
1080 			memcpy(buf, req->iov[j].iov_base, req->iov[j].iov_len);
1081 			buf += req->iov[j].iov_len;
1082 			total_len += req->iov[j].iov_len;
1083 		}
1084 
1085 		/* zero any remainder */
1086 		remainder = vol->params.chunk_size - total_len;
1087 		total_len += remainder;
1088 		if (req->rmw == false && remainder) {
1089 			memset(buf, 0, remainder);
1090 		}
1091 		assert(total_len == vol->params.chunk_size);
1092 	}
1093 
1094 	for (i = 0; i < req->num_io_units; i++) {
1095 		req->chunk->io_unit_index[i] = spdk_bit_array_find_first_clear(vol->allocated_backing_io_units, 0);
1096 		/* TODO: fail if no backing block found - but really this should also not
1097 		 * happen (see comment above).
1098 		 */
1099 		assert(req->chunk->io_unit_index[i] != UINT32_MAX);
1100 		spdk_bit_array_set(vol->allocated_backing_io_units, req->chunk->io_unit_index[i]);
1101 	}
1102 
1103 	_issue_backing_ops(req, vol, next_fn, true /* write */);
1104 }
1105 
1106 static void
1107 _write_compress_done(void *_req, int reduce_errno)
1108 {
1109 	struct spdk_reduce_vol_request *req = _req;
1110 
1111 	/* Negative reduce_errno indicates failure for compression operations.
1112 	 * Just write the uncompressed data instead.  Force this to happen
1113 	 * by just passing the full chunk size to _reduce_vol_write_chunk.
1114 	 * When it sees the data couldn't be compressed, it will just write
1115 	 * the uncompressed buffer to disk.
1116 	 */
1117 	if (reduce_errno < 0) {
1118 		reduce_errno = req->vol->params.chunk_size;
1119 	}
1120 
1121 	/* Positive reduce_errno indicates number of bytes in compressed buffer. */
1122 	_reduce_vol_write_chunk(req, _write_write_done, (uint32_t)reduce_errno);
1123 }
1124 
1125 static void
1126 _reduce_vol_compress_chunk(struct spdk_reduce_vol_request *req, reduce_request_fn next_fn)
1127 {
1128 	struct spdk_reduce_vol *vol = req->vol;
1129 
1130 	req->backing_cb_args.cb_fn = next_fn;
1131 	req->backing_cb_args.cb_arg = req;
1132 	req->comp_buf_iov[0].iov_base = req->comp_buf;
1133 	req->comp_buf_iov[0].iov_len = vol->params.chunk_size;
1134 	vol->backing_dev->compress(vol->backing_dev,
1135 				   &req->decomp_iov[0], req->decomp_iovcnt, req->comp_buf_iov, 1,
1136 				   &req->backing_cb_args);
1137 }
1138 
1139 static void
1140 _reduce_vol_decompress_chunk_scratch(struct spdk_reduce_vol_request *req, reduce_request_fn next_fn)
1141 {
1142 	struct spdk_reduce_vol *vol = req->vol;
1143 
1144 	req->backing_cb_args.cb_fn = next_fn;
1145 	req->backing_cb_args.cb_arg = req;
1146 	req->comp_buf_iov[0].iov_base = req->comp_buf;
1147 	req->comp_buf_iov[0].iov_len = req->chunk->compressed_size;
1148 	req->decomp_buf_iov[0].iov_base = req->decomp_buf;
1149 	req->decomp_buf_iov[0].iov_len = vol->params.chunk_size;
1150 	vol->backing_dev->decompress(vol->backing_dev,
1151 				     req->comp_buf_iov, 1, req->decomp_buf_iov, 1,
1152 				     &req->backing_cb_args);
1153 }
1154 
1155 static void
1156 _reduce_vol_decompress_chunk(struct spdk_reduce_vol_request *req, reduce_request_fn next_fn)
1157 {
1158 	struct spdk_reduce_vol *vol = req->vol;
1159 	uint64_t chunk_offset, remainder = 0;
1160 	uint64_t ttl_len = 0;
1161 	int i;
1162 
1163 	req->decomp_iovcnt = 0;
1164 	chunk_offset = req->offset % vol->logical_blocks_per_chunk;
1165 
1166 	if (chunk_offset) {
1167 		/* first iov point to our scratch buffer for any offset into the chunk */
1168 		req->decomp_iov[0].iov_base = req->decomp_buf;
1169 		req->decomp_iov[0].iov_len = chunk_offset * vol->params.logical_block_size;
1170 		ttl_len += req->decomp_iov[0].iov_len;
1171 		req->decomp_iovcnt = 1;
1172 	}
1173 
1174 	/* now the user data iov, direct to the user buffer */
1175 	for (i = 0; i < req->iovcnt; i++) {
1176 		req->decomp_iov[i + req->decomp_iovcnt].iov_base = req->iov[i].iov_base;
1177 		req->decomp_iov[i + req->decomp_iovcnt].iov_len = req->iov[i].iov_len;
1178 		ttl_len += req->decomp_iov[i + req->decomp_iovcnt].iov_len;
1179 	}
1180 	req->decomp_iovcnt += req->iovcnt;
1181 
1182 	/* send the rest of the chunk to our scratch buffer */
1183 	remainder = vol->params.chunk_size - ttl_len;
1184 	if (remainder) {
1185 		req->decomp_iov[req->decomp_iovcnt].iov_base = req->decomp_buf + ttl_len;
1186 		req->decomp_iov[req->decomp_iovcnt].iov_len = remainder;
1187 		ttl_len += req->decomp_iov[req->decomp_iovcnt].iov_len;
1188 		req->decomp_iovcnt++;
1189 	}
1190 	assert(ttl_len == vol->params.chunk_size);
1191 
1192 	req->backing_cb_args.cb_fn = next_fn;
1193 	req->backing_cb_args.cb_arg = req;
1194 	req->comp_buf_iov[0].iov_base = req->comp_buf;
1195 	req->comp_buf_iov[0].iov_len = req->chunk->compressed_size;
1196 	vol->backing_dev->decompress(vol->backing_dev,
1197 				     req->comp_buf_iov, 1, &req->decomp_iov[0], req->decomp_iovcnt,
1198 				     &req->backing_cb_args);
1199 }
1200 
1201 static void
1202 _write_decompress_done(void *_req, int reduce_errno)
1203 {
1204 	struct spdk_reduce_vol_request *req = _req;
1205 	struct spdk_reduce_vol *vol = req->vol;
1206 	uint64_t chunk_offset, remainder, ttl_len = 0;
1207 	int i;
1208 
1209 	/* Negative reduce_errno indicates failure for compression operations. */
1210 	if (reduce_errno < 0) {
1211 		_reduce_vol_complete_req(req, reduce_errno);
1212 		return;
1213 	}
1214 
1215 	/* Positive reduce_errno indicates number of bytes in decompressed
1216 	 *  buffer.  This should equal the chunk size - otherwise that's another
1217 	 *  type of failure.
1218 	 */
1219 	if ((uint32_t)reduce_errno != vol->params.chunk_size) {
1220 		_reduce_vol_complete_req(req, -EIO);
1221 		return;
1222 	}
1223 
1224 	req->decomp_iovcnt = 0;
1225 	chunk_offset = req->offset % vol->logical_blocks_per_chunk;
1226 
1227 	if (chunk_offset) {
1228 		req->decomp_iov[0].iov_base = req->decomp_buf;
1229 		req->decomp_iov[0].iov_len = chunk_offset * vol->params.logical_block_size;
1230 		ttl_len += req->decomp_iov[0].iov_len;
1231 		req->decomp_iovcnt = 1;
1232 	}
1233 
1234 	for (i = 0; i < req->iovcnt; i++) {
1235 		req->decomp_iov[i + req->decomp_iovcnt].iov_base = req->iov[i].iov_base;
1236 		req->decomp_iov[i + req->decomp_iovcnt].iov_len = req->iov[i].iov_len;
1237 		ttl_len += req->decomp_iov[i + req->decomp_iovcnt].iov_len;
1238 	}
1239 	req->decomp_iovcnt += req->iovcnt;
1240 
1241 	remainder = vol->params.chunk_size - ttl_len;
1242 	if (remainder) {
1243 		req->decomp_iov[req->decomp_iovcnt].iov_base = req->decomp_buf + ttl_len;
1244 		req->decomp_iov[req->decomp_iovcnt].iov_len = remainder;
1245 		ttl_len += req->decomp_iov[req->decomp_iovcnt].iov_len;
1246 		req->decomp_iovcnt++;
1247 	}
1248 	assert(ttl_len == vol->params.chunk_size);
1249 
1250 	_reduce_vol_compress_chunk(req, _write_compress_done);
1251 }
1252 
1253 static void
1254 _write_read_done(void *_req, int reduce_errno)
1255 {
1256 	struct spdk_reduce_vol_request *req = _req;
1257 
1258 	if (reduce_errno != 0) {
1259 		req->reduce_errno = reduce_errno;
1260 	}
1261 
1262 	assert(req->num_backing_ops > 0);
1263 	if (--req->num_backing_ops > 0) {
1264 		return;
1265 	}
1266 
1267 	if (req->reduce_errno != 0) {
1268 		_reduce_vol_complete_req(req, req->reduce_errno);
1269 		return;
1270 	}
1271 
1272 	if (req->chunk_is_compressed) {
1273 		_reduce_vol_decompress_chunk_scratch(req, _write_decompress_done);
1274 	} else {
1275 		_write_decompress_done(req, req->chunk->compressed_size);
1276 	}
1277 }
1278 
1279 static void
1280 _read_decompress_done(void *_req, int reduce_errno)
1281 {
1282 	struct spdk_reduce_vol_request *req = _req;
1283 	struct spdk_reduce_vol *vol = req->vol;
1284 
1285 	/* Negative reduce_errno indicates failure for compression operations. */
1286 	if (reduce_errno < 0) {
1287 		_reduce_vol_complete_req(req, reduce_errno);
1288 		return;
1289 	}
1290 
1291 	/* Positive reduce_errno indicates number of bytes in decompressed
1292 	 *  buffer.  This should equal the chunk size - otherwise that's another
1293 	 *  type of failure.
1294 	 */
1295 	if ((uint32_t)reduce_errno != vol->params.chunk_size) {
1296 		_reduce_vol_complete_req(req, -EIO);
1297 		return;
1298 	}
1299 
1300 	_reduce_vol_complete_req(req, 0);
1301 }
1302 
1303 static void
1304 _read_read_done(void *_req, int reduce_errno)
1305 {
1306 	struct spdk_reduce_vol_request *req = _req;
1307 	uint64_t chunk_offset;
1308 	uint8_t *buf;
1309 	int i;
1310 
1311 	if (reduce_errno != 0) {
1312 		req->reduce_errno = reduce_errno;
1313 	}
1314 
1315 	assert(req->num_backing_ops > 0);
1316 	if (--req->num_backing_ops > 0) {
1317 		return;
1318 	}
1319 
1320 	if (req->reduce_errno != 0) {
1321 		_reduce_vol_complete_req(req, req->reduce_errno);
1322 		return;
1323 	}
1324 
1325 	if (req->chunk_is_compressed) {
1326 		_reduce_vol_decompress_chunk(req, _read_decompress_done);
1327 	} else {
1328 
1329 		/* If the chunk was compressed, the data would have been sent to the
1330 		 *  host buffers by the decompression operation, if not we need to memcpy here.
1331 		 */
1332 		chunk_offset = req->offset % req->vol->logical_blocks_per_chunk;
1333 		buf = req->decomp_buf + chunk_offset * req->vol->params.logical_block_size;
1334 		for (i = 0; i < req->iovcnt; i++) {
1335 			memcpy(req->iov[i].iov_base, buf, req->iov[i].iov_len);
1336 			buf += req->iov[i].iov_len;
1337 		}
1338 
1339 		_read_decompress_done(req, req->chunk->compressed_size);
1340 	}
1341 }
1342 
1343 static void
1344 _reduce_vol_read_chunk(struct spdk_reduce_vol_request *req, reduce_request_fn next_fn)
1345 {
1346 	struct spdk_reduce_vol *vol = req->vol;
1347 
1348 	req->chunk_map_index = vol->pm_logical_map[req->logical_map_index];
1349 	assert(req->chunk_map_index != UINT32_MAX);
1350 
1351 	req->chunk = _reduce_vol_get_chunk_map(vol, req->chunk_map_index);
1352 	req->num_io_units = spdk_divide_round_up(req->chunk->compressed_size,
1353 			    vol->params.backing_io_unit_size);
1354 	req->chunk_is_compressed = (req->num_io_units != vol->backing_io_units_per_chunk);
1355 
1356 	_issue_backing_ops(req, vol, next_fn, false /* read */);
1357 }
1358 
1359 static bool
1360 _iov_array_is_valid(struct spdk_reduce_vol *vol, struct iovec *iov, int iovcnt,
1361 		    uint64_t length)
1362 {
1363 	uint64_t size = 0;
1364 	int i;
1365 
1366 	for (i = 0; i < iovcnt; i++) {
1367 		size += iov[i].iov_len;
1368 	}
1369 
1370 	return size == (length * vol->params.logical_block_size);
1371 }
1372 
1373 static bool
1374 _check_overlap(struct spdk_reduce_vol *vol, uint64_t logical_map_index)
1375 {
1376 	struct spdk_reduce_vol_request *req;
1377 
1378 	TAILQ_FOREACH(req, &vol->executing_requests, tailq) {
1379 		if (logical_map_index == req->logical_map_index) {
1380 			return true;
1381 		}
1382 	}
1383 
1384 	return false;
1385 }
1386 
1387 static void
1388 _start_readv_request(struct spdk_reduce_vol_request *req)
1389 {
1390 	TAILQ_INSERT_TAIL(&req->vol->executing_requests, req, tailq);
1391 	_reduce_vol_read_chunk(req, _read_read_done);
1392 }
1393 
1394 void
1395 spdk_reduce_vol_readv(struct spdk_reduce_vol *vol,
1396 		      struct iovec *iov, int iovcnt, uint64_t offset, uint64_t length,
1397 		      spdk_reduce_vol_op_complete cb_fn, void *cb_arg)
1398 {
1399 	struct spdk_reduce_vol_request *req;
1400 	uint64_t logical_map_index;
1401 	bool overlapped;
1402 	int i;
1403 
1404 	if (length == 0) {
1405 		cb_fn(cb_arg, 0);
1406 		return;
1407 	}
1408 
1409 	if (_request_spans_chunk_boundary(vol, offset, length)) {
1410 		cb_fn(cb_arg, -EINVAL);
1411 		return;
1412 	}
1413 
1414 	if (!_iov_array_is_valid(vol, iov, iovcnt, length)) {
1415 		cb_fn(cb_arg, -EINVAL);
1416 		return;
1417 	}
1418 
1419 	logical_map_index = offset / vol->logical_blocks_per_chunk;
1420 	overlapped = _check_overlap(vol, logical_map_index);
1421 
1422 	if (!overlapped && vol->pm_logical_map[logical_map_index] == REDUCE_EMPTY_MAP_ENTRY) {
1423 		/*
1424 		 * This chunk hasn't been allocated.  So treat the data as all
1425 		 * zeroes for this chunk - do the memset and immediately complete
1426 		 * the operation.
1427 		 */
1428 		for (i = 0; i < iovcnt; i++) {
1429 			memset(iov[i].iov_base, 0, iov[i].iov_len);
1430 		}
1431 		cb_fn(cb_arg, 0);
1432 		return;
1433 	}
1434 
1435 	req = TAILQ_FIRST(&vol->free_requests);
1436 	if (req == NULL) {
1437 		cb_fn(cb_arg, -ENOMEM);
1438 		return;
1439 	}
1440 
1441 	TAILQ_REMOVE(&vol->free_requests, req, tailq);
1442 	req->type = REDUCE_IO_READV;
1443 	req->vol = vol;
1444 	req->iov = iov;
1445 	req->iovcnt = iovcnt;
1446 	req->offset = offset;
1447 	req->logical_map_index = logical_map_index;
1448 	req->length = length;
1449 	req->cb_fn = cb_fn;
1450 	req->cb_arg = cb_arg;
1451 
1452 	if (!overlapped) {
1453 		_start_readv_request(req);
1454 	} else {
1455 		TAILQ_INSERT_TAIL(&vol->queued_requests, req, tailq);
1456 	}
1457 }
1458 
1459 static void
1460 _start_writev_request(struct spdk_reduce_vol_request *req)
1461 {
1462 	struct spdk_reduce_vol *vol = req->vol;
1463 	uint64_t chunk_offset, ttl_len = 0;
1464 	uint64_t remainder = 0;
1465 	uint32_t lbsize;
1466 	int i;
1467 
1468 	TAILQ_INSERT_TAIL(&req->vol->executing_requests, req, tailq);
1469 	if (vol->pm_logical_map[req->logical_map_index] != REDUCE_EMPTY_MAP_ENTRY) {
1470 		if ((req->length * vol->params.logical_block_size) < vol->params.chunk_size) {
1471 			/* Read old chunk, then overwrite with data from this write
1472 			 *  operation.
1473 			 */
1474 			req->rmw = true;
1475 			_reduce_vol_read_chunk(req, _write_read_done);
1476 			return;
1477 		}
1478 	}
1479 
1480 	lbsize = vol->params.logical_block_size;
1481 	req->decomp_iovcnt = 0;
1482 	req->rmw = false;
1483 
1484 	/* Note: point to our zero buf for offset into the chunk. */
1485 	chunk_offset = req->offset % vol->logical_blocks_per_chunk;
1486 	if (chunk_offset != 0) {
1487 		ttl_len += chunk_offset * lbsize;
1488 		req->decomp_iov[0].iov_base = g_zero_buf;
1489 		req->decomp_iov[0].iov_len = ttl_len;
1490 		req->decomp_iovcnt = 1;
1491 	}
1492 
1493 	/* now the user data iov, direct from the user buffer */
1494 	for (i = 0; i < req->iovcnt; i++) {
1495 		req->decomp_iov[i + req->decomp_iovcnt].iov_base = req->iov[i].iov_base;
1496 		req->decomp_iov[i + req->decomp_iovcnt].iov_len = req->iov[i].iov_len;
1497 		ttl_len += req->decomp_iov[i + req->decomp_iovcnt].iov_len;
1498 	}
1499 	req->decomp_iovcnt += req->iovcnt;
1500 
1501 	remainder = vol->params.chunk_size - ttl_len;
1502 	if (remainder) {
1503 		req->decomp_iov[req->decomp_iovcnt].iov_base = g_zero_buf;
1504 		req->decomp_iov[req->decomp_iovcnt].iov_len = remainder;
1505 		ttl_len += req->decomp_iov[req->decomp_iovcnt].iov_len;
1506 		req->decomp_iovcnt++;
1507 	}
1508 	assert(ttl_len == req->vol->params.chunk_size);
1509 
1510 	_reduce_vol_compress_chunk(req, _write_compress_done);
1511 }
1512 
1513 void
1514 spdk_reduce_vol_writev(struct spdk_reduce_vol *vol,
1515 		       struct iovec *iov, int iovcnt, uint64_t offset, uint64_t length,
1516 		       spdk_reduce_vol_op_complete cb_fn, void *cb_arg)
1517 {
1518 	struct spdk_reduce_vol_request *req;
1519 	uint64_t logical_map_index;
1520 	bool overlapped;
1521 
1522 	if (length == 0) {
1523 		cb_fn(cb_arg, 0);
1524 		return;
1525 	}
1526 
1527 	if (_request_spans_chunk_boundary(vol, offset, length)) {
1528 		cb_fn(cb_arg, -EINVAL);
1529 		return;
1530 	}
1531 
1532 	if (!_iov_array_is_valid(vol, iov, iovcnt, length)) {
1533 		cb_fn(cb_arg, -EINVAL);
1534 		return;
1535 	}
1536 
1537 	logical_map_index = offset / vol->logical_blocks_per_chunk;
1538 	overlapped = _check_overlap(vol, logical_map_index);
1539 
1540 	req = TAILQ_FIRST(&vol->free_requests);
1541 	if (req == NULL) {
1542 		cb_fn(cb_arg, -ENOMEM);
1543 		return;
1544 	}
1545 
1546 	TAILQ_REMOVE(&vol->free_requests, req, tailq);
1547 	req->type = REDUCE_IO_WRITEV;
1548 	req->vol = vol;
1549 	req->iov = iov;
1550 	req->iovcnt = iovcnt;
1551 	req->offset = offset;
1552 	req->logical_map_index = logical_map_index;
1553 	req->length = length;
1554 	req->cb_fn = cb_fn;
1555 	req->cb_arg = cb_arg;
1556 
1557 	if (!overlapped) {
1558 		_start_writev_request(req);
1559 	} else {
1560 		TAILQ_INSERT_TAIL(&vol->queued_requests, req, tailq);
1561 	}
1562 }
1563 
1564 const struct spdk_reduce_vol_params *
1565 spdk_reduce_vol_get_params(struct spdk_reduce_vol *vol)
1566 {
1567 	return &vol->params;
1568 }
1569 
1570 void spdk_reduce_vol_print_info(struct spdk_reduce_vol *vol)
1571 {
1572 	uint64_t logical_map_size, num_chunks, ttl_chunk_sz;
1573 	uint32_t struct_size;
1574 	uint64_t chunk_map_size;
1575 
1576 	SPDK_NOTICELOG("vol info:\n");
1577 	SPDK_NOTICELOG("\tvol->params.backing_io_unit_size = 0x%x\n", vol->params.backing_io_unit_size);
1578 	SPDK_NOTICELOG("\tvol->params.logical_block_size = 0x%x\n", vol->params.logical_block_size);
1579 	SPDK_NOTICELOG("\tvol->params.chunk_size = 0x%x\n", vol->params.chunk_size);
1580 	SPDK_NOTICELOG("\tvol->params.vol_size = 0x%" PRIx64 "\n", vol->params.vol_size);
1581 	num_chunks = _get_total_chunks(vol->params.vol_size, vol->params.chunk_size);
1582 	SPDK_NOTICELOG("\ttotal chunks (including extra) = 0x%" PRIx64 "\n", num_chunks);
1583 	SPDK_NOTICELOG("\ttotal chunks (excluding extra) = 0x%" PRIx64 "\n",
1584 		       vol->params.vol_size / vol->params.chunk_size);
1585 	ttl_chunk_sz = _get_pm_total_chunks_size(vol->params.vol_size, vol->params.chunk_size,
1586 			vol->params.backing_io_unit_size);
1587 	SPDK_NOTICELOG("\ttotal_chunks_size = 0x%" PRIx64 "\n", ttl_chunk_sz);
1588 	struct_size = _reduce_vol_get_chunk_struct_size(vol->backing_io_units_per_chunk);
1589 	SPDK_NOTICELOG("\tchunk_struct_size = 0x%x\n", struct_size);
1590 
1591 	SPDK_NOTICELOG("pmem info:\n");
1592 	SPDK_NOTICELOG("\tvol->pm_file.size = 0x%" PRIx64 "\n", vol->pm_file.size);
1593 	SPDK_NOTICELOG("\tvol->pm_file.pm_buf = %p\n", (void *)vol->pm_file.pm_buf);
1594 	SPDK_NOTICELOG("\tvol->pm_super = %p\n", (void *)vol->pm_super);
1595 	SPDK_NOTICELOG("\tvol->pm_logical_map = %p\n", (void *)vol->pm_logical_map);
1596 	logical_map_size = _get_pm_logical_map_size(vol->params.vol_size,
1597 			   vol->params.chunk_size);
1598 	SPDK_NOTICELOG("\tlogical_map_size = 0x%" PRIx64 "\n", logical_map_size);
1599 	SPDK_NOTICELOG("\tvol->pm_chunk_maps = %p\n", (void *)vol->pm_chunk_maps);
1600 	chunk_map_size = _get_pm_total_chunks_size(vol->params.vol_size, vol->params.chunk_size,
1601 			 vol->params.backing_io_unit_size);
1602 	SPDK_NOTICELOG("\tchunk_map_size = 0x%" PRIx64 "\n", chunk_map_size);
1603 }
1604 
1605 SPDK_LOG_REGISTER_COMPONENT("reduce", SPDK_LOG_REDUCE)
1606