xref: /spdk/lib/blobfs/blobfs.c (revision 1be7a80c2d0f5e323455efd701a18d8cc4ce7868)
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/blobfs.h"
37 #include "blobfs_internal.h"
38 
39 #include "spdk/queue.h"
40 #include "spdk/io_channel.h"
41 #include "spdk/assert.h"
42 #include "spdk/env.h"
43 #include "spdk/util.h"
44 #include "spdk_internal/log.h"
45 
46 #define BLOBFS_TRACE(file, str, args...) \
47 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s " str, file->name, ##args)
48 
49 #define BLOBFS_TRACE_RW(file, str, args...) \
50 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS_RW, "file=%s " str, file->name, ##args)
51 
52 #define BLOBFS_CACHE_SIZE (4ULL * 1024 * 1024 * 1024)
53 
54 static uint64_t g_fs_cache_size = BLOBFS_CACHE_SIZE;
55 static struct spdk_mempool *g_cache_pool;
56 static TAILQ_HEAD(, spdk_file) g_caches;
57 static int g_fs_count = 0;
58 static pthread_mutex_t g_cache_init_lock = PTHREAD_MUTEX_INITIALIZER;
59 static pthread_spinlock_t g_caches_lock;
60 
61 static void
62 __sem_post(void *arg, int bserrno)
63 {
64 	sem_t *sem = arg;
65 
66 	sem_post(sem);
67 }
68 
69 void
70 spdk_cache_buffer_free(struct cache_buffer *cache_buffer)
71 {
72 	spdk_mempool_put(g_cache_pool, cache_buffer->buf);
73 	free(cache_buffer);
74 }
75 
76 #define CACHE_READAHEAD_THRESHOLD	(128 * 1024)
77 
78 struct spdk_file {
79 	struct spdk_filesystem	*fs;
80 	struct spdk_blob	*blob;
81 	char			*name;
82 	uint64_t		length;
83 	bool			open_for_writing;
84 	uint64_t		length_flushed;
85 	uint64_t		append_pos;
86 	uint64_t		seq_byte_count;
87 	uint64_t		next_seq_offset;
88 	uint32_t		priority;
89 	TAILQ_ENTRY(spdk_file)	tailq;
90 	spdk_blob_id		blobid;
91 	uint32_t		ref_count;
92 	pthread_spinlock_t	lock;
93 	struct cache_buffer	*last;
94 	struct cache_tree	*tree;
95 	TAILQ_HEAD(open_requests_head, spdk_fs_request) open_requests;
96 	TAILQ_HEAD(sync_requests_head, spdk_fs_request) sync_requests;
97 	TAILQ_ENTRY(spdk_file)	cache_tailq;
98 };
99 
100 struct spdk_filesystem {
101 	struct spdk_blob_store	*bs;
102 	TAILQ_HEAD(, spdk_file)	files;
103 	struct spdk_bs_opts	bs_opts;
104 	struct spdk_bs_dev	*bdev;
105 	fs_send_request_fn	send_request;
106 
107 	struct {
108 		uint32_t		max_ops;
109 		struct spdk_io_channel	*sync_io_channel;
110 		struct spdk_fs_channel	*sync_fs_channel;
111 	} sync_target;
112 
113 	struct {
114 		uint32_t		max_ops;
115 		struct spdk_io_channel	*md_io_channel;
116 		struct spdk_fs_channel	*md_fs_channel;
117 	} md_target;
118 
119 	struct {
120 		uint32_t		max_ops;
121 	} io_target;
122 };
123 
124 struct spdk_fs_cb_args {
125 	union {
126 		spdk_fs_op_with_handle_complete		fs_op_with_handle;
127 		spdk_fs_op_complete			fs_op;
128 		spdk_file_op_with_handle_complete	file_op_with_handle;
129 		spdk_file_op_complete			file_op;
130 		spdk_file_stat_op_complete		stat_op;
131 	} fn;
132 	void *arg;
133 	sem_t *sem;
134 	struct spdk_filesystem *fs;
135 	struct spdk_file *file;
136 	int rc;
137 	bool from_request;
138 	union {
139 		struct {
140 			uint64_t	length;
141 		} truncate;
142 		struct {
143 			struct spdk_io_channel	*channel;
144 			void		*user_buf;
145 			void		*pin_buf;
146 			int		is_read;
147 			off_t		offset;
148 			size_t		length;
149 			uint64_t	start_page;
150 			uint64_t	num_pages;
151 			uint32_t	blocklen;
152 		} rw;
153 		struct {
154 			const char	*old_name;
155 			const char	*new_name;
156 		} rename;
157 		struct {
158 			struct cache_buffer	*cache_buffer;
159 			uint64_t		length;
160 		} flush;
161 		struct {
162 			struct cache_buffer	*cache_buffer;
163 			uint64_t		length;
164 			uint64_t		offset;
165 		} readahead;
166 		struct {
167 			uint64_t			offset;
168 			TAILQ_ENTRY(spdk_fs_request)	tailq;
169 			bool				xattr_in_progress;
170 		} sync;
171 		struct {
172 			uint32_t			num_clusters;
173 		} resize;
174 		struct {
175 			const char	*name;
176 			uint32_t	flags;
177 			TAILQ_ENTRY(spdk_fs_request)	tailq;
178 		} open;
179 		struct {
180 			const char	*name;
181 		} create;
182 		struct {
183 			const char	*name;
184 		} delete;
185 		struct {
186 			const char	*name;
187 		} stat;
188 	} op;
189 };
190 
191 static void cache_free_buffers(struct spdk_file *file);
192 
193 static void
194 __initialize_cache(void)
195 {
196 	assert(g_cache_pool == NULL);
197 
198 	g_cache_pool = spdk_mempool_create("spdk_fs_cache",
199 					   g_fs_cache_size / CACHE_BUFFER_SIZE,
200 					   CACHE_BUFFER_SIZE, -1, SPDK_ENV_SOCKET_ID_ANY);
201 	TAILQ_INIT(&g_caches);
202 	pthread_spin_init(&g_caches_lock, 0);
203 }
204 
205 static void
206 __free_cache(void)
207 {
208 	assert(g_cache_pool != NULL);
209 
210 	spdk_mempool_free(g_cache_pool);
211 	g_cache_pool = NULL;
212 }
213 
214 static uint64_t
215 __file_get_blob_size(struct spdk_file *file)
216 {
217 	uint64_t cluster_sz;
218 
219 	cluster_sz = file->fs->bs_opts.cluster_sz;
220 	return cluster_sz * spdk_blob_get_num_clusters(file->blob);
221 }
222 
223 struct spdk_fs_request {
224 	struct spdk_fs_cb_args		args;
225 	TAILQ_ENTRY(spdk_fs_request)	link;
226 	struct spdk_fs_channel		*channel;
227 };
228 
229 struct spdk_fs_channel {
230 	struct spdk_fs_request		*req_mem;
231 	TAILQ_HEAD(, spdk_fs_request)	reqs;
232 	sem_t				sem;
233 	struct spdk_filesystem		*fs;
234 	struct spdk_io_channel		*bs_channel;
235 	fs_send_request_fn		send_request;
236 	bool				sync;
237 	pthread_spinlock_t		lock;
238 };
239 
240 static struct spdk_fs_request *
241 alloc_fs_request(struct spdk_fs_channel *channel)
242 {
243 	struct spdk_fs_request *req;
244 
245 	if (channel->sync) {
246 		pthread_spin_lock(&channel->lock);
247 	}
248 
249 	req = TAILQ_FIRST(&channel->reqs);
250 	if (req) {
251 		TAILQ_REMOVE(&channel->reqs, req, link);
252 	}
253 
254 	if (channel->sync) {
255 		pthread_spin_unlock(&channel->lock);
256 	}
257 
258 	if (req == NULL) {
259 		return NULL;
260 	}
261 	memset(req, 0, sizeof(*req));
262 	req->channel = channel;
263 	req->args.from_request = true;
264 
265 	return req;
266 }
267 
268 static void
269 free_fs_request(struct spdk_fs_request *req)
270 {
271 	struct spdk_fs_channel *channel = req->channel;
272 
273 	if (channel->sync) {
274 		pthread_spin_lock(&channel->lock);
275 	}
276 
277 	TAILQ_INSERT_HEAD(&req->channel->reqs, req, link);
278 
279 	if (channel->sync) {
280 		pthread_spin_unlock(&channel->lock);
281 	}
282 }
283 
284 static int
285 _spdk_fs_channel_create(struct spdk_filesystem *fs, struct spdk_fs_channel *channel,
286 			uint32_t max_ops)
287 {
288 	uint32_t i;
289 
290 	channel->req_mem = calloc(max_ops, sizeof(struct spdk_fs_request));
291 	if (!channel->req_mem) {
292 		return -1;
293 	}
294 
295 	TAILQ_INIT(&channel->reqs);
296 	sem_init(&channel->sem, 0, 0);
297 
298 	for (i = 0; i < max_ops; i++) {
299 		TAILQ_INSERT_TAIL(&channel->reqs, &channel->req_mem[i], link);
300 	}
301 
302 	channel->fs = fs;
303 
304 	return 0;
305 }
306 
307 static int
308 _spdk_fs_md_channel_create(void *io_device, void *ctx_buf)
309 {
310 	struct spdk_filesystem		*fs;
311 	struct spdk_fs_channel		*channel = ctx_buf;
312 
313 	fs = SPDK_CONTAINEROF(io_device, struct spdk_filesystem, md_target);
314 
315 	return _spdk_fs_channel_create(fs, channel, fs->md_target.max_ops);
316 }
317 
318 static int
319 _spdk_fs_sync_channel_create(void *io_device, void *ctx_buf)
320 {
321 	struct spdk_filesystem		*fs;
322 	struct spdk_fs_channel		*channel = ctx_buf;
323 
324 	fs = SPDK_CONTAINEROF(io_device, struct spdk_filesystem, sync_target);
325 
326 	return _spdk_fs_channel_create(fs, channel, fs->sync_target.max_ops);
327 }
328 
329 static int
330 _spdk_fs_io_channel_create(void *io_device, void *ctx_buf)
331 {
332 	struct spdk_filesystem		*fs;
333 	struct spdk_fs_channel		*channel = ctx_buf;
334 
335 	fs = SPDK_CONTAINEROF(io_device, struct spdk_filesystem, io_target);
336 
337 	return _spdk_fs_channel_create(fs, channel, fs->io_target.max_ops);
338 }
339 
340 static void
341 _spdk_fs_channel_destroy(void *io_device, void *ctx_buf)
342 {
343 	struct spdk_fs_channel *channel = ctx_buf;
344 
345 	free(channel->req_mem);
346 	if (channel->bs_channel != NULL) {
347 		spdk_bs_free_io_channel(channel->bs_channel);
348 	}
349 }
350 
351 static void
352 __send_request_direct(fs_request_fn fn, void *arg)
353 {
354 	fn(arg);
355 }
356 
357 static void
358 common_fs_bs_init(struct spdk_filesystem *fs, struct spdk_blob_store *bs)
359 {
360 	fs->bs = bs;
361 	fs->bs_opts.cluster_sz = spdk_bs_get_cluster_size(bs);
362 	fs->md_target.md_fs_channel->bs_channel = spdk_bs_alloc_io_channel(fs->bs);
363 	fs->md_target.md_fs_channel->send_request = __send_request_direct;
364 	fs->sync_target.sync_fs_channel->bs_channel = spdk_bs_alloc_io_channel(fs->bs);
365 	fs->sync_target.sync_fs_channel->send_request = __send_request_direct;
366 
367 	pthread_mutex_lock(&g_cache_init_lock);
368 	if (g_fs_count == 0) {
369 		__initialize_cache();
370 	}
371 	g_fs_count++;
372 	pthread_mutex_unlock(&g_cache_init_lock);
373 }
374 
375 static void
376 init_cb(void *ctx, struct spdk_blob_store *bs, int bserrno)
377 {
378 	struct spdk_fs_request *req = ctx;
379 	struct spdk_fs_cb_args *args = &req->args;
380 	struct spdk_filesystem *fs = args->fs;
381 
382 	if (bserrno == 0) {
383 		common_fs_bs_init(fs, bs);
384 	} else {
385 		free(fs);
386 		fs = NULL;
387 	}
388 
389 	args->fn.fs_op_with_handle(args->arg, fs, bserrno);
390 	free_fs_request(req);
391 }
392 
393 static struct spdk_filesystem *
394 fs_alloc(struct spdk_bs_dev *dev, fs_send_request_fn send_request_fn)
395 {
396 	struct spdk_filesystem *fs;
397 
398 	fs = calloc(1, sizeof(*fs));
399 	if (fs == NULL) {
400 		return NULL;
401 	}
402 
403 	fs->bdev = dev;
404 	fs->send_request = send_request_fn;
405 	TAILQ_INIT(&fs->files);
406 
407 	fs->md_target.max_ops = 512;
408 	spdk_io_device_register(&fs->md_target, _spdk_fs_md_channel_create, _spdk_fs_channel_destroy,
409 				sizeof(struct spdk_fs_channel));
410 	fs->md_target.md_io_channel = spdk_get_io_channel(&fs->md_target);
411 	fs->md_target.md_fs_channel = spdk_io_channel_get_ctx(fs->md_target.md_io_channel);
412 
413 	fs->sync_target.max_ops = 512;
414 	spdk_io_device_register(&fs->sync_target, _spdk_fs_sync_channel_create, _spdk_fs_channel_destroy,
415 				sizeof(struct spdk_fs_channel));
416 	fs->sync_target.sync_io_channel = spdk_get_io_channel(&fs->sync_target);
417 	fs->sync_target.sync_fs_channel = spdk_io_channel_get_ctx(fs->sync_target.sync_io_channel);
418 
419 	fs->io_target.max_ops = 512;
420 	spdk_io_device_register(&fs->io_target, _spdk_fs_io_channel_create, _spdk_fs_channel_destroy,
421 				sizeof(struct spdk_fs_channel));
422 
423 	return fs;
424 }
425 
426 void
427 spdk_fs_init(struct spdk_bs_dev *dev, fs_send_request_fn send_request_fn,
428 	     spdk_fs_op_with_handle_complete cb_fn, void *cb_arg)
429 {
430 	struct spdk_filesystem *fs;
431 	struct spdk_fs_request *req;
432 	struct spdk_fs_cb_args *args;
433 
434 	fs = fs_alloc(dev, send_request_fn);
435 	if (fs == NULL) {
436 		cb_fn(cb_arg, NULL, -ENOMEM);
437 		return;
438 	}
439 
440 	req = alloc_fs_request(fs->md_target.md_fs_channel);
441 	if (req == NULL) {
442 		spdk_put_io_channel(fs->md_target.md_io_channel);
443 		spdk_io_device_unregister(&fs->md_target);
444 		spdk_put_io_channel(fs->sync_target.sync_io_channel);
445 		spdk_io_device_unregister(&fs->sync_target);
446 		spdk_io_device_unregister(&fs->io_target);
447 		free(fs);
448 		cb_fn(cb_arg, NULL, -ENOMEM);
449 		return;
450 	}
451 
452 	args = &req->args;
453 	args->fn.fs_op_with_handle = cb_fn;
454 	args->arg = cb_arg;
455 	args->fs = fs;
456 
457 	spdk_bs_init(dev, NULL, init_cb, req);
458 }
459 
460 static struct spdk_file *
461 file_alloc(struct spdk_filesystem *fs)
462 {
463 	struct spdk_file *file;
464 
465 	file = calloc(1, sizeof(*file));
466 	if (file == NULL) {
467 		return NULL;
468 	}
469 
470 	file->fs = fs;
471 	TAILQ_INIT(&file->open_requests);
472 	TAILQ_INIT(&file->sync_requests);
473 	pthread_spin_init(&file->lock, 0);
474 	file->tree = calloc(1, sizeof(*file->tree));
475 	TAILQ_INSERT_TAIL(&fs->files, file, tailq);
476 	file->priority = SPDK_FILE_PRIORITY_LOW;
477 	return file;
478 }
479 
480 static void
481 iter_cb(void *ctx, struct spdk_blob *blob, int rc)
482 {
483 	struct spdk_fs_request *req = ctx;
484 	struct spdk_fs_cb_args *args = &req->args;
485 	struct spdk_filesystem *fs = args->fs;
486 	struct spdk_file *f;
487 	uint64_t *length;
488 	const char *name;
489 	size_t value_len;
490 
491 	if (rc == -ENOENT) {
492 		/* Finished iterating */
493 		args->fn.fs_op_with_handle(args->arg, fs, 0);
494 		free_fs_request(req);
495 		return;
496 	} else if (rc < 0) {
497 		args->fn.fs_op_with_handle(args->arg, fs, rc);
498 		free_fs_request(req);
499 		return;
500 	}
501 
502 	rc = spdk_bs_md_get_xattr_value(blob, "name", (const void **)&name, &value_len);
503 	if (rc < 0) {
504 		args->fn.fs_op_with_handle(args->arg, fs, rc);
505 		free_fs_request(req);
506 		return;
507 	}
508 
509 	rc = spdk_bs_md_get_xattr_value(blob, "length", (const void **)&length, &value_len);
510 	if (rc < 0) {
511 		args->fn.fs_op_with_handle(args->arg, fs, rc);
512 		free_fs_request(req);
513 		return;
514 	}
515 	assert(value_len == 8);
516 
517 	f = file_alloc(fs);
518 	if (f == NULL) {
519 		args->fn.fs_op_with_handle(args->arg, fs, -ENOMEM);
520 		free_fs_request(req);
521 		return;
522 	}
523 
524 	f->name = strdup(name);
525 	f->blobid = spdk_blob_get_id(blob);
526 	f->length = *length;
527 	f->length_flushed = *length;
528 	f->append_pos = *length;
529 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "added file %s length=%ju\n", f->name, f->length);
530 
531 	spdk_bs_md_iter_next(fs->bs, &blob, iter_cb, req);
532 }
533 
534 static void
535 load_cb(void *ctx, struct spdk_blob_store *bs, int bserrno)
536 {
537 	struct spdk_fs_request *req = ctx;
538 	struct spdk_fs_cb_args *args = &req->args;
539 	struct spdk_filesystem *fs = args->fs;
540 
541 	if (bserrno != 0) {
542 		args->fn.fs_op_with_handle(args->arg, NULL, bserrno);
543 		free_fs_request(req);
544 		free(fs);
545 		return;
546 	}
547 
548 	common_fs_bs_init(fs, bs);
549 	spdk_bs_md_iter_first(fs->bs, iter_cb, req);
550 }
551 
552 void
553 spdk_fs_load(struct spdk_bs_dev *dev, fs_send_request_fn send_request_fn,
554 	     spdk_fs_op_with_handle_complete cb_fn, void *cb_arg)
555 {
556 	struct spdk_filesystem *fs;
557 	struct spdk_fs_cb_args *args;
558 	struct spdk_fs_request *req;
559 
560 	fs = fs_alloc(dev, send_request_fn);
561 	if (fs == NULL) {
562 		cb_fn(cb_arg, NULL, -ENOMEM);
563 		return;
564 	}
565 
566 	req = alloc_fs_request(fs->md_target.md_fs_channel);
567 	if (req == NULL) {
568 		spdk_put_io_channel(fs->md_target.md_io_channel);
569 		spdk_io_device_unregister(&fs->md_target);
570 		spdk_put_io_channel(fs->sync_target.sync_io_channel);
571 		spdk_io_device_unregister(&fs->sync_target);
572 		spdk_io_device_unregister(&fs->io_target);
573 		free(fs);
574 		cb_fn(cb_arg, NULL, -ENOMEM);
575 		return;
576 	}
577 
578 	args = &req->args;
579 	args->fn.fs_op_with_handle = cb_fn;
580 	args->arg = cb_arg;
581 	args->fs = fs;
582 
583 	spdk_bs_load(dev, load_cb, req);
584 }
585 
586 static void
587 unload_cb(void *ctx, int bserrno)
588 {
589 	struct spdk_fs_request *req = ctx;
590 	struct spdk_fs_cb_args *args = &req->args;
591 	struct spdk_filesystem *fs = args->fs;
592 
593 	pthread_mutex_lock(&g_cache_init_lock);
594 	g_fs_count--;
595 	if (g_fs_count == 0) {
596 		__free_cache();
597 	}
598 	pthread_mutex_unlock(&g_cache_init_lock);
599 
600 	args->fn.fs_op(args->arg, bserrno);
601 	free(req);
602 
603 	spdk_io_device_unregister(&fs->io_target);
604 	spdk_io_device_unregister(&fs->sync_target);
605 	spdk_io_device_unregister(&fs->md_target);
606 
607 	free(fs);
608 }
609 
610 void
611 spdk_fs_unload(struct spdk_filesystem *fs, spdk_fs_op_complete cb_fn, void *cb_arg)
612 {
613 	struct spdk_fs_request *req;
614 	struct spdk_fs_cb_args *args;
615 
616 	/*
617 	 * We must free the md_channel before unloading the blobstore, so just
618 	 *  allocate this request from the general heap.
619 	 */
620 	req = calloc(1, sizeof(*req));
621 	if (req == NULL) {
622 		cb_fn(cb_arg, -ENOMEM);
623 		return;
624 	}
625 
626 	args = &req->args;
627 	args->fn.fs_op = cb_fn;
628 	args->arg = cb_arg;
629 	args->fs = fs;
630 
631 	spdk_fs_free_io_channel(fs->md_target.md_io_channel);
632 	spdk_fs_free_io_channel(fs->sync_target.sync_io_channel);
633 	spdk_bs_unload(fs->bs, unload_cb, req);
634 }
635 
636 static struct spdk_file *
637 fs_find_file(struct spdk_filesystem *fs, const char *name)
638 {
639 	struct spdk_file *file;
640 
641 	TAILQ_FOREACH(file, &fs->files, tailq) {
642 		if (!strncmp(name, file->name, SPDK_FILE_NAME_MAX)) {
643 			return file;
644 		}
645 	}
646 
647 	return NULL;
648 }
649 
650 void
651 spdk_fs_file_stat_async(struct spdk_filesystem *fs, const char *name,
652 			spdk_file_stat_op_complete cb_fn, void *cb_arg)
653 {
654 	struct spdk_file_stat stat;
655 	struct spdk_file *f = NULL;
656 
657 	if (strnlen(name, SPDK_FILE_NAME_MAX + 1) == SPDK_FILE_NAME_MAX + 1) {
658 		cb_fn(cb_arg, NULL, -ENAMETOOLONG);
659 		return;
660 	}
661 
662 	f = fs_find_file(fs, name);
663 	if (f != NULL) {
664 		stat.blobid = f->blobid;
665 		stat.size = f->length;
666 		cb_fn(cb_arg, &stat, 0);
667 		return;
668 	}
669 
670 	cb_fn(cb_arg, NULL, -ENOENT);
671 }
672 
673 static void
674 __copy_stat(void *arg, struct spdk_file_stat *stat, int fserrno)
675 {
676 	struct spdk_fs_request *req = arg;
677 	struct spdk_fs_cb_args *args = &req->args;
678 
679 	args->rc = fserrno;
680 	if (fserrno == 0) {
681 		memcpy(args->arg, stat, sizeof(*stat));
682 	}
683 	sem_post(args->sem);
684 }
685 
686 static void
687 __file_stat(void *arg)
688 {
689 	struct spdk_fs_request *req = arg;
690 	struct spdk_fs_cb_args *args = &req->args;
691 
692 	spdk_fs_file_stat_async(args->fs, args->op.stat.name,
693 				args->fn.stat_op, req);
694 }
695 
696 int
697 spdk_fs_file_stat(struct spdk_filesystem *fs, struct spdk_io_channel *_channel,
698 		  const char *name, struct spdk_file_stat *stat)
699 {
700 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
701 	struct spdk_fs_request *req;
702 	int rc;
703 
704 	req = alloc_fs_request(channel);
705 	assert(req != NULL);
706 
707 	req->args.fs = fs;
708 	req->args.op.stat.name = name;
709 	req->args.fn.stat_op = __copy_stat;
710 	req->args.arg = stat;
711 	req->args.sem = &channel->sem;
712 	channel->send_request(__file_stat, req);
713 	sem_wait(&channel->sem);
714 
715 	rc = req->args.rc;
716 	free_fs_request(req);
717 
718 	return rc;
719 }
720 
721 static void
722 fs_create_blob_close_cb(void *ctx, int bserrno)
723 {
724 	struct spdk_fs_request *req = ctx;
725 	struct spdk_fs_cb_args *args = &req->args;
726 
727 	args->fn.file_op(args->arg, bserrno);
728 	free_fs_request(req);
729 }
730 
731 static void
732 fs_create_blob_open_cb(void *ctx, struct spdk_blob *blob, int bserrno)
733 {
734 	struct spdk_fs_request *req = ctx;
735 	struct spdk_fs_cb_args *args = &req->args;
736 	struct spdk_file *f = args->file;
737 	uint64_t length = 0;
738 
739 	f->blob = blob;
740 	spdk_bs_md_resize_blob(blob, 1);
741 	spdk_blob_md_set_xattr(blob, "name", f->name, strlen(f->name) + 1);
742 	spdk_blob_md_set_xattr(blob, "length", &length, sizeof(length));
743 
744 	spdk_bs_md_close_blob(&f->blob, fs_create_blob_close_cb, args);
745 }
746 
747 static void
748 fs_create_blob_create_cb(void *ctx, spdk_blob_id blobid, int bserrno)
749 {
750 	struct spdk_fs_request *req = ctx;
751 	struct spdk_fs_cb_args *args = &req->args;
752 	struct spdk_file *f = args->file;
753 
754 	f->blobid = blobid;
755 	spdk_bs_md_open_blob(f->fs->bs, blobid, fs_create_blob_open_cb, req);
756 }
757 
758 void
759 spdk_fs_create_file_async(struct spdk_filesystem *fs, const char *name,
760 			  spdk_file_op_complete cb_fn, void *cb_arg)
761 {
762 	struct spdk_file *file;
763 	struct spdk_fs_request *req;
764 	struct spdk_fs_cb_args *args;
765 
766 	if (strnlen(name, SPDK_FILE_NAME_MAX + 1) == SPDK_FILE_NAME_MAX + 1) {
767 		cb_fn(cb_arg, -ENAMETOOLONG);
768 		return;
769 	}
770 
771 	file = fs_find_file(fs, name);
772 	if (file != NULL) {
773 		cb_fn(cb_arg, -EEXIST);
774 		return;
775 	}
776 
777 	file = file_alloc(fs);
778 	if (file == NULL) {
779 		cb_fn(cb_arg, -ENOMEM);
780 		return;
781 	}
782 
783 	req = alloc_fs_request(fs->md_target.md_fs_channel);
784 	if (req == NULL) {
785 		cb_fn(cb_arg, -ENOMEM);
786 		return;
787 	}
788 
789 	args = &req->args;
790 	args->file = file;
791 	args->fn.file_op = cb_fn;
792 	args->arg = cb_arg;
793 
794 	file->name = strdup(name);
795 	spdk_bs_md_create_blob(fs->bs, fs_create_blob_create_cb, args);
796 }
797 
798 static void
799 __fs_create_file_done(void *arg, int fserrno)
800 {
801 	struct spdk_fs_request *req = arg;
802 	struct spdk_fs_cb_args *args = &req->args;
803 
804 	args->rc = fserrno;
805 	sem_post(args->sem);
806 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s\n", args->op.create.name);
807 }
808 
809 static void
810 __fs_create_file(void *arg)
811 {
812 	struct spdk_fs_request *req = arg;
813 	struct spdk_fs_cb_args *args = &req->args;
814 
815 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s\n", args->op.create.name);
816 	spdk_fs_create_file_async(args->fs, args->op.create.name, __fs_create_file_done, req);
817 }
818 
819 int
820 spdk_fs_create_file(struct spdk_filesystem *fs, struct spdk_io_channel *_channel, const char *name)
821 {
822 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
823 	struct spdk_fs_request *req;
824 	struct spdk_fs_cb_args *args;
825 	int rc;
826 
827 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s\n", name);
828 
829 	req = alloc_fs_request(channel);
830 	assert(req != NULL);
831 
832 	args = &req->args;
833 	args->fs = fs;
834 	args->op.create.name = name;
835 	args->sem = &channel->sem;
836 	fs->send_request(__fs_create_file, req);
837 	sem_wait(&channel->sem);
838 	rc = args->rc;
839 	free_fs_request(req);
840 
841 	return rc;
842 }
843 
844 static void
845 fs_open_blob_done(void *ctx, struct spdk_blob *blob, int bserrno)
846 {
847 	struct spdk_fs_request *req = ctx;
848 	struct spdk_fs_cb_args *args = &req->args;
849 	struct spdk_file *f = args->file;
850 
851 	f->blob = blob;
852 	while (!TAILQ_EMPTY(&f->open_requests)) {
853 		req = TAILQ_FIRST(&f->open_requests);
854 		args = &req->args;
855 		TAILQ_REMOVE(&f->open_requests, req, args.op.open.tailq);
856 		args->fn.file_op_with_handle(args->arg, f, bserrno);
857 		free_fs_request(req);
858 	}
859 }
860 
861 static void
862 fs_open_blob_create_cb(void *ctx, int bserrno)
863 {
864 	struct spdk_fs_request *req = ctx;
865 	struct spdk_fs_cb_args *args = &req->args;
866 	struct spdk_file *file = args->file;
867 	struct spdk_filesystem *fs = args->fs;
868 
869 	if (file == NULL) {
870 		/*
871 		 * This is from an open with CREATE flag - the file
872 		 *  is now created so look it up in the file list for this
873 		 *  filesystem.
874 		 */
875 		file = fs_find_file(fs, args->op.open.name);
876 		assert(file != NULL);
877 		args->file = file;
878 	}
879 
880 	file->ref_count++;
881 	TAILQ_INSERT_TAIL(&file->open_requests, req, args.op.open.tailq);
882 	if (file->ref_count == 1) {
883 		assert(file->blob == NULL);
884 		spdk_bs_md_open_blob(fs->bs, file->blobid, fs_open_blob_done, req);
885 	} else if (file->blob != NULL) {
886 		fs_open_blob_done(req, file->blob, 0);
887 	} else {
888 		/*
889 		 * The blob open for this file is in progress due to a previous
890 		 *  open request.  When that open completes, it will invoke the
891 		 *  open callback for this request.
892 		 */
893 	}
894 }
895 
896 void
897 spdk_fs_open_file_async(struct spdk_filesystem *fs, const char *name, uint32_t flags,
898 			spdk_file_op_with_handle_complete cb_fn, void *cb_arg)
899 {
900 	struct spdk_file *f = NULL;
901 	struct spdk_fs_request *req;
902 	struct spdk_fs_cb_args *args;
903 
904 	if (strnlen(name, SPDK_FILE_NAME_MAX + 1) == SPDK_FILE_NAME_MAX + 1) {
905 		cb_fn(cb_arg, NULL, -ENAMETOOLONG);
906 		return;
907 	}
908 
909 	f = fs_find_file(fs, name);
910 	if (f == NULL && !(flags & SPDK_BLOBFS_OPEN_CREATE)) {
911 		cb_fn(cb_arg, NULL, -ENOENT);
912 		return;
913 	}
914 
915 	req = alloc_fs_request(fs->md_target.md_fs_channel);
916 	if (req == NULL) {
917 		cb_fn(cb_arg, NULL, -ENOMEM);
918 		return;
919 	}
920 
921 	args = &req->args;
922 	args->fn.file_op_with_handle = cb_fn;
923 	args->arg = cb_arg;
924 	args->file = f;
925 	args->fs = fs;
926 	args->op.open.name = name;
927 
928 	if (f == NULL) {
929 		spdk_fs_create_file_async(fs, name, fs_open_blob_create_cb, req);
930 	} else {
931 		fs_open_blob_create_cb(req, 0);
932 	}
933 }
934 
935 static void
936 __fs_open_file_done(void *arg, struct spdk_file *file, int bserrno)
937 {
938 	struct spdk_fs_request *req = arg;
939 	struct spdk_fs_cb_args *args = &req->args;
940 
941 	args->file = file;
942 	args->rc = bserrno;
943 	sem_post(args->sem);
944 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s\n", args->op.open.name);
945 }
946 
947 static void
948 __fs_open_file(void *arg)
949 {
950 	struct spdk_fs_request *req = arg;
951 	struct spdk_fs_cb_args *args = &req->args;
952 
953 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s\n", args->op.open.name);
954 	spdk_fs_open_file_async(args->fs, args->op.open.name, args->op.open.flags,
955 				__fs_open_file_done, req);
956 }
957 
958 int
959 spdk_fs_open_file(struct spdk_filesystem *fs, struct spdk_io_channel *_channel,
960 		  const char *name, uint32_t flags, struct spdk_file **file)
961 {
962 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
963 	struct spdk_fs_request *req;
964 	struct spdk_fs_cb_args *args;
965 	int rc;
966 
967 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s\n", name);
968 
969 	req = alloc_fs_request(channel);
970 	assert(req != NULL);
971 
972 	args = &req->args;
973 	args->fs = fs;
974 	args->op.open.name = name;
975 	args->op.open.flags = flags;
976 	args->sem = &channel->sem;
977 	fs->send_request(__fs_open_file, req);
978 	sem_wait(&channel->sem);
979 	rc = args->rc;
980 	if (rc == 0) {
981 		*file = args->file;
982 	} else {
983 		*file = NULL;
984 	}
985 	free_fs_request(req);
986 
987 	return rc;
988 }
989 
990 static void
991 fs_rename_blob_close_cb(void *ctx, int bserrno)
992 {
993 	struct spdk_fs_request *req = ctx;
994 	struct spdk_fs_cb_args *args = &req->args;
995 
996 	args->fn.fs_op(args->arg, bserrno);
997 	free_fs_request(req);
998 }
999 
1000 static void
1001 fs_rename_blob_open_cb(void *ctx, struct spdk_blob *blob, int bserrno)
1002 {
1003 	struct spdk_fs_request *req = ctx;
1004 	struct spdk_fs_cb_args *args = &req->args;
1005 	struct spdk_file *f = args->file;
1006 	const char *new_name = args->op.rename.new_name;
1007 
1008 	f->blob = blob;
1009 	spdk_blob_md_set_xattr(blob, "name", new_name, strlen(new_name) + 1);
1010 	spdk_bs_md_close_blob(&f->blob, fs_rename_blob_close_cb, req);
1011 }
1012 
1013 static void
1014 __spdk_fs_md_rename_file(struct spdk_fs_request *req)
1015 {
1016 	struct spdk_fs_cb_args *args = &req->args;
1017 	struct spdk_file *f;
1018 
1019 	f = fs_find_file(args->fs, args->op.rename.old_name);
1020 	if (f == NULL) {
1021 		args->fn.fs_op(args->arg, -ENOENT);
1022 		free_fs_request(req);
1023 		return;
1024 	}
1025 
1026 	free(f->name);
1027 	f->name = strdup(args->op.rename.new_name);
1028 	args->file = f;
1029 	spdk_bs_md_open_blob(args->fs->bs, f->blobid, fs_rename_blob_open_cb, req);
1030 }
1031 
1032 static void
1033 fs_rename_delete_done(void *arg, int fserrno)
1034 {
1035 	__spdk_fs_md_rename_file(arg);
1036 }
1037 
1038 void
1039 spdk_fs_rename_file_async(struct spdk_filesystem *fs,
1040 			  const char *old_name, const char *new_name,
1041 			  spdk_file_op_complete cb_fn, void *cb_arg)
1042 {
1043 	struct spdk_file *f;
1044 	struct spdk_fs_request *req;
1045 	struct spdk_fs_cb_args *args;
1046 
1047 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "old=%s new=%s\n", old_name, new_name);
1048 	if (strnlen(new_name, SPDK_FILE_NAME_MAX + 1) == SPDK_FILE_NAME_MAX + 1) {
1049 		cb_fn(cb_arg, -ENAMETOOLONG);
1050 		return;
1051 	}
1052 
1053 	req = alloc_fs_request(fs->md_target.md_fs_channel);
1054 	if (req == NULL) {
1055 		cb_fn(cb_arg, -ENOMEM);
1056 		return;
1057 	}
1058 
1059 	args = &req->args;
1060 	args->fn.fs_op = cb_fn;
1061 	args->fs = fs;
1062 	args->arg = cb_arg;
1063 	args->op.rename.old_name = old_name;
1064 	args->op.rename.new_name = new_name;
1065 
1066 	f = fs_find_file(fs, new_name);
1067 	if (f == NULL) {
1068 		__spdk_fs_md_rename_file(req);
1069 		return;
1070 	}
1071 
1072 	/*
1073 	 * The rename overwrites an existing file.  So delete the existing file, then
1074 	 *  do the actual rename.
1075 	 */
1076 	spdk_fs_delete_file_async(fs, new_name, fs_rename_delete_done, req);
1077 }
1078 
1079 static void
1080 __fs_rename_file_done(void *arg, int fserrno)
1081 {
1082 	struct spdk_fs_request *req = arg;
1083 	struct spdk_fs_cb_args *args = &req->args;
1084 
1085 	args->rc = fserrno;
1086 	sem_post(args->sem);
1087 }
1088 
1089 static void
1090 __fs_rename_file(void *arg)
1091 {
1092 	struct spdk_fs_request *req = arg;
1093 	struct spdk_fs_cb_args *args = &req->args;
1094 
1095 	spdk_fs_rename_file_async(args->fs, args->op.rename.old_name, args->op.rename.new_name,
1096 				  __fs_rename_file_done, req);
1097 }
1098 
1099 int
1100 spdk_fs_rename_file(struct spdk_filesystem *fs, struct spdk_io_channel *_channel,
1101 		    const char *old_name, const char *new_name)
1102 {
1103 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
1104 	struct spdk_fs_request *req;
1105 	struct spdk_fs_cb_args *args;
1106 	int rc;
1107 
1108 	req = alloc_fs_request(channel);
1109 	assert(req != NULL);
1110 
1111 	args = &req->args;
1112 
1113 	args->fs = fs;
1114 	args->op.rename.old_name = old_name;
1115 	args->op.rename.new_name = new_name;
1116 	args->sem = &channel->sem;
1117 	fs->send_request(__fs_rename_file, req);
1118 	sem_wait(&channel->sem);
1119 	rc = args->rc;
1120 	free_fs_request(req);
1121 	return rc;
1122 }
1123 
1124 static void
1125 blob_delete_cb(void *ctx, int bserrno)
1126 {
1127 	struct spdk_fs_request *req = ctx;
1128 	struct spdk_fs_cb_args *args = &req->args;
1129 
1130 	args->fn.file_op(args->arg, bserrno);
1131 	free_fs_request(req);
1132 }
1133 
1134 void
1135 spdk_fs_delete_file_async(struct spdk_filesystem *fs, const char *name,
1136 			  spdk_file_op_complete cb_fn, void *cb_arg)
1137 {
1138 	struct spdk_file *f;
1139 	spdk_blob_id blobid;
1140 	struct spdk_fs_request *req;
1141 	struct spdk_fs_cb_args *args;
1142 
1143 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s\n", name);
1144 
1145 	if (strnlen(name, SPDK_FILE_NAME_MAX + 1) == SPDK_FILE_NAME_MAX + 1) {
1146 		cb_fn(cb_arg, -ENAMETOOLONG);
1147 		return;
1148 	}
1149 
1150 	f = fs_find_file(fs, name);
1151 	if (f == NULL) {
1152 		cb_fn(cb_arg, -ENOENT);
1153 		return;
1154 	}
1155 
1156 	if (f->ref_count > 0) {
1157 		/* For now, do not allow deleting files with open references. */
1158 		cb_fn(cb_arg, -EBUSY);
1159 		return;
1160 	}
1161 
1162 	req = alloc_fs_request(fs->md_target.md_fs_channel);
1163 	if (req == NULL) {
1164 		cb_fn(cb_arg, -ENOMEM);
1165 		return;
1166 	}
1167 
1168 	TAILQ_REMOVE(&fs->files, f, tailq);
1169 
1170 	cache_free_buffers(f);
1171 
1172 	blobid = f->blobid;
1173 
1174 	free(f->name);
1175 	free(f->tree);
1176 	free(f);
1177 
1178 	args = &req->args;
1179 	args->fn.file_op = cb_fn;
1180 	args->arg = cb_arg;
1181 	spdk_bs_md_delete_blob(fs->bs, blobid, blob_delete_cb, req);
1182 }
1183 
1184 static void
1185 __fs_delete_file_done(void *arg, int fserrno)
1186 {
1187 	struct spdk_fs_request *req = arg;
1188 	struct spdk_fs_cb_args *args = &req->args;
1189 
1190 	args->rc = fserrno;
1191 	sem_post(args->sem);
1192 }
1193 
1194 static void
1195 __fs_delete_file(void *arg)
1196 {
1197 	struct spdk_fs_request *req = arg;
1198 	struct spdk_fs_cb_args *args = &req->args;
1199 
1200 	spdk_fs_delete_file_async(args->fs, args->op.delete.name, __fs_delete_file_done, req);
1201 }
1202 
1203 int
1204 spdk_fs_delete_file(struct spdk_filesystem *fs, struct spdk_io_channel *_channel,
1205 		    const char *name)
1206 {
1207 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
1208 	struct spdk_fs_request *req;
1209 	struct spdk_fs_cb_args *args;
1210 	int rc;
1211 
1212 	req = alloc_fs_request(channel);
1213 	assert(req != NULL);
1214 
1215 	args = &req->args;
1216 	args->fs = fs;
1217 	args->op.delete.name = name;
1218 	args->sem = &channel->sem;
1219 	fs->send_request(__fs_delete_file, req);
1220 	sem_wait(&channel->sem);
1221 	rc = args->rc;
1222 	free_fs_request(req);
1223 
1224 	return rc;
1225 }
1226 
1227 spdk_fs_iter
1228 spdk_fs_iter_first(struct spdk_filesystem *fs)
1229 {
1230 	struct spdk_file *f;
1231 
1232 	f = TAILQ_FIRST(&fs->files);
1233 	return f;
1234 }
1235 
1236 spdk_fs_iter
1237 spdk_fs_iter_next(spdk_fs_iter iter)
1238 {
1239 	struct spdk_file *f = iter;
1240 
1241 	if (f == NULL) {
1242 		return NULL;
1243 	}
1244 
1245 	f = TAILQ_NEXT(f, tailq);
1246 	return f;
1247 }
1248 
1249 const char *
1250 spdk_file_get_name(struct spdk_file *file)
1251 {
1252 	return file->name;
1253 }
1254 
1255 uint64_t
1256 spdk_file_get_length(struct spdk_file *file)
1257 {
1258 	assert(file != NULL);
1259 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s length=0x%jx\n", file->name, file->length);
1260 	return file->length;
1261 }
1262 
1263 static void
1264 fs_truncate_complete_cb(void *ctx, int bserrno)
1265 {
1266 	struct spdk_fs_request *req = ctx;
1267 	struct spdk_fs_cb_args *args = &req->args;
1268 
1269 	args->fn.file_op(args->arg, bserrno);
1270 	free_fs_request(req);
1271 }
1272 
1273 static uint64_t
1274 __bytes_to_clusters(uint64_t length, uint64_t cluster_sz)
1275 {
1276 	return (length + cluster_sz - 1) / cluster_sz;
1277 }
1278 
1279 void
1280 spdk_file_truncate_async(struct spdk_file *file, uint64_t length,
1281 			 spdk_file_op_complete cb_fn, void *cb_arg)
1282 {
1283 	struct spdk_filesystem *fs;
1284 	size_t num_clusters;
1285 	struct spdk_fs_request *req;
1286 	struct spdk_fs_cb_args *args;
1287 
1288 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s old=0x%jx new=0x%jx\n", file->name, file->length, length);
1289 	if (length == file->length) {
1290 		cb_fn(cb_arg, 0);
1291 		return;
1292 	}
1293 
1294 	req = alloc_fs_request(file->fs->md_target.md_fs_channel);
1295 	if (req == NULL) {
1296 		cb_fn(cb_arg, -ENOMEM);
1297 		return;
1298 	}
1299 
1300 	args = &req->args;
1301 	args->fn.file_op = cb_fn;
1302 	args->arg = cb_arg;
1303 	args->file = file;
1304 	fs = file->fs;
1305 
1306 	num_clusters = __bytes_to_clusters(length, fs->bs_opts.cluster_sz);
1307 
1308 	spdk_bs_md_resize_blob(file->blob, num_clusters);
1309 	spdk_blob_md_set_xattr(file->blob, "length", &length, sizeof(length));
1310 
1311 	file->length = length;
1312 	if (file->append_pos > file->length) {
1313 		file->append_pos = file->length;
1314 	}
1315 
1316 	spdk_bs_md_sync_blob(file->blob, fs_truncate_complete_cb, args);
1317 }
1318 
1319 static void
1320 __truncate(void *arg)
1321 {
1322 	struct spdk_fs_request *req = arg;
1323 	struct spdk_fs_cb_args *args = &req->args;
1324 
1325 	spdk_file_truncate_async(args->file, args->op.truncate.length,
1326 				 args->fn.file_op, args->arg);
1327 }
1328 
1329 void
1330 spdk_file_truncate(struct spdk_file *file, struct spdk_io_channel *_channel,
1331 		   uint64_t length)
1332 {
1333 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
1334 	struct spdk_fs_request *req;
1335 	struct spdk_fs_cb_args *args;
1336 
1337 	req = alloc_fs_request(channel);
1338 	assert(req != NULL);
1339 
1340 	args = &req->args;
1341 
1342 	args->file = file;
1343 	args->op.truncate.length = length;
1344 	args->fn.file_op = __sem_post;
1345 	args->arg = &channel->sem;
1346 
1347 	channel->send_request(__truncate, req);
1348 	sem_wait(&channel->sem);
1349 	free_fs_request(req);
1350 }
1351 
1352 static void
1353 __rw_done(void *ctx, int bserrno)
1354 {
1355 	struct spdk_fs_request *req = ctx;
1356 	struct spdk_fs_cb_args *args = &req->args;
1357 
1358 	spdk_dma_free(args->op.rw.pin_buf);
1359 	args->fn.file_op(args->arg, bserrno);
1360 	free_fs_request(req);
1361 }
1362 
1363 static void
1364 __read_done(void *ctx, int bserrno)
1365 {
1366 	struct spdk_fs_request *req = ctx;
1367 	struct spdk_fs_cb_args *args = &req->args;
1368 
1369 	if (args->op.rw.is_read) {
1370 		memcpy(args->op.rw.user_buf,
1371 		       args->op.rw.pin_buf + (args->op.rw.offset & 0xFFF),
1372 		       args->op.rw.length);
1373 		__rw_done(req, 0);
1374 	} else {
1375 		memcpy(args->op.rw.pin_buf + (args->op.rw.offset & 0xFFF),
1376 		       args->op.rw.user_buf,
1377 		       args->op.rw.length);
1378 		spdk_bs_io_write_blob(args->file->blob, args->op.rw.channel,
1379 				      args->op.rw.pin_buf,
1380 				      args->op.rw.start_page, args->op.rw.num_pages,
1381 				      __rw_done, req);
1382 	}
1383 }
1384 
1385 static void
1386 __do_blob_read(void *ctx, int fserrno)
1387 {
1388 	struct spdk_fs_request *req = ctx;
1389 	struct spdk_fs_cb_args *args = &req->args;
1390 
1391 	spdk_bs_io_read_blob(args->file->blob, args->op.rw.channel,
1392 			     args->op.rw.pin_buf,
1393 			     args->op.rw.start_page, args->op.rw.num_pages,
1394 			     __read_done, req);
1395 }
1396 
1397 static void
1398 __get_page_parameters(struct spdk_file *file, uint64_t offset, uint64_t length,
1399 		      uint64_t *start_page, uint32_t *page_size, uint64_t *num_pages)
1400 {
1401 	uint64_t end_page;
1402 
1403 	*page_size = spdk_bs_get_page_size(file->fs->bs);
1404 	*start_page = offset / *page_size;
1405 	end_page = (offset + length - 1) / *page_size;
1406 	*num_pages = (end_page - *start_page + 1);
1407 }
1408 
1409 static void
1410 __readwrite(struct spdk_file *file, struct spdk_io_channel *_channel,
1411 	    void *payload, uint64_t offset, uint64_t length,
1412 	    spdk_file_op_complete cb_fn, void *cb_arg, int is_read)
1413 {
1414 	struct spdk_fs_request *req;
1415 	struct spdk_fs_cb_args *args;
1416 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
1417 	uint64_t start_page, num_pages, pin_buf_length;
1418 	uint32_t page_size;
1419 
1420 	if (is_read && offset + length > file->length) {
1421 		cb_fn(cb_arg, -EINVAL);
1422 		return;
1423 	}
1424 
1425 	req = alloc_fs_request(channel);
1426 	if (req == NULL) {
1427 		cb_fn(cb_arg, -ENOMEM);
1428 		return;
1429 	}
1430 
1431 	args = &req->args;
1432 	args->fn.file_op = cb_fn;
1433 	args->arg = cb_arg;
1434 	args->file = file;
1435 	args->op.rw.channel = channel->bs_channel;
1436 	args->op.rw.user_buf = payload;
1437 	args->op.rw.is_read = is_read;
1438 	args->op.rw.offset = offset;
1439 	args->op.rw.length = length;
1440 
1441 	__get_page_parameters(file, offset, length, &start_page, &page_size, &num_pages);
1442 	pin_buf_length = num_pages * page_size;
1443 	args->op.rw.pin_buf = spdk_dma_malloc(pin_buf_length, 4096, NULL);
1444 
1445 	args->op.rw.start_page = start_page;
1446 	args->op.rw.num_pages = num_pages;
1447 
1448 	if (!is_read && file->length < offset + length) {
1449 		spdk_file_truncate_async(file, offset + length, __do_blob_read, req);
1450 	} else {
1451 		__do_blob_read(req, 0);
1452 	}
1453 }
1454 
1455 void
1456 spdk_file_write_async(struct spdk_file *file, struct spdk_io_channel *channel,
1457 		      void *payload, uint64_t offset, uint64_t length,
1458 		      spdk_file_op_complete cb_fn, void *cb_arg)
1459 {
1460 	__readwrite(file, channel, payload, offset, length, cb_fn, cb_arg, 0);
1461 }
1462 
1463 void
1464 spdk_file_read_async(struct spdk_file *file, struct spdk_io_channel *channel,
1465 		     void *payload, uint64_t offset, uint64_t length,
1466 		     spdk_file_op_complete cb_fn, void *cb_arg)
1467 {
1468 	SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "file=%s offset=%jx length=%jx\n",
1469 		      file->name, offset, length);
1470 	__readwrite(file, channel, payload, offset, length, cb_fn, cb_arg, 1);
1471 }
1472 
1473 struct spdk_io_channel *
1474 spdk_fs_alloc_io_channel(struct spdk_filesystem *fs)
1475 {
1476 	struct spdk_io_channel *io_channel;
1477 	struct spdk_fs_channel *fs_channel;
1478 
1479 	io_channel = spdk_get_io_channel(&fs->io_target);
1480 	fs_channel = spdk_io_channel_get_ctx(io_channel);
1481 	fs_channel->bs_channel = spdk_bs_alloc_io_channel(fs->bs);
1482 	fs_channel->send_request = __send_request_direct;
1483 
1484 	return io_channel;
1485 }
1486 
1487 struct spdk_io_channel *
1488 spdk_fs_alloc_io_channel_sync(struct spdk_filesystem *fs)
1489 {
1490 	struct spdk_io_channel *io_channel;
1491 	struct spdk_fs_channel *fs_channel;
1492 
1493 	io_channel = spdk_get_io_channel(&fs->io_target);
1494 	fs_channel = spdk_io_channel_get_ctx(io_channel);
1495 	fs_channel->send_request = fs->send_request;
1496 	fs_channel->sync = 1;
1497 	pthread_spin_init(&fs_channel->lock, 0);
1498 
1499 	return io_channel;
1500 }
1501 
1502 void
1503 spdk_fs_free_io_channel(struct spdk_io_channel *channel)
1504 {
1505 	spdk_put_io_channel(channel);
1506 }
1507 
1508 void
1509 spdk_fs_set_cache_size(uint64_t size_in_mb)
1510 {
1511 	g_fs_cache_size = size_in_mb * 1024 * 1024;
1512 }
1513 
1514 uint64_t
1515 spdk_fs_get_cache_size(void)
1516 {
1517 	return g_fs_cache_size / (1024 * 1024);
1518 }
1519 
1520 static void __file_flush(void *_args);
1521 
1522 static void *
1523 alloc_cache_memory_buffer(struct spdk_file *context)
1524 {
1525 	struct spdk_file *file;
1526 	void *buf;
1527 
1528 	buf = spdk_mempool_get(g_cache_pool);
1529 	if (buf != NULL) {
1530 		return buf;
1531 	}
1532 
1533 	pthread_spin_lock(&g_caches_lock);
1534 	TAILQ_FOREACH(file, &g_caches, cache_tailq) {
1535 		if (!file->open_for_writing &&
1536 		    file->priority == SPDK_FILE_PRIORITY_LOW &&
1537 		    file != context) {
1538 			TAILQ_REMOVE(&g_caches, file, cache_tailq);
1539 			TAILQ_INSERT_TAIL(&g_caches, file, cache_tailq);
1540 			break;
1541 		}
1542 	}
1543 	pthread_spin_unlock(&g_caches_lock);
1544 	if (file != NULL) {
1545 		cache_free_buffers(file);
1546 		buf = spdk_mempool_get(g_cache_pool);
1547 		if (buf != NULL) {
1548 			return buf;
1549 		}
1550 	}
1551 
1552 	pthread_spin_lock(&g_caches_lock);
1553 	TAILQ_FOREACH(file, &g_caches, cache_tailq) {
1554 		if (!file->open_for_writing && file != context) {
1555 			TAILQ_REMOVE(&g_caches, file, cache_tailq);
1556 			TAILQ_INSERT_TAIL(&g_caches, file, cache_tailq);
1557 			break;
1558 		}
1559 	}
1560 	pthread_spin_unlock(&g_caches_lock);
1561 	if (file != NULL) {
1562 		cache_free_buffers(file);
1563 		buf = spdk_mempool_get(g_cache_pool);
1564 		if (buf != NULL) {
1565 			return buf;
1566 		}
1567 	}
1568 
1569 	pthread_spin_lock(&g_caches_lock);
1570 	TAILQ_FOREACH(file, &g_caches, cache_tailq) {
1571 		if (file != context) {
1572 			TAILQ_REMOVE(&g_caches, file, cache_tailq);
1573 			TAILQ_INSERT_TAIL(&g_caches, file, cache_tailq);
1574 			break;
1575 		}
1576 	}
1577 	pthread_spin_unlock(&g_caches_lock);
1578 	if (file != NULL) {
1579 		cache_free_buffers(file);
1580 		buf = spdk_mempool_get(g_cache_pool);
1581 		if (buf != NULL) {
1582 			return buf;
1583 		}
1584 	}
1585 
1586 	assert(false);
1587 	return NULL;
1588 }
1589 
1590 static struct cache_buffer *
1591 cache_insert_buffer(struct spdk_file *file, uint64_t offset)
1592 {
1593 	struct cache_buffer *buf;
1594 	int count = 0;
1595 
1596 	buf = calloc(1, sizeof(*buf));
1597 	if (buf == NULL) {
1598 		SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "calloc failed\n");
1599 		return NULL;
1600 	}
1601 
1602 	buf->buf = alloc_cache_memory_buffer(file);
1603 	if (buf->buf == NULL) {
1604 		while (buf->buf == NULL) {
1605 			/*
1606 			 * TODO: alloc_cache_memory_buffer() should eventually free
1607 			 *  some buffers.  Need a more sophisticated check here, instead
1608 			 *  of just bailing if 100 tries does not result in getting a
1609 			 *  free buffer.  This will involve using the sync channel's
1610 			 *  semaphore to block until a buffer becomes available.
1611 			 */
1612 			if (count++ == 100) {
1613 				SPDK_ERRLOG("could not allocate cache buffer\n");
1614 				assert(false);
1615 				free(buf);
1616 				return NULL;
1617 			}
1618 			buf->buf = alloc_cache_memory_buffer(file);
1619 		}
1620 	}
1621 
1622 	buf->buf_size = CACHE_BUFFER_SIZE;
1623 	buf->offset = offset;
1624 
1625 	pthread_spin_lock(&g_caches_lock);
1626 	if (file->tree->present_mask == 0) {
1627 		TAILQ_INSERT_TAIL(&g_caches, file, cache_tailq);
1628 	}
1629 	file->tree = spdk_tree_insert_buffer(file->tree, buf);
1630 	pthread_spin_unlock(&g_caches_lock);
1631 
1632 	return buf;
1633 }
1634 
1635 static struct cache_buffer *
1636 cache_append_buffer(struct spdk_file *file)
1637 {
1638 	struct cache_buffer *last;
1639 
1640 	assert(file->last == NULL || file->last->bytes_filled == file->last->buf_size);
1641 	assert((file->append_pos % CACHE_BUFFER_SIZE) == 0);
1642 
1643 	last = cache_insert_buffer(file, file->append_pos);
1644 	if (last == NULL) {
1645 		SPDK_TRACELOG(SPDK_TRACE_BLOBFS, "cache_insert_buffer failed\n");
1646 		return NULL;
1647 	}
1648 
1649 	if (file->last != NULL) {
1650 		file->last->next = last;
1651 	}
1652 	file->last = last;
1653 
1654 	return last;
1655 }
1656 
1657 static void
1658 __wake_caller(struct spdk_fs_cb_args *args)
1659 {
1660 	sem_post(args->sem);
1661 }
1662 
1663 static void __check_sync_reqs(struct spdk_file *file);
1664 
1665 static void
1666 __file_cache_finish_sync(struct spdk_file *file)
1667 {
1668 	struct spdk_fs_request *sync_req;
1669 	struct spdk_fs_cb_args *sync_args;
1670 
1671 	pthread_spin_lock(&file->lock);
1672 	sync_req = TAILQ_FIRST(&file->sync_requests);
1673 	sync_args = &sync_req->args;
1674 	assert(sync_args->op.sync.offset <= file->length_flushed);
1675 	BLOBFS_TRACE(file, "sync done offset=%jx\n", sync_args->op.sync.offset);
1676 	TAILQ_REMOVE(&file->sync_requests, sync_req, args.op.sync.tailq);
1677 	pthread_spin_unlock(&file->lock);
1678 
1679 	sync_args->fn.file_op(sync_args->arg, 0);
1680 	__check_sync_reqs(file);
1681 
1682 	pthread_spin_lock(&file->lock);
1683 	free_fs_request(sync_req);
1684 	pthread_spin_unlock(&file->lock);
1685 }
1686 
1687 static void
1688 __file_cache_finish_sync_bs_cb(void *ctx, int bserrno)
1689 {
1690 	struct spdk_file *file = ctx;
1691 
1692 	__file_cache_finish_sync(file);
1693 }
1694 
1695 static void
1696 __free_args(struct spdk_fs_cb_args *args)
1697 {
1698 	struct spdk_fs_request *req;
1699 
1700 	if (!args->from_request) {
1701 		free(args);
1702 	} else {
1703 		/* Depends on args being at the start of the spdk_fs_request structure. */
1704 		req = (struct spdk_fs_request *)args;
1705 		free_fs_request(req);
1706 	}
1707 }
1708 
1709 static void
1710 __check_sync_reqs(struct spdk_file *file)
1711 {
1712 	struct spdk_fs_request *sync_req;
1713 
1714 	pthread_spin_lock(&file->lock);
1715 
1716 	TAILQ_FOREACH(sync_req, &file->sync_requests, args.op.sync.tailq) {
1717 		if (sync_req->args.op.sync.offset <= file->length_flushed) {
1718 			break;
1719 		}
1720 	}
1721 
1722 	if (sync_req != NULL && !sync_req->args.op.sync.xattr_in_progress) {
1723 		BLOBFS_TRACE(file, "set xattr length 0x%jx\n", file->length_flushed);
1724 		sync_req->args.op.sync.xattr_in_progress = true;
1725 		spdk_blob_md_set_xattr(file->blob, "length", &file->length_flushed,
1726 				       sizeof(file->length_flushed));
1727 
1728 		pthread_spin_unlock(&file->lock);
1729 		spdk_bs_md_sync_blob(file->blob, __file_cache_finish_sync_bs_cb, file);
1730 	} else {
1731 		pthread_spin_unlock(&file->lock);
1732 	}
1733 }
1734 
1735 static void
1736 __file_flush_done(void *arg, int bserrno)
1737 {
1738 	struct spdk_fs_cb_args *args = arg;
1739 	struct spdk_file *file = args->file;
1740 	struct cache_buffer *next = args->op.flush.cache_buffer;
1741 
1742 	BLOBFS_TRACE(file, "length=%jx\n", args->op.flush.length);
1743 
1744 	pthread_spin_lock(&file->lock);
1745 	next->in_progress = false;
1746 	next->bytes_flushed += args->op.flush.length;
1747 	file->length_flushed += args->op.flush.length;
1748 	if (file->length_flushed > file->length) {
1749 		file->length = file->length_flushed;
1750 	}
1751 	if (next->bytes_flushed == next->buf_size) {
1752 		BLOBFS_TRACE(file, "write buffer fully flushed 0x%jx\n", file->length_flushed);
1753 		next = spdk_tree_find_buffer(file->tree, file->length_flushed);
1754 	}
1755 
1756 	/*
1757 	 * Assert that there is no cached data that extends past the end of the underlying
1758 	 *  blob.
1759 	 */
1760 	assert(next == NULL || next->offset < __file_get_blob_size(file) ||
1761 	       next->bytes_filled == 0);
1762 
1763 	pthread_spin_unlock(&file->lock);
1764 
1765 	__check_sync_reqs(file);
1766 
1767 	__file_flush(args);
1768 }
1769 
1770 static void
1771 __file_flush(void *_args)
1772 {
1773 	struct spdk_fs_cb_args *args = _args;
1774 	struct spdk_file *file = args->file;
1775 	struct cache_buffer *next;
1776 	uint64_t offset, length, start_page, num_pages;
1777 	uint32_t page_size;
1778 
1779 	pthread_spin_lock(&file->lock);
1780 	next = spdk_tree_find_buffer(file->tree, file->length_flushed);
1781 	if (next == NULL || next->in_progress) {
1782 		/*
1783 		 * There is either no data to flush, or a flush I/O is already in
1784 		 *  progress.  So return immediately - if a flush I/O is in
1785 		 *  progress we will flush more data after that is completed.
1786 		 */
1787 		__free_args(args);
1788 		pthread_spin_unlock(&file->lock);
1789 		return;
1790 	}
1791 
1792 	offset = next->offset + next->bytes_flushed;
1793 	length = next->bytes_filled - next->bytes_flushed;
1794 	if (length == 0) {
1795 		__free_args(args);
1796 		pthread_spin_unlock(&file->lock);
1797 		return;
1798 	}
1799 	args->op.flush.length = length;
1800 	args->op.flush.cache_buffer = next;
1801 
1802 	__get_page_parameters(file, offset, length, &start_page, &page_size, &num_pages);
1803 
1804 	next->in_progress = true;
1805 	BLOBFS_TRACE(file, "offset=%jx length=%jx page start=%jx num=%jx\n",
1806 		     offset, length, start_page, num_pages);
1807 	pthread_spin_unlock(&file->lock);
1808 	spdk_bs_io_write_blob(file->blob, file->fs->sync_target.sync_fs_channel->bs_channel,
1809 			      next->buf + (start_page * page_size) - next->offset,
1810 			      start_page, num_pages,
1811 			      __file_flush_done, args);
1812 }
1813 
1814 static void
1815 __file_extend_done(void *arg, int bserrno)
1816 {
1817 	struct spdk_fs_cb_args *args = arg;
1818 
1819 	__wake_caller(args);
1820 }
1821 
1822 static void
1823 __file_extend_blob(void *_args)
1824 {
1825 	struct spdk_fs_cb_args *args = _args;
1826 	struct spdk_file *file = args->file;
1827 
1828 	spdk_bs_md_resize_blob(file->blob, args->op.resize.num_clusters);
1829 
1830 	spdk_bs_md_sync_blob(file->blob, __file_extend_done, args);
1831 }
1832 
1833 static void
1834 __rw_from_file_done(void *arg, int bserrno)
1835 {
1836 	struct spdk_fs_cb_args *args = arg;
1837 
1838 	__wake_caller(args);
1839 	__free_args(args);
1840 }
1841 
1842 static void
1843 __rw_from_file(void *_args)
1844 {
1845 	struct spdk_fs_cb_args *args = _args;
1846 	struct spdk_file *file = args->file;
1847 
1848 	if (args->op.rw.is_read) {
1849 		spdk_file_read_async(file, file->fs->sync_target.sync_io_channel, args->op.rw.user_buf,
1850 				     args->op.rw.offset, args->op.rw.length,
1851 				     __rw_from_file_done, args);
1852 	} else {
1853 		spdk_file_write_async(file, file->fs->sync_target.sync_io_channel, args->op.rw.user_buf,
1854 				      args->op.rw.offset, args->op.rw.length,
1855 				      __rw_from_file_done, args);
1856 	}
1857 }
1858 
1859 static int
1860 __send_rw_from_file(struct spdk_file *file, sem_t *sem, void *payload,
1861 		    uint64_t offset, uint64_t length, bool is_read)
1862 {
1863 	struct spdk_fs_cb_args *args;
1864 
1865 	args = calloc(1, sizeof(*args));
1866 	if (args == NULL) {
1867 		sem_post(sem);
1868 		return -ENOMEM;
1869 	}
1870 
1871 	args->file = file;
1872 	args->sem = sem;
1873 	args->op.rw.user_buf = payload;
1874 	args->op.rw.offset = offset;
1875 	args->op.rw.length = length;
1876 	args->op.rw.is_read = is_read;
1877 	file->fs->send_request(__rw_from_file, args);
1878 	return 0;
1879 }
1880 
1881 int
1882 spdk_file_write(struct spdk_file *file, struct spdk_io_channel *_channel,
1883 		void *payload, uint64_t offset, uint64_t length)
1884 {
1885 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
1886 	struct spdk_fs_cb_args *args;
1887 	uint64_t rem_length, copy, blob_size, cluster_sz;
1888 	uint32_t cache_buffers_filled = 0;
1889 	uint8_t *cur_payload;
1890 	struct cache_buffer *last;
1891 
1892 	BLOBFS_TRACE_RW(file, "offset=%jx length=%jx\n", offset, length);
1893 
1894 	if (length == 0) {
1895 		return 0;
1896 	}
1897 
1898 	if (offset != file->append_pos) {
1899 		BLOBFS_TRACE(file, " error offset=%jx append_pos=%jx\n", offset, file->append_pos);
1900 		return -EINVAL;
1901 	}
1902 
1903 	pthread_spin_lock(&file->lock);
1904 	file->open_for_writing = true;
1905 
1906 	if (file->last == NULL) {
1907 		if (file->append_pos % CACHE_BUFFER_SIZE == 0) {
1908 			cache_append_buffer(file);
1909 		} else {
1910 			int rc;
1911 
1912 			file->append_pos += length;
1913 			pthread_spin_unlock(&file->lock);
1914 			rc = __send_rw_from_file(file, &channel->sem, payload,
1915 						 offset, length, false);
1916 			sem_wait(&channel->sem);
1917 			return rc;
1918 		}
1919 	}
1920 
1921 	blob_size = __file_get_blob_size(file);
1922 
1923 	if ((offset + length) > blob_size) {
1924 		struct spdk_fs_cb_args extend_args = {};
1925 
1926 		cluster_sz = file->fs->bs_opts.cluster_sz;
1927 		extend_args.sem = &channel->sem;
1928 		extend_args.op.resize.num_clusters = __bytes_to_clusters((offset + length), cluster_sz);
1929 		extend_args.file = file;
1930 		BLOBFS_TRACE(file, "start resize to %u clusters\n", extend_args.op.resize.num_clusters);
1931 		pthread_spin_unlock(&file->lock);
1932 		file->fs->send_request(__file_extend_blob, &extend_args);
1933 		sem_wait(&channel->sem);
1934 	}
1935 
1936 	last = file->last;
1937 	rem_length = length;
1938 	cur_payload = payload;
1939 	while (rem_length > 0) {
1940 		copy = last->buf_size - last->bytes_filled;
1941 		if (copy > rem_length) {
1942 			copy = rem_length;
1943 		}
1944 		BLOBFS_TRACE_RW(file, "  fill offset=%jx length=%jx\n", file->append_pos, copy);
1945 		memcpy(&last->buf[last->bytes_filled], cur_payload, copy);
1946 		file->append_pos += copy;
1947 		if (file->length < file->append_pos) {
1948 			file->length = file->append_pos;
1949 		}
1950 		cur_payload += copy;
1951 		last->bytes_filled += copy;
1952 		rem_length -= copy;
1953 		if (last->bytes_filled == last->buf_size) {
1954 			cache_buffers_filled++;
1955 			last = cache_append_buffer(file);
1956 			if (last == NULL) {
1957 				BLOBFS_TRACE(file, "nomem\n");
1958 				pthread_spin_unlock(&file->lock);
1959 				return -ENOMEM;
1960 			}
1961 		}
1962 	}
1963 
1964 	if (cache_buffers_filled == 0) {
1965 		pthread_spin_unlock(&file->lock);
1966 		return 0;
1967 	}
1968 
1969 	args = calloc(1, sizeof(*args));
1970 	if (args == NULL) {
1971 		pthread_spin_unlock(&file->lock);
1972 		return -ENOMEM;
1973 	}
1974 
1975 	args->file = file;
1976 	file->fs->send_request(__file_flush, args);
1977 	pthread_spin_unlock(&file->lock);
1978 	return 0;
1979 }
1980 
1981 static void
1982 __readahead_done(void *arg, int bserrno)
1983 {
1984 	struct spdk_fs_cb_args *args = arg;
1985 	struct cache_buffer *cache_buffer = args->op.readahead.cache_buffer;
1986 	struct spdk_file *file = args->file;
1987 
1988 	BLOBFS_TRACE(file, "offset=%jx\n", cache_buffer->offset);
1989 
1990 	pthread_spin_lock(&file->lock);
1991 	cache_buffer->bytes_filled = args->op.readahead.length;
1992 	cache_buffer->bytes_flushed = args->op.readahead.length;
1993 	cache_buffer->in_progress = false;
1994 	pthread_spin_unlock(&file->lock);
1995 
1996 	__free_args(args);
1997 }
1998 
1999 static void
2000 __readahead(void *_args)
2001 {
2002 	struct spdk_fs_cb_args *args = _args;
2003 	struct spdk_file *file = args->file;
2004 	uint64_t offset, length, start_page, num_pages;
2005 	uint32_t page_size;
2006 
2007 	offset = args->op.readahead.offset;
2008 	length = args->op.readahead.length;
2009 	assert(length > 0);
2010 
2011 	__get_page_parameters(file, offset, length, &start_page, &page_size, &num_pages);
2012 
2013 	BLOBFS_TRACE(file, "offset=%jx length=%jx page start=%jx num=%jx\n",
2014 		     offset, length, start_page, num_pages);
2015 	spdk_bs_io_read_blob(file->blob, file->fs->sync_target.sync_fs_channel->bs_channel,
2016 			     args->op.readahead.cache_buffer->buf,
2017 			     start_page, num_pages,
2018 			     __readahead_done, args);
2019 }
2020 
2021 static uint64_t
2022 __next_cache_buffer_offset(uint64_t offset)
2023 {
2024 	return (offset + CACHE_BUFFER_SIZE) & ~(CACHE_TREE_LEVEL_MASK(0));
2025 }
2026 
2027 static void
2028 check_readahead(struct spdk_file *file, uint64_t offset)
2029 {
2030 	struct spdk_fs_cb_args *args;
2031 
2032 	offset = __next_cache_buffer_offset(offset);
2033 	if (spdk_tree_find_buffer(file->tree, offset) != NULL || file->length <= offset) {
2034 		return;
2035 	}
2036 
2037 	args = calloc(1, sizeof(*args));
2038 	if (args == NULL) {
2039 		return;
2040 	}
2041 
2042 	BLOBFS_TRACE(file, "offset=%jx\n", offset);
2043 
2044 	args->file = file;
2045 	args->op.readahead.offset = offset;
2046 	args->op.readahead.cache_buffer = cache_insert_buffer(file, offset);
2047 	args->op.readahead.cache_buffer->in_progress = true;
2048 	if (file->length < (offset + CACHE_BUFFER_SIZE)) {
2049 		args->op.readahead.length = file->length & (CACHE_BUFFER_SIZE - 1);
2050 	} else {
2051 		args->op.readahead.length = CACHE_BUFFER_SIZE;
2052 	}
2053 	file->fs->send_request(__readahead, args);
2054 }
2055 
2056 static int
2057 __file_read(struct spdk_file *file, void *payload, uint64_t offset, uint64_t length, sem_t *sem)
2058 {
2059 	struct cache_buffer *buf;
2060 	int rc;
2061 
2062 	buf = spdk_tree_find_filled_buffer(file->tree, offset);
2063 	if (buf == NULL) {
2064 		pthread_spin_unlock(&file->lock);
2065 		rc = __send_rw_from_file(file, sem, payload, offset, length, true);
2066 		pthread_spin_lock(&file->lock);
2067 		return rc;
2068 	}
2069 
2070 	if ((offset + length) > (buf->offset + buf->bytes_filled)) {
2071 		length = buf->offset + buf->bytes_filled - offset;
2072 	}
2073 	BLOBFS_TRACE(file, "read %p offset=%ju length=%ju\n", payload, offset, length);
2074 	memcpy(payload, &buf->buf[offset - buf->offset], length);
2075 	if ((offset + length) % CACHE_BUFFER_SIZE == 0) {
2076 		pthread_spin_lock(&g_caches_lock);
2077 		spdk_tree_remove_buffer(file->tree, buf);
2078 		if (file->tree->present_mask == 0) {
2079 			TAILQ_REMOVE(&g_caches, file, cache_tailq);
2080 		}
2081 		pthread_spin_unlock(&g_caches_lock);
2082 	}
2083 
2084 	sem_post(sem);
2085 	return 0;
2086 }
2087 
2088 int64_t
2089 spdk_file_read(struct spdk_file *file, struct spdk_io_channel *_channel,
2090 	       void *payload, uint64_t offset, uint64_t length)
2091 {
2092 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
2093 	uint64_t final_offset, final_length;
2094 	uint32_t sub_reads = 0;
2095 	int rc = 0;
2096 
2097 	pthread_spin_lock(&file->lock);
2098 
2099 	BLOBFS_TRACE_RW(file, "offset=%ju length=%ju\n", offset, length);
2100 
2101 	file->open_for_writing = false;
2102 
2103 	if (length == 0 || offset >= file->append_pos) {
2104 		pthread_spin_unlock(&file->lock);
2105 		return 0;
2106 	}
2107 
2108 	if (offset + length > file->append_pos) {
2109 		length = file->append_pos - offset;
2110 	}
2111 
2112 	if (offset != file->next_seq_offset) {
2113 		file->seq_byte_count = 0;
2114 	}
2115 	file->seq_byte_count += length;
2116 	file->next_seq_offset = offset + length;
2117 	if (file->seq_byte_count >= CACHE_READAHEAD_THRESHOLD) {
2118 		check_readahead(file, offset);
2119 		check_readahead(file, offset + CACHE_BUFFER_SIZE);
2120 	}
2121 
2122 	final_length = 0;
2123 	final_offset = offset + length;
2124 	while (offset < final_offset) {
2125 		length = NEXT_CACHE_BUFFER_OFFSET(offset) - offset;
2126 		if (length > (final_offset - offset)) {
2127 			length = final_offset - offset;
2128 		}
2129 		rc = __file_read(file, payload, offset, length, &channel->sem);
2130 		if (rc == 0) {
2131 			final_length += length;
2132 		} else {
2133 			break;
2134 		}
2135 		payload += length;
2136 		offset += length;
2137 		sub_reads++;
2138 	}
2139 	pthread_spin_unlock(&file->lock);
2140 	while (sub_reads-- > 0) {
2141 		sem_wait(&channel->sem);
2142 	}
2143 	if (rc == 0) {
2144 		return final_length;
2145 	} else {
2146 		return rc;
2147 	}
2148 }
2149 
2150 static void
2151 _file_sync(struct spdk_file *file, struct spdk_fs_channel *channel,
2152 	   spdk_file_op_complete cb_fn, void *cb_arg)
2153 {
2154 	struct spdk_fs_request *sync_req;
2155 	struct spdk_fs_request *flush_req;
2156 	struct spdk_fs_cb_args *sync_args;
2157 	struct spdk_fs_cb_args *flush_args;
2158 
2159 	BLOBFS_TRACE(file, "offset=%jx\n", file->append_pos);
2160 
2161 	pthread_spin_lock(&file->lock);
2162 	if (file->append_pos <= file->length_flushed || file->last == NULL) {
2163 		BLOBFS_TRACE(file, "done - no data to flush\n");
2164 		pthread_spin_unlock(&file->lock);
2165 		cb_fn(cb_arg, 0);
2166 		return;
2167 	}
2168 
2169 	sync_req = alloc_fs_request(channel);
2170 	assert(sync_req != NULL);
2171 	sync_args = &sync_req->args;
2172 
2173 	flush_req = alloc_fs_request(channel);
2174 	assert(flush_req != NULL);
2175 	flush_args = &flush_req->args;
2176 
2177 	sync_args->file = file;
2178 	sync_args->fn.file_op = cb_fn;
2179 	sync_args->arg = cb_arg;
2180 	sync_args->op.sync.offset = file->append_pos;
2181 	sync_args->op.sync.xattr_in_progress = false;
2182 	TAILQ_INSERT_TAIL(&file->sync_requests, sync_req, args.op.sync.tailq);
2183 	pthread_spin_unlock(&file->lock);
2184 
2185 	flush_args->file = file;
2186 	channel->send_request(__file_flush, flush_args);
2187 }
2188 
2189 int
2190 spdk_file_sync(struct spdk_file *file, struct spdk_io_channel *_channel)
2191 {
2192 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
2193 
2194 	_file_sync(file, channel, __sem_post, &channel->sem);
2195 	sem_wait(&channel->sem);
2196 
2197 	return 0;
2198 }
2199 
2200 void
2201 spdk_file_sync_async(struct spdk_file *file, struct spdk_io_channel *_channel,
2202 		     spdk_file_op_complete cb_fn, void *cb_arg)
2203 {
2204 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
2205 
2206 	_file_sync(file, channel, cb_fn, cb_arg);
2207 }
2208 
2209 void
2210 spdk_file_set_priority(struct spdk_file *file, uint32_t priority)
2211 {
2212 	BLOBFS_TRACE(file, "priority=%u\n", priority);
2213 	file->priority = priority;
2214 
2215 }
2216 
2217 /*
2218  * Close routines
2219  */
2220 
2221 static void
2222 __file_close_async_done(void *ctx, int bserrno)
2223 {
2224 	struct spdk_fs_request *req = ctx;
2225 	struct spdk_fs_cb_args *args = &req->args;
2226 
2227 	args->fn.file_op(args->arg, bserrno);
2228 	free_fs_request(req);
2229 }
2230 
2231 static void
2232 __file_close_async(struct spdk_file *file, struct spdk_fs_request *req)
2233 {
2234 	pthread_spin_lock(&file->lock);
2235 	if (file->ref_count == 0) {
2236 		pthread_spin_unlock(&file->lock);
2237 		__file_close_async_done(req, -EBADF);
2238 		return;
2239 	}
2240 
2241 	file->ref_count--;
2242 	if (file->ref_count > 0) {
2243 		pthread_spin_unlock(&file->lock);
2244 		__file_close_async_done(req, 0);
2245 		return;
2246 	}
2247 
2248 	pthread_spin_unlock(&file->lock);
2249 
2250 	spdk_bs_md_close_blob(&file->blob, __file_close_async_done, req);
2251 }
2252 
2253 static void
2254 __file_close_async__sync_done(void *arg, int fserrno)
2255 {
2256 	struct spdk_fs_request *req = arg;
2257 	struct spdk_fs_cb_args *args = &req->args;
2258 
2259 	__file_close_async(args->file, req);
2260 }
2261 
2262 void
2263 spdk_file_close_async(struct spdk_file *file, spdk_file_op_complete cb_fn, void *cb_arg)
2264 {
2265 	struct spdk_fs_request *req;
2266 	struct spdk_fs_cb_args *args;
2267 
2268 	req = alloc_fs_request(file->fs->md_target.md_fs_channel);
2269 	if (req == NULL) {
2270 		cb_fn(cb_arg, -ENOMEM);
2271 		return;
2272 	}
2273 
2274 	args = &req->args;
2275 	args->file = file;
2276 	args->fn.file_op = cb_fn;
2277 	args->arg = cb_arg;
2278 
2279 	spdk_file_sync_async(file, file->fs->md_target.md_io_channel, __file_close_async__sync_done, req);
2280 }
2281 
2282 static void
2283 __file_close_done(void *arg, int fserrno)
2284 {
2285 	struct spdk_fs_cb_args *args = arg;
2286 
2287 	args->rc = fserrno;
2288 	sem_post(args->sem);
2289 }
2290 
2291 static void
2292 __file_close(void *arg)
2293 {
2294 	struct spdk_fs_request *req = arg;
2295 	struct spdk_fs_cb_args *args = &req->args;
2296 	struct spdk_file *file = args->file;
2297 
2298 	__file_close_async(file, req);
2299 }
2300 
2301 int
2302 spdk_file_close(struct spdk_file *file, struct spdk_io_channel *_channel)
2303 {
2304 	struct spdk_fs_channel *channel = spdk_io_channel_get_ctx(_channel);
2305 	struct spdk_fs_request *req;
2306 	struct spdk_fs_cb_args *args;
2307 
2308 	req = alloc_fs_request(channel);
2309 	assert(req != NULL);
2310 
2311 	args = &req->args;
2312 
2313 	spdk_file_sync(file, _channel);
2314 	BLOBFS_TRACE(file, "name=%s\n", file->name);
2315 	args->file = file;
2316 	args->sem = &channel->sem;
2317 	args->fn.file_op = __file_close_done;
2318 	args->arg = req;
2319 	channel->send_request(__file_close, req);
2320 	sem_wait(&channel->sem);
2321 
2322 	return args->rc;
2323 }
2324 
2325 static void
2326 cache_free_buffers(struct spdk_file *file)
2327 {
2328 	BLOBFS_TRACE(file, "free=%s\n", file->name);
2329 	pthread_spin_lock(&file->lock);
2330 	pthread_spin_lock(&g_caches_lock);
2331 	if (file->tree->present_mask == 0) {
2332 		pthread_spin_unlock(&g_caches_lock);
2333 		pthread_spin_unlock(&file->lock);
2334 		return;
2335 	}
2336 	spdk_tree_free_buffers(file->tree);
2337 	if (file->tree->present_mask == 0) {
2338 		TAILQ_REMOVE(&g_caches, file, cache_tailq);
2339 	}
2340 	file->last = NULL;
2341 	pthread_spin_unlock(&g_caches_lock);
2342 	pthread_spin_unlock(&file->lock);
2343 }
2344 
2345 SPDK_LOG_REGISTER_TRACE_FLAG("blobfs", SPDK_TRACE_BLOBFS);
2346 SPDK_LOG_REGISTER_TRACE_FLAG("blobfs_rw", SPDK_TRACE_BLOBFS_RW);
2347