xref: /spdk/module/bdev/rbd/bdev_rbd.c (revision de21d8f4e45b732c13ce5c7aa1872f73bffd38aa)
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 "bdev_rbd.h"
37 
38 #include <rbd/librbd.h>
39 #include <rados/librados.h>
40 #include <sys/eventfd.h>
41 #include <sys/epoll.h>
42 
43 #include "spdk/env.h"
44 #include "spdk/bdev.h"
45 #include "spdk/thread.h"
46 #include "spdk/json.h"
47 #include "spdk/string.h"
48 #include "spdk/util.h"
49 #include "spdk/likely.h"
50 
51 #include "spdk/bdev_module.h"
52 #include "spdk/log.h"
53 
54 #define SPDK_RBD_QUEUE_DEPTH 128
55 #define MAX_EVENTS_PER_POLL 128
56 
57 static int bdev_rbd_count = 0;
58 
59 struct bdev_rbd {
60 	struct spdk_bdev disk;
61 	char *rbd_name;
62 	char *user_id;
63 	char *pool_name;
64 	char **config;
65 	rbd_image_info_t info;
66 	TAILQ_ENTRY(bdev_rbd) tailq;
67 	struct spdk_poller *reset_timer;
68 	struct spdk_bdev_io *reset_bdev_io;
69 };
70 
71 struct bdev_rbd_group_channel {
72 	struct spdk_poller *poller;
73 	int epoll_fd;
74 };
75 
76 struct bdev_rbd_io_channel {
77 	rados_ioctx_t io_ctx;
78 	rados_t cluster;
79 	int pfd;
80 	rbd_image_t image;
81 	struct bdev_rbd *disk;
82 	struct bdev_rbd_group_channel *group_ch;
83 };
84 
85 struct bdev_rbd_io {
86 	size_t	total_len;
87 };
88 
89 static void
90 bdev_rbd_free(struct bdev_rbd *rbd)
91 {
92 	if (!rbd) {
93 		return;
94 	}
95 
96 	free(rbd->disk.name);
97 	free(rbd->rbd_name);
98 	free(rbd->user_id);
99 	free(rbd->pool_name);
100 	bdev_rbd_free_config(rbd->config);
101 	free(rbd);
102 }
103 
104 void
105 bdev_rbd_free_config(char **config)
106 {
107 	char **entry;
108 
109 	if (config) {
110 		for (entry = config; *entry; entry++) {
111 			free(*entry);
112 		}
113 		free(config);
114 	}
115 }
116 
117 char **
118 bdev_rbd_dup_config(const char *const *config)
119 {
120 	size_t count;
121 	char **copy;
122 
123 	if (!config) {
124 		return NULL;
125 	}
126 	for (count = 0; config[count]; count++) {}
127 	copy = calloc(count + 1, sizeof(*copy));
128 	if (!copy) {
129 		return NULL;
130 	}
131 	for (count = 0; config[count]; count++) {
132 		if (!(copy[count] = strdup(config[count]))) {
133 			bdev_rbd_free_config(copy);
134 			return NULL;
135 		}
136 	}
137 	return copy;
138 }
139 
140 static int
141 bdev_rados_context_init(const char *user_id, const char *rbd_pool_name, const char *const *config,
142 			rados_t *cluster, rados_ioctx_t *io_ctx)
143 {
144 	int ret;
145 
146 	ret = rados_create(cluster, user_id);
147 	if (ret < 0) {
148 		SPDK_ERRLOG("Failed to create rados_t struct\n");
149 		return -1;
150 	}
151 
152 	if (config) {
153 		const char *const *entry = config;
154 		while (*entry) {
155 			ret = rados_conf_set(*cluster, entry[0], entry[1]);
156 			if (ret < 0) {
157 				SPDK_ERRLOG("Failed to set %s = %s\n", entry[0], entry[1]);
158 				rados_shutdown(*cluster);
159 				return -1;
160 			}
161 			entry += 2;
162 		}
163 	} else {
164 		ret = rados_conf_read_file(*cluster, NULL);
165 		if (ret < 0) {
166 			SPDK_ERRLOG("Failed to read conf file\n");
167 			rados_shutdown(*cluster);
168 			return -1;
169 		}
170 	}
171 
172 	ret = rados_connect(*cluster);
173 	if (ret < 0) {
174 		SPDK_ERRLOG("Failed to connect to rbd_pool\n");
175 		rados_shutdown(*cluster);
176 		return -1;
177 	}
178 
179 	ret = rados_ioctx_create(*cluster, rbd_pool_name, io_ctx);
180 
181 	if (ret < 0) {
182 		SPDK_ERRLOG("Failed to create ioctx\n");
183 		rados_shutdown(*cluster);
184 		return -1;
185 	}
186 
187 	return 0;
188 }
189 
190 static int
191 bdev_rbd_init(const char *user_id, const char *rbd_pool_name, const char *const *config,
192 	      const char *rbd_name, rbd_image_info_t *info)
193 {
194 	int ret;
195 	rados_t cluster = NULL;
196 	rados_ioctx_t io_ctx = NULL;
197 	rbd_image_t image = NULL;
198 
199 	ret = bdev_rados_context_init(user_id, rbd_pool_name, config, &cluster, &io_ctx);
200 	if (ret < 0) {
201 		SPDK_ERRLOG("Failed to create rados context for user_id=%s and rbd_pool=%s\n",
202 			    user_id ? user_id : "admin (the default)", rbd_pool_name);
203 		return -1;
204 	}
205 
206 	ret = rbd_open(io_ctx, rbd_name, &image, NULL);
207 	if (ret < 0) {
208 		SPDK_ERRLOG("Failed to open specified rbd device\n");
209 		goto err;
210 	}
211 	ret = rbd_stat(image, info, sizeof(*info));
212 	rbd_close(image);
213 	if (ret < 0) {
214 		SPDK_ERRLOG("Failed to stat specified rbd device\n");
215 		goto err;
216 	}
217 
218 	rados_ioctx_destroy(io_ctx);
219 	return 0;
220 err:
221 	rados_ioctx_destroy(io_ctx);
222 	rados_shutdown(cluster);
223 	return -1;
224 }
225 
226 static void
227 bdev_rbd_exit(rbd_image_t image)
228 {
229 	rbd_flush(image);
230 	rbd_close(image);
231 }
232 
233 static void
234 bdev_rbd_finish_aiocb(rbd_completion_t cb, void *arg)
235 {
236 	/* Doing nothing here */
237 }
238 
239 static void
240 bdev_rbd_start_aio(struct spdk_io_channel *ch, struct spdk_bdev_io *bdev_io,
241 		   struct iovec *iov, int iovcnt, uint64_t offset, size_t len)
242 {
243 	struct bdev_rbd_io_channel *rbdio_ch = spdk_io_channel_get_ctx(ch);
244 	int ret;
245 	rbd_completion_t comp;
246 	struct bdev_rbd_io *rbd_io;
247 	rbd_image_t image = rbdio_ch->image;
248 
249 	ret = rbd_aio_create_completion(bdev_io, bdev_rbd_finish_aiocb,
250 					&comp);
251 	if (ret < 0) {
252 		goto err;
253 	}
254 
255 	if (bdev_io->type == SPDK_BDEV_IO_TYPE_READ) {
256 		rbd_io = (struct bdev_rbd_io *)bdev_io->driver_ctx;
257 		rbd_io->total_len = len;
258 		if (spdk_likely(iovcnt == 1)) {
259 			ret = rbd_aio_read(image, offset, iov[0].iov_len, iov[0].iov_base, comp);
260 		} else {
261 			ret = rbd_aio_readv(image, iov, iovcnt, offset, comp);
262 		}
263 	} else if (bdev_io->type == SPDK_BDEV_IO_TYPE_WRITE) {
264 		if (spdk_likely(iovcnt == 1)) {
265 			ret = rbd_aio_write(image, offset, iov[0].iov_len, iov[0].iov_base, comp);
266 		} else {
267 			ret = rbd_aio_writev(image, iov, iovcnt, offset, comp);
268 		}
269 	} else if (bdev_io->type == SPDK_BDEV_IO_TYPE_FLUSH) {
270 		ret = rbd_aio_flush(image, comp);
271 	}
272 
273 	if (ret < 0) {
274 		rbd_aio_release(comp);
275 		goto err;
276 	}
277 
278 	return;
279 
280 err:
281 	spdk_bdev_io_complete(bdev_io, SPDK_BDEV_IO_STATUS_FAILED);
282 }
283 
284 static int bdev_rbd_library_init(void);
285 
286 static void bdev_rbd_library_fini(void);
287 
288 static int
289 bdev_rbd_get_ctx_size(void)
290 {
291 	return sizeof(struct bdev_rbd_io);
292 }
293 
294 static struct spdk_bdev_module rbd_if = {
295 	.name = "rbd",
296 	.module_init = bdev_rbd_library_init,
297 	.module_fini = bdev_rbd_library_fini,
298 	.get_ctx_size = bdev_rbd_get_ctx_size,
299 
300 };
301 SPDK_BDEV_MODULE_REGISTER(rbd, &rbd_if)
302 
303 static int
304 bdev_rbd_reset_timer(void *arg)
305 {
306 	struct bdev_rbd *disk = arg;
307 
308 	/*
309 	 * TODO: This should check if any I/O is still in flight before completing the reset.
310 	 * For now, just complete after the timer expires.
311 	 */
312 	spdk_bdev_io_complete(disk->reset_bdev_io, SPDK_BDEV_IO_STATUS_SUCCESS);
313 	spdk_poller_unregister(&disk->reset_timer);
314 	disk->reset_bdev_io = NULL;
315 
316 	return SPDK_POLLER_BUSY;
317 }
318 
319 static void
320 bdev_rbd_reset(struct bdev_rbd *disk, struct spdk_bdev_io *bdev_io)
321 {
322 	/*
323 	 * HACK: Since librbd doesn't provide any way to cancel outstanding aio, just kick off a
324 	 * timer to wait for in-flight I/O to complete.
325 	 */
326 	assert(disk->reset_bdev_io == NULL);
327 	disk->reset_bdev_io = bdev_io;
328 	disk->reset_timer = SPDK_POLLER_REGISTER(bdev_rbd_reset_timer, disk, 1 * 1000 * 1000);
329 }
330 
331 static int
332 bdev_rbd_destruct(void *ctx)
333 {
334 	struct bdev_rbd *rbd = ctx;
335 
336 	spdk_io_device_unregister(rbd, NULL);
337 
338 	bdev_rbd_free(rbd);
339 	return 0;
340 }
341 
342 static void
343 bdev_rbd_get_buf_cb(struct spdk_io_channel *ch, struct spdk_bdev_io *bdev_io,
344 		    bool success)
345 {
346 	if (!success) {
347 		spdk_bdev_io_complete(bdev_io, SPDK_BDEV_IO_STATUS_FAILED);
348 		return;
349 	}
350 
351 	bdev_rbd_start_aio(ch,
352 			   bdev_io,
353 			   bdev_io->u.bdev.iovs,
354 			   bdev_io->u.bdev.iovcnt,
355 			   bdev_io->u.bdev.offset_blocks * bdev_io->bdev->blocklen,
356 			   bdev_io->u.bdev.num_blocks * bdev_io->bdev->blocklen);
357 }
358 
359 static void
360 bdev_rbd_submit_request(struct spdk_io_channel *ch, struct spdk_bdev_io *bdev_io)
361 {
362 	switch (bdev_io->type) {
363 	case SPDK_BDEV_IO_TYPE_READ:
364 		spdk_bdev_io_get_buf(bdev_io, bdev_rbd_get_buf_cb,
365 				     bdev_io->u.bdev.num_blocks * bdev_io->bdev->blocklen);
366 		break;
367 
368 	case SPDK_BDEV_IO_TYPE_WRITE:
369 	case SPDK_BDEV_IO_TYPE_FLUSH:
370 		bdev_rbd_start_aio(ch,
371 				   bdev_io,
372 				   bdev_io->u.bdev.iovs,
373 				   bdev_io->u.bdev.iovcnt,
374 				   bdev_io->u.bdev.offset_blocks * bdev_io->bdev->blocklen,
375 				   bdev_io->u.bdev.num_blocks * bdev_io->bdev->blocklen);
376 		break;
377 
378 	case SPDK_BDEV_IO_TYPE_RESET:
379 		bdev_rbd_reset((struct bdev_rbd *)bdev_io->bdev->ctxt,
380 			       bdev_io);
381 		break;
382 
383 	default:
384 		SPDK_ERRLOG("Unsupported IO type =%d\n", bdev_io->type);
385 		spdk_bdev_io_complete(bdev_io, SPDK_BDEV_IO_STATUS_FAILED);
386 		break;
387 	}
388 }
389 
390 static bool
391 bdev_rbd_io_type_supported(void *ctx, enum spdk_bdev_io_type io_type)
392 {
393 	switch (io_type) {
394 	case SPDK_BDEV_IO_TYPE_READ:
395 	case SPDK_BDEV_IO_TYPE_WRITE:
396 	case SPDK_BDEV_IO_TYPE_FLUSH:
397 	case SPDK_BDEV_IO_TYPE_RESET:
398 		return true;
399 
400 	default:
401 		return false;
402 	}
403 }
404 
405 static void
406 bdev_rbd_io_poll(struct bdev_rbd_io_channel *ch)
407 {
408 	int i, io_status, rc;
409 	rbd_completion_t comps[SPDK_RBD_QUEUE_DEPTH];
410 	struct spdk_bdev_io *bdev_io;
411 	struct bdev_rbd_io *rbd_io;
412 	enum spdk_bdev_io_status bio_status;
413 
414 	rc = rbd_poll_io_events(ch->image, comps, SPDK_RBD_QUEUE_DEPTH);
415 	for (i = 0; i < rc; i++) {
416 		bdev_io = rbd_aio_get_arg(comps[i]);
417 		rbd_io = (struct bdev_rbd_io *)bdev_io->driver_ctx;
418 		io_status = rbd_aio_get_return_value(comps[i]);
419 		bio_status = SPDK_BDEV_IO_STATUS_SUCCESS;
420 
421 		if (bdev_io->type == SPDK_BDEV_IO_TYPE_READ) {
422 			if ((int)rbd_io->total_len != io_status) {
423 				bio_status = SPDK_BDEV_IO_STATUS_FAILED;
424 			}
425 		} else {
426 			/* For others, 0 means success */
427 			if (io_status != 0) {
428 				bio_status = SPDK_BDEV_IO_STATUS_FAILED;
429 			}
430 		}
431 
432 		rbd_aio_release(comps[i]);
433 
434 		spdk_bdev_io_complete(bdev_io, bio_status);
435 	}
436 }
437 
438 static void
439 bdev_rbd_free_channel(struct bdev_rbd_io_channel *ch)
440 {
441 	if (!ch) {
442 		return;
443 	}
444 
445 	if (ch->image) {
446 		bdev_rbd_exit(ch->image);
447 	}
448 
449 	if (ch->io_ctx) {
450 		rados_ioctx_destroy(ch->io_ctx);
451 	}
452 
453 	if (ch->cluster) {
454 		rados_shutdown(ch->cluster);
455 	}
456 
457 	if (ch->pfd >= 0) {
458 		close(ch->pfd);
459 	}
460 
461 	if (ch->group_ch) {
462 		spdk_put_io_channel(spdk_io_channel_from_ctx(ch->group_ch));
463 	}
464 }
465 
466 static void *
467 bdev_rbd_handle(void *arg)
468 {
469 	struct bdev_rbd_io_channel *ch = arg;
470 	void *ret = arg;
471 	int rc;
472 
473 	rc = bdev_rados_context_init(ch->disk->user_id, ch->disk->pool_name,
474 				     (const char *const *)ch->disk->config,
475 				     &ch->cluster, &ch->io_ctx);
476 	if (rc < 0) {
477 		SPDK_ERRLOG("Failed to create rados context for user_id %s and rbd_pool=%s\n",
478 			    ch->disk->user_id ? ch->disk->user_id : "admin (the default)", ch->disk->pool_name);
479 		ret = NULL;
480 		goto end;
481 	}
482 
483 	if (rbd_open(ch->io_ctx, ch->disk->rbd_name, &ch->image, NULL) < 0) {
484 		SPDK_ERRLOG("Failed to open specified rbd device\n");
485 		ret = NULL;
486 	}
487 
488 end:
489 	return ret;
490 }
491 
492 static int
493 bdev_rbd_create_cb(void *io_device, void *ctx_buf)
494 {
495 	struct bdev_rbd_io_channel *ch = ctx_buf;
496 	int ret;
497 	struct epoll_event event;
498 
499 	ch->disk = io_device;
500 	ch->image = NULL;
501 	ch->io_ctx = NULL;
502 	ch->pfd = -1;
503 
504 	if (spdk_call_unaffinitized(bdev_rbd_handle, ch) == NULL) {
505 		goto err;
506 	}
507 
508 	ch->pfd = eventfd(0, EFD_NONBLOCK);
509 	if (ch->pfd < 0) {
510 		SPDK_ERRLOG("Failed to get eventfd\n");
511 		goto err;
512 	}
513 
514 	ret = rbd_set_image_notification(ch->image, ch->pfd, EVENT_TYPE_EVENTFD);
515 	if (ret < 0) {
516 		SPDK_ERRLOG("Failed to set rbd image notification\n");
517 		goto err;
518 	}
519 
520 	ch->group_ch = spdk_io_channel_get_ctx(spdk_get_io_channel(&rbd_if));
521 	assert(ch->group_ch != NULL);
522 	memset(&event, 0, sizeof(event));
523 	event.events = EPOLLIN;
524 	event.data.ptr = ch;
525 
526 	ret = epoll_ctl(ch->group_ch->epoll_fd, EPOLL_CTL_ADD, ch->pfd, &event);
527 	if (ret < 0) {
528 		SPDK_ERRLOG("Failed to add the fd of ch(%p) to the epoll group from group_ch=%p\n", ch,
529 			    ch->group_ch);
530 		goto err;
531 	}
532 
533 	return 0;
534 
535 err:
536 	bdev_rbd_free_channel(ch);
537 	return -1;
538 }
539 
540 static void
541 bdev_rbd_destroy_cb(void *io_device, void *ctx_buf)
542 {
543 	struct bdev_rbd_io_channel *io_channel = ctx_buf;
544 	int rc;
545 
546 	rc = epoll_ctl(io_channel->group_ch->epoll_fd, EPOLL_CTL_DEL,
547 		       io_channel->pfd, NULL);
548 	if (rc < 0) {
549 		SPDK_ERRLOG("Failed to remove fd on io_channel=%p from the polling group=%p\n",
550 			    io_channel, io_channel->group_ch);
551 	}
552 
553 	bdev_rbd_free_channel(io_channel);
554 }
555 
556 static struct spdk_io_channel *
557 bdev_rbd_get_io_channel(void *ctx)
558 {
559 	struct bdev_rbd *rbd_bdev = ctx;
560 
561 	return spdk_get_io_channel(rbd_bdev);
562 }
563 
564 static int
565 bdev_rbd_dump_info_json(void *ctx, struct spdk_json_write_ctx *w)
566 {
567 	struct bdev_rbd *rbd_bdev = ctx;
568 
569 	spdk_json_write_named_object_begin(w, "rbd");
570 
571 	spdk_json_write_named_string(w, "pool_name", rbd_bdev->pool_name);
572 
573 	spdk_json_write_named_string(w, "rbd_name", rbd_bdev->rbd_name);
574 
575 	if (rbd_bdev->user_id) {
576 		spdk_json_write_named_string(w, "user_id", rbd_bdev->user_id);
577 	}
578 
579 	if (rbd_bdev->config) {
580 		char **entry = rbd_bdev->config;
581 
582 		spdk_json_write_named_object_begin(w, "config");
583 		while (*entry) {
584 			spdk_json_write_named_string(w, entry[0], entry[1]);
585 			entry += 2;
586 		}
587 		spdk_json_write_object_end(w);
588 	}
589 
590 	spdk_json_write_object_end(w);
591 
592 	return 0;
593 }
594 
595 static void
596 bdev_rbd_write_config_json(struct spdk_bdev *bdev, struct spdk_json_write_ctx *w)
597 {
598 	struct bdev_rbd *rbd = bdev->ctxt;
599 
600 	spdk_json_write_object_begin(w);
601 
602 	spdk_json_write_named_string(w, "method", "bdev_rbd_create");
603 
604 	spdk_json_write_named_object_begin(w, "params");
605 	spdk_json_write_named_string(w, "name", bdev->name);
606 	spdk_json_write_named_string(w, "pool_name", rbd->pool_name);
607 	spdk_json_write_named_string(w, "rbd_name", rbd->rbd_name);
608 	spdk_json_write_named_uint32(w, "block_size", bdev->blocklen);
609 	if (rbd->user_id) {
610 		spdk_json_write_named_string(w, "user_id", rbd->user_id);
611 	}
612 
613 	if (rbd->config) {
614 		char **entry = rbd->config;
615 
616 		spdk_json_write_named_object_begin(w, "config");
617 		while (*entry) {
618 			spdk_json_write_named_string(w, entry[0], entry[1]);
619 			entry += 2;
620 		}
621 		spdk_json_write_object_end(w);
622 	}
623 
624 	spdk_json_write_object_end(w);
625 
626 	spdk_json_write_object_end(w);
627 }
628 
629 static const struct spdk_bdev_fn_table rbd_fn_table = {
630 	.destruct		= bdev_rbd_destruct,
631 	.submit_request		= bdev_rbd_submit_request,
632 	.io_type_supported	= bdev_rbd_io_type_supported,
633 	.get_io_channel		= bdev_rbd_get_io_channel,
634 	.dump_info_json		= bdev_rbd_dump_info_json,
635 	.write_config_json	= bdev_rbd_write_config_json,
636 };
637 
638 int
639 bdev_rbd_create(struct spdk_bdev **bdev, const char *name, const char *user_id,
640 		const char *pool_name,
641 		const char *const *config,
642 		const char *rbd_name,
643 		uint32_t block_size)
644 {
645 	struct bdev_rbd *rbd;
646 	int ret;
647 
648 	if ((pool_name == NULL) || (rbd_name == NULL)) {
649 		return -EINVAL;
650 	}
651 
652 	rbd = calloc(1, sizeof(struct bdev_rbd));
653 	if (rbd == NULL) {
654 		SPDK_ERRLOG("Failed to allocate bdev_rbd struct\n");
655 		return -ENOMEM;
656 	}
657 
658 	rbd->rbd_name = strdup(rbd_name);
659 	if (!rbd->rbd_name) {
660 		bdev_rbd_free(rbd);
661 		return -ENOMEM;
662 	}
663 
664 	if (user_id) {
665 		rbd->user_id = strdup(user_id);
666 		if (!rbd->user_id) {
667 			bdev_rbd_free(rbd);
668 			return -ENOMEM;
669 		}
670 	}
671 
672 	rbd->pool_name = strdup(pool_name);
673 	if (!rbd->pool_name) {
674 		bdev_rbd_free(rbd);
675 		return -ENOMEM;
676 	}
677 
678 	if (config && !(rbd->config = bdev_rbd_dup_config(config))) {
679 		bdev_rbd_free(rbd);
680 		return -ENOMEM;
681 	}
682 
683 	ret = bdev_rbd_init(rbd->user_id, rbd->pool_name,
684 			    (const char *const *)rbd->config,
685 			    rbd_name, &rbd->info);
686 	if (ret < 0) {
687 		bdev_rbd_free(rbd);
688 		SPDK_ERRLOG("Failed to init rbd device\n");
689 		return ret;
690 	}
691 
692 	if (name) {
693 		rbd->disk.name = strdup(name);
694 	} else {
695 		rbd->disk.name = spdk_sprintf_alloc("Ceph%d", bdev_rbd_count);
696 	}
697 	if (!rbd->disk.name) {
698 		bdev_rbd_free(rbd);
699 		return -ENOMEM;
700 	}
701 	rbd->disk.product_name = "Ceph Rbd Disk";
702 	bdev_rbd_count++;
703 
704 	rbd->disk.write_cache = 0;
705 	rbd->disk.blocklen = block_size;
706 	rbd->disk.blockcnt = rbd->info.size / rbd->disk.blocklen;
707 	rbd->disk.ctxt = rbd;
708 	rbd->disk.fn_table = &rbd_fn_table;
709 	rbd->disk.module = &rbd_if;
710 
711 	SPDK_NOTICELOG("Add %s rbd disk to lun\n", rbd->disk.name);
712 
713 	spdk_io_device_register(rbd, bdev_rbd_create_cb,
714 				bdev_rbd_destroy_cb,
715 				sizeof(struct bdev_rbd_io_channel),
716 				rbd_name);
717 	ret = spdk_bdev_register(&rbd->disk);
718 	if (ret) {
719 		spdk_io_device_unregister(rbd, NULL);
720 		bdev_rbd_free(rbd);
721 		return ret;
722 	}
723 
724 	*bdev = &(rbd->disk);
725 
726 	return ret;
727 }
728 
729 void
730 bdev_rbd_delete(struct spdk_bdev *bdev, spdk_delete_rbd_complete cb_fn, void *cb_arg)
731 {
732 	if (!bdev || bdev->module != &rbd_if) {
733 		cb_fn(cb_arg, -ENODEV);
734 		return;
735 	}
736 
737 	spdk_bdev_unregister(bdev, cb_fn, cb_arg);
738 }
739 
740 int
741 bdev_rbd_resize(struct spdk_bdev *bdev, const uint64_t new_size_in_mb)
742 {
743 	struct spdk_io_channel *ch;
744 	struct bdev_rbd_io_channel *rbd_io_ch;
745 	int rc;
746 	uint64_t new_size_in_byte;
747 	uint64_t current_size_in_mb;
748 
749 	if (bdev->module != &rbd_if) {
750 		return -EINVAL;
751 	}
752 
753 	current_size_in_mb = bdev->blocklen * bdev->blockcnt / (1024 * 1024);
754 	if (current_size_in_mb > new_size_in_mb) {
755 		SPDK_ERRLOG("The new bdev size must be lager than current bdev size.\n");
756 		return -EINVAL;
757 	}
758 
759 	ch = bdev_rbd_get_io_channel(bdev);
760 	rbd_io_ch = spdk_io_channel_get_ctx(ch);
761 	new_size_in_byte = new_size_in_mb * 1024 * 1024;
762 
763 	rc = rbd_resize(rbd_io_ch->image, new_size_in_byte);
764 	if (rc != 0) {
765 		SPDK_ERRLOG("failed to resize the ceph bdev.\n");
766 		return rc;
767 	}
768 
769 	rc = spdk_bdev_notify_blockcnt_change(bdev, new_size_in_byte / bdev->blocklen);
770 	if (rc != 0) {
771 		SPDK_ERRLOG("failed to notify block cnt change.\n");
772 		return rc;
773 	}
774 
775 	return rc;
776 }
777 
778 static int
779 bdev_rbd_group_poll(void *arg)
780 {
781 	struct bdev_rbd_group_channel *group_ch = arg;
782 	struct epoll_event events[MAX_EVENTS_PER_POLL];
783 	int num_events, i;
784 
785 	num_events = epoll_wait(group_ch->epoll_fd, events, MAX_EVENTS_PER_POLL, 0);
786 
787 	if (num_events <= 0) {
788 		return SPDK_POLLER_IDLE;
789 	}
790 
791 	for (i = 0; i < num_events; i++) {
792 		bdev_rbd_io_poll((struct bdev_rbd_io_channel *)events[i].data.ptr);
793 	}
794 
795 	return SPDK_POLLER_BUSY;
796 }
797 
798 static int
799 bdev_rbd_group_create_cb(void *io_device, void *ctx_buf)
800 {
801 	struct bdev_rbd_group_channel *ch = ctx_buf;
802 
803 	ch->epoll_fd = epoll_create1(0);
804 	if (ch->epoll_fd < 0) {
805 		SPDK_ERRLOG("Could not create epoll fd on io device=%p\n", io_device);
806 		return -1;
807 	}
808 
809 	ch->poller = SPDK_POLLER_REGISTER(bdev_rbd_group_poll, ch, 0);
810 
811 	return 0;
812 }
813 
814 static void
815 bdev_rbd_group_destroy_cb(void *io_device, void *ctx_buf)
816 {
817 	struct bdev_rbd_group_channel *ch = ctx_buf;
818 
819 	if (ch->epoll_fd >= 0) {
820 		close(ch->epoll_fd);
821 	}
822 
823 	spdk_poller_unregister(&ch->poller);
824 }
825 
826 static int
827 bdev_rbd_library_init(void)
828 {
829 	spdk_io_device_register(&rbd_if, bdev_rbd_group_create_cb, bdev_rbd_group_destroy_cb,
830 				sizeof(struct bdev_rbd_group_channel), "bdev_rbd_poll_groups");
831 
832 	return 0;
833 }
834 
835 static void
836 bdev_rbd_library_fini(void)
837 {
838 	spdk_io_device_unregister(&rbd_if, NULL);
839 }
840 
841 SPDK_LOG_REGISTER_COMPONENT(bdev_rbd)
842