Commit | Line | Data |
---|---|---|
602adf40 YS |
1 | /* |
2 | rbd.c -- Export ceph rados objects as a Linux block device | |
3 | ||
4 | ||
5 | based on drivers/block/osdblk.c: | |
6 | ||
7 | Copyright 2009 Red Hat, Inc. | |
8 | ||
9 | This program is free software; you can redistribute it and/or modify | |
10 | it under the terms of the GNU General Public License as published by | |
11 | the Free Software Foundation. | |
12 | ||
13 | This program is distributed in the hope that it will be useful, | |
14 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | GNU General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU General Public License | |
19 | along with this program; see the file COPYING. If not, write to | |
20 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
21 | ||
22 | ||
23 | ||
dfc5606d | 24 | For usage instructions, please refer to: |
602adf40 | 25 | |
dfc5606d | 26 | Documentation/ABI/testing/sysfs-bus-rbd |
602adf40 YS |
27 | |
28 | */ | |
29 | ||
30 | #include <linux/ceph/libceph.h> | |
31 | #include <linux/ceph/osd_client.h> | |
32 | #include <linux/ceph/mon_client.h> | |
33 | #include <linux/ceph/decode.h> | |
59c2be1e | 34 | #include <linux/parser.h> |
602adf40 YS |
35 | |
36 | #include <linux/kernel.h> | |
37 | #include <linux/device.h> | |
38 | #include <linux/module.h> | |
39 | #include <linux/fs.h> | |
40 | #include <linux/blkdev.h> | |
41 | ||
42 | #include "rbd_types.h" | |
43 | ||
44 | #define DRV_NAME "rbd" | |
45 | #define DRV_NAME_LONG "rbd (rados block device)" | |
46 | ||
47 | #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ | |
48 | ||
21079786 | 49 | #define RBD_MAX_MD_NAME_LEN (RBD_MAX_OBJ_NAME_LEN + sizeof(RBD_SUFFIX)) |
602adf40 YS |
50 | #define RBD_MAX_POOL_NAME_LEN 64 |
51 | #define RBD_MAX_SNAP_NAME_LEN 32 | |
52 | #define RBD_MAX_OPT_LEN 1024 | |
53 | ||
54 | #define RBD_SNAP_HEAD_NAME "-" | |
55 | ||
56 | #define DEV_NAME_LEN 32 | |
57 | ||
59c2be1e YS |
58 | #define RBD_NOTIFY_TIMEOUT_DEFAULT 10 |
59 | ||
602adf40 YS |
60 | /* |
61 | * block device image metadata (in-memory version) | |
62 | */ | |
63 | struct rbd_image_header { | |
64 | u64 image_size; | |
65 | char block_name[32]; | |
66 | __u8 obj_order; | |
67 | __u8 crypt_type; | |
68 | __u8 comp_type; | |
69 | struct rw_semaphore snap_rwsem; | |
70 | struct ceph_snap_context *snapc; | |
71 | size_t snap_names_len; | |
72 | u64 snap_seq; | |
73 | u32 total_snaps; | |
74 | ||
75 | char *snap_names; | |
76 | u64 *snap_sizes; | |
59c2be1e YS |
77 | |
78 | u64 obj_version; | |
79 | }; | |
80 | ||
81 | struct rbd_options { | |
82 | int notify_timeout; | |
602adf40 YS |
83 | }; |
84 | ||
85 | /* | |
86 | * an instance of the client. multiple devices may share a client. | |
87 | */ | |
88 | struct rbd_client { | |
89 | struct ceph_client *client; | |
59c2be1e | 90 | struct rbd_options *rbd_opts; |
602adf40 YS |
91 | struct kref kref; |
92 | struct list_head node; | |
93 | }; | |
94 | ||
1fec7093 YS |
95 | struct rbd_req_coll; |
96 | ||
602adf40 YS |
97 | /* |
98 | * a single io request | |
99 | */ | |
100 | struct rbd_request { | |
101 | struct request *rq; /* blk layer request */ | |
102 | struct bio *bio; /* cloned bio */ | |
103 | struct page **pages; /* list of used pages */ | |
104 | u64 len; | |
1fec7093 YS |
105 | int coll_index; |
106 | struct rbd_req_coll *coll; | |
107 | }; | |
108 | ||
109 | struct rbd_req_status { | |
110 | int done; | |
111 | int rc; | |
112 | u64 bytes; | |
113 | }; | |
114 | ||
115 | /* | |
116 | * a collection of requests | |
117 | */ | |
118 | struct rbd_req_coll { | |
119 | int total; | |
120 | int num_done; | |
121 | struct kref kref; | |
122 | struct rbd_req_status status[0]; | |
602adf40 YS |
123 | }; |
124 | ||
dfc5606d YS |
125 | struct rbd_snap { |
126 | struct device dev; | |
127 | const char *name; | |
128 | size_t size; | |
129 | struct list_head node; | |
130 | u64 id; | |
131 | }; | |
132 | ||
602adf40 YS |
133 | /* |
134 | * a single device | |
135 | */ | |
136 | struct rbd_device { | |
137 | int id; /* blkdev unique id */ | |
138 | ||
139 | int major; /* blkdev assigned major */ | |
140 | struct gendisk *disk; /* blkdev's gendisk and rq */ | |
141 | struct request_queue *q; | |
142 | ||
602adf40 YS |
143 | struct rbd_client *rbd_client; |
144 | ||
145 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ | |
146 | ||
147 | spinlock_t lock; /* queue lock */ | |
148 | ||
149 | struct rbd_image_header header; | |
150 | char obj[RBD_MAX_OBJ_NAME_LEN]; /* rbd image name */ | |
151 | int obj_len; | |
152 | char obj_md_name[RBD_MAX_MD_NAME_LEN]; /* hdr nm. */ | |
153 | char pool_name[RBD_MAX_POOL_NAME_LEN]; | |
154 | int poolid; | |
155 | ||
59c2be1e YS |
156 | struct ceph_osd_event *watch_event; |
157 | struct ceph_osd_request *watch_request; | |
158 | ||
602adf40 YS |
159 | char snap_name[RBD_MAX_SNAP_NAME_LEN]; |
160 | u32 cur_snap; /* index+1 of current snapshot within snap context | |
161 | 0 - for the head */ | |
162 | int read_only; | |
163 | ||
164 | struct list_head node; | |
dfc5606d YS |
165 | |
166 | /* list of snapshots */ | |
167 | struct list_head snaps; | |
168 | ||
169 | /* sysfs related */ | |
170 | struct device dev; | |
171 | }; | |
172 | ||
173 | static struct bus_type rbd_bus_type = { | |
174 | .name = "rbd", | |
602adf40 YS |
175 | }; |
176 | ||
21079786 | 177 | static DEFINE_SPINLOCK(node_lock); /* protects client get/put */ |
602adf40 | 178 | |
602adf40 YS |
179 | static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ |
180 | static LIST_HEAD(rbd_dev_list); /* devices */ | |
181 | static LIST_HEAD(rbd_client_list); /* clients */ | |
182 | ||
dfc5606d YS |
183 | static int __rbd_init_snaps_header(struct rbd_device *rbd_dev); |
184 | static void rbd_dev_release(struct device *dev); | |
dfc5606d YS |
185 | static ssize_t rbd_snap_add(struct device *dev, |
186 | struct device_attribute *attr, | |
187 | const char *buf, | |
188 | size_t count); | |
189 | static void __rbd_remove_snap_dev(struct rbd_device *rbd_dev, | |
69932487 | 190 | struct rbd_snap *snap); |
dfc5606d YS |
191 | |
192 | ||
193 | static struct rbd_device *dev_to_rbd(struct device *dev) | |
194 | { | |
195 | return container_of(dev, struct rbd_device, dev); | |
196 | } | |
197 | ||
198 | static struct device *rbd_get_dev(struct rbd_device *rbd_dev) | |
199 | { | |
200 | return get_device(&rbd_dev->dev); | |
201 | } | |
202 | ||
203 | static void rbd_put_dev(struct rbd_device *rbd_dev) | |
204 | { | |
205 | put_device(&rbd_dev->dev); | |
206 | } | |
602adf40 | 207 | |
59c2be1e YS |
208 | static int __rbd_update_snaps(struct rbd_device *rbd_dev); |
209 | ||
602adf40 YS |
210 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
211 | { | |
212 | struct gendisk *disk = bdev->bd_disk; | |
213 | struct rbd_device *rbd_dev = disk->private_data; | |
214 | ||
dfc5606d YS |
215 | rbd_get_dev(rbd_dev); |
216 | ||
602adf40 YS |
217 | set_device_ro(bdev, rbd_dev->read_only); |
218 | ||
219 | if ((mode & FMODE_WRITE) && rbd_dev->read_only) | |
220 | return -EROFS; | |
221 | ||
222 | return 0; | |
223 | } | |
224 | ||
dfc5606d YS |
225 | static int rbd_release(struct gendisk *disk, fmode_t mode) |
226 | { | |
227 | struct rbd_device *rbd_dev = disk->private_data; | |
228 | ||
229 | rbd_put_dev(rbd_dev); | |
230 | ||
231 | return 0; | |
232 | } | |
233 | ||
602adf40 YS |
234 | static const struct block_device_operations rbd_bd_ops = { |
235 | .owner = THIS_MODULE, | |
236 | .open = rbd_open, | |
dfc5606d | 237 | .release = rbd_release, |
602adf40 YS |
238 | }; |
239 | ||
240 | /* | |
241 | * Initialize an rbd client instance. | |
242 | * We own *opt. | |
243 | */ | |
59c2be1e YS |
244 | static struct rbd_client *rbd_client_create(struct ceph_options *opt, |
245 | struct rbd_options *rbd_opts) | |
602adf40 YS |
246 | { |
247 | struct rbd_client *rbdc; | |
248 | int ret = -ENOMEM; | |
249 | ||
250 | dout("rbd_client_create\n"); | |
251 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); | |
252 | if (!rbdc) | |
253 | goto out_opt; | |
254 | ||
255 | kref_init(&rbdc->kref); | |
256 | INIT_LIST_HEAD(&rbdc->node); | |
257 | ||
6ab00d46 | 258 | rbdc->client = ceph_create_client(opt, rbdc, 0, 0); |
602adf40 YS |
259 | if (IS_ERR(rbdc->client)) |
260 | goto out_rbdc; | |
28f259b7 | 261 | opt = NULL; /* Now rbdc->client is responsible for opt */ |
602adf40 YS |
262 | |
263 | ret = ceph_open_session(rbdc->client); | |
264 | if (ret < 0) | |
265 | goto out_err; | |
266 | ||
59c2be1e YS |
267 | rbdc->rbd_opts = rbd_opts; |
268 | ||
602adf40 YS |
269 | spin_lock(&node_lock); |
270 | list_add_tail(&rbdc->node, &rbd_client_list); | |
271 | spin_unlock(&node_lock); | |
272 | ||
273 | dout("rbd_client_create created %p\n", rbdc); | |
274 | return rbdc; | |
275 | ||
276 | out_err: | |
277 | ceph_destroy_client(rbdc->client); | |
602adf40 YS |
278 | out_rbdc: |
279 | kfree(rbdc); | |
280 | out_opt: | |
28f259b7 VK |
281 | if (opt) |
282 | ceph_destroy_options(opt); | |
283 | return ERR_PTR(ret); | |
602adf40 YS |
284 | } |
285 | ||
286 | /* | |
287 | * Find a ceph client with specific addr and configuration. | |
288 | */ | |
289 | static struct rbd_client *__rbd_client_find(struct ceph_options *opt) | |
290 | { | |
291 | struct rbd_client *client_node; | |
292 | ||
293 | if (opt->flags & CEPH_OPT_NOSHARE) | |
294 | return NULL; | |
295 | ||
296 | list_for_each_entry(client_node, &rbd_client_list, node) | |
297 | if (ceph_compare_options(opt, client_node->client) == 0) | |
298 | return client_node; | |
299 | return NULL; | |
300 | } | |
301 | ||
59c2be1e YS |
302 | /* |
303 | * mount options | |
304 | */ | |
305 | enum { | |
306 | Opt_notify_timeout, | |
307 | Opt_last_int, | |
308 | /* int args above */ | |
309 | Opt_last_string, | |
310 | /* string args above */ | |
311 | }; | |
312 | ||
313 | static match_table_t rbdopt_tokens = { | |
314 | {Opt_notify_timeout, "notify_timeout=%d"}, | |
315 | /* int args above */ | |
316 | /* string args above */ | |
317 | {-1, NULL} | |
318 | }; | |
319 | ||
320 | static int parse_rbd_opts_token(char *c, void *private) | |
321 | { | |
322 | struct rbd_options *rbdopt = private; | |
323 | substring_t argstr[MAX_OPT_ARGS]; | |
324 | int token, intval, ret; | |
325 | ||
21079786 | 326 | token = match_token(c, rbdopt_tokens, argstr); |
59c2be1e YS |
327 | if (token < 0) |
328 | return -EINVAL; | |
329 | ||
330 | if (token < Opt_last_int) { | |
331 | ret = match_int(&argstr[0], &intval); | |
332 | if (ret < 0) { | |
333 | pr_err("bad mount option arg (not int) " | |
334 | "at '%s'\n", c); | |
335 | return ret; | |
336 | } | |
337 | dout("got int token %d val %d\n", token, intval); | |
338 | } else if (token > Opt_last_int && token < Opt_last_string) { | |
339 | dout("got string token %d val %s\n", token, | |
340 | argstr[0].from); | |
341 | } else { | |
342 | dout("got token %d\n", token); | |
343 | } | |
344 | ||
345 | switch (token) { | |
346 | case Opt_notify_timeout: | |
347 | rbdopt->notify_timeout = intval; | |
348 | break; | |
349 | default: | |
350 | BUG_ON(token); | |
351 | } | |
352 | return 0; | |
353 | } | |
354 | ||
602adf40 YS |
355 | /* |
356 | * Get a ceph client with specific addr and configuration, if one does | |
357 | * not exist create it. | |
358 | */ | |
359 | static int rbd_get_client(struct rbd_device *rbd_dev, const char *mon_addr, | |
360 | char *options) | |
361 | { | |
362 | struct rbd_client *rbdc; | |
363 | struct ceph_options *opt; | |
364 | int ret; | |
59c2be1e YS |
365 | struct rbd_options *rbd_opts; |
366 | ||
367 | rbd_opts = kzalloc(sizeof(*rbd_opts), GFP_KERNEL); | |
368 | if (!rbd_opts) | |
369 | return -ENOMEM; | |
370 | ||
371 | rbd_opts->notify_timeout = RBD_NOTIFY_TIMEOUT_DEFAULT; | |
602adf40 | 372 | |
ee57741c | 373 | opt = ceph_parse_options(options, mon_addr, |
21079786 AE |
374 | mon_addr + strlen(mon_addr), |
375 | parse_rbd_opts_token, rbd_opts); | |
ee57741c AE |
376 | if (IS_ERR(opt)) { |
377 | ret = PTR_ERR(opt); | |
59c2be1e | 378 | goto done_err; |
ee57741c | 379 | } |
602adf40 YS |
380 | |
381 | spin_lock(&node_lock); | |
382 | rbdc = __rbd_client_find(opt); | |
383 | if (rbdc) { | |
384 | ceph_destroy_options(opt); | |
97bb59a0 | 385 | kfree(rbd_opts); |
602adf40 YS |
386 | |
387 | /* using an existing client */ | |
388 | kref_get(&rbdc->kref); | |
389 | rbd_dev->rbd_client = rbdc; | |
602adf40 YS |
390 | spin_unlock(&node_lock); |
391 | return 0; | |
392 | } | |
393 | spin_unlock(&node_lock); | |
394 | ||
59c2be1e YS |
395 | rbdc = rbd_client_create(opt, rbd_opts); |
396 | if (IS_ERR(rbdc)) { | |
397 | ret = PTR_ERR(rbdc); | |
398 | goto done_err; | |
399 | } | |
602adf40 YS |
400 | |
401 | rbd_dev->rbd_client = rbdc; | |
602adf40 | 402 | return 0; |
59c2be1e YS |
403 | done_err: |
404 | kfree(rbd_opts); | |
405 | return ret; | |
602adf40 YS |
406 | } |
407 | ||
408 | /* | |
409 | * Destroy ceph client | |
d23a4b3f AE |
410 | * |
411 | * Caller must hold node_lock. | |
602adf40 YS |
412 | */ |
413 | static void rbd_client_release(struct kref *kref) | |
414 | { | |
415 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); | |
416 | ||
417 | dout("rbd_release_client %p\n", rbdc); | |
602adf40 | 418 | list_del(&rbdc->node); |
602adf40 YS |
419 | |
420 | ceph_destroy_client(rbdc->client); | |
59c2be1e | 421 | kfree(rbdc->rbd_opts); |
602adf40 YS |
422 | kfree(rbdc); |
423 | } | |
424 | ||
425 | /* | |
426 | * Drop reference to ceph client node. If it's not referenced anymore, release | |
427 | * it. | |
428 | */ | |
429 | static void rbd_put_client(struct rbd_device *rbd_dev) | |
430 | { | |
d23a4b3f | 431 | spin_lock(&node_lock); |
602adf40 | 432 | kref_put(&rbd_dev->rbd_client->kref, rbd_client_release); |
d23a4b3f | 433 | spin_unlock(&node_lock); |
602adf40 | 434 | rbd_dev->rbd_client = NULL; |
602adf40 YS |
435 | } |
436 | ||
1fec7093 YS |
437 | /* |
438 | * Destroy requests collection | |
439 | */ | |
440 | static void rbd_coll_release(struct kref *kref) | |
441 | { | |
442 | struct rbd_req_coll *coll = | |
443 | container_of(kref, struct rbd_req_coll, kref); | |
444 | ||
445 | dout("rbd_coll_release %p\n", coll); | |
446 | kfree(coll); | |
447 | } | |
602adf40 YS |
448 | |
449 | /* | |
450 | * Create a new header structure, translate header format from the on-disk | |
451 | * header. | |
452 | */ | |
453 | static int rbd_header_from_disk(struct rbd_image_header *header, | |
454 | struct rbd_image_header_ondisk *ondisk, | |
455 | int allocated_snaps, | |
456 | gfp_t gfp_flags) | |
457 | { | |
458 | int i; | |
459 | u32 snap_count = le32_to_cpu(ondisk->snap_count); | |
460 | int ret = -ENOMEM; | |
461 | ||
21079786 | 462 | if (memcmp(ondisk, RBD_HEADER_TEXT, sizeof(RBD_HEADER_TEXT))) |
81e759fb | 463 | return -ENXIO; |
81e759fb | 464 | |
602adf40 | 465 | init_rwsem(&header->snap_rwsem); |
602adf40 YS |
466 | header->snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
467 | header->snapc = kmalloc(sizeof(struct ceph_snap_context) + | |
21079786 | 468 | snap_count * sizeof (*ondisk), |
602adf40 YS |
469 | gfp_flags); |
470 | if (!header->snapc) | |
471 | return -ENOMEM; | |
472 | if (snap_count) { | |
473 | header->snap_names = kmalloc(header->snap_names_len, | |
474 | GFP_KERNEL); | |
475 | if (!header->snap_names) | |
476 | goto err_snapc; | |
477 | header->snap_sizes = kmalloc(snap_count * sizeof(u64), | |
478 | GFP_KERNEL); | |
479 | if (!header->snap_sizes) | |
480 | goto err_names; | |
481 | } else { | |
482 | header->snap_names = NULL; | |
483 | header->snap_sizes = NULL; | |
484 | } | |
485 | memcpy(header->block_name, ondisk->block_name, | |
486 | sizeof(ondisk->block_name)); | |
487 | ||
488 | header->image_size = le64_to_cpu(ondisk->image_size); | |
489 | header->obj_order = ondisk->options.order; | |
490 | header->crypt_type = ondisk->options.crypt_type; | |
491 | header->comp_type = ondisk->options.comp_type; | |
492 | ||
493 | atomic_set(&header->snapc->nref, 1); | |
494 | header->snap_seq = le64_to_cpu(ondisk->snap_seq); | |
495 | header->snapc->num_snaps = snap_count; | |
496 | header->total_snaps = snap_count; | |
497 | ||
21079786 | 498 | if (snap_count && allocated_snaps == snap_count) { |
602adf40 YS |
499 | for (i = 0; i < snap_count; i++) { |
500 | header->snapc->snaps[i] = | |
501 | le64_to_cpu(ondisk->snaps[i].id); | |
502 | header->snap_sizes[i] = | |
503 | le64_to_cpu(ondisk->snaps[i].image_size); | |
504 | } | |
505 | ||
506 | /* copy snapshot names */ | |
507 | memcpy(header->snap_names, &ondisk->snaps[i], | |
508 | header->snap_names_len); | |
509 | } | |
510 | ||
511 | return 0; | |
512 | ||
513 | err_names: | |
514 | kfree(header->snap_names); | |
515 | err_snapc: | |
516 | kfree(header->snapc); | |
517 | return ret; | |
518 | } | |
519 | ||
520 | static int snap_index(struct rbd_image_header *header, int snap_num) | |
521 | { | |
522 | return header->total_snaps - snap_num; | |
523 | } | |
524 | ||
525 | static u64 cur_snap_id(struct rbd_device *rbd_dev) | |
526 | { | |
527 | struct rbd_image_header *header = &rbd_dev->header; | |
528 | ||
529 | if (!rbd_dev->cur_snap) | |
530 | return 0; | |
531 | ||
532 | return header->snapc->snaps[snap_index(header, rbd_dev->cur_snap)]; | |
533 | } | |
534 | ||
535 | static int snap_by_name(struct rbd_image_header *header, const char *snap_name, | |
536 | u64 *seq, u64 *size) | |
537 | { | |
538 | int i; | |
539 | char *p = header->snap_names; | |
540 | ||
541 | for (i = 0; i < header->total_snaps; i++, p += strlen(p) + 1) { | |
542 | if (strcmp(snap_name, p) == 0) | |
543 | break; | |
544 | } | |
545 | if (i == header->total_snaps) | |
546 | return -ENOENT; | |
547 | if (seq) | |
548 | *seq = header->snapc->snaps[i]; | |
549 | ||
550 | if (size) | |
551 | *size = header->snap_sizes[i]; | |
552 | ||
553 | return i; | |
554 | } | |
555 | ||
cc9d734c | 556 | static int rbd_header_set_snap(struct rbd_device *dev, u64 *size) |
602adf40 YS |
557 | { |
558 | struct rbd_image_header *header = &dev->header; | |
559 | struct ceph_snap_context *snapc = header->snapc; | |
560 | int ret = -ENOENT; | |
561 | ||
cc9d734c JD |
562 | BUILD_BUG_ON(sizeof (dev->snap_name) < sizeof (RBD_SNAP_HEAD_NAME)); |
563 | ||
602adf40 YS |
564 | down_write(&header->snap_rwsem); |
565 | ||
cc9d734c JD |
566 | if (!memcmp(dev->snap_name, RBD_SNAP_HEAD_NAME, |
567 | sizeof (RBD_SNAP_HEAD_NAME))) { | |
602adf40 YS |
568 | if (header->total_snaps) |
569 | snapc->seq = header->snap_seq; | |
570 | else | |
571 | snapc->seq = 0; | |
572 | dev->cur_snap = 0; | |
573 | dev->read_only = 0; | |
574 | if (size) | |
575 | *size = header->image_size; | |
576 | } else { | |
cc9d734c | 577 | ret = snap_by_name(header, dev->snap_name, &snapc->seq, size); |
602adf40 YS |
578 | if (ret < 0) |
579 | goto done; | |
580 | ||
581 | dev->cur_snap = header->total_snaps - ret; | |
582 | dev->read_only = 1; | |
583 | } | |
584 | ||
585 | ret = 0; | |
586 | done: | |
587 | up_write(&header->snap_rwsem); | |
588 | return ret; | |
589 | } | |
590 | ||
591 | static void rbd_header_free(struct rbd_image_header *header) | |
592 | { | |
593 | kfree(header->snapc); | |
594 | kfree(header->snap_names); | |
595 | kfree(header->snap_sizes); | |
596 | } | |
597 | ||
598 | /* | |
599 | * get the actual striped segment name, offset and length | |
600 | */ | |
601 | static u64 rbd_get_segment(struct rbd_image_header *header, | |
602 | const char *block_name, | |
603 | u64 ofs, u64 len, | |
604 | char *seg_name, u64 *segofs) | |
605 | { | |
606 | u64 seg = ofs >> header->obj_order; | |
607 | ||
608 | if (seg_name) | |
609 | snprintf(seg_name, RBD_MAX_SEG_NAME_LEN, | |
610 | "%s.%012llx", block_name, seg); | |
611 | ||
612 | ofs = ofs & ((1 << header->obj_order) - 1); | |
613 | len = min_t(u64, len, (1 << header->obj_order) - ofs); | |
614 | ||
615 | if (segofs) | |
616 | *segofs = ofs; | |
617 | ||
618 | return len; | |
619 | } | |
620 | ||
1fec7093 YS |
621 | static int rbd_get_num_segments(struct rbd_image_header *header, |
622 | u64 ofs, u64 len) | |
623 | { | |
624 | u64 start_seg = ofs >> header->obj_order; | |
625 | u64 end_seg = (ofs + len - 1) >> header->obj_order; | |
626 | return end_seg - start_seg + 1; | |
627 | } | |
628 | ||
029bcbd8 JD |
629 | /* |
630 | * returns the size of an object in the image | |
631 | */ | |
632 | static u64 rbd_obj_bytes(struct rbd_image_header *header) | |
633 | { | |
634 | return 1 << header->obj_order; | |
635 | } | |
636 | ||
602adf40 YS |
637 | /* |
638 | * bio helpers | |
639 | */ | |
640 | ||
641 | static void bio_chain_put(struct bio *chain) | |
642 | { | |
643 | struct bio *tmp; | |
644 | ||
645 | while (chain) { | |
646 | tmp = chain; | |
647 | chain = chain->bi_next; | |
648 | bio_put(tmp); | |
649 | } | |
650 | } | |
651 | ||
652 | /* | |
653 | * zeros a bio chain, starting at specific offset | |
654 | */ | |
655 | static void zero_bio_chain(struct bio *chain, int start_ofs) | |
656 | { | |
657 | struct bio_vec *bv; | |
658 | unsigned long flags; | |
659 | void *buf; | |
660 | int i; | |
661 | int pos = 0; | |
662 | ||
663 | while (chain) { | |
664 | bio_for_each_segment(bv, chain, i) { | |
665 | if (pos + bv->bv_len > start_ofs) { | |
666 | int remainder = max(start_ofs - pos, 0); | |
667 | buf = bvec_kmap_irq(bv, &flags); | |
668 | memset(buf + remainder, 0, | |
669 | bv->bv_len - remainder); | |
85b5aaa6 | 670 | bvec_kunmap_irq(buf, &flags); |
602adf40 YS |
671 | } |
672 | pos += bv->bv_len; | |
673 | } | |
674 | ||
675 | chain = chain->bi_next; | |
676 | } | |
677 | } | |
678 | ||
679 | /* | |
680 | * bio_chain_clone - clone a chain of bios up to a certain length. | |
681 | * might return a bio_pair that will need to be released. | |
682 | */ | |
683 | static struct bio *bio_chain_clone(struct bio **old, struct bio **next, | |
684 | struct bio_pair **bp, | |
685 | int len, gfp_t gfpmask) | |
686 | { | |
687 | struct bio *tmp, *old_chain = *old, *new_chain = NULL, *tail = NULL; | |
688 | int total = 0; | |
689 | ||
690 | if (*bp) { | |
691 | bio_pair_release(*bp); | |
692 | *bp = NULL; | |
693 | } | |
694 | ||
695 | while (old_chain && (total < len)) { | |
696 | tmp = bio_kmalloc(gfpmask, old_chain->bi_max_vecs); | |
697 | if (!tmp) | |
698 | goto err_out; | |
699 | ||
700 | if (total + old_chain->bi_size > len) { | |
701 | struct bio_pair *bp; | |
702 | ||
703 | /* | |
704 | * this split can only happen with a single paged bio, | |
705 | * split_bio will BUG_ON if this is not the case | |
706 | */ | |
707 | dout("bio_chain_clone split! total=%d remaining=%d" | |
708 | "bi_size=%d\n", | |
709 | (int)total, (int)len-total, | |
710 | (int)old_chain->bi_size); | |
711 | ||
712 | /* split the bio. We'll release it either in the next | |
713 | call, or it will have to be released outside */ | |
714 | bp = bio_split(old_chain, (len - total) / 512ULL); | |
715 | if (!bp) | |
716 | goto err_out; | |
717 | ||
718 | __bio_clone(tmp, &bp->bio1); | |
719 | ||
720 | *next = &bp->bio2; | |
721 | } else { | |
722 | __bio_clone(tmp, old_chain); | |
723 | *next = old_chain->bi_next; | |
724 | } | |
725 | ||
726 | tmp->bi_bdev = NULL; | |
727 | gfpmask &= ~__GFP_WAIT; | |
728 | tmp->bi_next = NULL; | |
729 | ||
730 | if (!new_chain) { | |
731 | new_chain = tail = tmp; | |
732 | } else { | |
733 | tail->bi_next = tmp; | |
734 | tail = tmp; | |
735 | } | |
736 | old_chain = old_chain->bi_next; | |
737 | ||
738 | total += tmp->bi_size; | |
739 | } | |
740 | ||
741 | BUG_ON(total < len); | |
742 | ||
743 | if (tail) | |
744 | tail->bi_next = NULL; | |
745 | ||
746 | *old = old_chain; | |
747 | ||
748 | return new_chain; | |
749 | ||
750 | err_out: | |
751 | dout("bio_chain_clone with err\n"); | |
752 | bio_chain_put(new_chain); | |
753 | return NULL; | |
754 | } | |
755 | ||
756 | /* | |
757 | * helpers for osd request op vectors. | |
758 | */ | |
759 | static int rbd_create_rw_ops(struct ceph_osd_req_op **ops, | |
760 | int num_ops, | |
761 | int opcode, | |
762 | u32 payload_len) | |
763 | { | |
764 | *ops = kzalloc(sizeof(struct ceph_osd_req_op) * (num_ops + 1), | |
765 | GFP_NOIO); | |
766 | if (!*ops) | |
767 | return -ENOMEM; | |
768 | (*ops)[0].op = opcode; | |
769 | /* | |
770 | * op extent offset and length will be set later on | |
771 | * in calc_raw_layout() | |
772 | */ | |
773 | (*ops)[0].payload_len = payload_len; | |
774 | return 0; | |
775 | } | |
776 | ||
777 | static void rbd_destroy_ops(struct ceph_osd_req_op *ops) | |
778 | { | |
779 | kfree(ops); | |
780 | } | |
781 | ||
1fec7093 YS |
782 | static void rbd_coll_end_req_index(struct request *rq, |
783 | struct rbd_req_coll *coll, | |
784 | int index, | |
785 | int ret, u64 len) | |
786 | { | |
787 | struct request_queue *q; | |
788 | int min, max, i; | |
789 | ||
790 | dout("rbd_coll_end_req_index %p index %d ret %d len %lld\n", | |
791 | coll, index, ret, len); | |
792 | ||
793 | if (!rq) | |
794 | return; | |
795 | ||
796 | if (!coll) { | |
797 | blk_end_request(rq, ret, len); | |
798 | return; | |
799 | } | |
800 | ||
801 | q = rq->q; | |
802 | ||
803 | spin_lock_irq(q->queue_lock); | |
804 | coll->status[index].done = 1; | |
805 | coll->status[index].rc = ret; | |
806 | coll->status[index].bytes = len; | |
807 | max = min = coll->num_done; | |
808 | while (max < coll->total && coll->status[max].done) | |
809 | max++; | |
810 | ||
811 | for (i = min; i<max; i++) { | |
812 | __blk_end_request(rq, coll->status[i].rc, | |
813 | coll->status[i].bytes); | |
814 | coll->num_done++; | |
815 | kref_put(&coll->kref, rbd_coll_release); | |
816 | } | |
817 | spin_unlock_irq(q->queue_lock); | |
818 | } | |
819 | ||
820 | static void rbd_coll_end_req(struct rbd_request *req, | |
821 | int ret, u64 len) | |
822 | { | |
823 | rbd_coll_end_req_index(req->rq, req->coll, req->coll_index, ret, len); | |
824 | } | |
825 | ||
602adf40 YS |
826 | /* |
827 | * Send ceph osd request | |
828 | */ | |
829 | static int rbd_do_request(struct request *rq, | |
830 | struct rbd_device *dev, | |
831 | struct ceph_snap_context *snapc, | |
832 | u64 snapid, | |
833 | const char *obj, u64 ofs, u64 len, | |
834 | struct bio *bio, | |
835 | struct page **pages, | |
836 | int num_pages, | |
837 | int flags, | |
838 | struct ceph_osd_req_op *ops, | |
839 | int num_reply, | |
1fec7093 YS |
840 | struct rbd_req_coll *coll, |
841 | int coll_index, | |
602adf40 | 842 | void (*rbd_cb)(struct ceph_osd_request *req, |
59c2be1e YS |
843 | struct ceph_msg *msg), |
844 | struct ceph_osd_request **linger_req, | |
845 | u64 *ver) | |
602adf40 YS |
846 | { |
847 | struct ceph_osd_request *req; | |
848 | struct ceph_file_layout *layout; | |
849 | int ret; | |
850 | u64 bno; | |
851 | struct timespec mtime = CURRENT_TIME; | |
852 | struct rbd_request *req_data; | |
853 | struct ceph_osd_request_head *reqhead; | |
854 | struct rbd_image_header *header = &dev->header; | |
1dbb4399 | 855 | struct ceph_osd_client *osdc; |
602adf40 | 856 | |
602adf40 | 857 | req_data = kzalloc(sizeof(*req_data), GFP_NOIO); |
1fec7093 YS |
858 | if (!req_data) { |
859 | if (coll) | |
860 | rbd_coll_end_req_index(rq, coll, coll_index, | |
861 | -ENOMEM, len); | |
862 | return -ENOMEM; | |
863 | } | |
864 | ||
865 | if (coll) { | |
866 | req_data->coll = coll; | |
867 | req_data->coll_index = coll_index; | |
868 | } | |
602adf40 | 869 | |
1fec7093 | 870 | dout("rbd_do_request obj=%s ofs=%lld len=%lld\n", obj, len, ofs); |
602adf40 YS |
871 | |
872 | down_read(&header->snap_rwsem); | |
873 | ||
1dbb4399 AE |
874 | osdc = &dev->rbd_client->client->osdc; |
875 | req = ceph_osdc_alloc_request(osdc, flags, snapc, ops, | |
876 | false, GFP_NOIO, pages, bio); | |
4ad12621 | 877 | if (!req) { |
602adf40 | 878 | up_read(&header->snap_rwsem); |
4ad12621 | 879 | ret = -ENOMEM; |
602adf40 YS |
880 | goto done_pages; |
881 | } | |
882 | ||
883 | req->r_callback = rbd_cb; | |
884 | ||
885 | req_data->rq = rq; | |
886 | req_data->bio = bio; | |
887 | req_data->pages = pages; | |
888 | req_data->len = len; | |
889 | ||
890 | req->r_priv = req_data; | |
891 | ||
892 | reqhead = req->r_request->front.iov_base; | |
893 | reqhead->snapid = cpu_to_le64(CEPH_NOSNAP); | |
894 | ||
895 | strncpy(req->r_oid, obj, sizeof(req->r_oid)); | |
896 | req->r_oid_len = strlen(req->r_oid); | |
897 | ||
898 | layout = &req->r_file_layout; | |
899 | memset(layout, 0, sizeof(*layout)); | |
900 | layout->fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
901 | layout->fl_stripe_count = cpu_to_le32(1); | |
902 | layout->fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
903 | layout->fl_pg_preferred = cpu_to_le32(-1); | |
904 | layout->fl_pg_pool = cpu_to_le32(dev->poolid); | |
1dbb4399 AE |
905 | ceph_calc_raw_layout(osdc, layout, snapid, ofs, &len, &bno, |
906 | req, ops); | |
602adf40 YS |
907 | |
908 | ceph_osdc_build_request(req, ofs, &len, | |
909 | ops, | |
910 | snapc, | |
911 | &mtime, | |
912 | req->r_oid, req->r_oid_len); | |
913 | up_read(&header->snap_rwsem); | |
914 | ||
59c2be1e | 915 | if (linger_req) { |
1dbb4399 | 916 | ceph_osdc_set_request_linger(osdc, req); |
59c2be1e YS |
917 | *linger_req = req; |
918 | } | |
919 | ||
1dbb4399 | 920 | ret = ceph_osdc_start_request(osdc, req, false); |
602adf40 YS |
921 | if (ret < 0) |
922 | goto done_err; | |
923 | ||
924 | if (!rbd_cb) { | |
1dbb4399 | 925 | ret = ceph_osdc_wait_request(osdc, req); |
59c2be1e YS |
926 | if (ver) |
927 | *ver = le64_to_cpu(req->r_reassert_version.version); | |
1fec7093 YS |
928 | dout("reassert_ver=%lld\n", |
929 | le64_to_cpu(req->r_reassert_version.version)); | |
602adf40 YS |
930 | ceph_osdc_put_request(req); |
931 | } | |
932 | return ret; | |
933 | ||
934 | done_err: | |
935 | bio_chain_put(req_data->bio); | |
936 | ceph_osdc_put_request(req); | |
937 | done_pages: | |
1fec7093 | 938 | rbd_coll_end_req(req_data, ret, len); |
602adf40 | 939 | kfree(req_data); |
602adf40 YS |
940 | return ret; |
941 | } | |
942 | ||
943 | /* | |
944 | * Ceph osd op callback | |
945 | */ | |
946 | static void rbd_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) | |
947 | { | |
948 | struct rbd_request *req_data = req->r_priv; | |
949 | struct ceph_osd_reply_head *replyhead; | |
950 | struct ceph_osd_op *op; | |
951 | __s32 rc; | |
952 | u64 bytes; | |
953 | int read_op; | |
954 | ||
955 | /* parse reply */ | |
956 | replyhead = msg->front.iov_base; | |
957 | WARN_ON(le32_to_cpu(replyhead->num_ops) == 0); | |
958 | op = (void *)(replyhead + 1); | |
959 | rc = le32_to_cpu(replyhead->result); | |
960 | bytes = le64_to_cpu(op->extent.length); | |
961 | read_op = (le32_to_cpu(op->op) == CEPH_OSD_OP_READ); | |
962 | ||
963 | dout("rbd_req_cb bytes=%lld readop=%d rc=%d\n", bytes, read_op, rc); | |
964 | ||
965 | if (rc == -ENOENT && read_op) { | |
966 | zero_bio_chain(req_data->bio, 0); | |
967 | rc = 0; | |
968 | } else if (rc == 0 && read_op && bytes < req_data->len) { | |
969 | zero_bio_chain(req_data->bio, bytes); | |
970 | bytes = req_data->len; | |
971 | } | |
972 | ||
1fec7093 | 973 | rbd_coll_end_req(req_data, rc, bytes); |
602adf40 YS |
974 | |
975 | if (req_data->bio) | |
976 | bio_chain_put(req_data->bio); | |
977 | ||
978 | ceph_osdc_put_request(req); | |
979 | kfree(req_data); | |
980 | } | |
981 | ||
59c2be1e YS |
982 | static void rbd_simple_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) |
983 | { | |
984 | ceph_osdc_put_request(req); | |
985 | } | |
986 | ||
602adf40 YS |
987 | /* |
988 | * Do a synchronous ceph osd operation | |
989 | */ | |
990 | static int rbd_req_sync_op(struct rbd_device *dev, | |
991 | struct ceph_snap_context *snapc, | |
992 | u64 snapid, | |
993 | int opcode, | |
994 | int flags, | |
995 | struct ceph_osd_req_op *orig_ops, | |
996 | int num_reply, | |
997 | const char *obj, | |
998 | u64 ofs, u64 len, | |
59c2be1e YS |
999 | char *buf, |
1000 | struct ceph_osd_request **linger_req, | |
1001 | u64 *ver) | |
602adf40 YS |
1002 | { |
1003 | int ret; | |
1004 | struct page **pages; | |
1005 | int num_pages; | |
1006 | struct ceph_osd_req_op *ops = orig_ops; | |
1007 | u32 payload_len; | |
1008 | ||
1009 | num_pages = calc_pages_for(ofs , len); | |
1010 | pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); | |
b8d0638a DC |
1011 | if (IS_ERR(pages)) |
1012 | return PTR_ERR(pages); | |
602adf40 YS |
1013 | |
1014 | if (!orig_ops) { | |
1015 | payload_len = (flags & CEPH_OSD_FLAG_WRITE ? len : 0); | |
1016 | ret = rbd_create_rw_ops(&ops, 1, opcode, payload_len); | |
1017 | if (ret < 0) | |
1018 | goto done; | |
1019 | ||
1020 | if ((flags & CEPH_OSD_FLAG_WRITE) && buf) { | |
1021 | ret = ceph_copy_to_page_vector(pages, buf, ofs, len); | |
1022 | if (ret < 0) | |
1023 | goto done_ops; | |
1024 | } | |
1025 | } | |
1026 | ||
1027 | ret = rbd_do_request(NULL, dev, snapc, snapid, | |
1028 | obj, ofs, len, NULL, | |
1029 | pages, num_pages, | |
1030 | flags, | |
1031 | ops, | |
1032 | 2, | |
1fec7093 | 1033 | NULL, 0, |
59c2be1e YS |
1034 | NULL, |
1035 | linger_req, ver); | |
602adf40 YS |
1036 | if (ret < 0) |
1037 | goto done_ops; | |
1038 | ||
1039 | if ((flags & CEPH_OSD_FLAG_READ) && buf) | |
1040 | ret = ceph_copy_from_page_vector(pages, buf, ofs, ret); | |
1041 | ||
1042 | done_ops: | |
1043 | if (!orig_ops) | |
1044 | rbd_destroy_ops(ops); | |
1045 | done: | |
1046 | ceph_release_page_vector(pages, num_pages); | |
1047 | return ret; | |
1048 | } | |
1049 | ||
1050 | /* | |
1051 | * Do an asynchronous ceph osd operation | |
1052 | */ | |
1053 | static int rbd_do_op(struct request *rq, | |
1054 | struct rbd_device *rbd_dev , | |
1055 | struct ceph_snap_context *snapc, | |
1056 | u64 snapid, | |
1057 | int opcode, int flags, int num_reply, | |
1058 | u64 ofs, u64 len, | |
1fec7093 YS |
1059 | struct bio *bio, |
1060 | struct rbd_req_coll *coll, | |
1061 | int coll_index) | |
602adf40 YS |
1062 | { |
1063 | char *seg_name; | |
1064 | u64 seg_ofs; | |
1065 | u64 seg_len; | |
1066 | int ret; | |
1067 | struct ceph_osd_req_op *ops; | |
1068 | u32 payload_len; | |
1069 | ||
1070 | seg_name = kmalloc(RBD_MAX_SEG_NAME_LEN + 1, GFP_NOIO); | |
1071 | if (!seg_name) | |
1072 | return -ENOMEM; | |
1073 | ||
1074 | seg_len = rbd_get_segment(&rbd_dev->header, | |
1075 | rbd_dev->header.block_name, | |
1076 | ofs, len, | |
1077 | seg_name, &seg_ofs); | |
602adf40 YS |
1078 | |
1079 | payload_len = (flags & CEPH_OSD_FLAG_WRITE ? seg_len : 0); | |
1080 | ||
1081 | ret = rbd_create_rw_ops(&ops, 1, opcode, payload_len); | |
1082 | if (ret < 0) | |
1083 | goto done; | |
1084 | ||
1085 | /* we've taken care of segment sizes earlier when we | |
1086 | cloned the bios. We should never have a segment | |
1087 | truncated at this point */ | |
1088 | BUG_ON(seg_len < len); | |
1089 | ||
1090 | ret = rbd_do_request(rq, rbd_dev, snapc, snapid, | |
1091 | seg_name, seg_ofs, seg_len, | |
1092 | bio, | |
1093 | NULL, 0, | |
1094 | flags, | |
1095 | ops, | |
1096 | num_reply, | |
1fec7093 | 1097 | coll, coll_index, |
59c2be1e | 1098 | rbd_req_cb, 0, NULL); |
11f77002 SW |
1099 | |
1100 | rbd_destroy_ops(ops); | |
602adf40 YS |
1101 | done: |
1102 | kfree(seg_name); | |
1103 | return ret; | |
1104 | } | |
1105 | ||
1106 | /* | |
1107 | * Request async osd write | |
1108 | */ | |
1109 | static int rbd_req_write(struct request *rq, | |
1110 | struct rbd_device *rbd_dev, | |
1111 | struct ceph_snap_context *snapc, | |
1112 | u64 ofs, u64 len, | |
1fec7093 YS |
1113 | struct bio *bio, |
1114 | struct rbd_req_coll *coll, | |
1115 | int coll_index) | |
602adf40 YS |
1116 | { |
1117 | return rbd_do_op(rq, rbd_dev, snapc, CEPH_NOSNAP, | |
1118 | CEPH_OSD_OP_WRITE, | |
1119 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, | |
1120 | 2, | |
1fec7093 | 1121 | ofs, len, bio, coll, coll_index); |
602adf40 YS |
1122 | } |
1123 | ||
1124 | /* | |
1125 | * Request async osd read | |
1126 | */ | |
1127 | static int rbd_req_read(struct request *rq, | |
1128 | struct rbd_device *rbd_dev, | |
1129 | u64 snapid, | |
1130 | u64 ofs, u64 len, | |
1fec7093 YS |
1131 | struct bio *bio, |
1132 | struct rbd_req_coll *coll, | |
1133 | int coll_index) | |
602adf40 YS |
1134 | { |
1135 | return rbd_do_op(rq, rbd_dev, NULL, | |
1136 | (snapid ? snapid : CEPH_NOSNAP), | |
1137 | CEPH_OSD_OP_READ, | |
1138 | CEPH_OSD_FLAG_READ, | |
1139 | 2, | |
1fec7093 | 1140 | ofs, len, bio, coll, coll_index); |
602adf40 YS |
1141 | } |
1142 | ||
1143 | /* | |
1144 | * Request sync osd read | |
1145 | */ | |
1146 | static int rbd_req_sync_read(struct rbd_device *dev, | |
1147 | struct ceph_snap_context *snapc, | |
1148 | u64 snapid, | |
1149 | const char *obj, | |
1150 | u64 ofs, u64 len, | |
59c2be1e YS |
1151 | char *buf, |
1152 | u64 *ver) | |
602adf40 YS |
1153 | { |
1154 | return rbd_req_sync_op(dev, NULL, | |
1155 | (snapid ? snapid : CEPH_NOSNAP), | |
1156 | CEPH_OSD_OP_READ, | |
1157 | CEPH_OSD_FLAG_READ, | |
1158 | NULL, | |
59c2be1e | 1159 | 1, obj, ofs, len, buf, NULL, ver); |
602adf40 YS |
1160 | } |
1161 | ||
1162 | /* | |
59c2be1e YS |
1163 | * Request sync osd watch |
1164 | */ | |
1165 | static int rbd_req_sync_notify_ack(struct rbd_device *dev, | |
1166 | u64 ver, | |
1167 | u64 notify_id, | |
1168 | const char *obj) | |
1169 | { | |
1170 | struct ceph_osd_req_op *ops; | |
1171 | struct page **pages = NULL; | |
11f77002 SW |
1172 | int ret; |
1173 | ||
1174 | ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_NOTIFY_ACK, 0); | |
59c2be1e YS |
1175 | if (ret < 0) |
1176 | return ret; | |
1177 | ||
1178 | ops[0].watch.ver = cpu_to_le64(dev->header.obj_version); | |
1179 | ops[0].watch.cookie = notify_id; | |
1180 | ops[0].watch.flag = 0; | |
1181 | ||
1182 | ret = rbd_do_request(NULL, dev, NULL, CEPH_NOSNAP, | |
1183 | obj, 0, 0, NULL, | |
1184 | pages, 0, | |
1185 | CEPH_OSD_FLAG_READ, | |
1186 | ops, | |
1187 | 1, | |
1fec7093 | 1188 | NULL, 0, |
59c2be1e YS |
1189 | rbd_simple_req_cb, 0, NULL); |
1190 | ||
1191 | rbd_destroy_ops(ops); | |
1192 | return ret; | |
1193 | } | |
1194 | ||
1195 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) | |
1196 | { | |
1197 | struct rbd_device *dev = (struct rbd_device *)data; | |
13143d2d SW |
1198 | int rc; |
1199 | ||
59c2be1e YS |
1200 | if (!dev) |
1201 | return; | |
1202 | ||
1203 | dout("rbd_watch_cb %s notify_id=%lld opcode=%d\n", dev->obj_md_name, | |
1204 | notify_id, (int)opcode); | |
1205 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
13143d2d | 1206 | rc = __rbd_update_snaps(dev); |
59c2be1e | 1207 | mutex_unlock(&ctl_mutex); |
13143d2d SW |
1208 | if (rc) |
1209 | pr_warning(DRV_NAME "%d got notification but failed to update" | |
1210 | " snaps: %d\n", dev->major, rc); | |
59c2be1e YS |
1211 | |
1212 | rbd_req_sync_notify_ack(dev, ver, notify_id, dev->obj_md_name); | |
1213 | } | |
1214 | ||
1215 | /* | |
1216 | * Request sync osd watch | |
1217 | */ | |
1218 | static int rbd_req_sync_watch(struct rbd_device *dev, | |
1219 | const char *obj, | |
1220 | u64 ver) | |
1221 | { | |
1222 | struct ceph_osd_req_op *ops; | |
1dbb4399 | 1223 | struct ceph_osd_client *osdc = &dev->rbd_client->client->osdc; |
59c2be1e YS |
1224 | |
1225 | int ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_WATCH, 0); | |
1226 | if (ret < 0) | |
1227 | return ret; | |
1228 | ||
1229 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, 0, | |
1230 | (void *)dev, &dev->watch_event); | |
1231 | if (ret < 0) | |
1232 | goto fail; | |
1233 | ||
1234 | ops[0].watch.ver = cpu_to_le64(ver); | |
1235 | ops[0].watch.cookie = cpu_to_le64(dev->watch_event->cookie); | |
1236 | ops[0].watch.flag = 1; | |
1237 | ||
1238 | ret = rbd_req_sync_op(dev, NULL, | |
1239 | CEPH_NOSNAP, | |
1240 | 0, | |
1241 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, | |
1242 | ops, | |
1243 | 1, obj, 0, 0, NULL, | |
1244 | &dev->watch_request, NULL); | |
1245 | ||
1246 | if (ret < 0) | |
1247 | goto fail_event; | |
1248 | ||
1249 | rbd_destroy_ops(ops); | |
1250 | return 0; | |
1251 | ||
1252 | fail_event: | |
1253 | ceph_osdc_cancel_event(dev->watch_event); | |
1254 | dev->watch_event = NULL; | |
1255 | fail: | |
1256 | rbd_destroy_ops(ops); | |
1257 | return ret; | |
1258 | } | |
1259 | ||
79e3057c YS |
1260 | /* |
1261 | * Request sync osd unwatch | |
1262 | */ | |
1263 | static int rbd_req_sync_unwatch(struct rbd_device *dev, | |
1264 | const char *obj) | |
1265 | { | |
1266 | struct ceph_osd_req_op *ops; | |
1267 | ||
1268 | int ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_WATCH, 0); | |
1269 | if (ret < 0) | |
1270 | return ret; | |
1271 | ||
1272 | ops[0].watch.ver = 0; | |
1273 | ops[0].watch.cookie = cpu_to_le64(dev->watch_event->cookie); | |
1274 | ops[0].watch.flag = 0; | |
1275 | ||
1276 | ret = rbd_req_sync_op(dev, NULL, | |
1277 | CEPH_NOSNAP, | |
1278 | 0, | |
1279 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, | |
1280 | ops, | |
1281 | 1, obj, 0, 0, NULL, NULL, NULL); | |
1282 | ||
1283 | rbd_destroy_ops(ops); | |
1284 | ceph_osdc_cancel_event(dev->watch_event); | |
1285 | dev->watch_event = NULL; | |
1286 | return ret; | |
1287 | } | |
1288 | ||
59c2be1e YS |
1289 | struct rbd_notify_info { |
1290 | struct rbd_device *dev; | |
1291 | }; | |
1292 | ||
1293 | static void rbd_notify_cb(u64 ver, u64 notify_id, u8 opcode, void *data) | |
1294 | { | |
1295 | struct rbd_device *dev = (struct rbd_device *)data; | |
1296 | if (!dev) | |
1297 | return; | |
1298 | ||
1299 | dout("rbd_notify_cb %s notify_id=%lld opcode=%d\n", dev->obj_md_name, | |
1300 | notify_id, (int)opcode); | |
1301 | } | |
1302 | ||
1303 | /* | |
1304 | * Request sync osd notify | |
1305 | */ | |
1306 | static int rbd_req_sync_notify(struct rbd_device *dev, | |
1307 | const char *obj) | |
1308 | { | |
1309 | struct ceph_osd_req_op *ops; | |
1dbb4399 | 1310 | struct ceph_osd_client *osdc = &dev->rbd_client->client->osdc; |
59c2be1e YS |
1311 | struct ceph_osd_event *event; |
1312 | struct rbd_notify_info info; | |
1313 | int payload_len = sizeof(u32) + sizeof(u32); | |
1314 | int ret; | |
1315 | ||
1316 | ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_NOTIFY, payload_len); | |
1317 | if (ret < 0) | |
1318 | return ret; | |
1319 | ||
1320 | info.dev = dev; | |
1321 | ||
1322 | ret = ceph_osdc_create_event(osdc, rbd_notify_cb, 1, | |
1323 | (void *)&info, &event); | |
1324 | if (ret < 0) | |
1325 | goto fail; | |
1326 | ||
1327 | ops[0].watch.ver = 1; | |
1328 | ops[0].watch.flag = 1; | |
1329 | ops[0].watch.cookie = event->cookie; | |
1330 | ops[0].watch.prot_ver = RADOS_NOTIFY_VER; | |
1331 | ops[0].watch.timeout = 12; | |
1332 | ||
1333 | ret = rbd_req_sync_op(dev, NULL, | |
1334 | CEPH_NOSNAP, | |
1335 | 0, | |
1336 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, | |
1337 | ops, | |
1338 | 1, obj, 0, 0, NULL, NULL, NULL); | |
1339 | if (ret < 0) | |
1340 | goto fail_event; | |
1341 | ||
1342 | ret = ceph_osdc_wait_event(event, CEPH_OSD_TIMEOUT_DEFAULT); | |
1343 | dout("ceph_osdc_wait_event returned %d\n", ret); | |
1344 | rbd_destroy_ops(ops); | |
1345 | return 0; | |
1346 | ||
1347 | fail_event: | |
1348 | ceph_osdc_cancel_event(event); | |
1349 | fail: | |
1350 | rbd_destroy_ops(ops); | |
1351 | return ret; | |
1352 | } | |
1353 | ||
602adf40 YS |
1354 | /* |
1355 | * Request sync osd read | |
1356 | */ | |
1357 | static int rbd_req_sync_exec(struct rbd_device *dev, | |
1358 | const char *obj, | |
1359 | const char *cls, | |
1360 | const char *method, | |
1361 | const char *data, | |
59c2be1e YS |
1362 | int len, |
1363 | u64 *ver) | |
602adf40 YS |
1364 | { |
1365 | struct ceph_osd_req_op *ops; | |
1366 | int cls_len = strlen(cls); | |
1367 | int method_len = strlen(method); | |
1368 | int ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_CALL, | |
1369 | cls_len + method_len + len); | |
1370 | if (ret < 0) | |
1371 | return ret; | |
1372 | ||
1373 | ops[0].cls.class_name = cls; | |
1374 | ops[0].cls.class_len = (__u8)cls_len; | |
1375 | ops[0].cls.method_name = method; | |
1376 | ops[0].cls.method_len = (__u8)method_len; | |
1377 | ops[0].cls.argc = 0; | |
1378 | ops[0].cls.indata = data; | |
1379 | ops[0].cls.indata_len = len; | |
1380 | ||
1381 | ret = rbd_req_sync_op(dev, NULL, | |
1382 | CEPH_NOSNAP, | |
1383 | 0, | |
1384 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, | |
1385 | ops, | |
59c2be1e | 1386 | 1, obj, 0, 0, NULL, NULL, ver); |
602adf40 YS |
1387 | |
1388 | rbd_destroy_ops(ops); | |
1389 | ||
1390 | dout("cls_exec returned %d\n", ret); | |
1391 | return ret; | |
1392 | } | |
1393 | ||
1fec7093 YS |
1394 | static struct rbd_req_coll *rbd_alloc_coll(int num_reqs) |
1395 | { | |
1396 | struct rbd_req_coll *coll = | |
1397 | kzalloc(sizeof(struct rbd_req_coll) + | |
1398 | sizeof(struct rbd_req_status) * num_reqs, | |
1399 | GFP_ATOMIC); | |
1400 | ||
1401 | if (!coll) | |
1402 | return NULL; | |
1403 | coll->total = num_reqs; | |
1404 | kref_init(&coll->kref); | |
1405 | return coll; | |
1406 | } | |
1407 | ||
602adf40 YS |
1408 | /* |
1409 | * block device queue callback | |
1410 | */ | |
1411 | static void rbd_rq_fn(struct request_queue *q) | |
1412 | { | |
1413 | struct rbd_device *rbd_dev = q->queuedata; | |
1414 | struct request *rq; | |
1415 | struct bio_pair *bp = NULL; | |
1416 | ||
1417 | rq = blk_fetch_request(q); | |
1418 | ||
1419 | while (1) { | |
1420 | struct bio *bio; | |
1421 | struct bio *rq_bio, *next_bio = NULL; | |
1422 | bool do_write; | |
1423 | int size, op_size = 0; | |
1424 | u64 ofs; | |
1fec7093 YS |
1425 | int num_segs, cur_seg = 0; |
1426 | struct rbd_req_coll *coll; | |
602adf40 YS |
1427 | |
1428 | /* peek at request from block layer */ | |
1429 | if (!rq) | |
1430 | break; | |
1431 | ||
1432 | dout("fetched request\n"); | |
1433 | ||
1434 | /* filter out block requests we don't understand */ | |
1435 | if ((rq->cmd_type != REQ_TYPE_FS)) { | |
1436 | __blk_end_request_all(rq, 0); | |
1437 | goto next; | |
1438 | } | |
1439 | ||
1440 | /* deduce our operation (read, write) */ | |
1441 | do_write = (rq_data_dir(rq) == WRITE); | |
1442 | ||
1443 | size = blk_rq_bytes(rq); | |
1444 | ofs = blk_rq_pos(rq) * 512ULL; | |
1445 | rq_bio = rq->bio; | |
1446 | if (do_write && rbd_dev->read_only) { | |
1447 | __blk_end_request_all(rq, -EROFS); | |
1448 | goto next; | |
1449 | } | |
1450 | ||
1451 | spin_unlock_irq(q->queue_lock); | |
1452 | ||
1453 | dout("%s 0x%x bytes at 0x%llx\n", | |
1454 | do_write ? "write" : "read", | |
1455 | size, blk_rq_pos(rq) * 512ULL); | |
1456 | ||
1fec7093 YS |
1457 | num_segs = rbd_get_num_segments(&rbd_dev->header, ofs, size); |
1458 | coll = rbd_alloc_coll(num_segs); | |
1459 | if (!coll) { | |
1460 | spin_lock_irq(q->queue_lock); | |
1461 | __blk_end_request_all(rq, -ENOMEM); | |
1462 | goto next; | |
1463 | } | |
1464 | ||
602adf40 YS |
1465 | do { |
1466 | /* a bio clone to be passed down to OSD req */ | |
1467 | dout("rq->bio->bi_vcnt=%d\n", rq->bio->bi_vcnt); | |
1468 | op_size = rbd_get_segment(&rbd_dev->header, | |
1469 | rbd_dev->header.block_name, | |
1470 | ofs, size, | |
1471 | NULL, NULL); | |
1fec7093 | 1472 | kref_get(&coll->kref); |
602adf40 YS |
1473 | bio = bio_chain_clone(&rq_bio, &next_bio, &bp, |
1474 | op_size, GFP_ATOMIC); | |
1475 | if (!bio) { | |
1fec7093 YS |
1476 | rbd_coll_end_req_index(rq, coll, cur_seg, |
1477 | -ENOMEM, op_size); | |
1478 | goto next_seg; | |
602adf40 YS |
1479 | } |
1480 | ||
1fec7093 | 1481 | |
602adf40 YS |
1482 | /* init OSD command: write or read */ |
1483 | if (do_write) | |
1484 | rbd_req_write(rq, rbd_dev, | |
1485 | rbd_dev->header.snapc, | |
1486 | ofs, | |
1fec7093 YS |
1487 | op_size, bio, |
1488 | coll, cur_seg); | |
602adf40 YS |
1489 | else |
1490 | rbd_req_read(rq, rbd_dev, | |
1491 | cur_snap_id(rbd_dev), | |
1492 | ofs, | |
1fec7093 YS |
1493 | op_size, bio, |
1494 | coll, cur_seg); | |
602adf40 | 1495 | |
1fec7093 | 1496 | next_seg: |
602adf40 YS |
1497 | size -= op_size; |
1498 | ofs += op_size; | |
1499 | ||
1fec7093 | 1500 | cur_seg++; |
602adf40 YS |
1501 | rq_bio = next_bio; |
1502 | } while (size > 0); | |
1fec7093 | 1503 | kref_put(&coll->kref, rbd_coll_release); |
602adf40 YS |
1504 | |
1505 | if (bp) | |
1506 | bio_pair_release(bp); | |
602adf40 YS |
1507 | spin_lock_irq(q->queue_lock); |
1508 | next: | |
1509 | rq = blk_fetch_request(q); | |
1510 | } | |
1511 | } | |
1512 | ||
1513 | /* | |
1514 | * a queue callback. Makes sure that we don't create a bio that spans across | |
1515 | * multiple osd objects. One exception would be with a single page bios, | |
1516 | * which we handle later at bio_chain_clone | |
1517 | */ | |
1518 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, | |
1519 | struct bio_vec *bvec) | |
1520 | { | |
1521 | struct rbd_device *rbd_dev = q->queuedata; | |
1522 | unsigned int chunk_sectors = 1 << (rbd_dev->header.obj_order - 9); | |
1523 | sector_t sector = bmd->bi_sector + get_start_sect(bmd->bi_bdev); | |
1524 | unsigned int bio_sectors = bmd->bi_size >> 9; | |
1525 | int max; | |
1526 | ||
1527 | max = (chunk_sectors - ((sector & (chunk_sectors - 1)) | |
1528 | + bio_sectors)) << 9; | |
1529 | if (max < 0) | |
1530 | max = 0; /* bio_add cannot handle a negative return */ | |
1531 | if (max <= bvec->bv_len && bio_sectors == 0) | |
1532 | return bvec->bv_len; | |
1533 | return max; | |
1534 | } | |
1535 | ||
1536 | static void rbd_free_disk(struct rbd_device *rbd_dev) | |
1537 | { | |
1538 | struct gendisk *disk = rbd_dev->disk; | |
1539 | ||
1540 | if (!disk) | |
1541 | return; | |
1542 | ||
1543 | rbd_header_free(&rbd_dev->header); | |
1544 | ||
1545 | if (disk->flags & GENHD_FL_UP) | |
1546 | del_gendisk(disk); | |
1547 | if (disk->queue) | |
1548 | blk_cleanup_queue(disk->queue); | |
1549 | put_disk(disk); | |
1550 | } | |
1551 | ||
1552 | /* | |
1553 | * reload the ondisk the header | |
1554 | */ | |
1555 | static int rbd_read_header(struct rbd_device *rbd_dev, | |
1556 | struct rbd_image_header *header) | |
1557 | { | |
1558 | ssize_t rc; | |
1559 | struct rbd_image_header_ondisk *dh; | |
1560 | int snap_count = 0; | |
1561 | u64 snap_names_len = 0; | |
59c2be1e | 1562 | u64 ver; |
602adf40 YS |
1563 | |
1564 | while (1) { | |
1565 | int len = sizeof(*dh) + | |
1566 | snap_count * sizeof(struct rbd_image_snap_ondisk) + | |
1567 | snap_names_len; | |
1568 | ||
1569 | rc = -ENOMEM; | |
1570 | dh = kmalloc(len, GFP_KERNEL); | |
1571 | if (!dh) | |
1572 | return -ENOMEM; | |
1573 | ||
1574 | rc = rbd_req_sync_read(rbd_dev, | |
1575 | NULL, CEPH_NOSNAP, | |
1576 | rbd_dev->obj_md_name, | |
1577 | 0, len, | |
59c2be1e | 1578 | (char *)dh, &ver); |
602adf40 YS |
1579 | if (rc < 0) |
1580 | goto out_dh; | |
1581 | ||
1582 | rc = rbd_header_from_disk(header, dh, snap_count, GFP_KERNEL); | |
81e759fb JD |
1583 | if (rc < 0) { |
1584 | if (rc == -ENXIO) { | |
1585 | pr_warning("unrecognized header format" | |
1586 | " for image %s", rbd_dev->obj); | |
1587 | } | |
602adf40 | 1588 | goto out_dh; |
81e759fb | 1589 | } |
602adf40 YS |
1590 | |
1591 | if (snap_count != header->total_snaps) { | |
1592 | snap_count = header->total_snaps; | |
1593 | snap_names_len = header->snap_names_len; | |
1594 | rbd_header_free(header); | |
1595 | kfree(dh); | |
1596 | continue; | |
1597 | } | |
1598 | break; | |
1599 | } | |
59c2be1e | 1600 | header->obj_version = ver; |
602adf40 YS |
1601 | |
1602 | out_dh: | |
1603 | kfree(dh); | |
1604 | return rc; | |
1605 | } | |
1606 | ||
1607 | /* | |
1608 | * create a snapshot | |
1609 | */ | |
1610 | static int rbd_header_add_snap(struct rbd_device *dev, | |
1611 | const char *snap_name, | |
1612 | gfp_t gfp_flags) | |
1613 | { | |
1614 | int name_len = strlen(snap_name); | |
1615 | u64 new_snapid; | |
1616 | int ret; | |
916d4d67 | 1617 | void *data, *p, *e; |
59c2be1e | 1618 | u64 ver; |
1dbb4399 | 1619 | struct ceph_mon_client *monc; |
602adf40 YS |
1620 | |
1621 | /* we should create a snapshot only if we're pointing at the head */ | |
1622 | if (dev->cur_snap) | |
1623 | return -EINVAL; | |
1624 | ||
1dbb4399 AE |
1625 | monc = &dev->rbd_client->client->monc; |
1626 | ret = ceph_monc_create_snapid(monc, dev->poolid, &new_snapid); | |
602adf40 YS |
1627 | dout("created snapid=%lld\n", new_snapid); |
1628 | if (ret < 0) | |
1629 | return ret; | |
1630 | ||
1631 | data = kmalloc(name_len + 16, gfp_flags); | |
1632 | if (!data) | |
1633 | return -ENOMEM; | |
1634 | ||
916d4d67 SW |
1635 | p = data; |
1636 | e = data + name_len + 16; | |
602adf40 | 1637 | |
916d4d67 SW |
1638 | ceph_encode_string_safe(&p, e, snap_name, name_len, bad); |
1639 | ceph_encode_64_safe(&p, e, new_snapid, bad); | |
602adf40 YS |
1640 | |
1641 | ret = rbd_req_sync_exec(dev, dev->obj_md_name, "rbd", "snap_add", | |
916d4d67 | 1642 | data, p - data, &ver); |
602adf40 | 1643 | |
916d4d67 | 1644 | kfree(data); |
602adf40 YS |
1645 | |
1646 | if (ret < 0) | |
1647 | return ret; | |
1648 | ||
1649 | dev->header.snapc->seq = new_snapid; | |
1650 | ||
1651 | return 0; | |
1652 | bad: | |
1653 | return -ERANGE; | |
1654 | } | |
1655 | ||
dfc5606d YS |
1656 | static void __rbd_remove_all_snaps(struct rbd_device *rbd_dev) |
1657 | { | |
1658 | struct rbd_snap *snap; | |
1659 | ||
1660 | while (!list_empty(&rbd_dev->snaps)) { | |
1661 | snap = list_first_entry(&rbd_dev->snaps, struct rbd_snap, node); | |
1662 | __rbd_remove_snap_dev(rbd_dev, snap); | |
1663 | } | |
1664 | } | |
1665 | ||
602adf40 YS |
1666 | /* |
1667 | * only read the first part of the ondisk header, without the snaps info | |
1668 | */ | |
dfc5606d | 1669 | static int __rbd_update_snaps(struct rbd_device *rbd_dev) |
602adf40 YS |
1670 | { |
1671 | int ret; | |
1672 | struct rbd_image_header h; | |
1673 | u64 snap_seq; | |
59c2be1e | 1674 | int follow_seq = 0; |
602adf40 YS |
1675 | |
1676 | ret = rbd_read_header(rbd_dev, &h); | |
1677 | if (ret < 0) | |
1678 | return ret; | |
1679 | ||
9db4b3e3 SW |
1680 | /* resized? */ |
1681 | set_capacity(rbd_dev->disk, h.image_size / 512ULL); | |
1682 | ||
602adf40 YS |
1683 | down_write(&rbd_dev->header.snap_rwsem); |
1684 | ||
1685 | snap_seq = rbd_dev->header.snapc->seq; | |
59c2be1e YS |
1686 | if (rbd_dev->header.total_snaps && |
1687 | rbd_dev->header.snapc->snaps[0] == snap_seq) | |
1688 | /* pointing at the head, will need to follow that | |
1689 | if head moves */ | |
1690 | follow_seq = 1; | |
602adf40 YS |
1691 | |
1692 | kfree(rbd_dev->header.snapc); | |
1693 | kfree(rbd_dev->header.snap_names); | |
1694 | kfree(rbd_dev->header.snap_sizes); | |
1695 | ||
1696 | rbd_dev->header.total_snaps = h.total_snaps; | |
1697 | rbd_dev->header.snapc = h.snapc; | |
1698 | rbd_dev->header.snap_names = h.snap_names; | |
dfc5606d | 1699 | rbd_dev->header.snap_names_len = h.snap_names_len; |
602adf40 | 1700 | rbd_dev->header.snap_sizes = h.snap_sizes; |
59c2be1e YS |
1701 | if (follow_seq) |
1702 | rbd_dev->header.snapc->seq = rbd_dev->header.snapc->snaps[0]; | |
1703 | else | |
1704 | rbd_dev->header.snapc->seq = snap_seq; | |
602adf40 | 1705 | |
dfc5606d YS |
1706 | ret = __rbd_init_snaps_header(rbd_dev); |
1707 | ||
602adf40 YS |
1708 | up_write(&rbd_dev->header.snap_rwsem); |
1709 | ||
dfc5606d | 1710 | return ret; |
602adf40 YS |
1711 | } |
1712 | ||
1713 | static int rbd_init_disk(struct rbd_device *rbd_dev) | |
1714 | { | |
1715 | struct gendisk *disk; | |
1716 | struct request_queue *q; | |
1717 | int rc; | |
1718 | u64 total_size = 0; | |
1719 | ||
1720 | /* contact OSD, request size info about the object being mapped */ | |
1721 | rc = rbd_read_header(rbd_dev, &rbd_dev->header); | |
1722 | if (rc) | |
1723 | return rc; | |
1724 | ||
dfc5606d YS |
1725 | /* no need to lock here, as rbd_dev is not registered yet */ |
1726 | rc = __rbd_init_snaps_header(rbd_dev); | |
1727 | if (rc) | |
1728 | return rc; | |
1729 | ||
cc9d734c | 1730 | rc = rbd_header_set_snap(rbd_dev, &total_size); |
602adf40 YS |
1731 | if (rc) |
1732 | return rc; | |
1733 | ||
1734 | /* create gendisk info */ | |
1735 | rc = -ENOMEM; | |
1736 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); | |
1737 | if (!disk) | |
1738 | goto out; | |
1739 | ||
aedfec59 SW |
1740 | snprintf(disk->disk_name, sizeof(disk->disk_name), DRV_NAME "%d", |
1741 | rbd_dev->id); | |
602adf40 YS |
1742 | disk->major = rbd_dev->major; |
1743 | disk->first_minor = 0; | |
1744 | disk->fops = &rbd_bd_ops; | |
1745 | disk->private_data = rbd_dev; | |
1746 | ||
1747 | /* init rq */ | |
1748 | rc = -ENOMEM; | |
1749 | q = blk_init_queue(rbd_rq_fn, &rbd_dev->lock); | |
1750 | if (!q) | |
1751 | goto out_disk; | |
029bcbd8 JD |
1752 | |
1753 | /* set io sizes to object size */ | |
1754 | blk_queue_max_hw_sectors(q, rbd_obj_bytes(&rbd_dev->header) / 512ULL); | |
1755 | blk_queue_max_segment_size(q, rbd_obj_bytes(&rbd_dev->header)); | |
1756 | blk_queue_io_min(q, rbd_obj_bytes(&rbd_dev->header)); | |
1757 | blk_queue_io_opt(q, rbd_obj_bytes(&rbd_dev->header)); | |
1758 | ||
602adf40 YS |
1759 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
1760 | disk->queue = q; | |
1761 | ||
1762 | q->queuedata = rbd_dev; | |
1763 | ||
1764 | rbd_dev->disk = disk; | |
1765 | rbd_dev->q = q; | |
1766 | ||
1767 | /* finally, announce the disk to the world */ | |
1768 | set_capacity(disk, total_size / 512ULL); | |
1769 | add_disk(disk); | |
1770 | ||
1771 | pr_info("%s: added with size 0x%llx\n", | |
1772 | disk->disk_name, (unsigned long long)total_size); | |
1773 | return 0; | |
1774 | ||
1775 | out_disk: | |
1776 | put_disk(disk); | |
1777 | out: | |
1778 | return rc; | |
1779 | } | |
1780 | ||
dfc5606d YS |
1781 | /* |
1782 | sysfs | |
1783 | */ | |
1784 | ||
1785 | static ssize_t rbd_size_show(struct device *dev, | |
1786 | struct device_attribute *attr, char *buf) | |
1787 | { | |
1788 | struct rbd_device *rbd_dev = dev_to_rbd(dev); | |
1789 | ||
1790 | return sprintf(buf, "%llu\n", (unsigned long long)rbd_dev->header.image_size); | |
1791 | } | |
1792 | ||
1793 | static ssize_t rbd_major_show(struct device *dev, | |
1794 | struct device_attribute *attr, char *buf) | |
1795 | { | |
1796 | struct rbd_device *rbd_dev = dev_to_rbd(dev); | |
602adf40 | 1797 | |
dfc5606d YS |
1798 | return sprintf(buf, "%d\n", rbd_dev->major); |
1799 | } | |
1800 | ||
1801 | static ssize_t rbd_client_id_show(struct device *dev, | |
1802 | struct device_attribute *attr, char *buf) | |
602adf40 | 1803 | { |
dfc5606d YS |
1804 | struct rbd_device *rbd_dev = dev_to_rbd(dev); |
1805 | ||
1dbb4399 AE |
1806 | return sprintf(buf, "client%lld\n", |
1807 | ceph_client_id(rbd_dev->rbd_client->client)); | |
602adf40 YS |
1808 | } |
1809 | ||
dfc5606d YS |
1810 | static ssize_t rbd_pool_show(struct device *dev, |
1811 | struct device_attribute *attr, char *buf) | |
602adf40 | 1812 | { |
dfc5606d YS |
1813 | struct rbd_device *rbd_dev = dev_to_rbd(dev); |
1814 | ||
1815 | return sprintf(buf, "%s\n", rbd_dev->pool_name); | |
1816 | } | |
1817 | ||
1818 | static ssize_t rbd_name_show(struct device *dev, | |
1819 | struct device_attribute *attr, char *buf) | |
1820 | { | |
1821 | struct rbd_device *rbd_dev = dev_to_rbd(dev); | |
1822 | ||
1823 | return sprintf(buf, "%s\n", rbd_dev->obj); | |
1824 | } | |
1825 | ||
1826 | static ssize_t rbd_snap_show(struct device *dev, | |
1827 | struct device_attribute *attr, | |
1828 | char *buf) | |
1829 | { | |
1830 | struct rbd_device *rbd_dev = dev_to_rbd(dev); | |
1831 | ||
1832 | return sprintf(buf, "%s\n", rbd_dev->snap_name); | |
1833 | } | |
1834 | ||
1835 | static ssize_t rbd_image_refresh(struct device *dev, | |
1836 | struct device_attribute *attr, | |
1837 | const char *buf, | |
1838 | size_t size) | |
1839 | { | |
1840 | struct rbd_device *rbd_dev = dev_to_rbd(dev); | |
1841 | int rc; | |
1842 | int ret = size; | |
602adf40 YS |
1843 | |
1844 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
1845 | ||
dfc5606d YS |
1846 | rc = __rbd_update_snaps(rbd_dev); |
1847 | if (rc < 0) | |
1848 | ret = rc; | |
602adf40 | 1849 | |
dfc5606d YS |
1850 | mutex_unlock(&ctl_mutex); |
1851 | return ret; | |
1852 | } | |
602adf40 | 1853 | |
dfc5606d YS |
1854 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
1855 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); | |
1856 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); | |
1857 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); | |
1858 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); | |
1859 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); | |
1860 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); | |
1861 | static DEVICE_ATTR(create_snap, S_IWUSR, NULL, rbd_snap_add); | |
dfc5606d YS |
1862 | |
1863 | static struct attribute *rbd_attrs[] = { | |
1864 | &dev_attr_size.attr, | |
1865 | &dev_attr_major.attr, | |
1866 | &dev_attr_client_id.attr, | |
1867 | &dev_attr_pool.attr, | |
1868 | &dev_attr_name.attr, | |
1869 | &dev_attr_current_snap.attr, | |
1870 | &dev_attr_refresh.attr, | |
1871 | &dev_attr_create_snap.attr, | |
dfc5606d YS |
1872 | NULL |
1873 | }; | |
1874 | ||
1875 | static struct attribute_group rbd_attr_group = { | |
1876 | .attrs = rbd_attrs, | |
1877 | }; | |
1878 | ||
1879 | static const struct attribute_group *rbd_attr_groups[] = { | |
1880 | &rbd_attr_group, | |
1881 | NULL | |
1882 | }; | |
1883 | ||
1884 | static void rbd_sysfs_dev_release(struct device *dev) | |
1885 | { | |
1886 | } | |
1887 | ||
1888 | static struct device_type rbd_device_type = { | |
1889 | .name = "rbd", | |
1890 | .groups = rbd_attr_groups, | |
1891 | .release = rbd_sysfs_dev_release, | |
1892 | }; | |
1893 | ||
1894 | ||
1895 | /* | |
1896 | sysfs - snapshots | |
1897 | */ | |
1898 | ||
1899 | static ssize_t rbd_snap_size_show(struct device *dev, | |
1900 | struct device_attribute *attr, | |
1901 | char *buf) | |
1902 | { | |
1903 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
1904 | ||
1905 | return sprintf(buf, "%lld\n", (long long)snap->size); | |
1906 | } | |
1907 | ||
1908 | static ssize_t rbd_snap_id_show(struct device *dev, | |
1909 | struct device_attribute *attr, | |
1910 | char *buf) | |
1911 | { | |
1912 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
1913 | ||
1914 | return sprintf(buf, "%lld\n", (long long)snap->id); | |
1915 | } | |
1916 | ||
1917 | static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); | |
1918 | static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); | |
1919 | ||
1920 | static struct attribute *rbd_snap_attrs[] = { | |
1921 | &dev_attr_snap_size.attr, | |
1922 | &dev_attr_snap_id.attr, | |
1923 | NULL, | |
1924 | }; | |
1925 | ||
1926 | static struct attribute_group rbd_snap_attr_group = { | |
1927 | .attrs = rbd_snap_attrs, | |
1928 | }; | |
1929 | ||
1930 | static void rbd_snap_dev_release(struct device *dev) | |
1931 | { | |
1932 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
1933 | kfree(snap->name); | |
1934 | kfree(snap); | |
1935 | } | |
1936 | ||
1937 | static const struct attribute_group *rbd_snap_attr_groups[] = { | |
1938 | &rbd_snap_attr_group, | |
1939 | NULL | |
1940 | }; | |
1941 | ||
1942 | static struct device_type rbd_snap_device_type = { | |
1943 | .groups = rbd_snap_attr_groups, | |
1944 | .release = rbd_snap_dev_release, | |
1945 | }; | |
1946 | ||
1947 | static void __rbd_remove_snap_dev(struct rbd_device *rbd_dev, | |
1948 | struct rbd_snap *snap) | |
1949 | { | |
1950 | list_del(&snap->node); | |
1951 | device_unregister(&snap->dev); | |
1952 | } | |
1953 | ||
1954 | static int rbd_register_snap_dev(struct rbd_device *rbd_dev, | |
1955 | struct rbd_snap *snap, | |
1956 | struct device *parent) | |
1957 | { | |
1958 | struct device *dev = &snap->dev; | |
1959 | int ret; | |
1960 | ||
1961 | dev->type = &rbd_snap_device_type; | |
1962 | dev->parent = parent; | |
1963 | dev->release = rbd_snap_dev_release; | |
1964 | dev_set_name(dev, "snap_%s", snap->name); | |
1965 | ret = device_register(dev); | |
1966 | ||
1967 | return ret; | |
1968 | } | |
1969 | ||
1970 | static int __rbd_add_snap_dev(struct rbd_device *rbd_dev, | |
1971 | int i, const char *name, | |
1972 | struct rbd_snap **snapp) | |
1973 | { | |
1974 | int ret; | |
1975 | struct rbd_snap *snap = kzalloc(sizeof(*snap), GFP_KERNEL); | |
1976 | if (!snap) | |
1977 | return -ENOMEM; | |
1978 | snap->name = kstrdup(name, GFP_KERNEL); | |
1979 | snap->size = rbd_dev->header.snap_sizes[i]; | |
1980 | snap->id = rbd_dev->header.snapc->snaps[i]; | |
1981 | if (device_is_registered(&rbd_dev->dev)) { | |
1982 | ret = rbd_register_snap_dev(rbd_dev, snap, | |
1983 | &rbd_dev->dev); | |
1984 | if (ret < 0) | |
1985 | goto err; | |
1986 | } | |
1987 | *snapp = snap; | |
1988 | return 0; | |
1989 | err: | |
1990 | kfree(snap->name); | |
1991 | kfree(snap); | |
1992 | return ret; | |
1993 | } | |
1994 | ||
1995 | /* | |
1996 | * search for the previous snap in a null delimited string list | |
1997 | */ | |
1998 | const char *rbd_prev_snap_name(const char *name, const char *start) | |
1999 | { | |
2000 | if (name < start + 2) | |
2001 | return NULL; | |
2002 | ||
2003 | name -= 2; | |
2004 | while (*name) { | |
2005 | if (name == start) | |
2006 | return start; | |
2007 | name--; | |
2008 | } | |
2009 | return name + 1; | |
2010 | } | |
2011 | ||
2012 | /* | |
2013 | * compare the old list of snapshots that we have to what's in the header | |
2014 | * and update it accordingly. Note that the header holds the snapshots | |
2015 | * in a reverse order (from newest to oldest) and we need to go from | |
2016 | * older to new so that we don't get a duplicate snap name when | |
2017 | * doing the process (e.g., removed snapshot and recreated a new | |
2018 | * one with the same name. | |
2019 | */ | |
2020 | static int __rbd_init_snaps_header(struct rbd_device *rbd_dev) | |
2021 | { | |
2022 | const char *name, *first_name; | |
2023 | int i = rbd_dev->header.total_snaps; | |
2024 | struct rbd_snap *snap, *old_snap = NULL; | |
2025 | int ret; | |
2026 | struct list_head *p, *n; | |
2027 | ||
2028 | first_name = rbd_dev->header.snap_names; | |
2029 | name = first_name + rbd_dev->header.snap_names_len; | |
2030 | ||
2031 | list_for_each_prev_safe(p, n, &rbd_dev->snaps) { | |
2032 | u64 cur_id; | |
2033 | ||
2034 | old_snap = list_entry(p, struct rbd_snap, node); | |
2035 | ||
2036 | if (i) | |
2037 | cur_id = rbd_dev->header.snapc->snaps[i - 1]; | |
2038 | ||
2039 | if (!i || old_snap->id < cur_id) { | |
2040 | /* old_snap->id was skipped, thus was removed */ | |
2041 | __rbd_remove_snap_dev(rbd_dev, old_snap); | |
2042 | continue; | |
2043 | } | |
2044 | if (old_snap->id == cur_id) { | |
2045 | /* we have this snapshot already */ | |
2046 | i--; | |
2047 | name = rbd_prev_snap_name(name, first_name); | |
2048 | continue; | |
2049 | } | |
2050 | for (; i > 0; | |
2051 | i--, name = rbd_prev_snap_name(name, first_name)) { | |
2052 | if (!name) { | |
2053 | WARN_ON(1); | |
2054 | return -EINVAL; | |
2055 | } | |
2056 | cur_id = rbd_dev->header.snapc->snaps[i]; | |
2057 | /* snapshot removal? handle it above */ | |
2058 | if (cur_id >= old_snap->id) | |
2059 | break; | |
2060 | /* a new snapshot */ | |
2061 | ret = __rbd_add_snap_dev(rbd_dev, i - 1, name, &snap); | |
2062 | if (ret < 0) | |
2063 | return ret; | |
2064 | ||
2065 | /* note that we add it backward so using n and not p */ | |
2066 | list_add(&snap->node, n); | |
2067 | p = &snap->node; | |
2068 | } | |
2069 | } | |
2070 | /* we're done going over the old snap list, just add what's left */ | |
2071 | for (; i > 0; i--) { | |
2072 | name = rbd_prev_snap_name(name, first_name); | |
2073 | if (!name) { | |
2074 | WARN_ON(1); | |
2075 | return -EINVAL; | |
2076 | } | |
2077 | ret = __rbd_add_snap_dev(rbd_dev, i - 1, name, &snap); | |
2078 | if (ret < 0) | |
2079 | return ret; | |
2080 | list_add(&snap->node, &rbd_dev->snaps); | |
2081 | } | |
2082 | ||
2083 | return 0; | |
2084 | } | |
2085 | ||
2086 | ||
2087 | static void rbd_root_dev_release(struct device *dev) | |
2088 | { | |
2089 | } | |
2090 | ||
2091 | static struct device rbd_root_dev = { | |
2092 | .init_name = "rbd", | |
2093 | .release = rbd_root_dev_release, | |
2094 | }; | |
2095 | ||
2096 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) | |
2097 | { | |
2098 | int ret = -ENOMEM; | |
2099 | struct device *dev; | |
2100 | struct rbd_snap *snap; | |
2101 | ||
2102 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2103 | dev = &rbd_dev->dev; | |
2104 | ||
2105 | dev->bus = &rbd_bus_type; | |
2106 | dev->type = &rbd_device_type; | |
2107 | dev->parent = &rbd_root_dev; | |
2108 | dev->release = rbd_dev_release; | |
2109 | dev_set_name(dev, "%d", rbd_dev->id); | |
2110 | ret = device_register(dev); | |
2111 | if (ret < 0) | |
2112 | goto done_free; | |
2113 | ||
2114 | list_for_each_entry(snap, &rbd_dev->snaps, node) { | |
2115 | ret = rbd_register_snap_dev(rbd_dev, snap, | |
2116 | &rbd_dev->dev); | |
2117 | if (ret < 0) | |
602adf40 YS |
2118 | break; |
2119 | } | |
2120 | ||
2121 | mutex_unlock(&ctl_mutex); | |
dfc5606d YS |
2122 | return 0; |
2123 | done_free: | |
2124 | mutex_unlock(&ctl_mutex); | |
2125 | return ret; | |
602adf40 YS |
2126 | } |
2127 | ||
dfc5606d YS |
2128 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
2129 | { | |
2130 | device_unregister(&rbd_dev->dev); | |
2131 | } | |
2132 | ||
59c2be1e YS |
2133 | static int rbd_init_watch_dev(struct rbd_device *rbd_dev) |
2134 | { | |
2135 | int ret, rc; | |
2136 | ||
2137 | do { | |
2138 | ret = rbd_req_sync_watch(rbd_dev, rbd_dev->obj_md_name, | |
2139 | rbd_dev->header.obj_version); | |
2140 | if (ret == -ERANGE) { | |
2141 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2142 | rc = __rbd_update_snaps(rbd_dev); | |
2143 | mutex_unlock(&ctl_mutex); | |
2144 | if (rc < 0) | |
2145 | return rc; | |
2146 | } | |
2147 | } while (ret == -ERANGE); | |
2148 | ||
2149 | return ret; | |
2150 | } | |
2151 | ||
1ddbe94e AE |
2152 | static atomic64_t rbd_id_max = ATOMIC64_INIT(0); |
2153 | ||
2154 | /* | |
2155 | * Get a unique rbd identifier. The minimum rbd id is 1. | |
2156 | */ | |
b7f23c36 AE |
2157 | static int rbd_id_get(void) |
2158 | { | |
1ddbe94e AE |
2159 | return atomic64_inc_return(&rbd_id_max); |
2160 | } | |
b7f23c36 | 2161 | |
1ddbe94e AE |
2162 | /* |
2163 | * Record that an rbd identifier is no longer in use. | |
2164 | */ | |
2165 | static void rbd_id_put(int rbd_id) | |
2166 | { | |
2167 | BUG_ON(rbd_id < 1); | |
b7f23c36 | 2168 | |
1ddbe94e AE |
2169 | /* |
2170 | * New id's are always one more than the current maximum. | |
2171 | * If the id being "put" *is* that maximum, decrement the | |
2172 | * maximum so the next one requested just reuses this one. | |
2173 | */ | |
2174 | atomic64_cmpxchg(&rbd_id_max, rbd_id, rbd_id - 1); | |
b7f23c36 AE |
2175 | } |
2176 | ||
59c2be1e YS |
2177 | static ssize_t rbd_add(struct bus_type *bus, |
2178 | const char *buf, | |
2179 | size_t count) | |
602adf40 YS |
2180 | { |
2181 | struct ceph_osd_client *osdc; | |
2182 | struct rbd_device *rbd_dev; | |
2183 | ssize_t rc = -ENOMEM; | |
b7f23c36 | 2184 | int irc; |
602adf40 YS |
2185 | char *mon_dev_name; |
2186 | char *options; | |
2187 | ||
2188 | if (!try_module_get(THIS_MODULE)) | |
2189 | return -ENODEV; | |
2190 | ||
2191 | mon_dev_name = kmalloc(RBD_MAX_OPT_LEN, GFP_KERNEL); | |
2192 | if (!mon_dev_name) | |
2193 | goto err_out_mod; | |
2194 | ||
2195 | options = kmalloc(RBD_MAX_OPT_LEN, GFP_KERNEL); | |
2196 | if (!options) | |
2197 | goto err_mon_dev; | |
2198 | ||
2199 | /* new rbd_device object */ | |
2200 | rbd_dev = kzalloc(sizeof(*rbd_dev), GFP_KERNEL); | |
2201 | if (!rbd_dev) | |
2202 | goto err_out_opt; | |
2203 | ||
2204 | /* static rbd_device initialization */ | |
2205 | spin_lock_init(&rbd_dev->lock); | |
2206 | INIT_LIST_HEAD(&rbd_dev->node); | |
dfc5606d | 2207 | INIT_LIST_HEAD(&rbd_dev->snaps); |
602adf40 | 2208 | |
0e805a1d AE |
2209 | init_rwsem(&rbd_dev->header.snap_rwsem); |
2210 | ||
1ddbe94e | 2211 | /* generate unique id: one more than highest used so far */ |
602adf40 YS |
2212 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
2213 | ||
b7f23c36 | 2214 | rbd_dev->id = rbd_id_get(); |
602adf40 YS |
2215 | |
2216 | /* add to global list */ | |
2217 | list_add_tail(&rbd_dev->node, &rbd_dev_list); | |
2218 | ||
2219 | /* parse add command */ | |
2220 | if (sscanf(buf, "%" __stringify(RBD_MAX_OPT_LEN) "s " | |
2221 | "%" __stringify(RBD_MAX_OPT_LEN) "s " | |
2222 | "%" __stringify(RBD_MAX_POOL_NAME_LEN) "s " | |
2223 | "%" __stringify(RBD_MAX_OBJ_NAME_LEN) "s" | |
2224 | "%" __stringify(RBD_MAX_SNAP_NAME_LEN) "s", | |
2225 | mon_dev_name, options, rbd_dev->pool_name, | |
2226 | rbd_dev->obj, rbd_dev->snap_name) < 4) { | |
2227 | rc = -EINVAL; | |
2228 | goto err_out_slot; | |
2229 | } | |
2230 | ||
2231 | if (rbd_dev->snap_name[0] == 0) | |
cc9d734c JD |
2232 | memcpy(rbd_dev->snap_name, RBD_SNAP_HEAD_NAME, |
2233 | sizeof (RBD_SNAP_HEAD_NAME)); | |
602adf40 YS |
2234 | |
2235 | rbd_dev->obj_len = strlen(rbd_dev->obj); | |
2236 | snprintf(rbd_dev->obj_md_name, sizeof(rbd_dev->obj_md_name), "%s%s", | |
2237 | rbd_dev->obj, RBD_SUFFIX); | |
2238 | ||
2239 | /* initialize rest of new object */ | |
2240 | snprintf(rbd_dev->name, DEV_NAME_LEN, DRV_NAME "%d", rbd_dev->id); | |
2241 | rc = rbd_get_client(rbd_dev, mon_dev_name, options); | |
2242 | if (rc < 0) | |
2243 | goto err_out_slot; | |
2244 | ||
2245 | mutex_unlock(&ctl_mutex); | |
2246 | ||
2247 | /* pick the pool */ | |
1dbb4399 | 2248 | osdc = &rbd_dev->rbd_client->client->osdc; |
602adf40 YS |
2249 | rc = ceph_pg_poolid_by_name(osdc->osdmap, rbd_dev->pool_name); |
2250 | if (rc < 0) | |
2251 | goto err_out_client; | |
2252 | rbd_dev->poolid = rc; | |
2253 | ||
2254 | /* register our block device */ | |
2255 | irc = register_blkdev(0, rbd_dev->name); | |
2256 | if (irc < 0) { | |
2257 | rc = irc; | |
2258 | goto err_out_client; | |
2259 | } | |
2260 | rbd_dev->major = irc; | |
2261 | ||
dfc5606d YS |
2262 | rc = rbd_bus_add_dev(rbd_dev); |
2263 | if (rc) | |
766fc439 YS |
2264 | goto err_out_blkdev; |
2265 | ||
602adf40 YS |
2266 | /* set up and announce blkdev mapping */ |
2267 | rc = rbd_init_disk(rbd_dev); | |
2268 | if (rc) | |
766fc439 | 2269 | goto err_out_bus; |
602adf40 | 2270 | |
59c2be1e YS |
2271 | rc = rbd_init_watch_dev(rbd_dev); |
2272 | if (rc) | |
2273 | goto err_out_bus; | |
2274 | ||
602adf40 YS |
2275 | return count; |
2276 | ||
766fc439 YS |
2277 | err_out_bus: |
2278 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2279 | list_del_init(&rbd_dev->node); | |
2280 | mutex_unlock(&ctl_mutex); | |
1ddbe94e | 2281 | rbd_id_put(target_id); |
766fc439 YS |
2282 | |
2283 | /* this will also clean up rest of rbd_dev stuff */ | |
2284 | ||
2285 | rbd_bus_del_dev(rbd_dev); | |
2286 | kfree(options); | |
2287 | kfree(mon_dev_name); | |
2288 | return rc; | |
2289 | ||
602adf40 YS |
2290 | err_out_blkdev: |
2291 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
2292 | err_out_client: | |
2293 | rbd_put_client(rbd_dev); | |
2294 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2295 | err_out_slot: | |
2296 | list_del_init(&rbd_dev->node); | |
2297 | mutex_unlock(&ctl_mutex); | |
1ddbe94e | 2298 | rbd_id_put(target_id); |
602adf40 YS |
2299 | |
2300 | kfree(rbd_dev); | |
2301 | err_out_opt: | |
2302 | kfree(options); | |
2303 | err_mon_dev: | |
2304 | kfree(mon_dev_name); | |
2305 | err_out_mod: | |
2306 | dout("Error adding device %s\n", buf); | |
2307 | module_put(THIS_MODULE); | |
2308 | return rc; | |
2309 | } | |
2310 | ||
2311 | static struct rbd_device *__rbd_get_dev(unsigned long id) | |
2312 | { | |
2313 | struct list_head *tmp; | |
2314 | struct rbd_device *rbd_dev; | |
2315 | ||
2316 | list_for_each(tmp, &rbd_dev_list) { | |
2317 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
2318 | if (rbd_dev->id == id) | |
2319 | return rbd_dev; | |
2320 | } | |
2321 | return NULL; | |
2322 | } | |
2323 | ||
dfc5606d | 2324 | static void rbd_dev_release(struct device *dev) |
602adf40 | 2325 | { |
dfc5606d YS |
2326 | struct rbd_device *rbd_dev = |
2327 | container_of(dev, struct rbd_device, dev); | |
602adf40 | 2328 | |
1dbb4399 AE |
2329 | if (rbd_dev->watch_request) { |
2330 | struct ceph_client *client = rbd_dev->rbd_client->client; | |
2331 | ||
2332 | ceph_osdc_unregister_linger_request(&client->osdc, | |
59c2be1e | 2333 | rbd_dev->watch_request); |
1dbb4399 | 2334 | } |
59c2be1e | 2335 | if (rbd_dev->watch_event) |
79e3057c | 2336 | rbd_req_sync_unwatch(rbd_dev, rbd_dev->obj_md_name); |
59c2be1e | 2337 | |
602adf40 YS |
2338 | rbd_put_client(rbd_dev); |
2339 | ||
2340 | /* clean up and free blkdev */ | |
2341 | rbd_free_disk(rbd_dev); | |
2342 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
2343 | kfree(rbd_dev); | |
2344 | ||
2345 | /* release module ref */ | |
2346 | module_put(THIS_MODULE); | |
602adf40 YS |
2347 | } |
2348 | ||
dfc5606d YS |
2349 | static ssize_t rbd_remove(struct bus_type *bus, |
2350 | const char *buf, | |
2351 | size_t count) | |
602adf40 YS |
2352 | { |
2353 | struct rbd_device *rbd_dev = NULL; | |
2354 | int target_id, rc; | |
2355 | unsigned long ul; | |
2356 | int ret = count; | |
2357 | ||
2358 | rc = strict_strtoul(buf, 10, &ul); | |
2359 | if (rc) | |
2360 | return rc; | |
2361 | ||
2362 | /* convert to int; abort if we lost anything in the conversion */ | |
2363 | target_id = (int) ul; | |
2364 | if (target_id != ul) | |
2365 | return -EINVAL; | |
2366 | ||
2367 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2368 | ||
2369 | rbd_dev = __rbd_get_dev(target_id); | |
2370 | if (!rbd_dev) { | |
2371 | ret = -ENOENT; | |
2372 | goto done; | |
2373 | } | |
2374 | ||
dfc5606d | 2375 | list_del_init(&rbd_dev->node); |
1ddbe94e | 2376 | rbd_id_put(target_id); |
dfc5606d YS |
2377 | |
2378 | __rbd_remove_all_snaps(rbd_dev); | |
2379 | rbd_bus_del_dev(rbd_dev); | |
602adf40 YS |
2380 | |
2381 | done: | |
2382 | mutex_unlock(&ctl_mutex); | |
2383 | return ret; | |
2384 | } | |
2385 | ||
dfc5606d YS |
2386 | static ssize_t rbd_snap_add(struct device *dev, |
2387 | struct device_attribute *attr, | |
2388 | const char *buf, | |
2389 | size_t count) | |
602adf40 | 2390 | { |
dfc5606d YS |
2391 | struct rbd_device *rbd_dev = dev_to_rbd(dev); |
2392 | int ret; | |
2393 | char *name = kmalloc(count + 1, GFP_KERNEL); | |
602adf40 YS |
2394 | if (!name) |
2395 | return -ENOMEM; | |
2396 | ||
dfc5606d | 2397 | snprintf(name, count, "%s", buf); |
602adf40 YS |
2398 | |
2399 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2400 | ||
602adf40 YS |
2401 | ret = rbd_header_add_snap(rbd_dev, |
2402 | name, GFP_KERNEL); | |
2403 | if (ret < 0) | |
59c2be1e | 2404 | goto err_unlock; |
602adf40 | 2405 | |
dfc5606d | 2406 | ret = __rbd_update_snaps(rbd_dev); |
602adf40 | 2407 | if (ret < 0) |
59c2be1e YS |
2408 | goto err_unlock; |
2409 | ||
2410 | /* shouldn't hold ctl_mutex when notifying.. notify might | |
2411 | trigger a watch callback that would need to get that mutex */ | |
2412 | mutex_unlock(&ctl_mutex); | |
2413 | ||
2414 | /* make a best effort, don't error if failed */ | |
2415 | rbd_req_sync_notify(rbd_dev, rbd_dev->obj_md_name); | |
602adf40 YS |
2416 | |
2417 | ret = count; | |
59c2be1e YS |
2418 | kfree(name); |
2419 | return ret; | |
2420 | ||
2421 | err_unlock: | |
602adf40 | 2422 | mutex_unlock(&ctl_mutex); |
602adf40 YS |
2423 | kfree(name); |
2424 | return ret; | |
2425 | } | |
2426 | ||
dfc5606d YS |
2427 | static struct bus_attribute rbd_bus_attrs[] = { |
2428 | __ATTR(add, S_IWUSR, NULL, rbd_add), | |
2429 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), | |
602adf40 YS |
2430 | __ATTR_NULL |
2431 | }; | |
2432 | ||
2433 | /* | |
2434 | * create control files in sysfs | |
dfc5606d | 2435 | * /sys/bus/rbd/... |
602adf40 YS |
2436 | */ |
2437 | static int rbd_sysfs_init(void) | |
2438 | { | |
dfc5606d | 2439 | int ret; |
602adf40 | 2440 | |
dfc5606d | 2441 | rbd_bus_type.bus_attrs = rbd_bus_attrs; |
602adf40 | 2442 | |
dfc5606d | 2443 | ret = bus_register(&rbd_bus_type); |
21079786 | 2444 | if (ret < 0) |
dfc5606d | 2445 | return ret; |
602adf40 | 2446 | |
dfc5606d | 2447 | ret = device_register(&rbd_root_dev); |
602adf40 | 2448 | |
602adf40 YS |
2449 | return ret; |
2450 | } | |
2451 | ||
2452 | static void rbd_sysfs_cleanup(void) | |
2453 | { | |
dfc5606d YS |
2454 | device_unregister(&rbd_root_dev); |
2455 | bus_unregister(&rbd_bus_type); | |
602adf40 YS |
2456 | } |
2457 | ||
2458 | int __init rbd_init(void) | |
2459 | { | |
2460 | int rc; | |
2461 | ||
2462 | rc = rbd_sysfs_init(); | |
2463 | if (rc) | |
2464 | return rc; | |
602adf40 YS |
2465 | pr_info("loaded " DRV_NAME_LONG "\n"); |
2466 | return 0; | |
2467 | } | |
2468 | ||
2469 | void __exit rbd_exit(void) | |
2470 | { | |
2471 | rbd_sysfs_cleanup(); | |
2472 | } | |
2473 | ||
2474 | module_init(rbd_init); | |
2475 | module_exit(rbd_exit); | |
2476 | ||
2477 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
2478 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
2479 | MODULE_DESCRIPTION("rados block device"); | |
2480 | ||
2481 | /* following authorship retained from original osdblk.c */ | |
2482 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); | |
2483 | ||
2484 | MODULE_LICENSE("GPL"); |