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 | ||
aafb230e AE |
44 | #define RBD_DEBUG /* Activate rbd_assert() calls */ |
45 | ||
593a9e7b AE |
46 | /* |
47 | * The basic unit of block I/O is a sector. It is interpreted in a | |
48 | * number of contexts in Linux (blk, bio, genhd), but the default is | |
49 | * universally 512 bytes. These symbols are just slightly more | |
50 | * meaningful than the bare numbers they represent. | |
51 | */ | |
52 | #define SECTOR_SHIFT 9 | |
53 | #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) | |
54 | ||
df111be6 AE |
55 | /* It might be useful to have this defined elsewhere too */ |
56 | ||
57 | #define U64_MAX ((u64) (~0ULL)) | |
58 | ||
f0f8cef5 AE |
59 | #define RBD_DRV_NAME "rbd" |
60 | #define RBD_DRV_NAME_LONG "rbd (rados block device)" | |
602adf40 YS |
61 | |
62 | #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ | |
63 | ||
d4b125e9 AE |
64 | #define RBD_SNAP_DEV_NAME_PREFIX "snap_" |
65 | #define RBD_MAX_SNAP_NAME_LEN \ | |
66 | (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1)) | |
67 | ||
35d489f9 | 68 | #define RBD_MAX_SNAP_COUNT 510 /* allows max snapc to fit in 4KB */ |
602adf40 YS |
69 | #define RBD_MAX_OPT_LEN 1024 |
70 | ||
71 | #define RBD_SNAP_HEAD_NAME "-" | |
72 | ||
1e130199 AE |
73 | #define RBD_IMAGE_ID_LEN_MAX 64 |
74 | #define RBD_OBJ_PREFIX_LEN_MAX 64 | |
589d30e0 | 75 | |
d889140c AE |
76 | /* Feature bits */ |
77 | ||
78 | #define RBD_FEATURE_LAYERING 1 | |
79 | ||
80 | /* Features supported by this (client software) implementation. */ | |
81 | ||
82 | #define RBD_FEATURES_ALL (0) | |
83 | ||
81a89793 AE |
84 | /* |
85 | * An RBD device name will be "rbd#", where the "rbd" comes from | |
86 | * RBD_DRV_NAME above, and # is a unique integer identifier. | |
87 | * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big | |
88 | * enough to hold all possible device names. | |
89 | */ | |
602adf40 | 90 | #define DEV_NAME_LEN 32 |
81a89793 | 91 | #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) |
602adf40 | 92 | |
cc0538b6 | 93 | #define RBD_READ_ONLY_DEFAULT false |
59c2be1e | 94 | |
602adf40 YS |
95 | /* |
96 | * block device image metadata (in-memory version) | |
97 | */ | |
98 | struct rbd_image_header { | |
f84344f3 | 99 | /* These four fields never change for a given rbd image */ |
849b4260 | 100 | char *object_prefix; |
34b13184 | 101 | u64 features; |
602adf40 YS |
102 | __u8 obj_order; |
103 | __u8 crypt_type; | |
104 | __u8 comp_type; | |
602adf40 | 105 | |
f84344f3 AE |
106 | /* The remaining fields need to be updated occasionally */ |
107 | u64 image_size; | |
108 | struct ceph_snap_context *snapc; | |
602adf40 YS |
109 | char *snap_names; |
110 | u64 *snap_sizes; | |
59c2be1e YS |
111 | |
112 | u64 obj_version; | |
113 | }; | |
114 | ||
0d7dbfce AE |
115 | /* |
116 | * An rbd image specification. | |
117 | * | |
118 | * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely | |
119 | * identify an image. | |
120 | */ | |
121 | struct rbd_spec { | |
122 | u64 pool_id; | |
123 | char *pool_name; | |
124 | ||
125 | char *image_id; | |
126 | size_t image_id_len; | |
127 | char *image_name; | |
128 | size_t image_name_len; | |
129 | ||
130 | u64 snap_id; | |
131 | char *snap_name; | |
132 | ||
133 | struct kref kref; | |
134 | }; | |
135 | ||
59c2be1e | 136 | struct rbd_options { |
cc0538b6 | 137 | bool read_only; |
602adf40 YS |
138 | }; |
139 | ||
140 | /* | |
f0f8cef5 | 141 | * an instance of the client. multiple devices may share an rbd client. |
602adf40 YS |
142 | */ |
143 | struct rbd_client { | |
144 | struct ceph_client *client; | |
145 | struct kref kref; | |
146 | struct list_head node; | |
147 | }; | |
148 | ||
149 | /* | |
f0f8cef5 | 150 | * a request completion status |
602adf40 | 151 | */ |
1fec7093 YS |
152 | struct rbd_req_status { |
153 | int done; | |
154 | int rc; | |
155 | u64 bytes; | |
156 | }; | |
157 | ||
158 | /* | |
159 | * a collection of requests | |
160 | */ | |
161 | struct rbd_req_coll { | |
162 | int total; | |
163 | int num_done; | |
164 | struct kref kref; | |
165 | struct rbd_req_status status[0]; | |
602adf40 YS |
166 | }; |
167 | ||
f0f8cef5 AE |
168 | /* |
169 | * a single io request | |
170 | */ | |
171 | struct rbd_request { | |
172 | struct request *rq; /* blk layer request */ | |
173 | struct bio *bio; /* cloned bio */ | |
174 | struct page **pages; /* list of used pages */ | |
175 | u64 len; | |
176 | int coll_index; | |
177 | struct rbd_req_coll *coll; | |
178 | }; | |
179 | ||
dfc5606d YS |
180 | struct rbd_snap { |
181 | struct device dev; | |
182 | const char *name; | |
3591538f | 183 | u64 size; |
dfc5606d YS |
184 | struct list_head node; |
185 | u64 id; | |
34b13184 | 186 | u64 features; |
dfc5606d YS |
187 | }; |
188 | ||
f84344f3 | 189 | struct rbd_mapping { |
99c1f08f | 190 | u64 size; |
34b13184 | 191 | u64 features; |
f84344f3 AE |
192 | bool read_only; |
193 | }; | |
194 | ||
602adf40 YS |
195 | /* |
196 | * a single device | |
197 | */ | |
198 | struct rbd_device { | |
de71a297 | 199 | int dev_id; /* blkdev unique id */ |
602adf40 YS |
200 | |
201 | int major; /* blkdev assigned major */ | |
202 | struct gendisk *disk; /* blkdev's gendisk and rq */ | |
602adf40 | 203 | |
a30b71b9 | 204 | u32 image_format; /* Either 1 or 2 */ |
602adf40 YS |
205 | struct rbd_client *rbd_client; |
206 | ||
207 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ | |
208 | ||
209 | spinlock_t lock; /* queue lock */ | |
210 | ||
211 | struct rbd_image_header header; | |
daba5fdb | 212 | bool exists; |
0d7dbfce | 213 | struct rbd_spec *spec; |
602adf40 | 214 | |
0d7dbfce | 215 | char *header_name; |
971f839a | 216 | |
59c2be1e YS |
217 | struct ceph_osd_event *watch_event; |
218 | struct ceph_osd_request *watch_request; | |
219 | ||
c666601a JD |
220 | /* protects updating the header */ |
221 | struct rw_semaphore header_rwsem; | |
f84344f3 AE |
222 | |
223 | struct rbd_mapping mapping; | |
602adf40 YS |
224 | |
225 | struct list_head node; | |
dfc5606d YS |
226 | |
227 | /* list of snapshots */ | |
228 | struct list_head snaps; | |
229 | ||
230 | /* sysfs related */ | |
231 | struct device dev; | |
232 | }; | |
233 | ||
602adf40 | 234 | static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ |
e124a82f | 235 | |
602adf40 | 236 | static LIST_HEAD(rbd_dev_list); /* devices */ |
e124a82f AE |
237 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
238 | ||
432b8587 AE |
239 | static LIST_HEAD(rbd_client_list); /* clients */ |
240 | static DEFINE_SPINLOCK(rbd_client_list_lock); | |
602adf40 | 241 | |
304f6808 AE |
242 | static int rbd_dev_snaps_update(struct rbd_device *rbd_dev); |
243 | static int rbd_dev_snaps_register(struct rbd_device *rbd_dev); | |
244 | ||
dfc5606d | 245 | static void rbd_dev_release(struct device *dev); |
41f38c2b | 246 | static void rbd_remove_snap_dev(struct rbd_snap *snap); |
dfc5606d | 247 | |
f0f8cef5 AE |
248 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
249 | size_t count); | |
250 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, | |
251 | size_t count); | |
252 | ||
253 | static struct bus_attribute rbd_bus_attrs[] = { | |
254 | __ATTR(add, S_IWUSR, NULL, rbd_add), | |
255 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), | |
256 | __ATTR_NULL | |
257 | }; | |
258 | ||
259 | static struct bus_type rbd_bus_type = { | |
260 | .name = "rbd", | |
261 | .bus_attrs = rbd_bus_attrs, | |
262 | }; | |
263 | ||
264 | static void rbd_root_dev_release(struct device *dev) | |
265 | { | |
266 | } | |
267 | ||
268 | static struct device rbd_root_dev = { | |
269 | .init_name = "rbd", | |
270 | .release = rbd_root_dev_release, | |
271 | }; | |
272 | ||
aafb230e AE |
273 | #ifdef RBD_DEBUG |
274 | #define rbd_assert(expr) \ | |
275 | if (unlikely(!(expr))) { \ | |
276 | printk(KERN_ERR "\nAssertion failure in %s() " \ | |
277 | "at line %d:\n\n" \ | |
278 | "\trbd_assert(%s);\n\n", \ | |
279 | __func__, __LINE__, #expr); \ | |
280 | BUG(); \ | |
281 | } | |
282 | #else /* !RBD_DEBUG */ | |
283 | # define rbd_assert(expr) ((void) 0) | |
284 | #endif /* !RBD_DEBUG */ | |
dfc5606d | 285 | |
dfc5606d YS |
286 | static struct device *rbd_get_dev(struct rbd_device *rbd_dev) |
287 | { | |
288 | return get_device(&rbd_dev->dev); | |
289 | } | |
290 | ||
291 | static void rbd_put_dev(struct rbd_device *rbd_dev) | |
292 | { | |
293 | put_device(&rbd_dev->dev); | |
294 | } | |
602adf40 | 295 | |
117973fb AE |
296 | static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver); |
297 | static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver); | |
59c2be1e | 298 | |
602adf40 YS |
299 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
300 | { | |
f0f8cef5 | 301 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
602adf40 | 302 | |
f84344f3 | 303 | if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only) |
602adf40 YS |
304 | return -EROFS; |
305 | ||
340c7a2b | 306 | rbd_get_dev(rbd_dev); |
f84344f3 | 307 | set_device_ro(bdev, rbd_dev->mapping.read_only); |
340c7a2b | 308 | |
602adf40 YS |
309 | return 0; |
310 | } | |
311 | ||
dfc5606d YS |
312 | static int rbd_release(struct gendisk *disk, fmode_t mode) |
313 | { | |
314 | struct rbd_device *rbd_dev = disk->private_data; | |
315 | ||
316 | rbd_put_dev(rbd_dev); | |
317 | ||
318 | return 0; | |
319 | } | |
320 | ||
602adf40 YS |
321 | static const struct block_device_operations rbd_bd_ops = { |
322 | .owner = THIS_MODULE, | |
323 | .open = rbd_open, | |
dfc5606d | 324 | .release = rbd_release, |
602adf40 YS |
325 | }; |
326 | ||
327 | /* | |
328 | * Initialize an rbd client instance. | |
43ae4701 | 329 | * We own *ceph_opts. |
602adf40 | 330 | */ |
f8c38929 | 331 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) |
602adf40 YS |
332 | { |
333 | struct rbd_client *rbdc; | |
334 | int ret = -ENOMEM; | |
335 | ||
336 | dout("rbd_client_create\n"); | |
337 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); | |
338 | if (!rbdc) | |
339 | goto out_opt; | |
340 | ||
341 | kref_init(&rbdc->kref); | |
342 | INIT_LIST_HEAD(&rbdc->node); | |
343 | ||
bc534d86 AE |
344 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
345 | ||
43ae4701 | 346 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
602adf40 | 347 | if (IS_ERR(rbdc->client)) |
bc534d86 | 348 | goto out_mutex; |
43ae4701 | 349 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
602adf40 YS |
350 | |
351 | ret = ceph_open_session(rbdc->client); | |
352 | if (ret < 0) | |
353 | goto out_err; | |
354 | ||
432b8587 | 355 | spin_lock(&rbd_client_list_lock); |
602adf40 | 356 | list_add_tail(&rbdc->node, &rbd_client_list); |
432b8587 | 357 | spin_unlock(&rbd_client_list_lock); |
602adf40 | 358 | |
bc534d86 AE |
359 | mutex_unlock(&ctl_mutex); |
360 | ||
602adf40 YS |
361 | dout("rbd_client_create created %p\n", rbdc); |
362 | return rbdc; | |
363 | ||
364 | out_err: | |
365 | ceph_destroy_client(rbdc->client); | |
bc534d86 AE |
366 | out_mutex: |
367 | mutex_unlock(&ctl_mutex); | |
602adf40 YS |
368 | kfree(rbdc); |
369 | out_opt: | |
43ae4701 AE |
370 | if (ceph_opts) |
371 | ceph_destroy_options(ceph_opts); | |
28f259b7 | 372 | return ERR_PTR(ret); |
602adf40 YS |
373 | } |
374 | ||
375 | /* | |
1f7ba331 AE |
376 | * Find a ceph client with specific addr and configuration. If |
377 | * found, bump its reference count. | |
602adf40 | 378 | */ |
1f7ba331 | 379 | static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) |
602adf40 YS |
380 | { |
381 | struct rbd_client *client_node; | |
1f7ba331 | 382 | bool found = false; |
602adf40 | 383 | |
43ae4701 | 384 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
602adf40 YS |
385 | return NULL; |
386 | ||
1f7ba331 AE |
387 | spin_lock(&rbd_client_list_lock); |
388 | list_for_each_entry(client_node, &rbd_client_list, node) { | |
389 | if (!ceph_compare_options(ceph_opts, client_node->client)) { | |
390 | kref_get(&client_node->kref); | |
391 | found = true; | |
392 | break; | |
393 | } | |
394 | } | |
395 | spin_unlock(&rbd_client_list_lock); | |
396 | ||
397 | return found ? client_node : NULL; | |
602adf40 YS |
398 | } |
399 | ||
59c2be1e YS |
400 | /* |
401 | * mount options | |
402 | */ | |
403 | enum { | |
59c2be1e YS |
404 | Opt_last_int, |
405 | /* int args above */ | |
406 | Opt_last_string, | |
407 | /* string args above */ | |
cc0538b6 AE |
408 | Opt_read_only, |
409 | Opt_read_write, | |
410 | /* Boolean args above */ | |
411 | Opt_last_bool, | |
59c2be1e YS |
412 | }; |
413 | ||
43ae4701 | 414 | static match_table_t rbd_opts_tokens = { |
59c2be1e YS |
415 | /* int args above */ |
416 | /* string args above */ | |
be466c1c | 417 | {Opt_read_only, "read_only"}, |
cc0538b6 AE |
418 | {Opt_read_only, "ro"}, /* Alternate spelling */ |
419 | {Opt_read_write, "read_write"}, | |
420 | {Opt_read_write, "rw"}, /* Alternate spelling */ | |
421 | /* Boolean args above */ | |
59c2be1e YS |
422 | {-1, NULL} |
423 | }; | |
424 | ||
425 | static int parse_rbd_opts_token(char *c, void *private) | |
426 | { | |
43ae4701 | 427 | struct rbd_options *rbd_opts = private; |
59c2be1e YS |
428 | substring_t argstr[MAX_OPT_ARGS]; |
429 | int token, intval, ret; | |
430 | ||
43ae4701 | 431 | token = match_token(c, rbd_opts_tokens, argstr); |
59c2be1e YS |
432 | if (token < 0) |
433 | return -EINVAL; | |
434 | ||
435 | if (token < Opt_last_int) { | |
436 | ret = match_int(&argstr[0], &intval); | |
437 | if (ret < 0) { | |
438 | pr_err("bad mount option arg (not int) " | |
439 | "at '%s'\n", c); | |
440 | return ret; | |
441 | } | |
442 | dout("got int token %d val %d\n", token, intval); | |
443 | } else if (token > Opt_last_int && token < Opt_last_string) { | |
444 | dout("got string token %d val %s\n", token, | |
445 | argstr[0].from); | |
cc0538b6 AE |
446 | } else if (token > Opt_last_string && token < Opt_last_bool) { |
447 | dout("got Boolean token %d\n", token); | |
59c2be1e YS |
448 | } else { |
449 | dout("got token %d\n", token); | |
450 | } | |
451 | ||
452 | switch (token) { | |
cc0538b6 AE |
453 | case Opt_read_only: |
454 | rbd_opts->read_only = true; | |
455 | break; | |
456 | case Opt_read_write: | |
457 | rbd_opts->read_only = false; | |
458 | break; | |
59c2be1e | 459 | default: |
aafb230e AE |
460 | rbd_assert(false); |
461 | break; | |
59c2be1e YS |
462 | } |
463 | return 0; | |
464 | } | |
465 | ||
602adf40 YS |
466 | /* |
467 | * Get a ceph client with specific addr and configuration, if one does | |
468 | * not exist create it. | |
469 | */ | |
78cea76e AE |
470 | static int rbd_get_client(struct rbd_device *rbd_dev, |
471 | struct ceph_options *ceph_opts) | |
602adf40 | 472 | { |
f8c38929 | 473 | struct rbd_client *rbdc; |
59c2be1e | 474 | |
1f7ba331 | 475 | rbdc = rbd_client_find(ceph_opts); |
602adf40 | 476 | if (rbdc) { |
602adf40 | 477 | /* using an existing client */ |
43ae4701 | 478 | ceph_destroy_options(ceph_opts); |
f8c38929 AE |
479 | } else { |
480 | rbdc = rbd_client_create(ceph_opts); | |
481 | if (IS_ERR(rbdc)) | |
482 | return PTR_ERR(rbdc); | |
602adf40 | 483 | } |
f8c38929 | 484 | rbd_dev->rbd_client = rbdc; |
602adf40 | 485 | |
f8c38929 | 486 | return 0; |
602adf40 YS |
487 | } |
488 | ||
489 | /* | |
490 | * Destroy ceph client | |
d23a4b3f | 491 | * |
432b8587 | 492 | * Caller must hold rbd_client_list_lock. |
602adf40 YS |
493 | */ |
494 | static void rbd_client_release(struct kref *kref) | |
495 | { | |
496 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); | |
497 | ||
498 | dout("rbd_release_client %p\n", rbdc); | |
cd9d9f5d | 499 | spin_lock(&rbd_client_list_lock); |
602adf40 | 500 | list_del(&rbdc->node); |
cd9d9f5d | 501 | spin_unlock(&rbd_client_list_lock); |
602adf40 YS |
502 | |
503 | ceph_destroy_client(rbdc->client); | |
504 | kfree(rbdc); | |
505 | } | |
506 | ||
507 | /* | |
508 | * Drop reference to ceph client node. If it's not referenced anymore, release | |
509 | * it. | |
510 | */ | |
511 | static void rbd_put_client(struct rbd_device *rbd_dev) | |
512 | { | |
513 | kref_put(&rbd_dev->rbd_client->kref, rbd_client_release); | |
514 | rbd_dev->rbd_client = NULL; | |
602adf40 YS |
515 | } |
516 | ||
1fec7093 YS |
517 | /* |
518 | * Destroy requests collection | |
519 | */ | |
520 | static void rbd_coll_release(struct kref *kref) | |
521 | { | |
522 | struct rbd_req_coll *coll = | |
523 | container_of(kref, struct rbd_req_coll, kref); | |
524 | ||
525 | dout("rbd_coll_release %p\n", coll); | |
526 | kfree(coll); | |
527 | } | |
602adf40 | 528 | |
a30b71b9 AE |
529 | static bool rbd_image_format_valid(u32 image_format) |
530 | { | |
531 | return image_format == 1 || image_format == 2; | |
532 | } | |
533 | ||
8e94af8e AE |
534 | static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) |
535 | { | |
103a150f AE |
536 | size_t size; |
537 | u32 snap_count; | |
538 | ||
539 | /* The header has to start with the magic rbd header text */ | |
540 | if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) | |
541 | return false; | |
542 | ||
db2388b6 AE |
543 | /* The bio layer requires at least sector-sized I/O */ |
544 | ||
545 | if (ondisk->options.order < SECTOR_SHIFT) | |
546 | return false; | |
547 | ||
548 | /* If we use u64 in a few spots we may be able to loosen this */ | |
549 | ||
550 | if (ondisk->options.order > 8 * sizeof (int) - 1) | |
551 | return false; | |
552 | ||
103a150f AE |
553 | /* |
554 | * The size of a snapshot header has to fit in a size_t, and | |
555 | * that limits the number of snapshots. | |
556 | */ | |
557 | snap_count = le32_to_cpu(ondisk->snap_count); | |
558 | size = SIZE_MAX - sizeof (struct ceph_snap_context); | |
559 | if (snap_count > size / sizeof (__le64)) | |
560 | return false; | |
561 | ||
562 | /* | |
563 | * Not only that, but the size of the entire the snapshot | |
564 | * header must also be representable in a size_t. | |
565 | */ | |
566 | size -= snap_count * sizeof (__le64); | |
567 | if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) | |
568 | return false; | |
569 | ||
570 | return true; | |
8e94af8e AE |
571 | } |
572 | ||
602adf40 YS |
573 | /* |
574 | * Create a new header structure, translate header format from the on-disk | |
575 | * header. | |
576 | */ | |
577 | static int rbd_header_from_disk(struct rbd_image_header *header, | |
4156d998 | 578 | struct rbd_image_header_ondisk *ondisk) |
602adf40 | 579 | { |
ccece235 | 580 | u32 snap_count; |
58c17b0e | 581 | size_t len; |
d2bb24e5 | 582 | size_t size; |
621901d6 | 583 | u32 i; |
602adf40 | 584 | |
6a52325f AE |
585 | memset(header, 0, sizeof (*header)); |
586 | ||
103a150f AE |
587 | snap_count = le32_to_cpu(ondisk->snap_count); |
588 | ||
58c17b0e AE |
589 | len = strnlen(ondisk->object_prefix, sizeof (ondisk->object_prefix)); |
590 | header->object_prefix = kmalloc(len + 1, GFP_KERNEL); | |
6a52325f | 591 | if (!header->object_prefix) |
602adf40 | 592 | return -ENOMEM; |
58c17b0e AE |
593 | memcpy(header->object_prefix, ondisk->object_prefix, len); |
594 | header->object_prefix[len] = '\0'; | |
00f1f36f | 595 | |
602adf40 | 596 | if (snap_count) { |
f785cc1d AE |
597 | u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
598 | ||
621901d6 AE |
599 | /* Save a copy of the snapshot names */ |
600 | ||
f785cc1d AE |
601 | if (snap_names_len > (u64) SIZE_MAX) |
602 | return -EIO; | |
603 | header->snap_names = kmalloc(snap_names_len, GFP_KERNEL); | |
602adf40 | 604 | if (!header->snap_names) |
6a52325f | 605 | goto out_err; |
f785cc1d AE |
606 | /* |
607 | * Note that rbd_dev_v1_header_read() guarantees | |
608 | * the ondisk buffer we're working with has | |
609 | * snap_names_len bytes beyond the end of the | |
610 | * snapshot id array, this memcpy() is safe. | |
611 | */ | |
612 | memcpy(header->snap_names, &ondisk->snaps[snap_count], | |
613 | snap_names_len); | |
6a52325f | 614 | |
621901d6 AE |
615 | /* Record each snapshot's size */ |
616 | ||
d2bb24e5 AE |
617 | size = snap_count * sizeof (*header->snap_sizes); |
618 | header->snap_sizes = kmalloc(size, GFP_KERNEL); | |
602adf40 | 619 | if (!header->snap_sizes) |
6a52325f | 620 | goto out_err; |
621901d6 AE |
621 | for (i = 0; i < snap_count; i++) |
622 | header->snap_sizes[i] = | |
623 | le64_to_cpu(ondisk->snaps[i].image_size); | |
602adf40 | 624 | } else { |
ccece235 | 625 | WARN_ON(ondisk->snap_names_len); |
602adf40 YS |
626 | header->snap_names = NULL; |
627 | header->snap_sizes = NULL; | |
628 | } | |
849b4260 | 629 | |
34b13184 | 630 | header->features = 0; /* No features support in v1 images */ |
602adf40 YS |
631 | header->obj_order = ondisk->options.order; |
632 | header->crypt_type = ondisk->options.crypt_type; | |
633 | header->comp_type = ondisk->options.comp_type; | |
6a52325f | 634 | |
621901d6 AE |
635 | /* Allocate and fill in the snapshot context */ |
636 | ||
f84344f3 | 637 | header->image_size = le64_to_cpu(ondisk->image_size); |
6a52325f AE |
638 | size = sizeof (struct ceph_snap_context); |
639 | size += snap_count * sizeof (header->snapc->snaps[0]); | |
640 | header->snapc = kzalloc(size, GFP_KERNEL); | |
641 | if (!header->snapc) | |
642 | goto out_err; | |
602adf40 YS |
643 | |
644 | atomic_set(&header->snapc->nref, 1); | |
505cbb9b | 645 | header->snapc->seq = le64_to_cpu(ondisk->snap_seq); |
602adf40 | 646 | header->snapc->num_snaps = snap_count; |
621901d6 AE |
647 | for (i = 0; i < snap_count; i++) |
648 | header->snapc->snaps[i] = | |
649 | le64_to_cpu(ondisk->snaps[i].id); | |
602adf40 YS |
650 | |
651 | return 0; | |
652 | ||
6a52325f | 653 | out_err: |
849b4260 | 654 | kfree(header->snap_sizes); |
ccece235 | 655 | header->snap_sizes = NULL; |
602adf40 | 656 | kfree(header->snap_names); |
ccece235 | 657 | header->snap_names = NULL; |
6a52325f AE |
658 | kfree(header->object_prefix); |
659 | header->object_prefix = NULL; | |
ccece235 | 660 | |
00f1f36f | 661 | return -ENOMEM; |
602adf40 YS |
662 | } |
663 | ||
8836b995 | 664 | static int snap_by_name(struct rbd_device *rbd_dev, const char *snap_name) |
602adf40 | 665 | { |
602adf40 | 666 | |
e86924a8 | 667 | struct rbd_snap *snap; |
602adf40 | 668 | |
e86924a8 AE |
669 | list_for_each_entry(snap, &rbd_dev->snaps, node) { |
670 | if (!strcmp(snap_name, snap->name)) { | |
0d7dbfce | 671 | rbd_dev->spec->snap_id = snap->id; |
e86924a8 | 672 | rbd_dev->mapping.size = snap->size; |
34b13184 | 673 | rbd_dev->mapping.features = snap->features; |
602adf40 | 674 | |
e86924a8 | 675 | return 0; |
00f1f36f | 676 | } |
00f1f36f | 677 | } |
e86924a8 | 678 | |
00f1f36f | 679 | return -ENOENT; |
602adf40 YS |
680 | } |
681 | ||
819d52bf | 682 | static int rbd_dev_set_mapping(struct rbd_device *rbd_dev) |
602adf40 | 683 | { |
78dc447d | 684 | int ret; |
602adf40 | 685 | |
0d7dbfce | 686 | if (!memcmp(rbd_dev->spec->snap_name, RBD_SNAP_HEAD_NAME, |
cc9d734c | 687 | sizeof (RBD_SNAP_HEAD_NAME))) { |
0d7dbfce | 688 | rbd_dev->spec->snap_id = CEPH_NOSNAP; |
99c1f08f | 689 | rbd_dev->mapping.size = rbd_dev->header.image_size; |
34b13184 | 690 | rbd_dev->mapping.features = rbd_dev->header.features; |
e86924a8 | 691 | ret = 0; |
602adf40 | 692 | } else { |
0d7dbfce | 693 | ret = snap_by_name(rbd_dev, rbd_dev->spec->snap_name); |
602adf40 YS |
694 | if (ret < 0) |
695 | goto done; | |
f84344f3 | 696 | rbd_dev->mapping.read_only = true; |
602adf40 | 697 | } |
daba5fdb | 698 | rbd_dev->exists = true; |
602adf40 | 699 | done: |
602adf40 YS |
700 | return ret; |
701 | } | |
702 | ||
703 | static void rbd_header_free(struct rbd_image_header *header) | |
704 | { | |
849b4260 | 705 | kfree(header->object_prefix); |
d78fd7ae | 706 | header->object_prefix = NULL; |
602adf40 | 707 | kfree(header->snap_sizes); |
d78fd7ae | 708 | header->snap_sizes = NULL; |
849b4260 | 709 | kfree(header->snap_names); |
d78fd7ae | 710 | header->snap_names = NULL; |
d1d25646 | 711 | ceph_put_snap_context(header->snapc); |
d78fd7ae | 712 | header->snapc = NULL; |
602adf40 YS |
713 | } |
714 | ||
65ccfe21 | 715 | static char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset) |
602adf40 | 716 | { |
65ccfe21 AE |
717 | char *name; |
718 | u64 segment; | |
719 | int ret; | |
602adf40 | 720 | |
65ccfe21 AE |
721 | name = kmalloc(RBD_MAX_SEG_NAME_LEN + 1, GFP_NOIO); |
722 | if (!name) | |
723 | return NULL; | |
724 | segment = offset >> rbd_dev->header.obj_order; | |
725 | ret = snprintf(name, RBD_MAX_SEG_NAME_LEN, "%s.%012llx", | |
726 | rbd_dev->header.object_prefix, segment); | |
727 | if (ret < 0 || ret >= RBD_MAX_SEG_NAME_LEN) { | |
728 | pr_err("error formatting segment name for #%llu (%d)\n", | |
729 | segment, ret); | |
730 | kfree(name); | |
731 | name = NULL; | |
732 | } | |
602adf40 | 733 | |
65ccfe21 AE |
734 | return name; |
735 | } | |
602adf40 | 736 | |
65ccfe21 AE |
737 | static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) |
738 | { | |
739 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; | |
602adf40 | 740 | |
65ccfe21 AE |
741 | return offset & (segment_size - 1); |
742 | } | |
743 | ||
744 | static u64 rbd_segment_length(struct rbd_device *rbd_dev, | |
745 | u64 offset, u64 length) | |
746 | { | |
747 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; | |
748 | ||
749 | offset &= segment_size - 1; | |
750 | ||
aafb230e | 751 | rbd_assert(length <= U64_MAX - offset); |
65ccfe21 AE |
752 | if (offset + length > segment_size) |
753 | length = segment_size - offset; | |
754 | ||
755 | return length; | |
602adf40 YS |
756 | } |
757 | ||
1fec7093 YS |
758 | static int rbd_get_num_segments(struct rbd_image_header *header, |
759 | u64 ofs, u64 len) | |
760 | { | |
df111be6 AE |
761 | u64 start_seg; |
762 | u64 end_seg; | |
763 | ||
764 | if (!len) | |
765 | return 0; | |
766 | if (len - 1 > U64_MAX - ofs) | |
767 | return -ERANGE; | |
768 | ||
769 | start_seg = ofs >> header->obj_order; | |
770 | end_seg = (ofs + len - 1) >> header->obj_order; | |
771 | ||
1fec7093 YS |
772 | return end_seg - start_seg + 1; |
773 | } | |
774 | ||
029bcbd8 JD |
775 | /* |
776 | * returns the size of an object in the image | |
777 | */ | |
778 | static u64 rbd_obj_bytes(struct rbd_image_header *header) | |
779 | { | |
780 | return 1 << header->obj_order; | |
781 | } | |
782 | ||
602adf40 YS |
783 | /* |
784 | * bio helpers | |
785 | */ | |
786 | ||
787 | static void bio_chain_put(struct bio *chain) | |
788 | { | |
789 | struct bio *tmp; | |
790 | ||
791 | while (chain) { | |
792 | tmp = chain; | |
793 | chain = chain->bi_next; | |
794 | bio_put(tmp); | |
795 | } | |
796 | } | |
797 | ||
798 | /* | |
799 | * zeros a bio chain, starting at specific offset | |
800 | */ | |
801 | static void zero_bio_chain(struct bio *chain, int start_ofs) | |
802 | { | |
803 | struct bio_vec *bv; | |
804 | unsigned long flags; | |
805 | void *buf; | |
806 | int i; | |
807 | int pos = 0; | |
808 | ||
809 | while (chain) { | |
810 | bio_for_each_segment(bv, chain, i) { | |
811 | if (pos + bv->bv_len > start_ofs) { | |
812 | int remainder = max(start_ofs - pos, 0); | |
813 | buf = bvec_kmap_irq(bv, &flags); | |
814 | memset(buf + remainder, 0, | |
815 | bv->bv_len - remainder); | |
85b5aaa6 | 816 | bvec_kunmap_irq(buf, &flags); |
602adf40 YS |
817 | } |
818 | pos += bv->bv_len; | |
819 | } | |
820 | ||
821 | chain = chain->bi_next; | |
822 | } | |
823 | } | |
824 | ||
825 | /* | |
f7760dad AE |
826 | * Clone a portion of a bio, starting at the given byte offset |
827 | * and continuing for the number of bytes indicated. | |
602adf40 | 828 | */ |
f7760dad AE |
829 | static struct bio *bio_clone_range(struct bio *bio_src, |
830 | unsigned int offset, | |
831 | unsigned int len, | |
832 | gfp_t gfpmask) | |
602adf40 | 833 | { |
f7760dad AE |
834 | struct bio_vec *bv; |
835 | unsigned int resid; | |
836 | unsigned short idx; | |
837 | unsigned int voff; | |
838 | unsigned short end_idx; | |
839 | unsigned short vcnt; | |
840 | struct bio *bio; | |
841 | ||
842 | /* Handle the easy case for the caller */ | |
843 | ||
844 | if (!offset && len == bio_src->bi_size) | |
845 | return bio_clone(bio_src, gfpmask); | |
846 | ||
847 | if (WARN_ON_ONCE(!len)) | |
848 | return NULL; | |
849 | if (WARN_ON_ONCE(len > bio_src->bi_size)) | |
850 | return NULL; | |
851 | if (WARN_ON_ONCE(offset > bio_src->bi_size - len)) | |
852 | return NULL; | |
853 | ||
854 | /* Find first affected segment... */ | |
855 | ||
856 | resid = offset; | |
857 | __bio_for_each_segment(bv, bio_src, idx, 0) { | |
858 | if (resid < bv->bv_len) | |
859 | break; | |
860 | resid -= bv->bv_len; | |
602adf40 | 861 | } |
f7760dad | 862 | voff = resid; |
602adf40 | 863 | |
f7760dad | 864 | /* ...and the last affected segment */ |
602adf40 | 865 | |
f7760dad AE |
866 | resid += len; |
867 | __bio_for_each_segment(bv, bio_src, end_idx, idx) { | |
868 | if (resid <= bv->bv_len) | |
869 | break; | |
870 | resid -= bv->bv_len; | |
871 | } | |
872 | vcnt = end_idx - idx + 1; | |
873 | ||
874 | /* Build the clone */ | |
875 | ||
876 | bio = bio_alloc(gfpmask, (unsigned int) vcnt); | |
877 | if (!bio) | |
878 | return NULL; /* ENOMEM */ | |
602adf40 | 879 | |
f7760dad AE |
880 | bio->bi_bdev = bio_src->bi_bdev; |
881 | bio->bi_sector = bio_src->bi_sector + (offset >> SECTOR_SHIFT); | |
882 | bio->bi_rw = bio_src->bi_rw; | |
883 | bio->bi_flags |= 1 << BIO_CLONED; | |
884 | ||
885 | /* | |
886 | * Copy over our part of the bio_vec, then update the first | |
887 | * and last (or only) entries. | |
888 | */ | |
889 | memcpy(&bio->bi_io_vec[0], &bio_src->bi_io_vec[idx], | |
890 | vcnt * sizeof (struct bio_vec)); | |
891 | bio->bi_io_vec[0].bv_offset += voff; | |
892 | if (vcnt > 1) { | |
893 | bio->bi_io_vec[0].bv_len -= voff; | |
894 | bio->bi_io_vec[vcnt - 1].bv_len = resid; | |
895 | } else { | |
896 | bio->bi_io_vec[0].bv_len = len; | |
602adf40 YS |
897 | } |
898 | ||
f7760dad AE |
899 | bio->bi_vcnt = vcnt; |
900 | bio->bi_size = len; | |
901 | bio->bi_idx = 0; | |
902 | ||
903 | return bio; | |
904 | } | |
905 | ||
906 | /* | |
907 | * Clone a portion of a bio chain, starting at the given byte offset | |
908 | * into the first bio in the source chain and continuing for the | |
909 | * number of bytes indicated. The result is another bio chain of | |
910 | * exactly the given length, or a null pointer on error. | |
911 | * | |
912 | * The bio_src and offset parameters are both in-out. On entry they | |
913 | * refer to the first source bio and the offset into that bio where | |
914 | * the start of data to be cloned is located. | |
915 | * | |
916 | * On return, bio_src is updated to refer to the bio in the source | |
917 | * chain that contains first un-cloned byte, and *offset will | |
918 | * contain the offset of that byte within that bio. | |
919 | */ | |
920 | static struct bio *bio_chain_clone_range(struct bio **bio_src, | |
921 | unsigned int *offset, | |
922 | unsigned int len, | |
923 | gfp_t gfpmask) | |
924 | { | |
925 | struct bio *bi = *bio_src; | |
926 | unsigned int off = *offset; | |
927 | struct bio *chain = NULL; | |
928 | struct bio **end; | |
929 | ||
930 | /* Build up a chain of clone bios up to the limit */ | |
931 | ||
932 | if (!bi || off >= bi->bi_size || !len) | |
933 | return NULL; /* Nothing to clone */ | |
602adf40 | 934 | |
f7760dad AE |
935 | end = &chain; |
936 | while (len) { | |
937 | unsigned int bi_size; | |
938 | struct bio *bio; | |
939 | ||
940 | if (!bi) | |
941 | goto out_err; /* EINVAL; ran out of bio's */ | |
942 | bi_size = min_t(unsigned int, bi->bi_size - off, len); | |
943 | bio = bio_clone_range(bi, off, bi_size, gfpmask); | |
944 | if (!bio) | |
945 | goto out_err; /* ENOMEM */ | |
946 | ||
947 | *end = bio; | |
948 | end = &bio->bi_next; | |
602adf40 | 949 | |
f7760dad AE |
950 | off += bi_size; |
951 | if (off == bi->bi_size) { | |
952 | bi = bi->bi_next; | |
953 | off = 0; | |
954 | } | |
955 | len -= bi_size; | |
956 | } | |
957 | *bio_src = bi; | |
958 | *offset = off; | |
959 | ||
960 | return chain; | |
961 | out_err: | |
962 | bio_chain_put(chain); | |
602adf40 | 963 | |
602adf40 YS |
964 | return NULL; |
965 | } | |
966 | ||
967 | /* | |
968 | * helpers for osd request op vectors. | |
969 | */ | |
57cfc106 AE |
970 | static struct ceph_osd_req_op *rbd_create_rw_ops(int num_ops, |
971 | int opcode, u32 payload_len) | |
602adf40 | 972 | { |
57cfc106 AE |
973 | struct ceph_osd_req_op *ops; |
974 | ||
975 | ops = kzalloc(sizeof (*ops) * (num_ops + 1), GFP_NOIO); | |
976 | if (!ops) | |
977 | return NULL; | |
978 | ||
979 | ops[0].op = opcode; | |
980 | ||
602adf40 YS |
981 | /* |
982 | * op extent offset and length will be set later on | |
983 | * in calc_raw_layout() | |
984 | */ | |
57cfc106 AE |
985 | ops[0].payload_len = payload_len; |
986 | ||
987 | return ops; | |
602adf40 YS |
988 | } |
989 | ||
990 | static void rbd_destroy_ops(struct ceph_osd_req_op *ops) | |
991 | { | |
992 | kfree(ops); | |
993 | } | |
994 | ||
1fec7093 YS |
995 | static void rbd_coll_end_req_index(struct request *rq, |
996 | struct rbd_req_coll *coll, | |
997 | int index, | |
998 | int ret, u64 len) | |
999 | { | |
1000 | struct request_queue *q; | |
1001 | int min, max, i; | |
1002 | ||
bd919d45 AE |
1003 | dout("rbd_coll_end_req_index %p index %d ret %d len %llu\n", |
1004 | coll, index, ret, (unsigned long long) len); | |
1fec7093 YS |
1005 | |
1006 | if (!rq) | |
1007 | return; | |
1008 | ||
1009 | if (!coll) { | |
1010 | blk_end_request(rq, ret, len); | |
1011 | return; | |
1012 | } | |
1013 | ||
1014 | q = rq->q; | |
1015 | ||
1016 | spin_lock_irq(q->queue_lock); | |
1017 | coll->status[index].done = 1; | |
1018 | coll->status[index].rc = ret; | |
1019 | coll->status[index].bytes = len; | |
1020 | max = min = coll->num_done; | |
1021 | while (max < coll->total && coll->status[max].done) | |
1022 | max++; | |
1023 | ||
1024 | for (i = min; i<max; i++) { | |
1025 | __blk_end_request(rq, coll->status[i].rc, | |
1026 | coll->status[i].bytes); | |
1027 | coll->num_done++; | |
1028 | kref_put(&coll->kref, rbd_coll_release); | |
1029 | } | |
1030 | spin_unlock_irq(q->queue_lock); | |
1031 | } | |
1032 | ||
1033 | static void rbd_coll_end_req(struct rbd_request *req, | |
1034 | int ret, u64 len) | |
1035 | { | |
1036 | rbd_coll_end_req_index(req->rq, req->coll, req->coll_index, ret, len); | |
1037 | } | |
1038 | ||
602adf40 YS |
1039 | /* |
1040 | * Send ceph osd request | |
1041 | */ | |
1042 | static int rbd_do_request(struct request *rq, | |
0ce1a794 | 1043 | struct rbd_device *rbd_dev, |
602adf40 YS |
1044 | struct ceph_snap_context *snapc, |
1045 | u64 snapid, | |
aded07ea | 1046 | const char *object_name, u64 ofs, u64 len, |
602adf40 YS |
1047 | struct bio *bio, |
1048 | struct page **pages, | |
1049 | int num_pages, | |
1050 | int flags, | |
1051 | struct ceph_osd_req_op *ops, | |
1fec7093 YS |
1052 | struct rbd_req_coll *coll, |
1053 | int coll_index, | |
602adf40 | 1054 | void (*rbd_cb)(struct ceph_osd_request *req, |
59c2be1e YS |
1055 | struct ceph_msg *msg), |
1056 | struct ceph_osd_request **linger_req, | |
1057 | u64 *ver) | |
602adf40 YS |
1058 | { |
1059 | struct ceph_osd_request *req; | |
1060 | struct ceph_file_layout *layout; | |
1061 | int ret; | |
1062 | u64 bno; | |
1063 | struct timespec mtime = CURRENT_TIME; | |
1064 | struct rbd_request *req_data; | |
1065 | struct ceph_osd_request_head *reqhead; | |
1dbb4399 | 1066 | struct ceph_osd_client *osdc; |
602adf40 | 1067 | |
602adf40 | 1068 | req_data = kzalloc(sizeof(*req_data), GFP_NOIO); |
1fec7093 YS |
1069 | if (!req_data) { |
1070 | if (coll) | |
1071 | rbd_coll_end_req_index(rq, coll, coll_index, | |
1072 | -ENOMEM, len); | |
1073 | return -ENOMEM; | |
1074 | } | |
1075 | ||
1076 | if (coll) { | |
1077 | req_data->coll = coll; | |
1078 | req_data->coll_index = coll_index; | |
1079 | } | |
602adf40 | 1080 | |
f7760dad AE |
1081 | dout("rbd_do_request object_name=%s ofs=%llu len=%llu coll=%p[%d]\n", |
1082 | object_name, (unsigned long long) ofs, | |
1083 | (unsigned long long) len, coll, coll_index); | |
602adf40 | 1084 | |
0ce1a794 | 1085 | osdc = &rbd_dev->rbd_client->client->osdc; |
1dbb4399 AE |
1086 | req = ceph_osdc_alloc_request(osdc, flags, snapc, ops, |
1087 | false, GFP_NOIO, pages, bio); | |
4ad12621 | 1088 | if (!req) { |
4ad12621 | 1089 | ret = -ENOMEM; |
602adf40 YS |
1090 | goto done_pages; |
1091 | } | |
1092 | ||
1093 | req->r_callback = rbd_cb; | |
1094 | ||
1095 | req_data->rq = rq; | |
1096 | req_data->bio = bio; | |
1097 | req_data->pages = pages; | |
1098 | req_data->len = len; | |
1099 | ||
1100 | req->r_priv = req_data; | |
1101 | ||
1102 | reqhead = req->r_request->front.iov_base; | |
1103 | reqhead->snapid = cpu_to_le64(CEPH_NOSNAP); | |
1104 | ||
aded07ea | 1105 | strncpy(req->r_oid, object_name, sizeof(req->r_oid)); |
602adf40 YS |
1106 | req->r_oid_len = strlen(req->r_oid); |
1107 | ||
1108 | layout = &req->r_file_layout; | |
1109 | memset(layout, 0, sizeof(*layout)); | |
1110 | layout->fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
1111 | layout->fl_stripe_count = cpu_to_le32(1); | |
1112 | layout->fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
0d7dbfce | 1113 | layout->fl_pg_pool = cpu_to_le32((int) rbd_dev->spec->pool_id); |
6cae3717 SW |
1114 | ret = ceph_calc_raw_layout(osdc, layout, snapid, ofs, &len, &bno, |
1115 | req, ops); | |
1116 | rbd_assert(ret == 0); | |
602adf40 YS |
1117 | |
1118 | ceph_osdc_build_request(req, ofs, &len, | |
1119 | ops, | |
1120 | snapc, | |
1121 | &mtime, | |
1122 | req->r_oid, req->r_oid_len); | |
602adf40 | 1123 | |
59c2be1e | 1124 | if (linger_req) { |
1dbb4399 | 1125 | ceph_osdc_set_request_linger(osdc, req); |
59c2be1e YS |
1126 | *linger_req = req; |
1127 | } | |
1128 | ||
1dbb4399 | 1129 | ret = ceph_osdc_start_request(osdc, req, false); |
602adf40 YS |
1130 | if (ret < 0) |
1131 | goto done_err; | |
1132 | ||
1133 | if (!rbd_cb) { | |
1dbb4399 | 1134 | ret = ceph_osdc_wait_request(osdc, req); |
59c2be1e YS |
1135 | if (ver) |
1136 | *ver = le64_to_cpu(req->r_reassert_version.version); | |
bd919d45 AE |
1137 | dout("reassert_ver=%llu\n", |
1138 | (unsigned long long) | |
1139 | le64_to_cpu(req->r_reassert_version.version)); | |
602adf40 YS |
1140 | ceph_osdc_put_request(req); |
1141 | } | |
1142 | return ret; | |
1143 | ||
1144 | done_err: | |
1145 | bio_chain_put(req_data->bio); | |
1146 | ceph_osdc_put_request(req); | |
1147 | done_pages: | |
1fec7093 | 1148 | rbd_coll_end_req(req_data, ret, len); |
602adf40 | 1149 | kfree(req_data); |
602adf40 YS |
1150 | return ret; |
1151 | } | |
1152 | ||
1153 | /* | |
1154 | * Ceph osd op callback | |
1155 | */ | |
1156 | static void rbd_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) | |
1157 | { | |
1158 | struct rbd_request *req_data = req->r_priv; | |
1159 | struct ceph_osd_reply_head *replyhead; | |
1160 | struct ceph_osd_op *op; | |
1161 | __s32 rc; | |
1162 | u64 bytes; | |
1163 | int read_op; | |
1164 | ||
1165 | /* parse reply */ | |
1166 | replyhead = msg->front.iov_base; | |
1167 | WARN_ON(le32_to_cpu(replyhead->num_ops) == 0); | |
1168 | op = (void *)(replyhead + 1); | |
1169 | rc = le32_to_cpu(replyhead->result); | |
1170 | bytes = le64_to_cpu(op->extent.length); | |
895cfcc8 | 1171 | read_op = (le16_to_cpu(op->op) == CEPH_OSD_OP_READ); |
602adf40 | 1172 | |
bd919d45 AE |
1173 | dout("rbd_req_cb bytes=%llu readop=%d rc=%d\n", |
1174 | (unsigned long long) bytes, read_op, (int) rc); | |
602adf40 YS |
1175 | |
1176 | if (rc == -ENOENT && read_op) { | |
1177 | zero_bio_chain(req_data->bio, 0); | |
1178 | rc = 0; | |
1179 | } else if (rc == 0 && read_op && bytes < req_data->len) { | |
1180 | zero_bio_chain(req_data->bio, bytes); | |
1181 | bytes = req_data->len; | |
1182 | } | |
1183 | ||
1fec7093 | 1184 | rbd_coll_end_req(req_data, rc, bytes); |
602adf40 YS |
1185 | |
1186 | if (req_data->bio) | |
1187 | bio_chain_put(req_data->bio); | |
1188 | ||
1189 | ceph_osdc_put_request(req); | |
1190 | kfree(req_data); | |
1191 | } | |
1192 | ||
59c2be1e YS |
1193 | static void rbd_simple_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) |
1194 | { | |
1195 | ceph_osdc_put_request(req); | |
1196 | } | |
1197 | ||
602adf40 YS |
1198 | /* |
1199 | * Do a synchronous ceph osd operation | |
1200 | */ | |
0ce1a794 | 1201 | static int rbd_req_sync_op(struct rbd_device *rbd_dev, |
602adf40 YS |
1202 | struct ceph_snap_context *snapc, |
1203 | u64 snapid, | |
602adf40 | 1204 | int flags, |
913d2fdc | 1205 | struct ceph_osd_req_op *ops, |
aded07ea | 1206 | const char *object_name, |
f8d4de6e AE |
1207 | u64 ofs, u64 inbound_size, |
1208 | char *inbound, | |
59c2be1e YS |
1209 | struct ceph_osd_request **linger_req, |
1210 | u64 *ver) | |
602adf40 YS |
1211 | { |
1212 | int ret; | |
1213 | struct page **pages; | |
1214 | int num_pages; | |
913d2fdc | 1215 | |
aafb230e | 1216 | rbd_assert(ops != NULL); |
602adf40 | 1217 | |
f8d4de6e | 1218 | num_pages = calc_pages_for(ofs, inbound_size); |
602adf40 | 1219 | pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); |
b8d0638a DC |
1220 | if (IS_ERR(pages)) |
1221 | return PTR_ERR(pages); | |
602adf40 | 1222 | |
0ce1a794 | 1223 | ret = rbd_do_request(NULL, rbd_dev, snapc, snapid, |
f8d4de6e | 1224 | object_name, ofs, inbound_size, NULL, |
602adf40 YS |
1225 | pages, num_pages, |
1226 | flags, | |
1227 | ops, | |
1fec7093 | 1228 | NULL, 0, |
59c2be1e YS |
1229 | NULL, |
1230 | linger_req, ver); | |
602adf40 | 1231 | if (ret < 0) |
913d2fdc | 1232 | goto done; |
602adf40 | 1233 | |
f8d4de6e AE |
1234 | if ((flags & CEPH_OSD_FLAG_READ) && inbound) |
1235 | ret = ceph_copy_from_page_vector(pages, inbound, ofs, ret); | |
602adf40 | 1236 | |
602adf40 YS |
1237 | done: |
1238 | ceph_release_page_vector(pages, num_pages); | |
1239 | return ret; | |
1240 | } | |
1241 | ||
1242 | /* | |
1243 | * Do an asynchronous ceph osd operation | |
1244 | */ | |
1245 | static int rbd_do_op(struct request *rq, | |
0ce1a794 | 1246 | struct rbd_device *rbd_dev, |
602adf40 | 1247 | struct ceph_snap_context *snapc, |
602adf40 | 1248 | u64 ofs, u64 len, |
1fec7093 YS |
1249 | struct bio *bio, |
1250 | struct rbd_req_coll *coll, | |
1251 | int coll_index) | |
602adf40 YS |
1252 | { |
1253 | char *seg_name; | |
1254 | u64 seg_ofs; | |
1255 | u64 seg_len; | |
1256 | int ret; | |
1257 | struct ceph_osd_req_op *ops; | |
1258 | u32 payload_len; | |
ff2e4bb5 AE |
1259 | int opcode; |
1260 | int flags; | |
4634246d | 1261 | u64 snapid; |
602adf40 | 1262 | |
65ccfe21 | 1263 | seg_name = rbd_segment_name(rbd_dev, ofs); |
602adf40 YS |
1264 | if (!seg_name) |
1265 | return -ENOMEM; | |
65ccfe21 AE |
1266 | seg_len = rbd_segment_length(rbd_dev, ofs, len); |
1267 | seg_ofs = rbd_segment_offset(rbd_dev, ofs); | |
602adf40 | 1268 | |
ff2e4bb5 AE |
1269 | if (rq_data_dir(rq) == WRITE) { |
1270 | opcode = CEPH_OSD_OP_WRITE; | |
1271 | flags = CEPH_OSD_FLAG_WRITE|CEPH_OSD_FLAG_ONDISK; | |
4634246d | 1272 | snapid = CEPH_NOSNAP; |
ff2e4bb5 AE |
1273 | payload_len = seg_len; |
1274 | } else { | |
1275 | opcode = CEPH_OSD_OP_READ; | |
1276 | flags = CEPH_OSD_FLAG_READ; | |
4634246d | 1277 | snapc = NULL; |
0d7dbfce | 1278 | snapid = rbd_dev->spec->snap_id; |
ff2e4bb5 AE |
1279 | payload_len = 0; |
1280 | } | |
602adf40 | 1281 | |
57cfc106 AE |
1282 | ret = -ENOMEM; |
1283 | ops = rbd_create_rw_ops(1, opcode, payload_len); | |
1284 | if (!ops) | |
602adf40 YS |
1285 | goto done; |
1286 | ||
1287 | /* we've taken care of segment sizes earlier when we | |
1288 | cloned the bios. We should never have a segment | |
1289 | truncated at this point */ | |
aafb230e | 1290 | rbd_assert(seg_len == len); |
602adf40 YS |
1291 | |
1292 | ret = rbd_do_request(rq, rbd_dev, snapc, snapid, | |
1293 | seg_name, seg_ofs, seg_len, | |
1294 | bio, | |
1295 | NULL, 0, | |
1296 | flags, | |
1297 | ops, | |
1fec7093 | 1298 | coll, coll_index, |
59c2be1e | 1299 | rbd_req_cb, 0, NULL); |
11f77002 SW |
1300 | |
1301 | rbd_destroy_ops(ops); | |
602adf40 YS |
1302 | done: |
1303 | kfree(seg_name); | |
1304 | return ret; | |
1305 | } | |
1306 | ||
602adf40 YS |
1307 | /* |
1308 | * Request sync osd read | |
1309 | */ | |
0ce1a794 | 1310 | static int rbd_req_sync_read(struct rbd_device *rbd_dev, |
602adf40 | 1311 | u64 snapid, |
aded07ea | 1312 | const char *object_name, |
602adf40 | 1313 | u64 ofs, u64 len, |
59c2be1e YS |
1314 | char *buf, |
1315 | u64 *ver) | |
602adf40 | 1316 | { |
913d2fdc AE |
1317 | struct ceph_osd_req_op *ops; |
1318 | int ret; | |
1319 | ||
1320 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_READ, 0); | |
1321 | if (!ops) | |
1322 | return -ENOMEM; | |
1323 | ||
1324 | ret = rbd_req_sync_op(rbd_dev, NULL, | |
b06e6a6b | 1325 | snapid, |
602adf40 | 1326 | CEPH_OSD_FLAG_READ, |
913d2fdc AE |
1327 | ops, object_name, ofs, len, buf, NULL, ver); |
1328 | rbd_destroy_ops(ops); | |
1329 | ||
1330 | return ret; | |
602adf40 YS |
1331 | } |
1332 | ||
1333 | /* | |
59c2be1e YS |
1334 | * Request sync osd watch |
1335 | */ | |
0ce1a794 | 1336 | static int rbd_req_sync_notify_ack(struct rbd_device *rbd_dev, |
59c2be1e | 1337 | u64 ver, |
7f0a24d8 | 1338 | u64 notify_id) |
59c2be1e YS |
1339 | { |
1340 | struct ceph_osd_req_op *ops; | |
11f77002 SW |
1341 | int ret; |
1342 | ||
57cfc106 AE |
1343 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_NOTIFY_ACK, 0); |
1344 | if (!ops) | |
1345 | return -ENOMEM; | |
59c2be1e | 1346 | |
a71b891b | 1347 | ops[0].watch.ver = cpu_to_le64(ver); |
59c2be1e YS |
1348 | ops[0].watch.cookie = notify_id; |
1349 | ops[0].watch.flag = 0; | |
1350 | ||
0ce1a794 | 1351 | ret = rbd_do_request(NULL, rbd_dev, NULL, CEPH_NOSNAP, |
7f0a24d8 | 1352 | rbd_dev->header_name, 0, 0, NULL, |
ad4f232f | 1353 | NULL, 0, |
59c2be1e YS |
1354 | CEPH_OSD_FLAG_READ, |
1355 | ops, | |
1fec7093 | 1356 | NULL, 0, |
59c2be1e YS |
1357 | rbd_simple_req_cb, 0, NULL); |
1358 | ||
1359 | rbd_destroy_ops(ops); | |
1360 | return ret; | |
1361 | } | |
1362 | ||
1363 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) | |
1364 | { | |
0ce1a794 | 1365 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
a71b891b | 1366 | u64 hver; |
13143d2d SW |
1367 | int rc; |
1368 | ||
0ce1a794 | 1369 | if (!rbd_dev) |
59c2be1e YS |
1370 | return; |
1371 | ||
bd919d45 AE |
1372 | dout("rbd_watch_cb %s notify_id=%llu opcode=%u\n", |
1373 | rbd_dev->header_name, (unsigned long long) notify_id, | |
1374 | (unsigned int) opcode); | |
117973fb | 1375 | rc = rbd_dev_refresh(rbd_dev, &hver); |
13143d2d | 1376 | if (rc) |
f0f8cef5 | 1377 | pr_warning(RBD_DRV_NAME "%d got notification but failed to " |
0ce1a794 | 1378 | " update snaps: %d\n", rbd_dev->major, rc); |
59c2be1e | 1379 | |
7f0a24d8 | 1380 | rbd_req_sync_notify_ack(rbd_dev, hver, notify_id); |
59c2be1e YS |
1381 | } |
1382 | ||
1383 | /* | |
1384 | * Request sync osd watch | |
1385 | */ | |
0e6f322d | 1386 | static int rbd_req_sync_watch(struct rbd_device *rbd_dev) |
59c2be1e YS |
1387 | { |
1388 | struct ceph_osd_req_op *ops; | |
0ce1a794 | 1389 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
57cfc106 | 1390 | int ret; |
59c2be1e | 1391 | |
57cfc106 AE |
1392 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_WATCH, 0); |
1393 | if (!ops) | |
1394 | return -ENOMEM; | |
59c2be1e YS |
1395 | |
1396 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, 0, | |
0ce1a794 | 1397 | (void *)rbd_dev, &rbd_dev->watch_event); |
59c2be1e YS |
1398 | if (ret < 0) |
1399 | goto fail; | |
1400 | ||
0e6f322d | 1401 | ops[0].watch.ver = cpu_to_le64(rbd_dev->header.obj_version); |
0ce1a794 | 1402 | ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); |
59c2be1e YS |
1403 | ops[0].watch.flag = 1; |
1404 | ||
0ce1a794 | 1405 | ret = rbd_req_sync_op(rbd_dev, NULL, |
59c2be1e | 1406 | CEPH_NOSNAP, |
59c2be1e YS |
1407 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
1408 | ops, | |
0e6f322d AE |
1409 | rbd_dev->header_name, |
1410 | 0, 0, NULL, | |
0ce1a794 | 1411 | &rbd_dev->watch_request, NULL); |
59c2be1e YS |
1412 | |
1413 | if (ret < 0) | |
1414 | goto fail_event; | |
1415 | ||
1416 | rbd_destroy_ops(ops); | |
1417 | return 0; | |
1418 | ||
1419 | fail_event: | |
0ce1a794 AE |
1420 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
1421 | rbd_dev->watch_event = NULL; | |
59c2be1e YS |
1422 | fail: |
1423 | rbd_destroy_ops(ops); | |
1424 | return ret; | |
1425 | } | |
1426 | ||
79e3057c YS |
1427 | /* |
1428 | * Request sync osd unwatch | |
1429 | */ | |
070c633f | 1430 | static int rbd_req_sync_unwatch(struct rbd_device *rbd_dev) |
79e3057c YS |
1431 | { |
1432 | struct ceph_osd_req_op *ops; | |
57cfc106 | 1433 | int ret; |
79e3057c | 1434 | |
57cfc106 AE |
1435 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_WATCH, 0); |
1436 | if (!ops) | |
1437 | return -ENOMEM; | |
79e3057c YS |
1438 | |
1439 | ops[0].watch.ver = 0; | |
0ce1a794 | 1440 | ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); |
79e3057c YS |
1441 | ops[0].watch.flag = 0; |
1442 | ||
0ce1a794 | 1443 | ret = rbd_req_sync_op(rbd_dev, NULL, |
79e3057c | 1444 | CEPH_NOSNAP, |
79e3057c YS |
1445 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
1446 | ops, | |
070c633f AE |
1447 | rbd_dev->header_name, |
1448 | 0, 0, NULL, NULL, NULL); | |
1449 | ||
79e3057c YS |
1450 | |
1451 | rbd_destroy_ops(ops); | |
0ce1a794 AE |
1452 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
1453 | rbd_dev->watch_event = NULL; | |
79e3057c YS |
1454 | return ret; |
1455 | } | |
1456 | ||
602adf40 | 1457 | /* |
3cb4a687 | 1458 | * Synchronous osd object method call |
602adf40 | 1459 | */ |
0ce1a794 | 1460 | static int rbd_req_sync_exec(struct rbd_device *rbd_dev, |
aded07ea AE |
1461 | const char *object_name, |
1462 | const char *class_name, | |
1463 | const char *method_name, | |
3cb4a687 AE |
1464 | const char *outbound, |
1465 | size_t outbound_size, | |
f8d4de6e AE |
1466 | char *inbound, |
1467 | size_t inbound_size, | |
3cb4a687 | 1468 | int flags, |
59c2be1e | 1469 | u64 *ver) |
602adf40 YS |
1470 | { |
1471 | struct ceph_osd_req_op *ops; | |
aded07ea AE |
1472 | int class_name_len = strlen(class_name); |
1473 | int method_name_len = strlen(method_name); | |
3cb4a687 | 1474 | int payload_size; |
57cfc106 AE |
1475 | int ret; |
1476 | ||
3cb4a687 AE |
1477 | /* |
1478 | * Any input parameters required by the method we're calling | |
1479 | * will be sent along with the class and method names as | |
1480 | * part of the message payload. That data and its size are | |
1481 | * supplied via the indata and indata_len fields (named from | |
1482 | * the perspective of the server side) in the OSD request | |
1483 | * operation. | |
1484 | */ | |
1485 | payload_size = class_name_len + method_name_len + outbound_size; | |
1486 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_CALL, payload_size); | |
57cfc106 AE |
1487 | if (!ops) |
1488 | return -ENOMEM; | |
602adf40 | 1489 | |
aded07ea AE |
1490 | ops[0].cls.class_name = class_name; |
1491 | ops[0].cls.class_len = (__u8) class_name_len; | |
1492 | ops[0].cls.method_name = method_name; | |
1493 | ops[0].cls.method_len = (__u8) method_name_len; | |
602adf40 | 1494 | ops[0].cls.argc = 0; |
3cb4a687 AE |
1495 | ops[0].cls.indata = outbound; |
1496 | ops[0].cls.indata_len = outbound_size; | |
602adf40 | 1497 | |
0ce1a794 | 1498 | ret = rbd_req_sync_op(rbd_dev, NULL, |
602adf40 | 1499 | CEPH_NOSNAP, |
3cb4a687 | 1500 | flags, ops, |
f8d4de6e AE |
1501 | object_name, 0, inbound_size, inbound, |
1502 | NULL, ver); | |
602adf40 YS |
1503 | |
1504 | rbd_destroy_ops(ops); | |
1505 | ||
1506 | dout("cls_exec returned %d\n", ret); | |
1507 | return ret; | |
1508 | } | |
1509 | ||
1fec7093 YS |
1510 | static struct rbd_req_coll *rbd_alloc_coll(int num_reqs) |
1511 | { | |
1512 | struct rbd_req_coll *coll = | |
1513 | kzalloc(sizeof(struct rbd_req_coll) + | |
1514 | sizeof(struct rbd_req_status) * num_reqs, | |
1515 | GFP_ATOMIC); | |
1516 | ||
1517 | if (!coll) | |
1518 | return NULL; | |
1519 | coll->total = num_reqs; | |
1520 | kref_init(&coll->kref); | |
1521 | return coll; | |
1522 | } | |
1523 | ||
602adf40 YS |
1524 | /* |
1525 | * block device queue callback | |
1526 | */ | |
1527 | static void rbd_rq_fn(struct request_queue *q) | |
1528 | { | |
1529 | struct rbd_device *rbd_dev = q->queuedata; | |
1530 | struct request *rq; | |
602adf40 | 1531 | |
00f1f36f | 1532 | while ((rq = blk_fetch_request(q))) { |
602adf40 | 1533 | struct bio *bio; |
602adf40 | 1534 | bool do_write; |
bd919d45 | 1535 | unsigned int size; |
602adf40 | 1536 | u64 ofs; |
1fec7093 YS |
1537 | int num_segs, cur_seg = 0; |
1538 | struct rbd_req_coll *coll; | |
d1d25646 | 1539 | struct ceph_snap_context *snapc; |
f7760dad | 1540 | unsigned int bio_offset; |
602adf40 | 1541 | |
602adf40 YS |
1542 | dout("fetched request\n"); |
1543 | ||
1544 | /* filter out block requests we don't understand */ | |
1545 | if ((rq->cmd_type != REQ_TYPE_FS)) { | |
1546 | __blk_end_request_all(rq, 0); | |
00f1f36f | 1547 | continue; |
602adf40 YS |
1548 | } |
1549 | ||
1550 | /* deduce our operation (read, write) */ | |
1551 | do_write = (rq_data_dir(rq) == WRITE); | |
f84344f3 | 1552 | if (do_write && rbd_dev->mapping.read_only) { |
602adf40 | 1553 | __blk_end_request_all(rq, -EROFS); |
00f1f36f | 1554 | continue; |
602adf40 YS |
1555 | } |
1556 | ||
1557 | spin_unlock_irq(q->queue_lock); | |
1558 | ||
d1d25646 | 1559 | down_read(&rbd_dev->header_rwsem); |
e88a36ec | 1560 | |
daba5fdb | 1561 | if (!rbd_dev->exists) { |
0d7dbfce | 1562 | rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); |
e88a36ec | 1563 | up_read(&rbd_dev->header_rwsem); |
d1d25646 JD |
1564 | dout("request for non-existent snapshot"); |
1565 | spin_lock_irq(q->queue_lock); | |
1566 | __blk_end_request_all(rq, -ENXIO); | |
1567 | continue; | |
e88a36ec JD |
1568 | } |
1569 | ||
d1d25646 JD |
1570 | snapc = ceph_get_snap_context(rbd_dev->header.snapc); |
1571 | ||
1572 | up_read(&rbd_dev->header_rwsem); | |
1573 | ||
f7760dad AE |
1574 | size = blk_rq_bytes(rq); |
1575 | ofs = blk_rq_pos(rq) * SECTOR_SIZE; | |
1576 | bio = rq->bio; | |
1577 | ||
602adf40 YS |
1578 | dout("%s 0x%x bytes at 0x%llx\n", |
1579 | do_write ? "write" : "read", | |
bd919d45 | 1580 | size, (unsigned long long) blk_rq_pos(rq) * SECTOR_SIZE); |
602adf40 | 1581 | |
1fec7093 | 1582 | num_segs = rbd_get_num_segments(&rbd_dev->header, ofs, size); |
df111be6 AE |
1583 | if (num_segs <= 0) { |
1584 | spin_lock_irq(q->queue_lock); | |
1585 | __blk_end_request_all(rq, num_segs); | |
1586 | ceph_put_snap_context(snapc); | |
1587 | continue; | |
1588 | } | |
1fec7093 YS |
1589 | coll = rbd_alloc_coll(num_segs); |
1590 | if (!coll) { | |
1591 | spin_lock_irq(q->queue_lock); | |
1592 | __blk_end_request_all(rq, -ENOMEM); | |
d1d25646 | 1593 | ceph_put_snap_context(snapc); |
00f1f36f | 1594 | continue; |
1fec7093 YS |
1595 | } |
1596 | ||
f7760dad | 1597 | bio_offset = 0; |
602adf40 | 1598 | do { |
f7760dad AE |
1599 | u64 limit = rbd_segment_length(rbd_dev, ofs, size); |
1600 | unsigned int chain_size; | |
1601 | struct bio *bio_chain; | |
1602 | ||
1603 | BUG_ON(limit > (u64) UINT_MAX); | |
1604 | chain_size = (unsigned int) limit; | |
bd919d45 | 1605 | dout("rq->bio->bi_vcnt=%hu\n", rq->bio->bi_vcnt); |
f7760dad | 1606 | |
1fec7093 | 1607 | kref_get(&coll->kref); |
f7760dad AE |
1608 | |
1609 | /* Pass a cloned bio chain via an osd request */ | |
1610 | ||
1611 | bio_chain = bio_chain_clone_range(&bio, | |
1612 | &bio_offset, chain_size, | |
1613 | GFP_ATOMIC); | |
1614 | if (bio_chain) | |
4634246d | 1615 | (void) rbd_do_op(rq, rbd_dev, snapc, |
f7760dad AE |
1616 | ofs, chain_size, |
1617 | bio_chain, coll, cur_seg); | |
4634246d | 1618 | else |
1fec7093 | 1619 | rbd_coll_end_req_index(rq, coll, cur_seg, |
f7760dad AE |
1620 | -ENOMEM, chain_size); |
1621 | size -= chain_size; | |
1622 | ofs += chain_size; | |
602adf40 | 1623 | |
1fec7093 | 1624 | cur_seg++; |
602adf40 | 1625 | } while (size > 0); |
1fec7093 | 1626 | kref_put(&coll->kref, rbd_coll_release); |
602adf40 | 1627 | |
602adf40 | 1628 | spin_lock_irq(q->queue_lock); |
d1d25646 JD |
1629 | |
1630 | ceph_put_snap_context(snapc); | |
602adf40 YS |
1631 | } |
1632 | } | |
1633 | ||
1634 | /* | |
1635 | * a queue callback. Makes sure that we don't create a bio that spans across | |
1636 | * multiple osd objects. One exception would be with a single page bios, | |
f7760dad | 1637 | * which we handle later at bio_chain_clone_range() |
602adf40 YS |
1638 | */ |
1639 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, | |
1640 | struct bio_vec *bvec) | |
1641 | { | |
1642 | struct rbd_device *rbd_dev = q->queuedata; | |
e5cfeed2 AE |
1643 | sector_t sector_offset; |
1644 | sector_t sectors_per_obj; | |
1645 | sector_t obj_sector_offset; | |
1646 | int ret; | |
1647 | ||
1648 | /* | |
1649 | * Find how far into its rbd object the partition-relative | |
1650 | * bio start sector is to offset relative to the enclosing | |
1651 | * device. | |
1652 | */ | |
1653 | sector_offset = get_start_sect(bmd->bi_bdev) + bmd->bi_sector; | |
1654 | sectors_per_obj = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); | |
1655 | obj_sector_offset = sector_offset & (sectors_per_obj - 1); | |
1656 | ||
1657 | /* | |
1658 | * Compute the number of bytes from that offset to the end | |
1659 | * of the object. Account for what's already used by the bio. | |
1660 | */ | |
1661 | ret = (int) (sectors_per_obj - obj_sector_offset) << SECTOR_SHIFT; | |
1662 | if (ret > bmd->bi_size) | |
1663 | ret -= bmd->bi_size; | |
1664 | else | |
1665 | ret = 0; | |
1666 | ||
1667 | /* | |
1668 | * Don't send back more than was asked for. And if the bio | |
1669 | * was empty, let the whole thing through because: "Note | |
1670 | * that a block device *must* allow a single page to be | |
1671 | * added to an empty bio." | |
1672 | */ | |
1673 | rbd_assert(bvec->bv_len <= PAGE_SIZE); | |
1674 | if (ret > (int) bvec->bv_len || !bmd->bi_size) | |
1675 | ret = (int) bvec->bv_len; | |
1676 | ||
1677 | return ret; | |
602adf40 YS |
1678 | } |
1679 | ||
1680 | static void rbd_free_disk(struct rbd_device *rbd_dev) | |
1681 | { | |
1682 | struct gendisk *disk = rbd_dev->disk; | |
1683 | ||
1684 | if (!disk) | |
1685 | return; | |
1686 | ||
602adf40 YS |
1687 | if (disk->flags & GENHD_FL_UP) |
1688 | del_gendisk(disk); | |
1689 | if (disk->queue) | |
1690 | blk_cleanup_queue(disk->queue); | |
1691 | put_disk(disk); | |
1692 | } | |
1693 | ||
1694 | /* | |
4156d998 AE |
1695 | * Read the complete header for the given rbd device. |
1696 | * | |
1697 | * Returns a pointer to a dynamically-allocated buffer containing | |
1698 | * the complete and validated header. Caller can pass the address | |
1699 | * of a variable that will be filled in with the version of the | |
1700 | * header object at the time it was read. | |
1701 | * | |
1702 | * Returns a pointer-coded errno if a failure occurs. | |
602adf40 | 1703 | */ |
4156d998 AE |
1704 | static struct rbd_image_header_ondisk * |
1705 | rbd_dev_v1_header_read(struct rbd_device *rbd_dev, u64 *version) | |
602adf40 | 1706 | { |
4156d998 | 1707 | struct rbd_image_header_ondisk *ondisk = NULL; |
50f7c4c9 | 1708 | u32 snap_count = 0; |
4156d998 AE |
1709 | u64 names_size = 0; |
1710 | u32 want_count; | |
1711 | int ret; | |
602adf40 | 1712 | |
00f1f36f | 1713 | /* |
4156d998 AE |
1714 | * The complete header will include an array of its 64-bit |
1715 | * snapshot ids, followed by the names of those snapshots as | |
1716 | * a contiguous block of NUL-terminated strings. Note that | |
1717 | * the number of snapshots could change by the time we read | |
1718 | * it in, in which case we re-read it. | |
00f1f36f | 1719 | */ |
4156d998 AE |
1720 | do { |
1721 | size_t size; | |
1722 | ||
1723 | kfree(ondisk); | |
1724 | ||
1725 | size = sizeof (*ondisk); | |
1726 | size += snap_count * sizeof (struct rbd_image_snap_ondisk); | |
1727 | size += names_size; | |
1728 | ondisk = kmalloc(size, GFP_KERNEL); | |
1729 | if (!ondisk) | |
1730 | return ERR_PTR(-ENOMEM); | |
1731 | ||
1732 | ret = rbd_req_sync_read(rbd_dev, CEPH_NOSNAP, | |
0bed54dc | 1733 | rbd_dev->header_name, |
4156d998 AE |
1734 | 0, size, |
1735 | (char *) ondisk, version); | |
1736 | ||
1737 | if (ret < 0) | |
1738 | goto out_err; | |
1739 | if (WARN_ON((size_t) ret < size)) { | |
1740 | ret = -ENXIO; | |
1741 | pr_warning("short header read for image %s" | |
1742 | " (want %zd got %d)\n", | |
0d7dbfce | 1743 | rbd_dev->spec->image_name, size, ret); |
4156d998 AE |
1744 | goto out_err; |
1745 | } | |
1746 | if (!rbd_dev_ondisk_valid(ondisk)) { | |
1747 | ret = -ENXIO; | |
1748 | pr_warning("invalid header for image %s\n", | |
0d7dbfce | 1749 | rbd_dev->spec->image_name); |
4156d998 | 1750 | goto out_err; |
81e759fb | 1751 | } |
602adf40 | 1752 | |
4156d998 AE |
1753 | names_size = le64_to_cpu(ondisk->snap_names_len); |
1754 | want_count = snap_count; | |
1755 | snap_count = le32_to_cpu(ondisk->snap_count); | |
1756 | } while (snap_count != want_count); | |
00f1f36f | 1757 | |
4156d998 | 1758 | return ondisk; |
00f1f36f | 1759 | |
4156d998 AE |
1760 | out_err: |
1761 | kfree(ondisk); | |
1762 | ||
1763 | return ERR_PTR(ret); | |
1764 | } | |
1765 | ||
1766 | /* | |
1767 | * reload the ondisk the header | |
1768 | */ | |
1769 | static int rbd_read_header(struct rbd_device *rbd_dev, | |
1770 | struct rbd_image_header *header) | |
1771 | { | |
1772 | struct rbd_image_header_ondisk *ondisk; | |
1773 | u64 ver = 0; | |
1774 | int ret; | |
602adf40 | 1775 | |
4156d998 AE |
1776 | ondisk = rbd_dev_v1_header_read(rbd_dev, &ver); |
1777 | if (IS_ERR(ondisk)) | |
1778 | return PTR_ERR(ondisk); | |
1779 | ret = rbd_header_from_disk(header, ondisk); | |
1780 | if (ret >= 0) | |
1781 | header->obj_version = ver; | |
1782 | kfree(ondisk); | |
1783 | ||
1784 | return ret; | |
602adf40 YS |
1785 | } |
1786 | ||
41f38c2b | 1787 | static void rbd_remove_all_snaps(struct rbd_device *rbd_dev) |
dfc5606d YS |
1788 | { |
1789 | struct rbd_snap *snap; | |
a0593290 | 1790 | struct rbd_snap *next; |
dfc5606d | 1791 | |
a0593290 | 1792 | list_for_each_entry_safe(snap, next, &rbd_dev->snaps, node) |
41f38c2b | 1793 | rbd_remove_snap_dev(snap); |
dfc5606d YS |
1794 | } |
1795 | ||
9478554a AE |
1796 | static void rbd_update_mapping_size(struct rbd_device *rbd_dev) |
1797 | { | |
1798 | sector_t size; | |
1799 | ||
0d7dbfce | 1800 | if (rbd_dev->spec->snap_id != CEPH_NOSNAP) |
9478554a AE |
1801 | return; |
1802 | ||
1803 | size = (sector_t) rbd_dev->header.image_size / SECTOR_SIZE; | |
1804 | dout("setting size to %llu sectors", (unsigned long long) size); | |
1805 | rbd_dev->mapping.size = (u64) size; | |
1806 | set_capacity(rbd_dev->disk, size); | |
1807 | } | |
1808 | ||
602adf40 YS |
1809 | /* |
1810 | * only read the first part of the ondisk header, without the snaps info | |
1811 | */ | |
117973fb | 1812 | static int rbd_dev_v1_refresh(struct rbd_device *rbd_dev, u64 *hver) |
602adf40 YS |
1813 | { |
1814 | int ret; | |
1815 | struct rbd_image_header h; | |
602adf40 YS |
1816 | |
1817 | ret = rbd_read_header(rbd_dev, &h); | |
1818 | if (ret < 0) | |
1819 | return ret; | |
1820 | ||
a51aa0c0 JD |
1821 | down_write(&rbd_dev->header_rwsem); |
1822 | ||
9478554a AE |
1823 | /* Update image size, and check for resize of mapped image */ |
1824 | rbd_dev->header.image_size = h.image_size; | |
1825 | rbd_update_mapping_size(rbd_dev); | |
9db4b3e3 | 1826 | |
849b4260 | 1827 | /* rbd_dev->header.object_prefix shouldn't change */ |
602adf40 | 1828 | kfree(rbd_dev->header.snap_sizes); |
849b4260 | 1829 | kfree(rbd_dev->header.snap_names); |
d1d25646 JD |
1830 | /* osd requests may still refer to snapc */ |
1831 | ceph_put_snap_context(rbd_dev->header.snapc); | |
602adf40 | 1832 | |
b813623a AE |
1833 | if (hver) |
1834 | *hver = h.obj_version; | |
a71b891b | 1835 | rbd_dev->header.obj_version = h.obj_version; |
93a24e08 | 1836 | rbd_dev->header.image_size = h.image_size; |
602adf40 YS |
1837 | rbd_dev->header.snapc = h.snapc; |
1838 | rbd_dev->header.snap_names = h.snap_names; | |
1839 | rbd_dev->header.snap_sizes = h.snap_sizes; | |
849b4260 AE |
1840 | /* Free the extra copy of the object prefix */ |
1841 | WARN_ON(strcmp(rbd_dev->header.object_prefix, h.object_prefix)); | |
1842 | kfree(h.object_prefix); | |
1843 | ||
304f6808 AE |
1844 | ret = rbd_dev_snaps_update(rbd_dev); |
1845 | if (!ret) | |
1846 | ret = rbd_dev_snaps_register(rbd_dev); | |
dfc5606d | 1847 | |
c666601a | 1848 | up_write(&rbd_dev->header_rwsem); |
602adf40 | 1849 | |
dfc5606d | 1850 | return ret; |
602adf40 YS |
1851 | } |
1852 | ||
117973fb | 1853 | static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver) |
1fe5e993 AE |
1854 | { |
1855 | int ret; | |
1856 | ||
117973fb | 1857 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
1fe5e993 | 1858 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
117973fb AE |
1859 | if (rbd_dev->image_format == 1) |
1860 | ret = rbd_dev_v1_refresh(rbd_dev, hver); | |
1861 | else | |
1862 | ret = rbd_dev_v2_refresh(rbd_dev, hver); | |
1fe5e993 AE |
1863 | mutex_unlock(&ctl_mutex); |
1864 | ||
1865 | return ret; | |
1866 | } | |
1867 | ||
602adf40 YS |
1868 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
1869 | { | |
1870 | struct gendisk *disk; | |
1871 | struct request_queue *q; | |
593a9e7b | 1872 | u64 segment_size; |
602adf40 | 1873 | |
602adf40 | 1874 | /* create gendisk info */ |
602adf40 YS |
1875 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); |
1876 | if (!disk) | |
1fcdb8aa | 1877 | return -ENOMEM; |
602adf40 | 1878 | |
f0f8cef5 | 1879 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
de71a297 | 1880 | rbd_dev->dev_id); |
602adf40 YS |
1881 | disk->major = rbd_dev->major; |
1882 | disk->first_minor = 0; | |
1883 | disk->fops = &rbd_bd_ops; | |
1884 | disk->private_data = rbd_dev; | |
1885 | ||
1886 | /* init rq */ | |
602adf40 YS |
1887 | q = blk_init_queue(rbd_rq_fn, &rbd_dev->lock); |
1888 | if (!q) | |
1889 | goto out_disk; | |
029bcbd8 | 1890 | |
593a9e7b AE |
1891 | /* We use the default size, but let's be explicit about it. */ |
1892 | blk_queue_physical_block_size(q, SECTOR_SIZE); | |
1893 | ||
029bcbd8 | 1894 | /* set io sizes to object size */ |
593a9e7b AE |
1895 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
1896 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); | |
1897 | blk_queue_max_segment_size(q, segment_size); | |
1898 | blk_queue_io_min(q, segment_size); | |
1899 | blk_queue_io_opt(q, segment_size); | |
029bcbd8 | 1900 | |
602adf40 YS |
1901 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
1902 | disk->queue = q; | |
1903 | ||
1904 | q->queuedata = rbd_dev; | |
1905 | ||
1906 | rbd_dev->disk = disk; | |
602adf40 | 1907 | |
12f02944 AE |
1908 | set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); |
1909 | ||
602adf40 | 1910 | return 0; |
602adf40 YS |
1911 | out_disk: |
1912 | put_disk(disk); | |
1fcdb8aa AE |
1913 | |
1914 | return -ENOMEM; | |
602adf40 YS |
1915 | } |
1916 | ||
dfc5606d YS |
1917 | /* |
1918 | sysfs | |
1919 | */ | |
1920 | ||
593a9e7b AE |
1921 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
1922 | { | |
1923 | return container_of(dev, struct rbd_device, dev); | |
1924 | } | |
1925 | ||
dfc5606d YS |
1926 | static ssize_t rbd_size_show(struct device *dev, |
1927 | struct device_attribute *attr, char *buf) | |
1928 | { | |
593a9e7b | 1929 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
a51aa0c0 JD |
1930 | sector_t size; |
1931 | ||
1932 | down_read(&rbd_dev->header_rwsem); | |
1933 | size = get_capacity(rbd_dev->disk); | |
1934 | up_read(&rbd_dev->header_rwsem); | |
dfc5606d | 1935 | |
a51aa0c0 | 1936 | return sprintf(buf, "%llu\n", (unsigned long long) size * SECTOR_SIZE); |
dfc5606d YS |
1937 | } |
1938 | ||
34b13184 AE |
1939 | /* |
1940 | * Note this shows the features for whatever's mapped, which is not | |
1941 | * necessarily the base image. | |
1942 | */ | |
1943 | static ssize_t rbd_features_show(struct device *dev, | |
1944 | struct device_attribute *attr, char *buf) | |
1945 | { | |
1946 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
1947 | ||
1948 | return sprintf(buf, "0x%016llx\n", | |
1949 | (unsigned long long) rbd_dev->mapping.features); | |
1950 | } | |
1951 | ||
dfc5606d YS |
1952 | static ssize_t rbd_major_show(struct device *dev, |
1953 | struct device_attribute *attr, char *buf) | |
1954 | { | |
593a9e7b | 1955 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 1956 | |
dfc5606d YS |
1957 | return sprintf(buf, "%d\n", rbd_dev->major); |
1958 | } | |
1959 | ||
1960 | static ssize_t rbd_client_id_show(struct device *dev, | |
1961 | struct device_attribute *attr, char *buf) | |
602adf40 | 1962 | { |
593a9e7b | 1963 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 1964 | |
1dbb4399 AE |
1965 | return sprintf(buf, "client%lld\n", |
1966 | ceph_client_id(rbd_dev->rbd_client->client)); | |
602adf40 YS |
1967 | } |
1968 | ||
dfc5606d YS |
1969 | static ssize_t rbd_pool_show(struct device *dev, |
1970 | struct device_attribute *attr, char *buf) | |
602adf40 | 1971 | { |
593a9e7b | 1972 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 1973 | |
0d7dbfce | 1974 | return sprintf(buf, "%s\n", rbd_dev->spec->pool_name); |
dfc5606d YS |
1975 | } |
1976 | ||
9bb2f334 AE |
1977 | static ssize_t rbd_pool_id_show(struct device *dev, |
1978 | struct device_attribute *attr, char *buf) | |
1979 | { | |
1980 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
1981 | ||
0d7dbfce AE |
1982 | return sprintf(buf, "%llu\n", |
1983 | (unsigned long long) rbd_dev->spec->pool_id); | |
9bb2f334 AE |
1984 | } |
1985 | ||
dfc5606d YS |
1986 | static ssize_t rbd_name_show(struct device *dev, |
1987 | struct device_attribute *attr, char *buf) | |
1988 | { | |
593a9e7b | 1989 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 1990 | |
0d7dbfce | 1991 | return sprintf(buf, "%s\n", rbd_dev->spec->image_name); |
dfc5606d YS |
1992 | } |
1993 | ||
589d30e0 AE |
1994 | static ssize_t rbd_image_id_show(struct device *dev, |
1995 | struct device_attribute *attr, char *buf) | |
1996 | { | |
1997 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
1998 | ||
0d7dbfce | 1999 | return sprintf(buf, "%s\n", rbd_dev->spec->image_id); |
589d30e0 AE |
2000 | } |
2001 | ||
34b13184 AE |
2002 | /* |
2003 | * Shows the name of the currently-mapped snapshot (or | |
2004 | * RBD_SNAP_HEAD_NAME for the base image). | |
2005 | */ | |
dfc5606d YS |
2006 | static ssize_t rbd_snap_show(struct device *dev, |
2007 | struct device_attribute *attr, | |
2008 | char *buf) | |
2009 | { | |
593a9e7b | 2010 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 2011 | |
0d7dbfce | 2012 | return sprintf(buf, "%s\n", rbd_dev->spec->snap_name); |
dfc5606d YS |
2013 | } |
2014 | ||
2015 | static ssize_t rbd_image_refresh(struct device *dev, | |
2016 | struct device_attribute *attr, | |
2017 | const char *buf, | |
2018 | size_t size) | |
2019 | { | |
593a9e7b | 2020 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
b813623a | 2021 | int ret; |
602adf40 | 2022 | |
117973fb | 2023 | ret = rbd_dev_refresh(rbd_dev, NULL); |
b813623a AE |
2024 | |
2025 | return ret < 0 ? ret : size; | |
dfc5606d | 2026 | } |
602adf40 | 2027 | |
dfc5606d | 2028 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
34b13184 | 2029 | static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL); |
dfc5606d YS |
2030 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); |
2031 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); | |
2032 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); | |
9bb2f334 | 2033 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
dfc5606d | 2034 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
589d30e0 | 2035 | static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL); |
dfc5606d YS |
2036 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); |
2037 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); | |
dfc5606d YS |
2038 | |
2039 | static struct attribute *rbd_attrs[] = { | |
2040 | &dev_attr_size.attr, | |
34b13184 | 2041 | &dev_attr_features.attr, |
dfc5606d YS |
2042 | &dev_attr_major.attr, |
2043 | &dev_attr_client_id.attr, | |
2044 | &dev_attr_pool.attr, | |
9bb2f334 | 2045 | &dev_attr_pool_id.attr, |
dfc5606d | 2046 | &dev_attr_name.attr, |
589d30e0 | 2047 | &dev_attr_image_id.attr, |
dfc5606d YS |
2048 | &dev_attr_current_snap.attr, |
2049 | &dev_attr_refresh.attr, | |
dfc5606d YS |
2050 | NULL |
2051 | }; | |
2052 | ||
2053 | static struct attribute_group rbd_attr_group = { | |
2054 | .attrs = rbd_attrs, | |
2055 | }; | |
2056 | ||
2057 | static const struct attribute_group *rbd_attr_groups[] = { | |
2058 | &rbd_attr_group, | |
2059 | NULL | |
2060 | }; | |
2061 | ||
2062 | static void rbd_sysfs_dev_release(struct device *dev) | |
2063 | { | |
2064 | } | |
2065 | ||
2066 | static struct device_type rbd_device_type = { | |
2067 | .name = "rbd", | |
2068 | .groups = rbd_attr_groups, | |
2069 | .release = rbd_sysfs_dev_release, | |
2070 | }; | |
2071 | ||
2072 | ||
2073 | /* | |
2074 | sysfs - snapshots | |
2075 | */ | |
2076 | ||
2077 | static ssize_t rbd_snap_size_show(struct device *dev, | |
2078 | struct device_attribute *attr, | |
2079 | char *buf) | |
2080 | { | |
2081 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2082 | ||
3591538f | 2083 | return sprintf(buf, "%llu\n", (unsigned long long)snap->size); |
dfc5606d YS |
2084 | } |
2085 | ||
2086 | static ssize_t rbd_snap_id_show(struct device *dev, | |
2087 | struct device_attribute *attr, | |
2088 | char *buf) | |
2089 | { | |
2090 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2091 | ||
3591538f | 2092 | return sprintf(buf, "%llu\n", (unsigned long long)snap->id); |
dfc5606d YS |
2093 | } |
2094 | ||
34b13184 AE |
2095 | static ssize_t rbd_snap_features_show(struct device *dev, |
2096 | struct device_attribute *attr, | |
2097 | char *buf) | |
2098 | { | |
2099 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2100 | ||
2101 | return sprintf(buf, "0x%016llx\n", | |
2102 | (unsigned long long) snap->features); | |
2103 | } | |
2104 | ||
dfc5606d YS |
2105 | static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); |
2106 | static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); | |
34b13184 | 2107 | static DEVICE_ATTR(snap_features, S_IRUGO, rbd_snap_features_show, NULL); |
dfc5606d YS |
2108 | |
2109 | static struct attribute *rbd_snap_attrs[] = { | |
2110 | &dev_attr_snap_size.attr, | |
2111 | &dev_attr_snap_id.attr, | |
34b13184 | 2112 | &dev_attr_snap_features.attr, |
dfc5606d YS |
2113 | NULL, |
2114 | }; | |
2115 | ||
2116 | static struct attribute_group rbd_snap_attr_group = { | |
2117 | .attrs = rbd_snap_attrs, | |
2118 | }; | |
2119 | ||
2120 | static void rbd_snap_dev_release(struct device *dev) | |
2121 | { | |
2122 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2123 | kfree(snap->name); | |
2124 | kfree(snap); | |
2125 | } | |
2126 | ||
2127 | static const struct attribute_group *rbd_snap_attr_groups[] = { | |
2128 | &rbd_snap_attr_group, | |
2129 | NULL | |
2130 | }; | |
2131 | ||
2132 | static struct device_type rbd_snap_device_type = { | |
2133 | .groups = rbd_snap_attr_groups, | |
2134 | .release = rbd_snap_dev_release, | |
2135 | }; | |
2136 | ||
8b8fb99c AE |
2137 | static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec) |
2138 | { | |
2139 | kref_get(&spec->kref); | |
2140 | ||
2141 | return spec; | |
2142 | } | |
2143 | ||
2144 | static void rbd_spec_free(struct kref *kref); | |
2145 | static void rbd_spec_put(struct rbd_spec *spec) | |
2146 | { | |
2147 | if (spec) | |
2148 | kref_put(&spec->kref, rbd_spec_free); | |
2149 | } | |
2150 | ||
2151 | static struct rbd_spec *rbd_spec_alloc(void) | |
2152 | { | |
2153 | struct rbd_spec *spec; | |
2154 | ||
2155 | spec = kzalloc(sizeof (*spec), GFP_KERNEL); | |
2156 | if (!spec) | |
2157 | return NULL; | |
2158 | kref_init(&spec->kref); | |
2159 | ||
2160 | rbd_spec_put(rbd_spec_get(spec)); /* TEMPORARY */ | |
2161 | ||
2162 | return spec; | |
2163 | } | |
2164 | ||
2165 | static void rbd_spec_free(struct kref *kref) | |
2166 | { | |
2167 | struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref); | |
2168 | ||
2169 | kfree(spec->pool_name); | |
2170 | kfree(spec->image_id); | |
2171 | kfree(spec->image_name); | |
2172 | kfree(spec->snap_name); | |
2173 | kfree(spec); | |
2174 | } | |
2175 | ||
304f6808 AE |
2176 | static bool rbd_snap_registered(struct rbd_snap *snap) |
2177 | { | |
2178 | bool ret = snap->dev.type == &rbd_snap_device_type; | |
2179 | bool reg = device_is_registered(&snap->dev); | |
2180 | ||
2181 | rbd_assert(!ret ^ reg); | |
2182 | ||
2183 | return ret; | |
2184 | } | |
2185 | ||
41f38c2b | 2186 | static void rbd_remove_snap_dev(struct rbd_snap *snap) |
dfc5606d YS |
2187 | { |
2188 | list_del(&snap->node); | |
304f6808 AE |
2189 | if (device_is_registered(&snap->dev)) |
2190 | device_unregister(&snap->dev); | |
dfc5606d YS |
2191 | } |
2192 | ||
14e7085d | 2193 | static int rbd_register_snap_dev(struct rbd_snap *snap, |
dfc5606d YS |
2194 | struct device *parent) |
2195 | { | |
2196 | struct device *dev = &snap->dev; | |
2197 | int ret; | |
2198 | ||
2199 | dev->type = &rbd_snap_device_type; | |
2200 | dev->parent = parent; | |
2201 | dev->release = rbd_snap_dev_release; | |
d4b125e9 | 2202 | dev_set_name(dev, "%s%s", RBD_SNAP_DEV_NAME_PREFIX, snap->name); |
304f6808 AE |
2203 | dout("%s: registering device for snapshot %s\n", __func__, snap->name); |
2204 | ||
dfc5606d YS |
2205 | ret = device_register(dev); |
2206 | ||
2207 | return ret; | |
2208 | } | |
2209 | ||
4e891e0a | 2210 | static struct rbd_snap *__rbd_add_snap_dev(struct rbd_device *rbd_dev, |
c8d18425 | 2211 | const char *snap_name, |
34b13184 AE |
2212 | u64 snap_id, u64 snap_size, |
2213 | u64 snap_features) | |
dfc5606d | 2214 | { |
4e891e0a | 2215 | struct rbd_snap *snap; |
dfc5606d | 2216 | int ret; |
4e891e0a AE |
2217 | |
2218 | snap = kzalloc(sizeof (*snap), GFP_KERNEL); | |
dfc5606d | 2219 | if (!snap) |
4e891e0a AE |
2220 | return ERR_PTR(-ENOMEM); |
2221 | ||
2222 | ret = -ENOMEM; | |
c8d18425 | 2223 | snap->name = kstrdup(snap_name, GFP_KERNEL); |
4e891e0a AE |
2224 | if (!snap->name) |
2225 | goto err; | |
2226 | ||
c8d18425 AE |
2227 | snap->id = snap_id; |
2228 | snap->size = snap_size; | |
34b13184 | 2229 | snap->features = snap_features; |
4e891e0a AE |
2230 | |
2231 | return snap; | |
2232 | ||
dfc5606d YS |
2233 | err: |
2234 | kfree(snap->name); | |
2235 | kfree(snap); | |
4e891e0a AE |
2236 | |
2237 | return ERR_PTR(ret); | |
dfc5606d YS |
2238 | } |
2239 | ||
cd892126 AE |
2240 | static char *rbd_dev_v1_snap_info(struct rbd_device *rbd_dev, u32 which, |
2241 | u64 *snap_size, u64 *snap_features) | |
2242 | { | |
2243 | char *snap_name; | |
2244 | ||
2245 | rbd_assert(which < rbd_dev->header.snapc->num_snaps); | |
2246 | ||
2247 | *snap_size = rbd_dev->header.snap_sizes[which]; | |
2248 | *snap_features = 0; /* No features for v1 */ | |
2249 | ||
2250 | /* Skip over names until we find the one we are looking for */ | |
2251 | ||
2252 | snap_name = rbd_dev->header.snap_names; | |
2253 | while (which--) | |
2254 | snap_name += strlen(snap_name) + 1; | |
2255 | ||
2256 | return snap_name; | |
2257 | } | |
2258 | ||
9d475de5 AE |
2259 | /* |
2260 | * Get the size and object order for an image snapshot, or if | |
2261 | * snap_id is CEPH_NOSNAP, gets this information for the base | |
2262 | * image. | |
2263 | */ | |
2264 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, | |
2265 | u8 *order, u64 *snap_size) | |
2266 | { | |
2267 | __le64 snapid = cpu_to_le64(snap_id); | |
2268 | int ret; | |
2269 | struct { | |
2270 | u8 order; | |
2271 | __le64 size; | |
2272 | } __attribute__ ((packed)) size_buf = { 0 }; | |
2273 | ||
2274 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2275 | "rbd", "get_size", | |
2276 | (char *) &snapid, sizeof (snapid), | |
2277 | (char *) &size_buf, sizeof (size_buf), | |
2278 | CEPH_OSD_FLAG_READ, NULL); | |
2279 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2280 | if (ret < 0) | |
2281 | return ret; | |
2282 | ||
2283 | *order = size_buf.order; | |
2284 | *snap_size = le64_to_cpu(size_buf.size); | |
2285 | ||
2286 | dout(" snap_id 0x%016llx order = %u, snap_size = %llu\n", | |
2287 | (unsigned long long) snap_id, (unsigned int) *order, | |
2288 | (unsigned long long) *snap_size); | |
2289 | ||
2290 | return 0; | |
2291 | } | |
2292 | ||
2293 | static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) | |
2294 | { | |
2295 | return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, | |
2296 | &rbd_dev->header.obj_order, | |
2297 | &rbd_dev->header.image_size); | |
2298 | } | |
2299 | ||
1e130199 AE |
2300 | static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) |
2301 | { | |
2302 | void *reply_buf; | |
2303 | int ret; | |
2304 | void *p; | |
2305 | ||
2306 | reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL); | |
2307 | if (!reply_buf) | |
2308 | return -ENOMEM; | |
2309 | ||
2310 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2311 | "rbd", "get_object_prefix", | |
2312 | NULL, 0, | |
2313 | reply_buf, RBD_OBJ_PREFIX_LEN_MAX, | |
2314 | CEPH_OSD_FLAG_READ, NULL); | |
2315 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2316 | if (ret < 0) | |
2317 | goto out; | |
a0ea3a40 | 2318 | ret = 0; /* rbd_req_sync_exec() can return positive */ |
1e130199 AE |
2319 | |
2320 | p = reply_buf; | |
2321 | rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, | |
2322 | p + RBD_OBJ_PREFIX_LEN_MAX, | |
2323 | NULL, GFP_NOIO); | |
2324 | ||
2325 | if (IS_ERR(rbd_dev->header.object_prefix)) { | |
2326 | ret = PTR_ERR(rbd_dev->header.object_prefix); | |
2327 | rbd_dev->header.object_prefix = NULL; | |
2328 | } else { | |
2329 | dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); | |
2330 | } | |
2331 | ||
2332 | out: | |
2333 | kfree(reply_buf); | |
2334 | ||
2335 | return ret; | |
2336 | } | |
2337 | ||
b1b5402a AE |
2338 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
2339 | u64 *snap_features) | |
2340 | { | |
2341 | __le64 snapid = cpu_to_le64(snap_id); | |
2342 | struct { | |
2343 | __le64 features; | |
2344 | __le64 incompat; | |
2345 | } features_buf = { 0 }; | |
d889140c | 2346 | u64 incompat; |
b1b5402a AE |
2347 | int ret; |
2348 | ||
2349 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2350 | "rbd", "get_features", | |
2351 | (char *) &snapid, sizeof (snapid), | |
2352 | (char *) &features_buf, sizeof (features_buf), | |
2353 | CEPH_OSD_FLAG_READ, NULL); | |
2354 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2355 | if (ret < 0) | |
2356 | return ret; | |
d889140c AE |
2357 | |
2358 | incompat = le64_to_cpu(features_buf.incompat); | |
2359 | if (incompat & ~RBD_FEATURES_ALL) | |
2360 | return -ENOTSUPP; | |
2361 | ||
b1b5402a AE |
2362 | *snap_features = le64_to_cpu(features_buf.features); |
2363 | ||
2364 | dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", | |
2365 | (unsigned long long) snap_id, | |
2366 | (unsigned long long) *snap_features, | |
2367 | (unsigned long long) le64_to_cpu(features_buf.incompat)); | |
2368 | ||
2369 | return 0; | |
2370 | } | |
2371 | ||
2372 | static int rbd_dev_v2_features(struct rbd_device *rbd_dev) | |
2373 | { | |
2374 | return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, | |
2375 | &rbd_dev->header.features); | |
2376 | } | |
2377 | ||
6e14b1a6 | 2378 | static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev, u64 *ver) |
35d489f9 AE |
2379 | { |
2380 | size_t size; | |
2381 | int ret; | |
2382 | void *reply_buf; | |
2383 | void *p; | |
2384 | void *end; | |
2385 | u64 seq; | |
2386 | u32 snap_count; | |
2387 | struct ceph_snap_context *snapc; | |
2388 | u32 i; | |
2389 | ||
2390 | /* | |
2391 | * We'll need room for the seq value (maximum snapshot id), | |
2392 | * snapshot count, and array of that many snapshot ids. | |
2393 | * For now we have a fixed upper limit on the number we're | |
2394 | * prepared to receive. | |
2395 | */ | |
2396 | size = sizeof (__le64) + sizeof (__le32) + | |
2397 | RBD_MAX_SNAP_COUNT * sizeof (__le64); | |
2398 | reply_buf = kzalloc(size, GFP_KERNEL); | |
2399 | if (!reply_buf) | |
2400 | return -ENOMEM; | |
2401 | ||
2402 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2403 | "rbd", "get_snapcontext", | |
2404 | NULL, 0, | |
2405 | reply_buf, size, | |
6e14b1a6 | 2406 | CEPH_OSD_FLAG_READ, ver); |
35d489f9 AE |
2407 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); |
2408 | if (ret < 0) | |
2409 | goto out; | |
2410 | ||
2411 | ret = -ERANGE; | |
2412 | p = reply_buf; | |
2413 | end = (char *) reply_buf + size; | |
2414 | ceph_decode_64_safe(&p, end, seq, out); | |
2415 | ceph_decode_32_safe(&p, end, snap_count, out); | |
2416 | ||
2417 | /* | |
2418 | * Make sure the reported number of snapshot ids wouldn't go | |
2419 | * beyond the end of our buffer. But before checking that, | |
2420 | * make sure the computed size of the snapshot context we | |
2421 | * allocate is representable in a size_t. | |
2422 | */ | |
2423 | if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) | |
2424 | / sizeof (u64)) { | |
2425 | ret = -EINVAL; | |
2426 | goto out; | |
2427 | } | |
2428 | if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) | |
2429 | goto out; | |
2430 | ||
2431 | size = sizeof (struct ceph_snap_context) + | |
2432 | snap_count * sizeof (snapc->snaps[0]); | |
2433 | snapc = kmalloc(size, GFP_KERNEL); | |
2434 | if (!snapc) { | |
2435 | ret = -ENOMEM; | |
2436 | goto out; | |
2437 | } | |
2438 | ||
2439 | atomic_set(&snapc->nref, 1); | |
2440 | snapc->seq = seq; | |
2441 | snapc->num_snaps = snap_count; | |
2442 | for (i = 0; i < snap_count; i++) | |
2443 | snapc->snaps[i] = ceph_decode_64(&p); | |
2444 | ||
2445 | rbd_dev->header.snapc = snapc; | |
2446 | ||
2447 | dout(" snap context seq = %llu, snap_count = %u\n", | |
2448 | (unsigned long long) seq, (unsigned int) snap_count); | |
2449 | ||
2450 | out: | |
2451 | kfree(reply_buf); | |
2452 | ||
2453 | return 0; | |
2454 | } | |
2455 | ||
b8b1e2db AE |
2456 | static char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, u32 which) |
2457 | { | |
2458 | size_t size; | |
2459 | void *reply_buf; | |
2460 | __le64 snap_id; | |
2461 | int ret; | |
2462 | void *p; | |
2463 | void *end; | |
b8b1e2db AE |
2464 | char *snap_name; |
2465 | ||
2466 | size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; | |
2467 | reply_buf = kmalloc(size, GFP_KERNEL); | |
2468 | if (!reply_buf) | |
2469 | return ERR_PTR(-ENOMEM); | |
2470 | ||
2471 | snap_id = cpu_to_le64(rbd_dev->header.snapc->snaps[which]); | |
2472 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2473 | "rbd", "get_snapshot_name", | |
2474 | (char *) &snap_id, sizeof (snap_id), | |
2475 | reply_buf, size, | |
2476 | CEPH_OSD_FLAG_READ, NULL); | |
2477 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2478 | if (ret < 0) | |
2479 | goto out; | |
2480 | ||
2481 | p = reply_buf; | |
2482 | end = (char *) reply_buf + size; | |
e5c35534 | 2483 | snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
b8b1e2db AE |
2484 | if (IS_ERR(snap_name)) { |
2485 | ret = PTR_ERR(snap_name); | |
2486 | goto out; | |
2487 | } else { | |
2488 | dout(" snap_id 0x%016llx snap_name = %s\n", | |
2489 | (unsigned long long) le64_to_cpu(snap_id), snap_name); | |
2490 | } | |
2491 | kfree(reply_buf); | |
2492 | ||
2493 | return snap_name; | |
2494 | out: | |
2495 | kfree(reply_buf); | |
2496 | ||
2497 | return ERR_PTR(ret); | |
2498 | } | |
2499 | ||
2500 | static char *rbd_dev_v2_snap_info(struct rbd_device *rbd_dev, u32 which, | |
2501 | u64 *snap_size, u64 *snap_features) | |
2502 | { | |
2503 | __le64 snap_id; | |
2504 | u8 order; | |
2505 | int ret; | |
2506 | ||
2507 | snap_id = rbd_dev->header.snapc->snaps[which]; | |
2508 | ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, &order, snap_size); | |
2509 | if (ret) | |
2510 | return ERR_PTR(ret); | |
2511 | ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, snap_features); | |
2512 | if (ret) | |
2513 | return ERR_PTR(ret); | |
2514 | ||
2515 | return rbd_dev_v2_snap_name(rbd_dev, which); | |
2516 | } | |
2517 | ||
2518 | static char *rbd_dev_snap_info(struct rbd_device *rbd_dev, u32 which, | |
2519 | u64 *snap_size, u64 *snap_features) | |
2520 | { | |
2521 | if (rbd_dev->image_format == 1) | |
2522 | return rbd_dev_v1_snap_info(rbd_dev, which, | |
2523 | snap_size, snap_features); | |
2524 | if (rbd_dev->image_format == 2) | |
2525 | return rbd_dev_v2_snap_info(rbd_dev, which, | |
2526 | snap_size, snap_features); | |
2527 | return ERR_PTR(-EINVAL); | |
2528 | } | |
2529 | ||
117973fb AE |
2530 | static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver) |
2531 | { | |
2532 | int ret; | |
2533 | __u8 obj_order; | |
2534 | ||
2535 | down_write(&rbd_dev->header_rwsem); | |
2536 | ||
2537 | /* Grab old order first, to see if it changes */ | |
2538 | ||
2539 | obj_order = rbd_dev->header.obj_order, | |
2540 | ret = rbd_dev_v2_image_size(rbd_dev); | |
2541 | if (ret) | |
2542 | goto out; | |
2543 | if (rbd_dev->header.obj_order != obj_order) { | |
2544 | ret = -EIO; | |
2545 | goto out; | |
2546 | } | |
2547 | rbd_update_mapping_size(rbd_dev); | |
2548 | ||
2549 | ret = rbd_dev_v2_snap_context(rbd_dev, hver); | |
2550 | dout("rbd_dev_v2_snap_context returned %d\n", ret); | |
2551 | if (ret) | |
2552 | goto out; | |
2553 | ret = rbd_dev_snaps_update(rbd_dev); | |
2554 | dout("rbd_dev_snaps_update returned %d\n", ret); | |
2555 | if (ret) | |
2556 | goto out; | |
2557 | ret = rbd_dev_snaps_register(rbd_dev); | |
2558 | dout("rbd_dev_snaps_register returned %d\n", ret); | |
2559 | out: | |
2560 | up_write(&rbd_dev->header_rwsem); | |
2561 | ||
2562 | return ret; | |
2563 | } | |
2564 | ||
dfc5606d | 2565 | /* |
35938150 AE |
2566 | * Scan the rbd device's current snapshot list and compare it to the |
2567 | * newly-received snapshot context. Remove any existing snapshots | |
2568 | * not present in the new snapshot context. Add a new snapshot for | |
2569 | * any snaphots in the snapshot context not in the current list. | |
2570 | * And verify there are no changes to snapshots we already know | |
2571 | * about. | |
2572 | * | |
2573 | * Assumes the snapshots in the snapshot context are sorted by | |
2574 | * snapshot id, highest id first. (Snapshots in the rbd_dev's list | |
2575 | * are also maintained in that order.) | |
dfc5606d | 2576 | */ |
304f6808 | 2577 | static int rbd_dev_snaps_update(struct rbd_device *rbd_dev) |
dfc5606d | 2578 | { |
35938150 AE |
2579 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
2580 | const u32 snap_count = snapc->num_snaps; | |
35938150 AE |
2581 | struct list_head *head = &rbd_dev->snaps; |
2582 | struct list_head *links = head->next; | |
2583 | u32 index = 0; | |
dfc5606d | 2584 | |
9fcbb800 | 2585 | dout("%s: snap count is %u\n", __func__, (unsigned int) snap_count); |
35938150 AE |
2586 | while (index < snap_count || links != head) { |
2587 | u64 snap_id; | |
2588 | struct rbd_snap *snap; | |
cd892126 AE |
2589 | char *snap_name; |
2590 | u64 snap_size = 0; | |
2591 | u64 snap_features = 0; | |
dfc5606d | 2592 | |
35938150 AE |
2593 | snap_id = index < snap_count ? snapc->snaps[index] |
2594 | : CEPH_NOSNAP; | |
2595 | snap = links != head ? list_entry(links, struct rbd_snap, node) | |
2596 | : NULL; | |
aafb230e | 2597 | rbd_assert(!snap || snap->id != CEPH_NOSNAP); |
dfc5606d | 2598 | |
35938150 AE |
2599 | if (snap_id == CEPH_NOSNAP || (snap && snap->id > snap_id)) { |
2600 | struct list_head *next = links->next; | |
dfc5606d | 2601 | |
35938150 | 2602 | /* Existing snapshot not in the new snap context */ |
dfc5606d | 2603 | |
0d7dbfce | 2604 | if (rbd_dev->spec->snap_id == snap->id) |
daba5fdb | 2605 | rbd_dev->exists = false; |
41f38c2b | 2606 | rbd_remove_snap_dev(snap); |
9fcbb800 | 2607 | dout("%ssnap id %llu has been removed\n", |
0d7dbfce AE |
2608 | rbd_dev->spec->snap_id == snap->id ? |
2609 | "mapped " : "", | |
9fcbb800 | 2610 | (unsigned long long) snap->id); |
35938150 AE |
2611 | |
2612 | /* Done with this list entry; advance */ | |
2613 | ||
2614 | links = next; | |
dfc5606d YS |
2615 | continue; |
2616 | } | |
35938150 | 2617 | |
b8b1e2db AE |
2618 | snap_name = rbd_dev_snap_info(rbd_dev, index, |
2619 | &snap_size, &snap_features); | |
cd892126 AE |
2620 | if (IS_ERR(snap_name)) |
2621 | return PTR_ERR(snap_name); | |
2622 | ||
9fcbb800 AE |
2623 | dout("entry %u: snap_id = %llu\n", (unsigned int) snap_count, |
2624 | (unsigned long long) snap_id); | |
35938150 AE |
2625 | if (!snap || (snap_id != CEPH_NOSNAP && snap->id < snap_id)) { |
2626 | struct rbd_snap *new_snap; | |
2627 | ||
2628 | /* We haven't seen this snapshot before */ | |
2629 | ||
c8d18425 | 2630 | new_snap = __rbd_add_snap_dev(rbd_dev, snap_name, |
cd892126 | 2631 | snap_id, snap_size, snap_features); |
9fcbb800 AE |
2632 | if (IS_ERR(new_snap)) { |
2633 | int err = PTR_ERR(new_snap); | |
2634 | ||
2635 | dout(" failed to add dev, error %d\n", err); | |
2636 | ||
2637 | return err; | |
2638 | } | |
35938150 AE |
2639 | |
2640 | /* New goes before existing, or at end of list */ | |
2641 | ||
9fcbb800 | 2642 | dout(" added dev%s\n", snap ? "" : " at end\n"); |
35938150 AE |
2643 | if (snap) |
2644 | list_add_tail(&new_snap->node, &snap->node); | |
2645 | else | |
523f3258 | 2646 | list_add_tail(&new_snap->node, head); |
35938150 AE |
2647 | } else { |
2648 | /* Already have this one */ | |
2649 | ||
9fcbb800 AE |
2650 | dout(" already present\n"); |
2651 | ||
cd892126 | 2652 | rbd_assert(snap->size == snap_size); |
aafb230e | 2653 | rbd_assert(!strcmp(snap->name, snap_name)); |
cd892126 | 2654 | rbd_assert(snap->features == snap_features); |
35938150 AE |
2655 | |
2656 | /* Done with this list entry; advance */ | |
2657 | ||
2658 | links = links->next; | |
dfc5606d | 2659 | } |
35938150 AE |
2660 | |
2661 | /* Advance to the next entry in the snapshot context */ | |
2662 | ||
2663 | index++; | |
dfc5606d | 2664 | } |
9fcbb800 | 2665 | dout("%s: done\n", __func__); |
dfc5606d YS |
2666 | |
2667 | return 0; | |
2668 | } | |
2669 | ||
304f6808 AE |
2670 | /* |
2671 | * Scan the list of snapshots and register the devices for any that | |
2672 | * have not already been registered. | |
2673 | */ | |
2674 | static int rbd_dev_snaps_register(struct rbd_device *rbd_dev) | |
2675 | { | |
2676 | struct rbd_snap *snap; | |
2677 | int ret = 0; | |
2678 | ||
2679 | dout("%s called\n", __func__); | |
86ff77bb AE |
2680 | if (WARN_ON(!device_is_registered(&rbd_dev->dev))) |
2681 | return -EIO; | |
304f6808 AE |
2682 | |
2683 | list_for_each_entry(snap, &rbd_dev->snaps, node) { | |
2684 | if (!rbd_snap_registered(snap)) { | |
2685 | ret = rbd_register_snap_dev(snap, &rbd_dev->dev); | |
2686 | if (ret < 0) | |
2687 | break; | |
2688 | } | |
2689 | } | |
2690 | dout("%s: returning %d\n", __func__, ret); | |
2691 | ||
2692 | return ret; | |
2693 | } | |
2694 | ||
dfc5606d YS |
2695 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
2696 | { | |
dfc5606d | 2697 | struct device *dev; |
cd789ab9 | 2698 | int ret; |
dfc5606d YS |
2699 | |
2700 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
dfc5606d | 2701 | |
cd789ab9 | 2702 | dev = &rbd_dev->dev; |
dfc5606d YS |
2703 | dev->bus = &rbd_bus_type; |
2704 | dev->type = &rbd_device_type; | |
2705 | dev->parent = &rbd_root_dev; | |
2706 | dev->release = rbd_dev_release; | |
de71a297 | 2707 | dev_set_name(dev, "%d", rbd_dev->dev_id); |
dfc5606d | 2708 | ret = device_register(dev); |
dfc5606d | 2709 | |
dfc5606d | 2710 | mutex_unlock(&ctl_mutex); |
cd789ab9 | 2711 | |
dfc5606d | 2712 | return ret; |
602adf40 YS |
2713 | } |
2714 | ||
dfc5606d YS |
2715 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
2716 | { | |
2717 | device_unregister(&rbd_dev->dev); | |
2718 | } | |
2719 | ||
59c2be1e YS |
2720 | static int rbd_init_watch_dev(struct rbd_device *rbd_dev) |
2721 | { | |
2722 | int ret, rc; | |
2723 | ||
2724 | do { | |
0e6f322d | 2725 | ret = rbd_req_sync_watch(rbd_dev); |
59c2be1e | 2726 | if (ret == -ERANGE) { |
117973fb | 2727 | rc = rbd_dev_refresh(rbd_dev, NULL); |
59c2be1e YS |
2728 | if (rc < 0) |
2729 | return rc; | |
2730 | } | |
2731 | } while (ret == -ERANGE); | |
2732 | ||
2733 | return ret; | |
2734 | } | |
2735 | ||
e2839308 | 2736 | static atomic64_t rbd_dev_id_max = ATOMIC64_INIT(0); |
1ddbe94e AE |
2737 | |
2738 | /* | |
499afd5b AE |
2739 | * Get a unique rbd identifier for the given new rbd_dev, and add |
2740 | * the rbd_dev to the global list. The minimum rbd id is 1. | |
1ddbe94e | 2741 | */ |
e2839308 | 2742 | static void rbd_dev_id_get(struct rbd_device *rbd_dev) |
b7f23c36 | 2743 | { |
e2839308 | 2744 | rbd_dev->dev_id = atomic64_inc_return(&rbd_dev_id_max); |
499afd5b AE |
2745 | |
2746 | spin_lock(&rbd_dev_list_lock); | |
2747 | list_add_tail(&rbd_dev->node, &rbd_dev_list); | |
2748 | spin_unlock(&rbd_dev_list_lock); | |
e2839308 AE |
2749 | dout("rbd_dev %p given dev id %llu\n", rbd_dev, |
2750 | (unsigned long long) rbd_dev->dev_id); | |
1ddbe94e | 2751 | } |
b7f23c36 | 2752 | |
1ddbe94e | 2753 | /* |
499afd5b AE |
2754 | * Remove an rbd_dev from the global list, and record that its |
2755 | * identifier is no longer in use. | |
1ddbe94e | 2756 | */ |
e2839308 | 2757 | static void rbd_dev_id_put(struct rbd_device *rbd_dev) |
1ddbe94e | 2758 | { |
d184f6bf | 2759 | struct list_head *tmp; |
de71a297 | 2760 | int rbd_id = rbd_dev->dev_id; |
d184f6bf AE |
2761 | int max_id; |
2762 | ||
aafb230e | 2763 | rbd_assert(rbd_id > 0); |
499afd5b | 2764 | |
e2839308 AE |
2765 | dout("rbd_dev %p released dev id %llu\n", rbd_dev, |
2766 | (unsigned long long) rbd_dev->dev_id); | |
499afd5b AE |
2767 | spin_lock(&rbd_dev_list_lock); |
2768 | list_del_init(&rbd_dev->node); | |
d184f6bf AE |
2769 | |
2770 | /* | |
2771 | * If the id being "put" is not the current maximum, there | |
2772 | * is nothing special we need to do. | |
2773 | */ | |
e2839308 | 2774 | if (rbd_id != atomic64_read(&rbd_dev_id_max)) { |
d184f6bf AE |
2775 | spin_unlock(&rbd_dev_list_lock); |
2776 | return; | |
2777 | } | |
2778 | ||
2779 | /* | |
2780 | * We need to update the current maximum id. Search the | |
2781 | * list to find out what it is. We're more likely to find | |
2782 | * the maximum at the end, so search the list backward. | |
2783 | */ | |
2784 | max_id = 0; | |
2785 | list_for_each_prev(tmp, &rbd_dev_list) { | |
2786 | struct rbd_device *rbd_dev; | |
2787 | ||
2788 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
b213e0b1 AE |
2789 | if (rbd_dev->dev_id > max_id) |
2790 | max_id = rbd_dev->dev_id; | |
d184f6bf | 2791 | } |
499afd5b | 2792 | spin_unlock(&rbd_dev_list_lock); |
b7f23c36 | 2793 | |
1ddbe94e | 2794 | /* |
e2839308 | 2795 | * The max id could have been updated by rbd_dev_id_get(), in |
d184f6bf AE |
2796 | * which case it now accurately reflects the new maximum. |
2797 | * Be careful not to overwrite the maximum value in that | |
2798 | * case. | |
1ddbe94e | 2799 | */ |
e2839308 AE |
2800 | atomic64_cmpxchg(&rbd_dev_id_max, rbd_id, max_id); |
2801 | dout(" max dev id has been reset\n"); | |
b7f23c36 AE |
2802 | } |
2803 | ||
e28fff26 AE |
2804 | /* |
2805 | * Skips over white space at *buf, and updates *buf to point to the | |
2806 | * first found non-space character (if any). Returns the length of | |
593a9e7b AE |
2807 | * the token (string of non-white space characters) found. Note |
2808 | * that *buf must be terminated with '\0'. | |
e28fff26 AE |
2809 | */ |
2810 | static inline size_t next_token(const char **buf) | |
2811 | { | |
2812 | /* | |
2813 | * These are the characters that produce nonzero for | |
2814 | * isspace() in the "C" and "POSIX" locales. | |
2815 | */ | |
2816 | const char *spaces = " \f\n\r\t\v"; | |
2817 | ||
2818 | *buf += strspn(*buf, spaces); /* Find start of token */ | |
2819 | ||
2820 | return strcspn(*buf, spaces); /* Return token length */ | |
2821 | } | |
2822 | ||
2823 | /* | |
2824 | * Finds the next token in *buf, and if the provided token buffer is | |
2825 | * big enough, copies the found token into it. The result, if | |
593a9e7b AE |
2826 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
2827 | * must be terminated with '\0' on entry. | |
e28fff26 AE |
2828 | * |
2829 | * Returns the length of the token found (not including the '\0'). | |
2830 | * Return value will be 0 if no token is found, and it will be >= | |
2831 | * token_size if the token would not fit. | |
2832 | * | |
593a9e7b | 2833 | * The *buf pointer will be updated to point beyond the end of the |
e28fff26 AE |
2834 | * found token. Note that this occurs even if the token buffer is |
2835 | * too small to hold it. | |
2836 | */ | |
2837 | static inline size_t copy_token(const char **buf, | |
2838 | char *token, | |
2839 | size_t token_size) | |
2840 | { | |
2841 | size_t len; | |
2842 | ||
2843 | len = next_token(buf); | |
2844 | if (len < token_size) { | |
2845 | memcpy(token, *buf, len); | |
2846 | *(token + len) = '\0'; | |
2847 | } | |
2848 | *buf += len; | |
2849 | ||
2850 | return len; | |
2851 | } | |
2852 | ||
ea3352f4 AE |
2853 | /* |
2854 | * Finds the next token in *buf, dynamically allocates a buffer big | |
2855 | * enough to hold a copy of it, and copies the token into the new | |
2856 | * buffer. The copy is guaranteed to be terminated with '\0'. Note | |
2857 | * that a duplicate buffer is created even for a zero-length token. | |
2858 | * | |
2859 | * Returns a pointer to the newly-allocated duplicate, or a null | |
2860 | * pointer if memory for the duplicate was not available. If | |
2861 | * the lenp argument is a non-null pointer, the length of the token | |
2862 | * (not including the '\0') is returned in *lenp. | |
2863 | * | |
2864 | * If successful, the *buf pointer will be updated to point beyond | |
2865 | * the end of the found token. | |
2866 | * | |
2867 | * Note: uses GFP_KERNEL for allocation. | |
2868 | */ | |
2869 | static inline char *dup_token(const char **buf, size_t *lenp) | |
2870 | { | |
2871 | char *dup; | |
2872 | size_t len; | |
2873 | ||
2874 | len = next_token(buf); | |
2875 | dup = kmalloc(len + 1, GFP_KERNEL); | |
2876 | if (!dup) | |
2877 | return NULL; | |
2878 | ||
2879 | memcpy(dup, *buf, len); | |
2880 | *(dup + len) = '\0'; | |
2881 | *buf += len; | |
2882 | ||
2883 | if (lenp) | |
2884 | *lenp = len; | |
2885 | ||
2886 | return dup; | |
2887 | } | |
2888 | ||
a725f65e | 2889 | /* |
3feeb894 AE |
2890 | * This fills in the pool_name, image_name, image_name_len, rbd_dev, |
2891 | * rbd_md_name, and name fields of the given rbd_dev, based on the | |
2892 | * list of monitor addresses and other options provided via | |
2893 | * /sys/bus/rbd/add. Returns a pointer to a dynamically-allocated | |
2894 | * copy of the snapshot name to map if successful, or a | |
2895 | * pointer-coded error otherwise. | |
d22f76e7 AE |
2896 | * |
2897 | * Note: rbd_dev is assumed to have been initially zero-filled. | |
a725f65e | 2898 | */ |
dc79b113 AE |
2899 | static int rbd_add_parse_args(struct rbd_device *rbd_dev, |
2900 | const char *buf, | |
2901 | struct ceph_options **ceph_opts, | |
2902 | struct rbd_options **opts) | |
e28fff26 | 2903 | { |
d22f76e7 | 2904 | size_t len; |
0ddebc0c AE |
2905 | const char *mon_addrs; |
2906 | size_t mon_addrs_size; | |
f28e565a | 2907 | char *options; |
4e9afeba | 2908 | struct rbd_options *rbd_opts = NULL; |
dc79b113 | 2909 | int ret; |
e28fff26 AE |
2910 | |
2911 | /* The first four tokens are required */ | |
2912 | ||
7ef3214a AE |
2913 | len = next_token(&buf); |
2914 | if (!len) | |
dc79b113 | 2915 | return -EINVAL; /* Missing monitor address(es) */ |
0ddebc0c | 2916 | mon_addrs = buf; |
f28e565a | 2917 | mon_addrs_size = len + 1; |
7ef3214a | 2918 | buf += len; |
a725f65e | 2919 | |
dc79b113 | 2920 | ret = -EINVAL; |
f28e565a AE |
2921 | options = dup_token(&buf, NULL); |
2922 | if (!options) | |
dc79b113 | 2923 | return -ENOMEM; |
f28e565a AE |
2924 | if (!*options) |
2925 | goto out_err; /* Missing options */ | |
e28fff26 | 2926 | |
0d7dbfce AE |
2927 | rbd_dev->spec->pool_name = dup_token(&buf, NULL); |
2928 | if (!rbd_dev->spec->pool_name) | |
f28e565a | 2929 | goto out_mem; |
0d7dbfce | 2930 | if (!*rbd_dev->spec->pool_name) |
f28e565a | 2931 | goto out_err; /* Missing pool name */ |
e28fff26 | 2932 | |
0d7dbfce AE |
2933 | rbd_dev->spec->image_name = |
2934 | dup_token(&buf, &rbd_dev->spec->image_name_len); | |
2935 | if (!rbd_dev->spec->image_name) | |
f28e565a | 2936 | goto out_mem; |
0d7dbfce | 2937 | if (!*rbd_dev->spec->image_name) |
f28e565a | 2938 | goto out_err; /* Missing image name */ |
d4b125e9 | 2939 | |
f28e565a AE |
2940 | /* |
2941 | * Snapshot name is optional; default is to use "-" | |
2942 | * (indicating the head/no snapshot). | |
2943 | */ | |
3feeb894 | 2944 | len = next_token(&buf); |
820a5f3e | 2945 | if (!len) { |
3feeb894 AE |
2946 | buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ |
2947 | len = sizeof (RBD_SNAP_HEAD_NAME) - 1; | |
f28e565a | 2948 | } else if (len > RBD_MAX_SNAP_NAME_LEN) { |
dc79b113 | 2949 | ret = -ENAMETOOLONG; |
f28e565a | 2950 | goto out_err; |
849b4260 | 2951 | } |
0d7dbfce AE |
2952 | rbd_dev->spec->snap_name = kmalloc(len + 1, GFP_KERNEL); |
2953 | if (!rbd_dev->spec->snap_name) | |
f28e565a | 2954 | goto out_mem; |
0d7dbfce AE |
2955 | memcpy(rbd_dev->spec->snap_name, buf, len); |
2956 | *(rbd_dev->spec->snap_name + len) = '\0'; | |
e5c35534 | 2957 | |
0ddebc0c | 2958 | /* Initialize all rbd options to the defaults */ |
e28fff26 | 2959 | |
4e9afeba AE |
2960 | rbd_opts = kzalloc(sizeof (*rbd_opts), GFP_KERNEL); |
2961 | if (!rbd_opts) | |
2962 | goto out_mem; | |
2963 | ||
2964 | rbd_opts->read_only = RBD_READ_ONLY_DEFAULT; | |
d22f76e7 | 2965 | |
dc79b113 | 2966 | *ceph_opts = ceph_parse_options(options, mon_addrs, |
0ddebc0c | 2967 | mon_addrs + mon_addrs_size - 1, |
4e9afeba | 2968 | parse_rbd_opts_token, rbd_opts); |
f28e565a | 2969 | kfree(options); |
dc79b113 AE |
2970 | if (IS_ERR(*ceph_opts)) { |
2971 | ret = PTR_ERR(*ceph_opts); | |
2972 | goto out_err; | |
2973 | } | |
4e9afeba | 2974 | *opts = rbd_opts; |
0ddebc0c | 2975 | |
dc79b113 | 2976 | return 0; |
f28e565a | 2977 | out_mem: |
dc79b113 | 2978 | ret = -ENOMEM; |
d22f76e7 | 2979 | out_err: |
0d7dbfce AE |
2980 | kfree(rbd_dev->spec->image_name); |
2981 | rbd_dev->spec->image_name = NULL; | |
2982 | rbd_dev->spec->image_name_len = 0; | |
2983 | kfree(rbd_dev->spec->pool_name); | |
2984 | rbd_dev->spec->pool_name = NULL; | |
f28e565a | 2985 | kfree(options); |
d22f76e7 | 2986 | |
dc79b113 | 2987 | return ret; |
a725f65e AE |
2988 | } |
2989 | ||
589d30e0 AE |
2990 | /* |
2991 | * An rbd format 2 image has a unique identifier, distinct from the | |
2992 | * name given to it by the user. Internally, that identifier is | |
2993 | * what's used to specify the names of objects related to the image. | |
2994 | * | |
2995 | * A special "rbd id" object is used to map an rbd image name to its | |
2996 | * id. If that object doesn't exist, then there is no v2 rbd image | |
2997 | * with the supplied name. | |
2998 | * | |
2999 | * This function will record the given rbd_dev's image_id field if | |
3000 | * it can be determined, and in that case will return 0. If any | |
3001 | * errors occur a negative errno will be returned and the rbd_dev's | |
3002 | * image_id field will be unchanged (and should be NULL). | |
3003 | */ | |
3004 | static int rbd_dev_image_id(struct rbd_device *rbd_dev) | |
3005 | { | |
3006 | int ret; | |
3007 | size_t size; | |
3008 | char *object_name; | |
3009 | void *response; | |
3010 | void *p; | |
3011 | ||
3012 | /* | |
3013 | * First, see if the format 2 image id file exists, and if | |
3014 | * so, get the image's persistent id from it. | |
3015 | */ | |
0d7dbfce | 3016 | size = sizeof (RBD_ID_PREFIX) + rbd_dev->spec->image_name_len; |
589d30e0 AE |
3017 | object_name = kmalloc(size, GFP_NOIO); |
3018 | if (!object_name) | |
3019 | return -ENOMEM; | |
0d7dbfce | 3020 | sprintf(object_name, "%s%s", RBD_ID_PREFIX, rbd_dev->spec->image_name); |
589d30e0 AE |
3021 | dout("rbd id object name is %s\n", object_name); |
3022 | ||
3023 | /* Response will be an encoded string, which includes a length */ | |
3024 | ||
3025 | size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; | |
3026 | response = kzalloc(size, GFP_NOIO); | |
3027 | if (!response) { | |
3028 | ret = -ENOMEM; | |
3029 | goto out; | |
3030 | } | |
3031 | ||
3032 | ret = rbd_req_sync_exec(rbd_dev, object_name, | |
3033 | "rbd", "get_id", | |
3034 | NULL, 0, | |
3035 | response, RBD_IMAGE_ID_LEN_MAX, | |
3036 | CEPH_OSD_FLAG_READ, NULL); | |
3037 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
3038 | if (ret < 0) | |
3039 | goto out; | |
a0ea3a40 | 3040 | ret = 0; /* rbd_req_sync_exec() can return positive */ |
589d30e0 AE |
3041 | |
3042 | p = response; | |
0d7dbfce | 3043 | rbd_dev->spec->image_id = ceph_extract_encoded_string(&p, |
589d30e0 | 3044 | p + RBD_IMAGE_ID_LEN_MAX, |
0d7dbfce | 3045 | &rbd_dev->spec->image_id_len, |
589d30e0 | 3046 | GFP_NOIO); |
0d7dbfce AE |
3047 | if (IS_ERR(rbd_dev->spec->image_id)) { |
3048 | ret = PTR_ERR(rbd_dev->spec->image_id); | |
3049 | rbd_dev->spec->image_id = NULL; | |
589d30e0 | 3050 | } else { |
0d7dbfce | 3051 | dout("image_id is %s\n", rbd_dev->spec->image_id); |
589d30e0 AE |
3052 | } |
3053 | out: | |
3054 | kfree(response); | |
3055 | kfree(object_name); | |
3056 | ||
3057 | return ret; | |
3058 | } | |
3059 | ||
a30b71b9 AE |
3060 | static int rbd_dev_v1_probe(struct rbd_device *rbd_dev) |
3061 | { | |
3062 | int ret; | |
3063 | size_t size; | |
3064 | ||
3065 | /* Version 1 images have no id; empty string is used */ | |
3066 | ||
0d7dbfce AE |
3067 | rbd_dev->spec->image_id = kstrdup("", GFP_KERNEL); |
3068 | if (!rbd_dev->spec->image_id) | |
a30b71b9 | 3069 | return -ENOMEM; |
0d7dbfce | 3070 | rbd_dev->spec->image_id_len = 0; |
a30b71b9 AE |
3071 | |
3072 | /* Record the header object name for this rbd image. */ | |
3073 | ||
0d7dbfce | 3074 | size = rbd_dev->spec->image_name_len + sizeof (RBD_SUFFIX); |
a30b71b9 AE |
3075 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); |
3076 | if (!rbd_dev->header_name) { | |
3077 | ret = -ENOMEM; | |
3078 | goto out_err; | |
3079 | } | |
0d7dbfce AE |
3080 | sprintf(rbd_dev->header_name, "%s%s", |
3081 | rbd_dev->spec->image_name, RBD_SUFFIX); | |
a30b71b9 AE |
3082 | |
3083 | /* Populate rbd image metadata */ | |
3084 | ||
3085 | ret = rbd_read_header(rbd_dev, &rbd_dev->header); | |
3086 | if (ret < 0) | |
3087 | goto out_err; | |
3088 | rbd_dev->image_format = 1; | |
3089 | ||
3090 | dout("discovered version 1 image, header name is %s\n", | |
3091 | rbd_dev->header_name); | |
3092 | ||
3093 | return 0; | |
3094 | ||
3095 | out_err: | |
3096 | kfree(rbd_dev->header_name); | |
3097 | rbd_dev->header_name = NULL; | |
0d7dbfce AE |
3098 | kfree(rbd_dev->spec->image_id); |
3099 | rbd_dev->spec->image_id = NULL; | |
a30b71b9 AE |
3100 | |
3101 | return ret; | |
3102 | } | |
3103 | ||
3104 | static int rbd_dev_v2_probe(struct rbd_device *rbd_dev) | |
3105 | { | |
3106 | size_t size; | |
9d475de5 | 3107 | int ret; |
6e14b1a6 | 3108 | u64 ver = 0; |
a30b71b9 AE |
3109 | |
3110 | /* | |
3111 | * Image id was filled in by the caller. Record the header | |
3112 | * object name for this rbd image. | |
3113 | */ | |
0d7dbfce | 3114 | size = sizeof (RBD_HEADER_PREFIX) + rbd_dev->spec->image_id_len; |
a30b71b9 AE |
3115 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); |
3116 | if (!rbd_dev->header_name) | |
3117 | return -ENOMEM; | |
3118 | sprintf(rbd_dev->header_name, "%s%s", | |
0d7dbfce | 3119 | RBD_HEADER_PREFIX, rbd_dev->spec->image_id); |
9d475de5 AE |
3120 | |
3121 | /* Get the size and object order for the image */ | |
3122 | ||
3123 | ret = rbd_dev_v2_image_size(rbd_dev); | |
1e130199 AE |
3124 | if (ret < 0) |
3125 | goto out_err; | |
3126 | ||
3127 | /* Get the object prefix (a.k.a. block_name) for the image */ | |
3128 | ||
3129 | ret = rbd_dev_v2_object_prefix(rbd_dev); | |
b1b5402a AE |
3130 | if (ret < 0) |
3131 | goto out_err; | |
3132 | ||
d889140c | 3133 | /* Get the and check features for the image */ |
b1b5402a AE |
3134 | |
3135 | ret = rbd_dev_v2_features(rbd_dev); | |
9d475de5 AE |
3136 | if (ret < 0) |
3137 | goto out_err; | |
35d489f9 | 3138 | |
6e14b1a6 AE |
3139 | /* crypto and compression type aren't (yet) supported for v2 images */ |
3140 | ||
3141 | rbd_dev->header.crypt_type = 0; | |
3142 | rbd_dev->header.comp_type = 0; | |
35d489f9 | 3143 | |
6e14b1a6 AE |
3144 | /* Get the snapshot context, plus the header version */ |
3145 | ||
3146 | ret = rbd_dev_v2_snap_context(rbd_dev, &ver); | |
35d489f9 AE |
3147 | if (ret) |
3148 | goto out_err; | |
6e14b1a6 AE |
3149 | rbd_dev->header.obj_version = ver; |
3150 | ||
a30b71b9 AE |
3151 | rbd_dev->image_format = 2; |
3152 | ||
3153 | dout("discovered version 2 image, header name is %s\n", | |
3154 | rbd_dev->header_name); | |
3155 | ||
35152979 | 3156 | return 0; |
9d475de5 AE |
3157 | out_err: |
3158 | kfree(rbd_dev->header_name); | |
3159 | rbd_dev->header_name = NULL; | |
1e130199 AE |
3160 | kfree(rbd_dev->header.object_prefix); |
3161 | rbd_dev->header.object_prefix = NULL; | |
9d475de5 AE |
3162 | |
3163 | return ret; | |
a30b71b9 AE |
3164 | } |
3165 | ||
3166 | /* | |
3167 | * Probe for the existence of the header object for the given rbd | |
3168 | * device. For format 2 images this includes determining the image | |
3169 | * id. | |
3170 | */ | |
3171 | static int rbd_dev_probe(struct rbd_device *rbd_dev) | |
3172 | { | |
3173 | int ret; | |
3174 | ||
3175 | /* | |
3176 | * Get the id from the image id object. If it's not a | |
3177 | * format 2 image, we'll get ENOENT back, and we'll assume | |
3178 | * it's a format 1 image. | |
3179 | */ | |
3180 | ret = rbd_dev_image_id(rbd_dev); | |
3181 | if (ret) | |
3182 | ret = rbd_dev_v1_probe(rbd_dev); | |
3183 | else | |
3184 | ret = rbd_dev_v2_probe(rbd_dev); | |
3185 | if (ret) | |
3186 | dout("probe failed, returning %d\n", ret); | |
3187 | ||
3188 | return ret; | |
3189 | } | |
3190 | ||
59c2be1e YS |
3191 | static ssize_t rbd_add(struct bus_type *bus, |
3192 | const char *buf, | |
3193 | size_t count) | |
602adf40 | 3194 | { |
cb8627c7 | 3195 | struct rbd_device *rbd_dev = NULL; |
dc79b113 | 3196 | struct ceph_options *ceph_opts = NULL; |
4e9afeba | 3197 | struct rbd_options *rbd_opts = NULL; |
27cc2594 AE |
3198 | struct ceph_osd_client *osdc; |
3199 | int rc = -ENOMEM; | |
602adf40 YS |
3200 | |
3201 | if (!try_module_get(THIS_MODULE)) | |
3202 | return -ENODEV; | |
3203 | ||
cb8627c7 AE |
3204 | rbd_dev = kzalloc(sizeof(*rbd_dev), GFP_KERNEL); |
3205 | if (!rbd_dev) | |
4e9afeba | 3206 | return -ENOMEM; |
8b8fb99c | 3207 | rbd_dev->spec = rbd_spec_alloc(); |
0d7dbfce AE |
3208 | if (!rbd_dev->spec) |
3209 | goto err_out_mem; | |
602adf40 YS |
3210 | |
3211 | /* static rbd_device initialization */ | |
3212 | spin_lock_init(&rbd_dev->lock); | |
3213 | INIT_LIST_HEAD(&rbd_dev->node); | |
dfc5606d | 3214 | INIT_LIST_HEAD(&rbd_dev->snaps); |
c666601a | 3215 | init_rwsem(&rbd_dev->header_rwsem); |
602adf40 | 3216 | |
602adf40 | 3217 | /* parse add command */ |
dc79b113 AE |
3218 | rc = rbd_add_parse_args(rbd_dev, buf, &ceph_opts, &rbd_opts); |
3219 | if (rc < 0) | |
0ddebc0c | 3220 | goto err_out_mem; |
4e9afeba | 3221 | rbd_dev->mapping.read_only = rbd_opts->read_only; |
78cea76e | 3222 | |
78cea76e AE |
3223 | rc = rbd_get_client(rbd_dev, ceph_opts); |
3224 | if (rc < 0) | |
0ddebc0c | 3225 | goto err_out_args; |
78cea76e | 3226 | ceph_opts = NULL; /* ceph_opts now owned by rbd_dev client */ |
602adf40 | 3227 | |
602adf40 | 3228 | /* pick the pool */ |
1dbb4399 | 3229 | osdc = &rbd_dev->rbd_client->client->osdc; |
0d7dbfce | 3230 | rc = ceph_pg_poolid_by_name(osdc->osdmap, rbd_dev->spec->pool_name); |
602adf40 YS |
3231 | if (rc < 0) |
3232 | goto err_out_client; | |
0d7dbfce | 3233 | rbd_dev->spec->pool_id = (u64) rc; |
602adf40 | 3234 | |
a30b71b9 AE |
3235 | rc = rbd_dev_probe(rbd_dev); |
3236 | if (rc < 0) | |
05fd6f6f AE |
3237 | goto err_out_client; |
3238 | ||
3239 | /* no need to lock here, as rbd_dev is not registered yet */ | |
3240 | rc = rbd_dev_snaps_update(rbd_dev); | |
3241 | if (rc) | |
41f38c2b | 3242 | goto err_out_probe; |
05fd6f6f | 3243 | |
819d52bf | 3244 | rc = rbd_dev_set_mapping(rbd_dev); |
05fd6f6f | 3245 | if (rc) |
41f38c2b | 3246 | goto err_out_snaps; |
05fd6f6f | 3247 | |
85ae8926 AE |
3248 | /* generate unique id: find highest unique id, add one */ |
3249 | rbd_dev_id_get(rbd_dev); | |
3250 | ||
3251 | /* Fill in the device name, now that we have its id. */ | |
3252 | BUILD_BUG_ON(DEV_NAME_LEN | |
3253 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); | |
3254 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); | |
3255 | ||
3256 | /* Get our block major device number. */ | |
3257 | ||
27cc2594 AE |
3258 | rc = register_blkdev(0, rbd_dev->name); |
3259 | if (rc < 0) | |
85ae8926 | 3260 | goto err_out_id; |
27cc2594 | 3261 | rbd_dev->major = rc; |
602adf40 | 3262 | |
0f308a31 AE |
3263 | /* Set up the blkdev mapping. */ |
3264 | ||
3265 | rc = rbd_init_disk(rbd_dev); | |
dfc5606d | 3266 | if (rc) |
766fc439 YS |
3267 | goto err_out_blkdev; |
3268 | ||
0f308a31 AE |
3269 | rc = rbd_bus_add_dev(rbd_dev); |
3270 | if (rc) | |
3271 | goto err_out_disk; | |
3272 | ||
32eec68d AE |
3273 | /* |
3274 | * At this point cleanup in the event of an error is the job | |
3275 | * of the sysfs code (initiated by rbd_bus_del_dev()). | |
32eec68d | 3276 | */ |
2ac4e75d | 3277 | |
4bb1f1ed | 3278 | down_write(&rbd_dev->header_rwsem); |
5ed16177 | 3279 | rc = rbd_dev_snaps_register(rbd_dev); |
4bb1f1ed | 3280 | up_write(&rbd_dev->header_rwsem); |
2ac4e75d AE |
3281 | if (rc) |
3282 | goto err_out_bus; | |
3283 | ||
3ee4001e AE |
3284 | rc = rbd_init_watch_dev(rbd_dev); |
3285 | if (rc) | |
3286 | goto err_out_bus; | |
3287 | ||
4e9afeba AE |
3288 | kfree(rbd_opts); |
3289 | ||
2ac4e75d AE |
3290 | /* Everything's ready. Announce the disk to the world. */ |
3291 | ||
2ac4e75d | 3292 | add_disk(rbd_dev->disk); |
3ee4001e | 3293 | |
2ac4e75d AE |
3294 | pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name, |
3295 | (unsigned long long) rbd_dev->mapping.size); | |
3296 | ||
602adf40 YS |
3297 | return count; |
3298 | ||
766fc439 | 3299 | err_out_bus: |
766fc439 YS |
3300 | /* this will also clean up rest of rbd_dev stuff */ |
3301 | ||
3302 | rbd_bus_del_dev(rbd_dev); | |
4e9afeba AE |
3303 | kfree(rbd_opts); |
3304 | ||
766fc439 YS |
3305 | return rc; |
3306 | ||
0f308a31 AE |
3307 | err_out_disk: |
3308 | rbd_free_disk(rbd_dev); | |
602adf40 YS |
3309 | err_out_blkdev: |
3310 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
85ae8926 AE |
3311 | err_out_id: |
3312 | rbd_dev_id_put(rbd_dev); | |
41f38c2b AE |
3313 | err_out_snaps: |
3314 | rbd_remove_all_snaps(rbd_dev); | |
3315 | err_out_probe: | |
05fd6f6f | 3316 | rbd_header_free(&rbd_dev->header); |
602adf40 | 3317 | err_out_client: |
3fcf2581 | 3318 | kfree(rbd_dev->header_name); |
602adf40 | 3319 | rbd_put_client(rbd_dev); |
0ddebc0c | 3320 | err_out_args: |
78cea76e AE |
3321 | if (ceph_opts) |
3322 | ceph_destroy_options(ceph_opts); | |
4e9afeba | 3323 | kfree(rbd_opts); |
85ae8926 | 3324 | err_out_mem: |
8b8fb99c | 3325 | rbd_spec_put(rbd_dev->spec); |
27cc2594 AE |
3326 | kfree(rbd_dev); |
3327 | ||
602adf40 YS |
3328 | dout("Error adding device %s\n", buf); |
3329 | module_put(THIS_MODULE); | |
27cc2594 AE |
3330 | |
3331 | return (ssize_t) rc; | |
602adf40 YS |
3332 | } |
3333 | ||
de71a297 | 3334 | static struct rbd_device *__rbd_get_dev(unsigned long dev_id) |
602adf40 YS |
3335 | { |
3336 | struct list_head *tmp; | |
3337 | struct rbd_device *rbd_dev; | |
3338 | ||
e124a82f | 3339 | spin_lock(&rbd_dev_list_lock); |
602adf40 YS |
3340 | list_for_each(tmp, &rbd_dev_list) { |
3341 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
de71a297 | 3342 | if (rbd_dev->dev_id == dev_id) { |
e124a82f | 3343 | spin_unlock(&rbd_dev_list_lock); |
602adf40 | 3344 | return rbd_dev; |
e124a82f | 3345 | } |
602adf40 | 3346 | } |
e124a82f | 3347 | spin_unlock(&rbd_dev_list_lock); |
602adf40 YS |
3348 | return NULL; |
3349 | } | |
3350 | ||
dfc5606d | 3351 | static void rbd_dev_release(struct device *dev) |
602adf40 | 3352 | { |
593a9e7b | 3353 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 3354 | |
1dbb4399 AE |
3355 | if (rbd_dev->watch_request) { |
3356 | struct ceph_client *client = rbd_dev->rbd_client->client; | |
3357 | ||
3358 | ceph_osdc_unregister_linger_request(&client->osdc, | |
59c2be1e | 3359 | rbd_dev->watch_request); |
1dbb4399 | 3360 | } |
59c2be1e | 3361 | if (rbd_dev->watch_event) |
070c633f | 3362 | rbd_req_sync_unwatch(rbd_dev); |
59c2be1e | 3363 | |
602adf40 YS |
3364 | rbd_put_client(rbd_dev); |
3365 | ||
3366 | /* clean up and free blkdev */ | |
3367 | rbd_free_disk(rbd_dev); | |
3368 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
32eec68d | 3369 | |
2ac4e75d AE |
3370 | /* release allocated disk header fields */ |
3371 | rbd_header_free(&rbd_dev->header); | |
3372 | ||
32eec68d | 3373 | /* done with the id, and with the rbd_dev */ |
0bed54dc | 3374 | kfree(rbd_dev->header_name); |
e2839308 | 3375 | rbd_dev_id_put(rbd_dev); |
8b8fb99c | 3376 | rbd_spec_put(rbd_dev->spec); |
602adf40 YS |
3377 | kfree(rbd_dev); |
3378 | ||
3379 | /* release module ref */ | |
3380 | module_put(THIS_MODULE); | |
602adf40 YS |
3381 | } |
3382 | ||
dfc5606d YS |
3383 | static ssize_t rbd_remove(struct bus_type *bus, |
3384 | const char *buf, | |
3385 | size_t count) | |
602adf40 YS |
3386 | { |
3387 | struct rbd_device *rbd_dev = NULL; | |
3388 | int target_id, rc; | |
3389 | unsigned long ul; | |
3390 | int ret = count; | |
3391 | ||
3392 | rc = strict_strtoul(buf, 10, &ul); | |
3393 | if (rc) | |
3394 | return rc; | |
3395 | ||
3396 | /* convert to int; abort if we lost anything in the conversion */ | |
3397 | target_id = (int) ul; | |
3398 | if (target_id != ul) | |
3399 | return -EINVAL; | |
3400 | ||
3401 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
3402 | ||
3403 | rbd_dev = __rbd_get_dev(target_id); | |
3404 | if (!rbd_dev) { | |
3405 | ret = -ENOENT; | |
3406 | goto done; | |
3407 | } | |
3408 | ||
41f38c2b | 3409 | rbd_remove_all_snaps(rbd_dev); |
dfc5606d | 3410 | rbd_bus_del_dev(rbd_dev); |
602adf40 YS |
3411 | |
3412 | done: | |
3413 | mutex_unlock(&ctl_mutex); | |
aafb230e | 3414 | |
602adf40 YS |
3415 | return ret; |
3416 | } | |
3417 | ||
602adf40 YS |
3418 | /* |
3419 | * create control files in sysfs | |
dfc5606d | 3420 | * /sys/bus/rbd/... |
602adf40 YS |
3421 | */ |
3422 | static int rbd_sysfs_init(void) | |
3423 | { | |
dfc5606d | 3424 | int ret; |
602adf40 | 3425 | |
fed4c143 | 3426 | ret = device_register(&rbd_root_dev); |
21079786 | 3427 | if (ret < 0) |
dfc5606d | 3428 | return ret; |
602adf40 | 3429 | |
fed4c143 AE |
3430 | ret = bus_register(&rbd_bus_type); |
3431 | if (ret < 0) | |
3432 | device_unregister(&rbd_root_dev); | |
602adf40 | 3433 | |
602adf40 YS |
3434 | return ret; |
3435 | } | |
3436 | ||
3437 | static void rbd_sysfs_cleanup(void) | |
3438 | { | |
dfc5606d | 3439 | bus_unregister(&rbd_bus_type); |
fed4c143 | 3440 | device_unregister(&rbd_root_dev); |
602adf40 YS |
3441 | } |
3442 | ||
3443 | int __init rbd_init(void) | |
3444 | { | |
3445 | int rc; | |
3446 | ||
3447 | rc = rbd_sysfs_init(); | |
3448 | if (rc) | |
3449 | return rc; | |
f0f8cef5 | 3450 | pr_info("loaded " RBD_DRV_NAME_LONG "\n"); |
602adf40 YS |
3451 | return 0; |
3452 | } | |
3453 | ||
3454 | void __exit rbd_exit(void) | |
3455 | { | |
3456 | rbd_sysfs_cleanup(); | |
3457 | } | |
3458 | ||
3459 | module_init(rbd_init); | |
3460 | module_exit(rbd_exit); | |
3461 | ||
3462 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
3463 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
3464 | MODULE_DESCRIPTION("rados block device"); | |
3465 | ||
3466 | /* following authorship retained from original osdblk.c */ | |
3467 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); | |
3468 | ||
3469 | MODULE_LICENSE("GPL"); |