]>
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 | ||
115 | struct rbd_options { | |
cc0538b6 | 116 | bool read_only; |
602adf40 YS |
117 | }; |
118 | ||
119 | /* | |
f0f8cef5 | 120 | * an instance of the client. multiple devices may share an rbd client. |
602adf40 YS |
121 | */ |
122 | struct rbd_client { | |
123 | struct ceph_client *client; | |
124 | struct kref kref; | |
125 | struct list_head node; | |
126 | }; | |
127 | ||
128 | /* | |
f0f8cef5 | 129 | * a request completion status |
602adf40 | 130 | */ |
1fec7093 YS |
131 | struct rbd_req_status { |
132 | int done; | |
133 | int rc; | |
134 | u64 bytes; | |
135 | }; | |
136 | ||
137 | /* | |
138 | * a collection of requests | |
139 | */ | |
140 | struct rbd_req_coll { | |
141 | int total; | |
142 | int num_done; | |
143 | struct kref kref; | |
144 | struct rbd_req_status status[0]; | |
602adf40 YS |
145 | }; |
146 | ||
f0f8cef5 AE |
147 | /* |
148 | * a single io request | |
149 | */ | |
150 | struct rbd_request { | |
151 | struct request *rq; /* blk layer request */ | |
152 | struct bio *bio; /* cloned bio */ | |
153 | struct page **pages; /* list of used pages */ | |
154 | u64 len; | |
155 | int coll_index; | |
156 | struct rbd_req_coll *coll; | |
157 | }; | |
158 | ||
dfc5606d YS |
159 | struct rbd_snap { |
160 | struct device dev; | |
161 | const char *name; | |
3591538f | 162 | u64 size; |
dfc5606d YS |
163 | struct list_head node; |
164 | u64 id; | |
34b13184 | 165 | u64 features; |
dfc5606d YS |
166 | }; |
167 | ||
f84344f3 | 168 | struct rbd_mapping { |
99c1f08f | 169 | u64 size; |
34b13184 | 170 | u64 features; |
f84344f3 AE |
171 | bool read_only; |
172 | }; | |
173 | ||
602adf40 YS |
174 | /* |
175 | * a single device | |
176 | */ | |
177 | struct rbd_device { | |
de71a297 | 178 | int dev_id; /* blkdev unique id */ |
602adf40 YS |
179 | |
180 | int major; /* blkdev assigned major */ | |
181 | struct gendisk *disk; /* blkdev's gendisk and rq */ | |
602adf40 | 182 | |
a30b71b9 | 183 | u32 image_format; /* Either 1 or 2 */ |
602adf40 YS |
184 | struct rbd_client *rbd_client; |
185 | ||
186 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ | |
187 | ||
188 | spinlock_t lock; /* queue lock */ | |
189 | ||
190 | struct rbd_image_header header; | |
daba5fdb | 191 | bool exists; |
589d30e0 AE |
192 | char *image_id; |
193 | size_t image_id_len; | |
0bed54dc AE |
194 | char *image_name; |
195 | size_t image_name_len; | |
196 | char *header_name; | |
d22f76e7 | 197 | char *pool_name; |
86992098 | 198 | u64 pool_id; |
602adf40 | 199 | |
971f839a AE |
200 | char *snap_name; |
201 | u64 snap_id; | |
202 | ||
59c2be1e YS |
203 | struct ceph_osd_event *watch_event; |
204 | struct ceph_osd_request *watch_request; | |
205 | ||
c666601a JD |
206 | /* protects updating the header */ |
207 | struct rw_semaphore header_rwsem; | |
f84344f3 AE |
208 | |
209 | struct rbd_mapping mapping; | |
602adf40 YS |
210 | |
211 | struct list_head node; | |
dfc5606d YS |
212 | |
213 | /* list of snapshots */ | |
214 | struct list_head snaps; | |
215 | ||
216 | /* sysfs related */ | |
217 | struct device dev; | |
218 | }; | |
219 | ||
602adf40 | 220 | static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ |
e124a82f | 221 | |
602adf40 | 222 | static LIST_HEAD(rbd_dev_list); /* devices */ |
e124a82f AE |
223 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
224 | ||
432b8587 AE |
225 | static LIST_HEAD(rbd_client_list); /* clients */ |
226 | static DEFINE_SPINLOCK(rbd_client_list_lock); | |
602adf40 | 227 | |
304f6808 AE |
228 | static int rbd_dev_snaps_update(struct rbd_device *rbd_dev); |
229 | static int rbd_dev_snaps_register(struct rbd_device *rbd_dev); | |
230 | ||
dfc5606d | 231 | static void rbd_dev_release(struct device *dev); |
41f38c2b | 232 | static void rbd_remove_snap_dev(struct rbd_snap *snap); |
dfc5606d | 233 | |
f0f8cef5 AE |
234 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
235 | size_t count); | |
236 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, | |
237 | size_t count); | |
238 | ||
239 | static struct bus_attribute rbd_bus_attrs[] = { | |
240 | __ATTR(add, S_IWUSR, NULL, rbd_add), | |
241 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), | |
242 | __ATTR_NULL | |
243 | }; | |
244 | ||
245 | static struct bus_type rbd_bus_type = { | |
246 | .name = "rbd", | |
247 | .bus_attrs = rbd_bus_attrs, | |
248 | }; | |
249 | ||
250 | static void rbd_root_dev_release(struct device *dev) | |
251 | { | |
252 | } | |
253 | ||
254 | static struct device rbd_root_dev = { | |
255 | .init_name = "rbd", | |
256 | .release = rbd_root_dev_release, | |
257 | }; | |
258 | ||
aafb230e AE |
259 | #ifdef RBD_DEBUG |
260 | #define rbd_assert(expr) \ | |
261 | if (unlikely(!(expr))) { \ | |
262 | printk(KERN_ERR "\nAssertion failure in %s() " \ | |
263 | "at line %d:\n\n" \ | |
264 | "\trbd_assert(%s);\n\n", \ | |
265 | __func__, __LINE__, #expr); \ | |
266 | BUG(); \ | |
267 | } | |
268 | #else /* !RBD_DEBUG */ | |
269 | # define rbd_assert(expr) ((void) 0) | |
270 | #endif /* !RBD_DEBUG */ | |
dfc5606d | 271 | |
dfc5606d YS |
272 | static struct device *rbd_get_dev(struct rbd_device *rbd_dev) |
273 | { | |
274 | return get_device(&rbd_dev->dev); | |
275 | } | |
276 | ||
277 | static void rbd_put_dev(struct rbd_device *rbd_dev) | |
278 | { | |
279 | put_device(&rbd_dev->dev); | |
280 | } | |
602adf40 | 281 | |
117973fb AE |
282 | static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver); |
283 | static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver); | |
59c2be1e | 284 | |
602adf40 YS |
285 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
286 | { | |
f0f8cef5 | 287 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
602adf40 | 288 | |
f84344f3 | 289 | if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only) |
602adf40 YS |
290 | return -EROFS; |
291 | ||
340c7a2b | 292 | rbd_get_dev(rbd_dev); |
f84344f3 | 293 | set_device_ro(bdev, rbd_dev->mapping.read_only); |
340c7a2b | 294 | |
602adf40 YS |
295 | return 0; |
296 | } | |
297 | ||
dfc5606d YS |
298 | static int rbd_release(struct gendisk *disk, fmode_t mode) |
299 | { | |
300 | struct rbd_device *rbd_dev = disk->private_data; | |
301 | ||
302 | rbd_put_dev(rbd_dev); | |
303 | ||
304 | return 0; | |
305 | } | |
306 | ||
602adf40 YS |
307 | static const struct block_device_operations rbd_bd_ops = { |
308 | .owner = THIS_MODULE, | |
309 | .open = rbd_open, | |
dfc5606d | 310 | .release = rbd_release, |
602adf40 YS |
311 | }; |
312 | ||
313 | /* | |
314 | * Initialize an rbd client instance. | |
43ae4701 | 315 | * We own *ceph_opts. |
602adf40 | 316 | */ |
f8c38929 | 317 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) |
602adf40 YS |
318 | { |
319 | struct rbd_client *rbdc; | |
320 | int ret = -ENOMEM; | |
321 | ||
322 | dout("rbd_client_create\n"); | |
323 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); | |
324 | if (!rbdc) | |
325 | goto out_opt; | |
326 | ||
327 | kref_init(&rbdc->kref); | |
328 | INIT_LIST_HEAD(&rbdc->node); | |
329 | ||
bc534d86 AE |
330 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
331 | ||
43ae4701 | 332 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
602adf40 | 333 | if (IS_ERR(rbdc->client)) |
bc534d86 | 334 | goto out_mutex; |
43ae4701 | 335 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
602adf40 YS |
336 | |
337 | ret = ceph_open_session(rbdc->client); | |
338 | if (ret < 0) | |
339 | goto out_err; | |
340 | ||
432b8587 | 341 | spin_lock(&rbd_client_list_lock); |
602adf40 | 342 | list_add_tail(&rbdc->node, &rbd_client_list); |
432b8587 | 343 | spin_unlock(&rbd_client_list_lock); |
602adf40 | 344 | |
bc534d86 AE |
345 | mutex_unlock(&ctl_mutex); |
346 | ||
602adf40 YS |
347 | dout("rbd_client_create created %p\n", rbdc); |
348 | return rbdc; | |
349 | ||
350 | out_err: | |
351 | ceph_destroy_client(rbdc->client); | |
bc534d86 AE |
352 | out_mutex: |
353 | mutex_unlock(&ctl_mutex); | |
602adf40 YS |
354 | kfree(rbdc); |
355 | out_opt: | |
43ae4701 AE |
356 | if (ceph_opts) |
357 | ceph_destroy_options(ceph_opts); | |
28f259b7 | 358 | return ERR_PTR(ret); |
602adf40 YS |
359 | } |
360 | ||
361 | /* | |
1f7ba331 AE |
362 | * Find a ceph client with specific addr and configuration. If |
363 | * found, bump its reference count. | |
602adf40 | 364 | */ |
1f7ba331 | 365 | static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) |
602adf40 YS |
366 | { |
367 | struct rbd_client *client_node; | |
1f7ba331 | 368 | bool found = false; |
602adf40 | 369 | |
43ae4701 | 370 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
602adf40 YS |
371 | return NULL; |
372 | ||
1f7ba331 AE |
373 | spin_lock(&rbd_client_list_lock); |
374 | list_for_each_entry(client_node, &rbd_client_list, node) { | |
375 | if (!ceph_compare_options(ceph_opts, client_node->client)) { | |
376 | kref_get(&client_node->kref); | |
377 | found = true; | |
378 | break; | |
379 | } | |
380 | } | |
381 | spin_unlock(&rbd_client_list_lock); | |
382 | ||
383 | return found ? client_node : NULL; | |
602adf40 YS |
384 | } |
385 | ||
59c2be1e YS |
386 | /* |
387 | * mount options | |
388 | */ | |
389 | enum { | |
59c2be1e YS |
390 | Opt_last_int, |
391 | /* int args above */ | |
392 | Opt_last_string, | |
393 | /* string args above */ | |
cc0538b6 AE |
394 | Opt_read_only, |
395 | Opt_read_write, | |
396 | /* Boolean args above */ | |
397 | Opt_last_bool, | |
59c2be1e YS |
398 | }; |
399 | ||
43ae4701 | 400 | static match_table_t rbd_opts_tokens = { |
59c2be1e YS |
401 | /* int args above */ |
402 | /* string args above */ | |
be466c1c | 403 | {Opt_read_only, "read_only"}, |
cc0538b6 AE |
404 | {Opt_read_only, "ro"}, /* Alternate spelling */ |
405 | {Opt_read_write, "read_write"}, | |
406 | {Opt_read_write, "rw"}, /* Alternate spelling */ | |
407 | /* Boolean args above */ | |
59c2be1e YS |
408 | {-1, NULL} |
409 | }; | |
410 | ||
411 | static int parse_rbd_opts_token(char *c, void *private) | |
412 | { | |
43ae4701 | 413 | struct rbd_options *rbd_opts = private; |
59c2be1e YS |
414 | substring_t argstr[MAX_OPT_ARGS]; |
415 | int token, intval, ret; | |
416 | ||
43ae4701 | 417 | token = match_token(c, rbd_opts_tokens, argstr); |
59c2be1e YS |
418 | if (token < 0) |
419 | return -EINVAL; | |
420 | ||
421 | if (token < Opt_last_int) { | |
422 | ret = match_int(&argstr[0], &intval); | |
423 | if (ret < 0) { | |
424 | pr_err("bad mount option arg (not int) " | |
425 | "at '%s'\n", c); | |
426 | return ret; | |
427 | } | |
428 | dout("got int token %d val %d\n", token, intval); | |
429 | } else if (token > Opt_last_int && token < Opt_last_string) { | |
430 | dout("got string token %d val %s\n", token, | |
431 | argstr[0].from); | |
cc0538b6 AE |
432 | } else if (token > Opt_last_string && token < Opt_last_bool) { |
433 | dout("got Boolean token %d\n", token); | |
59c2be1e YS |
434 | } else { |
435 | dout("got token %d\n", token); | |
436 | } | |
437 | ||
438 | switch (token) { | |
cc0538b6 AE |
439 | case Opt_read_only: |
440 | rbd_opts->read_only = true; | |
441 | break; | |
442 | case Opt_read_write: | |
443 | rbd_opts->read_only = false; | |
444 | break; | |
59c2be1e | 445 | default: |
aafb230e AE |
446 | rbd_assert(false); |
447 | break; | |
59c2be1e YS |
448 | } |
449 | return 0; | |
450 | } | |
451 | ||
602adf40 YS |
452 | /* |
453 | * Get a ceph client with specific addr and configuration, if one does | |
454 | * not exist create it. | |
455 | */ | |
78cea76e AE |
456 | static int rbd_get_client(struct rbd_device *rbd_dev, |
457 | struct ceph_options *ceph_opts) | |
602adf40 | 458 | { |
f8c38929 | 459 | struct rbd_client *rbdc; |
59c2be1e | 460 | |
1f7ba331 | 461 | rbdc = rbd_client_find(ceph_opts); |
602adf40 | 462 | if (rbdc) { |
602adf40 | 463 | /* using an existing client */ |
43ae4701 | 464 | ceph_destroy_options(ceph_opts); |
f8c38929 AE |
465 | } else { |
466 | rbdc = rbd_client_create(ceph_opts); | |
467 | if (IS_ERR(rbdc)) | |
468 | return PTR_ERR(rbdc); | |
602adf40 | 469 | } |
f8c38929 | 470 | rbd_dev->rbd_client = rbdc; |
602adf40 | 471 | |
f8c38929 | 472 | return 0; |
602adf40 YS |
473 | } |
474 | ||
475 | /* | |
476 | * Destroy ceph client | |
d23a4b3f | 477 | * |
432b8587 | 478 | * Caller must hold rbd_client_list_lock. |
602adf40 YS |
479 | */ |
480 | static void rbd_client_release(struct kref *kref) | |
481 | { | |
482 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); | |
483 | ||
484 | dout("rbd_release_client %p\n", rbdc); | |
cd9d9f5d | 485 | spin_lock(&rbd_client_list_lock); |
602adf40 | 486 | list_del(&rbdc->node); |
cd9d9f5d | 487 | spin_unlock(&rbd_client_list_lock); |
602adf40 YS |
488 | |
489 | ceph_destroy_client(rbdc->client); | |
490 | kfree(rbdc); | |
491 | } | |
492 | ||
493 | /* | |
494 | * Drop reference to ceph client node. If it's not referenced anymore, release | |
495 | * it. | |
496 | */ | |
497 | static void rbd_put_client(struct rbd_device *rbd_dev) | |
498 | { | |
499 | kref_put(&rbd_dev->rbd_client->kref, rbd_client_release); | |
500 | rbd_dev->rbd_client = NULL; | |
602adf40 YS |
501 | } |
502 | ||
1fec7093 YS |
503 | /* |
504 | * Destroy requests collection | |
505 | */ | |
506 | static void rbd_coll_release(struct kref *kref) | |
507 | { | |
508 | struct rbd_req_coll *coll = | |
509 | container_of(kref, struct rbd_req_coll, kref); | |
510 | ||
511 | dout("rbd_coll_release %p\n", coll); | |
512 | kfree(coll); | |
513 | } | |
602adf40 | 514 | |
a30b71b9 AE |
515 | static bool rbd_image_format_valid(u32 image_format) |
516 | { | |
517 | return image_format == 1 || image_format == 2; | |
518 | } | |
519 | ||
8e94af8e AE |
520 | static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) |
521 | { | |
103a150f AE |
522 | size_t size; |
523 | u32 snap_count; | |
524 | ||
525 | /* The header has to start with the magic rbd header text */ | |
526 | if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) | |
527 | return false; | |
528 | ||
db2388b6 AE |
529 | /* The bio layer requires at least sector-sized I/O */ |
530 | ||
531 | if (ondisk->options.order < SECTOR_SHIFT) | |
532 | return false; | |
533 | ||
534 | /* If we use u64 in a few spots we may be able to loosen this */ | |
535 | ||
536 | if (ondisk->options.order > 8 * sizeof (int) - 1) | |
537 | return false; | |
538 | ||
103a150f AE |
539 | /* |
540 | * The size of a snapshot header has to fit in a size_t, and | |
541 | * that limits the number of snapshots. | |
542 | */ | |
543 | snap_count = le32_to_cpu(ondisk->snap_count); | |
544 | size = SIZE_MAX - sizeof (struct ceph_snap_context); | |
545 | if (snap_count > size / sizeof (__le64)) | |
546 | return false; | |
547 | ||
548 | /* | |
549 | * Not only that, but the size of the entire the snapshot | |
550 | * header must also be representable in a size_t. | |
551 | */ | |
552 | size -= snap_count * sizeof (__le64); | |
553 | if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) | |
554 | return false; | |
555 | ||
556 | return true; | |
8e94af8e AE |
557 | } |
558 | ||
602adf40 YS |
559 | /* |
560 | * Create a new header structure, translate header format from the on-disk | |
561 | * header. | |
562 | */ | |
563 | static int rbd_header_from_disk(struct rbd_image_header *header, | |
4156d998 | 564 | struct rbd_image_header_ondisk *ondisk) |
602adf40 | 565 | { |
ccece235 | 566 | u32 snap_count; |
58c17b0e | 567 | size_t len; |
d2bb24e5 | 568 | size_t size; |
621901d6 | 569 | u32 i; |
602adf40 | 570 | |
6a52325f AE |
571 | memset(header, 0, sizeof (*header)); |
572 | ||
103a150f AE |
573 | snap_count = le32_to_cpu(ondisk->snap_count); |
574 | ||
58c17b0e AE |
575 | len = strnlen(ondisk->object_prefix, sizeof (ondisk->object_prefix)); |
576 | header->object_prefix = kmalloc(len + 1, GFP_KERNEL); | |
6a52325f | 577 | if (!header->object_prefix) |
602adf40 | 578 | return -ENOMEM; |
58c17b0e AE |
579 | memcpy(header->object_prefix, ondisk->object_prefix, len); |
580 | header->object_prefix[len] = '\0'; | |
00f1f36f | 581 | |
602adf40 | 582 | if (snap_count) { |
f785cc1d AE |
583 | u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
584 | ||
621901d6 AE |
585 | /* Save a copy of the snapshot names */ |
586 | ||
f785cc1d AE |
587 | if (snap_names_len > (u64) SIZE_MAX) |
588 | return -EIO; | |
589 | header->snap_names = kmalloc(snap_names_len, GFP_KERNEL); | |
602adf40 | 590 | if (!header->snap_names) |
6a52325f | 591 | goto out_err; |
f785cc1d AE |
592 | /* |
593 | * Note that rbd_dev_v1_header_read() guarantees | |
594 | * the ondisk buffer we're working with has | |
595 | * snap_names_len bytes beyond the end of the | |
596 | * snapshot id array, this memcpy() is safe. | |
597 | */ | |
598 | memcpy(header->snap_names, &ondisk->snaps[snap_count], | |
599 | snap_names_len); | |
6a52325f | 600 | |
621901d6 AE |
601 | /* Record each snapshot's size */ |
602 | ||
d2bb24e5 AE |
603 | size = snap_count * sizeof (*header->snap_sizes); |
604 | header->snap_sizes = kmalloc(size, GFP_KERNEL); | |
602adf40 | 605 | if (!header->snap_sizes) |
6a52325f | 606 | goto out_err; |
621901d6 AE |
607 | for (i = 0; i < snap_count; i++) |
608 | header->snap_sizes[i] = | |
609 | le64_to_cpu(ondisk->snaps[i].image_size); | |
602adf40 | 610 | } else { |
ccece235 | 611 | WARN_ON(ondisk->snap_names_len); |
602adf40 YS |
612 | header->snap_names = NULL; |
613 | header->snap_sizes = NULL; | |
614 | } | |
849b4260 | 615 | |
34b13184 | 616 | header->features = 0; /* No features support in v1 images */ |
602adf40 YS |
617 | header->obj_order = ondisk->options.order; |
618 | header->crypt_type = ondisk->options.crypt_type; | |
619 | header->comp_type = ondisk->options.comp_type; | |
6a52325f | 620 | |
621901d6 AE |
621 | /* Allocate and fill in the snapshot context */ |
622 | ||
f84344f3 | 623 | header->image_size = le64_to_cpu(ondisk->image_size); |
6a52325f AE |
624 | size = sizeof (struct ceph_snap_context); |
625 | size += snap_count * sizeof (header->snapc->snaps[0]); | |
626 | header->snapc = kzalloc(size, GFP_KERNEL); | |
627 | if (!header->snapc) | |
628 | goto out_err; | |
602adf40 YS |
629 | |
630 | atomic_set(&header->snapc->nref, 1); | |
505cbb9b | 631 | header->snapc->seq = le64_to_cpu(ondisk->snap_seq); |
602adf40 | 632 | header->snapc->num_snaps = snap_count; |
621901d6 AE |
633 | for (i = 0; i < snap_count; i++) |
634 | header->snapc->snaps[i] = | |
635 | le64_to_cpu(ondisk->snaps[i].id); | |
602adf40 YS |
636 | |
637 | return 0; | |
638 | ||
6a52325f | 639 | out_err: |
849b4260 | 640 | kfree(header->snap_sizes); |
ccece235 | 641 | header->snap_sizes = NULL; |
602adf40 | 642 | kfree(header->snap_names); |
ccece235 | 643 | header->snap_names = NULL; |
6a52325f AE |
644 | kfree(header->object_prefix); |
645 | header->object_prefix = NULL; | |
ccece235 | 646 | |
00f1f36f | 647 | return -ENOMEM; |
602adf40 YS |
648 | } |
649 | ||
8836b995 | 650 | static int snap_by_name(struct rbd_device *rbd_dev, const char *snap_name) |
602adf40 | 651 | { |
602adf40 | 652 | |
e86924a8 | 653 | struct rbd_snap *snap; |
602adf40 | 654 | |
e86924a8 AE |
655 | list_for_each_entry(snap, &rbd_dev->snaps, node) { |
656 | if (!strcmp(snap_name, snap->name)) { | |
971f839a | 657 | rbd_dev->snap_id = snap->id; |
e86924a8 | 658 | rbd_dev->mapping.size = snap->size; |
34b13184 | 659 | rbd_dev->mapping.features = snap->features; |
602adf40 | 660 | |
e86924a8 | 661 | return 0; |
00f1f36f | 662 | } |
00f1f36f | 663 | } |
e86924a8 | 664 | |
00f1f36f | 665 | return -ENOENT; |
602adf40 YS |
666 | } |
667 | ||
819d52bf | 668 | static int rbd_dev_set_mapping(struct rbd_device *rbd_dev) |
602adf40 | 669 | { |
78dc447d | 670 | int ret; |
602adf40 | 671 | |
819d52bf | 672 | if (!memcmp(rbd_dev->snap_name, RBD_SNAP_HEAD_NAME, |
cc9d734c | 673 | sizeof (RBD_SNAP_HEAD_NAME))) { |
971f839a | 674 | rbd_dev->snap_id = CEPH_NOSNAP; |
99c1f08f | 675 | rbd_dev->mapping.size = rbd_dev->header.image_size; |
34b13184 | 676 | rbd_dev->mapping.features = rbd_dev->header.features; |
e86924a8 | 677 | ret = 0; |
602adf40 | 678 | } else { |
819d52bf | 679 | ret = snap_by_name(rbd_dev, rbd_dev->snap_name); |
602adf40 YS |
680 | if (ret < 0) |
681 | goto done; | |
f84344f3 | 682 | rbd_dev->mapping.read_only = true; |
602adf40 | 683 | } |
daba5fdb | 684 | rbd_dev->exists = true; |
602adf40 | 685 | done: |
602adf40 YS |
686 | return ret; |
687 | } | |
688 | ||
689 | static void rbd_header_free(struct rbd_image_header *header) | |
690 | { | |
849b4260 | 691 | kfree(header->object_prefix); |
d78fd7ae | 692 | header->object_prefix = NULL; |
602adf40 | 693 | kfree(header->snap_sizes); |
d78fd7ae | 694 | header->snap_sizes = NULL; |
849b4260 | 695 | kfree(header->snap_names); |
d78fd7ae | 696 | header->snap_names = NULL; |
d1d25646 | 697 | ceph_put_snap_context(header->snapc); |
d78fd7ae | 698 | header->snapc = NULL; |
602adf40 YS |
699 | } |
700 | ||
65ccfe21 | 701 | static char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset) |
602adf40 | 702 | { |
65ccfe21 AE |
703 | char *name; |
704 | u64 segment; | |
705 | int ret; | |
602adf40 | 706 | |
65ccfe21 AE |
707 | name = kmalloc(RBD_MAX_SEG_NAME_LEN + 1, GFP_NOIO); |
708 | if (!name) | |
709 | return NULL; | |
710 | segment = offset >> rbd_dev->header.obj_order; | |
711 | ret = snprintf(name, RBD_MAX_SEG_NAME_LEN, "%s.%012llx", | |
712 | rbd_dev->header.object_prefix, segment); | |
713 | if (ret < 0 || ret >= RBD_MAX_SEG_NAME_LEN) { | |
714 | pr_err("error formatting segment name for #%llu (%d)\n", | |
715 | segment, ret); | |
716 | kfree(name); | |
717 | name = NULL; | |
718 | } | |
602adf40 | 719 | |
65ccfe21 AE |
720 | return name; |
721 | } | |
602adf40 | 722 | |
65ccfe21 AE |
723 | static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) |
724 | { | |
725 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; | |
602adf40 | 726 | |
65ccfe21 AE |
727 | return offset & (segment_size - 1); |
728 | } | |
729 | ||
730 | static u64 rbd_segment_length(struct rbd_device *rbd_dev, | |
731 | u64 offset, u64 length) | |
732 | { | |
733 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; | |
734 | ||
735 | offset &= segment_size - 1; | |
736 | ||
aafb230e | 737 | rbd_assert(length <= U64_MAX - offset); |
65ccfe21 AE |
738 | if (offset + length > segment_size) |
739 | length = segment_size - offset; | |
740 | ||
741 | return length; | |
602adf40 YS |
742 | } |
743 | ||
1fec7093 YS |
744 | static int rbd_get_num_segments(struct rbd_image_header *header, |
745 | u64 ofs, u64 len) | |
746 | { | |
df111be6 AE |
747 | u64 start_seg; |
748 | u64 end_seg; | |
749 | ||
750 | if (!len) | |
751 | return 0; | |
752 | if (len - 1 > U64_MAX - ofs) | |
753 | return -ERANGE; | |
754 | ||
755 | start_seg = ofs >> header->obj_order; | |
756 | end_seg = (ofs + len - 1) >> header->obj_order; | |
757 | ||
1fec7093 YS |
758 | return end_seg - start_seg + 1; |
759 | } | |
760 | ||
029bcbd8 JD |
761 | /* |
762 | * returns the size of an object in the image | |
763 | */ | |
764 | static u64 rbd_obj_bytes(struct rbd_image_header *header) | |
765 | { | |
766 | return 1 << header->obj_order; | |
767 | } | |
768 | ||
602adf40 YS |
769 | /* |
770 | * bio helpers | |
771 | */ | |
772 | ||
773 | static void bio_chain_put(struct bio *chain) | |
774 | { | |
775 | struct bio *tmp; | |
776 | ||
777 | while (chain) { | |
778 | tmp = chain; | |
779 | chain = chain->bi_next; | |
780 | bio_put(tmp); | |
781 | } | |
782 | } | |
783 | ||
784 | /* | |
785 | * zeros a bio chain, starting at specific offset | |
786 | */ | |
787 | static void zero_bio_chain(struct bio *chain, int start_ofs) | |
788 | { | |
789 | struct bio_vec *bv; | |
790 | unsigned long flags; | |
791 | void *buf; | |
792 | int i; | |
793 | int pos = 0; | |
794 | ||
795 | while (chain) { | |
796 | bio_for_each_segment(bv, chain, i) { | |
797 | if (pos + bv->bv_len > start_ofs) { | |
798 | int remainder = max(start_ofs - pos, 0); | |
799 | buf = bvec_kmap_irq(bv, &flags); | |
800 | memset(buf + remainder, 0, | |
801 | bv->bv_len - remainder); | |
85b5aaa6 | 802 | bvec_kunmap_irq(buf, &flags); |
602adf40 YS |
803 | } |
804 | pos += bv->bv_len; | |
805 | } | |
806 | ||
807 | chain = chain->bi_next; | |
808 | } | |
809 | } | |
810 | ||
811 | /* | |
f7760dad AE |
812 | * Clone a portion of a bio, starting at the given byte offset |
813 | * and continuing for the number of bytes indicated. | |
602adf40 | 814 | */ |
f7760dad AE |
815 | static struct bio *bio_clone_range(struct bio *bio_src, |
816 | unsigned int offset, | |
817 | unsigned int len, | |
818 | gfp_t gfpmask) | |
602adf40 | 819 | { |
f7760dad AE |
820 | struct bio_vec *bv; |
821 | unsigned int resid; | |
822 | unsigned short idx; | |
823 | unsigned int voff; | |
824 | unsigned short end_idx; | |
825 | unsigned short vcnt; | |
826 | struct bio *bio; | |
827 | ||
828 | /* Handle the easy case for the caller */ | |
829 | ||
830 | if (!offset && len == bio_src->bi_size) | |
831 | return bio_clone(bio_src, gfpmask); | |
832 | ||
833 | if (WARN_ON_ONCE(!len)) | |
834 | return NULL; | |
835 | if (WARN_ON_ONCE(len > bio_src->bi_size)) | |
836 | return NULL; | |
837 | if (WARN_ON_ONCE(offset > bio_src->bi_size - len)) | |
838 | return NULL; | |
839 | ||
840 | /* Find first affected segment... */ | |
841 | ||
842 | resid = offset; | |
843 | __bio_for_each_segment(bv, bio_src, idx, 0) { | |
844 | if (resid < bv->bv_len) | |
845 | break; | |
846 | resid -= bv->bv_len; | |
602adf40 | 847 | } |
f7760dad | 848 | voff = resid; |
602adf40 | 849 | |
f7760dad | 850 | /* ...and the last affected segment */ |
602adf40 | 851 | |
f7760dad AE |
852 | resid += len; |
853 | __bio_for_each_segment(bv, bio_src, end_idx, idx) { | |
854 | if (resid <= bv->bv_len) | |
855 | break; | |
856 | resid -= bv->bv_len; | |
857 | } | |
858 | vcnt = end_idx - idx + 1; | |
859 | ||
860 | /* Build the clone */ | |
861 | ||
862 | bio = bio_alloc(gfpmask, (unsigned int) vcnt); | |
863 | if (!bio) | |
864 | return NULL; /* ENOMEM */ | |
602adf40 | 865 | |
f7760dad AE |
866 | bio->bi_bdev = bio_src->bi_bdev; |
867 | bio->bi_sector = bio_src->bi_sector + (offset >> SECTOR_SHIFT); | |
868 | bio->bi_rw = bio_src->bi_rw; | |
869 | bio->bi_flags |= 1 << BIO_CLONED; | |
870 | ||
871 | /* | |
872 | * Copy over our part of the bio_vec, then update the first | |
873 | * and last (or only) entries. | |
874 | */ | |
875 | memcpy(&bio->bi_io_vec[0], &bio_src->bi_io_vec[idx], | |
876 | vcnt * sizeof (struct bio_vec)); | |
877 | bio->bi_io_vec[0].bv_offset += voff; | |
878 | if (vcnt > 1) { | |
879 | bio->bi_io_vec[0].bv_len -= voff; | |
880 | bio->bi_io_vec[vcnt - 1].bv_len = resid; | |
881 | } else { | |
882 | bio->bi_io_vec[0].bv_len = len; | |
602adf40 YS |
883 | } |
884 | ||
f7760dad AE |
885 | bio->bi_vcnt = vcnt; |
886 | bio->bi_size = len; | |
887 | bio->bi_idx = 0; | |
888 | ||
889 | return bio; | |
890 | } | |
891 | ||
892 | /* | |
893 | * Clone a portion of a bio chain, starting at the given byte offset | |
894 | * into the first bio in the source chain and continuing for the | |
895 | * number of bytes indicated. The result is another bio chain of | |
896 | * exactly the given length, or a null pointer on error. | |
897 | * | |
898 | * The bio_src and offset parameters are both in-out. On entry they | |
899 | * refer to the first source bio and the offset into that bio where | |
900 | * the start of data to be cloned is located. | |
901 | * | |
902 | * On return, bio_src is updated to refer to the bio in the source | |
903 | * chain that contains first un-cloned byte, and *offset will | |
904 | * contain the offset of that byte within that bio. | |
905 | */ | |
906 | static struct bio *bio_chain_clone_range(struct bio **bio_src, | |
907 | unsigned int *offset, | |
908 | unsigned int len, | |
909 | gfp_t gfpmask) | |
910 | { | |
911 | struct bio *bi = *bio_src; | |
912 | unsigned int off = *offset; | |
913 | struct bio *chain = NULL; | |
914 | struct bio **end; | |
915 | ||
916 | /* Build up a chain of clone bios up to the limit */ | |
917 | ||
918 | if (!bi || off >= bi->bi_size || !len) | |
919 | return NULL; /* Nothing to clone */ | |
602adf40 | 920 | |
f7760dad AE |
921 | end = &chain; |
922 | while (len) { | |
923 | unsigned int bi_size; | |
924 | struct bio *bio; | |
925 | ||
926 | if (!bi) | |
927 | goto out_err; /* EINVAL; ran out of bio's */ | |
928 | bi_size = min_t(unsigned int, bi->bi_size - off, len); | |
929 | bio = bio_clone_range(bi, off, bi_size, gfpmask); | |
930 | if (!bio) | |
931 | goto out_err; /* ENOMEM */ | |
932 | ||
933 | *end = bio; | |
934 | end = &bio->bi_next; | |
602adf40 | 935 | |
f7760dad AE |
936 | off += bi_size; |
937 | if (off == bi->bi_size) { | |
938 | bi = bi->bi_next; | |
939 | off = 0; | |
940 | } | |
941 | len -= bi_size; | |
942 | } | |
943 | *bio_src = bi; | |
944 | *offset = off; | |
945 | ||
946 | return chain; | |
947 | out_err: | |
948 | bio_chain_put(chain); | |
602adf40 | 949 | |
602adf40 YS |
950 | return NULL; |
951 | } | |
952 | ||
953 | /* | |
954 | * helpers for osd request op vectors. | |
955 | */ | |
57cfc106 AE |
956 | static struct ceph_osd_req_op *rbd_create_rw_ops(int num_ops, |
957 | int opcode, u32 payload_len) | |
602adf40 | 958 | { |
57cfc106 AE |
959 | struct ceph_osd_req_op *ops; |
960 | ||
961 | ops = kzalloc(sizeof (*ops) * (num_ops + 1), GFP_NOIO); | |
962 | if (!ops) | |
963 | return NULL; | |
964 | ||
965 | ops[0].op = opcode; | |
966 | ||
602adf40 YS |
967 | /* |
968 | * op extent offset and length will be set later on | |
969 | * in calc_raw_layout() | |
970 | */ | |
57cfc106 AE |
971 | ops[0].payload_len = payload_len; |
972 | ||
973 | return ops; | |
602adf40 YS |
974 | } |
975 | ||
976 | static void rbd_destroy_ops(struct ceph_osd_req_op *ops) | |
977 | { | |
978 | kfree(ops); | |
979 | } | |
980 | ||
1fec7093 YS |
981 | static void rbd_coll_end_req_index(struct request *rq, |
982 | struct rbd_req_coll *coll, | |
983 | int index, | |
984 | int ret, u64 len) | |
985 | { | |
986 | struct request_queue *q; | |
987 | int min, max, i; | |
988 | ||
bd919d45 AE |
989 | dout("rbd_coll_end_req_index %p index %d ret %d len %llu\n", |
990 | coll, index, ret, (unsigned long long) len); | |
1fec7093 YS |
991 | |
992 | if (!rq) | |
993 | return; | |
994 | ||
995 | if (!coll) { | |
996 | blk_end_request(rq, ret, len); | |
997 | return; | |
998 | } | |
999 | ||
1000 | q = rq->q; | |
1001 | ||
1002 | spin_lock_irq(q->queue_lock); | |
1003 | coll->status[index].done = 1; | |
1004 | coll->status[index].rc = ret; | |
1005 | coll->status[index].bytes = len; | |
1006 | max = min = coll->num_done; | |
1007 | while (max < coll->total && coll->status[max].done) | |
1008 | max++; | |
1009 | ||
1010 | for (i = min; i<max; i++) { | |
1011 | __blk_end_request(rq, coll->status[i].rc, | |
1012 | coll->status[i].bytes); | |
1013 | coll->num_done++; | |
1014 | kref_put(&coll->kref, rbd_coll_release); | |
1015 | } | |
1016 | spin_unlock_irq(q->queue_lock); | |
1017 | } | |
1018 | ||
1019 | static void rbd_coll_end_req(struct rbd_request *req, | |
1020 | int ret, u64 len) | |
1021 | { | |
1022 | rbd_coll_end_req_index(req->rq, req->coll, req->coll_index, ret, len); | |
1023 | } | |
1024 | ||
602adf40 YS |
1025 | /* |
1026 | * Send ceph osd request | |
1027 | */ | |
1028 | static int rbd_do_request(struct request *rq, | |
0ce1a794 | 1029 | struct rbd_device *rbd_dev, |
602adf40 YS |
1030 | struct ceph_snap_context *snapc, |
1031 | u64 snapid, | |
aded07ea | 1032 | const char *object_name, u64 ofs, u64 len, |
602adf40 YS |
1033 | struct bio *bio, |
1034 | struct page **pages, | |
1035 | int num_pages, | |
1036 | int flags, | |
1037 | struct ceph_osd_req_op *ops, | |
1fec7093 YS |
1038 | struct rbd_req_coll *coll, |
1039 | int coll_index, | |
602adf40 | 1040 | void (*rbd_cb)(struct ceph_osd_request *req, |
59c2be1e YS |
1041 | struct ceph_msg *msg), |
1042 | struct ceph_osd_request **linger_req, | |
1043 | u64 *ver) | |
602adf40 YS |
1044 | { |
1045 | struct ceph_osd_request *req; | |
1046 | struct ceph_file_layout *layout; | |
1047 | int ret; | |
1048 | u64 bno; | |
1049 | struct timespec mtime = CURRENT_TIME; | |
1050 | struct rbd_request *req_data; | |
1051 | struct ceph_osd_request_head *reqhead; | |
1dbb4399 | 1052 | struct ceph_osd_client *osdc; |
602adf40 | 1053 | |
602adf40 | 1054 | req_data = kzalloc(sizeof(*req_data), GFP_NOIO); |
1fec7093 YS |
1055 | if (!req_data) { |
1056 | if (coll) | |
1057 | rbd_coll_end_req_index(rq, coll, coll_index, | |
1058 | -ENOMEM, len); | |
1059 | return -ENOMEM; | |
1060 | } | |
1061 | ||
1062 | if (coll) { | |
1063 | req_data->coll = coll; | |
1064 | req_data->coll_index = coll_index; | |
1065 | } | |
602adf40 | 1066 | |
f7760dad AE |
1067 | dout("rbd_do_request object_name=%s ofs=%llu len=%llu coll=%p[%d]\n", |
1068 | object_name, (unsigned long long) ofs, | |
1069 | (unsigned long long) len, coll, coll_index); | |
602adf40 | 1070 | |
0ce1a794 | 1071 | osdc = &rbd_dev->rbd_client->client->osdc; |
1dbb4399 AE |
1072 | req = ceph_osdc_alloc_request(osdc, flags, snapc, ops, |
1073 | false, GFP_NOIO, pages, bio); | |
4ad12621 | 1074 | if (!req) { |
4ad12621 | 1075 | ret = -ENOMEM; |
602adf40 YS |
1076 | goto done_pages; |
1077 | } | |
1078 | ||
1079 | req->r_callback = rbd_cb; | |
1080 | ||
1081 | req_data->rq = rq; | |
1082 | req_data->bio = bio; | |
1083 | req_data->pages = pages; | |
1084 | req_data->len = len; | |
1085 | ||
1086 | req->r_priv = req_data; | |
1087 | ||
1088 | reqhead = req->r_request->front.iov_base; | |
1089 | reqhead->snapid = cpu_to_le64(CEPH_NOSNAP); | |
1090 | ||
aded07ea | 1091 | strncpy(req->r_oid, object_name, sizeof(req->r_oid)); |
602adf40 YS |
1092 | req->r_oid_len = strlen(req->r_oid); |
1093 | ||
1094 | layout = &req->r_file_layout; | |
1095 | memset(layout, 0, sizeof(*layout)); | |
1096 | layout->fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
1097 | layout->fl_stripe_count = cpu_to_le32(1); | |
1098 | layout->fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
86992098 | 1099 | layout->fl_pg_pool = cpu_to_le32((int) rbd_dev->pool_id); |
6cae3717 SW |
1100 | ret = ceph_calc_raw_layout(osdc, layout, snapid, ofs, &len, &bno, |
1101 | req, ops); | |
1102 | rbd_assert(ret == 0); | |
602adf40 YS |
1103 | |
1104 | ceph_osdc_build_request(req, ofs, &len, | |
1105 | ops, | |
1106 | snapc, | |
1107 | &mtime, | |
1108 | req->r_oid, req->r_oid_len); | |
602adf40 | 1109 | |
59c2be1e | 1110 | if (linger_req) { |
1dbb4399 | 1111 | ceph_osdc_set_request_linger(osdc, req); |
59c2be1e YS |
1112 | *linger_req = req; |
1113 | } | |
1114 | ||
1dbb4399 | 1115 | ret = ceph_osdc_start_request(osdc, req, false); |
602adf40 YS |
1116 | if (ret < 0) |
1117 | goto done_err; | |
1118 | ||
1119 | if (!rbd_cb) { | |
1dbb4399 | 1120 | ret = ceph_osdc_wait_request(osdc, req); |
59c2be1e YS |
1121 | if (ver) |
1122 | *ver = le64_to_cpu(req->r_reassert_version.version); | |
bd919d45 AE |
1123 | dout("reassert_ver=%llu\n", |
1124 | (unsigned long long) | |
1125 | le64_to_cpu(req->r_reassert_version.version)); | |
602adf40 YS |
1126 | ceph_osdc_put_request(req); |
1127 | } | |
1128 | return ret; | |
1129 | ||
1130 | done_err: | |
1131 | bio_chain_put(req_data->bio); | |
1132 | ceph_osdc_put_request(req); | |
1133 | done_pages: | |
1fec7093 | 1134 | rbd_coll_end_req(req_data, ret, len); |
602adf40 | 1135 | kfree(req_data); |
602adf40 YS |
1136 | return ret; |
1137 | } | |
1138 | ||
1139 | /* | |
1140 | * Ceph osd op callback | |
1141 | */ | |
1142 | static void rbd_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) | |
1143 | { | |
1144 | struct rbd_request *req_data = req->r_priv; | |
1145 | struct ceph_osd_reply_head *replyhead; | |
1146 | struct ceph_osd_op *op; | |
1147 | __s32 rc; | |
1148 | u64 bytes; | |
1149 | int read_op; | |
1150 | ||
1151 | /* parse reply */ | |
1152 | replyhead = msg->front.iov_base; | |
1153 | WARN_ON(le32_to_cpu(replyhead->num_ops) == 0); | |
1154 | op = (void *)(replyhead + 1); | |
1155 | rc = le32_to_cpu(replyhead->result); | |
1156 | bytes = le64_to_cpu(op->extent.length); | |
895cfcc8 | 1157 | read_op = (le16_to_cpu(op->op) == CEPH_OSD_OP_READ); |
602adf40 | 1158 | |
bd919d45 AE |
1159 | dout("rbd_req_cb bytes=%llu readop=%d rc=%d\n", |
1160 | (unsigned long long) bytes, read_op, (int) rc); | |
602adf40 YS |
1161 | |
1162 | if (rc == -ENOENT && read_op) { | |
1163 | zero_bio_chain(req_data->bio, 0); | |
1164 | rc = 0; | |
1165 | } else if (rc == 0 && read_op && bytes < req_data->len) { | |
1166 | zero_bio_chain(req_data->bio, bytes); | |
1167 | bytes = req_data->len; | |
1168 | } | |
1169 | ||
1fec7093 | 1170 | rbd_coll_end_req(req_data, rc, bytes); |
602adf40 YS |
1171 | |
1172 | if (req_data->bio) | |
1173 | bio_chain_put(req_data->bio); | |
1174 | ||
1175 | ceph_osdc_put_request(req); | |
1176 | kfree(req_data); | |
1177 | } | |
1178 | ||
59c2be1e YS |
1179 | static void rbd_simple_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) |
1180 | { | |
1181 | ceph_osdc_put_request(req); | |
1182 | } | |
1183 | ||
602adf40 YS |
1184 | /* |
1185 | * Do a synchronous ceph osd operation | |
1186 | */ | |
0ce1a794 | 1187 | static int rbd_req_sync_op(struct rbd_device *rbd_dev, |
602adf40 YS |
1188 | struct ceph_snap_context *snapc, |
1189 | u64 snapid, | |
602adf40 | 1190 | int flags, |
913d2fdc | 1191 | struct ceph_osd_req_op *ops, |
aded07ea | 1192 | const char *object_name, |
f8d4de6e AE |
1193 | u64 ofs, u64 inbound_size, |
1194 | char *inbound, | |
59c2be1e YS |
1195 | struct ceph_osd_request **linger_req, |
1196 | u64 *ver) | |
602adf40 YS |
1197 | { |
1198 | int ret; | |
1199 | struct page **pages; | |
1200 | int num_pages; | |
913d2fdc | 1201 | |
aafb230e | 1202 | rbd_assert(ops != NULL); |
602adf40 | 1203 | |
f8d4de6e | 1204 | num_pages = calc_pages_for(ofs, inbound_size); |
602adf40 | 1205 | pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); |
b8d0638a DC |
1206 | if (IS_ERR(pages)) |
1207 | return PTR_ERR(pages); | |
602adf40 | 1208 | |
0ce1a794 | 1209 | ret = rbd_do_request(NULL, rbd_dev, snapc, snapid, |
f8d4de6e | 1210 | object_name, ofs, inbound_size, NULL, |
602adf40 YS |
1211 | pages, num_pages, |
1212 | flags, | |
1213 | ops, | |
1fec7093 | 1214 | NULL, 0, |
59c2be1e YS |
1215 | NULL, |
1216 | linger_req, ver); | |
602adf40 | 1217 | if (ret < 0) |
913d2fdc | 1218 | goto done; |
602adf40 | 1219 | |
f8d4de6e AE |
1220 | if ((flags & CEPH_OSD_FLAG_READ) && inbound) |
1221 | ret = ceph_copy_from_page_vector(pages, inbound, ofs, ret); | |
602adf40 | 1222 | |
602adf40 YS |
1223 | done: |
1224 | ceph_release_page_vector(pages, num_pages); | |
1225 | return ret; | |
1226 | } | |
1227 | ||
1228 | /* | |
1229 | * Do an asynchronous ceph osd operation | |
1230 | */ | |
1231 | static int rbd_do_op(struct request *rq, | |
0ce1a794 | 1232 | struct rbd_device *rbd_dev, |
602adf40 | 1233 | struct ceph_snap_context *snapc, |
602adf40 | 1234 | u64 ofs, u64 len, |
1fec7093 YS |
1235 | struct bio *bio, |
1236 | struct rbd_req_coll *coll, | |
1237 | int coll_index) | |
602adf40 YS |
1238 | { |
1239 | char *seg_name; | |
1240 | u64 seg_ofs; | |
1241 | u64 seg_len; | |
1242 | int ret; | |
1243 | struct ceph_osd_req_op *ops; | |
1244 | u32 payload_len; | |
ff2e4bb5 AE |
1245 | int opcode; |
1246 | int flags; | |
4634246d | 1247 | u64 snapid; |
602adf40 | 1248 | |
65ccfe21 | 1249 | seg_name = rbd_segment_name(rbd_dev, ofs); |
602adf40 YS |
1250 | if (!seg_name) |
1251 | return -ENOMEM; | |
65ccfe21 AE |
1252 | seg_len = rbd_segment_length(rbd_dev, ofs, len); |
1253 | seg_ofs = rbd_segment_offset(rbd_dev, ofs); | |
602adf40 | 1254 | |
ff2e4bb5 AE |
1255 | if (rq_data_dir(rq) == WRITE) { |
1256 | opcode = CEPH_OSD_OP_WRITE; | |
1257 | flags = CEPH_OSD_FLAG_WRITE|CEPH_OSD_FLAG_ONDISK; | |
4634246d | 1258 | snapid = CEPH_NOSNAP; |
ff2e4bb5 AE |
1259 | payload_len = seg_len; |
1260 | } else { | |
1261 | opcode = CEPH_OSD_OP_READ; | |
1262 | flags = CEPH_OSD_FLAG_READ; | |
4634246d | 1263 | snapc = NULL; |
971f839a | 1264 | snapid = rbd_dev->snap_id; |
ff2e4bb5 AE |
1265 | payload_len = 0; |
1266 | } | |
602adf40 | 1267 | |
57cfc106 AE |
1268 | ret = -ENOMEM; |
1269 | ops = rbd_create_rw_ops(1, opcode, payload_len); | |
1270 | if (!ops) | |
602adf40 YS |
1271 | goto done; |
1272 | ||
1273 | /* we've taken care of segment sizes earlier when we | |
1274 | cloned the bios. We should never have a segment | |
1275 | truncated at this point */ | |
aafb230e | 1276 | rbd_assert(seg_len == len); |
602adf40 YS |
1277 | |
1278 | ret = rbd_do_request(rq, rbd_dev, snapc, snapid, | |
1279 | seg_name, seg_ofs, seg_len, | |
1280 | bio, | |
1281 | NULL, 0, | |
1282 | flags, | |
1283 | ops, | |
1fec7093 | 1284 | coll, coll_index, |
59c2be1e | 1285 | rbd_req_cb, 0, NULL); |
11f77002 SW |
1286 | |
1287 | rbd_destroy_ops(ops); | |
602adf40 YS |
1288 | done: |
1289 | kfree(seg_name); | |
1290 | return ret; | |
1291 | } | |
1292 | ||
602adf40 YS |
1293 | /* |
1294 | * Request sync osd read | |
1295 | */ | |
0ce1a794 | 1296 | static int rbd_req_sync_read(struct rbd_device *rbd_dev, |
602adf40 | 1297 | u64 snapid, |
aded07ea | 1298 | const char *object_name, |
602adf40 | 1299 | u64 ofs, u64 len, |
59c2be1e YS |
1300 | char *buf, |
1301 | u64 *ver) | |
602adf40 | 1302 | { |
913d2fdc AE |
1303 | struct ceph_osd_req_op *ops; |
1304 | int ret; | |
1305 | ||
1306 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_READ, 0); | |
1307 | if (!ops) | |
1308 | return -ENOMEM; | |
1309 | ||
1310 | ret = rbd_req_sync_op(rbd_dev, NULL, | |
b06e6a6b | 1311 | snapid, |
602adf40 | 1312 | CEPH_OSD_FLAG_READ, |
913d2fdc AE |
1313 | ops, object_name, ofs, len, buf, NULL, ver); |
1314 | rbd_destroy_ops(ops); | |
1315 | ||
1316 | return ret; | |
602adf40 YS |
1317 | } |
1318 | ||
1319 | /* | |
59c2be1e YS |
1320 | * Request sync osd watch |
1321 | */ | |
0ce1a794 | 1322 | static int rbd_req_sync_notify_ack(struct rbd_device *rbd_dev, |
59c2be1e | 1323 | u64 ver, |
7f0a24d8 | 1324 | u64 notify_id) |
59c2be1e YS |
1325 | { |
1326 | struct ceph_osd_req_op *ops; | |
11f77002 SW |
1327 | int ret; |
1328 | ||
57cfc106 AE |
1329 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_NOTIFY_ACK, 0); |
1330 | if (!ops) | |
1331 | return -ENOMEM; | |
59c2be1e | 1332 | |
a71b891b | 1333 | ops[0].watch.ver = cpu_to_le64(ver); |
59c2be1e YS |
1334 | ops[0].watch.cookie = notify_id; |
1335 | ops[0].watch.flag = 0; | |
1336 | ||
0ce1a794 | 1337 | ret = rbd_do_request(NULL, rbd_dev, NULL, CEPH_NOSNAP, |
7f0a24d8 | 1338 | rbd_dev->header_name, 0, 0, NULL, |
ad4f232f | 1339 | NULL, 0, |
59c2be1e YS |
1340 | CEPH_OSD_FLAG_READ, |
1341 | ops, | |
1fec7093 | 1342 | NULL, 0, |
59c2be1e YS |
1343 | rbd_simple_req_cb, 0, NULL); |
1344 | ||
1345 | rbd_destroy_ops(ops); | |
1346 | return ret; | |
1347 | } | |
1348 | ||
1349 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) | |
1350 | { | |
0ce1a794 | 1351 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
a71b891b | 1352 | u64 hver; |
13143d2d SW |
1353 | int rc; |
1354 | ||
0ce1a794 | 1355 | if (!rbd_dev) |
59c2be1e YS |
1356 | return; |
1357 | ||
bd919d45 AE |
1358 | dout("rbd_watch_cb %s notify_id=%llu opcode=%u\n", |
1359 | rbd_dev->header_name, (unsigned long long) notify_id, | |
1360 | (unsigned int) opcode); | |
117973fb | 1361 | rc = rbd_dev_refresh(rbd_dev, &hver); |
13143d2d | 1362 | if (rc) |
f0f8cef5 | 1363 | pr_warning(RBD_DRV_NAME "%d got notification but failed to " |
0ce1a794 | 1364 | " update snaps: %d\n", rbd_dev->major, rc); |
59c2be1e | 1365 | |
7f0a24d8 | 1366 | rbd_req_sync_notify_ack(rbd_dev, hver, notify_id); |
59c2be1e YS |
1367 | } |
1368 | ||
1369 | /* | |
1370 | * Request sync osd watch | |
1371 | */ | |
0e6f322d | 1372 | static int rbd_req_sync_watch(struct rbd_device *rbd_dev) |
59c2be1e YS |
1373 | { |
1374 | struct ceph_osd_req_op *ops; | |
0ce1a794 | 1375 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
57cfc106 | 1376 | int ret; |
59c2be1e | 1377 | |
57cfc106 AE |
1378 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_WATCH, 0); |
1379 | if (!ops) | |
1380 | return -ENOMEM; | |
59c2be1e YS |
1381 | |
1382 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, 0, | |
0ce1a794 | 1383 | (void *)rbd_dev, &rbd_dev->watch_event); |
59c2be1e YS |
1384 | if (ret < 0) |
1385 | goto fail; | |
1386 | ||
0e6f322d | 1387 | ops[0].watch.ver = cpu_to_le64(rbd_dev->header.obj_version); |
0ce1a794 | 1388 | ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); |
59c2be1e YS |
1389 | ops[0].watch.flag = 1; |
1390 | ||
0ce1a794 | 1391 | ret = rbd_req_sync_op(rbd_dev, NULL, |
59c2be1e | 1392 | CEPH_NOSNAP, |
59c2be1e YS |
1393 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
1394 | ops, | |
0e6f322d AE |
1395 | rbd_dev->header_name, |
1396 | 0, 0, NULL, | |
0ce1a794 | 1397 | &rbd_dev->watch_request, NULL); |
59c2be1e YS |
1398 | |
1399 | if (ret < 0) | |
1400 | goto fail_event; | |
1401 | ||
1402 | rbd_destroy_ops(ops); | |
1403 | return 0; | |
1404 | ||
1405 | fail_event: | |
0ce1a794 AE |
1406 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
1407 | rbd_dev->watch_event = NULL; | |
59c2be1e YS |
1408 | fail: |
1409 | rbd_destroy_ops(ops); | |
1410 | return ret; | |
1411 | } | |
1412 | ||
79e3057c YS |
1413 | /* |
1414 | * Request sync osd unwatch | |
1415 | */ | |
070c633f | 1416 | static int rbd_req_sync_unwatch(struct rbd_device *rbd_dev) |
79e3057c YS |
1417 | { |
1418 | struct ceph_osd_req_op *ops; | |
57cfc106 | 1419 | int ret; |
79e3057c | 1420 | |
57cfc106 AE |
1421 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_WATCH, 0); |
1422 | if (!ops) | |
1423 | return -ENOMEM; | |
79e3057c YS |
1424 | |
1425 | ops[0].watch.ver = 0; | |
0ce1a794 | 1426 | ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); |
79e3057c YS |
1427 | ops[0].watch.flag = 0; |
1428 | ||
0ce1a794 | 1429 | ret = rbd_req_sync_op(rbd_dev, NULL, |
79e3057c | 1430 | CEPH_NOSNAP, |
79e3057c YS |
1431 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
1432 | ops, | |
070c633f AE |
1433 | rbd_dev->header_name, |
1434 | 0, 0, NULL, NULL, NULL); | |
1435 | ||
79e3057c YS |
1436 | |
1437 | rbd_destroy_ops(ops); | |
0ce1a794 AE |
1438 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
1439 | rbd_dev->watch_event = NULL; | |
79e3057c YS |
1440 | return ret; |
1441 | } | |
1442 | ||
602adf40 | 1443 | /* |
3cb4a687 | 1444 | * Synchronous osd object method call |
602adf40 | 1445 | */ |
0ce1a794 | 1446 | static int rbd_req_sync_exec(struct rbd_device *rbd_dev, |
aded07ea AE |
1447 | const char *object_name, |
1448 | const char *class_name, | |
1449 | const char *method_name, | |
3cb4a687 AE |
1450 | const char *outbound, |
1451 | size_t outbound_size, | |
f8d4de6e AE |
1452 | char *inbound, |
1453 | size_t inbound_size, | |
3cb4a687 | 1454 | int flags, |
59c2be1e | 1455 | u64 *ver) |
602adf40 YS |
1456 | { |
1457 | struct ceph_osd_req_op *ops; | |
aded07ea AE |
1458 | int class_name_len = strlen(class_name); |
1459 | int method_name_len = strlen(method_name); | |
3cb4a687 | 1460 | int payload_size; |
57cfc106 AE |
1461 | int ret; |
1462 | ||
3cb4a687 AE |
1463 | /* |
1464 | * Any input parameters required by the method we're calling | |
1465 | * will be sent along with the class and method names as | |
1466 | * part of the message payload. That data and its size are | |
1467 | * supplied via the indata and indata_len fields (named from | |
1468 | * the perspective of the server side) in the OSD request | |
1469 | * operation. | |
1470 | */ | |
1471 | payload_size = class_name_len + method_name_len + outbound_size; | |
1472 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_CALL, payload_size); | |
57cfc106 AE |
1473 | if (!ops) |
1474 | return -ENOMEM; | |
602adf40 | 1475 | |
aded07ea AE |
1476 | ops[0].cls.class_name = class_name; |
1477 | ops[0].cls.class_len = (__u8) class_name_len; | |
1478 | ops[0].cls.method_name = method_name; | |
1479 | ops[0].cls.method_len = (__u8) method_name_len; | |
602adf40 | 1480 | ops[0].cls.argc = 0; |
3cb4a687 AE |
1481 | ops[0].cls.indata = outbound; |
1482 | ops[0].cls.indata_len = outbound_size; | |
602adf40 | 1483 | |
0ce1a794 | 1484 | ret = rbd_req_sync_op(rbd_dev, NULL, |
602adf40 | 1485 | CEPH_NOSNAP, |
3cb4a687 | 1486 | flags, ops, |
f8d4de6e AE |
1487 | object_name, 0, inbound_size, inbound, |
1488 | NULL, ver); | |
602adf40 YS |
1489 | |
1490 | rbd_destroy_ops(ops); | |
1491 | ||
1492 | dout("cls_exec returned %d\n", ret); | |
1493 | return ret; | |
1494 | } | |
1495 | ||
1fec7093 YS |
1496 | static struct rbd_req_coll *rbd_alloc_coll(int num_reqs) |
1497 | { | |
1498 | struct rbd_req_coll *coll = | |
1499 | kzalloc(sizeof(struct rbd_req_coll) + | |
1500 | sizeof(struct rbd_req_status) * num_reqs, | |
1501 | GFP_ATOMIC); | |
1502 | ||
1503 | if (!coll) | |
1504 | return NULL; | |
1505 | coll->total = num_reqs; | |
1506 | kref_init(&coll->kref); | |
1507 | return coll; | |
1508 | } | |
1509 | ||
602adf40 YS |
1510 | /* |
1511 | * block device queue callback | |
1512 | */ | |
1513 | static void rbd_rq_fn(struct request_queue *q) | |
1514 | { | |
1515 | struct rbd_device *rbd_dev = q->queuedata; | |
1516 | struct request *rq; | |
602adf40 | 1517 | |
00f1f36f | 1518 | while ((rq = blk_fetch_request(q))) { |
602adf40 | 1519 | struct bio *bio; |
602adf40 | 1520 | bool do_write; |
bd919d45 | 1521 | unsigned int size; |
602adf40 | 1522 | u64 ofs; |
1fec7093 YS |
1523 | int num_segs, cur_seg = 0; |
1524 | struct rbd_req_coll *coll; | |
d1d25646 | 1525 | struct ceph_snap_context *snapc; |
f7760dad | 1526 | unsigned int bio_offset; |
602adf40 | 1527 | |
602adf40 YS |
1528 | dout("fetched request\n"); |
1529 | ||
1530 | /* filter out block requests we don't understand */ | |
1531 | if ((rq->cmd_type != REQ_TYPE_FS)) { | |
1532 | __blk_end_request_all(rq, 0); | |
00f1f36f | 1533 | continue; |
602adf40 YS |
1534 | } |
1535 | ||
1536 | /* deduce our operation (read, write) */ | |
1537 | do_write = (rq_data_dir(rq) == WRITE); | |
f84344f3 | 1538 | if (do_write && rbd_dev->mapping.read_only) { |
602adf40 | 1539 | __blk_end_request_all(rq, -EROFS); |
00f1f36f | 1540 | continue; |
602adf40 YS |
1541 | } |
1542 | ||
1543 | spin_unlock_irq(q->queue_lock); | |
1544 | ||
d1d25646 | 1545 | down_read(&rbd_dev->header_rwsem); |
e88a36ec | 1546 | |
daba5fdb AE |
1547 | if (!rbd_dev->exists) { |
1548 | rbd_assert(rbd_dev->snap_id != CEPH_NOSNAP); | |
e88a36ec | 1549 | up_read(&rbd_dev->header_rwsem); |
d1d25646 JD |
1550 | dout("request for non-existent snapshot"); |
1551 | spin_lock_irq(q->queue_lock); | |
1552 | __blk_end_request_all(rq, -ENXIO); | |
1553 | continue; | |
e88a36ec JD |
1554 | } |
1555 | ||
d1d25646 JD |
1556 | snapc = ceph_get_snap_context(rbd_dev->header.snapc); |
1557 | ||
1558 | up_read(&rbd_dev->header_rwsem); | |
1559 | ||
f7760dad AE |
1560 | size = blk_rq_bytes(rq); |
1561 | ofs = blk_rq_pos(rq) * SECTOR_SIZE; | |
1562 | bio = rq->bio; | |
1563 | ||
602adf40 YS |
1564 | dout("%s 0x%x bytes at 0x%llx\n", |
1565 | do_write ? "write" : "read", | |
bd919d45 | 1566 | size, (unsigned long long) blk_rq_pos(rq) * SECTOR_SIZE); |
602adf40 | 1567 | |
1fec7093 | 1568 | num_segs = rbd_get_num_segments(&rbd_dev->header, ofs, size); |
df111be6 AE |
1569 | if (num_segs <= 0) { |
1570 | spin_lock_irq(q->queue_lock); | |
1571 | __blk_end_request_all(rq, num_segs); | |
1572 | ceph_put_snap_context(snapc); | |
1573 | continue; | |
1574 | } | |
1fec7093 YS |
1575 | coll = rbd_alloc_coll(num_segs); |
1576 | if (!coll) { | |
1577 | spin_lock_irq(q->queue_lock); | |
1578 | __blk_end_request_all(rq, -ENOMEM); | |
d1d25646 | 1579 | ceph_put_snap_context(snapc); |
00f1f36f | 1580 | continue; |
1fec7093 YS |
1581 | } |
1582 | ||
f7760dad | 1583 | bio_offset = 0; |
602adf40 | 1584 | do { |
f7760dad AE |
1585 | u64 limit = rbd_segment_length(rbd_dev, ofs, size); |
1586 | unsigned int chain_size; | |
1587 | struct bio *bio_chain; | |
1588 | ||
1589 | BUG_ON(limit > (u64) UINT_MAX); | |
1590 | chain_size = (unsigned int) limit; | |
bd919d45 | 1591 | dout("rq->bio->bi_vcnt=%hu\n", rq->bio->bi_vcnt); |
f7760dad | 1592 | |
1fec7093 | 1593 | kref_get(&coll->kref); |
f7760dad AE |
1594 | |
1595 | /* Pass a cloned bio chain via an osd request */ | |
1596 | ||
1597 | bio_chain = bio_chain_clone_range(&bio, | |
1598 | &bio_offset, chain_size, | |
1599 | GFP_ATOMIC); | |
1600 | if (bio_chain) | |
4634246d | 1601 | (void) rbd_do_op(rq, rbd_dev, snapc, |
f7760dad AE |
1602 | ofs, chain_size, |
1603 | bio_chain, coll, cur_seg); | |
4634246d | 1604 | else |
1fec7093 | 1605 | rbd_coll_end_req_index(rq, coll, cur_seg, |
f7760dad AE |
1606 | -ENOMEM, chain_size); |
1607 | size -= chain_size; | |
1608 | ofs += chain_size; | |
602adf40 | 1609 | |
1fec7093 | 1610 | cur_seg++; |
602adf40 | 1611 | } while (size > 0); |
1fec7093 | 1612 | kref_put(&coll->kref, rbd_coll_release); |
602adf40 | 1613 | |
602adf40 | 1614 | spin_lock_irq(q->queue_lock); |
d1d25646 JD |
1615 | |
1616 | ceph_put_snap_context(snapc); | |
602adf40 YS |
1617 | } |
1618 | } | |
1619 | ||
1620 | /* | |
1621 | * a queue callback. Makes sure that we don't create a bio that spans across | |
1622 | * multiple osd objects. One exception would be with a single page bios, | |
f7760dad | 1623 | * which we handle later at bio_chain_clone_range() |
602adf40 YS |
1624 | */ |
1625 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, | |
1626 | struct bio_vec *bvec) | |
1627 | { | |
1628 | struct rbd_device *rbd_dev = q->queuedata; | |
e5cfeed2 AE |
1629 | sector_t sector_offset; |
1630 | sector_t sectors_per_obj; | |
1631 | sector_t obj_sector_offset; | |
1632 | int ret; | |
1633 | ||
1634 | /* | |
1635 | * Find how far into its rbd object the partition-relative | |
1636 | * bio start sector is to offset relative to the enclosing | |
1637 | * device. | |
1638 | */ | |
1639 | sector_offset = get_start_sect(bmd->bi_bdev) + bmd->bi_sector; | |
1640 | sectors_per_obj = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); | |
1641 | obj_sector_offset = sector_offset & (sectors_per_obj - 1); | |
1642 | ||
1643 | /* | |
1644 | * Compute the number of bytes from that offset to the end | |
1645 | * of the object. Account for what's already used by the bio. | |
1646 | */ | |
1647 | ret = (int) (sectors_per_obj - obj_sector_offset) << SECTOR_SHIFT; | |
1648 | if (ret > bmd->bi_size) | |
1649 | ret -= bmd->bi_size; | |
1650 | else | |
1651 | ret = 0; | |
1652 | ||
1653 | /* | |
1654 | * Don't send back more than was asked for. And if the bio | |
1655 | * was empty, let the whole thing through because: "Note | |
1656 | * that a block device *must* allow a single page to be | |
1657 | * added to an empty bio." | |
1658 | */ | |
1659 | rbd_assert(bvec->bv_len <= PAGE_SIZE); | |
1660 | if (ret > (int) bvec->bv_len || !bmd->bi_size) | |
1661 | ret = (int) bvec->bv_len; | |
1662 | ||
1663 | return ret; | |
602adf40 YS |
1664 | } |
1665 | ||
1666 | static void rbd_free_disk(struct rbd_device *rbd_dev) | |
1667 | { | |
1668 | struct gendisk *disk = rbd_dev->disk; | |
1669 | ||
1670 | if (!disk) | |
1671 | return; | |
1672 | ||
602adf40 YS |
1673 | if (disk->flags & GENHD_FL_UP) |
1674 | del_gendisk(disk); | |
1675 | if (disk->queue) | |
1676 | blk_cleanup_queue(disk->queue); | |
1677 | put_disk(disk); | |
1678 | } | |
1679 | ||
1680 | /* | |
4156d998 AE |
1681 | * Read the complete header for the given rbd device. |
1682 | * | |
1683 | * Returns a pointer to a dynamically-allocated buffer containing | |
1684 | * the complete and validated header. Caller can pass the address | |
1685 | * of a variable that will be filled in with the version of the | |
1686 | * header object at the time it was read. | |
1687 | * | |
1688 | * Returns a pointer-coded errno if a failure occurs. | |
602adf40 | 1689 | */ |
4156d998 AE |
1690 | static struct rbd_image_header_ondisk * |
1691 | rbd_dev_v1_header_read(struct rbd_device *rbd_dev, u64 *version) | |
602adf40 | 1692 | { |
4156d998 | 1693 | struct rbd_image_header_ondisk *ondisk = NULL; |
50f7c4c9 | 1694 | u32 snap_count = 0; |
4156d998 AE |
1695 | u64 names_size = 0; |
1696 | u32 want_count; | |
1697 | int ret; | |
602adf40 | 1698 | |
00f1f36f | 1699 | /* |
4156d998 AE |
1700 | * The complete header will include an array of its 64-bit |
1701 | * snapshot ids, followed by the names of those snapshots as | |
1702 | * a contiguous block of NUL-terminated strings. Note that | |
1703 | * the number of snapshots could change by the time we read | |
1704 | * it in, in which case we re-read it. | |
00f1f36f | 1705 | */ |
4156d998 AE |
1706 | do { |
1707 | size_t size; | |
1708 | ||
1709 | kfree(ondisk); | |
1710 | ||
1711 | size = sizeof (*ondisk); | |
1712 | size += snap_count * sizeof (struct rbd_image_snap_ondisk); | |
1713 | size += names_size; | |
1714 | ondisk = kmalloc(size, GFP_KERNEL); | |
1715 | if (!ondisk) | |
1716 | return ERR_PTR(-ENOMEM); | |
1717 | ||
1718 | ret = rbd_req_sync_read(rbd_dev, CEPH_NOSNAP, | |
0bed54dc | 1719 | rbd_dev->header_name, |
4156d998 AE |
1720 | 0, size, |
1721 | (char *) ondisk, version); | |
1722 | ||
1723 | if (ret < 0) | |
1724 | goto out_err; | |
1725 | if (WARN_ON((size_t) ret < size)) { | |
1726 | ret = -ENXIO; | |
1727 | pr_warning("short header read for image %s" | |
1728 | " (want %zd got %d)\n", | |
1729 | rbd_dev->image_name, size, ret); | |
1730 | goto out_err; | |
1731 | } | |
1732 | if (!rbd_dev_ondisk_valid(ondisk)) { | |
1733 | ret = -ENXIO; | |
1734 | pr_warning("invalid header for image %s\n", | |
1735 | rbd_dev->image_name); | |
1736 | goto out_err; | |
81e759fb | 1737 | } |
602adf40 | 1738 | |
4156d998 AE |
1739 | names_size = le64_to_cpu(ondisk->snap_names_len); |
1740 | want_count = snap_count; | |
1741 | snap_count = le32_to_cpu(ondisk->snap_count); | |
1742 | } while (snap_count != want_count); | |
00f1f36f | 1743 | |
4156d998 | 1744 | return ondisk; |
00f1f36f | 1745 | |
4156d998 AE |
1746 | out_err: |
1747 | kfree(ondisk); | |
1748 | ||
1749 | return ERR_PTR(ret); | |
1750 | } | |
1751 | ||
1752 | /* | |
1753 | * reload the ondisk the header | |
1754 | */ | |
1755 | static int rbd_read_header(struct rbd_device *rbd_dev, | |
1756 | struct rbd_image_header *header) | |
1757 | { | |
1758 | struct rbd_image_header_ondisk *ondisk; | |
1759 | u64 ver = 0; | |
1760 | int ret; | |
602adf40 | 1761 | |
4156d998 AE |
1762 | ondisk = rbd_dev_v1_header_read(rbd_dev, &ver); |
1763 | if (IS_ERR(ondisk)) | |
1764 | return PTR_ERR(ondisk); | |
1765 | ret = rbd_header_from_disk(header, ondisk); | |
1766 | if (ret >= 0) | |
1767 | header->obj_version = ver; | |
1768 | kfree(ondisk); | |
1769 | ||
1770 | return ret; | |
602adf40 YS |
1771 | } |
1772 | ||
41f38c2b | 1773 | static void rbd_remove_all_snaps(struct rbd_device *rbd_dev) |
dfc5606d YS |
1774 | { |
1775 | struct rbd_snap *snap; | |
a0593290 | 1776 | struct rbd_snap *next; |
dfc5606d | 1777 | |
a0593290 | 1778 | list_for_each_entry_safe(snap, next, &rbd_dev->snaps, node) |
41f38c2b | 1779 | rbd_remove_snap_dev(snap); |
dfc5606d YS |
1780 | } |
1781 | ||
9478554a AE |
1782 | static void rbd_update_mapping_size(struct rbd_device *rbd_dev) |
1783 | { | |
1784 | sector_t size; | |
1785 | ||
971f839a | 1786 | if (rbd_dev->snap_id != CEPH_NOSNAP) |
9478554a AE |
1787 | return; |
1788 | ||
1789 | size = (sector_t) rbd_dev->header.image_size / SECTOR_SIZE; | |
1790 | dout("setting size to %llu sectors", (unsigned long long) size); | |
1791 | rbd_dev->mapping.size = (u64) size; | |
1792 | set_capacity(rbd_dev->disk, size); | |
1793 | } | |
1794 | ||
602adf40 YS |
1795 | /* |
1796 | * only read the first part of the ondisk header, without the snaps info | |
1797 | */ | |
117973fb | 1798 | static int rbd_dev_v1_refresh(struct rbd_device *rbd_dev, u64 *hver) |
602adf40 YS |
1799 | { |
1800 | int ret; | |
1801 | struct rbd_image_header h; | |
602adf40 YS |
1802 | |
1803 | ret = rbd_read_header(rbd_dev, &h); | |
1804 | if (ret < 0) | |
1805 | return ret; | |
1806 | ||
a51aa0c0 JD |
1807 | down_write(&rbd_dev->header_rwsem); |
1808 | ||
9478554a AE |
1809 | /* Update image size, and check for resize of mapped image */ |
1810 | rbd_dev->header.image_size = h.image_size; | |
1811 | rbd_update_mapping_size(rbd_dev); | |
9db4b3e3 | 1812 | |
849b4260 | 1813 | /* rbd_dev->header.object_prefix shouldn't change */ |
602adf40 | 1814 | kfree(rbd_dev->header.snap_sizes); |
849b4260 | 1815 | kfree(rbd_dev->header.snap_names); |
d1d25646 JD |
1816 | /* osd requests may still refer to snapc */ |
1817 | ceph_put_snap_context(rbd_dev->header.snapc); | |
602adf40 | 1818 | |
b813623a AE |
1819 | if (hver) |
1820 | *hver = h.obj_version; | |
a71b891b | 1821 | rbd_dev->header.obj_version = h.obj_version; |
93a24e08 | 1822 | rbd_dev->header.image_size = h.image_size; |
602adf40 YS |
1823 | rbd_dev->header.snapc = h.snapc; |
1824 | rbd_dev->header.snap_names = h.snap_names; | |
1825 | rbd_dev->header.snap_sizes = h.snap_sizes; | |
849b4260 AE |
1826 | /* Free the extra copy of the object prefix */ |
1827 | WARN_ON(strcmp(rbd_dev->header.object_prefix, h.object_prefix)); | |
1828 | kfree(h.object_prefix); | |
1829 | ||
304f6808 AE |
1830 | ret = rbd_dev_snaps_update(rbd_dev); |
1831 | if (!ret) | |
1832 | ret = rbd_dev_snaps_register(rbd_dev); | |
dfc5606d | 1833 | |
c666601a | 1834 | up_write(&rbd_dev->header_rwsem); |
602adf40 | 1835 | |
dfc5606d | 1836 | return ret; |
602adf40 YS |
1837 | } |
1838 | ||
117973fb | 1839 | static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver) |
1fe5e993 AE |
1840 | { |
1841 | int ret; | |
1842 | ||
117973fb | 1843 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
1fe5e993 | 1844 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
117973fb AE |
1845 | if (rbd_dev->image_format == 1) |
1846 | ret = rbd_dev_v1_refresh(rbd_dev, hver); | |
1847 | else | |
1848 | ret = rbd_dev_v2_refresh(rbd_dev, hver); | |
1fe5e993 AE |
1849 | mutex_unlock(&ctl_mutex); |
1850 | ||
1851 | return ret; | |
1852 | } | |
1853 | ||
602adf40 YS |
1854 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
1855 | { | |
1856 | struct gendisk *disk; | |
1857 | struct request_queue *q; | |
593a9e7b | 1858 | u64 segment_size; |
602adf40 | 1859 | |
602adf40 | 1860 | /* create gendisk info */ |
602adf40 YS |
1861 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); |
1862 | if (!disk) | |
1fcdb8aa | 1863 | return -ENOMEM; |
602adf40 | 1864 | |
f0f8cef5 | 1865 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
de71a297 | 1866 | rbd_dev->dev_id); |
602adf40 YS |
1867 | disk->major = rbd_dev->major; |
1868 | disk->first_minor = 0; | |
1869 | disk->fops = &rbd_bd_ops; | |
1870 | disk->private_data = rbd_dev; | |
1871 | ||
1872 | /* init rq */ | |
602adf40 YS |
1873 | q = blk_init_queue(rbd_rq_fn, &rbd_dev->lock); |
1874 | if (!q) | |
1875 | goto out_disk; | |
029bcbd8 | 1876 | |
593a9e7b AE |
1877 | /* We use the default size, but let's be explicit about it. */ |
1878 | blk_queue_physical_block_size(q, SECTOR_SIZE); | |
1879 | ||
029bcbd8 | 1880 | /* set io sizes to object size */ |
593a9e7b AE |
1881 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
1882 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); | |
1883 | blk_queue_max_segment_size(q, segment_size); | |
1884 | blk_queue_io_min(q, segment_size); | |
1885 | blk_queue_io_opt(q, segment_size); | |
029bcbd8 | 1886 | |
602adf40 YS |
1887 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
1888 | disk->queue = q; | |
1889 | ||
1890 | q->queuedata = rbd_dev; | |
1891 | ||
1892 | rbd_dev->disk = disk; | |
602adf40 | 1893 | |
12f02944 AE |
1894 | set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); |
1895 | ||
602adf40 | 1896 | return 0; |
602adf40 YS |
1897 | out_disk: |
1898 | put_disk(disk); | |
1fcdb8aa AE |
1899 | |
1900 | return -ENOMEM; | |
602adf40 YS |
1901 | } |
1902 | ||
dfc5606d YS |
1903 | /* |
1904 | sysfs | |
1905 | */ | |
1906 | ||
593a9e7b AE |
1907 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
1908 | { | |
1909 | return container_of(dev, struct rbd_device, dev); | |
1910 | } | |
1911 | ||
dfc5606d YS |
1912 | static ssize_t rbd_size_show(struct device *dev, |
1913 | struct device_attribute *attr, char *buf) | |
1914 | { | |
593a9e7b | 1915 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
a51aa0c0 JD |
1916 | sector_t size; |
1917 | ||
1918 | down_read(&rbd_dev->header_rwsem); | |
1919 | size = get_capacity(rbd_dev->disk); | |
1920 | up_read(&rbd_dev->header_rwsem); | |
dfc5606d | 1921 | |
a51aa0c0 | 1922 | return sprintf(buf, "%llu\n", (unsigned long long) size * SECTOR_SIZE); |
dfc5606d YS |
1923 | } |
1924 | ||
34b13184 AE |
1925 | /* |
1926 | * Note this shows the features for whatever's mapped, which is not | |
1927 | * necessarily the base image. | |
1928 | */ | |
1929 | static ssize_t rbd_features_show(struct device *dev, | |
1930 | struct device_attribute *attr, char *buf) | |
1931 | { | |
1932 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
1933 | ||
1934 | return sprintf(buf, "0x%016llx\n", | |
1935 | (unsigned long long) rbd_dev->mapping.features); | |
1936 | } | |
1937 | ||
dfc5606d YS |
1938 | static ssize_t rbd_major_show(struct device *dev, |
1939 | struct device_attribute *attr, char *buf) | |
1940 | { | |
593a9e7b | 1941 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 1942 | |
dfc5606d YS |
1943 | return sprintf(buf, "%d\n", rbd_dev->major); |
1944 | } | |
1945 | ||
1946 | static ssize_t rbd_client_id_show(struct device *dev, | |
1947 | struct device_attribute *attr, char *buf) | |
602adf40 | 1948 | { |
593a9e7b | 1949 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 1950 | |
1dbb4399 AE |
1951 | return sprintf(buf, "client%lld\n", |
1952 | ceph_client_id(rbd_dev->rbd_client->client)); | |
602adf40 YS |
1953 | } |
1954 | ||
dfc5606d YS |
1955 | static ssize_t rbd_pool_show(struct device *dev, |
1956 | struct device_attribute *attr, char *buf) | |
602adf40 | 1957 | { |
593a9e7b | 1958 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d YS |
1959 | |
1960 | return sprintf(buf, "%s\n", rbd_dev->pool_name); | |
1961 | } | |
1962 | ||
9bb2f334 AE |
1963 | static ssize_t rbd_pool_id_show(struct device *dev, |
1964 | struct device_attribute *attr, char *buf) | |
1965 | { | |
1966 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
1967 | ||
86992098 | 1968 | return sprintf(buf, "%llu\n", (unsigned long long) rbd_dev->pool_id); |
9bb2f334 AE |
1969 | } |
1970 | ||
dfc5606d YS |
1971 | static ssize_t rbd_name_show(struct device *dev, |
1972 | struct device_attribute *attr, char *buf) | |
1973 | { | |
593a9e7b | 1974 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 1975 | |
0bed54dc | 1976 | return sprintf(buf, "%s\n", rbd_dev->image_name); |
dfc5606d YS |
1977 | } |
1978 | ||
589d30e0 AE |
1979 | static ssize_t rbd_image_id_show(struct device *dev, |
1980 | struct device_attribute *attr, char *buf) | |
1981 | { | |
1982 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
1983 | ||
1984 | return sprintf(buf, "%s\n", rbd_dev->image_id); | |
1985 | } | |
1986 | ||
34b13184 AE |
1987 | /* |
1988 | * Shows the name of the currently-mapped snapshot (or | |
1989 | * RBD_SNAP_HEAD_NAME for the base image). | |
1990 | */ | |
dfc5606d YS |
1991 | static ssize_t rbd_snap_show(struct device *dev, |
1992 | struct device_attribute *attr, | |
1993 | char *buf) | |
1994 | { | |
593a9e7b | 1995 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 1996 | |
971f839a | 1997 | return sprintf(buf, "%s\n", rbd_dev->snap_name); |
dfc5606d YS |
1998 | } |
1999 | ||
2000 | static ssize_t rbd_image_refresh(struct device *dev, | |
2001 | struct device_attribute *attr, | |
2002 | const char *buf, | |
2003 | size_t size) | |
2004 | { | |
593a9e7b | 2005 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
b813623a | 2006 | int ret; |
602adf40 | 2007 | |
117973fb | 2008 | ret = rbd_dev_refresh(rbd_dev, NULL); |
b813623a AE |
2009 | |
2010 | return ret < 0 ? ret : size; | |
dfc5606d | 2011 | } |
602adf40 | 2012 | |
dfc5606d | 2013 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
34b13184 | 2014 | static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL); |
dfc5606d YS |
2015 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); |
2016 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); | |
2017 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); | |
9bb2f334 | 2018 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
dfc5606d | 2019 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
589d30e0 | 2020 | static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL); |
dfc5606d YS |
2021 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); |
2022 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); | |
dfc5606d YS |
2023 | |
2024 | static struct attribute *rbd_attrs[] = { | |
2025 | &dev_attr_size.attr, | |
34b13184 | 2026 | &dev_attr_features.attr, |
dfc5606d YS |
2027 | &dev_attr_major.attr, |
2028 | &dev_attr_client_id.attr, | |
2029 | &dev_attr_pool.attr, | |
9bb2f334 | 2030 | &dev_attr_pool_id.attr, |
dfc5606d | 2031 | &dev_attr_name.attr, |
589d30e0 | 2032 | &dev_attr_image_id.attr, |
dfc5606d YS |
2033 | &dev_attr_current_snap.attr, |
2034 | &dev_attr_refresh.attr, | |
dfc5606d YS |
2035 | NULL |
2036 | }; | |
2037 | ||
2038 | static struct attribute_group rbd_attr_group = { | |
2039 | .attrs = rbd_attrs, | |
2040 | }; | |
2041 | ||
2042 | static const struct attribute_group *rbd_attr_groups[] = { | |
2043 | &rbd_attr_group, | |
2044 | NULL | |
2045 | }; | |
2046 | ||
2047 | static void rbd_sysfs_dev_release(struct device *dev) | |
2048 | { | |
2049 | } | |
2050 | ||
2051 | static struct device_type rbd_device_type = { | |
2052 | .name = "rbd", | |
2053 | .groups = rbd_attr_groups, | |
2054 | .release = rbd_sysfs_dev_release, | |
2055 | }; | |
2056 | ||
2057 | ||
2058 | /* | |
2059 | sysfs - snapshots | |
2060 | */ | |
2061 | ||
2062 | static ssize_t rbd_snap_size_show(struct device *dev, | |
2063 | struct device_attribute *attr, | |
2064 | char *buf) | |
2065 | { | |
2066 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2067 | ||
3591538f | 2068 | return sprintf(buf, "%llu\n", (unsigned long long)snap->size); |
dfc5606d YS |
2069 | } |
2070 | ||
2071 | static ssize_t rbd_snap_id_show(struct device *dev, | |
2072 | struct device_attribute *attr, | |
2073 | char *buf) | |
2074 | { | |
2075 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2076 | ||
3591538f | 2077 | return sprintf(buf, "%llu\n", (unsigned long long)snap->id); |
dfc5606d YS |
2078 | } |
2079 | ||
34b13184 AE |
2080 | static ssize_t rbd_snap_features_show(struct device *dev, |
2081 | struct device_attribute *attr, | |
2082 | char *buf) | |
2083 | { | |
2084 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2085 | ||
2086 | return sprintf(buf, "0x%016llx\n", | |
2087 | (unsigned long long) snap->features); | |
2088 | } | |
2089 | ||
dfc5606d YS |
2090 | static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); |
2091 | static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); | |
34b13184 | 2092 | static DEVICE_ATTR(snap_features, S_IRUGO, rbd_snap_features_show, NULL); |
dfc5606d YS |
2093 | |
2094 | static struct attribute *rbd_snap_attrs[] = { | |
2095 | &dev_attr_snap_size.attr, | |
2096 | &dev_attr_snap_id.attr, | |
34b13184 | 2097 | &dev_attr_snap_features.attr, |
dfc5606d YS |
2098 | NULL, |
2099 | }; | |
2100 | ||
2101 | static struct attribute_group rbd_snap_attr_group = { | |
2102 | .attrs = rbd_snap_attrs, | |
2103 | }; | |
2104 | ||
2105 | static void rbd_snap_dev_release(struct device *dev) | |
2106 | { | |
2107 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2108 | kfree(snap->name); | |
2109 | kfree(snap); | |
2110 | } | |
2111 | ||
2112 | static const struct attribute_group *rbd_snap_attr_groups[] = { | |
2113 | &rbd_snap_attr_group, | |
2114 | NULL | |
2115 | }; | |
2116 | ||
2117 | static struct device_type rbd_snap_device_type = { | |
2118 | .groups = rbd_snap_attr_groups, | |
2119 | .release = rbd_snap_dev_release, | |
2120 | }; | |
2121 | ||
304f6808 AE |
2122 | static bool rbd_snap_registered(struct rbd_snap *snap) |
2123 | { | |
2124 | bool ret = snap->dev.type == &rbd_snap_device_type; | |
2125 | bool reg = device_is_registered(&snap->dev); | |
2126 | ||
2127 | rbd_assert(!ret ^ reg); | |
2128 | ||
2129 | return ret; | |
2130 | } | |
2131 | ||
41f38c2b | 2132 | static void rbd_remove_snap_dev(struct rbd_snap *snap) |
dfc5606d YS |
2133 | { |
2134 | list_del(&snap->node); | |
304f6808 AE |
2135 | if (device_is_registered(&snap->dev)) |
2136 | device_unregister(&snap->dev); | |
dfc5606d YS |
2137 | } |
2138 | ||
14e7085d | 2139 | static int rbd_register_snap_dev(struct rbd_snap *snap, |
dfc5606d YS |
2140 | struct device *parent) |
2141 | { | |
2142 | struct device *dev = &snap->dev; | |
2143 | int ret; | |
2144 | ||
2145 | dev->type = &rbd_snap_device_type; | |
2146 | dev->parent = parent; | |
2147 | dev->release = rbd_snap_dev_release; | |
d4b125e9 | 2148 | dev_set_name(dev, "%s%s", RBD_SNAP_DEV_NAME_PREFIX, snap->name); |
304f6808 AE |
2149 | dout("%s: registering device for snapshot %s\n", __func__, snap->name); |
2150 | ||
dfc5606d YS |
2151 | ret = device_register(dev); |
2152 | ||
2153 | return ret; | |
2154 | } | |
2155 | ||
4e891e0a | 2156 | static struct rbd_snap *__rbd_add_snap_dev(struct rbd_device *rbd_dev, |
c8d18425 | 2157 | const char *snap_name, |
34b13184 AE |
2158 | u64 snap_id, u64 snap_size, |
2159 | u64 snap_features) | |
dfc5606d | 2160 | { |
4e891e0a | 2161 | struct rbd_snap *snap; |
dfc5606d | 2162 | int ret; |
4e891e0a AE |
2163 | |
2164 | snap = kzalloc(sizeof (*snap), GFP_KERNEL); | |
dfc5606d | 2165 | if (!snap) |
4e891e0a AE |
2166 | return ERR_PTR(-ENOMEM); |
2167 | ||
2168 | ret = -ENOMEM; | |
c8d18425 | 2169 | snap->name = kstrdup(snap_name, GFP_KERNEL); |
4e891e0a AE |
2170 | if (!snap->name) |
2171 | goto err; | |
2172 | ||
c8d18425 AE |
2173 | snap->id = snap_id; |
2174 | snap->size = snap_size; | |
34b13184 | 2175 | snap->features = snap_features; |
4e891e0a AE |
2176 | |
2177 | return snap; | |
2178 | ||
dfc5606d YS |
2179 | err: |
2180 | kfree(snap->name); | |
2181 | kfree(snap); | |
4e891e0a AE |
2182 | |
2183 | return ERR_PTR(ret); | |
dfc5606d YS |
2184 | } |
2185 | ||
cd892126 AE |
2186 | static char *rbd_dev_v1_snap_info(struct rbd_device *rbd_dev, u32 which, |
2187 | u64 *snap_size, u64 *snap_features) | |
2188 | { | |
2189 | char *snap_name; | |
2190 | ||
2191 | rbd_assert(which < rbd_dev->header.snapc->num_snaps); | |
2192 | ||
2193 | *snap_size = rbd_dev->header.snap_sizes[which]; | |
2194 | *snap_features = 0; /* No features for v1 */ | |
2195 | ||
2196 | /* Skip over names until we find the one we are looking for */ | |
2197 | ||
2198 | snap_name = rbd_dev->header.snap_names; | |
2199 | while (which--) | |
2200 | snap_name += strlen(snap_name) + 1; | |
2201 | ||
2202 | return snap_name; | |
2203 | } | |
2204 | ||
9d475de5 AE |
2205 | /* |
2206 | * Get the size and object order for an image snapshot, or if | |
2207 | * snap_id is CEPH_NOSNAP, gets this information for the base | |
2208 | * image. | |
2209 | */ | |
2210 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, | |
2211 | u8 *order, u64 *snap_size) | |
2212 | { | |
2213 | __le64 snapid = cpu_to_le64(snap_id); | |
2214 | int ret; | |
2215 | struct { | |
2216 | u8 order; | |
2217 | __le64 size; | |
2218 | } __attribute__ ((packed)) size_buf = { 0 }; | |
2219 | ||
2220 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2221 | "rbd", "get_size", | |
2222 | (char *) &snapid, sizeof (snapid), | |
2223 | (char *) &size_buf, sizeof (size_buf), | |
2224 | CEPH_OSD_FLAG_READ, NULL); | |
2225 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2226 | if (ret < 0) | |
2227 | return ret; | |
2228 | ||
2229 | *order = size_buf.order; | |
2230 | *snap_size = le64_to_cpu(size_buf.size); | |
2231 | ||
2232 | dout(" snap_id 0x%016llx order = %u, snap_size = %llu\n", | |
2233 | (unsigned long long) snap_id, (unsigned int) *order, | |
2234 | (unsigned long long) *snap_size); | |
2235 | ||
2236 | return 0; | |
2237 | } | |
2238 | ||
2239 | static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) | |
2240 | { | |
2241 | return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, | |
2242 | &rbd_dev->header.obj_order, | |
2243 | &rbd_dev->header.image_size); | |
2244 | } | |
2245 | ||
1e130199 AE |
2246 | static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) |
2247 | { | |
2248 | void *reply_buf; | |
2249 | int ret; | |
2250 | void *p; | |
2251 | ||
2252 | reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL); | |
2253 | if (!reply_buf) | |
2254 | return -ENOMEM; | |
2255 | ||
2256 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2257 | "rbd", "get_object_prefix", | |
2258 | NULL, 0, | |
2259 | reply_buf, RBD_OBJ_PREFIX_LEN_MAX, | |
2260 | CEPH_OSD_FLAG_READ, NULL); | |
2261 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2262 | if (ret < 0) | |
2263 | goto out; | |
a0ea3a40 | 2264 | ret = 0; /* rbd_req_sync_exec() can return positive */ |
1e130199 AE |
2265 | |
2266 | p = reply_buf; | |
2267 | rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, | |
2268 | p + RBD_OBJ_PREFIX_LEN_MAX, | |
2269 | NULL, GFP_NOIO); | |
2270 | ||
2271 | if (IS_ERR(rbd_dev->header.object_prefix)) { | |
2272 | ret = PTR_ERR(rbd_dev->header.object_prefix); | |
2273 | rbd_dev->header.object_prefix = NULL; | |
2274 | } else { | |
2275 | dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); | |
2276 | } | |
2277 | ||
2278 | out: | |
2279 | kfree(reply_buf); | |
2280 | ||
2281 | return ret; | |
2282 | } | |
2283 | ||
b1b5402a AE |
2284 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
2285 | u64 *snap_features) | |
2286 | { | |
2287 | __le64 snapid = cpu_to_le64(snap_id); | |
2288 | struct { | |
2289 | __le64 features; | |
2290 | __le64 incompat; | |
2291 | } features_buf = { 0 }; | |
d889140c | 2292 | u64 incompat; |
b1b5402a AE |
2293 | int ret; |
2294 | ||
2295 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2296 | "rbd", "get_features", | |
2297 | (char *) &snapid, sizeof (snapid), | |
2298 | (char *) &features_buf, sizeof (features_buf), | |
2299 | CEPH_OSD_FLAG_READ, NULL); | |
2300 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2301 | if (ret < 0) | |
2302 | return ret; | |
d889140c AE |
2303 | |
2304 | incompat = le64_to_cpu(features_buf.incompat); | |
2305 | if (incompat & ~RBD_FEATURES_ALL) | |
2306 | return -ENOTSUPP; | |
2307 | ||
b1b5402a AE |
2308 | *snap_features = le64_to_cpu(features_buf.features); |
2309 | ||
2310 | dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", | |
2311 | (unsigned long long) snap_id, | |
2312 | (unsigned long long) *snap_features, | |
2313 | (unsigned long long) le64_to_cpu(features_buf.incompat)); | |
2314 | ||
2315 | return 0; | |
2316 | } | |
2317 | ||
2318 | static int rbd_dev_v2_features(struct rbd_device *rbd_dev) | |
2319 | { | |
2320 | return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, | |
2321 | &rbd_dev->header.features); | |
2322 | } | |
2323 | ||
6e14b1a6 | 2324 | static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev, u64 *ver) |
35d489f9 AE |
2325 | { |
2326 | size_t size; | |
2327 | int ret; | |
2328 | void *reply_buf; | |
2329 | void *p; | |
2330 | void *end; | |
2331 | u64 seq; | |
2332 | u32 snap_count; | |
2333 | struct ceph_snap_context *snapc; | |
2334 | u32 i; | |
2335 | ||
2336 | /* | |
2337 | * We'll need room for the seq value (maximum snapshot id), | |
2338 | * snapshot count, and array of that many snapshot ids. | |
2339 | * For now we have a fixed upper limit on the number we're | |
2340 | * prepared to receive. | |
2341 | */ | |
2342 | size = sizeof (__le64) + sizeof (__le32) + | |
2343 | RBD_MAX_SNAP_COUNT * sizeof (__le64); | |
2344 | reply_buf = kzalloc(size, GFP_KERNEL); | |
2345 | if (!reply_buf) | |
2346 | return -ENOMEM; | |
2347 | ||
2348 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2349 | "rbd", "get_snapcontext", | |
2350 | NULL, 0, | |
2351 | reply_buf, size, | |
6e14b1a6 | 2352 | CEPH_OSD_FLAG_READ, ver); |
35d489f9 AE |
2353 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); |
2354 | if (ret < 0) | |
2355 | goto out; | |
2356 | ||
2357 | ret = -ERANGE; | |
2358 | p = reply_buf; | |
2359 | end = (char *) reply_buf + size; | |
2360 | ceph_decode_64_safe(&p, end, seq, out); | |
2361 | ceph_decode_32_safe(&p, end, snap_count, out); | |
2362 | ||
2363 | /* | |
2364 | * Make sure the reported number of snapshot ids wouldn't go | |
2365 | * beyond the end of our buffer. But before checking that, | |
2366 | * make sure the computed size of the snapshot context we | |
2367 | * allocate is representable in a size_t. | |
2368 | */ | |
2369 | if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) | |
2370 | / sizeof (u64)) { | |
2371 | ret = -EINVAL; | |
2372 | goto out; | |
2373 | } | |
2374 | if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) | |
2375 | goto out; | |
2376 | ||
2377 | size = sizeof (struct ceph_snap_context) + | |
2378 | snap_count * sizeof (snapc->snaps[0]); | |
2379 | snapc = kmalloc(size, GFP_KERNEL); | |
2380 | if (!snapc) { | |
2381 | ret = -ENOMEM; | |
2382 | goto out; | |
2383 | } | |
2384 | ||
2385 | atomic_set(&snapc->nref, 1); | |
2386 | snapc->seq = seq; | |
2387 | snapc->num_snaps = snap_count; | |
2388 | for (i = 0; i < snap_count; i++) | |
2389 | snapc->snaps[i] = ceph_decode_64(&p); | |
2390 | ||
2391 | rbd_dev->header.snapc = snapc; | |
2392 | ||
2393 | dout(" snap context seq = %llu, snap_count = %u\n", | |
2394 | (unsigned long long) seq, (unsigned int) snap_count); | |
2395 | ||
2396 | out: | |
2397 | kfree(reply_buf); | |
2398 | ||
2399 | return 0; | |
2400 | } | |
2401 | ||
b8b1e2db AE |
2402 | static char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, u32 which) |
2403 | { | |
2404 | size_t size; | |
2405 | void *reply_buf; | |
2406 | __le64 snap_id; | |
2407 | int ret; | |
2408 | void *p; | |
2409 | void *end; | |
b8b1e2db AE |
2410 | char *snap_name; |
2411 | ||
2412 | size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; | |
2413 | reply_buf = kmalloc(size, GFP_KERNEL); | |
2414 | if (!reply_buf) | |
2415 | return ERR_PTR(-ENOMEM); | |
2416 | ||
2417 | snap_id = cpu_to_le64(rbd_dev->header.snapc->snaps[which]); | |
2418 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, | |
2419 | "rbd", "get_snapshot_name", | |
2420 | (char *) &snap_id, sizeof (snap_id), | |
2421 | reply_buf, size, | |
2422 | CEPH_OSD_FLAG_READ, NULL); | |
2423 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2424 | if (ret < 0) | |
2425 | goto out; | |
2426 | ||
2427 | p = reply_buf; | |
2428 | end = (char *) reply_buf + size; | |
e5c35534 | 2429 | snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
b8b1e2db AE |
2430 | if (IS_ERR(snap_name)) { |
2431 | ret = PTR_ERR(snap_name); | |
2432 | goto out; | |
2433 | } else { | |
2434 | dout(" snap_id 0x%016llx snap_name = %s\n", | |
2435 | (unsigned long long) le64_to_cpu(snap_id), snap_name); | |
2436 | } | |
2437 | kfree(reply_buf); | |
2438 | ||
2439 | return snap_name; | |
2440 | out: | |
2441 | kfree(reply_buf); | |
2442 | ||
2443 | return ERR_PTR(ret); | |
2444 | } | |
2445 | ||
2446 | static char *rbd_dev_v2_snap_info(struct rbd_device *rbd_dev, u32 which, | |
2447 | u64 *snap_size, u64 *snap_features) | |
2448 | { | |
2449 | __le64 snap_id; | |
2450 | u8 order; | |
2451 | int ret; | |
2452 | ||
2453 | snap_id = rbd_dev->header.snapc->snaps[which]; | |
2454 | ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, &order, snap_size); | |
2455 | if (ret) | |
2456 | return ERR_PTR(ret); | |
2457 | ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, snap_features); | |
2458 | if (ret) | |
2459 | return ERR_PTR(ret); | |
2460 | ||
2461 | return rbd_dev_v2_snap_name(rbd_dev, which); | |
2462 | } | |
2463 | ||
2464 | static char *rbd_dev_snap_info(struct rbd_device *rbd_dev, u32 which, | |
2465 | u64 *snap_size, u64 *snap_features) | |
2466 | { | |
2467 | if (rbd_dev->image_format == 1) | |
2468 | return rbd_dev_v1_snap_info(rbd_dev, which, | |
2469 | snap_size, snap_features); | |
2470 | if (rbd_dev->image_format == 2) | |
2471 | return rbd_dev_v2_snap_info(rbd_dev, which, | |
2472 | snap_size, snap_features); | |
2473 | return ERR_PTR(-EINVAL); | |
2474 | } | |
2475 | ||
117973fb AE |
2476 | static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver) |
2477 | { | |
2478 | int ret; | |
2479 | __u8 obj_order; | |
2480 | ||
2481 | down_write(&rbd_dev->header_rwsem); | |
2482 | ||
2483 | /* Grab old order first, to see if it changes */ | |
2484 | ||
2485 | obj_order = rbd_dev->header.obj_order, | |
2486 | ret = rbd_dev_v2_image_size(rbd_dev); | |
2487 | if (ret) | |
2488 | goto out; | |
2489 | if (rbd_dev->header.obj_order != obj_order) { | |
2490 | ret = -EIO; | |
2491 | goto out; | |
2492 | } | |
2493 | rbd_update_mapping_size(rbd_dev); | |
2494 | ||
2495 | ret = rbd_dev_v2_snap_context(rbd_dev, hver); | |
2496 | dout("rbd_dev_v2_snap_context returned %d\n", ret); | |
2497 | if (ret) | |
2498 | goto out; | |
2499 | ret = rbd_dev_snaps_update(rbd_dev); | |
2500 | dout("rbd_dev_snaps_update returned %d\n", ret); | |
2501 | if (ret) | |
2502 | goto out; | |
2503 | ret = rbd_dev_snaps_register(rbd_dev); | |
2504 | dout("rbd_dev_snaps_register returned %d\n", ret); | |
2505 | out: | |
2506 | up_write(&rbd_dev->header_rwsem); | |
2507 | ||
2508 | return ret; | |
2509 | } | |
2510 | ||
dfc5606d | 2511 | /* |
35938150 AE |
2512 | * Scan the rbd device's current snapshot list and compare it to the |
2513 | * newly-received snapshot context. Remove any existing snapshots | |
2514 | * not present in the new snapshot context. Add a new snapshot for | |
2515 | * any snaphots in the snapshot context not in the current list. | |
2516 | * And verify there are no changes to snapshots we already know | |
2517 | * about. | |
2518 | * | |
2519 | * Assumes the snapshots in the snapshot context are sorted by | |
2520 | * snapshot id, highest id first. (Snapshots in the rbd_dev's list | |
2521 | * are also maintained in that order.) | |
dfc5606d | 2522 | */ |
304f6808 | 2523 | static int rbd_dev_snaps_update(struct rbd_device *rbd_dev) |
dfc5606d | 2524 | { |
35938150 AE |
2525 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
2526 | const u32 snap_count = snapc->num_snaps; | |
35938150 AE |
2527 | struct list_head *head = &rbd_dev->snaps; |
2528 | struct list_head *links = head->next; | |
2529 | u32 index = 0; | |
dfc5606d | 2530 | |
9fcbb800 | 2531 | dout("%s: snap count is %u\n", __func__, (unsigned int) snap_count); |
35938150 AE |
2532 | while (index < snap_count || links != head) { |
2533 | u64 snap_id; | |
2534 | struct rbd_snap *snap; | |
cd892126 AE |
2535 | char *snap_name; |
2536 | u64 snap_size = 0; | |
2537 | u64 snap_features = 0; | |
dfc5606d | 2538 | |
35938150 AE |
2539 | snap_id = index < snap_count ? snapc->snaps[index] |
2540 | : CEPH_NOSNAP; | |
2541 | snap = links != head ? list_entry(links, struct rbd_snap, node) | |
2542 | : NULL; | |
aafb230e | 2543 | rbd_assert(!snap || snap->id != CEPH_NOSNAP); |
dfc5606d | 2544 | |
35938150 AE |
2545 | if (snap_id == CEPH_NOSNAP || (snap && snap->id > snap_id)) { |
2546 | struct list_head *next = links->next; | |
dfc5606d | 2547 | |
35938150 | 2548 | /* Existing snapshot not in the new snap context */ |
dfc5606d | 2549 | |
971f839a | 2550 | if (rbd_dev->snap_id == snap->id) |
daba5fdb | 2551 | rbd_dev->exists = false; |
41f38c2b | 2552 | rbd_remove_snap_dev(snap); |
9fcbb800 | 2553 | dout("%ssnap id %llu has been removed\n", |
971f839a | 2554 | rbd_dev->snap_id == snap->id ? "mapped " : "", |
9fcbb800 | 2555 | (unsigned long long) snap->id); |
35938150 AE |
2556 | |
2557 | /* Done with this list entry; advance */ | |
2558 | ||
2559 | links = next; | |
dfc5606d YS |
2560 | continue; |
2561 | } | |
35938150 | 2562 | |
b8b1e2db AE |
2563 | snap_name = rbd_dev_snap_info(rbd_dev, index, |
2564 | &snap_size, &snap_features); | |
cd892126 AE |
2565 | if (IS_ERR(snap_name)) |
2566 | return PTR_ERR(snap_name); | |
2567 | ||
9fcbb800 AE |
2568 | dout("entry %u: snap_id = %llu\n", (unsigned int) snap_count, |
2569 | (unsigned long long) snap_id); | |
35938150 AE |
2570 | if (!snap || (snap_id != CEPH_NOSNAP && snap->id < snap_id)) { |
2571 | struct rbd_snap *new_snap; | |
2572 | ||
2573 | /* We haven't seen this snapshot before */ | |
2574 | ||
c8d18425 | 2575 | new_snap = __rbd_add_snap_dev(rbd_dev, snap_name, |
cd892126 | 2576 | snap_id, snap_size, snap_features); |
9fcbb800 AE |
2577 | if (IS_ERR(new_snap)) { |
2578 | int err = PTR_ERR(new_snap); | |
2579 | ||
2580 | dout(" failed to add dev, error %d\n", err); | |
2581 | ||
2582 | return err; | |
2583 | } | |
35938150 AE |
2584 | |
2585 | /* New goes before existing, or at end of list */ | |
2586 | ||
9fcbb800 | 2587 | dout(" added dev%s\n", snap ? "" : " at end\n"); |
35938150 AE |
2588 | if (snap) |
2589 | list_add_tail(&new_snap->node, &snap->node); | |
2590 | else | |
523f3258 | 2591 | list_add_tail(&new_snap->node, head); |
35938150 AE |
2592 | } else { |
2593 | /* Already have this one */ | |
2594 | ||
9fcbb800 AE |
2595 | dout(" already present\n"); |
2596 | ||
cd892126 | 2597 | rbd_assert(snap->size == snap_size); |
aafb230e | 2598 | rbd_assert(!strcmp(snap->name, snap_name)); |
cd892126 | 2599 | rbd_assert(snap->features == snap_features); |
35938150 AE |
2600 | |
2601 | /* Done with this list entry; advance */ | |
2602 | ||
2603 | links = links->next; | |
dfc5606d | 2604 | } |
35938150 AE |
2605 | |
2606 | /* Advance to the next entry in the snapshot context */ | |
2607 | ||
2608 | index++; | |
dfc5606d | 2609 | } |
9fcbb800 | 2610 | dout("%s: done\n", __func__); |
dfc5606d YS |
2611 | |
2612 | return 0; | |
2613 | } | |
2614 | ||
304f6808 AE |
2615 | /* |
2616 | * Scan the list of snapshots and register the devices for any that | |
2617 | * have not already been registered. | |
2618 | */ | |
2619 | static int rbd_dev_snaps_register(struct rbd_device *rbd_dev) | |
2620 | { | |
2621 | struct rbd_snap *snap; | |
2622 | int ret = 0; | |
2623 | ||
2624 | dout("%s called\n", __func__); | |
86ff77bb AE |
2625 | if (WARN_ON(!device_is_registered(&rbd_dev->dev))) |
2626 | return -EIO; | |
304f6808 AE |
2627 | |
2628 | list_for_each_entry(snap, &rbd_dev->snaps, node) { | |
2629 | if (!rbd_snap_registered(snap)) { | |
2630 | ret = rbd_register_snap_dev(snap, &rbd_dev->dev); | |
2631 | if (ret < 0) | |
2632 | break; | |
2633 | } | |
2634 | } | |
2635 | dout("%s: returning %d\n", __func__, ret); | |
2636 | ||
2637 | return ret; | |
2638 | } | |
2639 | ||
dfc5606d YS |
2640 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
2641 | { | |
dfc5606d | 2642 | struct device *dev; |
cd789ab9 | 2643 | int ret; |
dfc5606d YS |
2644 | |
2645 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
dfc5606d | 2646 | |
cd789ab9 | 2647 | dev = &rbd_dev->dev; |
dfc5606d YS |
2648 | dev->bus = &rbd_bus_type; |
2649 | dev->type = &rbd_device_type; | |
2650 | dev->parent = &rbd_root_dev; | |
2651 | dev->release = rbd_dev_release; | |
de71a297 | 2652 | dev_set_name(dev, "%d", rbd_dev->dev_id); |
dfc5606d | 2653 | ret = device_register(dev); |
dfc5606d | 2654 | |
dfc5606d | 2655 | mutex_unlock(&ctl_mutex); |
cd789ab9 | 2656 | |
dfc5606d | 2657 | return ret; |
602adf40 YS |
2658 | } |
2659 | ||
dfc5606d YS |
2660 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
2661 | { | |
2662 | device_unregister(&rbd_dev->dev); | |
2663 | } | |
2664 | ||
59c2be1e YS |
2665 | static int rbd_init_watch_dev(struct rbd_device *rbd_dev) |
2666 | { | |
2667 | int ret, rc; | |
2668 | ||
2669 | do { | |
0e6f322d | 2670 | ret = rbd_req_sync_watch(rbd_dev); |
59c2be1e | 2671 | if (ret == -ERANGE) { |
117973fb | 2672 | rc = rbd_dev_refresh(rbd_dev, NULL); |
59c2be1e YS |
2673 | if (rc < 0) |
2674 | return rc; | |
2675 | } | |
2676 | } while (ret == -ERANGE); | |
2677 | ||
2678 | return ret; | |
2679 | } | |
2680 | ||
e2839308 | 2681 | static atomic64_t rbd_dev_id_max = ATOMIC64_INIT(0); |
1ddbe94e AE |
2682 | |
2683 | /* | |
499afd5b AE |
2684 | * Get a unique rbd identifier for the given new rbd_dev, and add |
2685 | * the rbd_dev to the global list. The minimum rbd id is 1. | |
1ddbe94e | 2686 | */ |
e2839308 | 2687 | static void rbd_dev_id_get(struct rbd_device *rbd_dev) |
b7f23c36 | 2688 | { |
e2839308 | 2689 | rbd_dev->dev_id = atomic64_inc_return(&rbd_dev_id_max); |
499afd5b AE |
2690 | |
2691 | spin_lock(&rbd_dev_list_lock); | |
2692 | list_add_tail(&rbd_dev->node, &rbd_dev_list); | |
2693 | spin_unlock(&rbd_dev_list_lock); | |
e2839308 AE |
2694 | dout("rbd_dev %p given dev id %llu\n", rbd_dev, |
2695 | (unsigned long long) rbd_dev->dev_id); | |
1ddbe94e | 2696 | } |
b7f23c36 | 2697 | |
1ddbe94e | 2698 | /* |
499afd5b AE |
2699 | * Remove an rbd_dev from the global list, and record that its |
2700 | * identifier is no longer in use. | |
1ddbe94e | 2701 | */ |
e2839308 | 2702 | static void rbd_dev_id_put(struct rbd_device *rbd_dev) |
1ddbe94e | 2703 | { |
d184f6bf | 2704 | struct list_head *tmp; |
de71a297 | 2705 | int rbd_id = rbd_dev->dev_id; |
d184f6bf AE |
2706 | int max_id; |
2707 | ||
aafb230e | 2708 | rbd_assert(rbd_id > 0); |
499afd5b | 2709 | |
e2839308 AE |
2710 | dout("rbd_dev %p released dev id %llu\n", rbd_dev, |
2711 | (unsigned long long) rbd_dev->dev_id); | |
499afd5b AE |
2712 | spin_lock(&rbd_dev_list_lock); |
2713 | list_del_init(&rbd_dev->node); | |
d184f6bf AE |
2714 | |
2715 | /* | |
2716 | * If the id being "put" is not the current maximum, there | |
2717 | * is nothing special we need to do. | |
2718 | */ | |
e2839308 | 2719 | if (rbd_id != atomic64_read(&rbd_dev_id_max)) { |
d184f6bf AE |
2720 | spin_unlock(&rbd_dev_list_lock); |
2721 | return; | |
2722 | } | |
2723 | ||
2724 | /* | |
2725 | * We need to update the current maximum id. Search the | |
2726 | * list to find out what it is. We're more likely to find | |
2727 | * the maximum at the end, so search the list backward. | |
2728 | */ | |
2729 | max_id = 0; | |
2730 | list_for_each_prev(tmp, &rbd_dev_list) { | |
2731 | struct rbd_device *rbd_dev; | |
2732 | ||
2733 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
b213e0b1 AE |
2734 | if (rbd_dev->dev_id > max_id) |
2735 | max_id = rbd_dev->dev_id; | |
d184f6bf | 2736 | } |
499afd5b | 2737 | spin_unlock(&rbd_dev_list_lock); |
b7f23c36 | 2738 | |
1ddbe94e | 2739 | /* |
e2839308 | 2740 | * The max id could have been updated by rbd_dev_id_get(), in |
d184f6bf AE |
2741 | * which case it now accurately reflects the new maximum. |
2742 | * Be careful not to overwrite the maximum value in that | |
2743 | * case. | |
1ddbe94e | 2744 | */ |
e2839308 AE |
2745 | atomic64_cmpxchg(&rbd_dev_id_max, rbd_id, max_id); |
2746 | dout(" max dev id has been reset\n"); | |
b7f23c36 AE |
2747 | } |
2748 | ||
e28fff26 AE |
2749 | /* |
2750 | * Skips over white space at *buf, and updates *buf to point to the | |
2751 | * first found non-space character (if any). Returns the length of | |
593a9e7b AE |
2752 | * the token (string of non-white space characters) found. Note |
2753 | * that *buf must be terminated with '\0'. | |
e28fff26 AE |
2754 | */ |
2755 | static inline size_t next_token(const char **buf) | |
2756 | { | |
2757 | /* | |
2758 | * These are the characters that produce nonzero for | |
2759 | * isspace() in the "C" and "POSIX" locales. | |
2760 | */ | |
2761 | const char *spaces = " \f\n\r\t\v"; | |
2762 | ||
2763 | *buf += strspn(*buf, spaces); /* Find start of token */ | |
2764 | ||
2765 | return strcspn(*buf, spaces); /* Return token length */ | |
2766 | } | |
2767 | ||
2768 | /* | |
2769 | * Finds the next token in *buf, and if the provided token buffer is | |
2770 | * big enough, copies the found token into it. The result, if | |
593a9e7b AE |
2771 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
2772 | * must be terminated with '\0' on entry. | |
e28fff26 AE |
2773 | * |
2774 | * Returns the length of the token found (not including the '\0'). | |
2775 | * Return value will be 0 if no token is found, and it will be >= | |
2776 | * token_size if the token would not fit. | |
2777 | * | |
593a9e7b | 2778 | * The *buf pointer will be updated to point beyond the end of the |
e28fff26 AE |
2779 | * found token. Note that this occurs even if the token buffer is |
2780 | * too small to hold it. | |
2781 | */ | |
2782 | static inline size_t copy_token(const char **buf, | |
2783 | char *token, | |
2784 | size_t token_size) | |
2785 | { | |
2786 | size_t len; | |
2787 | ||
2788 | len = next_token(buf); | |
2789 | if (len < token_size) { | |
2790 | memcpy(token, *buf, len); | |
2791 | *(token + len) = '\0'; | |
2792 | } | |
2793 | *buf += len; | |
2794 | ||
2795 | return len; | |
2796 | } | |
2797 | ||
ea3352f4 AE |
2798 | /* |
2799 | * Finds the next token in *buf, dynamically allocates a buffer big | |
2800 | * enough to hold a copy of it, and copies the token into the new | |
2801 | * buffer. The copy is guaranteed to be terminated with '\0'. Note | |
2802 | * that a duplicate buffer is created even for a zero-length token. | |
2803 | * | |
2804 | * Returns a pointer to the newly-allocated duplicate, or a null | |
2805 | * pointer if memory for the duplicate was not available. If | |
2806 | * the lenp argument is a non-null pointer, the length of the token | |
2807 | * (not including the '\0') is returned in *lenp. | |
2808 | * | |
2809 | * If successful, the *buf pointer will be updated to point beyond | |
2810 | * the end of the found token. | |
2811 | * | |
2812 | * Note: uses GFP_KERNEL for allocation. | |
2813 | */ | |
2814 | static inline char *dup_token(const char **buf, size_t *lenp) | |
2815 | { | |
2816 | char *dup; | |
2817 | size_t len; | |
2818 | ||
2819 | len = next_token(buf); | |
2820 | dup = kmalloc(len + 1, GFP_KERNEL); | |
2821 | if (!dup) | |
2822 | return NULL; | |
2823 | ||
2824 | memcpy(dup, *buf, len); | |
2825 | *(dup + len) = '\0'; | |
2826 | *buf += len; | |
2827 | ||
2828 | if (lenp) | |
2829 | *lenp = len; | |
2830 | ||
2831 | return dup; | |
2832 | } | |
2833 | ||
a725f65e | 2834 | /* |
3feeb894 AE |
2835 | * This fills in the pool_name, image_name, image_name_len, rbd_dev, |
2836 | * rbd_md_name, and name fields of the given rbd_dev, based on the | |
2837 | * list of monitor addresses and other options provided via | |
2838 | * /sys/bus/rbd/add. Returns a pointer to a dynamically-allocated | |
2839 | * copy of the snapshot name to map if successful, or a | |
2840 | * pointer-coded error otherwise. | |
d22f76e7 AE |
2841 | * |
2842 | * Note: rbd_dev is assumed to have been initially zero-filled. | |
a725f65e | 2843 | */ |
0ddebc0c | 2844 | static struct ceph_options *rbd_add_parse_args(struct rbd_device *rbd_dev, |
819d52bf | 2845 | const char *buf) |
e28fff26 | 2846 | { |
d22f76e7 | 2847 | size_t len; |
0ddebc0c AE |
2848 | const char *mon_addrs; |
2849 | size_t mon_addrs_size; | |
f28e565a AE |
2850 | char *options; |
2851 | struct ceph_options *err_ptr = ERR_PTR(-EINVAL); | |
0ddebc0c AE |
2852 | struct rbd_options rbd_opts; |
2853 | struct ceph_options *ceph_opts; | |
e28fff26 AE |
2854 | |
2855 | /* The first four tokens are required */ | |
2856 | ||
7ef3214a AE |
2857 | len = next_token(&buf); |
2858 | if (!len) | |
f28e565a | 2859 | return err_ptr; /* Missing monitor address(es) */ |
0ddebc0c | 2860 | mon_addrs = buf; |
f28e565a | 2861 | mon_addrs_size = len + 1; |
7ef3214a | 2862 | buf += len; |
a725f65e | 2863 | |
f28e565a AE |
2864 | options = dup_token(&buf, NULL); |
2865 | if (!options) | |
2866 | goto out_mem; | |
2867 | if (!*options) | |
2868 | goto out_err; /* Missing options */ | |
e28fff26 | 2869 | |
d22f76e7 AE |
2870 | rbd_dev->pool_name = dup_token(&buf, NULL); |
2871 | if (!rbd_dev->pool_name) | |
f28e565a AE |
2872 | goto out_mem; |
2873 | if (!*rbd_dev->pool_name) | |
2874 | goto out_err; /* Missing pool name */ | |
e28fff26 | 2875 | |
0bed54dc AE |
2876 | rbd_dev->image_name = dup_token(&buf, &rbd_dev->image_name_len); |
2877 | if (!rbd_dev->image_name) | |
f28e565a AE |
2878 | goto out_mem; |
2879 | if (!*rbd_dev->image_name) | |
2880 | goto out_err; /* Missing image name */ | |
d4b125e9 | 2881 | |
f28e565a AE |
2882 | /* |
2883 | * Snapshot name is optional; default is to use "-" | |
2884 | * (indicating the head/no snapshot). | |
2885 | */ | |
3feeb894 | 2886 | len = next_token(&buf); |
820a5f3e | 2887 | if (!len) { |
3feeb894 AE |
2888 | buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ |
2889 | len = sizeof (RBD_SNAP_HEAD_NAME) - 1; | |
f28e565a AE |
2890 | } else if (len > RBD_MAX_SNAP_NAME_LEN) { |
2891 | err_ptr = ERR_PTR(-ENAMETOOLONG); | |
2892 | goto out_err; | |
849b4260 | 2893 | } |
819d52bf AE |
2894 | rbd_dev->snap_name = kmalloc(len + 1, GFP_KERNEL); |
2895 | if (!rbd_dev->snap_name) | |
f28e565a | 2896 | goto out_mem; |
819d52bf AE |
2897 | memcpy(rbd_dev->snap_name, buf, len); |
2898 | *(rbd_dev->snap_name + len) = '\0'; | |
e5c35534 | 2899 | |
0ddebc0c | 2900 | /* Initialize all rbd options to the defaults */ |
e28fff26 | 2901 | |
0ddebc0c | 2902 | rbd_opts.read_only = RBD_READ_ONLY_DEFAULT; |
d22f76e7 | 2903 | |
0ddebc0c AE |
2904 | ceph_opts = ceph_parse_options(options, mon_addrs, |
2905 | mon_addrs + mon_addrs_size - 1, | |
2906 | parse_rbd_opts_token, &rbd_opts); | |
f28e565a | 2907 | kfree(options); |
0ddebc0c AE |
2908 | |
2909 | /* Record the parsed rbd options */ | |
2910 | ||
f28e565a | 2911 | if (!IS_ERR(ceph_opts)) |
0ddebc0c | 2912 | rbd_dev->mapping.read_only = rbd_opts.read_only; |
0ddebc0c AE |
2913 | |
2914 | return ceph_opts; | |
f28e565a AE |
2915 | out_mem: |
2916 | err_ptr = ERR_PTR(-ENOMEM); | |
d22f76e7 | 2917 | out_err: |
0bed54dc | 2918 | kfree(rbd_dev->image_name); |
d78fd7ae AE |
2919 | rbd_dev->image_name = NULL; |
2920 | rbd_dev->image_name_len = 0; | |
d22f76e7 AE |
2921 | kfree(rbd_dev->pool_name); |
2922 | rbd_dev->pool_name = NULL; | |
f28e565a | 2923 | kfree(options); |
d22f76e7 | 2924 | |
3feeb894 | 2925 | return err_ptr; |
a725f65e AE |
2926 | } |
2927 | ||
589d30e0 AE |
2928 | /* |
2929 | * An rbd format 2 image has a unique identifier, distinct from the | |
2930 | * name given to it by the user. Internally, that identifier is | |
2931 | * what's used to specify the names of objects related to the image. | |
2932 | * | |
2933 | * A special "rbd id" object is used to map an rbd image name to its | |
2934 | * id. If that object doesn't exist, then there is no v2 rbd image | |
2935 | * with the supplied name. | |
2936 | * | |
2937 | * This function will record the given rbd_dev's image_id field if | |
2938 | * it can be determined, and in that case will return 0. If any | |
2939 | * errors occur a negative errno will be returned and the rbd_dev's | |
2940 | * image_id field will be unchanged (and should be NULL). | |
2941 | */ | |
2942 | static int rbd_dev_image_id(struct rbd_device *rbd_dev) | |
2943 | { | |
2944 | int ret; | |
2945 | size_t size; | |
2946 | char *object_name; | |
2947 | void *response; | |
2948 | void *p; | |
2949 | ||
2950 | /* | |
2951 | * First, see if the format 2 image id file exists, and if | |
2952 | * so, get the image's persistent id from it. | |
2953 | */ | |
2954 | size = sizeof (RBD_ID_PREFIX) + rbd_dev->image_name_len; | |
2955 | object_name = kmalloc(size, GFP_NOIO); | |
2956 | if (!object_name) | |
2957 | return -ENOMEM; | |
2958 | sprintf(object_name, "%s%s", RBD_ID_PREFIX, rbd_dev->image_name); | |
2959 | dout("rbd id object name is %s\n", object_name); | |
2960 | ||
2961 | /* Response will be an encoded string, which includes a length */ | |
2962 | ||
2963 | size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; | |
2964 | response = kzalloc(size, GFP_NOIO); | |
2965 | if (!response) { | |
2966 | ret = -ENOMEM; | |
2967 | goto out; | |
2968 | } | |
2969 | ||
2970 | ret = rbd_req_sync_exec(rbd_dev, object_name, | |
2971 | "rbd", "get_id", | |
2972 | NULL, 0, | |
2973 | response, RBD_IMAGE_ID_LEN_MAX, | |
2974 | CEPH_OSD_FLAG_READ, NULL); | |
2975 | dout("%s: rbd_req_sync_exec returned %d\n", __func__, ret); | |
2976 | if (ret < 0) | |
2977 | goto out; | |
a0ea3a40 | 2978 | ret = 0; /* rbd_req_sync_exec() can return positive */ |
589d30e0 AE |
2979 | |
2980 | p = response; | |
2981 | rbd_dev->image_id = ceph_extract_encoded_string(&p, | |
2982 | p + RBD_IMAGE_ID_LEN_MAX, | |
2983 | &rbd_dev->image_id_len, | |
2984 | GFP_NOIO); | |
2985 | if (IS_ERR(rbd_dev->image_id)) { | |
2986 | ret = PTR_ERR(rbd_dev->image_id); | |
2987 | rbd_dev->image_id = NULL; | |
2988 | } else { | |
2989 | dout("image_id is %s\n", rbd_dev->image_id); | |
2990 | } | |
2991 | out: | |
2992 | kfree(response); | |
2993 | kfree(object_name); | |
2994 | ||
2995 | return ret; | |
2996 | } | |
2997 | ||
a30b71b9 AE |
2998 | static int rbd_dev_v1_probe(struct rbd_device *rbd_dev) |
2999 | { | |
3000 | int ret; | |
3001 | size_t size; | |
3002 | ||
3003 | /* Version 1 images have no id; empty string is used */ | |
3004 | ||
3005 | rbd_dev->image_id = kstrdup("", GFP_KERNEL); | |
3006 | if (!rbd_dev->image_id) | |
3007 | return -ENOMEM; | |
3008 | rbd_dev->image_id_len = 0; | |
3009 | ||
3010 | /* Record the header object name for this rbd image. */ | |
3011 | ||
3012 | size = rbd_dev->image_name_len + sizeof (RBD_SUFFIX); | |
3013 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); | |
3014 | if (!rbd_dev->header_name) { | |
3015 | ret = -ENOMEM; | |
3016 | goto out_err; | |
3017 | } | |
3018 | sprintf(rbd_dev->header_name, "%s%s", rbd_dev->image_name, RBD_SUFFIX); | |
3019 | ||
3020 | /* Populate rbd image metadata */ | |
3021 | ||
3022 | ret = rbd_read_header(rbd_dev, &rbd_dev->header); | |
3023 | if (ret < 0) | |
3024 | goto out_err; | |
3025 | rbd_dev->image_format = 1; | |
3026 | ||
3027 | dout("discovered version 1 image, header name is %s\n", | |
3028 | rbd_dev->header_name); | |
3029 | ||
3030 | return 0; | |
3031 | ||
3032 | out_err: | |
3033 | kfree(rbd_dev->header_name); | |
3034 | rbd_dev->header_name = NULL; | |
3035 | kfree(rbd_dev->image_id); | |
3036 | rbd_dev->image_id = NULL; | |
3037 | ||
3038 | return ret; | |
3039 | } | |
3040 | ||
3041 | static int rbd_dev_v2_probe(struct rbd_device *rbd_dev) | |
3042 | { | |
3043 | size_t size; | |
9d475de5 | 3044 | int ret; |
6e14b1a6 | 3045 | u64 ver = 0; |
a30b71b9 AE |
3046 | |
3047 | /* | |
3048 | * Image id was filled in by the caller. Record the header | |
3049 | * object name for this rbd image. | |
3050 | */ | |
3051 | size = sizeof (RBD_HEADER_PREFIX) + rbd_dev->image_id_len; | |
3052 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); | |
3053 | if (!rbd_dev->header_name) | |
3054 | return -ENOMEM; | |
3055 | sprintf(rbd_dev->header_name, "%s%s", | |
3056 | RBD_HEADER_PREFIX, rbd_dev->image_id); | |
9d475de5 AE |
3057 | |
3058 | /* Get the size and object order for the image */ | |
3059 | ||
3060 | ret = rbd_dev_v2_image_size(rbd_dev); | |
1e130199 AE |
3061 | if (ret < 0) |
3062 | goto out_err; | |
3063 | ||
3064 | /* Get the object prefix (a.k.a. block_name) for the image */ | |
3065 | ||
3066 | ret = rbd_dev_v2_object_prefix(rbd_dev); | |
b1b5402a AE |
3067 | if (ret < 0) |
3068 | goto out_err; | |
3069 | ||
d889140c | 3070 | /* Get the and check features for the image */ |
b1b5402a AE |
3071 | |
3072 | ret = rbd_dev_v2_features(rbd_dev); | |
9d475de5 AE |
3073 | if (ret < 0) |
3074 | goto out_err; | |
35d489f9 | 3075 | |
6e14b1a6 AE |
3076 | /* crypto and compression type aren't (yet) supported for v2 images */ |
3077 | ||
3078 | rbd_dev->header.crypt_type = 0; | |
3079 | rbd_dev->header.comp_type = 0; | |
35d489f9 | 3080 | |
6e14b1a6 AE |
3081 | /* Get the snapshot context, plus the header version */ |
3082 | ||
3083 | ret = rbd_dev_v2_snap_context(rbd_dev, &ver); | |
35d489f9 AE |
3084 | if (ret) |
3085 | goto out_err; | |
6e14b1a6 AE |
3086 | rbd_dev->header.obj_version = ver; |
3087 | ||
a30b71b9 AE |
3088 | rbd_dev->image_format = 2; |
3089 | ||
3090 | dout("discovered version 2 image, header name is %s\n", | |
3091 | rbd_dev->header_name); | |
3092 | ||
35152979 | 3093 | return 0; |
9d475de5 AE |
3094 | out_err: |
3095 | kfree(rbd_dev->header_name); | |
3096 | rbd_dev->header_name = NULL; | |
1e130199 AE |
3097 | kfree(rbd_dev->header.object_prefix); |
3098 | rbd_dev->header.object_prefix = NULL; | |
9d475de5 AE |
3099 | |
3100 | return ret; | |
a30b71b9 AE |
3101 | } |
3102 | ||
3103 | /* | |
3104 | * Probe for the existence of the header object for the given rbd | |
3105 | * device. For format 2 images this includes determining the image | |
3106 | * id. | |
3107 | */ | |
3108 | static int rbd_dev_probe(struct rbd_device *rbd_dev) | |
3109 | { | |
3110 | int ret; | |
3111 | ||
3112 | /* | |
3113 | * Get the id from the image id object. If it's not a | |
3114 | * format 2 image, we'll get ENOENT back, and we'll assume | |
3115 | * it's a format 1 image. | |
3116 | */ | |
3117 | ret = rbd_dev_image_id(rbd_dev); | |
3118 | if (ret) | |
3119 | ret = rbd_dev_v1_probe(rbd_dev); | |
3120 | else | |
3121 | ret = rbd_dev_v2_probe(rbd_dev); | |
3122 | if (ret) | |
3123 | dout("probe failed, returning %d\n", ret); | |
3124 | ||
3125 | return ret; | |
3126 | } | |
3127 | ||
59c2be1e YS |
3128 | static ssize_t rbd_add(struct bus_type *bus, |
3129 | const char *buf, | |
3130 | size_t count) | |
602adf40 | 3131 | { |
cb8627c7 | 3132 | struct rbd_device *rbd_dev = NULL; |
78cea76e | 3133 | struct ceph_options *ceph_opts; |
27cc2594 AE |
3134 | struct ceph_osd_client *osdc; |
3135 | int rc = -ENOMEM; | |
602adf40 YS |
3136 | |
3137 | if (!try_module_get(THIS_MODULE)) | |
3138 | return -ENODEV; | |
3139 | ||
cb8627c7 AE |
3140 | rbd_dev = kzalloc(sizeof(*rbd_dev), GFP_KERNEL); |
3141 | if (!rbd_dev) | |
85ae8926 | 3142 | goto err_out_mem; |
602adf40 YS |
3143 | |
3144 | /* static rbd_device initialization */ | |
3145 | spin_lock_init(&rbd_dev->lock); | |
3146 | INIT_LIST_HEAD(&rbd_dev->node); | |
dfc5606d | 3147 | INIT_LIST_HEAD(&rbd_dev->snaps); |
c666601a | 3148 | init_rwsem(&rbd_dev->header_rwsem); |
602adf40 | 3149 | |
602adf40 | 3150 | /* parse add command */ |
819d52bf | 3151 | ceph_opts = rbd_add_parse_args(rbd_dev, buf); |
78cea76e AE |
3152 | if (IS_ERR(ceph_opts)) { |
3153 | rc = PTR_ERR(ceph_opts); | |
0ddebc0c | 3154 | goto err_out_mem; |
78cea76e AE |
3155 | } |
3156 | ||
78cea76e AE |
3157 | rc = rbd_get_client(rbd_dev, ceph_opts); |
3158 | if (rc < 0) | |
0ddebc0c | 3159 | goto err_out_args; |
78cea76e | 3160 | ceph_opts = NULL; /* ceph_opts now owned by rbd_dev client */ |
602adf40 | 3161 | |
602adf40 | 3162 | /* pick the pool */ |
1dbb4399 | 3163 | osdc = &rbd_dev->rbd_client->client->osdc; |
602adf40 YS |
3164 | rc = ceph_pg_poolid_by_name(osdc->osdmap, rbd_dev->pool_name); |
3165 | if (rc < 0) | |
3166 | goto err_out_client; | |
86992098 | 3167 | rbd_dev->pool_id = (u64) rc; |
602adf40 | 3168 | |
a30b71b9 AE |
3169 | rc = rbd_dev_probe(rbd_dev); |
3170 | if (rc < 0) | |
05fd6f6f AE |
3171 | goto err_out_client; |
3172 | ||
3173 | /* no need to lock here, as rbd_dev is not registered yet */ | |
3174 | rc = rbd_dev_snaps_update(rbd_dev); | |
3175 | if (rc) | |
41f38c2b | 3176 | goto err_out_probe; |
05fd6f6f | 3177 | |
819d52bf | 3178 | rc = rbd_dev_set_mapping(rbd_dev); |
05fd6f6f | 3179 | if (rc) |
41f38c2b | 3180 | goto err_out_snaps; |
05fd6f6f | 3181 | |
85ae8926 AE |
3182 | /* generate unique id: find highest unique id, add one */ |
3183 | rbd_dev_id_get(rbd_dev); | |
3184 | ||
3185 | /* Fill in the device name, now that we have its id. */ | |
3186 | BUILD_BUG_ON(DEV_NAME_LEN | |
3187 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); | |
3188 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); | |
3189 | ||
3190 | /* Get our block major device number. */ | |
3191 | ||
27cc2594 AE |
3192 | rc = register_blkdev(0, rbd_dev->name); |
3193 | if (rc < 0) | |
85ae8926 | 3194 | goto err_out_id; |
27cc2594 | 3195 | rbd_dev->major = rc; |
602adf40 | 3196 | |
0f308a31 AE |
3197 | /* Set up the blkdev mapping. */ |
3198 | ||
3199 | rc = rbd_init_disk(rbd_dev); | |
dfc5606d | 3200 | if (rc) |
766fc439 YS |
3201 | goto err_out_blkdev; |
3202 | ||
0f308a31 AE |
3203 | rc = rbd_bus_add_dev(rbd_dev); |
3204 | if (rc) | |
3205 | goto err_out_disk; | |
3206 | ||
32eec68d AE |
3207 | /* |
3208 | * At this point cleanup in the event of an error is the job | |
3209 | * of the sysfs code (initiated by rbd_bus_del_dev()). | |
32eec68d | 3210 | */ |
2ac4e75d | 3211 | |
4bb1f1ed | 3212 | down_write(&rbd_dev->header_rwsem); |
5ed16177 | 3213 | rc = rbd_dev_snaps_register(rbd_dev); |
4bb1f1ed | 3214 | up_write(&rbd_dev->header_rwsem); |
2ac4e75d AE |
3215 | if (rc) |
3216 | goto err_out_bus; | |
3217 | ||
3ee4001e AE |
3218 | rc = rbd_init_watch_dev(rbd_dev); |
3219 | if (rc) | |
3220 | goto err_out_bus; | |
3221 | ||
2ac4e75d AE |
3222 | /* Everything's ready. Announce the disk to the world. */ |
3223 | ||
2ac4e75d | 3224 | add_disk(rbd_dev->disk); |
3ee4001e | 3225 | |
2ac4e75d AE |
3226 | pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name, |
3227 | (unsigned long long) rbd_dev->mapping.size); | |
3228 | ||
602adf40 YS |
3229 | return count; |
3230 | ||
766fc439 | 3231 | err_out_bus: |
766fc439 YS |
3232 | /* this will also clean up rest of rbd_dev stuff */ |
3233 | ||
3234 | rbd_bus_del_dev(rbd_dev); | |
766fc439 YS |
3235 | return rc; |
3236 | ||
0f308a31 AE |
3237 | err_out_disk: |
3238 | rbd_free_disk(rbd_dev); | |
602adf40 YS |
3239 | err_out_blkdev: |
3240 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
85ae8926 AE |
3241 | err_out_id: |
3242 | rbd_dev_id_put(rbd_dev); | |
41f38c2b AE |
3243 | err_out_snaps: |
3244 | rbd_remove_all_snaps(rbd_dev); | |
3245 | err_out_probe: | |
05fd6f6f | 3246 | rbd_header_free(&rbd_dev->header); |
602adf40 | 3247 | err_out_client: |
3fcf2581 | 3248 | kfree(rbd_dev->header_name); |
602adf40 | 3249 | rbd_put_client(rbd_dev); |
589d30e0 | 3250 | kfree(rbd_dev->image_id); |
0ddebc0c | 3251 | err_out_args: |
78cea76e AE |
3252 | if (ceph_opts) |
3253 | ceph_destroy_options(ceph_opts); | |
971f839a | 3254 | kfree(rbd_dev->snap_name); |
85ae8926 AE |
3255 | kfree(rbd_dev->image_name); |
3256 | kfree(rbd_dev->pool_name); | |
3257 | err_out_mem: | |
27cc2594 AE |
3258 | kfree(rbd_dev); |
3259 | ||
602adf40 YS |
3260 | dout("Error adding device %s\n", buf); |
3261 | module_put(THIS_MODULE); | |
27cc2594 AE |
3262 | |
3263 | return (ssize_t) rc; | |
602adf40 YS |
3264 | } |
3265 | ||
de71a297 | 3266 | static struct rbd_device *__rbd_get_dev(unsigned long dev_id) |
602adf40 YS |
3267 | { |
3268 | struct list_head *tmp; | |
3269 | struct rbd_device *rbd_dev; | |
3270 | ||
e124a82f | 3271 | spin_lock(&rbd_dev_list_lock); |
602adf40 YS |
3272 | list_for_each(tmp, &rbd_dev_list) { |
3273 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
de71a297 | 3274 | if (rbd_dev->dev_id == dev_id) { |
e124a82f | 3275 | spin_unlock(&rbd_dev_list_lock); |
602adf40 | 3276 | return rbd_dev; |
e124a82f | 3277 | } |
602adf40 | 3278 | } |
e124a82f | 3279 | spin_unlock(&rbd_dev_list_lock); |
602adf40 YS |
3280 | return NULL; |
3281 | } | |
3282 | ||
dfc5606d | 3283 | static void rbd_dev_release(struct device *dev) |
602adf40 | 3284 | { |
593a9e7b | 3285 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 3286 | |
1dbb4399 AE |
3287 | if (rbd_dev->watch_request) { |
3288 | struct ceph_client *client = rbd_dev->rbd_client->client; | |
3289 | ||
3290 | ceph_osdc_unregister_linger_request(&client->osdc, | |
59c2be1e | 3291 | rbd_dev->watch_request); |
1dbb4399 | 3292 | } |
59c2be1e | 3293 | if (rbd_dev->watch_event) |
070c633f | 3294 | rbd_req_sync_unwatch(rbd_dev); |
59c2be1e | 3295 | |
602adf40 YS |
3296 | rbd_put_client(rbd_dev); |
3297 | ||
3298 | /* clean up and free blkdev */ | |
3299 | rbd_free_disk(rbd_dev); | |
3300 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
32eec68d | 3301 | |
2ac4e75d AE |
3302 | /* release allocated disk header fields */ |
3303 | rbd_header_free(&rbd_dev->header); | |
3304 | ||
32eec68d | 3305 | /* done with the id, and with the rbd_dev */ |
971f839a | 3306 | kfree(rbd_dev->snap_name); |
589d30e0 | 3307 | kfree(rbd_dev->image_id); |
0bed54dc | 3308 | kfree(rbd_dev->header_name); |
d22f76e7 | 3309 | kfree(rbd_dev->pool_name); |
0bed54dc | 3310 | kfree(rbd_dev->image_name); |
e2839308 | 3311 | rbd_dev_id_put(rbd_dev); |
602adf40 YS |
3312 | kfree(rbd_dev); |
3313 | ||
3314 | /* release module ref */ | |
3315 | module_put(THIS_MODULE); | |
602adf40 YS |
3316 | } |
3317 | ||
dfc5606d YS |
3318 | static ssize_t rbd_remove(struct bus_type *bus, |
3319 | const char *buf, | |
3320 | size_t count) | |
602adf40 YS |
3321 | { |
3322 | struct rbd_device *rbd_dev = NULL; | |
3323 | int target_id, rc; | |
3324 | unsigned long ul; | |
3325 | int ret = count; | |
3326 | ||
3327 | rc = strict_strtoul(buf, 10, &ul); | |
3328 | if (rc) | |
3329 | return rc; | |
3330 | ||
3331 | /* convert to int; abort if we lost anything in the conversion */ | |
3332 | target_id = (int) ul; | |
3333 | if (target_id != ul) | |
3334 | return -EINVAL; | |
3335 | ||
3336 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
3337 | ||
3338 | rbd_dev = __rbd_get_dev(target_id); | |
3339 | if (!rbd_dev) { | |
3340 | ret = -ENOENT; | |
3341 | goto done; | |
3342 | } | |
3343 | ||
41f38c2b | 3344 | rbd_remove_all_snaps(rbd_dev); |
dfc5606d | 3345 | rbd_bus_del_dev(rbd_dev); |
602adf40 YS |
3346 | |
3347 | done: | |
3348 | mutex_unlock(&ctl_mutex); | |
aafb230e | 3349 | |
602adf40 YS |
3350 | return ret; |
3351 | } | |
3352 | ||
602adf40 YS |
3353 | /* |
3354 | * create control files in sysfs | |
dfc5606d | 3355 | * /sys/bus/rbd/... |
602adf40 YS |
3356 | */ |
3357 | static int rbd_sysfs_init(void) | |
3358 | { | |
dfc5606d | 3359 | int ret; |
602adf40 | 3360 | |
fed4c143 | 3361 | ret = device_register(&rbd_root_dev); |
21079786 | 3362 | if (ret < 0) |
dfc5606d | 3363 | return ret; |
602adf40 | 3364 | |
fed4c143 AE |
3365 | ret = bus_register(&rbd_bus_type); |
3366 | if (ret < 0) | |
3367 | device_unregister(&rbd_root_dev); | |
602adf40 | 3368 | |
602adf40 YS |
3369 | return ret; |
3370 | } | |
3371 | ||
3372 | static void rbd_sysfs_cleanup(void) | |
3373 | { | |
dfc5606d | 3374 | bus_unregister(&rbd_bus_type); |
fed4c143 | 3375 | device_unregister(&rbd_root_dev); |
602adf40 YS |
3376 | } |
3377 | ||
3378 | int __init rbd_init(void) | |
3379 | { | |
3380 | int rc; | |
3381 | ||
3382 | rc = rbd_sysfs_init(); | |
3383 | if (rc) | |
3384 | return rc; | |
f0f8cef5 | 3385 | pr_info("loaded " RBD_DRV_NAME_LONG "\n"); |
602adf40 YS |
3386 | return 0; |
3387 | } | |
3388 | ||
3389 | void __exit rbd_exit(void) | |
3390 | { | |
3391 | rbd_sysfs_cleanup(); | |
3392 | } | |
3393 | ||
3394 | module_init(rbd_init); | |
3395 | module_exit(rbd_exit); | |
3396 | ||
3397 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
3398 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
3399 | MODULE_DESCRIPTION("rados block device"); | |
3400 | ||
3401 | /* following authorship retained from original osdblk.c */ | |
3402 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); | |
3403 | ||
3404 | MODULE_LICENSE("GPL"); |