]>
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 | ||
593a9e7b AE |
44 | /* |
45 | * The basic unit of block I/O is a sector. It is interpreted in a | |
46 | * number of contexts in Linux (blk, bio, genhd), but the default is | |
47 | * universally 512 bytes. These symbols are just slightly more | |
48 | * meaningful than the bare numbers they represent. | |
49 | */ | |
50 | #define SECTOR_SHIFT 9 | |
51 | #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) | |
52 | ||
df111be6 AE |
53 | /* It might be useful to have this defined elsewhere too */ |
54 | ||
55 | #define U64_MAX ((u64) (~0ULL)) | |
56 | ||
f0f8cef5 AE |
57 | #define RBD_DRV_NAME "rbd" |
58 | #define RBD_DRV_NAME_LONG "rbd (rados block device)" | |
602adf40 YS |
59 | |
60 | #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ | |
61 | ||
602adf40 YS |
62 | #define RBD_MAX_SNAP_NAME_LEN 32 |
63 | #define RBD_MAX_OPT_LEN 1024 | |
64 | ||
65 | #define RBD_SNAP_HEAD_NAME "-" | |
66 | ||
81a89793 AE |
67 | /* |
68 | * An RBD device name will be "rbd#", where the "rbd" comes from | |
69 | * RBD_DRV_NAME above, and # is a unique integer identifier. | |
70 | * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big | |
71 | * enough to hold all possible device names. | |
72 | */ | |
602adf40 | 73 | #define DEV_NAME_LEN 32 |
81a89793 | 74 | #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) |
602adf40 | 75 | |
cc0538b6 | 76 | #define RBD_READ_ONLY_DEFAULT false |
59c2be1e | 77 | |
602adf40 YS |
78 | /* |
79 | * block device image metadata (in-memory version) | |
80 | */ | |
81 | struct rbd_image_header { | |
82 | u64 image_size; | |
849b4260 | 83 | char *object_prefix; |
602adf40 YS |
84 | __u8 obj_order; |
85 | __u8 crypt_type; | |
86 | __u8 comp_type; | |
602adf40 | 87 | struct ceph_snap_context *snapc; |
602adf40 YS |
88 | u32 total_snaps; |
89 | ||
90 | char *snap_names; | |
91 | u64 *snap_sizes; | |
59c2be1e YS |
92 | |
93 | u64 obj_version; | |
94 | }; | |
95 | ||
96 | struct rbd_options { | |
cc0538b6 | 97 | bool read_only; |
602adf40 YS |
98 | }; |
99 | ||
100 | /* | |
f0f8cef5 | 101 | * an instance of the client. multiple devices may share an rbd client. |
602adf40 YS |
102 | */ |
103 | struct rbd_client { | |
104 | struct ceph_client *client; | |
105 | struct kref kref; | |
106 | struct list_head node; | |
107 | }; | |
108 | ||
109 | /* | |
f0f8cef5 | 110 | * a request completion status |
602adf40 | 111 | */ |
1fec7093 YS |
112 | struct rbd_req_status { |
113 | int done; | |
114 | int rc; | |
115 | u64 bytes; | |
116 | }; | |
117 | ||
118 | /* | |
119 | * a collection of requests | |
120 | */ | |
121 | struct rbd_req_coll { | |
122 | int total; | |
123 | int num_done; | |
124 | struct kref kref; | |
125 | struct rbd_req_status status[0]; | |
602adf40 YS |
126 | }; |
127 | ||
f0f8cef5 AE |
128 | /* |
129 | * a single io request | |
130 | */ | |
131 | struct rbd_request { | |
132 | struct request *rq; /* blk layer request */ | |
133 | struct bio *bio; /* cloned bio */ | |
134 | struct page **pages; /* list of used pages */ | |
135 | u64 len; | |
136 | int coll_index; | |
137 | struct rbd_req_coll *coll; | |
138 | }; | |
139 | ||
dfc5606d YS |
140 | struct rbd_snap { |
141 | struct device dev; | |
142 | const char *name; | |
3591538f | 143 | u64 size; |
dfc5606d YS |
144 | struct list_head node; |
145 | u64 id; | |
146 | }; | |
147 | ||
602adf40 YS |
148 | /* |
149 | * a single device | |
150 | */ | |
151 | struct rbd_device { | |
de71a297 | 152 | int dev_id; /* blkdev unique id */ |
602adf40 YS |
153 | |
154 | int major; /* blkdev assigned major */ | |
155 | struct gendisk *disk; /* blkdev's gendisk and rq */ | |
156 | struct request_queue *q; | |
157 | ||
f8c38929 | 158 | struct rbd_options rbd_opts; |
602adf40 YS |
159 | struct rbd_client *rbd_client; |
160 | ||
161 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ | |
162 | ||
163 | spinlock_t lock; /* queue lock */ | |
164 | ||
165 | struct rbd_image_header header; | |
0bed54dc AE |
166 | char *image_name; |
167 | size_t image_name_len; | |
168 | char *header_name; | |
d22f76e7 | 169 | char *pool_name; |
9bb2f334 | 170 | int pool_id; |
602adf40 | 171 | |
59c2be1e YS |
172 | struct ceph_osd_event *watch_event; |
173 | struct ceph_osd_request *watch_request; | |
174 | ||
c666601a JD |
175 | /* protects updating the header */ |
176 | struct rw_semaphore header_rwsem; | |
e88a36ec | 177 | /* name of the snapshot this device reads from */ |
820a5f3e | 178 | char *snap_name; |
e88a36ec | 179 | /* id of the snapshot this device reads from */ |
77dfe99f | 180 | u64 snap_id; /* current snapshot id */ |
e88a36ec JD |
181 | /* whether the snap_id this device reads from still exists */ |
182 | bool snap_exists; | |
cc0538b6 | 183 | bool read_only; |
602adf40 YS |
184 | |
185 | struct list_head node; | |
dfc5606d YS |
186 | |
187 | /* list of snapshots */ | |
188 | struct list_head snaps; | |
189 | ||
190 | /* sysfs related */ | |
191 | struct device dev; | |
192 | }; | |
193 | ||
602adf40 | 194 | static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ |
e124a82f | 195 | |
602adf40 | 196 | static LIST_HEAD(rbd_dev_list); /* devices */ |
e124a82f AE |
197 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
198 | ||
432b8587 AE |
199 | static LIST_HEAD(rbd_client_list); /* clients */ |
200 | static DEFINE_SPINLOCK(rbd_client_list_lock); | |
602adf40 | 201 | |
dfc5606d YS |
202 | static int __rbd_init_snaps_header(struct rbd_device *rbd_dev); |
203 | static void rbd_dev_release(struct device *dev); | |
dfc5606d YS |
204 | static ssize_t rbd_snap_add(struct device *dev, |
205 | struct device_attribute *attr, | |
206 | const char *buf, | |
207 | size_t count); | |
14e7085d | 208 | static void __rbd_remove_snap_dev(struct rbd_snap *snap); |
dfc5606d | 209 | |
f0f8cef5 AE |
210 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
211 | size_t count); | |
212 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, | |
213 | size_t count); | |
214 | ||
215 | static struct bus_attribute rbd_bus_attrs[] = { | |
216 | __ATTR(add, S_IWUSR, NULL, rbd_add), | |
217 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), | |
218 | __ATTR_NULL | |
219 | }; | |
220 | ||
221 | static struct bus_type rbd_bus_type = { | |
222 | .name = "rbd", | |
223 | .bus_attrs = rbd_bus_attrs, | |
224 | }; | |
225 | ||
226 | static void rbd_root_dev_release(struct device *dev) | |
227 | { | |
228 | } | |
229 | ||
230 | static struct device rbd_root_dev = { | |
231 | .init_name = "rbd", | |
232 | .release = rbd_root_dev_release, | |
233 | }; | |
234 | ||
dfc5606d | 235 | |
dfc5606d YS |
236 | static struct device *rbd_get_dev(struct rbd_device *rbd_dev) |
237 | { | |
238 | return get_device(&rbd_dev->dev); | |
239 | } | |
240 | ||
241 | static void rbd_put_dev(struct rbd_device *rbd_dev) | |
242 | { | |
243 | put_device(&rbd_dev->dev); | |
244 | } | |
602adf40 | 245 | |
1fe5e993 | 246 | static int rbd_refresh_header(struct rbd_device *rbd_dev, u64 *hver); |
59c2be1e | 247 | |
602adf40 YS |
248 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
249 | { | |
f0f8cef5 | 250 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
602adf40 | 251 | |
602adf40 YS |
252 | if ((mode & FMODE_WRITE) && rbd_dev->read_only) |
253 | return -EROFS; | |
254 | ||
340c7a2b AE |
255 | rbd_get_dev(rbd_dev); |
256 | set_device_ro(bdev, rbd_dev->read_only); | |
257 | ||
602adf40 YS |
258 | return 0; |
259 | } | |
260 | ||
dfc5606d YS |
261 | static int rbd_release(struct gendisk *disk, fmode_t mode) |
262 | { | |
263 | struct rbd_device *rbd_dev = disk->private_data; | |
264 | ||
265 | rbd_put_dev(rbd_dev); | |
266 | ||
267 | return 0; | |
268 | } | |
269 | ||
602adf40 YS |
270 | static const struct block_device_operations rbd_bd_ops = { |
271 | .owner = THIS_MODULE, | |
272 | .open = rbd_open, | |
dfc5606d | 273 | .release = rbd_release, |
602adf40 YS |
274 | }; |
275 | ||
276 | /* | |
277 | * Initialize an rbd client instance. | |
43ae4701 | 278 | * We own *ceph_opts. |
602adf40 | 279 | */ |
f8c38929 | 280 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) |
602adf40 YS |
281 | { |
282 | struct rbd_client *rbdc; | |
283 | int ret = -ENOMEM; | |
284 | ||
285 | dout("rbd_client_create\n"); | |
286 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); | |
287 | if (!rbdc) | |
288 | goto out_opt; | |
289 | ||
290 | kref_init(&rbdc->kref); | |
291 | INIT_LIST_HEAD(&rbdc->node); | |
292 | ||
bc534d86 AE |
293 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
294 | ||
43ae4701 | 295 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
602adf40 | 296 | if (IS_ERR(rbdc->client)) |
bc534d86 | 297 | goto out_mutex; |
43ae4701 | 298 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
602adf40 YS |
299 | |
300 | ret = ceph_open_session(rbdc->client); | |
301 | if (ret < 0) | |
302 | goto out_err; | |
303 | ||
432b8587 | 304 | spin_lock(&rbd_client_list_lock); |
602adf40 | 305 | list_add_tail(&rbdc->node, &rbd_client_list); |
432b8587 | 306 | spin_unlock(&rbd_client_list_lock); |
602adf40 | 307 | |
bc534d86 AE |
308 | mutex_unlock(&ctl_mutex); |
309 | ||
602adf40 YS |
310 | dout("rbd_client_create created %p\n", rbdc); |
311 | return rbdc; | |
312 | ||
313 | out_err: | |
314 | ceph_destroy_client(rbdc->client); | |
bc534d86 AE |
315 | out_mutex: |
316 | mutex_unlock(&ctl_mutex); | |
602adf40 YS |
317 | kfree(rbdc); |
318 | out_opt: | |
43ae4701 AE |
319 | if (ceph_opts) |
320 | ceph_destroy_options(ceph_opts); | |
28f259b7 | 321 | return ERR_PTR(ret); |
602adf40 YS |
322 | } |
323 | ||
324 | /* | |
1f7ba331 AE |
325 | * Find a ceph client with specific addr and configuration. If |
326 | * found, bump its reference count. | |
602adf40 | 327 | */ |
1f7ba331 | 328 | static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) |
602adf40 YS |
329 | { |
330 | struct rbd_client *client_node; | |
1f7ba331 | 331 | bool found = false; |
602adf40 | 332 | |
43ae4701 | 333 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
602adf40 YS |
334 | return NULL; |
335 | ||
1f7ba331 AE |
336 | spin_lock(&rbd_client_list_lock); |
337 | list_for_each_entry(client_node, &rbd_client_list, node) { | |
338 | if (!ceph_compare_options(ceph_opts, client_node->client)) { | |
339 | kref_get(&client_node->kref); | |
340 | found = true; | |
341 | break; | |
342 | } | |
343 | } | |
344 | spin_unlock(&rbd_client_list_lock); | |
345 | ||
346 | return found ? client_node : NULL; | |
602adf40 YS |
347 | } |
348 | ||
59c2be1e YS |
349 | /* |
350 | * mount options | |
351 | */ | |
352 | enum { | |
59c2be1e YS |
353 | Opt_last_int, |
354 | /* int args above */ | |
355 | Opt_last_string, | |
356 | /* string args above */ | |
cc0538b6 AE |
357 | Opt_read_only, |
358 | Opt_read_write, | |
359 | /* Boolean args above */ | |
360 | Opt_last_bool, | |
59c2be1e YS |
361 | }; |
362 | ||
43ae4701 | 363 | static match_table_t rbd_opts_tokens = { |
59c2be1e YS |
364 | /* int args above */ |
365 | /* string args above */ | |
cc0538b6 AE |
366 | {Opt_read_only, "read_only"}, |
367 | {Opt_read_only, "ro"}, /* Alternate spelling */ | |
368 | {Opt_read_write, "read_write"}, | |
369 | {Opt_read_write, "rw"}, /* Alternate spelling */ | |
370 | /* Boolean args above */ | |
59c2be1e YS |
371 | {-1, NULL} |
372 | }; | |
373 | ||
374 | static int parse_rbd_opts_token(char *c, void *private) | |
375 | { | |
43ae4701 | 376 | struct rbd_options *rbd_opts = private; |
59c2be1e YS |
377 | substring_t argstr[MAX_OPT_ARGS]; |
378 | int token, intval, ret; | |
379 | ||
43ae4701 | 380 | token = match_token(c, rbd_opts_tokens, argstr); |
59c2be1e YS |
381 | if (token < 0) |
382 | return -EINVAL; | |
383 | ||
384 | if (token < Opt_last_int) { | |
385 | ret = match_int(&argstr[0], &intval); | |
386 | if (ret < 0) { | |
387 | pr_err("bad mount option arg (not int) " | |
388 | "at '%s'\n", c); | |
389 | return ret; | |
390 | } | |
391 | dout("got int token %d val %d\n", token, intval); | |
392 | } else if (token > Opt_last_int && token < Opt_last_string) { | |
393 | dout("got string token %d val %s\n", token, | |
394 | argstr[0].from); | |
cc0538b6 AE |
395 | } else if (token > Opt_last_string && token < Opt_last_bool) { |
396 | dout("got Boolean token %d\n", token); | |
59c2be1e YS |
397 | } else { |
398 | dout("got token %d\n", token); | |
399 | } | |
400 | ||
401 | switch (token) { | |
cc0538b6 AE |
402 | case Opt_read_only: |
403 | rbd_opts->read_only = true; | |
404 | break; | |
405 | case Opt_read_write: | |
406 | rbd_opts->read_only = false; | |
407 | break; | |
59c2be1e YS |
408 | default: |
409 | BUG_ON(token); | |
410 | } | |
411 | return 0; | |
412 | } | |
413 | ||
602adf40 YS |
414 | /* |
415 | * Get a ceph client with specific addr and configuration, if one does | |
416 | * not exist create it. | |
417 | */ | |
f8c38929 AE |
418 | static int rbd_get_client(struct rbd_device *rbd_dev, const char *mon_addr, |
419 | size_t mon_addr_len, char *options) | |
602adf40 | 420 | { |
f8c38929 | 421 | struct rbd_options *rbd_opts = &rbd_dev->rbd_opts; |
43ae4701 | 422 | struct ceph_options *ceph_opts; |
f8c38929 | 423 | struct rbd_client *rbdc; |
59c2be1e | 424 | |
cc0538b6 | 425 | rbd_opts->read_only = RBD_READ_ONLY_DEFAULT; |
602adf40 | 426 | |
43ae4701 AE |
427 | ceph_opts = ceph_parse_options(options, mon_addr, |
428 | mon_addr + mon_addr_len, | |
429 | parse_rbd_opts_token, rbd_opts); | |
f8c38929 AE |
430 | if (IS_ERR(ceph_opts)) |
431 | return PTR_ERR(ceph_opts); | |
602adf40 | 432 | |
1f7ba331 | 433 | rbdc = rbd_client_find(ceph_opts); |
602adf40 | 434 | if (rbdc) { |
602adf40 | 435 | /* using an existing client */ |
43ae4701 | 436 | ceph_destroy_options(ceph_opts); |
f8c38929 AE |
437 | } else { |
438 | rbdc = rbd_client_create(ceph_opts); | |
439 | if (IS_ERR(rbdc)) | |
440 | return PTR_ERR(rbdc); | |
602adf40 | 441 | } |
f8c38929 | 442 | rbd_dev->rbd_client = rbdc; |
602adf40 | 443 | |
f8c38929 | 444 | return 0; |
602adf40 YS |
445 | } |
446 | ||
447 | /* | |
448 | * Destroy ceph client | |
d23a4b3f | 449 | * |
432b8587 | 450 | * Caller must hold rbd_client_list_lock. |
602adf40 YS |
451 | */ |
452 | static void rbd_client_release(struct kref *kref) | |
453 | { | |
454 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); | |
455 | ||
456 | dout("rbd_release_client %p\n", rbdc); | |
cd9d9f5d | 457 | spin_lock(&rbd_client_list_lock); |
602adf40 | 458 | list_del(&rbdc->node); |
cd9d9f5d | 459 | spin_unlock(&rbd_client_list_lock); |
602adf40 YS |
460 | |
461 | ceph_destroy_client(rbdc->client); | |
462 | kfree(rbdc); | |
463 | } | |
464 | ||
465 | /* | |
466 | * Drop reference to ceph client node. If it's not referenced anymore, release | |
467 | * it. | |
468 | */ | |
469 | static void rbd_put_client(struct rbd_device *rbd_dev) | |
470 | { | |
471 | kref_put(&rbd_dev->rbd_client->kref, rbd_client_release); | |
472 | rbd_dev->rbd_client = NULL; | |
602adf40 YS |
473 | } |
474 | ||
1fec7093 YS |
475 | /* |
476 | * Destroy requests collection | |
477 | */ | |
478 | static void rbd_coll_release(struct kref *kref) | |
479 | { | |
480 | struct rbd_req_coll *coll = | |
481 | container_of(kref, struct rbd_req_coll, kref); | |
482 | ||
483 | dout("rbd_coll_release %p\n", coll); | |
484 | kfree(coll); | |
485 | } | |
602adf40 | 486 | |
8e94af8e AE |
487 | static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) |
488 | { | |
103a150f AE |
489 | size_t size; |
490 | u32 snap_count; | |
491 | ||
492 | /* The header has to start with the magic rbd header text */ | |
493 | if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) | |
494 | return false; | |
495 | ||
496 | /* | |
497 | * The size of a snapshot header has to fit in a size_t, and | |
498 | * that limits the number of snapshots. | |
499 | */ | |
500 | snap_count = le32_to_cpu(ondisk->snap_count); | |
501 | size = SIZE_MAX - sizeof (struct ceph_snap_context); | |
502 | if (snap_count > size / sizeof (__le64)) | |
503 | return false; | |
504 | ||
505 | /* | |
506 | * Not only that, but the size of the entire the snapshot | |
507 | * header must also be representable in a size_t. | |
508 | */ | |
509 | size -= snap_count * sizeof (__le64); | |
510 | if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) | |
511 | return false; | |
512 | ||
513 | return true; | |
8e94af8e AE |
514 | } |
515 | ||
602adf40 YS |
516 | /* |
517 | * Create a new header structure, translate header format from the on-disk | |
518 | * header. | |
519 | */ | |
520 | static int rbd_header_from_disk(struct rbd_image_header *header, | |
4156d998 | 521 | struct rbd_image_header_ondisk *ondisk) |
602adf40 | 522 | { |
ccece235 | 523 | u32 snap_count; |
58c17b0e | 524 | size_t len; |
d2bb24e5 | 525 | size_t size; |
621901d6 | 526 | u32 i; |
602adf40 | 527 | |
6a52325f AE |
528 | memset(header, 0, sizeof (*header)); |
529 | ||
103a150f AE |
530 | snap_count = le32_to_cpu(ondisk->snap_count); |
531 | ||
58c17b0e AE |
532 | len = strnlen(ondisk->object_prefix, sizeof (ondisk->object_prefix)); |
533 | header->object_prefix = kmalloc(len + 1, GFP_KERNEL); | |
6a52325f | 534 | if (!header->object_prefix) |
602adf40 | 535 | return -ENOMEM; |
58c17b0e AE |
536 | memcpy(header->object_prefix, ondisk->object_prefix, len); |
537 | header->object_prefix[len] = '\0'; | |
00f1f36f | 538 | |
602adf40 | 539 | if (snap_count) { |
f785cc1d AE |
540 | u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
541 | ||
621901d6 AE |
542 | /* Save a copy of the snapshot names */ |
543 | ||
f785cc1d AE |
544 | if (snap_names_len > (u64) SIZE_MAX) |
545 | return -EIO; | |
546 | header->snap_names = kmalloc(snap_names_len, GFP_KERNEL); | |
602adf40 | 547 | if (!header->snap_names) |
6a52325f | 548 | goto out_err; |
f785cc1d AE |
549 | /* |
550 | * Note that rbd_dev_v1_header_read() guarantees | |
551 | * the ondisk buffer we're working with has | |
552 | * snap_names_len bytes beyond the end of the | |
553 | * snapshot id array, this memcpy() is safe. | |
554 | */ | |
555 | memcpy(header->snap_names, &ondisk->snaps[snap_count], | |
556 | snap_names_len); | |
6a52325f | 557 | |
621901d6 AE |
558 | /* Record each snapshot's size */ |
559 | ||
d2bb24e5 AE |
560 | size = snap_count * sizeof (*header->snap_sizes); |
561 | header->snap_sizes = kmalloc(size, GFP_KERNEL); | |
602adf40 | 562 | if (!header->snap_sizes) |
6a52325f | 563 | goto out_err; |
621901d6 AE |
564 | for (i = 0; i < snap_count; i++) |
565 | header->snap_sizes[i] = | |
566 | le64_to_cpu(ondisk->snaps[i].image_size); | |
602adf40 | 567 | } else { |
ccece235 | 568 | WARN_ON(ondisk->snap_names_len); |
602adf40 YS |
569 | header->snap_names = NULL; |
570 | header->snap_sizes = NULL; | |
571 | } | |
849b4260 | 572 | |
602adf40 YS |
573 | header->image_size = le64_to_cpu(ondisk->image_size); |
574 | header->obj_order = ondisk->options.order; | |
575 | header->crypt_type = ondisk->options.crypt_type; | |
576 | header->comp_type = ondisk->options.comp_type; | |
6a52325f AE |
577 | header->total_snaps = snap_count; |
578 | ||
621901d6 AE |
579 | /* Allocate and fill in the snapshot context */ |
580 | ||
6a52325f AE |
581 | size = sizeof (struct ceph_snap_context); |
582 | size += snap_count * sizeof (header->snapc->snaps[0]); | |
583 | header->snapc = kzalloc(size, GFP_KERNEL); | |
584 | if (!header->snapc) | |
585 | goto out_err; | |
602adf40 YS |
586 | |
587 | atomic_set(&header->snapc->nref, 1); | |
505cbb9b | 588 | header->snapc->seq = le64_to_cpu(ondisk->snap_seq); |
602adf40 | 589 | header->snapc->num_snaps = snap_count; |
621901d6 AE |
590 | for (i = 0; i < snap_count; i++) |
591 | header->snapc->snaps[i] = | |
592 | le64_to_cpu(ondisk->snaps[i].id); | |
602adf40 YS |
593 | |
594 | return 0; | |
595 | ||
6a52325f | 596 | out_err: |
849b4260 | 597 | kfree(header->snap_sizes); |
ccece235 | 598 | header->snap_sizes = NULL; |
602adf40 | 599 | kfree(header->snap_names); |
ccece235 | 600 | header->snap_names = NULL; |
6a52325f AE |
601 | kfree(header->object_prefix); |
602 | header->object_prefix = NULL; | |
ccece235 | 603 | |
00f1f36f | 604 | return -ENOMEM; |
602adf40 YS |
605 | } |
606 | ||
602adf40 YS |
607 | static int snap_by_name(struct rbd_image_header *header, const char *snap_name, |
608 | u64 *seq, u64 *size) | |
609 | { | |
610 | int i; | |
611 | char *p = header->snap_names; | |
612 | ||
00f1f36f AE |
613 | for (i = 0; i < header->total_snaps; i++) { |
614 | if (!strcmp(snap_name, p)) { | |
602adf40 | 615 | |
00f1f36f | 616 | /* Found it. Pass back its id and/or size */ |
602adf40 | 617 | |
00f1f36f AE |
618 | if (seq) |
619 | *seq = header->snapc->snaps[i]; | |
620 | if (size) | |
621 | *size = header->snap_sizes[i]; | |
622 | return i; | |
623 | } | |
624 | p += strlen(p) + 1; /* Skip ahead to the next name */ | |
625 | } | |
626 | return -ENOENT; | |
602adf40 YS |
627 | } |
628 | ||
0ce1a794 | 629 | static int rbd_header_set_snap(struct rbd_device *rbd_dev, u64 *size) |
602adf40 | 630 | { |
78dc447d | 631 | int ret; |
602adf40 | 632 | |
0ce1a794 | 633 | down_write(&rbd_dev->header_rwsem); |
602adf40 | 634 | |
0ce1a794 | 635 | if (!memcmp(rbd_dev->snap_name, RBD_SNAP_HEAD_NAME, |
cc9d734c | 636 | sizeof (RBD_SNAP_HEAD_NAME))) { |
0ce1a794 | 637 | rbd_dev->snap_id = CEPH_NOSNAP; |
e88a36ec | 638 | rbd_dev->snap_exists = false; |
cc0538b6 | 639 | rbd_dev->read_only = rbd_dev->rbd_opts.read_only; |
602adf40 | 640 | if (size) |
78dc447d | 641 | *size = rbd_dev->header.image_size; |
602adf40 | 642 | } else { |
78dc447d AE |
643 | u64 snap_id = 0; |
644 | ||
645 | ret = snap_by_name(&rbd_dev->header, rbd_dev->snap_name, | |
646 | &snap_id, size); | |
602adf40 YS |
647 | if (ret < 0) |
648 | goto done; | |
78dc447d | 649 | rbd_dev->snap_id = snap_id; |
e88a36ec | 650 | rbd_dev->snap_exists = true; |
cc0538b6 | 651 | rbd_dev->read_only = true; /* No choice for snapshots */ |
602adf40 YS |
652 | } |
653 | ||
654 | ret = 0; | |
655 | done: | |
0ce1a794 | 656 | up_write(&rbd_dev->header_rwsem); |
602adf40 YS |
657 | return ret; |
658 | } | |
659 | ||
660 | static void rbd_header_free(struct rbd_image_header *header) | |
661 | { | |
849b4260 | 662 | kfree(header->object_prefix); |
d78fd7ae | 663 | header->object_prefix = NULL; |
602adf40 | 664 | kfree(header->snap_sizes); |
d78fd7ae | 665 | header->snap_sizes = NULL; |
849b4260 | 666 | kfree(header->snap_names); |
d78fd7ae | 667 | header->snap_names = NULL; |
d1d25646 | 668 | ceph_put_snap_context(header->snapc); |
d78fd7ae | 669 | header->snapc = NULL; |
602adf40 YS |
670 | } |
671 | ||
672 | /* | |
673 | * get the actual striped segment name, offset and length | |
674 | */ | |
675 | static u64 rbd_get_segment(struct rbd_image_header *header, | |
ca1e49a6 | 676 | const char *object_prefix, |
602adf40 YS |
677 | u64 ofs, u64 len, |
678 | char *seg_name, u64 *segofs) | |
679 | { | |
680 | u64 seg = ofs >> header->obj_order; | |
681 | ||
682 | if (seg_name) | |
683 | snprintf(seg_name, RBD_MAX_SEG_NAME_LEN, | |
ca1e49a6 | 684 | "%s.%012llx", object_prefix, seg); |
602adf40 YS |
685 | |
686 | ofs = ofs & ((1 << header->obj_order) - 1); | |
687 | len = min_t(u64, len, (1 << header->obj_order) - ofs); | |
688 | ||
689 | if (segofs) | |
690 | *segofs = ofs; | |
691 | ||
692 | return len; | |
693 | } | |
694 | ||
1fec7093 YS |
695 | static int rbd_get_num_segments(struct rbd_image_header *header, |
696 | u64 ofs, u64 len) | |
697 | { | |
df111be6 AE |
698 | u64 start_seg; |
699 | u64 end_seg; | |
700 | ||
701 | if (!len) | |
702 | return 0; | |
703 | if (len - 1 > U64_MAX - ofs) | |
704 | return -ERANGE; | |
705 | ||
706 | start_seg = ofs >> header->obj_order; | |
707 | end_seg = (ofs + len - 1) >> header->obj_order; | |
708 | ||
1fec7093 YS |
709 | return end_seg - start_seg + 1; |
710 | } | |
711 | ||
029bcbd8 JD |
712 | /* |
713 | * returns the size of an object in the image | |
714 | */ | |
715 | static u64 rbd_obj_bytes(struct rbd_image_header *header) | |
716 | { | |
717 | return 1 << header->obj_order; | |
718 | } | |
719 | ||
602adf40 YS |
720 | /* |
721 | * bio helpers | |
722 | */ | |
723 | ||
724 | static void bio_chain_put(struct bio *chain) | |
725 | { | |
726 | struct bio *tmp; | |
727 | ||
728 | while (chain) { | |
729 | tmp = chain; | |
730 | chain = chain->bi_next; | |
731 | bio_put(tmp); | |
732 | } | |
733 | } | |
734 | ||
735 | /* | |
736 | * zeros a bio chain, starting at specific offset | |
737 | */ | |
738 | static void zero_bio_chain(struct bio *chain, int start_ofs) | |
739 | { | |
740 | struct bio_vec *bv; | |
741 | unsigned long flags; | |
742 | void *buf; | |
743 | int i; | |
744 | int pos = 0; | |
745 | ||
746 | while (chain) { | |
747 | bio_for_each_segment(bv, chain, i) { | |
748 | if (pos + bv->bv_len > start_ofs) { | |
749 | int remainder = max(start_ofs - pos, 0); | |
750 | buf = bvec_kmap_irq(bv, &flags); | |
751 | memset(buf + remainder, 0, | |
752 | bv->bv_len - remainder); | |
85b5aaa6 | 753 | bvec_kunmap_irq(buf, &flags); |
602adf40 YS |
754 | } |
755 | pos += bv->bv_len; | |
756 | } | |
757 | ||
758 | chain = chain->bi_next; | |
759 | } | |
760 | } | |
761 | ||
762 | /* | |
763 | * bio_chain_clone - clone a chain of bios up to a certain length. | |
764 | * might return a bio_pair that will need to be released. | |
765 | */ | |
766 | static struct bio *bio_chain_clone(struct bio **old, struct bio **next, | |
767 | struct bio_pair **bp, | |
768 | int len, gfp_t gfpmask) | |
769 | { | |
542582fc AE |
770 | struct bio *old_chain = *old; |
771 | struct bio *new_chain = NULL; | |
772 | struct bio *tail; | |
602adf40 YS |
773 | int total = 0; |
774 | ||
775 | if (*bp) { | |
776 | bio_pair_release(*bp); | |
777 | *bp = NULL; | |
778 | } | |
779 | ||
780 | while (old_chain && (total < len)) { | |
542582fc AE |
781 | struct bio *tmp; |
782 | ||
602adf40 YS |
783 | tmp = bio_kmalloc(gfpmask, old_chain->bi_max_vecs); |
784 | if (!tmp) | |
785 | goto err_out; | |
542582fc | 786 | gfpmask &= ~__GFP_WAIT; /* can't wait after the first */ |
602adf40 YS |
787 | |
788 | if (total + old_chain->bi_size > len) { | |
789 | struct bio_pair *bp; | |
790 | ||
791 | /* | |
792 | * this split can only happen with a single paged bio, | |
793 | * split_bio will BUG_ON if this is not the case | |
794 | */ | |
795 | dout("bio_chain_clone split! total=%d remaining=%d" | |
bd919d45 AE |
796 | "bi_size=%u\n", |
797 | total, len - total, old_chain->bi_size); | |
602adf40 YS |
798 | |
799 | /* split the bio. We'll release it either in the next | |
800 | call, or it will have to be released outside */ | |
593a9e7b | 801 | bp = bio_split(old_chain, (len - total) / SECTOR_SIZE); |
602adf40 YS |
802 | if (!bp) |
803 | goto err_out; | |
804 | ||
805 | __bio_clone(tmp, &bp->bio1); | |
806 | ||
807 | *next = &bp->bio2; | |
808 | } else { | |
809 | __bio_clone(tmp, old_chain); | |
810 | *next = old_chain->bi_next; | |
811 | } | |
812 | ||
813 | tmp->bi_bdev = NULL; | |
602adf40 | 814 | tmp->bi_next = NULL; |
542582fc | 815 | if (new_chain) |
602adf40 | 816 | tail->bi_next = tmp; |
542582fc AE |
817 | else |
818 | new_chain = tmp; | |
819 | tail = tmp; | |
602adf40 YS |
820 | old_chain = old_chain->bi_next; |
821 | ||
822 | total += tmp->bi_size; | |
823 | } | |
824 | ||
825 | BUG_ON(total < len); | |
826 | ||
602adf40 YS |
827 | *old = old_chain; |
828 | ||
829 | return new_chain; | |
830 | ||
831 | err_out: | |
832 | dout("bio_chain_clone with err\n"); | |
833 | bio_chain_put(new_chain); | |
834 | return NULL; | |
835 | } | |
836 | ||
837 | /* | |
838 | * helpers for osd request op vectors. | |
839 | */ | |
57cfc106 AE |
840 | static struct ceph_osd_req_op *rbd_create_rw_ops(int num_ops, |
841 | int opcode, u32 payload_len) | |
602adf40 | 842 | { |
57cfc106 AE |
843 | struct ceph_osd_req_op *ops; |
844 | ||
845 | ops = kzalloc(sizeof (*ops) * (num_ops + 1), GFP_NOIO); | |
846 | if (!ops) | |
847 | return NULL; | |
848 | ||
849 | ops[0].op = opcode; | |
850 | ||
602adf40 YS |
851 | /* |
852 | * op extent offset and length will be set later on | |
853 | * in calc_raw_layout() | |
854 | */ | |
57cfc106 AE |
855 | ops[0].payload_len = payload_len; |
856 | ||
857 | return ops; | |
602adf40 YS |
858 | } |
859 | ||
860 | static void rbd_destroy_ops(struct ceph_osd_req_op *ops) | |
861 | { | |
862 | kfree(ops); | |
863 | } | |
864 | ||
1fec7093 YS |
865 | static void rbd_coll_end_req_index(struct request *rq, |
866 | struct rbd_req_coll *coll, | |
867 | int index, | |
868 | int ret, u64 len) | |
869 | { | |
870 | struct request_queue *q; | |
871 | int min, max, i; | |
872 | ||
bd919d45 AE |
873 | dout("rbd_coll_end_req_index %p index %d ret %d len %llu\n", |
874 | coll, index, ret, (unsigned long long) len); | |
1fec7093 YS |
875 | |
876 | if (!rq) | |
877 | return; | |
878 | ||
879 | if (!coll) { | |
880 | blk_end_request(rq, ret, len); | |
881 | return; | |
882 | } | |
883 | ||
884 | q = rq->q; | |
885 | ||
886 | spin_lock_irq(q->queue_lock); | |
887 | coll->status[index].done = 1; | |
888 | coll->status[index].rc = ret; | |
889 | coll->status[index].bytes = len; | |
890 | max = min = coll->num_done; | |
891 | while (max < coll->total && coll->status[max].done) | |
892 | max++; | |
893 | ||
894 | for (i = min; i<max; i++) { | |
895 | __blk_end_request(rq, coll->status[i].rc, | |
896 | coll->status[i].bytes); | |
897 | coll->num_done++; | |
898 | kref_put(&coll->kref, rbd_coll_release); | |
899 | } | |
900 | spin_unlock_irq(q->queue_lock); | |
901 | } | |
902 | ||
903 | static void rbd_coll_end_req(struct rbd_request *req, | |
904 | int ret, u64 len) | |
905 | { | |
906 | rbd_coll_end_req_index(req->rq, req->coll, req->coll_index, ret, len); | |
907 | } | |
908 | ||
602adf40 YS |
909 | /* |
910 | * Send ceph osd request | |
911 | */ | |
912 | static int rbd_do_request(struct request *rq, | |
0ce1a794 | 913 | struct rbd_device *rbd_dev, |
602adf40 YS |
914 | struct ceph_snap_context *snapc, |
915 | u64 snapid, | |
aded07ea | 916 | const char *object_name, u64 ofs, u64 len, |
602adf40 YS |
917 | struct bio *bio, |
918 | struct page **pages, | |
919 | int num_pages, | |
920 | int flags, | |
921 | struct ceph_osd_req_op *ops, | |
1fec7093 YS |
922 | struct rbd_req_coll *coll, |
923 | int coll_index, | |
602adf40 | 924 | void (*rbd_cb)(struct ceph_osd_request *req, |
59c2be1e YS |
925 | struct ceph_msg *msg), |
926 | struct ceph_osd_request **linger_req, | |
927 | u64 *ver) | |
602adf40 YS |
928 | { |
929 | struct ceph_osd_request *req; | |
930 | struct ceph_file_layout *layout; | |
931 | int ret; | |
932 | u64 bno; | |
933 | struct timespec mtime = CURRENT_TIME; | |
934 | struct rbd_request *req_data; | |
935 | struct ceph_osd_request_head *reqhead; | |
1dbb4399 | 936 | struct ceph_osd_client *osdc; |
602adf40 | 937 | |
602adf40 | 938 | req_data = kzalloc(sizeof(*req_data), GFP_NOIO); |
1fec7093 YS |
939 | if (!req_data) { |
940 | if (coll) | |
941 | rbd_coll_end_req_index(rq, coll, coll_index, | |
942 | -ENOMEM, len); | |
943 | return -ENOMEM; | |
944 | } | |
945 | ||
946 | if (coll) { | |
947 | req_data->coll = coll; | |
948 | req_data->coll_index = coll_index; | |
949 | } | |
602adf40 | 950 | |
bd919d45 AE |
951 | dout("rbd_do_request object_name=%s ofs=%llu len=%llu\n", object_name, |
952 | (unsigned long long) ofs, (unsigned long long) len); | |
602adf40 | 953 | |
0ce1a794 | 954 | osdc = &rbd_dev->rbd_client->client->osdc; |
1dbb4399 AE |
955 | req = ceph_osdc_alloc_request(osdc, flags, snapc, ops, |
956 | false, GFP_NOIO, pages, bio); | |
4ad12621 | 957 | if (!req) { |
4ad12621 | 958 | ret = -ENOMEM; |
602adf40 YS |
959 | goto done_pages; |
960 | } | |
961 | ||
962 | req->r_callback = rbd_cb; | |
963 | ||
964 | req_data->rq = rq; | |
965 | req_data->bio = bio; | |
966 | req_data->pages = pages; | |
967 | req_data->len = len; | |
968 | ||
969 | req->r_priv = req_data; | |
970 | ||
971 | reqhead = req->r_request->front.iov_base; | |
972 | reqhead->snapid = cpu_to_le64(CEPH_NOSNAP); | |
973 | ||
aded07ea | 974 | strncpy(req->r_oid, object_name, sizeof(req->r_oid)); |
602adf40 YS |
975 | req->r_oid_len = strlen(req->r_oid); |
976 | ||
977 | layout = &req->r_file_layout; | |
978 | memset(layout, 0, sizeof(*layout)); | |
979 | layout->fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
980 | layout->fl_stripe_count = cpu_to_le32(1); | |
981 | layout->fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
0ce1a794 | 982 | layout->fl_pg_pool = cpu_to_le32(rbd_dev->pool_id); |
1dbb4399 AE |
983 | ceph_calc_raw_layout(osdc, layout, snapid, ofs, &len, &bno, |
984 | req, ops); | |
602adf40 YS |
985 | |
986 | ceph_osdc_build_request(req, ofs, &len, | |
987 | ops, | |
988 | snapc, | |
989 | &mtime, | |
990 | req->r_oid, req->r_oid_len); | |
602adf40 | 991 | |
59c2be1e | 992 | if (linger_req) { |
1dbb4399 | 993 | ceph_osdc_set_request_linger(osdc, req); |
59c2be1e YS |
994 | *linger_req = req; |
995 | } | |
996 | ||
1dbb4399 | 997 | ret = ceph_osdc_start_request(osdc, req, false); |
602adf40 YS |
998 | if (ret < 0) |
999 | goto done_err; | |
1000 | ||
1001 | if (!rbd_cb) { | |
1dbb4399 | 1002 | ret = ceph_osdc_wait_request(osdc, req); |
59c2be1e YS |
1003 | if (ver) |
1004 | *ver = le64_to_cpu(req->r_reassert_version.version); | |
bd919d45 AE |
1005 | dout("reassert_ver=%llu\n", |
1006 | (unsigned long long) | |
1007 | le64_to_cpu(req->r_reassert_version.version)); | |
602adf40 YS |
1008 | ceph_osdc_put_request(req); |
1009 | } | |
1010 | return ret; | |
1011 | ||
1012 | done_err: | |
1013 | bio_chain_put(req_data->bio); | |
1014 | ceph_osdc_put_request(req); | |
1015 | done_pages: | |
1fec7093 | 1016 | rbd_coll_end_req(req_data, ret, len); |
602adf40 | 1017 | kfree(req_data); |
602adf40 YS |
1018 | return ret; |
1019 | } | |
1020 | ||
1021 | /* | |
1022 | * Ceph osd op callback | |
1023 | */ | |
1024 | static void rbd_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) | |
1025 | { | |
1026 | struct rbd_request *req_data = req->r_priv; | |
1027 | struct ceph_osd_reply_head *replyhead; | |
1028 | struct ceph_osd_op *op; | |
1029 | __s32 rc; | |
1030 | u64 bytes; | |
1031 | int read_op; | |
1032 | ||
1033 | /* parse reply */ | |
1034 | replyhead = msg->front.iov_base; | |
1035 | WARN_ON(le32_to_cpu(replyhead->num_ops) == 0); | |
1036 | op = (void *)(replyhead + 1); | |
1037 | rc = le32_to_cpu(replyhead->result); | |
1038 | bytes = le64_to_cpu(op->extent.length); | |
895cfcc8 | 1039 | read_op = (le16_to_cpu(op->op) == CEPH_OSD_OP_READ); |
602adf40 | 1040 | |
bd919d45 AE |
1041 | dout("rbd_req_cb bytes=%llu readop=%d rc=%d\n", |
1042 | (unsigned long long) bytes, read_op, (int) rc); | |
602adf40 YS |
1043 | |
1044 | if (rc == -ENOENT && read_op) { | |
1045 | zero_bio_chain(req_data->bio, 0); | |
1046 | rc = 0; | |
1047 | } else if (rc == 0 && read_op && bytes < req_data->len) { | |
1048 | zero_bio_chain(req_data->bio, bytes); | |
1049 | bytes = req_data->len; | |
1050 | } | |
1051 | ||
1fec7093 | 1052 | rbd_coll_end_req(req_data, rc, bytes); |
602adf40 YS |
1053 | |
1054 | if (req_data->bio) | |
1055 | bio_chain_put(req_data->bio); | |
1056 | ||
1057 | ceph_osdc_put_request(req); | |
1058 | kfree(req_data); | |
1059 | } | |
1060 | ||
59c2be1e YS |
1061 | static void rbd_simple_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) |
1062 | { | |
1063 | ceph_osdc_put_request(req); | |
1064 | } | |
1065 | ||
602adf40 YS |
1066 | /* |
1067 | * Do a synchronous ceph osd operation | |
1068 | */ | |
0ce1a794 | 1069 | static int rbd_req_sync_op(struct rbd_device *rbd_dev, |
602adf40 YS |
1070 | struct ceph_snap_context *snapc, |
1071 | u64 snapid, | |
602adf40 | 1072 | int flags, |
913d2fdc | 1073 | struct ceph_osd_req_op *ops, |
aded07ea | 1074 | const char *object_name, |
602adf40 | 1075 | u64 ofs, u64 len, |
59c2be1e YS |
1076 | char *buf, |
1077 | struct ceph_osd_request **linger_req, | |
1078 | u64 *ver) | |
602adf40 YS |
1079 | { |
1080 | int ret; | |
1081 | struct page **pages; | |
1082 | int num_pages; | |
913d2fdc AE |
1083 | |
1084 | BUG_ON(ops == NULL); | |
602adf40 YS |
1085 | |
1086 | num_pages = calc_pages_for(ofs , len); | |
1087 | pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); | |
b8d0638a DC |
1088 | if (IS_ERR(pages)) |
1089 | return PTR_ERR(pages); | |
602adf40 | 1090 | |
0ce1a794 | 1091 | ret = rbd_do_request(NULL, rbd_dev, snapc, snapid, |
aded07ea | 1092 | object_name, ofs, len, NULL, |
602adf40 YS |
1093 | pages, num_pages, |
1094 | flags, | |
1095 | ops, | |
1fec7093 | 1096 | NULL, 0, |
59c2be1e YS |
1097 | NULL, |
1098 | linger_req, ver); | |
602adf40 | 1099 | if (ret < 0) |
913d2fdc | 1100 | goto done; |
602adf40 YS |
1101 | |
1102 | if ((flags & CEPH_OSD_FLAG_READ) && buf) | |
1103 | ret = ceph_copy_from_page_vector(pages, buf, ofs, ret); | |
1104 | ||
602adf40 YS |
1105 | done: |
1106 | ceph_release_page_vector(pages, num_pages); | |
1107 | return ret; | |
1108 | } | |
1109 | ||
1110 | /* | |
1111 | * Do an asynchronous ceph osd operation | |
1112 | */ | |
1113 | static int rbd_do_op(struct request *rq, | |
0ce1a794 | 1114 | struct rbd_device *rbd_dev, |
602adf40 YS |
1115 | struct ceph_snap_context *snapc, |
1116 | u64 snapid, | |
d1f57ea6 | 1117 | int opcode, int flags, |
602adf40 | 1118 | u64 ofs, u64 len, |
1fec7093 YS |
1119 | struct bio *bio, |
1120 | struct rbd_req_coll *coll, | |
1121 | int coll_index) | |
602adf40 YS |
1122 | { |
1123 | char *seg_name; | |
1124 | u64 seg_ofs; | |
1125 | u64 seg_len; | |
1126 | int ret; | |
1127 | struct ceph_osd_req_op *ops; | |
1128 | u32 payload_len; | |
1129 | ||
1130 | seg_name = kmalloc(RBD_MAX_SEG_NAME_LEN + 1, GFP_NOIO); | |
1131 | if (!seg_name) | |
1132 | return -ENOMEM; | |
1133 | ||
1134 | seg_len = rbd_get_segment(&rbd_dev->header, | |
ca1e49a6 | 1135 | rbd_dev->header.object_prefix, |
602adf40 YS |
1136 | ofs, len, |
1137 | seg_name, &seg_ofs); | |
602adf40 YS |
1138 | |
1139 | payload_len = (flags & CEPH_OSD_FLAG_WRITE ? seg_len : 0); | |
1140 | ||
57cfc106 AE |
1141 | ret = -ENOMEM; |
1142 | ops = rbd_create_rw_ops(1, opcode, payload_len); | |
1143 | if (!ops) | |
602adf40 YS |
1144 | goto done; |
1145 | ||
1146 | /* we've taken care of segment sizes earlier when we | |
1147 | cloned the bios. We should never have a segment | |
1148 | truncated at this point */ | |
1149 | BUG_ON(seg_len < len); | |
1150 | ||
1151 | ret = rbd_do_request(rq, rbd_dev, snapc, snapid, | |
1152 | seg_name, seg_ofs, seg_len, | |
1153 | bio, | |
1154 | NULL, 0, | |
1155 | flags, | |
1156 | ops, | |
1fec7093 | 1157 | coll, coll_index, |
59c2be1e | 1158 | rbd_req_cb, 0, NULL); |
11f77002 SW |
1159 | |
1160 | rbd_destroy_ops(ops); | |
602adf40 YS |
1161 | done: |
1162 | kfree(seg_name); | |
1163 | return ret; | |
1164 | } | |
1165 | ||
1166 | /* | |
1167 | * Request async osd write | |
1168 | */ | |
1169 | static int rbd_req_write(struct request *rq, | |
1170 | struct rbd_device *rbd_dev, | |
1171 | struct ceph_snap_context *snapc, | |
1172 | u64 ofs, u64 len, | |
1fec7093 YS |
1173 | struct bio *bio, |
1174 | struct rbd_req_coll *coll, | |
1175 | int coll_index) | |
602adf40 YS |
1176 | { |
1177 | return rbd_do_op(rq, rbd_dev, snapc, CEPH_NOSNAP, | |
1178 | CEPH_OSD_OP_WRITE, | |
1179 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, | |
1fec7093 | 1180 | ofs, len, bio, coll, coll_index); |
602adf40 YS |
1181 | } |
1182 | ||
1183 | /* | |
1184 | * Request async osd read | |
1185 | */ | |
1186 | static int rbd_req_read(struct request *rq, | |
1187 | struct rbd_device *rbd_dev, | |
1188 | u64 snapid, | |
1189 | u64 ofs, u64 len, | |
1fec7093 YS |
1190 | struct bio *bio, |
1191 | struct rbd_req_coll *coll, | |
1192 | int coll_index) | |
602adf40 YS |
1193 | { |
1194 | return rbd_do_op(rq, rbd_dev, NULL, | |
b06e6a6b | 1195 | snapid, |
602adf40 YS |
1196 | CEPH_OSD_OP_READ, |
1197 | CEPH_OSD_FLAG_READ, | |
1fec7093 | 1198 | ofs, len, bio, coll, coll_index); |
602adf40 YS |
1199 | } |
1200 | ||
1201 | /* | |
1202 | * Request sync osd read | |
1203 | */ | |
0ce1a794 | 1204 | static int rbd_req_sync_read(struct rbd_device *rbd_dev, |
602adf40 | 1205 | u64 snapid, |
aded07ea | 1206 | const char *object_name, |
602adf40 | 1207 | u64 ofs, u64 len, |
59c2be1e YS |
1208 | char *buf, |
1209 | u64 *ver) | |
602adf40 | 1210 | { |
913d2fdc AE |
1211 | struct ceph_osd_req_op *ops; |
1212 | int ret; | |
1213 | ||
1214 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_READ, 0); | |
1215 | if (!ops) | |
1216 | return -ENOMEM; | |
1217 | ||
1218 | ret = rbd_req_sync_op(rbd_dev, NULL, | |
b06e6a6b | 1219 | snapid, |
602adf40 | 1220 | CEPH_OSD_FLAG_READ, |
913d2fdc AE |
1221 | ops, object_name, ofs, len, buf, NULL, ver); |
1222 | rbd_destroy_ops(ops); | |
1223 | ||
1224 | return ret; | |
602adf40 YS |
1225 | } |
1226 | ||
1227 | /* | |
59c2be1e YS |
1228 | * Request sync osd watch |
1229 | */ | |
0ce1a794 | 1230 | static int rbd_req_sync_notify_ack(struct rbd_device *rbd_dev, |
59c2be1e | 1231 | u64 ver, |
7f0a24d8 | 1232 | u64 notify_id) |
59c2be1e YS |
1233 | { |
1234 | struct ceph_osd_req_op *ops; | |
11f77002 SW |
1235 | int ret; |
1236 | ||
57cfc106 AE |
1237 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_NOTIFY_ACK, 0); |
1238 | if (!ops) | |
1239 | return -ENOMEM; | |
59c2be1e | 1240 | |
a71b891b | 1241 | ops[0].watch.ver = cpu_to_le64(ver); |
59c2be1e YS |
1242 | ops[0].watch.cookie = notify_id; |
1243 | ops[0].watch.flag = 0; | |
1244 | ||
0ce1a794 | 1245 | ret = rbd_do_request(NULL, rbd_dev, NULL, CEPH_NOSNAP, |
7f0a24d8 | 1246 | rbd_dev->header_name, 0, 0, NULL, |
ad4f232f | 1247 | NULL, 0, |
59c2be1e YS |
1248 | CEPH_OSD_FLAG_READ, |
1249 | ops, | |
1fec7093 | 1250 | NULL, 0, |
59c2be1e YS |
1251 | rbd_simple_req_cb, 0, NULL); |
1252 | ||
1253 | rbd_destroy_ops(ops); | |
1254 | return ret; | |
1255 | } | |
1256 | ||
1257 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) | |
1258 | { | |
0ce1a794 | 1259 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
a71b891b | 1260 | u64 hver; |
13143d2d SW |
1261 | int rc; |
1262 | ||
0ce1a794 | 1263 | if (!rbd_dev) |
59c2be1e YS |
1264 | return; |
1265 | ||
bd919d45 AE |
1266 | dout("rbd_watch_cb %s notify_id=%llu opcode=%u\n", |
1267 | rbd_dev->header_name, (unsigned long long) notify_id, | |
1268 | (unsigned int) opcode); | |
1fe5e993 | 1269 | rc = rbd_refresh_header(rbd_dev, &hver); |
13143d2d | 1270 | if (rc) |
f0f8cef5 | 1271 | pr_warning(RBD_DRV_NAME "%d got notification but failed to " |
0ce1a794 | 1272 | " update snaps: %d\n", rbd_dev->major, rc); |
59c2be1e | 1273 | |
7f0a24d8 | 1274 | rbd_req_sync_notify_ack(rbd_dev, hver, notify_id); |
59c2be1e YS |
1275 | } |
1276 | ||
1277 | /* | |
1278 | * Request sync osd watch | |
1279 | */ | |
0e6f322d | 1280 | static int rbd_req_sync_watch(struct rbd_device *rbd_dev) |
59c2be1e YS |
1281 | { |
1282 | struct ceph_osd_req_op *ops; | |
0ce1a794 | 1283 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
57cfc106 | 1284 | int ret; |
59c2be1e | 1285 | |
57cfc106 AE |
1286 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_WATCH, 0); |
1287 | if (!ops) | |
1288 | return -ENOMEM; | |
59c2be1e YS |
1289 | |
1290 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, 0, | |
0ce1a794 | 1291 | (void *)rbd_dev, &rbd_dev->watch_event); |
59c2be1e YS |
1292 | if (ret < 0) |
1293 | goto fail; | |
1294 | ||
0e6f322d | 1295 | ops[0].watch.ver = cpu_to_le64(rbd_dev->header.obj_version); |
0ce1a794 | 1296 | ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); |
59c2be1e YS |
1297 | ops[0].watch.flag = 1; |
1298 | ||
0ce1a794 | 1299 | ret = rbd_req_sync_op(rbd_dev, NULL, |
59c2be1e | 1300 | CEPH_NOSNAP, |
59c2be1e YS |
1301 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
1302 | ops, | |
0e6f322d AE |
1303 | rbd_dev->header_name, |
1304 | 0, 0, NULL, | |
0ce1a794 | 1305 | &rbd_dev->watch_request, NULL); |
59c2be1e YS |
1306 | |
1307 | if (ret < 0) | |
1308 | goto fail_event; | |
1309 | ||
1310 | rbd_destroy_ops(ops); | |
1311 | return 0; | |
1312 | ||
1313 | fail_event: | |
0ce1a794 AE |
1314 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
1315 | rbd_dev->watch_event = NULL; | |
59c2be1e YS |
1316 | fail: |
1317 | rbd_destroy_ops(ops); | |
1318 | return ret; | |
1319 | } | |
1320 | ||
79e3057c YS |
1321 | /* |
1322 | * Request sync osd unwatch | |
1323 | */ | |
070c633f | 1324 | static int rbd_req_sync_unwatch(struct rbd_device *rbd_dev) |
79e3057c YS |
1325 | { |
1326 | struct ceph_osd_req_op *ops; | |
57cfc106 | 1327 | int ret; |
79e3057c | 1328 | |
57cfc106 AE |
1329 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_WATCH, 0); |
1330 | if (!ops) | |
1331 | return -ENOMEM; | |
79e3057c YS |
1332 | |
1333 | ops[0].watch.ver = 0; | |
0ce1a794 | 1334 | ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); |
79e3057c YS |
1335 | ops[0].watch.flag = 0; |
1336 | ||
0ce1a794 | 1337 | ret = rbd_req_sync_op(rbd_dev, NULL, |
79e3057c | 1338 | CEPH_NOSNAP, |
79e3057c YS |
1339 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
1340 | ops, | |
070c633f AE |
1341 | rbd_dev->header_name, |
1342 | 0, 0, NULL, NULL, NULL); | |
1343 | ||
79e3057c YS |
1344 | |
1345 | rbd_destroy_ops(ops); | |
0ce1a794 AE |
1346 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
1347 | rbd_dev->watch_event = NULL; | |
79e3057c YS |
1348 | return ret; |
1349 | } | |
1350 | ||
59c2be1e | 1351 | struct rbd_notify_info { |
0ce1a794 | 1352 | struct rbd_device *rbd_dev; |
59c2be1e YS |
1353 | }; |
1354 | ||
1355 | static void rbd_notify_cb(u64 ver, u64 notify_id, u8 opcode, void *data) | |
1356 | { | |
0ce1a794 AE |
1357 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
1358 | if (!rbd_dev) | |
59c2be1e YS |
1359 | return; |
1360 | ||
bd919d45 AE |
1361 | dout("rbd_notify_cb %s notify_id=%llu opcode=%u\n", |
1362 | rbd_dev->header_name, (unsigned long long) notify_id, | |
1363 | (unsigned int) opcode); | |
59c2be1e YS |
1364 | } |
1365 | ||
1366 | /* | |
1367 | * Request sync osd notify | |
1368 | */ | |
4cb16250 | 1369 | static int rbd_req_sync_notify(struct rbd_device *rbd_dev) |
59c2be1e YS |
1370 | { |
1371 | struct ceph_osd_req_op *ops; | |
0ce1a794 | 1372 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
59c2be1e YS |
1373 | struct ceph_osd_event *event; |
1374 | struct rbd_notify_info info; | |
1375 | int payload_len = sizeof(u32) + sizeof(u32); | |
1376 | int ret; | |
1377 | ||
57cfc106 AE |
1378 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_NOTIFY, payload_len); |
1379 | if (!ops) | |
1380 | return -ENOMEM; | |
59c2be1e | 1381 | |
0ce1a794 | 1382 | info.rbd_dev = rbd_dev; |
59c2be1e YS |
1383 | |
1384 | ret = ceph_osdc_create_event(osdc, rbd_notify_cb, 1, | |
1385 | (void *)&info, &event); | |
1386 | if (ret < 0) | |
1387 | goto fail; | |
1388 | ||
1389 | ops[0].watch.ver = 1; | |
1390 | ops[0].watch.flag = 1; | |
1391 | ops[0].watch.cookie = event->cookie; | |
1392 | ops[0].watch.prot_ver = RADOS_NOTIFY_VER; | |
1393 | ops[0].watch.timeout = 12; | |
1394 | ||
0ce1a794 | 1395 | ret = rbd_req_sync_op(rbd_dev, NULL, |
59c2be1e | 1396 | CEPH_NOSNAP, |
59c2be1e YS |
1397 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
1398 | ops, | |
4cb16250 AE |
1399 | rbd_dev->header_name, |
1400 | 0, 0, NULL, NULL, NULL); | |
59c2be1e YS |
1401 | if (ret < 0) |
1402 | goto fail_event; | |
1403 | ||
1404 | ret = ceph_osdc_wait_event(event, CEPH_OSD_TIMEOUT_DEFAULT); | |
1405 | dout("ceph_osdc_wait_event returned %d\n", ret); | |
1406 | rbd_destroy_ops(ops); | |
1407 | return 0; | |
1408 | ||
1409 | fail_event: | |
1410 | ceph_osdc_cancel_event(event); | |
1411 | fail: | |
1412 | rbd_destroy_ops(ops); | |
1413 | return ret; | |
1414 | } | |
1415 | ||
602adf40 YS |
1416 | /* |
1417 | * Request sync osd read | |
1418 | */ | |
0ce1a794 | 1419 | static int rbd_req_sync_exec(struct rbd_device *rbd_dev, |
aded07ea AE |
1420 | const char *object_name, |
1421 | const char *class_name, | |
1422 | const char *method_name, | |
602adf40 | 1423 | const char *data, |
59c2be1e YS |
1424 | int len, |
1425 | u64 *ver) | |
602adf40 YS |
1426 | { |
1427 | struct ceph_osd_req_op *ops; | |
aded07ea AE |
1428 | int class_name_len = strlen(class_name); |
1429 | int method_name_len = strlen(method_name); | |
57cfc106 AE |
1430 | int ret; |
1431 | ||
1432 | ops = rbd_create_rw_ops(1, CEPH_OSD_OP_CALL, | |
aded07ea | 1433 | class_name_len + method_name_len + len); |
57cfc106 AE |
1434 | if (!ops) |
1435 | return -ENOMEM; | |
602adf40 | 1436 | |
aded07ea AE |
1437 | ops[0].cls.class_name = class_name; |
1438 | ops[0].cls.class_len = (__u8) class_name_len; | |
1439 | ops[0].cls.method_name = method_name; | |
1440 | ops[0].cls.method_len = (__u8) method_name_len; | |
602adf40 YS |
1441 | ops[0].cls.argc = 0; |
1442 | ops[0].cls.indata = data; | |
1443 | ops[0].cls.indata_len = len; | |
1444 | ||
0ce1a794 | 1445 | ret = rbd_req_sync_op(rbd_dev, NULL, |
602adf40 | 1446 | CEPH_NOSNAP, |
602adf40 YS |
1447 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
1448 | ops, | |
d1f57ea6 | 1449 | object_name, 0, 0, NULL, NULL, ver); |
602adf40 YS |
1450 | |
1451 | rbd_destroy_ops(ops); | |
1452 | ||
1453 | dout("cls_exec returned %d\n", ret); | |
1454 | return ret; | |
1455 | } | |
1456 | ||
1fec7093 YS |
1457 | static struct rbd_req_coll *rbd_alloc_coll(int num_reqs) |
1458 | { | |
1459 | struct rbd_req_coll *coll = | |
1460 | kzalloc(sizeof(struct rbd_req_coll) + | |
1461 | sizeof(struct rbd_req_status) * num_reqs, | |
1462 | GFP_ATOMIC); | |
1463 | ||
1464 | if (!coll) | |
1465 | return NULL; | |
1466 | coll->total = num_reqs; | |
1467 | kref_init(&coll->kref); | |
1468 | return coll; | |
1469 | } | |
1470 | ||
602adf40 YS |
1471 | /* |
1472 | * block device queue callback | |
1473 | */ | |
1474 | static void rbd_rq_fn(struct request_queue *q) | |
1475 | { | |
1476 | struct rbd_device *rbd_dev = q->queuedata; | |
1477 | struct request *rq; | |
1478 | struct bio_pair *bp = NULL; | |
1479 | ||
00f1f36f | 1480 | while ((rq = blk_fetch_request(q))) { |
602adf40 YS |
1481 | struct bio *bio; |
1482 | struct bio *rq_bio, *next_bio = NULL; | |
1483 | bool do_write; | |
bd919d45 AE |
1484 | unsigned int size; |
1485 | u64 op_size = 0; | |
602adf40 | 1486 | u64 ofs; |
1fec7093 YS |
1487 | int num_segs, cur_seg = 0; |
1488 | struct rbd_req_coll *coll; | |
d1d25646 | 1489 | struct ceph_snap_context *snapc; |
602adf40 | 1490 | |
602adf40 YS |
1491 | dout("fetched request\n"); |
1492 | ||
1493 | /* filter out block requests we don't understand */ | |
1494 | if ((rq->cmd_type != REQ_TYPE_FS)) { | |
1495 | __blk_end_request_all(rq, 0); | |
00f1f36f | 1496 | continue; |
602adf40 YS |
1497 | } |
1498 | ||
1499 | /* deduce our operation (read, write) */ | |
1500 | do_write = (rq_data_dir(rq) == WRITE); | |
1501 | ||
1502 | size = blk_rq_bytes(rq); | |
593a9e7b | 1503 | ofs = blk_rq_pos(rq) * SECTOR_SIZE; |
602adf40 YS |
1504 | rq_bio = rq->bio; |
1505 | if (do_write && rbd_dev->read_only) { | |
1506 | __blk_end_request_all(rq, -EROFS); | |
00f1f36f | 1507 | continue; |
602adf40 YS |
1508 | } |
1509 | ||
1510 | spin_unlock_irq(q->queue_lock); | |
1511 | ||
d1d25646 | 1512 | down_read(&rbd_dev->header_rwsem); |
e88a36ec | 1513 | |
d1d25646 | 1514 | if (rbd_dev->snap_id != CEPH_NOSNAP && !rbd_dev->snap_exists) { |
e88a36ec | 1515 | up_read(&rbd_dev->header_rwsem); |
d1d25646 JD |
1516 | dout("request for non-existent snapshot"); |
1517 | spin_lock_irq(q->queue_lock); | |
1518 | __blk_end_request_all(rq, -ENXIO); | |
1519 | continue; | |
e88a36ec JD |
1520 | } |
1521 | ||
d1d25646 JD |
1522 | snapc = ceph_get_snap_context(rbd_dev->header.snapc); |
1523 | ||
1524 | up_read(&rbd_dev->header_rwsem); | |
1525 | ||
602adf40 YS |
1526 | dout("%s 0x%x bytes at 0x%llx\n", |
1527 | do_write ? "write" : "read", | |
bd919d45 | 1528 | size, (unsigned long long) blk_rq_pos(rq) * SECTOR_SIZE); |
602adf40 | 1529 | |
1fec7093 | 1530 | num_segs = rbd_get_num_segments(&rbd_dev->header, ofs, size); |
df111be6 AE |
1531 | if (num_segs <= 0) { |
1532 | spin_lock_irq(q->queue_lock); | |
1533 | __blk_end_request_all(rq, num_segs); | |
1534 | ceph_put_snap_context(snapc); | |
1535 | continue; | |
1536 | } | |
1fec7093 YS |
1537 | coll = rbd_alloc_coll(num_segs); |
1538 | if (!coll) { | |
1539 | spin_lock_irq(q->queue_lock); | |
1540 | __blk_end_request_all(rq, -ENOMEM); | |
d1d25646 | 1541 | ceph_put_snap_context(snapc); |
00f1f36f | 1542 | continue; |
1fec7093 YS |
1543 | } |
1544 | ||
602adf40 YS |
1545 | do { |
1546 | /* a bio clone to be passed down to OSD req */ | |
bd919d45 | 1547 | dout("rq->bio->bi_vcnt=%hu\n", rq->bio->bi_vcnt); |
602adf40 | 1548 | op_size = rbd_get_segment(&rbd_dev->header, |
ca1e49a6 | 1549 | rbd_dev->header.object_prefix, |
602adf40 YS |
1550 | ofs, size, |
1551 | NULL, NULL); | |
1fec7093 | 1552 | kref_get(&coll->kref); |
602adf40 YS |
1553 | bio = bio_chain_clone(&rq_bio, &next_bio, &bp, |
1554 | op_size, GFP_ATOMIC); | |
1555 | if (!bio) { | |
1fec7093 YS |
1556 | rbd_coll_end_req_index(rq, coll, cur_seg, |
1557 | -ENOMEM, op_size); | |
1558 | goto next_seg; | |
602adf40 YS |
1559 | } |
1560 | ||
1fec7093 | 1561 | |
602adf40 YS |
1562 | /* init OSD command: write or read */ |
1563 | if (do_write) | |
1564 | rbd_req_write(rq, rbd_dev, | |
d1d25646 | 1565 | snapc, |
602adf40 | 1566 | ofs, |
1fec7093 YS |
1567 | op_size, bio, |
1568 | coll, cur_seg); | |
602adf40 YS |
1569 | else |
1570 | rbd_req_read(rq, rbd_dev, | |
77dfe99f | 1571 | rbd_dev->snap_id, |
602adf40 | 1572 | ofs, |
1fec7093 YS |
1573 | op_size, bio, |
1574 | coll, cur_seg); | |
602adf40 | 1575 | |
1fec7093 | 1576 | next_seg: |
602adf40 YS |
1577 | size -= op_size; |
1578 | ofs += op_size; | |
1579 | ||
1fec7093 | 1580 | cur_seg++; |
602adf40 YS |
1581 | rq_bio = next_bio; |
1582 | } while (size > 0); | |
1fec7093 | 1583 | kref_put(&coll->kref, rbd_coll_release); |
602adf40 YS |
1584 | |
1585 | if (bp) | |
1586 | bio_pair_release(bp); | |
602adf40 | 1587 | spin_lock_irq(q->queue_lock); |
d1d25646 JD |
1588 | |
1589 | ceph_put_snap_context(snapc); | |
602adf40 YS |
1590 | } |
1591 | } | |
1592 | ||
1593 | /* | |
1594 | * a queue callback. Makes sure that we don't create a bio that spans across | |
1595 | * multiple osd objects. One exception would be with a single page bios, | |
1596 | * which we handle later at bio_chain_clone | |
1597 | */ | |
1598 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, | |
1599 | struct bio_vec *bvec) | |
1600 | { | |
1601 | struct rbd_device *rbd_dev = q->queuedata; | |
593a9e7b AE |
1602 | unsigned int chunk_sectors; |
1603 | sector_t sector; | |
1604 | unsigned int bio_sectors; | |
602adf40 YS |
1605 | int max; |
1606 | ||
593a9e7b AE |
1607 | chunk_sectors = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); |
1608 | sector = bmd->bi_sector + get_start_sect(bmd->bi_bdev); | |
1609 | bio_sectors = bmd->bi_size >> SECTOR_SHIFT; | |
1610 | ||
602adf40 | 1611 | max = (chunk_sectors - ((sector & (chunk_sectors - 1)) |
593a9e7b | 1612 | + bio_sectors)) << SECTOR_SHIFT; |
602adf40 YS |
1613 | if (max < 0) |
1614 | max = 0; /* bio_add cannot handle a negative return */ | |
1615 | if (max <= bvec->bv_len && bio_sectors == 0) | |
1616 | return bvec->bv_len; | |
1617 | return max; | |
1618 | } | |
1619 | ||
1620 | static void rbd_free_disk(struct rbd_device *rbd_dev) | |
1621 | { | |
1622 | struct gendisk *disk = rbd_dev->disk; | |
1623 | ||
1624 | if (!disk) | |
1625 | return; | |
1626 | ||
1627 | rbd_header_free(&rbd_dev->header); | |
1628 | ||
1629 | if (disk->flags & GENHD_FL_UP) | |
1630 | del_gendisk(disk); | |
1631 | if (disk->queue) | |
1632 | blk_cleanup_queue(disk->queue); | |
1633 | put_disk(disk); | |
1634 | } | |
1635 | ||
1636 | /* | |
4156d998 AE |
1637 | * Read the complete header for the given rbd device. |
1638 | * | |
1639 | * Returns a pointer to a dynamically-allocated buffer containing | |
1640 | * the complete and validated header. Caller can pass the address | |
1641 | * of a variable that will be filled in with the version of the | |
1642 | * header object at the time it was read. | |
1643 | * | |
1644 | * Returns a pointer-coded errno if a failure occurs. | |
602adf40 | 1645 | */ |
4156d998 AE |
1646 | static struct rbd_image_header_ondisk * |
1647 | rbd_dev_v1_header_read(struct rbd_device *rbd_dev, u64 *version) | |
602adf40 | 1648 | { |
4156d998 | 1649 | struct rbd_image_header_ondisk *ondisk = NULL; |
50f7c4c9 | 1650 | u32 snap_count = 0; |
4156d998 AE |
1651 | u64 names_size = 0; |
1652 | u32 want_count; | |
1653 | int ret; | |
602adf40 | 1654 | |
00f1f36f | 1655 | /* |
4156d998 AE |
1656 | * The complete header will include an array of its 64-bit |
1657 | * snapshot ids, followed by the names of those snapshots as | |
1658 | * a contiguous block of NUL-terminated strings. Note that | |
1659 | * the number of snapshots could change by the time we read | |
1660 | * it in, in which case we re-read it. | |
00f1f36f | 1661 | */ |
4156d998 AE |
1662 | do { |
1663 | size_t size; | |
1664 | ||
1665 | kfree(ondisk); | |
1666 | ||
1667 | size = sizeof (*ondisk); | |
1668 | size += snap_count * sizeof (struct rbd_image_snap_ondisk); | |
1669 | size += names_size; | |
1670 | ondisk = kmalloc(size, GFP_KERNEL); | |
1671 | if (!ondisk) | |
1672 | return ERR_PTR(-ENOMEM); | |
1673 | ||
1674 | ret = rbd_req_sync_read(rbd_dev, CEPH_NOSNAP, | |
0bed54dc | 1675 | rbd_dev->header_name, |
4156d998 AE |
1676 | 0, size, |
1677 | (char *) ondisk, version); | |
1678 | ||
1679 | if (ret < 0) | |
1680 | goto out_err; | |
1681 | if (WARN_ON((size_t) ret < size)) { | |
1682 | ret = -ENXIO; | |
1683 | pr_warning("short header read for image %s" | |
1684 | " (want %zd got %d)\n", | |
1685 | rbd_dev->image_name, size, ret); | |
1686 | goto out_err; | |
1687 | } | |
1688 | if (!rbd_dev_ondisk_valid(ondisk)) { | |
1689 | ret = -ENXIO; | |
1690 | pr_warning("invalid header for image %s\n", | |
1691 | rbd_dev->image_name); | |
1692 | goto out_err; | |
81e759fb | 1693 | } |
602adf40 | 1694 | |
4156d998 AE |
1695 | names_size = le64_to_cpu(ondisk->snap_names_len); |
1696 | want_count = snap_count; | |
1697 | snap_count = le32_to_cpu(ondisk->snap_count); | |
1698 | } while (snap_count != want_count); | |
00f1f36f | 1699 | |
4156d998 | 1700 | return ondisk; |
00f1f36f | 1701 | |
4156d998 AE |
1702 | out_err: |
1703 | kfree(ondisk); | |
1704 | ||
1705 | return ERR_PTR(ret); | |
1706 | } | |
1707 | ||
1708 | /* | |
1709 | * reload the ondisk the header | |
1710 | */ | |
1711 | static int rbd_read_header(struct rbd_device *rbd_dev, | |
1712 | struct rbd_image_header *header) | |
1713 | { | |
1714 | struct rbd_image_header_ondisk *ondisk; | |
1715 | u64 ver = 0; | |
1716 | int ret; | |
602adf40 | 1717 | |
4156d998 AE |
1718 | ondisk = rbd_dev_v1_header_read(rbd_dev, &ver); |
1719 | if (IS_ERR(ondisk)) | |
1720 | return PTR_ERR(ondisk); | |
1721 | ret = rbd_header_from_disk(header, ondisk); | |
1722 | if (ret >= 0) | |
1723 | header->obj_version = ver; | |
1724 | kfree(ondisk); | |
1725 | ||
1726 | return ret; | |
602adf40 YS |
1727 | } |
1728 | ||
1729 | /* | |
1730 | * create a snapshot | |
1731 | */ | |
0ce1a794 | 1732 | static int rbd_header_add_snap(struct rbd_device *rbd_dev, |
602adf40 YS |
1733 | const char *snap_name, |
1734 | gfp_t gfp_flags) | |
1735 | { | |
1736 | int name_len = strlen(snap_name); | |
1737 | u64 new_snapid; | |
1738 | int ret; | |
916d4d67 | 1739 | void *data, *p, *e; |
1dbb4399 | 1740 | struct ceph_mon_client *monc; |
602adf40 YS |
1741 | |
1742 | /* we should create a snapshot only if we're pointing at the head */ | |
0ce1a794 | 1743 | if (rbd_dev->snap_id != CEPH_NOSNAP) |
602adf40 YS |
1744 | return -EINVAL; |
1745 | ||
0ce1a794 AE |
1746 | monc = &rbd_dev->rbd_client->client->monc; |
1747 | ret = ceph_monc_create_snapid(monc, rbd_dev->pool_id, &new_snapid); | |
bd919d45 | 1748 | dout("created snapid=%llu\n", (unsigned long long) new_snapid); |
602adf40 YS |
1749 | if (ret < 0) |
1750 | return ret; | |
1751 | ||
1752 | data = kmalloc(name_len + 16, gfp_flags); | |
1753 | if (!data) | |
1754 | return -ENOMEM; | |
1755 | ||
916d4d67 SW |
1756 | p = data; |
1757 | e = data + name_len + 16; | |
602adf40 | 1758 | |
916d4d67 SW |
1759 | ceph_encode_string_safe(&p, e, snap_name, name_len, bad); |
1760 | ceph_encode_64_safe(&p, e, new_snapid, bad); | |
602adf40 | 1761 | |
0bed54dc | 1762 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, |
0ce1a794 | 1763 | "rbd", "snap_add", |
d67d4be5 | 1764 | data, p - data, NULL); |
602adf40 | 1765 | |
916d4d67 | 1766 | kfree(data); |
602adf40 | 1767 | |
505cbb9b | 1768 | return ret < 0 ? ret : 0; |
602adf40 YS |
1769 | bad: |
1770 | return -ERANGE; | |
1771 | } | |
1772 | ||
dfc5606d YS |
1773 | static void __rbd_remove_all_snaps(struct rbd_device *rbd_dev) |
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) |
14e7085d | 1779 | __rbd_remove_snap_dev(snap); |
dfc5606d YS |
1780 | } |
1781 | ||
602adf40 YS |
1782 | /* |
1783 | * only read the first part of the ondisk header, without the snaps info | |
1784 | */ | |
b813623a | 1785 | static int __rbd_refresh_header(struct rbd_device *rbd_dev, u64 *hver) |
602adf40 YS |
1786 | { |
1787 | int ret; | |
1788 | struct rbd_image_header h; | |
602adf40 YS |
1789 | |
1790 | ret = rbd_read_header(rbd_dev, &h); | |
1791 | if (ret < 0) | |
1792 | return ret; | |
1793 | ||
a51aa0c0 JD |
1794 | down_write(&rbd_dev->header_rwsem); |
1795 | ||
9db4b3e3 | 1796 | /* resized? */ |
474ef7ce JD |
1797 | if (rbd_dev->snap_id == CEPH_NOSNAP) { |
1798 | sector_t size = (sector_t) h.image_size / SECTOR_SIZE; | |
1799 | ||
1800 | dout("setting size to %llu sectors", (unsigned long long) size); | |
1801 | set_capacity(rbd_dev->disk, size); | |
1802 | } | |
9db4b3e3 | 1803 | |
849b4260 | 1804 | /* rbd_dev->header.object_prefix shouldn't change */ |
602adf40 | 1805 | kfree(rbd_dev->header.snap_sizes); |
849b4260 | 1806 | kfree(rbd_dev->header.snap_names); |
d1d25646 JD |
1807 | /* osd requests may still refer to snapc */ |
1808 | ceph_put_snap_context(rbd_dev->header.snapc); | |
602adf40 | 1809 | |
b813623a AE |
1810 | if (hver) |
1811 | *hver = h.obj_version; | |
a71b891b | 1812 | rbd_dev->header.obj_version = h.obj_version; |
93a24e08 | 1813 | rbd_dev->header.image_size = h.image_size; |
602adf40 YS |
1814 | rbd_dev->header.total_snaps = h.total_snaps; |
1815 | rbd_dev->header.snapc = h.snapc; | |
1816 | rbd_dev->header.snap_names = h.snap_names; | |
1817 | rbd_dev->header.snap_sizes = h.snap_sizes; | |
849b4260 AE |
1818 | /* Free the extra copy of the object prefix */ |
1819 | WARN_ON(strcmp(rbd_dev->header.object_prefix, h.object_prefix)); | |
1820 | kfree(h.object_prefix); | |
1821 | ||
dfc5606d YS |
1822 | ret = __rbd_init_snaps_header(rbd_dev); |
1823 | ||
c666601a | 1824 | up_write(&rbd_dev->header_rwsem); |
602adf40 | 1825 | |
dfc5606d | 1826 | return ret; |
602adf40 YS |
1827 | } |
1828 | ||
1fe5e993 AE |
1829 | static int rbd_refresh_header(struct rbd_device *rbd_dev, u64 *hver) |
1830 | { | |
1831 | int ret; | |
1832 | ||
1833 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
1834 | ret = __rbd_refresh_header(rbd_dev, hver); | |
1835 | mutex_unlock(&ctl_mutex); | |
1836 | ||
1837 | return ret; | |
1838 | } | |
1839 | ||
602adf40 YS |
1840 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
1841 | { | |
1842 | struct gendisk *disk; | |
1843 | struct request_queue *q; | |
1844 | int rc; | |
593a9e7b | 1845 | u64 segment_size; |
602adf40 YS |
1846 | u64 total_size = 0; |
1847 | ||
1848 | /* contact OSD, request size info about the object being mapped */ | |
1849 | rc = rbd_read_header(rbd_dev, &rbd_dev->header); | |
1850 | if (rc) | |
1851 | return rc; | |
1852 | ||
dfc5606d YS |
1853 | /* no need to lock here, as rbd_dev is not registered yet */ |
1854 | rc = __rbd_init_snaps_header(rbd_dev); | |
1855 | if (rc) | |
1856 | return rc; | |
1857 | ||
cc9d734c | 1858 | rc = rbd_header_set_snap(rbd_dev, &total_size); |
602adf40 YS |
1859 | if (rc) |
1860 | return rc; | |
1861 | ||
1862 | /* create gendisk info */ | |
1863 | rc = -ENOMEM; | |
1864 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); | |
1865 | if (!disk) | |
1866 | goto out; | |
1867 | ||
f0f8cef5 | 1868 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
de71a297 | 1869 | rbd_dev->dev_id); |
602adf40 YS |
1870 | disk->major = rbd_dev->major; |
1871 | disk->first_minor = 0; | |
1872 | disk->fops = &rbd_bd_ops; | |
1873 | disk->private_data = rbd_dev; | |
1874 | ||
1875 | /* init rq */ | |
1876 | rc = -ENOMEM; | |
1877 | q = blk_init_queue(rbd_rq_fn, &rbd_dev->lock); | |
1878 | if (!q) | |
1879 | goto out_disk; | |
029bcbd8 | 1880 | |
593a9e7b AE |
1881 | /* We use the default size, but let's be explicit about it. */ |
1882 | blk_queue_physical_block_size(q, SECTOR_SIZE); | |
1883 | ||
029bcbd8 | 1884 | /* set io sizes to object size */ |
593a9e7b AE |
1885 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
1886 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); | |
1887 | blk_queue_max_segment_size(q, segment_size); | |
1888 | blk_queue_io_min(q, segment_size); | |
1889 | blk_queue_io_opt(q, segment_size); | |
029bcbd8 | 1890 | |
602adf40 YS |
1891 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
1892 | disk->queue = q; | |
1893 | ||
1894 | q->queuedata = rbd_dev; | |
1895 | ||
1896 | rbd_dev->disk = disk; | |
1897 | rbd_dev->q = q; | |
1898 | ||
1899 | /* finally, announce the disk to the world */ | |
593a9e7b | 1900 | set_capacity(disk, total_size / SECTOR_SIZE); |
602adf40 YS |
1901 | add_disk(disk); |
1902 | ||
1903 | pr_info("%s: added with size 0x%llx\n", | |
1904 | disk->disk_name, (unsigned long long)total_size); | |
1905 | return 0; | |
1906 | ||
1907 | out_disk: | |
1908 | put_disk(disk); | |
1909 | out: | |
1910 | return rc; | |
1911 | } | |
1912 | ||
dfc5606d YS |
1913 | /* |
1914 | sysfs | |
1915 | */ | |
1916 | ||
593a9e7b AE |
1917 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
1918 | { | |
1919 | return container_of(dev, struct rbd_device, dev); | |
1920 | } | |
1921 | ||
dfc5606d YS |
1922 | static ssize_t rbd_size_show(struct device *dev, |
1923 | struct device_attribute *attr, char *buf) | |
1924 | { | |
593a9e7b | 1925 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
a51aa0c0 JD |
1926 | sector_t size; |
1927 | ||
1928 | down_read(&rbd_dev->header_rwsem); | |
1929 | size = get_capacity(rbd_dev->disk); | |
1930 | up_read(&rbd_dev->header_rwsem); | |
dfc5606d | 1931 | |
a51aa0c0 | 1932 | return sprintf(buf, "%llu\n", (unsigned long long) size * SECTOR_SIZE); |
dfc5606d YS |
1933 | } |
1934 | ||
1935 | static ssize_t rbd_major_show(struct device *dev, | |
1936 | struct device_attribute *attr, char *buf) | |
1937 | { | |
593a9e7b | 1938 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 1939 | |
dfc5606d YS |
1940 | return sprintf(buf, "%d\n", rbd_dev->major); |
1941 | } | |
1942 | ||
1943 | static ssize_t rbd_client_id_show(struct device *dev, | |
1944 | struct device_attribute *attr, char *buf) | |
602adf40 | 1945 | { |
593a9e7b | 1946 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 1947 | |
1dbb4399 AE |
1948 | return sprintf(buf, "client%lld\n", |
1949 | ceph_client_id(rbd_dev->rbd_client->client)); | |
602adf40 YS |
1950 | } |
1951 | ||
dfc5606d YS |
1952 | static ssize_t rbd_pool_show(struct device *dev, |
1953 | struct device_attribute *attr, char *buf) | |
602adf40 | 1954 | { |
593a9e7b | 1955 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d YS |
1956 | |
1957 | return sprintf(buf, "%s\n", rbd_dev->pool_name); | |
1958 | } | |
1959 | ||
9bb2f334 AE |
1960 | static ssize_t rbd_pool_id_show(struct device *dev, |
1961 | struct device_attribute *attr, char *buf) | |
1962 | { | |
1963 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
1964 | ||
1965 | return sprintf(buf, "%d\n", rbd_dev->pool_id); | |
1966 | } | |
1967 | ||
dfc5606d YS |
1968 | static ssize_t rbd_name_show(struct device *dev, |
1969 | struct device_attribute *attr, char *buf) | |
1970 | { | |
593a9e7b | 1971 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 1972 | |
0bed54dc | 1973 | return sprintf(buf, "%s\n", rbd_dev->image_name); |
dfc5606d YS |
1974 | } |
1975 | ||
1976 | static ssize_t rbd_snap_show(struct device *dev, | |
1977 | struct device_attribute *attr, | |
1978 | char *buf) | |
1979 | { | |
593a9e7b | 1980 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d YS |
1981 | |
1982 | return sprintf(buf, "%s\n", rbd_dev->snap_name); | |
1983 | } | |
1984 | ||
1985 | static ssize_t rbd_image_refresh(struct device *dev, | |
1986 | struct device_attribute *attr, | |
1987 | const char *buf, | |
1988 | size_t size) | |
1989 | { | |
593a9e7b | 1990 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
b813623a | 1991 | int ret; |
602adf40 | 1992 | |
1fe5e993 | 1993 | ret = rbd_refresh_header(rbd_dev, NULL); |
b813623a AE |
1994 | |
1995 | return ret < 0 ? ret : size; | |
dfc5606d | 1996 | } |
602adf40 | 1997 | |
dfc5606d YS |
1998 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
1999 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); | |
2000 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); | |
2001 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); | |
9bb2f334 | 2002 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
dfc5606d YS |
2003 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
2004 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); | |
2005 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); | |
2006 | static DEVICE_ATTR(create_snap, S_IWUSR, NULL, rbd_snap_add); | |
dfc5606d YS |
2007 | |
2008 | static struct attribute *rbd_attrs[] = { | |
2009 | &dev_attr_size.attr, | |
2010 | &dev_attr_major.attr, | |
2011 | &dev_attr_client_id.attr, | |
2012 | &dev_attr_pool.attr, | |
9bb2f334 | 2013 | &dev_attr_pool_id.attr, |
dfc5606d YS |
2014 | &dev_attr_name.attr, |
2015 | &dev_attr_current_snap.attr, | |
2016 | &dev_attr_refresh.attr, | |
2017 | &dev_attr_create_snap.attr, | |
dfc5606d YS |
2018 | NULL |
2019 | }; | |
2020 | ||
2021 | static struct attribute_group rbd_attr_group = { | |
2022 | .attrs = rbd_attrs, | |
2023 | }; | |
2024 | ||
2025 | static const struct attribute_group *rbd_attr_groups[] = { | |
2026 | &rbd_attr_group, | |
2027 | NULL | |
2028 | }; | |
2029 | ||
2030 | static void rbd_sysfs_dev_release(struct device *dev) | |
2031 | { | |
2032 | } | |
2033 | ||
2034 | static struct device_type rbd_device_type = { | |
2035 | .name = "rbd", | |
2036 | .groups = rbd_attr_groups, | |
2037 | .release = rbd_sysfs_dev_release, | |
2038 | }; | |
2039 | ||
2040 | ||
2041 | /* | |
2042 | sysfs - snapshots | |
2043 | */ | |
2044 | ||
2045 | static ssize_t rbd_snap_size_show(struct device *dev, | |
2046 | struct device_attribute *attr, | |
2047 | char *buf) | |
2048 | { | |
2049 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2050 | ||
3591538f | 2051 | return sprintf(buf, "%llu\n", (unsigned long long)snap->size); |
dfc5606d YS |
2052 | } |
2053 | ||
2054 | static ssize_t rbd_snap_id_show(struct device *dev, | |
2055 | struct device_attribute *attr, | |
2056 | char *buf) | |
2057 | { | |
2058 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2059 | ||
3591538f | 2060 | return sprintf(buf, "%llu\n", (unsigned long long)snap->id); |
dfc5606d YS |
2061 | } |
2062 | ||
2063 | static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); | |
2064 | static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); | |
2065 | ||
2066 | static struct attribute *rbd_snap_attrs[] = { | |
2067 | &dev_attr_snap_size.attr, | |
2068 | &dev_attr_snap_id.attr, | |
2069 | NULL, | |
2070 | }; | |
2071 | ||
2072 | static struct attribute_group rbd_snap_attr_group = { | |
2073 | .attrs = rbd_snap_attrs, | |
2074 | }; | |
2075 | ||
2076 | static void rbd_snap_dev_release(struct device *dev) | |
2077 | { | |
2078 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2079 | kfree(snap->name); | |
2080 | kfree(snap); | |
2081 | } | |
2082 | ||
2083 | static const struct attribute_group *rbd_snap_attr_groups[] = { | |
2084 | &rbd_snap_attr_group, | |
2085 | NULL | |
2086 | }; | |
2087 | ||
2088 | static struct device_type rbd_snap_device_type = { | |
2089 | .groups = rbd_snap_attr_groups, | |
2090 | .release = rbd_snap_dev_release, | |
2091 | }; | |
2092 | ||
14e7085d | 2093 | static void __rbd_remove_snap_dev(struct rbd_snap *snap) |
dfc5606d YS |
2094 | { |
2095 | list_del(&snap->node); | |
2096 | device_unregister(&snap->dev); | |
2097 | } | |
2098 | ||
14e7085d | 2099 | static int rbd_register_snap_dev(struct rbd_snap *snap, |
dfc5606d YS |
2100 | struct device *parent) |
2101 | { | |
2102 | struct device *dev = &snap->dev; | |
2103 | int ret; | |
2104 | ||
2105 | dev->type = &rbd_snap_device_type; | |
2106 | dev->parent = parent; | |
2107 | dev->release = rbd_snap_dev_release; | |
2108 | dev_set_name(dev, "snap_%s", snap->name); | |
2109 | ret = device_register(dev); | |
2110 | ||
2111 | return ret; | |
2112 | } | |
2113 | ||
4e891e0a AE |
2114 | static struct rbd_snap *__rbd_add_snap_dev(struct rbd_device *rbd_dev, |
2115 | int i, const char *name) | |
dfc5606d | 2116 | { |
4e891e0a | 2117 | struct rbd_snap *snap; |
dfc5606d | 2118 | int ret; |
4e891e0a AE |
2119 | |
2120 | snap = kzalloc(sizeof (*snap), GFP_KERNEL); | |
dfc5606d | 2121 | if (!snap) |
4e891e0a AE |
2122 | return ERR_PTR(-ENOMEM); |
2123 | ||
2124 | ret = -ENOMEM; | |
dfc5606d | 2125 | snap->name = kstrdup(name, GFP_KERNEL); |
4e891e0a AE |
2126 | if (!snap->name) |
2127 | goto err; | |
2128 | ||
dfc5606d YS |
2129 | snap->size = rbd_dev->header.snap_sizes[i]; |
2130 | snap->id = rbd_dev->header.snapc->snaps[i]; | |
2131 | if (device_is_registered(&rbd_dev->dev)) { | |
14e7085d | 2132 | ret = rbd_register_snap_dev(snap, &rbd_dev->dev); |
dfc5606d YS |
2133 | if (ret < 0) |
2134 | goto err; | |
2135 | } | |
4e891e0a AE |
2136 | |
2137 | return snap; | |
2138 | ||
dfc5606d YS |
2139 | err: |
2140 | kfree(snap->name); | |
2141 | kfree(snap); | |
4e891e0a AE |
2142 | |
2143 | return ERR_PTR(ret); | |
dfc5606d YS |
2144 | } |
2145 | ||
2146 | /* | |
35938150 AE |
2147 | * Scan the rbd device's current snapshot list and compare it to the |
2148 | * newly-received snapshot context. Remove any existing snapshots | |
2149 | * not present in the new snapshot context. Add a new snapshot for | |
2150 | * any snaphots in the snapshot context not in the current list. | |
2151 | * And verify there are no changes to snapshots we already know | |
2152 | * about. | |
2153 | * | |
2154 | * Assumes the snapshots in the snapshot context are sorted by | |
2155 | * snapshot id, highest id first. (Snapshots in the rbd_dev's list | |
2156 | * are also maintained in that order.) | |
dfc5606d YS |
2157 | */ |
2158 | static int __rbd_init_snaps_header(struct rbd_device *rbd_dev) | |
2159 | { | |
35938150 AE |
2160 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
2161 | const u32 snap_count = snapc->num_snaps; | |
2162 | char *snap_name = rbd_dev->header.snap_names; | |
2163 | struct list_head *head = &rbd_dev->snaps; | |
2164 | struct list_head *links = head->next; | |
2165 | u32 index = 0; | |
dfc5606d | 2166 | |
35938150 AE |
2167 | while (index < snap_count || links != head) { |
2168 | u64 snap_id; | |
2169 | struct rbd_snap *snap; | |
dfc5606d | 2170 | |
35938150 AE |
2171 | snap_id = index < snap_count ? snapc->snaps[index] |
2172 | : CEPH_NOSNAP; | |
2173 | snap = links != head ? list_entry(links, struct rbd_snap, node) | |
2174 | : NULL; | |
2175 | BUG_ON(snap && snap->id == CEPH_NOSNAP); | |
dfc5606d | 2176 | |
35938150 AE |
2177 | if (snap_id == CEPH_NOSNAP || (snap && snap->id > snap_id)) { |
2178 | struct list_head *next = links->next; | |
dfc5606d | 2179 | |
35938150 | 2180 | /* Existing snapshot not in the new snap context */ |
dfc5606d | 2181 | |
35938150 | 2182 | if (rbd_dev->snap_id == snap->id) |
e88a36ec | 2183 | rbd_dev->snap_exists = false; |
35938150 AE |
2184 | __rbd_remove_snap_dev(snap); |
2185 | ||
2186 | /* Done with this list entry; advance */ | |
2187 | ||
2188 | links = next; | |
dfc5606d YS |
2189 | continue; |
2190 | } | |
35938150 AE |
2191 | |
2192 | if (!snap || (snap_id != CEPH_NOSNAP && snap->id < snap_id)) { | |
2193 | struct rbd_snap *new_snap; | |
2194 | ||
2195 | /* We haven't seen this snapshot before */ | |
2196 | ||
2197 | new_snap = __rbd_add_snap_dev(rbd_dev, index, | |
2198 | snap_name); | |
2199 | if (IS_ERR(new_snap)) | |
2200 | return PTR_ERR(new_snap); | |
2201 | ||
2202 | /* New goes before existing, or at end of list */ | |
2203 | ||
2204 | if (snap) | |
2205 | list_add_tail(&new_snap->node, &snap->node); | |
2206 | else | |
523f3258 | 2207 | list_add_tail(&new_snap->node, head); |
35938150 AE |
2208 | } else { |
2209 | /* Already have this one */ | |
2210 | ||
2211 | BUG_ON(snap->size != rbd_dev->header.snap_sizes[index]); | |
2212 | BUG_ON(strcmp(snap->name, snap_name)); | |
2213 | ||
2214 | /* Done with this list entry; advance */ | |
2215 | ||
2216 | links = links->next; | |
dfc5606d | 2217 | } |
35938150 AE |
2218 | |
2219 | /* Advance to the next entry in the snapshot context */ | |
2220 | ||
2221 | index++; | |
2222 | snap_name += strlen(snap_name) + 1; | |
dfc5606d YS |
2223 | } |
2224 | ||
2225 | return 0; | |
2226 | } | |
2227 | ||
dfc5606d YS |
2228 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
2229 | { | |
f0f8cef5 | 2230 | int ret; |
dfc5606d YS |
2231 | struct device *dev; |
2232 | struct rbd_snap *snap; | |
2233 | ||
2234 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2235 | dev = &rbd_dev->dev; | |
2236 | ||
2237 | dev->bus = &rbd_bus_type; | |
2238 | dev->type = &rbd_device_type; | |
2239 | dev->parent = &rbd_root_dev; | |
2240 | dev->release = rbd_dev_release; | |
de71a297 | 2241 | dev_set_name(dev, "%d", rbd_dev->dev_id); |
dfc5606d YS |
2242 | ret = device_register(dev); |
2243 | if (ret < 0) | |
f0f8cef5 | 2244 | goto out; |
dfc5606d YS |
2245 | |
2246 | list_for_each_entry(snap, &rbd_dev->snaps, node) { | |
14e7085d | 2247 | ret = rbd_register_snap_dev(snap, &rbd_dev->dev); |
dfc5606d | 2248 | if (ret < 0) |
602adf40 YS |
2249 | break; |
2250 | } | |
f0f8cef5 | 2251 | out: |
dfc5606d YS |
2252 | mutex_unlock(&ctl_mutex); |
2253 | return ret; | |
602adf40 YS |
2254 | } |
2255 | ||
dfc5606d YS |
2256 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
2257 | { | |
2258 | device_unregister(&rbd_dev->dev); | |
2259 | } | |
2260 | ||
59c2be1e YS |
2261 | static int rbd_init_watch_dev(struct rbd_device *rbd_dev) |
2262 | { | |
2263 | int ret, rc; | |
2264 | ||
2265 | do { | |
0e6f322d | 2266 | ret = rbd_req_sync_watch(rbd_dev); |
59c2be1e | 2267 | if (ret == -ERANGE) { |
1fe5e993 | 2268 | rc = rbd_refresh_header(rbd_dev, NULL); |
59c2be1e YS |
2269 | if (rc < 0) |
2270 | return rc; | |
2271 | } | |
2272 | } while (ret == -ERANGE); | |
2273 | ||
2274 | return ret; | |
2275 | } | |
2276 | ||
1ddbe94e AE |
2277 | static atomic64_t rbd_id_max = ATOMIC64_INIT(0); |
2278 | ||
2279 | /* | |
499afd5b AE |
2280 | * Get a unique rbd identifier for the given new rbd_dev, and add |
2281 | * the rbd_dev to the global list. The minimum rbd id is 1. | |
1ddbe94e | 2282 | */ |
499afd5b | 2283 | static void rbd_id_get(struct rbd_device *rbd_dev) |
b7f23c36 | 2284 | { |
de71a297 | 2285 | rbd_dev->dev_id = atomic64_inc_return(&rbd_id_max); |
499afd5b AE |
2286 | |
2287 | spin_lock(&rbd_dev_list_lock); | |
2288 | list_add_tail(&rbd_dev->node, &rbd_dev_list); | |
2289 | spin_unlock(&rbd_dev_list_lock); | |
1ddbe94e | 2290 | } |
b7f23c36 | 2291 | |
1ddbe94e | 2292 | /* |
499afd5b AE |
2293 | * Remove an rbd_dev from the global list, and record that its |
2294 | * identifier is no longer in use. | |
1ddbe94e | 2295 | */ |
499afd5b | 2296 | static void rbd_id_put(struct rbd_device *rbd_dev) |
1ddbe94e | 2297 | { |
d184f6bf | 2298 | struct list_head *tmp; |
de71a297 | 2299 | int rbd_id = rbd_dev->dev_id; |
d184f6bf AE |
2300 | int max_id; |
2301 | ||
2302 | BUG_ON(rbd_id < 1); | |
499afd5b AE |
2303 | |
2304 | spin_lock(&rbd_dev_list_lock); | |
2305 | list_del_init(&rbd_dev->node); | |
d184f6bf AE |
2306 | |
2307 | /* | |
2308 | * If the id being "put" is not the current maximum, there | |
2309 | * is nothing special we need to do. | |
2310 | */ | |
2311 | if (rbd_id != atomic64_read(&rbd_id_max)) { | |
2312 | spin_unlock(&rbd_dev_list_lock); | |
2313 | return; | |
2314 | } | |
2315 | ||
2316 | /* | |
2317 | * We need to update the current maximum id. Search the | |
2318 | * list to find out what it is. We're more likely to find | |
2319 | * the maximum at the end, so search the list backward. | |
2320 | */ | |
2321 | max_id = 0; | |
2322 | list_for_each_prev(tmp, &rbd_dev_list) { | |
2323 | struct rbd_device *rbd_dev; | |
2324 | ||
2325 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
2326 | if (rbd_id > max_id) | |
2327 | max_id = rbd_id; | |
2328 | } | |
499afd5b | 2329 | spin_unlock(&rbd_dev_list_lock); |
b7f23c36 | 2330 | |
1ddbe94e | 2331 | /* |
d184f6bf AE |
2332 | * The max id could have been updated by rbd_id_get(), in |
2333 | * which case it now accurately reflects the new maximum. | |
2334 | * Be careful not to overwrite the maximum value in that | |
2335 | * case. | |
1ddbe94e | 2336 | */ |
d184f6bf | 2337 | atomic64_cmpxchg(&rbd_id_max, rbd_id, max_id); |
b7f23c36 AE |
2338 | } |
2339 | ||
e28fff26 AE |
2340 | /* |
2341 | * Skips over white space at *buf, and updates *buf to point to the | |
2342 | * first found non-space character (if any). Returns the length of | |
593a9e7b AE |
2343 | * the token (string of non-white space characters) found. Note |
2344 | * that *buf must be terminated with '\0'. | |
e28fff26 AE |
2345 | */ |
2346 | static inline size_t next_token(const char **buf) | |
2347 | { | |
2348 | /* | |
2349 | * These are the characters that produce nonzero for | |
2350 | * isspace() in the "C" and "POSIX" locales. | |
2351 | */ | |
2352 | const char *spaces = " \f\n\r\t\v"; | |
2353 | ||
2354 | *buf += strspn(*buf, spaces); /* Find start of token */ | |
2355 | ||
2356 | return strcspn(*buf, spaces); /* Return token length */ | |
2357 | } | |
2358 | ||
2359 | /* | |
2360 | * Finds the next token in *buf, and if the provided token buffer is | |
2361 | * big enough, copies the found token into it. The result, if | |
593a9e7b AE |
2362 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
2363 | * must be terminated with '\0' on entry. | |
e28fff26 AE |
2364 | * |
2365 | * Returns the length of the token found (not including the '\0'). | |
2366 | * Return value will be 0 if no token is found, and it will be >= | |
2367 | * token_size if the token would not fit. | |
2368 | * | |
593a9e7b | 2369 | * The *buf pointer will be updated to point beyond the end of the |
e28fff26 AE |
2370 | * found token. Note that this occurs even if the token buffer is |
2371 | * too small to hold it. | |
2372 | */ | |
2373 | static inline size_t copy_token(const char **buf, | |
2374 | char *token, | |
2375 | size_t token_size) | |
2376 | { | |
2377 | size_t len; | |
2378 | ||
2379 | len = next_token(buf); | |
2380 | if (len < token_size) { | |
2381 | memcpy(token, *buf, len); | |
2382 | *(token + len) = '\0'; | |
2383 | } | |
2384 | *buf += len; | |
2385 | ||
2386 | return len; | |
2387 | } | |
2388 | ||
ea3352f4 AE |
2389 | /* |
2390 | * Finds the next token in *buf, dynamically allocates a buffer big | |
2391 | * enough to hold a copy of it, and copies the token into the new | |
2392 | * buffer. The copy is guaranteed to be terminated with '\0'. Note | |
2393 | * that a duplicate buffer is created even for a zero-length token. | |
2394 | * | |
2395 | * Returns a pointer to the newly-allocated duplicate, or a null | |
2396 | * pointer if memory for the duplicate was not available. If | |
2397 | * the lenp argument is a non-null pointer, the length of the token | |
2398 | * (not including the '\0') is returned in *lenp. | |
2399 | * | |
2400 | * If successful, the *buf pointer will be updated to point beyond | |
2401 | * the end of the found token. | |
2402 | * | |
2403 | * Note: uses GFP_KERNEL for allocation. | |
2404 | */ | |
2405 | static inline char *dup_token(const char **buf, size_t *lenp) | |
2406 | { | |
2407 | char *dup; | |
2408 | size_t len; | |
2409 | ||
2410 | len = next_token(buf); | |
2411 | dup = kmalloc(len + 1, GFP_KERNEL); | |
2412 | if (!dup) | |
2413 | return NULL; | |
2414 | ||
2415 | memcpy(dup, *buf, len); | |
2416 | *(dup + len) = '\0'; | |
2417 | *buf += len; | |
2418 | ||
2419 | if (lenp) | |
2420 | *lenp = len; | |
2421 | ||
2422 | return dup; | |
2423 | } | |
2424 | ||
a725f65e | 2425 | /* |
0bed54dc | 2426 | * This fills in the pool_name, image_name, image_name_len, snap_name, |
a725f65e AE |
2427 | * rbd_dev, rbd_md_name, and name fields of the given rbd_dev, based |
2428 | * on the list of monitor addresses and other options provided via | |
2429 | * /sys/bus/rbd/add. | |
d22f76e7 AE |
2430 | * |
2431 | * Note: rbd_dev is assumed to have been initially zero-filled. | |
a725f65e AE |
2432 | */ |
2433 | static int rbd_add_parse_args(struct rbd_device *rbd_dev, | |
2434 | const char *buf, | |
7ef3214a | 2435 | const char **mon_addrs, |
5214ecc4 | 2436 | size_t *mon_addrs_size, |
e28fff26 | 2437 | char *options, |
0bed54dc | 2438 | size_t options_size) |
e28fff26 | 2439 | { |
d22f76e7 AE |
2440 | size_t len; |
2441 | int ret; | |
e28fff26 AE |
2442 | |
2443 | /* The first four tokens are required */ | |
2444 | ||
7ef3214a AE |
2445 | len = next_token(&buf); |
2446 | if (!len) | |
a725f65e | 2447 | return -EINVAL; |
5214ecc4 | 2448 | *mon_addrs_size = len + 1; |
7ef3214a AE |
2449 | *mon_addrs = buf; |
2450 | ||
2451 | buf += len; | |
a725f65e | 2452 | |
e28fff26 AE |
2453 | len = copy_token(&buf, options, options_size); |
2454 | if (!len || len >= options_size) | |
2455 | return -EINVAL; | |
2456 | ||
bf3e5ae1 | 2457 | ret = -ENOMEM; |
d22f76e7 AE |
2458 | rbd_dev->pool_name = dup_token(&buf, NULL); |
2459 | if (!rbd_dev->pool_name) | |
d22f76e7 | 2460 | goto out_err; |
e28fff26 | 2461 | |
0bed54dc AE |
2462 | rbd_dev->image_name = dup_token(&buf, &rbd_dev->image_name_len); |
2463 | if (!rbd_dev->image_name) | |
bf3e5ae1 | 2464 | goto out_err; |
a725f65e | 2465 | |
cb8627c7 AE |
2466 | /* Create the name of the header object */ |
2467 | ||
0bed54dc | 2468 | rbd_dev->header_name = kmalloc(rbd_dev->image_name_len |
bf3e5ae1 AE |
2469 | + sizeof (RBD_SUFFIX), |
2470 | GFP_KERNEL); | |
0bed54dc | 2471 | if (!rbd_dev->header_name) |
cb8627c7 | 2472 | goto out_err; |
0bed54dc | 2473 | sprintf(rbd_dev->header_name, "%s%s", rbd_dev->image_name, RBD_SUFFIX); |
a725f65e | 2474 | |
e28fff26 | 2475 | /* |
820a5f3e AE |
2476 | * The snapshot name is optional. If none is is supplied, |
2477 | * we use the default value. | |
e28fff26 | 2478 | */ |
820a5f3e AE |
2479 | rbd_dev->snap_name = dup_token(&buf, &len); |
2480 | if (!rbd_dev->snap_name) | |
2481 | goto out_err; | |
2482 | if (!len) { | |
2483 | /* Replace the empty name with the default */ | |
2484 | kfree(rbd_dev->snap_name); | |
2485 | rbd_dev->snap_name | |
2486 | = kmalloc(sizeof (RBD_SNAP_HEAD_NAME), GFP_KERNEL); | |
2487 | if (!rbd_dev->snap_name) | |
2488 | goto out_err; | |
2489 | ||
e28fff26 AE |
2490 | memcpy(rbd_dev->snap_name, RBD_SNAP_HEAD_NAME, |
2491 | sizeof (RBD_SNAP_HEAD_NAME)); | |
849b4260 | 2492 | } |
e28fff26 | 2493 | |
a725f65e | 2494 | return 0; |
d22f76e7 AE |
2495 | |
2496 | out_err: | |
0bed54dc | 2497 | kfree(rbd_dev->header_name); |
d78fd7ae | 2498 | rbd_dev->header_name = NULL; |
0bed54dc | 2499 | kfree(rbd_dev->image_name); |
d78fd7ae AE |
2500 | rbd_dev->image_name = NULL; |
2501 | rbd_dev->image_name_len = 0; | |
d22f76e7 AE |
2502 | kfree(rbd_dev->pool_name); |
2503 | rbd_dev->pool_name = NULL; | |
2504 | ||
2505 | return ret; | |
a725f65e AE |
2506 | } |
2507 | ||
59c2be1e YS |
2508 | static ssize_t rbd_add(struct bus_type *bus, |
2509 | const char *buf, | |
2510 | size_t count) | |
602adf40 | 2511 | { |
cb8627c7 AE |
2512 | char *options; |
2513 | struct rbd_device *rbd_dev = NULL; | |
7ef3214a AE |
2514 | const char *mon_addrs = NULL; |
2515 | size_t mon_addrs_size = 0; | |
27cc2594 AE |
2516 | struct ceph_osd_client *osdc; |
2517 | int rc = -ENOMEM; | |
602adf40 YS |
2518 | |
2519 | if (!try_module_get(THIS_MODULE)) | |
2520 | return -ENODEV; | |
2521 | ||
60571c7d | 2522 | options = kmalloc(count, GFP_KERNEL); |
602adf40 | 2523 | if (!options) |
27cc2594 | 2524 | goto err_nomem; |
cb8627c7 AE |
2525 | rbd_dev = kzalloc(sizeof(*rbd_dev), GFP_KERNEL); |
2526 | if (!rbd_dev) | |
2527 | goto err_nomem; | |
602adf40 YS |
2528 | |
2529 | /* static rbd_device initialization */ | |
2530 | spin_lock_init(&rbd_dev->lock); | |
2531 | INIT_LIST_HEAD(&rbd_dev->node); | |
dfc5606d | 2532 | INIT_LIST_HEAD(&rbd_dev->snaps); |
c666601a | 2533 | init_rwsem(&rbd_dev->header_rwsem); |
602adf40 | 2534 | |
d184f6bf | 2535 | /* generate unique id: find highest unique id, add one */ |
499afd5b | 2536 | rbd_id_get(rbd_dev); |
602adf40 | 2537 | |
a725f65e | 2538 | /* Fill in the device name, now that we have its id. */ |
81a89793 AE |
2539 | BUILD_BUG_ON(DEV_NAME_LEN |
2540 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); | |
de71a297 | 2541 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); |
a725f65e | 2542 | |
602adf40 | 2543 | /* parse add command */ |
7ef3214a | 2544 | rc = rbd_add_parse_args(rbd_dev, buf, &mon_addrs, &mon_addrs_size, |
e28fff26 | 2545 | options, count); |
a725f65e | 2546 | if (rc) |
f0f8cef5 | 2547 | goto err_put_id; |
e124a82f | 2548 | |
f8c38929 AE |
2549 | rc = rbd_get_client(rbd_dev, mon_addrs, mon_addrs_size - 1, options); |
2550 | if (rc < 0) | |
f0f8cef5 | 2551 | goto err_put_id; |
602adf40 | 2552 | |
602adf40 | 2553 | /* pick the pool */ |
1dbb4399 | 2554 | osdc = &rbd_dev->rbd_client->client->osdc; |
602adf40 YS |
2555 | rc = ceph_pg_poolid_by_name(osdc->osdmap, rbd_dev->pool_name); |
2556 | if (rc < 0) | |
2557 | goto err_out_client; | |
9bb2f334 | 2558 | rbd_dev->pool_id = rc; |
602adf40 YS |
2559 | |
2560 | /* register our block device */ | |
27cc2594 AE |
2561 | rc = register_blkdev(0, rbd_dev->name); |
2562 | if (rc < 0) | |
602adf40 | 2563 | goto err_out_client; |
27cc2594 | 2564 | rbd_dev->major = rc; |
602adf40 | 2565 | |
dfc5606d YS |
2566 | rc = rbd_bus_add_dev(rbd_dev); |
2567 | if (rc) | |
766fc439 YS |
2568 | goto err_out_blkdev; |
2569 | ||
32eec68d AE |
2570 | /* |
2571 | * At this point cleanup in the event of an error is the job | |
2572 | * of the sysfs code (initiated by rbd_bus_del_dev()). | |
2573 | * | |
2574 | * Set up and announce blkdev mapping. | |
2575 | */ | |
602adf40 YS |
2576 | rc = rbd_init_disk(rbd_dev); |
2577 | if (rc) | |
766fc439 | 2578 | goto err_out_bus; |
602adf40 | 2579 | |
59c2be1e YS |
2580 | rc = rbd_init_watch_dev(rbd_dev); |
2581 | if (rc) | |
2582 | goto err_out_bus; | |
2583 | ||
602adf40 YS |
2584 | return count; |
2585 | ||
766fc439 | 2586 | err_out_bus: |
766fc439 YS |
2587 | /* this will also clean up rest of rbd_dev stuff */ |
2588 | ||
2589 | rbd_bus_del_dev(rbd_dev); | |
2590 | kfree(options); | |
766fc439 YS |
2591 | return rc; |
2592 | ||
602adf40 YS |
2593 | err_out_blkdev: |
2594 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
2595 | err_out_client: | |
2596 | rbd_put_client(rbd_dev); | |
f0f8cef5 | 2597 | err_put_id: |
cb8627c7 | 2598 | if (rbd_dev->pool_name) { |
820a5f3e | 2599 | kfree(rbd_dev->snap_name); |
0bed54dc AE |
2600 | kfree(rbd_dev->header_name); |
2601 | kfree(rbd_dev->image_name); | |
cb8627c7 AE |
2602 | kfree(rbd_dev->pool_name); |
2603 | } | |
499afd5b | 2604 | rbd_id_put(rbd_dev); |
27cc2594 | 2605 | err_nomem: |
27cc2594 | 2606 | kfree(rbd_dev); |
cb8627c7 | 2607 | kfree(options); |
27cc2594 | 2608 | |
602adf40 YS |
2609 | dout("Error adding device %s\n", buf); |
2610 | module_put(THIS_MODULE); | |
27cc2594 AE |
2611 | |
2612 | return (ssize_t) rc; | |
602adf40 YS |
2613 | } |
2614 | ||
de71a297 | 2615 | static struct rbd_device *__rbd_get_dev(unsigned long dev_id) |
602adf40 YS |
2616 | { |
2617 | struct list_head *tmp; | |
2618 | struct rbd_device *rbd_dev; | |
2619 | ||
e124a82f | 2620 | spin_lock(&rbd_dev_list_lock); |
602adf40 YS |
2621 | list_for_each(tmp, &rbd_dev_list) { |
2622 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
de71a297 | 2623 | if (rbd_dev->dev_id == dev_id) { |
e124a82f | 2624 | spin_unlock(&rbd_dev_list_lock); |
602adf40 | 2625 | return rbd_dev; |
e124a82f | 2626 | } |
602adf40 | 2627 | } |
e124a82f | 2628 | spin_unlock(&rbd_dev_list_lock); |
602adf40 YS |
2629 | return NULL; |
2630 | } | |
2631 | ||
dfc5606d | 2632 | static void rbd_dev_release(struct device *dev) |
602adf40 | 2633 | { |
593a9e7b | 2634 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 2635 | |
1dbb4399 AE |
2636 | if (rbd_dev->watch_request) { |
2637 | struct ceph_client *client = rbd_dev->rbd_client->client; | |
2638 | ||
2639 | ceph_osdc_unregister_linger_request(&client->osdc, | |
59c2be1e | 2640 | rbd_dev->watch_request); |
1dbb4399 | 2641 | } |
59c2be1e | 2642 | if (rbd_dev->watch_event) |
070c633f | 2643 | rbd_req_sync_unwatch(rbd_dev); |
59c2be1e | 2644 | |
602adf40 YS |
2645 | rbd_put_client(rbd_dev); |
2646 | ||
2647 | /* clean up and free blkdev */ | |
2648 | rbd_free_disk(rbd_dev); | |
2649 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
32eec68d AE |
2650 | |
2651 | /* done with the id, and with the rbd_dev */ | |
820a5f3e | 2652 | kfree(rbd_dev->snap_name); |
0bed54dc | 2653 | kfree(rbd_dev->header_name); |
d22f76e7 | 2654 | kfree(rbd_dev->pool_name); |
0bed54dc | 2655 | kfree(rbd_dev->image_name); |
32eec68d | 2656 | rbd_id_put(rbd_dev); |
602adf40 YS |
2657 | kfree(rbd_dev); |
2658 | ||
2659 | /* release module ref */ | |
2660 | module_put(THIS_MODULE); | |
602adf40 YS |
2661 | } |
2662 | ||
dfc5606d YS |
2663 | static ssize_t rbd_remove(struct bus_type *bus, |
2664 | const char *buf, | |
2665 | size_t count) | |
602adf40 YS |
2666 | { |
2667 | struct rbd_device *rbd_dev = NULL; | |
2668 | int target_id, rc; | |
2669 | unsigned long ul; | |
2670 | int ret = count; | |
2671 | ||
2672 | rc = strict_strtoul(buf, 10, &ul); | |
2673 | if (rc) | |
2674 | return rc; | |
2675 | ||
2676 | /* convert to int; abort if we lost anything in the conversion */ | |
2677 | target_id = (int) ul; | |
2678 | if (target_id != ul) | |
2679 | return -EINVAL; | |
2680 | ||
2681 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2682 | ||
2683 | rbd_dev = __rbd_get_dev(target_id); | |
2684 | if (!rbd_dev) { | |
2685 | ret = -ENOENT; | |
2686 | goto done; | |
2687 | } | |
2688 | ||
dfc5606d YS |
2689 | __rbd_remove_all_snaps(rbd_dev); |
2690 | rbd_bus_del_dev(rbd_dev); | |
602adf40 YS |
2691 | |
2692 | done: | |
2693 | mutex_unlock(&ctl_mutex); | |
2694 | return ret; | |
2695 | } | |
2696 | ||
dfc5606d YS |
2697 | static ssize_t rbd_snap_add(struct device *dev, |
2698 | struct device_attribute *attr, | |
2699 | const char *buf, | |
2700 | size_t count) | |
602adf40 | 2701 | { |
593a9e7b | 2702 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d YS |
2703 | int ret; |
2704 | char *name = kmalloc(count + 1, GFP_KERNEL); | |
602adf40 YS |
2705 | if (!name) |
2706 | return -ENOMEM; | |
2707 | ||
dfc5606d | 2708 | snprintf(name, count, "%s", buf); |
602adf40 YS |
2709 | |
2710 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
2711 | ||
602adf40 YS |
2712 | ret = rbd_header_add_snap(rbd_dev, |
2713 | name, GFP_KERNEL); | |
2714 | if (ret < 0) | |
59c2be1e | 2715 | goto err_unlock; |
602adf40 | 2716 | |
b813623a | 2717 | ret = __rbd_refresh_header(rbd_dev, NULL); |
602adf40 | 2718 | if (ret < 0) |
59c2be1e YS |
2719 | goto err_unlock; |
2720 | ||
2721 | /* shouldn't hold ctl_mutex when notifying.. notify might | |
2722 | trigger a watch callback that would need to get that mutex */ | |
2723 | mutex_unlock(&ctl_mutex); | |
2724 | ||
2725 | /* make a best effort, don't error if failed */ | |
4cb16250 | 2726 | rbd_req_sync_notify(rbd_dev); |
602adf40 YS |
2727 | |
2728 | ret = count; | |
59c2be1e YS |
2729 | kfree(name); |
2730 | return ret; | |
2731 | ||
2732 | err_unlock: | |
602adf40 | 2733 | mutex_unlock(&ctl_mutex); |
602adf40 YS |
2734 | kfree(name); |
2735 | return ret; | |
2736 | } | |
2737 | ||
602adf40 YS |
2738 | /* |
2739 | * create control files in sysfs | |
dfc5606d | 2740 | * /sys/bus/rbd/... |
602adf40 YS |
2741 | */ |
2742 | static int rbd_sysfs_init(void) | |
2743 | { | |
dfc5606d | 2744 | int ret; |
602adf40 | 2745 | |
fed4c143 | 2746 | ret = device_register(&rbd_root_dev); |
21079786 | 2747 | if (ret < 0) |
dfc5606d | 2748 | return ret; |
602adf40 | 2749 | |
fed4c143 AE |
2750 | ret = bus_register(&rbd_bus_type); |
2751 | if (ret < 0) | |
2752 | device_unregister(&rbd_root_dev); | |
602adf40 | 2753 | |
602adf40 YS |
2754 | return ret; |
2755 | } | |
2756 | ||
2757 | static void rbd_sysfs_cleanup(void) | |
2758 | { | |
dfc5606d | 2759 | bus_unregister(&rbd_bus_type); |
fed4c143 | 2760 | device_unregister(&rbd_root_dev); |
602adf40 YS |
2761 | } |
2762 | ||
2763 | int __init rbd_init(void) | |
2764 | { | |
2765 | int rc; | |
2766 | ||
2767 | rc = rbd_sysfs_init(); | |
2768 | if (rc) | |
2769 | return rc; | |
f0f8cef5 | 2770 | pr_info("loaded " RBD_DRV_NAME_LONG "\n"); |
602adf40 YS |
2771 | return 0; |
2772 | } | |
2773 | ||
2774 | void __exit rbd_exit(void) | |
2775 | { | |
2776 | rbd_sysfs_cleanup(); | |
2777 | } | |
2778 | ||
2779 | module_init(rbd_init); | |
2780 | module_exit(rbd_exit); | |
2781 | ||
2782 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
2783 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
2784 | MODULE_DESCRIPTION("rados block device"); | |
2785 | ||
2786 | /* following authorship retained from original osdblk.c */ | |
2787 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); | |
2788 | ||
2789 | MODULE_LICENSE("GPL"); |