]>
Commit | Line | Data |
---|---|---|
602adf40 YS |
1 | /* |
2 | rbd.c -- Export ceph rados objects as a Linux block device | |
3 | ||
4 | ||
5 | based on drivers/block/osdblk.c: | |
6 | ||
7 | Copyright 2009 Red Hat, Inc. | |
8 | ||
9 | This program is free software; you can redistribute it and/or modify | |
10 | it under the terms of the GNU General Public License as published by | |
11 | the Free Software Foundation. | |
12 | ||
13 | This program is distributed in the hope that it will be useful, | |
14 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | GNU General Public License for more details. | |
17 | ||
18 | You should have received a copy of the GNU General Public License | |
19 | along with this program; see the file COPYING. If not, write to | |
20 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
21 | ||
22 | ||
23 | ||
dfc5606d | 24 | For usage instructions, please refer to: |
602adf40 | 25 | |
dfc5606d | 26 | Documentation/ABI/testing/sysfs-bus-rbd |
602adf40 YS |
27 | |
28 | */ | |
29 | ||
30 | #include <linux/ceph/libceph.h> | |
31 | #include <linux/ceph/osd_client.h> | |
32 | #include <linux/ceph/mon_client.h> | |
33 | #include <linux/ceph/decode.h> | |
59c2be1e | 34 | #include <linux/parser.h> |
602adf40 YS |
35 | |
36 | #include <linux/kernel.h> | |
37 | #include <linux/device.h> | |
38 | #include <linux/module.h> | |
39 | #include <linux/fs.h> | |
40 | #include <linux/blkdev.h> | |
41 | ||
42 | #include "rbd_types.h" | |
43 | ||
aafb230e AE |
44 | #define RBD_DEBUG /* Activate rbd_assert() calls */ |
45 | ||
593a9e7b AE |
46 | /* |
47 | * The basic unit of block I/O is a sector. It is interpreted in a | |
48 | * number of contexts in Linux (blk, bio, genhd), but the default is | |
49 | * universally 512 bytes. These symbols are just slightly more | |
50 | * meaningful than the bare numbers they represent. | |
51 | */ | |
52 | #define SECTOR_SHIFT 9 | |
53 | #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) | |
54 | ||
f0f8cef5 AE |
55 | #define RBD_DRV_NAME "rbd" |
56 | #define RBD_DRV_NAME_LONG "rbd (rados block device)" | |
602adf40 YS |
57 | |
58 | #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ | |
59 | ||
d4b125e9 AE |
60 | #define RBD_SNAP_DEV_NAME_PREFIX "snap_" |
61 | #define RBD_MAX_SNAP_NAME_LEN \ | |
62 | (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1)) | |
63 | ||
35d489f9 | 64 | #define RBD_MAX_SNAP_COUNT 510 /* allows max snapc to fit in 4KB */ |
602adf40 YS |
65 | |
66 | #define RBD_SNAP_HEAD_NAME "-" | |
67 | ||
9e15b77d AE |
68 | /* This allows a single page to hold an image name sent by OSD */ |
69 | #define RBD_IMAGE_NAME_LEN_MAX (PAGE_SIZE - sizeof (__le32) - 1) | |
1e130199 | 70 | #define RBD_IMAGE_ID_LEN_MAX 64 |
9e15b77d | 71 | |
1e130199 | 72 | #define RBD_OBJ_PREFIX_LEN_MAX 64 |
589d30e0 | 73 | |
d889140c AE |
74 | /* Feature bits */ |
75 | ||
5cbf6f12 AE |
76 | #define RBD_FEATURE_LAYERING (1<<0) |
77 | #define RBD_FEATURE_STRIPINGV2 (1<<1) | |
78 | #define RBD_FEATURES_ALL \ | |
79 | (RBD_FEATURE_LAYERING | RBD_FEATURE_STRIPINGV2) | |
d889140c AE |
80 | |
81 | /* Features supported by this (client software) implementation. */ | |
82 | ||
5cbf6f12 | 83 | #define RBD_FEATURES_SUPPORTED (0) |
d889140c | 84 | |
81a89793 AE |
85 | /* |
86 | * An RBD device name will be "rbd#", where the "rbd" comes from | |
87 | * RBD_DRV_NAME above, and # is a unique integer identifier. | |
88 | * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big | |
89 | * enough to hold all possible device names. | |
90 | */ | |
602adf40 | 91 | #define DEV_NAME_LEN 32 |
81a89793 | 92 | #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) |
602adf40 YS |
93 | |
94 | /* | |
95 | * block device image metadata (in-memory version) | |
96 | */ | |
97 | struct rbd_image_header { | |
f84344f3 | 98 | /* These four fields never change for a given rbd image */ |
849b4260 | 99 | char *object_prefix; |
34b13184 | 100 | u64 features; |
602adf40 YS |
101 | __u8 obj_order; |
102 | __u8 crypt_type; | |
103 | __u8 comp_type; | |
602adf40 | 104 | |
f84344f3 AE |
105 | /* The remaining fields need to be updated occasionally */ |
106 | u64 image_size; | |
107 | struct ceph_snap_context *snapc; | |
602adf40 YS |
108 | char *snap_names; |
109 | u64 *snap_sizes; | |
59c2be1e YS |
110 | |
111 | u64 obj_version; | |
112 | }; | |
113 | ||
0d7dbfce AE |
114 | /* |
115 | * An rbd image specification. | |
116 | * | |
117 | * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely | |
c66c6e0c AE |
118 | * identify an image. Each rbd_dev structure includes a pointer to |
119 | * an rbd_spec structure that encapsulates this identity. | |
120 | * | |
121 | * Each of the id's in an rbd_spec has an associated name. For a | |
122 | * user-mapped image, the names are supplied and the id's associated | |
123 | * with them are looked up. For a layered image, a parent image is | |
124 | * defined by the tuple, and the names are looked up. | |
125 | * | |
126 | * An rbd_dev structure contains a parent_spec pointer which is | |
127 | * non-null if the image it represents is a child in a layered | |
128 | * image. This pointer will refer to the rbd_spec structure used | |
129 | * by the parent rbd_dev for its own identity (i.e., the structure | |
130 | * is shared between the parent and child). | |
131 | * | |
132 | * Since these structures are populated once, during the discovery | |
133 | * phase of image construction, they are effectively immutable so | |
134 | * we make no effort to synchronize access to them. | |
135 | * | |
136 | * Note that code herein does not assume the image name is known (it | |
137 | * could be a null pointer). | |
0d7dbfce AE |
138 | */ |
139 | struct rbd_spec { | |
140 | u64 pool_id; | |
141 | char *pool_name; | |
142 | ||
143 | char *image_id; | |
0d7dbfce | 144 | char *image_name; |
0d7dbfce AE |
145 | |
146 | u64 snap_id; | |
147 | char *snap_name; | |
148 | ||
149 | struct kref kref; | |
150 | }; | |
151 | ||
602adf40 | 152 | /* |
f0f8cef5 | 153 | * an instance of the client. multiple devices may share an rbd client. |
602adf40 YS |
154 | */ |
155 | struct rbd_client { | |
156 | struct ceph_client *client; | |
157 | struct kref kref; | |
158 | struct list_head node; | |
159 | }; | |
160 | ||
bf0d5f50 AE |
161 | struct rbd_img_request; |
162 | typedef void (*rbd_img_callback_t)(struct rbd_img_request *); | |
163 | ||
164 | #define BAD_WHICH U32_MAX /* Good which or bad which, which? */ | |
165 | ||
166 | struct rbd_obj_request; | |
167 | typedef void (*rbd_obj_callback_t)(struct rbd_obj_request *); | |
168 | ||
9969ebc5 AE |
169 | enum obj_request_type { |
170 | OBJ_REQUEST_NODATA, OBJ_REQUEST_BIO, OBJ_REQUEST_PAGES | |
171 | }; | |
bf0d5f50 | 172 | |
926f9b3f AE |
173 | enum obj_req_flags { |
174 | OBJ_REQ_DONE, /* completion flag: not done = 0, done = 1 */ | |
6365d33a | 175 | OBJ_REQ_IMG_DATA, /* object usage: standalone = 0, image = 1 */ |
5679c59f AE |
176 | OBJ_REQ_KNOWN, /* EXISTS flag valid: no = 0, yes = 1 */ |
177 | OBJ_REQ_EXISTS, /* target exists: no = 0, yes = 1 */ | |
926f9b3f AE |
178 | }; |
179 | ||
bf0d5f50 AE |
180 | struct rbd_obj_request { |
181 | const char *object_name; | |
182 | u64 offset; /* object start byte */ | |
183 | u64 length; /* bytes from offset */ | |
926f9b3f | 184 | unsigned long flags; |
bf0d5f50 AE |
185 | |
186 | struct rbd_img_request *img_request; | |
7da22d29 | 187 | u64 img_offset; /* image relative offset */ |
bf0d5f50 AE |
188 | struct list_head links; /* img_request->obj_requests */ |
189 | u32 which; /* posn image request list */ | |
190 | ||
191 | enum obj_request_type type; | |
788e2df3 AE |
192 | union { |
193 | struct bio *bio_list; | |
194 | struct { | |
195 | struct page **pages; | |
196 | u32 page_count; | |
197 | }; | |
198 | }; | |
bf0d5f50 AE |
199 | |
200 | struct ceph_osd_request *osd_req; | |
201 | ||
202 | u64 xferred; /* bytes transferred */ | |
203 | u64 version; | |
1b83bef2 | 204 | int result; |
bf0d5f50 AE |
205 | |
206 | rbd_obj_callback_t callback; | |
788e2df3 | 207 | struct completion completion; |
bf0d5f50 AE |
208 | |
209 | struct kref kref; | |
210 | }; | |
211 | ||
0c425248 | 212 | enum img_req_flags { |
9849e986 AE |
213 | IMG_REQ_WRITE, /* I/O direction: read = 0, write = 1 */ |
214 | IMG_REQ_CHILD, /* initiator: block = 0, child image = 1 */ | |
d0b2e944 | 215 | IMG_REQ_LAYERED, /* ENOENT handling: normal = 0, layered = 1 */ |
0c425248 AE |
216 | }; |
217 | ||
bf0d5f50 | 218 | struct rbd_img_request { |
bf0d5f50 AE |
219 | struct rbd_device *rbd_dev; |
220 | u64 offset; /* starting image byte offset */ | |
221 | u64 length; /* byte count from offset */ | |
0c425248 | 222 | unsigned long flags; |
bf0d5f50 | 223 | union { |
9849e986 | 224 | u64 snap_id; /* for reads */ |
bf0d5f50 | 225 | struct ceph_snap_context *snapc; /* for writes */ |
9849e986 AE |
226 | }; |
227 | union { | |
228 | struct request *rq; /* block request */ | |
229 | struct rbd_obj_request *obj_request; /* obj req initiator */ | |
bf0d5f50 AE |
230 | }; |
231 | spinlock_t completion_lock;/* protects next_completion */ | |
232 | u32 next_completion; | |
233 | rbd_img_callback_t callback; | |
55f27e09 | 234 | u64 xferred;/* aggregate bytes transferred */ |
a5a337d4 | 235 | int result; /* first nonzero obj_request result */ |
bf0d5f50 AE |
236 | |
237 | u32 obj_request_count; | |
238 | struct list_head obj_requests; /* rbd_obj_request structs */ | |
239 | ||
240 | struct kref kref; | |
241 | }; | |
242 | ||
243 | #define for_each_obj_request(ireq, oreq) \ | |
ef06f4d3 | 244 | list_for_each_entry(oreq, &(ireq)->obj_requests, links) |
bf0d5f50 | 245 | #define for_each_obj_request_from(ireq, oreq) \ |
ef06f4d3 | 246 | list_for_each_entry_from(oreq, &(ireq)->obj_requests, links) |
bf0d5f50 | 247 | #define for_each_obj_request_safe(ireq, oreq, n) \ |
ef06f4d3 | 248 | list_for_each_entry_safe_reverse(oreq, n, &(ireq)->obj_requests, links) |
bf0d5f50 | 249 | |
dfc5606d YS |
250 | struct rbd_snap { |
251 | struct device dev; | |
252 | const char *name; | |
3591538f | 253 | u64 size; |
dfc5606d YS |
254 | struct list_head node; |
255 | u64 id; | |
34b13184 | 256 | u64 features; |
dfc5606d YS |
257 | }; |
258 | ||
f84344f3 | 259 | struct rbd_mapping { |
99c1f08f | 260 | u64 size; |
34b13184 | 261 | u64 features; |
f84344f3 AE |
262 | bool read_only; |
263 | }; | |
264 | ||
602adf40 YS |
265 | /* |
266 | * a single device | |
267 | */ | |
268 | struct rbd_device { | |
de71a297 | 269 | int dev_id; /* blkdev unique id */ |
602adf40 YS |
270 | |
271 | int major; /* blkdev assigned major */ | |
272 | struct gendisk *disk; /* blkdev's gendisk and rq */ | |
602adf40 | 273 | |
a30b71b9 | 274 | u32 image_format; /* Either 1 or 2 */ |
602adf40 YS |
275 | struct rbd_client *rbd_client; |
276 | ||
277 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ | |
278 | ||
b82d167b | 279 | spinlock_t lock; /* queue, flags, open_count */ |
602adf40 YS |
280 | |
281 | struct rbd_image_header header; | |
b82d167b | 282 | unsigned long flags; /* possibly lock protected */ |
0d7dbfce | 283 | struct rbd_spec *spec; |
602adf40 | 284 | |
0d7dbfce | 285 | char *header_name; |
971f839a | 286 | |
0903e875 AE |
287 | struct ceph_file_layout layout; |
288 | ||
59c2be1e | 289 | struct ceph_osd_event *watch_event; |
975241af | 290 | struct rbd_obj_request *watch_request; |
59c2be1e | 291 | |
86b00e0d AE |
292 | struct rbd_spec *parent_spec; |
293 | u64 parent_overlap; | |
2f82ee54 | 294 | struct rbd_device *parent; |
86b00e0d | 295 | |
c666601a JD |
296 | /* protects updating the header */ |
297 | struct rw_semaphore header_rwsem; | |
f84344f3 AE |
298 | |
299 | struct rbd_mapping mapping; | |
602adf40 YS |
300 | |
301 | struct list_head node; | |
dfc5606d YS |
302 | |
303 | /* list of snapshots */ | |
304 | struct list_head snaps; | |
305 | ||
306 | /* sysfs related */ | |
307 | struct device dev; | |
b82d167b | 308 | unsigned long open_count; /* protected by lock */ |
dfc5606d YS |
309 | }; |
310 | ||
b82d167b AE |
311 | /* |
312 | * Flag bits for rbd_dev->flags. If atomicity is required, | |
313 | * rbd_dev->lock is used to protect access. | |
314 | * | |
315 | * Currently, only the "removing" flag (which is coupled with the | |
316 | * "open_count" field) requires atomic access. | |
317 | */ | |
6d292906 AE |
318 | enum rbd_dev_flags { |
319 | RBD_DEV_FLAG_EXISTS, /* mapped snapshot has not been deleted */ | |
b82d167b | 320 | RBD_DEV_FLAG_REMOVING, /* this mapping is being removed */ |
6d292906 AE |
321 | }; |
322 | ||
602adf40 | 323 | static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ |
e124a82f | 324 | |
602adf40 | 325 | static LIST_HEAD(rbd_dev_list); /* devices */ |
e124a82f AE |
326 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
327 | ||
432b8587 AE |
328 | static LIST_HEAD(rbd_client_list); /* clients */ |
329 | static DEFINE_SPINLOCK(rbd_client_list_lock); | |
602adf40 | 330 | |
304f6808 AE |
331 | static int rbd_dev_snaps_update(struct rbd_device *rbd_dev); |
332 | static int rbd_dev_snaps_register(struct rbd_device *rbd_dev); | |
333 | ||
dfc5606d | 334 | static void rbd_dev_release(struct device *dev); |
41f38c2b | 335 | static void rbd_remove_snap_dev(struct rbd_snap *snap); |
dfc5606d | 336 | |
f0f8cef5 AE |
337 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
338 | size_t count); | |
339 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, | |
340 | size_t count); | |
2f82ee54 | 341 | static int rbd_dev_probe(struct rbd_device *rbd_dev); |
f0f8cef5 AE |
342 | |
343 | static struct bus_attribute rbd_bus_attrs[] = { | |
344 | __ATTR(add, S_IWUSR, NULL, rbd_add), | |
345 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), | |
346 | __ATTR_NULL | |
347 | }; | |
348 | ||
349 | static struct bus_type rbd_bus_type = { | |
350 | .name = "rbd", | |
351 | .bus_attrs = rbd_bus_attrs, | |
352 | }; | |
353 | ||
354 | static void rbd_root_dev_release(struct device *dev) | |
355 | { | |
356 | } | |
357 | ||
358 | static struct device rbd_root_dev = { | |
359 | .init_name = "rbd", | |
360 | .release = rbd_root_dev_release, | |
361 | }; | |
362 | ||
06ecc6cb AE |
363 | static __printf(2, 3) |
364 | void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...) | |
365 | { | |
366 | struct va_format vaf; | |
367 | va_list args; | |
368 | ||
369 | va_start(args, fmt); | |
370 | vaf.fmt = fmt; | |
371 | vaf.va = &args; | |
372 | ||
373 | if (!rbd_dev) | |
374 | printk(KERN_WARNING "%s: %pV\n", RBD_DRV_NAME, &vaf); | |
375 | else if (rbd_dev->disk) | |
376 | printk(KERN_WARNING "%s: %s: %pV\n", | |
377 | RBD_DRV_NAME, rbd_dev->disk->disk_name, &vaf); | |
378 | else if (rbd_dev->spec && rbd_dev->spec->image_name) | |
379 | printk(KERN_WARNING "%s: image %s: %pV\n", | |
380 | RBD_DRV_NAME, rbd_dev->spec->image_name, &vaf); | |
381 | else if (rbd_dev->spec && rbd_dev->spec->image_id) | |
382 | printk(KERN_WARNING "%s: id %s: %pV\n", | |
383 | RBD_DRV_NAME, rbd_dev->spec->image_id, &vaf); | |
384 | else /* punt */ | |
385 | printk(KERN_WARNING "%s: rbd_dev %p: %pV\n", | |
386 | RBD_DRV_NAME, rbd_dev, &vaf); | |
387 | va_end(args); | |
388 | } | |
389 | ||
aafb230e AE |
390 | #ifdef RBD_DEBUG |
391 | #define rbd_assert(expr) \ | |
392 | if (unlikely(!(expr))) { \ | |
393 | printk(KERN_ERR "\nAssertion failure in %s() " \ | |
394 | "at line %d:\n\n" \ | |
395 | "\trbd_assert(%s);\n\n", \ | |
396 | __func__, __LINE__, #expr); \ | |
397 | BUG(); \ | |
398 | } | |
399 | #else /* !RBD_DEBUG */ | |
400 | # define rbd_assert(expr) ((void) 0) | |
401 | #endif /* !RBD_DEBUG */ | |
dfc5606d | 402 | |
8b3e1a56 AE |
403 | static void rbd_img_parent_read(struct rbd_obj_request *obj_request); |
404 | ||
117973fb AE |
405 | static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver); |
406 | static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver); | |
59c2be1e | 407 | |
602adf40 YS |
408 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
409 | { | |
f0f8cef5 | 410 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
b82d167b | 411 | bool removing = false; |
602adf40 | 412 | |
f84344f3 | 413 | if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only) |
602adf40 YS |
414 | return -EROFS; |
415 | ||
a14ea269 | 416 | spin_lock_irq(&rbd_dev->lock); |
b82d167b AE |
417 | if (test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags)) |
418 | removing = true; | |
419 | else | |
420 | rbd_dev->open_count++; | |
a14ea269 | 421 | spin_unlock_irq(&rbd_dev->lock); |
b82d167b AE |
422 | if (removing) |
423 | return -ENOENT; | |
424 | ||
42382b70 | 425 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
c3e946ce | 426 | (void) get_device(&rbd_dev->dev); |
f84344f3 | 427 | set_device_ro(bdev, rbd_dev->mapping.read_only); |
42382b70 | 428 | mutex_unlock(&ctl_mutex); |
340c7a2b | 429 | |
602adf40 YS |
430 | return 0; |
431 | } | |
432 | ||
dfc5606d YS |
433 | static int rbd_release(struct gendisk *disk, fmode_t mode) |
434 | { | |
435 | struct rbd_device *rbd_dev = disk->private_data; | |
b82d167b AE |
436 | unsigned long open_count_before; |
437 | ||
a14ea269 | 438 | spin_lock_irq(&rbd_dev->lock); |
b82d167b | 439 | open_count_before = rbd_dev->open_count--; |
a14ea269 | 440 | spin_unlock_irq(&rbd_dev->lock); |
b82d167b | 441 | rbd_assert(open_count_before > 0); |
dfc5606d | 442 | |
42382b70 | 443 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
c3e946ce | 444 | put_device(&rbd_dev->dev); |
42382b70 | 445 | mutex_unlock(&ctl_mutex); |
dfc5606d YS |
446 | |
447 | return 0; | |
448 | } | |
449 | ||
602adf40 YS |
450 | static const struct block_device_operations rbd_bd_ops = { |
451 | .owner = THIS_MODULE, | |
452 | .open = rbd_open, | |
dfc5606d | 453 | .release = rbd_release, |
602adf40 YS |
454 | }; |
455 | ||
456 | /* | |
457 | * Initialize an rbd client instance. | |
43ae4701 | 458 | * We own *ceph_opts. |
602adf40 | 459 | */ |
f8c38929 | 460 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) |
602adf40 YS |
461 | { |
462 | struct rbd_client *rbdc; | |
463 | int ret = -ENOMEM; | |
464 | ||
37206ee5 | 465 | dout("%s:\n", __func__); |
602adf40 YS |
466 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); |
467 | if (!rbdc) | |
468 | goto out_opt; | |
469 | ||
470 | kref_init(&rbdc->kref); | |
471 | INIT_LIST_HEAD(&rbdc->node); | |
472 | ||
bc534d86 AE |
473 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
474 | ||
43ae4701 | 475 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
602adf40 | 476 | if (IS_ERR(rbdc->client)) |
bc534d86 | 477 | goto out_mutex; |
43ae4701 | 478 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
602adf40 YS |
479 | |
480 | ret = ceph_open_session(rbdc->client); | |
481 | if (ret < 0) | |
482 | goto out_err; | |
483 | ||
432b8587 | 484 | spin_lock(&rbd_client_list_lock); |
602adf40 | 485 | list_add_tail(&rbdc->node, &rbd_client_list); |
432b8587 | 486 | spin_unlock(&rbd_client_list_lock); |
602adf40 | 487 | |
bc534d86 | 488 | mutex_unlock(&ctl_mutex); |
37206ee5 | 489 | dout("%s: rbdc %p\n", __func__, rbdc); |
bc534d86 | 490 | |
602adf40 YS |
491 | return rbdc; |
492 | ||
493 | out_err: | |
494 | ceph_destroy_client(rbdc->client); | |
bc534d86 AE |
495 | out_mutex: |
496 | mutex_unlock(&ctl_mutex); | |
602adf40 YS |
497 | kfree(rbdc); |
498 | out_opt: | |
43ae4701 AE |
499 | if (ceph_opts) |
500 | ceph_destroy_options(ceph_opts); | |
37206ee5 AE |
501 | dout("%s: error %d\n", __func__, ret); |
502 | ||
28f259b7 | 503 | return ERR_PTR(ret); |
602adf40 YS |
504 | } |
505 | ||
2f82ee54 AE |
506 | static struct rbd_client *__rbd_get_client(struct rbd_client *rbdc) |
507 | { | |
508 | kref_get(&rbdc->kref); | |
509 | ||
510 | return rbdc; | |
511 | } | |
512 | ||
602adf40 | 513 | /* |
1f7ba331 AE |
514 | * Find a ceph client with specific addr and configuration. If |
515 | * found, bump its reference count. | |
602adf40 | 516 | */ |
1f7ba331 | 517 | static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) |
602adf40 YS |
518 | { |
519 | struct rbd_client *client_node; | |
1f7ba331 | 520 | bool found = false; |
602adf40 | 521 | |
43ae4701 | 522 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
602adf40 YS |
523 | return NULL; |
524 | ||
1f7ba331 AE |
525 | spin_lock(&rbd_client_list_lock); |
526 | list_for_each_entry(client_node, &rbd_client_list, node) { | |
527 | if (!ceph_compare_options(ceph_opts, client_node->client)) { | |
2f82ee54 AE |
528 | __rbd_get_client(client_node); |
529 | ||
1f7ba331 AE |
530 | found = true; |
531 | break; | |
532 | } | |
533 | } | |
534 | spin_unlock(&rbd_client_list_lock); | |
535 | ||
536 | return found ? client_node : NULL; | |
602adf40 YS |
537 | } |
538 | ||
59c2be1e YS |
539 | /* |
540 | * mount options | |
541 | */ | |
542 | enum { | |
59c2be1e YS |
543 | Opt_last_int, |
544 | /* int args above */ | |
545 | Opt_last_string, | |
546 | /* string args above */ | |
cc0538b6 AE |
547 | Opt_read_only, |
548 | Opt_read_write, | |
549 | /* Boolean args above */ | |
550 | Opt_last_bool, | |
59c2be1e YS |
551 | }; |
552 | ||
43ae4701 | 553 | static match_table_t rbd_opts_tokens = { |
59c2be1e YS |
554 | /* int args above */ |
555 | /* string args above */ | |
be466c1c | 556 | {Opt_read_only, "read_only"}, |
cc0538b6 AE |
557 | {Opt_read_only, "ro"}, /* Alternate spelling */ |
558 | {Opt_read_write, "read_write"}, | |
559 | {Opt_read_write, "rw"}, /* Alternate spelling */ | |
560 | /* Boolean args above */ | |
59c2be1e YS |
561 | {-1, NULL} |
562 | }; | |
563 | ||
98571b5a AE |
564 | struct rbd_options { |
565 | bool read_only; | |
566 | }; | |
567 | ||
568 | #define RBD_READ_ONLY_DEFAULT false | |
569 | ||
59c2be1e YS |
570 | static int parse_rbd_opts_token(char *c, void *private) |
571 | { | |
43ae4701 | 572 | struct rbd_options *rbd_opts = private; |
59c2be1e YS |
573 | substring_t argstr[MAX_OPT_ARGS]; |
574 | int token, intval, ret; | |
575 | ||
43ae4701 | 576 | token = match_token(c, rbd_opts_tokens, argstr); |
59c2be1e YS |
577 | if (token < 0) |
578 | return -EINVAL; | |
579 | ||
580 | if (token < Opt_last_int) { | |
581 | ret = match_int(&argstr[0], &intval); | |
582 | if (ret < 0) { | |
583 | pr_err("bad mount option arg (not int) " | |
584 | "at '%s'\n", c); | |
585 | return ret; | |
586 | } | |
587 | dout("got int token %d val %d\n", token, intval); | |
588 | } else if (token > Opt_last_int && token < Opt_last_string) { | |
589 | dout("got string token %d val %s\n", token, | |
590 | argstr[0].from); | |
cc0538b6 AE |
591 | } else if (token > Opt_last_string && token < Opt_last_bool) { |
592 | dout("got Boolean token %d\n", token); | |
59c2be1e YS |
593 | } else { |
594 | dout("got token %d\n", token); | |
595 | } | |
596 | ||
597 | switch (token) { | |
cc0538b6 AE |
598 | case Opt_read_only: |
599 | rbd_opts->read_only = true; | |
600 | break; | |
601 | case Opt_read_write: | |
602 | rbd_opts->read_only = false; | |
603 | break; | |
59c2be1e | 604 | default: |
aafb230e AE |
605 | rbd_assert(false); |
606 | break; | |
59c2be1e YS |
607 | } |
608 | return 0; | |
609 | } | |
610 | ||
602adf40 YS |
611 | /* |
612 | * Get a ceph client with specific addr and configuration, if one does | |
613 | * not exist create it. | |
614 | */ | |
9d3997fd | 615 | static struct rbd_client *rbd_get_client(struct ceph_options *ceph_opts) |
602adf40 | 616 | { |
f8c38929 | 617 | struct rbd_client *rbdc; |
59c2be1e | 618 | |
1f7ba331 | 619 | rbdc = rbd_client_find(ceph_opts); |
9d3997fd | 620 | if (rbdc) /* using an existing client */ |
43ae4701 | 621 | ceph_destroy_options(ceph_opts); |
9d3997fd | 622 | else |
f8c38929 | 623 | rbdc = rbd_client_create(ceph_opts); |
602adf40 | 624 | |
9d3997fd | 625 | return rbdc; |
602adf40 YS |
626 | } |
627 | ||
628 | /* | |
629 | * Destroy ceph client | |
d23a4b3f | 630 | * |
432b8587 | 631 | * Caller must hold rbd_client_list_lock. |
602adf40 YS |
632 | */ |
633 | static void rbd_client_release(struct kref *kref) | |
634 | { | |
635 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); | |
636 | ||
37206ee5 | 637 | dout("%s: rbdc %p\n", __func__, rbdc); |
cd9d9f5d | 638 | spin_lock(&rbd_client_list_lock); |
602adf40 | 639 | list_del(&rbdc->node); |
cd9d9f5d | 640 | spin_unlock(&rbd_client_list_lock); |
602adf40 YS |
641 | |
642 | ceph_destroy_client(rbdc->client); | |
643 | kfree(rbdc); | |
644 | } | |
645 | ||
646 | /* | |
647 | * Drop reference to ceph client node. If it's not referenced anymore, release | |
648 | * it. | |
649 | */ | |
9d3997fd | 650 | static void rbd_put_client(struct rbd_client *rbdc) |
602adf40 | 651 | { |
c53d5893 AE |
652 | if (rbdc) |
653 | kref_put(&rbdc->kref, rbd_client_release); | |
602adf40 YS |
654 | } |
655 | ||
a30b71b9 AE |
656 | static bool rbd_image_format_valid(u32 image_format) |
657 | { | |
658 | return image_format == 1 || image_format == 2; | |
659 | } | |
660 | ||
8e94af8e AE |
661 | static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) |
662 | { | |
103a150f AE |
663 | size_t size; |
664 | u32 snap_count; | |
665 | ||
666 | /* The header has to start with the magic rbd header text */ | |
667 | if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) | |
668 | return false; | |
669 | ||
db2388b6 AE |
670 | /* The bio layer requires at least sector-sized I/O */ |
671 | ||
672 | if (ondisk->options.order < SECTOR_SHIFT) | |
673 | return false; | |
674 | ||
675 | /* If we use u64 in a few spots we may be able to loosen this */ | |
676 | ||
677 | if (ondisk->options.order > 8 * sizeof (int) - 1) | |
678 | return false; | |
679 | ||
103a150f AE |
680 | /* |
681 | * The size of a snapshot header has to fit in a size_t, and | |
682 | * that limits the number of snapshots. | |
683 | */ | |
684 | snap_count = le32_to_cpu(ondisk->snap_count); | |
685 | size = SIZE_MAX - sizeof (struct ceph_snap_context); | |
686 | if (snap_count > size / sizeof (__le64)) | |
687 | return false; | |
688 | ||
689 | /* | |
690 | * Not only that, but the size of the entire the snapshot | |
691 | * header must also be representable in a size_t. | |
692 | */ | |
693 | size -= snap_count * sizeof (__le64); | |
694 | if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) | |
695 | return false; | |
696 | ||
697 | return true; | |
8e94af8e AE |
698 | } |
699 | ||
602adf40 YS |
700 | /* |
701 | * Create a new header structure, translate header format from the on-disk | |
702 | * header. | |
703 | */ | |
704 | static int rbd_header_from_disk(struct rbd_image_header *header, | |
4156d998 | 705 | struct rbd_image_header_ondisk *ondisk) |
602adf40 | 706 | { |
ccece235 | 707 | u32 snap_count; |
58c17b0e | 708 | size_t len; |
d2bb24e5 | 709 | size_t size; |
621901d6 | 710 | u32 i; |
602adf40 | 711 | |
6a52325f AE |
712 | memset(header, 0, sizeof (*header)); |
713 | ||
103a150f AE |
714 | snap_count = le32_to_cpu(ondisk->snap_count); |
715 | ||
58c17b0e AE |
716 | len = strnlen(ondisk->object_prefix, sizeof (ondisk->object_prefix)); |
717 | header->object_prefix = kmalloc(len + 1, GFP_KERNEL); | |
6a52325f | 718 | if (!header->object_prefix) |
602adf40 | 719 | return -ENOMEM; |
58c17b0e AE |
720 | memcpy(header->object_prefix, ondisk->object_prefix, len); |
721 | header->object_prefix[len] = '\0'; | |
00f1f36f | 722 | |
602adf40 | 723 | if (snap_count) { |
f785cc1d AE |
724 | u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
725 | ||
621901d6 AE |
726 | /* Save a copy of the snapshot names */ |
727 | ||
f785cc1d AE |
728 | if (snap_names_len > (u64) SIZE_MAX) |
729 | return -EIO; | |
730 | header->snap_names = kmalloc(snap_names_len, GFP_KERNEL); | |
602adf40 | 731 | if (!header->snap_names) |
6a52325f | 732 | goto out_err; |
f785cc1d AE |
733 | /* |
734 | * Note that rbd_dev_v1_header_read() guarantees | |
735 | * the ondisk buffer we're working with has | |
736 | * snap_names_len bytes beyond the end of the | |
737 | * snapshot id array, this memcpy() is safe. | |
738 | */ | |
739 | memcpy(header->snap_names, &ondisk->snaps[snap_count], | |
740 | snap_names_len); | |
6a52325f | 741 | |
621901d6 AE |
742 | /* Record each snapshot's size */ |
743 | ||
d2bb24e5 AE |
744 | size = snap_count * sizeof (*header->snap_sizes); |
745 | header->snap_sizes = kmalloc(size, GFP_KERNEL); | |
602adf40 | 746 | if (!header->snap_sizes) |
6a52325f | 747 | goto out_err; |
621901d6 AE |
748 | for (i = 0; i < snap_count; i++) |
749 | header->snap_sizes[i] = | |
750 | le64_to_cpu(ondisk->snaps[i].image_size); | |
602adf40 | 751 | } else { |
ccece235 | 752 | WARN_ON(ondisk->snap_names_len); |
602adf40 YS |
753 | header->snap_names = NULL; |
754 | header->snap_sizes = NULL; | |
755 | } | |
849b4260 | 756 | |
34b13184 | 757 | header->features = 0; /* No features support in v1 images */ |
602adf40 YS |
758 | header->obj_order = ondisk->options.order; |
759 | header->crypt_type = ondisk->options.crypt_type; | |
760 | header->comp_type = ondisk->options.comp_type; | |
6a52325f | 761 | |
621901d6 AE |
762 | /* Allocate and fill in the snapshot context */ |
763 | ||
f84344f3 | 764 | header->image_size = le64_to_cpu(ondisk->image_size); |
6a52325f AE |
765 | size = sizeof (struct ceph_snap_context); |
766 | size += snap_count * sizeof (header->snapc->snaps[0]); | |
767 | header->snapc = kzalloc(size, GFP_KERNEL); | |
768 | if (!header->snapc) | |
769 | goto out_err; | |
602adf40 YS |
770 | |
771 | atomic_set(&header->snapc->nref, 1); | |
505cbb9b | 772 | header->snapc->seq = le64_to_cpu(ondisk->snap_seq); |
602adf40 | 773 | header->snapc->num_snaps = snap_count; |
621901d6 AE |
774 | for (i = 0; i < snap_count; i++) |
775 | header->snapc->snaps[i] = | |
776 | le64_to_cpu(ondisk->snaps[i].id); | |
602adf40 YS |
777 | |
778 | return 0; | |
779 | ||
6a52325f | 780 | out_err: |
849b4260 | 781 | kfree(header->snap_sizes); |
ccece235 | 782 | header->snap_sizes = NULL; |
602adf40 | 783 | kfree(header->snap_names); |
ccece235 | 784 | header->snap_names = NULL; |
6a52325f AE |
785 | kfree(header->object_prefix); |
786 | header->object_prefix = NULL; | |
ccece235 | 787 | |
00f1f36f | 788 | return -ENOMEM; |
602adf40 YS |
789 | } |
790 | ||
9e15b77d AE |
791 | static const char *rbd_snap_name(struct rbd_device *rbd_dev, u64 snap_id) |
792 | { | |
793 | struct rbd_snap *snap; | |
794 | ||
795 | if (snap_id == CEPH_NOSNAP) | |
796 | return RBD_SNAP_HEAD_NAME; | |
797 | ||
798 | list_for_each_entry(snap, &rbd_dev->snaps, node) | |
799 | if (snap_id == snap->id) | |
800 | return snap->name; | |
801 | ||
802 | return NULL; | |
803 | } | |
804 | ||
8836b995 | 805 | static int snap_by_name(struct rbd_device *rbd_dev, const char *snap_name) |
602adf40 | 806 | { |
602adf40 | 807 | |
e86924a8 | 808 | struct rbd_snap *snap; |
602adf40 | 809 | |
e86924a8 AE |
810 | list_for_each_entry(snap, &rbd_dev->snaps, node) { |
811 | if (!strcmp(snap_name, snap->name)) { | |
0d7dbfce | 812 | rbd_dev->spec->snap_id = snap->id; |
e86924a8 | 813 | rbd_dev->mapping.size = snap->size; |
34b13184 | 814 | rbd_dev->mapping.features = snap->features; |
602adf40 | 815 | |
e86924a8 | 816 | return 0; |
00f1f36f | 817 | } |
00f1f36f | 818 | } |
e86924a8 | 819 | |
00f1f36f | 820 | return -ENOENT; |
602adf40 YS |
821 | } |
822 | ||
819d52bf | 823 | static int rbd_dev_set_mapping(struct rbd_device *rbd_dev) |
602adf40 | 824 | { |
78dc447d | 825 | int ret; |
602adf40 | 826 | |
0d7dbfce | 827 | if (!memcmp(rbd_dev->spec->snap_name, RBD_SNAP_HEAD_NAME, |
cc9d734c | 828 | sizeof (RBD_SNAP_HEAD_NAME))) { |
0d7dbfce | 829 | rbd_dev->spec->snap_id = CEPH_NOSNAP; |
99c1f08f | 830 | rbd_dev->mapping.size = rbd_dev->header.image_size; |
34b13184 | 831 | rbd_dev->mapping.features = rbd_dev->header.features; |
e86924a8 | 832 | ret = 0; |
602adf40 | 833 | } else { |
0d7dbfce | 834 | ret = snap_by_name(rbd_dev, rbd_dev->spec->snap_name); |
602adf40 YS |
835 | if (ret < 0) |
836 | goto done; | |
f84344f3 | 837 | rbd_dev->mapping.read_only = true; |
602adf40 | 838 | } |
6d292906 AE |
839 | set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
840 | ||
602adf40 | 841 | done: |
602adf40 YS |
842 | return ret; |
843 | } | |
844 | ||
845 | static void rbd_header_free(struct rbd_image_header *header) | |
846 | { | |
849b4260 | 847 | kfree(header->object_prefix); |
d78fd7ae | 848 | header->object_prefix = NULL; |
602adf40 | 849 | kfree(header->snap_sizes); |
d78fd7ae | 850 | header->snap_sizes = NULL; |
849b4260 | 851 | kfree(header->snap_names); |
d78fd7ae | 852 | header->snap_names = NULL; |
d1d25646 | 853 | ceph_put_snap_context(header->snapc); |
d78fd7ae | 854 | header->snapc = NULL; |
602adf40 YS |
855 | } |
856 | ||
98571b5a | 857 | static const char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset) |
602adf40 | 858 | { |
65ccfe21 AE |
859 | char *name; |
860 | u64 segment; | |
861 | int ret; | |
602adf40 | 862 | |
2fd82b9e | 863 | name = kmalloc(MAX_OBJ_NAME_SIZE + 1, GFP_NOIO); |
65ccfe21 AE |
864 | if (!name) |
865 | return NULL; | |
866 | segment = offset >> rbd_dev->header.obj_order; | |
2fd82b9e | 867 | ret = snprintf(name, MAX_OBJ_NAME_SIZE + 1, "%s.%012llx", |
65ccfe21 | 868 | rbd_dev->header.object_prefix, segment); |
2fd82b9e | 869 | if (ret < 0 || ret > MAX_OBJ_NAME_SIZE) { |
65ccfe21 AE |
870 | pr_err("error formatting segment name for #%llu (%d)\n", |
871 | segment, ret); | |
872 | kfree(name); | |
873 | name = NULL; | |
874 | } | |
602adf40 | 875 | |
65ccfe21 AE |
876 | return name; |
877 | } | |
602adf40 | 878 | |
65ccfe21 AE |
879 | static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) |
880 | { | |
881 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; | |
602adf40 | 882 | |
65ccfe21 AE |
883 | return offset & (segment_size - 1); |
884 | } | |
885 | ||
886 | static u64 rbd_segment_length(struct rbd_device *rbd_dev, | |
887 | u64 offset, u64 length) | |
888 | { | |
889 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; | |
890 | ||
891 | offset &= segment_size - 1; | |
892 | ||
aafb230e | 893 | rbd_assert(length <= U64_MAX - offset); |
65ccfe21 AE |
894 | if (offset + length > segment_size) |
895 | length = segment_size - offset; | |
896 | ||
897 | return length; | |
602adf40 YS |
898 | } |
899 | ||
029bcbd8 JD |
900 | /* |
901 | * returns the size of an object in the image | |
902 | */ | |
903 | static u64 rbd_obj_bytes(struct rbd_image_header *header) | |
904 | { | |
905 | return 1 << header->obj_order; | |
906 | } | |
907 | ||
602adf40 YS |
908 | /* |
909 | * bio helpers | |
910 | */ | |
911 | ||
912 | static void bio_chain_put(struct bio *chain) | |
913 | { | |
914 | struct bio *tmp; | |
915 | ||
916 | while (chain) { | |
917 | tmp = chain; | |
918 | chain = chain->bi_next; | |
919 | bio_put(tmp); | |
920 | } | |
921 | } | |
922 | ||
923 | /* | |
924 | * zeros a bio chain, starting at specific offset | |
925 | */ | |
926 | static void zero_bio_chain(struct bio *chain, int start_ofs) | |
927 | { | |
928 | struct bio_vec *bv; | |
929 | unsigned long flags; | |
930 | void *buf; | |
931 | int i; | |
932 | int pos = 0; | |
933 | ||
934 | while (chain) { | |
935 | bio_for_each_segment(bv, chain, i) { | |
936 | if (pos + bv->bv_len > start_ofs) { | |
937 | int remainder = max(start_ofs - pos, 0); | |
938 | buf = bvec_kmap_irq(bv, &flags); | |
939 | memset(buf + remainder, 0, | |
940 | bv->bv_len - remainder); | |
85b5aaa6 | 941 | bvec_kunmap_irq(buf, &flags); |
602adf40 YS |
942 | } |
943 | pos += bv->bv_len; | |
944 | } | |
945 | ||
946 | chain = chain->bi_next; | |
947 | } | |
948 | } | |
949 | ||
950 | /* | |
f7760dad AE |
951 | * Clone a portion of a bio, starting at the given byte offset |
952 | * and continuing for the number of bytes indicated. | |
602adf40 | 953 | */ |
f7760dad AE |
954 | static struct bio *bio_clone_range(struct bio *bio_src, |
955 | unsigned int offset, | |
956 | unsigned int len, | |
957 | gfp_t gfpmask) | |
602adf40 | 958 | { |
f7760dad AE |
959 | struct bio_vec *bv; |
960 | unsigned int resid; | |
961 | unsigned short idx; | |
962 | unsigned int voff; | |
963 | unsigned short end_idx; | |
964 | unsigned short vcnt; | |
965 | struct bio *bio; | |
966 | ||
967 | /* Handle the easy case for the caller */ | |
968 | ||
969 | if (!offset && len == bio_src->bi_size) | |
970 | return bio_clone(bio_src, gfpmask); | |
971 | ||
972 | if (WARN_ON_ONCE(!len)) | |
973 | return NULL; | |
974 | if (WARN_ON_ONCE(len > bio_src->bi_size)) | |
975 | return NULL; | |
976 | if (WARN_ON_ONCE(offset > bio_src->bi_size - len)) | |
977 | return NULL; | |
978 | ||
979 | /* Find first affected segment... */ | |
980 | ||
981 | resid = offset; | |
982 | __bio_for_each_segment(bv, bio_src, idx, 0) { | |
983 | if (resid < bv->bv_len) | |
984 | break; | |
985 | resid -= bv->bv_len; | |
602adf40 | 986 | } |
f7760dad | 987 | voff = resid; |
602adf40 | 988 | |
f7760dad | 989 | /* ...and the last affected segment */ |
602adf40 | 990 | |
f7760dad AE |
991 | resid += len; |
992 | __bio_for_each_segment(bv, bio_src, end_idx, idx) { | |
993 | if (resid <= bv->bv_len) | |
994 | break; | |
995 | resid -= bv->bv_len; | |
996 | } | |
997 | vcnt = end_idx - idx + 1; | |
998 | ||
999 | /* Build the clone */ | |
1000 | ||
1001 | bio = bio_alloc(gfpmask, (unsigned int) vcnt); | |
1002 | if (!bio) | |
1003 | return NULL; /* ENOMEM */ | |
602adf40 | 1004 | |
f7760dad AE |
1005 | bio->bi_bdev = bio_src->bi_bdev; |
1006 | bio->bi_sector = bio_src->bi_sector + (offset >> SECTOR_SHIFT); | |
1007 | bio->bi_rw = bio_src->bi_rw; | |
1008 | bio->bi_flags |= 1 << BIO_CLONED; | |
1009 | ||
1010 | /* | |
1011 | * Copy over our part of the bio_vec, then update the first | |
1012 | * and last (or only) entries. | |
1013 | */ | |
1014 | memcpy(&bio->bi_io_vec[0], &bio_src->bi_io_vec[idx], | |
1015 | vcnt * sizeof (struct bio_vec)); | |
1016 | bio->bi_io_vec[0].bv_offset += voff; | |
1017 | if (vcnt > 1) { | |
1018 | bio->bi_io_vec[0].bv_len -= voff; | |
1019 | bio->bi_io_vec[vcnt - 1].bv_len = resid; | |
1020 | } else { | |
1021 | bio->bi_io_vec[0].bv_len = len; | |
602adf40 YS |
1022 | } |
1023 | ||
f7760dad AE |
1024 | bio->bi_vcnt = vcnt; |
1025 | bio->bi_size = len; | |
1026 | bio->bi_idx = 0; | |
1027 | ||
1028 | return bio; | |
1029 | } | |
1030 | ||
1031 | /* | |
1032 | * Clone a portion of a bio chain, starting at the given byte offset | |
1033 | * into the first bio in the source chain and continuing for the | |
1034 | * number of bytes indicated. The result is another bio chain of | |
1035 | * exactly the given length, or a null pointer on error. | |
1036 | * | |
1037 | * The bio_src and offset parameters are both in-out. On entry they | |
1038 | * refer to the first source bio and the offset into that bio where | |
1039 | * the start of data to be cloned is located. | |
1040 | * | |
1041 | * On return, bio_src is updated to refer to the bio in the source | |
1042 | * chain that contains first un-cloned byte, and *offset will | |
1043 | * contain the offset of that byte within that bio. | |
1044 | */ | |
1045 | static struct bio *bio_chain_clone_range(struct bio **bio_src, | |
1046 | unsigned int *offset, | |
1047 | unsigned int len, | |
1048 | gfp_t gfpmask) | |
1049 | { | |
1050 | struct bio *bi = *bio_src; | |
1051 | unsigned int off = *offset; | |
1052 | struct bio *chain = NULL; | |
1053 | struct bio **end; | |
1054 | ||
1055 | /* Build up a chain of clone bios up to the limit */ | |
1056 | ||
1057 | if (!bi || off >= bi->bi_size || !len) | |
1058 | return NULL; /* Nothing to clone */ | |
602adf40 | 1059 | |
f7760dad AE |
1060 | end = &chain; |
1061 | while (len) { | |
1062 | unsigned int bi_size; | |
1063 | struct bio *bio; | |
1064 | ||
f5400b7a AE |
1065 | if (!bi) { |
1066 | rbd_warn(NULL, "bio_chain exhausted with %u left", len); | |
f7760dad | 1067 | goto out_err; /* EINVAL; ran out of bio's */ |
f5400b7a | 1068 | } |
f7760dad AE |
1069 | bi_size = min_t(unsigned int, bi->bi_size - off, len); |
1070 | bio = bio_clone_range(bi, off, bi_size, gfpmask); | |
1071 | if (!bio) | |
1072 | goto out_err; /* ENOMEM */ | |
1073 | ||
1074 | *end = bio; | |
1075 | end = &bio->bi_next; | |
602adf40 | 1076 | |
f7760dad AE |
1077 | off += bi_size; |
1078 | if (off == bi->bi_size) { | |
1079 | bi = bi->bi_next; | |
1080 | off = 0; | |
1081 | } | |
1082 | len -= bi_size; | |
1083 | } | |
1084 | *bio_src = bi; | |
1085 | *offset = off; | |
1086 | ||
1087 | return chain; | |
1088 | out_err: | |
1089 | bio_chain_put(chain); | |
602adf40 | 1090 | |
602adf40 YS |
1091 | return NULL; |
1092 | } | |
1093 | ||
926f9b3f AE |
1094 | /* |
1095 | * The default/initial value for all object request flags is 0. For | |
1096 | * each flag, once its value is set to 1 it is never reset to 0 | |
1097 | * again. | |
1098 | */ | |
57acbaa7 | 1099 | static void obj_request_img_data_set(struct rbd_obj_request *obj_request) |
926f9b3f | 1100 | { |
57acbaa7 | 1101 | if (test_and_set_bit(OBJ_REQ_IMG_DATA, &obj_request->flags)) { |
926f9b3f AE |
1102 | struct rbd_device *rbd_dev; |
1103 | ||
57acbaa7 AE |
1104 | rbd_dev = obj_request->img_request->rbd_dev; |
1105 | rbd_warn(rbd_dev, "obj_request %p already marked img_data\n", | |
926f9b3f AE |
1106 | obj_request); |
1107 | } | |
1108 | } | |
1109 | ||
57acbaa7 | 1110 | static bool obj_request_img_data_test(struct rbd_obj_request *obj_request) |
926f9b3f AE |
1111 | { |
1112 | smp_mb(); | |
57acbaa7 | 1113 | return test_bit(OBJ_REQ_IMG_DATA, &obj_request->flags) != 0; |
926f9b3f AE |
1114 | } |
1115 | ||
57acbaa7 | 1116 | static void obj_request_done_set(struct rbd_obj_request *obj_request) |
6365d33a | 1117 | { |
57acbaa7 AE |
1118 | if (test_and_set_bit(OBJ_REQ_DONE, &obj_request->flags)) { |
1119 | struct rbd_device *rbd_dev = NULL; | |
6365d33a | 1120 | |
57acbaa7 AE |
1121 | if (obj_request_img_data_test(obj_request)) |
1122 | rbd_dev = obj_request->img_request->rbd_dev; | |
1123 | rbd_warn(rbd_dev, "obj_request %p already marked done\n", | |
6365d33a AE |
1124 | obj_request); |
1125 | } | |
1126 | } | |
1127 | ||
57acbaa7 | 1128 | static bool obj_request_done_test(struct rbd_obj_request *obj_request) |
6365d33a AE |
1129 | { |
1130 | smp_mb(); | |
57acbaa7 | 1131 | return test_bit(OBJ_REQ_DONE, &obj_request->flags) != 0; |
6365d33a AE |
1132 | } |
1133 | ||
5679c59f AE |
1134 | /* |
1135 | * This sets the KNOWN flag after (possibly) setting the EXISTS | |
1136 | * flag. The latter is set based on the "exists" value provided. | |
1137 | * | |
1138 | * Note that for our purposes once an object exists it never goes | |
1139 | * away again. It's possible that the response from two existence | |
1140 | * checks are separated by the creation of the target object, and | |
1141 | * the first ("doesn't exist") response arrives *after* the second | |
1142 | * ("does exist"). In that case we ignore the second one. | |
1143 | */ | |
1144 | static void obj_request_existence_set(struct rbd_obj_request *obj_request, | |
1145 | bool exists) | |
1146 | { | |
1147 | if (exists) | |
1148 | set_bit(OBJ_REQ_EXISTS, &obj_request->flags); | |
1149 | set_bit(OBJ_REQ_KNOWN, &obj_request->flags); | |
1150 | smp_mb(); | |
1151 | } | |
1152 | ||
1153 | static bool obj_request_known_test(struct rbd_obj_request *obj_request) | |
1154 | { | |
1155 | smp_mb(); | |
1156 | return test_bit(OBJ_REQ_KNOWN, &obj_request->flags) != 0; | |
1157 | } | |
1158 | ||
1159 | static bool obj_request_exists_test(struct rbd_obj_request *obj_request) | |
1160 | { | |
1161 | smp_mb(); | |
1162 | return test_bit(OBJ_REQ_EXISTS, &obj_request->flags) != 0; | |
1163 | } | |
1164 | ||
bf0d5f50 AE |
1165 | static void rbd_obj_request_get(struct rbd_obj_request *obj_request) |
1166 | { | |
37206ee5 AE |
1167 | dout("%s: obj %p (was %d)\n", __func__, obj_request, |
1168 | atomic_read(&obj_request->kref.refcount)); | |
bf0d5f50 AE |
1169 | kref_get(&obj_request->kref); |
1170 | } | |
1171 | ||
1172 | static void rbd_obj_request_destroy(struct kref *kref); | |
1173 | static void rbd_obj_request_put(struct rbd_obj_request *obj_request) | |
1174 | { | |
1175 | rbd_assert(obj_request != NULL); | |
37206ee5 AE |
1176 | dout("%s: obj %p (was %d)\n", __func__, obj_request, |
1177 | atomic_read(&obj_request->kref.refcount)); | |
bf0d5f50 AE |
1178 | kref_put(&obj_request->kref, rbd_obj_request_destroy); |
1179 | } | |
1180 | ||
1181 | static void rbd_img_request_get(struct rbd_img_request *img_request) | |
1182 | { | |
37206ee5 AE |
1183 | dout("%s: img %p (was %d)\n", __func__, img_request, |
1184 | atomic_read(&img_request->kref.refcount)); | |
bf0d5f50 AE |
1185 | kref_get(&img_request->kref); |
1186 | } | |
1187 | ||
1188 | static void rbd_img_request_destroy(struct kref *kref); | |
1189 | static void rbd_img_request_put(struct rbd_img_request *img_request) | |
1190 | { | |
1191 | rbd_assert(img_request != NULL); | |
37206ee5 AE |
1192 | dout("%s: img %p (was %d)\n", __func__, img_request, |
1193 | atomic_read(&img_request->kref.refcount)); | |
bf0d5f50 AE |
1194 | kref_put(&img_request->kref, rbd_img_request_destroy); |
1195 | } | |
1196 | ||
1197 | static inline void rbd_img_obj_request_add(struct rbd_img_request *img_request, | |
1198 | struct rbd_obj_request *obj_request) | |
1199 | { | |
25dcf954 AE |
1200 | rbd_assert(obj_request->img_request == NULL); |
1201 | ||
b155e86c | 1202 | /* Image request now owns object's original reference */ |
bf0d5f50 | 1203 | obj_request->img_request = img_request; |
25dcf954 | 1204 | obj_request->which = img_request->obj_request_count; |
6365d33a AE |
1205 | rbd_assert(!obj_request_img_data_test(obj_request)); |
1206 | obj_request_img_data_set(obj_request); | |
bf0d5f50 | 1207 | rbd_assert(obj_request->which != BAD_WHICH); |
25dcf954 AE |
1208 | img_request->obj_request_count++; |
1209 | list_add_tail(&obj_request->links, &img_request->obj_requests); | |
37206ee5 AE |
1210 | dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request, |
1211 | obj_request->which); | |
bf0d5f50 AE |
1212 | } |
1213 | ||
1214 | static inline void rbd_img_obj_request_del(struct rbd_img_request *img_request, | |
1215 | struct rbd_obj_request *obj_request) | |
1216 | { | |
1217 | rbd_assert(obj_request->which != BAD_WHICH); | |
25dcf954 | 1218 | |
37206ee5 AE |
1219 | dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request, |
1220 | obj_request->which); | |
bf0d5f50 | 1221 | list_del(&obj_request->links); |
25dcf954 AE |
1222 | rbd_assert(img_request->obj_request_count > 0); |
1223 | img_request->obj_request_count--; | |
1224 | rbd_assert(obj_request->which == img_request->obj_request_count); | |
1225 | obj_request->which = BAD_WHICH; | |
6365d33a | 1226 | rbd_assert(obj_request_img_data_test(obj_request)); |
bf0d5f50 | 1227 | rbd_assert(obj_request->img_request == img_request); |
bf0d5f50 | 1228 | obj_request->img_request = NULL; |
25dcf954 | 1229 | obj_request->callback = NULL; |
bf0d5f50 AE |
1230 | rbd_obj_request_put(obj_request); |
1231 | } | |
1232 | ||
1233 | static bool obj_request_type_valid(enum obj_request_type type) | |
1234 | { | |
1235 | switch (type) { | |
9969ebc5 | 1236 | case OBJ_REQUEST_NODATA: |
bf0d5f50 | 1237 | case OBJ_REQUEST_BIO: |
788e2df3 | 1238 | case OBJ_REQUEST_PAGES: |
bf0d5f50 AE |
1239 | return true; |
1240 | default: | |
1241 | return false; | |
1242 | } | |
1243 | } | |
1244 | ||
bf0d5f50 AE |
1245 | static int rbd_obj_request_submit(struct ceph_osd_client *osdc, |
1246 | struct rbd_obj_request *obj_request) | |
1247 | { | |
37206ee5 AE |
1248 | dout("%s: osdc %p obj %p\n", __func__, osdc, obj_request); |
1249 | ||
bf0d5f50 AE |
1250 | return ceph_osdc_start_request(osdc, obj_request->osd_req, false); |
1251 | } | |
1252 | ||
1253 | static void rbd_img_request_complete(struct rbd_img_request *img_request) | |
1254 | { | |
55f27e09 | 1255 | |
37206ee5 | 1256 | dout("%s: img %p\n", __func__, img_request); |
55f27e09 AE |
1257 | |
1258 | /* | |
1259 | * If no error occurred, compute the aggregate transfer | |
1260 | * count for the image request. We could instead use | |
1261 | * atomic64_cmpxchg() to update it as each object request | |
1262 | * completes; not clear which way is better off hand. | |
1263 | */ | |
1264 | if (!img_request->result) { | |
1265 | struct rbd_obj_request *obj_request; | |
1266 | u64 xferred = 0; | |
1267 | ||
1268 | for_each_obj_request(img_request, obj_request) | |
1269 | xferred += obj_request->xferred; | |
1270 | img_request->xferred = xferred; | |
1271 | } | |
1272 | ||
bf0d5f50 AE |
1273 | if (img_request->callback) |
1274 | img_request->callback(img_request); | |
1275 | else | |
1276 | rbd_img_request_put(img_request); | |
1277 | } | |
1278 | ||
788e2df3 AE |
1279 | /* Caller is responsible for rbd_obj_request_destroy(obj_request) */ |
1280 | ||
1281 | static int rbd_obj_request_wait(struct rbd_obj_request *obj_request) | |
1282 | { | |
37206ee5 AE |
1283 | dout("%s: obj %p\n", __func__, obj_request); |
1284 | ||
788e2df3 AE |
1285 | return wait_for_completion_interruptible(&obj_request->completion); |
1286 | } | |
1287 | ||
0c425248 AE |
1288 | /* |
1289 | * The default/initial value for all image request flags is 0. Each | |
1290 | * is conditionally set to 1 at image request initialization time | |
1291 | * and currently never change thereafter. | |
1292 | */ | |
1293 | static void img_request_write_set(struct rbd_img_request *img_request) | |
1294 | { | |
1295 | set_bit(IMG_REQ_WRITE, &img_request->flags); | |
1296 | smp_mb(); | |
1297 | } | |
1298 | ||
1299 | static bool img_request_write_test(struct rbd_img_request *img_request) | |
1300 | { | |
1301 | smp_mb(); | |
1302 | return test_bit(IMG_REQ_WRITE, &img_request->flags) != 0; | |
1303 | } | |
1304 | ||
9849e986 AE |
1305 | static void img_request_child_set(struct rbd_img_request *img_request) |
1306 | { | |
1307 | set_bit(IMG_REQ_CHILD, &img_request->flags); | |
1308 | smp_mb(); | |
1309 | } | |
1310 | ||
1311 | static bool img_request_child_test(struct rbd_img_request *img_request) | |
1312 | { | |
1313 | smp_mb(); | |
1314 | return test_bit(IMG_REQ_CHILD, &img_request->flags) != 0; | |
1315 | } | |
1316 | ||
d0b2e944 AE |
1317 | static void img_request_layered_set(struct rbd_img_request *img_request) |
1318 | { | |
1319 | set_bit(IMG_REQ_LAYERED, &img_request->flags); | |
1320 | smp_mb(); | |
1321 | } | |
1322 | ||
1323 | static bool img_request_layered_test(struct rbd_img_request *img_request) | |
1324 | { | |
1325 | smp_mb(); | |
1326 | return test_bit(IMG_REQ_LAYERED, &img_request->flags) != 0; | |
1327 | } | |
1328 | ||
6e2a4505 AE |
1329 | static void |
1330 | rbd_img_obj_request_read_callback(struct rbd_obj_request *obj_request) | |
1331 | { | |
1332 | dout("%s: obj %p img %p result %d %llu/%llu\n", __func__, | |
1333 | obj_request, obj_request->img_request, obj_request->result, | |
1334 | obj_request->xferred, obj_request->length); | |
1335 | /* | |
1336 | * ENOENT means a hole in the image. We zero-fill the | |
1337 | * entire length of the request. A short read also implies | |
1338 | * zero-fill to the end of the request. Either way we | |
1339 | * update the xferred count to indicate the whole request | |
1340 | * was satisfied. | |
1341 | */ | |
1342 | BUG_ON(obj_request->type != OBJ_REQUEST_BIO); | |
1343 | if (obj_request->result == -ENOENT) { | |
1344 | zero_bio_chain(obj_request->bio_list, 0); | |
1345 | obj_request->result = 0; | |
1346 | obj_request->xferred = obj_request->length; | |
1347 | } else if (obj_request->xferred < obj_request->length && | |
1348 | !obj_request->result) { | |
1349 | zero_bio_chain(obj_request->bio_list, obj_request->xferred); | |
1350 | obj_request->xferred = obj_request->length; | |
1351 | } | |
1352 | obj_request_done_set(obj_request); | |
1353 | } | |
1354 | ||
bf0d5f50 AE |
1355 | static void rbd_obj_request_complete(struct rbd_obj_request *obj_request) |
1356 | { | |
37206ee5 AE |
1357 | dout("%s: obj %p cb %p\n", __func__, obj_request, |
1358 | obj_request->callback); | |
bf0d5f50 AE |
1359 | if (obj_request->callback) |
1360 | obj_request->callback(obj_request); | |
788e2df3 AE |
1361 | else |
1362 | complete_all(&obj_request->completion); | |
bf0d5f50 AE |
1363 | } |
1364 | ||
c47f9371 | 1365 | static void rbd_osd_trivial_callback(struct rbd_obj_request *obj_request) |
39bf2c5d AE |
1366 | { |
1367 | dout("%s: obj %p\n", __func__, obj_request); | |
1368 | obj_request_done_set(obj_request); | |
1369 | } | |
1370 | ||
c47f9371 | 1371 | static void rbd_osd_read_callback(struct rbd_obj_request *obj_request) |
bf0d5f50 | 1372 | { |
57acbaa7 AE |
1373 | struct rbd_img_request *img_request = NULL; |
1374 | bool layered = false; | |
1375 | ||
1376 | if (obj_request_img_data_test(obj_request)) { | |
1377 | img_request = obj_request->img_request; | |
1378 | layered = img_request && img_request_layered_test(img_request); | |
1379 | } else { | |
1380 | img_request = NULL; | |
1381 | layered = false; | |
1382 | } | |
8b3e1a56 AE |
1383 | |
1384 | dout("%s: obj %p img %p result %d %llu/%llu\n", __func__, | |
1385 | obj_request, img_request, obj_request->result, | |
1386 | obj_request->xferred, obj_request->length); | |
1387 | if (layered && obj_request->result == -ENOENT) | |
1388 | rbd_img_parent_read(obj_request); | |
1389 | else if (img_request) | |
6e2a4505 AE |
1390 | rbd_img_obj_request_read_callback(obj_request); |
1391 | else | |
1392 | obj_request_done_set(obj_request); | |
bf0d5f50 AE |
1393 | } |
1394 | ||
c47f9371 | 1395 | static void rbd_osd_write_callback(struct rbd_obj_request *obj_request) |
bf0d5f50 | 1396 | { |
1b83bef2 SW |
1397 | dout("%s: obj %p result %d %llu\n", __func__, obj_request, |
1398 | obj_request->result, obj_request->length); | |
1399 | /* | |
8b3e1a56 AE |
1400 | * There is no such thing as a successful short write. Set |
1401 | * it to our originally-requested length. | |
1b83bef2 SW |
1402 | */ |
1403 | obj_request->xferred = obj_request->length; | |
07741308 | 1404 | obj_request_done_set(obj_request); |
bf0d5f50 AE |
1405 | } |
1406 | ||
fbfab539 AE |
1407 | /* |
1408 | * For a simple stat call there's nothing to do. We'll do more if | |
1409 | * this is part of a write sequence for a layered image. | |
1410 | */ | |
c47f9371 | 1411 | static void rbd_osd_stat_callback(struct rbd_obj_request *obj_request) |
fbfab539 | 1412 | { |
37206ee5 | 1413 | dout("%s: obj %p\n", __func__, obj_request); |
fbfab539 AE |
1414 | obj_request_done_set(obj_request); |
1415 | } | |
1416 | ||
bf0d5f50 AE |
1417 | static void rbd_osd_req_callback(struct ceph_osd_request *osd_req, |
1418 | struct ceph_msg *msg) | |
1419 | { | |
1420 | struct rbd_obj_request *obj_request = osd_req->r_priv; | |
bf0d5f50 AE |
1421 | u16 opcode; |
1422 | ||
37206ee5 | 1423 | dout("%s: osd_req %p msg %p\n", __func__, osd_req, msg); |
bf0d5f50 | 1424 | rbd_assert(osd_req == obj_request->osd_req); |
57acbaa7 AE |
1425 | if (obj_request_img_data_test(obj_request)) { |
1426 | rbd_assert(obj_request->img_request); | |
1427 | rbd_assert(obj_request->which != BAD_WHICH); | |
1428 | } else { | |
1429 | rbd_assert(obj_request->which == BAD_WHICH); | |
1430 | } | |
bf0d5f50 | 1431 | |
1b83bef2 SW |
1432 | if (osd_req->r_result < 0) |
1433 | obj_request->result = osd_req->r_result; | |
bf0d5f50 AE |
1434 | obj_request->version = le64_to_cpu(osd_req->r_reassert_version.version); |
1435 | ||
1b83bef2 | 1436 | WARN_ON(osd_req->r_num_ops != 1); /* For now */ |
bf0d5f50 | 1437 | |
c47f9371 AE |
1438 | /* |
1439 | * We support a 64-bit length, but ultimately it has to be | |
1440 | * passed to blk_end_request(), which takes an unsigned int. | |
1441 | */ | |
1b83bef2 | 1442 | obj_request->xferred = osd_req->r_reply_op_len[0]; |
8b3e1a56 | 1443 | rbd_assert(obj_request->xferred < (u64)UINT_MAX); |
79528734 | 1444 | opcode = osd_req->r_ops[0].op; |
bf0d5f50 AE |
1445 | switch (opcode) { |
1446 | case CEPH_OSD_OP_READ: | |
c47f9371 | 1447 | rbd_osd_read_callback(obj_request); |
bf0d5f50 AE |
1448 | break; |
1449 | case CEPH_OSD_OP_WRITE: | |
c47f9371 | 1450 | rbd_osd_write_callback(obj_request); |
bf0d5f50 | 1451 | break; |
fbfab539 | 1452 | case CEPH_OSD_OP_STAT: |
c47f9371 | 1453 | rbd_osd_stat_callback(obj_request); |
fbfab539 | 1454 | break; |
36be9a76 | 1455 | case CEPH_OSD_OP_CALL: |
b8d70035 | 1456 | case CEPH_OSD_OP_NOTIFY_ACK: |
9969ebc5 | 1457 | case CEPH_OSD_OP_WATCH: |
c47f9371 | 1458 | rbd_osd_trivial_callback(obj_request); |
9969ebc5 | 1459 | break; |
bf0d5f50 AE |
1460 | default: |
1461 | rbd_warn(NULL, "%s: unsupported op %hu\n", | |
1462 | obj_request->object_name, (unsigned short) opcode); | |
1463 | break; | |
1464 | } | |
1465 | ||
07741308 | 1466 | if (obj_request_done_test(obj_request)) |
bf0d5f50 AE |
1467 | rbd_obj_request_complete(obj_request); |
1468 | } | |
1469 | ||
2fa12320 | 1470 | static void rbd_osd_req_format(struct rbd_obj_request *obj_request, |
79528734 | 1471 | bool write_request) |
430c28c3 AE |
1472 | { |
1473 | struct rbd_img_request *img_request = obj_request->img_request; | |
8c042b0d | 1474 | struct ceph_osd_request *osd_req = obj_request->osd_req; |
430c28c3 AE |
1475 | struct ceph_snap_context *snapc = NULL; |
1476 | u64 snap_id = CEPH_NOSNAP; | |
1477 | struct timespec *mtime = NULL; | |
1478 | struct timespec now; | |
1479 | ||
8c042b0d | 1480 | rbd_assert(osd_req != NULL); |
430c28c3 AE |
1481 | |
1482 | if (write_request) { | |
1483 | now = CURRENT_TIME; | |
1484 | mtime = &now; | |
1485 | if (img_request) | |
1486 | snapc = img_request->snapc; | |
2fa12320 AE |
1487 | } else if (img_request) { |
1488 | snap_id = img_request->snap_id; | |
8c042b0d AE |
1489 | } |
1490 | ceph_osdc_build_request(osd_req, obj_request->offset, | |
79528734 | 1491 | snapc, snap_id, mtime); |
430c28c3 AE |
1492 | } |
1493 | ||
bf0d5f50 AE |
1494 | static struct ceph_osd_request *rbd_osd_req_create( |
1495 | struct rbd_device *rbd_dev, | |
1496 | bool write_request, | |
430c28c3 | 1497 | struct rbd_obj_request *obj_request) |
bf0d5f50 | 1498 | { |
bf0d5f50 AE |
1499 | struct ceph_snap_context *snapc = NULL; |
1500 | struct ceph_osd_client *osdc; | |
1501 | struct ceph_osd_request *osd_req; | |
bf0d5f50 | 1502 | |
6365d33a AE |
1503 | if (obj_request_img_data_test(obj_request)) { |
1504 | struct rbd_img_request *img_request = obj_request->img_request; | |
1505 | ||
0c425248 AE |
1506 | rbd_assert(write_request == |
1507 | img_request_write_test(img_request)); | |
1508 | if (write_request) | |
bf0d5f50 | 1509 | snapc = img_request->snapc; |
bf0d5f50 AE |
1510 | } |
1511 | ||
1512 | /* Allocate and initialize the request, for the single op */ | |
1513 | ||
1514 | osdc = &rbd_dev->rbd_client->client->osdc; | |
1515 | osd_req = ceph_osdc_alloc_request(osdc, snapc, 1, false, GFP_ATOMIC); | |
1516 | if (!osd_req) | |
1517 | return NULL; /* ENOMEM */ | |
bf0d5f50 | 1518 | |
430c28c3 | 1519 | if (write_request) |
bf0d5f50 | 1520 | osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; |
430c28c3 | 1521 | else |
bf0d5f50 | 1522 | osd_req->r_flags = CEPH_OSD_FLAG_READ; |
bf0d5f50 AE |
1523 | |
1524 | osd_req->r_callback = rbd_osd_req_callback; | |
1525 | osd_req->r_priv = obj_request; | |
1526 | ||
1527 | osd_req->r_oid_len = strlen(obj_request->object_name); | |
1528 | rbd_assert(osd_req->r_oid_len < sizeof (osd_req->r_oid)); | |
1529 | memcpy(osd_req->r_oid, obj_request->object_name, osd_req->r_oid_len); | |
1530 | ||
1531 | osd_req->r_file_layout = rbd_dev->layout; /* struct */ | |
1532 | ||
bf0d5f50 AE |
1533 | return osd_req; |
1534 | } | |
1535 | ||
1536 | static void rbd_osd_req_destroy(struct ceph_osd_request *osd_req) | |
1537 | { | |
1538 | ceph_osdc_put_request(osd_req); | |
1539 | } | |
1540 | ||
1541 | /* object_name is assumed to be a non-null pointer and NUL-terminated */ | |
1542 | ||
1543 | static struct rbd_obj_request *rbd_obj_request_create(const char *object_name, | |
1544 | u64 offset, u64 length, | |
1545 | enum obj_request_type type) | |
1546 | { | |
1547 | struct rbd_obj_request *obj_request; | |
1548 | size_t size; | |
1549 | char *name; | |
1550 | ||
1551 | rbd_assert(obj_request_type_valid(type)); | |
1552 | ||
1553 | size = strlen(object_name) + 1; | |
1554 | obj_request = kzalloc(sizeof (*obj_request) + size, GFP_KERNEL); | |
1555 | if (!obj_request) | |
1556 | return NULL; | |
1557 | ||
1558 | name = (char *)(obj_request + 1); | |
1559 | obj_request->object_name = memcpy(name, object_name, size); | |
1560 | obj_request->offset = offset; | |
1561 | obj_request->length = length; | |
926f9b3f | 1562 | obj_request->flags = 0; |
bf0d5f50 AE |
1563 | obj_request->which = BAD_WHICH; |
1564 | obj_request->type = type; | |
1565 | INIT_LIST_HEAD(&obj_request->links); | |
788e2df3 | 1566 | init_completion(&obj_request->completion); |
bf0d5f50 AE |
1567 | kref_init(&obj_request->kref); |
1568 | ||
37206ee5 AE |
1569 | dout("%s: \"%s\" %llu/%llu %d -> obj %p\n", __func__, object_name, |
1570 | offset, length, (int)type, obj_request); | |
1571 | ||
bf0d5f50 AE |
1572 | return obj_request; |
1573 | } | |
1574 | ||
1575 | static void rbd_obj_request_destroy(struct kref *kref) | |
1576 | { | |
1577 | struct rbd_obj_request *obj_request; | |
1578 | ||
1579 | obj_request = container_of(kref, struct rbd_obj_request, kref); | |
1580 | ||
37206ee5 AE |
1581 | dout("%s: obj %p\n", __func__, obj_request); |
1582 | ||
bf0d5f50 AE |
1583 | rbd_assert(obj_request->img_request == NULL); |
1584 | rbd_assert(obj_request->which == BAD_WHICH); | |
1585 | ||
1586 | if (obj_request->osd_req) | |
1587 | rbd_osd_req_destroy(obj_request->osd_req); | |
1588 | ||
1589 | rbd_assert(obj_request_type_valid(obj_request->type)); | |
1590 | switch (obj_request->type) { | |
9969ebc5 AE |
1591 | case OBJ_REQUEST_NODATA: |
1592 | break; /* Nothing to do */ | |
bf0d5f50 AE |
1593 | case OBJ_REQUEST_BIO: |
1594 | if (obj_request->bio_list) | |
1595 | bio_chain_put(obj_request->bio_list); | |
1596 | break; | |
788e2df3 AE |
1597 | case OBJ_REQUEST_PAGES: |
1598 | if (obj_request->pages) | |
1599 | ceph_release_page_vector(obj_request->pages, | |
1600 | obj_request->page_count); | |
1601 | break; | |
bf0d5f50 AE |
1602 | } |
1603 | ||
1604 | kfree(obj_request); | |
1605 | } | |
1606 | ||
1607 | /* | |
1608 | * Caller is responsible for filling in the list of object requests | |
1609 | * that comprises the image request, and the Linux request pointer | |
1610 | * (if there is one). | |
1611 | */ | |
cc344fa1 AE |
1612 | static struct rbd_img_request *rbd_img_request_create( |
1613 | struct rbd_device *rbd_dev, | |
bf0d5f50 | 1614 | u64 offset, u64 length, |
9849e986 AE |
1615 | bool write_request, |
1616 | bool child_request) | |
bf0d5f50 AE |
1617 | { |
1618 | struct rbd_img_request *img_request; | |
1619 | struct ceph_snap_context *snapc = NULL; | |
1620 | ||
1621 | img_request = kmalloc(sizeof (*img_request), GFP_ATOMIC); | |
1622 | if (!img_request) | |
1623 | return NULL; | |
1624 | ||
1625 | if (write_request) { | |
1626 | down_read(&rbd_dev->header_rwsem); | |
1627 | snapc = ceph_get_snap_context(rbd_dev->header.snapc); | |
1628 | up_read(&rbd_dev->header_rwsem); | |
1629 | if (WARN_ON(!snapc)) { | |
1630 | kfree(img_request); | |
1631 | return NULL; /* Shouldn't happen */ | |
1632 | } | |
0c425248 | 1633 | |
bf0d5f50 AE |
1634 | } |
1635 | ||
1636 | img_request->rq = NULL; | |
1637 | img_request->rbd_dev = rbd_dev; | |
1638 | img_request->offset = offset; | |
1639 | img_request->length = length; | |
0c425248 AE |
1640 | img_request->flags = 0; |
1641 | if (write_request) { | |
1642 | img_request_write_set(img_request); | |
bf0d5f50 | 1643 | img_request->snapc = snapc; |
0c425248 | 1644 | } else { |
bf0d5f50 | 1645 | img_request->snap_id = rbd_dev->spec->snap_id; |
0c425248 | 1646 | } |
9849e986 AE |
1647 | if (child_request) |
1648 | img_request_child_set(img_request); | |
d0b2e944 AE |
1649 | if (rbd_dev->parent_spec) |
1650 | img_request_layered_set(img_request); | |
bf0d5f50 AE |
1651 | spin_lock_init(&img_request->completion_lock); |
1652 | img_request->next_completion = 0; | |
1653 | img_request->callback = NULL; | |
a5a337d4 | 1654 | img_request->result = 0; |
bf0d5f50 AE |
1655 | img_request->obj_request_count = 0; |
1656 | INIT_LIST_HEAD(&img_request->obj_requests); | |
1657 | kref_init(&img_request->kref); | |
1658 | ||
5679c59f AE |
1659 | (void) obj_request_existence_set; |
1660 | (void) obj_request_known_test; | |
1661 | (void) obj_request_exists_test; | |
1662 | ||
bf0d5f50 AE |
1663 | rbd_img_request_get(img_request); /* Avoid a warning */ |
1664 | rbd_img_request_put(img_request); /* TEMPORARY */ | |
1665 | ||
37206ee5 AE |
1666 | dout("%s: rbd_dev %p %s %llu/%llu -> img %p\n", __func__, rbd_dev, |
1667 | write_request ? "write" : "read", offset, length, | |
1668 | img_request); | |
1669 | ||
bf0d5f50 AE |
1670 | return img_request; |
1671 | } | |
1672 | ||
1673 | static void rbd_img_request_destroy(struct kref *kref) | |
1674 | { | |
1675 | struct rbd_img_request *img_request; | |
1676 | struct rbd_obj_request *obj_request; | |
1677 | struct rbd_obj_request *next_obj_request; | |
1678 | ||
1679 | img_request = container_of(kref, struct rbd_img_request, kref); | |
1680 | ||
37206ee5 AE |
1681 | dout("%s: img %p\n", __func__, img_request); |
1682 | ||
bf0d5f50 AE |
1683 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) |
1684 | rbd_img_obj_request_del(img_request, obj_request); | |
25dcf954 | 1685 | rbd_assert(img_request->obj_request_count == 0); |
bf0d5f50 | 1686 | |
0c425248 | 1687 | if (img_request_write_test(img_request)) |
bf0d5f50 AE |
1688 | ceph_put_snap_context(img_request->snapc); |
1689 | ||
8b3e1a56 AE |
1690 | if (img_request_child_test(img_request)) |
1691 | rbd_obj_request_put(img_request->obj_request); | |
1692 | ||
bf0d5f50 AE |
1693 | kfree(img_request); |
1694 | } | |
1695 | ||
1217857f AE |
1696 | static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request) |
1697 | { | |
6365d33a | 1698 | struct rbd_img_request *img_request; |
1217857f AE |
1699 | unsigned int xferred; |
1700 | int result; | |
8b3e1a56 | 1701 | bool more; |
1217857f | 1702 | |
6365d33a AE |
1703 | rbd_assert(obj_request_img_data_test(obj_request)); |
1704 | img_request = obj_request->img_request; | |
1705 | ||
1217857f AE |
1706 | rbd_assert(obj_request->xferred <= (u64)UINT_MAX); |
1707 | xferred = (unsigned int)obj_request->xferred; | |
1708 | result = obj_request->result; | |
1709 | if (result) { | |
1710 | struct rbd_device *rbd_dev = img_request->rbd_dev; | |
1711 | ||
1712 | rbd_warn(rbd_dev, "%s %llx at %llx (%llx)\n", | |
1713 | img_request_write_test(img_request) ? "write" : "read", | |
1714 | obj_request->length, obj_request->img_offset, | |
1715 | obj_request->offset); | |
1716 | rbd_warn(rbd_dev, " result %d xferred %x\n", | |
1717 | result, xferred); | |
1718 | if (!img_request->result) | |
1719 | img_request->result = result; | |
1720 | } | |
1721 | ||
8b3e1a56 AE |
1722 | if (img_request_child_test(img_request)) { |
1723 | rbd_assert(img_request->obj_request != NULL); | |
1724 | more = obj_request->which < img_request->obj_request_count - 1; | |
1725 | } else { | |
1726 | rbd_assert(img_request->rq != NULL); | |
1727 | more = blk_end_request(img_request->rq, result, xferred); | |
1728 | } | |
1729 | ||
1730 | return more; | |
1217857f AE |
1731 | } |
1732 | ||
2169238d AE |
1733 | static void rbd_img_obj_callback(struct rbd_obj_request *obj_request) |
1734 | { | |
1735 | struct rbd_img_request *img_request; | |
1736 | u32 which = obj_request->which; | |
1737 | bool more = true; | |
1738 | ||
6365d33a | 1739 | rbd_assert(obj_request_img_data_test(obj_request)); |
2169238d AE |
1740 | img_request = obj_request->img_request; |
1741 | ||
1742 | dout("%s: img %p obj %p\n", __func__, img_request, obj_request); | |
1743 | rbd_assert(img_request != NULL); | |
2169238d AE |
1744 | rbd_assert(img_request->obj_request_count > 0); |
1745 | rbd_assert(which != BAD_WHICH); | |
1746 | rbd_assert(which < img_request->obj_request_count); | |
1747 | rbd_assert(which >= img_request->next_completion); | |
1748 | ||
1749 | spin_lock_irq(&img_request->completion_lock); | |
1750 | if (which != img_request->next_completion) | |
1751 | goto out; | |
1752 | ||
1753 | for_each_obj_request_from(img_request, obj_request) { | |
2169238d AE |
1754 | rbd_assert(more); |
1755 | rbd_assert(which < img_request->obj_request_count); | |
1756 | ||
1757 | if (!obj_request_done_test(obj_request)) | |
1758 | break; | |
1217857f | 1759 | more = rbd_img_obj_end_request(obj_request); |
2169238d AE |
1760 | which++; |
1761 | } | |
1762 | ||
1763 | rbd_assert(more ^ (which == img_request->obj_request_count)); | |
1764 | img_request->next_completion = which; | |
1765 | out: | |
1766 | spin_unlock_irq(&img_request->completion_lock); | |
1767 | ||
1768 | if (!more) | |
1769 | rbd_img_request_complete(img_request); | |
1770 | } | |
1771 | ||
bf0d5f50 AE |
1772 | static int rbd_img_request_fill_bio(struct rbd_img_request *img_request, |
1773 | struct bio *bio_list) | |
1774 | { | |
1775 | struct rbd_device *rbd_dev = img_request->rbd_dev; | |
1776 | struct rbd_obj_request *obj_request = NULL; | |
1777 | struct rbd_obj_request *next_obj_request; | |
0c425248 | 1778 | bool write_request = img_request_write_test(img_request); |
bf0d5f50 | 1779 | unsigned int bio_offset; |
7da22d29 | 1780 | u64 img_offset; |
bf0d5f50 AE |
1781 | u64 resid; |
1782 | u16 opcode; | |
1783 | ||
37206ee5 AE |
1784 | dout("%s: img %p bio %p\n", __func__, img_request, bio_list); |
1785 | ||
430c28c3 | 1786 | opcode = write_request ? CEPH_OSD_OP_WRITE : CEPH_OSD_OP_READ; |
bf0d5f50 | 1787 | bio_offset = 0; |
7da22d29 AE |
1788 | img_offset = img_request->offset; |
1789 | rbd_assert(img_offset == bio_list->bi_sector << SECTOR_SHIFT); | |
bf0d5f50 | 1790 | resid = img_request->length; |
4dda41d3 | 1791 | rbd_assert(resid > 0); |
bf0d5f50 | 1792 | while (resid) { |
2fa12320 | 1793 | struct ceph_osd_request *osd_req; |
bf0d5f50 AE |
1794 | const char *object_name; |
1795 | unsigned int clone_size; | |
bf0d5f50 AE |
1796 | u64 offset; |
1797 | u64 length; | |
1798 | ||
7da22d29 | 1799 | object_name = rbd_segment_name(rbd_dev, img_offset); |
bf0d5f50 AE |
1800 | if (!object_name) |
1801 | goto out_unwind; | |
7da22d29 AE |
1802 | offset = rbd_segment_offset(rbd_dev, img_offset); |
1803 | length = rbd_segment_length(rbd_dev, img_offset, resid); | |
bf0d5f50 AE |
1804 | obj_request = rbd_obj_request_create(object_name, |
1805 | offset, length, | |
1806 | OBJ_REQUEST_BIO); | |
1807 | kfree(object_name); /* object request has its own copy */ | |
1808 | if (!obj_request) | |
1809 | goto out_unwind; | |
1810 | ||
1811 | rbd_assert(length <= (u64) UINT_MAX); | |
1812 | clone_size = (unsigned int) length; | |
1813 | obj_request->bio_list = bio_chain_clone_range(&bio_list, | |
1814 | &bio_offset, clone_size, | |
1815 | GFP_ATOMIC); | |
1816 | if (!obj_request->bio_list) | |
1817 | goto out_partial; | |
1818 | ||
2fa12320 AE |
1819 | osd_req = rbd_osd_req_create(rbd_dev, write_request, |
1820 | obj_request); | |
1821 | if (!osd_req) | |
bf0d5f50 | 1822 | goto out_partial; |
2fa12320 | 1823 | obj_request->osd_req = osd_req; |
2169238d | 1824 | obj_request->callback = rbd_img_obj_callback; |
430c28c3 | 1825 | |
2fa12320 AE |
1826 | osd_req_op_extent_init(osd_req, 0, opcode, offset, length, |
1827 | 0, 0); | |
406e2c9f | 1828 | osd_req_op_extent_osd_data_bio(osd_req, 0, |
a4ce40a9 | 1829 | obj_request->bio_list, obj_request->length); |
2fa12320 | 1830 | rbd_osd_req_format(obj_request, write_request); |
430c28c3 | 1831 | |
7da22d29 | 1832 | obj_request->img_offset = img_offset; |
bf0d5f50 AE |
1833 | rbd_img_obj_request_add(img_request, obj_request); |
1834 | ||
7da22d29 | 1835 | img_offset += length; |
bf0d5f50 AE |
1836 | resid -= length; |
1837 | } | |
1838 | ||
1839 | return 0; | |
1840 | ||
1841 | out_partial: | |
1842 | rbd_obj_request_put(obj_request); | |
1843 | out_unwind: | |
1844 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) | |
1845 | rbd_obj_request_put(obj_request); | |
1846 | ||
1847 | return -ENOMEM; | |
1848 | } | |
1849 | ||
bf0d5f50 AE |
1850 | static int rbd_img_request_submit(struct rbd_img_request *img_request) |
1851 | { | |
1852 | struct rbd_device *rbd_dev = img_request->rbd_dev; | |
1853 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; | |
1854 | struct rbd_obj_request *obj_request; | |
46faeed4 | 1855 | struct rbd_obj_request *next_obj_request; |
bf0d5f50 | 1856 | |
37206ee5 | 1857 | dout("%s: img %p\n", __func__, img_request); |
46faeed4 | 1858 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) { |
bf0d5f50 AE |
1859 | int ret; |
1860 | ||
bf0d5f50 AE |
1861 | ret = rbd_obj_request_submit(osdc, obj_request); |
1862 | if (ret) | |
1863 | return ret; | |
bf0d5f50 AE |
1864 | } |
1865 | ||
1866 | return 0; | |
1867 | } | |
8b3e1a56 AE |
1868 | |
1869 | static void rbd_img_parent_read_callback(struct rbd_img_request *img_request) | |
1870 | { | |
1871 | struct rbd_obj_request *obj_request; | |
1872 | ||
1873 | rbd_assert(img_request_child_test(img_request)); | |
1874 | ||
1875 | obj_request = img_request->obj_request; | |
1876 | rbd_assert(obj_request != NULL); | |
1877 | obj_request->result = img_request->result; | |
1878 | obj_request->xferred = img_request->xferred; | |
1879 | ||
1880 | rbd_img_obj_request_read_callback(obj_request); | |
1881 | rbd_obj_request_complete(obj_request); | |
1882 | } | |
1883 | ||
1884 | static void rbd_img_parent_read(struct rbd_obj_request *obj_request) | |
1885 | { | |
1886 | struct rbd_device *rbd_dev; | |
1887 | struct rbd_img_request *img_request; | |
1888 | int result; | |
1889 | ||
1890 | rbd_assert(obj_request_img_data_test(obj_request)); | |
1891 | rbd_assert(obj_request->img_request != NULL); | |
1892 | rbd_assert(obj_request->result == (s32) -ENOENT); | |
1893 | rbd_assert(obj_request->type == OBJ_REQUEST_BIO); | |
1894 | ||
1895 | rbd_dev = obj_request->img_request->rbd_dev; | |
1896 | rbd_assert(rbd_dev->parent != NULL); | |
1897 | /* rbd_read_finish(obj_request, obj_request->length); */ | |
1898 | img_request = rbd_img_request_create(rbd_dev->parent, | |
1899 | obj_request->img_offset, | |
1900 | obj_request->length, | |
1901 | false, true); | |
1902 | result = -ENOMEM; | |
1903 | if (!img_request) | |
1904 | goto out_err; | |
1905 | ||
1906 | rbd_obj_request_get(obj_request); | |
1907 | img_request->obj_request = obj_request; | |
1908 | ||
1909 | result = rbd_img_request_fill_bio(img_request, obj_request->bio_list); | |
1910 | if (result) | |
1911 | goto out_err; | |
1912 | ||
1913 | img_request->callback = rbd_img_parent_read_callback; | |
1914 | result = rbd_img_request_submit(img_request); | |
1915 | if (result) | |
1916 | goto out_err; | |
1917 | ||
1918 | return; | |
1919 | out_err: | |
1920 | if (img_request) | |
1921 | rbd_img_request_put(img_request); | |
1922 | obj_request->result = result; | |
1923 | obj_request->xferred = 0; | |
1924 | obj_request_done_set(obj_request); | |
1925 | } | |
bf0d5f50 | 1926 | |
cf81b60e | 1927 | static int rbd_obj_notify_ack(struct rbd_device *rbd_dev, |
b8d70035 AE |
1928 | u64 ver, u64 notify_id) |
1929 | { | |
1930 | struct rbd_obj_request *obj_request; | |
2169238d | 1931 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
b8d70035 AE |
1932 | int ret; |
1933 | ||
1934 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, | |
1935 | OBJ_REQUEST_NODATA); | |
1936 | if (!obj_request) | |
1937 | return -ENOMEM; | |
1938 | ||
1939 | ret = -ENOMEM; | |
430c28c3 | 1940 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
b8d70035 AE |
1941 | if (!obj_request->osd_req) |
1942 | goto out; | |
2169238d | 1943 | obj_request->callback = rbd_obj_request_put; |
b8d70035 | 1944 | |
c99d2d4a AE |
1945 | osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_NOTIFY_ACK, |
1946 | notify_id, ver, 0); | |
2fa12320 | 1947 | rbd_osd_req_format(obj_request, false); |
430c28c3 | 1948 | |
b8d70035 | 1949 | ret = rbd_obj_request_submit(osdc, obj_request); |
b8d70035 | 1950 | out: |
cf81b60e AE |
1951 | if (ret) |
1952 | rbd_obj_request_put(obj_request); | |
b8d70035 AE |
1953 | |
1954 | return ret; | |
1955 | } | |
1956 | ||
1957 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) | |
1958 | { | |
1959 | struct rbd_device *rbd_dev = (struct rbd_device *)data; | |
1960 | u64 hver; | |
1961 | int rc; | |
1962 | ||
1963 | if (!rbd_dev) | |
1964 | return; | |
1965 | ||
37206ee5 | 1966 | dout("%s: \"%s\" notify_id %llu opcode %u\n", __func__, |
b8d70035 AE |
1967 | rbd_dev->header_name, (unsigned long long) notify_id, |
1968 | (unsigned int) opcode); | |
1969 | rc = rbd_dev_refresh(rbd_dev, &hver); | |
1970 | if (rc) | |
1971 | rbd_warn(rbd_dev, "got notification but failed to " | |
1972 | " update snaps: %d\n", rc); | |
1973 | ||
cf81b60e | 1974 | rbd_obj_notify_ack(rbd_dev, hver, notify_id); |
b8d70035 AE |
1975 | } |
1976 | ||
9969ebc5 AE |
1977 | /* |
1978 | * Request sync osd watch/unwatch. The value of "start" determines | |
1979 | * whether a watch request is being initiated or torn down. | |
1980 | */ | |
1981 | static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev, int start) | |
1982 | { | |
1983 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; | |
1984 | struct rbd_obj_request *obj_request; | |
9969ebc5 AE |
1985 | int ret; |
1986 | ||
1987 | rbd_assert(start ^ !!rbd_dev->watch_event); | |
1988 | rbd_assert(start ^ !!rbd_dev->watch_request); | |
1989 | ||
1990 | if (start) { | |
3c663bbd | 1991 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, rbd_dev, |
9969ebc5 AE |
1992 | &rbd_dev->watch_event); |
1993 | if (ret < 0) | |
1994 | return ret; | |
8eb87565 | 1995 | rbd_assert(rbd_dev->watch_event != NULL); |
9969ebc5 AE |
1996 | } |
1997 | ||
1998 | ret = -ENOMEM; | |
1999 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, | |
2000 | OBJ_REQUEST_NODATA); | |
2001 | if (!obj_request) | |
2002 | goto out_cancel; | |
2003 | ||
430c28c3 AE |
2004 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, true, obj_request); |
2005 | if (!obj_request->osd_req) | |
2006 | goto out_cancel; | |
2007 | ||
8eb87565 | 2008 | if (start) |
975241af | 2009 | ceph_osdc_set_request_linger(osdc, obj_request->osd_req); |
8eb87565 | 2010 | else |
6977c3f9 | 2011 | ceph_osdc_unregister_linger_request(osdc, |
975241af | 2012 | rbd_dev->watch_request->osd_req); |
2169238d AE |
2013 | |
2014 | osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_WATCH, | |
2015 | rbd_dev->watch_event->cookie, | |
2016 | rbd_dev->header.obj_version, start); | |
2017 | rbd_osd_req_format(obj_request, true); | |
2018 | ||
9969ebc5 AE |
2019 | ret = rbd_obj_request_submit(osdc, obj_request); |
2020 | if (ret) | |
2021 | goto out_cancel; | |
2022 | ret = rbd_obj_request_wait(obj_request); | |
2023 | if (ret) | |
2024 | goto out_cancel; | |
9969ebc5 AE |
2025 | ret = obj_request->result; |
2026 | if (ret) | |
2027 | goto out_cancel; | |
2028 | ||
8eb87565 AE |
2029 | /* |
2030 | * A watch request is set to linger, so the underlying osd | |
2031 | * request won't go away until we unregister it. We retain | |
2032 | * a pointer to the object request during that time (in | |
2033 | * rbd_dev->watch_request), so we'll keep a reference to | |
2034 | * it. We'll drop that reference (below) after we've | |
2035 | * unregistered it. | |
2036 | */ | |
2037 | if (start) { | |
2038 | rbd_dev->watch_request = obj_request; | |
2039 | ||
2040 | return 0; | |
2041 | } | |
2042 | ||
2043 | /* We have successfully torn down the watch request */ | |
2044 | ||
2045 | rbd_obj_request_put(rbd_dev->watch_request); | |
2046 | rbd_dev->watch_request = NULL; | |
9969ebc5 AE |
2047 | out_cancel: |
2048 | /* Cancel the event if we're tearing down, or on error */ | |
2049 | ceph_osdc_cancel_event(rbd_dev->watch_event); | |
2050 | rbd_dev->watch_event = NULL; | |
9969ebc5 AE |
2051 | if (obj_request) |
2052 | rbd_obj_request_put(obj_request); | |
2053 | ||
2054 | return ret; | |
2055 | } | |
2056 | ||
36be9a76 AE |
2057 | /* |
2058 | * Synchronous osd object method call | |
2059 | */ | |
2060 | static int rbd_obj_method_sync(struct rbd_device *rbd_dev, | |
2061 | const char *object_name, | |
2062 | const char *class_name, | |
2063 | const char *method_name, | |
2064 | const char *outbound, | |
2065 | size_t outbound_size, | |
2066 | char *inbound, | |
2067 | size_t inbound_size, | |
2068 | u64 *version) | |
2069 | { | |
2169238d | 2070 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
36be9a76 | 2071 | struct rbd_obj_request *obj_request; |
36be9a76 AE |
2072 | struct page **pages; |
2073 | u32 page_count; | |
2074 | int ret; | |
2075 | ||
2076 | /* | |
6010a451 AE |
2077 | * Method calls are ultimately read operations. The result |
2078 | * should placed into the inbound buffer provided. They | |
2079 | * also supply outbound data--parameters for the object | |
2080 | * method. Currently if this is present it will be a | |
2081 | * snapshot id. | |
36be9a76 AE |
2082 | */ |
2083 | page_count = (u32) calc_pages_for(0, inbound_size); | |
2084 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); | |
2085 | if (IS_ERR(pages)) | |
2086 | return PTR_ERR(pages); | |
2087 | ||
2088 | ret = -ENOMEM; | |
6010a451 | 2089 | obj_request = rbd_obj_request_create(object_name, 0, inbound_size, |
36be9a76 AE |
2090 | OBJ_REQUEST_PAGES); |
2091 | if (!obj_request) | |
2092 | goto out; | |
2093 | ||
2094 | obj_request->pages = pages; | |
2095 | obj_request->page_count = page_count; | |
2096 | ||
430c28c3 | 2097 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
36be9a76 AE |
2098 | if (!obj_request->osd_req) |
2099 | goto out; | |
2100 | ||
c99d2d4a | 2101 | osd_req_op_cls_init(obj_request->osd_req, 0, CEPH_OSD_OP_CALL, |
04017e29 AE |
2102 | class_name, method_name); |
2103 | if (outbound_size) { | |
2104 | struct ceph_pagelist *pagelist; | |
2105 | ||
2106 | pagelist = kmalloc(sizeof (*pagelist), GFP_NOFS); | |
2107 | if (!pagelist) | |
2108 | goto out; | |
2109 | ||
2110 | ceph_pagelist_init(pagelist); | |
2111 | ceph_pagelist_append(pagelist, outbound, outbound_size); | |
2112 | osd_req_op_cls_request_data_pagelist(obj_request->osd_req, 0, | |
2113 | pagelist); | |
2114 | } | |
a4ce40a9 AE |
2115 | osd_req_op_cls_response_data_pages(obj_request->osd_req, 0, |
2116 | obj_request->pages, inbound_size, | |
44cd188d | 2117 | 0, false, false); |
2fa12320 | 2118 | rbd_osd_req_format(obj_request, false); |
430c28c3 | 2119 | |
36be9a76 AE |
2120 | ret = rbd_obj_request_submit(osdc, obj_request); |
2121 | if (ret) | |
2122 | goto out; | |
2123 | ret = rbd_obj_request_wait(obj_request); | |
2124 | if (ret) | |
2125 | goto out; | |
2126 | ||
2127 | ret = obj_request->result; | |
2128 | if (ret < 0) | |
2129 | goto out; | |
23ed6e13 | 2130 | ret = 0; |
903bb32e | 2131 | ceph_copy_from_page_vector(pages, inbound, 0, obj_request->xferred); |
36be9a76 AE |
2132 | if (version) |
2133 | *version = obj_request->version; | |
2134 | out: | |
2135 | if (obj_request) | |
2136 | rbd_obj_request_put(obj_request); | |
2137 | else | |
2138 | ceph_release_page_vector(pages, page_count); | |
2139 | ||
2140 | return ret; | |
2141 | } | |
2142 | ||
bf0d5f50 | 2143 | static void rbd_request_fn(struct request_queue *q) |
cc344fa1 | 2144 | __releases(q->queue_lock) __acquires(q->queue_lock) |
bf0d5f50 AE |
2145 | { |
2146 | struct rbd_device *rbd_dev = q->queuedata; | |
2147 | bool read_only = rbd_dev->mapping.read_only; | |
2148 | struct request *rq; | |
2149 | int result; | |
2150 | ||
2151 | while ((rq = blk_fetch_request(q))) { | |
2152 | bool write_request = rq_data_dir(rq) == WRITE; | |
2153 | struct rbd_img_request *img_request; | |
2154 | u64 offset; | |
2155 | u64 length; | |
2156 | ||
2157 | /* Ignore any non-FS requests that filter through. */ | |
2158 | ||
2159 | if (rq->cmd_type != REQ_TYPE_FS) { | |
4dda41d3 AE |
2160 | dout("%s: non-fs request type %d\n", __func__, |
2161 | (int) rq->cmd_type); | |
2162 | __blk_end_request_all(rq, 0); | |
2163 | continue; | |
2164 | } | |
2165 | ||
2166 | /* Ignore/skip any zero-length requests */ | |
2167 | ||
2168 | offset = (u64) blk_rq_pos(rq) << SECTOR_SHIFT; | |
2169 | length = (u64) blk_rq_bytes(rq); | |
2170 | ||
2171 | if (!length) { | |
2172 | dout("%s: zero-length request\n", __func__); | |
bf0d5f50 AE |
2173 | __blk_end_request_all(rq, 0); |
2174 | continue; | |
2175 | } | |
2176 | ||
2177 | spin_unlock_irq(q->queue_lock); | |
2178 | ||
2179 | /* Disallow writes to a read-only device */ | |
2180 | ||
2181 | if (write_request) { | |
2182 | result = -EROFS; | |
2183 | if (read_only) | |
2184 | goto end_request; | |
2185 | rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP); | |
2186 | } | |
2187 | ||
6d292906 AE |
2188 | /* |
2189 | * Quit early if the mapped snapshot no longer | |
2190 | * exists. It's still possible the snapshot will | |
2191 | * have disappeared by the time our request arrives | |
2192 | * at the osd, but there's no sense in sending it if | |
2193 | * we already know. | |
2194 | */ | |
2195 | if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) { | |
bf0d5f50 AE |
2196 | dout("request for non-existent snapshot"); |
2197 | rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); | |
2198 | result = -ENXIO; | |
2199 | goto end_request; | |
2200 | } | |
2201 | ||
bf0d5f50 AE |
2202 | result = -EINVAL; |
2203 | if (WARN_ON(offset && length > U64_MAX - offset + 1)) | |
2204 | goto end_request; /* Shouldn't happen */ | |
2205 | ||
2206 | result = -ENOMEM; | |
2207 | img_request = rbd_img_request_create(rbd_dev, offset, length, | |
9849e986 | 2208 | write_request, false); |
bf0d5f50 AE |
2209 | if (!img_request) |
2210 | goto end_request; | |
2211 | ||
2212 | img_request->rq = rq; | |
2213 | ||
2214 | result = rbd_img_request_fill_bio(img_request, rq->bio); | |
2215 | if (!result) | |
2216 | result = rbd_img_request_submit(img_request); | |
2217 | if (result) | |
2218 | rbd_img_request_put(img_request); | |
2219 | end_request: | |
2220 | spin_lock_irq(q->queue_lock); | |
2221 | if (result < 0) { | |
7da22d29 AE |
2222 | rbd_warn(rbd_dev, "%s %llx at %llx result %d\n", |
2223 | write_request ? "write" : "read", | |
2224 | length, offset, result); | |
2225 | ||
bf0d5f50 AE |
2226 | __blk_end_request_all(rq, result); |
2227 | } | |
2228 | } | |
2229 | } | |
2230 | ||
602adf40 YS |
2231 | /* |
2232 | * a queue callback. Makes sure that we don't create a bio that spans across | |
2233 | * multiple osd objects. One exception would be with a single page bios, | |
f7760dad | 2234 | * which we handle later at bio_chain_clone_range() |
602adf40 YS |
2235 | */ |
2236 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, | |
2237 | struct bio_vec *bvec) | |
2238 | { | |
2239 | struct rbd_device *rbd_dev = q->queuedata; | |
e5cfeed2 AE |
2240 | sector_t sector_offset; |
2241 | sector_t sectors_per_obj; | |
2242 | sector_t obj_sector_offset; | |
2243 | int ret; | |
2244 | ||
2245 | /* | |
2246 | * Find how far into its rbd object the partition-relative | |
2247 | * bio start sector is to offset relative to the enclosing | |
2248 | * device. | |
2249 | */ | |
2250 | sector_offset = get_start_sect(bmd->bi_bdev) + bmd->bi_sector; | |
2251 | sectors_per_obj = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); | |
2252 | obj_sector_offset = sector_offset & (sectors_per_obj - 1); | |
2253 | ||
2254 | /* | |
2255 | * Compute the number of bytes from that offset to the end | |
2256 | * of the object. Account for what's already used by the bio. | |
2257 | */ | |
2258 | ret = (int) (sectors_per_obj - obj_sector_offset) << SECTOR_SHIFT; | |
2259 | if (ret > bmd->bi_size) | |
2260 | ret -= bmd->bi_size; | |
2261 | else | |
2262 | ret = 0; | |
2263 | ||
2264 | /* | |
2265 | * Don't send back more than was asked for. And if the bio | |
2266 | * was empty, let the whole thing through because: "Note | |
2267 | * that a block device *must* allow a single page to be | |
2268 | * added to an empty bio." | |
2269 | */ | |
2270 | rbd_assert(bvec->bv_len <= PAGE_SIZE); | |
2271 | if (ret > (int) bvec->bv_len || !bmd->bi_size) | |
2272 | ret = (int) bvec->bv_len; | |
2273 | ||
2274 | return ret; | |
602adf40 YS |
2275 | } |
2276 | ||
2277 | static void rbd_free_disk(struct rbd_device *rbd_dev) | |
2278 | { | |
2279 | struct gendisk *disk = rbd_dev->disk; | |
2280 | ||
2281 | if (!disk) | |
2282 | return; | |
2283 | ||
602adf40 YS |
2284 | if (disk->flags & GENHD_FL_UP) |
2285 | del_gendisk(disk); | |
2286 | if (disk->queue) | |
2287 | blk_cleanup_queue(disk->queue); | |
2288 | put_disk(disk); | |
2289 | } | |
2290 | ||
788e2df3 AE |
2291 | static int rbd_obj_read_sync(struct rbd_device *rbd_dev, |
2292 | const char *object_name, | |
2293 | u64 offset, u64 length, | |
2294 | char *buf, u64 *version) | |
2295 | ||
2296 | { | |
2169238d | 2297 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
788e2df3 | 2298 | struct rbd_obj_request *obj_request; |
788e2df3 AE |
2299 | struct page **pages = NULL; |
2300 | u32 page_count; | |
1ceae7ef | 2301 | size_t size; |
788e2df3 AE |
2302 | int ret; |
2303 | ||
2304 | page_count = (u32) calc_pages_for(offset, length); | |
2305 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); | |
2306 | if (IS_ERR(pages)) | |
2307 | ret = PTR_ERR(pages); | |
2308 | ||
2309 | ret = -ENOMEM; | |
2310 | obj_request = rbd_obj_request_create(object_name, offset, length, | |
36be9a76 | 2311 | OBJ_REQUEST_PAGES); |
788e2df3 AE |
2312 | if (!obj_request) |
2313 | goto out; | |
2314 | ||
2315 | obj_request->pages = pages; | |
2316 | obj_request->page_count = page_count; | |
2317 | ||
430c28c3 | 2318 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
788e2df3 AE |
2319 | if (!obj_request->osd_req) |
2320 | goto out; | |
2321 | ||
c99d2d4a AE |
2322 | osd_req_op_extent_init(obj_request->osd_req, 0, CEPH_OSD_OP_READ, |
2323 | offset, length, 0, 0); | |
406e2c9f | 2324 | osd_req_op_extent_osd_data_pages(obj_request->osd_req, 0, |
a4ce40a9 | 2325 | obj_request->pages, |
44cd188d AE |
2326 | obj_request->length, |
2327 | obj_request->offset & ~PAGE_MASK, | |
2328 | false, false); | |
2fa12320 | 2329 | rbd_osd_req_format(obj_request, false); |
430c28c3 | 2330 | |
788e2df3 AE |
2331 | ret = rbd_obj_request_submit(osdc, obj_request); |
2332 | if (ret) | |
2333 | goto out; | |
2334 | ret = rbd_obj_request_wait(obj_request); | |
2335 | if (ret) | |
2336 | goto out; | |
2337 | ||
2338 | ret = obj_request->result; | |
2339 | if (ret < 0) | |
2340 | goto out; | |
1ceae7ef AE |
2341 | |
2342 | rbd_assert(obj_request->xferred <= (u64) SIZE_MAX); | |
2343 | size = (size_t) obj_request->xferred; | |
903bb32e | 2344 | ceph_copy_from_page_vector(pages, buf, 0, size); |
23ed6e13 AE |
2345 | rbd_assert(size <= (size_t) INT_MAX); |
2346 | ret = (int) size; | |
788e2df3 AE |
2347 | if (version) |
2348 | *version = obj_request->version; | |
2349 | out: | |
2350 | if (obj_request) | |
2351 | rbd_obj_request_put(obj_request); | |
2352 | else | |
2353 | ceph_release_page_vector(pages, page_count); | |
2354 | ||
2355 | return ret; | |
2356 | } | |
2357 | ||
602adf40 | 2358 | /* |
4156d998 AE |
2359 | * Read the complete header for the given rbd device. |
2360 | * | |
2361 | * Returns a pointer to a dynamically-allocated buffer containing | |
2362 | * the complete and validated header. Caller can pass the address | |
2363 | * of a variable that will be filled in with the version of the | |
2364 | * header object at the time it was read. | |
2365 | * | |
2366 | * Returns a pointer-coded errno if a failure occurs. | |
602adf40 | 2367 | */ |
4156d998 AE |
2368 | static struct rbd_image_header_ondisk * |
2369 | rbd_dev_v1_header_read(struct rbd_device *rbd_dev, u64 *version) | |
602adf40 | 2370 | { |
4156d998 | 2371 | struct rbd_image_header_ondisk *ondisk = NULL; |
50f7c4c9 | 2372 | u32 snap_count = 0; |
4156d998 AE |
2373 | u64 names_size = 0; |
2374 | u32 want_count; | |
2375 | int ret; | |
602adf40 | 2376 | |
00f1f36f | 2377 | /* |
4156d998 AE |
2378 | * The complete header will include an array of its 64-bit |
2379 | * snapshot ids, followed by the names of those snapshots as | |
2380 | * a contiguous block of NUL-terminated strings. Note that | |
2381 | * the number of snapshots could change by the time we read | |
2382 | * it in, in which case we re-read it. | |
00f1f36f | 2383 | */ |
4156d998 AE |
2384 | do { |
2385 | size_t size; | |
2386 | ||
2387 | kfree(ondisk); | |
2388 | ||
2389 | size = sizeof (*ondisk); | |
2390 | size += snap_count * sizeof (struct rbd_image_snap_ondisk); | |
2391 | size += names_size; | |
2392 | ondisk = kmalloc(size, GFP_KERNEL); | |
2393 | if (!ondisk) | |
2394 | return ERR_PTR(-ENOMEM); | |
2395 | ||
788e2df3 | 2396 | ret = rbd_obj_read_sync(rbd_dev, rbd_dev->header_name, |
4156d998 AE |
2397 | 0, size, |
2398 | (char *) ondisk, version); | |
4156d998 AE |
2399 | if (ret < 0) |
2400 | goto out_err; | |
2401 | if (WARN_ON((size_t) ret < size)) { | |
2402 | ret = -ENXIO; | |
06ecc6cb AE |
2403 | rbd_warn(rbd_dev, "short header read (want %zd got %d)", |
2404 | size, ret); | |
4156d998 AE |
2405 | goto out_err; |
2406 | } | |
2407 | if (!rbd_dev_ondisk_valid(ondisk)) { | |
2408 | ret = -ENXIO; | |
06ecc6cb | 2409 | rbd_warn(rbd_dev, "invalid header"); |
4156d998 | 2410 | goto out_err; |
81e759fb | 2411 | } |
602adf40 | 2412 | |
4156d998 AE |
2413 | names_size = le64_to_cpu(ondisk->snap_names_len); |
2414 | want_count = snap_count; | |
2415 | snap_count = le32_to_cpu(ondisk->snap_count); | |
2416 | } while (snap_count != want_count); | |
00f1f36f | 2417 | |
4156d998 | 2418 | return ondisk; |
00f1f36f | 2419 | |
4156d998 AE |
2420 | out_err: |
2421 | kfree(ondisk); | |
2422 | ||
2423 | return ERR_PTR(ret); | |
2424 | } | |
2425 | ||
2426 | /* | |
2427 | * reload the ondisk the header | |
2428 | */ | |
2429 | static int rbd_read_header(struct rbd_device *rbd_dev, | |
2430 | struct rbd_image_header *header) | |
2431 | { | |
2432 | struct rbd_image_header_ondisk *ondisk; | |
2433 | u64 ver = 0; | |
2434 | int ret; | |
602adf40 | 2435 | |
4156d998 AE |
2436 | ondisk = rbd_dev_v1_header_read(rbd_dev, &ver); |
2437 | if (IS_ERR(ondisk)) | |
2438 | return PTR_ERR(ondisk); | |
2439 | ret = rbd_header_from_disk(header, ondisk); | |
2440 | if (ret >= 0) | |
2441 | header->obj_version = ver; | |
2442 | kfree(ondisk); | |
2443 | ||
2444 | return ret; | |
602adf40 YS |
2445 | } |
2446 | ||
41f38c2b | 2447 | static void rbd_remove_all_snaps(struct rbd_device *rbd_dev) |
dfc5606d YS |
2448 | { |
2449 | struct rbd_snap *snap; | |
a0593290 | 2450 | struct rbd_snap *next; |
dfc5606d | 2451 | |
a0593290 | 2452 | list_for_each_entry_safe(snap, next, &rbd_dev->snaps, node) |
41f38c2b | 2453 | rbd_remove_snap_dev(snap); |
dfc5606d YS |
2454 | } |
2455 | ||
9478554a AE |
2456 | static void rbd_update_mapping_size(struct rbd_device *rbd_dev) |
2457 | { | |
2458 | sector_t size; | |
2459 | ||
0d7dbfce | 2460 | if (rbd_dev->spec->snap_id != CEPH_NOSNAP) |
9478554a AE |
2461 | return; |
2462 | ||
2463 | size = (sector_t) rbd_dev->header.image_size / SECTOR_SIZE; | |
2464 | dout("setting size to %llu sectors", (unsigned long long) size); | |
2465 | rbd_dev->mapping.size = (u64) size; | |
2466 | set_capacity(rbd_dev->disk, size); | |
2467 | } | |
2468 | ||
602adf40 YS |
2469 | /* |
2470 | * only read the first part of the ondisk header, without the snaps info | |
2471 | */ | |
117973fb | 2472 | static int rbd_dev_v1_refresh(struct rbd_device *rbd_dev, u64 *hver) |
602adf40 YS |
2473 | { |
2474 | int ret; | |
2475 | struct rbd_image_header h; | |
602adf40 YS |
2476 | |
2477 | ret = rbd_read_header(rbd_dev, &h); | |
2478 | if (ret < 0) | |
2479 | return ret; | |
2480 | ||
a51aa0c0 JD |
2481 | down_write(&rbd_dev->header_rwsem); |
2482 | ||
9478554a AE |
2483 | /* Update image size, and check for resize of mapped image */ |
2484 | rbd_dev->header.image_size = h.image_size; | |
2485 | rbd_update_mapping_size(rbd_dev); | |
9db4b3e3 | 2486 | |
849b4260 | 2487 | /* rbd_dev->header.object_prefix shouldn't change */ |
602adf40 | 2488 | kfree(rbd_dev->header.snap_sizes); |
849b4260 | 2489 | kfree(rbd_dev->header.snap_names); |
d1d25646 JD |
2490 | /* osd requests may still refer to snapc */ |
2491 | ceph_put_snap_context(rbd_dev->header.snapc); | |
602adf40 | 2492 | |
b813623a AE |
2493 | if (hver) |
2494 | *hver = h.obj_version; | |
a71b891b | 2495 | rbd_dev->header.obj_version = h.obj_version; |
93a24e08 | 2496 | rbd_dev->header.image_size = h.image_size; |
602adf40 YS |
2497 | rbd_dev->header.snapc = h.snapc; |
2498 | rbd_dev->header.snap_names = h.snap_names; | |
2499 | rbd_dev->header.snap_sizes = h.snap_sizes; | |
849b4260 AE |
2500 | /* Free the extra copy of the object prefix */ |
2501 | WARN_ON(strcmp(rbd_dev->header.object_prefix, h.object_prefix)); | |
2502 | kfree(h.object_prefix); | |
2503 | ||
304f6808 AE |
2504 | ret = rbd_dev_snaps_update(rbd_dev); |
2505 | if (!ret) | |
2506 | ret = rbd_dev_snaps_register(rbd_dev); | |
dfc5606d | 2507 | |
c666601a | 2508 | up_write(&rbd_dev->header_rwsem); |
602adf40 | 2509 | |
dfc5606d | 2510 | return ret; |
602adf40 YS |
2511 | } |
2512 | ||
117973fb | 2513 | static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver) |
1fe5e993 AE |
2514 | { |
2515 | int ret; | |
2516 | ||
117973fb | 2517 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
1fe5e993 | 2518 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
117973fb AE |
2519 | if (rbd_dev->image_format == 1) |
2520 | ret = rbd_dev_v1_refresh(rbd_dev, hver); | |
2521 | else | |
2522 | ret = rbd_dev_v2_refresh(rbd_dev, hver); | |
1fe5e993 AE |
2523 | mutex_unlock(&ctl_mutex); |
2524 | ||
2525 | return ret; | |
2526 | } | |
2527 | ||
602adf40 YS |
2528 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
2529 | { | |
2530 | struct gendisk *disk; | |
2531 | struct request_queue *q; | |
593a9e7b | 2532 | u64 segment_size; |
602adf40 | 2533 | |
602adf40 | 2534 | /* create gendisk info */ |
602adf40 YS |
2535 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); |
2536 | if (!disk) | |
1fcdb8aa | 2537 | return -ENOMEM; |
602adf40 | 2538 | |
f0f8cef5 | 2539 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
de71a297 | 2540 | rbd_dev->dev_id); |
602adf40 YS |
2541 | disk->major = rbd_dev->major; |
2542 | disk->first_minor = 0; | |
2543 | disk->fops = &rbd_bd_ops; | |
2544 | disk->private_data = rbd_dev; | |
2545 | ||
bf0d5f50 | 2546 | q = blk_init_queue(rbd_request_fn, &rbd_dev->lock); |
602adf40 YS |
2547 | if (!q) |
2548 | goto out_disk; | |
029bcbd8 | 2549 | |
593a9e7b AE |
2550 | /* We use the default size, but let's be explicit about it. */ |
2551 | blk_queue_physical_block_size(q, SECTOR_SIZE); | |
2552 | ||
029bcbd8 | 2553 | /* set io sizes to object size */ |
593a9e7b AE |
2554 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
2555 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); | |
2556 | blk_queue_max_segment_size(q, segment_size); | |
2557 | blk_queue_io_min(q, segment_size); | |
2558 | blk_queue_io_opt(q, segment_size); | |
029bcbd8 | 2559 | |
602adf40 YS |
2560 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
2561 | disk->queue = q; | |
2562 | ||
2563 | q->queuedata = rbd_dev; | |
2564 | ||
2565 | rbd_dev->disk = disk; | |
602adf40 | 2566 | |
12f02944 AE |
2567 | set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); |
2568 | ||
602adf40 | 2569 | return 0; |
602adf40 YS |
2570 | out_disk: |
2571 | put_disk(disk); | |
1fcdb8aa AE |
2572 | |
2573 | return -ENOMEM; | |
602adf40 YS |
2574 | } |
2575 | ||
dfc5606d YS |
2576 | /* |
2577 | sysfs | |
2578 | */ | |
2579 | ||
593a9e7b AE |
2580 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
2581 | { | |
2582 | return container_of(dev, struct rbd_device, dev); | |
2583 | } | |
2584 | ||
dfc5606d YS |
2585 | static ssize_t rbd_size_show(struct device *dev, |
2586 | struct device_attribute *attr, char *buf) | |
2587 | { | |
593a9e7b | 2588 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
a51aa0c0 JD |
2589 | sector_t size; |
2590 | ||
2591 | down_read(&rbd_dev->header_rwsem); | |
2592 | size = get_capacity(rbd_dev->disk); | |
2593 | up_read(&rbd_dev->header_rwsem); | |
dfc5606d | 2594 | |
a51aa0c0 | 2595 | return sprintf(buf, "%llu\n", (unsigned long long) size * SECTOR_SIZE); |
dfc5606d YS |
2596 | } |
2597 | ||
34b13184 AE |
2598 | /* |
2599 | * Note this shows the features for whatever's mapped, which is not | |
2600 | * necessarily the base image. | |
2601 | */ | |
2602 | static ssize_t rbd_features_show(struct device *dev, | |
2603 | struct device_attribute *attr, char *buf) | |
2604 | { | |
2605 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
2606 | ||
2607 | return sprintf(buf, "0x%016llx\n", | |
2608 | (unsigned long long) rbd_dev->mapping.features); | |
2609 | } | |
2610 | ||
dfc5606d YS |
2611 | static ssize_t rbd_major_show(struct device *dev, |
2612 | struct device_attribute *attr, char *buf) | |
2613 | { | |
593a9e7b | 2614 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 2615 | |
dfc5606d YS |
2616 | return sprintf(buf, "%d\n", rbd_dev->major); |
2617 | } | |
2618 | ||
2619 | static ssize_t rbd_client_id_show(struct device *dev, | |
2620 | struct device_attribute *attr, char *buf) | |
602adf40 | 2621 | { |
593a9e7b | 2622 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 2623 | |
1dbb4399 AE |
2624 | return sprintf(buf, "client%lld\n", |
2625 | ceph_client_id(rbd_dev->rbd_client->client)); | |
602adf40 YS |
2626 | } |
2627 | ||
dfc5606d YS |
2628 | static ssize_t rbd_pool_show(struct device *dev, |
2629 | struct device_attribute *attr, char *buf) | |
602adf40 | 2630 | { |
593a9e7b | 2631 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 2632 | |
0d7dbfce | 2633 | return sprintf(buf, "%s\n", rbd_dev->spec->pool_name); |
dfc5606d YS |
2634 | } |
2635 | ||
9bb2f334 AE |
2636 | static ssize_t rbd_pool_id_show(struct device *dev, |
2637 | struct device_attribute *attr, char *buf) | |
2638 | { | |
2639 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
2640 | ||
0d7dbfce AE |
2641 | return sprintf(buf, "%llu\n", |
2642 | (unsigned long long) rbd_dev->spec->pool_id); | |
9bb2f334 AE |
2643 | } |
2644 | ||
dfc5606d YS |
2645 | static ssize_t rbd_name_show(struct device *dev, |
2646 | struct device_attribute *attr, char *buf) | |
2647 | { | |
593a9e7b | 2648 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 2649 | |
a92ffdf8 AE |
2650 | if (rbd_dev->spec->image_name) |
2651 | return sprintf(buf, "%s\n", rbd_dev->spec->image_name); | |
2652 | ||
2653 | return sprintf(buf, "(unknown)\n"); | |
dfc5606d YS |
2654 | } |
2655 | ||
589d30e0 AE |
2656 | static ssize_t rbd_image_id_show(struct device *dev, |
2657 | struct device_attribute *attr, char *buf) | |
2658 | { | |
2659 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
2660 | ||
0d7dbfce | 2661 | return sprintf(buf, "%s\n", rbd_dev->spec->image_id); |
589d30e0 AE |
2662 | } |
2663 | ||
34b13184 AE |
2664 | /* |
2665 | * Shows the name of the currently-mapped snapshot (or | |
2666 | * RBD_SNAP_HEAD_NAME for the base image). | |
2667 | */ | |
dfc5606d YS |
2668 | static ssize_t rbd_snap_show(struct device *dev, |
2669 | struct device_attribute *attr, | |
2670 | char *buf) | |
2671 | { | |
593a9e7b | 2672 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 2673 | |
0d7dbfce | 2674 | return sprintf(buf, "%s\n", rbd_dev->spec->snap_name); |
dfc5606d YS |
2675 | } |
2676 | ||
86b00e0d AE |
2677 | /* |
2678 | * For an rbd v2 image, shows the pool id, image id, and snapshot id | |
2679 | * for the parent image. If there is no parent, simply shows | |
2680 | * "(no parent image)". | |
2681 | */ | |
2682 | static ssize_t rbd_parent_show(struct device *dev, | |
2683 | struct device_attribute *attr, | |
2684 | char *buf) | |
2685 | { | |
2686 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
2687 | struct rbd_spec *spec = rbd_dev->parent_spec; | |
2688 | int count; | |
2689 | char *bufp = buf; | |
2690 | ||
2691 | if (!spec) | |
2692 | return sprintf(buf, "(no parent image)\n"); | |
2693 | ||
2694 | count = sprintf(bufp, "pool_id %llu\npool_name %s\n", | |
2695 | (unsigned long long) spec->pool_id, spec->pool_name); | |
2696 | if (count < 0) | |
2697 | return count; | |
2698 | bufp += count; | |
2699 | ||
2700 | count = sprintf(bufp, "image_id %s\nimage_name %s\n", spec->image_id, | |
2701 | spec->image_name ? spec->image_name : "(unknown)"); | |
2702 | if (count < 0) | |
2703 | return count; | |
2704 | bufp += count; | |
2705 | ||
2706 | count = sprintf(bufp, "snap_id %llu\nsnap_name %s\n", | |
2707 | (unsigned long long) spec->snap_id, spec->snap_name); | |
2708 | if (count < 0) | |
2709 | return count; | |
2710 | bufp += count; | |
2711 | ||
2712 | count = sprintf(bufp, "overlap %llu\n", rbd_dev->parent_overlap); | |
2713 | if (count < 0) | |
2714 | return count; | |
2715 | bufp += count; | |
2716 | ||
2717 | return (ssize_t) (bufp - buf); | |
2718 | } | |
2719 | ||
dfc5606d YS |
2720 | static ssize_t rbd_image_refresh(struct device *dev, |
2721 | struct device_attribute *attr, | |
2722 | const char *buf, | |
2723 | size_t size) | |
2724 | { | |
593a9e7b | 2725 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
b813623a | 2726 | int ret; |
602adf40 | 2727 | |
117973fb | 2728 | ret = rbd_dev_refresh(rbd_dev, NULL); |
b813623a AE |
2729 | |
2730 | return ret < 0 ? ret : size; | |
dfc5606d | 2731 | } |
602adf40 | 2732 | |
dfc5606d | 2733 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
34b13184 | 2734 | static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL); |
dfc5606d YS |
2735 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); |
2736 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); | |
2737 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); | |
9bb2f334 | 2738 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
dfc5606d | 2739 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
589d30e0 | 2740 | static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL); |
dfc5606d YS |
2741 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); |
2742 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); | |
86b00e0d | 2743 | static DEVICE_ATTR(parent, S_IRUGO, rbd_parent_show, NULL); |
dfc5606d YS |
2744 | |
2745 | static struct attribute *rbd_attrs[] = { | |
2746 | &dev_attr_size.attr, | |
34b13184 | 2747 | &dev_attr_features.attr, |
dfc5606d YS |
2748 | &dev_attr_major.attr, |
2749 | &dev_attr_client_id.attr, | |
2750 | &dev_attr_pool.attr, | |
9bb2f334 | 2751 | &dev_attr_pool_id.attr, |
dfc5606d | 2752 | &dev_attr_name.attr, |
589d30e0 | 2753 | &dev_attr_image_id.attr, |
dfc5606d | 2754 | &dev_attr_current_snap.attr, |
86b00e0d | 2755 | &dev_attr_parent.attr, |
dfc5606d | 2756 | &dev_attr_refresh.attr, |
dfc5606d YS |
2757 | NULL |
2758 | }; | |
2759 | ||
2760 | static struct attribute_group rbd_attr_group = { | |
2761 | .attrs = rbd_attrs, | |
2762 | }; | |
2763 | ||
2764 | static const struct attribute_group *rbd_attr_groups[] = { | |
2765 | &rbd_attr_group, | |
2766 | NULL | |
2767 | }; | |
2768 | ||
2769 | static void rbd_sysfs_dev_release(struct device *dev) | |
2770 | { | |
2771 | } | |
2772 | ||
2773 | static struct device_type rbd_device_type = { | |
2774 | .name = "rbd", | |
2775 | .groups = rbd_attr_groups, | |
2776 | .release = rbd_sysfs_dev_release, | |
2777 | }; | |
2778 | ||
2779 | ||
2780 | /* | |
2781 | sysfs - snapshots | |
2782 | */ | |
2783 | ||
2784 | static ssize_t rbd_snap_size_show(struct device *dev, | |
2785 | struct device_attribute *attr, | |
2786 | char *buf) | |
2787 | { | |
2788 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2789 | ||
3591538f | 2790 | return sprintf(buf, "%llu\n", (unsigned long long)snap->size); |
dfc5606d YS |
2791 | } |
2792 | ||
2793 | static ssize_t rbd_snap_id_show(struct device *dev, | |
2794 | struct device_attribute *attr, | |
2795 | char *buf) | |
2796 | { | |
2797 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2798 | ||
3591538f | 2799 | return sprintf(buf, "%llu\n", (unsigned long long)snap->id); |
dfc5606d YS |
2800 | } |
2801 | ||
34b13184 AE |
2802 | static ssize_t rbd_snap_features_show(struct device *dev, |
2803 | struct device_attribute *attr, | |
2804 | char *buf) | |
2805 | { | |
2806 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2807 | ||
2808 | return sprintf(buf, "0x%016llx\n", | |
2809 | (unsigned long long) snap->features); | |
2810 | } | |
2811 | ||
dfc5606d YS |
2812 | static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); |
2813 | static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); | |
34b13184 | 2814 | static DEVICE_ATTR(snap_features, S_IRUGO, rbd_snap_features_show, NULL); |
dfc5606d YS |
2815 | |
2816 | static struct attribute *rbd_snap_attrs[] = { | |
2817 | &dev_attr_snap_size.attr, | |
2818 | &dev_attr_snap_id.attr, | |
34b13184 | 2819 | &dev_attr_snap_features.attr, |
dfc5606d YS |
2820 | NULL, |
2821 | }; | |
2822 | ||
2823 | static struct attribute_group rbd_snap_attr_group = { | |
2824 | .attrs = rbd_snap_attrs, | |
2825 | }; | |
2826 | ||
2827 | static void rbd_snap_dev_release(struct device *dev) | |
2828 | { | |
2829 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); | |
2830 | kfree(snap->name); | |
2831 | kfree(snap); | |
2832 | } | |
2833 | ||
2834 | static const struct attribute_group *rbd_snap_attr_groups[] = { | |
2835 | &rbd_snap_attr_group, | |
2836 | NULL | |
2837 | }; | |
2838 | ||
2839 | static struct device_type rbd_snap_device_type = { | |
2840 | .groups = rbd_snap_attr_groups, | |
2841 | .release = rbd_snap_dev_release, | |
2842 | }; | |
2843 | ||
8b8fb99c AE |
2844 | static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec) |
2845 | { | |
2846 | kref_get(&spec->kref); | |
2847 | ||
2848 | return spec; | |
2849 | } | |
2850 | ||
2851 | static void rbd_spec_free(struct kref *kref); | |
2852 | static void rbd_spec_put(struct rbd_spec *spec) | |
2853 | { | |
2854 | if (spec) | |
2855 | kref_put(&spec->kref, rbd_spec_free); | |
2856 | } | |
2857 | ||
2858 | static struct rbd_spec *rbd_spec_alloc(void) | |
2859 | { | |
2860 | struct rbd_spec *spec; | |
2861 | ||
2862 | spec = kzalloc(sizeof (*spec), GFP_KERNEL); | |
2863 | if (!spec) | |
2864 | return NULL; | |
2865 | kref_init(&spec->kref); | |
2866 | ||
8b8fb99c AE |
2867 | return spec; |
2868 | } | |
2869 | ||
2870 | static void rbd_spec_free(struct kref *kref) | |
2871 | { | |
2872 | struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref); | |
2873 | ||
2874 | kfree(spec->pool_name); | |
2875 | kfree(spec->image_id); | |
2876 | kfree(spec->image_name); | |
2877 | kfree(spec->snap_name); | |
2878 | kfree(spec); | |
2879 | } | |
2880 | ||
cc344fa1 | 2881 | static struct rbd_device *rbd_dev_create(struct rbd_client *rbdc, |
c53d5893 AE |
2882 | struct rbd_spec *spec) |
2883 | { | |
2884 | struct rbd_device *rbd_dev; | |
2885 | ||
2886 | rbd_dev = kzalloc(sizeof (*rbd_dev), GFP_KERNEL); | |
2887 | if (!rbd_dev) | |
2888 | return NULL; | |
2889 | ||
2890 | spin_lock_init(&rbd_dev->lock); | |
6d292906 | 2891 | rbd_dev->flags = 0; |
c53d5893 AE |
2892 | INIT_LIST_HEAD(&rbd_dev->node); |
2893 | INIT_LIST_HEAD(&rbd_dev->snaps); | |
2894 | init_rwsem(&rbd_dev->header_rwsem); | |
2895 | ||
2896 | rbd_dev->spec = spec; | |
2897 | rbd_dev->rbd_client = rbdc; | |
2898 | ||
0903e875 AE |
2899 | /* Initialize the layout used for all rbd requests */ |
2900 | ||
2901 | rbd_dev->layout.fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
2902 | rbd_dev->layout.fl_stripe_count = cpu_to_le32(1); | |
2903 | rbd_dev->layout.fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
2904 | rbd_dev->layout.fl_pg_pool = cpu_to_le32((u32) spec->pool_id); | |
2905 | ||
c53d5893 AE |
2906 | return rbd_dev; |
2907 | } | |
2908 | ||
2909 | static void rbd_dev_destroy(struct rbd_device *rbd_dev) | |
2910 | { | |
86b00e0d | 2911 | rbd_spec_put(rbd_dev->parent_spec); |
c53d5893 AE |
2912 | kfree(rbd_dev->header_name); |
2913 | rbd_put_client(rbd_dev->rbd_client); | |
2914 | rbd_spec_put(rbd_dev->spec); | |
2915 | kfree(rbd_dev); | |
2916 | } | |
2917 | ||
304f6808 AE |
2918 | static bool rbd_snap_registered(struct rbd_snap *snap) |
2919 | { | |
2920 | bool ret = snap->dev.type == &rbd_snap_device_type; | |
2921 | bool reg = device_is_registered(&snap->dev); | |
2922 | ||
2923 | rbd_assert(!ret ^ reg); | |
2924 | ||
2925 | return ret; | |
2926 | } | |
2927 | ||
41f38c2b | 2928 | static void rbd_remove_snap_dev(struct rbd_snap *snap) |
dfc5606d YS |
2929 | { |
2930 | list_del(&snap->node); | |
304f6808 AE |
2931 | if (device_is_registered(&snap->dev)) |
2932 | device_unregister(&snap->dev); | |
dfc5606d YS |
2933 | } |
2934 | ||
14e7085d | 2935 | static int rbd_register_snap_dev(struct rbd_snap *snap, |
dfc5606d YS |
2936 | struct device *parent) |
2937 | { | |
2938 | struct device *dev = &snap->dev; | |
2939 | int ret; | |
2940 | ||
2941 | dev->type = &rbd_snap_device_type; | |
2942 | dev->parent = parent; | |
2943 | dev->release = rbd_snap_dev_release; | |
d4b125e9 | 2944 | dev_set_name(dev, "%s%s", RBD_SNAP_DEV_NAME_PREFIX, snap->name); |
304f6808 AE |
2945 | dout("%s: registering device for snapshot %s\n", __func__, snap->name); |
2946 | ||
dfc5606d YS |
2947 | ret = device_register(dev); |
2948 | ||
2949 | return ret; | |
2950 | } | |
2951 | ||
4e891e0a | 2952 | static struct rbd_snap *__rbd_add_snap_dev(struct rbd_device *rbd_dev, |
c8d18425 | 2953 | const char *snap_name, |
34b13184 AE |
2954 | u64 snap_id, u64 snap_size, |
2955 | u64 snap_features) | |
dfc5606d | 2956 | { |
4e891e0a | 2957 | struct rbd_snap *snap; |
dfc5606d | 2958 | int ret; |
4e891e0a AE |
2959 | |
2960 | snap = kzalloc(sizeof (*snap), GFP_KERNEL); | |
dfc5606d | 2961 | if (!snap) |
4e891e0a AE |
2962 | return ERR_PTR(-ENOMEM); |
2963 | ||
2964 | ret = -ENOMEM; | |
c8d18425 | 2965 | snap->name = kstrdup(snap_name, GFP_KERNEL); |
4e891e0a AE |
2966 | if (!snap->name) |
2967 | goto err; | |
2968 | ||
c8d18425 AE |
2969 | snap->id = snap_id; |
2970 | snap->size = snap_size; | |
34b13184 | 2971 | snap->features = snap_features; |
4e891e0a AE |
2972 | |
2973 | return snap; | |
2974 | ||
dfc5606d YS |
2975 | err: |
2976 | kfree(snap->name); | |
2977 | kfree(snap); | |
4e891e0a AE |
2978 | |
2979 | return ERR_PTR(ret); | |
dfc5606d YS |
2980 | } |
2981 | ||
cd892126 AE |
2982 | static char *rbd_dev_v1_snap_info(struct rbd_device *rbd_dev, u32 which, |
2983 | u64 *snap_size, u64 *snap_features) | |
2984 | { | |
2985 | char *snap_name; | |
2986 | ||
2987 | rbd_assert(which < rbd_dev->header.snapc->num_snaps); | |
2988 | ||
2989 | *snap_size = rbd_dev->header.snap_sizes[which]; | |
2990 | *snap_features = 0; /* No features for v1 */ | |
2991 | ||
2992 | /* Skip over names until we find the one we are looking for */ | |
2993 | ||
2994 | snap_name = rbd_dev->header.snap_names; | |
2995 | while (which--) | |
2996 | snap_name += strlen(snap_name) + 1; | |
2997 | ||
2998 | return snap_name; | |
2999 | } | |
3000 | ||
9d475de5 AE |
3001 | /* |
3002 | * Get the size and object order for an image snapshot, or if | |
3003 | * snap_id is CEPH_NOSNAP, gets this information for the base | |
3004 | * image. | |
3005 | */ | |
3006 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, | |
3007 | u8 *order, u64 *snap_size) | |
3008 | { | |
3009 | __le64 snapid = cpu_to_le64(snap_id); | |
3010 | int ret; | |
3011 | struct { | |
3012 | u8 order; | |
3013 | __le64 size; | |
3014 | } __attribute__ ((packed)) size_buf = { 0 }; | |
3015 | ||
36be9a76 | 3016 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
9d475de5 AE |
3017 | "rbd", "get_size", |
3018 | (char *) &snapid, sizeof (snapid), | |
07b2391f | 3019 | (char *) &size_buf, sizeof (size_buf), NULL); |
36be9a76 | 3020 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
9d475de5 AE |
3021 | if (ret < 0) |
3022 | return ret; | |
3023 | ||
3024 | *order = size_buf.order; | |
3025 | *snap_size = le64_to_cpu(size_buf.size); | |
3026 | ||
3027 | dout(" snap_id 0x%016llx order = %u, snap_size = %llu\n", | |
3028 | (unsigned long long) snap_id, (unsigned int) *order, | |
3029 | (unsigned long long) *snap_size); | |
3030 | ||
3031 | return 0; | |
3032 | } | |
3033 | ||
3034 | static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) | |
3035 | { | |
3036 | return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, | |
3037 | &rbd_dev->header.obj_order, | |
3038 | &rbd_dev->header.image_size); | |
3039 | } | |
3040 | ||
1e130199 AE |
3041 | static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) |
3042 | { | |
3043 | void *reply_buf; | |
3044 | int ret; | |
3045 | void *p; | |
3046 | ||
3047 | reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL); | |
3048 | if (!reply_buf) | |
3049 | return -ENOMEM; | |
3050 | ||
36be9a76 | 3051 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
1e130199 AE |
3052 | "rbd", "get_object_prefix", |
3053 | NULL, 0, | |
07b2391f | 3054 | reply_buf, RBD_OBJ_PREFIX_LEN_MAX, NULL); |
36be9a76 | 3055 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
1e130199 AE |
3056 | if (ret < 0) |
3057 | goto out; | |
3058 | ||
3059 | p = reply_buf; | |
3060 | rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, | |
3061 | p + RBD_OBJ_PREFIX_LEN_MAX, | |
3062 | NULL, GFP_NOIO); | |
3063 | ||
3064 | if (IS_ERR(rbd_dev->header.object_prefix)) { | |
3065 | ret = PTR_ERR(rbd_dev->header.object_prefix); | |
3066 | rbd_dev->header.object_prefix = NULL; | |
3067 | } else { | |
3068 | dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); | |
3069 | } | |
3070 | ||
3071 | out: | |
3072 | kfree(reply_buf); | |
3073 | ||
3074 | return ret; | |
3075 | } | |
3076 | ||
b1b5402a AE |
3077 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
3078 | u64 *snap_features) | |
3079 | { | |
3080 | __le64 snapid = cpu_to_le64(snap_id); | |
3081 | struct { | |
3082 | __le64 features; | |
3083 | __le64 incompat; | |
3084 | } features_buf = { 0 }; | |
d889140c | 3085 | u64 incompat; |
b1b5402a AE |
3086 | int ret; |
3087 | ||
36be9a76 | 3088 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
b1b5402a AE |
3089 | "rbd", "get_features", |
3090 | (char *) &snapid, sizeof (snapid), | |
3091 | (char *) &features_buf, sizeof (features_buf), | |
07b2391f | 3092 | NULL); |
36be9a76 | 3093 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
b1b5402a AE |
3094 | if (ret < 0) |
3095 | return ret; | |
d889140c AE |
3096 | |
3097 | incompat = le64_to_cpu(features_buf.incompat); | |
5cbf6f12 | 3098 | if (incompat & ~RBD_FEATURES_SUPPORTED) |
b8f5c6ed | 3099 | return -ENXIO; |
d889140c | 3100 | |
b1b5402a AE |
3101 | *snap_features = le64_to_cpu(features_buf.features); |
3102 | ||
3103 | dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", | |
3104 | (unsigned long long) snap_id, | |
3105 | (unsigned long long) *snap_features, | |
3106 | (unsigned long long) le64_to_cpu(features_buf.incompat)); | |
3107 | ||
3108 | return 0; | |
3109 | } | |
3110 | ||
3111 | static int rbd_dev_v2_features(struct rbd_device *rbd_dev) | |
3112 | { | |
3113 | return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, | |
3114 | &rbd_dev->header.features); | |
3115 | } | |
3116 | ||
86b00e0d AE |
3117 | static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) |
3118 | { | |
3119 | struct rbd_spec *parent_spec; | |
3120 | size_t size; | |
3121 | void *reply_buf = NULL; | |
3122 | __le64 snapid; | |
3123 | void *p; | |
3124 | void *end; | |
3125 | char *image_id; | |
3126 | u64 overlap; | |
86b00e0d AE |
3127 | int ret; |
3128 | ||
3129 | parent_spec = rbd_spec_alloc(); | |
3130 | if (!parent_spec) | |
3131 | return -ENOMEM; | |
3132 | ||
3133 | size = sizeof (__le64) + /* pool_id */ | |
3134 | sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX + /* image_id */ | |
3135 | sizeof (__le64) + /* snap_id */ | |
3136 | sizeof (__le64); /* overlap */ | |
3137 | reply_buf = kmalloc(size, GFP_KERNEL); | |
3138 | if (!reply_buf) { | |
3139 | ret = -ENOMEM; | |
3140 | goto out_err; | |
3141 | } | |
3142 | ||
3143 | snapid = cpu_to_le64(CEPH_NOSNAP); | |
36be9a76 | 3144 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
86b00e0d AE |
3145 | "rbd", "get_parent", |
3146 | (char *) &snapid, sizeof (snapid), | |
07b2391f | 3147 | (char *) reply_buf, size, NULL); |
36be9a76 | 3148 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
86b00e0d AE |
3149 | if (ret < 0) |
3150 | goto out_err; | |
3151 | ||
3152 | ret = -ERANGE; | |
3153 | p = reply_buf; | |
3154 | end = (char *) reply_buf + size; | |
3155 | ceph_decode_64_safe(&p, end, parent_spec->pool_id, out_err); | |
3156 | if (parent_spec->pool_id == CEPH_NOPOOL) | |
3157 | goto out; /* No parent? No problem. */ | |
3158 | ||
0903e875 AE |
3159 | /* The ceph file layout needs to fit pool id in 32 bits */ |
3160 | ||
3161 | ret = -EIO; | |
3162 | if (WARN_ON(parent_spec->pool_id > (u64) U32_MAX)) | |
3163 | goto out; | |
3164 | ||
979ed480 | 3165 | image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
86b00e0d AE |
3166 | if (IS_ERR(image_id)) { |
3167 | ret = PTR_ERR(image_id); | |
3168 | goto out_err; | |
3169 | } | |
3170 | parent_spec->image_id = image_id; | |
3171 | ceph_decode_64_safe(&p, end, parent_spec->snap_id, out_err); | |
3172 | ceph_decode_64_safe(&p, end, overlap, out_err); | |
3173 | ||
3174 | rbd_dev->parent_overlap = overlap; | |
3175 | rbd_dev->parent_spec = parent_spec; | |
3176 | parent_spec = NULL; /* rbd_dev now owns this */ | |
3177 | out: | |
3178 | ret = 0; | |
3179 | out_err: | |
3180 | kfree(reply_buf); | |
3181 | rbd_spec_put(parent_spec); | |
3182 | ||
3183 | return ret; | |
3184 | } | |
3185 | ||
9e15b77d AE |
3186 | static char *rbd_dev_image_name(struct rbd_device *rbd_dev) |
3187 | { | |
3188 | size_t image_id_size; | |
3189 | char *image_id; | |
3190 | void *p; | |
3191 | void *end; | |
3192 | size_t size; | |
3193 | void *reply_buf = NULL; | |
3194 | size_t len = 0; | |
3195 | char *image_name = NULL; | |
3196 | int ret; | |
3197 | ||
3198 | rbd_assert(!rbd_dev->spec->image_name); | |
3199 | ||
69e7a02f AE |
3200 | len = strlen(rbd_dev->spec->image_id); |
3201 | image_id_size = sizeof (__le32) + len; | |
9e15b77d AE |
3202 | image_id = kmalloc(image_id_size, GFP_KERNEL); |
3203 | if (!image_id) | |
3204 | return NULL; | |
3205 | ||
3206 | p = image_id; | |
3207 | end = (char *) image_id + image_id_size; | |
69e7a02f | 3208 | ceph_encode_string(&p, end, rbd_dev->spec->image_id, (u32) len); |
9e15b77d AE |
3209 | |
3210 | size = sizeof (__le32) + RBD_IMAGE_NAME_LEN_MAX; | |
3211 | reply_buf = kmalloc(size, GFP_KERNEL); | |
3212 | if (!reply_buf) | |
3213 | goto out; | |
3214 | ||
36be9a76 | 3215 | ret = rbd_obj_method_sync(rbd_dev, RBD_DIRECTORY, |
9e15b77d AE |
3216 | "rbd", "dir_get_name", |
3217 | image_id, image_id_size, | |
07b2391f | 3218 | (char *) reply_buf, size, NULL); |
9e15b77d AE |
3219 | if (ret < 0) |
3220 | goto out; | |
3221 | p = reply_buf; | |
3222 | end = (char *) reply_buf + size; | |
3223 | image_name = ceph_extract_encoded_string(&p, end, &len, GFP_KERNEL); | |
3224 | if (IS_ERR(image_name)) | |
3225 | image_name = NULL; | |
3226 | else | |
3227 | dout("%s: name is %s len is %zd\n", __func__, image_name, len); | |
3228 | out: | |
3229 | kfree(reply_buf); | |
3230 | kfree(image_id); | |
3231 | ||
3232 | return image_name; | |
3233 | } | |
3234 | ||
3235 | /* | |
3236 | * When a parent image gets probed, we only have the pool, image, | |
3237 | * and snapshot ids but not the names of any of them. This call | |
3238 | * is made later to fill in those names. It has to be done after | |
3239 | * rbd_dev_snaps_update() has completed because some of the | |
3240 | * information (in particular, snapshot name) is not available | |
3241 | * until then. | |
3242 | */ | |
3243 | static int rbd_dev_probe_update_spec(struct rbd_device *rbd_dev) | |
3244 | { | |
3245 | struct ceph_osd_client *osdc; | |
3246 | const char *name; | |
3247 | void *reply_buf = NULL; | |
3248 | int ret; | |
3249 | ||
3250 | if (rbd_dev->spec->pool_name) | |
3251 | return 0; /* Already have the names */ | |
3252 | ||
3253 | /* Look up the pool name */ | |
3254 | ||
3255 | osdc = &rbd_dev->rbd_client->client->osdc; | |
3256 | name = ceph_pg_pool_name_by_id(osdc->osdmap, rbd_dev->spec->pool_id); | |
935dc89f AE |
3257 | if (!name) { |
3258 | rbd_warn(rbd_dev, "there is no pool with id %llu", | |
3259 | rbd_dev->spec->pool_id); /* Really a BUG() */ | |
3260 | return -EIO; | |
3261 | } | |
9e15b77d AE |
3262 | |
3263 | rbd_dev->spec->pool_name = kstrdup(name, GFP_KERNEL); | |
3264 | if (!rbd_dev->spec->pool_name) | |
3265 | return -ENOMEM; | |
3266 | ||
3267 | /* Fetch the image name; tolerate failure here */ | |
3268 | ||
3269 | name = rbd_dev_image_name(rbd_dev); | |
69e7a02f | 3270 | if (name) |
9e15b77d | 3271 | rbd_dev->spec->image_name = (char *) name; |
69e7a02f | 3272 | else |
06ecc6cb | 3273 | rbd_warn(rbd_dev, "unable to get image name"); |
9e15b77d AE |
3274 | |
3275 | /* Look up the snapshot name. */ | |
3276 | ||
3277 | name = rbd_snap_name(rbd_dev, rbd_dev->spec->snap_id); | |
3278 | if (!name) { | |
935dc89f AE |
3279 | rbd_warn(rbd_dev, "no snapshot with id %llu", |
3280 | rbd_dev->spec->snap_id); /* Really a BUG() */ | |
9e15b77d AE |
3281 | ret = -EIO; |
3282 | goto out_err; | |
3283 | } | |
3284 | rbd_dev->spec->snap_name = kstrdup(name, GFP_KERNEL); | |
3285 | if(!rbd_dev->spec->snap_name) | |
3286 | goto out_err; | |
3287 | ||
3288 | return 0; | |
3289 | out_err: | |
3290 | kfree(reply_buf); | |
3291 | kfree(rbd_dev->spec->pool_name); | |
3292 | rbd_dev->spec->pool_name = NULL; | |
3293 | ||
3294 | return ret; | |
3295 | } | |
3296 | ||
6e14b1a6 | 3297 | static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev, u64 *ver) |
35d489f9 AE |
3298 | { |
3299 | size_t size; | |
3300 | int ret; | |
3301 | void *reply_buf; | |
3302 | void *p; | |
3303 | void *end; | |
3304 | u64 seq; | |
3305 | u32 snap_count; | |
3306 | struct ceph_snap_context *snapc; | |
3307 | u32 i; | |
3308 | ||
3309 | /* | |
3310 | * We'll need room for the seq value (maximum snapshot id), | |
3311 | * snapshot count, and array of that many snapshot ids. | |
3312 | * For now we have a fixed upper limit on the number we're | |
3313 | * prepared to receive. | |
3314 | */ | |
3315 | size = sizeof (__le64) + sizeof (__le32) + | |
3316 | RBD_MAX_SNAP_COUNT * sizeof (__le64); | |
3317 | reply_buf = kzalloc(size, GFP_KERNEL); | |
3318 | if (!reply_buf) | |
3319 | return -ENOMEM; | |
3320 | ||
36be9a76 | 3321 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
35d489f9 AE |
3322 | "rbd", "get_snapcontext", |
3323 | NULL, 0, | |
07b2391f | 3324 | reply_buf, size, ver); |
36be9a76 | 3325 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
35d489f9 AE |
3326 | if (ret < 0) |
3327 | goto out; | |
3328 | ||
3329 | ret = -ERANGE; | |
3330 | p = reply_buf; | |
3331 | end = (char *) reply_buf + size; | |
3332 | ceph_decode_64_safe(&p, end, seq, out); | |
3333 | ceph_decode_32_safe(&p, end, snap_count, out); | |
3334 | ||
3335 | /* | |
3336 | * Make sure the reported number of snapshot ids wouldn't go | |
3337 | * beyond the end of our buffer. But before checking that, | |
3338 | * make sure the computed size of the snapshot context we | |
3339 | * allocate is representable in a size_t. | |
3340 | */ | |
3341 | if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) | |
3342 | / sizeof (u64)) { | |
3343 | ret = -EINVAL; | |
3344 | goto out; | |
3345 | } | |
3346 | if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) | |
3347 | goto out; | |
3348 | ||
3349 | size = sizeof (struct ceph_snap_context) + | |
3350 | snap_count * sizeof (snapc->snaps[0]); | |
3351 | snapc = kmalloc(size, GFP_KERNEL); | |
3352 | if (!snapc) { | |
3353 | ret = -ENOMEM; | |
3354 | goto out; | |
3355 | } | |
3356 | ||
3357 | atomic_set(&snapc->nref, 1); | |
3358 | snapc->seq = seq; | |
3359 | snapc->num_snaps = snap_count; | |
3360 | for (i = 0; i < snap_count; i++) | |
3361 | snapc->snaps[i] = ceph_decode_64(&p); | |
3362 | ||
3363 | rbd_dev->header.snapc = snapc; | |
3364 | ||
3365 | dout(" snap context seq = %llu, snap_count = %u\n", | |
3366 | (unsigned long long) seq, (unsigned int) snap_count); | |
3367 | ||
3368 | out: | |
3369 | kfree(reply_buf); | |
3370 | ||
3371 | return 0; | |
3372 | } | |
3373 | ||
b8b1e2db AE |
3374 | static char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, u32 which) |
3375 | { | |
3376 | size_t size; | |
3377 | void *reply_buf; | |
3378 | __le64 snap_id; | |
3379 | int ret; | |
3380 | void *p; | |
3381 | void *end; | |
b8b1e2db AE |
3382 | char *snap_name; |
3383 | ||
3384 | size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; | |
3385 | reply_buf = kmalloc(size, GFP_KERNEL); | |
3386 | if (!reply_buf) | |
3387 | return ERR_PTR(-ENOMEM); | |
3388 | ||
3389 | snap_id = cpu_to_le64(rbd_dev->header.snapc->snaps[which]); | |
36be9a76 | 3390 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
b8b1e2db AE |
3391 | "rbd", "get_snapshot_name", |
3392 | (char *) &snap_id, sizeof (snap_id), | |
07b2391f | 3393 | reply_buf, size, NULL); |
36be9a76 | 3394 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
b8b1e2db AE |
3395 | if (ret < 0) |
3396 | goto out; | |
3397 | ||
3398 | p = reply_buf; | |
3399 | end = (char *) reply_buf + size; | |
e5c35534 | 3400 | snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
b8b1e2db AE |
3401 | if (IS_ERR(snap_name)) { |
3402 | ret = PTR_ERR(snap_name); | |
3403 | goto out; | |
3404 | } else { | |
3405 | dout(" snap_id 0x%016llx snap_name = %s\n", | |
3406 | (unsigned long long) le64_to_cpu(snap_id), snap_name); | |
3407 | } | |
3408 | kfree(reply_buf); | |
3409 | ||
3410 | return snap_name; | |
3411 | out: | |
3412 | kfree(reply_buf); | |
3413 | ||
3414 | return ERR_PTR(ret); | |
3415 | } | |
3416 | ||
3417 | static char *rbd_dev_v2_snap_info(struct rbd_device *rbd_dev, u32 which, | |
3418 | u64 *snap_size, u64 *snap_features) | |
3419 | { | |
e0b49868 | 3420 | u64 snap_id; |
b8b1e2db AE |
3421 | u8 order; |
3422 | int ret; | |
3423 | ||
3424 | snap_id = rbd_dev->header.snapc->snaps[which]; | |
3425 | ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, &order, snap_size); | |
3426 | if (ret) | |
3427 | return ERR_PTR(ret); | |
3428 | ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, snap_features); | |
3429 | if (ret) | |
3430 | return ERR_PTR(ret); | |
3431 | ||
3432 | return rbd_dev_v2_snap_name(rbd_dev, which); | |
3433 | } | |
3434 | ||
3435 | static char *rbd_dev_snap_info(struct rbd_device *rbd_dev, u32 which, | |
3436 | u64 *snap_size, u64 *snap_features) | |
3437 | { | |
3438 | if (rbd_dev->image_format == 1) | |
3439 | return rbd_dev_v1_snap_info(rbd_dev, which, | |
3440 | snap_size, snap_features); | |
3441 | if (rbd_dev->image_format == 2) | |
3442 | return rbd_dev_v2_snap_info(rbd_dev, which, | |
3443 | snap_size, snap_features); | |
3444 | return ERR_PTR(-EINVAL); | |
3445 | } | |
3446 | ||
117973fb AE |
3447 | static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver) |
3448 | { | |
3449 | int ret; | |
3450 | __u8 obj_order; | |
3451 | ||
3452 | down_write(&rbd_dev->header_rwsem); | |
3453 | ||
3454 | /* Grab old order first, to see if it changes */ | |
3455 | ||
3456 | obj_order = rbd_dev->header.obj_order, | |
3457 | ret = rbd_dev_v2_image_size(rbd_dev); | |
3458 | if (ret) | |
3459 | goto out; | |
3460 | if (rbd_dev->header.obj_order != obj_order) { | |
3461 | ret = -EIO; | |
3462 | goto out; | |
3463 | } | |
3464 | rbd_update_mapping_size(rbd_dev); | |
3465 | ||
3466 | ret = rbd_dev_v2_snap_context(rbd_dev, hver); | |
3467 | dout("rbd_dev_v2_snap_context returned %d\n", ret); | |
3468 | if (ret) | |
3469 | goto out; | |
3470 | ret = rbd_dev_snaps_update(rbd_dev); | |
3471 | dout("rbd_dev_snaps_update returned %d\n", ret); | |
3472 | if (ret) | |
3473 | goto out; | |
3474 | ret = rbd_dev_snaps_register(rbd_dev); | |
3475 | dout("rbd_dev_snaps_register returned %d\n", ret); | |
3476 | out: | |
3477 | up_write(&rbd_dev->header_rwsem); | |
3478 | ||
3479 | return ret; | |
3480 | } | |
3481 | ||
dfc5606d | 3482 | /* |
35938150 AE |
3483 | * Scan the rbd device's current snapshot list and compare it to the |
3484 | * newly-received snapshot context. Remove any existing snapshots | |
3485 | * not present in the new snapshot context. Add a new snapshot for | |
3486 | * any snaphots in the snapshot context not in the current list. | |
3487 | * And verify there are no changes to snapshots we already know | |
3488 | * about. | |
3489 | * | |
3490 | * Assumes the snapshots in the snapshot context are sorted by | |
3491 | * snapshot id, highest id first. (Snapshots in the rbd_dev's list | |
3492 | * are also maintained in that order.) | |
dfc5606d | 3493 | */ |
304f6808 | 3494 | static int rbd_dev_snaps_update(struct rbd_device *rbd_dev) |
dfc5606d | 3495 | { |
35938150 AE |
3496 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
3497 | const u32 snap_count = snapc->num_snaps; | |
35938150 AE |
3498 | struct list_head *head = &rbd_dev->snaps; |
3499 | struct list_head *links = head->next; | |
3500 | u32 index = 0; | |
dfc5606d | 3501 | |
9fcbb800 | 3502 | dout("%s: snap count is %u\n", __func__, (unsigned int) snap_count); |
35938150 AE |
3503 | while (index < snap_count || links != head) { |
3504 | u64 snap_id; | |
3505 | struct rbd_snap *snap; | |
cd892126 AE |
3506 | char *snap_name; |
3507 | u64 snap_size = 0; | |
3508 | u64 snap_features = 0; | |
dfc5606d | 3509 | |
35938150 AE |
3510 | snap_id = index < snap_count ? snapc->snaps[index] |
3511 | : CEPH_NOSNAP; | |
3512 | snap = links != head ? list_entry(links, struct rbd_snap, node) | |
3513 | : NULL; | |
aafb230e | 3514 | rbd_assert(!snap || snap->id != CEPH_NOSNAP); |
dfc5606d | 3515 | |
35938150 AE |
3516 | if (snap_id == CEPH_NOSNAP || (snap && snap->id > snap_id)) { |
3517 | struct list_head *next = links->next; | |
dfc5606d | 3518 | |
6d292906 AE |
3519 | /* |
3520 | * A previously-existing snapshot is not in | |
3521 | * the new snap context. | |
3522 | * | |
3523 | * If the now missing snapshot is the one the | |
3524 | * image is mapped to, clear its exists flag | |
3525 | * so we can avoid sending any more requests | |
3526 | * to it. | |
3527 | */ | |
0d7dbfce | 3528 | if (rbd_dev->spec->snap_id == snap->id) |
6d292906 | 3529 | clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
41f38c2b | 3530 | rbd_remove_snap_dev(snap); |
9fcbb800 | 3531 | dout("%ssnap id %llu has been removed\n", |
0d7dbfce AE |
3532 | rbd_dev->spec->snap_id == snap->id ? |
3533 | "mapped " : "", | |
9fcbb800 | 3534 | (unsigned long long) snap->id); |
35938150 AE |
3535 | |
3536 | /* Done with this list entry; advance */ | |
3537 | ||
3538 | links = next; | |
dfc5606d YS |
3539 | continue; |
3540 | } | |
35938150 | 3541 | |
b8b1e2db AE |
3542 | snap_name = rbd_dev_snap_info(rbd_dev, index, |
3543 | &snap_size, &snap_features); | |
cd892126 AE |
3544 | if (IS_ERR(snap_name)) |
3545 | return PTR_ERR(snap_name); | |
3546 | ||
9fcbb800 AE |
3547 | dout("entry %u: snap_id = %llu\n", (unsigned int) snap_count, |
3548 | (unsigned long long) snap_id); | |
35938150 AE |
3549 | if (!snap || (snap_id != CEPH_NOSNAP && snap->id < snap_id)) { |
3550 | struct rbd_snap *new_snap; | |
3551 | ||
3552 | /* We haven't seen this snapshot before */ | |
3553 | ||
c8d18425 | 3554 | new_snap = __rbd_add_snap_dev(rbd_dev, snap_name, |
cd892126 | 3555 | snap_id, snap_size, snap_features); |
9fcbb800 AE |
3556 | if (IS_ERR(new_snap)) { |
3557 | int err = PTR_ERR(new_snap); | |
3558 | ||
3559 | dout(" failed to add dev, error %d\n", err); | |
3560 | ||
3561 | return err; | |
3562 | } | |
35938150 AE |
3563 | |
3564 | /* New goes before existing, or at end of list */ | |
3565 | ||
9fcbb800 | 3566 | dout(" added dev%s\n", snap ? "" : " at end\n"); |
35938150 AE |
3567 | if (snap) |
3568 | list_add_tail(&new_snap->node, &snap->node); | |
3569 | else | |
523f3258 | 3570 | list_add_tail(&new_snap->node, head); |
35938150 AE |
3571 | } else { |
3572 | /* Already have this one */ | |
3573 | ||
9fcbb800 AE |
3574 | dout(" already present\n"); |
3575 | ||
cd892126 | 3576 | rbd_assert(snap->size == snap_size); |
aafb230e | 3577 | rbd_assert(!strcmp(snap->name, snap_name)); |
cd892126 | 3578 | rbd_assert(snap->features == snap_features); |
35938150 AE |
3579 | |
3580 | /* Done with this list entry; advance */ | |
3581 | ||
3582 | links = links->next; | |
dfc5606d | 3583 | } |
35938150 AE |
3584 | |
3585 | /* Advance to the next entry in the snapshot context */ | |
3586 | ||
3587 | index++; | |
dfc5606d | 3588 | } |
9fcbb800 | 3589 | dout("%s: done\n", __func__); |
dfc5606d YS |
3590 | |
3591 | return 0; | |
3592 | } | |
3593 | ||
304f6808 AE |
3594 | /* |
3595 | * Scan the list of snapshots and register the devices for any that | |
3596 | * have not already been registered. | |
3597 | */ | |
3598 | static int rbd_dev_snaps_register(struct rbd_device *rbd_dev) | |
3599 | { | |
3600 | struct rbd_snap *snap; | |
3601 | int ret = 0; | |
3602 | ||
37206ee5 | 3603 | dout("%s:\n", __func__); |
86ff77bb AE |
3604 | if (WARN_ON(!device_is_registered(&rbd_dev->dev))) |
3605 | return -EIO; | |
304f6808 AE |
3606 | |
3607 | list_for_each_entry(snap, &rbd_dev->snaps, node) { | |
3608 | if (!rbd_snap_registered(snap)) { | |
3609 | ret = rbd_register_snap_dev(snap, &rbd_dev->dev); | |
3610 | if (ret < 0) | |
3611 | break; | |
3612 | } | |
3613 | } | |
3614 | dout("%s: returning %d\n", __func__, ret); | |
3615 | ||
3616 | return ret; | |
3617 | } | |
3618 | ||
dfc5606d YS |
3619 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
3620 | { | |
dfc5606d | 3621 | struct device *dev; |
cd789ab9 | 3622 | int ret; |
dfc5606d YS |
3623 | |
3624 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
dfc5606d | 3625 | |
cd789ab9 | 3626 | dev = &rbd_dev->dev; |
dfc5606d YS |
3627 | dev->bus = &rbd_bus_type; |
3628 | dev->type = &rbd_device_type; | |
3629 | dev->parent = &rbd_root_dev; | |
3630 | dev->release = rbd_dev_release; | |
de71a297 | 3631 | dev_set_name(dev, "%d", rbd_dev->dev_id); |
dfc5606d | 3632 | ret = device_register(dev); |
dfc5606d | 3633 | |
dfc5606d | 3634 | mutex_unlock(&ctl_mutex); |
cd789ab9 | 3635 | |
dfc5606d | 3636 | return ret; |
602adf40 YS |
3637 | } |
3638 | ||
dfc5606d YS |
3639 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
3640 | { | |
3641 | device_unregister(&rbd_dev->dev); | |
3642 | } | |
3643 | ||
e2839308 | 3644 | static atomic64_t rbd_dev_id_max = ATOMIC64_INIT(0); |
1ddbe94e AE |
3645 | |
3646 | /* | |
499afd5b AE |
3647 | * Get a unique rbd identifier for the given new rbd_dev, and add |
3648 | * the rbd_dev to the global list. The minimum rbd id is 1. | |
1ddbe94e | 3649 | */ |
e2839308 | 3650 | static void rbd_dev_id_get(struct rbd_device *rbd_dev) |
b7f23c36 | 3651 | { |
e2839308 | 3652 | rbd_dev->dev_id = atomic64_inc_return(&rbd_dev_id_max); |
499afd5b AE |
3653 | |
3654 | spin_lock(&rbd_dev_list_lock); | |
3655 | list_add_tail(&rbd_dev->node, &rbd_dev_list); | |
3656 | spin_unlock(&rbd_dev_list_lock); | |
e2839308 AE |
3657 | dout("rbd_dev %p given dev id %llu\n", rbd_dev, |
3658 | (unsigned long long) rbd_dev->dev_id); | |
1ddbe94e | 3659 | } |
b7f23c36 | 3660 | |
1ddbe94e | 3661 | /* |
499afd5b AE |
3662 | * Remove an rbd_dev from the global list, and record that its |
3663 | * identifier is no longer in use. | |
1ddbe94e | 3664 | */ |
e2839308 | 3665 | static void rbd_dev_id_put(struct rbd_device *rbd_dev) |
1ddbe94e | 3666 | { |
d184f6bf | 3667 | struct list_head *tmp; |
de71a297 | 3668 | int rbd_id = rbd_dev->dev_id; |
d184f6bf AE |
3669 | int max_id; |
3670 | ||
aafb230e | 3671 | rbd_assert(rbd_id > 0); |
499afd5b | 3672 | |
e2839308 AE |
3673 | dout("rbd_dev %p released dev id %llu\n", rbd_dev, |
3674 | (unsigned long long) rbd_dev->dev_id); | |
499afd5b AE |
3675 | spin_lock(&rbd_dev_list_lock); |
3676 | list_del_init(&rbd_dev->node); | |
d184f6bf AE |
3677 | |
3678 | /* | |
3679 | * If the id being "put" is not the current maximum, there | |
3680 | * is nothing special we need to do. | |
3681 | */ | |
e2839308 | 3682 | if (rbd_id != atomic64_read(&rbd_dev_id_max)) { |
d184f6bf AE |
3683 | spin_unlock(&rbd_dev_list_lock); |
3684 | return; | |
3685 | } | |
3686 | ||
3687 | /* | |
3688 | * We need to update the current maximum id. Search the | |
3689 | * list to find out what it is. We're more likely to find | |
3690 | * the maximum at the end, so search the list backward. | |
3691 | */ | |
3692 | max_id = 0; | |
3693 | list_for_each_prev(tmp, &rbd_dev_list) { | |
3694 | struct rbd_device *rbd_dev; | |
3695 | ||
3696 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
b213e0b1 AE |
3697 | if (rbd_dev->dev_id > max_id) |
3698 | max_id = rbd_dev->dev_id; | |
d184f6bf | 3699 | } |
499afd5b | 3700 | spin_unlock(&rbd_dev_list_lock); |
b7f23c36 | 3701 | |
1ddbe94e | 3702 | /* |
e2839308 | 3703 | * The max id could have been updated by rbd_dev_id_get(), in |
d184f6bf AE |
3704 | * which case it now accurately reflects the new maximum. |
3705 | * Be careful not to overwrite the maximum value in that | |
3706 | * case. | |
1ddbe94e | 3707 | */ |
e2839308 AE |
3708 | atomic64_cmpxchg(&rbd_dev_id_max, rbd_id, max_id); |
3709 | dout(" max dev id has been reset\n"); | |
b7f23c36 AE |
3710 | } |
3711 | ||
e28fff26 AE |
3712 | /* |
3713 | * Skips over white space at *buf, and updates *buf to point to the | |
3714 | * first found non-space character (if any). Returns the length of | |
593a9e7b AE |
3715 | * the token (string of non-white space characters) found. Note |
3716 | * that *buf must be terminated with '\0'. | |
e28fff26 AE |
3717 | */ |
3718 | static inline size_t next_token(const char **buf) | |
3719 | { | |
3720 | /* | |
3721 | * These are the characters that produce nonzero for | |
3722 | * isspace() in the "C" and "POSIX" locales. | |
3723 | */ | |
3724 | const char *spaces = " \f\n\r\t\v"; | |
3725 | ||
3726 | *buf += strspn(*buf, spaces); /* Find start of token */ | |
3727 | ||
3728 | return strcspn(*buf, spaces); /* Return token length */ | |
3729 | } | |
3730 | ||
3731 | /* | |
3732 | * Finds the next token in *buf, and if the provided token buffer is | |
3733 | * big enough, copies the found token into it. The result, if | |
593a9e7b AE |
3734 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
3735 | * must be terminated with '\0' on entry. | |
e28fff26 AE |
3736 | * |
3737 | * Returns the length of the token found (not including the '\0'). | |
3738 | * Return value will be 0 if no token is found, and it will be >= | |
3739 | * token_size if the token would not fit. | |
3740 | * | |
593a9e7b | 3741 | * The *buf pointer will be updated to point beyond the end of the |
e28fff26 AE |
3742 | * found token. Note that this occurs even if the token buffer is |
3743 | * too small to hold it. | |
3744 | */ | |
3745 | static inline size_t copy_token(const char **buf, | |
3746 | char *token, | |
3747 | size_t token_size) | |
3748 | { | |
3749 | size_t len; | |
3750 | ||
3751 | len = next_token(buf); | |
3752 | if (len < token_size) { | |
3753 | memcpy(token, *buf, len); | |
3754 | *(token + len) = '\0'; | |
3755 | } | |
3756 | *buf += len; | |
3757 | ||
3758 | return len; | |
3759 | } | |
3760 | ||
ea3352f4 AE |
3761 | /* |
3762 | * Finds the next token in *buf, dynamically allocates a buffer big | |
3763 | * enough to hold a copy of it, and copies the token into the new | |
3764 | * buffer. The copy is guaranteed to be terminated with '\0'. Note | |
3765 | * that a duplicate buffer is created even for a zero-length token. | |
3766 | * | |
3767 | * Returns a pointer to the newly-allocated duplicate, or a null | |
3768 | * pointer if memory for the duplicate was not available. If | |
3769 | * the lenp argument is a non-null pointer, the length of the token | |
3770 | * (not including the '\0') is returned in *lenp. | |
3771 | * | |
3772 | * If successful, the *buf pointer will be updated to point beyond | |
3773 | * the end of the found token. | |
3774 | * | |
3775 | * Note: uses GFP_KERNEL for allocation. | |
3776 | */ | |
3777 | static inline char *dup_token(const char **buf, size_t *lenp) | |
3778 | { | |
3779 | char *dup; | |
3780 | size_t len; | |
3781 | ||
3782 | len = next_token(buf); | |
4caf35f9 | 3783 | dup = kmemdup(*buf, len + 1, GFP_KERNEL); |
ea3352f4 AE |
3784 | if (!dup) |
3785 | return NULL; | |
ea3352f4 AE |
3786 | *(dup + len) = '\0'; |
3787 | *buf += len; | |
3788 | ||
3789 | if (lenp) | |
3790 | *lenp = len; | |
3791 | ||
3792 | return dup; | |
3793 | } | |
3794 | ||
a725f65e | 3795 | /* |
859c31df AE |
3796 | * Parse the options provided for an "rbd add" (i.e., rbd image |
3797 | * mapping) request. These arrive via a write to /sys/bus/rbd/add, | |
3798 | * and the data written is passed here via a NUL-terminated buffer. | |
3799 | * Returns 0 if successful or an error code otherwise. | |
d22f76e7 | 3800 | * |
859c31df AE |
3801 | * The information extracted from these options is recorded in |
3802 | * the other parameters which return dynamically-allocated | |
3803 | * structures: | |
3804 | * ceph_opts | |
3805 | * The address of a pointer that will refer to a ceph options | |
3806 | * structure. Caller must release the returned pointer using | |
3807 | * ceph_destroy_options() when it is no longer needed. | |
3808 | * rbd_opts | |
3809 | * Address of an rbd options pointer. Fully initialized by | |
3810 | * this function; caller must release with kfree(). | |
3811 | * spec | |
3812 | * Address of an rbd image specification pointer. Fully | |
3813 | * initialized by this function based on parsed options. | |
3814 | * Caller must release with rbd_spec_put(). | |
3815 | * | |
3816 | * The options passed take this form: | |
3817 | * <mon_addrs> <options> <pool_name> <image_name> [<snap_id>] | |
3818 | * where: | |
3819 | * <mon_addrs> | |
3820 | * A comma-separated list of one or more monitor addresses. | |
3821 | * A monitor address is an ip address, optionally followed | |
3822 | * by a port number (separated by a colon). | |
3823 | * I.e.: ip1[:port1][,ip2[:port2]...] | |
3824 | * <options> | |
3825 | * A comma-separated list of ceph and/or rbd options. | |
3826 | * <pool_name> | |
3827 | * The name of the rados pool containing the rbd image. | |
3828 | * <image_name> | |
3829 | * The name of the image in that pool to map. | |
3830 | * <snap_id> | |
3831 | * An optional snapshot id. If provided, the mapping will | |
3832 | * present data from the image at the time that snapshot was | |
3833 | * created. The image head is used if no snapshot id is | |
3834 | * provided. Snapshot mappings are always read-only. | |
a725f65e | 3835 | */ |
859c31df | 3836 | static int rbd_add_parse_args(const char *buf, |
dc79b113 | 3837 | struct ceph_options **ceph_opts, |
859c31df AE |
3838 | struct rbd_options **opts, |
3839 | struct rbd_spec **rbd_spec) | |
e28fff26 | 3840 | { |
d22f76e7 | 3841 | size_t len; |
859c31df | 3842 | char *options; |
0ddebc0c AE |
3843 | const char *mon_addrs; |
3844 | size_t mon_addrs_size; | |
859c31df | 3845 | struct rbd_spec *spec = NULL; |
4e9afeba | 3846 | struct rbd_options *rbd_opts = NULL; |
859c31df | 3847 | struct ceph_options *copts; |
dc79b113 | 3848 | int ret; |
e28fff26 AE |
3849 | |
3850 | /* The first four tokens are required */ | |
3851 | ||
7ef3214a | 3852 | len = next_token(&buf); |
4fb5d671 AE |
3853 | if (!len) { |
3854 | rbd_warn(NULL, "no monitor address(es) provided"); | |
3855 | return -EINVAL; | |
3856 | } | |
0ddebc0c | 3857 | mon_addrs = buf; |
f28e565a | 3858 | mon_addrs_size = len + 1; |
7ef3214a | 3859 | buf += len; |
a725f65e | 3860 | |
dc79b113 | 3861 | ret = -EINVAL; |
f28e565a AE |
3862 | options = dup_token(&buf, NULL); |
3863 | if (!options) | |
dc79b113 | 3864 | return -ENOMEM; |
4fb5d671 AE |
3865 | if (!*options) { |
3866 | rbd_warn(NULL, "no options provided"); | |
3867 | goto out_err; | |
3868 | } | |
e28fff26 | 3869 | |
859c31df AE |
3870 | spec = rbd_spec_alloc(); |
3871 | if (!spec) | |
f28e565a | 3872 | goto out_mem; |
859c31df AE |
3873 | |
3874 | spec->pool_name = dup_token(&buf, NULL); | |
3875 | if (!spec->pool_name) | |
3876 | goto out_mem; | |
4fb5d671 AE |
3877 | if (!*spec->pool_name) { |
3878 | rbd_warn(NULL, "no pool name provided"); | |
3879 | goto out_err; | |
3880 | } | |
e28fff26 | 3881 | |
69e7a02f | 3882 | spec->image_name = dup_token(&buf, NULL); |
859c31df | 3883 | if (!spec->image_name) |
f28e565a | 3884 | goto out_mem; |
4fb5d671 AE |
3885 | if (!*spec->image_name) { |
3886 | rbd_warn(NULL, "no image name provided"); | |
3887 | goto out_err; | |
3888 | } | |
d4b125e9 | 3889 | |
f28e565a AE |
3890 | /* |
3891 | * Snapshot name is optional; default is to use "-" | |
3892 | * (indicating the head/no snapshot). | |
3893 | */ | |
3feeb894 | 3894 | len = next_token(&buf); |
820a5f3e | 3895 | if (!len) { |
3feeb894 AE |
3896 | buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ |
3897 | len = sizeof (RBD_SNAP_HEAD_NAME) - 1; | |
f28e565a | 3898 | } else if (len > RBD_MAX_SNAP_NAME_LEN) { |
dc79b113 | 3899 | ret = -ENAMETOOLONG; |
f28e565a | 3900 | goto out_err; |
849b4260 | 3901 | } |
4caf35f9 | 3902 | spec->snap_name = kmemdup(buf, len + 1, GFP_KERNEL); |
859c31df | 3903 | if (!spec->snap_name) |
f28e565a | 3904 | goto out_mem; |
859c31df | 3905 | *(spec->snap_name + len) = '\0'; |
e5c35534 | 3906 | |
0ddebc0c | 3907 | /* Initialize all rbd options to the defaults */ |
e28fff26 | 3908 | |
4e9afeba AE |
3909 | rbd_opts = kzalloc(sizeof (*rbd_opts), GFP_KERNEL); |
3910 | if (!rbd_opts) | |
3911 | goto out_mem; | |
3912 | ||
3913 | rbd_opts->read_only = RBD_READ_ONLY_DEFAULT; | |
d22f76e7 | 3914 | |
859c31df | 3915 | copts = ceph_parse_options(options, mon_addrs, |
0ddebc0c | 3916 | mon_addrs + mon_addrs_size - 1, |
4e9afeba | 3917 | parse_rbd_opts_token, rbd_opts); |
859c31df AE |
3918 | if (IS_ERR(copts)) { |
3919 | ret = PTR_ERR(copts); | |
dc79b113 AE |
3920 | goto out_err; |
3921 | } | |
859c31df AE |
3922 | kfree(options); |
3923 | ||
3924 | *ceph_opts = copts; | |
4e9afeba | 3925 | *opts = rbd_opts; |
859c31df | 3926 | *rbd_spec = spec; |
0ddebc0c | 3927 | |
dc79b113 | 3928 | return 0; |
f28e565a | 3929 | out_mem: |
dc79b113 | 3930 | ret = -ENOMEM; |
d22f76e7 | 3931 | out_err: |
859c31df AE |
3932 | kfree(rbd_opts); |
3933 | rbd_spec_put(spec); | |
f28e565a | 3934 | kfree(options); |
d22f76e7 | 3935 | |
dc79b113 | 3936 | return ret; |
a725f65e AE |
3937 | } |
3938 | ||
589d30e0 AE |
3939 | /* |
3940 | * An rbd format 2 image has a unique identifier, distinct from the | |
3941 | * name given to it by the user. Internally, that identifier is | |
3942 | * what's used to specify the names of objects related to the image. | |
3943 | * | |
3944 | * A special "rbd id" object is used to map an rbd image name to its | |
3945 | * id. If that object doesn't exist, then there is no v2 rbd image | |
3946 | * with the supplied name. | |
3947 | * | |
3948 | * This function will record the given rbd_dev's image_id field if | |
3949 | * it can be determined, and in that case will return 0. If any | |
3950 | * errors occur a negative errno will be returned and the rbd_dev's | |
3951 | * image_id field will be unchanged (and should be NULL). | |
3952 | */ | |
3953 | static int rbd_dev_image_id(struct rbd_device *rbd_dev) | |
3954 | { | |
3955 | int ret; | |
3956 | size_t size; | |
3957 | char *object_name; | |
3958 | void *response; | |
3959 | void *p; | |
3960 | ||
2f82ee54 AE |
3961 | /* If we already have it we don't need to look it up */ |
3962 | ||
3963 | if (rbd_dev->spec->image_id) | |
3964 | return 0; | |
3965 | ||
2c0d0a10 AE |
3966 | /* |
3967 | * When probing a parent image, the image id is already | |
3968 | * known (and the image name likely is not). There's no | |
3969 | * need to fetch the image id again in this case. | |
3970 | */ | |
3971 | if (rbd_dev->spec->image_id) | |
3972 | return 0; | |
3973 | ||
589d30e0 AE |
3974 | /* |
3975 | * First, see if the format 2 image id file exists, and if | |
3976 | * so, get the image's persistent id from it. | |
3977 | */ | |
69e7a02f | 3978 | size = sizeof (RBD_ID_PREFIX) + strlen(rbd_dev->spec->image_name); |
589d30e0 AE |
3979 | object_name = kmalloc(size, GFP_NOIO); |
3980 | if (!object_name) | |
3981 | return -ENOMEM; | |
0d7dbfce | 3982 | sprintf(object_name, "%s%s", RBD_ID_PREFIX, rbd_dev->spec->image_name); |
589d30e0 AE |
3983 | dout("rbd id object name is %s\n", object_name); |
3984 | ||
3985 | /* Response will be an encoded string, which includes a length */ | |
3986 | ||
3987 | size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; | |
3988 | response = kzalloc(size, GFP_NOIO); | |
3989 | if (!response) { | |
3990 | ret = -ENOMEM; | |
3991 | goto out; | |
3992 | } | |
3993 | ||
36be9a76 | 3994 | ret = rbd_obj_method_sync(rbd_dev, object_name, |
589d30e0 AE |
3995 | "rbd", "get_id", |
3996 | NULL, 0, | |
07b2391f | 3997 | response, RBD_IMAGE_ID_LEN_MAX, NULL); |
36be9a76 | 3998 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
589d30e0 AE |
3999 | if (ret < 0) |
4000 | goto out; | |
4001 | ||
4002 | p = response; | |
0d7dbfce | 4003 | rbd_dev->spec->image_id = ceph_extract_encoded_string(&p, |
589d30e0 | 4004 | p + RBD_IMAGE_ID_LEN_MAX, |
979ed480 | 4005 | NULL, GFP_NOIO); |
0d7dbfce AE |
4006 | if (IS_ERR(rbd_dev->spec->image_id)) { |
4007 | ret = PTR_ERR(rbd_dev->spec->image_id); | |
4008 | rbd_dev->spec->image_id = NULL; | |
589d30e0 | 4009 | } else { |
0d7dbfce | 4010 | dout("image_id is %s\n", rbd_dev->spec->image_id); |
589d30e0 AE |
4011 | } |
4012 | out: | |
4013 | kfree(response); | |
4014 | kfree(object_name); | |
4015 | ||
4016 | return ret; | |
4017 | } | |
4018 | ||
a30b71b9 AE |
4019 | static int rbd_dev_v1_probe(struct rbd_device *rbd_dev) |
4020 | { | |
4021 | int ret; | |
4022 | size_t size; | |
4023 | ||
4024 | /* Version 1 images have no id; empty string is used */ | |
4025 | ||
0d7dbfce AE |
4026 | rbd_dev->spec->image_id = kstrdup("", GFP_KERNEL); |
4027 | if (!rbd_dev->spec->image_id) | |
a30b71b9 | 4028 | return -ENOMEM; |
a30b71b9 AE |
4029 | |
4030 | /* Record the header object name for this rbd image. */ | |
4031 | ||
69e7a02f | 4032 | size = strlen(rbd_dev->spec->image_name) + sizeof (RBD_SUFFIX); |
a30b71b9 AE |
4033 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); |
4034 | if (!rbd_dev->header_name) { | |
4035 | ret = -ENOMEM; | |
4036 | goto out_err; | |
4037 | } | |
0d7dbfce AE |
4038 | sprintf(rbd_dev->header_name, "%s%s", |
4039 | rbd_dev->spec->image_name, RBD_SUFFIX); | |
a30b71b9 AE |
4040 | |
4041 | /* Populate rbd image metadata */ | |
4042 | ||
4043 | ret = rbd_read_header(rbd_dev, &rbd_dev->header); | |
4044 | if (ret < 0) | |
4045 | goto out_err; | |
86b00e0d AE |
4046 | |
4047 | /* Version 1 images have no parent (no layering) */ | |
4048 | ||
4049 | rbd_dev->parent_spec = NULL; | |
4050 | rbd_dev->parent_overlap = 0; | |
4051 | ||
a30b71b9 AE |
4052 | rbd_dev->image_format = 1; |
4053 | ||
4054 | dout("discovered version 1 image, header name is %s\n", | |
4055 | rbd_dev->header_name); | |
4056 | ||
4057 | return 0; | |
4058 | ||
4059 | out_err: | |
4060 | kfree(rbd_dev->header_name); | |
4061 | rbd_dev->header_name = NULL; | |
0d7dbfce AE |
4062 | kfree(rbd_dev->spec->image_id); |
4063 | rbd_dev->spec->image_id = NULL; | |
a30b71b9 AE |
4064 | |
4065 | return ret; | |
4066 | } | |
4067 | ||
4068 | static int rbd_dev_v2_probe(struct rbd_device *rbd_dev) | |
4069 | { | |
4070 | size_t size; | |
9d475de5 | 4071 | int ret; |
6e14b1a6 | 4072 | u64 ver = 0; |
a30b71b9 AE |
4073 | |
4074 | /* | |
4075 | * Image id was filled in by the caller. Record the header | |
4076 | * object name for this rbd image. | |
4077 | */ | |
979ed480 | 4078 | size = sizeof (RBD_HEADER_PREFIX) + strlen(rbd_dev->spec->image_id); |
a30b71b9 AE |
4079 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); |
4080 | if (!rbd_dev->header_name) | |
4081 | return -ENOMEM; | |
4082 | sprintf(rbd_dev->header_name, "%s%s", | |
0d7dbfce | 4083 | RBD_HEADER_PREFIX, rbd_dev->spec->image_id); |
9d475de5 AE |
4084 | |
4085 | /* Get the size and object order for the image */ | |
4086 | ||
4087 | ret = rbd_dev_v2_image_size(rbd_dev); | |
1e130199 AE |
4088 | if (ret < 0) |
4089 | goto out_err; | |
4090 | ||
4091 | /* Get the object prefix (a.k.a. block_name) for the image */ | |
4092 | ||
4093 | ret = rbd_dev_v2_object_prefix(rbd_dev); | |
b1b5402a AE |
4094 | if (ret < 0) |
4095 | goto out_err; | |
4096 | ||
d889140c | 4097 | /* Get the and check features for the image */ |
b1b5402a AE |
4098 | |
4099 | ret = rbd_dev_v2_features(rbd_dev); | |
9d475de5 AE |
4100 | if (ret < 0) |
4101 | goto out_err; | |
35d489f9 | 4102 | |
86b00e0d AE |
4103 | /* If the image supports layering, get the parent info */ |
4104 | ||
4105 | if (rbd_dev->header.features & RBD_FEATURE_LAYERING) { | |
4106 | ret = rbd_dev_v2_parent_info(rbd_dev); | |
4107 | if (ret < 0) | |
4108 | goto out_err; | |
4109 | } | |
4110 | ||
6e14b1a6 AE |
4111 | /* crypto and compression type aren't (yet) supported for v2 images */ |
4112 | ||
4113 | rbd_dev->header.crypt_type = 0; | |
4114 | rbd_dev->header.comp_type = 0; | |
35d489f9 | 4115 | |
6e14b1a6 AE |
4116 | /* Get the snapshot context, plus the header version */ |
4117 | ||
4118 | ret = rbd_dev_v2_snap_context(rbd_dev, &ver); | |
35d489f9 AE |
4119 | if (ret) |
4120 | goto out_err; | |
6e14b1a6 AE |
4121 | rbd_dev->header.obj_version = ver; |
4122 | ||
a30b71b9 AE |
4123 | rbd_dev->image_format = 2; |
4124 | ||
4125 | dout("discovered version 2 image, header name is %s\n", | |
4126 | rbd_dev->header_name); | |
4127 | ||
35152979 | 4128 | return 0; |
9d475de5 | 4129 | out_err: |
86b00e0d AE |
4130 | rbd_dev->parent_overlap = 0; |
4131 | rbd_spec_put(rbd_dev->parent_spec); | |
4132 | rbd_dev->parent_spec = NULL; | |
9d475de5 AE |
4133 | kfree(rbd_dev->header_name); |
4134 | rbd_dev->header_name = NULL; | |
1e130199 AE |
4135 | kfree(rbd_dev->header.object_prefix); |
4136 | rbd_dev->header.object_prefix = NULL; | |
9d475de5 AE |
4137 | |
4138 | return ret; | |
a30b71b9 AE |
4139 | } |
4140 | ||
83a06263 AE |
4141 | static int rbd_dev_probe_finish(struct rbd_device *rbd_dev) |
4142 | { | |
2f82ee54 AE |
4143 | struct rbd_device *parent = NULL; |
4144 | struct rbd_spec *parent_spec = NULL; | |
4145 | struct rbd_client *rbdc = NULL; | |
83a06263 AE |
4146 | int ret; |
4147 | ||
4148 | /* no need to lock here, as rbd_dev is not registered yet */ | |
4149 | ret = rbd_dev_snaps_update(rbd_dev); | |
4150 | if (ret) | |
4151 | return ret; | |
4152 | ||
9e15b77d AE |
4153 | ret = rbd_dev_probe_update_spec(rbd_dev); |
4154 | if (ret) | |
4155 | goto err_out_snaps; | |
4156 | ||
83a06263 AE |
4157 | ret = rbd_dev_set_mapping(rbd_dev); |
4158 | if (ret) | |
4159 | goto err_out_snaps; | |
4160 | ||
4161 | /* generate unique id: find highest unique id, add one */ | |
4162 | rbd_dev_id_get(rbd_dev); | |
4163 | ||
4164 | /* Fill in the device name, now that we have its id. */ | |
4165 | BUILD_BUG_ON(DEV_NAME_LEN | |
4166 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); | |
4167 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); | |
4168 | ||
4169 | /* Get our block major device number. */ | |
4170 | ||
4171 | ret = register_blkdev(0, rbd_dev->name); | |
4172 | if (ret < 0) | |
4173 | goto err_out_id; | |
4174 | rbd_dev->major = ret; | |
4175 | ||
4176 | /* Set up the blkdev mapping. */ | |
4177 | ||
4178 | ret = rbd_init_disk(rbd_dev); | |
4179 | if (ret) | |
4180 | goto err_out_blkdev; | |
4181 | ||
4182 | ret = rbd_bus_add_dev(rbd_dev); | |
4183 | if (ret) | |
4184 | goto err_out_disk; | |
4185 | ||
4186 | /* | |
4187 | * At this point cleanup in the event of an error is the job | |
4188 | * of the sysfs code (initiated by rbd_bus_del_dev()). | |
4189 | */ | |
2f82ee54 AE |
4190 | /* Probe the parent if there is one */ |
4191 | ||
4192 | if (rbd_dev->parent_spec) { | |
4193 | /* | |
4194 | * We need to pass a reference to the client and the | |
4195 | * parent spec when creating the parent rbd_dev. | |
4196 | * Images related by parent/child relationships | |
4197 | * always share both. | |
4198 | */ | |
4199 | parent_spec = rbd_spec_get(rbd_dev->parent_spec); | |
4200 | rbdc = __rbd_get_client(rbd_dev->rbd_client); | |
4201 | ||
4202 | parent = rbd_dev_create(rbdc, parent_spec); | |
4203 | if (!parent) { | |
4204 | ret = -ENOMEM; | |
4205 | goto err_out_spec; | |
4206 | } | |
4207 | rbdc = NULL; /* parent now owns reference */ | |
4208 | parent_spec = NULL; /* parent now owns reference */ | |
4209 | ret = rbd_dev_probe(parent); | |
4210 | if (ret < 0) | |
4211 | goto err_out_parent; | |
4212 | rbd_dev->parent = parent; | |
4213 | } | |
4214 | ||
83a06263 AE |
4215 | down_write(&rbd_dev->header_rwsem); |
4216 | ret = rbd_dev_snaps_register(rbd_dev); | |
4217 | up_write(&rbd_dev->header_rwsem); | |
4218 | if (ret) | |
4219 | goto err_out_bus; | |
4220 | ||
9969ebc5 | 4221 | ret = rbd_dev_header_watch_sync(rbd_dev, 1); |
83a06263 AE |
4222 | if (ret) |
4223 | goto err_out_bus; | |
4224 | ||
4225 | /* Everything's ready. Announce the disk to the world. */ | |
4226 | ||
4227 | add_disk(rbd_dev->disk); | |
4228 | ||
4229 | pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name, | |
4230 | (unsigned long long) rbd_dev->mapping.size); | |
4231 | ||
4232 | return ret; | |
2f82ee54 AE |
4233 | |
4234 | err_out_parent: | |
4235 | rbd_dev_destroy(parent); | |
4236 | err_out_spec: | |
4237 | rbd_spec_put(parent_spec); | |
4238 | rbd_put_client(rbdc); | |
83a06263 AE |
4239 | err_out_bus: |
4240 | /* this will also clean up rest of rbd_dev stuff */ | |
4241 | ||
4242 | rbd_bus_del_dev(rbd_dev); | |
4243 | ||
4244 | return ret; | |
4245 | err_out_disk: | |
4246 | rbd_free_disk(rbd_dev); | |
4247 | err_out_blkdev: | |
4248 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
4249 | err_out_id: | |
4250 | rbd_dev_id_put(rbd_dev); | |
4251 | err_out_snaps: | |
4252 | rbd_remove_all_snaps(rbd_dev); | |
4253 | ||
4254 | return ret; | |
4255 | } | |
4256 | ||
a30b71b9 AE |
4257 | /* |
4258 | * Probe for the existence of the header object for the given rbd | |
4259 | * device. For format 2 images this includes determining the image | |
4260 | * id. | |
4261 | */ | |
4262 | static int rbd_dev_probe(struct rbd_device *rbd_dev) | |
4263 | { | |
4264 | int ret; | |
4265 | ||
4266 | /* | |
4267 | * Get the id from the image id object. If it's not a | |
4268 | * format 2 image, we'll get ENOENT back, and we'll assume | |
4269 | * it's a format 1 image. | |
4270 | */ | |
4271 | ret = rbd_dev_image_id(rbd_dev); | |
4272 | if (ret) | |
4273 | ret = rbd_dev_v1_probe(rbd_dev); | |
4274 | else | |
4275 | ret = rbd_dev_v2_probe(rbd_dev); | |
83a06263 | 4276 | if (ret) { |
a30b71b9 AE |
4277 | dout("probe failed, returning %d\n", ret); |
4278 | ||
83a06263 AE |
4279 | return ret; |
4280 | } | |
4281 | ||
4282 | ret = rbd_dev_probe_finish(rbd_dev); | |
4283 | if (ret) | |
4284 | rbd_header_free(&rbd_dev->header); | |
4285 | ||
a30b71b9 AE |
4286 | return ret; |
4287 | } | |
4288 | ||
59c2be1e YS |
4289 | static ssize_t rbd_add(struct bus_type *bus, |
4290 | const char *buf, | |
4291 | size_t count) | |
602adf40 | 4292 | { |
cb8627c7 | 4293 | struct rbd_device *rbd_dev = NULL; |
dc79b113 | 4294 | struct ceph_options *ceph_opts = NULL; |
4e9afeba | 4295 | struct rbd_options *rbd_opts = NULL; |
859c31df | 4296 | struct rbd_spec *spec = NULL; |
9d3997fd | 4297 | struct rbd_client *rbdc; |
27cc2594 AE |
4298 | struct ceph_osd_client *osdc; |
4299 | int rc = -ENOMEM; | |
602adf40 YS |
4300 | |
4301 | if (!try_module_get(THIS_MODULE)) | |
4302 | return -ENODEV; | |
4303 | ||
602adf40 | 4304 | /* parse add command */ |
859c31df | 4305 | rc = rbd_add_parse_args(buf, &ceph_opts, &rbd_opts, &spec); |
dc79b113 | 4306 | if (rc < 0) |
bd4ba655 | 4307 | goto err_out_module; |
78cea76e | 4308 | |
9d3997fd AE |
4309 | rbdc = rbd_get_client(ceph_opts); |
4310 | if (IS_ERR(rbdc)) { | |
4311 | rc = PTR_ERR(rbdc); | |
0ddebc0c | 4312 | goto err_out_args; |
9d3997fd | 4313 | } |
c53d5893 | 4314 | ceph_opts = NULL; /* rbd_dev client now owns this */ |
602adf40 | 4315 | |
602adf40 | 4316 | /* pick the pool */ |
9d3997fd | 4317 | osdc = &rbdc->client->osdc; |
859c31df | 4318 | rc = ceph_pg_poolid_by_name(osdc->osdmap, spec->pool_name); |
602adf40 YS |
4319 | if (rc < 0) |
4320 | goto err_out_client; | |
859c31df AE |
4321 | spec->pool_id = (u64) rc; |
4322 | ||
0903e875 AE |
4323 | /* The ceph file layout needs to fit pool id in 32 bits */ |
4324 | ||
4325 | if (WARN_ON(spec->pool_id > (u64) U32_MAX)) { | |
4326 | rc = -EIO; | |
4327 | goto err_out_client; | |
4328 | } | |
4329 | ||
c53d5893 | 4330 | rbd_dev = rbd_dev_create(rbdc, spec); |
bd4ba655 AE |
4331 | if (!rbd_dev) |
4332 | goto err_out_client; | |
c53d5893 AE |
4333 | rbdc = NULL; /* rbd_dev now owns this */ |
4334 | spec = NULL; /* rbd_dev now owns this */ | |
602adf40 | 4335 | |
bd4ba655 | 4336 | rbd_dev->mapping.read_only = rbd_opts->read_only; |
c53d5893 AE |
4337 | kfree(rbd_opts); |
4338 | rbd_opts = NULL; /* done with this */ | |
bd4ba655 | 4339 | |
a30b71b9 AE |
4340 | rc = rbd_dev_probe(rbd_dev); |
4341 | if (rc < 0) | |
c53d5893 | 4342 | goto err_out_rbd_dev; |
05fd6f6f | 4343 | |
602adf40 | 4344 | return count; |
c53d5893 AE |
4345 | err_out_rbd_dev: |
4346 | rbd_dev_destroy(rbd_dev); | |
bd4ba655 | 4347 | err_out_client: |
9d3997fd | 4348 | rbd_put_client(rbdc); |
0ddebc0c | 4349 | err_out_args: |
78cea76e AE |
4350 | if (ceph_opts) |
4351 | ceph_destroy_options(ceph_opts); | |
4e9afeba | 4352 | kfree(rbd_opts); |
859c31df | 4353 | rbd_spec_put(spec); |
bd4ba655 AE |
4354 | err_out_module: |
4355 | module_put(THIS_MODULE); | |
27cc2594 | 4356 | |
602adf40 | 4357 | dout("Error adding device %s\n", buf); |
27cc2594 AE |
4358 | |
4359 | return (ssize_t) rc; | |
602adf40 YS |
4360 | } |
4361 | ||
de71a297 | 4362 | static struct rbd_device *__rbd_get_dev(unsigned long dev_id) |
602adf40 YS |
4363 | { |
4364 | struct list_head *tmp; | |
4365 | struct rbd_device *rbd_dev; | |
4366 | ||
e124a82f | 4367 | spin_lock(&rbd_dev_list_lock); |
602adf40 YS |
4368 | list_for_each(tmp, &rbd_dev_list) { |
4369 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
de71a297 | 4370 | if (rbd_dev->dev_id == dev_id) { |
e124a82f | 4371 | spin_unlock(&rbd_dev_list_lock); |
602adf40 | 4372 | return rbd_dev; |
e124a82f | 4373 | } |
602adf40 | 4374 | } |
e124a82f | 4375 | spin_unlock(&rbd_dev_list_lock); |
602adf40 YS |
4376 | return NULL; |
4377 | } | |
4378 | ||
dfc5606d | 4379 | static void rbd_dev_release(struct device *dev) |
602adf40 | 4380 | { |
593a9e7b | 4381 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 4382 | |
59c2be1e | 4383 | if (rbd_dev->watch_event) |
9969ebc5 | 4384 | rbd_dev_header_watch_sync(rbd_dev, 0); |
602adf40 YS |
4385 | |
4386 | /* clean up and free blkdev */ | |
4387 | rbd_free_disk(rbd_dev); | |
4388 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
32eec68d | 4389 | |
2ac4e75d AE |
4390 | /* release allocated disk header fields */ |
4391 | rbd_header_free(&rbd_dev->header); | |
4392 | ||
32eec68d | 4393 | /* done with the id, and with the rbd_dev */ |
e2839308 | 4394 | rbd_dev_id_put(rbd_dev); |
c53d5893 AE |
4395 | rbd_assert(rbd_dev->rbd_client != NULL); |
4396 | rbd_dev_destroy(rbd_dev); | |
602adf40 YS |
4397 | |
4398 | /* release module ref */ | |
4399 | module_put(THIS_MODULE); | |
602adf40 YS |
4400 | } |
4401 | ||
2f82ee54 AE |
4402 | static void __rbd_remove(struct rbd_device *rbd_dev) |
4403 | { | |
4404 | rbd_remove_all_snaps(rbd_dev); | |
4405 | rbd_bus_del_dev(rbd_dev); | |
4406 | } | |
4407 | ||
dfc5606d YS |
4408 | static ssize_t rbd_remove(struct bus_type *bus, |
4409 | const char *buf, | |
4410 | size_t count) | |
602adf40 YS |
4411 | { |
4412 | struct rbd_device *rbd_dev = NULL; | |
4413 | int target_id, rc; | |
4414 | unsigned long ul; | |
4415 | int ret = count; | |
4416 | ||
4417 | rc = strict_strtoul(buf, 10, &ul); | |
4418 | if (rc) | |
4419 | return rc; | |
4420 | ||
4421 | /* convert to int; abort if we lost anything in the conversion */ | |
4422 | target_id = (int) ul; | |
4423 | if (target_id != ul) | |
4424 | return -EINVAL; | |
4425 | ||
4426 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
4427 | ||
4428 | rbd_dev = __rbd_get_dev(target_id); | |
4429 | if (!rbd_dev) { | |
4430 | ret = -ENOENT; | |
4431 | goto done; | |
42382b70 AE |
4432 | } |
4433 | ||
a14ea269 | 4434 | spin_lock_irq(&rbd_dev->lock); |
b82d167b | 4435 | if (rbd_dev->open_count) |
42382b70 | 4436 | ret = -EBUSY; |
b82d167b AE |
4437 | else |
4438 | set_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags); | |
a14ea269 | 4439 | spin_unlock_irq(&rbd_dev->lock); |
b82d167b | 4440 | if (ret < 0) |
42382b70 | 4441 | goto done; |
602adf40 | 4442 | |
2f82ee54 AE |
4443 | while (rbd_dev->parent_spec) { |
4444 | struct rbd_device *first = rbd_dev; | |
4445 | struct rbd_device *second = first->parent; | |
4446 | struct rbd_device *third; | |
4447 | ||
4448 | /* | |
4449 | * Follow to the parent with no grandparent and | |
4450 | * remove it. | |
4451 | */ | |
4452 | while (second && (third = second->parent)) { | |
4453 | first = second; | |
4454 | second = third; | |
4455 | } | |
4456 | __rbd_remove(second); | |
4457 | rbd_spec_put(first->parent_spec); | |
4458 | first->parent_spec = NULL; | |
4459 | first->parent_overlap = 0; | |
4460 | first->parent = NULL; | |
4461 | } | |
4462 | __rbd_remove(rbd_dev); | |
602adf40 YS |
4463 | |
4464 | done: | |
4465 | mutex_unlock(&ctl_mutex); | |
aafb230e | 4466 | |
602adf40 YS |
4467 | return ret; |
4468 | } | |
4469 | ||
602adf40 YS |
4470 | /* |
4471 | * create control files in sysfs | |
dfc5606d | 4472 | * /sys/bus/rbd/... |
602adf40 YS |
4473 | */ |
4474 | static int rbd_sysfs_init(void) | |
4475 | { | |
dfc5606d | 4476 | int ret; |
602adf40 | 4477 | |
fed4c143 | 4478 | ret = device_register(&rbd_root_dev); |
21079786 | 4479 | if (ret < 0) |
dfc5606d | 4480 | return ret; |
602adf40 | 4481 | |
fed4c143 AE |
4482 | ret = bus_register(&rbd_bus_type); |
4483 | if (ret < 0) | |
4484 | device_unregister(&rbd_root_dev); | |
602adf40 | 4485 | |
602adf40 YS |
4486 | return ret; |
4487 | } | |
4488 | ||
4489 | static void rbd_sysfs_cleanup(void) | |
4490 | { | |
dfc5606d | 4491 | bus_unregister(&rbd_bus_type); |
fed4c143 | 4492 | device_unregister(&rbd_root_dev); |
602adf40 YS |
4493 | } |
4494 | ||
cc344fa1 | 4495 | static int __init rbd_init(void) |
602adf40 YS |
4496 | { |
4497 | int rc; | |
4498 | ||
1e32d34c AE |
4499 | if (!libceph_compatible(NULL)) { |
4500 | rbd_warn(NULL, "libceph incompatibility (quitting)"); | |
4501 | ||
4502 | return -EINVAL; | |
4503 | } | |
602adf40 YS |
4504 | rc = rbd_sysfs_init(); |
4505 | if (rc) | |
4506 | return rc; | |
f0f8cef5 | 4507 | pr_info("loaded " RBD_DRV_NAME_LONG "\n"); |
602adf40 YS |
4508 | return 0; |
4509 | } | |
4510 | ||
cc344fa1 | 4511 | static void __exit rbd_exit(void) |
602adf40 YS |
4512 | { |
4513 | rbd_sysfs_cleanup(); | |
4514 | } | |
4515 | ||
4516 | module_init(rbd_init); | |
4517 | module_exit(rbd_exit); | |
4518 | ||
4519 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
4520 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
4521 | MODULE_DESCRIPTION("rados block device"); | |
4522 | ||
4523 | /* following authorship retained from original osdblk.c */ | |
4524 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); | |
4525 | ||
4526 | MODULE_LICENSE("GPL"); |