]>
Commit | Line | Data |
---|---|---|
e2a58ee5 | 1 | |
602adf40 YS |
2 | /* |
3 | rbd.c -- Export ceph rados objects as a Linux block device | |
4 | ||
5 | ||
6 | based on drivers/block/osdblk.c: | |
7 | ||
8 | Copyright 2009 Red Hat, Inc. | |
9 | ||
10 | This program is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation. | |
13 | ||
14 | This program is distributed in the hope that it will be useful, | |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | GNU General Public License for more details. | |
18 | ||
19 | You should have received a copy of the GNU General Public License | |
20 | along with this program; see the file COPYING. If not, write to | |
21 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
22 | ||
23 | ||
24 | ||
dfc5606d | 25 | For usage instructions, please refer to: |
602adf40 | 26 | |
dfc5606d | 27 | Documentation/ABI/testing/sysfs-bus-rbd |
602adf40 YS |
28 | |
29 | */ | |
30 | ||
31 | #include <linux/ceph/libceph.h> | |
32 | #include <linux/ceph/osd_client.h> | |
33 | #include <linux/ceph/mon_client.h> | |
34 | #include <linux/ceph/decode.h> | |
59c2be1e | 35 | #include <linux/parser.h> |
30d1cff8 | 36 | #include <linux/bsearch.h> |
602adf40 YS |
37 | |
38 | #include <linux/kernel.h> | |
39 | #include <linux/device.h> | |
40 | #include <linux/module.h> | |
41 | #include <linux/fs.h> | |
42 | #include <linux/blkdev.h> | |
1c2a9dfe | 43 | #include <linux/slab.h> |
602adf40 YS |
44 | |
45 | #include "rbd_types.h" | |
46 | ||
aafb230e AE |
47 | #define RBD_DEBUG /* Activate rbd_assert() calls */ |
48 | ||
593a9e7b AE |
49 | /* |
50 | * The basic unit of block I/O is a sector. It is interpreted in a | |
51 | * number of contexts in Linux (blk, bio, genhd), but the default is | |
52 | * universally 512 bytes. These symbols are just slightly more | |
53 | * meaningful than the bare numbers they represent. | |
54 | */ | |
55 | #define SECTOR_SHIFT 9 | |
56 | #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) | |
57 | ||
f0f8cef5 AE |
58 | #define RBD_DRV_NAME "rbd" |
59 | #define RBD_DRV_NAME_LONG "rbd (rados block device)" | |
602adf40 YS |
60 | |
61 | #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ | |
62 | ||
d4b125e9 AE |
63 | #define RBD_SNAP_DEV_NAME_PREFIX "snap_" |
64 | #define RBD_MAX_SNAP_NAME_LEN \ | |
65 | (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1)) | |
66 | ||
35d489f9 | 67 | #define RBD_MAX_SNAP_COUNT 510 /* allows max snapc to fit in 4KB */ |
602adf40 YS |
68 | |
69 | #define RBD_SNAP_HEAD_NAME "-" | |
70 | ||
9682fc6d AE |
71 | #define BAD_SNAP_INDEX U32_MAX /* invalid index into snap array */ |
72 | ||
9e15b77d AE |
73 | /* This allows a single page to hold an image name sent by OSD */ |
74 | #define RBD_IMAGE_NAME_LEN_MAX (PAGE_SIZE - sizeof (__le32) - 1) | |
1e130199 | 75 | #define RBD_IMAGE_ID_LEN_MAX 64 |
9e15b77d | 76 | |
1e130199 | 77 | #define RBD_OBJ_PREFIX_LEN_MAX 64 |
589d30e0 | 78 | |
d889140c AE |
79 | /* Feature bits */ |
80 | ||
5cbf6f12 AE |
81 | #define RBD_FEATURE_LAYERING (1<<0) |
82 | #define RBD_FEATURE_STRIPINGV2 (1<<1) | |
83 | #define RBD_FEATURES_ALL \ | |
84 | (RBD_FEATURE_LAYERING | RBD_FEATURE_STRIPINGV2) | |
d889140c AE |
85 | |
86 | /* Features supported by this (client software) implementation. */ | |
87 | ||
770eba6e | 88 | #define RBD_FEATURES_SUPPORTED (RBD_FEATURES_ALL) |
d889140c | 89 | |
81a89793 AE |
90 | /* |
91 | * An RBD device name will be "rbd#", where the "rbd" comes from | |
92 | * RBD_DRV_NAME above, and # is a unique integer identifier. | |
93 | * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big | |
94 | * enough to hold all possible device names. | |
95 | */ | |
602adf40 | 96 | #define DEV_NAME_LEN 32 |
81a89793 | 97 | #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) |
602adf40 YS |
98 | |
99 | /* | |
100 | * block device image metadata (in-memory version) | |
101 | */ | |
102 | struct rbd_image_header { | |
f35a4dee | 103 | /* These six fields never change for a given rbd image */ |
849b4260 | 104 | char *object_prefix; |
602adf40 YS |
105 | __u8 obj_order; |
106 | __u8 crypt_type; | |
107 | __u8 comp_type; | |
f35a4dee AE |
108 | u64 stripe_unit; |
109 | u64 stripe_count; | |
110 | u64 features; /* Might be changeable someday? */ | |
602adf40 | 111 | |
f84344f3 AE |
112 | /* The remaining fields need to be updated occasionally */ |
113 | u64 image_size; | |
114 | struct ceph_snap_context *snapc; | |
f35a4dee AE |
115 | char *snap_names; /* format 1 only */ |
116 | u64 *snap_sizes; /* format 1 only */ | |
59c2be1e YS |
117 | }; |
118 | ||
0d7dbfce AE |
119 | /* |
120 | * An rbd image specification. | |
121 | * | |
122 | * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely | |
c66c6e0c AE |
123 | * identify an image. Each rbd_dev structure includes a pointer to |
124 | * an rbd_spec structure that encapsulates this identity. | |
125 | * | |
126 | * Each of the id's in an rbd_spec has an associated name. For a | |
127 | * user-mapped image, the names are supplied and the id's associated | |
128 | * with them are looked up. For a layered image, a parent image is | |
129 | * defined by the tuple, and the names are looked up. | |
130 | * | |
131 | * An rbd_dev structure contains a parent_spec pointer which is | |
132 | * non-null if the image it represents is a child in a layered | |
133 | * image. This pointer will refer to the rbd_spec structure used | |
134 | * by the parent rbd_dev for its own identity (i.e., the structure | |
135 | * is shared between the parent and child). | |
136 | * | |
137 | * Since these structures are populated once, during the discovery | |
138 | * phase of image construction, they are effectively immutable so | |
139 | * we make no effort to synchronize access to them. | |
140 | * | |
141 | * Note that code herein does not assume the image name is known (it | |
142 | * could be a null pointer). | |
0d7dbfce AE |
143 | */ |
144 | struct rbd_spec { | |
145 | u64 pool_id; | |
ecb4dc22 | 146 | const char *pool_name; |
0d7dbfce | 147 | |
ecb4dc22 AE |
148 | const char *image_id; |
149 | const char *image_name; | |
0d7dbfce AE |
150 | |
151 | u64 snap_id; | |
ecb4dc22 | 152 | const char *snap_name; |
0d7dbfce AE |
153 | |
154 | struct kref kref; | |
155 | }; | |
156 | ||
602adf40 | 157 | /* |
f0f8cef5 | 158 | * an instance of the client. multiple devices may share an rbd client. |
602adf40 YS |
159 | */ |
160 | struct rbd_client { | |
161 | struct ceph_client *client; | |
162 | struct kref kref; | |
163 | struct list_head node; | |
164 | }; | |
165 | ||
bf0d5f50 AE |
166 | struct rbd_img_request; |
167 | typedef void (*rbd_img_callback_t)(struct rbd_img_request *); | |
168 | ||
169 | #define BAD_WHICH U32_MAX /* Good which or bad which, which? */ | |
170 | ||
171 | struct rbd_obj_request; | |
172 | typedef void (*rbd_obj_callback_t)(struct rbd_obj_request *); | |
173 | ||
9969ebc5 AE |
174 | enum obj_request_type { |
175 | OBJ_REQUEST_NODATA, OBJ_REQUEST_BIO, OBJ_REQUEST_PAGES | |
176 | }; | |
bf0d5f50 | 177 | |
926f9b3f AE |
178 | enum obj_req_flags { |
179 | OBJ_REQ_DONE, /* completion flag: not done = 0, done = 1 */ | |
6365d33a | 180 | OBJ_REQ_IMG_DATA, /* object usage: standalone = 0, image = 1 */ |
5679c59f AE |
181 | OBJ_REQ_KNOWN, /* EXISTS flag valid: no = 0, yes = 1 */ |
182 | OBJ_REQ_EXISTS, /* target exists: no = 0, yes = 1 */ | |
926f9b3f AE |
183 | }; |
184 | ||
bf0d5f50 AE |
185 | struct rbd_obj_request { |
186 | const char *object_name; | |
187 | u64 offset; /* object start byte */ | |
188 | u64 length; /* bytes from offset */ | |
926f9b3f | 189 | unsigned long flags; |
bf0d5f50 | 190 | |
c5b5ef6c AE |
191 | /* |
192 | * An object request associated with an image will have its | |
193 | * img_data flag set; a standalone object request will not. | |
194 | * | |
195 | * A standalone object request will have which == BAD_WHICH | |
196 | * and a null obj_request pointer. | |
197 | * | |
198 | * An object request initiated in support of a layered image | |
199 | * object (to check for its existence before a write) will | |
200 | * have which == BAD_WHICH and a non-null obj_request pointer. | |
201 | * | |
202 | * Finally, an object request for rbd image data will have | |
203 | * which != BAD_WHICH, and will have a non-null img_request | |
204 | * pointer. The value of which will be in the range | |
205 | * 0..(img_request->obj_request_count-1). | |
206 | */ | |
207 | union { | |
208 | struct rbd_obj_request *obj_request; /* STAT op */ | |
209 | struct { | |
210 | struct rbd_img_request *img_request; | |
211 | u64 img_offset; | |
212 | /* links for img_request->obj_requests list */ | |
213 | struct list_head links; | |
214 | }; | |
215 | }; | |
bf0d5f50 AE |
216 | u32 which; /* posn image request list */ |
217 | ||
218 | enum obj_request_type type; | |
788e2df3 AE |
219 | union { |
220 | struct bio *bio_list; | |
221 | struct { | |
222 | struct page **pages; | |
223 | u32 page_count; | |
224 | }; | |
225 | }; | |
0eefd470 | 226 | struct page **copyup_pages; |
ebda6408 | 227 | u32 copyup_page_count; |
bf0d5f50 AE |
228 | |
229 | struct ceph_osd_request *osd_req; | |
230 | ||
231 | u64 xferred; /* bytes transferred */ | |
1b83bef2 | 232 | int result; |
bf0d5f50 AE |
233 | |
234 | rbd_obj_callback_t callback; | |
788e2df3 | 235 | struct completion completion; |
bf0d5f50 AE |
236 | |
237 | struct kref kref; | |
238 | }; | |
239 | ||
0c425248 | 240 | enum img_req_flags { |
9849e986 AE |
241 | IMG_REQ_WRITE, /* I/O direction: read = 0, write = 1 */ |
242 | IMG_REQ_CHILD, /* initiator: block = 0, child image = 1 */ | |
d0b2e944 | 243 | IMG_REQ_LAYERED, /* ENOENT handling: normal = 0, layered = 1 */ |
0c425248 AE |
244 | }; |
245 | ||
bf0d5f50 | 246 | struct rbd_img_request { |
bf0d5f50 AE |
247 | struct rbd_device *rbd_dev; |
248 | u64 offset; /* starting image byte offset */ | |
249 | u64 length; /* byte count from offset */ | |
0c425248 | 250 | unsigned long flags; |
bf0d5f50 | 251 | union { |
9849e986 | 252 | u64 snap_id; /* for reads */ |
bf0d5f50 | 253 | struct ceph_snap_context *snapc; /* for writes */ |
9849e986 AE |
254 | }; |
255 | union { | |
256 | struct request *rq; /* block request */ | |
257 | struct rbd_obj_request *obj_request; /* obj req initiator */ | |
bf0d5f50 | 258 | }; |
3d7efd18 | 259 | struct page **copyup_pages; |
ebda6408 | 260 | u32 copyup_page_count; |
bf0d5f50 AE |
261 | spinlock_t completion_lock;/* protects next_completion */ |
262 | u32 next_completion; | |
263 | rbd_img_callback_t callback; | |
55f27e09 | 264 | u64 xferred;/* aggregate bytes transferred */ |
a5a337d4 | 265 | int result; /* first nonzero obj_request result */ |
bf0d5f50 AE |
266 | |
267 | u32 obj_request_count; | |
268 | struct list_head obj_requests; /* rbd_obj_request structs */ | |
269 | ||
270 | struct kref kref; | |
271 | }; | |
272 | ||
273 | #define for_each_obj_request(ireq, oreq) \ | |
ef06f4d3 | 274 | list_for_each_entry(oreq, &(ireq)->obj_requests, links) |
bf0d5f50 | 275 | #define for_each_obj_request_from(ireq, oreq) \ |
ef06f4d3 | 276 | list_for_each_entry_from(oreq, &(ireq)->obj_requests, links) |
bf0d5f50 | 277 | #define for_each_obj_request_safe(ireq, oreq, n) \ |
ef06f4d3 | 278 | list_for_each_entry_safe_reverse(oreq, n, &(ireq)->obj_requests, links) |
bf0d5f50 | 279 | |
f84344f3 | 280 | struct rbd_mapping { |
99c1f08f | 281 | u64 size; |
34b13184 | 282 | u64 features; |
f84344f3 AE |
283 | bool read_only; |
284 | }; | |
285 | ||
602adf40 YS |
286 | /* |
287 | * a single device | |
288 | */ | |
289 | struct rbd_device { | |
de71a297 | 290 | int dev_id; /* blkdev unique id */ |
602adf40 YS |
291 | |
292 | int major; /* blkdev assigned major */ | |
293 | struct gendisk *disk; /* blkdev's gendisk and rq */ | |
602adf40 | 294 | |
a30b71b9 | 295 | u32 image_format; /* Either 1 or 2 */ |
602adf40 YS |
296 | struct rbd_client *rbd_client; |
297 | ||
298 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ | |
299 | ||
b82d167b | 300 | spinlock_t lock; /* queue, flags, open_count */ |
602adf40 YS |
301 | |
302 | struct rbd_image_header header; | |
b82d167b | 303 | unsigned long flags; /* possibly lock protected */ |
0d7dbfce | 304 | struct rbd_spec *spec; |
602adf40 | 305 | |
0d7dbfce | 306 | char *header_name; |
971f839a | 307 | |
0903e875 AE |
308 | struct ceph_file_layout layout; |
309 | ||
59c2be1e | 310 | struct ceph_osd_event *watch_event; |
975241af | 311 | struct rbd_obj_request *watch_request; |
59c2be1e | 312 | |
86b00e0d AE |
313 | struct rbd_spec *parent_spec; |
314 | u64 parent_overlap; | |
2f82ee54 | 315 | struct rbd_device *parent; |
86b00e0d | 316 | |
c666601a JD |
317 | /* protects updating the header */ |
318 | struct rw_semaphore header_rwsem; | |
f84344f3 AE |
319 | |
320 | struct rbd_mapping mapping; | |
602adf40 YS |
321 | |
322 | struct list_head node; | |
dfc5606d | 323 | |
dfc5606d YS |
324 | /* sysfs related */ |
325 | struct device dev; | |
b82d167b | 326 | unsigned long open_count; /* protected by lock */ |
dfc5606d YS |
327 | }; |
328 | ||
b82d167b AE |
329 | /* |
330 | * Flag bits for rbd_dev->flags. If atomicity is required, | |
331 | * rbd_dev->lock is used to protect access. | |
332 | * | |
333 | * Currently, only the "removing" flag (which is coupled with the | |
334 | * "open_count" field) requires atomic access. | |
335 | */ | |
6d292906 AE |
336 | enum rbd_dev_flags { |
337 | RBD_DEV_FLAG_EXISTS, /* mapped snapshot has not been deleted */ | |
b82d167b | 338 | RBD_DEV_FLAG_REMOVING, /* this mapping is being removed */ |
6d292906 AE |
339 | }; |
340 | ||
602adf40 | 341 | static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ |
e124a82f | 342 | |
602adf40 | 343 | static LIST_HEAD(rbd_dev_list); /* devices */ |
e124a82f AE |
344 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
345 | ||
432b8587 AE |
346 | static LIST_HEAD(rbd_client_list); /* clients */ |
347 | static DEFINE_SPINLOCK(rbd_client_list_lock); | |
602adf40 | 348 | |
78c2a44a AE |
349 | /* Slab caches for frequently-allocated structures */ |
350 | ||
1c2a9dfe | 351 | static struct kmem_cache *rbd_img_request_cache; |
868311b1 | 352 | static struct kmem_cache *rbd_obj_request_cache; |
78c2a44a | 353 | static struct kmem_cache *rbd_segment_name_cache; |
1c2a9dfe | 354 | |
3d7efd18 AE |
355 | static int rbd_img_request_submit(struct rbd_img_request *img_request); |
356 | ||
200a6a8b | 357 | static void rbd_dev_device_release(struct device *dev); |
dfc5606d | 358 | |
f0f8cef5 AE |
359 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
360 | size_t count); | |
361 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, | |
362 | size_t count); | |
1f3ef788 | 363 | static int rbd_dev_image_probe(struct rbd_device *rbd_dev, bool mapping); |
f0f8cef5 AE |
364 | |
365 | static struct bus_attribute rbd_bus_attrs[] = { | |
366 | __ATTR(add, S_IWUSR, NULL, rbd_add), | |
367 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), | |
368 | __ATTR_NULL | |
369 | }; | |
370 | ||
371 | static struct bus_type rbd_bus_type = { | |
372 | .name = "rbd", | |
373 | .bus_attrs = rbd_bus_attrs, | |
374 | }; | |
375 | ||
376 | static void rbd_root_dev_release(struct device *dev) | |
377 | { | |
378 | } | |
379 | ||
380 | static struct device rbd_root_dev = { | |
381 | .init_name = "rbd", | |
382 | .release = rbd_root_dev_release, | |
383 | }; | |
384 | ||
06ecc6cb AE |
385 | static __printf(2, 3) |
386 | void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...) | |
387 | { | |
388 | struct va_format vaf; | |
389 | va_list args; | |
390 | ||
391 | va_start(args, fmt); | |
392 | vaf.fmt = fmt; | |
393 | vaf.va = &args; | |
394 | ||
395 | if (!rbd_dev) | |
396 | printk(KERN_WARNING "%s: %pV\n", RBD_DRV_NAME, &vaf); | |
397 | else if (rbd_dev->disk) | |
398 | printk(KERN_WARNING "%s: %s: %pV\n", | |
399 | RBD_DRV_NAME, rbd_dev->disk->disk_name, &vaf); | |
400 | else if (rbd_dev->spec && rbd_dev->spec->image_name) | |
401 | printk(KERN_WARNING "%s: image %s: %pV\n", | |
402 | RBD_DRV_NAME, rbd_dev->spec->image_name, &vaf); | |
403 | else if (rbd_dev->spec && rbd_dev->spec->image_id) | |
404 | printk(KERN_WARNING "%s: id %s: %pV\n", | |
405 | RBD_DRV_NAME, rbd_dev->spec->image_id, &vaf); | |
406 | else /* punt */ | |
407 | printk(KERN_WARNING "%s: rbd_dev %p: %pV\n", | |
408 | RBD_DRV_NAME, rbd_dev, &vaf); | |
409 | va_end(args); | |
410 | } | |
411 | ||
aafb230e AE |
412 | #ifdef RBD_DEBUG |
413 | #define rbd_assert(expr) \ | |
414 | if (unlikely(!(expr))) { \ | |
415 | printk(KERN_ERR "\nAssertion failure in %s() " \ | |
416 | "at line %d:\n\n" \ | |
417 | "\trbd_assert(%s);\n\n", \ | |
418 | __func__, __LINE__, #expr); \ | |
419 | BUG(); \ | |
420 | } | |
421 | #else /* !RBD_DEBUG */ | |
422 | # define rbd_assert(expr) ((void) 0) | |
423 | #endif /* !RBD_DEBUG */ | |
dfc5606d | 424 | |
b454e36d | 425 | static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request); |
05a46afd AE |
426 | static void rbd_img_parent_read(struct rbd_obj_request *obj_request); |
427 | static void rbd_dev_remove_parent(struct rbd_device *rbd_dev); | |
8b3e1a56 | 428 | |
cc4a38bd | 429 | static int rbd_dev_refresh(struct rbd_device *rbd_dev); |
2df3fac7 AE |
430 | static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev); |
431 | static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev); | |
54cac61f AE |
432 | static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, |
433 | u64 snap_id); | |
2ad3d716 AE |
434 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
435 | u8 *order, u64 *snap_size); | |
436 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, | |
437 | u64 *snap_features); | |
438 | static u64 rbd_snap_id_by_name(struct rbd_device *rbd_dev, const char *name); | |
59c2be1e | 439 | |
602adf40 YS |
440 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
441 | { | |
f0f8cef5 | 442 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
b82d167b | 443 | bool removing = false; |
602adf40 | 444 | |
f84344f3 | 445 | if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only) |
602adf40 YS |
446 | return -EROFS; |
447 | ||
a14ea269 | 448 | spin_lock_irq(&rbd_dev->lock); |
b82d167b AE |
449 | if (test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags)) |
450 | removing = true; | |
451 | else | |
452 | rbd_dev->open_count++; | |
a14ea269 | 453 | spin_unlock_irq(&rbd_dev->lock); |
b82d167b AE |
454 | if (removing) |
455 | return -ENOENT; | |
456 | ||
42382b70 | 457 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
c3e946ce | 458 | (void) get_device(&rbd_dev->dev); |
f84344f3 | 459 | set_device_ro(bdev, rbd_dev->mapping.read_only); |
42382b70 | 460 | mutex_unlock(&ctl_mutex); |
340c7a2b | 461 | |
602adf40 YS |
462 | return 0; |
463 | } | |
464 | ||
dfc5606d YS |
465 | static int rbd_release(struct gendisk *disk, fmode_t mode) |
466 | { | |
467 | struct rbd_device *rbd_dev = disk->private_data; | |
b82d167b AE |
468 | unsigned long open_count_before; |
469 | ||
a14ea269 | 470 | spin_lock_irq(&rbd_dev->lock); |
b82d167b | 471 | open_count_before = rbd_dev->open_count--; |
a14ea269 | 472 | spin_unlock_irq(&rbd_dev->lock); |
b82d167b | 473 | rbd_assert(open_count_before > 0); |
dfc5606d | 474 | |
42382b70 | 475 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
c3e946ce | 476 | put_device(&rbd_dev->dev); |
42382b70 | 477 | mutex_unlock(&ctl_mutex); |
dfc5606d YS |
478 | |
479 | return 0; | |
480 | } | |
481 | ||
602adf40 YS |
482 | static const struct block_device_operations rbd_bd_ops = { |
483 | .owner = THIS_MODULE, | |
484 | .open = rbd_open, | |
dfc5606d | 485 | .release = rbd_release, |
602adf40 YS |
486 | }; |
487 | ||
488 | /* | |
489 | * Initialize an rbd client instance. | |
43ae4701 | 490 | * We own *ceph_opts. |
602adf40 | 491 | */ |
f8c38929 | 492 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) |
602adf40 YS |
493 | { |
494 | struct rbd_client *rbdc; | |
495 | int ret = -ENOMEM; | |
496 | ||
37206ee5 | 497 | dout("%s:\n", __func__); |
602adf40 YS |
498 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); |
499 | if (!rbdc) | |
500 | goto out_opt; | |
501 | ||
502 | kref_init(&rbdc->kref); | |
503 | INIT_LIST_HEAD(&rbdc->node); | |
504 | ||
bc534d86 AE |
505 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
506 | ||
43ae4701 | 507 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
602adf40 | 508 | if (IS_ERR(rbdc->client)) |
bc534d86 | 509 | goto out_mutex; |
43ae4701 | 510 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
602adf40 YS |
511 | |
512 | ret = ceph_open_session(rbdc->client); | |
513 | if (ret < 0) | |
514 | goto out_err; | |
515 | ||
432b8587 | 516 | spin_lock(&rbd_client_list_lock); |
602adf40 | 517 | list_add_tail(&rbdc->node, &rbd_client_list); |
432b8587 | 518 | spin_unlock(&rbd_client_list_lock); |
602adf40 | 519 | |
bc534d86 | 520 | mutex_unlock(&ctl_mutex); |
37206ee5 | 521 | dout("%s: rbdc %p\n", __func__, rbdc); |
bc534d86 | 522 | |
602adf40 YS |
523 | return rbdc; |
524 | ||
525 | out_err: | |
526 | ceph_destroy_client(rbdc->client); | |
bc534d86 AE |
527 | out_mutex: |
528 | mutex_unlock(&ctl_mutex); | |
602adf40 YS |
529 | kfree(rbdc); |
530 | out_opt: | |
43ae4701 AE |
531 | if (ceph_opts) |
532 | ceph_destroy_options(ceph_opts); | |
37206ee5 AE |
533 | dout("%s: error %d\n", __func__, ret); |
534 | ||
28f259b7 | 535 | return ERR_PTR(ret); |
602adf40 YS |
536 | } |
537 | ||
2f82ee54 AE |
538 | static struct rbd_client *__rbd_get_client(struct rbd_client *rbdc) |
539 | { | |
540 | kref_get(&rbdc->kref); | |
541 | ||
542 | return rbdc; | |
543 | } | |
544 | ||
602adf40 | 545 | /* |
1f7ba331 AE |
546 | * Find a ceph client with specific addr and configuration. If |
547 | * found, bump its reference count. | |
602adf40 | 548 | */ |
1f7ba331 | 549 | static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) |
602adf40 YS |
550 | { |
551 | struct rbd_client *client_node; | |
1f7ba331 | 552 | bool found = false; |
602adf40 | 553 | |
43ae4701 | 554 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
602adf40 YS |
555 | return NULL; |
556 | ||
1f7ba331 AE |
557 | spin_lock(&rbd_client_list_lock); |
558 | list_for_each_entry(client_node, &rbd_client_list, node) { | |
559 | if (!ceph_compare_options(ceph_opts, client_node->client)) { | |
2f82ee54 AE |
560 | __rbd_get_client(client_node); |
561 | ||
1f7ba331 AE |
562 | found = true; |
563 | break; | |
564 | } | |
565 | } | |
566 | spin_unlock(&rbd_client_list_lock); | |
567 | ||
568 | return found ? client_node : NULL; | |
602adf40 YS |
569 | } |
570 | ||
59c2be1e YS |
571 | /* |
572 | * mount options | |
573 | */ | |
574 | enum { | |
59c2be1e YS |
575 | Opt_last_int, |
576 | /* int args above */ | |
577 | Opt_last_string, | |
578 | /* string args above */ | |
cc0538b6 AE |
579 | Opt_read_only, |
580 | Opt_read_write, | |
581 | /* Boolean args above */ | |
582 | Opt_last_bool, | |
59c2be1e YS |
583 | }; |
584 | ||
43ae4701 | 585 | static match_table_t rbd_opts_tokens = { |
59c2be1e YS |
586 | /* int args above */ |
587 | /* string args above */ | |
be466c1c | 588 | {Opt_read_only, "read_only"}, |
cc0538b6 AE |
589 | {Opt_read_only, "ro"}, /* Alternate spelling */ |
590 | {Opt_read_write, "read_write"}, | |
591 | {Opt_read_write, "rw"}, /* Alternate spelling */ | |
592 | /* Boolean args above */ | |
59c2be1e YS |
593 | {-1, NULL} |
594 | }; | |
595 | ||
98571b5a AE |
596 | struct rbd_options { |
597 | bool read_only; | |
598 | }; | |
599 | ||
600 | #define RBD_READ_ONLY_DEFAULT false | |
601 | ||
59c2be1e YS |
602 | static int parse_rbd_opts_token(char *c, void *private) |
603 | { | |
43ae4701 | 604 | struct rbd_options *rbd_opts = private; |
59c2be1e YS |
605 | substring_t argstr[MAX_OPT_ARGS]; |
606 | int token, intval, ret; | |
607 | ||
43ae4701 | 608 | token = match_token(c, rbd_opts_tokens, argstr); |
59c2be1e YS |
609 | if (token < 0) |
610 | return -EINVAL; | |
611 | ||
612 | if (token < Opt_last_int) { | |
613 | ret = match_int(&argstr[0], &intval); | |
614 | if (ret < 0) { | |
615 | pr_err("bad mount option arg (not int) " | |
616 | "at '%s'\n", c); | |
617 | return ret; | |
618 | } | |
619 | dout("got int token %d val %d\n", token, intval); | |
620 | } else if (token > Opt_last_int && token < Opt_last_string) { | |
621 | dout("got string token %d val %s\n", token, | |
622 | argstr[0].from); | |
cc0538b6 AE |
623 | } else if (token > Opt_last_string && token < Opt_last_bool) { |
624 | dout("got Boolean token %d\n", token); | |
59c2be1e YS |
625 | } else { |
626 | dout("got token %d\n", token); | |
627 | } | |
628 | ||
629 | switch (token) { | |
cc0538b6 AE |
630 | case Opt_read_only: |
631 | rbd_opts->read_only = true; | |
632 | break; | |
633 | case Opt_read_write: | |
634 | rbd_opts->read_only = false; | |
635 | break; | |
59c2be1e | 636 | default: |
aafb230e AE |
637 | rbd_assert(false); |
638 | break; | |
59c2be1e YS |
639 | } |
640 | return 0; | |
641 | } | |
642 | ||
602adf40 YS |
643 | /* |
644 | * Get a ceph client with specific addr and configuration, if one does | |
645 | * not exist create it. | |
646 | */ | |
9d3997fd | 647 | static struct rbd_client *rbd_get_client(struct ceph_options *ceph_opts) |
602adf40 | 648 | { |
f8c38929 | 649 | struct rbd_client *rbdc; |
59c2be1e | 650 | |
1f7ba331 | 651 | rbdc = rbd_client_find(ceph_opts); |
9d3997fd | 652 | if (rbdc) /* using an existing client */ |
43ae4701 | 653 | ceph_destroy_options(ceph_opts); |
9d3997fd | 654 | else |
f8c38929 | 655 | rbdc = rbd_client_create(ceph_opts); |
602adf40 | 656 | |
9d3997fd | 657 | return rbdc; |
602adf40 YS |
658 | } |
659 | ||
660 | /* | |
661 | * Destroy ceph client | |
d23a4b3f | 662 | * |
432b8587 | 663 | * Caller must hold rbd_client_list_lock. |
602adf40 YS |
664 | */ |
665 | static void rbd_client_release(struct kref *kref) | |
666 | { | |
667 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); | |
668 | ||
37206ee5 | 669 | dout("%s: rbdc %p\n", __func__, rbdc); |
cd9d9f5d | 670 | spin_lock(&rbd_client_list_lock); |
602adf40 | 671 | list_del(&rbdc->node); |
cd9d9f5d | 672 | spin_unlock(&rbd_client_list_lock); |
602adf40 YS |
673 | |
674 | ceph_destroy_client(rbdc->client); | |
675 | kfree(rbdc); | |
676 | } | |
677 | ||
678 | /* | |
679 | * Drop reference to ceph client node. If it's not referenced anymore, release | |
680 | * it. | |
681 | */ | |
9d3997fd | 682 | static void rbd_put_client(struct rbd_client *rbdc) |
602adf40 | 683 | { |
c53d5893 AE |
684 | if (rbdc) |
685 | kref_put(&rbdc->kref, rbd_client_release); | |
602adf40 YS |
686 | } |
687 | ||
a30b71b9 AE |
688 | static bool rbd_image_format_valid(u32 image_format) |
689 | { | |
690 | return image_format == 1 || image_format == 2; | |
691 | } | |
692 | ||
8e94af8e AE |
693 | static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) |
694 | { | |
103a150f AE |
695 | size_t size; |
696 | u32 snap_count; | |
697 | ||
698 | /* The header has to start with the magic rbd header text */ | |
699 | if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) | |
700 | return false; | |
701 | ||
db2388b6 AE |
702 | /* The bio layer requires at least sector-sized I/O */ |
703 | ||
704 | if (ondisk->options.order < SECTOR_SHIFT) | |
705 | return false; | |
706 | ||
707 | /* If we use u64 in a few spots we may be able to loosen this */ | |
708 | ||
709 | if (ondisk->options.order > 8 * sizeof (int) - 1) | |
710 | return false; | |
711 | ||
103a150f AE |
712 | /* |
713 | * The size of a snapshot header has to fit in a size_t, and | |
714 | * that limits the number of snapshots. | |
715 | */ | |
716 | snap_count = le32_to_cpu(ondisk->snap_count); | |
717 | size = SIZE_MAX - sizeof (struct ceph_snap_context); | |
718 | if (snap_count > size / sizeof (__le64)) | |
719 | return false; | |
720 | ||
721 | /* | |
722 | * Not only that, but the size of the entire the snapshot | |
723 | * header must also be representable in a size_t. | |
724 | */ | |
725 | size -= snap_count * sizeof (__le64); | |
726 | if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) | |
727 | return false; | |
728 | ||
729 | return true; | |
8e94af8e AE |
730 | } |
731 | ||
602adf40 | 732 | /* |
bb23e37a AE |
733 | * Fill an rbd image header with information from the given format 1 |
734 | * on-disk header. | |
602adf40 | 735 | */ |
662518b1 | 736 | static int rbd_header_from_disk(struct rbd_device *rbd_dev, |
4156d998 | 737 | struct rbd_image_header_ondisk *ondisk) |
602adf40 | 738 | { |
662518b1 | 739 | struct rbd_image_header *header = &rbd_dev->header; |
bb23e37a AE |
740 | bool first_time = header->object_prefix == NULL; |
741 | struct ceph_snap_context *snapc; | |
742 | char *object_prefix = NULL; | |
743 | char *snap_names = NULL; | |
744 | u64 *snap_sizes = NULL; | |
ccece235 | 745 | u32 snap_count; |
d2bb24e5 | 746 | size_t size; |
bb23e37a | 747 | int ret = -ENOMEM; |
621901d6 | 748 | u32 i; |
602adf40 | 749 | |
bb23e37a AE |
750 | /* Allocate this now to avoid having to handle failure below */ |
751 | ||
752 | if (first_time) { | |
753 | size_t len; | |
754 | ||
755 | len = strnlen(ondisk->object_prefix, | |
756 | sizeof (ondisk->object_prefix)); | |
757 | object_prefix = kmalloc(len + 1, GFP_KERNEL); | |
758 | if (!object_prefix) | |
759 | return -ENOMEM; | |
760 | memcpy(object_prefix, ondisk->object_prefix, len); | |
761 | object_prefix[len] = '\0'; | |
762 | } | |
103a150f | 763 | |
bb23e37a | 764 | /* Allocate the snapshot context and fill it in */ |
00f1f36f | 765 | |
bb23e37a AE |
766 | snap_count = le32_to_cpu(ondisk->snap_count); |
767 | snapc = ceph_create_snap_context(snap_count, GFP_KERNEL); | |
768 | if (!snapc) | |
769 | goto out_err; | |
770 | snapc->seq = le64_to_cpu(ondisk->snap_seq); | |
602adf40 | 771 | if (snap_count) { |
bb23e37a | 772 | struct rbd_image_snap_ondisk *snaps; |
f785cc1d AE |
773 | u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
774 | ||
bb23e37a | 775 | /* We'll keep a copy of the snapshot names... */ |
621901d6 | 776 | |
bb23e37a AE |
777 | if (snap_names_len > (u64)SIZE_MAX) |
778 | goto out_2big; | |
779 | snap_names = kmalloc(snap_names_len, GFP_KERNEL); | |
780 | if (!snap_names) | |
6a52325f AE |
781 | goto out_err; |
782 | ||
bb23e37a | 783 | /* ...as well as the array of their sizes. */ |
621901d6 | 784 | |
d2bb24e5 | 785 | size = snap_count * sizeof (*header->snap_sizes); |
bb23e37a AE |
786 | snap_sizes = kmalloc(size, GFP_KERNEL); |
787 | if (!snap_sizes) | |
6a52325f | 788 | goto out_err; |
bb23e37a AE |
789 | |
790 | /* | |
791 | * Copy the names, and fill in each snapshot's id | |
792 | * and size. | |
793 | * | |
99a41ebc | 794 | * Note that rbd_dev_v1_header_info() guarantees the |
bb23e37a AE |
795 | * ondisk buffer we're working with has |
796 | * snap_names_len bytes beyond the end of the | |
797 | * snapshot id array, this memcpy() is safe. | |
798 | */ | |
799 | memcpy(snap_names, &ondisk->snaps[snap_count], snap_names_len); | |
800 | snaps = ondisk->snaps; | |
801 | for (i = 0; i < snap_count; i++) { | |
802 | snapc->snaps[i] = le64_to_cpu(snaps[i].id); | |
803 | snap_sizes[i] = le64_to_cpu(snaps[i].image_size); | |
804 | } | |
602adf40 | 805 | } |
849b4260 | 806 | |
bb23e37a AE |
807 | /* We won't fail any more, fill in the header */ |
808 | ||
662518b1 | 809 | down_write(&rbd_dev->header_rwsem); |
bb23e37a AE |
810 | if (first_time) { |
811 | header->object_prefix = object_prefix; | |
812 | header->obj_order = ondisk->options.order; | |
813 | header->crypt_type = ondisk->options.crypt_type; | |
814 | header->comp_type = ondisk->options.comp_type; | |
815 | /* The rest aren't used for format 1 images */ | |
816 | header->stripe_unit = 0; | |
817 | header->stripe_count = 0; | |
818 | header->features = 0; | |
662518b1 AE |
819 | } else { |
820 | ceph_put_snap_context(header->snapc); | |
821 | kfree(header->snap_names); | |
822 | kfree(header->snap_sizes); | |
bb23e37a | 823 | } |
6a52325f | 824 | |
bb23e37a | 825 | /* The remaining fields always get updated (when we refresh) */ |
621901d6 | 826 | |
f84344f3 | 827 | header->image_size = le64_to_cpu(ondisk->image_size); |
bb23e37a AE |
828 | header->snapc = snapc; |
829 | header->snap_names = snap_names; | |
830 | header->snap_sizes = snap_sizes; | |
602adf40 | 831 | |
662518b1 AE |
832 | /* Make sure mapping size is consistent with header info */ |
833 | ||
834 | if (rbd_dev->spec->snap_id == CEPH_NOSNAP || first_time) | |
835 | if (rbd_dev->mapping.size != header->image_size) | |
836 | rbd_dev->mapping.size = header->image_size; | |
837 | ||
838 | up_write(&rbd_dev->header_rwsem); | |
839 | ||
602adf40 | 840 | return 0; |
bb23e37a AE |
841 | out_2big: |
842 | ret = -EIO; | |
6a52325f | 843 | out_err: |
bb23e37a AE |
844 | kfree(snap_sizes); |
845 | kfree(snap_names); | |
846 | ceph_put_snap_context(snapc); | |
847 | kfree(object_prefix); | |
848 | ||
849 | return ret; | |
602adf40 YS |
850 | } |
851 | ||
9682fc6d AE |
852 | static const char *_rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, u32 which) |
853 | { | |
854 | const char *snap_name; | |
855 | ||
856 | rbd_assert(which < rbd_dev->header.snapc->num_snaps); | |
857 | ||
858 | /* Skip over names until we find the one we are looking for */ | |
859 | ||
860 | snap_name = rbd_dev->header.snap_names; | |
861 | while (which--) | |
862 | snap_name += strlen(snap_name) + 1; | |
863 | ||
864 | return kstrdup(snap_name, GFP_KERNEL); | |
865 | } | |
866 | ||
30d1cff8 AE |
867 | /* |
868 | * Snapshot id comparison function for use with qsort()/bsearch(). | |
869 | * Note that result is for snapshots in *descending* order. | |
870 | */ | |
871 | static int snapid_compare_reverse(const void *s1, const void *s2) | |
872 | { | |
873 | u64 snap_id1 = *(u64 *)s1; | |
874 | u64 snap_id2 = *(u64 *)s2; | |
875 | ||
876 | if (snap_id1 < snap_id2) | |
877 | return 1; | |
878 | return snap_id1 == snap_id2 ? 0 : -1; | |
879 | } | |
880 | ||
881 | /* | |
882 | * Search a snapshot context to see if the given snapshot id is | |
883 | * present. | |
884 | * | |
885 | * Returns the position of the snapshot id in the array if it's found, | |
886 | * or BAD_SNAP_INDEX otherwise. | |
887 | * | |
888 | * Note: The snapshot array is in kept sorted (by the osd) in | |
889 | * reverse order, highest snapshot id first. | |
890 | */ | |
9682fc6d AE |
891 | static u32 rbd_dev_snap_index(struct rbd_device *rbd_dev, u64 snap_id) |
892 | { | |
893 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; | |
30d1cff8 | 894 | u64 *found; |
9682fc6d | 895 | |
30d1cff8 AE |
896 | found = bsearch(&snap_id, &snapc->snaps, snapc->num_snaps, |
897 | sizeof (snap_id), snapid_compare_reverse); | |
9682fc6d | 898 | |
30d1cff8 | 899 | return found ? (u32)(found - &snapc->snaps[0]) : BAD_SNAP_INDEX; |
9682fc6d AE |
900 | } |
901 | ||
2ad3d716 AE |
902 | static const char *rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, |
903 | u64 snap_id) | |
9e15b77d | 904 | { |
54cac61f | 905 | u32 which; |
9e15b77d | 906 | |
54cac61f AE |
907 | which = rbd_dev_snap_index(rbd_dev, snap_id); |
908 | if (which == BAD_SNAP_INDEX) | |
909 | return NULL; | |
910 | ||
911 | return _rbd_dev_v1_snap_name(rbd_dev, which); | |
912 | } | |
913 | ||
914 | static const char *rbd_snap_name(struct rbd_device *rbd_dev, u64 snap_id) | |
915 | { | |
9e15b77d AE |
916 | if (snap_id == CEPH_NOSNAP) |
917 | return RBD_SNAP_HEAD_NAME; | |
918 | ||
54cac61f AE |
919 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
920 | if (rbd_dev->image_format == 1) | |
921 | return rbd_dev_v1_snap_name(rbd_dev, snap_id); | |
9e15b77d | 922 | |
54cac61f | 923 | return rbd_dev_v2_snap_name(rbd_dev, snap_id); |
9e15b77d AE |
924 | } |
925 | ||
2ad3d716 AE |
926 | static int rbd_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
927 | u64 *snap_size) | |
602adf40 | 928 | { |
2ad3d716 AE |
929 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
930 | if (snap_id == CEPH_NOSNAP) { | |
931 | *snap_size = rbd_dev->header.image_size; | |
932 | } else if (rbd_dev->image_format == 1) { | |
933 | u32 which; | |
602adf40 | 934 | |
2ad3d716 AE |
935 | which = rbd_dev_snap_index(rbd_dev, snap_id); |
936 | if (which == BAD_SNAP_INDEX) | |
937 | return -ENOENT; | |
e86924a8 | 938 | |
2ad3d716 AE |
939 | *snap_size = rbd_dev->header.snap_sizes[which]; |
940 | } else { | |
941 | u64 size = 0; | |
942 | int ret; | |
943 | ||
944 | ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, NULL, &size); | |
945 | if (ret) | |
946 | return ret; | |
947 | ||
948 | *snap_size = size; | |
949 | } | |
950 | return 0; | |
602adf40 YS |
951 | } |
952 | ||
2ad3d716 AE |
953 | static int rbd_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
954 | u64 *snap_features) | |
602adf40 | 955 | { |
2ad3d716 AE |
956 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
957 | if (snap_id == CEPH_NOSNAP) { | |
958 | *snap_features = rbd_dev->header.features; | |
959 | } else if (rbd_dev->image_format == 1) { | |
960 | *snap_features = 0; /* No features for format 1 */ | |
602adf40 | 961 | } else { |
2ad3d716 AE |
962 | u64 features = 0; |
963 | int ret; | |
8b0241f8 | 964 | |
2ad3d716 AE |
965 | ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, &features); |
966 | if (ret) | |
967 | return ret; | |
968 | ||
969 | *snap_features = features; | |
970 | } | |
971 | return 0; | |
972 | } | |
973 | ||
974 | static int rbd_dev_mapping_set(struct rbd_device *rbd_dev) | |
975 | { | |
8f4b7d98 | 976 | u64 snap_id = rbd_dev->spec->snap_id; |
2ad3d716 AE |
977 | u64 size = 0; |
978 | u64 features = 0; | |
979 | int ret; | |
980 | ||
2ad3d716 AE |
981 | ret = rbd_snap_size(rbd_dev, snap_id, &size); |
982 | if (ret) | |
983 | return ret; | |
984 | ret = rbd_snap_features(rbd_dev, snap_id, &features); | |
985 | if (ret) | |
986 | return ret; | |
987 | ||
988 | rbd_dev->mapping.size = size; | |
989 | rbd_dev->mapping.features = features; | |
990 | ||
8b0241f8 | 991 | return 0; |
602adf40 YS |
992 | } |
993 | ||
d1cf5788 AE |
994 | static void rbd_dev_mapping_clear(struct rbd_device *rbd_dev) |
995 | { | |
996 | rbd_dev->mapping.size = 0; | |
997 | rbd_dev->mapping.features = 0; | |
d1cf5788 AE |
998 | } |
999 | ||
98571b5a | 1000 | static const char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset) |
602adf40 | 1001 | { |
65ccfe21 AE |
1002 | char *name; |
1003 | u64 segment; | |
1004 | int ret; | |
602adf40 | 1005 | |
78c2a44a | 1006 | name = kmem_cache_alloc(rbd_segment_name_cache, GFP_NOIO); |
65ccfe21 AE |
1007 | if (!name) |
1008 | return NULL; | |
1009 | segment = offset >> rbd_dev->header.obj_order; | |
2fd82b9e | 1010 | ret = snprintf(name, MAX_OBJ_NAME_SIZE + 1, "%s.%012llx", |
65ccfe21 | 1011 | rbd_dev->header.object_prefix, segment); |
2fd82b9e | 1012 | if (ret < 0 || ret > MAX_OBJ_NAME_SIZE) { |
65ccfe21 AE |
1013 | pr_err("error formatting segment name for #%llu (%d)\n", |
1014 | segment, ret); | |
1015 | kfree(name); | |
1016 | name = NULL; | |
1017 | } | |
602adf40 | 1018 | |
65ccfe21 AE |
1019 | return name; |
1020 | } | |
602adf40 | 1021 | |
78c2a44a AE |
1022 | static void rbd_segment_name_free(const char *name) |
1023 | { | |
1024 | /* The explicit cast here is needed to drop the const qualifier */ | |
1025 | ||
1026 | kmem_cache_free(rbd_segment_name_cache, (void *)name); | |
1027 | } | |
1028 | ||
65ccfe21 AE |
1029 | static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) |
1030 | { | |
1031 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; | |
602adf40 | 1032 | |
65ccfe21 AE |
1033 | return offset & (segment_size - 1); |
1034 | } | |
1035 | ||
1036 | static u64 rbd_segment_length(struct rbd_device *rbd_dev, | |
1037 | u64 offset, u64 length) | |
1038 | { | |
1039 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; | |
1040 | ||
1041 | offset &= segment_size - 1; | |
1042 | ||
aafb230e | 1043 | rbd_assert(length <= U64_MAX - offset); |
65ccfe21 AE |
1044 | if (offset + length > segment_size) |
1045 | length = segment_size - offset; | |
1046 | ||
1047 | return length; | |
602adf40 YS |
1048 | } |
1049 | ||
029bcbd8 JD |
1050 | /* |
1051 | * returns the size of an object in the image | |
1052 | */ | |
1053 | static u64 rbd_obj_bytes(struct rbd_image_header *header) | |
1054 | { | |
1055 | return 1 << header->obj_order; | |
1056 | } | |
1057 | ||
602adf40 YS |
1058 | /* |
1059 | * bio helpers | |
1060 | */ | |
1061 | ||
1062 | static void bio_chain_put(struct bio *chain) | |
1063 | { | |
1064 | struct bio *tmp; | |
1065 | ||
1066 | while (chain) { | |
1067 | tmp = chain; | |
1068 | chain = chain->bi_next; | |
1069 | bio_put(tmp); | |
1070 | } | |
1071 | } | |
1072 | ||
1073 | /* | |
1074 | * zeros a bio chain, starting at specific offset | |
1075 | */ | |
1076 | static void zero_bio_chain(struct bio *chain, int start_ofs) | |
1077 | { | |
1078 | struct bio_vec *bv; | |
1079 | unsigned long flags; | |
1080 | void *buf; | |
1081 | int i; | |
1082 | int pos = 0; | |
1083 | ||
1084 | while (chain) { | |
1085 | bio_for_each_segment(bv, chain, i) { | |
1086 | if (pos + bv->bv_len > start_ofs) { | |
1087 | int remainder = max(start_ofs - pos, 0); | |
1088 | buf = bvec_kmap_irq(bv, &flags); | |
1089 | memset(buf + remainder, 0, | |
1090 | bv->bv_len - remainder); | |
85b5aaa6 | 1091 | bvec_kunmap_irq(buf, &flags); |
602adf40 YS |
1092 | } |
1093 | pos += bv->bv_len; | |
1094 | } | |
1095 | ||
1096 | chain = chain->bi_next; | |
1097 | } | |
1098 | } | |
1099 | ||
b9434c5b AE |
1100 | /* |
1101 | * similar to zero_bio_chain(), zeros data defined by a page array, | |
1102 | * starting at the given byte offset from the start of the array and | |
1103 | * continuing up to the given end offset. The pages array is | |
1104 | * assumed to be big enough to hold all bytes up to the end. | |
1105 | */ | |
1106 | static void zero_pages(struct page **pages, u64 offset, u64 end) | |
1107 | { | |
1108 | struct page **page = &pages[offset >> PAGE_SHIFT]; | |
1109 | ||
1110 | rbd_assert(end > offset); | |
1111 | rbd_assert(end - offset <= (u64)SIZE_MAX); | |
1112 | while (offset < end) { | |
1113 | size_t page_offset; | |
1114 | size_t length; | |
1115 | unsigned long flags; | |
1116 | void *kaddr; | |
1117 | ||
1118 | page_offset = (size_t)(offset & ~PAGE_MASK); | |
1119 | length = min(PAGE_SIZE - page_offset, (size_t)(end - offset)); | |
1120 | local_irq_save(flags); | |
1121 | kaddr = kmap_atomic(*page); | |
1122 | memset(kaddr + page_offset, 0, length); | |
1123 | kunmap_atomic(kaddr); | |
1124 | local_irq_restore(flags); | |
1125 | ||
1126 | offset += length; | |
1127 | page++; | |
1128 | } | |
1129 | } | |
1130 | ||
602adf40 | 1131 | /* |
f7760dad AE |
1132 | * Clone a portion of a bio, starting at the given byte offset |
1133 | * and continuing for the number of bytes indicated. | |
602adf40 | 1134 | */ |
f7760dad AE |
1135 | static struct bio *bio_clone_range(struct bio *bio_src, |
1136 | unsigned int offset, | |
1137 | unsigned int len, | |
1138 | gfp_t gfpmask) | |
602adf40 | 1139 | { |
f7760dad AE |
1140 | struct bio_vec *bv; |
1141 | unsigned int resid; | |
1142 | unsigned short idx; | |
1143 | unsigned int voff; | |
1144 | unsigned short end_idx; | |
1145 | unsigned short vcnt; | |
1146 | struct bio *bio; | |
1147 | ||
1148 | /* Handle the easy case for the caller */ | |
1149 | ||
1150 | if (!offset && len == bio_src->bi_size) | |
1151 | return bio_clone(bio_src, gfpmask); | |
1152 | ||
1153 | if (WARN_ON_ONCE(!len)) | |
1154 | return NULL; | |
1155 | if (WARN_ON_ONCE(len > bio_src->bi_size)) | |
1156 | return NULL; | |
1157 | if (WARN_ON_ONCE(offset > bio_src->bi_size - len)) | |
1158 | return NULL; | |
1159 | ||
1160 | /* Find first affected segment... */ | |
1161 | ||
1162 | resid = offset; | |
1163 | __bio_for_each_segment(bv, bio_src, idx, 0) { | |
1164 | if (resid < bv->bv_len) | |
1165 | break; | |
1166 | resid -= bv->bv_len; | |
602adf40 | 1167 | } |
f7760dad | 1168 | voff = resid; |
602adf40 | 1169 | |
f7760dad | 1170 | /* ...and the last affected segment */ |
602adf40 | 1171 | |
f7760dad AE |
1172 | resid += len; |
1173 | __bio_for_each_segment(bv, bio_src, end_idx, idx) { | |
1174 | if (resid <= bv->bv_len) | |
1175 | break; | |
1176 | resid -= bv->bv_len; | |
1177 | } | |
1178 | vcnt = end_idx - idx + 1; | |
1179 | ||
1180 | /* Build the clone */ | |
1181 | ||
1182 | bio = bio_alloc(gfpmask, (unsigned int) vcnt); | |
1183 | if (!bio) | |
1184 | return NULL; /* ENOMEM */ | |
602adf40 | 1185 | |
f7760dad AE |
1186 | bio->bi_bdev = bio_src->bi_bdev; |
1187 | bio->bi_sector = bio_src->bi_sector + (offset >> SECTOR_SHIFT); | |
1188 | bio->bi_rw = bio_src->bi_rw; | |
1189 | bio->bi_flags |= 1 << BIO_CLONED; | |
1190 | ||
1191 | /* | |
1192 | * Copy over our part of the bio_vec, then update the first | |
1193 | * and last (or only) entries. | |
1194 | */ | |
1195 | memcpy(&bio->bi_io_vec[0], &bio_src->bi_io_vec[idx], | |
1196 | vcnt * sizeof (struct bio_vec)); | |
1197 | bio->bi_io_vec[0].bv_offset += voff; | |
1198 | if (vcnt > 1) { | |
1199 | bio->bi_io_vec[0].bv_len -= voff; | |
1200 | bio->bi_io_vec[vcnt - 1].bv_len = resid; | |
1201 | } else { | |
1202 | bio->bi_io_vec[0].bv_len = len; | |
602adf40 YS |
1203 | } |
1204 | ||
f7760dad AE |
1205 | bio->bi_vcnt = vcnt; |
1206 | bio->bi_size = len; | |
1207 | bio->bi_idx = 0; | |
1208 | ||
1209 | return bio; | |
1210 | } | |
1211 | ||
1212 | /* | |
1213 | * Clone a portion of a bio chain, starting at the given byte offset | |
1214 | * into the first bio in the source chain and continuing for the | |
1215 | * number of bytes indicated. The result is another bio chain of | |
1216 | * exactly the given length, or a null pointer on error. | |
1217 | * | |
1218 | * The bio_src and offset parameters are both in-out. On entry they | |
1219 | * refer to the first source bio and the offset into that bio where | |
1220 | * the start of data to be cloned is located. | |
1221 | * | |
1222 | * On return, bio_src is updated to refer to the bio in the source | |
1223 | * chain that contains first un-cloned byte, and *offset will | |
1224 | * contain the offset of that byte within that bio. | |
1225 | */ | |
1226 | static struct bio *bio_chain_clone_range(struct bio **bio_src, | |
1227 | unsigned int *offset, | |
1228 | unsigned int len, | |
1229 | gfp_t gfpmask) | |
1230 | { | |
1231 | struct bio *bi = *bio_src; | |
1232 | unsigned int off = *offset; | |
1233 | struct bio *chain = NULL; | |
1234 | struct bio **end; | |
1235 | ||
1236 | /* Build up a chain of clone bios up to the limit */ | |
1237 | ||
1238 | if (!bi || off >= bi->bi_size || !len) | |
1239 | return NULL; /* Nothing to clone */ | |
602adf40 | 1240 | |
f7760dad AE |
1241 | end = &chain; |
1242 | while (len) { | |
1243 | unsigned int bi_size; | |
1244 | struct bio *bio; | |
1245 | ||
f5400b7a AE |
1246 | if (!bi) { |
1247 | rbd_warn(NULL, "bio_chain exhausted with %u left", len); | |
f7760dad | 1248 | goto out_err; /* EINVAL; ran out of bio's */ |
f5400b7a | 1249 | } |
f7760dad AE |
1250 | bi_size = min_t(unsigned int, bi->bi_size - off, len); |
1251 | bio = bio_clone_range(bi, off, bi_size, gfpmask); | |
1252 | if (!bio) | |
1253 | goto out_err; /* ENOMEM */ | |
1254 | ||
1255 | *end = bio; | |
1256 | end = &bio->bi_next; | |
602adf40 | 1257 | |
f7760dad AE |
1258 | off += bi_size; |
1259 | if (off == bi->bi_size) { | |
1260 | bi = bi->bi_next; | |
1261 | off = 0; | |
1262 | } | |
1263 | len -= bi_size; | |
1264 | } | |
1265 | *bio_src = bi; | |
1266 | *offset = off; | |
1267 | ||
1268 | return chain; | |
1269 | out_err: | |
1270 | bio_chain_put(chain); | |
602adf40 | 1271 | |
602adf40 YS |
1272 | return NULL; |
1273 | } | |
1274 | ||
926f9b3f AE |
1275 | /* |
1276 | * The default/initial value for all object request flags is 0. For | |
1277 | * each flag, once its value is set to 1 it is never reset to 0 | |
1278 | * again. | |
1279 | */ | |
57acbaa7 | 1280 | static void obj_request_img_data_set(struct rbd_obj_request *obj_request) |
926f9b3f | 1281 | { |
57acbaa7 | 1282 | if (test_and_set_bit(OBJ_REQ_IMG_DATA, &obj_request->flags)) { |
926f9b3f AE |
1283 | struct rbd_device *rbd_dev; |
1284 | ||
57acbaa7 AE |
1285 | rbd_dev = obj_request->img_request->rbd_dev; |
1286 | rbd_warn(rbd_dev, "obj_request %p already marked img_data\n", | |
926f9b3f AE |
1287 | obj_request); |
1288 | } | |
1289 | } | |
1290 | ||
57acbaa7 | 1291 | static bool obj_request_img_data_test(struct rbd_obj_request *obj_request) |
926f9b3f AE |
1292 | { |
1293 | smp_mb(); | |
57acbaa7 | 1294 | return test_bit(OBJ_REQ_IMG_DATA, &obj_request->flags) != 0; |
926f9b3f AE |
1295 | } |
1296 | ||
57acbaa7 | 1297 | static void obj_request_done_set(struct rbd_obj_request *obj_request) |
6365d33a | 1298 | { |
57acbaa7 AE |
1299 | if (test_and_set_bit(OBJ_REQ_DONE, &obj_request->flags)) { |
1300 | struct rbd_device *rbd_dev = NULL; | |
6365d33a | 1301 | |
57acbaa7 AE |
1302 | if (obj_request_img_data_test(obj_request)) |
1303 | rbd_dev = obj_request->img_request->rbd_dev; | |
1304 | rbd_warn(rbd_dev, "obj_request %p already marked done\n", | |
6365d33a AE |
1305 | obj_request); |
1306 | } | |
1307 | } | |
1308 | ||
57acbaa7 | 1309 | static bool obj_request_done_test(struct rbd_obj_request *obj_request) |
6365d33a AE |
1310 | { |
1311 | smp_mb(); | |
57acbaa7 | 1312 | return test_bit(OBJ_REQ_DONE, &obj_request->flags) != 0; |
6365d33a AE |
1313 | } |
1314 | ||
5679c59f AE |
1315 | /* |
1316 | * This sets the KNOWN flag after (possibly) setting the EXISTS | |
1317 | * flag. The latter is set based on the "exists" value provided. | |
1318 | * | |
1319 | * Note that for our purposes once an object exists it never goes | |
1320 | * away again. It's possible that the response from two existence | |
1321 | * checks are separated by the creation of the target object, and | |
1322 | * the first ("doesn't exist") response arrives *after* the second | |
1323 | * ("does exist"). In that case we ignore the second one. | |
1324 | */ | |
1325 | static void obj_request_existence_set(struct rbd_obj_request *obj_request, | |
1326 | bool exists) | |
1327 | { | |
1328 | if (exists) | |
1329 | set_bit(OBJ_REQ_EXISTS, &obj_request->flags); | |
1330 | set_bit(OBJ_REQ_KNOWN, &obj_request->flags); | |
1331 | smp_mb(); | |
1332 | } | |
1333 | ||
1334 | static bool obj_request_known_test(struct rbd_obj_request *obj_request) | |
1335 | { | |
1336 | smp_mb(); | |
1337 | return test_bit(OBJ_REQ_KNOWN, &obj_request->flags) != 0; | |
1338 | } | |
1339 | ||
1340 | static bool obj_request_exists_test(struct rbd_obj_request *obj_request) | |
1341 | { | |
1342 | smp_mb(); | |
1343 | return test_bit(OBJ_REQ_EXISTS, &obj_request->flags) != 0; | |
1344 | } | |
1345 | ||
bf0d5f50 AE |
1346 | static void rbd_obj_request_get(struct rbd_obj_request *obj_request) |
1347 | { | |
37206ee5 AE |
1348 | dout("%s: obj %p (was %d)\n", __func__, obj_request, |
1349 | atomic_read(&obj_request->kref.refcount)); | |
bf0d5f50 AE |
1350 | kref_get(&obj_request->kref); |
1351 | } | |
1352 | ||
1353 | static void rbd_obj_request_destroy(struct kref *kref); | |
1354 | static void rbd_obj_request_put(struct rbd_obj_request *obj_request) | |
1355 | { | |
1356 | rbd_assert(obj_request != NULL); | |
37206ee5 AE |
1357 | dout("%s: obj %p (was %d)\n", __func__, obj_request, |
1358 | atomic_read(&obj_request->kref.refcount)); | |
bf0d5f50 AE |
1359 | kref_put(&obj_request->kref, rbd_obj_request_destroy); |
1360 | } | |
1361 | ||
bf0d5f50 AE |
1362 | static void rbd_img_request_destroy(struct kref *kref); |
1363 | static void rbd_img_request_put(struct rbd_img_request *img_request) | |
1364 | { | |
1365 | rbd_assert(img_request != NULL); | |
37206ee5 AE |
1366 | dout("%s: img %p (was %d)\n", __func__, img_request, |
1367 | atomic_read(&img_request->kref.refcount)); | |
bf0d5f50 AE |
1368 | kref_put(&img_request->kref, rbd_img_request_destroy); |
1369 | } | |
1370 | ||
1371 | static inline void rbd_img_obj_request_add(struct rbd_img_request *img_request, | |
1372 | struct rbd_obj_request *obj_request) | |
1373 | { | |
25dcf954 AE |
1374 | rbd_assert(obj_request->img_request == NULL); |
1375 | ||
b155e86c | 1376 | /* Image request now owns object's original reference */ |
bf0d5f50 | 1377 | obj_request->img_request = img_request; |
25dcf954 | 1378 | obj_request->which = img_request->obj_request_count; |
6365d33a AE |
1379 | rbd_assert(!obj_request_img_data_test(obj_request)); |
1380 | obj_request_img_data_set(obj_request); | |
bf0d5f50 | 1381 | rbd_assert(obj_request->which != BAD_WHICH); |
25dcf954 AE |
1382 | img_request->obj_request_count++; |
1383 | list_add_tail(&obj_request->links, &img_request->obj_requests); | |
37206ee5 AE |
1384 | dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request, |
1385 | obj_request->which); | |
bf0d5f50 AE |
1386 | } |
1387 | ||
1388 | static inline void rbd_img_obj_request_del(struct rbd_img_request *img_request, | |
1389 | struct rbd_obj_request *obj_request) | |
1390 | { | |
1391 | rbd_assert(obj_request->which != BAD_WHICH); | |
25dcf954 | 1392 | |
37206ee5 AE |
1393 | dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request, |
1394 | obj_request->which); | |
bf0d5f50 | 1395 | list_del(&obj_request->links); |
25dcf954 AE |
1396 | rbd_assert(img_request->obj_request_count > 0); |
1397 | img_request->obj_request_count--; | |
1398 | rbd_assert(obj_request->which == img_request->obj_request_count); | |
1399 | obj_request->which = BAD_WHICH; | |
6365d33a | 1400 | rbd_assert(obj_request_img_data_test(obj_request)); |
bf0d5f50 | 1401 | rbd_assert(obj_request->img_request == img_request); |
bf0d5f50 | 1402 | obj_request->img_request = NULL; |
25dcf954 | 1403 | obj_request->callback = NULL; |
bf0d5f50 AE |
1404 | rbd_obj_request_put(obj_request); |
1405 | } | |
1406 | ||
1407 | static bool obj_request_type_valid(enum obj_request_type type) | |
1408 | { | |
1409 | switch (type) { | |
9969ebc5 | 1410 | case OBJ_REQUEST_NODATA: |
bf0d5f50 | 1411 | case OBJ_REQUEST_BIO: |
788e2df3 | 1412 | case OBJ_REQUEST_PAGES: |
bf0d5f50 AE |
1413 | return true; |
1414 | default: | |
1415 | return false; | |
1416 | } | |
1417 | } | |
1418 | ||
bf0d5f50 AE |
1419 | static int rbd_obj_request_submit(struct ceph_osd_client *osdc, |
1420 | struct rbd_obj_request *obj_request) | |
1421 | { | |
37206ee5 AE |
1422 | dout("%s: osdc %p obj %p\n", __func__, osdc, obj_request); |
1423 | ||
bf0d5f50 AE |
1424 | return ceph_osdc_start_request(osdc, obj_request->osd_req, false); |
1425 | } | |
1426 | ||
1427 | static void rbd_img_request_complete(struct rbd_img_request *img_request) | |
1428 | { | |
55f27e09 | 1429 | |
37206ee5 | 1430 | dout("%s: img %p\n", __func__, img_request); |
55f27e09 AE |
1431 | |
1432 | /* | |
1433 | * If no error occurred, compute the aggregate transfer | |
1434 | * count for the image request. We could instead use | |
1435 | * atomic64_cmpxchg() to update it as each object request | |
1436 | * completes; not clear which way is better off hand. | |
1437 | */ | |
1438 | if (!img_request->result) { | |
1439 | struct rbd_obj_request *obj_request; | |
1440 | u64 xferred = 0; | |
1441 | ||
1442 | for_each_obj_request(img_request, obj_request) | |
1443 | xferred += obj_request->xferred; | |
1444 | img_request->xferred = xferred; | |
1445 | } | |
1446 | ||
bf0d5f50 AE |
1447 | if (img_request->callback) |
1448 | img_request->callback(img_request); | |
1449 | else | |
1450 | rbd_img_request_put(img_request); | |
1451 | } | |
1452 | ||
788e2df3 AE |
1453 | /* Caller is responsible for rbd_obj_request_destroy(obj_request) */ |
1454 | ||
1455 | static int rbd_obj_request_wait(struct rbd_obj_request *obj_request) | |
1456 | { | |
37206ee5 AE |
1457 | dout("%s: obj %p\n", __func__, obj_request); |
1458 | ||
788e2df3 AE |
1459 | return wait_for_completion_interruptible(&obj_request->completion); |
1460 | } | |
1461 | ||
0c425248 AE |
1462 | /* |
1463 | * The default/initial value for all image request flags is 0. Each | |
1464 | * is conditionally set to 1 at image request initialization time | |
1465 | * and currently never change thereafter. | |
1466 | */ | |
1467 | static void img_request_write_set(struct rbd_img_request *img_request) | |
1468 | { | |
1469 | set_bit(IMG_REQ_WRITE, &img_request->flags); | |
1470 | smp_mb(); | |
1471 | } | |
1472 | ||
1473 | static bool img_request_write_test(struct rbd_img_request *img_request) | |
1474 | { | |
1475 | smp_mb(); | |
1476 | return test_bit(IMG_REQ_WRITE, &img_request->flags) != 0; | |
1477 | } | |
1478 | ||
9849e986 AE |
1479 | static void img_request_child_set(struct rbd_img_request *img_request) |
1480 | { | |
1481 | set_bit(IMG_REQ_CHILD, &img_request->flags); | |
1482 | smp_mb(); | |
1483 | } | |
1484 | ||
1485 | static bool img_request_child_test(struct rbd_img_request *img_request) | |
1486 | { | |
1487 | smp_mb(); | |
1488 | return test_bit(IMG_REQ_CHILD, &img_request->flags) != 0; | |
1489 | } | |
1490 | ||
d0b2e944 AE |
1491 | static void img_request_layered_set(struct rbd_img_request *img_request) |
1492 | { | |
1493 | set_bit(IMG_REQ_LAYERED, &img_request->flags); | |
1494 | smp_mb(); | |
1495 | } | |
1496 | ||
1497 | static bool img_request_layered_test(struct rbd_img_request *img_request) | |
1498 | { | |
1499 | smp_mb(); | |
1500 | return test_bit(IMG_REQ_LAYERED, &img_request->flags) != 0; | |
1501 | } | |
1502 | ||
6e2a4505 AE |
1503 | static void |
1504 | rbd_img_obj_request_read_callback(struct rbd_obj_request *obj_request) | |
1505 | { | |
b9434c5b AE |
1506 | u64 xferred = obj_request->xferred; |
1507 | u64 length = obj_request->length; | |
1508 | ||
6e2a4505 AE |
1509 | dout("%s: obj %p img %p result %d %llu/%llu\n", __func__, |
1510 | obj_request, obj_request->img_request, obj_request->result, | |
b9434c5b | 1511 | xferred, length); |
6e2a4505 AE |
1512 | /* |
1513 | * ENOENT means a hole in the image. We zero-fill the | |
1514 | * entire length of the request. A short read also implies | |
1515 | * zero-fill to the end of the request. Either way we | |
1516 | * update the xferred count to indicate the whole request | |
1517 | * was satisfied. | |
1518 | */ | |
b9434c5b | 1519 | rbd_assert(obj_request->type != OBJ_REQUEST_NODATA); |
6e2a4505 | 1520 | if (obj_request->result == -ENOENT) { |
b9434c5b AE |
1521 | if (obj_request->type == OBJ_REQUEST_BIO) |
1522 | zero_bio_chain(obj_request->bio_list, 0); | |
1523 | else | |
1524 | zero_pages(obj_request->pages, 0, length); | |
6e2a4505 | 1525 | obj_request->result = 0; |
b9434c5b AE |
1526 | obj_request->xferred = length; |
1527 | } else if (xferred < length && !obj_request->result) { | |
1528 | if (obj_request->type == OBJ_REQUEST_BIO) | |
1529 | zero_bio_chain(obj_request->bio_list, xferred); | |
1530 | else | |
1531 | zero_pages(obj_request->pages, xferred, length); | |
1532 | obj_request->xferred = length; | |
6e2a4505 AE |
1533 | } |
1534 | obj_request_done_set(obj_request); | |
1535 | } | |
1536 | ||
bf0d5f50 AE |
1537 | static void rbd_obj_request_complete(struct rbd_obj_request *obj_request) |
1538 | { | |
37206ee5 AE |
1539 | dout("%s: obj %p cb %p\n", __func__, obj_request, |
1540 | obj_request->callback); | |
bf0d5f50 AE |
1541 | if (obj_request->callback) |
1542 | obj_request->callback(obj_request); | |
788e2df3 AE |
1543 | else |
1544 | complete_all(&obj_request->completion); | |
bf0d5f50 AE |
1545 | } |
1546 | ||
c47f9371 | 1547 | static void rbd_osd_trivial_callback(struct rbd_obj_request *obj_request) |
39bf2c5d AE |
1548 | { |
1549 | dout("%s: obj %p\n", __func__, obj_request); | |
1550 | obj_request_done_set(obj_request); | |
1551 | } | |
1552 | ||
c47f9371 | 1553 | static void rbd_osd_read_callback(struct rbd_obj_request *obj_request) |
bf0d5f50 | 1554 | { |
57acbaa7 | 1555 | struct rbd_img_request *img_request = NULL; |
a9e8ba2c | 1556 | struct rbd_device *rbd_dev = NULL; |
57acbaa7 AE |
1557 | bool layered = false; |
1558 | ||
1559 | if (obj_request_img_data_test(obj_request)) { | |
1560 | img_request = obj_request->img_request; | |
1561 | layered = img_request && img_request_layered_test(img_request); | |
a9e8ba2c | 1562 | rbd_dev = img_request->rbd_dev; |
57acbaa7 | 1563 | } |
8b3e1a56 AE |
1564 | |
1565 | dout("%s: obj %p img %p result %d %llu/%llu\n", __func__, | |
1566 | obj_request, img_request, obj_request->result, | |
1567 | obj_request->xferred, obj_request->length); | |
a9e8ba2c AE |
1568 | if (layered && obj_request->result == -ENOENT && |
1569 | obj_request->img_offset < rbd_dev->parent_overlap) | |
8b3e1a56 AE |
1570 | rbd_img_parent_read(obj_request); |
1571 | else if (img_request) | |
6e2a4505 AE |
1572 | rbd_img_obj_request_read_callback(obj_request); |
1573 | else | |
1574 | obj_request_done_set(obj_request); | |
bf0d5f50 AE |
1575 | } |
1576 | ||
c47f9371 | 1577 | static void rbd_osd_write_callback(struct rbd_obj_request *obj_request) |
bf0d5f50 | 1578 | { |
1b83bef2 SW |
1579 | dout("%s: obj %p result %d %llu\n", __func__, obj_request, |
1580 | obj_request->result, obj_request->length); | |
1581 | /* | |
8b3e1a56 AE |
1582 | * There is no such thing as a successful short write. Set |
1583 | * it to our originally-requested length. | |
1b83bef2 SW |
1584 | */ |
1585 | obj_request->xferred = obj_request->length; | |
07741308 | 1586 | obj_request_done_set(obj_request); |
bf0d5f50 AE |
1587 | } |
1588 | ||
fbfab539 AE |
1589 | /* |
1590 | * For a simple stat call there's nothing to do. We'll do more if | |
1591 | * this is part of a write sequence for a layered image. | |
1592 | */ | |
c47f9371 | 1593 | static void rbd_osd_stat_callback(struct rbd_obj_request *obj_request) |
fbfab539 | 1594 | { |
37206ee5 | 1595 | dout("%s: obj %p\n", __func__, obj_request); |
fbfab539 AE |
1596 | obj_request_done_set(obj_request); |
1597 | } | |
1598 | ||
bf0d5f50 AE |
1599 | static void rbd_osd_req_callback(struct ceph_osd_request *osd_req, |
1600 | struct ceph_msg *msg) | |
1601 | { | |
1602 | struct rbd_obj_request *obj_request = osd_req->r_priv; | |
bf0d5f50 AE |
1603 | u16 opcode; |
1604 | ||
37206ee5 | 1605 | dout("%s: osd_req %p msg %p\n", __func__, osd_req, msg); |
bf0d5f50 | 1606 | rbd_assert(osd_req == obj_request->osd_req); |
57acbaa7 AE |
1607 | if (obj_request_img_data_test(obj_request)) { |
1608 | rbd_assert(obj_request->img_request); | |
1609 | rbd_assert(obj_request->which != BAD_WHICH); | |
1610 | } else { | |
1611 | rbd_assert(obj_request->which == BAD_WHICH); | |
1612 | } | |
bf0d5f50 | 1613 | |
1b83bef2 SW |
1614 | if (osd_req->r_result < 0) |
1615 | obj_request->result = osd_req->r_result; | |
bf0d5f50 | 1616 | |
0eefd470 | 1617 | BUG_ON(osd_req->r_num_ops > 2); |
bf0d5f50 | 1618 | |
c47f9371 AE |
1619 | /* |
1620 | * We support a 64-bit length, but ultimately it has to be | |
1621 | * passed to blk_end_request(), which takes an unsigned int. | |
1622 | */ | |
1b83bef2 | 1623 | obj_request->xferred = osd_req->r_reply_op_len[0]; |
8b3e1a56 | 1624 | rbd_assert(obj_request->xferred < (u64)UINT_MAX); |
79528734 | 1625 | opcode = osd_req->r_ops[0].op; |
bf0d5f50 AE |
1626 | switch (opcode) { |
1627 | case CEPH_OSD_OP_READ: | |
c47f9371 | 1628 | rbd_osd_read_callback(obj_request); |
bf0d5f50 AE |
1629 | break; |
1630 | case CEPH_OSD_OP_WRITE: | |
c47f9371 | 1631 | rbd_osd_write_callback(obj_request); |
bf0d5f50 | 1632 | break; |
fbfab539 | 1633 | case CEPH_OSD_OP_STAT: |
c47f9371 | 1634 | rbd_osd_stat_callback(obj_request); |
fbfab539 | 1635 | break; |
36be9a76 | 1636 | case CEPH_OSD_OP_CALL: |
b8d70035 | 1637 | case CEPH_OSD_OP_NOTIFY_ACK: |
9969ebc5 | 1638 | case CEPH_OSD_OP_WATCH: |
c47f9371 | 1639 | rbd_osd_trivial_callback(obj_request); |
9969ebc5 | 1640 | break; |
bf0d5f50 AE |
1641 | default: |
1642 | rbd_warn(NULL, "%s: unsupported op %hu\n", | |
1643 | obj_request->object_name, (unsigned short) opcode); | |
1644 | break; | |
1645 | } | |
1646 | ||
07741308 | 1647 | if (obj_request_done_test(obj_request)) |
bf0d5f50 AE |
1648 | rbd_obj_request_complete(obj_request); |
1649 | } | |
1650 | ||
9d4df01f | 1651 | static void rbd_osd_req_format_read(struct rbd_obj_request *obj_request) |
430c28c3 AE |
1652 | { |
1653 | struct rbd_img_request *img_request = obj_request->img_request; | |
8c042b0d | 1654 | struct ceph_osd_request *osd_req = obj_request->osd_req; |
9d4df01f | 1655 | u64 snap_id; |
430c28c3 | 1656 | |
8c042b0d | 1657 | rbd_assert(osd_req != NULL); |
430c28c3 | 1658 | |
9d4df01f | 1659 | snap_id = img_request ? img_request->snap_id : CEPH_NOSNAP; |
8c042b0d | 1660 | ceph_osdc_build_request(osd_req, obj_request->offset, |
9d4df01f AE |
1661 | NULL, snap_id, NULL); |
1662 | } | |
1663 | ||
1664 | static void rbd_osd_req_format_write(struct rbd_obj_request *obj_request) | |
1665 | { | |
1666 | struct rbd_img_request *img_request = obj_request->img_request; | |
1667 | struct ceph_osd_request *osd_req = obj_request->osd_req; | |
1668 | struct ceph_snap_context *snapc; | |
1669 | struct timespec mtime = CURRENT_TIME; | |
1670 | ||
1671 | rbd_assert(osd_req != NULL); | |
1672 | ||
1673 | snapc = img_request ? img_request->snapc : NULL; | |
1674 | ceph_osdc_build_request(osd_req, obj_request->offset, | |
1675 | snapc, CEPH_NOSNAP, &mtime); | |
430c28c3 AE |
1676 | } |
1677 | ||
bf0d5f50 AE |
1678 | static struct ceph_osd_request *rbd_osd_req_create( |
1679 | struct rbd_device *rbd_dev, | |
1680 | bool write_request, | |
430c28c3 | 1681 | struct rbd_obj_request *obj_request) |
bf0d5f50 | 1682 | { |
bf0d5f50 AE |
1683 | struct ceph_snap_context *snapc = NULL; |
1684 | struct ceph_osd_client *osdc; | |
1685 | struct ceph_osd_request *osd_req; | |
bf0d5f50 | 1686 | |
6365d33a AE |
1687 | if (obj_request_img_data_test(obj_request)) { |
1688 | struct rbd_img_request *img_request = obj_request->img_request; | |
1689 | ||
0c425248 AE |
1690 | rbd_assert(write_request == |
1691 | img_request_write_test(img_request)); | |
1692 | if (write_request) | |
bf0d5f50 | 1693 | snapc = img_request->snapc; |
bf0d5f50 AE |
1694 | } |
1695 | ||
1696 | /* Allocate and initialize the request, for the single op */ | |
1697 | ||
1698 | osdc = &rbd_dev->rbd_client->client->osdc; | |
1699 | osd_req = ceph_osdc_alloc_request(osdc, snapc, 1, false, GFP_ATOMIC); | |
1700 | if (!osd_req) | |
1701 | return NULL; /* ENOMEM */ | |
bf0d5f50 | 1702 | |
430c28c3 | 1703 | if (write_request) |
bf0d5f50 | 1704 | osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; |
430c28c3 | 1705 | else |
bf0d5f50 | 1706 | osd_req->r_flags = CEPH_OSD_FLAG_READ; |
bf0d5f50 AE |
1707 | |
1708 | osd_req->r_callback = rbd_osd_req_callback; | |
1709 | osd_req->r_priv = obj_request; | |
1710 | ||
1711 | osd_req->r_oid_len = strlen(obj_request->object_name); | |
1712 | rbd_assert(osd_req->r_oid_len < sizeof (osd_req->r_oid)); | |
1713 | memcpy(osd_req->r_oid, obj_request->object_name, osd_req->r_oid_len); | |
1714 | ||
1715 | osd_req->r_file_layout = rbd_dev->layout; /* struct */ | |
1716 | ||
bf0d5f50 AE |
1717 | return osd_req; |
1718 | } | |
1719 | ||
0eefd470 AE |
1720 | /* |
1721 | * Create a copyup osd request based on the information in the | |
1722 | * object request supplied. A copyup request has two osd ops, | |
1723 | * a copyup method call, and a "normal" write request. | |
1724 | */ | |
1725 | static struct ceph_osd_request * | |
1726 | rbd_osd_req_create_copyup(struct rbd_obj_request *obj_request) | |
1727 | { | |
1728 | struct rbd_img_request *img_request; | |
1729 | struct ceph_snap_context *snapc; | |
1730 | struct rbd_device *rbd_dev; | |
1731 | struct ceph_osd_client *osdc; | |
1732 | struct ceph_osd_request *osd_req; | |
1733 | ||
1734 | rbd_assert(obj_request_img_data_test(obj_request)); | |
1735 | img_request = obj_request->img_request; | |
1736 | rbd_assert(img_request); | |
1737 | rbd_assert(img_request_write_test(img_request)); | |
1738 | ||
1739 | /* Allocate and initialize the request, for the two ops */ | |
1740 | ||
1741 | snapc = img_request->snapc; | |
1742 | rbd_dev = img_request->rbd_dev; | |
1743 | osdc = &rbd_dev->rbd_client->client->osdc; | |
1744 | osd_req = ceph_osdc_alloc_request(osdc, snapc, 2, false, GFP_ATOMIC); | |
1745 | if (!osd_req) | |
1746 | return NULL; /* ENOMEM */ | |
1747 | ||
1748 | osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; | |
1749 | osd_req->r_callback = rbd_osd_req_callback; | |
1750 | osd_req->r_priv = obj_request; | |
1751 | ||
1752 | osd_req->r_oid_len = strlen(obj_request->object_name); | |
1753 | rbd_assert(osd_req->r_oid_len < sizeof (osd_req->r_oid)); | |
1754 | memcpy(osd_req->r_oid, obj_request->object_name, osd_req->r_oid_len); | |
1755 | ||
1756 | osd_req->r_file_layout = rbd_dev->layout; /* struct */ | |
1757 | ||
1758 | return osd_req; | |
1759 | } | |
1760 | ||
1761 | ||
bf0d5f50 AE |
1762 | static void rbd_osd_req_destroy(struct ceph_osd_request *osd_req) |
1763 | { | |
1764 | ceph_osdc_put_request(osd_req); | |
1765 | } | |
1766 | ||
1767 | /* object_name is assumed to be a non-null pointer and NUL-terminated */ | |
1768 | ||
1769 | static struct rbd_obj_request *rbd_obj_request_create(const char *object_name, | |
1770 | u64 offset, u64 length, | |
1771 | enum obj_request_type type) | |
1772 | { | |
1773 | struct rbd_obj_request *obj_request; | |
1774 | size_t size; | |
1775 | char *name; | |
1776 | ||
1777 | rbd_assert(obj_request_type_valid(type)); | |
1778 | ||
1779 | size = strlen(object_name) + 1; | |
f907ad55 AE |
1780 | name = kmalloc(size, GFP_KERNEL); |
1781 | if (!name) | |
bf0d5f50 AE |
1782 | return NULL; |
1783 | ||
868311b1 | 1784 | obj_request = kmem_cache_zalloc(rbd_obj_request_cache, GFP_KERNEL); |
f907ad55 AE |
1785 | if (!obj_request) { |
1786 | kfree(name); | |
1787 | return NULL; | |
1788 | } | |
1789 | ||
bf0d5f50 AE |
1790 | obj_request->object_name = memcpy(name, object_name, size); |
1791 | obj_request->offset = offset; | |
1792 | obj_request->length = length; | |
926f9b3f | 1793 | obj_request->flags = 0; |
bf0d5f50 AE |
1794 | obj_request->which = BAD_WHICH; |
1795 | obj_request->type = type; | |
1796 | INIT_LIST_HEAD(&obj_request->links); | |
788e2df3 | 1797 | init_completion(&obj_request->completion); |
bf0d5f50 AE |
1798 | kref_init(&obj_request->kref); |
1799 | ||
37206ee5 AE |
1800 | dout("%s: \"%s\" %llu/%llu %d -> obj %p\n", __func__, object_name, |
1801 | offset, length, (int)type, obj_request); | |
1802 | ||
bf0d5f50 AE |
1803 | return obj_request; |
1804 | } | |
1805 | ||
1806 | static void rbd_obj_request_destroy(struct kref *kref) | |
1807 | { | |
1808 | struct rbd_obj_request *obj_request; | |
1809 | ||
1810 | obj_request = container_of(kref, struct rbd_obj_request, kref); | |
1811 | ||
37206ee5 AE |
1812 | dout("%s: obj %p\n", __func__, obj_request); |
1813 | ||
bf0d5f50 AE |
1814 | rbd_assert(obj_request->img_request == NULL); |
1815 | rbd_assert(obj_request->which == BAD_WHICH); | |
1816 | ||
1817 | if (obj_request->osd_req) | |
1818 | rbd_osd_req_destroy(obj_request->osd_req); | |
1819 | ||
1820 | rbd_assert(obj_request_type_valid(obj_request->type)); | |
1821 | switch (obj_request->type) { | |
9969ebc5 AE |
1822 | case OBJ_REQUEST_NODATA: |
1823 | break; /* Nothing to do */ | |
bf0d5f50 AE |
1824 | case OBJ_REQUEST_BIO: |
1825 | if (obj_request->bio_list) | |
1826 | bio_chain_put(obj_request->bio_list); | |
1827 | break; | |
788e2df3 AE |
1828 | case OBJ_REQUEST_PAGES: |
1829 | if (obj_request->pages) | |
1830 | ceph_release_page_vector(obj_request->pages, | |
1831 | obj_request->page_count); | |
1832 | break; | |
bf0d5f50 AE |
1833 | } |
1834 | ||
f907ad55 | 1835 | kfree(obj_request->object_name); |
868311b1 AE |
1836 | obj_request->object_name = NULL; |
1837 | kmem_cache_free(rbd_obj_request_cache, obj_request); | |
bf0d5f50 AE |
1838 | } |
1839 | ||
1840 | /* | |
1841 | * Caller is responsible for filling in the list of object requests | |
1842 | * that comprises the image request, and the Linux request pointer | |
1843 | * (if there is one). | |
1844 | */ | |
cc344fa1 AE |
1845 | static struct rbd_img_request *rbd_img_request_create( |
1846 | struct rbd_device *rbd_dev, | |
bf0d5f50 | 1847 | u64 offset, u64 length, |
9849e986 AE |
1848 | bool write_request, |
1849 | bool child_request) | |
bf0d5f50 AE |
1850 | { |
1851 | struct rbd_img_request *img_request; | |
bf0d5f50 | 1852 | |
1c2a9dfe | 1853 | img_request = kmem_cache_alloc(rbd_img_request_cache, GFP_ATOMIC); |
bf0d5f50 AE |
1854 | if (!img_request) |
1855 | return NULL; | |
1856 | ||
1857 | if (write_request) { | |
1858 | down_read(&rbd_dev->header_rwsem); | |
812164f8 | 1859 | ceph_get_snap_context(rbd_dev->header.snapc); |
bf0d5f50 | 1860 | up_read(&rbd_dev->header_rwsem); |
bf0d5f50 AE |
1861 | } |
1862 | ||
1863 | img_request->rq = NULL; | |
1864 | img_request->rbd_dev = rbd_dev; | |
1865 | img_request->offset = offset; | |
1866 | img_request->length = length; | |
0c425248 AE |
1867 | img_request->flags = 0; |
1868 | if (write_request) { | |
1869 | img_request_write_set(img_request); | |
468521c1 | 1870 | img_request->snapc = rbd_dev->header.snapc; |
0c425248 | 1871 | } else { |
bf0d5f50 | 1872 | img_request->snap_id = rbd_dev->spec->snap_id; |
0c425248 | 1873 | } |
9849e986 AE |
1874 | if (child_request) |
1875 | img_request_child_set(img_request); | |
642a2537 | 1876 | if (rbd_dev->parent_overlap) |
d0b2e944 | 1877 | img_request_layered_set(img_request); |
bf0d5f50 AE |
1878 | spin_lock_init(&img_request->completion_lock); |
1879 | img_request->next_completion = 0; | |
1880 | img_request->callback = NULL; | |
a5a337d4 | 1881 | img_request->result = 0; |
bf0d5f50 AE |
1882 | img_request->obj_request_count = 0; |
1883 | INIT_LIST_HEAD(&img_request->obj_requests); | |
1884 | kref_init(&img_request->kref); | |
1885 | ||
37206ee5 AE |
1886 | dout("%s: rbd_dev %p %s %llu/%llu -> img %p\n", __func__, rbd_dev, |
1887 | write_request ? "write" : "read", offset, length, | |
1888 | img_request); | |
1889 | ||
bf0d5f50 AE |
1890 | return img_request; |
1891 | } | |
1892 | ||
1893 | static void rbd_img_request_destroy(struct kref *kref) | |
1894 | { | |
1895 | struct rbd_img_request *img_request; | |
1896 | struct rbd_obj_request *obj_request; | |
1897 | struct rbd_obj_request *next_obj_request; | |
1898 | ||
1899 | img_request = container_of(kref, struct rbd_img_request, kref); | |
1900 | ||
37206ee5 AE |
1901 | dout("%s: img %p\n", __func__, img_request); |
1902 | ||
bf0d5f50 AE |
1903 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) |
1904 | rbd_img_obj_request_del(img_request, obj_request); | |
25dcf954 | 1905 | rbd_assert(img_request->obj_request_count == 0); |
bf0d5f50 | 1906 | |
0c425248 | 1907 | if (img_request_write_test(img_request)) |
812164f8 | 1908 | ceph_put_snap_context(img_request->snapc); |
bf0d5f50 | 1909 | |
8b3e1a56 AE |
1910 | if (img_request_child_test(img_request)) |
1911 | rbd_obj_request_put(img_request->obj_request); | |
1912 | ||
1c2a9dfe | 1913 | kmem_cache_free(rbd_img_request_cache, img_request); |
bf0d5f50 AE |
1914 | } |
1915 | ||
1217857f AE |
1916 | static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request) |
1917 | { | |
6365d33a | 1918 | struct rbd_img_request *img_request; |
1217857f AE |
1919 | unsigned int xferred; |
1920 | int result; | |
8b3e1a56 | 1921 | bool more; |
1217857f | 1922 | |
6365d33a AE |
1923 | rbd_assert(obj_request_img_data_test(obj_request)); |
1924 | img_request = obj_request->img_request; | |
1925 | ||
1217857f AE |
1926 | rbd_assert(obj_request->xferred <= (u64)UINT_MAX); |
1927 | xferred = (unsigned int)obj_request->xferred; | |
1928 | result = obj_request->result; | |
1929 | if (result) { | |
1930 | struct rbd_device *rbd_dev = img_request->rbd_dev; | |
1931 | ||
1932 | rbd_warn(rbd_dev, "%s %llx at %llx (%llx)\n", | |
1933 | img_request_write_test(img_request) ? "write" : "read", | |
1934 | obj_request->length, obj_request->img_offset, | |
1935 | obj_request->offset); | |
1936 | rbd_warn(rbd_dev, " result %d xferred %x\n", | |
1937 | result, xferred); | |
1938 | if (!img_request->result) | |
1939 | img_request->result = result; | |
1940 | } | |
1941 | ||
f1a4739f AE |
1942 | /* Image object requests don't own their page array */ |
1943 | ||
1944 | if (obj_request->type == OBJ_REQUEST_PAGES) { | |
1945 | obj_request->pages = NULL; | |
1946 | obj_request->page_count = 0; | |
1947 | } | |
1948 | ||
8b3e1a56 AE |
1949 | if (img_request_child_test(img_request)) { |
1950 | rbd_assert(img_request->obj_request != NULL); | |
1951 | more = obj_request->which < img_request->obj_request_count - 1; | |
1952 | } else { | |
1953 | rbd_assert(img_request->rq != NULL); | |
1954 | more = blk_end_request(img_request->rq, result, xferred); | |
1955 | } | |
1956 | ||
1957 | return more; | |
1217857f AE |
1958 | } |
1959 | ||
2169238d AE |
1960 | static void rbd_img_obj_callback(struct rbd_obj_request *obj_request) |
1961 | { | |
1962 | struct rbd_img_request *img_request; | |
1963 | u32 which = obj_request->which; | |
1964 | bool more = true; | |
1965 | ||
6365d33a | 1966 | rbd_assert(obj_request_img_data_test(obj_request)); |
2169238d AE |
1967 | img_request = obj_request->img_request; |
1968 | ||
1969 | dout("%s: img %p obj %p\n", __func__, img_request, obj_request); | |
1970 | rbd_assert(img_request != NULL); | |
2169238d AE |
1971 | rbd_assert(img_request->obj_request_count > 0); |
1972 | rbd_assert(which != BAD_WHICH); | |
1973 | rbd_assert(which < img_request->obj_request_count); | |
1974 | rbd_assert(which >= img_request->next_completion); | |
1975 | ||
1976 | spin_lock_irq(&img_request->completion_lock); | |
1977 | if (which != img_request->next_completion) | |
1978 | goto out; | |
1979 | ||
1980 | for_each_obj_request_from(img_request, obj_request) { | |
2169238d AE |
1981 | rbd_assert(more); |
1982 | rbd_assert(which < img_request->obj_request_count); | |
1983 | ||
1984 | if (!obj_request_done_test(obj_request)) | |
1985 | break; | |
1217857f | 1986 | more = rbd_img_obj_end_request(obj_request); |
2169238d AE |
1987 | which++; |
1988 | } | |
1989 | ||
1990 | rbd_assert(more ^ (which == img_request->obj_request_count)); | |
1991 | img_request->next_completion = which; | |
1992 | out: | |
1993 | spin_unlock_irq(&img_request->completion_lock); | |
1994 | ||
1995 | if (!more) | |
1996 | rbd_img_request_complete(img_request); | |
1997 | } | |
1998 | ||
f1a4739f AE |
1999 | /* |
2000 | * Split up an image request into one or more object requests, each | |
2001 | * to a different object. The "type" parameter indicates whether | |
2002 | * "data_desc" is the pointer to the head of a list of bio | |
2003 | * structures, or the base of a page array. In either case this | |
2004 | * function assumes data_desc describes memory sufficient to hold | |
2005 | * all data described by the image request. | |
2006 | */ | |
2007 | static int rbd_img_request_fill(struct rbd_img_request *img_request, | |
2008 | enum obj_request_type type, | |
2009 | void *data_desc) | |
bf0d5f50 AE |
2010 | { |
2011 | struct rbd_device *rbd_dev = img_request->rbd_dev; | |
2012 | struct rbd_obj_request *obj_request = NULL; | |
2013 | struct rbd_obj_request *next_obj_request; | |
0c425248 | 2014 | bool write_request = img_request_write_test(img_request); |
f1a4739f AE |
2015 | struct bio *bio_list; |
2016 | unsigned int bio_offset = 0; | |
2017 | struct page **pages; | |
7da22d29 | 2018 | u64 img_offset; |
bf0d5f50 AE |
2019 | u64 resid; |
2020 | u16 opcode; | |
2021 | ||
f1a4739f AE |
2022 | dout("%s: img %p type %d data_desc %p\n", __func__, img_request, |
2023 | (int)type, data_desc); | |
37206ee5 | 2024 | |
430c28c3 | 2025 | opcode = write_request ? CEPH_OSD_OP_WRITE : CEPH_OSD_OP_READ; |
7da22d29 | 2026 | img_offset = img_request->offset; |
bf0d5f50 | 2027 | resid = img_request->length; |
4dda41d3 | 2028 | rbd_assert(resid > 0); |
f1a4739f AE |
2029 | |
2030 | if (type == OBJ_REQUEST_BIO) { | |
2031 | bio_list = data_desc; | |
2032 | rbd_assert(img_offset == bio_list->bi_sector << SECTOR_SHIFT); | |
2033 | } else { | |
2034 | rbd_assert(type == OBJ_REQUEST_PAGES); | |
2035 | pages = data_desc; | |
2036 | } | |
2037 | ||
bf0d5f50 | 2038 | while (resid) { |
2fa12320 | 2039 | struct ceph_osd_request *osd_req; |
bf0d5f50 | 2040 | const char *object_name; |
bf0d5f50 AE |
2041 | u64 offset; |
2042 | u64 length; | |
2043 | ||
7da22d29 | 2044 | object_name = rbd_segment_name(rbd_dev, img_offset); |
bf0d5f50 AE |
2045 | if (!object_name) |
2046 | goto out_unwind; | |
7da22d29 AE |
2047 | offset = rbd_segment_offset(rbd_dev, img_offset); |
2048 | length = rbd_segment_length(rbd_dev, img_offset, resid); | |
bf0d5f50 | 2049 | obj_request = rbd_obj_request_create(object_name, |
f1a4739f | 2050 | offset, length, type); |
78c2a44a AE |
2051 | /* object request has its own copy of the object name */ |
2052 | rbd_segment_name_free(object_name); | |
bf0d5f50 AE |
2053 | if (!obj_request) |
2054 | goto out_unwind; | |
2055 | ||
f1a4739f AE |
2056 | if (type == OBJ_REQUEST_BIO) { |
2057 | unsigned int clone_size; | |
2058 | ||
2059 | rbd_assert(length <= (u64)UINT_MAX); | |
2060 | clone_size = (unsigned int)length; | |
2061 | obj_request->bio_list = | |
2062 | bio_chain_clone_range(&bio_list, | |
2063 | &bio_offset, | |
2064 | clone_size, | |
2065 | GFP_ATOMIC); | |
2066 | if (!obj_request->bio_list) | |
2067 | goto out_partial; | |
2068 | } else { | |
2069 | unsigned int page_count; | |
2070 | ||
2071 | obj_request->pages = pages; | |
2072 | page_count = (u32)calc_pages_for(offset, length); | |
2073 | obj_request->page_count = page_count; | |
2074 | if ((offset + length) & ~PAGE_MASK) | |
2075 | page_count--; /* more on last page */ | |
2076 | pages += page_count; | |
2077 | } | |
bf0d5f50 | 2078 | |
2fa12320 AE |
2079 | osd_req = rbd_osd_req_create(rbd_dev, write_request, |
2080 | obj_request); | |
2081 | if (!osd_req) | |
bf0d5f50 | 2082 | goto out_partial; |
2fa12320 | 2083 | obj_request->osd_req = osd_req; |
2169238d | 2084 | obj_request->callback = rbd_img_obj_callback; |
430c28c3 | 2085 | |
2fa12320 AE |
2086 | osd_req_op_extent_init(osd_req, 0, opcode, offset, length, |
2087 | 0, 0); | |
f1a4739f AE |
2088 | if (type == OBJ_REQUEST_BIO) |
2089 | osd_req_op_extent_osd_data_bio(osd_req, 0, | |
2090 | obj_request->bio_list, length); | |
2091 | else | |
2092 | osd_req_op_extent_osd_data_pages(osd_req, 0, | |
2093 | obj_request->pages, length, | |
2094 | offset & ~PAGE_MASK, false, false); | |
9d4df01f AE |
2095 | |
2096 | if (write_request) | |
2097 | rbd_osd_req_format_write(obj_request); | |
2098 | else | |
2099 | rbd_osd_req_format_read(obj_request); | |
430c28c3 | 2100 | |
7da22d29 | 2101 | obj_request->img_offset = img_offset; |
bf0d5f50 AE |
2102 | rbd_img_obj_request_add(img_request, obj_request); |
2103 | ||
7da22d29 | 2104 | img_offset += length; |
bf0d5f50 AE |
2105 | resid -= length; |
2106 | } | |
2107 | ||
2108 | return 0; | |
2109 | ||
2110 | out_partial: | |
2111 | rbd_obj_request_put(obj_request); | |
2112 | out_unwind: | |
2113 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) | |
2114 | rbd_obj_request_put(obj_request); | |
2115 | ||
2116 | return -ENOMEM; | |
2117 | } | |
2118 | ||
0eefd470 AE |
2119 | static void |
2120 | rbd_img_obj_copyup_callback(struct rbd_obj_request *obj_request) | |
2121 | { | |
2122 | struct rbd_img_request *img_request; | |
2123 | struct rbd_device *rbd_dev; | |
ebda6408 | 2124 | struct page **pages; |
0eefd470 AE |
2125 | u32 page_count; |
2126 | ||
2127 | rbd_assert(obj_request->type == OBJ_REQUEST_BIO); | |
2128 | rbd_assert(obj_request_img_data_test(obj_request)); | |
2129 | img_request = obj_request->img_request; | |
2130 | rbd_assert(img_request); | |
2131 | ||
2132 | rbd_dev = img_request->rbd_dev; | |
2133 | rbd_assert(rbd_dev); | |
0eefd470 | 2134 | |
ebda6408 AE |
2135 | pages = obj_request->copyup_pages; |
2136 | rbd_assert(pages != NULL); | |
0eefd470 | 2137 | obj_request->copyup_pages = NULL; |
ebda6408 AE |
2138 | page_count = obj_request->copyup_page_count; |
2139 | rbd_assert(page_count); | |
2140 | obj_request->copyup_page_count = 0; | |
2141 | ceph_release_page_vector(pages, page_count); | |
0eefd470 AE |
2142 | |
2143 | /* | |
2144 | * We want the transfer count to reflect the size of the | |
2145 | * original write request. There is no such thing as a | |
2146 | * successful short write, so if the request was successful | |
2147 | * we can just set it to the originally-requested length. | |
2148 | */ | |
2149 | if (!obj_request->result) | |
2150 | obj_request->xferred = obj_request->length; | |
2151 | ||
2152 | /* Finish up with the normal image object callback */ | |
2153 | ||
2154 | rbd_img_obj_callback(obj_request); | |
2155 | } | |
2156 | ||
3d7efd18 AE |
2157 | static void |
2158 | rbd_img_obj_parent_read_full_callback(struct rbd_img_request *img_request) | |
2159 | { | |
2160 | struct rbd_obj_request *orig_request; | |
0eefd470 AE |
2161 | struct ceph_osd_request *osd_req; |
2162 | struct ceph_osd_client *osdc; | |
2163 | struct rbd_device *rbd_dev; | |
3d7efd18 | 2164 | struct page **pages; |
ebda6408 | 2165 | u32 page_count; |
3d7efd18 | 2166 | int result; |
ebda6408 | 2167 | u64 parent_length; |
b91f09f1 AE |
2168 | u64 offset; |
2169 | u64 length; | |
3d7efd18 AE |
2170 | |
2171 | rbd_assert(img_request_child_test(img_request)); | |
2172 | ||
2173 | /* First get what we need from the image request */ | |
2174 | ||
2175 | pages = img_request->copyup_pages; | |
2176 | rbd_assert(pages != NULL); | |
2177 | img_request->copyup_pages = NULL; | |
ebda6408 AE |
2178 | page_count = img_request->copyup_page_count; |
2179 | rbd_assert(page_count); | |
2180 | img_request->copyup_page_count = 0; | |
3d7efd18 AE |
2181 | |
2182 | orig_request = img_request->obj_request; | |
2183 | rbd_assert(orig_request != NULL); | |
b91f09f1 | 2184 | rbd_assert(obj_request_type_valid(orig_request->type)); |
3d7efd18 | 2185 | result = img_request->result; |
ebda6408 AE |
2186 | parent_length = img_request->length; |
2187 | rbd_assert(parent_length == img_request->xferred); | |
91c6febb | 2188 | rbd_img_request_put(img_request); |
3d7efd18 | 2189 | |
91c6febb AE |
2190 | rbd_assert(orig_request->img_request); |
2191 | rbd_dev = orig_request->img_request->rbd_dev; | |
0eefd470 | 2192 | rbd_assert(rbd_dev); |
0eefd470 | 2193 | |
0eefd470 AE |
2194 | if (result) |
2195 | goto out_err; | |
2196 | ||
8785b1d4 AE |
2197 | /* |
2198 | * The original osd request is of no use to use any more. | |
2199 | * We need a new one that can hold the two ops in a copyup | |
2200 | * request. Allocate the new copyup osd request for the | |
2201 | * original request, and release the old one. | |
2202 | */ | |
0eefd470 | 2203 | result = -ENOMEM; |
0eefd470 AE |
2204 | osd_req = rbd_osd_req_create_copyup(orig_request); |
2205 | if (!osd_req) | |
2206 | goto out_err; | |
8785b1d4 | 2207 | rbd_osd_req_destroy(orig_request->osd_req); |
0eefd470 AE |
2208 | orig_request->osd_req = osd_req; |
2209 | orig_request->copyup_pages = pages; | |
ebda6408 | 2210 | orig_request->copyup_page_count = page_count; |
3d7efd18 | 2211 | |
0eefd470 | 2212 | /* Initialize the copyup op */ |
3d7efd18 | 2213 | |
0eefd470 | 2214 | osd_req_op_cls_init(osd_req, 0, CEPH_OSD_OP_CALL, "rbd", "copyup"); |
ebda6408 | 2215 | osd_req_op_cls_request_data_pages(osd_req, 0, pages, parent_length, 0, |
0eefd470 | 2216 | false, false); |
3d7efd18 | 2217 | |
0eefd470 AE |
2218 | /* Then the original write request op */ |
2219 | ||
b91f09f1 AE |
2220 | offset = orig_request->offset; |
2221 | length = orig_request->length; | |
0eefd470 | 2222 | osd_req_op_extent_init(osd_req, 1, CEPH_OSD_OP_WRITE, |
b91f09f1 AE |
2223 | offset, length, 0, 0); |
2224 | if (orig_request->type == OBJ_REQUEST_BIO) | |
2225 | osd_req_op_extent_osd_data_bio(osd_req, 1, | |
2226 | orig_request->bio_list, length); | |
2227 | else | |
2228 | osd_req_op_extent_osd_data_pages(osd_req, 1, | |
2229 | orig_request->pages, length, | |
2230 | offset & ~PAGE_MASK, false, false); | |
0eefd470 AE |
2231 | |
2232 | rbd_osd_req_format_write(orig_request); | |
2233 | ||
2234 | /* All set, send it off. */ | |
2235 | ||
2236 | orig_request->callback = rbd_img_obj_copyup_callback; | |
2237 | osdc = &rbd_dev->rbd_client->client->osdc; | |
2238 | result = rbd_obj_request_submit(osdc, orig_request); | |
2239 | if (!result) | |
2240 | return; | |
2241 | out_err: | |
2242 | /* Record the error code and complete the request */ | |
2243 | ||
2244 | orig_request->result = result; | |
2245 | orig_request->xferred = 0; | |
2246 | obj_request_done_set(orig_request); | |
2247 | rbd_obj_request_complete(orig_request); | |
3d7efd18 AE |
2248 | } |
2249 | ||
2250 | /* | |
2251 | * Read from the parent image the range of data that covers the | |
2252 | * entire target of the given object request. This is used for | |
2253 | * satisfying a layered image write request when the target of an | |
2254 | * object request from the image request does not exist. | |
2255 | * | |
2256 | * A page array big enough to hold the returned data is allocated | |
2257 | * and supplied to rbd_img_request_fill() as the "data descriptor." | |
2258 | * When the read completes, this page array will be transferred to | |
2259 | * the original object request for the copyup operation. | |
2260 | * | |
2261 | * If an error occurs, record it as the result of the original | |
2262 | * object request and mark it done so it gets completed. | |
2263 | */ | |
2264 | static int rbd_img_obj_parent_read_full(struct rbd_obj_request *obj_request) | |
2265 | { | |
2266 | struct rbd_img_request *img_request = NULL; | |
2267 | struct rbd_img_request *parent_request = NULL; | |
2268 | struct rbd_device *rbd_dev; | |
2269 | u64 img_offset; | |
2270 | u64 length; | |
2271 | struct page **pages = NULL; | |
2272 | u32 page_count; | |
2273 | int result; | |
2274 | ||
2275 | rbd_assert(obj_request_img_data_test(obj_request)); | |
b91f09f1 | 2276 | rbd_assert(obj_request_type_valid(obj_request->type)); |
3d7efd18 AE |
2277 | |
2278 | img_request = obj_request->img_request; | |
2279 | rbd_assert(img_request != NULL); | |
2280 | rbd_dev = img_request->rbd_dev; | |
2281 | rbd_assert(rbd_dev->parent != NULL); | |
2282 | ||
2283 | /* | |
2284 | * Determine the byte range covered by the object in the | |
2285 | * child image to which the original request was to be sent. | |
2286 | */ | |
2287 | img_offset = obj_request->img_offset - obj_request->offset; | |
2288 | length = (u64)1 << rbd_dev->header.obj_order; | |
2289 | ||
a9e8ba2c AE |
2290 | /* |
2291 | * There is no defined parent data beyond the parent | |
2292 | * overlap, so limit what we read at that boundary if | |
2293 | * necessary. | |
2294 | */ | |
2295 | if (img_offset + length > rbd_dev->parent_overlap) { | |
2296 | rbd_assert(img_offset < rbd_dev->parent_overlap); | |
2297 | length = rbd_dev->parent_overlap - img_offset; | |
2298 | } | |
2299 | ||
3d7efd18 AE |
2300 | /* |
2301 | * Allocate a page array big enough to receive the data read | |
2302 | * from the parent. | |
2303 | */ | |
2304 | page_count = (u32)calc_pages_for(0, length); | |
2305 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); | |
2306 | if (IS_ERR(pages)) { | |
2307 | result = PTR_ERR(pages); | |
2308 | pages = NULL; | |
2309 | goto out_err; | |
2310 | } | |
2311 | ||
2312 | result = -ENOMEM; | |
2313 | parent_request = rbd_img_request_create(rbd_dev->parent, | |
2314 | img_offset, length, | |
2315 | false, true); | |
2316 | if (!parent_request) | |
2317 | goto out_err; | |
2318 | rbd_obj_request_get(obj_request); | |
2319 | parent_request->obj_request = obj_request; | |
2320 | ||
2321 | result = rbd_img_request_fill(parent_request, OBJ_REQUEST_PAGES, pages); | |
2322 | if (result) | |
2323 | goto out_err; | |
2324 | parent_request->copyup_pages = pages; | |
ebda6408 | 2325 | parent_request->copyup_page_count = page_count; |
3d7efd18 AE |
2326 | |
2327 | parent_request->callback = rbd_img_obj_parent_read_full_callback; | |
2328 | result = rbd_img_request_submit(parent_request); | |
2329 | if (!result) | |
2330 | return 0; | |
2331 | ||
2332 | parent_request->copyup_pages = NULL; | |
ebda6408 | 2333 | parent_request->copyup_page_count = 0; |
3d7efd18 AE |
2334 | parent_request->obj_request = NULL; |
2335 | rbd_obj_request_put(obj_request); | |
2336 | out_err: | |
2337 | if (pages) | |
2338 | ceph_release_page_vector(pages, page_count); | |
2339 | if (parent_request) | |
2340 | rbd_img_request_put(parent_request); | |
2341 | obj_request->result = result; | |
2342 | obj_request->xferred = 0; | |
2343 | obj_request_done_set(obj_request); | |
2344 | ||
2345 | return result; | |
2346 | } | |
2347 | ||
c5b5ef6c AE |
2348 | static void rbd_img_obj_exists_callback(struct rbd_obj_request *obj_request) |
2349 | { | |
c5b5ef6c AE |
2350 | struct rbd_obj_request *orig_request; |
2351 | int result; | |
2352 | ||
2353 | rbd_assert(!obj_request_img_data_test(obj_request)); | |
2354 | ||
2355 | /* | |
2356 | * All we need from the object request is the original | |
2357 | * request and the result of the STAT op. Grab those, then | |
2358 | * we're done with the request. | |
2359 | */ | |
2360 | orig_request = obj_request->obj_request; | |
2361 | obj_request->obj_request = NULL; | |
2362 | rbd_assert(orig_request); | |
2363 | rbd_assert(orig_request->img_request); | |
2364 | ||
2365 | result = obj_request->result; | |
2366 | obj_request->result = 0; | |
2367 | ||
2368 | dout("%s: obj %p for obj %p result %d %llu/%llu\n", __func__, | |
2369 | obj_request, orig_request, result, | |
2370 | obj_request->xferred, obj_request->length); | |
2371 | rbd_obj_request_put(obj_request); | |
2372 | ||
2373 | rbd_assert(orig_request); | |
2374 | rbd_assert(orig_request->img_request); | |
c5b5ef6c AE |
2375 | |
2376 | /* | |
2377 | * Our only purpose here is to determine whether the object | |
2378 | * exists, and we don't want to treat the non-existence as | |
2379 | * an error. If something else comes back, transfer the | |
2380 | * error to the original request and complete it now. | |
2381 | */ | |
2382 | if (!result) { | |
2383 | obj_request_existence_set(orig_request, true); | |
2384 | } else if (result == -ENOENT) { | |
2385 | obj_request_existence_set(orig_request, false); | |
2386 | } else if (result) { | |
2387 | orig_request->result = result; | |
3d7efd18 | 2388 | goto out; |
c5b5ef6c AE |
2389 | } |
2390 | ||
2391 | /* | |
2392 | * Resubmit the original request now that we have recorded | |
2393 | * whether the target object exists. | |
2394 | */ | |
b454e36d | 2395 | orig_request->result = rbd_img_obj_request_submit(orig_request); |
3d7efd18 | 2396 | out: |
c5b5ef6c AE |
2397 | if (orig_request->result) |
2398 | rbd_obj_request_complete(orig_request); | |
2399 | rbd_obj_request_put(orig_request); | |
2400 | } | |
2401 | ||
2402 | static int rbd_img_obj_exists_submit(struct rbd_obj_request *obj_request) | |
2403 | { | |
2404 | struct rbd_obj_request *stat_request; | |
2405 | struct rbd_device *rbd_dev; | |
2406 | struct ceph_osd_client *osdc; | |
2407 | struct page **pages = NULL; | |
2408 | u32 page_count; | |
2409 | size_t size; | |
2410 | int ret; | |
2411 | ||
2412 | /* | |
2413 | * The response data for a STAT call consists of: | |
2414 | * le64 length; | |
2415 | * struct { | |
2416 | * le32 tv_sec; | |
2417 | * le32 tv_nsec; | |
2418 | * } mtime; | |
2419 | */ | |
2420 | size = sizeof (__le64) + sizeof (__le32) + sizeof (__le32); | |
2421 | page_count = (u32)calc_pages_for(0, size); | |
2422 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); | |
2423 | if (IS_ERR(pages)) | |
2424 | return PTR_ERR(pages); | |
2425 | ||
2426 | ret = -ENOMEM; | |
2427 | stat_request = rbd_obj_request_create(obj_request->object_name, 0, 0, | |
2428 | OBJ_REQUEST_PAGES); | |
2429 | if (!stat_request) | |
2430 | goto out; | |
2431 | ||
2432 | rbd_obj_request_get(obj_request); | |
2433 | stat_request->obj_request = obj_request; | |
2434 | stat_request->pages = pages; | |
2435 | stat_request->page_count = page_count; | |
2436 | ||
2437 | rbd_assert(obj_request->img_request); | |
2438 | rbd_dev = obj_request->img_request->rbd_dev; | |
2439 | stat_request->osd_req = rbd_osd_req_create(rbd_dev, false, | |
2440 | stat_request); | |
2441 | if (!stat_request->osd_req) | |
2442 | goto out; | |
2443 | stat_request->callback = rbd_img_obj_exists_callback; | |
2444 | ||
2445 | osd_req_op_init(stat_request->osd_req, 0, CEPH_OSD_OP_STAT); | |
2446 | osd_req_op_raw_data_in_pages(stat_request->osd_req, 0, pages, size, 0, | |
2447 | false, false); | |
9d4df01f | 2448 | rbd_osd_req_format_read(stat_request); |
c5b5ef6c AE |
2449 | |
2450 | osdc = &rbd_dev->rbd_client->client->osdc; | |
2451 | ret = rbd_obj_request_submit(osdc, stat_request); | |
2452 | out: | |
2453 | if (ret) | |
2454 | rbd_obj_request_put(obj_request); | |
2455 | ||
2456 | return ret; | |
2457 | } | |
2458 | ||
b454e36d AE |
2459 | static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) |
2460 | { | |
2461 | struct rbd_img_request *img_request; | |
a9e8ba2c | 2462 | struct rbd_device *rbd_dev; |
3d7efd18 | 2463 | bool known; |
b454e36d AE |
2464 | |
2465 | rbd_assert(obj_request_img_data_test(obj_request)); | |
2466 | ||
2467 | img_request = obj_request->img_request; | |
2468 | rbd_assert(img_request); | |
a9e8ba2c | 2469 | rbd_dev = img_request->rbd_dev; |
b454e36d | 2470 | |
b454e36d | 2471 | /* |
a9e8ba2c AE |
2472 | * Only writes to layered images need special handling. |
2473 | * Reads and non-layered writes are simple object requests. | |
2474 | * Layered writes that start beyond the end of the overlap | |
2475 | * with the parent have no parent data, so they too are | |
2476 | * simple object requests. Finally, if the target object is | |
2477 | * known to already exist, its parent data has already been | |
2478 | * copied, so a write to the object can also be handled as a | |
2479 | * simple object request. | |
b454e36d AE |
2480 | */ |
2481 | if (!img_request_write_test(img_request) || | |
2482 | !img_request_layered_test(img_request) || | |
a9e8ba2c | 2483 | rbd_dev->parent_overlap <= obj_request->img_offset || |
3d7efd18 AE |
2484 | ((known = obj_request_known_test(obj_request)) && |
2485 | obj_request_exists_test(obj_request))) { | |
b454e36d AE |
2486 | |
2487 | struct rbd_device *rbd_dev; | |
2488 | struct ceph_osd_client *osdc; | |
2489 | ||
2490 | rbd_dev = obj_request->img_request->rbd_dev; | |
2491 | osdc = &rbd_dev->rbd_client->client->osdc; | |
2492 | ||
2493 | return rbd_obj_request_submit(osdc, obj_request); | |
2494 | } | |
2495 | ||
2496 | /* | |
3d7efd18 AE |
2497 | * It's a layered write. The target object might exist but |
2498 | * we may not know that yet. If we know it doesn't exist, | |
2499 | * start by reading the data for the full target object from | |
2500 | * the parent so we can use it for a copyup to the target. | |
b454e36d | 2501 | */ |
3d7efd18 AE |
2502 | if (known) |
2503 | return rbd_img_obj_parent_read_full(obj_request); | |
2504 | ||
2505 | /* We don't know whether the target exists. Go find out. */ | |
b454e36d AE |
2506 | |
2507 | return rbd_img_obj_exists_submit(obj_request); | |
2508 | } | |
2509 | ||
bf0d5f50 AE |
2510 | static int rbd_img_request_submit(struct rbd_img_request *img_request) |
2511 | { | |
bf0d5f50 | 2512 | struct rbd_obj_request *obj_request; |
46faeed4 | 2513 | struct rbd_obj_request *next_obj_request; |
bf0d5f50 | 2514 | |
37206ee5 | 2515 | dout("%s: img %p\n", __func__, img_request); |
46faeed4 | 2516 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) { |
bf0d5f50 AE |
2517 | int ret; |
2518 | ||
b454e36d | 2519 | ret = rbd_img_obj_request_submit(obj_request); |
bf0d5f50 AE |
2520 | if (ret) |
2521 | return ret; | |
bf0d5f50 AE |
2522 | } |
2523 | ||
2524 | return 0; | |
2525 | } | |
8b3e1a56 AE |
2526 | |
2527 | static void rbd_img_parent_read_callback(struct rbd_img_request *img_request) | |
2528 | { | |
2529 | struct rbd_obj_request *obj_request; | |
a9e8ba2c AE |
2530 | struct rbd_device *rbd_dev; |
2531 | u64 obj_end; | |
8b3e1a56 AE |
2532 | |
2533 | rbd_assert(img_request_child_test(img_request)); | |
2534 | ||
2535 | obj_request = img_request->obj_request; | |
a9e8ba2c AE |
2536 | rbd_assert(obj_request); |
2537 | rbd_assert(obj_request->img_request); | |
2538 | ||
8b3e1a56 | 2539 | obj_request->result = img_request->result; |
a9e8ba2c AE |
2540 | if (obj_request->result) |
2541 | goto out; | |
2542 | ||
2543 | /* | |
2544 | * We need to zero anything beyond the parent overlap | |
2545 | * boundary. Since rbd_img_obj_request_read_callback() | |
2546 | * will zero anything beyond the end of a short read, an | |
2547 | * easy way to do this is to pretend the data from the | |
2548 | * parent came up short--ending at the overlap boundary. | |
2549 | */ | |
2550 | rbd_assert(obj_request->img_offset < U64_MAX - obj_request->length); | |
2551 | obj_end = obj_request->img_offset + obj_request->length; | |
2552 | rbd_dev = obj_request->img_request->rbd_dev; | |
2553 | if (obj_end > rbd_dev->parent_overlap) { | |
2554 | u64 xferred = 0; | |
2555 | ||
2556 | if (obj_request->img_offset < rbd_dev->parent_overlap) | |
2557 | xferred = rbd_dev->parent_overlap - | |
2558 | obj_request->img_offset; | |
8b3e1a56 | 2559 | |
a9e8ba2c AE |
2560 | obj_request->xferred = min(img_request->xferred, xferred); |
2561 | } else { | |
2562 | obj_request->xferred = img_request->xferred; | |
2563 | } | |
2564 | out: | |
b5b09be3 | 2565 | rbd_img_request_put(img_request); |
8b3e1a56 AE |
2566 | rbd_img_obj_request_read_callback(obj_request); |
2567 | rbd_obj_request_complete(obj_request); | |
2568 | } | |
2569 | ||
2570 | static void rbd_img_parent_read(struct rbd_obj_request *obj_request) | |
2571 | { | |
2572 | struct rbd_device *rbd_dev; | |
2573 | struct rbd_img_request *img_request; | |
2574 | int result; | |
2575 | ||
2576 | rbd_assert(obj_request_img_data_test(obj_request)); | |
2577 | rbd_assert(obj_request->img_request != NULL); | |
2578 | rbd_assert(obj_request->result == (s32) -ENOENT); | |
5b2ab72d | 2579 | rbd_assert(obj_request_type_valid(obj_request->type)); |
8b3e1a56 AE |
2580 | |
2581 | rbd_dev = obj_request->img_request->rbd_dev; | |
2582 | rbd_assert(rbd_dev->parent != NULL); | |
2583 | /* rbd_read_finish(obj_request, obj_request->length); */ | |
2584 | img_request = rbd_img_request_create(rbd_dev->parent, | |
2585 | obj_request->img_offset, | |
2586 | obj_request->length, | |
2587 | false, true); | |
2588 | result = -ENOMEM; | |
2589 | if (!img_request) | |
2590 | goto out_err; | |
2591 | ||
2592 | rbd_obj_request_get(obj_request); | |
2593 | img_request->obj_request = obj_request; | |
2594 | ||
5b2ab72d AE |
2595 | if (obj_request->type == OBJ_REQUEST_BIO) |
2596 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO, | |
2597 | obj_request->bio_list); | |
2598 | else | |
2599 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_PAGES, | |
2600 | obj_request->pages); | |
8b3e1a56 AE |
2601 | if (result) |
2602 | goto out_err; | |
2603 | ||
2604 | img_request->callback = rbd_img_parent_read_callback; | |
2605 | result = rbd_img_request_submit(img_request); | |
2606 | if (result) | |
2607 | goto out_err; | |
2608 | ||
2609 | return; | |
2610 | out_err: | |
2611 | if (img_request) | |
2612 | rbd_img_request_put(img_request); | |
2613 | obj_request->result = result; | |
2614 | obj_request->xferred = 0; | |
2615 | obj_request_done_set(obj_request); | |
2616 | } | |
bf0d5f50 | 2617 | |
cc4a38bd | 2618 | static int rbd_obj_notify_ack(struct rbd_device *rbd_dev, u64 notify_id) |
b8d70035 AE |
2619 | { |
2620 | struct rbd_obj_request *obj_request; | |
2169238d | 2621 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
b8d70035 AE |
2622 | int ret; |
2623 | ||
2624 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, | |
2625 | OBJ_REQUEST_NODATA); | |
2626 | if (!obj_request) | |
2627 | return -ENOMEM; | |
2628 | ||
2629 | ret = -ENOMEM; | |
430c28c3 | 2630 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
b8d70035 AE |
2631 | if (!obj_request->osd_req) |
2632 | goto out; | |
2169238d | 2633 | obj_request->callback = rbd_obj_request_put; |
b8d70035 | 2634 | |
c99d2d4a | 2635 | osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_NOTIFY_ACK, |
cc4a38bd | 2636 | notify_id, 0, 0); |
9d4df01f | 2637 | rbd_osd_req_format_read(obj_request); |
430c28c3 | 2638 | |
b8d70035 | 2639 | ret = rbd_obj_request_submit(osdc, obj_request); |
b8d70035 | 2640 | out: |
cf81b60e AE |
2641 | if (ret) |
2642 | rbd_obj_request_put(obj_request); | |
b8d70035 AE |
2643 | |
2644 | return ret; | |
2645 | } | |
2646 | ||
2647 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) | |
2648 | { | |
2649 | struct rbd_device *rbd_dev = (struct rbd_device *)data; | |
e627db08 | 2650 | int ret; |
b8d70035 AE |
2651 | |
2652 | if (!rbd_dev) | |
2653 | return; | |
2654 | ||
37206ee5 | 2655 | dout("%s: \"%s\" notify_id %llu opcode %u\n", __func__, |
cc4a38bd AE |
2656 | rbd_dev->header_name, (unsigned long long)notify_id, |
2657 | (unsigned int)opcode); | |
e627db08 AE |
2658 | ret = rbd_dev_refresh(rbd_dev); |
2659 | if (ret) | |
2660 | rbd_warn(rbd_dev, ": header refresh error (%d)\n", ret); | |
b8d70035 | 2661 | |
cc4a38bd | 2662 | rbd_obj_notify_ack(rbd_dev, notify_id); |
b8d70035 AE |
2663 | } |
2664 | ||
9969ebc5 AE |
2665 | /* |
2666 | * Request sync osd watch/unwatch. The value of "start" determines | |
2667 | * whether a watch request is being initiated or torn down. | |
2668 | */ | |
1f3ef788 | 2669 | static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev, bool start) |
9969ebc5 AE |
2670 | { |
2671 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; | |
2672 | struct rbd_obj_request *obj_request; | |
9969ebc5 AE |
2673 | int ret; |
2674 | ||
2675 | rbd_assert(start ^ !!rbd_dev->watch_event); | |
2676 | rbd_assert(start ^ !!rbd_dev->watch_request); | |
2677 | ||
2678 | if (start) { | |
3c663bbd | 2679 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, rbd_dev, |
9969ebc5 AE |
2680 | &rbd_dev->watch_event); |
2681 | if (ret < 0) | |
2682 | return ret; | |
8eb87565 | 2683 | rbd_assert(rbd_dev->watch_event != NULL); |
9969ebc5 AE |
2684 | } |
2685 | ||
2686 | ret = -ENOMEM; | |
2687 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, | |
2688 | OBJ_REQUEST_NODATA); | |
2689 | if (!obj_request) | |
2690 | goto out_cancel; | |
2691 | ||
430c28c3 AE |
2692 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, true, obj_request); |
2693 | if (!obj_request->osd_req) | |
2694 | goto out_cancel; | |
2695 | ||
8eb87565 | 2696 | if (start) |
975241af | 2697 | ceph_osdc_set_request_linger(osdc, obj_request->osd_req); |
8eb87565 | 2698 | else |
6977c3f9 | 2699 | ceph_osdc_unregister_linger_request(osdc, |
975241af | 2700 | rbd_dev->watch_request->osd_req); |
2169238d AE |
2701 | |
2702 | osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_WATCH, | |
1f3ef788 | 2703 | rbd_dev->watch_event->cookie, 0, start ? 1 : 0); |
9d4df01f | 2704 | rbd_osd_req_format_write(obj_request); |
2169238d | 2705 | |
9969ebc5 AE |
2706 | ret = rbd_obj_request_submit(osdc, obj_request); |
2707 | if (ret) | |
2708 | goto out_cancel; | |
2709 | ret = rbd_obj_request_wait(obj_request); | |
2710 | if (ret) | |
2711 | goto out_cancel; | |
9969ebc5 AE |
2712 | ret = obj_request->result; |
2713 | if (ret) | |
2714 | goto out_cancel; | |
2715 | ||
8eb87565 AE |
2716 | /* |
2717 | * A watch request is set to linger, so the underlying osd | |
2718 | * request won't go away until we unregister it. We retain | |
2719 | * a pointer to the object request during that time (in | |
2720 | * rbd_dev->watch_request), so we'll keep a reference to | |
2721 | * it. We'll drop that reference (below) after we've | |
2722 | * unregistered it. | |
2723 | */ | |
2724 | if (start) { | |
2725 | rbd_dev->watch_request = obj_request; | |
2726 | ||
2727 | return 0; | |
2728 | } | |
2729 | ||
2730 | /* We have successfully torn down the watch request */ | |
2731 | ||
2732 | rbd_obj_request_put(rbd_dev->watch_request); | |
2733 | rbd_dev->watch_request = NULL; | |
9969ebc5 AE |
2734 | out_cancel: |
2735 | /* Cancel the event if we're tearing down, or on error */ | |
2736 | ceph_osdc_cancel_event(rbd_dev->watch_event); | |
2737 | rbd_dev->watch_event = NULL; | |
9969ebc5 AE |
2738 | if (obj_request) |
2739 | rbd_obj_request_put(obj_request); | |
2740 | ||
2741 | return ret; | |
2742 | } | |
2743 | ||
36be9a76 | 2744 | /* |
f40eb349 AE |
2745 | * Synchronous osd object method call. Returns the number of bytes |
2746 | * returned in the outbound buffer, or a negative error code. | |
36be9a76 AE |
2747 | */ |
2748 | static int rbd_obj_method_sync(struct rbd_device *rbd_dev, | |
2749 | const char *object_name, | |
2750 | const char *class_name, | |
2751 | const char *method_name, | |
4157976b | 2752 | const void *outbound, |
36be9a76 | 2753 | size_t outbound_size, |
4157976b | 2754 | void *inbound, |
e2a58ee5 | 2755 | size_t inbound_size) |
36be9a76 | 2756 | { |
2169238d | 2757 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
36be9a76 | 2758 | struct rbd_obj_request *obj_request; |
36be9a76 AE |
2759 | struct page **pages; |
2760 | u32 page_count; | |
2761 | int ret; | |
2762 | ||
2763 | /* | |
6010a451 AE |
2764 | * Method calls are ultimately read operations. The result |
2765 | * should placed into the inbound buffer provided. They | |
2766 | * also supply outbound data--parameters for the object | |
2767 | * method. Currently if this is present it will be a | |
2768 | * snapshot id. | |
36be9a76 | 2769 | */ |
57385b51 | 2770 | page_count = (u32)calc_pages_for(0, inbound_size); |
36be9a76 AE |
2771 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
2772 | if (IS_ERR(pages)) | |
2773 | return PTR_ERR(pages); | |
2774 | ||
2775 | ret = -ENOMEM; | |
6010a451 | 2776 | obj_request = rbd_obj_request_create(object_name, 0, inbound_size, |
36be9a76 AE |
2777 | OBJ_REQUEST_PAGES); |
2778 | if (!obj_request) | |
2779 | goto out; | |
2780 | ||
2781 | obj_request->pages = pages; | |
2782 | obj_request->page_count = page_count; | |
2783 | ||
430c28c3 | 2784 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
36be9a76 AE |
2785 | if (!obj_request->osd_req) |
2786 | goto out; | |
2787 | ||
c99d2d4a | 2788 | osd_req_op_cls_init(obj_request->osd_req, 0, CEPH_OSD_OP_CALL, |
04017e29 AE |
2789 | class_name, method_name); |
2790 | if (outbound_size) { | |
2791 | struct ceph_pagelist *pagelist; | |
2792 | ||
2793 | pagelist = kmalloc(sizeof (*pagelist), GFP_NOFS); | |
2794 | if (!pagelist) | |
2795 | goto out; | |
2796 | ||
2797 | ceph_pagelist_init(pagelist); | |
2798 | ceph_pagelist_append(pagelist, outbound, outbound_size); | |
2799 | osd_req_op_cls_request_data_pagelist(obj_request->osd_req, 0, | |
2800 | pagelist); | |
2801 | } | |
a4ce40a9 AE |
2802 | osd_req_op_cls_response_data_pages(obj_request->osd_req, 0, |
2803 | obj_request->pages, inbound_size, | |
44cd188d | 2804 | 0, false, false); |
9d4df01f | 2805 | rbd_osd_req_format_read(obj_request); |
430c28c3 | 2806 | |
36be9a76 AE |
2807 | ret = rbd_obj_request_submit(osdc, obj_request); |
2808 | if (ret) | |
2809 | goto out; | |
2810 | ret = rbd_obj_request_wait(obj_request); | |
2811 | if (ret) | |
2812 | goto out; | |
2813 | ||
2814 | ret = obj_request->result; | |
2815 | if (ret < 0) | |
2816 | goto out; | |
57385b51 AE |
2817 | |
2818 | rbd_assert(obj_request->xferred < (u64)INT_MAX); | |
2819 | ret = (int)obj_request->xferred; | |
903bb32e | 2820 | ceph_copy_from_page_vector(pages, inbound, 0, obj_request->xferred); |
36be9a76 AE |
2821 | out: |
2822 | if (obj_request) | |
2823 | rbd_obj_request_put(obj_request); | |
2824 | else | |
2825 | ceph_release_page_vector(pages, page_count); | |
2826 | ||
2827 | return ret; | |
2828 | } | |
2829 | ||
bf0d5f50 | 2830 | static void rbd_request_fn(struct request_queue *q) |
cc344fa1 | 2831 | __releases(q->queue_lock) __acquires(q->queue_lock) |
bf0d5f50 AE |
2832 | { |
2833 | struct rbd_device *rbd_dev = q->queuedata; | |
2834 | bool read_only = rbd_dev->mapping.read_only; | |
2835 | struct request *rq; | |
2836 | int result; | |
2837 | ||
2838 | while ((rq = blk_fetch_request(q))) { | |
2839 | bool write_request = rq_data_dir(rq) == WRITE; | |
2840 | struct rbd_img_request *img_request; | |
2841 | u64 offset; | |
2842 | u64 length; | |
2843 | ||
2844 | /* Ignore any non-FS requests that filter through. */ | |
2845 | ||
2846 | if (rq->cmd_type != REQ_TYPE_FS) { | |
4dda41d3 AE |
2847 | dout("%s: non-fs request type %d\n", __func__, |
2848 | (int) rq->cmd_type); | |
2849 | __blk_end_request_all(rq, 0); | |
2850 | continue; | |
2851 | } | |
2852 | ||
2853 | /* Ignore/skip any zero-length requests */ | |
2854 | ||
2855 | offset = (u64) blk_rq_pos(rq) << SECTOR_SHIFT; | |
2856 | length = (u64) blk_rq_bytes(rq); | |
2857 | ||
2858 | if (!length) { | |
2859 | dout("%s: zero-length request\n", __func__); | |
bf0d5f50 AE |
2860 | __blk_end_request_all(rq, 0); |
2861 | continue; | |
2862 | } | |
2863 | ||
2864 | spin_unlock_irq(q->queue_lock); | |
2865 | ||
2866 | /* Disallow writes to a read-only device */ | |
2867 | ||
2868 | if (write_request) { | |
2869 | result = -EROFS; | |
2870 | if (read_only) | |
2871 | goto end_request; | |
2872 | rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP); | |
2873 | } | |
2874 | ||
6d292906 AE |
2875 | /* |
2876 | * Quit early if the mapped snapshot no longer | |
2877 | * exists. It's still possible the snapshot will | |
2878 | * have disappeared by the time our request arrives | |
2879 | * at the osd, but there's no sense in sending it if | |
2880 | * we already know. | |
2881 | */ | |
2882 | if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) { | |
bf0d5f50 AE |
2883 | dout("request for non-existent snapshot"); |
2884 | rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); | |
2885 | result = -ENXIO; | |
2886 | goto end_request; | |
2887 | } | |
2888 | ||
bf0d5f50 | 2889 | result = -EINVAL; |
c0cd10db AE |
2890 | if (offset && length > U64_MAX - offset + 1) { |
2891 | rbd_warn(rbd_dev, "bad request range (%llu~%llu)\n", | |
2892 | offset, length); | |
bf0d5f50 | 2893 | goto end_request; /* Shouldn't happen */ |
c0cd10db | 2894 | } |
bf0d5f50 | 2895 | |
00a653e2 AE |
2896 | result = -EIO; |
2897 | if (offset + length > rbd_dev->mapping.size) { | |
2898 | rbd_warn(rbd_dev, "beyond EOD (%llu~%llu > %llu)\n", | |
2899 | offset, length, rbd_dev->mapping.size); | |
2900 | goto end_request; | |
2901 | } | |
2902 | ||
bf0d5f50 AE |
2903 | result = -ENOMEM; |
2904 | img_request = rbd_img_request_create(rbd_dev, offset, length, | |
9849e986 | 2905 | write_request, false); |
bf0d5f50 AE |
2906 | if (!img_request) |
2907 | goto end_request; | |
2908 | ||
2909 | img_request->rq = rq; | |
2910 | ||
f1a4739f AE |
2911 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO, |
2912 | rq->bio); | |
bf0d5f50 AE |
2913 | if (!result) |
2914 | result = rbd_img_request_submit(img_request); | |
2915 | if (result) | |
2916 | rbd_img_request_put(img_request); | |
2917 | end_request: | |
2918 | spin_lock_irq(q->queue_lock); | |
2919 | if (result < 0) { | |
7da22d29 AE |
2920 | rbd_warn(rbd_dev, "%s %llx at %llx result %d\n", |
2921 | write_request ? "write" : "read", | |
2922 | length, offset, result); | |
2923 | ||
bf0d5f50 AE |
2924 | __blk_end_request_all(rq, result); |
2925 | } | |
2926 | } | |
2927 | } | |
2928 | ||
602adf40 YS |
2929 | /* |
2930 | * a queue callback. Makes sure that we don't create a bio that spans across | |
2931 | * multiple osd objects. One exception would be with a single page bios, | |
f7760dad | 2932 | * which we handle later at bio_chain_clone_range() |
602adf40 YS |
2933 | */ |
2934 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, | |
2935 | struct bio_vec *bvec) | |
2936 | { | |
2937 | struct rbd_device *rbd_dev = q->queuedata; | |
e5cfeed2 AE |
2938 | sector_t sector_offset; |
2939 | sector_t sectors_per_obj; | |
2940 | sector_t obj_sector_offset; | |
2941 | int ret; | |
2942 | ||
2943 | /* | |
2944 | * Find how far into its rbd object the partition-relative | |
2945 | * bio start sector is to offset relative to the enclosing | |
2946 | * device. | |
2947 | */ | |
2948 | sector_offset = get_start_sect(bmd->bi_bdev) + bmd->bi_sector; | |
2949 | sectors_per_obj = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); | |
2950 | obj_sector_offset = sector_offset & (sectors_per_obj - 1); | |
2951 | ||
2952 | /* | |
2953 | * Compute the number of bytes from that offset to the end | |
2954 | * of the object. Account for what's already used by the bio. | |
2955 | */ | |
2956 | ret = (int) (sectors_per_obj - obj_sector_offset) << SECTOR_SHIFT; | |
2957 | if (ret > bmd->bi_size) | |
2958 | ret -= bmd->bi_size; | |
2959 | else | |
2960 | ret = 0; | |
2961 | ||
2962 | /* | |
2963 | * Don't send back more than was asked for. And if the bio | |
2964 | * was empty, let the whole thing through because: "Note | |
2965 | * that a block device *must* allow a single page to be | |
2966 | * added to an empty bio." | |
2967 | */ | |
2968 | rbd_assert(bvec->bv_len <= PAGE_SIZE); | |
2969 | if (ret > (int) bvec->bv_len || !bmd->bi_size) | |
2970 | ret = (int) bvec->bv_len; | |
2971 | ||
2972 | return ret; | |
602adf40 YS |
2973 | } |
2974 | ||
2975 | static void rbd_free_disk(struct rbd_device *rbd_dev) | |
2976 | { | |
2977 | struct gendisk *disk = rbd_dev->disk; | |
2978 | ||
2979 | if (!disk) | |
2980 | return; | |
2981 | ||
a0cab924 AE |
2982 | rbd_dev->disk = NULL; |
2983 | if (disk->flags & GENHD_FL_UP) { | |
602adf40 | 2984 | del_gendisk(disk); |
a0cab924 AE |
2985 | if (disk->queue) |
2986 | blk_cleanup_queue(disk->queue); | |
2987 | } | |
602adf40 YS |
2988 | put_disk(disk); |
2989 | } | |
2990 | ||
788e2df3 AE |
2991 | static int rbd_obj_read_sync(struct rbd_device *rbd_dev, |
2992 | const char *object_name, | |
7097f8df | 2993 | u64 offset, u64 length, void *buf) |
788e2df3 AE |
2994 | |
2995 | { | |
2169238d | 2996 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
788e2df3 | 2997 | struct rbd_obj_request *obj_request; |
788e2df3 AE |
2998 | struct page **pages = NULL; |
2999 | u32 page_count; | |
1ceae7ef | 3000 | size_t size; |
788e2df3 AE |
3001 | int ret; |
3002 | ||
3003 | page_count = (u32) calc_pages_for(offset, length); | |
3004 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); | |
3005 | if (IS_ERR(pages)) | |
3006 | ret = PTR_ERR(pages); | |
3007 | ||
3008 | ret = -ENOMEM; | |
3009 | obj_request = rbd_obj_request_create(object_name, offset, length, | |
36be9a76 | 3010 | OBJ_REQUEST_PAGES); |
788e2df3 AE |
3011 | if (!obj_request) |
3012 | goto out; | |
3013 | ||
3014 | obj_request->pages = pages; | |
3015 | obj_request->page_count = page_count; | |
3016 | ||
430c28c3 | 3017 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
788e2df3 AE |
3018 | if (!obj_request->osd_req) |
3019 | goto out; | |
3020 | ||
c99d2d4a AE |
3021 | osd_req_op_extent_init(obj_request->osd_req, 0, CEPH_OSD_OP_READ, |
3022 | offset, length, 0, 0); | |
406e2c9f | 3023 | osd_req_op_extent_osd_data_pages(obj_request->osd_req, 0, |
a4ce40a9 | 3024 | obj_request->pages, |
44cd188d AE |
3025 | obj_request->length, |
3026 | obj_request->offset & ~PAGE_MASK, | |
3027 | false, false); | |
9d4df01f | 3028 | rbd_osd_req_format_read(obj_request); |
430c28c3 | 3029 | |
788e2df3 AE |
3030 | ret = rbd_obj_request_submit(osdc, obj_request); |
3031 | if (ret) | |
3032 | goto out; | |
3033 | ret = rbd_obj_request_wait(obj_request); | |
3034 | if (ret) | |
3035 | goto out; | |
3036 | ||
3037 | ret = obj_request->result; | |
3038 | if (ret < 0) | |
3039 | goto out; | |
1ceae7ef AE |
3040 | |
3041 | rbd_assert(obj_request->xferred <= (u64) SIZE_MAX); | |
3042 | size = (size_t) obj_request->xferred; | |
903bb32e | 3043 | ceph_copy_from_page_vector(pages, buf, 0, size); |
7097f8df AE |
3044 | rbd_assert(size <= (size_t)INT_MAX); |
3045 | ret = (int)size; | |
788e2df3 AE |
3046 | out: |
3047 | if (obj_request) | |
3048 | rbd_obj_request_put(obj_request); | |
3049 | else | |
3050 | ceph_release_page_vector(pages, page_count); | |
3051 | ||
3052 | return ret; | |
3053 | } | |
3054 | ||
602adf40 | 3055 | /* |
662518b1 AE |
3056 | * Read the complete header for the given rbd device. On successful |
3057 | * return, the rbd_dev->header field will contain up-to-date | |
3058 | * information about the image. | |
602adf40 | 3059 | */ |
99a41ebc | 3060 | static int rbd_dev_v1_header_info(struct rbd_device *rbd_dev) |
602adf40 | 3061 | { |
4156d998 | 3062 | struct rbd_image_header_ondisk *ondisk = NULL; |
50f7c4c9 | 3063 | u32 snap_count = 0; |
4156d998 AE |
3064 | u64 names_size = 0; |
3065 | u32 want_count; | |
3066 | int ret; | |
602adf40 | 3067 | |
00f1f36f | 3068 | /* |
4156d998 AE |
3069 | * The complete header will include an array of its 64-bit |
3070 | * snapshot ids, followed by the names of those snapshots as | |
3071 | * a contiguous block of NUL-terminated strings. Note that | |
3072 | * the number of snapshots could change by the time we read | |
3073 | * it in, in which case we re-read it. | |
00f1f36f | 3074 | */ |
4156d998 AE |
3075 | do { |
3076 | size_t size; | |
3077 | ||
3078 | kfree(ondisk); | |
3079 | ||
3080 | size = sizeof (*ondisk); | |
3081 | size += snap_count * sizeof (struct rbd_image_snap_ondisk); | |
3082 | size += names_size; | |
3083 | ondisk = kmalloc(size, GFP_KERNEL); | |
3084 | if (!ondisk) | |
662518b1 | 3085 | return -ENOMEM; |
4156d998 | 3086 | |
788e2df3 | 3087 | ret = rbd_obj_read_sync(rbd_dev, rbd_dev->header_name, |
7097f8df | 3088 | 0, size, ondisk); |
4156d998 | 3089 | if (ret < 0) |
662518b1 | 3090 | goto out; |
c0cd10db | 3091 | if ((size_t)ret < size) { |
4156d998 | 3092 | ret = -ENXIO; |
06ecc6cb AE |
3093 | rbd_warn(rbd_dev, "short header read (want %zd got %d)", |
3094 | size, ret); | |
662518b1 | 3095 | goto out; |
4156d998 AE |
3096 | } |
3097 | if (!rbd_dev_ondisk_valid(ondisk)) { | |
3098 | ret = -ENXIO; | |
06ecc6cb | 3099 | rbd_warn(rbd_dev, "invalid header"); |
662518b1 | 3100 | goto out; |
81e759fb | 3101 | } |
602adf40 | 3102 | |
4156d998 AE |
3103 | names_size = le64_to_cpu(ondisk->snap_names_len); |
3104 | want_count = snap_count; | |
3105 | snap_count = le32_to_cpu(ondisk->snap_count); | |
3106 | } while (snap_count != want_count); | |
00f1f36f | 3107 | |
662518b1 AE |
3108 | ret = rbd_header_from_disk(rbd_dev, ondisk); |
3109 | out: | |
4156d998 AE |
3110 | kfree(ondisk); |
3111 | ||
3112 | return ret; | |
602adf40 YS |
3113 | } |
3114 | ||
15228ede AE |
3115 | /* |
3116 | * Clear the rbd device's EXISTS flag if the snapshot it's mapped to | |
3117 | * has disappeared from the (just updated) snapshot context. | |
3118 | */ | |
3119 | static void rbd_exists_validate(struct rbd_device *rbd_dev) | |
3120 | { | |
3121 | u64 snap_id; | |
3122 | ||
3123 | if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) | |
3124 | return; | |
3125 | ||
3126 | snap_id = rbd_dev->spec->snap_id; | |
3127 | if (snap_id == CEPH_NOSNAP) | |
3128 | return; | |
3129 | ||
3130 | if (rbd_dev_snap_index(rbd_dev, snap_id) == BAD_SNAP_INDEX) | |
3131 | clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); | |
3132 | } | |
3133 | ||
cc4a38bd | 3134 | static int rbd_dev_refresh(struct rbd_device *rbd_dev) |
1fe5e993 | 3135 | { |
e627db08 | 3136 | u64 mapping_size; |
1fe5e993 AE |
3137 | int ret; |
3138 | ||
117973fb | 3139 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
e627db08 | 3140 | mapping_size = rbd_dev->mapping.size; |
1fe5e993 | 3141 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
117973fb | 3142 | if (rbd_dev->image_format == 1) |
99a41ebc | 3143 | ret = rbd_dev_v1_header_info(rbd_dev); |
117973fb | 3144 | else |
2df3fac7 | 3145 | ret = rbd_dev_v2_header_info(rbd_dev); |
15228ede AE |
3146 | |
3147 | /* If it's a mapped snapshot, validate its EXISTS flag */ | |
3148 | ||
3149 | rbd_exists_validate(rbd_dev); | |
1fe5e993 | 3150 | mutex_unlock(&ctl_mutex); |
00a653e2 AE |
3151 | if (mapping_size != rbd_dev->mapping.size) { |
3152 | sector_t size; | |
3153 | ||
3154 | size = (sector_t)rbd_dev->mapping.size / SECTOR_SIZE; | |
3155 | dout("setting size to %llu sectors", (unsigned long long)size); | |
3156 | set_capacity(rbd_dev->disk, size); | |
a3fbe5d4 | 3157 | revalidate_disk(rbd_dev->disk); |
00a653e2 | 3158 | } |
1fe5e993 AE |
3159 | |
3160 | return ret; | |
3161 | } | |
3162 | ||
602adf40 YS |
3163 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
3164 | { | |
3165 | struct gendisk *disk; | |
3166 | struct request_queue *q; | |
593a9e7b | 3167 | u64 segment_size; |
602adf40 | 3168 | |
602adf40 | 3169 | /* create gendisk info */ |
602adf40 YS |
3170 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); |
3171 | if (!disk) | |
1fcdb8aa | 3172 | return -ENOMEM; |
602adf40 | 3173 | |
f0f8cef5 | 3174 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
de71a297 | 3175 | rbd_dev->dev_id); |
602adf40 YS |
3176 | disk->major = rbd_dev->major; |
3177 | disk->first_minor = 0; | |
3178 | disk->fops = &rbd_bd_ops; | |
3179 | disk->private_data = rbd_dev; | |
3180 | ||
bf0d5f50 | 3181 | q = blk_init_queue(rbd_request_fn, &rbd_dev->lock); |
602adf40 YS |
3182 | if (!q) |
3183 | goto out_disk; | |
029bcbd8 | 3184 | |
593a9e7b AE |
3185 | /* We use the default size, but let's be explicit about it. */ |
3186 | blk_queue_physical_block_size(q, SECTOR_SIZE); | |
3187 | ||
029bcbd8 | 3188 | /* set io sizes to object size */ |
593a9e7b AE |
3189 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
3190 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); | |
3191 | blk_queue_max_segment_size(q, segment_size); | |
3192 | blk_queue_io_min(q, segment_size); | |
3193 | blk_queue_io_opt(q, segment_size); | |
029bcbd8 | 3194 | |
602adf40 YS |
3195 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
3196 | disk->queue = q; | |
3197 | ||
3198 | q->queuedata = rbd_dev; | |
3199 | ||
3200 | rbd_dev->disk = disk; | |
602adf40 | 3201 | |
602adf40 | 3202 | return 0; |
602adf40 YS |
3203 | out_disk: |
3204 | put_disk(disk); | |
1fcdb8aa AE |
3205 | |
3206 | return -ENOMEM; | |
602adf40 YS |
3207 | } |
3208 | ||
dfc5606d YS |
3209 | /* |
3210 | sysfs | |
3211 | */ | |
3212 | ||
593a9e7b AE |
3213 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
3214 | { | |
3215 | return container_of(dev, struct rbd_device, dev); | |
3216 | } | |
3217 | ||
dfc5606d YS |
3218 | static ssize_t rbd_size_show(struct device *dev, |
3219 | struct device_attribute *attr, char *buf) | |
3220 | { | |
593a9e7b | 3221 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
a51aa0c0 | 3222 | |
fc71d833 AE |
3223 | return sprintf(buf, "%llu\n", |
3224 | (unsigned long long)rbd_dev->mapping.size); | |
dfc5606d YS |
3225 | } |
3226 | ||
34b13184 AE |
3227 | /* |
3228 | * Note this shows the features for whatever's mapped, which is not | |
3229 | * necessarily the base image. | |
3230 | */ | |
3231 | static ssize_t rbd_features_show(struct device *dev, | |
3232 | struct device_attribute *attr, char *buf) | |
3233 | { | |
3234 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
3235 | ||
3236 | return sprintf(buf, "0x%016llx\n", | |
fc71d833 | 3237 | (unsigned long long)rbd_dev->mapping.features); |
34b13184 AE |
3238 | } |
3239 | ||
dfc5606d YS |
3240 | static ssize_t rbd_major_show(struct device *dev, |
3241 | struct device_attribute *attr, char *buf) | |
3242 | { | |
593a9e7b | 3243 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 3244 | |
fc71d833 AE |
3245 | if (rbd_dev->major) |
3246 | return sprintf(buf, "%d\n", rbd_dev->major); | |
3247 | ||
3248 | return sprintf(buf, "(none)\n"); | |
3249 | ||
dfc5606d YS |
3250 | } |
3251 | ||
3252 | static ssize_t rbd_client_id_show(struct device *dev, | |
3253 | struct device_attribute *attr, char *buf) | |
602adf40 | 3254 | { |
593a9e7b | 3255 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 3256 | |
1dbb4399 AE |
3257 | return sprintf(buf, "client%lld\n", |
3258 | ceph_client_id(rbd_dev->rbd_client->client)); | |
602adf40 YS |
3259 | } |
3260 | ||
dfc5606d YS |
3261 | static ssize_t rbd_pool_show(struct device *dev, |
3262 | struct device_attribute *attr, char *buf) | |
602adf40 | 3263 | { |
593a9e7b | 3264 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 3265 | |
0d7dbfce | 3266 | return sprintf(buf, "%s\n", rbd_dev->spec->pool_name); |
dfc5606d YS |
3267 | } |
3268 | ||
9bb2f334 AE |
3269 | static ssize_t rbd_pool_id_show(struct device *dev, |
3270 | struct device_attribute *attr, char *buf) | |
3271 | { | |
3272 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
3273 | ||
0d7dbfce | 3274 | return sprintf(buf, "%llu\n", |
fc71d833 | 3275 | (unsigned long long) rbd_dev->spec->pool_id); |
9bb2f334 AE |
3276 | } |
3277 | ||
dfc5606d YS |
3278 | static ssize_t rbd_name_show(struct device *dev, |
3279 | struct device_attribute *attr, char *buf) | |
3280 | { | |
593a9e7b | 3281 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 3282 | |
a92ffdf8 AE |
3283 | if (rbd_dev->spec->image_name) |
3284 | return sprintf(buf, "%s\n", rbd_dev->spec->image_name); | |
3285 | ||
3286 | return sprintf(buf, "(unknown)\n"); | |
dfc5606d YS |
3287 | } |
3288 | ||
589d30e0 AE |
3289 | static ssize_t rbd_image_id_show(struct device *dev, |
3290 | struct device_attribute *attr, char *buf) | |
3291 | { | |
3292 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
3293 | ||
0d7dbfce | 3294 | return sprintf(buf, "%s\n", rbd_dev->spec->image_id); |
589d30e0 AE |
3295 | } |
3296 | ||
34b13184 AE |
3297 | /* |
3298 | * Shows the name of the currently-mapped snapshot (or | |
3299 | * RBD_SNAP_HEAD_NAME for the base image). | |
3300 | */ | |
dfc5606d YS |
3301 | static ssize_t rbd_snap_show(struct device *dev, |
3302 | struct device_attribute *attr, | |
3303 | char *buf) | |
3304 | { | |
593a9e7b | 3305 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
dfc5606d | 3306 | |
0d7dbfce | 3307 | return sprintf(buf, "%s\n", rbd_dev->spec->snap_name); |
dfc5606d YS |
3308 | } |
3309 | ||
86b00e0d AE |
3310 | /* |
3311 | * For an rbd v2 image, shows the pool id, image id, and snapshot id | |
3312 | * for the parent image. If there is no parent, simply shows | |
3313 | * "(no parent image)". | |
3314 | */ | |
3315 | static ssize_t rbd_parent_show(struct device *dev, | |
3316 | struct device_attribute *attr, | |
3317 | char *buf) | |
3318 | { | |
3319 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); | |
3320 | struct rbd_spec *spec = rbd_dev->parent_spec; | |
3321 | int count; | |
3322 | char *bufp = buf; | |
3323 | ||
3324 | if (!spec) | |
3325 | return sprintf(buf, "(no parent image)\n"); | |
3326 | ||
3327 | count = sprintf(bufp, "pool_id %llu\npool_name %s\n", | |
3328 | (unsigned long long) spec->pool_id, spec->pool_name); | |
3329 | if (count < 0) | |
3330 | return count; | |
3331 | bufp += count; | |
3332 | ||
3333 | count = sprintf(bufp, "image_id %s\nimage_name %s\n", spec->image_id, | |
3334 | spec->image_name ? spec->image_name : "(unknown)"); | |
3335 | if (count < 0) | |
3336 | return count; | |
3337 | bufp += count; | |
3338 | ||
3339 | count = sprintf(bufp, "snap_id %llu\nsnap_name %s\n", | |
3340 | (unsigned long long) spec->snap_id, spec->snap_name); | |
3341 | if (count < 0) | |
3342 | return count; | |
3343 | bufp += count; | |
3344 | ||
3345 | count = sprintf(bufp, "overlap %llu\n", rbd_dev->parent_overlap); | |
3346 | if (count < 0) | |
3347 | return count; | |
3348 | bufp += count; | |
3349 | ||
3350 | return (ssize_t) (bufp - buf); | |
3351 | } | |
3352 | ||
dfc5606d YS |
3353 | static ssize_t rbd_image_refresh(struct device *dev, |
3354 | struct device_attribute *attr, | |
3355 | const char *buf, | |
3356 | size_t size) | |
3357 | { | |
593a9e7b | 3358 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
b813623a | 3359 | int ret; |
602adf40 | 3360 | |
cc4a38bd | 3361 | ret = rbd_dev_refresh(rbd_dev); |
e627db08 AE |
3362 | if (ret) |
3363 | rbd_warn(rbd_dev, ": manual header refresh error (%d)\n", ret); | |
b813623a AE |
3364 | |
3365 | return ret < 0 ? ret : size; | |
dfc5606d | 3366 | } |
602adf40 | 3367 | |
dfc5606d | 3368 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
34b13184 | 3369 | static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL); |
dfc5606d YS |
3370 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); |
3371 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); | |
3372 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); | |
9bb2f334 | 3373 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
dfc5606d | 3374 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
589d30e0 | 3375 | static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL); |
dfc5606d YS |
3376 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); |
3377 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); | |
86b00e0d | 3378 | static DEVICE_ATTR(parent, S_IRUGO, rbd_parent_show, NULL); |
dfc5606d YS |
3379 | |
3380 | static struct attribute *rbd_attrs[] = { | |
3381 | &dev_attr_size.attr, | |
34b13184 | 3382 | &dev_attr_features.attr, |
dfc5606d YS |
3383 | &dev_attr_major.attr, |
3384 | &dev_attr_client_id.attr, | |
3385 | &dev_attr_pool.attr, | |
9bb2f334 | 3386 | &dev_attr_pool_id.attr, |
dfc5606d | 3387 | &dev_attr_name.attr, |
589d30e0 | 3388 | &dev_attr_image_id.attr, |
dfc5606d | 3389 | &dev_attr_current_snap.attr, |
86b00e0d | 3390 | &dev_attr_parent.attr, |
dfc5606d | 3391 | &dev_attr_refresh.attr, |
dfc5606d YS |
3392 | NULL |
3393 | }; | |
3394 | ||
3395 | static struct attribute_group rbd_attr_group = { | |
3396 | .attrs = rbd_attrs, | |
3397 | }; | |
3398 | ||
3399 | static const struct attribute_group *rbd_attr_groups[] = { | |
3400 | &rbd_attr_group, | |
3401 | NULL | |
3402 | }; | |
3403 | ||
3404 | static void rbd_sysfs_dev_release(struct device *dev) | |
3405 | { | |
3406 | } | |
3407 | ||
3408 | static struct device_type rbd_device_type = { | |
3409 | .name = "rbd", | |
3410 | .groups = rbd_attr_groups, | |
3411 | .release = rbd_sysfs_dev_release, | |
3412 | }; | |
3413 | ||
8b8fb99c AE |
3414 | static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec) |
3415 | { | |
3416 | kref_get(&spec->kref); | |
3417 | ||
3418 | return spec; | |
3419 | } | |
3420 | ||
3421 | static void rbd_spec_free(struct kref *kref); | |
3422 | static void rbd_spec_put(struct rbd_spec *spec) | |
3423 | { | |
3424 | if (spec) | |
3425 | kref_put(&spec->kref, rbd_spec_free); | |
3426 | } | |
3427 | ||
3428 | static struct rbd_spec *rbd_spec_alloc(void) | |
3429 | { | |
3430 | struct rbd_spec *spec; | |
3431 | ||
3432 | spec = kzalloc(sizeof (*spec), GFP_KERNEL); | |
3433 | if (!spec) | |
3434 | return NULL; | |
3435 | kref_init(&spec->kref); | |
3436 | ||
8b8fb99c AE |
3437 | return spec; |
3438 | } | |
3439 | ||
3440 | static void rbd_spec_free(struct kref *kref) | |
3441 | { | |
3442 | struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref); | |
3443 | ||
3444 | kfree(spec->pool_name); | |
3445 | kfree(spec->image_id); | |
3446 | kfree(spec->image_name); | |
3447 | kfree(spec->snap_name); | |
3448 | kfree(spec); | |
3449 | } | |
3450 | ||
cc344fa1 | 3451 | static struct rbd_device *rbd_dev_create(struct rbd_client *rbdc, |
c53d5893 AE |
3452 | struct rbd_spec *spec) |
3453 | { | |
3454 | struct rbd_device *rbd_dev; | |
3455 | ||
3456 | rbd_dev = kzalloc(sizeof (*rbd_dev), GFP_KERNEL); | |
3457 | if (!rbd_dev) | |
3458 | return NULL; | |
3459 | ||
3460 | spin_lock_init(&rbd_dev->lock); | |
6d292906 | 3461 | rbd_dev->flags = 0; |
c53d5893 | 3462 | INIT_LIST_HEAD(&rbd_dev->node); |
c53d5893 AE |
3463 | init_rwsem(&rbd_dev->header_rwsem); |
3464 | ||
3465 | rbd_dev->spec = spec; | |
3466 | rbd_dev->rbd_client = rbdc; | |
3467 | ||
0903e875 AE |
3468 | /* Initialize the layout used for all rbd requests */ |
3469 | ||
3470 | rbd_dev->layout.fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
3471 | rbd_dev->layout.fl_stripe_count = cpu_to_le32(1); | |
3472 | rbd_dev->layout.fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); | |
3473 | rbd_dev->layout.fl_pg_pool = cpu_to_le32((u32) spec->pool_id); | |
3474 | ||
c53d5893 AE |
3475 | return rbd_dev; |
3476 | } | |
3477 | ||
3478 | static void rbd_dev_destroy(struct rbd_device *rbd_dev) | |
3479 | { | |
c53d5893 AE |
3480 | rbd_put_client(rbd_dev->rbd_client); |
3481 | rbd_spec_put(rbd_dev->spec); | |
3482 | kfree(rbd_dev); | |
3483 | } | |
3484 | ||
9d475de5 AE |
3485 | /* |
3486 | * Get the size and object order for an image snapshot, or if | |
3487 | * snap_id is CEPH_NOSNAP, gets this information for the base | |
3488 | * image. | |
3489 | */ | |
3490 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, | |
3491 | u8 *order, u64 *snap_size) | |
3492 | { | |
3493 | __le64 snapid = cpu_to_le64(snap_id); | |
3494 | int ret; | |
3495 | struct { | |
3496 | u8 order; | |
3497 | __le64 size; | |
3498 | } __attribute__ ((packed)) size_buf = { 0 }; | |
3499 | ||
36be9a76 | 3500 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
9d475de5 | 3501 | "rbd", "get_size", |
4157976b | 3502 | &snapid, sizeof (snapid), |
e2a58ee5 | 3503 | &size_buf, sizeof (size_buf)); |
36be9a76 | 3504 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
9d475de5 AE |
3505 | if (ret < 0) |
3506 | return ret; | |
57385b51 AE |
3507 | if (ret < sizeof (size_buf)) |
3508 | return -ERANGE; | |
9d475de5 | 3509 | |
c86f86e9 AE |
3510 | if (order) |
3511 | *order = size_buf.order; | |
9d475de5 AE |
3512 | *snap_size = le64_to_cpu(size_buf.size); |
3513 | ||
3514 | dout(" snap_id 0x%016llx order = %u, snap_size = %llu\n", | |
57385b51 AE |
3515 | (unsigned long long)snap_id, (unsigned int)*order, |
3516 | (unsigned long long)*snap_size); | |
9d475de5 AE |
3517 | |
3518 | return 0; | |
3519 | } | |
3520 | ||
3521 | static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) | |
3522 | { | |
3523 | return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, | |
3524 | &rbd_dev->header.obj_order, | |
3525 | &rbd_dev->header.image_size); | |
3526 | } | |
3527 | ||
1e130199 AE |
3528 | static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) |
3529 | { | |
3530 | void *reply_buf; | |
3531 | int ret; | |
3532 | void *p; | |
3533 | ||
3534 | reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL); | |
3535 | if (!reply_buf) | |
3536 | return -ENOMEM; | |
3537 | ||
36be9a76 | 3538 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
4157976b | 3539 | "rbd", "get_object_prefix", NULL, 0, |
e2a58ee5 | 3540 | reply_buf, RBD_OBJ_PREFIX_LEN_MAX); |
36be9a76 | 3541 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
1e130199 AE |
3542 | if (ret < 0) |
3543 | goto out; | |
3544 | ||
3545 | p = reply_buf; | |
3546 | rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, | |
57385b51 AE |
3547 | p + ret, NULL, GFP_NOIO); |
3548 | ret = 0; | |
1e130199 AE |
3549 | |
3550 | if (IS_ERR(rbd_dev->header.object_prefix)) { | |
3551 | ret = PTR_ERR(rbd_dev->header.object_prefix); | |
3552 | rbd_dev->header.object_prefix = NULL; | |
3553 | } else { | |
3554 | dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); | |
3555 | } | |
1e130199 AE |
3556 | out: |
3557 | kfree(reply_buf); | |
3558 | ||
3559 | return ret; | |
3560 | } | |
3561 | ||
b1b5402a AE |
3562 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
3563 | u64 *snap_features) | |
3564 | { | |
3565 | __le64 snapid = cpu_to_le64(snap_id); | |
3566 | struct { | |
3567 | __le64 features; | |
3568 | __le64 incompat; | |
4157976b | 3569 | } __attribute__ ((packed)) features_buf = { 0 }; |
d889140c | 3570 | u64 incompat; |
b1b5402a AE |
3571 | int ret; |
3572 | ||
36be9a76 | 3573 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
b1b5402a | 3574 | "rbd", "get_features", |
4157976b | 3575 | &snapid, sizeof (snapid), |
e2a58ee5 | 3576 | &features_buf, sizeof (features_buf)); |
36be9a76 | 3577 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
b1b5402a AE |
3578 | if (ret < 0) |
3579 | return ret; | |
57385b51 AE |
3580 | if (ret < sizeof (features_buf)) |
3581 | return -ERANGE; | |
d889140c AE |
3582 | |
3583 | incompat = le64_to_cpu(features_buf.incompat); | |
5cbf6f12 | 3584 | if (incompat & ~RBD_FEATURES_SUPPORTED) |
b8f5c6ed | 3585 | return -ENXIO; |
d889140c | 3586 | |
b1b5402a AE |
3587 | *snap_features = le64_to_cpu(features_buf.features); |
3588 | ||
3589 | dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", | |
57385b51 AE |
3590 | (unsigned long long)snap_id, |
3591 | (unsigned long long)*snap_features, | |
3592 | (unsigned long long)le64_to_cpu(features_buf.incompat)); | |
b1b5402a AE |
3593 | |
3594 | return 0; | |
3595 | } | |
3596 | ||
3597 | static int rbd_dev_v2_features(struct rbd_device *rbd_dev) | |
3598 | { | |
3599 | return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, | |
3600 | &rbd_dev->header.features); | |
3601 | } | |
3602 | ||
86b00e0d AE |
3603 | static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) |
3604 | { | |
3605 | struct rbd_spec *parent_spec; | |
3606 | size_t size; | |
3607 | void *reply_buf = NULL; | |
3608 | __le64 snapid; | |
3609 | void *p; | |
3610 | void *end; | |
642a2537 | 3611 | u64 pool_id; |
86b00e0d AE |
3612 | char *image_id; |
3613 | u64 overlap; | |
86b00e0d AE |
3614 | int ret; |
3615 | ||
3616 | parent_spec = rbd_spec_alloc(); | |
3617 | if (!parent_spec) | |
3618 | return -ENOMEM; | |
3619 | ||
3620 | size = sizeof (__le64) + /* pool_id */ | |
3621 | sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX + /* image_id */ | |
3622 | sizeof (__le64) + /* snap_id */ | |
3623 | sizeof (__le64); /* overlap */ | |
3624 | reply_buf = kmalloc(size, GFP_KERNEL); | |
3625 | if (!reply_buf) { | |
3626 | ret = -ENOMEM; | |
3627 | goto out_err; | |
3628 | } | |
3629 | ||
3630 | snapid = cpu_to_le64(CEPH_NOSNAP); | |
36be9a76 | 3631 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
86b00e0d | 3632 | "rbd", "get_parent", |
4157976b | 3633 | &snapid, sizeof (snapid), |
e2a58ee5 | 3634 | reply_buf, size); |
36be9a76 | 3635 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
86b00e0d AE |
3636 | if (ret < 0) |
3637 | goto out_err; | |
3638 | ||
86b00e0d | 3639 | p = reply_buf; |
57385b51 AE |
3640 | end = reply_buf + ret; |
3641 | ret = -ERANGE; | |
642a2537 AE |
3642 | ceph_decode_64_safe(&p, end, pool_id, out_err); |
3643 | if (pool_id == CEPH_NOPOOL) | |
86b00e0d AE |
3644 | goto out; /* No parent? No problem. */ |
3645 | ||
0903e875 AE |
3646 | /* The ceph file layout needs to fit pool id in 32 bits */ |
3647 | ||
3648 | ret = -EIO; | |
642a2537 | 3649 | if (pool_id > (u64)U32_MAX) { |
c0cd10db | 3650 | rbd_warn(NULL, "parent pool id too large (%llu > %u)\n", |
642a2537 | 3651 | (unsigned long long)pool_id, U32_MAX); |
57385b51 | 3652 | goto out_err; |
c0cd10db | 3653 | } |
642a2537 | 3654 | parent_spec->pool_id = pool_id; |
0903e875 | 3655 | |
979ed480 | 3656 | image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
86b00e0d AE |
3657 | if (IS_ERR(image_id)) { |
3658 | ret = PTR_ERR(image_id); | |
3659 | goto out_err; | |
3660 | } | |
3661 | parent_spec->image_id = image_id; | |
3662 | ceph_decode_64_safe(&p, end, parent_spec->snap_id, out_err); | |
3663 | ceph_decode_64_safe(&p, end, overlap, out_err); | |
3664 | ||
70cf49cf | 3665 | if (overlap) { |
642a2537 | 3666 | rbd_spec_put(rbd_dev->parent_spec); |
70cf49cf AE |
3667 | rbd_dev->parent_spec = parent_spec; |
3668 | parent_spec = NULL; /* rbd_dev now owns this */ | |
3669 | rbd_dev->parent_overlap = overlap; | |
3670 | } else { | |
3671 | rbd_warn(rbd_dev, "ignoring parent of clone with overlap 0\n"); | |
3672 | } | |
86b00e0d AE |
3673 | out: |
3674 | ret = 0; | |
3675 | out_err: | |
3676 | kfree(reply_buf); | |
3677 | rbd_spec_put(parent_spec); | |
3678 | ||
3679 | return ret; | |
3680 | } | |
3681 | ||
cc070d59 AE |
3682 | static int rbd_dev_v2_striping_info(struct rbd_device *rbd_dev) |
3683 | { | |
3684 | struct { | |
3685 | __le64 stripe_unit; | |
3686 | __le64 stripe_count; | |
3687 | } __attribute__ ((packed)) striping_info_buf = { 0 }; | |
3688 | size_t size = sizeof (striping_info_buf); | |
3689 | void *p; | |
3690 | u64 obj_size; | |
3691 | u64 stripe_unit; | |
3692 | u64 stripe_count; | |
3693 | int ret; | |
3694 | ||
3695 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, | |
3696 | "rbd", "get_stripe_unit_count", NULL, 0, | |
e2a58ee5 | 3697 | (char *)&striping_info_buf, size); |
cc070d59 AE |
3698 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
3699 | if (ret < 0) | |
3700 | return ret; | |
3701 | if (ret < size) | |
3702 | return -ERANGE; | |
3703 | ||
3704 | /* | |
3705 | * We don't actually support the "fancy striping" feature | |
3706 | * (STRIPINGV2) yet, but if the striping sizes are the | |
3707 | * defaults the behavior is the same as before. So find | |
3708 | * out, and only fail if the image has non-default values. | |
3709 | */ | |
3710 | ret = -EINVAL; | |
3711 | obj_size = (u64)1 << rbd_dev->header.obj_order; | |
3712 | p = &striping_info_buf; | |
3713 | stripe_unit = ceph_decode_64(&p); | |
3714 | if (stripe_unit != obj_size) { | |
3715 | rbd_warn(rbd_dev, "unsupported stripe unit " | |
3716 | "(got %llu want %llu)", | |
3717 | stripe_unit, obj_size); | |
3718 | return -EINVAL; | |
3719 | } | |
3720 | stripe_count = ceph_decode_64(&p); | |
3721 | if (stripe_count != 1) { | |
3722 | rbd_warn(rbd_dev, "unsupported stripe count " | |
3723 | "(got %llu want 1)", stripe_count); | |
3724 | return -EINVAL; | |
3725 | } | |
500d0c0f AE |
3726 | rbd_dev->header.stripe_unit = stripe_unit; |
3727 | rbd_dev->header.stripe_count = stripe_count; | |
cc070d59 AE |
3728 | |
3729 | return 0; | |
3730 | } | |
3731 | ||
9e15b77d AE |
3732 | static char *rbd_dev_image_name(struct rbd_device *rbd_dev) |
3733 | { | |
3734 | size_t image_id_size; | |
3735 | char *image_id; | |
3736 | void *p; | |
3737 | void *end; | |
3738 | size_t size; | |
3739 | void *reply_buf = NULL; | |
3740 | size_t len = 0; | |
3741 | char *image_name = NULL; | |
3742 | int ret; | |
3743 | ||
3744 | rbd_assert(!rbd_dev->spec->image_name); | |
3745 | ||
69e7a02f AE |
3746 | len = strlen(rbd_dev->spec->image_id); |
3747 | image_id_size = sizeof (__le32) + len; | |
9e15b77d AE |
3748 | image_id = kmalloc(image_id_size, GFP_KERNEL); |
3749 | if (!image_id) | |
3750 | return NULL; | |
3751 | ||
3752 | p = image_id; | |
4157976b | 3753 | end = image_id + image_id_size; |
57385b51 | 3754 | ceph_encode_string(&p, end, rbd_dev->spec->image_id, (u32)len); |
9e15b77d AE |
3755 | |
3756 | size = sizeof (__le32) + RBD_IMAGE_NAME_LEN_MAX; | |
3757 | reply_buf = kmalloc(size, GFP_KERNEL); | |
3758 | if (!reply_buf) | |
3759 | goto out; | |
3760 | ||
36be9a76 | 3761 | ret = rbd_obj_method_sync(rbd_dev, RBD_DIRECTORY, |
9e15b77d AE |
3762 | "rbd", "dir_get_name", |
3763 | image_id, image_id_size, | |
e2a58ee5 | 3764 | reply_buf, size); |
9e15b77d AE |
3765 | if (ret < 0) |
3766 | goto out; | |
3767 | p = reply_buf; | |
f40eb349 AE |
3768 | end = reply_buf + ret; |
3769 | ||
9e15b77d AE |
3770 | image_name = ceph_extract_encoded_string(&p, end, &len, GFP_KERNEL); |
3771 | if (IS_ERR(image_name)) | |
3772 | image_name = NULL; | |
3773 | else | |
3774 | dout("%s: name is %s len is %zd\n", __func__, image_name, len); | |
3775 | out: | |
3776 | kfree(reply_buf); | |
3777 | kfree(image_id); | |
3778 | ||
3779 | return image_name; | |
3780 | } | |
3781 | ||
2ad3d716 AE |
3782 | static u64 rbd_v1_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) |
3783 | { | |
3784 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; | |
3785 | const char *snap_name; | |
3786 | u32 which = 0; | |
3787 | ||
3788 | /* Skip over names until we find the one we are looking for */ | |
3789 | ||
3790 | snap_name = rbd_dev->header.snap_names; | |
3791 | while (which < snapc->num_snaps) { | |
3792 | if (!strcmp(name, snap_name)) | |
3793 | return snapc->snaps[which]; | |
3794 | snap_name += strlen(snap_name) + 1; | |
3795 | which++; | |
3796 | } | |
3797 | return CEPH_NOSNAP; | |
3798 | } | |
3799 | ||
3800 | static u64 rbd_v2_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) | |
3801 | { | |
3802 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; | |
3803 | u32 which; | |
3804 | bool found = false; | |
3805 | u64 snap_id; | |
3806 | ||
3807 | for (which = 0; !found && which < snapc->num_snaps; which++) { | |
3808 | const char *snap_name; | |
3809 | ||
3810 | snap_id = snapc->snaps[which]; | |
3811 | snap_name = rbd_dev_v2_snap_name(rbd_dev, snap_id); | |
3812 | if (IS_ERR(snap_name)) | |
3813 | break; | |
3814 | found = !strcmp(name, snap_name); | |
3815 | kfree(snap_name); | |
3816 | } | |
3817 | return found ? snap_id : CEPH_NOSNAP; | |
3818 | } | |
3819 | ||
3820 | /* | |
3821 | * Assumes name is never RBD_SNAP_HEAD_NAME; returns CEPH_NOSNAP if | |
3822 | * no snapshot by that name is found, or if an error occurs. | |
3823 | */ | |
3824 | static u64 rbd_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) | |
3825 | { | |
3826 | if (rbd_dev->image_format == 1) | |
3827 | return rbd_v1_snap_id_by_name(rbd_dev, name); | |
3828 | ||
3829 | return rbd_v2_snap_id_by_name(rbd_dev, name); | |
3830 | } | |
3831 | ||
9e15b77d | 3832 | /* |
2e9f7f1c AE |
3833 | * When an rbd image has a parent image, it is identified by the |
3834 | * pool, image, and snapshot ids (not names). This function fills | |
3835 | * in the names for those ids. (It's OK if we can't figure out the | |
3836 | * name for an image id, but the pool and snapshot ids should always | |
3837 | * exist and have names.) All names in an rbd spec are dynamically | |
3838 | * allocated. | |
e1d4213f AE |
3839 | * |
3840 | * When an image being mapped (not a parent) is probed, we have the | |
3841 | * pool name and pool id, image name and image id, and the snapshot | |
3842 | * name. The only thing we're missing is the snapshot id. | |
9e15b77d | 3843 | */ |
2e9f7f1c | 3844 | static int rbd_dev_spec_update(struct rbd_device *rbd_dev) |
9e15b77d | 3845 | { |
2e9f7f1c AE |
3846 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
3847 | struct rbd_spec *spec = rbd_dev->spec; | |
3848 | const char *pool_name; | |
3849 | const char *image_name; | |
3850 | const char *snap_name; | |
9e15b77d AE |
3851 | int ret; |
3852 | ||
e1d4213f AE |
3853 | /* |
3854 | * An image being mapped will have the pool name (etc.), but | |
3855 | * we need to look up the snapshot id. | |
3856 | */ | |
2e9f7f1c AE |
3857 | if (spec->pool_name) { |
3858 | if (strcmp(spec->snap_name, RBD_SNAP_HEAD_NAME)) { | |
2ad3d716 | 3859 | u64 snap_id; |
e1d4213f | 3860 | |
2ad3d716 AE |
3861 | snap_id = rbd_snap_id_by_name(rbd_dev, spec->snap_name); |
3862 | if (snap_id == CEPH_NOSNAP) | |
e1d4213f | 3863 | return -ENOENT; |
2ad3d716 | 3864 | spec->snap_id = snap_id; |
e1d4213f | 3865 | } else { |
2e9f7f1c | 3866 | spec->snap_id = CEPH_NOSNAP; |
e1d4213f AE |
3867 | } |
3868 | ||
3869 | return 0; | |
3870 | } | |
9e15b77d | 3871 | |
2e9f7f1c | 3872 | /* Get the pool name; we have to make our own copy of this */ |
9e15b77d | 3873 | |
2e9f7f1c AE |
3874 | pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, spec->pool_id); |
3875 | if (!pool_name) { | |
3876 | rbd_warn(rbd_dev, "no pool with id %llu", spec->pool_id); | |
935dc89f AE |
3877 | return -EIO; |
3878 | } | |
2e9f7f1c AE |
3879 | pool_name = kstrdup(pool_name, GFP_KERNEL); |
3880 | if (!pool_name) | |
9e15b77d AE |
3881 | return -ENOMEM; |
3882 | ||
3883 | /* Fetch the image name; tolerate failure here */ | |
3884 | ||
2e9f7f1c AE |
3885 | image_name = rbd_dev_image_name(rbd_dev); |
3886 | if (!image_name) | |
06ecc6cb | 3887 | rbd_warn(rbd_dev, "unable to get image name"); |
9e15b77d | 3888 | |
2e9f7f1c | 3889 | /* Look up the snapshot name, and make a copy */ |
9e15b77d | 3890 | |
2e9f7f1c | 3891 | snap_name = rbd_snap_name(rbd_dev, spec->snap_id); |
2e9f7f1c AE |
3892 | if (!snap_name) { |
3893 | ret = -ENOMEM; | |
9e15b77d | 3894 | goto out_err; |
2e9f7f1c AE |
3895 | } |
3896 | ||
3897 | spec->pool_name = pool_name; | |
3898 | spec->image_name = image_name; | |
3899 | spec->snap_name = snap_name; | |
9e15b77d AE |
3900 | |
3901 | return 0; | |
3902 | out_err: | |
2e9f7f1c AE |
3903 | kfree(image_name); |
3904 | kfree(pool_name); | |
9e15b77d AE |
3905 | |
3906 | return ret; | |
3907 | } | |
3908 | ||
cc4a38bd | 3909 | static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev) |
35d489f9 AE |
3910 | { |
3911 | size_t size; | |
3912 | int ret; | |
3913 | void *reply_buf; | |
3914 | void *p; | |
3915 | void *end; | |
3916 | u64 seq; | |
3917 | u32 snap_count; | |
3918 | struct ceph_snap_context *snapc; | |
3919 | u32 i; | |
3920 | ||
3921 | /* | |
3922 | * We'll need room for the seq value (maximum snapshot id), | |
3923 | * snapshot count, and array of that many snapshot ids. | |
3924 | * For now we have a fixed upper limit on the number we're | |
3925 | * prepared to receive. | |
3926 | */ | |
3927 | size = sizeof (__le64) + sizeof (__le32) + | |
3928 | RBD_MAX_SNAP_COUNT * sizeof (__le64); | |
3929 | reply_buf = kzalloc(size, GFP_KERNEL); | |
3930 | if (!reply_buf) | |
3931 | return -ENOMEM; | |
3932 | ||
36be9a76 | 3933 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
4157976b | 3934 | "rbd", "get_snapcontext", NULL, 0, |
e2a58ee5 | 3935 | reply_buf, size); |
36be9a76 | 3936 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
35d489f9 AE |
3937 | if (ret < 0) |
3938 | goto out; | |
3939 | ||
35d489f9 | 3940 | p = reply_buf; |
57385b51 AE |
3941 | end = reply_buf + ret; |
3942 | ret = -ERANGE; | |
35d489f9 AE |
3943 | ceph_decode_64_safe(&p, end, seq, out); |
3944 | ceph_decode_32_safe(&p, end, snap_count, out); | |
3945 | ||
3946 | /* | |
3947 | * Make sure the reported number of snapshot ids wouldn't go | |
3948 | * beyond the end of our buffer. But before checking that, | |
3949 | * make sure the computed size of the snapshot context we | |
3950 | * allocate is representable in a size_t. | |
3951 | */ | |
3952 | if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) | |
3953 | / sizeof (u64)) { | |
3954 | ret = -EINVAL; | |
3955 | goto out; | |
3956 | } | |
3957 | if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) | |
3958 | goto out; | |
468521c1 | 3959 | ret = 0; |
35d489f9 | 3960 | |
812164f8 | 3961 | snapc = ceph_create_snap_context(snap_count, GFP_KERNEL); |
35d489f9 AE |
3962 | if (!snapc) { |
3963 | ret = -ENOMEM; | |
3964 | goto out; | |
3965 | } | |
35d489f9 | 3966 | snapc->seq = seq; |
35d489f9 AE |
3967 | for (i = 0; i < snap_count; i++) |
3968 | snapc->snaps[i] = ceph_decode_64(&p); | |
3969 | ||
49ece554 | 3970 | ceph_put_snap_context(rbd_dev->header.snapc); |
35d489f9 AE |
3971 | rbd_dev->header.snapc = snapc; |
3972 | ||
3973 | dout(" snap context seq = %llu, snap_count = %u\n", | |
57385b51 | 3974 | (unsigned long long)seq, (unsigned int)snap_count); |
35d489f9 AE |
3975 | out: |
3976 | kfree(reply_buf); | |
3977 | ||
57385b51 | 3978 | return ret; |
35d489f9 AE |
3979 | } |
3980 | ||
54cac61f AE |
3981 | static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, |
3982 | u64 snap_id) | |
b8b1e2db AE |
3983 | { |
3984 | size_t size; | |
3985 | void *reply_buf; | |
54cac61f | 3986 | __le64 snapid; |
b8b1e2db AE |
3987 | int ret; |
3988 | void *p; | |
3989 | void *end; | |
b8b1e2db AE |
3990 | char *snap_name; |
3991 | ||
3992 | size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; | |
3993 | reply_buf = kmalloc(size, GFP_KERNEL); | |
3994 | if (!reply_buf) | |
3995 | return ERR_PTR(-ENOMEM); | |
3996 | ||
54cac61f | 3997 | snapid = cpu_to_le64(snap_id); |
36be9a76 | 3998 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
b8b1e2db | 3999 | "rbd", "get_snapshot_name", |
54cac61f | 4000 | &snapid, sizeof (snapid), |
e2a58ee5 | 4001 | reply_buf, size); |
36be9a76 | 4002 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
f40eb349 AE |
4003 | if (ret < 0) { |
4004 | snap_name = ERR_PTR(ret); | |
b8b1e2db | 4005 | goto out; |
f40eb349 | 4006 | } |
b8b1e2db AE |
4007 | |
4008 | p = reply_buf; | |
f40eb349 | 4009 | end = reply_buf + ret; |
e5c35534 | 4010 | snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
f40eb349 | 4011 | if (IS_ERR(snap_name)) |
b8b1e2db | 4012 | goto out; |
b8b1e2db | 4013 | |
f40eb349 | 4014 | dout(" snap_id 0x%016llx snap_name = %s\n", |
54cac61f | 4015 | (unsigned long long)snap_id, snap_name); |
b8b1e2db AE |
4016 | out: |
4017 | kfree(reply_buf); | |
4018 | ||
f40eb349 | 4019 | return snap_name; |
b8b1e2db AE |
4020 | } |
4021 | ||
2df3fac7 | 4022 | static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev) |
117973fb | 4023 | { |
2df3fac7 | 4024 | bool first_time = rbd_dev->header.object_prefix == NULL; |
117973fb | 4025 | int ret; |
117973fb AE |
4026 | |
4027 | down_write(&rbd_dev->header_rwsem); | |
4028 | ||
2df3fac7 AE |
4029 | if (first_time) { |
4030 | ret = rbd_dev_v2_header_onetime(rbd_dev); | |
4031 | if (ret) | |
4032 | goto out; | |
4033 | } | |
4034 | ||
642a2537 AE |
4035 | /* |
4036 | * If the image supports layering, get the parent info. We | |
4037 | * need to probe the first time regardless. Thereafter we | |
4038 | * only need to if there's a parent, to see if it has | |
4039 | * disappeared due to the mapped image getting flattened. | |
4040 | */ | |
4041 | if (rbd_dev->header.features & RBD_FEATURE_LAYERING && | |
4042 | (first_time || rbd_dev->parent_spec)) { | |
4043 | bool warn; | |
4044 | ||
4045 | ret = rbd_dev_v2_parent_info(rbd_dev); | |
4046 | if (ret) | |
4047 | goto out; | |
4048 | ||
4049 | /* | |
4050 | * Print a warning if this is the initial probe and | |
4051 | * the image has a parent. Don't print it if the | |
4052 | * image now being probed is itself a parent. We | |
4053 | * can tell at this point because we won't know its | |
4054 | * pool name yet (just its pool id). | |
4055 | */ | |
4056 | warn = rbd_dev->parent_spec && rbd_dev->spec->pool_name; | |
4057 | if (first_time && warn) | |
4058 | rbd_warn(rbd_dev, "WARNING: kernel layering " | |
4059 | "is EXPERIMENTAL!"); | |
4060 | } | |
4061 | ||
117973fb AE |
4062 | ret = rbd_dev_v2_image_size(rbd_dev); |
4063 | if (ret) | |
4064 | goto out; | |
642a2537 | 4065 | |
29334ba4 AE |
4066 | if (rbd_dev->spec->snap_id == CEPH_NOSNAP) |
4067 | if (rbd_dev->mapping.size != rbd_dev->header.image_size) | |
4068 | rbd_dev->mapping.size = rbd_dev->header.image_size; | |
117973fb | 4069 | |
cc4a38bd | 4070 | ret = rbd_dev_v2_snap_context(rbd_dev); |
117973fb | 4071 | dout("rbd_dev_v2_snap_context returned %d\n", ret); |
117973fb AE |
4072 | out: |
4073 | up_write(&rbd_dev->header_rwsem); | |
4074 | ||
4075 | return ret; | |
4076 | } | |
4077 | ||
dfc5606d YS |
4078 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
4079 | { | |
dfc5606d | 4080 | struct device *dev; |
cd789ab9 | 4081 | int ret; |
dfc5606d YS |
4082 | |
4083 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
dfc5606d | 4084 | |
cd789ab9 | 4085 | dev = &rbd_dev->dev; |
dfc5606d YS |
4086 | dev->bus = &rbd_bus_type; |
4087 | dev->type = &rbd_device_type; | |
4088 | dev->parent = &rbd_root_dev; | |
200a6a8b | 4089 | dev->release = rbd_dev_device_release; |
de71a297 | 4090 | dev_set_name(dev, "%d", rbd_dev->dev_id); |
dfc5606d | 4091 | ret = device_register(dev); |
dfc5606d | 4092 | |
dfc5606d | 4093 | mutex_unlock(&ctl_mutex); |
cd789ab9 | 4094 | |
dfc5606d | 4095 | return ret; |
602adf40 YS |
4096 | } |
4097 | ||
dfc5606d YS |
4098 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
4099 | { | |
4100 | device_unregister(&rbd_dev->dev); | |
4101 | } | |
4102 | ||
e2839308 | 4103 | static atomic64_t rbd_dev_id_max = ATOMIC64_INIT(0); |
1ddbe94e AE |
4104 | |
4105 | /* | |
499afd5b AE |
4106 | * Get a unique rbd identifier for the given new rbd_dev, and add |
4107 | * the rbd_dev to the global list. The minimum rbd id is 1. | |
1ddbe94e | 4108 | */ |
e2839308 | 4109 | static void rbd_dev_id_get(struct rbd_device *rbd_dev) |
b7f23c36 | 4110 | { |
e2839308 | 4111 | rbd_dev->dev_id = atomic64_inc_return(&rbd_dev_id_max); |
499afd5b AE |
4112 | |
4113 | spin_lock(&rbd_dev_list_lock); | |
4114 | list_add_tail(&rbd_dev->node, &rbd_dev_list); | |
4115 | spin_unlock(&rbd_dev_list_lock); | |
e2839308 AE |
4116 | dout("rbd_dev %p given dev id %llu\n", rbd_dev, |
4117 | (unsigned long long) rbd_dev->dev_id); | |
1ddbe94e | 4118 | } |
b7f23c36 | 4119 | |
1ddbe94e | 4120 | /* |
499afd5b AE |
4121 | * Remove an rbd_dev from the global list, and record that its |
4122 | * identifier is no longer in use. | |
1ddbe94e | 4123 | */ |
e2839308 | 4124 | static void rbd_dev_id_put(struct rbd_device *rbd_dev) |
1ddbe94e | 4125 | { |
d184f6bf | 4126 | struct list_head *tmp; |
de71a297 | 4127 | int rbd_id = rbd_dev->dev_id; |
d184f6bf AE |
4128 | int max_id; |
4129 | ||
aafb230e | 4130 | rbd_assert(rbd_id > 0); |
499afd5b | 4131 | |
e2839308 AE |
4132 | dout("rbd_dev %p released dev id %llu\n", rbd_dev, |
4133 | (unsigned long long) rbd_dev->dev_id); | |
499afd5b AE |
4134 | spin_lock(&rbd_dev_list_lock); |
4135 | list_del_init(&rbd_dev->node); | |
d184f6bf AE |
4136 | |
4137 | /* | |
4138 | * If the id being "put" is not the current maximum, there | |
4139 | * is nothing special we need to do. | |
4140 | */ | |
e2839308 | 4141 | if (rbd_id != atomic64_read(&rbd_dev_id_max)) { |
d184f6bf AE |
4142 | spin_unlock(&rbd_dev_list_lock); |
4143 | return; | |
4144 | } | |
4145 | ||
4146 | /* | |
4147 | * We need to update the current maximum id. Search the | |
4148 | * list to find out what it is. We're more likely to find | |
4149 | * the maximum at the end, so search the list backward. | |
4150 | */ | |
4151 | max_id = 0; | |
4152 | list_for_each_prev(tmp, &rbd_dev_list) { | |
4153 | struct rbd_device *rbd_dev; | |
4154 | ||
4155 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
b213e0b1 AE |
4156 | if (rbd_dev->dev_id > max_id) |
4157 | max_id = rbd_dev->dev_id; | |
d184f6bf | 4158 | } |
499afd5b | 4159 | spin_unlock(&rbd_dev_list_lock); |
b7f23c36 | 4160 | |
1ddbe94e | 4161 | /* |
e2839308 | 4162 | * The max id could have been updated by rbd_dev_id_get(), in |
d184f6bf AE |
4163 | * which case it now accurately reflects the new maximum. |
4164 | * Be careful not to overwrite the maximum value in that | |
4165 | * case. | |
1ddbe94e | 4166 | */ |
e2839308 AE |
4167 | atomic64_cmpxchg(&rbd_dev_id_max, rbd_id, max_id); |
4168 | dout(" max dev id has been reset\n"); | |
b7f23c36 AE |
4169 | } |
4170 | ||
e28fff26 AE |
4171 | /* |
4172 | * Skips over white space at *buf, and updates *buf to point to the | |
4173 | * first found non-space character (if any). Returns the length of | |
593a9e7b AE |
4174 | * the token (string of non-white space characters) found. Note |
4175 | * that *buf must be terminated with '\0'. | |
e28fff26 AE |
4176 | */ |
4177 | static inline size_t next_token(const char **buf) | |
4178 | { | |
4179 | /* | |
4180 | * These are the characters that produce nonzero for | |
4181 | * isspace() in the "C" and "POSIX" locales. | |
4182 | */ | |
4183 | const char *spaces = " \f\n\r\t\v"; | |
4184 | ||
4185 | *buf += strspn(*buf, spaces); /* Find start of token */ | |
4186 | ||
4187 | return strcspn(*buf, spaces); /* Return token length */ | |
4188 | } | |
4189 | ||
4190 | /* | |
4191 | * Finds the next token in *buf, and if the provided token buffer is | |
4192 | * big enough, copies the found token into it. The result, if | |
593a9e7b AE |
4193 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
4194 | * must be terminated with '\0' on entry. | |
e28fff26 AE |
4195 | * |
4196 | * Returns the length of the token found (not including the '\0'). | |
4197 | * Return value will be 0 if no token is found, and it will be >= | |
4198 | * token_size if the token would not fit. | |
4199 | * | |
593a9e7b | 4200 | * The *buf pointer will be updated to point beyond the end of the |
e28fff26 AE |
4201 | * found token. Note that this occurs even if the token buffer is |
4202 | * too small to hold it. | |
4203 | */ | |
4204 | static inline size_t copy_token(const char **buf, | |
4205 | char *token, | |
4206 | size_t token_size) | |
4207 | { | |
4208 | size_t len; | |
4209 | ||
4210 | len = next_token(buf); | |
4211 | if (len < token_size) { | |
4212 | memcpy(token, *buf, len); | |
4213 | *(token + len) = '\0'; | |
4214 | } | |
4215 | *buf += len; | |
4216 | ||
4217 | return len; | |
4218 | } | |
4219 | ||
ea3352f4 AE |
4220 | /* |
4221 | * Finds the next token in *buf, dynamically allocates a buffer big | |
4222 | * enough to hold a copy of it, and copies the token into the new | |
4223 | * buffer. The copy is guaranteed to be terminated with '\0'. Note | |
4224 | * that a duplicate buffer is created even for a zero-length token. | |
4225 | * | |
4226 | * Returns a pointer to the newly-allocated duplicate, or a null | |
4227 | * pointer if memory for the duplicate was not available. If | |
4228 | * the lenp argument is a non-null pointer, the length of the token | |
4229 | * (not including the '\0') is returned in *lenp. | |
4230 | * | |
4231 | * If successful, the *buf pointer will be updated to point beyond | |
4232 | * the end of the found token. | |
4233 | * | |
4234 | * Note: uses GFP_KERNEL for allocation. | |
4235 | */ | |
4236 | static inline char *dup_token(const char **buf, size_t *lenp) | |
4237 | { | |
4238 | char *dup; | |
4239 | size_t len; | |
4240 | ||
4241 | len = next_token(buf); | |
4caf35f9 | 4242 | dup = kmemdup(*buf, len + 1, GFP_KERNEL); |
ea3352f4 AE |
4243 | if (!dup) |
4244 | return NULL; | |
ea3352f4 AE |
4245 | *(dup + len) = '\0'; |
4246 | *buf += len; | |
4247 | ||
4248 | if (lenp) | |
4249 | *lenp = len; | |
4250 | ||
4251 | return dup; | |
4252 | } | |
4253 | ||
a725f65e | 4254 | /* |
859c31df AE |
4255 | * Parse the options provided for an "rbd add" (i.e., rbd image |
4256 | * mapping) request. These arrive via a write to /sys/bus/rbd/add, | |
4257 | * and the data written is passed here via a NUL-terminated buffer. | |
4258 | * Returns 0 if successful or an error code otherwise. | |
d22f76e7 | 4259 | * |
859c31df AE |
4260 | * The information extracted from these options is recorded in |
4261 | * the other parameters which return dynamically-allocated | |
4262 | * structures: | |
4263 | * ceph_opts | |
4264 | * The address of a pointer that will refer to a ceph options | |
4265 | * structure. Caller must release the returned pointer using | |
4266 | * ceph_destroy_options() when it is no longer needed. | |
4267 | * rbd_opts | |
4268 | * Address of an rbd options pointer. Fully initialized by | |
4269 | * this function; caller must release with kfree(). | |
4270 | * spec | |
4271 | * Address of an rbd image specification pointer. Fully | |
4272 | * initialized by this function based on parsed options. | |
4273 | * Caller must release with rbd_spec_put(). | |
4274 | * | |
4275 | * The options passed take this form: | |
4276 | * <mon_addrs> <options> <pool_name> <image_name> [<snap_id>] | |
4277 | * where: | |
4278 | * <mon_addrs> | |
4279 | * A comma-separated list of one or more monitor addresses. | |
4280 | * A monitor address is an ip address, optionally followed | |
4281 | * by a port number (separated by a colon). | |
4282 | * I.e.: ip1[:port1][,ip2[:port2]...] | |
4283 | * <options> | |
4284 | * A comma-separated list of ceph and/or rbd options. | |
4285 | * <pool_name> | |
4286 | * The name of the rados pool containing the rbd image. | |
4287 | * <image_name> | |
4288 | * The name of the image in that pool to map. | |
4289 | * <snap_id> | |
4290 | * An optional snapshot id. If provided, the mapping will | |
4291 | * present data from the image at the time that snapshot was | |
4292 | * created. The image head is used if no snapshot id is | |
4293 | * provided. Snapshot mappings are always read-only. | |
a725f65e | 4294 | */ |
859c31df | 4295 | static int rbd_add_parse_args(const char *buf, |
dc79b113 | 4296 | struct ceph_options **ceph_opts, |
859c31df AE |
4297 | struct rbd_options **opts, |
4298 | struct rbd_spec **rbd_spec) | |
e28fff26 | 4299 | { |
d22f76e7 | 4300 | size_t len; |
859c31df | 4301 | char *options; |
0ddebc0c | 4302 | const char *mon_addrs; |
ecb4dc22 | 4303 | char *snap_name; |
0ddebc0c | 4304 | size_t mon_addrs_size; |
859c31df | 4305 | struct rbd_spec *spec = NULL; |
4e9afeba | 4306 | struct rbd_options *rbd_opts = NULL; |
859c31df | 4307 | struct ceph_options *copts; |
dc79b113 | 4308 | int ret; |
e28fff26 AE |
4309 | |
4310 | /* The first four tokens are required */ | |
4311 | ||
7ef3214a | 4312 | len = next_token(&buf); |
4fb5d671 AE |
4313 | if (!len) { |
4314 | rbd_warn(NULL, "no monitor address(es) provided"); | |
4315 | return -EINVAL; | |
4316 | } | |
0ddebc0c | 4317 | mon_addrs = buf; |
f28e565a | 4318 | mon_addrs_size = len + 1; |
7ef3214a | 4319 | buf += len; |
a725f65e | 4320 | |
dc79b113 | 4321 | ret = -EINVAL; |
f28e565a AE |
4322 | options = dup_token(&buf, NULL); |
4323 | if (!options) | |
dc79b113 | 4324 | return -ENOMEM; |
4fb5d671 AE |
4325 | if (!*options) { |
4326 | rbd_warn(NULL, "no options provided"); | |
4327 | goto out_err; | |
4328 | } | |
e28fff26 | 4329 | |
859c31df AE |
4330 | spec = rbd_spec_alloc(); |
4331 | if (!spec) | |
f28e565a | 4332 | goto out_mem; |
859c31df AE |
4333 | |
4334 | spec->pool_name = dup_token(&buf, NULL); | |
4335 | if (!spec->pool_name) | |
4336 | goto out_mem; | |
4fb5d671 AE |
4337 | if (!*spec->pool_name) { |
4338 | rbd_warn(NULL, "no pool name provided"); | |
4339 | goto out_err; | |
4340 | } | |
e28fff26 | 4341 | |
69e7a02f | 4342 | spec->image_name = dup_token(&buf, NULL); |
859c31df | 4343 | if (!spec->image_name) |
f28e565a | 4344 | goto out_mem; |
4fb5d671 AE |
4345 | if (!*spec->image_name) { |
4346 | rbd_warn(NULL, "no image name provided"); | |
4347 | goto out_err; | |
4348 | } | |
d4b125e9 | 4349 | |
f28e565a AE |
4350 | /* |
4351 | * Snapshot name is optional; default is to use "-" | |
4352 | * (indicating the head/no snapshot). | |
4353 | */ | |
3feeb894 | 4354 | len = next_token(&buf); |
820a5f3e | 4355 | if (!len) { |
3feeb894 AE |
4356 | buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ |
4357 | len = sizeof (RBD_SNAP_HEAD_NAME) - 1; | |
f28e565a | 4358 | } else if (len > RBD_MAX_SNAP_NAME_LEN) { |
dc79b113 | 4359 | ret = -ENAMETOOLONG; |
f28e565a | 4360 | goto out_err; |
849b4260 | 4361 | } |
ecb4dc22 AE |
4362 | snap_name = kmemdup(buf, len + 1, GFP_KERNEL); |
4363 | if (!snap_name) | |
f28e565a | 4364 | goto out_mem; |
ecb4dc22 AE |
4365 | *(snap_name + len) = '\0'; |
4366 | spec->snap_name = snap_name; | |
e5c35534 | 4367 | |
0ddebc0c | 4368 | /* Initialize all rbd options to the defaults */ |
e28fff26 | 4369 | |
4e9afeba AE |
4370 | rbd_opts = kzalloc(sizeof (*rbd_opts), GFP_KERNEL); |
4371 | if (!rbd_opts) | |
4372 | goto out_mem; | |
4373 | ||
4374 | rbd_opts->read_only = RBD_READ_ONLY_DEFAULT; | |
d22f76e7 | 4375 | |
859c31df | 4376 | copts = ceph_parse_options(options, mon_addrs, |
0ddebc0c | 4377 | mon_addrs + mon_addrs_size - 1, |
4e9afeba | 4378 | parse_rbd_opts_token, rbd_opts); |
859c31df AE |
4379 | if (IS_ERR(copts)) { |
4380 | ret = PTR_ERR(copts); | |
dc79b113 AE |
4381 | goto out_err; |
4382 | } | |
859c31df AE |
4383 | kfree(options); |
4384 | ||
4385 | *ceph_opts = copts; | |
4e9afeba | 4386 | *opts = rbd_opts; |
859c31df | 4387 | *rbd_spec = spec; |
0ddebc0c | 4388 | |
dc79b113 | 4389 | return 0; |
f28e565a | 4390 | out_mem: |
dc79b113 | 4391 | ret = -ENOMEM; |
d22f76e7 | 4392 | out_err: |
859c31df AE |
4393 | kfree(rbd_opts); |
4394 | rbd_spec_put(spec); | |
f28e565a | 4395 | kfree(options); |
d22f76e7 | 4396 | |
dc79b113 | 4397 | return ret; |
a725f65e AE |
4398 | } |
4399 | ||
589d30e0 AE |
4400 | /* |
4401 | * An rbd format 2 image has a unique identifier, distinct from the | |
4402 | * name given to it by the user. Internally, that identifier is | |
4403 | * what's used to specify the names of objects related to the image. | |
4404 | * | |
4405 | * A special "rbd id" object is used to map an rbd image name to its | |
4406 | * id. If that object doesn't exist, then there is no v2 rbd image | |
4407 | * with the supplied name. | |
4408 | * | |
4409 | * This function will record the given rbd_dev's image_id field if | |
4410 | * it can be determined, and in that case will return 0. If any | |
4411 | * errors occur a negative errno will be returned and the rbd_dev's | |
4412 | * image_id field will be unchanged (and should be NULL). | |
4413 | */ | |
4414 | static int rbd_dev_image_id(struct rbd_device *rbd_dev) | |
4415 | { | |
4416 | int ret; | |
4417 | size_t size; | |
4418 | char *object_name; | |
4419 | void *response; | |
c0fba368 | 4420 | char *image_id; |
2f82ee54 | 4421 | |
2c0d0a10 AE |
4422 | /* |
4423 | * When probing a parent image, the image id is already | |
4424 | * known (and the image name likely is not). There's no | |
c0fba368 AE |
4425 | * need to fetch the image id again in this case. We |
4426 | * do still need to set the image format though. | |
2c0d0a10 | 4427 | */ |
c0fba368 AE |
4428 | if (rbd_dev->spec->image_id) { |
4429 | rbd_dev->image_format = *rbd_dev->spec->image_id ? 2 : 1; | |
4430 | ||
2c0d0a10 | 4431 | return 0; |
c0fba368 | 4432 | } |
2c0d0a10 | 4433 | |
589d30e0 AE |
4434 | /* |
4435 | * First, see if the format 2 image id file exists, and if | |
4436 | * so, get the image's persistent id from it. | |
4437 | */ | |
69e7a02f | 4438 | size = sizeof (RBD_ID_PREFIX) + strlen(rbd_dev->spec->image_name); |
589d30e0 AE |
4439 | object_name = kmalloc(size, GFP_NOIO); |
4440 | if (!object_name) | |
4441 | return -ENOMEM; | |
0d7dbfce | 4442 | sprintf(object_name, "%s%s", RBD_ID_PREFIX, rbd_dev->spec->image_name); |
589d30e0 AE |
4443 | dout("rbd id object name is %s\n", object_name); |
4444 | ||
4445 | /* Response will be an encoded string, which includes a length */ | |
4446 | ||
4447 | size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; | |
4448 | response = kzalloc(size, GFP_NOIO); | |
4449 | if (!response) { | |
4450 | ret = -ENOMEM; | |
4451 | goto out; | |
4452 | } | |
4453 | ||
c0fba368 AE |
4454 | /* If it doesn't exist we'll assume it's a format 1 image */ |
4455 | ||
36be9a76 | 4456 | ret = rbd_obj_method_sync(rbd_dev, object_name, |
4157976b | 4457 | "rbd", "get_id", NULL, 0, |
e2a58ee5 | 4458 | response, RBD_IMAGE_ID_LEN_MAX); |
36be9a76 | 4459 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
c0fba368 AE |
4460 | if (ret == -ENOENT) { |
4461 | image_id = kstrdup("", GFP_KERNEL); | |
4462 | ret = image_id ? 0 : -ENOMEM; | |
4463 | if (!ret) | |
4464 | rbd_dev->image_format = 1; | |
4465 | } else if (ret > sizeof (__le32)) { | |
4466 | void *p = response; | |
4467 | ||
4468 | image_id = ceph_extract_encoded_string(&p, p + ret, | |
979ed480 | 4469 | NULL, GFP_NOIO); |
c0fba368 AE |
4470 | ret = IS_ERR(image_id) ? PTR_ERR(image_id) : 0; |
4471 | if (!ret) | |
4472 | rbd_dev->image_format = 2; | |
589d30e0 | 4473 | } else { |
c0fba368 AE |
4474 | ret = -EINVAL; |
4475 | } | |
4476 | ||
4477 | if (!ret) { | |
4478 | rbd_dev->spec->image_id = image_id; | |
4479 | dout("image_id is %s\n", image_id); | |
589d30e0 AE |
4480 | } |
4481 | out: | |
4482 | kfree(response); | |
4483 | kfree(object_name); | |
4484 | ||
4485 | return ret; | |
4486 | } | |
4487 | ||
6fd48b3b AE |
4488 | /* Undo whatever state changes are made by v1 or v2 image probe */ |
4489 | ||
4490 | static void rbd_dev_unprobe(struct rbd_device *rbd_dev) | |
4491 | { | |
4492 | struct rbd_image_header *header; | |
4493 | ||
4494 | rbd_dev_remove_parent(rbd_dev); | |
4495 | rbd_spec_put(rbd_dev->parent_spec); | |
4496 | rbd_dev->parent_spec = NULL; | |
4497 | rbd_dev->parent_overlap = 0; | |
4498 | ||
4499 | /* Free dynamic fields from the header, then zero it out */ | |
4500 | ||
4501 | header = &rbd_dev->header; | |
812164f8 | 4502 | ceph_put_snap_context(header->snapc); |
6fd48b3b AE |
4503 | kfree(header->snap_sizes); |
4504 | kfree(header->snap_names); | |
4505 | kfree(header->object_prefix); | |
4506 | memset(header, 0, sizeof (*header)); | |
4507 | } | |
4508 | ||
2df3fac7 | 4509 | static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev) |
a30b71b9 | 4510 | { |
9d475de5 | 4511 | int ret; |
a30b71b9 | 4512 | |
1e130199 | 4513 | ret = rbd_dev_v2_object_prefix(rbd_dev); |
57385b51 | 4514 | if (ret) |
b1b5402a AE |
4515 | goto out_err; |
4516 | ||
2df3fac7 AE |
4517 | /* |
4518 | * Get the and check features for the image. Currently the | |
4519 | * features are assumed to never change. | |
4520 | */ | |
b1b5402a | 4521 | ret = rbd_dev_v2_features(rbd_dev); |
57385b51 | 4522 | if (ret) |
9d475de5 | 4523 | goto out_err; |
35d489f9 | 4524 | |
cc070d59 AE |
4525 | /* If the image supports fancy striping, get its parameters */ |
4526 | ||
4527 | if (rbd_dev->header.features & RBD_FEATURE_STRIPINGV2) { | |
4528 | ret = rbd_dev_v2_striping_info(rbd_dev); | |
4529 | if (ret < 0) | |
4530 | goto out_err; | |
4531 | } | |
2df3fac7 | 4532 | /* No support for crypto and compression type format 2 images */ |
6e14b1a6 | 4533 | |
35152979 | 4534 | return 0; |
9d475de5 | 4535 | out_err: |
642a2537 | 4536 | rbd_dev->header.features = 0; |
1e130199 AE |
4537 | kfree(rbd_dev->header.object_prefix); |
4538 | rbd_dev->header.object_prefix = NULL; | |
9d475de5 AE |
4539 | |
4540 | return ret; | |
a30b71b9 AE |
4541 | } |
4542 | ||
124afba2 | 4543 | static int rbd_dev_probe_parent(struct rbd_device *rbd_dev) |
83a06263 | 4544 | { |
2f82ee54 | 4545 | struct rbd_device *parent = NULL; |
124afba2 AE |
4546 | struct rbd_spec *parent_spec; |
4547 | struct rbd_client *rbdc; | |
4548 | int ret; | |
4549 | ||
4550 | if (!rbd_dev->parent_spec) | |
4551 | return 0; | |
4552 | /* | |
4553 | * We need to pass a reference to the client and the parent | |
4554 | * spec when creating the parent rbd_dev. Images related by | |
4555 | * parent/child relationships always share both. | |
4556 | */ | |
4557 | parent_spec = rbd_spec_get(rbd_dev->parent_spec); | |
4558 | rbdc = __rbd_get_client(rbd_dev->rbd_client); | |
4559 | ||
4560 | ret = -ENOMEM; | |
4561 | parent = rbd_dev_create(rbdc, parent_spec); | |
4562 | if (!parent) | |
4563 | goto out_err; | |
4564 | ||
1f3ef788 | 4565 | ret = rbd_dev_image_probe(parent, false); |
124afba2 AE |
4566 | if (ret < 0) |
4567 | goto out_err; | |
4568 | rbd_dev->parent = parent; | |
4569 | ||
4570 | return 0; | |
4571 | out_err: | |
4572 | if (parent) { | |
4573 | rbd_spec_put(rbd_dev->parent_spec); | |
4574 | kfree(rbd_dev->header_name); | |
4575 | rbd_dev_destroy(parent); | |
4576 | } else { | |
4577 | rbd_put_client(rbdc); | |
4578 | rbd_spec_put(parent_spec); | |
4579 | } | |
4580 | ||
4581 | return ret; | |
4582 | } | |
4583 | ||
200a6a8b | 4584 | static int rbd_dev_device_setup(struct rbd_device *rbd_dev) |
124afba2 | 4585 | { |
83a06263 | 4586 | int ret; |
d1cf5788 | 4587 | |
83a06263 AE |
4588 | /* generate unique id: find highest unique id, add one */ |
4589 | rbd_dev_id_get(rbd_dev); | |
4590 | ||
4591 | /* Fill in the device name, now that we have its id. */ | |
4592 | BUILD_BUG_ON(DEV_NAME_LEN | |
4593 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); | |
4594 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); | |
4595 | ||
4596 | /* Get our block major device number. */ | |
4597 | ||
4598 | ret = register_blkdev(0, rbd_dev->name); | |
4599 | if (ret < 0) | |
4600 | goto err_out_id; | |
4601 | rbd_dev->major = ret; | |
4602 | ||
4603 | /* Set up the blkdev mapping. */ | |
4604 | ||
4605 | ret = rbd_init_disk(rbd_dev); | |
4606 | if (ret) | |
4607 | goto err_out_blkdev; | |
4608 | ||
f35a4dee | 4609 | ret = rbd_dev_mapping_set(rbd_dev); |
83a06263 AE |
4610 | if (ret) |
4611 | goto err_out_disk; | |
f35a4dee AE |
4612 | set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); |
4613 | ||
4614 | ret = rbd_bus_add_dev(rbd_dev); | |
4615 | if (ret) | |
4616 | goto err_out_mapping; | |
83a06263 | 4617 | |
83a06263 AE |
4618 | /* Everything's ready. Announce the disk to the world. */ |
4619 | ||
129b79d4 | 4620 | set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
83a06263 AE |
4621 | add_disk(rbd_dev->disk); |
4622 | ||
4623 | pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name, | |
4624 | (unsigned long long) rbd_dev->mapping.size); | |
4625 | ||
4626 | return ret; | |
2f82ee54 | 4627 | |
f35a4dee AE |
4628 | err_out_mapping: |
4629 | rbd_dev_mapping_clear(rbd_dev); | |
83a06263 AE |
4630 | err_out_disk: |
4631 | rbd_free_disk(rbd_dev); | |
4632 | err_out_blkdev: | |
4633 | unregister_blkdev(rbd_dev->major, rbd_dev->name); | |
4634 | err_out_id: | |
4635 | rbd_dev_id_put(rbd_dev); | |
d1cf5788 | 4636 | rbd_dev_mapping_clear(rbd_dev); |
83a06263 AE |
4637 | |
4638 | return ret; | |
4639 | } | |
4640 | ||
332bb12d AE |
4641 | static int rbd_dev_header_name(struct rbd_device *rbd_dev) |
4642 | { | |
4643 | struct rbd_spec *spec = rbd_dev->spec; | |
4644 | size_t size; | |
4645 | ||
4646 | /* Record the header object name for this rbd image. */ | |
4647 | ||
4648 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); | |
4649 | ||
4650 | if (rbd_dev->image_format == 1) | |
4651 | size = strlen(spec->image_name) + sizeof (RBD_SUFFIX); | |
4652 | else | |
4653 | size = sizeof (RBD_HEADER_PREFIX) + strlen(spec->image_id); | |
4654 | ||
4655 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); | |
4656 | if (!rbd_dev->header_name) | |
4657 | return -ENOMEM; | |
4658 | ||
4659 | if (rbd_dev->image_format == 1) | |
4660 | sprintf(rbd_dev->header_name, "%s%s", | |
4661 | spec->image_name, RBD_SUFFIX); | |
4662 | else | |
4663 | sprintf(rbd_dev->header_name, "%s%s", | |
4664 | RBD_HEADER_PREFIX, spec->image_id); | |
4665 | return 0; | |
4666 | } | |
4667 | ||
200a6a8b AE |
4668 | static void rbd_dev_image_release(struct rbd_device *rbd_dev) |
4669 | { | |
6fd48b3b | 4670 | rbd_dev_unprobe(rbd_dev); |
200a6a8b | 4671 | kfree(rbd_dev->header_name); |
6fd48b3b AE |
4672 | rbd_dev->header_name = NULL; |
4673 | rbd_dev->image_format = 0; | |
4674 | kfree(rbd_dev->spec->image_id); | |
4675 | rbd_dev->spec->image_id = NULL; | |
4676 | ||
200a6a8b AE |
4677 | rbd_dev_destroy(rbd_dev); |
4678 | } | |
4679 | ||
a30b71b9 AE |
4680 | /* |
4681 | * Probe for the existence of the header object for the given rbd | |
1f3ef788 AE |
4682 | * device. If this image is the one being mapped (i.e., not a |
4683 | * parent), initiate a watch on its header object before using that | |
4684 | * object to get detailed information about the rbd image. | |
a30b71b9 | 4685 | */ |
1f3ef788 | 4686 | static int rbd_dev_image_probe(struct rbd_device *rbd_dev, bool mapping) |
a30b71b9 AE |
4687 | { |
4688 | int ret; | |
b644de2b | 4689 | int tmp; |
a30b71b9 AE |
4690 | |
4691 | /* | |
4692 | * Get the id from the image id object. If it's not a | |
4693 | * format 2 image, we'll get ENOENT back, and we'll assume | |
4694 | * it's a format 1 image. | |
4695 | */ | |
4696 | ret = rbd_dev_image_id(rbd_dev); | |
4697 | if (ret) | |
c0fba368 AE |
4698 | return ret; |
4699 | rbd_assert(rbd_dev->spec->image_id); | |
4700 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); | |
4701 | ||
332bb12d AE |
4702 | ret = rbd_dev_header_name(rbd_dev); |
4703 | if (ret) | |
4704 | goto err_out_format; | |
4705 | ||
1f3ef788 AE |
4706 | if (mapping) { |
4707 | ret = rbd_dev_header_watch_sync(rbd_dev, true); | |
4708 | if (ret) | |
4709 | goto out_header_name; | |
4710 | } | |
b644de2b | 4711 | |
c0fba368 | 4712 | if (rbd_dev->image_format == 1) |
99a41ebc | 4713 | ret = rbd_dev_v1_header_info(rbd_dev); |
a30b71b9 | 4714 | else |
2df3fac7 | 4715 | ret = rbd_dev_v2_header_info(rbd_dev); |
5655c4d9 | 4716 | if (ret) |
b644de2b | 4717 | goto err_out_watch; |
83a06263 | 4718 | |
9bb81c9b AE |
4719 | ret = rbd_dev_spec_update(rbd_dev); |
4720 | if (ret) | |
33dca39f | 4721 | goto err_out_probe; |
9bb81c9b AE |
4722 | |
4723 | ret = rbd_dev_probe_parent(rbd_dev); | |
30d60ba2 AE |
4724 | if (ret) |
4725 | goto err_out_probe; | |
4726 | ||
4727 | dout("discovered format %u image, header name is %s\n", | |
4728 | rbd_dev->image_format, rbd_dev->header_name); | |
83a06263 | 4729 | |
30d60ba2 | 4730 | return 0; |
6fd48b3b AE |
4731 | err_out_probe: |
4732 | rbd_dev_unprobe(rbd_dev); | |
b644de2b | 4733 | err_out_watch: |
1f3ef788 AE |
4734 | if (mapping) { |
4735 | tmp = rbd_dev_header_watch_sync(rbd_dev, false); | |
4736 | if (tmp) | |
4737 | rbd_warn(rbd_dev, "unable to tear down " | |
4738 | "watch request (%d)\n", tmp); | |
4739 | } | |
332bb12d AE |
4740 | out_header_name: |
4741 | kfree(rbd_dev->header_name); | |
4742 | rbd_dev->header_name = NULL; | |
4743 | err_out_format: | |
4744 | rbd_dev->image_format = 0; | |
5655c4d9 AE |
4745 | kfree(rbd_dev->spec->image_id); |
4746 | rbd_dev->spec->image_id = NULL; | |
4747 | ||
4748 | dout("probe failed, returning %d\n", ret); | |
4749 | ||
a30b71b9 AE |
4750 | return ret; |
4751 | } | |
4752 | ||
59c2be1e YS |
4753 | static ssize_t rbd_add(struct bus_type *bus, |
4754 | const char *buf, | |
4755 | size_t count) | |
602adf40 | 4756 | { |
cb8627c7 | 4757 | struct rbd_device *rbd_dev = NULL; |
dc79b113 | 4758 | struct ceph_options *ceph_opts = NULL; |
4e9afeba | 4759 | struct rbd_options *rbd_opts = NULL; |
859c31df | 4760 | struct rbd_spec *spec = NULL; |
9d3997fd | 4761 | struct rbd_client *rbdc; |
27cc2594 | 4762 | struct ceph_osd_client *osdc; |
51344a38 | 4763 | bool read_only; |
27cc2594 | 4764 | int rc = -ENOMEM; |
602adf40 YS |
4765 | |
4766 | if (!try_module_get(THIS_MODULE)) | |
4767 | return -ENODEV; | |
4768 | ||
602adf40 | 4769 | /* parse add command */ |
859c31df | 4770 | rc = rbd_add_parse_args(buf, &ceph_opts, &rbd_opts, &spec); |
dc79b113 | 4771 | if (rc < 0) |
bd4ba655 | 4772 | goto err_out_module; |
51344a38 AE |
4773 | read_only = rbd_opts->read_only; |
4774 | kfree(rbd_opts); | |
4775 | rbd_opts = NULL; /* done with this */ | |
78cea76e | 4776 | |
9d3997fd AE |
4777 | rbdc = rbd_get_client(ceph_opts); |
4778 | if (IS_ERR(rbdc)) { | |
4779 | rc = PTR_ERR(rbdc); | |
0ddebc0c | 4780 | goto err_out_args; |
9d3997fd | 4781 | } |
c53d5893 | 4782 | ceph_opts = NULL; /* rbd_dev client now owns this */ |
602adf40 | 4783 | |
602adf40 | 4784 | /* pick the pool */ |
9d3997fd | 4785 | osdc = &rbdc->client->osdc; |
859c31df | 4786 | rc = ceph_pg_poolid_by_name(osdc->osdmap, spec->pool_name); |
602adf40 YS |
4787 | if (rc < 0) |
4788 | goto err_out_client; | |
c0cd10db | 4789 | spec->pool_id = (u64)rc; |
859c31df | 4790 | |
0903e875 AE |
4791 | /* The ceph file layout needs to fit pool id in 32 bits */ |
4792 | ||
c0cd10db AE |
4793 | if (spec->pool_id > (u64)U32_MAX) { |
4794 | rbd_warn(NULL, "pool id too large (%llu > %u)\n", | |
4795 | (unsigned long long)spec->pool_id, U32_MAX); | |
0903e875 AE |
4796 | rc = -EIO; |
4797 | goto err_out_client; | |
4798 | } | |
4799 | ||
c53d5893 | 4800 | rbd_dev = rbd_dev_create(rbdc, spec); |
bd4ba655 AE |
4801 | if (!rbd_dev) |
4802 | goto err_out_client; | |
c53d5893 AE |
4803 | rbdc = NULL; /* rbd_dev now owns this */ |
4804 | spec = NULL; /* rbd_dev now owns this */ | |
602adf40 | 4805 | |
1f3ef788 | 4806 | rc = rbd_dev_image_probe(rbd_dev, true); |
a30b71b9 | 4807 | if (rc < 0) |
c53d5893 | 4808 | goto err_out_rbd_dev; |
05fd6f6f | 4809 | |
7ce4eef7 AE |
4810 | /* If we are mapping a snapshot it must be marked read-only */ |
4811 | ||
4812 | if (rbd_dev->spec->snap_id != CEPH_NOSNAP) | |
4813 | read_only = true; | |
4814 | rbd_dev->mapping.read_only = read_only; | |
4815 | ||
b536f69a AE |
4816 | rc = rbd_dev_device_setup(rbd_dev); |
4817 | if (!rc) | |
4818 | return count; | |
4819 | ||
4820 | rbd_dev_image_release(rbd_dev); | |
c53d5893 AE |
4821 | err_out_rbd_dev: |
4822 | rbd_dev_destroy(rbd_dev); | |
bd4ba655 | 4823 | err_out_client: |
9d3997fd | 4824 | rbd_put_client(rbdc); |
0ddebc0c | 4825 | err_out_args: |
78cea76e AE |
4826 | if (ceph_opts) |
4827 | ceph_destroy_options(ceph_opts); | |
4e9afeba | 4828 | kfree(rbd_opts); |
859c31df | 4829 | rbd_spec_put(spec); |
bd4ba655 AE |
4830 | err_out_module: |
4831 | module_put(THIS_MODULE); | |
27cc2594 | 4832 | |
602adf40 | 4833 | dout("Error adding device %s\n", buf); |
27cc2594 | 4834 | |
c0cd10db | 4835 | return (ssize_t)rc; |
602adf40 YS |
4836 | } |
4837 | ||
de71a297 | 4838 | static struct rbd_device *__rbd_get_dev(unsigned long dev_id) |
602adf40 YS |
4839 | { |
4840 | struct list_head *tmp; | |
4841 | struct rbd_device *rbd_dev; | |
4842 | ||
e124a82f | 4843 | spin_lock(&rbd_dev_list_lock); |
602adf40 YS |
4844 | list_for_each(tmp, &rbd_dev_list) { |
4845 | rbd_dev = list_entry(tmp, struct rbd_device, node); | |
de71a297 | 4846 | if (rbd_dev->dev_id == dev_id) { |
e124a82f | 4847 | spin_unlock(&rbd_dev_list_lock); |
602adf40 | 4848 | return rbd_dev; |
e124a82f | 4849 | } |
602adf40 | 4850 | } |
e124a82f | 4851 | spin_unlock(&rbd_dev_list_lock); |
602adf40 YS |
4852 | return NULL; |
4853 | } | |
4854 | ||
200a6a8b | 4855 | static void rbd_dev_device_release(struct device *dev) |
602adf40 | 4856 | { |
593a9e7b | 4857 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
602adf40 | 4858 | |
602adf40 | 4859 | rbd_free_disk(rbd_dev); |
200a6a8b | 4860 | clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
6d80b130 | 4861 | rbd_dev_mapping_clear(rbd_dev); |
602adf40 | 4862 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
200a6a8b | 4863 | rbd_dev->major = 0; |
e2839308 | 4864 | rbd_dev_id_put(rbd_dev); |
d1cf5788 | 4865 | rbd_dev_mapping_clear(rbd_dev); |
602adf40 YS |
4866 | } |
4867 | ||
05a46afd AE |
4868 | static void rbd_dev_remove_parent(struct rbd_device *rbd_dev) |
4869 | { | |
ad945fc1 | 4870 | while (rbd_dev->parent) { |
05a46afd AE |
4871 | struct rbd_device *first = rbd_dev; |
4872 | struct rbd_device *second = first->parent; | |
4873 | struct rbd_device *third; | |
4874 | ||
4875 | /* | |
4876 | * Follow to the parent with no grandparent and | |
4877 | * remove it. | |
4878 | */ | |
4879 | while (second && (third = second->parent)) { | |
4880 | first = second; | |
4881 | second = third; | |
4882 | } | |
ad945fc1 | 4883 | rbd_assert(second); |
8ad42cd0 | 4884 | rbd_dev_image_release(second); |
ad945fc1 AE |
4885 | first->parent = NULL; |
4886 | first->parent_overlap = 0; | |
4887 | ||
4888 | rbd_assert(first->parent_spec); | |
05a46afd AE |
4889 | rbd_spec_put(first->parent_spec); |
4890 | first->parent_spec = NULL; | |
05a46afd AE |
4891 | } |
4892 | } | |
4893 | ||
dfc5606d YS |
4894 | static ssize_t rbd_remove(struct bus_type *bus, |
4895 | const char *buf, | |
4896 | size_t count) | |
602adf40 YS |
4897 | { |
4898 | struct rbd_device *rbd_dev = NULL; | |
0d8189e1 | 4899 | int target_id; |
602adf40 | 4900 | unsigned long ul; |
0d8189e1 | 4901 | int ret; |
602adf40 | 4902 | |
0d8189e1 AE |
4903 | ret = strict_strtoul(buf, 10, &ul); |
4904 | if (ret) | |
4905 | return ret; | |
602adf40 YS |
4906 | |
4907 | /* convert to int; abort if we lost anything in the conversion */ | |
4908 | target_id = (int) ul; | |
4909 | if (target_id != ul) | |
4910 | return -EINVAL; | |
4911 | ||
4912 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); | |
4913 | ||
4914 | rbd_dev = __rbd_get_dev(target_id); | |
4915 | if (!rbd_dev) { | |
4916 | ret = -ENOENT; | |
4917 | goto done; | |
42382b70 AE |
4918 | } |
4919 | ||
a14ea269 | 4920 | spin_lock_irq(&rbd_dev->lock); |
b82d167b | 4921 | if (rbd_dev->open_count) |
42382b70 | 4922 | ret = -EBUSY; |
b82d167b AE |
4923 | else |
4924 | set_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags); | |
a14ea269 | 4925 | spin_unlock_irq(&rbd_dev->lock); |
b82d167b | 4926 | if (ret < 0) |
42382b70 | 4927 | goto done; |
b480815a | 4928 | rbd_bus_del_dev(rbd_dev); |
1f3ef788 AE |
4929 | ret = rbd_dev_header_watch_sync(rbd_dev, false); |
4930 | if (ret) | |
4931 | rbd_warn(rbd_dev, "failed to cancel watch event (%d)\n", ret); | |
8ad42cd0 | 4932 | rbd_dev_image_release(rbd_dev); |
79ab7558 | 4933 | module_put(THIS_MODULE); |
1f3ef788 | 4934 | ret = count; |
602adf40 YS |
4935 | done: |
4936 | mutex_unlock(&ctl_mutex); | |
aafb230e | 4937 | |
602adf40 YS |
4938 | return ret; |
4939 | } | |
4940 | ||
602adf40 YS |
4941 | /* |
4942 | * create control files in sysfs | |
dfc5606d | 4943 | * /sys/bus/rbd/... |
602adf40 YS |
4944 | */ |
4945 | static int rbd_sysfs_init(void) | |
4946 | { | |
dfc5606d | 4947 | int ret; |
602adf40 | 4948 | |
fed4c143 | 4949 | ret = device_register(&rbd_root_dev); |
21079786 | 4950 | if (ret < 0) |
dfc5606d | 4951 | return ret; |
602adf40 | 4952 | |
fed4c143 AE |
4953 | ret = bus_register(&rbd_bus_type); |
4954 | if (ret < 0) | |
4955 | device_unregister(&rbd_root_dev); | |
602adf40 | 4956 | |
602adf40 YS |
4957 | return ret; |
4958 | } | |
4959 | ||
4960 | static void rbd_sysfs_cleanup(void) | |
4961 | { | |
dfc5606d | 4962 | bus_unregister(&rbd_bus_type); |
fed4c143 | 4963 | device_unregister(&rbd_root_dev); |
602adf40 YS |
4964 | } |
4965 | ||
1c2a9dfe AE |
4966 | static int rbd_slab_init(void) |
4967 | { | |
4968 | rbd_assert(!rbd_img_request_cache); | |
4969 | rbd_img_request_cache = kmem_cache_create("rbd_img_request", | |
4970 | sizeof (struct rbd_img_request), | |
4971 | __alignof__(struct rbd_img_request), | |
4972 | 0, NULL); | |
868311b1 AE |
4973 | if (!rbd_img_request_cache) |
4974 | return -ENOMEM; | |
4975 | ||
4976 | rbd_assert(!rbd_obj_request_cache); | |
4977 | rbd_obj_request_cache = kmem_cache_create("rbd_obj_request", | |
4978 | sizeof (struct rbd_obj_request), | |
4979 | __alignof__(struct rbd_obj_request), | |
4980 | 0, NULL); | |
78c2a44a AE |
4981 | if (!rbd_obj_request_cache) |
4982 | goto out_err; | |
4983 | ||
4984 | rbd_assert(!rbd_segment_name_cache); | |
4985 | rbd_segment_name_cache = kmem_cache_create("rbd_segment_name", | |
4986 | MAX_OBJ_NAME_SIZE + 1, 1, 0, NULL); | |
4987 | if (rbd_segment_name_cache) | |
1c2a9dfe | 4988 | return 0; |
78c2a44a AE |
4989 | out_err: |
4990 | if (rbd_obj_request_cache) { | |
4991 | kmem_cache_destroy(rbd_obj_request_cache); | |
4992 | rbd_obj_request_cache = NULL; | |
4993 | } | |
1c2a9dfe | 4994 | |
868311b1 AE |
4995 | kmem_cache_destroy(rbd_img_request_cache); |
4996 | rbd_img_request_cache = NULL; | |
4997 | ||
1c2a9dfe AE |
4998 | return -ENOMEM; |
4999 | } | |
5000 | ||
5001 | static void rbd_slab_exit(void) | |
5002 | { | |
78c2a44a AE |
5003 | rbd_assert(rbd_segment_name_cache); |
5004 | kmem_cache_destroy(rbd_segment_name_cache); | |
5005 | rbd_segment_name_cache = NULL; | |
5006 | ||
868311b1 AE |
5007 | rbd_assert(rbd_obj_request_cache); |
5008 | kmem_cache_destroy(rbd_obj_request_cache); | |
5009 | rbd_obj_request_cache = NULL; | |
5010 | ||
1c2a9dfe AE |
5011 | rbd_assert(rbd_img_request_cache); |
5012 | kmem_cache_destroy(rbd_img_request_cache); | |
5013 | rbd_img_request_cache = NULL; | |
5014 | } | |
5015 | ||
cc344fa1 | 5016 | static int __init rbd_init(void) |
602adf40 YS |
5017 | { |
5018 | int rc; | |
5019 | ||
1e32d34c AE |
5020 | if (!libceph_compatible(NULL)) { |
5021 | rbd_warn(NULL, "libceph incompatibility (quitting)"); | |
5022 | ||
5023 | return -EINVAL; | |
5024 | } | |
1c2a9dfe | 5025 | rc = rbd_slab_init(); |
602adf40 YS |
5026 | if (rc) |
5027 | return rc; | |
1c2a9dfe AE |
5028 | rc = rbd_sysfs_init(); |
5029 | if (rc) | |
5030 | rbd_slab_exit(); | |
5031 | else | |
5032 | pr_info("loaded " RBD_DRV_NAME_LONG "\n"); | |
5033 | ||
5034 | return rc; | |
602adf40 YS |
5035 | } |
5036 | ||
cc344fa1 | 5037 | static void __exit rbd_exit(void) |
602adf40 YS |
5038 | { |
5039 | rbd_sysfs_cleanup(); | |
1c2a9dfe | 5040 | rbd_slab_exit(); |
602adf40 YS |
5041 | } |
5042 | ||
5043 | module_init(rbd_init); | |
5044 | module_exit(rbd_exit); | |
5045 | ||
5046 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
5047 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
5048 | MODULE_DESCRIPTION("rados block device"); | |
5049 | ||
5050 | /* following authorship retained from original osdblk.c */ | |
5051 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); | |
5052 | ||
5053 | MODULE_LICENSE("GPL"); |