Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1 | /* |
| 2 | rbd.c -- Export ceph rados objects as a Linux block device |
| 3 | |
| 4 | |
| 5 | based on drivers/block/osdblk.c: |
| 6 | |
| 7 | Copyright 2009 Red Hat, Inc. |
| 8 | |
| 9 | This program is free software; you can redistribute it and/or modify |
| 10 | it under the terms of the GNU General Public License as published by |
| 11 | the Free Software Foundation. |
| 12 | |
| 13 | This program is distributed in the hope that it will be useful, |
| 14 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | GNU General Public License for more details. |
| 17 | |
| 18 | You should have received a copy of the GNU General Public License |
| 19 | along with this program; see the file COPYING. If not, write to |
| 20 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. |
| 21 | |
| 22 | |
| 23 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 24 | For usage instructions, please refer to: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 25 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 26 | Documentation/ABI/testing/sysfs-bus-rbd |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 27 | |
| 28 | */ |
| 29 | |
| 30 | #include <linux/ceph/libceph.h> |
| 31 | #include <linux/ceph/osd_client.h> |
| 32 | #include <linux/ceph/mon_client.h> |
| 33 | #include <linux/ceph/decode.h> |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 34 | #include <linux/parser.h> |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 35 | |
| 36 | #include <linux/kernel.h> |
| 37 | #include <linux/device.h> |
| 38 | #include <linux/module.h> |
| 39 | #include <linux/fs.h> |
| 40 | #include <linux/blkdev.h> |
| 41 | |
| 42 | #include "rbd_types.h" |
| 43 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 44 | #define RBD_DEBUG /* Activate rbd_assert() calls */ |
| 45 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 46 | /* |
| 47 | * The basic unit of block I/O is a sector. It is interpreted in a |
| 48 | * number of contexts in Linux (blk, bio, genhd), but the default is |
| 49 | * universally 512 bytes. These symbols are just slightly more |
| 50 | * meaningful than the bare numbers they represent. |
| 51 | */ |
| 52 | #define SECTOR_SHIFT 9 |
| 53 | #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) |
| 54 | |
Alex Elder | 2647ba3 | 2012-11-19 22:55:21 -0600 | [diff] [blame] | 55 | /* It might be useful to have these defined elsewhere */ |
Alex Elder | df111be | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 56 | |
Alex Elder | 2647ba3 | 2012-11-19 22:55:21 -0600 | [diff] [blame] | 57 | #define U8_MAX ((u8) (~0U)) |
| 58 | #define U16_MAX ((u16) (~0U)) |
| 59 | #define U32_MAX ((u32) (~0U)) |
| 60 | #define U64_MAX ((u64) (~0ULL)) |
Alex Elder | df111be | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 61 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 62 | #define RBD_DRV_NAME "rbd" |
| 63 | #define RBD_DRV_NAME_LONG "rbd (rados block device)" |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 64 | |
| 65 | #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ |
| 66 | |
Alex Elder | d4b125e | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 67 | #define RBD_SNAP_DEV_NAME_PREFIX "snap_" |
| 68 | #define RBD_MAX_SNAP_NAME_LEN \ |
| 69 | (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1)) |
| 70 | |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 71 | #define RBD_MAX_SNAP_COUNT 510 /* allows max snapc to fit in 4KB */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 72 | |
| 73 | #define RBD_SNAP_HEAD_NAME "-" |
| 74 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 75 | /* This allows a single page to hold an image name sent by OSD */ |
| 76 | #define RBD_IMAGE_NAME_LEN_MAX (PAGE_SIZE - sizeof (__le32) - 1) |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 77 | #define RBD_IMAGE_ID_LEN_MAX 64 |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 78 | |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 79 | #define RBD_OBJ_PREFIX_LEN_MAX 64 |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 80 | |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 81 | /* Feature bits */ |
| 82 | |
| 83 | #define RBD_FEATURE_LAYERING 1 |
| 84 | |
| 85 | /* Features supported by this (client software) implementation. */ |
| 86 | |
| 87 | #define RBD_FEATURES_ALL (0) |
| 88 | |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 89 | /* |
| 90 | * An RBD device name will be "rbd#", where the "rbd" comes from |
| 91 | * RBD_DRV_NAME above, and # is a unique integer identifier. |
| 92 | * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big |
| 93 | * enough to hold all possible device names. |
| 94 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 95 | #define DEV_NAME_LEN 32 |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 96 | #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 97 | |
| 98 | /* |
| 99 | * block device image metadata (in-memory version) |
| 100 | */ |
| 101 | struct rbd_image_header { |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 102 | /* These four fields never change for a given rbd image */ |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 103 | char *object_prefix; |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 104 | u64 features; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 105 | __u8 obj_order; |
| 106 | __u8 crypt_type; |
| 107 | __u8 comp_type; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 108 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 109 | /* The remaining fields need to be updated occasionally */ |
| 110 | u64 image_size; |
| 111 | struct ceph_snap_context *snapc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 112 | char *snap_names; |
| 113 | u64 *snap_sizes; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 114 | |
| 115 | u64 obj_version; |
| 116 | }; |
| 117 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 118 | /* |
| 119 | * An rbd image specification. |
| 120 | * |
| 121 | * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely |
Alex Elder | c66c6e0 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 122 | * identify an image. Each rbd_dev structure includes a pointer to |
| 123 | * an rbd_spec structure that encapsulates this identity. |
| 124 | * |
| 125 | * Each of the id's in an rbd_spec has an associated name. For a |
| 126 | * user-mapped image, the names are supplied and the id's associated |
| 127 | * with them are looked up. For a layered image, a parent image is |
| 128 | * defined by the tuple, and the names are looked up. |
| 129 | * |
| 130 | * An rbd_dev structure contains a parent_spec pointer which is |
| 131 | * non-null if the image it represents is a child in a layered |
| 132 | * image. This pointer will refer to the rbd_spec structure used |
| 133 | * by the parent rbd_dev for its own identity (i.e., the structure |
| 134 | * is shared between the parent and child). |
| 135 | * |
| 136 | * Since these structures are populated once, during the discovery |
| 137 | * phase of image construction, they are effectively immutable so |
| 138 | * we make no effort to synchronize access to them. |
| 139 | * |
| 140 | * Note that code herein does not assume the image name is known (it |
| 141 | * could be a null pointer). |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 142 | */ |
| 143 | struct rbd_spec { |
| 144 | u64 pool_id; |
| 145 | char *pool_name; |
| 146 | |
| 147 | char *image_id; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 148 | char *image_name; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 149 | |
| 150 | u64 snap_id; |
| 151 | char *snap_name; |
| 152 | |
| 153 | struct kref kref; |
| 154 | }; |
| 155 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 156 | /* |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 157 | * an instance of the client. multiple devices may share an rbd client. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 158 | */ |
| 159 | struct rbd_client { |
| 160 | struct ceph_client *client; |
| 161 | struct kref kref; |
| 162 | struct list_head node; |
| 163 | }; |
| 164 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 165 | struct rbd_img_request; |
| 166 | typedef void (*rbd_img_callback_t)(struct rbd_img_request *); |
| 167 | |
| 168 | #define BAD_WHICH U32_MAX /* Good which or bad which, which? */ |
| 169 | |
| 170 | struct rbd_obj_request; |
| 171 | typedef void (*rbd_obj_callback_t)(struct rbd_obj_request *); |
| 172 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 173 | enum obj_request_type { |
| 174 | OBJ_REQUEST_NODATA, OBJ_REQUEST_BIO, OBJ_REQUEST_PAGES |
| 175 | }; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 176 | |
| 177 | struct rbd_obj_request { |
| 178 | const char *object_name; |
| 179 | u64 offset; /* object start byte */ |
| 180 | u64 length; /* bytes from offset */ |
| 181 | |
| 182 | struct rbd_img_request *img_request; |
| 183 | struct list_head links; /* img_request->obj_requests */ |
| 184 | u32 which; /* posn image request list */ |
| 185 | |
| 186 | enum obj_request_type type; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 187 | union { |
| 188 | struct bio *bio_list; |
| 189 | struct { |
| 190 | struct page **pages; |
| 191 | u32 page_count; |
| 192 | }; |
| 193 | }; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 194 | |
| 195 | struct ceph_osd_request *osd_req; |
| 196 | |
| 197 | u64 xferred; /* bytes transferred */ |
| 198 | u64 version; |
| 199 | s32 result; |
| 200 | atomic_t done; |
| 201 | |
| 202 | rbd_obj_callback_t callback; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 203 | struct completion completion; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 204 | |
| 205 | struct kref kref; |
| 206 | }; |
| 207 | |
| 208 | struct rbd_img_request { |
| 209 | struct request *rq; |
| 210 | struct rbd_device *rbd_dev; |
| 211 | u64 offset; /* starting image byte offset */ |
| 212 | u64 length; /* byte count from offset */ |
| 213 | bool write_request; /* false for read */ |
| 214 | union { |
| 215 | struct ceph_snap_context *snapc; /* for writes */ |
| 216 | u64 snap_id; /* for reads */ |
| 217 | }; |
| 218 | spinlock_t completion_lock;/* protects next_completion */ |
| 219 | u32 next_completion; |
| 220 | rbd_img_callback_t callback; |
| 221 | |
| 222 | u32 obj_request_count; |
| 223 | struct list_head obj_requests; /* rbd_obj_request structs */ |
| 224 | |
| 225 | struct kref kref; |
| 226 | }; |
| 227 | |
| 228 | #define for_each_obj_request(ireq, oreq) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 229 | list_for_each_entry(oreq, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 230 | #define for_each_obj_request_from(ireq, oreq) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 231 | list_for_each_entry_from(oreq, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 232 | #define for_each_obj_request_safe(ireq, oreq, n) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 233 | list_for_each_entry_safe_reverse(oreq, n, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 234 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 235 | struct rbd_snap { |
| 236 | struct device dev; |
| 237 | const char *name; |
Josh Durgin | 3591538 | 2011-12-05 18:25:13 -0800 | [diff] [blame] | 238 | u64 size; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 239 | struct list_head node; |
| 240 | u64 id; |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 241 | u64 features; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 242 | }; |
| 243 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 244 | struct rbd_mapping { |
Alex Elder | 99c1f08 | 2012-08-30 14:42:15 -0500 | [diff] [blame] | 245 | u64 size; |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 246 | u64 features; |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 247 | bool read_only; |
| 248 | }; |
| 249 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 250 | /* |
| 251 | * a single device |
| 252 | */ |
| 253 | struct rbd_device { |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 254 | int dev_id; /* blkdev unique id */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 255 | |
| 256 | int major; /* blkdev assigned major */ |
| 257 | struct gendisk *disk; /* blkdev's gendisk and rq */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 258 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 259 | u32 image_format; /* Either 1 or 2 */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 260 | struct rbd_client *rbd_client; |
| 261 | |
| 262 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ |
| 263 | |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 264 | spinlock_t lock; /* queue, flags, open_count */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 265 | |
| 266 | struct rbd_image_header header; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 267 | unsigned long flags; /* possibly lock protected */ |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 268 | struct rbd_spec *spec; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 269 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 270 | char *header_name; |
Alex Elder | 971f839 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 271 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 272 | struct ceph_file_layout layout; |
| 273 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 274 | struct ceph_osd_event *watch_event; |
Alex Elder | 975241a | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 275 | struct rbd_obj_request *watch_request; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 276 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 277 | struct rbd_spec *parent_spec; |
| 278 | u64 parent_overlap; |
| 279 | |
Josh Durgin | c666601 | 2011-11-21 17:11:12 -0800 | [diff] [blame] | 280 | /* protects updating the header */ |
| 281 | struct rw_semaphore header_rwsem; |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 282 | |
| 283 | struct rbd_mapping mapping; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 284 | |
| 285 | struct list_head node; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 286 | |
| 287 | /* list of snapshots */ |
| 288 | struct list_head snaps; |
| 289 | |
| 290 | /* sysfs related */ |
| 291 | struct device dev; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 292 | unsigned long open_count; /* protected by lock */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 293 | }; |
| 294 | |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 295 | /* |
| 296 | * Flag bits for rbd_dev->flags. If atomicity is required, |
| 297 | * rbd_dev->lock is used to protect access. |
| 298 | * |
| 299 | * Currently, only the "removing" flag (which is coupled with the |
| 300 | * "open_count" field) requires atomic access. |
| 301 | */ |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 302 | enum rbd_dev_flags { |
| 303 | RBD_DEV_FLAG_EXISTS, /* mapped snapshot has not been deleted */ |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 304 | RBD_DEV_FLAG_REMOVING, /* this mapping is being removed */ |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 305 | }; |
| 306 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 307 | static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 308 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 309 | static LIST_HEAD(rbd_dev_list); /* devices */ |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 310 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
| 311 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 312 | static LIST_HEAD(rbd_client_list); /* clients */ |
| 313 | static DEFINE_SPINLOCK(rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 314 | |
Alex Elder | 304f680 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 315 | static int rbd_dev_snaps_update(struct rbd_device *rbd_dev); |
| 316 | static int rbd_dev_snaps_register(struct rbd_device *rbd_dev); |
| 317 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 318 | static void rbd_dev_release(struct device *dev); |
Alex Elder | 41f38c2 | 2012-10-25 23:34:40 -0500 | [diff] [blame] | 319 | static void rbd_remove_snap_dev(struct rbd_snap *snap); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 320 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 321 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
| 322 | size_t count); |
| 323 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, |
| 324 | size_t count); |
| 325 | |
| 326 | static struct bus_attribute rbd_bus_attrs[] = { |
| 327 | __ATTR(add, S_IWUSR, NULL, rbd_add), |
| 328 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), |
| 329 | __ATTR_NULL |
| 330 | }; |
| 331 | |
| 332 | static struct bus_type rbd_bus_type = { |
| 333 | .name = "rbd", |
| 334 | .bus_attrs = rbd_bus_attrs, |
| 335 | }; |
| 336 | |
| 337 | static void rbd_root_dev_release(struct device *dev) |
| 338 | { |
| 339 | } |
| 340 | |
| 341 | static struct device rbd_root_dev = { |
| 342 | .init_name = "rbd", |
| 343 | .release = rbd_root_dev_release, |
| 344 | }; |
| 345 | |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 346 | static __printf(2, 3) |
| 347 | void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...) |
| 348 | { |
| 349 | struct va_format vaf; |
| 350 | va_list args; |
| 351 | |
| 352 | va_start(args, fmt); |
| 353 | vaf.fmt = fmt; |
| 354 | vaf.va = &args; |
| 355 | |
| 356 | if (!rbd_dev) |
| 357 | printk(KERN_WARNING "%s: %pV\n", RBD_DRV_NAME, &vaf); |
| 358 | else if (rbd_dev->disk) |
| 359 | printk(KERN_WARNING "%s: %s: %pV\n", |
| 360 | RBD_DRV_NAME, rbd_dev->disk->disk_name, &vaf); |
| 361 | else if (rbd_dev->spec && rbd_dev->spec->image_name) |
| 362 | printk(KERN_WARNING "%s: image %s: %pV\n", |
| 363 | RBD_DRV_NAME, rbd_dev->spec->image_name, &vaf); |
| 364 | else if (rbd_dev->spec && rbd_dev->spec->image_id) |
| 365 | printk(KERN_WARNING "%s: id %s: %pV\n", |
| 366 | RBD_DRV_NAME, rbd_dev->spec->image_id, &vaf); |
| 367 | else /* punt */ |
| 368 | printk(KERN_WARNING "%s: rbd_dev %p: %pV\n", |
| 369 | RBD_DRV_NAME, rbd_dev, &vaf); |
| 370 | va_end(args); |
| 371 | } |
| 372 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 373 | #ifdef RBD_DEBUG |
| 374 | #define rbd_assert(expr) \ |
| 375 | if (unlikely(!(expr))) { \ |
| 376 | printk(KERN_ERR "\nAssertion failure in %s() " \ |
| 377 | "at line %d:\n\n" \ |
| 378 | "\trbd_assert(%s);\n\n", \ |
| 379 | __func__, __LINE__, #expr); \ |
| 380 | BUG(); \ |
| 381 | } |
| 382 | #else /* !RBD_DEBUG */ |
| 383 | # define rbd_assert(expr) ((void) 0) |
| 384 | #endif /* !RBD_DEBUG */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 385 | |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 386 | static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver); |
| 387 | static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 388 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 389 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
| 390 | { |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 391 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 392 | bool removing = false; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 393 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 394 | if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 395 | return -EROFS; |
| 396 | |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 397 | spin_lock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 398 | if (test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags)) |
| 399 | removing = true; |
| 400 | else |
| 401 | rbd_dev->open_count++; |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 402 | spin_unlock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 403 | if (removing) |
| 404 | return -ENOENT; |
| 405 | |
Alex Elder | 42382b7 | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 406 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
Alex Elder | c3e946c | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 407 | (void) get_device(&rbd_dev->dev); |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 408 | set_device_ro(bdev, rbd_dev->mapping.read_only); |
Alex Elder | 42382b7 | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 409 | mutex_unlock(&ctl_mutex); |
Alex Elder | 340c7a2 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 410 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 411 | return 0; |
| 412 | } |
| 413 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 414 | static int rbd_release(struct gendisk *disk, fmode_t mode) |
| 415 | { |
| 416 | struct rbd_device *rbd_dev = disk->private_data; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 417 | unsigned long open_count_before; |
| 418 | |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 419 | spin_lock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 420 | open_count_before = rbd_dev->open_count--; |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 421 | spin_unlock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 422 | rbd_assert(open_count_before > 0); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 423 | |
Alex Elder | 42382b7 | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 424 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
Alex Elder | c3e946c | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 425 | put_device(&rbd_dev->dev); |
Alex Elder | 42382b7 | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 426 | mutex_unlock(&ctl_mutex); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 427 | |
| 428 | return 0; |
| 429 | } |
| 430 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 431 | static const struct block_device_operations rbd_bd_ops = { |
| 432 | .owner = THIS_MODULE, |
| 433 | .open = rbd_open, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 434 | .release = rbd_release, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 435 | }; |
| 436 | |
| 437 | /* |
| 438 | * Initialize an rbd client instance. |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 439 | * We own *ceph_opts. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 440 | */ |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 441 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 442 | { |
| 443 | struct rbd_client *rbdc; |
| 444 | int ret = -ENOMEM; |
| 445 | |
| 446 | dout("rbd_client_create\n"); |
| 447 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); |
| 448 | if (!rbdc) |
| 449 | goto out_opt; |
| 450 | |
| 451 | kref_init(&rbdc->kref); |
| 452 | INIT_LIST_HEAD(&rbdc->node); |
| 453 | |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 454 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
| 455 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 456 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 457 | if (IS_ERR(rbdc->client)) |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 458 | goto out_mutex; |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 459 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 460 | |
| 461 | ret = ceph_open_session(rbdc->client); |
| 462 | if (ret < 0) |
| 463 | goto out_err; |
| 464 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 465 | spin_lock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 466 | list_add_tail(&rbdc->node, &rbd_client_list); |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 467 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 468 | |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 469 | mutex_unlock(&ctl_mutex); |
| 470 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 471 | dout("rbd_client_create created %p\n", rbdc); |
| 472 | return rbdc; |
| 473 | |
| 474 | out_err: |
| 475 | ceph_destroy_client(rbdc->client); |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 476 | out_mutex: |
| 477 | mutex_unlock(&ctl_mutex); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 478 | kfree(rbdc); |
| 479 | out_opt: |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 480 | if (ceph_opts) |
| 481 | ceph_destroy_options(ceph_opts); |
Vasiliy Kulikov | 28f259b | 2010-09-26 12:59:37 +0400 | [diff] [blame] | 482 | return ERR_PTR(ret); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 483 | } |
| 484 | |
| 485 | /* |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 486 | * Find a ceph client with specific addr and configuration. If |
| 487 | * found, bump its reference count. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 488 | */ |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 489 | static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 490 | { |
| 491 | struct rbd_client *client_node; |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 492 | bool found = false; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 493 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 494 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 495 | return NULL; |
| 496 | |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 497 | spin_lock(&rbd_client_list_lock); |
| 498 | list_for_each_entry(client_node, &rbd_client_list, node) { |
| 499 | if (!ceph_compare_options(ceph_opts, client_node->client)) { |
| 500 | kref_get(&client_node->kref); |
| 501 | found = true; |
| 502 | break; |
| 503 | } |
| 504 | } |
| 505 | spin_unlock(&rbd_client_list_lock); |
| 506 | |
| 507 | return found ? client_node : NULL; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 508 | } |
| 509 | |
| 510 | /* |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 511 | * mount options |
| 512 | */ |
| 513 | enum { |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 514 | Opt_last_int, |
| 515 | /* int args above */ |
| 516 | Opt_last_string, |
| 517 | /* string args above */ |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 518 | Opt_read_only, |
| 519 | Opt_read_write, |
| 520 | /* Boolean args above */ |
| 521 | Opt_last_bool, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 522 | }; |
| 523 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 524 | static match_table_t rbd_opts_tokens = { |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 525 | /* int args above */ |
| 526 | /* string args above */ |
Alex Elder | be466c1 | 2012-10-22 11:31:26 -0500 | [diff] [blame] | 527 | {Opt_read_only, "read_only"}, |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 528 | {Opt_read_only, "ro"}, /* Alternate spelling */ |
| 529 | {Opt_read_write, "read_write"}, |
| 530 | {Opt_read_write, "rw"}, /* Alternate spelling */ |
| 531 | /* Boolean args above */ |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 532 | {-1, NULL} |
| 533 | }; |
| 534 | |
Alex Elder | 98571b5 | 2013-01-20 14:44:42 -0600 | [diff] [blame] | 535 | struct rbd_options { |
| 536 | bool read_only; |
| 537 | }; |
| 538 | |
| 539 | #define RBD_READ_ONLY_DEFAULT false |
| 540 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 541 | static int parse_rbd_opts_token(char *c, void *private) |
| 542 | { |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 543 | struct rbd_options *rbd_opts = private; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 544 | substring_t argstr[MAX_OPT_ARGS]; |
| 545 | int token, intval, ret; |
| 546 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 547 | token = match_token(c, rbd_opts_tokens, argstr); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 548 | if (token < 0) |
| 549 | return -EINVAL; |
| 550 | |
| 551 | if (token < Opt_last_int) { |
| 552 | ret = match_int(&argstr[0], &intval); |
| 553 | if (ret < 0) { |
| 554 | pr_err("bad mount option arg (not int) " |
| 555 | "at '%s'\n", c); |
| 556 | return ret; |
| 557 | } |
| 558 | dout("got int token %d val %d\n", token, intval); |
| 559 | } else if (token > Opt_last_int && token < Opt_last_string) { |
| 560 | dout("got string token %d val %s\n", token, |
| 561 | argstr[0].from); |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 562 | } else if (token > Opt_last_string && token < Opt_last_bool) { |
| 563 | dout("got Boolean token %d\n", token); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 564 | } else { |
| 565 | dout("got token %d\n", token); |
| 566 | } |
| 567 | |
| 568 | switch (token) { |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 569 | case Opt_read_only: |
| 570 | rbd_opts->read_only = true; |
| 571 | break; |
| 572 | case Opt_read_write: |
| 573 | rbd_opts->read_only = false; |
| 574 | break; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 575 | default: |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 576 | rbd_assert(false); |
| 577 | break; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 578 | } |
| 579 | return 0; |
| 580 | } |
| 581 | |
| 582 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 583 | * Get a ceph client with specific addr and configuration, if one does |
| 584 | * not exist create it. |
| 585 | */ |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 586 | static struct rbd_client *rbd_get_client(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 587 | { |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 588 | struct rbd_client *rbdc; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 589 | |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 590 | rbdc = rbd_client_find(ceph_opts); |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 591 | if (rbdc) /* using an existing client */ |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 592 | ceph_destroy_options(ceph_opts); |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 593 | else |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 594 | rbdc = rbd_client_create(ceph_opts); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 595 | |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 596 | return rbdc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 597 | } |
| 598 | |
| 599 | /* |
| 600 | * Destroy ceph client |
Alex Elder | d23a4b3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 601 | * |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 602 | * Caller must hold rbd_client_list_lock. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 603 | */ |
| 604 | static void rbd_client_release(struct kref *kref) |
| 605 | { |
| 606 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); |
| 607 | |
| 608 | dout("rbd_release_client %p\n", rbdc); |
Alex Elder | cd9d9f5 | 2012-04-04 13:35:44 -0500 | [diff] [blame] | 609 | spin_lock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 610 | list_del(&rbdc->node); |
Alex Elder | cd9d9f5 | 2012-04-04 13:35:44 -0500 | [diff] [blame] | 611 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 612 | |
| 613 | ceph_destroy_client(rbdc->client); |
| 614 | kfree(rbdc); |
| 615 | } |
| 616 | |
| 617 | /* |
| 618 | * Drop reference to ceph client node. If it's not referenced anymore, release |
| 619 | * it. |
| 620 | */ |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 621 | static void rbd_put_client(struct rbd_client *rbdc) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 622 | { |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 623 | if (rbdc) |
| 624 | kref_put(&rbdc->kref, rbd_client_release); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 625 | } |
| 626 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 627 | static bool rbd_image_format_valid(u32 image_format) |
| 628 | { |
| 629 | return image_format == 1 || image_format == 2; |
| 630 | } |
| 631 | |
Alex Elder | 8e94af8 | 2012-07-25 09:32:40 -0500 | [diff] [blame] | 632 | static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) |
| 633 | { |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 634 | size_t size; |
| 635 | u32 snap_count; |
| 636 | |
| 637 | /* The header has to start with the magic rbd header text */ |
| 638 | if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) |
| 639 | return false; |
| 640 | |
Alex Elder | db2388b | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 641 | /* The bio layer requires at least sector-sized I/O */ |
| 642 | |
| 643 | if (ondisk->options.order < SECTOR_SHIFT) |
| 644 | return false; |
| 645 | |
| 646 | /* If we use u64 in a few spots we may be able to loosen this */ |
| 647 | |
| 648 | if (ondisk->options.order > 8 * sizeof (int) - 1) |
| 649 | return false; |
| 650 | |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 651 | /* |
| 652 | * The size of a snapshot header has to fit in a size_t, and |
| 653 | * that limits the number of snapshots. |
| 654 | */ |
| 655 | snap_count = le32_to_cpu(ondisk->snap_count); |
| 656 | size = SIZE_MAX - sizeof (struct ceph_snap_context); |
| 657 | if (snap_count > size / sizeof (__le64)) |
| 658 | return false; |
| 659 | |
| 660 | /* |
| 661 | * Not only that, but the size of the entire the snapshot |
| 662 | * header must also be representable in a size_t. |
| 663 | */ |
| 664 | size -= snap_count * sizeof (__le64); |
| 665 | if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) |
| 666 | return false; |
| 667 | |
| 668 | return true; |
Alex Elder | 8e94af8 | 2012-07-25 09:32:40 -0500 | [diff] [blame] | 669 | } |
| 670 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 671 | /* |
| 672 | * Create a new header structure, translate header format from the on-disk |
| 673 | * header. |
| 674 | */ |
| 675 | static int rbd_header_from_disk(struct rbd_image_header *header, |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 676 | struct rbd_image_header_ondisk *ondisk) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 677 | { |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 678 | u32 snap_count; |
Alex Elder | 58c17b0 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 679 | size_t len; |
Alex Elder | d2bb24e | 2012-07-26 23:37:14 -0500 | [diff] [blame] | 680 | size_t size; |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 681 | u32 i; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 682 | |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 683 | memset(header, 0, sizeof (*header)); |
| 684 | |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 685 | snap_count = le32_to_cpu(ondisk->snap_count); |
| 686 | |
Alex Elder | 58c17b0 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 687 | len = strnlen(ondisk->object_prefix, sizeof (ondisk->object_prefix)); |
| 688 | header->object_prefix = kmalloc(len + 1, GFP_KERNEL); |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 689 | if (!header->object_prefix) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 690 | return -ENOMEM; |
Alex Elder | 58c17b0 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 691 | memcpy(header->object_prefix, ondisk->object_prefix, len); |
| 692 | header->object_prefix[len] = '\0'; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 693 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 694 | if (snap_count) { |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 695 | u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
| 696 | |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 697 | /* Save a copy of the snapshot names */ |
| 698 | |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 699 | if (snap_names_len > (u64) SIZE_MAX) |
| 700 | return -EIO; |
| 701 | header->snap_names = kmalloc(snap_names_len, GFP_KERNEL); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 702 | if (!header->snap_names) |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 703 | goto out_err; |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 704 | /* |
| 705 | * Note that rbd_dev_v1_header_read() guarantees |
| 706 | * the ondisk buffer we're working with has |
| 707 | * snap_names_len bytes beyond the end of the |
| 708 | * snapshot id array, this memcpy() is safe. |
| 709 | */ |
| 710 | memcpy(header->snap_names, &ondisk->snaps[snap_count], |
| 711 | snap_names_len); |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 712 | |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 713 | /* Record each snapshot's size */ |
| 714 | |
Alex Elder | d2bb24e | 2012-07-26 23:37:14 -0500 | [diff] [blame] | 715 | size = snap_count * sizeof (*header->snap_sizes); |
| 716 | header->snap_sizes = kmalloc(size, GFP_KERNEL); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 717 | if (!header->snap_sizes) |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 718 | goto out_err; |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 719 | for (i = 0; i < snap_count; i++) |
| 720 | header->snap_sizes[i] = |
| 721 | le64_to_cpu(ondisk->snaps[i].image_size); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 722 | } else { |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 723 | WARN_ON(ondisk->snap_names_len); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 724 | header->snap_names = NULL; |
| 725 | header->snap_sizes = NULL; |
| 726 | } |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 727 | |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 728 | header->features = 0; /* No features support in v1 images */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 729 | header->obj_order = ondisk->options.order; |
| 730 | header->crypt_type = ondisk->options.crypt_type; |
| 731 | header->comp_type = ondisk->options.comp_type; |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 732 | |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 733 | /* Allocate and fill in the snapshot context */ |
| 734 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 735 | header->image_size = le64_to_cpu(ondisk->image_size); |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 736 | size = sizeof (struct ceph_snap_context); |
| 737 | size += snap_count * sizeof (header->snapc->snaps[0]); |
| 738 | header->snapc = kzalloc(size, GFP_KERNEL); |
| 739 | if (!header->snapc) |
| 740 | goto out_err; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 741 | |
| 742 | atomic_set(&header->snapc->nref, 1); |
Alex Elder | 505cbb9 | 2012-07-19 08:49:18 -0500 | [diff] [blame] | 743 | header->snapc->seq = le64_to_cpu(ondisk->snap_seq); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 744 | header->snapc->num_snaps = snap_count; |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 745 | for (i = 0; i < snap_count; i++) |
| 746 | header->snapc->snaps[i] = |
| 747 | le64_to_cpu(ondisk->snaps[i].id); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 748 | |
| 749 | return 0; |
| 750 | |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 751 | out_err: |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 752 | kfree(header->snap_sizes); |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 753 | header->snap_sizes = NULL; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 754 | kfree(header->snap_names); |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 755 | header->snap_names = NULL; |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 756 | kfree(header->object_prefix); |
| 757 | header->object_prefix = NULL; |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 758 | |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 759 | return -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 760 | } |
| 761 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 762 | static const char *rbd_snap_name(struct rbd_device *rbd_dev, u64 snap_id) |
| 763 | { |
| 764 | struct rbd_snap *snap; |
| 765 | |
| 766 | if (snap_id == CEPH_NOSNAP) |
| 767 | return RBD_SNAP_HEAD_NAME; |
| 768 | |
| 769 | list_for_each_entry(snap, &rbd_dev->snaps, node) |
| 770 | if (snap_id == snap->id) |
| 771 | return snap->name; |
| 772 | |
| 773 | return NULL; |
| 774 | } |
| 775 | |
Alex Elder | 8836b99 | 2012-08-30 14:42:15 -0500 | [diff] [blame] | 776 | static int snap_by_name(struct rbd_device *rbd_dev, const char *snap_name) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 777 | { |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 778 | |
Alex Elder | e86924a | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 779 | struct rbd_snap *snap; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 780 | |
Alex Elder | e86924a | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 781 | list_for_each_entry(snap, &rbd_dev->snaps, node) { |
| 782 | if (!strcmp(snap_name, snap->name)) { |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 783 | rbd_dev->spec->snap_id = snap->id; |
Alex Elder | e86924a | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 784 | rbd_dev->mapping.size = snap->size; |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 785 | rbd_dev->mapping.features = snap->features; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 786 | |
Alex Elder | e86924a | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 787 | return 0; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 788 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 789 | } |
Alex Elder | e86924a | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 790 | |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 791 | return -ENOENT; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 792 | } |
| 793 | |
Alex Elder | 819d52b | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 794 | static int rbd_dev_set_mapping(struct rbd_device *rbd_dev) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 795 | { |
Alex Elder | 78dc447 | 2012-07-19 08:49:18 -0500 | [diff] [blame] | 796 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 797 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 798 | if (!memcmp(rbd_dev->spec->snap_name, RBD_SNAP_HEAD_NAME, |
Josh Durgin | cc9d734 | 2011-11-21 18:19:13 -0800 | [diff] [blame] | 799 | sizeof (RBD_SNAP_HEAD_NAME))) { |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 800 | rbd_dev->spec->snap_id = CEPH_NOSNAP; |
Alex Elder | 99c1f08 | 2012-08-30 14:42:15 -0500 | [diff] [blame] | 801 | rbd_dev->mapping.size = rbd_dev->header.image_size; |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 802 | rbd_dev->mapping.features = rbd_dev->header.features; |
Alex Elder | e86924a | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 803 | ret = 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 804 | } else { |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 805 | ret = snap_by_name(rbd_dev, rbd_dev->spec->snap_name); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 806 | if (ret < 0) |
| 807 | goto done; |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 808 | rbd_dev->mapping.read_only = true; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 809 | } |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 810 | set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
| 811 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 812 | done: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 813 | return ret; |
| 814 | } |
| 815 | |
| 816 | static void rbd_header_free(struct rbd_image_header *header) |
| 817 | { |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 818 | kfree(header->object_prefix); |
Alex Elder | d78fd7a | 2012-07-26 23:37:14 -0500 | [diff] [blame] | 819 | header->object_prefix = NULL; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 820 | kfree(header->snap_sizes); |
Alex Elder | d78fd7a | 2012-07-26 23:37:14 -0500 | [diff] [blame] | 821 | header->snap_sizes = NULL; |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 822 | kfree(header->snap_names); |
Alex Elder | d78fd7a | 2012-07-26 23:37:14 -0500 | [diff] [blame] | 823 | header->snap_names = NULL; |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 824 | ceph_put_snap_context(header->snapc); |
Alex Elder | d78fd7a | 2012-07-26 23:37:14 -0500 | [diff] [blame] | 825 | header->snapc = NULL; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 826 | } |
| 827 | |
Alex Elder | 98571b5 | 2013-01-20 14:44:42 -0600 | [diff] [blame] | 828 | static const char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 829 | { |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 830 | char *name; |
| 831 | u64 segment; |
| 832 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 833 | |
Alex Elder | 2fd82b9 | 2012-11-09 15:05:54 -0600 | [diff] [blame] | 834 | name = kmalloc(MAX_OBJ_NAME_SIZE + 1, GFP_NOIO); |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 835 | if (!name) |
| 836 | return NULL; |
| 837 | segment = offset >> rbd_dev->header.obj_order; |
Alex Elder | 2fd82b9 | 2012-11-09 15:05:54 -0600 | [diff] [blame] | 838 | ret = snprintf(name, MAX_OBJ_NAME_SIZE + 1, "%s.%012llx", |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 839 | rbd_dev->header.object_prefix, segment); |
Alex Elder | 2fd82b9 | 2012-11-09 15:05:54 -0600 | [diff] [blame] | 840 | if (ret < 0 || ret > MAX_OBJ_NAME_SIZE) { |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 841 | pr_err("error formatting segment name for #%llu (%d)\n", |
| 842 | segment, ret); |
| 843 | kfree(name); |
| 844 | name = NULL; |
| 845 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 846 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 847 | return name; |
| 848 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 849 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 850 | static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) |
| 851 | { |
| 852 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 853 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 854 | return offset & (segment_size - 1); |
| 855 | } |
| 856 | |
| 857 | static u64 rbd_segment_length(struct rbd_device *rbd_dev, |
| 858 | u64 offset, u64 length) |
| 859 | { |
| 860 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; |
| 861 | |
| 862 | offset &= segment_size - 1; |
| 863 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 864 | rbd_assert(length <= U64_MAX - offset); |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 865 | if (offset + length > segment_size) |
| 866 | length = segment_size - offset; |
| 867 | |
| 868 | return length; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 869 | } |
| 870 | |
| 871 | /* |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 872 | * returns the size of an object in the image |
| 873 | */ |
| 874 | static u64 rbd_obj_bytes(struct rbd_image_header *header) |
| 875 | { |
| 876 | return 1 << header->obj_order; |
| 877 | } |
| 878 | |
| 879 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 880 | * bio helpers |
| 881 | */ |
| 882 | |
| 883 | static void bio_chain_put(struct bio *chain) |
| 884 | { |
| 885 | struct bio *tmp; |
| 886 | |
| 887 | while (chain) { |
| 888 | tmp = chain; |
| 889 | chain = chain->bi_next; |
| 890 | bio_put(tmp); |
| 891 | } |
| 892 | } |
| 893 | |
| 894 | /* |
| 895 | * zeros a bio chain, starting at specific offset |
| 896 | */ |
| 897 | static void zero_bio_chain(struct bio *chain, int start_ofs) |
| 898 | { |
| 899 | struct bio_vec *bv; |
| 900 | unsigned long flags; |
| 901 | void *buf; |
| 902 | int i; |
| 903 | int pos = 0; |
| 904 | |
| 905 | while (chain) { |
| 906 | bio_for_each_segment(bv, chain, i) { |
| 907 | if (pos + bv->bv_len > start_ofs) { |
| 908 | int remainder = max(start_ofs - pos, 0); |
| 909 | buf = bvec_kmap_irq(bv, &flags); |
| 910 | memset(buf + remainder, 0, |
| 911 | bv->bv_len - remainder); |
Dan Carpenter | 85b5aaa | 2010-10-11 21:15:11 +0200 | [diff] [blame] | 912 | bvec_kunmap_irq(buf, &flags); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 913 | } |
| 914 | pos += bv->bv_len; |
| 915 | } |
| 916 | |
| 917 | chain = chain->bi_next; |
| 918 | } |
| 919 | } |
| 920 | |
| 921 | /* |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 922 | * Clone a portion of a bio, starting at the given byte offset |
| 923 | * and continuing for the number of bytes indicated. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 924 | */ |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 925 | static struct bio *bio_clone_range(struct bio *bio_src, |
| 926 | unsigned int offset, |
| 927 | unsigned int len, |
| 928 | gfp_t gfpmask) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 929 | { |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 930 | struct bio_vec *bv; |
| 931 | unsigned int resid; |
| 932 | unsigned short idx; |
| 933 | unsigned int voff; |
| 934 | unsigned short end_idx; |
| 935 | unsigned short vcnt; |
| 936 | struct bio *bio; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 937 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 938 | /* Handle the easy case for the caller */ |
| 939 | |
| 940 | if (!offset && len == bio_src->bi_size) |
| 941 | return bio_clone(bio_src, gfpmask); |
| 942 | |
| 943 | if (WARN_ON_ONCE(!len)) |
| 944 | return NULL; |
| 945 | if (WARN_ON_ONCE(len > bio_src->bi_size)) |
| 946 | return NULL; |
| 947 | if (WARN_ON_ONCE(offset > bio_src->bi_size - len)) |
| 948 | return NULL; |
| 949 | |
| 950 | /* Find first affected segment... */ |
| 951 | |
| 952 | resid = offset; |
| 953 | __bio_for_each_segment(bv, bio_src, idx, 0) { |
| 954 | if (resid < bv->bv_len) |
| 955 | break; |
| 956 | resid -= bv->bv_len; |
| 957 | } |
| 958 | voff = resid; |
| 959 | |
| 960 | /* ...and the last affected segment */ |
| 961 | |
| 962 | resid += len; |
| 963 | __bio_for_each_segment(bv, bio_src, end_idx, idx) { |
| 964 | if (resid <= bv->bv_len) |
| 965 | break; |
| 966 | resid -= bv->bv_len; |
| 967 | } |
| 968 | vcnt = end_idx - idx + 1; |
| 969 | |
| 970 | /* Build the clone */ |
| 971 | |
| 972 | bio = bio_alloc(gfpmask, (unsigned int) vcnt); |
| 973 | if (!bio) |
| 974 | return NULL; /* ENOMEM */ |
| 975 | |
| 976 | bio->bi_bdev = bio_src->bi_bdev; |
| 977 | bio->bi_sector = bio_src->bi_sector + (offset >> SECTOR_SHIFT); |
| 978 | bio->bi_rw = bio_src->bi_rw; |
| 979 | bio->bi_flags |= 1 << BIO_CLONED; |
| 980 | |
| 981 | /* |
| 982 | * Copy over our part of the bio_vec, then update the first |
| 983 | * and last (or only) entries. |
| 984 | */ |
| 985 | memcpy(&bio->bi_io_vec[0], &bio_src->bi_io_vec[idx], |
| 986 | vcnt * sizeof (struct bio_vec)); |
| 987 | bio->bi_io_vec[0].bv_offset += voff; |
| 988 | if (vcnt > 1) { |
| 989 | bio->bi_io_vec[0].bv_len -= voff; |
| 990 | bio->bi_io_vec[vcnt - 1].bv_len = resid; |
| 991 | } else { |
| 992 | bio->bi_io_vec[0].bv_len = len; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 993 | } |
| 994 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 995 | bio->bi_vcnt = vcnt; |
| 996 | bio->bi_size = len; |
| 997 | bio->bi_idx = 0; |
Alex Elder | 542582f | 2012-08-09 10:33:25 -0700 | [diff] [blame] | 998 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 999 | return bio; |
| 1000 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1001 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1002 | /* |
| 1003 | * Clone a portion of a bio chain, starting at the given byte offset |
| 1004 | * into the first bio in the source chain and continuing for the |
| 1005 | * number of bytes indicated. The result is another bio chain of |
| 1006 | * exactly the given length, or a null pointer on error. |
| 1007 | * |
| 1008 | * The bio_src and offset parameters are both in-out. On entry they |
| 1009 | * refer to the first source bio and the offset into that bio where |
| 1010 | * the start of data to be cloned is located. |
| 1011 | * |
| 1012 | * On return, bio_src is updated to refer to the bio in the source |
| 1013 | * chain that contains first un-cloned byte, and *offset will |
| 1014 | * contain the offset of that byte within that bio. |
| 1015 | */ |
| 1016 | static struct bio *bio_chain_clone_range(struct bio **bio_src, |
| 1017 | unsigned int *offset, |
| 1018 | unsigned int len, |
| 1019 | gfp_t gfpmask) |
| 1020 | { |
| 1021 | struct bio *bi = *bio_src; |
| 1022 | unsigned int off = *offset; |
| 1023 | struct bio *chain = NULL; |
| 1024 | struct bio **end; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1025 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1026 | /* Build up a chain of clone bios up to the limit */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1027 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1028 | if (!bi || off >= bi->bi_size || !len) |
| 1029 | return NULL; /* Nothing to clone */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1030 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1031 | end = &chain; |
| 1032 | while (len) { |
| 1033 | unsigned int bi_size; |
| 1034 | struct bio *bio; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1035 | |
Alex Elder | f5400b7 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 1036 | if (!bi) { |
| 1037 | rbd_warn(NULL, "bio_chain exhausted with %u left", len); |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1038 | goto out_err; /* EINVAL; ran out of bio's */ |
Alex Elder | f5400b7 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 1039 | } |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1040 | bi_size = min_t(unsigned int, bi->bi_size - off, len); |
| 1041 | bio = bio_clone_range(bi, off, bi_size, gfpmask); |
| 1042 | if (!bio) |
| 1043 | goto out_err; /* ENOMEM */ |
| 1044 | |
| 1045 | *end = bio; |
| 1046 | end = &bio->bi_next; |
| 1047 | |
| 1048 | off += bi_size; |
| 1049 | if (off == bi->bi_size) { |
| 1050 | bi = bi->bi_next; |
| 1051 | off = 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1052 | } |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1053 | len -= bi_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1054 | } |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1055 | *bio_src = bi; |
| 1056 | *offset = off; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1057 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1058 | return chain; |
| 1059 | out_err: |
| 1060 | bio_chain_put(chain); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1061 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1062 | return NULL; |
| 1063 | } |
| 1064 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1065 | static void rbd_obj_request_get(struct rbd_obj_request *obj_request) |
| 1066 | { |
| 1067 | kref_get(&obj_request->kref); |
| 1068 | } |
| 1069 | |
| 1070 | static void rbd_obj_request_destroy(struct kref *kref); |
| 1071 | static void rbd_obj_request_put(struct rbd_obj_request *obj_request) |
| 1072 | { |
| 1073 | rbd_assert(obj_request != NULL); |
| 1074 | kref_put(&obj_request->kref, rbd_obj_request_destroy); |
| 1075 | } |
| 1076 | |
| 1077 | static void rbd_img_request_get(struct rbd_img_request *img_request) |
| 1078 | { |
| 1079 | kref_get(&img_request->kref); |
| 1080 | } |
| 1081 | |
| 1082 | static void rbd_img_request_destroy(struct kref *kref); |
| 1083 | static void rbd_img_request_put(struct rbd_img_request *img_request) |
| 1084 | { |
| 1085 | rbd_assert(img_request != NULL); |
| 1086 | kref_put(&img_request->kref, rbd_img_request_destroy); |
| 1087 | } |
| 1088 | |
| 1089 | static inline void rbd_img_obj_request_add(struct rbd_img_request *img_request, |
| 1090 | struct rbd_obj_request *obj_request) |
| 1091 | { |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1092 | rbd_assert(obj_request->img_request == NULL); |
| 1093 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1094 | rbd_obj_request_get(obj_request); |
| 1095 | obj_request->img_request = img_request; |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1096 | obj_request->which = img_request->obj_request_count; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1097 | rbd_assert(obj_request->which != BAD_WHICH); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1098 | img_request->obj_request_count++; |
| 1099 | list_add_tail(&obj_request->links, &img_request->obj_requests); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1100 | } |
| 1101 | |
| 1102 | static inline void rbd_img_obj_request_del(struct rbd_img_request *img_request, |
| 1103 | struct rbd_obj_request *obj_request) |
| 1104 | { |
| 1105 | rbd_assert(obj_request->which != BAD_WHICH); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1106 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1107 | list_del(&obj_request->links); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1108 | rbd_assert(img_request->obj_request_count > 0); |
| 1109 | img_request->obj_request_count--; |
| 1110 | rbd_assert(obj_request->which == img_request->obj_request_count); |
| 1111 | obj_request->which = BAD_WHICH; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1112 | rbd_assert(obj_request->img_request == img_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1113 | obj_request->img_request = NULL; |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1114 | obj_request->callback = NULL; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1115 | rbd_obj_request_put(obj_request); |
| 1116 | } |
| 1117 | |
| 1118 | static bool obj_request_type_valid(enum obj_request_type type) |
| 1119 | { |
| 1120 | switch (type) { |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1121 | case OBJ_REQUEST_NODATA: |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1122 | case OBJ_REQUEST_BIO: |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1123 | case OBJ_REQUEST_PAGES: |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1124 | return true; |
| 1125 | default: |
| 1126 | return false; |
| 1127 | } |
| 1128 | } |
| 1129 | |
Alex Elder | 8d23bf2 | 2012-11-19 22:55:21 -0600 | [diff] [blame] | 1130 | struct ceph_osd_req_op *rbd_osd_req_op_create(u16 opcode, ...) |
| 1131 | { |
| 1132 | struct ceph_osd_req_op *op; |
| 1133 | va_list args; |
Alex Elder | 2647ba3 | 2012-11-19 22:55:21 -0600 | [diff] [blame] | 1134 | size_t size; |
Alex Elder | 8d23bf2 | 2012-11-19 22:55:21 -0600 | [diff] [blame] | 1135 | |
| 1136 | op = kzalloc(sizeof (*op), GFP_NOIO); |
| 1137 | if (!op) |
| 1138 | return NULL; |
| 1139 | op->op = opcode; |
| 1140 | va_start(args, opcode); |
| 1141 | switch (opcode) { |
| 1142 | case CEPH_OSD_OP_READ: |
| 1143 | case CEPH_OSD_OP_WRITE: |
| 1144 | /* rbd_osd_req_op_create(READ, offset, length) */ |
| 1145 | /* rbd_osd_req_op_create(WRITE, offset, length) */ |
| 1146 | op->extent.offset = va_arg(args, u64); |
| 1147 | op->extent.length = va_arg(args, u64); |
| 1148 | if (opcode == CEPH_OSD_OP_WRITE) |
| 1149 | op->payload_len = op->extent.length; |
| 1150 | break; |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1151 | case CEPH_OSD_OP_STAT: |
| 1152 | break; |
Alex Elder | 2647ba3 | 2012-11-19 22:55:21 -0600 | [diff] [blame] | 1153 | case CEPH_OSD_OP_CALL: |
| 1154 | /* rbd_osd_req_op_create(CALL, class, method, data, datalen) */ |
| 1155 | op->cls.class_name = va_arg(args, char *); |
| 1156 | size = strlen(op->cls.class_name); |
| 1157 | rbd_assert(size <= (size_t) U8_MAX); |
| 1158 | op->cls.class_len = size; |
| 1159 | op->payload_len = size; |
| 1160 | |
| 1161 | op->cls.method_name = va_arg(args, char *); |
| 1162 | size = strlen(op->cls.method_name); |
| 1163 | rbd_assert(size <= (size_t) U8_MAX); |
| 1164 | op->cls.method_len = size; |
| 1165 | op->payload_len += size; |
| 1166 | |
| 1167 | op->cls.argc = 0; |
| 1168 | op->cls.indata = va_arg(args, void *); |
| 1169 | size = va_arg(args, size_t); |
| 1170 | rbd_assert(size <= (size_t) U32_MAX); |
| 1171 | op->cls.indata_len = (u32) size; |
| 1172 | op->payload_len += size; |
| 1173 | break; |
Alex Elder | 5efea49 | 2012-11-19 22:55:21 -0600 | [diff] [blame] | 1174 | case CEPH_OSD_OP_NOTIFY_ACK: |
| 1175 | case CEPH_OSD_OP_WATCH: |
| 1176 | /* rbd_osd_req_op_create(NOTIFY_ACK, cookie, version) */ |
| 1177 | /* rbd_osd_req_op_create(WATCH, cookie, version, flag) */ |
| 1178 | op->watch.cookie = va_arg(args, u64); |
| 1179 | op->watch.ver = va_arg(args, u64); |
| 1180 | op->watch.ver = cpu_to_le64(op->watch.ver); |
| 1181 | if (opcode == CEPH_OSD_OP_WATCH && va_arg(args, int)) |
| 1182 | op->watch.flag = (u8) 1; |
| 1183 | break; |
Alex Elder | 8d23bf2 | 2012-11-19 22:55:21 -0600 | [diff] [blame] | 1184 | default: |
| 1185 | rbd_warn(NULL, "unsupported opcode %hu\n", opcode); |
| 1186 | kfree(op); |
| 1187 | op = NULL; |
| 1188 | break; |
| 1189 | } |
| 1190 | va_end(args); |
| 1191 | |
| 1192 | return op; |
| 1193 | } |
| 1194 | |
| 1195 | static void rbd_osd_req_op_destroy(struct ceph_osd_req_op *op) |
| 1196 | { |
| 1197 | kfree(op); |
| 1198 | } |
| 1199 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1200 | static int rbd_obj_request_submit(struct ceph_osd_client *osdc, |
| 1201 | struct rbd_obj_request *obj_request) |
| 1202 | { |
| 1203 | return ceph_osdc_start_request(osdc, obj_request->osd_req, false); |
| 1204 | } |
| 1205 | |
| 1206 | static void rbd_img_request_complete(struct rbd_img_request *img_request) |
| 1207 | { |
| 1208 | if (img_request->callback) |
| 1209 | img_request->callback(img_request); |
| 1210 | else |
| 1211 | rbd_img_request_put(img_request); |
| 1212 | } |
| 1213 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1214 | /* Caller is responsible for rbd_obj_request_destroy(obj_request) */ |
| 1215 | |
| 1216 | static int rbd_obj_request_wait(struct rbd_obj_request *obj_request) |
| 1217 | { |
| 1218 | return wait_for_completion_interruptible(&obj_request->completion); |
| 1219 | } |
| 1220 | |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1221 | static void obj_request_done_init(struct rbd_obj_request *obj_request) |
| 1222 | { |
| 1223 | atomic_set(&obj_request->done, 0); |
| 1224 | smp_wmb(); |
| 1225 | } |
| 1226 | |
| 1227 | static void obj_request_done_set(struct rbd_obj_request *obj_request) |
| 1228 | { |
| 1229 | atomic_set(&obj_request->done, 1); |
| 1230 | smp_wmb(); |
| 1231 | } |
| 1232 | |
| 1233 | static bool obj_request_done_test(struct rbd_obj_request *obj_request) |
| 1234 | { |
| 1235 | smp_rmb(); |
| 1236 | return atomic_read(&obj_request->done) != 0; |
| 1237 | } |
| 1238 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1239 | static void rbd_osd_trivial_callback(struct rbd_obj_request *obj_request, |
| 1240 | struct ceph_osd_op *op) |
| 1241 | { |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1242 | obj_request_done_set(obj_request); |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1243 | } |
| 1244 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1245 | static void rbd_obj_request_complete(struct rbd_obj_request *obj_request) |
| 1246 | { |
| 1247 | if (obj_request->callback) |
| 1248 | obj_request->callback(obj_request); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1249 | else |
| 1250 | complete_all(&obj_request->completion); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1251 | } |
| 1252 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1253 | static void rbd_osd_read_callback(struct rbd_obj_request *obj_request, |
| 1254 | struct ceph_osd_op *op) |
| 1255 | { |
| 1256 | u64 xferred; |
| 1257 | |
| 1258 | /* |
| 1259 | * We support a 64-bit length, but ultimately it has to be |
| 1260 | * passed to blk_end_request(), which takes an unsigned int. |
| 1261 | */ |
| 1262 | xferred = le64_to_cpu(op->extent.length); |
| 1263 | rbd_assert(xferred < (u64) UINT_MAX); |
| 1264 | if (obj_request->result == (s32) -ENOENT) { |
| 1265 | zero_bio_chain(obj_request->bio_list, 0); |
| 1266 | obj_request->result = 0; |
| 1267 | } else if (xferred < obj_request->length && !obj_request->result) { |
| 1268 | zero_bio_chain(obj_request->bio_list, xferred); |
| 1269 | xferred = obj_request->length; |
| 1270 | } |
| 1271 | obj_request->xferred = xferred; |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1272 | obj_request_done_set(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1273 | } |
| 1274 | |
| 1275 | static void rbd_osd_write_callback(struct rbd_obj_request *obj_request, |
| 1276 | struct ceph_osd_op *op) |
| 1277 | { |
| 1278 | obj_request->xferred = le64_to_cpu(op->extent.length); |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1279 | obj_request_done_set(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1280 | } |
| 1281 | |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1282 | /* |
| 1283 | * For a simple stat call there's nothing to do. We'll do more if |
| 1284 | * this is part of a write sequence for a layered image. |
| 1285 | */ |
| 1286 | static void rbd_osd_stat_callback(struct rbd_obj_request *obj_request, |
| 1287 | struct ceph_osd_op *op) |
| 1288 | { |
| 1289 | obj_request_done_set(obj_request); |
| 1290 | } |
| 1291 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1292 | static void rbd_osd_req_callback(struct ceph_osd_request *osd_req, |
| 1293 | struct ceph_msg *msg) |
| 1294 | { |
| 1295 | struct rbd_obj_request *obj_request = osd_req->r_priv; |
| 1296 | struct ceph_osd_reply_head *reply_head; |
| 1297 | struct ceph_osd_op *op; |
| 1298 | u32 num_ops; |
| 1299 | u16 opcode; |
| 1300 | |
| 1301 | rbd_assert(osd_req == obj_request->osd_req); |
| 1302 | rbd_assert(!!obj_request->img_request ^ |
| 1303 | (obj_request->which == BAD_WHICH)); |
| 1304 | |
| 1305 | obj_request->xferred = le32_to_cpu(msg->hdr.data_len); |
| 1306 | reply_head = msg->front.iov_base; |
| 1307 | obj_request->result = (s32) le32_to_cpu(reply_head->result); |
| 1308 | obj_request->version = le64_to_cpu(osd_req->r_reassert_version.version); |
| 1309 | |
| 1310 | num_ops = le32_to_cpu(reply_head->num_ops); |
| 1311 | WARN_ON(num_ops != 1); /* For now */ |
| 1312 | |
| 1313 | op = &reply_head->ops[0]; |
| 1314 | opcode = le16_to_cpu(op->op); |
| 1315 | switch (opcode) { |
| 1316 | case CEPH_OSD_OP_READ: |
| 1317 | rbd_osd_read_callback(obj_request, op); |
| 1318 | break; |
| 1319 | case CEPH_OSD_OP_WRITE: |
| 1320 | rbd_osd_write_callback(obj_request, op); |
| 1321 | break; |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1322 | case CEPH_OSD_OP_STAT: |
| 1323 | rbd_osd_stat_callback(obj_request, op); |
| 1324 | break; |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 1325 | case CEPH_OSD_OP_CALL: |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 1326 | case CEPH_OSD_OP_NOTIFY_ACK: |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1327 | case CEPH_OSD_OP_WATCH: |
| 1328 | rbd_osd_trivial_callback(obj_request, op); |
| 1329 | break; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1330 | default: |
| 1331 | rbd_warn(NULL, "%s: unsupported op %hu\n", |
| 1332 | obj_request->object_name, (unsigned short) opcode); |
| 1333 | break; |
| 1334 | } |
| 1335 | |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1336 | if (obj_request_done_test(obj_request)) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1337 | rbd_obj_request_complete(obj_request); |
| 1338 | } |
| 1339 | |
| 1340 | static struct ceph_osd_request *rbd_osd_req_create( |
| 1341 | struct rbd_device *rbd_dev, |
| 1342 | bool write_request, |
| 1343 | struct rbd_obj_request *obj_request, |
| 1344 | struct ceph_osd_req_op *op) |
| 1345 | { |
| 1346 | struct rbd_img_request *img_request = obj_request->img_request; |
| 1347 | struct ceph_snap_context *snapc = NULL; |
| 1348 | struct ceph_osd_client *osdc; |
| 1349 | struct ceph_osd_request *osd_req; |
| 1350 | struct timespec now; |
| 1351 | struct timespec *mtime; |
| 1352 | u64 snap_id = CEPH_NOSNAP; |
| 1353 | u64 offset = obj_request->offset; |
| 1354 | u64 length = obj_request->length; |
| 1355 | |
| 1356 | if (img_request) { |
| 1357 | rbd_assert(img_request->write_request == write_request); |
| 1358 | if (img_request->write_request) |
| 1359 | snapc = img_request->snapc; |
| 1360 | else |
| 1361 | snap_id = img_request->snap_id; |
| 1362 | } |
| 1363 | |
| 1364 | /* Allocate and initialize the request, for the single op */ |
| 1365 | |
| 1366 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 1367 | osd_req = ceph_osdc_alloc_request(osdc, snapc, 1, false, GFP_ATOMIC); |
| 1368 | if (!osd_req) |
| 1369 | return NULL; /* ENOMEM */ |
| 1370 | |
| 1371 | rbd_assert(obj_request_type_valid(obj_request->type)); |
| 1372 | switch (obj_request->type) { |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1373 | case OBJ_REQUEST_NODATA: |
| 1374 | break; /* Nothing to do */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1375 | case OBJ_REQUEST_BIO: |
| 1376 | rbd_assert(obj_request->bio_list != NULL); |
| 1377 | osd_req->r_bio = obj_request->bio_list; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1378 | break; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1379 | case OBJ_REQUEST_PAGES: |
| 1380 | osd_req->r_pages = obj_request->pages; |
| 1381 | osd_req->r_num_pages = obj_request->page_count; |
| 1382 | osd_req->r_page_alignment = offset & ~PAGE_MASK; |
| 1383 | break; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1384 | } |
| 1385 | |
| 1386 | if (write_request) { |
| 1387 | osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; |
| 1388 | now = CURRENT_TIME; |
| 1389 | mtime = &now; |
| 1390 | } else { |
| 1391 | osd_req->r_flags = CEPH_OSD_FLAG_READ; |
| 1392 | mtime = NULL; /* not needed for reads */ |
| 1393 | offset = 0; /* These are not used... */ |
| 1394 | length = 0; /* ...for osd read requests */ |
| 1395 | } |
| 1396 | |
| 1397 | osd_req->r_callback = rbd_osd_req_callback; |
| 1398 | osd_req->r_priv = obj_request; |
| 1399 | |
| 1400 | osd_req->r_oid_len = strlen(obj_request->object_name); |
| 1401 | rbd_assert(osd_req->r_oid_len < sizeof (osd_req->r_oid)); |
| 1402 | memcpy(osd_req->r_oid, obj_request->object_name, osd_req->r_oid_len); |
| 1403 | |
| 1404 | osd_req->r_file_layout = rbd_dev->layout; /* struct */ |
| 1405 | |
| 1406 | /* osd_req will get its own reference to snapc (if non-null) */ |
| 1407 | |
| 1408 | ceph_osdc_build_request(osd_req, offset, length, 1, op, |
| 1409 | snapc, snap_id, mtime); |
| 1410 | |
| 1411 | return osd_req; |
| 1412 | } |
| 1413 | |
| 1414 | static void rbd_osd_req_destroy(struct ceph_osd_request *osd_req) |
| 1415 | { |
| 1416 | ceph_osdc_put_request(osd_req); |
| 1417 | } |
| 1418 | |
| 1419 | /* object_name is assumed to be a non-null pointer and NUL-terminated */ |
| 1420 | |
| 1421 | static struct rbd_obj_request *rbd_obj_request_create(const char *object_name, |
| 1422 | u64 offset, u64 length, |
| 1423 | enum obj_request_type type) |
| 1424 | { |
| 1425 | struct rbd_obj_request *obj_request; |
| 1426 | size_t size; |
| 1427 | char *name; |
| 1428 | |
| 1429 | rbd_assert(obj_request_type_valid(type)); |
| 1430 | |
| 1431 | size = strlen(object_name) + 1; |
| 1432 | obj_request = kzalloc(sizeof (*obj_request) + size, GFP_KERNEL); |
| 1433 | if (!obj_request) |
| 1434 | return NULL; |
| 1435 | |
| 1436 | name = (char *)(obj_request + 1); |
| 1437 | obj_request->object_name = memcpy(name, object_name, size); |
| 1438 | obj_request->offset = offset; |
| 1439 | obj_request->length = length; |
| 1440 | obj_request->which = BAD_WHICH; |
| 1441 | obj_request->type = type; |
| 1442 | INIT_LIST_HEAD(&obj_request->links); |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1443 | obj_request_done_init(obj_request); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1444 | init_completion(&obj_request->completion); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1445 | kref_init(&obj_request->kref); |
| 1446 | |
| 1447 | return obj_request; |
| 1448 | } |
| 1449 | |
| 1450 | static void rbd_obj_request_destroy(struct kref *kref) |
| 1451 | { |
| 1452 | struct rbd_obj_request *obj_request; |
| 1453 | |
| 1454 | obj_request = container_of(kref, struct rbd_obj_request, kref); |
| 1455 | |
| 1456 | rbd_assert(obj_request->img_request == NULL); |
| 1457 | rbd_assert(obj_request->which == BAD_WHICH); |
| 1458 | |
| 1459 | if (obj_request->osd_req) |
| 1460 | rbd_osd_req_destroy(obj_request->osd_req); |
| 1461 | |
| 1462 | rbd_assert(obj_request_type_valid(obj_request->type)); |
| 1463 | switch (obj_request->type) { |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1464 | case OBJ_REQUEST_NODATA: |
| 1465 | break; /* Nothing to do */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1466 | case OBJ_REQUEST_BIO: |
| 1467 | if (obj_request->bio_list) |
| 1468 | bio_chain_put(obj_request->bio_list); |
| 1469 | break; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1470 | case OBJ_REQUEST_PAGES: |
| 1471 | if (obj_request->pages) |
| 1472 | ceph_release_page_vector(obj_request->pages, |
| 1473 | obj_request->page_count); |
| 1474 | break; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1475 | } |
| 1476 | |
| 1477 | kfree(obj_request); |
| 1478 | } |
| 1479 | |
| 1480 | /* |
| 1481 | * Caller is responsible for filling in the list of object requests |
| 1482 | * that comprises the image request, and the Linux request pointer |
| 1483 | * (if there is one). |
| 1484 | */ |
| 1485 | struct rbd_img_request *rbd_img_request_create(struct rbd_device *rbd_dev, |
| 1486 | u64 offset, u64 length, |
| 1487 | bool write_request) |
| 1488 | { |
| 1489 | struct rbd_img_request *img_request; |
| 1490 | struct ceph_snap_context *snapc = NULL; |
| 1491 | |
| 1492 | img_request = kmalloc(sizeof (*img_request), GFP_ATOMIC); |
| 1493 | if (!img_request) |
| 1494 | return NULL; |
| 1495 | |
| 1496 | if (write_request) { |
| 1497 | down_read(&rbd_dev->header_rwsem); |
| 1498 | snapc = ceph_get_snap_context(rbd_dev->header.snapc); |
| 1499 | up_read(&rbd_dev->header_rwsem); |
| 1500 | if (WARN_ON(!snapc)) { |
| 1501 | kfree(img_request); |
| 1502 | return NULL; /* Shouldn't happen */ |
| 1503 | } |
| 1504 | } |
| 1505 | |
| 1506 | img_request->rq = NULL; |
| 1507 | img_request->rbd_dev = rbd_dev; |
| 1508 | img_request->offset = offset; |
| 1509 | img_request->length = length; |
| 1510 | img_request->write_request = write_request; |
| 1511 | if (write_request) |
| 1512 | img_request->snapc = snapc; |
| 1513 | else |
| 1514 | img_request->snap_id = rbd_dev->spec->snap_id; |
| 1515 | spin_lock_init(&img_request->completion_lock); |
| 1516 | img_request->next_completion = 0; |
| 1517 | img_request->callback = NULL; |
| 1518 | img_request->obj_request_count = 0; |
| 1519 | INIT_LIST_HEAD(&img_request->obj_requests); |
| 1520 | kref_init(&img_request->kref); |
| 1521 | |
| 1522 | rbd_img_request_get(img_request); /* Avoid a warning */ |
| 1523 | rbd_img_request_put(img_request); /* TEMPORARY */ |
| 1524 | |
| 1525 | return img_request; |
| 1526 | } |
| 1527 | |
| 1528 | static void rbd_img_request_destroy(struct kref *kref) |
| 1529 | { |
| 1530 | struct rbd_img_request *img_request; |
| 1531 | struct rbd_obj_request *obj_request; |
| 1532 | struct rbd_obj_request *next_obj_request; |
| 1533 | |
| 1534 | img_request = container_of(kref, struct rbd_img_request, kref); |
| 1535 | |
| 1536 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) |
| 1537 | rbd_img_obj_request_del(img_request, obj_request); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1538 | rbd_assert(img_request->obj_request_count == 0); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1539 | |
| 1540 | if (img_request->write_request) |
| 1541 | ceph_put_snap_context(img_request->snapc); |
| 1542 | |
| 1543 | kfree(img_request); |
| 1544 | } |
| 1545 | |
| 1546 | static int rbd_img_request_fill_bio(struct rbd_img_request *img_request, |
| 1547 | struct bio *bio_list) |
| 1548 | { |
| 1549 | struct rbd_device *rbd_dev = img_request->rbd_dev; |
| 1550 | struct rbd_obj_request *obj_request = NULL; |
| 1551 | struct rbd_obj_request *next_obj_request; |
| 1552 | unsigned int bio_offset; |
| 1553 | u64 image_offset; |
| 1554 | u64 resid; |
| 1555 | u16 opcode; |
| 1556 | |
| 1557 | opcode = img_request->write_request ? CEPH_OSD_OP_WRITE |
| 1558 | : CEPH_OSD_OP_READ; |
| 1559 | bio_offset = 0; |
| 1560 | image_offset = img_request->offset; |
| 1561 | rbd_assert(image_offset == bio_list->bi_sector << SECTOR_SHIFT); |
| 1562 | resid = img_request->length; |
| 1563 | while (resid) { |
| 1564 | const char *object_name; |
| 1565 | unsigned int clone_size; |
| 1566 | struct ceph_osd_req_op *op; |
| 1567 | u64 offset; |
| 1568 | u64 length; |
| 1569 | |
| 1570 | object_name = rbd_segment_name(rbd_dev, image_offset); |
| 1571 | if (!object_name) |
| 1572 | goto out_unwind; |
| 1573 | offset = rbd_segment_offset(rbd_dev, image_offset); |
| 1574 | length = rbd_segment_length(rbd_dev, image_offset, resid); |
| 1575 | obj_request = rbd_obj_request_create(object_name, |
| 1576 | offset, length, |
| 1577 | OBJ_REQUEST_BIO); |
| 1578 | kfree(object_name); /* object request has its own copy */ |
| 1579 | if (!obj_request) |
| 1580 | goto out_unwind; |
| 1581 | |
| 1582 | rbd_assert(length <= (u64) UINT_MAX); |
| 1583 | clone_size = (unsigned int) length; |
| 1584 | obj_request->bio_list = bio_chain_clone_range(&bio_list, |
| 1585 | &bio_offset, clone_size, |
| 1586 | GFP_ATOMIC); |
| 1587 | if (!obj_request->bio_list) |
| 1588 | goto out_partial; |
| 1589 | |
| 1590 | /* |
| 1591 | * Build up the op to use in building the osd |
| 1592 | * request. Note that the contents of the op are |
| 1593 | * copied by rbd_osd_req_create(). |
| 1594 | */ |
| 1595 | op = rbd_osd_req_op_create(opcode, offset, length); |
| 1596 | if (!op) |
| 1597 | goto out_partial; |
| 1598 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, |
| 1599 | img_request->write_request, |
| 1600 | obj_request, op); |
| 1601 | rbd_osd_req_op_destroy(op); |
| 1602 | if (!obj_request->osd_req) |
| 1603 | goto out_partial; |
| 1604 | /* status and version are initially zero-filled */ |
| 1605 | |
| 1606 | rbd_img_obj_request_add(img_request, obj_request); |
| 1607 | |
| 1608 | image_offset += length; |
| 1609 | resid -= length; |
| 1610 | } |
| 1611 | |
| 1612 | return 0; |
| 1613 | |
| 1614 | out_partial: |
| 1615 | rbd_obj_request_put(obj_request); |
| 1616 | out_unwind: |
| 1617 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) |
| 1618 | rbd_obj_request_put(obj_request); |
| 1619 | |
| 1620 | return -ENOMEM; |
| 1621 | } |
| 1622 | |
| 1623 | static void rbd_img_obj_callback(struct rbd_obj_request *obj_request) |
| 1624 | { |
| 1625 | struct rbd_img_request *img_request; |
| 1626 | u32 which = obj_request->which; |
| 1627 | bool more = true; |
| 1628 | |
| 1629 | img_request = obj_request->img_request; |
| 1630 | rbd_assert(img_request != NULL); |
| 1631 | rbd_assert(img_request->rq != NULL); |
| 1632 | rbd_assert(which != BAD_WHICH); |
| 1633 | rbd_assert(which < img_request->obj_request_count); |
| 1634 | rbd_assert(which >= img_request->next_completion); |
| 1635 | |
| 1636 | spin_lock_irq(&img_request->completion_lock); |
| 1637 | if (which != img_request->next_completion) |
| 1638 | goto out; |
| 1639 | |
| 1640 | for_each_obj_request_from(img_request, obj_request) { |
| 1641 | unsigned int xferred; |
| 1642 | int result; |
| 1643 | |
| 1644 | rbd_assert(more); |
| 1645 | rbd_assert(which < img_request->obj_request_count); |
| 1646 | |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1647 | if (!obj_request_done_test(obj_request)) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1648 | break; |
| 1649 | |
| 1650 | rbd_assert(obj_request->xferred <= (u64) UINT_MAX); |
| 1651 | xferred = (unsigned int) obj_request->xferred; |
| 1652 | result = (int) obj_request->result; |
| 1653 | if (result) |
| 1654 | rbd_warn(NULL, "obj_request %s result %d xferred %u\n", |
| 1655 | img_request->write_request ? "write" : "read", |
| 1656 | result, xferred); |
| 1657 | |
| 1658 | more = blk_end_request(img_request->rq, result, xferred); |
| 1659 | which++; |
| 1660 | } |
| 1661 | rbd_assert(more ^ (which == img_request->obj_request_count)); |
| 1662 | img_request->next_completion = which; |
| 1663 | out: |
| 1664 | spin_unlock_irq(&img_request->completion_lock); |
| 1665 | |
| 1666 | if (!more) |
| 1667 | rbd_img_request_complete(img_request); |
| 1668 | } |
| 1669 | |
| 1670 | static int rbd_img_request_submit(struct rbd_img_request *img_request) |
| 1671 | { |
| 1672 | struct rbd_device *rbd_dev = img_request->rbd_dev; |
| 1673 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
| 1674 | struct rbd_obj_request *obj_request; |
| 1675 | |
| 1676 | for_each_obj_request(img_request, obj_request) { |
| 1677 | int ret; |
| 1678 | |
| 1679 | obj_request->callback = rbd_img_obj_callback; |
| 1680 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 1681 | if (ret) |
| 1682 | return ret; |
| 1683 | /* |
| 1684 | * The image request has its own reference to each |
| 1685 | * of its object requests, so we can safely drop the |
| 1686 | * initial one here. |
| 1687 | */ |
| 1688 | rbd_obj_request_put(obj_request); |
| 1689 | } |
| 1690 | |
| 1691 | return 0; |
| 1692 | } |
| 1693 | |
Alex Elder | cf81b60 | 2013-01-17 12:18:46 -0600 | [diff] [blame] | 1694 | static int rbd_obj_notify_ack(struct rbd_device *rbd_dev, |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 1695 | u64 ver, u64 notify_id) |
| 1696 | { |
| 1697 | struct rbd_obj_request *obj_request; |
| 1698 | struct ceph_osd_req_op *op; |
| 1699 | struct ceph_osd_client *osdc; |
| 1700 | int ret; |
| 1701 | |
| 1702 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, |
| 1703 | OBJ_REQUEST_NODATA); |
| 1704 | if (!obj_request) |
| 1705 | return -ENOMEM; |
| 1706 | |
| 1707 | ret = -ENOMEM; |
| 1708 | op = rbd_osd_req_op_create(CEPH_OSD_OP_NOTIFY_ACK, notify_id, ver); |
| 1709 | if (!op) |
| 1710 | goto out; |
| 1711 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, |
| 1712 | obj_request, op); |
| 1713 | rbd_osd_req_op_destroy(op); |
| 1714 | if (!obj_request->osd_req) |
| 1715 | goto out; |
| 1716 | |
| 1717 | osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | cf81b60 | 2013-01-17 12:18:46 -0600 | [diff] [blame] | 1718 | obj_request->callback = rbd_obj_request_put; |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 1719 | ret = rbd_obj_request_submit(osdc, obj_request); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 1720 | out: |
Alex Elder | cf81b60 | 2013-01-17 12:18:46 -0600 | [diff] [blame] | 1721 | if (ret) |
| 1722 | rbd_obj_request_put(obj_request); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 1723 | |
| 1724 | return ret; |
| 1725 | } |
| 1726 | |
| 1727 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) |
| 1728 | { |
| 1729 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
| 1730 | u64 hver; |
| 1731 | int rc; |
| 1732 | |
| 1733 | if (!rbd_dev) |
| 1734 | return; |
| 1735 | |
| 1736 | dout("rbd_watch_cb %s notify_id=%llu opcode=%u\n", |
| 1737 | rbd_dev->header_name, (unsigned long long) notify_id, |
| 1738 | (unsigned int) opcode); |
| 1739 | rc = rbd_dev_refresh(rbd_dev, &hver); |
| 1740 | if (rc) |
| 1741 | rbd_warn(rbd_dev, "got notification but failed to " |
| 1742 | " update snaps: %d\n", rc); |
| 1743 | |
Alex Elder | cf81b60 | 2013-01-17 12:18:46 -0600 | [diff] [blame] | 1744 | rbd_obj_notify_ack(rbd_dev, hver, notify_id); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 1745 | } |
| 1746 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1747 | /* |
| 1748 | * Request sync osd watch/unwatch. The value of "start" determines |
| 1749 | * whether a watch request is being initiated or torn down. |
| 1750 | */ |
| 1751 | static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev, int start) |
| 1752 | { |
| 1753 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
| 1754 | struct rbd_obj_request *obj_request; |
| 1755 | struct ceph_osd_req_op *op; |
| 1756 | int ret; |
| 1757 | |
| 1758 | rbd_assert(start ^ !!rbd_dev->watch_event); |
| 1759 | rbd_assert(start ^ !!rbd_dev->watch_request); |
| 1760 | |
| 1761 | if (start) { |
Alex Elder | 3c663bb | 2013-02-15 11:42:30 -0600 | [diff] [blame] | 1762 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, rbd_dev, |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1763 | &rbd_dev->watch_event); |
| 1764 | if (ret < 0) |
| 1765 | return ret; |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1766 | rbd_assert(rbd_dev->watch_event != NULL); |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1767 | } |
| 1768 | |
| 1769 | ret = -ENOMEM; |
| 1770 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, |
| 1771 | OBJ_REQUEST_NODATA); |
| 1772 | if (!obj_request) |
| 1773 | goto out_cancel; |
| 1774 | |
| 1775 | op = rbd_osd_req_op_create(CEPH_OSD_OP_WATCH, |
| 1776 | rbd_dev->watch_event->cookie, |
| 1777 | rbd_dev->header.obj_version, start); |
| 1778 | if (!op) |
| 1779 | goto out_cancel; |
| 1780 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, true, |
| 1781 | obj_request, op); |
| 1782 | rbd_osd_req_op_destroy(op); |
| 1783 | if (!obj_request->osd_req) |
| 1784 | goto out_cancel; |
| 1785 | |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1786 | if (start) |
Alex Elder | 975241a | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1787 | ceph_osdc_set_request_linger(osdc, obj_request->osd_req); |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1788 | else |
Alex Elder | 6977c3f | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1789 | ceph_osdc_unregister_linger_request(osdc, |
Alex Elder | 975241a | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1790 | rbd_dev->watch_request->osd_req); |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1791 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 1792 | if (ret) |
| 1793 | goto out_cancel; |
| 1794 | ret = rbd_obj_request_wait(obj_request); |
| 1795 | if (ret) |
| 1796 | goto out_cancel; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1797 | ret = obj_request->result; |
| 1798 | if (ret) |
| 1799 | goto out_cancel; |
| 1800 | |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1801 | /* |
| 1802 | * A watch request is set to linger, so the underlying osd |
| 1803 | * request won't go away until we unregister it. We retain |
| 1804 | * a pointer to the object request during that time (in |
| 1805 | * rbd_dev->watch_request), so we'll keep a reference to |
| 1806 | * it. We'll drop that reference (below) after we've |
| 1807 | * unregistered it. |
| 1808 | */ |
| 1809 | if (start) { |
| 1810 | rbd_dev->watch_request = obj_request; |
| 1811 | |
| 1812 | return 0; |
| 1813 | } |
| 1814 | |
| 1815 | /* We have successfully torn down the watch request */ |
| 1816 | |
| 1817 | rbd_obj_request_put(rbd_dev->watch_request); |
| 1818 | rbd_dev->watch_request = NULL; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1819 | out_cancel: |
| 1820 | /* Cancel the event if we're tearing down, or on error */ |
| 1821 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
| 1822 | rbd_dev->watch_event = NULL; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1823 | if (obj_request) |
| 1824 | rbd_obj_request_put(obj_request); |
| 1825 | |
| 1826 | return ret; |
| 1827 | } |
| 1828 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 1829 | /* |
| 1830 | * Synchronous osd object method call |
| 1831 | */ |
| 1832 | static int rbd_obj_method_sync(struct rbd_device *rbd_dev, |
| 1833 | const char *object_name, |
| 1834 | const char *class_name, |
| 1835 | const char *method_name, |
| 1836 | const char *outbound, |
| 1837 | size_t outbound_size, |
| 1838 | char *inbound, |
| 1839 | size_t inbound_size, |
| 1840 | u64 *version) |
| 1841 | { |
| 1842 | struct rbd_obj_request *obj_request; |
| 1843 | struct ceph_osd_client *osdc; |
| 1844 | struct ceph_osd_req_op *op; |
| 1845 | struct page **pages; |
| 1846 | u32 page_count; |
| 1847 | int ret; |
| 1848 | |
| 1849 | /* |
| 1850 | * Method calls are ultimately read operations but they |
| 1851 | * don't involve object data (so no offset or length). |
| 1852 | * The result should placed into the inbound buffer |
| 1853 | * provided. They also supply outbound data--parameters for |
| 1854 | * the object method. Currently if this is present it will |
| 1855 | * be a snapshot id. |
| 1856 | */ |
| 1857 | page_count = (u32) calc_pages_for(0, inbound_size); |
| 1858 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 1859 | if (IS_ERR(pages)) |
| 1860 | return PTR_ERR(pages); |
| 1861 | |
| 1862 | ret = -ENOMEM; |
| 1863 | obj_request = rbd_obj_request_create(object_name, 0, 0, |
| 1864 | OBJ_REQUEST_PAGES); |
| 1865 | if (!obj_request) |
| 1866 | goto out; |
| 1867 | |
| 1868 | obj_request->pages = pages; |
| 1869 | obj_request->page_count = page_count; |
| 1870 | |
| 1871 | op = rbd_osd_req_op_create(CEPH_OSD_OP_CALL, class_name, |
| 1872 | method_name, outbound, outbound_size); |
| 1873 | if (!op) |
| 1874 | goto out; |
| 1875 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, |
| 1876 | obj_request, op); |
| 1877 | rbd_osd_req_op_destroy(op); |
| 1878 | if (!obj_request->osd_req) |
| 1879 | goto out; |
| 1880 | |
| 1881 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 1882 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 1883 | if (ret) |
| 1884 | goto out; |
| 1885 | ret = rbd_obj_request_wait(obj_request); |
| 1886 | if (ret) |
| 1887 | goto out; |
| 1888 | |
| 1889 | ret = obj_request->result; |
| 1890 | if (ret < 0) |
| 1891 | goto out; |
Alex Elder | 23ed6e1 | 2013-02-06 13:11:38 -0600 | [diff] [blame^] | 1892 | ret = 0; |
| 1893 | (void) ceph_copy_from_page_vector(pages, inbound, 0, |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 1894 | obj_request->xferred); |
| 1895 | if (version) |
| 1896 | *version = obj_request->version; |
| 1897 | out: |
| 1898 | if (obj_request) |
| 1899 | rbd_obj_request_put(obj_request); |
| 1900 | else |
| 1901 | ceph_release_page_vector(pages, page_count); |
| 1902 | |
| 1903 | return ret; |
| 1904 | } |
| 1905 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1906 | static void rbd_request_fn(struct request_queue *q) |
| 1907 | { |
| 1908 | struct rbd_device *rbd_dev = q->queuedata; |
| 1909 | bool read_only = rbd_dev->mapping.read_only; |
| 1910 | struct request *rq; |
| 1911 | int result; |
| 1912 | |
| 1913 | while ((rq = blk_fetch_request(q))) { |
| 1914 | bool write_request = rq_data_dir(rq) == WRITE; |
| 1915 | struct rbd_img_request *img_request; |
| 1916 | u64 offset; |
| 1917 | u64 length; |
| 1918 | |
| 1919 | /* Ignore any non-FS requests that filter through. */ |
| 1920 | |
| 1921 | if (rq->cmd_type != REQ_TYPE_FS) { |
| 1922 | __blk_end_request_all(rq, 0); |
| 1923 | continue; |
| 1924 | } |
| 1925 | |
| 1926 | spin_unlock_irq(q->queue_lock); |
| 1927 | |
| 1928 | /* Disallow writes to a read-only device */ |
| 1929 | |
| 1930 | if (write_request) { |
| 1931 | result = -EROFS; |
| 1932 | if (read_only) |
| 1933 | goto end_request; |
| 1934 | rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP); |
| 1935 | } |
| 1936 | |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 1937 | /* |
| 1938 | * Quit early if the mapped snapshot no longer |
| 1939 | * exists. It's still possible the snapshot will |
| 1940 | * have disappeared by the time our request arrives |
| 1941 | * at the osd, but there's no sense in sending it if |
| 1942 | * we already know. |
| 1943 | */ |
| 1944 | if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1945 | dout("request for non-existent snapshot"); |
| 1946 | rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); |
| 1947 | result = -ENXIO; |
| 1948 | goto end_request; |
| 1949 | } |
| 1950 | |
| 1951 | offset = (u64) blk_rq_pos(rq) << SECTOR_SHIFT; |
| 1952 | length = (u64) blk_rq_bytes(rq); |
| 1953 | |
| 1954 | result = -EINVAL; |
| 1955 | if (WARN_ON(offset && length > U64_MAX - offset + 1)) |
| 1956 | goto end_request; /* Shouldn't happen */ |
| 1957 | |
| 1958 | result = -ENOMEM; |
| 1959 | img_request = rbd_img_request_create(rbd_dev, offset, length, |
| 1960 | write_request); |
| 1961 | if (!img_request) |
| 1962 | goto end_request; |
| 1963 | |
| 1964 | img_request->rq = rq; |
| 1965 | |
| 1966 | result = rbd_img_request_fill_bio(img_request, rq->bio); |
| 1967 | if (!result) |
| 1968 | result = rbd_img_request_submit(img_request); |
| 1969 | if (result) |
| 1970 | rbd_img_request_put(img_request); |
| 1971 | end_request: |
| 1972 | spin_lock_irq(q->queue_lock); |
| 1973 | if (result < 0) { |
| 1974 | rbd_warn(rbd_dev, "obj_request %s result %d\n", |
| 1975 | write_request ? "write" : "read", result); |
| 1976 | __blk_end_request_all(rq, result); |
| 1977 | } |
| 1978 | } |
| 1979 | } |
| 1980 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1981 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1982 | * a queue callback. Makes sure that we don't create a bio that spans across |
| 1983 | * multiple osd objects. One exception would be with a single page bios, |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1984 | * which we handle later at bio_chain_clone_range() |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1985 | */ |
| 1986 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, |
| 1987 | struct bio_vec *bvec) |
| 1988 | { |
| 1989 | struct rbd_device *rbd_dev = q->queuedata; |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1990 | sector_t sector_offset; |
| 1991 | sector_t sectors_per_obj; |
| 1992 | sector_t obj_sector_offset; |
| 1993 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1994 | |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1995 | /* |
| 1996 | * Find how far into its rbd object the partition-relative |
| 1997 | * bio start sector is to offset relative to the enclosing |
| 1998 | * device. |
| 1999 | */ |
| 2000 | sector_offset = get_start_sect(bmd->bi_bdev) + bmd->bi_sector; |
| 2001 | sectors_per_obj = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); |
| 2002 | obj_sector_offset = sector_offset & (sectors_per_obj - 1); |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2003 | |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 2004 | /* |
| 2005 | * Compute the number of bytes from that offset to the end |
| 2006 | * of the object. Account for what's already used by the bio. |
| 2007 | */ |
| 2008 | ret = (int) (sectors_per_obj - obj_sector_offset) << SECTOR_SHIFT; |
| 2009 | if (ret > bmd->bi_size) |
| 2010 | ret -= bmd->bi_size; |
| 2011 | else |
| 2012 | ret = 0; |
| 2013 | |
| 2014 | /* |
| 2015 | * Don't send back more than was asked for. And if the bio |
| 2016 | * was empty, let the whole thing through because: "Note |
| 2017 | * that a block device *must* allow a single page to be |
| 2018 | * added to an empty bio." |
| 2019 | */ |
| 2020 | rbd_assert(bvec->bv_len <= PAGE_SIZE); |
| 2021 | if (ret > (int) bvec->bv_len || !bmd->bi_size) |
| 2022 | ret = (int) bvec->bv_len; |
| 2023 | |
| 2024 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2025 | } |
| 2026 | |
| 2027 | static void rbd_free_disk(struct rbd_device *rbd_dev) |
| 2028 | { |
| 2029 | struct gendisk *disk = rbd_dev->disk; |
| 2030 | |
| 2031 | if (!disk) |
| 2032 | return; |
| 2033 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2034 | if (disk->flags & GENHD_FL_UP) |
| 2035 | del_gendisk(disk); |
| 2036 | if (disk->queue) |
| 2037 | blk_cleanup_queue(disk->queue); |
| 2038 | put_disk(disk); |
| 2039 | } |
| 2040 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 2041 | static int rbd_obj_read_sync(struct rbd_device *rbd_dev, |
| 2042 | const char *object_name, |
| 2043 | u64 offset, u64 length, |
| 2044 | char *buf, u64 *version) |
| 2045 | |
| 2046 | { |
| 2047 | struct ceph_osd_req_op *op; |
| 2048 | struct rbd_obj_request *obj_request; |
| 2049 | struct ceph_osd_client *osdc; |
| 2050 | struct page **pages = NULL; |
| 2051 | u32 page_count; |
Alex Elder | 1ceae7e | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 2052 | size_t size; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 2053 | int ret; |
| 2054 | |
| 2055 | page_count = (u32) calc_pages_for(offset, length); |
| 2056 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 2057 | if (IS_ERR(pages)) |
| 2058 | ret = PTR_ERR(pages); |
| 2059 | |
| 2060 | ret = -ENOMEM; |
| 2061 | obj_request = rbd_obj_request_create(object_name, offset, length, |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2062 | OBJ_REQUEST_PAGES); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 2063 | if (!obj_request) |
| 2064 | goto out; |
| 2065 | |
| 2066 | obj_request->pages = pages; |
| 2067 | obj_request->page_count = page_count; |
| 2068 | |
| 2069 | op = rbd_osd_req_op_create(CEPH_OSD_OP_READ, offset, length); |
| 2070 | if (!op) |
| 2071 | goto out; |
| 2072 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, |
| 2073 | obj_request, op); |
| 2074 | rbd_osd_req_op_destroy(op); |
| 2075 | if (!obj_request->osd_req) |
| 2076 | goto out; |
| 2077 | |
| 2078 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2079 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 2080 | if (ret) |
| 2081 | goto out; |
| 2082 | ret = rbd_obj_request_wait(obj_request); |
| 2083 | if (ret) |
| 2084 | goto out; |
| 2085 | |
| 2086 | ret = obj_request->result; |
| 2087 | if (ret < 0) |
| 2088 | goto out; |
Alex Elder | 1ceae7e | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 2089 | |
| 2090 | rbd_assert(obj_request->xferred <= (u64) SIZE_MAX); |
| 2091 | size = (size_t) obj_request->xferred; |
Alex Elder | 23ed6e1 | 2013-02-06 13:11:38 -0600 | [diff] [blame^] | 2092 | (void) ceph_copy_from_page_vector(pages, buf, 0, size); |
| 2093 | rbd_assert(size <= (size_t) INT_MAX); |
| 2094 | ret = (int) size; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 2095 | if (version) |
| 2096 | *version = obj_request->version; |
| 2097 | out: |
| 2098 | if (obj_request) |
| 2099 | rbd_obj_request_put(obj_request); |
| 2100 | else |
| 2101 | ceph_release_page_vector(pages, page_count); |
| 2102 | |
| 2103 | return ret; |
| 2104 | } |
| 2105 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2106 | /* |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 2107 | * Read the complete header for the given rbd device. |
| 2108 | * |
| 2109 | * Returns a pointer to a dynamically-allocated buffer containing |
| 2110 | * the complete and validated header. Caller can pass the address |
| 2111 | * of a variable that will be filled in with the version of the |
| 2112 | * header object at the time it was read. |
| 2113 | * |
| 2114 | * Returns a pointer-coded errno if a failure occurs. |
| 2115 | */ |
| 2116 | static struct rbd_image_header_ondisk * |
| 2117 | rbd_dev_v1_header_read(struct rbd_device *rbd_dev, u64 *version) |
| 2118 | { |
| 2119 | struct rbd_image_header_ondisk *ondisk = NULL; |
| 2120 | u32 snap_count = 0; |
| 2121 | u64 names_size = 0; |
| 2122 | u32 want_count; |
| 2123 | int ret; |
| 2124 | |
| 2125 | /* |
| 2126 | * The complete header will include an array of its 64-bit |
| 2127 | * snapshot ids, followed by the names of those snapshots as |
| 2128 | * a contiguous block of NUL-terminated strings. Note that |
| 2129 | * the number of snapshots could change by the time we read |
| 2130 | * it in, in which case we re-read it. |
| 2131 | */ |
| 2132 | do { |
| 2133 | size_t size; |
| 2134 | |
| 2135 | kfree(ondisk); |
| 2136 | |
| 2137 | size = sizeof (*ondisk); |
| 2138 | size += snap_count * sizeof (struct rbd_image_snap_ondisk); |
| 2139 | size += names_size; |
| 2140 | ondisk = kmalloc(size, GFP_KERNEL); |
| 2141 | if (!ondisk) |
| 2142 | return ERR_PTR(-ENOMEM); |
| 2143 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 2144 | ret = rbd_obj_read_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 2145 | 0, size, |
| 2146 | (char *) ondisk, version); |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 2147 | if (ret < 0) |
| 2148 | goto out_err; |
| 2149 | if (WARN_ON((size_t) ret < size)) { |
| 2150 | ret = -ENXIO; |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 2151 | rbd_warn(rbd_dev, "short header read (want %zd got %d)", |
| 2152 | size, ret); |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 2153 | goto out_err; |
| 2154 | } |
| 2155 | if (!rbd_dev_ondisk_valid(ondisk)) { |
| 2156 | ret = -ENXIO; |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 2157 | rbd_warn(rbd_dev, "invalid header"); |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 2158 | goto out_err; |
| 2159 | } |
| 2160 | |
| 2161 | names_size = le64_to_cpu(ondisk->snap_names_len); |
| 2162 | want_count = snap_count; |
| 2163 | snap_count = le32_to_cpu(ondisk->snap_count); |
| 2164 | } while (snap_count != want_count); |
| 2165 | |
| 2166 | return ondisk; |
| 2167 | |
| 2168 | out_err: |
| 2169 | kfree(ondisk); |
| 2170 | |
| 2171 | return ERR_PTR(ret); |
| 2172 | } |
| 2173 | |
| 2174 | /* |
| 2175 | * reload the ondisk the header |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2176 | */ |
| 2177 | static int rbd_read_header(struct rbd_device *rbd_dev, |
| 2178 | struct rbd_image_header *header) |
| 2179 | { |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 2180 | struct rbd_image_header_ondisk *ondisk; |
| 2181 | u64 ver = 0; |
| 2182 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2183 | |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 2184 | ondisk = rbd_dev_v1_header_read(rbd_dev, &ver); |
| 2185 | if (IS_ERR(ondisk)) |
| 2186 | return PTR_ERR(ondisk); |
| 2187 | ret = rbd_header_from_disk(header, ondisk); |
| 2188 | if (ret >= 0) |
| 2189 | header->obj_version = ver; |
| 2190 | kfree(ondisk); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2191 | |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 2192 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2193 | } |
| 2194 | |
Alex Elder | 41f38c2 | 2012-10-25 23:34:40 -0500 | [diff] [blame] | 2195 | static void rbd_remove_all_snaps(struct rbd_device *rbd_dev) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2196 | { |
| 2197 | struct rbd_snap *snap; |
Alex Elder | a059329 | 2012-07-19 09:09:27 -0500 | [diff] [blame] | 2198 | struct rbd_snap *next; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2199 | |
Alex Elder | a059329 | 2012-07-19 09:09:27 -0500 | [diff] [blame] | 2200 | list_for_each_entry_safe(snap, next, &rbd_dev->snaps, node) |
Alex Elder | 41f38c2 | 2012-10-25 23:34:40 -0500 | [diff] [blame] | 2201 | rbd_remove_snap_dev(snap); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2202 | } |
| 2203 | |
Alex Elder | 9478554 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 2204 | static void rbd_update_mapping_size(struct rbd_device *rbd_dev) |
| 2205 | { |
| 2206 | sector_t size; |
| 2207 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 2208 | if (rbd_dev->spec->snap_id != CEPH_NOSNAP) |
Alex Elder | 9478554 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 2209 | return; |
| 2210 | |
| 2211 | size = (sector_t) rbd_dev->header.image_size / SECTOR_SIZE; |
| 2212 | dout("setting size to %llu sectors", (unsigned long long) size); |
| 2213 | rbd_dev->mapping.size = (u64) size; |
| 2214 | set_capacity(rbd_dev->disk, size); |
| 2215 | } |
| 2216 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2217 | /* |
| 2218 | * only read the first part of the ondisk header, without the snaps info |
| 2219 | */ |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 2220 | static int rbd_dev_v1_refresh(struct rbd_device *rbd_dev, u64 *hver) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2221 | { |
| 2222 | int ret; |
| 2223 | struct rbd_image_header h; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2224 | |
| 2225 | ret = rbd_read_header(rbd_dev, &h); |
| 2226 | if (ret < 0) |
| 2227 | return ret; |
| 2228 | |
Josh Durgin | a51aa0c | 2011-12-05 10:35:04 -0800 | [diff] [blame] | 2229 | down_write(&rbd_dev->header_rwsem); |
| 2230 | |
Alex Elder | 9478554 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 2231 | /* Update image size, and check for resize of mapped image */ |
| 2232 | rbd_dev->header.image_size = h.image_size; |
| 2233 | rbd_update_mapping_size(rbd_dev); |
Sage Weil | 9db4b3e | 2011-04-19 22:49:06 -0700 | [diff] [blame] | 2234 | |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 2235 | /* rbd_dev->header.object_prefix shouldn't change */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2236 | kfree(rbd_dev->header.snap_sizes); |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 2237 | kfree(rbd_dev->header.snap_names); |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 2238 | /* osd requests may still refer to snapc */ |
| 2239 | ceph_put_snap_context(rbd_dev->header.snapc); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2240 | |
Alex Elder | b813623 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 2241 | if (hver) |
| 2242 | *hver = h.obj_version; |
Josh Durgin | a71b891 | 2011-12-05 18:10:44 -0800 | [diff] [blame] | 2243 | rbd_dev->header.obj_version = h.obj_version; |
Josh Durgin | 93a24e0 | 2011-12-05 10:41:28 -0800 | [diff] [blame] | 2244 | rbd_dev->header.image_size = h.image_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2245 | rbd_dev->header.snapc = h.snapc; |
| 2246 | rbd_dev->header.snap_names = h.snap_names; |
| 2247 | rbd_dev->header.snap_sizes = h.snap_sizes; |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 2248 | /* Free the extra copy of the object prefix */ |
| 2249 | WARN_ON(strcmp(rbd_dev->header.object_prefix, h.object_prefix)); |
| 2250 | kfree(h.object_prefix); |
| 2251 | |
Alex Elder | 304f680 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 2252 | ret = rbd_dev_snaps_update(rbd_dev); |
| 2253 | if (!ret) |
| 2254 | ret = rbd_dev_snaps_register(rbd_dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2255 | |
Josh Durgin | c666601 | 2011-11-21 17:11:12 -0800 | [diff] [blame] | 2256 | up_write(&rbd_dev->header_rwsem); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2257 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2258 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2259 | } |
| 2260 | |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 2261 | static int rbd_dev_refresh(struct rbd_device *rbd_dev, u64 *hver) |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 2262 | { |
| 2263 | int ret; |
| 2264 | |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 2265 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 2266 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 2267 | if (rbd_dev->image_format == 1) |
| 2268 | ret = rbd_dev_v1_refresh(rbd_dev, hver); |
| 2269 | else |
| 2270 | ret = rbd_dev_v2_refresh(rbd_dev, hver); |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 2271 | mutex_unlock(&ctl_mutex); |
| 2272 | |
| 2273 | return ret; |
| 2274 | } |
| 2275 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2276 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
| 2277 | { |
| 2278 | struct gendisk *disk; |
| 2279 | struct request_queue *q; |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2280 | u64 segment_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2281 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2282 | /* create gendisk info */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2283 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); |
| 2284 | if (!disk) |
Alex Elder | 1fcdb8a | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 2285 | return -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2286 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2287 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2288 | rbd_dev->dev_id); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2289 | disk->major = rbd_dev->major; |
| 2290 | disk->first_minor = 0; |
| 2291 | disk->fops = &rbd_bd_ops; |
| 2292 | disk->private_data = rbd_dev; |
| 2293 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2294 | q = blk_init_queue(rbd_request_fn, &rbd_dev->lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2295 | if (!q) |
| 2296 | goto out_disk; |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 2297 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2298 | /* We use the default size, but let's be explicit about it. */ |
| 2299 | blk_queue_physical_block_size(q, SECTOR_SIZE); |
| 2300 | |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 2301 | /* set io sizes to object size */ |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2302 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
| 2303 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); |
| 2304 | blk_queue_max_segment_size(q, segment_size); |
| 2305 | blk_queue_io_min(q, segment_size); |
| 2306 | blk_queue_io_opt(q, segment_size); |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 2307 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2308 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
| 2309 | disk->queue = q; |
| 2310 | |
| 2311 | q->queuedata = rbd_dev; |
| 2312 | |
| 2313 | rbd_dev->disk = disk; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2314 | |
Alex Elder | 12f0294 | 2012-08-29 17:11:07 -0500 | [diff] [blame] | 2315 | set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); |
| 2316 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2317 | return 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2318 | out_disk: |
| 2319 | put_disk(disk); |
Alex Elder | 1fcdb8a | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 2320 | |
| 2321 | return -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2322 | } |
| 2323 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2324 | /* |
| 2325 | sysfs |
| 2326 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2327 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2328 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
| 2329 | { |
| 2330 | return container_of(dev, struct rbd_device, dev); |
| 2331 | } |
| 2332 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2333 | static ssize_t rbd_size_show(struct device *dev, |
| 2334 | struct device_attribute *attr, char *buf) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2335 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2336 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Josh Durgin | a51aa0c | 2011-12-05 10:35:04 -0800 | [diff] [blame] | 2337 | sector_t size; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2338 | |
Josh Durgin | a51aa0c | 2011-12-05 10:35:04 -0800 | [diff] [blame] | 2339 | down_read(&rbd_dev->header_rwsem); |
| 2340 | size = get_capacity(rbd_dev->disk); |
| 2341 | up_read(&rbd_dev->header_rwsem); |
| 2342 | |
| 2343 | return sprintf(buf, "%llu\n", (unsigned long long) size * SECTOR_SIZE); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2344 | } |
| 2345 | |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2346 | /* |
| 2347 | * Note this shows the features for whatever's mapped, which is not |
| 2348 | * necessarily the base image. |
| 2349 | */ |
| 2350 | static ssize_t rbd_features_show(struct device *dev, |
| 2351 | struct device_attribute *attr, char *buf) |
| 2352 | { |
| 2353 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 2354 | |
| 2355 | return sprintf(buf, "0x%016llx\n", |
| 2356 | (unsigned long long) rbd_dev->mapping.features); |
| 2357 | } |
| 2358 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2359 | static ssize_t rbd_major_show(struct device *dev, |
| 2360 | struct device_attribute *attr, char *buf) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2361 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2362 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2363 | |
| 2364 | return sprintf(buf, "%d\n", rbd_dev->major); |
| 2365 | } |
| 2366 | |
| 2367 | static ssize_t rbd_client_id_show(struct device *dev, |
| 2368 | struct device_attribute *attr, char *buf) |
| 2369 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2370 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2371 | |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 2372 | return sprintf(buf, "client%lld\n", |
| 2373 | ceph_client_id(rbd_dev->rbd_client->client)); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2374 | } |
| 2375 | |
| 2376 | static ssize_t rbd_pool_show(struct device *dev, |
| 2377 | struct device_attribute *attr, char *buf) |
| 2378 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2379 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2380 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 2381 | return sprintf(buf, "%s\n", rbd_dev->spec->pool_name); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2382 | } |
| 2383 | |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2384 | static ssize_t rbd_pool_id_show(struct device *dev, |
| 2385 | struct device_attribute *attr, char *buf) |
| 2386 | { |
| 2387 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 2388 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 2389 | return sprintf(buf, "%llu\n", |
| 2390 | (unsigned long long) rbd_dev->spec->pool_id); |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2391 | } |
| 2392 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2393 | static ssize_t rbd_name_show(struct device *dev, |
| 2394 | struct device_attribute *attr, char *buf) |
| 2395 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2396 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2397 | |
Alex Elder | a92ffdf | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 2398 | if (rbd_dev->spec->image_name) |
| 2399 | return sprintf(buf, "%s\n", rbd_dev->spec->image_name); |
| 2400 | |
| 2401 | return sprintf(buf, "(unknown)\n"); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2402 | } |
| 2403 | |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 2404 | static ssize_t rbd_image_id_show(struct device *dev, |
| 2405 | struct device_attribute *attr, char *buf) |
| 2406 | { |
| 2407 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 2408 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 2409 | return sprintf(buf, "%s\n", rbd_dev->spec->image_id); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 2410 | } |
| 2411 | |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2412 | /* |
| 2413 | * Shows the name of the currently-mapped snapshot (or |
| 2414 | * RBD_SNAP_HEAD_NAME for the base image). |
| 2415 | */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2416 | static ssize_t rbd_snap_show(struct device *dev, |
| 2417 | struct device_attribute *attr, |
| 2418 | char *buf) |
| 2419 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2420 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2421 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 2422 | return sprintf(buf, "%s\n", rbd_dev->spec->snap_name); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2423 | } |
| 2424 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2425 | /* |
| 2426 | * For an rbd v2 image, shows the pool id, image id, and snapshot id |
| 2427 | * for the parent image. If there is no parent, simply shows |
| 2428 | * "(no parent image)". |
| 2429 | */ |
| 2430 | static ssize_t rbd_parent_show(struct device *dev, |
| 2431 | struct device_attribute *attr, |
| 2432 | char *buf) |
| 2433 | { |
| 2434 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 2435 | struct rbd_spec *spec = rbd_dev->parent_spec; |
| 2436 | int count; |
| 2437 | char *bufp = buf; |
| 2438 | |
| 2439 | if (!spec) |
| 2440 | return sprintf(buf, "(no parent image)\n"); |
| 2441 | |
| 2442 | count = sprintf(bufp, "pool_id %llu\npool_name %s\n", |
| 2443 | (unsigned long long) spec->pool_id, spec->pool_name); |
| 2444 | if (count < 0) |
| 2445 | return count; |
| 2446 | bufp += count; |
| 2447 | |
| 2448 | count = sprintf(bufp, "image_id %s\nimage_name %s\n", spec->image_id, |
| 2449 | spec->image_name ? spec->image_name : "(unknown)"); |
| 2450 | if (count < 0) |
| 2451 | return count; |
| 2452 | bufp += count; |
| 2453 | |
| 2454 | count = sprintf(bufp, "snap_id %llu\nsnap_name %s\n", |
| 2455 | (unsigned long long) spec->snap_id, spec->snap_name); |
| 2456 | if (count < 0) |
| 2457 | return count; |
| 2458 | bufp += count; |
| 2459 | |
| 2460 | count = sprintf(bufp, "overlap %llu\n", rbd_dev->parent_overlap); |
| 2461 | if (count < 0) |
| 2462 | return count; |
| 2463 | bufp += count; |
| 2464 | |
| 2465 | return (ssize_t) (bufp - buf); |
| 2466 | } |
| 2467 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2468 | static ssize_t rbd_image_refresh(struct device *dev, |
| 2469 | struct device_attribute *attr, |
| 2470 | const char *buf, |
| 2471 | size_t size) |
| 2472 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2473 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Alex Elder | b813623 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 2474 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2475 | |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 2476 | ret = rbd_dev_refresh(rbd_dev, NULL); |
Alex Elder | b813623 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 2477 | |
| 2478 | return ret < 0 ? ret : size; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2479 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2480 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2481 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2482 | static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2483 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); |
| 2484 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); |
| 2485 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2486 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2487 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 2488 | static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2489 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); |
| 2490 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2491 | static DEVICE_ATTR(parent, S_IRUGO, rbd_parent_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2492 | |
| 2493 | static struct attribute *rbd_attrs[] = { |
| 2494 | &dev_attr_size.attr, |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2495 | &dev_attr_features.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2496 | &dev_attr_major.attr, |
| 2497 | &dev_attr_client_id.attr, |
| 2498 | &dev_attr_pool.attr, |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2499 | &dev_attr_pool_id.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2500 | &dev_attr_name.attr, |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 2501 | &dev_attr_image_id.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2502 | &dev_attr_current_snap.attr, |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2503 | &dev_attr_parent.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2504 | &dev_attr_refresh.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2505 | NULL |
| 2506 | }; |
| 2507 | |
| 2508 | static struct attribute_group rbd_attr_group = { |
| 2509 | .attrs = rbd_attrs, |
| 2510 | }; |
| 2511 | |
| 2512 | static const struct attribute_group *rbd_attr_groups[] = { |
| 2513 | &rbd_attr_group, |
| 2514 | NULL |
| 2515 | }; |
| 2516 | |
| 2517 | static void rbd_sysfs_dev_release(struct device *dev) |
| 2518 | { |
| 2519 | } |
| 2520 | |
| 2521 | static struct device_type rbd_device_type = { |
| 2522 | .name = "rbd", |
| 2523 | .groups = rbd_attr_groups, |
| 2524 | .release = rbd_sysfs_dev_release, |
| 2525 | }; |
| 2526 | |
| 2527 | |
| 2528 | /* |
| 2529 | sysfs - snapshots |
| 2530 | */ |
| 2531 | |
| 2532 | static ssize_t rbd_snap_size_show(struct device *dev, |
| 2533 | struct device_attribute *attr, |
| 2534 | char *buf) |
| 2535 | { |
| 2536 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); |
| 2537 | |
Josh Durgin | 3591538 | 2011-12-05 18:25:13 -0800 | [diff] [blame] | 2538 | return sprintf(buf, "%llu\n", (unsigned long long)snap->size); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2539 | } |
| 2540 | |
| 2541 | static ssize_t rbd_snap_id_show(struct device *dev, |
| 2542 | struct device_attribute *attr, |
| 2543 | char *buf) |
| 2544 | { |
| 2545 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); |
| 2546 | |
Josh Durgin | 3591538 | 2011-12-05 18:25:13 -0800 | [diff] [blame] | 2547 | return sprintf(buf, "%llu\n", (unsigned long long)snap->id); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2548 | } |
| 2549 | |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2550 | static ssize_t rbd_snap_features_show(struct device *dev, |
| 2551 | struct device_attribute *attr, |
| 2552 | char *buf) |
| 2553 | { |
| 2554 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); |
| 2555 | |
| 2556 | return sprintf(buf, "0x%016llx\n", |
| 2557 | (unsigned long long) snap->features); |
| 2558 | } |
| 2559 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2560 | static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); |
| 2561 | static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2562 | static DEVICE_ATTR(snap_features, S_IRUGO, rbd_snap_features_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2563 | |
| 2564 | static struct attribute *rbd_snap_attrs[] = { |
| 2565 | &dev_attr_snap_size.attr, |
| 2566 | &dev_attr_snap_id.attr, |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2567 | &dev_attr_snap_features.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2568 | NULL, |
| 2569 | }; |
| 2570 | |
| 2571 | static struct attribute_group rbd_snap_attr_group = { |
| 2572 | .attrs = rbd_snap_attrs, |
| 2573 | }; |
| 2574 | |
| 2575 | static void rbd_snap_dev_release(struct device *dev) |
| 2576 | { |
| 2577 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); |
| 2578 | kfree(snap->name); |
| 2579 | kfree(snap); |
| 2580 | } |
| 2581 | |
| 2582 | static const struct attribute_group *rbd_snap_attr_groups[] = { |
| 2583 | &rbd_snap_attr_group, |
| 2584 | NULL |
| 2585 | }; |
| 2586 | |
| 2587 | static struct device_type rbd_snap_device_type = { |
| 2588 | .groups = rbd_snap_attr_groups, |
| 2589 | .release = rbd_snap_dev_release, |
| 2590 | }; |
| 2591 | |
Alex Elder | 8b8fb99 | 2012-10-26 17:25:24 -0500 | [diff] [blame] | 2592 | static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec) |
| 2593 | { |
| 2594 | kref_get(&spec->kref); |
| 2595 | |
| 2596 | return spec; |
| 2597 | } |
| 2598 | |
| 2599 | static void rbd_spec_free(struct kref *kref); |
| 2600 | static void rbd_spec_put(struct rbd_spec *spec) |
| 2601 | { |
| 2602 | if (spec) |
| 2603 | kref_put(&spec->kref, rbd_spec_free); |
| 2604 | } |
| 2605 | |
| 2606 | static struct rbd_spec *rbd_spec_alloc(void) |
| 2607 | { |
| 2608 | struct rbd_spec *spec; |
| 2609 | |
| 2610 | spec = kzalloc(sizeof (*spec), GFP_KERNEL); |
| 2611 | if (!spec) |
| 2612 | return NULL; |
| 2613 | kref_init(&spec->kref); |
| 2614 | |
| 2615 | rbd_spec_put(rbd_spec_get(spec)); /* TEMPORARY */ |
| 2616 | |
| 2617 | return spec; |
| 2618 | } |
| 2619 | |
| 2620 | static void rbd_spec_free(struct kref *kref) |
| 2621 | { |
| 2622 | struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref); |
| 2623 | |
| 2624 | kfree(spec->pool_name); |
| 2625 | kfree(spec->image_id); |
| 2626 | kfree(spec->image_name); |
| 2627 | kfree(spec->snap_name); |
| 2628 | kfree(spec); |
| 2629 | } |
| 2630 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2631 | struct rbd_device *rbd_dev_create(struct rbd_client *rbdc, |
| 2632 | struct rbd_spec *spec) |
| 2633 | { |
| 2634 | struct rbd_device *rbd_dev; |
| 2635 | |
| 2636 | rbd_dev = kzalloc(sizeof (*rbd_dev), GFP_KERNEL); |
| 2637 | if (!rbd_dev) |
| 2638 | return NULL; |
| 2639 | |
| 2640 | spin_lock_init(&rbd_dev->lock); |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 2641 | rbd_dev->flags = 0; |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2642 | INIT_LIST_HEAD(&rbd_dev->node); |
| 2643 | INIT_LIST_HEAD(&rbd_dev->snaps); |
| 2644 | init_rwsem(&rbd_dev->header_rwsem); |
| 2645 | |
| 2646 | rbd_dev->spec = spec; |
| 2647 | rbd_dev->rbd_client = rbdc; |
| 2648 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 2649 | /* Initialize the layout used for all rbd requests */ |
| 2650 | |
| 2651 | rbd_dev->layout.fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); |
| 2652 | rbd_dev->layout.fl_stripe_count = cpu_to_le32(1); |
| 2653 | rbd_dev->layout.fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); |
| 2654 | rbd_dev->layout.fl_pg_pool = cpu_to_le32((u32) spec->pool_id); |
| 2655 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2656 | return rbd_dev; |
| 2657 | } |
| 2658 | |
| 2659 | static void rbd_dev_destroy(struct rbd_device *rbd_dev) |
| 2660 | { |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2661 | rbd_spec_put(rbd_dev->parent_spec); |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2662 | kfree(rbd_dev->header_name); |
| 2663 | rbd_put_client(rbd_dev->rbd_client); |
| 2664 | rbd_spec_put(rbd_dev->spec); |
| 2665 | kfree(rbd_dev); |
| 2666 | } |
| 2667 | |
Alex Elder | 304f680 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 2668 | static bool rbd_snap_registered(struct rbd_snap *snap) |
| 2669 | { |
| 2670 | bool ret = snap->dev.type == &rbd_snap_device_type; |
| 2671 | bool reg = device_is_registered(&snap->dev); |
| 2672 | |
| 2673 | rbd_assert(!ret ^ reg); |
| 2674 | |
| 2675 | return ret; |
| 2676 | } |
| 2677 | |
Alex Elder | 41f38c2 | 2012-10-25 23:34:40 -0500 | [diff] [blame] | 2678 | static void rbd_remove_snap_dev(struct rbd_snap *snap) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2679 | { |
| 2680 | list_del(&snap->node); |
Alex Elder | 304f680 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 2681 | if (device_is_registered(&snap->dev)) |
| 2682 | device_unregister(&snap->dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2683 | } |
| 2684 | |
Alex Elder | 14e7085 | 2012-07-19 09:09:27 -0500 | [diff] [blame] | 2685 | static int rbd_register_snap_dev(struct rbd_snap *snap, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2686 | struct device *parent) |
| 2687 | { |
| 2688 | struct device *dev = &snap->dev; |
| 2689 | int ret; |
| 2690 | |
| 2691 | dev->type = &rbd_snap_device_type; |
| 2692 | dev->parent = parent; |
| 2693 | dev->release = rbd_snap_dev_release; |
Alex Elder | d4b125e | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2694 | dev_set_name(dev, "%s%s", RBD_SNAP_DEV_NAME_PREFIX, snap->name); |
Alex Elder | 304f680 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 2695 | dout("%s: registering device for snapshot %s\n", __func__, snap->name); |
| 2696 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2697 | ret = device_register(dev); |
| 2698 | |
| 2699 | return ret; |
| 2700 | } |
| 2701 | |
Alex Elder | 4e891e0 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 2702 | static struct rbd_snap *__rbd_add_snap_dev(struct rbd_device *rbd_dev, |
Alex Elder | c8d1842 | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 2703 | const char *snap_name, |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2704 | u64 snap_id, u64 snap_size, |
| 2705 | u64 snap_features) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2706 | { |
Alex Elder | 4e891e0 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 2707 | struct rbd_snap *snap; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2708 | int ret; |
Alex Elder | 4e891e0 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 2709 | |
| 2710 | snap = kzalloc(sizeof (*snap), GFP_KERNEL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2711 | if (!snap) |
Alex Elder | 4e891e0 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 2712 | return ERR_PTR(-ENOMEM); |
| 2713 | |
| 2714 | ret = -ENOMEM; |
Alex Elder | c8d1842 | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 2715 | snap->name = kstrdup(snap_name, GFP_KERNEL); |
Alex Elder | 4e891e0 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 2716 | if (!snap->name) |
| 2717 | goto err; |
| 2718 | |
Alex Elder | c8d1842 | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 2719 | snap->id = snap_id; |
| 2720 | snap->size = snap_size; |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 2721 | snap->features = snap_features; |
Alex Elder | 4e891e0 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 2722 | |
| 2723 | return snap; |
| 2724 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2725 | err: |
| 2726 | kfree(snap->name); |
| 2727 | kfree(snap); |
Alex Elder | 4e891e0 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 2728 | |
| 2729 | return ERR_PTR(ret); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2730 | } |
| 2731 | |
Alex Elder | cd89212 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2732 | static char *rbd_dev_v1_snap_info(struct rbd_device *rbd_dev, u32 which, |
| 2733 | u64 *snap_size, u64 *snap_features) |
| 2734 | { |
| 2735 | char *snap_name; |
| 2736 | |
| 2737 | rbd_assert(which < rbd_dev->header.snapc->num_snaps); |
| 2738 | |
| 2739 | *snap_size = rbd_dev->header.snap_sizes[which]; |
| 2740 | *snap_features = 0; /* No features for v1 */ |
| 2741 | |
| 2742 | /* Skip over names until we find the one we are looking for */ |
| 2743 | |
| 2744 | snap_name = rbd_dev->header.snap_names; |
| 2745 | while (which--) |
| 2746 | snap_name += strlen(snap_name) + 1; |
| 2747 | |
| 2748 | return snap_name; |
| 2749 | } |
| 2750 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2751 | /* |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2752 | * Get the size and object order for an image snapshot, or if |
| 2753 | * snap_id is CEPH_NOSNAP, gets this information for the base |
| 2754 | * image. |
| 2755 | */ |
| 2756 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
| 2757 | u8 *order, u64 *snap_size) |
| 2758 | { |
| 2759 | __le64 snapid = cpu_to_le64(snap_id); |
| 2760 | int ret; |
| 2761 | struct { |
| 2762 | u8 order; |
| 2763 | __le64 size; |
| 2764 | } __attribute__ ((packed)) size_buf = { 0 }; |
| 2765 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2766 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2767 | "rbd", "get_size", |
| 2768 | (char *) &snapid, sizeof (snapid), |
Alex Elder | 07b2391 | 2012-11-09 08:43:16 -0600 | [diff] [blame] | 2769 | (char *) &size_buf, sizeof (size_buf), NULL); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2770 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2771 | if (ret < 0) |
| 2772 | return ret; |
| 2773 | |
| 2774 | *order = size_buf.order; |
| 2775 | *snap_size = le64_to_cpu(size_buf.size); |
| 2776 | |
| 2777 | dout(" snap_id 0x%016llx order = %u, snap_size = %llu\n", |
| 2778 | (unsigned long long) snap_id, (unsigned int) *order, |
| 2779 | (unsigned long long) *snap_size); |
| 2780 | |
| 2781 | return 0; |
| 2782 | } |
| 2783 | |
| 2784 | static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) |
| 2785 | { |
| 2786 | return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, |
| 2787 | &rbd_dev->header.obj_order, |
| 2788 | &rbd_dev->header.image_size); |
| 2789 | } |
| 2790 | |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2791 | static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) |
| 2792 | { |
| 2793 | void *reply_buf; |
| 2794 | int ret; |
| 2795 | void *p; |
| 2796 | |
| 2797 | reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL); |
| 2798 | if (!reply_buf) |
| 2799 | return -ENOMEM; |
| 2800 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2801 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2802 | "rbd", "get_object_prefix", |
| 2803 | NULL, 0, |
Alex Elder | 07b2391 | 2012-11-09 08:43:16 -0600 | [diff] [blame] | 2804 | reply_buf, RBD_OBJ_PREFIX_LEN_MAX, NULL); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2805 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2806 | if (ret < 0) |
| 2807 | goto out; |
| 2808 | |
| 2809 | p = reply_buf; |
| 2810 | rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, |
| 2811 | p + RBD_OBJ_PREFIX_LEN_MAX, |
| 2812 | NULL, GFP_NOIO); |
| 2813 | |
| 2814 | if (IS_ERR(rbd_dev->header.object_prefix)) { |
| 2815 | ret = PTR_ERR(rbd_dev->header.object_prefix); |
| 2816 | rbd_dev->header.object_prefix = NULL; |
| 2817 | } else { |
| 2818 | dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); |
| 2819 | } |
| 2820 | |
| 2821 | out: |
| 2822 | kfree(reply_buf); |
| 2823 | |
| 2824 | return ret; |
| 2825 | } |
| 2826 | |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2827 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
| 2828 | u64 *snap_features) |
| 2829 | { |
| 2830 | __le64 snapid = cpu_to_le64(snap_id); |
| 2831 | struct { |
| 2832 | __le64 features; |
| 2833 | __le64 incompat; |
| 2834 | } features_buf = { 0 }; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 2835 | u64 incompat; |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2836 | int ret; |
| 2837 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2838 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2839 | "rbd", "get_features", |
| 2840 | (char *) &snapid, sizeof (snapid), |
| 2841 | (char *) &features_buf, sizeof (features_buf), |
Alex Elder | 07b2391 | 2012-11-09 08:43:16 -0600 | [diff] [blame] | 2842 | NULL); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2843 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2844 | if (ret < 0) |
| 2845 | return ret; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 2846 | |
| 2847 | incompat = le64_to_cpu(features_buf.incompat); |
| 2848 | if (incompat & ~RBD_FEATURES_ALL) |
Alex Elder | b8f5c6e | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 2849 | return -ENXIO; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 2850 | |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 2851 | *snap_features = le64_to_cpu(features_buf.features); |
| 2852 | |
| 2853 | dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", |
| 2854 | (unsigned long long) snap_id, |
| 2855 | (unsigned long long) *snap_features, |
| 2856 | (unsigned long long) le64_to_cpu(features_buf.incompat)); |
| 2857 | |
| 2858 | return 0; |
| 2859 | } |
| 2860 | |
| 2861 | static int rbd_dev_v2_features(struct rbd_device *rbd_dev) |
| 2862 | { |
| 2863 | return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, |
| 2864 | &rbd_dev->header.features); |
| 2865 | } |
| 2866 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2867 | static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) |
| 2868 | { |
| 2869 | struct rbd_spec *parent_spec; |
| 2870 | size_t size; |
| 2871 | void *reply_buf = NULL; |
| 2872 | __le64 snapid; |
| 2873 | void *p; |
| 2874 | void *end; |
| 2875 | char *image_id; |
| 2876 | u64 overlap; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2877 | int ret; |
| 2878 | |
| 2879 | parent_spec = rbd_spec_alloc(); |
| 2880 | if (!parent_spec) |
| 2881 | return -ENOMEM; |
| 2882 | |
| 2883 | size = sizeof (__le64) + /* pool_id */ |
| 2884 | sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX + /* image_id */ |
| 2885 | sizeof (__le64) + /* snap_id */ |
| 2886 | sizeof (__le64); /* overlap */ |
| 2887 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 2888 | if (!reply_buf) { |
| 2889 | ret = -ENOMEM; |
| 2890 | goto out_err; |
| 2891 | } |
| 2892 | |
| 2893 | snapid = cpu_to_le64(CEPH_NOSNAP); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2894 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2895 | "rbd", "get_parent", |
| 2896 | (char *) &snapid, sizeof (snapid), |
Alex Elder | 07b2391 | 2012-11-09 08:43:16 -0600 | [diff] [blame] | 2897 | (char *) reply_buf, size, NULL); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2898 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2899 | if (ret < 0) |
| 2900 | goto out_err; |
| 2901 | |
| 2902 | ret = -ERANGE; |
| 2903 | p = reply_buf; |
| 2904 | end = (char *) reply_buf + size; |
| 2905 | ceph_decode_64_safe(&p, end, parent_spec->pool_id, out_err); |
| 2906 | if (parent_spec->pool_id == CEPH_NOPOOL) |
| 2907 | goto out; /* No parent? No problem. */ |
| 2908 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 2909 | /* The ceph file layout needs to fit pool id in 32 bits */ |
| 2910 | |
| 2911 | ret = -EIO; |
| 2912 | if (WARN_ON(parent_spec->pool_id > (u64) U32_MAX)) |
| 2913 | goto out; |
| 2914 | |
Alex Elder | 979ed48 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 2915 | image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 2916 | if (IS_ERR(image_id)) { |
| 2917 | ret = PTR_ERR(image_id); |
| 2918 | goto out_err; |
| 2919 | } |
| 2920 | parent_spec->image_id = image_id; |
| 2921 | ceph_decode_64_safe(&p, end, parent_spec->snap_id, out_err); |
| 2922 | ceph_decode_64_safe(&p, end, overlap, out_err); |
| 2923 | |
| 2924 | rbd_dev->parent_overlap = overlap; |
| 2925 | rbd_dev->parent_spec = parent_spec; |
| 2926 | parent_spec = NULL; /* rbd_dev now owns this */ |
| 2927 | out: |
| 2928 | ret = 0; |
| 2929 | out_err: |
| 2930 | kfree(reply_buf); |
| 2931 | rbd_spec_put(parent_spec); |
| 2932 | |
| 2933 | return ret; |
| 2934 | } |
| 2935 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 2936 | static char *rbd_dev_image_name(struct rbd_device *rbd_dev) |
| 2937 | { |
| 2938 | size_t image_id_size; |
| 2939 | char *image_id; |
| 2940 | void *p; |
| 2941 | void *end; |
| 2942 | size_t size; |
| 2943 | void *reply_buf = NULL; |
| 2944 | size_t len = 0; |
| 2945 | char *image_name = NULL; |
| 2946 | int ret; |
| 2947 | |
| 2948 | rbd_assert(!rbd_dev->spec->image_name); |
| 2949 | |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 2950 | len = strlen(rbd_dev->spec->image_id); |
| 2951 | image_id_size = sizeof (__le32) + len; |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 2952 | image_id = kmalloc(image_id_size, GFP_KERNEL); |
| 2953 | if (!image_id) |
| 2954 | return NULL; |
| 2955 | |
| 2956 | p = image_id; |
| 2957 | end = (char *) image_id + image_id_size; |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 2958 | ceph_encode_string(&p, end, rbd_dev->spec->image_id, (u32) len); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 2959 | |
| 2960 | size = sizeof (__le32) + RBD_IMAGE_NAME_LEN_MAX; |
| 2961 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 2962 | if (!reply_buf) |
| 2963 | goto out; |
| 2964 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2965 | ret = rbd_obj_method_sync(rbd_dev, RBD_DIRECTORY, |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 2966 | "rbd", "dir_get_name", |
| 2967 | image_id, image_id_size, |
Alex Elder | 07b2391 | 2012-11-09 08:43:16 -0600 | [diff] [blame] | 2968 | (char *) reply_buf, size, NULL); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 2969 | if (ret < 0) |
| 2970 | goto out; |
| 2971 | p = reply_buf; |
| 2972 | end = (char *) reply_buf + size; |
| 2973 | image_name = ceph_extract_encoded_string(&p, end, &len, GFP_KERNEL); |
| 2974 | if (IS_ERR(image_name)) |
| 2975 | image_name = NULL; |
| 2976 | else |
| 2977 | dout("%s: name is %s len is %zd\n", __func__, image_name, len); |
| 2978 | out: |
| 2979 | kfree(reply_buf); |
| 2980 | kfree(image_id); |
| 2981 | |
| 2982 | return image_name; |
| 2983 | } |
| 2984 | |
| 2985 | /* |
| 2986 | * When a parent image gets probed, we only have the pool, image, |
| 2987 | * and snapshot ids but not the names of any of them. This call |
| 2988 | * is made later to fill in those names. It has to be done after |
| 2989 | * rbd_dev_snaps_update() has completed because some of the |
| 2990 | * information (in particular, snapshot name) is not available |
| 2991 | * until then. |
| 2992 | */ |
| 2993 | static int rbd_dev_probe_update_spec(struct rbd_device *rbd_dev) |
| 2994 | { |
| 2995 | struct ceph_osd_client *osdc; |
| 2996 | const char *name; |
| 2997 | void *reply_buf = NULL; |
| 2998 | int ret; |
| 2999 | |
| 3000 | if (rbd_dev->spec->pool_name) |
| 3001 | return 0; /* Already have the names */ |
| 3002 | |
| 3003 | /* Look up the pool name */ |
| 3004 | |
| 3005 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 3006 | name = ceph_pg_pool_name_by_id(osdc->osdmap, rbd_dev->spec->pool_id); |
Alex Elder | 935dc89 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3007 | if (!name) { |
| 3008 | rbd_warn(rbd_dev, "there is no pool with id %llu", |
| 3009 | rbd_dev->spec->pool_id); /* Really a BUG() */ |
| 3010 | return -EIO; |
| 3011 | } |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3012 | |
| 3013 | rbd_dev->spec->pool_name = kstrdup(name, GFP_KERNEL); |
| 3014 | if (!rbd_dev->spec->pool_name) |
| 3015 | return -ENOMEM; |
| 3016 | |
| 3017 | /* Fetch the image name; tolerate failure here */ |
| 3018 | |
| 3019 | name = rbd_dev_image_name(rbd_dev); |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3020 | if (name) |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3021 | rbd_dev->spec->image_name = (char *) name; |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3022 | else |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3023 | rbd_warn(rbd_dev, "unable to get image name"); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3024 | |
| 3025 | /* Look up the snapshot name. */ |
| 3026 | |
| 3027 | name = rbd_snap_name(rbd_dev, rbd_dev->spec->snap_id); |
| 3028 | if (!name) { |
Alex Elder | 935dc89 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3029 | rbd_warn(rbd_dev, "no snapshot with id %llu", |
| 3030 | rbd_dev->spec->snap_id); /* Really a BUG() */ |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3031 | ret = -EIO; |
| 3032 | goto out_err; |
| 3033 | } |
| 3034 | rbd_dev->spec->snap_name = kstrdup(name, GFP_KERNEL); |
| 3035 | if(!rbd_dev->spec->snap_name) |
| 3036 | goto out_err; |
| 3037 | |
| 3038 | return 0; |
| 3039 | out_err: |
| 3040 | kfree(reply_buf); |
| 3041 | kfree(rbd_dev->spec->pool_name); |
| 3042 | rbd_dev->spec->pool_name = NULL; |
| 3043 | |
| 3044 | return ret; |
| 3045 | } |
| 3046 | |
Alex Elder | 6e14b1a | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3047 | static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev, u64 *ver) |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3048 | { |
| 3049 | size_t size; |
| 3050 | int ret; |
| 3051 | void *reply_buf; |
| 3052 | void *p; |
| 3053 | void *end; |
| 3054 | u64 seq; |
| 3055 | u32 snap_count; |
| 3056 | struct ceph_snap_context *snapc; |
| 3057 | u32 i; |
| 3058 | |
| 3059 | /* |
| 3060 | * We'll need room for the seq value (maximum snapshot id), |
| 3061 | * snapshot count, and array of that many snapshot ids. |
| 3062 | * For now we have a fixed upper limit on the number we're |
| 3063 | * prepared to receive. |
| 3064 | */ |
| 3065 | size = sizeof (__le64) + sizeof (__le32) + |
| 3066 | RBD_MAX_SNAP_COUNT * sizeof (__le64); |
| 3067 | reply_buf = kzalloc(size, GFP_KERNEL); |
| 3068 | if (!reply_buf) |
| 3069 | return -ENOMEM; |
| 3070 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3071 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3072 | "rbd", "get_snapcontext", |
| 3073 | NULL, 0, |
Alex Elder | 07b2391 | 2012-11-09 08:43:16 -0600 | [diff] [blame] | 3074 | reply_buf, size, ver); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3075 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3076 | if (ret < 0) |
| 3077 | goto out; |
| 3078 | |
| 3079 | ret = -ERANGE; |
| 3080 | p = reply_buf; |
| 3081 | end = (char *) reply_buf + size; |
| 3082 | ceph_decode_64_safe(&p, end, seq, out); |
| 3083 | ceph_decode_32_safe(&p, end, snap_count, out); |
| 3084 | |
| 3085 | /* |
| 3086 | * Make sure the reported number of snapshot ids wouldn't go |
| 3087 | * beyond the end of our buffer. But before checking that, |
| 3088 | * make sure the computed size of the snapshot context we |
| 3089 | * allocate is representable in a size_t. |
| 3090 | */ |
| 3091 | if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) |
| 3092 | / sizeof (u64)) { |
| 3093 | ret = -EINVAL; |
| 3094 | goto out; |
| 3095 | } |
| 3096 | if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) |
| 3097 | goto out; |
| 3098 | |
| 3099 | size = sizeof (struct ceph_snap_context) + |
| 3100 | snap_count * sizeof (snapc->snaps[0]); |
| 3101 | snapc = kmalloc(size, GFP_KERNEL); |
| 3102 | if (!snapc) { |
| 3103 | ret = -ENOMEM; |
| 3104 | goto out; |
| 3105 | } |
| 3106 | |
| 3107 | atomic_set(&snapc->nref, 1); |
| 3108 | snapc->seq = seq; |
| 3109 | snapc->num_snaps = snap_count; |
| 3110 | for (i = 0; i < snap_count; i++) |
| 3111 | snapc->snaps[i] = ceph_decode_64(&p); |
| 3112 | |
| 3113 | rbd_dev->header.snapc = snapc; |
| 3114 | |
| 3115 | dout(" snap context seq = %llu, snap_count = %u\n", |
| 3116 | (unsigned long long) seq, (unsigned int) snap_count); |
| 3117 | |
| 3118 | out: |
| 3119 | kfree(reply_buf); |
| 3120 | |
| 3121 | return 0; |
| 3122 | } |
| 3123 | |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3124 | static char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, u32 which) |
| 3125 | { |
| 3126 | size_t size; |
| 3127 | void *reply_buf; |
| 3128 | __le64 snap_id; |
| 3129 | int ret; |
| 3130 | void *p; |
| 3131 | void *end; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3132 | char *snap_name; |
| 3133 | |
| 3134 | size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; |
| 3135 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 3136 | if (!reply_buf) |
| 3137 | return ERR_PTR(-ENOMEM); |
| 3138 | |
| 3139 | snap_id = cpu_to_le64(rbd_dev->header.snapc->snaps[which]); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3140 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3141 | "rbd", "get_snapshot_name", |
| 3142 | (char *) &snap_id, sizeof (snap_id), |
Alex Elder | 07b2391 | 2012-11-09 08:43:16 -0600 | [diff] [blame] | 3143 | reply_buf, size, NULL); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3144 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3145 | if (ret < 0) |
| 3146 | goto out; |
| 3147 | |
| 3148 | p = reply_buf; |
| 3149 | end = (char *) reply_buf + size; |
Alex Elder | e5c3553 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3150 | snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3151 | if (IS_ERR(snap_name)) { |
| 3152 | ret = PTR_ERR(snap_name); |
| 3153 | goto out; |
| 3154 | } else { |
| 3155 | dout(" snap_id 0x%016llx snap_name = %s\n", |
| 3156 | (unsigned long long) le64_to_cpu(snap_id), snap_name); |
| 3157 | } |
| 3158 | kfree(reply_buf); |
| 3159 | |
| 3160 | return snap_name; |
| 3161 | out: |
| 3162 | kfree(reply_buf); |
| 3163 | |
| 3164 | return ERR_PTR(ret); |
| 3165 | } |
| 3166 | |
| 3167 | static char *rbd_dev_v2_snap_info(struct rbd_device *rbd_dev, u32 which, |
| 3168 | u64 *snap_size, u64 *snap_features) |
| 3169 | { |
Alex Elder | e0b4986 | 2013-01-09 14:44:18 -0600 | [diff] [blame] | 3170 | u64 snap_id; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3171 | u8 order; |
| 3172 | int ret; |
| 3173 | |
| 3174 | snap_id = rbd_dev->header.snapc->snaps[which]; |
| 3175 | ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, &order, snap_size); |
| 3176 | if (ret) |
| 3177 | return ERR_PTR(ret); |
| 3178 | ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, snap_features); |
| 3179 | if (ret) |
| 3180 | return ERR_PTR(ret); |
| 3181 | |
| 3182 | return rbd_dev_v2_snap_name(rbd_dev, which); |
| 3183 | } |
| 3184 | |
| 3185 | static char *rbd_dev_snap_info(struct rbd_device *rbd_dev, u32 which, |
| 3186 | u64 *snap_size, u64 *snap_features) |
| 3187 | { |
| 3188 | if (rbd_dev->image_format == 1) |
| 3189 | return rbd_dev_v1_snap_info(rbd_dev, which, |
| 3190 | snap_size, snap_features); |
| 3191 | if (rbd_dev->image_format == 2) |
| 3192 | return rbd_dev_v2_snap_info(rbd_dev, which, |
| 3193 | snap_size, snap_features); |
| 3194 | return ERR_PTR(-EINVAL); |
| 3195 | } |
| 3196 | |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 3197 | static int rbd_dev_v2_refresh(struct rbd_device *rbd_dev, u64 *hver) |
| 3198 | { |
| 3199 | int ret; |
| 3200 | __u8 obj_order; |
| 3201 | |
| 3202 | down_write(&rbd_dev->header_rwsem); |
| 3203 | |
| 3204 | /* Grab old order first, to see if it changes */ |
| 3205 | |
| 3206 | obj_order = rbd_dev->header.obj_order, |
| 3207 | ret = rbd_dev_v2_image_size(rbd_dev); |
| 3208 | if (ret) |
| 3209 | goto out; |
| 3210 | if (rbd_dev->header.obj_order != obj_order) { |
| 3211 | ret = -EIO; |
| 3212 | goto out; |
| 3213 | } |
| 3214 | rbd_update_mapping_size(rbd_dev); |
| 3215 | |
| 3216 | ret = rbd_dev_v2_snap_context(rbd_dev, hver); |
| 3217 | dout("rbd_dev_v2_snap_context returned %d\n", ret); |
| 3218 | if (ret) |
| 3219 | goto out; |
| 3220 | ret = rbd_dev_snaps_update(rbd_dev); |
| 3221 | dout("rbd_dev_snaps_update returned %d\n", ret); |
| 3222 | if (ret) |
| 3223 | goto out; |
| 3224 | ret = rbd_dev_snaps_register(rbd_dev); |
| 3225 | dout("rbd_dev_snaps_register returned %d\n", ret); |
| 3226 | out: |
| 3227 | up_write(&rbd_dev->header_rwsem); |
| 3228 | |
| 3229 | return ret; |
| 3230 | } |
| 3231 | |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3232 | /* |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3233 | * Scan the rbd device's current snapshot list and compare it to the |
| 3234 | * newly-received snapshot context. Remove any existing snapshots |
| 3235 | * not present in the new snapshot context. Add a new snapshot for |
| 3236 | * any snaphots in the snapshot context not in the current list. |
| 3237 | * And verify there are no changes to snapshots we already know |
| 3238 | * about. |
| 3239 | * |
| 3240 | * Assumes the snapshots in the snapshot context are sorted by |
| 3241 | * snapshot id, highest id first. (Snapshots in the rbd_dev's list |
| 3242 | * are also maintained in that order.) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3243 | */ |
Alex Elder | 304f680 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 3244 | static int rbd_dev_snaps_update(struct rbd_device *rbd_dev) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3245 | { |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3246 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
| 3247 | const u32 snap_count = snapc->num_snaps; |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3248 | struct list_head *head = &rbd_dev->snaps; |
| 3249 | struct list_head *links = head->next; |
| 3250 | u32 index = 0; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3251 | |
Alex Elder | 9fcbb80 | 2012-08-23 23:48:49 -0500 | [diff] [blame] | 3252 | dout("%s: snap count is %u\n", __func__, (unsigned int) snap_count); |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3253 | while (index < snap_count || links != head) { |
| 3254 | u64 snap_id; |
| 3255 | struct rbd_snap *snap; |
Alex Elder | cd89212 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3256 | char *snap_name; |
| 3257 | u64 snap_size = 0; |
| 3258 | u64 snap_features = 0; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3259 | |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3260 | snap_id = index < snap_count ? snapc->snaps[index] |
| 3261 | : CEPH_NOSNAP; |
| 3262 | snap = links != head ? list_entry(links, struct rbd_snap, node) |
| 3263 | : NULL; |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 3264 | rbd_assert(!snap || snap->id != CEPH_NOSNAP); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3265 | |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3266 | if (snap_id == CEPH_NOSNAP || (snap && snap->id > snap_id)) { |
| 3267 | struct list_head *next = links->next; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3268 | |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 3269 | /* |
| 3270 | * A previously-existing snapshot is not in |
| 3271 | * the new snap context. |
| 3272 | * |
| 3273 | * If the now missing snapshot is the one the |
| 3274 | * image is mapped to, clear its exists flag |
| 3275 | * so we can avoid sending any more requests |
| 3276 | * to it. |
| 3277 | */ |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3278 | if (rbd_dev->spec->snap_id == snap->id) |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 3279 | clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
Alex Elder | 41f38c2 | 2012-10-25 23:34:40 -0500 | [diff] [blame] | 3280 | rbd_remove_snap_dev(snap); |
Alex Elder | 9fcbb80 | 2012-08-23 23:48:49 -0500 | [diff] [blame] | 3281 | dout("%ssnap id %llu has been removed\n", |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3282 | rbd_dev->spec->snap_id == snap->id ? |
| 3283 | "mapped " : "", |
Alex Elder | 9fcbb80 | 2012-08-23 23:48:49 -0500 | [diff] [blame] | 3284 | (unsigned long long) snap->id); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3285 | |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3286 | /* Done with this list entry; advance */ |
| 3287 | |
| 3288 | links = next; |
| 3289 | continue; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3290 | } |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3291 | |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3292 | snap_name = rbd_dev_snap_info(rbd_dev, index, |
| 3293 | &snap_size, &snap_features); |
Alex Elder | cd89212 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3294 | if (IS_ERR(snap_name)) |
| 3295 | return PTR_ERR(snap_name); |
| 3296 | |
Alex Elder | 9fcbb80 | 2012-08-23 23:48:49 -0500 | [diff] [blame] | 3297 | dout("entry %u: snap_id = %llu\n", (unsigned int) snap_count, |
| 3298 | (unsigned long long) snap_id); |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3299 | if (!snap || (snap_id != CEPH_NOSNAP && snap->id < snap_id)) { |
| 3300 | struct rbd_snap *new_snap; |
| 3301 | |
| 3302 | /* We haven't seen this snapshot before */ |
| 3303 | |
Alex Elder | c8d1842 | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3304 | new_snap = __rbd_add_snap_dev(rbd_dev, snap_name, |
Alex Elder | cd89212 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3305 | snap_id, snap_size, snap_features); |
Alex Elder | 9fcbb80 | 2012-08-23 23:48:49 -0500 | [diff] [blame] | 3306 | if (IS_ERR(new_snap)) { |
| 3307 | int err = PTR_ERR(new_snap); |
| 3308 | |
| 3309 | dout(" failed to add dev, error %d\n", err); |
| 3310 | |
| 3311 | return err; |
| 3312 | } |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3313 | |
| 3314 | /* New goes before existing, or at end of list */ |
| 3315 | |
Alex Elder | 9fcbb80 | 2012-08-23 23:48:49 -0500 | [diff] [blame] | 3316 | dout(" added dev%s\n", snap ? "" : " at end\n"); |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3317 | if (snap) |
| 3318 | list_add_tail(&new_snap->node, &snap->node); |
| 3319 | else |
Alex Elder | 523f325 | 2012-08-30 00:16:37 -0500 | [diff] [blame] | 3320 | list_add_tail(&new_snap->node, head); |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3321 | } else { |
| 3322 | /* Already have this one */ |
| 3323 | |
Alex Elder | 9fcbb80 | 2012-08-23 23:48:49 -0500 | [diff] [blame] | 3324 | dout(" already present\n"); |
| 3325 | |
Alex Elder | cd89212 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3326 | rbd_assert(snap->size == snap_size); |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 3327 | rbd_assert(!strcmp(snap->name, snap_name)); |
Alex Elder | cd89212 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3328 | rbd_assert(snap->features == snap_features); |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3329 | |
| 3330 | /* Done with this list entry; advance */ |
| 3331 | |
| 3332 | links = links->next; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3333 | } |
Alex Elder | 3593815 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3334 | |
| 3335 | /* Advance to the next entry in the snapshot context */ |
| 3336 | |
| 3337 | index++; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3338 | } |
Alex Elder | 9fcbb80 | 2012-08-23 23:48:49 -0500 | [diff] [blame] | 3339 | dout("%s: done\n", __func__); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3340 | |
| 3341 | return 0; |
| 3342 | } |
| 3343 | |
Alex Elder | 304f680 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 3344 | /* |
| 3345 | * Scan the list of snapshots and register the devices for any that |
| 3346 | * have not already been registered. |
| 3347 | */ |
| 3348 | static int rbd_dev_snaps_register(struct rbd_device *rbd_dev) |
| 3349 | { |
| 3350 | struct rbd_snap *snap; |
| 3351 | int ret = 0; |
| 3352 | |
| 3353 | dout("%s called\n", __func__); |
Alex Elder | 86ff77b | 2012-08-31 17:29:53 -0500 | [diff] [blame] | 3354 | if (WARN_ON(!device_is_registered(&rbd_dev->dev))) |
| 3355 | return -EIO; |
Alex Elder | 304f680 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 3356 | |
| 3357 | list_for_each_entry(snap, &rbd_dev->snaps, node) { |
| 3358 | if (!rbd_snap_registered(snap)) { |
| 3359 | ret = rbd_register_snap_dev(snap, &rbd_dev->dev); |
| 3360 | if (ret < 0) |
| 3361 | break; |
| 3362 | } |
| 3363 | } |
| 3364 | dout("%s: returning %d\n", __func__, ret); |
| 3365 | |
| 3366 | return ret; |
| 3367 | } |
| 3368 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3369 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
| 3370 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3371 | struct device *dev; |
Alex Elder | cd789ab | 2012-08-30 00:16:38 -0500 | [diff] [blame] | 3372 | int ret; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3373 | |
| 3374 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3375 | |
Alex Elder | cd789ab | 2012-08-30 00:16:38 -0500 | [diff] [blame] | 3376 | dev = &rbd_dev->dev; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3377 | dev->bus = &rbd_bus_type; |
| 3378 | dev->type = &rbd_device_type; |
| 3379 | dev->parent = &rbd_root_dev; |
| 3380 | dev->release = rbd_dev_release; |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3381 | dev_set_name(dev, "%d", rbd_dev->dev_id); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3382 | ret = device_register(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3383 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3384 | mutex_unlock(&ctl_mutex); |
Alex Elder | cd789ab | 2012-08-30 00:16:38 -0500 | [diff] [blame] | 3385 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3386 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3387 | } |
| 3388 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3389 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
| 3390 | { |
| 3391 | device_unregister(&rbd_dev->dev); |
| 3392 | } |
| 3393 | |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3394 | static atomic64_t rbd_dev_id_max = ATOMIC64_INIT(0); |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3395 | |
| 3396 | /* |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 3397 | * Get a unique rbd identifier for the given new rbd_dev, and add |
| 3398 | * the rbd_dev to the global list. The minimum rbd id is 1. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3399 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3400 | static void rbd_dev_id_get(struct rbd_device *rbd_dev) |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 3401 | { |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3402 | rbd_dev->dev_id = atomic64_inc_return(&rbd_dev_id_max); |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 3403 | |
| 3404 | spin_lock(&rbd_dev_list_lock); |
| 3405 | list_add_tail(&rbd_dev->node, &rbd_dev_list); |
| 3406 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3407 | dout("rbd_dev %p given dev id %llu\n", rbd_dev, |
| 3408 | (unsigned long long) rbd_dev->dev_id); |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3409 | } |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 3410 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3411 | /* |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 3412 | * Remove an rbd_dev from the global list, and record that its |
| 3413 | * identifier is no longer in use. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3414 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3415 | static void rbd_dev_id_put(struct rbd_device *rbd_dev) |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3416 | { |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3417 | struct list_head *tmp; |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3418 | int rbd_id = rbd_dev->dev_id; |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3419 | int max_id; |
| 3420 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 3421 | rbd_assert(rbd_id > 0); |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 3422 | |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3423 | dout("rbd_dev %p released dev id %llu\n", rbd_dev, |
| 3424 | (unsigned long long) rbd_dev->dev_id); |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 3425 | spin_lock(&rbd_dev_list_lock); |
| 3426 | list_del_init(&rbd_dev->node); |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3427 | |
| 3428 | /* |
| 3429 | * If the id being "put" is not the current maximum, there |
| 3430 | * is nothing special we need to do. |
| 3431 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3432 | if (rbd_id != atomic64_read(&rbd_dev_id_max)) { |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3433 | spin_unlock(&rbd_dev_list_lock); |
| 3434 | return; |
| 3435 | } |
| 3436 | |
| 3437 | /* |
| 3438 | * We need to update the current maximum id. Search the |
| 3439 | * list to find out what it is. We're more likely to find |
| 3440 | * the maximum at the end, so search the list backward. |
| 3441 | */ |
| 3442 | max_id = 0; |
| 3443 | list_for_each_prev(tmp, &rbd_dev_list) { |
| 3444 | struct rbd_device *rbd_dev; |
| 3445 | |
| 3446 | rbd_dev = list_entry(tmp, struct rbd_device, node); |
Alex Elder | b213e0b | 2012-10-10 21:19:13 -0700 | [diff] [blame] | 3447 | if (rbd_dev->dev_id > max_id) |
| 3448 | max_id = rbd_dev->dev_id; |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3449 | } |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 3450 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 3451 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3452 | /* |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3453 | * The max id could have been updated by rbd_dev_id_get(), in |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3454 | * which case it now accurately reflects the new maximum. |
| 3455 | * Be careful not to overwrite the maximum value in that |
| 3456 | * case. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3457 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3458 | atomic64_cmpxchg(&rbd_dev_id_max, rbd_id, max_id); |
| 3459 | dout(" max dev id has been reset\n"); |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 3460 | } |
| 3461 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3462 | /* |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3463 | * Skips over white space at *buf, and updates *buf to point to the |
| 3464 | * first found non-space character (if any). Returns the length of |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3465 | * the token (string of non-white space characters) found. Note |
| 3466 | * that *buf must be terminated with '\0'. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3467 | */ |
| 3468 | static inline size_t next_token(const char **buf) |
| 3469 | { |
| 3470 | /* |
| 3471 | * These are the characters that produce nonzero for |
| 3472 | * isspace() in the "C" and "POSIX" locales. |
| 3473 | */ |
| 3474 | const char *spaces = " \f\n\r\t\v"; |
| 3475 | |
| 3476 | *buf += strspn(*buf, spaces); /* Find start of token */ |
| 3477 | |
| 3478 | return strcspn(*buf, spaces); /* Return token length */ |
| 3479 | } |
| 3480 | |
| 3481 | /* |
| 3482 | * Finds the next token in *buf, and if the provided token buffer is |
| 3483 | * big enough, copies the found token into it. The result, if |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3484 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
| 3485 | * must be terminated with '\0' on entry. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3486 | * |
| 3487 | * Returns the length of the token found (not including the '\0'). |
| 3488 | * Return value will be 0 if no token is found, and it will be >= |
| 3489 | * token_size if the token would not fit. |
| 3490 | * |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3491 | * The *buf pointer will be updated to point beyond the end of the |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3492 | * found token. Note that this occurs even if the token buffer is |
| 3493 | * too small to hold it. |
| 3494 | */ |
| 3495 | static inline size_t copy_token(const char **buf, |
| 3496 | char *token, |
| 3497 | size_t token_size) |
| 3498 | { |
| 3499 | size_t len; |
| 3500 | |
| 3501 | len = next_token(buf); |
| 3502 | if (len < token_size) { |
| 3503 | memcpy(token, *buf, len); |
| 3504 | *(token + len) = '\0'; |
| 3505 | } |
| 3506 | *buf += len; |
| 3507 | |
| 3508 | return len; |
| 3509 | } |
| 3510 | |
| 3511 | /* |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 3512 | * Finds the next token in *buf, dynamically allocates a buffer big |
| 3513 | * enough to hold a copy of it, and copies the token into the new |
| 3514 | * buffer. The copy is guaranteed to be terminated with '\0'. Note |
| 3515 | * that a duplicate buffer is created even for a zero-length token. |
| 3516 | * |
| 3517 | * Returns a pointer to the newly-allocated duplicate, or a null |
| 3518 | * pointer if memory for the duplicate was not available. If |
| 3519 | * the lenp argument is a non-null pointer, the length of the token |
| 3520 | * (not including the '\0') is returned in *lenp. |
| 3521 | * |
| 3522 | * If successful, the *buf pointer will be updated to point beyond |
| 3523 | * the end of the found token. |
| 3524 | * |
| 3525 | * Note: uses GFP_KERNEL for allocation. |
| 3526 | */ |
| 3527 | static inline char *dup_token(const char **buf, size_t *lenp) |
| 3528 | { |
| 3529 | char *dup; |
| 3530 | size_t len; |
| 3531 | |
| 3532 | len = next_token(buf); |
Alex Elder | 4caf35f | 2012-11-01 08:39:27 -0500 | [diff] [blame] | 3533 | dup = kmemdup(*buf, len + 1, GFP_KERNEL); |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 3534 | if (!dup) |
| 3535 | return NULL; |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 3536 | *(dup + len) = '\0'; |
| 3537 | *buf += len; |
| 3538 | |
| 3539 | if (lenp) |
| 3540 | *lenp = len; |
| 3541 | |
| 3542 | return dup; |
| 3543 | } |
| 3544 | |
| 3545 | /* |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3546 | * Parse the options provided for an "rbd add" (i.e., rbd image |
| 3547 | * mapping) request. These arrive via a write to /sys/bus/rbd/add, |
| 3548 | * and the data written is passed here via a NUL-terminated buffer. |
| 3549 | * Returns 0 if successful or an error code otherwise. |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3550 | * |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3551 | * The information extracted from these options is recorded in |
| 3552 | * the other parameters which return dynamically-allocated |
| 3553 | * structures: |
| 3554 | * ceph_opts |
| 3555 | * The address of a pointer that will refer to a ceph options |
| 3556 | * structure. Caller must release the returned pointer using |
| 3557 | * ceph_destroy_options() when it is no longer needed. |
| 3558 | * rbd_opts |
| 3559 | * Address of an rbd options pointer. Fully initialized by |
| 3560 | * this function; caller must release with kfree(). |
| 3561 | * spec |
| 3562 | * Address of an rbd image specification pointer. Fully |
| 3563 | * initialized by this function based on parsed options. |
| 3564 | * Caller must release with rbd_spec_put(). |
| 3565 | * |
| 3566 | * The options passed take this form: |
| 3567 | * <mon_addrs> <options> <pool_name> <image_name> [<snap_id>] |
| 3568 | * where: |
| 3569 | * <mon_addrs> |
| 3570 | * A comma-separated list of one or more monitor addresses. |
| 3571 | * A monitor address is an ip address, optionally followed |
| 3572 | * by a port number (separated by a colon). |
| 3573 | * I.e.: ip1[:port1][,ip2[:port2]...] |
| 3574 | * <options> |
| 3575 | * A comma-separated list of ceph and/or rbd options. |
| 3576 | * <pool_name> |
| 3577 | * The name of the rados pool containing the rbd image. |
| 3578 | * <image_name> |
| 3579 | * The name of the image in that pool to map. |
| 3580 | * <snap_id> |
| 3581 | * An optional snapshot id. If provided, the mapping will |
| 3582 | * present data from the image at the time that snapshot was |
| 3583 | * created. The image head is used if no snapshot id is |
| 3584 | * provided. Snapshot mappings are always read-only. |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3585 | */ |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3586 | static int rbd_add_parse_args(const char *buf, |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3587 | struct ceph_options **ceph_opts, |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3588 | struct rbd_options **opts, |
| 3589 | struct rbd_spec **rbd_spec) |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3590 | { |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3591 | size_t len; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3592 | char *options; |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3593 | const char *mon_addrs; |
| 3594 | size_t mon_addrs_size; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3595 | struct rbd_spec *spec = NULL; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3596 | struct rbd_options *rbd_opts = NULL; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3597 | struct ceph_options *copts; |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3598 | int ret; |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3599 | |
| 3600 | /* The first four tokens are required */ |
| 3601 | |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3602 | len = next_token(&buf); |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3603 | if (!len) { |
| 3604 | rbd_warn(NULL, "no monitor address(es) provided"); |
| 3605 | return -EINVAL; |
| 3606 | } |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3607 | mon_addrs = buf; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3608 | mon_addrs_size = len + 1; |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3609 | buf += len; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3610 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3611 | ret = -EINVAL; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3612 | options = dup_token(&buf, NULL); |
| 3613 | if (!options) |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3614 | return -ENOMEM; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3615 | if (!*options) { |
| 3616 | rbd_warn(NULL, "no options provided"); |
| 3617 | goto out_err; |
| 3618 | } |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3619 | |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3620 | spec = rbd_spec_alloc(); |
| 3621 | if (!spec) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3622 | goto out_mem; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3623 | |
| 3624 | spec->pool_name = dup_token(&buf, NULL); |
| 3625 | if (!spec->pool_name) |
| 3626 | goto out_mem; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3627 | if (!*spec->pool_name) { |
| 3628 | rbd_warn(NULL, "no pool name provided"); |
| 3629 | goto out_err; |
| 3630 | } |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3631 | |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3632 | spec->image_name = dup_token(&buf, NULL); |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3633 | if (!spec->image_name) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3634 | goto out_mem; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3635 | if (!*spec->image_name) { |
| 3636 | rbd_warn(NULL, "no image name provided"); |
| 3637 | goto out_err; |
| 3638 | } |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3639 | |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3640 | /* |
| 3641 | * Snapshot name is optional; default is to use "-" |
| 3642 | * (indicating the head/no snapshot). |
| 3643 | */ |
Alex Elder | 3feeb894 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 3644 | len = next_token(&buf); |
Alex Elder | 820a5f3 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 3645 | if (!len) { |
Alex Elder | 3feeb894 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 3646 | buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ |
| 3647 | len = sizeof (RBD_SNAP_HEAD_NAME) - 1; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3648 | } else if (len > RBD_MAX_SNAP_NAME_LEN) { |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3649 | ret = -ENAMETOOLONG; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3650 | goto out_err; |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 3651 | } |
Alex Elder | 4caf35f | 2012-11-01 08:39:27 -0500 | [diff] [blame] | 3652 | spec->snap_name = kmemdup(buf, len + 1, GFP_KERNEL); |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3653 | if (!spec->snap_name) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3654 | goto out_mem; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3655 | *(spec->snap_name + len) = '\0'; |
Alex Elder | e5c3553 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3656 | |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3657 | /* Initialize all rbd options to the defaults */ |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3658 | |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3659 | rbd_opts = kzalloc(sizeof (*rbd_opts), GFP_KERNEL); |
| 3660 | if (!rbd_opts) |
| 3661 | goto out_mem; |
| 3662 | |
| 3663 | rbd_opts->read_only = RBD_READ_ONLY_DEFAULT; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3664 | |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3665 | copts = ceph_parse_options(options, mon_addrs, |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3666 | mon_addrs + mon_addrs_size - 1, |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3667 | parse_rbd_opts_token, rbd_opts); |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3668 | if (IS_ERR(copts)) { |
| 3669 | ret = PTR_ERR(copts); |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3670 | goto out_err; |
| 3671 | } |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3672 | kfree(options); |
| 3673 | |
| 3674 | *ceph_opts = copts; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3675 | *opts = rbd_opts; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3676 | *rbd_spec = spec; |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3677 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3678 | return 0; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3679 | out_mem: |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3680 | ret = -ENOMEM; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3681 | out_err: |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3682 | kfree(rbd_opts); |
| 3683 | rbd_spec_put(spec); |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3684 | kfree(options); |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3685 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3686 | return ret; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 3687 | } |
| 3688 | |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3689 | /* |
| 3690 | * An rbd format 2 image has a unique identifier, distinct from the |
| 3691 | * name given to it by the user. Internally, that identifier is |
| 3692 | * what's used to specify the names of objects related to the image. |
| 3693 | * |
| 3694 | * A special "rbd id" object is used to map an rbd image name to its |
| 3695 | * id. If that object doesn't exist, then there is no v2 rbd image |
| 3696 | * with the supplied name. |
| 3697 | * |
| 3698 | * This function will record the given rbd_dev's image_id field if |
| 3699 | * it can be determined, and in that case will return 0. If any |
| 3700 | * errors occur a negative errno will be returned and the rbd_dev's |
| 3701 | * image_id field will be unchanged (and should be NULL). |
| 3702 | */ |
| 3703 | static int rbd_dev_image_id(struct rbd_device *rbd_dev) |
| 3704 | { |
| 3705 | int ret; |
| 3706 | size_t size; |
| 3707 | char *object_name; |
| 3708 | void *response; |
| 3709 | void *p; |
| 3710 | |
| 3711 | /* |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3712 | * When probing a parent image, the image id is already |
| 3713 | * known (and the image name likely is not). There's no |
| 3714 | * need to fetch the image id again in this case. |
| 3715 | */ |
| 3716 | if (rbd_dev->spec->image_id) |
| 3717 | return 0; |
| 3718 | |
| 3719 | /* |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3720 | * First, see if the format 2 image id file exists, and if |
| 3721 | * so, get the image's persistent id from it. |
| 3722 | */ |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3723 | size = sizeof (RBD_ID_PREFIX) + strlen(rbd_dev->spec->image_name); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3724 | object_name = kmalloc(size, GFP_NOIO); |
| 3725 | if (!object_name) |
| 3726 | return -ENOMEM; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3727 | sprintf(object_name, "%s%s", RBD_ID_PREFIX, rbd_dev->spec->image_name); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3728 | dout("rbd id object name is %s\n", object_name); |
| 3729 | |
| 3730 | /* Response will be an encoded string, which includes a length */ |
| 3731 | |
| 3732 | size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; |
| 3733 | response = kzalloc(size, GFP_NOIO); |
| 3734 | if (!response) { |
| 3735 | ret = -ENOMEM; |
| 3736 | goto out; |
| 3737 | } |
| 3738 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3739 | ret = rbd_obj_method_sync(rbd_dev, object_name, |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3740 | "rbd", "get_id", |
| 3741 | NULL, 0, |
Alex Elder | 07b2391 | 2012-11-09 08:43:16 -0600 | [diff] [blame] | 3742 | response, RBD_IMAGE_ID_LEN_MAX, NULL); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3743 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3744 | if (ret < 0) |
| 3745 | goto out; |
| 3746 | |
| 3747 | p = response; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3748 | rbd_dev->spec->image_id = ceph_extract_encoded_string(&p, |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3749 | p + RBD_IMAGE_ID_LEN_MAX, |
Alex Elder | 979ed48 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3750 | NULL, GFP_NOIO); |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3751 | if (IS_ERR(rbd_dev->spec->image_id)) { |
| 3752 | ret = PTR_ERR(rbd_dev->spec->image_id); |
| 3753 | rbd_dev->spec->image_id = NULL; |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3754 | } else { |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3755 | dout("image_id is %s\n", rbd_dev->spec->image_id); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3756 | } |
| 3757 | out: |
| 3758 | kfree(response); |
| 3759 | kfree(object_name); |
| 3760 | |
| 3761 | return ret; |
| 3762 | } |
| 3763 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3764 | static int rbd_dev_v1_probe(struct rbd_device *rbd_dev) |
| 3765 | { |
| 3766 | int ret; |
| 3767 | size_t size; |
| 3768 | |
| 3769 | /* Version 1 images have no id; empty string is used */ |
| 3770 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3771 | rbd_dev->spec->image_id = kstrdup("", GFP_KERNEL); |
| 3772 | if (!rbd_dev->spec->image_id) |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3773 | return -ENOMEM; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3774 | |
| 3775 | /* Record the header object name for this rbd image. */ |
| 3776 | |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3777 | size = strlen(rbd_dev->spec->image_name) + sizeof (RBD_SUFFIX); |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3778 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); |
| 3779 | if (!rbd_dev->header_name) { |
| 3780 | ret = -ENOMEM; |
| 3781 | goto out_err; |
| 3782 | } |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3783 | sprintf(rbd_dev->header_name, "%s%s", |
| 3784 | rbd_dev->spec->image_name, RBD_SUFFIX); |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3785 | |
| 3786 | /* Populate rbd image metadata */ |
| 3787 | |
| 3788 | ret = rbd_read_header(rbd_dev, &rbd_dev->header); |
| 3789 | if (ret < 0) |
| 3790 | goto out_err; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3791 | |
| 3792 | /* Version 1 images have no parent (no layering) */ |
| 3793 | |
| 3794 | rbd_dev->parent_spec = NULL; |
| 3795 | rbd_dev->parent_overlap = 0; |
| 3796 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3797 | rbd_dev->image_format = 1; |
| 3798 | |
| 3799 | dout("discovered version 1 image, header name is %s\n", |
| 3800 | rbd_dev->header_name); |
| 3801 | |
| 3802 | return 0; |
| 3803 | |
| 3804 | out_err: |
| 3805 | kfree(rbd_dev->header_name); |
| 3806 | rbd_dev->header_name = NULL; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3807 | kfree(rbd_dev->spec->image_id); |
| 3808 | rbd_dev->spec->image_id = NULL; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3809 | |
| 3810 | return ret; |
| 3811 | } |
| 3812 | |
| 3813 | static int rbd_dev_v2_probe(struct rbd_device *rbd_dev) |
| 3814 | { |
| 3815 | size_t size; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3816 | int ret; |
Alex Elder | 6e14b1a | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3817 | u64 ver = 0; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3818 | |
| 3819 | /* |
| 3820 | * Image id was filled in by the caller. Record the header |
| 3821 | * object name for this rbd image. |
| 3822 | */ |
Alex Elder | 979ed48 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3823 | size = sizeof (RBD_HEADER_PREFIX) + strlen(rbd_dev->spec->image_id); |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3824 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); |
| 3825 | if (!rbd_dev->header_name) |
| 3826 | return -ENOMEM; |
| 3827 | sprintf(rbd_dev->header_name, "%s%s", |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3828 | RBD_HEADER_PREFIX, rbd_dev->spec->image_id); |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3829 | |
| 3830 | /* Get the size and object order for the image */ |
| 3831 | |
| 3832 | ret = rbd_dev_v2_image_size(rbd_dev); |
| 3833 | if (ret < 0) |
| 3834 | goto out_err; |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3835 | |
| 3836 | /* Get the object prefix (a.k.a. block_name) for the image */ |
| 3837 | |
| 3838 | ret = rbd_dev_v2_object_prefix(rbd_dev); |
| 3839 | if (ret < 0) |
| 3840 | goto out_err; |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3841 | |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 3842 | /* Get the and check features for the image */ |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3843 | |
| 3844 | ret = rbd_dev_v2_features(rbd_dev); |
| 3845 | if (ret < 0) |
| 3846 | goto out_err; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3847 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3848 | /* If the image supports layering, get the parent info */ |
| 3849 | |
| 3850 | if (rbd_dev->header.features & RBD_FEATURE_LAYERING) { |
| 3851 | ret = rbd_dev_v2_parent_info(rbd_dev); |
| 3852 | if (ret < 0) |
| 3853 | goto out_err; |
| 3854 | } |
| 3855 | |
Alex Elder | 6e14b1a | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3856 | /* crypto and compression type aren't (yet) supported for v2 images */ |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3857 | |
Alex Elder | 6e14b1a | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3858 | rbd_dev->header.crypt_type = 0; |
| 3859 | rbd_dev->header.comp_type = 0; |
| 3860 | |
| 3861 | /* Get the snapshot context, plus the header version */ |
| 3862 | |
| 3863 | ret = rbd_dev_v2_snap_context(rbd_dev, &ver); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3864 | if (ret) |
| 3865 | goto out_err; |
Alex Elder | 6e14b1a | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3866 | rbd_dev->header.obj_version = ver; |
| 3867 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3868 | rbd_dev->image_format = 2; |
| 3869 | |
| 3870 | dout("discovered version 2 image, header name is %s\n", |
| 3871 | rbd_dev->header_name); |
| 3872 | |
Alex Elder | 3515297 | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 3873 | return 0; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3874 | out_err: |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3875 | rbd_dev->parent_overlap = 0; |
| 3876 | rbd_spec_put(rbd_dev->parent_spec); |
| 3877 | rbd_dev->parent_spec = NULL; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3878 | kfree(rbd_dev->header_name); |
| 3879 | rbd_dev->header_name = NULL; |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3880 | kfree(rbd_dev->header.object_prefix); |
| 3881 | rbd_dev->header.object_prefix = NULL; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3882 | |
| 3883 | return ret; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3884 | } |
| 3885 | |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 3886 | static int rbd_dev_probe_finish(struct rbd_device *rbd_dev) |
| 3887 | { |
| 3888 | int ret; |
| 3889 | |
| 3890 | /* no need to lock here, as rbd_dev is not registered yet */ |
| 3891 | ret = rbd_dev_snaps_update(rbd_dev); |
| 3892 | if (ret) |
| 3893 | return ret; |
| 3894 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3895 | ret = rbd_dev_probe_update_spec(rbd_dev); |
| 3896 | if (ret) |
| 3897 | goto err_out_snaps; |
| 3898 | |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 3899 | ret = rbd_dev_set_mapping(rbd_dev); |
| 3900 | if (ret) |
| 3901 | goto err_out_snaps; |
| 3902 | |
| 3903 | /* generate unique id: find highest unique id, add one */ |
| 3904 | rbd_dev_id_get(rbd_dev); |
| 3905 | |
| 3906 | /* Fill in the device name, now that we have its id. */ |
| 3907 | BUILD_BUG_ON(DEV_NAME_LEN |
| 3908 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); |
| 3909 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); |
| 3910 | |
| 3911 | /* Get our block major device number. */ |
| 3912 | |
| 3913 | ret = register_blkdev(0, rbd_dev->name); |
| 3914 | if (ret < 0) |
| 3915 | goto err_out_id; |
| 3916 | rbd_dev->major = ret; |
| 3917 | |
| 3918 | /* Set up the blkdev mapping. */ |
| 3919 | |
| 3920 | ret = rbd_init_disk(rbd_dev); |
| 3921 | if (ret) |
| 3922 | goto err_out_blkdev; |
| 3923 | |
| 3924 | ret = rbd_bus_add_dev(rbd_dev); |
| 3925 | if (ret) |
| 3926 | goto err_out_disk; |
| 3927 | |
| 3928 | /* |
| 3929 | * At this point cleanup in the event of an error is the job |
| 3930 | * of the sysfs code (initiated by rbd_bus_del_dev()). |
| 3931 | */ |
| 3932 | down_write(&rbd_dev->header_rwsem); |
| 3933 | ret = rbd_dev_snaps_register(rbd_dev); |
| 3934 | up_write(&rbd_dev->header_rwsem); |
| 3935 | if (ret) |
| 3936 | goto err_out_bus; |
| 3937 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 3938 | ret = rbd_dev_header_watch_sync(rbd_dev, 1); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 3939 | if (ret) |
| 3940 | goto err_out_bus; |
| 3941 | |
| 3942 | /* Everything's ready. Announce the disk to the world. */ |
| 3943 | |
| 3944 | add_disk(rbd_dev->disk); |
| 3945 | |
| 3946 | pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name, |
| 3947 | (unsigned long long) rbd_dev->mapping.size); |
| 3948 | |
| 3949 | return ret; |
| 3950 | err_out_bus: |
| 3951 | /* this will also clean up rest of rbd_dev stuff */ |
| 3952 | |
| 3953 | rbd_bus_del_dev(rbd_dev); |
| 3954 | |
| 3955 | return ret; |
| 3956 | err_out_disk: |
| 3957 | rbd_free_disk(rbd_dev); |
| 3958 | err_out_blkdev: |
| 3959 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
| 3960 | err_out_id: |
| 3961 | rbd_dev_id_put(rbd_dev); |
| 3962 | err_out_snaps: |
| 3963 | rbd_remove_all_snaps(rbd_dev); |
| 3964 | |
| 3965 | return ret; |
| 3966 | } |
| 3967 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3968 | /* |
| 3969 | * Probe for the existence of the header object for the given rbd |
| 3970 | * device. For format 2 images this includes determining the image |
| 3971 | * id. |
| 3972 | */ |
| 3973 | static int rbd_dev_probe(struct rbd_device *rbd_dev) |
| 3974 | { |
| 3975 | int ret; |
| 3976 | |
| 3977 | /* |
| 3978 | * Get the id from the image id object. If it's not a |
| 3979 | * format 2 image, we'll get ENOENT back, and we'll assume |
| 3980 | * it's a format 1 image. |
| 3981 | */ |
| 3982 | ret = rbd_dev_image_id(rbd_dev); |
| 3983 | if (ret) |
| 3984 | ret = rbd_dev_v1_probe(rbd_dev); |
| 3985 | else |
| 3986 | ret = rbd_dev_v2_probe(rbd_dev); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 3987 | if (ret) { |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3988 | dout("probe failed, returning %d\n", ret); |
| 3989 | |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 3990 | return ret; |
| 3991 | } |
| 3992 | |
| 3993 | ret = rbd_dev_probe_finish(rbd_dev); |
| 3994 | if (ret) |
| 3995 | rbd_header_free(&rbd_dev->header); |
| 3996 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3997 | return ret; |
| 3998 | } |
| 3999 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 4000 | static ssize_t rbd_add(struct bus_type *bus, |
| 4001 | const char *buf, |
| 4002 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4003 | { |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 4004 | struct rbd_device *rbd_dev = NULL; |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4005 | struct ceph_options *ceph_opts = NULL; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4006 | struct rbd_options *rbd_opts = NULL; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4007 | struct rbd_spec *spec = NULL; |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4008 | struct rbd_client *rbdc; |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4009 | struct ceph_osd_client *osdc; |
| 4010 | int rc = -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4011 | |
| 4012 | if (!try_module_get(THIS_MODULE)) |
| 4013 | return -ENODEV; |
| 4014 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4015 | /* parse add command */ |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4016 | rc = rbd_add_parse_args(buf, &ceph_opts, &rbd_opts, &spec); |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4017 | if (rc < 0) |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4018 | goto err_out_module; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4019 | |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4020 | rbdc = rbd_get_client(ceph_opts); |
| 4021 | if (IS_ERR(rbdc)) { |
| 4022 | rc = PTR_ERR(rbdc); |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4023 | goto err_out_args; |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4024 | } |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4025 | ceph_opts = NULL; /* rbd_dev client now owns this */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4026 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4027 | /* pick the pool */ |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4028 | osdc = &rbdc->client->osdc; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4029 | rc = ceph_pg_poolid_by_name(osdc->osdmap, spec->pool_name); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4030 | if (rc < 0) |
| 4031 | goto err_out_client; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4032 | spec->pool_id = (u64) rc; |
| 4033 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 4034 | /* The ceph file layout needs to fit pool id in 32 bits */ |
| 4035 | |
| 4036 | if (WARN_ON(spec->pool_id > (u64) U32_MAX)) { |
| 4037 | rc = -EIO; |
| 4038 | goto err_out_client; |
| 4039 | } |
| 4040 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4041 | rbd_dev = rbd_dev_create(rbdc, spec); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4042 | if (!rbd_dev) |
| 4043 | goto err_out_client; |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4044 | rbdc = NULL; /* rbd_dev now owns this */ |
| 4045 | spec = NULL; /* rbd_dev now owns this */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4046 | |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4047 | rbd_dev->mapping.read_only = rbd_opts->read_only; |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4048 | kfree(rbd_opts); |
| 4049 | rbd_opts = NULL; /* done with this */ |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4050 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4051 | rc = rbd_dev_probe(rbd_dev); |
| 4052 | if (rc < 0) |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4053 | goto err_out_rbd_dev; |
Alex Elder | 05fd6f6 | 2012-08-29 17:11:07 -0500 | [diff] [blame] | 4054 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4055 | return count; |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4056 | err_out_rbd_dev: |
| 4057 | rbd_dev_destroy(rbd_dev); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4058 | err_out_client: |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4059 | rbd_put_client(rbdc); |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4060 | err_out_args: |
Alex Elder | 78cea76 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4061 | if (ceph_opts) |
| 4062 | ceph_destroy_options(ceph_opts); |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4063 | kfree(rbd_opts); |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4064 | rbd_spec_put(spec); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4065 | err_out_module: |
| 4066 | module_put(THIS_MODULE); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4067 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4068 | dout("Error adding device %s\n", buf); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4069 | |
| 4070 | return (ssize_t) rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4071 | } |
| 4072 | |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4073 | static struct rbd_device *__rbd_get_dev(unsigned long dev_id) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4074 | { |
| 4075 | struct list_head *tmp; |
| 4076 | struct rbd_device *rbd_dev; |
| 4077 | |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4078 | spin_lock(&rbd_dev_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4079 | list_for_each(tmp, &rbd_dev_list) { |
| 4080 | rbd_dev = list_entry(tmp, struct rbd_device, node); |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4081 | if (rbd_dev->dev_id == dev_id) { |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4082 | spin_unlock(&rbd_dev_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4083 | return rbd_dev; |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4084 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4085 | } |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4086 | spin_unlock(&rbd_dev_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4087 | return NULL; |
| 4088 | } |
| 4089 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4090 | static void rbd_dev_release(struct device *dev) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4091 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 4092 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4093 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 4094 | if (rbd_dev->watch_event) |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 4095 | rbd_dev_header_watch_sync(rbd_dev, 0); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4096 | |
| 4097 | /* clean up and free blkdev */ |
| 4098 | rbd_free_disk(rbd_dev); |
| 4099 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
Alex Elder | 32eec68 | 2012-02-08 16:11:14 -0600 | [diff] [blame] | 4100 | |
Alex Elder | 2ac4e75 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 4101 | /* release allocated disk header fields */ |
| 4102 | rbd_header_free(&rbd_dev->header); |
| 4103 | |
Alex Elder | 32eec68 | 2012-02-08 16:11:14 -0600 | [diff] [blame] | 4104 | /* done with the id, and with the rbd_dev */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4105 | rbd_dev_id_put(rbd_dev); |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4106 | rbd_assert(rbd_dev->rbd_client != NULL); |
| 4107 | rbd_dev_destroy(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4108 | |
| 4109 | /* release module ref */ |
| 4110 | module_put(THIS_MODULE); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4111 | } |
| 4112 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4113 | static ssize_t rbd_remove(struct bus_type *bus, |
| 4114 | const char *buf, |
| 4115 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4116 | { |
| 4117 | struct rbd_device *rbd_dev = NULL; |
| 4118 | int target_id, rc; |
| 4119 | unsigned long ul; |
| 4120 | int ret = count; |
| 4121 | |
| 4122 | rc = strict_strtoul(buf, 10, &ul); |
| 4123 | if (rc) |
| 4124 | return rc; |
| 4125 | |
| 4126 | /* convert to int; abort if we lost anything in the conversion */ |
| 4127 | target_id = (int) ul; |
| 4128 | if (target_id != ul) |
| 4129 | return -EINVAL; |
| 4130 | |
| 4131 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
| 4132 | |
| 4133 | rbd_dev = __rbd_get_dev(target_id); |
| 4134 | if (!rbd_dev) { |
| 4135 | ret = -ENOENT; |
| 4136 | goto done; |
| 4137 | } |
| 4138 | |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 4139 | spin_lock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 4140 | if (rbd_dev->open_count) |
Alex Elder | 42382b7 | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 4141 | ret = -EBUSY; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 4142 | else |
| 4143 | set_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags); |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 4144 | spin_unlock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 4145 | if (ret < 0) |
Alex Elder | 42382b7 | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 4146 | goto done; |
Alex Elder | 42382b7 | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 4147 | |
Alex Elder | 41f38c2 | 2012-10-25 23:34:40 -0500 | [diff] [blame] | 4148 | rbd_remove_all_snaps(rbd_dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4149 | rbd_bus_del_dev(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4150 | |
| 4151 | done: |
| 4152 | mutex_unlock(&ctl_mutex); |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 4153 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4154 | return ret; |
| 4155 | } |
| 4156 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4157 | /* |
| 4158 | * create control files in sysfs |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4159 | * /sys/bus/rbd/... |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4160 | */ |
| 4161 | static int rbd_sysfs_init(void) |
| 4162 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4163 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4164 | |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 4165 | ret = device_register(&rbd_root_dev); |
Alex Elder | 2107978 | 2012-01-24 10:08:36 -0600 | [diff] [blame] | 4166 | if (ret < 0) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4167 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4168 | |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 4169 | ret = bus_register(&rbd_bus_type); |
| 4170 | if (ret < 0) |
| 4171 | device_unregister(&rbd_root_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4172 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4173 | return ret; |
| 4174 | } |
| 4175 | |
| 4176 | static void rbd_sysfs_cleanup(void) |
| 4177 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4178 | bus_unregister(&rbd_bus_type); |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 4179 | device_unregister(&rbd_root_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4180 | } |
| 4181 | |
| 4182 | int __init rbd_init(void) |
| 4183 | { |
| 4184 | int rc; |
| 4185 | |
Alex Elder | 1e32d34 | 2013-01-30 11:13:33 -0600 | [diff] [blame] | 4186 | if (!libceph_compatible(NULL)) { |
| 4187 | rbd_warn(NULL, "libceph incompatibility (quitting)"); |
| 4188 | |
| 4189 | return -EINVAL; |
| 4190 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4191 | rc = rbd_sysfs_init(); |
| 4192 | if (rc) |
| 4193 | return rc; |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4194 | pr_info("loaded " RBD_DRV_NAME_LONG "\n"); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4195 | return 0; |
| 4196 | } |
| 4197 | |
| 4198 | void __exit rbd_exit(void) |
| 4199 | { |
| 4200 | rbd_sysfs_cleanup(); |
| 4201 | } |
| 4202 | |
| 4203 | module_init(rbd_init); |
| 4204 | module_exit(rbd_exit); |
| 4205 | |
| 4206 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); |
| 4207 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); |
| 4208 | MODULE_DESCRIPTION("rados block device"); |
| 4209 | |
| 4210 | /* following authorship retained from original osdblk.c */ |
| 4211 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); |
| 4212 | |
| 4213 | MODULE_LICENSE("GPL"); |