Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2 | /* |
| 3 | rbd.c -- Export ceph rados objects as a Linux block device |
| 4 | |
| 5 | |
| 6 | based on drivers/block/osdblk.c: |
| 7 | |
| 8 | Copyright 2009 Red Hat, Inc. |
| 9 | |
| 10 | This program is free software; you can redistribute it and/or modify |
| 11 | it under the terms of the GNU General Public License as published by |
| 12 | the Free Software Foundation. |
| 13 | |
| 14 | This program is distributed in the hope that it will be useful, |
| 15 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 17 | GNU General Public License for more details. |
| 18 | |
| 19 | You should have received a copy of the GNU General Public License |
| 20 | along with this program; see the file COPYING. If not, write to |
| 21 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. |
| 22 | |
| 23 | |
| 24 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 25 | For usage instructions, please refer to: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 26 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 27 | Documentation/ABI/testing/sysfs-bus-rbd |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 28 | |
| 29 | */ |
| 30 | |
| 31 | #include <linux/ceph/libceph.h> |
| 32 | #include <linux/ceph/osd_client.h> |
| 33 | #include <linux/ceph/mon_client.h> |
| 34 | #include <linux/ceph/decode.h> |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 35 | #include <linux/parser.h> |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 36 | #include <linux/bsearch.h> |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 37 | |
| 38 | #include <linux/kernel.h> |
| 39 | #include <linux/device.h> |
| 40 | #include <linux/module.h> |
| 41 | #include <linux/fs.h> |
| 42 | #include <linux/blkdev.h> |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 43 | #include <linux/slab.h> |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 44 | #include <linux/idr.h> |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 45 | |
| 46 | #include "rbd_types.h" |
| 47 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 48 | #define RBD_DEBUG /* Activate rbd_assert() calls */ |
| 49 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 50 | /* |
| 51 | * The basic unit of block I/O is a sector. It is interpreted in a |
| 52 | * number of contexts in Linux (blk, bio, genhd), but the default is |
| 53 | * universally 512 bytes. These symbols are just slightly more |
| 54 | * meaningful than the bare numbers they represent. |
| 55 | */ |
| 56 | #define SECTOR_SHIFT 9 |
| 57 | #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) |
| 58 | |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 59 | /* |
| 60 | * Increment the given counter and return its updated value. |
| 61 | * If the counter is already 0 it will not be incremented. |
| 62 | * If the counter is already at its maximum value returns |
| 63 | * -EINVAL without updating it. |
| 64 | */ |
| 65 | static int atomic_inc_return_safe(atomic_t *v) |
| 66 | { |
| 67 | unsigned int counter; |
| 68 | |
| 69 | counter = (unsigned int)__atomic_add_unless(v, 1, 0); |
| 70 | if (counter <= (unsigned int)INT_MAX) |
| 71 | return (int)counter; |
| 72 | |
| 73 | atomic_dec(v); |
| 74 | |
| 75 | return -EINVAL; |
| 76 | } |
| 77 | |
| 78 | /* Decrement the counter. Return the resulting value, or -EINVAL */ |
| 79 | static int atomic_dec_return_safe(atomic_t *v) |
| 80 | { |
| 81 | int counter; |
| 82 | |
| 83 | counter = atomic_dec_return(v); |
| 84 | if (counter >= 0) |
| 85 | return counter; |
| 86 | |
| 87 | atomic_inc(v); |
| 88 | |
| 89 | return -EINVAL; |
| 90 | } |
| 91 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 92 | #define RBD_DRV_NAME "rbd" |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 93 | |
Ilya Dryomov | 7e513d4 | 2013-12-16 19:26:32 +0200 | [diff] [blame] | 94 | #define RBD_MINORS_PER_MAJOR 256 |
| 95 | #define RBD_SINGLE_MAJOR_PART_SHIFT 4 |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 96 | |
Alex Elder | d4b125e | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 97 | #define RBD_SNAP_DEV_NAME_PREFIX "snap_" |
| 98 | #define RBD_MAX_SNAP_NAME_LEN \ |
| 99 | (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1)) |
| 100 | |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 101 | #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] | 102 | |
| 103 | #define RBD_SNAP_HEAD_NAME "-" |
| 104 | |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 105 | #define BAD_SNAP_INDEX U32_MAX /* invalid index into snap array */ |
| 106 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 107 | /* This allows a single page to hold an image name sent by OSD */ |
| 108 | #define RBD_IMAGE_NAME_LEN_MAX (PAGE_SIZE - sizeof (__le32) - 1) |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 109 | #define RBD_IMAGE_ID_LEN_MAX 64 |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 110 | |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 111 | #define RBD_OBJ_PREFIX_LEN_MAX 64 |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 112 | |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 113 | /* Feature bits */ |
| 114 | |
Alex Elder | 5cbf6f12 | 2013-04-11 09:29:48 -0500 | [diff] [blame] | 115 | #define RBD_FEATURE_LAYERING (1<<0) |
| 116 | #define RBD_FEATURE_STRIPINGV2 (1<<1) |
| 117 | #define RBD_FEATURES_ALL \ |
| 118 | (RBD_FEATURE_LAYERING | RBD_FEATURE_STRIPINGV2) |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 119 | |
| 120 | /* Features supported by this (client software) implementation. */ |
| 121 | |
Alex Elder | 770eba6 | 2012-10-25 23:34:40 -0500 | [diff] [blame] | 122 | #define RBD_FEATURES_SUPPORTED (RBD_FEATURES_ALL) |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 123 | |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 124 | /* |
| 125 | * An RBD device name will be "rbd#", where the "rbd" comes from |
| 126 | * RBD_DRV_NAME above, and # is a unique integer identifier. |
| 127 | * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big |
| 128 | * enough to hold all possible device names. |
| 129 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 130 | #define DEV_NAME_LEN 32 |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 131 | #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 132 | |
| 133 | /* |
| 134 | * block device image metadata (in-memory version) |
| 135 | */ |
| 136 | struct rbd_image_header { |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 137 | /* These six fields never change for a given rbd image */ |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 138 | char *object_prefix; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 139 | __u8 obj_order; |
| 140 | __u8 crypt_type; |
| 141 | __u8 comp_type; |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 142 | u64 stripe_unit; |
| 143 | u64 stripe_count; |
| 144 | u64 features; /* Might be changeable someday? */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 145 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 146 | /* The remaining fields need to be updated occasionally */ |
| 147 | u64 image_size; |
| 148 | struct ceph_snap_context *snapc; |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 149 | char *snap_names; /* format 1 only */ |
| 150 | u64 *snap_sizes; /* format 1 only */ |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 151 | }; |
| 152 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 153 | /* |
| 154 | * An rbd image specification. |
| 155 | * |
| 156 | * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely |
Alex Elder | c66c6e0 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 157 | * identify an image. Each rbd_dev structure includes a pointer to |
| 158 | * an rbd_spec structure that encapsulates this identity. |
| 159 | * |
| 160 | * Each of the id's in an rbd_spec has an associated name. For a |
| 161 | * user-mapped image, the names are supplied and the id's associated |
| 162 | * with them are looked up. For a layered image, a parent image is |
| 163 | * defined by the tuple, and the names are looked up. |
| 164 | * |
| 165 | * An rbd_dev structure contains a parent_spec pointer which is |
| 166 | * non-null if the image it represents is a child in a layered |
| 167 | * image. This pointer will refer to the rbd_spec structure used |
| 168 | * by the parent rbd_dev for its own identity (i.e., the structure |
| 169 | * is shared between the parent and child). |
| 170 | * |
| 171 | * Since these structures are populated once, during the discovery |
| 172 | * phase of image construction, they are effectively immutable so |
| 173 | * we make no effort to synchronize access to them. |
| 174 | * |
| 175 | * Note that code herein does not assume the image name is known (it |
| 176 | * could be a null pointer). |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 177 | */ |
| 178 | struct rbd_spec { |
| 179 | u64 pool_id; |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 180 | const char *pool_name; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 181 | |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 182 | const char *image_id; |
| 183 | const char *image_name; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 184 | |
| 185 | u64 snap_id; |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 186 | const char *snap_name; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 187 | |
| 188 | struct kref kref; |
| 189 | }; |
| 190 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 191 | /* |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 192 | * an instance of the client. multiple devices may share an rbd client. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 193 | */ |
| 194 | struct rbd_client { |
| 195 | struct ceph_client *client; |
| 196 | struct kref kref; |
| 197 | struct list_head node; |
| 198 | }; |
| 199 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 200 | struct rbd_img_request; |
| 201 | typedef void (*rbd_img_callback_t)(struct rbd_img_request *); |
| 202 | |
| 203 | #define BAD_WHICH U32_MAX /* Good which or bad which, which? */ |
| 204 | |
| 205 | struct rbd_obj_request; |
| 206 | typedef void (*rbd_obj_callback_t)(struct rbd_obj_request *); |
| 207 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 208 | enum obj_request_type { |
| 209 | OBJ_REQUEST_NODATA, OBJ_REQUEST_BIO, OBJ_REQUEST_PAGES |
| 210 | }; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 211 | |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 212 | enum obj_req_flags { |
| 213 | OBJ_REQ_DONE, /* completion flag: not done = 0, done = 1 */ |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 214 | OBJ_REQ_IMG_DATA, /* object usage: standalone = 0, image = 1 */ |
Alex Elder | 5679c59 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 215 | OBJ_REQ_KNOWN, /* EXISTS flag valid: no = 0, yes = 1 */ |
| 216 | OBJ_REQ_EXISTS, /* target exists: no = 0, yes = 1 */ |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 217 | }; |
| 218 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 219 | struct rbd_obj_request { |
| 220 | const char *object_name; |
| 221 | u64 offset; /* object start byte */ |
| 222 | u64 length; /* bytes from offset */ |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 223 | unsigned long flags; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 224 | |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 225 | /* |
| 226 | * An object request associated with an image will have its |
| 227 | * img_data flag set; a standalone object request will not. |
| 228 | * |
| 229 | * A standalone object request will have which == BAD_WHICH |
| 230 | * and a null obj_request pointer. |
| 231 | * |
| 232 | * An object request initiated in support of a layered image |
| 233 | * object (to check for its existence before a write) will |
| 234 | * have which == BAD_WHICH and a non-null obj_request pointer. |
| 235 | * |
| 236 | * Finally, an object request for rbd image data will have |
| 237 | * which != BAD_WHICH, and will have a non-null img_request |
| 238 | * pointer. The value of which will be in the range |
| 239 | * 0..(img_request->obj_request_count-1). |
| 240 | */ |
| 241 | union { |
| 242 | struct rbd_obj_request *obj_request; /* STAT op */ |
| 243 | struct { |
| 244 | struct rbd_img_request *img_request; |
| 245 | u64 img_offset; |
| 246 | /* links for img_request->obj_requests list */ |
| 247 | struct list_head links; |
| 248 | }; |
| 249 | }; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 250 | u32 which; /* posn image request list */ |
| 251 | |
| 252 | enum obj_request_type type; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 253 | union { |
| 254 | struct bio *bio_list; |
| 255 | struct { |
| 256 | struct page **pages; |
| 257 | u32 page_count; |
| 258 | }; |
| 259 | }; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 260 | struct page **copyup_pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 261 | u32 copyup_page_count; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 262 | |
| 263 | struct ceph_osd_request *osd_req; |
| 264 | |
| 265 | u64 xferred; /* bytes transferred */ |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 266 | int result; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 267 | |
| 268 | rbd_obj_callback_t callback; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 269 | struct completion completion; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 270 | |
| 271 | struct kref kref; |
| 272 | }; |
| 273 | |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 274 | enum img_req_flags { |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 275 | IMG_REQ_WRITE, /* I/O direction: read = 0, write = 1 */ |
| 276 | IMG_REQ_CHILD, /* initiator: block = 0, child image = 1 */ |
Alex Elder | d0b2e94 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 277 | IMG_REQ_LAYERED, /* ENOENT handling: normal = 0, layered = 1 */ |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 278 | }; |
| 279 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 280 | struct rbd_img_request { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 281 | struct rbd_device *rbd_dev; |
| 282 | u64 offset; /* starting image byte offset */ |
| 283 | u64 length; /* byte count from offset */ |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 284 | unsigned long flags; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 285 | union { |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 286 | u64 snap_id; /* for reads */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 287 | struct ceph_snap_context *snapc; /* for writes */ |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 288 | }; |
| 289 | union { |
| 290 | struct request *rq; /* block request */ |
| 291 | struct rbd_obj_request *obj_request; /* obj req initiator */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 292 | }; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 293 | struct page **copyup_pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 294 | u32 copyup_page_count; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 295 | spinlock_t completion_lock;/* protects next_completion */ |
| 296 | u32 next_completion; |
| 297 | rbd_img_callback_t callback; |
Alex Elder | 55f27e0 | 2013-04-10 12:34:25 -0500 | [diff] [blame] | 298 | u64 xferred;/* aggregate bytes transferred */ |
Alex Elder | a5a337d | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 299 | int result; /* first nonzero obj_request result */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 300 | |
| 301 | u32 obj_request_count; |
| 302 | struct list_head obj_requests; /* rbd_obj_request structs */ |
| 303 | |
| 304 | struct kref kref; |
| 305 | }; |
| 306 | |
| 307 | #define for_each_obj_request(ireq, oreq) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 308 | list_for_each_entry(oreq, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 309 | #define for_each_obj_request_from(ireq, oreq) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 310 | list_for_each_entry_from(oreq, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 311 | #define for_each_obj_request_safe(ireq, oreq, n) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 312 | list_for_each_entry_safe_reverse(oreq, n, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 313 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 314 | struct rbd_mapping { |
Alex Elder | 99c1f08 | 2012-08-30 14:42:15 -0500 | [diff] [blame] | 315 | u64 size; |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 316 | u64 features; |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 317 | bool read_only; |
| 318 | }; |
| 319 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 320 | /* |
| 321 | * a single device |
| 322 | */ |
| 323 | struct rbd_device { |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 324 | int dev_id; /* blkdev unique id */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 325 | |
| 326 | int major; /* blkdev assigned major */ |
Ilya Dryomov | dd82fff | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 327 | int minor; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 328 | struct gendisk *disk; /* blkdev's gendisk and rq */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 329 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 330 | u32 image_format; /* Either 1 or 2 */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 331 | struct rbd_client *rbd_client; |
| 332 | |
| 333 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ |
| 334 | |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 335 | spinlock_t lock; /* queue, flags, open_count */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 336 | |
| 337 | struct rbd_image_header header; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 338 | unsigned long flags; /* possibly lock protected */ |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 339 | struct rbd_spec *spec; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 340 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 341 | char *header_name; |
Alex Elder | 971f839 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 342 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 343 | struct ceph_file_layout layout; |
| 344 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 345 | struct ceph_osd_event *watch_event; |
Alex Elder | 975241a | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 346 | struct rbd_obj_request *watch_request; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 347 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 348 | struct rbd_spec *parent_spec; |
| 349 | u64 parent_overlap; |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 350 | atomic_t parent_ref; |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 351 | struct rbd_device *parent; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 352 | |
Josh Durgin | c666601 | 2011-11-21 17:11:12 -0800 | [diff] [blame] | 353 | /* protects updating the header */ |
| 354 | struct rw_semaphore header_rwsem; |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 355 | |
| 356 | struct rbd_mapping mapping; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 357 | |
| 358 | struct list_head node; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 359 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 360 | /* sysfs related */ |
| 361 | struct device dev; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 362 | unsigned long open_count; /* protected by lock */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 363 | }; |
| 364 | |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 365 | /* |
| 366 | * Flag bits for rbd_dev->flags. If atomicity is required, |
| 367 | * rbd_dev->lock is used to protect access. |
| 368 | * |
| 369 | * Currently, only the "removing" flag (which is coupled with the |
| 370 | * "open_count" field) requires atomic access. |
| 371 | */ |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 372 | enum rbd_dev_flags { |
| 373 | RBD_DEV_FLAG_EXISTS, /* mapped snapshot has not been deleted */ |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 374 | RBD_DEV_FLAG_REMOVING, /* this mapping is being removed */ |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 375 | }; |
| 376 | |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 377 | static DEFINE_MUTEX(client_mutex); /* Serialize client creation */ |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 378 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 379 | static LIST_HEAD(rbd_dev_list); /* devices */ |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 380 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
| 381 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 382 | static LIST_HEAD(rbd_client_list); /* clients */ |
| 383 | static DEFINE_SPINLOCK(rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 384 | |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 385 | /* Slab caches for frequently-allocated structures */ |
| 386 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 387 | static struct kmem_cache *rbd_img_request_cache; |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 388 | static struct kmem_cache *rbd_obj_request_cache; |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 389 | static struct kmem_cache *rbd_segment_name_cache; |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 390 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 391 | static int rbd_major; |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 392 | static DEFINE_IDA(rbd_dev_id_ida); |
| 393 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 394 | /* |
| 395 | * Default to false for now, as single-major requires >= 0.75 version of |
| 396 | * userspace rbd utility. |
| 397 | */ |
| 398 | static bool single_major = false; |
| 399 | module_param(single_major, bool, S_IRUGO); |
| 400 | MODULE_PARM_DESC(single_major, "Use a single major number for all rbd devices (default: false)"); |
| 401 | |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 402 | static int rbd_img_request_submit(struct rbd_img_request *img_request); |
| 403 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 404 | static void rbd_dev_device_release(struct device *dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 405 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 406 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
| 407 | size_t count); |
| 408 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, |
| 409 | size_t count); |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 410 | static ssize_t rbd_add_single_major(struct bus_type *bus, const char *buf, |
| 411 | size_t count); |
| 412 | static ssize_t rbd_remove_single_major(struct bus_type *bus, const char *buf, |
| 413 | size_t count); |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 414 | static int rbd_dev_image_probe(struct rbd_device *rbd_dev, bool mapping); |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 415 | static void rbd_spec_put(struct rbd_spec *spec); |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 416 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 417 | static int rbd_dev_id_to_minor(int dev_id) |
| 418 | { |
Ilya Dryomov | 7e513d4 | 2013-12-16 19:26:32 +0200 | [diff] [blame] | 419 | return dev_id << RBD_SINGLE_MAJOR_PART_SHIFT; |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 420 | } |
| 421 | |
| 422 | static int minor_to_rbd_dev_id(int minor) |
| 423 | { |
Ilya Dryomov | 7e513d4 | 2013-12-16 19:26:32 +0200 | [diff] [blame] | 424 | return minor >> RBD_SINGLE_MAJOR_PART_SHIFT; |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 425 | } |
| 426 | |
Greg Kroah-Hartman | b15a21d | 2013-08-23 14:24:28 -0700 | [diff] [blame] | 427 | static BUS_ATTR(add, S_IWUSR, NULL, rbd_add); |
| 428 | static BUS_ATTR(remove, S_IWUSR, NULL, rbd_remove); |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 429 | static BUS_ATTR(add_single_major, S_IWUSR, NULL, rbd_add_single_major); |
| 430 | static BUS_ATTR(remove_single_major, S_IWUSR, NULL, rbd_remove_single_major); |
Greg Kroah-Hartman | b15a21d | 2013-08-23 14:24:28 -0700 | [diff] [blame] | 431 | |
| 432 | static struct attribute *rbd_bus_attrs[] = { |
| 433 | &bus_attr_add.attr, |
| 434 | &bus_attr_remove.attr, |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 435 | &bus_attr_add_single_major.attr, |
| 436 | &bus_attr_remove_single_major.attr, |
Greg Kroah-Hartman | b15a21d | 2013-08-23 14:24:28 -0700 | [diff] [blame] | 437 | NULL, |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 438 | }; |
Ilya Dryomov | 92c76dc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 439 | |
| 440 | static umode_t rbd_bus_is_visible(struct kobject *kobj, |
| 441 | struct attribute *attr, int index) |
| 442 | { |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 443 | if (!single_major && |
| 444 | (attr == &bus_attr_add_single_major.attr || |
| 445 | attr == &bus_attr_remove_single_major.attr)) |
| 446 | return 0; |
| 447 | |
Ilya Dryomov | 92c76dc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 448 | return attr->mode; |
| 449 | } |
| 450 | |
| 451 | static const struct attribute_group rbd_bus_group = { |
| 452 | .attrs = rbd_bus_attrs, |
| 453 | .is_visible = rbd_bus_is_visible, |
| 454 | }; |
| 455 | __ATTRIBUTE_GROUPS(rbd_bus); |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 456 | |
| 457 | static struct bus_type rbd_bus_type = { |
| 458 | .name = "rbd", |
Greg Kroah-Hartman | b15a21d | 2013-08-23 14:24:28 -0700 | [diff] [blame] | 459 | .bus_groups = rbd_bus_groups, |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 460 | }; |
| 461 | |
| 462 | static void rbd_root_dev_release(struct device *dev) |
| 463 | { |
| 464 | } |
| 465 | |
| 466 | static struct device rbd_root_dev = { |
| 467 | .init_name = "rbd", |
| 468 | .release = rbd_root_dev_release, |
| 469 | }; |
| 470 | |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 471 | static __printf(2, 3) |
| 472 | void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...) |
| 473 | { |
| 474 | struct va_format vaf; |
| 475 | va_list args; |
| 476 | |
| 477 | va_start(args, fmt); |
| 478 | vaf.fmt = fmt; |
| 479 | vaf.va = &args; |
| 480 | |
| 481 | if (!rbd_dev) |
| 482 | printk(KERN_WARNING "%s: %pV\n", RBD_DRV_NAME, &vaf); |
| 483 | else if (rbd_dev->disk) |
| 484 | printk(KERN_WARNING "%s: %s: %pV\n", |
| 485 | RBD_DRV_NAME, rbd_dev->disk->disk_name, &vaf); |
| 486 | else if (rbd_dev->spec && rbd_dev->spec->image_name) |
| 487 | printk(KERN_WARNING "%s: image %s: %pV\n", |
| 488 | RBD_DRV_NAME, rbd_dev->spec->image_name, &vaf); |
| 489 | else if (rbd_dev->spec && rbd_dev->spec->image_id) |
| 490 | printk(KERN_WARNING "%s: id %s: %pV\n", |
| 491 | RBD_DRV_NAME, rbd_dev->spec->image_id, &vaf); |
| 492 | else /* punt */ |
| 493 | printk(KERN_WARNING "%s: rbd_dev %p: %pV\n", |
| 494 | RBD_DRV_NAME, rbd_dev, &vaf); |
| 495 | va_end(args); |
| 496 | } |
| 497 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 498 | #ifdef RBD_DEBUG |
| 499 | #define rbd_assert(expr) \ |
| 500 | if (unlikely(!(expr))) { \ |
| 501 | printk(KERN_ERR "\nAssertion failure in %s() " \ |
| 502 | "at line %d:\n\n" \ |
| 503 | "\trbd_assert(%s);\n\n", \ |
| 504 | __func__, __LINE__, #expr); \ |
| 505 | BUG(); \ |
| 506 | } |
| 507 | #else /* !RBD_DEBUG */ |
| 508 | # define rbd_assert(expr) ((void) 0) |
| 509 | #endif /* !RBD_DEBUG */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 510 | |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 511 | static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request); |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 512 | static void rbd_img_parent_read(struct rbd_obj_request *obj_request); |
| 513 | static void rbd_dev_remove_parent(struct rbd_device *rbd_dev); |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 514 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 515 | static int rbd_dev_refresh(struct rbd_device *rbd_dev); |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 516 | static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev); |
| 517 | static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev); |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 518 | static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, |
| 519 | u64 snap_id); |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 520 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
| 521 | u8 *order, u64 *snap_size); |
| 522 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
| 523 | u64 *snap_features); |
| 524 | static u64 rbd_snap_id_by_name(struct rbd_device *rbd_dev, const char *name); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 525 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 526 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
| 527 | { |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 528 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 529 | bool removing = false; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 530 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 531 | if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 532 | return -EROFS; |
| 533 | |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 534 | spin_lock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 535 | if (test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags)) |
| 536 | removing = true; |
| 537 | else |
| 538 | rbd_dev->open_count++; |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 539 | spin_unlock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 540 | if (removing) |
| 541 | return -ENOENT; |
| 542 | |
Alex Elder | c3e946c | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 543 | (void) get_device(&rbd_dev->dev); |
Alex Elder | 340c7a2 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 544 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 545 | return 0; |
| 546 | } |
| 547 | |
Al Viro | db2a144 | 2013-05-05 21:52:57 -0400 | [diff] [blame] | 548 | static void rbd_release(struct gendisk *disk, fmode_t mode) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 549 | { |
| 550 | struct rbd_device *rbd_dev = disk->private_data; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 551 | unsigned long open_count_before; |
| 552 | |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 553 | spin_lock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 554 | open_count_before = rbd_dev->open_count--; |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 555 | spin_unlock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 556 | rbd_assert(open_count_before > 0); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 557 | |
Alex Elder | c3e946c | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 558 | put_device(&rbd_dev->dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 559 | } |
| 560 | |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 561 | static int rbd_ioctl_set_ro(struct rbd_device *rbd_dev, unsigned long arg) |
| 562 | { |
Josh Durgin | 77f33c0 | 2013-09-30 17:09:54 -0700 | [diff] [blame] | 563 | int ret = 0; |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 564 | int val; |
| 565 | bool ro; |
Josh Durgin | 77f33c0 | 2013-09-30 17:09:54 -0700 | [diff] [blame] | 566 | bool ro_changed = false; |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 567 | |
Josh Durgin | 77f33c0 | 2013-09-30 17:09:54 -0700 | [diff] [blame] | 568 | /* get_user() may sleep, so call it before taking rbd_dev->lock */ |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 569 | if (get_user(val, (int __user *)(arg))) |
| 570 | return -EFAULT; |
| 571 | |
| 572 | ro = val ? true : false; |
| 573 | /* Snapshot doesn't allow to write*/ |
| 574 | if (rbd_dev->spec->snap_id != CEPH_NOSNAP && !ro) |
| 575 | return -EROFS; |
| 576 | |
Josh Durgin | 77f33c0 | 2013-09-30 17:09:54 -0700 | [diff] [blame] | 577 | spin_lock_irq(&rbd_dev->lock); |
| 578 | /* prevent others open this device */ |
| 579 | if (rbd_dev->open_count > 1) { |
| 580 | ret = -EBUSY; |
| 581 | goto out; |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 582 | } |
| 583 | |
Josh Durgin | 77f33c0 | 2013-09-30 17:09:54 -0700 | [diff] [blame] | 584 | if (rbd_dev->mapping.read_only != ro) { |
| 585 | rbd_dev->mapping.read_only = ro; |
| 586 | ro_changed = true; |
| 587 | } |
| 588 | |
| 589 | out: |
| 590 | spin_unlock_irq(&rbd_dev->lock); |
| 591 | /* set_disk_ro() may sleep, so call it after releasing rbd_dev->lock */ |
| 592 | if (ret == 0 && ro_changed) |
| 593 | set_disk_ro(rbd_dev->disk, ro ? 1 : 0); |
| 594 | |
| 595 | return ret; |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 596 | } |
| 597 | |
| 598 | static int rbd_ioctl(struct block_device *bdev, fmode_t mode, |
| 599 | unsigned int cmd, unsigned long arg) |
| 600 | { |
| 601 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
| 602 | int ret = 0; |
| 603 | |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 604 | switch (cmd) { |
| 605 | case BLKROSET: |
| 606 | ret = rbd_ioctl_set_ro(rbd_dev, arg); |
| 607 | break; |
| 608 | default: |
| 609 | ret = -ENOTTY; |
| 610 | } |
| 611 | |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 612 | return ret; |
| 613 | } |
| 614 | |
| 615 | #ifdef CONFIG_COMPAT |
| 616 | static int rbd_compat_ioctl(struct block_device *bdev, fmode_t mode, |
| 617 | unsigned int cmd, unsigned long arg) |
| 618 | { |
| 619 | return rbd_ioctl(bdev, mode, cmd, arg); |
| 620 | } |
| 621 | #endif /* CONFIG_COMPAT */ |
| 622 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 623 | static const struct block_device_operations rbd_bd_ops = { |
| 624 | .owner = THIS_MODULE, |
| 625 | .open = rbd_open, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 626 | .release = rbd_release, |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 627 | .ioctl = rbd_ioctl, |
| 628 | #ifdef CONFIG_COMPAT |
| 629 | .compat_ioctl = rbd_compat_ioctl, |
| 630 | #endif |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 631 | }; |
| 632 | |
| 633 | /* |
Alex Elder | 7262cfc | 2013-05-16 15:04:20 -0500 | [diff] [blame] | 634 | * Initialize an rbd client instance. Success or not, this function |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 635 | * consumes ceph_opts. Caller holds client_mutex. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 636 | */ |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 637 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 638 | { |
| 639 | struct rbd_client *rbdc; |
| 640 | int ret = -ENOMEM; |
| 641 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 642 | dout("%s:\n", __func__); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 643 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); |
| 644 | if (!rbdc) |
| 645 | goto out_opt; |
| 646 | |
| 647 | kref_init(&rbdc->kref); |
| 648 | INIT_LIST_HEAD(&rbdc->node); |
| 649 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 650 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 651 | if (IS_ERR(rbdc->client)) |
Alex Elder | 08f7546 | 2013-05-29 11:19:00 -0500 | [diff] [blame] | 652 | goto out_rbdc; |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 653 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 654 | |
| 655 | ret = ceph_open_session(rbdc->client); |
| 656 | if (ret < 0) |
Alex Elder | 08f7546 | 2013-05-29 11:19:00 -0500 | [diff] [blame] | 657 | goto out_client; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 658 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 659 | spin_lock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 660 | list_add_tail(&rbdc->node, &rbd_client_list); |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 661 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 662 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 663 | dout("%s: rbdc %p\n", __func__, rbdc); |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 664 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 665 | return rbdc; |
Alex Elder | 08f7546 | 2013-05-29 11:19:00 -0500 | [diff] [blame] | 666 | out_client: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 667 | ceph_destroy_client(rbdc->client); |
Alex Elder | 08f7546 | 2013-05-29 11:19:00 -0500 | [diff] [blame] | 668 | out_rbdc: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 669 | kfree(rbdc); |
| 670 | out_opt: |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 671 | if (ceph_opts) |
| 672 | ceph_destroy_options(ceph_opts); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 673 | dout("%s: error %d\n", __func__, ret); |
| 674 | |
Vasiliy Kulikov | 28f259b | 2010-09-26 12:59:37 +0400 | [diff] [blame] | 675 | return ERR_PTR(ret); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 676 | } |
| 677 | |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 678 | static struct rbd_client *__rbd_get_client(struct rbd_client *rbdc) |
| 679 | { |
| 680 | kref_get(&rbdc->kref); |
| 681 | |
| 682 | return rbdc; |
| 683 | } |
| 684 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 685 | /* |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 686 | * Find a ceph client with specific addr and configuration. If |
| 687 | * found, bump its reference count. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 688 | */ |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 689 | static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 690 | { |
| 691 | struct rbd_client *client_node; |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 692 | bool found = false; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 693 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 694 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 695 | return NULL; |
| 696 | |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 697 | spin_lock(&rbd_client_list_lock); |
| 698 | list_for_each_entry(client_node, &rbd_client_list, node) { |
| 699 | if (!ceph_compare_options(ceph_opts, client_node->client)) { |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 700 | __rbd_get_client(client_node); |
| 701 | |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 702 | found = true; |
| 703 | break; |
| 704 | } |
| 705 | } |
| 706 | spin_unlock(&rbd_client_list_lock); |
| 707 | |
| 708 | return found ? client_node : NULL; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 709 | } |
| 710 | |
| 711 | /* |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 712 | * mount options |
| 713 | */ |
| 714 | enum { |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 715 | Opt_last_int, |
| 716 | /* int args above */ |
| 717 | Opt_last_string, |
| 718 | /* string args above */ |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 719 | Opt_read_only, |
| 720 | Opt_read_write, |
| 721 | /* Boolean args above */ |
| 722 | Opt_last_bool, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 723 | }; |
| 724 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 725 | static match_table_t rbd_opts_tokens = { |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 726 | /* int args above */ |
| 727 | /* string args above */ |
Alex Elder | be466c1 | 2012-10-22 11:31:26 -0500 | [diff] [blame] | 728 | {Opt_read_only, "read_only"}, |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 729 | {Opt_read_only, "ro"}, /* Alternate spelling */ |
| 730 | {Opt_read_write, "read_write"}, |
| 731 | {Opt_read_write, "rw"}, /* Alternate spelling */ |
| 732 | /* Boolean args above */ |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 733 | {-1, NULL} |
| 734 | }; |
| 735 | |
Alex Elder | 98571b5 | 2013-01-20 14:44:42 -0600 | [diff] [blame] | 736 | struct rbd_options { |
| 737 | bool read_only; |
| 738 | }; |
| 739 | |
| 740 | #define RBD_READ_ONLY_DEFAULT false |
| 741 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 742 | static int parse_rbd_opts_token(char *c, void *private) |
| 743 | { |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 744 | struct rbd_options *rbd_opts = private; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 745 | substring_t argstr[MAX_OPT_ARGS]; |
| 746 | int token, intval, ret; |
| 747 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 748 | token = match_token(c, rbd_opts_tokens, argstr); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 749 | if (token < 0) |
| 750 | return -EINVAL; |
| 751 | |
| 752 | if (token < Opt_last_int) { |
| 753 | ret = match_int(&argstr[0], &intval); |
| 754 | if (ret < 0) { |
| 755 | pr_err("bad mount option arg (not int) " |
| 756 | "at '%s'\n", c); |
| 757 | return ret; |
| 758 | } |
| 759 | dout("got int token %d val %d\n", token, intval); |
| 760 | } else if (token > Opt_last_int && token < Opt_last_string) { |
| 761 | dout("got string token %d val %s\n", token, |
| 762 | argstr[0].from); |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 763 | } else if (token > Opt_last_string && token < Opt_last_bool) { |
| 764 | dout("got Boolean token %d\n", token); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 765 | } else { |
| 766 | dout("got token %d\n", token); |
| 767 | } |
| 768 | |
| 769 | switch (token) { |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 770 | case Opt_read_only: |
| 771 | rbd_opts->read_only = true; |
| 772 | break; |
| 773 | case Opt_read_write: |
| 774 | rbd_opts->read_only = false; |
| 775 | break; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 776 | default: |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 777 | rbd_assert(false); |
| 778 | break; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 779 | } |
| 780 | return 0; |
| 781 | } |
| 782 | |
| 783 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 784 | * Get a ceph client with specific addr and configuration, if one does |
Alex Elder | 7262cfc | 2013-05-16 15:04:20 -0500 | [diff] [blame] | 785 | * not exist create it. Either way, ceph_opts is consumed by this |
| 786 | * function. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 787 | */ |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 788 | static struct rbd_client *rbd_get_client(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 789 | { |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 790 | struct rbd_client *rbdc; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 791 | |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 792 | mutex_lock_nested(&client_mutex, SINGLE_DEPTH_NESTING); |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 793 | rbdc = rbd_client_find(ceph_opts); |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 794 | if (rbdc) /* using an existing client */ |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 795 | ceph_destroy_options(ceph_opts); |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 796 | else |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 797 | rbdc = rbd_client_create(ceph_opts); |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 798 | mutex_unlock(&client_mutex); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 799 | |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 800 | return rbdc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 801 | } |
| 802 | |
| 803 | /* |
| 804 | * Destroy ceph client |
Alex Elder | d23a4b3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 805 | * |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 806 | * Caller must hold rbd_client_list_lock. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 807 | */ |
| 808 | static void rbd_client_release(struct kref *kref) |
| 809 | { |
| 810 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); |
| 811 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 812 | dout("%s: rbdc %p\n", __func__, rbdc); |
Alex Elder | cd9d9f5 | 2012-04-04 13:35:44 -0500 | [diff] [blame] | 813 | spin_lock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 814 | list_del(&rbdc->node); |
Alex Elder | cd9d9f5 | 2012-04-04 13:35:44 -0500 | [diff] [blame] | 815 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 816 | |
| 817 | ceph_destroy_client(rbdc->client); |
| 818 | kfree(rbdc); |
| 819 | } |
| 820 | |
| 821 | /* |
| 822 | * Drop reference to ceph client node. If it's not referenced anymore, release |
| 823 | * it. |
| 824 | */ |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 825 | static void rbd_put_client(struct rbd_client *rbdc) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 826 | { |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 827 | if (rbdc) |
| 828 | kref_put(&rbdc->kref, rbd_client_release); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 829 | } |
| 830 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 831 | static bool rbd_image_format_valid(u32 image_format) |
| 832 | { |
| 833 | return image_format == 1 || image_format == 2; |
| 834 | } |
| 835 | |
Alex Elder | 8e94af8 | 2012-07-25 09:32:40 -0500 | [diff] [blame] | 836 | static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) |
| 837 | { |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 838 | size_t size; |
| 839 | u32 snap_count; |
| 840 | |
| 841 | /* The header has to start with the magic rbd header text */ |
| 842 | if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) |
| 843 | return false; |
| 844 | |
Alex Elder | db2388b | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 845 | /* The bio layer requires at least sector-sized I/O */ |
| 846 | |
| 847 | if (ondisk->options.order < SECTOR_SHIFT) |
| 848 | return false; |
| 849 | |
| 850 | /* If we use u64 in a few spots we may be able to loosen this */ |
| 851 | |
| 852 | if (ondisk->options.order > 8 * sizeof (int) - 1) |
| 853 | return false; |
| 854 | |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 855 | /* |
| 856 | * The size of a snapshot header has to fit in a size_t, and |
| 857 | * that limits the number of snapshots. |
| 858 | */ |
| 859 | snap_count = le32_to_cpu(ondisk->snap_count); |
| 860 | size = SIZE_MAX - sizeof (struct ceph_snap_context); |
| 861 | if (snap_count > size / sizeof (__le64)) |
| 862 | return false; |
| 863 | |
| 864 | /* |
| 865 | * Not only that, but the size of the entire the snapshot |
| 866 | * header must also be representable in a size_t. |
| 867 | */ |
| 868 | size -= snap_count * sizeof (__le64); |
| 869 | if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) |
| 870 | return false; |
| 871 | |
| 872 | return true; |
Alex Elder | 8e94af8 | 2012-07-25 09:32:40 -0500 | [diff] [blame] | 873 | } |
| 874 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 875 | /* |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 876 | * Fill an rbd image header with information from the given format 1 |
| 877 | * on-disk header. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 878 | */ |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 879 | static int rbd_header_from_disk(struct rbd_device *rbd_dev, |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 880 | struct rbd_image_header_ondisk *ondisk) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 881 | { |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 882 | struct rbd_image_header *header = &rbd_dev->header; |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 883 | bool first_time = header->object_prefix == NULL; |
| 884 | struct ceph_snap_context *snapc; |
| 885 | char *object_prefix = NULL; |
| 886 | char *snap_names = NULL; |
| 887 | u64 *snap_sizes = NULL; |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 888 | u32 snap_count; |
Alex Elder | d2bb24e | 2012-07-26 23:37:14 -0500 | [diff] [blame] | 889 | size_t size; |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 890 | int ret = -ENOMEM; |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 891 | u32 i; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 892 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 893 | /* Allocate this now to avoid having to handle failure below */ |
| 894 | |
| 895 | if (first_time) { |
| 896 | size_t len; |
| 897 | |
| 898 | len = strnlen(ondisk->object_prefix, |
| 899 | sizeof (ondisk->object_prefix)); |
| 900 | object_prefix = kmalloc(len + 1, GFP_KERNEL); |
| 901 | if (!object_prefix) |
| 902 | return -ENOMEM; |
| 903 | memcpy(object_prefix, ondisk->object_prefix, len); |
| 904 | object_prefix[len] = '\0'; |
| 905 | } |
| 906 | |
| 907 | /* Allocate the snapshot context and fill it in */ |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 908 | |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 909 | snap_count = le32_to_cpu(ondisk->snap_count); |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 910 | snapc = ceph_create_snap_context(snap_count, GFP_KERNEL); |
| 911 | if (!snapc) |
| 912 | goto out_err; |
| 913 | snapc->seq = le64_to_cpu(ondisk->snap_seq); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 914 | if (snap_count) { |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 915 | struct rbd_image_snap_ondisk *snaps; |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 916 | u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
| 917 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 918 | /* We'll keep a copy of the snapshot names... */ |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 919 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 920 | if (snap_names_len > (u64)SIZE_MAX) |
| 921 | goto out_2big; |
| 922 | snap_names = kmalloc(snap_names_len, GFP_KERNEL); |
| 923 | if (!snap_names) |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 924 | goto out_err; |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 925 | |
| 926 | /* ...as well as the array of their sizes. */ |
| 927 | |
| 928 | size = snap_count * sizeof (*header->snap_sizes); |
| 929 | snap_sizes = kmalloc(size, GFP_KERNEL); |
| 930 | if (!snap_sizes) |
| 931 | goto out_err; |
| 932 | |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 933 | /* |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 934 | * Copy the names, and fill in each snapshot's id |
| 935 | * and size. |
| 936 | * |
Alex Elder | 99a41eb | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 937 | * Note that rbd_dev_v1_header_info() guarantees the |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 938 | * ondisk buffer we're working with has |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 939 | * snap_names_len bytes beyond the end of the |
| 940 | * snapshot id array, this memcpy() is safe. |
| 941 | */ |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 942 | memcpy(snap_names, &ondisk->snaps[snap_count], snap_names_len); |
| 943 | snaps = ondisk->snaps; |
| 944 | for (i = 0; i < snap_count; i++) { |
| 945 | snapc->snaps[i] = le64_to_cpu(snaps[i].id); |
| 946 | snap_sizes[i] = le64_to_cpu(snaps[i].image_size); |
| 947 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 948 | } |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 949 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 950 | /* We won't fail any more, fill in the header */ |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 951 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 952 | if (first_time) { |
| 953 | header->object_prefix = object_prefix; |
| 954 | header->obj_order = ondisk->options.order; |
| 955 | header->crypt_type = ondisk->options.crypt_type; |
| 956 | header->comp_type = ondisk->options.comp_type; |
| 957 | /* The rest aren't used for format 1 images */ |
| 958 | header->stripe_unit = 0; |
| 959 | header->stripe_count = 0; |
| 960 | header->features = 0; |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 961 | } else { |
| 962 | ceph_put_snap_context(header->snapc); |
| 963 | kfree(header->snap_names); |
| 964 | kfree(header->snap_sizes); |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 965 | } |
| 966 | |
| 967 | /* The remaining fields always get updated (when we refresh) */ |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 968 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 969 | header->image_size = le64_to_cpu(ondisk->image_size); |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 970 | header->snapc = snapc; |
| 971 | header->snap_names = snap_names; |
| 972 | header->snap_sizes = snap_sizes; |
Alex Elder | 468521c | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 973 | |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 974 | /* Make sure mapping size is consistent with header info */ |
| 975 | |
| 976 | if (rbd_dev->spec->snap_id == CEPH_NOSNAP || first_time) |
| 977 | if (rbd_dev->mapping.size != header->image_size) |
| 978 | rbd_dev->mapping.size = header->image_size; |
| 979 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 980 | return 0; |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 981 | out_2big: |
| 982 | ret = -EIO; |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 983 | out_err: |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 984 | kfree(snap_sizes); |
| 985 | kfree(snap_names); |
| 986 | ceph_put_snap_context(snapc); |
| 987 | kfree(object_prefix); |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 988 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 989 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 990 | } |
| 991 | |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 992 | static const char *_rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, u32 which) |
| 993 | { |
| 994 | const char *snap_name; |
| 995 | |
| 996 | rbd_assert(which < rbd_dev->header.snapc->num_snaps); |
| 997 | |
| 998 | /* Skip over names until we find the one we are looking for */ |
| 999 | |
| 1000 | snap_name = rbd_dev->header.snap_names; |
| 1001 | while (which--) |
| 1002 | snap_name += strlen(snap_name) + 1; |
| 1003 | |
| 1004 | return kstrdup(snap_name, GFP_KERNEL); |
| 1005 | } |
| 1006 | |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1007 | /* |
| 1008 | * Snapshot id comparison function for use with qsort()/bsearch(). |
| 1009 | * Note that result is for snapshots in *descending* order. |
| 1010 | */ |
| 1011 | static int snapid_compare_reverse(const void *s1, const void *s2) |
| 1012 | { |
| 1013 | u64 snap_id1 = *(u64 *)s1; |
| 1014 | u64 snap_id2 = *(u64 *)s2; |
| 1015 | |
| 1016 | if (snap_id1 < snap_id2) |
| 1017 | return 1; |
| 1018 | return snap_id1 == snap_id2 ? 0 : -1; |
| 1019 | } |
| 1020 | |
| 1021 | /* |
| 1022 | * Search a snapshot context to see if the given snapshot id is |
| 1023 | * present. |
| 1024 | * |
| 1025 | * Returns the position of the snapshot id in the array if it's found, |
| 1026 | * or BAD_SNAP_INDEX otherwise. |
| 1027 | * |
| 1028 | * Note: The snapshot array is in kept sorted (by the osd) in |
| 1029 | * reverse order, highest snapshot id first. |
| 1030 | */ |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1031 | static u32 rbd_dev_snap_index(struct rbd_device *rbd_dev, u64 snap_id) |
| 1032 | { |
| 1033 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1034 | u64 *found; |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1035 | |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1036 | found = bsearch(&snap_id, &snapc->snaps, snapc->num_snaps, |
| 1037 | sizeof (snap_id), snapid_compare_reverse); |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1038 | |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1039 | return found ? (u32)(found - &snapc->snaps[0]) : BAD_SNAP_INDEX; |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1040 | } |
| 1041 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1042 | static const char *rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, |
| 1043 | u64 snap_id) |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1044 | { |
| 1045 | u32 which; |
Josh Durgin | da6a6b6 | 2013-09-04 17:57:31 -0700 | [diff] [blame] | 1046 | const char *snap_name; |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1047 | |
| 1048 | which = rbd_dev_snap_index(rbd_dev, snap_id); |
| 1049 | if (which == BAD_SNAP_INDEX) |
Josh Durgin | da6a6b6 | 2013-09-04 17:57:31 -0700 | [diff] [blame] | 1050 | return ERR_PTR(-ENOENT); |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1051 | |
Josh Durgin | da6a6b6 | 2013-09-04 17:57:31 -0700 | [diff] [blame] | 1052 | snap_name = _rbd_dev_v1_snap_name(rbd_dev, which); |
| 1053 | return snap_name ? snap_name : ERR_PTR(-ENOMEM); |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1054 | } |
| 1055 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 1056 | static const char *rbd_snap_name(struct rbd_device *rbd_dev, u64 snap_id) |
| 1057 | { |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 1058 | if (snap_id == CEPH_NOSNAP) |
| 1059 | return RBD_SNAP_HEAD_NAME; |
| 1060 | |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1061 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 1062 | if (rbd_dev->image_format == 1) |
| 1063 | return rbd_dev_v1_snap_name(rbd_dev, snap_id); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 1064 | |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1065 | return rbd_dev_v2_snap_name(rbd_dev, snap_id); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 1066 | } |
| 1067 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1068 | static int rbd_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
| 1069 | u64 *snap_size) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1070 | { |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1071 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 1072 | if (snap_id == CEPH_NOSNAP) { |
| 1073 | *snap_size = rbd_dev->header.image_size; |
| 1074 | } else if (rbd_dev->image_format == 1) { |
| 1075 | u32 which; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1076 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1077 | which = rbd_dev_snap_index(rbd_dev, snap_id); |
| 1078 | if (which == BAD_SNAP_INDEX) |
| 1079 | return -ENOENT; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1080 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1081 | *snap_size = rbd_dev->header.snap_sizes[which]; |
| 1082 | } else { |
| 1083 | u64 size = 0; |
| 1084 | int ret; |
| 1085 | |
| 1086 | ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, NULL, &size); |
| 1087 | if (ret) |
| 1088 | return ret; |
| 1089 | |
| 1090 | *snap_size = size; |
| 1091 | } |
| 1092 | return 0; |
| 1093 | } |
| 1094 | |
| 1095 | static int rbd_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
| 1096 | u64 *snap_features) |
| 1097 | { |
| 1098 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 1099 | if (snap_id == CEPH_NOSNAP) { |
| 1100 | *snap_features = rbd_dev->header.features; |
| 1101 | } else if (rbd_dev->image_format == 1) { |
| 1102 | *snap_features = 0; /* No features for format 1 */ |
| 1103 | } else { |
| 1104 | u64 features = 0; |
| 1105 | int ret; |
| 1106 | |
| 1107 | ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, &features); |
| 1108 | if (ret) |
| 1109 | return ret; |
| 1110 | |
| 1111 | *snap_features = features; |
| 1112 | } |
| 1113 | return 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1114 | } |
| 1115 | |
Alex Elder | d1cf578 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 1116 | static int rbd_dev_mapping_set(struct rbd_device *rbd_dev) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1117 | { |
Alex Elder | 8f4b7d9 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 1118 | u64 snap_id = rbd_dev->spec->snap_id; |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1119 | u64 size = 0; |
| 1120 | u64 features = 0; |
| 1121 | int ret; |
Alex Elder | 8b0241f | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 1122 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1123 | ret = rbd_snap_size(rbd_dev, snap_id, &size); |
| 1124 | if (ret) |
| 1125 | return ret; |
| 1126 | ret = rbd_snap_features(rbd_dev, snap_id, &features); |
| 1127 | if (ret) |
| 1128 | return ret; |
| 1129 | |
| 1130 | rbd_dev->mapping.size = size; |
| 1131 | rbd_dev->mapping.features = features; |
| 1132 | |
Alex Elder | 8b0241f | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 1133 | return 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1134 | } |
| 1135 | |
Alex Elder | d1cf578 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 1136 | static void rbd_dev_mapping_clear(struct rbd_device *rbd_dev) |
| 1137 | { |
| 1138 | rbd_dev->mapping.size = 0; |
| 1139 | rbd_dev->mapping.features = 0; |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 1140 | } |
| 1141 | |
Alex Elder | 98571b5 | 2013-01-20 14:44:42 -0600 | [diff] [blame] | 1142 | 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] | 1143 | { |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1144 | char *name; |
| 1145 | u64 segment; |
| 1146 | int ret; |
Josh Durgin | 3a96d5c | 2013-06-12 19:15:06 -0700 | [diff] [blame] | 1147 | char *name_format; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1148 | |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 1149 | name = kmem_cache_alloc(rbd_segment_name_cache, GFP_NOIO); |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1150 | if (!name) |
| 1151 | return NULL; |
| 1152 | segment = offset >> rbd_dev->header.obj_order; |
Josh Durgin | 3a96d5c | 2013-06-12 19:15:06 -0700 | [diff] [blame] | 1153 | name_format = "%s.%012llx"; |
| 1154 | if (rbd_dev->image_format == 2) |
| 1155 | name_format = "%s.%016llx"; |
Ilya Dryomov | 2d0ebc5 | 2014-01-27 17:40:18 +0200 | [diff] [blame] | 1156 | ret = snprintf(name, CEPH_MAX_OID_NAME_LEN + 1, name_format, |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1157 | rbd_dev->header.object_prefix, segment); |
Ilya Dryomov | 2d0ebc5 | 2014-01-27 17:40:18 +0200 | [diff] [blame] | 1158 | if (ret < 0 || ret > CEPH_MAX_OID_NAME_LEN) { |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1159 | pr_err("error formatting segment name for #%llu (%d)\n", |
| 1160 | segment, ret); |
| 1161 | kfree(name); |
| 1162 | name = NULL; |
| 1163 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1164 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1165 | return name; |
| 1166 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1167 | |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 1168 | static void rbd_segment_name_free(const char *name) |
| 1169 | { |
| 1170 | /* The explicit cast here is needed to drop the const qualifier */ |
| 1171 | |
| 1172 | kmem_cache_free(rbd_segment_name_cache, (void *)name); |
| 1173 | } |
| 1174 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1175 | static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) |
| 1176 | { |
| 1177 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1178 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1179 | return offset & (segment_size - 1); |
| 1180 | } |
| 1181 | |
| 1182 | static u64 rbd_segment_length(struct rbd_device *rbd_dev, |
| 1183 | u64 offset, u64 length) |
| 1184 | { |
| 1185 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; |
| 1186 | |
| 1187 | offset &= segment_size - 1; |
| 1188 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 1189 | rbd_assert(length <= U64_MAX - offset); |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1190 | if (offset + length > segment_size) |
| 1191 | length = segment_size - offset; |
| 1192 | |
| 1193 | return length; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1194 | } |
| 1195 | |
| 1196 | /* |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 1197 | * returns the size of an object in the image |
| 1198 | */ |
| 1199 | static u64 rbd_obj_bytes(struct rbd_image_header *header) |
| 1200 | { |
| 1201 | return 1 << header->obj_order; |
| 1202 | } |
| 1203 | |
| 1204 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1205 | * bio helpers |
| 1206 | */ |
| 1207 | |
| 1208 | static void bio_chain_put(struct bio *chain) |
| 1209 | { |
| 1210 | struct bio *tmp; |
| 1211 | |
| 1212 | while (chain) { |
| 1213 | tmp = chain; |
| 1214 | chain = chain->bi_next; |
| 1215 | bio_put(tmp); |
| 1216 | } |
| 1217 | } |
| 1218 | |
| 1219 | /* |
| 1220 | * zeros a bio chain, starting at specific offset |
| 1221 | */ |
| 1222 | static void zero_bio_chain(struct bio *chain, int start_ofs) |
| 1223 | { |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1224 | struct bio_vec bv; |
| 1225 | struct bvec_iter iter; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1226 | unsigned long flags; |
| 1227 | void *buf; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1228 | int pos = 0; |
| 1229 | |
| 1230 | while (chain) { |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1231 | bio_for_each_segment(bv, chain, iter) { |
| 1232 | if (pos + bv.bv_len > start_ofs) { |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1233 | int remainder = max(start_ofs - pos, 0); |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1234 | buf = bvec_kmap_irq(&bv, &flags); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1235 | memset(buf + remainder, 0, |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1236 | bv.bv_len - remainder); |
| 1237 | flush_dcache_page(bv.bv_page); |
Dan Carpenter | 85b5aaa | 2010-10-11 21:15:11 +0200 | [diff] [blame] | 1238 | bvec_kunmap_irq(buf, &flags); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1239 | } |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1240 | pos += bv.bv_len; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1241 | } |
| 1242 | |
| 1243 | chain = chain->bi_next; |
| 1244 | } |
| 1245 | } |
| 1246 | |
| 1247 | /* |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1248 | * similar to zero_bio_chain(), zeros data defined by a page array, |
| 1249 | * starting at the given byte offset from the start of the array and |
| 1250 | * continuing up to the given end offset. The pages array is |
| 1251 | * assumed to be big enough to hold all bytes up to the end. |
| 1252 | */ |
| 1253 | static void zero_pages(struct page **pages, u64 offset, u64 end) |
| 1254 | { |
| 1255 | struct page **page = &pages[offset >> PAGE_SHIFT]; |
| 1256 | |
| 1257 | rbd_assert(end > offset); |
| 1258 | rbd_assert(end - offset <= (u64)SIZE_MAX); |
| 1259 | while (offset < end) { |
| 1260 | size_t page_offset; |
| 1261 | size_t length; |
| 1262 | unsigned long flags; |
| 1263 | void *kaddr; |
| 1264 | |
Geert Uytterhoeven | 491205a | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 1265 | page_offset = offset & ~PAGE_MASK; |
| 1266 | length = min_t(size_t, PAGE_SIZE - page_offset, end - offset); |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1267 | local_irq_save(flags); |
| 1268 | kaddr = kmap_atomic(*page); |
| 1269 | memset(kaddr + page_offset, 0, length); |
Alex Elder | e215605 | 2013-05-22 20:54:25 -0500 | [diff] [blame] | 1270 | flush_dcache_page(*page); |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1271 | kunmap_atomic(kaddr); |
| 1272 | local_irq_restore(flags); |
| 1273 | |
| 1274 | offset += length; |
| 1275 | page++; |
| 1276 | } |
| 1277 | } |
| 1278 | |
| 1279 | /* |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1280 | * Clone a portion of a bio, starting at the given byte offset |
| 1281 | * and continuing for the number of bytes indicated. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1282 | */ |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1283 | static struct bio *bio_clone_range(struct bio *bio_src, |
| 1284 | unsigned int offset, |
| 1285 | unsigned int len, |
| 1286 | gfp_t gfpmask) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1287 | { |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1288 | struct bio *bio; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1289 | |
Kent Overstreet | 5341a62 | 2013-08-07 14:31:11 -0700 | [diff] [blame] | 1290 | bio = bio_clone(bio_src, gfpmask); |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1291 | if (!bio) |
| 1292 | return NULL; /* ENOMEM */ |
| 1293 | |
Kent Overstreet | 5341a62 | 2013-08-07 14:31:11 -0700 | [diff] [blame] | 1294 | bio_advance(bio, offset); |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1295 | bio->bi_iter.bi_size = len; |
Alex Elder | 542582f | 2012-08-09 10:33:25 -0700 | [diff] [blame] | 1296 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1297 | return bio; |
| 1298 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1299 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1300 | /* |
| 1301 | * Clone a portion of a bio chain, starting at the given byte offset |
| 1302 | * into the first bio in the source chain and continuing for the |
| 1303 | * number of bytes indicated. The result is another bio chain of |
| 1304 | * exactly the given length, or a null pointer on error. |
| 1305 | * |
| 1306 | * The bio_src and offset parameters are both in-out. On entry they |
| 1307 | * refer to the first source bio and the offset into that bio where |
| 1308 | * the start of data to be cloned is located. |
| 1309 | * |
| 1310 | * On return, bio_src is updated to refer to the bio in the source |
| 1311 | * chain that contains first un-cloned byte, and *offset will |
| 1312 | * contain the offset of that byte within that bio. |
| 1313 | */ |
| 1314 | static struct bio *bio_chain_clone_range(struct bio **bio_src, |
| 1315 | unsigned int *offset, |
| 1316 | unsigned int len, |
| 1317 | gfp_t gfpmask) |
| 1318 | { |
| 1319 | struct bio *bi = *bio_src; |
| 1320 | unsigned int off = *offset; |
| 1321 | struct bio *chain = NULL; |
| 1322 | struct bio **end; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1323 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1324 | /* Build up a chain of clone bios up to the limit */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1325 | |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1326 | if (!bi || off >= bi->bi_iter.bi_size || !len) |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1327 | return NULL; /* Nothing to clone */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1328 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1329 | end = &chain; |
| 1330 | while (len) { |
| 1331 | unsigned int bi_size; |
| 1332 | struct bio *bio; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1333 | |
Alex Elder | f5400b7 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 1334 | if (!bi) { |
| 1335 | rbd_warn(NULL, "bio_chain exhausted with %u left", len); |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1336 | goto out_err; /* EINVAL; ran out of bio's */ |
Alex Elder | f5400b7 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 1337 | } |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1338 | bi_size = min_t(unsigned int, bi->bi_iter.bi_size - off, len); |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1339 | bio = bio_clone_range(bi, off, bi_size, gfpmask); |
| 1340 | if (!bio) |
| 1341 | goto out_err; /* ENOMEM */ |
| 1342 | |
| 1343 | *end = bio; |
| 1344 | end = &bio->bi_next; |
| 1345 | |
| 1346 | off += bi_size; |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1347 | if (off == bi->bi_iter.bi_size) { |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1348 | bi = bi->bi_next; |
| 1349 | off = 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1350 | } |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1351 | len -= bi_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1352 | } |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1353 | *bio_src = bi; |
| 1354 | *offset = off; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1355 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1356 | return chain; |
| 1357 | out_err: |
| 1358 | bio_chain_put(chain); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1359 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1360 | return NULL; |
| 1361 | } |
| 1362 | |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1363 | /* |
| 1364 | * The default/initial value for all object request flags is 0. For |
| 1365 | * each flag, once its value is set to 1 it is never reset to 0 |
| 1366 | * again. |
| 1367 | */ |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1368 | static void obj_request_img_data_set(struct rbd_obj_request *obj_request) |
| 1369 | { |
| 1370 | if (test_and_set_bit(OBJ_REQ_IMG_DATA, &obj_request->flags)) { |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1371 | struct rbd_device *rbd_dev; |
| 1372 | |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1373 | rbd_dev = obj_request->img_request->rbd_dev; |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1374 | rbd_warn(rbd_dev, "obj_request %p already marked img_data\n", |
| 1375 | obj_request); |
| 1376 | } |
| 1377 | } |
| 1378 | |
| 1379 | static bool obj_request_img_data_test(struct rbd_obj_request *obj_request) |
| 1380 | { |
| 1381 | smp_mb(); |
| 1382 | return test_bit(OBJ_REQ_IMG_DATA, &obj_request->flags) != 0; |
| 1383 | } |
| 1384 | |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1385 | static void obj_request_done_set(struct rbd_obj_request *obj_request) |
| 1386 | { |
| 1387 | if (test_and_set_bit(OBJ_REQ_DONE, &obj_request->flags)) { |
| 1388 | struct rbd_device *rbd_dev = NULL; |
| 1389 | |
| 1390 | if (obj_request_img_data_test(obj_request)) |
| 1391 | rbd_dev = obj_request->img_request->rbd_dev; |
| 1392 | rbd_warn(rbd_dev, "obj_request %p already marked done\n", |
| 1393 | obj_request); |
| 1394 | } |
| 1395 | } |
| 1396 | |
| 1397 | static bool obj_request_done_test(struct rbd_obj_request *obj_request) |
| 1398 | { |
| 1399 | smp_mb(); |
| 1400 | return test_bit(OBJ_REQ_DONE, &obj_request->flags) != 0; |
| 1401 | } |
| 1402 | |
Alex Elder | 5679c59 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1403 | /* |
| 1404 | * This sets the KNOWN flag after (possibly) setting the EXISTS |
| 1405 | * flag. The latter is set based on the "exists" value provided. |
| 1406 | * |
| 1407 | * Note that for our purposes once an object exists it never goes |
| 1408 | * away again. It's possible that the response from two existence |
| 1409 | * checks are separated by the creation of the target object, and |
| 1410 | * the first ("doesn't exist") response arrives *after* the second |
| 1411 | * ("does exist"). In that case we ignore the second one. |
| 1412 | */ |
| 1413 | static void obj_request_existence_set(struct rbd_obj_request *obj_request, |
| 1414 | bool exists) |
| 1415 | { |
| 1416 | if (exists) |
| 1417 | set_bit(OBJ_REQ_EXISTS, &obj_request->flags); |
| 1418 | set_bit(OBJ_REQ_KNOWN, &obj_request->flags); |
| 1419 | smp_mb(); |
| 1420 | } |
| 1421 | |
| 1422 | static bool obj_request_known_test(struct rbd_obj_request *obj_request) |
| 1423 | { |
| 1424 | smp_mb(); |
| 1425 | return test_bit(OBJ_REQ_KNOWN, &obj_request->flags) != 0; |
| 1426 | } |
| 1427 | |
| 1428 | static bool obj_request_exists_test(struct rbd_obj_request *obj_request) |
| 1429 | { |
| 1430 | smp_mb(); |
| 1431 | return test_bit(OBJ_REQ_EXISTS, &obj_request->flags) != 0; |
| 1432 | } |
| 1433 | |
Ilya Dryomov | 9638556 | 2014-06-10 13:53:29 +0400 | [diff] [blame] | 1434 | static bool obj_request_overlaps_parent(struct rbd_obj_request *obj_request) |
| 1435 | { |
| 1436 | struct rbd_device *rbd_dev = obj_request->img_request->rbd_dev; |
| 1437 | |
| 1438 | return obj_request->img_offset < |
| 1439 | round_up(rbd_dev->parent_overlap, rbd_obj_bytes(&rbd_dev->header)); |
| 1440 | } |
| 1441 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1442 | static void rbd_obj_request_get(struct rbd_obj_request *obj_request) |
| 1443 | { |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1444 | dout("%s: obj %p (was %d)\n", __func__, obj_request, |
| 1445 | atomic_read(&obj_request->kref.refcount)); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1446 | kref_get(&obj_request->kref); |
| 1447 | } |
| 1448 | |
| 1449 | static void rbd_obj_request_destroy(struct kref *kref); |
| 1450 | static void rbd_obj_request_put(struct rbd_obj_request *obj_request) |
| 1451 | { |
| 1452 | rbd_assert(obj_request != NULL); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1453 | dout("%s: obj %p (was %d)\n", __func__, obj_request, |
| 1454 | atomic_read(&obj_request->kref.refcount)); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1455 | kref_put(&obj_request->kref, rbd_obj_request_destroy); |
| 1456 | } |
| 1457 | |
Alex Elder | 0f2d5be | 2014-04-26 14:21:44 +0400 | [diff] [blame] | 1458 | static void rbd_img_request_get(struct rbd_img_request *img_request) |
| 1459 | { |
| 1460 | dout("%s: img %p (was %d)\n", __func__, img_request, |
| 1461 | atomic_read(&img_request->kref.refcount)); |
| 1462 | kref_get(&img_request->kref); |
| 1463 | } |
| 1464 | |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1465 | static bool img_request_child_test(struct rbd_img_request *img_request); |
| 1466 | static void rbd_parent_request_destroy(struct kref *kref); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1467 | static void rbd_img_request_destroy(struct kref *kref); |
| 1468 | static void rbd_img_request_put(struct rbd_img_request *img_request) |
| 1469 | { |
| 1470 | rbd_assert(img_request != NULL); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1471 | dout("%s: img %p (was %d)\n", __func__, img_request, |
| 1472 | atomic_read(&img_request->kref.refcount)); |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1473 | if (img_request_child_test(img_request)) |
| 1474 | kref_put(&img_request->kref, rbd_parent_request_destroy); |
| 1475 | else |
| 1476 | kref_put(&img_request->kref, rbd_img_request_destroy); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1477 | } |
| 1478 | |
| 1479 | static inline void rbd_img_obj_request_add(struct rbd_img_request *img_request, |
| 1480 | struct rbd_obj_request *obj_request) |
| 1481 | { |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1482 | rbd_assert(obj_request->img_request == NULL); |
| 1483 | |
Alex Elder | b155e86 | 2013-04-15 14:50:37 -0500 | [diff] [blame] | 1484 | /* Image request now owns object's original reference */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1485 | obj_request->img_request = img_request; |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1486 | obj_request->which = img_request->obj_request_count; |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1487 | rbd_assert(!obj_request_img_data_test(obj_request)); |
| 1488 | obj_request_img_data_set(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1489 | rbd_assert(obj_request->which != BAD_WHICH); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1490 | img_request->obj_request_count++; |
| 1491 | list_add_tail(&obj_request->links, &img_request->obj_requests); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1492 | dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request, |
| 1493 | obj_request->which); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1494 | } |
| 1495 | |
| 1496 | static inline void rbd_img_obj_request_del(struct rbd_img_request *img_request, |
| 1497 | struct rbd_obj_request *obj_request) |
| 1498 | { |
| 1499 | rbd_assert(obj_request->which != BAD_WHICH); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1500 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1501 | dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request, |
| 1502 | obj_request->which); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1503 | list_del(&obj_request->links); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1504 | rbd_assert(img_request->obj_request_count > 0); |
| 1505 | img_request->obj_request_count--; |
| 1506 | rbd_assert(obj_request->which == img_request->obj_request_count); |
| 1507 | obj_request->which = BAD_WHICH; |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1508 | rbd_assert(obj_request_img_data_test(obj_request)); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1509 | rbd_assert(obj_request->img_request == img_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1510 | obj_request->img_request = NULL; |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1511 | obj_request->callback = NULL; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1512 | rbd_obj_request_put(obj_request); |
| 1513 | } |
| 1514 | |
| 1515 | static bool obj_request_type_valid(enum obj_request_type type) |
| 1516 | { |
| 1517 | switch (type) { |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1518 | case OBJ_REQUEST_NODATA: |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1519 | case OBJ_REQUEST_BIO: |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1520 | case OBJ_REQUEST_PAGES: |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1521 | return true; |
| 1522 | default: |
| 1523 | return false; |
| 1524 | } |
| 1525 | } |
| 1526 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1527 | static int rbd_obj_request_submit(struct ceph_osd_client *osdc, |
| 1528 | struct rbd_obj_request *obj_request) |
| 1529 | { |
Ilya Dryomov | 71c20a0 | 2014-06-19 11:38:14 +0400 | [diff] [blame] | 1530 | dout("%s %p\n", __func__, obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1531 | return ceph_osdc_start_request(osdc, obj_request->osd_req, false); |
| 1532 | } |
| 1533 | |
Ilya Dryomov | 71c20a0 | 2014-06-19 11:38:14 +0400 | [diff] [blame] | 1534 | static void rbd_obj_request_end(struct rbd_obj_request *obj_request) |
| 1535 | { |
| 1536 | dout("%s %p\n", __func__, obj_request); |
| 1537 | ceph_osdc_cancel_request(obj_request->osd_req); |
| 1538 | } |
| 1539 | |
| 1540 | /* |
| 1541 | * Wait for an object request to complete. If interrupted, cancel the |
| 1542 | * underlying osd request. |
| 1543 | */ |
| 1544 | static int rbd_obj_request_wait(struct rbd_obj_request *obj_request) |
| 1545 | { |
| 1546 | int ret; |
| 1547 | |
| 1548 | dout("%s %p\n", __func__, obj_request); |
| 1549 | |
| 1550 | ret = wait_for_completion_interruptible(&obj_request->completion); |
| 1551 | if (ret < 0) { |
| 1552 | dout("%s %p interrupted\n", __func__, obj_request); |
| 1553 | rbd_obj_request_end(obj_request); |
| 1554 | return ret; |
| 1555 | } |
| 1556 | |
| 1557 | dout("%s %p done\n", __func__, obj_request); |
| 1558 | return 0; |
| 1559 | } |
| 1560 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1561 | static void rbd_img_request_complete(struct rbd_img_request *img_request) |
| 1562 | { |
Alex Elder | 55f27e0 | 2013-04-10 12:34:25 -0500 | [diff] [blame] | 1563 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1564 | dout("%s: img %p\n", __func__, img_request); |
Alex Elder | 55f27e0 | 2013-04-10 12:34:25 -0500 | [diff] [blame] | 1565 | |
| 1566 | /* |
| 1567 | * If no error occurred, compute the aggregate transfer |
| 1568 | * count for the image request. We could instead use |
| 1569 | * atomic64_cmpxchg() to update it as each object request |
| 1570 | * completes; not clear which way is better off hand. |
| 1571 | */ |
| 1572 | if (!img_request->result) { |
| 1573 | struct rbd_obj_request *obj_request; |
| 1574 | u64 xferred = 0; |
| 1575 | |
| 1576 | for_each_obj_request(img_request, obj_request) |
| 1577 | xferred += obj_request->xferred; |
| 1578 | img_request->xferred = xferred; |
| 1579 | } |
| 1580 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1581 | if (img_request->callback) |
| 1582 | img_request->callback(img_request); |
| 1583 | else |
| 1584 | rbd_img_request_put(img_request); |
| 1585 | } |
| 1586 | |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 1587 | /* |
| 1588 | * The default/initial value for all image request flags is 0. Each |
| 1589 | * is conditionally set to 1 at image request initialization time |
| 1590 | * and currently never change thereafter. |
| 1591 | */ |
| 1592 | static void img_request_write_set(struct rbd_img_request *img_request) |
| 1593 | { |
| 1594 | set_bit(IMG_REQ_WRITE, &img_request->flags); |
| 1595 | smp_mb(); |
| 1596 | } |
| 1597 | |
| 1598 | static bool img_request_write_test(struct rbd_img_request *img_request) |
| 1599 | { |
| 1600 | smp_mb(); |
| 1601 | return test_bit(IMG_REQ_WRITE, &img_request->flags) != 0; |
| 1602 | } |
| 1603 | |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1604 | static void img_request_child_set(struct rbd_img_request *img_request) |
| 1605 | { |
| 1606 | set_bit(IMG_REQ_CHILD, &img_request->flags); |
| 1607 | smp_mb(); |
| 1608 | } |
| 1609 | |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1610 | static void img_request_child_clear(struct rbd_img_request *img_request) |
| 1611 | { |
| 1612 | clear_bit(IMG_REQ_CHILD, &img_request->flags); |
| 1613 | smp_mb(); |
| 1614 | } |
| 1615 | |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1616 | static bool img_request_child_test(struct rbd_img_request *img_request) |
| 1617 | { |
| 1618 | smp_mb(); |
| 1619 | return test_bit(IMG_REQ_CHILD, &img_request->flags) != 0; |
| 1620 | } |
| 1621 | |
Alex Elder | d0b2e94 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1622 | static void img_request_layered_set(struct rbd_img_request *img_request) |
| 1623 | { |
| 1624 | set_bit(IMG_REQ_LAYERED, &img_request->flags); |
| 1625 | smp_mb(); |
| 1626 | } |
| 1627 | |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1628 | static void img_request_layered_clear(struct rbd_img_request *img_request) |
| 1629 | { |
| 1630 | clear_bit(IMG_REQ_LAYERED, &img_request->flags); |
| 1631 | smp_mb(); |
| 1632 | } |
| 1633 | |
Alex Elder | d0b2e94 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1634 | static bool img_request_layered_test(struct rbd_img_request *img_request) |
| 1635 | { |
| 1636 | smp_mb(); |
| 1637 | return test_bit(IMG_REQ_LAYERED, &img_request->flags) != 0; |
| 1638 | } |
| 1639 | |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1640 | static void |
| 1641 | rbd_img_obj_request_read_callback(struct rbd_obj_request *obj_request) |
| 1642 | { |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1643 | u64 xferred = obj_request->xferred; |
| 1644 | u64 length = obj_request->length; |
| 1645 | |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1646 | dout("%s: obj %p img %p result %d %llu/%llu\n", __func__, |
| 1647 | obj_request, obj_request->img_request, obj_request->result, |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1648 | xferred, length); |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1649 | /* |
Josh Durgin | 17c1cc1 | 2013-08-26 17:55:38 -0700 | [diff] [blame] | 1650 | * ENOENT means a hole in the image. We zero-fill the entire |
| 1651 | * length of the request. A short read also implies zero-fill |
| 1652 | * to the end of the request. An error requires the whole |
| 1653 | * length of the request to be reported finished with an error |
| 1654 | * to the block layer. In each case we update the xferred |
| 1655 | * count to indicate the whole request was satisfied. |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1656 | */ |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1657 | rbd_assert(obj_request->type != OBJ_REQUEST_NODATA); |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1658 | if (obj_request->result == -ENOENT) { |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1659 | if (obj_request->type == OBJ_REQUEST_BIO) |
| 1660 | zero_bio_chain(obj_request->bio_list, 0); |
| 1661 | else |
| 1662 | zero_pages(obj_request->pages, 0, length); |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1663 | obj_request->result = 0; |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1664 | } else if (xferred < length && !obj_request->result) { |
| 1665 | if (obj_request->type == OBJ_REQUEST_BIO) |
| 1666 | zero_bio_chain(obj_request->bio_list, xferred); |
| 1667 | else |
| 1668 | zero_pages(obj_request->pages, xferred, length); |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1669 | } |
Josh Durgin | 17c1cc1 | 2013-08-26 17:55:38 -0700 | [diff] [blame] | 1670 | obj_request->xferred = length; |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1671 | obj_request_done_set(obj_request); |
| 1672 | } |
| 1673 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1674 | static void rbd_obj_request_complete(struct rbd_obj_request *obj_request) |
| 1675 | { |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1676 | dout("%s: obj %p cb %p\n", __func__, obj_request, |
| 1677 | obj_request->callback); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1678 | if (obj_request->callback) |
| 1679 | obj_request->callback(obj_request); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1680 | else |
| 1681 | complete_all(&obj_request->completion); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1682 | } |
| 1683 | |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1684 | static void rbd_osd_trivial_callback(struct rbd_obj_request *obj_request) |
Alex Elder | 39bf2c5 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1685 | { |
| 1686 | dout("%s: obj %p\n", __func__, obj_request); |
| 1687 | obj_request_done_set(obj_request); |
| 1688 | } |
| 1689 | |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1690 | static void rbd_osd_read_callback(struct rbd_obj_request *obj_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1691 | { |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1692 | struct rbd_img_request *img_request = NULL; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 1693 | struct rbd_device *rbd_dev = NULL; |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1694 | bool layered = false; |
| 1695 | |
| 1696 | if (obj_request_img_data_test(obj_request)) { |
| 1697 | img_request = obj_request->img_request; |
| 1698 | layered = img_request && img_request_layered_test(img_request); |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 1699 | rbd_dev = img_request->rbd_dev; |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1700 | } |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1701 | |
| 1702 | dout("%s: obj %p img %p result %d %llu/%llu\n", __func__, |
| 1703 | obj_request, img_request, obj_request->result, |
| 1704 | obj_request->xferred, obj_request->length); |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 1705 | if (layered && obj_request->result == -ENOENT && |
| 1706 | obj_request->img_offset < rbd_dev->parent_overlap) |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1707 | rbd_img_parent_read(obj_request); |
| 1708 | else if (img_request) |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1709 | rbd_img_obj_request_read_callback(obj_request); |
| 1710 | else |
| 1711 | obj_request_done_set(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1712 | } |
| 1713 | |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1714 | static void rbd_osd_write_callback(struct rbd_obj_request *obj_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1715 | { |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 1716 | dout("%s: obj %p result %d %llu\n", __func__, obj_request, |
| 1717 | obj_request->result, obj_request->length); |
| 1718 | /* |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1719 | * There is no such thing as a successful short write. Set |
| 1720 | * it to our originally-requested length. |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 1721 | */ |
| 1722 | obj_request->xferred = obj_request->length; |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1723 | obj_request_done_set(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1724 | } |
| 1725 | |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1726 | /* |
| 1727 | * For a simple stat call there's nothing to do. We'll do more if |
| 1728 | * this is part of a write sequence for a layered image. |
| 1729 | */ |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1730 | static void rbd_osd_stat_callback(struct rbd_obj_request *obj_request) |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1731 | { |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1732 | dout("%s: obj %p\n", __func__, obj_request); |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1733 | obj_request_done_set(obj_request); |
| 1734 | } |
| 1735 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1736 | static void rbd_osd_req_callback(struct ceph_osd_request *osd_req, |
| 1737 | struct ceph_msg *msg) |
| 1738 | { |
| 1739 | struct rbd_obj_request *obj_request = osd_req->r_priv; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1740 | u16 opcode; |
| 1741 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1742 | dout("%s: osd_req %p msg %p\n", __func__, osd_req, msg); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1743 | rbd_assert(osd_req == obj_request->osd_req); |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1744 | if (obj_request_img_data_test(obj_request)) { |
| 1745 | rbd_assert(obj_request->img_request); |
| 1746 | rbd_assert(obj_request->which != BAD_WHICH); |
| 1747 | } else { |
| 1748 | rbd_assert(obj_request->which == BAD_WHICH); |
| 1749 | } |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1750 | |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 1751 | if (osd_req->r_result < 0) |
| 1752 | obj_request->result = osd_req->r_result; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1753 | |
Ilya Dryomov | 7cc69d4 | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 1754 | rbd_assert(osd_req->r_num_ops <= CEPH_OSD_MAX_OP); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1755 | |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1756 | /* |
| 1757 | * We support a 64-bit length, but ultimately it has to be |
| 1758 | * passed to blk_end_request(), which takes an unsigned int. |
| 1759 | */ |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 1760 | obj_request->xferred = osd_req->r_reply_op_len[0]; |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1761 | rbd_assert(obj_request->xferred < (u64)UINT_MAX); |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 1762 | |
Alex Elder | 7952873 | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1763 | opcode = osd_req->r_ops[0].op; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1764 | switch (opcode) { |
| 1765 | case CEPH_OSD_OP_READ: |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1766 | rbd_osd_read_callback(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1767 | break; |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 1768 | case CEPH_OSD_OP_SETALLOCHINT: |
| 1769 | rbd_assert(osd_req->r_ops[1].op == CEPH_OSD_OP_WRITE); |
| 1770 | /* fall through */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1771 | case CEPH_OSD_OP_WRITE: |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1772 | rbd_osd_write_callback(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1773 | break; |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1774 | case CEPH_OSD_OP_STAT: |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1775 | rbd_osd_stat_callback(obj_request); |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1776 | break; |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 1777 | case CEPH_OSD_OP_CALL: |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 1778 | case CEPH_OSD_OP_NOTIFY_ACK: |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1779 | case CEPH_OSD_OP_WATCH: |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1780 | rbd_osd_trivial_callback(obj_request); |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1781 | break; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1782 | default: |
| 1783 | rbd_warn(NULL, "%s: unsupported op %hu\n", |
| 1784 | obj_request->object_name, (unsigned short) opcode); |
| 1785 | break; |
| 1786 | } |
| 1787 | |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1788 | if (obj_request_done_test(obj_request)) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1789 | rbd_obj_request_complete(obj_request); |
| 1790 | } |
| 1791 | |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1792 | static void rbd_osd_req_format_read(struct rbd_obj_request *obj_request) |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1793 | { |
| 1794 | struct rbd_img_request *img_request = obj_request->img_request; |
Alex Elder | 8c042b0 | 2013-04-03 01:28:58 -0500 | [diff] [blame] | 1795 | struct ceph_osd_request *osd_req = obj_request->osd_req; |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1796 | u64 snap_id; |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1797 | |
Alex Elder | 8c042b0 | 2013-04-03 01:28:58 -0500 | [diff] [blame] | 1798 | rbd_assert(osd_req != NULL); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1799 | |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1800 | snap_id = img_request ? img_request->snap_id : CEPH_NOSNAP; |
Alex Elder | 8c042b0 | 2013-04-03 01:28:58 -0500 | [diff] [blame] | 1801 | ceph_osdc_build_request(osd_req, obj_request->offset, |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1802 | NULL, snap_id, NULL); |
| 1803 | } |
| 1804 | |
| 1805 | static void rbd_osd_req_format_write(struct rbd_obj_request *obj_request) |
| 1806 | { |
| 1807 | struct rbd_img_request *img_request = obj_request->img_request; |
| 1808 | struct ceph_osd_request *osd_req = obj_request->osd_req; |
| 1809 | struct ceph_snap_context *snapc; |
| 1810 | struct timespec mtime = CURRENT_TIME; |
| 1811 | |
| 1812 | rbd_assert(osd_req != NULL); |
| 1813 | |
| 1814 | snapc = img_request ? img_request->snapc : NULL; |
| 1815 | ceph_osdc_build_request(osd_req, obj_request->offset, |
| 1816 | snapc, CEPH_NOSNAP, &mtime); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1817 | } |
| 1818 | |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 1819 | /* |
| 1820 | * Create an osd request. A read request has one osd op (read). |
| 1821 | * A write request has either one (watch) or two (hint+write) osd ops. |
| 1822 | * (All rbd data writes are prefixed with an allocation hint op, but |
| 1823 | * technically osd watch is a write request, hence this distinction.) |
| 1824 | */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1825 | static struct ceph_osd_request *rbd_osd_req_create( |
| 1826 | struct rbd_device *rbd_dev, |
| 1827 | bool write_request, |
Ilya Dryomov | deb236b | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 1828 | unsigned int num_ops, |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1829 | struct rbd_obj_request *obj_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1830 | { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1831 | struct ceph_snap_context *snapc = NULL; |
| 1832 | struct ceph_osd_client *osdc; |
| 1833 | struct ceph_osd_request *osd_req; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1834 | |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1835 | if (obj_request_img_data_test(obj_request)) { |
| 1836 | struct rbd_img_request *img_request = obj_request->img_request; |
| 1837 | |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 1838 | rbd_assert(write_request == |
| 1839 | img_request_write_test(img_request)); |
| 1840 | if (write_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1841 | snapc = img_request->snapc; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1842 | } |
| 1843 | |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 1844 | rbd_assert(num_ops == 1 || (write_request && num_ops == 2)); |
Ilya Dryomov | deb236b | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 1845 | |
| 1846 | /* Allocate and initialize the request, for the num_ops ops */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1847 | |
| 1848 | osdc = &rbd_dev->rbd_client->client->osdc; |
Ilya Dryomov | deb236b | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 1849 | osd_req = ceph_osdc_alloc_request(osdc, snapc, num_ops, false, |
| 1850 | GFP_ATOMIC); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1851 | if (!osd_req) |
| 1852 | return NULL; /* ENOMEM */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1853 | |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1854 | if (write_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1855 | osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1856 | else |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1857 | osd_req->r_flags = CEPH_OSD_FLAG_READ; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1858 | |
| 1859 | osd_req->r_callback = rbd_osd_req_callback; |
| 1860 | osd_req->r_priv = obj_request; |
| 1861 | |
Ilya Dryomov | 3c972c9 | 2014-01-27 17:40:20 +0200 | [diff] [blame] | 1862 | osd_req->r_base_oloc.pool = ceph_file_layout_pg_pool(rbd_dev->layout); |
| 1863 | ceph_oid_set_name(&osd_req->r_base_oid, obj_request->object_name); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1864 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1865 | return osd_req; |
| 1866 | } |
| 1867 | |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1868 | /* |
| 1869 | * Create a copyup osd request based on the information in the |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 1870 | * object request supplied. A copyup request has three osd ops, |
| 1871 | * a copyup method call, a hint op, and a write op. |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1872 | */ |
| 1873 | static struct ceph_osd_request * |
| 1874 | rbd_osd_req_create_copyup(struct rbd_obj_request *obj_request) |
| 1875 | { |
| 1876 | struct rbd_img_request *img_request; |
| 1877 | struct ceph_snap_context *snapc; |
| 1878 | struct rbd_device *rbd_dev; |
| 1879 | struct ceph_osd_client *osdc; |
| 1880 | struct ceph_osd_request *osd_req; |
| 1881 | |
| 1882 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 1883 | img_request = obj_request->img_request; |
| 1884 | rbd_assert(img_request); |
| 1885 | rbd_assert(img_request_write_test(img_request)); |
| 1886 | |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 1887 | /* Allocate and initialize the request, for the three ops */ |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1888 | |
| 1889 | snapc = img_request->snapc; |
| 1890 | rbd_dev = img_request->rbd_dev; |
| 1891 | osdc = &rbd_dev->rbd_client->client->osdc; |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 1892 | osd_req = ceph_osdc_alloc_request(osdc, snapc, 3, false, GFP_ATOMIC); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1893 | if (!osd_req) |
| 1894 | return NULL; /* ENOMEM */ |
| 1895 | |
| 1896 | osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; |
| 1897 | osd_req->r_callback = rbd_osd_req_callback; |
| 1898 | osd_req->r_priv = obj_request; |
| 1899 | |
Ilya Dryomov | 3c972c9 | 2014-01-27 17:40:20 +0200 | [diff] [blame] | 1900 | osd_req->r_base_oloc.pool = ceph_file_layout_pg_pool(rbd_dev->layout); |
| 1901 | ceph_oid_set_name(&osd_req->r_base_oid, obj_request->object_name); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1902 | |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1903 | return osd_req; |
| 1904 | } |
| 1905 | |
| 1906 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1907 | static void rbd_osd_req_destroy(struct ceph_osd_request *osd_req) |
| 1908 | { |
| 1909 | ceph_osdc_put_request(osd_req); |
| 1910 | } |
| 1911 | |
| 1912 | /* object_name is assumed to be a non-null pointer and NUL-terminated */ |
| 1913 | |
| 1914 | static struct rbd_obj_request *rbd_obj_request_create(const char *object_name, |
| 1915 | u64 offset, u64 length, |
| 1916 | enum obj_request_type type) |
| 1917 | { |
| 1918 | struct rbd_obj_request *obj_request; |
| 1919 | size_t size; |
| 1920 | char *name; |
| 1921 | |
| 1922 | rbd_assert(obj_request_type_valid(type)); |
| 1923 | |
| 1924 | size = strlen(object_name) + 1; |
Alex Elder | f907ad5 | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1925 | name = kmalloc(size, GFP_KERNEL); |
| 1926 | if (!name) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1927 | return NULL; |
| 1928 | |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1929 | obj_request = kmem_cache_zalloc(rbd_obj_request_cache, GFP_KERNEL); |
Alex Elder | f907ad5 | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1930 | if (!obj_request) { |
| 1931 | kfree(name); |
| 1932 | return NULL; |
| 1933 | } |
| 1934 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1935 | obj_request->object_name = memcpy(name, object_name, size); |
| 1936 | obj_request->offset = offset; |
| 1937 | obj_request->length = length; |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1938 | obj_request->flags = 0; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1939 | obj_request->which = BAD_WHICH; |
| 1940 | obj_request->type = type; |
| 1941 | INIT_LIST_HEAD(&obj_request->links); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1942 | init_completion(&obj_request->completion); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1943 | kref_init(&obj_request->kref); |
| 1944 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1945 | dout("%s: \"%s\" %llu/%llu %d -> obj %p\n", __func__, object_name, |
| 1946 | offset, length, (int)type, obj_request); |
| 1947 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1948 | return obj_request; |
| 1949 | } |
| 1950 | |
| 1951 | static void rbd_obj_request_destroy(struct kref *kref) |
| 1952 | { |
| 1953 | struct rbd_obj_request *obj_request; |
| 1954 | |
| 1955 | obj_request = container_of(kref, struct rbd_obj_request, kref); |
| 1956 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1957 | dout("%s: obj %p\n", __func__, obj_request); |
| 1958 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1959 | rbd_assert(obj_request->img_request == NULL); |
| 1960 | rbd_assert(obj_request->which == BAD_WHICH); |
| 1961 | |
| 1962 | if (obj_request->osd_req) |
| 1963 | rbd_osd_req_destroy(obj_request->osd_req); |
| 1964 | |
| 1965 | rbd_assert(obj_request_type_valid(obj_request->type)); |
| 1966 | switch (obj_request->type) { |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1967 | case OBJ_REQUEST_NODATA: |
| 1968 | break; /* Nothing to do */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1969 | case OBJ_REQUEST_BIO: |
| 1970 | if (obj_request->bio_list) |
| 1971 | bio_chain_put(obj_request->bio_list); |
| 1972 | break; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1973 | case OBJ_REQUEST_PAGES: |
| 1974 | if (obj_request->pages) |
| 1975 | ceph_release_page_vector(obj_request->pages, |
| 1976 | obj_request->page_count); |
| 1977 | break; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1978 | } |
| 1979 | |
Alex Elder | f907ad5 | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1980 | kfree(obj_request->object_name); |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1981 | obj_request->object_name = NULL; |
| 1982 | kmem_cache_free(rbd_obj_request_cache, obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1983 | } |
| 1984 | |
Alex Elder | fb65d228 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1985 | /* It's OK to call this for a device with no parent */ |
| 1986 | |
| 1987 | static void rbd_spec_put(struct rbd_spec *spec); |
| 1988 | static void rbd_dev_unparent(struct rbd_device *rbd_dev) |
| 1989 | { |
| 1990 | rbd_dev_remove_parent(rbd_dev); |
| 1991 | rbd_spec_put(rbd_dev->parent_spec); |
| 1992 | rbd_dev->parent_spec = NULL; |
| 1993 | rbd_dev->parent_overlap = 0; |
| 1994 | } |
| 1995 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1996 | /* |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1997 | * Parent image reference counting is used to determine when an |
| 1998 | * image's parent fields can be safely torn down--after there are no |
| 1999 | * more in-flight requests to the parent image. When the last |
| 2000 | * reference is dropped, cleaning them up is safe. |
| 2001 | */ |
| 2002 | static void rbd_dev_parent_put(struct rbd_device *rbd_dev) |
| 2003 | { |
| 2004 | int counter; |
| 2005 | |
| 2006 | if (!rbd_dev->parent_spec) |
| 2007 | return; |
| 2008 | |
| 2009 | counter = atomic_dec_return_safe(&rbd_dev->parent_ref); |
| 2010 | if (counter > 0) |
| 2011 | return; |
| 2012 | |
| 2013 | /* Last reference; clean up parent data structures */ |
| 2014 | |
| 2015 | if (!counter) |
| 2016 | rbd_dev_unparent(rbd_dev); |
| 2017 | else |
| 2018 | rbd_warn(rbd_dev, "parent reference underflow\n"); |
| 2019 | } |
| 2020 | |
| 2021 | /* |
| 2022 | * If an image has a non-zero parent overlap, get a reference to its |
| 2023 | * parent. |
| 2024 | * |
Alex Elder | 392a9da | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2025 | * We must get the reference before checking for the overlap to |
| 2026 | * coordinate properly with zeroing the parent overlap in |
| 2027 | * rbd_dev_v2_parent_info() when an image gets flattened. We |
| 2028 | * drop it again if there is no overlap. |
| 2029 | * |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2030 | * Returns true if the rbd device has a parent with a non-zero |
| 2031 | * overlap and a reference for it was successfully taken, or |
| 2032 | * false otherwise. |
| 2033 | */ |
| 2034 | static bool rbd_dev_parent_get(struct rbd_device *rbd_dev) |
| 2035 | { |
| 2036 | int counter; |
| 2037 | |
| 2038 | if (!rbd_dev->parent_spec) |
| 2039 | return false; |
| 2040 | |
| 2041 | counter = atomic_inc_return_safe(&rbd_dev->parent_ref); |
| 2042 | if (counter > 0 && rbd_dev->parent_overlap) |
| 2043 | return true; |
| 2044 | |
| 2045 | /* Image was flattened, but parent is not yet torn down */ |
| 2046 | |
| 2047 | if (counter < 0) |
| 2048 | rbd_warn(rbd_dev, "parent reference overflow\n"); |
| 2049 | |
| 2050 | return false; |
| 2051 | } |
| 2052 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2053 | /* |
| 2054 | * Caller is responsible for filling in the list of object requests |
| 2055 | * that comprises the image request, and the Linux request pointer |
| 2056 | * (if there is one). |
| 2057 | */ |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 2058 | static struct rbd_img_request *rbd_img_request_create( |
| 2059 | struct rbd_device *rbd_dev, |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2060 | u64 offset, u64 length, |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2061 | bool write_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2062 | { |
| 2063 | struct rbd_img_request *img_request; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2064 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 2065 | img_request = kmem_cache_alloc(rbd_img_request_cache, GFP_ATOMIC); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2066 | if (!img_request) |
| 2067 | return NULL; |
| 2068 | |
| 2069 | if (write_request) { |
| 2070 | down_read(&rbd_dev->header_rwsem); |
Alex Elder | 812164f8 | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 2071 | ceph_get_snap_context(rbd_dev->header.snapc); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2072 | up_read(&rbd_dev->header_rwsem); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2073 | } |
| 2074 | |
| 2075 | img_request->rq = NULL; |
| 2076 | img_request->rbd_dev = rbd_dev; |
| 2077 | img_request->offset = offset; |
| 2078 | img_request->length = length; |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2079 | img_request->flags = 0; |
| 2080 | if (write_request) { |
| 2081 | img_request_write_set(img_request); |
Alex Elder | 468521c | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 2082 | img_request->snapc = rbd_dev->header.snapc; |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2083 | } else { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2084 | img_request->snap_id = rbd_dev->spec->snap_id; |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2085 | } |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2086 | if (rbd_dev_parent_get(rbd_dev)) |
Alex Elder | d0b2e94 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2087 | img_request_layered_set(img_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2088 | spin_lock_init(&img_request->completion_lock); |
| 2089 | img_request->next_completion = 0; |
| 2090 | img_request->callback = NULL; |
Alex Elder | a5a337d | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2091 | img_request->result = 0; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2092 | img_request->obj_request_count = 0; |
| 2093 | INIT_LIST_HEAD(&img_request->obj_requests); |
| 2094 | kref_init(&img_request->kref); |
| 2095 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2096 | dout("%s: rbd_dev %p %s %llu/%llu -> img %p\n", __func__, rbd_dev, |
| 2097 | write_request ? "write" : "read", offset, length, |
| 2098 | img_request); |
| 2099 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2100 | return img_request; |
| 2101 | } |
| 2102 | |
| 2103 | static void rbd_img_request_destroy(struct kref *kref) |
| 2104 | { |
| 2105 | struct rbd_img_request *img_request; |
| 2106 | struct rbd_obj_request *obj_request; |
| 2107 | struct rbd_obj_request *next_obj_request; |
| 2108 | |
| 2109 | img_request = container_of(kref, struct rbd_img_request, kref); |
| 2110 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2111 | dout("%s: img %p\n", __func__, img_request); |
| 2112 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2113 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) |
| 2114 | rbd_img_obj_request_del(img_request, obj_request); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2115 | rbd_assert(img_request->obj_request_count == 0); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2116 | |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2117 | if (img_request_layered_test(img_request)) { |
| 2118 | img_request_layered_clear(img_request); |
| 2119 | rbd_dev_parent_put(img_request->rbd_dev); |
| 2120 | } |
| 2121 | |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2122 | if (img_request_write_test(img_request)) |
Alex Elder | 812164f8 | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 2123 | ceph_put_snap_context(img_request->snapc); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2124 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 2125 | kmem_cache_free(rbd_img_request_cache, img_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2126 | } |
| 2127 | |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2128 | static struct rbd_img_request *rbd_parent_request_create( |
| 2129 | struct rbd_obj_request *obj_request, |
| 2130 | u64 img_offset, u64 length) |
| 2131 | { |
| 2132 | struct rbd_img_request *parent_request; |
| 2133 | struct rbd_device *rbd_dev; |
| 2134 | |
| 2135 | rbd_assert(obj_request->img_request); |
| 2136 | rbd_dev = obj_request->img_request->rbd_dev; |
| 2137 | |
| 2138 | parent_request = rbd_img_request_create(rbd_dev->parent, |
| 2139 | img_offset, length, false); |
| 2140 | if (!parent_request) |
| 2141 | return NULL; |
| 2142 | |
| 2143 | img_request_child_set(parent_request); |
| 2144 | rbd_obj_request_get(obj_request); |
| 2145 | parent_request->obj_request = obj_request; |
| 2146 | |
| 2147 | return parent_request; |
| 2148 | } |
| 2149 | |
| 2150 | static void rbd_parent_request_destroy(struct kref *kref) |
| 2151 | { |
| 2152 | struct rbd_img_request *parent_request; |
| 2153 | struct rbd_obj_request *orig_request; |
| 2154 | |
| 2155 | parent_request = container_of(kref, struct rbd_img_request, kref); |
| 2156 | orig_request = parent_request->obj_request; |
| 2157 | |
| 2158 | parent_request->obj_request = NULL; |
| 2159 | rbd_obj_request_put(orig_request); |
| 2160 | img_request_child_clear(parent_request); |
| 2161 | |
| 2162 | rbd_img_request_destroy(kref); |
| 2163 | } |
| 2164 | |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2165 | static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request) |
| 2166 | { |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2167 | struct rbd_img_request *img_request; |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2168 | unsigned int xferred; |
| 2169 | int result; |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2170 | bool more; |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2171 | |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2172 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 2173 | img_request = obj_request->img_request; |
| 2174 | |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2175 | rbd_assert(obj_request->xferred <= (u64)UINT_MAX); |
| 2176 | xferred = (unsigned int)obj_request->xferred; |
| 2177 | result = obj_request->result; |
| 2178 | if (result) { |
| 2179 | struct rbd_device *rbd_dev = img_request->rbd_dev; |
| 2180 | |
| 2181 | rbd_warn(rbd_dev, "%s %llx at %llx (%llx)\n", |
| 2182 | img_request_write_test(img_request) ? "write" : "read", |
| 2183 | obj_request->length, obj_request->img_offset, |
| 2184 | obj_request->offset); |
| 2185 | rbd_warn(rbd_dev, " result %d xferred %x\n", |
| 2186 | result, xferred); |
| 2187 | if (!img_request->result) |
| 2188 | img_request->result = result; |
| 2189 | } |
| 2190 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2191 | /* Image object requests don't own their page array */ |
| 2192 | |
| 2193 | if (obj_request->type == OBJ_REQUEST_PAGES) { |
| 2194 | obj_request->pages = NULL; |
| 2195 | obj_request->page_count = 0; |
| 2196 | } |
| 2197 | |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2198 | if (img_request_child_test(img_request)) { |
| 2199 | rbd_assert(img_request->obj_request != NULL); |
| 2200 | more = obj_request->which < img_request->obj_request_count - 1; |
| 2201 | } else { |
| 2202 | rbd_assert(img_request->rq != NULL); |
| 2203 | more = blk_end_request(img_request->rq, result, xferred); |
| 2204 | } |
| 2205 | |
| 2206 | return more; |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2207 | } |
| 2208 | |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2209 | static void rbd_img_obj_callback(struct rbd_obj_request *obj_request) |
| 2210 | { |
| 2211 | struct rbd_img_request *img_request; |
| 2212 | u32 which = obj_request->which; |
| 2213 | bool more = true; |
| 2214 | |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2215 | rbd_assert(obj_request_img_data_test(obj_request)); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2216 | img_request = obj_request->img_request; |
| 2217 | |
| 2218 | dout("%s: img %p obj %p\n", __func__, img_request, obj_request); |
| 2219 | rbd_assert(img_request != NULL); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2220 | rbd_assert(img_request->obj_request_count > 0); |
| 2221 | rbd_assert(which != BAD_WHICH); |
| 2222 | rbd_assert(which < img_request->obj_request_count); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2223 | |
| 2224 | spin_lock_irq(&img_request->completion_lock); |
| 2225 | if (which != img_request->next_completion) |
| 2226 | goto out; |
| 2227 | |
| 2228 | for_each_obj_request_from(img_request, obj_request) { |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2229 | rbd_assert(more); |
| 2230 | rbd_assert(which < img_request->obj_request_count); |
| 2231 | |
| 2232 | if (!obj_request_done_test(obj_request)) |
| 2233 | break; |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2234 | more = rbd_img_obj_end_request(obj_request); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2235 | which++; |
| 2236 | } |
| 2237 | |
| 2238 | rbd_assert(more ^ (which == img_request->obj_request_count)); |
| 2239 | img_request->next_completion = which; |
| 2240 | out: |
| 2241 | spin_unlock_irq(&img_request->completion_lock); |
Alex Elder | 0f2d5be | 2014-04-26 14:21:44 +0400 | [diff] [blame] | 2242 | rbd_img_request_put(img_request); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2243 | |
| 2244 | if (!more) |
| 2245 | rbd_img_request_complete(img_request); |
| 2246 | } |
| 2247 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2248 | /* |
| 2249 | * Split up an image request into one or more object requests, each |
| 2250 | * to a different object. The "type" parameter indicates whether |
| 2251 | * "data_desc" is the pointer to the head of a list of bio |
| 2252 | * structures, or the base of a page array. In either case this |
| 2253 | * function assumes data_desc describes memory sufficient to hold |
| 2254 | * all data described by the image request. |
| 2255 | */ |
| 2256 | static int rbd_img_request_fill(struct rbd_img_request *img_request, |
| 2257 | enum obj_request_type type, |
| 2258 | void *data_desc) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2259 | { |
| 2260 | struct rbd_device *rbd_dev = img_request->rbd_dev; |
| 2261 | struct rbd_obj_request *obj_request = NULL; |
| 2262 | struct rbd_obj_request *next_obj_request; |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2263 | bool write_request = img_request_write_test(img_request); |
Jingoo Han | a158073 | 2013-08-09 13:04:35 +0900 | [diff] [blame] | 2264 | struct bio *bio_list = NULL; |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2265 | unsigned int bio_offset = 0; |
Jingoo Han | a158073 | 2013-08-09 13:04:35 +0900 | [diff] [blame] | 2266 | struct page **pages = NULL; |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2267 | u64 img_offset; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2268 | u64 resid; |
| 2269 | u16 opcode; |
| 2270 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2271 | dout("%s: img %p type %d data_desc %p\n", __func__, img_request, |
| 2272 | (int)type, data_desc); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2273 | |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2274 | opcode = write_request ? CEPH_OSD_OP_WRITE : CEPH_OSD_OP_READ; |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2275 | img_offset = img_request->offset; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2276 | resid = img_request->length; |
Alex Elder | 4dda41d | 2013-02-20 21:59:33 -0600 | [diff] [blame] | 2277 | rbd_assert(resid > 0); |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2278 | |
| 2279 | if (type == OBJ_REQUEST_BIO) { |
| 2280 | bio_list = data_desc; |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 2281 | rbd_assert(img_offset == |
| 2282 | bio_list->bi_iter.bi_sector << SECTOR_SHIFT); |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2283 | } else { |
| 2284 | rbd_assert(type == OBJ_REQUEST_PAGES); |
| 2285 | pages = data_desc; |
| 2286 | } |
| 2287 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2288 | while (resid) { |
Alex Elder | 2fa1232 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2289 | struct ceph_osd_request *osd_req; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2290 | const char *object_name; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2291 | u64 offset; |
| 2292 | u64 length; |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2293 | unsigned int which = 0; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2294 | |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2295 | object_name = rbd_segment_name(rbd_dev, img_offset); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2296 | if (!object_name) |
| 2297 | goto out_unwind; |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2298 | offset = rbd_segment_offset(rbd_dev, img_offset); |
| 2299 | length = rbd_segment_length(rbd_dev, img_offset, resid); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2300 | obj_request = rbd_obj_request_create(object_name, |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2301 | offset, length, type); |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 2302 | /* object request has its own copy of the object name */ |
| 2303 | rbd_segment_name_free(object_name); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2304 | if (!obj_request) |
| 2305 | goto out_unwind; |
Ilya Dryomov | 62054da | 2014-03-04 11:57:17 +0200 | [diff] [blame] | 2306 | |
Josh Durgin | 03507db | 2013-08-27 14:45:46 -0700 | [diff] [blame] | 2307 | /* |
| 2308 | * set obj_request->img_request before creating the |
| 2309 | * osd_request so that it gets the right snapc |
| 2310 | */ |
| 2311 | rbd_img_obj_request_add(img_request, obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2312 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2313 | if (type == OBJ_REQUEST_BIO) { |
| 2314 | unsigned int clone_size; |
| 2315 | |
| 2316 | rbd_assert(length <= (u64)UINT_MAX); |
| 2317 | clone_size = (unsigned int)length; |
| 2318 | obj_request->bio_list = |
| 2319 | bio_chain_clone_range(&bio_list, |
| 2320 | &bio_offset, |
| 2321 | clone_size, |
| 2322 | GFP_ATOMIC); |
| 2323 | if (!obj_request->bio_list) |
Ilya Dryomov | 62054da | 2014-03-04 11:57:17 +0200 | [diff] [blame] | 2324 | goto out_unwind; |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2325 | } else { |
| 2326 | unsigned int page_count; |
| 2327 | |
| 2328 | obj_request->pages = pages; |
| 2329 | page_count = (u32)calc_pages_for(offset, length); |
| 2330 | obj_request->page_count = page_count; |
| 2331 | if ((offset + length) & ~PAGE_MASK) |
| 2332 | page_count--; /* more on last page */ |
| 2333 | pages += page_count; |
| 2334 | } |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2335 | |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2336 | osd_req = rbd_osd_req_create(rbd_dev, write_request, |
| 2337 | (write_request ? 2 : 1), |
Ilya Dryomov | deb236b | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 2338 | obj_request); |
Alex Elder | 2fa1232 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2339 | if (!osd_req) |
Ilya Dryomov | 62054da | 2014-03-04 11:57:17 +0200 | [diff] [blame] | 2340 | goto out_unwind; |
Alex Elder | 2fa1232 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2341 | obj_request->osd_req = osd_req; |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2342 | obj_request->callback = rbd_img_obj_callback; |
Alex Elder | 0f2d5be | 2014-04-26 14:21:44 +0400 | [diff] [blame] | 2343 | rbd_img_request_get(img_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2344 | |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2345 | if (write_request) { |
| 2346 | osd_req_op_alloc_hint_init(osd_req, which, |
| 2347 | rbd_obj_bytes(&rbd_dev->header), |
| 2348 | rbd_obj_bytes(&rbd_dev->header)); |
| 2349 | which++; |
| 2350 | } |
| 2351 | |
| 2352 | osd_req_op_extent_init(osd_req, which, opcode, offset, length, |
| 2353 | 0, 0); |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2354 | if (type == OBJ_REQUEST_BIO) |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2355 | osd_req_op_extent_osd_data_bio(osd_req, which, |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2356 | obj_request->bio_list, length); |
| 2357 | else |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2358 | osd_req_op_extent_osd_data_pages(osd_req, which, |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2359 | obj_request->pages, length, |
| 2360 | offset & ~PAGE_MASK, false, false); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2361 | |
| 2362 | if (write_request) |
| 2363 | rbd_osd_req_format_write(obj_request); |
| 2364 | else |
| 2365 | rbd_osd_req_format_read(obj_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2366 | |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2367 | obj_request->img_offset = img_offset; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2368 | |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2369 | img_offset += length; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2370 | resid -= length; |
| 2371 | } |
| 2372 | |
| 2373 | return 0; |
| 2374 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2375 | out_unwind: |
| 2376 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) |
Ilya Dryomov | 42dd037 | 2014-03-04 11:57:17 +0200 | [diff] [blame] | 2377 | rbd_img_obj_request_del(img_request, obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2378 | |
| 2379 | return -ENOMEM; |
| 2380 | } |
| 2381 | |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2382 | static void |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2383 | rbd_img_obj_copyup_callback(struct rbd_obj_request *obj_request) |
| 2384 | { |
| 2385 | struct rbd_img_request *img_request; |
| 2386 | struct rbd_device *rbd_dev; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2387 | struct page **pages; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2388 | u32 page_count; |
| 2389 | |
| 2390 | rbd_assert(obj_request->type == OBJ_REQUEST_BIO); |
| 2391 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 2392 | img_request = obj_request->img_request; |
| 2393 | rbd_assert(img_request); |
| 2394 | |
| 2395 | rbd_dev = img_request->rbd_dev; |
| 2396 | rbd_assert(rbd_dev); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2397 | |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2398 | pages = obj_request->copyup_pages; |
| 2399 | rbd_assert(pages != NULL); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2400 | obj_request->copyup_pages = NULL; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2401 | page_count = obj_request->copyup_page_count; |
| 2402 | rbd_assert(page_count); |
| 2403 | obj_request->copyup_page_count = 0; |
| 2404 | ceph_release_page_vector(pages, page_count); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2405 | |
| 2406 | /* |
| 2407 | * We want the transfer count to reflect the size of the |
| 2408 | * original write request. There is no such thing as a |
| 2409 | * successful short write, so if the request was successful |
| 2410 | * we can just set it to the originally-requested length. |
| 2411 | */ |
| 2412 | if (!obj_request->result) |
| 2413 | obj_request->xferred = obj_request->length; |
| 2414 | |
| 2415 | /* Finish up with the normal image object callback */ |
| 2416 | |
| 2417 | rbd_img_obj_callback(obj_request); |
| 2418 | } |
| 2419 | |
| 2420 | static void |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2421 | rbd_img_obj_parent_read_full_callback(struct rbd_img_request *img_request) |
| 2422 | { |
| 2423 | struct rbd_obj_request *orig_request; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2424 | struct ceph_osd_request *osd_req; |
| 2425 | struct ceph_osd_client *osdc; |
| 2426 | struct rbd_device *rbd_dev; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2427 | struct page **pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2428 | u32 page_count; |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2429 | int img_result; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2430 | u64 parent_length; |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2431 | u64 offset; |
| 2432 | u64 length; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2433 | |
| 2434 | rbd_assert(img_request_child_test(img_request)); |
| 2435 | |
| 2436 | /* First get what we need from the image request */ |
| 2437 | |
| 2438 | pages = img_request->copyup_pages; |
| 2439 | rbd_assert(pages != NULL); |
| 2440 | img_request->copyup_pages = NULL; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2441 | page_count = img_request->copyup_page_count; |
| 2442 | rbd_assert(page_count); |
| 2443 | img_request->copyup_page_count = 0; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2444 | |
| 2445 | orig_request = img_request->obj_request; |
| 2446 | rbd_assert(orig_request != NULL); |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2447 | rbd_assert(obj_request_type_valid(orig_request->type)); |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2448 | img_result = img_request->result; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2449 | parent_length = img_request->length; |
| 2450 | rbd_assert(parent_length == img_request->xferred); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2451 | rbd_img_request_put(img_request); |
| 2452 | |
Alex Elder | 91c6feb | 2013-05-06 17:40:32 -0500 | [diff] [blame] | 2453 | rbd_assert(orig_request->img_request); |
| 2454 | rbd_dev = orig_request->img_request->rbd_dev; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2455 | rbd_assert(rbd_dev); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2456 | |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2457 | /* |
| 2458 | * If the overlap has become 0 (most likely because the |
| 2459 | * image has been flattened) we need to free the pages |
| 2460 | * and re-submit the original write request. |
| 2461 | */ |
| 2462 | if (!rbd_dev->parent_overlap) { |
| 2463 | struct ceph_osd_client *osdc; |
| 2464 | |
| 2465 | ceph_release_page_vector(pages, page_count); |
| 2466 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2467 | img_result = rbd_obj_request_submit(osdc, orig_request); |
| 2468 | if (!img_result) |
| 2469 | return; |
| 2470 | } |
| 2471 | |
| 2472 | if (img_result) |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2473 | goto out_err; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2474 | |
Alex Elder | 8785b1d | 2013-05-09 10:08:49 -0500 | [diff] [blame] | 2475 | /* |
| 2476 | * The original osd request is of no use to use any more. |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2477 | * We need a new one that can hold the three ops in a copyup |
Alex Elder | 8785b1d | 2013-05-09 10:08:49 -0500 | [diff] [blame] | 2478 | * request. Allocate the new copyup osd request for the |
| 2479 | * original request, and release the old one. |
| 2480 | */ |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2481 | img_result = -ENOMEM; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2482 | osd_req = rbd_osd_req_create_copyup(orig_request); |
| 2483 | if (!osd_req) |
| 2484 | goto out_err; |
Alex Elder | 8785b1d | 2013-05-09 10:08:49 -0500 | [diff] [blame] | 2485 | rbd_osd_req_destroy(orig_request->osd_req); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2486 | orig_request->osd_req = osd_req; |
| 2487 | orig_request->copyup_pages = pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2488 | orig_request->copyup_page_count = page_count; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2489 | |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2490 | /* Initialize the copyup op */ |
| 2491 | |
| 2492 | osd_req_op_cls_init(osd_req, 0, CEPH_OSD_OP_CALL, "rbd", "copyup"); |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2493 | osd_req_op_cls_request_data_pages(osd_req, 0, pages, parent_length, 0, |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2494 | false, false); |
| 2495 | |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2496 | /* Then the hint op */ |
| 2497 | |
| 2498 | osd_req_op_alloc_hint_init(osd_req, 1, rbd_obj_bytes(&rbd_dev->header), |
| 2499 | rbd_obj_bytes(&rbd_dev->header)); |
| 2500 | |
| 2501 | /* And the original write request op */ |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2502 | |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2503 | offset = orig_request->offset; |
| 2504 | length = orig_request->length; |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2505 | osd_req_op_extent_init(osd_req, 2, CEPH_OSD_OP_WRITE, |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2506 | offset, length, 0, 0); |
| 2507 | if (orig_request->type == OBJ_REQUEST_BIO) |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2508 | osd_req_op_extent_osd_data_bio(osd_req, 2, |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2509 | orig_request->bio_list, length); |
| 2510 | else |
Ilya Dryomov | 0ccd592 | 2014-02-25 16:22:28 +0200 | [diff] [blame] | 2511 | osd_req_op_extent_osd_data_pages(osd_req, 2, |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2512 | orig_request->pages, length, |
| 2513 | offset & ~PAGE_MASK, false, false); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2514 | |
| 2515 | rbd_osd_req_format_write(orig_request); |
| 2516 | |
| 2517 | /* All set, send it off. */ |
| 2518 | |
| 2519 | orig_request->callback = rbd_img_obj_copyup_callback; |
| 2520 | osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2521 | img_result = rbd_obj_request_submit(osdc, orig_request); |
| 2522 | if (!img_result) |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2523 | return; |
| 2524 | out_err: |
| 2525 | /* Record the error code and complete the request */ |
| 2526 | |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2527 | orig_request->result = img_result; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2528 | orig_request->xferred = 0; |
| 2529 | obj_request_done_set(orig_request); |
| 2530 | rbd_obj_request_complete(orig_request); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2531 | } |
| 2532 | |
| 2533 | /* |
| 2534 | * Read from the parent image the range of data that covers the |
| 2535 | * entire target of the given object request. This is used for |
| 2536 | * satisfying a layered image write request when the target of an |
| 2537 | * object request from the image request does not exist. |
| 2538 | * |
| 2539 | * A page array big enough to hold the returned data is allocated |
| 2540 | * and supplied to rbd_img_request_fill() as the "data descriptor." |
| 2541 | * When the read completes, this page array will be transferred to |
| 2542 | * the original object request for the copyup operation. |
| 2543 | * |
| 2544 | * If an error occurs, record it as the result of the original |
| 2545 | * object request and mark it done so it gets completed. |
| 2546 | */ |
| 2547 | static int rbd_img_obj_parent_read_full(struct rbd_obj_request *obj_request) |
| 2548 | { |
| 2549 | struct rbd_img_request *img_request = NULL; |
| 2550 | struct rbd_img_request *parent_request = NULL; |
| 2551 | struct rbd_device *rbd_dev; |
| 2552 | u64 img_offset; |
| 2553 | u64 length; |
| 2554 | struct page **pages = NULL; |
| 2555 | u32 page_count; |
| 2556 | int result; |
| 2557 | |
| 2558 | rbd_assert(obj_request_img_data_test(obj_request)); |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2559 | rbd_assert(obj_request_type_valid(obj_request->type)); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2560 | |
| 2561 | img_request = obj_request->img_request; |
| 2562 | rbd_assert(img_request != NULL); |
| 2563 | rbd_dev = img_request->rbd_dev; |
| 2564 | rbd_assert(rbd_dev->parent != NULL); |
| 2565 | |
| 2566 | /* |
| 2567 | * Determine the byte range covered by the object in the |
| 2568 | * child image to which the original request was to be sent. |
| 2569 | */ |
| 2570 | img_offset = obj_request->img_offset - obj_request->offset; |
| 2571 | length = (u64)1 << rbd_dev->header.obj_order; |
| 2572 | |
| 2573 | /* |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2574 | * There is no defined parent data beyond the parent |
| 2575 | * overlap, so limit what we read at that boundary if |
| 2576 | * necessary. |
| 2577 | */ |
| 2578 | if (img_offset + length > rbd_dev->parent_overlap) { |
| 2579 | rbd_assert(img_offset < rbd_dev->parent_overlap); |
| 2580 | length = rbd_dev->parent_overlap - img_offset; |
| 2581 | } |
| 2582 | |
| 2583 | /* |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2584 | * Allocate a page array big enough to receive the data read |
| 2585 | * from the parent. |
| 2586 | */ |
| 2587 | page_count = (u32)calc_pages_for(0, length); |
| 2588 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 2589 | if (IS_ERR(pages)) { |
| 2590 | result = PTR_ERR(pages); |
| 2591 | pages = NULL; |
| 2592 | goto out_err; |
| 2593 | } |
| 2594 | |
| 2595 | result = -ENOMEM; |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2596 | parent_request = rbd_parent_request_create(obj_request, |
| 2597 | img_offset, length); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2598 | if (!parent_request) |
| 2599 | goto out_err; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2600 | |
| 2601 | result = rbd_img_request_fill(parent_request, OBJ_REQUEST_PAGES, pages); |
| 2602 | if (result) |
| 2603 | goto out_err; |
| 2604 | parent_request->copyup_pages = pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2605 | parent_request->copyup_page_count = page_count; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2606 | |
| 2607 | parent_request->callback = rbd_img_obj_parent_read_full_callback; |
| 2608 | result = rbd_img_request_submit(parent_request); |
| 2609 | if (!result) |
| 2610 | return 0; |
| 2611 | |
| 2612 | parent_request->copyup_pages = NULL; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2613 | parent_request->copyup_page_count = 0; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2614 | parent_request->obj_request = NULL; |
| 2615 | rbd_obj_request_put(obj_request); |
| 2616 | out_err: |
| 2617 | if (pages) |
| 2618 | ceph_release_page_vector(pages, page_count); |
| 2619 | if (parent_request) |
| 2620 | rbd_img_request_put(parent_request); |
| 2621 | obj_request->result = result; |
| 2622 | obj_request->xferred = 0; |
| 2623 | obj_request_done_set(obj_request); |
| 2624 | |
| 2625 | return result; |
| 2626 | } |
| 2627 | |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2628 | static void rbd_img_obj_exists_callback(struct rbd_obj_request *obj_request) |
| 2629 | { |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2630 | struct rbd_obj_request *orig_request; |
Alex Elder | 638f5ab | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2631 | struct rbd_device *rbd_dev; |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2632 | int result; |
| 2633 | |
| 2634 | rbd_assert(!obj_request_img_data_test(obj_request)); |
| 2635 | |
| 2636 | /* |
| 2637 | * All we need from the object request is the original |
| 2638 | * request and the result of the STAT op. Grab those, then |
| 2639 | * we're done with the request. |
| 2640 | */ |
| 2641 | orig_request = obj_request->obj_request; |
| 2642 | obj_request->obj_request = NULL; |
Alex Elder | 912c317 | 2013-05-13 20:35:38 -0500 | [diff] [blame] | 2643 | rbd_obj_request_put(orig_request); |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2644 | rbd_assert(orig_request); |
| 2645 | rbd_assert(orig_request->img_request); |
| 2646 | |
| 2647 | result = obj_request->result; |
| 2648 | obj_request->result = 0; |
| 2649 | |
| 2650 | dout("%s: obj %p for obj %p result %d %llu/%llu\n", __func__, |
| 2651 | obj_request, orig_request, result, |
| 2652 | obj_request->xferred, obj_request->length); |
| 2653 | rbd_obj_request_put(obj_request); |
| 2654 | |
Alex Elder | 638f5ab | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2655 | /* |
| 2656 | * If the overlap has become 0 (most likely because the |
| 2657 | * image has been flattened) we need to free the pages |
| 2658 | * and re-submit the original write request. |
| 2659 | */ |
| 2660 | rbd_dev = orig_request->img_request->rbd_dev; |
| 2661 | if (!rbd_dev->parent_overlap) { |
| 2662 | struct ceph_osd_client *osdc; |
| 2663 | |
Alex Elder | 638f5ab | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2664 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2665 | result = rbd_obj_request_submit(osdc, orig_request); |
| 2666 | if (!result) |
| 2667 | return; |
| 2668 | } |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2669 | |
| 2670 | /* |
| 2671 | * Our only purpose here is to determine whether the object |
| 2672 | * exists, and we don't want to treat the non-existence as |
| 2673 | * an error. If something else comes back, transfer the |
| 2674 | * error to the original request and complete it now. |
| 2675 | */ |
| 2676 | if (!result) { |
| 2677 | obj_request_existence_set(orig_request, true); |
| 2678 | } else if (result == -ENOENT) { |
| 2679 | obj_request_existence_set(orig_request, false); |
| 2680 | } else if (result) { |
| 2681 | orig_request->result = result; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2682 | goto out; |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2683 | } |
| 2684 | |
| 2685 | /* |
| 2686 | * Resubmit the original request now that we have recorded |
| 2687 | * whether the target object exists. |
| 2688 | */ |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2689 | orig_request->result = rbd_img_obj_request_submit(orig_request); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2690 | out: |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2691 | if (orig_request->result) |
| 2692 | rbd_obj_request_complete(orig_request); |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2693 | } |
| 2694 | |
| 2695 | static int rbd_img_obj_exists_submit(struct rbd_obj_request *obj_request) |
| 2696 | { |
| 2697 | struct rbd_obj_request *stat_request; |
| 2698 | struct rbd_device *rbd_dev; |
| 2699 | struct ceph_osd_client *osdc; |
| 2700 | struct page **pages = NULL; |
| 2701 | u32 page_count; |
| 2702 | size_t size; |
| 2703 | int ret; |
| 2704 | |
| 2705 | /* |
| 2706 | * The response data for a STAT call consists of: |
| 2707 | * le64 length; |
| 2708 | * struct { |
| 2709 | * le32 tv_sec; |
| 2710 | * le32 tv_nsec; |
| 2711 | * } mtime; |
| 2712 | */ |
| 2713 | size = sizeof (__le64) + sizeof (__le32) + sizeof (__le32); |
| 2714 | page_count = (u32)calc_pages_for(0, size); |
| 2715 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 2716 | if (IS_ERR(pages)) |
| 2717 | return PTR_ERR(pages); |
| 2718 | |
| 2719 | ret = -ENOMEM; |
| 2720 | stat_request = rbd_obj_request_create(obj_request->object_name, 0, 0, |
| 2721 | OBJ_REQUEST_PAGES); |
| 2722 | if (!stat_request) |
| 2723 | goto out; |
| 2724 | |
| 2725 | rbd_obj_request_get(obj_request); |
| 2726 | stat_request->obj_request = obj_request; |
| 2727 | stat_request->pages = pages; |
| 2728 | stat_request->page_count = page_count; |
| 2729 | |
| 2730 | rbd_assert(obj_request->img_request); |
| 2731 | rbd_dev = obj_request->img_request->rbd_dev; |
Ilya Dryomov | deb236b | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 2732 | stat_request->osd_req = rbd_osd_req_create(rbd_dev, false, 1, |
| 2733 | stat_request); |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2734 | if (!stat_request->osd_req) |
| 2735 | goto out; |
| 2736 | stat_request->callback = rbd_img_obj_exists_callback; |
| 2737 | |
| 2738 | osd_req_op_init(stat_request->osd_req, 0, CEPH_OSD_OP_STAT); |
| 2739 | osd_req_op_raw_data_in_pages(stat_request->osd_req, 0, pages, size, 0, |
| 2740 | false, false); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2741 | rbd_osd_req_format_read(stat_request); |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2742 | |
| 2743 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2744 | ret = rbd_obj_request_submit(osdc, stat_request); |
| 2745 | out: |
| 2746 | if (ret) |
| 2747 | rbd_obj_request_put(obj_request); |
| 2748 | |
| 2749 | return ret; |
| 2750 | } |
| 2751 | |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2752 | static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) |
| 2753 | { |
| 2754 | struct rbd_img_request *img_request; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2755 | struct rbd_device *rbd_dev; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2756 | bool known; |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2757 | |
| 2758 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 2759 | |
| 2760 | img_request = obj_request->img_request; |
| 2761 | rbd_assert(img_request); |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2762 | rbd_dev = img_request->rbd_dev; |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2763 | |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2764 | /* |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2765 | * Only writes to layered images need special handling. |
| 2766 | * Reads and non-layered writes are simple object requests. |
| 2767 | * Layered writes that start beyond the end of the overlap |
| 2768 | * with the parent have no parent data, so they too are |
| 2769 | * simple object requests. Finally, if the target object is |
| 2770 | * known to already exist, its parent data has already been |
| 2771 | * copied, so a write to the object can also be handled as a |
| 2772 | * simple object request. |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2773 | */ |
| 2774 | if (!img_request_write_test(img_request) || |
| 2775 | !img_request_layered_test(img_request) || |
Ilya Dryomov | 9638556 | 2014-06-10 13:53:29 +0400 | [diff] [blame] | 2776 | !obj_request_overlaps_parent(obj_request) || |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2777 | ((known = obj_request_known_test(obj_request)) && |
| 2778 | obj_request_exists_test(obj_request))) { |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2779 | |
| 2780 | struct rbd_device *rbd_dev; |
| 2781 | struct ceph_osd_client *osdc; |
| 2782 | |
| 2783 | rbd_dev = obj_request->img_request->rbd_dev; |
| 2784 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2785 | |
| 2786 | return rbd_obj_request_submit(osdc, obj_request); |
| 2787 | } |
| 2788 | |
| 2789 | /* |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2790 | * It's a layered write. The target object might exist but |
| 2791 | * we may not know that yet. If we know it doesn't exist, |
| 2792 | * start by reading the data for the full target object from |
| 2793 | * the parent so we can use it for a copyup to the target. |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2794 | */ |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2795 | if (known) |
| 2796 | return rbd_img_obj_parent_read_full(obj_request); |
| 2797 | |
| 2798 | /* We don't know whether the target exists. Go find out. */ |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2799 | |
| 2800 | return rbd_img_obj_exists_submit(obj_request); |
| 2801 | } |
| 2802 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2803 | static int rbd_img_request_submit(struct rbd_img_request *img_request) |
| 2804 | { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2805 | struct rbd_obj_request *obj_request; |
Alex Elder | 46faeed | 2013-04-10 17:47:46 -0500 | [diff] [blame] | 2806 | struct rbd_obj_request *next_obj_request; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2807 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2808 | dout("%s: img %p\n", __func__, img_request); |
Alex Elder | 46faeed | 2013-04-10 17:47:46 -0500 | [diff] [blame] | 2809 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2810 | int ret; |
| 2811 | |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2812 | ret = rbd_img_obj_request_submit(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2813 | if (ret) |
| 2814 | return ret; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2815 | } |
| 2816 | |
| 2817 | return 0; |
| 2818 | } |
| 2819 | |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2820 | static void rbd_img_parent_read_callback(struct rbd_img_request *img_request) |
| 2821 | { |
| 2822 | struct rbd_obj_request *obj_request; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2823 | struct rbd_device *rbd_dev; |
| 2824 | u64 obj_end; |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2825 | u64 img_xferred; |
| 2826 | int img_result; |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2827 | |
| 2828 | rbd_assert(img_request_child_test(img_request)); |
| 2829 | |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2830 | /* First get what we need from the image request and release it */ |
| 2831 | |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2832 | obj_request = img_request->obj_request; |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2833 | img_xferred = img_request->xferred; |
| 2834 | img_result = img_request->result; |
| 2835 | rbd_img_request_put(img_request); |
| 2836 | |
| 2837 | /* |
| 2838 | * If the overlap has become 0 (most likely because the |
| 2839 | * image has been flattened) we need to re-submit the |
| 2840 | * original request. |
| 2841 | */ |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2842 | rbd_assert(obj_request); |
| 2843 | rbd_assert(obj_request->img_request); |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2844 | rbd_dev = obj_request->img_request->rbd_dev; |
| 2845 | if (!rbd_dev->parent_overlap) { |
| 2846 | struct ceph_osd_client *osdc; |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2847 | |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2848 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2849 | img_result = rbd_obj_request_submit(osdc, obj_request); |
| 2850 | if (!img_result) |
| 2851 | return; |
| 2852 | } |
| 2853 | |
| 2854 | obj_request->result = img_result; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2855 | if (obj_request->result) |
| 2856 | goto out; |
| 2857 | |
| 2858 | /* |
| 2859 | * We need to zero anything beyond the parent overlap |
| 2860 | * boundary. Since rbd_img_obj_request_read_callback() |
| 2861 | * will zero anything beyond the end of a short read, an |
| 2862 | * easy way to do this is to pretend the data from the |
| 2863 | * parent came up short--ending at the overlap boundary. |
| 2864 | */ |
| 2865 | rbd_assert(obj_request->img_offset < U64_MAX - obj_request->length); |
| 2866 | obj_end = obj_request->img_offset + obj_request->length; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2867 | if (obj_end > rbd_dev->parent_overlap) { |
| 2868 | u64 xferred = 0; |
| 2869 | |
| 2870 | if (obj_request->img_offset < rbd_dev->parent_overlap) |
| 2871 | xferred = rbd_dev->parent_overlap - |
| 2872 | obj_request->img_offset; |
| 2873 | |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2874 | obj_request->xferred = min(img_xferred, xferred); |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2875 | } else { |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2876 | obj_request->xferred = img_xferred; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2877 | } |
| 2878 | out: |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2879 | rbd_img_obj_request_read_callback(obj_request); |
| 2880 | rbd_obj_request_complete(obj_request); |
| 2881 | } |
| 2882 | |
| 2883 | static void rbd_img_parent_read(struct rbd_obj_request *obj_request) |
| 2884 | { |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2885 | struct rbd_img_request *img_request; |
| 2886 | int result; |
| 2887 | |
| 2888 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 2889 | rbd_assert(obj_request->img_request != NULL); |
| 2890 | rbd_assert(obj_request->result == (s32) -ENOENT); |
Alex Elder | 5b2ab72 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2891 | rbd_assert(obj_request_type_valid(obj_request->type)); |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2892 | |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2893 | /* rbd_read_finish(obj_request, obj_request->length); */ |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2894 | img_request = rbd_parent_request_create(obj_request, |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2895 | obj_request->img_offset, |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2896 | obj_request->length); |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2897 | result = -ENOMEM; |
| 2898 | if (!img_request) |
| 2899 | goto out_err; |
| 2900 | |
Alex Elder | 5b2ab72 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2901 | if (obj_request->type == OBJ_REQUEST_BIO) |
| 2902 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO, |
| 2903 | obj_request->bio_list); |
| 2904 | else |
| 2905 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_PAGES, |
| 2906 | obj_request->pages); |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2907 | if (result) |
| 2908 | goto out_err; |
| 2909 | |
| 2910 | img_request->callback = rbd_img_parent_read_callback; |
| 2911 | result = rbd_img_request_submit(img_request); |
| 2912 | if (result) |
| 2913 | goto out_err; |
| 2914 | |
| 2915 | return; |
| 2916 | out_err: |
| 2917 | if (img_request) |
| 2918 | rbd_img_request_put(img_request); |
| 2919 | obj_request->result = result; |
| 2920 | obj_request->xferred = 0; |
| 2921 | obj_request_done_set(obj_request); |
| 2922 | } |
| 2923 | |
Josh Durgin | 20e0af6 | 2013-08-29 17:36:03 -0700 | [diff] [blame] | 2924 | static int rbd_obj_notify_ack_sync(struct rbd_device *rbd_dev, u64 notify_id) |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2925 | { |
| 2926 | struct rbd_obj_request *obj_request; |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2927 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2928 | int ret; |
| 2929 | |
| 2930 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, |
| 2931 | OBJ_REQUEST_NODATA); |
| 2932 | if (!obj_request) |
| 2933 | return -ENOMEM; |
| 2934 | |
| 2935 | ret = -ENOMEM; |
Ilya Dryomov | deb236b | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 2936 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, 1, |
| 2937 | obj_request); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2938 | if (!obj_request->osd_req) |
| 2939 | goto out; |
| 2940 | |
Alex Elder | c99d2d4 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 2941 | osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_NOTIFY_ACK, |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 2942 | notify_id, 0, 0); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2943 | rbd_osd_req_format_read(obj_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2944 | |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2945 | ret = rbd_obj_request_submit(osdc, obj_request); |
Alex Elder | cf81b60 | 2013-01-17 12:18:46 -0600 | [diff] [blame] | 2946 | if (ret) |
Josh Durgin | 20e0af6 | 2013-08-29 17:36:03 -0700 | [diff] [blame] | 2947 | goto out; |
| 2948 | ret = rbd_obj_request_wait(obj_request); |
| 2949 | out: |
| 2950 | rbd_obj_request_put(obj_request); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2951 | |
| 2952 | return ret; |
| 2953 | } |
| 2954 | |
| 2955 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) |
| 2956 | { |
| 2957 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
Alex Elder | e627db0 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 2958 | int ret; |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2959 | |
| 2960 | if (!rbd_dev) |
| 2961 | return; |
| 2962 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2963 | dout("%s: \"%s\" notify_id %llu opcode %u\n", __func__, |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 2964 | rbd_dev->header_name, (unsigned long long)notify_id, |
| 2965 | (unsigned int)opcode); |
Alex Elder | e627db0 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 2966 | ret = rbd_dev_refresh(rbd_dev); |
| 2967 | if (ret) |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 2968 | rbd_warn(rbd_dev, "header refresh error (%d)\n", ret); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2969 | |
Josh Durgin | 20e0af6 | 2013-08-29 17:36:03 -0700 | [diff] [blame] | 2970 | rbd_obj_notify_ack_sync(rbd_dev, notify_id); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2971 | } |
| 2972 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2973 | /* |
Ilya Dryomov | bb040aa | 2014-06-19 11:38:14 +0400 | [diff] [blame] | 2974 | * Send a (un)watch request and wait for the ack. Return a request |
| 2975 | * with a ref held on success or error. |
| 2976 | */ |
| 2977 | static struct rbd_obj_request *rbd_obj_watch_request_helper( |
| 2978 | struct rbd_device *rbd_dev, |
| 2979 | bool watch) |
| 2980 | { |
| 2981 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
| 2982 | struct rbd_obj_request *obj_request; |
| 2983 | int ret; |
| 2984 | |
| 2985 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, |
| 2986 | OBJ_REQUEST_NODATA); |
| 2987 | if (!obj_request) |
| 2988 | return ERR_PTR(-ENOMEM); |
| 2989 | |
| 2990 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, true, 1, |
| 2991 | obj_request); |
| 2992 | if (!obj_request->osd_req) { |
| 2993 | ret = -ENOMEM; |
| 2994 | goto out; |
| 2995 | } |
| 2996 | |
| 2997 | osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_WATCH, |
| 2998 | rbd_dev->watch_event->cookie, 0, watch); |
| 2999 | rbd_osd_req_format_write(obj_request); |
| 3000 | |
| 3001 | if (watch) |
| 3002 | ceph_osdc_set_request_linger(osdc, obj_request->osd_req); |
| 3003 | |
| 3004 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 3005 | if (ret) |
| 3006 | goto out; |
| 3007 | |
| 3008 | ret = rbd_obj_request_wait(obj_request); |
| 3009 | if (ret) |
| 3010 | goto out; |
| 3011 | |
| 3012 | ret = obj_request->result; |
| 3013 | if (ret) { |
| 3014 | if (watch) |
| 3015 | rbd_obj_request_end(obj_request); |
| 3016 | goto out; |
| 3017 | } |
| 3018 | |
| 3019 | return obj_request; |
| 3020 | |
| 3021 | out: |
| 3022 | rbd_obj_request_put(obj_request); |
| 3023 | return ERR_PTR(ret); |
| 3024 | } |
| 3025 | |
| 3026 | /* |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3027 | * Initiate a watch request, synchronously. |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 3028 | */ |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3029 | static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev) |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 3030 | { |
| 3031 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
| 3032 | struct rbd_obj_request *obj_request; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 3033 | int ret; |
| 3034 | |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3035 | rbd_assert(!rbd_dev->watch_event); |
| 3036 | rbd_assert(!rbd_dev->watch_request); |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 3037 | |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3038 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, rbd_dev, |
| 3039 | &rbd_dev->watch_event); |
| 3040 | if (ret < 0) |
| 3041 | return ret; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 3042 | |
Ilya Dryomov | 76756a5 | 2014-06-20 18:29:20 +0400 | [diff] [blame^] | 3043 | obj_request = rbd_obj_watch_request_helper(rbd_dev, true); |
| 3044 | if (IS_ERR(obj_request)) { |
| 3045 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
| 3046 | rbd_dev->watch_event = NULL; |
| 3047 | return PTR_ERR(obj_request); |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3048 | } |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 3049 | |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 3050 | /* |
| 3051 | * A watch request is set to linger, so the underlying osd |
| 3052 | * request won't go away until we unregister it. We retain |
| 3053 | * a pointer to the object request during that time (in |
Ilya Dryomov | 76756a5 | 2014-06-20 18:29:20 +0400 | [diff] [blame^] | 3054 | * rbd_dev->watch_request), so we'll keep a reference to it. |
| 3055 | * We'll drop that reference after we've unregistered it in |
| 3056 | * rbd_dev_header_unwatch_sync(). |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 3057 | */ |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3058 | rbd_dev->watch_request = obj_request; |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 3059 | |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3060 | return 0; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 3061 | } |
| 3062 | |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3063 | /* |
| 3064 | * Tear down a watch request, synchronously. |
| 3065 | */ |
Ilya Dryomov | 76756a5 | 2014-06-20 18:29:20 +0400 | [diff] [blame^] | 3066 | static void rbd_dev_header_unwatch_sync(struct rbd_device *rbd_dev) |
Ilya Dryomov | fca2706 | 2013-12-16 18:02:40 +0200 | [diff] [blame] | 3067 | { |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3068 | struct rbd_obj_request *obj_request; |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3069 | |
| 3070 | rbd_assert(rbd_dev->watch_event); |
| 3071 | rbd_assert(rbd_dev->watch_request); |
| 3072 | |
Ilya Dryomov | 76756a5 | 2014-06-20 18:29:20 +0400 | [diff] [blame^] | 3073 | rbd_obj_request_end(rbd_dev->watch_request); |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3074 | rbd_obj_request_put(rbd_dev->watch_request); |
| 3075 | rbd_dev->watch_request = NULL; |
| 3076 | |
Ilya Dryomov | 76756a5 | 2014-06-20 18:29:20 +0400 | [diff] [blame^] | 3077 | obj_request = rbd_obj_watch_request_helper(rbd_dev, false); |
| 3078 | if (!IS_ERR(obj_request)) |
| 3079 | rbd_obj_request_put(obj_request); |
| 3080 | else |
| 3081 | rbd_warn(rbd_dev, "unable to tear down watch request (%ld)", |
| 3082 | PTR_ERR(obj_request)); |
| 3083 | |
Ilya Dryomov | b30a01f | 2014-05-22 19:28:52 +0400 | [diff] [blame] | 3084 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
| 3085 | rbd_dev->watch_event = NULL; |
Ilya Dryomov | fca2706 | 2013-12-16 18:02:40 +0200 | [diff] [blame] | 3086 | } |
| 3087 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3088 | /* |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 3089 | * Synchronous osd object method call. Returns the number of bytes |
| 3090 | * returned in the outbound buffer, or a negative error code. |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3091 | */ |
| 3092 | static int rbd_obj_method_sync(struct rbd_device *rbd_dev, |
| 3093 | const char *object_name, |
| 3094 | const char *class_name, |
| 3095 | const char *method_name, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3096 | const void *outbound, |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3097 | size_t outbound_size, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3098 | void *inbound, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3099 | size_t inbound_size) |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3100 | { |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3101 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3102 | struct rbd_obj_request *obj_request; |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3103 | struct page **pages; |
| 3104 | u32 page_count; |
| 3105 | int ret; |
| 3106 | |
| 3107 | /* |
Alex Elder | 6010a45 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 3108 | * Method calls are ultimately read operations. The result |
| 3109 | * should placed into the inbound buffer provided. They |
| 3110 | * also supply outbound data--parameters for the object |
| 3111 | * method. Currently if this is present it will be a |
| 3112 | * snapshot id. |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3113 | */ |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3114 | page_count = (u32)calc_pages_for(0, inbound_size); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3115 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 3116 | if (IS_ERR(pages)) |
| 3117 | return PTR_ERR(pages); |
| 3118 | |
| 3119 | ret = -ENOMEM; |
Alex Elder | 6010a45 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 3120 | obj_request = rbd_obj_request_create(object_name, 0, inbound_size, |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3121 | OBJ_REQUEST_PAGES); |
| 3122 | if (!obj_request) |
| 3123 | goto out; |
| 3124 | |
| 3125 | obj_request->pages = pages; |
| 3126 | obj_request->page_count = page_count; |
| 3127 | |
Ilya Dryomov | deb236b | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 3128 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, 1, |
| 3129 | obj_request); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3130 | if (!obj_request->osd_req) |
| 3131 | goto out; |
| 3132 | |
Alex Elder | c99d2d4 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 3133 | osd_req_op_cls_init(obj_request->osd_req, 0, CEPH_OSD_OP_CALL, |
Alex Elder | 04017e2 | 2013-04-05 14:46:02 -0500 | [diff] [blame] | 3134 | class_name, method_name); |
| 3135 | if (outbound_size) { |
| 3136 | struct ceph_pagelist *pagelist; |
| 3137 | |
| 3138 | pagelist = kmalloc(sizeof (*pagelist), GFP_NOFS); |
| 3139 | if (!pagelist) |
| 3140 | goto out; |
| 3141 | |
| 3142 | ceph_pagelist_init(pagelist); |
| 3143 | ceph_pagelist_append(pagelist, outbound, outbound_size); |
| 3144 | osd_req_op_cls_request_data_pagelist(obj_request->osd_req, 0, |
| 3145 | pagelist); |
| 3146 | } |
Alex Elder | a4ce40a | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3147 | osd_req_op_cls_response_data_pages(obj_request->osd_req, 0, |
| 3148 | obj_request->pages, inbound_size, |
Alex Elder | 44cd188 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3149 | 0, false, false); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 3150 | rbd_osd_req_format_read(obj_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 3151 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3152 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 3153 | if (ret) |
| 3154 | goto out; |
| 3155 | ret = rbd_obj_request_wait(obj_request); |
| 3156 | if (ret) |
| 3157 | goto out; |
| 3158 | |
| 3159 | ret = obj_request->result; |
| 3160 | if (ret < 0) |
| 3161 | goto out; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3162 | |
| 3163 | rbd_assert(obj_request->xferred < (u64)INT_MAX); |
| 3164 | ret = (int)obj_request->xferred; |
Alex Elder | 903bb32 | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 3165 | ceph_copy_from_page_vector(pages, inbound, 0, obj_request->xferred); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3166 | out: |
| 3167 | if (obj_request) |
| 3168 | rbd_obj_request_put(obj_request); |
| 3169 | else |
| 3170 | ceph_release_page_vector(pages, page_count); |
| 3171 | |
| 3172 | return ret; |
| 3173 | } |
| 3174 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3175 | static void rbd_request_fn(struct request_queue *q) |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 3176 | __releases(q->queue_lock) __acquires(q->queue_lock) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3177 | { |
| 3178 | struct rbd_device *rbd_dev = q->queuedata; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3179 | struct request *rq; |
| 3180 | int result; |
| 3181 | |
| 3182 | while ((rq = blk_fetch_request(q))) { |
| 3183 | bool write_request = rq_data_dir(rq) == WRITE; |
| 3184 | struct rbd_img_request *img_request; |
| 3185 | u64 offset; |
| 3186 | u64 length; |
| 3187 | |
| 3188 | /* Ignore any non-FS requests that filter through. */ |
| 3189 | |
| 3190 | if (rq->cmd_type != REQ_TYPE_FS) { |
Alex Elder | 4dda41d | 2013-02-20 21:59:33 -0600 | [diff] [blame] | 3191 | dout("%s: non-fs request type %d\n", __func__, |
| 3192 | (int) rq->cmd_type); |
| 3193 | __blk_end_request_all(rq, 0); |
| 3194 | continue; |
| 3195 | } |
| 3196 | |
| 3197 | /* Ignore/skip any zero-length requests */ |
| 3198 | |
| 3199 | offset = (u64) blk_rq_pos(rq) << SECTOR_SHIFT; |
| 3200 | length = (u64) blk_rq_bytes(rq); |
| 3201 | |
| 3202 | if (!length) { |
| 3203 | dout("%s: zero-length request\n", __func__); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3204 | __blk_end_request_all(rq, 0); |
| 3205 | continue; |
| 3206 | } |
| 3207 | |
| 3208 | spin_unlock_irq(q->queue_lock); |
| 3209 | |
| 3210 | /* Disallow writes to a read-only device */ |
| 3211 | |
| 3212 | if (write_request) { |
| 3213 | result = -EROFS; |
Guangliang Zhao | 131fd9f | 2013-09-24 11:25:36 +0800 | [diff] [blame] | 3214 | if (rbd_dev->mapping.read_only) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3215 | goto end_request; |
| 3216 | rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP); |
| 3217 | } |
| 3218 | |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 3219 | /* |
| 3220 | * Quit early if the mapped snapshot no longer |
| 3221 | * exists. It's still possible the snapshot will |
| 3222 | * have disappeared by the time our request arrives |
| 3223 | * at the osd, but there's no sense in sending it if |
| 3224 | * we already know. |
| 3225 | */ |
| 3226 | if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3227 | dout("request for non-existent snapshot"); |
| 3228 | rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); |
| 3229 | result = -ENXIO; |
| 3230 | goto end_request; |
| 3231 | } |
| 3232 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3233 | result = -EINVAL; |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3234 | if (offset && length > U64_MAX - offset + 1) { |
| 3235 | rbd_warn(rbd_dev, "bad request range (%llu~%llu)\n", |
| 3236 | offset, length); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3237 | goto end_request; /* Shouldn't happen */ |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3238 | } |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3239 | |
Alex Elder | 00a653e | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3240 | result = -EIO; |
| 3241 | if (offset + length > rbd_dev->mapping.size) { |
| 3242 | rbd_warn(rbd_dev, "beyond EOD (%llu~%llu > %llu)\n", |
| 3243 | offset, length, rbd_dev->mapping.size); |
| 3244 | goto end_request; |
| 3245 | } |
| 3246 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3247 | result = -ENOMEM; |
| 3248 | img_request = rbd_img_request_create(rbd_dev, offset, length, |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 3249 | write_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3250 | if (!img_request) |
| 3251 | goto end_request; |
| 3252 | |
| 3253 | img_request->rq = rq; |
| 3254 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 3255 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO, |
| 3256 | rq->bio); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3257 | if (!result) |
| 3258 | result = rbd_img_request_submit(img_request); |
| 3259 | if (result) |
| 3260 | rbd_img_request_put(img_request); |
| 3261 | end_request: |
| 3262 | spin_lock_irq(q->queue_lock); |
| 3263 | if (result < 0) { |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 3264 | rbd_warn(rbd_dev, "%s %llx at %llx result %d\n", |
| 3265 | write_request ? "write" : "read", |
| 3266 | length, offset, result); |
| 3267 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3268 | __blk_end_request_all(rq, result); |
| 3269 | } |
| 3270 | } |
| 3271 | } |
| 3272 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3273 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3274 | * a queue callback. Makes sure that we don't create a bio that spans across |
| 3275 | * multiple osd objects. One exception would be with a single page bios, |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 3276 | * which we handle later at bio_chain_clone_range() |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3277 | */ |
| 3278 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, |
| 3279 | struct bio_vec *bvec) |
| 3280 | { |
| 3281 | struct rbd_device *rbd_dev = q->queuedata; |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 3282 | sector_t sector_offset; |
| 3283 | sector_t sectors_per_obj; |
| 3284 | sector_t obj_sector_offset; |
| 3285 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3286 | |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 3287 | /* |
| 3288 | * Find how far into its rbd object the partition-relative |
| 3289 | * bio start sector is to offset relative to the enclosing |
| 3290 | * device. |
| 3291 | */ |
| 3292 | sector_offset = get_start_sect(bmd->bi_bdev) + bmd->bi_sector; |
| 3293 | sectors_per_obj = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); |
| 3294 | obj_sector_offset = sector_offset & (sectors_per_obj - 1); |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3295 | |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 3296 | /* |
| 3297 | * Compute the number of bytes from that offset to the end |
| 3298 | * of the object. Account for what's already used by the bio. |
| 3299 | */ |
| 3300 | ret = (int) (sectors_per_obj - obj_sector_offset) << SECTOR_SHIFT; |
| 3301 | if (ret > bmd->bi_size) |
| 3302 | ret -= bmd->bi_size; |
| 3303 | else |
| 3304 | ret = 0; |
| 3305 | |
| 3306 | /* |
| 3307 | * Don't send back more than was asked for. And if the bio |
| 3308 | * was empty, let the whole thing through because: "Note |
| 3309 | * that a block device *must* allow a single page to be |
| 3310 | * added to an empty bio." |
| 3311 | */ |
| 3312 | rbd_assert(bvec->bv_len <= PAGE_SIZE); |
| 3313 | if (ret > (int) bvec->bv_len || !bmd->bi_size) |
| 3314 | ret = (int) bvec->bv_len; |
| 3315 | |
| 3316 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3317 | } |
| 3318 | |
| 3319 | static void rbd_free_disk(struct rbd_device *rbd_dev) |
| 3320 | { |
| 3321 | struct gendisk *disk = rbd_dev->disk; |
| 3322 | |
| 3323 | if (!disk) |
| 3324 | return; |
| 3325 | |
Alex Elder | a0cab92 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 3326 | rbd_dev->disk = NULL; |
| 3327 | if (disk->flags & GENHD_FL_UP) { |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3328 | del_gendisk(disk); |
Alex Elder | a0cab92 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 3329 | if (disk->queue) |
| 3330 | blk_cleanup_queue(disk->queue); |
| 3331 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3332 | put_disk(disk); |
| 3333 | } |
| 3334 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3335 | static int rbd_obj_read_sync(struct rbd_device *rbd_dev, |
| 3336 | const char *object_name, |
Alex Elder | 7097f8d | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3337 | u64 offset, u64 length, void *buf) |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3338 | |
| 3339 | { |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3340 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3341 | struct rbd_obj_request *obj_request; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3342 | struct page **pages = NULL; |
| 3343 | u32 page_count; |
Alex Elder | 1ceae7e | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 3344 | size_t size; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3345 | int ret; |
| 3346 | |
| 3347 | page_count = (u32) calc_pages_for(offset, length); |
| 3348 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 3349 | if (IS_ERR(pages)) |
| 3350 | ret = PTR_ERR(pages); |
| 3351 | |
| 3352 | ret = -ENOMEM; |
| 3353 | obj_request = rbd_obj_request_create(object_name, offset, length, |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3354 | OBJ_REQUEST_PAGES); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3355 | if (!obj_request) |
| 3356 | goto out; |
| 3357 | |
| 3358 | obj_request->pages = pages; |
| 3359 | obj_request->page_count = page_count; |
| 3360 | |
Ilya Dryomov | deb236b | 2014-02-25 16:22:27 +0200 | [diff] [blame] | 3361 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, 1, |
| 3362 | obj_request); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3363 | if (!obj_request->osd_req) |
| 3364 | goto out; |
| 3365 | |
Alex Elder | c99d2d4 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 3366 | osd_req_op_extent_init(obj_request->osd_req, 0, CEPH_OSD_OP_READ, |
| 3367 | offset, length, 0, 0); |
Alex Elder | 406e2c9 | 2013-04-15 14:50:36 -0500 | [diff] [blame] | 3368 | osd_req_op_extent_osd_data_pages(obj_request->osd_req, 0, |
Alex Elder | a4ce40a | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3369 | obj_request->pages, |
Alex Elder | 44cd188 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3370 | obj_request->length, |
| 3371 | obj_request->offset & ~PAGE_MASK, |
| 3372 | false, false); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 3373 | rbd_osd_req_format_read(obj_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 3374 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3375 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 3376 | if (ret) |
| 3377 | goto out; |
| 3378 | ret = rbd_obj_request_wait(obj_request); |
| 3379 | if (ret) |
| 3380 | goto out; |
| 3381 | |
| 3382 | ret = obj_request->result; |
| 3383 | if (ret < 0) |
| 3384 | goto out; |
Alex Elder | 1ceae7e | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 3385 | |
| 3386 | rbd_assert(obj_request->xferred <= (u64) SIZE_MAX); |
| 3387 | size = (size_t) obj_request->xferred; |
Alex Elder | 903bb32 | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 3388 | ceph_copy_from_page_vector(pages, buf, 0, size); |
Alex Elder | 7097f8d | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3389 | rbd_assert(size <= (size_t)INT_MAX); |
| 3390 | ret = (int)size; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3391 | out: |
| 3392 | if (obj_request) |
| 3393 | rbd_obj_request_put(obj_request); |
| 3394 | else |
| 3395 | ceph_release_page_vector(pages, page_count); |
| 3396 | |
| 3397 | return ret; |
| 3398 | } |
| 3399 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3400 | /* |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3401 | * Read the complete header for the given rbd device. On successful |
| 3402 | * return, the rbd_dev->header field will contain up-to-date |
| 3403 | * information about the image. |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3404 | */ |
Alex Elder | 99a41eb | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 3405 | static int rbd_dev_v1_header_info(struct rbd_device *rbd_dev) |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3406 | { |
| 3407 | struct rbd_image_header_ondisk *ondisk = NULL; |
| 3408 | u32 snap_count = 0; |
| 3409 | u64 names_size = 0; |
| 3410 | u32 want_count; |
| 3411 | int ret; |
| 3412 | |
| 3413 | /* |
| 3414 | * The complete header will include an array of its 64-bit |
| 3415 | * snapshot ids, followed by the names of those snapshots as |
| 3416 | * a contiguous block of NUL-terminated strings. Note that |
| 3417 | * the number of snapshots could change by the time we read |
| 3418 | * it in, in which case we re-read it. |
| 3419 | */ |
| 3420 | do { |
| 3421 | size_t size; |
| 3422 | |
| 3423 | kfree(ondisk); |
| 3424 | |
| 3425 | size = sizeof (*ondisk); |
| 3426 | size += snap_count * sizeof (struct rbd_image_snap_ondisk); |
| 3427 | size += names_size; |
| 3428 | ondisk = kmalloc(size, GFP_KERNEL); |
| 3429 | if (!ondisk) |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3430 | return -ENOMEM; |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3431 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3432 | ret = rbd_obj_read_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 7097f8d | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3433 | 0, size, ondisk); |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3434 | if (ret < 0) |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3435 | goto out; |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3436 | if ((size_t)ret < size) { |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3437 | ret = -ENXIO; |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3438 | rbd_warn(rbd_dev, "short header read (want %zd got %d)", |
| 3439 | size, ret); |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3440 | goto out; |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3441 | } |
| 3442 | if (!rbd_dev_ondisk_valid(ondisk)) { |
| 3443 | ret = -ENXIO; |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3444 | rbd_warn(rbd_dev, "invalid header"); |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3445 | goto out; |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3446 | } |
| 3447 | |
| 3448 | names_size = le64_to_cpu(ondisk->snap_names_len); |
| 3449 | want_count = snap_count; |
| 3450 | snap_count = le32_to_cpu(ondisk->snap_count); |
| 3451 | } while (snap_count != want_count); |
| 3452 | |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3453 | ret = rbd_header_from_disk(rbd_dev, ondisk); |
| 3454 | out: |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3455 | kfree(ondisk); |
| 3456 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3457 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3458 | } |
| 3459 | |
Alex Elder | 15228ed | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 3460 | /* |
| 3461 | * Clear the rbd device's EXISTS flag if the snapshot it's mapped to |
| 3462 | * has disappeared from the (just updated) snapshot context. |
| 3463 | */ |
| 3464 | static void rbd_exists_validate(struct rbd_device *rbd_dev) |
| 3465 | { |
| 3466 | u64 snap_id; |
| 3467 | |
| 3468 | if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) |
| 3469 | return; |
| 3470 | |
| 3471 | snap_id = rbd_dev->spec->snap_id; |
| 3472 | if (snap_id == CEPH_NOSNAP) |
| 3473 | return; |
| 3474 | |
| 3475 | if (rbd_dev_snap_index(rbd_dev, snap_id) == BAD_SNAP_INDEX) |
| 3476 | clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
| 3477 | } |
| 3478 | |
Josh Durgin | 9875201 | 2013-08-29 17:26:31 -0700 | [diff] [blame] | 3479 | static void rbd_dev_update_size(struct rbd_device *rbd_dev) |
| 3480 | { |
| 3481 | sector_t size; |
| 3482 | bool removing; |
| 3483 | |
| 3484 | /* |
| 3485 | * Don't hold the lock while doing disk operations, |
| 3486 | * or lock ordering will conflict with the bdev mutex via: |
| 3487 | * rbd_add() -> blkdev_get() -> rbd_open() |
| 3488 | */ |
| 3489 | spin_lock_irq(&rbd_dev->lock); |
| 3490 | removing = test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags); |
| 3491 | spin_unlock_irq(&rbd_dev->lock); |
| 3492 | /* |
| 3493 | * If the device is being removed, rbd_dev->disk has |
| 3494 | * been destroyed, so don't try to update its size |
| 3495 | */ |
| 3496 | if (!removing) { |
| 3497 | size = (sector_t)rbd_dev->mapping.size / SECTOR_SIZE; |
| 3498 | dout("setting size to %llu sectors", (unsigned long long)size); |
| 3499 | set_capacity(rbd_dev->disk, size); |
| 3500 | revalidate_disk(rbd_dev->disk); |
| 3501 | } |
| 3502 | } |
| 3503 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3504 | static int rbd_dev_refresh(struct rbd_device *rbd_dev) |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3505 | { |
Alex Elder | e627db0 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3506 | u64 mapping_size; |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3507 | int ret; |
| 3508 | |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 3509 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 3510 | down_write(&rbd_dev->header_rwsem); |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 3511 | mapping_size = rbd_dev->mapping.size; |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 3512 | if (rbd_dev->image_format == 1) |
Alex Elder | 99a41eb | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 3513 | ret = rbd_dev_v1_header_info(rbd_dev); |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 3514 | else |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 3515 | ret = rbd_dev_v2_header_info(rbd_dev); |
Alex Elder | 15228ed | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 3516 | |
| 3517 | /* If it's a mapped snapshot, validate its EXISTS flag */ |
| 3518 | |
| 3519 | rbd_exists_validate(rbd_dev); |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 3520 | up_write(&rbd_dev->header_rwsem); |
| 3521 | |
Alex Elder | 00a653e | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3522 | if (mapping_size != rbd_dev->mapping.size) { |
Josh Durgin | 9875201 | 2013-08-29 17:26:31 -0700 | [diff] [blame] | 3523 | rbd_dev_update_size(rbd_dev); |
Alex Elder | 00a653e | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3524 | } |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3525 | |
| 3526 | return ret; |
| 3527 | } |
| 3528 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3529 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
| 3530 | { |
| 3531 | struct gendisk *disk; |
| 3532 | struct request_queue *q; |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3533 | u64 segment_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3534 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3535 | /* create gendisk info */ |
Ilya Dryomov | 7e513d4 | 2013-12-16 19:26:32 +0200 | [diff] [blame] | 3536 | disk = alloc_disk(single_major ? |
| 3537 | (1 << RBD_SINGLE_MAJOR_PART_SHIFT) : |
| 3538 | RBD_MINORS_PER_MAJOR); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3539 | if (!disk) |
Alex Elder | 1fcdb8a | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3540 | return -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3541 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3542 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3543 | rbd_dev->dev_id); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3544 | disk->major = rbd_dev->major; |
Ilya Dryomov | dd82fff | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 3545 | disk->first_minor = rbd_dev->minor; |
Ilya Dryomov | 7e513d4 | 2013-12-16 19:26:32 +0200 | [diff] [blame] | 3546 | if (single_major) |
| 3547 | disk->flags |= GENHD_FL_EXT_DEVT; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3548 | disk->fops = &rbd_bd_ops; |
| 3549 | disk->private_data = rbd_dev; |
| 3550 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3551 | q = blk_init_queue(rbd_request_fn, &rbd_dev->lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3552 | if (!q) |
| 3553 | goto out_disk; |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 3554 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3555 | /* We use the default size, but let's be explicit about it. */ |
| 3556 | blk_queue_physical_block_size(q, SECTOR_SIZE); |
| 3557 | |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 3558 | /* set io sizes to object size */ |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3559 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
| 3560 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); |
| 3561 | blk_queue_max_segment_size(q, segment_size); |
| 3562 | blk_queue_io_min(q, segment_size); |
| 3563 | blk_queue_io_opt(q, segment_size); |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 3564 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3565 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
| 3566 | disk->queue = q; |
| 3567 | |
| 3568 | q->queuedata = rbd_dev; |
| 3569 | |
| 3570 | rbd_dev->disk = disk; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3571 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3572 | return 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3573 | out_disk: |
| 3574 | put_disk(disk); |
Alex Elder | 1fcdb8a | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3575 | |
| 3576 | return -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3577 | } |
| 3578 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3579 | /* |
| 3580 | sysfs |
| 3581 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3582 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3583 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
| 3584 | { |
| 3585 | return container_of(dev, struct rbd_device, dev); |
| 3586 | } |
| 3587 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3588 | static ssize_t rbd_size_show(struct device *dev, |
| 3589 | struct device_attribute *attr, char *buf) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3590 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3591 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3592 | |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3593 | return sprintf(buf, "%llu\n", |
| 3594 | (unsigned long long)rbd_dev->mapping.size); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3595 | } |
| 3596 | |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3597 | /* |
| 3598 | * Note this shows the features for whatever's mapped, which is not |
| 3599 | * necessarily the base image. |
| 3600 | */ |
| 3601 | static ssize_t rbd_features_show(struct device *dev, |
| 3602 | struct device_attribute *attr, char *buf) |
| 3603 | { |
| 3604 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3605 | |
| 3606 | return sprintf(buf, "0x%016llx\n", |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3607 | (unsigned long long)rbd_dev->mapping.features); |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3608 | } |
| 3609 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3610 | static ssize_t rbd_major_show(struct device *dev, |
| 3611 | struct device_attribute *attr, char *buf) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3612 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3613 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3614 | |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3615 | if (rbd_dev->major) |
| 3616 | return sprintf(buf, "%d\n", rbd_dev->major); |
| 3617 | |
| 3618 | return sprintf(buf, "(none)\n"); |
Ilya Dryomov | dd82fff | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 3619 | } |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3620 | |
Ilya Dryomov | dd82fff | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 3621 | static ssize_t rbd_minor_show(struct device *dev, |
| 3622 | struct device_attribute *attr, char *buf) |
| 3623 | { |
| 3624 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3625 | |
| 3626 | return sprintf(buf, "%d\n", rbd_dev->minor); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3627 | } |
| 3628 | |
| 3629 | static ssize_t rbd_client_id_show(struct device *dev, |
| 3630 | struct device_attribute *attr, char *buf) |
| 3631 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3632 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3633 | |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 3634 | return sprintf(buf, "client%lld\n", |
| 3635 | ceph_client_id(rbd_dev->rbd_client->client)); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3636 | } |
| 3637 | |
| 3638 | static ssize_t rbd_pool_show(struct device *dev, |
| 3639 | struct device_attribute *attr, char *buf) |
| 3640 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3641 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3642 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3643 | return sprintf(buf, "%s\n", rbd_dev->spec->pool_name); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3644 | } |
| 3645 | |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3646 | static ssize_t rbd_pool_id_show(struct device *dev, |
| 3647 | struct device_attribute *attr, char *buf) |
| 3648 | { |
| 3649 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3650 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3651 | return sprintf(buf, "%llu\n", |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3652 | (unsigned long long) rbd_dev->spec->pool_id); |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3653 | } |
| 3654 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3655 | static ssize_t rbd_name_show(struct device *dev, |
| 3656 | struct device_attribute *attr, char *buf) |
| 3657 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3658 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3659 | |
Alex Elder | a92ffdf | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3660 | if (rbd_dev->spec->image_name) |
| 3661 | return sprintf(buf, "%s\n", rbd_dev->spec->image_name); |
| 3662 | |
| 3663 | return sprintf(buf, "(unknown)\n"); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3664 | } |
| 3665 | |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3666 | static ssize_t rbd_image_id_show(struct device *dev, |
| 3667 | struct device_attribute *attr, char *buf) |
| 3668 | { |
| 3669 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3670 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3671 | return sprintf(buf, "%s\n", rbd_dev->spec->image_id); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3672 | } |
| 3673 | |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3674 | /* |
| 3675 | * Shows the name of the currently-mapped snapshot (or |
| 3676 | * RBD_SNAP_HEAD_NAME for the base image). |
| 3677 | */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3678 | static ssize_t rbd_snap_show(struct device *dev, |
| 3679 | struct device_attribute *attr, |
| 3680 | char *buf) |
| 3681 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3682 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3683 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3684 | return sprintf(buf, "%s\n", rbd_dev->spec->snap_name); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3685 | } |
| 3686 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3687 | /* |
| 3688 | * For an rbd v2 image, shows the pool id, image id, and snapshot id |
| 3689 | * for the parent image. If there is no parent, simply shows |
| 3690 | * "(no parent image)". |
| 3691 | */ |
| 3692 | static ssize_t rbd_parent_show(struct device *dev, |
| 3693 | struct device_attribute *attr, |
| 3694 | char *buf) |
| 3695 | { |
| 3696 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3697 | struct rbd_spec *spec = rbd_dev->parent_spec; |
| 3698 | int count; |
| 3699 | char *bufp = buf; |
| 3700 | |
| 3701 | if (!spec) |
| 3702 | return sprintf(buf, "(no parent image)\n"); |
| 3703 | |
| 3704 | count = sprintf(bufp, "pool_id %llu\npool_name %s\n", |
| 3705 | (unsigned long long) spec->pool_id, spec->pool_name); |
| 3706 | if (count < 0) |
| 3707 | return count; |
| 3708 | bufp += count; |
| 3709 | |
| 3710 | count = sprintf(bufp, "image_id %s\nimage_name %s\n", spec->image_id, |
| 3711 | spec->image_name ? spec->image_name : "(unknown)"); |
| 3712 | if (count < 0) |
| 3713 | return count; |
| 3714 | bufp += count; |
| 3715 | |
| 3716 | count = sprintf(bufp, "snap_id %llu\nsnap_name %s\n", |
| 3717 | (unsigned long long) spec->snap_id, spec->snap_name); |
| 3718 | if (count < 0) |
| 3719 | return count; |
| 3720 | bufp += count; |
| 3721 | |
| 3722 | count = sprintf(bufp, "overlap %llu\n", rbd_dev->parent_overlap); |
| 3723 | if (count < 0) |
| 3724 | return count; |
| 3725 | bufp += count; |
| 3726 | |
| 3727 | return (ssize_t) (bufp - buf); |
| 3728 | } |
| 3729 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3730 | static ssize_t rbd_image_refresh(struct device *dev, |
| 3731 | struct device_attribute *attr, |
| 3732 | const char *buf, |
| 3733 | size_t size) |
| 3734 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3735 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Alex Elder | b813623 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3736 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3737 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3738 | ret = rbd_dev_refresh(rbd_dev); |
Alex Elder | e627db0 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3739 | if (ret) |
| 3740 | rbd_warn(rbd_dev, ": manual header refresh error (%d)\n", ret); |
Alex Elder | b813623 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3741 | |
| 3742 | return ret < 0 ? ret : size; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3743 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3744 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3745 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3746 | static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3747 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); |
Ilya Dryomov | dd82fff | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 3748 | static DEVICE_ATTR(minor, S_IRUGO, rbd_minor_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3749 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); |
| 3750 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3751 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3752 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3753 | static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3754 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); |
| 3755 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3756 | static DEVICE_ATTR(parent, S_IRUGO, rbd_parent_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3757 | |
| 3758 | static struct attribute *rbd_attrs[] = { |
| 3759 | &dev_attr_size.attr, |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3760 | &dev_attr_features.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3761 | &dev_attr_major.attr, |
Ilya Dryomov | dd82fff | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 3762 | &dev_attr_minor.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3763 | &dev_attr_client_id.attr, |
| 3764 | &dev_attr_pool.attr, |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3765 | &dev_attr_pool_id.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3766 | &dev_attr_name.attr, |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3767 | &dev_attr_image_id.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3768 | &dev_attr_current_snap.attr, |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3769 | &dev_attr_parent.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3770 | &dev_attr_refresh.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3771 | NULL |
| 3772 | }; |
| 3773 | |
| 3774 | static struct attribute_group rbd_attr_group = { |
| 3775 | .attrs = rbd_attrs, |
| 3776 | }; |
| 3777 | |
| 3778 | static const struct attribute_group *rbd_attr_groups[] = { |
| 3779 | &rbd_attr_group, |
| 3780 | NULL |
| 3781 | }; |
| 3782 | |
| 3783 | static void rbd_sysfs_dev_release(struct device *dev) |
| 3784 | { |
| 3785 | } |
| 3786 | |
| 3787 | static struct device_type rbd_device_type = { |
| 3788 | .name = "rbd", |
| 3789 | .groups = rbd_attr_groups, |
| 3790 | .release = rbd_sysfs_dev_release, |
| 3791 | }; |
| 3792 | |
Alex Elder | 8b8fb99 | 2012-10-26 17:25:24 -0500 | [diff] [blame] | 3793 | static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec) |
| 3794 | { |
| 3795 | kref_get(&spec->kref); |
| 3796 | |
| 3797 | return spec; |
| 3798 | } |
| 3799 | |
| 3800 | static void rbd_spec_free(struct kref *kref); |
| 3801 | static void rbd_spec_put(struct rbd_spec *spec) |
| 3802 | { |
| 3803 | if (spec) |
| 3804 | kref_put(&spec->kref, rbd_spec_free); |
| 3805 | } |
| 3806 | |
| 3807 | static struct rbd_spec *rbd_spec_alloc(void) |
| 3808 | { |
| 3809 | struct rbd_spec *spec; |
| 3810 | |
| 3811 | spec = kzalloc(sizeof (*spec), GFP_KERNEL); |
| 3812 | if (!spec) |
| 3813 | return NULL; |
| 3814 | kref_init(&spec->kref); |
| 3815 | |
Alex Elder | 8b8fb99 | 2012-10-26 17:25:24 -0500 | [diff] [blame] | 3816 | return spec; |
| 3817 | } |
| 3818 | |
| 3819 | static void rbd_spec_free(struct kref *kref) |
| 3820 | { |
| 3821 | struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref); |
| 3822 | |
| 3823 | kfree(spec->pool_name); |
| 3824 | kfree(spec->image_id); |
| 3825 | kfree(spec->image_name); |
| 3826 | kfree(spec->snap_name); |
| 3827 | kfree(spec); |
| 3828 | } |
| 3829 | |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 3830 | static struct rbd_device *rbd_dev_create(struct rbd_client *rbdc, |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3831 | struct rbd_spec *spec) |
| 3832 | { |
| 3833 | struct rbd_device *rbd_dev; |
| 3834 | |
| 3835 | rbd_dev = kzalloc(sizeof (*rbd_dev), GFP_KERNEL); |
| 3836 | if (!rbd_dev) |
| 3837 | return NULL; |
| 3838 | |
| 3839 | spin_lock_init(&rbd_dev->lock); |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 3840 | rbd_dev->flags = 0; |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 3841 | atomic_set(&rbd_dev->parent_ref, 0); |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3842 | INIT_LIST_HEAD(&rbd_dev->node); |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3843 | init_rwsem(&rbd_dev->header_rwsem); |
| 3844 | |
| 3845 | rbd_dev->spec = spec; |
| 3846 | rbd_dev->rbd_client = rbdc; |
| 3847 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 3848 | /* Initialize the layout used for all rbd requests */ |
| 3849 | |
| 3850 | rbd_dev->layout.fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); |
| 3851 | rbd_dev->layout.fl_stripe_count = cpu_to_le32(1); |
| 3852 | rbd_dev->layout.fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); |
| 3853 | rbd_dev->layout.fl_pg_pool = cpu_to_le32((u32) spec->pool_id); |
| 3854 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3855 | return rbd_dev; |
| 3856 | } |
| 3857 | |
| 3858 | static void rbd_dev_destroy(struct rbd_device *rbd_dev) |
| 3859 | { |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3860 | rbd_put_client(rbd_dev->rbd_client); |
| 3861 | rbd_spec_put(rbd_dev->spec); |
| 3862 | kfree(rbd_dev); |
| 3863 | } |
| 3864 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3865 | /* |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3866 | * Get the size and object order for an image snapshot, or if |
| 3867 | * snap_id is CEPH_NOSNAP, gets this information for the base |
| 3868 | * image. |
| 3869 | */ |
| 3870 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
| 3871 | u8 *order, u64 *snap_size) |
| 3872 | { |
| 3873 | __le64 snapid = cpu_to_le64(snap_id); |
| 3874 | int ret; |
| 3875 | struct { |
| 3876 | u8 order; |
| 3877 | __le64 size; |
| 3878 | } __attribute__ ((packed)) size_buf = { 0 }; |
| 3879 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3880 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3881 | "rbd", "get_size", |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3882 | &snapid, sizeof (snapid), |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3883 | &size_buf, sizeof (size_buf)); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3884 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3885 | if (ret < 0) |
| 3886 | return ret; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3887 | if (ret < sizeof (size_buf)) |
| 3888 | return -ERANGE; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3889 | |
Josh Durgin | c354557 | 2013-08-28 17:08:10 -0700 | [diff] [blame] | 3890 | if (order) { |
Alex Elder | c86f86e | 2013-04-25 15:09:41 -0500 | [diff] [blame] | 3891 | *order = size_buf.order; |
Josh Durgin | c354557 | 2013-08-28 17:08:10 -0700 | [diff] [blame] | 3892 | dout(" order %u", (unsigned int)*order); |
| 3893 | } |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3894 | *snap_size = le64_to_cpu(size_buf.size); |
| 3895 | |
Josh Durgin | c354557 | 2013-08-28 17:08:10 -0700 | [diff] [blame] | 3896 | dout(" snap_id 0x%016llx snap_size = %llu\n", |
| 3897 | (unsigned long long)snap_id, |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3898 | (unsigned long long)*snap_size); |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3899 | |
| 3900 | return 0; |
| 3901 | } |
| 3902 | |
| 3903 | static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) |
| 3904 | { |
| 3905 | return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, |
| 3906 | &rbd_dev->header.obj_order, |
| 3907 | &rbd_dev->header.image_size); |
| 3908 | } |
| 3909 | |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3910 | static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) |
| 3911 | { |
| 3912 | void *reply_buf; |
| 3913 | int ret; |
| 3914 | void *p; |
| 3915 | |
| 3916 | reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL); |
| 3917 | if (!reply_buf) |
| 3918 | return -ENOMEM; |
| 3919 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3920 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3921 | "rbd", "get_object_prefix", NULL, 0, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3922 | reply_buf, RBD_OBJ_PREFIX_LEN_MAX); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3923 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3924 | if (ret < 0) |
| 3925 | goto out; |
| 3926 | |
| 3927 | p = reply_buf; |
| 3928 | rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3929 | p + ret, NULL, GFP_NOIO); |
| 3930 | ret = 0; |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3931 | |
| 3932 | if (IS_ERR(rbd_dev->header.object_prefix)) { |
| 3933 | ret = PTR_ERR(rbd_dev->header.object_prefix); |
| 3934 | rbd_dev->header.object_prefix = NULL; |
| 3935 | } else { |
| 3936 | dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); |
| 3937 | } |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3938 | out: |
| 3939 | kfree(reply_buf); |
| 3940 | |
| 3941 | return ret; |
| 3942 | } |
| 3943 | |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3944 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
| 3945 | u64 *snap_features) |
| 3946 | { |
| 3947 | __le64 snapid = cpu_to_le64(snap_id); |
| 3948 | struct { |
| 3949 | __le64 features; |
| 3950 | __le64 incompat; |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3951 | } __attribute__ ((packed)) features_buf = { 0 }; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 3952 | u64 incompat; |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3953 | int ret; |
| 3954 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3955 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3956 | "rbd", "get_features", |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3957 | &snapid, sizeof (snapid), |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3958 | &features_buf, sizeof (features_buf)); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3959 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3960 | if (ret < 0) |
| 3961 | return ret; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3962 | if (ret < sizeof (features_buf)) |
| 3963 | return -ERANGE; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 3964 | |
| 3965 | incompat = le64_to_cpu(features_buf.incompat); |
Alex Elder | 5cbf6f12 | 2013-04-11 09:29:48 -0500 | [diff] [blame] | 3966 | if (incompat & ~RBD_FEATURES_SUPPORTED) |
Alex Elder | b8f5c6e | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3967 | return -ENXIO; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 3968 | |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3969 | *snap_features = le64_to_cpu(features_buf.features); |
| 3970 | |
| 3971 | dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3972 | (unsigned long long)snap_id, |
| 3973 | (unsigned long long)*snap_features, |
| 3974 | (unsigned long long)le64_to_cpu(features_buf.incompat)); |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3975 | |
| 3976 | return 0; |
| 3977 | } |
| 3978 | |
| 3979 | static int rbd_dev_v2_features(struct rbd_device *rbd_dev) |
| 3980 | { |
| 3981 | return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, |
| 3982 | &rbd_dev->header.features); |
| 3983 | } |
| 3984 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3985 | static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) |
| 3986 | { |
| 3987 | struct rbd_spec *parent_spec; |
| 3988 | size_t size; |
| 3989 | void *reply_buf = NULL; |
| 3990 | __le64 snapid; |
| 3991 | void *p; |
| 3992 | void *end; |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3993 | u64 pool_id; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3994 | char *image_id; |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 3995 | u64 snap_id; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3996 | u64 overlap; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3997 | int ret; |
| 3998 | |
| 3999 | parent_spec = rbd_spec_alloc(); |
| 4000 | if (!parent_spec) |
| 4001 | return -ENOMEM; |
| 4002 | |
| 4003 | size = sizeof (__le64) + /* pool_id */ |
| 4004 | sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX + /* image_id */ |
| 4005 | sizeof (__le64) + /* snap_id */ |
| 4006 | sizeof (__le64); /* overlap */ |
| 4007 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 4008 | if (!reply_buf) { |
| 4009 | ret = -ENOMEM; |
| 4010 | goto out_err; |
| 4011 | } |
| 4012 | |
| 4013 | snapid = cpu_to_le64(CEPH_NOSNAP); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4014 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4015 | "rbd", "get_parent", |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4016 | &snapid, sizeof (snapid), |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4017 | reply_buf, size); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4018 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4019 | if (ret < 0) |
| 4020 | goto out_err; |
| 4021 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4022 | p = reply_buf; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4023 | end = reply_buf + ret; |
| 4024 | ret = -ERANGE; |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4025 | ceph_decode_64_safe(&p, end, pool_id, out_err); |
Alex Elder | 392a9da | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4026 | if (pool_id == CEPH_NOPOOL) { |
| 4027 | /* |
| 4028 | * Either the parent never existed, or we have |
| 4029 | * record of it but the image got flattened so it no |
| 4030 | * longer has a parent. When the parent of a |
| 4031 | * layered image disappears we immediately set the |
| 4032 | * overlap to 0. The effect of this is that all new |
| 4033 | * requests will be treated as if the image had no |
| 4034 | * parent. |
| 4035 | */ |
| 4036 | if (rbd_dev->parent_overlap) { |
| 4037 | rbd_dev->parent_overlap = 0; |
| 4038 | smp_mb(); |
| 4039 | rbd_dev_parent_put(rbd_dev); |
| 4040 | pr_info("%s: clone image has been flattened\n", |
| 4041 | rbd_dev->disk->disk_name); |
| 4042 | } |
| 4043 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4044 | goto out; /* No parent? No problem. */ |
Alex Elder | 392a9da | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4045 | } |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4046 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 4047 | /* The ceph file layout needs to fit pool id in 32 bits */ |
| 4048 | |
| 4049 | ret = -EIO; |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4050 | if (pool_id > (u64)U32_MAX) { |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4051 | rbd_warn(NULL, "parent pool id too large (%llu > %u)\n", |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4052 | (unsigned long long)pool_id, U32_MAX); |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4053 | goto out_err; |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4054 | } |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 4055 | |
Alex Elder | 979ed48 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 4056 | image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4057 | if (IS_ERR(image_id)) { |
| 4058 | ret = PTR_ERR(image_id); |
| 4059 | goto out_err; |
| 4060 | } |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 4061 | ceph_decode_64_safe(&p, end, snap_id, out_err); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4062 | ceph_decode_64_safe(&p, end, overlap, out_err); |
| 4063 | |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 4064 | /* |
| 4065 | * The parent won't change (except when the clone is |
| 4066 | * flattened, already handled that). So we only need to |
| 4067 | * record the parent spec we have not already done so. |
| 4068 | */ |
| 4069 | if (!rbd_dev->parent_spec) { |
| 4070 | parent_spec->pool_id = pool_id; |
| 4071 | parent_spec->image_id = image_id; |
| 4072 | parent_spec->snap_id = snap_id; |
Alex Elder | 70cf49c | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4073 | rbd_dev->parent_spec = parent_spec; |
| 4074 | parent_spec = NULL; /* rbd_dev now owns this */ |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 4075 | } |
| 4076 | |
| 4077 | /* |
| 4078 | * We always update the parent overlap. If it's zero we |
| 4079 | * treat it specially. |
| 4080 | */ |
| 4081 | rbd_dev->parent_overlap = overlap; |
| 4082 | smp_mb(); |
| 4083 | if (!overlap) { |
| 4084 | |
| 4085 | /* A null parent_spec indicates it's the initial probe */ |
| 4086 | |
| 4087 | if (parent_spec) { |
| 4088 | /* |
| 4089 | * The overlap has become zero, so the clone |
| 4090 | * must have been resized down to 0 at some |
| 4091 | * point. Treat this the same as a flatten. |
| 4092 | */ |
| 4093 | rbd_dev_parent_put(rbd_dev); |
| 4094 | pr_info("%s: clone image now standalone\n", |
| 4095 | rbd_dev->disk->disk_name); |
| 4096 | } else { |
| 4097 | /* |
| 4098 | * For the initial probe, if we find the |
| 4099 | * overlap is zero we just pretend there was |
| 4100 | * no parent image. |
| 4101 | */ |
| 4102 | rbd_warn(rbd_dev, "ignoring parent of " |
| 4103 | "clone with overlap 0\n"); |
| 4104 | } |
Alex Elder | 70cf49c | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4105 | } |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4106 | out: |
| 4107 | ret = 0; |
| 4108 | out_err: |
| 4109 | kfree(reply_buf); |
| 4110 | rbd_spec_put(parent_spec); |
| 4111 | |
| 4112 | return ret; |
| 4113 | } |
| 4114 | |
Alex Elder | cc070d5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4115 | static int rbd_dev_v2_striping_info(struct rbd_device *rbd_dev) |
| 4116 | { |
| 4117 | struct { |
| 4118 | __le64 stripe_unit; |
| 4119 | __le64 stripe_count; |
| 4120 | } __attribute__ ((packed)) striping_info_buf = { 0 }; |
| 4121 | size_t size = sizeof (striping_info_buf); |
| 4122 | void *p; |
| 4123 | u64 obj_size; |
| 4124 | u64 stripe_unit; |
| 4125 | u64 stripe_count; |
| 4126 | int ret; |
| 4127 | |
| 4128 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
| 4129 | "rbd", "get_stripe_unit_count", NULL, 0, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4130 | (char *)&striping_info_buf, size); |
Alex Elder | cc070d5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4131 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
| 4132 | if (ret < 0) |
| 4133 | return ret; |
| 4134 | if (ret < size) |
| 4135 | return -ERANGE; |
| 4136 | |
| 4137 | /* |
| 4138 | * We don't actually support the "fancy striping" feature |
| 4139 | * (STRIPINGV2) yet, but if the striping sizes are the |
| 4140 | * defaults the behavior is the same as before. So find |
| 4141 | * out, and only fail if the image has non-default values. |
| 4142 | */ |
| 4143 | ret = -EINVAL; |
| 4144 | obj_size = (u64)1 << rbd_dev->header.obj_order; |
| 4145 | p = &striping_info_buf; |
| 4146 | stripe_unit = ceph_decode_64(&p); |
| 4147 | if (stripe_unit != obj_size) { |
| 4148 | rbd_warn(rbd_dev, "unsupported stripe unit " |
| 4149 | "(got %llu want %llu)", |
| 4150 | stripe_unit, obj_size); |
| 4151 | return -EINVAL; |
| 4152 | } |
| 4153 | stripe_count = ceph_decode_64(&p); |
| 4154 | if (stripe_count != 1) { |
| 4155 | rbd_warn(rbd_dev, "unsupported stripe count " |
| 4156 | "(got %llu want 1)", stripe_count); |
| 4157 | return -EINVAL; |
| 4158 | } |
Alex Elder | 500d0c0 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4159 | rbd_dev->header.stripe_unit = stripe_unit; |
| 4160 | rbd_dev->header.stripe_count = stripe_count; |
Alex Elder | cc070d5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4161 | |
| 4162 | return 0; |
| 4163 | } |
| 4164 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4165 | static char *rbd_dev_image_name(struct rbd_device *rbd_dev) |
| 4166 | { |
| 4167 | size_t image_id_size; |
| 4168 | char *image_id; |
| 4169 | void *p; |
| 4170 | void *end; |
| 4171 | size_t size; |
| 4172 | void *reply_buf = NULL; |
| 4173 | size_t len = 0; |
| 4174 | char *image_name = NULL; |
| 4175 | int ret; |
| 4176 | |
| 4177 | rbd_assert(!rbd_dev->spec->image_name); |
| 4178 | |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 4179 | len = strlen(rbd_dev->spec->image_id); |
| 4180 | image_id_size = sizeof (__le32) + len; |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4181 | image_id = kmalloc(image_id_size, GFP_KERNEL); |
| 4182 | if (!image_id) |
| 4183 | return NULL; |
| 4184 | |
| 4185 | p = image_id; |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4186 | end = image_id + image_id_size; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4187 | ceph_encode_string(&p, end, rbd_dev->spec->image_id, (u32)len); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4188 | |
| 4189 | size = sizeof (__le32) + RBD_IMAGE_NAME_LEN_MAX; |
| 4190 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 4191 | if (!reply_buf) |
| 4192 | goto out; |
| 4193 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4194 | ret = rbd_obj_method_sync(rbd_dev, RBD_DIRECTORY, |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4195 | "rbd", "dir_get_name", |
| 4196 | image_id, image_id_size, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4197 | reply_buf, size); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4198 | if (ret < 0) |
| 4199 | goto out; |
| 4200 | p = reply_buf; |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4201 | end = reply_buf + ret; |
| 4202 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4203 | image_name = ceph_extract_encoded_string(&p, end, &len, GFP_KERNEL); |
| 4204 | if (IS_ERR(image_name)) |
| 4205 | image_name = NULL; |
| 4206 | else |
| 4207 | dout("%s: name is %s len is %zd\n", __func__, image_name, len); |
| 4208 | out: |
| 4209 | kfree(reply_buf); |
| 4210 | kfree(image_id); |
| 4211 | |
| 4212 | return image_name; |
| 4213 | } |
| 4214 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4215 | static u64 rbd_v1_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) |
| 4216 | { |
| 4217 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
| 4218 | const char *snap_name; |
| 4219 | u32 which = 0; |
| 4220 | |
| 4221 | /* Skip over names until we find the one we are looking for */ |
| 4222 | |
| 4223 | snap_name = rbd_dev->header.snap_names; |
| 4224 | while (which < snapc->num_snaps) { |
| 4225 | if (!strcmp(name, snap_name)) |
| 4226 | return snapc->snaps[which]; |
| 4227 | snap_name += strlen(snap_name) + 1; |
| 4228 | which++; |
| 4229 | } |
| 4230 | return CEPH_NOSNAP; |
| 4231 | } |
| 4232 | |
| 4233 | static u64 rbd_v2_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) |
| 4234 | { |
| 4235 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
| 4236 | u32 which; |
| 4237 | bool found = false; |
| 4238 | u64 snap_id; |
| 4239 | |
| 4240 | for (which = 0; !found && which < snapc->num_snaps; which++) { |
| 4241 | const char *snap_name; |
| 4242 | |
| 4243 | snap_id = snapc->snaps[which]; |
| 4244 | snap_name = rbd_dev_v2_snap_name(rbd_dev, snap_id); |
Josh Durgin | efadc98 | 2013-08-29 19:16:42 -0700 | [diff] [blame] | 4245 | if (IS_ERR(snap_name)) { |
| 4246 | /* ignore no-longer existing snapshots */ |
| 4247 | if (PTR_ERR(snap_name) == -ENOENT) |
| 4248 | continue; |
| 4249 | else |
| 4250 | break; |
| 4251 | } |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4252 | found = !strcmp(name, snap_name); |
| 4253 | kfree(snap_name); |
| 4254 | } |
| 4255 | return found ? snap_id : CEPH_NOSNAP; |
| 4256 | } |
| 4257 | |
| 4258 | /* |
| 4259 | * Assumes name is never RBD_SNAP_HEAD_NAME; returns CEPH_NOSNAP if |
| 4260 | * no snapshot by that name is found, or if an error occurs. |
| 4261 | */ |
| 4262 | static u64 rbd_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) |
| 4263 | { |
| 4264 | if (rbd_dev->image_format == 1) |
| 4265 | return rbd_v1_snap_id_by_name(rbd_dev, name); |
| 4266 | |
| 4267 | return rbd_v2_snap_id_by_name(rbd_dev, name); |
| 4268 | } |
| 4269 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4270 | /* |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4271 | * When an rbd image has a parent image, it is identified by the |
| 4272 | * pool, image, and snapshot ids (not names). This function fills |
| 4273 | * in the names for those ids. (It's OK if we can't figure out the |
| 4274 | * name for an image id, but the pool and snapshot ids should always |
| 4275 | * exist and have names.) All names in an rbd spec are dynamically |
| 4276 | * allocated. |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4277 | * |
| 4278 | * When an image being mapped (not a parent) is probed, we have the |
| 4279 | * pool name and pool id, image name and image id, and the snapshot |
| 4280 | * name. The only thing we're missing is the snapshot id. |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4281 | */ |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4282 | static int rbd_dev_spec_update(struct rbd_device *rbd_dev) |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4283 | { |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4284 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
| 4285 | struct rbd_spec *spec = rbd_dev->spec; |
| 4286 | const char *pool_name; |
| 4287 | const char *image_name; |
| 4288 | const char *snap_name; |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4289 | int ret; |
| 4290 | |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4291 | /* |
| 4292 | * An image being mapped will have the pool name (etc.), but |
| 4293 | * we need to look up the snapshot id. |
| 4294 | */ |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4295 | if (spec->pool_name) { |
| 4296 | if (strcmp(spec->snap_name, RBD_SNAP_HEAD_NAME)) { |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4297 | u64 snap_id; |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4298 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4299 | snap_id = rbd_snap_id_by_name(rbd_dev, spec->snap_name); |
| 4300 | if (snap_id == CEPH_NOSNAP) |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4301 | return -ENOENT; |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4302 | spec->snap_id = snap_id; |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4303 | } else { |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4304 | spec->snap_id = CEPH_NOSNAP; |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4305 | } |
| 4306 | |
| 4307 | return 0; |
| 4308 | } |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4309 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4310 | /* Get the pool name; we have to make our own copy of this */ |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4311 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4312 | pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, spec->pool_id); |
| 4313 | if (!pool_name) { |
| 4314 | rbd_warn(rbd_dev, "no pool with id %llu", spec->pool_id); |
Alex Elder | 935dc89 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4315 | return -EIO; |
| 4316 | } |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4317 | pool_name = kstrdup(pool_name, GFP_KERNEL); |
| 4318 | if (!pool_name) |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4319 | return -ENOMEM; |
| 4320 | |
| 4321 | /* Fetch the image name; tolerate failure here */ |
| 4322 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4323 | image_name = rbd_dev_image_name(rbd_dev); |
| 4324 | if (!image_name) |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4325 | rbd_warn(rbd_dev, "unable to get image name"); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4326 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4327 | /* Look up the snapshot name, and make a copy */ |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4328 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4329 | snap_name = rbd_snap_name(rbd_dev, spec->snap_id); |
Josh Durgin | da6a6b6 | 2013-09-04 17:57:31 -0700 | [diff] [blame] | 4330 | if (IS_ERR(snap_name)) { |
| 4331 | ret = PTR_ERR(snap_name); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4332 | goto out_err; |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4333 | } |
| 4334 | |
| 4335 | spec->pool_name = pool_name; |
| 4336 | spec->image_name = image_name; |
| 4337 | spec->snap_name = snap_name; |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4338 | |
| 4339 | return 0; |
| 4340 | out_err: |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4341 | kfree(image_name); |
| 4342 | kfree(pool_name); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4343 | |
| 4344 | return ret; |
| 4345 | } |
| 4346 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4347 | static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev) |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4348 | { |
| 4349 | size_t size; |
| 4350 | int ret; |
| 4351 | void *reply_buf; |
| 4352 | void *p; |
| 4353 | void *end; |
| 4354 | u64 seq; |
| 4355 | u32 snap_count; |
| 4356 | struct ceph_snap_context *snapc; |
| 4357 | u32 i; |
| 4358 | |
| 4359 | /* |
| 4360 | * We'll need room for the seq value (maximum snapshot id), |
| 4361 | * snapshot count, and array of that many snapshot ids. |
| 4362 | * For now we have a fixed upper limit on the number we're |
| 4363 | * prepared to receive. |
| 4364 | */ |
| 4365 | size = sizeof (__le64) + sizeof (__le32) + |
| 4366 | RBD_MAX_SNAP_COUNT * sizeof (__le64); |
| 4367 | reply_buf = kzalloc(size, GFP_KERNEL); |
| 4368 | if (!reply_buf) |
| 4369 | return -ENOMEM; |
| 4370 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4371 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4372 | "rbd", "get_snapcontext", NULL, 0, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4373 | reply_buf, size); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4374 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4375 | if (ret < 0) |
| 4376 | goto out; |
| 4377 | |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4378 | p = reply_buf; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4379 | end = reply_buf + ret; |
| 4380 | ret = -ERANGE; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4381 | ceph_decode_64_safe(&p, end, seq, out); |
| 4382 | ceph_decode_32_safe(&p, end, snap_count, out); |
| 4383 | |
| 4384 | /* |
| 4385 | * Make sure the reported number of snapshot ids wouldn't go |
| 4386 | * beyond the end of our buffer. But before checking that, |
| 4387 | * make sure the computed size of the snapshot context we |
| 4388 | * allocate is representable in a size_t. |
| 4389 | */ |
| 4390 | if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) |
| 4391 | / sizeof (u64)) { |
| 4392 | ret = -EINVAL; |
| 4393 | goto out; |
| 4394 | } |
| 4395 | if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) |
| 4396 | goto out; |
Alex Elder | 468521c | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4397 | ret = 0; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4398 | |
Alex Elder | 812164f8 | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 4399 | snapc = ceph_create_snap_context(snap_count, GFP_KERNEL); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4400 | if (!snapc) { |
| 4401 | ret = -ENOMEM; |
| 4402 | goto out; |
| 4403 | } |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4404 | snapc->seq = seq; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4405 | for (i = 0; i < snap_count; i++) |
| 4406 | snapc->snaps[i] = ceph_decode_64(&p); |
| 4407 | |
Alex Elder | 49ece55 | 2013-05-06 08:37:00 -0500 | [diff] [blame] | 4408 | ceph_put_snap_context(rbd_dev->header.snapc); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4409 | rbd_dev->header.snapc = snapc; |
| 4410 | |
| 4411 | dout(" snap context seq = %llu, snap_count = %u\n", |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4412 | (unsigned long long)seq, (unsigned int)snap_count); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4413 | out: |
| 4414 | kfree(reply_buf); |
| 4415 | |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4416 | return ret; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4417 | } |
| 4418 | |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4419 | static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, |
| 4420 | u64 snap_id) |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4421 | { |
| 4422 | size_t size; |
| 4423 | void *reply_buf; |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4424 | __le64 snapid; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4425 | int ret; |
| 4426 | void *p; |
| 4427 | void *end; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4428 | char *snap_name; |
| 4429 | |
| 4430 | size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; |
| 4431 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 4432 | if (!reply_buf) |
| 4433 | return ERR_PTR(-ENOMEM); |
| 4434 | |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4435 | snapid = cpu_to_le64(snap_id); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4436 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4437 | "rbd", "get_snapshot_name", |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4438 | &snapid, sizeof (snapid), |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4439 | reply_buf, size); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4440 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4441 | if (ret < 0) { |
| 4442 | snap_name = ERR_PTR(ret); |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4443 | goto out; |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4444 | } |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4445 | |
| 4446 | p = reply_buf; |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4447 | end = reply_buf + ret; |
Alex Elder | e5c3553 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4448 | snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4449 | if (IS_ERR(snap_name)) |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4450 | goto out; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4451 | |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4452 | dout(" snap_id 0x%016llx snap_name = %s\n", |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4453 | (unsigned long long)snap_id, snap_name); |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4454 | out: |
| 4455 | kfree(reply_buf); |
| 4456 | |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4457 | return snap_name; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4458 | } |
| 4459 | |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4460 | static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev) |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4461 | { |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4462 | bool first_time = rbd_dev->header.object_prefix == NULL; |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4463 | int ret; |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4464 | |
Josh Durgin | 1617e40 | 2013-06-12 14:43:10 -0700 | [diff] [blame] | 4465 | ret = rbd_dev_v2_image_size(rbd_dev); |
| 4466 | if (ret) |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 4467 | return ret; |
Josh Durgin | 1617e40 | 2013-06-12 14:43:10 -0700 | [diff] [blame] | 4468 | |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4469 | if (first_time) { |
| 4470 | ret = rbd_dev_v2_header_onetime(rbd_dev); |
| 4471 | if (ret) |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 4472 | return ret; |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4473 | } |
| 4474 | |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4475 | /* |
| 4476 | * If the image supports layering, get the parent info. We |
| 4477 | * need to probe the first time regardless. Thereafter we |
| 4478 | * only need to if there's a parent, to see if it has |
| 4479 | * disappeared due to the mapped image getting flattened. |
| 4480 | */ |
| 4481 | if (rbd_dev->header.features & RBD_FEATURE_LAYERING && |
| 4482 | (first_time || rbd_dev->parent_spec)) { |
| 4483 | bool warn; |
| 4484 | |
| 4485 | ret = rbd_dev_v2_parent_info(rbd_dev); |
| 4486 | if (ret) |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 4487 | return ret; |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4488 | |
| 4489 | /* |
| 4490 | * Print a warning if this is the initial probe and |
| 4491 | * the image has a parent. Don't print it if the |
| 4492 | * image now being probed is itself a parent. We |
| 4493 | * can tell at this point because we won't know its |
| 4494 | * pool name yet (just its pool id). |
| 4495 | */ |
| 4496 | warn = rbd_dev->parent_spec && rbd_dev->spec->pool_name; |
| 4497 | if (first_time && warn) |
| 4498 | rbd_warn(rbd_dev, "WARNING: kernel layering " |
| 4499 | "is EXPERIMENTAL!"); |
| 4500 | } |
| 4501 | |
Alex Elder | 29334ba | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 4502 | if (rbd_dev->spec->snap_id == CEPH_NOSNAP) |
| 4503 | if (rbd_dev->mapping.size != rbd_dev->header.image_size) |
| 4504 | rbd_dev->mapping.size = rbd_dev->header.image_size; |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4505 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4506 | ret = rbd_dev_v2_snap_context(rbd_dev); |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4507 | dout("rbd_dev_v2_snap_context returned %d\n", ret); |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4508 | |
| 4509 | return ret; |
| 4510 | } |
| 4511 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4512 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
| 4513 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4514 | struct device *dev; |
Alex Elder | cd789ab | 2012-08-30 00:16:38 -0500 | [diff] [blame] | 4515 | int ret; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4516 | |
Alex Elder | cd789ab | 2012-08-30 00:16:38 -0500 | [diff] [blame] | 4517 | dev = &rbd_dev->dev; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4518 | dev->bus = &rbd_bus_type; |
| 4519 | dev->type = &rbd_device_type; |
| 4520 | dev->parent = &rbd_root_dev; |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4521 | dev->release = rbd_dev_device_release; |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4522 | dev_set_name(dev, "%d", rbd_dev->dev_id); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4523 | ret = device_register(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4524 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4525 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4526 | } |
| 4527 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4528 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
| 4529 | { |
| 4530 | device_unregister(&rbd_dev->dev); |
| 4531 | } |
| 4532 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4533 | /* |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4534 | * Get a unique rbd identifier for the given new rbd_dev, and add |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 4535 | * the rbd_dev to the global list. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4536 | */ |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 4537 | static int rbd_dev_id_get(struct rbd_device *rbd_dev) |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 4538 | { |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 4539 | int new_dev_id; |
| 4540 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 4541 | new_dev_id = ida_simple_get(&rbd_dev_id_ida, |
| 4542 | 0, minor_to_rbd_dev_id(1 << MINORBITS), |
| 4543 | GFP_KERNEL); |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 4544 | if (new_dev_id < 0) |
| 4545 | return new_dev_id; |
| 4546 | |
| 4547 | rbd_dev->dev_id = new_dev_id; |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4548 | |
| 4549 | spin_lock(&rbd_dev_list_lock); |
| 4550 | list_add_tail(&rbd_dev->node, &rbd_dev_list); |
| 4551 | spin_unlock(&rbd_dev_list_lock); |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 4552 | |
Ilya Dryomov | 70eebd2 | 2013-12-13 15:28:56 +0200 | [diff] [blame] | 4553 | dout("rbd_dev %p given dev id %d\n", rbd_dev, rbd_dev->dev_id); |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 4554 | |
| 4555 | return 0; |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4556 | } |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 4557 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4558 | /* |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4559 | * Remove an rbd_dev from the global list, and record that its |
| 4560 | * identifier is no longer in use. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4561 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4562 | static void rbd_dev_id_put(struct rbd_device *rbd_dev) |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4563 | { |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4564 | spin_lock(&rbd_dev_list_lock); |
| 4565 | list_del_init(&rbd_dev->node); |
| 4566 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 4567 | |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 4568 | ida_simple_remove(&rbd_dev_id_ida, rbd_dev->dev_id); |
| 4569 | |
| 4570 | dout("rbd_dev %p released dev id %d\n", rbd_dev, rbd_dev->dev_id); |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 4571 | } |
| 4572 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4573 | /* |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4574 | * Skips over white space at *buf, and updates *buf to point to the |
| 4575 | * first found non-space character (if any). Returns the length of |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 4576 | * the token (string of non-white space characters) found. Note |
| 4577 | * that *buf must be terminated with '\0'. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4578 | */ |
| 4579 | static inline size_t next_token(const char **buf) |
| 4580 | { |
| 4581 | /* |
| 4582 | * These are the characters that produce nonzero for |
| 4583 | * isspace() in the "C" and "POSIX" locales. |
| 4584 | */ |
| 4585 | const char *spaces = " \f\n\r\t\v"; |
| 4586 | |
| 4587 | *buf += strspn(*buf, spaces); /* Find start of token */ |
| 4588 | |
| 4589 | return strcspn(*buf, spaces); /* Return token length */ |
| 4590 | } |
| 4591 | |
| 4592 | /* |
| 4593 | * Finds the next token in *buf, and if the provided token buffer is |
| 4594 | * big enough, copies the found token into it. The result, if |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 4595 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
| 4596 | * must be terminated with '\0' on entry. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4597 | * |
| 4598 | * Returns the length of the token found (not including the '\0'). |
| 4599 | * Return value will be 0 if no token is found, and it will be >= |
| 4600 | * token_size if the token would not fit. |
| 4601 | * |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 4602 | * 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] | 4603 | * found token. Note that this occurs even if the token buffer is |
| 4604 | * too small to hold it. |
| 4605 | */ |
| 4606 | static inline size_t copy_token(const char **buf, |
| 4607 | char *token, |
| 4608 | size_t token_size) |
| 4609 | { |
| 4610 | size_t len; |
| 4611 | |
| 4612 | len = next_token(buf); |
| 4613 | if (len < token_size) { |
| 4614 | memcpy(token, *buf, len); |
| 4615 | *(token + len) = '\0'; |
| 4616 | } |
| 4617 | *buf += len; |
| 4618 | |
| 4619 | return len; |
| 4620 | } |
| 4621 | |
| 4622 | /* |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 4623 | * Finds the next token in *buf, dynamically allocates a buffer big |
| 4624 | * enough to hold a copy of it, and copies the token into the new |
| 4625 | * buffer. The copy is guaranteed to be terminated with '\0'. Note |
| 4626 | * that a duplicate buffer is created even for a zero-length token. |
| 4627 | * |
| 4628 | * Returns a pointer to the newly-allocated duplicate, or a null |
| 4629 | * pointer if memory for the duplicate was not available. If |
| 4630 | * the lenp argument is a non-null pointer, the length of the token |
| 4631 | * (not including the '\0') is returned in *lenp. |
| 4632 | * |
| 4633 | * If successful, the *buf pointer will be updated to point beyond |
| 4634 | * the end of the found token. |
| 4635 | * |
| 4636 | * Note: uses GFP_KERNEL for allocation. |
| 4637 | */ |
| 4638 | static inline char *dup_token(const char **buf, size_t *lenp) |
| 4639 | { |
| 4640 | char *dup; |
| 4641 | size_t len; |
| 4642 | |
| 4643 | len = next_token(buf); |
Alex Elder | 4caf35f | 2012-11-01 08:39:27 -0500 | [diff] [blame] | 4644 | dup = kmemdup(*buf, len + 1, GFP_KERNEL); |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 4645 | if (!dup) |
| 4646 | return NULL; |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 4647 | *(dup + len) = '\0'; |
| 4648 | *buf += len; |
| 4649 | |
| 4650 | if (lenp) |
| 4651 | *lenp = len; |
| 4652 | |
| 4653 | return dup; |
| 4654 | } |
| 4655 | |
| 4656 | /* |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4657 | * Parse the options provided for an "rbd add" (i.e., rbd image |
| 4658 | * mapping) request. These arrive via a write to /sys/bus/rbd/add, |
| 4659 | * and the data written is passed here via a NUL-terminated buffer. |
| 4660 | * Returns 0 if successful or an error code otherwise. |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4661 | * |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4662 | * The information extracted from these options is recorded in |
| 4663 | * the other parameters which return dynamically-allocated |
| 4664 | * structures: |
| 4665 | * ceph_opts |
| 4666 | * The address of a pointer that will refer to a ceph options |
| 4667 | * structure. Caller must release the returned pointer using |
| 4668 | * ceph_destroy_options() when it is no longer needed. |
| 4669 | * rbd_opts |
| 4670 | * Address of an rbd options pointer. Fully initialized by |
| 4671 | * this function; caller must release with kfree(). |
| 4672 | * spec |
| 4673 | * Address of an rbd image specification pointer. Fully |
| 4674 | * initialized by this function based on parsed options. |
| 4675 | * Caller must release with rbd_spec_put(). |
| 4676 | * |
| 4677 | * The options passed take this form: |
| 4678 | * <mon_addrs> <options> <pool_name> <image_name> [<snap_id>] |
| 4679 | * where: |
| 4680 | * <mon_addrs> |
| 4681 | * A comma-separated list of one or more monitor addresses. |
| 4682 | * A monitor address is an ip address, optionally followed |
| 4683 | * by a port number (separated by a colon). |
| 4684 | * I.e.: ip1[:port1][,ip2[:port2]...] |
| 4685 | * <options> |
| 4686 | * A comma-separated list of ceph and/or rbd options. |
| 4687 | * <pool_name> |
| 4688 | * The name of the rados pool containing the rbd image. |
| 4689 | * <image_name> |
| 4690 | * The name of the image in that pool to map. |
| 4691 | * <snap_id> |
| 4692 | * An optional snapshot id. If provided, the mapping will |
| 4693 | * present data from the image at the time that snapshot was |
| 4694 | * created. The image head is used if no snapshot id is |
| 4695 | * provided. Snapshot mappings are always read-only. |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4696 | */ |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4697 | static int rbd_add_parse_args(const char *buf, |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4698 | struct ceph_options **ceph_opts, |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4699 | struct rbd_options **opts, |
| 4700 | struct rbd_spec **rbd_spec) |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4701 | { |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4702 | size_t len; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4703 | char *options; |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4704 | const char *mon_addrs; |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4705 | char *snap_name; |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4706 | size_t mon_addrs_size; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4707 | struct rbd_spec *spec = NULL; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4708 | struct rbd_options *rbd_opts = NULL; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4709 | struct ceph_options *copts; |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4710 | int ret; |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4711 | |
| 4712 | /* The first four tokens are required */ |
| 4713 | |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4714 | len = next_token(&buf); |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4715 | if (!len) { |
| 4716 | rbd_warn(NULL, "no monitor address(es) provided"); |
| 4717 | return -EINVAL; |
| 4718 | } |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4719 | mon_addrs = buf; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4720 | mon_addrs_size = len + 1; |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4721 | buf += len; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4722 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4723 | ret = -EINVAL; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4724 | options = dup_token(&buf, NULL); |
| 4725 | if (!options) |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4726 | return -ENOMEM; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4727 | if (!*options) { |
| 4728 | rbd_warn(NULL, "no options provided"); |
| 4729 | goto out_err; |
| 4730 | } |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4731 | |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4732 | spec = rbd_spec_alloc(); |
| 4733 | if (!spec) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4734 | goto out_mem; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4735 | |
| 4736 | spec->pool_name = dup_token(&buf, NULL); |
| 4737 | if (!spec->pool_name) |
| 4738 | goto out_mem; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4739 | if (!*spec->pool_name) { |
| 4740 | rbd_warn(NULL, "no pool name provided"); |
| 4741 | goto out_err; |
| 4742 | } |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4743 | |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 4744 | spec->image_name = dup_token(&buf, NULL); |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4745 | if (!spec->image_name) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4746 | goto out_mem; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4747 | if (!*spec->image_name) { |
| 4748 | rbd_warn(NULL, "no image name provided"); |
| 4749 | goto out_err; |
| 4750 | } |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4751 | |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4752 | /* |
| 4753 | * Snapshot name is optional; default is to use "-" |
| 4754 | * (indicating the head/no snapshot). |
| 4755 | */ |
Alex Elder | 3feeb894 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 4756 | len = next_token(&buf); |
Alex Elder | 820a5f3 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 4757 | if (!len) { |
Alex Elder | 3feeb894 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 4758 | buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ |
| 4759 | len = sizeof (RBD_SNAP_HEAD_NAME) - 1; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4760 | } else if (len > RBD_MAX_SNAP_NAME_LEN) { |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4761 | ret = -ENAMETOOLONG; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4762 | goto out_err; |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 4763 | } |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4764 | snap_name = kmemdup(buf, len + 1, GFP_KERNEL); |
| 4765 | if (!snap_name) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4766 | goto out_mem; |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4767 | *(snap_name + len) = '\0'; |
| 4768 | spec->snap_name = snap_name; |
Alex Elder | e5c3553 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4769 | |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4770 | /* Initialize all rbd options to the defaults */ |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4771 | |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4772 | rbd_opts = kzalloc(sizeof (*rbd_opts), GFP_KERNEL); |
| 4773 | if (!rbd_opts) |
| 4774 | goto out_mem; |
| 4775 | |
| 4776 | rbd_opts->read_only = RBD_READ_ONLY_DEFAULT; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4777 | |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4778 | copts = ceph_parse_options(options, mon_addrs, |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4779 | mon_addrs + mon_addrs_size - 1, |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4780 | parse_rbd_opts_token, rbd_opts); |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4781 | if (IS_ERR(copts)) { |
| 4782 | ret = PTR_ERR(copts); |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4783 | goto out_err; |
| 4784 | } |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4785 | kfree(options); |
| 4786 | |
| 4787 | *ceph_opts = copts; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4788 | *opts = rbd_opts; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4789 | *rbd_spec = spec; |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4790 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4791 | return 0; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4792 | out_mem: |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4793 | ret = -ENOMEM; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4794 | out_err: |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4795 | kfree(rbd_opts); |
| 4796 | rbd_spec_put(spec); |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4797 | kfree(options); |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4798 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4799 | return ret; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4800 | } |
| 4801 | |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4802 | /* |
Ilya Dryomov | 30ba1f0 | 2014-05-13 11:19:27 +0400 | [diff] [blame] | 4803 | * Return pool id (>= 0) or a negative error code. |
| 4804 | */ |
| 4805 | static int rbd_add_get_pool_id(struct rbd_client *rbdc, const char *pool_name) |
| 4806 | { |
| 4807 | u64 newest_epoch; |
| 4808 | unsigned long timeout = rbdc->client->options->mount_timeout * HZ; |
| 4809 | int tries = 0; |
| 4810 | int ret; |
| 4811 | |
| 4812 | again: |
| 4813 | ret = ceph_pg_poolid_by_name(rbdc->client->osdc.osdmap, pool_name); |
| 4814 | if (ret == -ENOENT && tries++ < 1) { |
| 4815 | ret = ceph_monc_do_get_version(&rbdc->client->monc, "osdmap", |
| 4816 | &newest_epoch); |
| 4817 | if (ret < 0) |
| 4818 | return ret; |
| 4819 | |
| 4820 | if (rbdc->client->osdc.osdmap->epoch < newest_epoch) { |
| 4821 | ceph_monc_request_next_osdmap(&rbdc->client->monc); |
| 4822 | (void) ceph_monc_wait_osdmap(&rbdc->client->monc, |
| 4823 | newest_epoch, timeout); |
| 4824 | goto again; |
| 4825 | } else { |
| 4826 | /* the osdmap we have is new enough */ |
| 4827 | return -ENOENT; |
| 4828 | } |
| 4829 | } |
| 4830 | |
| 4831 | return ret; |
| 4832 | } |
| 4833 | |
| 4834 | /* |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4835 | * An rbd format 2 image has a unique identifier, distinct from the |
| 4836 | * name given to it by the user. Internally, that identifier is |
| 4837 | * what's used to specify the names of objects related to the image. |
| 4838 | * |
| 4839 | * A special "rbd id" object is used to map an rbd image name to its |
| 4840 | * id. If that object doesn't exist, then there is no v2 rbd image |
| 4841 | * with the supplied name. |
| 4842 | * |
| 4843 | * This function will record the given rbd_dev's image_id field if |
| 4844 | * it can be determined, and in that case will return 0. If any |
| 4845 | * errors occur a negative errno will be returned and the rbd_dev's |
| 4846 | * image_id field will be unchanged (and should be NULL). |
| 4847 | */ |
| 4848 | static int rbd_dev_image_id(struct rbd_device *rbd_dev) |
| 4849 | { |
| 4850 | int ret; |
| 4851 | size_t size; |
| 4852 | char *object_name; |
| 4853 | void *response; |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4854 | char *image_id; |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4855 | |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4856 | /* |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4857 | * When probing a parent image, the image id is already |
| 4858 | * known (and the image name likely is not). There's no |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4859 | * need to fetch the image id again in this case. We |
| 4860 | * do still need to set the image format though. |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4861 | */ |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4862 | if (rbd_dev->spec->image_id) { |
| 4863 | rbd_dev->image_format = *rbd_dev->spec->image_id ? 2 : 1; |
| 4864 | |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4865 | return 0; |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4866 | } |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4867 | |
| 4868 | /* |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4869 | * First, see if the format 2 image id file exists, and if |
| 4870 | * so, get the image's persistent id from it. |
| 4871 | */ |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 4872 | size = sizeof (RBD_ID_PREFIX) + strlen(rbd_dev->spec->image_name); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4873 | object_name = kmalloc(size, GFP_NOIO); |
| 4874 | if (!object_name) |
| 4875 | return -ENOMEM; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4876 | 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] | 4877 | dout("rbd id object name is %s\n", object_name); |
| 4878 | |
| 4879 | /* Response will be an encoded string, which includes a length */ |
| 4880 | |
| 4881 | size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; |
| 4882 | response = kzalloc(size, GFP_NOIO); |
| 4883 | if (!response) { |
| 4884 | ret = -ENOMEM; |
| 4885 | goto out; |
| 4886 | } |
| 4887 | |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4888 | /* If it doesn't exist we'll assume it's a format 1 image */ |
| 4889 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4890 | ret = rbd_obj_method_sync(rbd_dev, object_name, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4891 | "rbd", "get_id", NULL, 0, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4892 | response, RBD_IMAGE_ID_LEN_MAX); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4893 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4894 | if (ret == -ENOENT) { |
| 4895 | image_id = kstrdup("", GFP_KERNEL); |
| 4896 | ret = image_id ? 0 : -ENOMEM; |
| 4897 | if (!ret) |
| 4898 | rbd_dev->image_format = 1; |
| 4899 | } else if (ret > sizeof (__le32)) { |
| 4900 | void *p = response; |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4901 | |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4902 | image_id = ceph_extract_encoded_string(&p, p + ret, |
Alex Elder | 979ed48 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 4903 | NULL, GFP_NOIO); |
Duan Jiong | 461f758 | 2014-04-11 16:38:12 +0800 | [diff] [blame] | 4904 | ret = PTR_ERR_OR_ZERO(image_id); |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4905 | if (!ret) |
| 4906 | rbd_dev->image_format = 2; |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4907 | } else { |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4908 | ret = -EINVAL; |
| 4909 | } |
| 4910 | |
| 4911 | if (!ret) { |
| 4912 | rbd_dev->spec->image_id = image_id; |
| 4913 | dout("image_id is %s\n", image_id); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4914 | } |
| 4915 | out: |
| 4916 | kfree(response); |
| 4917 | kfree(object_name); |
| 4918 | |
| 4919 | return ret; |
| 4920 | } |
| 4921 | |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 4922 | /* |
| 4923 | * Undo whatever state changes are made by v1 or v2 header info |
| 4924 | * call. |
| 4925 | */ |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4926 | static void rbd_dev_unprobe(struct rbd_device *rbd_dev) |
| 4927 | { |
| 4928 | struct rbd_image_header *header; |
| 4929 | |
Alex Elder | 392a9da | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4930 | /* Drop parent reference unless it's already been done (or none) */ |
| 4931 | |
| 4932 | if (rbd_dev->parent_overlap) |
| 4933 | rbd_dev_parent_put(rbd_dev); |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4934 | |
| 4935 | /* Free dynamic fields from the header, then zero it out */ |
| 4936 | |
| 4937 | header = &rbd_dev->header; |
Alex Elder | 812164f8 | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 4938 | ceph_put_snap_context(header->snapc); |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4939 | kfree(header->snap_sizes); |
| 4940 | kfree(header->snap_names); |
| 4941 | kfree(header->object_prefix); |
| 4942 | memset(header, 0, sizeof (*header)); |
| 4943 | } |
| 4944 | |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4945 | static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev) |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4946 | { |
| 4947 | int ret; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4948 | |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4949 | ret = rbd_dev_v2_object_prefix(rbd_dev); |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4950 | if (ret) |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4951 | goto out_err; |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4952 | |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4953 | /* |
| 4954 | * Get the and check features for the image. Currently the |
| 4955 | * features are assumed to never change. |
| 4956 | */ |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4957 | ret = rbd_dev_v2_features(rbd_dev); |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4958 | if (ret) |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4959 | goto out_err; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4960 | |
Alex Elder | cc070d5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4961 | /* If the image supports fancy striping, get its parameters */ |
| 4962 | |
| 4963 | if (rbd_dev->header.features & RBD_FEATURE_STRIPINGV2) { |
| 4964 | ret = rbd_dev_v2_striping_info(rbd_dev); |
| 4965 | if (ret < 0) |
| 4966 | goto out_err; |
| 4967 | } |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4968 | /* No support for crypto and compression type format 2 images */ |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4969 | |
Alex Elder | 3515297 | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4970 | return 0; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4971 | out_err: |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4972 | rbd_dev->header.features = 0; |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4973 | kfree(rbd_dev->header.object_prefix); |
| 4974 | rbd_dev->header.object_prefix = NULL; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4975 | |
| 4976 | return ret; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4977 | } |
| 4978 | |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4979 | static int rbd_dev_probe_parent(struct rbd_device *rbd_dev) |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4980 | { |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4981 | struct rbd_device *parent = NULL; |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4982 | struct rbd_spec *parent_spec; |
| 4983 | struct rbd_client *rbdc; |
| 4984 | int ret; |
| 4985 | |
| 4986 | if (!rbd_dev->parent_spec) |
| 4987 | return 0; |
| 4988 | /* |
| 4989 | * We need to pass a reference to the client and the parent |
| 4990 | * spec when creating the parent rbd_dev. Images related by |
| 4991 | * parent/child relationships always share both. |
| 4992 | */ |
| 4993 | parent_spec = rbd_spec_get(rbd_dev->parent_spec); |
| 4994 | rbdc = __rbd_get_client(rbd_dev->rbd_client); |
| 4995 | |
| 4996 | ret = -ENOMEM; |
| 4997 | parent = rbd_dev_create(rbdc, parent_spec); |
| 4998 | if (!parent) |
| 4999 | goto out_err; |
| 5000 | |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5001 | ret = rbd_dev_image_probe(parent, false); |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5002 | if (ret < 0) |
| 5003 | goto out_err; |
| 5004 | rbd_dev->parent = parent; |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 5005 | atomic_set(&rbd_dev->parent_ref, 1); |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5006 | |
| 5007 | return 0; |
| 5008 | out_err: |
| 5009 | if (parent) { |
Alex Elder | fb65d228 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 5010 | rbd_dev_unparent(rbd_dev); |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5011 | kfree(rbd_dev->header_name); |
| 5012 | rbd_dev_destroy(parent); |
| 5013 | } else { |
| 5014 | rbd_put_client(rbdc); |
| 5015 | rbd_spec_put(parent_spec); |
| 5016 | } |
| 5017 | |
| 5018 | return ret; |
| 5019 | } |
| 5020 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5021 | static int rbd_dev_device_setup(struct rbd_device *rbd_dev) |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5022 | { |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5023 | int ret; |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5024 | |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5025 | /* Get an id and fill in device name. */ |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5026 | |
Ilya Dryomov | f8a22fc | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5027 | ret = rbd_dev_id_get(rbd_dev); |
| 5028 | if (ret) |
| 5029 | return ret; |
| 5030 | |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5031 | BUILD_BUG_ON(DEV_NAME_LEN |
| 5032 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); |
| 5033 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); |
| 5034 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5035 | /* Record our major and minor device numbers. */ |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5036 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5037 | if (!single_major) { |
| 5038 | ret = register_blkdev(0, rbd_dev->name); |
| 5039 | if (ret < 0) |
| 5040 | goto err_out_id; |
| 5041 | |
| 5042 | rbd_dev->major = ret; |
| 5043 | rbd_dev->minor = 0; |
| 5044 | } else { |
| 5045 | rbd_dev->major = rbd_major; |
| 5046 | rbd_dev->minor = rbd_dev_id_to_minor(rbd_dev->dev_id); |
| 5047 | } |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5048 | |
| 5049 | /* Set up the blkdev mapping. */ |
| 5050 | |
| 5051 | ret = rbd_init_disk(rbd_dev); |
| 5052 | if (ret) |
| 5053 | goto err_out_blkdev; |
| 5054 | |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 5055 | ret = rbd_dev_mapping_set(rbd_dev); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5056 | if (ret) |
| 5057 | goto err_out_disk; |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 5058 | set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); |
Josh Durgin | 22001f6 | 2013-09-30 20:10:04 -0700 | [diff] [blame] | 5059 | set_disk_ro(rbd_dev->disk, rbd_dev->mapping.read_only); |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 5060 | |
| 5061 | ret = rbd_bus_add_dev(rbd_dev); |
| 5062 | if (ret) |
| 5063 | goto err_out_mapping; |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5064 | |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5065 | /* Everything's ready. Announce the disk to the world. */ |
| 5066 | |
Alex Elder | 129b79d | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5067 | set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5068 | add_disk(rbd_dev->disk); |
| 5069 | |
| 5070 | pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name, |
| 5071 | (unsigned long long) rbd_dev->mapping.size); |
| 5072 | |
| 5073 | return ret; |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 5074 | |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 5075 | err_out_mapping: |
| 5076 | rbd_dev_mapping_clear(rbd_dev); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5077 | err_out_disk: |
| 5078 | rbd_free_disk(rbd_dev); |
| 5079 | err_out_blkdev: |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5080 | if (!single_major) |
| 5081 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5082 | err_out_id: |
| 5083 | rbd_dev_id_put(rbd_dev); |
Alex Elder | d1cf578 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5084 | rbd_dev_mapping_clear(rbd_dev); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5085 | |
| 5086 | return ret; |
| 5087 | } |
| 5088 | |
Alex Elder | 332bb12 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5089 | static int rbd_dev_header_name(struct rbd_device *rbd_dev) |
| 5090 | { |
| 5091 | struct rbd_spec *spec = rbd_dev->spec; |
| 5092 | size_t size; |
| 5093 | |
| 5094 | /* Record the header object name for this rbd image. */ |
| 5095 | |
| 5096 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 5097 | |
| 5098 | if (rbd_dev->image_format == 1) |
| 5099 | size = strlen(spec->image_name) + sizeof (RBD_SUFFIX); |
| 5100 | else |
| 5101 | size = sizeof (RBD_HEADER_PREFIX) + strlen(spec->image_id); |
| 5102 | |
| 5103 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); |
| 5104 | if (!rbd_dev->header_name) |
| 5105 | return -ENOMEM; |
| 5106 | |
| 5107 | if (rbd_dev->image_format == 1) |
| 5108 | sprintf(rbd_dev->header_name, "%s%s", |
| 5109 | spec->image_name, RBD_SUFFIX); |
| 5110 | else |
| 5111 | sprintf(rbd_dev->header_name, "%s%s", |
| 5112 | RBD_HEADER_PREFIX, spec->image_id); |
| 5113 | return 0; |
| 5114 | } |
| 5115 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5116 | static void rbd_dev_image_release(struct rbd_device *rbd_dev) |
| 5117 | { |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5118 | rbd_dev_unprobe(rbd_dev); |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5119 | kfree(rbd_dev->header_name); |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5120 | rbd_dev->header_name = NULL; |
| 5121 | rbd_dev->image_format = 0; |
| 5122 | kfree(rbd_dev->spec->image_id); |
| 5123 | rbd_dev->spec->image_id = NULL; |
| 5124 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5125 | rbd_dev_destroy(rbd_dev); |
| 5126 | } |
| 5127 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 5128 | /* |
| 5129 | * Probe for the existence of the header object for the given rbd |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5130 | * device. If this image is the one being mapped (i.e., not a |
| 5131 | * parent), initiate a watch on its header object before using that |
| 5132 | * object to get detailed information about the rbd image. |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 5133 | */ |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5134 | static int rbd_dev_image_probe(struct rbd_device *rbd_dev, bool mapping) |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 5135 | { |
| 5136 | int ret; |
| 5137 | |
| 5138 | /* |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 5139 | * Get the id from the image id object. Unless there's an |
| 5140 | * error, rbd_dev->spec->image_id will be filled in with |
| 5141 | * a dynamically-allocated string, and rbd_dev->image_format |
| 5142 | * will be set to either 1 or 2. |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 5143 | */ |
| 5144 | ret = rbd_dev_image_id(rbd_dev); |
| 5145 | if (ret) |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 5146 | return ret; |
| 5147 | rbd_assert(rbd_dev->spec->image_id); |
| 5148 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 5149 | |
Alex Elder | 332bb12 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5150 | ret = rbd_dev_header_name(rbd_dev); |
| 5151 | if (ret) |
| 5152 | goto err_out_format; |
| 5153 | |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5154 | if (mapping) { |
Ilya Dryomov | fca2706 | 2013-12-16 18:02:40 +0200 | [diff] [blame] | 5155 | ret = rbd_dev_header_watch_sync(rbd_dev); |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5156 | if (ret) |
| 5157 | goto out_header_name; |
| 5158 | } |
Alex Elder | b644de2 | 2013-04-27 09:59:31 -0500 | [diff] [blame] | 5159 | |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 5160 | if (rbd_dev->image_format == 1) |
Alex Elder | 99a41eb | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 5161 | ret = rbd_dev_v1_header_info(rbd_dev); |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 5162 | else |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 5163 | ret = rbd_dev_v2_header_info(rbd_dev); |
Alex Elder | 5655c4d | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 5164 | if (ret) |
Alex Elder | b644de2 | 2013-04-27 09:59:31 -0500 | [diff] [blame] | 5165 | goto err_out_watch; |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5166 | |
Alex Elder | 9bb81c9 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5167 | ret = rbd_dev_spec_update(rbd_dev); |
| 5168 | if (ret) |
Alex Elder | 33dca39 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 5169 | goto err_out_probe; |
Alex Elder | 9bb81c9 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5170 | |
| 5171 | ret = rbd_dev_probe_parent(rbd_dev); |
Alex Elder | 30d60ba | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 5172 | if (ret) |
| 5173 | goto err_out_probe; |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 5174 | |
Alex Elder | 30d60ba | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 5175 | dout("discovered format %u image, header name is %s\n", |
| 5176 | rbd_dev->image_format, rbd_dev->header_name); |
| 5177 | |
| 5178 | return 0; |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5179 | err_out_probe: |
| 5180 | rbd_dev_unprobe(rbd_dev); |
Alex Elder | b644de2 | 2013-04-27 09:59:31 -0500 | [diff] [blame] | 5181 | err_out_watch: |
Ilya Dryomov | fca2706 | 2013-12-16 18:02:40 +0200 | [diff] [blame] | 5182 | if (mapping) |
| 5183 | rbd_dev_header_unwatch_sync(rbd_dev); |
Alex Elder | 332bb12 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5184 | out_header_name: |
| 5185 | kfree(rbd_dev->header_name); |
| 5186 | rbd_dev->header_name = NULL; |
| 5187 | err_out_format: |
| 5188 | rbd_dev->image_format = 0; |
Alex Elder | 5655c4d | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 5189 | kfree(rbd_dev->spec->image_id); |
| 5190 | rbd_dev->spec->image_id = NULL; |
| 5191 | |
| 5192 | dout("probe failed, returning %d\n", ret); |
| 5193 | |
| 5194 | return ret; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 5195 | } |
| 5196 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5197 | static ssize_t do_rbd_add(struct bus_type *bus, |
| 5198 | const char *buf, |
| 5199 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5200 | { |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 5201 | struct rbd_device *rbd_dev = NULL; |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5202 | struct ceph_options *ceph_opts = NULL; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5203 | struct rbd_options *rbd_opts = NULL; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5204 | struct rbd_spec *spec = NULL; |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5205 | struct rbd_client *rbdc; |
Alex Elder | 51344a3 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 5206 | bool read_only; |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5207 | int rc = -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5208 | |
| 5209 | if (!try_module_get(THIS_MODULE)) |
| 5210 | return -ENODEV; |
| 5211 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5212 | /* parse add command */ |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5213 | rc = rbd_add_parse_args(buf, &ceph_opts, &rbd_opts, &spec); |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5214 | if (rc < 0) |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5215 | goto err_out_module; |
Alex Elder | 51344a3 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 5216 | read_only = rbd_opts->read_only; |
| 5217 | kfree(rbd_opts); |
| 5218 | rbd_opts = NULL; /* done with this */ |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5219 | |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5220 | rbdc = rbd_get_client(ceph_opts); |
| 5221 | if (IS_ERR(rbdc)) { |
| 5222 | rc = PTR_ERR(rbdc); |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5223 | goto err_out_args; |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5224 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5225 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5226 | /* pick the pool */ |
Ilya Dryomov | 30ba1f0 | 2014-05-13 11:19:27 +0400 | [diff] [blame] | 5227 | rc = rbd_add_get_pool_id(rbdc, spec->pool_name); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5228 | if (rc < 0) |
| 5229 | goto err_out_client; |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 5230 | spec->pool_id = (u64)rc; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5231 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 5232 | /* The ceph file layout needs to fit pool id in 32 bits */ |
| 5233 | |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 5234 | if (spec->pool_id > (u64)U32_MAX) { |
| 5235 | rbd_warn(NULL, "pool id too large (%llu > %u)\n", |
| 5236 | (unsigned long long)spec->pool_id, U32_MAX); |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 5237 | rc = -EIO; |
| 5238 | goto err_out_client; |
| 5239 | } |
| 5240 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5241 | rbd_dev = rbd_dev_create(rbdc, spec); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5242 | if (!rbd_dev) |
| 5243 | goto err_out_client; |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5244 | rbdc = NULL; /* rbd_dev now owns this */ |
| 5245 | spec = NULL; /* rbd_dev now owns this */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5246 | |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5247 | rc = rbd_dev_image_probe(rbd_dev, true); |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 5248 | if (rc < 0) |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5249 | goto err_out_rbd_dev; |
Alex Elder | 05fd6f6 | 2012-08-29 17:11:07 -0500 | [diff] [blame] | 5250 | |
Alex Elder | 7ce4eef | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5251 | /* If we are mapping a snapshot it must be marked read-only */ |
| 5252 | |
| 5253 | if (rbd_dev->spec->snap_id != CEPH_NOSNAP) |
| 5254 | read_only = true; |
| 5255 | rbd_dev->mapping.read_only = read_only; |
| 5256 | |
Alex Elder | b536f69 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5257 | rc = rbd_dev_device_setup(rbd_dev); |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 5258 | if (rc) { |
Ilya Dryomov | e37180c | 2013-12-16 18:02:41 +0200 | [diff] [blame] | 5259 | /* |
| 5260 | * rbd_dev_header_unwatch_sync() can't be moved into |
| 5261 | * rbd_dev_image_release() without refactoring, see |
| 5262 | * commit 1f3ef78861ac. |
| 5263 | */ |
| 5264 | rbd_dev_header_unwatch_sync(rbd_dev); |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 5265 | rbd_dev_image_release(rbd_dev); |
| 5266 | goto err_out_module; |
| 5267 | } |
Alex Elder | b536f69 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5268 | |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 5269 | return count; |
| 5270 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5271 | err_out_rbd_dev: |
| 5272 | rbd_dev_destroy(rbd_dev); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5273 | err_out_client: |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5274 | rbd_put_client(rbdc); |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5275 | err_out_args: |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5276 | rbd_spec_put(spec); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5277 | err_out_module: |
| 5278 | module_put(THIS_MODULE); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5279 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5280 | dout("Error adding device %s\n", buf); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5281 | |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 5282 | return (ssize_t)rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5283 | } |
| 5284 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5285 | static ssize_t rbd_add(struct bus_type *bus, |
| 5286 | const char *buf, |
| 5287 | size_t count) |
| 5288 | { |
| 5289 | if (single_major) |
| 5290 | return -EINVAL; |
| 5291 | |
| 5292 | return do_rbd_add(bus, buf, count); |
| 5293 | } |
| 5294 | |
| 5295 | static ssize_t rbd_add_single_major(struct bus_type *bus, |
| 5296 | const char *buf, |
| 5297 | size_t count) |
| 5298 | { |
| 5299 | return do_rbd_add(bus, buf, count); |
| 5300 | } |
| 5301 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5302 | static void rbd_dev_device_release(struct device *dev) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5303 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 5304 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5305 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5306 | rbd_free_disk(rbd_dev); |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5307 | clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
Alex Elder | 6d80b13 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 5308 | rbd_dev_mapping_clear(rbd_dev); |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5309 | if (!single_major) |
| 5310 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 5311 | rbd_dev_id_put(rbd_dev); |
Alex Elder | d1cf578 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5312 | rbd_dev_mapping_clear(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5313 | } |
| 5314 | |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5315 | static void rbd_dev_remove_parent(struct rbd_device *rbd_dev) |
| 5316 | { |
Alex Elder | ad945fc | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5317 | while (rbd_dev->parent) { |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5318 | struct rbd_device *first = rbd_dev; |
| 5319 | struct rbd_device *second = first->parent; |
| 5320 | struct rbd_device *third; |
| 5321 | |
| 5322 | /* |
| 5323 | * Follow to the parent with no grandparent and |
| 5324 | * remove it. |
| 5325 | */ |
| 5326 | while (second && (third = second->parent)) { |
| 5327 | first = second; |
| 5328 | second = third; |
| 5329 | } |
Alex Elder | ad945fc | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5330 | rbd_assert(second); |
Alex Elder | 8ad42cd | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5331 | rbd_dev_image_release(second); |
Alex Elder | ad945fc | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5332 | first->parent = NULL; |
| 5333 | first->parent_overlap = 0; |
| 5334 | |
| 5335 | rbd_assert(first->parent_spec); |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5336 | rbd_spec_put(first->parent_spec); |
| 5337 | first->parent_spec = NULL; |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5338 | } |
| 5339 | } |
| 5340 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5341 | static ssize_t do_rbd_remove(struct bus_type *bus, |
| 5342 | const char *buf, |
| 5343 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5344 | { |
| 5345 | struct rbd_device *rbd_dev = NULL; |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5346 | struct list_head *tmp; |
| 5347 | int dev_id; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5348 | unsigned long ul; |
Alex Elder | 82a442d | 2013-05-31 17:40:44 -0500 | [diff] [blame] | 5349 | bool already = false; |
Alex Elder | 0d8189e | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5350 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5351 | |
Jingoo Han | bb8e0e8 | 2013-09-11 14:20:07 -0700 | [diff] [blame] | 5352 | ret = kstrtoul(buf, 10, &ul); |
Alex Elder | 0d8189e | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5353 | if (ret) |
| 5354 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5355 | |
| 5356 | /* convert to int; abort if we lost anything in the conversion */ |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5357 | dev_id = (int)ul; |
| 5358 | if (dev_id != ul) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5359 | return -EINVAL; |
| 5360 | |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5361 | ret = -ENOENT; |
| 5362 | spin_lock(&rbd_dev_list_lock); |
| 5363 | list_for_each(tmp, &rbd_dev_list) { |
| 5364 | rbd_dev = list_entry(tmp, struct rbd_device, node); |
| 5365 | if (rbd_dev->dev_id == dev_id) { |
| 5366 | ret = 0; |
| 5367 | break; |
| 5368 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5369 | } |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5370 | if (!ret) { |
| 5371 | spin_lock_irq(&rbd_dev->lock); |
| 5372 | if (rbd_dev->open_count) |
| 5373 | ret = -EBUSY; |
| 5374 | else |
Alex Elder | 82a442d | 2013-05-31 17:40:44 -0500 | [diff] [blame] | 5375 | already = test_and_set_bit(RBD_DEV_FLAG_REMOVING, |
| 5376 | &rbd_dev->flags); |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5377 | spin_unlock_irq(&rbd_dev->lock); |
| 5378 | } |
| 5379 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | 82a442d | 2013-05-31 17:40:44 -0500 | [diff] [blame] | 5380 | if (ret < 0 || already) |
Alex Elder | 1ba0f1e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5381 | return ret; |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5382 | |
Ilya Dryomov | fca2706 | 2013-12-16 18:02:40 +0200 | [diff] [blame] | 5383 | rbd_dev_header_unwatch_sync(rbd_dev); |
Josh Durgin | 9abc599 | 2013-08-29 17:31:03 -0700 | [diff] [blame] | 5384 | /* |
| 5385 | * flush remaining watch callbacks - these must be complete |
| 5386 | * before the osd_client is shutdown |
| 5387 | */ |
| 5388 | dout("%s: flushing notifies", __func__); |
| 5389 | ceph_osdc_flush_notifies(&rbd_dev->rbd_client->client->osdc); |
Ilya Dryomov | fca2706 | 2013-12-16 18:02:40 +0200 | [diff] [blame] | 5390 | |
Josh Durgin | 9875201 | 2013-08-29 17:26:31 -0700 | [diff] [blame] | 5391 | /* |
| 5392 | * Don't free anything from rbd_dev->disk until after all |
| 5393 | * notifies are completely processed. Otherwise |
| 5394 | * rbd_bus_del_dev() will race with rbd_watch_cb(), resulting |
| 5395 | * in a potential use after free of rbd_dev->disk or rbd_dev. |
| 5396 | */ |
| 5397 | rbd_bus_del_dev(rbd_dev); |
Alex Elder | 8ad42cd | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5398 | rbd_dev_image_release(rbd_dev); |
Alex Elder | 79ab755 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5399 | module_put(THIS_MODULE); |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 5400 | |
Alex Elder | 1ba0f1e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5401 | return count; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5402 | } |
| 5403 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5404 | static ssize_t rbd_remove(struct bus_type *bus, |
| 5405 | const char *buf, |
| 5406 | size_t count) |
| 5407 | { |
| 5408 | if (single_major) |
| 5409 | return -EINVAL; |
| 5410 | |
| 5411 | return do_rbd_remove(bus, buf, count); |
| 5412 | } |
| 5413 | |
| 5414 | static ssize_t rbd_remove_single_major(struct bus_type *bus, |
| 5415 | const char *buf, |
| 5416 | size_t count) |
| 5417 | { |
| 5418 | return do_rbd_remove(bus, buf, count); |
| 5419 | } |
| 5420 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5421 | /* |
| 5422 | * create control files in sysfs |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5423 | * /sys/bus/rbd/... |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5424 | */ |
| 5425 | static int rbd_sysfs_init(void) |
| 5426 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5427 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5428 | |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 5429 | ret = device_register(&rbd_root_dev); |
Alex Elder | 2107978 | 2012-01-24 10:08:36 -0600 | [diff] [blame] | 5430 | if (ret < 0) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5431 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5432 | |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 5433 | ret = bus_register(&rbd_bus_type); |
| 5434 | if (ret < 0) |
| 5435 | device_unregister(&rbd_root_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5436 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5437 | return ret; |
| 5438 | } |
| 5439 | |
| 5440 | static void rbd_sysfs_cleanup(void) |
| 5441 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5442 | bus_unregister(&rbd_bus_type); |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 5443 | device_unregister(&rbd_root_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5444 | } |
| 5445 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5446 | static int rbd_slab_init(void) |
| 5447 | { |
| 5448 | rbd_assert(!rbd_img_request_cache); |
| 5449 | rbd_img_request_cache = kmem_cache_create("rbd_img_request", |
| 5450 | sizeof (struct rbd_img_request), |
| 5451 | __alignof__(struct rbd_img_request), |
| 5452 | 0, NULL); |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5453 | if (!rbd_img_request_cache) |
| 5454 | return -ENOMEM; |
| 5455 | |
| 5456 | rbd_assert(!rbd_obj_request_cache); |
| 5457 | rbd_obj_request_cache = kmem_cache_create("rbd_obj_request", |
| 5458 | sizeof (struct rbd_obj_request), |
| 5459 | __alignof__(struct rbd_obj_request), |
| 5460 | 0, NULL); |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 5461 | if (!rbd_obj_request_cache) |
| 5462 | goto out_err; |
| 5463 | |
| 5464 | rbd_assert(!rbd_segment_name_cache); |
| 5465 | rbd_segment_name_cache = kmem_cache_create("rbd_segment_name", |
Ilya Dryomov | 2d0ebc5 | 2014-01-27 17:40:18 +0200 | [diff] [blame] | 5466 | CEPH_MAX_OID_NAME_LEN + 1, 1, 0, NULL); |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 5467 | if (rbd_segment_name_cache) |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5468 | return 0; |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 5469 | out_err: |
| 5470 | if (rbd_obj_request_cache) { |
| 5471 | kmem_cache_destroy(rbd_obj_request_cache); |
| 5472 | rbd_obj_request_cache = NULL; |
| 5473 | } |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5474 | |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5475 | kmem_cache_destroy(rbd_img_request_cache); |
| 5476 | rbd_img_request_cache = NULL; |
| 5477 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5478 | return -ENOMEM; |
| 5479 | } |
| 5480 | |
| 5481 | static void rbd_slab_exit(void) |
| 5482 | { |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 5483 | rbd_assert(rbd_segment_name_cache); |
| 5484 | kmem_cache_destroy(rbd_segment_name_cache); |
| 5485 | rbd_segment_name_cache = NULL; |
| 5486 | |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5487 | rbd_assert(rbd_obj_request_cache); |
| 5488 | kmem_cache_destroy(rbd_obj_request_cache); |
| 5489 | rbd_obj_request_cache = NULL; |
| 5490 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5491 | rbd_assert(rbd_img_request_cache); |
| 5492 | kmem_cache_destroy(rbd_img_request_cache); |
| 5493 | rbd_img_request_cache = NULL; |
| 5494 | } |
| 5495 | |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 5496 | static int __init rbd_init(void) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5497 | { |
| 5498 | int rc; |
| 5499 | |
Alex Elder | 1e32d34 | 2013-01-30 11:13:33 -0600 | [diff] [blame] | 5500 | if (!libceph_compatible(NULL)) { |
| 5501 | rbd_warn(NULL, "libceph incompatibility (quitting)"); |
Alex Elder | 1e32d34 | 2013-01-30 11:13:33 -0600 | [diff] [blame] | 5502 | return -EINVAL; |
| 5503 | } |
Ilya Dryomov | e1b4d96 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5504 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5505 | rc = rbd_slab_init(); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5506 | if (rc) |
| 5507 | return rc; |
Ilya Dryomov | e1b4d96 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5508 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5509 | if (single_major) { |
| 5510 | rbd_major = register_blkdev(0, RBD_DRV_NAME); |
| 5511 | if (rbd_major < 0) { |
| 5512 | rc = rbd_major; |
| 5513 | goto err_out_slab; |
| 5514 | } |
| 5515 | } |
| 5516 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5517 | rc = rbd_sysfs_init(); |
| 5518 | if (rc) |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5519 | goto err_out_blkdev; |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5520 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5521 | if (single_major) |
| 5522 | pr_info("loaded (major %d)\n", rbd_major); |
| 5523 | else |
| 5524 | pr_info("loaded\n"); |
| 5525 | |
Ilya Dryomov | e1b4d96 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5526 | return 0; |
| 5527 | |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5528 | err_out_blkdev: |
| 5529 | if (single_major) |
| 5530 | unregister_blkdev(rbd_major, RBD_DRV_NAME); |
Ilya Dryomov | e1b4d96 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5531 | err_out_slab: |
| 5532 | rbd_slab_exit(); |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5533 | return rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5534 | } |
| 5535 | |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 5536 | static void __exit rbd_exit(void) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5537 | { |
Ilya Dryomov | ffe312c | 2014-05-20 15:46:04 +0400 | [diff] [blame] | 5538 | ida_destroy(&rbd_dev_id_ida); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5539 | rbd_sysfs_cleanup(); |
Ilya Dryomov | 9b60e70 | 2013-12-13 15:28:57 +0200 | [diff] [blame] | 5540 | if (single_major) |
| 5541 | unregister_blkdev(rbd_major, RBD_DRV_NAME); |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5542 | rbd_slab_exit(); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5543 | } |
| 5544 | |
| 5545 | module_init(rbd_init); |
| 5546 | module_exit(rbd_exit); |
| 5547 | |
Alex Elder | d552c61 | 2013-05-31 20:13:09 -0500 | [diff] [blame] | 5548 | MODULE_AUTHOR("Alex Elder <elder@inktank.com>"); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5549 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); |
| 5550 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5551 | /* following authorship retained from original osdblk.c */ |
| 5552 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); |
| 5553 | |
Ilya Dryomov | 90da258 | 2013-12-13 15:28:56 +0200 | [diff] [blame] | 5554 | MODULE_DESCRIPTION("RADOS Block Device (RBD) driver"); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5555 | MODULE_LICENSE("GPL"); |