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> |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 44 | |
| 45 | #include "rbd_types.h" |
| 46 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 47 | #define RBD_DEBUG /* Activate rbd_assert() calls */ |
| 48 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 49 | /* |
| 50 | * The basic unit of block I/O is a sector. It is interpreted in a |
| 51 | * number of contexts in Linux (blk, bio, genhd), but the default is |
| 52 | * universally 512 bytes. These symbols are just slightly more |
| 53 | * meaningful than the bare numbers they represent. |
| 54 | */ |
| 55 | #define SECTOR_SHIFT 9 |
| 56 | #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) |
| 57 | |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 58 | /* |
| 59 | * Increment the given counter and return its updated value. |
| 60 | * If the counter is already 0 it will not be incremented. |
| 61 | * If the counter is already at its maximum value returns |
| 62 | * -EINVAL without updating it. |
| 63 | */ |
| 64 | static int atomic_inc_return_safe(atomic_t *v) |
| 65 | { |
| 66 | unsigned int counter; |
| 67 | |
| 68 | counter = (unsigned int)__atomic_add_unless(v, 1, 0); |
| 69 | if (counter <= (unsigned int)INT_MAX) |
| 70 | return (int)counter; |
| 71 | |
| 72 | atomic_dec(v); |
| 73 | |
| 74 | return -EINVAL; |
| 75 | } |
| 76 | |
| 77 | /* Decrement the counter. Return the resulting value, or -EINVAL */ |
| 78 | static int atomic_dec_return_safe(atomic_t *v) |
| 79 | { |
| 80 | int counter; |
| 81 | |
| 82 | counter = atomic_dec_return(v); |
| 83 | if (counter >= 0) |
| 84 | return counter; |
| 85 | |
| 86 | atomic_inc(v); |
| 87 | |
| 88 | return -EINVAL; |
| 89 | } |
| 90 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 91 | #define RBD_DRV_NAME "rbd" |
| 92 | #define RBD_DRV_NAME_LONG "rbd (rados block device)" |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 93 | |
| 94 | #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ |
| 95 | |
Alex Elder | d4b125e | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 96 | #define RBD_SNAP_DEV_NAME_PREFIX "snap_" |
| 97 | #define RBD_MAX_SNAP_NAME_LEN \ |
| 98 | (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1)) |
| 99 | |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 100 | #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] | 101 | |
| 102 | #define RBD_SNAP_HEAD_NAME "-" |
| 103 | |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 104 | #define BAD_SNAP_INDEX U32_MAX /* invalid index into snap array */ |
| 105 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 106 | /* This allows a single page to hold an image name sent by OSD */ |
| 107 | #define RBD_IMAGE_NAME_LEN_MAX (PAGE_SIZE - sizeof (__le32) - 1) |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 108 | #define RBD_IMAGE_ID_LEN_MAX 64 |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 109 | |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 110 | #define RBD_OBJ_PREFIX_LEN_MAX 64 |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 111 | |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 112 | /* Feature bits */ |
| 113 | |
Alex Elder | 5cbf6f12 | 2013-04-11 09:29:48 -0500 | [diff] [blame] | 114 | #define RBD_FEATURE_LAYERING (1<<0) |
| 115 | #define RBD_FEATURE_STRIPINGV2 (1<<1) |
| 116 | #define RBD_FEATURES_ALL \ |
| 117 | (RBD_FEATURE_LAYERING | RBD_FEATURE_STRIPINGV2) |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 118 | |
| 119 | /* Features supported by this (client software) implementation. */ |
| 120 | |
Alex Elder | 770eba6 | 2012-10-25 23:34:40 -0500 | [diff] [blame] | 121 | #define RBD_FEATURES_SUPPORTED (RBD_FEATURES_ALL) |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 122 | |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 123 | /* |
| 124 | * An RBD device name will be "rbd#", where the "rbd" comes from |
| 125 | * RBD_DRV_NAME above, and # is a unique integer identifier. |
| 126 | * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big |
| 127 | * enough to hold all possible device names. |
| 128 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 129 | #define DEV_NAME_LEN 32 |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 130 | #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 131 | |
| 132 | /* |
| 133 | * block device image metadata (in-memory version) |
| 134 | */ |
| 135 | struct rbd_image_header { |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 136 | /* These six fields never change for a given rbd image */ |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 137 | char *object_prefix; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 138 | __u8 obj_order; |
| 139 | __u8 crypt_type; |
| 140 | __u8 comp_type; |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 141 | u64 stripe_unit; |
| 142 | u64 stripe_count; |
| 143 | u64 features; /* Might be changeable someday? */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 144 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 145 | /* The remaining fields need to be updated occasionally */ |
| 146 | u64 image_size; |
| 147 | struct ceph_snap_context *snapc; |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 148 | char *snap_names; /* format 1 only */ |
| 149 | u64 *snap_sizes; /* format 1 only */ |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 150 | }; |
| 151 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 152 | /* |
| 153 | * An rbd image specification. |
| 154 | * |
| 155 | * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely |
Alex Elder | c66c6e0 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 156 | * identify an image. Each rbd_dev structure includes a pointer to |
| 157 | * an rbd_spec structure that encapsulates this identity. |
| 158 | * |
| 159 | * Each of the id's in an rbd_spec has an associated name. For a |
| 160 | * user-mapped image, the names are supplied and the id's associated |
| 161 | * with them are looked up. For a layered image, a parent image is |
| 162 | * defined by the tuple, and the names are looked up. |
| 163 | * |
| 164 | * An rbd_dev structure contains a parent_spec pointer which is |
| 165 | * non-null if the image it represents is a child in a layered |
| 166 | * image. This pointer will refer to the rbd_spec structure used |
| 167 | * by the parent rbd_dev for its own identity (i.e., the structure |
| 168 | * is shared between the parent and child). |
| 169 | * |
| 170 | * Since these structures are populated once, during the discovery |
| 171 | * phase of image construction, they are effectively immutable so |
| 172 | * we make no effort to synchronize access to them. |
| 173 | * |
| 174 | * Note that code herein does not assume the image name is known (it |
| 175 | * could be a null pointer). |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 176 | */ |
| 177 | struct rbd_spec { |
| 178 | u64 pool_id; |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 179 | const char *pool_name; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 180 | |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 181 | const char *image_id; |
| 182 | const char *image_name; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 183 | |
| 184 | u64 snap_id; |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 185 | const char *snap_name; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 186 | |
| 187 | struct kref kref; |
| 188 | }; |
| 189 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 190 | /* |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 191 | * an instance of the client. multiple devices may share an rbd client. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 192 | */ |
| 193 | struct rbd_client { |
| 194 | struct ceph_client *client; |
| 195 | struct kref kref; |
| 196 | struct list_head node; |
| 197 | }; |
| 198 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 199 | struct rbd_img_request; |
| 200 | typedef void (*rbd_img_callback_t)(struct rbd_img_request *); |
| 201 | |
| 202 | #define BAD_WHICH U32_MAX /* Good which or bad which, which? */ |
| 203 | |
| 204 | struct rbd_obj_request; |
| 205 | typedef void (*rbd_obj_callback_t)(struct rbd_obj_request *); |
| 206 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 207 | enum obj_request_type { |
| 208 | OBJ_REQUEST_NODATA, OBJ_REQUEST_BIO, OBJ_REQUEST_PAGES |
| 209 | }; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 210 | |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 211 | enum obj_req_flags { |
| 212 | OBJ_REQ_DONE, /* completion flag: not done = 0, done = 1 */ |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 213 | OBJ_REQ_IMG_DATA, /* object usage: standalone = 0, image = 1 */ |
Alex Elder | 5679c59 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 214 | OBJ_REQ_KNOWN, /* EXISTS flag valid: no = 0, yes = 1 */ |
| 215 | OBJ_REQ_EXISTS, /* target exists: no = 0, yes = 1 */ |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 216 | }; |
| 217 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 218 | struct rbd_obj_request { |
| 219 | const char *object_name; |
| 220 | u64 offset; /* object start byte */ |
| 221 | u64 length; /* bytes from offset */ |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 222 | unsigned long flags; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 223 | |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 224 | /* |
| 225 | * An object request associated with an image will have its |
| 226 | * img_data flag set; a standalone object request will not. |
| 227 | * |
| 228 | * A standalone object request will have which == BAD_WHICH |
| 229 | * and a null obj_request pointer. |
| 230 | * |
| 231 | * An object request initiated in support of a layered image |
| 232 | * object (to check for its existence before a write) will |
| 233 | * have which == BAD_WHICH and a non-null obj_request pointer. |
| 234 | * |
| 235 | * Finally, an object request for rbd image data will have |
| 236 | * which != BAD_WHICH, and will have a non-null img_request |
| 237 | * pointer. The value of which will be in the range |
| 238 | * 0..(img_request->obj_request_count-1). |
| 239 | */ |
| 240 | union { |
| 241 | struct rbd_obj_request *obj_request; /* STAT op */ |
| 242 | struct { |
| 243 | struct rbd_img_request *img_request; |
| 244 | u64 img_offset; |
| 245 | /* links for img_request->obj_requests list */ |
| 246 | struct list_head links; |
| 247 | }; |
| 248 | }; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 249 | u32 which; /* posn image request list */ |
| 250 | |
| 251 | enum obj_request_type type; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 252 | union { |
| 253 | struct bio *bio_list; |
| 254 | struct { |
| 255 | struct page **pages; |
| 256 | u32 page_count; |
| 257 | }; |
| 258 | }; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 259 | struct page **copyup_pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 260 | u32 copyup_page_count; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 261 | |
| 262 | struct ceph_osd_request *osd_req; |
| 263 | |
| 264 | u64 xferred; /* bytes transferred */ |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 265 | int result; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 266 | |
| 267 | rbd_obj_callback_t callback; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 268 | struct completion completion; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 269 | |
| 270 | struct kref kref; |
| 271 | }; |
| 272 | |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 273 | enum img_req_flags { |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 274 | IMG_REQ_WRITE, /* I/O direction: read = 0, write = 1 */ |
| 275 | IMG_REQ_CHILD, /* initiator: block = 0, child image = 1 */ |
Alex Elder | d0b2e94 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 276 | IMG_REQ_LAYERED, /* ENOENT handling: normal = 0, layered = 1 */ |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 277 | }; |
| 278 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 279 | struct rbd_img_request { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 280 | struct rbd_device *rbd_dev; |
| 281 | u64 offset; /* starting image byte offset */ |
| 282 | u64 length; /* byte count from offset */ |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 283 | unsigned long flags; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 284 | union { |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 285 | u64 snap_id; /* for reads */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 286 | struct ceph_snap_context *snapc; /* for writes */ |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 287 | }; |
| 288 | union { |
| 289 | struct request *rq; /* block request */ |
| 290 | struct rbd_obj_request *obj_request; /* obj req initiator */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 291 | }; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 292 | struct page **copyup_pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 293 | u32 copyup_page_count; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 294 | spinlock_t completion_lock;/* protects next_completion */ |
| 295 | u32 next_completion; |
| 296 | rbd_img_callback_t callback; |
Alex Elder | 55f27e0 | 2013-04-10 12:34:25 -0500 | [diff] [blame] | 297 | u64 xferred;/* aggregate bytes transferred */ |
Alex Elder | a5a337d | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 298 | int result; /* first nonzero obj_request result */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 299 | |
| 300 | u32 obj_request_count; |
| 301 | struct list_head obj_requests; /* rbd_obj_request structs */ |
| 302 | |
| 303 | struct kref kref; |
| 304 | }; |
| 305 | |
| 306 | #define for_each_obj_request(ireq, oreq) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 307 | list_for_each_entry(oreq, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 308 | #define for_each_obj_request_from(ireq, oreq) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 309 | list_for_each_entry_from(oreq, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 310 | #define for_each_obj_request_safe(ireq, oreq, n) \ |
Alex Elder | ef06f4d3 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 311 | list_for_each_entry_safe_reverse(oreq, n, &(ireq)->obj_requests, links) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 312 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 313 | struct rbd_mapping { |
Alex Elder | 99c1f08 | 2012-08-30 14:42:15 -0500 | [diff] [blame] | 314 | u64 size; |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 315 | u64 features; |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 316 | bool read_only; |
| 317 | }; |
| 318 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 319 | /* |
| 320 | * a single device |
| 321 | */ |
| 322 | struct rbd_device { |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 323 | int dev_id; /* blkdev unique id */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 324 | |
| 325 | int major; /* blkdev assigned major */ |
| 326 | struct gendisk *disk; /* blkdev's gendisk and rq */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 327 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 328 | u32 image_format; /* Either 1 or 2 */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 329 | struct rbd_client *rbd_client; |
| 330 | |
| 331 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ |
| 332 | |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 333 | spinlock_t lock; /* queue, flags, open_count */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 334 | |
| 335 | struct rbd_image_header header; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 336 | unsigned long flags; /* possibly lock protected */ |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 337 | struct rbd_spec *spec; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 338 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 339 | char *header_name; |
Alex Elder | 971f839 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 340 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 341 | struct ceph_file_layout layout; |
| 342 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 343 | struct ceph_osd_event *watch_event; |
Alex Elder | 975241a | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 344 | struct rbd_obj_request *watch_request; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 345 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 346 | struct rbd_spec *parent_spec; |
| 347 | u64 parent_overlap; |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 348 | atomic_t parent_ref; |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 349 | struct rbd_device *parent; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 350 | |
Josh Durgin | c666601 | 2011-11-21 17:11:12 -0800 | [diff] [blame] | 351 | /* protects updating the header */ |
| 352 | struct rw_semaphore header_rwsem; |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 353 | |
| 354 | struct rbd_mapping mapping; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 355 | |
| 356 | struct list_head node; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 357 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 358 | /* sysfs related */ |
| 359 | struct device dev; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 360 | unsigned long open_count; /* protected by lock */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 361 | }; |
| 362 | |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 363 | /* |
| 364 | * Flag bits for rbd_dev->flags. If atomicity is required, |
| 365 | * rbd_dev->lock is used to protect access. |
| 366 | * |
| 367 | * Currently, only the "removing" flag (which is coupled with the |
| 368 | * "open_count" field) requires atomic access. |
| 369 | */ |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 370 | enum rbd_dev_flags { |
| 371 | RBD_DEV_FLAG_EXISTS, /* mapped snapshot has not been deleted */ |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 372 | RBD_DEV_FLAG_REMOVING, /* this mapping is being removed */ |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 373 | }; |
| 374 | |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 375 | static DEFINE_MUTEX(client_mutex); /* Serialize client creation */ |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 376 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 377 | static LIST_HEAD(rbd_dev_list); /* devices */ |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 378 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
| 379 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 380 | static LIST_HEAD(rbd_client_list); /* clients */ |
| 381 | static DEFINE_SPINLOCK(rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 382 | |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 383 | /* Slab caches for frequently-allocated structures */ |
| 384 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 385 | static struct kmem_cache *rbd_img_request_cache; |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 386 | static struct kmem_cache *rbd_obj_request_cache; |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 387 | static struct kmem_cache *rbd_segment_name_cache; |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 388 | |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 389 | static int rbd_img_request_submit(struct rbd_img_request *img_request); |
| 390 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 391 | static void rbd_dev_device_release(struct device *dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 392 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 393 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
| 394 | size_t count); |
| 395 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, |
| 396 | size_t count); |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 397 | 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] | 398 | static void rbd_spec_put(struct rbd_spec *spec); |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 399 | |
| 400 | static struct bus_attribute rbd_bus_attrs[] = { |
| 401 | __ATTR(add, S_IWUSR, NULL, rbd_add), |
| 402 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), |
| 403 | __ATTR_NULL |
| 404 | }; |
| 405 | |
| 406 | static struct bus_type rbd_bus_type = { |
| 407 | .name = "rbd", |
| 408 | .bus_attrs = rbd_bus_attrs, |
| 409 | }; |
| 410 | |
| 411 | static void rbd_root_dev_release(struct device *dev) |
| 412 | { |
| 413 | } |
| 414 | |
| 415 | static struct device rbd_root_dev = { |
| 416 | .init_name = "rbd", |
| 417 | .release = rbd_root_dev_release, |
| 418 | }; |
| 419 | |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 420 | static __printf(2, 3) |
| 421 | void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...) |
| 422 | { |
| 423 | struct va_format vaf; |
| 424 | va_list args; |
| 425 | |
| 426 | va_start(args, fmt); |
| 427 | vaf.fmt = fmt; |
| 428 | vaf.va = &args; |
| 429 | |
| 430 | if (!rbd_dev) |
| 431 | printk(KERN_WARNING "%s: %pV\n", RBD_DRV_NAME, &vaf); |
| 432 | else if (rbd_dev->disk) |
| 433 | printk(KERN_WARNING "%s: %s: %pV\n", |
| 434 | RBD_DRV_NAME, rbd_dev->disk->disk_name, &vaf); |
| 435 | else if (rbd_dev->spec && rbd_dev->spec->image_name) |
| 436 | printk(KERN_WARNING "%s: image %s: %pV\n", |
| 437 | RBD_DRV_NAME, rbd_dev->spec->image_name, &vaf); |
| 438 | else if (rbd_dev->spec && rbd_dev->spec->image_id) |
| 439 | printk(KERN_WARNING "%s: id %s: %pV\n", |
| 440 | RBD_DRV_NAME, rbd_dev->spec->image_id, &vaf); |
| 441 | else /* punt */ |
| 442 | printk(KERN_WARNING "%s: rbd_dev %p: %pV\n", |
| 443 | RBD_DRV_NAME, rbd_dev, &vaf); |
| 444 | va_end(args); |
| 445 | } |
| 446 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 447 | #ifdef RBD_DEBUG |
| 448 | #define rbd_assert(expr) \ |
| 449 | if (unlikely(!(expr))) { \ |
| 450 | printk(KERN_ERR "\nAssertion failure in %s() " \ |
| 451 | "at line %d:\n\n" \ |
| 452 | "\trbd_assert(%s);\n\n", \ |
| 453 | __func__, __LINE__, #expr); \ |
| 454 | BUG(); \ |
| 455 | } |
| 456 | #else /* !RBD_DEBUG */ |
| 457 | # define rbd_assert(expr) ((void) 0) |
| 458 | #endif /* !RBD_DEBUG */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 459 | |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 460 | 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] | 461 | static void rbd_img_parent_read(struct rbd_obj_request *obj_request); |
| 462 | static void rbd_dev_remove_parent(struct rbd_device *rbd_dev); |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 463 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 464 | static int rbd_dev_refresh(struct rbd_device *rbd_dev); |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 465 | static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev); |
| 466 | static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev); |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 467 | static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, |
| 468 | u64 snap_id); |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 469 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
| 470 | u8 *order, u64 *snap_size); |
| 471 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
| 472 | u64 *snap_features); |
| 473 | 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] | 474 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 475 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
| 476 | { |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 477 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 478 | bool removing = false; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 479 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 480 | if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 481 | return -EROFS; |
| 482 | |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 483 | spin_lock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 484 | if (test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags)) |
| 485 | removing = true; |
| 486 | else |
| 487 | rbd_dev->open_count++; |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 488 | spin_unlock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 489 | if (removing) |
| 490 | return -ENOENT; |
| 491 | |
Alex Elder | c3e946c | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 492 | (void) get_device(&rbd_dev->dev); |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 493 | set_device_ro(bdev, rbd_dev->mapping.read_only); |
Alex Elder | 340c7a2 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 494 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 495 | return 0; |
| 496 | } |
| 497 | |
Al Viro | db2a144 | 2013-05-05 21:52:57 -0400 | [diff] [blame] | 498 | static void rbd_release(struct gendisk *disk, fmode_t mode) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 499 | { |
| 500 | struct rbd_device *rbd_dev = disk->private_data; |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 501 | unsigned long open_count_before; |
| 502 | |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 503 | spin_lock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 504 | open_count_before = rbd_dev->open_count--; |
Alex Elder | a14ea26 | 2013-02-05 13:23:12 -0600 | [diff] [blame] | 505 | spin_unlock_irq(&rbd_dev->lock); |
Alex Elder | b82d167 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 506 | rbd_assert(open_count_before > 0); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 507 | |
Alex Elder | c3e946c | 2012-11-16 09:29:16 -0600 | [diff] [blame] | 508 | put_device(&rbd_dev->dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 509 | } |
| 510 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 511 | static const struct block_device_operations rbd_bd_ops = { |
| 512 | .owner = THIS_MODULE, |
| 513 | .open = rbd_open, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 514 | .release = rbd_release, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 515 | }; |
| 516 | |
| 517 | /* |
Alex Elder | 7262cfc | 2013-05-16 15:04:20 -0500 | [diff] [blame] | 518 | * Initialize an rbd client instance. Success or not, this function |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 519 | * consumes ceph_opts. Caller holds client_mutex. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 520 | */ |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 521 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 522 | { |
| 523 | struct rbd_client *rbdc; |
| 524 | int ret = -ENOMEM; |
| 525 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 526 | dout("%s:\n", __func__); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 527 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); |
| 528 | if (!rbdc) |
| 529 | goto out_opt; |
| 530 | |
| 531 | kref_init(&rbdc->kref); |
| 532 | INIT_LIST_HEAD(&rbdc->node); |
| 533 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 534 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 535 | if (IS_ERR(rbdc->client)) |
Alex Elder | 08f7546 | 2013-05-29 11:19:00 -0500 | [diff] [blame] | 536 | goto out_rbdc; |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 537 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 538 | |
| 539 | ret = ceph_open_session(rbdc->client); |
| 540 | if (ret < 0) |
Alex Elder | 08f7546 | 2013-05-29 11:19:00 -0500 | [diff] [blame] | 541 | goto out_client; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 542 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 543 | spin_lock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 544 | list_add_tail(&rbdc->node, &rbd_client_list); |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 545 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 546 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 547 | dout("%s: rbdc %p\n", __func__, rbdc); |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 548 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 549 | return rbdc; |
Alex Elder | 08f7546 | 2013-05-29 11:19:00 -0500 | [diff] [blame] | 550 | out_client: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 551 | ceph_destroy_client(rbdc->client); |
Alex Elder | 08f7546 | 2013-05-29 11:19:00 -0500 | [diff] [blame] | 552 | out_rbdc: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 553 | kfree(rbdc); |
| 554 | out_opt: |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 555 | if (ceph_opts) |
| 556 | ceph_destroy_options(ceph_opts); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 557 | dout("%s: error %d\n", __func__, ret); |
| 558 | |
Vasiliy Kulikov | 28f259b | 2010-09-26 12:59:37 +0400 | [diff] [blame] | 559 | return ERR_PTR(ret); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 560 | } |
| 561 | |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 562 | static struct rbd_client *__rbd_get_client(struct rbd_client *rbdc) |
| 563 | { |
| 564 | kref_get(&rbdc->kref); |
| 565 | |
| 566 | return rbdc; |
| 567 | } |
| 568 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 569 | /* |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 570 | * Find a ceph client with specific addr and configuration. If |
| 571 | * found, bump its reference count. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 572 | */ |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 573 | static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 574 | { |
| 575 | struct rbd_client *client_node; |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 576 | bool found = false; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 577 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 578 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 579 | return NULL; |
| 580 | |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 581 | spin_lock(&rbd_client_list_lock); |
| 582 | list_for_each_entry(client_node, &rbd_client_list, node) { |
| 583 | if (!ceph_compare_options(ceph_opts, client_node->client)) { |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 584 | __rbd_get_client(client_node); |
| 585 | |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 586 | found = true; |
| 587 | break; |
| 588 | } |
| 589 | } |
| 590 | spin_unlock(&rbd_client_list_lock); |
| 591 | |
| 592 | return found ? client_node : NULL; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 593 | } |
| 594 | |
| 595 | /* |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 596 | * mount options |
| 597 | */ |
| 598 | enum { |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 599 | Opt_last_int, |
| 600 | /* int args above */ |
| 601 | Opt_last_string, |
| 602 | /* string args above */ |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 603 | Opt_read_only, |
| 604 | Opt_read_write, |
| 605 | /* Boolean args above */ |
| 606 | Opt_last_bool, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 607 | }; |
| 608 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 609 | static match_table_t rbd_opts_tokens = { |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 610 | /* int args above */ |
| 611 | /* string args above */ |
Alex Elder | be466c1 | 2012-10-22 11:31:26 -0500 | [diff] [blame] | 612 | {Opt_read_only, "read_only"}, |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 613 | {Opt_read_only, "ro"}, /* Alternate spelling */ |
| 614 | {Opt_read_write, "read_write"}, |
| 615 | {Opt_read_write, "rw"}, /* Alternate spelling */ |
| 616 | /* Boolean args above */ |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 617 | {-1, NULL} |
| 618 | }; |
| 619 | |
Alex Elder | 98571b5 | 2013-01-20 14:44:42 -0600 | [diff] [blame] | 620 | struct rbd_options { |
| 621 | bool read_only; |
| 622 | }; |
| 623 | |
| 624 | #define RBD_READ_ONLY_DEFAULT false |
| 625 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 626 | static int parse_rbd_opts_token(char *c, void *private) |
| 627 | { |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 628 | struct rbd_options *rbd_opts = private; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 629 | substring_t argstr[MAX_OPT_ARGS]; |
| 630 | int token, intval, ret; |
| 631 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 632 | token = match_token(c, rbd_opts_tokens, argstr); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 633 | if (token < 0) |
| 634 | return -EINVAL; |
| 635 | |
| 636 | if (token < Opt_last_int) { |
| 637 | ret = match_int(&argstr[0], &intval); |
| 638 | if (ret < 0) { |
| 639 | pr_err("bad mount option arg (not int) " |
| 640 | "at '%s'\n", c); |
| 641 | return ret; |
| 642 | } |
| 643 | dout("got int token %d val %d\n", token, intval); |
| 644 | } else if (token > Opt_last_int && token < Opt_last_string) { |
| 645 | dout("got string token %d val %s\n", token, |
| 646 | argstr[0].from); |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 647 | } else if (token > Opt_last_string && token < Opt_last_bool) { |
| 648 | dout("got Boolean token %d\n", token); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 649 | } else { |
| 650 | dout("got token %d\n", token); |
| 651 | } |
| 652 | |
| 653 | switch (token) { |
Alex Elder | cc0538b | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 654 | case Opt_read_only: |
| 655 | rbd_opts->read_only = true; |
| 656 | break; |
| 657 | case Opt_read_write: |
| 658 | rbd_opts->read_only = false; |
| 659 | break; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 660 | default: |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 661 | rbd_assert(false); |
| 662 | break; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 663 | } |
| 664 | return 0; |
| 665 | } |
| 666 | |
| 667 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 668 | * Get a ceph client with specific addr and configuration, if one does |
Alex Elder | 7262cfc | 2013-05-16 15:04:20 -0500 | [diff] [blame] | 669 | * not exist create it. Either way, ceph_opts is consumed by this |
| 670 | * function. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 671 | */ |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 672 | static struct rbd_client *rbd_get_client(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 673 | { |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 674 | struct rbd_client *rbdc; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 675 | |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 676 | mutex_lock_nested(&client_mutex, SINGLE_DEPTH_NESTING); |
Alex Elder | 1f7ba33 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 677 | rbdc = rbd_client_find(ceph_opts); |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 678 | if (rbdc) /* using an existing client */ |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 679 | ceph_destroy_options(ceph_opts); |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 680 | else |
Alex Elder | f8c3892 | 2012-08-10 13:12:07 -0700 | [diff] [blame] | 681 | rbdc = rbd_client_create(ceph_opts); |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 682 | mutex_unlock(&client_mutex); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 683 | |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 684 | return rbdc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 685 | } |
| 686 | |
| 687 | /* |
| 688 | * Destroy ceph client |
Alex Elder | d23a4b3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 689 | * |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 690 | * Caller must hold rbd_client_list_lock. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 691 | */ |
| 692 | static void rbd_client_release(struct kref *kref) |
| 693 | { |
| 694 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); |
| 695 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 696 | dout("%s: rbdc %p\n", __func__, rbdc); |
Alex Elder | cd9d9f5 | 2012-04-04 13:35:44 -0500 | [diff] [blame] | 697 | spin_lock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 698 | list_del(&rbdc->node); |
Alex Elder | cd9d9f5 | 2012-04-04 13:35:44 -0500 | [diff] [blame] | 699 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 700 | |
| 701 | ceph_destroy_client(rbdc->client); |
| 702 | kfree(rbdc); |
| 703 | } |
| 704 | |
| 705 | /* |
| 706 | * Drop reference to ceph client node. If it's not referenced anymore, release |
| 707 | * it. |
| 708 | */ |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 709 | static void rbd_put_client(struct rbd_client *rbdc) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 710 | { |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 711 | if (rbdc) |
| 712 | kref_put(&rbdc->kref, rbd_client_release); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 713 | } |
| 714 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 715 | static bool rbd_image_format_valid(u32 image_format) |
| 716 | { |
| 717 | return image_format == 1 || image_format == 2; |
| 718 | } |
| 719 | |
Alex Elder | 8e94af8 | 2012-07-25 09:32:40 -0500 | [diff] [blame] | 720 | static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) |
| 721 | { |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 722 | size_t size; |
| 723 | u32 snap_count; |
| 724 | |
| 725 | /* The header has to start with the magic rbd header text */ |
| 726 | if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) |
| 727 | return false; |
| 728 | |
Alex Elder | db2388b | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 729 | /* The bio layer requires at least sector-sized I/O */ |
| 730 | |
| 731 | if (ondisk->options.order < SECTOR_SHIFT) |
| 732 | return false; |
| 733 | |
| 734 | /* If we use u64 in a few spots we may be able to loosen this */ |
| 735 | |
| 736 | if (ondisk->options.order > 8 * sizeof (int) - 1) |
| 737 | return false; |
| 738 | |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 739 | /* |
| 740 | * The size of a snapshot header has to fit in a size_t, and |
| 741 | * that limits the number of snapshots. |
| 742 | */ |
| 743 | snap_count = le32_to_cpu(ondisk->snap_count); |
| 744 | size = SIZE_MAX - sizeof (struct ceph_snap_context); |
| 745 | if (snap_count > size / sizeof (__le64)) |
| 746 | return false; |
| 747 | |
| 748 | /* |
| 749 | * Not only that, but the size of the entire the snapshot |
| 750 | * header must also be representable in a size_t. |
| 751 | */ |
| 752 | size -= snap_count * sizeof (__le64); |
| 753 | if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) |
| 754 | return false; |
| 755 | |
| 756 | return true; |
Alex Elder | 8e94af8 | 2012-07-25 09:32:40 -0500 | [diff] [blame] | 757 | } |
| 758 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 759 | /* |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 760 | * Fill an rbd image header with information from the given format 1 |
| 761 | * on-disk header. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 762 | */ |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 763 | static int rbd_header_from_disk(struct rbd_device *rbd_dev, |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 764 | struct rbd_image_header_ondisk *ondisk) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 765 | { |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 766 | struct rbd_image_header *header = &rbd_dev->header; |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 767 | bool first_time = header->object_prefix == NULL; |
| 768 | struct ceph_snap_context *snapc; |
| 769 | char *object_prefix = NULL; |
| 770 | char *snap_names = NULL; |
| 771 | u64 *snap_sizes = NULL; |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 772 | u32 snap_count; |
Alex Elder | d2bb24e | 2012-07-26 23:37:14 -0500 | [diff] [blame] | 773 | size_t size; |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 774 | int ret = -ENOMEM; |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 775 | u32 i; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 776 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 777 | /* Allocate this now to avoid having to handle failure below */ |
| 778 | |
| 779 | if (first_time) { |
| 780 | size_t len; |
| 781 | |
| 782 | len = strnlen(ondisk->object_prefix, |
| 783 | sizeof (ondisk->object_prefix)); |
| 784 | object_prefix = kmalloc(len + 1, GFP_KERNEL); |
| 785 | if (!object_prefix) |
| 786 | return -ENOMEM; |
| 787 | memcpy(object_prefix, ondisk->object_prefix, len); |
| 788 | object_prefix[len] = '\0'; |
| 789 | } |
| 790 | |
| 791 | /* Allocate the snapshot context and fill it in */ |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 792 | |
Alex Elder | 103a150 | 2012-08-02 11:29:45 -0500 | [diff] [blame] | 793 | snap_count = le32_to_cpu(ondisk->snap_count); |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 794 | snapc = ceph_create_snap_context(snap_count, GFP_KERNEL); |
| 795 | if (!snapc) |
| 796 | goto out_err; |
| 797 | snapc->seq = le64_to_cpu(ondisk->snap_seq); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 798 | if (snap_count) { |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 799 | struct rbd_image_snap_ondisk *snaps; |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 800 | u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
| 801 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 802 | /* We'll keep a copy of the snapshot names... */ |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 803 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 804 | if (snap_names_len > (u64)SIZE_MAX) |
| 805 | goto out_2big; |
| 806 | snap_names = kmalloc(snap_names_len, GFP_KERNEL); |
| 807 | if (!snap_names) |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 808 | goto out_err; |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 809 | |
| 810 | /* ...as well as the array of their sizes. */ |
| 811 | |
| 812 | size = snap_count * sizeof (*header->snap_sizes); |
| 813 | snap_sizes = kmalloc(size, GFP_KERNEL); |
| 814 | if (!snap_sizes) |
| 815 | goto out_err; |
| 816 | |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 817 | /* |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 818 | * Copy the names, and fill in each snapshot's id |
| 819 | * and size. |
| 820 | * |
Alex Elder | 99a41eb | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 821 | * Note that rbd_dev_v1_header_info() guarantees the |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 822 | * ondisk buffer we're working with has |
Alex Elder | f785cc1 | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 823 | * snap_names_len bytes beyond the end of the |
| 824 | * snapshot id array, this memcpy() is safe. |
| 825 | */ |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 826 | memcpy(snap_names, &ondisk->snaps[snap_count], snap_names_len); |
| 827 | snaps = ondisk->snaps; |
| 828 | for (i = 0; i < snap_count; i++) { |
| 829 | snapc->snaps[i] = le64_to_cpu(snaps[i].id); |
| 830 | snap_sizes[i] = le64_to_cpu(snaps[i].image_size); |
| 831 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 832 | } |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 833 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 834 | /* We won't fail any more, fill in the header */ |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 835 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 836 | if (first_time) { |
| 837 | header->object_prefix = object_prefix; |
| 838 | header->obj_order = ondisk->options.order; |
| 839 | header->crypt_type = ondisk->options.crypt_type; |
| 840 | header->comp_type = ondisk->options.comp_type; |
| 841 | /* The rest aren't used for format 1 images */ |
| 842 | header->stripe_unit = 0; |
| 843 | header->stripe_count = 0; |
| 844 | header->features = 0; |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 845 | } else { |
| 846 | ceph_put_snap_context(header->snapc); |
| 847 | kfree(header->snap_names); |
| 848 | kfree(header->snap_sizes); |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 849 | } |
| 850 | |
| 851 | /* The remaining fields always get updated (when we refresh) */ |
Alex Elder | 621901d | 2012-08-23 23:22:06 -0500 | [diff] [blame] | 852 | |
Alex Elder | f84344f | 2012-08-31 17:29:51 -0500 | [diff] [blame] | 853 | header->image_size = le64_to_cpu(ondisk->image_size); |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 854 | header->snapc = snapc; |
| 855 | header->snap_names = snap_names; |
| 856 | header->snap_sizes = snap_sizes; |
Alex Elder | 468521c | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 857 | |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 858 | /* Make sure mapping size is consistent with header info */ |
| 859 | |
| 860 | if (rbd_dev->spec->snap_id == CEPH_NOSNAP || first_time) |
| 861 | if (rbd_dev->mapping.size != header->image_size) |
| 862 | rbd_dev->mapping.size = header->image_size; |
| 863 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 864 | return 0; |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 865 | out_2big: |
| 866 | ret = -EIO; |
Alex Elder | 6a52325 | 2012-07-19 17:12:59 -0500 | [diff] [blame] | 867 | out_err: |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 868 | kfree(snap_sizes); |
| 869 | kfree(snap_names); |
| 870 | ceph_put_snap_context(snapc); |
| 871 | kfree(object_prefix); |
Alex Elder | ccece23 | 2012-07-10 20:30:10 -0500 | [diff] [blame] | 872 | |
Alex Elder | bb23e37 | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 873 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 874 | } |
| 875 | |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 876 | static const char *_rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, u32 which) |
| 877 | { |
| 878 | const char *snap_name; |
| 879 | |
| 880 | rbd_assert(which < rbd_dev->header.snapc->num_snaps); |
| 881 | |
| 882 | /* Skip over names until we find the one we are looking for */ |
| 883 | |
| 884 | snap_name = rbd_dev->header.snap_names; |
| 885 | while (which--) |
| 886 | snap_name += strlen(snap_name) + 1; |
| 887 | |
| 888 | return kstrdup(snap_name, GFP_KERNEL); |
| 889 | } |
| 890 | |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 891 | /* |
| 892 | * Snapshot id comparison function for use with qsort()/bsearch(). |
| 893 | * Note that result is for snapshots in *descending* order. |
| 894 | */ |
| 895 | static int snapid_compare_reverse(const void *s1, const void *s2) |
| 896 | { |
| 897 | u64 snap_id1 = *(u64 *)s1; |
| 898 | u64 snap_id2 = *(u64 *)s2; |
| 899 | |
| 900 | if (snap_id1 < snap_id2) |
| 901 | return 1; |
| 902 | return snap_id1 == snap_id2 ? 0 : -1; |
| 903 | } |
| 904 | |
| 905 | /* |
| 906 | * Search a snapshot context to see if the given snapshot id is |
| 907 | * present. |
| 908 | * |
| 909 | * Returns the position of the snapshot id in the array if it's found, |
| 910 | * or BAD_SNAP_INDEX otherwise. |
| 911 | * |
| 912 | * Note: The snapshot array is in kept sorted (by the osd) in |
| 913 | * reverse order, highest snapshot id first. |
| 914 | */ |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 915 | static u32 rbd_dev_snap_index(struct rbd_device *rbd_dev, u64 snap_id) |
| 916 | { |
| 917 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 918 | u64 *found; |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 919 | |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 920 | found = bsearch(&snap_id, &snapc->snaps, snapc->num_snaps, |
| 921 | sizeof (snap_id), snapid_compare_reverse); |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 922 | |
Alex Elder | 30d1cff | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 923 | return found ? (u32)(found - &snapc->snaps[0]) : BAD_SNAP_INDEX; |
Alex Elder | 9682fc6 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 924 | } |
| 925 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 926 | static const char *rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, |
| 927 | u64 snap_id) |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 928 | { |
| 929 | u32 which; |
| 930 | |
| 931 | which = rbd_dev_snap_index(rbd_dev, snap_id); |
| 932 | if (which == BAD_SNAP_INDEX) |
| 933 | return NULL; |
| 934 | |
| 935 | return _rbd_dev_v1_snap_name(rbd_dev, which); |
| 936 | } |
| 937 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 938 | static const char *rbd_snap_name(struct rbd_device *rbd_dev, u64 snap_id) |
| 939 | { |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 940 | if (snap_id == CEPH_NOSNAP) |
| 941 | return RBD_SNAP_HEAD_NAME; |
| 942 | |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 943 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 944 | if (rbd_dev->image_format == 1) |
| 945 | return rbd_dev_v1_snap_name(rbd_dev, snap_id); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 946 | |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 947 | return rbd_dev_v2_snap_name(rbd_dev, snap_id); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 948 | } |
| 949 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 950 | static int rbd_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
| 951 | u64 *snap_size) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 952 | { |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 953 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 954 | if (snap_id == CEPH_NOSNAP) { |
| 955 | *snap_size = rbd_dev->header.image_size; |
| 956 | } else if (rbd_dev->image_format == 1) { |
| 957 | u32 which; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 958 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 959 | which = rbd_dev_snap_index(rbd_dev, snap_id); |
| 960 | if (which == BAD_SNAP_INDEX) |
| 961 | return -ENOENT; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 962 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 963 | *snap_size = rbd_dev->header.snap_sizes[which]; |
| 964 | } else { |
| 965 | u64 size = 0; |
| 966 | int ret; |
| 967 | |
| 968 | ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, NULL, &size); |
| 969 | if (ret) |
| 970 | return ret; |
| 971 | |
| 972 | *snap_size = size; |
| 973 | } |
| 974 | return 0; |
| 975 | } |
| 976 | |
| 977 | static int rbd_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
| 978 | u64 *snap_features) |
| 979 | { |
| 980 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 981 | if (snap_id == CEPH_NOSNAP) { |
| 982 | *snap_features = rbd_dev->header.features; |
| 983 | } else if (rbd_dev->image_format == 1) { |
| 984 | *snap_features = 0; /* No features for format 1 */ |
| 985 | } else { |
| 986 | u64 features = 0; |
| 987 | int ret; |
| 988 | |
| 989 | ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, &features); |
| 990 | if (ret) |
| 991 | return ret; |
| 992 | |
| 993 | *snap_features = features; |
| 994 | } |
| 995 | return 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 996 | } |
| 997 | |
Alex Elder | d1cf578 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 998 | static int rbd_dev_mapping_set(struct rbd_device *rbd_dev) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 999 | { |
Alex Elder | 8f4b7d9 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 1000 | u64 snap_id = rbd_dev->spec->snap_id; |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1001 | u64 size = 0; |
| 1002 | u64 features = 0; |
| 1003 | int ret; |
Alex Elder | 8b0241f | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 1004 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 1005 | ret = rbd_snap_size(rbd_dev, snap_id, &size); |
| 1006 | if (ret) |
| 1007 | return ret; |
| 1008 | ret = rbd_snap_features(rbd_dev, snap_id, &features); |
| 1009 | if (ret) |
| 1010 | return ret; |
| 1011 | |
| 1012 | rbd_dev->mapping.size = size; |
| 1013 | rbd_dev->mapping.features = features; |
| 1014 | |
Alex Elder | 8b0241f | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 1015 | return 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1016 | } |
| 1017 | |
Alex Elder | d1cf578 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 1018 | static void rbd_dev_mapping_clear(struct rbd_device *rbd_dev) |
| 1019 | { |
| 1020 | rbd_dev->mapping.size = 0; |
| 1021 | rbd_dev->mapping.features = 0; |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 1022 | } |
| 1023 | |
Alex Elder | 98571b5 | 2013-01-20 14:44:42 -0600 | [diff] [blame] | 1024 | 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] | 1025 | { |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1026 | char *name; |
| 1027 | u64 segment; |
| 1028 | int ret; |
Josh Durgin | 3a96d5c | 2013-06-12 19:15:06 -0700 | [diff] [blame] | 1029 | char *name_format; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1030 | |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 1031 | name = kmem_cache_alloc(rbd_segment_name_cache, GFP_NOIO); |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1032 | if (!name) |
| 1033 | return NULL; |
| 1034 | segment = offset >> rbd_dev->header.obj_order; |
Josh Durgin | 3a96d5c | 2013-06-12 19:15:06 -0700 | [diff] [blame] | 1035 | name_format = "%s.%012llx"; |
| 1036 | if (rbd_dev->image_format == 2) |
| 1037 | name_format = "%s.%016llx"; |
| 1038 | ret = snprintf(name, MAX_OBJ_NAME_SIZE + 1, name_format, |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1039 | rbd_dev->header.object_prefix, segment); |
Alex Elder | 2fd82b9 | 2012-11-09 15:05:54 -0600 | [diff] [blame] | 1040 | if (ret < 0 || ret > MAX_OBJ_NAME_SIZE) { |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1041 | pr_err("error formatting segment name for #%llu (%d)\n", |
| 1042 | segment, ret); |
| 1043 | kfree(name); |
| 1044 | name = NULL; |
| 1045 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1046 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1047 | return name; |
| 1048 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1049 | |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 1050 | static void rbd_segment_name_free(const char *name) |
| 1051 | { |
| 1052 | /* The explicit cast here is needed to drop the const qualifier */ |
| 1053 | |
| 1054 | kmem_cache_free(rbd_segment_name_cache, (void *)name); |
| 1055 | } |
| 1056 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1057 | static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset) |
| 1058 | { |
| 1059 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1060 | |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1061 | return offset & (segment_size - 1); |
| 1062 | } |
| 1063 | |
| 1064 | static u64 rbd_segment_length(struct rbd_device *rbd_dev, |
| 1065 | u64 offset, u64 length) |
| 1066 | { |
| 1067 | u64 segment_size = (u64) 1 << rbd_dev->header.obj_order; |
| 1068 | |
| 1069 | offset &= segment_size - 1; |
| 1070 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 1071 | rbd_assert(length <= U64_MAX - offset); |
Alex Elder | 65ccfe2 | 2012-08-09 10:33:26 -0700 | [diff] [blame] | 1072 | if (offset + length > segment_size) |
| 1073 | length = segment_size - offset; |
| 1074 | |
| 1075 | return length; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1076 | } |
| 1077 | |
| 1078 | /* |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 1079 | * returns the size of an object in the image |
| 1080 | */ |
| 1081 | static u64 rbd_obj_bytes(struct rbd_image_header *header) |
| 1082 | { |
| 1083 | return 1 << header->obj_order; |
| 1084 | } |
| 1085 | |
| 1086 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1087 | * bio helpers |
| 1088 | */ |
| 1089 | |
| 1090 | static void bio_chain_put(struct bio *chain) |
| 1091 | { |
| 1092 | struct bio *tmp; |
| 1093 | |
| 1094 | while (chain) { |
| 1095 | tmp = chain; |
| 1096 | chain = chain->bi_next; |
| 1097 | bio_put(tmp); |
| 1098 | } |
| 1099 | } |
| 1100 | |
| 1101 | /* |
| 1102 | * zeros a bio chain, starting at specific offset |
| 1103 | */ |
| 1104 | static void zero_bio_chain(struct bio *chain, int start_ofs) |
| 1105 | { |
| 1106 | struct bio_vec *bv; |
| 1107 | unsigned long flags; |
| 1108 | void *buf; |
| 1109 | int i; |
| 1110 | int pos = 0; |
| 1111 | |
| 1112 | while (chain) { |
| 1113 | bio_for_each_segment(bv, chain, i) { |
| 1114 | if (pos + bv->bv_len > start_ofs) { |
| 1115 | int remainder = max(start_ofs - pos, 0); |
| 1116 | buf = bvec_kmap_irq(bv, &flags); |
| 1117 | memset(buf + remainder, 0, |
| 1118 | bv->bv_len - remainder); |
Alex Elder | e215605 | 2013-05-22 20:54:25 -0500 | [diff] [blame] | 1119 | flush_dcache_page(bv->bv_page); |
Dan Carpenter | 85b5aaa | 2010-10-11 21:15:11 +0200 | [diff] [blame] | 1120 | bvec_kunmap_irq(buf, &flags); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1121 | } |
| 1122 | pos += bv->bv_len; |
| 1123 | } |
| 1124 | |
| 1125 | chain = chain->bi_next; |
| 1126 | } |
| 1127 | } |
| 1128 | |
| 1129 | /* |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1130 | * similar to zero_bio_chain(), zeros data defined by a page array, |
| 1131 | * starting at the given byte offset from the start of the array and |
| 1132 | * continuing up to the given end offset. The pages array is |
| 1133 | * assumed to be big enough to hold all bytes up to the end. |
| 1134 | */ |
| 1135 | static void zero_pages(struct page **pages, u64 offset, u64 end) |
| 1136 | { |
| 1137 | struct page **page = &pages[offset >> PAGE_SHIFT]; |
| 1138 | |
| 1139 | rbd_assert(end > offset); |
| 1140 | rbd_assert(end - offset <= (u64)SIZE_MAX); |
| 1141 | while (offset < end) { |
| 1142 | size_t page_offset; |
| 1143 | size_t length; |
| 1144 | unsigned long flags; |
| 1145 | void *kaddr; |
| 1146 | |
Geert Uytterhoeven | 491205a | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 1147 | page_offset = offset & ~PAGE_MASK; |
| 1148 | length = min_t(size_t, PAGE_SIZE - page_offset, end - offset); |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1149 | local_irq_save(flags); |
| 1150 | kaddr = kmap_atomic(*page); |
| 1151 | memset(kaddr + page_offset, 0, length); |
Alex Elder | e215605 | 2013-05-22 20:54:25 -0500 | [diff] [blame] | 1152 | flush_dcache_page(*page); |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1153 | kunmap_atomic(kaddr); |
| 1154 | local_irq_restore(flags); |
| 1155 | |
| 1156 | offset += length; |
| 1157 | page++; |
| 1158 | } |
| 1159 | } |
| 1160 | |
| 1161 | /* |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1162 | * Clone a portion of a bio, starting at the given byte offset |
| 1163 | * and continuing for the number of bytes indicated. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1164 | */ |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1165 | static struct bio *bio_clone_range(struct bio *bio_src, |
| 1166 | unsigned int offset, |
| 1167 | unsigned int len, |
| 1168 | gfp_t gfpmask) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1169 | { |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1170 | struct bio_vec *bv; |
| 1171 | unsigned int resid; |
| 1172 | unsigned short idx; |
| 1173 | unsigned int voff; |
| 1174 | unsigned short end_idx; |
| 1175 | unsigned short vcnt; |
| 1176 | struct bio *bio; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1177 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1178 | /* Handle the easy case for the caller */ |
| 1179 | |
| 1180 | if (!offset && len == bio_src->bi_size) |
| 1181 | return bio_clone(bio_src, gfpmask); |
| 1182 | |
| 1183 | if (WARN_ON_ONCE(!len)) |
| 1184 | return NULL; |
| 1185 | if (WARN_ON_ONCE(len > bio_src->bi_size)) |
| 1186 | return NULL; |
| 1187 | if (WARN_ON_ONCE(offset > bio_src->bi_size - len)) |
| 1188 | return NULL; |
| 1189 | |
| 1190 | /* Find first affected segment... */ |
| 1191 | |
| 1192 | resid = offset; |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1193 | bio_for_each_segment(bv, bio_src, idx) { |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1194 | if (resid < bv->bv_len) |
| 1195 | break; |
| 1196 | resid -= bv->bv_len; |
| 1197 | } |
| 1198 | voff = resid; |
| 1199 | |
| 1200 | /* ...and the last affected segment */ |
| 1201 | |
| 1202 | resid += len; |
| 1203 | __bio_for_each_segment(bv, bio_src, end_idx, idx) { |
| 1204 | if (resid <= bv->bv_len) |
| 1205 | break; |
| 1206 | resid -= bv->bv_len; |
| 1207 | } |
| 1208 | vcnt = end_idx - idx + 1; |
| 1209 | |
| 1210 | /* Build the clone */ |
| 1211 | |
| 1212 | bio = bio_alloc(gfpmask, (unsigned int) vcnt); |
| 1213 | if (!bio) |
| 1214 | return NULL; /* ENOMEM */ |
| 1215 | |
| 1216 | bio->bi_bdev = bio_src->bi_bdev; |
| 1217 | bio->bi_sector = bio_src->bi_sector + (offset >> SECTOR_SHIFT); |
| 1218 | bio->bi_rw = bio_src->bi_rw; |
| 1219 | bio->bi_flags |= 1 << BIO_CLONED; |
| 1220 | |
| 1221 | /* |
| 1222 | * Copy over our part of the bio_vec, then update the first |
| 1223 | * and last (or only) entries. |
| 1224 | */ |
| 1225 | memcpy(&bio->bi_io_vec[0], &bio_src->bi_io_vec[idx], |
| 1226 | vcnt * sizeof (struct bio_vec)); |
| 1227 | bio->bi_io_vec[0].bv_offset += voff; |
| 1228 | if (vcnt > 1) { |
| 1229 | bio->bi_io_vec[0].bv_len -= voff; |
| 1230 | bio->bi_io_vec[vcnt - 1].bv_len = resid; |
| 1231 | } else { |
| 1232 | bio->bi_io_vec[0].bv_len = len; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1233 | } |
| 1234 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1235 | bio->bi_vcnt = vcnt; |
| 1236 | bio->bi_size = len; |
| 1237 | bio->bi_idx = 0; |
Alex Elder | 542582f | 2012-08-09 10:33:25 -0700 | [diff] [blame] | 1238 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1239 | return bio; |
| 1240 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1241 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1242 | /* |
| 1243 | * Clone a portion of a bio chain, starting at the given byte offset |
| 1244 | * into the first bio in the source chain and continuing for the |
| 1245 | * number of bytes indicated. The result is another bio chain of |
| 1246 | * exactly the given length, or a null pointer on error. |
| 1247 | * |
| 1248 | * The bio_src and offset parameters are both in-out. On entry they |
| 1249 | * refer to the first source bio and the offset into that bio where |
| 1250 | * the start of data to be cloned is located. |
| 1251 | * |
| 1252 | * On return, bio_src is updated to refer to the bio in the source |
| 1253 | * chain that contains first un-cloned byte, and *offset will |
| 1254 | * contain the offset of that byte within that bio. |
| 1255 | */ |
| 1256 | static struct bio *bio_chain_clone_range(struct bio **bio_src, |
| 1257 | unsigned int *offset, |
| 1258 | unsigned int len, |
| 1259 | gfp_t gfpmask) |
| 1260 | { |
| 1261 | struct bio *bi = *bio_src; |
| 1262 | unsigned int off = *offset; |
| 1263 | struct bio *chain = NULL; |
| 1264 | struct bio **end; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1265 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1266 | /* Build up a chain of clone bios up to the limit */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1267 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1268 | if (!bi || off >= bi->bi_size || !len) |
| 1269 | return NULL; /* Nothing to clone */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1270 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1271 | end = &chain; |
| 1272 | while (len) { |
| 1273 | unsigned int bi_size; |
| 1274 | struct bio *bio; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1275 | |
Alex Elder | f5400b7 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 1276 | if (!bi) { |
| 1277 | rbd_warn(NULL, "bio_chain exhausted with %u left", len); |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1278 | goto out_err; /* EINVAL; ran out of bio's */ |
Alex Elder | f5400b7 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 1279 | } |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1280 | bi_size = min_t(unsigned int, bi->bi_size - off, len); |
| 1281 | bio = bio_clone_range(bi, off, bi_size, gfpmask); |
| 1282 | if (!bio) |
| 1283 | goto out_err; /* ENOMEM */ |
| 1284 | |
| 1285 | *end = bio; |
| 1286 | end = &bio->bi_next; |
| 1287 | |
| 1288 | off += bi_size; |
| 1289 | if (off == bi->bi_size) { |
| 1290 | bi = bi->bi_next; |
| 1291 | off = 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1292 | } |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1293 | len -= bi_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1294 | } |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1295 | *bio_src = bi; |
| 1296 | *offset = off; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1297 | |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 1298 | return chain; |
| 1299 | out_err: |
| 1300 | bio_chain_put(chain); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1301 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1302 | return NULL; |
| 1303 | } |
| 1304 | |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1305 | /* |
| 1306 | * The default/initial value for all object request flags is 0. For |
| 1307 | * each flag, once its value is set to 1 it is never reset to 0 |
| 1308 | * again. |
| 1309 | */ |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1310 | static void obj_request_img_data_set(struct rbd_obj_request *obj_request) |
| 1311 | { |
| 1312 | if (test_and_set_bit(OBJ_REQ_IMG_DATA, &obj_request->flags)) { |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1313 | struct rbd_device *rbd_dev; |
| 1314 | |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1315 | rbd_dev = obj_request->img_request->rbd_dev; |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1316 | rbd_warn(rbd_dev, "obj_request %p already marked img_data\n", |
| 1317 | obj_request); |
| 1318 | } |
| 1319 | } |
| 1320 | |
| 1321 | static bool obj_request_img_data_test(struct rbd_obj_request *obj_request) |
| 1322 | { |
| 1323 | smp_mb(); |
| 1324 | return test_bit(OBJ_REQ_IMG_DATA, &obj_request->flags) != 0; |
| 1325 | } |
| 1326 | |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1327 | static void obj_request_done_set(struct rbd_obj_request *obj_request) |
| 1328 | { |
| 1329 | if (test_and_set_bit(OBJ_REQ_DONE, &obj_request->flags)) { |
| 1330 | struct rbd_device *rbd_dev = NULL; |
| 1331 | |
| 1332 | if (obj_request_img_data_test(obj_request)) |
| 1333 | rbd_dev = obj_request->img_request->rbd_dev; |
| 1334 | rbd_warn(rbd_dev, "obj_request %p already marked done\n", |
| 1335 | obj_request); |
| 1336 | } |
| 1337 | } |
| 1338 | |
| 1339 | static bool obj_request_done_test(struct rbd_obj_request *obj_request) |
| 1340 | { |
| 1341 | smp_mb(); |
| 1342 | return test_bit(OBJ_REQ_DONE, &obj_request->flags) != 0; |
| 1343 | } |
| 1344 | |
Alex Elder | 5679c59 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1345 | /* |
| 1346 | * This sets the KNOWN flag after (possibly) setting the EXISTS |
| 1347 | * flag. The latter is set based on the "exists" value provided. |
| 1348 | * |
| 1349 | * Note that for our purposes once an object exists it never goes |
| 1350 | * away again. It's possible that the response from two existence |
| 1351 | * checks are separated by the creation of the target object, and |
| 1352 | * the first ("doesn't exist") response arrives *after* the second |
| 1353 | * ("does exist"). In that case we ignore the second one. |
| 1354 | */ |
| 1355 | static void obj_request_existence_set(struct rbd_obj_request *obj_request, |
| 1356 | bool exists) |
| 1357 | { |
| 1358 | if (exists) |
| 1359 | set_bit(OBJ_REQ_EXISTS, &obj_request->flags); |
| 1360 | set_bit(OBJ_REQ_KNOWN, &obj_request->flags); |
| 1361 | smp_mb(); |
| 1362 | } |
| 1363 | |
| 1364 | static bool obj_request_known_test(struct rbd_obj_request *obj_request) |
| 1365 | { |
| 1366 | smp_mb(); |
| 1367 | return test_bit(OBJ_REQ_KNOWN, &obj_request->flags) != 0; |
| 1368 | } |
| 1369 | |
| 1370 | static bool obj_request_exists_test(struct rbd_obj_request *obj_request) |
| 1371 | { |
| 1372 | smp_mb(); |
| 1373 | return test_bit(OBJ_REQ_EXISTS, &obj_request->flags) != 0; |
| 1374 | } |
| 1375 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1376 | static void rbd_obj_request_get(struct rbd_obj_request *obj_request) |
| 1377 | { |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1378 | dout("%s: obj %p (was %d)\n", __func__, obj_request, |
| 1379 | atomic_read(&obj_request->kref.refcount)); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1380 | kref_get(&obj_request->kref); |
| 1381 | } |
| 1382 | |
| 1383 | static void rbd_obj_request_destroy(struct kref *kref); |
| 1384 | static void rbd_obj_request_put(struct rbd_obj_request *obj_request) |
| 1385 | { |
| 1386 | rbd_assert(obj_request != NULL); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1387 | dout("%s: obj %p (was %d)\n", __func__, obj_request, |
| 1388 | atomic_read(&obj_request->kref.refcount)); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1389 | kref_put(&obj_request->kref, rbd_obj_request_destroy); |
| 1390 | } |
| 1391 | |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1392 | static bool img_request_child_test(struct rbd_img_request *img_request); |
| 1393 | static void rbd_parent_request_destroy(struct kref *kref); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1394 | static void rbd_img_request_destroy(struct kref *kref); |
| 1395 | static void rbd_img_request_put(struct rbd_img_request *img_request) |
| 1396 | { |
| 1397 | rbd_assert(img_request != NULL); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1398 | dout("%s: img %p (was %d)\n", __func__, img_request, |
| 1399 | atomic_read(&img_request->kref.refcount)); |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1400 | if (img_request_child_test(img_request)) |
| 1401 | kref_put(&img_request->kref, rbd_parent_request_destroy); |
| 1402 | else |
| 1403 | kref_put(&img_request->kref, rbd_img_request_destroy); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1404 | } |
| 1405 | |
| 1406 | static inline void rbd_img_obj_request_add(struct rbd_img_request *img_request, |
| 1407 | struct rbd_obj_request *obj_request) |
| 1408 | { |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1409 | rbd_assert(obj_request->img_request == NULL); |
| 1410 | |
Alex Elder | b155e86 | 2013-04-15 14:50:37 -0500 | [diff] [blame] | 1411 | /* Image request now owns object's original reference */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1412 | obj_request->img_request = img_request; |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1413 | obj_request->which = img_request->obj_request_count; |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1414 | rbd_assert(!obj_request_img_data_test(obj_request)); |
| 1415 | obj_request_img_data_set(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1416 | rbd_assert(obj_request->which != BAD_WHICH); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1417 | img_request->obj_request_count++; |
| 1418 | list_add_tail(&obj_request->links, &img_request->obj_requests); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1419 | dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request, |
| 1420 | obj_request->which); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1421 | } |
| 1422 | |
| 1423 | static inline void rbd_img_obj_request_del(struct rbd_img_request *img_request, |
| 1424 | struct rbd_obj_request *obj_request) |
| 1425 | { |
| 1426 | rbd_assert(obj_request->which != BAD_WHICH); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1427 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1428 | dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request, |
| 1429 | obj_request->which); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1430 | list_del(&obj_request->links); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1431 | rbd_assert(img_request->obj_request_count > 0); |
| 1432 | img_request->obj_request_count--; |
| 1433 | rbd_assert(obj_request->which == img_request->obj_request_count); |
| 1434 | obj_request->which = BAD_WHICH; |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1435 | rbd_assert(obj_request_img_data_test(obj_request)); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1436 | rbd_assert(obj_request->img_request == img_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1437 | obj_request->img_request = NULL; |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 1438 | obj_request->callback = NULL; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1439 | rbd_obj_request_put(obj_request); |
| 1440 | } |
| 1441 | |
| 1442 | static bool obj_request_type_valid(enum obj_request_type type) |
| 1443 | { |
| 1444 | switch (type) { |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1445 | case OBJ_REQUEST_NODATA: |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1446 | case OBJ_REQUEST_BIO: |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1447 | case OBJ_REQUEST_PAGES: |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1448 | return true; |
| 1449 | default: |
| 1450 | return false; |
| 1451 | } |
| 1452 | } |
| 1453 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1454 | static int rbd_obj_request_submit(struct ceph_osd_client *osdc, |
| 1455 | struct rbd_obj_request *obj_request) |
| 1456 | { |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1457 | dout("%s: osdc %p obj %p\n", __func__, osdc, obj_request); |
| 1458 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1459 | return ceph_osdc_start_request(osdc, obj_request->osd_req, false); |
| 1460 | } |
| 1461 | |
| 1462 | static void rbd_img_request_complete(struct rbd_img_request *img_request) |
| 1463 | { |
Alex Elder | 55f27e0 | 2013-04-10 12:34:25 -0500 | [diff] [blame] | 1464 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1465 | dout("%s: img %p\n", __func__, img_request); |
Alex Elder | 55f27e0 | 2013-04-10 12:34:25 -0500 | [diff] [blame] | 1466 | |
| 1467 | /* |
| 1468 | * If no error occurred, compute the aggregate transfer |
| 1469 | * count for the image request. We could instead use |
| 1470 | * atomic64_cmpxchg() to update it as each object request |
| 1471 | * completes; not clear which way is better off hand. |
| 1472 | */ |
| 1473 | if (!img_request->result) { |
| 1474 | struct rbd_obj_request *obj_request; |
| 1475 | u64 xferred = 0; |
| 1476 | |
| 1477 | for_each_obj_request(img_request, obj_request) |
| 1478 | xferred += obj_request->xferred; |
| 1479 | img_request->xferred = xferred; |
| 1480 | } |
| 1481 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1482 | if (img_request->callback) |
| 1483 | img_request->callback(img_request); |
| 1484 | else |
| 1485 | rbd_img_request_put(img_request); |
| 1486 | } |
| 1487 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1488 | /* Caller is responsible for rbd_obj_request_destroy(obj_request) */ |
| 1489 | |
| 1490 | static int rbd_obj_request_wait(struct rbd_obj_request *obj_request) |
| 1491 | { |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1492 | dout("%s: obj %p\n", __func__, obj_request); |
| 1493 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1494 | return wait_for_completion_interruptible(&obj_request->completion); |
| 1495 | } |
| 1496 | |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 1497 | /* |
| 1498 | * The default/initial value for all image request flags is 0. Each |
| 1499 | * is conditionally set to 1 at image request initialization time |
| 1500 | * and currently never change thereafter. |
| 1501 | */ |
| 1502 | static void img_request_write_set(struct rbd_img_request *img_request) |
| 1503 | { |
| 1504 | set_bit(IMG_REQ_WRITE, &img_request->flags); |
| 1505 | smp_mb(); |
| 1506 | } |
| 1507 | |
| 1508 | static bool img_request_write_test(struct rbd_img_request *img_request) |
| 1509 | { |
| 1510 | smp_mb(); |
| 1511 | return test_bit(IMG_REQ_WRITE, &img_request->flags) != 0; |
| 1512 | } |
| 1513 | |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1514 | static void img_request_child_set(struct rbd_img_request *img_request) |
| 1515 | { |
| 1516 | set_bit(IMG_REQ_CHILD, &img_request->flags); |
| 1517 | smp_mb(); |
| 1518 | } |
| 1519 | |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1520 | static void img_request_child_clear(struct rbd_img_request *img_request) |
| 1521 | { |
| 1522 | clear_bit(IMG_REQ_CHILD, &img_request->flags); |
| 1523 | smp_mb(); |
| 1524 | } |
| 1525 | |
Alex Elder | 9849e98 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1526 | static bool img_request_child_test(struct rbd_img_request *img_request) |
| 1527 | { |
| 1528 | smp_mb(); |
| 1529 | return test_bit(IMG_REQ_CHILD, &img_request->flags) != 0; |
| 1530 | } |
| 1531 | |
Alex Elder | d0b2e94 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1532 | static void img_request_layered_set(struct rbd_img_request *img_request) |
| 1533 | { |
| 1534 | set_bit(IMG_REQ_LAYERED, &img_request->flags); |
| 1535 | smp_mb(); |
| 1536 | } |
| 1537 | |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1538 | static void img_request_layered_clear(struct rbd_img_request *img_request) |
| 1539 | { |
| 1540 | clear_bit(IMG_REQ_LAYERED, &img_request->flags); |
| 1541 | smp_mb(); |
| 1542 | } |
| 1543 | |
Alex Elder | d0b2e94 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1544 | static bool img_request_layered_test(struct rbd_img_request *img_request) |
| 1545 | { |
| 1546 | smp_mb(); |
| 1547 | return test_bit(IMG_REQ_LAYERED, &img_request->flags) != 0; |
| 1548 | } |
| 1549 | |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1550 | static void |
| 1551 | rbd_img_obj_request_read_callback(struct rbd_obj_request *obj_request) |
| 1552 | { |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1553 | u64 xferred = obj_request->xferred; |
| 1554 | u64 length = obj_request->length; |
| 1555 | |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1556 | dout("%s: obj %p img %p result %d %llu/%llu\n", __func__, |
| 1557 | obj_request, obj_request->img_request, obj_request->result, |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1558 | xferred, length); |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1559 | /* |
Josh Durgin | 17c1cc1 | 2013-08-26 17:55:38 -0700 | [diff] [blame] | 1560 | * ENOENT means a hole in the image. We zero-fill the entire |
| 1561 | * length of the request. A short read also implies zero-fill |
| 1562 | * to the end of the request. An error requires the whole |
| 1563 | * length of the request to be reported finished with an error |
| 1564 | * to the block layer. In each case we update the xferred |
| 1565 | * count to indicate the whole request was satisfied. |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1566 | */ |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1567 | rbd_assert(obj_request->type != OBJ_REQUEST_NODATA); |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1568 | if (obj_request->result == -ENOENT) { |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1569 | if (obj_request->type == OBJ_REQUEST_BIO) |
| 1570 | zero_bio_chain(obj_request->bio_list, 0); |
| 1571 | else |
| 1572 | zero_pages(obj_request->pages, 0, length); |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1573 | obj_request->result = 0; |
Alex Elder | b9434c5 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1574 | } else if (xferred < length && !obj_request->result) { |
| 1575 | if (obj_request->type == OBJ_REQUEST_BIO) |
| 1576 | zero_bio_chain(obj_request->bio_list, xferred); |
| 1577 | else |
| 1578 | zero_pages(obj_request->pages, xferred, length); |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1579 | } |
Josh Durgin | 17c1cc1 | 2013-08-26 17:55:38 -0700 | [diff] [blame] | 1580 | obj_request->xferred = length; |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1581 | obj_request_done_set(obj_request); |
| 1582 | } |
| 1583 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1584 | static void rbd_obj_request_complete(struct rbd_obj_request *obj_request) |
| 1585 | { |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1586 | dout("%s: obj %p cb %p\n", __func__, obj_request, |
| 1587 | obj_request->callback); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1588 | if (obj_request->callback) |
| 1589 | obj_request->callback(obj_request); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1590 | else |
| 1591 | complete_all(&obj_request->completion); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1592 | } |
| 1593 | |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1594 | static void rbd_osd_trivial_callback(struct rbd_obj_request *obj_request) |
Alex Elder | 39bf2c5 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1595 | { |
| 1596 | dout("%s: obj %p\n", __func__, obj_request); |
| 1597 | obj_request_done_set(obj_request); |
| 1598 | } |
| 1599 | |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1600 | static void rbd_osd_read_callback(struct rbd_obj_request *obj_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1601 | { |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1602 | struct rbd_img_request *img_request = NULL; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 1603 | struct rbd_device *rbd_dev = NULL; |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1604 | bool layered = false; |
| 1605 | |
| 1606 | if (obj_request_img_data_test(obj_request)) { |
| 1607 | img_request = obj_request->img_request; |
| 1608 | layered = img_request && img_request_layered_test(img_request); |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 1609 | rbd_dev = img_request->rbd_dev; |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1610 | } |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1611 | |
| 1612 | dout("%s: obj %p img %p result %d %llu/%llu\n", __func__, |
| 1613 | obj_request, img_request, obj_request->result, |
| 1614 | obj_request->xferred, obj_request->length); |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 1615 | if (layered && obj_request->result == -ENOENT && |
| 1616 | obj_request->img_offset < rbd_dev->parent_overlap) |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1617 | rbd_img_parent_read(obj_request); |
| 1618 | else if (img_request) |
Alex Elder | 6e2a450 | 2013-03-27 09:16:30 -0500 | [diff] [blame] | 1619 | rbd_img_obj_request_read_callback(obj_request); |
| 1620 | else |
| 1621 | obj_request_done_set(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1622 | } |
| 1623 | |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1624 | static void rbd_osd_write_callback(struct rbd_obj_request *obj_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1625 | { |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 1626 | dout("%s: obj %p result %d %llu\n", __func__, obj_request, |
| 1627 | obj_request->result, obj_request->length); |
| 1628 | /* |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1629 | * There is no such thing as a successful short write. Set |
| 1630 | * it to our originally-requested length. |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 1631 | */ |
| 1632 | obj_request->xferred = obj_request->length; |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1633 | obj_request_done_set(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1634 | } |
| 1635 | |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1636 | /* |
| 1637 | * For a simple stat call there's nothing to do. We'll do more if |
| 1638 | * this is part of a write sequence for a layered image. |
| 1639 | */ |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1640 | static void rbd_osd_stat_callback(struct rbd_obj_request *obj_request) |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1641 | { |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1642 | dout("%s: obj %p\n", __func__, obj_request); |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1643 | obj_request_done_set(obj_request); |
| 1644 | } |
| 1645 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1646 | static void rbd_osd_req_callback(struct ceph_osd_request *osd_req, |
| 1647 | struct ceph_msg *msg) |
| 1648 | { |
| 1649 | struct rbd_obj_request *obj_request = osd_req->r_priv; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1650 | u16 opcode; |
| 1651 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1652 | dout("%s: osd_req %p msg %p\n", __func__, osd_req, msg); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1653 | rbd_assert(osd_req == obj_request->osd_req); |
Alex Elder | 57acbaa | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1654 | if (obj_request_img_data_test(obj_request)) { |
| 1655 | rbd_assert(obj_request->img_request); |
| 1656 | rbd_assert(obj_request->which != BAD_WHICH); |
| 1657 | } else { |
| 1658 | rbd_assert(obj_request->which == BAD_WHICH); |
| 1659 | } |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1660 | |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 1661 | if (osd_req->r_result < 0) |
| 1662 | obj_request->result = osd_req->r_result; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1663 | |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1664 | BUG_ON(osd_req->r_num_ops > 2); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1665 | |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1666 | /* |
| 1667 | * We support a 64-bit length, but ultimately it has to be |
| 1668 | * passed to blk_end_request(), which takes an unsigned int. |
| 1669 | */ |
Sage Weil | 1b83bef | 2013-02-25 16:11:12 -0800 | [diff] [blame] | 1670 | obj_request->xferred = osd_req->r_reply_op_len[0]; |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1671 | rbd_assert(obj_request->xferred < (u64)UINT_MAX); |
Alex Elder | 7952873 | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1672 | opcode = osd_req->r_ops[0].op; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1673 | switch (opcode) { |
| 1674 | case CEPH_OSD_OP_READ: |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1675 | rbd_osd_read_callback(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1676 | break; |
| 1677 | case CEPH_OSD_OP_WRITE: |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1678 | rbd_osd_write_callback(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1679 | break; |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1680 | case CEPH_OSD_OP_STAT: |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1681 | rbd_osd_stat_callback(obj_request); |
Alex Elder | fbfab53 | 2013-02-08 09:55:48 -0600 | [diff] [blame] | 1682 | break; |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 1683 | case CEPH_OSD_OP_CALL: |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 1684 | case CEPH_OSD_OP_NOTIFY_ACK: |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1685 | case CEPH_OSD_OP_WATCH: |
Alex Elder | c47f937 | 2013-02-26 14:23:07 -0600 | [diff] [blame] | 1686 | rbd_osd_trivial_callback(obj_request); |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1687 | break; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1688 | default: |
| 1689 | rbd_warn(NULL, "%s: unsupported op %hu\n", |
| 1690 | obj_request->object_name, (unsigned short) opcode); |
| 1691 | break; |
| 1692 | } |
| 1693 | |
Alex Elder | 0774130 | 2013-02-05 23:41:50 -0600 | [diff] [blame] | 1694 | if (obj_request_done_test(obj_request)) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1695 | rbd_obj_request_complete(obj_request); |
| 1696 | } |
| 1697 | |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1698 | 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] | 1699 | { |
| 1700 | struct rbd_img_request *img_request = obj_request->img_request; |
Alex Elder | 8c042b0 | 2013-04-03 01:28:58 -0500 | [diff] [blame] | 1701 | struct ceph_osd_request *osd_req = obj_request->osd_req; |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1702 | u64 snap_id; |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1703 | |
Alex Elder | 8c042b0 | 2013-04-03 01:28:58 -0500 | [diff] [blame] | 1704 | rbd_assert(osd_req != NULL); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1705 | |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1706 | snap_id = img_request ? img_request->snap_id : CEPH_NOSNAP; |
Alex Elder | 8c042b0 | 2013-04-03 01:28:58 -0500 | [diff] [blame] | 1707 | ceph_osdc_build_request(osd_req, obj_request->offset, |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1708 | NULL, snap_id, NULL); |
| 1709 | } |
| 1710 | |
| 1711 | static void rbd_osd_req_format_write(struct rbd_obj_request *obj_request) |
| 1712 | { |
| 1713 | struct rbd_img_request *img_request = obj_request->img_request; |
| 1714 | struct ceph_osd_request *osd_req = obj_request->osd_req; |
| 1715 | struct ceph_snap_context *snapc; |
| 1716 | struct timespec mtime = CURRENT_TIME; |
| 1717 | |
| 1718 | rbd_assert(osd_req != NULL); |
| 1719 | |
| 1720 | snapc = img_request ? img_request->snapc : NULL; |
| 1721 | ceph_osdc_build_request(osd_req, obj_request->offset, |
| 1722 | snapc, CEPH_NOSNAP, &mtime); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1723 | } |
| 1724 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1725 | static struct ceph_osd_request *rbd_osd_req_create( |
| 1726 | struct rbd_device *rbd_dev, |
| 1727 | bool write_request, |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1728 | struct rbd_obj_request *obj_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1729 | { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1730 | struct ceph_snap_context *snapc = NULL; |
| 1731 | struct ceph_osd_client *osdc; |
| 1732 | struct ceph_osd_request *osd_req; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1733 | |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1734 | if (obj_request_img_data_test(obj_request)) { |
| 1735 | struct rbd_img_request *img_request = obj_request->img_request; |
| 1736 | |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 1737 | rbd_assert(write_request == |
| 1738 | img_request_write_test(img_request)); |
| 1739 | if (write_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1740 | snapc = img_request->snapc; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1741 | } |
| 1742 | |
| 1743 | /* Allocate and initialize the request, for the single op */ |
| 1744 | |
| 1745 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 1746 | osd_req = ceph_osdc_alloc_request(osdc, snapc, 1, false, GFP_ATOMIC); |
| 1747 | if (!osd_req) |
| 1748 | return NULL; /* ENOMEM */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1749 | |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1750 | if (write_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1751 | osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 1752 | else |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1753 | osd_req->r_flags = CEPH_OSD_FLAG_READ; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1754 | |
| 1755 | osd_req->r_callback = rbd_osd_req_callback; |
| 1756 | osd_req->r_priv = obj_request; |
| 1757 | |
| 1758 | osd_req->r_oid_len = strlen(obj_request->object_name); |
| 1759 | rbd_assert(osd_req->r_oid_len < sizeof (osd_req->r_oid)); |
| 1760 | memcpy(osd_req->r_oid, obj_request->object_name, osd_req->r_oid_len); |
| 1761 | |
| 1762 | osd_req->r_file_layout = rbd_dev->layout; /* struct */ |
| 1763 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1764 | return osd_req; |
| 1765 | } |
| 1766 | |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 1767 | /* |
| 1768 | * Create a copyup osd request based on the information in the |
| 1769 | * object request supplied. A copyup request has two osd ops, |
| 1770 | * a copyup method call, and a "normal" write request. |
| 1771 | */ |
| 1772 | static struct ceph_osd_request * |
| 1773 | rbd_osd_req_create_copyup(struct rbd_obj_request *obj_request) |
| 1774 | { |
| 1775 | struct rbd_img_request *img_request; |
| 1776 | struct ceph_snap_context *snapc; |
| 1777 | struct rbd_device *rbd_dev; |
| 1778 | struct ceph_osd_client *osdc; |
| 1779 | struct ceph_osd_request *osd_req; |
| 1780 | |
| 1781 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 1782 | img_request = obj_request->img_request; |
| 1783 | rbd_assert(img_request); |
| 1784 | rbd_assert(img_request_write_test(img_request)); |
| 1785 | |
| 1786 | /* Allocate and initialize the request, for the two ops */ |
| 1787 | |
| 1788 | snapc = img_request->snapc; |
| 1789 | rbd_dev = img_request->rbd_dev; |
| 1790 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 1791 | osd_req = ceph_osdc_alloc_request(osdc, snapc, 2, false, GFP_ATOMIC); |
| 1792 | if (!osd_req) |
| 1793 | return NULL; /* ENOMEM */ |
| 1794 | |
| 1795 | osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK; |
| 1796 | osd_req->r_callback = rbd_osd_req_callback; |
| 1797 | osd_req->r_priv = obj_request; |
| 1798 | |
| 1799 | osd_req->r_oid_len = strlen(obj_request->object_name); |
| 1800 | rbd_assert(osd_req->r_oid_len < sizeof (osd_req->r_oid)); |
| 1801 | memcpy(osd_req->r_oid, obj_request->object_name, osd_req->r_oid_len); |
| 1802 | |
| 1803 | osd_req->r_file_layout = rbd_dev->layout; /* struct */ |
| 1804 | |
| 1805 | return osd_req; |
| 1806 | } |
| 1807 | |
| 1808 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1809 | static void rbd_osd_req_destroy(struct ceph_osd_request *osd_req) |
| 1810 | { |
| 1811 | ceph_osdc_put_request(osd_req); |
| 1812 | } |
| 1813 | |
| 1814 | /* object_name is assumed to be a non-null pointer and NUL-terminated */ |
| 1815 | |
| 1816 | static struct rbd_obj_request *rbd_obj_request_create(const char *object_name, |
| 1817 | u64 offset, u64 length, |
| 1818 | enum obj_request_type type) |
| 1819 | { |
| 1820 | struct rbd_obj_request *obj_request; |
| 1821 | size_t size; |
| 1822 | char *name; |
| 1823 | |
| 1824 | rbd_assert(obj_request_type_valid(type)); |
| 1825 | |
| 1826 | size = strlen(object_name) + 1; |
Alex Elder | f907ad5 | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1827 | name = kmalloc(size, GFP_KERNEL); |
| 1828 | if (!name) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1829 | return NULL; |
| 1830 | |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1831 | obj_request = kmem_cache_zalloc(rbd_obj_request_cache, GFP_KERNEL); |
Alex Elder | f907ad5 | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1832 | if (!obj_request) { |
| 1833 | kfree(name); |
| 1834 | return NULL; |
| 1835 | } |
| 1836 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1837 | obj_request->object_name = memcpy(name, object_name, size); |
| 1838 | obj_request->offset = offset; |
| 1839 | obj_request->length = length; |
Alex Elder | 926f9b3 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 1840 | obj_request->flags = 0; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1841 | obj_request->which = BAD_WHICH; |
| 1842 | obj_request->type = type; |
| 1843 | INIT_LIST_HEAD(&obj_request->links); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1844 | init_completion(&obj_request->completion); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1845 | kref_init(&obj_request->kref); |
| 1846 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1847 | dout("%s: \"%s\" %llu/%llu %d -> obj %p\n", __func__, object_name, |
| 1848 | offset, length, (int)type, obj_request); |
| 1849 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1850 | return obj_request; |
| 1851 | } |
| 1852 | |
| 1853 | static void rbd_obj_request_destroy(struct kref *kref) |
| 1854 | { |
| 1855 | struct rbd_obj_request *obj_request; |
| 1856 | |
| 1857 | obj_request = container_of(kref, struct rbd_obj_request, kref); |
| 1858 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1859 | dout("%s: obj %p\n", __func__, obj_request); |
| 1860 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1861 | rbd_assert(obj_request->img_request == NULL); |
| 1862 | rbd_assert(obj_request->which == BAD_WHICH); |
| 1863 | |
| 1864 | if (obj_request->osd_req) |
| 1865 | rbd_osd_req_destroy(obj_request->osd_req); |
| 1866 | |
| 1867 | rbd_assert(obj_request_type_valid(obj_request->type)); |
| 1868 | switch (obj_request->type) { |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 1869 | case OBJ_REQUEST_NODATA: |
| 1870 | break; /* Nothing to do */ |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1871 | case OBJ_REQUEST_BIO: |
| 1872 | if (obj_request->bio_list) |
| 1873 | bio_chain_put(obj_request->bio_list); |
| 1874 | break; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 1875 | case OBJ_REQUEST_PAGES: |
| 1876 | if (obj_request->pages) |
| 1877 | ceph_release_page_vector(obj_request->pages, |
| 1878 | obj_request->page_count); |
| 1879 | break; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1880 | } |
| 1881 | |
Alex Elder | f907ad5 | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1882 | kfree(obj_request->object_name); |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1883 | obj_request->object_name = NULL; |
| 1884 | kmem_cache_free(rbd_obj_request_cache, obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1885 | } |
| 1886 | |
Alex Elder | fb65d228 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1887 | /* It's OK to call this for a device with no parent */ |
| 1888 | |
| 1889 | static void rbd_spec_put(struct rbd_spec *spec); |
| 1890 | static void rbd_dev_unparent(struct rbd_device *rbd_dev) |
| 1891 | { |
| 1892 | rbd_dev_remove_parent(rbd_dev); |
| 1893 | rbd_spec_put(rbd_dev->parent_spec); |
| 1894 | rbd_dev->parent_spec = NULL; |
| 1895 | rbd_dev->parent_overlap = 0; |
| 1896 | } |
| 1897 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1898 | /* |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1899 | * Parent image reference counting is used to determine when an |
| 1900 | * image's parent fields can be safely torn down--after there are no |
| 1901 | * more in-flight requests to the parent image. When the last |
| 1902 | * reference is dropped, cleaning them up is safe. |
| 1903 | */ |
| 1904 | static void rbd_dev_parent_put(struct rbd_device *rbd_dev) |
| 1905 | { |
| 1906 | int counter; |
| 1907 | |
| 1908 | if (!rbd_dev->parent_spec) |
| 1909 | return; |
| 1910 | |
| 1911 | counter = atomic_dec_return_safe(&rbd_dev->parent_ref); |
| 1912 | if (counter > 0) |
| 1913 | return; |
| 1914 | |
| 1915 | /* Last reference; clean up parent data structures */ |
| 1916 | |
| 1917 | if (!counter) |
| 1918 | rbd_dev_unparent(rbd_dev); |
| 1919 | else |
| 1920 | rbd_warn(rbd_dev, "parent reference underflow\n"); |
| 1921 | } |
| 1922 | |
| 1923 | /* |
| 1924 | * If an image has a non-zero parent overlap, get a reference to its |
| 1925 | * parent. |
| 1926 | * |
Alex Elder | 392a9da | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 1927 | * We must get the reference before checking for the overlap to |
| 1928 | * coordinate properly with zeroing the parent overlap in |
| 1929 | * rbd_dev_v2_parent_info() when an image gets flattened. We |
| 1930 | * drop it again if there is no overlap. |
| 1931 | * |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1932 | * Returns true if the rbd device has a parent with a non-zero |
| 1933 | * overlap and a reference for it was successfully taken, or |
| 1934 | * false otherwise. |
| 1935 | */ |
| 1936 | static bool rbd_dev_parent_get(struct rbd_device *rbd_dev) |
| 1937 | { |
| 1938 | int counter; |
| 1939 | |
| 1940 | if (!rbd_dev->parent_spec) |
| 1941 | return false; |
| 1942 | |
| 1943 | counter = atomic_inc_return_safe(&rbd_dev->parent_ref); |
| 1944 | if (counter > 0 && rbd_dev->parent_overlap) |
| 1945 | return true; |
| 1946 | |
| 1947 | /* Image was flattened, but parent is not yet torn down */ |
| 1948 | |
| 1949 | if (counter < 0) |
| 1950 | rbd_warn(rbd_dev, "parent reference overflow\n"); |
| 1951 | |
| 1952 | return false; |
| 1953 | } |
| 1954 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1955 | /* |
| 1956 | * Caller is responsible for filling in the list of object requests |
| 1957 | * that comprises the image request, and the Linux request pointer |
| 1958 | * (if there is one). |
| 1959 | */ |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 1960 | static struct rbd_img_request *rbd_img_request_create( |
| 1961 | struct rbd_device *rbd_dev, |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1962 | u64 offset, u64 length, |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1963 | bool write_request) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1964 | { |
| 1965 | struct rbd_img_request *img_request; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1966 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 1967 | img_request = kmem_cache_alloc(rbd_img_request_cache, GFP_ATOMIC); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1968 | if (!img_request) |
| 1969 | return NULL; |
| 1970 | |
| 1971 | if (write_request) { |
| 1972 | down_read(&rbd_dev->header_rwsem); |
Alex Elder | 812164f8 | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 1973 | ceph_get_snap_context(rbd_dev->header.snapc); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1974 | up_read(&rbd_dev->header_rwsem); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1975 | } |
| 1976 | |
| 1977 | img_request->rq = NULL; |
| 1978 | img_request->rbd_dev = rbd_dev; |
| 1979 | img_request->offset = offset; |
| 1980 | img_request->length = length; |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 1981 | img_request->flags = 0; |
| 1982 | if (write_request) { |
| 1983 | img_request_write_set(img_request); |
Alex Elder | 468521c | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 1984 | img_request->snapc = rbd_dev->header.snapc; |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 1985 | } else { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1986 | img_request->snap_id = rbd_dev->spec->snap_id; |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 1987 | } |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 1988 | if (rbd_dev_parent_get(rbd_dev)) |
Alex Elder | d0b2e94 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1989 | img_request_layered_set(img_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1990 | spin_lock_init(&img_request->completion_lock); |
| 1991 | img_request->next_completion = 0; |
| 1992 | img_request->callback = NULL; |
Alex Elder | a5a337d | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 1993 | img_request->result = 0; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 1994 | img_request->obj_request_count = 0; |
| 1995 | INIT_LIST_HEAD(&img_request->obj_requests); |
| 1996 | kref_init(&img_request->kref); |
| 1997 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 1998 | dout("%s: rbd_dev %p %s %llu/%llu -> img %p\n", __func__, rbd_dev, |
| 1999 | write_request ? "write" : "read", offset, length, |
| 2000 | img_request); |
| 2001 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2002 | return img_request; |
| 2003 | } |
| 2004 | |
| 2005 | static void rbd_img_request_destroy(struct kref *kref) |
| 2006 | { |
| 2007 | struct rbd_img_request *img_request; |
| 2008 | struct rbd_obj_request *obj_request; |
| 2009 | struct rbd_obj_request *next_obj_request; |
| 2010 | |
| 2011 | img_request = container_of(kref, struct rbd_img_request, kref); |
| 2012 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2013 | dout("%s: img %p\n", __func__, img_request); |
| 2014 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2015 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) |
| 2016 | rbd_img_obj_request_del(img_request, obj_request); |
Alex Elder | 25dcf95 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2017 | rbd_assert(img_request->obj_request_count == 0); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2018 | |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2019 | if (img_request_layered_test(img_request)) { |
| 2020 | img_request_layered_clear(img_request); |
| 2021 | rbd_dev_parent_put(img_request->rbd_dev); |
| 2022 | } |
| 2023 | |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2024 | if (img_request_write_test(img_request)) |
Alex Elder | 812164f8 | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 2025 | ceph_put_snap_context(img_request->snapc); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2026 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 2027 | kmem_cache_free(rbd_img_request_cache, img_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2028 | } |
| 2029 | |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2030 | static struct rbd_img_request *rbd_parent_request_create( |
| 2031 | struct rbd_obj_request *obj_request, |
| 2032 | u64 img_offset, u64 length) |
| 2033 | { |
| 2034 | struct rbd_img_request *parent_request; |
| 2035 | struct rbd_device *rbd_dev; |
| 2036 | |
| 2037 | rbd_assert(obj_request->img_request); |
| 2038 | rbd_dev = obj_request->img_request->rbd_dev; |
| 2039 | |
| 2040 | parent_request = rbd_img_request_create(rbd_dev->parent, |
| 2041 | img_offset, length, false); |
| 2042 | if (!parent_request) |
| 2043 | return NULL; |
| 2044 | |
| 2045 | img_request_child_set(parent_request); |
| 2046 | rbd_obj_request_get(obj_request); |
| 2047 | parent_request->obj_request = obj_request; |
| 2048 | |
| 2049 | return parent_request; |
| 2050 | } |
| 2051 | |
| 2052 | static void rbd_parent_request_destroy(struct kref *kref) |
| 2053 | { |
| 2054 | struct rbd_img_request *parent_request; |
| 2055 | struct rbd_obj_request *orig_request; |
| 2056 | |
| 2057 | parent_request = container_of(kref, struct rbd_img_request, kref); |
| 2058 | orig_request = parent_request->obj_request; |
| 2059 | |
| 2060 | parent_request->obj_request = NULL; |
| 2061 | rbd_obj_request_put(orig_request); |
| 2062 | img_request_child_clear(parent_request); |
| 2063 | |
| 2064 | rbd_img_request_destroy(kref); |
| 2065 | } |
| 2066 | |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2067 | static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request) |
| 2068 | { |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2069 | struct rbd_img_request *img_request; |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2070 | unsigned int xferred; |
| 2071 | int result; |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2072 | bool more; |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2073 | |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2074 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 2075 | img_request = obj_request->img_request; |
| 2076 | |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2077 | rbd_assert(obj_request->xferred <= (u64)UINT_MAX); |
| 2078 | xferred = (unsigned int)obj_request->xferred; |
| 2079 | result = obj_request->result; |
| 2080 | if (result) { |
| 2081 | struct rbd_device *rbd_dev = img_request->rbd_dev; |
| 2082 | |
| 2083 | rbd_warn(rbd_dev, "%s %llx at %llx (%llx)\n", |
| 2084 | img_request_write_test(img_request) ? "write" : "read", |
| 2085 | obj_request->length, obj_request->img_offset, |
| 2086 | obj_request->offset); |
| 2087 | rbd_warn(rbd_dev, " result %d xferred %x\n", |
| 2088 | result, xferred); |
| 2089 | if (!img_request->result) |
| 2090 | img_request->result = result; |
| 2091 | } |
| 2092 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2093 | /* Image object requests don't own their page array */ |
| 2094 | |
| 2095 | if (obj_request->type == OBJ_REQUEST_PAGES) { |
| 2096 | obj_request->pages = NULL; |
| 2097 | obj_request->page_count = 0; |
| 2098 | } |
| 2099 | |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2100 | if (img_request_child_test(img_request)) { |
| 2101 | rbd_assert(img_request->obj_request != NULL); |
| 2102 | more = obj_request->which < img_request->obj_request_count - 1; |
| 2103 | } else { |
| 2104 | rbd_assert(img_request->rq != NULL); |
| 2105 | more = blk_end_request(img_request->rq, result, xferred); |
| 2106 | } |
| 2107 | |
| 2108 | return more; |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2109 | } |
| 2110 | |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2111 | static void rbd_img_obj_callback(struct rbd_obj_request *obj_request) |
| 2112 | { |
| 2113 | struct rbd_img_request *img_request; |
| 2114 | u32 which = obj_request->which; |
| 2115 | bool more = true; |
| 2116 | |
Alex Elder | 6365d33 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2117 | rbd_assert(obj_request_img_data_test(obj_request)); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2118 | img_request = obj_request->img_request; |
| 2119 | |
| 2120 | dout("%s: img %p obj %p\n", __func__, img_request, obj_request); |
| 2121 | rbd_assert(img_request != NULL); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2122 | rbd_assert(img_request->obj_request_count > 0); |
| 2123 | rbd_assert(which != BAD_WHICH); |
| 2124 | rbd_assert(which < img_request->obj_request_count); |
| 2125 | rbd_assert(which >= img_request->next_completion); |
| 2126 | |
| 2127 | spin_lock_irq(&img_request->completion_lock); |
| 2128 | if (which != img_request->next_completion) |
| 2129 | goto out; |
| 2130 | |
| 2131 | for_each_obj_request_from(img_request, obj_request) { |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2132 | rbd_assert(more); |
| 2133 | rbd_assert(which < img_request->obj_request_count); |
| 2134 | |
| 2135 | if (!obj_request_done_test(obj_request)) |
| 2136 | break; |
Alex Elder | 1217857f | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2137 | more = rbd_img_obj_end_request(obj_request); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2138 | which++; |
| 2139 | } |
| 2140 | |
| 2141 | rbd_assert(more ^ (which == img_request->obj_request_count)); |
| 2142 | img_request->next_completion = which; |
| 2143 | out: |
| 2144 | spin_unlock_irq(&img_request->completion_lock); |
| 2145 | |
| 2146 | if (!more) |
| 2147 | rbd_img_request_complete(img_request); |
| 2148 | } |
| 2149 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2150 | /* |
| 2151 | * Split up an image request into one or more object requests, each |
| 2152 | * to a different object. The "type" parameter indicates whether |
| 2153 | * "data_desc" is the pointer to the head of a list of bio |
| 2154 | * structures, or the base of a page array. In either case this |
| 2155 | * function assumes data_desc describes memory sufficient to hold |
| 2156 | * all data described by the image request. |
| 2157 | */ |
| 2158 | static int rbd_img_request_fill(struct rbd_img_request *img_request, |
| 2159 | enum obj_request_type type, |
| 2160 | void *data_desc) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2161 | { |
| 2162 | struct rbd_device *rbd_dev = img_request->rbd_dev; |
| 2163 | struct rbd_obj_request *obj_request = NULL; |
| 2164 | struct rbd_obj_request *next_obj_request; |
Alex Elder | 0c42524 | 2013-02-08 09:55:49 -0600 | [diff] [blame] | 2165 | bool write_request = img_request_write_test(img_request); |
Jingoo Han | a158073 | 2013-08-09 13:04:35 +0900 | [diff] [blame] | 2166 | struct bio *bio_list = NULL; |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2167 | unsigned int bio_offset = 0; |
Jingoo Han | a158073 | 2013-08-09 13:04:35 +0900 | [diff] [blame] | 2168 | struct page **pages = NULL; |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2169 | u64 img_offset; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2170 | u64 resid; |
| 2171 | u16 opcode; |
| 2172 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2173 | dout("%s: img %p type %d data_desc %p\n", __func__, img_request, |
| 2174 | (int)type, data_desc); |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2175 | |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2176 | opcode = write_request ? CEPH_OSD_OP_WRITE : CEPH_OSD_OP_READ; |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2177 | img_offset = img_request->offset; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2178 | resid = img_request->length; |
Alex Elder | 4dda41d | 2013-02-20 21:59:33 -0600 | [diff] [blame] | 2179 | rbd_assert(resid > 0); |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2180 | |
| 2181 | if (type == OBJ_REQUEST_BIO) { |
| 2182 | bio_list = data_desc; |
| 2183 | rbd_assert(img_offset == bio_list->bi_sector << SECTOR_SHIFT); |
| 2184 | } else { |
| 2185 | rbd_assert(type == OBJ_REQUEST_PAGES); |
| 2186 | pages = data_desc; |
| 2187 | } |
| 2188 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2189 | while (resid) { |
Alex Elder | 2fa1232 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2190 | struct ceph_osd_request *osd_req; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2191 | const char *object_name; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2192 | u64 offset; |
| 2193 | u64 length; |
| 2194 | |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2195 | object_name = rbd_segment_name(rbd_dev, img_offset); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2196 | if (!object_name) |
| 2197 | goto out_unwind; |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2198 | offset = rbd_segment_offset(rbd_dev, img_offset); |
| 2199 | length = rbd_segment_length(rbd_dev, img_offset, resid); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2200 | obj_request = rbd_obj_request_create(object_name, |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2201 | offset, length, type); |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 2202 | /* object request has its own copy of the object name */ |
| 2203 | rbd_segment_name_free(object_name); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2204 | if (!obj_request) |
| 2205 | goto out_unwind; |
Josh Durgin | 03507db | 2013-08-27 14:45:46 -0700 | [diff] [blame] | 2206 | /* |
| 2207 | * set obj_request->img_request before creating the |
| 2208 | * osd_request so that it gets the right snapc |
| 2209 | */ |
| 2210 | rbd_img_obj_request_add(img_request, obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2211 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2212 | if (type == OBJ_REQUEST_BIO) { |
| 2213 | unsigned int clone_size; |
| 2214 | |
| 2215 | rbd_assert(length <= (u64)UINT_MAX); |
| 2216 | clone_size = (unsigned int)length; |
| 2217 | obj_request->bio_list = |
| 2218 | bio_chain_clone_range(&bio_list, |
| 2219 | &bio_offset, |
| 2220 | clone_size, |
| 2221 | GFP_ATOMIC); |
| 2222 | if (!obj_request->bio_list) |
| 2223 | goto out_partial; |
| 2224 | } else { |
| 2225 | unsigned int page_count; |
| 2226 | |
| 2227 | obj_request->pages = pages; |
| 2228 | page_count = (u32)calc_pages_for(offset, length); |
| 2229 | obj_request->page_count = page_count; |
| 2230 | if ((offset + length) & ~PAGE_MASK) |
| 2231 | page_count--; /* more on last page */ |
| 2232 | pages += page_count; |
| 2233 | } |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2234 | |
Alex Elder | 2fa1232 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2235 | osd_req = rbd_osd_req_create(rbd_dev, write_request, |
| 2236 | obj_request); |
| 2237 | if (!osd_req) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2238 | goto out_partial; |
Alex Elder | 2fa1232 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2239 | obj_request->osd_req = osd_req; |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2240 | obj_request->callback = rbd_img_obj_callback; |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2241 | |
Alex Elder | 2fa1232 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2242 | osd_req_op_extent_init(osd_req, 0, opcode, offset, length, |
| 2243 | 0, 0); |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2244 | if (type == OBJ_REQUEST_BIO) |
| 2245 | osd_req_op_extent_osd_data_bio(osd_req, 0, |
| 2246 | obj_request->bio_list, length); |
| 2247 | else |
| 2248 | osd_req_op_extent_osd_data_pages(osd_req, 0, |
| 2249 | obj_request->pages, length, |
| 2250 | offset & ~PAGE_MASK, false, false); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2251 | |
| 2252 | if (write_request) |
| 2253 | rbd_osd_req_format_write(obj_request); |
| 2254 | else |
| 2255 | rbd_osd_req_format_read(obj_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2256 | |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2257 | obj_request->img_offset = img_offset; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2258 | |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2259 | img_offset += length; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2260 | resid -= length; |
| 2261 | } |
| 2262 | |
| 2263 | return 0; |
| 2264 | |
| 2265 | out_partial: |
| 2266 | rbd_obj_request_put(obj_request); |
| 2267 | out_unwind: |
| 2268 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) |
| 2269 | rbd_obj_request_put(obj_request); |
| 2270 | |
| 2271 | return -ENOMEM; |
| 2272 | } |
| 2273 | |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2274 | static void |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2275 | rbd_img_obj_copyup_callback(struct rbd_obj_request *obj_request) |
| 2276 | { |
| 2277 | struct rbd_img_request *img_request; |
| 2278 | struct rbd_device *rbd_dev; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2279 | struct page **pages; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2280 | u32 page_count; |
| 2281 | |
| 2282 | rbd_assert(obj_request->type == OBJ_REQUEST_BIO); |
| 2283 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 2284 | img_request = obj_request->img_request; |
| 2285 | rbd_assert(img_request); |
| 2286 | |
| 2287 | rbd_dev = img_request->rbd_dev; |
| 2288 | rbd_assert(rbd_dev); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2289 | |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2290 | pages = obj_request->copyup_pages; |
| 2291 | rbd_assert(pages != NULL); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2292 | obj_request->copyup_pages = NULL; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2293 | page_count = obj_request->copyup_page_count; |
| 2294 | rbd_assert(page_count); |
| 2295 | obj_request->copyup_page_count = 0; |
| 2296 | ceph_release_page_vector(pages, page_count); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2297 | |
| 2298 | /* |
| 2299 | * We want the transfer count to reflect the size of the |
| 2300 | * original write request. There is no such thing as a |
| 2301 | * successful short write, so if the request was successful |
| 2302 | * we can just set it to the originally-requested length. |
| 2303 | */ |
| 2304 | if (!obj_request->result) |
| 2305 | obj_request->xferred = obj_request->length; |
| 2306 | |
| 2307 | /* Finish up with the normal image object callback */ |
| 2308 | |
| 2309 | rbd_img_obj_callback(obj_request); |
| 2310 | } |
| 2311 | |
| 2312 | static void |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2313 | rbd_img_obj_parent_read_full_callback(struct rbd_img_request *img_request) |
| 2314 | { |
| 2315 | struct rbd_obj_request *orig_request; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2316 | struct ceph_osd_request *osd_req; |
| 2317 | struct ceph_osd_client *osdc; |
| 2318 | struct rbd_device *rbd_dev; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2319 | struct page **pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2320 | u32 page_count; |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2321 | int img_result; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2322 | u64 parent_length; |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2323 | u64 offset; |
| 2324 | u64 length; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2325 | |
| 2326 | rbd_assert(img_request_child_test(img_request)); |
| 2327 | |
| 2328 | /* First get what we need from the image request */ |
| 2329 | |
| 2330 | pages = img_request->copyup_pages; |
| 2331 | rbd_assert(pages != NULL); |
| 2332 | img_request->copyup_pages = NULL; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2333 | page_count = img_request->copyup_page_count; |
| 2334 | rbd_assert(page_count); |
| 2335 | img_request->copyup_page_count = 0; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2336 | |
| 2337 | orig_request = img_request->obj_request; |
| 2338 | rbd_assert(orig_request != NULL); |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2339 | rbd_assert(obj_request_type_valid(orig_request->type)); |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2340 | img_result = img_request->result; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2341 | parent_length = img_request->length; |
| 2342 | rbd_assert(parent_length == img_request->xferred); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2343 | rbd_img_request_put(img_request); |
| 2344 | |
Alex Elder | 91c6feb | 2013-05-06 17:40:32 -0500 | [diff] [blame] | 2345 | rbd_assert(orig_request->img_request); |
| 2346 | rbd_dev = orig_request->img_request->rbd_dev; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2347 | rbd_assert(rbd_dev); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2348 | |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2349 | /* |
| 2350 | * If the overlap has become 0 (most likely because the |
| 2351 | * image has been flattened) we need to free the pages |
| 2352 | * and re-submit the original write request. |
| 2353 | */ |
| 2354 | if (!rbd_dev->parent_overlap) { |
| 2355 | struct ceph_osd_client *osdc; |
| 2356 | |
| 2357 | ceph_release_page_vector(pages, page_count); |
| 2358 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2359 | img_result = rbd_obj_request_submit(osdc, orig_request); |
| 2360 | if (!img_result) |
| 2361 | return; |
| 2362 | } |
| 2363 | |
| 2364 | if (img_result) |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2365 | goto out_err; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2366 | |
Alex Elder | 8785b1d | 2013-05-09 10:08:49 -0500 | [diff] [blame] | 2367 | /* |
| 2368 | * The original osd request is of no use to use any more. |
| 2369 | * We need a new one that can hold the two ops in a copyup |
| 2370 | * request. Allocate the new copyup osd request for the |
| 2371 | * original request, and release the old one. |
| 2372 | */ |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2373 | img_result = -ENOMEM; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2374 | osd_req = rbd_osd_req_create_copyup(orig_request); |
| 2375 | if (!osd_req) |
| 2376 | goto out_err; |
Alex Elder | 8785b1d | 2013-05-09 10:08:49 -0500 | [diff] [blame] | 2377 | rbd_osd_req_destroy(orig_request->osd_req); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2378 | orig_request->osd_req = osd_req; |
| 2379 | orig_request->copyup_pages = pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2380 | orig_request->copyup_page_count = page_count; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2381 | |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2382 | /* Initialize the copyup op */ |
| 2383 | |
| 2384 | 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] | 2385 | 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] | 2386 | false, false); |
| 2387 | |
| 2388 | /* Then the original write request op */ |
| 2389 | |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2390 | offset = orig_request->offset; |
| 2391 | length = orig_request->length; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2392 | osd_req_op_extent_init(osd_req, 1, CEPH_OSD_OP_WRITE, |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2393 | offset, length, 0, 0); |
| 2394 | if (orig_request->type == OBJ_REQUEST_BIO) |
| 2395 | osd_req_op_extent_osd_data_bio(osd_req, 1, |
| 2396 | orig_request->bio_list, length); |
| 2397 | else |
| 2398 | osd_req_op_extent_osd_data_pages(osd_req, 1, |
| 2399 | orig_request->pages, length, |
| 2400 | offset & ~PAGE_MASK, false, false); |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2401 | |
| 2402 | rbd_osd_req_format_write(orig_request); |
| 2403 | |
| 2404 | /* All set, send it off. */ |
| 2405 | |
| 2406 | orig_request->callback = rbd_img_obj_copyup_callback; |
| 2407 | osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2408 | img_result = rbd_obj_request_submit(osdc, orig_request); |
| 2409 | if (!img_result) |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2410 | return; |
| 2411 | out_err: |
| 2412 | /* Record the error code and complete the request */ |
| 2413 | |
Alex Elder | bbea1c1 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2414 | orig_request->result = img_result; |
Alex Elder | 0eefd47 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2415 | orig_request->xferred = 0; |
| 2416 | obj_request_done_set(orig_request); |
| 2417 | rbd_obj_request_complete(orig_request); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2418 | } |
| 2419 | |
| 2420 | /* |
| 2421 | * Read from the parent image the range of data that covers the |
| 2422 | * entire target of the given object request. This is used for |
| 2423 | * satisfying a layered image write request when the target of an |
| 2424 | * object request from the image request does not exist. |
| 2425 | * |
| 2426 | * A page array big enough to hold the returned data is allocated |
| 2427 | * and supplied to rbd_img_request_fill() as the "data descriptor." |
| 2428 | * When the read completes, this page array will be transferred to |
| 2429 | * the original object request for the copyup operation. |
| 2430 | * |
| 2431 | * If an error occurs, record it as the result of the original |
| 2432 | * object request and mark it done so it gets completed. |
| 2433 | */ |
| 2434 | static int rbd_img_obj_parent_read_full(struct rbd_obj_request *obj_request) |
| 2435 | { |
| 2436 | struct rbd_img_request *img_request = NULL; |
| 2437 | struct rbd_img_request *parent_request = NULL; |
| 2438 | struct rbd_device *rbd_dev; |
| 2439 | u64 img_offset; |
| 2440 | u64 length; |
| 2441 | struct page **pages = NULL; |
| 2442 | u32 page_count; |
| 2443 | int result; |
| 2444 | |
| 2445 | rbd_assert(obj_request_img_data_test(obj_request)); |
Alex Elder | b91f09f | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2446 | rbd_assert(obj_request_type_valid(obj_request->type)); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2447 | |
| 2448 | img_request = obj_request->img_request; |
| 2449 | rbd_assert(img_request != NULL); |
| 2450 | rbd_dev = img_request->rbd_dev; |
| 2451 | rbd_assert(rbd_dev->parent != NULL); |
| 2452 | |
| 2453 | /* |
| 2454 | * Determine the byte range covered by the object in the |
| 2455 | * child image to which the original request was to be sent. |
| 2456 | */ |
| 2457 | img_offset = obj_request->img_offset - obj_request->offset; |
| 2458 | length = (u64)1 << rbd_dev->header.obj_order; |
| 2459 | |
| 2460 | /* |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2461 | * There is no defined parent data beyond the parent |
| 2462 | * overlap, so limit what we read at that boundary if |
| 2463 | * necessary. |
| 2464 | */ |
| 2465 | if (img_offset + length > rbd_dev->parent_overlap) { |
| 2466 | rbd_assert(img_offset < rbd_dev->parent_overlap); |
| 2467 | length = rbd_dev->parent_overlap - img_offset; |
| 2468 | } |
| 2469 | |
| 2470 | /* |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2471 | * Allocate a page array big enough to receive the data read |
| 2472 | * from the parent. |
| 2473 | */ |
| 2474 | page_count = (u32)calc_pages_for(0, length); |
| 2475 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 2476 | if (IS_ERR(pages)) { |
| 2477 | result = PTR_ERR(pages); |
| 2478 | pages = NULL; |
| 2479 | goto out_err; |
| 2480 | } |
| 2481 | |
| 2482 | result = -ENOMEM; |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2483 | parent_request = rbd_parent_request_create(obj_request, |
| 2484 | img_offset, length); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2485 | if (!parent_request) |
| 2486 | goto out_err; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2487 | |
| 2488 | result = rbd_img_request_fill(parent_request, OBJ_REQUEST_PAGES, pages); |
| 2489 | if (result) |
| 2490 | goto out_err; |
| 2491 | parent_request->copyup_pages = pages; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2492 | parent_request->copyup_page_count = page_count; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2493 | |
| 2494 | parent_request->callback = rbd_img_obj_parent_read_full_callback; |
| 2495 | result = rbd_img_request_submit(parent_request); |
| 2496 | if (!result) |
| 2497 | return 0; |
| 2498 | |
| 2499 | parent_request->copyup_pages = NULL; |
Alex Elder | ebda640 | 2013-05-10 16:29:22 -0500 | [diff] [blame] | 2500 | parent_request->copyup_page_count = 0; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2501 | parent_request->obj_request = NULL; |
| 2502 | rbd_obj_request_put(obj_request); |
| 2503 | out_err: |
| 2504 | if (pages) |
| 2505 | ceph_release_page_vector(pages, page_count); |
| 2506 | if (parent_request) |
| 2507 | rbd_img_request_put(parent_request); |
| 2508 | obj_request->result = result; |
| 2509 | obj_request->xferred = 0; |
| 2510 | obj_request_done_set(obj_request); |
| 2511 | |
| 2512 | return result; |
| 2513 | } |
| 2514 | |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2515 | static void rbd_img_obj_exists_callback(struct rbd_obj_request *obj_request) |
| 2516 | { |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2517 | struct rbd_obj_request *orig_request; |
Alex Elder | 638f5ab | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2518 | struct rbd_device *rbd_dev; |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2519 | int result; |
| 2520 | |
| 2521 | rbd_assert(!obj_request_img_data_test(obj_request)); |
| 2522 | |
| 2523 | /* |
| 2524 | * All we need from the object request is the original |
| 2525 | * request and the result of the STAT op. Grab those, then |
| 2526 | * we're done with the request. |
| 2527 | */ |
| 2528 | orig_request = obj_request->obj_request; |
| 2529 | obj_request->obj_request = NULL; |
Alex Elder | 912c317 | 2013-05-13 20:35:38 -0500 | [diff] [blame] | 2530 | rbd_obj_request_put(orig_request); |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2531 | rbd_assert(orig_request); |
| 2532 | rbd_assert(orig_request->img_request); |
| 2533 | |
| 2534 | result = obj_request->result; |
| 2535 | obj_request->result = 0; |
| 2536 | |
| 2537 | dout("%s: obj %p for obj %p result %d %llu/%llu\n", __func__, |
| 2538 | obj_request, orig_request, result, |
| 2539 | obj_request->xferred, obj_request->length); |
| 2540 | rbd_obj_request_put(obj_request); |
| 2541 | |
Alex Elder | 638f5ab | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2542 | /* |
| 2543 | * If the overlap has become 0 (most likely because the |
| 2544 | * image has been flattened) we need to free the pages |
| 2545 | * and re-submit the original write request. |
| 2546 | */ |
| 2547 | rbd_dev = orig_request->img_request->rbd_dev; |
| 2548 | if (!rbd_dev->parent_overlap) { |
| 2549 | struct ceph_osd_client *osdc; |
| 2550 | |
Alex Elder | 638f5ab | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2551 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2552 | result = rbd_obj_request_submit(osdc, orig_request); |
| 2553 | if (!result) |
| 2554 | return; |
| 2555 | } |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2556 | |
| 2557 | /* |
| 2558 | * Our only purpose here is to determine whether the object |
| 2559 | * exists, and we don't want to treat the non-existence as |
| 2560 | * an error. If something else comes back, transfer the |
| 2561 | * error to the original request and complete it now. |
| 2562 | */ |
| 2563 | if (!result) { |
| 2564 | obj_request_existence_set(orig_request, true); |
| 2565 | } else if (result == -ENOENT) { |
| 2566 | obj_request_existence_set(orig_request, false); |
| 2567 | } else if (result) { |
| 2568 | orig_request->result = result; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2569 | goto out; |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2570 | } |
| 2571 | |
| 2572 | /* |
| 2573 | * Resubmit the original request now that we have recorded |
| 2574 | * whether the target object exists. |
| 2575 | */ |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2576 | orig_request->result = rbd_img_obj_request_submit(orig_request); |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2577 | out: |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2578 | if (orig_request->result) |
| 2579 | rbd_obj_request_complete(orig_request); |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2580 | } |
| 2581 | |
| 2582 | static int rbd_img_obj_exists_submit(struct rbd_obj_request *obj_request) |
| 2583 | { |
| 2584 | struct rbd_obj_request *stat_request; |
| 2585 | struct rbd_device *rbd_dev; |
| 2586 | struct ceph_osd_client *osdc; |
| 2587 | struct page **pages = NULL; |
| 2588 | u32 page_count; |
| 2589 | size_t size; |
| 2590 | int ret; |
| 2591 | |
| 2592 | /* |
| 2593 | * The response data for a STAT call consists of: |
| 2594 | * le64 length; |
| 2595 | * struct { |
| 2596 | * le32 tv_sec; |
| 2597 | * le32 tv_nsec; |
| 2598 | * } mtime; |
| 2599 | */ |
| 2600 | size = sizeof (__le64) + sizeof (__le32) + sizeof (__le32); |
| 2601 | page_count = (u32)calc_pages_for(0, size); |
| 2602 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 2603 | if (IS_ERR(pages)) |
| 2604 | return PTR_ERR(pages); |
| 2605 | |
| 2606 | ret = -ENOMEM; |
| 2607 | stat_request = rbd_obj_request_create(obj_request->object_name, 0, 0, |
| 2608 | OBJ_REQUEST_PAGES); |
| 2609 | if (!stat_request) |
| 2610 | goto out; |
| 2611 | |
| 2612 | rbd_obj_request_get(obj_request); |
| 2613 | stat_request->obj_request = obj_request; |
| 2614 | stat_request->pages = pages; |
| 2615 | stat_request->page_count = page_count; |
| 2616 | |
| 2617 | rbd_assert(obj_request->img_request); |
| 2618 | rbd_dev = obj_request->img_request->rbd_dev; |
| 2619 | stat_request->osd_req = rbd_osd_req_create(rbd_dev, false, |
| 2620 | stat_request); |
| 2621 | if (!stat_request->osd_req) |
| 2622 | goto out; |
| 2623 | stat_request->callback = rbd_img_obj_exists_callback; |
| 2624 | |
| 2625 | osd_req_op_init(stat_request->osd_req, 0, CEPH_OSD_OP_STAT); |
| 2626 | osd_req_op_raw_data_in_pages(stat_request->osd_req, 0, pages, size, 0, |
| 2627 | false, false); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2628 | rbd_osd_req_format_read(stat_request); |
Alex Elder | c5b5ef6 | 2013-02-11 12:33:24 -0600 | [diff] [blame] | 2629 | |
| 2630 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2631 | ret = rbd_obj_request_submit(osdc, stat_request); |
| 2632 | out: |
| 2633 | if (ret) |
| 2634 | rbd_obj_request_put(obj_request); |
| 2635 | |
| 2636 | return ret; |
| 2637 | } |
| 2638 | |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2639 | static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request) |
| 2640 | { |
| 2641 | struct rbd_img_request *img_request; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2642 | struct rbd_device *rbd_dev; |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2643 | bool known; |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2644 | |
| 2645 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 2646 | |
| 2647 | img_request = obj_request->img_request; |
| 2648 | rbd_assert(img_request); |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2649 | rbd_dev = img_request->rbd_dev; |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2650 | |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2651 | /* |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2652 | * Only writes to layered images need special handling. |
| 2653 | * Reads and non-layered writes are simple object requests. |
| 2654 | * Layered writes that start beyond the end of the overlap |
| 2655 | * with the parent have no parent data, so they too are |
| 2656 | * simple object requests. Finally, if the target object is |
| 2657 | * known to already exist, its parent data has already been |
| 2658 | * copied, so a write to the object can also be handled as a |
| 2659 | * simple object request. |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2660 | */ |
| 2661 | if (!img_request_write_test(img_request) || |
| 2662 | !img_request_layered_test(img_request) || |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2663 | rbd_dev->parent_overlap <= obj_request->img_offset || |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2664 | ((known = obj_request_known_test(obj_request)) && |
| 2665 | obj_request_exists_test(obj_request))) { |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2666 | |
| 2667 | struct rbd_device *rbd_dev; |
| 2668 | struct ceph_osd_client *osdc; |
| 2669 | |
| 2670 | rbd_dev = obj_request->img_request->rbd_dev; |
| 2671 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2672 | |
| 2673 | return rbd_obj_request_submit(osdc, obj_request); |
| 2674 | } |
| 2675 | |
| 2676 | /* |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2677 | * It's a layered write. The target object might exist but |
| 2678 | * we may not know that yet. If we know it doesn't exist, |
| 2679 | * start by reading the data for the full target object from |
| 2680 | * 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] | 2681 | */ |
Alex Elder | 3d7efd1 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2682 | if (known) |
| 2683 | return rbd_img_obj_parent_read_full(obj_request); |
| 2684 | |
| 2685 | /* We don't know whether the target exists. Go find out. */ |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2686 | |
| 2687 | return rbd_img_obj_exists_submit(obj_request); |
| 2688 | } |
| 2689 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2690 | static int rbd_img_request_submit(struct rbd_img_request *img_request) |
| 2691 | { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2692 | struct rbd_obj_request *obj_request; |
Alex Elder | 46faeed | 2013-04-10 17:47:46 -0500 | [diff] [blame] | 2693 | struct rbd_obj_request *next_obj_request; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2694 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2695 | dout("%s: img %p\n", __func__, img_request); |
Alex Elder | 46faeed | 2013-04-10 17:47:46 -0500 | [diff] [blame] | 2696 | for_each_obj_request_safe(img_request, obj_request, next_obj_request) { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2697 | int ret; |
| 2698 | |
Alex Elder | b454e36 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2699 | ret = rbd_img_obj_request_submit(obj_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2700 | if (ret) |
| 2701 | return ret; |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 2702 | } |
| 2703 | |
| 2704 | return 0; |
| 2705 | } |
| 2706 | |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2707 | static void rbd_img_parent_read_callback(struct rbd_img_request *img_request) |
| 2708 | { |
| 2709 | struct rbd_obj_request *obj_request; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2710 | struct rbd_device *rbd_dev; |
| 2711 | u64 obj_end; |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2712 | u64 img_xferred; |
| 2713 | int img_result; |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2714 | |
| 2715 | rbd_assert(img_request_child_test(img_request)); |
| 2716 | |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2717 | /* First get what we need from the image request and release it */ |
| 2718 | |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2719 | obj_request = img_request->obj_request; |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2720 | img_xferred = img_request->xferred; |
| 2721 | img_result = img_request->result; |
| 2722 | rbd_img_request_put(img_request); |
| 2723 | |
| 2724 | /* |
| 2725 | * If the overlap has become 0 (most likely because the |
| 2726 | * image has been flattened) we need to re-submit the |
| 2727 | * original request. |
| 2728 | */ |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2729 | rbd_assert(obj_request); |
| 2730 | rbd_assert(obj_request->img_request); |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2731 | rbd_dev = obj_request->img_request->rbd_dev; |
| 2732 | if (!rbd_dev->parent_overlap) { |
| 2733 | struct ceph_osd_client *osdc; |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2734 | |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2735 | osdc = &rbd_dev->rbd_client->client->osdc; |
| 2736 | img_result = rbd_obj_request_submit(osdc, obj_request); |
| 2737 | if (!img_result) |
| 2738 | return; |
| 2739 | } |
| 2740 | |
| 2741 | obj_request->result = img_result; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2742 | if (obj_request->result) |
| 2743 | goto out; |
| 2744 | |
| 2745 | /* |
| 2746 | * We need to zero anything beyond the parent overlap |
| 2747 | * boundary. Since rbd_img_obj_request_read_callback() |
| 2748 | * will zero anything beyond the end of a short read, an |
| 2749 | * easy way to do this is to pretend the data from the |
| 2750 | * parent came up short--ending at the overlap boundary. |
| 2751 | */ |
| 2752 | rbd_assert(obj_request->img_offset < U64_MAX - obj_request->length); |
| 2753 | obj_end = obj_request->img_offset + obj_request->length; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2754 | if (obj_end > rbd_dev->parent_overlap) { |
| 2755 | u64 xferred = 0; |
| 2756 | |
| 2757 | if (obj_request->img_offset < rbd_dev->parent_overlap) |
| 2758 | xferred = rbd_dev->parent_overlap - |
| 2759 | obj_request->img_offset; |
| 2760 | |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2761 | obj_request->xferred = min(img_xferred, xferred); |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2762 | } else { |
Alex Elder | 02c74fb | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2763 | obj_request->xferred = img_xferred; |
Alex Elder | a9e8ba2c | 2013-04-21 00:32:07 -0500 | [diff] [blame] | 2764 | } |
| 2765 | out: |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2766 | rbd_img_obj_request_read_callback(obj_request); |
| 2767 | rbd_obj_request_complete(obj_request); |
| 2768 | } |
| 2769 | |
| 2770 | static void rbd_img_parent_read(struct rbd_obj_request *obj_request) |
| 2771 | { |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2772 | struct rbd_img_request *img_request; |
| 2773 | int result; |
| 2774 | |
| 2775 | rbd_assert(obj_request_img_data_test(obj_request)); |
| 2776 | rbd_assert(obj_request->img_request != NULL); |
| 2777 | rbd_assert(obj_request->result == (s32) -ENOENT); |
Alex Elder | 5b2ab72 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2778 | rbd_assert(obj_request_type_valid(obj_request->type)); |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2779 | |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2780 | /* rbd_read_finish(obj_request, obj_request->length); */ |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2781 | img_request = rbd_parent_request_create(obj_request, |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2782 | obj_request->img_offset, |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 2783 | obj_request->length); |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2784 | result = -ENOMEM; |
| 2785 | if (!img_request) |
| 2786 | goto out_err; |
| 2787 | |
Alex Elder | 5b2ab72 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2788 | if (obj_request->type == OBJ_REQUEST_BIO) |
| 2789 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO, |
| 2790 | obj_request->bio_list); |
| 2791 | else |
| 2792 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_PAGES, |
| 2793 | obj_request->pages); |
Alex Elder | 8b3e1a5 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 2794 | if (result) |
| 2795 | goto out_err; |
| 2796 | |
| 2797 | img_request->callback = rbd_img_parent_read_callback; |
| 2798 | result = rbd_img_request_submit(img_request); |
| 2799 | if (result) |
| 2800 | goto out_err; |
| 2801 | |
| 2802 | return; |
| 2803 | out_err: |
| 2804 | if (img_request) |
| 2805 | rbd_img_request_put(img_request); |
| 2806 | obj_request->result = result; |
| 2807 | obj_request->xferred = 0; |
| 2808 | obj_request_done_set(obj_request); |
| 2809 | } |
| 2810 | |
Josh Durgin | 20e0af6 | 2013-08-29 17:36:03 -0700 | [diff] [blame^] | 2811 | 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] | 2812 | { |
| 2813 | struct rbd_obj_request *obj_request; |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2814 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2815 | int ret; |
| 2816 | |
| 2817 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, |
| 2818 | OBJ_REQUEST_NODATA); |
| 2819 | if (!obj_request) |
| 2820 | return -ENOMEM; |
| 2821 | |
| 2822 | ret = -ENOMEM; |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2823 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2824 | if (!obj_request->osd_req) |
| 2825 | goto out; |
| 2826 | |
Alex Elder | c99d2d4 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 2827 | 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] | 2828 | notify_id, 0, 0); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2829 | rbd_osd_req_format_read(obj_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2830 | |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2831 | ret = rbd_obj_request_submit(osdc, obj_request); |
Alex Elder | cf81b60 | 2013-01-17 12:18:46 -0600 | [diff] [blame] | 2832 | if (ret) |
Josh Durgin | 20e0af6 | 2013-08-29 17:36:03 -0700 | [diff] [blame^] | 2833 | goto out; |
| 2834 | ret = rbd_obj_request_wait(obj_request); |
| 2835 | out: |
| 2836 | rbd_obj_request_put(obj_request); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2837 | |
| 2838 | return ret; |
| 2839 | } |
| 2840 | |
| 2841 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) |
| 2842 | { |
| 2843 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
Alex Elder | e627db0 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 2844 | int ret; |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2845 | |
| 2846 | if (!rbd_dev) |
| 2847 | return; |
| 2848 | |
Alex Elder | 37206ee | 2013-02-20 17:32:08 -0600 | [diff] [blame] | 2849 | dout("%s: \"%s\" notify_id %llu opcode %u\n", __func__, |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 2850 | rbd_dev->header_name, (unsigned long long)notify_id, |
| 2851 | (unsigned int)opcode); |
Alex Elder | e627db0 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 2852 | ret = rbd_dev_refresh(rbd_dev); |
| 2853 | if (ret) |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 2854 | rbd_warn(rbd_dev, "header refresh error (%d)\n", ret); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2855 | |
Josh Durgin | 20e0af6 | 2013-08-29 17:36:03 -0700 | [diff] [blame^] | 2856 | rbd_obj_notify_ack_sync(rbd_dev, notify_id); |
Alex Elder | b8d7003 | 2012-11-30 17:53:04 -0600 | [diff] [blame] | 2857 | } |
| 2858 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2859 | /* |
| 2860 | * Request sync osd watch/unwatch. The value of "start" determines |
| 2861 | * whether a watch request is being initiated or torn down. |
| 2862 | */ |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2863 | static int rbd_dev_header_watch_sync(struct rbd_device *rbd_dev, bool start) |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2864 | { |
| 2865 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
| 2866 | struct rbd_obj_request *obj_request; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2867 | int ret; |
| 2868 | |
| 2869 | rbd_assert(start ^ !!rbd_dev->watch_event); |
| 2870 | rbd_assert(start ^ !!rbd_dev->watch_request); |
| 2871 | |
| 2872 | if (start) { |
Alex Elder | 3c663bb | 2013-02-15 11:42:30 -0600 | [diff] [blame] | 2873 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, rbd_dev, |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2874 | &rbd_dev->watch_event); |
| 2875 | if (ret < 0) |
| 2876 | return ret; |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2877 | rbd_assert(rbd_dev->watch_event != NULL); |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2878 | } |
| 2879 | |
| 2880 | ret = -ENOMEM; |
| 2881 | obj_request = rbd_obj_request_create(rbd_dev->header_name, 0, 0, |
| 2882 | OBJ_REQUEST_NODATA); |
| 2883 | if (!obj_request) |
| 2884 | goto out_cancel; |
| 2885 | |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2886 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, true, obj_request); |
| 2887 | if (!obj_request->osd_req) |
| 2888 | goto out_cancel; |
| 2889 | |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2890 | if (start) |
Alex Elder | 975241a | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2891 | ceph_osdc_set_request_linger(osdc, obj_request->osd_req); |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2892 | else |
Alex Elder | 6977c3f | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2893 | ceph_osdc_unregister_linger_request(osdc, |
Alex Elder | 975241a | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2894 | rbd_dev->watch_request->osd_req); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2895 | |
| 2896 | osd_req_op_watch_init(obj_request->osd_req, 0, CEPH_OSD_OP_WATCH, |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 2897 | rbd_dev->watch_event->cookie, 0, start ? 1 : 0); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2898 | rbd_osd_req_format_write(obj_request); |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2899 | |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2900 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 2901 | if (ret) |
| 2902 | goto out_cancel; |
| 2903 | ret = rbd_obj_request_wait(obj_request); |
| 2904 | if (ret) |
| 2905 | goto out_cancel; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2906 | ret = obj_request->result; |
| 2907 | if (ret) |
| 2908 | goto out_cancel; |
| 2909 | |
Alex Elder | 8eb8756 | 2013-01-25 17:08:55 -0600 | [diff] [blame] | 2910 | /* |
| 2911 | * A watch request is set to linger, so the underlying osd |
| 2912 | * request won't go away until we unregister it. We retain |
| 2913 | * a pointer to the object request during that time (in |
| 2914 | * rbd_dev->watch_request), so we'll keep a reference to |
| 2915 | * it. We'll drop that reference (below) after we've |
| 2916 | * unregistered it. |
| 2917 | */ |
| 2918 | if (start) { |
| 2919 | rbd_dev->watch_request = obj_request; |
| 2920 | |
| 2921 | return 0; |
| 2922 | } |
| 2923 | |
| 2924 | /* We have successfully torn down the watch request */ |
| 2925 | |
| 2926 | rbd_obj_request_put(rbd_dev->watch_request); |
| 2927 | rbd_dev->watch_request = NULL; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2928 | out_cancel: |
| 2929 | /* Cancel the event if we're tearing down, or on error */ |
| 2930 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
| 2931 | rbd_dev->watch_event = NULL; |
Alex Elder | 9969ebc | 2013-01-18 12:31:10 -0600 | [diff] [blame] | 2932 | if (obj_request) |
| 2933 | rbd_obj_request_put(obj_request); |
| 2934 | |
| 2935 | return ret; |
| 2936 | } |
| 2937 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2938 | /* |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 2939 | * Synchronous osd object method call. Returns the number of bytes |
| 2940 | * returned in the outbound buffer, or a negative error code. |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2941 | */ |
| 2942 | static int rbd_obj_method_sync(struct rbd_device *rbd_dev, |
| 2943 | const char *object_name, |
| 2944 | const char *class_name, |
| 2945 | const char *method_name, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 2946 | const void *outbound, |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2947 | size_t outbound_size, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 2948 | void *inbound, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 2949 | size_t inbound_size) |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2950 | { |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2951 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2952 | struct rbd_obj_request *obj_request; |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2953 | struct page **pages; |
| 2954 | u32 page_count; |
| 2955 | int ret; |
| 2956 | |
| 2957 | /* |
Alex Elder | 6010a45 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 2958 | * Method calls are ultimately read operations. The result |
| 2959 | * should placed into the inbound buffer provided. They |
| 2960 | * also supply outbound data--parameters for the object |
| 2961 | * method. Currently if this is present it will be a |
| 2962 | * snapshot id. |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2963 | */ |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 2964 | page_count = (u32)calc_pages_for(0, inbound_size); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2965 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 2966 | if (IS_ERR(pages)) |
| 2967 | return PTR_ERR(pages); |
| 2968 | |
| 2969 | ret = -ENOMEM; |
Alex Elder | 6010a45 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 2970 | obj_request = rbd_obj_request_create(object_name, 0, inbound_size, |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2971 | OBJ_REQUEST_PAGES); |
| 2972 | if (!obj_request) |
| 2973 | goto out; |
| 2974 | |
| 2975 | obj_request->pages = pages; |
| 2976 | obj_request->page_count = page_count; |
| 2977 | |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 2978 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 2979 | if (!obj_request->osd_req) |
| 2980 | goto out; |
| 2981 | |
Alex Elder | c99d2d4 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 2982 | 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] | 2983 | class_name, method_name); |
| 2984 | if (outbound_size) { |
| 2985 | struct ceph_pagelist *pagelist; |
| 2986 | |
| 2987 | pagelist = kmalloc(sizeof (*pagelist), GFP_NOFS); |
| 2988 | if (!pagelist) |
| 2989 | goto out; |
| 2990 | |
| 2991 | ceph_pagelist_init(pagelist); |
| 2992 | ceph_pagelist_append(pagelist, outbound, outbound_size); |
| 2993 | osd_req_op_cls_request_data_pagelist(obj_request->osd_req, 0, |
| 2994 | pagelist); |
| 2995 | } |
Alex Elder | a4ce40a | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2996 | osd_req_op_cls_response_data_pages(obj_request->osd_req, 0, |
| 2997 | obj_request->pages, inbound_size, |
Alex Elder | 44cd188 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 2998 | 0, false, false); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 2999 | rbd_osd_req_format_read(obj_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 3000 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3001 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 3002 | if (ret) |
| 3003 | goto out; |
| 3004 | ret = rbd_obj_request_wait(obj_request); |
| 3005 | if (ret) |
| 3006 | goto out; |
| 3007 | |
| 3008 | ret = obj_request->result; |
| 3009 | if (ret < 0) |
| 3010 | goto out; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3011 | |
| 3012 | rbd_assert(obj_request->xferred < (u64)INT_MAX); |
| 3013 | ret = (int)obj_request->xferred; |
Alex Elder | 903bb32 | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 3014 | ceph_copy_from_page_vector(pages, inbound, 0, obj_request->xferred); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3015 | out: |
| 3016 | if (obj_request) |
| 3017 | rbd_obj_request_put(obj_request); |
| 3018 | else |
| 3019 | ceph_release_page_vector(pages, page_count); |
| 3020 | |
| 3021 | return ret; |
| 3022 | } |
| 3023 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3024 | static void rbd_request_fn(struct request_queue *q) |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 3025 | __releases(q->queue_lock) __acquires(q->queue_lock) |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3026 | { |
| 3027 | struct rbd_device *rbd_dev = q->queuedata; |
| 3028 | bool read_only = rbd_dev->mapping.read_only; |
| 3029 | struct request *rq; |
| 3030 | int result; |
| 3031 | |
| 3032 | while ((rq = blk_fetch_request(q))) { |
| 3033 | bool write_request = rq_data_dir(rq) == WRITE; |
| 3034 | struct rbd_img_request *img_request; |
| 3035 | u64 offset; |
| 3036 | u64 length; |
| 3037 | |
| 3038 | /* Ignore any non-FS requests that filter through. */ |
| 3039 | |
| 3040 | if (rq->cmd_type != REQ_TYPE_FS) { |
Alex Elder | 4dda41d | 2013-02-20 21:59:33 -0600 | [diff] [blame] | 3041 | dout("%s: non-fs request type %d\n", __func__, |
| 3042 | (int) rq->cmd_type); |
| 3043 | __blk_end_request_all(rq, 0); |
| 3044 | continue; |
| 3045 | } |
| 3046 | |
| 3047 | /* Ignore/skip any zero-length requests */ |
| 3048 | |
| 3049 | offset = (u64) blk_rq_pos(rq) << SECTOR_SHIFT; |
| 3050 | length = (u64) blk_rq_bytes(rq); |
| 3051 | |
| 3052 | if (!length) { |
| 3053 | dout("%s: zero-length request\n", __func__); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3054 | __blk_end_request_all(rq, 0); |
| 3055 | continue; |
| 3056 | } |
| 3057 | |
| 3058 | spin_unlock_irq(q->queue_lock); |
| 3059 | |
| 3060 | /* Disallow writes to a read-only device */ |
| 3061 | |
| 3062 | if (write_request) { |
| 3063 | result = -EROFS; |
| 3064 | if (read_only) |
| 3065 | goto end_request; |
| 3066 | rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP); |
| 3067 | } |
| 3068 | |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 3069 | /* |
| 3070 | * Quit early if the mapped snapshot no longer |
| 3071 | * exists. It's still possible the snapshot will |
| 3072 | * have disappeared by the time our request arrives |
| 3073 | * at the osd, but there's no sense in sending it if |
| 3074 | * we already know. |
| 3075 | */ |
| 3076 | if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) { |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3077 | dout("request for non-existent snapshot"); |
| 3078 | rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP); |
| 3079 | result = -ENXIO; |
| 3080 | goto end_request; |
| 3081 | } |
| 3082 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3083 | result = -EINVAL; |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3084 | if (offset && length > U64_MAX - offset + 1) { |
| 3085 | rbd_warn(rbd_dev, "bad request range (%llu~%llu)\n", |
| 3086 | offset, length); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3087 | goto end_request; /* Shouldn't happen */ |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3088 | } |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3089 | |
Alex Elder | 00a653e | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3090 | result = -EIO; |
| 3091 | if (offset + length > rbd_dev->mapping.size) { |
| 3092 | rbd_warn(rbd_dev, "beyond EOD (%llu~%llu > %llu)\n", |
| 3093 | offset, length, rbd_dev->mapping.size); |
| 3094 | goto end_request; |
| 3095 | } |
| 3096 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3097 | result = -ENOMEM; |
| 3098 | img_request = rbd_img_request_create(rbd_dev, offset, length, |
Alex Elder | e93f315 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 3099 | write_request); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3100 | if (!img_request) |
| 3101 | goto end_request; |
| 3102 | |
| 3103 | img_request->rq = rq; |
| 3104 | |
Alex Elder | f1a4739 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 3105 | result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO, |
| 3106 | rq->bio); |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3107 | if (!result) |
| 3108 | result = rbd_img_request_submit(img_request); |
| 3109 | if (result) |
| 3110 | rbd_img_request_put(img_request); |
| 3111 | end_request: |
| 3112 | spin_lock_irq(q->queue_lock); |
| 3113 | if (result < 0) { |
Alex Elder | 7da22d2 | 2013-01-24 16:13:36 -0600 | [diff] [blame] | 3114 | rbd_warn(rbd_dev, "%s %llx at %llx result %d\n", |
| 3115 | write_request ? "write" : "read", |
| 3116 | length, offset, result); |
| 3117 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3118 | __blk_end_request_all(rq, result); |
| 3119 | } |
| 3120 | } |
| 3121 | } |
| 3122 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3123 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3124 | * a queue callback. Makes sure that we don't create a bio that spans across |
| 3125 | * multiple osd objects. One exception would be with a single page bios, |
Alex Elder | f7760da | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 3126 | * which we handle later at bio_chain_clone_range() |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3127 | */ |
| 3128 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, |
| 3129 | struct bio_vec *bvec) |
| 3130 | { |
| 3131 | struct rbd_device *rbd_dev = q->queuedata; |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 3132 | sector_t sector_offset; |
| 3133 | sector_t sectors_per_obj; |
| 3134 | sector_t obj_sector_offset; |
| 3135 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3136 | |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 3137 | /* |
| 3138 | * Find how far into its rbd object the partition-relative |
| 3139 | * bio start sector is to offset relative to the enclosing |
| 3140 | * device. |
| 3141 | */ |
| 3142 | sector_offset = get_start_sect(bmd->bi_bdev) + bmd->bi_sector; |
| 3143 | sectors_per_obj = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); |
| 3144 | obj_sector_offset = sector_offset & (sectors_per_obj - 1); |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3145 | |
Alex Elder | e5cfeed2 | 2012-10-20 22:17:27 -0500 | [diff] [blame] | 3146 | /* |
| 3147 | * Compute the number of bytes from that offset to the end |
| 3148 | * of the object. Account for what's already used by the bio. |
| 3149 | */ |
| 3150 | ret = (int) (sectors_per_obj - obj_sector_offset) << SECTOR_SHIFT; |
| 3151 | if (ret > bmd->bi_size) |
| 3152 | ret -= bmd->bi_size; |
| 3153 | else |
| 3154 | ret = 0; |
| 3155 | |
| 3156 | /* |
| 3157 | * Don't send back more than was asked for. And if the bio |
| 3158 | * was empty, let the whole thing through because: "Note |
| 3159 | * that a block device *must* allow a single page to be |
| 3160 | * added to an empty bio." |
| 3161 | */ |
| 3162 | rbd_assert(bvec->bv_len <= PAGE_SIZE); |
| 3163 | if (ret > (int) bvec->bv_len || !bmd->bi_size) |
| 3164 | ret = (int) bvec->bv_len; |
| 3165 | |
| 3166 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3167 | } |
| 3168 | |
| 3169 | static void rbd_free_disk(struct rbd_device *rbd_dev) |
| 3170 | { |
| 3171 | struct gendisk *disk = rbd_dev->disk; |
| 3172 | |
| 3173 | if (!disk) |
| 3174 | return; |
| 3175 | |
Alex Elder | a0cab92 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 3176 | rbd_dev->disk = NULL; |
| 3177 | if (disk->flags & GENHD_FL_UP) { |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3178 | del_gendisk(disk); |
Alex Elder | a0cab92 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 3179 | if (disk->queue) |
| 3180 | blk_cleanup_queue(disk->queue); |
| 3181 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3182 | put_disk(disk); |
| 3183 | } |
| 3184 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3185 | static int rbd_obj_read_sync(struct rbd_device *rbd_dev, |
| 3186 | const char *object_name, |
Alex Elder | 7097f8d | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3187 | u64 offset, u64 length, void *buf) |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3188 | |
| 3189 | { |
Alex Elder | 2169238 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3190 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3191 | struct rbd_obj_request *obj_request; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3192 | struct page **pages = NULL; |
| 3193 | u32 page_count; |
Alex Elder | 1ceae7e | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 3194 | size_t size; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3195 | int ret; |
| 3196 | |
| 3197 | page_count = (u32) calc_pages_for(offset, length); |
| 3198 | pages = ceph_alloc_page_vector(page_count, GFP_KERNEL); |
| 3199 | if (IS_ERR(pages)) |
| 3200 | ret = PTR_ERR(pages); |
| 3201 | |
| 3202 | ret = -ENOMEM; |
| 3203 | obj_request = rbd_obj_request_create(object_name, offset, length, |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3204 | OBJ_REQUEST_PAGES); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3205 | if (!obj_request) |
| 3206 | goto out; |
| 3207 | |
| 3208 | obj_request->pages = pages; |
| 3209 | obj_request->page_count = page_count; |
| 3210 | |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 3211 | obj_request->osd_req = rbd_osd_req_create(rbd_dev, false, obj_request); |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3212 | if (!obj_request->osd_req) |
| 3213 | goto out; |
| 3214 | |
Alex Elder | c99d2d4 | 2013-04-05 01:27:11 -0500 | [diff] [blame] | 3215 | osd_req_op_extent_init(obj_request->osd_req, 0, CEPH_OSD_OP_READ, |
| 3216 | offset, length, 0, 0); |
Alex Elder | 406e2c9 | 2013-04-15 14:50:36 -0500 | [diff] [blame] | 3217 | osd_req_op_extent_osd_data_pages(obj_request->osd_req, 0, |
Alex Elder | a4ce40a | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3218 | obj_request->pages, |
Alex Elder | 44cd188 | 2013-04-05 01:27:12 -0500 | [diff] [blame] | 3219 | obj_request->length, |
| 3220 | obj_request->offset & ~PAGE_MASK, |
| 3221 | false, false); |
Alex Elder | 9d4df01 | 2013-04-19 15:34:50 -0500 | [diff] [blame] | 3222 | rbd_osd_req_format_read(obj_request); |
Alex Elder | 430c28c | 2013-04-03 21:32:51 -0500 | [diff] [blame] | 3223 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3224 | ret = rbd_obj_request_submit(osdc, obj_request); |
| 3225 | if (ret) |
| 3226 | goto out; |
| 3227 | ret = rbd_obj_request_wait(obj_request); |
| 3228 | if (ret) |
| 3229 | goto out; |
| 3230 | |
| 3231 | ret = obj_request->result; |
| 3232 | if (ret < 0) |
| 3233 | goto out; |
Alex Elder | 1ceae7e | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 3234 | |
| 3235 | rbd_assert(obj_request->xferred <= (u64) SIZE_MAX); |
| 3236 | size = (size_t) obj_request->xferred; |
Alex Elder | 903bb32 | 2013-02-06 13:11:38 -0600 | [diff] [blame] | 3237 | ceph_copy_from_page_vector(pages, buf, 0, size); |
Alex Elder | 7097f8d | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3238 | rbd_assert(size <= (size_t)INT_MAX); |
| 3239 | ret = (int)size; |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3240 | out: |
| 3241 | if (obj_request) |
| 3242 | rbd_obj_request_put(obj_request); |
| 3243 | else |
| 3244 | ceph_release_page_vector(pages, page_count); |
| 3245 | |
| 3246 | return ret; |
| 3247 | } |
| 3248 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3249 | /* |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3250 | * Read the complete header for the given rbd device. On successful |
| 3251 | * return, the rbd_dev->header field will contain up-to-date |
| 3252 | * information about the image. |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3253 | */ |
Alex Elder | 99a41eb | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 3254 | static int rbd_dev_v1_header_info(struct rbd_device *rbd_dev) |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3255 | { |
| 3256 | struct rbd_image_header_ondisk *ondisk = NULL; |
| 3257 | u32 snap_count = 0; |
| 3258 | u64 names_size = 0; |
| 3259 | u32 want_count; |
| 3260 | int ret; |
| 3261 | |
| 3262 | /* |
| 3263 | * The complete header will include an array of its 64-bit |
| 3264 | * snapshot ids, followed by the names of those snapshots as |
| 3265 | * a contiguous block of NUL-terminated strings. Note that |
| 3266 | * the number of snapshots could change by the time we read |
| 3267 | * it in, in which case we re-read it. |
| 3268 | */ |
| 3269 | do { |
| 3270 | size_t size; |
| 3271 | |
| 3272 | kfree(ondisk); |
| 3273 | |
| 3274 | size = sizeof (*ondisk); |
| 3275 | size += snap_count * sizeof (struct rbd_image_snap_ondisk); |
| 3276 | size += names_size; |
| 3277 | ondisk = kmalloc(size, GFP_KERNEL); |
| 3278 | if (!ondisk) |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3279 | return -ENOMEM; |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3280 | |
Alex Elder | 788e2df | 2013-01-17 12:25:27 -0600 | [diff] [blame] | 3281 | ret = rbd_obj_read_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 7097f8d | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3282 | 0, size, ondisk); |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3283 | if (ret < 0) |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3284 | goto out; |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3285 | if ((size_t)ret < size) { |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3286 | ret = -ENXIO; |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3287 | rbd_warn(rbd_dev, "short header read (want %zd got %d)", |
| 3288 | size, ret); |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3289 | goto out; |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3290 | } |
| 3291 | if (!rbd_dev_ondisk_valid(ondisk)) { |
| 3292 | ret = -ENXIO; |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 3293 | rbd_warn(rbd_dev, "invalid header"); |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3294 | goto out; |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3295 | } |
| 3296 | |
| 3297 | names_size = le64_to_cpu(ondisk->snap_names_len); |
| 3298 | want_count = snap_count; |
| 3299 | snap_count = le32_to_cpu(ondisk->snap_count); |
| 3300 | } while (snap_count != want_count); |
| 3301 | |
Alex Elder | 662518b | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 3302 | ret = rbd_header_from_disk(rbd_dev, ondisk); |
| 3303 | out: |
Alex Elder | 4156d99 | 2012-08-02 11:29:46 -0500 | [diff] [blame] | 3304 | kfree(ondisk); |
| 3305 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3306 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3307 | } |
| 3308 | |
Alex Elder | 15228ed | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 3309 | /* |
| 3310 | * Clear the rbd device's EXISTS flag if the snapshot it's mapped to |
| 3311 | * has disappeared from the (just updated) snapshot context. |
| 3312 | */ |
| 3313 | static void rbd_exists_validate(struct rbd_device *rbd_dev) |
| 3314 | { |
| 3315 | u64 snap_id; |
| 3316 | |
| 3317 | if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) |
| 3318 | return; |
| 3319 | |
| 3320 | snap_id = rbd_dev->spec->snap_id; |
| 3321 | if (snap_id == CEPH_NOSNAP) |
| 3322 | return; |
| 3323 | |
| 3324 | if (rbd_dev_snap_index(rbd_dev, snap_id) == BAD_SNAP_INDEX) |
| 3325 | clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
| 3326 | } |
| 3327 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3328 | static int rbd_dev_refresh(struct rbd_device *rbd_dev) |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3329 | { |
Alex Elder | e627db0 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3330 | u64 mapping_size; |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3331 | int ret; |
| 3332 | |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 3333 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 3334 | down_write(&rbd_dev->header_rwsem); |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 3335 | mapping_size = rbd_dev->mapping.size; |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 3336 | if (rbd_dev->image_format == 1) |
Alex Elder | 99a41eb | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 3337 | ret = rbd_dev_v1_header_info(rbd_dev); |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 3338 | else |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 3339 | ret = rbd_dev_v2_header_info(rbd_dev); |
Alex Elder | 15228ed | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 3340 | |
| 3341 | /* If it's a mapped snapshot, validate its EXISTS flag */ |
| 3342 | |
| 3343 | rbd_exists_validate(rbd_dev); |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 3344 | up_write(&rbd_dev->header_rwsem); |
| 3345 | |
Alex Elder | 00a653e | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3346 | if (mapping_size != rbd_dev->mapping.size) { |
| 3347 | sector_t size; |
| 3348 | |
| 3349 | size = (sector_t)rbd_dev->mapping.size / SECTOR_SIZE; |
| 3350 | dout("setting size to %llu sectors", (unsigned long long)size); |
| 3351 | set_capacity(rbd_dev->disk, size); |
Alex Elder | a3fbe5d | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 3352 | revalidate_disk(rbd_dev->disk); |
Alex Elder | 00a653e | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3353 | } |
Alex Elder | 1fe5e99 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3354 | |
| 3355 | return ret; |
| 3356 | } |
| 3357 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3358 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
| 3359 | { |
| 3360 | struct gendisk *disk; |
| 3361 | struct request_queue *q; |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3362 | u64 segment_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3363 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3364 | /* create gendisk info */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3365 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); |
| 3366 | if (!disk) |
Alex Elder | 1fcdb8a | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3367 | return -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3368 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 3369 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3370 | rbd_dev->dev_id); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3371 | disk->major = rbd_dev->major; |
| 3372 | disk->first_minor = 0; |
| 3373 | disk->fops = &rbd_bd_ops; |
| 3374 | disk->private_data = rbd_dev; |
| 3375 | |
Alex Elder | bf0d5f50 | 2012-11-22 00:00:08 -0600 | [diff] [blame] | 3376 | q = blk_init_queue(rbd_request_fn, &rbd_dev->lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3377 | if (!q) |
| 3378 | goto out_disk; |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 3379 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3380 | /* We use the default size, but let's be explicit about it. */ |
| 3381 | blk_queue_physical_block_size(q, SECTOR_SIZE); |
| 3382 | |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 3383 | /* set io sizes to object size */ |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3384 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
| 3385 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); |
| 3386 | blk_queue_max_segment_size(q, segment_size); |
| 3387 | blk_queue_io_min(q, segment_size); |
| 3388 | blk_queue_io_opt(q, segment_size); |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 3389 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3390 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
| 3391 | disk->queue = q; |
| 3392 | |
| 3393 | q->queuedata = rbd_dev; |
| 3394 | |
| 3395 | rbd_dev->disk = disk; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3396 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3397 | return 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3398 | out_disk: |
| 3399 | put_disk(disk); |
Alex Elder | 1fcdb8a | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 3400 | |
| 3401 | return -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3402 | } |
| 3403 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3404 | /* |
| 3405 | sysfs |
| 3406 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3407 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3408 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
| 3409 | { |
| 3410 | return container_of(dev, struct rbd_device, dev); |
| 3411 | } |
| 3412 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3413 | static ssize_t rbd_size_show(struct device *dev, |
| 3414 | struct device_attribute *attr, char *buf) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3415 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3416 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3417 | |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3418 | return sprintf(buf, "%llu\n", |
| 3419 | (unsigned long long)rbd_dev->mapping.size); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3420 | } |
| 3421 | |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3422 | /* |
| 3423 | * Note this shows the features for whatever's mapped, which is not |
| 3424 | * necessarily the base image. |
| 3425 | */ |
| 3426 | static ssize_t rbd_features_show(struct device *dev, |
| 3427 | struct device_attribute *attr, char *buf) |
| 3428 | { |
| 3429 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3430 | |
| 3431 | return sprintf(buf, "0x%016llx\n", |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3432 | (unsigned long long)rbd_dev->mapping.features); |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3433 | } |
| 3434 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3435 | static ssize_t rbd_major_show(struct device *dev, |
| 3436 | struct device_attribute *attr, char *buf) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3437 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3438 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3439 | |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3440 | if (rbd_dev->major) |
| 3441 | return sprintf(buf, "%d\n", rbd_dev->major); |
| 3442 | |
| 3443 | return sprintf(buf, "(none)\n"); |
| 3444 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3445 | } |
| 3446 | |
| 3447 | static ssize_t rbd_client_id_show(struct device *dev, |
| 3448 | struct device_attribute *attr, char *buf) |
| 3449 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3450 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3451 | |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 3452 | return sprintf(buf, "client%lld\n", |
| 3453 | ceph_client_id(rbd_dev->rbd_client->client)); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3454 | } |
| 3455 | |
| 3456 | static ssize_t rbd_pool_show(struct device *dev, |
| 3457 | struct device_attribute *attr, char *buf) |
| 3458 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3459 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3460 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3461 | return sprintf(buf, "%s\n", rbd_dev->spec->pool_name); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3462 | } |
| 3463 | |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3464 | static ssize_t rbd_pool_id_show(struct device *dev, |
| 3465 | struct device_attribute *attr, char *buf) |
| 3466 | { |
| 3467 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3468 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3469 | return sprintf(buf, "%llu\n", |
Alex Elder | fc71d83 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 3470 | (unsigned long long) rbd_dev->spec->pool_id); |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3471 | } |
| 3472 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3473 | static ssize_t rbd_name_show(struct device *dev, |
| 3474 | struct device_attribute *attr, char *buf) |
| 3475 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3476 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3477 | |
Alex Elder | a92ffdf | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3478 | if (rbd_dev->spec->image_name) |
| 3479 | return sprintf(buf, "%s\n", rbd_dev->spec->image_name); |
| 3480 | |
| 3481 | return sprintf(buf, "(unknown)\n"); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3482 | } |
| 3483 | |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3484 | static ssize_t rbd_image_id_show(struct device *dev, |
| 3485 | struct device_attribute *attr, char *buf) |
| 3486 | { |
| 3487 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3488 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3489 | return sprintf(buf, "%s\n", rbd_dev->spec->image_id); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3490 | } |
| 3491 | |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3492 | /* |
| 3493 | * Shows the name of the currently-mapped snapshot (or |
| 3494 | * RBD_SNAP_HEAD_NAME for the base image). |
| 3495 | */ |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3496 | static ssize_t rbd_snap_show(struct device *dev, |
| 3497 | struct device_attribute *attr, |
| 3498 | char *buf) |
| 3499 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3500 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3501 | |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 3502 | return sprintf(buf, "%s\n", rbd_dev->spec->snap_name); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3503 | } |
| 3504 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3505 | /* |
| 3506 | * For an rbd v2 image, shows the pool id, image id, and snapshot id |
| 3507 | * for the parent image. If there is no parent, simply shows |
| 3508 | * "(no parent image)". |
| 3509 | */ |
| 3510 | static ssize_t rbd_parent_show(struct device *dev, |
| 3511 | struct device_attribute *attr, |
| 3512 | char *buf) |
| 3513 | { |
| 3514 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 3515 | struct rbd_spec *spec = rbd_dev->parent_spec; |
| 3516 | int count; |
| 3517 | char *bufp = buf; |
| 3518 | |
| 3519 | if (!spec) |
| 3520 | return sprintf(buf, "(no parent image)\n"); |
| 3521 | |
| 3522 | count = sprintf(bufp, "pool_id %llu\npool_name %s\n", |
| 3523 | (unsigned long long) spec->pool_id, spec->pool_name); |
| 3524 | if (count < 0) |
| 3525 | return count; |
| 3526 | bufp += count; |
| 3527 | |
| 3528 | count = sprintf(bufp, "image_id %s\nimage_name %s\n", spec->image_id, |
| 3529 | spec->image_name ? spec->image_name : "(unknown)"); |
| 3530 | if (count < 0) |
| 3531 | return count; |
| 3532 | bufp += count; |
| 3533 | |
| 3534 | count = sprintf(bufp, "snap_id %llu\nsnap_name %s\n", |
| 3535 | (unsigned long long) spec->snap_id, spec->snap_name); |
| 3536 | if (count < 0) |
| 3537 | return count; |
| 3538 | bufp += count; |
| 3539 | |
| 3540 | count = sprintf(bufp, "overlap %llu\n", rbd_dev->parent_overlap); |
| 3541 | if (count < 0) |
| 3542 | return count; |
| 3543 | bufp += count; |
| 3544 | |
| 3545 | return (ssize_t) (bufp - buf); |
| 3546 | } |
| 3547 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3548 | static ssize_t rbd_image_refresh(struct device *dev, |
| 3549 | struct device_attribute *attr, |
| 3550 | const char *buf, |
| 3551 | size_t size) |
| 3552 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 3553 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Alex Elder | b813623 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3554 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3555 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3556 | ret = rbd_dev_refresh(rbd_dev); |
Alex Elder | e627db0 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 3557 | if (ret) |
| 3558 | rbd_warn(rbd_dev, ": manual header refresh error (%d)\n", ret); |
Alex Elder | b813623 | 2012-07-25 09:32:41 -0500 | [diff] [blame] | 3559 | |
| 3560 | return ret < 0 ? ret : size; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3561 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 3562 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3563 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3564 | static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3565 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); |
| 3566 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); |
| 3567 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3568 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3569 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3570 | static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3571 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); |
| 3572 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3573 | static DEVICE_ATTR(parent, S_IRUGO, rbd_parent_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3574 | |
| 3575 | static struct attribute *rbd_attrs[] = { |
| 3576 | &dev_attr_size.attr, |
Alex Elder | 34b1318 | 2012-07-13 20:35:12 -0500 | [diff] [blame] | 3577 | &dev_attr_features.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3578 | &dev_attr_major.attr, |
| 3579 | &dev_attr_client_id.attr, |
| 3580 | &dev_attr_pool.attr, |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 3581 | &dev_attr_pool_id.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3582 | &dev_attr_name.attr, |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 3583 | &dev_attr_image_id.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3584 | &dev_attr_current_snap.attr, |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3585 | &dev_attr_parent.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3586 | &dev_attr_refresh.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3587 | NULL |
| 3588 | }; |
| 3589 | |
| 3590 | static struct attribute_group rbd_attr_group = { |
| 3591 | .attrs = rbd_attrs, |
| 3592 | }; |
| 3593 | |
| 3594 | static const struct attribute_group *rbd_attr_groups[] = { |
| 3595 | &rbd_attr_group, |
| 3596 | NULL |
| 3597 | }; |
| 3598 | |
| 3599 | static void rbd_sysfs_dev_release(struct device *dev) |
| 3600 | { |
| 3601 | } |
| 3602 | |
| 3603 | static struct device_type rbd_device_type = { |
| 3604 | .name = "rbd", |
| 3605 | .groups = rbd_attr_groups, |
| 3606 | .release = rbd_sysfs_dev_release, |
| 3607 | }; |
| 3608 | |
Alex Elder | 8b8fb99 | 2012-10-26 17:25:24 -0500 | [diff] [blame] | 3609 | static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec) |
| 3610 | { |
| 3611 | kref_get(&spec->kref); |
| 3612 | |
| 3613 | return spec; |
| 3614 | } |
| 3615 | |
| 3616 | static void rbd_spec_free(struct kref *kref); |
| 3617 | static void rbd_spec_put(struct rbd_spec *spec) |
| 3618 | { |
| 3619 | if (spec) |
| 3620 | kref_put(&spec->kref, rbd_spec_free); |
| 3621 | } |
| 3622 | |
| 3623 | static struct rbd_spec *rbd_spec_alloc(void) |
| 3624 | { |
| 3625 | struct rbd_spec *spec; |
| 3626 | |
| 3627 | spec = kzalloc(sizeof (*spec), GFP_KERNEL); |
| 3628 | if (!spec) |
| 3629 | return NULL; |
| 3630 | kref_init(&spec->kref); |
| 3631 | |
Alex Elder | 8b8fb99 | 2012-10-26 17:25:24 -0500 | [diff] [blame] | 3632 | return spec; |
| 3633 | } |
| 3634 | |
| 3635 | static void rbd_spec_free(struct kref *kref) |
| 3636 | { |
| 3637 | struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref); |
| 3638 | |
| 3639 | kfree(spec->pool_name); |
| 3640 | kfree(spec->image_id); |
| 3641 | kfree(spec->image_name); |
| 3642 | kfree(spec->snap_name); |
| 3643 | kfree(spec); |
| 3644 | } |
| 3645 | |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 3646 | static struct rbd_device *rbd_dev_create(struct rbd_client *rbdc, |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3647 | struct rbd_spec *spec) |
| 3648 | { |
| 3649 | struct rbd_device *rbd_dev; |
| 3650 | |
| 3651 | rbd_dev = kzalloc(sizeof (*rbd_dev), GFP_KERNEL); |
| 3652 | if (!rbd_dev) |
| 3653 | return NULL; |
| 3654 | |
| 3655 | spin_lock_init(&rbd_dev->lock); |
Alex Elder | 6d29290 | 2013-01-14 12:43:31 -0600 | [diff] [blame] | 3656 | rbd_dev->flags = 0; |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 3657 | atomic_set(&rbd_dev->parent_ref, 0); |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3658 | INIT_LIST_HEAD(&rbd_dev->node); |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3659 | init_rwsem(&rbd_dev->header_rwsem); |
| 3660 | |
| 3661 | rbd_dev->spec = spec; |
| 3662 | rbd_dev->rbd_client = rbdc; |
| 3663 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 3664 | /* Initialize the layout used for all rbd requests */ |
| 3665 | |
| 3666 | rbd_dev->layout.fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); |
| 3667 | rbd_dev->layout.fl_stripe_count = cpu_to_le32(1); |
| 3668 | rbd_dev->layout.fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); |
| 3669 | rbd_dev->layout.fl_pg_pool = cpu_to_le32((u32) spec->pool_id); |
| 3670 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3671 | return rbd_dev; |
| 3672 | } |
| 3673 | |
| 3674 | static void rbd_dev_destroy(struct rbd_device *rbd_dev) |
| 3675 | { |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3676 | rbd_put_client(rbd_dev->rbd_client); |
| 3677 | rbd_spec_put(rbd_dev->spec); |
| 3678 | kfree(rbd_dev); |
| 3679 | } |
| 3680 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 3681 | /* |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3682 | * Get the size and object order for an image snapshot, or if |
| 3683 | * snap_id is CEPH_NOSNAP, gets this information for the base |
| 3684 | * image. |
| 3685 | */ |
| 3686 | static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, |
| 3687 | u8 *order, u64 *snap_size) |
| 3688 | { |
| 3689 | __le64 snapid = cpu_to_le64(snap_id); |
| 3690 | int ret; |
| 3691 | struct { |
| 3692 | u8 order; |
| 3693 | __le64 size; |
| 3694 | } __attribute__ ((packed)) size_buf = { 0 }; |
| 3695 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3696 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3697 | "rbd", "get_size", |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3698 | &snapid, sizeof (snapid), |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3699 | &size_buf, sizeof (size_buf)); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3700 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3701 | if (ret < 0) |
| 3702 | return ret; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3703 | if (ret < sizeof (size_buf)) |
| 3704 | return -ERANGE; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3705 | |
Josh Durgin | c354557 | 2013-08-28 17:08:10 -0700 | [diff] [blame] | 3706 | if (order) { |
Alex Elder | c86f86e | 2013-04-25 15:09:41 -0500 | [diff] [blame] | 3707 | *order = size_buf.order; |
Josh Durgin | c354557 | 2013-08-28 17:08:10 -0700 | [diff] [blame] | 3708 | dout(" order %u", (unsigned int)*order); |
| 3709 | } |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3710 | *snap_size = le64_to_cpu(size_buf.size); |
| 3711 | |
Josh Durgin | c354557 | 2013-08-28 17:08:10 -0700 | [diff] [blame] | 3712 | dout(" snap_id 0x%016llx snap_size = %llu\n", |
| 3713 | (unsigned long long)snap_id, |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3714 | (unsigned long long)*snap_size); |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3715 | |
| 3716 | return 0; |
| 3717 | } |
| 3718 | |
| 3719 | static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) |
| 3720 | { |
| 3721 | return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, |
| 3722 | &rbd_dev->header.obj_order, |
| 3723 | &rbd_dev->header.image_size); |
| 3724 | } |
| 3725 | |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3726 | static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) |
| 3727 | { |
| 3728 | void *reply_buf; |
| 3729 | int ret; |
| 3730 | void *p; |
| 3731 | |
| 3732 | reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL); |
| 3733 | if (!reply_buf) |
| 3734 | return -ENOMEM; |
| 3735 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3736 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3737 | "rbd", "get_object_prefix", NULL, 0, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3738 | reply_buf, RBD_OBJ_PREFIX_LEN_MAX); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3739 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3740 | if (ret < 0) |
| 3741 | goto out; |
| 3742 | |
| 3743 | p = reply_buf; |
| 3744 | rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3745 | p + ret, NULL, GFP_NOIO); |
| 3746 | ret = 0; |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3747 | |
| 3748 | if (IS_ERR(rbd_dev->header.object_prefix)) { |
| 3749 | ret = PTR_ERR(rbd_dev->header.object_prefix); |
| 3750 | rbd_dev->header.object_prefix = NULL; |
| 3751 | } else { |
| 3752 | dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); |
| 3753 | } |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3754 | out: |
| 3755 | kfree(reply_buf); |
| 3756 | |
| 3757 | return ret; |
| 3758 | } |
| 3759 | |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3760 | static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, |
| 3761 | u64 *snap_features) |
| 3762 | { |
| 3763 | __le64 snapid = cpu_to_le64(snap_id); |
| 3764 | struct { |
| 3765 | __le64 features; |
| 3766 | __le64 incompat; |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3767 | } __attribute__ ((packed)) features_buf = { 0 }; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 3768 | u64 incompat; |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3769 | int ret; |
| 3770 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3771 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3772 | "rbd", "get_features", |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3773 | &snapid, sizeof (snapid), |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3774 | &features_buf, sizeof (features_buf)); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3775 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3776 | if (ret < 0) |
| 3777 | return ret; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3778 | if (ret < sizeof (features_buf)) |
| 3779 | return -ERANGE; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 3780 | |
| 3781 | incompat = le64_to_cpu(features_buf.incompat); |
Alex Elder | 5cbf6f12 | 2013-04-11 09:29:48 -0500 | [diff] [blame] | 3782 | if (incompat & ~RBD_FEATURES_SUPPORTED) |
Alex Elder | b8f5c6e | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3783 | return -ENXIO; |
Alex Elder | d889140 | 2012-10-09 13:50:17 -0700 | [diff] [blame] | 3784 | |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3785 | *snap_features = le64_to_cpu(features_buf.features); |
| 3786 | |
| 3787 | dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3788 | (unsigned long long)snap_id, |
| 3789 | (unsigned long long)*snap_features, |
| 3790 | (unsigned long long)le64_to_cpu(features_buf.incompat)); |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 3791 | |
| 3792 | return 0; |
| 3793 | } |
| 3794 | |
| 3795 | static int rbd_dev_v2_features(struct rbd_device *rbd_dev) |
| 3796 | { |
| 3797 | return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, |
| 3798 | &rbd_dev->header.features); |
| 3799 | } |
| 3800 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3801 | static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) |
| 3802 | { |
| 3803 | struct rbd_spec *parent_spec; |
| 3804 | size_t size; |
| 3805 | void *reply_buf = NULL; |
| 3806 | __le64 snapid; |
| 3807 | void *p; |
| 3808 | void *end; |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3809 | u64 pool_id; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3810 | char *image_id; |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 3811 | u64 snap_id; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3812 | u64 overlap; |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3813 | int ret; |
| 3814 | |
| 3815 | parent_spec = rbd_spec_alloc(); |
| 3816 | if (!parent_spec) |
| 3817 | return -ENOMEM; |
| 3818 | |
| 3819 | size = sizeof (__le64) + /* pool_id */ |
| 3820 | sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX + /* image_id */ |
| 3821 | sizeof (__le64) + /* snap_id */ |
| 3822 | sizeof (__le64); /* overlap */ |
| 3823 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 3824 | if (!reply_buf) { |
| 3825 | ret = -ENOMEM; |
| 3826 | goto out_err; |
| 3827 | } |
| 3828 | |
| 3829 | snapid = cpu_to_le64(CEPH_NOSNAP); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3830 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3831 | "rbd", "get_parent", |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3832 | &snapid, sizeof (snapid), |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3833 | reply_buf, size); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 3834 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3835 | if (ret < 0) |
| 3836 | goto out_err; |
| 3837 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3838 | p = reply_buf; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3839 | end = reply_buf + ret; |
| 3840 | ret = -ERANGE; |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3841 | ceph_decode_64_safe(&p, end, pool_id, out_err); |
Alex Elder | 392a9da | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3842 | if (pool_id == CEPH_NOPOOL) { |
| 3843 | /* |
| 3844 | * Either the parent never existed, or we have |
| 3845 | * record of it but the image got flattened so it no |
| 3846 | * longer has a parent. When the parent of a |
| 3847 | * layered image disappears we immediately set the |
| 3848 | * overlap to 0. The effect of this is that all new |
| 3849 | * requests will be treated as if the image had no |
| 3850 | * parent. |
| 3851 | */ |
| 3852 | if (rbd_dev->parent_overlap) { |
| 3853 | rbd_dev->parent_overlap = 0; |
| 3854 | smp_mb(); |
| 3855 | rbd_dev_parent_put(rbd_dev); |
| 3856 | pr_info("%s: clone image has been flattened\n", |
| 3857 | rbd_dev->disk->disk_name); |
| 3858 | } |
| 3859 | |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3860 | goto out; /* No parent? No problem. */ |
Alex Elder | 392a9da | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3861 | } |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3862 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 3863 | /* The ceph file layout needs to fit pool id in 32 bits */ |
| 3864 | |
| 3865 | ret = -EIO; |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3866 | if (pool_id > (u64)U32_MAX) { |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3867 | rbd_warn(NULL, "parent pool id too large (%llu > %u)\n", |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3868 | (unsigned long long)pool_id, U32_MAX); |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3869 | goto out_err; |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3870 | } |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 3871 | |
Alex Elder | 979ed48 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3872 | image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3873 | if (IS_ERR(image_id)) { |
| 3874 | ret = PTR_ERR(image_id); |
| 3875 | goto out_err; |
| 3876 | } |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 3877 | ceph_decode_64_safe(&p, end, snap_id, out_err); |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3878 | ceph_decode_64_safe(&p, end, overlap, out_err); |
| 3879 | |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 3880 | /* |
| 3881 | * The parent won't change (except when the clone is |
| 3882 | * flattened, already handled that). So we only need to |
| 3883 | * record the parent spec we have not already done so. |
| 3884 | */ |
| 3885 | if (!rbd_dev->parent_spec) { |
| 3886 | parent_spec->pool_id = pool_id; |
| 3887 | parent_spec->image_id = image_id; |
| 3888 | parent_spec->snap_id = snap_id; |
Alex Elder | 70cf49c | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3889 | rbd_dev->parent_spec = parent_spec; |
| 3890 | parent_spec = NULL; /* rbd_dev now owns this */ |
Alex Elder | 3b5cf2a | 2013-05-29 11:18:59 -0500 | [diff] [blame] | 3891 | } |
| 3892 | |
| 3893 | /* |
| 3894 | * We always update the parent overlap. If it's zero we |
| 3895 | * treat it specially. |
| 3896 | */ |
| 3897 | rbd_dev->parent_overlap = overlap; |
| 3898 | smp_mb(); |
| 3899 | if (!overlap) { |
| 3900 | |
| 3901 | /* A null parent_spec indicates it's the initial probe */ |
| 3902 | |
| 3903 | if (parent_spec) { |
| 3904 | /* |
| 3905 | * The overlap has become zero, so the clone |
| 3906 | * must have been resized down to 0 at some |
| 3907 | * point. Treat this the same as a flatten. |
| 3908 | */ |
| 3909 | rbd_dev_parent_put(rbd_dev); |
| 3910 | pr_info("%s: clone image now standalone\n", |
| 3911 | rbd_dev->disk->disk_name); |
| 3912 | } else { |
| 3913 | /* |
| 3914 | * For the initial probe, if we find the |
| 3915 | * overlap is zero we just pretend there was |
| 3916 | * no parent image. |
| 3917 | */ |
| 3918 | rbd_warn(rbd_dev, "ignoring parent of " |
| 3919 | "clone with overlap 0\n"); |
| 3920 | } |
Alex Elder | 70cf49c | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 3921 | } |
Alex Elder | 86b00e0 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 3922 | out: |
| 3923 | ret = 0; |
| 3924 | out_err: |
| 3925 | kfree(reply_buf); |
| 3926 | rbd_spec_put(parent_spec); |
| 3927 | |
| 3928 | return ret; |
| 3929 | } |
| 3930 | |
Alex Elder | cc070d5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3931 | static int rbd_dev_v2_striping_info(struct rbd_device *rbd_dev) |
| 3932 | { |
| 3933 | struct { |
| 3934 | __le64 stripe_unit; |
| 3935 | __le64 stripe_count; |
| 3936 | } __attribute__ ((packed)) striping_info_buf = { 0 }; |
| 3937 | size_t size = sizeof (striping_info_buf); |
| 3938 | void *p; |
| 3939 | u64 obj_size; |
| 3940 | u64 stripe_unit; |
| 3941 | u64 stripe_count; |
| 3942 | int ret; |
| 3943 | |
| 3944 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
| 3945 | "rbd", "get_stripe_unit_count", NULL, 0, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 3946 | (char *)&striping_info_buf, size); |
Alex Elder | cc070d5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3947 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
| 3948 | if (ret < 0) |
| 3949 | return ret; |
| 3950 | if (ret < size) |
| 3951 | return -ERANGE; |
| 3952 | |
| 3953 | /* |
| 3954 | * We don't actually support the "fancy striping" feature |
| 3955 | * (STRIPINGV2) yet, but if the striping sizes are the |
| 3956 | * defaults the behavior is the same as before. So find |
| 3957 | * out, and only fail if the image has non-default values. |
| 3958 | */ |
| 3959 | ret = -EINVAL; |
| 3960 | obj_size = (u64)1 << rbd_dev->header.obj_order; |
| 3961 | p = &striping_info_buf; |
| 3962 | stripe_unit = ceph_decode_64(&p); |
| 3963 | if (stripe_unit != obj_size) { |
| 3964 | rbd_warn(rbd_dev, "unsupported stripe unit " |
| 3965 | "(got %llu want %llu)", |
| 3966 | stripe_unit, obj_size); |
| 3967 | return -EINVAL; |
| 3968 | } |
| 3969 | stripe_count = ceph_decode_64(&p); |
| 3970 | if (stripe_count != 1) { |
| 3971 | rbd_warn(rbd_dev, "unsupported stripe count " |
| 3972 | "(got %llu want 1)", stripe_count); |
| 3973 | return -EINVAL; |
| 3974 | } |
Alex Elder | 500d0c0 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 3975 | rbd_dev->header.stripe_unit = stripe_unit; |
| 3976 | rbd_dev->header.stripe_count = stripe_count; |
Alex Elder | cc070d5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 3977 | |
| 3978 | return 0; |
| 3979 | } |
| 3980 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3981 | static char *rbd_dev_image_name(struct rbd_device *rbd_dev) |
| 3982 | { |
| 3983 | size_t image_id_size; |
| 3984 | char *image_id; |
| 3985 | void *p; |
| 3986 | void *end; |
| 3987 | size_t size; |
| 3988 | void *reply_buf = NULL; |
| 3989 | size_t len = 0; |
| 3990 | char *image_name = NULL; |
| 3991 | int ret; |
| 3992 | |
| 3993 | rbd_assert(!rbd_dev->spec->image_name); |
| 3994 | |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 3995 | len = strlen(rbd_dev->spec->image_id); |
| 3996 | image_id_size = sizeof (__le32) + len; |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 3997 | image_id = kmalloc(image_id_size, GFP_KERNEL); |
| 3998 | if (!image_id) |
| 3999 | return NULL; |
| 4000 | |
| 4001 | p = image_id; |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4002 | end = image_id + image_id_size; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4003 | ceph_encode_string(&p, end, rbd_dev->spec->image_id, (u32)len); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4004 | |
| 4005 | size = sizeof (__le32) + RBD_IMAGE_NAME_LEN_MAX; |
| 4006 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 4007 | if (!reply_buf) |
| 4008 | goto out; |
| 4009 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4010 | ret = rbd_obj_method_sync(rbd_dev, RBD_DIRECTORY, |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4011 | "rbd", "dir_get_name", |
| 4012 | image_id, image_id_size, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4013 | reply_buf, size); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4014 | if (ret < 0) |
| 4015 | goto out; |
| 4016 | p = reply_buf; |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4017 | end = reply_buf + ret; |
| 4018 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4019 | image_name = ceph_extract_encoded_string(&p, end, &len, GFP_KERNEL); |
| 4020 | if (IS_ERR(image_name)) |
| 4021 | image_name = NULL; |
| 4022 | else |
| 4023 | dout("%s: name is %s len is %zd\n", __func__, image_name, len); |
| 4024 | out: |
| 4025 | kfree(reply_buf); |
| 4026 | kfree(image_id); |
| 4027 | |
| 4028 | return image_name; |
| 4029 | } |
| 4030 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4031 | static u64 rbd_v1_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) |
| 4032 | { |
| 4033 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
| 4034 | const char *snap_name; |
| 4035 | u32 which = 0; |
| 4036 | |
| 4037 | /* Skip over names until we find the one we are looking for */ |
| 4038 | |
| 4039 | snap_name = rbd_dev->header.snap_names; |
| 4040 | while (which < snapc->num_snaps) { |
| 4041 | if (!strcmp(name, snap_name)) |
| 4042 | return snapc->snaps[which]; |
| 4043 | snap_name += strlen(snap_name) + 1; |
| 4044 | which++; |
| 4045 | } |
| 4046 | return CEPH_NOSNAP; |
| 4047 | } |
| 4048 | |
| 4049 | static u64 rbd_v2_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) |
| 4050 | { |
| 4051 | struct ceph_snap_context *snapc = rbd_dev->header.snapc; |
| 4052 | u32 which; |
| 4053 | bool found = false; |
| 4054 | u64 snap_id; |
| 4055 | |
| 4056 | for (which = 0; !found && which < snapc->num_snaps; which++) { |
| 4057 | const char *snap_name; |
| 4058 | |
| 4059 | snap_id = snapc->snaps[which]; |
| 4060 | snap_name = rbd_dev_v2_snap_name(rbd_dev, snap_id); |
| 4061 | if (IS_ERR(snap_name)) |
| 4062 | break; |
| 4063 | found = !strcmp(name, snap_name); |
| 4064 | kfree(snap_name); |
| 4065 | } |
| 4066 | return found ? snap_id : CEPH_NOSNAP; |
| 4067 | } |
| 4068 | |
| 4069 | /* |
| 4070 | * Assumes name is never RBD_SNAP_HEAD_NAME; returns CEPH_NOSNAP if |
| 4071 | * no snapshot by that name is found, or if an error occurs. |
| 4072 | */ |
| 4073 | static u64 rbd_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) |
| 4074 | { |
| 4075 | if (rbd_dev->image_format == 1) |
| 4076 | return rbd_v1_snap_id_by_name(rbd_dev, name); |
| 4077 | |
| 4078 | return rbd_v2_snap_id_by_name(rbd_dev, name); |
| 4079 | } |
| 4080 | |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4081 | /* |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4082 | * When an rbd image has a parent image, it is identified by the |
| 4083 | * pool, image, and snapshot ids (not names). This function fills |
| 4084 | * in the names for those ids. (It's OK if we can't figure out the |
| 4085 | * name for an image id, but the pool and snapshot ids should always |
| 4086 | * exist and have names.) All names in an rbd spec are dynamically |
| 4087 | * allocated. |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4088 | * |
| 4089 | * When an image being mapped (not a parent) is probed, we have the |
| 4090 | * pool name and pool id, image name and image id, and the snapshot |
| 4091 | * name. The only thing we're missing is the snapshot id. |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4092 | */ |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4093 | static int rbd_dev_spec_update(struct rbd_device *rbd_dev) |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4094 | { |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4095 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
| 4096 | struct rbd_spec *spec = rbd_dev->spec; |
| 4097 | const char *pool_name; |
| 4098 | const char *image_name; |
| 4099 | const char *snap_name; |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4100 | int ret; |
| 4101 | |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4102 | /* |
| 4103 | * An image being mapped will have the pool name (etc.), but |
| 4104 | * we need to look up the snapshot id. |
| 4105 | */ |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4106 | if (spec->pool_name) { |
| 4107 | if (strcmp(spec->snap_name, RBD_SNAP_HEAD_NAME)) { |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4108 | u64 snap_id; |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4109 | |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4110 | snap_id = rbd_snap_id_by_name(rbd_dev, spec->snap_name); |
| 4111 | if (snap_id == CEPH_NOSNAP) |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4112 | return -ENOENT; |
Alex Elder | 2ad3d71 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4113 | spec->snap_id = snap_id; |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4114 | } else { |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4115 | spec->snap_id = CEPH_NOSNAP; |
Alex Elder | e1d4213 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4116 | } |
| 4117 | |
| 4118 | return 0; |
| 4119 | } |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4120 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4121 | /* 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] | 4122 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4123 | pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, spec->pool_id); |
| 4124 | if (!pool_name) { |
| 4125 | rbd_warn(rbd_dev, "no pool with id %llu", spec->pool_id); |
Alex Elder | 935dc89 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4126 | return -EIO; |
| 4127 | } |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4128 | pool_name = kstrdup(pool_name, GFP_KERNEL); |
| 4129 | if (!pool_name) |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4130 | return -ENOMEM; |
| 4131 | |
| 4132 | /* Fetch the image name; tolerate failure here */ |
| 4133 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4134 | image_name = rbd_dev_image_name(rbd_dev); |
| 4135 | if (!image_name) |
Alex Elder | 06ecc6c | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4136 | rbd_warn(rbd_dev, "unable to get image name"); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4137 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4138 | /* Look up the snapshot name, and make a copy */ |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4139 | |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4140 | snap_name = rbd_snap_name(rbd_dev, spec->snap_id); |
| 4141 | if (!snap_name) { |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4142 | ret = -ENOMEM; |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4143 | goto out_err; |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4144 | } |
| 4145 | |
| 4146 | spec->pool_name = pool_name; |
| 4147 | spec->image_name = image_name; |
| 4148 | spec->snap_name = snap_name; |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4149 | |
| 4150 | return 0; |
| 4151 | out_err: |
Alex Elder | 2e9f7f1 | 2013-04-26 09:43:48 -0500 | [diff] [blame] | 4152 | kfree(image_name); |
| 4153 | kfree(pool_name); |
Alex Elder | 9e15b77 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4154 | |
| 4155 | return ret; |
| 4156 | } |
| 4157 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4158 | static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev) |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4159 | { |
| 4160 | size_t size; |
| 4161 | int ret; |
| 4162 | void *reply_buf; |
| 4163 | void *p; |
| 4164 | void *end; |
| 4165 | u64 seq; |
| 4166 | u32 snap_count; |
| 4167 | struct ceph_snap_context *snapc; |
| 4168 | u32 i; |
| 4169 | |
| 4170 | /* |
| 4171 | * We'll need room for the seq value (maximum snapshot id), |
| 4172 | * snapshot count, and array of that many snapshot ids. |
| 4173 | * For now we have a fixed upper limit on the number we're |
| 4174 | * prepared to receive. |
| 4175 | */ |
| 4176 | size = sizeof (__le64) + sizeof (__le32) + |
| 4177 | RBD_MAX_SNAP_COUNT * sizeof (__le64); |
| 4178 | reply_buf = kzalloc(size, GFP_KERNEL); |
| 4179 | if (!reply_buf) |
| 4180 | return -ENOMEM; |
| 4181 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4182 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4183 | "rbd", "get_snapcontext", NULL, 0, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4184 | reply_buf, size); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4185 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4186 | if (ret < 0) |
| 4187 | goto out; |
| 4188 | |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4189 | p = reply_buf; |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4190 | end = reply_buf + ret; |
| 4191 | ret = -ERANGE; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4192 | ceph_decode_64_safe(&p, end, seq, out); |
| 4193 | ceph_decode_32_safe(&p, end, snap_count, out); |
| 4194 | |
| 4195 | /* |
| 4196 | * Make sure the reported number of snapshot ids wouldn't go |
| 4197 | * beyond the end of our buffer. But before checking that, |
| 4198 | * make sure the computed size of the snapshot context we |
| 4199 | * allocate is representable in a size_t. |
| 4200 | */ |
| 4201 | if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) |
| 4202 | / sizeof (u64)) { |
| 4203 | ret = -EINVAL; |
| 4204 | goto out; |
| 4205 | } |
| 4206 | if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) |
| 4207 | goto out; |
Alex Elder | 468521c | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4208 | ret = 0; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4209 | |
Alex Elder | 812164f8 | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 4210 | snapc = ceph_create_snap_context(snap_count, GFP_KERNEL); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4211 | if (!snapc) { |
| 4212 | ret = -ENOMEM; |
| 4213 | goto out; |
| 4214 | } |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4215 | snapc->seq = seq; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4216 | for (i = 0; i < snap_count; i++) |
| 4217 | snapc->snaps[i] = ceph_decode_64(&p); |
| 4218 | |
Alex Elder | 49ece55 | 2013-05-06 08:37:00 -0500 | [diff] [blame] | 4219 | ceph_put_snap_context(rbd_dev->header.snapc); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4220 | rbd_dev->header.snapc = snapc; |
| 4221 | |
| 4222 | dout(" snap context seq = %llu, snap_count = %u\n", |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4223 | (unsigned long long)seq, (unsigned int)snap_count); |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4224 | out: |
| 4225 | kfree(reply_buf); |
| 4226 | |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4227 | return ret; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4228 | } |
| 4229 | |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4230 | static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, |
| 4231 | u64 snap_id) |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4232 | { |
| 4233 | size_t size; |
| 4234 | void *reply_buf; |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4235 | __le64 snapid; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4236 | int ret; |
| 4237 | void *p; |
| 4238 | void *end; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4239 | char *snap_name; |
| 4240 | |
| 4241 | size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; |
| 4242 | reply_buf = kmalloc(size, GFP_KERNEL); |
| 4243 | if (!reply_buf) |
| 4244 | return ERR_PTR(-ENOMEM); |
| 4245 | |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4246 | snapid = cpu_to_le64(snap_id); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4247 | ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_name, |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4248 | "rbd", "get_snapshot_name", |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4249 | &snapid, sizeof (snapid), |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4250 | reply_buf, size); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4251 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4252 | if (ret < 0) { |
| 4253 | snap_name = ERR_PTR(ret); |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4254 | goto out; |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4255 | } |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4256 | |
| 4257 | p = reply_buf; |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4258 | end = reply_buf + ret; |
Alex Elder | e5c3553 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4259 | snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4260 | if (IS_ERR(snap_name)) |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4261 | goto out; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4262 | |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4263 | dout(" snap_id 0x%016llx snap_name = %s\n", |
Alex Elder | 54cac61 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4264 | (unsigned long long)snap_id, snap_name); |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4265 | out: |
| 4266 | kfree(reply_buf); |
| 4267 | |
Alex Elder | f40eb34 | 2013-04-25 15:09:42 -0500 | [diff] [blame] | 4268 | return snap_name; |
Alex Elder | b8b1e2d | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4269 | } |
| 4270 | |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4271 | static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev) |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4272 | { |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4273 | bool first_time = rbd_dev->header.object_prefix == NULL; |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4274 | int ret; |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4275 | |
Josh Durgin | 1617e40 | 2013-06-12 14:43:10 -0700 | [diff] [blame] | 4276 | ret = rbd_dev_v2_image_size(rbd_dev); |
| 4277 | if (ret) |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 4278 | return ret; |
Josh Durgin | 1617e40 | 2013-06-12 14:43:10 -0700 | [diff] [blame] | 4279 | |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4280 | if (first_time) { |
| 4281 | ret = rbd_dev_v2_header_onetime(rbd_dev); |
| 4282 | if (ret) |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 4283 | return ret; |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4284 | } |
| 4285 | |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4286 | /* |
| 4287 | * If the image supports layering, get the parent info. We |
| 4288 | * need to probe the first time regardless. Thereafter we |
| 4289 | * only need to if there's a parent, to see if it has |
| 4290 | * disappeared due to the mapped image getting flattened. |
| 4291 | */ |
| 4292 | if (rbd_dev->header.features & RBD_FEATURE_LAYERING && |
| 4293 | (first_time || rbd_dev->parent_spec)) { |
| 4294 | bool warn; |
| 4295 | |
| 4296 | ret = rbd_dev_v2_parent_info(rbd_dev); |
| 4297 | if (ret) |
Alex Elder | cfbf637 | 2013-05-31 17:40:45 -0500 | [diff] [blame] | 4298 | return ret; |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4299 | |
| 4300 | /* |
| 4301 | * Print a warning if this is the initial probe and |
| 4302 | * the image has a parent. Don't print it if the |
| 4303 | * image now being probed is itself a parent. We |
| 4304 | * can tell at this point because we won't know its |
| 4305 | * pool name yet (just its pool id). |
| 4306 | */ |
| 4307 | warn = rbd_dev->parent_spec && rbd_dev->spec->pool_name; |
| 4308 | if (first_time && warn) |
| 4309 | rbd_warn(rbd_dev, "WARNING: kernel layering " |
| 4310 | "is EXPERIMENTAL!"); |
| 4311 | } |
| 4312 | |
Alex Elder | 29334ba | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 4313 | if (rbd_dev->spec->snap_id == CEPH_NOSNAP) |
| 4314 | if (rbd_dev->mapping.size != rbd_dev->header.image_size) |
| 4315 | rbd_dev->mapping.size = rbd_dev->header.image_size; |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4316 | |
Alex Elder | cc4a38bd | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4317 | ret = rbd_dev_v2_snap_context(rbd_dev); |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4318 | dout("rbd_dev_v2_snap_context returned %d\n", ret); |
Alex Elder | 117973f | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4319 | |
| 4320 | return ret; |
| 4321 | } |
| 4322 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4323 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
| 4324 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4325 | struct device *dev; |
Alex Elder | cd789ab | 2012-08-30 00:16:38 -0500 | [diff] [blame] | 4326 | int ret; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4327 | |
Alex Elder | cd789ab | 2012-08-30 00:16:38 -0500 | [diff] [blame] | 4328 | dev = &rbd_dev->dev; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4329 | dev->bus = &rbd_bus_type; |
| 4330 | dev->type = &rbd_device_type; |
| 4331 | dev->parent = &rbd_root_dev; |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4332 | dev->release = rbd_dev_device_release; |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4333 | dev_set_name(dev, "%d", rbd_dev->dev_id); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4334 | ret = device_register(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4335 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4336 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 4337 | } |
| 4338 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 4339 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
| 4340 | { |
| 4341 | device_unregister(&rbd_dev->dev); |
| 4342 | } |
| 4343 | |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4344 | static atomic64_t rbd_dev_id_max = ATOMIC64_INIT(0); |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4345 | |
| 4346 | /* |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4347 | * Get a unique rbd identifier for the given new rbd_dev, and add |
| 4348 | * the rbd_dev to the global list. The minimum rbd id is 1. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4349 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4350 | static void rbd_dev_id_get(struct rbd_device *rbd_dev) |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 4351 | { |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4352 | rbd_dev->dev_id = atomic64_inc_return(&rbd_dev_id_max); |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4353 | |
| 4354 | spin_lock(&rbd_dev_list_lock); |
| 4355 | list_add_tail(&rbd_dev->node, &rbd_dev_list); |
| 4356 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4357 | dout("rbd_dev %p given dev id %llu\n", rbd_dev, |
| 4358 | (unsigned long long) rbd_dev->dev_id); |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4359 | } |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 4360 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4361 | /* |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4362 | * Remove an rbd_dev from the global list, and record that its |
| 4363 | * identifier is no longer in use. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4364 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4365 | static void rbd_dev_id_put(struct rbd_device *rbd_dev) |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4366 | { |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4367 | struct list_head *tmp; |
Alex Elder | de71a29 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4368 | int rbd_id = rbd_dev->dev_id; |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4369 | int max_id; |
| 4370 | |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 4371 | rbd_assert(rbd_id > 0); |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4372 | |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4373 | dout("rbd_dev %p released dev id %llu\n", rbd_dev, |
| 4374 | (unsigned long long) rbd_dev->dev_id); |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4375 | spin_lock(&rbd_dev_list_lock); |
| 4376 | list_del_init(&rbd_dev->node); |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4377 | |
| 4378 | /* |
| 4379 | * If the id being "put" is not the current maximum, there |
| 4380 | * is nothing special we need to do. |
| 4381 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4382 | if (rbd_id != atomic64_read(&rbd_dev_id_max)) { |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4383 | spin_unlock(&rbd_dev_list_lock); |
| 4384 | return; |
| 4385 | } |
| 4386 | |
| 4387 | /* |
| 4388 | * We need to update the current maximum id. Search the |
| 4389 | * list to find out what it is. We're more likely to find |
| 4390 | * the maximum at the end, so search the list backward. |
| 4391 | */ |
| 4392 | max_id = 0; |
| 4393 | list_for_each_prev(tmp, &rbd_dev_list) { |
| 4394 | struct rbd_device *rbd_dev; |
| 4395 | |
| 4396 | rbd_dev = list_entry(tmp, struct rbd_device, node); |
Alex Elder | b213e0b | 2012-10-10 21:19:13 -0700 | [diff] [blame] | 4397 | if (rbd_dev->dev_id > max_id) |
| 4398 | max_id = rbd_dev->dev_id; |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4399 | } |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 4400 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 4401 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4402 | /* |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4403 | * The max id could have been updated by rbd_dev_id_get(), in |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4404 | * which case it now accurately reflects the new maximum. |
| 4405 | * Be careful not to overwrite the maximum value in that |
| 4406 | * case. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 4407 | */ |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 4408 | atomic64_cmpxchg(&rbd_dev_id_max, rbd_id, max_id); |
| 4409 | dout(" max dev id has been reset\n"); |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 4410 | } |
| 4411 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4412 | /* |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4413 | * Skips over white space at *buf, and updates *buf to point to the |
| 4414 | * first found non-space character (if any). Returns the length of |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 4415 | * the token (string of non-white space characters) found. Note |
| 4416 | * that *buf must be terminated with '\0'. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4417 | */ |
| 4418 | static inline size_t next_token(const char **buf) |
| 4419 | { |
| 4420 | /* |
| 4421 | * These are the characters that produce nonzero for |
| 4422 | * isspace() in the "C" and "POSIX" locales. |
| 4423 | */ |
| 4424 | const char *spaces = " \f\n\r\t\v"; |
| 4425 | |
| 4426 | *buf += strspn(*buf, spaces); /* Find start of token */ |
| 4427 | |
| 4428 | return strcspn(*buf, spaces); /* Return token length */ |
| 4429 | } |
| 4430 | |
| 4431 | /* |
| 4432 | * Finds the next token in *buf, and if the provided token buffer is |
| 4433 | * big enough, copies the found token into it. The result, if |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 4434 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
| 4435 | * must be terminated with '\0' on entry. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4436 | * |
| 4437 | * Returns the length of the token found (not including the '\0'). |
| 4438 | * Return value will be 0 if no token is found, and it will be >= |
| 4439 | * token_size if the token would not fit. |
| 4440 | * |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 4441 | * 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] | 4442 | * found token. Note that this occurs even if the token buffer is |
| 4443 | * too small to hold it. |
| 4444 | */ |
| 4445 | static inline size_t copy_token(const char **buf, |
| 4446 | char *token, |
| 4447 | size_t token_size) |
| 4448 | { |
| 4449 | size_t len; |
| 4450 | |
| 4451 | len = next_token(buf); |
| 4452 | if (len < token_size) { |
| 4453 | memcpy(token, *buf, len); |
| 4454 | *(token + len) = '\0'; |
| 4455 | } |
| 4456 | *buf += len; |
| 4457 | |
| 4458 | return len; |
| 4459 | } |
| 4460 | |
| 4461 | /* |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 4462 | * Finds the next token in *buf, dynamically allocates a buffer big |
| 4463 | * enough to hold a copy of it, and copies the token into the new |
| 4464 | * buffer. The copy is guaranteed to be terminated with '\0'. Note |
| 4465 | * that a duplicate buffer is created even for a zero-length token. |
| 4466 | * |
| 4467 | * Returns a pointer to the newly-allocated duplicate, or a null |
| 4468 | * pointer if memory for the duplicate was not available. If |
| 4469 | * the lenp argument is a non-null pointer, the length of the token |
| 4470 | * (not including the '\0') is returned in *lenp. |
| 4471 | * |
| 4472 | * If successful, the *buf pointer will be updated to point beyond |
| 4473 | * the end of the found token. |
| 4474 | * |
| 4475 | * Note: uses GFP_KERNEL for allocation. |
| 4476 | */ |
| 4477 | static inline char *dup_token(const char **buf, size_t *lenp) |
| 4478 | { |
| 4479 | char *dup; |
| 4480 | size_t len; |
| 4481 | |
| 4482 | len = next_token(buf); |
Alex Elder | 4caf35f | 2012-11-01 08:39:27 -0500 | [diff] [blame] | 4483 | dup = kmemdup(*buf, len + 1, GFP_KERNEL); |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 4484 | if (!dup) |
| 4485 | return NULL; |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 4486 | *(dup + len) = '\0'; |
| 4487 | *buf += len; |
| 4488 | |
| 4489 | if (lenp) |
| 4490 | *lenp = len; |
| 4491 | |
| 4492 | return dup; |
| 4493 | } |
| 4494 | |
| 4495 | /* |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4496 | * Parse the options provided for an "rbd add" (i.e., rbd image |
| 4497 | * mapping) request. These arrive via a write to /sys/bus/rbd/add, |
| 4498 | * and the data written is passed here via a NUL-terminated buffer. |
| 4499 | * Returns 0 if successful or an error code otherwise. |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4500 | * |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4501 | * The information extracted from these options is recorded in |
| 4502 | * the other parameters which return dynamically-allocated |
| 4503 | * structures: |
| 4504 | * ceph_opts |
| 4505 | * The address of a pointer that will refer to a ceph options |
| 4506 | * structure. Caller must release the returned pointer using |
| 4507 | * ceph_destroy_options() when it is no longer needed. |
| 4508 | * rbd_opts |
| 4509 | * Address of an rbd options pointer. Fully initialized by |
| 4510 | * this function; caller must release with kfree(). |
| 4511 | * spec |
| 4512 | * Address of an rbd image specification pointer. Fully |
| 4513 | * initialized by this function based on parsed options. |
| 4514 | * Caller must release with rbd_spec_put(). |
| 4515 | * |
| 4516 | * The options passed take this form: |
| 4517 | * <mon_addrs> <options> <pool_name> <image_name> [<snap_id>] |
| 4518 | * where: |
| 4519 | * <mon_addrs> |
| 4520 | * A comma-separated list of one or more monitor addresses. |
| 4521 | * A monitor address is an ip address, optionally followed |
| 4522 | * by a port number (separated by a colon). |
| 4523 | * I.e.: ip1[:port1][,ip2[:port2]...] |
| 4524 | * <options> |
| 4525 | * A comma-separated list of ceph and/or rbd options. |
| 4526 | * <pool_name> |
| 4527 | * The name of the rados pool containing the rbd image. |
| 4528 | * <image_name> |
| 4529 | * The name of the image in that pool to map. |
| 4530 | * <snap_id> |
| 4531 | * An optional snapshot id. If provided, the mapping will |
| 4532 | * present data from the image at the time that snapshot was |
| 4533 | * created. The image head is used if no snapshot id is |
| 4534 | * provided. Snapshot mappings are always read-only. |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4535 | */ |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4536 | static int rbd_add_parse_args(const char *buf, |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4537 | struct ceph_options **ceph_opts, |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4538 | struct rbd_options **opts, |
| 4539 | struct rbd_spec **rbd_spec) |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4540 | { |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4541 | size_t len; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4542 | char *options; |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4543 | const char *mon_addrs; |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4544 | char *snap_name; |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4545 | size_t mon_addrs_size; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4546 | struct rbd_spec *spec = NULL; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4547 | struct rbd_options *rbd_opts = NULL; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4548 | struct ceph_options *copts; |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4549 | int ret; |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4550 | |
| 4551 | /* The first four tokens are required */ |
| 4552 | |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4553 | len = next_token(&buf); |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4554 | if (!len) { |
| 4555 | rbd_warn(NULL, "no monitor address(es) provided"); |
| 4556 | return -EINVAL; |
| 4557 | } |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4558 | mon_addrs = buf; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4559 | mon_addrs_size = len + 1; |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4560 | buf += len; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4561 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4562 | ret = -EINVAL; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4563 | options = dup_token(&buf, NULL); |
| 4564 | if (!options) |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4565 | return -ENOMEM; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4566 | if (!*options) { |
| 4567 | rbd_warn(NULL, "no options provided"); |
| 4568 | goto out_err; |
| 4569 | } |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4570 | |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4571 | spec = rbd_spec_alloc(); |
| 4572 | if (!spec) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4573 | goto out_mem; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4574 | |
| 4575 | spec->pool_name = dup_token(&buf, NULL); |
| 4576 | if (!spec->pool_name) |
| 4577 | goto out_mem; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4578 | if (!*spec->pool_name) { |
| 4579 | rbd_warn(NULL, "no pool name provided"); |
| 4580 | goto out_err; |
| 4581 | } |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4582 | |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 4583 | spec->image_name = dup_token(&buf, NULL); |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4584 | if (!spec->image_name) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4585 | goto out_mem; |
Alex Elder | 4fb5d671 | 2012-11-01 10:17:15 -0500 | [diff] [blame] | 4586 | if (!*spec->image_name) { |
| 4587 | rbd_warn(NULL, "no image name provided"); |
| 4588 | goto out_err; |
| 4589 | } |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4590 | |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4591 | /* |
| 4592 | * Snapshot name is optional; default is to use "-" |
| 4593 | * (indicating the head/no snapshot). |
| 4594 | */ |
Alex Elder | 3feeb894 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 4595 | len = next_token(&buf); |
Alex Elder | 820a5f3 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 4596 | if (!len) { |
Alex Elder | 3feeb894 | 2012-08-31 17:29:52 -0500 | [diff] [blame] | 4597 | buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ |
| 4598 | len = sizeof (RBD_SNAP_HEAD_NAME) - 1; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4599 | } else if (len > RBD_MAX_SNAP_NAME_LEN) { |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4600 | ret = -ENAMETOOLONG; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4601 | goto out_err; |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 4602 | } |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4603 | snap_name = kmemdup(buf, len + 1, GFP_KERNEL); |
| 4604 | if (!snap_name) |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4605 | goto out_mem; |
Alex Elder | ecb4dc2 | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 4606 | *(snap_name + len) = '\0'; |
| 4607 | spec->snap_name = snap_name; |
Alex Elder | e5c3553 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4608 | |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4609 | /* Initialize all rbd options to the defaults */ |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4610 | |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4611 | rbd_opts = kzalloc(sizeof (*rbd_opts), GFP_KERNEL); |
| 4612 | if (!rbd_opts) |
| 4613 | goto out_mem; |
| 4614 | |
| 4615 | rbd_opts->read_only = RBD_READ_ONLY_DEFAULT; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4616 | |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4617 | copts = ceph_parse_options(options, mon_addrs, |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4618 | mon_addrs + mon_addrs_size - 1, |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4619 | parse_rbd_opts_token, rbd_opts); |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4620 | if (IS_ERR(copts)) { |
| 4621 | ret = PTR_ERR(copts); |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4622 | goto out_err; |
| 4623 | } |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4624 | kfree(options); |
| 4625 | |
| 4626 | *ceph_opts = copts; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4627 | *opts = rbd_opts; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4628 | *rbd_spec = spec; |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4629 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4630 | return 0; |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4631 | out_mem: |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4632 | ret = -ENOMEM; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4633 | out_err: |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 4634 | kfree(rbd_opts); |
| 4635 | rbd_spec_put(spec); |
Alex Elder | f28e565 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4636 | kfree(options); |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 4637 | |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4638 | return ret; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 4639 | } |
| 4640 | |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4641 | /* |
| 4642 | * An rbd format 2 image has a unique identifier, distinct from the |
| 4643 | * name given to it by the user. Internally, that identifier is |
| 4644 | * what's used to specify the names of objects related to the image. |
| 4645 | * |
| 4646 | * A special "rbd id" object is used to map an rbd image name to its |
| 4647 | * id. If that object doesn't exist, then there is no v2 rbd image |
| 4648 | * with the supplied name. |
| 4649 | * |
| 4650 | * This function will record the given rbd_dev's image_id field if |
| 4651 | * it can be determined, and in that case will return 0. If any |
| 4652 | * errors occur a negative errno will be returned and the rbd_dev's |
| 4653 | * image_id field will be unchanged (and should be NULL). |
| 4654 | */ |
| 4655 | static int rbd_dev_image_id(struct rbd_device *rbd_dev) |
| 4656 | { |
| 4657 | int ret; |
| 4658 | size_t size; |
| 4659 | char *object_name; |
| 4660 | void *response; |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4661 | char *image_id; |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4662 | |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4663 | /* |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4664 | * When probing a parent image, the image id is already |
| 4665 | * known (and the image name likely is not). There's no |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4666 | * need to fetch the image id again in this case. We |
| 4667 | * do still need to set the image format though. |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4668 | */ |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4669 | if (rbd_dev->spec->image_id) { |
| 4670 | rbd_dev->image_format = *rbd_dev->spec->image_id ? 2 : 1; |
| 4671 | |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4672 | return 0; |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4673 | } |
Alex Elder | 2c0d0a1 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4674 | |
| 4675 | /* |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4676 | * First, see if the format 2 image id file exists, and if |
| 4677 | * so, get the image's persistent id from it. |
| 4678 | */ |
Alex Elder | 69e7a02 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 4679 | size = sizeof (RBD_ID_PREFIX) + strlen(rbd_dev->spec->image_name); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4680 | object_name = kmalloc(size, GFP_NOIO); |
| 4681 | if (!object_name) |
| 4682 | return -ENOMEM; |
Alex Elder | 0d7dbfc | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 4683 | 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] | 4684 | dout("rbd id object name is %s\n", object_name); |
| 4685 | |
| 4686 | /* Response will be an encoded string, which includes a length */ |
| 4687 | |
| 4688 | size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; |
| 4689 | response = kzalloc(size, GFP_NOIO); |
| 4690 | if (!response) { |
| 4691 | ret = -ENOMEM; |
| 4692 | goto out; |
| 4693 | } |
| 4694 | |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4695 | /* If it doesn't exist we'll assume it's a format 1 image */ |
| 4696 | |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4697 | ret = rbd_obj_method_sync(rbd_dev, object_name, |
Alex Elder | 4157976 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4698 | "rbd", "get_id", NULL, 0, |
Alex Elder | e2a58ee | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4699 | response, RBD_IMAGE_ID_LEN_MAX); |
Alex Elder | 36be9a7 | 2013-01-19 00:30:28 -0600 | [diff] [blame] | 4700 | dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4701 | if (ret == -ENOENT) { |
| 4702 | image_id = kstrdup("", GFP_KERNEL); |
| 4703 | ret = image_id ? 0 : -ENOMEM; |
| 4704 | if (!ret) |
| 4705 | rbd_dev->image_format = 1; |
| 4706 | } else if (ret > sizeof (__le32)) { |
| 4707 | void *p = response; |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4708 | |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4709 | image_id = ceph_extract_encoded_string(&p, p + ret, |
Alex Elder | 979ed48 | 2012-11-01 08:39:26 -0500 | [diff] [blame] | 4710 | NULL, GFP_NOIO); |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4711 | ret = IS_ERR(image_id) ? PTR_ERR(image_id) : 0; |
| 4712 | if (!ret) |
| 4713 | rbd_dev->image_format = 2; |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4714 | } else { |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4715 | ret = -EINVAL; |
| 4716 | } |
| 4717 | |
| 4718 | if (!ret) { |
| 4719 | rbd_dev->spec->image_id = image_id; |
| 4720 | dout("image_id is %s\n", image_id); |
Alex Elder | 589d30e | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4721 | } |
| 4722 | out: |
| 4723 | kfree(response); |
| 4724 | kfree(object_name); |
| 4725 | |
| 4726 | return ret; |
| 4727 | } |
| 4728 | |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 4729 | /* |
| 4730 | * Undo whatever state changes are made by v1 or v2 header info |
| 4731 | * call. |
| 4732 | */ |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4733 | static void rbd_dev_unprobe(struct rbd_device *rbd_dev) |
| 4734 | { |
| 4735 | struct rbd_image_header *header; |
| 4736 | |
Alex Elder | 392a9da | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4737 | /* Drop parent reference unless it's already been done (or none) */ |
| 4738 | |
| 4739 | if (rbd_dev->parent_overlap) |
| 4740 | rbd_dev_parent_put(rbd_dev); |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4741 | |
| 4742 | /* Free dynamic fields from the header, then zero it out */ |
| 4743 | |
| 4744 | header = &rbd_dev->header; |
Alex Elder | 812164f8 | 2013-04-30 00:44:32 -0500 | [diff] [blame] | 4745 | ceph_put_snap_context(header->snapc); |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4746 | kfree(header->snap_sizes); |
| 4747 | kfree(header->snap_names); |
| 4748 | kfree(header->object_prefix); |
| 4749 | memset(header, 0, sizeof (*header)); |
| 4750 | } |
| 4751 | |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4752 | static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev) |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4753 | { |
| 4754 | int ret; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4755 | |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4756 | ret = rbd_dev_v2_object_prefix(rbd_dev); |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4757 | if (ret) |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4758 | goto out_err; |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4759 | |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4760 | /* |
| 4761 | * Get the and check features for the image. Currently the |
| 4762 | * features are assumed to never change. |
| 4763 | */ |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4764 | ret = rbd_dev_v2_features(rbd_dev); |
Alex Elder | 57385b5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4765 | if (ret) |
Alex Elder | b1b5402 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4766 | goto out_err; |
Alex Elder | 35d489f | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4767 | |
Alex Elder | cc070d5 | 2013-04-21 12:14:45 -0500 | [diff] [blame] | 4768 | /* If the image supports fancy striping, get its parameters */ |
| 4769 | |
| 4770 | if (rbd_dev->header.features & RBD_FEATURE_STRIPINGV2) { |
| 4771 | ret = rbd_dev_v2_striping_info(rbd_dev); |
| 4772 | if (ret < 0) |
| 4773 | goto out_err; |
| 4774 | } |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4775 | /* No support for crypto and compression type format 2 images */ |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4776 | |
Alex Elder | 3515297 | 2012-08-31 17:29:55 -0500 | [diff] [blame] | 4777 | return 0; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4778 | out_err: |
Alex Elder | 642a253 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4779 | rbd_dev->header.features = 0; |
Alex Elder | 1e13019 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4780 | kfree(rbd_dev->header.object_prefix); |
| 4781 | rbd_dev->header.object_prefix = NULL; |
Alex Elder | 9d475de | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 4782 | |
| 4783 | return ret; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4784 | } |
| 4785 | |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4786 | static int rbd_dev_probe_parent(struct rbd_device *rbd_dev) |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4787 | { |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4788 | struct rbd_device *parent = NULL; |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4789 | struct rbd_spec *parent_spec; |
| 4790 | struct rbd_client *rbdc; |
| 4791 | int ret; |
| 4792 | |
| 4793 | if (!rbd_dev->parent_spec) |
| 4794 | return 0; |
| 4795 | /* |
| 4796 | * We need to pass a reference to the client and the parent |
| 4797 | * spec when creating the parent rbd_dev. Images related by |
| 4798 | * parent/child relationships always share both. |
| 4799 | */ |
| 4800 | parent_spec = rbd_spec_get(rbd_dev->parent_spec); |
| 4801 | rbdc = __rbd_get_client(rbd_dev->rbd_client); |
| 4802 | |
| 4803 | ret = -ENOMEM; |
| 4804 | parent = rbd_dev_create(rbdc, parent_spec); |
| 4805 | if (!parent) |
| 4806 | goto out_err; |
| 4807 | |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4808 | ret = rbd_dev_image_probe(parent, false); |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4809 | if (ret < 0) |
| 4810 | goto out_err; |
| 4811 | rbd_dev->parent = parent; |
Alex Elder | a2acd00 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 4812 | atomic_set(&rbd_dev->parent_ref, 1); |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4813 | |
| 4814 | return 0; |
| 4815 | out_err: |
| 4816 | if (parent) { |
Alex Elder | fb65d228 | 2013-05-08 22:50:04 -0500 | [diff] [blame] | 4817 | rbd_dev_unparent(rbd_dev); |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4818 | kfree(rbd_dev->header_name); |
| 4819 | rbd_dev_destroy(parent); |
| 4820 | } else { |
| 4821 | rbd_put_client(rbdc); |
| 4822 | rbd_spec_put(parent_spec); |
| 4823 | } |
| 4824 | |
| 4825 | return ret; |
| 4826 | } |
| 4827 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4828 | static int rbd_dev_device_setup(struct rbd_device *rbd_dev) |
Alex Elder | 124afba | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4829 | { |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4830 | int ret; |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4831 | |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4832 | /* generate unique id: find highest unique id, add one */ |
| 4833 | rbd_dev_id_get(rbd_dev); |
| 4834 | |
| 4835 | /* Fill in the device name, now that we have its id. */ |
| 4836 | BUILD_BUG_ON(DEV_NAME_LEN |
| 4837 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); |
| 4838 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->dev_id); |
| 4839 | |
| 4840 | /* Get our block major device number. */ |
| 4841 | |
| 4842 | ret = register_blkdev(0, rbd_dev->name); |
| 4843 | if (ret < 0) |
| 4844 | goto err_out_id; |
| 4845 | rbd_dev->major = ret; |
| 4846 | |
| 4847 | /* Set up the blkdev mapping. */ |
| 4848 | |
| 4849 | ret = rbd_init_disk(rbd_dev); |
| 4850 | if (ret) |
| 4851 | goto err_out_blkdev; |
| 4852 | |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 4853 | ret = rbd_dev_mapping_set(rbd_dev); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4854 | if (ret) |
| 4855 | goto err_out_disk; |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 4856 | set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); |
| 4857 | |
| 4858 | ret = rbd_bus_add_dev(rbd_dev); |
| 4859 | if (ret) |
| 4860 | goto err_out_mapping; |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4861 | |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4862 | /* Everything's ready. Announce the disk to the world. */ |
| 4863 | |
Alex Elder | 129b79d | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 4864 | set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4865 | add_disk(rbd_dev->disk); |
| 4866 | |
| 4867 | pr_info("%s: added with size 0x%llx\n", rbd_dev->disk->disk_name, |
| 4868 | (unsigned long long) rbd_dev->mapping.size); |
| 4869 | |
| 4870 | return ret; |
Alex Elder | 2f82ee5 | 2012-10-30 19:40:33 -0500 | [diff] [blame] | 4871 | |
Alex Elder | f35a4de | 2013-05-06 09:51:29 -0500 | [diff] [blame] | 4872 | err_out_mapping: |
| 4873 | rbd_dev_mapping_clear(rbd_dev); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4874 | err_out_disk: |
| 4875 | rbd_free_disk(rbd_dev); |
| 4876 | err_out_blkdev: |
| 4877 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
| 4878 | err_out_id: |
| 4879 | rbd_dev_id_put(rbd_dev); |
Alex Elder | d1cf578 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 4880 | rbd_dev_mapping_clear(rbd_dev); |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4881 | |
| 4882 | return ret; |
| 4883 | } |
| 4884 | |
Alex Elder | 332bb12 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 4885 | static int rbd_dev_header_name(struct rbd_device *rbd_dev) |
| 4886 | { |
| 4887 | struct rbd_spec *spec = rbd_dev->spec; |
| 4888 | size_t size; |
| 4889 | |
| 4890 | /* Record the header object name for this rbd image. */ |
| 4891 | |
| 4892 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 4893 | |
| 4894 | if (rbd_dev->image_format == 1) |
| 4895 | size = strlen(spec->image_name) + sizeof (RBD_SUFFIX); |
| 4896 | else |
| 4897 | size = sizeof (RBD_HEADER_PREFIX) + strlen(spec->image_id); |
| 4898 | |
| 4899 | rbd_dev->header_name = kmalloc(size, GFP_KERNEL); |
| 4900 | if (!rbd_dev->header_name) |
| 4901 | return -ENOMEM; |
| 4902 | |
| 4903 | if (rbd_dev->image_format == 1) |
| 4904 | sprintf(rbd_dev->header_name, "%s%s", |
| 4905 | spec->image_name, RBD_SUFFIX); |
| 4906 | else |
| 4907 | sprintf(rbd_dev->header_name, "%s%s", |
| 4908 | RBD_HEADER_PREFIX, spec->image_id); |
| 4909 | return 0; |
| 4910 | } |
| 4911 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4912 | static void rbd_dev_image_release(struct rbd_device *rbd_dev) |
| 4913 | { |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4914 | rbd_dev_unprobe(rbd_dev); |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4915 | kfree(rbd_dev->header_name); |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4916 | rbd_dev->header_name = NULL; |
| 4917 | rbd_dev->image_format = 0; |
| 4918 | kfree(rbd_dev->spec->image_id); |
| 4919 | rbd_dev->spec->image_id = NULL; |
| 4920 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4921 | rbd_dev_destroy(rbd_dev); |
| 4922 | } |
| 4923 | |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4924 | /* |
| 4925 | * Probe for the existence of the header object for the given rbd |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4926 | * device. If this image is the one being mapped (i.e., not a |
| 4927 | * parent), initiate a watch on its header object before using that |
| 4928 | * object to get detailed information about the rbd image. |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4929 | */ |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4930 | 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] | 4931 | { |
| 4932 | int ret; |
Alex Elder | b644de2 | 2013-04-27 09:59:31 -0500 | [diff] [blame] | 4933 | int tmp; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4934 | |
| 4935 | /* |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 4936 | * Get the id from the image id object. Unless there's an |
| 4937 | * error, rbd_dev->spec->image_id will be filled in with |
| 4938 | * a dynamically-allocated string, and rbd_dev->image_format |
| 4939 | * will be set to either 1 or 2. |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4940 | */ |
| 4941 | ret = rbd_dev_image_id(rbd_dev); |
| 4942 | if (ret) |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4943 | return ret; |
| 4944 | rbd_assert(rbd_dev->spec->image_id); |
| 4945 | rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); |
| 4946 | |
Alex Elder | 332bb12 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 4947 | ret = rbd_dev_header_name(rbd_dev); |
| 4948 | if (ret) |
| 4949 | goto err_out_format; |
| 4950 | |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4951 | if (mapping) { |
| 4952 | ret = rbd_dev_header_watch_sync(rbd_dev, true); |
| 4953 | if (ret) |
| 4954 | goto out_header_name; |
| 4955 | } |
Alex Elder | b644de2 | 2013-04-27 09:59:31 -0500 | [diff] [blame] | 4956 | |
Alex Elder | c0fba36 | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4957 | if (rbd_dev->image_format == 1) |
Alex Elder | 99a41eb | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4958 | ret = rbd_dev_v1_header_info(rbd_dev); |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4959 | else |
Alex Elder | 2df3fac | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4960 | ret = rbd_dev_v2_header_info(rbd_dev); |
Alex Elder | 5655c4d | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4961 | if (ret) |
Alex Elder | b644de2 | 2013-04-27 09:59:31 -0500 | [diff] [blame] | 4962 | goto err_out_watch; |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4963 | |
Alex Elder | 9bb81c9 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 4964 | ret = rbd_dev_spec_update(rbd_dev); |
| 4965 | if (ret) |
Alex Elder | 33dca39 | 2013-04-30 00:44:33 -0500 | [diff] [blame] | 4966 | goto err_out_probe; |
Alex Elder | 9bb81c9 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 4967 | |
| 4968 | ret = rbd_dev_probe_parent(rbd_dev); |
Alex Elder | 30d60ba | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4969 | if (ret) |
| 4970 | goto err_out_probe; |
Alex Elder | 83a0626 | 2012-10-30 15:47:17 -0500 | [diff] [blame] | 4971 | |
Alex Elder | 30d60ba | 2013-05-06 09:51:30 -0500 | [diff] [blame] | 4972 | dout("discovered format %u image, header name is %s\n", |
| 4973 | rbd_dev->image_format, rbd_dev->header_name); |
| 4974 | |
| 4975 | return 0; |
Alex Elder | 6fd48b3 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 4976 | err_out_probe: |
| 4977 | rbd_dev_unprobe(rbd_dev); |
Alex Elder | b644de2 | 2013-04-27 09:59:31 -0500 | [diff] [blame] | 4978 | err_out_watch: |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 4979 | if (mapping) { |
| 4980 | tmp = rbd_dev_header_watch_sync(rbd_dev, false); |
| 4981 | if (tmp) |
| 4982 | rbd_warn(rbd_dev, "unable to tear down " |
| 4983 | "watch request (%d)\n", tmp); |
| 4984 | } |
Alex Elder | 332bb12 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 4985 | out_header_name: |
| 4986 | kfree(rbd_dev->header_name); |
| 4987 | rbd_dev->header_name = NULL; |
| 4988 | err_out_format: |
| 4989 | rbd_dev->image_format = 0; |
Alex Elder | 5655c4d | 2013-04-25 23:15:08 -0500 | [diff] [blame] | 4990 | kfree(rbd_dev->spec->image_id); |
| 4991 | rbd_dev->spec->image_id = NULL; |
| 4992 | |
| 4993 | dout("probe failed, returning %d\n", ret); |
| 4994 | |
| 4995 | return ret; |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 4996 | } |
| 4997 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 4998 | static ssize_t rbd_add(struct bus_type *bus, |
| 4999 | const char *buf, |
| 5000 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5001 | { |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 5002 | struct rbd_device *rbd_dev = NULL; |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5003 | struct ceph_options *ceph_opts = NULL; |
Alex Elder | 4e9afeb | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5004 | struct rbd_options *rbd_opts = NULL; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5005 | struct rbd_spec *spec = NULL; |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5006 | struct rbd_client *rbdc; |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5007 | struct ceph_osd_client *osdc; |
Alex Elder | 51344a3 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 5008 | bool read_only; |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5009 | int rc = -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5010 | |
| 5011 | if (!try_module_get(THIS_MODULE)) |
| 5012 | return -ENODEV; |
| 5013 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5014 | /* parse add command */ |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5015 | rc = rbd_add_parse_args(buf, &ceph_opts, &rbd_opts, &spec); |
Alex Elder | dc79b11 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5016 | if (rc < 0) |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5017 | goto err_out_module; |
Alex Elder | 51344a3 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 5018 | read_only = rbd_opts->read_only; |
| 5019 | kfree(rbd_opts); |
| 5020 | rbd_opts = NULL; /* done with this */ |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5021 | |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5022 | rbdc = rbd_get_client(ceph_opts); |
| 5023 | if (IS_ERR(rbdc)) { |
| 5024 | rc = PTR_ERR(rbdc); |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5025 | goto err_out_args; |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5026 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5027 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5028 | /* pick the pool */ |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5029 | osdc = &rbdc->client->osdc; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5030 | rc = ceph_pg_poolid_by_name(osdc->osdmap, spec->pool_name); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5031 | if (rc < 0) |
| 5032 | goto err_out_client; |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 5033 | spec->pool_id = (u64)rc; |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5034 | |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 5035 | /* The ceph file layout needs to fit pool id in 32 bits */ |
| 5036 | |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 5037 | if (spec->pool_id > (u64)U32_MAX) { |
| 5038 | rbd_warn(NULL, "pool id too large (%llu > %u)\n", |
| 5039 | (unsigned long long)spec->pool_id, U32_MAX); |
Alex Elder | 0903e87 | 2012-11-14 12:25:19 -0600 | [diff] [blame] | 5040 | rc = -EIO; |
| 5041 | goto err_out_client; |
| 5042 | } |
| 5043 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5044 | rbd_dev = rbd_dev_create(rbdc, spec); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5045 | if (!rbd_dev) |
| 5046 | goto err_out_client; |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5047 | rbdc = NULL; /* rbd_dev now owns this */ |
| 5048 | spec = NULL; /* rbd_dev now owns this */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5049 | |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5050 | rc = rbd_dev_image_probe(rbd_dev, true); |
Alex Elder | a30b71b | 2012-07-10 20:30:11 -0500 | [diff] [blame] | 5051 | if (rc < 0) |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5052 | goto err_out_rbd_dev; |
Alex Elder | 05fd6f6 | 2012-08-29 17:11:07 -0500 | [diff] [blame] | 5053 | |
Alex Elder | 7ce4eef | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5054 | /* If we are mapping a snapshot it must be marked read-only */ |
| 5055 | |
| 5056 | if (rbd_dev->spec->snap_id != CEPH_NOSNAP) |
| 5057 | read_only = true; |
| 5058 | rbd_dev->mapping.read_only = read_only; |
| 5059 | |
Alex Elder | b536f69 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5060 | rc = rbd_dev_device_setup(rbd_dev); |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 5061 | if (rc) { |
| 5062 | rbd_dev_image_release(rbd_dev); |
| 5063 | goto err_out_module; |
| 5064 | } |
Alex Elder | b536f69 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5065 | |
Alex Elder | 3abef3b | 2013-05-13 20:35:37 -0500 | [diff] [blame] | 5066 | return count; |
| 5067 | |
Alex Elder | c53d589 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5068 | err_out_rbd_dev: |
| 5069 | rbd_dev_destroy(rbd_dev); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5070 | err_out_client: |
Alex Elder | 9d3997f | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5071 | rbd_put_client(rbdc); |
Alex Elder | 0ddebc0 | 2012-10-25 23:34:41 -0500 | [diff] [blame] | 5072 | err_out_args: |
Alex Elder | 859c31d | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5073 | rbd_spec_put(spec); |
Alex Elder | bd4ba65 | 2012-10-25 23:34:42 -0500 | [diff] [blame] | 5074 | err_out_module: |
| 5075 | module_put(THIS_MODULE); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5076 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5077 | dout("Error adding device %s\n", buf); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 5078 | |
Alex Elder | c0cd10db | 2013-04-26 09:43:47 -0500 | [diff] [blame] | 5079 | return (ssize_t)rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5080 | } |
| 5081 | |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5082 | static void rbd_dev_device_release(struct device *dev) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5083 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 5084 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5085 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5086 | rbd_free_disk(rbd_dev); |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5087 | clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); |
Alex Elder | 6d80b13 | 2013-05-06 07:40:30 -0500 | [diff] [blame] | 5088 | rbd_dev_mapping_clear(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5089 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
Alex Elder | 200a6a8 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5090 | rbd_dev->major = 0; |
Alex Elder | e283930 | 2012-08-29 17:11:06 -0500 | [diff] [blame] | 5091 | rbd_dev_id_put(rbd_dev); |
Alex Elder | d1cf578 | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5092 | rbd_dev_mapping_clear(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5093 | } |
| 5094 | |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5095 | static void rbd_dev_remove_parent(struct rbd_device *rbd_dev) |
| 5096 | { |
Alex Elder | ad945fc | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5097 | while (rbd_dev->parent) { |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5098 | struct rbd_device *first = rbd_dev; |
| 5099 | struct rbd_device *second = first->parent; |
| 5100 | struct rbd_device *third; |
| 5101 | |
| 5102 | /* |
| 5103 | * Follow to the parent with no grandparent and |
| 5104 | * remove it. |
| 5105 | */ |
| 5106 | while (second && (third = second->parent)) { |
| 5107 | first = second; |
| 5108 | second = third; |
| 5109 | } |
Alex Elder | ad945fc | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5110 | rbd_assert(second); |
Alex Elder | 8ad42cd | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5111 | rbd_dev_image_release(second); |
Alex Elder | ad945fc | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5112 | first->parent = NULL; |
| 5113 | first->parent_overlap = 0; |
| 5114 | |
| 5115 | rbd_assert(first->parent_spec); |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5116 | rbd_spec_put(first->parent_spec); |
| 5117 | first->parent_spec = NULL; |
Alex Elder | 05a46af | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5118 | } |
| 5119 | } |
| 5120 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5121 | static ssize_t rbd_remove(struct bus_type *bus, |
| 5122 | const char *buf, |
| 5123 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5124 | { |
| 5125 | struct rbd_device *rbd_dev = NULL; |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5126 | struct list_head *tmp; |
| 5127 | int dev_id; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5128 | unsigned long ul; |
Alex Elder | 82a442d | 2013-05-31 17:40:44 -0500 | [diff] [blame] | 5129 | bool already = false; |
Alex Elder | 0d8189e | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5130 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5131 | |
Alex Elder | 0d8189e | 2013-04-27 09:59:30 -0500 | [diff] [blame] | 5132 | ret = strict_strtoul(buf, 10, &ul); |
| 5133 | if (ret) |
| 5134 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5135 | |
| 5136 | /* convert to int; abort if we lost anything in the conversion */ |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5137 | dev_id = (int)ul; |
| 5138 | if (dev_id != ul) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5139 | return -EINVAL; |
| 5140 | |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5141 | ret = -ENOENT; |
| 5142 | spin_lock(&rbd_dev_list_lock); |
| 5143 | list_for_each(tmp, &rbd_dev_list) { |
| 5144 | rbd_dev = list_entry(tmp, struct rbd_device, node); |
| 5145 | if (rbd_dev->dev_id == dev_id) { |
| 5146 | ret = 0; |
| 5147 | break; |
| 5148 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5149 | } |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5150 | if (!ret) { |
| 5151 | spin_lock_irq(&rbd_dev->lock); |
| 5152 | if (rbd_dev->open_count) |
| 5153 | ret = -EBUSY; |
| 5154 | else |
Alex Elder | 82a442d | 2013-05-31 17:40:44 -0500 | [diff] [blame] | 5155 | already = test_and_set_bit(RBD_DEV_FLAG_REMOVING, |
| 5156 | &rbd_dev->flags); |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5157 | spin_unlock_irq(&rbd_dev->lock); |
| 5158 | } |
| 5159 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | 82a442d | 2013-05-31 17:40:44 -0500 | [diff] [blame] | 5160 | if (ret < 0 || already) |
Alex Elder | 1ba0f1e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5161 | return ret; |
Alex Elder | 751cc0e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5162 | |
Alex Elder | b480815 | 2013-04-26 15:44:36 -0500 | [diff] [blame] | 5163 | rbd_bus_del_dev(rbd_dev); |
Alex Elder | 1f3ef78 | 2013-05-06 17:40:33 -0500 | [diff] [blame] | 5164 | ret = rbd_dev_header_watch_sync(rbd_dev, false); |
| 5165 | if (ret) |
| 5166 | rbd_warn(rbd_dev, "failed to cancel watch event (%d)\n", ret); |
Josh Durgin | 9abc599 | 2013-08-29 17:31:03 -0700 | [diff] [blame] | 5167 | |
| 5168 | /* |
| 5169 | * flush remaining watch callbacks - these must be complete |
| 5170 | * before the osd_client is shutdown |
| 5171 | */ |
| 5172 | dout("%s: flushing notifies", __func__); |
| 5173 | ceph_osdc_flush_notifies(&rbd_dev->rbd_client->client->osdc); |
Alex Elder | 8ad42cd | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5174 | rbd_dev_image_release(rbd_dev); |
Alex Elder | 79ab755 | 2013-04-28 23:32:34 -0500 | [diff] [blame] | 5175 | module_put(THIS_MODULE); |
Alex Elder | aafb230e | 2012-09-06 16:00:54 -0500 | [diff] [blame] | 5176 | |
Alex Elder | 1ba0f1e | 2013-05-31 15:17:01 -0500 | [diff] [blame] | 5177 | return count; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5178 | } |
| 5179 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5180 | /* |
| 5181 | * create control files in sysfs |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5182 | * /sys/bus/rbd/... |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5183 | */ |
| 5184 | static int rbd_sysfs_init(void) |
| 5185 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5186 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5187 | |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 5188 | ret = device_register(&rbd_root_dev); |
Alex Elder | 2107978 | 2012-01-24 10:08:36 -0600 | [diff] [blame] | 5189 | if (ret < 0) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5190 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5191 | |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 5192 | ret = bus_register(&rbd_bus_type); |
| 5193 | if (ret < 0) |
| 5194 | device_unregister(&rbd_root_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5195 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5196 | return ret; |
| 5197 | } |
| 5198 | |
| 5199 | static void rbd_sysfs_cleanup(void) |
| 5200 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 5201 | bus_unregister(&rbd_bus_type); |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 5202 | device_unregister(&rbd_root_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5203 | } |
| 5204 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5205 | static int rbd_slab_init(void) |
| 5206 | { |
| 5207 | rbd_assert(!rbd_img_request_cache); |
| 5208 | rbd_img_request_cache = kmem_cache_create("rbd_img_request", |
| 5209 | sizeof (struct rbd_img_request), |
| 5210 | __alignof__(struct rbd_img_request), |
| 5211 | 0, NULL); |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5212 | if (!rbd_img_request_cache) |
| 5213 | return -ENOMEM; |
| 5214 | |
| 5215 | rbd_assert(!rbd_obj_request_cache); |
| 5216 | rbd_obj_request_cache = kmem_cache_create("rbd_obj_request", |
| 5217 | sizeof (struct rbd_obj_request), |
| 5218 | __alignof__(struct rbd_obj_request), |
| 5219 | 0, NULL); |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 5220 | if (!rbd_obj_request_cache) |
| 5221 | goto out_err; |
| 5222 | |
| 5223 | rbd_assert(!rbd_segment_name_cache); |
| 5224 | rbd_segment_name_cache = kmem_cache_create("rbd_segment_name", |
| 5225 | MAX_OBJ_NAME_SIZE + 1, 1, 0, NULL); |
| 5226 | if (rbd_segment_name_cache) |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5227 | return 0; |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 5228 | out_err: |
| 5229 | if (rbd_obj_request_cache) { |
| 5230 | kmem_cache_destroy(rbd_obj_request_cache); |
| 5231 | rbd_obj_request_cache = NULL; |
| 5232 | } |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5233 | |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5234 | kmem_cache_destroy(rbd_img_request_cache); |
| 5235 | rbd_img_request_cache = NULL; |
| 5236 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5237 | return -ENOMEM; |
| 5238 | } |
| 5239 | |
| 5240 | static void rbd_slab_exit(void) |
| 5241 | { |
Alex Elder | 78c2a44 | 2013-05-01 12:43:04 -0500 | [diff] [blame] | 5242 | rbd_assert(rbd_segment_name_cache); |
| 5243 | kmem_cache_destroy(rbd_segment_name_cache); |
| 5244 | rbd_segment_name_cache = NULL; |
| 5245 | |
Alex Elder | 868311b | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5246 | rbd_assert(rbd_obj_request_cache); |
| 5247 | kmem_cache_destroy(rbd_obj_request_cache); |
| 5248 | rbd_obj_request_cache = NULL; |
| 5249 | |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5250 | rbd_assert(rbd_img_request_cache); |
| 5251 | kmem_cache_destroy(rbd_img_request_cache); |
| 5252 | rbd_img_request_cache = NULL; |
| 5253 | } |
| 5254 | |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 5255 | static int __init rbd_init(void) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5256 | { |
| 5257 | int rc; |
| 5258 | |
Alex Elder | 1e32d34 | 2013-01-30 11:13:33 -0600 | [diff] [blame] | 5259 | if (!libceph_compatible(NULL)) { |
| 5260 | rbd_warn(NULL, "libceph incompatibility (quitting)"); |
| 5261 | |
| 5262 | return -EINVAL; |
| 5263 | } |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5264 | rc = rbd_slab_init(); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5265 | if (rc) |
| 5266 | return rc; |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5267 | rc = rbd_sysfs_init(); |
| 5268 | if (rc) |
| 5269 | rbd_slab_exit(); |
| 5270 | else |
| 5271 | pr_info("loaded " RBD_DRV_NAME_LONG "\n"); |
| 5272 | |
| 5273 | return rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5274 | } |
| 5275 | |
Alex Elder | cc344fa | 2013-02-19 12:25:56 -0600 | [diff] [blame] | 5276 | static void __exit rbd_exit(void) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5277 | { |
| 5278 | rbd_sysfs_cleanup(); |
Alex Elder | 1c2a9df | 2013-05-01 12:43:03 -0500 | [diff] [blame] | 5279 | rbd_slab_exit(); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5280 | } |
| 5281 | |
| 5282 | module_init(rbd_init); |
| 5283 | module_exit(rbd_exit); |
| 5284 | |
Alex Elder | d552c61 | 2013-05-31 20:13:09 -0500 | [diff] [blame] | 5285 | MODULE_AUTHOR("Alex Elder <elder@inktank.com>"); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 5286 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); |
| 5287 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); |
| 5288 | MODULE_DESCRIPTION("rados block device"); |
| 5289 | |
| 5290 | /* following authorship retained from original osdblk.c */ |
| 5291 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); |
| 5292 | |
| 5293 | MODULE_LICENSE("GPL"); |