Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1 | /* |
| 2 | rbd.c -- Export ceph rados objects as a Linux block device |
| 3 | |
| 4 | |
| 5 | based on drivers/block/osdblk.c: |
| 6 | |
| 7 | Copyright 2009 Red Hat, Inc. |
| 8 | |
| 9 | This program is free software; you can redistribute it and/or modify |
| 10 | it under the terms of the GNU General Public License as published by |
| 11 | the Free Software Foundation. |
| 12 | |
| 13 | This program is distributed in the hope that it will be useful, |
| 14 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | GNU General Public License for more details. |
| 17 | |
| 18 | You should have received a copy of the GNU General Public License |
| 19 | along with this program; see the file COPYING. If not, write to |
| 20 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. |
| 21 | |
| 22 | |
| 23 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 24 | For usage instructions, please refer to: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 25 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 26 | Documentation/ABI/testing/sysfs-bus-rbd |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 27 | |
| 28 | */ |
| 29 | |
| 30 | #include <linux/ceph/libceph.h> |
| 31 | #include <linux/ceph/osd_client.h> |
| 32 | #include <linux/ceph/mon_client.h> |
| 33 | #include <linux/ceph/decode.h> |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 34 | #include <linux/parser.h> |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 35 | |
| 36 | #include <linux/kernel.h> |
| 37 | #include <linux/device.h> |
| 38 | #include <linux/module.h> |
| 39 | #include <linux/fs.h> |
| 40 | #include <linux/blkdev.h> |
| 41 | |
| 42 | #include "rbd_types.h" |
| 43 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 44 | /* |
| 45 | * The basic unit of block I/O is a sector. It is interpreted in a |
| 46 | * number of contexts in Linux (blk, bio, genhd), but the default is |
| 47 | * universally 512 bytes. These symbols are just slightly more |
| 48 | * meaningful than the bare numbers they represent. |
| 49 | */ |
| 50 | #define SECTOR_SHIFT 9 |
| 51 | #define SECTOR_SIZE (1ULL << SECTOR_SHIFT) |
| 52 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 53 | #define RBD_DRV_NAME "rbd" |
| 54 | #define RBD_DRV_NAME_LONG "rbd (rados block device)" |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 55 | |
| 56 | #define RBD_MINORS_PER_MAJOR 256 /* max minors per blkdev */ |
| 57 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 58 | #define RBD_MAX_SNAP_NAME_LEN 32 |
| 59 | #define RBD_MAX_OPT_LEN 1024 |
| 60 | |
| 61 | #define RBD_SNAP_HEAD_NAME "-" |
| 62 | |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 63 | /* |
| 64 | * An RBD device name will be "rbd#", where the "rbd" comes from |
| 65 | * RBD_DRV_NAME above, and # is a unique integer identifier. |
| 66 | * MAX_INT_FORMAT_WIDTH is used in ensuring DEV_NAME_LEN is big |
| 67 | * enough to hold all possible device names. |
| 68 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 69 | #define DEV_NAME_LEN 32 |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 70 | #define MAX_INT_FORMAT_WIDTH ((5 * sizeof (int)) / 2 + 1) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 71 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 72 | #define RBD_NOTIFY_TIMEOUT_DEFAULT 10 |
| 73 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 74 | /* |
| 75 | * block device image metadata (in-memory version) |
| 76 | */ |
| 77 | struct rbd_image_header { |
| 78 | u64 image_size; |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 79 | char *object_prefix; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 80 | __u8 obj_order; |
| 81 | __u8 crypt_type; |
| 82 | __u8 comp_type; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 83 | struct ceph_snap_context *snapc; |
| 84 | size_t snap_names_len; |
| 85 | u64 snap_seq; |
| 86 | u32 total_snaps; |
| 87 | |
| 88 | char *snap_names; |
| 89 | u64 *snap_sizes; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 90 | |
| 91 | u64 obj_version; |
| 92 | }; |
| 93 | |
| 94 | struct rbd_options { |
| 95 | int notify_timeout; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 96 | }; |
| 97 | |
| 98 | /* |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 99 | * an instance of the client. multiple devices may share an rbd client. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 100 | */ |
| 101 | struct rbd_client { |
| 102 | struct ceph_client *client; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 103 | struct rbd_options *rbd_opts; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 104 | struct kref kref; |
| 105 | struct list_head node; |
| 106 | }; |
| 107 | |
| 108 | /* |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 109 | * a request completion status |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 110 | */ |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 111 | struct rbd_req_status { |
| 112 | int done; |
| 113 | int rc; |
| 114 | u64 bytes; |
| 115 | }; |
| 116 | |
| 117 | /* |
| 118 | * a collection of requests |
| 119 | */ |
| 120 | struct rbd_req_coll { |
| 121 | int total; |
| 122 | int num_done; |
| 123 | struct kref kref; |
| 124 | struct rbd_req_status status[0]; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 125 | }; |
| 126 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 127 | /* |
| 128 | * a single io request |
| 129 | */ |
| 130 | struct rbd_request { |
| 131 | struct request *rq; /* blk layer request */ |
| 132 | struct bio *bio; /* cloned bio */ |
| 133 | struct page **pages; /* list of used pages */ |
| 134 | u64 len; |
| 135 | int coll_index; |
| 136 | struct rbd_req_coll *coll; |
| 137 | }; |
| 138 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 139 | struct rbd_snap { |
| 140 | struct device dev; |
| 141 | const char *name; |
Josh Durgin | 3591538 | 2011-12-05 18:25:13 -0800 | [diff] [blame] | 142 | u64 size; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 143 | struct list_head node; |
| 144 | u64 id; |
| 145 | }; |
| 146 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 147 | /* |
| 148 | * a single device |
| 149 | */ |
| 150 | struct rbd_device { |
| 151 | int id; /* blkdev unique id */ |
| 152 | |
| 153 | int major; /* blkdev assigned major */ |
| 154 | struct gendisk *disk; /* blkdev's gendisk and rq */ |
| 155 | struct request_queue *q; |
| 156 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 157 | struct rbd_client *rbd_client; |
| 158 | |
| 159 | char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ |
| 160 | |
| 161 | spinlock_t lock; /* queue lock */ |
| 162 | |
| 163 | struct rbd_image_header header; |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 164 | char *image_name; |
| 165 | size_t image_name_len; |
| 166 | char *header_name; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 167 | char *pool_name; |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 168 | int pool_id; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 169 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 170 | struct ceph_osd_event *watch_event; |
| 171 | struct ceph_osd_request *watch_request; |
| 172 | |
Josh Durgin | c666601 | 2011-11-21 17:11:12 -0800 | [diff] [blame] | 173 | /* protects updating the header */ |
| 174 | struct rw_semaphore header_rwsem; |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 175 | /* name of the snapshot this device reads from */ |
Alex Elder | 820a5f3 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 176 | char *snap_name; |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 177 | /* id of the snapshot this device reads from */ |
Josh Durgin | 77dfe99 | 2011-11-21 13:04:42 -0800 | [diff] [blame] | 178 | u64 snap_id; /* current snapshot id */ |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 179 | /* whether the snap_id this device reads from still exists */ |
| 180 | bool snap_exists; |
| 181 | int read_only; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 182 | |
| 183 | struct list_head node; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 184 | |
| 185 | /* list of snapshots */ |
| 186 | struct list_head snaps; |
| 187 | |
| 188 | /* sysfs related */ |
| 189 | struct device dev; |
| 190 | }; |
| 191 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 192 | static DEFINE_MUTEX(ctl_mutex); /* Serialize open/close/setup/teardown */ |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 193 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 194 | static LIST_HEAD(rbd_dev_list); /* devices */ |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 195 | static DEFINE_SPINLOCK(rbd_dev_list_lock); |
| 196 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 197 | static LIST_HEAD(rbd_client_list); /* clients */ |
| 198 | static DEFINE_SPINLOCK(rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 199 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 200 | static int __rbd_init_snaps_header(struct rbd_device *rbd_dev); |
| 201 | static void rbd_dev_release(struct device *dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 202 | static ssize_t rbd_snap_add(struct device *dev, |
| 203 | struct device_attribute *attr, |
| 204 | const char *buf, |
| 205 | size_t count); |
| 206 | static void __rbd_remove_snap_dev(struct rbd_device *rbd_dev, |
Justin P. Mattock | 6993248 | 2011-07-26 23:06:29 -0700 | [diff] [blame] | 207 | struct rbd_snap *snap); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 208 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 209 | static ssize_t rbd_add(struct bus_type *bus, const char *buf, |
| 210 | size_t count); |
| 211 | static ssize_t rbd_remove(struct bus_type *bus, const char *buf, |
| 212 | size_t count); |
| 213 | |
| 214 | static struct bus_attribute rbd_bus_attrs[] = { |
| 215 | __ATTR(add, S_IWUSR, NULL, rbd_add), |
| 216 | __ATTR(remove, S_IWUSR, NULL, rbd_remove), |
| 217 | __ATTR_NULL |
| 218 | }; |
| 219 | |
| 220 | static struct bus_type rbd_bus_type = { |
| 221 | .name = "rbd", |
| 222 | .bus_attrs = rbd_bus_attrs, |
| 223 | }; |
| 224 | |
| 225 | static void rbd_root_dev_release(struct device *dev) |
| 226 | { |
| 227 | } |
| 228 | |
| 229 | static struct device rbd_root_dev = { |
| 230 | .init_name = "rbd", |
| 231 | .release = rbd_root_dev_release, |
| 232 | }; |
| 233 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 234 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 235 | static struct device *rbd_get_dev(struct rbd_device *rbd_dev) |
| 236 | { |
| 237 | return get_device(&rbd_dev->dev); |
| 238 | } |
| 239 | |
| 240 | static void rbd_put_dev(struct rbd_device *rbd_dev) |
| 241 | { |
| 242 | put_device(&rbd_dev->dev); |
| 243 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 244 | |
Josh Durgin | 263c6ca | 2011-12-05 10:43:42 -0800 | [diff] [blame] | 245 | static int __rbd_refresh_header(struct rbd_device *rbd_dev); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 246 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 247 | static int rbd_open(struct block_device *bdev, fmode_t mode) |
| 248 | { |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 249 | struct rbd_device *rbd_dev = bdev->bd_disk->private_data; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 250 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 251 | rbd_get_dev(rbd_dev); |
| 252 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 253 | set_device_ro(bdev, rbd_dev->read_only); |
| 254 | |
| 255 | if ((mode & FMODE_WRITE) && rbd_dev->read_only) |
| 256 | return -EROFS; |
| 257 | |
| 258 | return 0; |
| 259 | } |
| 260 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 261 | static int rbd_release(struct gendisk *disk, fmode_t mode) |
| 262 | { |
| 263 | struct rbd_device *rbd_dev = disk->private_data; |
| 264 | |
| 265 | rbd_put_dev(rbd_dev); |
| 266 | |
| 267 | return 0; |
| 268 | } |
| 269 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 270 | static const struct block_device_operations rbd_bd_ops = { |
| 271 | .owner = THIS_MODULE, |
| 272 | .open = rbd_open, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 273 | .release = rbd_release, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 274 | }; |
| 275 | |
| 276 | /* |
| 277 | * Initialize an rbd client instance. |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 278 | * We own *ceph_opts. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 279 | */ |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 280 | static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 281 | struct rbd_options *rbd_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 282 | { |
| 283 | struct rbd_client *rbdc; |
| 284 | int ret = -ENOMEM; |
| 285 | |
| 286 | dout("rbd_client_create\n"); |
| 287 | rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); |
| 288 | if (!rbdc) |
| 289 | goto out_opt; |
| 290 | |
| 291 | kref_init(&rbdc->kref); |
| 292 | INIT_LIST_HEAD(&rbdc->node); |
| 293 | |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 294 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
| 295 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 296 | rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 297 | if (IS_ERR(rbdc->client)) |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 298 | goto out_mutex; |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 299 | ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 300 | |
| 301 | ret = ceph_open_session(rbdc->client); |
| 302 | if (ret < 0) |
| 303 | goto out_err; |
| 304 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 305 | rbdc->rbd_opts = rbd_opts; |
| 306 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 307 | spin_lock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 308 | list_add_tail(&rbdc->node, &rbd_client_list); |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 309 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 310 | |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 311 | mutex_unlock(&ctl_mutex); |
| 312 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 313 | dout("rbd_client_create created %p\n", rbdc); |
| 314 | return rbdc; |
| 315 | |
| 316 | out_err: |
| 317 | ceph_destroy_client(rbdc->client); |
Alex Elder | bc534d8 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 318 | out_mutex: |
| 319 | mutex_unlock(&ctl_mutex); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 320 | kfree(rbdc); |
| 321 | out_opt: |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 322 | if (ceph_opts) |
| 323 | ceph_destroy_options(ceph_opts); |
Vasiliy Kulikov | 28f259b | 2010-09-26 12:59:37 +0400 | [diff] [blame] | 324 | return ERR_PTR(ret); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 325 | } |
| 326 | |
| 327 | /* |
| 328 | * Find a ceph client with specific addr and configuration. |
| 329 | */ |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 330 | static struct rbd_client *__rbd_client_find(struct ceph_options *ceph_opts) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 331 | { |
| 332 | struct rbd_client *client_node; |
| 333 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 334 | if (ceph_opts->flags & CEPH_OPT_NOSHARE) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 335 | return NULL; |
| 336 | |
| 337 | list_for_each_entry(client_node, &rbd_client_list, node) |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 338 | if (!ceph_compare_options(ceph_opts, client_node->client)) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 339 | return client_node; |
| 340 | return NULL; |
| 341 | } |
| 342 | |
| 343 | /* |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 344 | * mount options |
| 345 | */ |
| 346 | enum { |
| 347 | Opt_notify_timeout, |
| 348 | Opt_last_int, |
| 349 | /* int args above */ |
| 350 | Opt_last_string, |
| 351 | /* string args above */ |
| 352 | }; |
| 353 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 354 | static match_table_t rbd_opts_tokens = { |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 355 | {Opt_notify_timeout, "notify_timeout=%d"}, |
| 356 | /* int args above */ |
| 357 | /* string args above */ |
| 358 | {-1, NULL} |
| 359 | }; |
| 360 | |
| 361 | static int parse_rbd_opts_token(char *c, void *private) |
| 362 | { |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 363 | struct rbd_options *rbd_opts = private; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 364 | substring_t argstr[MAX_OPT_ARGS]; |
| 365 | int token, intval, ret; |
| 366 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 367 | token = match_token(c, rbd_opts_tokens, argstr); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 368 | if (token < 0) |
| 369 | return -EINVAL; |
| 370 | |
| 371 | if (token < Opt_last_int) { |
| 372 | ret = match_int(&argstr[0], &intval); |
| 373 | if (ret < 0) { |
| 374 | pr_err("bad mount option arg (not int) " |
| 375 | "at '%s'\n", c); |
| 376 | return ret; |
| 377 | } |
| 378 | dout("got int token %d val %d\n", token, intval); |
| 379 | } else if (token > Opt_last_int && token < Opt_last_string) { |
| 380 | dout("got string token %d val %s\n", token, |
| 381 | argstr[0].from); |
| 382 | } else { |
| 383 | dout("got token %d\n", token); |
| 384 | } |
| 385 | |
| 386 | switch (token) { |
| 387 | case Opt_notify_timeout: |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 388 | rbd_opts->notify_timeout = intval; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 389 | break; |
| 390 | default: |
| 391 | BUG_ON(token); |
| 392 | } |
| 393 | return 0; |
| 394 | } |
| 395 | |
| 396 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 397 | * Get a ceph client with specific addr and configuration, if one does |
| 398 | * not exist create it. |
| 399 | */ |
Alex Elder | 5214ecc | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 400 | static struct rbd_client *rbd_get_client(const char *mon_addr, |
| 401 | size_t mon_addr_len, |
| 402 | char *options) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 403 | { |
| 404 | struct rbd_client *rbdc; |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 405 | struct ceph_options *ceph_opts; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 406 | struct rbd_options *rbd_opts; |
| 407 | |
| 408 | rbd_opts = kzalloc(sizeof(*rbd_opts), GFP_KERNEL); |
| 409 | if (!rbd_opts) |
Alex Elder | d720bcb | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 410 | return ERR_PTR(-ENOMEM); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 411 | |
| 412 | rbd_opts->notify_timeout = RBD_NOTIFY_TIMEOUT_DEFAULT; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 413 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 414 | ceph_opts = ceph_parse_options(options, mon_addr, |
| 415 | mon_addr + mon_addr_len, |
| 416 | parse_rbd_opts_token, rbd_opts); |
| 417 | if (IS_ERR(ceph_opts)) { |
Alex Elder | d720bcb | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 418 | kfree(rbd_opts); |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 419 | return ERR_CAST(ceph_opts); |
Alex Elder | ee57741 | 2012-01-24 10:08:36 -0600 | [diff] [blame] | 420 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 421 | |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 422 | spin_lock(&rbd_client_list_lock); |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 423 | rbdc = __rbd_client_find(ceph_opts); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 424 | if (rbdc) { |
Alex Elder | e6994d3 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 425 | /* using an existing client */ |
| 426 | kref_get(&rbdc->kref); |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 427 | spin_unlock(&rbd_client_list_lock); |
Alex Elder | e6994d3 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 428 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 429 | ceph_destroy_options(ceph_opts); |
Alex Elder | 97bb59a | 2012-01-24 10:08:36 -0600 | [diff] [blame] | 430 | kfree(rbd_opts); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 431 | |
Alex Elder | d720bcb | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 432 | return rbdc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 433 | } |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 434 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 435 | |
Alex Elder | 43ae470 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 436 | rbdc = rbd_client_create(ceph_opts, rbd_opts); |
Alex Elder | d97081b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 437 | |
Alex Elder | d720bcb | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 438 | if (IS_ERR(rbdc)) |
| 439 | kfree(rbd_opts); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 440 | |
Alex Elder | d720bcb | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 441 | return rbdc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 442 | } |
| 443 | |
| 444 | /* |
| 445 | * Destroy ceph client |
Alex Elder | d23a4b3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 446 | * |
Alex Elder | 432b858 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 447 | * Caller must hold rbd_client_list_lock. |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 448 | */ |
| 449 | static void rbd_client_release(struct kref *kref) |
| 450 | { |
| 451 | struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); |
| 452 | |
| 453 | dout("rbd_release_client %p\n", rbdc); |
Alex Elder | cd9d9f5 | 2012-04-04 13:35:44 -0500 | [diff] [blame] | 454 | spin_lock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 455 | list_del(&rbdc->node); |
Alex Elder | cd9d9f5 | 2012-04-04 13:35:44 -0500 | [diff] [blame] | 456 | spin_unlock(&rbd_client_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 457 | |
| 458 | ceph_destroy_client(rbdc->client); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 459 | kfree(rbdc->rbd_opts); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 460 | kfree(rbdc); |
| 461 | } |
| 462 | |
| 463 | /* |
| 464 | * Drop reference to ceph client node. If it's not referenced anymore, release |
| 465 | * it. |
| 466 | */ |
| 467 | static void rbd_put_client(struct rbd_device *rbd_dev) |
| 468 | { |
| 469 | kref_put(&rbd_dev->rbd_client->kref, rbd_client_release); |
| 470 | rbd_dev->rbd_client = NULL; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 471 | } |
| 472 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 473 | /* |
| 474 | * Destroy requests collection |
| 475 | */ |
| 476 | static void rbd_coll_release(struct kref *kref) |
| 477 | { |
| 478 | struct rbd_req_coll *coll = |
| 479 | container_of(kref, struct rbd_req_coll, kref); |
| 480 | |
| 481 | dout("rbd_coll_release %p\n", coll); |
| 482 | kfree(coll); |
| 483 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 484 | |
| 485 | /* |
| 486 | * Create a new header structure, translate header format from the on-disk |
| 487 | * header. |
| 488 | */ |
| 489 | static int rbd_header_from_disk(struct rbd_image_header *header, |
| 490 | struct rbd_image_header_ondisk *ondisk, |
Xi Wang | 50f7c4c | 2012-04-20 15:49:44 -0500 | [diff] [blame] | 491 | u32 allocated_snaps, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 492 | gfp_t gfp_flags) |
| 493 | { |
Xi Wang | 50f7c4c | 2012-04-20 15:49:44 -0500 | [diff] [blame] | 494 | u32 i, snap_count; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 495 | |
Alex Elder | 2107978 | 2012-01-24 10:08:36 -0600 | [diff] [blame] | 496 | if (memcmp(ondisk, RBD_HEADER_TEXT, sizeof(RBD_HEADER_TEXT))) |
Josh Durgin | 81e759f | 2011-11-15 14:49:53 -0800 | [diff] [blame] | 497 | return -ENXIO; |
Josh Durgin | 81e759f | 2011-11-15 14:49:53 -0800 | [diff] [blame] | 498 | |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 499 | snap_count = le32_to_cpu(ondisk->snap_count); |
Xi Wang | 50f7c4c | 2012-04-20 15:49:44 -0500 | [diff] [blame] | 500 | if (snap_count > (UINT_MAX - sizeof(struct ceph_snap_context)) |
| 501 | / sizeof (*ondisk)) |
| 502 | return -EINVAL; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 503 | header->snapc = kmalloc(sizeof(struct ceph_snap_context) + |
Yan, Zheng | f9f9a19 | 2012-06-06 09:15:33 -0500 | [diff] [blame] | 504 | snap_count * sizeof(u64), |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 505 | gfp_flags); |
| 506 | if (!header->snapc) |
| 507 | return -ENOMEM; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 508 | |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 509 | header->snap_names_len = le64_to_cpu(ondisk->snap_names_len); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 510 | if (snap_count) { |
| 511 | header->snap_names = kmalloc(header->snap_names_len, |
Dan Carpenter | f8ad495 | 2012-04-20 15:49:44 -0500 | [diff] [blame] | 512 | gfp_flags); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 513 | if (!header->snap_names) |
| 514 | goto err_snapc; |
| 515 | header->snap_sizes = kmalloc(snap_count * sizeof(u64), |
Dan Carpenter | f8ad495 | 2012-04-20 15:49:44 -0500 | [diff] [blame] | 516 | gfp_flags); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 517 | if (!header->snap_sizes) |
| 518 | goto err_names; |
| 519 | } else { |
| 520 | header->snap_names = NULL; |
| 521 | header->snap_sizes = NULL; |
| 522 | } |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 523 | |
| 524 | header->object_prefix = kmalloc(sizeof (ondisk->block_name) + 1, |
| 525 | gfp_flags); |
| 526 | if (!header->object_prefix) |
| 527 | goto err_sizes; |
| 528 | |
Alex Elder | ca1e49a | 2012-07-10 20:30:09 -0500 | [diff] [blame] | 529 | memcpy(header->object_prefix, ondisk->block_name, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 530 | sizeof(ondisk->block_name)); |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 531 | header->object_prefix[sizeof (ondisk->block_name)] = '\0'; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 532 | |
| 533 | header->image_size = le64_to_cpu(ondisk->image_size); |
| 534 | header->obj_order = ondisk->options.order; |
| 535 | header->crypt_type = ondisk->options.crypt_type; |
| 536 | header->comp_type = ondisk->options.comp_type; |
| 537 | |
| 538 | atomic_set(&header->snapc->nref, 1); |
| 539 | header->snap_seq = le64_to_cpu(ondisk->snap_seq); |
| 540 | header->snapc->num_snaps = snap_count; |
| 541 | header->total_snaps = snap_count; |
| 542 | |
Alex Elder | 2107978 | 2012-01-24 10:08:36 -0600 | [diff] [blame] | 543 | if (snap_count && allocated_snaps == snap_count) { |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 544 | for (i = 0; i < snap_count; i++) { |
| 545 | header->snapc->snaps[i] = |
| 546 | le64_to_cpu(ondisk->snaps[i].id); |
| 547 | header->snap_sizes[i] = |
| 548 | le64_to_cpu(ondisk->snaps[i].image_size); |
| 549 | } |
| 550 | |
| 551 | /* copy snapshot names */ |
| 552 | memcpy(header->snap_names, &ondisk->snaps[i], |
| 553 | header->snap_names_len); |
| 554 | } |
| 555 | |
| 556 | return 0; |
| 557 | |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 558 | err_sizes: |
| 559 | kfree(header->snap_sizes); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 560 | err_names: |
| 561 | kfree(header->snap_names); |
| 562 | err_snapc: |
| 563 | kfree(header->snapc); |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 564 | return -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 565 | } |
| 566 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 567 | static int snap_by_name(struct rbd_image_header *header, const char *snap_name, |
| 568 | u64 *seq, u64 *size) |
| 569 | { |
| 570 | int i; |
| 571 | char *p = header->snap_names; |
| 572 | |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 573 | for (i = 0; i < header->total_snaps; i++) { |
| 574 | if (!strcmp(snap_name, p)) { |
| 575 | |
| 576 | /* Found it. Pass back its id and/or size */ |
| 577 | |
| 578 | if (seq) |
| 579 | *seq = header->snapc->snaps[i]; |
| 580 | if (size) |
| 581 | *size = header->snap_sizes[i]; |
| 582 | return i; |
| 583 | } |
| 584 | p += strlen(p) + 1; /* Skip ahead to the next name */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 585 | } |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 586 | return -ENOENT; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 587 | } |
| 588 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 589 | static int rbd_header_set_snap(struct rbd_device *rbd_dev, u64 *size) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 590 | { |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 591 | struct rbd_image_header *header = &rbd_dev->header; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 592 | struct ceph_snap_context *snapc = header->snapc; |
| 593 | int ret = -ENOENT; |
| 594 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 595 | down_write(&rbd_dev->header_rwsem); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 596 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 597 | if (!memcmp(rbd_dev->snap_name, RBD_SNAP_HEAD_NAME, |
Josh Durgin | cc9d734 | 2011-11-21 18:19:13 -0800 | [diff] [blame] | 598 | sizeof (RBD_SNAP_HEAD_NAME))) { |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 599 | if (header->total_snaps) |
| 600 | snapc->seq = header->snap_seq; |
| 601 | else |
| 602 | snapc->seq = 0; |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 603 | rbd_dev->snap_id = CEPH_NOSNAP; |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 604 | rbd_dev->snap_exists = false; |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 605 | rbd_dev->read_only = 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 606 | if (size) |
| 607 | *size = header->image_size; |
| 608 | } else { |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 609 | ret = snap_by_name(header, rbd_dev->snap_name, |
| 610 | &snapc->seq, size); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 611 | if (ret < 0) |
| 612 | goto done; |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 613 | rbd_dev->snap_id = snapc->seq; |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 614 | rbd_dev->snap_exists = true; |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 615 | rbd_dev->read_only = 1; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 616 | } |
| 617 | |
| 618 | ret = 0; |
| 619 | done: |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 620 | up_write(&rbd_dev->header_rwsem); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 621 | return ret; |
| 622 | } |
| 623 | |
| 624 | static void rbd_header_free(struct rbd_image_header *header) |
| 625 | { |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 626 | kfree(header->object_prefix); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 627 | kfree(header->snap_sizes); |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 628 | kfree(header->snap_names); |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 629 | ceph_put_snap_context(header->snapc); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 630 | } |
| 631 | |
| 632 | /* |
| 633 | * get the actual striped segment name, offset and length |
| 634 | */ |
| 635 | static u64 rbd_get_segment(struct rbd_image_header *header, |
Alex Elder | ca1e49a | 2012-07-10 20:30:09 -0500 | [diff] [blame] | 636 | const char *object_prefix, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 637 | u64 ofs, u64 len, |
| 638 | char *seg_name, u64 *segofs) |
| 639 | { |
| 640 | u64 seg = ofs >> header->obj_order; |
| 641 | |
| 642 | if (seg_name) |
| 643 | snprintf(seg_name, RBD_MAX_SEG_NAME_LEN, |
Alex Elder | ca1e49a | 2012-07-10 20:30:09 -0500 | [diff] [blame] | 644 | "%s.%012llx", object_prefix, seg); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 645 | |
| 646 | ofs = ofs & ((1 << header->obj_order) - 1); |
| 647 | len = min_t(u64, len, (1 << header->obj_order) - ofs); |
| 648 | |
| 649 | if (segofs) |
| 650 | *segofs = ofs; |
| 651 | |
| 652 | return len; |
| 653 | } |
| 654 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 655 | static int rbd_get_num_segments(struct rbd_image_header *header, |
| 656 | u64 ofs, u64 len) |
| 657 | { |
| 658 | u64 start_seg = ofs >> header->obj_order; |
| 659 | u64 end_seg = (ofs + len - 1) >> header->obj_order; |
| 660 | return end_seg - start_seg + 1; |
| 661 | } |
| 662 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 663 | /* |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 664 | * returns the size of an object in the image |
| 665 | */ |
| 666 | static u64 rbd_obj_bytes(struct rbd_image_header *header) |
| 667 | { |
| 668 | return 1 << header->obj_order; |
| 669 | } |
| 670 | |
| 671 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 672 | * bio helpers |
| 673 | */ |
| 674 | |
| 675 | static void bio_chain_put(struct bio *chain) |
| 676 | { |
| 677 | struct bio *tmp; |
| 678 | |
| 679 | while (chain) { |
| 680 | tmp = chain; |
| 681 | chain = chain->bi_next; |
| 682 | bio_put(tmp); |
| 683 | } |
| 684 | } |
| 685 | |
| 686 | /* |
| 687 | * zeros a bio chain, starting at specific offset |
| 688 | */ |
| 689 | static void zero_bio_chain(struct bio *chain, int start_ofs) |
| 690 | { |
| 691 | struct bio_vec *bv; |
| 692 | unsigned long flags; |
| 693 | void *buf; |
| 694 | int i; |
| 695 | int pos = 0; |
| 696 | |
| 697 | while (chain) { |
| 698 | bio_for_each_segment(bv, chain, i) { |
| 699 | if (pos + bv->bv_len > start_ofs) { |
| 700 | int remainder = max(start_ofs - pos, 0); |
| 701 | buf = bvec_kmap_irq(bv, &flags); |
| 702 | memset(buf + remainder, 0, |
| 703 | bv->bv_len - remainder); |
Dan Carpenter | 85b5aaa | 2010-10-11 21:15:11 +0200 | [diff] [blame] | 704 | bvec_kunmap_irq(buf, &flags); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 705 | } |
| 706 | pos += bv->bv_len; |
| 707 | } |
| 708 | |
| 709 | chain = chain->bi_next; |
| 710 | } |
| 711 | } |
| 712 | |
| 713 | /* |
| 714 | * bio_chain_clone - clone a chain of bios up to a certain length. |
| 715 | * might return a bio_pair that will need to be released. |
| 716 | */ |
| 717 | static struct bio *bio_chain_clone(struct bio **old, struct bio **next, |
| 718 | struct bio_pair **bp, |
| 719 | int len, gfp_t gfpmask) |
| 720 | { |
| 721 | struct bio *tmp, *old_chain = *old, *new_chain = NULL, *tail = NULL; |
| 722 | int total = 0; |
| 723 | |
| 724 | if (*bp) { |
| 725 | bio_pair_release(*bp); |
| 726 | *bp = NULL; |
| 727 | } |
| 728 | |
| 729 | while (old_chain && (total < len)) { |
| 730 | tmp = bio_kmalloc(gfpmask, old_chain->bi_max_vecs); |
| 731 | if (!tmp) |
| 732 | goto err_out; |
| 733 | |
| 734 | if (total + old_chain->bi_size > len) { |
| 735 | struct bio_pair *bp; |
| 736 | |
| 737 | /* |
| 738 | * this split can only happen with a single paged bio, |
| 739 | * split_bio will BUG_ON if this is not the case |
| 740 | */ |
| 741 | dout("bio_chain_clone split! total=%d remaining=%d" |
| 742 | "bi_size=%d\n", |
| 743 | (int)total, (int)len-total, |
| 744 | (int)old_chain->bi_size); |
| 745 | |
| 746 | /* split the bio. We'll release it either in the next |
| 747 | call, or it will have to be released outside */ |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 748 | bp = bio_split(old_chain, (len - total) / SECTOR_SIZE); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 749 | if (!bp) |
| 750 | goto err_out; |
| 751 | |
| 752 | __bio_clone(tmp, &bp->bio1); |
| 753 | |
| 754 | *next = &bp->bio2; |
| 755 | } else { |
| 756 | __bio_clone(tmp, old_chain); |
| 757 | *next = old_chain->bi_next; |
| 758 | } |
| 759 | |
| 760 | tmp->bi_bdev = NULL; |
| 761 | gfpmask &= ~__GFP_WAIT; |
| 762 | tmp->bi_next = NULL; |
| 763 | |
| 764 | if (!new_chain) { |
| 765 | new_chain = tail = tmp; |
| 766 | } else { |
| 767 | tail->bi_next = tmp; |
| 768 | tail = tmp; |
| 769 | } |
| 770 | old_chain = old_chain->bi_next; |
| 771 | |
| 772 | total += tmp->bi_size; |
| 773 | } |
| 774 | |
| 775 | BUG_ON(total < len); |
| 776 | |
| 777 | if (tail) |
| 778 | tail->bi_next = NULL; |
| 779 | |
| 780 | *old = old_chain; |
| 781 | |
| 782 | return new_chain; |
| 783 | |
| 784 | err_out: |
| 785 | dout("bio_chain_clone with err\n"); |
| 786 | bio_chain_put(new_chain); |
| 787 | return NULL; |
| 788 | } |
| 789 | |
| 790 | /* |
| 791 | * helpers for osd request op vectors. |
| 792 | */ |
| 793 | static int rbd_create_rw_ops(struct ceph_osd_req_op **ops, |
| 794 | int num_ops, |
| 795 | int opcode, |
| 796 | u32 payload_len) |
| 797 | { |
| 798 | *ops = kzalloc(sizeof(struct ceph_osd_req_op) * (num_ops + 1), |
| 799 | GFP_NOIO); |
| 800 | if (!*ops) |
| 801 | return -ENOMEM; |
| 802 | (*ops)[0].op = opcode; |
| 803 | /* |
| 804 | * op extent offset and length will be set later on |
| 805 | * in calc_raw_layout() |
| 806 | */ |
| 807 | (*ops)[0].payload_len = payload_len; |
| 808 | return 0; |
| 809 | } |
| 810 | |
| 811 | static void rbd_destroy_ops(struct ceph_osd_req_op *ops) |
| 812 | { |
| 813 | kfree(ops); |
| 814 | } |
| 815 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 816 | static void rbd_coll_end_req_index(struct request *rq, |
| 817 | struct rbd_req_coll *coll, |
| 818 | int index, |
| 819 | int ret, u64 len) |
| 820 | { |
| 821 | struct request_queue *q; |
| 822 | int min, max, i; |
| 823 | |
| 824 | dout("rbd_coll_end_req_index %p index %d ret %d len %lld\n", |
| 825 | coll, index, ret, len); |
| 826 | |
| 827 | if (!rq) |
| 828 | return; |
| 829 | |
| 830 | if (!coll) { |
| 831 | blk_end_request(rq, ret, len); |
| 832 | return; |
| 833 | } |
| 834 | |
| 835 | q = rq->q; |
| 836 | |
| 837 | spin_lock_irq(q->queue_lock); |
| 838 | coll->status[index].done = 1; |
| 839 | coll->status[index].rc = ret; |
| 840 | coll->status[index].bytes = len; |
| 841 | max = min = coll->num_done; |
| 842 | while (max < coll->total && coll->status[max].done) |
| 843 | max++; |
| 844 | |
| 845 | for (i = min; i<max; i++) { |
| 846 | __blk_end_request(rq, coll->status[i].rc, |
| 847 | coll->status[i].bytes); |
| 848 | coll->num_done++; |
| 849 | kref_put(&coll->kref, rbd_coll_release); |
| 850 | } |
| 851 | spin_unlock_irq(q->queue_lock); |
| 852 | } |
| 853 | |
| 854 | static void rbd_coll_end_req(struct rbd_request *req, |
| 855 | int ret, u64 len) |
| 856 | { |
| 857 | rbd_coll_end_req_index(req->rq, req->coll, req->coll_index, ret, len); |
| 858 | } |
| 859 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 860 | /* |
| 861 | * Send ceph osd request |
| 862 | */ |
| 863 | static int rbd_do_request(struct request *rq, |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 864 | struct rbd_device *rbd_dev, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 865 | struct ceph_snap_context *snapc, |
| 866 | u64 snapid, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 867 | const char *object_name, u64 ofs, u64 len, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 868 | struct bio *bio, |
| 869 | struct page **pages, |
| 870 | int num_pages, |
| 871 | int flags, |
| 872 | struct ceph_osd_req_op *ops, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 873 | struct rbd_req_coll *coll, |
| 874 | int coll_index, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 875 | void (*rbd_cb)(struct ceph_osd_request *req, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 876 | struct ceph_msg *msg), |
| 877 | struct ceph_osd_request **linger_req, |
| 878 | u64 *ver) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 879 | { |
| 880 | struct ceph_osd_request *req; |
| 881 | struct ceph_file_layout *layout; |
| 882 | int ret; |
| 883 | u64 bno; |
| 884 | struct timespec mtime = CURRENT_TIME; |
| 885 | struct rbd_request *req_data; |
| 886 | struct ceph_osd_request_head *reqhead; |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 887 | struct ceph_osd_client *osdc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 888 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 889 | req_data = kzalloc(sizeof(*req_data), GFP_NOIO); |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 890 | if (!req_data) { |
| 891 | if (coll) |
| 892 | rbd_coll_end_req_index(rq, coll, coll_index, |
| 893 | -ENOMEM, len); |
| 894 | return -ENOMEM; |
| 895 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 896 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 897 | if (coll) { |
| 898 | req_data->coll = coll; |
| 899 | req_data->coll_index = coll_index; |
| 900 | } |
| 901 | |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 902 | dout("rbd_do_request object_name=%s ofs=%lld len=%lld\n", |
| 903 | object_name, len, ofs); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 904 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 905 | osdc = &rbd_dev->rbd_client->client->osdc; |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 906 | req = ceph_osdc_alloc_request(osdc, flags, snapc, ops, |
| 907 | false, GFP_NOIO, pages, bio); |
Sage Weil | 4ad1262 | 2011-05-03 09:23:36 -0700 | [diff] [blame] | 908 | if (!req) { |
Sage Weil | 4ad1262 | 2011-05-03 09:23:36 -0700 | [diff] [blame] | 909 | ret = -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 910 | goto done_pages; |
| 911 | } |
| 912 | |
| 913 | req->r_callback = rbd_cb; |
| 914 | |
| 915 | req_data->rq = rq; |
| 916 | req_data->bio = bio; |
| 917 | req_data->pages = pages; |
| 918 | req_data->len = len; |
| 919 | |
| 920 | req->r_priv = req_data; |
| 921 | |
| 922 | reqhead = req->r_request->front.iov_base; |
| 923 | reqhead->snapid = cpu_to_le64(CEPH_NOSNAP); |
| 924 | |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 925 | strncpy(req->r_oid, object_name, sizeof(req->r_oid)); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 926 | req->r_oid_len = strlen(req->r_oid); |
| 927 | |
| 928 | layout = &req->r_file_layout; |
| 929 | memset(layout, 0, sizeof(*layout)); |
| 930 | layout->fl_stripe_unit = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); |
| 931 | layout->fl_stripe_count = cpu_to_le32(1); |
| 932 | layout->fl_object_size = cpu_to_le32(1 << RBD_MAX_OBJ_ORDER); |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 933 | layout->fl_pg_pool = cpu_to_le32(rbd_dev->pool_id); |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 934 | ceph_calc_raw_layout(osdc, layout, snapid, ofs, &len, &bno, |
| 935 | req, ops); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 936 | |
| 937 | ceph_osdc_build_request(req, ofs, &len, |
| 938 | ops, |
| 939 | snapc, |
| 940 | &mtime, |
| 941 | req->r_oid, req->r_oid_len); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 942 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 943 | if (linger_req) { |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 944 | ceph_osdc_set_request_linger(osdc, req); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 945 | *linger_req = req; |
| 946 | } |
| 947 | |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 948 | ret = ceph_osdc_start_request(osdc, req, false); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 949 | if (ret < 0) |
| 950 | goto done_err; |
| 951 | |
| 952 | if (!rbd_cb) { |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 953 | ret = ceph_osdc_wait_request(osdc, req); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 954 | if (ver) |
| 955 | *ver = le64_to_cpu(req->r_reassert_version.version); |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 956 | dout("reassert_ver=%lld\n", |
| 957 | le64_to_cpu(req->r_reassert_version.version)); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 958 | ceph_osdc_put_request(req); |
| 959 | } |
| 960 | return ret; |
| 961 | |
| 962 | done_err: |
| 963 | bio_chain_put(req_data->bio); |
| 964 | ceph_osdc_put_request(req); |
| 965 | done_pages: |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 966 | rbd_coll_end_req(req_data, ret, len); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 967 | kfree(req_data); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 968 | return ret; |
| 969 | } |
| 970 | |
| 971 | /* |
| 972 | * Ceph osd op callback |
| 973 | */ |
| 974 | static void rbd_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) |
| 975 | { |
| 976 | struct rbd_request *req_data = req->r_priv; |
| 977 | struct ceph_osd_reply_head *replyhead; |
| 978 | struct ceph_osd_op *op; |
| 979 | __s32 rc; |
| 980 | u64 bytes; |
| 981 | int read_op; |
| 982 | |
| 983 | /* parse reply */ |
| 984 | replyhead = msg->front.iov_base; |
| 985 | WARN_ON(le32_to_cpu(replyhead->num_ops) == 0); |
| 986 | op = (void *)(replyhead + 1); |
| 987 | rc = le32_to_cpu(replyhead->result); |
| 988 | bytes = le64_to_cpu(op->extent.length); |
Dan Carpenter | 895cfcc | 2012-06-06 09:15:33 -0500 | [diff] [blame] | 989 | read_op = (le16_to_cpu(op->op) == CEPH_OSD_OP_READ); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 990 | |
| 991 | dout("rbd_req_cb bytes=%lld readop=%d rc=%d\n", bytes, read_op, rc); |
| 992 | |
| 993 | if (rc == -ENOENT && read_op) { |
| 994 | zero_bio_chain(req_data->bio, 0); |
| 995 | rc = 0; |
| 996 | } else if (rc == 0 && read_op && bytes < req_data->len) { |
| 997 | zero_bio_chain(req_data->bio, bytes); |
| 998 | bytes = req_data->len; |
| 999 | } |
| 1000 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1001 | rbd_coll_end_req(req_data, rc, bytes); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1002 | |
| 1003 | if (req_data->bio) |
| 1004 | bio_chain_put(req_data->bio); |
| 1005 | |
| 1006 | ceph_osdc_put_request(req); |
| 1007 | kfree(req_data); |
| 1008 | } |
| 1009 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1010 | static void rbd_simple_req_cb(struct ceph_osd_request *req, struct ceph_msg *msg) |
| 1011 | { |
| 1012 | ceph_osdc_put_request(req); |
| 1013 | } |
| 1014 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1015 | /* |
| 1016 | * Do a synchronous ceph osd operation |
| 1017 | */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1018 | static int rbd_req_sync_op(struct rbd_device *rbd_dev, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1019 | struct ceph_snap_context *snapc, |
| 1020 | u64 snapid, |
| 1021 | int opcode, |
| 1022 | int flags, |
| 1023 | struct ceph_osd_req_op *orig_ops, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1024 | const char *object_name, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1025 | u64 ofs, u64 len, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1026 | char *buf, |
| 1027 | struct ceph_osd_request **linger_req, |
| 1028 | u64 *ver) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1029 | { |
| 1030 | int ret; |
| 1031 | struct page **pages; |
| 1032 | int num_pages; |
| 1033 | struct ceph_osd_req_op *ops = orig_ops; |
| 1034 | u32 payload_len; |
| 1035 | |
| 1036 | num_pages = calc_pages_for(ofs , len); |
| 1037 | pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); |
Dan Carpenter | b8d0638 | 2010-10-11 21:14:23 +0200 | [diff] [blame] | 1038 | if (IS_ERR(pages)) |
| 1039 | return PTR_ERR(pages); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1040 | |
| 1041 | if (!orig_ops) { |
| 1042 | payload_len = (flags & CEPH_OSD_FLAG_WRITE ? len : 0); |
| 1043 | ret = rbd_create_rw_ops(&ops, 1, opcode, payload_len); |
| 1044 | if (ret < 0) |
| 1045 | goto done; |
| 1046 | |
| 1047 | if ((flags & CEPH_OSD_FLAG_WRITE) && buf) { |
| 1048 | ret = ceph_copy_to_page_vector(pages, buf, ofs, len); |
| 1049 | if (ret < 0) |
| 1050 | goto done_ops; |
| 1051 | } |
| 1052 | } |
| 1053 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1054 | ret = rbd_do_request(NULL, rbd_dev, snapc, snapid, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1055 | object_name, ofs, len, NULL, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1056 | pages, num_pages, |
| 1057 | flags, |
| 1058 | ops, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1059 | NULL, 0, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1060 | NULL, |
| 1061 | linger_req, ver); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1062 | if (ret < 0) |
| 1063 | goto done_ops; |
| 1064 | |
| 1065 | if ((flags & CEPH_OSD_FLAG_READ) && buf) |
| 1066 | ret = ceph_copy_from_page_vector(pages, buf, ofs, ret); |
| 1067 | |
| 1068 | done_ops: |
| 1069 | if (!orig_ops) |
| 1070 | rbd_destroy_ops(ops); |
| 1071 | done: |
| 1072 | ceph_release_page_vector(pages, num_pages); |
| 1073 | return ret; |
| 1074 | } |
| 1075 | |
| 1076 | /* |
| 1077 | * Do an asynchronous ceph osd operation |
| 1078 | */ |
| 1079 | static int rbd_do_op(struct request *rq, |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1080 | struct rbd_device *rbd_dev, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1081 | struct ceph_snap_context *snapc, |
| 1082 | u64 snapid, |
Alex Elder | d1f57ea | 2012-06-26 12:57:03 -0700 | [diff] [blame] | 1083 | int opcode, int flags, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1084 | u64 ofs, u64 len, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1085 | struct bio *bio, |
| 1086 | struct rbd_req_coll *coll, |
| 1087 | int coll_index) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1088 | { |
| 1089 | char *seg_name; |
| 1090 | u64 seg_ofs; |
| 1091 | u64 seg_len; |
| 1092 | int ret; |
| 1093 | struct ceph_osd_req_op *ops; |
| 1094 | u32 payload_len; |
| 1095 | |
| 1096 | seg_name = kmalloc(RBD_MAX_SEG_NAME_LEN + 1, GFP_NOIO); |
| 1097 | if (!seg_name) |
| 1098 | return -ENOMEM; |
| 1099 | |
| 1100 | seg_len = rbd_get_segment(&rbd_dev->header, |
Alex Elder | ca1e49a | 2012-07-10 20:30:09 -0500 | [diff] [blame] | 1101 | rbd_dev->header.object_prefix, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1102 | ofs, len, |
| 1103 | seg_name, &seg_ofs); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1104 | |
| 1105 | payload_len = (flags & CEPH_OSD_FLAG_WRITE ? seg_len : 0); |
| 1106 | |
| 1107 | ret = rbd_create_rw_ops(&ops, 1, opcode, payload_len); |
| 1108 | if (ret < 0) |
| 1109 | goto done; |
| 1110 | |
| 1111 | /* we've taken care of segment sizes earlier when we |
| 1112 | cloned the bios. We should never have a segment |
| 1113 | truncated at this point */ |
| 1114 | BUG_ON(seg_len < len); |
| 1115 | |
| 1116 | ret = rbd_do_request(rq, rbd_dev, snapc, snapid, |
| 1117 | seg_name, seg_ofs, seg_len, |
| 1118 | bio, |
| 1119 | NULL, 0, |
| 1120 | flags, |
| 1121 | ops, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1122 | coll, coll_index, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1123 | rbd_req_cb, 0, NULL); |
Sage Weil | 11f7700 | 2011-05-12 16:13:54 -0700 | [diff] [blame] | 1124 | |
| 1125 | rbd_destroy_ops(ops); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1126 | done: |
| 1127 | kfree(seg_name); |
| 1128 | return ret; |
| 1129 | } |
| 1130 | |
| 1131 | /* |
| 1132 | * Request async osd write |
| 1133 | */ |
| 1134 | static int rbd_req_write(struct request *rq, |
| 1135 | struct rbd_device *rbd_dev, |
| 1136 | struct ceph_snap_context *snapc, |
| 1137 | u64 ofs, u64 len, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1138 | struct bio *bio, |
| 1139 | struct rbd_req_coll *coll, |
| 1140 | int coll_index) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1141 | { |
| 1142 | return rbd_do_op(rq, rbd_dev, snapc, CEPH_NOSNAP, |
| 1143 | CEPH_OSD_OP_WRITE, |
| 1144 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1145 | ofs, len, bio, coll, coll_index); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1146 | } |
| 1147 | |
| 1148 | /* |
| 1149 | * Request async osd read |
| 1150 | */ |
| 1151 | static int rbd_req_read(struct request *rq, |
| 1152 | struct rbd_device *rbd_dev, |
| 1153 | u64 snapid, |
| 1154 | u64 ofs, u64 len, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1155 | struct bio *bio, |
| 1156 | struct rbd_req_coll *coll, |
| 1157 | int coll_index) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1158 | { |
| 1159 | return rbd_do_op(rq, rbd_dev, NULL, |
Josh Durgin | b06e6a6 | 2011-11-21 18:16:52 -0800 | [diff] [blame] | 1160 | snapid, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1161 | CEPH_OSD_OP_READ, |
| 1162 | CEPH_OSD_FLAG_READ, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1163 | ofs, len, bio, coll, coll_index); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1164 | } |
| 1165 | |
| 1166 | /* |
| 1167 | * Request sync osd read |
| 1168 | */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1169 | static int rbd_req_sync_read(struct rbd_device *rbd_dev, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1170 | struct ceph_snap_context *snapc, |
| 1171 | u64 snapid, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1172 | const char *object_name, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1173 | u64 ofs, u64 len, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1174 | char *buf, |
| 1175 | u64 *ver) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1176 | { |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1177 | return rbd_req_sync_op(rbd_dev, NULL, |
Josh Durgin | b06e6a6 | 2011-11-21 18:16:52 -0800 | [diff] [blame] | 1178 | snapid, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1179 | CEPH_OSD_OP_READ, |
| 1180 | CEPH_OSD_FLAG_READ, |
| 1181 | NULL, |
Alex Elder | d1f57ea | 2012-06-26 12:57:03 -0700 | [diff] [blame] | 1182 | object_name, ofs, len, buf, NULL, ver); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1183 | } |
| 1184 | |
| 1185 | /* |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1186 | * Request sync osd watch |
| 1187 | */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1188 | static int rbd_req_sync_notify_ack(struct rbd_device *rbd_dev, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1189 | u64 ver, |
| 1190 | u64 notify_id, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1191 | const char *object_name) |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1192 | { |
| 1193 | struct ceph_osd_req_op *ops; |
Sage Weil | 11f7700 | 2011-05-12 16:13:54 -0700 | [diff] [blame] | 1194 | int ret; |
| 1195 | |
| 1196 | ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_NOTIFY_ACK, 0); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1197 | if (ret < 0) |
| 1198 | return ret; |
| 1199 | |
Josh Durgin | a71b891 | 2011-12-05 18:10:44 -0800 | [diff] [blame^] | 1200 | ops[0].watch.ver = cpu_to_le64(ver); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1201 | ops[0].watch.cookie = notify_id; |
| 1202 | ops[0].watch.flag = 0; |
| 1203 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1204 | ret = rbd_do_request(NULL, rbd_dev, NULL, CEPH_NOSNAP, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1205 | object_name, 0, 0, NULL, |
Alex Elder | ad4f232 | 2012-07-03 16:01:19 -0500 | [diff] [blame] | 1206 | NULL, 0, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1207 | CEPH_OSD_FLAG_READ, |
| 1208 | ops, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1209 | NULL, 0, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1210 | rbd_simple_req_cb, 0, NULL); |
| 1211 | |
| 1212 | rbd_destroy_ops(ops); |
| 1213 | return ret; |
| 1214 | } |
| 1215 | |
| 1216 | static void rbd_watch_cb(u64 ver, u64 notify_id, u8 opcode, void *data) |
| 1217 | { |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1218 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
Josh Durgin | a71b891 | 2011-12-05 18:10:44 -0800 | [diff] [blame^] | 1219 | u64 hver; |
Sage Weil | 13143d2 | 2011-05-12 16:08:30 -0700 | [diff] [blame] | 1220 | int rc; |
| 1221 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1222 | if (!rbd_dev) |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1223 | return; |
| 1224 | |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1225 | dout("rbd_watch_cb %s notify_id=%lld opcode=%d\n", |
| 1226 | rbd_dev->header_name, notify_id, (int) opcode); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1227 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1228 | rc = __rbd_refresh_header(rbd_dev); |
Josh Durgin | a71b891 | 2011-12-05 18:10:44 -0800 | [diff] [blame^] | 1229 | hver = rbd_dev->header.obj_version; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1230 | mutex_unlock(&ctl_mutex); |
Sage Weil | 13143d2 | 2011-05-12 16:08:30 -0700 | [diff] [blame] | 1231 | if (rc) |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 1232 | pr_warning(RBD_DRV_NAME "%d got notification but failed to " |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1233 | " update snaps: %d\n", rbd_dev->major, rc); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1234 | |
Josh Durgin | a71b891 | 2011-12-05 18:10:44 -0800 | [diff] [blame^] | 1235 | rbd_req_sync_notify_ack(rbd_dev, hver, notify_id, rbd_dev->header_name); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1236 | } |
| 1237 | |
| 1238 | /* |
| 1239 | * Request sync osd watch |
| 1240 | */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1241 | static int rbd_req_sync_watch(struct rbd_device *rbd_dev, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1242 | const char *object_name, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1243 | u64 ver) |
| 1244 | { |
| 1245 | struct ceph_osd_req_op *ops; |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1246 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1247 | |
| 1248 | int ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_WATCH, 0); |
| 1249 | if (ret < 0) |
| 1250 | return ret; |
| 1251 | |
| 1252 | ret = ceph_osdc_create_event(osdc, rbd_watch_cb, 0, |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1253 | (void *)rbd_dev, &rbd_dev->watch_event); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1254 | if (ret < 0) |
| 1255 | goto fail; |
| 1256 | |
| 1257 | ops[0].watch.ver = cpu_to_le64(ver); |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1258 | ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1259 | ops[0].watch.flag = 1; |
| 1260 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1261 | ret = rbd_req_sync_op(rbd_dev, NULL, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1262 | CEPH_NOSNAP, |
| 1263 | 0, |
| 1264 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
| 1265 | ops, |
Alex Elder | d1f57ea | 2012-06-26 12:57:03 -0700 | [diff] [blame] | 1266 | object_name, 0, 0, NULL, |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1267 | &rbd_dev->watch_request, NULL); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1268 | |
| 1269 | if (ret < 0) |
| 1270 | goto fail_event; |
| 1271 | |
| 1272 | rbd_destroy_ops(ops); |
| 1273 | return 0; |
| 1274 | |
| 1275 | fail_event: |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1276 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
| 1277 | rbd_dev->watch_event = NULL; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1278 | fail: |
| 1279 | rbd_destroy_ops(ops); |
| 1280 | return ret; |
| 1281 | } |
| 1282 | |
Yehuda Sadeh | 79e3057c | 2011-07-12 16:56:57 -0700 | [diff] [blame] | 1283 | /* |
| 1284 | * Request sync osd unwatch |
| 1285 | */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1286 | static int rbd_req_sync_unwatch(struct rbd_device *rbd_dev, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1287 | const char *object_name) |
Yehuda Sadeh | 79e3057c | 2011-07-12 16:56:57 -0700 | [diff] [blame] | 1288 | { |
| 1289 | struct ceph_osd_req_op *ops; |
| 1290 | |
| 1291 | int ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_WATCH, 0); |
| 1292 | if (ret < 0) |
| 1293 | return ret; |
| 1294 | |
| 1295 | ops[0].watch.ver = 0; |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1296 | ops[0].watch.cookie = cpu_to_le64(rbd_dev->watch_event->cookie); |
Yehuda Sadeh | 79e3057c | 2011-07-12 16:56:57 -0700 | [diff] [blame] | 1297 | ops[0].watch.flag = 0; |
| 1298 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1299 | ret = rbd_req_sync_op(rbd_dev, NULL, |
Yehuda Sadeh | 79e3057c | 2011-07-12 16:56:57 -0700 | [diff] [blame] | 1300 | CEPH_NOSNAP, |
| 1301 | 0, |
| 1302 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
| 1303 | ops, |
Alex Elder | d1f57ea | 2012-06-26 12:57:03 -0700 | [diff] [blame] | 1304 | object_name, 0, 0, NULL, NULL, NULL); |
Yehuda Sadeh | 79e3057c | 2011-07-12 16:56:57 -0700 | [diff] [blame] | 1305 | |
| 1306 | rbd_destroy_ops(ops); |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1307 | ceph_osdc_cancel_event(rbd_dev->watch_event); |
| 1308 | rbd_dev->watch_event = NULL; |
Yehuda Sadeh | 79e3057c | 2011-07-12 16:56:57 -0700 | [diff] [blame] | 1309 | return ret; |
| 1310 | } |
| 1311 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1312 | struct rbd_notify_info { |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1313 | struct rbd_device *rbd_dev; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1314 | }; |
| 1315 | |
| 1316 | static void rbd_notify_cb(u64 ver, u64 notify_id, u8 opcode, void *data) |
| 1317 | { |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1318 | struct rbd_device *rbd_dev = (struct rbd_device *)data; |
| 1319 | if (!rbd_dev) |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1320 | return; |
| 1321 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1322 | dout("rbd_notify_cb %s notify_id=%lld opcode=%d\n", |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1323 | rbd_dev->header_name, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1324 | notify_id, (int)opcode); |
| 1325 | } |
| 1326 | |
| 1327 | /* |
| 1328 | * Request sync osd notify |
| 1329 | */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1330 | static int rbd_req_sync_notify(struct rbd_device *rbd_dev, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1331 | const char *object_name) |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1332 | { |
| 1333 | struct ceph_osd_req_op *ops; |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1334 | struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1335 | struct ceph_osd_event *event; |
| 1336 | struct rbd_notify_info info; |
| 1337 | int payload_len = sizeof(u32) + sizeof(u32); |
| 1338 | int ret; |
| 1339 | |
| 1340 | ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_NOTIFY, payload_len); |
| 1341 | if (ret < 0) |
| 1342 | return ret; |
| 1343 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1344 | info.rbd_dev = rbd_dev; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1345 | |
| 1346 | ret = ceph_osdc_create_event(osdc, rbd_notify_cb, 1, |
| 1347 | (void *)&info, &event); |
| 1348 | if (ret < 0) |
| 1349 | goto fail; |
| 1350 | |
| 1351 | ops[0].watch.ver = 1; |
| 1352 | ops[0].watch.flag = 1; |
| 1353 | ops[0].watch.cookie = event->cookie; |
| 1354 | ops[0].watch.prot_ver = RADOS_NOTIFY_VER; |
| 1355 | ops[0].watch.timeout = 12; |
| 1356 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1357 | ret = rbd_req_sync_op(rbd_dev, NULL, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1358 | CEPH_NOSNAP, |
| 1359 | 0, |
| 1360 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
| 1361 | ops, |
Alex Elder | d1f57ea | 2012-06-26 12:57:03 -0700 | [diff] [blame] | 1362 | object_name, 0, 0, NULL, NULL, NULL); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1363 | if (ret < 0) |
| 1364 | goto fail_event; |
| 1365 | |
| 1366 | ret = ceph_osdc_wait_event(event, CEPH_OSD_TIMEOUT_DEFAULT); |
| 1367 | dout("ceph_osdc_wait_event returned %d\n", ret); |
| 1368 | rbd_destroy_ops(ops); |
| 1369 | return 0; |
| 1370 | |
| 1371 | fail_event: |
| 1372 | ceph_osdc_cancel_event(event); |
| 1373 | fail: |
| 1374 | rbd_destroy_ops(ops); |
| 1375 | return ret; |
| 1376 | } |
| 1377 | |
| 1378 | /* |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1379 | * Request sync osd read |
| 1380 | */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1381 | static int rbd_req_sync_exec(struct rbd_device *rbd_dev, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1382 | const char *object_name, |
| 1383 | const char *class_name, |
| 1384 | const char *method_name, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1385 | const char *data, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1386 | int len, |
| 1387 | u64 *ver) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1388 | { |
| 1389 | struct ceph_osd_req_op *ops; |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1390 | int class_name_len = strlen(class_name); |
| 1391 | int method_name_len = strlen(method_name); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1392 | int ret = rbd_create_rw_ops(&ops, 1, CEPH_OSD_OP_CALL, |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1393 | class_name_len + method_name_len + len); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1394 | if (ret < 0) |
| 1395 | return ret; |
| 1396 | |
Alex Elder | aded07e | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1397 | ops[0].cls.class_name = class_name; |
| 1398 | ops[0].cls.class_len = (__u8) class_name_len; |
| 1399 | ops[0].cls.method_name = method_name; |
| 1400 | ops[0].cls.method_len = (__u8) method_name_len; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1401 | ops[0].cls.argc = 0; |
| 1402 | ops[0].cls.indata = data; |
| 1403 | ops[0].cls.indata_len = len; |
| 1404 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1405 | ret = rbd_req_sync_op(rbd_dev, NULL, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1406 | CEPH_NOSNAP, |
| 1407 | 0, |
| 1408 | CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK, |
| 1409 | ops, |
Alex Elder | d1f57ea | 2012-06-26 12:57:03 -0700 | [diff] [blame] | 1410 | object_name, 0, 0, NULL, NULL, ver); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1411 | |
| 1412 | rbd_destroy_ops(ops); |
| 1413 | |
| 1414 | dout("cls_exec returned %d\n", ret); |
| 1415 | return ret; |
| 1416 | } |
| 1417 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1418 | static struct rbd_req_coll *rbd_alloc_coll(int num_reqs) |
| 1419 | { |
| 1420 | struct rbd_req_coll *coll = |
| 1421 | kzalloc(sizeof(struct rbd_req_coll) + |
| 1422 | sizeof(struct rbd_req_status) * num_reqs, |
| 1423 | GFP_ATOMIC); |
| 1424 | |
| 1425 | if (!coll) |
| 1426 | return NULL; |
| 1427 | coll->total = num_reqs; |
| 1428 | kref_init(&coll->kref); |
| 1429 | return coll; |
| 1430 | } |
| 1431 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1432 | /* |
| 1433 | * block device queue callback |
| 1434 | */ |
| 1435 | static void rbd_rq_fn(struct request_queue *q) |
| 1436 | { |
| 1437 | struct rbd_device *rbd_dev = q->queuedata; |
| 1438 | struct request *rq; |
| 1439 | struct bio_pair *bp = NULL; |
| 1440 | |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1441 | while ((rq = blk_fetch_request(q))) { |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1442 | struct bio *bio; |
| 1443 | struct bio *rq_bio, *next_bio = NULL; |
| 1444 | bool do_write; |
| 1445 | int size, op_size = 0; |
| 1446 | u64 ofs; |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1447 | int num_segs, cur_seg = 0; |
| 1448 | struct rbd_req_coll *coll; |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1449 | struct ceph_snap_context *snapc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1450 | |
| 1451 | /* peek at request from block layer */ |
| 1452 | if (!rq) |
| 1453 | break; |
| 1454 | |
| 1455 | dout("fetched request\n"); |
| 1456 | |
| 1457 | /* filter out block requests we don't understand */ |
| 1458 | if ((rq->cmd_type != REQ_TYPE_FS)) { |
| 1459 | __blk_end_request_all(rq, 0); |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1460 | continue; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1461 | } |
| 1462 | |
| 1463 | /* deduce our operation (read, write) */ |
| 1464 | do_write = (rq_data_dir(rq) == WRITE); |
| 1465 | |
| 1466 | size = blk_rq_bytes(rq); |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1467 | ofs = blk_rq_pos(rq) * SECTOR_SIZE; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1468 | rq_bio = rq->bio; |
| 1469 | if (do_write && rbd_dev->read_only) { |
| 1470 | __blk_end_request_all(rq, -EROFS); |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1471 | continue; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1472 | } |
| 1473 | |
| 1474 | spin_unlock_irq(q->queue_lock); |
| 1475 | |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1476 | down_read(&rbd_dev->header_rwsem); |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 1477 | |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1478 | if (rbd_dev->snap_id != CEPH_NOSNAP && !rbd_dev->snap_exists) { |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 1479 | up_read(&rbd_dev->header_rwsem); |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1480 | dout("request for non-existent snapshot"); |
| 1481 | spin_lock_irq(q->queue_lock); |
| 1482 | __blk_end_request_all(rq, -ENXIO); |
| 1483 | continue; |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 1484 | } |
| 1485 | |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1486 | snapc = ceph_get_snap_context(rbd_dev->header.snapc); |
| 1487 | |
| 1488 | up_read(&rbd_dev->header_rwsem); |
| 1489 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1490 | dout("%s 0x%x bytes at 0x%llx\n", |
| 1491 | do_write ? "write" : "read", |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1492 | size, blk_rq_pos(rq) * SECTOR_SIZE); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1493 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1494 | num_segs = rbd_get_num_segments(&rbd_dev->header, ofs, size); |
| 1495 | coll = rbd_alloc_coll(num_segs); |
| 1496 | if (!coll) { |
| 1497 | spin_lock_irq(q->queue_lock); |
| 1498 | __blk_end_request_all(rq, -ENOMEM); |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1499 | ceph_put_snap_context(snapc); |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1500 | continue; |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1501 | } |
| 1502 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1503 | do { |
| 1504 | /* a bio clone to be passed down to OSD req */ |
| 1505 | dout("rq->bio->bi_vcnt=%d\n", rq->bio->bi_vcnt); |
| 1506 | op_size = rbd_get_segment(&rbd_dev->header, |
Alex Elder | ca1e49a | 2012-07-10 20:30:09 -0500 | [diff] [blame] | 1507 | rbd_dev->header.object_prefix, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1508 | ofs, size, |
| 1509 | NULL, NULL); |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1510 | kref_get(&coll->kref); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1511 | bio = bio_chain_clone(&rq_bio, &next_bio, &bp, |
| 1512 | op_size, GFP_ATOMIC); |
| 1513 | if (!bio) { |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1514 | rbd_coll_end_req_index(rq, coll, cur_seg, |
| 1515 | -ENOMEM, op_size); |
| 1516 | goto next_seg; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1517 | } |
| 1518 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1519 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1520 | /* init OSD command: write or read */ |
| 1521 | if (do_write) |
| 1522 | rbd_req_write(rq, rbd_dev, |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1523 | snapc, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1524 | ofs, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1525 | op_size, bio, |
| 1526 | coll, cur_seg); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1527 | else |
| 1528 | rbd_req_read(rq, rbd_dev, |
Josh Durgin | 77dfe99 | 2011-11-21 13:04:42 -0800 | [diff] [blame] | 1529 | rbd_dev->snap_id, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1530 | ofs, |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1531 | op_size, bio, |
| 1532 | coll, cur_seg); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1533 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1534 | next_seg: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1535 | size -= op_size; |
| 1536 | ofs += op_size; |
| 1537 | |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1538 | cur_seg++; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1539 | rq_bio = next_bio; |
| 1540 | } while (size > 0); |
Yehuda Sadeh | 1fec709 | 2011-05-13 13:52:56 -0700 | [diff] [blame] | 1541 | kref_put(&coll->kref, rbd_coll_release); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1542 | |
| 1543 | if (bp) |
| 1544 | bio_pair_release(bp); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1545 | spin_lock_irq(q->queue_lock); |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1546 | |
| 1547 | ceph_put_snap_context(snapc); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1548 | } |
| 1549 | } |
| 1550 | |
| 1551 | /* |
| 1552 | * a queue callback. Makes sure that we don't create a bio that spans across |
| 1553 | * multiple osd objects. One exception would be with a single page bios, |
| 1554 | * which we handle later at bio_chain_clone |
| 1555 | */ |
| 1556 | static int rbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bmd, |
| 1557 | struct bio_vec *bvec) |
| 1558 | { |
| 1559 | struct rbd_device *rbd_dev = q->queuedata; |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1560 | unsigned int chunk_sectors; |
| 1561 | sector_t sector; |
| 1562 | unsigned int bio_sectors; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1563 | int max; |
| 1564 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1565 | chunk_sectors = 1 << (rbd_dev->header.obj_order - SECTOR_SHIFT); |
| 1566 | sector = bmd->bi_sector + get_start_sect(bmd->bi_bdev); |
| 1567 | bio_sectors = bmd->bi_size >> SECTOR_SHIFT; |
| 1568 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1569 | max = (chunk_sectors - ((sector & (chunk_sectors - 1)) |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1570 | + bio_sectors)) << SECTOR_SHIFT; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1571 | if (max < 0) |
| 1572 | max = 0; /* bio_add cannot handle a negative return */ |
| 1573 | if (max <= bvec->bv_len && bio_sectors == 0) |
| 1574 | return bvec->bv_len; |
| 1575 | return max; |
| 1576 | } |
| 1577 | |
| 1578 | static void rbd_free_disk(struct rbd_device *rbd_dev) |
| 1579 | { |
| 1580 | struct gendisk *disk = rbd_dev->disk; |
| 1581 | |
| 1582 | if (!disk) |
| 1583 | return; |
| 1584 | |
| 1585 | rbd_header_free(&rbd_dev->header); |
| 1586 | |
| 1587 | if (disk->flags & GENHD_FL_UP) |
| 1588 | del_gendisk(disk); |
| 1589 | if (disk->queue) |
| 1590 | blk_cleanup_queue(disk->queue); |
| 1591 | put_disk(disk); |
| 1592 | } |
| 1593 | |
| 1594 | /* |
| 1595 | * reload the ondisk the header |
| 1596 | */ |
| 1597 | static int rbd_read_header(struct rbd_device *rbd_dev, |
| 1598 | struct rbd_image_header *header) |
| 1599 | { |
| 1600 | ssize_t rc; |
| 1601 | struct rbd_image_header_ondisk *dh; |
Xi Wang | 50f7c4c | 2012-04-20 15:49:44 -0500 | [diff] [blame] | 1602 | u32 snap_count = 0; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1603 | u64 ver; |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1604 | size_t len; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1605 | |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1606 | /* |
| 1607 | * First reads the fixed-size header to determine the number |
| 1608 | * of snapshots, then re-reads it, along with all snapshot |
| 1609 | * records as well as their stored names. |
| 1610 | */ |
| 1611 | len = sizeof (*dh); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1612 | while (1) { |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1613 | dh = kmalloc(len, GFP_KERNEL); |
| 1614 | if (!dh) |
| 1615 | return -ENOMEM; |
| 1616 | |
| 1617 | rc = rbd_req_sync_read(rbd_dev, |
| 1618 | NULL, CEPH_NOSNAP, |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1619 | rbd_dev->header_name, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1620 | 0, len, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1621 | (char *)dh, &ver); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1622 | if (rc < 0) |
| 1623 | goto out_dh; |
| 1624 | |
| 1625 | rc = rbd_header_from_disk(header, dh, snap_count, GFP_KERNEL); |
Josh Durgin | 81e759f | 2011-11-15 14:49:53 -0800 | [diff] [blame] | 1626 | if (rc < 0) { |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1627 | if (rc == -ENXIO) |
Josh Durgin | 81e759f | 2011-11-15 14:49:53 -0800 | [diff] [blame] | 1628 | pr_warning("unrecognized header format" |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1629 | " for image %s\n", |
| 1630 | rbd_dev->image_name); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1631 | goto out_dh; |
Josh Durgin | 81e759f | 2011-11-15 14:49:53 -0800 | [diff] [blame] | 1632 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1633 | |
Alex Elder | 00f1f36 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 1634 | if (snap_count == header->total_snaps) |
| 1635 | break; |
| 1636 | |
| 1637 | snap_count = header->total_snaps; |
| 1638 | len = sizeof (*dh) + |
| 1639 | snap_count * sizeof(struct rbd_image_snap_ondisk) + |
| 1640 | header->snap_names_len; |
| 1641 | |
| 1642 | rbd_header_free(header); |
| 1643 | kfree(dh); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1644 | } |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1645 | header->obj_version = ver; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1646 | |
| 1647 | out_dh: |
| 1648 | kfree(dh); |
| 1649 | return rc; |
| 1650 | } |
| 1651 | |
| 1652 | /* |
| 1653 | * create a snapshot |
| 1654 | */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1655 | static int rbd_header_add_snap(struct rbd_device *rbd_dev, |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1656 | const char *snap_name, |
| 1657 | gfp_t gfp_flags) |
| 1658 | { |
| 1659 | int name_len = strlen(snap_name); |
| 1660 | u64 new_snapid; |
| 1661 | int ret; |
Sage Weil | 916d4d6 | 2011-05-12 16:10:50 -0700 | [diff] [blame] | 1662 | void *data, *p, *e; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1663 | u64 ver; |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 1664 | struct ceph_mon_client *monc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1665 | |
| 1666 | /* we should create a snapshot only if we're pointing at the head */ |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1667 | if (rbd_dev->snap_id != CEPH_NOSNAP) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1668 | return -EINVAL; |
| 1669 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1670 | monc = &rbd_dev->rbd_client->client->monc; |
| 1671 | ret = ceph_monc_create_snapid(monc, rbd_dev->pool_id, &new_snapid); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1672 | dout("created snapid=%lld\n", new_snapid); |
| 1673 | if (ret < 0) |
| 1674 | return ret; |
| 1675 | |
| 1676 | data = kmalloc(name_len + 16, gfp_flags); |
| 1677 | if (!data) |
| 1678 | return -ENOMEM; |
| 1679 | |
Sage Weil | 916d4d6 | 2011-05-12 16:10:50 -0700 | [diff] [blame] | 1680 | p = data; |
| 1681 | e = data + name_len + 16; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1682 | |
Sage Weil | 916d4d6 | 2011-05-12 16:10:50 -0700 | [diff] [blame] | 1683 | ceph_encode_string_safe(&p, e, snap_name, name_len, bad); |
| 1684 | ceph_encode_64_safe(&p, e, new_snapid, bad); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1685 | |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1686 | ret = rbd_req_sync_exec(rbd_dev, rbd_dev->header_name, |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1687 | "rbd", "snap_add", |
Sage Weil | 916d4d6 | 2011-05-12 16:10:50 -0700 | [diff] [blame] | 1688 | data, p - data, &ver); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1689 | |
Sage Weil | 916d4d6 | 2011-05-12 16:10:50 -0700 | [diff] [blame] | 1690 | kfree(data); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1691 | |
| 1692 | if (ret < 0) |
| 1693 | return ret; |
| 1694 | |
Alex Elder | 0ce1a79 | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1695 | down_write(&rbd_dev->header_rwsem); |
| 1696 | rbd_dev->header.snapc->seq = new_snapid; |
| 1697 | up_write(&rbd_dev->header_rwsem); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1698 | |
| 1699 | return 0; |
| 1700 | bad: |
| 1701 | return -ERANGE; |
| 1702 | } |
| 1703 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1704 | static void __rbd_remove_all_snaps(struct rbd_device *rbd_dev) |
| 1705 | { |
| 1706 | struct rbd_snap *snap; |
| 1707 | |
| 1708 | while (!list_empty(&rbd_dev->snaps)) { |
| 1709 | snap = list_first_entry(&rbd_dev->snaps, struct rbd_snap, node); |
| 1710 | __rbd_remove_snap_dev(rbd_dev, snap); |
| 1711 | } |
| 1712 | } |
| 1713 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1714 | /* |
| 1715 | * only read the first part of the ondisk header, without the snaps info |
| 1716 | */ |
Josh Durgin | 263c6ca | 2011-12-05 10:43:42 -0800 | [diff] [blame] | 1717 | static int __rbd_refresh_header(struct rbd_device *rbd_dev) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1718 | { |
| 1719 | int ret; |
| 1720 | struct rbd_image_header h; |
| 1721 | u64 snap_seq; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1722 | int follow_seq = 0; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1723 | |
| 1724 | ret = rbd_read_header(rbd_dev, &h); |
| 1725 | if (ret < 0) |
| 1726 | return ret; |
| 1727 | |
Josh Durgin | a51aa0c | 2011-12-05 10:35:04 -0800 | [diff] [blame] | 1728 | down_write(&rbd_dev->header_rwsem); |
| 1729 | |
Sage Weil | 9db4b3e | 2011-04-19 22:49:06 -0700 | [diff] [blame] | 1730 | /* resized? */ |
Josh Durgin | 474ef7c | 2011-11-21 17:13:54 -0800 | [diff] [blame] | 1731 | if (rbd_dev->snap_id == CEPH_NOSNAP) { |
| 1732 | sector_t size = (sector_t) h.image_size / SECTOR_SIZE; |
| 1733 | |
| 1734 | dout("setting size to %llu sectors", (unsigned long long) size); |
| 1735 | set_capacity(rbd_dev->disk, size); |
| 1736 | } |
Sage Weil | 9db4b3e | 2011-04-19 22:49:06 -0700 | [diff] [blame] | 1737 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1738 | snap_seq = rbd_dev->header.snapc->seq; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1739 | if (rbd_dev->header.total_snaps && |
| 1740 | rbd_dev->header.snapc->snaps[0] == snap_seq) |
| 1741 | /* pointing at the head, will need to follow that |
| 1742 | if head moves */ |
| 1743 | follow_seq = 1; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1744 | |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 1745 | /* rbd_dev->header.object_prefix shouldn't change */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1746 | kfree(rbd_dev->header.snap_sizes); |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 1747 | kfree(rbd_dev->header.snap_names); |
Josh Durgin | d1d2564 | 2011-12-05 14:03:05 -0800 | [diff] [blame] | 1748 | /* osd requests may still refer to snapc */ |
| 1749 | ceph_put_snap_context(rbd_dev->header.snapc); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1750 | |
Josh Durgin | a71b891 | 2011-12-05 18:10:44 -0800 | [diff] [blame^] | 1751 | rbd_dev->header.obj_version = h.obj_version; |
Josh Durgin | 93a24e0 | 2011-12-05 10:41:28 -0800 | [diff] [blame] | 1752 | rbd_dev->header.image_size = h.image_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1753 | rbd_dev->header.total_snaps = h.total_snaps; |
| 1754 | rbd_dev->header.snapc = h.snapc; |
| 1755 | rbd_dev->header.snap_names = h.snap_names; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1756 | rbd_dev->header.snap_names_len = h.snap_names_len; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1757 | rbd_dev->header.snap_sizes = h.snap_sizes; |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 1758 | /* Free the extra copy of the object prefix */ |
| 1759 | WARN_ON(strcmp(rbd_dev->header.object_prefix, h.object_prefix)); |
| 1760 | kfree(h.object_prefix); |
| 1761 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 1762 | if (follow_seq) |
| 1763 | rbd_dev->header.snapc->seq = rbd_dev->header.snapc->snaps[0]; |
| 1764 | else |
| 1765 | rbd_dev->header.snapc->seq = snap_seq; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1766 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1767 | ret = __rbd_init_snaps_header(rbd_dev); |
| 1768 | |
Josh Durgin | c666601 | 2011-11-21 17:11:12 -0800 | [diff] [blame] | 1769 | up_write(&rbd_dev->header_rwsem); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1770 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1771 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1772 | } |
| 1773 | |
| 1774 | static int rbd_init_disk(struct rbd_device *rbd_dev) |
| 1775 | { |
| 1776 | struct gendisk *disk; |
| 1777 | struct request_queue *q; |
| 1778 | int rc; |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1779 | u64 segment_size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1780 | u64 total_size = 0; |
| 1781 | |
| 1782 | /* contact OSD, request size info about the object being mapped */ |
| 1783 | rc = rbd_read_header(rbd_dev, &rbd_dev->header); |
| 1784 | if (rc) |
| 1785 | return rc; |
| 1786 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1787 | /* no need to lock here, as rbd_dev is not registered yet */ |
| 1788 | rc = __rbd_init_snaps_header(rbd_dev); |
| 1789 | if (rc) |
| 1790 | return rc; |
| 1791 | |
Josh Durgin | cc9d734 | 2011-11-21 18:19:13 -0800 | [diff] [blame] | 1792 | rc = rbd_header_set_snap(rbd_dev, &total_size); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1793 | if (rc) |
| 1794 | return rc; |
| 1795 | |
| 1796 | /* create gendisk info */ |
| 1797 | rc = -ENOMEM; |
| 1798 | disk = alloc_disk(RBD_MINORS_PER_MAJOR); |
| 1799 | if (!disk) |
| 1800 | goto out; |
| 1801 | |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 1802 | snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", |
Sage Weil | aedfec5 | 2011-05-12 20:57:03 -0700 | [diff] [blame] | 1803 | rbd_dev->id); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1804 | disk->major = rbd_dev->major; |
| 1805 | disk->first_minor = 0; |
| 1806 | disk->fops = &rbd_bd_ops; |
| 1807 | disk->private_data = rbd_dev; |
| 1808 | |
| 1809 | /* init rq */ |
| 1810 | rc = -ENOMEM; |
| 1811 | q = blk_init_queue(rbd_rq_fn, &rbd_dev->lock); |
| 1812 | if (!q) |
| 1813 | goto out_disk; |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 1814 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1815 | /* We use the default size, but let's be explicit about it. */ |
| 1816 | blk_queue_physical_block_size(q, SECTOR_SIZE); |
| 1817 | |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 1818 | /* set io sizes to object size */ |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1819 | segment_size = rbd_obj_bytes(&rbd_dev->header); |
| 1820 | blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE); |
| 1821 | blk_queue_max_segment_size(q, segment_size); |
| 1822 | blk_queue_io_min(q, segment_size); |
| 1823 | blk_queue_io_opt(q, segment_size); |
Josh Durgin | 029bcbd | 2011-07-22 11:35:23 -0700 | [diff] [blame] | 1824 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1825 | blk_queue_merge_bvec(q, rbd_merge_bvec); |
| 1826 | disk->queue = q; |
| 1827 | |
| 1828 | q->queuedata = rbd_dev; |
| 1829 | |
| 1830 | rbd_dev->disk = disk; |
| 1831 | rbd_dev->q = q; |
| 1832 | |
| 1833 | /* finally, announce the disk to the world */ |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1834 | set_capacity(disk, total_size / SECTOR_SIZE); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1835 | add_disk(disk); |
| 1836 | |
| 1837 | pr_info("%s: added with size 0x%llx\n", |
| 1838 | disk->disk_name, (unsigned long long)total_size); |
| 1839 | return 0; |
| 1840 | |
| 1841 | out_disk: |
| 1842 | put_disk(disk); |
| 1843 | out: |
| 1844 | return rc; |
| 1845 | } |
| 1846 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1847 | /* |
| 1848 | sysfs |
| 1849 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1850 | |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1851 | static struct rbd_device *dev_to_rbd_dev(struct device *dev) |
| 1852 | { |
| 1853 | return container_of(dev, struct rbd_device, dev); |
| 1854 | } |
| 1855 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1856 | static ssize_t rbd_size_show(struct device *dev, |
| 1857 | struct device_attribute *attr, char *buf) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1858 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1859 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Josh Durgin | a51aa0c | 2011-12-05 10:35:04 -0800 | [diff] [blame] | 1860 | sector_t size; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1861 | |
Josh Durgin | a51aa0c | 2011-12-05 10:35:04 -0800 | [diff] [blame] | 1862 | down_read(&rbd_dev->header_rwsem); |
| 1863 | size = get_capacity(rbd_dev->disk); |
| 1864 | up_read(&rbd_dev->header_rwsem); |
| 1865 | |
| 1866 | return sprintf(buf, "%llu\n", (unsigned long long) size * SECTOR_SIZE); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1867 | } |
| 1868 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1869 | static ssize_t rbd_major_show(struct device *dev, |
| 1870 | struct device_attribute *attr, char *buf) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1871 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1872 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1873 | |
| 1874 | return sprintf(buf, "%d\n", rbd_dev->major); |
| 1875 | } |
| 1876 | |
| 1877 | static ssize_t rbd_client_id_show(struct device *dev, |
| 1878 | struct device_attribute *attr, char *buf) |
| 1879 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1880 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1881 | |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 1882 | return sprintf(buf, "client%lld\n", |
| 1883 | ceph_client_id(rbd_dev->rbd_client->client)); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1884 | } |
| 1885 | |
| 1886 | static ssize_t rbd_pool_show(struct device *dev, |
| 1887 | struct device_attribute *attr, char *buf) |
| 1888 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1889 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1890 | |
| 1891 | return sprintf(buf, "%s\n", rbd_dev->pool_name); |
| 1892 | } |
| 1893 | |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 1894 | static ssize_t rbd_pool_id_show(struct device *dev, |
| 1895 | struct device_attribute *attr, char *buf) |
| 1896 | { |
| 1897 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
| 1898 | |
| 1899 | return sprintf(buf, "%d\n", rbd_dev->pool_id); |
| 1900 | } |
| 1901 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1902 | static ssize_t rbd_name_show(struct device *dev, |
| 1903 | struct device_attribute *attr, char *buf) |
| 1904 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1905 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1906 | |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 1907 | return sprintf(buf, "%s\n", rbd_dev->image_name); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1908 | } |
| 1909 | |
| 1910 | static ssize_t rbd_snap_show(struct device *dev, |
| 1911 | struct device_attribute *attr, |
| 1912 | char *buf) |
| 1913 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1914 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1915 | |
| 1916 | return sprintf(buf, "%s\n", rbd_dev->snap_name); |
| 1917 | } |
| 1918 | |
| 1919 | static ssize_t rbd_image_refresh(struct device *dev, |
| 1920 | struct device_attribute *attr, |
| 1921 | const char *buf, |
| 1922 | size_t size) |
| 1923 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 1924 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1925 | int rc; |
| 1926 | int ret = size; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1927 | |
| 1928 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
| 1929 | |
Josh Durgin | 263c6ca | 2011-12-05 10:43:42 -0800 | [diff] [blame] | 1930 | rc = __rbd_refresh_header(rbd_dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1931 | if (rc < 0) |
| 1932 | ret = rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1933 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1934 | mutex_unlock(&ctl_mutex); |
| 1935 | return ret; |
| 1936 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 1937 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1938 | static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL); |
| 1939 | static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL); |
| 1940 | static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL); |
| 1941 | static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL); |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 1942 | static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1943 | static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL); |
| 1944 | static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh); |
| 1945 | static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL); |
| 1946 | static DEVICE_ATTR(create_snap, S_IWUSR, NULL, rbd_snap_add); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1947 | |
| 1948 | static struct attribute *rbd_attrs[] = { |
| 1949 | &dev_attr_size.attr, |
| 1950 | &dev_attr_major.attr, |
| 1951 | &dev_attr_client_id.attr, |
| 1952 | &dev_attr_pool.attr, |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 1953 | &dev_attr_pool_id.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1954 | &dev_attr_name.attr, |
| 1955 | &dev_attr_current_snap.attr, |
| 1956 | &dev_attr_refresh.attr, |
| 1957 | &dev_attr_create_snap.attr, |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1958 | NULL |
| 1959 | }; |
| 1960 | |
| 1961 | static struct attribute_group rbd_attr_group = { |
| 1962 | .attrs = rbd_attrs, |
| 1963 | }; |
| 1964 | |
| 1965 | static const struct attribute_group *rbd_attr_groups[] = { |
| 1966 | &rbd_attr_group, |
| 1967 | NULL |
| 1968 | }; |
| 1969 | |
| 1970 | static void rbd_sysfs_dev_release(struct device *dev) |
| 1971 | { |
| 1972 | } |
| 1973 | |
| 1974 | static struct device_type rbd_device_type = { |
| 1975 | .name = "rbd", |
| 1976 | .groups = rbd_attr_groups, |
| 1977 | .release = rbd_sysfs_dev_release, |
| 1978 | }; |
| 1979 | |
| 1980 | |
| 1981 | /* |
| 1982 | sysfs - snapshots |
| 1983 | */ |
| 1984 | |
| 1985 | static ssize_t rbd_snap_size_show(struct device *dev, |
| 1986 | struct device_attribute *attr, |
| 1987 | char *buf) |
| 1988 | { |
| 1989 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); |
| 1990 | |
Josh Durgin | 3591538 | 2011-12-05 18:25:13 -0800 | [diff] [blame] | 1991 | return sprintf(buf, "%llu\n", (unsigned long long)snap->size); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 1992 | } |
| 1993 | |
| 1994 | static ssize_t rbd_snap_id_show(struct device *dev, |
| 1995 | struct device_attribute *attr, |
| 1996 | char *buf) |
| 1997 | { |
| 1998 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); |
| 1999 | |
Josh Durgin | 3591538 | 2011-12-05 18:25:13 -0800 | [diff] [blame] | 2000 | return sprintf(buf, "%llu\n", (unsigned long long)snap->id); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2001 | } |
| 2002 | |
| 2003 | static DEVICE_ATTR(snap_size, S_IRUGO, rbd_snap_size_show, NULL); |
| 2004 | static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL); |
| 2005 | |
| 2006 | static struct attribute *rbd_snap_attrs[] = { |
| 2007 | &dev_attr_snap_size.attr, |
| 2008 | &dev_attr_snap_id.attr, |
| 2009 | NULL, |
| 2010 | }; |
| 2011 | |
| 2012 | static struct attribute_group rbd_snap_attr_group = { |
| 2013 | .attrs = rbd_snap_attrs, |
| 2014 | }; |
| 2015 | |
| 2016 | static void rbd_snap_dev_release(struct device *dev) |
| 2017 | { |
| 2018 | struct rbd_snap *snap = container_of(dev, struct rbd_snap, dev); |
| 2019 | kfree(snap->name); |
| 2020 | kfree(snap); |
| 2021 | } |
| 2022 | |
| 2023 | static const struct attribute_group *rbd_snap_attr_groups[] = { |
| 2024 | &rbd_snap_attr_group, |
| 2025 | NULL |
| 2026 | }; |
| 2027 | |
| 2028 | static struct device_type rbd_snap_device_type = { |
| 2029 | .groups = rbd_snap_attr_groups, |
| 2030 | .release = rbd_snap_dev_release, |
| 2031 | }; |
| 2032 | |
| 2033 | static void __rbd_remove_snap_dev(struct rbd_device *rbd_dev, |
| 2034 | struct rbd_snap *snap) |
| 2035 | { |
| 2036 | list_del(&snap->node); |
| 2037 | device_unregister(&snap->dev); |
| 2038 | } |
| 2039 | |
| 2040 | static int rbd_register_snap_dev(struct rbd_device *rbd_dev, |
| 2041 | struct rbd_snap *snap, |
| 2042 | struct device *parent) |
| 2043 | { |
| 2044 | struct device *dev = &snap->dev; |
| 2045 | int ret; |
| 2046 | |
| 2047 | dev->type = &rbd_snap_device_type; |
| 2048 | dev->parent = parent; |
| 2049 | dev->release = rbd_snap_dev_release; |
| 2050 | dev_set_name(dev, "snap_%s", snap->name); |
| 2051 | ret = device_register(dev); |
| 2052 | |
| 2053 | return ret; |
| 2054 | } |
| 2055 | |
| 2056 | static int __rbd_add_snap_dev(struct rbd_device *rbd_dev, |
| 2057 | int i, const char *name, |
| 2058 | struct rbd_snap **snapp) |
| 2059 | { |
| 2060 | int ret; |
| 2061 | struct rbd_snap *snap = kzalloc(sizeof(*snap), GFP_KERNEL); |
| 2062 | if (!snap) |
| 2063 | return -ENOMEM; |
| 2064 | snap->name = kstrdup(name, GFP_KERNEL); |
| 2065 | snap->size = rbd_dev->header.snap_sizes[i]; |
| 2066 | snap->id = rbd_dev->header.snapc->snaps[i]; |
| 2067 | if (device_is_registered(&rbd_dev->dev)) { |
| 2068 | ret = rbd_register_snap_dev(rbd_dev, snap, |
| 2069 | &rbd_dev->dev); |
| 2070 | if (ret < 0) |
| 2071 | goto err; |
| 2072 | } |
| 2073 | *snapp = snap; |
| 2074 | return 0; |
| 2075 | err: |
| 2076 | kfree(snap->name); |
| 2077 | kfree(snap); |
| 2078 | return ret; |
| 2079 | } |
| 2080 | |
| 2081 | /* |
| 2082 | * search for the previous snap in a null delimited string list |
| 2083 | */ |
| 2084 | const char *rbd_prev_snap_name(const char *name, const char *start) |
| 2085 | { |
| 2086 | if (name < start + 2) |
| 2087 | return NULL; |
| 2088 | |
| 2089 | name -= 2; |
| 2090 | while (*name) { |
| 2091 | if (name == start) |
| 2092 | return start; |
| 2093 | name--; |
| 2094 | } |
| 2095 | return name + 1; |
| 2096 | } |
| 2097 | |
| 2098 | /* |
| 2099 | * compare the old list of snapshots that we have to what's in the header |
| 2100 | * and update it accordingly. Note that the header holds the snapshots |
| 2101 | * in a reverse order (from newest to oldest) and we need to go from |
| 2102 | * older to new so that we don't get a duplicate snap name when |
| 2103 | * doing the process (e.g., removed snapshot and recreated a new |
| 2104 | * one with the same name. |
| 2105 | */ |
| 2106 | static int __rbd_init_snaps_header(struct rbd_device *rbd_dev) |
| 2107 | { |
| 2108 | const char *name, *first_name; |
| 2109 | int i = rbd_dev->header.total_snaps; |
| 2110 | struct rbd_snap *snap, *old_snap = NULL; |
| 2111 | int ret; |
| 2112 | struct list_head *p, *n; |
| 2113 | |
| 2114 | first_name = rbd_dev->header.snap_names; |
| 2115 | name = first_name + rbd_dev->header.snap_names_len; |
| 2116 | |
| 2117 | list_for_each_prev_safe(p, n, &rbd_dev->snaps) { |
| 2118 | u64 cur_id; |
| 2119 | |
| 2120 | old_snap = list_entry(p, struct rbd_snap, node); |
| 2121 | |
| 2122 | if (i) |
| 2123 | cur_id = rbd_dev->header.snapc->snaps[i - 1]; |
| 2124 | |
| 2125 | if (!i || old_snap->id < cur_id) { |
Josh Durgin | e88a36e | 2011-11-21 18:14:25 -0800 | [diff] [blame] | 2126 | /* |
| 2127 | * old_snap->id was skipped, thus was |
| 2128 | * removed. If this rbd_dev is mapped to |
| 2129 | * the removed snapshot, record that it no |
| 2130 | * longer exists, to prevent further I/O. |
| 2131 | */ |
| 2132 | if (rbd_dev->snap_id == old_snap->id) |
| 2133 | rbd_dev->snap_exists = false; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2134 | __rbd_remove_snap_dev(rbd_dev, old_snap); |
| 2135 | continue; |
| 2136 | } |
| 2137 | if (old_snap->id == cur_id) { |
| 2138 | /* we have this snapshot already */ |
| 2139 | i--; |
| 2140 | name = rbd_prev_snap_name(name, first_name); |
| 2141 | continue; |
| 2142 | } |
| 2143 | for (; i > 0; |
| 2144 | i--, name = rbd_prev_snap_name(name, first_name)) { |
| 2145 | if (!name) { |
| 2146 | WARN_ON(1); |
| 2147 | return -EINVAL; |
| 2148 | } |
| 2149 | cur_id = rbd_dev->header.snapc->snaps[i]; |
| 2150 | /* snapshot removal? handle it above */ |
| 2151 | if (cur_id >= old_snap->id) |
| 2152 | break; |
| 2153 | /* a new snapshot */ |
| 2154 | ret = __rbd_add_snap_dev(rbd_dev, i - 1, name, &snap); |
| 2155 | if (ret < 0) |
| 2156 | return ret; |
| 2157 | |
| 2158 | /* note that we add it backward so using n and not p */ |
| 2159 | list_add(&snap->node, n); |
| 2160 | p = &snap->node; |
| 2161 | } |
| 2162 | } |
| 2163 | /* we're done going over the old snap list, just add what's left */ |
| 2164 | for (; i > 0; i--) { |
| 2165 | name = rbd_prev_snap_name(name, first_name); |
| 2166 | if (!name) { |
| 2167 | WARN_ON(1); |
| 2168 | return -EINVAL; |
| 2169 | } |
| 2170 | ret = __rbd_add_snap_dev(rbd_dev, i - 1, name, &snap); |
| 2171 | if (ret < 0) |
| 2172 | return ret; |
| 2173 | list_add(&snap->node, &rbd_dev->snaps); |
| 2174 | } |
| 2175 | |
| 2176 | return 0; |
| 2177 | } |
| 2178 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2179 | static int rbd_bus_add_dev(struct rbd_device *rbd_dev) |
| 2180 | { |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2181 | int ret; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2182 | struct device *dev; |
| 2183 | struct rbd_snap *snap; |
| 2184 | |
| 2185 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
| 2186 | dev = &rbd_dev->dev; |
| 2187 | |
| 2188 | dev->bus = &rbd_bus_type; |
| 2189 | dev->type = &rbd_device_type; |
| 2190 | dev->parent = &rbd_root_dev; |
| 2191 | dev->release = rbd_dev_release; |
| 2192 | dev_set_name(dev, "%d", rbd_dev->id); |
| 2193 | ret = device_register(dev); |
| 2194 | if (ret < 0) |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2195 | goto out; |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2196 | |
| 2197 | list_for_each_entry(snap, &rbd_dev->snaps, node) { |
| 2198 | ret = rbd_register_snap_dev(rbd_dev, snap, |
| 2199 | &rbd_dev->dev); |
| 2200 | if (ret < 0) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2201 | break; |
| 2202 | } |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2203 | out: |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2204 | mutex_unlock(&ctl_mutex); |
| 2205 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2206 | } |
| 2207 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2208 | static void rbd_bus_del_dev(struct rbd_device *rbd_dev) |
| 2209 | { |
| 2210 | device_unregister(&rbd_dev->dev); |
| 2211 | } |
| 2212 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2213 | static int rbd_init_watch_dev(struct rbd_device *rbd_dev) |
| 2214 | { |
| 2215 | int ret, rc; |
| 2216 | |
| 2217 | do { |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2218 | ret = rbd_req_sync_watch(rbd_dev, rbd_dev->header_name, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2219 | rbd_dev->header.obj_version); |
| 2220 | if (ret == -ERANGE) { |
| 2221 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
Josh Durgin | 263c6ca | 2011-12-05 10:43:42 -0800 | [diff] [blame] | 2222 | rc = __rbd_refresh_header(rbd_dev); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2223 | mutex_unlock(&ctl_mutex); |
| 2224 | if (rc < 0) |
| 2225 | return rc; |
| 2226 | } |
| 2227 | } while (ret == -ERANGE); |
| 2228 | |
| 2229 | return ret; |
| 2230 | } |
| 2231 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2232 | static atomic64_t rbd_id_max = ATOMIC64_INIT(0); |
| 2233 | |
| 2234 | /* |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2235 | * Get a unique rbd identifier for the given new rbd_dev, and add |
| 2236 | * 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] | 2237 | */ |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2238 | static void rbd_id_get(struct rbd_device *rbd_dev) |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 2239 | { |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2240 | rbd_dev->id = atomic64_inc_return(&rbd_id_max); |
| 2241 | |
| 2242 | spin_lock(&rbd_dev_list_lock); |
| 2243 | list_add_tail(&rbd_dev->node, &rbd_dev_list); |
| 2244 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2245 | } |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 2246 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2247 | /* |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2248 | * Remove an rbd_dev from the global list, and record that its |
| 2249 | * identifier is no longer in use. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2250 | */ |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2251 | static void rbd_id_put(struct rbd_device *rbd_dev) |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2252 | { |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2253 | struct list_head *tmp; |
| 2254 | int rbd_id = rbd_dev->id; |
| 2255 | int max_id; |
| 2256 | |
| 2257 | BUG_ON(rbd_id < 1); |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2258 | |
| 2259 | spin_lock(&rbd_dev_list_lock); |
| 2260 | list_del_init(&rbd_dev->node); |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2261 | |
| 2262 | /* |
| 2263 | * If the id being "put" is not the current maximum, there |
| 2264 | * is nothing special we need to do. |
| 2265 | */ |
| 2266 | if (rbd_id != atomic64_read(&rbd_id_max)) { |
| 2267 | spin_unlock(&rbd_dev_list_lock); |
| 2268 | return; |
| 2269 | } |
| 2270 | |
| 2271 | /* |
| 2272 | * We need to update the current maximum id. Search the |
| 2273 | * list to find out what it is. We're more likely to find |
| 2274 | * the maximum at the end, so search the list backward. |
| 2275 | */ |
| 2276 | max_id = 0; |
| 2277 | list_for_each_prev(tmp, &rbd_dev_list) { |
| 2278 | struct rbd_device *rbd_dev; |
| 2279 | |
| 2280 | rbd_dev = list_entry(tmp, struct rbd_device, node); |
| 2281 | if (rbd_id > max_id) |
| 2282 | max_id = rbd_id; |
| 2283 | } |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2284 | spin_unlock(&rbd_dev_list_lock); |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 2285 | |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2286 | /* |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2287 | * The max id could have been updated by rbd_id_get(), in |
| 2288 | * which case it now accurately reflects the new maximum. |
| 2289 | * Be careful not to overwrite the maximum value in that |
| 2290 | * case. |
Alex Elder | 1ddbe94 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2291 | */ |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2292 | atomic64_cmpxchg(&rbd_id_max, rbd_id, max_id); |
Alex Elder | b7f23c3 | 2012-01-29 13:57:43 -0600 | [diff] [blame] | 2293 | } |
| 2294 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2295 | /* |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2296 | * Skips over white space at *buf, and updates *buf to point to the |
| 2297 | * first found non-space character (if any). Returns the length of |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2298 | * the token (string of non-white space characters) found. Note |
| 2299 | * that *buf must be terminated with '\0'. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2300 | */ |
| 2301 | static inline size_t next_token(const char **buf) |
| 2302 | { |
| 2303 | /* |
| 2304 | * These are the characters that produce nonzero for |
| 2305 | * isspace() in the "C" and "POSIX" locales. |
| 2306 | */ |
| 2307 | const char *spaces = " \f\n\r\t\v"; |
| 2308 | |
| 2309 | *buf += strspn(*buf, spaces); /* Find start of token */ |
| 2310 | |
| 2311 | return strcspn(*buf, spaces); /* Return token length */ |
| 2312 | } |
| 2313 | |
| 2314 | /* |
| 2315 | * Finds the next token in *buf, and if the provided token buffer is |
| 2316 | * big enough, copies the found token into it. The result, if |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2317 | * copied, is guaranteed to be terminated with '\0'. Note that *buf |
| 2318 | * must be terminated with '\0' on entry. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2319 | * |
| 2320 | * Returns the length of the token found (not including the '\0'). |
| 2321 | * Return value will be 0 if no token is found, and it will be >= |
| 2322 | * token_size if the token would not fit. |
| 2323 | * |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2324 | * 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] | 2325 | * found token. Note that this occurs even if the token buffer is |
| 2326 | * too small to hold it. |
| 2327 | */ |
| 2328 | static inline size_t copy_token(const char **buf, |
| 2329 | char *token, |
| 2330 | size_t token_size) |
| 2331 | { |
| 2332 | size_t len; |
| 2333 | |
| 2334 | len = next_token(buf); |
| 2335 | if (len < token_size) { |
| 2336 | memcpy(token, *buf, len); |
| 2337 | *(token + len) = '\0'; |
| 2338 | } |
| 2339 | *buf += len; |
| 2340 | |
| 2341 | return len; |
| 2342 | } |
| 2343 | |
| 2344 | /* |
Alex Elder | ea3352f | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2345 | * Finds the next token in *buf, dynamically allocates a buffer big |
| 2346 | * enough to hold a copy of it, and copies the token into the new |
| 2347 | * buffer. The copy is guaranteed to be terminated with '\0'. Note |
| 2348 | * that a duplicate buffer is created even for a zero-length token. |
| 2349 | * |
| 2350 | * Returns a pointer to the newly-allocated duplicate, or a null |
| 2351 | * pointer if memory for the duplicate was not available. If |
| 2352 | * the lenp argument is a non-null pointer, the length of the token |
| 2353 | * (not including the '\0') is returned in *lenp. |
| 2354 | * |
| 2355 | * If successful, the *buf pointer will be updated to point beyond |
| 2356 | * the end of the found token. |
| 2357 | * |
| 2358 | * Note: uses GFP_KERNEL for allocation. |
| 2359 | */ |
| 2360 | static inline char *dup_token(const char **buf, size_t *lenp) |
| 2361 | { |
| 2362 | char *dup; |
| 2363 | size_t len; |
| 2364 | |
| 2365 | len = next_token(buf); |
| 2366 | dup = kmalloc(len + 1, GFP_KERNEL); |
| 2367 | if (!dup) |
| 2368 | return NULL; |
| 2369 | |
| 2370 | memcpy(dup, *buf, len); |
| 2371 | *(dup + len) = '\0'; |
| 2372 | *buf += len; |
| 2373 | |
| 2374 | if (lenp) |
| 2375 | *lenp = len; |
| 2376 | |
| 2377 | return dup; |
| 2378 | } |
| 2379 | |
| 2380 | /* |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2381 | * This fills in the pool_name, image_name, image_name_len, snap_name, |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2382 | * rbd_dev, rbd_md_name, and name fields of the given rbd_dev, based |
| 2383 | * on the list of monitor addresses and other options provided via |
| 2384 | * /sys/bus/rbd/add. |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2385 | * |
| 2386 | * Note: rbd_dev is assumed to have been initially zero-filled. |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2387 | */ |
| 2388 | static int rbd_add_parse_args(struct rbd_device *rbd_dev, |
| 2389 | const char *buf, |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2390 | const char **mon_addrs, |
Alex Elder | 5214ecc | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2391 | size_t *mon_addrs_size, |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2392 | char *options, |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2393 | size_t options_size) |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2394 | { |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2395 | size_t len; |
| 2396 | int ret; |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2397 | |
| 2398 | /* The first four tokens are required */ |
| 2399 | |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2400 | len = next_token(&buf); |
| 2401 | if (!len) |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2402 | return -EINVAL; |
Alex Elder | 5214ecc | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2403 | *mon_addrs_size = len + 1; |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2404 | *mon_addrs = buf; |
| 2405 | |
| 2406 | buf += len; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2407 | |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2408 | len = copy_token(&buf, options, options_size); |
| 2409 | if (!len || len >= options_size) |
| 2410 | return -EINVAL; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2411 | |
Alex Elder | bf3e5ae | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2412 | ret = -ENOMEM; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2413 | rbd_dev->pool_name = dup_token(&buf, NULL); |
| 2414 | if (!rbd_dev->pool_name) |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2415 | goto out_err; |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2416 | |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2417 | rbd_dev->image_name = dup_token(&buf, &rbd_dev->image_name_len); |
| 2418 | if (!rbd_dev->image_name) |
Alex Elder | bf3e5ae | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2419 | goto out_err; |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2420 | |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2421 | /* Create the name of the header object */ |
| 2422 | |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2423 | rbd_dev->header_name = kmalloc(rbd_dev->image_name_len |
Alex Elder | bf3e5ae | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2424 | + sizeof (RBD_SUFFIX), |
| 2425 | GFP_KERNEL); |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2426 | if (!rbd_dev->header_name) |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2427 | goto out_err; |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2428 | sprintf(rbd_dev->header_name, "%s%s", rbd_dev->image_name, RBD_SUFFIX); |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2429 | |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2430 | /* |
Alex Elder | 820a5f3 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 2431 | * The snapshot name is optional. If none is is supplied, |
| 2432 | * we use the default value. |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2433 | */ |
Alex Elder | 820a5f3 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 2434 | rbd_dev->snap_name = dup_token(&buf, &len); |
| 2435 | if (!rbd_dev->snap_name) |
| 2436 | goto out_err; |
| 2437 | if (!len) { |
| 2438 | /* Replace the empty name with the default */ |
| 2439 | kfree(rbd_dev->snap_name); |
| 2440 | rbd_dev->snap_name |
| 2441 | = kmalloc(sizeof (RBD_SNAP_HEAD_NAME), GFP_KERNEL); |
| 2442 | if (!rbd_dev->snap_name) |
| 2443 | goto out_err; |
| 2444 | |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2445 | memcpy(rbd_dev->snap_name, RBD_SNAP_HEAD_NAME, |
| 2446 | sizeof (RBD_SNAP_HEAD_NAME)); |
Alex Elder | 849b426 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 2447 | } |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2448 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2449 | return 0; |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2450 | |
| 2451 | out_err: |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2452 | kfree(rbd_dev->header_name); |
| 2453 | kfree(rbd_dev->image_name); |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2454 | kfree(rbd_dev->pool_name); |
| 2455 | rbd_dev->pool_name = NULL; |
| 2456 | |
| 2457 | return ret; |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2458 | } |
| 2459 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2460 | static ssize_t rbd_add(struct bus_type *bus, |
| 2461 | const char *buf, |
| 2462 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2463 | { |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2464 | char *options; |
| 2465 | struct rbd_device *rbd_dev = NULL; |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2466 | const char *mon_addrs = NULL; |
| 2467 | size_t mon_addrs_size = 0; |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2468 | struct ceph_osd_client *osdc; |
| 2469 | int rc = -ENOMEM; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2470 | |
| 2471 | if (!try_module_get(THIS_MODULE)) |
| 2472 | return -ENODEV; |
| 2473 | |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2474 | options = kmalloc(count, GFP_KERNEL); |
| 2475 | if (!options) |
| 2476 | goto err_nomem; |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2477 | rbd_dev = kzalloc(sizeof(*rbd_dev), GFP_KERNEL); |
| 2478 | if (!rbd_dev) |
| 2479 | goto err_nomem; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2480 | |
| 2481 | /* static rbd_device initialization */ |
| 2482 | spin_lock_init(&rbd_dev->lock); |
| 2483 | INIT_LIST_HEAD(&rbd_dev->node); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2484 | INIT_LIST_HEAD(&rbd_dev->snaps); |
Josh Durgin | c666601 | 2011-11-21 17:11:12 -0800 | [diff] [blame] | 2485 | init_rwsem(&rbd_dev->header_rwsem); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2486 | |
Josh Durgin | c666601 | 2011-11-21 17:11:12 -0800 | [diff] [blame] | 2487 | init_rwsem(&rbd_dev->header_rwsem); |
Alex Elder | 0e805a1 | 2012-01-11 19:42:15 -0800 | [diff] [blame] | 2488 | |
Alex Elder | d184f6b | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2489 | /* generate unique id: find highest unique id, add one */ |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2490 | rbd_id_get(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2491 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2492 | /* Fill in the device name, now that we have its id. */ |
Alex Elder | 81a8979 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2493 | BUILD_BUG_ON(DEV_NAME_LEN |
| 2494 | < sizeof (RBD_DRV_NAME) + MAX_INT_FORMAT_WIDTH); |
| 2495 | sprintf(rbd_dev->name, "%s%d", RBD_DRV_NAME, rbd_dev->id); |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2496 | |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2497 | /* parse add command */ |
Alex Elder | 7ef3214 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2498 | rc = rbd_add_parse_args(rbd_dev, buf, &mon_addrs, &mon_addrs_size, |
Alex Elder | e28fff26 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2499 | options, count); |
Alex Elder | a725f65e | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2500 | if (rc) |
| 2501 | goto err_put_id; |
| 2502 | |
Alex Elder | 5214ecc | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2503 | rbd_dev->rbd_client = rbd_get_client(mon_addrs, mon_addrs_size - 1, |
| 2504 | options); |
Alex Elder | d720bcb | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2505 | if (IS_ERR(rbd_dev->rbd_client)) { |
| 2506 | rc = PTR_ERR(rbd_dev->rbd_client); |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2507 | goto err_put_id; |
Alex Elder | d720bcb | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2508 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2509 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2510 | /* pick the pool */ |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 2511 | osdc = &rbd_dev->rbd_client->client->osdc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2512 | rc = ceph_pg_poolid_by_name(osdc->osdmap, rbd_dev->pool_name); |
| 2513 | if (rc < 0) |
| 2514 | goto err_out_client; |
Alex Elder | 9bb2f33 | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2515 | rbd_dev->pool_id = rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2516 | |
| 2517 | /* register our block device */ |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2518 | rc = register_blkdev(0, rbd_dev->name); |
| 2519 | if (rc < 0) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2520 | goto err_out_client; |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2521 | rbd_dev->major = rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2522 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2523 | rc = rbd_bus_add_dev(rbd_dev); |
| 2524 | if (rc) |
Yehuda Sadeh | 766fc43 | 2011-01-07 14:58:42 -0800 | [diff] [blame] | 2525 | goto err_out_blkdev; |
| 2526 | |
Alex Elder | 32eec68 | 2012-02-08 16:11:14 -0600 | [diff] [blame] | 2527 | /* |
| 2528 | * At this point cleanup in the event of an error is the job |
| 2529 | * of the sysfs code (initiated by rbd_bus_del_dev()). |
| 2530 | * |
| 2531 | * Set up and announce blkdev mapping. |
| 2532 | */ |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2533 | rc = rbd_init_disk(rbd_dev); |
| 2534 | if (rc) |
Yehuda Sadeh | 766fc43 | 2011-01-07 14:58:42 -0800 | [diff] [blame] | 2535 | goto err_out_bus; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2536 | |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2537 | rc = rbd_init_watch_dev(rbd_dev); |
| 2538 | if (rc) |
| 2539 | goto err_out_bus; |
| 2540 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2541 | return count; |
| 2542 | |
Yehuda Sadeh | 766fc43 | 2011-01-07 14:58:42 -0800 | [diff] [blame] | 2543 | err_out_bus: |
Yehuda Sadeh | 766fc43 | 2011-01-07 14:58:42 -0800 | [diff] [blame] | 2544 | /* this will also clean up rest of rbd_dev stuff */ |
| 2545 | |
| 2546 | rbd_bus_del_dev(rbd_dev); |
| 2547 | kfree(options); |
Yehuda Sadeh | 766fc43 | 2011-01-07 14:58:42 -0800 | [diff] [blame] | 2548 | return rc; |
| 2549 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2550 | err_out_blkdev: |
| 2551 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
| 2552 | err_out_client: |
| 2553 | rbd_put_client(rbd_dev); |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2554 | err_put_id: |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2555 | if (rbd_dev->pool_name) { |
Alex Elder | 820a5f3 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 2556 | kfree(rbd_dev->snap_name); |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2557 | kfree(rbd_dev->header_name); |
| 2558 | kfree(rbd_dev->image_name); |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2559 | kfree(rbd_dev->pool_name); |
| 2560 | } |
Alex Elder | 499afd5 | 2012-02-02 08:13:29 -0600 | [diff] [blame] | 2561 | rbd_id_put(rbd_dev); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2562 | err_nomem: |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2563 | kfree(rbd_dev); |
Alex Elder | cb8627c | 2012-07-09 21:04:23 -0500 | [diff] [blame] | 2564 | kfree(options); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2565 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2566 | dout("Error adding device %s\n", buf); |
| 2567 | module_put(THIS_MODULE); |
Alex Elder | 27cc259 | 2012-02-02 08:13:30 -0600 | [diff] [blame] | 2568 | |
| 2569 | return (ssize_t) rc; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2570 | } |
| 2571 | |
| 2572 | static struct rbd_device *__rbd_get_dev(unsigned long id) |
| 2573 | { |
| 2574 | struct list_head *tmp; |
| 2575 | struct rbd_device *rbd_dev; |
| 2576 | |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2577 | spin_lock(&rbd_dev_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2578 | list_for_each(tmp, &rbd_dev_list) { |
| 2579 | rbd_dev = list_entry(tmp, struct rbd_device, node); |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2580 | if (rbd_dev->id == id) { |
| 2581 | spin_unlock(&rbd_dev_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2582 | return rbd_dev; |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2583 | } |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2584 | } |
Alex Elder | e124a82 | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2585 | spin_unlock(&rbd_dev_list_lock); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2586 | return NULL; |
| 2587 | } |
| 2588 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2589 | static void rbd_dev_release(struct device *dev) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2590 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2591 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2592 | |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 2593 | if (rbd_dev->watch_request) { |
| 2594 | struct ceph_client *client = rbd_dev->rbd_client->client; |
| 2595 | |
| 2596 | ceph_osdc_unregister_linger_request(&client->osdc, |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2597 | rbd_dev->watch_request); |
Alex Elder | 1dbb439 | 2012-01-24 10:08:37 -0600 | [diff] [blame] | 2598 | } |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2599 | if (rbd_dev->watch_event) |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2600 | rbd_req_sync_unwatch(rbd_dev, rbd_dev->header_name); |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2601 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2602 | rbd_put_client(rbd_dev); |
| 2603 | |
| 2604 | /* clean up and free blkdev */ |
| 2605 | rbd_free_disk(rbd_dev); |
| 2606 | unregister_blkdev(rbd_dev->major, rbd_dev->name); |
Alex Elder | 32eec68 | 2012-02-08 16:11:14 -0600 | [diff] [blame] | 2607 | |
| 2608 | /* done with the id, and with the rbd_dev */ |
Alex Elder | 820a5f3 | 2012-07-09 21:04:24 -0500 | [diff] [blame] | 2609 | kfree(rbd_dev->snap_name); |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2610 | kfree(rbd_dev->header_name); |
Alex Elder | d22f76e | 2012-07-12 10:46:35 -0500 | [diff] [blame] | 2611 | kfree(rbd_dev->pool_name); |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2612 | kfree(rbd_dev->image_name); |
Alex Elder | 32eec68 | 2012-02-08 16:11:14 -0600 | [diff] [blame] | 2613 | rbd_id_put(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2614 | kfree(rbd_dev); |
| 2615 | |
| 2616 | /* release module ref */ |
| 2617 | module_put(THIS_MODULE); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2618 | } |
| 2619 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2620 | static ssize_t rbd_remove(struct bus_type *bus, |
| 2621 | const char *buf, |
| 2622 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2623 | { |
| 2624 | struct rbd_device *rbd_dev = NULL; |
| 2625 | int target_id, rc; |
| 2626 | unsigned long ul; |
| 2627 | int ret = count; |
| 2628 | |
| 2629 | rc = strict_strtoul(buf, 10, &ul); |
| 2630 | if (rc) |
| 2631 | return rc; |
| 2632 | |
| 2633 | /* convert to int; abort if we lost anything in the conversion */ |
| 2634 | target_id = (int) ul; |
| 2635 | if (target_id != ul) |
| 2636 | return -EINVAL; |
| 2637 | |
| 2638 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
| 2639 | |
| 2640 | rbd_dev = __rbd_get_dev(target_id); |
| 2641 | if (!rbd_dev) { |
| 2642 | ret = -ENOENT; |
| 2643 | goto done; |
| 2644 | } |
| 2645 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2646 | __rbd_remove_all_snaps(rbd_dev); |
| 2647 | rbd_bus_del_dev(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2648 | |
| 2649 | done: |
| 2650 | mutex_unlock(&ctl_mutex); |
| 2651 | return ret; |
| 2652 | } |
| 2653 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2654 | static ssize_t rbd_snap_add(struct device *dev, |
| 2655 | struct device_attribute *attr, |
| 2656 | const char *buf, |
| 2657 | size_t count) |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2658 | { |
Alex Elder | 593a9e7 | 2012-02-07 12:03:37 -0600 | [diff] [blame] | 2659 | struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2660 | int ret; |
| 2661 | char *name = kmalloc(count + 1, GFP_KERNEL); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2662 | if (!name) |
| 2663 | return -ENOMEM; |
| 2664 | |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2665 | snprintf(name, count, "%s", buf); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2666 | |
| 2667 | mutex_lock_nested(&ctl_mutex, SINGLE_DEPTH_NESTING); |
| 2668 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2669 | ret = rbd_header_add_snap(rbd_dev, |
| 2670 | name, GFP_KERNEL); |
| 2671 | if (ret < 0) |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2672 | goto err_unlock; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2673 | |
Josh Durgin | 263c6ca | 2011-12-05 10:43:42 -0800 | [diff] [blame] | 2674 | ret = __rbd_refresh_header(rbd_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2675 | if (ret < 0) |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2676 | goto err_unlock; |
| 2677 | |
| 2678 | /* shouldn't hold ctl_mutex when notifying.. notify might |
| 2679 | trigger a watch callback that would need to get that mutex */ |
| 2680 | mutex_unlock(&ctl_mutex); |
| 2681 | |
| 2682 | /* make a best effort, don't error if failed */ |
Alex Elder | 0bed54d | 2012-07-03 16:01:18 -0500 | [diff] [blame] | 2683 | rbd_req_sync_notify(rbd_dev, rbd_dev->header_name); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2684 | |
| 2685 | ret = count; |
Yehuda Sadeh | 59c2be1 | 2011-03-21 15:10:11 -0700 | [diff] [blame] | 2686 | kfree(name); |
| 2687 | return ret; |
| 2688 | |
| 2689 | err_unlock: |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2690 | mutex_unlock(&ctl_mutex); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2691 | kfree(name); |
| 2692 | return ret; |
| 2693 | } |
| 2694 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2695 | /* |
| 2696 | * create control files in sysfs |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2697 | * /sys/bus/rbd/... |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2698 | */ |
| 2699 | static int rbd_sysfs_init(void) |
| 2700 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2701 | int ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2702 | |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 2703 | ret = device_register(&rbd_root_dev); |
Alex Elder | 2107978 | 2012-01-24 10:08:36 -0600 | [diff] [blame] | 2704 | if (ret < 0) |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2705 | return ret; |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2706 | |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 2707 | ret = bus_register(&rbd_bus_type); |
| 2708 | if (ret < 0) |
| 2709 | device_unregister(&rbd_root_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2710 | |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2711 | return ret; |
| 2712 | } |
| 2713 | |
| 2714 | static void rbd_sysfs_cleanup(void) |
| 2715 | { |
Yehuda Sadeh | dfc5606 | 2010-11-19 14:51:04 -0800 | [diff] [blame] | 2716 | bus_unregister(&rbd_bus_type); |
Alex Elder | fed4c14 | 2012-02-07 12:03:36 -0600 | [diff] [blame] | 2717 | device_unregister(&rbd_root_dev); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2718 | } |
| 2719 | |
| 2720 | int __init rbd_init(void) |
| 2721 | { |
| 2722 | int rc; |
| 2723 | |
| 2724 | rc = rbd_sysfs_init(); |
| 2725 | if (rc) |
| 2726 | return rc; |
Alex Elder | f0f8cef | 2012-01-29 13:57:44 -0600 | [diff] [blame] | 2727 | pr_info("loaded " RBD_DRV_NAME_LONG "\n"); |
Yehuda Sadeh | 602adf4 | 2010-08-12 16:11:25 -0700 | [diff] [blame] | 2728 | return 0; |
| 2729 | } |
| 2730 | |
| 2731 | void __exit rbd_exit(void) |
| 2732 | { |
| 2733 | rbd_sysfs_cleanup(); |
| 2734 | } |
| 2735 | |
| 2736 | module_init(rbd_init); |
| 2737 | module_exit(rbd_exit); |
| 2738 | |
| 2739 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); |
| 2740 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); |
| 2741 | MODULE_DESCRIPTION("rados block device"); |
| 2742 | |
| 2743 | /* following authorship retained from original osdblk.c */ |
| 2744 | MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); |
| 2745 | |
| 2746 | MODULE_LICENSE("GPL"); |