1 2 /* 3 rbd.c -- Export ceph rados objects as a Linux block device 4 5 6 based on drivers/block/osdblk.c: 7 8 Copyright 2009 Red Hat, Inc. 9 10 This program is free software; you can redistribute it and/or modify 11 it under the terms of the GNU General Public License as published by 12 the Free Software Foundation. 13 14 This program is distributed in the hope that it will be useful, 15 but WITHOUT ANY WARRANTY; without even the implied warranty of 16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 17 GNU General Public License for more details. 18 19 You should have received a copy of the GNU General Public License 20 along with this program; see the file COPYING. If not, write to 21 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 22 23 24 25 For usage instructions, please refer to: 26 27 Documentation/ABI/testing/sysfs-bus-rbd 28 29 */ 30 31 #include <linux/ceph/libceph.h> 32 #include <linux/ceph/osd_client.h> 33 #include <linux/ceph/mon_client.h> 34 #include <linux/ceph/cls_lock_client.h> 35 #include <linux/ceph/striper.h> 36 #include <linux/ceph/decode.h> 37 #include <linux/fs_parser.h> 38 #include <linux/bsearch.h> 39 40 #include <linux/kernel.h> 41 #include <linux/device.h> 42 #include <linux/module.h> 43 #include <linux/blk-mq.h> 44 #include <linux/fs.h> 45 #include <linux/blkdev.h> 46 #include <linux/slab.h> 47 #include <linux/idr.h> 48 #include <linux/workqueue.h> 49 50 #include "rbd_types.h" 51 52 #define RBD_DEBUG /* Activate rbd_assert() calls */ 53 54 /* 55 * Increment the given counter and return its updated value. 56 * If the counter is already 0 it will not be incremented. 57 * If the counter is already at its maximum value returns 58 * -EINVAL without updating it. 59 */ 60 static int atomic_inc_return_safe(atomic_t *v) 61 { 62 unsigned int counter; 63 64 counter = (unsigned int)atomic_fetch_add_unless(v, 1, 0); 65 if (counter <= (unsigned int)INT_MAX) 66 return (int)counter; 67 68 atomic_dec(v); 69 70 return -EINVAL; 71 } 72 73 /* Decrement the counter. Return the resulting value, or -EINVAL */ 74 static int atomic_dec_return_safe(atomic_t *v) 75 { 76 int counter; 77 78 counter = atomic_dec_return(v); 79 if (counter >= 0) 80 return counter; 81 82 atomic_inc(v); 83 84 return -EINVAL; 85 } 86 87 #define RBD_DRV_NAME "rbd" 88 89 #define RBD_MINORS_PER_MAJOR 256 90 #define RBD_SINGLE_MAJOR_PART_SHIFT 4 91 92 #define RBD_MAX_PARENT_CHAIN_LEN 16 93 94 #define RBD_SNAP_DEV_NAME_PREFIX "snap_" 95 #define RBD_MAX_SNAP_NAME_LEN \ 96 (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1)) 97 98 #define RBD_MAX_SNAP_COUNT 510 /* allows max snapc to fit in 4KB */ 99 100 #define RBD_SNAP_HEAD_NAME "-" 101 102 #define BAD_SNAP_INDEX U32_MAX /* invalid index into snap array */ 103 104 /* This allows a single page to hold an image name sent by OSD */ 105 #define RBD_IMAGE_NAME_LEN_MAX (PAGE_SIZE - sizeof (__le32) - 1) 106 #define RBD_IMAGE_ID_LEN_MAX 64 107 108 #define RBD_OBJ_PREFIX_LEN_MAX 64 109 110 #define RBD_NOTIFY_TIMEOUT 5 /* seconds */ 111 #define RBD_RETRY_DELAY msecs_to_jiffies(1000) 112 113 /* Feature bits */ 114 115 #define RBD_FEATURE_LAYERING (1ULL<<0) 116 #define RBD_FEATURE_STRIPINGV2 (1ULL<<1) 117 #define RBD_FEATURE_EXCLUSIVE_LOCK (1ULL<<2) 118 #define RBD_FEATURE_OBJECT_MAP (1ULL<<3) 119 #define RBD_FEATURE_FAST_DIFF (1ULL<<4) 120 #define RBD_FEATURE_DEEP_FLATTEN (1ULL<<5) 121 #define RBD_FEATURE_DATA_POOL (1ULL<<7) 122 #define RBD_FEATURE_OPERATIONS (1ULL<<8) 123 124 #define RBD_FEATURES_ALL (RBD_FEATURE_LAYERING | \ 125 RBD_FEATURE_STRIPINGV2 | \ 126 RBD_FEATURE_EXCLUSIVE_LOCK | \ 127 RBD_FEATURE_OBJECT_MAP | \ 128 RBD_FEATURE_FAST_DIFF | \ 129 RBD_FEATURE_DEEP_FLATTEN | \ 130 RBD_FEATURE_DATA_POOL | \ 131 RBD_FEATURE_OPERATIONS) 132 133 /* Features supported by this (client software) implementation. */ 134 135 #define RBD_FEATURES_SUPPORTED (RBD_FEATURES_ALL) 136 137 /* 138 * An RBD device name will be "rbd#", where the "rbd" comes from 139 * RBD_DRV_NAME above, and # is a unique integer identifier. 140 */ 141 #define DEV_NAME_LEN 32 142 143 /* 144 * block device image metadata (in-memory version) 145 */ 146 struct rbd_image_header { 147 /* These six fields never change for a given rbd image */ 148 char *object_prefix; 149 __u8 obj_order; 150 u64 stripe_unit; 151 u64 stripe_count; 152 s64 data_pool_id; 153 u64 features; /* Might be changeable someday? */ 154 155 /* The remaining fields need to be updated occasionally */ 156 u64 image_size; 157 struct ceph_snap_context *snapc; 158 char *snap_names; /* format 1 only */ 159 u64 *snap_sizes; /* format 1 only */ 160 }; 161 162 /* 163 * An rbd image specification. 164 * 165 * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely 166 * identify an image. Each rbd_dev structure includes a pointer to 167 * an rbd_spec structure that encapsulates this identity. 168 * 169 * Each of the id's in an rbd_spec has an associated name. For a 170 * user-mapped image, the names are supplied and the id's associated 171 * with them are looked up. For a layered image, a parent image is 172 * defined by the tuple, and the names are looked up. 173 * 174 * An rbd_dev structure contains a parent_spec pointer which is 175 * non-null if the image it represents is a child in a layered 176 * image. This pointer will refer to the rbd_spec structure used 177 * by the parent rbd_dev for its own identity (i.e., the structure 178 * is shared between the parent and child). 179 * 180 * Since these structures are populated once, during the discovery 181 * phase of image construction, they are effectively immutable so 182 * we make no effort to synchronize access to them. 183 * 184 * Note that code herein does not assume the image name is known (it 185 * could be a null pointer). 186 */ 187 struct rbd_spec { 188 u64 pool_id; 189 const char *pool_name; 190 const char *pool_ns; /* NULL if default, never "" */ 191 192 const char *image_id; 193 const char *image_name; 194 195 u64 snap_id; 196 const char *snap_name; 197 198 struct kref kref; 199 }; 200 201 /* 202 * an instance of the client. multiple devices may share an rbd client. 203 */ 204 struct rbd_client { 205 struct ceph_client *client; 206 struct kref kref; 207 struct list_head node; 208 }; 209 210 struct pending_result { 211 int result; /* first nonzero result */ 212 int num_pending; 213 }; 214 215 struct rbd_img_request; 216 217 enum obj_request_type { 218 OBJ_REQUEST_NODATA = 1, 219 OBJ_REQUEST_BIO, /* pointer into provided bio (list) */ 220 OBJ_REQUEST_BVECS, /* pointer into provided bio_vec array */ 221 OBJ_REQUEST_OWN_BVECS, /* private bio_vec array, doesn't own pages */ 222 }; 223 224 enum obj_operation_type { 225 OBJ_OP_READ = 1, 226 OBJ_OP_WRITE, 227 OBJ_OP_DISCARD, 228 OBJ_OP_ZEROOUT, 229 }; 230 231 #define RBD_OBJ_FLAG_DELETION (1U << 0) 232 #define RBD_OBJ_FLAG_COPYUP_ENABLED (1U << 1) 233 #define RBD_OBJ_FLAG_COPYUP_ZEROS (1U << 2) 234 #define RBD_OBJ_FLAG_MAY_EXIST (1U << 3) 235 #define RBD_OBJ_FLAG_NOOP_FOR_NONEXISTENT (1U << 4) 236 237 enum rbd_obj_read_state { 238 RBD_OBJ_READ_START = 1, 239 RBD_OBJ_READ_OBJECT, 240 RBD_OBJ_READ_PARENT, 241 }; 242 243 /* 244 * Writes go through the following state machine to deal with 245 * layering: 246 * 247 * . . . . . RBD_OBJ_WRITE_GUARD. . . . . . . . . . . . . . 248 * . | . 249 * . v . 250 * . RBD_OBJ_WRITE_READ_FROM_PARENT. . . . 251 * . | . . 252 * . v v (deep-copyup . 253 * (image . RBD_OBJ_WRITE_COPYUP_EMPTY_SNAPC . not needed) . 254 * flattened) v | . . 255 * . v . . 256 * . . . .RBD_OBJ_WRITE_COPYUP_OPS. . . . . (copyup . 257 * | not needed) v 258 * v . 259 * done . . . . . . . . . . . . . . . . . . 260 * ^ 261 * | 262 * RBD_OBJ_WRITE_FLAT 263 * 264 * Writes start in RBD_OBJ_WRITE_GUARD or _FLAT, depending on whether 265 * assert_exists guard is needed or not (in some cases it's not needed 266 * even if there is a parent). 267 */ 268 enum rbd_obj_write_state { 269 RBD_OBJ_WRITE_START = 1, 270 RBD_OBJ_WRITE_PRE_OBJECT_MAP, 271 RBD_OBJ_WRITE_OBJECT, 272 __RBD_OBJ_WRITE_COPYUP, 273 RBD_OBJ_WRITE_COPYUP, 274 RBD_OBJ_WRITE_POST_OBJECT_MAP, 275 }; 276 277 enum rbd_obj_copyup_state { 278 RBD_OBJ_COPYUP_START = 1, 279 RBD_OBJ_COPYUP_READ_PARENT, 280 __RBD_OBJ_COPYUP_OBJECT_MAPS, 281 RBD_OBJ_COPYUP_OBJECT_MAPS, 282 __RBD_OBJ_COPYUP_WRITE_OBJECT, 283 RBD_OBJ_COPYUP_WRITE_OBJECT, 284 }; 285 286 struct rbd_obj_request { 287 struct ceph_object_extent ex; 288 unsigned int flags; /* RBD_OBJ_FLAG_* */ 289 union { 290 enum rbd_obj_read_state read_state; /* for reads */ 291 enum rbd_obj_write_state write_state; /* for writes */ 292 }; 293 294 struct rbd_img_request *img_request; 295 struct ceph_file_extent *img_extents; 296 u32 num_img_extents; 297 298 union { 299 struct ceph_bio_iter bio_pos; 300 struct { 301 struct ceph_bvec_iter bvec_pos; 302 u32 bvec_count; 303 u32 bvec_idx; 304 }; 305 }; 306 307 enum rbd_obj_copyup_state copyup_state; 308 struct bio_vec *copyup_bvecs; 309 u32 copyup_bvec_count; 310 311 struct list_head osd_reqs; /* w/ r_private_item */ 312 313 struct mutex state_mutex; 314 struct pending_result pending; 315 struct kref kref; 316 }; 317 318 enum img_req_flags { 319 IMG_REQ_CHILD, /* initiator: block = 0, child image = 1 */ 320 IMG_REQ_LAYERED, /* ENOENT handling: normal = 0, layered = 1 */ 321 }; 322 323 enum rbd_img_state { 324 RBD_IMG_START = 1, 325 RBD_IMG_EXCLUSIVE_LOCK, 326 __RBD_IMG_OBJECT_REQUESTS, 327 RBD_IMG_OBJECT_REQUESTS, 328 }; 329 330 struct rbd_img_request { 331 struct rbd_device *rbd_dev; 332 enum obj_operation_type op_type; 333 enum obj_request_type data_type; 334 unsigned long flags; 335 enum rbd_img_state state; 336 union { 337 u64 snap_id; /* for reads */ 338 struct ceph_snap_context *snapc; /* for writes */ 339 }; 340 struct rbd_obj_request *obj_request; /* obj req initiator */ 341 342 struct list_head lock_item; 343 struct list_head object_extents; /* obj_req.ex structs */ 344 345 struct mutex state_mutex; 346 struct pending_result pending; 347 struct work_struct work; 348 int work_result; 349 }; 350 351 #define for_each_obj_request(ireq, oreq) \ 352 list_for_each_entry(oreq, &(ireq)->object_extents, ex.oe_item) 353 #define for_each_obj_request_safe(ireq, oreq, n) \ 354 list_for_each_entry_safe(oreq, n, &(ireq)->object_extents, ex.oe_item) 355 356 enum rbd_watch_state { 357 RBD_WATCH_STATE_UNREGISTERED, 358 RBD_WATCH_STATE_REGISTERED, 359 RBD_WATCH_STATE_ERROR, 360 }; 361 362 enum rbd_lock_state { 363 RBD_LOCK_STATE_UNLOCKED, 364 RBD_LOCK_STATE_LOCKED, 365 RBD_LOCK_STATE_RELEASING, 366 }; 367 368 /* WatchNotify::ClientId */ 369 struct rbd_client_id { 370 u64 gid; 371 u64 handle; 372 }; 373 374 struct rbd_mapping { 375 u64 size; 376 }; 377 378 /* 379 * a single device 380 */ 381 struct rbd_device { 382 int dev_id; /* blkdev unique id */ 383 384 int major; /* blkdev assigned major */ 385 int minor; 386 struct gendisk *disk; /* blkdev's gendisk and rq */ 387 388 u32 image_format; /* Either 1 or 2 */ 389 struct rbd_client *rbd_client; 390 391 char name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */ 392 393 spinlock_t lock; /* queue, flags, open_count */ 394 395 struct rbd_image_header header; 396 unsigned long flags; /* possibly lock protected */ 397 struct rbd_spec *spec; 398 struct rbd_options *opts; 399 char *config_info; /* add{,_single_major} string */ 400 401 struct ceph_object_id header_oid; 402 struct ceph_object_locator header_oloc; 403 404 struct ceph_file_layout layout; /* used for all rbd requests */ 405 406 struct mutex watch_mutex; 407 enum rbd_watch_state watch_state; 408 struct ceph_osd_linger_request *watch_handle; 409 u64 watch_cookie; 410 struct delayed_work watch_dwork; 411 412 struct rw_semaphore lock_rwsem; 413 enum rbd_lock_state lock_state; 414 char lock_cookie[32]; 415 struct rbd_client_id owner_cid; 416 struct work_struct acquired_lock_work; 417 struct work_struct released_lock_work; 418 struct delayed_work lock_dwork; 419 struct work_struct unlock_work; 420 spinlock_t lock_lists_lock; 421 struct list_head acquiring_list; 422 struct list_head running_list; 423 struct completion acquire_wait; 424 int acquire_err; 425 struct completion releasing_wait; 426 427 spinlock_t object_map_lock; 428 u8 *object_map; 429 u64 object_map_size; /* in objects */ 430 u64 object_map_flags; 431 432 struct workqueue_struct *task_wq; 433 434 struct rbd_spec *parent_spec; 435 u64 parent_overlap; 436 atomic_t parent_ref; 437 struct rbd_device *parent; 438 439 /* Block layer tags. */ 440 struct blk_mq_tag_set tag_set; 441 442 /* protects updating the header */ 443 struct rw_semaphore header_rwsem; 444 445 struct rbd_mapping mapping; 446 447 struct list_head node; 448 449 /* sysfs related */ 450 struct device dev; 451 unsigned long open_count; /* protected by lock */ 452 }; 453 454 /* 455 * Flag bits for rbd_dev->flags: 456 * - REMOVING (which is coupled with rbd_dev->open_count) is protected 457 * by rbd_dev->lock 458 */ 459 enum rbd_dev_flags { 460 RBD_DEV_FLAG_EXISTS, /* rbd_dev_device_setup() ran */ 461 RBD_DEV_FLAG_REMOVING, /* this mapping is being removed */ 462 RBD_DEV_FLAG_READONLY, /* -o ro or snapshot */ 463 }; 464 465 static DEFINE_MUTEX(client_mutex); /* Serialize client creation */ 466 467 static LIST_HEAD(rbd_dev_list); /* devices */ 468 static DEFINE_SPINLOCK(rbd_dev_list_lock); 469 470 static LIST_HEAD(rbd_client_list); /* clients */ 471 static DEFINE_SPINLOCK(rbd_client_list_lock); 472 473 /* Slab caches for frequently-allocated structures */ 474 475 static struct kmem_cache *rbd_img_request_cache; 476 static struct kmem_cache *rbd_obj_request_cache; 477 478 static int rbd_major; 479 static DEFINE_IDA(rbd_dev_id_ida); 480 481 static struct workqueue_struct *rbd_wq; 482 483 static struct ceph_snap_context rbd_empty_snapc = { 484 .nref = REFCOUNT_INIT(1), 485 }; 486 487 /* 488 * single-major requires >= 0.75 version of userspace rbd utility. 489 */ 490 static bool single_major = true; 491 module_param(single_major, bool, 0444); 492 MODULE_PARM_DESC(single_major, "Use a single major number for all rbd devices (default: true)"); 493 494 static ssize_t add_store(struct bus_type *bus, const char *buf, size_t count); 495 static ssize_t remove_store(struct bus_type *bus, const char *buf, 496 size_t count); 497 static ssize_t add_single_major_store(struct bus_type *bus, const char *buf, 498 size_t count); 499 static ssize_t remove_single_major_store(struct bus_type *bus, const char *buf, 500 size_t count); 501 static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth); 502 503 static int rbd_dev_id_to_minor(int dev_id) 504 { 505 return dev_id << RBD_SINGLE_MAJOR_PART_SHIFT; 506 } 507 508 static int minor_to_rbd_dev_id(int minor) 509 { 510 return minor >> RBD_SINGLE_MAJOR_PART_SHIFT; 511 } 512 513 static bool rbd_is_ro(struct rbd_device *rbd_dev) 514 { 515 return test_bit(RBD_DEV_FLAG_READONLY, &rbd_dev->flags); 516 } 517 518 static bool rbd_is_snap(struct rbd_device *rbd_dev) 519 { 520 return rbd_dev->spec->snap_id != CEPH_NOSNAP; 521 } 522 523 static bool __rbd_is_lock_owner(struct rbd_device *rbd_dev) 524 { 525 lockdep_assert_held(&rbd_dev->lock_rwsem); 526 527 return rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED || 528 rbd_dev->lock_state == RBD_LOCK_STATE_RELEASING; 529 } 530 531 static bool rbd_is_lock_owner(struct rbd_device *rbd_dev) 532 { 533 bool is_lock_owner; 534 535 down_read(&rbd_dev->lock_rwsem); 536 is_lock_owner = __rbd_is_lock_owner(rbd_dev); 537 up_read(&rbd_dev->lock_rwsem); 538 return is_lock_owner; 539 } 540 541 static ssize_t supported_features_show(struct bus_type *bus, char *buf) 542 { 543 return sprintf(buf, "0x%llx\n", RBD_FEATURES_SUPPORTED); 544 } 545 546 static BUS_ATTR_WO(add); 547 static BUS_ATTR_WO(remove); 548 static BUS_ATTR_WO(add_single_major); 549 static BUS_ATTR_WO(remove_single_major); 550 static BUS_ATTR_RO(supported_features); 551 552 static struct attribute *rbd_bus_attrs[] = { 553 &bus_attr_add.attr, 554 &bus_attr_remove.attr, 555 &bus_attr_add_single_major.attr, 556 &bus_attr_remove_single_major.attr, 557 &bus_attr_supported_features.attr, 558 NULL, 559 }; 560 561 static umode_t rbd_bus_is_visible(struct kobject *kobj, 562 struct attribute *attr, int index) 563 { 564 if (!single_major && 565 (attr == &bus_attr_add_single_major.attr || 566 attr == &bus_attr_remove_single_major.attr)) 567 return 0; 568 569 return attr->mode; 570 } 571 572 static const struct attribute_group rbd_bus_group = { 573 .attrs = rbd_bus_attrs, 574 .is_visible = rbd_bus_is_visible, 575 }; 576 __ATTRIBUTE_GROUPS(rbd_bus); 577 578 static struct bus_type rbd_bus_type = { 579 .name = "rbd", 580 .bus_groups = rbd_bus_groups, 581 }; 582 583 static void rbd_root_dev_release(struct device *dev) 584 { 585 } 586 587 static struct device rbd_root_dev = { 588 .init_name = "rbd", 589 .release = rbd_root_dev_release, 590 }; 591 592 static __printf(2, 3) 593 void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...) 594 { 595 struct va_format vaf; 596 va_list args; 597 598 va_start(args, fmt); 599 vaf.fmt = fmt; 600 vaf.va = &args; 601 602 if (!rbd_dev) 603 printk(KERN_WARNING "%s: %pV\n", RBD_DRV_NAME, &vaf); 604 else if (rbd_dev->disk) 605 printk(KERN_WARNING "%s: %s: %pV\n", 606 RBD_DRV_NAME, rbd_dev->disk->disk_name, &vaf); 607 else if (rbd_dev->spec && rbd_dev->spec->image_name) 608 printk(KERN_WARNING "%s: image %s: %pV\n", 609 RBD_DRV_NAME, rbd_dev->spec->image_name, &vaf); 610 else if (rbd_dev->spec && rbd_dev->spec->image_id) 611 printk(KERN_WARNING "%s: id %s: %pV\n", 612 RBD_DRV_NAME, rbd_dev->spec->image_id, &vaf); 613 else /* punt */ 614 printk(KERN_WARNING "%s: rbd_dev %p: %pV\n", 615 RBD_DRV_NAME, rbd_dev, &vaf); 616 va_end(args); 617 } 618 619 #ifdef RBD_DEBUG 620 #define rbd_assert(expr) \ 621 if (unlikely(!(expr))) { \ 622 printk(KERN_ERR "\nAssertion failure in %s() " \ 623 "at line %d:\n\n" \ 624 "\trbd_assert(%s);\n\n", \ 625 __func__, __LINE__, #expr); \ 626 BUG(); \ 627 } 628 #else /* !RBD_DEBUG */ 629 # define rbd_assert(expr) ((void) 0) 630 #endif /* !RBD_DEBUG */ 631 632 static void rbd_dev_remove_parent(struct rbd_device *rbd_dev); 633 634 static int rbd_dev_refresh(struct rbd_device *rbd_dev); 635 static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev); 636 static int rbd_dev_header_info(struct rbd_device *rbd_dev); 637 static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev); 638 static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, 639 u64 snap_id); 640 static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, 641 u8 *order, u64 *snap_size); 642 static int rbd_dev_v2_get_flags(struct rbd_device *rbd_dev); 643 644 static void rbd_obj_handle_request(struct rbd_obj_request *obj_req, int result); 645 static void rbd_img_handle_request(struct rbd_img_request *img_req, int result); 646 647 /* 648 * Return true if nothing else is pending. 649 */ 650 static bool pending_result_dec(struct pending_result *pending, int *result) 651 { 652 rbd_assert(pending->num_pending > 0); 653 654 if (*result && !pending->result) 655 pending->result = *result; 656 if (--pending->num_pending) 657 return false; 658 659 *result = pending->result; 660 return true; 661 } 662 663 static int rbd_open(struct block_device *bdev, fmode_t mode) 664 { 665 struct rbd_device *rbd_dev = bdev->bd_disk->private_data; 666 bool removing = false; 667 668 spin_lock_irq(&rbd_dev->lock); 669 if (test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags)) 670 removing = true; 671 else 672 rbd_dev->open_count++; 673 spin_unlock_irq(&rbd_dev->lock); 674 if (removing) 675 return -ENOENT; 676 677 (void) get_device(&rbd_dev->dev); 678 679 return 0; 680 } 681 682 static void rbd_release(struct gendisk *disk, fmode_t mode) 683 { 684 struct rbd_device *rbd_dev = disk->private_data; 685 unsigned long open_count_before; 686 687 spin_lock_irq(&rbd_dev->lock); 688 open_count_before = rbd_dev->open_count--; 689 spin_unlock_irq(&rbd_dev->lock); 690 rbd_assert(open_count_before > 0); 691 692 put_device(&rbd_dev->dev); 693 } 694 695 static const struct block_device_operations rbd_bd_ops = { 696 .owner = THIS_MODULE, 697 .open = rbd_open, 698 .release = rbd_release, 699 }; 700 701 /* 702 * Initialize an rbd client instance. Success or not, this function 703 * consumes ceph_opts. Caller holds client_mutex. 704 */ 705 static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts) 706 { 707 struct rbd_client *rbdc; 708 int ret = -ENOMEM; 709 710 dout("%s:\n", __func__); 711 rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL); 712 if (!rbdc) 713 goto out_opt; 714 715 kref_init(&rbdc->kref); 716 INIT_LIST_HEAD(&rbdc->node); 717 718 rbdc->client = ceph_create_client(ceph_opts, rbdc); 719 if (IS_ERR(rbdc->client)) 720 goto out_rbdc; 721 ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */ 722 723 ret = ceph_open_session(rbdc->client); 724 if (ret < 0) 725 goto out_client; 726 727 spin_lock(&rbd_client_list_lock); 728 list_add_tail(&rbdc->node, &rbd_client_list); 729 spin_unlock(&rbd_client_list_lock); 730 731 dout("%s: rbdc %p\n", __func__, rbdc); 732 733 return rbdc; 734 out_client: 735 ceph_destroy_client(rbdc->client); 736 out_rbdc: 737 kfree(rbdc); 738 out_opt: 739 if (ceph_opts) 740 ceph_destroy_options(ceph_opts); 741 dout("%s: error %d\n", __func__, ret); 742 743 return ERR_PTR(ret); 744 } 745 746 static struct rbd_client *__rbd_get_client(struct rbd_client *rbdc) 747 { 748 kref_get(&rbdc->kref); 749 750 return rbdc; 751 } 752 753 /* 754 * Find a ceph client with specific addr and configuration. If 755 * found, bump its reference count. 756 */ 757 static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts) 758 { 759 struct rbd_client *client_node; 760 bool found = false; 761 762 if (ceph_opts->flags & CEPH_OPT_NOSHARE) 763 return NULL; 764 765 spin_lock(&rbd_client_list_lock); 766 list_for_each_entry(client_node, &rbd_client_list, node) { 767 if (!ceph_compare_options(ceph_opts, client_node->client)) { 768 __rbd_get_client(client_node); 769 770 found = true; 771 break; 772 } 773 } 774 spin_unlock(&rbd_client_list_lock); 775 776 return found ? client_node : NULL; 777 } 778 779 /* 780 * (Per device) rbd map options 781 */ 782 enum { 783 Opt_queue_depth, 784 Opt_alloc_size, 785 Opt_lock_timeout, 786 /* int args above */ 787 Opt_pool_ns, 788 Opt_compression_hint, 789 /* string args above */ 790 Opt_read_only, 791 Opt_read_write, 792 Opt_lock_on_read, 793 Opt_exclusive, 794 Opt_notrim, 795 }; 796 797 enum { 798 Opt_compression_hint_none, 799 Opt_compression_hint_compressible, 800 Opt_compression_hint_incompressible, 801 }; 802 803 static const struct constant_table rbd_param_compression_hint[] = { 804 {"none", Opt_compression_hint_none}, 805 {"compressible", Opt_compression_hint_compressible}, 806 {"incompressible", Opt_compression_hint_incompressible}, 807 {} 808 }; 809 810 static const struct fs_parameter_spec rbd_parameters[] = { 811 fsparam_u32 ("alloc_size", Opt_alloc_size), 812 fsparam_enum ("compression_hint", Opt_compression_hint, 813 rbd_param_compression_hint), 814 fsparam_flag ("exclusive", Opt_exclusive), 815 fsparam_flag ("lock_on_read", Opt_lock_on_read), 816 fsparam_u32 ("lock_timeout", Opt_lock_timeout), 817 fsparam_flag ("notrim", Opt_notrim), 818 fsparam_string ("_pool_ns", Opt_pool_ns), 819 fsparam_u32 ("queue_depth", Opt_queue_depth), 820 fsparam_flag ("read_only", Opt_read_only), 821 fsparam_flag ("read_write", Opt_read_write), 822 fsparam_flag ("ro", Opt_read_only), 823 fsparam_flag ("rw", Opt_read_write), 824 {} 825 }; 826 827 struct rbd_options { 828 int queue_depth; 829 int alloc_size; 830 unsigned long lock_timeout; 831 bool read_only; 832 bool lock_on_read; 833 bool exclusive; 834 bool trim; 835 836 u32 alloc_hint_flags; /* CEPH_OSD_OP_ALLOC_HINT_FLAG_* */ 837 }; 838 839 #define RBD_QUEUE_DEPTH_DEFAULT BLKDEV_DEFAULT_RQ 840 #define RBD_ALLOC_SIZE_DEFAULT (64 * 1024) 841 #define RBD_LOCK_TIMEOUT_DEFAULT 0 /* no timeout */ 842 #define RBD_READ_ONLY_DEFAULT false 843 #define RBD_LOCK_ON_READ_DEFAULT false 844 #define RBD_EXCLUSIVE_DEFAULT false 845 #define RBD_TRIM_DEFAULT true 846 847 struct rbd_parse_opts_ctx { 848 struct rbd_spec *spec; 849 struct ceph_options *copts; 850 struct rbd_options *opts; 851 }; 852 853 static char* obj_op_name(enum obj_operation_type op_type) 854 { 855 switch (op_type) { 856 case OBJ_OP_READ: 857 return "read"; 858 case OBJ_OP_WRITE: 859 return "write"; 860 case OBJ_OP_DISCARD: 861 return "discard"; 862 case OBJ_OP_ZEROOUT: 863 return "zeroout"; 864 default: 865 return "???"; 866 } 867 } 868 869 /* 870 * Destroy ceph client 871 * 872 * Caller must hold rbd_client_list_lock. 873 */ 874 static void rbd_client_release(struct kref *kref) 875 { 876 struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref); 877 878 dout("%s: rbdc %p\n", __func__, rbdc); 879 spin_lock(&rbd_client_list_lock); 880 list_del(&rbdc->node); 881 spin_unlock(&rbd_client_list_lock); 882 883 ceph_destroy_client(rbdc->client); 884 kfree(rbdc); 885 } 886 887 /* 888 * Drop reference to ceph client node. If it's not referenced anymore, release 889 * it. 890 */ 891 static void rbd_put_client(struct rbd_client *rbdc) 892 { 893 if (rbdc) 894 kref_put(&rbdc->kref, rbd_client_release); 895 } 896 897 /* 898 * Get a ceph client with specific addr and configuration, if one does 899 * not exist create it. Either way, ceph_opts is consumed by this 900 * function. 901 */ 902 static struct rbd_client *rbd_get_client(struct ceph_options *ceph_opts) 903 { 904 struct rbd_client *rbdc; 905 int ret; 906 907 mutex_lock(&client_mutex); 908 rbdc = rbd_client_find(ceph_opts); 909 if (rbdc) { 910 ceph_destroy_options(ceph_opts); 911 912 /* 913 * Using an existing client. Make sure ->pg_pools is up to 914 * date before we look up the pool id in do_rbd_add(). 915 */ 916 ret = ceph_wait_for_latest_osdmap(rbdc->client, 917 rbdc->client->options->mount_timeout); 918 if (ret) { 919 rbd_warn(NULL, "failed to get latest osdmap: %d", ret); 920 rbd_put_client(rbdc); 921 rbdc = ERR_PTR(ret); 922 } 923 } else { 924 rbdc = rbd_client_create(ceph_opts); 925 } 926 mutex_unlock(&client_mutex); 927 928 return rbdc; 929 } 930 931 static bool rbd_image_format_valid(u32 image_format) 932 { 933 return image_format == 1 || image_format == 2; 934 } 935 936 static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk) 937 { 938 size_t size; 939 u32 snap_count; 940 941 /* The header has to start with the magic rbd header text */ 942 if (memcmp(&ondisk->text, RBD_HEADER_TEXT, sizeof (RBD_HEADER_TEXT))) 943 return false; 944 945 /* The bio layer requires at least sector-sized I/O */ 946 947 if (ondisk->options.order < SECTOR_SHIFT) 948 return false; 949 950 /* If we use u64 in a few spots we may be able to loosen this */ 951 952 if (ondisk->options.order > 8 * sizeof (int) - 1) 953 return false; 954 955 /* 956 * The size of a snapshot header has to fit in a size_t, and 957 * that limits the number of snapshots. 958 */ 959 snap_count = le32_to_cpu(ondisk->snap_count); 960 size = SIZE_MAX - sizeof (struct ceph_snap_context); 961 if (snap_count > size / sizeof (__le64)) 962 return false; 963 964 /* 965 * Not only that, but the size of the entire the snapshot 966 * header must also be representable in a size_t. 967 */ 968 size -= snap_count * sizeof (__le64); 969 if ((u64) size < le64_to_cpu(ondisk->snap_names_len)) 970 return false; 971 972 return true; 973 } 974 975 /* 976 * returns the size of an object in the image 977 */ 978 static u32 rbd_obj_bytes(struct rbd_image_header *header) 979 { 980 return 1U << header->obj_order; 981 } 982 983 static void rbd_init_layout(struct rbd_device *rbd_dev) 984 { 985 if (rbd_dev->header.stripe_unit == 0 || 986 rbd_dev->header.stripe_count == 0) { 987 rbd_dev->header.stripe_unit = rbd_obj_bytes(&rbd_dev->header); 988 rbd_dev->header.stripe_count = 1; 989 } 990 991 rbd_dev->layout.stripe_unit = rbd_dev->header.stripe_unit; 992 rbd_dev->layout.stripe_count = rbd_dev->header.stripe_count; 993 rbd_dev->layout.object_size = rbd_obj_bytes(&rbd_dev->header); 994 rbd_dev->layout.pool_id = rbd_dev->header.data_pool_id == CEPH_NOPOOL ? 995 rbd_dev->spec->pool_id : rbd_dev->header.data_pool_id; 996 RCU_INIT_POINTER(rbd_dev->layout.pool_ns, NULL); 997 } 998 999 /* 1000 * Fill an rbd image header with information from the given format 1 1001 * on-disk header. 1002 */ 1003 static int rbd_header_from_disk(struct rbd_device *rbd_dev, 1004 struct rbd_image_header_ondisk *ondisk) 1005 { 1006 struct rbd_image_header *header = &rbd_dev->header; 1007 bool first_time = header->object_prefix == NULL; 1008 struct ceph_snap_context *snapc; 1009 char *object_prefix = NULL; 1010 char *snap_names = NULL; 1011 u64 *snap_sizes = NULL; 1012 u32 snap_count; 1013 int ret = -ENOMEM; 1014 u32 i; 1015 1016 /* Allocate this now to avoid having to handle failure below */ 1017 1018 if (first_time) { 1019 object_prefix = kstrndup(ondisk->object_prefix, 1020 sizeof(ondisk->object_prefix), 1021 GFP_KERNEL); 1022 if (!object_prefix) 1023 return -ENOMEM; 1024 } 1025 1026 /* Allocate the snapshot context and fill it in */ 1027 1028 snap_count = le32_to_cpu(ondisk->snap_count); 1029 snapc = ceph_create_snap_context(snap_count, GFP_KERNEL); 1030 if (!snapc) 1031 goto out_err; 1032 snapc->seq = le64_to_cpu(ondisk->snap_seq); 1033 if (snap_count) { 1034 struct rbd_image_snap_ondisk *snaps; 1035 u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len); 1036 1037 /* We'll keep a copy of the snapshot names... */ 1038 1039 if (snap_names_len > (u64)SIZE_MAX) 1040 goto out_2big; 1041 snap_names = kmalloc(snap_names_len, GFP_KERNEL); 1042 if (!snap_names) 1043 goto out_err; 1044 1045 /* ...as well as the array of their sizes. */ 1046 snap_sizes = kmalloc_array(snap_count, 1047 sizeof(*header->snap_sizes), 1048 GFP_KERNEL); 1049 if (!snap_sizes) 1050 goto out_err; 1051 1052 /* 1053 * Copy the names, and fill in each snapshot's id 1054 * and size. 1055 * 1056 * Note that rbd_dev_v1_header_info() guarantees the 1057 * ondisk buffer we're working with has 1058 * snap_names_len bytes beyond the end of the 1059 * snapshot id array, this memcpy() is safe. 1060 */ 1061 memcpy(snap_names, &ondisk->snaps[snap_count], snap_names_len); 1062 snaps = ondisk->snaps; 1063 for (i = 0; i < snap_count; i++) { 1064 snapc->snaps[i] = le64_to_cpu(snaps[i].id); 1065 snap_sizes[i] = le64_to_cpu(snaps[i].image_size); 1066 } 1067 } 1068 1069 /* We won't fail any more, fill in the header */ 1070 1071 if (first_time) { 1072 header->object_prefix = object_prefix; 1073 header->obj_order = ondisk->options.order; 1074 rbd_init_layout(rbd_dev); 1075 } else { 1076 ceph_put_snap_context(header->snapc); 1077 kfree(header->snap_names); 1078 kfree(header->snap_sizes); 1079 } 1080 1081 /* The remaining fields always get updated (when we refresh) */ 1082 1083 header->image_size = le64_to_cpu(ondisk->image_size); 1084 header->snapc = snapc; 1085 header->snap_names = snap_names; 1086 header->snap_sizes = snap_sizes; 1087 1088 return 0; 1089 out_2big: 1090 ret = -EIO; 1091 out_err: 1092 kfree(snap_sizes); 1093 kfree(snap_names); 1094 ceph_put_snap_context(snapc); 1095 kfree(object_prefix); 1096 1097 return ret; 1098 } 1099 1100 static const char *_rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, u32 which) 1101 { 1102 const char *snap_name; 1103 1104 rbd_assert(which < rbd_dev->header.snapc->num_snaps); 1105 1106 /* Skip over names until we find the one we are looking for */ 1107 1108 snap_name = rbd_dev->header.snap_names; 1109 while (which--) 1110 snap_name += strlen(snap_name) + 1; 1111 1112 return kstrdup(snap_name, GFP_KERNEL); 1113 } 1114 1115 /* 1116 * Snapshot id comparison function for use with qsort()/bsearch(). 1117 * Note that result is for snapshots in *descending* order. 1118 */ 1119 static int snapid_compare_reverse(const void *s1, const void *s2) 1120 { 1121 u64 snap_id1 = *(u64 *)s1; 1122 u64 snap_id2 = *(u64 *)s2; 1123 1124 if (snap_id1 < snap_id2) 1125 return 1; 1126 return snap_id1 == snap_id2 ? 0 : -1; 1127 } 1128 1129 /* 1130 * Search a snapshot context to see if the given snapshot id is 1131 * present. 1132 * 1133 * Returns the position of the snapshot id in the array if it's found, 1134 * or BAD_SNAP_INDEX otherwise. 1135 * 1136 * Note: The snapshot array is in kept sorted (by the osd) in 1137 * reverse order, highest snapshot id first. 1138 */ 1139 static u32 rbd_dev_snap_index(struct rbd_device *rbd_dev, u64 snap_id) 1140 { 1141 struct ceph_snap_context *snapc = rbd_dev->header.snapc; 1142 u64 *found; 1143 1144 found = bsearch(&snap_id, &snapc->snaps, snapc->num_snaps, 1145 sizeof (snap_id), snapid_compare_reverse); 1146 1147 return found ? (u32)(found - &snapc->snaps[0]) : BAD_SNAP_INDEX; 1148 } 1149 1150 static const char *rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, 1151 u64 snap_id) 1152 { 1153 u32 which; 1154 const char *snap_name; 1155 1156 which = rbd_dev_snap_index(rbd_dev, snap_id); 1157 if (which == BAD_SNAP_INDEX) 1158 return ERR_PTR(-ENOENT); 1159 1160 snap_name = _rbd_dev_v1_snap_name(rbd_dev, which); 1161 return snap_name ? snap_name : ERR_PTR(-ENOMEM); 1162 } 1163 1164 static const char *rbd_snap_name(struct rbd_device *rbd_dev, u64 snap_id) 1165 { 1166 if (snap_id == CEPH_NOSNAP) 1167 return RBD_SNAP_HEAD_NAME; 1168 1169 rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); 1170 if (rbd_dev->image_format == 1) 1171 return rbd_dev_v1_snap_name(rbd_dev, snap_id); 1172 1173 return rbd_dev_v2_snap_name(rbd_dev, snap_id); 1174 } 1175 1176 static int rbd_snap_size(struct rbd_device *rbd_dev, u64 snap_id, 1177 u64 *snap_size) 1178 { 1179 rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); 1180 if (snap_id == CEPH_NOSNAP) { 1181 *snap_size = rbd_dev->header.image_size; 1182 } else if (rbd_dev->image_format == 1) { 1183 u32 which; 1184 1185 which = rbd_dev_snap_index(rbd_dev, snap_id); 1186 if (which == BAD_SNAP_INDEX) 1187 return -ENOENT; 1188 1189 *snap_size = rbd_dev->header.snap_sizes[which]; 1190 } else { 1191 u64 size = 0; 1192 int ret; 1193 1194 ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, NULL, &size); 1195 if (ret) 1196 return ret; 1197 1198 *snap_size = size; 1199 } 1200 return 0; 1201 } 1202 1203 static int rbd_dev_mapping_set(struct rbd_device *rbd_dev) 1204 { 1205 u64 snap_id = rbd_dev->spec->snap_id; 1206 u64 size = 0; 1207 int ret; 1208 1209 ret = rbd_snap_size(rbd_dev, snap_id, &size); 1210 if (ret) 1211 return ret; 1212 1213 rbd_dev->mapping.size = size; 1214 return 0; 1215 } 1216 1217 static void rbd_dev_mapping_clear(struct rbd_device *rbd_dev) 1218 { 1219 rbd_dev->mapping.size = 0; 1220 } 1221 1222 static void zero_bios(struct ceph_bio_iter *bio_pos, u32 off, u32 bytes) 1223 { 1224 struct ceph_bio_iter it = *bio_pos; 1225 1226 ceph_bio_iter_advance(&it, off); 1227 ceph_bio_iter_advance_step(&it, bytes, ({ 1228 memzero_bvec(&bv); 1229 })); 1230 } 1231 1232 static void zero_bvecs(struct ceph_bvec_iter *bvec_pos, u32 off, u32 bytes) 1233 { 1234 struct ceph_bvec_iter it = *bvec_pos; 1235 1236 ceph_bvec_iter_advance(&it, off); 1237 ceph_bvec_iter_advance_step(&it, bytes, ({ 1238 memzero_bvec(&bv); 1239 })); 1240 } 1241 1242 /* 1243 * Zero a range in @obj_req data buffer defined by a bio (list) or 1244 * (private) bio_vec array. 1245 * 1246 * @off is relative to the start of the data buffer. 1247 */ 1248 static void rbd_obj_zero_range(struct rbd_obj_request *obj_req, u32 off, 1249 u32 bytes) 1250 { 1251 dout("%s %p data buf %u~%u\n", __func__, obj_req, off, bytes); 1252 1253 switch (obj_req->img_request->data_type) { 1254 case OBJ_REQUEST_BIO: 1255 zero_bios(&obj_req->bio_pos, off, bytes); 1256 break; 1257 case OBJ_REQUEST_BVECS: 1258 case OBJ_REQUEST_OWN_BVECS: 1259 zero_bvecs(&obj_req->bvec_pos, off, bytes); 1260 break; 1261 default: 1262 BUG(); 1263 } 1264 } 1265 1266 static void rbd_obj_request_destroy(struct kref *kref); 1267 static void rbd_obj_request_put(struct rbd_obj_request *obj_request) 1268 { 1269 rbd_assert(obj_request != NULL); 1270 dout("%s: obj %p (was %d)\n", __func__, obj_request, 1271 kref_read(&obj_request->kref)); 1272 kref_put(&obj_request->kref, rbd_obj_request_destroy); 1273 } 1274 1275 static inline void rbd_img_obj_request_add(struct rbd_img_request *img_request, 1276 struct rbd_obj_request *obj_request) 1277 { 1278 rbd_assert(obj_request->img_request == NULL); 1279 1280 /* Image request now owns object's original reference */ 1281 obj_request->img_request = img_request; 1282 dout("%s: img %p obj %p\n", __func__, img_request, obj_request); 1283 } 1284 1285 static inline void rbd_img_obj_request_del(struct rbd_img_request *img_request, 1286 struct rbd_obj_request *obj_request) 1287 { 1288 dout("%s: img %p obj %p\n", __func__, img_request, obj_request); 1289 list_del(&obj_request->ex.oe_item); 1290 rbd_assert(obj_request->img_request == img_request); 1291 rbd_obj_request_put(obj_request); 1292 } 1293 1294 static void rbd_osd_submit(struct ceph_osd_request *osd_req) 1295 { 1296 struct rbd_obj_request *obj_req = osd_req->r_priv; 1297 1298 dout("%s osd_req %p for obj_req %p objno %llu %llu~%llu\n", 1299 __func__, osd_req, obj_req, obj_req->ex.oe_objno, 1300 obj_req->ex.oe_off, obj_req->ex.oe_len); 1301 ceph_osdc_start_request(osd_req->r_osdc, osd_req, false); 1302 } 1303 1304 /* 1305 * The default/initial value for all image request flags is 0. Each 1306 * is conditionally set to 1 at image request initialization time 1307 * and currently never change thereafter. 1308 */ 1309 static void img_request_layered_set(struct rbd_img_request *img_request) 1310 { 1311 set_bit(IMG_REQ_LAYERED, &img_request->flags); 1312 } 1313 1314 static bool img_request_layered_test(struct rbd_img_request *img_request) 1315 { 1316 return test_bit(IMG_REQ_LAYERED, &img_request->flags) != 0; 1317 } 1318 1319 static bool rbd_obj_is_entire(struct rbd_obj_request *obj_req) 1320 { 1321 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 1322 1323 return !obj_req->ex.oe_off && 1324 obj_req->ex.oe_len == rbd_dev->layout.object_size; 1325 } 1326 1327 static bool rbd_obj_is_tail(struct rbd_obj_request *obj_req) 1328 { 1329 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 1330 1331 return obj_req->ex.oe_off + obj_req->ex.oe_len == 1332 rbd_dev->layout.object_size; 1333 } 1334 1335 /* 1336 * Must be called after rbd_obj_calc_img_extents(). 1337 */ 1338 static bool rbd_obj_copyup_enabled(struct rbd_obj_request *obj_req) 1339 { 1340 if (!obj_req->num_img_extents || 1341 (rbd_obj_is_entire(obj_req) && 1342 !obj_req->img_request->snapc->num_snaps)) 1343 return false; 1344 1345 return true; 1346 } 1347 1348 static u64 rbd_obj_img_extents_bytes(struct rbd_obj_request *obj_req) 1349 { 1350 return ceph_file_extents_bytes(obj_req->img_extents, 1351 obj_req->num_img_extents); 1352 } 1353 1354 static bool rbd_img_is_write(struct rbd_img_request *img_req) 1355 { 1356 switch (img_req->op_type) { 1357 case OBJ_OP_READ: 1358 return false; 1359 case OBJ_OP_WRITE: 1360 case OBJ_OP_DISCARD: 1361 case OBJ_OP_ZEROOUT: 1362 return true; 1363 default: 1364 BUG(); 1365 } 1366 } 1367 1368 static void rbd_osd_req_callback(struct ceph_osd_request *osd_req) 1369 { 1370 struct rbd_obj_request *obj_req = osd_req->r_priv; 1371 int result; 1372 1373 dout("%s osd_req %p result %d for obj_req %p\n", __func__, osd_req, 1374 osd_req->r_result, obj_req); 1375 1376 /* 1377 * Writes aren't allowed to return a data payload. In some 1378 * guarded write cases (e.g. stat + zero on an empty object) 1379 * a stat response makes it through, but we don't care. 1380 */ 1381 if (osd_req->r_result > 0 && rbd_img_is_write(obj_req->img_request)) 1382 result = 0; 1383 else 1384 result = osd_req->r_result; 1385 1386 rbd_obj_handle_request(obj_req, result); 1387 } 1388 1389 static void rbd_osd_format_read(struct ceph_osd_request *osd_req) 1390 { 1391 struct rbd_obj_request *obj_request = osd_req->r_priv; 1392 struct rbd_device *rbd_dev = obj_request->img_request->rbd_dev; 1393 struct ceph_options *opt = rbd_dev->rbd_client->client->options; 1394 1395 osd_req->r_flags = CEPH_OSD_FLAG_READ | opt->read_from_replica; 1396 osd_req->r_snapid = obj_request->img_request->snap_id; 1397 } 1398 1399 static void rbd_osd_format_write(struct ceph_osd_request *osd_req) 1400 { 1401 struct rbd_obj_request *obj_request = osd_req->r_priv; 1402 1403 osd_req->r_flags = CEPH_OSD_FLAG_WRITE; 1404 ktime_get_real_ts64(&osd_req->r_mtime); 1405 osd_req->r_data_offset = obj_request->ex.oe_off; 1406 } 1407 1408 static struct ceph_osd_request * 1409 __rbd_obj_add_osd_request(struct rbd_obj_request *obj_req, 1410 struct ceph_snap_context *snapc, int num_ops) 1411 { 1412 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 1413 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 1414 struct ceph_osd_request *req; 1415 const char *name_format = rbd_dev->image_format == 1 ? 1416 RBD_V1_DATA_FORMAT : RBD_V2_DATA_FORMAT; 1417 int ret; 1418 1419 req = ceph_osdc_alloc_request(osdc, snapc, num_ops, false, GFP_NOIO); 1420 if (!req) 1421 return ERR_PTR(-ENOMEM); 1422 1423 list_add_tail(&req->r_private_item, &obj_req->osd_reqs); 1424 req->r_callback = rbd_osd_req_callback; 1425 req->r_priv = obj_req; 1426 1427 /* 1428 * Data objects may be stored in a separate pool, but always in 1429 * the same namespace in that pool as the header in its pool. 1430 */ 1431 ceph_oloc_copy(&req->r_base_oloc, &rbd_dev->header_oloc); 1432 req->r_base_oloc.pool = rbd_dev->layout.pool_id; 1433 1434 ret = ceph_oid_aprintf(&req->r_base_oid, GFP_NOIO, name_format, 1435 rbd_dev->header.object_prefix, 1436 obj_req->ex.oe_objno); 1437 if (ret) 1438 return ERR_PTR(ret); 1439 1440 return req; 1441 } 1442 1443 static struct ceph_osd_request * 1444 rbd_obj_add_osd_request(struct rbd_obj_request *obj_req, int num_ops) 1445 { 1446 return __rbd_obj_add_osd_request(obj_req, obj_req->img_request->snapc, 1447 num_ops); 1448 } 1449 1450 static struct rbd_obj_request *rbd_obj_request_create(void) 1451 { 1452 struct rbd_obj_request *obj_request; 1453 1454 obj_request = kmem_cache_zalloc(rbd_obj_request_cache, GFP_NOIO); 1455 if (!obj_request) 1456 return NULL; 1457 1458 ceph_object_extent_init(&obj_request->ex); 1459 INIT_LIST_HEAD(&obj_request->osd_reqs); 1460 mutex_init(&obj_request->state_mutex); 1461 kref_init(&obj_request->kref); 1462 1463 dout("%s %p\n", __func__, obj_request); 1464 return obj_request; 1465 } 1466 1467 static void rbd_obj_request_destroy(struct kref *kref) 1468 { 1469 struct rbd_obj_request *obj_request; 1470 struct ceph_osd_request *osd_req; 1471 u32 i; 1472 1473 obj_request = container_of(kref, struct rbd_obj_request, kref); 1474 1475 dout("%s: obj %p\n", __func__, obj_request); 1476 1477 while (!list_empty(&obj_request->osd_reqs)) { 1478 osd_req = list_first_entry(&obj_request->osd_reqs, 1479 struct ceph_osd_request, r_private_item); 1480 list_del_init(&osd_req->r_private_item); 1481 ceph_osdc_put_request(osd_req); 1482 } 1483 1484 switch (obj_request->img_request->data_type) { 1485 case OBJ_REQUEST_NODATA: 1486 case OBJ_REQUEST_BIO: 1487 case OBJ_REQUEST_BVECS: 1488 break; /* Nothing to do */ 1489 case OBJ_REQUEST_OWN_BVECS: 1490 kfree(obj_request->bvec_pos.bvecs); 1491 break; 1492 default: 1493 BUG(); 1494 } 1495 1496 kfree(obj_request->img_extents); 1497 if (obj_request->copyup_bvecs) { 1498 for (i = 0; i < obj_request->copyup_bvec_count; i++) { 1499 if (obj_request->copyup_bvecs[i].bv_page) 1500 __free_page(obj_request->copyup_bvecs[i].bv_page); 1501 } 1502 kfree(obj_request->copyup_bvecs); 1503 } 1504 1505 kmem_cache_free(rbd_obj_request_cache, obj_request); 1506 } 1507 1508 /* It's OK to call this for a device with no parent */ 1509 1510 static void rbd_spec_put(struct rbd_spec *spec); 1511 static void rbd_dev_unparent(struct rbd_device *rbd_dev) 1512 { 1513 rbd_dev_remove_parent(rbd_dev); 1514 rbd_spec_put(rbd_dev->parent_spec); 1515 rbd_dev->parent_spec = NULL; 1516 rbd_dev->parent_overlap = 0; 1517 } 1518 1519 /* 1520 * Parent image reference counting is used to determine when an 1521 * image's parent fields can be safely torn down--after there are no 1522 * more in-flight requests to the parent image. When the last 1523 * reference is dropped, cleaning them up is safe. 1524 */ 1525 static void rbd_dev_parent_put(struct rbd_device *rbd_dev) 1526 { 1527 int counter; 1528 1529 if (!rbd_dev->parent_spec) 1530 return; 1531 1532 counter = atomic_dec_return_safe(&rbd_dev->parent_ref); 1533 if (counter > 0) 1534 return; 1535 1536 /* Last reference; clean up parent data structures */ 1537 1538 if (!counter) 1539 rbd_dev_unparent(rbd_dev); 1540 else 1541 rbd_warn(rbd_dev, "parent reference underflow"); 1542 } 1543 1544 /* 1545 * If an image has a non-zero parent overlap, get a reference to its 1546 * parent. 1547 * 1548 * Returns true if the rbd device has a parent with a non-zero 1549 * overlap and a reference for it was successfully taken, or 1550 * false otherwise. 1551 */ 1552 static bool rbd_dev_parent_get(struct rbd_device *rbd_dev) 1553 { 1554 int counter = 0; 1555 1556 if (!rbd_dev->parent_spec) 1557 return false; 1558 1559 if (rbd_dev->parent_overlap) 1560 counter = atomic_inc_return_safe(&rbd_dev->parent_ref); 1561 1562 if (counter < 0) 1563 rbd_warn(rbd_dev, "parent reference overflow"); 1564 1565 return counter > 0; 1566 } 1567 1568 static void rbd_img_request_init(struct rbd_img_request *img_request, 1569 struct rbd_device *rbd_dev, 1570 enum obj_operation_type op_type) 1571 { 1572 memset(img_request, 0, sizeof(*img_request)); 1573 1574 img_request->rbd_dev = rbd_dev; 1575 img_request->op_type = op_type; 1576 1577 INIT_LIST_HEAD(&img_request->lock_item); 1578 INIT_LIST_HEAD(&img_request->object_extents); 1579 mutex_init(&img_request->state_mutex); 1580 } 1581 1582 static void rbd_img_capture_header(struct rbd_img_request *img_req) 1583 { 1584 struct rbd_device *rbd_dev = img_req->rbd_dev; 1585 1586 lockdep_assert_held(&rbd_dev->header_rwsem); 1587 1588 if (rbd_img_is_write(img_req)) 1589 img_req->snapc = ceph_get_snap_context(rbd_dev->header.snapc); 1590 else 1591 img_req->snap_id = rbd_dev->spec->snap_id; 1592 1593 if (rbd_dev_parent_get(rbd_dev)) 1594 img_request_layered_set(img_req); 1595 } 1596 1597 static void rbd_img_request_destroy(struct rbd_img_request *img_request) 1598 { 1599 struct rbd_obj_request *obj_request; 1600 struct rbd_obj_request *next_obj_request; 1601 1602 dout("%s: img %p\n", __func__, img_request); 1603 1604 WARN_ON(!list_empty(&img_request->lock_item)); 1605 for_each_obj_request_safe(img_request, obj_request, next_obj_request) 1606 rbd_img_obj_request_del(img_request, obj_request); 1607 1608 if (img_request_layered_test(img_request)) 1609 rbd_dev_parent_put(img_request->rbd_dev); 1610 1611 if (rbd_img_is_write(img_request)) 1612 ceph_put_snap_context(img_request->snapc); 1613 1614 if (test_bit(IMG_REQ_CHILD, &img_request->flags)) 1615 kmem_cache_free(rbd_img_request_cache, img_request); 1616 } 1617 1618 #define BITS_PER_OBJ 2 1619 #define OBJS_PER_BYTE (BITS_PER_BYTE / BITS_PER_OBJ) 1620 #define OBJ_MASK ((1 << BITS_PER_OBJ) - 1) 1621 1622 static void __rbd_object_map_index(struct rbd_device *rbd_dev, u64 objno, 1623 u64 *index, u8 *shift) 1624 { 1625 u32 off; 1626 1627 rbd_assert(objno < rbd_dev->object_map_size); 1628 *index = div_u64_rem(objno, OBJS_PER_BYTE, &off); 1629 *shift = (OBJS_PER_BYTE - off - 1) * BITS_PER_OBJ; 1630 } 1631 1632 static u8 __rbd_object_map_get(struct rbd_device *rbd_dev, u64 objno) 1633 { 1634 u64 index; 1635 u8 shift; 1636 1637 lockdep_assert_held(&rbd_dev->object_map_lock); 1638 __rbd_object_map_index(rbd_dev, objno, &index, &shift); 1639 return (rbd_dev->object_map[index] >> shift) & OBJ_MASK; 1640 } 1641 1642 static void __rbd_object_map_set(struct rbd_device *rbd_dev, u64 objno, u8 val) 1643 { 1644 u64 index; 1645 u8 shift; 1646 u8 *p; 1647 1648 lockdep_assert_held(&rbd_dev->object_map_lock); 1649 rbd_assert(!(val & ~OBJ_MASK)); 1650 1651 __rbd_object_map_index(rbd_dev, objno, &index, &shift); 1652 p = &rbd_dev->object_map[index]; 1653 *p = (*p & ~(OBJ_MASK << shift)) | (val << shift); 1654 } 1655 1656 static u8 rbd_object_map_get(struct rbd_device *rbd_dev, u64 objno) 1657 { 1658 u8 state; 1659 1660 spin_lock(&rbd_dev->object_map_lock); 1661 state = __rbd_object_map_get(rbd_dev, objno); 1662 spin_unlock(&rbd_dev->object_map_lock); 1663 return state; 1664 } 1665 1666 static bool use_object_map(struct rbd_device *rbd_dev) 1667 { 1668 /* 1669 * An image mapped read-only can't use the object map -- it isn't 1670 * loaded because the header lock isn't acquired. Someone else can 1671 * write to the image and update the object map behind our back. 1672 * 1673 * A snapshot can't be written to, so using the object map is always 1674 * safe. 1675 */ 1676 if (!rbd_is_snap(rbd_dev) && rbd_is_ro(rbd_dev)) 1677 return false; 1678 1679 return ((rbd_dev->header.features & RBD_FEATURE_OBJECT_MAP) && 1680 !(rbd_dev->object_map_flags & RBD_FLAG_OBJECT_MAP_INVALID)); 1681 } 1682 1683 static bool rbd_object_map_may_exist(struct rbd_device *rbd_dev, u64 objno) 1684 { 1685 u8 state; 1686 1687 /* fall back to default logic if object map is disabled or invalid */ 1688 if (!use_object_map(rbd_dev)) 1689 return true; 1690 1691 state = rbd_object_map_get(rbd_dev, objno); 1692 return state != OBJECT_NONEXISTENT; 1693 } 1694 1695 static void rbd_object_map_name(struct rbd_device *rbd_dev, u64 snap_id, 1696 struct ceph_object_id *oid) 1697 { 1698 if (snap_id == CEPH_NOSNAP) 1699 ceph_oid_printf(oid, "%s%s", RBD_OBJECT_MAP_PREFIX, 1700 rbd_dev->spec->image_id); 1701 else 1702 ceph_oid_printf(oid, "%s%s.%016llx", RBD_OBJECT_MAP_PREFIX, 1703 rbd_dev->spec->image_id, snap_id); 1704 } 1705 1706 static int rbd_object_map_lock(struct rbd_device *rbd_dev) 1707 { 1708 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 1709 CEPH_DEFINE_OID_ONSTACK(oid); 1710 u8 lock_type; 1711 char *lock_tag; 1712 struct ceph_locker *lockers; 1713 u32 num_lockers; 1714 bool broke_lock = false; 1715 int ret; 1716 1717 rbd_object_map_name(rbd_dev, CEPH_NOSNAP, &oid); 1718 1719 again: 1720 ret = ceph_cls_lock(osdc, &oid, &rbd_dev->header_oloc, RBD_LOCK_NAME, 1721 CEPH_CLS_LOCK_EXCLUSIVE, "", "", "", 0); 1722 if (ret != -EBUSY || broke_lock) { 1723 if (ret == -EEXIST) 1724 ret = 0; /* already locked by myself */ 1725 if (ret) 1726 rbd_warn(rbd_dev, "failed to lock object map: %d", ret); 1727 return ret; 1728 } 1729 1730 ret = ceph_cls_lock_info(osdc, &oid, &rbd_dev->header_oloc, 1731 RBD_LOCK_NAME, &lock_type, &lock_tag, 1732 &lockers, &num_lockers); 1733 if (ret) { 1734 if (ret == -ENOENT) 1735 goto again; 1736 1737 rbd_warn(rbd_dev, "failed to get object map lockers: %d", ret); 1738 return ret; 1739 } 1740 1741 kfree(lock_tag); 1742 if (num_lockers == 0) 1743 goto again; 1744 1745 rbd_warn(rbd_dev, "breaking object map lock owned by %s%llu", 1746 ENTITY_NAME(lockers[0].id.name)); 1747 1748 ret = ceph_cls_break_lock(osdc, &oid, &rbd_dev->header_oloc, 1749 RBD_LOCK_NAME, lockers[0].id.cookie, 1750 &lockers[0].id.name); 1751 ceph_free_lockers(lockers, num_lockers); 1752 if (ret) { 1753 if (ret == -ENOENT) 1754 goto again; 1755 1756 rbd_warn(rbd_dev, "failed to break object map lock: %d", ret); 1757 return ret; 1758 } 1759 1760 broke_lock = true; 1761 goto again; 1762 } 1763 1764 static void rbd_object_map_unlock(struct rbd_device *rbd_dev) 1765 { 1766 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 1767 CEPH_DEFINE_OID_ONSTACK(oid); 1768 int ret; 1769 1770 rbd_object_map_name(rbd_dev, CEPH_NOSNAP, &oid); 1771 1772 ret = ceph_cls_unlock(osdc, &oid, &rbd_dev->header_oloc, RBD_LOCK_NAME, 1773 ""); 1774 if (ret && ret != -ENOENT) 1775 rbd_warn(rbd_dev, "failed to unlock object map: %d", ret); 1776 } 1777 1778 static int decode_object_map_header(void **p, void *end, u64 *object_map_size) 1779 { 1780 u8 struct_v; 1781 u32 struct_len; 1782 u32 header_len; 1783 void *header_end; 1784 int ret; 1785 1786 ceph_decode_32_safe(p, end, header_len, e_inval); 1787 header_end = *p + header_len; 1788 1789 ret = ceph_start_decoding(p, end, 1, "BitVector header", &struct_v, 1790 &struct_len); 1791 if (ret) 1792 return ret; 1793 1794 ceph_decode_64_safe(p, end, *object_map_size, e_inval); 1795 1796 *p = header_end; 1797 return 0; 1798 1799 e_inval: 1800 return -EINVAL; 1801 } 1802 1803 static int __rbd_object_map_load(struct rbd_device *rbd_dev) 1804 { 1805 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 1806 CEPH_DEFINE_OID_ONSTACK(oid); 1807 struct page **pages; 1808 void *p, *end; 1809 size_t reply_len; 1810 u64 num_objects; 1811 u64 object_map_bytes; 1812 u64 object_map_size; 1813 int num_pages; 1814 int ret; 1815 1816 rbd_assert(!rbd_dev->object_map && !rbd_dev->object_map_size); 1817 1818 num_objects = ceph_get_num_objects(&rbd_dev->layout, 1819 rbd_dev->mapping.size); 1820 object_map_bytes = DIV_ROUND_UP_ULL(num_objects * BITS_PER_OBJ, 1821 BITS_PER_BYTE); 1822 num_pages = calc_pages_for(0, object_map_bytes) + 1; 1823 pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); 1824 if (IS_ERR(pages)) 1825 return PTR_ERR(pages); 1826 1827 reply_len = num_pages * PAGE_SIZE; 1828 rbd_object_map_name(rbd_dev, rbd_dev->spec->snap_id, &oid); 1829 ret = ceph_osdc_call(osdc, &oid, &rbd_dev->header_oloc, 1830 "rbd", "object_map_load", CEPH_OSD_FLAG_READ, 1831 NULL, 0, pages, &reply_len); 1832 if (ret) 1833 goto out; 1834 1835 p = page_address(pages[0]); 1836 end = p + min(reply_len, (size_t)PAGE_SIZE); 1837 ret = decode_object_map_header(&p, end, &object_map_size); 1838 if (ret) 1839 goto out; 1840 1841 if (object_map_size != num_objects) { 1842 rbd_warn(rbd_dev, "object map size mismatch: %llu vs %llu", 1843 object_map_size, num_objects); 1844 ret = -EINVAL; 1845 goto out; 1846 } 1847 1848 if (offset_in_page(p) + object_map_bytes > reply_len) { 1849 ret = -EINVAL; 1850 goto out; 1851 } 1852 1853 rbd_dev->object_map = kvmalloc(object_map_bytes, GFP_KERNEL); 1854 if (!rbd_dev->object_map) { 1855 ret = -ENOMEM; 1856 goto out; 1857 } 1858 1859 rbd_dev->object_map_size = object_map_size; 1860 ceph_copy_from_page_vector(pages, rbd_dev->object_map, 1861 offset_in_page(p), object_map_bytes); 1862 1863 out: 1864 ceph_release_page_vector(pages, num_pages); 1865 return ret; 1866 } 1867 1868 static void rbd_object_map_free(struct rbd_device *rbd_dev) 1869 { 1870 kvfree(rbd_dev->object_map); 1871 rbd_dev->object_map = NULL; 1872 rbd_dev->object_map_size = 0; 1873 } 1874 1875 static int rbd_object_map_load(struct rbd_device *rbd_dev) 1876 { 1877 int ret; 1878 1879 ret = __rbd_object_map_load(rbd_dev); 1880 if (ret) 1881 return ret; 1882 1883 ret = rbd_dev_v2_get_flags(rbd_dev); 1884 if (ret) { 1885 rbd_object_map_free(rbd_dev); 1886 return ret; 1887 } 1888 1889 if (rbd_dev->object_map_flags & RBD_FLAG_OBJECT_MAP_INVALID) 1890 rbd_warn(rbd_dev, "object map is invalid"); 1891 1892 return 0; 1893 } 1894 1895 static int rbd_object_map_open(struct rbd_device *rbd_dev) 1896 { 1897 int ret; 1898 1899 ret = rbd_object_map_lock(rbd_dev); 1900 if (ret) 1901 return ret; 1902 1903 ret = rbd_object_map_load(rbd_dev); 1904 if (ret) { 1905 rbd_object_map_unlock(rbd_dev); 1906 return ret; 1907 } 1908 1909 return 0; 1910 } 1911 1912 static void rbd_object_map_close(struct rbd_device *rbd_dev) 1913 { 1914 rbd_object_map_free(rbd_dev); 1915 rbd_object_map_unlock(rbd_dev); 1916 } 1917 1918 /* 1919 * This function needs snap_id (or more precisely just something to 1920 * distinguish between HEAD and snapshot object maps), new_state and 1921 * current_state that were passed to rbd_object_map_update(). 1922 * 1923 * To avoid allocating and stashing a context we piggyback on the OSD 1924 * request. A HEAD update has two ops (assert_locked). For new_state 1925 * and current_state we decode our own object_map_update op, encoded in 1926 * rbd_cls_object_map_update(). 1927 */ 1928 static int rbd_object_map_update_finish(struct rbd_obj_request *obj_req, 1929 struct ceph_osd_request *osd_req) 1930 { 1931 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 1932 struct ceph_osd_data *osd_data; 1933 u64 objno; 1934 u8 state, new_state, current_state; 1935 bool has_current_state; 1936 void *p; 1937 1938 if (osd_req->r_result) 1939 return osd_req->r_result; 1940 1941 /* 1942 * Nothing to do for a snapshot object map. 1943 */ 1944 if (osd_req->r_num_ops == 1) 1945 return 0; 1946 1947 /* 1948 * Update in-memory HEAD object map. 1949 */ 1950 rbd_assert(osd_req->r_num_ops == 2); 1951 osd_data = osd_req_op_data(osd_req, 1, cls, request_data); 1952 rbd_assert(osd_data->type == CEPH_OSD_DATA_TYPE_PAGES); 1953 1954 p = page_address(osd_data->pages[0]); 1955 objno = ceph_decode_64(&p); 1956 rbd_assert(objno == obj_req->ex.oe_objno); 1957 rbd_assert(ceph_decode_64(&p) == objno + 1); 1958 new_state = ceph_decode_8(&p); 1959 has_current_state = ceph_decode_8(&p); 1960 if (has_current_state) 1961 current_state = ceph_decode_8(&p); 1962 1963 spin_lock(&rbd_dev->object_map_lock); 1964 state = __rbd_object_map_get(rbd_dev, objno); 1965 if (!has_current_state || current_state == state || 1966 (current_state == OBJECT_EXISTS && state == OBJECT_EXISTS_CLEAN)) 1967 __rbd_object_map_set(rbd_dev, objno, new_state); 1968 spin_unlock(&rbd_dev->object_map_lock); 1969 1970 return 0; 1971 } 1972 1973 static void rbd_object_map_callback(struct ceph_osd_request *osd_req) 1974 { 1975 struct rbd_obj_request *obj_req = osd_req->r_priv; 1976 int result; 1977 1978 dout("%s osd_req %p result %d for obj_req %p\n", __func__, osd_req, 1979 osd_req->r_result, obj_req); 1980 1981 result = rbd_object_map_update_finish(obj_req, osd_req); 1982 rbd_obj_handle_request(obj_req, result); 1983 } 1984 1985 static bool update_needed(struct rbd_device *rbd_dev, u64 objno, u8 new_state) 1986 { 1987 u8 state = rbd_object_map_get(rbd_dev, objno); 1988 1989 if (state == new_state || 1990 (new_state == OBJECT_PENDING && state == OBJECT_NONEXISTENT) || 1991 (new_state == OBJECT_NONEXISTENT && state != OBJECT_PENDING)) 1992 return false; 1993 1994 return true; 1995 } 1996 1997 static int rbd_cls_object_map_update(struct ceph_osd_request *req, 1998 int which, u64 objno, u8 new_state, 1999 const u8 *current_state) 2000 { 2001 struct page **pages; 2002 void *p, *start; 2003 int ret; 2004 2005 ret = osd_req_op_cls_init(req, which, "rbd", "object_map_update"); 2006 if (ret) 2007 return ret; 2008 2009 pages = ceph_alloc_page_vector(1, GFP_NOIO); 2010 if (IS_ERR(pages)) 2011 return PTR_ERR(pages); 2012 2013 p = start = page_address(pages[0]); 2014 ceph_encode_64(&p, objno); 2015 ceph_encode_64(&p, objno + 1); 2016 ceph_encode_8(&p, new_state); 2017 if (current_state) { 2018 ceph_encode_8(&p, 1); 2019 ceph_encode_8(&p, *current_state); 2020 } else { 2021 ceph_encode_8(&p, 0); 2022 } 2023 2024 osd_req_op_cls_request_data_pages(req, which, pages, p - start, 0, 2025 false, true); 2026 return 0; 2027 } 2028 2029 /* 2030 * Return: 2031 * 0 - object map update sent 2032 * 1 - object map update isn't needed 2033 * <0 - error 2034 */ 2035 static int rbd_object_map_update(struct rbd_obj_request *obj_req, u64 snap_id, 2036 u8 new_state, const u8 *current_state) 2037 { 2038 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 2039 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 2040 struct ceph_osd_request *req; 2041 int num_ops = 1; 2042 int which = 0; 2043 int ret; 2044 2045 if (snap_id == CEPH_NOSNAP) { 2046 if (!update_needed(rbd_dev, obj_req->ex.oe_objno, new_state)) 2047 return 1; 2048 2049 num_ops++; /* assert_locked */ 2050 } 2051 2052 req = ceph_osdc_alloc_request(osdc, NULL, num_ops, false, GFP_NOIO); 2053 if (!req) 2054 return -ENOMEM; 2055 2056 list_add_tail(&req->r_private_item, &obj_req->osd_reqs); 2057 req->r_callback = rbd_object_map_callback; 2058 req->r_priv = obj_req; 2059 2060 rbd_object_map_name(rbd_dev, snap_id, &req->r_base_oid); 2061 ceph_oloc_copy(&req->r_base_oloc, &rbd_dev->header_oloc); 2062 req->r_flags = CEPH_OSD_FLAG_WRITE; 2063 ktime_get_real_ts64(&req->r_mtime); 2064 2065 if (snap_id == CEPH_NOSNAP) { 2066 /* 2067 * Protect against possible race conditions during lock 2068 * ownership transitions. 2069 */ 2070 ret = ceph_cls_assert_locked(req, which++, RBD_LOCK_NAME, 2071 CEPH_CLS_LOCK_EXCLUSIVE, "", ""); 2072 if (ret) 2073 return ret; 2074 } 2075 2076 ret = rbd_cls_object_map_update(req, which, obj_req->ex.oe_objno, 2077 new_state, current_state); 2078 if (ret) 2079 return ret; 2080 2081 ret = ceph_osdc_alloc_messages(req, GFP_NOIO); 2082 if (ret) 2083 return ret; 2084 2085 ceph_osdc_start_request(osdc, req, false); 2086 return 0; 2087 } 2088 2089 static void prune_extents(struct ceph_file_extent *img_extents, 2090 u32 *num_img_extents, u64 overlap) 2091 { 2092 u32 cnt = *num_img_extents; 2093 2094 /* drop extents completely beyond the overlap */ 2095 while (cnt && img_extents[cnt - 1].fe_off >= overlap) 2096 cnt--; 2097 2098 if (cnt) { 2099 struct ceph_file_extent *ex = &img_extents[cnt - 1]; 2100 2101 /* trim final overlapping extent */ 2102 if (ex->fe_off + ex->fe_len > overlap) 2103 ex->fe_len = overlap - ex->fe_off; 2104 } 2105 2106 *num_img_extents = cnt; 2107 } 2108 2109 /* 2110 * Determine the byte range(s) covered by either just the object extent 2111 * or the entire object in the parent image. 2112 */ 2113 static int rbd_obj_calc_img_extents(struct rbd_obj_request *obj_req, 2114 bool entire) 2115 { 2116 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 2117 int ret; 2118 2119 if (!rbd_dev->parent_overlap) 2120 return 0; 2121 2122 ret = ceph_extent_to_file(&rbd_dev->layout, obj_req->ex.oe_objno, 2123 entire ? 0 : obj_req->ex.oe_off, 2124 entire ? rbd_dev->layout.object_size : 2125 obj_req->ex.oe_len, 2126 &obj_req->img_extents, 2127 &obj_req->num_img_extents); 2128 if (ret) 2129 return ret; 2130 2131 prune_extents(obj_req->img_extents, &obj_req->num_img_extents, 2132 rbd_dev->parent_overlap); 2133 return 0; 2134 } 2135 2136 static void rbd_osd_setup_data(struct ceph_osd_request *osd_req, int which) 2137 { 2138 struct rbd_obj_request *obj_req = osd_req->r_priv; 2139 2140 switch (obj_req->img_request->data_type) { 2141 case OBJ_REQUEST_BIO: 2142 osd_req_op_extent_osd_data_bio(osd_req, which, 2143 &obj_req->bio_pos, 2144 obj_req->ex.oe_len); 2145 break; 2146 case OBJ_REQUEST_BVECS: 2147 case OBJ_REQUEST_OWN_BVECS: 2148 rbd_assert(obj_req->bvec_pos.iter.bi_size == 2149 obj_req->ex.oe_len); 2150 rbd_assert(obj_req->bvec_idx == obj_req->bvec_count); 2151 osd_req_op_extent_osd_data_bvec_pos(osd_req, which, 2152 &obj_req->bvec_pos); 2153 break; 2154 default: 2155 BUG(); 2156 } 2157 } 2158 2159 static int rbd_osd_setup_stat(struct ceph_osd_request *osd_req, int which) 2160 { 2161 struct page **pages; 2162 2163 /* 2164 * The response data for a STAT call consists of: 2165 * le64 length; 2166 * struct { 2167 * le32 tv_sec; 2168 * le32 tv_nsec; 2169 * } mtime; 2170 */ 2171 pages = ceph_alloc_page_vector(1, GFP_NOIO); 2172 if (IS_ERR(pages)) 2173 return PTR_ERR(pages); 2174 2175 osd_req_op_init(osd_req, which, CEPH_OSD_OP_STAT, 0); 2176 osd_req_op_raw_data_in_pages(osd_req, which, pages, 2177 8 + sizeof(struct ceph_timespec), 2178 0, false, true); 2179 return 0; 2180 } 2181 2182 static int rbd_osd_setup_copyup(struct ceph_osd_request *osd_req, int which, 2183 u32 bytes) 2184 { 2185 struct rbd_obj_request *obj_req = osd_req->r_priv; 2186 int ret; 2187 2188 ret = osd_req_op_cls_init(osd_req, which, "rbd", "copyup"); 2189 if (ret) 2190 return ret; 2191 2192 osd_req_op_cls_request_data_bvecs(osd_req, which, obj_req->copyup_bvecs, 2193 obj_req->copyup_bvec_count, bytes); 2194 return 0; 2195 } 2196 2197 static int rbd_obj_init_read(struct rbd_obj_request *obj_req) 2198 { 2199 obj_req->read_state = RBD_OBJ_READ_START; 2200 return 0; 2201 } 2202 2203 static void __rbd_osd_setup_write_ops(struct ceph_osd_request *osd_req, 2204 int which) 2205 { 2206 struct rbd_obj_request *obj_req = osd_req->r_priv; 2207 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 2208 u16 opcode; 2209 2210 if (!use_object_map(rbd_dev) || 2211 !(obj_req->flags & RBD_OBJ_FLAG_MAY_EXIST)) { 2212 osd_req_op_alloc_hint_init(osd_req, which++, 2213 rbd_dev->layout.object_size, 2214 rbd_dev->layout.object_size, 2215 rbd_dev->opts->alloc_hint_flags); 2216 } 2217 2218 if (rbd_obj_is_entire(obj_req)) 2219 opcode = CEPH_OSD_OP_WRITEFULL; 2220 else 2221 opcode = CEPH_OSD_OP_WRITE; 2222 2223 osd_req_op_extent_init(osd_req, which, opcode, 2224 obj_req->ex.oe_off, obj_req->ex.oe_len, 0, 0); 2225 rbd_osd_setup_data(osd_req, which); 2226 } 2227 2228 static int rbd_obj_init_write(struct rbd_obj_request *obj_req) 2229 { 2230 int ret; 2231 2232 /* reverse map the entire object onto the parent */ 2233 ret = rbd_obj_calc_img_extents(obj_req, true); 2234 if (ret) 2235 return ret; 2236 2237 if (rbd_obj_copyup_enabled(obj_req)) 2238 obj_req->flags |= RBD_OBJ_FLAG_COPYUP_ENABLED; 2239 2240 obj_req->write_state = RBD_OBJ_WRITE_START; 2241 return 0; 2242 } 2243 2244 static u16 truncate_or_zero_opcode(struct rbd_obj_request *obj_req) 2245 { 2246 return rbd_obj_is_tail(obj_req) ? CEPH_OSD_OP_TRUNCATE : 2247 CEPH_OSD_OP_ZERO; 2248 } 2249 2250 static void __rbd_osd_setup_discard_ops(struct ceph_osd_request *osd_req, 2251 int which) 2252 { 2253 struct rbd_obj_request *obj_req = osd_req->r_priv; 2254 2255 if (rbd_obj_is_entire(obj_req) && !obj_req->num_img_extents) { 2256 rbd_assert(obj_req->flags & RBD_OBJ_FLAG_DELETION); 2257 osd_req_op_init(osd_req, which, CEPH_OSD_OP_DELETE, 0); 2258 } else { 2259 osd_req_op_extent_init(osd_req, which, 2260 truncate_or_zero_opcode(obj_req), 2261 obj_req->ex.oe_off, obj_req->ex.oe_len, 2262 0, 0); 2263 } 2264 } 2265 2266 static int rbd_obj_init_discard(struct rbd_obj_request *obj_req) 2267 { 2268 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 2269 u64 off, next_off; 2270 int ret; 2271 2272 /* 2273 * Align the range to alloc_size boundary and punt on discards 2274 * that are too small to free up any space. 2275 * 2276 * alloc_size == object_size && is_tail() is a special case for 2277 * filestore with filestore_punch_hole = false, needed to allow 2278 * truncate (in addition to delete). 2279 */ 2280 if (rbd_dev->opts->alloc_size != rbd_dev->layout.object_size || 2281 !rbd_obj_is_tail(obj_req)) { 2282 off = round_up(obj_req->ex.oe_off, rbd_dev->opts->alloc_size); 2283 next_off = round_down(obj_req->ex.oe_off + obj_req->ex.oe_len, 2284 rbd_dev->opts->alloc_size); 2285 if (off >= next_off) 2286 return 1; 2287 2288 dout("%s %p %llu~%llu -> %llu~%llu\n", __func__, 2289 obj_req, obj_req->ex.oe_off, obj_req->ex.oe_len, 2290 off, next_off - off); 2291 obj_req->ex.oe_off = off; 2292 obj_req->ex.oe_len = next_off - off; 2293 } 2294 2295 /* reverse map the entire object onto the parent */ 2296 ret = rbd_obj_calc_img_extents(obj_req, true); 2297 if (ret) 2298 return ret; 2299 2300 obj_req->flags |= RBD_OBJ_FLAG_NOOP_FOR_NONEXISTENT; 2301 if (rbd_obj_is_entire(obj_req) && !obj_req->num_img_extents) 2302 obj_req->flags |= RBD_OBJ_FLAG_DELETION; 2303 2304 obj_req->write_state = RBD_OBJ_WRITE_START; 2305 return 0; 2306 } 2307 2308 static void __rbd_osd_setup_zeroout_ops(struct ceph_osd_request *osd_req, 2309 int which) 2310 { 2311 struct rbd_obj_request *obj_req = osd_req->r_priv; 2312 u16 opcode; 2313 2314 if (rbd_obj_is_entire(obj_req)) { 2315 if (obj_req->num_img_extents) { 2316 if (!(obj_req->flags & RBD_OBJ_FLAG_COPYUP_ENABLED)) 2317 osd_req_op_init(osd_req, which++, 2318 CEPH_OSD_OP_CREATE, 0); 2319 opcode = CEPH_OSD_OP_TRUNCATE; 2320 } else { 2321 rbd_assert(obj_req->flags & RBD_OBJ_FLAG_DELETION); 2322 osd_req_op_init(osd_req, which++, 2323 CEPH_OSD_OP_DELETE, 0); 2324 opcode = 0; 2325 } 2326 } else { 2327 opcode = truncate_or_zero_opcode(obj_req); 2328 } 2329 2330 if (opcode) 2331 osd_req_op_extent_init(osd_req, which, opcode, 2332 obj_req->ex.oe_off, obj_req->ex.oe_len, 2333 0, 0); 2334 } 2335 2336 static int rbd_obj_init_zeroout(struct rbd_obj_request *obj_req) 2337 { 2338 int ret; 2339 2340 /* reverse map the entire object onto the parent */ 2341 ret = rbd_obj_calc_img_extents(obj_req, true); 2342 if (ret) 2343 return ret; 2344 2345 if (rbd_obj_copyup_enabled(obj_req)) 2346 obj_req->flags |= RBD_OBJ_FLAG_COPYUP_ENABLED; 2347 if (!obj_req->num_img_extents) { 2348 obj_req->flags |= RBD_OBJ_FLAG_NOOP_FOR_NONEXISTENT; 2349 if (rbd_obj_is_entire(obj_req)) 2350 obj_req->flags |= RBD_OBJ_FLAG_DELETION; 2351 } 2352 2353 obj_req->write_state = RBD_OBJ_WRITE_START; 2354 return 0; 2355 } 2356 2357 static int count_write_ops(struct rbd_obj_request *obj_req) 2358 { 2359 struct rbd_img_request *img_req = obj_req->img_request; 2360 2361 switch (img_req->op_type) { 2362 case OBJ_OP_WRITE: 2363 if (!use_object_map(img_req->rbd_dev) || 2364 !(obj_req->flags & RBD_OBJ_FLAG_MAY_EXIST)) 2365 return 2; /* setallochint + write/writefull */ 2366 2367 return 1; /* write/writefull */ 2368 case OBJ_OP_DISCARD: 2369 return 1; /* delete/truncate/zero */ 2370 case OBJ_OP_ZEROOUT: 2371 if (rbd_obj_is_entire(obj_req) && obj_req->num_img_extents && 2372 !(obj_req->flags & RBD_OBJ_FLAG_COPYUP_ENABLED)) 2373 return 2; /* create + truncate */ 2374 2375 return 1; /* delete/truncate/zero */ 2376 default: 2377 BUG(); 2378 } 2379 } 2380 2381 static void rbd_osd_setup_write_ops(struct ceph_osd_request *osd_req, 2382 int which) 2383 { 2384 struct rbd_obj_request *obj_req = osd_req->r_priv; 2385 2386 switch (obj_req->img_request->op_type) { 2387 case OBJ_OP_WRITE: 2388 __rbd_osd_setup_write_ops(osd_req, which); 2389 break; 2390 case OBJ_OP_DISCARD: 2391 __rbd_osd_setup_discard_ops(osd_req, which); 2392 break; 2393 case OBJ_OP_ZEROOUT: 2394 __rbd_osd_setup_zeroout_ops(osd_req, which); 2395 break; 2396 default: 2397 BUG(); 2398 } 2399 } 2400 2401 /* 2402 * Prune the list of object requests (adjust offset and/or length, drop 2403 * redundant requests). Prepare object request state machines and image 2404 * request state machine for execution. 2405 */ 2406 static int __rbd_img_fill_request(struct rbd_img_request *img_req) 2407 { 2408 struct rbd_obj_request *obj_req, *next_obj_req; 2409 int ret; 2410 2411 for_each_obj_request_safe(img_req, obj_req, next_obj_req) { 2412 switch (img_req->op_type) { 2413 case OBJ_OP_READ: 2414 ret = rbd_obj_init_read(obj_req); 2415 break; 2416 case OBJ_OP_WRITE: 2417 ret = rbd_obj_init_write(obj_req); 2418 break; 2419 case OBJ_OP_DISCARD: 2420 ret = rbd_obj_init_discard(obj_req); 2421 break; 2422 case OBJ_OP_ZEROOUT: 2423 ret = rbd_obj_init_zeroout(obj_req); 2424 break; 2425 default: 2426 BUG(); 2427 } 2428 if (ret < 0) 2429 return ret; 2430 if (ret > 0) { 2431 rbd_img_obj_request_del(img_req, obj_req); 2432 continue; 2433 } 2434 } 2435 2436 img_req->state = RBD_IMG_START; 2437 return 0; 2438 } 2439 2440 union rbd_img_fill_iter { 2441 struct ceph_bio_iter bio_iter; 2442 struct ceph_bvec_iter bvec_iter; 2443 }; 2444 2445 struct rbd_img_fill_ctx { 2446 enum obj_request_type pos_type; 2447 union rbd_img_fill_iter *pos; 2448 union rbd_img_fill_iter iter; 2449 ceph_object_extent_fn_t set_pos_fn; 2450 ceph_object_extent_fn_t count_fn; 2451 ceph_object_extent_fn_t copy_fn; 2452 }; 2453 2454 static struct ceph_object_extent *alloc_object_extent(void *arg) 2455 { 2456 struct rbd_img_request *img_req = arg; 2457 struct rbd_obj_request *obj_req; 2458 2459 obj_req = rbd_obj_request_create(); 2460 if (!obj_req) 2461 return NULL; 2462 2463 rbd_img_obj_request_add(img_req, obj_req); 2464 return &obj_req->ex; 2465 } 2466 2467 /* 2468 * While su != os && sc == 1 is technically not fancy (it's the same 2469 * layout as su == os && sc == 1), we can't use the nocopy path for it 2470 * because ->set_pos_fn() should be called only once per object. 2471 * ceph_file_to_extents() invokes action_fn once per stripe unit, so 2472 * treat su != os && sc == 1 as fancy. 2473 */ 2474 static bool rbd_layout_is_fancy(struct ceph_file_layout *l) 2475 { 2476 return l->stripe_unit != l->object_size; 2477 } 2478 2479 static int rbd_img_fill_request_nocopy(struct rbd_img_request *img_req, 2480 struct ceph_file_extent *img_extents, 2481 u32 num_img_extents, 2482 struct rbd_img_fill_ctx *fctx) 2483 { 2484 u32 i; 2485 int ret; 2486 2487 img_req->data_type = fctx->pos_type; 2488 2489 /* 2490 * Create object requests and set each object request's starting 2491 * position in the provided bio (list) or bio_vec array. 2492 */ 2493 fctx->iter = *fctx->pos; 2494 for (i = 0; i < num_img_extents; i++) { 2495 ret = ceph_file_to_extents(&img_req->rbd_dev->layout, 2496 img_extents[i].fe_off, 2497 img_extents[i].fe_len, 2498 &img_req->object_extents, 2499 alloc_object_extent, img_req, 2500 fctx->set_pos_fn, &fctx->iter); 2501 if (ret) 2502 return ret; 2503 } 2504 2505 return __rbd_img_fill_request(img_req); 2506 } 2507 2508 /* 2509 * Map a list of image extents to a list of object extents, create the 2510 * corresponding object requests (normally each to a different object, 2511 * but not always) and add them to @img_req. For each object request, 2512 * set up its data descriptor to point to the corresponding chunk(s) of 2513 * @fctx->pos data buffer. 2514 * 2515 * Because ceph_file_to_extents() will merge adjacent object extents 2516 * together, each object request's data descriptor may point to multiple 2517 * different chunks of @fctx->pos data buffer. 2518 * 2519 * @fctx->pos data buffer is assumed to be large enough. 2520 */ 2521 static int rbd_img_fill_request(struct rbd_img_request *img_req, 2522 struct ceph_file_extent *img_extents, 2523 u32 num_img_extents, 2524 struct rbd_img_fill_ctx *fctx) 2525 { 2526 struct rbd_device *rbd_dev = img_req->rbd_dev; 2527 struct rbd_obj_request *obj_req; 2528 u32 i; 2529 int ret; 2530 2531 if (fctx->pos_type == OBJ_REQUEST_NODATA || 2532 !rbd_layout_is_fancy(&rbd_dev->layout)) 2533 return rbd_img_fill_request_nocopy(img_req, img_extents, 2534 num_img_extents, fctx); 2535 2536 img_req->data_type = OBJ_REQUEST_OWN_BVECS; 2537 2538 /* 2539 * Create object requests and determine ->bvec_count for each object 2540 * request. Note that ->bvec_count sum over all object requests may 2541 * be greater than the number of bio_vecs in the provided bio (list) 2542 * or bio_vec array because when mapped, those bio_vecs can straddle 2543 * stripe unit boundaries. 2544 */ 2545 fctx->iter = *fctx->pos; 2546 for (i = 0; i < num_img_extents; i++) { 2547 ret = ceph_file_to_extents(&rbd_dev->layout, 2548 img_extents[i].fe_off, 2549 img_extents[i].fe_len, 2550 &img_req->object_extents, 2551 alloc_object_extent, img_req, 2552 fctx->count_fn, &fctx->iter); 2553 if (ret) 2554 return ret; 2555 } 2556 2557 for_each_obj_request(img_req, obj_req) { 2558 obj_req->bvec_pos.bvecs = kmalloc_array(obj_req->bvec_count, 2559 sizeof(*obj_req->bvec_pos.bvecs), 2560 GFP_NOIO); 2561 if (!obj_req->bvec_pos.bvecs) 2562 return -ENOMEM; 2563 } 2564 2565 /* 2566 * Fill in each object request's private bio_vec array, splitting and 2567 * rearranging the provided bio_vecs in stripe unit chunks as needed. 2568 */ 2569 fctx->iter = *fctx->pos; 2570 for (i = 0; i < num_img_extents; i++) { 2571 ret = ceph_iterate_extents(&rbd_dev->layout, 2572 img_extents[i].fe_off, 2573 img_extents[i].fe_len, 2574 &img_req->object_extents, 2575 fctx->copy_fn, &fctx->iter); 2576 if (ret) 2577 return ret; 2578 } 2579 2580 return __rbd_img_fill_request(img_req); 2581 } 2582 2583 static int rbd_img_fill_nodata(struct rbd_img_request *img_req, 2584 u64 off, u64 len) 2585 { 2586 struct ceph_file_extent ex = { off, len }; 2587 union rbd_img_fill_iter dummy = {}; 2588 struct rbd_img_fill_ctx fctx = { 2589 .pos_type = OBJ_REQUEST_NODATA, 2590 .pos = &dummy, 2591 }; 2592 2593 return rbd_img_fill_request(img_req, &ex, 1, &fctx); 2594 } 2595 2596 static void set_bio_pos(struct ceph_object_extent *ex, u32 bytes, void *arg) 2597 { 2598 struct rbd_obj_request *obj_req = 2599 container_of(ex, struct rbd_obj_request, ex); 2600 struct ceph_bio_iter *it = arg; 2601 2602 dout("%s objno %llu bytes %u\n", __func__, ex->oe_objno, bytes); 2603 obj_req->bio_pos = *it; 2604 ceph_bio_iter_advance(it, bytes); 2605 } 2606 2607 static void count_bio_bvecs(struct ceph_object_extent *ex, u32 bytes, void *arg) 2608 { 2609 struct rbd_obj_request *obj_req = 2610 container_of(ex, struct rbd_obj_request, ex); 2611 struct ceph_bio_iter *it = arg; 2612 2613 dout("%s objno %llu bytes %u\n", __func__, ex->oe_objno, bytes); 2614 ceph_bio_iter_advance_step(it, bytes, ({ 2615 obj_req->bvec_count++; 2616 })); 2617 2618 } 2619 2620 static void copy_bio_bvecs(struct ceph_object_extent *ex, u32 bytes, void *arg) 2621 { 2622 struct rbd_obj_request *obj_req = 2623 container_of(ex, struct rbd_obj_request, ex); 2624 struct ceph_bio_iter *it = arg; 2625 2626 dout("%s objno %llu bytes %u\n", __func__, ex->oe_objno, bytes); 2627 ceph_bio_iter_advance_step(it, bytes, ({ 2628 obj_req->bvec_pos.bvecs[obj_req->bvec_idx++] = bv; 2629 obj_req->bvec_pos.iter.bi_size += bv.bv_len; 2630 })); 2631 } 2632 2633 static int __rbd_img_fill_from_bio(struct rbd_img_request *img_req, 2634 struct ceph_file_extent *img_extents, 2635 u32 num_img_extents, 2636 struct ceph_bio_iter *bio_pos) 2637 { 2638 struct rbd_img_fill_ctx fctx = { 2639 .pos_type = OBJ_REQUEST_BIO, 2640 .pos = (union rbd_img_fill_iter *)bio_pos, 2641 .set_pos_fn = set_bio_pos, 2642 .count_fn = count_bio_bvecs, 2643 .copy_fn = copy_bio_bvecs, 2644 }; 2645 2646 return rbd_img_fill_request(img_req, img_extents, num_img_extents, 2647 &fctx); 2648 } 2649 2650 static int rbd_img_fill_from_bio(struct rbd_img_request *img_req, 2651 u64 off, u64 len, struct bio *bio) 2652 { 2653 struct ceph_file_extent ex = { off, len }; 2654 struct ceph_bio_iter it = { .bio = bio, .iter = bio->bi_iter }; 2655 2656 return __rbd_img_fill_from_bio(img_req, &ex, 1, &it); 2657 } 2658 2659 static void set_bvec_pos(struct ceph_object_extent *ex, u32 bytes, void *arg) 2660 { 2661 struct rbd_obj_request *obj_req = 2662 container_of(ex, struct rbd_obj_request, ex); 2663 struct ceph_bvec_iter *it = arg; 2664 2665 obj_req->bvec_pos = *it; 2666 ceph_bvec_iter_shorten(&obj_req->bvec_pos, bytes); 2667 ceph_bvec_iter_advance(it, bytes); 2668 } 2669 2670 static void count_bvecs(struct ceph_object_extent *ex, u32 bytes, void *arg) 2671 { 2672 struct rbd_obj_request *obj_req = 2673 container_of(ex, struct rbd_obj_request, ex); 2674 struct ceph_bvec_iter *it = arg; 2675 2676 ceph_bvec_iter_advance_step(it, bytes, ({ 2677 obj_req->bvec_count++; 2678 })); 2679 } 2680 2681 static void copy_bvecs(struct ceph_object_extent *ex, u32 bytes, void *arg) 2682 { 2683 struct rbd_obj_request *obj_req = 2684 container_of(ex, struct rbd_obj_request, ex); 2685 struct ceph_bvec_iter *it = arg; 2686 2687 ceph_bvec_iter_advance_step(it, bytes, ({ 2688 obj_req->bvec_pos.bvecs[obj_req->bvec_idx++] = bv; 2689 obj_req->bvec_pos.iter.bi_size += bv.bv_len; 2690 })); 2691 } 2692 2693 static int __rbd_img_fill_from_bvecs(struct rbd_img_request *img_req, 2694 struct ceph_file_extent *img_extents, 2695 u32 num_img_extents, 2696 struct ceph_bvec_iter *bvec_pos) 2697 { 2698 struct rbd_img_fill_ctx fctx = { 2699 .pos_type = OBJ_REQUEST_BVECS, 2700 .pos = (union rbd_img_fill_iter *)bvec_pos, 2701 .set_pos_fn = set_bvec_pos, 2702 .count_fn = count_bvecs, 2703 .copy_fn = copy_bvecs, 2704 }; 2705 2706 return rbd_img_fill_request(img_req, img_extents, num_img_extents, 2707 &fctx); 2708 } 2709 2710 static int rbd_img_fill_from_bvecs(struct rbd_img_request *img_req, 2711 struct ceph_file_extent *img_extents, 2712 u32 num_img_extents, 2713 struct bio_vec *bvecs) 2714 { 2715 struct ceph_bvec_iter it = { 2716 .bvecs = bvecs, 2717 .iter = { .bi_size = ceph_file_extents_bytes(img_extents, 2718 num_img_extents) }, 2719 }; 2720 2721 return __rbd_img_fill_from_bvecs(img_req, img_extents, num_img_extents, 2722 &it); 2723 } 2724 2725 static void rbd_img_handle_request_work(struct work_struct *work) 2726 { 2727 struct rbd_img_request *img_req = 2728 container_of(work, struct rbd_img_request, work); 2729 2730 rbd_img_handle_request(img_req, img_req->work_result); 2731 } 2732 2733 static void rbd_img_schedule(struct rbd_img_request *img_req, int result) 2734 { 2735 INIT_WORK(&img_req->work, rbd_img_handle_request_work); 2736 img_req->work_result = result; 2737 queue_work(rbd_wq, &img_req->work); 2738 } 2739 2740 static bool rbd_obj_may_exist(struct rbd_obj_request *obj_req) 2741 { 2742 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 2743 2744 if (rbd_object_map_may_exist(rbd_dev, obj_req->ex.oe_objno)) { 2745 obj_req->flags |= RBD_OBJ_FLAG_MAY_EXIST; 2746 return true; 2747 } 2748 2749 dout("%s %p objno %llu assuming dne\n", __func__, obj_req, 2750 obj_req->ex.oe_objno); 2751 return false; 2752 } 2753 2754 static int rbd_obj_read_object(struct rbd_obj_request *obj_req) 2755 { 2756 struct ceph_osd_request *osd_req; 2757 int ret; 2758 2759 osd_req = __rbd_obj_add_osd_request(obj_req, NULL, 1); 2760 if (IS_ERR(osd_req)) 2761 return PTR_ERR(osd_req); 2762 2763 osd_req_op_extent_init(osd_req, 0, CEPH_OSD_OP_READ, 2764 obj_req->ex.oe_off, obj_req->ex.oe_len, 0, 0); 2765 rbd_osd_setup_data(osd_req, 0); 2766 rbd_osd_format_read(osd_req); 2767 2768 ret = ceph_osdc_alloc_messages(osd_req, GFP_NOIO); 2769 if (ret) 2770 return ret; 2771 2772 rbd_osd_submit(osd_req); 2773 return 0; 2774 } 2775 2776 static int rbd_obj_read_from_parent(struct rbd_obj_request *obj_req) 2777 { 2778 struct rbd_img_request *img_req = obj_req->img_request; 2779 struct rbd_device *parent = img_req->rbd_dev->parent; 2780 struct rbd_img_request *child_img_req; 2781 int ret; 2782 2783 child_img_req = kmem_cache_alloc(rbd_img_request_cache, GFP_NOIO); 2784 if (!child_img_req) 2785 return -ENOMEM; 2786 2787 rbd_img_request_init(child_img_req, parent, OBJ_OP_READ); 2788 __set_bit(IMG_REQ_CHILD, &child_img_req->flags); 2789 child_img_req->obj_request = obj_req; 2790 2791 down_read(&parent->header_rwsem); 2792 rbd_img_capture_header(child_img_req); 2793 up_read(&parent->header_rwsem); 2794 2795 dout("%s child_img_req %p for obj_req %p\n", __func__, child_img_req, 2796 obj_req); 2797 2798 if (!rbd_img_is_write(img_req)) { 2799 switch (img_req->data_type) { 2800 case OBJ_REQUEST_BIO: 2801 ret = __rbd_img_fill_from_bio(child_img_req, 2802 obj_req->img_extents, 2803 obj_req->num_img_extents, 2804 &obj_req->bio_pos); 2805 break; 2806 case OBJ_REQUEST_BVECS: 2807 case OBJ_REQUEST_OWN_BVECS: 2808 ret = __rbd_img_fill_from_bvecs(child_img_req, 2809 obj_req->img_extents, 2810 obj_req->num_img_extents, 2811 &obj_req->bvec_pos); 2812 break; 2813 default: 2814 BUG(); 2815 } 2816 } else { 2817 ret = rbd_img_fill_from_bvecs(child_img_req, 2818 obj_req->img_extents, 2819 obj_req->num_img_extents, 2820 obj_req->copyup_bvecs); 2821 } 2822 if (ret) { 2823 rbd_img_request_destroy(child_img_req); 2824 return ret; 2825 } 2826 2827 /* avoid parent chain recursion */ 2828 rbd_img_schedule(child_img_req, 0); 2829 return 0; 2830 } 2831 2832 static bool rbd_obj_advance_read(struct rbd_obj_request *obj_req, int *result) 2833 { 2834 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 2835 int ret; 2836 2837 again: 2838 switch (obj_req->read_state) { 2839 case RBD_OBJ_READ_START: 2840 rbd_assert(!*result); 2841 2842 if (!rbd_obj_may_exist(obj_req)) { 2843 *result = -ENOENT; 2844 obj_req->read_state = RBD_OBJ_READ_OBJECT; 2845 goto again; 2846 } 2847 2848 ret = rbd_obj_read_object(obj_req); 2849 if (ret) { 2850 *result = ret; 2851 return true; 2852 } 2853 obj_req->read_state = RBD_OBJ_READ_OBJECT; 2854 return false; 2855 case RBD_OBJ_READ_OBJECT: 2856 if (*result == -ENOENT && rbd_dev->parent_overlap) { 2857 /* reverse map this object extent onto the parent */ 2858 ret = rbd_obj_calc_img_extents(obj_req, false); 2859 if (ret) { 2860 *result = ret; 2861 return true; 2862 } 2863 if (obj_req->num_img_extents) { 2864 ret = rbd_obj_read_from_parent(obj_req); 2865 if (ret) { 2866 *result = ret; 2867 return true; 2868 } 2869 obj_req->read_state = RBD_OBJ_READ_PARENT; 2870 return false; 2871 } 2872 } 2873 2874 /* 2875 * -ENOENT means a hole in the image -- zero-fill the entire 2876 * length of the request. A short read also implies zero-fill 2877 * to the end of the request. 2878 */ 2879 if (*result == -ENOENT) { 2880 rbd_obj_zero_range(obj_req, 0, obj_req->ex.oe_len); 2881 *result = 0; 2882 } else if (*result >= 0) { 2883 if (*result < obj_req->ex.oe_len) 2884 rbd_obj_zero_range(obj_req, *result, 2885 obj_req->ex.oe_len - *result); 2886 else 2887 rbd_assert(*result == obj_req->ex.oe_len); 2888 *result = 0; 2889 } 2890 return true; 2891 case RBD_OBJ_READ_PARENT: 2892 /* 2893 * The parent image is read only up to the overlap -- zero-fill 2894 * from the overlap to the end of the request. 2895 */ 2896 if (!*result) { 2897 u32 obj_overlap = rbd_obj_img_extents_bytes(obj_req); 2898 2899 if (obj_overlap < obj_req->ex.oe_len) 2900 rbd_obj_zero_range(obj_req, obj_overlap, 2901 obj_req->ex.oe_len - obj_overlap); 2902 } 2903 return true; 2904 default: 2905 BUG(); 2906 } 2907 } 2908 2909 static bool rbd_obj_write_is_noop(struct rbd_obj_request *obj_req) 2910 { 2911 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 2912 2913 if (rbd_object_map_may_exist(rbd_dev, obj_req->ex.oe_objno)) 2914 obj_req->flags |= RBD_OBJ_FLAG_MAY_EXIST; 2915 2916 if (!(obj_req->flags & RBD_OBJ_FLAG_MAY_EXIST) && 2917 (obj_req->flags & RBD_OBJ_FLAG_NOOP_FOR_NONEXISTENT)) { 2918 dout("%s %p noop for nonexistent\n", __func__, obj_req); 2919 return true; 2920 } 2921 2922 return false; 2923 } 2924 2925 /* 2926 * Return: 2927 * 0 - object map update sent 2928 * 1 - object map update isn't needed 2929 * <0 - error 2930 */ 2931 static int rbd_obj_write_pre_object_map(struct rbd_obj_request *obj_req) 2932 { 2933 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 2934 u8 new_state; 2935 2936 if (!(rbd_dev->header.features & RBD_FEATURE_OBJECT_MAP)) 2937 return 1; 2938 2939 if (obj_req->flags & RBD_OBJ_FLAG_DELETION) 2940 new_state = OBJECT_PENDING; 2941 else 2942 new_state = OBJECT_EXISTS; 2943 2944 return rbd_object_map_update(obj_req, CEPH_NOSNAP, new_state, NULL); 2945 } 2946 2947 static int rbd_obj_write_object(struct rbd_obj_request *obj_req) 2948 { 2949 struct ceph_osd_request *osd_req; 2950 int num_ops = count_write_ops(obj_req); 2951 int which = 0; 2952 int ret; 2953 2954 if (obj_req->flags & RBD_OBJ_FLAG_COPYUP_ENABLED) 2955 num_ops++; /* stat */ 2956 2957 osd_req = rbd_obj_add_osd_request(obj_req, num_ops); 2958 if (IS_ERR(osd_req)) 2959 return PTR_ERR(osd_req); 2960 2961 if (obj_req->flags & RBD_OBJ_FLAG_COPYUP_ENABLED) { 2962 ret = rbd_osd_setup_stat(osd_req, which++); 2963 if (ret) 2964 return ret; 2965 } 2966 2967 rbd_osd_setup_write_ops(osd_req, which); 2968 rbd_osd_format_write(osd_req); 2969 2970 ret = ceph_osdc_alloc_messages(osd_req, GFP_NOIO); 2971 if (ret) 2972 return ret; 2973 2974 rbd_osd_submit(osd_req); 2975 return 0; 2976 } 2977 2978 /* 2979 * copyup_bvecs pages are never highmem pages 2980 */ 2981 static bool is_zero_bvecs(struct bio_vec *bvecs, u32 bytes) 2982 { 2983 struct ceph_bvec_iter it = { 2984 .bvecs = bvecs, 2985 .iter = { .bi_size = bytes }, 2986 }; 2987 2988 ceph_bvec_iter_advance_step(&it, bytes, ({ 2989 if (memchr_inv(bvec_virt(&bv), 0, bv.bv_len)) 2990 return false; 2991 })); 2992 return true; 2993 } 2994 2995 #define MODS_ONLY U32_MAX 2996 2997 static int rbd_obj_copyup_empty_snapc(struct rbd_obj_request *obj_req, 2998 u32 bytes) 2999 { 3000 struct ceph_osd_request *osd_req; 3001 int ret; 3002 3003 dout("%s obj_req %p bytes %u\n", __func__, obj_req, bytes); 3004 rbd_assert(bytes > 0 && bytes != MODS_ONLY); 3005 3006 osd_req = __rbd_obj_add_osd_request(obj_req, &rbd_empty_snapc, 1); 3007 if (IS_ERR(osd_req)) 3008 return PTR_ERR(osd_req); 3009 3010 ret = rbd_osd_setup_copyup(osd_req, 0, bytes); 3011 if (ret) 3012 return ret; 3013 3014 rbd_osd_format_write(osd_req); 3015 3016 ret = ceph_osdc_alloc_messages(osd_req, GFP_NOIO); 3017 if (ret) 3018 return ret; 3019 3020 rbd_osd_submit(osd_req); 3021 return 0; 3022 } 3023 3024 static int rbd_obj_copyup_current_snapc(struct rbd_obj_request *obj_req, 3025 u32 bytes) 3026 { 3027 struct ceph_osd_request *osd_req; 3028 int num_ops = count_write_ops(obj_req); 3029 int which = 0; 3030 int ret; 3031 3032 dout("%s obj_req %p bytes %u\n", __func__, obj_req, bytes); 3033 3034 if (bytes != MODS_ONLY) 3035 num_ops++; /* copyup */ 3036 3037 osd_req = rbd_obj_add_osd_request(obj_req, num_ops); 3038 if (IS_ERR(osd_req)) 3039 return PTR_ERR(osd_req); 3040 3041 if (bytes != MODS_ONLY) { 3042 ret = rbd_osd_setup_copyup(osd_req, which++, bytes); 3043 if (ret) 3044 return ret; 3045 } 3046 3047 rbd_osd_setup_write_ops(osd_req, which); 3048 rbd_osd_format_write(osd_req); 3049 3050 ret = ceph_osdc_alloc_messages(osd_req, GFP_NOIO); 3051 if (ret) 3052 return ret; 3053 3054 rbd_osd_submit(osd_req); 3055 return 0; 3056 } 3057 3058 static int setup_copyup_bvecs(struct rbd_obj_request *obj_req, u64 obj_overlap) 3059 { 3060 u32 i; 3061 3062 rbd_assert(!obj_req->copyup_bvecs); 3063 obj_req->copyup_bvec_count = calc_pages_for(0, obj_overlap); 3064 obj_req->copyup_bvecs = kcalloc(obj_req->copyup_bvec_count, 3065 sizeof(*obj_req->copyup_bvecs), 3066 GFP_NOIO); 3067 if (!obj_req->copyup_bvecs) 3068 return -ENOMEM; 3069 3070 for (i = 0; i < obj_req->copyup_bvec_count; i++) { 3071 unsigned int len = min(obj_overlap, (u64)PAGE_SIZE); 3072 3073 obj_req->copyup_bvecs[i].bv_page = alloc_page(GFP_NOIO); 3074 if (!obj_req->copyup_bvecs[i].bv_page) 3075 return -ENOMEM; 3076 3077 obj_req->copyup_bvecs[i].bv_offset = 0; 3078 obj_req->copyup_bvecs[i].bv_len = len; 3079 obj_overlap -= len; 3080 } 3081 3082 rbd_assert(!obj_overlap); 3083 return 0; 3084 } 3085 3086 /* 3087 * The target object doesn't exist. Read the data for the entire 3088 * target object up to the overlap point (if any) from the parent, 3089 * so we can use it for a copyup. 3090 */ 3091 static int rbd_obj_copyup_read_parent(struct rbd_obj_request *obj_req) 3092 { 3093 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 3094 int ret; 3095 3096 rbd_assert(obj_req->num_img_extents); 3097 prune_extents(obj_req->img_extents, &obj_req->num_img_extents, 3098 rbd_dev->parent_overlap); 3099 if (!obj_req->num_img_extents) { 3100 /* 3101 * The overlap has become 0 (most likely because the 3102 * image has been flattened). Re-submit the original write 3103 * request -- pass MODS_ONLY since the copyup isn't needed 3104 * anymore. 3105 */ 3106 return rbd_obj_copyup_current_snapc(obj_req, MODS_ONLY); 3107 } 3108 3109 ret = setup_copyup_bvecs(obj_req, rbd_obj_img_extents_bytes(obj_req)); 3110 if (ret) 3111 return ret; 3112 3113 return rbd_obj_read_from_parent(obj_req); 3114 } 3115 3116 static void rbd_obj_copyup_object_maps(struct rbd_obj_request *obj_req) 3117 { 3118 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 3119 struct ceph_snap_context *snapc = obj_req->img_request->snapc; 3120 u8 new_state; 3121 u32 i; 3122 int ret; 3123 3124 rbd_assert(!obj_req->pending.result && !obj_req->pending.num_pending); 3125 3126 if (!(rbd_dev->header.features & RBD_FEATURE_OBJECT_MAP)) 3127 return; 3128 3129 if (obj_req->flags & RBD_OBJ_FLAG_COPYUP_ZEROS) 3130 return; 3131 3132 for (i = 0; i < snapc->num_snaps; i++) { 3133 if ((rbd_dev->header.features & RBD_FEATURE_FAST_DIFF) && 3134 i + 1 < snapc->num_snaps) 3135 new_state = OBJECT_EXISTS_CLEAN; 3136 else 3137 new_state = OBJECT_EXISTS; 3138 3139 ret = rbd_object_map_update(obj_req, snapc->snaps[i], 3140 new_state, NULL); 3141 if (ret < 0) { 3142 obj_req->pending.result = ret; 3143 return; 3144 } 3145 3146 rbd_assert(!ret); 3147 obj_req->pending.num_pending++; 3148 } 3149 } 3150 3151 static void rbd_obj_copyup_write_object(struct rbd_obj_request *obj_req) 3152 { 3153 u32 bytes = rbd_obj_img_extents_bytes(obj_req); 3154 int ret; 3155 3156 rbd_assert(!obj_req->pending.result && !obj_req->pending.num_pending); 3157 3158 /* 3159 * Only send non-zero copyup data to save some I/O and network 3160 * bandwidth -- zero copyup data is equivalent to the object not 3161 * existing. 3162 */ 3163 if (obj_req->flags & RBD_OBJ_FLAG_COPYUP_ZEROS) 3164 bytes = 0; 3165 3166 if (obj_req->img_request->snapc->num_snaps && bytes > 0) { 3167 /* 3168 * Send a copyup request with an empty snapshot context to 3169 * deep-copyup the object through all existing snapshots. 3170 * A second request with the current snapshot context will be 3171 * sent for the actual modification. 3172 */ 3173 ret = rbd_obj_copyup_empty_snapc(obj_req, bytes); 3174 if (ret) { 3175 obj_req->pending.result = ret; 3176 return; 3177 } 3178 3179 obj_req->pending.num_pending++; 3180 bytes = MODS_ONLY; 3181 } 3182 3183 ret = rbd_obj_copyup_current_snapc(obj_req, bytes); 3184 if (ret) { 3185 obj_req->pending.result = ret; 3186 return; 3187 } 3188 3189 obj_req->pending.num_pending++; 3190 } 3191 3192 static bool rbd_obj_advance_copyup(struct rbd_obj_request *obj_req, int *result) 3193 { 3194 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 3195 int ret; 3196 3197 again: 3198 switch (obj_req->copyup_state) { 3199 case RBD_OBJ_COPYUP_START: 3200 rbd_assert(!*result); 3201 3202 ret = rbd_obj_copyup_read_parent(obj_req); 3203 if (ret) { 3204 *result = ret; 3205 return true; 3206 } 3207 if (obj_req->num_img_extents) 3208 obj_req->copyup_state = RBD_OBJ_COPYUP_READ_PARENT; 3209 else 3210 obj_req->copyup_state = RBD_OBJ_COPYUP_WRITE_OBJECT; 3211 return false; 3212 case RBD_OBJ_COPYUP_READ_PARENT: 3213 if (*result) 3214 return true; 3215 3216 if (is_zero_bvecs(obj_req->copyup_bvecs, 3217 rbd_obj_img_extents_bytes(obj_req))) { 3218 dout("%s %p detected zeros\n", __func__, obj_req); 3219 obj_req->flags |= RBD_OBJ_FLAG_COPYUP_ZEROS; 3220 } 3221 3222 rbd_obj_copyup_object_maps(obj_req); 3223 if (!obj_req->pending.num_pending) { 3224 *result = obj_req->pending.result; 3225 obj_req->copyup_state = RBD_OBJ_COPYUP_OBJECT_MAPS; 3226 goto again; 3227 } 3228 obj_req->copyup_state = __RBD_OBJ_COPYUP_OBJECT_MAPS; 3229 return false; 3230 case __RBD_OBJ_COPYUP_OBJECT_MAPS: 3231 if (!pending_result_dec(&obj_req->pending, result)) 3232 return false; 3233 fallthrough; 3234 case RBD_OBJ_COPYUP_OBJECT_MAPS: 3235 if (*result) { 3236 rbd_warn(rbd_dev, "snap object map update failed: %d", 3237 *result); 3238 return true; 3239 } 3240 3241 rbd_obj_copyup_write_object(obj_req); 3242 if (!obj_req->pending.num_pending) { 3243 *result = obj_req->pending.result; 3244 obj_req->copyup_state = RBD_OBJ_COPYUP_WRITE_OBJECT; 3245 goto again; 3246 } 3247 obj_req->copyup_state = __RBD_OBJ_COPYUP_WRITE_OBJECT; 3248 return false; 3249 case __RBD_OBJ_COPYUP_WRITE_OBJECT: 3250 if (!pending_result_dec(&obj_req->pending, result)) 3251 return false; 3252 fallthrough; 3253 case RBD_OBJ_COPYUP_WRITE_OBJECT: 3254 return true; 3255 default: 3256 BUG(); 3257 } 3258 } 3259 3260 /* 3261 * Return: 3262 * 0 - object map update sent 3263 * 1 - object map update isn't needed 3264 * <0 - error 3265 */ 3266 static int rbd_obj_write_post_object_map(struct rbd_obj_request *obj_req) 3267 { 3268 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 3269 u8 current_state = OBJECT_PENDING; 3270 3271 if (!(rbd_dev->header.features & RBD_FEATURE_OBJECT_MAP)) 3272 return 1; 3273 3274 if (!(obj_req->flags & RBD_OBJ_FLAG_DELETION)) 3275 return 1; 3276 3277 return rbd_object_map_update(obj_req, CEPH_NOSNAP, OBJECT_NONEXISTENT, 3278 ¤t_state); 3279 } 3280 3281 static bool rbd_obj_advance_write(struct rbd_obj_request *obj_req, int *result) 3282 { 3283 struct rbd_device *rbd_dev = obj_req->img_request->rbd_dev; 3284 int ret; 3285 3286 again: 3287 switch (obj_req->write_state) { 3288 case RBD_OBJ_WRITE_START: 3289 rbd_assert(!*result); 3290 3291 if (rbd_obj_write_is_noop(obj_req)) 3292 return true; 3293 3294 ret = rbd_obj_write_pre_object_map(obj_req); 3295 if (ret < 0) { 3296 *result = ret; 3297 return true; 3298 } 3299 obj_req->write_state = RBD_OBJ_WRITE_PRE_OBJECT_MAP; 3300 if (ret > 0) 3301 goto again; 3302 return false; 3303 case RBD_OBJ_WRITE_PRE_OBJECT_MAP: 3304 if (*result) { 3305 rbd_warn(rbd_dev, "pre object map update failed: %d", 3306 *result); 3307 return true; 3308 } 3309 ret = rbd_obj_write_object(obj_req); 3310 if (ret) { 3311 *result = ret; 3312 return true; 3313 } 3314 obj_req->write_state = RBD_OBJ_WRITE_OBJECT; 3315 return false; 3316 case RBD_OBJ_WRITE_OBJECT: 3317 if (*result == -ENOENT) { 3318 if (obj_req->flags & RBD_OBJ_FLAG_COPYUP_ENABLED) { 3319 *result = 0; 3320 obj_req->copyup_state = RBD_OBJ_COPYUP_START; 3321 obj_req->write_state = __RBD_OBJ_WRITE_COPYUP; 3322 goto again; 3323 } 3324 /* 3325 * On a non-existent object: 3326 * delete - -ENOENT, truncate/zero - 0 3327 */ 3328 if (obj_req->flags & RBD_OBJ_FLAG_DELETION) 3329 *result = 0; 3330 } 3331 if (*result) 3332 return true; 3333 3334 obj_req->write_state = RBD_OBJ_WRITE_COPYUP; 3335 goto again; 3336 case __RBD_OBJ_WRITE_COPYUP: 3337 if (!rbd_obj_advance_copyup(obj_req, result)) 3338 return false; 3339 fallthrough; 3340 case RBD_OBJ_WRITE_COPYUP: 3341 if (*result) { 3342 rbd_warn(rbd_dev, "copyup failed: %d", *result); 3343 return true; 3344 } 3345 ret = rbd_obj_write_post_object_map(obj_req); 3346 if (ret < 0) { 3347 *result = ret; 3348 return true; 3349 } 3350 obj_req->write_state = RBD_OBJ_WRITE_POST_OBJECT_MAP; 3351 if (ret > 0) 3352 goto again; 3353 return false; 3354 case RBD_OBJ_WRITE_POST_OBJECT_MAP: 3355 if (*result) 3356 rbd_warn(rbd_dev, "post object map update failed: %d", 3357 *result); 3358 return true; 3359 default: 3360 BUG(); 3361 } 3362 } 3363 3364 /* 3365 * Return true if @obj_req is completed. 3366 */ 3367 static bool __rbd_obj_handle_request(struct rbd_obj_request *obj_req, 3368 int *result) 3369 { 3370 struct rbd_img_request *img_req = obj_req->img_request; 3371 struct rbd_device *rbd_dev = img_req->rbd_dev; 3372 bool done; 3373 3374 mutex_lock(&obj_req->state_mutex); 3375 if (!rbd_img_is_write(img_req)) 3376 done = rbd_obj_advance_read(obj_req, result); 3377 else 3378 done = rbd_obj_advance_write(obj_req, result); 3379 mutex_unlock(&obj_req->state_mutex); 3380 3381 if (done && *result) { 3382 rbd_assert(*result < 0); 3383 rbd_warn(rbd_dev, "%s at objno %llu %llu~%llu result %d", 3384 obj_op_name(img_req->op_type), obj_req->ex.oe_objno, 3385 obj_req->ex.oe_off, obj_req->ex.oe_len, *result); 3386 } 3387 return done; 3388 } 3389 3390 /* 3391 * This is open-coded in rbd_img_handle_request() to avoid parent chain 3392 * recursion. 3393 */ 3394 static void rbd_obj_handle_request(struct rbd_obj_request *obj_req, int result) 3395 { 3396 if (__rbd_obj_handle_request(obj_req, &result)) 3397 rbd_img_handle_request(obj_req->img_request, result); 3398 } 3399 3400 static bool need_exclusive_lock(struct rbd_img_request *img_req) 3401 { 3402 struct rbd_device *rbd_dev = img_req->rbd_dev; 3403 3404 if (!(rbd_dev->header.features & RBD_FEATURE_EXCLUSIVE_LOCK)) 3405 return false; 3406 3407 if (rbd_is_ro(rbd_dev)) 3408 return false; 3409 3410 rbd_assert(!test_bit(IMG_REQ_CHILD, &img_req->flags)); 3411 if (rbd_dev->opts->lock_on_read || 3412 (rbd_dev->header.features & RBD_FEATURE_OBJECT_MAP)) 3413 return true; 3414 3415 return rbd_img_is_write(img_req); 3416 } 3417 3418 static bool rbd_lock_add_request(struct rbd_img_request *img_req) 3419 { 3420 struct rbd_device *rbd_dev = img_req->rbd_dev; 3421 bool locked; 3422 3423 lockdep_assert_held(&rbd_dev->lock_rwsem); 3424 locked = rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED; 3425 spin_lock(&rbd_dev->lock_lists_lock); 3426 rbd_assert(list_empty(&img_req->lock_item)); 3427 if (!locked) 3428 list_add_tail(&img_req->lock_item, &rbd_dev->acquiring_list); 3429 else 3430 list_add_tail(&img_req->lock_item, &rbd_dev->running_list); 3431 spin_unlock(&rbd_dev->lock_lists_lock); 3432 return locked; 3433 } 3434 3435 static void rbd_lock_del_request(struct rbd_img_request *img_req) 3436 { 3437 struct rbd_device *rbd_dev = img_req->rbd_dev; 3438 bool need_wakeup; 3439 3440 lockdep_assert_held(&rbd_dev->lock_rwsem); 3441 spin_lock(&rbd_dev->lock_lists_lock); 3442 rbd_assert(!list_empty(&img_req->lock_item)); 3443 list_del_init(&img_req->lock_item); 3444 need_wakeup = (rbd_dev->lock_state == RBD_LOCK_STATE_RELEASING && 3445 list_empty(&rbd_dev->running_list)); 3446 spin_unlock(&rbd_dev->lock_lists_lock); 3447 if (need_wakeup) 3448 complete(&rbd_dev->releasing_wait); 3449 } 3450 3451 static int rbd_img_exclusive_lock(struct rbd_img_request *img_req) 3452 { 3453 struct rbd_device *rbd_dev = img_req->rbd_dev; 3454 3455 if (!need_exclusive_lock(img_req)) 3456 return 1; 3457 3458 if (rbd_lock_add_request(img_req)) 3459 return 1; 3460 3461 if (rbd_dev->opts->exclusive) { 3462 WARN_ON(1); /* lock got released? */ 3463 return -EROFS; 3464 } 3465 3466 /* 3467 * Note the use of mod_delayed_work() in rbd_acquire_lock() 3468 * and cancel_delayed_work() in wake_lock_waiters(). 3469 */ 3470 dout("%s rbd_dev %p queueing lock_dwork\n", __func__, rbd_dev); 3471 queue_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork, 0); 3472 return 0; 3473 } 3474 3475 static void rbd_img_object_requests(struct rbd_img_request *img_req) 3476 { 3477 struct rbd_obj_request *obj_req; 3478 3479 rbd_assert(!img_req->pending.result && !img_req->pending.num_pending); 3480 3481 for_each_obj_request(img_req, obj_req) { 3482 int result = 0; 3483 3484 if (__rbd_obj_handle_request(obj_req, &result)) { 3485 if (result) { 3486 img_req->pending.result = result; 3487 return; 3488 } 3489 } else { 3490 img_req->pending.num_pending++; 3491 } 3492 } 3493 } 3494 3495 static bool rbd_img_advance(struct rbd_img_request *img_req, int *result) 3496 { 3497 struct rbd_device *rbd_dev = img_req->rbd_dev; 3498 int ret; 3499 3500 again: 3501 switch (img_req->state) { 3502 case RBD_IMG_START: 3503 rbd_assert(!*result); 3504 3505 ret = rbd_img_exclusive_lock(img_req); 3506 if (ret < 0) { 3507 *result = ret; 3508 return true; 3509 } 3510 img_req->state = RBD_IMG_EXCLUSIVE_LOCK; 3511 if (ret > 0) 3512 goto again; 3513 return false; 3514 case RBD_IMG_EXCLUSIVE_LOCK: 3515 if (*result) 3516 return true; 3517 3518 rbd_assert(!need_exclusive_lock(img_req) || 3519 __rbd_is_lock_owner(rbd_dev)); 3520 3521 rbd_img_object_requests(img_req); 3522 if (!img_req->pending.num_pending) { 3523 *result = img_req->pending.result; 3524 img_req->state = RBD_IMG_OBJECT_REQUESTS; 3525 goto again; 3526 } 3527 img_req->state = __RBD_IMG_OBJECT_REQUESTS; 3528 return false; 3529 case __RBD_IMG_OBJECT_REQUESTS: 3530 if (!pending_result_dec(&img_req->pending, result)) 3531 return false; 3532 fallthrough; 3533 case RBD_IMG_OBJECT_REQUESTS: 3534 return true; 3535 default: 3536 BUG(); 3537 } 3538 } 3539 3540 /* 3541 * Return true if @img_req is completed. 3542 */ 3543 static bool __rbd_img_handle_request(struct rbd_img_request *img_req, 3544 int *result) 3545 { 3546 struct rbd_device *rbd_dev = img_req->rbd_dev; 3547 bool done; 3548 3549 if (need_exclusive_lock(img_req)) { 3550 down_read(&rbd_dev->lock_rwsem); 3551 mutex_lock(&img_req->state_mutex); 3552 done = rbd_img_advance(img_req, result); 3553 if (done) 3554 rbd_lock_del_request(img_req); 3555 mutex_unlock(&img_req->state_mutex); 3556 up_read(&rbd_dev->lock_rwsem); 3557 } else { 3558 mutex_lock(&img_req->state_mutex); 3559 done = rbd_img_advance(img_req, result); 3560 mutex_unlock(&img_req->state_mutex); 3561 } 3562 3563 if (done && *result) { 3564 rbd_assert(*result < 0); 3565 rbd_warn(rbd_dev, "%s%s result %d", 3566 test_bit(IMG_REQ_CHILD, &img_req->flags) ? "child " : "", 3567 obj_op_name(img_req->op_type), *result); 3568 } 3569 return done; 3570 } 3571 3572 static void rbd_img_handle_request(struct rbd_img_request *img_req, int result) 3573 { 3574 again: 3575 if (!__rbd_img_handle_request(img_req, &result)) 3576 return; 3577 3578 if (test_bit(IMG_REQ_CHILD, &img_req->flags)) { 3579 struct rbd_obj_request *obj_req = img_req->obj_request; 3580 3581 rbd_img_request_destroy(img_req); 3582 if (__rbd_obj_handle_request(obj_req, &result)) { 3583 img_req = obj_req->img_request; 3584 goto again; 3585 } 3586 } else { 3587 struct request *rq = blk_mq_rq_from_pdu(img_req); 3588 3589 rbd_img_request_destroy(img_req); 3590 blk_mq_end_request(rq, errno_to_blk_status(result)); 3591 } 3592 } 3593 3594 static const struct rbd_client_id rbd_empty_cid; 3595 3596 static bool rbd_cid_equal(const struct rbd_client_id *lhs, 3597 const struct rbd_client_id *rhs) 3598 { 3599 return lhs->gid == rhs->gid && lhs->handle == rhs->handle; 3600 } 3601 3602 static struct rbd_client_id rbd_get_cid(struct rbd_device *rbd_dev) 3603 { 3604 struct rbd_client_id cid; 3605 3606 mutex_lock(&rbd_dev->watch_mutex); 3607 cid.gid = ceph_client_gid(rbd_dev->rbd_client->client); 3608 cid.handle = rbd_dev->watch_cookie; 3609 mutex_unlock(&rbd_dev->watch_mutex); 3610 return cid; 3611 } 3612 3613 /* 3614 * lock_rwsem must be held for write 3615 */ 3616 static void rbd_set_owner_cid(struct rbd_device *rbd_dev, 3617 const struct rbd_client_id *cid) 3618 { 3619 dout("%s rbd_dev %p %llu-%llu -> %llu-%llu\n", __func__, rbd_dev, 3620 rbd_dev->owner_cid.gid, rbd_dev->owner_cid.handle, 3621 cid->gid, cid->handle); 3622 rbd_dev->owner_cid = *cid; /* struct */ 3623 } 3624 3625 static void format_lock_cookie(struct rbd_device *rbd_dev, char *buf) 3626 { 3627 mutex_lock(&rbd_dev->watch_mutex); 3628 sprintf(buf, "%s %llu", RBD_LOCK_COOKIE_PREFIX, rbd_dev->watch_cookie); 3629 mutex_unlock(&rbd_dev->watch_mutex); 3630 } 3631 3632 static void __rbd_lock(struct rbd_device *rbd_dev, const char *cookie) 3633 { 3634 struct rbd_client_id cid = rbd_get_cid(rbd_dev); 3635 3636 rbd_dev->lock_state = RBD_LOCK_STATE_LOCKED; 3637 strcpy(rbd_dev->lock_cookie, cookie); 3638 rbd_set_owner_cid(rbd_dev, &cid); 3639 queue_work(rbd_dev->task_wq, &rbd_dev->acquired_lock_work); 3640 } 3641 3642 /* 3643 * lock_rwsem must be held for write 3644 */ 3645 static int rbd_lock(struct rbd_device *rbd_dev) 3646 { 3647 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 3648 char cookie[32]; 3649 int ret; 3650 3651 WARN_ON(__rbd_is_lock_owner(rbd_dev) || 3652 rbd_dev->lock_cookie[0] != '\0'); 3653 3654 format_lock_cookie(rbd_dev, cookie); 3655 ret = ceph_cls_lock(osdc, &rbd_dev->header_oid, &rbd_dev->header_oloc, 3656 RBD_LOCK_NAME, CEPH_CLS_LOCK_EXCLUSIVE, cookie, 3657 RBD_LOCK_TAG, "", 0); 3658 if (ret) 3659 return ret; 3660 3661 __rbd_lock(rbd_dev, cookie); 3662 return 0; 3663 } 3664 3665 /* 3666 * lock_rwsem must be held for write 3667 */ 3668 static void rbd_unlock(struct rbd_device *rbd_dev) 3669 { 3670 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 3671 int ret; 3672 3673 WARN_ON(!__rbd_is_lock_owner(rbd_dev) || 3674 rbd_dev->lock_cookie[0] == '\0'); 3675 3676 ret = ceph_cls_unlock(osdc, &rbd_dev->header_oid, &rbd_dev->header_oloc, 3677 RBD_LOCK_NAME, rbd_dev->lock_cookie); 3678 if (ret && ret != -ENOENT) 3679 rbd_warn(rbd_dev, "failed to unlock header: %d", ret); 3680 3681 /* treat errors as the image is unlocked */ 3682 rbd_dev->lock_state = RBD_LOCK_STATE_UNLOCKED; 3683 rbd_dev->lock_cookie[0] = '\0'; 3684 rbd_set_owner_cid(rbd_dev, &rbd_empty_cid); 3685 queue_work(rbd_dev->task_wq, &rbd_dev->released_lock_work); 3686 } 3687 3688 static int __rbd_notify_op_lock(struct rbd_device *rbd_dev, 3689 enum rbd_notify_op notify_op, 3690 struct page ***preply_pages, 3691 size_t *preply_len) 3692 { 3693 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 3694 struct rbd_client_id cid = rbd_get_cid(rbd_dev); 3695 char buf[4 + 8 + 8 + CEPH_ENCODING_START_BLK_LEN]; 3696 int buf_size = sizeof(buf); 3697 void *p = buf; 3698 3699 dout("%s rbd_dev %p notify_op %d\n", __func__, rbd_dev, notify_op); 3700 3701 /* encode *LockPayload NotifyMessage (op + ClientId) */ 3702 ceph_start_encoding(&p, 2, 1, buf_size - CEPH_ENCODING_START_BLK_LEN); 3703 ceph_encode_32(&p, notify_op); 3704 ceph_encode_64(&p, cid.gid); 3705 ceph_encode_64(&p, cid.handle); 3706 3707 return ceph_osdc_notify(osdc, &rbd_dev->header_oid, 3708 &rbd_dev->header_oloc, buf, buf_size, 3709 RBD_NOTIFY_TIMEOUT, preply_pages, preply_len); 3710 } 3711 3712 static void rbd_notify_op_lock(struct rbd_device *rbd_dev, 3713 enum rbd_notify_op notify_op) 3714 { 3715 __rbd_notify_op_lock(rbd_dev, notify_op, NULL, NULL); 3716 } 3717 3718 static void rbd_notify_acquired_lock(struct work_struct *work) 3719 { 3720 struct rbd_device *rbd_dev = container_of(work, struct rbd_device, 3721 acquired_lock_work); 3722 3723 rbd_notify_op_lock(rbd_dev, RBD_NOTIFY_OP_ACQUIRED_LOCK); 3724 } 3725 3726 static void rbd_notify_released_lock(struct work_struct *work) 3727 { 3728 struct rbd_device *rbd_dev = container_of(work, struct rbd_device, 3729 released_lock_work); 3730 3731 rbd_notify_op_lock(rbd_dev, RBD_NOTIFY_OP_RELEASED_LOCK); 3732 } 3733 3734 static int rbd_request_lock(struct rbd_device *rbd_dev) 3735 { 3736 struct page **reply_pages; 3737 size_t reply_len; 3738 bool lock_owner_responded = false; 3739 int ret; 3740 3741 dout("%s rbd_dev %p\n", __func__, rbd_dev); 3742 3743 ret = __rbd_notify_op_lock(rbd_dev, RBD_NOTIFY_OP_REQUEST_LOCK, 3744 &reply_pages, &reply_len); 3745 if (ret && ret != -ETIMEDOUT) { 3746 rbd_warn(rbd_dev, "failed to request lock: %d", ret); 3747 goto out; 3748 } 3749 3750 if (reply_len > 0 && reply_len <= PAGE_SIZE) { 3751 void *p = page_address(reply_pages[0]); 3752 void *const end = p + reply_len; 3753 u32 n; 3754 3755 ceph_decode_32_safe(&p, end, n, e_inval); /* num_acks */ 3756 while (n--) { 3757 u8 struct_v; 3758 u32 len; 3759 3760 ceph_decode_need(&p, end, 8 + 8, e_inval); 3761 p += 8 + 8; /* skip gid and cookie */ 3762 3763 ceph_decode_32_safe(&p, end, len, e_inval); 3764 if (!len) 3765 continue; 3766 3767 if (lock_owner_responded) { 3768 rbd_warn(rbd_dev, 3769 "duplicate lock owners detected"); 3770 ret = -EIO; 3771 goto out; 3772 } 3773 3774 lock_owner_responded = true; 3775 ret = ceph_start_decoding(&p, end, 1, "ResponseMessage", 3776 &struct_v, &len); 3777 if (ret) { 3778 rbd_warn(rbd_dev, 3779 "failed to decode ResponseMessage: %d", 3780 ret); 3781 goto e_inval; 3782 } 3783 3784 ret = ceph_decode_32(&p); 3785 } 3786 } 3787 3788 if (!lock_owner_responded) { 3789 rbd_warn(rbd_dev, "no lock owners detected"); 3790 ret = -ETIMEDOUT; 3791 } 3792 3793 out: 3794 ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len)); 3795 return ret; 3796 3797 e_inval: 3798 ret = -EINVAL; 3799 goto out; 3800 } 3801 3802 /* 3803 * Either image request state machine(s) or rbd_add_acquire_lock() 3804 * (i.e. "rbd map"). 3805 */ 3806 static void wake_lock_waiters(struct rbd_device *rbd_dev, int result) 3807 { 3808 struct rbd_img_request *img_req; 3809 3810 dout("%s rbd_dev %p result %d\n", __func__, rbd_dev, result); 3811 lockdep_assert_held_write(&rbd_dev->lock_rwsem); 3812 3813 cancel_delayed_work(&rbd_dev->lock_dwork); 3814 if (!completion_done(&rbd_dev->acquire_wait)) { 3815 rbd_assert(list_empty(&rbd_dev->acquiring_list) && 3816 list_empty(&rbd_dev->running_list)); 3817 rbd_dev->acquire_err = result; 3818 complete_all(&rbd_dev->acquire_wait); 3819 return; 3820 } 3821 3822 list_for_each_entry(img_req, &rbd_dev->acquiring_list, lock_item) { 3823 mutex_lock(&img_req->state_mutex); 3824 rbd_assert(img_req->state == RBD_IMG_EXCLUSIVE_LOCK); 3825 rbd_img_schedule(img_req, result); 3826 mutex_unlock(&img_req->state_mutex); 3827 } 3828 3829 list_splice_tail_init(&rbd_dev->acquiring_list, &rbd_dev->running_list); 3830 } 3831 3832 static int get_lock_owner_info(struct rbd_device *rbd_dev, 3833 struct ceph_locker **lockers, u32 *num_lockers) 3834 { 3835 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 3836 u8 lock_type; 3837 char *lock_tag; 3838 int ret; 3839 3840 dout("%s rbd_dev %p\n", __func__, rbd_dev); 3841 3842 ret = ceph_cls_lock_info(osdc, &rbd_dev->header_oid, 3843 &rbd_dev->header_oloc, RBD_LOCK_NAME, 3844 &lock_type, &lock_tag, lockers, num_lockers); 3845 if (ret) 3846 return ret; 3847 3848 if (*num_lockers == 0) { 3849 dout("%s rbd_dev %p no lockers detected\n", __func__, rbd_dev); 3850 goto out; 3851 } 3852 3853 if (strcmp(lock_tag, RBD_LOCK_TAG)) { 3854 rbd_warn(rbd_dev, "locked by external mechanism, tag %s", 3855 lock_tag); 3856 ret = -EBUSY; 3857 goto out; 3858 } 3859 3860 if (lock_type == CEPH_CLS_LOCK_SHARED) { 3861 rbd_warn(rbd_dev, "shared lock type detected"); 3862 ret = -EBUSY; 3863 goto out; 3864 } 3865 3866 if (strncmp((*lockers)[0].id.cookie, RBD_LOCK_COOKIE_PREFIX, 3867 strlen(RBD_LOCK_COOKIE_PREFIX))) { 3868 rbd_warn(rbd_dev, "locked by external mechanism, cookie %s", 3869 (*lockers)[0].id.cookie); 3870 ret = -EBUSY; 3871 goto out; 3872 } 3873 3874 out: 3875 kfree(lock_tag); 3876 return ret; 3877 } 3878 3879 static int find_watcher(struct rbd_device *rbd_dev, 3880 const struct ceph_locker *locker) 3881 { 3882 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 3883 struct ceph_watch_item *watchers; 3884 u32 num_watchers; 3885 u64 cookie; 3886 int i; 3887 int ret; 3888 3889 ret = ceph_osdc_list_watchers(osdc, &rbd_dev->header_oid, 3890 &rbd_dev->header_oloc, &watchers, 3891 &num_watchers); 3892 if (ret) 3893 return ret; 3894 3895 sscanf(locker->id.cookie, RBD_LOCK_COOKIE_PREFIX " %llu", &cookie); 3896 for (i = 0; i < num_watchers; i++) { 3897 /* 3898 * Ignore addr->type while comparing. This mimics 3899 * entity_addr_t::get_legacy_str() + strcmp(). 3900 */ 3901 if (ceph_addr_equal_no_type(&watchers[i].addr, 3902 &locker->info.addr) && 3903 watchers[i].cookie == cookie) { 3904 struct rbd_client_id cid = { 3905 .gid = le64_to_cpu(watchers[i].name.num), 3906 .handle = cookie, 3907 }; 3908 3909 dout("%s rbd_dev %p found cid %llu-%llu\n", __func__, 3910 rbd_dev, cid.gid, cid.handle); 3911 rbd_set_owner_cid(rbd_dev, &cid); 3912 ret = 1; 3913 goto out; 3914 } 3915 } 3916 3917 dout("%s rbd_dev %p no watchers\n", __func__, rbd_dev); 3918 ret = 0; 3919 out: 3920 kfree(watchers); 3921 return ret; 3922 } 3923 3924 /* 3925 * lock_rwsem must be held for write 3926 */ 3927 static int rbd_try_lock(struct rbd_device *rbd_dev) 3928 { 3929 struct ceph_client *client = rbd_dev->rbd_client->client; 3930 struct ceph_locker *lockers; 3931 u32 num_lockers; 3932 int ret; 3933 3934 for (;;) { 3935 ret = rbd_lock(rbd_dev); 3936 if (ret != -EBUSY) 3937 return ret; 3938 3939 /* determine if the current lock holder is still alive */ 3940 ret = get_lock_owner_info(rbd_dev, &lockers, &num_lockers); 3941 if (ret) 3942 return ret; 3943 3944 if (num_lockers == 0) 3945 goto again; 3946 3947 ret = find_watcher(rbd_dev, lockers); 3948 if (ret) 3949 goto out; /* request lock or error */ 3950 3951 rbd_warn(rbd_dev, "breaking header lock owned by %s%llu", 3952 ENTITY_NAME(lockers[0].id.name)); 3953 3954 ret = ceph_monc_blocklist_add(&client->monc, 3955 &lockers[0].info.addr); 3956 if (ret) { 3957 rbd_warn(rbd_dev, "blocklist of %s%llu failed: %d", 3958 ENTITY_NAME(lockers[0].id.name), ret); 3959 goto out; 3960 } 3961 3962 ret = ceph_cls_break_lock(&client->osdc, &rbd_dev->header_oid, 3963 &rbd_dev->header_oloc, RBD_LOCK_NAME, 3964 lockers[0].id.cookie, 3965 &lockers[0].id.name); 3966 if (ret && ret != -ENOENT) 3967 goto out; 3968 3969 again: 3970 ceph_free_lockers(lockers, num_lockers); 3971 } 3972 3973 out: 3974 ceph_free_lockers(lockers, num_lockers); 3975 return ret; 3976 } 3977 3978 static int rbd_post_acquire_action(struct rbd_device *rbd_dev) 3979 { 3980 int ret; 3981 3982 if (rbd_dev->header.features & RBD_FEATURE_OBJECT_MAP) { 3983 ret = rbd_object_map_open(rbd_dev); 3984 if (ret) 3985 return ret; 3986 } 3987 3988 return 0; 3989 } 3990 3991 /* 3992 * Return: 3993 * 0 - lock acquired 3994 * 1 - caller should call rbd_request_lock() 3995 * <0 - error 3996 */ 3997 static int rbd_try_acquire_lock(struct rbd_device *rbd_dev) 3998 { 3999 int ret; 4000 4001 down_read(&rbd_dev->lock_rwsem); 4002 dout("%s rbd_dev %p read lock_state %d\n", __func__, rbd_dev, 4003 rbd_dev->lock_state); 4004 if (__rbd_is_lock_owner(rbd_dev)) { 4005 up_read(&rbd_dev->lock_rwsem); 4006 return 0; 4007 } 4008 4009 up_read(&rbd_dev->lock_rwsem); 4010 down_write(&rbd_dev->lock_rwsem); 4011 dout("%s rbd_dev %p write lock_state %d\n", __func__, rbd_dev, 4012 rbd_dev->lock_state); 4013 if (__rbd_is_lock_owner(rbd_dev)) { 4014 up_write(&rbd_dev->lock_rwsem); 4015 return 0; 4016 } 4017 4018 ret = rbd_try_lock(rbd_dev); 4019 if (ret < 0) { 4020 rbd_warn(rbd_dev, "failed to lock header: %d", ret); 4021 if (ret == -EBLOCKLISTED) 4022 goto out; 4023 4024 ret = 1; /* request lock anyway */ 4025 } 4026 if (ret > 0) { 4027 up_write(&rbd_dev->lock_rwsem); 4028 return ret; 4029 } 4030 4031 rbd_assert(rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED); 4032 rbd_assert(list_empty(&rbd_dev->running_list)); 4033 4034 ret = rbd_post_acquire_action(rbd_dev); 4035 if (ret) { 4036 rbd_warn(rbd_dev, "post-acquire action failed: %d", ret); 4037 /* 4038 * Can't stay in RBD_LOCK_STATE_LOCKED because 4039 * rbd_lock_add_request() would let the request through, 4040 * assuming that e.g. object map is locked and loaded. 4041 */ 4042 rbd_unlock(rbd_dev); 4043 } 4044 4045 out: 4046 wake_lock_waiters(rbd_dev, ret); 4047 up_write(&rbd_dev->lock_rwsem); 4048 return ret; 4049 } 4050 4051 static void rbd_acquire_lock(struct work_struct *work) 4052 { 4053 struct rbd_device *rbd_dev = container_of(to_delayed_work(work), 4054 struct rbd_device, lock_dwork); 4055 int ret; 4056 4057 dout("%s rbd_dev %p\n", __func__, rbd_dev); 4058 again: 4059 ret = rbd_try_acquire_lock(rbd_dev); 4060 if (ret <= 0) { 4061 dout("%s rbd_dev %p ret %d - done\n", __func__, rbd_dev, ret); 4062 return; 4063 } 4064 4065 ret = rbd_request_lock(rbd_dev); 4066 if (ret == -ETIMEDOUT) { 4067 goto again; /* treat this as a dead client */ 4068 } else if (ret == -EROFS) { 4069 rbd_warn(rbd_dev, "peer will not release lock"); 4070 down_write(&rbd_dev->lock_rwsem); 4071 wake_lock_waiters(rbd_dev, ret); 4072 up_write(&rbd_dev->lock_rwsem); 4073 } else if (ret < 0) { 4074 rbd_warn(rbd_dev, "error requesting lock: %d", ret); 4075 mod_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork, 4076 RBD_RETRY_DELAY); 4077 } else { 4078 /* 4079 * lock owner acked, but resend if we don't see them 4080 * release the lock 4081 */ 4082 dout("%s rbd_dev %p requeuing lock_dwork\n", __func__, 4083 rbd_dev); 4084 mod_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork, 4085 msecs_to_jiffies(2 * RBD_NOTIFY_TIMEOUT * MSEC_PER_SEC)); 4086 } 4087 } 4088 4089 static bool rbd_quiesce_lock(struct rbd_device *rbd_dev) 4090 { 4091 dout("%s rbd_dev %p\n", __func__, rbd_dev); 4092 lockdep_assert_held_write(&rbd_dev->lock_rwsem); 4093 4094 if (rbd_dev->lock_state != RBD_LOCK_STATE_LOCKED) 4095 return false; 4096 4097 /* 4098 * Ensure that all in-flight IO is flushed. 4099 */ 4100 rbd_dev->lock_state = RBD_LOCK_STATE_RELEASING; 4101 rbd_assert(!completion_done(&rbd_dev->releasing_wait)); 4102 if (list_empty(&rbd_dev->running_list)) 4103 return true; 4104 4105 up_write(&rbd_dev->lock_rwsem); 4106 wait_for_completion(&rbd_dev->releasing_wait); 4107 4108 down_write(&rbd_dev->lock_rwsem); 4109 if (rbd_dev->lock_state != RBD_LOCK_STATE_RELEASING) 4110 return false; 4111 4112 rbd_assert(list_empty(&rbd_dev->running_list)); 4113 return true; 4114 } 4115 4116 static void rbd_pre_release_action(struct rbd_device *rbd_dev) 4117 { 4118 if (rbd_dev->header.features & RBD_FEATURE_OBJECT_MAP) 4119 rbd_object_map_close(rbd_dev); 4120 } 4121 4122 static void __rbd_release_lock(struct rbd_device *rbd_dev) 4123 { 4124 rbd_assert(list_empty(&rbd_dev->running_list)); 4125 4126 rbd_pre_release_action(rbd_dev); 4127 rbd_unlock(rbd_dev); 4128 } 4129 4130 /* 4131 * lock_rwsem must be held for write 4132 */ 4133 static void rbd_release_lock(struct rbd_device *rbd_dev) 4134 { 4135 if (!rbd_quiesce_lock(rbd_dev)) 4136 return; 4137 4138 __rbd_release_lock(rbd_dev); 4139 4140 /* 4141 * Give others a chance to grab the lock - we would re-acquire 4142 * almost immediately if we got new IO while draining the running 4143 * list otherwise. We need to ack our own notifications, so this 4144 * lock_dwork will be requeued from rbd_handle_released_lock() by 4145 * way of maybe_kick_acquire(). 4146 */ 4147 cancel_delayed_work(&rbd_dev->lock_dwork); 4148 } 4149 4150 static void rbd_release_lock_work(struct work_struct *work) 4151 { 4152 struct rbd_device *rbd_dev = container_of(work, struct rbd_device, 4153 unlock_work); 4154 4155 down_write(&rbd_dev->lock_rwsem); 4156 rbd_release_lock(rbd_dev); 4157 up_write(&rbd_dev->lock_rwsem); 4158 } 4159 4160 static void maybe_kick_acquire(struct rbd_device *rbd_dev) 4161 { 4162 bool have_requests; 4163 4164 dout("%s rbd_dev %p\n", __func__, rbd_dev); 4165 if (__rbd_is_lock_owner(rbd_dev)) 4166 return; 4167 4168 spin_lock(&rbd_dev->lock_lists_lock); 4169 have_requests = !list_empty(&rbd_dev->acquiring_list); 4170 spin_unlock(&rbd_dev->lock_lists_lock); 4171 if (have_requests || delayed_work_pending(&rbd_dev->lock_dwork)) { 4172 dout("%s rbd_dev %p kicking lock_dwork\n", __func__, rbd_dev); 4173 mod_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork, 0); 4174 } 4175 } 4176 4177 static void rbd_handle_acquired_lock(struct rbd_device *rbd_dev, u8 struct_v, 4178 void **p) 4179 { 4180 struct rbd_client_id cid = { 0 }; 4181 4182 if (struct_v >= 2) { 4183 cid.gid = ceph_decode_64(p); 4184 cid.handle = ceph_decode_64(p); 4185 } 4186 4187 dout("%s rbd_dev %p cid %llu-%llu\n", __func__, rbd_dev, cid.gid, 4188 cid.handle); 4189 if (!rbd_cid_equal(&cid, &rbd_empty_cid)) { 4190 down_write(&rbd_dev->lock_rwsem); 4191 if (rbd_cid_equal(&cid, &rbd_dev->owner_cid)) { 4192 dout("%s rbd_dev %p cid %llu-%llu == owner_cid\n", 4193 __func__, rbd_dev, cid.gid, cid.handle); 4194 } else { 4195 rbd_set_owner_cid(rbd_dev, &cid); 4196 } 4197 downgrade_write(&rbd_dev->lock_rwsem); 4198 } else { 4199 down_read(&rbd_dev->lock_rwsem); 4200 } 4201 4202 maybe_kick_acquire(rbd_dev); 4203 up_read(&rbd_dev->lock_rwsem); 4204 } 4205 4206 static void rbd_handle_released_lock(struct rbd_device *rbd_dev, u8 struct_v, 4207 void **p) 4208 { 4209 struct rbd_client_id cid = { 0 }; 4210 4211 if (struct_v >= 2) { 4212 cid.gid = ceph_decode_64(p); 4213 cid.handle = ceph_decode_64(p); 4214 } 4215 4216 dout("%s rbd_dev %p cid %llu-%llu\n", __func__, rbd_dev, cid.gid, 4217 cid.handle); 4218 if (!rbd_cid_equal(&cid, &rbd_empty_cid)) { 4219 down_write(&rbd_dev->lock_rwsem); 4220 if (!rbd_cid_equal(&cid, &rbd_dev->owner_cid)) { 4221 dout("%s rbd_dev %p cid %llu-%llu != owner_cid %llu-%llu\n", 4222 __func__, rbd_dev, cid.gid, cid.handle, 4223 rbd_dev->owner_cid.gid, rbd_dev->owner_cid.handle); 4224 } else { 4225 rbd_set_owner_cid(rbd_dev, &rbd_empty_cid); 4226 } 4227 downgrade_write(&rbd_dev->lock_rwsem); 4228 } else { 4229 down_read(&rbd_dev->lock_rwsem); 4230 } 4231 4232 maybe_kick_acquire(rbd_dev); 4233 up_read(&rbd_dev->lock_rwsem); 4234 } 4235 4236 /* 4237 * Returns result for ResponseMessage to be encoded (<= 0), or 1 if no 4238 * ResponseMessage is needed. 4239 */ 4240 static int rbd_handle_request_lock(struct rbd_device *rbd_dev, u8 struct_v, 4241 void **p) 4242 { 4243 struct rbd_client_id my_cid = rbd_get_cid(rbd_dev); 4244 struct rbd_client_id cid = { 0 }; 4245 int result = 1; 4246 4247 if (struct_v >= 2) { 4248 cid.gid = ceph_decode_64(p); 4249 cid.handle = ceph_decode_64(p); 4250 } 4251 4252 dout("%s rbd_dev %p cid %llu-%llu\n", __func__, rbd_dev, cid.gid, 4253 cid.handle); 4254 if (rbd_cid_equal(&cid, &my_cid)) 4255 return result; 4256 4257 down_read(&rbd_dev->lock_rwsem); 4258 if (__rbd_is_lock_owner(rbd_dev)) { 4259 if (rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED && 4260 rbd_cid_equal(&rbd_dev->owner_cid, &rbd_empty_cid)) 4261 goto out_unlock; 4262 4263 /* 4264 * encode ResponseMessage(0) so the peer can detect 4265 * a missing owner 4266 */ 4267 result = 0; 4268 4269 if (rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED) { 4270 if (!rbd_dev->opts->exclusive) { 4271 dout("%s rbd_dev %p queueing unlock_work\n", 4272 __func__, rbd_dev); 4273 queue_work(rbd_dev->task_wq, 4274 &rbd_dev->unlock_work); 4275 } else { 4276 /* refuse to release the lock */ 4277 result = -EROFS; 4278 } 4279 } 4280 } 4281 4282 out_unlock: 4283 up_read(&rbd_dev->lock_rwsem); 4284 return result; 4285 } 4286 4287 static void __rbd_acknowledge_notify(struct rbd_device *rbd_dev, 4288 u64 notify_id, u64 cookie, s32 *result) 4289 { 4290 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 4291 char buf[4 + CEPH_ENCODING_START_BLK_LEN]; 4292 int buf_size = sizeof(buf); 4293 int ret; 4294 4295 if (result) { 4296 void *p = buf; 4297 4298 /* encode ResponseMessage */ 4299 ceph_start_encoding(&p, 1, 1, 4300 buf_size - CEPH_ENCODING_START_BLK_LEN); 4301 ceph_encode_32(&p, *result); 4302 } else { 4303 buf_size = 0; 4304 } 4305 4306 ret = ceph_osdc_notify_ack(osdc, &rbd_dev->header_oid, 4307 &rbd_dev->header_oloc, notify_id, cookie, 4308 buf, buf_size); 4309 if (ret) 4310 rbd_warn(rbd_dev, "acknowledge_notify failed: %d", ret); 4311 } 4312 4313 static void rbd_acknowledge_notify(struct rbd_device *rbd_dev, u64 notify_id, 4314 u64 cookie) 4315 { 4316 dout("%s rbd_dev %p\n", __func__, rbd_dev); 4317 __rbd_acknowledge_notify(rbd_dev, notify_id, cookie, NULL); 4318 } 4319 4320 static void rbd_acknowledge_notify_result(struct rbd_device *rbd_dev, 4321 u64 notify_id, u64 cookie, s32 result) 4322 { 4323 dout("%s rbd_dev %p result %d\n", __func__, rbd_dev, result); 4324 __rbd_acknowledge_notify(rbd_dev, notify_id, cookie, &result); 4325 } 4326 4327 static void rbd_watch_cb(void *arg, u64 notify_id, u64 cookie, 4328 u64 notifier_id, void *data, size_t data_len) 4329 { 4330 struct rbd_device *rbd_dev = arg; 4331 void *p = data; 4332 void *const end = p + data_len; 4333 u8 struct_v = 0; 4334 u32 len; 4335 u32 notify_op; 4336 int ret; 4337 4338 dout("%s rbd_dev %p cookie %llu notify_id %llu data_len %zu\n", 4339 __func__, rbd_dev, cookie, notify_id, data_len); 4340 if (data_len) { 4341 ret = ceph_start_decoding(&p, end, 1, "NotifyMessage", 4342 &struct_v, &len); 4343 if (ret) { 4344 rbd_warn(rbd_dev, "failed to decode NotifyMessage: %d", 4345 ret); 4346 return; 4347 } 4348 4349 notify_op = ceph_decode_32(&p); 4350 } else { 4351 /* legacy notification for header updates */ 4352 notify_op = RBD_NOTIFY_OP_HEADER_UPDATE; 4353 len = 0; 4354 } 4355 4356 dout("%s rbd_dev %p notify_op %u\n", __func__, rbd_dev, notify_op); 4357 switch (notify_op) { 4358 case RBD_NOTIFY_OP_ACQUIRED_LOCK: 4359 rbd_handle_acquired_lock(rbd_dev, struct_v, &p); 4360 rbd_acknowledge_notify(rbd_dev, notify_id, cookie); 4361 break; 4362 case RBD_NOTIFY_OP_RELEASED_LOCK: 4363 rbd_handle_released_lock(rbd_dev, struct_v, &p); 4364 rbd_acknowledge_notify(rbd_dev, notify_id, cookie); 4365 break; 4366 case RBD_NOTIFY_OP_REQUEST_LOCK: 4367 ret = rbd_handle_request_lock(rbd_dev, struct_v, &p); 4368 if (ret <= 0) 4369 rbd_acknowledge_notify_result(rbd_dev, notify_id, 4370 cookie, ret); 4371 else 4372 rbd_acknowledge_notify(rbd_dev, notify_id, cookie); 4373 break; 4374 case RBD_NOTIFY_OP_HEADER_UPDATE: 4375 ret = rbd_dev_refresh(rbd_dev); 4376 if (ret) 4377 rbd_warn(rbd_dev, "refresh failed: %d", ret); 4378 4379 rbd_acknowledge_notify(rbd_dev, notify_id, cookie); 4380 break; 4381 default: 4382 if (rbd_is_lock_owner(rbd_dev)) 4383 rbd_acknowledge_notify_result(rbd_dev, notify_id, 4384 cookie, -EOPNOTSUPP); 4385 else 4386 rbd_acknowledge_notify(rbd_dev, notify_id, cookie); 4387 break; 4388 } 4389 } 4390 4391 static void __rbd_unregister_watch(struct rbd_device *rbd_dev); 4392 4393 static void rbd_watch_errcb(void *arg, u64 cookie, int err) 4394 { 4395 struct rbd_device *rbd_dev = arg; 4396 4397 rbd_warn(rbd_dev, "encountered watch error: %d", err); 4398 4399 down_write(&rbd_dev->lock_rwsem); 4400 rbd_set_owner_cid(rbd_dev, &rbd_empty_cid); 4401 up_write(&rbd_dev->lock_rwsem); 4402 4403 mutex_lock(&rbd_dev->watch_mutex); 4404 if (rbd_dev->watch_state == RBD_WATCH_STATE_REGISTERED) { 4405 __rbd_unregister_watch(rbd_dev); 4406 rbd_dev->watch_state = RBD_WATCH_STATE_ERROR; 4407 4408 queue_delayed_work(rbd_dev->task_wq, &rbd_dev->watch_dwork, 0); 4409 } 4410 mutex_unlock(&rbd_dev->watch_mutex); 4411 } 4412 4413 /* 4414 * watch_mutex must be locked 4415 */ 4416 static int __rbd_register_watch(struct rbd_device *rbd_dev) 4417 { 4418 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 4419 struct ceph_osd_linger_request *handle; 4420 4421 rbd_assert(!rbd_dev->watch_handle); 4422 dout("%s rbd_dev %p\n", __func__, rbd_dev); 4423 4424 handle = ceph_osdc_watch(osdc, &rbd_dev->header_oid, 4425 &rbd_dev->header_oloc, rbd_watch_cb, 4426 rbd_watch_errcb, rbd_dev); 4427 if (IS_ERR(handle)) 4428 return PTR_ERR(handle); 4429 4430 rbd_dev->watch_handle = handle; 4431 return 0; 4432 } 4433 4434 /* 4435 * watch_mutex must be locked 4436 */ 4437 static void __rbd_unregister_watch(struct rbd_device *rbd_dev) 4438 { 4439 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 4440 int ret; 4441 4442 rbd_assert(rbd_dev->watch_handle); 4443 dout("%s rbd_dev %p\n", __func__, rbd_dev); 4444 4445 ret = ceph_osdc_unwatch(osdc, rbd_dev->watch_handle); 4446 if (ret) 4447 rbd_warn(rbd_dev, "failed to unwatch: %d", ret); 4448 4449 rbd_dev->watch_handle = NULL; 4450 } 4451 4452 static int rbd_register_watch(struct rbd_device *rbd_dev) 4453 { 4454 int ret; 4455 4456 mutex_lock(&rbd_dev->watch_mutex); 4457 rbd_assert(rbd_dev->watch_state == RBD_WATCH_STATE_UNREGISTERED); 4458 ret = __rbd_register_watch(rbd_dev); 4459 if (ret) 4460 goto out; 4461 4462 rbd_dev->watch_state = RBD_WATCH_STATE_REGISTERED; 4463 rbd_dev->watch_cookie = rbd_dev->watch_handle->linger_id; 4464 4465 out: 4466 mutex_unlock(&rbd_dev->watch_mutex); 4467 return ret; 4468 } 4469 4470 static void cancel_tasks_sync(struct rbd_device *rbd_dev) 4471 { 4472 dout("%s rbd_dev %p\n", __func__, rbd_dev); 4473 4474 cancel_work_sync(&rbd_dev->acquired_lock_work); 4475 cancel_work_sync(&rbd_dev->released_lock_work); 4476 cancel_delayed_work_sync(&rbd_dev->lock_dwork); 4477 cancel_work_sync(&rbd_dev->unlock_work); 4478 } 4479 4480 /* 4481 * header_rwsem must not be held to avoid a deadlock with 4482 * rbd_dev_refresh() when flushing notifies. 4483 */ 4484 static void rbd_unregister_watch(struct rbd_device *rbd_dev) 4485 { 4486 cancel_tasks_sync(rbd_dev); 4487 4488 mutex_lock(&rbd_dev->watch_mutex); 4489 if (rbd_dev->watch_state == RBD_WATCH_STATE_REGISTERED) 4490 __rbd_unregister_watch(rbd_dev); 4491 rbd_dev->watch_state = RBD_WATCH_STATE_UNREGISTERED; 4492 mutex_unlock(&rbd_dev->watch_mutex); 4493 4494 cancel_delayed_work_sync(&rbd_dev->watch_dwork); 4495 ceph_osdc_flush_notifies(&rbd_dev->rbd_client->client->osdc); 4496 } 4497 4498 /* 4499 * lock_rwsem must be held for write 4500 */ 4501 static void rbd_reacquire_lock(struct rbd_device *rbd_dev) 4502 { 4503 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 4504 char cookie[32]; 4505 int ret; 4506 4507 if (!rbd_quiesce_lock(rbd_dev)) 4508 return; 4509 4510 format_lock_cookie(rbd_dev, cookie); 4511 ret = ceph_cls_set_cookie(osdc, &rbd_dev->header_oid, 4512 &rbd_dev->header_oloc, RBD_LOCK_NAME, 4513 CEPH_CLS_LOCK_EXCLUSIVE, rbd_dev->lock_cookie, 4514 RBD_LOCK_TAG, cookie); 4515 if (ret) { 4516 if (ret != -EOPNOTSUPP) 4517 rbd_warn(rbd_dev, "failed to update lock cookie: %d", 4518 ret); 4519 4520 /* 4521 * Lock cookie cannot be updated on older OSDs, so do 4522 * a manual release and queue an acquire. 4523 */ 4524 __rbd_release_lock(rbd_dev); 4525 queue_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork, 0); 4526 } else { 4527 __rbd_lock(rbd_dev, cookie); 4528 wake_lock_waiters(rbd_dev, 0); 4529 } 4530 } 4531 4532 static void rbd_reregister_watch(struct work_struct *work) 4533 { 4534 struct rbd_device *rbd_dev = container_of(to_delayed_work(work), 4535 struct rbd_device, watch_dwork); 4536 int ret; 4537 4538 dout("%s rbd_dev %p\n", __func__, rbd_dev); 4539 4540 mutex_lock(&rbd_dev->watch_mutex); 4541 if (rbd_dev->watch_state != RBD_WATCH_STATE_ERROR) { 4542 mutex_unlock(&rbd_dev->watch_mutex); 4543 return; 4544 } 4545 4546 ret = __rbd_register_watch(rbd_dev); 4547 if (ret) { 4548 rbd_warn(rbd_dev, "failed to reregister watch: %d", ret); 4549 if (ret != -EBLOCKLISTED && ret != -ENOENT) { 4550 queue_delayed_work(rbd_dev->task_wq, 4551 &rbd_dev->watch_dwork, 4552 RBD_RETRY_DELAY); 4553 mutex_unlock(&rbd_dev->watch_mutex); 4554 return; 4555 } 4556 4557 mutex_unlock(&rbd_dev->watch_mutex); 4558 down_write(&rbd_dev->lock_rwsem); 4559 wake_lock_waiters(rbd_dev, ret); 4560 up_write(&rbd_dev->lock_rwsem); 4561 return; 4562 } 4563 4564 rbd_dev->watch_state = RBD_WATCH_STATE_REGISTERED; 4565 rbd_dev->watch_cookie = rbd_dev->watch_handle->linger_id; 4566 mutex_unlock(&rbd_dev->watch_mutex); 4567 4568 down_write(&rbd_dev->lock_rwsem); 4569 if (rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED) 4570 rbd_reacquire_lock(rbd_dev); 4571 up_write(&rbd_dev->lock_rwsem); 4572 4573 ret = rbd_dev_refresh(rbd_dev); 4574 if (ret) 4575 rbd_warn(rbd_dev, "reregistration refresh failed: %d", ret); 4576 } 4577 4578 /* 4579 * Synchronous osd object method call. Returns the number of bytes 4580 * returned in the outbound buffer, or a negative error code. 4581 */ 4582 static int rbd_obj_method_sync(struct rbd_device *rbd_dev, 4583 struct ceph_object_id *oid, 4584 struct ceph_object_locator *oloc, 4585 const char *method_name, 4586 const void *outbound, 4587 size_t outbound_size, 4588 void *inbound, 4589 size_t inbound_size) 4590 { 4591 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 4592 struct page *req_page = NULL; 4593 struct page *reply_page; 4594 int ret; 4595 4596 /* 4597 * Method calls are ultimately read operations. The result 4598 * should placed into the inbound buffer provided. They 4599 * also supply outbound data--parameters for the object 4600 * method. Currently if this is present it will be a 4601 * snapshot id. 4602 */ 4603 if (outbound) { 4604 if (outbound_size > PAGE_SIZE) 4605 return -E2BIG; 4606 4607 req_page = alloc_page(GFP_KERNEL); 4608 if (!req_page) 4609 return -ENOMEM; 4610 4611 memcpy(page_address(req_page), outbound, outbound_size); 4612 } 4613 4614 reply_page = alloc_page(GFP_KERNEL); 4615 if (!reply_page) { 4616 if (req_page) 4617 __free_page(req_page); 4618 return -ENOMEM; 4619 } 4620 4621 ret = ceph_osdc_call(osdc, oid, oloc, RBD_DRV_NAME, method_name, 4622 CEPH_OSD_FLAG_READ, req_page, outbound_size, 4623 &reply_page, &inbound_size); 4624 if (!ret) { 4625 memcpy(inbound, page_address(reply_page), inbound_size); 4626 ret = inbound_size; 4627 } 4628 4629 if (req_page) 4630 __free_page(req_page); 4631 __free_page(reply_page); 4632 return ret; 4633 } 4634 4635 static void rbd_queue_workfn(struct work_struct *work) 4636 { 4637 struct rbd_img_request *img_request = 4638 container_of(work, struct rbd_img_request, work); 4639 struct rbd_device *rbd_dev = img_request->rbd_dev; 4640 enum obj_operation_type op_type = img_request->op_type; 4641 struct request *rq = blk_mq_rq_from_pdu(img_request); 4642 u64 offset = (u64)blk_rq_pos(rq) << SECTOR_SHIFT; 4643 u64 length = blk_rq_bytes(rq); 4644 u64 mapping_size; 4645 int result; 4646 4647 /* Ignore/skip any zero-length requests */ 4648 if (!length) { 4649 dout("%s: zero-length request\n", __func__); 4650 result = 0; 4651 goto err_img_request; 4652 } 4653 4654 blk_mq_start_request(rq); 4655 4656 down_read(&rbd_dev->header_rwsem); 4657 mapping_size = rbd_dev->mapping.size; 4658 rbd_img_capture_header(img_request); 4659 up_read(&rbd_dev->header_rwsem); 4660 4661 if (offset + length > mapping_size) { 4662 rbd_warn(rbd_dev, "beyond EOD (%llu~%llu > %llu)", offset, 4663 length, mapping_size); 4664 result = -EIO; 4665 goto err_img_request; 4666 } 4667 4668 dout("%s rbd_dev %p img_req %p %s %llu~%llu\n", __func__, rbd_dev, 4669 img_request, obj_op_name(op_type), offset, length); 4670 4671 if (op_type == OBJ_OP_DISCARD || op_type == OBJ_OP_ZEROOUT) 4672 result = rbd_img_fill_nodata(img_request, offset, length); 4673 else 4674 result = rbd_img_fill_from_bio(img_request, offset, length, 4675 rq->bio); 4676 if (result) 4677 goto err_img_request; 4678 4679 rbd_img_handle_request(img_request, 0); 4680 return; 4681 4682 err_img_request: 4683 rbd_img_request_destroy(img_request); 4684 if (result) 4685 rbd_warn(rbd_dev, "%s %llx at %llx result %d", 4686 obj_op_name(op_type), length, offset, result); 4687 blk_mq_end_request(rq, errno_to_blk_status(result)); 4688 } 4689 4690 static blk_status_t rbd_queue_rq(struct blk_mq_hw_ctx *hctx, 4691 const struct blk_mq_queue_data *bd) 4692 { 4693 struct rbd_device *rbd_dev = hctx->queue->queuedata; 4694 struct rbd_img_request *img_req = blk_mq_rq_to_pdu(bd->rq); 4695 enum obj_operation_type op_type; 4696 4697 switch (req_op(bd->rq)) { 4698 case REQ_OP_DISCARD: 4699 op_type = OBJ_OP_DISCARD; 4700 break; 4701 case REQ_OP_WRITE_ZEROES: 4702 op_type = OBJ_OP_ZEROOUT; 4703 break; 4704 case REQ_OP_WRITE: 4705 op_type = OBJ_OP_WRITE; 4706 break; 4707 case REQ_OP_READ: 4708 op_type = OBJ_OP_READ; 4709 break; 4710 default: 4711 rbd_warn(rbd_dev, "unknown req_op %d", req_op(bd->rq)); 4712 return BLK_STS_IOERR; 4713 } 4714 4715 rbd_img_request_init(img_req, rbd_dev, op_type); 4716 4717 if (rbd_img_is_write(img_req)) { 4718 if (rbd_is_ro(rbd_dev)) { 4719 rbd_warn(rbd_dev, "%s on read-only mapping", 4720 obj_op_name(img_req->op_type)); 4721 return BLK_STS_IOERR; 4722 } 4723 rbd_assert(!rbd_is_snap(rbd_dev)); 4724 } 4725 4726 INIT_WORK(&img_req->work, rbd_queue_workfn); 4727 queue_work(rbd_wq, &img_req->work); 4728 return BLK_STS_OK; 4729 } 4730 4731 static void rbd_free_disk(struct rbd_device *rbd_dev) 4732 { 4733 blk_cleanup_disk(rbd_dev->disk); 4734 blk_mq_free_tag_set(&rbd_dev->tag_set); 4735 rbd_dev->disk = NULL; 4736 } 4737 4738 static int rbd_obj_read_sync(struct rbd_device *rbd_dev, 4739 struct ceph_object_id *oid, 4740 struct ceph_object_locator *oloc, 4741 void *buf, int buf_len) 4742 4743 { 4744 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 4745 struct ceph_osd_request *req; 4746 struct page **pages; 4747 int num_pages = calc_pages_for(0, buf_len); 4748 int ret; 4749 4750 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_KERNEL); 4751 if (!req) 4752 return -ENOMEM; 4753 4754 ceph_oid_copy(&req->r_base_oid, oid); 4755 ceph_oloc_copy(&req->r_base_oloc, oloc); 4756 req->r_flags = CEPH_OSD_FLAG_READ; 4757 4758 pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL); 4759 if (IS_ERR(pages)) { 4760 ret = PTR_ERR(pages); 4761 goto out_req; 4762 } 4763 4764 osd_req_op_extent_init(req, 0, CEPH_OSD_OP_READ, 0, buf_len, 0, 0); 4765 osd_req_op_extent_osd_data_pages(req, 0, pages, buf_len, 0, false, 4766 true); 4767 4768 ret = ceph_osdc_alloc_messages(req, GFP_KERNEL); 4769 if (ret) 4770 goto out_req; 4771 4772 ceph_osdc_start_request(osdc, req, false); 4773 ret = ceph_osdc_wait_request(osdc, req); 4774 if (ret >= 0) 4775 ceph_copy_from_page_vector(pages, buf, 0, ret); 4776 4777 out_req: 4778 ceph_osdc_put_request(req); 4779 return ret; 4780 } 4781 4782 /* 4783 * Read the complete header for the given rbd device. On successful 4784 * return, the rbd_dev->header field will contain up-to-date 4785 * information about the image. 4786 */ 4787 static int rbd_dev_v1_header_info(struct rbd_device *rbd_dev) 4788 { 4789 struct rbd_image_header_ondisk *ondisk = NULL; 4790 u32 snap_count = 0; 4791 u64 names_size = 0; 4792 u32 want_count; 4793 int ret; 4794 4795 /* 4796 * The complete header will include an array of its 64-bit 4797 * snapshot ids, followed by the names of those snapshots as 4798 * a contiguous block of NUL-terminated strings. Note that 4799 * the number of snapshots could change by the time we read 4800 * it in, in which case we re-read it. 4801 */ 4802 do { 4803 size_t size; 4804 4805 kfree(ondisk); 4806 4807 size = sizeof (*ondisk); 4808 size += snap_count * sizeof (struct rbd_image_snap_ondisk); 4809 size += names_size; 4810 ondisk = kmalloc(size, GFP_KERNEL); 4811 if (!ondisk) 4812 return -ENOMEM; 4813 4814 ret = rbd_obj_read_sync(rbd_dev, &rbd_dev->header_oid, 4815 &rbd_dev->header_oloc, ondisk, size); 4816 if (ret < 0) 4817 goto out; 4818 if ((size_t)ret < size) { 4819 ret = -ENXIO; 4820 rbd_warn(rbd_dev, "short header read (want %zd got %d)", 4821 size, ret); 4822 goto out; 4823 } 4824 if (!rbd_dev_ondisk_valid(ondisk)) { 4825 ret = -ENXIO; 4826 rbd_warn(rbd_dev, "invalid header"); 4827 goto out; 4828 } 4829 4830 names_size = le64_to_cpu(ondisk->snap_names_len); 4831 want_count = snap_count; 4832 snap_count = le32_to_cpu(ondisk->snap_count); 4833 } while (snap_count != want_count); 4834 4835 ret = rbd_header_from_disk(rbd_dev, ondisk); 4836 out: 4837 kfree(ondisk); 4838 4839 return ret; 4840 } 4841 4842 static void rbd_dev_update_size(struct rbd_device *rbd_dev) 4843 { 4844 sector_t size; 4845 4846 /* 4847 * If EXISTS is not set, rbd_dev->disk may be NULL, so don't 4848 * try to update its size. If REMOVING is set, updating size 4849 * is just useless work since the device can't be opened. 4850 */ 4851 if (test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags) && 4852 !test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags)) { 4853 size = (sector_t)rbd_dev->mapping.size / SECTOR_SIZE; 4854 dout("setting size to %llu sectors", (unsigned long long)size); 4855 set_capacity_and_notify(rbd_dev->disk, size); 4856 } 4857 } 4858 4859 static int rbd_dev_refresh(struct rbd_device *rbd_dev) 4860 { 4861 u64 mapping_size; 4862 int ret; 4863 4864 down_write(&rbd_dev->header_rwsem); 4865 mapping_size = rbd_dev->mapping.size; 4866 4867 ret = rbd_dev_header_info(rbd_dev); 4868 if (ret) 4869 goto out; 4870 4871 /* 4872 * If there is a parent, see if it has disappeared due to the 4873 * mapped image getting flattened. 4874 */ 4875 if (rbd_dev->parent) { 4876 ret = rbd_dev_v2_parent_info(rbd_dev); 4877 if (ret) 4878 goto out; 4879 } 4880 4881 rbd_assert(!rbd_is_snap(rbd_dev)); 4882 rbd_dev->mapping.size = rbd_dev->header.image_size; 4883 4884 out: 4885 up_write(&rbd_dev->header_rwsem); 4886 if (!ret && mapping_size != rbd_dev->mapping.size) 4887 rbd_dev_update_size(rbd_dev); 4888 4889 return ret; 4890 } 4891 4892 static const struct blk_mq_ops rbd_mq_ops = { 4893 .queue_rq = rbd_queue_rq, 4894 }; 4895 4896 static int rbd_init_disk(struct rbd_device *rbd_dev) 4897 { 4898 struct gendisk *disk; 4899 struct request_queue *q; 4900 unsigned int objset_bytes = 4901 rbd_dev->layout.object_size * rbd_dev->layout.stripe_count; 4902 int err; 4903 4904 memset(&rbd_dev->tag_set, 0, sizeof(rbd_dev->tag_set)); 4905 rbd_dev->tag_set.ops = &rbd_mq_ops; 4906 rbd_dev->tag_set.queue_depth = rbd_dev->opts->queue_depth; 4907 rbd_dev->tag_set.numa_node = NUMA_NO_NODE; 4908 rbd_dev->tag_set.flags = BLK_MQ_F_SHOULD_MERGE; 4909 rbd_dev->tag_set.nr_hw_queues = num_present_cpus(); 4910 rbd_dev->tag_set.cmd_size = sizeof(struct rbd_img_request); 4911 4912 err = blk_mq_alloc_tag_set(&rbd_dev->tag_set); 4913 if (err) 4914 return err; 4915 4916 disk = blk_mq_alloc_disk(&rbd_dev->tag_set, rbd_dev); 4917 if (IS_ERR(disk)) { 4918 err = PTR_ERR(disk); 4919 goto out_tag_set; 4920 } 4921 q = disk->queue; 4922 4923 snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", 4924 rbd_dev->dev_id); 4925 disk->major = rbd_dev->major; 4926 disk->first_minor = rbd_dev->minor; 4927 if (single_major) 4928 disk->minors = (1 << RBD_SINGLE_MAJOR_PART_SHIFT); 4929 else 4930 disk->minors = RBD_MINORS_PER_MAJOR; 4931 disk->fops = &rbd_bd_ops; 4932 disk->private_data = rbd_dev; 4933 4934 blk_queue_flag_set(QUEUE_FLAG_NONROT, q); 4935 /* QUEUE_FLAG_ADD_RANDOM is off by default for blk-mq */ 4936 4937 blk_queue_max_hw_sectors(q, objset_bytes >> SECTOR_SHIFT); 4938 q->limits.max_sectors = queue_max_hw_sectors(q); 4939 blk_queue_max_segments(q, USHRT_MAX); 4940 blk_queue_max_segment_size(q, UINT_MAX); 4941 blk_queue_io_min(q, rbd_dev->opts->alloc_size); 4942 blk_queue_io_opt(q, rbd_dev->opts->alloc_size); 4943 4944 if (rbd_dev->opts->trim) { 4945 blk_queue_flag_set(QUEUE_FLAG_DISCARD, q); 4946 q->limits.discard_granularity = rbd_dev->opts->alloc_size; 4947 blk_queue_max_discard_sectors(q, objset_bytes >> SECTOR_SHIFT); 4948 blk_queue_max_write_zeroes_sectors(q, objset_bytes >> SECTOR_SHIFT); 4949 } 4950 4951 if (!ceph_test_opt(rbd_dev->rbd_client->client, NOCRC)) 4952 blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, q); 4953 4954 rbd_dev->disk = disk; 4955 4956 return 0; 4957 out_tag_set: 4958 blk_mq_free_tag_set(&rbd_dev->tag_set); 4959 return err; 4960 } 4961 4962 /* 4963 sysfs 4964 */ 4965 4966 static struct rbd_device *dev_to_rbd_dev(struct device *dev) 4967 { 4968 return container_of(dev, struct rbd_device, dev); 4969 } 4970 4971 static ssize_t rbd_size_show(struct device *dev, 4972 struct device_attribute *attr, char *buf) 4973 { 4974 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 4975 4976 return sprintf(buf, "%llu\n", 4977 (unsigned long long)rbd_dev->mapping.size); 4978 } 4979 4980 static ssize_t rbd_features_show(struct device *dev, 4981 struct device_attribute *attr, char *buf) 4982 { 4983 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 4984 4985 return sprintf(buf, "0x%016llx\n", rbd_dev->header.features); 4986 } 4987 4988 static ssize_t rbd_major_show(struct device *dev, 4989 struct device_attribute *attr, char *buf) 4990 { 4991 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 4992 4993 if (rbd_dev->major) 4994 return sprintf(buf, "%d\n", rbd_dev->major); 4995 4996 return sprintf(buf, "(none)\n"); 4997 } 4998 4999 static ssize_t rbd_minor_show(struct device *dev, 5000 struct device_attribute *attr, char *buf) 5001 { 5002 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5003 5004 return sprintf(buf, "%d\n", rbd_dev->minor); 5005 } 5006 5007 static ssize_t rbd_client_addr_show(struct device *dev, 5008 struct device_attribute *attr, char *buf) 5009 { 5010 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5011 struct ceph_entity_addr *client_addr = 5012 ceph_client_addr(rbd_dev->rbd_client->client); 5013 5014 return sprintf(buf, "%pISpc/%u\n", &client_addr->in_addr, 5015 le32_to_cpu(client_addr->nonce)); 5016 } 5017 5018 static ssize_t rbd_client_id_show(struct device *dev, 5019 struct device_attribute *attr, char *buf) 5020 { 5021 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5022 5023 return sprintf(buf, "client%lld\n", 5024 ceph_client_gid(rbd_dev->rbd_client->client)); 5025 } 5026 5027 static ssize_t rbd_cluster_fsid_show(struct device *dev, 5028 struct device_attribute *attr, char *buf) 5029 { 5030 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5031 5032 return sprintf(buf, "%pU\n", &rbd_dev->rbd_client->client->fsid); 5033 } 5034 5035 static ssize_t rbd_config_info_show(struct device *dev, 5036 struct device_attribute *attr, char *buf) 5037 { 5038 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5039 5040 if (!capable(CAP_SYS_ADMIN)) 5041 return -EPERM; 5042 5043 return sprintf(buf, "%s\n", rbd_dev->config_info); 5044 } 5045 5046 static ssize_t rbd_pool_show(struct device *dev, 5047 struct device_attribute *attr, char *buf) 5048 { 5049 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5050 5051 return sprintf(buf, "%s\n", rbd_dev->spec->pool_name); 5052 } 5053 5054 static ssize_t rbd_pool_id_show(struct device *dev, 5055 struct device_attribute *attr, char *buf) 5056 { 5057 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5058 5059 return sprintf(buf, "%llu\n", 5060 (unsigned long long) rbd_dev->spec->pool_id); 5061 } 5062 5063 static ssize_t rbd_pool_ns_show(struct device *dev, 5064 struct device_attribute *attr, char *buf) 5065 { 5066 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5067 5068 return sprintf(buf, "%s\n", rbd_dev->spec->pool_ns ?: ""); 5069 } 5070 5071 static ssize_t rbd_name_show(struct device *dev, 5072 struct device_attribute *attr, char *buf) 5073 { 5074 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5075 5076 if (rbd_dev->spec->image_name) 5077 return sprintf(buf, "%s\n", rbd_dev->spec->image_name); 5078 5079 return sprintf(buf, "(unknown)\n"); 5080 } 5081 5082 static ssize_t rbd_image_id_show(struct device *dev, 5083 struct device_attribute *attr, char *buf) 5084 { 5085 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5086 5087 return sprintf(buf, "%s\n", rbd_dev->spec->image_id); 5088 } 5089 5090 /* 5091 * Shows the name of the currently-mapped snapshot (or 5092 * RBD_SNAP_HEAD_NAME for the base image). 5093 */ 5094 static ssize_t rbd_snap_show(struct device *dev, 5095 struct device_attribute *attr, 5096 char *buf) 5097 { 5098 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5099 5100 return sprintf(buf, "%s\n", rbd_dev->spec->snap_name); 5101 } 5102 5103 static ssize_t rbd_snap_id_show(struct device *dev, 5104 struct device_attribute *attr, char *buf) 5105 { 5106 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5107 5108 return sprintf(buf, "%llu\n", rbd_dev->spec->snap_id); 5109 } 5110 5111 /* 5112 * For a v2 image, shows the chain of parent images, separated by empty 5113 * lines. For v1 images or if there is no parent, shows "(no parent 5114 * image)". 5115 */ 5116 static ssize_t rbd_parent_show(struct device *dev, 5117 struct device_attribute *attr, 5118 char *buf) 5119 { 5120 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5121 ssize_t count = 0; 5122 5123 if (!rbd_dev->parent) 5124 return sprintf(buf, "(no parent image)\n"); 5125 5126 for ( ; rbd_dev->parent; rbd_dev = rbd_dev->parent) { 5127 struct rbd_spec *spec = rbd_dev->parent_spec; 5128 5129 count += sprintf(&buf[count], "%s" 5130 "pool_id %llu\npool_name %s\n" 5131 "pool_ns %s\n" 5132 "image_id %s\nimage_name %s\n" 5133 "snap_id %llu\nsnap_name %s\n" 5134 "overlap %llu\n", 5135 !count ? "" : "\n", /* first? */ 5136 spec->pool_id, spec->pool_name, 5137 spec->pool_ns ?: "", 5138 spec->image_id, spec->image_name ?: "(unknown)", 5139 spec->snap_id, spec->snap_name, 5140 rbd_dev->parent_overlap); 5141 } 5142 5143 return count; 5144 } 5145 5146 static ssize_t rbd_image_refresh(struct device *dev, 5147 struct device_attribute *attr, 5148 const char *buf, 5149 size_t size) 5150 { 5151 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5152 int ret; 5153 5154 if (!capable(CAP_SYS_ADMIN)) 5155 return -EPERM; 5156 5157 ret = rbd_dev_refresh(rbd_dev); 5158 if (ret) 5159 return ret; 5160 5161 return size; 5162 } 5163 5164 static DEVICE_ATTR(size, 0444, rbd_size_show, NULL); 5165 static DEVICE_ATTR(features, 0444, rbd_features_show, NULL); 5166 static DEVICE_ATTR(major, 0444, rbd_major_show, NULL); 5167 static DEVICE_ATTR(minor, 0444, rbd_minor_show, NULL); 5168 static DEVICE_ATTR(client_addr, 0444, rbd_client_addr_show, NULL); 5169 static DEVICE_ATTR(client_id, 0444, rbd_client_id_show, NULL); 5170 static DEVICE_ATTR(cluster_fsid, 0444, rbd_cluster_fsid_show, NULL); 5171 static DEVICE_ATTR(config_info, 0400, rbd_config_info_show, NULL); 5172 static DEVICE_ATTR(pool, 0444, rbd_pool_show, NULL); 5173 static DEVICE_ATTR(pool_id, 0444, rbd_pool_id_show, NULL); 5174 static DEVICE_ATTR(pool_ns, 0444, rbd_pool_ns_show, NULL); 5175 static DEVICE_ATTR(name, 0444, rbd_name_show, NULL); 5176 static DEVICE_ATTR(image_id, 0444, rbd_image_id_show, NULL); 5177 static DEVICE_ATTR(refresh, 0200, NULL, rbd_image_refresh); 5178 static DEVICE_ATTR(current_snap, 0444, rbd_snap_show, NULL); 5179 static DEVICE_ATTR(snap_id, 0444, rbd_snap_id_show, NULL); 5180 static DEVICE_ATTR(parent, 0444, rbd_parent_show, NULL); 5181 5182 static struct attribute *rbd_attrs[] = { 5183 &dev_attr_size.attr, 5184 &dev_attr_features.attr, 5185 &dev_attr_major.attr, 5186 &dev_attr_minor.attr, 5187 &dev_attr_client_addr.attr, 5188 &dev_attr_client_id.attr, 5189 &dev_attr_cluster_fsid.attr, 5190 &dev_attr_config_info.attr, 5191 &dev_attr_pool.attr, 5192 &dev_attr_pool_id.attr, 5193 &dev_attr_pool_ns.attr, 5194 &dev_attr_name.attr, 5195 &dev_attr_image_id.attr, 5196 &dev_attr_current_snap.attr, 5197 &dev_attr_snap_id.attr, 5198 &dev_attr_parent.attr, 5199 &dev_attr_refresh.attr, 5200 NULL 5201 }; 5202 5203 static struct attribute_group rbd_attr_group = { 5204 .attrs = rbd_attrs, 5205 }; 5206 5207 static const struct attribute_group *rbd_attr_groups[] = { 5208 &rbd_attr_group, 5209 NULL 5210 }; 5211 5212 static void rbd_dev_release(struct device *dev); 5213 5214 static const struct device_type rbd_device_type = { 5215 .name = "rbd", 5216 .groups = rbd_attr_groups, 5217 .release = rbd_dev_release, 5218 }; 5219 5220 static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec) 5221 { 5222 kref_get(&spec->kref); 5223 5224 return spec; 5225 } 5226 5227 static void rbd_spec_free(struct kref *kref); 5228 static void rbd_spec_put(struct rbd_spec *spec) 5229 { 5230 if (spec) 5231 kref_put(&spec->kref, rbd_spec_free); 5232 } 5233 5234 static struct rbd_spec *rbd_spec_alloc(void) 5235 { 5236 struct rbd_spec *spec; 5237 5238 spec = kzalloc(sizeof (*spec), GFP_KERNEL); 5239 if (!spec) 5240 return NULL; 5241 5242 spec->pool_id = CEPH_NOPOOL; 5243 spec->snap_id = CEPH_NOSNAP; 5244 kref_init(&spec->kref); 5245 5246 return spec; 5247 } 5248 5249 static void rbd_spec_free(struct kref *kref) 5250 { 5251 struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref); 5252 5253 kfree(spec->pool_name); 5254 kfree(spec->pool_ns); 5255 kfree(spec->image_id); 5256 kfree(spec->image_name); 5257 kfree(spec->snap_name); 5258 kfree(spec); 5259 } 5260 5261 static void rbd_dev_free(struct rbd_device *rbd_dev) 5262 { 5263 WARN_ON(rbd_dev->watch_state != RBD_WATCH_STATE_UNREGISTERED); 5264 WARN_ON(rbd_dev->lock_state != RBD_LOCK_STATE_UNLOCKED); 5265 5266 ceph_oid_destroy(&rbd_dev->header_oid); 5267 ceph_oloc_destroy(&rbd_dev->header_oloc); 5268 kfree(rbd_dev->config_info); 5269 5270 rbd_put_client(rbd_dev->rbd_client); 5271 rbd_spec_put(rbd_dev->spec); 5272 kfree(rbd_dev->opts); 5273 kfree(rbd_dev); 5274 } 5275 5276 static void rbd_dev_release(struct device *dev) 5277 { 5278 struct rbd_device *rbd_dev = dev_to_rbd_dev(dev); 5279 bool need_put = !!rbd_dev->opts; 5280 5281 if (need_put) { 5282 destroy_workqueue(rbd_dev->task_wq); 5283 ida_simple_remove(&rbd_dev_id_ida, rbd_dev->dev_id); 5284 } 5285 5286 rbd_dev_free(rbd_dev); 5287 5288 /* 5289 * This is racy, but way better than putting module outside of 5290 * the release callback. The race window is pretty small, so 5291 * doing something similar to dm (dm-builtin.c) is overkill. 5292 */ 5293 if (need_put) 5294 module_put(THIS_MODULE); 5295 } 5296 5297 static struct rbd_device *__rbd_dev_create(struct rbd_client *rbdc, 5298 struct rbd_spec *spec) 5299 { 5300 struct rbd_device *rbd_dev; 5301 5302 rbd_dev = kzalloc(sizeof(*rbd_dev), GFP_KERNEL); 5303 if (!rbd_dev) 5304 return NULL; 5305 5306 spin_lock_init(&rbd_dev->lock); 5307 INIT_LIST_HEAD(&rbd_dev->node); 5308 init_rwsem(&rbd_dev->header_rwsem); 5309 5310 rbd_dev->header.data_pool_id = CEPH_NOPOOL; 5311 ceph_oid_init(&rbd_dev->header_oid); 5312 rbd_dev->header_oloc.pool = spec->pool_id; 5313 if (spec->pool_ns) { 5314 WARN_ON(!*spec->pool_ns); 5315 rbd_dev->header_oloc.pool_ns = 5316 ceph_find_or_create_string(spec->pool_ns, 5317 strlen(spec->pool_ns)); 5318 } 5319 5320 mutex_init(&rbd_dev->watch_mutex); 5321 rbd_dev->watch_state = RBD_WATCH_STATE_UNREGISTERED; 5322 INIT_DELAYED_WORK(&rbd_dev->watch_dwork, rbd_reregister_watch); 5323 5324 init_rwsem(&rbd_dev->lock_rwsem); 5325 rbd_dev->lock_state = RBD_LOCK_STATE_UNLOCKED; 5326 INIT_WORK(&rbd_dev->acquired_lock_work, rbd_notify_acquired_lock); 5327 INIT_WORK(&rbd_dev->released_lock_work, rbd_notify_released_lock); 5328 INIT_DELAYED_WORK(&rbd_dev->lock_dwork, rbd_acquire_lock); 5329 INIT_WORK(&rbd_dev->unlock_work, rbd_release_lock_work); 5330 spin_lock_init(&rbd_dev->lock_lists_lock); 5331 INIT_LIST_HEAD(&rbd_dev->acquiring_list); 5332 INIT_LIST_HEAD(&rbd_dev->running_list); 5333 init_completion(&rbd_dev->acquire_wait); 5334 init_completion(&rbd_dev->releasing_wait); 5335 5336 spin_lock_init(&rbd_dev->object_map_lock); 5337 5338 rbd_dev->dev.bus = &rbd_bus_type; 5339 rbd_dev->dev.type = &rbd_device_type; 5340 rbd_dev->dev.parent = &rbd_root_dev; 5341 device_initialize(&rbd_dev->dev); 5342 5343 rbd_dev->rbd_client = rbdc; 5344 rbd_dev->spec = spec; 5345 5346 return rbd_dev; 5347 } 5348 5349 /* 5350 * Create a mapping rbd_dev. 5351 */ 5352 static struct rbd_device *rbd_dev_create(struct rbd_client *rbdc, 5353 struct rbd_spec *spec, 5354 struct rbd_options *opts) 5355 { 5356 struct rbd_device *rbd_dev; 5357 5358 rbd_dev = __rbd_dev_create(rbdc, spec); 5359 if (!rbd_dev) 5360 return NULL; 5361 5362 rbd_dev->opts = opts; 5363 5364 /* get an id and fill in device name */ 5365 rbd_dev->dev_id = ida_simple_get(&rbd_dev_id_ida, 0, 5366 minor_to_rbd_dev_id(1 << MINORBITS), 5367 GFP_KERNEL); 5368 if (rbd_dev->dev_id < 0) 5369 goto fail_rbd_dev; 5370 5371 sprintf(rbd_dev->name, RBD_DRV_NAME "%d", rbd_dev->dev_id); 5372 rbd_dev->task_wq = alloc_ordered_workqueue("%s-tasks", WQ_MEM_RECLAIM, 5373 rbd_dev->name); 5374 if (!rbd_dev->task_wq) 5375 goto fail_dev_id; 5376 5377 /* we have a ref from do_rbd_add() */ 5378 __module_get(THIS_MODULE); 5379 5380 dout("%s rbd_dev %p dev_id %d\n", __func__, rbd_dev, rbd_dev->dev_id); 5381 return rbd_dev; 5382 5383 fail_dev_id: 5384 ida_simple_remove(&rbd_dev_id_ida, rbd_dev->dev_id); 5385 fail_rbd_dev: 5386 rbd_dev_free(rbd_dev); 5387 return NULL; 5388 } 5389 5390 static void rbd_dev_destroy(struct rbd_device *rbd_dev) 5391 { 5392 if (rbd_dev) 5393 put_device(&rbd_dev->dev); 5394 } 5395 5396 /* 5397 * Get the size and object order for an image snapshot, or if 5398 * snap_id is CEPH_NOSNAP, gets this information for the base 5399 * image. 5400 */ 5401 static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id, 5402 u8 *order, u64 *snap_size) 5403 { 5404 __le64 snapid = cpu_to_le64(snap_id); 5405 int ret; 5406 struct { 5407 u8 order; 5408 __le64 size; 5409 } __attribute__ ((packed)) size_buf = { 0 }; 5410 5411 ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid, 5412 &rbd_dev->header_oloc, "get_size", 5413 &snapid, sizeof(snapid), 5414 &size_buf, sizeof(size_buf)); 5415 dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); 5416 if (ret < 0) 5417 return ret; 5418 if (ret < sizeof (size_buf)) 5419 return -ERANGE; 5420 5421 if (order) { 5422 *order = size_buf.order; 5423 dout(" order %u", (unsigned int)*order); 5424 } 5425 *snap_size = le64_to_cpu(size_buf.size); 5426 5427 dout(" snap_id 0x%016llx snap_size = %llu\n", 5428 (unsigned long long)snap_id, 5429 (unsigned long long)*snap_size); 5430 5431 return 0; 5432 } 5433 5434 static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev) 5435 { 5436 return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP, 5437 &rbd_dev->header.obj_order, 5438 &rbd_dev->header.image_size); 5439 } 5440 5441 static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev) 5442 { 5443 size_t size; 5444 void *reply_buf; 5445 int ret; 5446 void *p; 5447 5448 /* Response will be an encoded string, which includes a length */ 5449 size = sizeof(__le32) + RBD_OBJ_PREFIX_LEN_MAX; 5450 reply_buf = kzalloc(size, GFP_KERNEL); 5451 if (!reply_buf) 5452 return -ENOMEM; 5453 5454 ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid, 5455 &rbd_dev->header_oloc, "get_object_prefix", 5456 NULL, 0, reply_buf, size); 5457 dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); 5458 if (ret < 0) 5459 goto out; 5460 5461 p = reply_buf; 5462 rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p, 5463 p + ret, NULL, GFP_NOIO); 5464 ret = 0; 5465 5466 if (IS_ERR(rbd_dev->header.object_prefix)) { 5467 ret = PTR_ERR(rbd_dev->header.object_prefix); 5468 rbd_dev->header.object_prefix = NULL; 5469 } else { 5470 dout(" object_prefix = %s\n", rbd_dev->header.object_prefix); 5471 } 5472 out: 5473 kfree(reply_buf); 5474 5475 return ret; 5476 } 5477 5478 static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id, 5479 bool read_only, u64 *snap_features) 5480 { 5481 struct { 5482 __le64 snap_id; 5483 u8 read_only; 5484 } features_in; 5485 struct { 5486 __le64 features; 5487 __le64 incompat; 5488 } __attribute__ ((packed)) features_buf = { 0 }; 5489 u64 unsup; 5490 int ret; 5491 5492 features_in.snap_id = cpu_to_le64(snap_id); 5493 features_in.read_only = read_only; 5494 5495 ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid, 5496 &rbd_dev->header_oloc, "get_features", 5497 &features_in, sizeof(features_in), 5498 &features_buf, sizeof(features_buf)); 5499 dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); 5500 if (ret < 0) 5501 return ret; 5502 if (ret < sizeof (features_buf)) 5503 return -ERANGE; 5504 5505 unsup = le64_to_cpu(features_buf.incompat) & ~RBD_FEATURES_SUPPORTED; 5506 if (unsup) { 5507 rbd_warn(rbd_dev, "image uses unsupported features: 0x%llx", 5508 unsup); 5509 return -ENXIO; 5510 } 5511 5512 *snap_features = le64_to_cpu(features_buf.features); 5513 5514 dout(" snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n", 5515 (unsigned long long)snap_id, 5516 (unsigned long long)*snap_features, 5517 (unsigned long long)le64_to_cpu(features_buf.incompat)); 5518 5519 return 0; 5520 } 5521 5522 static int rbd_dev_v2_features(struct rbd_device *rbd_dev) 5523 { 5524 return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP, 5525 rbd_is_ro(rbd_dev), 5526 &rbd_dev->header.features); 5527 } 5528 5529 /* 5530 * These are generic image flags, but since they are used only for 5531 * object map, store them in rbd_dev->object_map_flags. 5532 * 5533 * For the same reason, this function is called only on object map 5534 * (re)load and not on header refresh. 5535 */ 5536 static int rbd_dev_v2_get_flags(struct rbd_device *rbd_dev) 5537 { 5538 __le64 snapid = cpu_to_le64(rbd_dev->spec->snap_id); 5539 __le64 flags; 5540 int ret; 5541 5542 ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid, 5543 &rbd_dev->header_oloc, "get_flags", 5544 &snapid, sizeof(snapid), 5545 &flags, sizeof(flags)); 5546 if (ret < 0) 5547 return ret; 5548 if (ret < sizeof(flags)) 5549 return -EBADMSG; 5550 5551 rbd_dev->object_map_flags = le64_to_cpu(flags); 5552 return 0; 5553 } 5554 5555 struct parent_image_info { 5556 u64 pool_id; 5557 const char *pool_ns; 5558 const char *image_id; 5559 u64 snap_id; 5560 5561 bool has_overlap; 5562 u64 overlap; 5563 }; 5564 5565 /* 5566 * The caller is responsible for @pii. 5567 */ 5568 static int decode_parent_image_spec(void **p, void *end, 5569 struct parent_image_info *pii) 5570 { 5571 u8 struct_v; 5572 u32 struct_len; 5573 int ret; 5574 5575 ret = ceph_start_decoding(p, end, 1, "ParentImageSpec", 5576 &struct_v, &struct_len); 5577 if (ret) 5578 return ret; 5579 5580 ceph_decode_64_safe(p, end, pii->pool_id, e_inval); 5581 pii->pool_ns = ceph_extract_encoded_string(p, end, NULL, GFP_KERNEL); 5582 if (IS_ERR(pii->pool_ns)) { 5583 ret = PTR_ERR(pii->pool_ns); 5584 pii->pool_ns = NULL; 5585 return ret; 5586 } 5587 pii->image_id = ceph_extract_encoded_string(p, end, NULL, GFP_KERNEL); 5588 if (IS_ERR(pii->image_id)) { 5589 ret = PTR_ERR(pii->image_id); 5590 pii->image_id = NULL; 5591 return ret; 5592 } 5593 ceph_decode_64_safe(p, end, pii->snap_id, e_inval); 5594 return 0; 5595 5596 e_inval: 5597 return -EINVAL; 5598 } 5599 5600 static int __get_parent_info(struct rbd_device *rbd_dev, 5601 struct page *req_page, 5602 struct page *reply_page, 5603 struct parent_image_info *pii) 5604 { 5605 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 5606 size_t reply_len = PAGE_SIZE; 5607 void *p, *end; 5608 int ret; 5609 5610 ret = ceph_osdc_call(osdc, &rbd_dev->header_oid, &rbd_dev->header_oloc, 5611 "rbd", "parent_get", CEPH_OSD_FLAG_READ, 5612 req_page, sizeof(u64), &reply_page, &reply_len); 5613 if (ret) 5614 return ret == -EOPNOTSUPP ? 1 : ret; 5615 5616 p = page_address(reply_page); 5617 end = p + reply_len; 5618 ret = decode_parent_image_spec(&p, end, pii); 5619 if (ret) 5620 return ret; 5621 5622 ret = ceph_osdc_call(osdc, &rbd_dev->header_oid, &rbd_dev->header_oloc, 5623 "rbd", "parent_overlap_get", CEPH_OSD_FLAG_READ, 5624 req_page, sizeof(u64), &reply_page, &reply_len); 5625 if (ret) 5626 return ret; 5627 5628 p = page_address(reply_page); 5629 end = p + reply_len; 5630 ceph_decode_8_safe(&p, end, pii->has_overlap, e_inval); 5631 if (pii->has_overlap) 5632 ceph_decode_64_safe(&p, end, pii->overlap, e_inval); 5633 5634 return 0; 5635 5636 e_inval: 5637 return -EINVAL; 5638 } 5639 5640 /* 5641 * The caller is responsible for @pii. 5642 */ 5643 static int __get_parent_info_legacy(struct rbd_device *rbd_dev, 5644 struct page *req_page, 5645 struct page *reply_page, 5646 struct parent_image_info *pii) 5647 { 5648 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 5649 size_t reply_len = PAGE_SIZE; 5650 void *p, *end; 5651 int ret; 5652 5653 ret = ceph_osdc_call(osdc, &rbd_dev->header_oid, &rbd_dev->header_oloc, 5654 "rbd", "get_parent", CEPH_OSD_FLAG_READ, 5655 req_page, sizeof(u64), &reply_page, &reply_len); 5656 if (ret) 5657 return ret; 5658 5659 p = page_address(reply_page); 5660 end = p + reply_len; 5661 ceph_decode_64_safe(&p, end, pii->pool_id, e_inval); 5662 pii->image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); 5663 if (IS_ERR(pii->image_id)) { 5664 ret = PTR_ERR(pii->image_id); 5665 pii->image_id = NULL; 5666 return ret; 5667 } 5668 ceph_decode_64_safe(&p, end, pii->snap_id, e_inval); 5669 pii->has_overlap = true; 5670 ceph_decode_64_safe(&p, end, pii->overlap, e_inval); 5671 5672 return 0; 5673 5674 e_inval: 5675 return -EINVAL; 5676 } 5677 5678 static int get_parent_info(struct rbd_device *rbd_dev, 5679 struct parent_image_info *pii) 5680 { 5681 struct page *req_page, *reply_page; 5682 void *p; 5683 int ret; 5684 5685 req_page = alloc_page(GFP_KERNEL); 5686 if (!req_page) 5687 return -ENOMEM; 5688 5689 reply_page = alloc_page(GFP_KERNEL); 5690 if (!reply_page) { 5691 __free_page(req_page); 5692 return -ENOMEM; 5693 } 5694 5695 p = page_address(req_page); 5696 ceph_encode_64(&p, rbd_dev->spec->snap_id); 5697 ret = __get_parent_info(rbd_dev, req_page, reply_page, pii); 5698 if (ret > 0) 5699 ret = __get_parent_info_legacy(rbd_dev, req_page, reply_page, 5700 pii); 5701 5702 __free_page(req_page); 5703 __free_page(reply_page); 5704 return ret; 5705 } 5706 5707 static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev) 5708 { 5709 struct rbd_spec *parent_spec; 5710 struct parent_image_info pii = { 0 }; 5711 int ret; 5712 5713 parent_spec = rbd_spec_alloc(); 5714 if (!parent_spec) 5715 return -ENOMEM; 5716 5717 ret = get_parent_info(rbd_dev, &pii); 5718 if (ret) 5719 goto out_err; 5720 5721 dout("%s pool_id %llu pool_ns %s image_id %s snap_id %llu has_overlap %d overlap %llu\n", 5722 __func__, pii.pool_id, pii.pool_ns, pii.image_id, pii.snap_id, 5723 pii.has_overlap, pii.overlap); 5724 5725 if (pii.pool_id == CEPH_NOPOOL || !pii.has_overlap) { 5726 /* 5727 * Either the parent never existed, or we have 5728 * record of it but the image got flattened so it no 5729 * longer has a parent. When the parent of a 5730 * layered image disappears we immediately set the 5731 * overlap to 0. The effect of this is that all new 5732 * requests will be treated as if the image had no 5733 * parent. 5734 * 5735 * If !pii.has_overlap, the parent image spec is not 5736 * applicable. It's there to avoid duplication in each 5737 * snapshot record. 5738 */ 5739 if (rbd_dev->parent_overlap) { 5740 rbd_dev->parent_overlap = 0; 5741 rbd_dev_parent_put(rbd_dev); 5742 pr_info("%s: clone image has been flattened\n", 5743 rbd_dev->disk->disk_name); 5744 } 5745 5746 goto out; /* No parent? No problem. */ 5747 } 5748 5749 /* The ceph file layout needs to fit pool id in 32 bits */ 5750 5751 ret = -EIO; 5752 if (pii.pool_id > (u64)U32_MAX) { 5753 rbd_warn(NULL, "parent pool id too large (%llu > %u)", 5754 (unsigned long long)pii.pool_id, U32_MAX); 5755 goto out_err; 5756 } 5757 5758 /* 5759 * The parent won't change (except when the clone is 5760 * flattened, already handled that). So we only need to 5761 * record the parent spec we have not already done so. 5762 */ 5763 if (!rbd_dev->parent_spec) { 5764 parent_spec->pool_id = pii.pool_id; 5765 if (pii.pool_ns && *pii.pool_ns) { 5766 parent_spec->pool_ns = pii.pool_ns; 5767 pii.pool_ns = NULL; 5768 } 5769 parent_spec->image_id = pii.image_id; 5770 pii.image_id = NULL; 5771 parent_spec->snap_id = pii.snap_id; 5772 5773 rbd_dev->parent_spec = parent_spec; 5774 parent_spec = NULL; /* rbd_dev now owns this */ 5775 } 5776 5777 /* 5778 * We always update the parent overlap. If it's zero we issue 5779 * a warning, as we will proceed as if there was no parent. 5780 */ 5781 if (!pii.overlap) { 5782 if (parent_spec) { 5783 /* refresh, careful to warn just once */ 5784 if (rbd_dev->parent_overlap) 5785 rbd_warn(rbd_dev, 5786 "clone now standalone (overlap became 0)"); 5787 } else { 5788 /* initial probe */ 5789 rbd_warn(rbd_dev, "clone is standalone (overlap 0)"); 5790 } 5791 } 5792 rbd_dev->parent_overlap = pii.overlap; 5793 5794 out: 5795 ret = 0; 5796 out_err: 5797 kfree(pii.pool_ns); 5798 kfree(pii.image_id); 5799 rbd_spec_put(parent_spec); 5800 return ret; 5801 } 5802 5803 static int rbd_dev_v2_striping_info(struct rbd_device *rbd_dev) 5804 { 5805 struct { 5806 __le64 stripe_unit; 5807 __le64 stripe_count; 5808 } __attribute__ ((packed)) striping_info_buf = { 0 }; 5809 size_t size = sizeof (striping_info_buf); 5810 void *p; 5811 int ret; 5812 5813 ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid, 5814 &rbd_dev->header_oloc, "get_stripe_unit_count", 5815 NULL, 0, &striping_info_buf, size); 5816 dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); 5817 if (ret < 0) 5818 return ret; 5819 if (ret < size) 5820 return -ERANGE; 5821 5822 p = &striping_info_buf; 5823 rbd_dev->header.stripe_unit = ceph_decode_64(&p); 5824 rbd_dev->header.stripe_count = ceph_decode_64(&p); 5825 return 0; 5826 } 5827 5828 static int rbd_dev_v2_data_pool(struct rbd_device *rbd_dev) 5829 { 5830 __le64 data_pool_id; 5831 int ret; 5832 5833 ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid, 5834 &rbd_dev->header_oloc, "get_data_pool", 5835 NULL, 0, &data_pool_id, sizeof(data_pool_id)); 5836 if (ret < 0) 5837 return ret; 5838 if (ret < sizeof(data_pool_id)) 5839 return -EBADMSG; 5840 5841 rbd_dev->header.data_pool_id = le64_to_cpu(data_pool_id); 5842 WARN_ON(rbd_dev->header.data_pool_id == CEPH_NOPOOL); 5843 return 0; 5844 } 5845 5846 static char *rbd_dev_image_name(struct rbd_device *rbd_dev) 5847 { 5848 CEPH_DEFINE_OID_ONSTACK(oid); 5849 size_t image_id_size; 5850 char *image_id; 5851 void *p; 5852 void *end; 5853 size_t size; 5854 void *reply_buf = NULL; 5855 size_t len = 0; 5856 char *image_name = NULL; 5857 int ret; 5858 5859 rbd_assert(!rbd_dev->spec->image_name); 5860 5861 len = strlen(rbd_dev->spec->image_id); 5862 image_id_size = sizeof (__le32) + len; 5863 image_id = kmalloc(image_id_size, GFP_KERNEL); 5864 if (!image_id) 5865 return NULL; 5866 5867 p = image_id; 5868 end = image_id + image_id_size; 5869 ceph_encode_string(&p, end, rbd_dev->spec->image_id, (u32)len); 5870 5871 size = sizeof (__le32) + RBD_IMAGE_NAME_LEN_MAX; 5872 reply_buf = kmalloc(size, GFP_KERNEL); 5873 if (!reply_buf) 5874 goto out; 5875 5876 ceph_oid_printf(&oid, "%s", RBD_DIRECTORY); 5877 ret = rbd_obj_method_sync(rbd_dev, &oid, &rbd_dev->header_oloc, 5878 "dir_get_name", image_id, image_id_size, 5879 reply_buf, size); 5880 if (ret < 0) 5881 goto out; 5882 p = reply_buf; 5883 end = reply_buf + ret; 5884 5885 image_name = ceph_extract_encoded_string(&p, end, &len, GFP_KERNEL); 5886 if (IS_ERR(image_name)) 5887 image_name = NULL; 5888 else 5889 dout("%s: name is %s len is %zd\n", __func__, image_name, len); 5890 out: 5891 kfree(reply_buf); 5892 kfree(image_id); 5893 5894 return image_name; 5895 } 5896 5897 static u64 rbd_v1_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) 5898 { 5899 struct ceph_snap_context *snapc = rbd_dev->header.snapc; 5900 const char *snap_name; 5901 u32 which = 0; 5902 5903 /* Skip over names until we find the one we are looking for */ 5904 5905 snap_name = rbd_dev->header.snap_names; 5906 while (which < snapc->num_snaps) { 5907 if (!strcmp(name, snap_name)) 5908 return snapc->snaps[which]; 5909 snap_name += strlen(snap_name) + 1; 5910 which++; 5911 } 5912 return CEPH_NOSNAP; 5913 } 5914 5915 static u64 rbd_v2_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) 5916 { 5917 struct ceph_snap_context *snapc = rbd_dev->header.snapc; 5918 u32 which; 5919 bool found = false; 5920 u64 snap_id; 5921 5922 for (which = 0; !found && which < snapc->num_snaps; which++) { 5923 const char *snap_name; 5924 5925 snap_id = snapc->snaps[which]; 5926 snap_name = rbd_dev_v2_snap_name(rbd_dev, snap_id); 5927 if (IS_ERR(snap_name)) { 5928 /* ignore no-longer existing snapshots */ 5929 if (PTR_ERR(snap_name) == -ENOENT) 5930 continue; 5931 else 5932 break; 5933 } 5934 found = !strcmp(name, snap_name); 5935 kfree(snap_name); 5936 } 5937 return found ? snap_id : CEPH_NOSNAP; 5938 } 5939 5940 /* 5941 * Assumes name is never RBD_SNAP_HEAD_NAME; returns CEPH_NOSNAP if 5942 * no snapshot by that name is found, or if an error occurs. 5943 */ 5944 static u64 rbd_snap_id_by_name(struct rbd_device *rbd_dev, const char *name) 5945 { 5946 if (rbd_dev->image_format == 1) 5947 return rbd_v1_snap_id_by_name(rbd_dev, name); 5948 5949 return rbd_v2_snap_id_by_name(rbd_dev, name); 5950 } 5951 5952 /* 5953 * An image being mapped will have everything but the snap id. 5954 */ 5955 static int rbd_spec_fill_snap_id(struct rbd_device *rbd_dev) 5956 { 5957 struct rbd_spec *spec = rbd_dev->spec; 5958 5959 rbd_assert(spec->pool_id != CEPH_NOPOOL && spec->pool_name); 5960 rbd_assert(spec->image_id && spec->image_name); 5961 rbd_assert(spec->snap_name); 5962 5963 if (strcmp(spec->snap_name, RBD_SNAP_HEAD_NAME)) { 5964 u64 snap_id; 5965 5966 snap_id = rbd_snap_id_by_name(rbd_dev, spec->snap_name); 5967 if (snap_id == CEPH_NOSNAP) 5968 return -ENOENT; 5969 5970 spec->snap_id = snap_id; 5971 } else { 5972 spec->snap_id = CEPH_NOSNAP; 5973 } 5974 5975 return 0; 5976 } 5977 5978 /* 5979 * A parent image will have all ids but none of the names. 5980 * 5981 * All names in an rbd spec are dynamically allocated. It's OK if we 5982 * can't figure out the name for an image id. 5983 */ 5984 static int rbd_spec_fill_names(struct rbd_device *rbd_dev) 5985 { 5986 struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc; 5987 struct rbd_spec *spec = rbd_dev->spec; 5988 const char *pool_name; 5989 const char *image_name; 5990 const char *snap_name; 5991 int ret; 5992 5993 rbd_assert(spec->pool_id != CEPH_NOPOOL); 5994 rbd_assert(spec->image_id); 5995 rbd_assert(spec->snap_id != CEPH_NOSNAP); 5996 5997 /* Get the pool name; we have to make our own copy of this */ 5998 5999 pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, spec->pool_id); 6000 if (!pool_name) { 6001 rbd_warn(rbd_dev, "no pool with id %llu", spec->pool_id); 6002 return -EIO; 6003 } 6004 pool_name = kstrdup(pool_name, GFP_KERNEL); 6005 if (!pool_name) 6006 return -ENOMEM; 6007 6008 /* Fetch the image name; tolerate failure here */ 6009 6010 image_name = rbd_dev_image_name(rbd_dev); 6011 if (!image_name) 6012 rbd_warn(rbd_dev, "unable to get image name"); 6013 6014 /* Fetch the snapshot name */ 6015 6016 snap_name = rbd_snap_name(rbd_dev, spec->snap_id); 6017 if (IS_ERR(snap_name)) { 6018 ret = PTR_ERR(snap_name); 6019 goto out_err; 6020 } 6021 6022 spec->pool_name = pool_name; 6023 spec->image_name = image_name; 6024 spec->snap_name = snap_name; 6025 6026 return 0; 6027 6028 out_err: 6029 kfree(image_name); 6030 kfree(pool_name); 6031 return ret; 6032 } 6033 6034 static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev) 6035 { 6036 size_t size; 6037 int ret; 6038 void *reply_buf; 6039 void *p; 6040 void *end; 6041 u64 seq; 6042 u32 snap_count; 6043 struct ceph_snap_context *snapc; 6044 u32 i; 6045 6046 /* 6047 * We'll need room for the seq value (maximum snapshot id), 6048 * snapshot count, and array of that many snapshot ids. 6049 * For now we have a fixed upper limit on the number we're 6050 * prepared to receive. 6051 */ 6052 size = sizeof (__le64) + sizeof (__le32) + 6053 RBD_MAX_SNAP_COUNT * sizeof (__le64); 6054 reply_buf = kzalloc(size, GFP_KERNEL); 6055 if (!reply_buf) 6056 return -ENOMEM; 6057 6058 ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid, 6059 &rbd_dev->header_oloc, "get_snapcontext", 6060 NULL, 0, reply_buf, size); 6061 dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); 6062 if (ret < 0) 6063 goto out; 6064 6065 p = reply_buf; 6066 end = reply_buf + ret; 6067 ret = -ERANGE; 6068 ceph_decode_64_safe(&p, end, seq, out); 6069 ceph_decode_32_safe(&p, end, snap_count, out); 6070 6071 /* 6072 * Make sure the reported number of snapshot ids wouldn't go 6073 * beyond the end of our buffer. But before checking that, 6074 * make sure the computed size of the snapshot context we 6075 * allocate is representable in a size_t. 6076 */ 6077 if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context)) 6078 / sizeof (u64)) { 6079 ret = -EINVAL; 6080 goto out; 6081 } 6082 if (!ceph_has_room(&p, end, snap_count * sizeof (__le64))) 6083 goto out; 6084 ret = 0; 6085 6086 snapc = ceph_create_snap_context(snap_count, GFP_KERNEL); 6087 if (!snapc) { 6088 ret = -ENOMEM; 6089 goto out; 6090 } 6091 snapc->seq = seq; 6092 for (i = 0; i < snap_count; i++) 6093 snapc->snaps[i] = ceph_decode_64(&p); 6094 6095 ceph_put_snap_context(rbd_dev->header.snapc); 6096 rbd_dev->header.snapc = snapc; 6097 6098 dout(" snap context seq = %llu, snap_count = %u\n", 6099 (unsigned long long)seq, (unsigned int)snap_count); 6100 out: 6101 kfree(reply_buf); 6102 6103 return ret; 6104 } 6105 6106 static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev, 6107 u64 snap_id) 6108 { 6109 size_t size; 6110 void *reply_buf; 6111 __le64 snapid; 6112 int ret; 6113 void *p; 6114 void *end; 6115 char *snap_name; 6116 6117 size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN; 6118 reply_buf = kmalloc(size, GFP_KERNEL); 6119 if (!reply_buf) 6120 return ERR_PTR(-ENOMEM); 6121 6122 snapid = cpu_to_le64(snap_id); 6123 ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid, 6124 &rbd_dev->header_oloc, "get_snapshot_name", 6125 &snapid, sizeof(snapid), reply_buf, size); 6126 dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); 6127 if (ret < 0) { 6128 snap_name = ERR_PTR(ret); 6129 goto out; 6130 } 6131 6132 p = reply_buf; 6133 end = reply_buf + ret; 6134 snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL); 6135 if (IS_ERR(snap_name)) 6136 goto out; 6137 6138 dout(" snap_id 0x%016llx snap_name = %s\n", 6139 (unsigned long long)snap_id, snap_name); 6140 out: 6141 kfree(reply_buf); 6142 6143 return snap_name; 6144 } 6145 6146 static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev) 6147 { 6148 bool first_time = rbd_dev->header.object_prefix == NULL; 6149 int ret; 6150 6151 ret = rbd_dev_v2_image_size(rbd_dev); 6152 if (ret) 6153 return ret; 6154 6155 if (first_time) { 6156 ret = rbd_dev_v2_header_onetime(rbd_dev); 6157 if (ret) 6158 return ret; 6159 } 6160 6161 ret = rbd_dev_v2_snap_context(rbd_dev); 6162 if (ret && first_time) { 6163 kfree(rbd_dev->header.object_prefix); 6164 rbd_dev->header.object_prefix = NULL; 6165 } 6166 6167 return ret; 6168 } 6169 6170 static int rbd_dev_header_info(struct rbd_device *rbd_dev) 6171 { 6172 rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); 6173 6174 if (rbd_dev->image_format == 1) 6175 return rbd_dev_v1_header_info(rbd_dev); 6176 6177 return rbd_dev_v2_header_info(rbd_dev); 6178 } 6179 6180 /* 6181 * Skips over white space at *buf, and updates *buf to point to the 6182 * first found non-space character (if any). Returns the length of 6183 * the token (string of non-white space characters) found. Note 6184 * that *buf must be terminated with '\0'. 6185 */ 6186 static inline size_t next_token(const char **buf) 6187 { 6188 /* 6189 * These are the characters that produce nonzero for 6190 * isspace() in the "C" and "POSIX" locales. 6191 */ 6192 static const char spaces[] = " \f\n\r\t\v"; 6193 6194 *buf += strspn(*buf, spaces); /* Find start of token */ 6195 6196 return strcspn(*buf, spaces); /* Return token length */ 6197 } 6198 6199 /* 6200 * Finds the next token in *buf, dynamically allocates a buffer big 6201 * enough to hold a copy of it, and copies the token into the new 6202 * buffer. The copy is guaranteed to be terminated with '\0'. Note 6203 * that a duplicate buffer is created even for a zero-length token. 6204 * 6205 * Returns a pointer to the newly-allocated duplicate, or a null 6206 * pointer if memory for the duplicate was not available. If 6207 * the lenp argument is a non-null pointer, the length of the token 6208 * (not including the '\0') is returned in *lenp. 6209 * 6210 * If successful, the *buf pointer will be updated to point beyond 6211 * the end of the found token. 6212 * 6213 * Note: uses GFP_KERNEL for allocation. 6214 */ 6215 static inline char *dup_token(const char **buf, size_t *lenp) 6216 { 6217 char *dup; 6218 size_t len; 6219 6220 len = next_token(buf); 6221 dup = kmemdup(*buf, len + 1, GFP_KERNEL); 6222 if (!dup) 6223 return NULL; 6224 *(dup + len) = '\0'; 6225 *buf += len; 6226 6227 if (lenp) 6228 *lenp = len; 6229 6230 return dup; 6231 } 6232 6233 static int rbd_parse_param(struct fs_parameter *param, 6234 struct rbd_parse_opts_ctx *pctx) 6235 { 6236 struct rbd_options *opt = pctx->opts; 6237 struct fs_parse_result result; 6238 struct p_log log = {.prefix = "rbd"}; 6239 int token, ret; 6240 6241 ret = ceph_parse_param(param, pctx->copts, NULL); 6242 if (ret != -ENOPARAM) 6243 return ret; 6244 6245 token = __fs_parse(&log, rbd_parameters, param, &result); 6246 dout("%s fs_parse '%s' token %d\n", __func__, param->key, token); 6247 if (token < 0) { 6248 if (token == -ENOPARAM) 6249 return inval_plog(&log, "Unknown parameter '%s'", 6250 param->key); 6251 return token; 6252 } 6253 6254 switch (token) { 6255 case Opt_queue_depth: 6256 if (result.uint_32 < 1) 6257 goto out_of_range; 6258 opt->queue_depth = result.uint_32; 6259 break; 6260 case Opt_alloc_size: 6261 if (result.uint_32 < SECTOR_SIZE) 6262 goto out_of_range; 6263 if (!is_power_of_2(result.uint_32)) 6264 return inval_plog(&log, "alloc_size must be a power of 2"); 6265 opt->alloc_size = result.uint_32; 6266 break; 6267 case Opt_lock_timeout: 6268 /* 0 is "wait forever" (i.e. infinite timeout) */ 6269 if (result.uint_32 > INT_MAX / 1000) 6270 goto out_of_range; 6271 opt->lock_timeout = msecs_to_jiffies(result.uint_32 * 1000); 6272 break; 6273 case Opt_pool_ns: 6274 kfree(pctx->spec->pool_ns); 6275 pctx->spec->pool_ns = param->string; 6276 param->string = NULL; 6277 break; 6278 case Opt_compression_hint: 6279 switch (result.uint_32) { 6280 case Opt_compression_hint_none: 6281 opt->alloc_hint_flags &= 6282 ~(CEPH_OSD_ALLOC_HINT_FLAG_COMPRESSIBLE | 6283 CEPH_OSD_ALLOC_HINT_FLAG_INCOMPRESSIBLE); 6284 break; 6285 case Opt_compression_hint_compressible: 6286 opt->alloc_hint_flags |= 6287 CEPH_OSD_ALLOC_HINT_FLAG_COMPRESSIBLE; 6288 opt->alloc_hint_flags &= 6289 ~CEPH_OSD_ALLOC_HINT_FLAG_INCOMPRESSIBLE; 6290 break; 6291 case Opt_compression_hint_incompressible: 6292 opt->alloc_hint_flags |= 6293 CEPH_OSD_ALLOC_HINT_FLAG_INCOMPRESSIBLE; 6294 opt->alloc_hint_flags &= 6295 ~CEPH_OSD_ALLOC_HINT_FLAG_COMPRESSIBLE; 6296 break; 6297 default: 6298 BUG(); 6299 } 6300 break; 6301 case Opt_read_only: 6302 opt->read_only = true; 6303 break; 6304 case Opt_read_write: 6305 opt->read_only = false; 6306 break; 6307 case Opt_lock_on_read: 6308 opt->lock_on_read = true; 6309 break; 6310 case Opt_exclusive: 6311 opt->exclusive = true; 6312 break; 6313 case Opt_notrim: 6314 opt->trim = false; 6315 break; 6316 default: 6317 BUG(); 6318 } 6319 6320 return 0; 6321 6322 out_of_range: 6323 return inval_plog(&log, "%s out of range", param->key); 6324 } 6325 6326 /* 6327 * This duplicates most of generic_parse_monolithic(), untying it from 6328 * fs_context and skipping standard superblock and security options. 6329 */ 6330 static int rbd_parse_options(char *options, struct rbd_parse_opts_ctx *pctx) 6331 { 6332 char *key; 6333 int ret = 0; 6334 6335 dout("%s '%s'\n", __func__, options); 6336 while ((key = strsep(&options, ",")) != NULL) { 6337 if (*key) { 6338 struct fs_parameter param = { 6339 .key = key, 6340 .type = fs_value_is_flag, 6341 }; 6342 char *value = strchr(key, '='); 6343 size_t v_len = 0; 6344 6345 if (value) { 6346 if (value == key) 6347 continue; 6348 *value++ = 0; 6349 v_len = strlen(value); 6350 param.string = kmemdup_nul(value, v_len, 6351 GFP_KERNEL); 6352 if (!param.string) 6353 return -ENOMEM; 6354 param.type = fs_value_is_string; 6355 } 6356 param.size = v_len; 6357 6358 ret = rbd_parse_param(¶m, pctx); 6359 kfree(param.string); 6360 if (ret) 6361 break; 6362 } 6363 } 6364 6365 return ret; 6366 } 6367 6368 /* 6369 * Parse the options provided for an "rbd add" (i.e., rbd image 6370 * mapping) request. These arrive via a write to /sys/bus/rbd/add, 6371 * and the data written is passed here via a NUL-terminated buffer. 6372 * Returns 0 if successful or an error code otherwise. 6373 * 6374 * The information extracted from these options is recorded in 6375 * the other parameters which return dynamically-allocated 6376 * structures: 6377 * ceph_opts 6378 * The address of a pointer that will refer to a ceph options 6379 * structure. Caller must release the returned pointer using 6380 * ceph_destroy_options() when it is no longer needed. 6381 * rbd_opts 6382 * Address of an rbd options pointer. Fully initialized by 6383 * this function; caller must release with kfree(). 6384 * spec 6385 * Address of an rbd image specification pointer. Fully 6386 * initialized by this function based on parsed options. 6387 * Caller must release with rbd_spec_put(). 6388 * 6389 * The options passed take this form: 6390 * <mon_addrs> <options> <pool_name> <image_name> [<snap_id>] 6391 * where: 6392 * <mon_addrs> 6393 * A comma-separated list of one or more monitor addresses. 6394 * A monitor address is an ip address, optionally followed 6395 * by a port number (separated by a colon). 6396 * I.e.: ip1[:port1][,ip2[:port2]...] 6397 * <options> 6398 * A comma-separated list of ceph and/or rbd options. 6399 * <pool_name> 6400 * The name of the rados pool containing the rbd image. 6401 * <image_name> 6402 * The name of the image in that pool to map. 6403 * <snap_id> 6404 * An optional snapshot id. If provided, the mapping will 6405 * present data from the image at the time that snapshot was 6406 * created. The image head is used if no snapshot id is 6407 * provided. Snapshot mappings are always read-only. 6408 */ 6409 static int rbd_add_parse_args(const char *buf, 6410 struct ceph_options **ceph_opts, 6411 struct rbd_options **opts, 6412 struct rbd_spec **rbd_spec) 6413 { 6414 size_t len; 6415 char *options; 6416 const char *mon_addrs; 6417 char *snap_name; 6418 size_t mon_addrs_size; 6419 struct rbd_parse_opts_ctx pctx = { 0 }; 6420 int ret; 6421 6422 /* The first four tokens are required */ 6423 6424 len = next_token(&buf); 6425 if (!len) { 6426 rbd_warn(NULL, "no monitor address(es) provided"); 6427 return -EINVAL; 6428 } 6429 mon_addrs = buf; 6430 mon_addrs_size = len; 6431 buf += len; 6432 6433 ret = -EINVAL; 6434 options = dup_token(&buf, NULL); 6435 if (!options) 6436 return -ENOMEM; 6437 if (!*options) { 6438 rbd_warn(NULL, "no options provided"); 6439 goto out_err; 6440 } 6441 6442 pctx.spec = rbd_spec_alloc(); 6443 if (!pctx.spec) 6444 goto out_mem; 6445 6446 pctx.spec->pool_name = dup_token(&buf, NULL); 6447 if (!pctx.spec->pool_name) 6448 goto out_mem; 6449 if (!*pctx.spec->pool_name) { 6450 rbd_warn(NULL, "no pool name provided"); 6451 goto out_err; 6452 } 6453 6454 pctx.spec->image_name = dup_token(&buf, NULL); 6455 if (!pctx.spec->image_name) 6456 goto out_mem; 6457 if (!*pctx.spec->image_name) { 6458 rbd_warn(NULL, "no image name provided"); 6459 goto out_err; 6460 } 6461 6462 /* 6463 * Snapshot name is optional; default is to use "-" 6464 * (indicating the head/no snapshot). 6465 */ 6466 len = next_token(&buf); 6467 if (!len) { 6468 buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */ 6469 len = sizeof (RBD_SNAP_HEAD_NAME) - 1; 6470 } else if (len > RBD_MAX_SNAP_NAME_LEN) { 6471 ret = -ENAMETOOLONG; 6472 goto out_err; 6473 } 6474 snap_name = kmemdup(buf, len + 1, GFP_KERNEL); 6475 if (!snap_name) 6476 goto out_mem; 6477 *(snap_name + len) = '\0'; 6478 pctx.spec->snap_name = snap_name; 6479 6480 pctx.copts = ceph_alloc_options(); 6481 if (!pctx.copts) 6482 goto out_mem; 6483 6484 /* Initialize all rbd options to the defaults */ 6485 6486 pctx.opts = kzalloc(sizeof(*pctx.opts), GFP_KERNEL); 6487 if (!pctx.opts) 6488 goto out_mem; 6489 6490 pctx.opts->read_only = RBD_READ_ONLY_DEFAULT; 6491 pctx.opts->queue_depth = RBD_QUEUE_DEPTH_DEFAULT; 6492 pctx.opts->alloc_size = RBD_ALLOC_SIZE_DEFAULT; 6493 pctx.opts->lock_timeout = RBD_LOCK_TIMEOUT_DEFAULT; 6494 pctx.opts->lock_on_read = RBD_LOCK_ON_READ_DEFAULT; 6495 pctx.opts->exclusive = RBD_EXCLUSIVE_DEFAULT; 6496 pctx.opts->trim = RBD_TRIM_DEFAULT; 6497 6498 ret = ceph_parse_mon_ips(mon_addrs, mon_addrs_size, pctx.copts, NULL, 6499 ','); 6500 if (ret) 6501 goto out_err; 6502 6503 ret = rbd_parse_options(options, &pctx); 6504 if (ret) 6505 goto out_err; 6506 6507 *ceph_opts = pctx.copts; 6508 *opts = pctx.opts; 6509 *rbd_spec = pctx.spec; 6510 kfree(options); 6511 return 0; 6512 6513 out_mem: 6514 ret = -ENOMEM; 6515 out_err: 6516 kfree(pctx.opts); 6517 ceph_destroy_options(pctx.copts); 6518 rbd_spec_put(pctx.spec); 6519 kfree(options); 6520 return ret; 6521 } 6522 6523 static void rbd_dev_image_unlock(struct rbd_device *rbd_dev) 6524 { 6525 down_write(&rbd_dev->lock_rwsem); 6526 if (__rbd_is_lock_owner(rbd_dev)) 6527 __rbd_release_lock(rbd_dev); 6528 up_write(&rbd_dev->lock_rwsem); 6529 } 6530 6531 /* 6532 * If the wait is interrupted, an error is returned even if the lock 6533 * was successfully acquired. rbd_dev_image_unlock() will release it 6534 * if needed. 6535 */ 6536 static int rbd_add_acquire_lock(struct rbd_device *rbd_dev) 6537 { 6538 long ret; 6539 6540 if (!(rbd_dev->header.features & RBD_FEATURE_EXCLUSIVE_LOCK)) { 6541 if (!rbd_dev->opts->exclusive && !rbd_dev->opts->lock_on_read) 6542 return 0; 6543 6544 rbd_warn(rbd_dev, "exclusive-lock feature is not enabled"); 6545 return -EINVAL; 6546 } 6547 6548 if (rbd_is_ro(rbd_dev)) 6549 return 0; 6550 6551 rbd_assert(!rbd_is_lock_owner(rbd_dev)); 6552 queue_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork, 0); 6553 ret = wait_for_completion_killable_timeout(&rbd_dev->acquire_wait, 6554 ceph_timeout_jiffies(rbd_dev->opts->lock_timeout)); 6555 if (ret > 0) { 6556 ret = rbd_dev->acquire_err; 6557 } else { 6558 cancel_delayed_work_sync(&rbd_dev->lock_dwork); 6559 if (!ret) 6560 ret = -ETIMEDOUT; 6561 } 6562 6563 if (ret) { 6564 rbd_warn(rbd_dev, "failed to acquire exclusive lock: %ld", ret); 6565 return ret; 6566 } 6567 6568 /* 6569 * The lock may have been released by now, unless automatic lock 6570 * transitions are disabled. 6571 */ 6572 rbd_assert(!rbd_dev->opts->exclusive || rbd_is_lock_owner(rbd_dev)); 6573 return 0; 6574 } 6575 6576 /* 6577 * An rbd format 2 image has a unique identifier, distinct from the 6578 * name given to it by the user. Internally, that identifier is 6579 * what's used to specify the names of objects related to the image. 6580 * 6581 * A special "rbd id" object is used to map an rbd image name to its 6582 * id. If that object doesn't exist, then there is no v2 rbd image 6583 * with the supplied name. 6584 * 6585 * This function will record the given rbd_dev's image_id field if 6586 * it can be determined, and in that case will return 0. If any 6587 * errors occur a negative errno will be returned and the rbd_dev's 6588 * image_id field will be unchanged (and should be NULL). 6589 */ 6590 static int rbd_dev_image_id(struct rbd_device *rbd_dev) 6591 { 6592 int ret; 6593 size_t size; 6594 CEPH_DEFINE_OID_ONSTACK(oid); 6595 void *response; 6596 char *image_id; 6597 6598 /* 6599 * When probing a parent image, the image id is already 6600 * known (and the image name likely is not). There's no 6601 * need to fetch the image id again in this case. We 6602 * do still need to set the image format though. 6603 */ 6604 if (rbd_dev->spec->image_id) { 6605 rbd_dev->image_format = *rbd_dev->spec->image_id ? 2 : 1; 6606 6607 return 0; 6608 } 6609 6610 /* 6611 * First, see if the format 2 image id file exists, and if 6612 * so, get the image's persistent id from it. 6613 */ 6614 ret = ceph_oid_aprintf(&oid, GFP_KERNEL, "%s%s", RBD_ID_PREFIX, 6615 rbd_dev->spec->image_name); 6616 if (ret) 6617 return ret; 6618 6619 dout("rbd id object name is %s\n", oid.name); 6620 6621 /* Response will be an encoded string, which includes a length */ 6622 size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX; 6623 response = kzalloc(size, GFP_NOIO); 6624 if (!response) { 6625 ret = -ENOMEM; 6626 goto out; 6627 } 6628 6629 /* If it doesn't exist we'll assume it's a format 1 image */ 6630 6631 ret = rbd_obj_method_sync(rbd_dev, &oid, &rbd_dev->header_oloc, 6632 "get_id", NULL, 0, 6633 response, size); 6634 dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret); 6635 if (ret == -ENOENT) { 6636 image_id = kstrdup("", GFP_KERNEL); 6637 ret = image_id ? 0 : -ENOMEM; 6638 if (!ret) 6639 rbd_dev->image_format = 1; 6640 } else if (ret >= 0) { 6641 void *p = response; 6642 6643 image_id = ceph_extract_encoded_string(&p, p + ret, 6644 NULL, GFP_NOIO); 6645 ret = PTR_ERR_OR_ZERO(image_id); 6646 if (!ret) 6647 rbd_dev->image_format = 2; 6648 } 6649 6650 if (!ret) { 6651 rbd_dev->spec->image_id = image_id; 6652 dout("image_id is %s\n", image_id); 6653 } 6654 out: 6655 kfree(response); 6656 ceph_oid_destroy(&oid); 6657 return ret; 6658 } 6659 6660 /* 6661 * Undo whatever state changes are made by v1 or v2 header info 6662 * call. 6663 */ 6664 static void rbd_dev_unprobe(struct rbd_device *rbd_dev) 6665 { 6666 struct rbd_image_header *header; 6667 6668 rbd_dev_parent_put(rbd_dev); 6669 rbd_object_map_free(rbd_dev); 6670 rbd_dev_mapping_clear(rbd_dev); 6671 6672 /* Free dynamic fields from the header, then zero it out */ 6673 6674 header = &rbd_dev->header; 6675 ceph_put_snap_context(header->snapc); 6676 kfree(header->snap_sizes); 6677 kfree(header->snap_names); 6678 kfree(header->object_prefix); 6679 memset(header, 0, sizeof (*header)); 6680 } 6681 6682 static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev) 6683 { 6684 int ret; 6685 6686 ret = rbd_dev_v2_object_prefix(rbd_dev); 6687 if (ret) 6688 goto out_err; 6689 6690 /* 6691 * Get the and check features for the image. Currently the 6692 * features are assumed to never change. 6693 */ 6694 ret = rbd_dev_v2_features(rbd_dev); 6695 if (ret) 6696 goto out_err; 6697 6698 /* If the image supports fancy striping, get its parameters */ 6699 6700 if (rbd_dev->header.features & RBD_FEATURE_STRIPINGV2) { 6701 ret = rbd_dev_v2_striping_info(rbd_dev); 6702 if (ret < 0) 6703 goto out_err; 6704 } 6705 6706 if (rbd_dev->header.features & RBD_FEATURE_DATA_POOL) { 6707 ret = rbd_dev_v2_data_pool(rbd_dev); 6708 if (ret) 6709 goto out_err; 6710 } 6711 6712 rbd_init_layout(rbd_dev); 6713 return 0; 6714 6715 out_err: 6716 rbd_dev->header.features = 0; 6717 kfree(rbd_dev->header.object_prefix); 6718 rbd_dev->header.object_prefix = NULL; 6719 return ret; 6720 } 6721 6722 /* 6723 * @depth is rbd_dev_image_probe() -> rbd_dev_probe_parent() -> 6724 * rbd_dev_image_probe() recursion depth, which means it's also the 6725 * length of the already discovered part of the parent chain. 6726 */ 6727 static int rbd_dev_probe_parent(struct rbd_device *rbd_dev, int depth) 6728 { 6729 struct rbd_device *parent = NULL; 6730 int ret; 6731 6732 if (!rbd_dev->parent_spec) 6733 return 0; 6734 6735 if (++depth > RBD_MAX_PARENT_CHAIN_LEN) { 6736 pr_info("parent chain is too long (%d)\n", depth); 6737 ret = -EINVAL; 6738 goto out_err; 6739 } 6740 6741 parent = __rbd_dev_create(rbd_dev->rbd_client, rbd_dev->parent_spec); 6742 if (!parent) { 6743 ret = -ENOMEM; 6744 goto out_err; 6745 } 6746 6747 /* 6748 * Images related by parent/child relationships always share 6749 * rbd_client and spec/parent_spec, so bump their refcounts. 6750 */ 6751 __rbd_get_client(rbd_dev->rbd_client); 6752 rbd_spec_get(rbd_dev->parent_spec); 6753 6754 __set_bit(RBD_DEV_FLAG_READONLY, &parent->flags); 6755 6756 ret = rbd_dev_image_probe(parent, depth); 6757 if (ret < 0) 6758 goto out_err; 6759 6760 rbd_dev->parent = parent; 6761 atomic_set(&rbd_dev->parent_ref, 1); 6762 return 0; 6763 6764 out_err: 6765 rbd_dev_unparent(rbd_dev); 6766 rbd_dev_destroy(parent); 6767 return ret; 6768 } 6769 6770 static void rbd_dev_device_release(struct rbd_device *rbd_dev) 6771 { 6772 clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); 6773 rbd_free_disk(rbd_dev); 6774 if (!single_major) 6775 unregister_blkdev(rbd_dev->major, rbd_dev->name); 6776 } 6777 6778 /* 6779 * rbd_dev->header_rwsem must be locked for write and will be unlocked 6780 * upon return. 6781 */ 6782 static int rbd_dev_device_setup(struct rbd_device *rbd_dev) 6783 { 6784 int ret; 6785 6786 /* Record our major and minor device numbers. */ 6787 6788 if (!single_major) { 6789 ret = register_blkdev(0, rbd_dev->name); 6790 if (ret < 0) 6791 goto err_out_unlock; 6792 6793 rbd_dev->major = ret; 6794 rbd_dev->minor = 0; 6795 } else { 6796 rbd_dev->major = rbd_major; 6797 rbd_dev->minor = rbd_dev_id_to_minor(rbd_dev->dev_id); 6798 } 6799 6800 /* Set up the blkdev mapping. */ 6801 6802 ret = rbd_init_disk(rbd_dev); 6803 if (ret) 6804 goto err_out_blkdev; 6805 6806 set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE); 6807 set_disk_ro(rbd_dev->disk, rbd_is_ro(rbd_dev)); 6808 6809 ret = dev_set_name(&rbd_dev->dev, "%d", rbd_dev->dev_id); 6810 if (ret) 6811 goto err_out_disk; 6812 6813 set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags); 6814 up_write(&rbd_dev->header_rwsem); 6815 return 0; 6816 6817 err_out_disk: 6818 rbd_free_disk(rbd_dev); 6819 err_out_blkdev: 6820 if (!single_major) 6821 unregister_blkdev(rbd_dev->major, rbd_dev->name); 6822 err_out_unlock: 6823 up_write(&rbd_dev->header_rwsem); 6824 return ret; 6825 } 6826 6827 static int rbd_dev_header_name(struct rbd_device *rbd_dev) 6828 { 6829 struct rbd_spec *spec = rbd_dev->spec; 6830 int ret; 6831 6832 /* Record the header object name for this rbd image. */ 6833 6834 rbd_assert(rbd_image_format_valid(rbd_dev->image_format)); 6835 if (rbd_dev->image_format == 1) 6836 ret = ceph_oid_aprintf(&rbd_dev->header_oid, GFP_KERNEL, "%s%s", 6837 spec->image_name, RBD_SUFFIX); 6838 else 6839 ret = ceph_oid_aprintf(&rbd_dev->header_oid, GFP_KERNEL, "%s%s", 6840 RBD_HEADER_PREFIX, spec->image_id); 6841 6842 return ret; 6843 } 6844 6845 static void rbd_print_dne(struct rbd_device *rbd_dev, bool is_snap) 6846 { 6847 if (!is_snap) { 6848 pr_info("image %s/%s%s%s does not exist\n", 6849 rbd_dev->spec->pool_name, 6850 rbd_dev->spec->pool_ns ?: "", 6851 rbd_dev->spec->pool_ns ? "/" : "", 6852 rbd_dev->spec->image_name); 6853 } else { 6854 pr_info("snap %s/%s%s%s@%s does not exist\n", 6855 rbd_dev->spec->pool_name, 6856 rbd_dev->spec->pool_ns ?: "", 6857 rbd_dev->spec->pool_ns ? "/" : "", 6858 rbd_dev->spec->image_name, 6859 rbd_dev->spec->snap_name); 6860 } 6861 } 6862 6863 static void rbd_dev_image_release(struct rbd_device *rbd_dev) 6864 { 6865 if (!rbd_is_ro(rbd_dev)) 6866 rbd_unregister_watch(rbd_dev); 6867 6868 rbd_dev_unprobe(rbd_dev); 6869 rbd_dev->image_format = 0; 6870 kfree(rbd_dev->spec->image_id); 6871 rbd_dev->spec->image_id = NULL; 6872 } 6873 6874 /* 6875 * Probe for the existence of the header object for the given rbd 6876 * device. If this image is the one being mapped (i.e., not a 6877 * parent), initiate a watch on its header object before using that 6878 * object to get detailed information about the rbd image. 6879 * 6880 * On success, returns with header_rwsem held for write if called 6881 * with @depth == 0. 6882 */ 6883 static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth) 6884 { 6885 bool need_watch = !rbd_is_ro(rbd_dev); 6886 int ret; 6887 6888 /* 6889 * Get the id from the image id object. Unless there's an 6890 * error, rbd_dev->spec->image_id will be filled in with 6891 * a dynamically-allocated string, and rbd_dev->image_format 6892 * will be set to either 1 or 2. 6893 */ 6894 ret = rbd_dev_image_id(rbd_dev); 6895 if (ret) 6896 return ret; 6897 6898 ret = rbd_dev_header_name(rbd_dev); 6899 if (ret) 6900 goto err_out_format; 6901 6902 if (need_watch) { 6903 ret = rbd_register_watch(rbd_dev); 6904 if (ret) { 6905 if (ret == -ENOENT) 6906 rbd_print_dne(rbd_dev, false); 6907 goto err_out_format; 6908 } 6909 } 6910 6911 if (!depth) 6912 down_write(&rbd_dev->header_rwsem); 6913 6914 ret = rbd_dev_header_info(rbd_dev); 6915 if (ret) { 6916 if (ret == -ENOENT && !need_watch) 6917 rbd_print_dne(rbd_dev, false); 6918 goto err_out_probe; 6919 } 6920 6921 /* 6922 * If this image is the one being mapped, we have pool name and 6923 * id, image name and id, and snap name - need to fill snap id. 6924 * Otherwise this is a parent image, identified by pool, image 6925 * and snap ids - need to fill in names for those ids. 6926 */ 6927 if (!depth) 6928 ret = rbd_spec_fill_snap_id(rbd_dev); 6929 else 6930 ret = rbd_spec_fill_names(rbd_dev); 6931 if (ret) { 6932 if (ret == -ENOENT) 6933 rbd_print_dne(rbd_dev, true); 6934 goto err_out_probe; 6935 } 6936 6937 ret = rbd_dev_mapping_set(rbd_dev); 6938 if (ret) 6939 goto err_out_probe; 6940 6941 if (rbd_is_snap(rbd_dev) && 6942 (rbd_dev->header.features & RBD_FEATURE_OBJECT_MAP)) { 6943 ret = rbd_object_map_load(rbd_dev); 6944 if (ret) 6945 goto err_out_probe; 6946 } 6947 6948 if (rbd_dev->header.features & RBD_FEATURE_LAYERING) { 6949 ret = rbd_dev_v2_parent_info(rbd_dev); 6950 if (ret) 6951 goto err_out_probe; 6952 } 6953 6954 ret = rbd_dev_probe_parent(rbd_dev, depth); 6955 if (ret) 6956 goto err_out_probe; 6957 6958 dout("discovered format %u image, header name is %s\n", 6959 rbd_dev->image_format, rbd_dev->header_oid.name); 6960 return 0; 6961 6962 err_out_probe: 6963 if (!depth) 6964 up_write(&rbd_dev->header_rwsem); 6965 if (need_watch) 6966 rbd_unregister_watch(rbd_dev); 6967 rbd_dev_unprobe(rbd_dev); 6968 err_out_format: 6969 rbd_dev->image_format = 0; 6970 kfree(rbd_dev->spec->image_id); 6971 rbd_dev->spec->image_id = NULL; 6972 return ret; 6973 } 6974 6975 static ssize_t do_rbd_add(struct bus_type *bus, 6976 const char *buf, 6977 size_t count) 6978 { 6979 struct rbd_device *rbd_dev = NULL; 6980 struct ceph_options *ceph_opts = NULL; 6981 struct rbd_options *rbd_opts = NULL; 6982 struct rbd_spec *spec = NULL; 6983 struct rbd_client *rbdc; 6984 int rc; 6985 6986 if (!capable(CAP_SYS_ADMIN)) 6987 return -EPERM; 6988 6989 if (!try_module_get(THIS_MODULE)) 6990 return -ENODEV; 6991 6992 /* parse add command */ 6993 rc = rbd_add_parse_args(buf, &ceph_opts, &rbd_opts, &spec); 6994 if (rc < 0) 6995 goto out; 6996 6997 rbdc = rbd_get_client(ceph_opts); 6998 if (IS_ERR(rbdc)) { 6999 rc = PTR_ERR(rbdc); 7000 goto err_out_args; 7001 } 7002 7003 /* pick the pool */ 7004 rc = ceph_pg_poolid_by_name(rbdc->client->osdc.osdmap, spec->pool_name); 7005 if (rc < 0) { 7006 if (rc == -ENOENT) 7007 pr_info("pool %s does not exist\n", spec->pool_name); 7008 goto err_out_client; 7009 } 7010 spec->pool_id = (u64)rc; 7011 7012 rbd_dev = rbd_dev_create(rbdc, spec, rbd_opts); 7013 if (!rbd_dev) { 7014 rc = -ENOMEM; 7015 goto err_out_client; 7016 } 7017 rbdc = NULL; /* rbd_dev now owns this */ 7018 spec = NULL; /* rbd_dev now owns this */ 7019 rbd_opts = NULL; /* rbd_dev now owns this */ 7020 7021 /* if we are mapping a snapshot it will be a read-only mapping */ 7022 if (rbd_dev->opts->read_only || 7023 strcmp(rbd_dev->spec->snap_name, RBD_SNAP_HEAD_NAME)) 7024 __set_bit(RBD_DEV_FLAG_READONLY, &rbd_dev->flags); 7025 7026 rbd_dev->config_info = kstrdup(buf, GFP_KERNEL); 7027 if (!rbd_dev->config_info) { 7028 rc = -ENOMEM; 7029 goto err_out_rbd_dev; 7030 } 7031 7032 rc = rbd_dev_image_probe(rbd_dev, 0); 7033 if (rc < 0) 7034 goto err_out_rbd_dev; 7035 7036 if (rbd_dev->opts->alloc_size > rbd_dev->layout.object_size) { 7037 rbd_warn(rbd_dev, "alloc_size adjusted to %u", 7038 rbd_dev->layout.object_size); 7039 rbd_dev->opts->alloc_size = rbd_dev->layout.object_size; 7040 } 7041 7042 rc = rbd_dev_device_setup(rbd_dev); 7043 if (rc) 7044 goto err_out_image_probe; 7045 7046 rc = rbd_add_acquire_lock(rbd_dev); 7047 if (rc) 7048 goto err_out_image_lock; 7049 7050 /* Everything's ready. Announce the disk to the world. */ 7051 7052 rc = device_add(&rbd_dev->dev); 7053 if (rc) 7054 goto err_out_image_lock; 7055 7056 rc = device_add_disk(&rbd_dev->dev, rbd_dev->disk, NULL); 7057 if (rc) 7058 goto err_out_cleanup_disk; 7059 7060 spin_lock(&rbd_dev_list_lock); 7061 list_add_tail(&rbd_dev->node, &rbd_dev_list); 7062 spin_unlock(&rbd_dev_list_lock); 7063 7064 pr_info("%s: capacity %llu features 0x%llx\n", rbd_dev->disk->disk_name, 7065 (unsigned long long)get_capacity(rbd_dev->disk) << SECTOR_SHIFT, 7066 rbd_dev->header.features); 7067 rc = count; 7068 out: 7069 module_put(THIS_MODULE); 7070 return rc; 7071 7072 err_out_cleanup_disk: 7073 rbd_free_disk(rbd_dev); 7074 err_out_image_lock: 7075 rbd_dev_image_unlock(rbd_dev); 7076 rbd_dev_device_release(rbd_dev); 7077 err_out_image_probe: 7078 rbd_dev_image_release(rbd_dev); 7079 err_out_rbd_dev: 7080 rbd_dev_destroy(rbd_dev); 7081 err_out_client: 7082 rbd_put_client(rbdc); 7083 err_out_args: 7084 rbd_spec_put(spec); 7085 kfree(rbd_opts); 7086 goto out; 7087 } 7088 7089 static ssize_t add_store(struct bus_type *bus, const char *buf, size_t count) 7090 { 7091 if (single_major) 7092 return -EINVAL; 7093 7094 return do_rbd_add(bus, buf, count); 7095 } 7096 7097 static ssize_t add_single_major_store(struct bus_type *bus, const char *buf, 7098 size_t count) 7099 { 7100 return do_rbd_add(bus, buf, count); 7101 } 7102 7103 static void rbd_dev_remove_parent(struct rbd_device *rbd_dev) 7104 { 7105 while (rbd_dev->parent) { 7106 struct rbd_device *first = rbd_dev; 7107 struct rbd_device *second = first->parent; 7108 struct rbd_device *third; 7109 7110 /* 7111 * Follow to the parent with no grandparent and 7112 * remove it. 7113 */ 7114 while (second && (third = second->parent)) { 7115 first = second; 7116 second = third; 7117 } 7118 rbd_assert(second); 7119 rbd_dev_image_release(second); 7120 rbd_dev_destroy(second); 7121 first->parent = NULL; 7122 first->parent_overlap = 0; 7123 7124 rbd_assert(first->parent_spec); 7125 rbd_spec_put(first->parent_spec); 7126 first->parent_spec = NULL; 7127 } 7128 } 7129 7130 static ssize_t do_rbd_remove(struct bus_type *bus, 7131 const char *buf, 7132 size_t count) 7133 { 7134 struct rbd_device *rbd_dev = NULL; 7135 struct list_head *tmp; 7136 int dev_id; 7137 char opt_buf[6]; 7138 bool force = false; 7139 int ret; 7140 7141 if (!capable(CAP_SYS_ADMIN)) 7142 return -EPERM; 7143 7144 dev_id = -1; 7145 opt_buf[0] = '\0'; 7146 sscanf(buf, "%d %5s", &dev_id, opt_buf); 7147 if (dev_id < 0) { 7148 pr_err("dev_id out of range\n"); 7149 return -EINVAL; 7150 } 7151 if (opt_buf[0] != '\0') { 7152 if (!strcmp(opt_buf, "force")) { 7153 force = true; 7154 } else { 7155 pr_err("bad remove option at '%s'\n", opt_buf); 7156 return -EINVAL; 7157 } 7158 } 7159 7160 ret = -ENOENT; 7161 spin_lock(&rbd_dev_list_lock); 7162 list_for_each(tmp, &rbd_dev_list) { 7163 rbd_dev = list_entry(tmp, struct rbd_device, node); 7164 if (rbd_dev->dev_id == dev_id) { 7165 ret = 0; 7166 break; 7167 } 7168 } 7169 if (!ret) { 7170 spin_lock_irq(&rbd_dev->lock); 7171 if (rbd_dev->open_count && !force) 7172 ret = -EBUSY; 7173 else if (test_and_set_bit(RBD_DEV_FLAG_REMOVING, 7174 &rbd_dev->flags)) 7175 ret = -EINPROGRESS; 7176 spin_unlock_irq(&rbd_dev->lock); 7177 } 7178 spin_unlock(&rbd_dev_list_lock); 7179 if (ret) 7180 return ret; 7181 7182 if (force) { 7183 /* 7184 * Prevent new IO from being queued and wait for existing 7185 * IO to complete/fail. 7186 */ 7187 blk_mq_freeze_queue(rbd_dev->disk->queue); 7188 blk_set_queue_dying(rbd_dev->disk->queue); 7189 } 7190 7191 del_gendisk(rbd_dev->disk); 7192 spin_lock(&rbd_dev_list_lock); 7193 list_del_init(&rbd_dev->node); 7194 spin_unlock(&rbd_dev_list_lock); 7195 device_del(&rbd_dev->dev); 7196 7197 rbd_dev_image_unlock(rbd_dev); 7198 rbd_dev_device_release(rbd_dev); 7199 rbd_dev_image_release(rbd_dev); 7200 rbd_dev_destroy(rbd_dev); 7201 return count; 7202 } 7203 7204 static ssize_t remove_store(struct bus_type *bus, const char *buf, size_t count) 7205 { 7206 if (single_major) 7207 return -EINVAL; 7208 7209 return do_rbd_remove(bus, buf, count); 7210 } 7211 7212 static ssize_t remove_single_major_store(struct bus_type *bus, const char *buf, 7213 size_t count) 7214 { 7215 return do_rbd_remove(bus, buf, count); 7216 } 7217 7218 /* 7219 * create control files in sysfs 7220 * /sys/bus/rbd/... 7221 */ 7222 static int __init rbd_sysfs_init(void) 7223 { 7224 int ret; 7225 7226 ret = device_register(&rbd_root_dev); 7227 if (ret < 0) 7228 return ret; 7229 7230 ret = bus_register(&rbd_bus_type); 7231 if (ret < 0) 7232 device_unregister(&rbd_root_dev); 7233 7234 return ret; 7235 } 7236 7237 static void __exit rbd_sysfs_cleanup(void) 7238 { 7239 bus_unregister(&rbd_bus_type); 7240 device_unregister(&rbd_root_dev); 7241 } 7242 7243 static int __init rbd_slab_init(void) 7244 { 7245 rbd_assert(!rbd_img_request_cache); 7246 rbd_img_request_cache = KMEM_CACHE(rbd_img_request, 0); 7247 if (!rbd_img_request_cache) 7248 return -ENOMEM; 7249 7250 rbd_assert(!rbd_obj_request_cache); 7251 rbd_obj_request_cache = KMEM_CACHE(rbd_obj_request, 0); 7252 if (!rbd_obj_request_cache) 7253 goto out_err; 7254 7255 return 0; 7256 7257 out_err: 7258 kmem_cache_destroy(rbd_img_request_cache); 7259 rbd_img_request_cache = NULL; 7260 return -ENOMEM; 7261 } 7262 7263 static void rbd_slab_exit(void) 7264 { 7265 rbd_assert(rbd_obj_request_cache); 7266 kmem_cache_destroy(rbd_obj_request_cache); 7267 rbd_obj_request_cache = NULL; 7268 7269 rbd_assert(rbd_img_request_cache); 7270 kmem_cache_destroy(rbd_img_request_cache); 7271 rbd_img_request_cache = NULL; 7272 } 7273 7274 static int __init rbd_init(void) 7275 { 7276 int rc; 7277 7278 if (!libceph_compatible(NULL)) { 7279 rbd_warn(NULL, "libceph incompatibility (quitting)"); 7280 return -EINVAL; 7281 } 7282 7283 rc = rbd_slab_init(); 7284 if (rc) 7285 return rc; 7286 7287 /* 7288 * The number of active work items is limited by the number of 7289 * rbd devices * queue depth, so leave @max_active at default. 7290 */ 7291 rbd_wq = alloc_workqueue(RBD_DRV_NAME, WQ_MEM_RECLAIM, 0); 7292 if (!rbd_wq) { 7293 rc = -ENOMEM; 7294 goto err_out_slab; 7295 } 7296 7297 if (single_major) { 7298 rbd_major = register_blkdev(0, RBD_DRV_NAME); 7299 if (rbd_major < 0) { 7300 rc = rbd_major; 7301 goto err_out_wq; 7302 } 7303 } 7304 7305 rc = rbd_sysfs_init(); 7306 if (rc) 7307 goto err_out_blkdev; 7308 7309 if (single_major) 7310 pr_info("loaded (major %d)\n", rbd_major); 7311 else 7312 pr_info("loaded\n"); 7313 7314 return 0; 7315 7316 err_out_blkdev: 7317 if (single_major) 7318 unregister_blkdev(rbd_major, RBD_DRV_NAME); 7319 err_out_wq: 7320 destroy_workqueue(rbd_wq); 7321 err_out_slab: 7322 rbd_slab_exit(); 7323 return rc; 7324 } 7325 7326 static void __exit rbd_exit(void) 7327 { 7328 ida_destroy(&rbd_dev_id_ida); 7329 rbd_sysfs_cleanup(); 7330 if (single_major) 7331 unregister_blkdev(rbd_major, RBD_DRV_NAME); 7332 destroy_workqueue(rbd_wq); 7333 rbd_slab_exit(); 7334 } 7335 7336 module_init(rbd_init); 7337 module_exit(rbd_exit); 7338 7339 MODULE_AUTHOR("Alex Elder <elder@inktank.com>"); 7340 MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); 7341 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); 7342 /* following authorship retained from original osdblk.c */ 7343 MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>"); 7344 7345 MODULE_DESCRIPTION("RADOS Block Device (RBD) driver"); 7346 MODULE_LICENSE("GPL"); 7347