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