drivers/block/rbd.c

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