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/decode.h>
  36 #include <linux/parser.h>
  37 #include <linux/bsearch.h>
  38
  39 #include <linux/kernel.h>
  40 #include <linux/device.h>
  41 #include <linux/module.h>
  42 #include <linux/blk-mq.h>
  43 #include <linux/fs.h>
  44 #include <linux/blkdev.h>
  45 #include <linux/slab.h>
  46 #include <linux/idr.h>
  47 #include <linux/workqueue.h>
  48
  49 #include "rbd_types.h"
  50
  51 #define RBD_DEBUG       /* Activate rbd_assert() calls */
  52
  53 /*
  54  * The basic unit of block I/O is a sector.  It is interpreted in a
  55  * number of contexts in Linux (blk, bio, genhd), but the default is
  56  * universally 512 bytes.  These symbols are just slightly more
  57  * meaningful than the bare numbers they represent.
  58  */
  59 #define SECTOR_SHIFT    9
  60 #define SECTOR_SIZE     (1ULL << SECTOR_SHIFT)
  61
  62 /*
  63  * Increment the given counter and return its updated value.
  64  * If the counter is already 0 it will not be incremented.
  65  * If the counter is already at its maximum value returns
  66  * -EINVAL without updating it.
  67  */
  68 static int atomic_inc_return_safe(atomic_t *v)
  69 {
  70         unsigned int counter;
  71
  72         counter = (unsigned int)__atomic_add_unless(v, 1, 0);
  73         if (counter <= (unsigned int)INT_MAX)
  74                 return (int)counter;
  75
  76         atomic_dec(v);
  77
  78         return -EINVAL;
  79 }
  80
  81 /* Decrement the counter.  Return the resulting value, or -EINVAL */
  82 static int atomic_dec_return_safe(atomic_t *v)
  83 {
  84         int counter;
  85
  86         counter = atomic_dec_return(v);
  87         if (counter >= 0)
  88                 return counter;
  89
  90         atomic_inc(v);
  91
  92         return -EINVAL;
  93 }
  94
  95 #define RBD_DRV_NAME "rbd"
  96
  97 #define RBD_MINORS_PER_MAJOR            256
  98 #define RBD_SINGLE_MAJOR_PART_SHIFT     4
  99
 100 #define RBD_MAX_PARENT_CHAIN_LEN        16
 101
 102 #define RBD_SNAP_DEV_NAME_PREFIX        "snap_"
 103 #define RBD_MAX_SNAP_NAME_LEN   \
 104                         (NAME_MAX - (sizeof (RBD_SNAP_DEV_NAME_PREFIX) - 1))
 105
 106 #define RBD_MAX_SNAP_COUNT      510     /* allows max snapc to fit in 4KB */
 107
 108 #define RBD_SNAP_HEAD_NAME      "-"
 109
 110 #define BAD_SNAP_INDEX  U32_MAX         /* invalid index into snap array */
 111
 112 /* This allows a single page to hold an image name sent by OSD */
 113 #define RBD_IMAGE_NAME_LEN_MAX  (PAGE_SIZE - sizeof (__le32) - 1)
 114 #define RBD_IMAGE_ID_LEN_MAX    64
 115
 116 #define RBD_OBJ_PREFIX_LEN_MAX  64
 117
 118 #define RBD_NOTIFY_TIMEOUT      5       /* seconds */
 119 #define RBD_RETRY_DELAY         msecs_to_jiffies(1000)
 120
 121 /* Feature bits */
 122
 123 #define RBD_FEATURE_LAYERING    (1<<0)
 124 #define RBD_FEATURE_STRIPINGV2  (1<<1)
 125 #define RBD_FEATURE_EXCLUSIVE_LOCK (1<<2)
 126 #define RBD_FEATURES_ALL        (RBD_FEATURE_LAYERING |         \
 127                                  RBD_FEATURE_STRIPINGV2 |       \
 128                                  RBD_FEATURE_EXCLUSIVE_LOCK)
 129
 130 /* Features supported by this (client software) implementation. */
 131
 132 #define RBD_FEATURES_SUPPORTED  (RBD_FEATURES_ALL)
 133
 134 /*
 135  * An RBD device name will be "rbd#", where the "rbd" comes from
 136  * RBD_DRV_NAME above, and # is a unique integer identifier.
 137  */
 138 #define DEV_NAME_LEN            32
 139
 140 /*
 141  * block device image metadata (in-memory version)
 142  */
 143 struct rbd_image_header {
 144         /* These six fields never change for a given rbd image */
 145         char *object_prefix;
 146         __u8 obj_order;
 147         __u8 crypt_type;
 148         __u8 comp_type;
 149         u64 stripe_unit;
 150         u64 stripe_count;
 151         u64 features;           /* Might be changeable someday? */
 152
 153         /* The remaining fields need to be updated occasionally */
 154         u64 image_size;
 155         struct ceph_snap_context *snapc;
 156         char *snap_names;       /* format 1 only */
 157         u64 *snap_sizes;        /* format 1 only */
 158 };
 159
 160 /*
 161  * An rbd image specification.
 162  *
 163  * The tuple (pool_id, image_id, snap_id) is sufficient to uniquely
 164  * identify an image.  Each rbd_dev structure includes a pointer to
 165  * an rbd_spec structure that encapsulates this identity.
 166  *
 167  * Each of the id's in an rbd_spec has an associated name.  For a
 168  * user-mapped image, the names are supplied and the id's associated
 169  * with them are looked up.  For a layered image, a parent image is
 170  * defined by the tuple, and the names are looked up.
 171  *
 172  * An rbd_dev structure contains a parent_spec pointer which is
 173  * non-null if the image it represents is a child in a layered
 174  * image.  This pointer will refer to the rbd_spec structure used
 175  * by the parent rbd_dev for its own identity (i.e., the structure
 176  * is shared between the parent and child).
 177  *
 178  * Since these structures are populated once, during the discovery
 179  * phase of image construction, they are effectively immutable so
 180  * we make no effort to synchronize access to them.
 181  *
 182  * Note that code herein does not assume the image name is known (it
 183  * could be a null pointer).
 184  */
 185 struct rbd_spec {
 186         u64             pool_id;
 187         const char      *pool_name;
 188
 189         const char      *image_id;
 190         const char      *image_name;
 191
 192         u64             snap_id;
 193         const char      *snap_name;
 194
 195         struct kref     kref;
 196 };
 197
 198 /*
 199  * an instance of the client.  multiple devices may share an rbd client.
 200  */
 201 struct rbd_client {
 202         struct ceph_client      *client;
 203         struct kref             kref;
 204         struct list_head        node;
 205 };
 206
 207 struct rbd_img_request;
 208 typedef void (*rbd_img_callback_t)(struct rbd_img_request *);
 209
 210 #define BAD_WHICH       U32_MAX         /* Good which or bad which, which? */
 211
 212 struct rbd_obj_request;
 213 typedef void (*rbd_obj_callback_t)(struct rbd_obj_request *);
 214
 215 enum obj_request_type {
 216         OBJ_REQUEST_NODATA, OBJ_REQUEST_BIO, OBJ_REQUEST_PAGES
 217 };
 218
 219 enum obj_operation_type {
 220         OBJ_OP_WRITE,
 221         OBJ_OP_READ,
 222         OBJ_OP_DISCARD,
 223 };
 224
 225 enum obj_req_flags {
 226         OBJ_REQ_DONE,           /* completion flag: not done = 0, done = 1 */
 227         OBJ_REQ_IMG_DATA,       /* object usage: standalone = 0, image = 1 */
 228         OBJ_REQ_KNOWN,          /* EXISTS flag valid: no = 0, yes = 1 */
 229         OBJ_REQ_EXISTS,         /* target exists: no = 0, yes = 1 */
 230 };
 231
 232 struct rbd_obj_request {
 233         const char              *object_name;
 234         u64                     offset;         /* object start byte */
 235         u64                     length;         /* bytes from offset */
 236         unsigned long           flags;
 237
 238         /*
 239          * An object request associated with an image will have its
 240          * img_data flag set; a standalone object request will not.
 241          *
 242          * A standalone object request will have which == BAD_WHICH
 243          * and a null obj_request pointer.
 244          *
 245          * An object request initiated in support of a layered image
 246          * object (to check for its existence before a write) will
 247          * have which == BAD_WHICH and a non-null obj_request pointer.
 248          *
 249          * Finally, an object request for rbd image data will have
 250          * which != BAD_WHICH, and will have a non-null img_request
 251          * pointer.  The value of which will be in the range
 252          * 0..(img_request->obj_request_count-1).
 253          */
 254         union {
 255                 struct rbd_obj_request  *obj_request;   /* STAT op */
 256                 struct {
 257                         struct rbd_img_request  *img_request;
 258                         u64                     img_offset;
 259                         /* links for img_request->obj_requests list */
 260                         struct list_head        links;
 261                 };
 262         };
 263         u32                     which;          /* posn image request list */
 264
 265         enum obj_request_type   type;
 266         union {
 267                 struct bio      *bio_list;
 268                 struct {
 269                         struct page     **pages;
 270                         u32             page_count;
 271                 };
 272         };
 273         struct page             **copyup_pages;
 274         u32                     copyup_page_count;
 275
 276         struct ceph_osd_request *osd_req;
 277
 278         u64                     xferred;        /* bytes transferred */
 279         int                     result;
 280
 281         rbd_obj_callback_t      callback;
 282         struct completion       completion;
 283
 284         struct kref             kref;
 285 };
 286
 287 enum img_req_flags {
 288         IMG_REQ_WRITE,          /* I/O direction: read = 0, write = 1 */
 289         IMG_REQ_CHILD,          /* initiator: block = 0, child image = 1 */
 290         IMG_REQ_LAYERED,        /* ENOENT handling: normal = 0, layered = 1 */
 291         IMG_REQ_DISCARD,        /* discard: normal = 0, discard request = 1 */
 292 };
 293
 294 struct rbd_img_request {
 295         struct rbd_device       *rbd_dev;
 296         u64                     offset; /* starting image byte offset */
 297         u64                     length; /* byte count from offset */
 298         unsigned long           flags;
 299         union {
 300                 u64                     snap_id;        /* for reads */
 301                 struct ceph_snap_context *snapc;        /* for writes */
 302         };
 303         union {
 304                 struct request          *rq;            /* block request */
 305                 struct rbd_obj_request  *obj_request;   /* obj req initiator */
 306         };
 307         struct page             **copyup_pages;
 308         u32                     copyup_page_count;
 309         spinlock_t              completion_lock;/* protects next_completion */
 310         u32                     next_completion;
 311         rbd_img_callback_t      callback;
 312         u64                     xferred;/* aggregate bytes transferred */
 313         int                     result; /* first nonzero obj_request result */
 314
 315         u32                     obj_request_count;
 316         struct list_head        obj_requests;   /* rbd_obj_request structs */
 317
 318         struct kref             kref;
 319 };
 320
 321 #define for_each_obj_request(ireq, oreq) \
 322         list_for_each_entry(oreq, &(ireq)->obj_requests, links)
 323 #define for_each_obj_request_from(ireq, oreq) \
 324         list_for_each_entry_from(oreq, &(ireq)->obj_requests, links)
 325 #define for_each_obj_request_safe(ireq, oreq, n) \
 326         list_for_each_entry_safe_reverse(oreq, n, &(ireq)->obj_requests, links)
 327
 328 enum rbd_watch_state {
 329         RBD_WATCH_STATE_UNREGISTERED,
 330         RBD_WATCH_STATE_REGISTERED,
 331         RBD_WATCH_STATE_ERROR,
 332 };
 333
 334 enum rbd_lock_state {
 335         RBD_LOCK_STATE_UNLOCKED,
 336         RBD_LOCK_STATE_LOCKED,
 337         RBD_LOCK_STATE_RELEASING,
 338 };
 339
 340 /* WatchNotify::ClientId */
 341 struct rbd_client_id {
 342         u64 gid;
 343         u64 handle;
 344 };
 345
 346 struct rbd_mapping {
 347         u64                     size;
 348         u64                     features;
 349         bool                    read_only;
 350 };
 351
 352 /*
 353  * a single device
 354  */
 355 struct rbd_device {
 356         int                     dev_id;         /* blkdev unique id */
 357
 358         int                     major;          /* blkdev assigned major */
 359         int                     minor;
 360         struct gendisk          *disk;          /* blkdev's gendisk and rq */
 361
 362         u32                     image_format;   /* Either 1 or 2 */
 363         struct rbd_client       *rbd_client;
 364
 365         char                    name[DEV_NAME_LEN]; /* blkdev name, e.g. rbd3 */
 366
 367         spinlock_t              lock;           /* queue, flags, open_count */
 368
 369         struct rbd_image_header header;
 370         unsigned long           flags;          /* possibly lock protected */
 371         struct rbd_spec         *spec;
 372         struct rbd_options      *opts;
 373         char                    *config_info;   /* add{,_single_major} string */
 374
 375         struct ceph_object_id   header_oid;
 376         struct ceph_object_locator header_oloc;
 377
 378         struct ceph_file_layout layout;         /* used for all rbd requests */
 379
 380         struct mutex            watch_mutex;
 381         enum rbd_watch_state    watch_state;
 382         struct ceph_osd_linger_request *watch_handle;
 383         u64                     watch_cookie;
 384         struct delayed_work     watch_dwork;
 385
 386         struct rw_semaphore     lock_rwsem;
 387         enum rbd_lock_state     lock_state;
 388         struct rbd_client_id    owner_cid;
 389         struct work_struct      acquired_lock_work;
 390         struct work_struct      released_lock_work;
 391         struct delayed_work     lock_dwork;
 392         struct work_struct      unlock_work;
 393         wait_queue_head_t       lock_waitq;
 394
 395         struct workqueue_struct *task_wq;
 396
 397         struct rbd_spec         *parent_spec;
 398         u64                     parent_overlap;
 399         atomic_t                parent_ref;
 400         struct rbd_device       *parent;
 401
 402         /* Block layer tags. */
 403         struct blk_mq_tag_set   tag_set;
 404
 405         /* protects updating the header */
 406         struct rw_semaphore     header_rwsem;
 407
 408         struct rbd_mapping      mapping;
 409
 410         struct list_head        node;
 411
 412         /* sysfs related */
 413         struct device           dev;
 414         unsigned long           open_count;     /* protected by lock */
 415 };
 416
 417 /*
 418  * Flag bits for rbd_dev->flags.  If atomicity is required,
 419  * rbd_dev->lock is used to protect access.
 420  *
 421  * Currently, only the "removing" flag (which is coupled with the
 422  * "open_count" field) requires atomic access.
 423  */
 424 enum rbd_dev_flags {
 425         RBD_DEV_FLAG_EXISTS,    /* mapped snapshot has not been deleted */
 426         RBD_DEV_FLAG_REMOVING,  /* this mapping is being removed */
 427 };
 428
 429 static DEFINE_MUTEX(client_mutex);      /* Serialize client creation */
 430
 431 static LIST_HEAD(rbd_dev_list);    /* devices */
 432 static DEFINE_SPINLOCK(rbd_dev_list_lock);
 433
 434 static LIST_HEAD(rbd_client_list);              /* clients */
 435 static DEFINE_SPINLOCK(rbd_client_list_lock);
 436
 437 /* Slab caches for frequently-allocated structures */
 438
 439 static struct kmem_cache        *rbd_img_request_cache;
 440 static struct kmem_cache        *rbd_obj_request_cache;
 441 static struct kmem_cache        *rbd_segment_name_cache;
 442
 443 static int rbd_major;
 444 static DEFINE_IDA(rbd_dev_id_ida);
 445
 446 static struct workqueue_struct *rbd_wq;
 447
 448 /*
 449  * Default to false for now, as single-major requires >= 0.75 version of
 450  * userspace rbd utility.
 451  */
 452 static bool single_major = false;
 453 module_param(single_major, bool, S_IRUGO);
 454 MODULE_PARM_DESC(single_major, "Use a single major number for all rbd devices (default: false)");
 455
 456 static int rbd_img_request_submit(struct rbd_img_request *img_request);
 457
 458 static ssize_t rbd_add(struct bus_type *bus, const char *buf,
 459                        size_t count);
 460 static ssize_t rbd_remove(struct bus_type *bus, const char *buf,
 461                           size_t count);
 462 static ssize_t rbd_add_single_major(struct bus_type *bus, const char *buf,
 463                                     size_t count);
 464 static ssize_t rbd_remove_single_major(struct bus_type *bus, const char *buf,
 465                                        size_t count);
 466 static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth);
 467 static void rbd_spec_put(struct rbd_spec *spec);
 468
 469 static int rbd_dev_id_to_minor(int dev_id)
 470 {
 471         return dev_id << RBD_SINGLE_MAJOR_PART_SHIFT;
 472 }
 473
 474 static int minor_to_rbd_dev_id(int minor)
 475 {
 476         return minor >> RBD_SINGLE_MAJOR_PART_SHIFT;
 477 }
 478
 479 static bool rbd_is_lock_supported(struct rbd_device *rbd_dev)
 480 {
 481         return (rbd_dev->header.features & RBD_FEATURE_EXCLUSIVE_LOCK) &&
 482                rbd_dev->spec->snap_id == CEPH_NOSNAP &&
 483                !rbd_dev->mapping.read_only;
 484 }
 485
 486 static bool __rbd_is_lock_owner(struct rbd_device *rbd_dev)
 487 {
 488         return rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED ||
 489                rbd_dev->lock_state == RBD_LOCK_STATE_RELEASING;
 490 }
 491
 492 static bool rbd_is_lock_owner(struct rbd_device *rbd_dev)
 493 {
 494         bool is_lock_owner;
 495
 496         down_read(&rbd_dev->lock_rwsem);
 497         is_lock_owner = __rbd_is_lock_owner(rbd_dev);
 498         up_read(&rbd_dev->lock_rwsem);
 499         return is_lock_owner;
 500 }
 501
 502 static BUS_ATTR(add, S_IWUSR, NULL, rbd_add);
 503 static BUS_ATTR(remove, S_IWUSR, NULL, rbd_remove);
 504 static BUS_ATTR(add_single_major, S_IWUSR, NULL, rbd_add_single_major);
 505 static BUS_ATTR(remove_single_major, S_IWUSR, NULL, rbd_remove_single_major);
 506
 507 static struct attribute *rbd_bus_attrs[] = {
 508         &bus_attr_add.attr,
 509         &bus_attr_remove.attr,
 510         &bus_attr_add_single_major.attr,
 511         &bus_attr_remove_single_major.attr,
 512         NULL,
 513 };
 514
 515 static umode_t rbd_bus_is_visible(struct kobject *kobj,
 516                                   struct attribute *attr, int index)
 517 {
 518         if (!single_major &&
 519             (attr == &bus_attr_add_single_major.attr ||
 520              attr == &bus_attr_remove_single_major.attr))
 521                 return 0;
 522
 523         return attr->mode;
 524 }
 525
 526 static const struct attribute_group rbd_bus_group = {
 527         .attrs = rbd_bus_attrs,
 528         .is_visible = rbd_bus_is_visible,
 529 };
 530 __ATTRIBUTE_GROUPS(rbd_bus);
 531
 532 static struct bus_type rbd_bus_type = {
 533         .name           = "rbd",
 534         .bus_groups     = rbd_bus_groups,
 535 };
 536
 537 static void rbd_root_dev_release(struct device *dev)
 538 {
 539 }
 540
 541 static struct device rbd_root_dev = {
 542         .init_name =    "rbd",
 543         .release =      rbd_root_dev_release,
 544 };
 545
 546 static __printf(2, 3)
 547 void rbd_warn(struct rbd_device *rbd_dev, const char *fmt, ...)
 548 {
 549         struct va_format vaf;
 550         va_list args;
 551
 552         va_start(args, fmt);
 553         vaf.fmt = fmt;
 554         vaf.va = &args;
 555
 556         if (!rbd_dev)
 557                 printk(KERN_WARNING "%s: %pV\n", RBD_DRV_NAME, &vaf);
 558         else if (rbd_dev->disk)
 559                 printk(KERN_WARNING "%s: %s: %pV\n",
 560                         RBD_DRV_NAME, rbd_dev->disk->disk_name, &vaf);
 561         else if (rbd_dev->spec && rbd_dev->spec->image_name)
 562                 printk(KERN_WARNING "%s: image %s: %pV\n",
 563                         RBD_DRV_NAME, rbd_dev->spec->image_name, &vaf);
 564         else if (rbd_dev->spec && rbd_dev->spec->image_id)
 565                 printk(KERN_WARNING "%s: id %s: %pV\n",
 566                         RBD_DRV_NAME, rbd_dev->spec->image_id, &vaf);
 567         else    /* punt */
 568                 printk(KERN_WARNING "%s: rbd_dev %p: %pV\n",
 569                         RBD_DRV_NAME, rbd_dev, &vaf);
 570         va_end(args);
 571 }
 572
 573 #ifdef RBD_DEBUG
 574 #define rbd_assert(expr)                                                \
 575                 if (unlikely(!(expr))) {                                \
 576                         printk(KERN_ERR "\nAssertion failure in %s() "  \
 577                                                 "at line %d:\n\n"       \
 578                                         "\trbd_assert(%s);\n\n",        \
 579                                         __func__, __LINE__, #expr);     \
 580                         BUG();                                          \
 581                 }
 582 #else /* !RBD_DEBUG */
 583 #  define rbd_assert(expr)      ((void) 0)
 584 #endif /* !RBD_DEBUG */
 585
 586 static void rbd_osd_copyup_callback(struct rbd_obj_request *obj_request);
 587 static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request);
 588 static void rbd_img_parent_read(struct rbd_obj_request *obj_request);
 589 static void rbd_dev_remove_parent(struct rbd_device *rbd_dev);
 590
 591 static int rbd_dev_refresh(struct rbd_device *rbd_dev);
 592 static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev);
 593 static int rbd_dev_header_info(struct rbd_device *rbd_dev);
 594 static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev);
 595 static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev,
 596                                         u64 snap_id);
 597 static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id,
 598                                 u8 *order, u64 *snap_size);
 599 static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id,
 600                 u64 *snap_features);
 601
 602 static int rbd_open(struct block_device *bdev, fmode_t mode)
 603 {
 604         struct rbd_device *rbd_dev = bdev->bd_disk->private_data;
 605         bool removing = false;
 606
 607         if ((mode & FMODE_WRITE) && rbd_dev->mapping.read_only)
 608                 return -EROFS;
 609
 610         spin_lock_irq(&rbd_dev->lock);
 611         if (test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags))
 612                 removing = true;
 613         else
 614                 rbd_dev->open_count++;
 615         spin_unlock_irq(&rbd_dev->lock);
 616         if (removing)
 617                 return -ENOENT;
 618
 619         (void) get_device(&rbd_dev->dev);
 620
 621         return 0;
 622 }
 623
 624 static void rbd_release(struct gendisk *disk, fmode_t mode)
 625 {
 626         struct rbd_device *rbd_dev = disk->private_data;
 627         unsigned long open_count_before;
 628
 629         spin_lock_irq(&rbd_dev->lock);
 630         open_count_before = rbd_dev->open_count--;
 631         spin_unlock_irq(&rbd_dev->lock);
 632         rbd_assert(open_count_before > 0);
 633
 634         put_device(&rbd_dev->dev);
 635 }
 636
 637 static int rbd_ioctl_set_ro(struct rbd_device *rbd_dev, unsigned long arg)
 638 {
 639         int ret = 0;
 640         int val;
 641         bool ro;
 642         bool ro_changed = false;
 643
 644         /* get_user() may sleep, so call it before taking rbd_dev->lock */
 645         if (get_user(val, (int __user *)(arg)))
 646                 return -EFAULT;
 647
 648         ro = val ? true : false;
 649         /* Snapshot doesn't allow to write*/
 650         if (rbd_dev->spec->snap_id != CEPH_NOSNAP && !ro)
 651                 return -EROFS;
 652
 653         spin_lock_irq(&rbd_dev->lock);
 654         /* prevent others open this device */
 655         if (rbd_dev->open_count > 1) {
 656                 ret = -EBUSY;
 657                 goto out;
 658         }
 659
 660         if (rbd_dev->mapping.read_only != ro) {
 661                 rbd_dev->mapping.read_only = ro;
 662                 ro_changed = true;
 663         }
 664
 665 out:
 666         spin_unlock_irq(&rbd_dev->lock);
 667         /* set_disk_ro() may sleep, so call it after releasing rbd_dev->lock */
 668         if (ret == 0 && ro_changed)
 669                 set_disk_ro(rbd_dev->disk, ro ? 1 : 0);
 670
 671         return ret;
 672 }
 673
 674 static int rbd_ioctl(struct block_device *bdev, fmode_t mode,
 675                         unsigned int cmd, unsigned long arg)
 676 {
 677         struct rbd_device *rbd_dev = bdev->bd_disk->private_data;
 678         int ret = 0;
 679
 680         switch (cmd) {
 681         case BLKROSET:
 682                 ret = rbd_ioctl_set_ro(rbd_dev, arg);
 683                 break;
 684         default:
 685                 ret = -ENOTTY;
 686         }
 687
 688         return ret;
 689 }
 690
 691 #ifdef CONFIG_COMPAT
 692 static int rbd_compat_ioctl(struct block_device *bdev, fmode_t mode,
 693                                 unsigned int cmd, unsigned long arg)
 694 {
 695         return rbd_ioctl(bdev, mode, cmd, arg);
 696 }
 697 #endif /* CONFIG_COMPAT */
 698
 699 static const struct block_device_operations rbd_bd_ops = {
 700         .owner                  = THIS_MODULE,
 701         .open                   = rbd_open,
 702         .release                = rbd_release,
 703         .ioctl                  = rbd_ioctl,
 704 #ifdef CONFIG_COMPAT
 705         .compat_ioctl           = rbd_compat_ioctl,
 706 #endif
 707 };
 708
 709 /*
 710  * Initialize an rbd client instance.  Success or not, this function
 711  * consumes ceph_opts.  Caller holds client_mutex.
 712  */
 713 static struct rbd_client *rbd_client_create(struct ceph_options *ceph_opts)
 714 {
 715         struct rbd_client *rbdc;
 716         int ret = -ENOMEM;
 717
 718         dout("%s:\n", __func__);
 719         rbdc = kmalloc(sizeof(struct rbd_client), GFP_KERNEL);
 720         if (!rbdc)
 721                 goto out_opt;
 722
 723         kref_init(&rbdc->kref);
 724         INIT_LIST_HEAD(&rbdc->node);
 725
 726         rbdc->client = ceph_create_client(ceph_opts, rbdc, 0, 0);
 727         if (IS_ERR(rbdc->client))
 728                 goto out_rbdc;
 729         ceph_opts = NULL; /* Now rbdc->client is responsible for ceph_opts */
 730
 731         ret = ceph_open_session(rbdc->client);
 732         if (ret < 0)
 733                 goto out_client;
 734
 735         spin_lock(&rbd_client_list_lock);
 736         list_add_tail(&rbdc->node, &rbd_client_list);
 737         spin_unlock(&rbd_client_list_lock);
 738
 739         dout("%s: rbdc %p\n", __func__, rbdc);
 740
 741         return rbdc;
 742 out_client:
 743         ceph_destroy_client(rbdc->client);
 744 out_rbdc:
 745         kfree(rbdc);
 746 out_opt:
 747         if (ceph_opts)
 748                 ceph_destroy_options(ceph_opts);
 749         dout("%s: error %d\n", __func__, ret);
 750
 751         return ERR_PTR(ret);
 752 }
 753
 754 static struct rbd_client *__rbd_get_client(struct rbd_client *rbdc)
 755 {
 756         kref_get(&rbdc->kref);
 757
 758         return rbdc;
 759 }
 760
 761 /*
 762  * Find a ceph client with specific addr and configuration.  If
 763  * found, bump its reference count.
 764  */
 765 static struct rbd_client *rbd_client_find(struct ceph_options *ceph_opts)
 766 {
 767         struct rbd_client *client_node;
 768         bool found = false;
 769
 770         if (ceph_opts->flags & CEPH_OPT_NOSHARE)
 771                 return NULL;
 772
 773         spin_lock(&rbd_client_list_lock);
 774         list_for_each_entry(client_node, &rbd_client_list, node) {
 775                 if (!ceph_compare_options(ceph_opts, client_node->client)) {
 776                         __rbd_get_client(client_node);
 777
 778                         found = true;
 779                         break;
 780                 }
 781         }
 782         spin_unlock(&rbd_client_list_lock);
 783
 784         return found ? client_node : NULL;
 785 }
 786
 787 /*
 788  * (Per device) rbd map options
 789  */
 790 enum {
 791         Opt_queue_depth,
 792         Opt_last_int,
 793         /* int args above */
 794         Opt_last_string,
 795         /* string args above */
 796         Opt_read_only,
 797         Opt_read_write,
 798         Opt_lock_on_read,
 799         Opt_err
 800 };
 801
 802 static match_table_t rbd_opts_tokens = {
 803         {Opt_queue_depth, "queue_depth=%d"},
 804         /* int args above */
 805         /* string args above */
 806         {Opt_read_only, "read_only"},
 807         {Opt_read_only, "ro"},          /* Alternate spelling */
 808         {Opt_read_write, "read_write"},
 809         {Opt_read_write, "rw"},         /* Alternate spelling */
 810         {Opt_lock_on_read, "lock_on_read"},
 811         {Opt_err, NULL}
 812 };
 813
 814 struct rbd_options {
 815         int     queue_depth;
 816         bool    read_only;
 817         bool    lock_on_read;
 818 };
 819
 820 #define RBD_QUEUE_DEPTH_DEFAULT BLKDEV_MAX_RQ
 821 #define RBD_READ_ONLY_DEFAULT   false
 822 #define RBD_LOCK_ON_READ_DEFAULT false
 823
 824 static int parse_rbd_opts_token(char *c, void *private)
 825 {
 826         struct rbd_options *rbd_opts = private;
 827         substring_t argstr[MAX_OPT_ARGS];
 828         int token, intval, ret;
 829
 830         token = match_token(c, rbd_opts_tokens, argstr);
 831         if (token < Opt_last_int) {
 832                 ret = match_int(&argstr[0], &intval);
 833                 if (ret < 0) {
 834                         pr_err("bad mount option arg (not int) at '%s'\n", c);
 835                         return ret;
 836                 }
 837                 dout("got int token %d val %d\n", token, intval);
 838         } else if (token > Opt_last_int && token < Opt_last_string) {
 839                 dout("got string token %d val %s\n", token, argstr[0].from);
 840         } else {
 841                 dout("got token %d\n", token);
 842         }
 843
 844         switch (token) {
 845         case Opt_queue_depth:
 846                 if (intval < 1) {
 847                         pr_err("queue_depth out of range\n");
 848                         return -EINVAL;
 849                 }
 850                 rbd_opts->queue_depth = intval;
 851                 break;
 852         case Opt_read_only:
 853                 rbd_opts->read_only = true;
 854                 break;
 855         case Opt_read_write:
 856                 rbd_opts->read_only = false;
 857                 break;
 858         case Opt_lock_on_read:
 859                 rbd_opts->lock_on_read = true;
 860                 break;
 861         default:
 862                 /* libceph prints "bad option" msg */
 863                 return -EINVAL;
 864         }
 865
 866         return 0;
 867 }
 868
 869 static char* obj_op_name(enum obj_operation_type op_type)
 870 {
 871         switch (op_type) {
 872         case OBJ_OP_READ:
 873                 return "read";
 874         case OBJ_OP_WRITE:
 875                 return "write";
 876         case OBJ_OP_DISCARD:
 877                 return "discard";
 878         default:
 879                 return "???";
 880         }
 881 }
 882
 883 /*
 884  * Get a ceph client with specific addr and configuration, if one does
 885  * not exist create it.  Either way, ceph_opts is consumed by this
 886  * function.
 887  */
 888 static struct rbd_client *rbd_get_client(struct ceph_options *ceph_opts)
 889 {
 890         struct rbd_client *rbdc;
 891
 892         mutex_lock_nested(&client_mutex, SINGLE_DEPTH_NESTING);
 893         rbdc = rbd_client_find(ceph_opts);
 894         if (rbdc)       /* using an existing client */
 895                 ceph_destroy_options(ceph_opts);
 896         else
 897                 rbdc = rbd_client_create(ceph_opts);
 898         mutex_unlock(&client_mutex);
 899
 900         return rbdc;
 901 }
 902
 903 /*
 904  * Destroy ceph client
 905  *
 906  * Caller must hold rbd_client_list_lock.
 907  */
 908 static void rbd_client_release(struct kref *kref)
 909 {
 910         struct rbd_client *rbdc = container_of(kref, struct rbd_client, kref);
 911
 912         dout("%s: rbdc %p\n", __func__, rbdc);
 913         spin_lock(&rbd_client_list_lock);
 914         list_del(&rbdc->node);
 915         spin_unlock(&rbd_client_list_lock);
 916
 917         ceph_destroy_client(rbdc->client);
 918         kfree(rbdc);
 919 }
 920
 921 /*
 922  * Drop reference to ceph client node. If it's not referenced anymore, release
 923  * it.
 924  */
 925 static void rbd_put_client(struct rbd_client *rbdc)
 926 {
 927         if (rbdc)
 928                 kref_put(&rbdc->kref, rbd_client_release);
 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  * Fill an rbd image header with information from the given format 1
 977  * on-disk header.
 978  */
 979 static int rbd_header_from_disk(struct rbd_device *rbd_dev,
 980                                  struct rbd_image_header_ondisk *ondisk)
 981 {
 982         struct rbd_image_header *header = &rbd_dev->header;
 983         bool first_time = header->object_prefix == NULL;
 984         struct ceph_snap_context *snapc;
 985         char *object_prefix = NULL;
 986         char *snap_names = NULL;
 987         u64 *snap_sizes = NULL;
 988         u32 snap_count;
 989         size_t size;
 990         int ret = -ENOMEM;
 991         u32 i;
 992
 993         /* Allocate this now to avoid having to handle failure below */
 994
 995         if (first_time) {
 996                 size_t len;
 997
 998                 len = strnlen(ondisk->object_prefix,
 999                                 sizeof (ondisk->object_prefix));
1000                 object_prefix = kmalloc(len + 1, GFP_KERNEL);
1001                 if (!object_prefix)
1002                         return -ENOMEM;
1003                 memcpy(object_prefix, ondisk->object_prefix, len);
1004                 object_prefix[len] = '\0';
1005         }
1006
1007         /* Allocate the snapshot context and fill it in */
1008
1009         snap_count = le32_to_cpu(ondisk->snap_count);
1010         snapc = ceph_create_snap_context(snap_count, GFP_KERNEL);
1011         if (!snapc)
1012                 goto out_err;
1013         snapc->seq = le64_to_cpu(ondisk->snap_seq);
1014         if (snap_count) {
1015                 struct rbd_image_snap_ondisk *snaps;
1016                 u64 snap_names_len = le64_to_cpu(ondisk->snap_names_len);
1017
1018                 /* We'll keep a copy of the snapshot names... */
1019
1020                 if (snap_names_len > (u64)SIZE_MAX)
1021                         goto out_2big;
1022                 snap_names = kmalloc(snap_names_len, GFP_KERNEL);
1023                 if (!snap_names)
1024                         goto out_err;
1025
1026                 /* ...as well as the array of their sizes. */
1027
1028                 size = snap_count * sizeof (*header->snap_sizes);
1029                 snap_sizes = kmalloc(size, GFP_KERNEL);
1030                 if (!snap_sizes)
1031                         goto out_err;
1032
1033                 /*
1034                  * Copy the names, and fill in each snapshot's id
1035                  * and size.
1036                  *
1037                  * Note that rbd_dev_v1_header_info() guarantees the
1038                  * ondisk buffer we're working with has
1039                  * snap_names_len bytes beyond the end of the
1040                  * snapshot id array, this memcpy() is safe.
1041                  */
1042                 memcpy(snap_names, &ondisk->snaps[snap_count], snap_names_len);
1043                 snaps = ondisk->snaps;
1044                 for (i = 0; i < snap_count; i++) {
1045                         snapc->snaps[i] = le64_to_cpu(snaps[i].id);
1046                         snap_sizes[i] = le64_to_cpu(snaps[i].image_size);
1047                 }
1048         }
1049
1050         /* We won't fail any more, fill in the header */
1051
1052         if (first_time) {
1053                 header->object_prefix = object_prefix;
1054                 header->obj_order = ondisk->options.order;
1055                 header->crypt_type = ondisk->options.crypt_type;
1056                 header->comp_type = ondisk->options.comp_type;
1057                 /* The rest aren't used for format 1 images */
1058                 header->stripe_unit = 0;
1059                 header->stripe_count = 0;
1060                 header->features = 0;
1061         } else {
1062                 ceph_put_snap_context(header->snapc);
1063                 kfree(header->snap_names);
1064                 kfree(header->snap_sizes);
1065         }
1066
1067         /* The remaining fields always get updated (when we refresh) */
1068
1069         header->image_size = le64_to_cpu(ondisk->image_size);
1070         header->snapc = snapc;
1071         header->snap_names = snap_names;
1072         header->snap_sizes = snap_sizes;
1073
1074         return 0;
1075 out_2big:
1076         ret = -EIO;
1077 out_err:
1078         kfree(snap_sizes);
1079         kfree(snap_names);
1080         ceph_put_snap_context(snapc);
1081         kfree(object_prefix);
1082
1083         return ret;
1084 }
1085
1086 static const char *_rbd_dev_v1_snap_name(struct rbd_device *rbd_dev, u32 which)
1087 {
1088         const char *snap_name;
1089
1090         rbd_assert(which < rbd_dev->header.snapc->num_snaps);
1091
1092         /* Skip over names until we find the one we are looking for */
1093
1094         snap_name = rbd_dev->header.snap_names;
1095         while (which--)
1096                 snap_name += strlen(snap_name) + 1;
1097
1098         return kstrdup(snap_name, GFP_KERNEL);
1099 }
1100
1101 /*
1102  * Snapshot id comparison function for use with qsort()/bsearch().
1103  * Note that result is for snapshots in *descending* order.
1104  */
1105 static int snapid_compare_reverse(const void *s1, const void *s2)
1106 {
1107         u64 snap_id1 = *(u64 *)s1;
1108         u64 snap_id2 = *(u64 *)s2;
1109
1110         if (snap_id1 < snap_id2)
1111                 return 1;
1112         return snap_id1 == snap_id2 ? 0 : -1;
1113 }
1114
1115 /*
1116  * Search a snapshot context to see if the given snapshot id is
1117  * present.
1118  *
1119  * Returns the position of the snapshot id in the array if it's found,
1120  * or BAD_SNAP_INDEX otherwise.
1121  *
1122  * Note: The snapshot array is in kept sorted (by the osd) in
1123  * reverse order, highest snapshot id first.
1124  */
1125 static u32 rbd_dev_snap_index(struct rbd_device *rbd_dev, u64 snap_id)
1126 {
1127         struct ceph_snap_context *snapc = rbd_dev->header.snapc;
1128         u64 *found;
1129
1130         found = bsearch(&snap_id, &snapc->snaps, snapc->num_snaps,
1131                                 sizeof (snap_id), snapid_compare_reverse);
1132
1133         return found ? (u32)(found - &snapc->snaps[0]) : BAD_SNAP_INDEX;
1134 }
1135
1136 static const char *rbd_dev_v1_snap_name(struct rbd_device *rbd_dev,
1137                                         u64 snap_id)
1138 {
1139         u32 which;
1140         const char *snap_name;
1141
1142         which = rbd_dev_snap_index(rbd_dev, snap_id);
1143         if (which == BAD_SNAP_INDEX)
1144                 return ERR_PTR(-ENOENT);
1145
1146         snap_name = _rbd_dev_v1_snap_name(rbd_dev, which);
1147         return snap_name ? snap_name : ERR_PTR(-ENOMEM);
1148 }
1149
1150 static const char *rbd_snap_name(struct rbd_device *rbd_dev, u64 snap_id)
1151 {
1152         if (snap_id == CEPH_NOSNAP)
1153                 return RBD_SNAP_HEAD_NAME;
1154
1155         rbd_assert(rbd_image_format_valid(rbd_dev->image_format));
1156         if (rbd_dev->image_format == 1)
1157                 return rbd_dev_v1_snap_name(rbd_dev, snap_id);
1158
1159         return rbd_dev_v2_snap_name(rbd_dev, snap_id);
1160 }
1161
1162 static int rbd_snap_size(struct rbd_device *rbd_dev, u64 snap_id,
1163                                 u64 *snap_size)
1164 {
1165         rbd_assert(rbd_image_format_valid(rbd_dev->image_format));
1166         if (snap_id == CEPH_NOSNAP) {
1167                 *snap_size = rbd_dev->header.image_size;
1168         } else if (rbd_dev->image_format == 1) {
1169                 u32 which;
1170
1171                 which = rbd_dev_snap_index(rbd_dev, snap_id);
1172                 if (which == BAD_SNAP_INDEX)
1173                         return -ENOENT;
1174
1175                 *snap_size = rbd_dev->header.snap_sizes[which];
1176         } else {
1177                 u64 size = 0;
1178                 int ret;
1179
1180                 ret = _rbd_dev_v2_snap_size(rbd_dev, snap_id, NULL, &size);
1181                 if (ret)
1182                         return ret;
1183
1184                 *snap_size = size;
1185         }
1186         return 0;
1187 }
1188
1189 static int rbd_snap_features(struct rbd_device *rbd_dev, u64 snap_id,
1190                         u64 *snap_features)
1191 {
1192         rbd_assert(rbd_image_format_valid(rbd_dev->image_format));
1193         if (snap_id == CEPH_NOSNAP) {
1194                 *snap_features = rbd_dev->header.features;
1195         } else if (rbd_dev->image_format == 1) {
1196                 *snap_features = 0;     /* No features for format 1 */
1197         } else {
1198                 u64 features = 0;
1199                 int ret;
1200
1201                 ret = _rbd_dev_v2_snap_features(rbd_dev, snap_id, &features);
1202                 if (ret)
1203                         return ret;
1204
1205                 *snap_features = features;
1206         }
1207         return 0;
1208 }
1209
1210 static int rbd_dev_mapping_set(struct rbd_device *rbd_dev)
1211 {
1212         u64 snap_id = rbd_dev->spec->snap_id;
1213         u64 size = 0;
1214         u64 features = 0;
1215         int ret;
1216
1217         ret = rbd_snap_size(rbd_dev, snap_id, &size);
1218         if (ret)
1219                 return ret;
1220         ret = rbd_snap_features(rbd_dev, snap_id, &features);
1221         if (ret)
1222                 return ret;
1223
1224         rbd_dev->mapping.size = size;
1225         rbd_dev->mapping.features = features;
1226
1227         return 0;
1228 }
1229
1230 static void rbd_dev_mapping_clear(struct rbd_device *rbd_dev)
1231 {
1232         rbd_dev->mapping.size = 0;
1233         rbd_dev->mapping.features = 0;
1234 }
1235
1236 static void rbd_segment_name_free(const char *name)
1237 {
1238         /* The explicit cast here is needed to drop the const qualifier */
1239
1240         kmem_cache_free(rbd_segment_name_cache, (void *)name);
1241 }
1242
1243 static const char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset)
1244 {
1245         char *name;
1246         u64 segment;
1247         int ret;
1248         char *name_format;
1249
1250         name = kmem_cache_alloc(rbd_segment_name_cache, GFP_NOIO);
1251         if (!name)
1252                 return NULL;
1253         segment = offset >> rbd_dev->header.obj_order;
1254         name_format = "%s.%012llx";
1255         if (rbd_dev->image_format == 2)
1256                 name_format = "%s.%016llx";
1257         ret = snprintf(name, CEPH_MAX_OID_NAME_LEN + 1, name_format,
1258                         rbd_dev->header.object_prefix, segment);
1259         if (ret < 0 || ret > CEPH_MAX_OID_NAME_LEN) {
1260                 pr_err("error formatting segment name for #%llu (%d)\n",
1261                         segment, ret);
1262                 rbd_segment_name_free(name);
1263                 name = NULL;
1264         }
1265
1266         return name;
1267 }
1268
1269 static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset)
1270 {
1271         u64 segment_size = (u64) 1 << rbd_dev->header.obj_order;
1272
1273         return offset & (segment_size - 1);
1274 }
1275
1276 static u64 rbd_segment_length(struct rbd_device *rbd_dev,
1277                                 u64 offset, u64 length)
1278 {
1279         u64 segment_size = (u64) 1 << rbd_dev->header.obj_order;
1280
1281         offset &= segment_size - 1;
1282
1283         rbd_assert(length <= U64_MAX - offset);
1284         if (offset + length > segment_size)
1285                 length = segment_size - offset;
1286
1287         return length;
1288 }
1289
1290 /*
1291  * returns the size of an object in the image
1292  */
1293 static u64 rbd_obj_bytes(struct rbd_image_header *header)
1294 {
1295         return 1 << header->obj_order;
1296 }
1297
1298 /*
1299  * bio helpers
1300  */
1301
1302 static void bio_chain_put(struct bio *chain)
1303 {
1304         struct bio *tmp;
1305
1306         while (chain) {
1307                 tmp = chain;
1308                 chain = chain->bi_next;
1309                 bio_put(tmp);
1310         }
1311 }
1312
1313 /*
1314  * zeros a bio chain, starting at specific offset
1315  */
1316 static void zero_bio_chain(struct bio *chain, int start_ofs)
1317 {
1318         struct bio_vec bv;
1319         struct bvec_iter iter;
1320         unsigned long flags;
1321         void *buf;
1322         int pos = 0;
1323
1324         while (chain) {
1325                 bio_for_each_segment(bv, chain, iter) {
1326                         if (pos + bv.bv_len > start_ofs) {
1327                                 int remainder = max(start_ofs - pos, 0);
1328                                 buf = bvec_kmap_irq(&bv, &flags);
1329                                 memset(buf + remainder, 0,
1330                                        bv.bv_len - remainder);
1331                                 flush_dcache_page(bv.bv_page);
1332                                 bvec_kunmap_irq(buf, &flags);
1333                         }
1334                         pos += bv.bv_len;
1335                 }
1336
1337                 chain = chain->bi_next;
1338         }
1339 }
1340
1341 /*
1342  * similar to zero_bio_chain(), zeros data defined by a page array,
1343  * starting at the given byte offset from the start of the array and
1344  * continuing up to the given end offset.  The pages array is
1345  * assumed to be big enough to hold all bytes up to the end.
1346  */
1347 static void zero_pages(struct page **pages, u64 offset, u64 end)
1348 {
1349         struct page **page = &pages[offset >> PAGE_SHIFT];
1350
1351         rbd_assert(end > offset);
1352         rbd_assert(end - offset <= (u64)SIZE_MAX);
1353         while (offset < end) {
1354                 size_t page_offset;
1355                 size_t length;
1356                 unsigned long flags;
1357                 void *kaddr;
1358
1359                 page_offset = offset & ~PAGE_MASK;
1360                 length = min_t(size_t, PAGE_SIZE - page_offset, end - offset);
1361                 local_irq_save(flags);
1362                 kaddr = kmap_atomic(*page);
1363                 memset(kaddr + page_offset, 0, length);
1364                 flush_dcache_page(*page);
1365                 kunmap_atomic(kaddr);
1366                 local_irq_restore(flags);
1367
1368                 offset += length;
1369                 page++;
1370         }
1371 }
1372
1373 /*
1374  * Clone a portion of a bio, starting at the given byte offset
1375  * and continuing for the number of bytes indicated.
1376  */
1377 static struct bio *bio_clone_range(struct bio *bio_src,
1378                                         unsigned int offset,
1379                                         unsigned int len,
1380                                         gfp_t gfpmask)
1381 {
1382         struct bio *bio;
1383
1384         bio = bio_clone(bio_src, gfpmask);
1385         if (!bio)
1386                 return NULL;    /* ENOMEM */
1387
1388         bio_advance(bio, offset);
1389         bio->bi_iter.bi_size = len;
1390
1391         return bio;
1392 }
1393
1394 /*
1395  * Clone a portion of a bio chain, starting at the given byte offset
1396  * into the first bio in the source chain and continuing for the
1397  * number of bytes indicated.  The result is another bio chain of
1398  * exactly the given length, or a null pointer on error.
1399  *
1400  * The bio_src and offset parameters are both in-out.  On entry they
1401  * refer to the first source bio and the offset into that bio where
1402  * the start of data to be cloned is located.
1403  *
1404  * On return, bio_src is updated to refer to the bio in the source
1405  * chain that contains first un-cloned byte, and *offset will
1406  * contain the offset of that byte within that bio.
1407  */
1408 static struct bio *bio_chain_clone_range(struct bio **bio_src,
1409                                         unsigned int *offset,
1410                                         unsigned int len,
1411                                         gfp_t gfpmask)
1412 {
1413         struct bio *bi = *bio_src;
1414         unsigned int off = *offset;
1415         struct bio *chain = NULL;
1416         struct bio **end;
1417
1418         /* Build up a chain of clone bios up to the limit */
1419
1420         if (!bi || off >= bi->bi_iter.bi_size || !len)
1421                 return NULL;            /* Nothing to clone */
1422
1423         end = &chain;
1424         while (len) {
1425                 unsigned int bi_size;
1426                 struct bio *bio;
1427
1428                 if (!bi) {
1429                         rbd_warn(NULL, "bio_chain exhausted with %u left", len);
1430                         goto out_err;   /* EINVAL; ran out of bio's */
1431                 }
1432                 bi_size = min_t(unsigned int, bi->bi_iter.bi_size - off, len);
1433                 bio = bio_clone_range(bi, off, bi_size, gfpmask);
1434                 if (!bio)
1435                         goto out_err;   /* ENOMEM */
1436
1437                 *end = bio;
1438                 end = &bio->bi_next;
1439
1440                 off += bi_size;
1441                 if (off == bi->bi_iter.bi_size) {
1442                         bi = bi->bi_next;
1443                         off = 0;
1444                 }
1445                 len -= bi_size;
1446         }
1447         *bio_src = bi;
1448         *offset = off;
1449
1450         return chain;
1451 out_err:
1452         bio_chain_put(chain);
1453
1454         return NULL;
1455 }
1456
1457 /*
1458  * The default/initial value for all object request flags is 0.  For
1459  * each flag, once its value is set to 1 it is never reset to 0
1460  * again.
1461  */
1462 static void obj_request_img_data_set(struct rbd_obj_request *obj_request)
1463 {
1464         if (test_and_set_bit(OBJ_REQ_IMG_DATA, &obj_request->flags)) {
1465                 struct rbd_device *rbd_dev;
1466
1467                 rbd_dev = obj_request->img_request->rbd_dev;
1468                 rbd_warn(rbd_dev, "obj_request %p already marked img_data",
1469                         obj_request);
1470         }
1471 }
1472
1473 static bool obj_request_img_data_test(struct rbd_obj_request *obj_request)
1474 {
1475         smp_mb();
1476         return test_bit(OBJ_REQ_IMG_DATA, &obj_request->flags) != 0;
1477 }
1478
1479 static void obj_request_done_set(struct rbd_obj_request *obj_request)
1480 {
1481         if (test_and_set_bit(OBJ_REQ_DONE, &obj_request->flags)) {
1482                 struct rbd_device *rbd_dev = NULL;
1483
1484                 if (obj_request_img_data_test(obj_request))
1485                         rbd_dev = obj_request->img_request->rbd_dev;
1486                 rbd_warn(rbd_dev, "obj_request %p already marked done",
1487                         obj_request);
1488         }
1489 }
1490
1491 static bool obj_request_done_test(struct rbd_obj_request *obj_request)
1492 {
1493         smp_mb();
1494         return test_bit(OBJ_REQ_DONE, &obj_request->flags) != 0;
1495 }
1496
1497 /*
1498  * This sets the KNOWN flag after (possibly) setting the EXISTS
1499  * flag.  The latter is set based on the "exists" value provided.
1500  *
1501  * Note that for our purposes once an object exists it never goes
1502  * away again.  It's possible that the response from two existence
1503  * checks are separated by the creation of the target object, and
1504  * the first ("doesn't exist") response arrives *after* the second
1505  * ("does exist").  In that case we ignore the second one.
1506  */
1507 static void obj_request_existence_set(struct rbd_obj_request *obj_request,
1508                                 bool exists)
1509 {
1510         if (exists)
1511                 set_bit(OBJ_REQ_EXISTS, &obj_request->flags);
1512         set_bit(OBJ_REQ_KNOWN, &obj_request->flags);
1513         smp_mb();
1514 }
1515
1516 static bool obj_request_known_test(struct rbd_obj_request *obj_request)
1517 {
1518         smp_mb();
1519         return test_bit(OBJ_REQ_KNOWN, &obj_request->flags) != 0;
1520 }
1521
1522 static bool obj_request_exists_test(struct rbd_obj_request *obj_request)
1523 {
1524         smp_mb();
1525         return test_bit(OBJ_REQ_EXISTS, &obj_request->flags) != 0;
1526 }
1527
1528 static bool obj_request_overlaps_parent(struct rbd_obj_request *obj_request)
1529 {
1530         struct rbd_device *rbd_dev = obj_request->img_request->rbd_dev;
1531
1532         return obj_request->img_offset <
1533             round_up(rbd_dev->parent_overlap, rbd_obj_bytes(&rbd_dev->header));
1534 }
1535
1536 static void rbd_obj_request_get(struct rbd_obj_request *obj_request)
1537 {
1538         dout("%s: obj %p (was %d)\n", __func__, obj_request,
1539                 atomic_read(&obj_request->kref.refcount));
1540         kref_get(&obj_request->kref);
1541 }
1542
1543 static void rbd_obj_request_destroy(struct kref *kref);
1544 static void rbd_obj_request_put(struct rbd_obj_request *obj_request)
1545 {
1546         rbd_assert(obj_request != NULL);
1547         dout("%s: obj %p (was %d)\n", __func__, obj_request,
1548                 atomic_read(&obj_request->kref.refcount));
1549         kref_put(&obj_request->kref, rbd_obj_request_destroy);
1550 }
1551
1552 static void rbd_img_request_get(struct rbd_img_request *img_request)
1553 {
1554         dout("%s: img %p (was %d)\n", __func__, img_request,
1555              atomic_read(&img_request->kref.refcount));
1556         kref_get(&img_request->kref);
1557 }
1558
1559 static bool img_request_child_test(struct rbd_img_request *img_request);
1560 static void rbd_parent_request_destroy(struct kref *kref);
1561 static void rbd_img_request_destroy(struct kref *kref);
1562 static void rbd_img_request_put(struct rbd_img_request *img_request)
1563 {
1564         rbd_assert(img_request != NULL);
1565         dout("%s: img %p (was %d)\n", __func__, img_request,
1566                 atomic_read(&img_request->kref.refcount));
1567         if (img_request_child_test(img_request))
1568                 kref_put(&img_request->kref, rbd_parent_request_destroy);
1569         else
1570                 kref_put(&img_request->kref, rbd_img_request_destroy);
1571 }
1572
1573 static inline void rbd_img_obj_request_add(struct rbd_img_request *img_request,
1574                                         struct rbd_obj_request *obj_request)
1575 {
1576         rbd_assert(obj_request->img_request == NULL);
1577
1578         /* Image request now owns object's original reference */
1579         obj_request->img_request = img_request;
1580         obj_request->which = img_request->obj_request_count;
1581         rbd_assert(!obj_request_img_data_test(obj_request));
1582         obj_request_img_data_set(obj_request);
1583         rbd_assert(obj_request->which != BAD_WHICH);
1584         img_request->obj_request_count++;
1585         list_add_tail(&obj_request->links, &img_request->obj_requests);
1586         dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request,
1587                 obj_request->which);
1588 }
1589
1590 static inline void rbd_img_obj_request_del(struct rbd_img_request *img_request,
1591                                         struct rbd_obj_request *obj_request)
1592 {
1593         rbd_assert(obj_request->which != BAD_WHICH);
1594
1595         dout("%s: img %p obj %p w=%u\n", __func__, img_request, obj_request,
1596                 obj_request->which);
1597         list_del(&obj_request->links);
1598         rbd_assert(img_request->obj_request_count > 0);
1599         img_request->obj_request_count--;
1600         rbd_assert(obj_request->which == img_request->obj_request_count);
1601         obj_request->which = BAD_WHICH;
1602         rbd_assert(obj_request_img_data_test(obj_request));
1603         rbd_assert(obj_request->img_request == img_request);
1604         obj_request->img_request = NULL;
1605         obj_request->callback = NULL;
1606         rbd_obj_request_put(obj_request);
1607 }
1608
1609 static bool obj_request_type_valid(enum obj_request_type type)
1610 {
1611         switch (type) {
1612         case OBJ_REQUEST_NODATA:
1613         case OBJ_REQUEST_BIO:
1614         case OBJ_REQUEST_PAGES:
1615                 return true;
1616         default:
1617                 return false;
1618         }
1619 }
1620
1621 static void rbd_obj_request_submit(struct rbd_obj_request *obj_request)
1622 {
1623         struct ceph_osd_request *osd_req = obj_request->osd_req;
1624
1625         dout("%s %p osd_req %p\n", __func__, obj_request, osd_req);
1626         ceph_osdc_start_request(osd_req->r_osdc, osd_req, false);
1627 }
1628
1629 static void rbd_obj_request_end(struct rbd_obj_request *obj_request)
1630 {
1631         dout("%s %p\n", __func__, obj_request);
1632         ceph_osdc_cancel_request(obj_request->osd_req);
1633 }
1634
1635 /*
1636  * Wait for an object request to complete.  If interrupted, cancel the
1637  * underlying osd request.
1638  *
1639  * @timeout: in jiffies, 0 means "wait forever"
1640  */
1641 static int __rbd_obj_request_wait(struct rbd_obj_request *obj_request,
1642                                   unsigned long timeout)
1643 {
1644         long ret;
1645
1646         dout("%s %p\n", __func__, obj_request);
1647         ret = wait_for_completion_interruptible_timeout(
1648                                         &obj_request->completion,
1649                                         ceph_timeout_jiffies(timeout));
1650         if (ret <= 0) {
1651                 if (ret == 0)
1652                         ret = -ETIMEDOUT;
1653                 rbd_obj_request_end(obj_request);
1654         } else {
1655                 ret = 0;
1656         }
1657
1658         dout("%s %p ret %d\n", __func__, obj_request, (int)ret);
1659         return ret;
1660 }
1661
1662 static int rbd_obj_request_wait(struct rbd_obj_request *obj_request)
1663 {
1664         return __rbd_obj_request_wait(obj_request, 0);
1665 }
1666
1667 static void rbd_img_request_complete(struct rbd_img_request *img_request)
1668 {
1669
1670         dout("%s: img %p\n", __func__, img_request);
1671
1672         /*
1673          * If no error occurred, compute the aggregate transfer
1674          * count for the image request.  We could instead use
1675          * atomic64_cmpxchg() to update it as each object request
1676          * completes; not clear which way is better off hand.
1677          */
1678         if (!img_request->result) {
1679                 struct rbd_obj_request *obj_request;
1680                 u64 xferred = 0;
1681
1682                 for_each_obj_request(img_request, obj_request)
1683                         xferred += obj_request->xferred;
1684                 img_request->xferred = xferred;
1685         }
1686
1687         if (img_request->callback)
1688                 img_request->callback(img_request);
1689         else
1690                 rbd_img_request_put(img_request);
1691 }
1692
1693 /*
1694  * The default/initial value for all image request flags is 0.  Each
1695  * is conditionally set to 1 at image request initialization time
1696  * and currently never change thereafter.
1697  */
1698 static void img_request_write_set(struct rbd_img_request *img_request)
1699 {
1700         set_bit(IMG_REQ_WRITE, &img_request->flags);
1701         smp_mb();
1702 }
1703
1704 static bool img_request_write_test(struct rbd_img_request *img_request)
1705 {
1706         smp_mb();
1707         return test_bit(IMG_REQ_WRITE, &img_request->flags) != 0;
1708 }
1709
1710 /*
1711  * Set the discard flag when the img_request is an discard request
1712  */
1713 static void img_request_discard_set(struct rbd_img_request *img_request)
1714 {
1715         set_bit(IMG_REQ_DISCARD, &img_request->flags);
1716         smp_mb();
1717 }
1718
1719 static bool img_request_discard_test(struct rbd_img_request *img_request)
1720 {
1721         smp_mb();
1722         return test_bit(IMG_REQ_DISCARD, &img_request->flags) != 0;
1723 }
1724
1725 static void img_request_child_set(struct rbd_img_request *img_request)
1726 {
1727         set_bit(IMG_REQ_CHILD, &img_request->flags);
1728         smp_mb();
1729 }
1730
1731 static void img_request_child_clear(struct rbd_img_request *img_request)
1732 {
1733         clear_bit(IMG_REQ_CHILD, &img_request->flags);
1734         smp_mb();
1735 }
1736
1737 static bool img_request_child_test(struct rbd_img_request *img_request)
1738 {
1739         smp_mb();
1740         return test_bit(IMG_REQ_CHILD, &img_request->flags) != 0;
1741 }
1742
1743 static void img_request_layered_set(struct rbd_img_request *img_request)
1744 {
1745         set_bit(IMG_REQ_LAYERED, &img_request->flags);
1746         smp_mb();
1747 }
1748
1749 static void img_request_layered_clear(struct rbd_img_request *img_request)
1750 {
1751         clear_bit(IMG_REQ_LAYERED, &img_request->flags);
1752         smp_mb();
1753 }
1754
1755 static bool img_request_layered_test(struct rbd_img_request *img_request)
1756 {
1757         smp_mb();
1758         return test_bit(IMG_REQ_LAYERED, &img_request->flags) != 0;
1759 }
1760
1761 static enum obj_operation_type
1762 rbd_img_request_op_type(struct rbd_img_request *img_request)
1763 {
1764         if (img_request_write_test(img_request))
1765                 return OBJ_OP_WRITE;
1766         else if (img_request_discard_test(img_request))
1767                 return OBJ_OP_DISCARD;
1768         else
1769                 return OBJ_OP_READ;
1770 }
1771
1772 static void
1773 rbd_img_obj_request_read_callback(struct rbd_obj_request *obj_request)
1774 {
1775         u64 xferred = obj_request->xferred;
1776         u64 length = obj_request->length;
1777
1778         dout("%s: obj %p img %p result %d %llu/%llu\n", __func__,
1779                 obj_request, obj_request->img_request, obj_request->result,
1780                 xferred, length);
1781         /*
1782          * ENOENT means a hole in the image.  We zero-fill the entire
1783          * length of the request.  A short read also implies zero-fill
1784          * to the end of the request.  An error requires the whole
1785          * length of the request to be reported finished with an error
1786          * to the block layer.  In each case we update the xferred
1787          * count to indicate the whole request was satisfied.
1788          */
1789         rbd_assert(obj_request->type != OBJ_REQUEST_NODATA);
1790         if (obj_request->result == -ENOENT) {
1791                 if (obj_request->type == OBJ_REQUEST_BIO)
1792                         zero_bio_chain(obj_request->bio_list, 0);
1793                 else
1794                         zero_pages(obj_request->pages, 0, length);
1795                 obj_request->result = 0;
1796         } else if (xferred < length && !obj_request->result) {
1797                 if (obj_request->type == OBJ_REQUEST_BIO)
1798                         zero_bio_chain(obj_request->bio_list, xferred);
1799                 else
1800                         zero_pages(obj_request->pages, xferred, length);
1801         }
1802         obj_request->xferred = length;
1803         obj_request_done_set(obj_request);
1804 }
1805
1806 static void rbd_obj_request_complete(struct rbd_obj_request *obj_request)
1807 {
1808         dout("%s: obj %p cb %p\n", __func__, obj_request,
1809                 obj_request->callback);
1810         if (obj_request->callback)
1811                 obj_request->callback(obj_request);
1812         else
1813                 complete_all(&obj_request->completion);
1814 }
1815
1816 static void rbd_osd_read_callback(struct rbd_obj_request *obj_request)
1817 {
1818         struct rbd_img_request *img_request = NULL;
1819         struct rbd_device *rbd_dev = NULL;
1820         bool layered = false;
1821
1822         if (obj_request_img_data_test(obj_request)) {
1823                 img_request = obj_request->img_request;
1824                 layered = img_request && img_request_layered_test(img_request);
1825                 rbd_dev = img_request->rbd_dev;
1826         }
1827
1828         dout("%s: obj %p img %p result %d %llu/%llu\n", __func__,
1829                 obj_request, img_request, obj_request->result,
1830                 obj_request->xferred, obj_request->length);
1831         if (layered && obj_request->result == -ENOENT &&
1832                         obj_request->img_offset < rbd_dev->parent_overlap)
1833                 rbd_img_parent_read(obj_request);
1834         else if (img_request)
1835                 rbd_img_obj_request_read_callback(obj_request);
1836         else
1837                 obj_request_done_set(obj_request);
1838 }
1839
1840 static void rbd_osd_write_callback(struct rbd_obj_request *obj_request)
1841 {
1842         dout("%s: obj %p result %d %llu\n", __func__, obj_request,
1843                 obj_request->result, obj_request->length);
1844         /*
1845          * There is no such thing as a successful short write.  Set
1846          * it to our originally-requested length.
1847          */
1848         obj_request->xferred = obj_request->length;
1849         obj_request_done_set(obj_request);
1850 }
1851
1852 static void rbd_osd_discard_callback(struct rbd_obj_request *obj_request)
1853 {
1854         dout("%s: obj %p result %d %llu\n", __func__, obj_request,
1855                 obj_request->result, obj_request->length);
1856         /*
1857          * There is no such thing as a successful short discard.  Set
1858          * it to our originally-requested length.
1859          */
1860         obj_request->xferred = obj_request->length;
1861         /* discarding a non-existent object is not a problem */
1862         if (obj_request->result == -ENOENT)
1863                 obj_request->result = 0;
1864         obj_request_done_set(obj_request);
1865 }
1866
1867 /*
1868  * For a simple stat call there's nothing to do.  We'll do more if
1869  * this is part of a write sequence for a layered image.
1870  */
1871 static void rbd_osd_stat_callback(struct rbd_obj_request *obj_request)
1872 {
1873         dout("%s: obj %p\n", __func__, obj_request);
1874         obj_request_done_set(obj_request);
1875 }
1876
1877 static void rbd_osd_call_callback(struct rbd_obj_request *obj_request)
1878 {
1879         dout("%s: obj %p\n", __func__, obj_request);
1880
1881         if (obj_request_img_data_test(obj_request))
1882                 rbd_osd_copyup_callback(obj_request);
1883         else
1884                 obj_request_done_set(obj_request);
1885 }
1886
1887 static void rbd_osd_req_callback(struct ceph_osd_request *osd_req)
1888 {
1889         struct rbd_obj_request *obj_request = osd_req->r_priv;
1890         u16 opcode;
1891
1892         dout("%s: osd_req %p\n", __func__, osd_req);
1893         rbd_assert(osd_req == obj_request->osd_req);
1894         if (obj_request_img_data_test(obj_request)) {
1895                 rbd_assert(obj_request->img_request);
1896                 rbd_assert(obj_request->which != BAD_WHICH);
1897         } else {
1898                 rbd_assert(obj_request->which == BAD_WHICH);
1899         }
1900
1901         if (osd_req->r_result < 0)
1902                 obj_request->result = osd_req->r_result;
1903
1904         /*
1905          * We support a 64-bit length, but ultimately it has to be
1906          * passed to the block layer, which just supports a 32-bit
1907          * length field.
1908          */
1909         obj_request->xferred = osd_req->r_ops[0].outdata_len;
1910         rbd_assert(obj_request->xferred < (u64)UINT_MAX);
1911
1912         opcode = osd_req->r_ops[0].op;
1913         switch (opcode) {
1914         case CEPH_OSD_OP_READ:
1915                 rbd_osd_read_callback(obj_request);
1916                 break;
1917         case CEPH_OSD_OP_SETALLOCHINT:
1918                 rbd_assert(osd_req->r_ops[1].op == CEPH_OSD_OP_WRITE ||
1919                            osd_req->r_ops[1].op == CEPH_OSD_OP_WRITEFULL);
1920                 /* fall through */
1921         case CEPH_OSD_OP_WRITE:
1922         case CEPH_OSD_OP_WRITEFULL:
1923                 rbd_osd_write_callback(obj_request);
1924                 break;
1925         case CEPH_OSD_OP_STAT:
1926                 rbd_osd_stat_callback(obj_request);
1927                 break;
1928         case CEPH_OSD_OP_DELETE:
1929         case CEPH_OSD_OP_TRUNCATE:
1930         case CEPH_OSD_OP_ZERO:
1931                 rbd_osd_discard_callback(obj_request);
1932                 break;
1933         case CEPH_OSD_OP_CALL:
1934                 rbd_osd_call_callback(obj_request);
1935                 break;
1936         default:
1937                 rbd_warn(NULL, "%s: unsupported op %hu",
1938                         obj_request->object_name, (unsigned short) opcode);
1939                 break;
1940         }
1941
1942         if (obj_request_done_test(obj_request))
1943                 rbd_obj_request_complete(obj_request);
1944 }
1945
1946 static void rbd_osd_req_format_read(struct rbd_obj_request *obj_request)
1947 {
1948         struct rbd_img_request *img_request = obj_request->img_request;
1949         struct ceph_osd_request *osd_req = obj_request->osd_req;
1950
1951         if (img_request)
1952                 osd_req->r_snapid = img_request->snap_id;
1953 }
1954
1955 static void rbd_osd_req_format_write(struct rbd_obj_request *obj_request)
1956 {
1957         struct ceph_osd_request *osd_req = obj_request->osd_req;
1958
1959         osd_req->r_mtime = CURRENT_TIME;
1960         osd_req->r_data_offset = obj_request->offset;
1961 }
1962
1963 /*
1964  * Create an osd request.  A read request has one osd op (read).
1965  * A write request has either one (watch) or two (hint+write) osd ops.
1966  * (All rbd data writes are prefixed with an allocation hint op, but
1967  * technically osd watch is a write request, hence this distinction.)
1968  */
1969 static struct ceph_osd_request *rbd_osd_req_create(
1970                                         struct rbd_device *rbd_dev,
1971                                         enum obj_operation_type op_type,
1972                                         unsigned int num_ops,
1973                                         struct rbd_obj_request *obj_request)
1974 {
1975         struct ceph_snap_context *snapc = NULL;
1976         struct ceph_osd_client *osdc;
1977         struct ceph_osd_request *osd_req;
1978
1979         if (obj_request_img_data_test(obj_request) &&
1980                 (op_type == OBJ_OP_DISCARD || op_type == OBJ_OP_WRITE)) {
1981                 struct rbd_img_request *img_request = obj_request->img_request;
1982                 if (op_type == OBJ_OP_WRITE) {
1983                         rbd_assert(img_request_write_test(img_request));
1984                 } else {
1985                         rbd_assert(img_request_discard_test(img_request));
1986                 }
1987                 snapc = img_request->snapc;
1988         }
1989
1990         rbd_assert(num_ops == 1 || ((op_type == OBJ_OP_WRITE) && num_ops == 2));
1991
1992         /* Allocate and initialize the request, for the num_ops ops */
1993
1994         osdc = &rbd_dev->rbd_client->client->osdc;
1995         osd_req = ceph_osdc_alloc_request(osdc, snapc, num_ops, false,
1996                                           GFP_NOIO);
1997         if (!osd_req)
1998                 goto fail;
1999
2000         if (op_type == OBJ_OP_WRITE || op_type == OBJ_OP_DISCARD)
2001                 osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK;
2002         else
2003                 osd_req->r_flags = CEPH_OSD_FLAG_READ;
2004
2005         osd_req->r_callback = rbd_osd_req_callback;
2006         osd_req->r_priv = obj_request;
2007
2008         osd_req->r_base_oloc.pool = rbd_dev->layout.pool_id;
2009         if (ceph_oid_aprintf(&osd_req->r_base_oid, GFP_NOIO, "%s",
2010                              obj_request->object_name))
2011                 goto fail;
2012
2013         if (ceph_osdc_alloc_messages(osd_req, GFP_NOIO))
2014                 goto fail;
2015
2016         return osd_req;
2017
2018 fail:
2019         ceph_osdc_put_request(osd_req);
2020         return NULL;
2021 }
2022
2023 /*
2024  * Create a copyup osd request based on the information in the object
2025  * request supplied.  A copyup request has two or three osd ops, a
2026  * copyup method call, potentially a hint op, and a write or truncate
2027  * or zero op.
2028  */
2029 static struct ceph_osd_request *
2030 rbd_osd_req_create_copyup(struct rbd_obj_request *obj_request)
2031 {
2032         struct rbd_img_request *img_request;
2033         struct ceph_snap_context *snapc;
2034         struct rbd_device *rbd_dev;
2035         struct ceph_osd_client *osdc;
2036         struct ceph_osd_request *osd_req;
2037         int num_osd_ops = 3;
2038
2039         rbd_assert(obj_request_img_data_test(obj_request));
2040         img_request = obj_request->img_request;
2041         rbd_assert(img_request);
2042         rbd_assert(img_request_write_test(img_request) ||
2043                         img_request_discard_test(img_request));
2044
2045         if (img_request_discard_test(img_request))
2046                 num_osd_ops = 2;
2047
2048         /* Allocate and initialize the request, for all the ops */
2049
2050         snapc = img_request->snapc;
2051         rbd_dev = img_request->rbd_dev;
2052         osdc = &rbd_dev->rbd_client->client->osdc;
2053         osd_req = ceph_osdc_alloc_request(osdc, snapc, num_osd_ops,
2054                                                 false, GFP_NOIO);
2055         if (!osd_req)
2056                 goto fail;
2057
2058         osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK;
2059         osd_req->r_callback = rbd_osd_req_callback;
2060         osd_req->r_priv = obj_request;
2061
2062         osd_req->r_base_oloc.pool = rbd_dev->layout.pool_id;
2063         if (ceph_oid_aprintf(&osd_req->r_base_oid, GFP_NOIO, "%s",
2064                              obj_request->object_name))
2065                 goto fail;
2066
2067         if (ceph_osdc_alloc_messages(osd_req, GFP_NOIO))
2068                 goto fail;
2069
2070         return osd_req;
2071
2072 fail:
2073         ceph_osdc_put_request(osd_req);
2074         return NULL;
2075 }
2076
2077
2078 static void rbd_osd_req_destroy(struct ceph_osd_request *osd_req)
2079 {
2080         ceph_osdc_put_request(osd_req);
2081 }
2082
2083 /* object_name is assumed to be a non-null pointer and NUL-terminated */
2084
2085 static struct rbd_obj_request *rbd_obj_request_create(const char *object_name,
2086                                                 u64 offset, u64 length,
2087                                                 enum obj_request_type type)
2088 {
2089         struct rbd_obj_request *obj_request;
2090         size_t size;
2091         char *name;
2092
2093         rbd_assert(obj_request_type_valid(type));
2094
2095         size = strlen(object_name) + 1;
2096         name = kmalloc(size, GFP_NOIO);
2097         if (!name)
2098                 return NULL;
2099
2100         obj_request = kmem_cache_zalloc(rbd_obj_request_cache, GFP_NOIO);
2101         if (!obj_request) {
2102                 kfree(name);
2103                 return NULL;
2104         }
2105
2106         obj_request->object_name = memcpy(name, object_name, size);
2107         obj_request->offset = offset;
2108         obj_request->length = length;
2109         obj_request->flags = 0;
2110         obj_request->which = BAD_WHICH;
2111         obj_request->type = type;
2112         INIT_LIST_HEAD(&obj_request->links);
2113         init_completion(&obj_request->completion);
2114         kref_init(&obj_request->kref);
2115
2116         dout("%s: \"%s\" %llu/%llu %d -> obj %p\n", __func__, object_name,
2117                 offset, length, (int)type, obj_request);
2118
2119         return obj_request;
2120 }
2121
2122 static void rbd_obj_request_destroy(struct kref *kref)
2123 {
2124         struct rbd_obj_request *obj_request;
2125
2126         obj_request = container_of(kref, struct rbd_obj_request, kref);
2127
2128         dout("%s: obj %p\n", __func__, obj_request);
2129
2130         rbd_assert(obj_request->img_request == NULL);
2131         rbd_assert(obj_request->which == BAD_WHICH);
2132
2133         if (obj_request->osd_req)
2134                 rbd_osd_req_destroy(obj_request->osd_req);
2135
2136         rbd_assert(obj_request_type_valid(obj_request->type));
2137         switch (obj_request->type) {
2138         case OBJ_REQUEST_NODATA:
2139                 break;          /* Nothing to do */
2140         case OBJ_REQUEST_BIO:
2141                 if (obj_request->bio_list)
2142                         bio_chain_put(obj_request->bio_list);
2143                 break;
2144         case OBJ_REQUEST_PAGES:
2145                 if (obj_request->pages)
2146                         ceph_release_page_vector(obj_request->pages,
2147                                                 obj_request->page_count);
2148                 break;
2149         }
2150
2151         kfree(obj_request->object_name);
2152         obj_request->object_name = NULL;
2153         kmem_cache_free(rbd_obj_request_cache, obj_request);
2154 }
2155
2156 /* It's OK to call this for a device with no parent */
2157
2158 static void rbd_spec_put(struct rbd_spec *spec);
2159 static void rbd_dev_unparent(struct rbd_device *rbd_dev)
2160 {
2161         rbd_dev_remove_parent(rbd_dev);
2162         rbd_spec_put(rbd_dev->parent_spec);
2163         rbd_dev->parent_spec = NULL;
2164         rbd_dev->parent_overlap = 0;
2165 }
2166
2167 /*
2168  * Parent image reference counting is used to determine when an
2169  * image's parent fields can be safely torn down--after there are no
2170  * more in-flight requests to the parent image.  When the last
2171  * reference is dropped, cleaning them up is safe.
2172  */
2173 static void rbd_dev_parent_put(struct rbd_device *rbd_dev)
2174 {
2175         int counter;
2176
2177         if (!rbd_dev->parent_spec)
2178                 return;
2179
2180         counter = atomic_dec_return_safe(&rbd_dev->parent_ref);
2181         if (counter > 0)
2182                 return;
2183
2184         /* Last reference; clean up parent data structures */
2185
2186         if (!counter)
2187                 rbd_dev_unparent(rbd_dev);
2188         else
2189                 rbd_warn(rbd_dev, "parent reference underflow");
2190 }
2191
2192 /*
2193  * If an image has a non-zero parent overlap, get a reference to its
2194  * parent.
2195  *
2196  * Returns true if the rbd device has a parent with a non-zero
2197  * overlap and a reference for it was successfully taken, or
2198  * false otherwise.
2199  */
2200 static bool rbd_dev_parent_get(struct rbd_device *rbd_dev)
2201 {
2202         int counter = 0;
2203
2204         if (!rbd_dev->parent_spec)
2205                 return false;
2206
2207         down_read(&rbd_dev->header_rwsem);
2208         if (rbd_dev->parent_overlap)
2209                 counter = atomic_inc_return_safe(&rbd_dev->parent_ref);
2210         up_read(&rbd_dev->header_rwsem);
2211
2212         if (counter < 0)
2213                 rbd_warn(rbd_dev, "parent reference overflow");
2214
2215         return counter > 0;
2216 }
2217
2218 /*
2219  * Caller is responsible for filling in the list of object requests
2220  * that comprises the image request, and the Linux request pointer
2221  * (if there is one).
2222  */
2223 static struct rbd_img_request *rbd_img_request_create(
2224                                         struct rbd_device *rbd_dev,
2225                                         u64 offset, u64 length,
2226                                         enum obj_operation_type op_type,
2227                                         struct ceph_snap_context *snapc)
2228 {
2229         struct rbd_img_request *img_request;
2230
2231         img_request = kmem_cache_alloc(rbd_img_request_cache, GFP_NOIO);
2232         if (!img_request)
2233                 return NULL;
2234
2235         img_request->rq = NULL;
2236         img_request->rbd_dev = rbd_dev;
2237         img_request->offset = offset;
2238         img_request->length = length;
2239         img_request->flags = 0;
2240         if (op_type == OBJ_OP_DISCARD) {
2241                 img_request_discard_set(img_request);
2242                 img_request->snapc = snapc;
2243         } else if (op_type == OBJ_OP_WRITE) {
2244                 img_request_write_set(img_request);
2245                 img_request->snapc = snapc;
2246         } else {
2247                 img_request->snap_id = rbd_dev->spec->snap_id;
2248         }
2249         if (rbd_dev_parent_get(rbd_dev))
2250                 img_request_layered_set(img_request);
2251         spin_lock_init(&img_request->completion_lock);
2252         img_request->next_completion = 0;
2253         img_request->callback = NULL;
2254         img_request->result = 0;
2255         img_request->obj_request_count = 0;
2256         INIT_LIST_HEAD(&img_request->obj_requests);
2257         kref_init(&img_request->kref);
2258
2259         dout("%s: rbd_dev %p %s %llu/%llu -> img %p\n", __func__, rbd_dev,
2260                 obj_op_name(op_type), offset, length, img_request);
2261
2262         return img_request;
2263 }
2264
2265 static void rbd_img_request_destroy(struct kref *kref)
2266 {
2267         struct rbd_img_request *img_request;
2268         struct rbd_obj_request *obj_request;
2269         struct rbd_obj_request *next_obj_request;
2270
2271         img_request = container_of(kref, struct rbd_img_request, kref);
2272
2273         dout("%s: img %p\n", __func__, img_request);
2274
2275         for_each_obj_request_safe(img_request, obj_request, next_obj_request)
2276                 rbd_img_obj_request_del(img_request, obj_request);
2277         rbd_assert(img_request->obj_request_count == 0);
2278
2279         if (img_request_layered_test(img_request)) {
2280                 img_request_layered_clear(img_request);
2281                 rbd_dev_parent_put(img_request->rbd_dev);
2282         }
2283
2284         if (img_request_write_test(img_request) ||
2285                 img_request_discard_test(img_request))
2286                 ceph_put_snap_context(img_request->snapc);
2287
2288         kmem_cache_free(rbd_img_request_cache, img_request);
2289 }
2290
2291 static struct rbd_img_request *rbd_parent_request_create(
2292                                         struct rbd_obj_request *obj_request,
2293                                         u64 img_offset, u64 length)
2294 {
2295         struct rbd_img_request *parent_request;
2296         struct rbd_device *rbd_dev;
2297
2298         rbd_assert(obj_request->img_request);
2299         rbd_dev = obj_request->img_request->rbd_dev;
2300
2301         parent_request = rbd_img_request_create(rbd_dev->parent, img_offset,
2302                                                 length, OBJ_OP_READ, NULL);
2303         if (!parent_request)
2304                 return NULL;
2305
2306         img_request_child_set(parent_request);
2307         rbd_obj_request_get(obj_request);
2308         parent_request->obj_request = obj_request;
2309
2310         return parent_request;
2311 }
2312
2313 static void rbd_parent_request_destroy(struct kref *kref)
2314 {
2315         struct rbd_img_request *parent_request;
2316         struct rbd_obj_request *orig_request;
2317
2318         parent_request = container_of(kref, struct rbd_img_request, kref);
2319         orig_request = parent_request->obj_request;
2320
2321         parent_request->obj_request = NULL;
2322         rbd_obj_request_put(orig_request);
2323         img_request_child_clear(parent_request);
2324
2325         rbd_img_request_destroy(kref);
2326 }
2327
2328 static bool rbd_img_obj_end_request(struct rbd_obj_request *obj_request)
2329 {
2330         struct rbd_img_request *img_request;
2331         unsigned int xferred;
2332         int result;
2333         bool more;
2334
2335         rbd_assert(obj_request_img_data_test(obj_request));
2336         img_request = obj_request->img_request;
2337
2338         rbd_assert(obj_request->xferred <= (u64)UINT_MAX);
2339         xferred = (unsigned int)obj_request->xferred;
2340         result = obj_request->result;
2341         if (result) {
2342                 struct rbd_device *rbd_dev = img_request->rbd_dev;
2343                 enum obj_operation_type op_type;
2344
2345                 if (img_request_discard_test(img_request))
2346                         op_type = OBJ_OP_DISCARD;
2347                 else if (img_request_write_test(img_request))
2348                         op_type = OBJ_OP_WRITE;
2349                 else
2350                         op_type = OBJ_OP_READ;
2351
2352                 rbd_warn(rbd_dev, "%s %llx at %llx (%llx)",
2353                         obj_op_name(op_type), obj_request->length,
2354                         obj_request->img_offset, obj_request->offset);
2355                 rbd_warn(rbd_dev, "  result %d xferred %x",
2356                         result, xferred);
2357                 if (!img_request->result)
2358                         img_request->result = result;
2359                 /*
2360                  * Need to end I/O on the entire obj_request worth of
2361                  * bytes in case of error.
2362                  */
2363                 xferred = obj_request->length;
2364         }
2365
2366         /* Image object requests don't own their page array */
2367
2368         if (obj_request->type == OBJ_REQUEST_PAGES) {
2369                 obj_request->pages = NULL;
2370                 obj_request->page_count = 0;
2371         }
2372
2373         if (img_request_child_test(img_request)) {
2374                 rbd_assert(img_request->obj_request != NULL);
2375                 more = obj_request->which < img_request->obj_request_count - 1;
2376         } else {
2377                 rbd_assert(img_request->rq != NULL);
2378
2379                 more = blk_update_request(img_request->rq, result, xferred);
2380                 if (!more)
2381                         __blk_mq_end_request(img_request->rq, result);
2382         }
2383
2384         return more;
2385 }
2386
2387 static void rbd_img_obj_callback(struct rbd_obj_request *obj_request)
2388 {
2389         struct rbd_img_request *img_request;
2390         u32 which = obj_request->which;
2391         bool more = true;
2392
2393         rbd_assert(obj_request_img_data_test(obj_request));
2394         img_request = obj_request->img_request;
2395
2396         dout("%s: img %p obj %p\n", __func__, img_request, obj_request);
2397         rbd_assert(img_request != NULL);
2398         rbd_assert(img_request->obj_request_count > 0);
2399         rbd_assert(which != BAD_WHICH);
2400         rbd_assert(which < img_request->obj_request_count);
2401
2402         spin_lock_irq(&img_request->completion_lock);
2403         if (which != img_request->next_completion)
2404                 goto out;
2405
2406         for_each_obj_request_from(img_request, obj_request) {
2407                 rbd_assert(more);
2408                 rbd_assert(which < img_request->obj_request_count);
2409
2410                 if (!obj_request_done_test(obj_request))
2411                         break;
2412                 more = rbd_img_obj_end_request(obj_request);
2413                 which++;
2414         }
2415
2416         rbd_assert(more ^ (which == img_request->obj_request_count));
2417         img_request->next_completion = which;
2418 out:
2419         spin_unlock_irq(&img_request->completion_lock);
2420         rbd_img_request_put(img_request);
2421
2422         if (!more)
2423                 rbd_img_request_complete(img_request);
2424 }
2425
2426 /*
2427  * Add individual osd ops to the given ceph_osd_request and prepare
2428  * them for submission. num_ops is the current number of
2429  * osd operations already to the object request.
2430  */
2431 static void rbd_img_obj_request_fill(struct rbd_obj_request *obj_request,
2432                                 struct ceph_osd_request *osd_request,
2433                                 enum obj_operation_type op_type,
2434                                 unsigned int num_ops)
2435 {
2436         struct rbd_img_request *img_request = obj_request->img_request;
2437         struct rbd_device *rbd_dev = img_request->rbd_dev;
2438         u64 object_size = rbd_obj_bytes(&rbd_dev->header);
2439         u64 offset = obj_request->offset;
2440         u64 length = obj_request->length;
2441         u64 img_end;
2442         u16 opcode;
2443
2444         if (op_type == OBJ_OP_DISCARD) {
2445                 if (!offset && length == object_size &&
2446                     (!img_request_layered_test(img_request) ||
2447                      !obj_request_overlaps_parent(obj_request))) {
2448                         opcode = CEPH_OSD_OP_DELETE;
2449                 } else if ((offset + length == object_size)) {
2450                         opcode = CEPH_OSD_OP_TRUNCATE;
2451                 } else {
2452                         down_read(&rbd_dev->header_rwsem);
2453                         img_end = rbd_dev->header.image_size;
2454                         up_read(&rbd_dev->header_rwsem);
2455
2456                         if (obj_request->img_offset + length == img_end)
2457                                 opcode = CEPH_OSD_OP_TRUNCATE;
2458                         else
2459                                 opcode = CEPH_OSD_OP_ZERO;
2460                 }
2461         } else if (op_type == OBJ_OP_WRITE) {
2462                 if (!offset && length == object_size)
2463                         opcode = CEPH_OSD_OP_WRITEFULL;
2464                 else
2465                         opcode = CEPH_OSD_OP_WRITE;
2466                 osd_req_op_alloc_hint_init(osd_request, num_ops,
2467                                         object_size, object_size);
2468                 num_ops++;
2469         } else {
2470                 opcode = CEPH_OSD_OP_READ;
2471         }
2472
2473         if (opcode == CEPH_OSD_OP_DELETE)
2474                 osd_req_op_init(osd_request, num_ops, opcode, 0);
2475         else
2476                 osd_req_op_extent_init(osd_request, num_ops, opcode,
2477                                        offset, length, 0, 0);
2478
2479         if (obj_request->type == OBJ_REQUEST_BIO)
2480                 osd_req_op_extent_osd_data_bio(osd_request, num_ops,
2481                                         obj_request->bio_list, length);
2482         else if (obj_request->type == OBJ_REQUEST_PAGES)
2483                 osd_req_op_extent_osd_data_pages(osd_request, num_ops,
2484                                         obj_request->pages, length,
2485                                         offset & ~PAGE_MASK, false, false);
2486
2487         /* Discards are also writes */
2488         if (op_type == OBJ_OP_WRITE || op_type == OBJ_OP_DISCARD)
2489                 rbd_osd_req_format_write(obj_request);
2490         else
2491                 rbd_osd_req_format_read(obj_request);
2492 }
2493
2494 /*
2495  * Split up an image request into one or more object requests, each
2496  * to a different object.  The "type" parameter indicates whether
2497  * "data_desc" is the pointer to the head of a list of bio
2498  * structures, or the base of a page array.  In either case this
2499  * function assumes data_desc describes memory sufficient to hold
2500  * all data described by the image request.
2501  */
2502 static int rbd_img_request_fill(struct rbd_img_request *img_request,
2503                                         enum obj_request_type type,
2504                                         void *data_desc)
2505 {
2506         struct rbd_device *rbd_dev = img_request->rbd_dev;
2507         struct rbd_obj_request *obj_request = NULL;
2508         struct rbd_obj_request *next_obj_request;
2509         struct bio *bio_list = NULL;
2510         unsigned int bio_offset = 0;
2511         struct page **pages = NULL;
2512         enum obj_operation_type op_type;
2513         u64 img_offset;
2514         u64 resid;
2515
2516         dout("%s: img %p type %d data_desc %p\n", __func__, img_request,
2517                 (int)type, data_desc);
2518
2519         img_offset = img_request->offset;
2520         resid = img_request->length;
2521         rbd_assert(resid > 0);
2522         op_type = rbd_img_request_op_type(img_request);
2523
2524         if (type == OBJ_REQUEST_BIO) {
2525                 bio_list = data_desc;
2526                 rbd_assert(img_offset ==
2527                            bio_list->bi_iter.bi_sector << SECTOR_SHIFT);
2528         } else if (type == OBJ_REQUEST_PAGES) {
2529                 pages = data_desc;
2530         }
2531
2532         while (resid) {
2533                 struct ceph_osd_request *osd_req;
2534                 const char *object_name;
2535                 u64 offset;
2536                 u64 length;
2537
2538                 object_name = rbd_segment_name(rbd_dev, img_offset);
2539                 if (!object_name)
2540                         goto out_unwind;
2541                 offset = rbd_segment_offset(rbd_dev, img_offset);
2542                 length = rbd_segment_length(rbd_dev, img_offset, resid);
2543                 obj_request = rbd_obj_request_create(object_name,
2544                                                 offset, length, type);
2545                 /* object request has its own copy of the object name */
2546                 rbd_segment_name_free(object_name);
2547                 if (!obj_request)
2548                         goto out_unwind;
2549
2550                 /*
2551                  * set obj_request->img_request before creating the
2552                  * osd_request so that it gets the right snapc
2553                  */
2554                 rbd_img_obj_request_add(img_request, obj_request);
2555
2556                 if (type == OBJ_REQUEST_BIO) {
2557                         unsigned int clone_size;
2558
2559                         rbd_assert(length <= (u64)UINT_MAX);
2560                         clone_size = (unsigned int)length;
2561                         obj_request->bio_list =
2562                                         bio_chain_clone_range(&bio_list,
2563                                                                 &bio_offset,
2564                                                                 clone_size,
2565                                                                 GFP_NOIO);
2566                         if (!obj_request->bio_list)
2567                                 goto out_unwind;
2568                 } else if (type == OBJ_REQUEST_PAGES) {
2569                         unsigned int page_count;
2570
2571                         obj_request->pages = pages;
2572                         page_count = (u32)calc_pages_for(offset, length);
2573                         obj_request->page_count = page_count;
2574                         if ((offset + length) & ~PAGE_MASK)
2575                                 page_count--;   /* more on last page */
2576                         pages += page_count;
2577                 }
2578
2579                 osd_req = rbd_osd_req_create(rbd_dev, op_type,
2580                                         (op_type == OBJ_OP_WRITE) ? 2 : 1,
2581                                         obj_request);
2582                 if (!osd_req)
2583                         goto out_unwind;
2584
2585                 obj_request->osd_req = osd_req;
2586                 obj_request->callback = rbd_img_obj_callback;
2587                 obj_request->img_offset = img_offset;
2588
2589                 rbd_img_obj_request_fill(obj_request, osd_req, op_type, 0);
2590
2591                 rbd_img_request_get(img_request);
2592
2593                 img_offset += length;
2594                 resid -= length;
2595         }
2596
2597         return 0;
2598
2599 out_unwind:
2600         for_each_obj_request_safe(img_request, obj_request, next_obj_request)
2601                 rbd_img_obj_request_del(img_request, obj_request);
2602
2603         return -ENOMEM;
2604 }
2605
2606 static void
2607 rbd_osd_copyup_callback(struct rbd_obj_request *obj_request)
2608 {
2609         struct rbd_img_request *img_request;
2610         struct rbd_device *rbd_dev;
2611         struct page **pages;
2612         u32 page_count;
2613
2614         dout("%s: obj %p\n", __func__, obj_request);
2615
2616         rbd_assert(obj_request->type == OBJ_REQUEST_BIO ||
2617                 obj_request->type == OBJ_REQUEST_NODATA);
2618         rbd_assert(obj_request_img_data_test(obj_request));
2619         img_request = obj_request->img_request;
2620         rbd_assert(img_request);
2621
2622         rbd_dev = img_request->rbd_dev;
2623         rbd_assert(rbd_dev);
2624
2625         pages = obj_request->copyup_pages;
2626         rbd_assert(pages != NULL);
2627         obj_request->copyup_pages = NULL;
2628         page_count = obj_request->copyup_page_count;
2629         rbd_assert(page_count);
2630         obj_request->copyup_page_count = 0;
2631         ceph_release_page_vector(pages, page_count);
2632
2633         /*
2634          * We want the transfer count to reflect the size of the
2635          * original write request.  There is no such thing as a
2636          * successful short write, so if the request was successful
2637          * we can just set it to the originally-requested length.
2638          */
2639         if (!obj_request->result)
2640                 obj_request->xferred = obj_request->length;
2641
2642         obj_request_done_set(obj_request);
2643 }
2644
2645 static void
2646 rbd_img_obj_parent_read_full_callback(struct rbd_img_request *img_request)
2647 {
2648         struct rbd_obj_request *orig_request;
2649         struct ceph_osd_request *osd_req;
2650         struct rbd_device *rbd_dev;
2651         struct page **pages;
2652         enum obj_operation_type op_type;
2653         u32 page_count;
2654         int img_result;
2655         u64 parent_length;
2656
2657         rbd_assert(img_request_child_test(img_request));
2658
2659         /* First get what we need from the image request */
2660
2661         pages = img_request->copyup_pages;
2662         rbd_assert(pages != NULL);
2663         img_request->copyup_pages = NULL;
2664         page_count = img_request->copyup_page_count;
2665         rbd_assert(page_count);
2666         img_request->copyup_page_count = 0;
2667
2668         orig_request = img_request->obj_request;
2669         rbd_assert(orig_request != NULL);
2670         rbd_assert(obj_request_type_valid(orig_request->type));
2671         img_result = img_request->result;
2672         parent_length = img_request->length;
2673         rbd_assert(parent_length == img_request->xferred);
2674         rbd_img_request_put(img_request);
2675
2676         rbd_assert(orig_request->img_request);
2677         rbd_dev = orig_request->img_request->rbd_dev;
2678         rbd_assert(rbd_dev);
2679
2680         /*
2681          * If the overlap has become 0 (most likely because the
2682          * image has been flattened) we need to free the pages
2683          * and re-submit the original write request.
2684          */
2685         if (!rbd_dev->parent_overlap) {
2686                 ceph_release_page_vector(pages, page_count);
2687                 rbd_obj_request_submit(orig_request);
2688                 return;
2689         }
2690
2691         if (img_result)
2692                 goto out_err;
2693
2694         /*
2695          * The original osd request is of no use to use any more.
2696          * We need a new one that can hold the three ops in a copyup
2697          * request.  Allocate the new copyup osd request for the
2698          * original request, and release the old one.
2699          */
2700         img_result = -ENOMEM;
2701         osd_req = rbd_osd_req_create_copyup(orig_request);
2702         if (!osd_req)
2703                 goto out_err;
2704         rbd_osd_req_destroy(orig_request->osd_req);
2705         orig_request->osd_req = osd_req;
2706         orig_request->copyup_pages = pages;
2707         orig_request->copyup_page_count = page_count;
2708
2709         /* Initialize the copyup op */
2710
2711         osd_req_op_cls_init(osd_req, 0, CEPH_OSD_OP_CALL, "rbd", "copyup");
2712         osd_req_op_cls_request_data_pages(osd_req, 0, pages, parent_length, 0,
2713                                                 false, false);
2714
2715         /* Add the other op(s) */
2716
2717         op_type = rbd_img_request_op_type(orig_request->img_request);
2718         rbd_img_obj_request_fill(orig_request, osd_req, op_type, 1);
2719
2720         /* All set, send it off. */
2721
2722         rbd_obj_request_submit(orig_request);
2723         return;
2724
2725 out_err:
2726         /* Record the error code and complete the request */
2727
2728         orig_request->result = img_result;
2729         orig_request->xferred = 0;
2730         obj_request_done_set(orig_request);
2731         rbd_obj_request_complete(orig_request);
2732 }
2733
2734 /*
2735  * Read from the parent image the range of data that covers the
2736  * entire target of the given object request.  This is used for
2737  * satisfying a layered image write request when the target of an
2738  * object request from the image request does not exist.
2739  *
2740  * A page array big enough to hold the returned data is allocated
2741  * and supplied to rbd_img_request_fill() as the "data descriptor."
2742  * When the read completes, this page array will be transferred to
2743  * the original object request for the copyup operation.
2744  *
2745  * If an error occurs, record it as the result of the original
2746  * object request and mark it done so it gets completed.
2747  */
2748 static int rbd_img_obj_parent_read_full(struct rbd_obj_request *obj_request)
2749 {
2750         struct rbd_device *rbd_dev = obj_request->img_request->rbd_dev;
2751         struct rbd_img_request *parent_request = NULL;
2752         u64 img_offset;
2753         u64 length;
2754         struct page **pages = NULL;
2755         u32 page_count;
2756         int result;
2757
2758         rbd_assert(rbd_dev->parent != NULL);
2759
2760         /*
2761          * Determine the byte range covered by the object in the
2762          * child image to which the original request was to be sent.
2763          */
2764         img_offset = obj_request->img_offset - obj_request->offset;
2765         length = (u64)1 << rbd_dev->header.obj_order;
2766
2767         /*
2768          * There is no defined parent data beyond the parent
2769          * overlap, so limit what we read at that boundary if
2770          * necessary.
2771          */
2772         if (img_offset + length > rbd_dev->parent_overlap) {
2773                 rbd_assert(img_offset < rbd_dev->parent_overlap);
2774                 length = rbd_dev->parent_overlap - img_offset;
2775         }
2776
2777         /*
2778          * Allocate a page array big enough to receive the data read
2779          * from the parent.
2780          */
2781         page_count = (u32)calc_pages_for(0, length);
2782         pages = ceph_alloc_page_vector(page_count, GFP_KERNEL);
2783         if (IS_ERR(pages)) {
2784                 result = PTR_ERR(pages);
2785                 pages = NULL;
2786                 goto out_err;
2787         }
2788
2789         result = -ENOMEM;
2790         parent_request = rbd_parent_request_create(obj_request,
2791                                                 img_offset, length);
2792         if (!parent_request)
2793                 goto out_err;
2794
2795         result = rbd_img_request_fill(parent_request, OBJ_REQUEST_PAGES, pages);
2796         if (result)
2797                 goto out_err;
2798
2799         parent_request->copyup_pages = pages;
2800         parent_request->copyup_page_count = page_count;
2801         parent_request->callback = rbd_img_obj_parent_read_full_callback;
2802
2803         result = rbd_img_request_submit(parent_request);
2804         if (!result)
2805                 return 0;
2806
2807         parent_request->copyup_pages = NULL;
2808         parent_request->copyup_page_count = 0;
2809         parent_request->obj_request = NULL;
2810         rbd_obj_request_put(obj_request);
2811 out_err:
2812         if (pages)
2813                 ceph_release_page_vector(pages, page_count);
2814         if (parent_request)
2815                 rbd_img_request_put(parent_request);
2816         obj_request->result = result;
2817         obj_request->xferred = 0;
2818         obj_request_done_set(obj_request);
2819
2820         return result;
2821 }
2822
2823 static void rbd_img_obj_exists_callback(struct rbd_obj_request *obj_request)
2824 {
2825         struct rbd_obj_request *orig_request;
2826         struct rbd_device *rbd_dev;
2827         int result;
2828
2829         rbd_assert(!obj_request_img_data_test(obj_request));
2830
2831         /*
2832          * All we need from the object request is the original
2833          * request and the result of the STAT op.  Grab those, then
2834          * we're done with the request.
2835          */
2836         orig_request = obj_request->obj_request;
2837         obj_request->obj_request = NULL;
2838         rbd_obj_request_put(orig_request);
2839         rbd_assert(orig_request);
2840         rbd_assert(orig_request->img_request);
2841
2842         result = obj_request->result;
2843         obj_request->result = 0;
2844
2845         dout("%s: obj %p for obj %p result %d %llu/%llu\n", __func__,
2846                 obj_request, orig_request, result,
2847                 obj_request->xferred, obj_request->length);
2848         rbd_obj_request_put(obj_request);
2849
2850         /*
2851          * If the overlap has become 0 (most likely because the
2852          * image has been flattened) we need to re-submit the
2853          * original request.
2854          */
2855         rbd_dev = orig_request->img_request->rbd_dev;
2856         if (!rbd_dev->parent_overlap) {
2857                 rbd_obj_request_submit(orig_request);
2858                 return;
2859         }
2860
2861         /*
2862          * Our only purpose here is to determine whether the object
2863          * exists, and we don't want to treat the non-existence as
2864          * an error.  If something else comes back, transfer the
2865          * error to the original request and complete it now.
2866          */
2867         if (!result) {
2868                 obj_request_existence_set(orig_request, true);
2869         } else if (result == -ENOENT) {
2870                 obj_request_existence_set(orig_request, false);
2871         } else if (result) {
2872                 orig_request->result = result;
2873                 goto out;
2874         }
2875
2876         /*
2877          * Resubmit the original request now that we have recorded
2878          * whether the target object exists.
2879          */
2880         orig_request->result = rbd_img_obj_request_submit(orig_request);
2881 out:
2882         if (orig_request->result)
2883                 rbd_obj_request_complete(orig_request);
2884 }
2885
2886 static int rbd_img_obj_exists_submit(struct rbd_obj_request *obj_request)
2887 {
2888         struct rbd_device *rbd_dev = obj_request->img_request->rbd_dev;
2889         struct rbd_obj_request *stat_request;
2890         struct page **pages = NULL;
2891         u32 page_count;
2892         size_t size;
2893         int ret;
2894
2895         /*
2896          * The response data for a STAT call consists of:
2897          *     le64 length;
2898          *     struct {
2899          *         le32 tv_sec;
2900          *         le32 tv_nsec;
2901          *     } mtime;
2902          */
2903         size = sizeof (__le64) + sizeof (__le32) + sizeof (__le32);
2904         page_count = (u32)calc_pages_for(0, size);
2905         pages = ceph_alloc_page_vector(page_count, GFP_KERNEL);
2906         if (IS_ERR(pages))
2907                 return PTR_ERR(pages);
2908
2909         ret = -ENOMEM;
2910         stat_request = rbd_obj_request_create(obj_request->object_name, 0, 0,
2911                                                         OBJ_REQUEST_PAGES);
2912         if (!stat_request)
2913                 goto out;
2914
2915         rbd_obj_request_get(obj_request);
2916         stat_request->obj_request = obj_request;
2917         stat_request->pages = pages;
2918         stat_request->page_count = page_count;
2919
2920         stat_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1,
2921                                                    stat_request);
2922         if (!stat_request->osd_req)
2923                 goto out;
2924         stat_request->callback = rbd_img_obj_exists_callback;
2925
2926         osd_req_op_init(stat_request->osd_req, 0, CEPH_OSD_OP_STAT, 0);
2927         osd_req_op_raw_data_in_pages(stat_request->osd_req, 0, pages, size, 0,
2928                                         false, false);
2929         rbd_osd_req_format_read(stat_request);
2930
2931         rbd_obj_request_submit(stat_request);
2932         return 0;
2933
2934 out:
2935         if (ret)
2936                 rbd_obj_request_put(obj_request);
2937
2938         return ret;
2939 }
2940
2941 static bool img_obj_request_simple(struct rbd_obj_request *obj_request)
2942 {
2943         struct rbd_img_request *img_request = obj_request->img_request;
2944         struct rbd_device *rbd_dev = img_request->rbd_dev;
2945
2946         /* Reads */
2947         if (!img_request_write_test(img_request) &&
2948             !img_request_discard_test(img_request))
2949                 return true;
2950
2951         /* Non-layered writes */
2952         if (!img_request_layered_test(img_request))
2953                 return true;
2954
2955         /*
2956          * Layered writes outside of the parent overlap range don't
2957          * share any data with the parent.
2958          */
2959         if (!obj_request_overlaps_parent(obj_request))
2960                 return true;
2961
2962         /*
2963          * Entire-object layered writes - we will overwrite whatever
2964          * parent data there is anyway.
2965          */
2966         if (!obj_request->offset &&
2967             obj_request->length == rbd_obj_bytes(&rbd_dev->header))
2968                 return true;
2969
2970         /*
2971          * If the object is known to already exist, its parent data has
2972          * already been copied.
2973          */
2974         if (obj_request_known_test(obj_request) &&
2975             obj_request_exists_test(obj_request))
2976                 return true;
2977
2978         return false;
2979 }
2980
2981 static int rbd_img_obj_request_submit(struct rbd_obj_request *obj_request)
2982 {
2983         rbd_assert(obj_request_img_data_test(obj_request));
2984         rbd_assert(obj_request_type_valid(obj_request->type));
2985         rbd_assert(obj_request->img_request);
2986
2987         if (img_obj_request_simple(obj_request)) {
2988                 rbd_obj_request_submit(obj_request);
2989                 return 0;
2990         }
2991
2992         /*
2993          * It's a layered write.  The target object might exist but
2994          * we may not know that yet.  If we know it doesn't exist,
2995          * start by reading the data for the full target object from
2996          * the parent so we can use it for a copyup to the target.
2997          */
2998         if (obj_request_known_test(obj_request))
2999                 return rbd_img_obj_parent_read_full(obj_request);
3000
3001         /* We don't know whether the target exists.  Go find out. */
3002
3003         return rbd_img_obj_exists_submit(obj_request);
3004 }
3005
3006 static int rbd_img_request_submit(struct rbd_img_request *img_request)
3007 {
3008         struct rbd_obj_request *obj_request;
3009         struct rbd_obj_request *next_obj_request;
3010         int ret = 0;
3011
3012         dout("%s: img %p\n", __func__, img_request);
3013
3014         rbd_img_request_get(img_request);
3015         for_each_obj_request_safe(img_request, obj_request, next_obj_request) {
3016                 ret = rbd_img_obj_request_submit(obj_request);
3017                 if (ret)
3018                         goto out_put_ireq;
3019         }
3020
3021 out_put_ireq:
3022         rbd_img_request_put(img_request);
3023         return ret;
3024 }
3025
3026 static void rbd_img_parent_read_callback(struct rbd_img_request *img_request)
3027 {
3028         struct rbd_obj_request *obj_request;
3029         struct rbd_device *rbd_dev;
3030         u64 obj_end;
3031         u64 img_xferred;
3032         int img_result;
3033
3034         rbd_assert(img_request_child_test(img_request));
3035
3036         /* First get what we need from the image request and release it */
3037
3038         obj_request = img_request->obj_request;
3039         img_xferred = img_request->xferred;
3040         img_result = img_request->result;
3041         rbd_img_request_put(img_request);
3042
3043         /*
3044          * If the overlap has become 0 (most likely because the
3045          * image has been flattened) we need to re-submit the
3046          * original request.
3047          */
3048         rbd_assert(obj_request);
3049         rbd_assert(obj_request->img_request);
3050         rbd_dev = obj_request->img_request->rbd_dev;
3051         if (!rbd_dev->parent_overlap) {
3052                 rbd_obj_request_submit(obj_request);
3053                 return;
3054         }
3055
3056         obj_request->result = img_result;
3057         if (obj_request->result)
3058                 goto out;
3059
3060         /*
3061          * We need to zero anything beyond the parent overlap
3062          * boundary.  Since rbd_img_obj_request_read_callback()
3063          * will zero anything beyond the end of a short read, an
3064          * easy way to do this is to pretend the data from the
3065          * parent came up short--ending at the overlap boundary.
3066          */
3067         rbd_assert(obj_request->img_offset < U64_MAX - obj_request->length);
3068         obj_end = obj_request->img_offset + obj_request->length;
3069         if (obj_end > rbd_dev->parent_overlap) {
3070                 u64 xferred = 0;
3071
3072                 if (obj_request->img_offset < rbd_dev->parent_overlap)
3073                         xferred = rbd_dev->parent_overlap -
3074                                         obj_request->img_offset;
3075
3076                 obj_request->xferred = min(img_xferred, xferred);
3077         } else {
3078                 obj_request->xferred = img_xferred;
3079         }
3080 out:
3081         rbd_img_obj_request_read_callback(obj_request);
3082         rbd_obj_request_complete(obj_request);
3083 }
3084
3085 static void rbd_img_parent_read(struct rbd_obj_request *obj_request)
3086 {
3087         struct rbd_img_request *img_request;
3088         int result;
3089
3090         rbd_assert(obj_request_img_data_test(obj_request));
3091         rbd_assert(obj_request->img_request != NULL);
3092         rbd_assert(obj_request->result == (s32) -ENOENT);
3093         rbd_assert(obj_request_type_valid(obj_request->type));
3094
3095         /* rbd_read_finish(obj_request, obj_request->length); */
3096         img_request = rbd_parent_request_create(obj_request,
3097                                                 obj_request->img_offset,
3098                                                 obj_request->length);
3099         result = -ENOMEM;
3100         if (!img_request)
3101                 goto out_err;
3102
3103         if (obj_request->type == OBJ_REQUEST_BIO)
3104                 result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO,
3105                                                 obj_request->bio_list);
3106         else
3107                 result = rbd_img_request_fill(img_request, OBJ_REQUEST_PAGES,
3108                                                 obj_request->pages);
3109         if (result)
3110                 goto out_err;
3111
3112         img_request->callback = rbd_img_parent_read_callback;
3113         result = rbd_img_request_submit(img_request);
3114         if (result)
3115                 goto out_err;
3116
3117         return;
3118 out_err:
3119         if (img_request)
3120                 rbd_img_request_put(img_request);
3121         obj_request->result = result;
3122         obj_request->xferred = 0;
3123         obj_request_done_set(obj_request);
3124 }
3125
3126 static const struct rbd_client_id rbd_empty_cid;
3127
3128 static bool rbd_cid_equal(const struct rbd_client_id *lhs,
3129                           const struct rbd_client_id *rhs)
3130 {
3131         return lhs->gid == rhs->gid && lhs->handle == rhs->handle;
3132 }
3133
3134 static struct rbd_client_id rbd_get_cid(struct rbd_device *rbd_dev)
3135 {
3136         struct rbd_client_id cid;
3137
3138         mutex_lock(&rbd_dev->watch_mutex);
3139         cid.gid = ceph_client_gid(rbd_dev->rbd_client->client);
3140         cid.handle = rbd_dev->watch_cookie;
3141         mutex_unlock(&rbd_dev->watch_mutex);
3142         return cid;
3143 }
3144
3145 /*
3146  * lock_rwsem must be held for write
3147  */
3148 static void rbd_set_owner_cid(struct rbd_device *rbd_dev,
3149                               const struct rbd_client_id *cid)
3150 {
3151         dout("%s rbd_dev %p %llu-%llu -> %llu-%llu\n", __func__, rbd_dev,
3152              rbd_dev->owner_cid.gid, rbd_dev->owner_cid.handle,
3153              cid->gid, cid->handle);
3154         rbd_dev->owner_cid = *cid; /* struct */
3155 }
3156
3157 static void format_lock_cookie(struct rbd_device *rbd_dev, char *buf)
3158 {
3159         mutex_lock(&rbd_dev->watch_mutex);
3160         sprintf(buf, "%s %llu", RBD_LOCK_COOKIE_PREFIX, rbd_dev->watch_cookie);
3161         mutex_unlock(&rbd_dev->watch_mutex);
3162 }
3163
3164 /*
3165  * lock_rwsem must be held for write
3166  */
3167 static int rbd_lock(struct rbd_device *rbd_dev)
3168 {
3169         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
3170         struct rbd_client_id cid = rbd_get_cid(rbd_dev);
3171         char cookie[32];
3172         int ret;
3173
3174         WARN_ON(__rbd_is_lock_owner(rbd_dev));
3175
3176         format_lock_cookie(rbd_dev, cookie);
3177         ret = ceph_cls_lock(osdc, &rbd_dev->header_oid, &rbd_dev->header_oloc,
3178                             RBD_LOCK_NAME, CEPH_CLS_LOCK_EXCLUSIVE, cookie,
3179                             RBD_LOCK_TAG, "", 0);
3180         if (ret)
3181                 return ret;
3182
3183         rbd_dev->lock_state = RBD_LOCK_STATE_LOCKED;
3184         rbd_set_owner_cid(rbd_dev, &cid);
3185         queue_work(rbd_dev->task_wq, &rbd_dev->acquired_lock_work);
3186         return 0;
3187 }
3188
3189 /*
3190  * lock_rwsem must be held for write
3191  */
3192 static int rbd_unlock(struct rbd_device *rbd_dev)
3193 {
3194         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
3195         char cookie[32];
3196         int ret;
3197
3198         WARN_ON(!__rbd_is_lock_owner(rbd_dev));
3199
3200         rbd_dev->lock_state = RBD_LOCK_STATE_UNLOCKED;
3201
3202         format_lock_cookie(rbd_dev, cookie);
3203         ret = ceph_cls_unlock(osdc, &rbd_dev->header_oid, &rbd_dev->header_oloc,
3204                               RBD_LOCK_NAME, cookie);
3205         if (ret && ret != -ENOENT) {
3206                 rbd_warn(rbd_dev, "cls_unlock failed: %d", ret);
3207                 return ret;
3208         }
3209
3210         rbd_set_owner_cid(rbd_dev, &rbd_empty_cid);
3211         queue_work(rbd_dev->task_wq, &rbd_dev->released_lock_work);
3212         return 0;
3213 }
3214
3215 static int __rbd_notify_op_lock(struct rbd_device *rbd_dev,
3216                                 enum rbd_notify_op notify_op,
3217                                 struct page ***preply_pages,
3218                                 size_t *preply_len)
3219 {
3220         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
3221         struct rbd_client_id cid = rbd_get_cid(rbd_dev);
3222         int buf_size = 4 + 8 + 8 + CEPH_ENCODING_START_BLK_LEN;
3223         char buf[buf_size];
3224         void *p = buf;
3225
3226         dout("%s rbd_dev %p notify_op %d\n", __func__, rbd_dev, notify_op);
3227
3228         /* encode *LockPayload NotifyMessage (op + ClientId) */
3229         ceph_start_encoding(&p, 2, 1, buf_size - CEPH_ENCODING_START_BLK_LEN);
3230         ceph_encode_32(&p, notify_op);
3231         ceph_encode_64(&p, cid.gid);
3232         ceph_encode_64(&p, cid.handle);
3233
3234         return ceph_osdc_notify(osdc, &rbd_dev->header_oid,
3235                                 &rbd_dev->header_oloc, buf, buf_size,
3236                                 RBD_NOTIFY_TIMEOUT, preply_pages, preply_len);
3237 }
3238
3239 static void rbd_notify_op_lock(struct rbd_device *rbd_dev,
3240                                enum rbd_notify_op notify_op)
3241 {
3242         struct page **reply_pages;
3243         size_t reply_len;
3244
3245         __rbd_notify_op_lock(rbd_dev, notify_op, &reply_pages, &reply_len);
3246         ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len));
3247 }
3248
3249 static void rbd_notify_acquired_lock(struct work_struct *work)
3250 {
3251         struct rbd_device *rbd_dev = container_of(work, struct rbd_device,
3252                                                   acquired_lock_work);
3253
3254         rbd_notify_op_lock(rbd_dev, RBD_NOTIFY_OP_ACQUIRED_LOCK);
3255 }
3256
3257 static void rbd_notify_released_lock(struct work_struct *work)
3258 {
3259         struct rbd_device *rbd_dev = container_of(work, struct rbd_device,
3260                                                   released_lock_work);
3261
3262         rbd_notify_op_lock(rbd_dev, RBD_NOTIFY_OP_RELEASED_LOCK);
3263 }
3264
3265 static int rbd_request_lock(struct rbd_device *rbd_dev)
3266 {
3267         struct page **reply_pages;
3268         size_t reply_len;
3269         bool lock_owner_responded = false;
3270         int ret;
3271
3272         dout("%s rbd_dev %p\n", __func__, rbd_dev);
3273
3274         ret = __rbd_notify_op_lock(rbd_dev, RBD_NOTIFY_OP_REQUEST_LOCK,
3275                                    &reply_pages, &reply_len);
3276         if (ret && ret != -ETIMEDOUT) {
3277                 rbd_warn(rbd_dev, "failed to request lock: %d", ret);
3278                 goto out;
3279         }
3280
3281         if (reply_len > 0 && reply_len <= PAGE_SIZE) {
3282                 void *p = page_address(reply_pages[0]);
3283                 void *const end = p + reply_len;
3284                 u32 n;
3285
3286                 ceph_decode_32_safe(&p, end, n, e_inval); /* num_acks */
3287                 while (n--) {
3288                         u8 struct_v;
3289                         u32 len;
3290
3291                         ceph_decode_need(&p, end, 8 + 8, e_inval);
3292                         p += 8 + 8; /* skip gid and cookie */
3293
3294                         ceph_decode_32_safe(&p, end, len, e_inval);
3295                         if (!len)
3296                                 continue;
3297
3298                         if (lock_owner_responded) {
3299                                 rbd_warn(rbd_dev,
3300                                          "duplicate lock owners detected");
3301                                 ret = -EIO;
3302                                 goto out;
3303                         }
3304
3305                         lock_owner_responded = true;
3306                         ret = ceph_start_decoding(&p, end, 1, "ResponseMessage",
3307                                                   &struct_v, &len);
3308                         if (ret) {
3309                                 rbd_warn(rbd_dev,
3310                                          "failed to decode ResponseMessage: %d",
3311                                          ret);
3312                                 goto e_inval;
3313                         }
3314
3315                         ret = ceph_decode_32(&p);
3316                 }
3317         }
3318
3319         if (!lock_owner_responded) {
3320                 rbd_warn(rbd_dev, "no lock owners detected");
3321                 ret = -ETIMEDOUT;
3322         }
3323
3324 out:
3325         ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len));
3326         return ret;
3327
3328 e_inval:
3329         ret = -EINVAL;
3330         goto out;
3331 }
3332
3333 static void wake_requests(struct rbd_device *rbd_dev, bool wake_all)
3334 {
3335         dout("%s rbd_dev %p wake_all %d\n", __func__, rbd_dev, wake_all);
3336
3337         cancel_delayed_work(&rbd_dev->lock_dwork);
3338         if (wake_all)
3339                 wake_up_all(&rbd_dev->lock_waitq);
3340         else
3341                 wake_up(&rbd_dev->lock_waitq);
3342 }
3343
3344 static int get_lock_owner_info(struct rbd_device *rbd_dev,
3345                                struct ceph_locker **lockers, u32 *num_lockers)
3346 {
3347         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
3348         u8 lock_type;
3349         char *lock_tag;
3350         int ret;
3351
3352         dout("%s rbd_dev %p\n", __func__, rbd_dev);
3353
3354         ret = ceph_cls_lock_info(osdc, &rbd_dev->header_oid,
3355                                  &rbd_dev->header_oloc, RBD_LOCK_NAME,
3356                                  &lock_type, &lock_tag, lockers, num_lockers);
3357         if (ret)
3358                 return ret;
3359
3360         if (*num_lockers == 0) {
3361                 dout("%s rbd_dev %p no lockers detected\n", __func__, rbd_dev);
3362                 goto out;
3363         }
3364
3365         if (strcmp(lock_tag, RBD_LOCK_TAG)) {
3366                 rbd_warn(rbd_dev, "locked by external mechanism, tag %s",
3367                          lock_tag);
3368                 ret = -EBUSY;
3369                 goto out;
3370         }
3371
3372         if (lock_type == CEPH_CLS_LOCK_SHARED) {
3373                 rbd_warn(rbd_dev, "shared lock type detected");
3374                 ret = -EBUSY;
3375                 goto out;
3376         }
3377
3378         if (strncmp((*lockers)[0].id.cookie, RBD_LOCK_COOKIE_PREFIX,
3379                     strlen(RBD_LOCK_COOKIE_PREFIX))) {
3380                 rbd_warn(rbd_dev, "locked by external mechanism, cookie %s",
3381                          (*lockers)[0].id.cookie);
3382                 ret = -EBUSY;
3383                 goto out;
3384         }
3385
3386 out:
3387         kfree(lock_tag);
3388         return ret;
3389 }
3390
3391 static int find_watcher(struct rbd_device *rbd_dev,
3392                         const struct ceph_locker *locker)
3393 {
3394         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
3395         struct ceph_watch_item *watchers;
3396         u32 num_watchers;
3397         u64 cookie;
3398         int i;
3399         int ret;
3400
3401         ret = ceph_osdc_list_watchers(osdc, &rbd_dev->header_oid,
3402                                       &rbd_dev->header_oloc, &watchers,
3403                                       &num_watchers);
3404         if (ret)
3405                 return ret;
3406
3407         sscanf(locker->id.cookie, RBD_LOCK_COOKIE_PREFIX " %llu", &cookie);
3408         for (i = 0; i < num_watchers; i++) {
3409                 if (!memcmp(&watchers[i].addr, &locker->info.addr,
3410                             sizeof(locker->info.addr)) &&
3411                     watchers[i].cookie == cookie) {
3412                         struct rbd_client_id cid = {
3413                                 .gid = le64_to_cpu(watchers[i].name.num),
3414                                 .handle = cookie,
3415                         };
3416
3417                         dout("%s rbd_dev %p found cid %llu-%llu\n", __func__,
3418                              rbd_dev, cid.gid, cid.handle);
3419                         rbd_set_owner_cid(rbd_dev, &cid);
3420                         ret = 1;
3421                         goto out;
3422                 }
3423         }
3424
3425         dout("%s rbd_dev %p no watchers\n", __func__, rbd_dev);
3426         ret = 0;
3427 out:
3428         kfree(watchers);
3429         return ret;
3430 }
3431
3432 /*
3433  * lock_rwsem must be held for write
3434  */
3435 static int rbd_try_lock(struct rbd_device *rbd_dev)
3436 {
3437         struct ceph_client *client = rbd_dev->rbd_client->client;
3438         struct ceph_locker *lockers;
3439         u32 num_lockers;
3440         int ret;
3441
3442         for (;;) {
3443                 ret = rbd_lock(rbd_dev);
3444                 if (ret != -EBUSY)
3445                         return ret;
3446
3447                 /* determine if the current lock holder is still alive */
3448                 ret = get_lock_owner_info(rbd_dev, &lockers, &num_lockers);
3449                 if (ret)
3450                         return ret;
3451
3452                 if (num_lockers == 0)
3453                         goto again;
3454
3455                 ret = find_watcher(rbd_dev, lockers);
3456                 if (ret) {
3457                         if (ret > 0)
3458                                 ret = 0; /* have to request lock */
3459                         goto out;
3460                 }
3461
3462                 rbd_warn(rbd_dev, "%s%llu seems dead, breaking lock",
3463                          ENTITY_NAME(lockers[0].id.name));
3464
3465                 ret = ceph_monc_blacklist_add(&client->monc,
3466                                               &lockers[0].info.addr);
3467                 if (ret) {
3468                         rbd_warn(rbd_dev, "blacklist of %s%llu failed: %d",
3469                                  ENTITY_NAME(lockers[0].id.name), ret);
3470                         goto out;
3471                 }
3472
3473                 ret = ceph_cls_break_lock(&client->osdc, &rbd_dev->header_oid,
3474                                           &rbd_dev->header_oloc, RBD_LOCK_NAME,
3475                                           lockers[0].id.cookie,
3476                                           &lockers[0].id.name);
3477                 if (ret && ret != -ENOENT)
3478                         goto out;
3479
3480 again:
3481                 ceph_free_lockers(lockers, num_lockers);
3482         }
3483
3484 out:
3485         ceph_free_lockers(lockers, num_lockers);
3486         return ret;
3487 }
3488
3489 /*
3490  * ret is set only if lock_state is RBD_LOCK_STATE_UNLOCKED
3491  */
3492 static enum rbd_lock_state rbd_try_acquire_lock(struct rbd_device *rbd_dev,
3493                                                 int *pret)
3494 {
3495         enum rbd_lock_state lock_state;
3496
3497         down_read(&rbd_dev->lock_rwsem);
3498         dout("%s rbd_dev %p read lock_state %d\n", __func__, rbd_dev,
3499              rbd_dev->lock_state);
3500         if (__rbd_is_lock_owner(rbd_dev)) {
3501                 lock_state = rbd_dev->lock_state;
3502                 up_read(&rbd_dev->lock_rwsem);
3503                 return lock_state;
3504         }
3505
3506         up_read(&rbd_dev->lock_rwsem);
3507         down_write(&rbd_dev->lock_rwsem);
3508         dout("%s rbd_dev %p write lock_state %d\n", __func__, rbd_dev,
3509              rbd_dev->lock_state);
3510         if (!__rbd_is_lock_owner(rbd_dev)) {
3511                 *pret = rbd_try_lock(rbd_dev);
3512                 if (*pret)
3513                         rbd_warn(rbd_dev, "failed to acquire lock: %d", *pret);
3514         }
3515
3516         lock_state = rbd_dev->lock_state;
3517         up_write(&rbd_dev->lock_rwsem);
3518         return lock_state;
3519 }
3520
3521 static void rbd_acquire_lock(struct work_struct *work)
3522 {
3523         struct rbd_device *rbd_dev = container_of(to_delayed_work(work),
3524                                             struct rbd_device, lock_dwork);
3525         enum rbd_lock_state lock_state;
3526         int ret;
3527
3528         dout("%s rbd_dev %p\n", __func__, rbd_dev);
3529 again:
3530         lock_state = rbd_try_acquire_lock(rbd_dev, &ret);
3531         if (lock_state != RBD_LOCK_STATE_UNLOCKED || ret == -EBLACKLISTED) {
3532                 if (lock_state == RBD_LOCK_STATE_LOCKED)
3533                         wake_requests(rbd_dev, true);
3534                 dout("%s rbd_dev %p lock_state %d ret %d - done\n", __func__,
3535                      rbd_dev, lock_state, ret);
3536                 return;
3537         }
3538
3539         ret = rbd_request_lock(rbd_dev);
3540         if (ret == -ETIMEDOUT) {
3541                 goto again; /* treat this as a dead client */
3542         } else if (ret < 0) {
3543                 rbd_warn(rbd_dev, "error requesting lock: %d", ret);
3544                 mod_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork,
3545                                  RBD_RETRY_DELAY);
3546         } else {
3547                 /*
3548                  * lock owner acked, but resend if we don't see them
3549                  * release the lock
3550                  */
3551                 dout("%s rbd_dev %p requeueing lock_dwork\n", __func__,
3552                      rbd_dev);
3553                 mod_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork,
3554                     msecs_to_jiffies(2 * RBD_NOTIFY_TIMEOUT * MSEC_PER_SEC));
3555         }
3556 }
3557
3558 /*
3559  * lock_rwsem must be held for write
3560  */
3561 static bool rbd_release_lock(struct rbd_device *rbd_dev)
3562 {
3563         dout("%s rbd_dev %p read lock_state %d\n", __func__, rbd_dev,
3564              rbd_dev->lock_state);
3565         if (rbd_dev->lock_state != RBD_LOCK_STATE_LOCKED)
3566                 return false;
3567
3568         rbd_dev->lock_state = RBD_LOCK_STATE_RELEASING;
3569         downgrade_write(&rbd_dev->lock_rwsem);
3570         /*
3571          * Ensure that all in-flight IO is flushed.
3572          *
3573          * FIXME: ceph_osdc_sync() flushes the entire OSD client, which
3574          * may be shared with other devices.
3575          */
3576         ceph_osdc_sync(&rbd_dev->rbd_client->client->osdc);
3577         up_read(&rbd_dev->lock_rwsem);
3578
3579         down_write(&rbd_dev->lock_rwsem);
3580         dout("%s rbd_dev %p write lock_state %d\n", __func__, rbd_dev,
3581              rbd_dev->lock_state);
3582         if (rbd_dev->lock_state != RBD_LOCK_STATE_RELEASING)
3583                 return false;
3584
3585         if (!rbd_unlock(rbd_dev))
3586                 /*
3587                  * Give others a chance to grab the lock - we would re-acquire
3588                  * almost immediately if we got new IO during ceph_osdc_sync()
3589                  * otherwise.  We need to ack our own notifications, so this
3590                  * lock_dwork will be requeued from rbd_wait_state_locked()
3591                  * after wake_requests() in rbd_handle_released_lock().
3592                  */
3593                 cancel_delayed_work(&rbd_dev->lock_dwork);
3594
3595         return true;
3596 }
3597
3598 static void rbd_release_lock_work(struct work_struct *work)
3599 {
3600         struct rbd_device *rbd_dev = container_of(work, struct rbd_device,
3601                                                   unlock_work);
3602
3603         down_write(&rbd_dev->lock_rwsem);
3604         rbd_release_lock(rbd_dev);
3605         up_write(&rbd_dev->lock_rwsem);
3606 }
3607
3608 static void rbd_handle_acquired_lock(struct rbd_device *rbd_dev, u8 struct_v,
3609                                      void **p)
3610 {
3611         struct rbd_client_id cid = { 0 };
3612
3613         if (struct_v >= 2) {
3614                 cid.gid = ceph_decode_64(p);
3615                 cid.handle = ceph_decode_64(p);
3616         }
3617
3618         dout("%s rbd_dev %p cid %llu-%llu\n", __func__, rbd_dev, cid.gid,
3619              cid.handle);
3620         if (!rbd_cid_equal(&cid, &rbd_empty_cid)) {
3621                 down_write(&rbd_dev->lock_rwsem);
3622                 if (rbd_cid_equal(&cid, &rbd_dev->owner_cid)) {
3623                         /*
3624                          * we already know that the remote client is
3625                          * the owner
3626                          */
3627                         up_write(&rbd_dev->lock_rwsem);
3628                         return;
3629                 }
3630
3631                 rbd_set_owner_cid(rbd_dev, &cid);
3632                 downgrade_write(&rbd_dev->lock_rwsem);
3633         } else {
3634                 down_read(&rbd_dev->lock_rwsem);
3635         }
3636
3637         if (!__rbd_is_lock_owner(rbd_dev))
3638                 wake_requests(rbd_dev, false);
3639         up_read(&rbd_dev->lock_rwsem);
3640 }
3641
3642 static void rbd_handle_released_lock(struct rbd_device *rbd_dev, u8 struct_v,
3643                                      void **p)
3644 {
3645         struct rbd_client_id cid = { 0 };
3646
3647         if (struct_v >= 2) {
3648                 cid.gid = ceph_decode_64(p);
3649                 cid.handle = ceph_decode_64(p);
3650         }
3651
3652         dout("%s rbd_dev %p cid %llu-%llu\n", __func__, rbd_dev, cid.gid,
3653              cid.handle);
3654         if (!rbd_cid_equal(&cid, &rbd_empty_cid)) {
3655                 down_write(&rbd_dev->lock_rwsem);
3656                 if (!rbd_cid_equal(&cid, &rbd_dev->owner_cid)) {
3657                         dout("%s rbd_dev %p unexpected owner, cid %llu-%llu != owner_cid %llu-%llu\n",
3658                              __func__, rbd_dev, cid.gid, cid.handle,
3659                              rbd_dev->owner_cid.gid, rbd_dev->owner_cid.handle);
3660                         up_write(&rbd_dev->lock_rwsem);
3661                         return;
3662                 }
3663
3664                 rbd_set_owner_cid(rbd_dev, &rbd_empty_cid);
3665                 downgrade_write(&rbd_dev->lock_rwsem);
3666         } else {
3667                 down_read(&rbd_dev->lock_rwsem);
3668         }
3669
3670         if (!__rbd_is_lock_owner(rbd_dev))
3671                 wake_requests(rbd_dev, false);
3672         up_read(&rbd_dev->lock_rwsem);
3673 }
3674
3675 static bool rbd_handle_request_lock(struct rbd_device *rbd_dev, u8 struct_v,
3676                                     void **p)
3677 {
3678         struct rbd_client_id my_cid = rbd_get_cid(rbd_dev);
3679         struct rbd_client_id cid = { 0 };
3680         bool need_to_send;
3681
3682         if (struct_v >= 2) {
3683                 cid.gid = ceph_decode_64(p);
3684                 cid.handle = ceph_decode_64(p);
3685         }
3686
3687         dout("%s rbd_dev %p cid %llu-%llu\n", __func__, rbd_dev, cid.gid,
3688              cid.handle);
3689         if (rbd_cid_equal(&cid, &my_cid))
3690                 return false;
3691
3692         down_read(&rbd_dev->lock_rwsem);
3693         need_to_send = __rbd_is_lock_owner(rbd_dev);
3694         if (rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED) {
3695                 if (!rbd_cid_equal(&rbd_dev->owner_cid, &rbd_empty_cid)) {
3696                         dout("%s rbd_dev %p queueing unlock_work\n", __func__,
3697                              rbd_dev);
3698                         queue_work(rbd_dev->task_wq, &rbd_dev->unlock_work);
3699                 }
3700         }
3701         up_read(&rbd_dev->lock_rwsem);
3702         return need_to_send;
3703 }
3704
3705 static void __rbd_acknowledge_notify(struct rbd_device *rbd_dev,
3706                                      u64 notify_id, u64 cookie, s32 *result)
3707 {
3708         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
3709         int buf_size = 4 + CEPH_ENCODING_START_BLK_LEN;
3710         char buf[buf_size];
3711         int ret;
3712
3713         if (result) {
3714                 void *p = buf;
3715
3716                 /* encode ResponseMessage */
3717                 ceph_start_encoding(&p, 1, 1,
3718                                     buf_size - CEPH_ENCODING_START_BLK_LEN);
3719                 ceph_encode_32(&p, *result);
3720         } else {
3721                 buf_size = 0;
3722         }
3723
3724         ret = ceph_osdc_notify_ack(osdc, &rbd_dev->header_oid,
3725                                    &rbd_dev->header_oloc, notify_id, cookie,
3726                                    buf, buf_size);
3727         if (ret)
3728                 rbd_warn(rbd_dev, "acknowledge_notify failed: %d", ret);
3729 }
3730
3731 static void rbd_acknowledge_notify(struct rbd_device *rbd_dev, u64 notify_id,
3732                                    u64 cookie)
3733 {
3734         dout("%s rbd_dev %p\n", __func__, rbd_dev);
3735         __rbd_acknowledge_notify(rbd_dev, notify_id, cookie, NULL);
3736 }
3737
3738 static void rbd_acknowledge_notify_result(struct rbd_device *rbd_dev,
3739                                           u64 notify_id, u64 cookie, s32 result)
3740 {
3741         dout("%s rbd_dev %p result %d\n", __func__, rbd_dev, result);
3742         __rbd_acknowledge_notify(rbd_dev, notify_id, cookie, &result);
3743 }
3744
3745 static void rbd_watch_cb(void *arg, u64 notify_id, u64 cookie,
3746                          u64 notifier_id, void *data, size_t data_len)
3747 {
3748         struct rbd_device *rbd_dev = arg;
3749         void *p = data;
3750         void *const end = p + data_len;
3751         u8 struct_v;
3752         u32 len;
3753         u32 notify_op;
3754         int ret;
3755
3756         dout("%s rbd_dev %p cookie %llu notify_id %llu data_len %zu\n",
3757              __func__, rbd_dev, cookie, notify_id, data_len);
3758         if (data_len) {
3759                 ret = ceph_start_decoding(&p, end, 1, "NotifyMessage",
3760                                           &struct_v, &len);
3761                 if (ret) {
3762                         rbd_warn(rbd_dev, "failed to decode NotifyMessage: %d",
3763                                  ret);
3764                         return;
3765                 }
3766
3767                 notify_op = ceph_decode_32(&p);
3768         } else {
3769                 /* legacy notification for header updates */
3770                 notify_op = RBD_NOTIFY_OP_HEADER_UPDATE;
3771                 len = 0;
3772         }
3773
3774         dout("%s rbd_dev %p notify_op %u\n", __func__, rbd_dev, notify_op);
3775         switch (notify_op) {
3776         case RBD_NOTIFY_OP_ACQUIRED_LOCK:
3777                 rbd_handle_acquired_lock(rbd_dev, struct_v, &p);
3778                 rbd_acknowledge_notify(rbd_dev, notify_id, cookie);
3779                 break;
3780         case RBD_NOTIFY_OP_RELEASED_LOCK:
3781                 rbd_handle_released_lock(rbd_dev, struct_v, &p);
3782                 rbd_acknowledge_notify(rbd_dev, notify_id, cookie);
3783                 break;
3784         case RBD_NOTIFY_OP_REQUEST_LOCK:
3785                 if (rbd_handle_request_lock(rbd_dev, struct_v, &p))
3786                         /*
3787                          * send ResponseMessage(0) back so the client
3788                          * can detect a missing owner
3789                          */
3790                         rbd_acknowledge_notify_result(rbd_dev, notify_id,
3791                                                       cookie, 0);
3792                 else
3793                         rbd_acknowledge_notify(rbd_dev, notify_id, cookie);
3794                 break;
3795         case RBD_NOTIFY_OP_HEADER_UPDATE:
3796                 ret = rbd_dev_refresh(rbd_dev);
3797                 if (ret)
3798                         rbd_warn(rbd_dev, "refresh failed: %d", ret);
3799
3800                 rbd_acknowledge_notify(rbd_dev, notify_id, cookie);
3801                 break;
3802         default:
3803                 if (rbd_is_lock_owner(rbd_dev))
3804                         rbd_acknowledge_notify_result(rbd_dev, notify_id,
3805                                                       cookie, -EOPNOTSUPP);
3806                 else
3807                         rbd_acknowledge_notify(rbd_dev, notify_id, cookie);
3808                 break;
3809         }
3810 }
3811
3812 static void __rbd_unregister_watch(struct rbd_device *rbd_dev);
3813
3814 static void rbd_watch_errcb(void *arg, u64 cookie, int err)
3815 {
3816         struct rbd_device *rbd_dev = arg;
3817
3818         rbd_warn(rbd_dev, "encountered watch error: %d", err);
3819
3820         down_write(&rbd_dev->lock_rwsem);
3821         rbd_set_owner_cid(rbd_dev, &rbd_empty_cid);
3822         up_write(&rbd_dev->lock_rwsem);
3823
3824         mutex_lock(&rbd_dev->watch_mutex);
3825         if (rbd_dev->watch_state == RBD_WATCH_STATE_REGISTERED) {
3826                 __rbd_unregister_watch(rbd_dev);
3827                 rbd_dev->watch_state = RBD_WATCH_STATE_ERROR;
3828
3829                 queue_delayed_work(rbd_dev->task_wq, &rbd_dev->watch_dwork, 0);
3830         }
3831         mutex_unlock(&rbd_dev->watch_mutex);
3832 }
3833
3834 /*
3835  * watch_mutex must be locked
3836  */
3837 static int __rbd_register_watch(struct rbd_device *rbd_dev)
3838 {
3839         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
3840         struct ceph_osd_linger_request *handle;
3841
3842         rbd_assert(!rbd_dev->watch_handle);
3843         dout("%s rbd_dev %p\n", __func__, rbd_dev);
3844
3845         handle = ceph_osdc_watch(osdc, &rbd_dev->header_oid,
3846                                  &rbd_dev->header_oloc, rbd_watch_cb,
3847                                  rbd_watch_errcb, rbd_dev);
3848         if (IS_ERR(handle))
3849                 return PTR_ERR(handle);
3850
3851         rbd_dev->watch_handle = handle;
3852         return 0;
3853 }
3854
3855 /*
3856  * watch_mutex must be locked
3857  */
3858 static void __rbd_unregister_watch(struct rbd_device *rbd_dev)
3859 {
3860         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
3861         int ret;
3862
3863         rbd_assert(rbd_dev->watch_handle);
3864         dout("%s rbd_dev %p\n", __func__, rbd_dev);
3865
3866         ret = ceph_osdc_unwatch(osdc, rbd_dev->watch_handle);
3867         if (ret)
3868                 rbd_warn(rbd_dev, "failed to unwatch: %d", ret);
3869
3870         rbd_dev->watch_handle = NULL;
3871 }
3872
3873 static int rbd_register_watch(struct rbd_device *rbd_dev)
3874 {
3875         int ret;
3876
3877         mutex_lock(&rbd_dev->watch_mutex);
3878         rbd_assert(rbd_dev->watch_state == RBD_WATCH_STATE_UNREGISTERED);
3879         ret = __rbd_register_watch(rbd_dev);
3880         if (ret)
3881                 goto out;
3882
3883         rbd_dev->watch_state = RBD_WATCH_STATE_REGISTERED;
3884         rbd_dev->watch_cookie = rbd_dev->watch_handle->linger_id;
3885
3886 out:
3887         mutex_unlock(&rbd_dev->watch_mutex);
3888         return ret;
3889 }
3890
3891 static void cancel_tasks_sync(struct rbd_device *rbd_dev)
3892 {
3893         dout("%s rbd_dev %p\n", __func__, rbd_dev);
3894
3895         cancel_delayed_work_sync(&rbd_dev->watch_dwork);
3896         cancel_work_sync(&rbd_dev->acquired_lock_work);
3897         cancel_work_sync(&rbd_dev->released_lock_work);
3898         cancel_delayed_work_sync(&rbd_dev->lock_dwork);
3899         cancel_work_sync(&rbd_dev->unlock_work);
3900 }
3901
3902 static void rbd_unregister_watch(struct rbd_device *rbd_dev)
3903 {
3904         WARN_ON(waitqueue_active(&rbd_dev->lock_waitq));
3905         cancel_tasks_sync(rbd_dev);
3906
3907         mutex_lock(&rbd_dev->watch_mutex);
3908         if (rbd_dev->watch_state == RBD_WATCH_STATE_REGISTERED)
3909                 __rbd_unregister_watch(rbd_dev);
3910         rbd_dev->watch_state = RBD_WATCH_STATE_UNREGISTERED;
3911         mutex_unlock(&rbd_dev->watch_mutex);
3912
3913         ceph_osdc_flush_notifies(&rbd_dev->rbd_client->client->osdc);
3914 }
3915
3916 static void rbd_reregister_watch(struct work_struct *work)
3917 {
3918         struct rbd_device *rbd_dev = container_of(to_delayed_work(work),
3919                                             struct rbd_device, watch_dwork);
3920         bool was_lock_owner = false;
3921         int ret;
3922
3923         dout("%s rbd_dev %p\n", __func__, rbd_dev);
3924
3925         down_write(&rbd_dev->lock_rwsem);
3926         if (rbd_dev->lock_state == RBD_LOCK_STATE_LOCKED)
3927                 was_lock_owner = rbd_release_lock(rbd_dev);
3928
3929         mutex_lock(&rbd_dev->watch_mutex);
3930         if (rbd_dev->watch_state != RBD_WATCH_STATE_ERROR)
3931                 goto fail_unlock;
3932
3933         ret = __rbd_register_watch(rbd_dev);
3934         if (ret) {
3935                 rbd_warn(rbd_dev, "failed to reregister watch: %d", ret);
3936                 if (ret != -EBLACKLISTED)
3937                         queue_delayed_work(rbd_dev->task_wq,
3938                                            &rbd_dev->watch_dwork,
3939                                            RBD_RETRY_DELAY);
3940                 goto fail_unlock;
3941         }
3942
3943         rbd_dev->watch_state = RBD_WATCH_STATE_REGISTERED;
3944         rbd_dev->watch_cookie = rbd_dev->watch_handle->linger_id;
3945         mutex_unlock(&rbd_dev->watch_mutex);
3946
3947         ret = rbd_dev_refresh(rbd_dev);
3948         if (ret)
3949                 rbd_warn(rbd_dev, "reregisteration refresh failed: %d", ret);
3950
3951         if (was_lock_owner) {
3952                 ret = rbd_try_lock(rbd_dev);
3953                 if (ret)
3954                         rbd_warn(rbd_dev, "reregisteration lock failed: %d",
3955                                  ret);
3956         }
3957
3958         up_write(&rbd_dev->lock_rwsem);
3959         wake_requests(rbd_dev, true);
3960         return;
3961
3962 fail_unlock:
3963         mutex_unlock(&rbd_dev->watch_mutex);
3964         up_write(&rbd_dev->lock_rwsem);
3965 }
3966
3967 /*
3968  * Synchronous osd object method call.  Returns the number of bytes
3969  * returned in the outbound buffer, or a negative error code.
3970  */
3971 static int rbd_obj_method_sync(struct rbd_device *rbd_dev,
3972                              const char *object_name,
3973                              const char *class_name,
3974                              const char *method_name,
3975                              const void *outbound,
3976                              size_t outbound_size,
3977                              void *inbound,
3978                              size_t inbound_size)
3979 {
3980         struct rbd_obj_request *obj_request;
3981         struct page **pages;
3982         u32 page_count;
3983         int ret;
3984
3985         /*
3986          * Method calls are ultimately read operations.  The result
3987          * should placed into the inbound buffer provided.  They
3988          * also supply outbound data--parameters for the object
3989          * method.  Currently if this is present it will be a
3990          * snapshot id.
3991          */
3992         page_count = (u32)calc_pages_for(0, inbound_size);
3993         pages = ceph_alloc_page_vector(page_count, GFP_KERNEL);
3994         if (IS_ERR(pages))
3995                 return PTR_ERR(pages);
3996
3997         ret = -ENOMEM;
3998         obj_request = rbd_obj_request_create(object_name, 0, inbound_size,
3999                                                         OBJ_REQUEST_PAGES);
4000         if (!obj_request)
4001                 goto out;
4002
4003         obj_request->pages = pages;
4004         obj_request->page_count = page_count;
4005
4006         obj_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1,
4007                                                   obj_request);
4008         if (!obj_request->osd_req)
4009                 goto out;
4010
4011         osd_req_op_cls_init(obj_request->osd_req, 0, CEPH_OSD_OP_CALL,
4012                                         class_name, method_name);
4013         if (outbound_size) {
4014                 struct ceph_pagelist *pagelist;
4015
4016                 pagelist = kmalloc(sizeof (*pagelist), GFP_NOFS);
4017                 if (!pagelist)
4018                         goto out;
4019
4020                 ceph_pagelist_init(pagelist);
4021                 ceph_pagelist_append(pagelist, outbound, outbound_size);
4022                 osd_req_op_cls_request_data_pagelist(obj_request->osd_req, 0,
4023                                                 pagelist);
4024         }
4025         osd_req_op_cls_response_data_pages(obj_request->osd_req, 0,
4026                                         obj_request->pages, inbound_size,
4027                                         0, false, false);
4028         rbd_osd_req_format_read(obj_request);
4029
4030         rbd_obj_request_submit(obj_request);
4031         ret = rbd_obj_request_wait(obj_request);
4032         if (ret)
4033                 goto out;
4034
4035         ret = obj_request->result;
4036         if (ret < 0)
4037                 goto out;
4038
4039         rbd_assert(obj_request->xferred < (u64)INT_MAX);
4040         ret = (int)obj_request->xferred;
4041         ceph_copy_from_page_vector(pages, inbound, 0, obj_request->xferred);
4042 out:
4043         if (obj_request)
4044                 rbd_obj_request_put(obj_request);
4045         else
4046                 ceph_release_page_vector(pages, page_count);
4047
4048         return ret;
4049 }
4050
4051 /*
4052  * lock_rwsem must be held for read
4053  */
4054 static void rbd_wait_state_locked(struct rbd_device *rbd_dev)
4055 {
4056         DEFINE_WAIT(wait);
4057
4058         do {
4059                 /*
4060                  * Note the use of mod_delayed_work() in rbd_acquire_lock()
4061                  * and cancel_delayed_work() in wake_requests().
4062                  */
4063                 dout("%s rbd_dev %p queueing lock_dwork\n", __func__, rbd_dev);
4064                 queue_delayed_work(rbd_dev->task_wq, &rbd_dev->lock_dwork, 0);
4065                 prepare_to_wait_exclusive(&rbd_dev->lock_waitq, &wait,
4066                                           TASK_UNINTERRUPTIBLE);
4067                 up_read(&rbd_dev->lock_rwsem);
4068                 schedule();
4069                 down_read(&rbd_dev->lock_rwsem);
4070         } while (rbd_dev->lock_state != RBD_LOCK_STATE_LOCKED);
4071         finish_wait(&rbd_dev->lock_waitq, &wait);
4072 }
4073
4074 static void rbd_queue_workfn(struct work_struct *work)
4075 {
4076         struct request *rq = blk_mq_rq_from_pdu(work);
4077         struct rbd_device *rbd_dev = rq->q->queuedata;
4078         struct rbd_img_request *img_request;
4079         struct ceph_snap_context *snapc = NULL;
4080         u64 offset = (u64)blk_rq_pos(rq) << SECTOR_SHIFT;
4081         u64 length = blk_rq_bytes(rq);
4082         enum obj_operation_type op_type;
4083         u64 mapping_size;
4084         bool must_be_locked;
4085         int result;
4086
4087         if (rq->cmd_type != REQ_TYPE_FS) {
4088                 dout("%s: non-fs request type %d\n", __func__,
4089                         (int) rq->cmd_type);
4090                 result = -EIO;
4091                 goto err;
4092         }
4093
4094         if (req_op(rq) == REQ_OP_DISCARD)
4095                 op_type = OBJ_OP_DISCARD;
4096         else if (req_op(rq) == REQ_OP_WRITE)
4097                 op_type = OBJ_OP_WRITE;
4098         else
4099                 op_type = OBJ_OP_READ;
4100
4101         /* Ignore/skip any zero-length requests */
4102
4103         if (!length) {
4104                 dout("%s: zero-length request\n", __func__);
4105                 result = 0;
4106                 goto err_rq;
4107         }
4108
4109         /* Only reads are allowed to a read-only device */
4110
4111         if (op_type != OBJ_OP_READ) {
4112                 if (rbd_dev->mapping.read_only) {
4113                         result = -EROFS;
4114                         goto err_rq;
4115                 }
4116                 rbd_assert(rbd_dev->spec->snap_id == CEPH_NOSNAP);
4117         }
4118
4119         /*
4120          * Quit early if the mapped snapshot no longer exists.  It's
4121          * still possible the snapshot will have disappeared by the
4122          * time our request arrives at the osd, but there's no sense in
4123          * sending it if we already know.
4124          */
4125         if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags)) {
4126                 dout("request for non-existent snapshot");
4127                 rbd_assert(rbd_dev->spec->snap_id != CEPH_NOSNAP);
4128                 result = -ENXIO;
4129                 goto err_rq;
4130         }
4131
4132         if (offset && length > U64_MAX - offset + 1) {
4133                 rbd_warn(rbd_dev, "bad request range (%llu~%llu)", offset,
4134                          length);
4135                 result = -EINVAL;
4136                 goto err_rq;    /* Shouldn't happen */
4137         }
4138
4139         blk_mq_start_request(rq);
4140
4141         down_read(&rbd_dev->header_rwsem);
4142         mapping_size = rbd_dev->mapping.size;
4143         if (op_type != OBJ_OP_READ) {
4144                 snapc = rbd_dev->header.snapc;
4145                 ceph_get_snap_context(snapc);
4146                 must_be_locked = rbd_is_lock_supported(rbd_dev);
4147         } else {
4148                 must_be_locked = rbd_dev->opts->lock_on_read &&
4149                                         rbd_is_lock_supported(rbd_dev);
4150         }
4151         up_read(&rbd_dev->header_rwsem);
4152
4153         if (offset + length > mapping_size) {
4154                 rbd_warn(rbd_dev, "beyond EOD (%llu~%llu > %llu)", offset,
4155                          length, mapping_size);
4156                 result = -EIO;
4157                 goto err_rq;
4158         }
4159
4160         if (must_be_locked) {
4161                 down_read(&rbd_dev->lock_rwsem);
4162                 if (rbd_dev->lock_state != RBD_LOCK_STATE_LOCKED)
4163                         rbd_wait_state_locked(rbd_dev);
4164         }
4165
4166         img_request = rbd_img_request_create(rbd_dev, offset, length, op_type,
4167                                              snapc);
4168         if (!img_request) {
4169                 result = -ENOMEM;
4170                 goto err_unlock;
4171         }
4172         img_request->rq = rq;
4173         snapc = NULL; /* img_request consumes a ref */
4174
4175         if (op_type == OBJ_OP_DISCARD)
4176                 result = rbd_img_request_fill(img_request, OBJ_REQUEST_NODATA,
4177                                               NULL);
4178         else
4179                 result = rbd_img_request_fill(img_request, OBJ_REQUEST_BIO,
4180                                               rq->bio);
4181         if (result)
4182                 goto err_img_request;
4183
4184         result = rbd_img_request_submit(img_request);
4185         if (result)
4186                 goto err_img_request;
4187
4188         if (must_be_locked)
4189                 up_read(&rbd_dev->lock_rwsem);
4190         return;
4191
4192 err_img_request:
4193         rbd_img_request_put(img_request);
4194 err_unlock:
4195         if (must_be_locked)
4196                 up_read(&rbd_dev->lock_rwsem);
4197 err_rq:
4198         if (result)
4199                 rbd_warn(rbd_dev, "%s %llx at %llx result %d",
4200                          obj_op_name(op_type), length, offset, result);
4201         ceph_put_snap_context(snapc);
4202 err:
4203         blk_mq_end_request(rq, result);
4204 }
4205
4206 static int rbd_queue_rq(struct blk_mq_hw_ctx *hctx,
4207                 const struct blk_mq_queue_data *bd)
4208 {
4209         struct request *rq = bd->rq;
4210         struct work_struct *work = blk_mq_rq_to_pdu(rq);
4211
4212         queue_work(rbd_wq, work);
4213         return BLK_MQ_RQ_QUEUE_OK;
4214 }
4215
4216 static void rbd_free_disk(struct rbd_device *rbd_dev)
4217 {
4218         struct gendisk *disk = rbd_dev->disk;
4219
4220         if (!disk)
4221                 return;
4222
4223         rbd_dev->disk = NULL;
4224         if (disk->flags & GENHD_FL_UP) {
4225                 del_gendisk(disk);
4226                 if (disk->queue)
4227                         blk_cleanup_queue(disk->queue);
4228                 blk_mq_free_tag_set(&rbd_dev->tag_set);
4229         }
4230         put_disk(disk);
4231 }
4232
4233 static int rbd_obj_read_sync(struct rbd_device *rbd_dev,
4234                                 const char *object_name,
4235                                 u64 offset, u64 length, void *buf)
4236
4237 {
4238         struct rbd_obj_request *obj_request;
4239         struct page **pages = NULL;
4240         u32 page_count;
4241         size_t size;
4242         int ret;
4243
4244         page_count = (u32) calc_pages_for(offset, length);
4245         pages = ceph_alloc_page_vector(page_count, GFP_KERNEL);
4246         if (IS_ERR(pages))
4247                 return PTR_ERR(pages);
4248
4249         ret = -ENOMEM;
4250         obj_request = rbd_obj_request_create(object_name, offset, length,
4251                                                         OBJ_REQUEST_PAGES);
4252         if (!obj_request)
4253                 goto out;
4254
4255         obj_request->pages = pages;
4256         obj_request->page_count = page_count;
4257
4258         obj_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1,
4259                                                   obj_request);
4260         if (!obj_request->osd_req)
4261                 goto out;
4262
4263         osd_req_op_extent_init(obj_request->osd_req, 0, CEPH_OSD_OP_READ,
4264                                         offset, length, 0, 0);
4265         osd_req_op_extent_osd_data_pages(obj_request->osd_req, 0,
4266                                         obj_request->pages,
4267                                         obj_request->length,
4268                                         obj_request->offset & ~PAGE_MASK,
4269                                         false, false);
4270         rbd_osd_req_format_read(obj_request);
4271
4272         rbd_obj_request_submit(obj_request);
4273         ret = rbd_obj_request_wait(obj_request);
4274         if (ret)
4275                 goto out;
4276
4277         ret = obj_request->result;
4278         if (ret < 0)
4279                 goto out;
4280
4281         rbd_assert(obj_request->xferred <= (u64) SIZE_MAX);
4282         size = (size_t) obj_request->xferred;
4283         ceph_copy_from_page_vector(pages, buf, 0, size);
4284         rbd_assert(size <= (size_t)INT_MAX);
4285         ret = (int)size;
4286 out:
4287         if (obj_request)
4288                 rbd_obj_request_put(obj_request);
4289         else
4290                 ceph_release_page_vector(pages, page_count);
4291
4292         return ret;
4293 }
4294
4295 /*
4296  * Read the complete header for the given rbd device.  On successful
4297  * return, the rbd_dev->header field will contain up-to-date
4298  * information about the image.
4299  */
4300 static int rbd_dev_v1_header_info(struct rbd_device *rbd_dev)
4301 {
4302         struct rbd_image_header_ondisk *ondisk = NULL;
4303         u32 snap_count = 0;
4304         u64 names_size = 0;
4305         u32 want_count;
4306         int ret;
4307
4308         /*
4309          * The complete header will include an array of its 64-bit
4310          * snapshot ids, followed by the names of those snapshots as
4311          * a contiguous block of NUL-terminated strings.  Note that
4312          * the number of snapshots could change by the time we read
4313          * it in, in which case we re-read it.
4314          */
4315         do {
4316                 size_t size;
4317
4318                 kfree(ondisk);
4319
4320                 size = sizeof (*ondisk);
4321                 size += snap_count * sizeof (struct rbd_image_snap_ondisk);
4322                 size += names_size;
4323                 ondisk = kmalloc(size, GFP_KERNEL);
4324                 if (!ondisk)
4325                         return -ENOMEM;
4326
4327                 ret = rbd_obj_read_sync(rbd_dev, rbd_dev->header_oid.name,
4328                                        0, size, ondisk);
4329                 if (ret < 0)
4330                         goto out;
4331                 if ((size_t)ret < size) {
4332                         ret = -ENXIO;
4333                         rbd_warn(rbd_dev, "short header read (want %zd got %d)",
4334                                 size, ret);
4335                         goto out;
4336                 }
4337                 if (!rbd_dev_ondisk_valid(ondisk)) {
4338                         ret = -ENXIO;
4339                         rbd_warn(rbd_dev, "invalid header");
4340                         goto out;
4341                 }
4342
4343                 names_size = le64_to_cpu(ondisk->snap_names_len);
4344                 want_count = snap_count;
4345                 snap_count = le32_to_cpu(ondisk->snap_count);
4346         } while (snap_count != want_count);
4347
4348         ret = rbd_header_from_disk(rbd_dev, ondisk);
4349 out:
4350         kfree(ondisk);
4351
4352         return ret;
4353 }
4354
4355 /*
4356  * Clear the rbd device's EXISTS flag if the snapshot it's mapped to
4357  * has disappeared from the (just updated) snapshot context.
4358  */
4359 static void rbd_exists_validate(struct rbd_device *rbd_dev)
4360 {
4361         u64 snap_id;
4362
4363         if (!test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags))
4364                 return;
4365
4366         snap_id = rbd_dev->spec->snap_id;
4367         if (snap_id == CEPH_NOSNAP)
4368                 return;
4369
4370         if (rbd_dev_snap_index(rbd_dev, snap_id) == BAD_SNAP_INDEX)
4371                 clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
4372 }
4373
4374 static void rbd_dev_update_size(struct rbd_device *rbd_dev)
4375 {
4376         sector_t size;
4377
4378         /*
4379          * If EXISTS is not set, rbd_dev->disk may be NULL, so don't
4380          * try to update its size.  If REMOVING is set, updating size
4381          * is just useless work since the device can't be opened.
4382          */
4383         if (test_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags) &&
4384             !test_bit(RBD_DEV_FLAG_REMOVING, &rbd_dev->flags)) {
4385                 size = (sector_t)rbd_dev->mapping.size / SECTOR_SIZE;
4386                 dout("setting size to %llu sectors", (unsigned long long)size);
4387                 set_capacity(rbd_dev->disk, size);
4388                 revalidate_disk(rbd_dev->disk);
4389         }
4390 }
4391
4392 static int rbd_dev_refresh(struct rbd_device *rbd_dev)
4393 {
4394         u64 mapping_size;
4395         int ret;
4396
4397         down_write(&rbd_dev->header_rwsem);
4398         mapping_size = rbd_dev->mapping.size;
4399
4400         ret = rbd_dev_header_info(rbd_dev);
4401         if (ret)
4402                 goto out;
4403
4404         /*
4405          * If there is a parent, see if it has disappeared due to the
4406          * mapped image getting flattened.
4407          */
4408         if (rbd_dev->parent) {
4409                 ret = rbd_dev_v2_parent_info(rbd_dev);
4410                 if (ret)
4411                         goto out;
4412         }
4413
4414         if (rbd_dev->spec->snap_id == CEPH_NOSNAP) {
4415                 rbd_dev->mapping.size = rbd_dev->header.image_size;
4416         } else {
4417                 /* validate mapped snapshot's EXISTS flag */
4418                 rbd_exists_validate(rbd_dev);
4419         }
4420
4421 out:
4422         up_write(&rbd_dev->header_rwsem);
4423         if (!ret && mapping_size != rbd_dev->mapping.size)
4424                 rbd_dev_update_size(rbd_dev);
4425
4426         return ret;
4427 }
4428
4429 static int rbd_init_request(void *data, struct request *rq,
4430                 unsigned int hctx_idx, unsigned int request_idx,
4431                 unsigned int numa_node)
4432 {
4433         struct work_struct *work = blk_mq_rq_to_pdu(rq);
4434
4435         INIT_WORK(work, rbd_queue_workfn);
4436         return 0;
4437 }
4438
4439 static struct blk_mq_ops rbd_mq_ops = {
4440         .queue_rq       = rbd_queue_rq,
4441         .map_queue      = blk_mq_map_queue,
4442         .init_request   = rbd_init_request,
4443 };
4444
4445 static int rbd_init_disk(struct rbd_device *rbd_dev)
4446 {
4447         struct gendisk *disk;
4448         struct request_queue *q;
4449         u64 segment_size;
4450         int err;
4451
4452         /* create gendisk info */
4453         disk = alloc_disk(single_major ?
4454                           (1 << RBD_SINGLE_MAJOR_PART_SHIFT) :
4455                           RBD_MINORS_PER_MAJOR);
4456         if (!disk)
4457                 return -ENOMEM;
4458
4459         snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d",
4460                  rbd_dev->dev_id);
4461         disk->major = rbd_dev->major;
4462         disk->first_minor = rbd_dev->minor;
4463         if (single_major)
4464                 disk->flags |= GENHD_FL_EXT_DEVT;
4465         disk->fops = &rbd_bd_ops;
4466         disk->private_data = rbd_dev;
4467
4468         memset(&rbd_dev->tag_set, 0, sizeof(rbd_dev->tag_set));
4469         rbd_dev->tag_set.ops = &rbd_mq_ops;
4470         rbd_dev->tag_set.queue_depth = rbd_dev->opts->queue_depth;
4471         rbd_dev->tag_set.numa_node = NUMA_NO_NODE;
4472         rbd_dev->tag_set.flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE;
4473         rbd_dev->tag_set.nr_hw_queues = 1;
4474         rbd_dev->tag_set.cmd_size = sizeof(struct work_struct);
4475
4476         err = blk_mq_alloc_tag_set(&rbd_dev->tag_set);
4477         if (err)
4478                 goto out_disk;
4479
4480         q = blk_mq_init_queue(&rbd_dev->tag_set);
4481         if (IS_ERR(q)) {
4482                 err = PTR_ERR(q);
4483                 goto out_tag_set;
4484         }
4485
4486         queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q);
4487         /* QUEUE_FLAG_ADD_RANDOM is off by default for blk-mq */
4488
4489         /* set io sizes to object size */
4490         segment_size = rbd_obj_bytes(&rbd_dev->header);
4491         blk_queue_max_hw_sectors(q, segment_size / SECTOR_SIZE);
4492         q->limits.max_sectors = queue_max_hw_sectors(q);
4493         blk_queue_max_segments(q, segment_size / SECTOR_SIZE);
4494         blk_queue_max_segment_size(q, segment_size);
4495         blk_queue_io_min(q, segment_size);
4496         blk_queue_io_opt(q, segment_size);
4497
4498         /* enable the discard support */
4499         queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q);
4500         q->limits.discard_granularity = segment_size;
4501         q->limits.discard_alignment = segment_size;
4502         blk_queue_max_discard_sectors(q, segment_size / SECTOR_SIZE);
4503         q->limits.discard_zeroes_data = 1;
4504
4505         if (!ceph_test_opt(rbd_dev->rbd_client->client, NOCRC))
4506                 q->backing_dev_info.capabilities |= BDI_CAP_STABLE_WRITES;
4507
4508         disk->queue = q;
4509
4510         q->queuedata = rbd_dev;
4511
4512         rbd_dev->disk = disk;
4513
4514         return 0;
4515 out_tag_set:
4516         blk_mq_free_tag_set(&rbd_dev->tag_set);
4517 out_disk:
4518         put_disk(disk);
4519         return err;
4520 }
4521
4522 /*
4523   sysfs
4524 */
4525
4526 static struct rbd_device *dev_to_rbd_dev(struct device *dev)
4527 {
4528         return container_of(dev, struct rbd_device, dev);
4529 }
4530
4531 static ssize_t rbd_size_show(struct device *dev,
4532                              struct device_attribute *attr, char *buf)
4533 {
4534         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4535
4536         return sprintf(buf, "%llu\n",
4537                 (unsigned long long)rbd_dev->mapping.size);
4538 }
4539
4540 /*
4541  * Note this shows the features for whatever's mapped, which is not
4542  * necessarily the base image.
4543  */
4544 static ssize_t rbd_features_show(struct device *dev,
4545                              struct device_attribute *attr, char *buf)
4546 {
4547         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4548
4549         return sprintf(buf, "0x%016llx\n",
4550                         (unsigned long long)rbd_dev->mapping.features);
4551 }
4552
4553 static ssize_t rbd_major_show(struct device *dev,
4554                               struct device_attribute *attr, char *buf)
4555 {
4556         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4557
4558         if (rbd_dev->major)
4559                 return sprintf(buf, "%d\n", rbd_dev->major);
4560
4561         return sprintf(buf, "(none)\n");
4562 }
4563
4564 static ssize_t rbd_minor_show(struct device *dev,
4565                               struct device_attribute *attr, char *buf)
4566 {
4567         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4568
4569         return sprintf(buf, "%d\n", rbd_dev->minor);
4570 }
4571
4572 static ssize_t rbd_client_addr_show(struct device *dev,
4573                                     struct device_attribute *attr, char *buf)
4574 {
4575         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4576         struct ceph_entity_addr *client_addr =
4577             ceph_client_addr(rbd_dev->rbd_client->client);
4578
4579         return sprintf(buf, "%pISpc/%u\n", &client_addr->in_addr,
4580                        le32_to_cpu(client_addr->nonce));
4581 }
4582
4583 static ssize_t rbd_client_id_show(struct device *dev,
4584                                   struct device_attribute *attr, char *buf)
4585 {
4586         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4587
4588         return sprintf(buf, "client%lld\n",
4589                        ceph_client_gid(rbd_dev->rbd_client->client));
4590 }
4591
4592 static ssize_t rbd_cluster_fsid_show(struct device *dev,
4593                                      struct device_attribute *attr, char *buf)
4594 {
4595         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4596
4597         return sprintf(buf, "%pU\n", &rbd_dev->rbd_client->client->fsid);
4598 }
4599
4600 static ssize_t rbd_config_info_show(struct device *dev,
4601                                     struct device_attribute *attr, char *buf)
4602 {
4603         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4604
4605         return sprintf(buf, "%s\n", rbd_dev->config_info);
4606 }
4607
4608 static ssize_t rbd_pool_show(struct device *dev,
4609                              struct device_attribute *attr, char *buf)
4610 {
4611         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4612
4613         return sprintf(buf, "%s\n", rbd_dev->spec->pool_name);
4614 }
4615
4616 static ssize_t rbd_pool_id_show(struct device *dev,
4617                              struct device_attribute *attr, char *buf)
4618 {
4619         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4620
4621         return sprintf(buf, "%llu\n",
4622                         (unsigned long long) rbd_dev->spec->pool_id);
4623 }
4624
4625 static ssize_t rbd_name_show(struct device *dev,
4626                              struct device_attribute *attr, char *buf)
4627 {
4628         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4629
4630         if (rbd_dev->spec->image_name)
4631                 return sprintf(buf, "%s\n", rbd_dev->spec->image_name);
4632
4633         return sprintf(buf, "(unknown)\n");
4634 }
4635
4636 static ssize_t rbd_image_id_show(struct device *dev,
4637                              struct device_attribute *attr, char *buf)
4638 {
4639         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4640
4641         return sprintf(buf, "%s\n", rbd_dev->spec->image_id);
4642 }
4643
4644 /*
4645  * Shows the name of the currently-mapped snapshot (or
4646  * RBD_SNAP_HEAD_NAME for the base image).
4647  */
4648 static ssize_t rbd_snap_show(struct device *dev,
4649                              struct device_attribute *attr,
4650                              char *buf)
4651 {
4652         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4653
4654         return sprintf(buf, "%s\n", rbd_dev->spec->snap_name);
4655 }
4656
4657 static ssize_t rbd_snap_id_show(struct device *dev,
4658                                 struct device_attribute *attr, char *buf)
4659 {
4660         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4661
4662         return sprintf(buf, "%llu\n", rbd_dev->spec->snap_id);
4663 }
4664
4665 /*
4666  * For a v2 image, shows the chain of parent images, separated by empty
4667  * lines.  For v1 images or if there is no parent, shows "(no parent
4668  * image)".
4669  */
4670 static ssize_t rbd_parent_show(struct device *dev,
4671                                struct device_attribute *attr,
4672                                char *buf)
4673 {
4674         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4675         ssize_t count = 0;
4676
4677         if (!rbd_dev->parent)
4678                 return sprintf(buf, "(no parent image)\n");
4679
4680         for ( ; rbd_dev->parent; rbd_dev = rbd_dev->parent) {
4681                 struct rbd_spec *spec = rbd_dev->parent_spec;
4682
4683                 count += sprintf(&buf[count], "%s"
4684                             "pool_id %llu\npool_name %s\n"
4685                             "image_id %s\nimage_name %s\n"
4686                             "snap_id %llu\nsnap_name %s\n"
4687                             "overlap %llu\n",
4688                             !count ? "" : "\n", /* first? */
4689                             spec->pool_id, spec->pool_name,
4690                             spec->image_id, spec->image_name ?: "(unknown)",
4691                             spec->snap_id, spec->snap_name,
4692                             rbd_dev->parent_overlap);
4693         }
4694
4695         return count;
4696 }
4697
4698 static ssize_t rbd_image_refresh(struct device *dev,
4699                                  struct device_attribute *attr,
4700                                  const char *buf,
4701                                  size_t size)
4702 {
4703         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4704         int ret;
4705
4706         ret = rbd_dev_refresh(rbd_dev);
4707         if (ret)
4708                 return ret;
4709
4710         return size;
4711 }
4712
4713 static DEVICE_ATTR(size, S_IRUGO, rbd_size_show, NULL);
4714 static DEVICE_ATTR(features, S_IRUGO, rbd_features_show, NULL);
4715 static DEVICE_ATTR(major, S_IRUGO, rbd_major_show, NULL);
4716 static DEVICE_ATTR(minor, S_IRUGO, rbd_minor_show, NULL);
4717 static DEVICE_ATTR(client_addr, S_IRUGO, rbd_client_addr_show, NULL);
4718 static DEVICE_ATTR(client_id, S_IRUGO, rbd_client_id_show, NULL);
4719 static DEVICE_ATTR(cluster_fsid, S_IRUGO, rbd_cluster_fsid_show, NULL);
4720 static DEVICE_ATTR(config_info, S_IRUSR, rbd_config_info_show, NULL);
4721 static DEVICE_ATTR(pool, S_IRUGO, rbd_pool_show, NULL);
4722 static DEVICE_ATTR(pool_id, S_IRUGO, rbd_pool_id_show, NULL);
4723 static DEVICE_ATTR(name, S_IRUGO, rbd_name_show, NULL);
4724 static DEVICE_ATTR(image_id, S_IRUGO, rbd_image_id_show, NULL);
4725 static DEVICE_ATTR(refresh, S_IWUSR, NULL, rbd_image_refresh);
4726 static DEVICE_ATTR(current_snap, S_IRUGO, rbd_snap_show, NULL);
4727 static DEVICE_ATTR(snap_id, S_IRUGO, rbd_snap_id_show, NULL);
4728 static DEVICE_ATTR(parent, S_IRUGO, rbd_parent_show, NULL);
4729
4730 static struct attribute *rbd_attrs[] = {
4731         &dev_attr_size.attr,
4732         &dev_attr_features.attr,
4733         &dev_attr_major.attr,
4734         &dev_attr_minor.attr,
4735         &dev_attr_client_addr.attr,
4736         &dev_attr_client_id.attr,
4737         &dev_attr_cluster_fsid.attr,
4738         &dev_attr_config_info.attr,
4739         &dev_attr_pool.attr,
4740         &dev_attr_pool_id.attr,
4741         &dev_attr_name.attr,
4742         &dev_attr_image_id.attr,
4743         &dev_attr_current_snap.attr,
4744         &dev_attr_snap_id.attr,
4745         &dev_attr_parent.attr,
4746         &dev_attr_refresh.attr,
4747         NULL
4748 };
4749
4750 static struct attribute_group rbd_attr_group = {
4751         .attrs = rbd_attrs,
4752 };
4753
4754 static const struct attribute_group *rbd_attr_groups[] = {
4755         &rbd_attr_group,
4756         NULL
4757 };
4758
4759 static void rbd_dev_release(struct device *dev);
4760
4761 static struct device_type rbd_device_type = {
4762         .name           = "rbd",
4763         .groups         = rbd_attr_groups,
4764         .release        = rbd_dev_release,
4765 };
4766
4767 static struct rbd_spec *rbd_spec_get(struct rbd_spec *spec)
4768 {
4769         kref_get(&spec->kref);
4770
4771         return spec;
4772 }
4773
4774 static void rbd_spec_free(struct kref *kref);
4775 static void rbd_spec_put(struct rbd_spec *spec)
4776 {
4777         if (spec)
4778                 kref_put(&spec->kref, rbd_spec_free);
4779 }
4780
4781 static struct rbd_spec *rbd_spec_alloc(void)
4782 {
4783         struct rbd_spec *spec;
4784
4785         spec = kzalloc(sizeof (*spec), GFP_KERNEL);
4786         if (!spec)
4787                 return NULL;
4788
4789         spec->pool_id = CEPH_NOPOOL;
4790         spec->snap_id = CEPH_NOSNAP;
4791         kref_init(&spec->kref);
4792
4793         return spec;
4794 }
4795
4796 static void rbd_spec_free(struct kref *kref)
4797 {
4798         struct rbd_spec *spec = container_of(kref, struct rbd_spec, kref);
4799
4800         kfree(spec->pool_name);
4801         kfree(spec->image_id);
4802         kfree(spec->image_name);
4803         kfree(spec->snap_name);
4804         kfree(spec);
4805 }
4806
4807 static void rbd_dev_free(struct rbd_device *rbd_dev)
4808 {
4809         WARN_ON(rbd_dev->watch_state != RBD_WATCH_STATE_UNREGISTERED);
4810         WARN_ON(rbd_dev->lock_state != RBD_LOCK_STATE_UNLOCKED);
4811
4812         ceph_oid_destroy(&rbd_dev->header_oid);
4813         ceph_oloc_destroy(&rbd_dev->header_oloc);
4814         kfree(rbd_dev->config_info);
4815
4816         rbd_put_client(rbd_dev->rbd_client);
4817         rbd_spec_put(rbd_dev->spec);
4818         kfree(rbd_dev->opts);
4819         kfree(rbd_dev);
4820 }
4821
4822 static void rbd_dev_release(struct device *dev)
4823 {
4824         struct rbd_device *rbd_dev = dev_to_rbd_dev(dev);
4825         bool need_put = !!rbd_dev->opts;
4826
4827         if (need_put) {
4828                 destroy_workqueue(rbd_dev->task_wq);
4829                 ida_simple_remove(&rbd_dev_id_ida, rbd_dev->dev_id);
4830         }
4831
4832         rbd_dev_free(rbd_dev);
4833
4834         /*
4835          * This is racy, but way better than putting module outside of
4836          * the release callback.  The race window is pretty small, so
4837          * doing something similar to dm (dm-builtin.c) is overkill.
4838          */
4839         if (need_put)
4840                 module_put(THIS_MODULE);
4841 }
4842
4843 static struct rbd_device *__rbd_dev_create(struct rbd_client *rbdc,
4844                                            struct rbd_spec *spec)
4845 {
4846         struct rbd_device *rbd_dev;
4847
4848         rbd_dev = kzalloc(sizeof(*rbd_dev), GFP_KERNEL);
4849         if (!rbd_dev)
4850                 return NULL;
4851
4852         spin_lock_init(&rbd_dev->lock);
4853         INIT_LIST_HEAD(&rbd_dev->node);
4854         init_rwsem(&rbd_dev->header_rwsem);
4855
4856         ceph_oid_init(&rbd_dev->header_oid);
4857         ceph_oloc_init(&rbd_dev->header_oloc);
4858
4859         mutex_init(&rbd_dev->watch_mutex);
4860         rbd_dev->watch_state = RBD_WATCH_STATE_UNREGISTERED;
4861         INIT_DELAYED_WORK(&rbd_dev->watch_dwork, rbd_reregister_watch);
4862
4863         init_rwsem(&rbd_dev->lock_rwsem);
4864         rbd_dev->lock_state = RBD_LOCK_STATE_UNLOCKED;
4865         INIT_WORK(&rbd_dev->acquired_lock_work, rbd_notify_acquired_lock);
4866         INIT_WORK(&rbd_dev->released_lock_work, rbd_notify_released_lock);
4867         INIT_DELAYED_WORK(&rbd_dev->lock_dwork, rbd_acquire_lock);
4868         INIT_WORK(&rbd_dev->unlock_work, rbd_release_lock_work);
4869         init_waitqueue_head(&rbd_dev->lock_waitq);
4870
4871         rbd_dev->dev.bus = &rbd_bus_type;
4872         rbd_dev->dev.type = &rbd_device_type;
4873         rbd_dev->dev.parent = &rbd_root_dev;
4874         device_initialize(&rbd_dev->dev);
4875
4876         rbd_dev->rbd_client = rbdc;
4877         rbd_dev->spec = spec;
4878
4879         rbd_dev->layout.stripe_unit = 1 << RBD_MAX_OBJ_ORDER;
4880         rbd_dev->layout.stripe_count = 1;
4881         rbd_dev->layout.object_size = 1 << RBD_MAX_OBJ_ORDER;
4882         rbd_dev->layout.pool_id = spec->pool_id;
4883         RCU_INIT_POINTER(rbd_dev->layout.pool_ns, NULL);
4884
4885         return rbd_dev;
4886 }
4887
4888 /*
4889  * Create a mapping rbd_dev.
4890  */
4891 static struct rbd_device *rbd_dev_create(struct rbd_client *rbdc,
4892                                          struct rbd_spec *spec,
4893                                          struct rbd_options *opts)
4894 {
4895         struct rbd_device *rbd_dev;
4896
4897         rbd_dev = __rbd_dev_create(rbdc, spec);
4898         if (!rbd_dev)
4899                 return NULL;
4900
4901         rbd_dev->opts = opts;
4902
4903         /* get an id and fill in device name */
4904         rbd_dev->dev_id = ida_simple_get(&rbd_dev_id_ida, 0,
4905                                          minor_to_rbd_dev_id(1 << MINORBITS),
4906                                          GFP_KERNEL);
4907         if (rbd_dev->dev_id < 0)
4908                 goto fail_rbd_dev;
4909
4910         sprintf(rbd_dev->name, RBD_DRV_NAME "%d", rbd_dev->dev_id);
4911         rbd_dev->task_wq = alloc_ordered_workqueue("%s-tasks", WQ_MEM_RECLAIM,
4912                                                    rbd_dev->name);
4913         if (!rbd_dev->task_wq)
4914                 goto fail_dev_id;
4915
4916         /* we have a ref from do_rbd_add() */
4917         __module_get(THIS_MODULE);
4918
4919         dout("%s rbd_dev %p dev_id %d\n", __func__, rbd_dev, rbd_dev->dev_id);
4920         return rbd_dev;
4921
4922 fail_dev_id:
4923         ida_simple_remove(&rbd_dev_id_ida, rbd_dev->dev_id);
4924 fail_rbd_dev:
4925         rbd_dev_free(rbd_dev);
4926         return NULL;
4927 }
4928
4929 static void rbd_dev_destroy(struct rbd_device *rbd_dev)
4930 {
4931         if (rbd_dev)
4932                 put_device(&rbd_dev->dev);
4933 }
4934
4935 /*
4936  * Get the size and object order for an image snapshot, or if
4937  * snap_id is CEPH_NOSNAP, gets this information for the base
4938  * image.
4939  */
4940 static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id,
4941                                 u8 *order, u64 *snap_size)
4942 {
4943         __le64 snapid = cpu_to_le64(snap_id);
4944         int ret;
4945         struct {
4946                 u8 order;
4947                 __le64 size;
4948         } __attribute__ ((packed)) size_buf = { 0 };
4949
4950         ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
4951                                 "rbd", "get_size",
4952                                 &snapid, sizeof (snapid),
4953                                 &size_buf, sizeof (size_buf));
4954         dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
4955         if (ret < 0)
4956                 return ret;
4957         if (ret < sizeof (size_buf))
4958                 return -ERANGE;
4959
4960         if (order) {
4961                 *order = size_buf.order;
4962                 dout("  order %u", (unsigned int)*order);
4963         }
4964         *snap_size = le64_to_cpu(size_buf.size);
4965
4966         dout("  snap_id 0x%016llx snap_size = %llu\n",
4967                 (unsigned long long)snap_id,
4968                 (unsigned long long)*snap_size);
4969
4970         return 0;
4971 }
4972
4973 static int rbd_dev_v2_image_size(struct rbd_device *rbd_dev)
4974 {
4975         return _rbd_dev_v2_snap_size(rbd_dev, CEPH_NOSNAP,
4976                                         &rbd_dev->header.obj_order,
4977                                         &rbd_dev->header.image_size);
4978 }
4979
4980 static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev)
4981 {
4982         void *reply_buf;
4983         int ret;
4984         void *p;
4985
4986         reply_buf = kzalloc(RBD_OBJ_PREFIX_LEN_MAX, GFP_KERNEL);
4987         if (!reply_buf)
4988                 return -ENOMEM;
4989
4990         ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
4991                                 "rbd", "get_object_prefix", NULL, 0,
4992                                 reply_buf, RBD_OBJ_PREFIX_LEN_MAX);
4993         dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
4994         if (ret < 0)
4995                 goto out;
4996
4997         p = reply_buf;
4998         rbd_dev->header.object_prefix = ceph_extract_encoded_string(&p,
4999                                                 p + ret, NULL, GFP_NOIO);
5000         ret = 0;
5001
5002         if (IS_ERR(rbd_dev->header.object_prefix)) {
5003                 ret = PTR_ERR(rbd_dev->header.object_prefix);
5004                 rbd_dev->header.object_prefix = NULL;
5005         } else {
5006                 dout("  object_prefix = %s\n", rbd_dev->header.object_prefix);
5007         }
5008 out:
5009         kfree(reply_buf);
5010
5011         return ret;
5012 }
5013
5014 static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id,
5015                 u64 *snap_features)
5016 {
5017         __le64 snapid = cpu_to_le64(snap_id);
5018         struct {
5019                 __le64 features;
5020                 __le64 incompat;
5021         } __attribute__ ((packed)) features_buf = { 0 };
5022         u64 unsup;
5023         int ret;
5024
5025         ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
5026                                 "rbd", "get_features",
5027                                 &snapid, sizeof (snapid),
5028                                 &features_buf, sizeof (features_buf));
5029         dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
5030         if (ret < 0)
5031                 return ret;
5032         if (ret < sizeof (features_buf))
5033                 return -ERANGE;
5034
5035         unsup = le64_to_cpu(features_buf.incompat) & ~RBD_FEATURES_SUPPORTED;
5036         if (unsup) {
5037                 rbd_warn(rbd_dev, "image uses unsupported features: 0x%llx",
5038                          unsup);
5039                 return -ENXIO;
5040         }
5041
5042         *snap_features = le64_to_cpu(features_buf.features);
5043
5044         dout("  snap_id 0x%016llx features = 0x%016llx incompat = 0x%016llx\n",
5045                 (unsigned long long)snap_id,
5046                 (unsigned long long)*snap_features,
5047                 (unsigned long long)le64_to_cpu(features_buf.incompat));
5048
5049         return 0;
5050 }
5051
5052 static int rbd_dev_v2_features(struct rbd_device *rbd_dev)
5053 {
5054         return _rbd_dev_v2_snap_features(rbd_dev, CEPH_NOSNAP,
5055                                                 &rbd_dev->header.features);
5056 }
5057
5058 static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev)
5059 {
5060         struct rbd_spec *parent_spec;
5061         size_t size;
5062         void *reply_buf = NULL;
5063         __le64 snapid;
5064         void *p;
5065         void *end;
5066         u64 pool_id;
5067         char *image_id;
5068         u64 snap_id;
5069         u64 overlap;
5070         int ret;
5071
5072         parent_spec = rbd_spec_alloc();
5073         if (!parent_spec)
5074                 return -ENOMEM;
5075
5076         size = sizeof (__le64) +                                /* pool_id */
5077                 sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX +        /* image_id */
5078                 sizeof (__le64) +                               /* snap_id */
5079                 sizeof (__le64);                                /* overlap */
5080         reply_buf = kmalloc(size, GFP_KERNEL);
5081         if (!reply_buf) {
5082                 ret = -ENOMEM;
5083                 goto out_err;
5084         }
5085
5086         snapid = cpu_to_le64(rbd_dev->spec->snap_id);
5087         ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
5088                                 "rbd", "get_parent",
5089                                 &snapid, sizeof (snapid),
5090                                 reply_buf, size);
5091         dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
5092         if (ret < 0)
5093                 goto out_err;
5094
5095         p = reply_buf;
5096         end = reply_buf + ret;
5097         ret = -ERANGE;
5098         ceph_decode_64_safe(&p, end, pool_id, out_err);
5099         if (pool_id == CEPH_NOPOOL) {
5100                 /*
5101                  * Either the parent never existed, or we have
5102                  * record of it but the image got flattened so it no
5103                  * longer has a parent.  When the parent of a
5104                  * layered image disappears we immediately set the
5105                  * overlap to 0.  The effect of this is that all new
5106                  * requests will be treated as if the image had no
5107                  * parent.
5108                  */
5109                 if (rbd_dev->parent_overlap) {
5110                         rbd_dev->parent_overlap = 0;
5111                         rbd_dev_parent_put(rbd_dev);
5112                         pr_info("%s: clone image has been flattened\n",
5113                                 rbd_dev->disk->disk_name);
5114                 }
5115
5116                 goto out;       /* No parent?  No problem. */
5117         }
5118
5119         /* The ceph file layout needs to fit pool id in 32 bits */
5120
5121         ret = -EIO;
5122         if (pool_id > (u64)U32_MAX) {
5123                 rbd_warn(NULL, "parent pool id too large (%llu > %u)",
5124                         (unsigned long long)pool_id, U32_MAX);
5125                 goto out_err;
5126         }
5127
5128         image_id = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL);
5129         if (IS_ERR(image_id)) {
5130                 ret = PTR_ERR(image_id);
5131                 goto out_err;
5132         }
5133         ceph_decode_64_safe(&p, end, snap_id, out_err);
5134         ceph_decode_64_safe(&p, end, overlap, out_err);
5135
5136         /*
5137          * The parent won't change (except when the clone is
5138          * flattened, already handled that).  So we only need to
5139          * record the parent spec we have not already done so.
5140          */
5141         if (!rbd_dev->parent_spec) {
5142                 parent_spec->pool_id = pool_id;
5143                 parent_spec->image_id = image_id;
5144                 parent_spec->snap_id = snap_id;
5145                 rbd_dev->parent_spec = parent_spec;
5146                 parent_spec = NULL;     /* rbd_dev now owns this */
5147         } else {
5148                 kfree(image_id);
5149         }
5150
5151         /*
5152          * We always update the parent overlap.  If it's zero we issue
5153          * a warning, as we will proceed as if there was no parent.
5154          */
5155         if (!overlap) {
5156                 if (parent_spec) {
5157                         /* refresh, careful to warn just once */
5158                         if (rbd_dev->parent_overlap)
5159                                 rbd_warn(rbd_dev,
5160                                     "clone now standalone (overlap became 0)");
5161                 } else {
5162                         /* initial probe */
5163                         rbd_warn(rbd_dev, "clone is standalone (overlap 0)");
5164                 }
5165         }
5166         rbd_dev->parent_overlap = overlap;
5167
5168 out:
5169         ret = 0;
5170 out_err:
5171         kfree(reply_buf);
5172         rbd_spec_put(parent_spec);
5173
5174         return ret;
5175 }
5176
5177 static int rbd_dev_v2_striping_info(struct rbd_device *rbd_dev)
5178 {
5179         struct {
5180                 __le64 stripe_unit;
5181                 __le64 stripe_count;
5182         } __attribute__ ((packed)) striping_info_buf = { 0 };
5183         size_t size = sizeof (striping_info_buf);
5184         void *p;
5185         u64 obj_size;
5186         u64 stripe_unit;
5187         u64 stripe_count;
5188         int ret;
5189
5190         ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
5191                                 "rbd", "get_stripe_unit_count", NULL, 0,
5192                                 (char *)&striping_info_buf, size);
5193         dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
5194         if (ret < 0)
5195                 return ret;
5196         if (ret < size)
5197                 return -ERANGE;
5198
5199         /*
5200          * We don't actually support the "fancy striping" feature
5201          * (STRIPINGV2) yet, but if the striping sizes are the
5202          * defaults the behavior is the same as before.  So find
5203          * out, and only fail if the image has non-default values.
5204          */
5205         ret = -EINVAL;
5206         obj_size = (u64)1 << rbd_dev->header.obj_order;
5207         p = &striping_info_buf;
5208         stripe_unit = ceph_decode_64(&p);
5209         if (stripe_unit != obj_size) {
5210                 rbd_warn(rbd_dev, "unsupported stripe unit "
5211                                 "(got %llu want %llu)",
5212                                 stripe_unit, obj_size);
5213                 return -EINVAL;
5214         }
5215         stripe_count = ceph_decode_64(&p);
5216         if (stripe_count != 1) {
5217                 rbd_warn(rbd_dev, "unsupported stripe count "
5218                                 "(got %llu want 1)", stripe_count);
5219                 return -EINVAL;
5220         }
5221         rbd_dev->header.stripe_unit = stripe_unit;
5222         rbd_dev->header.stripe_count = stripe_count;
5223
5224         return 0;
5225 }
5226
5227 static char *rbd_dev_image_name(struct rbd_device *rbd_dev)
5228 {
5229         size_t image_id_size;
5230         char *image_id;
5231         void *p;
5232         void *end;
5233         size_t size;
5234         void *reply_buf = NULL;
5235         size_t len = 0;
5236         char *image_name = NULL;
5237         int ret;
5238
5239         rbd_assert(!rbd_dev->spec->image_name);
5240
5241         len = strlen(rbd_dev->spec->image_id);
5242         image_id_size = sizeof (__le32) + len;
5243         image_id = kmalloc(image_id_size, GFP_KERNEL);
5244         if (!image_id)
5245                 return NULL;
5246
5247         p = image_id;
5248         end = image_id + image_id_size;
5249         ceph_encode_string(&p, end, rbd_dev->spec->image_id, (u32)len);
5250
5251         size = sizeof (__le32) + RBD_IMAGE_NAME_LEN_MAX;
5252         reply_buf = kmalloc(size, GFP_KERNEL);
5253         if (!reply_buf)
5254                 goto out;
5255
5256         ret = rbd_obj_method_sync(rbd_dev, RBD_DIRECTORY,
5257                                 "rbd", "dir_get_name",
5258                                 image_id, image_id_size,
5259                                 reply_buf, size);
5260         if (ret < 0)
5261                 goto out;
5262         p = reply_buf;
5263         end = reply_buf + ret;
5264
5265         image_name = ceph_extract_encoded_string(&p, end, &len, GFP_KERNEL);
5266         if (IS_ERR(image_name))
5267                 image_name = NULL;
5268         else
5269                 dout("%s: name is %s len is %zd\n", __func__, image_name, len);
5270 out:
5271         kfree(reply_buf);
5272         kfree(image_id);
5273
5274         return image_name;
5275 }
5276
5277 static u64 rbd_v1_snap_id_by_name(struct rbd_device *rbd_dev, const char *name)
5278 {
5279         struct ceph_snap_context *snapc = rbd_dev->header.snapc;
5280         const char *snap_name;
5281         u32 which = 0;
5282
5283         /* Skip over names until we find the one we are looking for */
5284
5285         snap_name = rbd_dev->header.snap_names;
5286         while (which < snapc->num_snaps) {
5287                 if (!strcmp(name, snap_name))
5288                         return snapc->snaps[which];
5289                 snap_name += strlen(snap_name) + 1;
5290                 which++;
5291         }
5292         return CEPH_NOSNAP;
5293 }
5294
5295 static u64 rbd_v2_snap_id_by_name(struct rbd_device *rbd_dev, const char *name)
5296 {
5297         struct ceph_snap_context *snapc = rbd_dev->header.snapc;
5298         u32 which;
5299         bool found = false;
5300         u64 snap_id;
5301
5302         for (which = 0; !found && which < snapc->num_snaps; which++) {
5303                 const char *snap_name;
5304
5305                 snap_id = snapc->snaps[which];
5306                 snap_name = rbd_dev_v2_snap_name(rbd_dev, snap_id);
5307                 if (IS_ERR(snap_name)) {
5308                         /* ignore no-longer existing snapshots */
5309                         if (PTR_ERR(snap_name) == -ENOENT)
5310                                 continue;
5311                         else
5312                                 break;
5313                 }
5314                 found = !strcmp(name, snap_name);
5315                 kfree(snap_name);
5316         }
5317         return found ? snap_id : CEPH_NOSNAP;
5318 }
5319
5320 /*
5321  * Assumes name is never RBD_SNAP_HEAD_NAME; returns CEPH_NOSNAP if
5322  * no snapshot by that name is found, or if an error occurs.
5323  */
5324 static u64 rbd_snap_id_by_name(struct rbd_device *rbd_dev, const char *name)
5325 {
5326         if (rbd_dev->image_format == 1)
5327                 return rbd_v1_snap_id_by_name(rbd_dev, name);
5328
5329         return rbd_v2_snap_id_by_name(rbd_dev, name);
5330 }
5331
5332 /*
5333  * An image being mapped will have everything but the snap id.
5334  */
5335 static int rbd_spec_fill_snap_id(struct rbd_device *rbd_dev)
5336 {
5337         struct rbd_spec *spec = rbd_dev->spec;
5338
5339         rbd_assert(spec->pool_id != CEPH_NOPOOL && spec->pool_name);
5340         rbd_assert(spec->image_id && spec->image_name);
5341         rbd_assert(spec->snap_name);
5342
5343         if (strcmp(spec->snap_name, RBD_SNAP_HEAD_NAME)) {
5344                 u64 snap_id;
5345
5346                 snap_id = rbd_snap_id_by_name(rbd_dev, spec->snap_name);
5347                 if (snap_id == CEPH_NOSNAP)
5348                         return -ENOENT;
5349
5350                 spec->snap_id = snap_id;
5351         } else {
5352                 spec->snap_id = CEPH_NOSNAP;
5353         }
5354
5355         return 0;
5356 }
5357
5358 /*
5359  * A parent image will have all ids but none of the names.
5360  *
5361  * All names in an rbd spec are dynamically allocated.  It's OK if we
5362  * can't figure out the name for an image id.
5363  */
5364 static int rbd_spec_fill_names(struct rbd_device *rbd_dev)
5365 {
5366         struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
5367         struct rbd_spec *spec = rbd_dev->spec;
5368         const char *pool_name;
5369         const char *image_name;
5370         const char *snap_name;
5371         int ret;
5372
5373         rbd_assert(spec->pool_id != CEPH_NOPOOL);
5374         rbd_assert(spec->image_id);
5375         rbd_assert(spec->snap_id != CEPH_NOSNAP);
5376
5377         /* Get the pool name; we have to make our own copy of this */
5378
5379         pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, spec->pool_id);
5380         if (!pool_name) {
5381                 rbd_warn(rbd_dev, "no pool with id %llu", spec->pool_id);
5382                 return -EIO;
5383         }
5384         pool_name = kstrdup(pool_name, GFP_KERNEL);
5385         if (!pool_name)
5386                 return -ENOMEM;
5387
5388         /* Fetch the image name; tolerate failure here */
5389
5390         image_name = rbd_dev_image_name(rbd_dev);
5391         if (!image_name)
5392                 rbd_warn(rbd_dev, "unable to get image name");
5393
5394         /* Fetch the snapshot name */
5395
5396         snap_name = rbd_snap_name(rbd_dev, spec->snap_id);
5397         if (IS_ERR(snap_name)) {
5398                 ret = PTR_ERR(snap_name);
5399                 goto out_err;
5400         }
5401
5402         spec->pool_name = pool_name;
5403         spec->image_name = image_name;
5404         spec->snap_name = snap_name;
5405
5406         return 0;
5407
5408 out_err:
5409         kfree(image_name);
5410         kfree(pool_name);
5411         return ret;
5412 }
5413
5414 static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev)
5415 {
5416         size_t size;
5417         int ret;
5418         void *reply_buf;
5419         void *p;
5420         void *end;
5421         u64 seq;
5422         u32 snap_count;
5423         struct ceph_snap_context *snapc;
5424         u32 i;
5425
5426         /*
5427          * We'll need room for the seq value (maximum snapshot id),
5428          * snapshot count, and array of that many snapshot ids.
5429          * For now we have a fixed upper limit on the number we're
5430          * prepared to receive.
5431          */
5432         size = sizeof (__le64) + sizeof (__le32) +
5433                         RBD_MAX_SNAP_COUNT * sizeof (__le64);
5434         reply_buf = kzalloc(size, GFP_KERNEL);
5435         if (!reply_buf)
5436                 return -ENOMEM;
5437
5438         ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
5439                                 "rbd", "get_snapcontext", NULL, 0,
5440                                 reply_buf, size);
5441         dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
5442         if (ret < 0)
5443                 goto out;
5444
5445         p = reply_buf;
5446         end = reply_buf + ret;
5447         ret = -ERANGE;
5448         ceph_decode_64_safe(&p, end, seq, out);
5449         ceph_decode_32_safe(&p, end, snap_count, out);
5450
5451         /*
5452          * Make sure the reported number of snapshot ids wouldn't go
5453          * beyond the end of our buffer.  But before checking that,
5454          * make sure the computed size of the snapshot context we
5455          * allocate is representable in a size_t.
5456          */
5457         if (snap_count > (SIZE_MAX - sizeof (struct ceph_snap_context))
5458                                  / sizeof (u64)) {
5459                 ret = -EINVAL;
5460                 goto out;
5461         }
5462         if (!ceph_has_room(&p, end, snap_count * sizeof (__le64)))
5463                 goto out;
5464         ret = 0;
5465
5466         snapc = ceph_create_snap_context(snap_count, GFP_KERNEL);
5467         if (!snapc) {
5468                 ret = -ENOMEM;
5469                 goto out;
5470         }
5471         snapc->seq = seq;
5472         for (i = 0; i < snap_count; i++)
5473                 snapc->snaps[i] = ceph_decode_64(&p);
5474
5475         ceph_put_snap_context(rbd_dev->header.snapc);
5476         rbd_dev->header.snapc = snapc;
5477
5478         dout("  snap context seq = %llu, snap_count = %u\n",
5479                 (unsigned long long)seq, (unsigned int)snap_count);
5480 out:
5481         kfree(reply_buf);
5482
5483         return ret;
5484 }
5485
5486 static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev,
5487                                         u64 snap_id)
5488 {
5489         size_t size;
5490         void *reply_buf;
5491         __le64 snapid;
5492         int ret;
5493         void *p;
5494         void *end;
5495         char *snap_name;
5496
5497         size = sizeof (__le32) + RBD_MAX_SNAP_NAME_LEN;
5498         reply_buf = kmalloc(size, GFP_KERNEL);
5499         if (!reply_buf)
5500                 return ERR_PTR(-ENOMEM);
5501
5502         snapid = cpu_to_le64(snap_id);
5503         ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
5504                                 "rbd", "get_snapshot_name",
5505                                 &snapid, sizeof (snapid),
5506                                 reply_buf, size);
5507         dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
5508         if (ret < 0) {
5509                 snap_name = ERR_PTR(ret);
5510                 goto out;
5511         }
5512
5513         p = reply_buf;
5514         end = reply_buf + ret;
5515         snap_name = ceph_extract_encoded_string(&p, end, NULL, GFP_KERNEL);
5516         if (IS_ERR(snap_name))
5517                 goto out;
5518
5519         dout("  snap_id 0x%016llx snap_name = %s\n",
5520                 (unsigned long long)snap_id, snap_name);
5521 out:
5522         kfree(reply_buf);
5523
5524         return snap_name;
5525 }
5526
5527 static int rbd_dev_v2_header_info(struct rbd_device *rbd_dev)
5528 {
5529         bool first_time = rbd_dev->header.object_prefix == NULL;
5530         int ret;
5531
5532         ret = rbd_dev_v2_image_size(rbd_dev);
5533         if (ret)
5534                 return ret;
5535
5536         if (first_time) {
5537                 ret = rbd_dev_v2_header_onetime(rbd_dev);
5538                 if (ret)
5539                         return ret;
5540         }
5541
5542         ret = rbd_dev_v2_snap_context(rbd_dev);
5543         if (ret && first_time) {
5544                 kfree(rbd_dev->header.object_prefix);
5545                 rbd_dev->header.object_prefix = NULL;
5546         }
5547
5548         return ret;
5549 }
5550
5551 static int rbd_dev_header_info(struct rbd_device *rbd_dev)
5552 {
5553         rbd_assert(rbd_image_format_valid(rbd_dev->image_format));
5554
5555         if (rbd_dev->image_format == 1)
5556                 return rbd_dev_v1_header_info(rbd_dev);
5557
5558         return rbd_dev_v2_header_info(rbd_dev);
5559 }
5560
5561 /*
5562  * Skips over white space at *buf, and updates *buf to point to the
5563  * first found non-space character (if any). Returns the length of
5564  * the token (string of non-white space characters) found.  Note
5565  * that *buf must be terminated with '\0'.
5566  */
5567 static inline size_t next_token(const char **buf)
5568 {
5569         /*
5570         * These are the characters that produce nonzero for
5571         * isspace() in the "C" and "POSIX" locales.
5572         */
5573         const char *spaces = " \f\n\r\t\v";
5574
5575         *buf += strspn(*buf, spaces);   /* Find start of token */
5576
5577         return strcspn(*buf, spaces);   /* Return token length */
5578 }
5579
5580 /*
5581  * Finds the next token in *buf, dynamically allocates a buffer big
5582  * enough to hold a copy of it, and copies the token into the new
5583  * buffer.  The copy is guaranteed to be terminated with '\0'.  Note
5584  * that a duplicate buffer is created even for a zero-length token.
5585  *
5586  * Returns a pointer to the newly-allocated duplicate, or a null
5587  * pointer if memory for the duplicate was not available.  If
5588  * the lenp argument is a non-null pointer, the length of the token
5589  * (not including the '\0') is returned in *lenp.
5590  *
5591  * If successful, the *buf pointer will be updated to point beyond
5592  * the end of the found token.
5593  *
5594  * Note: uses GFP_KERNEL for allocation.
5595  */
5596 static inline char *dup_token(const char **buf, size_t *lenp)
5597 {
5598         char *dup;
5599         size_t len;
5600
5601         len = next_token(buf);
5602         dup = kmemdup(*buf, len + 1, GFP_KERNEL);
5603         if (!dup)
5604                 return NULL;
5605         *(dup + len) = '\0';
5606         *buf += len;
5607
5608         if (lenp)
5609                 *lenp = len;
5610
5611         return dup;
5612 }
5613
5614 /*
5615  * Parse the options provided for an "rbd add" (i.e., rbd image
5616  * mapping) request.  These arrive via a write to /sys/bus/rbd/add,
5617  * and the data written is passed here via a NUL-terminated buffer.
5618  * Returns 0 if successful or an error code otherwise.
5619  *
5620  * The information extracted from these options is recorded in
5621  * the other parameters which return dynamically-allocated
5622  * structures:
5623  *  ceph_opts
5624  *      The address of a pointer that will refer to a ceph options
5625  *      structure.  Caller must release the returned pointer using
5626  *      ceph_destroy_options() when it is no longer needed.
5627  *  rbd_opts
5628  *      Address of an rbd options pointer.  Fully initialized by
5629  *      this function; caller must release with kfree().
5630  *  spec
5631  *      Address of an rbd image specification pointer.  Fully
5632  *      initialized by this function based on parsed options.
5633  *      Caller must release with rbd_spec_put().
5634  *
5635  * The options passed take this form:
5636  *  <mon_addrs> <options> <pool_name> <image_name> [<snap_id>]
5637  * where:
5638  *  <mon_addrs>
5639  *      A comma-separated list of one or more monitor addresses.
5640  *      A monitor address is an ip address, optionally followed
5641  *      by a port number (separated by a colon).
5642  *        I.e.:  ip1[:port1][,ip2[:port2]...]
5643  *  <options>
5644  *      A comma-separated list of ceph and/or rbd options.
5645  *  <pool_name>
5646  *      The name of the rados pool containing the rbd image.
5647  *  <image_name>
5648  *      The name of the image in that pool to map.
5649  *  <snap_id>
5650  *      An optional snapshot id.  If provided, the mapping will
5651  *      present data from the image at the time that snapshot was
5652  *      created.  The image head is used if no snapshot id is
5653  *      provided.  Snapshot mappings are always read-only.
5654  */
5655 static int rbd_add_parse_args(const char *buf,
5656                                 struct ceph_options **ceph_opts,
5657                                 struct rbd_options **opts,
5658                                 struct rbd_spec **rbd_spec)
5659 {
5660         size_t len;
5661         char *options;
5662         const char *mon_addrs;
5663         char *snap_name;
5664         size_t mon_addrs_size;
5665         struct rbd_spec *spec = NULL;
5666         struct rbd_options *rbd_opts = NULL;
5667         struct ceph_options *copts;
5668         int ret;
5669
5670         /* The first four tokens are required */
5671
5672         len = next_token(&buf);
5673         if (!len) {
5674                 rbd_warn(NULL, "no monitor address(es) provided");
5675                 return -EINVAL;
5676         }
5677         mon_addrs = buf;
5678         mon_addrs_size = len + 1;
5679         buf += len;
5680
5681         ret = -EINVAL;
5682         options = dup_token(&buf, NULL);
5683         if (!options)
5684                 return -ENOMEM;
5685         if (!*options) {
5686                 rbd_warn(NULL, "no options provided");
5687                 goto out_err;
5688         }
5689
5690         spec = rbd_spec_alloc();
5691         if (!spec)
5692                 goto out_mem;
5693
5694         spec->pool_name = dup_token(&buf, NULL);
5695         if (!spec->pool_name)
5696                 goto out_mem;
5697         if (!*spec->pool_name) {
5698                 rbd_warn(NULL, "no pool name provided");
5699                 goto out_err;
5700         }
5701
5702         spec->image_name = dup_token(&buf, NULL);
5703         if (!spec->image_name)
5704                 goto out_mem;
5705         if (!*spec->image_name) {
5706                 rbd_warn(NULL, "no image name provided");
5707                 goto out_err;
5708         }
5709
5710         /*
5711          * Snapshot name is optional; default is to use "-"
5712          * (indicating the head/no snapshot).
5713          */
5714         len = next_token(&buf);
5715         if (!len) {
5716                 buf = RBD_SNAP_HEAD_NAME; /* No snapshot supplied */
5717                 len = sizeof (RBD_SNAP_HEAD_NAME) - 1;
5718         } else if (len > RBD_MAX_SNAP_NAME_LEN) {
5719                 ret = -ENAMETOOLONG;
5720                 goto out_err;
5721         }
5722         snap_name = kmemdup(buf, len + 1, GFP_KERNEL);
5723         if (!snap_name)
5724                 goto out_mem;
5725         *(snap_name + len) = '\0';
5726         spec->snap_name = snap_name;
5727
5728         /* Initialize all rbd options to the defaults */
5729
5730         rbd_opts = kzalloc(sizeof (*rbd_opts), GFP_KERNEL);
5731         if (!rbd_opts)
5732                 goto out_mem;
5733
5734         rbd_opts->read_only = RBD_READ_ONLY_DEFAULT;
5735         rbd_opts->queue_depth = RBD_QUEUE_DEPTH_DEFAULT;
5736         rbd_opts->lock_on_read = RBD_LOCK_ON_READ_DEFAULT;
5737
5738         copts = ceph_parse_options(options, mon_addrs,
5739                                         mon_addrs + mon_addrs_size - 1,
5740                                         parse_rbd_opts_token, rbd_opts);
5741         if (IS_ERR(copts)) {
5742                 ret = PTR_ERR(copts);
5743                 goto out_err;
5744         }
5745         kfree(options);
5746
5747         *ceph_opts = copts;
5748         *opts = rbd_opts;
5749         *rbd_spec = spec;
5750
5751         return 0;
5752 out_mem:
5753         ret = -ENOMEM;
5754 out_err:
5755         kfree(rbd_opts);
5756         rbd_spec_put(spec);
5757         kfree(options);
5758
5759         return ret;
5760 }
5761
5762 /*
5763  * Return pool id (>= 0) or a negative error code.
5764  */
5765 static int rbd_add_get_pool_id(struct rbd_client *rbdc, const char *pool_name)
5766 {
5767         struct ceph_options *opts = rbdc->client->options;
5768         u64 newest_epoch;
5769         int tries = 0;
5770         int ret;
5771
5772 again:
5773         ret = ceph_pg_poolid_by_name(rbdc->client->osdc.osdmap, pool_name);
5774         if (ret == -ENOENT && tries++ < 1) {
5775                 ret = ceph_monc_get_version(&rbdc->client->monc, "osdmap",
5776                                             &newest_epoch);
5777                 if (ret < 0)
5778                         return ret;
5779
5780                 if (rbdc->client->osdc.osdmap->epoch < newest_epoch) {
5781                         ceph_osdc_maybe_request_map(&rbdc->client->osdc);
5782                         (void) ceph_monc_wait_osdmap(&rbdc->client->monc,
5783                                                      newest_epoch,
5784                                                      opts->mount_timeout);
5785                         goto again;
5786                 } else {
5787                         /* the osdmap we have is new enough */
5788                         return -ENOENT;
5789                 }
5790         }
5791
5792         return ret;
5793 }
5794
5795 /*
5796  * An rbd format 2 image has a unique identifier, distinct from the
5797  * name given to it by the user.  Internally, that identifier is
5798  * what's used to specify the names of objects related to the image.
5799  *
5800  * A special "rbd id" object is used to map an rbd image name to its
5801  * id.  If that object doesn't exist, then there is no v2 rbd image
5802  * with the supplied name.
5803  *
5804  * This function will record the given rbd_dev's image_id field if
5805  * it can be determined, and in that case will return 0.  If any
5806  * errors occur a negative errno will be returned and the rbd_dev's
5807  * image_id field will be unchanged (and should be NULL).
5808  */
5809 static int rbd_dev_image_id(struct rbd_device *rbd_dev)
5810 {
5811         int ret;
5812         size_t size;
5813         char *object_name;
5814         void *response;
5815         char *image_id;
5816
5817         /*
5818          * When probing a parent image, the image id is already
5819          * known (and the image name likely is not).  There's no
5820          * need to fetch the image id again in this case.  We
5821          * do still need to set the image format though.
5822          */
5823         if (rbd_dev->spec->image_id) {
5824                 rbd_dev->image_format = *rbd_dev->spec->image_id ? 2 : 1;
5825
5826                 return 0;
5827         }
5828
5829         /*
5830          * First, see if the format 2 image id file exists, and if
5831          * so, get the image's persistent id from it.
5832          */
5833         size = sizeof (RBD_ID_PREFIX) + strlen(rbd_dev->spec->image_name);
5834         object_name = kmalloc(size, GFP_NOIO);
5835         if (!object_name)
5836                 return -ENOMEM;
5837         sprintf(object_name, "%s%s", RBD_ID_PREFIX, rbd_dev->spec->image_name);
5838         dout("rbd id object name is %s\n", object_name);
5839
5840         /* Response will be an encoded string, which includes a length */
5841
5842         size = sizeof (__le32) + RBD_IMAGE_ID_LEN_MAX;
5843         response = kzalloc(size, GFP_NOIO);
5844         if (!response) {
5845                 ret = -ENOMEM;
5846                 goto out;
5847         }
5848
5849         /* If it doesn't exist we'll assume it's a format 1 image */
5850
5851         ret = rbd_obj_method_sync(rbd_dev, object_name,
5852                                 "rbd", "get_id", NULL, 0,
5853                                 response, RBD_IMAGE_ID_LEN_MAX);
5854         dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
5855         if (ret == -ENOENT) {
5856                 image_id = kstrdup("", GFP_KERNEL);
5857                 ret = image_id ? 0 : -ENOMEM;
5858                 if (!ret)
5859                         rbd_dev->image_format = 1;
5860         } else if (ret >= 0) {
5861                 void *p = response;
5862
5863                 image_id = ceph_extract_encoded_string(&p, p + ret,
5864                                                 NULL, GFP_NOIO);
5865                 ret = PTR_ERR_OR_ZERO(image_id);
5866                 if (!ret)
5867                         rbd_dev->image_format = 2;
5868         }
5869
5870         if (!ret) {
5871                 rbd_dev->spec->image_id = image_id;
5872                 dout("image_id is %s\n", image_id);
5873         }
5874 out:
5875         kfree(response);
5876         kfree(object_name);
5877
5878         return ret;
5879 }
5880
5881 /*
5882  * Undo whatever state changes are made by v1 or v2 header info
5883  * call.
5884  */
5885 static void rbd_dev_unprobe(struct rbd_device *rbd_dev)
5886 {
5887         struct rbd_image_header *header;
5888
5889         rbd_dev_parent_put(rbd_dev);
5890
5891         /* Free dynamic fields from the header, then zero it out */
5892
5893         header = &rbd_dev->header;
5894         ceph_put_snap_context(header->snapc);
5895         kfree(header->snap_sizes);
5896         kfree(header->snap_names);
5897         kfree(header->object_prefix);
5898         memset(header, 0, sizeof (*header));
5899 }
5900
5901 static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev)
5902 {
5903         int ret;
5904
5905         ret = rbd_dev_v2_object_prefix(rbd_dev);
5906         if (ret)
5907                 goto out_err;
5908
5909         /*
5910          * Get the and check features for the image.  Currently the
5911          * features are assumed to never change.
5912          */
5913         ret = rbd_dev_v2_features(rbd_dev);
5914         if (ret)
5915                 goto out_err;
5916
5917         /* If the image supports fancy striping, get its parameters */
5918
5919         if (rbd_dev->header.features & RBD_FEATURE_STRIPINGV2) {
5920                 ret = rbd_dev_v2_striping_info(rbd_dev);
5921                 if (ret < 0)
5922                         goto out_err;
5923         }
5924         /* No support for crypto and compression type format 2 images */
5925
5926         return 0;
5927 out_err:
5928         rbd_dev->header.features = 0;
5929         kfree(rbd_dev->header.object_prefix);
5930         rbd_dev->header.object_prefix = NULL;
5931
5932         return ret;
5933 }
5934
5935 /*
5936  * @depth is rbd_dev_image_probe() -> rbd_dev_probe_parent() ->
5937  * rbd_dev_image_probe() recursion depth, which means it's also the
5938  * length of the already discovered part of the parent chain.
5939  */
5940 static int rbd_dev_probe_parent(struct rbd_device *rbd_dev, int depth)
5941 {
5942         struct rbd_device *parent = NULL;
5943         int ret;
5944
5945         if (!rbd_dev->parent_spec)
5946                 return 0;
5947
5948         if (++depth > RBD_MAX_PARENT_CHAIN_LEN) {
5949                 pr_info("parent chain is too long (%d)\n", depth);
5950                 ret = -EINVAL;
5951                 goto out_err;
5952         }
5953
5954         parent = __rbd_dev_create(rbd_dev->rbd_client, rbd_dev->parent_spec);
5955         if (!parent) {
5956                 ret = -ENOMEM;
5957                 goto out_err;
5958         }
5959
5960         /*
5961          * Images related by parent/child relationships always share
5962          * rbd_client and spec/parent_spec, so bump their refcounts.
5963          */
5964         __rbd_get_client(rbd_dev->rbd_client);
5965         rbd_spec_get(rbd_dev->parent_spec);
5966
5967         ret = rbd_dev_image_probe(parent, depth);
5968         if (ret < 0)
5969                 goto out_err;
5970
5971         rbd_dev->parent = parent;
5972         atomic_set(&rbd_dev->parent_ref, 1);
5973         return 0;
5974
5975 out_err:
5976         rbd_dev_unparent(rbd_dev);
5977         rbd_dev_destroy(parent);
5978         return ret;
5979 }
5980
5981 /*
5982  * rbd_dev->header_rwsem must be locked for write and will be unlocked
5983  * upon return.
5984  */
5985 static int rbd_dev_device_setup(struct rbd_device *rbd_dev)
5986 {
5987         int ret;
5988
5989         /* Record our major and minor device numbers. */
5990
5991         if (!single_major) {
5992                 ret = register_blkdev(0, rbd_dev->name);
5993                 if (ret < 0)
5994                         goto err_out_unlock;
5995
5996                 rbd_dev->major = ret;
5997                 rbd_dev->minor = 0;
5998         } else {
5999                 rbd_dev->major = rbd_major;
6000                 rbd_dev->minor = rbd_dev_id_to_minor(rbd_dev->dev_id);
6001         }
6002
6003         /* Set up the blkdev mapping. */
6004
6005         ret = rbd_init_disk(rbd_dev);
6006         if (ret)
6007                 goto err_out_blkdev;
6008
6009         ret = rbd_dev_mapping_set(rbd_dev);
6010         if (ret)
6011                 goto err_out_disk;
6012
6013         set_capacity(rbd_dev->disk, rbd_dev->mapping.size / SECTOR_SIZE);
6014         set_disk_ro(rbd_dev->disk, rbd_dev->mapping.read_only);
6015
6016         dev_set_name(&rbd_dev->dev, "%d", rbd_dev->dev_id);
6017         ret = device_add(&rbd_dev->dev);
6018         if (ret)
6019                 goto err_out_mapping;
6020
6021         /* Everything's ready.  Announce the disk to the world. */
6022
6023         set_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
6024         up_write(&rbd_dev->header_rwsem);
6025
6026         spin_lock(&rbd_dev_list_lock);
6027         list_add_tail(&rbd_dev->node, &rbd_dev_list);
6028         spin_unlock(&rbd_dev_list_lock);
6029
6030         add_disk(rbd_dev->disk);
6031         pr_info("%s: capacity %llu features 0x%llx\n", rbd_dev->disk->disk_name,
6032                 (unsigned long long)get_capacity(rbd_dev->disk) << SECTOR_SHIFT,
6033                 rbd_dev->header.features);
6034
6035         return ret;
6036
6037 err_out_mapping:
6038         rbd_dev_mapping_clear(rbd_dev);
6039 err_out_disk:
6040         rbd_free_disk(rbd_dev);
6041 err_out_blkdev:
6042         if (!single_major)
6043                 unregister_blkdev(rbd_dev->major, rbd_dev->name);
6044 err_out_unlock:
6045         up_write(&rbd_dev->header_rwsem);
6046         return ret;
6047 }
6048
6049 static int rbd_dev_header_name(struct rbd_device *rbd_dev)
6050 {
6051         struct rbd_spec *spec = rbd_dev->spec;
6052         int ret;
6053
6054         /* Record the header object name for this rbd image. */
6055
6056         rbd_assert(rbd_image_format_valid(rbd_dev->image_format));
6057
6058         rbd_dev->header_oloc.pool = rbd_dev->layout.pool_id;
6059         if (rbd_dev->image_format == 1)
6060                 ret = ceph_oid_aprintf(&rbd_dev->header_oid, GFP_KERNEL, "%s%s",
6061                                        spec->image_name, RBD_SUFFIX);
6062         else
6063                 ret = ceph_oid_aprintf(&rbd_dev->header_oid, GFP_KERNEL, "%s%s",
6064                                        RBD_HEADER_PREFIX, spec->image_id);
6065
6066         return ret;
6067 }
6068
6069 static void rbd_dev_image_release(struct rbd_device *rbd_dev)
6070 {
6071         rbd_dev_unprobe(rbd_dev);
6072         rbd_dev->image_format = 0;
6073         kfree(rbd_dev->spec->image_id);
6074         rbd_dev->spec->image_id = NULL;
6075
6076         rbd_dev_destroy(rbd_dev);
6077 }
6078
6079 /*
6080  * Probe for the existence of the header object for the given rbd
6081  * device.  If this image is the one being mapped (i.e., not a
6082  * parent), initiate a watch on its header object before using that
6083  * object to get detailed information about the rbd image.
6084  */
6085 static int rbd_dev_image_probe(struct rbd_device *rbd_dev, int depth)
6086 {
6087         int ret;
6088
6089         /*
6090          * Get the id from the image id object.  Unless there's an
6091          * error, rbd_dev->spec->image_id will be filled in with
6092          * a dynamically-allocated string, and rbd_dev->image_format
6093          * will be set to either 1 or 2.
6094          */
6095         ret = rbd_dev_image_id(rbd_dev);
6096         if (ret)
6097                 return ret;
6098
6099         ret = rbd_dev_header_name(rbd_dev);
6100         if (ret)
6101                 goto err_out_format;
6102
6103         if (!depth) {
6104                 ret = rbd_register_watch(rbd_dev);
6105                 if (ret) {
6106                         if (ret == -ENOENT)
6107                                 pr_info("image %s/%s does not exist\n",
6108                                         rbd_dev->spec->pool_name,
6109                                         rbd_dev->spec->image_name);
6110                         goto err_out_format;
6111                 }
6112         }
6113
6114         ret = rbd_dev_header_info(rbd_dev);
6115         if (ret)
6116                 goto err_out_watch;
6117
6118         /*
6119          * If this image is the one being mapped, we have pool name and
6120          * id, image name and id, and snap name - need to fill snap id.
6121          * Otherwise this is a parent image, identified by pool, image
6122          * and snap ids - need to fill in names for those ids.
6123          */
6124         if (!depth)
6125                 ret = rbd_spec_fill_snap_id(rbd_dev);
6126         else
6127                 ret = rbd_spec_fill_names(rbd_dev);
6128         if (ret) {
6129                 if (ret == -ENOENT)
6130                         pr_info("snap %s/%s@%s does not exist\n",
6131                                 rbd_dev->spec->pool_name,
6132                                 rbd_dev->spec->image_name,
6133                                 rbd_dev->spec->snap_name);
6134                 goto err_out_probe;
6135         }
6136
6137         if (rbd_dev->header.features & RBD_FEATURE_LAYERING) {
6138                 ret = rbd_dev_v2_parent_info(rbd_dev);
6139                 if (ret)
6140                         goto err_out_probe;
6141
6142                 /*
6143                  * Need to warn users if this image is the one being
6144                  * mapped and has a parent.
6145                  */
6146                 if (!depth && rbd_dev->parent_spec)
6147                         rbd_warn(rbd_dev,
6148                                  "WARNING: kernel layering is EXPERIMENTAL!");
6149         }
6150
6151         ret = rbd_dev_probe_parent(rbd_dev, depth);
6152         if (ret)
6153                 goto err_out_probe;
6154
6155         dout("discovered format %u image, header name is %s\n",
6156                 rbd_dev->image_format, rbd_dev->header_oid.name);
6157         return 0;
6158
6159 err_out_probe:
6160         rbd_dev_unprobe(rbd_dev);
6161 err_out_watch:
6162         if (!depth)
6163                 rbd_unregister_watch(rbd_dev);
6164 err_out_format:
6165         rbd_dev->image_format = 0;
6166         kfree(rbd_dev->spec->image_id);
6167         rbd_dev->spec->image_id = NULL;
6168         return ret;
6169 }
6170
6171 static ssize_t do_rbd_add(struct bus_type *bus,
6172                           const char *buf,
6173                           size_t count)
6174 {
6175         struct rbd_device *rbd_dev = NULL;
6176         struct ceph_options *ceph_opts = NULL;
6177         struct rbd_options *rbd_opts = NULL;
6178         struct rbd_spec *spec = NULL;
6179         struct rbd_client *rbdc;
6180         bool read_only;
6181         int rc;
6182
6183         if (!try_module_get(THIS_MODULE))
6184                 return -ENODEV;
6185
6186         /* parse add command */
6187         rc = rbd_add_parse_args(buf, &ceph_opts, &rbd_opts, &spec);
6188         if (rc < 0)
6189                 goto out;
6190
6191         rbdc = rbd_get_client(ceph_opts);
6192         if (IS_ERR(rbdc)) {
6193                 rc = PTR_ERR(rbdc);
6194                 goto err_out_args;
6195         }
6196
6197         /* pick the pool */
6198         rc = rbd_add_get_pool_id(rbdc, spec->pool_name);
6199         if (rc < 0) {
6200                 if (rc == -ENOENT)
6201                         pr_info("pool %s does not exist\n", spec->pool_name);
6202                 goto err_out_client;
6203         }
6204         spec->pool_id = (u64)rc;
6205
6206         rbd_dev = rbd_dev_create(rbdc, spec, rbd_opts);
6207         if (!rbd_dev) {
6208                 rc = -ENOMEM;
6209                 goto err_out_client;
6210         }
6211         rbdc = NULL;            /* rbd_dev now owns this */
6212         spec = NULL;            /* rbd_dev now owns this */
6213         rbd_opts = NULL;        /* rbd_dev now owns this */
6214
6215         rbd_dev->config_info = kstrdup(buf, GFP_KERNEL);
6216         if (!rbd_dev->config_info) {
6217                 rc = -ENOMEM;
6218                 goto err_out_rbd_dev;
6219         }
6220
6221         down_write(&rbd_dev->header_rwsem);
6222         rc = rbd_dev_image_probe(rbd_dev, 0);
6223         if (rc < 0) {
6224                 up_write(&rbd_dev->header_rwsem);
6225                 goto err_out_rbd_dev;
6226         }
6227
6228         /* If we are mapping a snapshot it must be marked read-only */
6229
6230         read_only = rbd_dev->opts->read_only;
6231         if (rbd_dev->spec->snap_id != CEPH_NOSNAP)
6232                 read_only = true;
6233         rbd_dev->mapping.read_only = read_only;
6234
6235         rc = rbd_dev_device_setup(rbd_dev);
6236         if (rc) {
6237                 /*
6238                  * rbd_unregister_watch() can't be moved into
6239                  * rbd_dev_image_release() without refactoring, see
6240                  * commit 1f3ef78861ac.
6241                  */
6242                 rbd_unregister_watch(rbd_dev);
6243                 rbd_dev_image_release(rbd_dev);
6244                 goto out;
6245         }
6246
6247         rc = count;
6248 out:
6249         module_put(THIS_MODULE);
6250         return rc;
6251
6252 err_out_rbd_dev:
6253         rbd_dev_destroy(rbd_dev);
6254 err_out_client:
6255         rbd_put_client(rbdc);
6256 err_out_args:
6257         rbd_spec_put(spec);
6258         kfree(rbd_opts);
6259         goto out;
6260 }
6261
6262 static ssize_t rbd_add(struct bus_type *bus,
6263                        const char *buf,
6264                        size_t count)
6265 {
6266         if (single_major)
6267                 return -EINVAL;
6268
6269         return do_rbd_add(bus, buf, count);
6270 }
6271
6272 static ssize_t rbd_add_single_major(struct bus_type *bus,
6273                                     const char *buf,
6274                                     size_t count)
6275 {
6276         return do_rbd_add(bus, buf, count);
6277 }
6278
6279 static void rbd_dev_device_release(struct rbd_device *rbd_dev)
6280 {
6281         rbd_free_disk(rbd_dev);
6282
6283         spin_lock(&rbd_dev_list_lock);
6284         list_del_init(&rbd_dev->node);
6285         spin_unlock(&rbd_dev_list_lock);
6286
6287         clear_bit(RBD_DEV_FLAG_EXISTS, &rbd_dev->flags);
6288         device_del(&rbd_dev->dev);
6289         rbd_dev_mapping_clear(rbd_dev);
6290         if (!single_major)
6291                 unregister_blkdev(rbd_dev->major, rbd_dev->name);
6292 }
6293
6294 static void rbd_dev_remove_parent(struct rbd_device *rbd_dev)
6295 {
6296         while (rbd_dev->parent) {
6297                 struct rbd_device *first = rbd_dev;
6298                 struct rbd_device *second = first->parent;
6299                 struct rbd_device *third;
6300
6301                 /*
6302                  * Follow to the parent with no grandparent and
6303                  * remove it.
6304                  */
6305                 while (second && (third = second->parent)) {
6306                         first = second;
6307                         second = third;
6308                 }
6309                 rbd_assert(second);
6310                 rbd_dev_image_release(second);
6311                 first->parent = NULL;
6312                 first->parent_overlap = 0;
6313
6314                 rbd_assert(first->parent_spec);
6315                 rbd_spec_put(first->parent_spec);
6316                 first->parent_spec = NULL;
6317         }
6318 }
6319
6320 static ssize_t do_rbd_remove(struct bus_type *bus,
6321                              const char *buf,
6322                              size_t count)
6323 {
6324         struct rbd_device *rbd_dev = NULL;
6325         struct list_head *tmp;
6326         int dev_id;
6327         char opt_buf[6];
6328         bool already = false;
6329         bool force = false;
6330         int ret;
6331
6332         dev_id = -1;
6333         opt_buf[0] = '\0';
6334         sscanf(buf, "%d %5s", &dev_id, opt_buf);
6335         if (dev_id < 0) {
6336                 pr_err("dev_id out of range\n");
6337                 return -EINVAL;
6338         }
6339         if (opt_buf[0] != '\0') {
6340                 if (!strcmp(opt_buf, "force")) {
6341                         force = true;
6342                 } else {
6343                         pr_err("bad remove option at '%s'\n", opt_buf);
6344                         return -EINVAL;
6345                 }
6346         }
6347
6348         ret = -ENOENT;
6349         spin_lock(&rbd_dev_list_lock);
6350         list_for_each(tmp, &rbd_dev_list) {
6351                 rbd_dev = list_entry(tmp, struct rbd_device, node);
6352                 if (rbd_dev->dev_id == dev_id) {
6353                         ret = 0;
6354                         break;
6355                 }
6356         }
6357         if (!ret) {
6358                 spin_lock_irq(&rbd_dev->lock);
6359                 if (rbd_dev->open_count && !force)
6360                         ret = -EBUSY;
6361                 else
6362                         already = test_and_set_bit(RBD_DEV_FLAG_REMOVING,
6363                                                         &rbd_dev->flags);
6364                 spin_unlock_irq(&rbd_dev->lock);
6365         }
6366         spin_unlock(&rbd_dev_list_lock);
6367         if (ret < 0 || already)
6368                 return ret;
6369
6370         if (force) {
6371                 /*
6372                  * Prevent new IO from being queued and wait for existing
6373                  * IO to complete/fail.
6374                  */
6375                 blk_mq_freeze_queue(rbd_dev->disk->queue);
6376                 blk_set_queue_dying(rbd_dev->disk->queue);
6377         }
6378
6379         down_write(&rbd_dev->lock_rwsem);
6380         if (__rbd_is_lock_owner(rbd_dev))
6381                 rbd_unlock(rbd_dev);
6382         up_write(&rbd_dev->lock_rwsem);
6383         rbd_unregister_watch(rbd_dev);
6384
6385         /*
6386          * Don't free anything from rbd_dev->disk until after all
6387          * notifies are completely processed. Otherwise
6388          * rbd_bus_del_dev() will race with rbd_watch_cb(), resulting
6389          * in a potential use after free of rbd_dev->disk or rbd_dev.
6390          */
6391         rbd_dev_device_release(rbd_dev);
6392         rbd_dev_image_release(rbd_dev);
6393
6394         return count;
6395 }
6396
6397 static ssize_t rbd_remove(struct bus_type *bus,
6398                           const char *buf,
6399                           size_t count)
6400 {
6401         if (single_major)
6402                 return -EINVAL;
6403
6404         return do_rbd_remove(bus, buf, count);
6405 }
6406
6407 static ssize_t rbd_remove_single_major(struct bus_type *bus,
6408                                        const char *buf,
6409                                        size_t count)
6410 {
6411         return do_rbd_remove(bus, buf, count);
6412 }
6413
6414 /*
6415  * create control files in sysfs
6416  * /sys/bus/rbd/...
6417  */
6418 static int rbd_sysfs_init(void)
6419 {
6420         int ret;
6421
6422         ret = device_register(&rbd_root_dev);
6423         if (ret < 0)
6424                 return ret;
6425
6426         ret = bus_register(&rbd_bus_type);
6427         if (ret < 0)
6428                 device_unregister(&rbd_root_dev);
6429
6430         return ret;
6431 }
6432
6433 static void rbd_sysfs_cleanup(void)
6434 {
6435         bus_unregister(&rbd_bus_type);
6436         device_unregister(&rbd_root_dev);
6437 }
6438
6439 static int rbd_slab_init(void)
6440 {
6441         rbd_assert(!rbd_img_request_cache);
6442         rbd_img_request_cache = KMEM_CACHE(rbd_img_request, 0);
6443         if (!rbd_img_request_cache)
6444                 return -ENOMEM;
6445
6446         rbd_assert(!rbd_obj_request_cache);
6447         rbd_obj_request_cache = KMEM_CACHE(rbd_obj_request, 0);
6448         if (!rbd_obj_request_cache)
6449                 goto out_err;
6450
6451         rbd_assert(!rbd_segment_name_cache);
6452         rbd_segment_name_cache = kmem_cache_create("rbd_segment_name",
6453                                         CEPH_MAX_OID_NAME_LEN + 1, 1, 0, NULL);
6454         if (rbd_segment_name_cache)
6455                 return 0;
6456 out_err:
6457         kmem_cache_destroy(rbd_obj_request_cache);
6458         rbd_obj_request_cache = NULL;
6459
6460         kmem_cache_destroy(rbd_img_request_cache);
6461         rbd_img_request_cache = NULL;
6462
6463         return -ENOMEM;
6464 }
6465
6466 static void rbd_slab_exit(void)
6467 {
6468         rbd_assert(rbd_segment_name_cache);
6469         kmem_cache_destroy(rbd_segment_name_cache);
6470         rbd_segment_name_cache = NULL;
6471
6472         rbd_assert(rbd_obj_request_cache);
6473         kmem_cache_destroy(rbd_obj_request_cache);
6474         rbd_obj_request_cache = NULL;
6475
6476         rbd_assert(rbd_img_request_cache);
6477         kmem_cache_destroy(rbd_img_request_cache);
6478         rbd_img_request_cache = NULL;
6479 }
6480
6481 static int __init rbd_init(void)
6482 {
6483         int rc;
6484
6485         if (!libceph_compatible(NULL)) {
6486                 rbd_warn(NULL, "libceph incompatibility (quitting)");
6487                 return -EINVAL;
6488         }
6489
6490         rc = rbd_slab_init();
6491         if (rc)
6492                 return rc;
6493
6494         /*
6495          * The number of active work items is limited by the number of
6496          * rbd devices * queue depth, so leave @max_active at default.
6497          */
6498         rbd_wq = alloc_workqueue(RBD_DRV_NAME, WQ_MEM_RECLAIM, 0);
6499         if (!rbd_wq) {
6500                 rc = -ENOMEM;
6501                 goto err_out_slab;
6502         }
6503
6504         if (single_major) {
6505                 rbd_major = register_blkdev(0, RBD_DRV_NAME);
6506                 if (rbd_major < 0) {
6507                         rc = rbd_major;
6508                         goto err_out_wq;
6509                 }
6510         }
6511
6512         rc = rbd_sysfs_init();
6513         if (rc)
6514                 goto err_out_blkdev;
6515
6516         if (single_major)
6517                 pr_info("loaded (major %d)\n", rbd_major);
6518         else
6519                 pr_info("loaded\n");
6520
6521         return 0;
6522
6523 err_out_blkdev:
6524         if (single_major)
6525                 unregister_blkdev(rbd_major, RBD_DRV_NAME);
6526 err_out_wq:
6527         destroy_workqueue(rbd_wq);
6528 err_out_slab:
6529         rbd_slab_exit();
6530         return rc;
6531 }
6532
6533 static void __exit rbd_exit(void)
6534 {
6535         ida_destroy(&rbd_dev_id_ida);
6536         rbd_sysfs_cleanup();
6537         if (single_major)
6538                 unregister_blkdev(rbd_major, RBD_DRV_NAME);
6539         destroy_workqueue(rbd_wq);
6540         rbd_slab_exit();
6541 }
6542
6543 module_init(rbd_init);
6544 module_exit(rbd_exit);
6545
6546 MODULE_AUTHOR("Alex Elder <elder@inktank.com>");
6547 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
6548 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
6549 /* following authorship retained from original osdblk.c */
6550 MODULE_AUTHOR("Jeff Garzik <jeff@garzik.org>");
6551
6552 MODULE_DESCRIPTION("RADOS Block Device (RBD) driver");
6553 MODULE_LICENSE("GPL");