include/linux/ceph/ceph_fs.h

   1 /*
   2  * ceph_fs.h - Ceph constants and data types to share between kernel and
   3  * user space.
   4  *
   5  * Most types in this file are defined as little-endian, and are
   6  * primarily intended to describe data structures that pass over the
   7  * wire or that are stored on disk.
   8  *
   9  * LGPL2
  10  */
  11
  12 #ifndef CEPH_FS_H
  13 #define CEPH_FS_H
  14
  15 #include <linux/ceph/msgr.h>
  16 #include <linux/ceph/rados.h>
  17
  18 /*
  19  * subprotocol versions.  when specific messages types or high-level
  20  * protocols change, bump the affected components.  we keep rev
  21  * internal cluster protocols separately from the public,
  22  * client-facing protocol.
  23  */
  24 #define CEPH_OSDC_PROTOCOL   24 /* server/client */
  25 #define CEPH_MDSC_PROTOCOL   32 /* server/client */
  26 #define CEPH_MONC_PROTOCOL   15 /* server/client */
  27
  28
  29 #define CEPH_INO_ROOT   1
  30 #define CEPH_INO_CEPH   2       /* hidden .ceph dir */
  31 #define CEPH_INO_DOTDOT 3       /* used by ceph fuse for parent (..) */
  32
  33 /* arbitrary limit on max # of monitors (cluster of 3 is typical) */
  34 #define CEPH_MAX_MON   31
  35
  36 /*
  37  * legacy ceph_file_layoute
  38  */
  39 struct ceph_file_layout_legacy {
  40         /* file -> object mapping */
  41         __le32 fl_stripe_unit;     /* stripe unit, in bytes.  must be multiple
  42                                       of page size. */
  43         __le32 fl_stripe_count;    /* over this many objects */
  44         __le32 fl_object_size;     /* until objects are this big, then move to
  45                                       new objects */
  46         __le32 fl_cas_hash;        /* UNUSED.  0 = none; 1 = sha256 */
  47
  48         /* pg -> disk layout */
  49         __le32 fl_object_stripe_unit;  /* UNUSED.  for per-object parity, if any */
  50
  51         /* object -> pg layout */
  52         __le32 fl_unused;       /* unused; used to be preferred primary for pg (-1 for none) */
  53         __le32 fl_pg_pool;      /* namespace, crush ruleset, rep level */
  54 } __attribute__ ((packed));
  55
  56 struct ceph_string;
  57 /*
  58  * ceph_file_layout - describe data layout for a file/inode
  59  */
  60 struct ceph_file_layout {
  61         /* file -> object mapping */
  62         u32 stripe_unit;   /* stripe unit, in bytes */
  63         u32 stripe_count;  /* over this many objects */
  64         u32 object_size;   /* until objects are this big */
  65         s64 pool_id;        /* rados pool id */
  66         struct ceph_string __rcu *pool_ns; /* rados pool namespace */
  67 };
  68
  69 extern int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
  70 extern void ceph_file_layout_from_legacy(struct ceph_file_layout *fl,
  71                                 struct ceph_file_layout_legacy *legacy);
  72 extern void ceph_file_layout_to_legacy(struct ceph_file_layout *fl,
  73                                 struct ceph_file_layout_legacy *legacy);
  74
  75 #define CEPH_MIN_STRIPE_UNIT 65536
  76
  77 struct ceph_dir_layout {
  78         __u8   dl_dir_hash;   /* see ceph_hash.h for ids */
  79         __u8   dl_unused1;
  80         __u16  dl_unused2;
  81         __u32  dl_unused3;
  82 } __attribute__ ((packed));
  83
  84 /* crypto algorithms */
  85 #define CEPH_CRYPTO_NONE 0x0
  86 #define CEPH_CRYPTO_AES  0x1
  87
  88 #define CEPH_AES_IV "cephsageyudagreg"
  89
  90 /* security/authentication protocols */
  91 #define CEPH_AUTH_UNKNOWN       0x0
  92 #define CEPH_AUTH_NONE          0x1
  93 #define CEPH_AUTH_CEPHX         0x2
  94
  95 #define CEPH_AUTH_UID_DEFAULT ((__u64) -1)
  96
  97
  98 /*********************************************
  99  * message layer
 100  */
 101
 102 /*
 103  * message types
 104  */
 105
 106 /* misc */
 107 #define CEPH_MSG_SHUTDOWN               1
 108 #define CEPH_MSG_PING                   2
 109
 110 /* client <-> monitor */
 111 #define CEPH_MSG_MON_MAP                4
 112 #define CEPH_MSG_MON_GET_MAP            5
 113 #define CEPH_MSG_STATFS                 13
 114 #define CEPH_MSG_STATFS_REPLY           14
 115 #define CEPH_MSG_MON_SUBSCRIBE          15
 116 #define CEPH_MSG_MON_SUBSCRIBE_ACK      16
 117 #define CEPH_MSG_AUTH                   17
 118 #define CEPH_MSG_AUTH_REPLY             18
 119 #define CEPH_MSG_MON_GET_VERSION        19
 120 #define CEPH_MSG_MON_GET_VERSION_REPLY  20
 121
 122 /* client <-> mds */
 123 #define CEPH_MSG_MDS_MAP                21
 124 #define CEPH_MSG_FS_MAP_USER            103
 125
 126 #define CEPH_MSG_CLIENT_SESSION         22
 127 #define CEPH_MSG_CLIENT_RECONNECT       23
 128
 129 #define CEPH_MSG_CLIENT_REQUEST         24
 130 #define CEPH_MSG_CLIENT_REQUEST_FORWARD 25
 131 #define CEPH_MSG_CLIENT_REPLY           26
 132 #define CEPH_MSG_CLIENT_CAPS            0x310
 133 #define CEPH_MSG_CLIENT_LEASE           0x311
 134 #define CEPH_MSG_CLIENT_SNAP            0x312
 135 #define CEPH_MSG_CLIENT_CAPRELEASE      0x313
 136
 137 /* pool ops */
 138 #define CEPH_MSG_POOLOP_REPLY           48
 139 #define CEPH_MSG_POOLOP                 49
 140
 141
 142 /* osd */
 143 #define CEPH_MSG_OSD_MAP                41
 144 #define CEPH_MSG_OSD_OP                 42
 145 #define CEPH_MSG_OSD_OPREPLY            43
 146 #define CEPH_MSG_WATCH_NOTIFY           44
 147
 148
 149 /* watch-notify operations */
 150 enum {
 151         CEPH_WATCH_EVENT_NOTIFY           = 1, /* notifying watcher */
 152         CEPH_WATCH_EVENT_NOTIFY_COMPLETE  = 2, /* notifier notified when done */
 153         CEPH_WATCH_EVENT_DISCONNECT       = 3, /* we were disconnected */
 154 };
 155
 156
 157 struct ceph_mon_request_header {
 158         __le64 have_version;
 159         __le16 session_mon;
 160         __le64 session_mon_tid;
 161 } __attribute__ ((packed));
 162
 163 struct ceph_mon_statfs {
 164         struct ceph_mon_request_header monhdr;
 165         struct ceph_fsid fsid;
 166 } __attribute__ ((packed));
 167
 168 struct ceph_statfs {
 169         __le64 kb, kb_used, kb_avail;
 170         __le64 num_objects;
 171 } __attribute__ ((packed));
 172
 173 struct ceph_mon_statfs_reply {
 174         struct ceph_fsid fsid;
 175         __le64 version;
 176         struct ceph_statfs st;
 177 } __attribute__ ((packed));
 178
 179 struct ceph_osd_getmap {
 180         struct ceph_mon_request_header monhdr;
 181         struct ceph_fsid fsid;
 182         __le32 start;
 183 } __attribute__ ((packed));
 184
 185 struct ceph_mds_getmap {
 186         struct ceph_mon_request_header monhdr;
 187         struct ceph_fsid fsid;
 188 } __attribute__ ((packed));
 189
 190 struct ceph_client_mount {
 191         struct ceph_mon_request_header monhdr;
 192 } __attribute__ ((packed));
 193
 194 #define CEPH_SUBSCRIBE_ONETIME    1  /* i want only 1 update after have */
 195
 196 struct ceph_mon_subscribe_item {
 197         __le64 start;
 198         __u8 flags;
 199 } __attribute__ ((packed));
 200
 201 struct ceph_mon_subscribe_ack {
 202         __le32 duration;         /* seconds */
 203         struct ceph_fsid fsid;
 204 } __attribute__ ((packed));
 205
 206 #define CEPH_FS_CLUSTER_ID_NONE  -1
 207
 208 /*
 209  * mdsmap flags
 210  */
 211 #define CEPH_MDSMAP_DOWN    (1<<0)  /* cluster deliberately down */
 212
 213 /*
 214  * mds states
 215  *   > 0 -> in
 216  *  <= 0 -> out
 217  */
 218 #define CEPH_MDS_STATE_DNE          0  /* down, does not exist. */
 219 #define CEPH_MDS_STATE_STOPPED     -1  /* down, once existed, but no subtrees.
 220                                           empty log. */
 221 #define CEPH_MDS_STATE_BOOT        -4  /* up, boot announcement. */
 222 #define CEPH_MDS_STATE_STANDBY     -5  /* up, idle.  waiting for assignment. */
 223 #define CEPH_MDS_STATE_CREATING    -6  /* up, creating MDS instance. */
 224 #define CEPH_MDS_STATE_STARTING    -7  /* up, starting previously stopped mds */
 225 #define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
 226 #define CEPH_MDS_STATE_REPLAYONCE   -9 /* up, replaying an active node's journal */
 227
 228 #define CEPH_MDS_STATE_REPLAY       8  /* up, replaying journal. */
 229 #define CEPH_MDS_STATE_RESOLVE      9  /* up, disambiguating distributed
 230                                           operations (import, rename, etc.) */
 231 #define CEPH_MDS_STATE_RECONNECT    10 /* up, reconnect to clients */
 232 #define CEPH_MDS_STATE_REJOIN       11 /* up, rejoining distributed cache */
 233 #define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */
 234 #define CEPH_MDS_STATE_ACTIVE       13 /* up, active */
 235 #define CEPH_MDS_STATE_STOPPING     14 /* up, but exporting metadata */
 236
 237 extern const char *ceph_mds_state_name(int s);
 238
 239
 240 /*
 241  * metadata lock types.
 242  *  - these are bitmasks.. we can compose them
 243  *  - they also define the lock ordering by the MDS
 244  *  - a few of these are internal to the mds
 245  */
 246 #define CEPH_LOCK_DVERSION    1
 247 #define CEPH_LOCK_DN          2
 248 #define CEPH_LOCK_ISNAP       16
 249 #define CEPH_LOCK_IVERSION    32    /* mds internal */
 250 #define CEPH_LOCK_IFILE       64
 251 #define CEPH_LOCK_IAUTH       128
 252 #define CEPH_LOCK_ILINK       256
 253 #define CEPH_LOCK_IDFT        512   /* dir frag tree */
 254 #define CEPH_LOCK_INEST       1024  /* mds internal */
 255 #define CEPH_LOCK_IXATTR      2048
 256 #define CEPH_LOCK_IFLOCK      4096  /* advisory file locks */
 257 #define CEPH_LOCK_INO         8192  /* immutable inode bits; not a lock */
 258 #define CEPH_LOCK_IPOLICY     16384 /* policy lock on dirs. MDS internal */
 259
 260 /* client_session ops */
 261 enum {
 262         CEPH_SESSION_REQUEST_OPEN,
 263         CEPH_SESSION_OPEN,
 264         CEPH_SESSION_REQUEST_CLOSE,
 265         CEPH_SESSION_CLOSE,
 266         CEPH_SESSION_REQUEST_RENEWCAPS,
 267         CEPH_SESSION_RENEWCAPS,
 268         CEPH_SESSION_STALE,
 269         CEPH_SESSION_RECALL_STATE,
 270         CEPH_SESSION_FLUSHMSG,
 271         CEPH_SESSION_FLUSHMSG_ACK,
 272         CEPH_SESSION_FORCE_RO,
 273 };
 274
 275 extern const char *ceph_session_op_name(int op);
 276
 277 struct ceph_mds_session_head {
 278         __le32 op;
 279         __le64 seq;
 280         struct ceph_timespec stamp;
 281         __le32 max_caps, max_leases;
 282 } __attribute__ ((packed));
 283
 284 /* client_request */
 285 /*
 286  * metadata ops.
 287  *  & 0x001000 -> write op
 288  *  & 0x010000 -> follow symlink (e.g. stat(), not lstat()).
 289  &  & 0x100000 -> use weird ino/path trace
 290  */
 291 #define CEPH_MDS_OP_WRITE        0x001000
 292 enum {
 293         CEPH_MDS_OP_LOOKUP     = 0x00100,
 294         CEPH_MDS_OP_GETATTR    = 0x00101,
 295         CEPH_MDS_OP_LOOKUPHASH = 0x00102,
 296         CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
 297         CEPH_MDS_OP_LOOKUPINO  = 0x00104,
 298         CEPH_MDS_OP_LOOKUPNAME = 0x00105,
 299
 300         CEPH_MDS_OP_SETXATTR   = 0x01105,
 301         CEPH_MDS_OP_RMXATTR    = 0x01106,
 302         CEPH_MDS_OP_SETLAYOUT  = 0x01107,
 303         CEPH_MDS_OP_SETATTR    = 0x01108,
 304         CEPH_MDS_OP_SETFILELOCK= 0x01109,
 305         CEPH_MDS_OP_GETFILELOCK= 0x00110,
 306         CEPH_MDS_OP_SETDIRLAYOUT=0x0110a,
 307
 308         CEPH_MDS_OP_MKNOD      = 0x01201,
 309         CEPH_MDS_OP_LINK       = 0x01202,
 310         CEPH_MDS_OP_UNLINK     = 0x01203,
 311         CEPH_MDS_OP_RENAME     = 0x01204,
 312         CEPH_MDS_OP_MKDIR      = 0x01220,
 313         CEPH_MDS_OP_RMDIR      = 0x01221,
 314         CEPH_MDS_OP_SYMLINK    = 0x01222,
 315
 316         CEPH_MDS_OP_CREATE     = 0x01301,
 317         CEPH_MDS_OP_OPEN       = 0x00302,
 318         CEPH_MDS_OP_READDIR    = 0x00305,
 319
 320         CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
 321         CEPH_MDS_OP_MKSNAP     = 0x01400,
 322         CEPH_MDS_OP_RMSNAP     = 0x01401,
 323         CEPH_MDS_OP_LSSNAP     = 0x00402,
 324         CEPH_MDS_OP_RENAMESNAP = 0x01403,
 325 };
 326
 327 extern const char *ceph_mds_op_name(int op);
 328
 329
 330 #define CEPH_SETATTR_MODE   1
 331 #define CEPH_SETATTR_UID    2
 332 #define CEPH_SETATTR_GID    4
 333 #define CEPH_SETATTR_MTIME  8
 334 #define CEPH_SETATTR_ATIME 16
 335 #define CEPH_SETATTR_SIZE  32
 336 #define CEPH_SETATTR_CTIME 64
 337
 338 /*
 339  * Ceph setxattr request flags.
 340  */
 341 #define CEPH_XATTR_CREATE  (1 << 0)
 342 #define CEPH_XATTR_REPLACE (1 << 1)
 343 #define CEPH_XATTR_REMOVE  (1 << 31)
 344
 345 /*
 346  * readdir request flags;
 347  */
 348 #define CEPH_READDIR_REPLY_BITFLAGS     (1<<0)
 349
 350 /*
 351  * readdir reply flags.
 352  */
 353 #define CEPH_READDIR_FRAG_END           (1<<0)
 354 #define CEPH_READDIR_FRAG_COMPLETE      (1<<8)
 355 #define CEPH_READDIR_HASH_ORDER         (1<<9)
 356
 357 union ceph_mds_request_args {
 358         struct {
 359                 __le32 mask;                 /* CEPH_CAP_* */
 360         } __attribute__ ((packed)) getattr;
 361         struct {
 362                 __le32 mode;
 363                 __le32 uid;
 364                 __le32 gid;
 365                 struct ceph_timespec mtime;
 366                 struct ceph_timespec atime;
 367                 __le64 size, old_size;       /* old_size needed by truncate */
 368                 __le32 mask;                 /* CEPH_SETATTR_* */
 369         } __attribute__ ((packed)) setattr;
 370         struct {
 371                 __le32 frag;                 /* which dir fragment */
 372                 __le32 max_entries;          /* how many dentries to grab */
 373                 __le32 max_bytes;
 374                 __le16 flags;
 375         } __attribute__ ((packed)) readdir;
 376         struct {
 377                 __le32 mode;
 378                 __le32 rdev;
 379         } __attribute__ ((packed)) mknod;
 380         struct {
 381                 __le32 mode;
 382         } __attribute__ ((packed)) mkdir;
 383         struct {
 384                 __le32 flags;
 385                 __le32 mode;
 386                 __le32 stripe_unit;          /* layout for newly created file */
 387                 __le32 stripe_count;         /* ... */
 388                 __le32 object_size;
 389                 __le32 file_replication;
 390                __le32 mask;                 /* CEPH_CAP_* */
 391                __le32 old_size;
 392         } __attribute__ ((packed)) open;
 393         struct {
 394                 __le32 flags;
 395         } __attribute__ ((packed)) setxattr;
 396         struct {
 397                 struct ceph_file_layout_legacy layout;
 398         } __attribute__ ((packed)) setlayout;
 399         struct {
 400                 __u8 rule; /* currently fcntl or flock */
 401                 __u8 type; /* shared, exclusive, remove*/
 402                 __le64 owner; /* owner of the lock */
 403                 __le64 pid; /* process id requesting the lock */
 404                 __le64 start; /* initial location to lock */
 405                 __le64 length; /* num bytes to lock from start */
 406                 __u8 wait; /* will caller wait for lock to become available? */
 407         } __attribute__ ((packed)) filelock_change;
 408 } __attribute__ ((packed));
 409
 410 #define CEPH_MDS_FLAG_REPLAY        1  /* this is a replayed op */
 411 #define CEPH_MDS_FLAG_WANT_DENTRY   2  /* want dentry in reply */
 412
 413 struct ceph_mds_request_head {
 414         __le64 oldest_client_tid;
 415         __le32 mdsmap_epoch;           /* on client */
 416         __le32 flags;                  /* CEPH_MDS_FLAG_* */
 417         __u8 num_retry, num_fwd;       /* count retry, fwd attempts */
 418         __le16 num_releases;           /* # include cap/lease release records */
 419         __le32 op;                     /* mds op code */
 420         __le32 caller_uid, caller_gid;
 421         __le64 ino;                    /* use this ino for openc, mkdir, mknod,
 422                                           etc. (if replaying) */
 423         union ceph_mds_request_args args;
 424 } __attribute__ ((packed));
 425
 426 /* cap/lease release record */
 427 struct ceph_mds_request_release {
 428         __le64 ino, cap_id;            /* ino and unique cap id */
 429         __le32 caps, wanted;           /* new issued, wanted */
 430         __le32 seq, issue_seq, mseq;
 431         __le32 dname_seq;              /* if releasing a dentry lease, a */
 432         __le32 dname_len;              /* string follows. */
 433 } __attribute__ ((packed));
 434
 435 /* client reply */
 436 struct ceph_mds_reply_head {
 437         __le32 op;
 438         __le32 result;
 439         __le32 mdsmap_epoch;
 440         __u8 safe;                     /* true if committed to disk */
 441         __u8 is_dentry, is_target;     /* true if dentry, target inode records
 442                                           are included with reply */
 443 } __attribute__ ((packed));
 444
 445 /* one for each node split */
 446 struct ceph_frag_tree_split {
 447         __le32 frag;                   /* this frag splits... */
 448         __le32 by;                     /* ...by this many bits */
 449 } __attribute__ ((packed));
 450
 451 struct ceph_frag_tree_head {
 452         __le32 nsplits;                /* num ceph_frag_tree_split records */
 453         struct ceph_frag_tree_split splits[];
 454 } __attribute__ ((packed));
 455
 456 /* capability issue, for bundling with mds reply */
 457 struct ceph_mds_reply_cap {
 458         __le32 caps, wanted;           /* caps issued, wanted */
 459         __le64 cap_id;
 460         __le32 seq, mseq;
 461         __le64 realm;                  /* snap realm */
 462         __u8 flags;                    /* CEPH_CAP_FLAG_* */
 463 } __attribute__ ((packed));
 464
 465 #define CEPH_CAP_FLAG_AUTH      (1 << 0)  /* cap is issued by auth mds */
 466 #define CEPH_CAP_FLAG_RELEASE   (1 << 1)  /* release the cap */
 467
 468 /* inode record, for bundling with mds reply */
 469 struct ceph_mds_reply_inode {
 470         __le64 ino;
 471         __le64 snapid;
 472         __le32 rdev;
 473         __le64 version;                /* inode version */
 474         __le64 xattr_version;          /* version for xattr blob */
 475         struct ceph_mds_reply_cap cap; /* caps issued for this inode */
 476         struct ceph_file_layout_legacy layout;
 477         struct ceph_timespec ctime, mtime, atime;
 478         __le32 time_warp_seq;
 479         __le64 size, max_size, truncate_size;
 480         __le32 truncate_seq;
 481         __le32 mode, uid, gid;
 482         __le32 nlink;
 483         __le64 files, subdirs, rbytes, rfiles, rsubdirs;  /* dir stats */
 484         struct ceph_timespec rctime;
 485         struct ceph_frag_tree_head fragtree;  /* (must be at end of struct) */
 486 } __attribute__ ((packed));
 487 /* followed by frag array, symlink string, dir layout, xattr blob */
 488
 489 /* reply_lease follows dname, and reply_inode */
 490 struct ceph_mds_reply_lease {
 491         __le16 mask;            /* lease type(s) */
 492         __le32 duration_ms;     /* lease duration */
 493         __le32 seq;
 494 } __attribute__ ((packed));
 495
 496 struct ceph_mds_reply_dirfrag {
 497         __le32 frag;            /* fragment */
 498         __le32 auth;            /* auth mds, if this is a delegation point */
 499         __le32 ndist;           /* number of mds' this is replicated on */
 500         __le32 dist[];
 501 } __attribute__ ((packed));
 502
 503 #define CEPH_LOCK_FCNTL         1
 504 #define CEPH_LOCK_FLOCK         2
 505 #define CEPH_LOCK_FCNTL_INTR    3
 506 #define CEPH_LOCK_FLOCK_INTR    4
 507
 508
 509 #define CEPH_LOCK_SHARED   1
 510 #define CEPH_LOCK_EXCL     2
 511 #define CEPH_LOCK_UNLOCK   4
 512
 513 struct ceph_filelock {
 514         __le64 start;/* file offset to start lock at */
 515         __le64 length; /* num bytes to lock; 0 for all following start */
 516         __le64 client; /* which client holds the lock */
 517         __le64 owner; /* owner the lock */
 518         __le64 pid; /* process id holding the lock on the client */
 519         __u8 type; /* shared lock, exclusive lock, or unlock */
 520 } __attribute__ ((packed));
 521
 522
 523 /* file access modes */
 524 #define CEPH_FILE_MODE_PIN        0
 525 #define CEPH_FILE_MODE_RD         1
 526 #define CEPH_FILE_MODE_WR         2
 527 #define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
 528 #define CEPH_FILE_MODE_LAZY       4  /* lazy io */
 529 #define CEPH_FILE_MODE_BITS       4
 530
 531 int ceph_flags_to_mode(int flags);
 532
 533 #define CEPH_INLINE_NONE        ((__u64)-1)
 534
 535 /* capability bits */
 536 #define CEPH_CAP_PIN         1  /* no specific capabilities beyond the pin */
 537
 538 /* generic cap bits */
 539 #define CEPH_CAP_GSHARED     1  /* client can reads */
 540 #define CEPH_CAP_GEXCL       2  /* client can read and update */
 541 #define CEPH_CAP_GCACHE      4  /* (file) client can cache reads */
 542 #define CEPH_CAP_GRD         8  /* (file) client can read */
 543 #define CEPH_CAP_GWR        16  /* (file) client can write */
 544 #define CEPH_CAP_GBUFFER    32  /* (file) client can buffer writes */
 545 #define CEPH_CAP_GWREXTEND  64  /* (file) client can extend EOF */
 546 #define CEPH_CAP_GLAZYIO   128  /* (file) client can perform lazy io */
 547
 548 #define CEPH_CAP_SIMPLE_BITS  2
 549 #define CEPH_CAP_FILE_BITS    8
 550
 551 /* per-lock shift */
 552 #define CEPH_CAP_SAUTH      2
 553 #define CEPH_CAP_SLINK      4
 554 #define CEPH_CAP_SXATTR     6
 555 #define CEPH_CAP_SFILE      8
 556 #define CEPH_CAP_SFLOCK    20
 557
 558 #define CEPH_CAP_BITS      22
 559
 560 /* composed values */
 561 #define CEPH_CAP_AUTH_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SAUTH)
 562 #define CEPH_CAP_AUTH_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SAUTH)
 563 #define CEPH_CAP_LINK_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SLINK)
 564 #define CEPH_CAP_LINK_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SLINK)
 565 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED  << CEPH_CAP_SXATTR)
 566 #define CEPH_CAP_XATTR_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SXATTR)
 567 #define CEPH_CAP_FILE(x)    (x << CEPH_CAP_SFILE)
 568 #define CEPH_CAP_FILE_SHARED   (CEPH_CAP_GSHARED   << CEPH_CAP_SFILE)
 569 #define CEPH_CAP_FILE_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SFILE)
 570 #define CEPH_CAP_FILE_CACHE    (CEPH_CAP_GCACHE    << CEPH_CAP_SFILE)
 571 #define CEPH_CAP_FILE_RD       (CEPH_CAP_GRD       << CEPH_CAP_SFILE)
 572 #define CEPH_CAP_FILE_WR       (CEPH_CAP_GWR       << CEPH_CAP_SFILE)
 573 #define CEPH_CAP_FILE_BUFFER   (CEPH_CAP_GBUFFER   << CEPH_CAP_SFILE)
 574 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
 575 #define CEPH_CAP_FILE_LAZYIO   (CEPH_CAP_GLAZYIO   << CEPH_CAP_SFILE)
 576 #define CEPH_CAP_FLOCK_SHARED  (CEPH_CAP_GSHARED   << CEPH_CAP_SFLOCK)
 577 #define CEPH_CAP_FLOCK_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SFLOCK)
 578
 579
 580 /* cap masks (for getattr) */
 581 #define CEPH_STAT_CAP_INODE    CEPH_CAP_PIN
 582 #define CEPH_STAT_CAP_TYPE     CEPH_CAP_PIN  /* mode >> 12 */
 583 #define CEPH_STAT_CAP_SYMLINK  CEPH_CAP_PIN
 584 #define CEPH_STAT_CAP_UID      CEPH_CAP_AUTH_SHARED
 585 #define CEPH_STAT_CAP_GID      CEPH_CAP_AUTH_SHARED
 586 #define CEPH_STAT_CAP_MODE     CEPH_CAP_AUTH_SHARED
 587 #define CEPH_STAT_CAP_NLINK    CEPH_CAP_LINK_SHARED
 588 #define CEPH_STAT_CAP_LAYOUT   CEPH_CAP_FILE_SHARED
 589 #define CEPH_STAT_CAP_MTIME    CEPH_CAP_FILE_SHARED
 590 #define CEPH_STAT_CAP_SIZE     CEPH_CAP_FILE_SHARED
 591 #define CEPH_STAT_CAP_ATIME    CEPH_CAP_FILE_SHARED  /* fixme */
 592 #define CEPH_STAT_CAP_XATTR    CEPH_CAP_XATTR_SHARED
 593 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN |                 \
 594                                  CEPH_CAP_AUTH_SHARED | \
 595                                  CEPH_CAP_LINK_SHARED | \
 596                                  CEPH_CAP_FILE_SHARED | \
 597                                  CEPH_CAP_XATTR_SHARED)
 598 #define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \
 599                                    CEPH_CAP_FILE_RD)
 600
 601 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED |                     \
 602                               CEPH_CAP_LINK_SHARED |                    \
 603                               CEPH_CAP_XATTR_SHARED |                   \
 604                               CEPH_CAP_FILE_SHARED)
 605 #define CEPH_CAP_ANY_RD   (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD |     \
 606                            CEPH_CAP_FILE_CACHE)
 607
 608 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL |         \
 609                            CEPH_CAP_LINK_EXCL |         \
 610                            CEPH_CAP_XATTR_EXCL |        \
 611                            CEPH_CAP_FILE_EXCL)
 612 #define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \
 613                               CEPH_CAP_FILE_SHARED)
 614 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \
 615                               CEPH_CAP_FILE_EXCL)
 616 #define CEPH_CAP_ANY_WR   (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
 617 #define CEPH_CAP_ANY      (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
 618                            CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
 619                            CEPH_CAP_PIN)
 620
 621 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
 622                         CEPH_LOCK_IXATTR)
 623
 624 int ceph_caps_for_mode(int mode);
 625
 626 enum {
 627         CEPH_CAP_OP_GRANT,         /* mds->client grant */
 628         CEPH_CAP_OP_REVOKE,        /* mds->client revoke */
 629         CEPH_CAP_OP_TRUNC,         /* mds->client trunc notify */
 630         CEPH_CAP_OP_EXPORT,        /* mds has exported the cap */
 631         CEPH_CAP_OP_IMPORT,        /* mds has imported the cap */
 632         CEPH_CAP_OP_UPDATE,        /* client->mds update */
 633         CEPH_CAP_OP_DROP,          /* client->mds drop cap bits */
 634         CEPH_CAP_OP_FLUSH,         /* client->mds cap writeback */
 635         CEPH_CAP_OP_FLUSH_ACK,     /* mds->client flushed */
 636         CEPH_CAP_OP_FLUSHSNAP,     /* client->mds flush snapped metadata */
 637         CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
 638         CEPH_CAP_OP_RELEASE,       /* client->mds release (clean) cap */
 639         CEPH_CAP_OP_RENEW,         /* client->mds renewal request */
 640 };
 641
 642 extern const char *ceph_cap_op_name(int op);
 643
 644 /*
 645  * caps message, used for capability callbacks, acks, requests, etc.
 646  */
 647 struct ceph_mds_caps {
 648         __le32 op;                  /* CEPH_CAP_OP_* */
 649         __le64 ino, realm;
 650         __le64 cap_id;
 651         __le32 seq, issue_seq;
 652         __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
 653         __le32 migrate_seq;
 654         __le64 snap_follows;
 655         __le32 snap_trace_len;
 656
 657         /* authlock */
 658         __le32 uid, gid, mode;
 659
 660         /* linklock */
 661         __le32 nlink;
 662
 663         /* xattrlock */
 664         __le32 xattr_len;
 665         __le64 xattr_version;
 666
 667         /* filelock */
 668         __le64 size, max_size, truncate_size;
 669         __le32 truncate_seq;
 670         struct ceph_timespec mtime, atime, ctime;
 671         struct ceph_file_layout_legacy layout;
 672         __le32 time_warp_seq;
 673 } __attribute__ ((packed));
 674
 675 struct ceph_mds_cap_peer {
 676         __le64 cap_id;
 677         __le32 seq;
 678         __le32 mseq;
 679         __le32 mds;
 680         __u8   flags;
 681 } __attribute__ ((packed));
 682
 683 /* cap release msg head */
 684 struct ceph_mds_cap_release {
 685         __le32 num;                /* number of cap_items that follow */
 686 } __attribute__ ((packed));
 687
 688 struct ceph_mds_cap_item {
 689         __le64 ino;
 690         __le64 cap_id;
 691         __le32 migrate_seq, seq;
 692 } __attribute__ ((packed));
 693
 694 #define CEPH_MDS_LEASE_REVOKE           1  /*    mds  -> client */
 695 #define CEPH_MDS_LEASE_RELEASE          2  /* client  -> mds    */
 696 #define CEPH_MDS_LEASE_RENEW            3  /* client <-> mds    */
 697 #define CEPH_MDS_LEASE_REVOKE_ACK       4  /* client  -> mds    */
 698
 699 extern const char *ceph_lease_op_name(int o);
 700
 701 /* lease msg header */
 702 struct ceph_mds_lease {
 703         __u8 action;            /* CEPH_MDS_LEASE_* */
 704         __le16 mask;            /* which lease */
 705         __le64 ino;
 706         __le64 first, last;     /* snap range */
 707         __le32 seq;
 708         __le32 duration_ms;     /* duration of renewal */
 709 } __attribute__ ((packed));
 710 /* followed by a __le32+string for dname */
 711
 712 /* client reconnect */
 713 struct ceph_mds_cap_reconnect {
 714         __le64 cap_id;
 715         __le32 wanted;
 716         __le32 issued;
 717         __le64 snaprealm;
 718         __le64 pathbase;        /* base ino for our path to this ino */
 719         __le32 flock_len;       /* size of flock state blob, if any */
 720 } __attribute__ ((packed));
 721 /* followed by flock blob */
 722
 723 struct ceph_mds_cap_reconnect_v1 {
 724         __le64 cap_id;
 725         __le32 wanted;
 726         __le32 issued;
 727         __le64 size;
 728         struct ceph_timespec mtime, atime;
 729         __le64 snaprealm;
 730         __le64 pathbase;        /* base ino for our path to this ino */
 731 } __attribute__ ((packed));
 732
 733 struct ceph_mds_snaprealm_reconnect {
 734         __le64 ino;     /* snap realm base */
 735         __le64 seq;     /* snap seq for this snap realm */
 736         __le64 parent;  /* parent realm */
 737 } __attribute__ ((packed));
 738
 739 /*
 740  * snaps
 741  */
 742 enum {
 743         CEPH_SNAP_OP_UPDATE,  /* CREATE or DESTROY */
 744         CEPH_SNAP_OP_CREATE,
 745         CEPH_SNAP_OP_DESTROY,
 746         CEPH_SNAP_OP_SPLIT,
 747 };
 748
 749 extern const char *ceph_snap_op_name(int o);
 750
 751 /* snap msg header */
 752 struct ceph_mds_snap_head {
 753         __le32 op;                /* CEPH_SNAP_OP_* */
 754         __le64 split;             /* ino to split off, if any */
 755         __le32 num_split_inos;    /* # inos belonging to new child realm */
 756         __le32 num_split_realms;  /* # child realms udner new child realm */
 757         __le32 trace_len;         /* size of snap trace blob */
 758 } __attribute__ ((packed));
 759 /* followed by split ino list, then split realms, then the trace blob */
 760
 761 /*
 762  * encode info about a snaprealm, as viewed by a client
 763  */
 764 struct ceph_mds_snap_realm {
 765         __le64 ino;           /* ino */
 766         __le64 created;       /* snap: when created */
 767         __le64 parent;        /* ino: parent realm */
 768         __le64 parent_since;  /* snap: same parent since */
 769         __le64 seq;           /* snap: version */
 770         __le32 num_snaps;
 771         __le32 num_prior_parent_snaps;
 772 } __attribute__ ((packed));
 773 /* followed by my snap list, then prior parent snap list */
 774
 775 #endif