ceph/src/include/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 "msgr.h"
  16 #include "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_LOST_AND_FOUND 4       /* reserved ino for use in recovery */
  32
  33 /* arbitrary limit on max # of monitors (cluster of 3 is typical) */
  34 #define CEPH_MAX_MON   31
  35
  36 /*
  37  * ceph_file_layout - describe data layout for a file/inode
  38  */
  39 struct ceph_file_layout {
  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 #define CEPH_MIN_STRIPE_UNIT 65536
  57
  58 struct ceph_dir_layout {
  59         __u8   dl_dir_hash;   /* see ceph_hash.h for ids */
  60         __u8   dl_unused1;
  61         __u16  dl_unused2;
  62         __u32  dl_unused3;
  63 } __attribute__ ((packed));
  64
  65 /* crypto algorithms */
  66 #define CEPH_CRYPTO_NONE 0x0
  67 #define CEPH_CRYPTO_AES  0x1
  68
  69 #define CEPH_AES_IV "cephsageyudagreg"
  70
  71 /* security/authentication protocols */
  72 #define CEPH_AUTH_UNKNOWN       0x0
  73 #define CEPH_AUTH_NONE          0x1
  74 #define CEPH_AUTH_CEPHX         0x2
  75
  76 #define CEPH_AUTH_UID_DEFAULT ((__u64) -1)
  77
  78
  79 /*********************************************
  80  * message layer
  81  */
  82
  83 /*
  84  * message types
  85  */
  86
  87 /* misc */
  88 #define CEPH_MSG_SHUTDOWN               1
  89 #define CEPH_MSG_PING                   2
  90
  91 /* client <-> monitor */
  92 #define CEPH_MSG_MON_MAP                4
  93 #define CEPH_MSG_MON_GET_MAP            5
  94 #define CEPH_MSG_MON_GET_OSDMAP         6
  95 #define CEPH_MSG_MON_METADATA           7
  96 #define CEPH_MSG_STATFS                 13
  97 #define CEPH_MSG_STATFS_REPLY           14
  98 #define CEPH_MSG_MON_SUBSCRIBE          15
  99 #define CEPH_MSG_MON_SUBSCRIBE_ACK      16
 100 #define CEPH_MSG_AUTH                   17
 101 #define CEPH_MSG_AUTH_REPLY             18
 102 #define CEPH_MSG_MON_GET_VERSION        19
 103 #define CEPH_MSG_MON_GET_VERSION_REPLY  20
 104
 105 /* client <-> mds */
 106 #define CEPH_MSG_MDS_MAP                21
 107
 108 #define CEPH_MSG_CLIENT_SESSION         22
 109 #define CEPH_MSG_CLIENT_RECONNECT       23
 110
 111 #define CEPH_MSG_CLIENT_REQUEST         24
 112 #define CEPH_MSG_CLIENT_REQUEST_FORWARD 25
 113 #define CEPH_MSG_CLIENT_REPLY           26
 114 #define CEPH_MSG_CLIENT_CAPS            0x310
 115 #define CEPH_MSG_CLIENT_LEASE           0x311
 116 #define CEPH_MSG_CLIENT_SNAP            0x312
 117 #define CEPH_MSG_CLIENT_CAPRELEASE      0x313
 118 #define CEPH_MSG_CLIENT_QUOTA           0x314
 119
 120 /* pool ops */
 121 #define CEPH_MSG_POOLOP_REPLY           48
 122 #define CEPH_MSG_POOLOP                 49
 123
 124
 125 /* osd */
 126 #define CEPH_MSG_OSD_MAP                41
 127 #define CEPH_MSG_OSD_OP                 42
 128 #define CEPH_MSG_OSD_OPREPLY            43
 129 #define CEPH_MSG_WATCH_NOTIFY           44
 130 #define CEPH_MSG_OSD_BACKOFF            61
 131
 132 /* FSMap subscribers (see all MDS clusters at once) */
 133 #define CEPH_MSG_FS_MAP                 45
 134 /* FSMapUser subscribers (get MDS clusters name->ID mapping) */
 135 #define CEPH_MSG_FS_MAP_USER            103
 136
 137 /* watch-notify operations */
 138 enum {
 139         CEPH_WATCH_EVENT_NOTIFY           = 1, /* notifying watcher */
 140         CEPH_WATCH_EVENT_NOTIFY_COMPLETE  = 2, /* notifier notified when done */
 141         CEPH_WATCH_EVENT_DISCONNECT       = 3, /* we were disconnected */
 142 };
 143
 144 const char *ceph_watch_event_name(int o);
 145
 146 /* pool operations */
 147 enum {
 148   POOL_OP_CREATE                        = 0x01,
 149   POOL_OP_DELETE                        = 0x02,
 150   POOL_OP_AUID_CHANGE                   = 0x03,
 151   POOL_OP_CREATE_SNAP                   = 0x11,
 152   POOL_OP_DELETE_SNAP                   = 0x12,
 153   POOL_OP_CREATE_UNMANAGED_SNAP         = 0x21,
 154   POOL_OP_DELETE_UNMANAGED_SNAP         = 0x22,
 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 const char *ceph_pool_op_name(int op);
 180
 181 struct ceph_mon_poolop {
 182         struct ceph_mon_request_header monhdr;
 183         struct ceph_fsid fsid;
 184         __le32 pool;
 185         __le32 op;
 186         __le64 auid;
 187         __le64 snapid;
 188         __le32 name_len;
 189 } __attribute__ ((packed));
 190
 191 struct ceph_mon_poolop_reply {
 192         struct ceph_mon_request_header monhdr;
 193         struct ceph_fsid fsid;
 194         __le32 reply_code;
 195         __le32 epoch;
 196         char has_data;
 197         char data[0];
 198 } __attribute__ ((packed));
 199
 200 struct ceph_mon_unmanaged_snap {
 201         __le64 snapid;
 202 } __attribute__ ((packed));
 203
 204 struct ceph_osd_getmap {
 205         struct ceph_mon_request_header monhdr;
 206         struct ceph_fsid fsid;
 207         __le32 start;
 208 } __attribute__ ((packed));
 209
 210 struct ceph_mds_getmap {
 211         struct ceph_mon_request_header monhdr;
 212         struct ceph_fsid fsid;
 213 } __attribute__ ((packed));
 214
 215 struct ceph_client_mount {
 216         struct ceph_mon_request_header monhdr;
 217 } __attribute__ ((packed));
 218
 219 #define CEPH_SUBSCRIBE_ONETIME    1  /* i want only 1 update after have */
 220
 221 struct ceph_mon_subscribe_item {
 222         __le64 start;
 223         __u8 flags;
 224 } __attribute__ ((packed));
 225
 226 struct ceph_mon_subscribe_ack {
 227         __le32 duration;         /* seconds */
 228         struct ceph_fsid fsid;
 229 } __attribute__ ((packed));
 230
 231 /*
 232  * mdsmap flags
 233  */
 234 #define CEPH_MDSMAP_DOWN    (1<<0)  /* cluster deliberately down */
 235 #define CEPH_MDSMAP_ALLOW_SNAPS   (1<<1)  /* cluster allowed to create snapshots */
 236 #define CEPH_MDSMAP_ALLOW_MULTIMDS (1<<2) /* cluster allowed to have >1 active MDS */
 237 #define CEPH_MDSMAP_ALLOW_DIRFRAGS (1<<3) /* cluster allowed to fragment directories */
 238
 239 #define CEPH_MDSMAP_ALLOW_CLASSICS (CEPH_MDSMAP_ALLOW_SNAPS | CEPH_MDSMAP_ALLOW_MULTIMDS | \
 240                                     CEPH_MDSMAP_ALLOW_DIRFRAGS)
 241
 242 #define CEPH_MDSMAP_DEFAULTS CEPH_MDSMAP_ALLOW_DIRFRAGS | CEPH_MDSMAP_ALLOW_MULTIMDS
 243
 244 /*
 245  * mds states
 246  *   > 0 -> in
 247  *  <= 0 -> out
 248  */
 249 #define CEPH_MDS_STATE_DNE          0  /* down, does not exist. */
 250 #define CEPH_MDS_STATE_STOPPED     -1  /* down, once existed, but no subtrees.
 251                                           empty log. */
 252 #define CEPH_MDS_STATE_BOOT        -4  /* up, boot announcement. */
 253 #define CEPH_MDS_STATE_STANDBY     -5  /* up, idle.  waiting for assignment. */
 254 #define CEPH_MDS_STATE_CREATING    -6  /* up, creating MDS instance. */
 255 #define CEPH_MDS_STATE_STARTING    -7  /* up, starting previously stopped mds */
 256 #define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
 257 #define CEPH_MDS_STATE_REPLAYONCE   -9 /* Legacy, unused */
 258 #define CEPH_MDS_STATE_NULL         -10
 259
 260 #define CEPH_MDS_STATE_REPLAY       8  /* up, replaying journal. */
 261 #define CEPH_MDS_STATE_RESOLVE      9  /* up, disambiguating distributed
 262                                           operations (import, rename, etc.) */
 263 #define CEPH_MDS_STATE_RECONNECT    10 /* up, reconnect to clients */
 264 #define CEPH_MDS_STATE_REJOIN       11 /* up, rejoining distributed cache */
 265 #define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */
 266 #define CEPH_MDS_STATE_ACTIVE       13 /* up, active */
 267 #define CEPH_MDS_STATE_STOPPING     14 /* up, but exporting metadata */
 268 #define CEPH_MDS_STATE_DAMAGED      15 /* rank not replayable, need repair */
 269
 270 extern const char *ceph_mds_state_name(int s);
 271
 272
 273 /*
 274  * metadata lock types.
 275  *  - these are bitmasks.. we can compose them
 276  *  - they also define the lock ordering by the MDS
 277  *  - a few of these are internal to the mds
 278  */
 279 #define CEPH_LOCK_DVERSION    1
 280 #define CEPH_LOCK_DN          2
 281 #define CEPH_LOCK_ISNAP       16
 282 #define CEPH_LOCK_IVERSION    32    /* mds internal */
 283 #define CEPH_LOCK_IFILE       64
 284 #define CEPH_LOCK_IAUTH       128
 285 #define CEPH_LOCK_ILINK       256
 286 #define CEPH_LOCK_IDFT        512   /* dir frag tree */
 287 #define CEPH_LOCK_INEST       1024  /* mds internal */
 288 #define CEPH_LOCK_IXATTR      2048
 289 #define CEPH_LOCK_IFLOCK      4096  /* advisory file locks */
 290 #define CEPH_LOCK_INO         8192  /* immutable inode bits; not a lock */
 291 #define CEPH_LOCK_IPOLICY     16384 /* policy lock on dirs. MDS internal */
 292
 293 /* client_session ops */
 294 enum {
 295         CEPH_SESSION_REQUEST_OPEN,
 296         CEPH_SESSION_OPEN,
 297         CEPH_SESSION_REQUEST_CLOSE,
 298         CEPH_SESSION_CLOSE,
 299         CEPH_SESSION_REQUEST_RENEWCAPS,
 300         CEPH_SESSION_RENEWCAPS,
 301         CEPH_SESSION_STALE,
 302         CEPH_SESSION_RECALL_STATE,
 303         CEPH_SESSION_FLUSHMSG,
 304         CEPH_SESSION_FLUSHMSG_ACK,
 305         CEPH_SESSION_FORCE_RO,
 306     // A response to REQUEST_OPEN indicating that the client should
 307     // permanently desist from contacting the MDS
 308         CEPH_SESSION_REJECT,
 309         CEPH_SESSION_REQUEST_FLUSH_MDLOG
 310 };
 311
 312 extern const char *ceph_session_op_name(int op);
 313
 314 struct ceph_mds_session_head {
 315         __le32 op;
 316         __le64 seq;
 317         struct ceph_timespec stamp;
 318         __le32 max_caps, max_leases;
 319 } __attribute__ ((packed));
 320
 321 /* client_request */
 322 /*
 323  * metadata ops.
 324  *  & 0x001000 -> write op
 325  *  & 0x010000 -> follow symlink (e.g. stat(), not lstat()).
 326  &  & 0x100000 -> use weird ino/path trace
 327  */
 328 #define CEPH_MDS_OP_WRITE        0x001000
 329 enum {
 330         CEPH_MDS_OP_LOOKUP     = 0x00100,
 331         CEPH_MDS_OP_GETATTR    = 0x00101,
 332         CEPH_MDS_OP_LOOKUPHASH = 0x00102,
 333         CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
 334         CEPH_MDS_OP_LOOKUPINO  = 0x00104,
 335         CEPH_MDS_OP_LOOKUPNAME = 0x00105,
 336
 337         CEPH_MDS_OP_SETXATTR   = 0x01105,
 338         CEPH_MDS_OP_RMXATTR    = 0x01106,
 339         CEPH_MDS_OP_SETLAYOUT  = 0x01107,
 340         CEPH_MDS_OP_SETATTR    = 0x01108,
 341         CEPH_MDS_OP_SETFILELOCK= 0x01109,
 342         CEPH_MDS_OP_GETFILELOCK= 0x00110,
 343         CEPH_MDS_OP_SETDIRLAYOUT=0x0110a,
 344
 345         CEPH_MDS_OP_MKNOD      = 0x01201,
 346         CEPH_MDS_OP_LINK       = 0x01202,
 347         CEPH_MDS_OP_UNLINK     = 0x01203,
 348         CEPH_MDS_OP_RENAME     = 0x01204,
 349         CEPH_MDS_OP_MKDIR      = 0x01220,
 350         CEPH_MDS_OP_RMDIR      = 0x01221,
 351         CEPH_MDS_OP_SYMLINK    = 0x01222,
 352
 353         CEPH_MDS_OP_CREATE     = 0x01301,
 354         CEPH_MDS_OP_OPEN       = 0x00302,
 355         CEPH_MDS_OP_READDIR    = 0x00305,
 356
 357         CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
 358         CEPH_MDS_OP_MKSNAP     = 0x01400,
 359         CEPH_MDS_OP_RMSNAP     = 0x01401,
 360         CEPH_MDS_OP_LSSNAP     = 0x00402,
 361         CEPH_MDS_OP_RENAMESNAP = 0x01403,
 362
 363         // internal op
 364         CEPH_MDS_OP_FRAGMENTDIR= 0x01500,
 365         CEPH_MDS_OP_EXPORTDIR  = 0x01501,
 366         CEPH_MDS_OP_FLUSH      = 0x01502,
 367         CEPH_MDS_OP_ENQUEUE_SCRUB  = 0x01503,
 368         CEPH_MDS_OP_REPAIR_FRAGSTATS = 0x01504,
 369         CEPH_MDS_OP_REPAIR_INODESTATS = 0x01505
 370 };
 371
 372 extern const char *ceph_mds_op_name(int op);
 373
 374 #ifndef CEPH_SETATTR_MODE
 375 #define CEPH_SETATTR_MODE       (1 << 0)
 376 #define CEPH_SETATTR_UID        (1 << 1)
 377 #define CEPH_SETATTR_GID        (1 << 2)
 378 #define CEPH_SETATTR_MTIME      (1 << 3)
 379 #define CEPH_SETATTR_ATIME      (1 << 4)
 380 #define CEPH_SETATTR_SIZE       (1 << 5)
 381 #define CEPH_SETATTR_CTIME      (1 << 6)
 382 #define CEPH_SETATTR_BTIME      (1 << 9)
 383 #endif
 384 #define CEPH_SETATTR_MTIME_NOW  (1 << 7)
 385 #define CEPH_SETATTR_ATIME_NOW  (1 << 8)
 386 #define CEPH_SETATTR_KILL_SGUID (1 << 10)
 387
 388 /*
 389  * open request flags
 390  */
 391 #define CEPH_O_RDONLY          00000000
 392 #define CEPH_O_WRONLY          00000001
 393 #define CEPH_O_RDWR            00000002
 394 #define CEPH_O_CREAT           00000100
 395 #define CEPH_O_EXCL            00000200
 396 #define CEPH_O_TRUNC           00001000
 397 #define CEPH_O_DIRECTORY       00200000
 398 #define CEPH_O_NOFOLLOW        00400000
 399
 400 int ceph_flags_sys2wire(int flags);
 401
 402 /*
 403  * Ceph setxattr request flags.
 404  */
 405 #define CEPH_XATTR_CREATE  (1 << 0)
 406 #define CEPH_XATTR_REPLACE (1 << 1)
 407 #define CEPH_XATTR_REMOVE  (1 << 31)
 408
 409 /*
 410  * readdir request flags;
 411  */
 412 #define CEPH_READDIR_REPLY_BITFLAGS     (1<<0)
 413
 414 /*
 415  * readdir reply flags.
 416  */
 417 #define CEPH_READDIR_FRAG_END           (1<<0)
 418 #define CEPH_READDIR_FRAG_COMPLETE      (1<<8)
 419 #define CEPH_READDIR_HASH_ORDER         (1<<9)
 420 #define CEPH_READDIR_OFFSET_HASH       (1<<10)
 421
 422 /* Note that this is embedded wthin ceph_mds_request_head_legacy. */
 423 union ceph_mds_request_args_legacy {
 424         struct {
 425                 __le32 mask;                 /* CEPH_CAP_* */
 426         } __attribute__ ((packed)) getattr;
 427         struct {
 428                 __le32 mode;
 429                 __le32 uid;
 430                 __le32 gid;
 431                 struct ceph_timespec mtime;
 432                 struct ceph_timespec atime;
 433                 __le64 size, old_size;       /* old_size needed by truncate */
 434                 __le32 mask;                 /* CEPH_SETATTR_* */
 435         } __attribute__ ((packed)) setattr;
 436         struct {
 437                 __le32 frag;                 /* which dir fragment */
 438                 __le32 max_entries;          /* how many dentries to grab */
 439                 __le32 max_bytes;
 440                 __le16 flags;
 441                __le32 offset_hash;
 442         } __attribute__ ((packed)) readdir;
 443         struct {
 444                 __le32 mode;
 445                 __le32 rdev;
 446         } __attribute__ ((packed)) mknod;
 447         struct {
 448                 __le32 mode;
 449         } __attribute__ ((packed)) mkdir;
 450         struct {
 451                 __le32 flags;
 452                 __le32 mode;
 453                 __le32 stripe_unit;          /* layout for newly created file */
 454                 __le32 stripe_count;         /* ... */
 455                 __le32 object_size;
 456                 __le32 pool;                 /* if >= 0 and CREATEPOOLID feature */
 457                 __le32 mask;                 /* CEPH_CAP_* */
 458                 __le64 old_size;             /* if O_TRUNC */
 459         } __attribute__ ((packed)) open;
 460         struct {
 461                 __le32 flags;
 462                 __le32 osdmap_epoch;        /* use for set file/dir layout */
 463         } __attribute__ ((packed)) setxattr;
 464         struct {
 465                 struct ceph_file_layout layout;
 466         } __attribute__ ((packed)) setlayout;
 467         struct {
 468                 __u8 rule; /* currently fcntl or flock */
 469                 __u8 type; /* shared, exclusive, remove*/
 470                 __le64 owner; /* who requests/holds the lock */
 471                 __le64 pid; /* process id requesting the lock */
 472                 __le64 start; /* initial location to lock */
 473                 __le64 length; /* num bytes to lock from start */
 474                 __u8 wait; /* will caller wait for lock to become available? */
 475         } __attribute__ ((packed)) filelock_change;
 476 } __attribute__ ((packed));
 477
 478 #define CEPH_MDS_FLAG_REPLAY        1  /* this is a replayed op */
 479 #define CEPH_MDS_FLAG_WANT_DENTRY   2  /* want dentry in reply */
 480
 481 struct ceph_mds_request_head_legacy {
 482         __le64 oldest_client_tid;
 483         __le32 mdsmap_epoch;           /* on client */
 484         __le32 flags;                  /* CEPH_MDS_FLAG_* */
 485         __u8 num_retry, num_fwd;       /* count retry, fwd attempts */
 486         __le16 num_releases;           /* # include cap/lease release records */
 487         __le32 op;                     /* mds op code */
 488         __le32 caller_uid, caller_gid;
 489         __le64 ino;                    /* use this ino for openc, mkdir, mknod,
 490                                           etc. (if replaying) */
 491         union ceph_mds_request_args_legacy args;
 492 } __attribute__ ((packed));
 493
 494 /*
 495  * Note that this is embedded wthin ceph_mds_request_head. Also, compatability
 496  * with the ceph_mds_request_args_legacy must be maintained!
 497  */
 498 union ceph_mds_request_args {
 499         struct {
 500                 __le32 mask;                 /* CEPH_CAP_* */
 501         } __attribute__ ((packed)) getattr;
 502         struct {
 503                 __le32 mode;
 504                 __le32 uid;
 505                 __le32 gid;
 506                 struct ceph_timespec mtime;
 507                 struct ceph_timespec atime;
 508                 __le64 size, old_size;       /* old_size needed by truncate */
 509                 __le32 mask;                 /* CEPH_SETATTR_* */
 510                 struct ceph_timespec btime;
 511         } __attribute__ ((packed)) setattr;
 512         struct {
 513                 __le32 frag;                 /* which dir fragment */
 514                 __le32 max_entries;          /* how many dentries to grab */
 515                 __le32 max_bytes;
 516                 __le16 flags;
 517                __le32 offset_hash;
 518         } __attribute__ ((packed)) readdir;
 519         struct {
 520                 __le32 mode;
 521                 __le32 rdev;
 522         } __attribute__ ((packed)) mknod;
 523         struct {
 524                 __le32 mode;
 525         } __attribute__ ((packed)) mkdir;
 526         struct {
 527                 __le32 flags;
 528                 __le32 mode;
 529                 __le32 stripe_unit;          /* layout for newly created file */
 530                 __le32 stripe_count;         /* ... */
 531                 __le32 object_size;
 532                 __le32 pool;                 /* if >= 0 and CREATEPOOLID feature */
 533                 __le32 mask;                 /* CEPH_CAP_* */
 534                 __le64 old_size;             /* if O_TRUNC */
 535         } __attribute__ ((packed)) open;
 536         struct {
 537                 __le32 flags;
 538                 __le32 osdmap_epoch;        /* use for set file/dir layout */
 539         } __attribute__ ((packed)) setxattr;
 540         struct {
 541                 struct ceph_file_layout layout;
 542         } __attribute__ ((packed)) setlayout;
 543         struct {
 544                 __u8 rule; /* currently fcntl or flock */
 545                 __u8 type; /* shared, exclusive, remove*/
 546                 __le64 owner; /* who requests/holds the lock */
 547                 __le64 pid; /* process id requesting the lock */
 548                 __le64 start; /* initial location to lock */
 549                 __le64 length; /* num bytes to lock from start */
 550                 __u8 wait; /* will caller wait for lock to become available? */
 551         } __attribute__ ((packed)) filelock_change;
 552 } __attribute__ ((packed));
 553
 554 #define CEPH_MDS_REQUEST_HEAD_VERSION   1
 555
 556 /*
 557  * Note that any change to this structure must ensure that it is compatible
 558  * with ceph_mds_request_head_legacy.
 559  */
 560 struct ceph_mds_request_head {
 561         __le16 version;
 562         __le64 oldest_client_tid;
 563         __le32 mdsmap_epoch;           /* on client */
 564         __le32 flags;                  /* CEPH_MDS_FLAG_* */
 565         __u8 num_retry, num_fwd;       /* count retry, fwd attempts */
 566         __le16 num_releases;           /* # include cap/lease release records */
 567         __le32 op;                     /* mds op code */
 568         __le32 caller_uid, caller_gid;
 569         __le64 ino;                    /* use this ino for openc, mkdir, mknod,
 570                                           etc. (if replaying) */
 571         union ceph_mds_request_args args;
 572 } __attribute__ ((packed));
 573
 574 /* cap/lease release record */
 575 struct ceph_mds_request_release {
 576         __le64 ino, cap_id;            /* ino and unique cap id */
 577         __le32 caps, wanted;           /* new issued, wanted */
 578         __le32 seq, issue_seq, mseq;
 579         __le32 dname_seq;              /* if releasing a dentry lease, a */
 580         __le32 dname_len;              /* string follows. */
 581 } __attribute__ ((packed));
 582
 583 static inline void
 584 copy_from_legacy_head(struct ceph_mds_request_head *head,
 585                         struct ceph_mds_request_head_legacy *legacy)
 586 {
 587         memcpy(&(head->oldest_client_tid), legacy, sizeof(*legacy));
 588 }
 589
 590 static inline void
 591 copy_to_legacy_head(struct ceph_mds_request_head_legacy *legacy,
 592                         struct ceph_mds_request_head *head)
 593 {
 594         memcpy(legacy, &(head->oldest_client_tid), sizeof(*legacy));
 595 }
 596
 597 /* client reply */
 598 struct ceph_mds_reply_head {
 599         __le32 op;
 600         __le32 result;
 601         __le32 mdsmap_epoch;
 602         __u8 safe;                     /* true if committed to disk */
 603         __u8 is_dentry, is_target;     /* true if dentry, target inode records
 604                                           are included with reply */
 605 } __attribute__ ((packed));
 606
 607 /* one for each node split */
 608 struct ceph_frag_tree_split {
 609         __le32 frag;                   /* this frag splits... */
 610         __le32 by;                     /* ...by this many bits */
 611 } __attribute__ ((packed));
 612
 613 struct ceph_frag_tree_head {
 614         __le32 nsplits;                /* num ceph_frag_tree_split records */
 615         struct ceph_frag_tree_split splits[];
 616 } __attribute__ ((packed));
 617
 618 /* capability issue, for bundling with mds reply */
 619 struct ceph_mds_reply_cap {
 620         __le32 caps, wanted;           /* caps issued, wanted */
 621         __le64 cap_id;
 622         __le32 seq, mseq;
 623         __le64 realm;                  /* snap realm */
 624         __u8 flags;                    /* CEPH_CAP_FLAG_* */
 625 } __attribute__ ((packed));
 626
 627 #define CEPH_CAP_FLAG_AUTH      (1 << 0)        /* cap is issued by auth mds */
 628 #define CEPH_CAP_FLAG_RELEASE   (1 << 1)        /* ask client to release the cap */
 629
 630 /* reply_lease follows dname, and reply_inode */
 631 struct ceph_mds_reply_lease {
 632         __le16 mask;            /* lease type(s) */
 633         __le32 duration_ms;     /* lease duration */
 634         __le32 seq;
 635 } __attribute__ ((packed));
 636
 637 struct ceph_mds_reply_dirfrag {
 638         __le32 frag;            /* fragment */
 639         __le32 auth;            /* auth mds, if this is a delegation point */
 640         __le32 ndist;           /* number of mds' this is replicated on */
 641         __le32 dist[];
 642 } __attribute__ ((packed));
 643
 644 #define CEPH_LOCK_FCNTL         1
 645 #define CEPH_LOCK_FLOCK         2
 646 #define CEPH_LOCK_FCNTL_INTR    3
 647 #define CEPH_LOCK_FLOCK_INTR    4
 648
 649 #define CEPH_LOCK_SHARED   1
 650 #define CEPH_LOCK_EXCL     2
 651 #define CEPH_LOCK_UNLOCK   4
 652
 653 struct ceph_filelock {
 654         __le64 start;/* file offset to start lock at */
 655         __le64 length; /* num bytes to lock; 0 for all following start */
 656         __le64 client; /* which client holds the lock */
 657         __le64 owner; /* who requests/holds the lock */
 658         __le64 pid; /* process id holding the lock on the client */
 659         __u8 type; /* shared lock, exclusive lock, or unlock */
 660 } __attribute__ ((packed));
 661
 662
 663 /* file access modes */
 664 #define CEPH_FILE_MODE_PIN        0
 665 #define CEPH_FILE_MODE_RD         1
 666 #define CEPH_FILE_MODE_WR         2
 667 #define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
 668 #define CEPH_FILE_MODE_LAZY       4  /* lazy io */
 669 #define CEPH_FILE_MODE_NUM        8  /* bc these are bit fields.. mostly */
 670
 671 int ceph_flags_to_mode(int flags);
 672
 673 /* inline data state */
 674 #define CEPH_INLINE_NONE        ((__u64)-1)
 675 #define CEPH_INLINE_MAX_SIZE    CEPH_MIN_STRIPE_UNIT
 676
 677 /* capability bits */
 678 #define CEPH_CAP_PIN         1  /* no specific capabilities beyond the pin */
 679
 680 /* generic cap bits */
 681 #define CEPH_CAP_GSHARED     1  /* client can reads */
 682 #define CEPH_CAP_GEXCL       2  /* client can read and update */
 683 #define CEPH_CAP_GCACHE      4  /* (file) client can cache reads */
 684 #define CEPH_CAP_GRD         8  /* (file) client can read */
 685 #define CEPH_CAP_GWR        16  /* (file) client can write */
 686 #define CEPH_CAP_GBUFFER    32  /* (file) client can buffer writes */
 687 #define CEPH_CAP_GWREXTEND  64  /* (file) client can extend EOF */
 688 #define CEPH_CAP_GLAZYIO   128  /* (file) client can perform lazy io */
 689
 690 #define CEPH_CAP_SIMPLE_BITS  2
 691 #define CEPH_CAP_FILE_BITS    8
 692
 693 /* per-lock shift */
 694 #define CEPH_CAP_SAUTH      2
 695 #define CEPH_CAP_SLINK      4
 696 #define CEPH_CAP_SXATTR     6
 697 #define CEPH_CAP_SFILE      8
 698
 699 /* composed values */
 700 #define CEPH_CAP_AUTH_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SAUTH)
 701 #define CEPH_CAP_AUTH_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SAUTH)
 702 #define CEPH_CAP_LINK_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SLINK)
 703 #define CEPH_CAP_LINK_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SLINK)
 704 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED  << CEPH_CAP_SXATTR)
 705 #define CEPH_CAP_XATTR_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SXATTR)
 706 #define CEPH_CAP_FILE(x)    (x << CEPH_CAP_SFILE)
 707 #define CEPH_CAP_FILE_SHARED   (CEPH_CAP_GSHARED   << CEPH_CAP_SFILE)
 708 #define CEPH_CAP_FILE_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SFILE)
 709 #define CEPH_CAP_FILE_CACHE    (CEPH_CAP_GCACHE    << CEPH_CAP_SFILE)
 710 #define CEPH_CAP_FILE_RD       (CEPH_CAP_GRD       << CEPH_CAP_SFILE)
 711 #define CEPH_CAP_FILE_WR       (CEPH_CAP_GWR       << CEPH_CAP_SFILE)
 712 #define CEPH_CAP_FILE_BUFFER   (CEPH_CAP_GBUFFER   << CEPH_CAP_SFILE)
 713 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
 714 #define CEPH_CAP_FILE_LAZYIO   (CEPH_CAP_GLAZYIO   << CEPH_CAP_SFILE)
 715
 716 /* cap masks (for getattr) */
 717 #define CEPH_STAT_CAP_INODE    CEPH_CAP_PIN
 718 #define CEPH_STAT_CAP_TYPE     CEPH_CAP_PIN  /* mode >> 12 */
 719 #define CEPH_STAT_CAP_SYMLINK  CEPH_CAP_PIN
 720 #define CEPH_STAT_CAP_UID      CEPH_CAP_AUTH_SHARED
 721 #define CEPH_STAT_CAP_GID      CEPH_CAP_AUTH_SHARED
 722 #define CEPH_STAT_CAP_MODE     CEPH_CAP_AUTH_SHARED
 723 #define CEPH_STAT_CAP_NLINK    CEPH_CAP_LINK_SHARED
 724 #define CEPH_STAT_CAP_LAYOUT   CEPH_CAP_FILE_SHARED
 725 #define CEPH_STAT_CAP_MTIME    CEPH_CAP_FILE_SHARED
 726 #define CEPH_STAT_CAP_SIZE     CEPH_CAP_FILE_SHARED
 727 #define CEPH_STAT_CAP_ATIME    CEPH_CAP_FILE_SHARED  /* fixme */
 728 #define CEPH_STAT_CAP_XATTR    CEPH_CAP_XATTR_SHARED
 729 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN |                 \
 730                                  CEPH_CAP_AUTH_SHARED | \
 731                                  CEPH_CAP_LINK_SHARED | \
 732                                  CEPH_CAP_FILE_SHARED | \
 733                                  CEPH_CAP_XATTR_SHARED)
 734 #define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \
 735                                    CEPH_CAP_FILE_RD)
 736 #define CEPH_STAT_RSTAT        CEPH_CAP_FILE_WREXTEND
 737
 738 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED |                     \
 739                               CEPH_CAP_LINK_SHARED |                    \
 740                               CEPH_CAP_XATTR_SHARED |                   \
 741                               CEPH_CAP_FILE_SHARED)
 742 #define CEPH_CAP_ANY_RD   (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD |     \
 743                            CEPH_CAP_FILE_CACHE)
 744
 745 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL |         \
 746                            CEPH_CAP_LINK_EXCL |         \
 747                            CEPH_CAP_XATTR_EXCL |        \
 748                            CEPH_CAP_FILE_EXCL)
 749 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \
 750                               CEPH_CAP_FILE_EXCL)
 751 #define CEPH_CAP_ANY_WR   (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
 752 #define CEPH_CAP_ANY      (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
 753                            CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
 754                            CEPH_CAP_PIN)
 755
 756 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
 757                         CEPH_LOCK_IXATTR)
 758
 759 int ceph_caps_for_mode(int mode);
 760
 761 enum {
 762         CEPH_CAP_OP_GRANT,         /* mds->client grant */
 763         CEPH_CAP_OP_REVOKE,        /* mds->client revoke */
 764         CEPH_CAP_OP_TRUNC,         /* mds->client trunc notify */
 765         CEPH_CAP_OP_EXPORT,        /* mds has exported the cap */
 766         CEPH_CAP_OP_IMPORT,        /* mds has imported the cap */
 767         CEPH_CAP_OP_UPDATE,        /* client->mds update */
 768         CEPH_CAP_OP_DROP,          /* client->mds drop cap bits */
 769         CEPH_CAP_OP_FLUSH,         /* client->mds cap writeback */
 770         CEPH_CAP_OP_FLUSH_ACK,     /* mds->client flushed */
 771         CEPH_CAP_OP_FLUSHSNAP,     /* client->mds flush snapped metadata */
 772         CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
 773         CEPH_CAP_OP_RELEASE,       /* client->mds release (clean) cap */
 774         CEPH_CAP_OP_RENEW,         /* client->mds renewal request */
 775 };
 776
 777 extern const char *ceph_cap_op_name(int op);
 778
 779 /* extra info for cap import/export */
 780 struct ceph_mds_cap_peer {
 781         __le64 cap_id;
 782         __le32 seq;
 783         __le32 mseq;
 784         __le32 mds;
 785         __u8   flags;
 786 } __attribute__ ((packed));
 787
 788 /*
 789  * caps message, used for capability callbacks, acks, requests, etc.
 790  */
 791 struct ceph_mds_caps_head {
 792         __le32 op;                  /* CEPH_CAP_OP_* */
 793         __le64 ino, realm;
 794         __le64 cap_id;
 795         __le32 seq, issue_seq;
 796         __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
 797         __le32 migrate_seq;
 798         __le64 snap_follows;
 799         __le32 snap_trace_len;
 800
 801         /* authlock */
 802         __le32 uid, gid, mode;
 803
 804         /* linklock */
 805         __le32 nlink;
 806
 807         /* xattrlock */
 808         __le32 xattr_len;
 809         __le64 xattr_version;
 810 } __attribute__ ((packed));
 811
 812 struct ceph_mds_caps_body_legacy {
 813         union {
 814                 /* all except export */
 815                 struct {
 816                         /* filelock */
 817                         __le64 size, max_size, truncate_size;
 818                         __le32 truncate_seq;
 819                         struct ceph_timespec mtime, atime, ctime;
 820                         struct ceph_file_layout layout;
 821                         __le32 time_warp_seq;
 822                 };
 823                 /* export message */
 824                 struct ceph_mds_cap_peer peer;
 825         };
 826 } __attribute__ ((packed));
 827
 828 /* cap release msg head */
 829 struct ceph_mds_cap_release {
 830         __le32 num;                /* number of cap_items that follow */
 831 } __attribute__ ((packed));
 832
 833 struct ceph_mds_cap_item {
 834         __le64 ino;
 835         __le64 cap_id;
 836         __le32 migrate_seq, seq;
 837 } __attribute__ ((packed));
 838
 839 #define CEPH_MDS_LEASE_REVOKE           1  /*    mds  -> client */
 840 #define CEPH_MDS_LEASE_RELEASE          2  /* client  -> mds    */
 841 #define CEPH_MDS_LEASE_RENEW            3  /* client <-> mds    */
 842 #define CEPH_MDS_LEASE_REVOKE_ACK       4  /* client  -> mds    */
 843
 844 extern const char *ceph_lease_op_name(int o);
 845
 846 /* lease msg header */
 847 struct ceph_mds_lease {
 848         __u8 action;            /* CEPH_MDS_LEASE_* */
 849         __le16 mask;            /* which lease */
 850         __le64 ino;
 851         __le64 first, last;     /* snap range */
 852         __le32 seq;
 853         __le32 duration_ms;     /* duration of renewal */
 854 } __attribute__ ((packed));
 855 /* followed by a __le32+string for dname */
 856
 857 /* client reconnect */
 858 struct ceph_mds_cap_reconnect {
 859         __le64 cap_id;
 860         __le32 wanted;
 861         __le32 issued;
 862         __le64 snaprealm;
 863         __le64 pathbase;        /* base ino for our path to this ino */
 864         __le32 flock_len;       /* size of flock state blob, if any */
 865 } __attribute__ ((packed));
 866 /* followed by flock blob */
 867
 868 struct ceph_mds_cap_reconnect_v1 {
 869         __le64 cap_id;
 870         __le32 wanted;
 871         __le32 issued;
 872         __le64 size;
 873         struct ceph_timespec mtime, atime;
 874         __le64 snaprealm;
 875         __le64 pathbase;        /* base ino for our path to this ino */
 876 } __attribute__ ((packed));
 877
 878 struct ceph_mds_snaprealm_reconnect {
 879         __le64 ino;     /* snap realm base */
 880         __le64 seq;     /* snap seq for this snap realm */
 881         __le64 parent;  /* parent realm */
 882 } __attribute__ ((packed));
 883
 884 /*
 885  * snaps
 886  */
 887 enum {
 888         CEPH_SNAP_OP_UPDATE,  /* CREATE or DESTROY */
 889         CEPH_SNAP_OP_CREATE,
 890         CEPH_SNAP_OP_DESTROY,
 891         CEPH_SNAP_OP_SPLIT,
 892 };
 893
 894 extern const char *ceph_snap_op_name(int o);
 895
 896 /* snap msg header */
 897 struct ceph_mds_snap_head {
 898         __le32 op;                /* CEPH_SNAP_OP_* */
 899         __le64 split;             /* ino to split off, if any */
 900         __le32 num_split_inos;    /* # inos belonging to new child realm */
 901         __le32 num_split_realms;  /* # child realms udner new child realm */
 902         __le32 trace_len;         /* size of snap trace blob */
 903 } __attribute__ ((packed));
 904 /* followed by split ino list, then split realms, then the trace blob */
 905
 906 /*
 907  * encode info about a snaprealm, as viewed by a client
 908  */
 909 struct ceph_mds_snap_realm {
 910         __le64 ino;           /* ino */
 911         __le64 created;       /* snap: when created */
 912         __le64 parent;        /* ino: parent realm */
 913         __le64 parent_since;  /* snap: same parent since */
 914         __le64 seq;           /* snap: version */
 915         __le32 num_snaps;
 916         __le32 num_prior_parent_snaps;
 917 } __attribute__ ((packed));
 918 /* followed by my snap list, then prior parent snap list */
 919
 920 #endif