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 /* mon commands */
 142 #define CEPH_MSG_MON_COMMAND            50
 143 #define CEPH_MSG_MON_COMMAND_ACK        51
 144
 145 /* osd */
 146 #define CEPH_MSG_OSD_MAP                41
 147 #define CEPH_MSG_OSD_OP                 42
 148 #define CEPH_MSG_OSD_OPREPLY            43
 149 #define CEPH_MSG_WATCH_NOTIFY           44
 150
 151
 152 /* watch-notify operations */
 153 enum {
 154         CEPH_WATCH_EVENT_NOTIFY           = 1, /* notifying watcher */
 155         CEPH_WATCH_EVENT_NOTIFY_COMPLETE  = 2, /* notifier notified when done */
 156         CEPH_WATCH_EVENT_DISCONNECT       = 3, /* we were disconnected */
 157 };
 158
 159
 160 struct ceph_mon_request_header {
 161         __le64 have_version;
 162         __le16 session_mon;
 163         __le64 session_mon_tid;
 164 } __attribute__ ((packed));
 165
 166 struct ceph_mon_statfs {
 167         struct ceph_mon_request_header monhdr;
 168         struct ceph_fsid fsid;
 169 } __attribute__ ((packed));
 170
 171 struct ceph_statfs {
 172         __le64 kb, kb_used, kb_avail;
 173         __le64 num_objects;
 174 } __attribute__ ((packed));
 175
 176 struct ceph_mon_statfs_reply {
 177         struct ceph_fsid fsid;
 178         __le64 version;
 179         struct ceph_statfs st;
 180 } __attribute__ ((packed));
 181
 182 struct ceph_mon_command {
 183         struct ceph_mon_request_header monhdr;
 184         struct ceph_fsid fsid;
 185         __le32 num_strs;         /* always 1 */
 186         __le32 str_len;
 187         char str[];
 188 } __attribute__ ((packed));
 189
 190 struct ceph_osd_getmap {
 191         struct ceph_mon_request_header monhdr;
 192         struct ceph_fsid fsid;
 193         __le32 start;
 194 } __attribute__ ((packed));
 195
 196 struct ceph_mds_getmap {
 197         struct ceph_mon_request_header monhdr;
 198         struct ceph_fsid fsid;
 199 } __attribute__ ((packed));
 200
 201 struct ceph_client_mount {
 202         struct ceph_mon_request_header monhdr;
 203 } __attribute__ ((packed));
 204
 205 #define CEPH_SUBSCRIBE_ONETIME    1  /* i want only 1 update after have */
 206
 207 struct ceph_mon_subscribe_item {
 208         __le64 start;
 209         __u8 flags;
 210 } __attribute__ ((packed));
 211
 212 struct ceph_mon_subscribe_ack {
 213         __le32 duration;         /* seconds */
 214         struct ceph_fsid fsid;
 215 } __attribute__ ((packed));
 216
 217 #define CEPH_FS_CLUSTER_ID_NONE  -1
 218
 219 /*
 220  * mdsmap flags
 221  */
 222 #define CEPH_MDSMAP_DOWN    (1<<0)  /* cluster deliberately down */
 223
 224 /*
 225  * mds states
 226  *   > 0 -> in
 227  *  <= 0 -> out
 228  */
 229 #define CEPH_MDS_STATE_DNE          0  /* down, does not exist. */
 230 #define CEPH_MDS_STATE_STOPPED     -1  /* down, once existed, but no subtrees.
 231                                           empty log. */
 232 #define CEPH_MDS_STATE_BOOT        -4  /* up, boot announcement. */
 233 #define CEPH_MDS_STATE_STANDBY     -5  /* up, idle.  waiting for assignment. */
 234 #define CEPH_MDS_STATE_CREATING    -6  /* up, creating MDS instance. */
 235 #define CEPH_MDS_STATE_STARTING    -7  /* up, starting previously stopped mds */
 236 #define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
 237 #define CEPH_MDS_STATE_REPLAYONCE   -9 /* up, replaying an active node's journal */
 238
 239 #define CEPH_MDS_STATE_REPLAY       8  /* up, replaying journal. */
 240 #define CEPH_MDS_STATE_RESOLVE      9  /* up, disambiguating distributed
 241                                           operations (import, rename, etc.) */
 242 #define CEPH_MDS_STATE_RECONNECT    10 /* up, reconnect to clients */
 243 #define CEPH_MDS_STATE_REJOIN       11 /* up, rejoining distributed cache */
 244 #define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */
 245 #define CEPH_MDS_STATE_ACTIVE       13 /* up, active */
 246 #define CEPH_MDS_STATE_STOPPING     14 /* up, but exporting metadata */
 247
 248 extern const char *ceph_mds_state_name(int s);
 249
 250
 251 /*
 252  * metadata lock types.
 253  *  - these are bitmasks.. we can compose them
 254  *  - they also define the lock ordering by the MDS
 255  *  - a few of these are internal to the mds
 256  */
 257 #define CEPH_LOCK_DVERSION    1
 258 #define CEPH_LOCK_DN          2
 259 #define CEPH_LOCK_ISNAP       16
 260 #define CEPH_LOCK_IVERSION    32    /* mds internal */
 261 #define CEPH_LOCK_IFILE       64
 262 #define CEPH_LOCK_IAUTH       128
 263 #define CEPH_LOCK_ILINK       256
 264 #define CEPH_LOCK_IDFT        512   /* dir frag tree */
 265 #define CEPH_LOCK_INEST       1024  /* mds internal */
 266 #define CEPH_LOCK_IXATTR      2048
 267 #define CEPH_LOCK_IFLOCK      4096  /* advisory file locks */
 268 #define CEPH_LOCK_INO         8192  /* immutable inode bits; not a lock */
 269 #define CEPH_LOCK_IPOLICY     16384 /* policy lock on dirs. MDS internal */
 270
 271 /* client_session ops */
 272 enum {
 273         CEPH_SESSION_REQUEST_OPEN,
 274         CEPH_SESSION_OPEN,
 275         CEPH_SESSION_REQUEST_CLOSE,
 276         CEPH_SESSION_CLOSE,
 277         CEPH_SESSION_REQUEST_RENEWCAPS,
 278         CEPH_SESSION_RENEWCAPS,
 279         CEPH_SESSION_STALE,
 280         CEPH_SESSION_RECALL_STATE,
 281         CEPH_SESSION_FLUSHMSG,
 282         CEPH_SESSION_FLUSHMSG_ACK,
 283         CEPH_SESSION_FORCE_RO,
 284 };
 285
 286 extern const char *ceph_session_op_name(int op);
 287
 288 struct ceph_mds_session_head {
 289         __le32 op;
 290         __le64 seq;
 291         struct ceph_timespec stamp;
 292         __le32 max_caps, max_leases;
 293 } __attribute__ ((packed));
 294
 295 /* client_request */
 296 /*
 297  * metadata ops.
 298  *  & 0x001000 -> write op
 299  *  & 0x010000 -> follow symlink (e.g. stat(), not lstat()).
 300  &  & 0x100000 -> use weird ino/path trace
 301  */
 302 #define CEPH_MDS_OP_WRITE        0x001000
 303 enum {
 304         CEPH_MDS_OP_LOOKUP     = 0x00100,
 305         CEPH_MDS_OP_GETATTR    = 0x00101,
 306         CEPH_MDS_OP_LOOKUPHASH = 0x00102,
 307         CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
 308         CEPH_MDS_OP_LOOKUPINO  = 0x00104,
 309         CEPH_MDS_OP_LOOKUPNAME = 0x00105,
 310
 311         CEPH_MDS_OP_SETXATTR   = 0x01105,
 312         CEPH_MDS_OP_RMXATTR    = 0x01106,
 313         CEPH_MDS_OP_SETLAYOUT  = 0x01107,
 314         CEPH_MDS_OP_SETATTR    = 0x01108,
 315         CEPH_MDS_OP_SETFILELOCK= 0x01109,
 316         CEPH_MDS_OP_GETFILELOCK= 0x00110,
 317         CEPH_MDS_OP_SETDIRLAYOUT=0x0110a,
 318
 319         CEPH_MDS_OP_MKNOD      = 0x01201,
 320         CEPH_MDS_OP_LINK       = 0x01202,
 321         CEPH_MDS_OP_UNLINK     = 0x01203,
 322         CEPH_MDS_OP_RENAME     = 0x01204,
 323         CEPH_MDS_OP_MKDIR      = 0x01220,
 324         CEPH_MDS_OP_RMDIR      = 0x01221,
 325         CEPH_MDS_OP_SYMLINK    = 0x01222,
 326
 327         CEPH_MDS_OP_CREATE     = 0x01301,
 328         CEPH_MDS_OP_OPEN       = 0x00302,
 329         CEPH_MDS_OP_READDIR    = 0x00305,
 330
 331         CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
 332         CEPH_MDS_OP_MKSNAP     = 0x01400,
 333         CEPH_MDS_OP_RMSNAP     = 0x01401,
 334         CEPH_MDS_OP_LSSNAP     = 0x00402,
 335         CEPH_MDS_OP_RENAMESNAP = 0x01403,
 336 };
 337
 338 extern const char *ceph_mds_op_name(int op);
 339
 340
 341 #define CEPH_SETATTR_MODE   1
 342 #define CEPH_SETATTR_UID    2
 343 #define CEPH_SETATTR_GID    4
 344 #define CEPH_SETATTR_MTIME  8
 345 #define CEPH_SETATTR_ATIME 16
 346 #define CEPH_SETATTR_SIZE  32
 347 #define CEPH_SETATTR_CTIME 64
 348
 349 /*
 350  * Ceph setxattr request flags.
 351  */
 352 #define CEPH_XATTR_CREATE  (1 << 0)
 353 #define CEPH_XATTR_REPLACE (1 << 1)
 354 #define CEPH_XATTR_REMOVE  (1 << 31)
 355
 356 /*
 357  * readdir request flags;
 358  */
 359 #define CEPH_READDIR_REPLY_BITFLAGS     (1<<0)
 360
 361 /*
 362  * readdir reply flags.
 363  */
 364 #define CEPH_READDIR_FRAG_END           (1<<0)
 365 #define CEPH_READDIR_FRAG_COMPLETE      (1<<8)
 366 #define CEPH_READDIR_HASH_ORDER         (1<<9)
 367
 368 union ceph_mds_request_args {
 369         struct {
 370                 __le32 mask;                 /* CEPH_CAP_* */
 371         } __attribute__ ((packed)) getattr;
 372         struct {
 373                 __le32 mode;
 374                 __le32 uid;
 375                 __le32 gid;
 376                 struct ceph_timespec mtime;
 377                 struct ceph_timespec atime;
 378                 __le64 size, old_size;       /* old_size needed by truncate */
 379                 __le32 mask;                 /* CEPH_SETATTR_* */
 380         } __attribute__ ((packed)) setattr;
 381         struct {
 382                 __le32 frag;                 /* which dir fragment */
 383                 __le32 max_entries;          /* how many dentries to grab */
 384                 __le32 max_bytes;
 385                 __le16 flags;
 386         } __attribute__ ((packed)) readdir;
 387         struct {
 388                 __le32 mode;
 389                 __le32 rdev;
 390         } __attribute__ ((packed)) mknod;
 391         struct {
 392                 __le32 mode;
 393         } __attribute__ ((packed)) mkdir;
 394         struct {
 395                 __le32 flags;
 396                 __le32 mode;
 397                 __le32 stripe_unit;          /* layout for newly created file */
 398                 __le32 stripe_count;         /* ... */
 399                 __le32 object_size;
 400                 __le32 file_replication;
 401                __le32 mask;                 /* CEPH_CAP_* */
 402                __le32 old_size;
 403         } __attribute__ ((packed)) open;
 404         struct {
 405                 __le32 flags;
 406         } __attribute__ ((packed)) setxattr;
 407         struct {
 408                 struct ceph_file_layout_legacy layout;
 409         } __attribute__ ((packed)) setlayout;
 410         struct {
 411                 __u8 rule; /* currently fcntl or flock */
 412                 __u8 type; /* shared, exclusive, remove*/
 413                 __le64 owner; /* owner of the lock */
 414                 __le64 pid; /* process id requesting the lock */
 415                 __le64 start; /* initial location to lock */
 416                 __le64 length; /* num bytes to lock from start */
 417                 __u8 wait; /* will caller wait for lock to become available? */
 418         } __attribute__ ((packed)) filelock_change;
 419 } __attribute__ ((packed));
 420
 421 #define CEPH_MDS_FLAG_REPLAY        1  /* this is a replayed op */
 422 #define CEPH_MDS_FLAG_WANT_DENTRY   2  /* want dentry in reply */
 423
 424 struct ceph_mds_request_head {
 425         __le64 oldest_client_tid;
 426         __le32 mdsmap_epoch;           /* on client */
 427         __le32 flags;                  /* CEPH_MDS_FLAG_* */
 428         __u8 num_retry, num_fwd;       /* count retry, fwd attempts */
 429         __le16 num_releases;           /* # include cap/lease release records */
 430         __le32 op;                     /* mds op code */
 431         __le32 caller_uid, caller_gid;
 432         __le64 ino;                    /* use this ino for openc, mkdir, mknod,
 433                                           etc. (if replaying) */
 434         union ceph_mds_request_args args;
 435 } __attribute__ ((packed));
 436
 437 /* cap/lease release record */
 438 struct ceph_mds_request_release {
 439         __le64 ino, cap_id;            /* ino and unique cap id */
 440         __le32 caps, wanted;           /* new issued, wanted */
 441         __le32 seq, issue_seq, mseq;
 442         __le32 dname_seq;              /* if releasing a dentry lease, a */
 443         __le32 dname_len;              /* string follows. */
 444 } __attribute__ ((packed));
 445
 446 /* client reply */
 447 struct ceph_mds_reply_head {
 448         __le32 op;
 449         __le32 result;
 450         __le32 mdsmap_epoch;
 451         __u8 safe;                     /* true if committed to disk */
 452         __u8 is_dentry, is_target;     /* true if dentry, target inode records
 453                                           are included with reply */
 454 } __attribute__ ((packed));
 455
 456 /* one for each node split */
 457 struct ceph_frag_tree_split {
 458         __le32 frag;                   /* this frag splits... */
 459         __le32 by;                     /* ...by this many bits */
 460 } __attribute__ ((packed));
 461
 462 struct ceph_frag_tree_head {
 463         __le32 nsplits;                /* num ceph_frag_tree_split records */
 464         struct ceph_frag_tree_split splits[];
 465 } __attribute__ ((packed));
 466
 467 /* capability issue, for bundling with mds reply */
 468 struct ceph_mds_reply_cap {
 469         __le32 caps, wanted;           /* caps issued, wanted */
 470         __le64 cap_id;
 471         __le32 seq, mseq;
 472         __le64 realm;                  /* snap realm */
 473         __u8 flags;                    /* CEPH_CAP_FLAG_* */
 474 } __attribute__ ((packed));
 475
 476 #define CEPH_CAP_FLAG_AUTH      (1 << 0)  /* cap is issued by auth mds */
 477 #define CEPH_CAP_FLAG_RELEASE   (1 << 1)  /* release the cap */
 478
 479 /* inode record, for bundling with mds reply */
 480 struct ceph_mds_reply_inode {
 481         __le64 ino;
 482         __le64 snapid;
 483         __le32 rdev;
 484         __le64 version;                /* inode version */
 485         __le64 xattr_version;          /* version for xattr blob */
 486         struct ceph_mds_reply_cap cap; /* caps issued for this inode */
 487         struct ceph_file_layout_legacy layout;
 488         struct ceph_timespec ctime, mtime, atime;
 489         __le32 time_warp_seq;
 490         __le64 size, max_size, truncate_size;
 491         __le32 truncate_seq;
 492         __le32 mode, uid, gid;
 493         __le32 nlink;
 494         __le64 files, subdirs, rbytes, rfiles, rsubdirs;  /* dir stats */
 495         struct ceph_timespec rctime;
 496         struct ceph_frag_tree_head fragtree;  /* (must be at end of struct) */
 497 } __attribute__ ((packed));
 498 /* followed by frag array, symlink string, dir layout, xattr blob */
 499
 500 /* reply_lease follows dname, and reply_inode */
 501 struct ceph_mds_reply_lease {
 502         __le16 mask;            /* lease type(s) */
 503         __le32 duration_ms;     /* lease duration */
 504         __le32 seq;
 505 } __attribute__ ((packed));
 506
 507 struct ceph_mds_reply_dirfrag {
 508         __le32 frag;            /* fragment */
 509         __le32 auth;            /* auth mds, if this is a delegation point */
 510         __le32 ndist;           /* number of mds' this is replicated on */
 511         __le32 dist[];
 512 } __attribute__ ((packed));
 513
 514 #define CEPH_LOCK_FCNTL         1
 515 #define CEPH_LOCK_FLOCK         2
 516 #define CEPH_LOCK_FCNTL_INTR    3
 517 #define CEPH_LOCK_FLOCK_INTR    4
 518
 519
 520 #define CEPH_LOCK_SHARED   1
 521 #define CEPH_LOCK_EXCL     2
 522 #define CEPH_LOCK_UNLOCK   4
 523
 524 struct ceph_filelock {
 525         __le64 start;/* file offset to start lock at */
 526         __le64 length; /* num bytes to lock; 0 for all following start */
 527         __le64 client; /* which client holds the lock */
 528         __le64 owner; /* owner the lock */
 529         __le64 pid; /* process id holding the lock on the client */
 530         __u8 type; /* shared lock, exclusive lock, or unlock */
 531 } __attribute__ ((packed));
 532
 533
 534 /* file access modes */
 535 #define CEPH_FILE_MODE_PIN        0
 536 #define CEPH_FILE_MODE_RD         1
 537 #define CEPH_FILE_MODE_WR         2
 538 #define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
 539 #define CEPH_FILE_MODE_LAZY       4  /* lazy io */
 540 #define CEPH_FILE_MODE_BITS       4
 541
 542 int ceph_flags_to_mode(int flags);
 543
 544 #define CEPH_INLINE_NONE        ((__u64)-1)
 545
 546 /* capability bits */
 547 #define CEPH_CAP_PIN         1  /* no specific capabilities beyond the pin */
 548
 549 /* generic cap bits */
 550 #define CEPH_CAP_GSHARED     1  /* client can reads */
 551 #define CEPH_CAP_GEXCL       2  /* client can read and update */
 552 #define CEPH_CAP_GCACHE      4  /* (file) client can cache reads */
 553 #define CEPH_CAP_GRD         8  /* (file) client can read */
 554 #define CEPH_CAP_GWR        16  /* (file) client can write */
 555 #define CEPH_CAP_GBUFFER    32  /* (file) client can buffer writes */
 556 #define CEPH_CAP_GWREXTEND  64  /* (file) client can extend EOF */
 557 #define CEPH_CAP_GLAZYIO   128  /* (file) client can perform lazy io */
 558
 559 #define CEPH_CAP_SIMPLE_BITS  2
 560 #define CEPH_CAP_FILE_BITS    8
 561
 562 /* per-lock shift */
 563 #define CEPH_CAP_SAUTH      2
 564 #define CEPH_CAP_SLINK      4
 565 #define CEPH_CAP_SXATTR     6
 566 #define CEPH_CAP_SFILE      8
 567 #define CEPH_CAP_SFLOCK    20
 568
 569 #define CEPH_CAP_BITS      22
 570
 571 /* composed values */
 572 #define CEPH_CAP_AUTH_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SAUTH)
 573 #define CEPH_CAP_AUTH_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SAUTH)
 574 #define CEPH_CAP_LINK_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SLINK)
 575 #define CEPH_CAP_LINK_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SLINK)
 576 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED  << CEPH_CAP_SXATTR)
 577 #define CEPH_CAP_XATTR_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SXATTR)
 578 #define CEPH_CAP_FILE(x)    (x << CEPH_CAP_SFILE)
 579 #define CEPH_CAP_FILE_SHARED   (CEPH_CAP_GSHARED   << CEPH_CAP_SFILE)
 580 #define CEPH_CAP_FILE_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SFILE)
 581 #define CEPH_CAP_FILE_CACHE    (CEPH_CAP_GCACHE    << CEPH_CAP_SFILE)
 582 #define CEPH_CAP_FILE_RD       (CEPH_CAP_GRD       << CEPH_CAP_SFILE)
 583 #define CEPH_CAP_FILE_WR       (CEPH_CAP_GWR       << CEPH_CAP_SFILE)
 584 #define CEPH_CAP_FILE_BUFFER   (CEPH_CAP_GBUFFER   << CEPH_CAP_SFILE)
 585 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
 586 #define CEPH_CAP_FILE_LAZYIO   (CEPH_CAP_GLAZYIO   << CEPH_CAP_SFILE)
 587 #define CEPH_CAP_FLOCK_SHARED  (CEPH_CAP_GSHARED   << CEPH_CAP_SFLOCK)
 588 #define CEPH_CAP_FLOCK_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SFLOCK)
 589
 590
 591 /* cap masks (for getattr) */
 592 #define CEPH_STAT_CAP_INODE    CEPH_CAP_PIN
 593 #define CEPH_STAT_CAP_TYPE     CEPH_CAP_PIN  /* mode >> 12 */
 594 #define CEPH_STAT_CAP_SYMLINK  CEPH_CAP_PIN
 595 #define CEPH_STAT_CAP_UID      CEPH_CAP_AUTH_SHARED
 596 #define CEPH_STAT_CAP_GID      CEPH_CAP_AUTH_SHARED
 597 #define CEPH_STAT_CAP_MODE     CEPH_CAP_AUTH_SHARED
 598 #define CEPH_STAT_CAP_NLINK    CEPH_CAP_LINK_SHARED
 599 #define CEPH_STAT_CAP_LAYOUT   CEPH_CAP_FILE_SHARED
 600 #define CEPH_STAT_CAP_MTIME    CEPH_CAP_FILE_SHARED
 601 #define CEPH_STAT_CAP_SIZE     CEPH_CAP_FILE_SHARED
 602 #define CEPH_STAT_CAP_ATIME    CEPH_CAP_FILE_SHARED  /* fixme */
 603 #define CEPH_STAT_CAP_XATTR    CEPH_CAP_XATTR_SHARED
 604 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN |                 \
 605                                  CEPH_CAP_AUTH_SHARED | \
 606                                  CEPH_CAP_LINK_SHARED | \
 607                                  CEPH_CAP_FILE_SHARED | \
 608                                  CEPH_CAP_XATTR_SHARED)
 609 #define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \
 610                                    CEPH_CAP_FILE_RD)
 611
 612 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED |                     \
 613                               CEPH_CAP_LINK_SHARED |                    \
 614                               CEPH_CAP_XATTR_SHARED |                   \
 615                               CEPH_CAP_FILE_SHARED)
 616 #define CEPH_CAP_ANY_RD   (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD |     \
 617                            CEPH_CAP_FILE_CACHE)
 618
 619 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL |         \
 620                            CEPH_CAP_LINK_EXCL |         \
 621                            CEPH_CAP_XATTR_EXCL |        \
 622                            CEPH_CAP_FILE_EXCL)
 623 #define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \
 624                               CEPH_CAP_FILE_SHARED)
 625 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \
 626                               CEPH_CAP_FILE_EXCL)
 627 #define CEPH_CAP_ANY_WR   (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
 628 #define CEPH_CAP_ANY      (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
 629                            CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
 630                            CEPH_CAP_PIN)
 631
 632 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
 633                         CEPH_LOCK_IXATTR)
 634
 635 int ceph_caps_for_mode(int mode);
 636
 637 enum {
 638         CEPH_CAP_OP_GRANT,         /* mds->client grant */
 639         CEPH_CAP_OP_REVOKE,        /* mds->client revoke */
 640         CEPH_CAP_OP_TRUNC,         /* mds->client trunc notify */
 641         CEPH_CAP_OP_EXPORT,        /* mds has exported the cap */
 642         CEPH_CAP_OP_IMPORT,        /* mds has imported the cap */
 643         CEPH_CAP_OP_UPDATE,        /* client->mds update */
 644         CEPH_CAP_OP_DROP,          /* client->mds drop cap bits */
 645         CEPH_CAP_OP_FLUSH,         /* client->mds cap writeback */
 646         CEPH_CAP_OP_FLUSH_ACK,     /* mds->client flushed */
 647         CEPH_CAP_OP_FLUSHSNAP,     /* client->mds flush snapped metadata */
 648         CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
 649         CEPH_CAP_OP_RELEASE,       /* client->mds release (clean) cap */
 650         CEPH_CAP_OP_RENEW,         /* client->mds renewal request */
 651 };
 652
 653 extern const char *ceph_cap_op_name(int op);
 654
 655 /*
 656  * caps message, used for capability callbacks, acks, requests, etc.
 657  */
 658 struct ceph_mds_caps {
 659         __le32 op;                  /* CEPH_CAP_OP_* */
 660         __le64 ino, realm;
 661         __le64 cap_id;
 662         __le32 seq, issue_seq;
 663         __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
 664         __le32 migrate_seq;
 665         __le64 snap_follows;
 666         __le32 snap_trace_len;
 667
 668         /* authlock */
 669         __le32 uid, gid, mode;
 670
 671         /* linklock */
 672         __le32 nlink;
 673
 674         /* xattrlock */
 675         __le32 xattr_len;
 676         __le64 xattr_version;
 677
 678         /* filelock */
 679         __le64 size, max_size, truncate_size;
 680         __le32 truncate_seq;
 681         struct ceph_timespec mtime, atime, ctime;
 682         struct ceph_file_layout_legacy layout;
 683         __le32 time_warp_seq;
 684 } __attribute__ ((packed));
 685
 686 struct ceph_mds_cap_peer {
 687         __le64 cap_id;
 688         __le32 seq;
 689         __le32 mseq;
 690         __le32 mds;
 691         __u8   flags;
 692 } __attribute__ ((packed));
 693
 694 /* cap release msg head */
 695 struct ceph_mds_cap_release {
 696         __le32 num;                /* number of cap_items that follow */
 697 } __attribute__ ((packed));
 698
 699 struct ceph_mds_cap_item {
 700         __le64 ino;
 701         __le64 cap_id;
 702         __le32 migrate_seq, seq;
 703 } __attribute__ ((packed));
 704
 705 #define CEPH_MDS_LEASE_REVOKE           1  /*    mds  -> client */
 706 #define CEPH_MDS_LEASE_RELEASE          2  /* client  -> mds    */
 707 #define CEPH_MDS_LEASE_RENEW            3  /* client <-> mds    */
 708 #define CEPH_MDS_LEASE_REVOKE_ACK       4  /* client  -> mds    */
 709
 710 extern const char *ceph_lease_op_name(int o);
 711
 712 /* lease msg header */
 713 struct ceph_mds_lease {
 714         __u8 action;            /* CEPH_MDS_LEASE_* */
 715         __le16 mask;            /* which lease */
 716         __le64 ino;
 717         __le64 first, last;     /* snap range */
 718         __le32 seq;
 719         __le32 duration_ms;     /* duration of renewal */
 720 } __attribute__ ((packed));
 721 /* followed by a __le32+string for dname */
 722
 723 /* client reconnect */
 724 struct ceph_mds_cap_reconnect {
 725         __le64 cap_id;
 726         __le32 wanted;
 727         __le32 issued;
 728         __le64 snaprealm;
 729         __le64 pathbase;        /* base ino for our path to this ino */
 730         __le32 flock_len;       /* size of flock state blob, if any */
 731 } __attribute__ ((packed));
 732 /* followed by flock blob */
 733
 734 struct ceph_mds_cap_reconnect_v1 {
 735         __le64 cap_id;
 736         __le32 wanted;
 737         __le32 issued;
 738         __le64 size;
 739         struct ceph_timespec mtime, atime;
 740         __le64 snaprealm;
 741         __le64 pathbase;        /* base ino for our path to this ino */
 742 } __attribute__ ((packed));
 743
 744 struct ceph_mds_snaprealm_reconnect {
 745         __le64 ino;     /* snap realm base */
 746         __le64 seq;     /* snap seq for this snap realm */
 747         __le64 parent;  /* parent realm */
 748 } __attribute__ ((packed));
 749
 750 /*
 751  * snaps
 752  */
 753 enum {
 754         CEPH_SNAP_OP_UPDATE,  /* CREATE or DESTROY */
 755         CEPH_SNAP_OP_CREATE,
 756         CEPH_SNAP_OP_DESTROY,
 757         CEPH_SNAP_OP_SPLIT,
 758 };
 759
 760 extern const char *ceph_snap_op_name(int o);
 761
 762 /* snap msg header */
 763 struct ceph_mds_snap_head {
 764         __le32 op;                /* CEPH_SNAP_OP_* */
 765         __le64 split;             /* ino to split off, if any */
 766         __le32 num_split_inos;    /* # inos belonging to new child realm */
 767         __le32 num_split_realms;  /* # child realms udner new child realm */
 768         __le32 trace_len;         /* size of snap trace blob */
 769 } __attribute__ ((packed));
 770 /* followed by split ino list, then split realms, then the trace blob */
 771
 772 /*
 773  * encode info about a snaprealm, as viewed by a client
 774  */
 775 struct ceph_mds_snap_realm {
 776         __le64 ino;           /* ino */
 777         __le64 created;       /* snap: when created */
 778         __le64 parent;        /* ino: parent realm */
 779         __le64 parent_since;  /* snap: same parent since */
 780         __le64 seq;           /* snap: version */
 781         __le32 num_snaps;
 782         __le32 num_prior_parent_snaps;
 783 } __attribute__ ((packed));
 784 /* followed by my snap list, then prior parent snap list */
 785
 786 #endif