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 "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_OSD_PROTOCOL     8 /* cluster internal */
  25 #define CEPH_MDS_PROTOCOL    12 /* cluster internal */
  26 #define CEPH_MON_PROTOCOL     5 /* cluster internal */
  27 #define CEPH_OSDC_PROTOCOL   24 /* server/client */
  28 #define CEPH_MDSC_PROTOCOL   32 /* server/client */
  29 #define CEPH_MONC_PROTOCOL   15 /* server/client */
  30
  31
  32 #define CEPH_INO_ROOT  1
  33 #define CEPH_INO_CEPH  2        /* hidden .ceph dir */
  34
  35 /* arbitrary limit on max # of monitors (cluster of 3 is typical) */
  36 #define CEPH_MAX_MON   31
  37
  38
  39 /*
  40  * feature bits
  41  */
  42 #define CEPH_FEATURE_UID            (1<<0)
  43 #define CEPH_FEATURE_NOSRCADDR      (1<<1)
  44 #define CEPH_FEATURE_MONCLOCKCHECK  (1<<2)
  45 #define CEPH_FEATURE_FLOCK          (1<<3)
  46 #define CEPH_FEATURE_SUBSCRIBE2     (1<<4)
  47 #define CEPH_FEATURE_MONNAMES       (1<<5)
  48 #define CEPH_FEATURE_RECONNECT_SEQ  (1<<6)
  49 #define CEPH_FEATURE_DIRLAYOUTHASH  (1<<7)
  50
  51
  52 /*
  53  * ceph_file_layout - describe data layout for a file/inode
  54  */
  55 struct ceph_file_layout {
  56         /* file -> object mapping */
  57         __le32 fl_stripe_unit;     /* stripe unit, in bytes.  must be multiple
  58                                       of page size. */
  59         __le32 fl_stripe_count;    /* over this many objects */
  60         __le32 fl_object_size;     /* until objects are this big, then move to
  61                                       new objects */
  62         __le32 fl_cas_hash;        /* UNUSED.  0 = none; 1 = sha256 */
  63
  64         /* pg -> disk layout */
  65         __le32 fl_object_stripe_unit;  /* UNUSED.  for per-object parity, if any */
  66
  67         /* object -> pg layout */
  68         __le32 fl_pg_preferred; /* preferred primary for pg (-1 for none) */
  69         __le32 fl_pg_pool;      /* namespace, crush ruleset, rep level */
  70 } __attribute__ ((packed));
  71
  72 #define CEPH_MIN_STRIPE_UNIT 65536
  73
  74 int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
  75
  76 struct ceph_dir_layout {
  77         __u8   dl_dir_hash;   /* see ceph_hash.h for ids */
  78         __u8   dl_unused1;
  79         __u16  dl_unused2;
  80         __u32  dl_unused3;
  81 } __attribute__ ((packed));
  82
  83 /* crypto algorithms */
  84 #define CEPH_CRYPTO_NONE 0x0
  85 #define CEPH_CRYPTO_AES  0x1
  86
  87 #define CEPH_AES_IV "cephsageyudagreg"
  88
  89 /* security/authentication protocols */
  90 #define CEPH_AUTH_UNKNOWN       0x0
  91 #define CEPH_AUTH_NONE          0x1
  92 #define CEPH_AUTH_CEPHX         0x2
  93
  94 #define CEPH_AUTH_UID_DEFAULT ((__u64) -1)
  95
  96
  97 /*********************************************
  98  * message layer
  99  */
 100
 101 /*
 102  * message types
 103  */
 104
 105 /* misc */
 106 #define CEPH_MSG_SHUTDOWN               1
 107 #define CEPH_MSG_PING                   2
 108
 109 /* client <-> monitor */
 110 #define CEPH_MSG_MON_MAP                4
 111 #define CEPH_MSG_MON_GET_MAP            5
 112 #define CEPH_MSG_STATFS                 13
 113 #define CEPH_MSG_STATFS_REPLY           14
 114 #define CEPH_MSG_MON_SUBSCRIBE          15
 115 #define CEPH_MSG_MON_SUBSCRIBE_ACK      16
 116 #define CEPH_MSG_AUTH                   17
 117 #define CEPH_MSG_AUTH_REPLY             18
 118
 119 /* client <-> mds */
 120 #define CEPH_MSG_MDS_MAP                21
 121
 122 #define CEPH_MSG_CLIENT_SESSION         22
 123 #define CEPH_MSG_CLIENT_RECONNECT       23
 124
 125 #define CEPH_MSG_CLIENT_REQUEST         24
 126 #define CEPH_MSG_CLIENT_REQUEST_FORWARD 25
 127 #define CEPH_MSG_CLIENT_REPLY           26
 128 #define CEPH_MSG_CLIENT_CAPS            0x310
 129 #define CEPH_MSG_CLIENT_LEASE           0x311
 130 #define CEPH_MSG_CLIENT_SNAP            0x312
 131 #define CEPH_MSG_CLIENT_CAPRELEASE      0x313
 132
 133 /* pool ops */
 134 #define CEPH_MSG_POOLOP_REPLY           48
 135 #define CEPH_MSG_POOLOP                 49
 136
 137
 138 /* osd */
 139 #define CEPH_MSG_OSD_MAP          41
 140 #define CEPH_MSG_OSD_OP           42
 141 #define CEPH_MSG_OSD_OPREPLY      43
 142
 143 /* pool operations */
 144 enum {
 145   POOL_OP_CREATE                        = 0x01,
 146   POOL_OP_DELETE                        = 0x02,
 147   POOL_OP_AUID_CHANGE                   = 0x03,
 148   POOL_OP_CREATE_SNAP                   = 0x11,
 149   POOL_OP_DELETE_SNAP                   = 0x12,
 150   POOL_OP_CREATE_UNMANAGED_SNAP         = 0x21,
 151   POOL_OP_DELETE_UNMANAGED_SNAP         = 0x22,
 152 };
 153
 154 struct ceph_mon_request_header {
 155         __le64 have_version;
 156         __le16 session_mon;
 157         __le64 session_mon_tid;
 158 } __attribute__ ((packed));
 159
 160 struct ceph_mon_statfs {
 161         struct ceph_mon_request_header monhdr;
 162         struct ceph_fsid fsid;
 163 } __attribute__ ((packed));
 164
 165 struct ceph_statfs {
 166         __le64 kb, kb_used, kb_avail;
 167         __le64 num_objects;
 168 } __attribute__ ((packed));
 169
 170 struct ceph_mon_statfs_reply {
 171         struct ceph_fsid fsid;
 172         __le64 version;
 173         struct ceph_statfs st;
 174 } __attribute__ ((packed));
 175
 176 const char *ceph_pool_op_name(int op);
 177
 178 struct ceph_mon_poolop {
 179         struct ceph_mon_request_header monhdr;
 180         struct ceph_fsid fsid;
 181         __le32 pool;
 182         __le32 op;
 183         __le64 auid;
 184         __le64 snapid;
 185         __le32 name_len;
 186 } __attribute__ ((packed));
 187
 188 struct ceph_mon_poolop_reply {
 189         struct ceph_mon_request_header monhdr;
 190         struct ceph_fsid fsid;
 191         __le32 reply_code;
 192         __le32 epoch;
 193         char has_data;
 194         char data[0];
 195 } __attribute__ ((packed));
 196
 197 struct ceph_mon_unmanaged_snap {
 198         __le64 snapid;
 199 } __attribute__ ((packed));
 200
 201 struct ceph_osd_getmap {
 202         struct ceph_mon_request_header monhdr;
 203         struct ceph_fsid fsid;
 204         __le32 start;
 205 } __attribute__ ((packed));
 206
 207 struct ceph_mds_getmap {
 208         struct ceph_mon_request_header monhdr;
 209         struct ceph_fsid fsid;
 210 } __attribute__ ((packed));
 211
 212 struct ceph_client_mount {
 213         struct ceph_mon_request_header monhdr;
 214 } __attribute__ ((packed));
 215
 216 struct ceph_mon_subscribe_item {
 217         __le64 have_version;    __le64 have;
 218         __u8 onetime;
 219 } __attribute__ ((packed));
 220
 221 struct ceph_mon_subscribe_ack {
 222         __le32 duration;         /* seconds */
 223         struct ceph_fsid fsid;
 224 } __attribute__ ((packed));
 225
 226 /*
 227  * mds states
 228  *   > 0 -> in
 229  *  <= 0 -> out
 230  */
 231 #define CEPH_MDS_STATE_DNE          0  /* down, does not exist. */
 232 #define CEPH_MDS_STATE_STOPPED     -1  /* down, once existed, but no subtrees.
 233                                           empty log. */
 234 #define CEPH_MDS_STATE_BOOT        -4  /* up, boot announcement. */
 235 #define CEPH_MDS_STATE_STANDBY     -5  /* up, idle.  waiting for assignment. */
 236 #define CEPH_MDS_STATE_CREATING    -6  /* up, creating MDS instance. */
 237 #define CEPH_MDS_STATE_STARTING    -7  /* up, starting previously stopped mds */
 238 #define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
 239
 240 #define CEPH_MDS_STATE_REPLAY       8  /* up, replaying journal. */
 241 #define CEPH_MDS_STATE_RESOLVE      9  /* up, disambiguating distributed
 242                                           operations (import, rename, etc.) */
 243 #define CEPH_MDS_STATE_RECONNECT    10 /* up, reconnect to clients */
 244 #define CEPH_MDS_STATE_REJOIN       11 /* up, rejoining distributed cache */
 245 #define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */
 246 #define CEPH_MDS_STATE_ACTIVE       13 /* up, active */
 247 #define CEPH_MDS_STATE_STOPPING     14 /* up, but exporting metadata */
 248
 249 extern const char *ceph_mds_state_name(int s);
 250
 251
 252 /*
 253  * metadata lock types.
 254  *  - these are bitmasks.. we can compose them
 255  *  - they also define the lock ordering by the MDS
 256  *  - a few of these are internal to the mds
 257  */
 258 #define CEPH_LOCK_DVERSION    1
 259 #define CEPH_LOCK_DN          2
 260 #define CEPH_LOCK_ISNAP       16
 261 #define CEPH_LOCK_IVERSION    32    /* mds internal */
 262 #define CEPH_LOCK_IFILE       64
 263 #define CEPH_LOCK_IAUTH       128
 264 #define CEPH_LOCK_ILINK       256
 265 #define CEPH_LOCK_IDFT        512   /* dir frag tree */
 266 #define CEPH_LOCK_INEST       1024  /* mds internal */
 267 #define CEPH_LOCK_IXATTR      2048
 268 #define CEPH_LOCK_IFLOCK      4096  /* advisory file locks */
 269 #define CEPH_LOCK_INO         8192  /* immutable inode bits; not a lock */
 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 };
 282
 283 extern const char *ceph_session_op_name(int op);
 284
 285 struct ceph_mds_session_head {
 286         __le32 op;
 287         __le64 seq;
 288         struct ceph_timespec stamp;
 289         __le32 max_caps, max_leases;
 290 } __attribute__ ((packed));
 291
 292 /* client_request */
 293 /*
 294  * metadata ops.
 295  *  & 0x001000 -> write op
 296  *  & 0x010000 -> follow symlink (e.g. stat(), not lstat()).
 297  &  & 0x100000 -> use weird ino/path trace
 298  */
 299 #define CEPH_MDS_OP_WRITE        0x001000
 300 enum {
 301         CEPH_MDS_OP_LOOKUP     = 0x00100,
 302         CEPH_MDS_OP_GETATTR    = 0x00101,
 303         CEPH_MDS_OP_LOOKUPHASH = 0x00102,
 304         CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
 305
 306         CEPH_MDS_OP_SETXATTR   = 0x01105,
 307         CEPH_MDS_OP_RMXATTR    = 0x01106,
 308         CEPH_MDS_OP_SETLAYOUT  = 0x01107,
 309         CEPH_MDS_OP_SETATTR    = 0x01108,
 310         CEPH_MDS_OP_SETFILELOCK= 0x01109,
 311         CEPH_MDS_OP_GETFILELOCK= 0x00110,
 312         CEPH_MDS_OP_SETDIRLAYOUT=0x0110a,
 313
 314         CEPH_MDS_OP_MKNOD      = 0x01201,
 315         CEPH_MDS_OP_LINK       = 0x01202,
 316         CEPH_MDS_OP_UNLINK     = 0x01203,
 317         CEPH_MDS_OP_RENAME     = 0x01204,
 318         CEPH_MDS_OP_MKDIR      = 0x01220,
 319         CEPH_MDS_OP_RMDIR      = 0x01221,
 320         CEPH_MDS_OP_SYMLINK    = 0x01222,
 321
 322         CEPH_MDS_OP_CREATE     = 0x01301,
 323         CEPH_MDS_OP_OPEN       = 0x00302,
 324         CEPH_MDS_OP_READDIR    = 0x00305,
 325
 326         CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
 327         CEPH_MDS_OP_MKSNAP     = 0x01400,
 328         CEPH_MDS_OP_RMSNAP     = 0x01401,
 329         CEPH_MDS_OP_LSSNAP     = 0x00402,
 330 };
 331
 332 extern const char *ceph_mds_op_name(int op);
 333
 334
 335 #define CEPH_SETATTR_MODE   1
 336 #define CEPH_SETATTR_UID    2
 337 #define CEPH_SETATTR_GID    4
 338 #define CEPH_SETATTR_MTIME  8
 339 #define CEPH_SETATTR_ATIME 16
 340 #define CEPH_SETATTR_SIZE  32
 341 #define CEPH_SETATTR_CTIME 64
 342
 343 union ceph_mds_request_args {
 344         struct {
 345                 __le32 mask;                 /* CEPH_CAP_* */
 346         } __attribute__ ((packed)) getattr;
 347         struct {
 348                 __le32 mode;
 349                 __le32 uid;
 350                 __le32 gid;
 351                 struct ceph_timespec mtime;
 352                 struct ceph_timespec atime;
 353                 __le64 size, old_size;       /* old_size needed by truncate */
 354                 __le32 mask;                 /* CEPH_SETATTR_* */
 355         } __attribute__ ((packed)) setattr;
 356         struct {
 357                 __le32 frag;                 /* which dir fragment */
 358                 __le32 max_entries;          /* how many dentries to grab */
 359                 __le32 max_bytes;
 360         } __attribute__ ((packed)) readdir;
 361         struct {
 362                 __le32 mode;
 363                 __le32 rdev;
 364         } __attribute__ ((packed)) mknod;
 365         struct {
 366                 __le32 mode;
 367         } __attribute__ ((packed)) mkdir;
 368         struct {
 369                 __le32 flags;
 370                 __le32 mode;
 371                 __le32 stripe_unit;          /* layout for newly created file */
 372                 __le32 stripe_count;         /* ... */
 373                 __le32 object_size;
 374                 __le32 file_replication;
 375                 __le32 preferred;
 376         } __attribute__ ((packed)) open;
 377         struct {
 378                 __le32 flags;
 379         } __attribute__ ((packed)) setxattr;
 380         struct {
 381                 struct ceph_file_layout layout;
 382         } __attribute__ ((packed)) setlayout;
 383         struct {
 384                 __u8 rule; /* currently fcntl or flock */
 385                 __u8 type; /* shared, exclusive, remove*/
 386                 __le64 pid; /* process id requesting the lock */
 387                 __le64 pid_namespace;
 388                 __le64 start; /* initial location to lock */
 389                 __le64 length; /* num bytes to lock from start */
 390                 __u8 wait; /* will caller wait for lock to become available? */
 391         } __attribute__ ((packed)) filelock_change;
 392 } __attribute__ ((packed));
 393
 394 #define CEPH_MDS_FLAG_REPLAY        1  /* this is a replayed op */
 395 #define CEPH_MDS_FLAG_WANT_DENTRY   2  /* want dentry in reply */
 396
 397 struct ceph_mds_request_head {
 398         __le64 oldest_client_tid;
 399         __le32 mdsmap_epoch;           /* on client */
 400         __le32 flags;                  /* CEPH_MDS_FLAG_* */
 401         __u8 num_retry, num_fwd;       /* count retry, fwd attempts */
 402         __le16 num_releases;           /* # include cap/lease release records */
 403         __le32 op;                     /* mds op code */
 404         __le32 caller_uid, caller_gid;
 405         __le64 ino;                    /* use this ino for openc, mkdir, mknod,
 406                                           etc. (if replaying) */
 407         union ceph_mds_request_args args;
 408 } __attribute__ ((packed));
 409
 410 /* cap/lease release record */
 411 struct ceph_mds_request_release {
 412         __le64 ino, cap_id;            /* ino and unique cap id */
 413         __le32 caps, wanted;           /* new issued, wanted */
 414         __le32 seq, issue_seq, mseq;
 415         __le32 dname_seq;              /* if releasing a dentry lease, a */
 416         __le32 dname_len;              /* string follows. */
 417 } __attribute__ ((packed));
 418
 419 /* client reply */
 420 struct ceph_mds_reply_head {
 421         __le32 op;
 422         __le32 result;
 423         __le32 mdsmap_epoch;
 424         __u8 safe;                     /* true if committed to disk */
 425         __u8 is_dentry, is_target;     /* true if dentry, target inode records
 426                                           are included with reply */
 427 } __attribute__ ((packed));
 428
 429 /* one for each node split */
 430 struct ceph_frag_tree_split {
 431         __le32 frag;                   /* this frag splits... */
 432         __le32 by;                     /* ...by this many bits */
 433 } __attribute__ ((packed));
 434
 435 struct ceph_frag_tree_head {
 436         __le32 nsplits;                /* num ceph_frag_tree_split records */
 437         struct ceph_frag_tree_split splits[];
 438 } __attribute__ ((packed));
 439
 440 /* capability issue, for bundling with mds reply */
 441 struct ceph_mds_reply_cap {
 442         __le32 caps, wanted;           /* caps issued, wanted */
 443         __le64 cap_id;
 444         __le32 seq, mseq;
 445         __le64 realm;                  /* snap realm */
 446         __u8 flags;                    /* CEPH_CAP_FLAG_* */
 447 } __attribute__ ((packed));
 448
 449 #define CEPH_CAP_FLAG_AUTH  1          /* cap is issued by auth mds */
 450
 451 /* inode record, for bundling with mds reply */
 452 struct ceph_mds_reply_inode {
 453         __le64 ino;
 454         __le64 snapid;
 455         __le32 rdev;
 456         __le64 version;                /* inode version */
 457         __le64 xattr_version;          /* version for xattr blob */
 458         struct ceph_mds_reply_cap cap; /* caps issued for this inode */
 459         struct ceph_file_layout layout;
 460         struct ceph_timespec ctime, mtime, atime;
 461         __le32 time_warp_seq;
 462         __le64 size, max_size, truncate_size;
 463         __le32 truncate_seq;
 464         __le32 mode, uid, gid;
 465         __le32 nlink;
 466         __le64 files, subdirs, rbytes, rfiles, rsubdirs;  /* dir stats */
 467         struct ceph_timespec rctime;
 468         struct ceph_frag_tree_head fragtree;  /* (must be at end of struct) */
 469 } __attribute__ ((packed));
 470 /* followed by frag array, symlink string, dir layout, xattr blob */
 471
 472 /* reply_lease follows dname, and reply_inode */
 473 struct ceph_mds_reply_lease {
 474         __le16 mask;            /* lease type(s) */
 475         __le32 duration_ms;     /* lease duration */
 476         __le32 seq;
 477 } __attribute__ ((packed));
 478
 479 struct ceph_mds_reply_dirfrag {
 480         __le32 frag;            /* fragment */
 481         __le32 auth;            /* auth mds, if this is a delegation point */
 482         __le32 ndist;           /* number of mds' this is replicated on */
 483         __le32 dist[];
 484 } __attribute__ ((packed));
 485
 486 #define CEPH_LOCK_FCNTL    1
 487 #define CEPH_LOCK_FLOCK    2
 488
 489 #define CEPH_LOCK_SHARED   1
 490 #define CEPH_LOCK_EXCL     2
 491 #define CEPH_LOCK_UNLOCK   4
 492
 493 struct ceph_filelock {
 494         __le64 start;/* file offset to start lock at */
 495         __le64 length; /* num bytes to lock; 0 for all following start */
 496         __le64 client; /* which client holds the lock */
 497         __le64 pid; /* process id holding the lock on the client */
 498         __le64 pid_namespace;
 499         __u8 type; /* shared lock, exclusive lock, or unlock */
 500 } __attribute__ ((packed));
 501
 502
 503 /* file access modes */
 504 #define CEPH_FILE_MODE_PIN        0
 505 #define CEPH_FILE_MODE_RD         1
 506 #define CEPH_FILE_MODE_WR         2
 507 #define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
 508 #define CEPH_FILE_MODE_LAZY       4  /* lazy io */
 509 #define CEPH_FILE_MODE_NUM        8  /* bc these are bit fields.. mostly */
 510
 511 int ceph_flags_to_mode(int flags);
 512
 513
 514 /* capability bits */
 515 #define CEPH_CAP_PIN         1  /* no specific capabilities beyond the pin */
 516
 517 /* generic cap bits */
 518 #define CEPH_CAP_GSHARED     1  /* client can reads */
 519 #define CEPH_CAP_GEXCL       2  /* client can read and update */
 520 #define CEPH_CAP_GCACHE      4  /* (file) client can cache reads */
 521 #define CEPH_CAP_GRD         8  /* (file) client can read */
 522 #define CEPH_CAP_GWR        16  /* (file) client can write */
 523 #define CEPH_CAP_GBUFFER    32  /* (file) client can buffer writes */
 524 #define CEPH_CAP_GWREXTEND  64  /* (file) client can extend EOF */
 525 #define CEPH_CAP_GLAZYIO   128  /* (file) client can perform lazy io */
 526
 527 /* per-lock shift */
 528 #define CEPH_CAP_SAUTH      2
 529 #define CEPH_CAP_SLINK      4
 530 #define CEPH_CAP_SXATTR     6
 531 #define CEPH_CAP_SFILE      8
 532 #define CEPH_CAP_SFLOCK    20
 533
 534 #define CEPH_CAP_BITS       22
 535
 536 /* composed values */
 537 #define CEPH_CAP_AUTH_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SAUTH)
 538 #define CEPH_CAP_AUTH_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SAUTH)
 539 #define CEPH_CAP_LINK_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SLINK)
 540 #define CEPH_CAP_LINK_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SLINK)
 541 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED  << CEPH_CAP_SXATTR)
 542 #define CEPH_CAP_XATTR_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SXATTR)
 543 #define CEPH_CAP_FILE(x)    (x << CEPH_CAP_SFILE)
 544 #define CEPH_CAP_FILE_SHARED   (CEPH_CAP_GSHARED   << CEPH_CAP_SFILE)
 545 #define CEPH_CAP_FILE_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SFILE)
 546 #define CEPH_CAP_FILE_CACHE    (CEPH_CAP_GCACHE    << CEPH_CAP_SFILE)
 547 #define CEPH_CAP_FILE_RD       (CEPH_CAP_GRD       << CEPH_CAP_SFILE)
 548 #define CEPH_CAP_FILE_WR       (CEPH_CAP_GWR       << CEPH_CAP_SFILE)
 549 #define CEPH_CAP_FILE_BUFFER   (CEPH_CAP_GBUFFER   << CEPH_CAP_SFILE)
 550 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
 551 #define CEPH_CAP_FILE_LAZYIO   (CEPH_CAP_GLAZYIO   << CEPH_CAP_SFILE)
 552 #define CEPH_CAP_FLOCK_SHARED  (CEPH_CAP_GSHARED   << CEPH_CAP_SFLOCK)
 553 #define CEPH_CAP_FLOCK_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SFLOCK)
 554
 555
 556 /* cap masks (for getattr) */
 557 #define CEPH_STAT_CAP_INODE    CEPH_CAP_PIN
 558 #define CEPH_STAT_CAP_TYPE     CEPH_CAP_PIN  /* mode >> 12 */
 559 #define CEPH_STAT_CAP_SYMLINK  CEPH_CAP_PIN
 560 #define CEPH_STAT_CAP_UID      CEPH_CAP_AUTH_SHARED
 561 #define CEPH_STAT_CAP_GID      CEPH_CAP_AUTH_SHARED
 562 #define CEPH_STAT_CAP_MODE     CEPH_CAP_AUTH_SHARED
 563 #define CEPH_STAT_CAP_NLINK    CEPH_CAP_LINK_SHARED
 564 #define CEPH_STAT_CAP_LAYOUT   CEPH_CAP_FILE_SHARED
 565 #define CEPH_STAT_CAP_MTIME    CEPH_CAP_FILE_SHARED
 566 #define CEPH_STAT_CAP_SIZE     CEPH_CAP_FILE_SHARED
 567 #define CEPH_STAT_CAP_ATIME    CEPH_CAP_FILE_SHARED  /* fixme */
 568 #define CEPH_STAT_CAP_XATTR    CEPH_CAP_XATTR_SHARED
 569 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN |                 \
 570                                  CEPH_CAP_AUTH_SHARED | \
 571                                  CEPH_CAP_LINK_SHARED | \
 572                                  CEPH_CAP_FILE_SHARED | \
 573                                  CEPH_CAP_XATTR_SHARED)
 574
 575 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED |                     \
 576                               CEPH_CAP_LINK_SHARED |                    \
 577                               CEPH_CAP_XATTR_SHARED |                   \
 578                               CEPH_CAP_FILE_SHARED)
 579 #define CEPH_CAP_ANY_RD   (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD |     \
 580                            CEPH_CAP_FILE_CACHE)
 581
 582 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL |         \
 583                            CEPH_CAP_LINK_EXCL |         \
 584                            CEPH_CAP_XATTR_EXCL |        \
 585                            CEPH_CAP_FILE_EXCL)
 586 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \
 587                               CEPH_CAP_FILE_EXCL)
 588 #define CEPH_CAP_ANY_WR   (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
 589 #define CEPH_CAP_ANY      (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
 590                            CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
 591                            CEPH_CAP_PIN)
 592
 593 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
 594                         CEPH_LOCK_IXATTR)
 595
 596 int ceph_caps_for_mode(int mode);
 597
 598 enum {
 599         CEPH_CAP_OP_GRANT,         /* mds->client grant */
 600         CEPH_CAP_OP_REVOKE,        /* mds->client revoke */
 601         CEPH_CAP_OP_TRUNC,         /* mds->client trunc notify */
 602         CEPH_CAP_OP_EXPORT,        /* mds has exported the cap */
 603         CEPH_CAP_OP_IMPORT,        /* mds has imported the cap */
 604         CEPH_CAP_OP_UPDATE,        /* client->mds update */
 605         CEPH_CAP_OP_DROP,          /* client->mds drop cap bits */
 606         CEPH_CAP_OP_FLUSH,         /* client->mds cap writeback */
 607         CEPH_CAP_OP_FLUSH_ACK,     /* mds->client flushed */
 608         CEPH_CAP_OP_FLUSHSNAP,     /* client->mds flush snapped metadata */
 609         CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
 610         CEPH_CAP_OP_RELEASE,       /* client->mds release (clean) cap */
 611         CEPH_CAP_OP_RENEW,         /* client->mds renewal request */
 612 };
 613
 614 extern const char *ceph_cap_op_name(int op);
 615
 616 /*
 617  * caps message, used for capability callbacks, acks, requests, etc.
 618  */
 619 struct ceph_mds_caps {
 620         __le32 op;                  /* CEPH_CAP_OP_* */
 621         __le64 ino, realm;
 622         __le64 cap_id;
 623         __le32 seq, issue_seq;
 624         __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
 625         __le32 migrate_seq;
 626         __le64 snap_follows;
 627         __le32 snap_trace_len;
 628
 629         /* authlock */
 630         __le32 uid, gid, mode;
 631
 632         /* linklock */
 633         __le32 nlink;
 634
 635         /* xattrlock */
 636         __le32 xattr_len;
 637         __le64 xattr_version;
 638
 639         /* filelock */
 640         __le64 size, max_size, truncate_size;
 641         __le32 truncate_seq;
 642         struct ceph_timespec mtime, atime, ctime;
 643         struct ceph_file_layout layout;
 644         __le32 time_warp_seq;
 645 } __attribute__ ((packed));
 646
 647 /* cap release msg head */
 648 struct ceph_mds_cap_release {
 649         __le32 num;                /* number of cap_items that follow */
 650 } __attribute__ ((packed));
 651
 652 struct ceph_mds_cap_item {
 653         __le64 ino;
 654         __le64 cap_id;
 655         __le32 migrate_seq, seq;
 656 } __attribute__ ((packed));
 657
 658 #define CEPH_MDS_LEASE_REVOKE           1  /*    mds  -> client */
 659 #define CEPH_MDS_LEASE_RELEASE          2  /* client  -> mds    */
 660 #define CEPH_MDS_LEASE_RENEW            3  /* client <-> mds    */
 661 #define CEPH_MDS_LEASE_REVOKE_ACK       4  /* client  -> mds    */
 662
 663 extern const char *ceph_lease_op_name(int o);
 664
 665 /* lease msg header */
 666 struct ceph_mds_lease {
 667         __u8 action;            /* CEPH_MDS_LEASE_* */
 668         __le16 mask;            /* which lease */
 669         __le64 ino;
 670         __le64 first, last;     /* snap range */
 671         __le32 seq;
 672         __le32 duration_ms;     /* duration of renewal */
 673 } __attribute__ ((packed));
 674 /* followed by a __le32+string for dname */
 675
 676 /* client reconnect */
 677 struct ceph_mds_cap_reconnect {
 678         __le64 cap_id;
 679         __le32 wanted;
 680         __le32 issued;
 681         __le64 snaprealm;
 682         __le64 pathbase;        /* base ino for our path to this ino */
 683         __le32 flock_len;       /* size of flock state blob, if any */
 684 } __attribute__ ((packed));
 685 /* followed by flock blob */
 686
 687 struct ceph_mds_cap_reconnect_v1 {
 688         __le64 cap_id;
 689         __le32 wanted;
 690         __le32 issued;
 691         __le64 size;
 692         struct ceph_timespec mtime, atime;
 693         __le64 snaprealm;
 694         __le64 pathbase;        /* base ino for our path to this ino */
 695 } __attribute__ ((packed));
 696
 697 struct ceph_mds_snaprealm_reconnect {
 698         __le64 ino;     /* snap realm base */
 699         __le64 seq;     /* snap seq for this snap realm */
 700         __le64 parent;  /* parent realm */
 701 } __attribute__ ((packed));
 702
 703 /*
 704  * snaps
 705  */
 706 enum {
 707         CEPH_SNAP_OP_UPDATE,  /* CREATE or DESTROY */
 708         CEPH_SNAP_OP_CREATE,
 709         CEPH_SNAP_OP_DESTROY,
 710         CEPH_SNAP_OP_SPLIT,
 711 };
 712
 713 extern const char *ceph_snap_op_name(int o);
 714
 715 /* snap msg header */
 716 struct ceph_mds_snap_head {
 717         __le32 op;                /* CEPH_SNAP_OP_* */
 718         __le64 split;             /* ino to split off, if any */
 719         __le32 num_split_inos;    /* # inos belonging to new child realm */
 720         __le32 num_split_realms;  /* # child realms udner new child realm */
 721         __le32 trace_len;         /* size of snap trace blob */
 722 } __attribute__ ((packed));
 723 /* followed by split ino list, then split realms, then the trace blob */
 724
 725 /*
 726  * encode info about a snaprealm, as viewed by a client
 727  */
 728 struct ceph_mds_snap_realm {
 729         __le64 ino;           /* ino */
 730         __le64 created;       /* snap: when created */
 731         __le64 parent;        /* ino: parent realm */
 732         __le64 parent_since;  /* snap: same parent since */
 733         __le64 seq;           /* snap: version */
 734         __le32 num_snaps;
 735         __le32 num_prior_parent_snaps;
 736 } __attribute__ ((packed));
 737 /* followed by my snap list, then prior parent snap list */
 738
 739 #endif