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  * 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_file_layout_su(l) ((__s32)le32_to_cpu((l).fl_stripe_unit))
  57 #define ceph_file_layout_stripe_count(l) \
  58         ((__s32)le32_to_cpu((l).fl_stripe_count))
  59 #define ceph_file_layout_object_size(l) ((__s32)le32_to_cpu((l).fl_object_size))
  60 #define ceph_file_layout_cas_hash(l) ((__s32)le32_to_cpu((l).fl_cas_hash))
  61 #define ceph_file_layout_object_su(l) \
  62         ((__s32)le32_to_cpu((l).fl_object_stripe_unit))
  63 #define ceph_file_layout_pg_pool(l) \
  64         ((__s32)le32_to_cpu((l).fl_pg_pool))
  65
  66 static inline unsigned ceph_file_layout_stripe_width(struct ceph_file_layout *l)
  67 {
  68         return le32_to_cpu(l->fl_stripe_unit) *
  69                 le32_to_cpu(l->fl_stripe_count);
  70 }
  71
  72 /* "period" == bytes before i start on a new set of objects */
  73 static inline unsigned ceph_file_layout_period(struct ceph_file_layout *l)
  74 {
  75         return le32_to_cpu(l->fl_object_size) *
  76                 le32_to_cpu(l->fl_stripe_count);
  77 }
  78
  79 #define CEPH_MIN_STRIPE_UNIT 65536
  80
  81 int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
  82
  83 struct ceph_dir_layout {
  84         __u8   dl_dir_hash;   /* see ceph_hash.h for ids */
  85         __u8   dl_unused1;
  86         __u16  dl_unused2;
  87         __u32  dl_unused3;
  88 } __attribute__ ((packed));
  89
  90 /* crypto algorithms */
  91 #define CEPH_CRYPTO_NONE 0x0
  92 #define CEPH_CRYPTO_AES  0x1
  93
  94 #define CEPH_AES_IV "cephsageyudagreg"
  95
  96 /* security/authentication protocols */
  97 #define CEPH_AUTH_UNKNOWN       0x0
  98 #define CEPH_AUTH_NONE          0x1
  99 #define CEPH_AUTH_CEPHX         0x2
 100
 101 #define CEPH_AUTH_UID_DEFAULT ((__u64) -1)
 102
 103
 104 /*********************************************
 105  * message layer
 106  */
 107
 108 /*
 109  * message types
 110  */
 111
 112 /* misc */
 113 #define CEPH_MSG_SHUTDOWN               1
 114 #define CEPH_MSG_PING                   2
 115
 116 /* client <-> monitor */
 117 #define CEPH_MSG_MON_MAP                4
 118 #define CEPH_MSG_MON_GET_MAP            5
 119 #define CEPH_MSG_STATFS                 13
 120 #define CEPH_MSG_STATFS_REPLY           14
 121 #define CEPH_MSG_MON_SUBSCRIBE          15
 122 #define CEPH_MSG_MON_SUBSCRIBE_ACK      16
 123 #define CEPH_MSG_AUTH                   17
 124 #define CEPH_MSG_AUTH_REPLY             18
 125 #define CEPH_MSG_MON_GET_VERSION        19
 126 #define CEPH_MSG_MON_GET_VERSION_REPLY  20
 127
 128 /* client <-> mds */
 129 #define CEPH_MSG_MDS_MAP                21
 130
 131 #define CEPH_MSG_CLIENT_SESSION         22
 132 #define CEPH_MSG_CLIENT_RECONNECT       23
 133
 134 #define CEPH_MSG_CLIENT_REQUEST         24
 135 #define CEPH_MSG_CLIENT_REQUEST_FORWARD 25
 136 #define CEPH_MSG_CLIENT_REPLY           26
 137 #define CEPH_MSG_CLIENT_CAPS            0x310
 138 #define CEPH_MSG_CLIENT_LEASE           0x311
 139 #define CEPH_MSG_CLIENT_SNAP            0x312
 140 #define CEPH_MSG_CLIENT_CAPRELEASE      0x313
 141
 142 /* pool ops */
 143 #define CEPH_MSG_POOLOP_REPLY           48
 144 #define CEPH_MSG_POOLOP                 49
 145
 146
 147 /* osd */
 148 #define CEPH_MSG_OSD_MAP                41
 149 #define CEPH_MSG_OSD_OP                 42
 150 #define CEPH_MSG_OSD_OPREPLY            43
 151 #define CEPH_MSG_WATCH_NOTIFY           44
 152
 153
 154 /* watch-notify operations */
 155 enum {
 156   WATCH_NOTIFY                          = 1, /* notifying watcher */
 157   WATCH_NOTIFY_COMPLETE                 = 2, /* notifier notified when done */
 158 };
 159
 160
 161 /* pool operations */
 162 enum {
 163   POOL_OP_CREATE                        = 0x01,
 164   POOL_OP_DELETE                        = 0x02,
 165   POOL_OP_AUID_CHANGE                   = 0x03,
 166   POOL_OP_CREATE_SNAP                   = 0x11,
 167   POOL_OP_DELETE_SNAP                   = 0x12,
 168   POOL_OP_CREATE_UNMANAGED_SNAP         = 0x21,
 169   POOL_OP_DELETE_UNMANAGED_SNAP         = 0x22,
 170 };
 171
 172 struct ceph_mon_request_header {
 173         __le64 have_version;
 174         __le16 session_mon;
 175         __le64 session_mon_tid;
 176 } __attribute__ ((packed));
 177
 178 struct ceph_mon_statfs {
 179         struct ceph_mon_request_header monhdr;
 180         struct ceph_fsid fsid;
 181 } __attribute__ ((packed));
 182
 183 struct ceph_statfs {
 184         __le64 kb, kb_used, kb_avail;
 185         __le64 num_objects;
 186 } __attribute__ ((packed));
 187
 188 struct ceph_mon_statfs_reply {
 189         struct ceph_fsid fsid;
 190         __le64 version;
 191         struct ceph_statfs st;
 192 } __attribute__ ((packed));
 193
 194 const char *ceph_pool_op_name(int op);
 195
 196 struct ceph_mon_poolop {
 197         struct ceph_mon_request_header monhdr;
 198         struct ceph_fsid fsid;
 199         __le32 pool;
 200         __le32 op;
 201         __le64 auid;
 202         __le64 snapid;
 203         __le32 name_len;
 204 } __attribute__ ((packed));
 205
 206 struct ceph_mon_poolop_reply {
 207         struct ceph_mon_request_header monhdr;
 208         struct ceph_fsid fsid;
 209         __le32 reply_code;
 210         __le32 epoch;
 211         char has_data;
 212         char data[0];
 213 } __attribute__ ((packed));
 214
 215 struct ceph_mon_unmanaged_snap {
 216         __le64 snapid;
 217 } __attribute__ ((packed));
 218
 219 struct ceph_osd_getmap {
 220         struct ceph_mon_request_header monhdr;
 221         struct ceph_fsid fsid;
 222         __le32 start;
 223 } __attribute__ ((packed));
 224
 225 struct ceph_mds_getmap {
 226         struct ceph_mon_request_header monhdr;
 227         struct ceph_fsid fsid;
 228 } __attribute__ ((packed));
 229
 230 struct ceph_client_mount {
 231         struct ceph_mon_request_header monhdr;
 232 } __attribute__ ((packed));
 233
 234 #define CEPH_SUBSCRIBE_ONETIME    1  /* i want only 1 update after have */
 235
 236 struct ceph_mon_subscribe_item {
 237         __le64 have_version;    __le64 have;
 238         __u8 onetime;
 239 } __attribute__ ((packed));
 240
 241 struct ceph_mon_subscribe_ack {
 242         __le32 duration;         /* seconds */
 243         struct ceph_fsid fsid;
 244 } __attribute__ ((packed));
 245
 246 /*
 247  * mdsmap flags
 248  */
 249 #define CEPH_MDSMAP_DOWN    (1<<0)  /* cluster deliberately down */
 250
 251 /*
 252  * mds states
 253  *   > 0 -> in
 254  *  <= 0 -> out
 255  */
 256 #define CEPH_MDS_STATE_DNE          0  /* down, does not exist. */
 257 #define CEPH_MDS_STATE_STOPPED     -1  /* down, once existed, but no subtrees.
 258                                           empty log. */
 259 #define CEPH_MDS_STATE_BOOT        -4  /* up, boot announcement. */
 260 #define CEPH_MDS_STATE_STANDBY     -5  /* up, idle.  waiting for assignment. */
 261 #define CEPH_MDS_STATE_CREATING    -6  /* up, creating MDS instance. */
 262 #define CEPH_MDS_STATE_STARTING    -7  /* up, starting previously stopped mds */
 263 #define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
 264 #define CEPH_MDS_STATE_REPLAYONCE   -9 /* up, replaying an active node's journal */
 265
 266 #define CEPH_MDS_STATE_REPLAY       8  /* up, replaying journal. */
 267 #define CEPH_MDS_STATE_RESOLVE      9  /* up, disambiguating distributed
 268                                           operations (import, rename, etc.) */
 269 #define CEPH_MDS_STATE_RECONNECT    10 /* up, reconnect to clients */
 270 #define CEPH_MDS_STATE_REJOIN       11 /* up, rejoining distributed cache */
 271 #define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */
 272 #define CEPH_MDS_STATE_ACTIVE       13 /* up, active */
 273 #define CEPH_MDS_STATE_STOPPING     14 /* up, but exporting metadata */
 274
 275 extern const char *ceph_mds_state_name(int s);
 276
 277
 278 /*
 279  * metadata lock types.
 280  *  - these are bitmasks.. we can compose them
 281  *  - they also define the lock ordering by the MDS
 282  *  - a few of these are internal to the mds
 283  */
 284 #define CEPH_LOCK_DVERSION    1
 285 #define CEPH_LOCK_DN          2
 286 #define CEPH_LOCK_ISNAP       16
 287 #define CEPH_LOCK_IVERSION    32    /* mds internal */
 288 #define CEPH_LOCK_IFILE       64
 289 #define CEPH_LOCK_IAUTH       128
 290 #define CEPH_LOCK_ILINK       256
 291 #define CEPH_LOCK_IDFT        512   /* dir frag tree */
 292 #define CEPH_LOCK_INEST       1024  /* mds internal */
 293 #define CEPH_LOCK_IXATTR      2048
 294 #define CEPH_LOCK_IFLOCK      4096  /* advisory file locks */
 295 #define CEPH_LOCK_INO         8192  /* immutable inode bits; not a lock */
 296 #define CEPH_LOCK_IPOLICY     16384 /* policy lock on dirs. MDS internal */
 297
 298 /* client_session ops */
 299 enum {
 300         CEPH_SESSION_REQUEST_OPEN,
 301         CEPH_SESSION_OPEN,
 302         CEPH_SESSION_REQUEST_CLOSE,
 303         CEPH_SESSION_CLOSE,
 304         CEPH_SESSION_REQUEST_RENEWCAPS,
 305         CEPH_SESSION_RENEWCAPS,
 306         CEPH_SESSION_STALE,
 307         CEPH_SESSION_RECALL_STATE,
 308         CEPH_SESSION_FLUSHMSG,
 309         CEPH_SESSION_FLUSHMSG_ACK,
 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
 336         CEPH_MDS_OP_SETXATTR   = 0x01105,
 337         CEPH_MDS_OP_RMXATTR    = 0x01106,
 338         CEPH_MDS_OP_SETLAYOUT  = 0x01107,
 339         CEPH_MDS_OP_SETATTR    = 0x01108,
 340         CEPH_MDS_OP_SETFILELOCK= 0x01109,
 341         CEPH_MDS_OP_GETFILELOCK= 0x00110,
 342         CEPH_MDS_OP_SETDIRLAYOUT=0x0110a,
 343
 344         CEPH_MDS_OP_MKNOD      = 0x01201,
 345         CEPH_MDS_OP_LINK       = 0x01202,
 346         CEPH_MDS_OP_UNLINK     = 0x01203,
 347         CEPH_MDS_OP_RENAME     = 0x01204,
 348         CEPH_MDS_OP_MKDIR      = 0x01220,
 349         CEPH_MDS_OP_RMDIR      = 0x01221,
 350         CEPH_MDS_OP_SYMLINK    = 0x01222,
 351
 352         CEPH_MDS_OP_CREATE     = 0x01301,
 353         CEPH_MDS_OP_OPEN       = 0x00302,
 354         CEPH_MDS_OP_READDIR    = 0x00305,
 355
 356         CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
 357         CEPH_MDS_OP_MKSNAP     = 0x01400,
 358         CEPH_MDS_OP_RMSNAP     = 0x01401,
 359         CEPH_MDS_OP_LSSNAP     = 0x00402,
 360 };
 361
 362 extern const char *ceph_mds_op_name(int op);
 363
 364
 365 #define CEPH_SETATTR_MODE   1
 366 #define CEPH_SETATTR_UID    2
 367 #define CEPH_SETATTR_GID    4
 368 #define CEPH_SETATTR_MTIME  8
 369 #define CEPH_SETATTR_ATIME 16
 370 #define CEPH_SETATTR_SIZE  32
 371 #define CEPH_SETATTR_CTIME 64
 372
 373 /*
 374  * Ceph setxattr request flags.
 375  */
 376 #define CEPH_XATTR_CREATE  1
 377 #define CEPH_XATTR_REPLACE 2
 378
 379 union ceph_mds_request_args {
 380         struct {
 381                 __le32 mask;                 /* CEPH_CAP_* */
 382         } __attribute__ ((packed)) getattr;
 383         struct {
 384                 __le32 mode;
 385                 __le32 uid;
 386                 __le32 gid;
 387                 struct ceph_timespec mtime;
 388                 struct ceph_timespec atime;
 389                 __le64 size, old_size;       /* old_size needed by truncate */
 390                 __le32 mask;                 /* CEPH_SETATTR_* */
 391         } __attribute__ ((packed)) setattr;
 392         struct {
 393                 __le32 frag;                 /* which dir fragment */
 394                 __le32 max_entries;          /* how many dentries to grab */
 395                 __le32 max_bytes;
 396         } __attribute__ ((packed)) readdir;
 397         struct {
 398                 __le32 mode;
 399                 __le32 rdev;
 400         } __attribute__ ((packed)) mknod;
 401         struct {
 402                 __le32 mode;
 403         } __attribute__ ((packed)) mkdir;
 404         struct {
 405                 __le32 flags;
 406                 __le32 mode;
 407                 __le32 stripe_unit;          /* layout for newly created file */
 408                 __le32 stripe_count;         /* ... */
 409                 __le32 object_size;
 410                 __le32 file_replication;
 411                 __le32 unused;               /* used to be preferred osd */
 412         } __attribute__ ((packed)) open;
 413         struct {
 414                 __le32 flags;
 415         } __attribute__ ((packed)) setxattr;
 416         struct {
 417                 struct ceph_file_layout layout;
 418         } __attribute__ ((packed)) setlayout;
 419         struct {
 420                 __u8 rule; /* currently fcntl or flock */
 421                 __u8 type; /* shared, exclusive, remove*/
 422                 __le64 pid; /* process id requesting the lock */
 423                 __le64 pid_namespace;
 424                 __le64 start; /* initial location to lock */
 425                 __le64 length; /* num bytes to lock from start */
 426                 __u8 wait; /* will caller wait for lock to become available? */
 427         } __attribute__ ((packed)) filelock_change;
 428 } __attribute__ ((packed));
 429
 430 #define CEPH_MDS_FLAG_REPLAY        1  /* this is a replayed op */
 431 #define CEPH_MDS_FLAG_WANT_DENTRY   2  /* want dentry in reply */
 432
 433 struct ceph_mds_request_head {
 434         __le64 oldest_client_tid;
 435         __le32 mdsmap_epoch;           /* on client */
 436         __le32 flags;                  /* CEPH_MDS_FLAG_* */
 437         __u8 num_retry, num_fwd;       /* count retry, fwd attempts */
 438         __le16 num_releases;           /* # include cap/lease release records */
 439         __le32 op;                     /* mds op code */
 440         __le32 caller_uid, caller_gid;
 441         __le64 ino;                    /* use this ino for openc, mkdir, mknod,
 442                                           etc. (if replaying) */
 443         union ceph_mds_request_args args;
 444 } __attribute__ ((packed));
 445
 446 /* cap/lease release record */
 447 struct ceph_mds_request_release {
 448         __le64 ino, cap_id;            /* ino and unique cap id */
 449         __le32 caps, wanted;           /* new issued, wanted */
 450         __le32 seq, issue_seq, mseq;
 451         __le32 dname_seq;              /* if releasing a dentry lease, a */
 452         __le32 dname_len;              /* string follows. */
 453 } __attribute__ ((packed));
 454
 455 /* client reply */
 456 struct ceph_mds_reply_head {
 457         __le32 op;
 458         __le32 result;
 459         __le32 mdsmap_epoch;
 460         __u8 safe;                     /* true if committed to disk */
 461         __u8 is_dentry, is_target;     /* true if dentry, target inode records
 462                                           are included with reply */
 463 } __attribute__ ((packed));
 464
 465 /* one for each node split */
 466 struct ceph_frag_tree_split {
 467         __le32 frag;                   /* this frag splits... */
 468         __le32 by;                     /* ...by this many bits */
 469 } __attribute__ ((packed));
 470
 471 struct ceph_frag_tree_head {
 472         __le32 nsplits;                /* num ceph_frag_tree_split records */
 473         struct ceph_frag_tree_split splits[];
 474 } __attribute__ ((packed));
 475
 476 /* capability issue, for bundling with mds reply */
 477 struct ceph_mds_reply_cap {
 478         __le32 caps, wanted;           /* caps issued, wanted */
 479         __le64 cap_id;
 480         __le32 seq, mseq;
 481         __le64 realm;                  /* snap realm */
 482         __u8 flags;                    /* CEPH_CAP_FLAG_* */
 483 } __attribute__ ((packed));
 484
 485 #define CEPH_CAP_FLAG_AUTH      (1 << 0)  /* cap is issued by auth mds */
 486 #define CEPH_CAP_FLAG_RELEASE   (1 << 1)  /* release the cap */
 487
 488 /* inode record, for bundling with mds reply */
 489 struct ceph_mds_reply_inode {
 490         __le64 ino;
 491         __le64 snapid;
 492         __le32 rdev;
 493         __le64 version;                /* inode version */
 494         __le64 xattr_version;          /* version for xattr blob */
 495         struct ceph_mds_reply_cap cap; /* caps issued for this inode */
 496         struct ceph_file_layout layout;
 497         struct ceph_timespec ctime, mtime, atime;
 498         __le32 time_warp_seq;
 499         __le64 size, max_size, truncate_size;
 500         __le32 truncate_seq;
 501         __le32 mode, uid, gid;
 502         __le32 nlink;
 503         __le64 files, subdirs, rbytes, rfiles, rsubdirs;  /* dir stats */
 504         struct ceph_timespec rctime;
 505         struct ceph_frag_tree_head fragtree;  /* (must be at end of struct) */
 506 } __attribute__ ((packed));
 507 /* followed by frag array, symlink string, dir layout, xattr blob */
 508
 509 /* reply_lease follows dname, and reply_inode */
 510 struct ceph_mds_reply_lease {
 511         __le16 mask;            /* lease type(s) */
 512         __le32 duration_ms;     /* lease duration */
 513         __le32 seq;
 514 } __attribute__ ((packed));
 515
 516 struct ceph_mds_reply_dirfrag {
 517         __le32 frag;            /* fragment */
 518         __le32 auth;            /* auth mds, if this is a delegation point */
 519         __le32 ndist;           /* number of mds' this is replicated on */
 520         __le32 dist[];
 521 } __attribute__ ((packed));
 522
 523 #define CEPH_LOCK_FCNTL    1
 524 #define CEPH_LOCK_FLOCK    2
 525
 526 #define CEPH_LOCK_SHARED   1
 527 #define CEPH_LOCK_EXCL     2
 528 #define CEPH_LOCK_UNLOCK   4
 529
 530 struct ceph_filelock {
 531         __le64 start;/* file offset to start lock at */
 532         __le64 length; /* num bytes to lock; 0 for all following start */
 533         __le64 client; /* which client holds the lock */
 534         __le64 pid; /* process id holding the lock on the client */
 535         __le64 pid_namespace;
 536         __u8 type; /* shared lock, exclusive lock, or unlock */
 537 } __attribute__ ((packed));
 538
 539
 540 /* file access modes */
 541 #define CEPH_FILE_MODE_PIN        0
 542 #define CEPH_FILE_MODE_RD         1
 543 #define CEPH_FILE_MODE_WR         2
 544 #define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
 545 #define CEPH_FILE_MODE_LAZY       4  /* lazy io */
 546 #define CEPH_FILE_MODE_NUM        8  /* bc these are bit fields.. mostly */
 547
 548 int ceph_flags_to_mode(int flags);
 549
 550
 551 /* capability bits */
 552 #define CEPH_CAP_PIN         1  /* no specific capabilities beyond the pin */
 553
 554 /* generic cap bits */
 555 #define CEPH_CAP_GSHARED     1  /* client can reads */
 556 #define CEPH_CAP_GEXCL       2  /* client can read and update */
 557 #define CEPH_CAP_GCACHE      4  /* (file) client can cache reads */
 558 #define CEPH_CAP_GRD         8  /* (file) client can read */
 559 #define CEPH_CAP_GWR        16  /* (file) client can write */
 560 #define CEPH_CAP_GBUFFER    32  /* (file) client can buffer writes */
 561 #define CEPH_CAP_GWREXTEND  64  /* (file) client can extend EOF */
 562 #define CEPH_CAP_GLAZYIO   128  /* (file) client can perform lazy io */
 563
 564 #define CEPH_CAP_SIMPLE_BITS  2
 565 #define CEPH_CAP_FILE_BITS    8
 566
 567 /* per-lock shift */
 568 #define CEPH_CAP_SAUTH      2
 569 #define CEPH_CAP_SLINK      4
 570 #define CEPH_CAP_SXATTR     6
 571 #define CEPH_CAP_SFILE      8
 572 #define CEPH_CAP_SFLOCK    20
 573
 574 #define CEPH_CAP_BITS      22
 575
 576 /* composed values */
 577 #define CEPH_CAP_AUTH_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SAUTH)
 578 #define CEPH_CAP_AUTH_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SAUTH)
 579 #define CEPH_CAP_LINK_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SLINK)
 580 #define CEPH_CAP_LINK_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SLINK)
 581 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED  << CEPH_CAP_SXATTR)
 582 #define CEPH_CAP_XATTR_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SXATTR)
 583 #define CEPH_CAP_FILE(x)    (x << CEPH_CAP_SFILE)
 584 #define CEPH_CAP_FILE_SHARED   (CEPH_CAP_GSHARED   << CEPH_CAP_SFILE)
 585 #define CEPH_CAP_FILE_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SFILE)
 586 #define CEPH_CAP_FILE_CACHE    (CEPH_CAP_GCACHE    << CEPH_CAP_SFILE)
 587 #define CEPH_CAP_FILE_RD       (CEPH_CAP_GRD       << CEPH_CAP_SFILE)
 588 #define CEPH_CAP_FILE_WR       (CEPH_CAP_GWR       << CEPH_CAP_SFILE)
 589 #define CEPH_CAP_FILE_BUFFER   (CEPH_CAP_GBUFFER   << CEPH_CAP_SFILE)
 590 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
 591 #define CEPH_CAP_FILE_LAZYIO   (CEPH_CAP_GLAZYIO   << CEPH_CAP_SFILE)
 592 #define CEPH_CAP_FLOCK_SHARED  (CEPH_CAP_GSHARED   << CEPH_CAP_SFLOCK)
 593 #define CEPH_CAP_FLOCK_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SFLOCK)
 594
 595
 596 /* cap masks (for getattr) */
 597 #define CEPH_STAT_CAP_INODE    CEPH_CAP_PIN
 598 #define CEPH_STAT_CAP_TYPE     CEPH_CAP_PIN  /* mode >> 12 */
 599 #define CEPH_STAT_CAP_SYMLINK  CEPH_CAP_PIN
 600 #define CEPH_STAT_CAP_UID      CEPH_CAP_AUTH_SHARED
 601 #define CEPH_STAT_CAP_GID      CEPH_CAP_AUTH_SHARED
 602 #define CEPH_STAT_CAP_MODE     CEPH_CAP_AUTH_SHARED
 603 #define CEPH_STAT_CAP_NLINK    CEPH_CAP_LINK_SHARED
 604 #define CEPH_STAT_CAP_LAYOUT   CEPH_CAP_FILE_SHARED
 605 #define CEPH_STAT_CAP_MTIME    CEPH_CAP_FILE_SHARED
 606 #define CEPH_STAT_CAP_SIZE     CEPH_CAP_FILE_SHARED
 607 #define CEPH_STAT_CAP_ATIME    CEPH_CAP_FILE_SHARED  /* fixme */
 608 #define CEPH_STAT_CAP_XATTR    CEPH_CAP_XATTR_SHARED
 609 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN |                 \
 610                                  CEPH_CAP_AUTH_SHARED | \
 611                                  CEPH_CAP_LINK_SHARED | \
 612                                  CEPH_CAP_FILE_SHARED | \
 613                                  CEPH_CAP_XATTR_SHARED)
 614
 615 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED |                     \
 616                               CEPH_CAP_LINK_SHARED |                    \
 617                               CEPH_CAP_XATTR_SHARED |                   \
 618                               CEPH_CAP_FILE_SHARED)
 619 #define CEPH_CAP_ANY_RD   (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD |     \
 620                            CEPH_CAP_FILE_CACHE)
 621
 622 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL |         \
 623                            CEPH_CAP_LINK_EXCL |         \
 624                            CEPH_CAP_XATTR_EXCL |        \
 625                            CEPH_CAP_FILE_EXCL)
 626 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \
 627                               CEPH_CAP_FILE_EXCL)
 628 #define CEPH_CAP_ANY_WR   (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
 629 #define CEPH_CAP_ANY      (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
 630                            CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
 631                            CEPH_CAP_PIN)
 632
 633 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
 634                         CEPH_LOCK_IXATTR)
 635
 636 int ceph_caps_for_mode(int mode);
 637
 638 enum {
 639         CEPH_CAP_OP_GRANT,         /* mds->client grant */
 640         CEPH_CAP_OP_REVOKE,        /* mds->client revoke */
 641         CEPH_CAP_OP_TRUNC,         /* mds->client trunc notify */
 642         CEPH_CAP_OP_EXPORT,        /* mds has exported the cap */
 643         CEPH_CAP_OP_IMPORT,        /* mds has imported the cap */
 644         CEPH_CAP_OP_UPDATE,        /* client->mds update */
 645         CEPH_CAP_OP_DROP,          /* client->mds drop cap bits */
 646         CEPH_CAP_OP_FLUSH,         /* client->mds cap writeback */
 647         CEPH_CAP_OP_FLUSH_ACK,     /* mds->client flushed */
 648         CEPH_CAP_OP_FLUSHSNAP,     /* client->mds flush snapped metadata */
 649         CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
 650         CEPH_CAP_OP_RELEASE,       /* client->mds release (clean) cap */
 651         CEPH_CAP_OP_RENEW,         /* client->mds renewal request */
 652 };
 653
 654 extern const char *ceph_cap_op_name(int op);
 655
 656 /*
 657  * caps message, used for capability callbacks, acks, requests, etc.
 658  */
 659 struct ceph_mds_caps {
 660         __le32 op;                  /* CEPH_CAP_OP_* */
 661         __le64 ino, realm;
 662         __le64 cap_id;
 663         __le32 seq, issue_seq;
 664         __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
 665         __le32 migrate_seq;
 666         __le64 snap_follows;
 667         __le32 snap_trace_len;
 668
 669         /* authlock */
 670         __le32 uid, gid, mode;
 671
 672         /* linklock */
 673         __le32 nlink;
 674
 675         /* xattrlock */
 676         __le32 xattr_len;
 677         __le64 xattr_version;
 678
 679         /* filelock */
 680         __le64 size, max_size, truncate_size;
 681         __le32 truncate_seq;
 682         struct ceph_timespec mtime, atime, ctime;
 683         struct ceph_file_layout layout;
 684         __le32 time_warp_seq;
 685 } __attribute__ ((packed));
 686
 687 struct ceph_mds_cap_peer {
 688         __le64 cap_id;
 689         __le32 seq;
 690         __le32 mseq;
 691         __le32 mds;
 692         __u8   flags;
 693 } __attribute__ ((packed));
 694
 695 /* cap release msg head */
 696 struct ceph_mds_cap_release {
 697         __le32 num;                /* number of cap_items that follow */
 698 } __attribute__ ((packed));
 699
 700 struct ceph_mds_cap_item {
 701         __le64 ino;
 702         __le64 cap_id;
 703         __le32 migrate_seq, seq;
 704 } __attribute__ ((packed));
 705
 706 #define CEPH_MDS_LEASE_REVOKE           1  /*    mds  -> client */
 707 #define CEPH_MDS_LEASE_RELEASE          2  /* client  -> mds    */
 708 #define CEPH_MDS_LEASE_RENEW            3  /* client <-> mds    */
 709 #define CEPH_MDS_LEASE_REVOKE_ACK       4  /* client  -> mds    */
 710
 711 extern const char *ceph_lease_op_name(int o);
 712
 713 /* lease msg header */
 714 struct ceph_mds_lease {
 715         __u8 action;            /* CEPH_MDS_LEASE_* */
 716         __le16 mask;            /* which lease */
 717         __le64 ino;
 718         __le64 first, last;     /* snap range */
 719         __le32 seq;
 720         __le32 duration_ms;     /* duration of renewal */
 721 } __attribute__ ((packed));
 722 /* followed by a __le32+string for dname */
 723
 724 /* client reconnect */
 725 struct ceph_mds_cap_reconnect {
 726         __le64 cap_id;
 727         __le32 wanted;
 728         __le32 issued;
 729         __le64 snaprealm;
 730         __le64 pathbase;        /* base ino for our path to this ino */
 731         __le32 flock_len;       /* size of flock state blob, if any */
 732 } __attribute__ ((packed));
 733 /* followed by flock blob */
 734
 735 struct ceph_mds_cap_reconnect_v1 {
 736         __le64 cap_id;
 737         __le32 wanted;
 738         __le32 issued;
 739         __le64 size;
 740         struct ceph_timespec mtime, atime;
 741         __le64 snaprealm;
 742         __le64 pathbase;        /* base ino for our path to this ino */
 743 } __attribute__ ((packed));
 744
 745 struct ceph_mds_snaprealm_reconnect {
 746         __le64 ino;     /* snap realm base */
 747         __le64 seq;     /* snap seq for this snap realm */
 748         __le64 parent;  /* parent realm */
 749 } __attribute__ ((packed));
 750
 751 /*
 752  * snaps
 753  */
 754 enum {
 755         CEPH_SNAP_OP_UPDATE,  /* CREATE or DESTROY */
 756         CEPH_SNAP_OP_CREATE,
 757         CEPH_SNAP_OP_DESTROY,
 758         CEPH_SNAP_OP_SPLIT,
 759 };
 760
 761 extern const char *ceph_snap_op_name(int o);
 762
 763 /* snap msg header */
 764 struct ceph_mds_snap_head {
 765         __le32 op;                /* CEPH_SNAP_OP_* */
 766         __le64 split;             /* ino to split off, if any */
 767         __le32 num_split_inos;    /* # inos belonging to new child realm */
 768         __le32 num_split_realms;  /* # child realms udner new child realm */
 769         __le32 trace_len;         /* size of snap trace blob */
 770 } __attribute__ ((packed));
 771 /* followed by split ino list, then split realms, then the trace blob */
 772
 773 /*
 774  * encode info about a snaprealm, as viewed by a client
 775  */
 776 struct ceph_mds_snap_realm {
 777         __le64 ino;           /* ino */
 778         __le64 created;       /* snap: when created */
 779         __le64 parent;        /* ino: parent realm */
 780         __le64 parent_since;  /* snap: same parent since */
 781         __le64 seq;           /* snap: version */
 782         __le32 num_snaps;
 783         __le32 num_prior_parent_snaps;
 784 } __attribute__ ((packed));
 785 /* followed by my snap list, then prior parent snap list */
 786
 787 #endif