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