1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * ceph_fs.h - Ceph constants and data types to share between kernel and 4 * user space. 5 * 6 * Most types in this file are defined as little-endian, and are 7 * primarily intended to describe data structures that pass over the 8 * wire or that are stored on disk. 9 * 10 * LGPL2 11 */ 12 13 #ifndef CEPH_FS_H 14 #define CEPH_FS_H 15 16 #include <linux/ceph/msgr.h> 17 #include <linux/ceph/rados.h> 18 19 /* 20 * subprotocol versions. when specific messages types or high-level 21 * protocols change, bump the affected components. we keep rev 22 * internal cluster protocols separately from the public, 23 * client-facing protocol. 24 */ 25 #define CEPH_OSDC_PROTOCOL 24 /* server/client */ 26 #define CEPH_MDSC_PROTOCOL 32 /* server/client */ 27 #define CEPH_MONC_PROTOCOL 15 /* server/client */ 28 29 30 #define CEPH_INO_ROOT 1 31 #define CEPH_INO_CEPH 2 /* hidden .ceph dir */ 32 #define CEPH_INO_DOTDOT 3 /* used by ceph fuse for parent (..) */ 33 34 /* arbitrary limit on max # of monitors (cluster of 3 is typical) */ 35 #define CEPH_MAX_MON 31 36 37 /* 38 * legacy ceph_file_layoute 39 */ 40 struct ceph_file_layout_legacy { 41 /* file -> object mapping */ 42 __le32 fl_stripe_unit; /* stripe unit, in bytes. must be multiple 43 of page size. */ 44 __le32 fl_stripe_count; /* over this many objects */ 45 __le32 fl_object_size; /* until objects are this big, then move to 46 new objects */ 47 __le32 fl_cas_hash; /* UNUSED. 0 = none; 1 = sha256 */ 48 49 /* pg -> disk layout */ 50 __le32 fl_object_stripe_unit; /* UNUSED. for per-object parity, if any */ 51 52 /* object -> pg layout */ 53 __le32 fl_unused; /* unused; used to be preferred primary for pg (-1 for none) */ 54 __le32 fl_pg_pool; /* namespace, crush ruleset, rep level */ 55 } __attribute__ ((packed)); 56 57 struct ceph_string; 58 /* 59 * ceph_file_layout - describe data layout for a file/inode 60 */ 61 struct ceph_file_layout { 62 /* file -> object mapping */ 63 u32 stripe_unit; /* stripe unit, in bytes */ 64 u32 stripe_count; /* over this many objects */ 65 u32 object_size; /* until objects are this big */ 66 s64 pool_id; /* rados pool id */ 67 struct ceph_string __rcu *pool_ns; /* rados pool namespace */ 68 }; 69 70 extern int ceph_file_layout_is_valid(const struct ceph_file_layout *layout); 71 extern void ceph_file_layout_from_legacy(struct ceph_file_layout *fl, 72 struct ceph_file_layout_legacy *legacy); 73 extern void ceph_file_layout_to_legacy(struct ceph_file_layout *fl, 74 struct ceph_file_layout_legacy *legacy); 75 76 #define CEPH_MIN_STRIPE_UNIT 65536 77 78 struct ceph_dir_layout { 79 __u8 dl_dir_hash; /* see ceph_hash.h for ids */ 80 __u8 dl_unused1; 81 __u16 dl_unused2; 82 __u32 dl_unused3; 83 } __attribute__ ((packed)); 84 85 /* crypto algorithms */ 86 #define CEPH_CRYPTO_NONE 0x0 87 #define CEPH_CRYPTO_AES 0x1 88 89 #define CEPH_AES_IV "cephsageyudagreg" 90 91 /* security/authentication protocols */ 92 #define CEPH_AUTH_UNKNOWN 0x0 93 #define CEPH_AUTH_NONE 0x1 94 #define CEPH_AUTH_CEPHX 0x2 95 96 #define CEPH_AUTH_UID_DEFAULT ((__u64) -1) 97 98 99 /********************************************* 100 * message layer 101 */ 102 103 /* 104 * message types 105 */ 106 107 /* misc */ 108 #define CEPH_MSG_SHUTDOWN 1 109 #define CEPH_MSG_PING 2 110 111 /* client <-> monitor */ 112 #define CEPH_MSG_MON_MAP 4 113 #define CEPH_MSG_MON_GET_MAP 5 114 #define CEPH_MSG_STATFS 13 115 #define CEPH_MSG_STATFS_REPLY 14 116 #define CEPH_MSG_MON_SUBSCRIBE 15 117 #define CEPH_MSG_MON_SUBSCRIBE_ACK 16 118 #define CEPH_MSG_AUTH 17 119 #define CEPH_MSG_AUTH_REPLY 18 120 #define CEPH_MSG_MON_GET_VERSION 19 121 #define CEPH_MSG_MON_GET_VERSION_REPLY 20 122 123 /* client <-> mds */ 124 #define CEPH_MSG_MDS_MAP 21 125 #define CEPH_MSG_FS_MAP_USER 103 126 127 #define CEPH_MSG_CLIENT_SESSION 22 128 #define CEPH_MSG_CLIENT_RECONNECT 23 129 130 #define CEPH_MSG_CLIENT_REQUEST 24 131 #define CEPH_MSG_CLIENT_REQUEST_FORWARD 25 132 #define CEPH_MSG_CLIENT_REPLY 26 133 #define CEPH_MSG_CLIENT_METRICS 29 134 #define CEPH_MSG_CLIENT_CAPS 0x310 135 #define CEPH_MSG_CLIENT_LEASE 0x311 136 #define CEPH_MSG_CLIENT_SNAP 0x312 137 #define CEPH_MSG_CLIENT_CAPRELEASE 0x313 138 #define CEPH_MSG_CLIENT_QUOTA 0x314 139 140 /* pool ops */ 141 #define CEPH_MSG_POOLOP_REPLY 48 142 #define CEPH_MSG_POOLOP 49 143 144 /* mon commands */ 145 #define CEPH_MSG_MON_COMMAND 50 146 #define CEPH_MSG_MON_COMMAND_ACK 51 147 148 /* osd */ 149 #define CEPH_MSG_OSD_MAP 41 150 #define CEPH_MSG_OSD_OP 42 151 #define CEPH_MSG_OSD_OPREPLY 43 152 #define CEPH_MSG_WATCH_NOTIFY 44 153 #define CEPH_MSG_OSD_BACKOFF 61 154 155 156 /* watch-notify operations */ 157 enum { 158 CEPH_WATCH_EVENT_NOTIFY = 1, /* notifying watcher */ 159 CEPH_WATCH_EVENT_NOTIFY_COMPLETE = 2, /* notifier notified when done */ 160 CEPH_WATCH_EVENT_DISCONNECT = 3, /* we were disconnected */ 161 }; 162 163 164 struct ceph_mon_request_header { 165 __le64 have_version; 166 __le16 session_mon; 167 __le64 session_mon_tid; 168 } __attribute__ ((packed)); 169 170 struct ceph_mon_statfs { 171 struct ceph_mon_request_header monhdr; 172 struct ceph_fsid fsid; 173 __u8 contains_data_pool; 174 __le64 data_pool; 175 } __attribute__ ((packed)); 176 177 struct ceph_statfs { 178 __le64 kb, kb_used, kb_avail; 179 __le64 num_objects; 180 } __attribute__ ((packed)); 181 182 struct ceph_mon_statfs_reply { 183 struct ceph_fsid fsid; 184 __le64 version; 185 struct ceph_statfs st; 186 } __attribute__ ((packed)); 187 188 struct ceph_mon_command { 189 struct ceph_mon_request_header monhdr; 190 struct ceph_fsid fsid; 191 __le32 num_strs; /* always 1 */ 192 __le32 str_len; 193 char str[]; 194 } __attribute__ ((packed)); 195 196 struct ceph_osd_getmap { 197 struct ceph_mon_request_header monhdr; 198 struct ceph_fsid fsid; 199 __le32 start; 200 } __attribute__ ((packed)); 201 202 struct ceph_mds_getmap { 203 struct ceph_mon_request_header monhdr; 204 struct ceph_fsid fsid; 205 } __attribute__ ((packed)); 206 207 struct ceph_client_mount { 208 struct ceph_mon_request_header monhdr; 209 } __attribute__ ((packed)); 210 211 #define CEPH_SUBSCRIBE_ONETIME 1 /* i want only 1 update after have */ 212 213 struct ceph_mon_subscribe_item { 214 __le64 start; 215 __u8 flags; 216 } __attribute__ ((packed)); 217 218 struct ceph_mon_subscribe_ack { 219 __le32 duration; /* seconds */ 220 struct ceph_fsid fsid; 221 } __attribute__ ((packed)); 222 223 #define CEPH_FS_CLUSTER_ID_NONE -1 224 225 /* 226 * mdsmap flags 227 */ 228 #define CEPH_MDSMAP_DOWN (1<<0) /* cluster deliberately down */ 229 230 /* 231 * mds states 232 * > 0 -> in 233 * <= 0 -> out 234 */ 235 #define CEPH_MDS_STATE_DNE 0 /* down, does not exist. */ 236 #define CEPH_MDS_STATE_STOPPED -1 /* down, once existed, but no subtrees. 237 empty log. */ 238 #define CEPH_MDS_STATE_BOOT -4 /* up, boot announcement. */ 239 #define CEPH_MDS_STATE_STANDBY -5 /* up, idle. waiting for assignment. */ 240 #define CEPH_MDS_STATE_CREATING -6 /* up, creating MDS instance. */ 241 #define CEPH_MDS_STATE_STARTING -7 /* up, starting previously stopped mds */ 242 #define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */ 243 #define CEPH_MDS_STATE_REPLAYONCE -9 /* up, replaying an active node's journal */ 244 245 #define CEPH_MDS_STATE_REPLAY 8 /* up, replaying journal. */ 246 #define CEPH_MDS_STATE_RESOLVE 9 /* up, disambiguating distributed 247 operations (import, rename, etc.) */ 248 #define CEPH_MDS_STATE_RECONNECT 10 /* up, reconnect to clients */ 249 #define CEPH_MDS_STATE_REJOIN 11 /* up, rejoining distributed cache */ 250 #define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */ 251 #define CEPH_MDS_STATE_ACTIVE 13 /* up, active */ 252 #define CEPH_MDS_STATE_STOPPING 14 /* up, but exporting metadata */ 253 254 extern const char *ceph_mds_state_name(int s); 255 256 257 /* 258 * metadata lock types. 259 * - these are bitmasks.. we can compose them 260 * - they also define the lock ordering by the MDS 261 * - a few of these are internal to the mds 262 */ 263 #define CEPH_LOCK_DVERSION 1 264 #define CEPH_LOCK_DN 2 265 #define CEPH_LOCK_ISNAP 16 266 #define CEPH_LOCK_IVERSION 32 /* mds internal */ 267 #define CEPH_LOCK_IFILE 64 268 #define CEPH_LOCK_IAUTH 128 269 #define CEPH_LOCK_ILINK 256 270 #define CEPH_LOCK_IDFT 512 /* dir frag tree */ 271 #define CEPH_LOCK_INEST 1024 /* mds internal */ 272 #define CEPH_LOCK_IXATTR 2048 273 #define CEPH_LOCK_IFLOCK 4096 /* advisory file locks */ 274 #define CEPH_LOCK_INO 8192 /* immutable inode bits; not a lock */ 275 #define CEPH_LOCK_IPOLICY 16384 /* policy lock on dirs. MDS internal */ 276 277 /* client_session ops */ 278 enum { 279 CEPH_SESSION_REQUEST_OPEN, 280 CEPH_SESSION_OPEN, 281 CEPH_SESSION_REQUEST_CLOSE, 282 CEPH_SESSION_CLOSE, 283 CEPH_SESSION_REQUEST_RENEWCAPS, 284 CEPH_SESSION_RENEWCAPS, 285 CEPH_SESSION_STALE, 286 CEPH_SESSION_RECALL_STATE, 287 CEPH_SESSION_FLUSHMSG, 288 CEPH_SESSION_FLUSHMSG_ACK, 289 CEPH_SESSION_FORCE_RO, 290 CEPH_SESSION_REJECT, 291 CEPH_SESSION_REQUEST_FLUSH_MDLOG, 292 }; 293 294 extern const char *ceph_session_op_name(int op); 295 296 struct ceph_mds_session_head { 297 __le32 op; 298 __le64 seq; 299 struct ceph_timespec stamp; 300 __le32 max_caps, max_leases; 301 } __attribute__ ((packed)); 302 303 /* client_request */ 304 /* 305 * metadata ops. 306 * & 0x001000 -> write op 307 * & 0x010000 -> follow symlink (e.g. stat(), not lstat()). 308 & & 0x100000 -> use weird ino/path trace 309 */ 310 #define CEPH_MDS_OP_WRITE 0x001000 311 enum { 312 CEPH_MDS_OP_LOOKUP = 0x00100, 313 CEPH_MDS_OP_GETATTR = 0x00101, 314 CEPH_MDS_OP_LOOKUPHASH = 0x00102, 315 CEPH_MDS_OP_LOOKUPPARENT = 0x00103, 316 CEPH_MDS_OP_LOOKUPINO = 0x00104, 317 CEPH_MDS_OP_LOOKUPNAME = 0x00105, 318 319 CEPH_MDS_OP_SETXATTR = 0x01105, 320 CEPH_MDS_OP_RMXATTR = 0x01106, 321 CEPH_MDS_OP_SETLAYOUT = 0x01107, 322 CEPH_MDS_OP_SETATTR = 0x01108, 323 CEPH_MDS_OP_SETFILELOCK= 0x01109, 324 CEPH_MDS_OP_GETFILELOCK= 0x00110, 325 CEPH_MDS_OP_SETDIRLAYOUT=0x0110a, 326 327 CEPH_MDS_OP_MKNOD = 0x01201, 328 CEPH_MDS_OP_LINK = 0x01202, 329 CEPH_MDS_OP_UNLINK = 0x01203, 330 CEPH_MDS_OP_RENAME = 0x01204, 331 CEPH_MDS_OP_MKDIR = 0x01220, 332 CEPH_MDS_OP_RMDIR = 0x01221, 333 CEPH_MDS_OP_SYMLINK = 0x01222, 334 335 CEPH_MDS_OP_CREATE = 0x01301, 336 CEPH_MDS_OP_OPEN = 0x00302, 337 CEPH_MDS_OP_READDIR = 0x00305, 338 339 CEPH_MDS_OP_LOOKUPSNAP = 0x00400, 340 CEPH_MDS_OP_MKSNAP = 0x01400, 341 CEPH_MDS_OP_RMSNAP = 0x01401, 342 CEPH_MDS_OP_LSSNAP = 0x00402, 343 CEPH_MDS_OP_RENAMESNAP = 0x01403, 344 }; 345 346 extern const char *ceph_mds_op_name(int op); 347 348 349 #define CEPH_SETATTR_MODE 1 350 #define CEPH_SETATTR_UID 2 351 #define CEPH_SETATTR_GID 4 352 #define CEPH_SETATTR_MTIME 8 353 #define CEPH_SETATTR_ATIME 16 354 #define CEPH_SETATTR_SIZE 32 355 #define CEPH_SETATTR_CTIME 64 356 357 /* 358 * Ceph setxattr request flags. 359 */ 360 #define CEPH_XATTR_CREATE (1 << 0) 361 #define CEPH_XATTR_REPLACE (1 << 1) 362 #define CEPH_XATTR_REMOVE (1 << 31) 363 364 /* 365 * readdir request flags; 366 */ 367 #define CEPH_READDIR_REPLY_BITFLAGS (1<<0) 368 369 /* 370 * readdir reply flags. 371 */ 372 #define CEPH_READDIR_FRAG_END (1<<0) 373 #define CEPH_READDIR_FRAG_COMPLETE (1<<8) 374 #define CEPH_READDIR_HASH_ORDER (1<<9) 375 #define CEPH_READDIR_OFFSET_HASH (1<<10) 376 377 /* 378 * open request flags 379 */ 380 #define CEPH_O_RDONLY 00000000 381 #define CEPH_O_WRONLY 00000001 382 #define CEPH_O_RDWR 00000002 383 #define CEPH_O_CREAT 00000100 384 #define CEPH_O_EXCL 00000200 385 #define CEPH_O_TRUNC 00001000 386 #define CEPH_O_DIRECTORY 00200000 387 #define CEPH_O_NOFOLLOW 00400000 388 389 union ceph_mds_request_args { 390 struct { 391 __le32 mask; /* CEPH_CAP_* */ 392 } __attribute__ ((packed)) getattr; 393 struct { 394 __le32 mode; 395 __le32 uid; 396 __le32 gid; 397 struct ceph_timespec mtime; 398 struct ceph_timespec atime; 399 __le64 size, old_size; /* old_size needed by truncate */ 400 __le32 mask; /* CEPH_SETATTR_* */ 401 } __attribute__ ((packed)) setattr; 402 struct { 403 __le32 frag; /* which dir fragment */ 404 __le32 max_entries; /* how many dentries to grab */ 405 __le32 max_bytes; 406 __le16 flags; 407 __le32 offset_hash; 408 } __attribute__ ((packed)) readdir; 409 struct { 410 __le32 mode; 411 __le32 rdev; 412 } __attribute__ ((packed)) mknod; 413 struct { 414 __le32 mode; 415 } __attribute__ ((packed)) mkdir; 416 struct { 417 __le32 flags; 418 __le32 mode; 419 __le32 stripe_unit; /* layout for newly created file */ 420 __le32 stripe_count; /* ... */ 421 __le32 object_size; 422 __le32 file_replication; 423 __le32 mask; /* CEPH_CAP_* */ 424 __le32 old_size; 425 } __attribute__ ((packed)) open; 426 struct { 427 __le32 flags; 428 } __attribute__ ((packed)) setxattr; 429 struct { 430 struct ceph_file_layout_legacy layout; 431 } __attribute__ ((packed)) setlayout; 432 struct { 433 __u8 rule; /* currently fcntl or flock */ 434 __u8 type; /* shared, exclusive, remove*/ 435 __le64 owner; /* owner of the lock */ 436 __le64 pid; /* process id requesting the lock */ 437 __le64 start; /* initial location to lock */ 438 __le64 length; /* num bytes to lock from start */ 439 __u8 wait; /* will caller wait for lock to become available? */ 440 } __attribute__ ((packed)) filelock_change; 441 struct { 442 __le32 mask; /* CEPH_CAP_* */ 443 __le64 snapid; 444 __le64 parent; 445 __le32 hash; 446 } __attribute__ ((packed)) lookupino; 447 } __attribute__ ((packed)); 448 449 #define CEPH_MDS_FLAG_REPLAY 1 /* this is a replayed op */ 450 #define CEPH_MDS_FLAG_WANT_DENTRY 2 /* want dentry in reply */ 451 #define CEPH_MDS_FLAG_ASYNC 4 /* request is asynchronous */ 452 453 struct ceph_mds_request_head { 454 __le64 oldest_client_tid; 455 __le32 mdsmap_epoch; /* on client */ 456 __le32 flags; /* CEPH_MDS_FLAG_* */ 457 __u8 num_retry, num_fwd; /* count retry, fwd attempts */ 458 __le16 num_releases; /* # include cap/lease release records */ 459 __le32 op; /* mds op code */ 460 __le32 caller_uid, caller_gid; 461 __le64 ino; /* use this ino for openc, mkdir, mknod, 462 etc. (if replaying) */ 463 union ceph_mds_request_args args; 464 } __attribute__ ((packed)); 465 466 /* cap/lease release record */ 467 struct ceph_mds_request_release { 468 __le64 ino, cap_id; /* ino and unique cap id */ 469 __le32 caps, wanted; /* new issued, wanted */ 470 __le32 seq, issue_seq, mseq; 471 __le32 dname_seq; /* if releasing a dentry lease, a */ 472 __le32 dname_len; /* string follows. */ 473 } __attribute__ ((packed)); 474 475 /* client reply */ 476 struct ceph_mds_reply_head { 477 __le32 op; 478 __le32 result; 479 __le32 mdsmap_epoch; 480 __u8 safe; /* true if committed to disk */ 481 __u8 is_dentry, is_target; /* true if dentry, target inode records 482 are included with reply */ 483 } __attribute__ ((packed)); 484 485 /* one for each node split */ 486 struct ceph_frag_tree_split { 487 __le32 frag; /* this frag splits... */ 488 __le32 by; /* ...by this many bits */ 489 } __attribute__ ((packed)); 490 491 struct ceph_frag_tree_head { 492 __le32 nsplits; /* num ceph_frag_tree_split records */ 493 struct ceph_frag_tree_split splits[]; 494 } __attribute__ ((packed)); 495 496 /* capability issue, for bundling with mds reply */ 497 struct ceph_mds_reply_cap { 498 __le32 caps, wanted; /* caps issued, wanted */ 499 __le64 cap_id; 500 __le32 seq, mseq; 501 __le64 realm; /* snap realm */ 502 __u8 flags; /* CEPH_CAP_FLAG_* */ 503 } __attribute__ ((packed)); 504 505 #define CEPH_CAP_FLAG_AUTH (1 << 0) /* cap is issued by auth mds */ 506 #define CEPH_CAP_FLAG_RELEASE (1 << 1) /* release the cap */ 507 508 /* inode record, for bundling with mds reply */ 509 struct ceph_mds_reply_inode { 510 __le64 ino; 511 __le64 snapid; 512 __le32 rdev; 513 __le64 version; /* inode version */ 514 __le64 xattr_version; /* version for xattr blob */ 515 struct ceph_mds_reply_cap cap; /* caps issued for this inode */ 516 struct ceph_file_layout_legacy layout; 517 struct ceph_timespec ctime, mtime, atime; 518 __le32 time_warp_seq; 519 __le64 size, max_size, truncate_size; 520 __le32 truncate_seq; 521 __le32 mode, uid, gid; 522 __le32 nlink; 523 __le64 files, subdirs, rbytes, rfiles, rsubdirs; /* dir stats */ 524 struct ceph_timespec rctime; 525 struct ceph_frag_tree_head fragtree; /* (must be at end of struct) */ 526 } __attribute__ ((packed)); 527 /* followed by frag array, symlink string, dir layout, xattr blob */ 528 529 /* reply_lease follows dname, and reply_inode */ 530 struct ceph_mds_reply_lease { 531 __le16 mask; /* lease type(s) */ 532 __le32 duration_ms; /* lease duration */ 533 __le32 seq; 534 } __attribute__ ((packed)); 535 536 #define CEPH_LEASE_VALID (1 | 2) /* old and new bit values */ 537 #define CEPH_LEASE_PRIMARY_LINK 4 /* primary linkage */ 538 539 struct ceph_mds_reply_dirfrag { 540 __le32 frag; /* fragment */ 541 __le32 auth; /* auth mds, if this is a delegation point */ 542 __le32 ndist; /* number of mds' this is replicated on */ 543 __le32 dist[]; 544 } __attribute__ ((packed)); 545 546 #define CEPH_LOCK_FCNTL 1 547 #define CEPH_LOCK_FLOCK 2 548 #define CEPH_LOCK_FCNTL_INTR 3 549 #define CEPH_LOCK_FLOCK_INTR 4 550 551 552 #define CEPH_LOCK_SHARED 1 553 #define CEPH_LOCK_EXCL 2 554 #define CEPH_LOCK_UNLOCK 4 555 556 struct ceph_filelock { 557 __le64 start;/* file offset to start lock at */ 558 __le64 length; /* num bytes to lock; 0 for all following start */ 559 __le64 client; /* which client holds the lock */ 560 __le64 owner; /* owner the lock */ 561 __le64 pid; /* process id holding the lock on the client */ 562 __u8 type; /* shared lock, exclusive lock, or unlock */ 563 } __attribute__ ((packed)); 564 565 566 /* file access modes */ 567 #define CEPH_FILE_MODE_PIN 0 568 #define CEPH_FILE_MODE_RD 1 569 #define CEPH_FILE_MODE_WR 2 570 #define CEPH_FILE_MODE_RDWR 3 /* RD | WR */ 571 #define CEPH_FILE_MODE_LAZY 4 /* lazy io */ 572 #define CEPH_FILE_MODE_BITS 4 573 #define CEPH_FILE_MODE_MASK ((1 << CEPH_FILE_MODE_BITS) - 1) 574 575 int ceph_flags_to_mode(int flags); 576 577 #define CEPH_INLINE_NONE ((__u64)-1) 578 579 /* capability bits */ 580 #define CEPH_CAP_PIN 1 /* no specific capabilities beyond the pin */ 581 582 /* generic cap bits */ 583 #define CEPH_CAP_GSHARED 1 /* client can reads */ 584 #define CEPH_CAP_GEXCL 2 /* client can read and update */ 585 #define CEPH_CAP_GCACHE 4 /* (file) client can cache reads */ 586 #define CEPH_CAP_GRD 8 /* (file) client can read */ 587 #define CEPH_CAP_GWR 16 /* (file) client can write */ 588 #define CEPH_CAP_GBUFFER 32 /* (file) client can buffer writes */ 589 #define CEPH_CAP_GWREXTEND 64 /* (file) client can extend EOF */ 590 #define CEPH_CAP_GLAZYIO 128 /* (file) client can perform lazy io */ 591 592 #define CEPH_CAP_SIMPLE_BITS 2 593 #define CEPH_CAP_FILE_BITS 8 594 595 /* per-lock shift */ 596 #define CEPH_CAP_SAUTH 2 597 #define CEPH_CAP_SLINK 4 598 #define CEPH_CAP_SXATTR 6 599 #define CEPH_CAP_SFILE 8 600 #define CEPH_CAP_SFLOCK 20 601 602 #define CEPH_CAP_BITS 22 603 604 /* composed values */ 605 #define CEPH_CAP_AUTH_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SAUTH) 606 #define CEPH_CAP_AUTH_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SAUTH) 607 #define CEPH_CAP_LINK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SLINK) 608 #define CEPH_CAP_LINK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SLINK) 609 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SXATTR) 610 #define CEPH_CAP_XATTR_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SXATTR) 611 #define CEPH_CAP_FILE(x) (x << CEPH_CAP_SFILE) 612 #define CEPH_CAP_FILE_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFILE) 613 #define CEPH_CAP_FILE_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFILE) 614 #define CEPH_CAP_FILE_CACHE (CEPH_CAP_GCACHE << CEPH_CAP_SFILE) 615 #define CEPH_CAP_FILE_RD (CEPH_CAP_GRD << CEPH_CAP_SFILE) 616 #define CEPH_CAP_FILE_WR (CEPH_CAP_GWR << CEPH_CAP_SFILE) 617 #define CEPH_CAP_FILE_BUFFER (CEPH_CAP_GBUFFER << CEPH_CAP_SFILE) 618 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE) 619 #define CEPH_CAP_FILE_LAZYIO (CEPH_CAP_GLAZYIO << CEPH_CAP_SFILE) 620 #define CEPH_CAP_FLOCK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFLOCK) 621 #define CEPH_CAP_FLOCK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFLOCK) 622 623 624 /* cap masks (for getattr) */ 625 #define CEPH_STAT_CAP_INODE CEPH_CAP_PIN 626 #define CEPH_STAT_CAP_TYPE CEPH_CAP_PIN /* mode >> 12 */ 627 #define CEPH_STAT_CAP_SYMLINK CEPH_CAP_PIN 628 #define CEPH_STAT_CAP_UID CEPH_CAP_AUTH_SHARED 629 #define CEPH_STAT_CAP_GID CEPH_CAP_AUTH_SHARED 630 #define CEPH_STAT_CAP_MODE CEPH_CAP_AUTH_SHARED 631 #define CEPH_STAT_CAP_NLINK CEPH_CAP_LINK_SHARED 632 #define CEPH_STAT_CAP_LAYOUT CEPH_CAP_FILE_SHARED 633 #define CEPH_STAT_CAP_MTIME CEPH_CAP_FILE_SHARED 634 #define CEPH_STAT_CAP_SIZE CEPH_CAP_FILE_SHARED 635 #define CEPH_STAT_CAP_ATIME CEPH_CAP_FILE_SHARED /* fixme */ 636 #define CEPH_STAT_CAP_XATTR CEPH_CAP_XATTR_SHARED 637 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN | \ 638 CEPH_CAP_AUTH_SHARED | \ 639 CEPH_CAP_LINK_SHARED | \ 640 CEPH_CAP_FILE_SHARED | \ 641 CEPH_CAP_XATTR_SHARED) 642 #define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \ 643 CEPH_CAP_FILE_RD) 644 #define CEPH_STAT_RSTAT CEPH_CAP_FILE_WREXTEND 645 646 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED | \ 647 CEPH_CAP_LINK_SHARED | \ 648 CEPH_CAP_XATTR_SHARED | \ 649 CEPH_CAP_FILE_SHARED) 650 #define CEPH_CAP_ANY_RD (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD | \ 651 CEPH_CAP_FILE_CACHE) 652 653 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL | \ 654 CEPH_CAP_LINK_EXCL | \ 655 CEPH_CAP_XATTR_EXCL | \ 656 CEPH_CAP_FILE_EXCL) 657 #define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \ 658 CEPH_CAP_FILE_SHARED) 659 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \ 660 CEPH_CAP_FILE_EXCL) 661 #define CEPH_CAP_ANY_WR (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR) 662 #define CEPH_CAP_ANY (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \ 663 CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \ 664 CEPH_CAP_PIN) 665 #define CEPH_CAP_ALL_FILE (CEPH_CAP_PIN | CEPH_CAP_ANY_SHARED | \ 666 CEPH_CAP_AUTH_EXCL | CEPH_CAP_XATTR_EXCL | \ 667 CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR) 668 669 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \ 670 CEPH_LOCK_IXATTR) 671 672 /* cap masks async dir operations */ 673 #define CEPH_CAP_DIR_CREATE CEPH_CAP_FILE_CACHE 674 #define CEPH_CAP_DIR_UNLINK CEPH_CAP_FILE_RD 675 #define CEPH_CAP_ANY_DIR_OPS (CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_RD | \ 676 CEPH_CAP_FILE_WREXTEND | CEPH_CAP_FILE_LAZYIO) 677 678 int ceph_caps_for_mode(int mode); 679 680 enum { 681 CEPH_CAP_OP_GRANT, /* mds->client grant */ 682 CEPH_CAP_OP_REVOKE, /* mds->client revoke */ 683 CEPH_CAP_OP_TRUNC, /* mds->client trunc notify */ 684 CEPH_CAP_OP_EXPORT, /* mds has exported the cap */ 685 CEPH_CAP_OP_IMPORT, /* mds has imported the cap */ 686 CEPH_CAP_OP_UPDATE, /* client->mds update */ 687 CEPH_CAP_OP_DROP, /* client->mds drop cap bits */ 688 CEPH_CAP_OP_FLUSH, /* client->mds cap writeback */ 689 CEPH_CAP_OP_FLUSH_ACK, /* mds->client flushed */ 690 CEPH_CAP_OP_FLUSHSNAP, /* client->mds flush snapped metadata */ 691 CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */ 692 CEPH_CAP_OP_RELEASE, /* client->mds release (clean) cap */ 693 CEPH_CAP_OP_RENEW, /* client->mds renewal request */ 694 }; 695 696 extern const char *ceph_cap_op_name(int op); 697 698 /* flags field in client cap messages (version >= 10) */ 699 #define CEPH_CLIENT_CAPS_SYNC (1<<0) 700 #define CEPH_CLIENT_CAPS_NO_CAPSNAP (1<<1) 701 #define CEPH_CLIENT_CAPS_PENDING_CAPSNAP (1<<2) 702 703 /* 704 * caps message, used for capability callbacks, acks, requests, etc. 705 */ 706 struct ceph_mds_caps { 707 __le32 op; /* CEPH_CAP_OP_* */ 708 __le64 ino, realm; 709 __le64 cap_id; 710 __le32 seq, issue_seq; 711 __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */ 712 __le32 migrate_seq; 713 __le64 snap_follows; 714 __le32 snap_trace_len; 715 716 /* authlock */ 717 __le32 uid, gid, mode; 718 719 /* linklock */ 720 __le32 nlink; 721 722 /* xattrlock */ 723 __le32 xattr_len; 724 __le64 xattr_version; 725 726 /* filelock */ 727 __le64 size, max_size, truncate_size; 728 __le32 truncate_seq; 729 struct ceph_timespec mtime, atime, ctime; 730 struct ceph_file_layout_legacy layout; 731 __le32 time_warp_seq; 732 } __attribute__ ((packed)); 733 734 struct ceph_mds_cap_peer { 735 __le64 cap_id; 736 __le32 seq; 737 __le32 mseq; 738 __le32 mds; 739 __u8 flags; 740 } __attribute__ ((packed)); 741 742 /* cap release msg head */ 743 struct ceph_mds_cap_release { 744 __le32 num; /* number of cap_items that follow */ 745 } __attribute__ ((packed)); 746 747 struct ceph_mds_cap_item { 748 __le64 ino; 749 __le64 cap_id; 750 __le32 migrate_seq, seq; 751 } __attribute__ ((packed)); 752 753 #define CEPH_MDS_LEASE_REVOKE 1 /* mds -> client */ 754 #define CEPH_MDS_LEASE_RELEASE 2 /* client -> mds */ 755 #define CEPH_MDS_LEASE_RENEW 3 /* client <-> mds */ 756 #define CEPH_MDS_LEASE_REVOKE_ACK 4 /* client -> mds */ 757 758 extern const char *ceph_lease_op_name(int o); 759 760 /* lease msg header */ 761 struct ceph_mds_lease { 762 __u8 action; /* CEPH_MDS_LEASE_* */ 763 __le16 mask; /* which lease */ 764 __le64 ino; 765 __le64 first, last; /* snap range */ 766 __le32 seq; 767 __le32 duration_ms; /* duration of renewal */ 768 } __attribute__ ((packed)); 769 /* followed by a __le32+string for dname */ 770 771 /* client reconnect */ 772 struct ceph_mds_cap_reconnect { 773 __le64 cap_id; 774 __le32 wanted; 775 __le32 issued; 776 __le64 snaprealm; 777 __le64 pathbase; /* base ino for our path to this ino */ 778 __le32 flock_len; /* size of flock state blob, if any */ 779 } __attribute__ ((packed)); 780 /* followed by flock blob */ 781 782 struct ceph_mds_cap_reconnect_v1 { 783 __le64 cap_id; 784 __le32 wanted; 785 __le32 issued; 786 __le64 size; 787 struct ceph_timespec mtime, atime; 788 __le64 snaprealm; 789 __le64 pathbase; /* base ino for our path to this ino */ 790 } __attribute__ ((packed)); 791 792 struct ceph_mds_snaprealm_reconnect { 793 __le64 ino; /* snap realm base */ 794 __le64 seq; /* snap seq for this snap realm */ 795 __le64 parent; /* parent realm */ 796 } __attribute__ ((packed)); 797 798 /* 799 * snaps 800 */ 801 enum { 802 CEPH_SNAP_OP_UPDATE, /* CREATE or DESTROY */ 803 CEPH_SNAP_OP_CREATE, 804 CEPH_SNAP_OP_DESTROY, 805 CEPH_SNAP_OP_SPLIT, 806 }; 807 808 extern const char *ceph_snap_op_name(int o); 809 810 /* snap msg header */ 811 struct ceph_mds_snap_head { 812 __le32 op; /* CEPH_SNAP_OP_* */ 813 __le64 split; /* ino to split off, if any */ 814 __le32 num_split_inos; /* # inos belonging to new child realm */ 815 __le32 num_split_realms; /* # child realms udner new child realm */ 816 __le32 trace_len; /* size of snap trace blob */ 817 } __attribute__ ((packed)); 818 /* followed by split ino list, then split realms, then the trace blob */ 819 820 /* 821 * encode info about a snaprealm, as viewed by a client 822 */ 823 struct ceph_mds_snap_realm { 824 __le64 ino; /* ino */ 825 __le64 created; /* snap: when created */ 826 __le64 parent; /* ino: parent realm */ 827 __le64 parent_since; /* snap: same parent since */ 828 __le64 seq; /* snap: version */ 829 __le32 num_snaps; 830 __le32 num_prior_parent_snaps; 831 } __attribute__ ((packed)); 832 /* followed by my snap list, then prior parent snap list */ 833 834 /* 835 * quotas 836 */ 837 struct ceph_mds_quota { 838 __le64 ino; /* ino */ 839 struct ceph_timespec rctime; 840 __le64 rbytes; /* dir stats */ 841 __le64 rfiles; 842 __le64 rsubdirs; 843 __u8 struct_v; /* compat */ 844 __u8 struct_compat; 845 __le32 struct_len; 846 __le64 max_bytes; /* quota max. bytes */ 847 __le64 max_files; /* quota max. files */ 848 } __attribute__ ((packed)); 849 850 #endif 851