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 }; 292 293 extern const char *ceph_session_op_name(int op); 294 295 struct ceph_mds_session_head { 296 __le32 op; 297 __le64 seq; 298 struct ceph_timespec stamp; 299 __le32 max_caps, max_leases; 300 } __attribute__ ((packed)); 301 302 /* client_request */ 303 /* 304 * metadata ops. 305 * & 0x001000 -> write op 306 * & 0x010000 -> follow symlink (e.g. stat(), not lstat()). 307 & & 0x100000 -> use weird ino/path trace 308 */ 309 #define CEPH_MDS_OP_WRITE 0x001000 310 enum { 311 CEPH_MDS_OP_LOOKUP = 0x00100, 312 CEPH_MDS_OP_GETATTR = 0x00101, 313 CEPH_MDS_OP_LOOKUPHASH = 0x00102, 314 CEPH_MDS_OP_LOOKUPPARENT = 0x00103, 315 CEPH_MDS_OP_LOOKUPINO = 0x00104, 316 CEPH_MDS_OP_LOOKUPNAME = 0x00105, 317 318 CEPH_MDS_OP_SETXATTR = 0x01105, 319 CEPH_MDS_OP_RMXATTR = 0x01106, 320 CEPH_MDS_OP_SETLAYOUT = 0x01107, 321 CEPH_MDS_OP_SETATTR = 0x01108, 322 CEPH_MDS_OP_SETFILELOCK= 0x01109, 323 CEPH_MDS_OP_GETFILELOCK= 0x00110, 324 CEPH_MDS_OP_SETDIRLAYOUT=0x0110a, 325 326 CEPH_MDS_OP_MKNOD = 0x01201, 327 CEPH_MDS_OP_LINK = 0x01202, 328 CEPH_MDS_OP_UNLINK = 0x01203, 329 CEPH_MDS_OP_RENAME = 0x01204, 330 CEPH_MDS_OP_MKDIR = 0x01220, 331 CEPH_MDS_OP_RMDIR = 0x01221, 332 CEPH_MDS_OP_SYMLINK = 0x01222, 333 334 CEPH_MDS_OP_CREATE = 0x01301, 335 CEPH_MDS_OP_OPEN = 0x00302, 336 CEPH_MDS_OP_READDIR = 0x00305, 337 338 CEPH_MDS_OP_LOOKUPSNAP = 0x00400, 339 CEPH_MDS_OP_MKSNAP = 0x01400, 340 CEPH_MDS_OP_RMSNAP = 0x01401, 341 CEPH_MDS_OP_LSSNAP = 0x00402, 342 CEPH_MDS_OP_RENAMESNAP = 0x01403, 343 }; 344 345 extern const char *ceph_mds_op_name(int op); 346 347 348 #define CEPH_SETATTR_MODE 1 349 #define CEPH_SETATTR_UID 2 350 #define CEPH_SETATTR_GID 4 351 #define CEPH_SETATTR_MTIME 8 352 #define CEPH_SETATTR_ATIME 16 353 #define CEPH_SETATTR_SIZE 32 354 #define CEPH_SETATTR_CTIME 64 355 356 /* 357 * Ceph setxattr request flags. 358 */ 359 #define CEPH_XATTR_CREATE (1 << 0) 360 #define CEPH_XATTR_REPLACE (1 << 1) 361 #define CEPH_XATTR_REMOVE (1 << 31) 362 363 /* 364 * readdir request flags; 365 */ 366 #define CEPH_READDIR_REPLY_BITFLAGS (1<<0) 367 368 /* 369 * readdir reply flags. 370 */ 371 #define CEPH_READDIR_FRAG_END (1<<0) 372 #define CEPH_READDIR_FRAG_COMPLETE (1<<8) 373 #define CEPH_READDIR_HASH_ORDER (1<<9) 374 #define CEPH_READDIR_OFFSET_HASH (1<<10) 375 376 /* 377 * open request flags 378 */ 379 #define CEPH_O_RDONLY 00000000 380 #define CEPH_O_WRONLY 00000001 381 #define CEPH_O_RDWR 00000002 382 #define CEPH_O_CREAT 00000100 383 #define CEPH_O_EXCL 00000200 384 #define CEPH_O_TRUNC 00001000 385 #define CEPH_O_DIRECTORY 00200000 386 #define CEPH_O_NOFOLLOW 00400000 387 388 union ceph_mds_request_args { 389 struct { 390 __le32 mask; /* CEPH_CAP_* */ 391 } __attribute__ ((packed)) getattr; 392 struct { 393 __le32 mode; 394 __le32 uid; 395 __le32 gid; 396 struct ceph_timespec mtime; 397 struct ceph_timespec atime; 398 __le64 size, old_size; /* old_size needed by truncate */ 399 __le32 mask; /* CEPH_SETATTR_* */ 400 } __attribute__ ((packed)) setattr; 401 struct { 402 __le32 frag; /* which dir fragment */ 403 __le32 max_entries; /* how many dentries to grab */ 404 __le32 max_bytes; 405 __le16 flags; 406 __le32 offset_hash; 407 } __attribute__ ((packed)) readdir; 408 struct { 409 __le32 mode; 410 __le32 rdev; 411 } __attribute__ ((packed)) mknod; 412 struct { 413 __le32 mode; 414 } __attribute__ ((packed)) mkdir; 415 struct { 416 __le32 flags; 417 __le32 mode; 418 __le32 stripe_unit; /* layout for newly created file */ 419 __le32 stripe_count; /* ... */ 420 __le32 object_size; 421 __le32 file_replication; 422 __le32 mask; /* CEPH_CAP_* */ 423 __le32 old_size; 424 } __attribute__ ((packed)) open; 425 struct { 426 __le32 flags; 427 } __attribute__ ((packed)) setxattr; 428 struct { 429 struct ceph_file_layout_legacy layout; 430 } __attribute__ ((packed)) setlayout; 431 struct { 432 __u8 rule; /* currently fcntl or flock */ 433 __u8 type; /* shared, exclusive, remove*/ 434 __le64 owner; /* owner of the lock */ 435 __le64 pid; /* process id requesting the lock */ 436 __le64 start; /* initial location to lock */ 437 __le64 length; /* num bytes to lock from start */ 438 __u8 wait; /* will caller wait for lock to become available? */ 439 } __attribute__ ((packed)) filelock_change; 440 struct { 441 __le32 mask; /* CEPH_CAP_* */ 442 __le64 snapid; 443 __le64 parent; 444 __le32 hash; 445 } __attribute__ ((packed)) lookupino; 446 } __attribute__ ((packed)); 447 448 #define CEPH_MDS_FLAG_REPLAY 1 /* this is a replayed op */ 449 #define CEPH_MDS_FLAG_WANT_DENTRY 2 /* want dentry in reply */ 450 #define CEPH_MDS_FLAG_ASYNC 4 /* request is asynchronous */ 451 452 struct ceph_mds_request_head { 453 __le64 oldest_client_tid; 454 __le32 mdsmap_epoch; /* on client */ 455 __le32 flags; /* CEPH_MDS_FLAG_* */ 456 __u8 num_retry, num_fwd; /* count retry, fwd attempts */ 457 __le16 num_releases; /* # include cap/lease release records */ 458 __le32 op; /* mds op code */ 459 __le32 caller_uid, caller_gid; 460 __le64 ino; /* use this ino for openc, mkdir, mknod, 461 etc. (if replaying) */ 462 union ceph_mds_request_args args; 463 } __attribute__ ((packed)); 464 465 /* cap/lease release record */ 466 struct ceph_mds_request_release { 467 __le64 ino, cap_id; /* ino and unique cap id */ 468 __le32 caps, wanted; /* new issued, wanted */ 469 __le32 seq, issue_seq, mseq; 470 __le32 dname_seq; /* if releasing a dentry lease, a */ 471 __le32 dname_len; /* string follows. */ 472 } __attribute__ ((packed)); 473 474 /* client reply */ 475 struct ceph_mds_reply_head { 476 __le32 op; 477 __le32 result; 478 __le32 mdsmap_epoch; 479 __u8 safe; /* true if committed to disk */ 480 __u8 is_dentry, is_target; /* true if dentry, target inode records 481 are included with reply */ 482 } __attribute__ ((packed)); 483 484 /* one for each node split */ 485 struct ceph_frag_tree_split { 486 __le32 frag; /* this frag splits... */ 487 __le32 by; /* ...by this many bits */ 488 } __attribute__ ((packed)); 489 490 struct ceph_frag_tree_head { 491 __le32 nsplits; /* num ceph_frag_tree_split records */ 492 struct ceph_frag_tree_split splits[]; 493 } __attribute__ ((packed)); 494 495 /* capability issue, for bundling with mds reply */ 496 struct ceph_mds_reply_cap { 497 __le32 caps, wanted; /* caps issued, wanted */ 498 __le64 cap_id; 499 __le32 seq, mseq; 500 __le64 realm; /* snap realm */ 501 __u8 flags; /* CEPH_CAP_FLAG_* */ 502 } __attribute__ ((packed)); 503 504 #define CEPH_CAP_FLAG_AUTH (1 << 0) /* cap is issued by auth mds */ 505 #define CEPH_CAP_FLAG_RELEASE (1 << 1) /* release the cap */ 506 507 /* inode record, for bundling with mds reply */ 508 struct ceph_mds_reply_inode { 509 __le64 ino; 510 __le64 snapid; 511 __le32 rdev; 512 __le64 version; /* inode version */ 513 __le64 xattr_version; /* version for xattr blob */ 514 struct ceph_mds_reply_cap cap; /* caps issued for this inode */ 515 struct ceph_file_layout_legacy layout; 516 struct ceph_timespec ctime, mtime, atime; 517 __le32 time_warp_seq; 518 __le64 size, max_size, truncate_size; 519 __le32 truncate_seq; 520 __le32 mode, uid, gid; 521 __le32 nlink; 522 __le64 files, subdirs, rbytes, rfiles, rsubdirs; /* dir stats */ 523 struct ceph_timespec rctime; 524 struct ceph_frag_tree_head fragtree; /* (must be at end of struct) */ 525 } __attribute__ ((packed)); 526 /* followed by frag array, symlink string, dir layout, xattr blob */ 527 528 /* reply_lease follows dname, and reply_inode */ 529 struct ceph_mds_reply_lease { 530 __le16 mask; /* lease type(s) */ 531 __le32 duration_ms; /* lease duration */ 532 __le32 seq; 533 } __attribute__ ((packed)); 534 535 #define CEPH_LEASE_VALID (1 | 2) /* old and new bit values */ 536 #define CEPH_LEASE_PRIMARY_LINK 4 /* primary linkage */ 537 538 struct ceph_mds_reply_dirfrag { 539 __le32 frag; /* fragment */ 540 __le32 auth; /* auth mds, if this is a delegation point */ 541 __le32 ndist; /* number of mds' this is replicated on */ 542 __le32 dist[]; 543 } __attribute__ ((packed)); 544 545 #define CEPH_LOCK_FCNTL 1 546 #define CEPH_LOCK_FLOCK 2 547 #define CEPH_LOCK_FCNTL_INTR 3 548 #define CEPH_LOCK_FLOCK_INTR 4 549 550 551 #define CEPH_LOCK_SHARED 1 552 #define CEPH_LOCK_EXCL 2 553 #define CEPH_LOCK_UNLOCK 4 554 555 struct ceph_filelock { 556 __le64 start;/* file offset to start lock at */ 557 __le64 length; /* num bytes to lock; 0 for all following start */ 558 __le64 client; /* which client holds the lock */ 559 __le64 owner; /* owner the lock */ 560 __le64 pid; /* process id holding the lock on the client */ 561 __u8 type; /* shared lock, exclusive lock, or unlock */ 562 } __attribute__ ((packed)); 563 564 565 /* file access modes */ 566 #define CEPH_FILE_MODE_PIN 0 567 #define CEPH_FILE_MODE_RD 1 568 #define CEPH_FILE_MODE_WR 2 569 #define CEPH_FILE_MODE_RDWR 3 /* RD | WR */ 570 #define CEPH_FILE_MODE_LAZY 4 /* lazy io */ 571 #define CEPH_FILE_MODE_BITS 4 572 #define CEPH_FILE_MODE_MASK ((1 << CEPH_FILE_MODE_BITS) - 1) 573 574 int ceph_flags_to_mode(int flags); 575 576 #define CEPH_INLINE_NONE ((__u64)-1) 577 578 /* capability bits */ 579 #define CEPH_CAP_PIN 1 /* no specific capabilities beyond the pin */ 580 581 /* generic cap bits */ 582 #define CEPH_CAP_GSHARED 1 /* client can reads */ 583 #define CEPH_CAP_GEXCL 2 /* client can read and update */ 584 #define CEPH_CAP_GCACHE 4 /* (file) client can cache reads */ 585 #define CEPH_CAP_GRD 8 /* (file) client can read */ 586 #define CEPH_CAP_GWR 16 /* (file) client can write */ 587 #define CEPH_CAP_GBUFFER 32 /* (file) client can buffer writes */ 588 #define CEPH_CAP_GWREXTEND 64 /* (file) client can extend EOF */ 589 #define CEPH_CAP_GLAZYIO 128 /* (file) client can perform lazy io */ 590 591 #define CEPH_CAP_SIMPLE_BITS 2 592 #define CEPH_CAP_FILE_BITS 8 593 594 /* per-lock shift */ 595 #define CEPH_CAP_SAUTH 2 596 #define CEPH_CAP_SLINK 4 597 #define CEPH_CAP_SXATTR 6 598 #define CEPH_CAP_SFILE 8 599 #define CEPH_CAP_SFLOCK 20 600 601 #define CEPH_CAP_BITS 22 602 603 /* composed values */ 604 #define CEPH_CAP_AUTH_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SAUTH) 605 #define CEPH_CAP_AUTH_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SAUTH) 606 #define CEPH_CAP_LINK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SLINK) 607 #define CEPH_CAP_LINK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SLINK) 608 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SXATTR) 609 #define CEPH_CAP_XATTR_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SXATTR) 610 #define CEPH_CAP_FILE(x) (x << CEPH_CAP_SFILE) 611 #define CEPH_CAP_FILE_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFILE) 612 #define CEPH_CAP_FILE_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFILE) 613 #define CEPH_CAP_FILE_CACHE (CEPH_CAP_GCACHE << CEPH_CAP_SFILE) 614 #define CEPH_CAP_FILE_RD (CEPH_CAP_GRD << CEPH_CAP_SFILE) 615 #define CEPH_CAP_FILE_WR (CEPH_CAP_GWR << CEPH_CAP_SFILE) 616 #define CEPH_CAP_FILE_BUFFER (CEPH_CAP_GBUFFER << CEPH_CAP_SFILE) 617 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE) 618 #define CEPH_CAP_FILE_LAZYIO (CEPH_CAP_GLAZYIO << CEPH_CAP_SFILE) 619 #define CEPH_CAP_FLOCK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFLOCK) 620 #define CEPH_CAP_FLOCK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFLOCK) 621 622 623 /* cap masks (for getattr) */ 624 #define CEPH_STAT_CAP_INODE CEPH_CAP_PIN 625 #define CEPH_STAT_CAP_TYPE CEPH_CAP_PIN /* mode >> 12 */ 626 #define CEPH_STAT_CAP_SYMLINK CEPH_CAP_PIN 627 #define CEPH_STAT_CAP_UID CEPH_CAP_AUTH_SHARED 628 #define CEPH_STAT_CAP_GID CEPH_CAP_AUTH_SHARED 629 #define CEPH_STAT_CAP_MODE CEPH_CAP_AUTH_SHARED 630 #define CEPH_STAT_CAP_NLINK CEPH_CAP_LINK_SHARED 631 #define CEPH_STAT_CAP_LAYOUT CEPH_CAP_FILE_SHARED 632 #define CEPH_STAT_CAP_MTIME CEPH_CAP_FILE_SHARED 633 #define CEPH_STAT_CAP_SIZE CEPH_CAP_FILE_SHARED 634 #define CEPH_STAT_CAP_ATIME CEPH_CAP_FILE_SHARED /* fixme */ 635 #define CEPH_STAT_CAP_XATTR CEPH_CAP_XATTR_SHARED 636 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN | \ 637 CEPH_CAP_AUTH_SHARED | \ 638 CEPH_CAP_LINK_SHARED | \ 639 CEPH_CAP_FILE_SHARED | \ 640 CEPH_CAP_XATTR_SHARED) 641 #define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \ 642 CEPH_CAP_FILE_RD) 643 #define CEPH_STAT_RSTAT CEPH_CAP_FILE_WREXTEND 644 645 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED | \ 646 CEPH_CAP_LINK_SHARED | \ 647 CEPH_CAP_XATTR_SHARED | \ 648 CEPH_CAP_FILE_SHARED) 649 #define CEPH_CAP_ANY_RD (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD | \ 650 CEPH_CAP_FILE_CACHE) 651 652 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL | \ 653 CEPH_CAP_LINK_EXCL | \ 654 CEPH_CAP_XATTR_EXCL | \ 655 CEPH_CAP_FILE_EXCL) 656 #define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \ 657 CEPH_CAP_FILE_SHARED) 658 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \ 659 CEPH_CAP_FILE_EXCL) 660 #define CEPH_CAP_ANY_WR (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR) 661 #define CEPH_CAP_ANY (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \ 662 CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \ 663 CEPH_CAP_PIN) 664 #define CEPH_CAP_ALL_FILE (CEPH_CAP_PIN | CEPH_CAP_ANY_SHARED | \ 665 CEPH_CAP_AUTH_EXCL | CEPH_CAP_XATTR_EXCL | \ 666 CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR) 667 668 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \ 669 CEPH_LOCK_IXATTR) 670 671 /* cap masks async dir operations */ 672 #define CEPH_CAP_DIR_CREATE CEPH_CAP_FILE_CACHE 673 #define CEPH_CAP_DIR_UNLINK CEPH_CAP_FILE_RD 674 #define CEPH_CAP_ANY_DIR_OPS (CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_RD | \ 675 CEPH_CAP_FILE_WREXTEND | CEPH_CAP_FILE_LAZYIO) 676 677 int ceph_caps_for_mode(int mode); 678 679 enum { 680 CEPH_CAP_OP_GRANT, /* mds->client grant */ 681 CEPH_CAP_OP_REVOKE, /* mds->client revoke */ 682 CEPH_CAP_OP_TRUNC, /* mds->client trunc notify */ 683 CEPH_CAP_OP_EXPORT, /* mds has exported the cap */ 684 CEPH_CAP_OP_IMPORT, /* mds has imported the cap */ 685 CEPH_CAP_OP_UPDATE, /* client->mds update */ 686 CEPH_CAP_OP_DROP, /* client->mds drop cap bits */ 687 CEPH_CAP_OP_FLUSH, /* client->mds cap writeback */ 688 CEPH_CAP_OP_FLUSH_ACK, /* mds->client flushed */ 689 CEPH_CAP_OP_FLUSHSNAP, /* client->mds flush snapped metadata */ 690 CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */ 691 CEPH_CAP_OP_RELEASE, /* client->mds release (clean) cap */ 692 CEPH_CAP_OP_RENEW, /* client->mds renewal request */ 693 }; 694 695 extern const char *ceph_cap_op_name(int op); 696 697 /* flags field in client cap messages (version >= 10) */ 698 #define CEPH_CLIENT_CAPS_SYNC (1<<0) 699 #define CEPH_CLIENT_CAPS_NO_CAPSNAP (1<<1) 700 #define CEPH_CLIENT_CAPS_PENDING_CAPSNAP (1<<2) 701 702 /* 703 * caps message, used for capability callbacks, acks, requests, etc. 704 */ 705 struct ceph_mds_caps { 706 __le32 op; /* CEPH_CAP_OP_* */ 707 __le64 ino, realm; 708 __le64 cap_id; 709 __le32 seq, issue_seq; 710 __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */ 711 __le32 migrate_seq; 712 __le64 snap_follows; 713 __le32 snap_trace_len; 714 715 /* authlock */ 716 __le32 uid, gid, mode; 717 718 /* linklock */ 719 __le32 nlink; 720 721 /* xattrlock */ 722 __le32 xattr_len; 723 __le64 xattr_version; 724 725 /* filelock */ 726 __le64 size, max_size, truncate_size; 727 __le32 truncate_seq; 728 struct ceph_timespec mtime, atime, ctime; 729 struct ceph_file_layout_legacy layout; 730 __le32 time_warp_seq; 731 } __attribute__ ((packed)); 732 733 struct ceph_mds_cap_peer { 734 __le64 cap_id; 735 __le32 seq; 736 __le32 mseq; 737 __le32 mds; 738 __u8 flags; 739 } __attribute__ ((packed)); 740 741 /* cap release msg head */ 742 struct ceph_mds_cap_release { 743 __le32 num; /* number of cap_items that follow */ 744 } __attribute__ ((packed)); 745 746 struct ceph_mds_cap_item { 747 __le64 ino; 748 __le64 cap_id; 749 __le32 migrate_seq, seq; 750 } __attribute__ ((packed)); 751 752 #define CEPH_MDS_LEASE_REVOKE 1 /* mds -> client */ 753 #define CEPH_MDS_LEASE_RELEASE 2 /* client -> mds */ 754 #define CEPH_MDS_LEASE_RENEW 3 /* client <-> mds */ 755 #define CEPH_MDS_LEASE_REVOKE_ACK 4 /* client -> mds */ 756 757 extern const char *ceph_lease_op_name(int o); 758 759 /* lease msg header */ 760 struct ceph_mds_lease { 761 __u8 action; /* CEPH_MDS_LEASE_* */ 762 __le16 mask; /* which lease */ 763 __le64 ino; 764 __le64 first, last; /* snap range */ 765 __le32 seq; 766 __le32 duration_ms; /* duration of renewal */ 767 } __attribute__ ((packed)); 768 /* followed by a __le32+string for dname */ 769 770 /* client reconnect */ 771 struct ceph_mds_cap_reconnect { 772 __le64 cap_id; 773 __le32 wanted; 774 __le32 issued; 775 __le64 snaprealm; 776 __le64 pathbase; /* base ino for our path to this ino */ 777 __le32 flock_len; /* size of flock state blob, if any */ 778 } __attribute__ ((packed)); 779 /* followed by flock blob */ 780 781 struct ceph_mds_cap_reconnect_v1 { 782 __le64 cap_id; 783 __le32 wanted; 784 __le32 issued; 785 __le64 size; 786 struct ceph_timespec mtime, atime; 787 __le64 snaprealm; 788 __le64 pathbase; /* base ino for our path to this ino */ 789 } __attribute__ ((packed)); 790 791 struct ceph_mds_snaprealm_reconnect { 792 __le64 ino; /* snap realm base */ 793 __le64 seq; /* snap seq for this snap realm */ 794 __le64 parent; /* parent realm */ 795 } __attribute__ ((packed)); 796 797 /* 798 * snaps 799 */ 800 enum { 801 CEPH_SNAP_OP_UPDATE, /* CREATE or DESTROY */ 802 CEPH_SNAP_OP_CREATE, 803 CEPH_SNAP_OP_DESTROY, 804 CEPH_SNAP_OP_SPLIT, 805 }; 806 807 extern const char *ceph_snap_op_name(int o); 808 809 /* snap msg header */ 810 struct ceph_mds_snap_head { 811 __le32 op; /* CEPH_SNAP_OP_* */ 812 __le64 split; /* ino to split off, if any */ 813 __le32 num_split_inos; /* # inos belonging to new child realm */ 814 __le32 num_split_realms; /* # child realms udner new child realm */ 815 __le32 trace_len; /* size of snap trace blob */ 816 } __attribute__ ((packed)); 817 /* followed by split ino list, then split realms, then the trace blob */ 818 819 /* 820 * encode info about a snaprealm, as viewed by a client 821 */ 822 struct ceph_mds_snap_realm { 823 __le64 ino; /* ino */ 824 __le64 created; /* snap: when created */ 825 __le64 parent; /* ino: parent realm */ 826 __le64 parent_since; /* snap: same parent since */ 827 __le64 seq; /* snap: version */ 828 __le32 num_snaps; 829 __le32 num_prior_parent_snaps; 830 } __attribute__ ((packed)); 831 /* followed by my snap list, then prior parent snap list */ 832 833 /* 834 * quotas 835 */ 836 struct ceph_mds_quota { 837 __le64 ino; /* ino */ 838 struct ceph_timespec rctime; 839 __le64 rbytes; /* dir stats */ 840 __le64 rfiles; 841 __le64 rsubdirs; 842 __u8 struct_v; /* compat */ 843 __u8 struct_compat; 844 __le32 struct_len; 845 __le64 max_bytes; /* quota max. bytes */ 846 __le64 max_files; /* quota max. files */ 847 } __attribute__ ((packed)); 848 849 #endif 850
