ceph_fs.h (28424B)
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_GLOBAL_SNAPREALM 3 /* global dummy snaprealm */ 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 */ 40struct 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 57struct ceph_string; 58/* 59 * ceph_file_layout - describe data layout for a file/inode 60 */ 61struct 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 70extern int ceph_file_layout_is_valid(const struct ceph_file_layout *layout); 71extern void ceph_file_layout_from_legacy(struct ceph_file_layout *fl, 72 struct ceph_file_layout_legacy *legacy); 73extern 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 78struct 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_MODE_NONE 0 97#define CEPH_AUTH_MODE_AUTHORIZER 1 98#define CEPH_AUTH_MODE_MON 10 99 100/* msgr2 protocol modes */ 101#define CEPH_CON_MODE_UNKNOWN 0x0 102#define CEPH_CON_MODE_CRC 0x1 103#define CEPH_CON_MODE_SECURE 0x2 104 105#define CEPH_AUTH_UID_DEFAULT ((__u64) -1) 106 107const char *ceph_auth_proto_name(int proto); 108const char *ceph_con_mode_name(int mode); 109 110/********************************************* 111 * message layer 112 */ 113 114/* 115 * message types 116 */ 117 118/* misc */ 119#define CEPH_MSG_SHUTDOWN 1 120#define CEPH_MSG_PING 2 121 122/* client <-> monitor */ 123#define CEPH_MSG_MON_MAP 4 124#define CEPH_MSG_MON_GET_MAP 5 125#define CEPH_MSG_STATFS 13 126#define CEPH_MSG_STATFS_REPLY 14 127#define CEPH_MSG_MON_SUBSCRIBE 15 128#define CEPH_MSG_MON_SUBSCRIBE_ACK 16 129#define CEPH_MSG_AUTH 17 130#define CEPH_MSG_AUTH_REPLY 18 131#define CEPH_MSG_MON_GET_VERSION 19 132#define CEPH_MSG_MON_GET_VERSION_REPLY 20 133 134/* client <-> mds */ 135#define CEPH_MSG_MDS_MAP 21 136#define CEPH_MSG_FS_MAP_USER 103 137 138#define CEPH_MSG_CLIENT_SESSION 22 139#define CEPH_MSG_CLIENT_RECONNECT 23 140 141#define CEPH_MSG_CLIENT_REQUEST 24 142#define CEPH_MSG_CLIENT_REQUEST_FORWARD 25 143#define CEPH_MSG_CLIENT_REPLY 26 144#define CEPH_MSG_CLIENT_METRICS 29 145#define CEPH_MSG_CLIENT_CAPS 0x310 146#define CEPH_MSG_CLIENT_LEASE 0x311 147#define CEPH_MSG_CLIENT_SNAP 0x312 148#define CEPH_MSG_CLIENT_CAPRELEASE 0x313 149#define CEPH_MSG_CLIENT_QUOTA 0x314 150 151/* pool ops */ 152#define CEPH_MSG_POOLOP_REPLY 48 153#define CEPH_MSG_POOLOP 49 154 155/* mon commands */ 156#define CEPH_MSG_MON_COMMAND 50 157#define CEPH_MSG_MON_COMMAND_ACK 51 158 159/* osd */ 160#define CEPH_MSG_OSD_MAP 41 161#define CEPH_MSG_OSD_OP 42 162#define CEPH_MSG_OSD_OPREPLY 43 163#define CEPH_MSG_WATCH_NOTIFY 44 164#define CEPH_MSG_OSD_BACKOFF 61 165 166 167/* watch-notify operations */ 168enum { 169 CEPH_WATCH_EVENT_NOTIFY = 1, /* notifying watcher */ 170 CEPH_WATCH_EVENT_NOTIFY_COMPLETE = 2, /* notifier notified when done */ 171 CEPH_WATCH_EVENT_DISCONNECT = 3, /* we were disconnected */ 172}; 173 174 175struct ceph_mon_request_header { 176 __le64 have_version; 177 __le16 session_mon; 178 __le64 session_mon_tid; 179} __attribute__ ((packed)); 180 181struct ceph_mon_statfs { 182 struct ceph_mon_request_header monhdr; 183 struct ceph_fsid fsid; 184 __u8 contains_data_pool; 185 __le64 data_pool; 186} __attribute__ ((packed)); 187 188struct ceph_statfs { 189 __le64 kb, kb_used, kb_avail; 190 __le64 num_objects; 191} __attribute__ ((packed)); 192 193struct ceph_mon_statfs_reply { 194 struct ceph_fsid fsid; 195 __le64 version; 196 struct ceph_statfs st; 197} __attribute__ ((packed)); 198 199struct ceph_mon_command { 200 struct ceph_mon_request_header monhdr; 201 struct ceph_fsid fsid; 202 __le32 num_strs; /* always 1 */ 203 __le32 str_len; 204 char str[]; 205} __attribute__ ((packed)); 206 207struct ceph_osd_getmap { 208 struct ceph_mon_request_header monhdr; 209 struct ceph_fsid fsid; 210 __le32 start; 211} __attribute__ ((packed)); 212 213struct ceph_mds_getmap { 214 struct ceph_mon_request_header monhdr; 215 struct ceph_fsid fsid; 216} __attribute__ ((packed)); 217 218struct ceph_client_mount { 219 struct ceph_mon_request_header monhdr; 220} __attribute__ ((packed)); 221 222#define CEPH_SUBSCRIBE_ONETIME 1 /* i want only 1 update after have */ 223 224struct ceph_mon_subscribe_item { 225 __le64 start; 226 __u8 flags; 227} __attribute__ ((packed)); 228 229struct ceph_mon_subscribe_ack { 230 __le32 duration; /* seconds */ 231 struct ceph_fsid fsid; 232} __attribute__ ((packed)); 233 234#define CEPH_FS_CLUSTER_ID_NONE -1 235 236/* 237 * mdsmap flags 238 */ 239#define CEPH_MDSMAP_DOWN (1<<0) /* cluster deliberately down */ 240 241/* 242 * mds states 243 * > 0 -> in 244 * <= 0 -> out 245 */ 246#define CEPH_MDS_STATE_DNE 0 /* down, does not exist. */ 247#define CEPH_MDS_STATE_STOPPED -1 /* down, once existed, but no subtrees. 248 empty log. */ 249#define CEPH_MDS_STATE_BOOT -4 /* up, boot announcement. */ 250#define CEPH_MDS_STATE_STANDBY -5 /* up, idle. waiting for assignment. */ 251#define CEPH_MDS_STATE_CREATING -6 /* up, creating MDS instance. */ 252#define CEPH_MDS_STATE_STARTING -7 /* up, starting previously stopped mds */ 253#define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */ 254#define CEPH_MDS_STATE_REPLAYONCE -9 /* up, replaying an active node's journal */ 255 256#define CEPH_MDS_STATE_REPLAY 8 /* up, replaying journal. */ 257#define CEPH_MDS_STATE_RESOLVE 9 /* up, disambiguating distributed 258 operations (import, rename, etc.) */ 259#define CEPH_MDS_STATE_RECONNECT 10 /* up, reconnect to clients */ 260#define CEPH_MDS_STATE_REJOIN 11 /* up, rejoining distributed cache */ 261#define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */ 262#define CEPH_MDS_STATE_ACTIVE 13 /* up, active */ 263#define CEPH_MDS_STATE_STOPPING 14 /* up, but exporting metadata */ 264 265extern const char *ceph_mds_state_name(int s); 266 267 268/* 269 * metadata lock types. 270 * - these are bitmasks.. we can compose them 271 * - they also define the lock ordering by the MDS 272 * - a few of these are internal to the mds 273 */ 274#define CEPH_LOCK_DVERSION 1 275#define CEPH_LOCK_DN 2 276#define CEPH_LOCK_ISNAP 16 277#define CEPH_LOCK_IVERSION 32 /* mds internal */ 278#define CEPH_LOCK_IFILE 64 279#define CEPH_LOCK_IAUTH 128 280#define CEPH_LOCK_ILINK 256 281#define CEPH_LOCK_IDFT 512 /* dir frag tree */ 282#define CEPH_LOCK_INEST 1024 /* mds internal */ 283#define CEPH_LOCK_IXATTR 2048 284#define CEPH_LOCK_IFLOCK 4096 /* advisory file locks */ 285#define CEPH_LOCK_INO 8192 /* immutable inode bits; not a lock */ 286#define CEPH_LOCK_IPOLICY 16384 /* policy lock on dirs. MDS internal */ 287 288/* client_session ops */ 289enum { 290 CEPH_SESSION_REQUEST_OPEN, 291 CEPH_SESSION_OPEN, 292 CEPH_SESSION_REQUEST_CLOSE, 293 CEPH_SESSION_CLOSE, 294 CEPH_SESSION_REQUEST_RENEWCAPS, 295 CEPH_SESSION_RENEWCAPS, 296 CEPH_SESSION_STALE, 297 CEPH_SESSION_RECALL_STATE, 298 CEPH_SESSION_FLUSHMSG, 299 CEPH_SESSION_FLUSHMSG_ACK, 300 CEPH_SESSION_FORCE_RO, 301 CEPH_SESSION_REJECT, 302 CEPH_SESSION_REQUEST_FLUSH_MDLOG, 303}; 304 305#define CEPH_SESSION_BLOCKLISTED (1 << 0) /* session blocklisted */ 306 307extern const char *ceph_session_op_name(int op); 308 309struct ceph_mds_session_head { 310 __le32 op; 311 __le64 seq; 312 struct ceph_timespec stamp; 313 __le32 max_caps, max_leases; 314} __attribute__ ((packed)); 315 316/* client_request */ 317/* 318 * metadata ops. 319 * & 0x001000 -> write op 320 * & 0x010000 -> follow symlink (e.g. stat(), not lstat()). 321 & & 0x100000 -> use weird ino/path trace 322 */ 323#define CEPH_MDS_OP_WRITE 0x001000 324enum { 325 CEPH_MDS_OP_LOOKUP = 0x00100, 326 CEPH_MDS_OP_GETATTR = 0x00101, 327 CEPH_MDS_OP_LOOKUPHASH = 0x00102, 328 CEPH_MDS_OP_LOOKUPPARENT = 0x00103, 329 CEPH_MDS_OP_LOOKUPINO = 0x00104, 330 CEPH_MDS_OP_LOOKUPNAME = 0x00105, 331 CEPH_MDS_OP_GETVXATTR = 0x00106, 332 333 CEPH_MDS_OP_SETXATTR = 0x01105, 334 CEPH_MDS_OP_RMXATTR = 0x01106, 335 CEPH_MDS_OP_SETLAYOUT = 0x01107, 336 CEPH_MDS_OP_SETATTR = 0x01108, 337 CEPH_MDS_OP_SETFILELOCK= 0x01109, 338 CEPH_MDS_OP_GETFILELOCK= 0x00110, 339 CEPH_MDS_OP_SETDIRLAYOUT=0x0110a, 340 341 CEPH_MDS_OP_MKNOD = 0x01201, 342 CEPH_MDS_OP_LINK = 0x01202, 343 CEPH_MDS_OP_UNLINK = 0x01203, 344 CEPH_MDS_OP_RENAME = 0x01204, 345 CEPH_MDS_OP_MKDIR = 0x01220, 346 CEPH_MDS_OP_RMDIR = 0x01221, 347 CEPH_MDS_OP_SYMLINK = 0x01222, 348 349 CEPH_MDS_OP_CREATE = 0x01301, 350 CEPH_MDS_OP_OPEN = 0x00302, 351 CEPH_MDS_OP_READDIR = 0x00305, 352 353 CEPH_MDS_OP_LOOKUPSNAP = 0x00400, 354 CEPH_MDS_OP_MKSNAP = 0x01400, 355 CEPH_MDS_OP_RMSNAP = 0x01401, 356 CEPH_MDS_OP_LSSNAP = 0x00402, 357 CEPH_MDS_OP_RENAMESNAP = 0x01403, 358}; 359 360extern const char *ceph_mds_op_name(int op); 361 362 363#define CEPH_SETATTR_MODE 1 364#define CEPH_SETATTR_UID 2 365#define CEPH_SETATTR_GID 4 366#define CEPH_SETATTR_MTIME 8 367#define CEPH_SETATTR_ATIME 16 368#define CEPH_SETATTR_SIZE 32 369#define CEPH_SETATTR_CTIME 64 370 371/* 372 * Ceph setxattr request flags. 373 */ 374#define CEPH_XATTR_CREATE (1 << 0) 375#define CEPH_XATTR_REPLACE (1 << 1) 376#define CEPH_XATTR_REMOVE (1 << 31) 377 378/* 379 * readdir request flags; 380 */ 381#define CEPH_READDIR_REPLY_BITFLAGS (1<<0) 382 383/* 384 * readdir reply flags. 385 */ 386#define CEPH_READDIR_FRAG_END (1<<0) 387#define CEPH_READDIR_FRAG_COMPLETE (1<<8) 388#define CEPH_READDIR_HASH_ORDER (1<<9) 389#define CEPH_READDIR_OFFSET_HASH (1<<10) 390 391/* 392 * open request flags 393 */ 394#define CEPH_O_RDONLY 00000000 395#define CEPH_O_WRONLY 00000001 396#define CEPH_O_RDWR 00000002 397#define CEPH_O_CREAT 00000100 398#define CEPH_O_EXCL 00000200 399#define CEPH_O_TRUNC 00001000 400#define CEPH_O_DIRECTORY 00200000 401#define CEPH_O_NOFOLLOW 00400000 402 403union ceph_mds_request_args { 404 struct { 405 __le32 mask; /* CEPH_CAP_* */ 406 } __attribute__ ((packed)) getattr; 407 struct { 408 __le32 mode; 409 __le32 uid; 410 __le32 gid; 411 struct ceph_timespec mtime; 412 struct ceph_timespec atime; 413 __le64 size, old_size; /* old_size needed by truncate */ 414 __le32 mask; /* CEPH_SETATTR_* */ 415 } __attribute__ ((packed)) setattr; 416 struct { 417 __le32 frag; /* which dir fragment */ 418 __le32 max_entries; /* how many dentries to grab */ 419 __le32 max_bytes; 420 __le16 flags; 421 __le32 offset_hash; 422 } __attribute__ ((packed)) readdir; 423 struct { 424 __le32 mode; 425 __le32 rdev; 426 } __attribute__ ((packed)) mknod; 427 struct { 428 __le32 mode; 429 } __attribute__ ((packed)) mkdir; 430 struct { 431 __le32 flags; 432 __le32 mode; 433 __le32 stripe_unit; /* layout for newly created file */ 434 __le32 stripe_count; /* ... */ 435 __le32 object_size; 436 __le32 file_replication; 437 __le32 mask; /* CEPH_CAP_* */ 438 __le32 old_size; 439 } __attribute__ ((packed)) open; 440 struct { 441 __le32 flags; 442 __le32 osdmap_epoch; /* used for setting file/dir layouts */ 443 } __attribute__ ((packed)) setxattr; 444 struct { 445 struct ceph_file_layout_legacy layout; 446 } __attribute__ ((packed)) setlayout; 447 struct { 448 __u8 rule; /* currently fcntl or flock */ 449 __u8 type; /* shared, exclusive, remove*/ 450 __le64 owner; /* owner of the lock */ 451 __le64 pid; /* process id requesting the lock */ 452 __le64 start; /* initial location to lock */ 453 __le64 length; /* num bytes to lock from start */ 454 __u8 wait; /* will caller wait for lock to become available? */ 455 } __attribute__ ((packed)) filelock_change; 456 struct { 457 __le32 mask; /* CEPH_CAP_* */ 458 __le64 snapid; 459 __le64 parent; 460 __le32 hash; 461 } __attribute__ ((packed)) lookupino; 462} __attribute__ ((packed)); 463 464union ceph_mds_request_args_ext { 465 union ceph_mds_request_args old; 466 struct { 467 __le32 mode; 468 __le32 uid; 469 __le32 gid; 470 struct ceph_timespec mtime; 471 struct ceph_timespec atime; 472 __le64 size, old_size; /* old_size needed by truncate */ 473 __le32 mask; /* CEPH_SETATTR_* */ 474 struct ceph_timespec btime; 475 } __attribute__ ((packed)) setattr_ext; 476}; 477 478#define CEPH_MDS_FLAG_REPLAY 1 /* this is a replayed op */ 479#define CEPH_MDS_FLAG_WANT_DENTRY 2 /* want dentry in reply */ 480#define CEPH_MDS_FLAG_ASYNC 4 /* request is asynchronous */ 481 482struct ceph_mds_request_head_old { 483 __le64 oldest_client_tid; 484 __le32 mdsmap_epoch; /* on client */ 485 __le32 flags; /* CEPH_MDS_FLAG_* */ 486 __u8 num_retry, num_fwd; /* count retry, fwd attempts */ 487 __le16 num_releases; /* # include cap/lease release records */ 488 __le32 op; /* mds op code */ 489 __le32 caller_uid, caller_gid; 490 __le64 ino; /* use this ino for openc, mkdir, mknod, 491 etc. (if replaying) */ 492 union ceph_mds_request_args args; 493} __attribute__ ((packed)); 494 495#define CEPH_MDS_REQUEST_HEAD_VERSION 1 496 497struct ceph_mds_request_head { 498 __le16 version; /* struct version */ 499 __le64 oldest_client_tid; 500 __le32 mdsmap_epoch; /* on client */ 501 __le32 flags; /* CEPH_MDS_FLAG_* */ 502 __u8 num_retry, num_fwd; /* count retry, fwd attempts */ 503 __le16 num_releases; /* # include cap/lease release records */ 504 __le32 op; /* mds op code */ 505 __le32 caller_uid, caller_gid; 506 __le64 ino; /* use this ino for openc, mkdir, mknod, 507 etc. (if replaying) */ 508 union ceph_mds_request_args_ext args; 509} __attribute__ ((packed)); 510 511/* cap/lease release record */ 512struct ceph_mds_request_release { 513 __le64 ino, cap_id; /* ino and unique cap id */ 514 __le32 caps, wanted; /* new issued, wanted */ 515 __le32 seq, issue_seq, mseq; 516 __le32 dname_seq; /* if releasing a dentry lease, a */ 517 __le32 dname_len; /* string follows. */ 518} __attribute__ ((packed)); 519 520/* client reply */ 521struct ceph_mds_reply_head { 522 __le32 op; 523 __le32 result; 524 __le32 mdsmap_epoch; 525 __u8 safe; /* true if committed to disk */ 526 __u8 is_dentry, is_target; /* true if dentry, target inode records 527 are included with reply */ 528} __attribute__ ((packed)); 529 530/* one for each node split */ 531struct ceph_frag_tree_split { 532 __le32 frag; /* this frag splits... */ 533 __le32 by; /* ...by this many bits */ 534} __attribute__ ((packed)); 535 536struct ceph_frag_tree_head { 537 __le32 nsplits; /* num ceph_frag_tree_split records */ 538 struct ceph_frag_tree_split splits[]; 539} __attribute__ ((packed)); 540 541/* capability issue, for bundling with mds reply */ 542struct ceph_mds_reply_cap { 543 __le32 caps, wanted; /* caps issued, wanted */ 544 __le64 cap_id; 545 __le32 seq, mseq; 546 __le64 realm; /* snap realm */ 547 __u8 flags; /* CEPH_CAP_FLAG_* */ 548} __attribute__ ((packed)); 549 550#define CEPH_CAP_FLAG_AUTH (1 << 0) /* cap is issued by auth mds */ 551#define CEPH_CAP_FLAG_RELEASE (1 << 1) /* release the cap */ 552 553/* inode record, for bundling with mds reply */ 554struct ceph_mds_reply_inode { 555 __le64 ino; 556 __le64 snapid; 557 __le32 rdev; 558 __le64 version; /* inode version */ 559 __le64 xattr_version; /* version for xattr blob */ 560 struct ceph_mds_reply_cap cap; /* caps issued for this inode */ 561 struct ceph_file_layout_legacy layout; 562 struct ceph_timespec ctime, mtime, atime; 563 __le32 time_warp_seq; 564 __le64 size, max_size, truncate_size; 565 __le32 truncate_seq; 566 __le32 mode, uid, gid; 567 __le32 nlink; 568 __le64 files, subdirs, rbytes, rfiles, rsubdirs; /* dir stats */ 569 struct ceph_timespec rctime; 570 struct ceph_frag_tree_head fragtree; /* (must be at end of struct) */ 571} __attribute__ ((packed)); 572/* followed by frag array, symlink string, dir layout, xattr blob */ 573 574/* reply_lease follows dname, and reply_inode */ 575struct ceph_mds_reply_lease { 576 __le16 mask; /* lease type(s) */ 577 __le32 duration_ms; /* lease duration */ 578 __le32 seq; 579} __attribute__ ((packed)); 580 581#define CEPH_LEASE_VALID (1 | 2) /* old and new bit values */ 582#define CEPH_LEASE_PRIMARY_LINK 4 /* primary linkage */ 583 584struct ceph_mds_reply_dirfrag { 585 __le32 frag; /* fragment */ 586 __le32 auth; /* auth mds, if this is a delegation point */ 587 __le32 ndist; /* number of mds' this is replicated on */ 588 __le32 dist[]; 589} __attribute__ ((packed)); 590 591#define CEPH_LOCK_FCNTL 1 592#define CEPH_LOCK_FLOCK 2 593#define CEPH_LOCK_FCNTL_INTR 3 594#define CEPH_LOCK_FLOCK_INTR 4 595 596 597#define CEPH_LOCK_SHARED 1 598#define CEPH_LOCK_EXCL 2 599#define CEPH_LOCK_UNLOCK 4 600 601struct ceph_filelock { 602 __le64 start;/* file offset to start lock at */ 603 __le64 length; /* num bytes to lock; 0 for all following start */ 604 __le64 client; /* which client holds the lock */ 605 __le64 owner; /* owner the lock */ 606 __le64 pid; /* process id holding the lock on the client */ 607 __u8 type; /* shared lock, exclusive lock, or unlock */ 608} __attribute__ ((packed)); 609 610 611/* file access modes */ 612#define CEPH_FILE_MODE_PIN 0 613#define CEPH_FILE_MODE_RD 1 614#define CEPH_FILE_MODE_WR 2 615#define CEPH_FILE_MODE_RDWR 3 /* RD | WR */ 616#define CEPH_FILE_MODE_LAZY 4 /* lazy io */ 617#define CEPH_FILE_MODE_BITS 4 618#define CEPH_FILE_MODE_MASK ((1 << CEPH_FILE_MODE_BITS) - 1) 619 620int ceph_flags_to_mode(int flags); 621 622#define CEPH_INLINE_NONE ((__u64)-1) 623 624/* capability bits */ 625#define CEPH_CAP_PIN 1 /* no specific capabilities beyond the pin */ 626 627/* generic cap bits */ 628#define CEPH_CAP_GSHARED 1 /* client can reads */ 629#define CEPH_CAP_GEXCL 2 /* client can read and update */ 630#define CEPH_CAP_GCACHE 4 /* (file) client can cache reads */ 631#define CEPH_CAP_GRD 8 /* (file) client can read */ 632#define CEPH_CAP_GWR 16 /* (file) client can write */ 633#define CEPH_CAP_GBUFFER 32 /* (file) client can buffer writes */ 634#define CEPH_CAP_GWREXTEND 64 /* (file) client can extend EOF */ 635#define CEPH_CAP_GLAZYIO 128 /* (file) client can perform lazy io */ 636 637#define CEPH_CAP_SIMPLE_BITS 2 638#define CEPH_CAP_FILE_BITS 8 639 640/* per-lock shift */ 641#define CEPH_CAP_SAUTH 2 642#define CEPH_CAP_SLINK 4 643#define CEPH_CAP_SXATTR 6 644#define CEPH_CAP_SFILE 8 645#define CEPH_CAP_SFLOCK 20 646 647#define CEPH_CAP_BITS 22 648 649/* composed values */ 650#define CEPH_CAP_AUTH_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SAUTH) 651#define CEPH_CAP_AUTH_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SAUTH) 652#define CEPH_CAP_LINK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SLINK) 653#define CEPH_CAP_LINK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SLINK) 654#define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SXATTR) 655#define CEPH_CAP_XATTR_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SXATTR) 656#define CEPH_CAP_FILE(x) (x << CEPH_CAP_SFILE) 657#define CEPH_CAP_FILE_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFILE) 658#define CEPH_CAP_FILE_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFILE) 659#define CEPH_CAP_FILE_CACHE (CEPH_CAP_GCACHE << CEPH_CAP_SFILE) 660#define CEPH_CAP_FILE_RD (CEPH_CAP_GRD << CEPH_CAP_SFILE) 661#define CEPH_CAP_FILE_WR (CEPH_CAP_GWR << CEPH_CAP_SFILE) 662#define CEPH_CAP_FILE_BUFFER (CEPH_CAP_GBUFFER << CEPH_CAP_SFILE) 663#define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE) 664#define CEPH_CAP_FILE_LAZYIO (CEPH_CAP_GLAZYIO << CEPH_CAP_SFILE) 665#define CEPH_CAP_FLOCK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFLOCK) 666#define CEPH_CAP_FLOCK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFLOCK) 667 668 669/* cap masks (for getattr) */ 670#define CEPH_STAT_CAP_INODE CEPH_CAP_PIN 671#define CEPH_STAT_CAP_TYPE CEPH_CAP_PIN /* mode >> 12 */ 672#define CEPH_STAT_CAP_SYMLINK CEPH_CAP_PIN 673#define CEPH_STAT_CAP_UID CEPH_CAP_AUTH_SHARED 674#define CEPH_STAT_CAP_GID CEPH_CAP_AUTH_SHARED 675#define CEPH_STAT_CAP_MODE CEPH_CAP_AUTH_SHARED 676#define CEPH_STAT_CAP_NLINK CEPH_CAP_LINK_SHARED 677#define CEPH_STAT_CAP_LAYOUT CEPH_CAP_FILE_SHARED 678#define CEPH_STAT_CAP_MTIME CEPH_CAP_FILE_SHARED 679#define CEPH_STAT_CAP_SIZE CEPH_CAP_FILE_SHARED 680#define CEPH_STAT_CAP_ATIME CEPH_CAP_FILE_SHARED /* fixme */ 681#define CEPH_STAT_CAP_XATTR CEPH_CAP_XATTR_SHARED 682#define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN | \ 683 CEPH_CAP_AUTH_SHARED | \ 684 CEPH_CAP_LINK_SHARED | \ 685 CEPH_CAP_FILE_SHARED | \ 686 CEPH_CAP_XATTR_SHARED) 687#define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \ 688 CEPH_CAP_FILE_RD) 689#define CEPH_STAT_RSTAT CEPH_CAP_FILE_WREXTEND 690 691#define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED | \ 692 CEPH_CAP_LINK_SHARED | \ 693 CEPH_CAP_XATTR_SHARED | \ 694 CEPH_CAP_FILE_SHARED) 695#define CEPH_CAP_ANY_RD (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD | \ 696 CEPH_CAP_FILE_CACHE) 697 698#define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL | \ 699 CEPH_CAP_LINK_EXCL | \ 700 CEPH_CAP_XATTR_EXCL | \ 701 CEPH_CAP_FILE_EXCL) 702#define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \ 703 CEPH_CAP_FILE_SHARED) 704#define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \ 705 CEPH_CAP_FILE_EXCL) 706#define CEPH_CAP_ANY_WR (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR) 707#define CEPH_CAP_ANY (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \ 708 CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \ 709 CEPH_CAP_PIN) 710#define CEPH_CAP_ALL_FILE (CEPH_CAP_PIN | CEPH_CAP_ANY_SHARED | \ 711 CEPH_CAP_AUTH_EXCL | CEPH_CAP_XATTR_EXCL | \ 712 CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR) 713 714#define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \ 715 CEPH_LOCK_IXATTR) 716 717/* cap masks async dir operations */ 718#define CEPH_CAP_DIR_CREATE CEPH_CAP_FILE_CACHE 719#define CEPH_CAP_DIR_UNLINK CEPH_CAP_FILE_RD 720#define CEPH_CAP_ANY_DIR_OPS (CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_RD | \ 721 CEPH_CAP_FILE_WREXTEND | CEPH_CAP_FILE_LAZYIO) 722 723int ceph_caps_for_mode(int mode); 724 725enum { 726 CEPH_CAP_OP_GRANT, /* mds->client grant */ 727 CEPH_CAP_OP_REVOKE, /* mds->client revoke */ 728 CEPH_CAP_OP_TRUNC, /* mds->client trunc notify */ 729 CEPH_CAP_OP_EXPORT, /* mds has exported the cap */ 730 CEPH_CAP_OP_IMPORT, /* mds has imported the cap */ 731 CEPH_CAP_OP_UPDATE, /* client->mds update */ 732 CEPH_CAP_OP_DROP, /* client->mds drop cap bits */ 733 CEPH_CAP_OP_FLUSH, /* client->mds cap writeback */ 734 CEPH_CAP_OP_FLUSH_ACK, /* mds->client flushed */ 735 CEPH_CAP_OP_FLUSHSNAP, /* client->mds flush snapped metadata */ 736 CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */ 737 CEPH_CAP_OP_RELEASE, /* client->mds release (clean) cap */ 738 CEPH_CAP_OP_RENEW, /* client->mds renewal request */ 739}; 740 741extern const char *ceph_cap_op_name(int op); 742 743/* flags field in client cap messages (version >= 10) */ 744#define CEPH_CLIENT_CAPS_SYNC (1<<0) 745#define CEPH_CLIENT_CAPS_NO_CAPSNAP (1<<1) 746#define CEPH_CLIENT_CAPS_PENDING_CAPSNAP (1<<2) 747 748/* 749 * caps message, used for capability callbacks, acks, requests, etc. 750 */ 751struct ceph_mds_caps { 752 __le32 op; /* CEPH_CAP_OP_* */ 753 __le64 ino, realm; 754 __le64 cap_id; 755 __le32 seq, issue_seq; 756 __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */ 757 __le32 migrate_seq; 758 __le64 snap_follows; 759 __le32 snap_trace_len; 760 761 /* authlock */ 762 __le32 uid, gid, mode; 763 764 /* linklock */ 765 __le32 nlink; 766 767 /* xattrlock */ 768 __le32 xattr_len; 769 __le64 xattr_version; 770 771 /* filelock */ 772 __le64 size, max_size, truncate_size; 773 __le32 truncate_seq; 774 struct ceph_timespec mtime, atime, ctime; 775 struct ceph_file_layout_legacy layout; 776 __le32 time_warp_seq; 777} __attribute__ ((packed)); 778 779struct ceph_mds_cap_peer { 780 __le64 cap_id; 781 __le32 seq; 782 __le32 mseq; 783 __le32 mds; 784 __u8 flags; 785} __attribute__ ((packed)); 786 787/* cap release msg head */ 788struct ceph_mds_cap_release { 789 __le32 num; /* number of cap_items that follow */ 790} __attribute__ ((packed)); 791 792struct ceph_mds_cap_item { 793 __le64 ino; 794 __le64 cap_id; 795 __le32 migrate_seq, seq; 796} __attribute__ ((packed)); 797 798#define CEPH_MDS_LEASE_REVOKE 1 /* mds -> client */ 799#define CEPH_MDS_LEASE_RELEASE 2 /* client -> mds */ 800#define CEPH_MDS_LEASE_RENEW 3 /* client <-> mds */ 801#define CEPH_MDS_LEASE_REVOKE_ACK 4 /* client -> mds */ 802 803extern const char *ceph_lease_op_name(int o); 804 805/* lease msg header */ 806struct ceph_mds_lease { 807 __u8 action; /* CEPH_MDS_LEASE_* */ 808 __le16 mask; /* which lease */ 809 __le64 ino; 810 __le64 first, last; /* snap range */ 811 __le32 seq; 812 __le32 duration_ms; /* duration of renewal */ 813} __attribute__ ((packed)); 814/* followed by a __le32+string for dname */ 815 816/* client reconnect */ 817struct ceph_mds_cap_reconnect { 818 __le64 cap_id; 819 __le32 wanted; 820 __le32 issued; 821 __le64 snaprealm; 822 __le64 pathbase; /* base ino for our path to this ino */ 823 __le32 flock_len; /* size of flock state blob, if any */ 824} __attribute__ ((packed)); 825/* followed by flock blob */ 826 827struct ceph_mds_cap_reconnect_v1 { 828 __le64 cap_id; 829 __le32 wanted; 830 __le32 issued; 831 __le64 size; 832 struct ceph_timespec mtime, atime; 833 __le64 snaprealm; 834 __le64 pathbase; /* base ino for our path to this ino */ 835} __attribute__ ((packed)); 836 837struct ceph_mds_snaprealm_reconnect { 838 __le64 ino; /* snap realm base */ 839 __le64 seq; /* snap seq for this snap realm */ 840 __le64 parent; /* parent realm */ 841} __attribute__ ((packed)); 842 843/* 844 * snaps 845 */ 846enum { 847 CEPH_SNAP_OP_UPDATE, /* CREATE or DESTROY */ 848 CEPH_SNAP_OP_CREATE, 849 CEPH_SNAP_OP_DESTROY, 850 CEPH_SNAP_OP_SPLIT, 851}; 852 853extern const char *ceph_snap_op_name(int o); 854 855/* snap msg header */ 856struct ceph_mds_snap_head { 857 __le32 op; /* CEPH_SNAP_OP_* */ 858 __le64 split; /* ino to split off, if any */ 859 __le32 num_split_inos; /* # inos belonging to new child realm */ 860 __le32 num_split_realms; /* # child realms udner new child realm */ 861 __le32 trace_len; /* size of snap trace blob */ 862} __attribute__ ((packed)); 863/* followed by split ino list, then split realms, then the trace blob */ 864 865/* 866 * encode info about a snaprealm, as viewed by a client 867 */ 868struct ceph_mds_snap_realm { 869 __le64 ino; /* ino */ 870 __le64 created; /* snap: when created */ 871 __le64 parent; /* ino: parent realm */ 872 __le64 parent_since; /* snap: same parent since */ 873 __le64 seq; /* snap: version */ 874 __le32 num_snaps; 875 __le32 num_prior_parent_snaps; 876} __attribute__ ((packed)); 877/* followed by my snap list, then prior parent snap list */ 878 879/* 880 * quotas 881 */ 882struct ceph_mds_quota { 883 __le64 ino; /* ino */ 884 struct ceph_timespec rctime; 885 __le64 rbytes; /* dir stats */ 886 __le64 rfiles; 887 __le64 rsubdirs; 888 __u8 struct_v; /* compat */ 889 __u8 struct_compat; 890 __le32 struct_len; 891 __le64 max_bytes; /* quota max. bytes */ 892 __le64 max_files; /* quota max. files */ 893} __attribute__ ((packed)); 894 895#endif