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