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 * ceph_file_layout - describe data layout for a file/inode
38 */
39 struct ceph_file_layout {
40 /* file -> object mapping */
41 __le32 fl_stripe_unit; /* stripe unit, in bytes. must be multiple
42 of page size. */
43 __le32 fl_stripe_count; /* over this many objects */
44 __le32 fl_object_size; /* until objects are this big, then move to
45 new objects */
46 __le32 fl_cas_hash; /* UNUSED. 0 = none; 1 = sha256 */
47
48 /* pg -> disk layout */
49 __le32 fl_object_stripe_unit; /* UNUSED. for per-object parity, if any */
50
51 /* object -> pg layout */
52 __le32 fl_unused; /* unused; used to be preferred primary for pg (-1 for none) */
53 __le32 fl_pg_pool; /* namespace, crush ruleset, rep level */
54 } __attribute__ ((packed));
55
56 #define ceph_file_layout_su(l) ((__s32)le32_to_cpu((l).fl_stripe_unit))
57 #define ceph_file_layout_stripe_count(l) \
58 ((__s32)le32_to_cpu((l).fl_stripe_count))
59 #define ceph_file_layout_object_size(l) ((__s32)le32_to_cpu((l).fl_object_size))
60 #define ceph_file_layout_cas_hash(l) ((__s32)le32_to_cpu((l).fl_cas_hash))
61 #define ceph_file_layout_object_su(l) \
62 ((__s32)le32_to_cpu((l).fl_object_stripe_unit))
63 #define ceph_file_layout_pg_pool(l) \
64 ((__s32)le32_to_cpu((l).fl_pg_pool))
65
ceph_file_layout_stripe_width(struct ceph_file_layout * l)66 static inline unsigned ceph_file_layout_stripe_width(struct ceph_file_layout *l)
67 {
68 return le32_to_cpu(l->fl_stripe_unit) *
69 le32_to_cpu(l->fl_stripe_count);
70 }
71
72 /* "period" == bytes before i start on a new set of objects */
ceph_file_layout_period(struct ceph_file_layout * l)73 static inline unsigned ceph_file_layout_period(struct ceph_file_layout *l)
74 {
75 return le32_to_cpu(l->fl_object_size) *
76 le32_to_cpu(l->fl_stripe_count);
77 }
78
79 #define CEPH_MIN_STRIPE_UNIT 65536
80
81 int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
82
83 struct ceph_dir_layout {
84 __u8 dl_dir_hash; /* see ceph_hash.h for ids */
85 __u8 dl_unused1;
86 __u16 dl_unused2;
87 __u32 dl_unused3;
88 } __attribute__ ((packed));
89
90 /* crypto algorithms */
91 #define CEPH_CRYPTO_NONE 0x0
92 #define CEPH_CRYPTO_AES 0x1
93
94 #define CEPH_AES_IV "cephsageyudagreg"
95
96 /* security/authentication protocols */
97 #define CEPH_AUTH_UNKNOWN 0x0
98 #define CEPH_AUTH_NONE 0x1
99 #define CEPH_AUTH_CEPHX 0x2
100
101 #define CEPH_AUTH_UID_DEFAULT ((__u64) -1)
102
103
104 /*********************************************
105 * message layer
106 */
107
108 /*
109 * message types
110 */
111
112 /* misc */
113 #define CEPH_MSG_SHUTDOWN 1
114 #define CEPH_MSG_PING 2
115
116 /* client <-> monitor */
117 #define CEPH_MSG_MON_MAP 4
118 #define CEPH_MSG_MON_GET_MAP 5
119 #define CEPH_MSG_STATFS 13
120 #define CEPH_MSG_STATFS_REPLY 14
121 #define CEPH_MSG_MON_SUBSCRIBE 15
122 #define CEPH_MSG_MON_SUBSCRIBE_ACK 16
123 #define CEPH_MSG_AUTH 17
124 #define CEPH_MSG_AUTH_REPLY 18
125 #define CEPH_MSG_MON_GET_VERSION 19
126 #define CEPH_MSG_MON_GET_VERSION_REPLY 20
127
128 /* client <-> mds */
129 #define CEPH_MSG_MDS_MAP 21
130
131 #define CEPH_MSG_CLIENT_SESSION 22
132 #define CEPH_MSG_CLIENT_RECONNECT 23
133
134 #define CEPH_MSG_CLIENT_REQUEST 24
135 #define CEPH_MSG_CLIENT_REQUEST_FORWARD 25
136 #define CEPH_MSG_CLIENT_REPLY 26
137 #define CEPH_MSG_CLIENT_CAPS 0x310
138 #define CEPH_MSG_CLIENT_LEASE 0x311
139 #define CEPH_MSG_CLIENT_SNAP 0x312
140 #define CEPH_MSG_CLIENT_CAPRELEASE 0x313
141
142 /* pool ops */
143 #define CEPH_MSG_POOLOP_REPLY 48
144 #define CEPH_MSG_POOLOP 49
145
146
147 /* osd */
148 #define CEPH_MSG_OSD_MAP 41
149 #define CEPH_MSG_OSD_OP 42
150 #define CEPH_MSG_OSD_OPREPLY 43
151 #define CEPH_MSG_WATCH_NOTIFY 44
152
153
154 /* watch-notify operations */
155 enum {
156 WATCH_NOTIFY = 1, /* notifying watcher */
157 WATCH_NOTIFY_COMPLETE = 2, /* notifier notified when done */
158 };
159
160
161 /* pool operations */
162 enum {
163 POOL_OP_CREATE = 0x01,
164 POOL_OP_DELETE = 0x02,
165 POOL_OP_AUID_CHANGE = 0x03,
166 POOL_OP_CREATE_SNAP = 0x11,
167 POOL_OP_DELETE_SNAP = 0x12,
168 POOL_OP_CREATE_UNMANAGED_SNAP = 0x21,
169 POOL_OP_DELETE_UNMANAGED_SNAP = 0x22,
170 };
171
172 struct ceph_mon_request_header {
173 __le64 have_version;
174 __le16 session_mon;
175 __le64 session_mon_tid;
176 } __attribute__ ((packed));
177
178 struct ceph_mon_statfs {
179 struct ceph_mon_request_header monhdr;
180 struct ceph_fsid fsid;
181 } __attribute__ ((packed));
182
183 struct ceph_statfs {
184 __le64 kb, kb_used, kb_avail;
185 __le64 num_objects;
186 } __attribute__ ((packed));
187
188 struct ceph_mon_statfs_reply {
189 struct ceph_fsid fsid;
190 __le64 version;
191 struct ceph_statfs st;
192 } __attribute__ ((packed));
193
194 const char *ceph_pool_op_name(int op);
195
196 struct ceph_mon_poolop {
197 struct ceph_mon_request_header monhdr;
198 struct ceph_fsid fsid;
199 __le32 pool;
200 __le32 op;
201 __le64 auid;
202 __le64 snapid;
203 __le32 name_len;
204 } __attribute__ ((packed));
205
206 struct ceph_mon_poolop_reply {
207 struct ceph_mon_request_header monhdr;
208 struct ceph_fsid fsid;
209 __le32 reply_code;
210 __le32 epoch;
211 char has_data;
212 char data[0];
213 } __attribute__ ((packed));
214
215 struct ceph_mon_unmanaged_snap {
216 __le64 snapid;
217 } __attribute__ ((packed));
218
219 struct ceph_osd_getmap {
220 struct ceph_mon_request_header monhdr;
221 struct ceph_fsid fsid;
222 __le32 start;
223 } __attribute__ ((packed));
224
225 struct ceph_mds_getmap {
226 struct ceph_mon_request_header monhdr;
227 struct ceph_fsid fsid;
228 } __attribute__ ((packed));
229
230 struct ceph_client_mount {
231 struct ceph_mon_request_header monhdr;
232 } __attribute__ ((packed));
233
234 #define CEPH_SUBSCRIBE_ONETIME 1 /* i want only 1 update after have */
235
236 struct ceph_mon_subscribe_item {
237 __le64 have_version; __le64 have;
238 __u8 onetime;
239 } __attribute__ ((packed));
240
241 struct ceph_mon_subscribe_ack {
242 __le32 duration; /* seconds */
243 struct ceph_fsid fsid;
244 } __attribute__ ((packed));
245
246 /*
247 * mdsmap flags
248 */
249 #define CEPH_MDSMAP_DOWN (1<<0) /* cluster deliberately down */
250
251 /*
252 * mds states
253 * > 0 -> in
254 * <= 0 -> out
255 */
256 #define CEPH_MDS_STATE_DNE 0 /* down, does not exist. */
257 #define CEPH_MDS_STATE_STOPPED -1 /* down, once existed, but no subtrees.
258 empty log. */
259 #define CEPH_MDS_STATE_BOOT -4 /* up, boot announcement. */
260 #define CEPH_MDS_STATE_STANDBY -5 /* up, idle. waiting for assignment. */
261 #define CEPH_MDS_STATE_CREATING -6 /* up, creating MDS instance. */
262 #define CEPH_MDS_STATE_STARTING -7 /* up, starting previously stopped mds */
263 #define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
264 #define CEPH_MDS_STATE_REPLAYONCE -9 /* up, replaying an active node's journal */
265
266 #define CEPH_MDS_STATE_REPLAY 8 /* up, replaying journal. */
267 #define CEPH_MDS_STATE_RESOLVE 9 /* up, disambiguating distributed
268 operations (import, rename, etc.) */
269 #define CEPH_MDS_STATE_RECONNECT 10 /* up, reconnect to clients */
270 #define CEPH_MDS_STATE_REJOIN 11 /* up, rejoining distributed cache */
271 #define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */
272 #define CEPH_MDS_STATE_ACTIVE 13 /* up, active */
273 #define CEPH_MDS_STATE_STOPPING 14 /* up, but exporting metadata */
274
275 extern const char *ceph_mds_state_name(int s);
276
277
278 /*
279 * metadata lock types.
280 * - these are bitmasks.. we can compose them
281 * - they also define the lock ordering by the MDS
282 * - a few of these are internal to the mds
283 */
284 #define CEPH_LOCK_DVERSION 1
285 #define CEPH_LOCK_DN 2
286 #define CEPH_LOCK_ISNAP 16
287 #define CEPH_LOCK_IVERSION 32 /* mds internal */
288 #define CEPH_LOCK_IFILE 64
289 #define CEPH_LOCK_IAUTH 128
290 #define CEPH_LOCK_ILINK 256
291 #define CEPH_LOCK_IDFT 512 /* dir frag tree */
292 #define CEPH_LOCK_INEST 1024 /* mds internal */
293 #define CEPH_LOCK_IXATTR 2048
294 #define CEPH_LOCK_IFLOCK 4096 /* advisory file locks */
295 #define CEPH_LOCK_INO 8192 /* immutable inode bits; not a lock */
296 #define CEPH_LOCK_IPOLICY 16384 /* policy lock on dirs. MDS internal */
297
298 /* client_session ops */
299 enum {
300 CEPH_SESSION_REQUEST_OPEN,
301 CEPH_SESSION_OPEN,
302 CEPH_SESSION_REQUEST_CLOSE,
303 CEPH_SESSION_CLOSE,
304 CEPH_SESSION_REQUEST_RENEWCAPS,
305 CEPH_SESSION_RENEWCAPS,
306 CEPH_SESSION_STALE,
307 CEPH_SESSION_RECALL_STATE,
308 CEPH_SESSION_FLUSHMSG,
309 CEPH_SESSION_FLUSHMSG_ACK,
310 };
311
312 extern const char *ceph_session_op_name(int op);
313
314 struct ceph_mds_session_head {
315 __le32 op;
316 __le64 seq;
317 struct ceph_timespec stamp;
318 __le32 max_caps, max_leases;
319 } __attribute__ ((packed));
320
321 /* client_request */
322 /*
323 * metadata ops.
324 * & 0x001000 -> write op
325 * & 0x010000 -> follow symlink (e.g. stat(), not lstat()).
326 & & 0x100000 -> use weird ino/path trace
327 */
328 #define CEPH_MDS_OP_WRITE 0x001000
329 enum {
330 CEPH_MDS_OP_LOOKUP = 0x00100,
331 CEPH_MDS_OP_GETATTR = 0x00101,
332 CEPH_MDS_OP_LOOKUPHASH = 0x00102,
333 CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
334 CEPH_MDS_OP_LOOKUPINO = 0x00104,
335 CEPH_MDS_OP_LOOKUPNAME = 0x00105,
336
337 CEPH_MDS_OP_SETXATTR = 0x01105,
338 CEPH_MDS_OP_RMXATTR = 0x01106,
339 CEPH_MDS_OP_SETLAYOUT = 0x01107,
340 CEPH_MDS_OP_SETATTR = 0x01108,
341 CEPH_MDS_OP_SETFILELOCK= 0x01109,
342 CEPH_MDS_OP_GETFILELOCK= 0x00110,
343 CEPH_MDS_OP_SETDIRLAYOUT=0x0110a,
344
345 CEPH_MDS_OP_MKNOD = 0x01201,
346 CEPH_MDS_OP_LINK = 0x01202,
347 CEPH_MDS_OP_UNLINK = 0x01203,
348 CEPH_MDS_OP_RENAME = 0x01204,
349 CEPH_MDS_OP_MKDIR = 0x01220,
350 CEPH_MDS_OP_RMDIR = 0x01221,
351 CEPH_MDS_OP_SYMLINK = 0x01222,
352
353 CEPH_MDS_OP_CREATE = 0x01301,
354 CEPH_MDS_OP_OPEN = 0x00302,
355 CEPH_MDS_OP_READDIR = 0x00305,
356
357 CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
358 CEPH_MDS_OP_MKSNAP = 0x01400,
359 CEPH_MDS_OP_RMSNAP = 0x01401,
360 CEPH_MDS_OP_LSSNAP = 0x00402,
361 };
362
363 extern const char *ceph_mds_op_name(int op);
364
365
366 #define CEPH_SETATTR_MODE 1
367 #define CEPH_SETATTR_UID 2
368 #define CEPH_SETATTR_GID 4
369 #define CEPH_SETATTR_MTIME 8
370 #define CEPH_SETATTR_ATIME 16
371 #define CEPH_SETATTR_SIZE 32
372 #define CEPH_SETATTR_CTIME 64
373
374 /*
375 * Ceph setxattr request flags.
376 */
377 #define CEPH_XATTR_CREATE (1 << 0)
378 #define CEPH_XATTR_REPLACE (1 << 1)
379 #define CEPH_XATTR_REMOVE (1 << 31)
380
381 union ceph_mds_request_args {
382 struct {
383 __le32 mask; /* CEPH_CAP_* */
384 } __attribute__ ((packed)) getattr;
385 struct {
386 __le32 mode;
387 __le32 uid;
388 __le32 gid;
389 struct ceph_timespec mtime;
390 struct ceph_timespec atime;
391 __le64 size, old_size; /* old_size needed by truncate */
392 __le32 mask; /* CEPH_SETATTR_* */
393 } __attribute__ ((packed)) setattr;
394 struct {
395 __le32 frag; /* which dir fragment */
396 __le32 max_entries; /* how many dentries to grab */
397 __le32 max_bytes;
398 } __attribute__ ((packed)) readdir;
399 struct {
400 __le32 mode;
401 __le32 rdev;
402 } __attribute__ ((packed)) mknod;
403 struct {
404 __le32 mode;
405 } __attribute__ ((packed)) mkdir;
406 struct {
407 __le32 flags;
408 __le32 mode;
409 __le32 stripe_unit; /* layout for newly created file */
410 __le32 stripe_count; /* ... */
411 __le32 object_size;
412 __le32 file_replication;
413 __le32 unused; /* used to be preferred osd */
414 } __attribute__ ((packed)) open;
415 struct {
416 __le32 flags;
417 } __attribute__ ((packed)) setxattr;
418 struct {
419 struct ceph_file_layout layout;
420 } __attribute__ ((packed)) setlayout;
421 struct {
422 __u8 rule; /* currently fcntl or flock */
423 __u8 type; /* shared, exclusive, remove*/
424 __le64 owner; /* owner of the lock */
425 __le64 pid; /* process id requesting the lock */
426 __le64 start; /* initial location to lock */
427 __le64 length; /* num bytes to lock from start */
428 __u8 wait; /* will caller wait for lock to become available? */
429 } __attribute__ ((packed)) filelock_change;
430 } __attribute__ ((packed));
431
432 #define CEPH_MDS_FLAG_REPLAY 1 /* this is a replayed op */
433 #define CEPH_MDS_FLAG_WANT_DENTRY 2 /* want dentry in reply */
434
435 struct ceph_mds_request_head {
436 __le64 oldest_client_tid;
437 __le32 mdsmap_epoch; /* on client */
438 __le32 flags; /* CEPH_MDS_FLAG_* */
439 __u8 num_retry, num_fwd; /* count retry, fwd attempts */
440 __le16 num_releases; /* # include cap/lease release records */
441 __le32 op; /* mds op code */
442 __le32 caller_uid, caller_gid;
443 __le64 ino; /* use this ino for openc, mkdir, mknod,
444 etc. (if replaying) */
445 union ceph_mds_request_args args;
446 } __attribute__ ((packed));
447
448 /* cap/lease release record */
449 struct ceph_mds_request_release {
450 __le64 ino, cap_id; /* ino and unique cap id */
451 __le32 caps, wanted; /* new issued, wanted */
452 __le32 seq, issue_seq, mseq;
453 __le32 dname_seq; /* if releasing a dentry lease, a */
454 __le32 dname_len; /* string follows. */
455 } __attribute__ ((packed));
456
457 /* client reply */
458 struct ceph_mds_reply_head {
459 __le32 op;
460 __le32 result;
461 __le32 mdsmap_epoch;
462 __u8 safe; /* true if committed to disk */
463 __u8 is_dentry, is_target; /* true if dentry, target inode records
464 are included with reply */
465 } __attribute__ ((packed));
466
467 /* one for each node split */
468 struct ceph_frag_tree_split {
469 __le32 frag; /* this frag splits... */
470 __le32 by; /* ...by this many bits */
471 } __attribute__ ((packed));
472
473 struct ceph_frag_tree_head {
474 __le32 nsplits; /* num ceph_frag_tree_split records */
475 struct ceph_frag_tree_split splits[];
476 } __attribute__ ((packed));
477
478 /* capability issue, for bundling with mds reply */
479 struct ceph_mds_reply_cap {
480 __le32 caps, wanted; /* caps issued, wanted */
481 __le64 cap_id;
482 __le32 seq, mseq;
483 __le64 realm; /* snap realm */
484 __u8 flags; /* CEPH_CAP_FLAG_* */
485 } __attribute__ ((packed));
486
487 #define CEPH_CAP_FLAG_AUTH (1 << 0) /* cap is issued by auth mds */
488 #define CEPH_CAP_FLAG_RELEASE (1 << 1) /* release the cap */
489
490 /* inode record, for bundling with mds reply */
491 struct ceph_mds_reply_inode {
492 __le64 ino;
493 __le64 snapid;
494 __le32 rdev;
495 __le64 version; /* inode version */
496 __le64 xattr_version; /* version for xattr blob */
497 struct ceph_mds_reply_cap cap; /* caps issued for this inode */
498 struct ceph_file_layout layout;
499 struct ceph_timespec ctime, mtime, atime;
500 __le32 time_warp_seq;
501 __le64 size, max_size, truncate_size;
502 __le32 truncate_seq;
503 __le32 mode, uid, gid;
504 __le32 nlink;
505 __le64 files, subdirs, rbytes, rfiles, rsubdirs; /* dir stats */
506 struct ceph_timespec rctime;
507 struct ceph_frag_tree_head fragtree; /* (must be at end of struct) */
508 } __attribute__ ((packed));
509 /* followed by frag array, symlink string, dir layout, xattr blob */
510
511 /* reply_lease follows dname, and reply_inode */
512 struct ceph_mds_reply_lease {
513 __le16 mask; /* lease type(s) */
514 __le32 duration_ms; /* lease duration */
515 __le32 seq;
516 } __attribute__ ((packed));
517
518 struct ceph_mds_reply_dirfrag {
519 __le32 frag; /* fragment */
520 __le32 auth; /* auth mds, if this is a delegation point */
521 __le32 ndist; /* number of mds' this is replicated on */
522 __le32 dist[];
523 } __attribute__ ((packed));
524
525 #define CEPH_LOCK_FCNTL 1
526 #define CEPH_LOCK_FLOCK 2
527
528 #define CEPH_LOCK_SHARED 1
529 #define CEPH_LOCK_EXCL 2
530 #define CEPH_LOCK_UNLOCK 4
531
532 struct ceph_filelock {
533 __le64 start;/* file offset to start lock at */
534 __le64 length; /* num bytes to lock; 0 for all following start */
535 __le64 client; /* which client holds the lock */
536 __le64 owner; /* owner the lock */
537 __le64 pid; /* process id holding the lock on the client */
538 __u8 type; /* shared lock, exclusive lock, or unlock */
539 } __attribute__ ((packed));
540
541
542 /* file access modes */
543 #define CEPH_FILE_MODE_PIN 0
544 #define CEPH_FILE_MODE_RD 1
545 #define CEPH_FILE_MODE_WR 2
546 #define CEPH_FILE_MODE_RDWR 3 /* RD | WR */
547 #define CEPH_FILE_MODE_LAZY 4 /* lazy io */
548 #define CEPH_FILE_MODE_NUM 8 /* bc these are bit fields.. mostly */
549
550 int ceph_flags_to_mode(int flags);
551
552
553 /* capability bits */
554 #define CEPH_CAP_PIN 1 /* no specific capabilities beyond the pin */
555
556 /* generic cap bits */
557 #define CEPH_CAP_GSHARED 1 /* client can reads */
558 #define CEPH_CAP_GEXCL 2 /* client can read and update */
559 #define CEPH_CAP_GCACHE 4 /* (file) client can cache reads */
560 #define CEPH_CAP_GRD 8 /* (file) client can read */
561 #define CEPH_CAP_GWR 16 /* (file) client can write */
562 #define CEPH_CAP_GBUFFER 32 /* (file) client can buffer writes */
563 #define CEPH_CAP_GWREXTEND 64 /* (file) client can extend EOF */
564 #define CEPH_CAP_GLAZYIO 128 /* (file) client can perform lazy io */
565
566 #define CEPH_CAP_SIMPLE_BITS 2
567 #define CEPH_CAP_FILE_BITS 8
568
569 /* per-lock shift */
570 #define CEPH_CAP_SAUTH 2
571 #define CEPH_CAP_SLINK 4
572 #define CEPH_CAP_SXATTR 6
573 #define CEPH_CAP_SFILE 8
574 #define CEPH_CAP_SFLOCK 20
575
576 #define CEPH_CAP_BITS 22
577
578 /* composed values */
579 #define CEPH_CAP_AUTH_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SAUTH)
580 #define CEPH_CAP_AUTH_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SAUTH)
581 #define CEPH_CAP_LINK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SLINK)
582 #define CEPH_CAP_LINK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SLINK)
583 #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SXATTR)
584 #define CEPH_CAP_XATTR_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SXATTR)
585 #define CEPH_CAP_FILE(x) (x << CEPH_CAP_SFILE)
586 #define CEPH_CAP_FILE_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFILE)
587 #define CEPH_CAP_FILE_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFILE)
588 #define CEPH_CAP_FILE_CACHE (CEPH_CAP_GCACHE << CEPH_CAP_SFILE)
589 #define CEPH_CAP_FILE_RD (CEPH_CAP_GRD << CEPH_CAP_SFILE)
590 #define CEPH_CAP_FILE_WR (CEPH_CAP_GWR << CEPH_CAP_SFILE)
591 #define CEPH_CAP_FILE_BUFFER (CEPH_CAP_GBUFFER << CEPH_CAP_SFILE)
592 #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
593 #define CEPH_CAP_FILE_LAZYIO (CEPH_CAP_GLAZYIO << CEPH_CAP_SFILE)
594 #define CEPH_CAP_FLOCK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFLOCK)
595 #define CEPH_CAP_FLOCK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFLOCK)
596
597
598 /* cap masks (for getattr) */
599 #define CEPH_STAT_CAP_INODE CEPH_CAP_PIN
600 #define CEPH_STAT_CAP_TYPE CEPH_CAP_PIN /* mode >> 12 */
601 #define CEPH_STAT_CAP_SYMLINK CEPH_CAP_PIN
602 #define CEPH_STAT_CAP_UID CEPH_CAP_AUTH_SHARED
603 #define CEPH_STAT_CAP_GID CEPH_CAP_AUTH_SHARED
604 #define CEPH_STAT_CAP_MODE CEPH_CAP_AUTH_SHARED
605 #define CEPH_STAT_CAP_NLINK CEPH_CAP_LINK_SHARED
606 #define CEPH_STAT_CAP_LAYOUT CEPH_CAP_FILE_SHARED
607 #define CEPH_STAT_CAP_MTIME CEPH_CAP_FILE_SHARED
608 #define CEPH_STAT_CAP_SIZE CEPH_CAP_FILE_SHARED
609 #define CEPH_STAT_CAP_ATIME CEPH_CAP_FILE_SHARED /* fixme */
610 #define CEPH_STAT_CAP_XATTR CEPH_CAP_XATTR_SHARED
611 #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN | \
612 CEPH_CAP_AUTH_SHARED | \
613 CEPH_CAP_LINK_SHARED | \
614 CEPH_CAP_FILE_SHARED | \
615 CEPH_CAP_XATTR_SHARED)
616
617 #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED | \
618 CEPH_CAP_LINK_SHARED | \
619 CEPH_CAP_XATTR_SHARED | \
620 CEPH_CAP_FILE_SHARED)
621 #define CEPH_CAP_ANY_RD (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD | \
622 CEPH_CAP_FILE_CACHE)
623
624 #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL | \
625 CEPH_CAP_LINK_EXCL | \
626 CEPH_CAP_XATTR_EXCL | \
627 CEPH_CAP_FILE_EXCL)
628 #define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \
629 CEPH_CAP_FILE_SHARED)
630 #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \
631 CEPH_CAP_FILE_EXCL)
632 #define CEPH_CAP_ANY_WR (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
633 #define CEPH_CAP_ANY (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
634 CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
635 CEPH_CAP_PIN)
636
637 #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
638 CEPH_LOCK_IXATTR)
639
640 int ceph_caps_for_mode(int mode);
641
642 enum {
643 CEPH_CAP_OP_GRANT, /* mds->client grant */
644 CEPH_CAP_OP_REVOKE, /* mds->client revoke */
645 CEPH_CAP_OP_TRUNC, /* mds->client trunc notify */
646 CEPH_CAP_OP_EXPORT, /* mds has exported the cap */
647 CEPH_CAP_OP_IMPORT, /* mds has imported the cap */
648 CEPH_CAP_OP_UPDATE, /* client->mds update */
649 CEPH_CAP_OP_DROP, /* client->mds drop cap bits */
650 CEPH_CAP_OP_FLUSH, /* client->mds cap writeback */
651 CEPH_CAP_OP_FLUSH_ACK, /* mds->client flushed */
652 CEPH_CAP_OP_FLUSHSNAP, /* client->mds flush snapped metadata */
653 CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
654 CEPH_CAP_OP_RELEASE, /* client->mds release (clean) cap */
655 CEPH_CAP_OP_RENEW, /* client->mds renewal request */
656 };
657
658 extern const char *ceph_cap_op_name(int op);
659
660 /*
661 * caps message, used for capability callbacks, acks, requests, etc.
662 */
663 struct ceph_mds_caps {
664 __le32 op; /* CEPH_CAP_OP_* */
665 __le64 ino, realm;
666 __le64 cap_id;
667 __le32 seq, issue_seq;
668 __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
669 __le32 migrate_seq;
670 __le64 snap_follows;
671 __le32 snap_trace_len;
672
673 /* authlock */
674 __le32 uid, gid, mode;
675
676 /* linklock */
677 __le32 nlink;
678
679 /* xattrlock */
680 __le32 xattr_len;
681 __le64 xattr_version;
682
683 /* filelock */
684 __le64 size, max_size, truncate_size;
685 __le32 truncate_seq;
686 struct ceph_timespec mtime, atime, ctime;
687 struct ceph_file_layout layout;
688 __le32 time_warp_seq;
689 } __attribute__ ((packed));
690
691 struct ceph_mds_cap_peer {
692 __le64 cap_id;
693 __le32 seq;
694 __le32 mseq;
695 __le32 mds;
696 __u8 flags;
697 } __attribute__ ((packed));
698
699 /* cap release msg head */
700 struct ceph_mds_cap_release {
701 __le32 num; /* number of cap_items that follow */
702 } __attribute__ ((packed));
703
704 struct ceph_mds_cap_item {
705 __le64 ino;
706 __le64 cap_id;
707 __le32 migrate_seq, seq;
708 } __attribute__ ((packed));
709
710 #define CEPH_MDS_LEASE_REVOKE 1 /* mds -> client */
711 #define CEPH_MDS_LEASE_RELEASE 2 /* client -> mds */
712 #define CEPH_MDS_LEASE_RENEW 3 /* client <-> mds */
713 #define CEPH_MDS_LEASE_REVOKE_ACK 4 /* client -> mds */
714
715 extern const char *ceph_lease_op_name(int o);
716
717 /* lease msg header */
718 struct ceph_mds_lease {
719 __u8 action; /* CEPH_MDS_LEASE_* */
720 __le16 mask; /* which lease */
721 __le64 ino;
722 __le64 first, last; /* snap range */
723 __le32 seq;
724 __le32 duration_ms; /* duration of renewal */
725 } __attribute__ ((packed));
726 /* followed by a __le32+string for dname */
727
728 /* client reconnect */
729 struct ceph_mds_cap_reconnect {
730 __le64 cap_id;
731 __le32 wanted;
732 __le32 issued;
733 __le64 snaprealm;
734 __le64 pathbase; /* base ino for our path to this ino */
735 __le32 flock_len; /* size of flock state blob, if any */
736 } __attribute__ ((packed));
737 /* followed by flock blob */
738
739 struct ceph_mds_cap_reconnect_v1 {
740 __le64 cap_id;
741 __le32 wanted;
742 __le32 issued;
743 __le64 size;
744 struct ceph_timespec mtime, atime;
745 __le64 snaprealm;
746 __le64 pathbase; /* base ino for our path to this ino */
747 } __attribute__ ((packed));
748
749 struct ceph_mds_snaprealm_reconnect {
750 __le64 ino; /* snap realm base */
751 __le64 seq; /* snap seq for this snap realm */
752 __le64 parent; /* parent realm */
753 } __attribute__ ((packed));
754
755 /*
756 * snaps
757 */
758 enum {
759 CEPH_SNAP_OP_UPDATE, /* CREATE or DESTROY */
760 CEPH_SNAP_OP_CREATE,
761 CEPH_SNAP_OP_DESTROY,
762 CEPH_SNAP_OP_SPLIT,
763 };
764
765 extern const char *ceph_snap_op_name(int o);
766
767 /* snap msg header */
768 struct ceph_mds_snap_head {
769 __le32 op; /* CEPH_SNAP_OP_* */
770 __le64 split; /* ino to split off, if any */
771 __le32 num_split_inos; /* # inos belonging to new child realm */
772 __le32 num_split_realms; /* # child realms udner new child realm */
773 __le32 trace_len; /* size of snap trace blob */
774 } __attribute__ ((packed));
775 /* followed by split ino list, then split realms, then the trace blob */
776
777 /*
778 * encode info about a snaprealm, as viewed by a client
779 */
780 struct ceph_mds_snap_realm {
781 __le64 ino; /* ino */
782 __le64 created; /* snap: when created */
783 __le64 parent; /* ino: parent realm */
784 __le64 parent_since; /* snap: same parent since */
785 __le64 seq; /* snap: version */
786 __le32 num_snaps;
787 __le32 num_prior_parent_snaps;
788 } __attribute__ ((packed));
789 /* followed by my snap list, then prior parent snap list */
790
791 #endif
792