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 | */ |
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_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 | |
107 | const char *ceph_auth_proto_name(int proto); |
108 | const 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 */ |
168 | enum { |
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 | |
175 | struct { |
176 | __le64 ; |
177 | __le16 ; |
178 | __le64 ; |
179 | } __attribute__ ((packed)); |
180 | |
181 | struct 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 | |
188 | struct ceph_statfs { |
189 | __le64 kb, kb_used, kb_avail; |
190 | __le64 num_objects; |
191 | } __attribute__ ((packed)); |
192 | |
193 | struct ceph_mon_statfs_reply { |
194 | struct ceph_fsid fsid; |
195 | __le64 version; |
196 | struct ceph_statfs st; |
197 | } __attribute__ ((packed)); |
198 | |
199 | struct 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 | |
207 | struct ceph_osd_getmap { |
208 | struct ceph_mon_request_header monhdr; |
209 | struct ceph_fsid fsid; |
210 | __le32 start; |
211 | } __attribute__ ((packed)); |
212 | |
213 | struct ceph_mds_getmap { |
214 | struct ceph_mon_request_header monhdr; |
215 | struct ceph_fsid fsid; |
216 | } __attribute__ ((packed)); |
217 | |
218 | struct 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 | |
224 | struct ceph_mon_subscribe_item { |
225 | __le64 start; |
226 | __u8 flags; |
227 | } __attribute__ ((packed)); |
228 | |
229 | struct 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 | |
265 | extern 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 */ |
289 | enum { |
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 | |
307 | extern const char *ceph_session_op_name(int op); |
308 | |
309 | struct 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 |
324 | enum { |
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 | |
360 | extern const char *ceph_mds_op_name(int op); |
361 | |
362 | #define CEPH_SETATTR_MODE (1 << 0) |
363 | #define CEPH_SETATTR_UID (1 << 1) |
364 | #define CEPH_SETATTR_GID (1 << 2) |
365 | #define CEPH_SETATTR_MTIME (1 << 3) |
366 | #define CEPH_SETATTR_ATIME (1 << 4) |
367 | #define CEPH_SETATTR_SIZE (1 << 5) |
368 | #define CEPH_SETATTR_CTIME (1 << 6) |
369 | #define CEPH_SETATTR_MTIME_NOW (1 << 7) |
370 | #define CEPH_SETATTR_ATIME_NOW (1 << 8) |
371 | #define CEPH_SETATTR_BTIME (1 << 9) |
372 | #define CEPH_SETATTR_KILL_SGUID (1 << 10) |
373 | #define CEPH_SETATTR_FSCRYPT_AUTH (1 << 11) |
374 | #define CEPH_SETATTR_FSCRYPT_FILE (1 << 12) |
375 | |
376 | /* |
377 | * Ceph setxattr request flags. |
378 | */ |
379 | #define CEPH_XATTR_CREATE (1 << 0) |
380 | #define CEPH_XATTR_REPLACE (1 << 1) |
381 | #define CEPH_XATTR_REMOVE (1 << 31) |
382 | |
383 | /* |
384 | * readdir request flags; |
385 | */ |
386 | #define CEPH_READDIR_REPLY_BITFLAGS (1<<0) |
387 | |
388 | /* |
389 | * readdir reply flags. |
390 | */ |
391 | #define CEPH_READDIR_FRAG_END (1<<0) |
392 | #define CEPH_READDIR_FRAG_COMPLETE (1<<8) |
393 | #define CEPH_READDIR_HASH_ORDER (1<<9) |
394 | #define CEPH_READDIR_OFFSET_HASH (1<<10) |
395 | |
396 | /* |
397 | * open request flags |
398 | */ |
399 | #define CEPH_O_RDONLY 00000000 |
400 | #define CEPH_O_WRONLY 00000001 |
401 | #define CEPH_O_RDWR 00000002 |
402 | #define CEPH_O_CREAT 00000100 |
403 | #define CEPH_O_EXCL 00000200 |
404 | #define CEPH_O_TRUNC 00001000 |
405 | #define CEPH_O_DIRECTORY 00200000 |
406 | #define CEPH_O_NOFOLLOW 00400000 |
407 | |
408 | union ceph_mds_request_args { |
409 | struct { |
410 | __le32 mask; /* CEPH_CAP_* */ |
411 | } __attribute__ ((packed)) getattr; |
412 | struct { |
413 | __le32 mode; |
414 | __le32 uid; |
415 | __le32 gid; |
416 | struct ceph_timespec mtime; |
417 | struct ceph_timespec atime; |
418 | __le64 size, old_size; /* old_size needed by truncate */ |
419 | __le32 mask; /* CEPH_SETATTR_* */ |
420 | } __attribute__ ((packed)) setattr; |
421 | struct { |
422 | __le32 frag; /* which dir fragment */ |
423 | __le32 max_entries; /* how many dentries to grab */ |
424 | __le32 max_bytes; |
425 | __le16 flags; |
426 | __le32 offset_hash; |
427 | } __attribute__ ((packed)) readdir; |
428 | struct { |
429 | __le32 mode; |
430 | __le32 rdev; |
431 | } __attribute__ ((packed)) mknod; |
432 | struct { |
433 | __le32 mode; |
434 | } __attribute__ ((packed)) mkdir; |
435 | struct { |
436 | __le32 flags; |
437 | __le32 mode; |
438 | __le32 stripe_unit; /* layout for newly created file */ |
439 | __le32 stripe_count; /* ... */ |
440 | __le32 object_size; |
441 | __le32 pool; |
442 | __le32 mask; /* CEPH_CAP_* */ |
443 | __le64 old_size; |
444 | } __attribute__ ((packed)) open; |
445 | struct { |
446 | __le32 flags; |
447 | __le32 osdmap_epoch; /* used for setting file/dir layouts */ |
448 | } __attribute__ ((packed)) setxattr; |
449 | struct { |
450 | struct ceph_file_layout_legacy layout; |
451 | } __attribute__ ((packed)) setlayout; |
452 | struct { |
453 | __u8 rule; /* currently fcntl or flock */ |
454 | __u8 type; /* shared, exclusive, remove*/ |
455 | __le64 owner; /* owner of the lock */ |
456 | __le64 pid; /* process id requesting the lock */ |
457 | __le64 start; /* initial location to lock */ |
458 | __le64 length; /* num bytes to lock from start */ |
459 | __u8 wait; /* will caller wait for lock to become available? */ |
460 | } __attribute__ ((packed)) filelock_change; |
461 | struct { |
462 | __le32 mask; /* CEPH_CAP_* */ |
463 | __le64 snapid; |
464 | __le64 parent; |
465 | __le32 hash; |
466 | } __attribute__ ((packed)) lookupino; |
467 | } __attribute__ ((packed)); |
468 | |
469 | union ceph_mds_request_args_ext { |
470 | union ceph_mds_request_args old; |
471 | struct { |
472 | __le32 mode; |
473 | __le32 uid; |
474 | __le32 gid; |
475 | struct ceph_timespec mtime; |
476 | struct ceph_timespec atime; |
477 | __le64 size, old_size; /* old_size needed by truncate */ |
478 | __le32 mask; /* CEPH_SETATTR_* */ |
479 | struct ceph_timespec btime; |
480 | } __attribute__ ((packed)) setattr_ext; |
481 | }; |
482 | |
483 | #define CEPH_MDS_FLAG_REPLAY 1 /* this is a replayed op */ |
484 | #define CEPH_MDS_FLAG_WANT_DENTRY 2 /* want dentry in reply */ |
485 | #define CEPH_MDS_FLAG_ASYNC 4 /* request is asynchronous */ |
486 | |
487 | struct ceph_mds_request_head_legacy { |
488 | __le64 oldest_client_tid; |
489 | __le32 mdsmap_epoch; /* on client */ |
490 | __le32 flags; /* CEPH_MDS_FLAG_* */ |
491 | __u8 num_retry, num_fwd; /* count retry, fwd attempts */ |
492 | __le16 num_releases; /* # include cap/lease release records */ |
493 | __le32 op; /* mds op code */ |
494 | __le32 caller_uid, caller_gid; |
495 | __le64 ino; /* use this ino for openc, mkdir, mknod, |
496 | etc. (if replaying) */ |
497 | union ceph_mds_request_args args; |
498 | } __attribute__ ((packed)); |
499 | |
500 | #define CEPH_MDS_REQUEST_HEAD_VERSION 2 |
501 | |
502 | struct ceph_mds_request_head_old { |
503 | __le16 version; /* struct version */ |
504 | __le64 oldest_client_tid; |
505 | __le32 mdsmap_epoch; /* on client */ |
506 | __le32 flags; /* CEPH_MDS_FLAG_* */ |
507 | __u8 num_retry, num_fwd; /* count retry, fwd attempts */ |
508 | __le16 num_releases; /* # include cap/lease release records */ |
509 | __le32 op; /* mds op code */ |
510 | __le32 caller_uid, caller_gid; |
511 | __le64 ino; /* use this ino for openc, mkdir, mknod, |
512 | etc. (if replaying) */ |
513 | union ceph_mds_request_args_ext args; |
514 | } __attribute__ ((packed)); |
515 | |
516 | struct ceph_mds_request_head { |
517 | __le16 version; /* struct version */ |
518 | __le64 oldest_client_tid; |
519 | __le32 mdsmap_epoch; /* on client */ |
520 | __le32 flags; /* CEPH_MDS_FLAG_* */ |
521 | __u8 num_retry, num_fwd; /* legacy count retry and fwd attempts */ |
522 | __le16 num_releases; /* # include cap/lease release records */ |
523 | __le32 op; /* mds op code */ |
524 | __le32 caller_uid, caller_gid; |
525 | __le64 ino; /* use this ino for openc, mkdir, mknod, |
526 | etc. (if replaying) */ |
527 | union ceph_mds_request_args_ext args; |
528 | |
529 | __le32 ext_num_retry; /* new count retry attempts */ |
530 | __le32 ext_num_fwd; /* new count fwd attempts */ |
531 | } __attribute__ ((packed)); |
532 | |
533 | /* cap/lease release record */ |
534 | struct ceph_mds_request_release { |
535 | __le64 ino, cap_id; /* ino and unique cap id */ |
536 | __le32 caps, wanted; /* new issued, wanted */ |
537 | __le32 seq, issue_seq, mseq; |
538 | __le32 dname_seq; /* if releasing a dentry lease, a */ |
539 | __le32 dname_len; /* string follows. */ |
540 | } __attribute__ ((packed)); |
541 | |
542 | /* client reply */ |
543 | struct ceph_mds_reply_head { |
544 | __le32 op; |
545 | __le32 result; |
546 | __le32 mdsmap_epoch; |
547 | __u8 safe; /* true if committed to disk */ |
548 | __u8 is_dentry, is_target; /* true if dentry, target inode records |
549 | are included with reply */ |
550 | } __attribute__ ((packed)); |
551 | |
552 | /* one for each node split */ |
553 | struct ceph_frag_tree_split { |
554 | __le32 frag; /* this frag splits... */ |
555 | __le32 by; /* ...by this many bits */ |
556 | } __attribute__ ((packed)); |
557 | |
558 | struct ceph_frag_tree_head { |
559 | __le32 nsplits; /* num ceph_frag_tree_split records */ |
560 | struct ceph_frag_tree_split splits[]; |
561 | } __attribute__ ((packed)); |
562 | |
563 | /* capability issue, for bundling with mds reply */ |
564 | struct ceph_mds_reply_cap { |
565 | __le32 caps, wanted; /* caps issued, wanted */ |
566 | __le64 cap_id; |
567 | __le32 seq, mseq; |
568 | __le64 realm; /* snap realm */ |
569 | __u8 flags; /* CEPH_CAP_FLAG_* */ |
570 | } __attribute__ ((packed)); |
571 | |
572 | #define CEPH_CAP_FLAG_AUTH (1 << 0) /* cap is issued by auth mds */ |
573 | #define CEPH_CAP_FLAG_RELEASE (1 << 1) /* release the cap */ |
574 | |
575 | /* inode record, for bundling with mds reply */ |
576 | struct ceph_mds_reply_inode { |
577 | __le64 ino; |
578 | __le64 snapid; |
579 | __le32 rdev; |
580 | __le64 version; /* inode version */ |
581 | __le64 xattr_version; /* version for xattr blob */ |
582 | struct ceph_mds_reply_cap cap; /* caps issued for this inode */ |
583 | struct ceph_file_layout_legacy layout; |
584 | struct ceph_timespec ctime, mtime, atime; |
585 | __le32 time_warp_seq; |
586 | __le64 size, max_size, truncate_size; |
587 | __le32 truncate_seq; |
588 | __le32 mode, uid, gid; |
589 | __le32 nlink; |
590 | __le64 files, subdirs, rbytes, rfiles, rsubdirs; /* dir stats */ |
591 | struct ceph_timespec rctime; |
592 | struct ceph_frag_tree_head fragtree; /* (must be at end of struct) */ |
593 | } __attribute__ ((packed)); |
594 | /* followed by frag array, symlink string, dir layout, xattr blob */ |
595 | |
596 | /* reply_lease follows dname, and reply_inode */ |
597 | struct ceph_mds_reply_lease { |
598 | __le16 mask; /* lease type(s) */ |
599 | __le32 duration_ms; /* lease duration */ |
600 | __le32 seq; |
601 | } __attribute__ ((packed)); |
602 | |
603 | #define CEPH_LEASE_VALID (1 | 2) /* old and new bit values */ |
604 | #define CEPH_LEASE_PRIMARY_LINK 4 /* primary linkage */ |
605 | |
606 | struct ceph_mds_reply_dirfrag { |
607 | __le32 frag; /* fragment */ |
608 | __le32 auth; /* auth mds, if this is a delegation point */ |
609 | __le32 ndist; /* number of mds' this is replicated on */ |
610 | __le32 dist[]; |
611 | } __attribute__ ((packed)); |
612 | |
613 | #define CEPH_LOCK_FCNTL 1 |
614 | #define CEPH_LOCK_FLOCK 2 |
615 | #define CEPH_LOCK_FCNTL_INTR 3 |
616 | #define CEPH_LOCK_FLOCK_INTR 4 |
617 | |
618 | |
619 | #define CEPH_LOCK_SHARED 1 |
620 | #define CEPH_LOCK_EXCL 2 |
621 | #define CEPH_LOCK_UNLOCK 4 |
622 | |
623 | struct ceph_filelock { |
624 | __le64 start;/* file offset to start lock at */ |
625 | __le64 length; /* num bytes to lock; 0 for all following start */ |
626 | __le64 client; /* which client holds the lock */ |
627 | __le64 owner; /* owner the lock */ |
628 | __le64 pid; /* process id holding the lock on the client */ |
629 | __u8 type; /* shared lock, exclusive lock, or unlock */ |
630 | } __attribute__ ((packed)); |
631 | |
632 | |
633 | /* file access modes */ |
634 | #define CEPH_FILE_MODE_PIN 0 |
635 | #define CEPH_FILE_MODE_RD 1 |
636 | #define CEPH_FILE_MODE_WR 2 |
637 | #define CEPH_FILE_MODE_RDWR 3 /* RD | WR */ |
638 | #define CEPH_FILE_MODE_LAZY 4 /* lazy io */ |
639 | #define CEPH_FILE_MODE_BITS 4 |
640 | #define CEPH_FILE_MODE_MASK ((1 << CEPH_FILE_MODE_BITS) - 1) |
641 | |
642 | int ceph_flags_to_mode(int flags); |
643 | |
644 | #define CEPH_INLINE_NONE ((__u64)-1) |
645 | |
646 | /* capability bits */ |
647 | #define CEPH_CAP_PIN 1 /* no specific capabilities beyond the pin */ |
648 | |
649 | /* generic cap bits */ |
650 | #define CEPH_CAP_GSHARED 1 /* client can reads */ |
651 | #define CEPH_CAP_GEXCL 2 /* client can read and update */ |
652 | #define CEPH_CAP_GCACHE 4 /* (file) client can cache reads */ |
653 | #define CEPH_CAP_GRD 8 /* (file) client can read */ |
654 | #define CEPH_CAP_GWR 16 /* (file) client can write */ |
655 | #define CEPH_CAP_GBUFFER 32 /* (file) client can buffer writes */ |
656 | #define CEPH_CAP_GWREXTEND 64 /* (file) client can extend EOF */ |
657 | #define CEPH_CAP_GLAZYIO 128 /* (file) client can perform lazy io */ |
658 | |
659 | #define CEPH_CAP_SIMPLE_BITS 2 |
660 | #define CEPH_CAP_FILE_BITS 8 |
661 | |
662 | /* per-lock shift */ |
663 | #define CEPH_CAP_SAUTH 2 |
664 | #define CEPH_CAP_SLINK 4 |
665 | #define CEPH_CAP_SXATTR 6 |
666 | #define CEPH_CAP_SFILE 8 |
667 | #define CEPH_CAP_SFLOCK 20 |
668 | |
669 | #define CEPH_CAP_BITS 22 |
670 | |
671 | /* composed values */ |
672 | #define CEPH_CAP_AUTH_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SAUTH) |
673 | #define CEPH_CAP_AUTH_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SAUTH) |
674 | #define CEPH_CAP_LINK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SLINK) |
675 | #define CEPH_CAP_LINK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SLINK) |
676 | #define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SXATTR) |
677 | #define CEPH_CAP_XATTR_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SXATTR) |
678 | #define CEPH_CAP_FILE(x) (x << CEPH_CAP_SFILE) |
679 | #define CEPH_CAP_FILE_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFILE) |
680 | #define CEPH_CAP_FILE_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFILE) |
681 | #define CEPH_CAP_FILE_CACHE (CEPH_CAP_GCACHE << CEPH_CAP_SFILE) |
682 | #define CEPH_CAP_FILE_RD (CEPH_CAP_GRD << CEPH_CAP_SFILE) |
683 | #define CEPH_CAP_FILE_WR (CEPH_CAP_GWR << CEPH_CAP_SFILE) |
684 | #define CEPH_CAP_FILE_BUFFER (CEPH_CAP_GBUFFER << CEPH_CAP_SFILE) |
685 | #define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE) |
686 | #define CEPH_CAP_FILE_LAZYIO (CEPH_CAP_GLAZYIO << CEPH_CAP_SFILE) |
687 | #define CEPH_CAP_FLOCK_SHARED (CEPH_CAP_GSHARED << CEPH_CAP_SFLOCK) |
688 | #define CEPH_CAP_FLOCK_EXCL (CEPH_CAP_GEXCL << CEPH_CAP_SFLOCK) |
689 | |
690 | |
691 | /* cap masks (for getattr) */ |
692 | #define CEPH_STAT_CAP_INODE CEPH_CAP_PIN |
693 | #define CEPH_STAT_CAP_TYPE CEPH_CAP_PIN /* mode >> 12 */ |
694 | #define CEPH_STAT_CAP_SYMLINK CEPH_CAP_PIN |
695 | #define CEPH_STAT_CAP_UID CEPH_CAP_AUTH_SHARED |
696 | #define CEPH_STAT_CAP_GID CEPH_CAP_AUTH_SHARED |
697 | #define CEPH_STAT_CAP_MODE CEPH_CAP_AUTH_SHARED |
698 | #define CEPH_STAT_CAP_NLINK CEPH_CAP_LINK_SHARED |
699 | #define CEPH_STAT_CAP_LAYOUT CEPH_CAP_FILE_SHARED |
700 | #define CEPH_STAT_CAP_MTIME CEPH_CAP_FILE_SHARED |
701 | #define CEPH_STAT_CAP_SIZE CEPH_CAP_FILE_SHARED |
702 | #define CEPH_STAT_CAP_ATIME CEPH_CAP_FILE_SHARED /* fixme */ |
703 | #define CEPH_STAT_CAP_XATTR CEPH_CAP_XATTR_SHARED |
704 | #define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN | \ |
705 | CEPH_CAP_AUTH_SHARED | \ |
706 | CEPH_CAP_LINK_SHARED | \ |
707 | CEPH_CAP_FILE_SHARED | \ |
708 | CEPH_CAP_XATTR_SHARED) |
709 | #define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \ |
710 | CEPH_CAP_FILE_RD) |
711 | #define CEPH_STAT_RSTAT CEPH_CAP_FILE_WREXTEND |
712 | |
713 | #define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED | \ |
714 | CEPH_CAP_LINK_SHARED | \ |
715 | CEPH_CAP_XATTR_SHARED | \ |
716 | CEPH_CAP_FILE_SHARED) |
717 | #define CEPH_CAP_ANY_RD (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD | \ |
718 | CEPH_CAP_FILE_CACHE) |
719 | |
720 | #define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL | \ |
721 | CEPH_CAP_LINK_EXCL | \ |
722 | CEPH_CAP_XATTR_EXCL | \ |
723 | CEPH_CAP_FILE_EXCL) |
724 | #define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \ |
725 | CEPH_CAP_FILE_SHARED) |
726 | #define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER | \ |
727 | CEPH_CAP_FILE_EXCL) |
728 | #define CEPH_CAP_ANY_WR (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR) |
729 | #define CEPH_CAP_ANY (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \ |
730 | CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \ |
731 | CEPH_CAP_PIN) |
732 | #define CEPH_CAP_ALL_FILE (CEPH_CAP_PIN | CEPH_CAP_ANY_SHARED | \ |
733 | CEPH_CAP_AUTH_EXCL | CEPH_CAP_XATTR_EXCL | \ |
734 | CEPH_CAP_ANY_FILE_RD | CEPH_CAP_ANY_FILE_WR) |
735 | |
736 | #define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \ |
737 | CEPH_LOCK_IXATTR) |
738 | |
739 | /* cap masks async dir operations */ |
740 | #define CEPH_CAP_DIR_CREATE CEPH_CAP_FILE_CACHE |
741 | #define CEPH_CAP_DIR_UNLINK CEPH_CAP_FILE_RD |
742 | #define CEPH_CAP_ANY_DIR_OPS (CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_RD | \ |
743 | CEPH_CAP_FILE_WREXTEND | CEPH_CAP_FILE_LAZYIO) |
744 | |
745 | int ceph_caps_for_mode(int mode); |
746 | |
747 | enum { |
748 | CEPH_CAP_OP_GRANT, /* mds->client grant */ |
749 | CEPH_CAP_OP_REVOKE, /* mds->client revoke */ |
750 | CEPH_CAP_OP_TRUNC, /* mds->client trunc notify */ |
751 | CEPH_CAP_OP_EXPORT, /* mds has exported the cap */ |
752 | CEPH_CAP_OP_IMPORT, /* mds has imported the cap */ |
753 | CEPH_CAP_OP_UPDATE, /* client->mds update */ |
754 | CEPH_CAP_OP_DROP, /* client->mds drop cap bits */ |
755 | CEPH_CAP_OP_FLUSH, /* client->mds cap writeback */ |
756 | CEPH_CAP_OP_FLUSH_ACK, /* mds->client flushed */ |
757 | CEPH_CAP_OP_FLUSHSNAP, /* client->mds flush snapped metadata */ |
758 | CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */ |
759 | CEPH_CAP_OP_RELEASE, /* client->mds release (clean) cap */ |
760 | CEPH_CAP_OP_RENEW, /* client->mds renewal request */ |
761 | }; |
762 | |
763 | extern const char *ceph_cap_op_name(int op); |
764 | |
765 | /* flags field in client cap messages (version >= 10) */ |
766 | #define CEPH_CLIENT_CAPS_SYNC (1<<0) |
767 | #define CEPH_CLIENT_CAPS_NO_CAPSNAP (1<<1) |
768 | #define CEPH_CLIENT_CAPS_PENDING_CAPSNAP (1<<2) |
769 | |
770 | /* |
771 | * caps message, used for capability callbacks, acks, requests, etc. |
772 | */ |
773 | struct ceph_mds_caps { |
774 | __le32 op; /* CEPH_CAP_OP_* */ |
775 | __le64 ino, realm; |
776 | __le64 cap_id; |
777 | __le32 seq, issue_seq; |
778 | __le32 caps, wanted, dirty; /* latest issued/wanted/dirty */ |
779 | __le32 migrate_seq; |
780 | __le64 snap_follows; |
781 | __le32 snap_trace_len; |
782 | |
783 | /* authlock */ |
784 | __le32 uid, gid, mode; |
785 | |
786 | /* linklock */ |
787 | __le32 nlink; |
788 | |
789 | /* xattrlock */ |
790 | __le32 xattr_len; |
791 | __le64 xattr_version; |
792 | |
793 | /* a union of non-export and export bodies. */ |
794 | __le64 size, max_size, truncate_size; |
795 | __le32 truncate_seq; |
796 | struct ceph_timespec mtime, atime, ctime; |
797 | struct ceph_file_layout_legacy layout; |
798 | __le32 time_warp_seq; |
799 | } __attribute__ ((packed)); |
800 | |
801 | struct ceph_mds_cap_peer { |
802 | __le64 cap_id; |
803 | __le32 seq; |
804 | __le32 mseq; |
805 | __le32 mds; |
806 | __u8 flags; |
807 | } __attribute__ ((packed)); |
808 | |
809 | /* cap release msg head */ |
810 | struct ceph_mds_cap_release { |
811 | __le32 num; /* number of cap_items that follow */ |
812 | } __attribute__ ((packed)); |
813 | |
814 | struct ceph_mds_cap_item { |
815 | __le64 ino; |
816 | __le64 cap_id; |
817 | __le32 migrate_seq, seq; |
818 | } __attribute__ ((packed)); |
819 | |
820 | #define CEPH_MDS_LEASE_REVOKE 1 /* mds -> client */ |
821 | #define CEPH_MDS_LEASE_RELEASE 2 /* client -> mds */ |
822 | #define CEPH_MDS_LEASE_RENEW 3 /* client <-> mds */ |
823 | #define CEPH_MDS_LEASE_REVOKE_ACK 4 /* client -> mds */ |
824 | |
825 | extern const char *ceph_lease_op_name(int o); |
826 | |
827 | /* lease msg header */ |
828 | struct ceph_mds_lease { |
829 | __u8 action; /* CEPH_MDS_LEASE_* */ |
830 | __le16 mask; /* which lease */ |
831 | __le64 ino; |
832 | __le64 first, last; /* snap range */ |
833 | __le32 seq; |
834 | __le32 duration_ms; /* duration of renewal */ |
835 | } __attribute__ ((packed)); |
836 | /* followed by a __le32+string for dname */ |
837 | |
838 | /* client reconnect */ |
839 | struct ceph_mds_cap_reconnect { |
840 | __le64 cap_id; |
841 | __le32 wanted; |
842 | __le32 issued; |
843 | __le64 snaprealm; |
844 | __le64 pathbase; /* base ino for our path to this ino */ |
845 | __le32 flock_len; /* size of flock state blob, if any */ |
846 | } __attribute__ ((packed)); |
847 | /* followed by flock blob */ |
848 | |
849 | struct ceph_mds_cap_reconnect_v1 { |
850 | __le64 cap_id; |
851 | __le32 wanted; |
852 | __le32 issued; |
853 | __le64 size; |
854 | struct ceph_timespec mtime, atime; |
855 | __le64 snaprealm; |
856 | __le64 pathbase; /* base ino for our path to this ino */ |
857 | } __attribute__ ((packed)); |
858 | |
859 | struct ceph_mds_snaprealm_reconnect { |
860 | __le64 ino; /* snap realm base */ |
861 | __le64 seq; /* snap seq for this snap realm */ |
862 | __le64 parent; /* parent realm */ |
863 | } __attribute__ ((packed)); |
864 | |
865 | /* |
866 | * snaps |
867 | */ |
868 | enum { |
869 | CEPH_SNAP_OP_UPDATE, /* CREATE or DESTROY */ |
870 | CEPH_SNAP_OP_CREATE, |
871 | CEPH_SNAP_OP_DESTROY, |
872 | CEPH_SNAP_OP_SPLIT, |
873 | }; |
874 | |
875 | extern const char *ceph_snap_op_name(int o); |
876 | |
877 | /* snap msg header */ |
878 | struct ceph_mds_snap_head { |
879 | __le32 op; /* CEPH_SNAP_OP_* */ |
880 | __le64 split; /* ino to split off, if any */ |
881 | __le32 num_split_inos; /* # inos belonging to new child realm */ |
882 | __le32 num_split_realms; /* # child realms udner new child realm */ |
883 | __le32 trace_len; /* size of snap trace blob */ |
884 | } __attribute__ ((packed)); |
885 | /* followed by split ino list, then split realms, then the trace blob */ |
886 | |
887 | /* |
888 | * encode info about a snaprealm, as viewed by a client |
889 | */ |
890 | struct ceph_mds_snap_realm { |
891 | __le64 ino; /* ino */ |
892 | __le64 created; /* snap: when created */ |
893 | __le64 parent; /* ino: parent realm */ |
894 | __le64 parent_since; /* snap: same parent since */ |
895 | __le64 seq; /* snap: version */ |
896 | __le32 num_snaps; |
897 | __le32 num_prior_parent_snaps; |
898 | } __attribute__ ((packed)); |
899 | /* followed by my snap list, then prior parent snap list */ |
900 | |
901 | /* |
902 | * quotas |
903 | */ |
904 | struct ceph_mds_quota { |
905 | __le64 ino; /* ino */ |
906 | struct ceph_timespec rctime; |
907 | __le64 rbytes; /* dir stats */ |
908 | __le64 rfiles; |
909 | __le64 rsubdirs; |
910 | __u8 struct_v; /* compat */ |
911 | __u8 struct_compat; |
912 | __le32 struct_len; |
913 | __le64 max_bytes; /* quota max. bytes */ |
914 | __le64 max_files; /* quota max. files */ |
915 | } __attribute__ ((packed)); |
916 | |
917 | #endif |
918 | |