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 "msgr.h"
16#include "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_OSD_PROTOCOL     8 /* cluster internal */
25#define CEPH_MDS_PROTOCOL    12 /* cluster internal */
26#define CEPH_MON_PROTOCOL     5 /* cluster internal */
27#define CEPH_OSDC_PROTOCOL   24 /* server/client */
28#define CEPH_MDSC_PROTOCOL   32 /* server/client */
29#define CEPH_MONC_PROTOCOL   15 /* server/client */
30
31
32#define CEPH_INO_ROOT  1
33#define CEPH_INO_CEPH  2        /* hidden .ceph dir */
34
35/* arbitrary limit on max # of monitors (cluster of 3 is typical) */
36#define CEPH_MAX_MON   31
37
38
39/*
40 * feature bits
41 */
42#define CEPH_FEATURE_UID            (1<<0)
43#define CEPH_FEATURE_NOSRCADDR      (1<<1)
44#define CEPH_FEATURE_MONCLOCKCHECK  (1<<2)
45#define CEPH_FEATURE_FLOCK          (1<<3)
46
47
48/*
49 * ceph_file_layout - describe data layout for a file/inode
50 */
51struct ceph_file_layout {
52	/* file -> object mapping */
53	__le32 fl_stripe_unit;     /* stripe unit, in bytes.  must be multiple
54				      of page size. */
55	__le32 fl_stripe_count;    /* over this many objects */
56	__le32 fl_object_size;     /* until objects are this big, then move to
57				      new objects */
58	__le32 fl_cas_hash;        /* 0 = none; 1 = sha256 */
59
60	/* pg -> disk layout */
61	__le32 fl_object_stripe_unit;  /* for per-object parity, if any */
62
63	/* object -> pg layout */
64	__le32 fl_pg_preferred; /* preferred primary for pg (-1 for none) */
65	__le32 fl_pg_pool;      /* namespace, crush ruleset, rep level */
66} __attribute__ ((packed));
67
68#define CEPH_MIN_STRIPE_UNIT 65536
69
70int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
71
72
73/* crypto algorithms */
74#define CEPH_CRYPTO_NONE 0x0
75#define CEPH_CRYPTO_AES  0x1
76
77#define CEPH_AES_IV "cephsageyudagreg"
78
79/* security/authentication protocols */
80#define CEPH_AUTH_UNKNOWN	0x0
81#define CEPH_AUTH_NONE	 	0x1
82#define CEPH_AUTH_CEPHX	 	0x2
83
84#define CEPH_AUTH_UID_DEFAULT ((__u64) -1)
85
86
87/*********************************************
88 * message layer
89 */
90
91/*
92 * message types
93 */
94
95/* misc */
96#define CEPH_MSG_SHUTDOWN               1
97#define CEPH_MSG_PING                   2
98
99/* client <-> monitor */
100#define CEPH_MSG_MON_MAP                4
101#define CEPH_MSG_MON_GET_MAP            5
102#define CEPH_MSG_STATFS                 13
103#define CEPH_MSG_STATFS_REPLY           14
104#define CEPH_MSG_MON_SUBSCRIBE          15
105#define CEPH_MSG_MON_SUBSCRIBE_ACK      16
106#define CEPH_MSG_AUTH			17
107#define CEPH_MSG_AUTH_REPLY		18
108
109/* client <-> mds */
110#define CEPH_MSG_MDS_MAP                21
111
112#define CEPH_MSG_CLIENT_SESSION         22
113#define CEPH_MSG_CLIENT_RECONNECT       23
114
115#define CEPH_MSG_CLIENT_REQUEST         24
116#define CEPH_MSG_CLIENT_REQUEST_FORWARD 25
117#define CEPH_MSG_CLIENT_REPLY           26
118#define CEPH_MSG_CLIENT_CAPS            0x310
119#define CEPH_MSG_CLIENT_LEASE           0x311
120#define CEPH_MSG_CLIENT_SNAP            0x312
121#define CEPH_MSG_CLIENT_CAPRELEASE      0x313
122
123/* pool ops */
124#define CEPH_MSG_POOLOP_REPLY           48
125#define CEPH_MSG_POOLOP                 49
126
127
128/* osd */
129#define CEPH_MSG_OSD_MAP          41
130#define CEPH_MSG_OSD_OP           42
131#define CEPH_MSG_OSD_OPREPLY      43
132
133/* pool operations */
134enum {
135  POOL_OP_CREATE			= 0x01,
136  POOL_OP_DELETE			= 0x02,
137  POOL_OP_AUID_CHANGE			= 0x03,
138  POOL_OP_CREATE_SNAP			= 0x11,
139  POOL_OP_DELETE_SNAP			= 0x12,
140  POOL_OP_CREATE_UNMANAGED_SNAP		= 0x21,
141  POOL_OP_DELETE_UNMANAGED_SNAP		= 0x22,
142};
143
144struct ceph_mon_request_header {
145	__le64 have_version;
146	__le16 session_mon;
147	__le64 session_mon_tid;
148} __attribute__ ((packed));
149
150struct ceph_mon_statfs {
151	struct ceph_mon_request_header monhdr;
152	struct ceph_fsid fsid;
153} __attribute__ ((packed));
154
155struct ceph_statfs {
156	__le64 kb, kb_used, kb_avail;
157	__le64 num_objects;
158} __attribute__ ((packed));
159
160struct ceph_mon_statfs_reply {
161	struct ceph_fsid fsid;
162	__le64 version;
163	struct ceph_statfs st;
164} __attribute__ ((packed));
165
166const char *ceph_pool_op_name(int op);
167
168struct ceph_mon_poolop {
169	struct ceph_mon_request_header monhdr;
170	struct ceph_fsid fsid;
171	__le32 pool;
172	__le32 op;
173	__le64 auid;
174	__le64 snapid;
175	__le32 name_len;
176} __attribute__ ((packed));
177
178struct ceph_mon_poolop_reply {
179	struct ceph_mon_request_header monhdr;
180	struct ceph_fsid fsid;
181	__le32 reply_code;
182	__le32 epoch;
183	char has_data;
184	char data[0];
185} __attribute__ ((packed));
186
187struct ceph_mon_unmanaged_snap {
188	__le64 snapid;
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
206struct ceph_mon_subscribe_item {
207	__le64 have_version;	__le64 have;
208	__u8 onetime;
209} __attribute__ ((packed));
210
211struct ceph_mon_subscribe_ack {
212	__le32 duration;         /* seconds */
213	struct ceph_fsid fsid;
214} __attribute__ ((packed));
215
216/*
217 * mds states
218 *   > 0 -> in
219 *  <= 0 -> out
220 */
221#define CEPH_MDS_STATE_DNE          0  /* down, does not exist. */
222#define CEPH_MDS_STATE_STOPPED     -1  /* down, once existed, but no subtrees.
223					  empty log. */
224#define CEPH_MDS_STATE_BOOT        -4  /* up, boot announcement. */
225#define CEPH_MDS_STATE_STANDBY     -5  /* up, idle.  waiting for assignment. */
226#define CEPH_MDS_STATE_CREATING    -6  /* up, creating MDS instance. */
227#define CEPH_MDS_STATE_STARTING    -7  /* up, starting previously stopped mds */
228#define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
229
230#define CEPH_MDS_STATE_REPLAY       8  /* up, replaying journal. */
231#define CEPH_MDS_STATE_RESOLVE      9  /* up, disambiguating distributed
232					  operations (import, rename, etc.) */
233#define CEPH_MDS_STATE_RECONNECT    10 /* up, reconnect to clients */
234#define CEPH_MDS_STATE_REJOIN       11 /* up, rejoining distributed cache */
235#define CEPH_MDS_STATE_CLIENTREPLAY 12 /* up, replaying client operations */
236#define CEPH_MDS_STATE_ACTIVE       13 /* up, active */
237#define CEPH_MDS_STATE_STOPPING     14 /* up, but exporting metadata */
238
239extern const char *ceph_mds_state_name(int s);
240
241
242/*
243 * metadata lock types.
244 *  - these are bitmasks.. we can compose them
245 *  - they also define the lock ordering by the MDS
246 *  - a few of these are internal to the mds
247 */
248#define CEPH_LOCK_DVERSION    1
249#define CEPH_LOCK_DN          2
250#define CEPH_LOCK_ISNAP       16
251#define CEPH_LOCK_IVERSION    32    /* mds internal */
252#define CEPH_LOCK_IFILE       64
253#define CEPH_LOCK_IAUTH       128
254#define CEPH_LOCK_ILINK       256
255#define CEPH_LOCK_IDFT        512   /* dir frag tree */
256#define CEPH_LOCK_INEST       1024  /* mds internal */
257#define CEPH_LOCK_IXATTR      2048
258#define CEPH_LOCK_IFLOCK      4096  /* advisory file locks */
259#define CEPH_LOCK_INO         8192  /* immutable inode bits; not a lock */
260
261/* client_session ops */
262enum {
263	CEPH_SESSION_REQUEST_OPEN,
264	CEPH_SESSION_OPEN,
265	CEPH_SESSION_REQUEST_CLOSE,
266	CEPH_SESSION_CLOSE,
267	CEPH_SESSION_REQUEST_RENEWCAPS,
268	CEPH_SESSION_RENEWCAPS,
269	CEPH_SESSION_STALE,
270	CEPH_SESSION_RECALL_STATE,
271};
272
273extern const char *ceph_session_op_name(int op);
274
275struct ceph_mds_session_head {
276	__le32 op;
277	__le64 seq;
278	struct ceph_timespec stamp;
279	__le32 max_caps, max_leases;
280} __attribute__ ((packed));
281
282/* client_request */
283/*
284 * metadata ops.
285 *  & 0x001000 -> write op
286 *  & 0x010000 -> follow symlink (e.g. stat(), not lstat()).
287 &  & 0x100000 -> use weird ino/path trace
288 */
289#define CEPH_MDS_OP_WRITE        0x001000
290enum {
291	CEPH_MDS_OP_LOOKUP     = 0x00100,
292	CEPH_MDS_OP_GETATTR    = 0x00101,
293	CEPH_MDS_OP_LOOKUPHASH = 0x00102,
294	CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
295
296	CEPH_MDS_OP_SETXATTR   = 0x01105,
297	CEPH_MDS_OP_RMXATTR    = 0x01106,
298	CEPH_MDS_OP_SETLAYOUT  = 0x01107,
299	CEPH_MDS_OP_SETATTR    = 0x01108,
300	CEPH_MDS_OP_SETFILELOCK= 0x01109,
301	CEPH_MDS_OP_GETFILELOCK= 0x00110,
302
303	CEPH_MDS_OP_MKNOD      = 0x01201,
304	CEPH_MDS_OP_LINK       = 0x01202,
305	CEPH_MDS_OP_UNLINK     = 0x01203,
306	CEPH_MDS_OP_RENAME     = 0x01204,
307	CEPH_MDS_OP_MKDIR      = 0x01220,
308	CEPH_MDS_OP_RMDIR      = 0x01221,
309	CEPH_MDS_OP_SYMLINK    = 0x01222,
310
311	CEPH_MDS_OP_CREATE     = 0x01301,
312	CEPH_MDS_OP_OPEN       = 0x00302,
313	CEPH_MDS_OP_READDIR    = 0x00305,
314
315	CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
316	CEPH_MDS_OP_MKSNAP     = 0x01400,
317	CEPH_MDS_OP_RMSNAP     = 0x01401,
318	CEPH_MDS_OP_LSSNAP     = 0x00402,
319};
320
321extern const char *ceph_mds_op_name(int op);
322
323
324#define CEPH_SETATTR_MODE   1
325#define CEPH_SETATTR_UID    2
326#define CEPH_SETATTR_GID    4
327#define CEPH_SETATTR_MTIME  8
328#define CEPH_SETATTR_ATIME 16
329#define CEPH_SETATTR_SIZE  32
330#define CEPH_SETATTR_CTIME 64
331
332union ceph_mds_request_args {
333	struct {
334		__le32 mask;                 /* CEPH_CAP_* */
335	} __attribute__ ((packed)) getattr;
336	struct {
337		__le32 mode;
338		__le32 uid;
339		__le32 gid;
340		struct ceph_timespec mtime;
341		struct ceph_timespec atime;
342		__le64 size, old_size;       /* old_size needed by truncate */
343		__le32 mask;                 /* CEPH_SETATTR_* */
344	} __attribute__ ((packed)) setattr;
345	struct {
346		__le32 frag;                 /* which dir fragment */
347		__le32 max_entries;          /* how many dentries to grab */
348		__le32 max_bytes;
349	} __attribute__ ((packed)) readdir;
350	struct {
351		__le32 mode;
352		__le32 rdev;
353	} __attribute__ ((packed)) mknod;
354	struct {
355		__le32 mode;
356	} __attribute__ ((packed)) mkdir;
357	struct {
358		__le32 flags;
359		__le32 mode;
360		__le32 stripe_unit;          /* layout for newly created file */
361		__le32 stripe_count;         /* ... */
362		__le32 object_size;
363		__le32 file_replication;
364		__le32 preferred;
365	} __attribute__ ((packed)) open;
366	struct {
367		__le32 flags;
368	} __attribute__ ((packed)) setxattr;
369	struct {
370		struct ceph_file_layout layout;
371	} __attribute__ ((packed)) setlayout;
372	struct {
373		__u8 rule; /* currently fcntl or flock */
374		__u8 type; /* shared, exclusive, remove*/
375		__le64 pid; /* process id requesting the lock */
376		__le64 pid_namespace;
377		__le64 start; /* initial location to lock */
378		__le64 length; /* num bytes to lock from start */
379		__u8 wait; /* will caller wait for lock to become available? */
380	} __attribute__ ((packed)) filelock_change;
381} __attribute__ ((packed));
382
383#define CEPH_MDS_FLAG_REPLAY        1  /* this is a replayed op */
384#define CEPH_MDS_FLAG_WANT_DENTRY   2  /* want dentry in reply */
385
386struct ceph_mds_request_head {
387	__le64 oldest_client_tid;
388	__le32 mdsmap_epoch;           /* on client */
389	__le32 flags;                  /* CEPH_MDS_FLAG_* */
390	__u8 num_retry, num_fwd;       /* count retry, fwd attempts */
391	__le16 num_releases;           /* # include cap/lease release records */
392	__le32 op;                     /* mds op code */
393	__le32 caller_uid, caller_gid;
394	__le64 ino;                    /* use this ino for openc, mkdir, mknod,
395					  etc. (if replaying) */
396	union ceph_mds_request_args args;
397} __attribute__ ((packed));
398
399/* cap/lease release record */
400struct ceph_mds_request_release {
401	__le64 ino, cap_id;            /* ino and unique cap id */
402	__le32 caps, wanted;           /* new issued, wanted */
403	__le32 seq, issue_seq, mseq;
404	__le32 dname_seq;              /* if releasing a dentry lease, a */
405	__le32 dname_len;              /* string follows. */
406} __attribute__ ((packed));
407
408/* client reply */
409struct ceph_mds_reply_head {
410	__le32 op;
411	__le32 result;
412	__le32 mdsmap_epoch;
413	__u8 safe;                     /* true if committed to disk */
414	__u8 is_dentry, is_target;     /* true if dentry, target inode records
415					  are included with reply */
416} __attribute__ ((packed));
417
418/* one for each node split */
419struct ceph_frag_tree_split {
420	__le32 frag;                   /* this frag splits... */
421	__le32 by;                     /* ...by this many bits */
422} __attribute__ ((packed));
423
424struct ceph_frag_tree_head {
425	__le32 nsplits;                /* num ceph_frag_tree_split records */
426	struct ceph_frag_tree_split splits[];
427} __attribute__ ((packed));
428
429/* capability issue, for bundling with mds reply */
430struct ceph_mds_reply_cap {
431	__le32 caps, wanted;           /* caps issued, wanted */
432	__le64 cap_id;
433	__le32 seq, mseq;
434	__le64 realm;                  /* snap realm */
435	__u8 flags;                    /* CEPH_CAP_FLAG_* */
436} __attribute__ ((packed));
437
438#define CEPH_CAP_FLAG_AUTH  1          /* cap is issued by auth mds */
439
440/* inode record, for bundling with mds reply */
441struct ceph_mds_reply_inode {
442	__le64 ino;
443	__le64 snapid;
444	__le32 rdev;
445	__le64 version;                /* inode version */
446	__le64 xattr_version;          /* version for xattr blob */
447	struct ceph_mds_reply_cap cap; /* caps issued for this inode */
448	struct ceph_file_layout layout;
449	struct ceph_timespec ctime, mtime, atime;
450	__le32 time_warp_seq;
451	__le64 size, max_size, truncate_size;
452	__le32 truncate_seq;
453	__le32 mode, uid, gid;
454	__le32 nlink;
455	__le64 files, subdirs, rbytes, rfiles, rsubdirs;  /* dir stats */
456	struct ceph_timespec rctime;
457	struct ceph_frag_tree_head fragtree;  /* (must be at end of struct) */
458} __attribute__ ((packed));
459/* followed by frag array, then symlink string, then xattr blob */
460
461/* reply_lease follows dname, and reply_inode */
462struct ceph_mds_reply_lease {
463	__le16 mask;            /* lease type(s) */
464	__le32 duration_ms;     /* lease duration */
465	__le32 seq;
466} __attribute__ ((packed));
467
468struct ceph_mds_reply_dirfrag {
469	__le32 frag;            /* fragment */
470	__le32 auth;            /* auth mds, if this is a delegation point */
471	__le32 ndist;           /* number of mds' this is replicated on */
472	__le32 dist[];
473} __attribute__ ((packed));
474
475#define CEPH_LOCK_FCNTL    1
476#define CEPH_LOCK_FLOCK    2
477
478#define CEPH_LOCK_SHARED   1
479#define CEPH_LOCK_EXCL     2
480#define CEPH_LOCK_UNLOCK   4
481
482struct ceph_filelock {
483	__le64 start;/* file offset to start lock at */
484	__le64 length; /* num bytes to lock; 0 for all following start */
485	__le64 client; /* which client holds the lock */
486	__le64 pid; /* process id holding the lock on the client */
487	__le64 pid_namespace;
488	__u8 type; /* shared lock, exclusive lock, or unlock */
489} __attribute__ ((packed));
490
491
492/* file access modes */
493#define CEPH_FILE_MODE_PIN        0
494#define CEPH_FILE_MODE_RD         1
495#define CEPH_FILE_MODE_WR         2
496#define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
497#define CEPH_FILE_MODE_LAZY       4  /* lazy io */
498#define CEPH_FILE_MODE_NUM        8  /* bc these are bit fields.. mostly */
499
500int ceph_flags_to_mode(int flags);
501
502
503/* capability bits */
504#define CEPH_CAP_PIN         1  /* no specific capabilities beyond the pin */
505
506/* generic cap bits */
507#define CEPH_CAP_GSHARED     1  /* client can reads */
508#define CEPH_CAP_GEXCL       2  /* client can read and update */
509#define CEPH_CAP_GCACHE      4  /* (file) client can cache reads */
510#define CEPH_CAP_GRD         8  /* (file) client can read */
511#define CEPH_CAP_GWR        16  /* (file) client can write */
512#define CEPH_CAP_GBUFFER    32  /* (file) client can buffer writes */
513#define CEPH_CAP_GWREXTEND  64  /* (file) client can extend EOF */
514#define CEPH_CAP_GLAZYIO   128  /* (file) client can perform lazy io */
515
516/* per-lock shift */
517#define CEPH_CAP_SAUTH      2
518#define CEPH_CAP_SLINK      4
519#define CEPH_CAP_SXATTR     6
520#define CEPH_CAP_SFILE      8
521#define CEPH_CAP_SFLOCK    20
522
523#define CEPH_CAP_BITS       22
524
525/* composed values */
526#define CEPH_CAP_AUTH_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SAUTH)
527#define CEPH_CAP_AUTH_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SAUTH)
528#define CEPH_CAP_LINK_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SLINK)
529#define CEPH_CAP_LINK_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SLINK)
530#define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED  << CEPH_CAP_SXATTR)
531#define CEPH_CAP_XATTR_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SXATTR)
532#define CEPH_CAP_FILE(x)    (x << CEPH_CAP_SFILE)
533#define CEPH_CAP_FILE_SHARED   (CEPH_CAP_GSHARED   << CEPH_CAP_SFILE)
534#define CEPH_CAP_FILE_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SFILE)
535#define CEPH_CAP_FILE_CACHE    (CEPH_CAP_GCACHE    << CEPH_CAP_SFILE)
536#define CEPH_CAP_FILE_RD       (CEPH_CAP_GRD       << CEPH_CAP_SFILE)
537#define CEPH_CAP_FILE_WR       (CEPH_CAP_GWR       << CEPH_CAP_SFILE)
538#define CEPH_CAP_FILE_BUFFER   (CEPH_CAP_GBUFFER   << CEPH_CAP_SFILE)
539#define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
540#define CEPH_CAP_FILE_LAZYIO   (CEPH_CAP_GLAZYIO   << CEPH_CAP_SFILE)
541#define CEPH_CAP_FLOCK_SHARED  (CEPH_CAP_GSHARED   << CEPH_CAP_SFLOCK)
542#define CEPH_CAP_FLOCK_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SFLOCK)
543
544
545/* cap masks (for getattr) */
546#define CEPH_STAT_CAP_INODE    CEPH_CAP_PIN
547#define CEPH_STAT_CAP_TYPE     CEPH_CAP_PIN  /* mode >> 12 */
548#define CEPH_STAT_CAP_SYMLINK  CEPH_CAP_PIN
549#define CEPH_STAT_CAP_UID      CEPH_CAP_AUTH_SHARED
550#define CEPH_STAT_CAP_GID      CEPH_CAP_AUTH_SHARED
551#define CEPH_STAT_CAP_MODE     CEPH_CAP_AUTH_SHARED
552#define CEPH_STAT_CAP_NLINK    CEPH_CAP_LINK_SHARED
553#define CEPH_STAT_CAP_LAYOUT   CEPH_CAP_FILE_SHARED
554#define CEPH_STAT_CAP_MTIME    CEPH_CAP_FILE_SHARED
555#define CEPH_STAT_CAP_SIZE     CEPH_CAP_FILE_SHARED
556#define CEPH_STAT_CAP_ATIME    CEPH_CAP_FILE_SHARED
557#define CEPH_STAT_CAP_XATTR    CEPH_CAP_XATTR_SHARED
558#define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN |			\
559				 CEPH_CAP_AUTH_SHARED |	\
560				 CEPH_CAP_LINK_SHARED |	\
561				 CEPH_CAP_FILE_SHARED |	\
562				 CEPH_CAP_XATTR_SHARED)
563
564#define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED |			\
565			      CEPH_CAP_LINK_SHARED |			\
566			      CEPH_CAP_XATTR_SHARED |			\
567			      CEPH_CAP_FILE_SHARED)
568#define CEPH_CAP_ANY_RD   (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD |	\
569			   CEPH_CAP_FILE_CACHE)
570
571#define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL |		\
572			   CEPH_CAP_LINK_EXCL |		\
573			   CEPH_CAP_XATTR_EXCL |	\
574			   CEPH_CAP_FILE_EXCL)
575#define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER |	\
576			      CEPH_CAP_FILE_EXCL)
577#define CEPH_CAP_ANY_WR   (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
578#define CEPH_CAP_ANY      (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
579			   CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
580			   CEPH_CAP_PIN)
581
582#define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
583			CEPH_LOCK_IXATTR)
584
585int ceph_caps_for_mode(int mode);
586
587enum {
588	CEPH_CAP_OP_GRANT,         /* mds->client grant */
589	CEPH_CAP_OP_REVOKE,        /* mds->client revoke */
590	CEPH_CAP_OP_TRUNC,         /* mds->client trunc notify */
591	CEPH_CAP_OP_EXPORT,        /* mds has exported the cap */
592	CEPH_CAP_OP_IMPORT,        /* mds has imported the cap */
593	CEPH_CAP_OP_UPDATE,        /* client->mds update */
594	CEPH_CAP_OP_DROP,          /* client->mds drop cap bits */
595	CEPH_CAP_OP_FLUSH,         /* client->mds cap writeback */
596	CEPH_CAP_OP_FLUSH_ACK,     /* mds->client flushed */
597	CEPH_CAP_OP_FLUSHSNAP,     /* client->mds flush snapped metadata */
598	CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
599	CEPH_CAP_OP_RELEASE,       /* client->mds release (clean) cap */
600	CEPH_CAP_OP_RENEW,         /* client->mds renewal request */
601};
602
603extern const char *ceph_cap_op_name(int op);
604
605/*
606 * caps message, used for capability callbacks, acks, requests, etc.
607 */
608struct ceph_mds_caps {
609	__le32 op;                  /* CEPH_CAP_OP_* */
610	__le64 ino, realm;
611	__le64 cap_id;
612	__le32 seq, issue_seq;
613	__le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
614	__le32 migrate_seq;
615	__le64 snap_follows;
616	__le32 snap_trace_len;
617
618	/* authlock */
619	__le32 uid, gid, mode;
620
621	/* linklock */
622	__le32 nlink;
623
624	/* xattrlock */
625	__le32 xattr_len;
626	__le64 xattr_version;
627
628	/* filelock */
629	__le64 size, max_size, truncate_size;
630	__le32 truncate_seq;
631	struct ceph_timespec mtime, atime, ctime;
632	struct ceph_file_layout layout;
633	__le32 time_warp_seq;
634} __attribute__ ((packed));
635
636/* cap release msg head */
637struct ceph_mds_cap_release {
638	__le32 num;                /* number of cap_items that follow */
639} __attribute__ ((packed));
640
641struct ceph_mds_cap_item {
642	__le64 ino;
643	__le64 cap_id;
644	__le32 migrate_seq, seq;
645} __attribute__ ((packed));
646
647#define CEPH_MDS_LEASE_REVOKE           1  /*    mds  -> client */
648#define CEPH_MDS_LEASE_RELEASE          2  /* client  -> mds    */
649#define CEPH_MDS_LEASE_RENEW            3  /* client <-> mds    */
650#define CEPH_MDS_LEASE_REVOKE_ACK       4  /* client  -> mds    */
651
652extern const char *ceph_lease_op_name(int o);
653
654/* lease msg header */
655struct ceph_mds_lease {
656	__u8 action;            /* CEPH_MDS_LEASE_* */
657	__le16 mask;            /* which lease */
658	__le64 ino;
659	__le64 first, last;     /* snap range */
660	__le32 seq;
661	__le32 duration_ms;     /* duration of renewal */
662} __attribute__ ((packed));
663/* followed by a __le32+string for dname */
664
665/* client reconnect */
666struct ceph_mds_cap_reconnect {
667	__le64 cap_id;
668	__le32 wanted;
669	__le32 issued;
670	__le64 snaprealm;
671	__le64 pathbase;        /* base ino for our path to this ino */
672	__le32 flock_len;       /* size of flock state blob, if any */
673} __attribute__ ((packed));
674/* followed by flock blob */
675
676struct ceph_mds_cap_reconnect_v1 {
677	__le64 cap_id;
678	__le32 wanted;
679	__le32 issued;
680	__le64 size;
681	struct ceph_timespec mtime, atime;
682	__le64 snaprealm;
683	__le64 pathbase;        /* base ino for our path to this ino */
684} __attribute__ ((packed));
685
686struct ceph_mds_snaprealm_reconnect {
687	__le64 ino;     /* snap realm base */
688	__le64 seq;     /* snap seq for this snap realm */
689	__le64 parent;  /* parent realm */
690} __attribute__ ((packed));
691
692/*
693 * snaps
694 */
695enum {
696	CEPH_SNAP_OP_UPDATE,  /* CREATE or DESTROY */
697	CEPH_SNAP_OP_CREATE,
698	CEPH_SNAP_OP_DESTROY,
699	CEPH_SNAP_OP_SPLIT,
700};
701
702extern const char *ceph_snap_op_name(int o);
703
704/* snap msg header */
705struct ceph_mds_snap_head {
706	__le32 op;                /* CEPH_SNAP_OP_* */
707	__le64 split;             /* ino to split off, if any */
708	__le32 num_split_inos;    /* # inos belonging to new child realm */
709	__le32 num_split_realms;  /* # child realms udner new child realm */
710	__le32 trace_len;         /* size of snap trace blob */
711} __attribute__ ((packed));
712/* followed by split ino list, then split realms, then the trace blob */
713
714/*
715 * encode info about a snaprealm, as viewed by a client
716 */
717struct ceph_mds_snap_realm {
718	__le64 ino;           /* ino */
719	__le64 created;       /* snap: when created */
720	__le64 parent;        /* ino: parent realm */
721	__le64 parent_since;  /* snap: same parent since */
722	__le64 seq;           /* snap: version */
723	__le32 num_snaps;
724	__le32 num_prior_parent_snaps;
725} __attribute__ ((packed));
726/* followed by my snap list, then prior parent snap list */
727
728#endif
729