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 */
39struct 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
66static 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 */
73static 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
81int ceph_file_layout_is_valid(const struct ceph_file_layout *layout);
82
83struct 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 */
155enum {
156  WATCH_NOTIFY				= 1, /* notifying watcher */
157  WATCH_NOTIFY_COMPLETE			= 2, /* notifier notified when done */
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_osd_getmap {
184	struct ceph_mon_request_header monhdr;
185	struct ceph_fsid fsid;
186	__le32 start;
187} __attribute__ ((packed));
188
189struct ceph_mds_getmap {
190	struct ceph_mon_request_header monhdr;
191	struct ceph_fsid fsid;
192} __attribute__ ((packed));
193
194struct ceph_client_mount {
195	struct ceph_mon_request_header monhdr;
196} __attribute__ ((packed));
197
198#define CEPH_SUBSCRIBE_ONETIME    1  /* i want only 1 update after have */
199
200struct ceph_mon_subscribe_item {
201	__le64 have_version;    __le64 have;
202	__u8 onetime;
203} __attribute__ ((packed));
204
205struct ceph_mon_subscribe_ack {
206	__le32 duration;         /* seconds */
207	struct ceph_fsid fsid;
208} __attribute__ ((packed));
209
210/*
211 * mdsmap flags
212 */
213#define CEPH_MDSMAP_DOWN    (1<<0)  /* cluster deliberately down */
214
215/*
216 * mds states
217 *   > 0 -> in
218 *  <= 0 -> out
219 */
220#define CEPH_MDS_STATE_DNE          0  /* down, does not exist. */
221#define CEPH_MDS_STATE_STOPPED     -1  /* down, once existed, but no subtrees.
222					  empty log. */
223#define CEPH_MDS_STATE_BOOT        -4  /* up, boot announcement. */
224#define CEPH_MDS_STATE_STANDBY     -5  /* up, idle.  waiting for assignment. */
225#define CEPH_MDS_STATE_CREATING    -6  /* up, creating MDS instance. */
226#define CEPH_MDS_STATE_STARTING    -7  /* up, starting previously stopped mds */
227#define CEPH_MDS_STATE_STANDBY_REPLAY -8 /* up, tailing active node's journal */
228#define CEPH_MDS_STATE_REPLAYONCE   -9 /* up, replaying an 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#define CEPH_LOCK_IPOLICY     16384 /* policy lock on dirs. MDS internal */
261
262/* client_session ops */
263enum {
264	CEPH_SESSION_REQUEST_OPEN,
265	CEPH_SESSION_OPEN,
266	CEPH_SESSION_REQUEST_CLOSE,
267	CEPH_SESSION_CLOSE,
268	CEPH_SESSION_REQUEST_RENEWCAPS,
269	CEPH_SESSION_RENEWCAPS,
270	CEPH_SESSION_STALE,
271	CEPH_SESSION_RECALL_STATE,
272	CEPH_SESSION_FLUSHMSG,
273	CEPH_SESSION_FLUSHMSG_ACK,
274	CEPH_SESSION_FORCE_RO,
275};
276
277extern const char *ceph_session_op_name(int op);
278
279struct ceph_mds_session_head {
280	__le32 op;
281	__le64 seq;
282	struct ceph_timespec stamp;
283	__le32 max_caps, max_leases;
284} __attribute__ ((packed));
285
286/* client_request */
287/*
288 * metadata ops.
289 *  & 0x001000 -> write op
290 *  & 0x010000 -> follow symlink (e.g. stat(), not lstat()).
291 &  & 0x100000 -> use weird ino/path trace
292 */
293#define CEPH_MDS_OP_WRITE        0x001000
294enum {
295	CEPH_MDS_OP_LOOKUP     = 0x00100,
296	CEPH_MDS_OP_GETATTR    = 0x00101,
297	CEPH_MDS_OP_LOOKUPHASH = 0x00102,
298	CEPH_MDS_OP_LOOKUPPARENT = 0x00103,
299	CEPH_MDS_OP_LOOKUPINO  = 0x00104,
300	CEPH_MDS_OP_LOOKUPNAME = 0x00105,
301
302	CEPH_MDS_OP_SETXATTR   = 0x01105,
303	CEPH_MDS_OP_RMXATTR    = 0x01106,
304	CEPH_MDS_OP_SETLAYOUT  = 0x01107,
305	CEPH_MDS_OP_SETATTR    = 0x01108,
306	CEPH_MDS_OP_SETFILELOCK= 0x01109,
307	CEPH_MDS_OP_GETFILELOCK= 0x00110,
308	CEPH_MDS_OP_SETDIRLAYOUT=0x0110a,
309
310	CEPH_MDS_OP_MKNOD      = 0x01201,
311	CEPH_MDS_OP_LINK       = 0x01202,
312	CEPH_MDS_OP_UNLINK     = 0x01203,
313	CEPH_MDS_OP_RENAME     = 0x01204,
314	CEPH_MDS_OP_MKDIR      = 0x01220,
315	CEPH_MDS_OP_RMDIR      = 0x01221,
316	CEPH_MDS_OP_SYMLINK    = 0x01222,
317
318	CEPH_MDS_OP_CREATE     = 0x01301,
319	CEPH_MDS_OP_OPEN       = 0x00302,
320	CEPH_MDS_OP_READDIR    = 0x00305,
321
322	CEPH_MDS_OP_LOOKUPSNAP = 0x00400,
323	CEPH_MDS_OP_MKSNAP     = 0x01400,
324	CEPH_MDS_OP_RMSNAP     = 0x01401,
325	CEPH_MDS_OP_LSSNAP     = 0x00402,
326	CEPH_MDS_OP_RENAMESNAP = 0x01403,
327};
328
329extern const char *ceph_mds_op_name(int op);
330
331
332#define CEPH_SETATTR_MODE   1
333#define CEPH_SETATTR_UID    2
334#define CEPH_SETATTR_GID    4
335#define CEPH_SETATTR_MTIME  8
336#define CEPH_SETATTR_ATIME 16
337#define CEPH_SETATTR_SIZE  32
338#define CEPH_SETATTR_CTIME 64
339
340/*
341 * Ceph setxattr request flags.
342 */
343#define CEPH_XATTR_CREATE  (1 << 0)
344#define CEPH_XATTR_REPLACE (1 << 1)
345#define CEPH_XATTR_REMOVE  (1 << 31)
346
347union ceph_mds_request_args {
348	struct {
349		__le32 mask;                 /* CEPH_CAP_* */
350	} __attribute__ ((packed)) getattr;
351	struct {
352		__le32 mode;
353		__le32 uid;
354		__le32 gid;
355		struct ceph_timespec mtime;
356		struct ceph_timespec atime;
357		__le64 size, old_size;       /* old_size needed by truncate */
358		__le32 mask;                 /* CEPH_SETATTR_* */
359	} __attribute__ ((packed)) setattr;
360	struct {
361		__le32 frag;                 /* which dir fragment */
362		__le32 max_entries;          /* how many dentries to grab */
363		__le32 max_bytes;
364	} __attribute__ ((packed)) readdir;
365	struct {
366		__le32 mode;
367		__le32 rdev;
368	} __attribute__ ((packed)) mknod;
369	struct {
370		__le32 mode;
371	} __attribute__ ((packed)) mkdir;
372	struct {
373		__le32 flags;
374		__le32 mode;
375		__le32 stripe_unit;          /* layout for newly created file */
376		__le32 stripe_count;         /* ... */
377		__le32 object_size;
378		__le32 file_replication;
379		__le32 unused;               /* used to be preferred osd */
380	} __attribute__ ((packed)) open;
381	struct {
382		__le32 flags;
383	} __attribute__ ((packed)) setxattr;
384	struct {
385		struct ceph_file_layout layout;
386	} __attribute__ ((packed)) setlayout;
387	struct {
388		__u8 rule; /* currently fcntl or flock */
389		__u8 type; /* shared, exclusive, remove*/
390		__le64 owner; /* owner of the lock */
391		__le64 pid; /* process id requesting the lock */
392		__le64 start; /* initial location to lock */
393		__le64 length; /* num bytes to lock from start */
394		__u8 wait; /* will caller wait for lock to become available? */
395	} __attribute__ ((packed)) filelock_change;
396} __attribute__ ((packed));
397
398#define CEPH_MDS_FLAG_REPLAY        1  /* this is a replayed op */
399#define CEPH_MDS_FLAG_WANT_DENTRY   2  /* want dentry in reply */
400
401struct ceph_mds_request_head {
402	__le64 oldest_client_tid;
403	__le32 mdsmap_epoch;           /* on client */
404	__le32 flags;                  /* CEPH_MDS_FLAG_* */
405	__u8 num_retry, num_fwd;       /* count retry, fwd attempts */
406	__le16 num_releases;           /* # include cap/lease release records */
407	__le32 op;                     /* mds op code */
408	__le32 caller_uid, caller_gid;
409	__le64 ino;                    /* use this ino for openc, mkdir, mknod,
410					  etc. (if replaying) */
411	union ceph_mds_request_args args;
412} __attribute__ ((packed));
413
414/* cap/lease release record */
415struct ceph_mds_request_release {
416	__le64 ino, cap_id;            /* ino and unique cap id */
417	__le32 caps, wanted;           /* new issued, wanted */
418	__le32 seq, issue_seq, mseq;
419	__le32 dname_seq;              /* if releasing a dentry lease, a */
420	__le32 dname_len;              /* string follows. */
421} __attribute__ ((packed));
422
423/* client reply */
424struct ceph_mds_reply_head {
425	__le32 op;
426	__le32 result;
427	__le32 mdsmap_epoch;
428	__u8 safe;                     /* true if committed to disk */
429	__u8 is_dentry, is_target;     /* true if dentry, target inode records
430					  are included with reply */
431} __attribute__ ((packed));
432
433/* one for each node split */
434struct ceph_frag_tree_split {
435	__le32 frag;                   /* this frag splits... */
436	__le32 by;                     /* ...by this many bits */
437} __attribute__ ((packed));
438
439struct ceph_frag_tree_head {
440	__le32 nsplits;                /* num ceph_frag_tree_split records */
441	struct ceph_frag_tree_split splits[];
442} __attribute__ ((packed));
443
444/* capability issue, for bundling with mds reply */
445struct ceph_mds_reply_cap {
446	__le32 caps, wanted;           /* caps issued, wanted */
447	__le64 cap_id;
448	__le32 seq, mseq;
449	__le64 realm;                  /* snap realm */
450	__u8 flags;                    /* CEPH_CAP_FLAG_* */
451} __attribute__ ((packed));
452
453#define CEPH_CAP_FLAG_AUTH	(1 << 0)  /* cap is issued by auth mds */
454#define CEPH_CAP_FLAG_RELEASE	(1 << 1)  /* release the cap */
455
456/* inode record, for bundling with mds reply */
457struct ceph_mds_reply_inode {
458	__le64 ino;
459	__le64 snapid;
460	__le32 rdev;
461	__le64 version;                /* inode version */
462	__le64 xattr_version;          /* version for xattr blob */
463	struct ceph_mds_reply_cap cap; /* caps issued for this inode */
464	struct ceph_file_layout layout;
465	struct ceph_timespec ctime, mtime, atime;
466	__le32 time_warp_seq;
467	__le64 size, max_size, truncate_size;
468	__le32 truncate_seq;
469	__le32 mode, uid, gid;
470	__le32 nlink;
471	__le64 files, subdirs, rbytes, rfiles, rsubdirs;  /* dir stats */
472	struct ceph_timespec rctime;
473	struct ceph_frag_tree_head fragtree;  /* (must be at end of struct) */
474} __attribute__ ((packed));
475/* followed by frag array, symlink string, dir layout, xattr blob */
476
477/* reply_lease follows dname, and reply_inode */
478struct ceph_mds_reply_lease {
479	__le16 mask;            /* lease type(s) */
480	__le32 duration_ms;     /* lease duration */
481	__le32 seq;
482} __attribute__ ((packed));
483
484struct ceph_mds_reply_dirfrag {
485	__le32 frag;            /* fragment */
486	__le32 auth;            /* auth mds, if this is a delegation point */
487	__le32 ndist;           /* number of mds' this is replicated on */
488	__le32 dist[];
489} __attribute__ ((packed));
490
491#define CEPH_LOCK_FCNTL		1
492#define CEPH_LOCK_FLOCK		2
493#define CEPH_LOCK_FCNTL_INTR    3
494#define CEPH_LOCK_FLOCK_INTR    4
495
496
497#define CEPH_LOCK_SHARED   1
498#define CEPH_LOCK_EXCL     2
499#define CEPH_LOCK_UNLOCK   4
500
501struct ceph_filelock {
502	__le64 start;/* file offset to start lock at */
503	__le64 length; /* num bytes to lock; 0 for all following start */
504	__le64 client; /* which client holds the lock */
505	__le64 owner; /* owner the lock */
506	__le64 pid; /* process id holding the lock on the client */
507	__u8 type; /* shared lock, exclusive lock, or unlock */
508} __attribute__ ((packed));
509
510
511/* file access modes */
512#define CEPH_FILE_MODE_PIN        0
513#define CEPH_FILE_MODE_RD         1
514#define CEPH_FILE_MODE_WR         2
515#define CEPH_FILE_MODE_RDWR       3  /* RD | WR */
516#define CEPH_FILE_MODE_LAZY       4  /* lazy io */
517#define CEPH_FILE_MODE_NUM        8  /* bc these are bit fields.. mostly */
518
519int ceph_flags_to_mode(int flags);
520
521#define CEPH_INLINE_NONE	((__u64)-1)
522
523/* capability bits */
524#define CEPH_CAP_PIN         1  /* no specific capabilities beyond the pin */
525
526/* generic cap bits */
527#define CEPH_CAP_GSHARED     1  /* client can reads */
528#define CEPH_CAP_GEXCL       2  /* client can read and update */
529#define CEPH_CAP_GCACHE      4  /* (file) client can cache reads */
530#define CEPH_CAP_GRD         8  /* (file) client can read */
531#define CEPH_CAP_GWR        16  /* (file) client can write */
532#define CEPH_CAP_GBUFFER    32  /* (file) client can buffer writes */
533#define CEPH_CAP_GWREXTEND  64  /* (file) client can extend EOF */
534#define CEPH_CAP_GLAZYIO   128  /* (file) client can perform lazy io */
535
536#define CEPH_CAP_SIMPLE_BITS  2
537#define CEPH_CAP_FILE_BITS    8
538
539/* per-lock shift */
540#define CEPH_CAP_SAUTH      2
541#define CEPH_CAP_SLINK      4
542#define CEPH_CAP_SXATTR     6
543#define CEPH_CAP_SFILE      8
544#define CEPH_CAP_SFLOCK    20
545
546#define CEPH_CAP_BITS      22
547
548/* composed values */
549#define CEPH_CAP_AUTH_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SAUTH)
550#define CEPH_CAP_AUTH_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SAUTH)
551#define CEPH_CAP_LINK_SHARED  (CEPH_CAP_GSHARED  << CEPH_CAP_SLINK)
552#define CEPH_CAP_LINK_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SLINK)
553#define CEPH_CAP_XATTR_SHARED (CEPH_CAP_GSHARED  << CEPH_CAP_SXATTR)
554#define CEPH_CAP_XATTR_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SXATTR)
555#define CEPH_CAP_FILE(x)    (x << CEPH_CAP_SFILE)
556#define CEPH_CAP_FILE_SHARED   (CEPH_CAP_GSHARED   << CEPH_CAP_SFILE)
557#define CEPH_CAP_FILE_EXCL     (CEPH_CAP_GEXCL     << CEPH_CAP_SFILE)
558#define CEPH_CAP_FILE_CACHE    (CEPH_CAP_GCACHE    << CEPH_CAP_SFILE)
559#define CEPH_CAP_FILE_RD       (CEPH_CAP_GRD       << CEPH_CAP_SFILE)
560#define CEPH_CAP_FILE_WR       (CEPH_CAP_GWR       << CEPH_CAP_SFILE)
561#define CEPH_CAP_FILE_BUFFER   (CEPH_CAP_GBUFFER   << CEPH_CAP_SFILE)
562#define CEPH_CAP_FILE_WREXTEND (CEPH_CAP_GWREXTEND << CEPH_CAP_SFILE)
563#define CEPH_CAP_FILE_LAZYIO   (CEPH_CAP_GLAZYIO   << CEPH_CAP_SFILE)
564#define CEPH_CAP_FLOCK_SHARED  (CEPH_CAP_GSHARED   << CEPH_CAP_SFLOCK)
565#define CEPH_CAP_FLOCK_EXCL    (CEPH_CAP_GEXCL     << CEPH_CAP_SFLOCK)
566
567
568/* cap masks (for getattr) */
569#define CEPH_STAT_CAP_INODE    CEPH_CAP_PIN
570#define CEPH_STAT_CAP_TYPE     CEPH_CAP_PIN  /* mode >> 12 */
571#define CEPH_STAT_CAP_SYMLINK  CEPH_CAP_PIN
572#define CEPH_STAT_CAP_UID      CEPH_CAP_AUTH_SHARED
573#define CEPH_STAT_CAP_GID      CEPH_CAP_AUTH_SHARED
574#define CEPH_STAT_CAP_MODE     CEPH_CAP_AUTH_SHARED
575#define CEPH_STAT_CAP_NLINK    CEPH_CAP_LINK_SHARED
576#define CEPH_STAT_CAP_LAYOUT   CEPH_CAP_FILE_SHARED
577#define CEPH_STAT_CAP_MTIME    CEPH_CAP_FILE_SHARED
578#define CEPH_STAT_CAP_SIZE     CEPH_CAP_FILE_SHARED
579#define CEPH_STAT_CAP_ATIME    CEPH_CAP_FILE_SHARED  /* fixme */
580#define CEPH_STAT_CAP_XATTR    CEPH_CAP_XATTR_SHARED
581#define CEPH_STAT_CAP_INODE_ALL (CEPH_CAP_PIN |			\
582				 CEPH_CAP_AUTH_SHARED |	\
583				 CEPH_CAP_LINK_SHARED |	\
584				 CEPH_CAP_FILE_SHARED |	\
585				 CEPH_CAP_XATTR_SHARED)
586#define CEPH_STAT_CAP_INLINE_DATA (CEPH_CAP_FILE_SHARED | \
587				   CEPH_CAP_FILE_RD)
588
589#define CEPH_CAP_ANY_SHARED (CEPH_CAP_AUTH_SHARED |			\
590			      CEPH_CAP_LINK_SHARED |			\
591			      CEPH_CAP_XATTR_SHARED |			\
592			      CEPH_CAP_FILE_SHARED)
593#define CEPH_CAP_ANY_RD   (CEPH_CAP_ANY_SHARED | CEPH_CAP_FILE_RD |	\
594			   CEPH_CAP_FILE_CACHE)
595
596#define CEPH_CAP_ANY_EXCL (CEPH_CAP_AUTH_EXCL |		\
597			   CEPH_CAP_LINK_EXCL |		\
598			   CEPH_CAP_XATTR_EXCL |	\
599			   CEPH_CAP_FILE_EXCL)
600#define CEPH_CAP_ANY_FILE_RD (CEPH_CAP_FILE_RD | CEPH_CAP_FILE_CACHE | \
601			      CEPH_CAP_FILE_SHARED)
602#define CEPH_CAP_ANY_FILE_WR (CEPH_CAP_FILE_WR | CEPH_CAP_FILE_BUFFER |	\
603			      CEPH_CAP_FILE_EXCL)
604#define CEPH_CAP_ANY_WR   (CEPH_CAP_ANY_EXCL | CEPH_CAP_ANY_FILE_WR)
605#define CEPH_CAP_ANY      (CEPH_CAP_ANY_RD | CEPH_CAP_ANY_EXCL | \
606			   CEPH_CAP_ANY_FILE_WR | CEPH_CAP_FILE_LAZYIO | \
607			   CEPH_CAP_PIN)
608
609#define CEPH_CAP_LOCKS (CEPH_LOCK_IFILE | CEPH_LOCK_IAUTH | CEPH_LOCK_ILINK | \
610			CEPH_LOCK_IXATTR)
611
612int ceph_caps_for_mode(int mode);
613
614enum {
615	CEPH_CAP_OP_GRANT,         /* mds->client grant */
616	CEPH_CAP_OP_REVOKE,        /* mds->client revoke */
617	CEPH_CAP_OP_TRUNC,         /* mds->client trunc notify */
618	CEPH_CAP_OP_EXPORT,        /* mds has exported the cap */
619	CEPH_CAP_OP_IMPORT,        /* mds has imported the cap */
620	CEPH_CAP_OP_UPDATE,        /* client->mds update */
621	CEPH_CAP_OP_DROP,          /* client->mds drop cap bits */
622	CEPH_CAP_OP_FLUSH,         /* client->mds cap writeback */
623	CEPH_CAP_OP_FLUSH_ACK,     /* mds->client flushed */
624	CEPH_CAP_OP_FLUSHSNAP,     /* client->mds flush snapped metadata */
625	CEPH_CAP_OP_FLUSHSNAP_ACK, /* mds->client flushed snapped metadata */
626	CEPH_CAP_OP_RELEASE,       /* client->mds release (clean) cap */
627	CEPH_CAP_OP_RENEW,         /* client->mds renewal request */
628};
629
630extern const char *ceph_cap_op_name(int op);
631
632/*
633 * caps message, used for capability callbacks, acks, requests, etc.
634 */
635struct ceph_mds_caps {
636	__le32 op;                  /* CEPH_CAP_OP_* */
637	__le64 ino, realm;
638	__le64 cap_id;
639	__le32 seq, issue_seq;
640	__le32 caps, wanted, dirty; /* latest issued/wanted/dirty */
641	__le32 migrate_seq;
642	__le64 snap_follows;
643	__le32 snap_trace_len;
644
645	/* authlock */
646	__le32 uid, gid, mode;
647
648	/* linklock */
649	__le32 nlink;
650
651	/* xattrlock */
652	__le32 xattr_len;
653	__le64 xattr_version;
654
655	/* filelock */
656	__le64 size, max_size, truncate_size;
657	__le32 truncate_seq;
658	struct ceph_timespec mtime, atime, ctime;
659	struct ceph_file_layout layout;
660	__le32 time_warp_seq;
661} __attribute__ ((packed));
662
663struct ceph_mds_cap_peer {
664	__le64 cap_id;
665	__le32 seq;
666	__le32 mseq;
667	__le32 mds;
668	__u8   flags;
669} __attribute__ ((packed));
670
671/* cap release msg head */
672struct ceph_mds_cap_release {
673	__le32 num;                /* number of cap_items that follow */
674} __attribute__ ((packed));
675
676struct ceph_mds_cap_item {
677	__le64 ino;
678	__le64 cap_id;
679	__le32 migrate_seq, seq;
680} __attribute__ ((packed));
681
682#define CEPH_MDS_LEASE_REVOKE           1  /*    mds  -> client */
683#define CEPH_MDS_LEASE_RELEASE          2  /* client  -> mds    */
684#define CEPH_MDS_LEASE_RENEW            3  /* client <-> mds    */
685#define CEPH_MDS_LEASE_REVOKE_ACK       4  /* client  -> mds    */
686
687extern const char *ceph_lease_op_name(int o);
688
689/* lease msg header */
690struct ceph_mds_lease {
691	__u8 action;            /* CEPH_MDS_LEASE_* */
692	__le16 mask;            /* which lease */
693	__le64 ino;
694	__le64 first, last;     /* snap range */
695	__le32 seq;
696	__le32 duration_ms;     /* duration of renewal */
697} __attribute__ ((packed));
698/* followed by a __le32+string for dname */
699
700/* client reconnect */
701struct ceph_mds_cap_reconnect {
702	__le64 cap_id;
703	__le32 wanted;
704	__le32 issued;
705	__le64 snaprealm;
706	__le64 pathbase;        /* base ino for our path to this ino */
707	__le32 flock_len;       /* size of flock state blob, if any */
708} __attribute__ ((packed));
709/* followed by flock blob */
710
711struct ceph_mds_cap_reconnect_v1 {
712	__le64 cap_id;
713	__le32 wanted;
714	__le32 issued;
715	__le64 size;
716	struct ceph_timespec mtime, atime;
717	__le64 snaprealm;
718	__le64 pathbase;        /* base ino for our path to this ino */
719} __attribute__ ((packed));
720
721struct ceph_mds_snaprealm_reconnect {
722	__le64 ino;     /* snap realm base */
723	__le64 seq;     /* snap seq for this snap realm */
724	__le64 parent;  /* parent realm */
725} __attribute__ ((packed));
726
727/*
728 * snaps
729 */
730enum {
731	CEPH_SNAP_OP_UPDATE,  /* CREATE or DESTROY */
732	CEPH_SNAP_OP_CREATE,
733	CEPH_SNAP_OP_DESTROY,
734	CEPH_SNAP_OP_SPLIT,
735};
736
737extern const char *ceph_snap_op_name(int o);
738
739/* snap msg header */
740struct ceph_mds_snap_head {
741	__le32 op;                /* CEPH_SNAP_OP_* */
742	__le64 split;             /* ino to split off, if any */
743	__le32 num_split_inos;    /* # inos belonging to new child realm */
744	__le32 num_split_realms;  /* # child realms udner new child realm */
745	__le32 trace_len;         /* size of snap trace blob */
746} __attribute__ ((packed));
747/* followed by split ino list, then split realms, then the trace blob */
748
749/*
750 * encode info about a snaprealm, as viewed by a client
751 */
752struct ceph_mds_snap_realm {
753	__le64 ino;           /* ino */
754	__le64 created;       /* snap: when created */
755	__le64 parent;        /* ino: parent realm */
756	__le64 parent_since;  /* snap: same parent since */
757	__le64 seq;           /* snap: version */
758	__le32 num_snaps;
759	__le32 num_prior_parent_snaps;
760} __attribute__ ((packed));
761/* followed by my snap list, then prior parent snap list */
762
763#endif
764