5 * Data types for the Ceph distributed object storage layer RADOS
6 * (Reliable Autonomic Distributed Object Store).
9 #include <linux/ceph/msgr.h>
12 * osdmap encoding versions
14 #define CEPH_OSDMAP_INC_VERSION 5
15 #define CEPH_OSDMAP_INC_VERSION_EXT 6
16 #define CEPH_OSDMAP_VERSION 5
17 #define CEPH_OSDMAP_VERSION_EXT 6
23 unsigned char fsid[16];
26 static inline int ceph_fsid_compare(const struct ceph_fsid *a,
27 const struct ceph_fsid *b)
29 return memcmp(a, b, sizeof(*a));
35 typedef __le64 ceph_snapid_t;
36 #define CEPH_SNAPDIR ((__u64)(-1)) /* reserved for hidden .snap dir */
37 #define CEPH_NOSNAP ((__u64)(-2)) /* "head", "live" revision */
38 #define CEPH_MAXSNAP ((__u64)(-3)) /* largest valid snapid */
40 struct ceph_timespec {
43 } __attribute__ ((packed));
47 * object layout - how objects are mapped into PGs
49 #define CEPH_OBJECT_LAYOUT_HASH 1
50 #define CEPH_OBJECT_LAYOUT_LINEAR 2
51 #define CEPH_OBJECT_LAYOUT_HASHINO 3
54 * pg layout -- how PGs are mapped onto (sets of) OSDs
56 #define CEPH_PG_LAYOUT_CRUSH 0
57 #define CEPH_PG_LAYOUT_HASH 1
58 #define CEPH_PG_LAYOUT_LINEAR 2
59 #define CEPH_PG_LAYOUT_HYBRID 3
61 #define CEPH_PG_MAX_SIZE 16 /* max # osds in a single pg */
65 * we encode this into one __le64.
68 __le16 preferred; /* preferred primary osd */
69 __le16 ps; /* placement seed */
70 __le32 pool; /* object pool */
71 } __attribute__ ((packed));
74 * pg_pool is a set of pgs storing a pool of objects
76 * pg_num -- base number of pseudorandomly placed pgs
78 * pgp_num -- effective number when calculating pg placement. this
79 * is used for pg_num increases. new pgs result in data being "split"
80 * into new pgs. for this to proceed smoothly, new pgs are intiially
81 * colocated with their parents; that is, pgp_num doesn't increase
82 * until the new pgs have successfully split. only _then_ are the new
83 * pgs placed independently.
85 * lpg_num -- localized pg count (per device). replicas are randomly
88 * lpgp_num -- as above.
90 #define CEPH_NOPOOL ((__u64) (-1)) /* pool id not defined */
92 #define CEPH_PG_TYPE_REP 1
93 #define CEPH_PG_TYPE_RAID4 2
94 #define CEPH_PG_POOL_VERSION 2
96 __u8 type; /* CEPH_PG_TYPE_* */
97 __u8 size; /* number of osds in each pg */
98 __u8 crush_ruleset; /* crush placement rule */
99 __u8 object_hash; /* hash mapping object name to ps */
100 __le32 pg_num, pgp_num; /* number of pg's */
101 __le32 lpg_num, lpgp_num; /* number of localized pg's */
102 __le32 last_change; /* most recent epoch changed */
103 __le64 snap_seq; /* seq for per-pool snapshot */
104 __le32 snap_epoch; /* epoch of last snap */
106 __le32 num_removed_snap_intervals; /* if non-empty, NO per-pool snaps */
107 __le64 auid; /* who owns the pg */
108 } __attribute__ ((packed));
111 * stable_mod func is used to control number of placement groups.
112 * similar to straight-up modulo, but produces a stable mapping as b
113 * increases over time. b is the number of bins, and bmask is the
114 * containing power of 2 minus 1.
116 * b <= bmask and bmask=(2**n)-1
117 * e.g., b=12 -> bmask=15, b=123 -> bmask=127
119 static inline int ceph_stable_mod(int x, int b, int bmask)
124 return x & (bmask >> 1);
128 * object layout - how a given object should be stored.
130 struct ceph_object_layout {
131 struct ceph_pg ol_pgid; /* raw pg, with _full_ ps precision. */
132 __le32 ol_stripe_unit; /* for per-object parity, if any */
133 } __attribute__ ((packed));
136 * compound epoch+version, used by storage layer to serialize mutations
138 struct ceph_eversion {
141 } __attribute__ ((packed));
148 #define CEPH_OSD_EXISTS 1
149 #define CEPH_OSD_UP 2
151 /* osd weights. fixed point value: 0x10000 == 1.0 ("in"), 0 == "out" */
152 #define CEPH_OSD_IN 0x10000
153 #define CEPH_OSD_OUT 0
159 #define CEPH_OSDMAP_NEARFULL (1<<0) /* sync writes (near ENOSPC) */
160 #define CEPH_OSDMAP_FULL (1<<1) /* no data writes (ENOSPC) */
161 #define CEPH_OSDMAP_PAUSERD (1<<2) /* pause all reads */
162 #define CEPH_OSDMAP_PAUSEWR (1<<3) /* pause all writes */
163 #define CEPH_OSDMAP_PAUSEREC (1<<4) /* pause recovery */
168 #define CEPH_OSD_OP_MODE 0xf000
169 #define CEPH_OSD_OP_MODE_RD 0x1000
170 #define CEPH_OSD_OP_MODE_WR 0x2000
171 #define CEPH_OSD_OP_MODE_RMW 0x3000
172 #define CEPH_OSD_OP_MODE_SUB 0x4000
174 #define CEPH_OSD_OP_TYPE 0x0f00
175 #define CEPH_OSD_OP_TYPE_LOCK 0x0100
176 #define CEPH_OSD_OP_TYPE_DATA 0x0200
177 #define CEPH_OSD_OP_TYPE_ATTR 0x0300
178 #define CEPH_OSD_OP_TYPE_EXEC 0x0400
179 #define CEPH_OSD_OP_TYPE_PG 0x0500
184 CEPH_OSD_OP_READ = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 1,
185 CEPH_OSD_OP_STAT = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 2,
186 CEPH_OSD_OP_MAPEXT = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 3,
189 CEPH_OSD_OP_MASKTRUNC = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 4,
190 CEPH_OSD_OP_SPARSE_READ = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 5,
192 CEPH_OSD_OP_NOTIFY = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 6,
193 CEPH_OSD_OP_NOTIFY_ACK = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 7,
196 CEPH_OSD_OP_ASSERT_VER = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 8,
199 CEPH_OSD_OP_WRITE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 1,
200 CEPH_OSD_OP_WRITEFULL = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 2,
201 CEPH_OSD_OP_TRUNCATE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 3,
202 CEPH_OSD_OP_ZERO = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 4,
203 CEPH_OSD_OP_DELETE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 5,
206 CEPH_OSD_OP_APPEND = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 6,
207 CEPH_OSD_OP_STARTSYNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 7,
208 CEPH_OSD_OP_SETTRUNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 8,
209 CEPH_OSD_OP_TRIMTRUNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 9,
211 CEPH_OSD_OP_TMAPUP = CEPH_OSD_OP_MODE_RMW | CEPH_OSD_OP_TYPE_DATA | 10,
212 CEPH_OSD_OP_TMAPPUT = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 11,
213 CEPH_OSD_OP_TMAPGET = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 12,
215 CEPH_OSD_OP_CREATE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 13,
216 CEPH_OSD_OP_ROLLBACK= CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 14,
218 CEPH_OSD_OP_WATCH = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 15,
222 CEPH_OSD_OP_GETXATTR = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_ATTR | 1,
223 CEPH_OSD_OP_GETXATTRS = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_ATTR | 2,
224 CEPH_OSD_OP_CMPXATTR = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_ATTR | 3,
227 CEPH_OSD_OP_SETXATTR = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 1,
228 CEPH_OSD_OP_SETXATTRS = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 2,
229 CEPH_OSD_OP_RESETXATTRS = CEPH_OSD_OP_MODE_WR|CEPH_OSD_OP_TYPE_ATTR | 3,
230 CEPH_OSD_OP_RMXATTR = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 4,
233 CEPH_OSD_OP_PULL = CEPH_OSD_OP_MODE_SUB | 1,
234 CEPH_OSD_OP_PUSH = CEPH_OSD_OP_MODE_SUB | 2,
235 CEPH_OSD_OP_BALANCEREADS = CEPH_OSD_OP_MODE_SUB | 3,
236 CEPH_OSD_OP_UNBALANCEREADS = CEPH_OSD_OP_MODE_SUB | 4,
237 CEPH_OSD_OP_SCRUB = CEPH_OSD_OP_MODE_SUB | 5,
238 CEPH_OSD_OP_SCRUB_RESERVE = CEPH_OSD_OP_MODE_SUB | 6,
239 CEPH_OSD_OP_SCRUB_UNRESERVE = CEPH_OSD_OP_MODE_SUB | 7,
240 CEPH_OSD_OP_SCRUB_STOP = CEPH_OSD_OP_MODE_SUB | 8,
243 CEPH_OSD_OP_WRLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 1,
244 CEPH_OSD_OP_WRUNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 2,
245 CEPH_OSD_OP_RDLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 3,
246 CEPH_OSD_OP_RDUNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 4,
247 CEPH_OSD_OP_UPLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 5,
248 CEPH_OSD_OP_DNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 6,
251 CEPH_OSD_OP_CALL = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_EXEC | 1,
254 CEPH_OSD_OP_PGLS = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_PG | 1,
257 static inline int ceph_osd_op_type_lock(int op)
259 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_LOCK;
261 static inline int ceph_osd_op_type_data(int op)
263 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_DATA;
265 static inline int ceph_osd_op_type_attr(int op)
267 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_ATTR;
269 static inline int ceph_osd_op_type_exec(int op)
271 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_EXEC;
273 static inline int ceph_osd_op_type_pg(int op)
275 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_PG;
278 static inline int ceph_osd_op_mode_subop(int op)
280 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_SUB;
282 static inline int ceph_osd_op_mode_read(int op)
284 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_RD;
286 static inline int ceph_osd_op_mode_modify(int op)
288 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_WR;
292 * note that the following tmap stuff is also defined in the ceph librados.h
293 * any modification here needs to be updated there
295 #define CEPH_OSD_TMAP_HDR 'h'
296 #define CEPH_OSD_TMAP_SET 's'
297 #define CEPH_OSD_TMAP_RM 'r'
299 extern const char *ceph_osd_op_name(int op);
305 * An op may be READ, WRITE, or READ|WRITE.
308 CEPH_OSD_FLAG_ACK = 1, /* want (or is) "ack" ack */
309 CEPH_OSD_FLAG_ONNVRAM = 2, /* want (or is) "onnvram" ack */
310 CEPH_OSD_FLAG_ONDISK = 4, /* want (or is) "ondisk" ack */
311 CEPH_OSD_FLAG_RETRY = 8, /* resend attempt */
312 CEPH_OSD_FLAG_READ = 16, /* op may read */
313 CEPH_OSD_FLAG_WRITE = 32, /* op may write */
314 CEPH_OSD_FLAG_ORDERSNAP = 64, /* EOLDSNAP if snapc is out of order */
315 CEPH_OSD_FLAG_PEERSTAT = 128, /* msg includes osd_peer_stat */
316 CEPH_OSD_FLAG_BALANCE_READS = 256,
317 CEPH_OSD_FLAG_PARALLELEXEC = 512, /* execute op in parallel */
318 CEPH_OSD_FLAG_PGOP = 1024, /* pg op, no object */
319 CEPH_OSD_FLAG_EXEC = 2048, /* op may exec */
320 CEPH_OSD_FLAG_EXEC_PUBLIC = 4096, /* op may exec (public) */
324 CEPH_OSD_OP_FLAG_EXCL = 1, /* EXCL object create */
327 #define EOLDSNAPC ERESTART /* ORDERSNAP flag set; writer has old snapc*/
328 #define EBLACKLISTED ESHUTDOWN /* blacklisted */
330 /* xattr comparison */
332 CEPH_OSD_CMPXATTR_OP_NOP = 0,
333 CEPH_OSD_CMPXATTR_OP_EQ = 1,
334 CEPH_OSD_CMPXATTR_OP_NE = 2,
335 CEPH_OSD_CMPXATTR_OP_GT = 3,
336 CEPH_OSD_CMPXATTR_OP_GTE = 4,
337 CEPH_OSD_CMPXATTR_OP_LT = 5,
338 CEPH_OSD_CMPXATTR_OP_LTE = 6
342 CEPH_OSD_CMPXATTR_MODE_STRING = 1,
343 CEPH_OSD_CMPXATTR_MODE_U64 = 2
346 #define RADOS_NOTIFY_VER 1
349 * an individual object operation. each may be accompanied by some data
353 __le16 op; /* CEPH_OSD_OP_* */
354 __le32 flags; /* CEPH_OSD_FLAG_* */
357 __le64 offset, length;
358 __le64 truncate_size;
360 } __attribute__ ((packed)) extent;
364 __u8 cmp_op; /* CEPH_OSD_CMPXATTR_OP_* */
365 __u8 cmp_mode; /* CEPH_OSD_CMPXATTR_MODE_* */
366 } __attribute__ ((packed)) xattr;
372 } __attribute__ ((packed)) cls;
374 __le64 cookie, count;
375 } __attribute__ ((packed)) pgls;
378 } __attribute__ ((packed)) snap;
382 __u8 flag; /* 0 = unwatch, 1 = watch */
383 } __attribute__ ((packed)) watch;
386 } __attribute__ ((packed));
389 * osd request message header. each request may include multiple
390 * ceph_osd_op object operations.
392 struct ceph_osd_request_head {
393 __le32 client_inc; /* client incarnation */
394 struct ceph_object_layout layout; /* pgid */
395 __le32 osdmap_epoch; /* client's osdmap epoch */
399 struct ceph_timespec mtime; /* for mutations only */
400 struct ceph_eversion reassert_version; /* if we are replaying op */
402 __le32 object_len; /* length of object name */
404 __le64 snapid; /* snapid to read */
405 __le64 snap_seq; /* writer's snap context */
409 struct ceph_osd_op ops[]; /* followed by ops[], obj, ticket, snaps */
410 } __attribute__ ((packed));
412 struct ceph_osd_reply_head {
413 __le32 client_inc; /* client incarnation */
415 struct ceph_object_layout layout;
417 struct ceph_eversion reassert_version; /* for replaying uncommitted */
419 __le32 result; /* result code */
421 __le32 object_len; /* length of object name */
423 struct ceph_osd_op ops[0]; /* ops[], object */
424 } __attribute__ ((packed));