5 * Data types for the Ceph distributed object storage layer RADOS
6 * (Reliable Autonomic Distributed Object Store).
15 unsigned char fsid[16];
18 static inline int ceph_fsid_compare(const struct ceph_fsid *a,
19 const struct ceph_fsid *b)
21 return memcmp(a, b, sizeof(*a));
27 typedef __le64 ceph_snapid_t;
28 #define CEPH_SNAPDIR ((__u64)(-1)) /* reserved for hidden .snap dir */
29 #define CEPH_NOSNAP ((__u64)(-2)) /* "head", "live" revision */
30 #define CEPH_MAXSNAP ((__u64)(-3)) /* largest valid snapid */
32 struct ceph_timespec {
35 } __attribute__ ((packed));
39 * object layout - how objects are mapped into PGs
41 #define CEPH_OBJECT_LAYOUT_HASH 1
42 #define CEPH_OBJECT_LAYOUT_LINEAR 2
43 #define CEPH_OBJECT_LAYOUT_HASHINO 3
46 * pg layout -- how PGs are mapped onto (sets of) OSDs
48 #define CEPH_PG_LAYOUT_CRUSH 0
49 #define CEPH_PG_LAYOUT_HASH 1
50 #define CEPH_PG_LAYOUT_LINEAR 2
51 #define CEPH_PG_LAYOUT_HYBRID 3
56 * we encode this into one __le64.
59 __le16 preferred; /* preferred primary osd */
60 __le16 ps; /* placement seed */
61 __le32 pool; /* object pool */
62 } __attribute__ ((packed));
65 * pg_pool is a set of pgs storing a pool of objects
67 * pg_num -- base number of pseudorandomly placed pgs
69 * pgp_num -- effective number when calculating pg placement. this
70 * is used for pg_num increases. new pgs result in data being "split"
71 * into new pgs. for this to proceed smoothly, new pgs are intiially
72 * colocated with their parents; that is, pgp_num doesn't increase
73 * until the new pgs have successfully split. only _then_ are the new
74 * pgs placed independently.
76 * lpg_num -- localized pg count (per device). replicas are randomly
79 * lpgp_num -- as above.
81 #define CEPH_PG_TYPE_REP 1
82 #define CEPH_PG_TYPE_RAID4 2
84 __u8 type; /* CEPH_PG_TYPE_* */
85 __u8 size; /* number of osds in each pg */
86 __u8 crush_ruleset; /* crush placement rule */
87 __u8 object_hash; /* hash mapping object name to ps */
88 __le32 pg_num, pgp_num; /* number of pg's */
89 __le32 lpg_num, lpgp_num; /* number of localized pg's */
90 __le32 last_change; /* most recent epoch changed */
91 __le64 snap_seq; /* seq for per-pool snapshot */
92 __le32 snap_epoch; /* epoch of last snap */
94 __le32 num_removed_snap_intervals;
95 } __attribute__ ((packed));
98 * stable_mod func is used to control number of placement groups.
99 * similar to straight-up modulo, but produces a stable mapping as b
100 * increases over time. b is the number of bins, and bmask is the
101 * containing power of 2 minus 1.
103 * b <= bmask and bmask=(2**n)-1
104 * e.g., b=12 -> bmask=15, b=123 -> bmask=127
106 static inline int ceph_stable_mod(int x, int b, int bmask)
111 return x & (bmask >> 1);
115 * object layout - how a given object should be stored.
117 struct ceph_object_layout {
118 struct ceph_pg ol_pgid; /* raw pg, with _full_ ps precision. */
119 __le32 ol_stripe_unit; /* for per-object parity, if any */
120 } __attribute__ ((packed));
123 * compound epoch+version, used by storage layer to serialize mutations
125 struct ceph_eversion {
128 } __attribute__ ((packed));
135 #define CEPH_OSD_EXISTS 1
136 #define CEPH_OSD_UP 2
138 /* osd weights. fixed point value: 0x10000 == 1.0 ("in"), 0 == "out" */
139 #define CEPH_OSD_IN 0x10000
140 #define CEPH_OSD_OUT 0
146 #define CEPH_OSDMAP_NEARFULL (1<<0) /* sync writes (near ENOSPC) */
147 #define CEPH_OSDMAP_FULL (1<<1) /* no data writes (ENOSPC) */
148 #define CEPH_OSDMAP_PAUSERD (1<<2) /* pause all reads */
149 #define CEPH_OSDMAP_PAUSEWR (1<<3) /* pause all writes */
150 #define CEPH_OSDMAP_PAUSEREC (1<<4) /* pause recovery */
155 #define CEPH_OSD_OP_MODE 0xf000
156 #define CEPH_OSD_OP_MODE_RD 0x1000
157 #define CEPH_OSD_OP_MODE_WR 0x2000
158 #define CEPH_OSD_OP_MODE_RMW 0x3000
159 #define CEPH_OSD_OP_MODE_SUB 0x4000
161 #define CEPH_OSD_OP_TYPE 0x0f00
162 #define CEPH_OSD_OP_TYPE_LOCK 0x0100
163 #define CEPH_OSD_OP_TYPE_DATA 0x0200
164 #define CEPH_OSD_OP_TYPE_ATTR 0x0300
165 #define CEPH_OSD_OP_TYPE_EXEC 0x0400
166 #define CEPH_OSD_OP_TYPE_PG 0x0500
171 CEPH_OSD_OP_READ = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 1,
172 CEPH_OSD_OP_STAT = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 2,
175 CEPH_OSD_OP_MASKTRUNC = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 4,
178 CEPH_OSD_OP_WRITE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 1,
179 CEPH_OSD_OP_WRITEFULL = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 2,
180 CEPH_OSD_OP_TRUNCATE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 3,
181 CEPH_OSD_OP_ZERO = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 4,
182 CEPH_OSD_OP_DELETE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 5,
185 CEPH_OSD_OP_APPEND = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 6,
186 CEPH_OSD_OP_STARTSYNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 7,
187 CEPH_OSD_OP_SETTRUNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 8,
188 CEPH_OSD_OP_TRIMTRUNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 9,
190 CEPH_OSD_OP_TMAPUP = CEPH_OSD_OP_MODE_RMW | CEPH_OSD_OP_TYPE_DATA | 10,
191 CEPH_OSD_OP_TMAPPUT = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 11,
192 CEPH_OSD_OP_TMAPGET = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 12,
194 CEPH_OSD_OP_CREATE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 13,
198 CEPH_OSD_OP_GETXATTR = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_ATTR | 1,
199 CEPH_OSD_OP_GETXATTRS = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_ATTR | 2,
202 CEPH_OSD_OP_SETXATTR = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 1,
203 CEPH_OSD_OP_SETXATTRS = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 2,
204 CEPH_OSD_OP_RESETXATTRS = CEPH_OSD_OP_MODE_WR|CEPH_OSD_OP_TYPE_ATTR | 3,
205 CEPH_OSD_OP_RMXATTR = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 4,
208 CEPH_OSD_OP_PULL = CEPH_OSD_OP_MODE_SUB | 1,
209 CEPH_OSD_OP_PUSH = CEPH_OSD_OP_MODE_SUB | 2,
210 CEPH_OSD_OP_BALANCEREADS = CEPH_OSD_OP_MODE_SUB | 3,
211 CEPH_OSD_OP_UNBALANCEREADS = CEPH_OSD_OP_MODE_SUB | 4,
212 CEPH_OSD_OP_SCRUB = CEPH_OSD_OP_MODE_SUB | 5,
215 CEPH_OSD_OP_WRLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 1,
216 CEPH_OSD_OP_WRUNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 2,
217 CEPH_OSD_OP_RDLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 3,
218 CEPH_OSD_OP_RDUNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 4,
219 CEPH_OSD_OP_UPLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 5,
220 CEPH_OSD_OP_DNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 6,
223 CEPH_OSD_OP_CALL = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_EXEC | 1,
226 CEPH_OSD_OP_PGLS = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_PG | 1,
229 static inline int ceph_osd_op_type_lock(int op)
231 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_LOCK;
233 static inline int ceph_osd_op_type_data(int op)
235 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_DATA;
237 static inline int ceph_osd_op_type_attr(int op)
239 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_ATTR;
241 static inline int ceph_osd_op_type_exec(int op)
243 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_EXEC;
245 static inline int ceph_osd_op_type_pg(int op)
247 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_PG;
250 static inline int ceph_osd_op_mode_subop(int op)
252 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_SUB;
254 static inline int ceph_osd_op_mode_read(int op)
256 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_RD;
258 static inline int ceph_osd_op_mode_modify(int op)
260 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_WR;
263 #define CEPH_OSD_TMAP_HDR 'h'
264 #define CEPH_OSD_TMAP_SET 's'
265 #define CEPH_OSD_TMAP_RM 'r'
267 extern const char *ceph_osd_op_name(int op);
273 * An op may be READ, WRITE, or READ|WRITE.
276 CEPH_OSD_FLAG_ACK = 1, /* want (or is) "ack" ack */
277 CEPH_OSD_FLAG_ONNVRAM = 2, /* want (or is) "onnvram" ack */
278 CEPH_OSD_FLAG_ONDISK = 4, /* want (or is) "ondisk" ack */
279 CEPH_OSD_FLAG_RETRY = 8, /* resend attempt */
280 CEPH_OSD_FLAG_READ = 16, /* op may read */
281 CEPH_OSD_FLAG_WRITE = 32, /* op may write */
282 CEPH_OSD_FLAG_ORDERSNAP = 64, /* EOLDSNAP if snapc is out of order */
283 CEPH_OSD_FLAG_PEERSTAT = 128, /* msg includes osd_peer_stat */
284 CEPH_OSD_FLAG_BALANCE_READS = 256,
285 CEPH_OSD_FLAG_PARALLELEXEC = 512, /* execute op in parallel */
286 CEPH_OSD_FLAG_PGOP = 1024, /* pg op, no object */
287 CEPH_OSD_FLAG_EXEC = 2048, /* op may exec */
291 CEPH_OSD_OP_FLAG_EXCL = 1, /* EXCL object create */
294 #define EOLDSNAPC ERESTART /* ORDERSNAP flag set; writer has old snapc*/
295 #define EBLACKLISTED ESHUTDOWN /* blacklisted */
298 * an individual object operation. each may be accompanied by some data
302 __le16 op; /* CEPH_OSD_OP_* */
303 __le32 flags; /* CEPH_OSD_FLAG_* */
306 __le64 offset, length;
307 } __attribute__ ((packed)) extent;
311 } __attribute__ ((packed)) xattr;
313 __le64 truncate_size;
315 } __attribute__ ((packed)) trunc;
321 } __attribute__ ((packed)) cls;
323 __le64 cookie, count;
324 } __attribute__ ((packed)) pgls;
327 } __attribute__ ((packed));
330 * osd request message header. each request may include multiple
331 * ceph_osd_op object operations.
333 struct ceph_osd_request_head {
334 __le64 tid; /* transaction id */
335 __le32 client_inc; /* client incarnation */
336 struct ceph_object_layout layout; /* pgid */
337 __le32 osdmap_epoch; /* client's osdmap epoch */
341 struct ceph_timespec mtime; /* for mutations only */
342 struct ceph_eversion reassert_version; /* if we are replaying op */
344 __le32 object_len; /* length of object name */
346 __le64 snapid; /* snapid to read */
347 __le64 snap_seq; /* writer's snap context */
351 struct ceph_osd_op ops[]; /* followed by ops[], obj, ticket, snaps */
352 } __attribute__ ((packed));
354 struct ceph_osd_reply_head {
355 __le64 tid; /* transaction id */
356 __le32 client_inc; /* client incarnation */
358 struct ceph_object_layout layout;
360 struct ceph_eversion reassert_version; /* for replaying uncommitted */
362 __le32 result; /* result code */
364 __le32 object_len; /* length of object name */
366 struct ceph_osd_op ops[0]; /* ops[], object */
367 } __attribute__ ((packed));