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.
61 __s16 preferred; /* preferred primary osd */
62 __u16 ps; /* placement seed */
63 __u32 pool; /* object pool */
64 } __attribute__ ((packed)) pg;
65 } __attribute__ ((packed));
68 * pg_pool is a set of pgs storing a pool of objects
70 * pg_num -- base number of pseudorandomly placed pgs
72 * pgp_num -- effective number when calculating pg placement. this
73 * is used for pg_num increases. new pgs result in data being "split"
74 * into new pgs. for this to proceed smoothly, new pgs are intiially
75 * colocated with their parents; that is, pgp_num doesn't increase
76 * until the new pgs have successfully split. only _then_ are the new
77 * pgs placed independently.
79 * lpg_num -- localized pg count (per device). replicas are randomly
82 * lpgp_num -- as above.
84 #define CEPH_PG_TYPE_REP 1
85 #define CEPH_PG_TYPE_RAID4 2
87 __u8 type; /* CEPH_PG_TYPE_* */
88 __u8 size; /* number of osds in each pg */
89 __u8 crush_ruleset; /* crush placement rule */
90 __le32 pg_num, pgp_num; /* number of pg's */
91 __le32 lpg_num, lpgp_num; /* number of localized pg's */
92 __le32 last_change; /* most recent epoch changed */
93 __le64 snap_seq; /* seq for per-pool snapshot */
94 __le32 snap_epoch; /* epoch of last snap */
96 __le32 num_removed_snap_intervals;
97 } __attribute__ ((packed));
100 * stable_mod func is used to control number of placement groups.
101 * similar to straight-up modulo, but produces a stable mapping as b
102 * increases over time. b is the number of bins, and bmask is the
103 * containing power of 2 minus 1.
105 * b <= bmask and bmask=(2**n)-1
106 * e.g., b=12 -> bmask=15, b=123 -> bmask=127
108 static inline int ceph_stable_mod(int x, int b, int bmask)
113 return x & (bmask >> 1);
117 * object layout - how a given object should be stored.
119 struct ceph_object_layout {
120 __le64 ol_pgid; /* raw pg, with _full_ ps precision. */
121 __le32 ol_stripe_unit; /* for per-object parity, if any */
122 } __attribute__ ((packed));
125 * compound epoch+version, used by storage layer to serialize mutations
127 struct ceph_eversion {
130 } __attribute__ ((packed));
137 #define CEPH_OSD_EXISTS 1
138 #define CEPH_OSD_UP 2
140 /* osd weights. fixed point value: 0x10000 == 1.0 ("in"), 0 == "out" */
141 #define CEPH_OSD_IN 0x10000
142 #define CEPH_OSD_OUT 0
148 #define CEPH_OSDMAP_NEARFULL (1<<0) /* sync writes (near ENOSPC) */
149 #define CEPH_OSDMAP_FULL (1<<1) /* no data writes (ENOSPC) */
150 #define CEPH_OSDMAP_PAUSERD (1<<2) /* pause all reads */
151 #define CEPH_OSDMAP_PAUSEWR (1<<3) /* pause all writes */
152 #define CEPH_OSDMAP_PAUSEREC (1<<4) /* pause recovery */
157 #define CEPH_OSD_OP_MODE 0xf000
158 #define CEPH_OSD_OP_MODE_RD 0x1000
159 #define CEPH_OSD_OP_MODE_WR 0x2000
160 #define CEPH_OSD_OP_MODE_RMW 0x3000
161 #define CEPH_OSD_OP_MODE_SUB 0x4000
162 #define CEPH_OSD_OP_MODE_EXEC 0x8000
164 #define CEPH_OSD_OP_TYPE 0x0f00
165 #define CEPH_OSD_OP_TYPE_LOCK 0x0100
166 #define CEPH_OSD_OP_TYPE_DATA 0x0200
167 #define CEPH_OSD_OP_TYPE_ATTR 0x0300
168 #define CEPH_OSD_OP_TYPE_EXEC 0x0400
169 #define CEPH_OSD_OP_TYPE_PG 0x0500
174 CEPH_OSD_OP_READ = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 1,
175 CEPH_OSD_OP_STAT = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 2,
178 CEPH_OSD_OP_MASKTRUNC = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 4,
181 CEPH_OSD_OP_WRITE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 1,
182 CEPH_OSD_OP_WRITEFULL = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 2,
183 CEPH_OSD_OP_TRUNCATE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 3,
184 CEPH_OSD_OP_ZERO = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 4,
185 CEPH_OSD_OP_DELETE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 5,
188 CEPH_OSD_OP_APPEND = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 6,
189 CEPH_OSD_OP_STARTSYNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 7,
190 CEPH_OSD_OP_SETTRUNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 8,
191 CEPH_OSD_OP_TRIMTRUNC = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 9,
193 CEPH_OSD_OP_TMAPUP = CEPH_OSD_OP_MODE_RMW | CEPH_OSD_OP_TYPE_DATA | 10,
194 CEPH_OSD_OP_TMAPPUT = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 11,
195 CEPH_OSD_OP_TMAPGET = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_DATA | 12,
197 CEPH_OSD_OP_CREATE = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_DATA | 13,
201 CEPH_OSD_OP_GETXATTR = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_ATTR | 1,
202 CEPH_OSD_OP_GETXATTRS = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_ATTR | 2,
205 CEPH_OSD_OP_SETXATTR = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 1,
206 CEPH_OSD_OP_SETXATTRS = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 2,
207 CEPH_OSD_OP_RESETXATTRS = CEPH_OSD_OP_MODE_WR|CEPH_OSD_OP_TYPE_ATTR | 3,
208 CEPH_OSD_OP_RMXATTR = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_ATTR | 4,
211 CEPH_OSD_OP_PULL = CEPH_OSD_OP_MODE_SUB | 1,
212 CEPH_OSD_OP_PUSH = CEPH_OSD_OP_MODE_SUB | 2,
213 CEPH_OSD_OP_BALANCEREADS = CEPH_OSD_OP_MODE_SUB | 3,
214 CEPH_OSD_OP_UNBALANCEREADS = CEPH_OSD_OP_MODE_SUB | 4,
215 CEPH_OSD_OP_SCRUB = CEPH_OSD_OP_MODE_SUB | 5,
218 CEPH_OSD_OP_WRLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 1,
219 CEPH_OSD_OP_WRUNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 2,
220 CEPH_OSD_OP_RDLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 3,
221 CEPH_OSD_OP_RDUNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 4,
222 CEPH_OSD_OP_UPLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 5,
223 CEPH_OSD_OP_DNLOCK = CEPH_OSD_OP_MODE_WR | CEPH_OSD_OP_TYPE_LOCK | 6,
226 CEPH_OSD_OP_CALL = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_EXEC | 1,
229 CEPH_OSD_OP_PGLS = CEPH_OSD_OP_MODE_RD | CEPH_OSD_OP_TYPE_PG | 1,
232 static inline int ceph_osd_op_type_lock(int op)
234 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_LOCK;
236 static inline int ceph_osd_op_type_data(int op)
238 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_DATA;
240 static inline int ceph_osd_op_type_attr(int op)
242 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_ATTR;
244 static inline int ceph_osd_op_type_exec(int op)
246 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_EXEC;
248 static inline int ceph_osd_op_type_pg(int op)
250 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_PG;
253 static inline int ceph_osd_op_mode_subop(int op)
255 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_SUB;
257 static inline int ceph_osd_op_mode_read(int op)
259 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_RD;
261 static inline int ceph_osd_op_mode_modify(int op)
263 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_WR;
266 #define CEPH_OSD_TMAP_HDR 'h'
267 #define CEPH_OSD_TMAP_SET 's'
268 #define CEPH_OSD_TMAP_RM 'r'
270 extern const char *ceph_osd_op_name(int op);
276 * An op may be READ, WRITE, or READ|WRITE.
279 CEPH_OSD_FLAG_ACK = 1, /* want (or is) "ack" ack */
280 CEPH_OSD_FLAG_ONNVRAM = 2, /* want (or is) "onnvram" ack */
281 CEPH_OSD_FLAG_ONDISK = 4, /* want (or is) "ondisk" ack */
282 CEPH_OSD_FLAG_RETRY = 8, /* resend attempt */
283 CEPH_OSD_FLAG_READ = 16, /* op may read */
284 CEPH_OSD_FLAG_WRITE = 32, /* op may write */
285 CEPH_OSD_FLAG_ORDERSNAP = 64, /* EOLDSNAP if snapc is out of order */
286 CEPH_OSD_FLAG_PEERSTAT = 128, /* msg includes osd_peer_stat */
287 CEPH_OSD_FLAG_BALANCE_READS = 256,
288 CEPH_OSD_FLAG_PARALLELEXEC = 512, /* execute op in parallel */
289 CEPH_OSD_FLAG_PGOP = 1024, /* pg op, no object */
293 CEPH_OSD_OP_FLAG_EXCL = 1, /* EXCL object create */
296 #define EOLDSNAPC ERESTART /* ORDERSNAP flag set; writer has old snapc*/
297 #define EBLACKLISTED ESHUTDOWN /* blacklisted */
300 * an individual object operation. each may be accompanied by some data
304 __le16 op; /* CEPH_OSD_OP_* */
305 __le32 flags; /* CEPH_OSD_FLAG_* */
308 __le64 offset, length;
309 } __attribute__ ((packed)) extent;
313 } __attribute__ ((packed)) xattr;
315 __le64 truncate_size;
317 } __attribute__ ((packed)) trunc;
323 } __attribute__ ((packed)) cls;
325 __le64 cookie, count;
326 } __attribute__ ((packed)) pgls;
329 } __attribute__ ((packed));
332 * osd request message header. each request may include multiple
333 * ceph_osd_op object operations.
335 struct ceph_osd_request_head {
336 __le64 tid; /* transaction id */
337 __le32 client_inc; /* client incarnation */
338 struct ceph_object_layout layout; /* pgid */
339 __le32 osdmap_epoch; /* client's osdmap epoch */
343 struct ceph_timespec mtime; /* for mutations only */
344 struct ceph_eversion reassert_version; /* if we are replaying op */
346 __le32 object_len; /* length of object name */
348 __le64 snapid; /* snapid to read */
349 __le64 snap_seq; /* writer's snap context */
353 struct ceph_osd_op ops[]; /* followed by ops[], obj, ticket, snaps */
354 } __attribute__ ((packed));
356 struct ceph_osd_reply_head {
357 __le64 tid; /* transaction id */
358 __le32 client_inc; /* client incarnation */
360 struct ceph_object_layout layout;
362 struct ceph_eversion reassert_version; /* for replaying uncommitted */
364 __le32 result; /* result code */
366 __le32 object_len; /* length of object name */
368 struct ceph_osd_op ops[0]; /* ops[], object */
369 } __attribute__ ((packed));