1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef CEPH_RADOS_H
3 #define CEPH_RADOS_H
4
5 /*
6 * Data types for the Ceph distributed object storage layer RADOS
7 * (Reliable Autonomic Distributed Object Store).
8 */
9
10 #include <linux/ceph/msgr.h>
11
12 /*
13 * fs id
14 */
15 struct ceph_fsid {
16 unsigned char fsid[16];
17 };
18
ceph_fsid_compare(const struct ceph_fsid * a,const struct ceph_fsid * b)19 static inline int ceph_fsid_compare(const struct ceph_fsid *a,
20 const struct ceph_fsid *b)
21 {
22 return memcmp(a, b, sizeof(*a));
23 }
24
25 /*
26 * ino, object, etc.
27 */
28 typedef __le64 ceph_snapid_t;
29 #define CEPH_SNAPDIR ((__u64)(-1)) /* reserved for hidden .snap dir */
30 #define CEPH_NOSNAP ((__u64)(-2)) /* "head", "live" revision */
31 #define CEPH_MAXSNAP ((__u64)(-3)) /* largest valid snapid */
32
33 struct ceph_timespec {
34 __le32 tv_sec;
35 __le32 tv_nsec;
36 } __attribute__ ((packed));
37
38
39 /*
40 * object layout - how objects are mapped into PGs
41 */
42 #define CEPH_OBJECT_LAYOUT_HASH 1
43 #define CEPH_OBJECT_LAYOUT_LINEAR 2
44 #define CEPH_OBJECT_LAYOUT_HASHINO 3
45
46 /*
47 * pg layout -- how PGs are mapped onto (sets of) OSDs
48 */
49 #define CEPH_PG_LAYOUT_CRUSH 0
50 #define CEPH_PG_LAYOUT_HASH 1
51 #define CEPH_PG_LAYOUT_LINEAR 2
52 #define CEPH_PG_LAYOUT_HYBRID 3
53
54 #define CEPH_PG_MAX_SIZE 32 /* max # osds in a single pg */
55
56 /*
57 * placement group.
58 * we encode this into one __le64.
59 */
60 struct ceph_pg_v1 {
61 __le16 preferred; /* preferred primary osd */
62 __le16 ps; /* placement seed */
63 __le32 pool; /* object pool */
64 } __attribute__ ((packed));
65
66 /*
67 * pg_pool is a set of pgs storing a pool of objects
68 *
69 * pg_num -- base number of pseudorandomly placed pgs
70 *
71 * pgp_num -- effective number when calculating pg placement. this
72 * is used for pg_num increases. new pgs result in data being "split"
73 * into new pgs. for this to proceed smoothly, new pgs are intiially
74 * colocated with their parents; that is, pgp_num doesn't increase
75 * until the new pgs have successfully split. only _then_ are the new
76 * pgs placed independently.
77 *
78 * lpg_num -- localized pg count (per device). replicas are randomly
79 * selected.
80 *
81 * lpgp_num -- as above.
82 */
83 #define CEPH_NOPOOL ((__u64) (-1)) /* pool id not defined */
84
85 #define CEPH_POOL_TYPE_REP 1
86 #define CEPH_POOL_TYPE_RAID4 2 /* never implemented */
87 #define CEPH_POOL_TYPE_EC 3
88
89 /*
90 * stable_mod func is used to control number of placement groups.
91 * similar to straight-up modulo, but produces a stable mapping as b
92 * increases over time. b is the number of bins, and bmask is the
93 * containing power of 2 minus 1.
94 *
95 * b <= bmask and bmask=(2**n)-1
96 * e.g., b=12 -> bmask=15, b=123 -> bmask=127
97 */
ceph_stable_mod(int x,int b,int bmask)98 static inline int ceph_stable_mod(int x, int b, int bmask)
99 {
100 if ((x & bmask) < b)
101 return x & bmask;
102 else
103 return x & (bmask >> 1);
104 }
105
106 /*
107 * object layout - how a given object should be stored.
108 */
109 struct ceph_object_layout {
110 struct ceph_pg_v1 ol_pgid; /* raw pg, with _full_ ps precision. */
111 __le32 ol_stripe_unit; /* for per-object parity, if any */
112 } __attribute__ ((packed));
113
114 /*
115 * compound epoch+version, used by storage layer to serialize mutations
116 */
117 struct ceph_eversion {
118 __le64 version;
119 __le32 epoch;
120 } __attribute__ ((packed));
121
122 /*
123 * osd map bits
124 */
125
126 /* status bits */
127 #define CEPH_OSD_EXISTS (1<<0)
128 #define CEPH_OSD_UP (1<<1)
129 #define CEPH_OSD_AUTOOUT (1<<2) /* osd was automatically marked out */
130 #define CEPH_OSD_NEW (1<<3) /* osd is new, never marked in */
131
132 extern const char *ceph_osd_state_name(int s);
133
134 /* osd weights. fixed point value: 0x10000 == 1.0 ("in"), 0 == "out" */
135 #define CEPH_OSD_IN 0x10000
136 #define CEPH_OSD_OUT 0
137
138 /* osd primary-affinity. fixed point value: 0x10000 == baseline */
139 #define CEPH_OSD_MAX_PRIMARY_AFFINITY 0x10000
140 #define CEPH_OSD_DEFAULT_PRIMARY_AFFINITY 0x10000
141
142
143 /*
144 * osd map flag bits
145 */
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 */
151 #define CEPH_OSDMAP_NOUP (1<<5) /* block osd boot */
152 #define CEPH_OSDMAP_NODOWN (1<<6) /* block osd mark-down/failure */
153 #define CEPH_OSDMAP_NOOUT (1<<7) /* block osd auto mark-out */
154 #define CEPH_OSDMAP_NOIN (1<<8) /* block osd auto mark-in */
155 #define CEPH_OSDMAP_NOBACKFILL (1<<9) /* block osd backfill */
156 #define CEPH_OSDMAP_NORECOVER (1<<10) /* block osd recovery and backfill */
157 #define CEPH_OSDMAP_NOSCRUB (1<<11) /* block periodic scrub */
158 #define CEPH_OSDMAP_NODEEP_SCRUB (1<<12) /* block periodic deep-scrub */
159 #define CEPH_OSDMAP_NOTIERAGENT (1<<13) /* disable tiering agent */
160 #define CEPH_OSDMAP_NOREBALANCE (1<<14) /* block osd backfill unless pg is degraded */
161 #define CEPH_OSDMAP_SORTBITWISE (1<<15) /* use bitwise hobject_t sort */
162 #define CEPH_OSDMAP_REQUIRE_JEWEL (1<<16) /* require jewel for booting osds */
163 #define CEPH_OSDMAP_REQUIRE_KRAKEN (1<<17) /* require kraken for booting osds */
164 #define CEPH_OSDMAP_REQUIRE_LUMINOUS (1<<18) /* require l for booting osds */
165 #define CEPH_OSDMAP_RECOVERY_DELETES (1<<19) /* deletes performed during recovery instead of peering */
166
167 /*
168 * The error code to return when an OSD can't handle a write
169 * because it is too large.
170 */
171 #define OSD_WRITETOOBIG EMSGSIZE
172
173 /*
174 * osd ops
175 *
176 * WARNING: do not use these op codes directly. Use the helpers
177 * defined below instead. In certain cases, op code behavior was
178 * redefined, resulting in special-cases in the helpers.
179 */
180 #define CEPH_OSD_OP_MODE 0xf000
181 #define CEPH_OSD_OP_MODE_RD 0x1000
182 #define CEPH_OSD_OP_MODE_WR 0x2000
183 #define CEPH_OSD_OP_MODE_RMW 0x3000
184 #define CEPH_OSD_OP_MODE_SUB 0x4000
185 #define CEPH_OSD_OP_MODE_CACHE 0x8000
186
187 #define CEPH_OSD_OP_TYPE 0x0f00
188 #define CEPH_OSD_OP_TYPE_LOCK 0x0100
189 #define CEPH_OSD_OP_TYPE_DATA 0x0200
190 #define CEPH_OSD_OP_TYPE_ATTR 0x0300
191 #define CEPH_OSD_OP_TYPE_EXEC 0x0400
192 #define CEPH_OSD_OP_TYPE_PG 0x0500
193 #define CEPH_OSD_OP_TYPE_MULTI 0x0600 /* multiobject */
194
195 #define __CEPH_OSD_OP1(mode, nr) \
196 (CEPH_OSD_OP_MODE_##mode | (nr))
197
198 #define __CEPH_OSD_OP(mode, type, nr) \
199 (CEPH_OSD_OP_MODE_##mode | CEPH_OSD_OP_TYPE_##type | (nr))
200
201 #define __CEPH_FORALL_OSD_OPS(f) \
202 /** data **/ \
203 /* read */ \
204 f(READ, __CEPH_OSD_OP(RD, DATA, 1), "read") \
205 f(STAT, __CEPH_OSD_OP(RD, DATA, 2), "stat") \
206 f(MAPEXT, __CEPH_OSD_OP(RD, DATA, 3), "mapext") \
207 \
208 /* fancy read */ \
209 f(MASKTRUNC, __CEPH_OSD_OP(RD, DATA, 4), "masktrunc") \
210 f(SPARSE_READ, __CEPH_OSD_OP(RD, DATA, 5), "sparse-read") \
211 \
212 f(NOTIFY, __CEPH_OSD_OP(RD, DATA, 6), "notify") \
213 f(NOTIFY_ACK, __CEPH_OSD_OP(RD, DATA, 7), "notify-ack") \
214 \
215 /* versioning */ \
216 f(ASSERT_VER, __CEPH_OSD_OP(RD, DATA, 8), "assert-version") \
217 \
218 f(LIST_WATCHERS, __CEPH_OSD_OP(RD, DATA, 9), "list-watchers") \
219 \
220 f(LIST_SNAPS, __CEPH_OSD_OP(RD, DATA, 10), "list-snaps") \
221 \
222 /* sync */ \
223 f(SYNC_READ, __CEPH_OSD_OP(RD, DATA, 11), "sync_read") \
224 \
225 /* write */ \
226 f(WRITE, __CEPH_OSD_OP(WR, DATA, 1), "write") \
227 f(WRITEFULL, __CEPH_OSD_OP(WR, DATA, 2), "writefull") \
228 f(TRUNCATE, __CEPH_OSD_OP(WR, DATA, 3), "truncate") \
229 f(ZERO, __CEPH_OSD_OP(WR, DATA, 4), "zero") \
230 f(DELETE, __CEPH_OSD_OP(WR, DATA, 5), "delete") \
231 \
232 /* fancy write */ \
233 f(APPEND, __CEPH_OSD_OP(WR, DATA, 6), "append") \
234 f(SETTRUNC, __CEPH_OSD_OP(WR, DATA, 8), "settrunc") \
235 f(TRIMTRUNC, __CEPH_OSD_OP(WR, DATA, 9), "trimtrunc") \
236 \
237 f(TMAPUP, __CEPH_OSD_OP(RMW, DATA, 10), "tmapup") \
238 f(TMAPPUT, __CEPH_OSD_OP(WR, DATA, 11), "tmapput") \
239 f(TMAPGET, __CEPH_OSD_OP(RD, DATA, 12), "tmapget") \
240 \
241 f(CREATE, __CEPH_OSD_OP(WR, DATA, 13), "create") \
242 f(ROLLBACK, __CEPH_OSD_OP(WR, DATA, 14), "rollback") \
243 \
244 f(WATCH, __CEPH_OSD_OP(WR, DATA, 15), "watch") \
245 \
246 /* omap */ \
247 f(OMAPGETKEYS, __CEPH_OSD_OP(RD, DATA, 17), "omap-get-keys") \
248 f(OMAPGETVALS, __CEPH_OSD_OP(RD, DATA, 18), "omap-get-vals") \
249 f(OMAPGETHEADER, __CEPH_OSD_OP(RD, DATA, 19), "omap-get-header") \
250 f(OMAPGETVALSBYKEYS, __CEPH_OSD_OP(RD, DATA, 20), "omap-get-vals-by-keys") \
251 f(OMAPSETVALS, __CEPH_OSD_OP(WR, DATA, 21), "omap-set-vals") \
252 f(OMAPSETHEADER, __CEPH_OSD_OP(WR, DATA, 22), "omap-set-header") \
253 f(OMAPCLEAR, __CEPH_OSD_OP(WR, DATA, 23), "omap-clear") \
254 f(OMAPRMKEYS, __CEPH_OSD_OP(WR, DATA, 24), "omap-rm-keys") \
255 f(OMAP_CMP, __CEPH_OSD_OP(RD, DATA, 25), "omap-cmp") \
256 \
257 /* tiering */ \
258 f(COPY_FROM, __CEPH_OSD_OP(WR, DATA, 26), "copy-from") \
259 f(COPY_GET_CLASSIC, __CEPH_OSD_OP(RD, DATA, 27), "copy-get-classic") \
260 f(UNDIRTY, __CEPH_OSD_OP(WR, DATA, 28), "undirty") \
261 f(ISDIRTY, __CEPH_OSD_OP(RD, DATA, 29), "isdirty") \
262 f(COPY_GET, __CEPH_OSD_OP(RD, DATA, 30), "copy-get") \
263 f(CACHE_FLUSH, __CEPH_OSD_OP(CACHE, DATA, 31), "cache-flush") \
264 f(CACHE_EVICT, __CEPH_OSD_OP(CACHE, DATA, 32), "cache-evict") \
265 f(CACHE_TRY_FLUSH, __CEPH_OSD_OP(CACHE, DATA, 33), "cache-try-flush") \
266 \
267 /* convert tmap to omap */ \
268 f(TMAP2OMAP, __CEPH_OSD_OP(RMW, DATA, 34), "tmap2omap") \
269 \
270 /* hints */ \
271 f(SETALLOCHINT, __CEPH_OSD_OP(WR, DATA, 35), "set-alloc-hint") \
272 \
273 /** multi **/ \
274 f(CLONERANGE, __CEPH_OSD_OP(WR, MULTI, 1), "clonerange") \
275 f(ASSERT_SRC_VERSION, __CEPH_OSD_OP(RD, MULTI, 2), "assert-src-version") \
276 f(SRC_CMPXATTR, __CEPH_OSD_OP(RD, MULTI, 3), "src-cmpxattr") \
277 \
278 /** attrs **/ \
279 /* read */ \
280 f(GETXATTR, __CEPH_OSD_OP(RD, ATTR, 1), "getxattr") \
281 f(GETXATTRS, __CEPH_OSD_OP(RD, ATTR, 2), "getxattrs") \
282 f(CMPXATTR, __CEPH_OSD_OP(RD, ATTR, 3), "cmpxattr") \
283 \
284 /* write */ \
285 f(SETXATTR, __CEPH_OSD_OP(WR, ATTR, 1), "setxattr") \
286 f(SETXATTRS, __CEPH_OSD_OP(WR, ATTR, 2), "setxattrs") \
287 f(RESETXATTRS, __CEPH_OSD_OP(WR, ATTR, 3), "resetxattrs") \
288 f(RMXATTR, __CEPH_OSD_OP(WR, ATTR, 4), "rmxattr") \
289 \
290 /** subop **/ \
291 f(PULL, __CEPH_OSD_OP1(SUB, 1), "pull") \
292 f(PUSH, __CEPH_OSD_OP1(SUB, 2), "push") \
293 f(BALANCEREADS, __CEPH_OSD_OP1(SUB, 3), "balance-reads") \
294 f(UNBALANCEREADS, __CEPH_OSD_OP1(SUB, 4), "unbalance-reads") \
295 f(SCRUB, __CEPH_OSD_OP1(SUB, 5), "scrub") \
296 f(SCRUB_RESERVE, __CEPH_OSD_OP1(SUB, 6), "scrub-reserve") \
297 f(SCRUB_UNRESERVE, __CEPH_OSD_OP1(SUB, 7), "scrub-unreserve") \
298 f(SCRUB_STOP, __CEPH_OSD_OP1(SUB, 8), "scrub-stop") \
299 f(SCRUB_MAP, __CEPH_OSD_OP1(SUB, 9), "scrub-map") \
300 \
301 /** lock **/ \
302 f(WRLOCK, __CEPH_OSD_OP(WR, LOCK, 1), "wrlock") \
303 f(WRUNLOCK, __CEPH_OSD_OP(WR, LOCK, 2), "wrunlock") \
304 f(RDLOCK, __CEPH_OSD_OP(WR, LOCK, 3), "rdlock") \
305 f(RDUNLOCK, __CEPH_OSD_OP(WR, LOCK, 4), "rdunlock") \
306 f(UPLOCK, __CEPH_OSD_OP(WR, LOCK, 5), "uplock") \
307 f(DNLOCK, __CEPH_OSD_OP(WR, LOCK, 6), "dnlock") \
308 \
309 /** exec **/ \
310 /* note: the RD bit here is wrong; see special-case below in helper */ \
311 f(CALL, __CEPH_OSD_OP(RD, EXEC, 1), "call") \
312 \
313 /** pg **/ \
314 f(PGLS, __CEPH_OSD_OP(RD, PG, 1), "pgls") \
315 f(PGLS_FILTER, __CEPH_OSD_OP(RD, PG, 2), "pgls-filter") \
316 f(PG_HITSET_LS, __CEPH_OSD_OP(RD, PG, 3), "pg-hitset-ls") \
317 f(PG_HITSET_GET, __CEPH_OSD_OP(RD, PG, 4), "pg-hitset-get")
318
319 enum {
320 #define GENERATE_ENUM_ENTRY(op, opcode, str) CEPH_OSD_OP_##op = (opcode),
321 __CEPH_FORALL_OSD_OPS(GENERATE_ENUM_ENTRY)
322 #undef GENERATE_ENUM_ENTRY
323 };
324
ceph_osd_op_type_lock(int op)325 static inline int ceph_osd_op_type_lock(int op)
326 {
327 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_LOCK;
328 }
ceph_osd_op_type_data(int op)329 static inline int ceph_osd_op_type_data(int op)
330 {
331 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_DATA;
332 }
ceph_osd_op_type_attr(int op)333 static inline int ceph_osd_op_type_attr(int op)
334 {
335 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_ATTR;
336 }
ceph_osd_op_type_exec(int op)337 static inline int ceph_osd_op_type_exec(int op)
338 {
339 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_EXEC;
340 }
ceph_osd_op_type_pg(int op)341 static inline int ceph_osd_op_type_pg(int op)
342 {
343 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_PG;
344 }
ceph_osd_op_type_multi(int op)345 static inline int ceph_osd_op_type_multi(int op)
346 {
347 return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_MULTI;
348 }
349
ceph_osd_op_mode_subop(int op)350 static inline int ceph_osd_op_mode_subop(int op)
351 {
352 return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_SUB;
353 }
ceph_osd_op_mode_read(int op)354 static inline int ceph_osd_op_mode_read(int op)
355 {
356 return (op & CEPH_OSD_OP_MODE_RD) &&
357 op != CEPH_OSD_OP_CALL;
358 }
ceph_osd_op_mode_modify(int op)359 static inline int ceph_osd_op_mode_modify(int op)
360 {
361 return op & CEPH_OSD_OP_MODE_WR;
362 }
363
364 /*
365 * note that the following tmap stuff is also defined in the ceph librados.h
366 * any modification here needs to be updated there
367 */
368 #define CEPH_OSD_TMAP_HDR 'h'
369 #define CEPH_OSD_TMAP_SET 's'
370 #define CEPH_OSD_TMAP_CREATE 'c' /* create key */
371 #define CEPH_OSD_TMAP_RM 'r'
372 #define CEPH_OSD_TMAP_RMSLOPPY 'R'
373
374 extern const char *ceph_osd_op_name(int op);
375
376 /*
377 * osd op flags
378 *
379 * An op may be READ, WRITE, or READ|WRITE.
380 */
381 enum {
382 CEPH_OSD_FLAG_ACK = 0x0001, /* want (or is) "ack" ack */
383 CEPH_OSD_FLAG_ONNVRAM = 0x0002, /* want (or is) "onnvram" ack */
384 CEPH_OSD_FLAG_ONDISK = 0x0004, /* want (or is) "ondisk" ack */
385 CEPH_OSD_FLAG_RETRY = 0x0008, /* resend attempt */
386 CEPH_OSD_FLAG_READ = 0x0010, /* op may read */
387 CEPH_OSD_FLAG_WRITE = 0x0020, /* op may write */
388 CEPH_OSD_FLAG_ORDERSNAP = 0x0040, /* EOLDSNAP if snapc is out of order */
389 CEPH_OSD_FLAG_PEERSTAT_OLD = 0x0080, /* DEPRECATED msg includes osd_peer_stat */
390 CEPH_OSD_FLAG_BALANCE_READS = 0x0100,
391 CEPH_OSD_FLAG_PARALLELEXEC = 0x0200, /* execute op in parallel */
392 CEPH_OSD_FLAG_PGOP = 0x0400, /* pg op, no object */
393 CEPH_OSD_FLAG_EXEC = 0x0800, /* op may exec */
394 CEPH_OSD_FLAG_EXEC_PUBLIC = 0x1000, /* DEPRECATED op may exec (public) */
395 CEPH_OSD_FLAG_LOCALIZE_READS = 0x2000, /* read from nearby replica, if any */
396 CEPH_OSD_FLAG_RWORDERED = 0x4000, /* order wrt concurrent reads */
397 CEPH_OSD_FLAG_IGNORE_CACHE = 0x8000, /* ignore cache logic */
398 CEPH_OSD_FLAG_SKIPRWLOCKS = 0x10000, /* skip rw locks */
399 CEPH_OSD_FLAG_IGNORE_OVERLAY = 0x20000, /* ignore pool overlay */
400 CEPH_OSD_FLAG_FLUSH = 0x40000, /* this is part of flush */
401 CEPH_OSD_FLAG_MAP_SNAP_CLONE = 0x80000, /* map snap direct to clone id */
402 CEPH_OSD_FLAG_ENFORCE_SNAPC = 0x100000, /* use snapc provided even if
403 pool uses pool snaps */
404 CEPH_OSD_FLAG_REDIRECTED = 0x200000, /* op has been redirected */
405 CEPH_OSD_FLAG_KNOWN_REDIR = 0x400000, /* redirect bit is authoritative */
406 CEPH_OSD_FLAG_FULL_TRY = 0x800000, /* try op despite full flag */
407 CEPH_OSD_FLAG_FULL_FORCE = 0x1000000, /* force op despite full flag */
408 };
409
410 enum {
411 CEPH_OSD_OP_FLAG_EXCL = 1, /* EXCL object create */
412 CEPH_OSD_OP_FLAG_FAILOK = 2, /* continue despite failure */
413 CEPH_OSD_OP_FLAG_FADVISE_RANDOM = 0x4, /* the op is random */
414 CEPH_OSD_OP_FLAG_FADVISE_SEQUENTIAL = 0x8, /* the op is sequential */
415 CEPH_OSD_OP_FLAG_FADVISE_WILLNEED = 0x10,/* data will be accessed in
416 the near future */
417 CEPH_OSD_OP_FLAG_FADVISE_DONTNEED = 0x20,/* data will not be accessed
418 in the near future */
419 CEPH_OSD_OP_FLAG_FADVISE_NOCACHE = 0x40,/* data will be accessed only
420 once by this client */
421 };
422
423 #define EOLDSNAPC ERESTART /* ORDERSNAP flag set; writer has old snapc*/
424 #define EBLACKLISTED ESHUTDOWN /* blacklisted */
425
426 /* xattr comparison */
427 enum {
428 CEPH_OSD_CMPXATTR_OP_NOP = 0,
429 CEPH_OSD_CMPXATTR_OP_EQ = 1,
430 CEPH_OSD_CMPXATTR_OP_NE = 2,
431 CEPH_OSD_CMPXATTR_OP_GT = 3,
432 CEPH_OSD_CMPXATTR_OP_GTE = 4,
433 CEPH_OSD_CMPXATTR_OP_LT = 5,
434 CEPH_OSD_CMPXATTR_OP_LTE = 6
435 };
436
437 enum {
438 CEPH_OSD_CMPXATTR_MODE_STRING = 1,
439 CEPH_OSD_CMPXATTR_MODE_U64 = 2
440 };
441
442 enum {
443 CEPH_OSD_COPY_FROM_FLAG_FLUSH = 1, /* part of a flush operation */
444 CEPH_OSD_COPY_FROM_FLAG_IGNORE_OVERLAY = 2, /* ignore pool overlay */
445 CEPH_OSD_COPY_FROM_FLAG_IGNORE_CACHE = 4, /* ignore osd cache logic */
446 CEPH_OSD_COPY_FROM_FLAG_MAP_SNAP_CLONE = 8, /* map snap direct to
447 * cloneid */
448 CEPH_OSD_COPY_FROM_FLAG_RWORDERED = 16, /* order with write */
449 };
450
451 enum {
452 CEPH_OSD_WATCH_OP_UNWATCH = 0,
453 CEPH_OSD_WATCH_OP_LEGACY_WATCH = 1,
454 /* note: use only ODD ids to prevent pre-giant code from
455 interpreting the op as UNWATCH */
456 CEPH_OSD_WATCH_OP_WATCH = 3,
457 CEPH_OSD_WATCH_OP_RECONNECT = 5,
458 CEPH_OSD_WATCH_OP_PING = 7,
459 };
460
461 const char *ceph_osd_watch_op_name(int o);
462
463 enum {
464 CEPH_OSD_BACKOFF_OP_BLOCK = 1,
465 CEPH_OSD_BACKOFF_OP_ACK_BLOCK = 2,
466 CEPH_OSD_BACKOFF_OP_UNBLOCK = 3,
467 };
468
469 /*
470 * an individual object operation. each may be accompanied by some data
471 * payload
472 */
473 struct ceph_osd_op {
474 __le16 op; /* CEPH_OSD_OP_* */
475 __le32 flags; /* CEPH_OSD_OP_FLAG_* */
476 union {
477 struct {
478 __le64 offset, length;
479 __le64 truncate_size;
480 __le32 truncate_seq;
481 } __attribute__ ((packed)) extent;
482 struct {
483 __le32 name_len;
484 __le32 value_len;
485 __u8 cmp_op; /* CEPH_OSD_CMPXATTR_OP_* */
486 __u8 cmp_mode; /* CEPH_OSD_CMPXATTR_MODE_* */
487 } __attribute__ ((packed)) xattr;
488 struct {
489 __u8 class_len;
490 __u8 method_len;
491 __u8 argc;
492 __le32 indata_len;
493 } __attribute__ ((packed)) cls;
494 struct {
495 __le64 cookie, count;
496 } __attribute__ ((packed)) pgls;
497 struct {
498 __le64 snapid;
499 } __attribute__ ((packed)) snap;
500 struct {
501 __le64 cookie;
502 __le64 ver; /* no longer used */
503 __u8 op; /* CEPH_OSD_WATCH_OP_* */
504 __le32 gen; /* registration generation */
505 } __attribute__ ((packed)) watch;
506 struct {
507 __le64 cookie;
508 } __attribute__ ((packed)) notify;
509 struct {
510 __le64 offset, length;
511 __le64 src_offset;
512 } __attribute__ ((packed)) clonerange;
513 struct {
514 __le64 expected_object_size;
515 __le64 expected_write_size;
516 } __attribute__ ((packed)) alloc_hint;
517 struct {
518 __le64 snapid;
519 __le64 src_version;
520 __u8 flags; /* CEPH_OSD_COPY_FROM_FLAG_* */
521 /*
522 * CEPH_OSD_OP_FLAG_FADVISE_*: fadvise flags
523 * for src object, flags for dest object are in
524 * ceph_osd_op::flags.
525 */
526 __le32 src_fadvise_flags;
527 } __attribute__ ((packed)) copy_from;
528 };
529 __le32 payload_len;
530 } __attribute__ ((packed));
531
532
533 #endif
534
