ceph/src/include/rados.h

   1 #ifndef CEPH_RADOS_H
   2 #define CEPH_RADOS_H
   3
   4 /*
   5  * Data types for the Ceph distributed object storage layer RADOS
   6  * (Reliable Autonomic Distributed Object Store).
   7  */
   8
   9 #include <string.h>
  10 #include "msgr.h"
  11
  12 /*
  13  * fs id
  14  */
  15 struct ceph_fsid {
  16         unsigned char fsid[16];
  17 };
  18
  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      16  /* max # osds in a single pg */
  55
  56 /*
  57  * placement group.
  58  * we encode this into one __le64.
  59  */
  60 struct ceph_pg {
  61         __le16 preferred; /* preferred primary osd */
  62         __le16 ps;        /* placement seed */
  63         __le32 pool;      /* object pool */
  64 } __attribute__ ((packed));
  65
  66 /*
  67  * pg pool types
  68  *
  69  * NOTE: These map 1:1 on to the pg_pool_t::TYPE_* values.  They are
  70  * duplicated here only for CrushCompiler's benefit.
  71  */
  72 #define CEPH_PG_TYPE_REPLICATED 1
  73 /* #define CEPH_PG_TYPE_RAID4   2   never implemented */
  74 #define CEPH_PG_TYPE_ERASURE 3
  75
  76 /*
  77  * stable_mod func is used to control number of placement groups.
  78  * similar to straight-up modulo, but produces a stable mapping as b
  79  * increases over time.  b is the number of bins, and bmask is the
  80  * containing power of 2 minus 1.
  81  *
  82  * b <= bmask and bmask=(2**n)-1
  83  * e.g., b=12 -> bmask=15, b=123 -> bmask=127
  84  */
  85 static inline int ceph_stable_mod(int x, int b, int bmask)
  86 {
  87         if ((x & bmask) < b)
  88                 return x & bmask;
  89         else
  90                 return x & (bmask >> 1);
  91 }
  92
  93 /*
  94  * object layout - how a given object should be stored.
  95  */
  96 struct ceph_object_layout {
  97         struct ceph_pg ol_pgid;   /* raw pg, with _full_ ps precision. */
  98         __le32 ol_stripe_unit;    /* for per-object parity, if any */
  99 } __attribute__ ((packed));
 100
 101 /*
 102  * compound epoch+version, used by storage layer to serialize mutations
 103  */
 104 struct ceph_eversion {
 105         __le32 epoch;
 106         __le64 version;
 107 } __attribute__ ((packed));
 108
 109 /*
 110  * osd map bits
 111  */
 112
 113 /* status bits */
 114 #define CEPH_OSD_EXISTS       (1<<0)
 115 #define CEPH_OSD_UP           (1<<1)
 116 #define CEPH_OSD_AUTOOUT      (1<<2)  /* osd was automatically marked out */
 117 #define CEPH_OSD_NEW          (1<<3)  /* osd is new, never marked in */
 118 #define CEPH_OSD_FULL         (1<<4)  /* osd is at or above full threshold */
 119 #define CEPH_OSD_NEARFULL     (1<<5)  /* osd is at or above nearfull threshold */
 120 #define CEPH_OSD_BACKFILLFULL (1<<6)  /* osd is at or above backfillfull threshold */
 121 #define CEPH_OSD_DESTROYED    (1<<7)  /* osd has been destroyed */
 122 #define CEPH_OSD_NOUP         (1<<8)  /* osd can not be marked up */
 123 #define CEPH_OSD_NODOWN       (1<<9)  /* osd can not be marked down */
 124 #define CEPH_OSD_NOIN         (1<<10) /* osd can not be marked in */
 125 #define CEPH_OSD_NOOUT        (1<<11) /* osd can not be marked out */
 126
 127 extern const char *ceph_osd_state_name(int s);
 128
 129 /* osd weights.  fixed point value: 0x10000 == 1.0 ("in"), 0 == "out" */
 130 #define CEPH_OSD_IN  0x10000
 131 #define CEPH_OSD_OUT 0
 132
 133 #define CEPH_OSD_MAX_PRIMARY_AFFINITY 0x10000
 134 #define CEPH_OSD_DEFAULT_PRIMARY_AFFINITY 0x10000
 135
 136
 137 /*
 138  * osd map flag bits
 139  */
 140 #define CEPH_OSDMAP_NEARFULL         (1<<0)  /* sync writes (near ENOSPC) */
 141 #define CEPH_OSDMAP_FULL             (1<<1)  /* no data writes (ENOSPC) */
 142 #define CEPH_OSDMAP_PAUSERD          (1<<2)  /* pause all reads */
 143 #define CEPH_OSDMAP_PAUSEWR          (1<<3)  /* pause all writes */
 144 #define CEPH_OSDMAP_PAUSEREC         (1<<4)  /* pause recovery */
 145 #define CEPH_OSDMAP_NOUP             (1<<5)  /* block osd boot */
 146 #define CEPH_OSDMAP_NODOWN           (1<<6)  /* block osd mark-down/failure */
 147 #define CEPH_OSDMAP_NOOUT            (1<<7)  /* block osd auto mark-out */
 148 #define CEPH_OSDMAP_NOIN             (1<<8)  /* block osd auto mark-in */
 149 #define CEPH_OSDMAP_NOBACKFILL       (1<<9)  /* block osd backfill */
 150 #define CEPH_OSDMAP_NORECOVER        (1<<10) /* block osd recovery and backfill */
 151 #define CEPH_OSDMAP_NOSCRUB          (1<<11) /* block periodic scrub */
 152 #define CEPH_OSDMAP_NODEEP_SCRUB     (1<<12) /* block periodic deep-scrub */
 153 #define CEPH_OSDMAP_NOTIERAGENT      (1<<13) /* disable tiering agent */
 154 #define CEPH_OSDMAP_NOREBALANCE      (1<<14) /* block osd backfill unless pg is degraded */
 155 #define CEPH_OSDMAP_SORTBITWISE      (1<<15) /* use bitwise hobject_t sort */
 156 #define CEPH_OSDMAP_REQUIRE_JEWEL    (1<<16) /* require jewel for booting osds */
 157 #define CEPH_OSDMAP_REQUIRE_KRAKEN   (1<<17) /* require kraken for booting osds */
 158 #define CEPH_OSDMAP_REQUIRE_LUMINOUS (1<<18) /* require l for booting osds */
 159 #define CEPH_OSDMAP_RECOVERY_DELETES (1<<19) /* deletes performed during recovery instead of peering */
 160 #define CEPH_OSDMAP_PURGED_SNAPDIRS  (1<<20) /* osds have converted snapsets */
 161 #define CEPH_OSDMAP_NOSNAPTRIM       (1<<21) /* disable snap trimming */
 162 #define CEPH_OSDMAP_PGLOG_HARDLIMIT  (1<<22) /* put a hard limit on pg log length */
 163
 164 /* these are hidden in 'ceph status' view */
 165 #define CEPH_OSDMAP_SEMIHIDDEN_FLAGS (CEPH_OSDMAP_REQUIRE_JEWEL|        \
 166                                       CEPH_OSDMAP_REQUIRE_KRAKEN |      \
 167                                       CEPH_OSDMAP_REQUIRE_LUMINOUS |    \
 168                                       CEPH_OSDMAP_RECOVERY_DELETES |    \
 169                                       CEPH_OSDMAP_SORTBITWISE |         \
 170                                       CEPH_OSDMAP_PURGED_SNAPDIRS |     \
 171                                       CEPH_OSDMAP_PGLOG_HARDLIMIT)
 172 #define CEPH_OSDMAP_LEGACY_REQUIRE_FLAGS (CEPH_OSDMAP_REQUIRE_JEWEL |   \
 173                                           CEPH_OSDMAP_REQUIRE_KRAKEN |  \
 174                                           CEPH_OSDMAP_REQUIRE_LUMINOUS)
 175
 176 /*
 177  * major ceph release numbers
 178  */
 179 #define CEPH_RELEASE_ARGONAUT    1
 180 #define CEPH_RELEASE_BOBTAIL     2
 181 #define CEPH_RELEASE_CUTTLEFISH  3
 182 #define CEPH_RELEASE_DUMPLING    4
 183 #define CEPH_RELEASE_EMPEROR     5
 184 #define CEPH_RELEASE_FIREFLY     6
 185 #define CEPH_RELEASE_GIANT       7
 186 #define CEPH_RELEASE_HAMMER      8
 187 #define CEPH_RELEASE_INFERNALIS  9
 188 #define CEPH_RELEASE_JEWEL      10
 189 #define CEPH_RELEASE_KRAKEN     11
 190 #define CEPH_RELEASE_LUMINOUS   12
 191 #define CEPH_RELEASE_MIMIC      13
 192 #define CEPH_RELEASE_NAUTILUS   14
 193 #define CEPH_RELEASE_MAX        15  /* highest + 1 */
 194
 195 extern const char *ceph_release_name(int r);
 196 extern int ceph_release_from_name(const char *s);
 197 extern uint64_t ceph_release_features(int r);
 198 extern int ceph_release_from_features(uint64_t features);
 199
 200 /*
 201  * The error code to return when an OSD can't handle a write
 202  * because it is too large.
 203  */
 204 #define OSD_WRITETOOBIG EMSGSIZE
 205
 206 /*
 207  * osd ops
 208  *
 209  * WARNING: do not use these op codes directly.  Use the helpers
 210  * defined below instead.  In certain cases, op code behavior was
 211  * redefined, resulting in special-cases in the helpers.
 212  */
 213 #define CEPH_OSD_OP_MODE       0xf000
 214 #define CEPH_OSD_OP_MODE_RD    0x1000
 215 #define CEPH_OSD_OP_MODE_WR    0x2000
 216 #define CEPH_OSD_OP_MODE_RMW   0x3000
 217 #define CEPH_OSD_OP_MODE_SUB   0x4000
 218 #define CEPH_OSD_OP_MODE_CACHE 0x8000
 219
 220 #define CEPH_OSD_OP_TYPE       0x0f00
 221 #define CEPH_OSD_OP_TYPE_DATA  0x0200
 222 #define CEPH_OSD_OP_TYPE_ATTR  0x0300
 223 #define CEPH_OSD_OP_TYPE_EXEC  0x0400
 224 #define CEPH_OSD_OP_TYPE_PG    0x0500
 225 //      LEAVE UNUSED           0x0600 used to be multiobject ops
 226
 227 #define __CEPH_OSD_OP1(mode, nr) \
 228         (CEPH_OSD_OP_MODE_##mode | (nr))
 229
 230 #define __CEPH_OSD_OP(mode, type, nr) \
 231         (CEPH_OSD_OP_MODE_##mode | CEPH_OSD_OP_TYPE_##type | (nr))
 232
 233 #define __CEPH_FORALL_OSD_OPS(f)                                            \
 234         /** data **/                                                        \
 235         /* read */                                                          \
 236         f(READ,         __CEPH_OSD_OP(RD, DATA, 1),     "read")             \
 237         f(STAT,         __CEPH_OSD_OP(RD, DATA, 2),     "stat")             \
 238         f(MAPEXT,       __CEPH_OSD_OP(RD, DATA, 3),     "mapext")           \
 239         f(CHECKSUM,     __CEPH_OSD_OP(RD, DATA, 31),    "checksum")         \
 240                                                                             \
 241         /* fancy read */                                                    \
 242         f(MASKTRUNC,    __CEPH_OSD_OP(RD, DATA, 4),     "masktrunc")        \
 243         f(SPARSE_READ,  __CEPH_OSD_OP(RD, DATA, 5),     "sparse-read")      \
 244                                                                             \
 245         f(NOTIFY,       __CEPH_OSD_OP(RD, DATA, 6),     "notify")           \
 246         f(NOTIFY_ACK,   __CEPH_OSD_OP(RD, DATA, 7),     "notify-ack")       \
 247                                                                             \
 248         /* versioning */                                                    \
 249         f(ASSERT_VER,   __CEPH_OSD_OP(RD, DATA, 8),     "assert-version")   \
 250                                                                             \
 251         f(LIST_WATCHERS, __CEPH_OSD_OP(RD, DATA, 9),    "list-watchers")    \
 252                                                                             \
 253         f(LIST_SNAPS,   __CEPH_OSD_OP(RD, DATA, 10),    "list-snaps")       \
 254                                                                             \
 255         /* sync */                                                          \
 256         f(SYNC_READ,    __CEPH_OSD_OP(RD, DATA, 11),    "sync_read")        \
 257                                                                             \
 258         /* write */                                                         \
 259         f(WRITE,        __CEPH_OSD_OP(WR, DATA, 1),     "write")            \
 260         f(WRITEFULL,    __CEPH_OSD_OP(WR, DATA, 2),     "writefull")        \
 261         f(TRUNCATE,     __CEPH_OSD_OP(WR, DATA, 3),     "truncate")         \
 262         f(ZERO,         __CEPH_OSD_OP(WR, DATA, 4),     "zero")             \
 263         f(DELETE,       __CEPH_OSD_OP(WR, DATA, 5),     "delete")           \
 264                                                                             \
 265         /* fancy write */                                                   \
 266         f(APPEND,       __CEPH_OSD_OP(WR, DATA, 6),     "append")           \
 267         f(STARTSYNC,    __CEPH_OSD_OP(WR, DATA, 7),     "startsync")        \
 268         f(SETTRUNC,     __CEPH_OSD_OP(WR, DATA, 8),     "settrunc")         \
 269         f(TRIMTRUNC,    __CEPH_OSD_OP(WR, DATA, 9),     "trimtrunc")        \
 270                                                                             \
 271         f(TMAPUP,       __CEPH_OSD_OP(RMW, DATA, 10),   "tmapup")           \
 272         f(TMAPPUT,      __CEPH_OSD_OP(WR, DATA, 11),    "tmapput")          \
 273         f(TMAPGET,      __CEPH_OSD_OP(RD, DATA, 12),    "tmapget")          \
 274                                                                             \
 275         f(CREATE,       __CEPH_OSD_OP(WR, DATA, 13),    "create")           \
 276         f(ROLLBACK,     __CEPH_OSD_OP(WR, DATA, 14),    "rollback")         \
 277                                                                             \
 278         f(WATCH,        __CEPH_OSD_OP(WR, DATA, 15),    "watch")            \
 279                                                                             \
 280         /* omap */                                                          \
 281         f(OMAPGETKEYS,  __CEPH_OSD_OP(RD, DATA, 17),    "omap-get-keys")    \
 282         f(OMAPGETVALS,  __CEPH_OSD_OP(RD, DATA, 18),    "omap-get-vals")    \
 283         f(OMAPGETHEADER, __CEPH_OSD_OP(RD, DATA, 19),   "omap-get-header")  \
 284         f(OMAPGETVALSBYKEYS, __CEPH_OSD_OP(RD, DATA, 20), "omap-get-vals-by-keys") \
 285         f(OMAPSETVALS,  __CEPH_OSD_OP(WR, DATA, 21),    "omap-set-vals")    \
 286         f(OMAPSETHEADER, __CEPH_OSD_OP(WR, DATA, 22),   "omap-set-header")  \
 287         f(OMAPCLEAR,    __CEPH_OSD_OP(WR, DATA, 23),    "omap-clear")       \
 288         f(OMAPRMKEYS,   __CEPH_OSD_OP(WR, DATA, 24),    "omap-rm-keys")     \
 289         f(OMAP_CMP,     __CEPH_OSD_OP(RD, DATA, 25),    "omap-cmp")         \
 290                                                                             \
 291         /* tiering */                                                       \
 292         f(COPY_FROM,    __CEPH_OSD_OP(WR, DATA, 26),    "copy-from")        \
 293         /* was copy-get-classic */                                      \
 294         f(UNDIRTY,      __CEPH_OSD_OP(WR, DATA, 28),    "undirty")          \
 295         f(ISDIRTY,      __CEPH_OSD_OP(RD, DATA, 29),    "isdirty")          \
 296         f(COPY_GET,     __CEPH_OSD_OP(RD, DATA, 30),    "copy-get")         \
 297         f(CACHE_FLUSH,  __CEPH_OSD_OP(CACHE, DATA, 31), "cache-flush")      \
 298         f(CACHE_EVICT,  __CEPH_OSD_OP(CACHE, DATA, 32), "cache-evict")      \
 299         f(CACHE_TRY_FLUSH, __CEPH_OSD_OP(CACHE, DATA, 33), "cache-try-flush") \
 300                                                                             \
 301         /* convert tmap to omap */                                          \
 302         f(TMAP2OMAP,    __CEPH_OSD_OP(RMW, DATA, 34),   "tmap2omap")        \
 303                                                                             \
 304         /* hints */                                                         \
 305         f(SETALLOCHINT, __CEPH_OSD_OP(WR, DATA, 35),    "set-alloc-hint")   \
 306                                                                             \
 307         /* cache pin/unpin */                                               \
 308         f(CACHE_PIN,    __CEPH_OSD_OP(WR, DATA, 36),    "cache-pin")        \
 309         f(CACHE_UNPIN,  __CEPH_OSD_OP(WR, DATA, 37),    "cache-unpin")      \
 310                                                                             \
 311         /* ESX/SCSI */                                                      \
 312         f(WRITESAME,    __CEPH_OSD_OP(WR, DATA, 38),    "write-same")       \
 313         f(CMPEXT,       __CEPH_OSD_OP(RD, DATA, 32),    "cmpext")           \
 314                                                                             \
 315         /* Extensible */                                                    \
 316         f(SET_REDIRECT, __CEPH_OSD_OP(WR, DATA, 39),    "set-redirect")     \
 317         f(SET_CHUNK,    __CEPH_OSD_OP(WR, DATA, 40),    "set-chunk")        \
 318         f(TIER_PROMOTE, __CEPH_OSD_OP(WR, DATA, 41),    "tier-promote")     \
 319         f(UNSET_MANIFEST, __CEPH_OSD_OP(WR, DATA, 42),  "unset-manifest")   \
 320                                                                             \
 321         /** attrs **/                                                       \
 322         /* read */                                                          \
 323         f(GETXATTR,     __CEPH_OSD_OP(RD, ATTR, 1),     "getxattr")         \
 324         f(GETXATTRS,    __CEPH_OSD_OP(RD, ATTR, 2),     "getxattrs")        \
 325         f(CMPXATTR,     __CEPH_OSD_OP(RD, ATTR, 3),     "cmpxattr")         \
 326                                                                             \
 327         /* write */                                                         \
 328         f(SETXATTR,     __CEPH_OSD_OP(WR, ATTR, 1),     "setxattr")         \
 329         f(SETXATTRS,    __CEPH_OSD_OP(WR, ATTR, 2),     "setxattrs")        \
 330         f(RESETXATTRS,  __CEPH_OSD_OP(WR, ATTR, 3),     "resetxattrs")      \
 331         f(RMXATTR,      __CEPH_OSD_OP(WR, ATTR, 4),     "rmxattr")          \
 332                                                                             \
 333         /** subop **/                                                       \
 334         f(PULL,         __CEPH_OSD_OP1(SUB, 1),         "pull")             \
 335         f(PUSH,         __CEPH_OSD_OP1(SUB, 2),         "push")             \
 336         f(BALANCEREADS, __CEPH_OSD_OP1(SUB, 3),         "balance-reads")    \
 337         f(UNBALANCEREADS, __CEPH_OSD_OP1(SUB, 4),       "unbalance-reads")  \
 338         f(SCRUB,        __CEPH_OSD_OP1(SUB, 5),         "scrub")            \
 339         f(SCRUB_RESERVE, __CEPH_OSD_OP1(SUB, 6),        "scrub-reserve")    \
 340         f(SCRUB_UNRESERVE, __CEPH_OSD_OP1(SUB, 7),      "scrub-unreserve")  \
 341         /* 8 used to be scrub-stop */                                   \
 342         f(SCRUB_MAP,    __CEPH_OSD_OP1(SUB, 9),         "scrub-map")        \
 343                                                                             \
 344         /** exec **/                                                        \
 345         /* note: the RD bit here is wrong; see special-case below in helper */ \
 346         f(CALL,         __CEPH_OSD_OP(RD, EXEC, 1),     "call")             \
 347                                                                             \
 348         /** pg **/                                                          \
 349         f(PGLS,         __CEPH_OSD_OP(RD, PG, 1),       "pgls")             \
 350         f(PGLS_FILTER,  __CEPH_OSD_OP(RD, PG, 2),       "pgls-filter")      \
 351         f(PG_HITSET_LS, __CEPH_OSD_OP(RD, PG, 3),       "pg-hitset-ls")     \
 352         f(PG_HITSET_GET, __CEPH_OSD_OP(RD, PG, 4),      "pg-hitset-get")    \
 353         f(PGNLS,        __CEPH_OSD_OP(RD, PG, 5),       "pgnls")            \
 354         f(PGNLS_FILTER, __CEPH_OSD_OP(RD, PG, 6),       "pgnls-filter")     \
 355         f(SCRUBLS, __CEPH_OSD_OP(RD, PG, 7), "scrubls")
 356
 357 enum {
 358 #define GENERATE_ENUM_ENTRY(op, opcode, str)    CEPH_OSD_OP_##op = (opcode),
 359 __CEPH_FORALL_OSD_OPS(GENERATE_ENUM_ENTRY)
 360 #undef GENERATE_ENUM_ENTRY
 361 };
 362
 363 static inline int ceph_osd_op_type_data(int op)
 364 {
 365         return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_DATA;
 366 }
 367 static inline int ceph_osd_op_type_attr(int op)
 368 {
 369         return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_ATTR;
 370 }
 371 static inline int ceph_osd_op_type_exec(int op)
 372 {
 373         return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_EXEC;
 374 }
 375 static inline int ceph_osd_op_type_pg(int op)
 376 {
 377         return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_PG;
 378 }
 379
 380 static inline int ceph_osd_op_mode_subop(int op)
 381 {
 382         return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_SUB;
 383 }
 384 static inline int ceph_osd_op_mode_read(int op)
 385 {
 386         return (op & CEPH_OSD_OP_MODE_RD) &&
 387                 op != CEPH_OSD_OP_CALL;
 388 }
 389 static inline int ceph_osd_op_mode_modify(int op)
 390 {
 391         return op & CEPH_OSD_OP_MODE_WR;
 392 }
 393 static inline int ceph_osd_op_mode_cache(int op)
 394 {
 395         return op & CEPH_OSD_OP_MODE_CACHE;
 396 }
 397 static inline int ceph_osd_op_uses_extent(int op)
 398 {
 399         switch(op) {
 400         case CEPH_OSD_OP_READ:
 401         case CEPH_OSD_OP_MAPEXT:
 402         case CEPH_OSD_OP_MASKTRUNC:
 403         case CEPH_OSD_OP_SPARSE_READ:
 404         case CEPH_OSD_OP_SYNC_READ:
 405         case CEPH_OSD_OP_WRITE:
 406         case CEPH_OSD_OP_WRITEFULL:
 407         case CEPH_OSD_OP_TRUNCATE:
 408         case CEPH_OSD_OP_ZERO:
 409         case CEPH_OSD_OP_APPEND:
 410         case CEPH_OSD_OP_TRIMTRUNC:
 411         case CEPH_OSD_OP_CMPEXT:
 412                 return true;
 413         default:
 414                 return false;
 415         }
 416 }
 417
 418 /*
 419  * note that the following tmap stuff is also defined in the ceph librados.h
 420  * and objclass.h. Any modification here needs to be updated there
 421  */
 422 #define CEPH_OSD_TMAP_HDR 'h'
 423 #define CEPH_OSD_TMAP_SET 's'
 424 #define CEPH_OSD_TMAP_CREATE 'c' /* create key */
 425 #define CEPH_OSD_TMAP_RM  'r'
 426 #define CEPH_OSD_TMAP_RMSLOPPY 'R'
 427
 428 extern const char *ceph_osd_op_name(int op);
 429
 430 /*
 431  * osd op flags
 432  *
 433  * An op may be READ, WRITE, or READ|WRITE.
 434  */
 435 enum {
 436         CEPH_OSD_FLAG_ACK =            0x0001,  /* want (or is) "ack" ack */
 437         CEPH_OSD_FLAG_ONNVRAM =        0x0002,  /* want (or is) "onnvram" ack */
 438         CEPH_OSD_FLAG_ONDISK =         0x0004,  /* want (or is) "ondisk" ack */
 439         CEPH_OSD_FLAG_RETRY =          0x0008,  /* resend attempt */
 440         CEPH_OSD_FLAG_READ =           0x0010,  /* op may read */
 441         CEPH_OSD_FLAG_WRITE =          0x0020,  /* op may write */
 442         CEPH_OSD_FLAG_ORDERSNAP =      0x0040,  /* EOLDSNAP if snapc is out of order */
 443         CEPH_OSD_FLAG_PEERSTAT_OLD =   0x0080,  /* DEPRECATED msg includes osd_peer_stat */
 444         CEPH_OSD_FLAG_BALANCE_READS =  0x0100,
 445         CEPH_OSD_FLAG_PARALLELEXEC =   0x0200,  /* execute op in parallel */
 446         CEPH_OSD_FLAG_PGOP =           0x0400,  /* pg op, no object */
 447         CEPH_OSD_FLAG_EXEC =           0x0800,  /* op may exec */
 448         CEPH_OSD_FLAG_EXEC_PUBLIC =    0x1000,  /* DEPRECATED op may exec (public) */
 449         CEPH_OSD_FLAG_LOCALIZE_READS = 0x2000,  /* read from nearby replica, if any */
 450         CEPH_OSD_FLAG_RWORDERED =      0x4000,  /* order wrt concurrent reads */
 451         CEPH_OSD_FLAG_IGNORE_CACHE =   0x8000,  /* ignore cache logic */
 452         CEPH_OSD_FLAG_SKIPRWLOCKS =   0x10000,  /* skip rw locks */
 453         CEPH_OSD_FLAG_IGNORE_OVERLAY =0x20000,  /* ignore pool overlay */
 454         CEPH_OSD_FLAG_FLUSH =         0x40000,  /* this is part of flush */
 455         CEPH_OSD_FLAG_MAP_SNAP_CLONE =0x80000,  /* map snap direct to clone id
 456                                                  */
 457         CEPH_OSD_FLAG_ENFORCE_SNAPC    =0x100000,  /* use snapc provided even if
 458                                                       pool uses pool snaps */
 459         CEPH_OSD_FLAG_REDIRECTED   = 0x200000,  /* op has been redirected */
 460         CEPH_OSD_FLAG_KNOWN_REDIR = 0x400000,  /* redirect bit is authoritative */
 461         CEPH_OSD_FLAG_FULL_TRY =    0x800000,  /* try op despite full flag */
 462         CEPH_OSD_FLAG_FULL_FORCE = 0x1000000,  /* force op despite full flag */
 463         CEPH_OSD_FLAG_IGNORE_REDIRECT = 0x2000000,  /* ignore redirection */
 464 };
 465
 466 enum {
 467         CEPH_OSD_OP_FLAG_EXCL = 0x1,      /* EXCL object create */
 468         CEPH_OSD_OP_FLAG_FAILOK = 0x2,    /* continue despite failure */
 469         CEPH_OSD_OP_FLAG_FADVISE_RANDOM     = 0x4, /* the op is random */
 470         CEPH_OSD_OP_FLAG_FADVISE_SEQUENTIAL = 0x8, /* the op is sequential */
 471         CEPH_OSD_OP_FLAG_FADVISE_WILLNEED   = 0x10,/* data will be accessed in the near future */
 472         CEPH_OSD_OP_FLAG_FADVISE_DONTNEED   = 0x20,/* data will not be accessed in the near future */
 473         CEPH_OSD_OP_FLAG_FADVISE_NOCACHE   = 0x40, /* data will be accessed only once by this client */
 474         CEPH_OSD_OP_FLAG_WITH_REFERENCE   = 0x80, /* need reference couting */
 475         CEPH_OSD_OP_FLAG_BYPASS_CLEAN_CACHE = 0x100, /* bypass ObjectStore cache, mainly for deep-scrub */
 476 };
 477
 478 #define EOLDSNAPC    85  /* ORDERSNAP flag set; writer has old snapc*/
 479 #define EBLACKLISTED 108 /* blacklisted */
 480
 481 /* xattr comparison */
 482 enum {
 483         CEPH_OSD_CMPXATTR_OP_EQ  = 1,
 484         CEPH_OSD_CMPXATTR_OP_NE  = 2,
 485         CEPH_OSD_CMPXATTR_OP_GT  = 3,
 486         CEPH_OSD_CMPXATTR_OP_GTE = 4,
 487         CEPH_OSD_CMPXATTR_OP_LT  = 5,
 488         CEPH_OSD_CMPXATTR_OP_LTE = 6
 489 };
 490
 491 enum {
 492         CEPH_OSD_CMPXATTR_MODE_STRING = 1,
 493         CEPH_OSD_CMPXATTR_MODE_U64    = 2
 494 };
 495
 496 enum {
 497         CEPH_OSD_COPY_FROM_FLAG_FLUSH = 1,     /* part of a flush operation */
 498         CEPH_OSD_COPY_FROM_FLAG_IGNORE_OVERLAY = 2,  /* ignore pool overlay */
 499         CEPH_OSD_COPY_FROM_FLAG_IGNORE_CACHE = 4, /* ignore osd cache logic */
 500         CEPH_OSD_COPY_FROM_FLAG_MAP_SNAP_CLONE = 8, /* map snap direct to
 501                                                      * cloneid */
 502         CEPH_OSD_COPY_FROM_FLAG_RWORDERED = 16, /* order with write */
 503 };
 504
 505 enum {
 506         CEPH_OSD_TMAP2OMAP_NULLOK = 1,
 507 };
 508
 509 enum {
 510         CEPH_OSD_WATCH_OP_UNWATCH = 0,
 511         CEPH_OSD_WATCH_OP_LEGACY_WATCH = 1,
 512         /* note: use only ODD ids to prevent pre-giant code from
 513            interpreting the op as UNWATCH */
 514         CEPH_OSD_WATCH_OP_WATCH = 3,
 515         CEPH_OSD_WATCH_OP_RECONNECT = 5,
 516         CEPH_OSD_WATCH_OP_PING = 7,
 517 };
 518
 519 enum {
 520         CEPH_OSD_CHECKSUM_OP_TYPE_XXHASH32 = 0,
 521         CEPH_OSD_CHECKSUM_OP_TYPE_XXHASH64 = 1,
 522         CEPH_OSD_CHECKSUM_OP_TYPE_CRC32C   = 2
 523 };
 524
 525 const char *ceph_osd_watch_op_name(int o);
 526
 527 enum {
 528         CEPH_OSD_ALLOC_HINT_FLAG_SEQUENTIAL_WRITE = 1,
 529         CEPH_OSD_ALLOC_HINT_FLAG_RANDOM_WRITE = 2,
 530         CEPH_OSD_ALLOC_HINT_FLAG_SEQUENTIAL_READ = 4,
 531         CEPH_OSD_ALLOC_HINT_FLAG_RANDOM_READ = 8,
 532         CEPH_OSD_ALLOC_HINT_FLAG_APPEND_ONLY = 16,
 533         CEPH_OSD_ALLOC_HINT_FLAG_IMMUTABLE = 32,
 534         CEPH_OSD_ALLOC_HINT_FLAG_SHORTLIVED = 64,
 535         CEPH_OSD_ALLOC_HINT_FLAG_LONGLIVED = 128,
 536         CEPH_OSD_ALLOC_HINT_FLAG_COMPRESSIBLE = 256,
 537         CEPH_OSD_ALLOC_HINT_FLAG_INCOMPRESSIBLE = 512,
 538 };
 539
 540 const char *ceph_osd_alloc_hint_flag_name(int f);
 541
 542 enum {
 543         CEPH_OSD_BACKOFF_OP_BLOCK = 1,
 544         CEPH_OSD_BACKOFF_OP_ACK_BLOCK = 2,
 545         CEPH_OSD_BACKOFF_OP_UNBLOCK = 3,
 546 };
 547
 548 const char *ceph_osd_backoff_op_name(int op);
 549
 550 /*
 551  * an individual object operation.  each may be accompanied by some data
 552  * payload
 553  */
 554 struct ceph_osd_op {
 555         __le16 op;           /* CEPH_OSD_OP_* */
 556         __le32 flags;        /* CEPH_OSD_OP_FLAG_* */
 557         union {
 558                 struct {
 559                         __le64 offset, length;
 560                         __le64 truncate_size;
 561                         __le32 truncate_seq;
 562                 } __attribute__ ((packed)) extent;
 563                 struct {
 564                         __le32 name_len;
 565                         __le32 value_len;
 566                         __u8 cmp_op;       /* CEPH_OSD_CMPXATTR_OP_* */
 567                         __u8 cmp_mode;     /* CEPH_OSD_CMPXATTR_MODE_* */
 568                 } __attribute__ ((packed)) xattr;
 569                 struct {
 570                         __u8 class_len;
 571                         __u8 method_len;
 572                         __u8 argc;
 573                         __le32 indata_len;
 574                 } __attribute__ ((packed)) cls;
 575                 struct {
 576                         __le64 count;
 577                         __le32 start_epoch; /* for the pgls sequence */
 578                 } __attribute__ ((packed)) pgls;
 579                 struct {
 580                         __le64 snapid;
 581                 } __attribute__ ((packed)) snap;
 582                 struct {
 583                         __le64 cookie;
 584                         __le64 ver;     /* no longer used */
 585                         __u8 op;        /* CEPH_OSD_WATCH_OP_* */
 586                         __u32 gen;      /* registration generation */
 587                         __u32 timeout; /* connection timeout */
 588                 } __attribute__ ((packed)) watch;
 589                 struct {
 590                         __le64 cookie;
 591                 } __attribute__ ((packed)) notify;
 592                 struct {
 593                         __le64 unused;
 594                         __le64 ver;
 595                 } __attribute__ ((packed)) assert_ver;
 596                 struct {
 597                         __le64 offset, length;
 598                         __le64 src_offset;
 599                 } __attribute__ ((packed)) clonerange;
 600                 struct {
 601                         __le64 max;     /* max data in reply */
 602                 } __attribute__ ((packed)) copy_get;
 603                 struct {
 604                         __le64 snapid;
 605                         __le64 src_version;
 606                         __u8 flags; /* CEPH_OSD_COPY_FROM_FLAG_* */
 607                         /*
 608                          * CEPH_OSD_OP_FLAG_FADVISE_*: fadvise flags
 609                          * for src object, flags for dest object are in
 610                          * ceph_osd_op::flags.
 611                          */
 612                         __le32 src_fadvise_flags;
 613                 } __attribute__ ((packed)) copy_from;
 614                 struct {
 615                         struct ceph_timespec stamp;
 616                 } __attribute__ ((packed)) hit_set_get;
 617                 struct {
 618                         __u8 flags;
 619                 } __attribute__ ((packed)) tmap2omap;
 620                 struct {
 621                         __le64 expected_object_size;
 622                         __le64 expected_write_size;
 623                         __le32 flags;  /* CEPH_OSD_OP_ALLOC_HINT_FLAG_* */
 624                 } __attribute__ ((packed)) alloc_hint;
 625                 struct {
 626                         __le64 offset;
 627                         __le64 length;
 628                         __le64 data_length;
 629                 } __attribute__ ((packed)) writesame;
 630                 struct {
 631                         __le64 offset;
 632                         __le64 length;
 633                         __le32 chunk_size;
 634                         __u8 type;              /* CEPH_OSD_CHECKSUM_OP_TYPE_* */
 635                 } __attribute__ ((packed)) checksum;
 636         };
 637         __le32 payload_len;
 638 } __attribute__ ((packed));
 639
 640 /*
 641  * Check the compatibility of struct ceph_osd_op
 642  *  (2+4+(2*8+8+4)+4) = (sizeof(ceph_osd_op::op) +
 643  *                     sizeof(ceph_osd_op::flags) +
 644  *                     sizeof(ceph_osd_op::extent) +
 645  *                     sizeof(ceph_osd_op::payload_len))
 646  */
 647 #ifdef __cplusplus
 648 static_assert(sizeof(ceph_osd_op) == (2+4+(2*8+8+4)+4),
 649               "sizeof(ceph_osd_op) breaks the compatibility");
 650 #endif
 651
 652 struct ceph_osd_reply_head {
 653         __le32 client_inc;                /* client incarnation */
 654         __le32 flags;
 655         struct ceph_object_layout layout;
 656         __le32 osdmap_epoch;
 657         struct ceph_eversion reassert_version; /* for replaying uncommitted */
 658
 659         __le32 result;                    /* result code */
 660
 661         __le32 object_len;                /* length of object name */
 662         __le32 num_ops;
 663         struct ceph_osd_op ops[0];  /* ops[], object */
 664 } __attribute__ ((packed));
 665
 666
 667 #endif