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 "msgr.h"
  10
  11 /*
  12  * fs id
  13  */
  14 struct ceph_fsid {
  15         unsigned char fsid[16];
  16 };
  17
  18 static inline int ceph_fsid_compare(const struct ceph_fsid *a,
  19                                     const struct ceph_fsid *b)
  20 {
  21         return memcmp(a, b, sizeof(*a));
  22 }
  23
  24 /*
  25  * ino, object, etc.
  26  */
  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 */
  31
  32 struct ceph_timespec {
  33         __le32 tv_sec;
  34         __le32 tv_nsec;
  35 } __attribute__ ((packed));
  36
  37
  38 /*
  39  * object layout - how objects are mapped into PGs
  40  */
  41 #define CEPH_OBJECT_LAYOUT_HASH     1
  42 #define CEPH_OBJECT_LAYOUT_LINEAR   2
  43 #define CEPH_OBJECT_LAYOUT_HASHINO  3
  44
  45 /*
  46  * pg layout -- how PGs are mapped onto (sets of) OSDs
  47  */
  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
  52
  53 #define CEPH_PG_MAX_SIZE      16  /* max # osds in a single pg */
  54
  55 /*
  56  * placement group.
  57  * we encode this into one __le64.
  58  */
  59 struct ceph_pg {
  60         __le16 preferred; /* preferred primary osd */
  61         __le16 ps;        /* placement seed */
  62         __le32 pool;      /* object pool */
  63 } __attribute__ ((packed));
  64
  65 /*
  66  * pg pool types
  67  *
  68  * NOTE: These map 1:1 on to the pg_pool_t::TYPE_* values.  They are
  69  * duplicated here only for CrushCompiler's benefit.
  70  */
  71 #define CEPH_PG_TYPE_REPLICATED 1
  72 /* #define CEPH_PG_TYPE_RAID4   2   never implemented */
  73 #define CEPH_PG_TYPE_ERASURE 3
  74
  75 /*
  76  * stable_mod func is used to control number of placement groups.
  77  * similar to straight-up modulo, but produces a stable mapping as b
  78  * increases over time.  b is the number of bins, and bmask is the
  79  * containing power of 2 minus 1.
  80  *
  81  * b <= bmask and bmask=(2**n)-1
  82  * e.g., b=12 -> bmask=15, b=123 -> bmask=127
  83  */
  84 static inline int ceph_stable_mod(int x, int b, int bmask)
  85 {
  86         if ((x & bmask) < b)
  87                 return x & bmask;
  88         else
  89                 return x & (bmask >> 1);
  90 }
  91
  92 /*
  93  * object layout - how a given object should be stored.
  94  */
  95 struct ceph_object_layout {
  96         struct ceph_pg ol_pgid;   /* raw pg, with _full_ ps precision. */
  97         __le32 ol_stripe_unit;    /* for per-object parity, if any */
  98 } __attribute__ ((packed));
  99
 100 /*
 101  * compound epoch+version, used by storage layer to serialize mutations
 102  */
 103 struct ceph_eversion {
 104         __le32 epoch;
 105         __le64 version;
 106 } __attribute__ ((packed));
 107
 108 /*
 109  * osd map bits
 110  */
 111
 112 /* status bits */
 113 #define CEPH_OSD_EXISTS  (1<<0)
 114 #define CEPH_OSD_UP      (1<<1)
 115 #define CEPH_OSD_AUTOOUT (1<<2)  /* osd was automatically marked out */
 116 #define CEPH_OSD_NEW     (1<<3)  /* osd is new, never marked in */
 117 #define CEPH_OSD_FULL    (1<<4)  /* osd is at or above full threshold */
 118 #define CEPH_OSD_NEARFULL (1<<5) /* osd is at or above nearfull threshold */
 119 #define CEPH_OSD_BACKFILLFULL (1<<6) /* osd is at or above backfillfull threshold */
 120
 121 extern const char *ceph_osd_state_name(int s);
 122
 123 /* osd weights.  fixed point value: 0x10000 == 1.0 ("in"), 0 == "out" */
 124 #define CEPH_OSD_IN  0x10000
 125 #define CEPH_OSD_OUT 0
 126
 127 #define CEPH_OSD_MAX_PRIMARY_AFFINITY 0x10000
 128 #define CEPH_OSD_DEFAULT_PRIMARY_AFFINITY 0x10000
 129
 130
 131 /*
 132  * osd map flag bits
 133  */
 134 #define CEPH_OSDMAP_NEARFULL (1<<0)  /* sync writes (near ENOSPC) */
 135 #define CEPH_OSDMAP_FULL     (1<<1)  /* no data writes (ENOSPC) */
 136 #define CEPH_OSDMAP_PAUSERD  (1<<2)  /* pause all reads */
 137 #define CEPH_OSDMAP_PAUSEWR  (1<<3)  /* pause all writes */
 138 #define CEPH_OSDMAP_PAUSEREC (1<<4)  /* pause recovery */
 139 #define CEPH_OSDMAP_NOUP     (1<<5)  /* block osd boot */
 140 #define CEPH_OSDMAP_NODOWN   (1<<6)  /* block osd mark-down/failure */
 141 #define CEPH_OSDMAP_NOOUT    (1<<7)  /* block osd auto mark-out */
 142 #define CEPH_OSDMAP_NOIN     (1<<8)  /* block osd auto mark-in */
 143 #define CEPH_OSDMAP_NOBACKFILL (1<<9) /* block osd backfill */
 144 #define CEPH_OSDMAP_NORECOVER (1<<10) /* block osd recovery and backfill */
 145 #define CEPH_OSDMAP_NOSCRUB  (1<<11) /* block periodic scrub */
 146 #define CEPH_OSDMAP_NODEEP_SCRUB (1<<12) /* block periodic deep-scrub */
 147 #define CEPH_OSDMAP_NOTIERAGENT (1<<13) /* disable tiering agent */
 148 #define CEPH_OSDMAP_NOREBALANCE (1<<14) /* block osd backfill unless pg is degraded */
 149 #define CEPH_OSDMAP_SORTBITWISE (1<<15) /* use bitwise hobject_t sort */
 150 #define CEPH_OSDMAP_REQUIRE_JEWEL (1<<16) /* require jewel for booting osds */
 151 #define CEPH_OSDMAP_REQUIRE_KRAKEN (1<<17) /* require kraken for booting osds */
 152 #define CEPH_OSDMAP_REQUIRE_LUMINOUS (1<<18) /* require l for booting osds */
 153
 154 /* these are hidden in 'ceph status' view */
 155 #define CEPH_OSDMAP_SEMIHIDDEN_FLAGS (CEPH_OSDMAP_REQUIRE_JEWEL|        \
 156                                       CEPH_OSDMAP_REQUIRE_KRAKEN |      \
 157                                       CEPH_OSDMAP_REQUIRE_LUMINOUS |    \
 158                                       CEPH_OSDMAP_SORTBITWISE)
 159
 160 /*
 161  * The error code to return when an OSD can't handle a write
 162  * because it is too large.
 163  */
 164 #define OSD_WRITETOOBIG EMSGSIZE
 165
 166 /*
 167  * osd ops
 168  *
 169  * WARNING: do not use these op codes directly.  Use the helpers
 170  * defined below instead.  In certain cases, op code behavior was
 171  * redefined, resulting in special-cases in the helpers.
 172  */
 173 #define CEPH_OSD_OP_MODE       0xf000
 174 #define CEPH_OSD_OP_MODE_RD    0x1000
 175 #define CEPH_OSD_OP_MODE_WR    0x2000
 176 #define CEPH_OSD_OP_MODE_RMW   0x3000
 177 #define CEPH_OSD_OP_MODE_SUB   0x4000
 178 #define CEPH_OSD_OP_MODE_CACHE 0x8000
 179
 180 #define CEPH_OSD_OP_TYPE       0x0f00
 181 #define CEPH_OSD_OP_TYPE_DATA  0x0200
 182 #define CEPH_OSD_OP_TYPE_ATTR  0x0300
 183 #define CEPH_OSD_OP_TYPE_EXEC  0x0400
 184 #define CEPH_OSD_OP_TYPE_PG    0x0500
 185 //      LEAVE UNUSED           0x0600 used to be multiobject ops
 186
 187 #define __CEPH_OSD_OP1(mode, nr) \
 188         (CEPH_OSD_OP_MODE_##mode | (nr))
 189
 190 #define __CEPH_OSD_OP(mode, type, nr) \
 191         (CEPH_OSD_OP_MODE_##mode | CEPH_OSD_OP_TYPE_##type | (nr))
 192
 193 #define __CEPH_FORALL_OSD_OPS(f)                                            \
 194         /** data **/                                                        \
 195         /* read */                                                          \
 196         f(READ,         __CEPH_OSD_OP(RD, DATA, 1),     "read")             \
 197         f(STAT,         __CEPH_OSD_OP(RD, DATA, 2),     "stat")             \
 198         f(MAPEXT,       __CEPH_OSD_OP(RD, DATA, 3),     "mapext")           \
 199         f(CHECKSUM,     __CEPH_OSD_OP(RD, DATA, 31),    "checksum")         \
 200                                                                             \
 201         /* fancy read */                                                    \
 202         f(MASKTRUNC,    __CEPH_OSD_OP(RD, DATA, 4),     "masktrunc")        \
 203         f(SPARSE_READ,  __CEPH_OSD_OP(RD, DATA, 5),     "sparse-read")      \
 204                                                                             \
 205         f(NOTIFY,       __CEPH_OSD_OP(RD, DATA, 6),     "notify")           \
 206         f(NOTIFY_ACK,   __CEPH_OSD_OP(RD, DATA, 7),     "notify-ack")       \
 207                                                                             \
 208         /* versioning */                                                    \
 209         f(ASSERT_VER,   __CEPH_OSD_OP(RD, DATA, 8),     "assert-version")   \
 210                                                                             \
 211         f(LIST_WATCHERS, __CEPH_OSD_OP(RD, DATA, 9),    "list-watchers")    \
 212                                                                             \
 213         f(LIST_SNAPS,   __CEPH_OSD_OP(RD, DATA, 10),    "list-snaps")       \
 214                                                                             \
 215         /* sync */                                                          \
 216         f(SYNC_READ,    __CEPH_OSD_OP(RD, DATA, 11),    "sync_read")        \
 217                                                                             \
 218         /* write */                                                         \
 219         f(WRITE,        __CEPH_OSD_OP(WR, DATA, 1),     "write")            \
 220         f(WRITEFULL,    __CEPH_OSD_OP(WR, DATA, 2),     "writefull")        \
 221         f(TRUNCATE,     __CEPH_OSD_OP(WR, DATA, 3),     "truncate")         \
 222         f(ZERO,         __CEPH_OSD_OP(WR, DATA, 4),     "zero")             \
 223         f(DELETE,       __CEPH_OSD_OP(WR, DATA, 5),     "delete")           \
 224                                                                             \
 225         /* fancy write */                                                   \
 226         f(APPEND,       __CEPH_OSD_OP(WR, DATA, 6),     "append")           \
 227         f(STARTSYNC,    __CEPH_OSD_OP(WR, DATA, 7),     "startsync")        \
 228         f(SETTRUNC,     __CEPH_OSD_OP(WR, DATA, 8),     "settrunc")         \
 229         f(TRIMTRUNC,    __CEPH_OSD_OP(WR, DATA, 9),     "trimtrunc")        \
 230                                                                             \
 231         f(TMAPUP,       __CEPH_OSD_OP(RMW, DATA, 10),   "tmapup")           \
 232         f(TMAPPUT,      __CEPH_OSD_OP(WR, DATA, 11),    "tmapput")          \
 233         f(TMAPGET,      __CEPH_OSD_OP(RD, DATA, 12),    "tmapget")          \
 234                                                                             \
 235         f(CREATE,       __CEPH_OSD_OP(WR, DATA, 13),    "create")           \
 236         f(ROLLBACK,     __CEPH_OSD_OP(WR, DATA, 14),    "rollback")         \
 237                                                                             \
 238         f(WATCH,        __CEPH_OSD_OP(WR, DATA, 15),    "watch")            \
 239                                                                             \
 240         /* omap */                                                          \
 241         f(OMAPGETKEYS,  __CEPH_OSD_OP(RD, DATA, 17),    "omap-get-keys")    \
 242         f(OMAPGETVALS,  __CEPH_OSD_OP(RD, DATA, 18),    "omap-get-vals")    \
 243         f(OMAPGETHEADER, __CEPH_OSD_OP(RD, DATA, 19),   "omap-get-header")  \
 244         f(OMAPGETVALSBYKEYS, __CEPH_OSD_OP(RD, DATA, 20), "omap-get-vals-by-keys") \
 245         f(OMAPSETVALS,  __CEPH_OSD_OP(WR, DATA, 21),    "omap-set-vals")    \
 246         f(OMAPSETHEADER, __CEPH_OSD_OP(WR, DATA, 22),   "omap-set-header")  \
 247         f(OMAPCLEAR,    __CEPH_OSD_OP(WR, DATA, 23),    "omap-clear")       \
 248         f(OMAPRMKEYS,   __CEPH_OSD_OP(WR, DATA, 24),    "omap-rm-keys")     \
 249         f(OMAP_CMP,     __CEPH_OSD_OP(RD, DATA, 25),    "omap-cmp")         \
 250                                                                             \
 251         /* tiering */                                                       \
 252         f(COPY_FROM,    __CEPH_OSD_OP(WR, DATA, 26),    "copy-from")        \
 253         /* was copy-get-classic */                                      \
 254         f(UNDIRTY,      __CEPH_OSD_OP(WR, DATA, 28),    "undirty")          \
 255         f(ISDIRTY,      __CEPH_OSD_OP(RD, DATA, 29),    "isdirty")          \
 256         f(COPY_GET,     __CEPH_OSD_OP(RD, DATA, 30),    "copy-get")         \
 257         f(CACHE_FLUSH,  __CEPH_OSD_OP(CACHE, DATA, 31), "cache-flush")      \
 258         f(CACHE_EVICT,  __CEPH_OSD_OP(CACHE, DATA, 32), "cache-evict")      \
 259         f(CACHE_TRY_FLUSH, __CEPH_OSD_OP(CACHE, DATA, 33), "cache-try-flush") \
 260                                                                             \
 261         /* convert tmap to omap */                                          \
 262         f(TMAP2OMAP,    __CEPH_OSD_OP(RMW, DATA, 34),   "tmap2omap")        \
 263                                                                             \
 264         /* hints */                                                         \
 265         f(SETALLOCHINT, __CEPH_OSD_OP(WR, DATA, 35),    "set-alloc-hint")   \
 266                                                                             \
 267         /* cache pin/unpin */                                               \
 268         f(CACHE_PIN,    __CEPH_OSD_OP(WR, DATA, 36),    "cache-pin")        \
 269         f(CACHE_UNPIN,  __CEPH_OSD_OP(WR, DATA, 37),    "cache-unpin")      \
 270                                                                             \
 271         /* ESX/SCSI */                                                      \
 272         f(WRITESAME,    __CEPH_OSD_OP(WR, DATA, 38),    "write-same")       \
 273         f(CMPEXT,       __CEPH_OSD_OP(RD, DATA, 32),    "cmpext")           \
 274                                                                             \
 275         /** attrs **/                                                       \
 276         /* read */                                                          \
 277         f(GETXATTR,     __CEPH_OSD_OP(RD, ATTR, 1),     "getxattr")         \
 278         f(GETXATTRS,    __CEPH_OSD_OP(RD, ATTR, 2),     "getxattrs")        \
 279         f(CMPXATTR,     __CEPH_OSD_OP(RD, ATTR, 3),     "cmpxattr")         \
 280                                                                             \
 281         /* write */                                                         \
 282         f(SETXATTR,     __CEPH_OSD_OP(WR, ATTR, 1),     "setxattr")         \
 283         f(SETXATTRS,    __CEPH_OSD_OP(WR, ATTR, 2),     "setxattrs")        \
 284         f(RESETXATTRS,  __CEPH_OSD_OP(WR, ATTR, 3),     "resetxattrs")      \
 285         f(RMXATTR,      __CEPH_OSD_OP(WR, ATTR, 4),     "rmxattr")          \
 286                                                                             \
 287         /** subop **/                                                       \
 288         f(PULL,         __CEPH_OSD_OP1(SUB, 1),         "pull")             \
 289         f(PUSH,         __CEPH_OSD_OP1(SUB, 2),         "push")             \
 290         f(BALANCEREADS, __CEPH_OSD_OP1(SUB, 3),         "balance-reads")    \
 291         f(UNBALANCEREADS, __CEPH_OSD_OP1(SUB, 4),       "unbalance-reads")  \
 292         f(SCRUB,        __CEPH_OSD_OP1(SUB, 5),         "scrub")            \
 293         f(SCRUB_RESERVE, __CEPH_OSD_OP1(SUB, 6),        "scrub-reserve")    \
 294         f(SCRUB_UNRESERVE, __CEPH_OSD_OP1(SUB, 7),      "scrub-unreserve")  \
 295         /* 8 used to be scrub-stop */                                   \
 296         f(SCRUB_MAP,    __CEPH_OSD_OP1(SUB, 9),         "scrub-map")        \
 297                                                                             \
 298         /** exec **/                                                        \
 299         /* note: the RD bit here is wrong; see special-case below in helper */ \
 300         f(CALL,         __CEPH_OSD_OP(RD, EXEC, 1),     "call")             \
 301                                                                             \
 302         /** pg **/                                                          \
 303         f(PGLS,         __CEPH_OSD_OP(RD, PG, 1),       "pgls")             \
 304         f(PGLS_FILTER,  __CEPH_OSD_OP(RD, PG, 2),       "pgls-filter")      \
 305         f(PG_HITSET_LS, __CEPH_OSD_OP(RD, PG, 3),       "pg-hitset-ls")     \
 306         f(PG_HITSET_GET, __CEPH_OSD_OP(RD, PG, 4),      "pg-hitset-get")    \
 307         f(PGNLS,        __CEPH_OSD_OP(RD, PG, 5),       "pgnls")            \
 308         f(PGNLS_FILTER, __CEPH_OSD_OP(RD, PG, 6),       "pgnls-filter")     \
 309         f(SCRUBLS, __CEPH_OSD_OP(RD, PG, 7), "scrubls")
 310
 311 enum {
 312 #define GENERATE_ENUM_ENTRY(op, opcode, str)    CEPH_OSD_OP_##op = (opcode),
 313 __CEPH_FORALL_OSD_OPS(GENERATE_ENUM_ENTRY)
 314 #undef GENERATE_ENUM_ENTRY
 315 };
 316
 317 static inline int ceph_osd_op_type_data(int op)
 318 {
 319         return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_DATA;
 320 }
 321 static inline int ceph_osd_op_type_attr(int op)
 322 {
 323         return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_ATTR;
 324 }
 325 static inline int ceph_osd_op_type_exec(int op)
 326 {
 327         return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_EXEC;
 328 }
 329 static inline int ceph_osd_op_type_pg(int op)
 330 {
 331         return (op & CEPH_OSD_OP_TYPE) == CEPH_OSD_OP_TYPE_PG;
 332 }
 333
 334 static inline int ceph_osd_op_mode_subop(int op)
 335 {
 336         return (op & CEPH_OSD_OP_MODE) == CEPH_OSD_OP_MODE_SUB;
 337 }
 338 static inline int ceph_osd_op_mode_read(int op)
 339 {
 340         return (op & CEPH_OSD_OP_MODE_RD) &&
 341                 op != CEPH_OSD_OP_CALL;
 342 }
 343 static inline int ceph_osd_op_mode_modify(int op)
 344 {
 345         return op & CEPH_OSD_OP_MODE_WR;
 346 }
 347 static inline int ceph_osd_op_mode_cache(int op)
 348 {
 349         return op & CEPH_OSD_OP_MODE_CACHE;
 350 }
 351 static inline int ceph_osd_op_uses_extent(int op)
 352 {
 353         switch(op) {
 354         case CEPH_OSD_OP_READ:
 355         case CEPH_OSD_OP_MAPEXT:
 356         case CEPH_OSD_OP_MASKTRUNC:
 357         case CEPH_OSD_OP_SPARSE_READ:
 358         case CEPH_OSD_OP_SYNC_READ:
 359         case CEPH_OSD_OP_WRITE:
 360         case CEPH_OSD_OP_WRITEFULL:
 361         case CEPH_OSD_OP_TRUNCATE:
 362         case CEPH_OSD_OP_ZERO:
 363         case CEPH_OSD_OP_APPEND:
 364         case CEPH_OSD_OP_TRIMTRUNC:
 365         case CEPH_OSD_OP_CMPEXT:
 366                 return true;
 367         default:
 368                 return false;
 369         }
 370 }
 371
 372 /*
 373  * note that the following tmap stuff is also defined in the ceph librados.h
 374  * and objclass.h. Any modification here needs to be updated there
 375  */
 376 #define CEPH_OSD_TMAP_HDR 'h'
 377 #define CEPH_OSD_TMAP_SET 's'
 378 #define CEPH_OSD_TMAP_CREATE 'c' /* create key */
 379 #define CEPH_OSD_TMAP_RM  'r'
 380 #define CEPH_OSD_TMAP_RMSLOPPY 'R'
 381
 382 extern const char *ceph_osd_op_name(int op);
 383
 384 /*
 385  * osd op flags
 386  *
 387  * An op may be READ, WRITE, or READ|WRITE.
 388  */
 389 enum {
 390         CEPH_OSD_FLAG_ACK =            0x0001,  /* want (or is) "ack" ack */
 391         CEPH_OSD_FLAG_ONNVRAM =        0x0002,  /* want (or is) "onnvram" ack */
 392         CEPH_OSD_FLAG_ONDISK =         0x0004,  /* want (or is) "ondisk" ack */
 393         CEPH_OSD_FLAG_RETRY =          0x0008,  /* resend attempt */
 394         CEPH_OSD_FLAG_READ =           0x0010,  /* op may read */
 395         CEPH_OSD_FLAG_WRITE =          0x0020,  /* op may write */
 396         CEPH_OSD_FLAG_ORDERSNAP =      0x0040,  /* EOLDSNAP if snapc is out of order */
 397         CEPH_OSD_FLAG_PEERSTAT_OLD =   0x0080,  /* DEPRECATED msg includes osd_peer_stat */
 398         CEPH_OSD_FLAG_BALANCE_READS =  0x0100,
 399         CEPH_OSD_FLAG_PARALLELEXEC =   0x0200,  /* execute op in parallel */
 400         CEPH_OSD_FLAG_PGOP =           0x0400,  /* pg op, no object */
 401         CEPH_OSD_FLAG_EXEC =           0x0800,  /* op may exec */
 402         CEPH_OSD_FLAG_EXEC_PUBLIC =    0x1000,  /* DEPRECATED op may exec (public) */
 403         CEPH_OSD_FLAG_LOCALIZE_READS = 0x2000,  /* read from nearby replica, if any */
 404         CEPH_OSD_FLAG_RWORDERED =      0x4000,  /* order wrt concurrent reads */
 405         CEPH_OSD_FLAG_IGNORE_CACHE =   0x8000,  /* ignore cache logic */
 406         CEPH_OSD_FLAG_SKIPRWLOCKS =   0x10000,  /* skip rw locks */
 407         CEPH_OSD_FLAG_IGNORE_OVERLAY =0x20000,  /* ignore pool overlay */
 408         CEPH_OSD_FLAG_FLUSH =         0x40000,  /* this is part of flush */
 409         CEPH_OSD_FLAG_MAP_SNAP_CLONE =0x80000,  /* map snap direct to clone id
 410                                                  */
 411         CEPH_OSD_FLAG_ENFORCE_SNAPC    =0x100000,  /* use snapc provided even if
 412                                                       pool uses pool snaps */
 413         CEPH_OSD_FLAG_REDIRECTED   = 0x200000,  /* op has been redirected */
 414         CEPH_OSD_FLAG_KNOWN_REDIR = 0x400000,  /* redirect bit is authoritative */
 415         CEPH_OSD_FLAG_FULL_TRY =    0x800000,  /* try op despite full flag */
 416         CEPH_OSD_FLAG_FULL_FORCE = 0x1000000,  /* force op despite full flag */
 417 };
 418
 419 enum {
 420         CEPH_OSD_OP_FLAG_EXCL = 0x1,      /* EXCL object create */
 421         CEPH_OSD_OP_FLAG_FAILOK = 0x2,    /* continue despite failure */
 422         CEPH_OSD_OP_FLAG_FADVISE_RANDOM     = 0x4, /* the op is random */
 423         CEPH_OSD_OP_FLAG_FADVISE_SEQUENTIAL = 0x8, /* the op is sequential */
 424         CEPH_OSD_OP_FLAG_FADVISE_WILLNEED   = 0x10,/* data will be accessed in the near future */
 425         CEPH_OSD_OP_FLAG_FADVISE_DONTNEED   = 0x20,/* data will not be accessed in the near future */
 426         CEPH_OSD_OP_FLAG_FADVISE_NOCACHE   = 0x40, /* data will be accessed only once by this client */
 427 };
 428
 429 #define EOLDSNAPC    85  /* ORDERSNAP flag set; writer has old snapc*/
 430 #define EBLACKLISTED 108 /* blacklisted */
 431
 432 /* xattr comparison */
 433 enum {
 434         CEPH_OSD_CMPXATTR_OP_EQ  = 1,
 435         CEPH_OSD_CMPXATTR_OP_NE  = 2,
 436         CEPH_OSD_CMPXATTR_OP_GT  = 3,
 437         CEPH_OSD_CMPXATTR_OP_GTE = 4,
 438         CEPH_OSD_CMPXATTR_OP_LT  = 5,
 439         CEPH_OSD_CMPXATTR_OP_LTE = 6
 440 };
 441
 442 enum {
 443         CEPH_OSD_CMPXATTR_MODE_STRING = 1,
 444         CEPH_OSD_CMPXATTR_MODE_U64    = 2
 445 };
 446
 447 enum {
 448         CEPH_OSD_COPY_FROM_FLAG_FLUSH = 1,     /* part of a flush operation */
 449         CEPH_OSD_COPY_FROM_FLAG_IGNORE_OVERLAY = 2,  /* ignore pool overlay */
 450         CEPH_OSD_COPY_FROM_FLAG_IGNORE_CACHE = 4, /* ignore osd cache logic */
 451         CEPH_OSD_COPY_FROM_FLAG_MAP_SNAP_CLONE = 8, /* map snap direct to
 452                                                      * cloneid */
 453         CEPH_OSD_COPY_FROM_FLAG_RWORDERED = 16, /* order with write */
 454 };
 455
 456 enum {
 457         CEPH_OSD_TMAP2OMAP_NULLOK = 1,
 458 };
 459
 460 enum {
 461         CEPH_OSD_WATCH_OP_UNWATCH = 0,
 462         CEPH_OSD_WATCH_OP_LEGACY_WATCH = 1,
 463         /* note: use only ODD ids to prevent pre-giant code from
 464            interpreting the op as UNWATCH */
 465         CEPH_OSD_WATCH_OP_WATCH = 3,
 466         CEPH_OSD_WATCH_OP_RECONNECT = 5,
 467         CEPH_OSD_WATCH_OP_PING = 7,
 468 };
 469
 470 enum {
 471         CEPH_OSD_CHECKSUM_OP_TYPE_XXHASH32 = 0,
 472         CEPH_OSD_CHECKSUM_OP_TYPE_XXHASH64 = 1,
 473         CEPH_OSD_CHECKSUM_OP_TYPE_CRC32C   = 2
 474 };
 475
 476 const char *ceph_osd_watch_op_name(int o);
 477
 478 enum {
 479         CEPH_OSD_ALLOC_HINT_FLAG_SEQUENTIAL_WRITE = 1,
 480         CEPH_OSD_ALLOC_HINT_FLAG_RANDOM_WRITE = 2,
 481         CEPH_OSD_ALLOC_HINT_FLAG_SEQUENTIAL_READ = 4,
 482         CEPH_OSD_ALLOC_HINT_FLAG_RANDOM_READ = 8,
 483         CEPH_OSD_ALLOC_HINT_FLAG_APPEND_ONLY = 16,
 484         CEPH_OSD_ALLOC_HINT_FLAG_IMMUTABLE = 32,
 485         CEPH_OSD_ALLOC_HINT_FLAG_SHORTLIVED = 64,
 486         CEPH_OSD_ALLOC_HINT_FLAG_LONGLIVED = 128,
 487         CEPH_OSD_ALLOC_HINT_FLAG_COMPRESSIBLE = 256,
 488         CEPH_OSD_ALLOC_HINT_FLAG_INCOMPRESSIBLE = 512,
 489 };
 490
 491 const char *ceph_osd_alloc_hint_flag_name(int f);
 492
 493 enum {
 494         CEPH_OSD_BACKOFF_OP_BLOCK = 1,
 495         CEPH_OSD_BACKOFF_OP_ACK_BLOCK = 2,
 496         CEPH_OSD_BACKOFF_OP_UNBLOCK = 3,
 497 };
 498
 499 const char *ceph_osd_backoff_op_name(int op);
 500
 501 /*
 502  * an individual object operation.  each may be accompanied by some data
 503  * payload
 504  */
 505 struct ceph_osd_op {
 506         __le16 op;           /* CEPH_OSD_OP_* */
 507         __le32 flags;        /* CEPH_OSD_OP_FLAG_* */
 508         union {
 509                 struct {
 510                         __le64 offset, length;
 511                         __le64 truncate_size;
 512                         __le32 truncate_seq;
 513                 } __attribute__ ((packed)) extent;
 514                 struct {
 515                         __le32 name_len;
 516                         __le32 value_len;
 517                         __u8 cmp_op;       /* CEPH_OSD_CMPXATTR_OP_* */
 518                         __u8 cmp_mode;     /* CEPH_OSD_CMPXATTR_MODE_* */
 519                 } __attribute__ ((packed)) xattr;
 520                 struct {
 521                         __u8 class_len;
 522                         __u8 method_len;
 523                         __u8 argc;
 524                         __le32 indata_len;
 525                 } __attribute__ ((packed)) cls;
 526                 struct {
 527                         __le64 count;
 528                         __le32 start_epoch; /* for the pgls sequence */
 529                 } __attribute__ ((packed)) pgls;
 530                 struct {
 531                         __le64 snapid;
 532                 } __attribute__ ((packed)) snap;
 533                 struct {
 534                         __le64 cookie;
 535                         __le64 ver;     /* no longer used */
 536                         __u8 op;        /* CEPH_OSD_WATCH_OP_* */
 537                         __u32 gen;      /* registration generation */
 538                         __u32 timeout; /* connection timeout */
 539                 } __attribute__ ((packed)) watch;
 540                 struct {
 541                         __le64 cookie;
 542                 } __attribute__ ((packed)) notify;
 543                 struct {
 544                         __le64 unused;
 545                         __le64 ver;
 546                 } __attribute__ ((packed)) assert_ver;
 547                 struct {
 548                         __le64 offset, length;
 549                         __le64 src_offset;
 550                 } __attribute__ ((packed)) clonerange;
 551                 struct {
 552                         __le64 max;     /* max data in reply */
 553                 } __attribute__ ((packed)) copy_get;
 554                 struct {
 555                         __le64 snapid;
 556                         __le64 src_version;
 557                         __u8 flags;
 558                         /*
 559                          * __le32 flags: CEPH_OSD_OP_FLAG_FADVISE_: mean the fadvise flags for dest object
 560                          * src_fadvise_flags mean the fadvise flags for src object
 561                          */
 562                         __le32 src_fadvise_flags;
 563                 } __attribute__ ((packed)) copy_from;
 564                 struct {
 565                         struct ceph_timespec stamp;
 566                 } __attribute__ ((packed)) hit_set_get;
 567                 struct {
 568                         __u8 flags;
 569                 } __attribute__ ((packed)) tmap2omap;
 570                 struct {
 571                         __le64 expected_object_size;
 572                         __le64 expected_write_size;
 573                         __le32 flags;  /* CEPH_OSD_OP_ALLOC_HINT_FLAG_* */
 574                 } __attribute__ ((packed)) alloc_hint;
 575                 struct {
 576                         __le64 offset;
 577                         __le64 length;
 578                         __le64 data_length;
 579                 } __attribute__ ((packed)) writesame;
 580                 struct {
 581                         __le64 offset;
 582                         __le64 length;
 583                         __le32 chunk_size;
 584                         __u8 type;              /* CEPH_OSD_CHECKSUM_OP_TYPE_* */
 585                 } __attribute__ ((packed)) checksum;
 586         };
 587         __le32 payload_len;
 588 } __attribute__ ((packed));
 589
 590 /*
 591  * Check the compatibility of struct ceph_osd_op
 592  *  (2+4+(2*8+8+4)+4) = (sizeof(ceph_osd_op::op) +
 593  *                     sizeof(ceph_osd_op::flags) +
 594  *                     sizeof(ceph_osd_op::extent) +
 595  *                     sizeof(ceph_osd_op::payload_len))
 596  */
 597 #ifdef __cplusplus
 598 static_assert(sizeof(ceph_osd_op) == (2+4+(2*8+8+4)+4),
 599               "sizeof(ceph_osd_op) breaks the compatibility");
 600 #endif
 601
 602 struct ceph_osd_reply_head {
 603         __le32 client_inc;                /* client incarnation */
 604         __le32 flags;
 605         struct ceph_object_layout layout;
 606         __le32 osdmap_epoch;
 607         struct ceph_eversion reassert_version; /* for replaying uncommitted */
 608
 609         __le32 result;                    /* result code */
 610
 611         __le32 object_len;                /* length of object name */
 612         __le32 num_ops;
 613         struct ceph_osd_op ops[0];  /* ops[], object */
 614 } __attribute__ ((packed));
 615
 616
 617 #endif