include/linux/ceph/osdmap.h

   1 #ifndef _FS_CEPH_OSDMAP_H
   2 #define _FS_CEPH_OSDMAP_H
   3
   4 #include <linux/rbtree.h>
   5 #include <linux/ceph/types.h>
   6 #include <linux/ceph/decode.h>
   7 #include <linux/ceph/ceph_fs.h>
   8 #include <linux/crush/crush.h>
   9
  10 /*
  11  * The osd map describes the current membership of the osd cluster and
  12  * specifies the mapping of objects to placement groups and placement
  13  * groups to (sets of) osds.  That is, it completely specifies the
  14  * (desired) distribution of all data objects in the system at some
  15  * point in time.
  16  *
  17  * Each map version is identified by an epoch, which increases monotonically.
  18  *
  19  * The map can be updated either via an incremental map (diff) describing
  20  * the change between two successive epochs, or as a fully encoded map.
  21  */
  22 struct ceph_pg {
  23         uint64_t pool;
  24         uint32_t seed;
  25 };
  26
  27 #define CEPH_POOL_FLAG_HASHPSPOOL  1
  28
  29 struct ceph_pg_pool_info {
  30         struct rb_node node;
  31         s64 id;
  32         u8 type;
  33         u8 size;
  34         u8 crush_ruleset;
  35         u8 object_hash;
  36         u32 pg_num, pgp_num;
  37         int pg_num_mask, pgp_num_mask;
  38         s64 read_tier;
  39         s64 write_tier; /* wins for read+write ops */
  40         u64 flags;
  41         char *name;
  42 };
  43
  44 struct ceph_object_locator {
  45         s64 pool;
  46 };
  47
  48 /*
  49  * Maximum supported by kernel client object name length
  50  *
  51  * (probably outdated: must be >= RBD_MAX_MD_NAME_LEN -- currently 100)
  52  */
  53 #define CEPH_MAX_OID_NAME_LEN 100
  54
  55 struct ceph_object_id {
  56         char name[CEPH_MAX_OID_NAME_LEN];
  57         int name_len;
  58 };
  59
  60 struct ceph_pg_mapping {
  61         struct rb_node node;
  62         struct ceph_pg pgid;
  63         int len;
  64         int osds[];
  65 };
  66
  67 struct ceph_osdmap {
  68         struct ceph_fsid fsid;
  69         u32 epoch;
  70         u32 mkfs_epoch;
  71         struct ceph_timespec created, modified;
  72
  73         u32 flags;         /* CEPH_OSDMAP_* */
  74
  75         u32 max_osd;       /* size of osd_state, _offload, _addr arrays */
  76         u8 *osd_state;     /* CEPH_OSD_* */
  77         u32 *osd_weight;   /* 0 = failed, 0x10000 = 100% normal */
  78         struct ceph_entity_addr *osd_addr;
  79
  80         struct rb_root pg_temp;
  81         struct rb_root pg_pools;
  82         u32 pool_max;
  83
  84         /* the CRUSH map specifies the mapping of placement groups to
  85          * the list of osds that store+replicate them. */
  86         struct crush_map *crush;
  87 };
  88
  89 static inline void ceph_oid_set_name(struct ceph_object_id *oid,
  90                                      const char *name)
  91 {
  92         int len;
  93
  94         len = strlen(name);
  95         if (len > sizeof(oid->name)) {
  96                 WARN(1, "ceph_oid_set_name '%s' len %d vs %zu, truncating\n",
  97                      name, len, sizeof(oid->name));
  98                 len = sizeof(oid->name);
  99         }
 100
 101         memcpy(oid->name, name, len);
 102         oid->name_len = len;
 103 }
 104
 105 static inline void ceph_oid_copy(struct ceph_object_id *dest,
 106                                  struct ceph_object_id *src)
 107 {
 108         BUG_ON(src->name_len > sizeof(dest->name));
 109         memcpy(dest->name, src->name, src->name_len);
 110         dest->name_len = src->name_len;
 111 }
 112
 113 static inline int ceph_osd_is_up(struct ceph_osdmap *map, int osd)
 114 {
 115         return (osd < map->max_osd) && (map->osd_state[osd] & CEPH_OSD_UP);
 116 }
 117
 118 static inline bool ceph_osdmap_flag(struct ceph_osdmap *map, int flag)
 119 {
 120         return map && (map->flags & flag);
 121 }
 122
 123 extern char *ceph_osdmap_state_str(char *str, int len, int state);
 124
 125 static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
 126                                                      int osd)
 127 {
 128         if (osd >= map->max_osd)
 129                 return NULL;
 130         return &map->osd_addr[osd];
 131 }
 132
 133 static inline int ceph_decode_pgid(void **p, void *end, struct ceph_pg *pgid)
 134 {
 135         __u8 version;
 136
 137         if (!ceph_has_room(p, end, 1 + 8 + 4 + 4)) {
 138                 pr_warning("incomplete pg encoding");
 139
 140                 return -EINVAL;
 141         }
 142         version = ceph_decode_8(p);
 143         if (version > 1) {
 144                 pr_warning("do not understand pg encoding %d > 1",
 145                         (int)version);
 146                 return -EINVAL;
 147         }
 148
 149         pgid->pool = ceph_decode_64(p);
 150         pgid->seed = ceph_decode_32(p);
 151         *p += 4;        /* skip deprecated preferred value */
 152
 153         return 0;
 154 }
 155
 156 extern struct ceph_osdmap *osdmap_decode(void **p, void *end);
 157 extern struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 158                                             struct ceph_osdmap *map,
 159                                             struct ceph_messenger *msgr);
 160 extern void ceph_osdmap_destroy(struct ceph_osdmap *map);
 161
 162 /* calculate mapping of a file extent to an object */
 163 extern int ceph_calc_file_object_mapping(struct ceph_file_layout *layout,
 164                                          u64 off, u64 len,
 165                                          u64 *bno, u64 *oxoff, u64 *oxlen);
 166
 167 /* calculate mapping of object to a placement group */
 168 extern int ceph_oloc_oid_to_pg(struct ceph_osdmap *osdmap,
 169                                struct ceph_object_locator *oloc,
 170                                struct ceph_object_id *oid,
 171                                struct ceph_pg *pg_out);
 172
 173 extern int ceph_calc_pg_acting(struct ceph_osdmap *osdmap,
 174                                struct ceph_pg pgid,
 175                                int *acting);
 176 extern int ceph_calc_pg_primary(struct ceph_osdmap *osdmap,
 177                                 struct ceph_pg pgid);
 178
 179 extern struct ceph_pg_pool_info *ceph_pg_pool_by_id(struct ceph_osdmap *map,
 180                                                     u64 id);
 181
 182 extern const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id);
 183 extern int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name);
 184
 185 #endif