1 #ifndef _FS_CEPH_OSDMAP_H
2 #define _FS_CEPH_OSDMAP_H
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>
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
17 * Each map version is identified by an epoch, which increases monotonically.
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.
27 #define CEPH_POOL_FLAG_HASHPSPOOL 1
29 struct ceph_pg_pool_info {
37 int pg_num_mask, pgp_num_mask;
39 s64 write_tier; /* wins for read+write ops */
44 static inline bool ceph_can_shift_osds(struct ceph_pg_pool_info *pool)
47 case CEPH_POOL_TYPE_REP:
49 case CEPH_POOL_TYPE_EC:
56 struct ceph_object_locator {
61 * Maximum supported by kernel client object name length
63 * (probably outdated: must be >= RBD_MAX_MD_NAME_LEN -- currently 100)
65 #define CEPH_MAX_OID_NAME_LEN 100
68 * 51-char inline_name is long enough for all cephfs and all but one
69 * rbd requests: <imgname> in "<imgname>.rbd"/"rbd_id.<imgname>" can be
70 * arbitrarily long (~PAGE_SIZE). It's done once during rbd map; all
71 * other rbd requests fit into inline_name.
73 * Makes ceph_object_id 64 bytes on 64-bit.
75 #define CEPH_OID_INLINE_LEN 52
78 * Both inline and external buffers have space for a NUL-terminator,
79 * which is carried around. It's not required though - RADOS object
80 * names don't have to be NUL-terminated and may contain NULs.
82 struct ceph_object_id {
84 char inline_name[CEPH_OID_INLINE_LEN];
88 static inline void ceph_oid_init(struct ceph_object_id *oid)
90 oid->name = oid->inline_name;
94 static inline bool ceph_oid_empty(const struct ceph_object_id *oid)
96 return oid->name == oid->inline_name && !oid->name_len;
99 void ceph_oid_copy(struct ceph_object_id *dest,
100 const struct ceph_object_id *src);
102 void ceph_oid_printf(struct ceph_object_id *oid, const char *fmt, ...);
104 int ceph_oid_aprintf(struct ceph_object_id *oid, gfp_t gfp,
105 const char *fmt, ...);
106 void ceph_oid_destroy(struct ceph_object_id *oid);
108 struct ceph_pg_mapping {
124 struct ceph_fsid fsid;
126 struct ceph_timespec created, modified;
128 u32 flags; /* CEPH_OSDMAP_* */
130 u32 max_osd; /* size of osd_state, _offload, _addr arrays */
131 u8 *osd_state; /* CEPH_OSD_* */
132 u32 *osd_weight; /* 0 = failed, 0x10000 = 100% normal */
133 struct ceph_entity_addr *osd_addr;
135 struct rb_root pg_temp;
136 struct rb_root primary_temp;
138 u32 *osd_primary_affinity;
140 struct rb_root pg_pools;
143 /* the CRUSH map specifies the mapping of placement groups to
144 * the list of osds that store+replicate them. */
145 struct crush_map *crush;
147 struct mutex crush_scratch_mutex;
148 int crush_scratch_ary[CEPH_PG_MAX_SIZE * 3];
151 static inline int ceph_osd_exists(struct ceph_osdmap *map, int osd)
153 return osd >= 0 && osd < map->max_osd &&
154 (map->osd_state[osd] & CEPH_OSD_EXISTS);
157 static inline int ceph_osd_is_up(struct ceph_osdmap *map, int osd)
159 return ceph_osd_exists(map, osd) &&
160 (map->osd_state[osd] & CEPH_OSD_UP);
163 static inline int ceph_osd_is_down(struct ceph_osdmap *map, int osd)
165 return !ceph_osd_is_up(map, osd);
168 static inline bool ceph_osdmap_flag(struct ceph_osdmap *map, int flag)
170 return map && (map->flags & flag);
173 extern char *ceph_osdmap_state_str(char *str, int len, int state);
174 extern u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd);
176 static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
179 if (osd >= map->max_osd)
181 return &map->osd_addr[osd];
184 static inline int ceph_decode_pgid(void **p, void *end, struct ceph_pg *pgid)
188 if (!ceph_has_room(p, end, 1 + 8 + 4 + 4)) {
189 pr_warn("incomplete pg encoding\n");
192 version = ceph_decode_8(p);
194 pr_warn("do not understand pg encoding %d > 1\n",
199 pgid->pool = ceph_decode_64(p);
200 pgid->seed = ceph_decode_32(p);
201 *p += 4; /* skip deprecated preferred value */
206 extern struct ceph_osdmap *ceph_osdmap_decode(void **p, void *end);
207 struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
208 struct ceph_osdmap *map);
209 extern void ceph_osdmap_destroy(struct ceph_osdmap *map);
211 /* calculate mapping of a file extent to an object */
212 extern int ceph_calc_file_object_mapping(struct ceph_file_layout *layout,
214 u64 *bno, u64 *oxoff, u64 *oxlen);
216 int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap,
217 struct ceph_object_id *oid,
218 struct ceph_object_locator *oloc,
219 struct ceph_pg *raw_pgid);
221 extern int ceph_calc_pg_acting(struct ceph_osdmap *osdmap,
223 int *osds, int *primary);
224 extern int ceph_calc_pg_primary(struct ceph_osdmap *osdmap,
225 struct ceph_pg pgid);
227 extern struct ceph_pg_pool_info *ceph_pg_pool_by_id(struct ceph_osdmap *map,
230 extern const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id);
231 extern int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name);