1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _FS_CEPH_OSDMAP_H
3 #define _FS_CEPH_OSDMAP_H
5 #include <linux/rbtree.h>
6 #include <linux/ceph/types.h>
7 #include <linux/ceph/decode.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_SPG_NOSHARD -1
34 int ceph_pg_compare(const struct ceph_pg *lhs, const struct ceph_pg *rhs);
35 int ceph_spg_compare(const struct ceph_spg *lhs, const struct ceph_spg *rhs);
37 #define CEPH_POOL_FLAG_HASHPSPOOL (1ULL << 0) /* hash pg seed and pool id
39 #define CEPH_POOL_FLAG_FULL (1ULL << 1) /* pool is full */
40 #define CEPH_POOL_FLAG_FULL_QUOTA (1ULL << 10) /* pool ran out of quota,
42 #define CEPH_POOL_FLAG_NEARFULL (1ULL << 11) /* pool is nearfull */
44 struct ceph_pg_pool_info {
47 u8 type; /* CEPH_POOL_TYPE_* */
52 u32 last_force_request_resend;
54 int pg_num_mask, pgp_num_mask;
56 s64 write_tier; /* wins for read+write ops */
57 u64 flags; /* CEPH_POOL_FLAG_* */
60 bool was_full; /* for handle_one_map() */
63 static inline bool ceph_can_shift_osds(struct ceph_pg_pool_info *pool)
66 case CEPH_POOL_TYPE_REP:
68 case CEPH_POOL_TYPE_EC:
75 struct ceph_object_locator {
77 struct ceph_string *pool_ns;
80 static inline void ceph_oloc_init(struct ceph_object_locator *oloc)
86 static inline bool ceph_oloc_empty(const struct ceph_object_locator *oloc)
88 return oloc->pool == -1;
91 void ceph_oloc_copy(struct ceph_object_locator *dest,
92 const struct ceph_object_locator *src);
93 void ceph_oloc_destroy(struct ceph_object_locator *oloc);
96 * 51-char inline_name is long enough for all cephfs and all but one
97 * rbd requests: <imgname> in "<imgname>.rbd"/"rbd_id.<imgname>" can be
98 * arbitrarily long (~PAGE_SIZE). It's done once during rbd map; all
99 * other rbd requests fit into inline_name.
101 * Makes ceph_object_id 64 bytes on 64-bit.
103 #define CEPH_OID_INLINE_LEN 52
106 * Both inline and external buffers have space for a NUL-terminator,
107 * which is carried around. It's not required though - RADOS object
108 * names don't have to be NUL-terminated and may contain NULs.
110 struct ceph_object_id {
112 char inline_name[CEPH_OID_INLINE_LEN];
116 #define __CEPH_OID_INITIALIZER(oid) { .name = (oid).inline_name }
118 #define CEPH_DEFINE_OID_ONSTACK(oid) \
119 struct ceph_object_id oid = __CEPH_OID_INITIALIZER(oid)
121 static inline void ceph_oid_init(struct ceph_object_id *oid)
123 *oid = (struct ceph_object_id) __CEPH_OID_INITIALIZER(*oid);
126 static inline bool ceph_oid_empty(const struct ceph_object_id *oid)
128 return oid->name == oid->inline_name && !oid->name_len;
131 void ceph_oid_copy(struct ceph_object_id *dest,
132 const struct ceph_object_id *src);
134 void ceph_oid_printf(struct ceph_object_id *oid, const char *fmt, ...);
136 int ceph_oid_aprintf(struct ceph_object_id *oid, gfp_t gfp,
137 const char *fmt, ...);
138 void ceph_oid_destroy(struct ceph_object_id *oid);
140 struct ceph_pg_mapping {
160 struct ceph_fsid fsid;
162 struct ceph_timespec created, modified;
164 u32 flags; /* CEPH_OSDMAP_* */
166 u32 max_osd; /* size of osd_state, _offload, _addr arrays */
167 u32 *osd_state; /* CEPH_OSD_* */
168 u32 *osd_weight; /* 0 = failed, 0x10000 = 100% normal */
169 struct ceph_entity_addr *osd_addr;
171 struct rb_root pg_temp;
172 struct rb_root primary_temp;
174 /* remap (post-CRUSH, pre-up) */
175 struct rb_root pg_upmap; /* PG := raw set */
176 struct rb_root pg_upmap_items; /* from -> to within raw set */
178 u32 *osd_primary_affinity;
180 struct rb_root pg_pools;
183 /* the CRUSH map specifies the mapping of placement groups to
184 * the list of osds that store+replicate them. */
185 struct crush_map *crush;
187 struct mutex crush_workspace_mutex;
188 void *crush_workspace;
191 static inline bool ceph_osd_exists(struct ceph_osdmap *map, int osd)
193 return osd >= 0 && osd < map->max_osd &&
194 (map->osd_state[osd] & CEPH_OSD_EXISTS);
197 static inline bool ceph_osd_is_up(struct ceph_osdmap *map, int osd)
199 return ceph_osd_exists(map, osd) &&
200 (map->osd_state[osd] & CEPH_OSD_UP);
203 static inline bool ceph_osd_is_down(struct ceph_osdmap *map, int osd)
205 return !ceph_osd_is_up(map, osd);
208 char *ceph_osdmap_state_str(char *str, int len, u32 state);
209 extern u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd);
211 static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
214 if (osd >= map->max_osd)
216 return &map->osd_addr[osd];
219 #define CEPH_PGID_ENCODING_LEN (1 + 8 + 4 + 4)
221 static inline int ceph_decode_pgid(void **p, void *end, struct ceph_pg *pgid)
225 if (!ceph_has_room(p, end, CEPH_PGID_ENCODING_LEN)) {
226 pr_warn("incomplete pg encoding\n");
229 version = ceph_decode_8(p);
231 pr_warn("do not understand pg encoding %d > 1\n",
236 pgid->pool = ceph_decode_64(p);
237 pgid->seed = ceph_decode_32(p);
238 *p += 4; /* skip deprecated preferred value */
243 struct ceph_osdmap *ceph_osdmap_alloc(void);
244 extern struct ceph_osdmap *ceph_osdmap_decode(void **p, void *end);
245 struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
246 struct ceph_osdmap *map);
247 extern void ceph_osdmap_destroy(struct ceph_osdmap *map);
250 int osds[CEPH_PG_MAX_SIZE];
252 int primary; /* id, NOT index */
255 static inline void ceph_osds_init(struct ceph_osds *set)
261 void ceph_osds_copy(struct ceph_osds *dest, const struct ceph_osds *src);
263 bool ceph_pg_is_split(const struct ceph_pg *pgid, u32 old_pg_num,
265 bool ceph_is_new_interval(const struct ceph_osds *old_acting,
266 const struct ceph_osds *new_acting,
267 const struct ceph_osds *old_up,
268 const struct ceph_osds *new_up,
275 bool old_sort_bitwise,
276 bool new_sort_bitwise,
277 bool old_recovery_deletes,
278 bool new_recovery_deletes,
279 const struct ceph_pg *pgid);
280 bool ceph_osds_changed(const struct ceph_osds *old_acting,
281 const struct ceph_osds *new_acting,
284 void __ceph_object_locator_to_pg(struct ceph_pg_pool_info *pi,
285 const struct ceph_object_id *oid,
286 const struct ceph_object_locator *oloc,
287 struct ceph_pg *raw_pgid);
288 int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap,
289 const struct ceph_object_id *oid,
290 const struct ceph_object_locator *oloc,
291 struct ceph_pg *raw_pgid);
293 void ceph_pg_to_up_acting_osds(struct ceph_osdmap *osdmap,
294 struct ceph_pg_pool_info *pi,
295 const struct ceph_pg *raw_pgid,
296 struct ceph_osds *up,
297 struct ceph_osds *acting);
298 bool ceph_pg_to_primary_shard(struct ceph_osdmap *osdmap,
299 struct ceph_pg_pool_info *pi,
300 const struct ceph_pg *raw_pgid,
301 struct ceph_spg *spgid);
302 int ceph_pg_to_acting_primary(struct ceph_osdmap *osdmap,
303 const struct ceph_pg *raw_pgid);
310 struct crush_loc_node {
311 struct rb_node cl_node;
312 struct crush_loc cl_loc; /* pointers into cl_data */
316 int ceph_parse_crush_location(char *crush_location, struct rb_root *locs);
317 int ceph_compare_crush_locs(struct rb_root *locs1, struct rb_root *locs2);
318 void ceph_clear_crush_locs(struct rb_root *locs);
320 int ceph_get_crush_locality(struct ceph_osdmap *osdmap, int id,
321 struct rb_root *locs);
323 extern struct ceph_pg_pool_info *ceph_pg_pool_by_id(struct ceph_osdmap *map,
325 extern const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id);
326 extern int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name);
327 u64 ceph_pg_pool_flags(struct ceph_osdmap *map, u64 id);