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 int ceph_pg_compare(const struct ceph_pg
*lhs
, const struct ceph_pg
*rhs
);
29 #define CEPH_POOL_FLAG_HASHPSPOOL (1ULL << 0) /* hash pg seed and pool id
31 #define CEPH_POOL_FLAG_FULL (1ULL << 1) /* pool is full */
33 struct ceph_pg_pool_info
{
36 u8 type
; /* CEPH_POOL_TYPE_* */
41 u32 last_force_request_resend
;
43 int pg_num_mask
, pgp_num_mask
;
45 s64 write_tier
; /* wins for read+write ops */
46 u64 flags
; /* CEPH_POOL_FLAG_* */
49 bool was_full
; /* for handle_one_map() */
52 static inline bool ceph_can_shift_osds(struct ceph_pg_pool_info
*pool
)
55 case CEPH_POOL_TYPE_REP
:
57 case CEPH_POOL_TYPE_EC
:
64 struct ceph_object_locator
{
66 struct ceph_string
*pool_ns
;
69 static inline void ceph_oloc_init(struct ceph_object_locator
*oloc
)
75 static inline bool ceph_oloc_empty(const struct ceph_object_locator
*oloc
)
77 return oloc
->pool
== -1;
80 void ceph_oloc_copy(struct ceph_object_locator
*dest
,
81 const struct ceph_object_locator
*src
);
82 void ceph_oloc_destroy(struct ceph_object_locator
*oloc
);
85 * Maximum supported by kernel client object name length
87 * (probably outdated: must be >= RBD_MAX_MD_NAME_LEN -- currently 100)
89 #define CEPH_MAX_OID_NAME_LEN 100
92 * 51-char inline_name is long enough for all cephfs and all but one
93 * rbd requests: <imgname> in "<imgname>.rbd"/"rbd_id.<imgname>" can be
94 * arbitrarily long (~PAGE_SIZE). It's done once during rbd map; all
95 * other rbd requests fit into inline_name.
97 * Makes ceph_object_id 64 bytes on 64-bit.
99 #define CEPH_OID_INLINE_LEN 52
102 * Both inline and external buffers have space for a NUL-terminator,
103 * which is carried around. It's not required though - RADOS object
104 * names don't have to be NUL-terminated and may contain NULs.
106 struct ceph_object_id
{
108 char inline_name
[CEPH_OID_INLINE_LEN
];
112 static inline void ceph_oid_init(struct ceph_object_id
*oid
)
114 oid
->name
= oid
->inline_name
;
118 #define CEPH_OID_INIT_ONSTACK(oid) \
119 ({ ceph_oid_init(&oid); oid; })
120 #define CEPH_DEFINE_OID_ONSTACK(oid) \
121 struct ceph_object_id oid = CEPH_OID_INIT_ONSTACK(oid)
123 static inline bool ceph_oid_empty(const struct ceph_object_id
*oid
)
125 return oid
->name
== oid
->inline_name
&& !oid
->name_len
;
128 void ceph_oid_copy(struct ceph_object_id
*dest
,
129 const struct ceph_object_id
*src
);
131 void ceph_oid_printf(struct ceph_object_id
*oid
, const char *fmt
, ...);
133 int ceph_oid_aprintf(struct ceph_object_id
*oid
, gfp_t gfp
,
134 const char *fmt
, ...);
135 void ceph_oid_destroy(struct ceph_object_id
*oid
);
137 struct ceph_pg_mapping
{
153 struct ceph_fsid fsid
;
155 struct ceph_timespec created
, modified
;
157 u32 flags
; /* CEPH_OSDMAP_* */
159 u32 max_osd
; /* size of osd_state, _offload, _addr arrays */
160 u8
*osd_state
; /* CEPH_OSD_* */
161 u32
*osd_weight
; /* 0 = failed, 0x10000 = 100% normal */
162 struct ceph_entity_addr
*osd_addr
;
164 struct rb_root pg_temp
;
165 struct rb_root primary_temp
;
167 u32
*osd_primary_affinity
;
169 struct rb_root pg_pools
;
172 /* the CRUSH map specifies the mapping of placement groups to
173 * the list of osds that store+replicate them. */
174 struct crush_map
*crush
;
176 struct mutex crush_scratch_mutex
;
177 int crush_scratch_ary
[CEPH_PG_MAX_SIZE
* 3];
180 static inline bool ceph_osd_exists(struct ceph_osdmap
*map
, int osd
)
182 return osd
>= 0 && osd
< map
->max_osd
&&
183 (map
->osd_state
[osd
] & CEPH_OSD_EXISTS
);
186 static inline bool ceph_osd_is_up(struct ceph_osdmap
*map
, int osd
)
188 return ceph_osd_exists(map
, osd
) &&
189 (map
->osd_state
[osd
] & CEPH_OSD_UP
);
192 static inline bool ceph_osd_is_down(struct ceph_osdmap
*map
, int osd
)
194 return !ceph_osd_is_up(map
, osd
);
197 extern char *ceph_osdmap_state_str(char *str
, int len
, int state
);
198 extern u32
ceph_get_primary_affinity(struct ceph_osdmap
*map
, int osd
);
200 static inline struct ceph_entity_addr
*ceph_osd_addr(struct ceph_osdmap
*map
,
203 if (osd
>= map
->max_osd
)
205 return &map
->osd_addr
[osd
];
208 static inline int ceph_decode_pgid(void **p
, void *end
, struct ceph_pg
*pgid
)
212 if (!ceph_has_room(p
, end
, 1 + 8 + 4 + 4)) {
213 pr_warn("incomplete pg encoding\n");
216 version
= ceph_decode_8(p
);
218 pr_warn("do not understand pg encoding %d > 1\n",
223 pgid
->pool
= ceph_decode_64(p
);
224 pgid
->seed
= ceph_decode_32(p
);
225 *p
+= 4; /* skip deprecated preferred value */
230 struct ceph_osdmap
*ceph_osdmap_alloc(void);
231 extern struct ceph_osdmap
*ceph_osdmap_decode(void **p
, void *end
);
232 struct ceph_osdmap
*osdmap_apply_incremental(void **p
, void *end
,
233 struct ceph_osdmap
*map
);
234 extern void ceph_osdmap_destroy(struct ceph_osdmap
*map
);
237 int osds
[CEPH_PG_MAX_SIZE
];
239 int primary
; /* id, NOT index */
242 static inline void ceph_osds_init(struct ceph_osds
*set
)
248 void ceph_osds_copy(struct ceph_osds
*dest
, const struct ceph_osds
*src
);
250 bool ceph_is_new_interval(const struct ceph_osds
*old_acting
,
251 const struct ceph_osds
*new_acting
,
252 const struct ceph_osds
*old_up
,
253 const struct ceph_osds
*new_up
,
260 bool old_sort_bitwise
,
261 bool new_sort_bitwise
,
262 const struct ceph_pg
*pgid
);
263 bool ceph_osds_changed(const struct ceph_osds
*old_acting
,
264 const struct ceph_osds
*new_acting
,
267 /* calculate mapping of a file extent to an object */
268 extern int ceph_calc_file_object_mapping(struct ceph_file_layout
*layout
,
270 u64
*bno
, u64
*oxoff
, u64
*oxlen
);
272 int ceph_object_locator_to_pg(struct ceph_osdmap
*osdmap
,
273 struct ceph_object_id
*oid
,
274 struct ceph_object_locator
*oloc
,
275 struct ceph_pg
*raw_pgid
);
277 void ceph_pg_to_up_acting_osds(struct ceph_osdmap
*osdmap
,
278 const struct ceph_pg
*raw_pgid
,
279 struct ceph_osds
*up
,
280 struct ceph_osds
*acting
);
281 int ceph_pg_to_acting_primary(struct ceph_osdmap
*osdmap
,
282 const struct ceph_pg
*raw_pgid
);
284 extern struct ceph_pg_pool_info
*ceph_pg_pool_by_id(struct ceph_osdmap
*map
,
287 extern const char *ceph_pg_pool_name_by_id(struct ceph_osdmap
*map
, u64 id
);
288 extern int ceph_pg_poolid_by_name(struct ceph_osdmap
*map
, const char *name
);