blob: db4fb6322aae080c744a93659b4b1b13a3f295b5 [file] [log] [blame]
Sage Weilf24e9982009-10-06 11:31:10 -07001#ifndef _FS_CEPH_OSDMAP_H
2#define _FS_CEPH_OSDMAP_H
3
4#include <linux/rbtree.h>
David Howellsa1ce3922012-10-02 18:01:25 +01005#include <linux/ceph/types.h>
Alex Elderef4859d2013-04-01 18:58:26 -05006#include <linux/ceph/decode.h>
David Howellsa1ce3922012-10-02 18:01:25 +01007#include <linux/ceph/ceph_fs.h>
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07008#include <linux/crush/crush.h>
Sage Weilf24e9982009-10-06 11:31:10 -07009
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 */
Sage Weil5b191d92013-02-23 10:38:16 -080022struct ceph_pg {
23 uint64_t pool;
24 uint32_t seed;
25};
26
Sage Weil83ca14f2013-02-26 10:39:09 -080027#define CEPH_POOL_FLAG_HASHPSPOOL 1
28
Sage Weilf24e9982009-10-06 11:31:10 -070029struct ceph_pg_pool_info {
Sage Weil4fc51be2010-02-16 15:55:03 -080030 struct rb_node node;
Sage Weil4f6a7e52013-02-23 10:41:09 -080031 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;
Ilya Dryomov17a13e402014-01-27 17:40:19 +020038 s64 read_tier;
39 s64 write_tier; /* wins for read+write ops */
Sage Weil4f6a7e52013-02-23 10:41:09 -080040 u64 flags;
Sage Weil2844a762010-04-09 15:46:42 -070041 char *name;
Sage Weilf24e9982009-10-06 11:31:10 -070042};
43
Sage Weil4f6a7e52013-02-23 10:41:09 -080044struct ceph_object_locator {
Ilya Dryomov221165252014-01-27 17:40:18 +020045 s64 pool;
Sage Weil4f6a7e52013-02-23 10:41:09 -080046};
47
Ilya Dryomov4295f222014-01-27 17:40:18 +020048/*
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51 * (probably outdated: must be 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
55struct ceph_object_id {
56 char name[CEPH_MAX_OID_NAME_LEN];
57 int name_len;
58};
59
Sage Weilf24e9982009-10-06 11:31:10 -070060struct ceph_pg_mapping {
61 struct rb_node node;
Sage Weil5b191d92013-02-23 10:38:16 -080062 struct ceph_pg pgid;
Ilya Dryomov35a935d2014-03-21 19:05:29 +020063
64 union {
65 struct {
66 int len;
67 int osds[];
68 } pg_temp;
Ilya Dryomov9686f942014-03-21 19:05:29 +020069 struct {
70 int osd;
71 } primary_temp;
Ilya Dryomov35a935d2014-03-21 19:05:29 +020072 };
Sage Weilf24e9982009-10-06 11:31:10 -070073};
74
75struct ceph_osdmap {
76 struct ceph_fsid fsid;
77 u32 epoch;
78 u32 mkfs_epoch;
79 struct ceph_timespec created, modified;
80
81 u32 flags; /* CEPH_OSDMAP_* */
82
83 u32 max_osd; /* size of osd_state, _offload, _addr arrays */
84 u8 *osd_state; /* CEPH_OSD_* */
85 u32 *osd_weight; /* 0 = failed, 0x10000 = 100% normal */
86 struct ceph_entity_addr *osd_addr;
87
88 struct rb_root pg_temp;
Ilya Dryomov9686f942014-03-21 19:05:29 +020089 struct rb_root primary_temp;
90
Sage Weil4fc51be2010-02-16 15:55:03 -080091 struct rb_root pg_pools;
92 u32 pool_max;
Sage Weilf24e9982009-10-06 11:31:10 -070093
94 /* the CRUSH map specifies the mapping of placement groups to
95 * the list of osds that store+replicate them. */
96 struct crush_map *crush;
Ilya Dryomov9d521472014-01-31 17:54:26 +020097
98 struct mutex crush_scratch_mutex;
99 int crush_scratch_ary[CEPH_PG_MAX_SIZE * 3];
Sage Weilf24e9982009-10-06 11:31:10 -0700100};
101
Ilya Dryomov4295f222014-01-27 17:40:18 +0200102static inline void ceph_oid_set_name(struct ceph_object_id *oid,
103 const char *name)
104{
105 int len;
106
107 len = strlen(name);
108 if (len > sizeof(oid->name)) {
109 WARN(1, "ceph_oid_set_name '%s' len %d vs %zu, truncating\n",
110 name, len, sizeof(oid->name));
111 len = sizeof(oid->name);
112 }
113
114 memcpy(oid->name, name, len);
115 oid->name_len = len;
116}
117
118static inline void ceph_oid_copy(struct ceph_object_id *dest,
119 struct ceph_object_id *src)
120{
121 BUG_ON(src->name_len > sizeof(dest->name));
122 memcpy(dest->name, src->name, src->name_len);
123 dest->name_len = src->name_len;
124}
125
Sage Weilf24e9982009-10-06 11:31:10 -0700126static inline int ceph_osd_is_up(struct ceph_osdmap *map, int osd)
127{
128 return (osd < map->max_osd) && (map->osd_state[osd] & CEPH_OSD_UP);
129}
130
131static inline bool ceph_osdmap_flag(struct ceph_osdmap *map, int flag)
132{
133 return map && (map->flags & flag);
134}
135
136extern char *ceph_osdmap_state_str(char *str, int len, int state);
137
138static inline struct ceph_entity_addr *ceph_osd_addr(struct ceph_osdmap *map,
139 int osd)
140{
141 if (osd >= map->max_osd)
142 return NULL;
143 return &map->osd_addr[osd];
144}
145
Alex Elderef4859d2013-04-01 18:58:26 -0500146static inline int ceph_decode_pgid(void **p, void *end, struct ceph_pg *pgid)
147{
148 __u8 version;
149
150 if (!ceph_has_room(p, end, 1 + 8 + 4 + 4)) {
151 pr_warning("incomplete pg encoding");
152
153 return -EINVAL;
154 }
155 version = ceph_decode_8(p);
156 if (version > 1) {
157 pr_warning("do not understand pg encoding %d > 1",
158 (int)version);
159 return -EINVAL;
160 }
161
162 pgid->pool = ceph_decode_64(p);
163 pgid->seed = ceph_decode_32(p);
164 *p += 4; /* skip deprecated preferred value */
165
166 return 0;
167}
168
Ilya Dryomova2505d62014-03-13 16:36:13 +0200169extern struct ceph_osdmap *ceph_osdmap_decode(void **p, void *end);
Sage Weilf24e9982009-10-06 11:31:10 -0700170extern struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
171 struct ceph_osdmap *map,
172 struct ceph_messenger *msgr);
173extern void ceph_osdmap_destroy(struct ceph_osdmap *map);
174
175/* calculate mapping of a file extent to an object */
Sage Weild63b77f2012-09-24 20:59:48 -0700176extern int ceph_calc_file_object_mapping(struct ceph_file_layout *layout,
Alex Eldere8afad62012-11-14 09:38:19 -0600177 u64 off, u64 len,
Sage Weild63b77f2012-09-24 20:59:48 -0700178 u64 *bno, u64 *oxoff, u64 *oxlen);
Sage Weilf24e9982009-10-06 11:31:10 -0700179
180/* calculate mapping of object to a placement group */
Ilya Dryomov7c13cb62014-01-27 17:40:19 +0200181extern int ceph_oloc_oid_to_pg(struct ceph_osdmap *osdmap,
182 struct ceph_object_locator *oloc,
183 struct ceph_object_id *oid,
184 struct ceph_pg *pg_out);
185
Sage Weil129793542013-01-08 09:15:10 -0800186extern int ceph_calc_pg_acting(struct ceph_osdmap *osdmap,
Sage Weil5b191d92013-02-23 10:38:16 -0800187 struct ceph_pg pgid,
Sage Weild85b7052010-05-10 10:24:48 -0700188 int *acting);
Sage Weil51042122009-11-04 11:39:12 -0800189extern int ceph_calc_pg_primary(struct ceph_osdmap *osdmap,
Sage Weil5b191d92013-02-23 10:38:16 -0800190 struct ceph_pg pgid);
Sage Weilf24e9982009-10-06 11:31:10 -0700191
Ilya Dryomovce7f6a22014-01-27 17:40:19 +0200192extern struct ceph_pg_pool_info *ceph_pg_pool_by_id(struct ceph_osdmap *map,
193 u64 id);
194
Alex Elder72afc712012-10-30 19:40:33 -0500195extern const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id);
Yehuda Sadeh7669a2c2010-05-17 12:31:35 -0700196extern int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name);
197
Sage Weilf24e9982009-10-06 11:31:10 -0700198#endif