2 #include <linux/ceph/ceph_debug.h>
4 #include <linux/module.h>
5 #include <linux/slab.h>
8 #include <linux/ceph/libceph.h>
9 #include <linux/ceph/osdmap.h>
10 #include <linux/ceph/decode.h>
11 #include <linux/crush/hash.h>
12 #include <linux/crush/mapper.h>
14 char *ceph_osdmap_state_str(char *str, int len, int state)
19 if ((state & CEPH_OSD_EXISTS) && (state & CEPH_OSD_UP))
20 snprintf(str, len, "exists, up");
21 else if (state & CEPH_OSD_EXISTS)
22 snprintf(str, len, "exists");
23 else if (state & CEPH_OSD_UP)
24 snprintf(str, len, "up");
26 snprintf(str, len, "doesn't exist");
33 static int calc_bits_of(unsigned int t)
44 * the foo_mask is the smallest value 2^n-1 that is >= foo.
46 static void calc_pg_masks(struct ceph_pg_pool_info *pi)
48 pi->pg_num_mask = (1 << calc_bits_of(pi->pg_num-1)) - 1;
49 pi->pgp_num_mask = (1 << calc_bits_of(pi->pgp_num-1)) - 1;
55 static int crush_decode_uniform_bucket(void **p, void *end,
56 struct crush_bucket_uniform *b)
58 dout("crush_decode_uniform_bucket %p to %p\n", *p, end);
59 ceph_decode_need(p, end, (1+b->h.size) * sizeof(u32), bad);
60 b->item_weight = ceph_decode_32(p);
66 static int crush_decode_list_bucket(void **p, void *end,
67 struct crush_bucket_list *b)
70 dout("crush_decode_list_bucket %p to %p\n", *p, end);
71 b->item_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS);
72 if (b->item_weights == NULL)
74 b->sum_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS);
75 if (b->sum_weights == NULL)
77 ceph_decode_need(p, end, 2 * b->h.size * sizeof(u32), bad);
78 for (j = 0; j < b->h.size; j++) {
79 b->item_weights[j] = ceph_decode_32(p);
80 b->sum_weights[j] = ceph_decode_32(p);
87 static int crush_decode_tree_bucket(void **p, void *end,
88 struct crush_bucket_tree *b)
91 dout("crush_decode_tree_bucket %p to %p\n", *p, end);
92 ceph_decode_8_safe(p, end, b->num_nodes, bad);
93 b->node_weights = kcalloc(b->num_nodes, sizeof(u32), GFP_NOFS);
94 if (b->node_weights == NULL)
96 ceph_decode_need(p, end, b->num_nodes * sizeof(u32), bad);
97 for (j = 0; j < b->num_nodes; j++)
98 b->node_weights[j] = ceph_decode_32(p);
104 static int crush_decode_straw_bucket(void **p, void *end,
105 struct crush_bucket_straw *b)
108 dout("crush_decode_straw_bucket %p to %p\n", *p, end);
109 b->item_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS);
110 if (b->item_weights == NULL)
112 b->straws = kcalloc(b->h.size, sizeof(u32), GFP_NOFS);
113 if (b->straws == NULL)
115 ceph_decode_need(p, end, 2 * b->h.size * sizeof(u32), bad);
116 for (j = 0; j < b->h.size; j++) {
117 b->item_weights[j] = ceph_decode_32(p);
118 b->straws[j] = ceph_decode_32(p);
125 static int crush_decode_straw2_bucket(void **p, void *end,
126 struct crush_bucket_straw2 *b)
129 dout("crush_decode_straw2_bucket %p to %p\n", *p, end);
130 b->item_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS);
131 if (b->item_weights == NULL)
133 ceph_decode_need(p, end, b->h.size * sizeof(u32), bad);
134 for (j = 0; j < b->h.size; j++)
135 b->item_weights[j] = ceph_decode_32(p);
141 static int skip_name_map(void **p, void *end)
144 ceph_decode_32_safe(p, end, len ,bad);
148 ceph_decode_32_safe(p, end, strlen, bad);
156 static void crush_finalize(struct crush_map *c)
160 /* Space for the array of pointers to per-bucket workspace */
161 c->working_size = sizeof(struct crush_work) +
162 c->max_buckets * sizeof(struct crush_work_bucket *);
164 for (b = 0; b < c->max_buckets; b++) {
168 switch (c->buckets[b]->alg) {
171 * The base case, permutation variables and
172 * the pointer to the permutation array.
174 c->working_size += sizeof(struct crush_work_bucket);
177 /* Every bucket has a permutation array. */
178 c->working_size += c->buckets[b]->size * sizeof(__u32);
182 static struct crush_map *crush_decode(void *pbyval, void *end)
188 void *start = pbyval;
192 dout("crush_decode %p to %p len %d\n", *p, end, (int)(end - *p));
194 c = kzalloc(sizeof(*c), GFP_NOFS);
196 return ERR_PTR(-ENOMEM);
198 /* set tunables to default values */
199 c->choose_local_tries = 2;
200 c->choose_local_fallback_tries = 5;
201 c->choose_total_tries = 19;
202 c->chooseleaf_descend_once = 0;
204 ceph_decode_need(p, end, 4*sizeof(u32), bad);
205 magic = ceph_decode_32(p);
206 if (magic != CRUSH_MAGIC) {
207 pr_err("crush_decode magic %x != current %x\n",
208 (unsigned int)magic, (unsigned int)CRUSH_MAGIC);
211 c->max_buckets = ceph_decode_32(p);
212 c->max_rules = ceph_decode_32(p);
213 c->max_devices = ceph_decode_32(p);
215 c->buckets = kcalloc(c->max_buckets, sizeof(*c->buckets), GFP_NOFS);
216 if (c->buckets == NULL)
218 c->rules = kcalloc(c->max_rules, sizeof(*c->rules), GFP_NOFS);
219 if (c->rules == NULL)
223 for (i = 0; i < c->max_buckets; i++) {
226 struct crush_bucket *b;
228 ceph_decode_32_safe(p, end, alg, bad);
230 c->buckets[i] = NULL;
233 dout("crush_decode bucket %d off %x %p to %p\n",
234 i, (int)(*p-start), *p, end);
237 case CRUSH_BUCKET_UNIFORM:
238 size = sizeof(struct crush_bucket_uniform);
240 case CRUSH_BUCKET_LIST:
241 size = sizeof(struct crush_bucket_list);
243 case CRUSH_BUCKET_TREE:
244 size = sizeof(struct crush_bucket_tree);
246 case CRUSH_BUCKET_STRAW:
247 size = sizeof(struct crush_bucket_straw);
249 case CRUSH_BUCKET_STRAW2:
250 size = sizeof(struct crush_bucket_straw2);
257 b = c->buckets[i] = kzalloc(size, GFP_NOFS);
261 ceph_decode_need(p, end, 4*sizeof(u32), bad);
262 b->id = ceph_decode_32(p);
263 b->type = ceph_decode_16(p);
264 b->alg = ceph_decode_8(p);
265 b->hash = ceph_decode_8(p);
266 b->weight = ceph_decode_32(p);
267 b->size = ceph_decode_32(p);
269 dout("crush_decode bucket size %d off %x %p to %p\n",
270 b->size, (int)(*p-start), *p, end);
272 b->items = kcalloc(b->size, sizeof(__s32), GFP_NOFS);
273 if (b->items == NULL)
276 ceph_decode_need(p, end, b->size*sizeof(u32), bad);
277 for (j = 0; j < b->size; j++)
278 b->items[j] = ceph_decode_32(p);
281 case CRUSH_BUCKET_UNIFORM:
282 err = crush_decode_uniform_bucket(p, end,
283 (struct crush_bucket_uniform *)b);
287 case CRUSH_BUCKET_LIST:
288 err = crush_decode_list_bucket(p, end,
289 (struct crush_bucket_list *)b);
293 case CRUSH_BUCKET_TREE:
294 err = crush_decode_tree_bucket(p, end,
295 (struct crush_bucket_tree *)b);
299 case CRUSH_BUCKET_STRAW:
300 err = crush_decode_straw_bucket(p, end,
301 (struct crush_bucket_straw *)b);
305 case CRUSH_BUCKET_STRAW2:
306 err = crush_decode_straw2_bucket(p, end,
307 (struct crush_bucket_straw2 *)b);
315 dout("rule vec is %p\n", c->rules);
316 for (i = 0; i < c->max_rules; i++) {
318 struct crush_rule *r;
321 ceph_decode_32_safe(p, end, yes, bad);
323 dout("crush_decode NO rule %d off %x %p to %p\n",
324 i, (int)(*p-start), *p, end);
329 dout("crush_decode rule %d off %x %p to %p\n",
330 i, (int)(*p-start), *p, end);
333 ceph_decode_32_safe(p, end, yes, bad);
334 #if BITS_PER_LONG == 32
336 if (yes > (ULONG_MAX - sizeof(*r))
337 / sizeof(struct crush_rule_step))
340 r = c->rules[i] = kmalloc(sizeof(*r) +
341 yes*sizeof(struct crush_rule_step),
345 dout(" rule %d is at %p\n", i, r);
347 ceph_decode_copy_safe(p, end, &r->mask, 4, bad); /* 4 u8's */
348 ceph_decode_need(p, end, r->len*3*sizeof(u32), bad);
349 for (j = 0; j < r->len; j++) {
350 r->steps[j].op = ceph_decode_32(p);
351 r->steps[j].arg1 = ceph_decode_32(p);
352 r->steps[j].arg2 = ceph_decode_32(p);
356 /* ignore trailing name maps. */
357 for (num_name_maps = 0; num_name_maps < 3; num_name_maps++) {
358 err = skip_name_map(p, end);
364 ceph_decode_need(p, end, 3*sizeof(u32), done);
365 c->choose_local_tries = ceph_decode_32(p);
366 c->choose_local_fallback_tries = ceph_decode_32(p);
367 c->choose_total_tries = ceph_decode_32(p);
368 dout("crush decode tunable choose_local_tries = %d\n",
369 c->choose_local_tries);
370 dout("crush decode tunable choose_local_fallback_tries = %d\n",
371 c->choose_local_fallback_tries);
372 dout("crush decode tunable choose_total_tries = %d\n",
373 c->choose_total_tries);
375 ceph_decode_need(p, end, sizeof(u32), done);
376 c->chooseleaf_descend_once = ceph_decode_32(p);
377 dout("crush decode tunable chooseleaf_descend_once = %d\n",
378 c->chooseleaf_descend_once);
380 ceph_decode_need(p, end, sizeof(u8), done);
381 c->chooseleaf_vary_r = ceph_decode_8(p);
382 dout("crush decode tunable chooseleaf_vary_r = %d\n",
383 c->chooseleaf_vary_r);
385 /* skip straw_calc_version, allowed_bucket_algs */
386 ceph_decode_need(p, end, sizeof(u8) + sizeof(u32), done);
387 *p += sizeof(u8) + sizeof(u32);
389 ceph_decode_need(p, end, sizeof(u8), done);
390 c->chooseleaf_stable = ceph_decode_8(p);
391 dout("crush decode tunable chooseleaf_stable = %d\n",
392 c->chooseleaf_stable);
396 dout("crush_decode success\n");
402 dout("crush_decode fail %d\n", err);
407 int ceph_pg_compare(const struct ceph_pg *lhs, const struct ceph_pg *rhs)
409 if (lhs->pool < rhs->pool)
411 if (lhs->pool > rhs->pool)
413 if (lhs->seed < rhs->seed)
415 if (lhs->seed > rhs->seed)
422 * rbtree of pg_mapping for handling pg_temp (explicit mapping of pgid
423 * to a set of osds) and primary_temp (explicit primary setting)
425 static int __insert_pg_mapping(struct ceph_pg_mapping *new,
426 struct rb_root *root)
428 struct rb_node **p = &root->rb_node;
429 struct rb_node *parent = NULL;
430 struct ceph_pg_mapping *pg = NULL;
433 dout("__insert_pg_mapping %llx %p\n", *(u64 *)&new->pgid, new);
436 pg = rb_entry(parent, struct ceph_pg_mapping, node);
437 c = ceph_pg_compare(&new->pgid, &pg->pgid);
446 rb_link_node(&new->node, parent, p);
447 rb_insert_color(&new->node, root);
451 static struct ceph_pg_mapping *__lookup_pg_mapping(struct rb_root *root,
454 struct rb_node *n = root->rb_node;
455 struct ceph_pg_mapping *pg;
459 pg = rb_entry(n, struct ceph_pg_mapping, node);
460 c = ceph_pg_compare(&pgid, &pg->pgid);
466 dout("__lookup_pg_mapping %lld.%x got %p\n",
467 pgid.pool, pgid.seed, pg);
474 static int __remove_pg_mapping(struct rb_root *root, struct ceph_pg pgid)
476 struct ceph_pg_mapping *pg = __lookup_pg_mapping(root, pgid);
479 dout("__remove_pg_mapping %lld.%x %p\n", pgid.pool, pgid.seed,
481 rb_erase(&pg->node, root);
485 dout("__remove_pg_mapping %lld.%x dne\n", pgid.pool, pgid.seed);
490 * rbtree of pg pool info
492 static int __insert_pg_pool(struct rb_root *root, struct ceph_pg_pool_info *new)
494 struct rb_node **p = &root->rb_node;
495 struct rb_node *parent = NULL;
496 struct ceph_pg_pool_info *pi = NULL;
500 pi = rb_entry(parent, struct ceph_pg_pool_info, node);
501 if (new->id < pi->id)
503 else if (new->id > pi->id)
509 rb_link_node(&new->node, parent, p);
510 rb_insert_color(&new->node, root);
514 static struct ceph_pg_pool_info *__lookup_pg_pool(struct rb_root *root, u64 id)
516 struct ceph_pg_pool_info *pi;
517 struct rb_node *n = root->rb_node;
520 pi = rb_entry(n, struct ceph_pg_pool_info, node);
523 else if (id > pi->id)
531 struct ceph_pg_pool_info *ceph_pg_pool_by_id(struct ceph_osdmap *map, u64 id)
533 return __lookup_pg_pool(&map->pg_pools, id);
536 const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id)
538 struct ceph_pg_pool_info *pi;
540 if (id == CEPH_NOPOOL)
543 if (WARN_ON_ONCE(id > (u64) INT_MAX))
546 pi = __lookup_pg_pool(&map->pg_pools, (int) id);
548 return pi ? pi->name : NULL;
550 EXPORT_SYMBOL(ceph_pg_pool_name_by_id);
552 int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name)
556 for (rbp = rb_first(&map->pg_pools); rbp; rbp = rb_next(rbp)) {
557 struct ceph_pg_pool_info *pi =
558 rb_entry(rbp, struct ceph_pg_pool_info, node);
559 if (pi->name && strcmp(pi->name, name) == 0)
564 EXPORT_SYMBOL(ceph_pg_poolid_by_name);
566 static void __remove_pg_pool(struct rb_root *root, struct ceph_pg_pool_info *pi)
568 rb_erase(&pi->node, root);
573 static int decode_pool(void **p, void *end, struct ceph_pg_pool_info *pi)
579 ceph_decode_need(p, end, 2 + 4, bad);
580 ev = ceph_decode_8(p); /* encoding version */
581 cv = ceph_decode_8(p); /* compat version */
583 pr_warn("got v %d < 5 cv %d of ceph_pg_pool\n", ev, cv);
587 pr_warn("got v %d cv %d > 9 of ceph_pg_pool\n", ev, cv);
590 len = ceph_decode_32(p);
591 ceph_decode_need(p, end, len, bad);
594 pi->type = ceph_decode_8(p);
595 pi->size = ceph_decode_8(p);
596 pi->crush_ruleset = ceph_decode_8(p);
597 pi->object_hash = ceph_decode_8(p);
599 pi->pg_num = ceph_decode_32(p);
600 pi->pgp_num = ceph_decode_32(p);
602 *p += 4 + 4; /* skip lpg* */
603 *p += 4; /* skip last_change */
604 *p += 8 + 4; /* skip snap_seq, snap_epoch */
607 num = ceph_decode_32(p);
609 *p += 8; /* snapid key */
610 *p += 1 + 1; /* versions */
611 len = ceph_decode_32(p);
615 /* skip removed_snaps */
616 num = ceph_decode_32(p);
619 *p += 8; /* skip auid */
620 pi->flags = ceph_decode_64(p);
621 *p += 4; /* skip crash_replay_interval */
624 pi->min_size = ceph_decode_8(p);
626 pi->min_size = pi->size - pi->size / 2;
629 *p += 8 + 8; /* skip quota_max_* */
633 num = ceph_decode_32(p);
636 *p += 8; /* skip tier_of */
637 *p += 1; /* skip cache_mode */
639 pi->read_tier = ceph_decode_64(p);
640 pi->write_tier = ceph_decode_64(p);
647 /* skip properties */
648 num = ceph_decode_32(p);
650 len = ceph_decode_32(p);
652 len = ceph_decode_32(p);
658 /* skip hit_set_params */
659 *p += 1 + 1; /* versions */
660 len = ceph_decode_32(p);
663 *p += 4; /* skip hit_set_period */
664 *p += 4; /* skip hit_set_count */
668 *p += 4; /* skip stripe_width */
671 *p += 8; /* skip target_max_bytes */
672 *p += 8; /* skip target_max_objects */
673 *p += 4; /* skip cache_target_dirty_ratio_micro */
674 *p += 4; /* skip cache_target_full_ratio_micro */
675 *p += 4; /* skip cache_min_flush_age */
676 *p += 4; /* skip cache_min_evict_age */
680 /* skip erasure_code_profile */
681 len = ceph_decode_32(p);
686 pi->last_force_request_resend = ceph_decode_32(p);
688 pi->last_force_request_resend = 0;
690 /* ignore the rest */
700 static int decode_pool_names(void **p, void *end, struct ceph_osdmap *map)
702 struct ceph_pg_pool_info *pi;
706 ceph_decode_32_safe(p, end, num, bad);
707 dout(" %d pool names\n", num);
709 ceph_decode_64_safe(p, end, pool, bad);
710 ceph_decode_32_safe(p, end, len, bad);
711 dout(" pool %llu len %d\n", pool, len);
712 ceph_decode_need(p, end, len, bad);
713 pi = __lookup_pg_pool(&map->pg_pools, pool);
715 char *name = kstrndup(*p, len, GFP_NOFS);
721 dout(" name is %s\n", pi->name);
734 struct ceph_osdmap *ceph_osdmap_alloc(void)
736 struct ceph_osdmap *map;
738 map = kzalloc(sizeof(*map), GFP_NOIO);
742 map->pg_pools = RB_ROOT;
744 map->pg_temp = RB_ROOT;
745 map->primary_temp = RB_ROOT;
746 mutex_init(&map->crush_workspace_mutex);
751 void ceph_osdmap_destroy(struct ceph_osdmap *map)
753 dout("osdmap_destroy %p\n", map);
755 crush_destroy(map->crush);
756 while (!RB_EMPTY_ROOT(&map->pg_temp)) {
757 struct ceph_pg_mapping *pg =
758 rb_entry(rb_first(&map->pg_temp),
759 struct ceph_pg_mapping, node);
760 rb_erase(&pg->node, &map->pg_temp);
763 while (!RB_EMPTY_ROOT(&map->primary_temp)) {
764 struct ceph_pg_mapping *pg =
765 rb_entry(rb_first(&map->primary_temp),
766 struct ceph_pg_mapping, node);
767 rb_erase(&pg->node, &map->primary_temp);
770 while (!RB_EMPTY_ROOT(&map->pg_pools)) {
771 struct ceph_pg_pool_info *pi =
772 rb_entry(rb_first(&map->pg_pools),
773 struct ceph_pg_pool_info, node);
774 __remove_pg_pool(&map->pg_pools, pi);
776 kfree(map->osd_state);
777 kfree(map->osd_weight);
778 kfree(map->osd_addr);
779 kfree(map->osd_primary_affinity);
780 kfree(map->crush_workspace);
785 * Adjust max_osd value, (re)allocate arrays.
787 * The new elements are properly initialized.
789 static int osdmap_set_max_osd(struct ceph_osdmap *map, int max)
793 struct ceph_entity_addr *addr;
796 state = krealloc(map->osd_state, max*sizeof(*state), GFP_NOFS);
799 map->osd_state = state;
801 weight = krealloc(map->osd_weight, max*sizeof(*weight), GFP_NOFS);
804 map->osd_weight = weight;
806 addr = krealloc(map->osd_addr, max*sizeof(*addr), GFP_NOFS);
809 map->osd_addr = addr;
811 for (i = map->max_osd; i < max; i++) {
812 map->osd_state[i] = 0;
813 map->osd_weight[i] = CEPH_OSD_OUT;
814 memset(map->osd_addr + i, 0, sizeof(*map->osd_addr));
817 if (map->osd_primary_affinity) {
820 affinity = krealloc(map->osd_primary_affinity,
821 max*sizeof(*affinity), GFP_NOFS);
824 map->osd_primary_affinity = affinity;
826 for (i = map->max_osd; i < max; i++)
827 map->osd_primary_affinity[i] =
828 CEPH_OSD_DEFAULT_PRIMARY_AFFINITY;
836 static int osdmap_set_crush(struct ceph_osdmap *map, struct crush_map *crush)
842 return PTR_ERR(crush);
844 work_size = crush_work_size(crush, CEPH_PG_MAX_SIZE);
845 dout("%s work_size %zu bytes\n", __func__, work_size);
846 workspace = kmalloc(work_size, GFP_NOIO);
848 crush_destroy(crush);
851 crush_init_workspace(crush, workspace);
854 crush_destroy(map->crush);
855 kfree(map->crush_workspace);
857 map->crush_workspace = workspace;
861 #define OSDMAP_WRAPPER_COMPAT_VER 7
862 #define OSDMAP_CLIENT_DATA_COMPAT_VER 1
865 * Return 0 or error. On success, *v is set to 0 for old (v6) osdmaps,
866 * to struct_v of the client_data section for new (v7 and above)
869 static int get_osdmap_client_data_v(void **p, void *end,
870 const char *prefix, u8 *v)
874 ceph_decode_8_safe(p, end, struct_v, e_inval);
878 ceph_decode_8_safe(p, end, struct_compat, e_inval);
879 if (struct_compat > OSDMAP_WRAPPER_COMPAT_VER) {
880 pr_warn("got v %d cv %d > %d of %s ceph_osdmap\n",
881 struct_v, struct_compat,
882 OSDMAP_WRAPPER_COMPAT_VER, prefix);
885 *p += 4; /* ignore wrapper struct_len */
887 ceph_decode_8_safe(p, end, struct_v, e_inval);
888 ceph_decode_8_safe(p, end, struct_compat, e_inval);
889 if (struct_compat > OSDMAP_CLIENT_DATA_COMPAT_VER) {
890 pr_warn("got v %d cv %d > %d of %s ceph_osdmap client data\n",
891 struct_v, struct_compat,
892 OSDMAP_CLIENT_DATA_COMPAT_VER, prefix);
895 *p += 4; /* ignore client data struct_len */
900 ceph_decode_16_safe(p, end, version, e_inval);
902 pr_warn("got v %d < 6 of %s ceph_osdmap\n",
907 /* old osdmap enconding */
918 static int __decode_pools(void **p, void *end, struct ceph_osdmap *map,
923 ceph_decode_32_safe(p, end, n, e_inval);
925 struct ceph_pg_pool_info *pi;
929 ceph_decode_64_safe(p, end, pool, e_inval);
931 pi = __lookup_pg_pool(&map->pg_pools, pool);
932 if (!incremental || !pi) {
933 pi = kzalloc(sizeof(*pi), GFP_NOFS);
939 ret = __insert_pg_pool(&map->pg_pools, pi);
946 ret = decode_pool(p, end, pi);
957 static int decode_pools(void **p, void *end, struct ceph_osdmap *map)
959 return __decode_pools(p, end, map, false);
962 static int decode_new_pools(void **p, void *end, struct ceph_osdmap *map)
964 return __decode_pools(p, end, map, true);
967 static int __decode_pg_temp(void **p, void *end, struct ceph_osdmap *map,
972 ceph_decode_32_safe(p, end, n, e_inval);
978 ret = ceph_decode_pgid(p, end, &pgid);
982 ceph_decode_32_safe(p, end, len, e_inval);
984 ret = __remove_pg_mapping(&map->pg_temp, pgid);
985 BUG_ON(!incremental && ret != -ENOENT);
987 if (!incremental || len > 0) {
988 struct ceph_pg_mapping *pg;
990 ceph_decode_need(p, end, len*sizeof(u32), e_inval);
992 if (len > (UINT_MAX - sizeof(*pg)) / sizeof(u32))
995 pg = kzalloc(sizeof(*pg) + len*sizeof(u32), GFP_NOFS);
1000 pg->pg_temp.len = len;
1001 for (i = 0; i < len; i++)
1002 pg->pg_temp.osds[i] = ceph_decode_32(p);
1004 ret = __insert_pg_mapping(pg, &map->pg_temp);
1018 static int decode_pg_temp(void **p, void *end, struct ceph_osdmap *map)
1020 return __decode_pg_temp(p, end, map, false);
1023 static int decode_new_pg_temp(void **p, void *end, struct ceph_osdmap *map)
1025 return __decode_pg_temp(p, end, map, true);
1028 static int __decode_primary_temp(void **p, void *end, struct ceph_osdmap *map,
1033 ceph_decode_32_safe(p, end, n, e_inval);
1035 struct ceph_pg pgid;
1039 ret = ceph_decode_pgid(p, end, &pgid);
1043 ceph_decode_32_safe(p, end, osd, e_inval);
1045 ret = __remove_pg_mapping(&map->primary_temp, pgid);
1046 BUG_ON(!incremental && ret != -ENOENT);
1048 if (!incremental || osd != (u32)-1) {
1049 struct ceph_pg_mapping *pg;
1051 pg = kzalloc(sizeof(*pg), GFP_NOFS);
1056 pg->primary_temp.osd = osd;
1058 ret = __insert_pg_mapping(pg, &map->primary_temp);
1072 static int decode_primary_temp(void **p, void *end, struct ceph_osdmap *map)
1074 return __decode_primary_temp(p, end, map, false);
1077 static int decode_new_primary_temp(void **p, void *end,
1078 struct ceph_osdmap *map)
1080 return __decode_primary_temp(p, end, map, true);
1083 u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd)
1085 BUG_ON(osd >= map->max_osd);
1087 if (!map->osd_primary_affinity)
1088 return CEPH_OSD_DEFAULT_PRIMARY_AFFINITY;
1090 return map->osd_primary_affinity[osd];
1093 static int set_primary_affinity(struct ceph_osdmap *map, int osd, u32 aff)
1095 BUG_ON(osd >= map->max_osd);
1097 if (!map->osd_primary_affinity) {
1100 map->osd_primary_affinity = kmalloc(map->max_osd*sizeof(u32),
1102 if (!map->osd_primary_affinity)
1105 for (i = 0; i < map->max_osd; i++)
1106 map->osd_primary_affinity[i] =
1107 CEPH_OSD_DEFAULT_PRIMARY_AFFINITY;
1110 map->osd_primary_affinity[osd] = aff;
1115 static int decode_primary_affinity(void **p, void *end,
1116 struct ceph_osdmap *map)
1120 ceph_decode_32_safe(p, end, len, e_inval);
1122 kfree(map->osd_primary_affinity);
1123 map->osd_primary_affinity = NULL;
1126 if (len != map->max_osd)
1129 ceph_decode_need(p, end, map->max_osd*sizeof(u32), e_inval);
1131 for (i = 0; i < map->max_osd; i++) {
1134 ret = set_primary_affinity(map, i, ceph_decode_32(p));
1145 static int decode_new_primary_affinity(void **p, void *end,
1146 struct ceph_osdmap *map)
1150 ceph_decode_32_safe(p, end, n, e_inval);
1155 ceph_decode_32_safe(p, end, osd, e_inval);
1156 ceph_decode_32_safe(p, end, aff, e_inval);
1158 ret = set_primary_affinity(map, osd, aff);
1162 pr_info("osd%d primary-affinity 0x%x\n", osd, aff);
1172 * decode a full map.
1174 static int osdmap_decode(void **p, void *end, struct ceph_osdmap *map)
1183 dout("%s %p to %p len %d\n", __func__, *p, end, (int)(end - *p));
1185 err = get_osdmap_client_data_v(p, end, "full", &struct_v);
1189 /* fsid, epoch, created, modified */
1190 ceph_decode_need(p, end, sizeof(map->fsid) + sizeof(u32) +
1191 sizeof(map->created) + sizeof(map->modified), e_inval);
1192 ceph_decode_copy(p, &map->fsid, sizeof(map->fsid));
1193 epoch = map->epoch = ceph_decode_32(p);
1194 ceph_decode_copy(p, &map->created, sizeof(map->created));
1195 ceph_decode_copy(p, &map->modified, sizeof(map->modified));
1198 err = decode_pools(p, end, map);
1203 err = decode_pool_names(p, end, map);
1207 ceph_decode_32_safe(p, end, map->pool_max, e_inval);
1209 ceph_decode_32_safe(p, end, map->flags, e_inval);
1212 ceph_decode_32_safe(p, end, max, e_inval);
1214 /* (re)alloc osd arrays */
1215 err = osdmap_set_max_osd(map, max);
1219 /* osd_state, osd_weight, osd_addrs->client_addr */
1220 ceph_decode_need(p, end, 3*sizeof(u32) +
1221 map->max_osd*(1 + sizeof(*map->osd_weight) +
1222 sizeof(*map->osd_addr)), e_inval);
1224 if (ceph_decode_32(p) != map->max_osd)
1227 ceph_decode_copy(p, map->osd_state, map->max_osd);
1229 if (ceph_decode_32(p) != map->max_osd)
1232 for (i = 0; i < map->max_osd; i++)
1233 map->osd_weight[i] = ceph_decode_32(p);
1235 if (ceph_decode_32(p) != map->max_osd)
1238 ceph_decode_copy(p, map->osd_addr, map->max_osd*sizeof(*map->osd_addr));
1239 for (i = 0; i < map->max_osd; i++)
1240 ceph_decode_addr(&map->osd_addr[i]);
1243 err = decode_pg_temp(p, end, map);
1248 if (struct_v >= 1) {
1249 err = decode_primary_temp(p, end, map);
1254 /* primary_affinity */
1255 if (struct_v >= 2) {
1256 err = decode_primary_affinity(p, end, map);
1260 /* XXX can this happen? */
1261 kfree(map->osd_primary_affinity);
1262 map->osd_primary_affinity = NULL;
1266 ceph_decode_32_safe(p, end, len, e_inval);
1267 err = osdmap_set_crush(map, crush_decode(*p, min(*p + len, end)));
1271 /* ignore the rest */
1274 dout("full osdmap epoch %d max_osd %d\n", map->epoch, map->max_osd);
1280 pr_err("corrupt full osdmap (%d) epoch %d off %d (%p of %p-%p)\n",
1281 err, epoch, (int)(*p - start), *p, start, end);
1282 print_hex_dump(KERN_DEBUG, "osdmap: ",
1283 DUMP_PREFIX_OFFSET, 16, 1,
1284 start, end - start, true);
1289 * Allocate and decode a full map.
1291 struct ceph_osdmap *ceph_osdmap_decode(void **p, void *end)
1293 struct ceph_osdmap *map;
1296 map = ceph_osdmap_alloc();
1298 return ERR_PTR(-ENOMEM);
1300 ret = osdmap_decode(p, end, map);
1302 ceph_osdmap_destroy(map);
1303 return ERR_PTR(ret);
1310 * Encoding order is (new_up_client, new_state, new_weight). Need to
1311 * apply in the (new_weight, new_state, new_up_client) order, because
1312 * an incremental map may look like e.g.
1314 * new_up_client: { osd=6, addr=... } # set osd_state and addr
1315 * new_state: { osd=6, xorstate=EXISTS } # clear osd_state
1317 static int decode_new_up_state_weight(void **p, void *end,
1318 struct ceph_osdmap *map)
1320 void *new_up_client;
1322 void *new_weight_end;
1326 ceph_decode_32_safe(p, end, len, e_inval);
1327 len *= sizeof(u32) + sizeof(struct ceph_entity_addr);
1328 ceph_decode_need(p, end, len, e_inval);
1332 ceph_decode_32_safe(p, end, len, e_inval);
1333 len *= sizeof(u32) + sizeof(u8);
1334 ceph_decode_need(p, end, len, e_inval);
1338 ceph_decode_32_safe(p, end, len, e_inval);
1343 ceph_decode_need(p, end, 2*sizeof(u32), e_inval);
1344 osd = ceph_decode_32(p);
1345 w = ceph_decode_32(p);
1346 BUG_ON(osd >= map->max_osd);
1347 pr_info("osd%d weight 0x%x %s\n", osd, w,
1348 w == CEPH_OSD_IN ? "(in)" :
1349 (w == CEPH_OSD_OUT ? "(out)" : ""));
1350 map->osd_weight[osd] = w;
1353 * If we are marking in, set the EXISTS, and clear the
1354 * AUTOOUT and NEW bits.
1357 map->osd_state[osd] |= CEPH_OSD_EXISTS;
1358 map->osd_state[osd] &= ~(CEPH_OSD_AUTOOUT |
1362 new_weight_end = *p;
1364 /* new_state (up/down) */
1366 len = ceph_decode_32(p);
1372 osd = ceph_decode_32(p);
1373 xorstate = ceph_decode_8(p);
1375 xorstate = CEPH_OSD_UP;
1376 BUG_ON(osd >= map->max_osd);
1377 if ((map->osd_state[osd] & CEPH_OSD_UP) &&
1378 (xorstate & CEPH_OSD_UP))
1379 pr_info("osd%d down\n", osd);
1380 if ((map->osd_state[osd] & CEPH_OSD_EXISTS) &&
1381 (xorstate & CEPH_OSD_EXISTS)) {
1382 pr_info("osd%d does not exist\n", osd);
1383 ret = set_primary_affinity(map, osd,
1384 CEPH_OSD_DEFAULT_PRIMARY_AFFINITY);
1387 memset(map->osd_addr + osd, 0, sizeof(*map->osd_addr));
1388 map->osd_state[osd] = 0;
1390 map->osd_state[osd] ^= xorstate;
1396 len = ceph_decode_32(p);
1399 struct ceph_entity_addr addr;
1401 osd = ceph_decode_32(p);
1402 ceph_decode_copy(p, &addr, sizeof(addr));
1403 ceph_decode_addr(&addr);
1404 BUG_ON(osd >= map->max_osd);
1405 pr_info("osd%d up\n", osd);
1406 map->osd_state[osd] |= CEPH_OSD_EXISTS | CEPH_OSD_UP;
1407 map->osd_addr[osd] = addr;
1410 *p = new_weight_end;
1418 * decode and apply an incremental map update.
1420 struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
1421 struct ceph_osdmap *map)
1423 struct ceph_fsid fsid;
1425 struct ceph_timespec modified;
1429 __s32 new_flags, max;
1434 dout("%s %p to %p len %d\n", __func__, *p, end, (int)(end - *p));
1436 err = get_osdmap_client_data_v(p, end, "inc", &struct_v);
1440 /* fsid, epoch, modified, new_pool_max, new_flags */
1441 ceph_decode_need(p, end, sizeof(fsid) + sizeof(u32) + sizeof(modified) +
1442 sizeof(u64) + sizeof(u32), e_inval);
1443 ceph_decode_copy(p, &fsid, sizeof(fsid));
1444 epoch = ceph_decode_32(p);
1445 BUG_ON(epoch != map->epoch+1);
1446 ceph_decode_copy(p, &modified, sizeof(modified));
1447 new_pool_max = ceph_decode_64(p);
1448 new_flags = ceph_decode_32(p);
1451 ceph_decode_32_safe(p, end, len, e_inval);
1453 dout("apply_incremental full map len %d, %p to %p\n",
1455 return ceph_osdmap_decode(p, min(*p+len, end));
1459 ceph_decode_32_safe(p, end, len, e_inval);
1461 err = osdmap_set_crush(map,
1462 crush_decode(*p, min(*p + len, end)));
1470 map->flags = new_flags;
1471 if (new_pool_max >= 0)
1472 map->pool_max = new_pool_max;
1475 ceph_decode_32_safe(p, end, max, e_inval);
1477 err = osdmap_set_max_osd(map, max);
1483 map->modified = modified;
1486 err = decode_new_pools(p, end, map);
1490 /* new_pool_names */
1491 err = decode_pool_names(p, end, map);
1496 ceph_decode_32_safe(p, end, len, e_inval);
1498 struct ceph_pg_pool_info *pi;
1500 ceph_decode_64_safe(p, end, pool, e_inval);
1501 pi = __lookup_pg_pool(&map->pg_pools, pool);
1503 __remove_pg_pool(&map->pg_pools, pi);
1506 /* new_up_client, new_state, new_weight */
1507 err = decode_new_up_state_weight(p, end, map);
1512 err = decode_new_pg_temp(p, end, map);
1516 /* new_primary_temp */
1517 if (struct_v >= 1) {
1518 err = decode_new_primary_temp(p, end, map);
1523 /* new_primary_affinity */
1524 if (struct_v >= 2) {
1525 err = decode_new_primary_affinity(p, end, map);
1530 /* ignore the rest */
1533 dout("inc osdmap epoch %d max_osd %d\n", map->epoch, map->max_osd);
1539 pr_err("corrupt inc osdmap (%d) epoch %d off %d (%p of %p-%p)\n",
1540 err, epoch, (int)(*p - start), *p, start, end);
1541 print_hex_dump(KERN_DEBUG, "osdmap: ",
1542 DUMP_PREFIX_OFFSET, 16, 1,
1543 start, end - start, true);
1544 return ERR_PTR(err);
1547 void ceph_oloc_copy(struct ceph_object_locator *dest,
1548 const struct ceph_object_locator *src)
1550 WARN_ON(!ceph_oloc_empty(dest));
1551 WARN_ON(dest->pool_ns); /* empty() only covers ->pool */
1553 dest->pool = src->pool;
1555 dest->pool_ns = ceph_get_string(src->pool_ns);
1557 EXPORT_SYMBOL(ceph_oloc_copy);
1559 void ceph_oloc_destroy(struct ceph_object_locator *oloc)
1561 ceph_put_string(oloc->pool_ns);
1563 EXPORT_SYMBOL(ceph_oloc_destroy);
1565 void ceph_oid_copy(struct ceph_object_id *dest,
1566 const struct ceph_object_id *src)
1568 WARN_ON(!ceph_oid_empty(dest));
1570 if (src->name != src->inline_name) {
1571 /* very rare, see ceph_object_id definition */
1572 dest->name = kmalloc(src->name_len + 1,
1573 GFP_NOIO | __GFP_NOFAIL);
1576 memcpy(dest->name, src->name, src->name_len + 1);
1577 dest->name_len = src->name_len;
1579 EXPORT_SYMBOL(ceph_oid_copy);
1581 static __printf(2, 0)
1582 int oid_printf_vargs(struct ceph_object_id *oid, const char *fmt, va_list ap)
1586 WARN_ON(!ceph_oid_empty(oid));
1588 len = vsnprintf(oid->inline_name, sizeof(oid->inline_name), fmt, ap);
1589 if (len >= sizeof(oid->inline_name))
1592 oid->name_len = len;
1597 * If oid doesn't fit into inline buffer, BUG.
1599 void ceph_oid_printf(struct ceph_object_id *oid, const char *fmt, ...)
1604 BUG_ON(oid_printf_vargs(oid, fmt, ap));
1607 EXPORT_SYMBOL(ceph_oid_printf);
1609 static __printf(3, 0)
1610 int oid_aprintf_vargs(struct ceph_object_id *oid, gfp_t gfp,
1611 const char *fmt, va_list ap)
1617 len = oid_printf_vargs(oid, fmt, aq);
1621 char *external_name;
1623 external_name = kmalloc(len + 1, gfp);
1627 oid->name = external_name;
1628 WARN_ON(vsnprintf(oid->name, len + 1, fmt, ap) != len);
1629 oid->name_len = len;
1636 * If oid doesn't fit into inline buffer, allocate.
1638 int ceph_oid_aprintf(struct ceph_object_id *oid, gfp_t gfp,
1639 const char *fmt, ...)
1645 ret = oid_aprintf_vargs(oid, gfp, fmt, ap);
1650 EXPORT_SYMBOL(ceph_oid_aprintf);
1652 void ceph_oid_destroy(struct ceph_object_id *oid)
1654 if (oid->name != oid->inline_name)
1657 EXPORT_SYMBOL(ceph_oid_destroy);
1662 static bool __osds_equal(const struct ceph_osds *lhs,
1663 const struct ceph_osds *rhs)
1665 if (lhs->size == rhs->size &&
1666 !memcmp(lhs->osds, rhs->osds, rhs->size * sizeof(rhs->osds[0])))
1675 static bool osds_equal(const struct ceph_osds *lhs,
1676 const struct ceph_osds *rhs)
1678 if (__osds_equal(lhs, rhs) &&
1679 lhs->primary == rhs->primary)
1685 static bool osds_valid(const struct ceph_osds *set)
1688 if (set->size > 0 && set->primary >= 0)
1691 /* empty can_shift_osds set */
1692 if (!set->size && set->primary == -1)
1695 /* empty !can_shift_osds set - all NONE */
1696 if (set->size > 0 && set->primary == -1) {
1699 for (i = 0; i < set->size; i++) {
1700 if (set->osds[i] != CRUSH_ITEM_NONE)
1710 void ceph_osds_copy(struct ceph_osds *dest, const struct ceph_osds *src)
1712 memcpy(dest->osds, src->osds, src->size * sizeof(src->osds[0]));
1713 dest->size = src->size;
1714 dest->primary = src->primary;
1717 static bool is_split(const struct ceph_pg *pgid,
1721 int old_bits = calc_bits_of(old_pg_num);
1722 int old_mask = (1 << old_bits) - 1;
1725 WARN_ON(pgid->seed >= old_pg_num);
1726 if (new_pg_num <= old_pg_num)
1729 for (n = 1; ; n++) {
1730 int next_bit = n << (old_bits - 1);
1731 u32 s = next_bit | pgid->seed;
1733 if (s < old_pg_num || s == pgid->seed)
1735 if (s >= new_pg_num)
1738 s = ceph_stable_mod(s, old_pg_num, old_mask);
1739 if (s == pgid->seed)
1746 bool ceph_is_new_interval(const struct ceph_osds *old_acting,
1747 const struct ceph_osds *new_acting,
1748 const struct ceph_osds *old_up,
1749 const struct ceph_osds *new_up,
1756 bool old_sort_bitwise,
1757 bool new_sort_bitwise,
1758 const struct ceph_pg *pgid)
1760 return !osds_equal(old_acting, new_acting) ||
1761 !osds_equal(old_up, new_up) ||
1762 old_size != new_size ||
1763 old_min_size != new_min_size ||
1764 is_split(pgid, old_pg_num, new_pg_num) ||
1765 old_sort_bitwise != new_sort_bitwise;
1768 static int calc_pg_rank(int osd, const struct ceph_osds *acting)
1772 for (i = 0; i < acting->size; i++) {
1773 if (acting->osds[i] == osd)
1780 static bool primary_changed(const struct ceph_osds *old_acting,
1781 const struct ceph_osds *new_acting)
1783 if (!old_acting->size && !new_acting->size)
1784 return false; /* both still empty */
1786 if (!old_acting->size ^ !new_acting->size)
1787 return true; /* was empty, now not, or vice versa */
1789 if (old_acting->primary != new_acting->primary)
1790 return true; /* primary changed */
1792 if (calc_pg_rank(old_acting->primary, old_acting) !=
1793 calc_pg_rank(new_acting->primary, new_acting))
1796 return false; /* same primary (tho replicas may have changed) */
1799 bool ceph_osds_changed(const struct ceph_osds *old_acting,
1800 const struct ceph_osds *new_acting,
1803 if (primary_changed(old_acting, new_acting))
1806 if (any_change && !__osds_equal(old_acting, new_acting))
1813 * calculate file layout from given offset, length.
1814 * fill in correct oid, logical length, and object extent
1817 * for now, we write only a single su, until we can
1818 * pass a stride back to the caller.
1820 int ceph_calc_file_object_mapping(struct ceph_file_layout *layout,
1823 u64 *oxoff, u64 *oxlen)
1825 u32 osize = layout->object_size;
1826 u32 su = layout->stripe_unit;
1827 u32 sc = layout->stripe_count;
1828 u32 bl, stripeno, stripepos, objsetno;
1832 dout("mapping %llu~%llu osize %u fl_su %u\n", off, len,
1834 if (su == 0 || sc == 0)
1836 su_per_object = osize / su;
1837 if (su_per_object == 0)
1839 dout("osize %u / su %u = su_per_object %u\n", osize, su,
1842 if ((su & ~PAGE_MASK) != 0)
1845 /* bl = *off / su; */
1849 dout("off %llu / su %u = bl %u\n", off, su, bl);
1852 stripepos = bl % sc;
1853 objsetno = stripeno / su_per_object;
1855 *ono = objsetno * sc + stripepos;
1856 dout("objset %u * sc %u = ono %u\n", objsetno, sc, (unsigned int)*ono);
1858 /* *oxoff = *off % layout->fl_stripe_unit; # offset in su */
1860 su_offset = do_div(t, su);
1861 *oxoff = su_offset + (stripeno % su_per_object) * su;
1864 * Calculate the length of the extent being written to the selected
1865 * object. This is the minimum of the full length requested (len) or
1866 * the remainder of the current stripe being written to.
1868 *oxlen = min_t(u64, len, su - su_offset);
1870 dout(" obj extent %llu~%llu\n", *oxoff, *oxlen);
1874 dout(" invalid layout\n");
1880 EXPORT_SYMBOL(ceph_calc_file_object_mapping);
1883 * Map an object into a PG.
1885 * Should only be called with target_oid and target_oloc (as opposed to
1886 * base_oid and base_oloc), since tiering isn't taken into account.
1888 int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap,
1889 struct ceph_object_id *oid,
1890 struct ceph_object_locator *oloc,
1891 struct ceph_pg *raw_pgid)
1893 struct ceph_pg_pool_info *pi;
1895 pi = ceph_pg_pool_by_id(osdmap, oloc->pool);
1899 if (!oloc->pool_ns) {
1900 raw_pgid->pool = oloc->pool;
1901 raw_pgid->seed = ceph_str_hash(pi->object_hash, oid->name,
1903 dout("%s %s -> raw_pgid %llu.%x\n", __func__, oid->name,
1904 raw_pgid->pool, raw_pgid->seed);
1906 char stack_buf[256];
1907 char *buf = stack_buf;
1908 int nsl = oloc->pool_ns->len;
1909 size_t total = nsl + 1 + oid->name_len;
1911 if (total > sizeof(stack_buf)) {
1912 buf = kmalloc(total, GFP_NOIO);
1916 memcpy(buf, oloc->pool_ns->str, nsl);
1918 memcpy(buf + nsl + 1, oid->name, oid->name_len);
1919 raw_pgid->pool = oloc->pool;
1920 raw_pgid->seed = ceph_str_hash(pi->object_hash, buf, total);
1921 if (buf != stack_buf)
1923 dout("%s %s ns %.*s -> raw_pgid %llu.%x\n", __func__,
1924 oid->name, nsl, oloc->pool_ns->str,
1925 raw_pgid->pool, raw_pgid->seed);
1929 EXPORT_SYMBOL(ceph_object_locator_to_pg);
1932 * Map a raw PG (full precision ps) into an actual PG.
1934 static void raw_pg_to_pg(struct ceph_pg_pool_info *pi,
1935 const struct ceph_pg *raw_pgid,
1936 struct ceph_pg *pgid)
1938 pgid->pool = raw_pgid->pool;
1939 pgid->seed = ceph_stable_mod(raw_pgid->seed, pi->pg_num,
1944 * Map a raw PG (full precision ps) into a placement ps (placement
1945 * seed). Include pool id in that value so that different pools don't
1946 * use the same seeds.
1948 static u32 raw_pg_to_pps(struct ceph_pg_pool_info *pi,
1949 const struct ceph_pg *raw_pgid)
1951 if (pi->flags & CEPH_POOL_FLAG_HASHPSPOOL) {
1952 /* hash pool id and seed so that pool PGs do not overlap */
1953 return crush_hash32_2(CRUSH_HASH_RJENKINS1,
1954 ceph_stable_mod(raw_pgid->seed,
1960 * legacy behavior: add ps and pool together. this is
1961 * not a great approach because the PGs from each pool
1962 * will overlap on top of each other: 0.5 == 1.4 ==
1965 return ceph_stable_mod(raw_pgid->seed, pi->pgp_num,
1967 (unsigned)raw_pgid->pool;
1971 static int do_crush(struct ceph_osdmap *map, int ruleno, int x,
1972 int *result, int result_max,
1973 const __u32 *weight, int weight_max)
1977 BUG_ON(result_max > CEPH_PG_MAX_SIZE);
1979 mutex_lock(&map->crush_workspace_mutex);
1980 r = crush_do_rule(map->crush, ruleno, x, result, result_max,
1981 weight, weight_max, map->crush_workspace);
1982 mutex_unlock(&map->crush_workspace_mutex);
1988 * Calculate raw set (CRUSH output) for given PG. The result may
1989 * contain nonexistent OSDs. ->primary is undefined for a raw set.
1991 * Placement seed (CRUSH input) is returned through @ppps.
1993 static void pg_to_raw_osds(struct ceph_osdmap *osdmap,
1994 struct ceph_pg_pool_info *pi,
1995 const struct ceph_pg *raw_pgid,
1996 struct ceph_osds *raw,
1999 u32 pps = raw_pg_to_pps(pi, raw_pgid);
2003 ceph_osds_init(raw);
2007 ruleno = crush_find_rule(osdmap->crush, pi->crush_ruleset, pi->type,
2010 pr_err("no crush rule: pool %lld ruleset %d type %d size %d\n",
2011 pi->id, pi->crush_ruleset, pi->type, pi->size);
2015 if (pi->size > ARRAY_SIZE(raw->osds)) {
2016 pr_err_ratelimited("pool %lld ruleset %d type %d too wide: size %d > %zu\n",
2017 pi->id, pi->crush_ruleset, pi->type, pi->size,
2018 ARRAY_SIZE(raw->osds));
2022 len = do_crush(osdmap, ruleno, pps, raw->osds, pi->size,
2023 osdmap->osd_weight, osdmap->max_osd);
2025 pr_err("error %d from crush rule %d: pool %lld ruleset %d type %d size %d\n",
2026 len, ruleno, pi->id, pi->crush_ruleset, pi->type,
2035 * Given raw set, calculate up set and up primary. By definition of an
2036 * up set, the result won't contain nonexistent or down OSDs.
2038 * This is done in-place - on return @set is the up set. If it's
2039 * empty, ->primary will remain undefined.
2041 static void raw_to_up_osds(struct ceph_osdmap *osdmap,
2042 struct ceph_pg_pool_info *pi,
2043 struct ceph_osds *set)
2047 /* ->primary is undefined for a raw set */
2048 BUG_ON(set->primary != -1);
2050 if (ceph_can_shift_osds(pi)) {
2054 for (i = 0; i < set->size; i++) {
2055 if (ceph_osd_is_down(osdmap, set->osds[i])) {
2060 set->osds[i - removed] = set->osds[i];
2062 set->size -= removed;
2064 set->primary = set->osds[0];
2066 /* set down/dne devices to NONE */
2067 for (i = set->size - 1; i >= 0; i--) {
2068 if (ceph_osd_is_down(osdmap, set->osds[i]))
2069 set->osds[i] = CRUSH_ITEM_NONE;
2071 set->primary = set->osds[i];
2076 static void apply_primary_affinity(struct ceph_osdmap *osdmap,
2077 struct ceph_pg_pool_info *pi,
2079 struct ceph_osds *up)
2085 * Do we have any non-default primary_affinity values for these
2088 if (!osdmap->osd_primary_affinity)
2091 for (i = 0; i < up->size; i++) {
2092 int osd = up->osds[i];
2094 if (osd != CRUSH_ITEM_NONE &&
2095 osdmap->osd_primary_affinity[osd] !=
2096 CEPH_OSD_DEFAULT_PRIMARY_AFFINITY) {
2104 * Pick the primary. Feed both the seed (for the pg) and the
2105 * osd into the hash/rng so that a proportional fraction of an
2106 * osd's pgs get rejected as primary.
2108 for (i = 0; i < up->size; i++) {
2109 int osd = up->osds[i];
2112 if (osd == CRUSH_ITEM_NONE)
2115 aff = osdmap->osd_primary_affinity[osd];
2116 if (aff < CEPH_OSD_MAX_PRIMARY_AFFINITY &&
2117 (crush_hash32_2(CRUSH_HASH_RJENKINS1,
2118 pps, osd) >> 16) >= aff) {
2120 * We chose not to use this primary. Note it
2121 * anyway as a fallback in case we don't pick
2122 * anyone else, but keep looking.
2134 up->primary = up->osds[pos];
2136 if (ceph_can_shift_osds(pi) && pos > 0) {
2137 /* move the new primary to the front */
2138 for (i = pos; i > 0; i--)
2139 up->osds[i] = up->osds[i - 1];
2140 up->osds[0] = up->primary;
2145 * Get pg_temp and primary_temp mappings for given PG.
2147 * Note that a PG may have none, only pg_temp, only primary_temp or
2148 * both pg_temp and primary_temp mappings. This means @temp isn't
2149 * always a valid OSD set on return: in the "only primary_temp" case,
2150 * @temp will have its ->primary >= 0 but ->size == 0.
2152 static void get_temp_osds(struct ceph_osdmap *osdmap,
2153 struct ceph_pg_pool_info *pi,
2154 const struct ceph_pg *raw_pgid,
2155 struct ceph_osds *temp)
2157 struct ceph_pg pgid;
2158 struct ceph_pg_mapping *pg;
2161 raw_pg_to_pg(pi, raw_pgid, &pgid);
2162 ceph_osds_init(temp);
2165 pg = __lookup_pg_mapping(&osdmap->pg_temp, pgid);
2167 for (i = 0; i < pg->pg_temp.len; i++) {
2168 if (ceph_osd_is_down(osdmap, pg->pg_temp.osds[i])) {
2169 if (ceph_can_shift_osds(pi))
2172 temp->osds[temp->size++] = CRUSH_ITEM_NONE;
2174 temp->osds[temp->size++] = pg->pg_temp.osds[i];
2178 /* apply pg_temp's primary */
2179 for (i = 0; i < temp->size; i++) {
2180 if (temp->osds[i] != CRUSH_ITEM_NONE) {
2181 temp->primary = temp->osds[i];
2188 pg = __lookup_pg_mapping(&osdmap->primary_temp, pgid);
2190 temp->primary = pg->primary_temp.osd;
2194 * Map a PG to its acting set as well as its up set.
2196 * Acting set is used for data mapping purposes, while up set can be
2197 * recorded for detecting interval changes and deciding whether to
2200 void ceph_pg_to_up_acting_osds(struct ceph_osdmap *osdmap,
2201 const struct ceph_pg *raw_pgid,
2202 struct ceph_osds *up,
2203 struct ceph_osds *acting)
2205 struct ceph_pg_pool_info *pi;
2208 pi = ceph_pg_pool_by_id(osdmap, raw_pgid->pool);
2211 ceph_osds_init(acting);
2215 pg_to_raw_osds(osdmap, pi, raw_pgid, up, &pps);
2216 raw_to_up_osds(osdmap, pi, up);
2217 apply_primary_affinity(osdmap, pi, pps, up);
2218 get_temp_osds(osdmap, pi, raw_pgid, acting);
2219 if (!acting->size) {
2220 memcpy(acting->osds, up->osds, up->size * sizeof(up->osds[0]));
2221 acting->size = up->size;
2222 if (acting->primary == -1)
2223 acting->primary = up->primary;
2226 WARN_ON(!osds_valid(up) || !osds_valid(acting));
2230 * Return acting primary for given PG, or -1 if none.
2232 int ceph_pg_to_acting_primary(struct ceph_osdmap *osdmap,
2233 const struct ceph_pg *raw_pgid)
2235 struct ceph_osds up, acting;
2237 ceph_pg_to_up_acting_osds(osdmap, raw_pgid, &up, &acting);
2238 return acting.primary;
2240 EXPORT_SYMBOL(ceph_pg_to_acting_primary);