~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ceph/osdmap.c

Version: ~ [ linux-5.13-rc5 ] ~ [ linux-5.12.9 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.42 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.124 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.193 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.235 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.271 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.271 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 
  2 #include <linux/ceph/ceph_debug.h>
  3 
  4 #include <linux/module.h>
  5 #include <linux/slab.h>
  6 #include <asm/div64.h>
  7 
  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>
 13 
 14 char *ceph_osdmap_state_str(char *str, int len, int state)
 15 {
 16         if (!len)
 17                 return str;
 18 
 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");
 25         else
 26                 snprintf(str, len, "doesn't exist");
 27 
 28         return str;
 29 }
 30 
 31 /* maps */
 32 
 33 static int calc_bits_of(unsigned int t)
 34 {
 35         int b = 0;
 36         while (t) {
 37                 t = t >> 1;
 38                 b++;
 39         }
 40         return b;
 41 }
 42 
 43 /*
 44  * the foo_mask is the smallest value 2^n-1 that is >= foo.
 45  */
 46 static void calc_pg_masks(struct ceph_pg_pool_info *pi)
 47 {
 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;
 50 }
 51 
 52 /*
 53  * decode crush map
 54  */
 55 static int crush_decode_uniform_bucket(void **p, void *end,
 56                                        struct crush_bucket_uniform *b)
 57 {
 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);
 61         return 0;
 62 bad:
 63         return -EINVAL;
 64 }
 65 
 66 static int crush_decode_list_bucket(void **p, void *end,
 67                                     struct crush_bucket_list *b)
 68 {
 69         int j;
 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)
 73                 return -ENOMEM;
 74         b->sum_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS);
 75         if (b->sum_weights == NULL)
 76                 return -ENOMEM;
 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);
 81         }
 82         return 0;
 83 bad:
 84         return -EINVAL;
 85 }
 86 
 87 static int crush_decode_tree_bucket(void **p, void *end,
 88                                     struct crush_bucket_tree *b)
 89 {
 90         int j;
 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)
 95                 return -ENOMEM;
 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);
 99         return 0;
100 bad:
101         return -EINVAL;
102 }
103 
104 static int crush_decode_straw_bucket(void **p, void *end,
105                                      struct crush_bucket_straw *b)
106 {
107         int j;
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)
111                 return -ENOMEM;
112         b->straws = kcalloc(b->h.size, sizeof(u32), GFP_NOFS);
113         if (b->straws == NULL)
114                 return -ENOMEM;
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);
119         }
120         return 0;
121 bad:
122         return -EINVAL;
123 }
124 
125 static int skip_name_map(void **p, void *end)
126 {
127         int len;
128         ceph_decode_32_safe(p, end, len ,bad);
129         while (len--) {
130                 int strlen;
131                 *p += sizeof(u32);
132                 ceph_decode_32_safe(p, end, strlen, bad);
133                 *p += strlen;
134 }
135         return 0;
136 bad:
137         return -EINVAL;
138 }
139 
140 static struct crush_map *crush_decode(void *pbyval, void *end)
141 {
142         struct crush_map *c;
143         int err = -EINVAL;
144         int i, j;
145         void **p = &pbyval;
146         void *start = pbyval;
147         u32 magic;
148         u32 num_name_maps;
149 
150         dout("crush_decode %p to %p len %d\n", *p, end, (int)(end - *p));
151 
152         c = kzalloc(sizeof(*c), GFP_NOFS);
153         if (c == NULL)
154                 return ERR_PTR(-ENOMEM);
155 
156         /* set tunables to default values */
157         c->choose_local_tries = 2;
158         c->choose_local_fallback_tries = 5;
159         c->choose_total_tries = 19;
160         c->chooseleaf_descend_once = 0;
161 
162         ceph_decode_need(p, end, 4*sizeof(u32), bad);
163         magic = ceph_decode_32(p);
164         if (magic != CRUSH_MAGIC) {
165                 pr_err("crush_decode magic %x != current %x\n",
166                        (unsigned int)magic, (unsigned int)CRUSH_MAGIC);
167                 goto bad;
168         }
169         c->max_buckets = ceph_decode_32(p);
170         c->max_rules = ceph_decode_32(p);
171         c->max_devices = ceph_decode_32(p);
172 
173         c->buckets = kcalloc(c->max_buckets, sizeof(*c->buckets), GFP_NOFS);
174         if (c->buckets == NULL)
175                 goto badmem;
176         c->rules = kcalloc(c->max_rules, sizeof(*c->rules), GFP_NOFS);
177         if (c->rules == NULL)
178                 goto badmem;
179 
180         /* buckets */
181         for (i = 0; i < c->max_buckets; i++) {
182                 int size = 0;
183                 u32 alg;
184                 struct crush_bucket *b;
185 
186                 ceph_decode_32_safe(p, end, alg, bad);
187                 if (alg == 0) {
188                         c->buckets[i] = NULL;
189                         continue;
190                 }
191                 dout("crush_decode bucket %d off %x %p to %p\n",
192                      i, (int)(*p-start), *p, end);
193 
194                 switch (alg) {
195                 case CRUSH_BUCKET_UNIFORM:
196                         size = sizeof(struct crush_bucket_uniform);
197                         break;
198                 case CRUSH_BUCKET_LIST:
199                         size = sizeof(struct crush_bucket_list);
200                         break;
201                 case CRUSH_BUCKET_TREE:
202                         size = sizeof(struct crush_bucket_tree);
203                         break;
204                 case CRUSH_BUCKET_STRAW:
205                         size = sizeof(struct crush_bucket_straw);
206                         break;
207                 default:
208                         err = -EINVAL;
209                         goto bad;
210                 }
211                 BUG_ON(size == 0);
212                 b = c->buckets[i] = kzalloc(size, GFP_NOFS);
213                 if (b == NULL)
214                         goto badmem;
215 
216                 ceph_decode_need(p, end, 4*sizeof(u32), bad);
217                 b->id = ceph_decode_32(p);
218                 b->type = ceph_decode_16(p);
219                 b->alg = ceph_decode_8(p);
220                 b->hash = ceph_decode_8(p);
221                 b->weight = ceph_decode_32(p);
222                 b->size = ceph_decode_32(p);
223 
224                 dout("crush_decode bucket size %d off %x %p to %p\n",
225                      b->size, (int)(*p-start), *p, end);
226 
227                 b->items = kcalloc(b->size, sizeof(__s32), GFP_NOFS);
228                 if (b->items == NULL)
229                         goto badmem;
230                 b->perm = kcalloc(b->size, sizeof(u32), GFP_NOFS);
231                 if (b->perm == NULL)
232                         goto badmem;
233                 b->perm_n = 0;
234 
235                 ceph_decode_need(p, end, b->size*sizeof(u32), bad);
236                 for (j = 0; j < b->size; j++)
237                         b->items[j] = ceph_decode_32(p);
238 
239                 switch (b->alg) {
240                 case CRUSH_BUCKET_UNIFORM:
241                         err = crush_decode_uniform_bucket(p, end,
242                                   (struct crush_bucket_uniform *)b);
243                         if (err < 0)
244                                 goto bad;
245                         break;
246                 case CRUSH_BUCKET_LIST:
247                         err = crush_decode_list_bucket(p, end,
248                                (struct crush_bucket_list *)b);
249                         if (err < 0)
250                                 goto bad;
251                         break;
252                 case CRUSH_BUCKET_TREE:
253                         err = crush_decode_tree_bucket(p, end,
254                                 (struct crush_bucket_tree *)b);
255                         if (err < 0)
256                                 goto bad;
257                         break;
258                 case CRUSH_BUCKET_STRAW:
259                         err = crush_decode_straw_bucket(p, end,
260                                 (struct crush_bucket_straw *)b);
261                         if (err < 0)
262                                 goto bad;
263                         break;
264                 }
265         }
266 
267         /* rules */
268         dout("rule vec is %p\n", c->rules);
269         for (i = 0; i < c->max_rules; i++) {
270                 u32 yes;
271                 struct crush_rule *r;
272 
273                 ceph_decode_32_safe(p, end, yes, bad);
274                 if (!yes) {
275                         dout("crush_decode NO rule %d off %x %p to %p\n",
276                              i, (int)(*p-start), *p, end);
277                         c->rules[i] = NULL;
278                         continue;
279                 }
280 
281                 dout("crush_decode rule %d off %x %p to %p\n",
282                      i, (int)(*p-start), *p, end);
283 
284                 /* len */
285                 ceph_decode_32_safe(p, end, yes, bad);
286 #if BITS_PER_LONG == 32
287                 err = -EINVAL;
288                 if (yes > (ULONG_MAX - sizeof(*r))
289                           / sizeof(struct crush_rule_step))
290                         goto bad;
291 #endif
292                 r = c->rules[i] = kmalloc(sizeof(*r) +
293                                           yes*sizeof(struct crush_rule_step),
294                                           GFP_NOFS);
295                 if (r == NULL)
296                         goto badmem;
297                 dout(" rule %d is at %p\n", i, r);
298                 r->len = yes;
299                 ceph_decode_copy_safe(p, end, &r->mask, 4, bad); /* 4 u8's */
300                 ceph_decode_need(p, end, r->len*3*sizeof(u32), bad);
301                 for (j = 0; j < r->len; j++) {
302                         r->steps[j].op = ceph_decode_32(p);
303                         r->steps[j].arg1 = ceph_decode_32(p);
304                         r->steps[j].arg2 = ceph_decode_32(p);
305                 }
306         }
307 
308         /* ignore trailing name maps. */
309         for (num_name_maps = 0; num_name_maps < 3; num_name_maps++) {
310                 err = skip_name_map(p, end);
311                 if (err < 0)
312                         goto done;
313         }
314 
315         /* tunables */
316         ceph_decode_need(p, end, 3*sizeof(u32), done);
317         c->choose_local_tries = ceph_decode_32(p);
318         c->choose_local_fallback_tries =  ceph_decode_32(p);
319         c->choose_total_tries = ceph_decode_32(p);
320         dout("crush decode tunable choose_local_tries = %d",
321              c->choose_local_tries);
322         dout("crush decode tunable choose_local_fallback_tries = %d",
323              c->choose_local_fallback_tries);
324         dout("crush decode tunable choose_total_tries = %d",
325              c->choose_total_tries);
326 
327         ceph_decode_need(p, end, sizeof(u32), done);
328         c->chooseleaf_descend_once = ceph_decode_32(p);
329         dout("crush decode tunable chooseleaf_descend_once = %d",
330              c->chooseleaf_descend_once);
331 
332 done:
333         dout("crush_decode success\n");
334         return c;
335 
336 badmem:
337         err = -ENOMEM;
338 bad:
339         dout("crush_decode fail %d\n", err);
340         crush_destroy(c);
341         return ERR_PTR(err);
342 }
343 
344 /*
345  * rbtree of pg_mapping for handling pg_temp (explicit mapping of pgid
346  * to a set of osds)
347  */
348 static int pgid_cmp(struct ceph_pg l, struct ceph_pg r)
349 {
350         if (l.pool < r.pool)
351                 return -1;
352         if (l.pool > r.pool)
353                 return 1;
354         if (l.seed < r.seed)
355                 return -1;
356         if (l.seed > r.seed)
357                 return 1;
358         return 0;
359 }
360 
361 static int __insert_pg_mapping(struct ceph_pg_mapping *new,
362                                struct rb_root *root)
363 {
364         struct rb_node **p = &root->rb_node;
365         struct rb_node *parent = NULL;
366         struct ceph_pg_mapping *pg = NULL;
367         int c;
368 
369         dout("__insert_pg_mapping %llx %p\n", *(u64 *)&new->pgid, new);
370         while (*p) {
371                 parent = *p;
372                 pg = rb_entry(parent, struct ceph_pg_mapping, node);
373                 c = pgid_cmp(new->pgid, pg->pgid);
374                 if (c < 0)
375                         p = &(*p)->rb_left;
376                 else if (c > 0)
377                         p = &(*p)->rb_right;
378                 else
379                         return -EEXIST;
380         }
381 
382         rb_link_node(&new->node, parent, p);
383         rb_insert_color(&new->node, root);
384         return 0;
385 }
386 
387 static struct ceph_pg_mapping *__lookup_pg_mapping(struct rb_root *root,
388                                                    struct ceph_pg pgid)
389 {
390         struct rb_node *n = root->rb_node;
391         struct ceph_pg_mapping *pg;
392         int c;
393 
394         while (n) {
395                 pg = rb_entry(n, struct ceph_pg_mapping, node);
396                 c = pgid_cmp(pgid, pg->pgid);
397                 if (c < 0) {
398                         n = n->rb_left;
399                 } else if (c > 0) {
400                         n = n->rb_right;
401                 } else {
402                         dout("__lookup_pg_mapping %lld.%x got %p\n",
403                              pgid.pool, pgid.seed, pg);
404                         return pg;
405                 }
406         }
407         return NULL;
408 }
409 
410 static int __remove_pg_mapping(struct rb_root *root, struct ceph_pg pgid)
411 {
412         struct ceph_pg_mapping *pg = __lookup_pg_mapping(root, pgid);
413 
414         if (pg) {
415                 dout("__remove_pg_mapping %lld.%x %p\n", pgid.pool, pgid.seed,
416                      pg);
417                 rb_erase(&pg->node, root);
418                 kfree(pg);
419                 return 0;
420         }
421         dout("__remove_pg_mapping %lld.%x dne\n", pgid.pool, pgid.seed);
422         return -ENOENT;
423 }
424 
425 /*
426  * rbtree of pg pool info
427  */
428 static int __insert_pg_pool(struct rb_root *root, struct ceph_pg_pool_info *new)
429 {
430         struct rb_node **p = &root->rb_node;
431         struct rb_node *parent = NULL;
432         struct ceph_pg_pool_info *pi = NULL;
433 
434         while (*p) {
435                 parent = *p;
436                 pi = rb_entry(parent, struct ceph_pg_pool_info, node);
437                 if (new->id < pi->id)
438                         p = &(*p)->rb_left;
439                 else if (new->id > pi->id)
440                         p = &(*p)->rb_right;
441                 else
442                         return -EEXIST;
443         }
444 
445         rb_link_node(&new->node, parent, p);
446         rb_insert_color(&new->node, root);
447         return 0;
448 }
449 
450 static struct ceph_pg_pool_info *__lookup_pg_pool(struct rb_root *root, u64 id)
451 {
452         struct ceph_pg_pool_info *pi;
453         struct rb_node *n = root->rb_node;
454 
455         while (n) {
456                 pi = rb_entry(n, struct ceph_pg_pool_info, node);
457                 if (id < pi->id)
458                         n = n->rb_left;
459                 else if (id > pi->id)
460                         n = n->rb_right;
461                 else
462                         return pi;
463         }
464         return NULL;
465 }
466 
467 struct ceph_pg_pool_info *ceph_pg_pool_by_id(struct ceph_osdmap *map, u64 id)
468 {
469         return __lookup_pg_pool(&map->pg_pools, id);
470 }
471 
472 const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id)
473 {
474         struct ceph_pg_pool_info *pi;
475 
476         if (id == CEPH_NOPOOL)
477                 return NULL;
478 
479         if (WARN_ON_ONCE(id > (u64) INT_MAX))
480                 return NULL;
481 
482         pi = __lookup_pg_pool(&map->pg_pools, (int) id);
483 
484         return pi ? pi->name : NULL;
485 }
486 EXPORT_SYMBOL(ceph_pg_pool_name_by_id);
487 
488 int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name)
489 {
490         struct rb_node *rbp;
491 
492         for (rbp = rb_first(&map->pg_pools); rbp; rbp = rb_next(rbp)) {
493                 struct ceph_pg_pool_info *pi =
494                         rb_entry(rbp, struct ceph_pg_pool_info, node);
495                 if (pi->name && strcmp(pi->name, name) == 0)
496                         return pi->id;
497         }
498         return -ENOENT;
499 }
500 EXPORT_SYMBOL(ceph_pg_poolid_by_name);
501 
502 static void __remove_pg_pool(struct rb_root *root, struct ceph_pg_pool_info *pi)
503 {
504         rb_erase(&pi->node, root);
505         kfree(pi->name);
506         kfree(pi);
507 }
508 
509 static int __decode_pool(void **p, void *end, struct ceph_pg_pool_info *pi)
510 {
511         u8 ev, cv;
512         unsigned len, num;
513         void *pool_end;
514 
515         ceph_decode_need(p, end, 2 + 4, bad);
516         ev = ceph_decode_8(p);  /* encoding version */
517         cv = ceph_decode_8(p); /* compat version */
518         if (ev < 5) {
519                 pr_warning("got v %d < 5 cv %d of ceph_pg_pool\n", ev, cv);
520                 return -EINVAL;
521         }
522         if (cv > 9) {
523                 pr_warning("got v %d cv %d > 9 of ceph_pg_pool\n", ev, cv);
524                 return -EINVAL;
525         }
526         len = ceph_decode_32(p);
527         ceph_decode_need(p, end, len, bad);
528         pool_end = *p + len;
529 
530         pi->type = ceph_decode_8(p);
531         pi->size = ceph_decode_8(p);
532         pi->crush_ruleset = ceph_decode_8(p);
533         pi->object_hash = ceph_decode_8(p);
534 
535         pi->pg_num = ceph_decode_32(p);
536         pi->pgp_num = ceph_decode_32(p);
537 
538         *p += 4 + 4;  /* skip lpg* */
539         *p += 4;      /* skip last_change */
540         *p += 8 + 4;  /* skip snap_seq, snap_epoch */
541 
542         /* skip snaps */
543         num = ceph_decode_32(p);
544         while (num--) {
545                 *p += 8;  /* snapid key */
546                 *p += 1 + 1; /* versions */
547                 len = ceph_decode_32(p);
548                 *p += len;
549         }
550 
551         /* skip removed_snaps */
552         num = ceph_decode_32(p);
553         *p += num * (8 + 8);
554 
555         *p += 8;  /* skip auid */
556         pi->flags = ceph_decode_64(p);
557         *p += 4;  /* skip crash_replay_interval */
558 
559         if (ev >= 7)
560                 *p += 1;  /* skip min_size */
561 
562         if (ev >= 8)
563                 *p += 8 + 8;  /* skip quota_max_* */
564 
565         if (ev >= 9) {
566                 /* skip tiers */
567                 num = ceph_decode_32(p);
568                 *p += num * 8;
569 
570                 *p += 8;  /* skip tier_of */
571                 *p += 1;  /* skip cache_mode */
572 
573                 pi->read_tier = ceph_decode_64(p);
574                 pi->write_tier = ceph_decode_64(p);
575         } else {
576                 pi->read_tier = -1;
577                 pi->write_tier = -1;
578         }
579 
580         /* ignore the rest */
581 
582         *p = pool_end;
583         calc_pg_masks(pi);
584         return 0;
585 
586 bad:
587         return -EINVAL;
588 }
589 
590 static int __decode_pool_names(void **p, void *end, struct ceph_osdmap *map)
591 {
592         struct ceph_pg_pool_info *pi;
593         u32 num, len;
594         u64 pool;
595 
596         ceph_decode_32_safe(p, end, num, bad);
597         dout(" %d pool names\n", num);
598         while (num--) {
599                 ceph_decode_64_safe(p, end, pool, bad);
600                 ceph_decode_32_safe(p, end, len, bad);
601                 dout("  pool %llu len %d\n", pool, len);
602                 ceph_decode_need(p, end, len, bad);
603                 pi = __lookup_pg_pool(&map->pg_pools, pool);
604                 if (pi) {
605                         char *name = kstrndup(*p, len, GFP_NOFS);
606 
607                         if (!name)
608                                 return -ENOMEM;
609                         kfree(pi->name);
610                         pi->name = name;
611                         dout("  name is %s\n", pi->name);
612                 }
613                 *p += len;
614         }
615         return 0;
616 
617 bad:
618         return -EINVAL;
619 }
620 
621 /*
622  * osd map
623  */
624 void ceph_osdmap_destroy(struct ceph_osdmap *map)
625 {
626         dout("osdmap_destroy %p\n", map);
627         if (map->crush)
628                 crush_destroy(map->crush);
629         while (!RB_EMPTY_ROOT(&map->pg_temp)) {
630                 struct ceph_pg_mapping *pg =
631                         rb_entry(rb_first(&map->pg_temp),
632                                  struct ceph_pg_mapping, node);
633                 rb_erase(&pg->node, &map->pg_temp);
634                 kfree(pg);
635         }
636         while (!RB_EMPTY_ROOT(&map->pg_pools)) {
637                 struct ceph_pg_pool_info *pi =
638                         rb_entry(rb_first(&map->pg_pools),
639                                  struct ceph_pg_pool_info, node);
640                 __remove_pg_pool(&map->pg_pools, pi);
641         }
642         kfree(map->osd_state);
643         kfree(map->osd_weight);
644         kfree(map->osd_addr);
645         kfree(map);
646 }
647 
648 /*
649  * adjust max osd value.  reallocate arrays.
650  */
651 static int osdmap_set_max_osd(struct ceph_osdmap *map, int max)
652 {
653         u8 *state;
654         struct ceph_entity_addr *addr;
655         u32 *weight;
656 
657         state = kcalloc(max, sizeof(*state), GFP_NOFS);
658         addr = kcalloc(max, sizeof(*addr), GFP_NOFS);
659         weight = kcalloc(max, sizeof(*weight), GFP_NOFS);
660         if (state == NULL || addr == NULL || weight == NULL) {
661                 kfree(state);
662                 kfree(addr);
663                 kfree(weight);
664                 return -ENOMEM;
665         }
666 
667         /* copy old? */
668         if (map->osd_state) {
669                 memcpy(state, map->osd_state, map->max_osd*sizeof(*state));
670                 memcpy(addr, map->osd_addr, map->max_osd*sizeof(*addr));
671                 memcpy(weight, map->osd_weight, map->max_osd*sizeof(*weight));
672                 kfree(map->osd_state);
673                 kfree(map->osd_addr);
674                 kfree(map->osd_weight);
675         }
676 
677         map->osd_state = state;
678         map->osd_weight = weight;
679         map->osd_addr = addr;
680         map->max_osd = max;
681         return 0;
682 }
683 
684 /*
685  * decode a full map.
686  */
687 struct ceph_osdmap *osdmap_decode(void **p, void *end)
688 {
689         struct ceph_osdmap *map;
690         u16 version;
691         u32 len, max, i;
692         int err = -EINVAL;
693         void *start = *p;
694         struct ceph_pg_pool_info *pi;
695 
696         dout("osdmap_decode %p to %p len %d\n", *p, end, (int)(end - *p));
697 
698         map = kzalloc(sizeof(*map), GFP_NOFS);
699         if (map == NULL)
700                 return ERR_PTR(-ENOMEM);
701         map->pg_temp = RB_ROOT;
702 
703         ceph_decode_16_safe(p, end, version, bad);
704         if (version > 6) {
705                 pr_warning("got unknown v %d > 6 of osdmap\n", version);
706                 goto bad;
707         }
708         if (version < 6) {
709                 pr_warning("got old v %d < 6 of osdmap\n", version);
710                 goto bad;
711         }
712 
713         ceph_decode_need(p, end, 2*sizeof(u64)+6*sizeof(u32), bad);
714         ceph_decode_copy(p, &map->fsid, sizeof(map->fsid));
715         map->epoch = ceph_decode_32(p);
716         ceph_decode_copy(p, &map->created, sizeof(map->created));
717         ceph_decode_copy(p, &map->modified, sizeof(map->modified));
718 
719         ceph_decode_32_safe(p, end, max, bad);
720         while (max--) {
721                 ceph_decode_need(p, end, 8 + 2, bad);
722                 err = -ENOMEM;
723                 pi = kzalloc(sizeof(*pi), GFP_NOFS);
724                 if (!pi)
725                         goto bad;
726                 pi->id = ceph_decode_64(p);
727                 err = __decode_pool(p, end, pi);
728                 if (err < 0) {
729                         kfree(pi);
730                         goto bad;
731                 }
732                 __insert_pg_pool(&map->pg_pools, pi);
733         }
734 
735         err = __decode_pool_names(p, end, map);
736         if (err < 0) {
737                 dout("fail to decode pool names");
738                 goto bad;
739         }
740 
741         ceph_decode_32_safe(p, end, map->pool_max, bad);
742 
743         ceph_decode_32_safe(p, end, map->flags, bad);
744 
745         max = ceph_decode_32(p);
746 
747         /* (re)alloc osd arrays */
748         err = osdmap_set_max_osd(map, max);
749         if (err < 0)
750                 goto bad;
751         dout("osdmap_decode max_osd = %d\n", map->max_osd);
752 
753         /* osds */
754         err = -EINVAL;
755         ceph_decode_need(p, end, 3*sizeof(u32) +
756                          map->max_osd*(1 + sizeof(*map->osd_weight) +
757                                        sizeof(*map->osd_addr)), bad);
758         *p += 4; /* skip length field (should match max) */
759         ceph_decode_copy(p, map->osd_state, map->max_osd);
760 
761         *p += 4; /* skip length field (should match max) */
762         for (i = 0; i < map->max_osd; i++)
763                 map->osd_weight[i] = ceph_decode_32(p);
764 
765         *p += 4; /* skip length field (should match max) */
766         ceph_decode_copy(p, map->osd_addr, map->max_osd*sizeof(*map->osd_addr));
767         for (i = 0; i < map->max_osd; i++)
768                 ceph_decode_addr(&map->osd_addr[i]);
769 
770         /* pg_temp */
771         ceph_decode_32_safe(p, end, len, bad);
772         for (i = 0; i < len; i++) {
773                 int n, j;
774                 struct ceph_pg pgid;
775                 struct ceph_pg_mapping *pg;
776 
777                 err = ceph_decode_pgid(p, end, &pgid);
778                 if (err)
779                         goto bad;
780                 ceph_decode_need(p, end, sizeof(u32), bad);
781                 n = ceph_decode_32(p);
782                 err = -EINVAL;
783                 if (n > (UINT_MAX - sizeof(*pg)) / sizeof(u32))
784                         goto bad;
785                 ceph_decode_need(p, end, n * sizeof(u32), bad);
786                 err = -ENOMEM;
787                 pg = kmalloc(sizeof(*pg) + n*sizeof(u32), GFP_NOFS);
788                 if (!pg)
789                         goto bad;
790                 pg->pgid = pgid;
791                 pg->len = n;
792                 for (j = 0; j < n; j++)
793                         pg->osds[j] = ceph_decode_32(p);
794 
795                 err = __insert_pg_mapping(pg, &map->pg_temp);
796                 if (err)
797                         goto bad;
798                 dout(" added pg_temp %lld.%x len %d\n", pgid.pool, pgid.seed,
799                      len);
800         }
801 
802         /* crush */
803         ceph_decode_32_safe(p, end, len, bad);
804         dout("osdmap_decode crush len %d from off 0x%x\n", len,
805              (int)(*p - start));
806         ceph_decode_need(p, end, len, bad);
807         map->crush = crush_decode(*p, end);
808         *p += len;
809         if (IS_ERR(map->crush)) {
810                 err = PTR_ERR(map->crush);
811                 map->crush = NULL;
812                 goto bad;
813         }
814 
815         /* ignore the rest of the map */
816         *p = end;
817 
818         dout("osdmap_decode done %p %p\n", *p, end);
819         return map;
820 
821 bad:
822         dout("osdmap_decode fail err %d\n", err);
823         ceph_osdmap_destroy(map);
824         return ERR_PTR(err);
825 }
826 
827 /*
828  * Encoding order is (new_up_client, new_state, new_weight).  Need to
829  * apply in the (new_weight, new_state, new_up_client) order, because
830  * an incremental map may look like e.g.
831  *
832  *     new_up_client: { osd=6, addr=... } # set osd_state and addr
833  *     new_state: { osd=6, xorstate=EXISTS } # clear osd_state
834  */
835 static int decode_new_up_state_weight(void **p, void *end,
836                                       struct ceph_osdmap *map)
837 {
838         void *new_up_client;
839         void *new_state;
840         void *new_weight_end;
841         u32 len;
842 
843         new_up_client = *p;
844         ceph_decode_32_safe(p, end, len, e_inval);
845         len *= sizeof(u32) + sizeof(struct ceph_entity_addr);
846         ceph_decode_need(p, end, len, e_inval);
847         *p += len;
848 
849         new_state = *p;
850         ceph_decode_32_safe(p, end, len, e_inval);
851         len *= sizeof(u32) + sizeof(u8);
852         ceph_decode_need(p, end, len, e_inval);
853         *p += len;
854 
855         /* new_weight */
856         ceph_decode_32_safe(p, end, len, e_inval);
857         while (len--) {
858                 s32 osd;
859                 u32 w;
860 
861                 ceph_decode_need(p, end, 2*sizeof(u32), e_inval);
862                 osd = ceph_decode_32(p);
863                 w = ceph_decode_32(p);
864                 BUG_ON(osd >= map->max_osd);
865                 pr_info("osd%d weight 0x%x %s\n", osd, w,
866                      w == CEPH_OSD_IN ? "(in)" :
867                      (w == CEPH_OSD_OUT ? "(out)" : ""));
868                 map->osd_weight[osd] = w;
869 
870                 /*
871                  * If we are marking in, set the EXISTS, and clear the
872                  * AUTOOUT and NEW bits.
873                  */
874                 if (w) {
875                         map->osd_state[osd] |= CEPH_OSD_EXISTS;
876                         map->osd_state[osd] &= ~(CEPH_OSD_AUTOOUT |
877                                                  CEPH_OSD_NEW);
878                 }
879         }
880         new_weight_end = *p;
881 
882         /* new_state (up/down) */
883         *p = new_state;
884         len = ceph_decode_32(p);
885         while (len--) {
886                 s32 osd;
887                 u8 xorstate;
888 
889                 osd = ceph_decode_32(p);
890                 xorstate = ceph_decode_8(p);
891                 if (xorstate == 0)
892                         xorstate = CEPH_OSD_UP;
893                 BUG_ON(osd >= map->max_osd);
894                 if ((map->osd_state[osd] & CEPH_OSD_UP) &&
895                     (xorstate & CEPH_OSD_UP))
896                         pr_info("osd%d down\n", osd);
897                 if ((map->osd_state[osd] & CEPH_OSD_EXISTS) &&
898                     (xorstate & CEPH_OSD_EXISTS)) {
899                         pr_info("osd%d does not exist\n", osd);
900                         map->osd_weight[osd] = CEPH_OSD_IN;
901                         memset(map->osd_addr + osd, 0, sizeof(*map->osd_addr));
902                         map->osd_state[osd] = 0;
903                 } else {
904                         map->osd_state[osd] ^= xorstate;
905                 }
906         }
907 
908         /* new_up_client */
909         *p = new_up_client;
910         len = ceph_decode_32(p);
911         while (len--) {
912                 s32 osd;
913                 struct ceph_entity_addr addr;
914 
915                 osd = ceph_decode_32(p);
916                 ceph_decode_copy(p, &addr, sizeof(addr));
917                 ceph_decode_addr(&addr);
918                 BUG_ON(osd >= map->max_osd);
919                 pr_info("osd%d up\n", osd);
920                 map->osd_state[osd] |= CEPH_OSD_EXISTS | CEPH_OSD_UP;
921                 map->osd_addr[osd] = addr;
922         }
923 
924         *p = new_weight_end;
925         return 0;
926 
927 e_inval:
928         return -EINVAL;
929 }
930 
931 /*
932  * decode and apply an incremental map update.
933  */
934 struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
935                                              struct ceph_osdmap *map,
936                                              struct ceph_messenger *msgr)
937 {
938         struct crush_map *newcrush = NULL;
939         struct ceph_fsid fsid;
940         u32 epoch = 0;
941         struct ceph_timespec modified;
942         s32 len;
943         u64 pool;
944         __s64 new_pool_max;
945         __s32 new_flags, max;
946         void *start = *p;
947         int err = -EINVAL;
948         u16 version;
949 
950         ceph_decode_16_safe(p, end, version, bad);
951         if (version != 6) {
952                 pr_warning("got unknown v %d != 6 of inc osdmap\n", version);
953                 goto bad;
954         }
955 
956         ceph_decode_need(p, end, sizeof(fsid)+sizeof(modified)+2*sizeof(u32),
957                          bad);
958         ceph_decode_copy(p, &fsid, sizeof(fsid));
959         epoch = ceph_decode_32(p);
960         BUG_ON(epoch != map->epoch+1);
961         ceph_decode_copy(p, &modified, sizeof(modified));
962         new_pool_max = ceph_decode_64(p);
963         new_flags = ceph_decode_32(p);
964 
965         /* full map? */
966         ceph_decode_32_safe(p, end, len, bad);
967         if (len > 0) {
968                 dout("apply_incremental full map len %d, %p to %p\n",
969                      len, *p, end);
970                 return osdmap_decode(p, min(*p+len, end));
971         }
972 
973         /* new crush? */
974         ceph_decode_32_safe(p, end, len, bad);
975         if (len > 0) {
976                 dout("apply_incremental new crush map len %d, %p to %p\n",
977                      len, *p, end);
978                 newcrush = crush_decode(*p, min(*p+len, end));
979                 if (IS_ERR(newcrush))
980                         return ERR_CAST(newcrush);
981                 *p += len;
982         }
983 
984         /* new flags? */
985         if (new_flags >= 0)
986                 map->flags = new_flags;
987         if (new_pool_max >= 0)
988                 map->pool_max = new_pool_max;
989 
990         ceph_decode_need(p, end, 5*sizeof(u32), bad);
991 
992         /* new max? */
993         max = ceph_decode_32(p);
994         if (max >= 0) {
995                 err = osdmap_set_max_osd(map, max);
996                 if (err < 0)
997                         goto bad;
998         }
999 
1000         map->epoch++;
1001         map->modified = modified;
1002         if (newcrush) {
1003                 if (map->crush)
1004                         crush_destroy(map->crush);
1005                 map->crush = newcrush;
1006                 newcrush = NULL;
1007         }
1008 
1009         /* new_pool */
1010         ceph_decode_32_safe(p, end, len, bad);
1011         while (len--) {
1012                 struct ceph_pg_pool_info *pi;
1013 
1014                 ceph_decode_64_safe(p, end, pool, bad);
1015                 pi = __lookup_pg_pool(&map->pg_pools, pool);
1016                 if (!pi) {
1017                         pi = kzalloc(sizeof(*pi), GFP_NOFS);
1018                         if (!pi) {
1019                                 err = -ENOMEM;
1020                                 goto bad;
1021                         }
1022                         pi->id = pool;
1023                         __insert_pg_pool(&map->pg_pools, pi);
1024                 }
1025                 err = __decode_pool(p, end, pi);
1026                 if (err < 0)
1027                         goto bad;
1028         }
1029         if (version >= 5) {
1030                 err = __decode_pool_names(p, end, map);
1031                 if (err < 0)
1032                         goto bad;
1033         }
1034 
1035         /* old_pool */
1036         ceph_decode_32_safe(p, end, len, bad);
1037         while (len--) {
1038                 struct ceph_pg_pool_info *pi;
1039 
1040                 ceph_decode_64_safe(p, end, pool, bad);
1041                 pi = __lookup_pg_pool(&map->pg_pools, pool);
1042                 if (pi)
1043                         __remove_pg_pool(&map->pg_pools, pi);
1044         }
1045 
1046         /* new_up_client, new_state, new_weight */
1047         err = decode_new_up_state_weight(p, end, map);
1048         if (err)
1049                 goto bad;
1050 
1051         /* new_pg_temp */
1052         ceph_decode_32_safe(p, end, len, bad);
1053         while (len--) {
1054                 struct ceph_pg_mapping *pg;
1055                 int j;
1056                 struct ceph_pg pgid;
1057                 u32 pglen;
1058 
1059                 err = ceph_decode_pgid(p, end, &pgid);
1060                 if (err)
1061                         goto bad;
1062                 ceph_decode_need(p, end, sizeof(u32), bad);
1063                 pglen = ceph_decode_32(p);
1064                 if (pglen) {
1065                         ceph_decode_need(p, end, pglen*sizeof(u32), bad);
1066 
1067                         /* removing existing (if any) */
1068                         (void) __remove_pg_mapping(&map->pg_temp, pgid);
1069 
1070                         /* insert */
1071                         err = -EINVAL;
1072                         if (pglen > (UINT_MAX - sizeof(*pg)) / sizeof(u32))
1073                                 goto bad;
1074                         err = -ENOMEM;
1075                         pg = kmalloc(sizeof(*pg) + sizeof(u32)*pglen, GFP_NOFS);
1076                         if (!pg)
1077                                 goto bad;
1078                         pg->pgid = pgid;
1079                         pg->len = pglen;
1080                         for (j = 0; j < pglen; j++)
1081                                 pg->osds[j] = ceph_decode_32(p);
1082                         err = __insert_pg_mapping(pg, &map->pg_temp);
1083                         if (err) {
1084                                 kfree(pg);
1085                                 goto bad;
1086                         }
1087                         dout(" added pg_temp %lld.%x len %d\n", pgid.pool,
1088                              pgid.seed, pglen);
1089                 } else {
1090                         /* remove */
1091                         __remove_pg_mapping(&map->pg_temp, pgid);
1092                 }
1093         }
1094 
1095         /* ignore the rest */
1096         *p = end;
1097         return map;
1098 
1099 bad:
1100         pr_err("corrupt inc osdmap epoch %d off %d (%p of %p-%p)\n",
1101                epoch, (int)(*p - start), *p, start, end);
1102         print_hex_dump(KERN_DEBUG, "osdmap: ",
1103                        DUMP_PREFIX_OFFSET, 16, 1,
1104                        start, end - start, true);
1105         if (newcrush)
1106                 crush_destroy(newcrush);
1107         return ERR_PTR(err);
1108 }
1109 
1110 
1111 
1112 
1113 /*
1114  * calculate file layout from given offset, length.
1115  * fill in correct oid, logical length, and object extent
1116  * offset, length.
1117  *
1118  * for now, we write only a single su, until we can
1119  * pass a stride back to the caller.
1120  */
1121 int ceph_calc_file_object_mapping(struct ceph_file_layout *layout,
1122                                    u64 off, u64 len,
1123                                    u64 *ono,
1124                                    u64 *oxoff, u64 *oxlen)
1125 {
1126         u32 osize = le32_to_cpu(layout->fl_object_size);
1127         u32 su = le32_to_cpu(layout->fl_stripe_unit);
1128         u32 sc = le32_to_cpu(layout->fl_stripe_count);
1129         u32 bl, stripeno, stripepos, objsetno;
1130         u32 su_per_object;
1131         u64 t, su_offset;
1132 
1133         dout("mapping %llu~%llu  osize %u fl_su %u\n", off, len,
1134              osize, su);
1135         if (su == 0 || sc == 0)
1136                 goto invalid;
1137         su_per_object = osize / su;
1138         if (su_per_object == 0)
1139                 goto invalid;
1140         dout("osize %u / su %u = su_per_object %u\n", osize, su,
1141              su_per_object);
1142 
1143         if ((su & ~PAGE_MASK) != 0)
1144                 goto invalid;
1145 
1146         /* bl = *off / su; */
1147         t = off;
1148         do_div(t, su);
1149         bl = t;
1150         dout("off %llu / su %u = bl %u\n", off, su, bl);
1151 
1152         stripeno = bl / sc;
1153         stripepos = bl % sc;
1154         objsetno = stripeno / su_per_object;
1155 
1156         *ono = objsetno * sc + stripepos;
1157         dout("objset %u * sc %u = ono %u\n", objsetno, sc, (unsigned int)*ono);
1158 
1159         /* *oxoff = *off % layout->fl_stripe_unit;  # offset in su */
1160         t = off;
1161         su_offset = do_div(t, su);
1162         *oxoff = su_offset + (stripeno % su_per_object) * su;
1163 
1164         /*
1165          * Calculate the length of the extent being written to the selected
1166          * object. This is the minimum of the full length requested (len) or
1167          * the remainder of the current stripe being written to.
1168          */
1169         *oxlen = min_t(u64, len, su - su_offset);
1170 
1171         dout(" obj extent %llu~%llu\n", *oxoff, *oxlen);
1172         return 0;
1173 
1174 invalid:
1175         dout(" invalid layout\n");
1176         *ono = 0;
1177         *oxoff = 0;
1178         *oxlen = 0;
1179         return -EINVAL;
1180 }
1181 EXPORT_SYMBOL(ceph_calc_file_object_mapping);
1182 
1183 /*
1184  * Calculate mapping of a (oloc, oid) pair to a PG.  Should only be
1185  * called with target's (oloc, oid), since tiering isn't taken into
1186  * account.
1187  */
1188 int ceph_oloc_oid_to_pg(struct ceph_osdmap *osdmap,
1189                         struct ceph_object_locator *oloc,
1190                         struct ceph_object_id *oid,
1191                         struct ceph_pg *pg_out)
1192 {
1193         struct ceph_pg_pool_info *pi;
1194 
1195         pi = __lookup_pg_pool(&osdmap->pg_pools, oloc->pool);
1196         if (!pi)
1197                 return -EIO;
1198 
1199         pg_out->pool = oloc->pool;
1200         pg_out->seed = ceph_str_hash(pi->object_hash, oid->name,
1201                                      oid->name_len);
1202 
1203         dout("%s '%.*s' pgid %llu.%x\n", __func__, oid->name_len, oid->name,
1204              pg_out->pool, pg_out->seed);
1205         return 0;
1206 }
1207 EXPORT_SYMBOL(ceph_oloc_oid_to_pg);
1208 
1209 static int crush_do_rule_ary(const struct crush_map *map, int ruleno, int x,
1210                              int *result, int result_max,
1211                              const __u32 *weight, int weight_max)
1212 {
1213         int scratch[result_max * 3];
1214 
1215         return crush_do_rule(map, ruleno, x, result, result_max,
1216                              weight, weight_max, scratch);
1217 }
1218 
1219 /*
1220  * Calculate raw osd vector for the given pgid.  Return pointer to osd
1221  * array, or NULL on failure.
1222  */
1223 static int *calc_pg_raw(struct ceph_osdmap *osdmap, struct ceph_pg pgid,
1224                         int *osds, int *num)
1225 {
1226         struct ceph_pg_mapping *pg;
1227         struct ceph_pg_pool_info *pool;
1228         int ruleno;
1229         int r;
1230         u32 pps;
1231 
1232         pool = __lookup_pg_pool(&osdmap->pg_pools, pgid.pool);
1233         if (!pool)
1234                 return NULL;
1235 
1236         /* pg_temp? */
1237         pgid.seed = ceph_stable_mod(pgid.seed, pool->pg_num,
1238                                     pool->pg_num_mask);
1239         pg = __lookup_pg_mapping(&osdmap->pg_temp, pgid);
1240         if (pg) {
1241                 *num = pg->len;
1242                 return pg->osds;
1243         }
1244 
1245         /* crush */
1246         ruleno = crush_find_rule(osdmap->crush, pool->crush_ruleset,
1247                                  pool->type, pool->size);
1248         if (ruleno < 0) {
1249                 pr_err("no crush rule pool %lld ruleset %d type %d size %d\n",
1250                        pgid.pool, pool->crush_ruleset, pool->type,
1251                        pool->size);
1252                 return NULL;
1253         }
1254 
1255         if (pool->flags & CEPH_POOL_FLAG_HASHPSPOOL) {
1256                 /* hash pool id and seed sothat pool PGs do not overlap */
1257                 pps = crush_hash32_2(CRUSH_HASH_RJENKINS1,
1258                                      ceph_stable_mod(pgid.seed, pool->pgp_num,
1259                                                      pool->pgp_num_mask),
1260                                      pgid.pool);
1261         } else {
1262                 /*
1263                  * legacy ehavior: add ps and pool together.  this is
1264                  * not a great approach because the PGs from each pool
1265                  * will overlap on top of each other: 0.5 == 1.4 ==
1266                  * 2.3 == ...
1267                  */
1268                 pps = ceph_stable_mod(pgid.seed, pool->pgp_num,
1269                                       pool->pgp_num_mask) +
1270                         (unsigned)pgid.pool;
1271         }
1272         r = crush_do_rule_ary(osdmap->crush, ruleno, pps,
1273                               osds, min_t(int, pool->size, *num),
1274                               osdmap->osd_weight, osdmap->max_osd);
1275         if (r < 0) {
1276                 pr_err("error %d from crush rule: pool %lld ruleset %d type %d"
1277                        " size %d\n", r, pgid.pool, pool->crush_ruleset,
1278                        pool->type, pool->size);
1279                 return NULL;
1280         }
1281         *num = r;
1282         return osds;
1283 }
1284 
1285 /*
1286  * Return acting set for given pgid.
1287  */
1288 int ceph_calc_pg_acting(struct ceph_osdmap *osdmap, struct ceph_pg pgid,
1289                         int *acting)
1290 {
1291         int rawosds[CEPH_PG_MAX_SIZE], *osds;
1292         int i, o, num = CEPH_PG_MAX_SIZE;
1293 
1294         osds = calc_pg_raw(osdmap, pgid, rawosds, &num);
1295         if (!osds)
1296                 return -1;
1297 
1298         /* primary is first up osd */
1299         o = 0;
1300         for (i = 0; i < num; i++)
1301                 if (ceph_osd_is_up(osdmap, osds[i]))
1302                         acting[o++] = osds[i];
1303         return o;
1304 }
1305 
1306 /*
1307  * Return primary osd for given pgid, or -1 if none.
1308  */
1309 int ceph_calc_pg_primary(struct ceph_osdmap *osdmap, struct ceph_pg pgid)
1310 {
1311         int rawosds[CEPH_PG_MAX_SIZE], *osds;
1312         int i, num = CEPH_PG_MAX_SIZE;
1313 
1314         osds = calc_pg_raw(osdmap, pgid, rawosds, &num);
1315         if (!osds)
1316                 return -1;
1317 
1318         /* primary is first up osd */
1319         for (i = 0; i < num; i++)
1320                 if (ceph_osd_is_up(osdmap, osds[i]))
1321                         return osds[i];
1322         return -1;
1323 }
1324 EXPORT_SYMBOL(ceph_calc_pg_primary);
1325 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp