]>
Commit | Line | Data |
---|---|---|
1 | // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*- | |
2 | // vim: ts=8 sw=2 smarttab | |
3 | ||
4 | #include <errno.h> | |
5 | ||
6 | #include <string> | |
7 | #include <map> | |
8 | #include <sstream> | |
9 | ||
10 | #include <boost/utility/string_ref.hpp> | |
11 | #include <boost/format.hpp> | |
12 | ||
13 | #include "common/errno.h" | |
14 | #include "common/ceph_json.h" | |
15 | #include "rgw_rados.h" | |
16 | #include "rgw_acl.h" | |
17 | #include "rgw_acl_s3.h" | |
18 | ||
19 | #include "include/types.h" | |
20 | #include "rgw_bucket.h" | |
21 | #include "rgw_user.h" | |
22 | #include "rgw_string.h" | |
23 | #include "rgw_multi.h" | |
24 | ||
25 | #include "include/rados/librados.hpp" | |
26 | // until everything is moved from rgw_common | |
27 | #include "rgw_common.h" | |
28 | ||
29 | #include "cls/user/cls_user_types.h" | |
30 | ||
31 | #define dout_context g_ceph_context | |
32 | #define dout_subsys ceph_subsys_rgw | |
33 | ||
34 | #define BUCKET_TAG_TIMEOUT 30 | |
35 | ||
36 | using namespace std; | |
37 | ||
38 | static RGWMetadataHandler *bucket_meta_handler = NULL; | |
39 | static RGWMetadataHandler *bucket_instance_meta_handler = NULL; | |
40 | ||
41 | // define as static when RGWBucket implementation compete | |
42 | void rgw_get_buckets_obj(const rgw_user& user_id, string& buckets_obj_id) | |
43 | { | |
44 | buckets_obj_id = user_id.to_str(); | |
45 | buckets_obj_id += RGW_BUCKETS_OBJ_SUFFIX; | |
46 | } | |
47 | ||
48 | /* | |
49 | * Note that this is not a reversal of parse_bucket(). That one deals | |
50 | * with the syntax we need in metadata and such. This one deals with | |
51 | * the representation in RADOS pools. We chose '/' because it's not | |
52 | * acceptable in bucket names and thus qualified buckets cannot conflict | |
53 | * with the legacy or S3 buckets. | |
54 | */ | |
55 | std::string rgw_make_bucket_entry_name(const std::string& tenant_name, | |
56 | const std::string& bucket_name) { | |
57 | std::string bucket_entry; | |
58 | ||
59 | if (bucket_name.empty()) { | |
60 | bucket_entry.clear(); | |
61 | } else if (tenant_name.empty()) { | |
62 | bucket_entry = bucket_name; | |
63 | } else { | |
64 | bucket_entry = tenant_name + "/" + bucket_name; | |
65 | } | |
66 | ||
67 | return bucket_entry; | |
68 | } | |
69 | ||
70 | /* | |
71 | * Tenants are separated from buckets in URLs by a colon in S3. | |
72 | * This function is not to be used on Swift URLs, not even for COPY arguments. | |
73 | */ | |
74 | void rgw_parse_url_bucket(const string &bucket, const string& auth_tenant, | |
75 | string &tenant_name, string &bucket_name) { | |
76 | ||
77 | int pos = bucket.find(':'); | |
78 | if (pos >= 0) { | |
79 | /* | |
80 | * N.B.: We allow ":bucket" syntax with explicit empty tenant in order | |
81 | * to refer to the legacy tenant, in case users in new named tenants | |
82 | * want to access old global buckets. | |
83 | */ | |
84 | tenant_name = bucket.substr(0, pos); | |
85 | bucket_name = bucket.substr(pos + 1); | |
86 | } else { | |
87 | tenant_name = auth_tenant; | |
88 | bucket_name = bucket; | |
89 | } | |
90 | } | |
91 | ||
92 | /** | |
93 | * Get all the buckets owned by a user and fill up an RGWUserBuckets with them. | |
94 | * Returns: 0 on success, -ERR# on failure. | |
95 | */ | |
96 | int rgw_read_user_buckets(RGWRados * store, | |
97 | const rgw_user& user_id, | |
98 | RGWUserBuckets& buckets, | |
99 | const string& marker, | |
100 | const string& end_marker, | |
101 | uint64_t max, | |
102 | bool need_stats, | |
103 | bool *is_truncated, | |
104 | uint64_t default_amount) | |
105 | { | |
106 | int ret; | |
107 | buckets.clear(); | |
108 | string buckets_obj_id; | |
109 | rgw_get_buckets_obj(user_id, buckets_obj_id); | |
110 | rgw_raw_obj obj(store->get_zone_params().user_uid_pool, buckets_obj_id); | |
111 | list<cls_user_bucket_entry> entries; | |
112 | ||
113 | bool truncated = false; | |
114 | string m = marker; | |
115 | ||
116 | uint64_t total = 0; | |
117 | ||
118 | if (!max) { | |
119 | max = default_amount; | |
120 | } | |
121 | ||
122 | do { | |
123 | ret = store->cls_user_list_buckets(obj, m, end_marker, max - total, entries, &m, &truncated); | |
124 | if (ret == -ENOENT) | |
125 | ret = 0; | |
126 | ||
127 | if (ret < 0) | |
128 | return ret; | |
129 | ||
130 | for (const auto& entry : entries) { | |
131 | buckets.add(RGWBucketEnt(user_id, entry)); | |
132 | total++; | |
133 | } | |
134 | ||
135 | } while (truncated && total < max); | |
136 | ||
137 | if (is_truncated != nullptr) { | |
138 | *is_truncated = truncated; | |
139 | } | |
140 | ||
141 | if (need_stats) { | |
142 | map<string, RGWBucketEnt>& m = buckets.get_buckets(); | |
143 | ret = store->update_containers_stats(m); | |
144 | if (ret < 0 && ret != -ENOENT) { | |
145 | ldout(store->ctx(), 0) << "ERROR: could not get stats for buckets" << dendl; | |
146 | return ret; | |
147 | } | |
148 | } | |
149 | return 0; | |
150 | } | |
151 | ||
152 | int rgw_bucket_sync_user_stats(RGWRados *store, const rgw_user& user_id, const RGWBucketInfo& bucket_info) | |
153 | { | |
154 | string buckets_obj_id; | |
155 | rgw_get_buckets_obj(user_id, buckets_obj_id); | |
156 | rgw_raw_obj obj(store->get_zone_params().user_uid_pool, buckets_obj_id); | |
157 | ||
158 | return store->cls_user_sync_bucket_stats(obj, bucket_info); | |
159 | } | |
160 | ||
161 | int rgw_bucket_sync_user_stats(RGWRados *store, const string& tenant_name, const string& bucket_name) | |
162 | { | |
163 | RGWBucketInfo bucket_info; | |
164 | RGWObjectCtx obj_ctx(store); | |
165 | int ret = store->get_bucket_info(obj_ctx, tenant_name, bucket_name, bucket_info, NULL); | |
166 | if (ret < 0) { | |
167 | ldout(store->ctx(), 0) << "ERROR: could not fetch bucket info: ret=" << ret << dendl; | |
168 | return ret; | |
169 | } | |
170 | ||
171 | ret = rgw_bucket_sync_user_stats(store, bucket_info.owner, bucket_info); | |
172 | if (ret < 0) { | |
173 | ldout(store->ctx(), 0) << "ERROR: could not sync user stats for bucket " << bucket_name << ": ret=" << ret << dendl; | |
174 | return ret; | |
175 | } | |
176 | ||
177 | return 0; | |
178 | } | |
179 | ||
180 | int rgw_link_bucket(RGWRados *store, const rgw_user& user_id, rgw_bucket& bucket, real_time creation_time, bool update_entrypoint) | |
181 | { | |
182 | int ret; | |
183 | string& tenant_name = bucket.tenant; | |
184 | string& bucket_name = bucket.name; | |
185 | ||
186 | cls_user_bucket_entry new_bucket; | |
187 | ||
188 | RGWBucketEntryPoint ep; | |
189 | RGWObjVersionTracker ot; | |
190 | ||
191 | bucket.convert(&new_bucket.bucket); | |
192 | new_bucket.size = 0; | |
193 | if (real_clock::is_zero(creation_time)) | |
194 | new_bucket.creation_time = real_clock::now(); | |
195 | else | |
196 | new_bucket.creation_time = creation_time; | |
197 | ||
198 | map<string, bufferlist> attrs; | |
199 | RGWObjectCtx obj_ctx(store); | |
200 | ||
201 | if (update_entrypoint) { | |
202 | ret = store->get_bucket_entrypoint_info(obj_ctx, tenant_name, bucket_name, ep, &ot, NULL, &attrs); | |
203 | if (ret < 0 && ret != -ENOENT) { | |
204 | ldout(store->ctx(), 0) << "ERROR: store->get_bucket_entrypoint_info() returned: " | |
205 | << cpp_strerror(-ret) << dendl; | |
206 | } | |
207 | } | |
208 | ||
209 | string buckets_obj_id; | |
210 | rgw_get_buckets_obj(user_id, buckets_obj_id); | |
211 | ||
212 | rgw_raw_obj obj(store->get_zone_params().user_uid_pool, buckets_obj_id); | |
213 | ret = store->cls_user_add_bucket(obj, new_bucket); | |
214 | if (ret < 0) { | |
215 | ldout(store->ctx(), 0) << "ERROR: error adding bucket to directory: " | |
216 | << cpp_strerror(-ret) << dendl; | |
217 | goto done_err; | |
218 | } | |
219 | ||
220 | if (!update_entrypoint) | |
221 | return 0; | |
222 | ||
223 | ep.linked = true; | |
224 | ep.owner = user_id; | |
225 | ep.bucket = bucket; | |
226 | ret = store->put_bucket_entrypoint_info(tenant_name, bucket_name, ep, false, ot, real_time(), &attrs); | |
227 | if (ret < 0) | |
228 | goto done_err; | |
229 | ||
230 | return 0; | |
231 | done_err: | |
232 | int r = rgw_unlink_bucket(store, user_id, bucket.tenant, bucket.name); | |
233 | if (r < 0) { | |
234 | ldout(store->ctx(), 0) << "ERROR: failed unlinking bucket on error cleanup: " | |
235 | << cpp_strerror(-r) << dendl; | |
236 | } | |
237 | return ret; | |
238 | } | |
239 | ||
240 | int rgw_unlink_bucket(RGWRados *store, const rgw_user& user_id, const string& tenant_name, const string& bucket_name, bool update_entrypoint) | |
241 | { | |
242 | int ret; | |
243 | ||
244 | string buckets_obj_id; | |
245 | rgw_get_buckets_obj(user_id, buckets_obj_id); | |
246 | ||
247 | cls_user_bucket bucket; | |
248 | bucket.name = bucket_name; | |
249 | rgw_raw_obj obj(store->get_zone_params().user_uid_pool, buckets_obj_id); | |
250 | ret = store->cls_user_remove_bucket(obj, bucket); | |
251 | if (ret < 0) { | |
252 | ldout(store->ctx(), 0) << "ERROR: error removing bucket from directory: " | |
253 | << cpp_strerror(-ret)<< dendl; | |
254 | } | |
255 | ||
256 | if (!update_entrypoint) | |
257 | return 0; | |
258 | ||
259 | RGWBucketEntryPoint ep; | |
260 | RGWObjVersionTracker ot; | |
261 | map<string, bufferlist> attrs; | |
262 | RGWObjectCtx obj_ctx(store); | |
263 | ret = store->get_bucket_entrypoint_info(obj_ctx, tenant_name, bucket_name, ep, &ot, NULL, &attrs); | |
264 | if (ret == -ENOENT) | |
265 | return 0; | |
266 | if (ret < 0) | |
267 | return ret; | |
268 | ||
269 | if (!ep.linked) | |
270 | return 0; | |
271 | ||
272 | if (ep.owner != user_id) { | |
273 | ldout(store->ctx(), 0) << "bucket entry point user mismatch, can't unlink bucket: " << ep.owner << " != " << user_id << dendl; | |
274 | return -EINVAL; | |
275 | } | |
276 | ||
277 | ep.linked = false; | |
278 | return store->put_bucket_entrypoint_info(tenant_name, bucket_name, ep, false, ot, real_time(), &attrs); | |
279 | } | |
280 | ||
281 | int rgw_bucket_store_info(RGWRados *store, const string& bucket_name, bufferlist& bl, bool exclusive, | |
282 | map<string, bufferlist> *pattrs, RGWObjVersionTracker *objv_tracker, | |
283 | real_time mtime) { | |
284 | return store->meta_mgr->put_entry(bucket_meta_handler, bucket_name, bl, exclusive, objv_tracker, mtime, pattrs); | |
285 | } | |
286 | ||
287 | int rgw_bucket_instance_store_info(RGWRados *store, string& entry, bufferlist& bl, bool exclusive, | |
288 | map<string, bufferlist> *pattrs, RGWObjVersionTracker *objv_tracker, | |
289 | real_time mtime) { | |
290 | return store->meta_mgr->put_entry(bucket_instance_meta_handler, entry, bl, exclusive, objv_tracker, mtime, pattrs); | |
291 | } | |
292 | ||
293 | int rgw_bucket_instance_remove_entry(RGWRados *store, string& entry, RGWObjVersionTracker *objv_tracker) { | |
294 | return store->meta_mgr->remove_entry(bucket_instance_meta_handler, entry, objv_tracker); | |
295 | } | |
296 | ||
297 | // 'tenant/' is used in bucket instance keys for sync to avoid parsing ambiguity | |
298 | // with the existing instance[:shard] format. once we parse the shard, the / is | |
299 | // replaced with a : to match the [tenant:]instance format | |
300 | void rgw_bucket_instance_key_to_oid(string& key) | |
301 | { | |
302 | // replace tenant/ with tenant: | |
303 | auto c = key.find('/'); | |
304 | if (c != string::npos) { | |
305 | key[c] = ':'; | |
306 | } | |
307 | } | |
308 | ||
309 | // convert bucket instance oids back to the tenant/ format for metadata keys. | |
310 | // it's safe to parse 'tenant:' only for oids, because they won't contain the | |
311 | // optional :shard at the end | |
312 | void rgw_bucket_instance_oid_to_key(string& oid) | |
313 | { | |
314 | // find first : (could be tenant:bucket or bucket:instance) | |
315 | auto c = oid.find(':'); | |
316 | if (c != string::npos) { | |
317 | // if we find another :, the first one was for tenant | |
318 | if (oid.find(':', c + 1) != string::npos) { | |
319 | oid[c] = '/'; | |
320 | } | |
321 | } | |
322 | } | |
323 | ||
324 | int rgw_bucket_parse_bucket_instance(const string& bucket_instance, string *target_bucket_instance, int *shard_id) | |
325 | { | |
326 | ssize_t pos = bucket_instance.rfind(':'); | |
327 | if (pos < 0) { | |
328 | return -EINVAL; | |
329 | } | |
330 | ||
331 | string first = bucket_instance.substr(0, pos); | |
332 | string second = bucket_instance.substr(pos + 1); | |
333 | ||
334 | if (first.find(':') == string::npos) { | |
335 | *shard_id = -1; | |
336 | *target_bucket_instance = bucket_instance; | |
337 | return 0; | |
338 | } | |
339 | ||
340 | *target_bucket_instance = first; | |
341 | string err; | |
342 | *shard_id = strict_strtol(second.c_str(), 10, &err); | |
343 | if (!err.empty()) { | |
344 | return -EINVAL; | |
345 | } | |
346 | ||
347 | return 0; | |
348 | } | |
349 | ||
350 | // parse key in format: [tenant/]name:instance[:shard_id] | |
351 | int rgw_bucket_parse_bucket_key(CephContext *cct, const string& key, | |
352 | rgw_bucket *bucket, int *shard_id) | |
353 | { | |
354 | boost::string_ref name{key}; | |
355 | boost::string_ref instance; | |
356 | ||
357 | // split tenant/name | |
358 | auto pos = name.find('/'); | |
359 | if (pos != boost::string_ref::npos) { | |
360 | auto tenant = name.substr(0, pos); | |
361 | bucket->tenant.assign(tenant.begin(), tenant.end()); | |
362 | name = name.substr(pos + 1); | |
363 | } | |
364 | ||
365 | // split name:instance | |
366 | pos = name.find(':'); | |
367 | if (pos != boost::string_ref::npos) { | |
368 | instance = name.substr(pos + 1); | |
369 | name = name.substr(0, pos); | |
370 | } | |
371 | bucket->name.assign(name.begin(), name.end()); | |
372 | ||
373 | // split instance:shard | |
374 | pos = instance.find(':'); | |
375 | if (pos == boost::string_ref::npos) { | |
376 | bucket->bucket_id.assign(instance.begin(), instance.end()); | |
377 | *shard_id = -1; | |
378 | return 0; | |
379 | } | |
380 | ||
381 | // parse shard id | |
382 | auto shard = instance.substr(pos + 1); | |
383 | string err; | |
384 | auto id = strict_strtol(shard.data(), 10, &err); | |
385 | if (!err.empty()) { | |
386 | ldout(cct, 0) << "ERROR: failed to parse bucket shard '" | |
387 | << instance.data() << "': " << err << dendl; | |
388 | return -EINVAL; | |
389 | } | |
390 | ||
391 | *shard_id = id; | |
392 | instance = instance.substr(0, pos); | |
393 | bucket->bucket_id.assign(instance.begin(), instance.end()); | |
394 | return 0; | |
395 | } | |
396 | ||
397 | int rgw_bucket_set_attrs(RGWRados *store, RGWBucketInfo& bucket_info, | |
398 | map<string, bufferlist>& attrs, | |
399 | RGWObjVersionTracker *objv_tracker) | |
400 | { | |
401 | rgw_bucket& bucket = bucket_info.bucket; | |
402 | ||
403 | if (!bucket_info.has_instance_obj) { | |
404 | /* an old bucket object, need to convert it */ | |
405 | RGWObjectCtx obj_ctx(store); | |
406 | int ret = store->convert_old_bucket_info(obj_ctx, bucket.tenant, bucket.name); | |
407 | if (ret < 0) { | |
408 | ldout(store->ctx(), 0) << "ERROR: failed converting old bucket info: " << ret << dendl; | |
409 | return ret; | |
410 | } | |
411 | } | |
412 | ||
413 | /* we want the bucket instance name without the oid prefix cruft */ | |
414 | string key = bucket.get_key(); | |
415 | bufferlist bl; | |
416 | ||
417 | ::encode(bucket_info, bl); | |
418 | ||
419 | return rgw_bucket_instance_store_info(store, key, bl, false, &attrs, objv_tracker, real_time()); | |
420 | } | |
421 | ||
422 | static void dump_mulipart_index_results(list<rgw_obj_index_key>& objs_to_unlink, | |
423 | Formatter *f) | |
424 | { | |
425 | for (const auto& o : objs_to_unlink) { | |
426 | f->dump_string("object", o.name); | |
427 | } | |
428 | } | |
429 | ||
430 | void check_bad_user_bucket_mapping(RGWRados *store, const rgw_user& user_id, | |
431 | bool fix) | |
432 | { | |
433 | RGWUserBuckets user_buckets; | |
434 | bool is_truncated = false; | |
435 | string marker; | |
436 | ||
437 | CephContext *cct = store->ctx(); | |
438 | ||
439 | size_t max_entries = cct->_conf->rgw_list_buckets_max_chunk; | |
440 | ||
441 | do { | |
442 | int ret = rgw_read_user_buckets(store, user_id, user_buckets, marker, | |
443 | string(), max_entries, false, | |
444 | &is_truncated); | |
445 | if (ret < 0) { | |
446 | ldout(store->ctx(), 0) << "failed to read user buckets: " | |
447 | << cpp_strerror(-ret) << dendl; | |
448 | return; | |
449 | } | |
450 | ||
451 | map<string, RGWBucketEnt>& buckets = user_buckets.get_buckets(); | |
452 | for (map<string, RGWBucketEnt>::iterator i = buckets.begin(); | |
453 | i != buckets.end(); | |
454 | ++i) { | |
455 | marker = i->first; | |
456 | ||
457 | RGWBucketEnt& bucket_ent = i->second; | |
458 | rgw_bucket& bucket = bucket_ent.bucket; | |
459 | ||
460 | RGWBucketInfo bucket_info; | |
461 | real_time mtime; | |
462 | RGWObjectCtx obj_ctx(store); | |
463 | int r = store->get_bucket_info(obj_ctx, user_id.tenant, bucket.name, bucket_info, &mtime); | |
464 | if (r < 0) { | |
465 | ldout(store->ctx(), 0) << "could not get bucket info for bucket=" << bucket << dendl; | |
466 | continue; | |
467 | } | |
468 | ||
469 | rgw_bucket& actual_bucket = bucket_info.bucket; | |
470 | ||
471 | if (actual_bucket.name.compare(bucket.name) != 0 || | |
472 | actual_bucket.tenant.compare(bucket.tenant) != 0 || | |
473 | actual_bucket.marker.compare(bucket.marker) != 0 || | |
474 | actual_bucket.bucket_id.compare(bucket.bucket_id) != 0) { | |
475 | cout << "bucket info mismatch: expected " << actual_bucket << " got " << bucket << std::endl; | |
476 | if (fix) { | |
477 | cout << "fixing" << std::endl; | |
478 | r = rgw_link_bucket(store, user_id, actual_bucket, bucket_info.creation_time); | |
479 | if (r < 0) { | |
480 | cerr << "failed to fix bucket: " << cpp_strerror(-r) << std::endl; | |
481 | } | |
482 | } | |
483 | } | |
484 | } | |
485 | } while (is_truncated); | |
486 | } | |
487 | ||
488 | static bool bucket_object_check_filter(const string& oid) | |
489 | { | |
490 | rgw_obj_key key; | |
491 | string ns; | |
492 | return rgw_obj_key::oid_to_key_in_ns(oid, &key, ns); | |
493 | } | |
494 | ||
495 | int rgw_remove_object(RGWRados *store, RGWBucketInfo& bucket_info, rgw_bucket& bucket, rgw_obj_key& key) | |
496 | { | |
497 | RGWObjectCtx rctx(store); | |
498 | ||
499 | if (key.instance.empty()) { | |
500 | key.instance = "null"; | |
501 | } | |
502 | ||
503 | rgw_obj obj(bucket, key); | |
504 | ||
505 | return store->delete_obj(rctx, bucket_info, obj, bucket_info.versioning_status()); | |
506 | } | |
507 | ||
508 | int rgw_remove_bucket(RGWRados *store, rgw_bucket& bucket, bool delete_children) | |
509 | { | |
510 | int ret; | |
511 | map<RGWObjCategory, RGWStorageStats> stats; | |
512 | std::vector<rgw_bucket_dir_entry> objs; | |
513 | map<string, bool> common_prefixes; | |
514 | RGWBucketInfo info; | |
515 | RGWObjectCtx obj_ctx(store); | |
516 | ||
517 | string bucket_ver, master_ver; | |
518 | ||
519 | ret = store->get_bucket_info(obj_ctx, bucket.tenant, bucket.name, info, NULL); | |
520 | if (ret < 0) | |
521 | return ret; | |
522 | ||
523 | ret = store->get_bucket_stats(info, RGW_NO_SHARD, &bucket_ver, &master_ver, stats, NULL); | |
524 | if (ret < 0) | |
525 | return ret; | |
526 | ||
527 | RGWRados::Bucket target(store, info); | |
528 | RGWRados::Bucket::List list_op(&target); | |
529 | CephContext *cct = store->ctx(); | |
530 | int max = 1000; | |
531 | ||
532 | list_op.params.list_versions = true; | |
533 | ||
534 | do { | |
535 | objs.clear(); | |
536 | ||
537 | ret = list_op.list_objects(max, &objs, &common_prefixes, NULL); | |
538 | if (ret < 0) | |
539 | return ret; | |
540 | ||
541 | if (!objs.empty() && !delete_children) { | |
542 | lderr(store->ctx()) << "ERROR: could not remove non-empty bucket " << bucket.name << dendl; | |
543 | return -ENOTEMPTY; | |
544 | } | |
545 | ||
546 | for (const auto& obj : objs) { | |
547 | rgw_obj_key key(obj.key); | |
548 | ret = rgw_remove_object(store, info, bucket, key); | |
549 | if (ret < 0) | |
550 | return ret; | |
551 | } | |
552 | ||
553 | } while (!objs.empty()); | |
554 | ||
555 | string prefix, delimiter; | |
556 | ||
557 | ret = abort_bucket_multiparts(store, cct, info, prefix, delimiter); | |
558 | if (ret < 0) { | |
559 | return ret; | |
560 | } | |
561 | ||
562 | ret = rgw_bucket_sync_user_stats(store, bucket.tenant, info); | |
563 | if ( ret < 0) { | |
564 | dout(1) << "WARNING: failed sync user stats before bucket delete. ret=" << ret << dendl; | |
565 | } | |
566 | ||
567 | RGWObjVersionTracker objv_tracker; | |
568 | ||
569 | ret = store->delete_bucket(info, objv_tracker); | |
570 | if (ret < 0) { | |
571 | lderr(store->ctx()) << "ERROR: could not remove bucket " << bucket.name << dendl; | |
572 | return ret; | |
573 | } | |
574 | ||
575 | ret = rgw_unlink_bucket(store, info.owner, bucket.tenant, bucket.name, false); | |
576 | if (ret < 0) { | |
577 | lderr(store->ctx()) << "ERROR: unable to remove user bucket information" << dendl; | |
578 | } | |
579 | ||
580 | return ret; | |
581 | } | |
582 | ||
583 | static int aio_wait(librados::AioCompletion *handle) | |
584 | { | |
585 | librados::AioCompletion *c = (librados::AioCompletion *)handle; | |
586 | c->wait_for_safe(); | |
587 | int ret = c->get_return_value(); | |
588 | c->release(); | |
589 | return ret; | |
590 | } | |
591 | ||
592 | static int drain_handles(list<librados::AioCompletion *>& pending) | |
593 | { | |
594 | int ret = 0; | |
595 | while (!pending.empty()) { | |
596 | librados::AioCompletion *handle = pending.front(); | |
597 | pending.pop_front(); | |
598 | int r = aio_wait(handle); | |
599 | if (r < 0) { | |
600 | ret = r; | |
601 | } | |
602 | } | |
603 | return ret; | |
604 | } | |
605 | ||
606 | int rgw_remove_bucket_bypass_gc(RGWRados *store, rgw_bucket& bucket, | |
607 | int concurrent_max, bool keep_index_consistent) | |
608 | { | |
609 | int ret; | |
610 | map<RGWObjCategory, RGWStorageStats> stats; | |
611 | std::vector<rgw_bucket_dir_entry> objs; | |
612 | map<string, bool> common_prefixes; | |
613 | RGWBucketInfo info; | |
614 | RGWObjectCtx obj_ctx(store); | |
615 | CephContext *cct = store->ctx(); | |
616 | ||
617 | string bucket_ver, master_ver; | |
618 | ||
619 | ret = store->get_bucket_info(obj_ctx, bucket.tenant, bucket.name, info, NULL); | |
620 | if (ret < 0) | |
621 | return ret; | |
622 | ||
623 | ret = store->get_bucket_stats(info, RGW_NO_SHARD, &bucket_ver, &master_ver, stats, NULL); | |
624 | if (ret < 0) | |
625 | return ret; | |
626 | ||
627 | string prefix, delimiter; | |
628 | ||
629 | ret = abort_bucket_multiparts(store, cct, info, prefix, delimiter); | |
630 | if (ret < 0) { | |
631 | return ret; | |
632 | } | |
633 | ||
634 | RGWRados::Bucket target(store, info); | |
635 | RGWRados::Bucket::List list_op(&target); | |
636 | ||
637 | list_op.params.list_versions = true; | |
638 | ||
639 | std::list<librados::AioCompletion*> handles; | |
640 | ||
641 | int max = 1000; | |
642 | int max_aio = concurrent_max; | |
643 | ret = list_op.list_objects(max, &objs, &common_prefixes, NULL); | |
644 | if (ret < 0) | |
645 | return ret; | |
646 | ||
647 | while (!objs.empty()) { | |
648 | std::vector<rgw_bucket_dir_entry>::iterator it = objs.begin(); | |
649 | for (; it != objs.end(); ++it) { | |
650 | RGWObjState *astate = NULL; | |
651 | rgw_obj obj(bucket, (*it).key); | |
652 | ||
653 | ret = store->get_obj_state(&obj_ctx, info, obj, &astate, false); | |
654 | if (ret == -ENOENT) { | |
655 | dout(1) << "WARNING: cannot find obj state for obj " << obj.get_oid() << dendl; | |
656 | continue; | |
657 | } | |
658 | if (ret < 0) { | |
659 | lderr(store->ctx()) << "ERROR: get obj state returned with error " << ret << dendl; | |
660 | return ret; | |
661 | } | |
662 | ||
663 | if (astate->has_manifest) { | |
664 | RGWObjManifest& manifest = astate->manifest; | |
665 | RGWObjManifest::obj_iterator miter = manifest.obj_begin(); | |
666 | rgw_obj head_obj = manifest.get_obj(); | |
667 | rgw_raw_obj raw_head_obj; | |
668 | store->obj_to_raw(info.placement_rule, head_obj, &raw_head_obj); | |
669 | ||
670 | ||
671 | for (; miter != manifest.obj_end() && max_aio--; ++miter) { | |
672 | if (!max_aio) { | |
673 | ret = drain_handles(handles); | |
674 | if (ret < 0) { | |
675 | lderr(store->ctx()) << "ERROR: could not drain handles as aio completion returned with " << ret << dendl; | |
676 | return ret; | |
677 | } | |
678 | max_aio = concurrent_max; | |
679 | } | |
680 | ||
681 | rgw_raw_obj last_obj = miter.get_location().get_raw_obj(store); | |
682 | if (last_obj == raw_head_obj) { | |
683 | // have the head obj deleted at the end | |
684 | continue; | |
685 | } | |
686 | ||
687 | ret = store->delete_raw_obj_aio(last_obj, handles); | |
688 | if (ret < 0) { | |
689 | lderr(store->ctx()) << "ERROR: delete obj aio failed with " << ret << dendl; | |
690 | return ret; | |
691 | } | |
692 | } // for all shadow objs | |
693 | ||
694 | ret = store->delete_obj_aio(head_obj, info, astate, handles, keep_index_consistent); | |
695 | if (ret < 0) { | |
696 | lderr(store->ctx()) << "ERROR: delete obj aio failed with " << ret << dendl; | |
697 | return ret; | |
698 | } | |
699 | } | |
700 | ||
701 | if (!max_aio) { | |
702 | ret = drain_handles(handles); | |
703 | if (ret < 0) { | |
704 | lderr(store->ctx()) << "ERROR: could not drain handles as aio completion returned with " << ret << dendl; | |
705 | return ret; | |
706 | } | |
707 | max_aio = concurrent_max; | |
708 | } | |
709 | } // for all RGW objects | |
710 | objs.clear(); | |
711 | ||
712 | ret = list_op.list_objects(max, &objs, &common_prefixes, NULL); | |
713 | if (ret < 0) | |
714 | return ret; | |
715 | } | |
716 | ||
717 | ret = drain_handles(handles); | |
718 | if (ret < 0) { | |
719 | lderr(store->ctx()) << "ERROR: could not drain handles as aio completion returned with " << ret << dendl; | |
720 | return ret; | |
721 | } | |
722 | ||
723 | ret = rgw_bucket_sync_user_stats(store, bucket.tenant, info); | |
724 | if (ret < 0) { | |
725 | dout(1) << "WARNING: failed sync user stats before bucket delete. ret=" << ret << dendl; | |
726 | } | |
727 | ||
728 | RGWObjVersionTracker objv_tracker; | |
729 | ||
730 | ret = rgw_bucket_delete_bucket_obj(store, bucket.tenant, bucket.name, objv_tracker); | |
731 | if (ret < 0) { | |
732 | lderr(store->ctx()) << "ERROR: could not remove bucket " << bucket.name << "with ret as " << ret << dendl; | |
733 | return ret; | |
734 | } | |
735 | ||
736 | if (!store->is_syncing_bucket_meta(bucket)) { | |
737 | RGWObjVersionTracker objv_tracker; | |
738 | string entry = bucket.get_key(); | |
739 | ret = rgw_bucket_instance_remove_entry(store, entry, &objv_tracker); | |
740 | if (ret < 0) { | |
741 | lderr(store->ctx()) << "ERROR: could not remove bucket instance entry" << bucket.name << "with ret as " << ret << dendl; | |
742 | return ret; | |
743 | } | |
744 | } | |
745 | ||
746 | ret = rgw_unlink_bucket(store, info.owner, bucket.tenant, bucket.name, false); | |
747 | if (ret < 0) { | |
748 | lderr(store->ctx()) << "ERROR: unable to remove user bucket information" << dendl; | |
749 | } | |
750 | ||
751 | return ret; | |
752 | } | |
753 | ||
754 | int rgw_bucket_delete_bucket_obj(RGWRados *store, | |
755 | const string& tenant_name, | |
756 | const string& bucket_name, | |
757 | RGWObjVersionTracker& objv_tracker) | |
758 | { | |
759 | string key; | |
760 | ||
761 | rgw_make_bucket_entry_name(tenant_name, bucket_name, key); | |
762 | return store->meta_mgr->remove_entry(bucket_meta_handler, key, &objv_tracker); | |
763 | } | |
764 | ||
765 | static void set_err_msg(std::string *sink, std::string msg) | |
766 | { | |
767 | if (sink && !msg.empty()) | |
768 | *sink = msg; | |
769 | } | |
770 | ||
771 | int RGWBucket::init(RGWRados *storage, RGWBucketAdminOpState& op_state) | |
772 | { | |
773 | if (!storage) | |
774 | return -EINVAL; | |
775 | ||
776 | store = storage; | |
777 | ||
778 | rgw_user user_id = op_state.get_user_id(); | |
779 | tenant = user_id.tenant; | |
780 | bucket_name = op_state.get_bucket_name(); | |
781 | RGWUserBuckets user_buckets; | |
782 | RGWObjectCtx obj_ctx(store); | |
783 | ||
784 | if (bucket_name.empty() && user_id.empty()) | |
785 | return -EINVAL; | |
786 | ||
787 | if (!bucket_name.empty()) { | |
788 | int r = store->get_bucket_info(obj_ctx, tenant, bucket_name, bucket_info, NULL); | |
789 | if (r < 0) { | |
790 | ldout(store->ctx(), 0) << "could not get bucket info for bucket=" << bucket_name << dendl; | |
791 | return r; | |
792 | } | |
793 | ||
794 | op_state.set_bucket(bucket_info.bucket); | |
795 | } | |
796 | ||
797 | if (!user_id.empty()) { | |
798 | int r = rgw_get_user_info_by_uid(store, user_id, user_info); | |
799 | if (r < 0) | |
800 | return r; | |
801 | ||
802 | op_state.display_name = user_info.display_name; | |
803 | } | |
804 | ||
805 | clear_failure(); | |
806 | return 0; | |
807 | } | |
808 | ||
809 | int RGWBucket::link(RGWBucketAdminOpState& op_state, std::string *err_msg) | |
810 | { | |
811 | if (!op_state.is_user_op()) { | |
812 | set_err_msg(err_msg, "empty user id"); | |
813 | return -EINVAL; | |
814 | } | |
815 | ||
816 | string bucket_id = op_state.get_bucket_id(); | |
817 | if (bucket_id.empty()) { | |
818 | set_err_msg(err_msg, "empty bucket instance id"); | |
819 | return -EINVAL; | |
820 | } | |
821 | ||
822 | std::string display_name = op_state.get_user_display_name(); | |
823 | rgw_bucket bucket = op_state.get_bucket(); | |
824 | ||
825 | const rgw_pool& root_pool = store->get_zone_params().domain_root; | |
826 | rgw_raw_obj obj(root_pool, bucket.name); | |
827 | RGWObjVersionTracker objv_tracker; | |
828 | ||
829 | map<string, bufferlist> attrs; | |
830 | RGWBucketInfo bucket_info; | |
831 | ||
832 | string key = bucket.name + ":" + bucket_id; | |
833 | RGWObjectCtx obj_ctx(store); | |
834 | int r = store->get_bucket_instance_info(obj_ctx, key, bucket_info, NULL, &attrs); | |
835 | if (r < 0) { | |
836 | return r; | |
837 | } | |
838 | ||
839 | rgw_user user_id = op_state.get_user_id(); | |
840 | ||
841 | map<string, bufferlist>::iterator aiter = attrs.find(RGW_ATTR_ACL); | |
842 | if (aiter != attrs.end()) { | |
843 | bufferlist aclbl = aiter->second; | |
844 | RGWAccessControlPolicy policy; | |
845 | ACLOwner owner; | |
846 | try { | |
847 | bufferlist::iterator iter = aclbl.begin(); | |
848 | ::decode(policy, iter); | |
849 | owner = policy.get_owner(); | |
850 | } catch (buffer::error& err) { | |
851 | set_err_msg(err_msg, "couldn't decode policy"); | |
852 | return -EIO; | |
853 | } | |
854 | ||
855 | r = rgw_unlink_bucket(store, owner.get_id(), bucket.tenant, bucket.name, false); | |
856 | if (r < 0) { | |
857 | set_err_msg(err_msg, "could not unlink policy from user " + owner.get_id().to_str()); | |
858 | return r; | |
859 | } | |
860 | ||
861 | // now update the user for the bucket... | |
862 | if (display_name.empty()) { | |
863 | ldout(store->ctx(), 0) << "WARNING: user " << user_info.user_id << " has no display name set" << dendl; | |
864 | } | |
865 | policy.create_default(user_info.user_id, display_name); | |
866 | ||
867 | owner = policy.get_owner(); | |
868 | r = store->set_bucket_owner(bucket_info.bucket, owner); | |
869 | if (r < 0) { | |
870 | set_err_msg(err_msg, "failed to set bucket owner: " + cpp_strerror(-r)); | |
871 | return r; | |
872 | } | |
873 | ||
874 | // ...and encode the acl | |
875 | aclbl.clear(); | |
876 | policy.encode(aclbl); | |
877 | ||
878 | r = store->system_obj_set_attr(NULL, obj, RGW_ATTR_ACL, aclbl, &objv_tracker); | |
879 | if (r < 0) { | |
880 | return r; | |
881 | } | |
882 | ||
883 | RGWAccessControlPolicy policy_instance; | |
884 | policy_instance.create_default(user_info.user_id, display_name); | |
885 | aclbl.clear(); | |
886 | policy_instance.encode(aclbl); | |
887 | ||
888 | string oid_bucket_instance = RGW_BUCKET_INSTANCE_MD_PREFIX + key; | |
889 | rgw_raw_obj obj_bucket_instance(root_pool, oid_bucket_instance); | |
890 | r = store->system_obj_set_attr(NULL, obj_bucket_instance, RGW_ATTR_ACL, aclbl, &objv_tracker); | |
891 | if (r < 0) { | |
892 | return r; | |
893 | } | |
894 | ||
895 | r = rgw_link_bucket(store, user_info.user_id, bucket_info.bucket, real_time()); | |
896 | if (r < 0) { | |
897 | return r; | |
898 | } | |
899 | } | |
900 | ||
901 | return 0; | |
902 | } | |
903 | ||
904 | int RGWBucket::unlink(RGWBucketAdminOpState& op_state, std::string *err_msg) | |
905 | { | |
906 | rgw_bucket bucket = op_state.get_bucket(); | |
907 | ||
908 | if (!op_state.is_user_op()) { | |
909 | set_err_msg(err_msg, "could not fetch user or user bucket info"); | |
910 | return -EINVAL; | |
911 | } | |
912 | ||
913 | int r = rgw_unlink_bucket(store, user_info.user_id, bucket.tenant, bucket.name); | |
914 | if (r < 0) { | |
915 | set_err_msg(err_msg, "error unlinking bucket" + cpp_strerror(-r)); | |
916 | } | |
917 | ||
918 | return r; | |
919 | } | |
920 | ||
921 | int RGWBucket::remove(RGWBucketAdminOpState& op_state, bool bypass_gc, | |
922 | bool keep_index_consistent, std::string *err_msg) | |
923 | { | |
924 | bool delete_children = op_state.will_delete_children(); | |
925 | rgw_bucket bucket = op_state.get_bucket(); | |
926 | int ret; | |
927 | ||
928 | if (bypass_gc) { | |
929 | if (delete_children) { | |
930 | ret = rgw_remove_bucket_bypass_gc(store, bucket, op_state.get_max_aio(), keep_index_consistent); | |
931 | } else { | |
932 | set_err_msg(err_msg, "purge objects should be set for gc to be bypassed"); | |
933 | return -EINVAL; | |
934 | } | |
935 | } else { | |
936 | ret = rgw_remove_bucket(store, bucket, delete_children); | |
937 | } | |
938 | ||
939 | if (ret < 0) { | |
940 | set_err_msg(err_msg, "unable to remove bucket" + cpp_strerror(-ret)); | |
941 | return ret; | |
942 | } | |
943 | ||
944 | return 0; | |
945 | } | |
946 | ||
947 | int RGWBucket::remove_object(RGWBucketAdminOpState& op_state, std::string *err_msg) | |
948 | { | |
949 | rgw_bucket bucket = op_state.get_bucket(); | |
950 | std::string object_name = op_state.get_object_name(); | |
951 | ||
952 | rgw_obj_key key(object_name); | |
953 | ||
954 | int ret = rgw_remove_object(store, bucket_info, bucket, key); | |
955 | if (ret < 0) { | |
956 | set_err_msg(err_msg, "unable to remove object" + cpp_strerror(-ret)); | |
957 | return ret; | |
958 | } | |
959 | ||
960 | return 0; | |
961 | } | |
962 | ||
963 | static void dump_bucket_index(map<string, rgw_bucket_dir_entry> result, Formatter *f) | |
964 | { | |
965 | map<string, rgw_bucket_dir_entry>::iterator iter; | |
966 | for (iter = result.begin(); iter != result.end(); ++iter) { | |
967 | f->dump_string("object", iter->first); | |
968 | } | |
969 | } | |
970 | ||
971 | static void dump_bucket_usage(map<RGWObjCategory, RGWStorageStats>& stats, Formatter *formatter) | |
972 | { | |
973 | map<RGWObjCategory, RGWStorageStats>::iterator iter; | |
974 | ||
975 | formatter->open_object_section("usage"); | |
976 | for (iter = stats.begin(); iter != stats.end(); ++iter) { | |
977 | RGWStorageStats& s = iter->second; | |
978 | const char *cat_name = rgw_obj_category_name(iter->first); | |
979 | formatter->open_object_section(cat_name); | |
980 | s.dump(formatter); | |
981 | formatter->close_section(); | |
982 | } | |
983 | formatter->close_section(); | |
984 | } | |
985 | ||
986 | static void dump_index_check(map<RGWObjCategory, RGWStorageStats> existing_stats, | |
987 | map<RGWObjCategory, RGWStorageStats> calculated_stats, | |
988 | Formatter *formatter) | |
989 | { | |
990 | formatter->open_object_section("check_result"); | |
991 | formatter->open_object_section("existing_header"); | |
992 | dump_bucket_usage(existing_stats, formatter); | |
993 | formatter->close_section(); | |
994 | formatter->open_object_section("calculated_header"); | |
995 | dump_bucket_usage(calculated_stats, formatter); | |
996 | formatter->close_section(); | |
997 | formatter->close_section(); | |
998 | } | |
999 | ||
1000 | int RGWBucket::check_bad_index_multipart(RGWBucketAdminOpState& op_state, | |
1001 | RGWFormatterFlusher& flusher ,std::string *err_msg) | |
1002 | { | |
1003 | bool fix_index = op_state.will_fix_index(); | |
1004 | rgw_bucket bucket = op_state.get_bucket(); | |
1005 | ||
1006 | size_t max = 1000; | |
1007 | ||
1008 | map<string, bool> common_prefixes; | |
1009 | ||
1010 | bool is_truncated; | |
1011 | map<string, bool> meta_objs; | |
1012 | map<rgw_obj_index_key, string> all_objs; | |
1013 | ||
1014 | RGWBucketInfo bucket_info; | |
1015 | RGWObjectCtx obj_ctx(store); | |
1016 | int r = store->get_bucket_instance_info(obj_ctx, bucket, bucket_info, nullptr, nullptr); | |
1017 | if (r < 0) { | |
1018 | ldout(store->ctx(), 0) << "ERROR: " << __func__ << "(): get_bucket_instance_info(bucket=" << bucket << ") returned r=" << r << dendl; | |
1019 | return r; | |
1020 | } | |
1021 | ||
1022 | RGWRados::Bucket target(store, bucket_info); | |
1023 | RGWRados::Bucket::List list_op(&target); | |
1024 | ||
1025 | list_op.params.list_versions = true; | |
1026 | list_op.params.ns = RGW_OBJ_NS_MULTIPART; | |
1027 | ||
1028 | do { | |
1029 | vector<rgw_bucket_dir_entry> result; | |
1030 | int r = list_op.list_objects(max, &result, &common_prefixes, &is_truncated); | |
1031 | if (r < 0) { | |
1032 | set_err_msg(err_msg, "failed to list objects in bucket=" + bucket.name + | |
1033 | " err=" + cpp_strerror(-r)); | |
1034 | ||
1035 | return r; | |
1036 | } | |
1037 | ||
1038 | vector<rgw_bucket_dir_entry>::iterator iter; | |
1039 | for (iter = result.begin(); iter != result.end(); ++iter) { | |
1040 | rgw_obj_index_key key = iter->key; | |
1041 | rgw_obj obj(bucket, key); | |
1042 | string oid = obj.get_oid(); | |
1043 | ||
1044 | int pos = oid.find_last_of('.'); | |
1045 | if (pos < 0) { | |
1046 | /* obj has no suffix */ | |
1047 | all_objs[key] = oid; | |
1048 | } else { | |
1049 | /* obj has suffix */ | |
1050 | string name = oid.substr(0, pos); | |
1051 | string suffix = oid.substr(pos + 1); | |
1052 | ||
1053 | if (suffix.compare("meta") == 0) { | |
1054 | meta_objs[name] = true; | |
1055 | } else { | |
1056 | all_objs[key] = name; | |
1057 | } | |
1058 | } | |
1059 | } | |
1060 | ||
1061 | } while (is_truncated); | |
1062 | ||
1063 | list<rgw_obj_index_key> objs_to_unlink; | |
1064 | Formatter *f = flusher.get_formatter(); | |
1065 | ||
1066 | f->open_array_section("invalid_multipart_entries"); | |
1067 | ||
1068 | for (auto aiter = all_objs.begin(); aiter != all_objs.end(); ++aiter) { | |
1069 | string& name = aiter->second; | |
1070 | ||
1071 | if (meta_objs.find(name) == meta_objs.end()) { | |
1072 | objs_to_unlink.push_back(aiter->first); | |
1073 | } | |
1074 | ||
1075 | if (objs_to_unlink.size() > max) { | |
1076 | if (fix_index) { | |
1077 | int r = store->remove_objs_from_index(bucket_info, objs_to_unlink); | |
1078 | if (r < 0) { | |
1079 | set_err_msg(err_msg, "ERROR: remove_obj_from_index() returned error: " + | |
1080 | cpp_strerror(-r)); | |
1081 | return r; | |
1082 | } | |
1083 | } | |
1084 | ||
1085 | dump_mulipart_index_results(objs_to_unlink, flusher.get_formatter()); | |
1086 | flusher.flush(); | |
1087 | objs_to_unlink.clear(); | |
1088 | } | |
1089 | } | |
1090 | ||
1091 | if (fix_index) { | |
1092 | int r = store->remove_objs_from_index(bucket_info, objs_to_unlink); | |
1093 | if (r < 0) { | |
1094 | set_err_msg(err_msg, "ERROR: remove_obj_from_index() returned error: " + | |
1095 | cpp_strerror(-r)); | |
1096 | ||
1097 | return r; | |
1098 | } | |
1099 | } | |
1100 | ||
1101 | dump_mulipart_index_results(objs_to_unlink, f); | |
1102 | f->close_section(); | |
1103 | flusher.flush(); | |
1104 | ||
1105 | return 0; | |
1106 | } | |
1107 | ||
1108 | int RGWBucket::check_object_index(RGWBucketAdminOpState& op_state, | |
1109 | RGWFormatterFlusher& flusher, | |
1110 | std::string *err_msg) | |
1111 | { | |
1112 | ||
1113 | bool fix_index = op_state.will_fix_index(); | |
1114 | ||
1115 | rgw_bucket bucket = op_state.get_bucket(); | |
1116 | ||
1117 | if (!fix_index) { | |
1118 | set_err_msg(err_msg, "check-objects flag requires fix index enabled"); | |
1119 | return -EINVAL; | |
1120 | } | |
1121 | ||
1122 | store->cls_obj_set_bucket_tag_timeout(bucket_info, BUCKET_TAG_TIMEOUT); | |
1123 | ||
1124 | string prefix; | |
1125 | rgw_obj_index_key marker; | |
1126 | bool is_truncated = true; | |
1127 | ||
1128 | Formatter *formatter = flusher.get_formatter(); | |
1129 | formatter->open_object_section("objects"); | |
1130 | while (is_truncated) { | |
1131 | map<string, rgw_bucket_dir_entry> result; | |
1132 | ||
1133 | int r = store->cls_bucket_list(bucket_info, RGW_NO_SHARD, marker, prefix, 1000, true, | |
1134 | result, &is_truncated, &marker, | |
1135 | bucket_object_check_filter); | |
1136 | if (r == -ENOENT) { | |
1137 | break; | |
1138 | } else if (r < 0 && r != -ENOENT) { | |
1139 | set_err_msg(err_msg, "ERROR: failed operation r=" + cpp_strerror(-r)); | |
1140 | } | |
1141 | ||
1142 | ||
1143 | dump_bucket_index(result, formatter); | |
1144 | flusher.flush(); | |
1145 | ||
1146 | } | |
1147 | ||
1148 | formatter->close_section(); | |
1149 | ||
1150 | store->cls_obj_set_bucket_tag_timeout(bucket_info, 0); | |
1151 | ||
1152 | return 0; | |
1153 | } | |
1154 | ||
1155 | ||
1156 | int RGWBucket::check_index(RGWBucketAdminOpState& op_state, | |
1157 | map<RGWObjCategory, RGWStorageStats>& existing_stats, | |
1158 | map<RGWObjCategory, RGWStorageStats>& calculated_stats, | |
1159 | std::string *err_msg) | |
1160 | { | |
1161 | rgw_bucket bucket = op_state.get_bucket(); | |
1162 | bool fix_index = op_state.will_fix_index(); | |
1163 | ||
1164 | int r = store->bucket_check_index(bucket_info, &existing_stats, &calculated_stats); | |
1165 | if (r < 0) { | |
1166 | set_err_msg(err_msg, "failed to check index error=" + cpp_strerror(-r)); | |
1167 | return r; | |
1168 | } | |
1169 | ||
1170 | if (fix_index) { | |
1171 | r = store->bucket_rebuild_index(bucket_info); | |
1172 | if (r < 0) { | |
1173 | set_err_msg(err_msg, "failed to rebuild index err=" + cpp_strerror(-r)); | |
1174 | return r; | |
1175 | } | |
1176 | } | |
1177 | ||
1178 | return 0; | |
1179 | } | |
1180 | ||
1181 | ||
1182 | int RGWBucket::policy_bl_to_stream(bufferlist& bl, ostream& o) | |
1183 | { | |
1184 | RGWAccessControlPolicy_S3 policy(g_ceph_context); | |
1185 | bufferlist::iterator iter = bl.begin(); | |
1186 | try { | |
1187 | policy.decode(iter); | |
1188 | } catch (buffer::error& err) { | |
1189 | dout(0) << "ERROR: caught buffer::error, could not decode policy" << dendl; | |
1190 | return -EIO; | |
1191 | } | |
1192 | policy.to_xml(o); | |
1193 | return 0; | |
1194 | } | |
1195 | ||
1196 | static int policy_decode(RGWRados *store, bufferlist& bl, RGWAccessControlPolicy& policy) | |
1197 | { | |
1198 | bufferlist::iterator iter = bl.begin(); | |
1199 | try { | |
1200 | policy.decode(iter); | |
1201 | } catch (buffer::error& err) { | |
1202 | ldout(store->ctx(), 0) << "ERROR: caught buffer::error, could not decode policy" << dendl; | |
1203 | return -EIO; | |
1204 | } | |
1205 | return 0; | |
1206 | } | |
1207 | ||
1208 | int RGWBucket::get_policy(RGWBucketAdminOpState& op_state, RGWAccessControlPolicy& policy) | |
1209 | { | |
1210 | std::string object_name = op_state.get_object_name(); | |
1211 | rgw_bucket bucket = op_state.get_bucket(); | |
1212 | RGWObjectCtx obj_ctx(store); | |
1213 | ||
1214 | RGWBucketInfo bucket_info; | |
1215 | map<string, bufferlist> attrs; | |
1216 | int ret = store->get_bucket_info(obj_ctx, bucket.tenant, bucket.name, bucket_info, NULL, &attrs); | |
1217 | if (ret < 0) { | |
1218 | return ret; | |
1219 | } | |
1220 | ||
1221 | if (!object_name.empty()) { | |
1222 | bufferlist bl; | |
1223 | rgw_obj obj(bucket, object_name); | |
1224 | ||
1225 | RGWRados::Object op_target(store, bucket_info, obj_ctx, obj); | |
1226 | RGWRados::Object::Read rop(&op_target); | |
1227 | ||
1228 | int ret = rop.get_attr(RGW_ATTR_ACL, bl); | |
1229 | if (ret < 0) | |
1230 | return ret; | |
1231 | ||
1232 | return policy_decode(store, bl, policy); | |
1233 | } | |
1234 | ||
1235 | map<string, bufferlist>::iterator aiter = attrs.find(RGW_ATTR_ACL); | |
1236 | if (aiter == attrs.end()) { | |
1237 | return -ENOENT; | |
1238 | } | |
1239 | ||
1240 | return policy_decode(store, aiter->second, policy); | |
1241 | } | |
1242 | ||
1243 | ||
1244 | int RGWBucketAdminOp::get_policy(RGWRados *store, RGWBucketAdminOpState& op_state, | |
1245 | RGWAccessControlPolicy& policy) | |
1246 | { | |
1247 | RGWBucket bucket; | |
1248 | ||
1249 | int ret = bucket.init(store, op_state); | |
1250 | if (ret < 0) | |
1251 | return ret; | |
1252 | ||
1253 | ret = bucket.get_policy(op_state, policy); | |
1254 | if (ret < 0) | |
1255 | return ret; | |
1256 | ||
1257 | return 0; | |
1258 | } | |
1259 | ||
1260 | /* Wrappers to facilitate RESTful interface */ | |
1261 | ||
1262 | ||
1263 | int RGWBucketAdminOp::get_policy(RGWRados *store, RGWBucketAdminOpState& op_state, | |
1264 | RGWFormatterFlusher& flusher) | |
1265 | { | |
1266 | RGWAccessControlPolicy policy(store->ctx()); | |
1267 | ||
1268 | int ret = get_policy(store, op_state, policy); | |
1269 | if (ret < 0) | |
1270 | return ret; | |
1271 | ||
1272 | Formatter *formatter = flusher.get_formatter(); | |
1273 | ||
1274 | flusher.start(0); | |
1275 | ||
1276 | formatter->open_object_section("policy"); | |
1277 | policy.dump(formatter); | |
1278 | formatter->close_section(); | |
1279 | ||
1280 | flusher.flush(); | |
1281 | ||
1282 | return 0; | |
1283 | } | |
1284 | ||
1285 | int RGWBucketAdminOp::dump_s3_policy(RGWRados *store, RGWBucketAdminOpState& op_state, | |
1286 | ostream& os) | |
1287 | { | |
1288 | RGWAccessControlPolicy_S3 policy(store->ctx()); | |
1289 | ||
1290 | int ret = get_policy(store, op_state, policy); | |
1291 | if (ret < 0) | |
1292 | return ret; | |
1293 | ||
1294 | policy.to_xml(os); | |
1295 | ||
1296 | return 0; | |
1297 | } | |
1298 | ||
1299 | int RGWBucketAdminOp::unlink(RGWRados *store, RGWBucketAdminOpState& op_state) | |
1300 | { | |
1301 | RGWBucket bucket; | |
1302 | ||
1303 | int ret = bucket.init(store, op_state); | |
1304 | if (ret < 0) | |
1305 | return ret; | |
1306 | ||
1307 | return bucket.unlink(op_state); | |
1308 | } | |
1309 | ||
1310 | int RGWBucketAdminOp::link(RGWRados *store, RGWBucketAdminOpState& op_state, string *err) | |
1311 | { | |
1312 | RGWBucket bucket; | |
1313 | ||
1314 | int ret = bucket.init(store, op_state); | |
1315 | if (ret < 0) | |
1316 | return ret; | |
1317 | ||
1318 | return bucket.link(op_state, err); | |
1319 | ||
1320 | } | |
1321 | ||
1322 | int RGWBucketAdminOp::check_index(RGWRados *store, RGWBucketAdminOpState& op_state, | |
1323 | RGWFormatterFlusher& flusher) | |
1324 | { | |
1325 | int ret; | |
1326 | map<RGWObjCategory, RGWStorageStats> existing_stats; | |
1327 | map<RGWObjCategory, RGWStorageStats> calculated_stats; | |
1328 | ||
1329 | ||
1330 | RGWBucket bucket; | |
1331 | ||
1332 | ret = bucket.init(store, op_state); | |
1333 | if (ret < 0) | |
1334 | return ret; | |
1335 | ||
1336 | Formatter *formatter = flusher.get_formatter(); | |
1337 | flusher.start(0); | |
1338 | ||
1339 | ret = bucket.check_bad_index_multipart(op_state, flusher); | |
1340 | if (ret < 0) | |
1341 | return ret; | |
1342 | ||
1343 | ret = bucket.check_object_index(op_state, flusher); | |
1344 | if (ret < 0) | |
1345 | return ret; | |
1346 | ||
1347 | ret = bucket.check_index(op_state, existing_stats, calculated_stats); | |
1348 | if (ret < 0) | |
1349 | return ret; | |
1350 | ||
1351 | dump_index_check(existing_stats, calculated_stats, formatter); | |
1352 | flusher.flush(); | |
1353 | ||
1354 | return 0; | |
1355 | } | |
1356 | ||
1357 | int RGWBucketAdminOp::remove_bucket(RGWRados *store, RGWBucketAdminOpState& op_state, | |
1358 | bool bypass_gc, bool keep_index_consistent) | |
1359 | { | |
1360 | RGWBucket bucket; | |
1361 | ||
1362 | int ret = bucket.init(store, op_state); | |
1363 | if (ret < 0) | |
1364 | return ret; | |
1365 | ||
1366 | std::string err_msg; | |
1367 | ret = bucket.remove(op_state, bypass_gc, keep_index_consistent, &err_msg); | |
1368 | if (!err_msg.empty()) { | |
1369 | lderr(store->ctx()) << "ERROR: " << err_msg << dendl; | |
1370 | } | |
1371 | return ret; | |
1372 | } | |
1373 | ||
1374 | int RGWBucketAdminOp::remove_object(RGWRados *store, RGWBucketAdminOpState& op_state) | |
1375 | { | |
1376 | RGWBucket bucket; | |
1377 | ||
1378 | int ret = bucket.init(store, op_state); | |
1379 | if (ret < 0) | |
1380 | return ret; | |
1381 | ||
1382 | return bucket.remove_object(op_state); | |
1383 | } | |
1384 | ||
1385 | static int bucket_stats(RGWRados *store, const std::string& tenant_name, std::string& bucket_name, Formatter *formatter) | |
1386 | { | |
1387 | RGWBucketInfo bucket_info; | |
1388 | map<RGWObjCategory, RGWStorageStats> stats; | |
1389 | ||
1390 | real_time mtime; | |
1391 | RGWObjectCtx obj_ctx(store); | |
1392 | int r = store->get_bucket_info(obj_ctx, tenant_name, bucket_name, bucket_info, &mtime); | |
1393 | if (r < 0) | |
1394 | return r; | |
1395 | ||
1396 | rgw_bucket& bucket = bucket_info.bucket; | |
1397 | ||
1398 | string bucket_ver, master_ver; | |
1399 | string max_marker; | |
1400 | int ret = store->get_bucket_stats(bucket_info, RGW_NO_SHARD, &bucket_ver, &master_ver, stats, &max_marker); | |
1401 | if (ret < 0) { | |
1402 | cerr << "error getting bucket stats ret=" << ret << std::endl; | |
1403 | return ret; | |
1404 | } | |
1405 | ||
1406 | utime_t ut(mtime); | |
1407 | ||
1408 | formatter->open_object_section("stats"); | |
1409 | formatter->dump_string("bucket", bucket.name); | |
1410 | formatter->dump_string("zonegroup", bucket_info.zonegroup); | |
1411 | formatter->dump_string("placement_rule", bucket_info.placement_rule); | |
1412 | ::encode_json("explicit_placement", bucket.explicit_placement, formatter); | |
1413 | formatter->dump_string("id", bucket.bucket_id); | |
1414 | formatter->dump_string("marker", bucket.marker); | |
1415 | formatter->dump_stream("index_type") << bucket_info.index_type; | |
1416 | ::encode_json("owner", bucket_info.owner, formatter); | |
1417 | formatter->dump_string("ver", bucket_ver); | |
1418 | formatter->dump_string("master_ver", master_ver); | |
1419 | formatter->dump_stream("mtime") << ut; | |
1420 | formatter->dump_string("max_marker", max_marker); | |
1421 | dump_bucket_usage(stats, formatter); | |
1422 | encode_json("bucket_quota", bucket_info.quota, formatter); | |
1423 | formatter->close_section(); | |
1424 | ||
1425 | return 0; | |
1426 | } | |
1427 | ||
1428 | int RGWBucketAdminOp::limit_check(RGWRados *store, | |
1429 | RGWBucketAdminOpState& op_state, | |
1430 | const std::list<std::string>& user_ids, | |
1431 | RGWFormatterFlusher& flusher, | |
1432 | bool warnings_only) | |
1433 | { | |
1434 | int ret = 0; | |
1435 | const size_t max_entries = | |
1436 | store->ctx()->_conf->rgw_list_buckets_max_chunk; | |
1437 | ||
1438 | const size_t safe_max_objs_per_shard = | |
1439 | store->ctx()->_conf->rgw_safe_max_objects_per_shard; | |
1440 | ||
1441 | uint16_t shard_warn_pct = | |
1442 | store->ctx()->_conf->rgw_shard_warning_threshold; | |
1443 | if (shard_warn_pct > 100) | |
1444 | shard_warn_pct = 90; | |
1445 | ||
1446 | Formatter *formatter = flusher.get_formatter(); | |
1447 | flusher.start(0); | |
1448 | ||
1449 | formatter->open_array_section("users"); | |
1450 | ||
1451 | for (const auto& user_id : user_ids) { | |
1452 | formatter->open_object_section("user"); | |
1453 | formatter->dump_string("user_id", user_id); | |
1454 | bool done; | |
1455 | formatter->open_array_section("buckets"); | |
1456 | do { | |
1457 | RGWUserBuckets buckets; | |
1458 | string marker; | |
1459 | bool is_truncated; | |
1460 | ||
1461 | ret = rgw_read_user_buckets(store, user_id, buckets, | |
1462 | marker, string(), max_entries, false, | |
1463 | &is_truncated); | |
1464 | if (ret < 0) | |
1465 | return ret; | |
1466 | ||
1467 | map<string, RGWBucketEnt>& m_buckets = buckets.get_buckets(); | |
1468 | ||
1469 | for (const auto& iter : m_buckets) { | |
1470 | auto& bucket = iter.second.bucket; | |
1471 | uint32_t num_shards = 1; | |
1472 | uint64_t num_objects = 0; | |
1473 | ||
1474 | /* need info for num_shards */ | |
1475 | RGWBucketInfo info; | |
1476 | RGWObjectCtx obj_ctx(store); | |
1477 | ||
1478 | marker = bucket.name; /* Casey's location for marker update, | |
1479 | * as we may now not reach the end of | |
1480 | * the loop body */ | |
1481 | ||
1482 | ret = store->get_bucket_info(obj_ctx, bucket.tenant, bucket.name, | |
1483 | info, nullptr); | |
1484 | if (ret < 0) | |
1485 | continue; | |
1486 | ||
1487 | /* need stats for num_entries */ | |
1488 | string bucket_ver, master_ver; | |
1489 | std::map<RGWObjCategory, RGWStorageStats> stats; | |
1490 | ret = store->get_bucket_stats(info, RGW_NO_SHARD, &bucket_ver, | |
1491 | &master_ver, stats, nullptr); | |
1492 | ||
1493 | if (ret < 0) | |
1494 | continue; | |
1495 | ||
1496 | for (const auto& s : stats) { | |
1497 | num_objects += s.second.num_objects; | |
1498 | } | |
1499 | ||
1500 | num_shards = info.num_shards; | |
1501 | uint64_t objs_per_shard = | |
1502 | (num_shards) ? num_objects/num_shards : num_objects; | |
1503 | { | |
1504 | bool warn = false; | |
1505 | stringstream ss; | |
1506 | if (objs_per_shard > safe_max_objs_per_shard) { | |
1507 | double over = | |
1508 | 100 - (safe_max_objs_per_shard/objs_per_shard * 100); | |
1509 | ss << boost::format("OVER %4f%%") % over; | |
1510 | warn = true; | |
1511 | } else { | |
1512 | double fill_pct = | |
1513 | objs_per_shard / safe_max_objs_per_shard * 100; | |
1514 | if (fill_pct >= shard_warn_pct) { | |
1515 | ss << boost::format("WARN %4f%%") % fill_pct; | |
1516 | warn = true; | |
1517 | } else { | |
1518 | ss << "OK"; | |
1519 | } | |
1520 | } | |
1521 | ||
1522 | if (warn || (! warnings_only)) { | |
1523 | formatter->open_object_section("bucket"); | |
1524 | formatter->dump_string("bucket", bucket.name); | |
1525 | formatter->dump_string("tenant", bucket.tenant); | |
1526 | formatter->dump_int("num_objects", num_objects); | |
1527 | formatter->dump_int("num_shards", num_shards); | |
1528 | formatter->dump_int("objects_per_shard", objs_per_shard); | |
1529 | formatter->dump_string("fill_status", ss.str()); | |
1530 | formatter->close_section(); | |
1531 | } | |
1532 | } | |
1533 | } | |
1534 | ||
1535 | done = (m_buckets.size() < max_entries); | |
1536 | } while (!done); /* foreach: bucket */ | |
1537 | ||
1538 | formatter->close_section(); | |
1539 | formatter->close_section(); | |
1540 | formatter->flush(cout); | |
1541 | ||
1542 | } /* foreach: user_id */ | |
1543 | ||
1544 | formatter->close_section(); | |
1545 | formatter->flush(cout); | |
1546 | ||
1547 | return ret; | |
1548 | } /* RGWBucketAdminOp::limit_check */ | |
1549 | ||
1550 | int RGWBucketAdminOp::info(RGWRados *store, RGWBucketAdminOpState& op_state, | |
1551 | RGWFormatterFlusher& flusher) | |
1552 | { | |
1553 | RGWBucket bucket; | |
1554 | int ret; | |
1555 | ||
1556 | string bucket_name = op_state.get_bucket_name(); | |
1557 | ||
1558 | if (!bucket_name.empty()) { | |
1559 | ret = bucket.init(store, op_state); | |
1560 | if (ret < 0) | |
1561 | return ret; | |
1562 | } | |
1563 | ||
1564 | Formatter *formatter = flusher.get_formatter(); | |
1565 | flusher.start(0); | |
1566 | ||
1567 | CephContext *cct = store->ctx(); | |
1568 | ||
1569 | const size_t max_entries = cct->_conf->rgw_list_buckets_max_chunk; | |
1570 | ||
1571 | bool show_stats = op_state.will_fetch_stats(); | |
1572 | rgw_user user_id = op_state.get_user_id(); | |
1573 | if (op_state.is_user_op()) { | |
1574 | formatter->open_array_section("buckets"); | |
1575 | ||
1576 | RGWUserBuckets buckets; | |
1577 | string marker; | |
1578 | bool is_truncated = false; | |
1579 | ||
1580 | do { | |
1581 | ret = rgw_read_user_buckets(store, op_state.get_user_id(), buckets, | |
1582 | marker, string(), max_entries, false, | |
1583 | &is_truncated); | |
1584 | if (ret < 0) | |
1585 | return ret; | |
1586 | ||
1587 | map<string, RGWBucketEnt>& m = buckets.get_buckets(); | |
1588 | map<string, RGWBucketEnt>::iterator iter; | |
1589 | ||
1590 | for (iter = m.begin(); iter != m.end(); ++iter) { | |
1591 | std::string obj_name = iter->first; | |
1592 | if (show_stats) | |
1593 | bucket_stats(store, user_id.tenant, obj_name, formatter); | |
1594 | else | |
1595 | formatter->dump_string("bucket", obj_name); | |
1596 | ||
1597 | marker = obj_name; | |
1598 | } | |
1599 | ||
1600 | flusher.flush(); | |
1601 | } while (is_truncated); | |
1602 | ||
1603 | formatter->close_section(); | |
1604 | } else if (!bucket_name.empty()) { | |
1605 | bucket_stats(store, user_id.tenant, bucket_name, formatter); | |
1606 | } else { | |
1607 | RGWAccessHandle handle; | |
1608 | ||
1609 | formatter->open_array_section("buckets"); | |
1610 | if (store->list_buckets_init(&handle) >= 0) { | |
1611 | rgw_bucket_dir_entry obj; | |
1612 | while (store->list_buckets_next(obj, &handle) >= 0) { | |
1613 | if (show_stats) | |
1614 | bucket_stats(store, user_id.tenant, obj.key.name, formatter); | |
1615 | else | |
1616 | formatter->dump_string("bucket", obj.key.name); | |
1617 | } | |
1618 | } | |
1619 | ||
1620 | formatter->close_section(); | |
1621 | } | |
1622 | ||
1623 | flusher.flush(); | |
1624 | ||
1625 | return 0; | |
1626 | } | |
1627 | ||
1628 | ||
1629 | void rgw_data_change::dump(Formatter *f) const | |
1630 | { | |
1631 | string type; | |
1632 | switch (entity_type) { | |
1633 | case ENTITY_TYPE_BUCKET: | |
1634 | type = "bucket"; | |
1635 | break; | |
1636 | default: | |
1637 | type = "unknown"; | |
1638 | } | |
1639 | encode_json("entity_type", type, f); | |
1640 | encode_json("key", key, f); | |
1641 | utime_t ut(timestamp); | |
1642 | encode_json("timestamp", ut, f); | |
1643 | } | |
1644 | ||
1645 | void rgw_data_change::decode_json(JSONObj *obj) { | |
1646 | string s; | |
1647 | JSONDecoder::decode_json("entity_type", s, obj); | |
1648 | if (s == "bucket") { | |
1649 | entity_type = ENTITY_TYPE_BUCKET; | |
1650 | } else { | |
1651 | entity_type = ENTITY_TYPE_UNKNOWN; | |
1652 | } | |
1653 | JSONDecoder::decode_json("key", key, obj); | |
1654 | utime_t ut; | |
1655 | JSONDecoder::decode_json("timestamp", ut, obj); | |
1656 | timestamp = ut.to_real_time(); | |
1657 | } | |
1658 | ||
1659 | void rgw_data_change_log_entry::dump(Formatter *f) const | |
1660 | { | |
1661 | encode_json("log_id", log_id, f); | |
1662 | utime_t ut(log_timestamp); | |
1663 | encode_json("log_timestamp", ut, f); | |
1664 | encode_json("entry", entry, f); | |
1665 | } | |
1666 | ||
1667 | void rgw_data_change_log_entry::decode_json(JSONObj *obj) { | |
1668 | JSONDecoder::decode_json("log_id", log_id, obj); | |
1669 | utime_t ut; | |
1670 | JSONDecoder::decode_json("log_timestamp", ut, obj); | |
1671 | log_timestamp = ut.to_real_time(); | |
1672 | JSONDecoder::decode_json("entry", entry, obj); | |
1673 | } | |
1674 | ||
1675 | int RGWDataChangesLog::choose_oid(const rgw_bucket_shard& bs) { | |
1676 | const string& name = bs.bucket.name; | |
1677 | int shard_shift = (bs.shard_id > 0 ? bs.shard_id : 0); | |
1678 | uint32_t r = (ceph_str_hash_linux(name.c_str(), name.size()) + shard_shift) % num_shards; | |
1679 | ||
1680 | return (int)r; | |
1681 | } | |
1682 | ||
1683 | int RGWDataChangesLog::renew_entries() | |
1684 | { | |
1685 | if (!store->need_to_log_data()) | |
1686 | return 0; | |
1687 | ||
1688 | /* we can't keep the bucket name as part of the cls_log_entry, and we need | |
1689 | * it later, so we keep two lists under the map */ | |
1690 | map<int, pair<list<rgw_bucket_shard>, list<cls_log_entry> > > m; | |
1691 | ||
1692 | lock.Lock(); | |
1693 | map<rgw_bucket_shard, bool> entries; | |
1694 | entries.swap(cur_cycle); | |
1695 | lock.Unlock(); | |
1696 | ||
1697 | map<rgw_bucket_shard, bool>::iterator iter; | |
1698 | string section; | |
1699 | real_time ut = real_clock::now(); | |
1700 | for (iter = entries.begin(); iter != entries.end(); ++iter) { | |
1701 | const rgw_bucket_shard& bs = iter->first; | |
1702 | ||
1703 | int index = choose_oid(bs); | |
1704 | ||
1705 | cls_log_entry entry; | |
1706 | ||
1707 | rgw_data_change change; | |
1708 | bufferlist bl; | |
1709 | change.entity_type = ENTITY_TYPE_BUCKET; | |
1710 | change.key = bs.get_key(); | |
1711 | change.timestamp = ut; | |
1712 | ::encode(change, bl); | |
1713 | ||
1714 | store->time_log_prepare_entry(entry, ut, section, change.key, bl); | |
1715 | ||
1716 | m[index].first.push_back(bs); | |
1717 | m[index].second.emplace_back(std::move(entry)); | |
1718 | } | |
1719 | ||
1720 | map<int, pair<list<rgw_bucket_shard>, list<cls_log_entry> > >::iterator miter; | |
1721 | for (miter = m.begin(); miter != m.end(); ++miter) { | |
1722 | list<cls_log_entry>& entries = miter->second.second; | |
1723 | ||
1724 | real_time now = real_clock::now(); | |
1725 | ||
1726 | int ret = store->time_log_add(oids[miter->first], entries, NULL); | |
1727 | if (ret < 0) { | |
1728 | /* we don't really need to have a special handling for failed cases here, | |
1729 | * as this is just an optimization. */ | |
1730 | lderr(cct) << "ERROR: store->time_log_add() returned " << ret << dendl; | |
1731 | return ret; | |
1732 | } | |
1733 | ||
1734 | real_time expiration = now; | |
1735 | expiration += make_timespan(cct->_conf->rgw_data_log_window); | |
1736 | ||
1737 | list<rgw_bucket_shard>& buckets = miter->second.first; | |
1738 | list<rgw_bucket_shard>::iterator liter; | |
1739 | for (liter = buckets.begin(); liter != buckets.end(); ++liter) { | |
1740 | update_renewed(*liter, expiration); | |
1741 | } | |
1742 | } | |
1743 | ||
1744 | return 0; | |
1745 | } | |
1746 | ||
1747 | void RGWDataChangesLog::_get_change(const rgw_bucket_shard& bs, ChangeStatusPtr& status) | |
1748 | { | |
1749 | assert(lock.is_locked()); | |
1750 | if (!changes.find(bs, status)) { | |
1751 | status = ChangeStatusPtr(new ChangeStatus); | |
1752 | changes.add(bs, status); | |
1753 | } | |
1754 | } | |
1755 | ||
1756 | void RGWDataChangesLog::register_renew(rgw_bucket_shard& bs) | |
1757 | { | |
1758 | Mutex::Locker l(lock); | |
1759 | cur_cycle[bs] = true; | |
1760 | } | |
1761 | ||
1762 | void RGWDataChangesLog::update_renewed(rgw_bucket_shard& bs, real_time& expiration) | |
1763 | { | |
1764 | Mutex::Locker l(lock); | |
1765 | ChangeStatusPtr status; | |
1766 | _get_change(bs, status); | |
1767 | ||
1768 | ldout(cct, 20) << "RGWDataChangesLog::update_renewd() bucket_name=" << bs.bucket.name << " shard_id=" << bs.shard_id << " expiration=" << expiration << dendl; | |
1769 | status->cur_expiration = expiration; | |
1770 | } | |
1771 | ||
1772 | int RGWDataChangesLog::get_log_shard_id(rgw_bucket& bucket, int shard_id) { | |
1773 | rgw_bucket_shard bs(bucket, shard_id); | |
1774 | ||
1775 | return choose_oid(bs); | |
1776 | } | |
1777 | ||
1778 | int RGWDataChangesLog::add_entry(rgw_bucket& bucket, int shard_id) { | |
1779 | if (!store->need_to_log_data()) | |
1780 | return 0; | |
1781 | ||
1782 | rgw_bucket_shard bs(bucket, shard_id); | |
1783 | ||
1784 | int index = choose_oid(bs); | |
1785 | mark_modified(index, bs); | |
1786 | ||
1787 | lock.Lock(); | |
1788 | ||
1789 | ChangeStatusPtr status; | |
1790 | _get_change(bs, status); | |
1791 | ||
1792 | lock.Unlock(); | |
1793 | ||
1794 | real_time now = real_clock::now(); | |
1795 | ||
1796 | status->lock->Lock(); | |
1797 | ||
1798 | ldout(cct, 20) << "RGWDataChangesLog::add_entry() bucket.name=" << bucket.name << " shard_id=" << shard_id << " now=" << now << " cur_expiration=" << status->cur_expiration << dendl; | |
1799 | ||
1800 | if (now < status->cur_expiration) { | |
1801 | /* no need to send, recently completed */ | |
1802 | status->lock->Unlock(); | |
1803 | ||
1804 | register_renew(bs); | |
1805 | return 0; | |
1806 | } | |
1807 | ||
1808 | RefCountedCond *cond; | |
1809 | ||
1810 | if (status->pending) { | |
1811 | cond = status->cond; | |
1812 | ||
1813 | assert(cond); | |
1814 | ||
1815 | status->cond->get(); | |
1816 | status->lock->Unlock(); | |
1817 | ||
1818 | int ret = cond->wait(); | |
1819 | cond->put(); | |
1820 | if (!ret) { | |
1821 | register_renew(bs); | |
1822 | } | |
1823 | return ret; | |
1824 | } | |
1825 | ||
1826 | status->cond = new RefCountedCond; | |
1827 | status->pending = true; | |
1828 | ||
1829 | string& oid = oids[index]; | |
1830 | real_time expiration; | |
1831 | ||
1832 | int ret; | |
1833 | ||
1834 | do { | |
1835 | status->cur_sent = now; | |
1836 | ||
1837 | expiration = now; | |
1838 | expiration += ceph::make_timespan(cct->_conf->rgw_data_log_window); | |
1839 | ||
1840 | status->lock->Unlock(); | |
1841 | ||
1842 | bufferlist bl; | |
1843 | rgw_data_change change; | |
1844 | change.entity_type = ENTITY_TYPE_BUCKET; | |
1845 | change.key = bs.get_key(); | |
1846 | change.timestamp = now; | |
1847 | ::encode(change, bl); | |
1848 | string section; | |
1849 | ||
1850 | ldout(cct, 20) << "RGWDataChangesLog::add_entry() sending update with now=" << now << " cur_expiration=" << expiration << dendl; | |
1851 | ||
1852 | ret = store->time_log_add(oid, now, section, change.key, bl); | |
1853 | ||
1854 | now = real_clock::now(); | |
1855 | ||
1856 | status->lock->Lock(); | |
1857 | ||
1858 | } while (!ret && real_clock::now() > expiration); | |
1859 | ||
1860 | cond = status->cond; | |
1861 | ||
1862 | status->pending = false; | |
1863 | status->cur_expiration = status->cur_sent; /* time of when operation started, not completed */ | |
1864 | status->cur_expiration += make_timespan(cct->_conf->rgw_data_log_window); | |
1865 | status->cond = NULL; | |
1866 | status->lock->Unlock(); | |
1867 | ||
1868 | cond->done(ret); | |
1869 | cond->put(); | |
1870 | ||
1871 | return ret; | |
1872 | } | |
1873 | ||
1874 | int RGWDataChangesLog::list_entries(int shard, const real_time& start_time, const real_time& end_time, int max_entries, | |
1875 | list<rgw_data_change_log_entry>& entries, | |
1876 | const string& marker, | |
1877 | string *out_marker, | |
1878 | bool *truncated) { | |
1879 | if (shard >= num_shards) | |
1880 | return -EINVAL; | |
1881 | ||
1882 | list<cls_log_entry> log_entries; | |
1883 | ||
1884 | int ret = store->time_log_list(oids[shard], start_time, end_time, | |
1885 | max_entries, log_entries, marker, | |
1886 | out_marker, truncated); | |
1887 | if (ret < 0) | |
1888 | return ret; | |
1889 | ||
1890 | list<cls_log_entry>::iterator iter; | |
1891 | for (iter = log_entries.begin(); iter != log_entries.end(); ++iter) { | |
1892 | rgw_data_change_log_entry log_entry; | |
1893 | log_entry.log_id = iter->id; | |
1894 | real_time rt = iter->timestamp.to_real_time(); | |
1895 | log_entry.log_timestamp = rt; | |
1896 | bufferlist::iterator liter = iter->data.begin(); | |
1897 | try { | |
1898 | ::decode(log_entry.entry, liter); | |
1899 | } catch (buffer::error& err) { | |
1900 | lderr(cct) << "ERROR: failed to decode data changes log entry" << dendl; | |
1901 | return -EIO; | |
1902 | } | |
1903 | entries.push_back(log_entry); | |
1904 | } | |
1905 | ||
1906 | return 0; | |
1907 | } | |
1908 | ||
1909 | int RGWDataChangesLog::list_entries(const real_time& start_time, const real_time& end_time, int max_entries, | |
1910 | list<rgw_data_change_log_entry>& entries, LogMarker& marker, bool *ptruncated) { | |
1911 | bool truncated; | |
1912 | entries.clear(); | |
1913 | ||
1914 | for (; marker.shard < num_shards && (int)entries.size() < max_entries; | |
1915 | marker.shard++, marker.marker.clear()) { | |
1916 | int ret = list_entries(marker.shard, start_time, end_time, max_entries - entries.size(), entries, | |
1917 | marker.marker, NULL, &truncated); | |
1918 | if (ret == -ENOENT) { | |
1919 | continue; | |
1920 | } | |
1921 | if (ret < 0) { | |
1922 | return ret; | |
1923 | } | |
1924 | if (truncated) { | |
1925 | *ptruncated = true; | |
1926 | return 0; | |
1927 | } | |
1928 | } | |
1929 | ||
1930 | *ptruncated = (marker.shard < num_shards); | |
1931 | ||
1932 | return 0; | |
1933 | } | |
1934 | ||
1935 | int RGWDataChangesLog::get_info(int shard_id, RGWDataChangesLogInfo *info) | |
1936 | { | |
1937 | if (shard_id >= num_shards) | |
1938 | return -EINVAL; | |
1939 | ||
1940 | string oid = oids[shard_id]; | |
1941 | ||
1942 | cls_log_header header; | |
1943 | ||
1944 | int ret = store->time_log_info(oid, &header); | |
1945 | if ((ret < 0) && (ret != -ENOENT)) | |
1946 | return ret; | |
1947 | ||
1948 | info->marker = header.max_marker; | |
1949 | info->last_update = header.max_time.to_real_time(); | |
1950 | ||
1951 | return 0; | |
1952 | } | |
1953 | ||
1954 | int RGWDataChangesLog::trim_entries(int shard_id, const real_time& start_time, const real_time& end_time, | |
1955 | const string& start_marker, const string& end_marker) | |
1956 | { | |
1957 | int ret; | |
1958 | ||
1959 | if (shard_id > num_shards) | |
1960 | return -EINVAL; | |
1961 | ||
1962 | ret = store->time_log_trim(oids[shard_id], start_time, end_time, start_marker, end_marker); | |
1963 | ||
1964 | if (ret == -ENOENT || ret == -ENODATA) | |
1965 | ret = 0; | |
1966 | ||
1967 | return ret; | |
1968 | } | |
1969 | ||
1970 | int RGWDataChangesLog::trim_entries(const real_time& start_time, const real_time& end_time, | |
1971 | const string& start_marker, const string& end_marker) | |
1972 | { | |
1973 | for (int shard = 0; shard < num_shards; shard++) { | |
1974 | int ret = store->time_log_trim(oids[shard], start_time, end_time, start_marker, end_marker); | |
1975 | if (ret == -ENOENT || ret == -ENODATA) { | |
1976 | continue; | |
1977 | } | |
1978 | if (ret < 0) | |
1979 | return ret; | |
1980 | } | |
1981 | ||
1982 | return 0; | |
1983 | } | |
1984 | ||
1985 | bool RGWDataChangesLog::going_down() | |
1986 | { | |
1987 | return down_flag; | |
1988 | } | |
1989 | ||
1990 | RGWDataChangesLog::~RGWDataChangesLog() { | |
1991 | down_flag = true; | |
1992 | renew_thread->stop(); | |
1993 | renew_thread->join(); | |
1994 | delete renew_thread; | |
1995 | delete[] oids; | |
1996 | } | |
1997 | ||
1998 | void *RGWDataChangesLog::ChangesRenewThread::entry() { | |
1999 | do { | |
2000 | dout(2) << "RGWDataChangesLog::ChangesRenewThread: start" << dendl; | |
2001 | int r = log->renew_entries(); | |
2002 | if (r < 0) { | |
2003 | dout(0) << "ERROR: RGWDataChangesLog::renew_entries returned error r=" << r << dendl; | |
2004 | } | |
2005 | ||
2006 | if (log->going_down()) | |
2007 | break; | |
2008 | ||
2009 | int interval = cct->_conf->rgw_data_log_window * 3 / 4; | |
2010 | lock.Lock(); | |
2011 | cond.WaitInterval(lock, utime_t(interval, 0)); | |
2012 | lock.Unlock(); | |
2013 | } while (!log->going_down()); | |
2014 | ||
2015 | return NULL; | |
2016 | } | |
2017 | ||
2018 | void RGWDataChangesLog::ChangesRenewThread::stop() | |
2019 | { | |
2020 | Mutex::Locker l(lock); | |
2021 | cond.Signal(); | |
2022 | } | |
2023 | ||
2024 | void RGWDataChangesLog::mark_modified(int shard_id, const rgw_bucket_shard& bs) | |
2025 | { | |
2026 | auto key = bs.get_key(); | |
2027 | modified_lock.get_read(); | |
2028 | map<int, set<string> >::iterator iter = modified_shards.find(shard_id); | |
2029 | if (iter != modified_shards.end()) { | |
2030 | set<string>& keys = iter->second; | |
2031 | if (keys.find(key) != keys.end()) { | |
2032 | modified_lock.unlock(); | |
2033 | return; | |
2034 | } | |
2035 | } | |
2036 | modified_lock.unlock(); | |
2037 | ||
2038 | RWLock::WLocker wl(modified_lock); | |
2039 | modified_shards[shard_id].insert(key); | |
2040 | } | |
2041 | ||
2042 | void RGWDataChangesLog::read_clear_modified(map<int, set<string> > &modified) | |
2043 | { | |
2044 | RWLock::WLocker wl(modified_lock); | |
2045 | modified.swap(modified_shards); | |
2046 | modified_shards.clear(); | |
2047 | } | |
2048 | ||
2049 | void RGWBucketCompleteInfo::dump(Formatter *f) const { | |
2050 | encode_json("bucket_info", info, f); | |
2051 | encode_json("attrs", attrs, f); | |
2052 | } | |
2053 | ||
2054 | void RGWBucketCompleteInfo::decode_json(JSONObj *obj) { | |
2055 | JSONDecoder::decode_json("bucket_info", info, obj); | |
2056 | JSONDecoder::decode_json("attrs", attrs, obj); | |
2057 | } | |
2058 | ||
2059 | class RGWBucketMetadataHandler : public RGWMetadataHandler { | |
2060 | ||
2061 | public: | |
2062 | string get_type() override { return "bucket"; } | |
2063 | ||
2064 | int get(RGWRados *store, string& entry, RGWMetadataObject **obj) override { | |
2065 | RGWObjVersionTracker ot; | |
2066 | RGWBucketEntryPoint be; | |
2067 | ||
2068 | real_time mtime; | |
2069 | map<string, bufferlist> attrs; | |
2070 | RGWObjectCtx obj_ctx(store); | |
2071 | ||
2072 | string tenant_name, bucket_name; | |
2073 | parse_bucket(entry, &tenant_name, &bucket_name); | |
2074 | int ret = store->get_bucket_entrypoint_info(obj_ctx, tenant_name, bucket_name, be, &ot, &mtime, &attrs); | |
2075 | if (ret < 0) | |
2076 | return ret; | |
2077 | ||
2078 | RGWBucketEntryMetadataObject *mdo = new RGWBucketEntryMetadataObject(be, ot.read_version, mtime); | |
2079 | ||
2080 | *obj = mdo; | |
2081 | ||
2082 | return 0; | |
2083 | } | |
2084 | ||
2085 | int put(RGWRados *store, string& entry, RGWObjVersionTracker& objv_tracker, | |
2086 | real_time mtime, JSONObj *obj, sync_type_t sync_type) override { | |
2087 | RGWBucketEntryPoint be, old_be; | |
2088 | try { | |
2089 | decode_json_obj(be, obj); | |
2090 | } catch (JSONDecoder::err& e) { | |
2091 | return -EINVAL; | |
2092 | } | |
2093 | ||
2094 | real_time orig_mtime; | |
2095 | map<string, bufferlist> attrs; | |
2096 | ||
2097 | RGWObjVersionTracker old_ot; | |
2098 | RGWObjectCtx obj_ctx(store); | |
2099 | ||
2100 | string tenant_name, bucket_name; | |
2101 | parse_bucket(entry, &tenant_name, &bucket_name); | |
2102 | int ret = store->get_bucket_entrypoint_info(obj_ctx, tenant_name, bucket_name, old_be, &old_ot, &orig_mtime, &attrs); | |
2103 | if (ret < 0 && ret != -ENOENT) | |
2104 | return ret; | |
2105 | ||
2106 | // are we actually going to perform this put, or is it too old? | |
2107 | if (ret != -ENOENT && | |
2108 | !check_versions(old_ot.read_version, orig_mtime, | |
2109 | objv_tracker.write_version, mtime, sync_type)) { | |
2110 | return STATUS_NO_APPLY; | |
2111 | } | |
2112 | ||
2113 | objv_tracker.read_version = old_ot.read_version; /* maintain the obj version we just read */ | |
2114 | ||
2115 | ret = store->put_bucket_entrypoint_info(tenant_name, bucket_name, be, false, objv_tracker, mtime, &attrs); | |
2116 | if (ret < 0) | |
2117 | return ret; | |
2118 | ||
2119 | /* link bucket */ | |
2120 | if (be.linked) { | |
2121 | ret = rgw_link_bucket(store, be.owner, be.bucket, be.creation_time, false); | |
2122 | } else { | |
2123 | ret = rgw_unlink_bucket(store, be.owner, be.bucket.tenant, be.bucket.name, false); | |
2124 | } | |
2125 | ||
2126 | return ret; | |
2127 | } | |
2128 | ||
2129 | struct list_keys_info { | |
2130 | RGWRados *store; | |
2131 | RGWListRawObjsCtx ctx; | |
2132 | }; | |
2133 | ||
2134 | int remove(RGWRados *store, string& entry, RGWObjVersionTracker& objv_tracker) override { | |
2135 | RGWBucketEntryPoint be; | |
2136 | RGWObjectCtx obj_ctx(store); | |
2137 | ||
2138 | string tenant_name, bucket_name; | |
2139 | parse_bucket(entry, &tenant_name, &bucket_name); | |
2140 | int ret = store->get_bucket_entrypoint_info(obj_ctx, tenant_name, bucket_name, be, &objv_tracker, NULL, NULL); | |
2141 | if (ret < 0) | |
2142 | return ret; | |
2143 | ||
2144 | /* | |
2145 | * We're unlinking the bucket but we don't want to update the entrypoint here - we're removing | |
2146 | * it immediately and don't want to invalidate our cached objv_version or the bucket obj removal | |
2147 | * will incorrectly fail. | |
2148 | */ | |
2149 | ret = rgw_unlink_bucket(store, be.owner, tenant_name, bucket_name, false); | |
2150 | if (ret < 0) { | |
2151 | lderr(store->ctx()) << "could not unlink bucket=" << entry << " owner=" << be.owner << dendl; | |
2152 | } | |
2153 | ||
2154 | ret = rgw_bucket_delete_bucket_obj(store, tenant_name, bucket_name, objv_tracker); | |
2155 | if (ret < 0) { | |
2156 | lderr(store->ctx()) << "could not delete bucket=" << entry << dendl; | |
2157 | } | |
2158 | /* idempotent */ | |
2159 | return 0; | |
2160 | } | |
2161 | ||
2162 | void get_pool_and_oid(RGWRados *store, const string& key, rgw_pool& pool, string& oid) override { | |
2163 | oid = key; | |
2164 | pool = store->get_zone_params().domain_root; | |
2165 | } | |
2166 | ||
2167 | int list_keys_init(RGWRados *store, void **phandle) override | |
2168 | { | |
2169 | list_keys_info *info = new list_keys_info; | |
2170 | ||
2171 | info->store = store; | |
2172 | ||
2173 | *phandle = (void *)info; | |
2174 | ||
2175 | return 0; | |
2176 | } | |
2177 | ||
2178 | int list_keys_next(void *handle, int max, list<string>& keys, bool *truncated) override { | |
2179 | list_keys_info *info = static_cast<list_keys_info *>(handle); | |
2180 | ||
2181 | string no_filter; | |
2182 | ||
2183 | keys.clear(); | |
2184 | ||
2185 | RGWRados *store = info->store; | |
2186 | ||
2187 | list<string> unfiltered_keys; | |
2188 | ||
2189 | int ret = store->list_raw_objects(store->get_zone_params().domain_root, no_filter, | |
2190 | max, info->ctx, unfiltered_keys, truncated); | |
2191 | if (ret < 0 && ret != -ENOENT) | |
2192 | return ret; | |
2193 | if (ret == -ENOENT) { | |
2194 | if (truncated) | |
2195 | *truncated = false; | |
2196 | return 0; | |
2197 | } | |
2198 | ||
2199 | // now filter out the system entries | |
2200 | list<string>::iterator iter; | |
2201 | for (iter = unfiltered_keys.begin(); iter != unfiltered_keys.end(); ++iter) { | |
2202 | string& k = *iter; | |
2203 | ||
2204 | if (k[0] != '.') { | |
2205 | keys.push_back(k); | |
2206 | } | |
2207 | } | |
2208 | ||
2209 | return 0; | |
2210 | } | |
2211 | ||
2212 | void list_keys_complete(void *handle) override { | |
2213 | list_keys_info *info = static_cast<list_keys_info *>(handle); | |
2214 | delete info; | |
2215 | } | |
2216 | }; | |
2217 | ||
2218 | class RGWBucketInstanceMetadataHandler : public RGWMetadataHandler { | |
2219 | ||
2220 | public: | |
2221 | string get_type() override { return "bucket.instance"; } | |
2222 | ||
2223 | int get(RGWRados *store, string& oid, RGWMetadataObject **obj) override { | |
2224 | RGWBucketCompleteInfo bci; | |
2225 | ||
2226 | real_time mtime; | |
2227 | RGWObjectCtx obj_ctx(store); | |
2228 | ||
2229 | int ret = store->get_bucket_instance_info(obj_ctx, oid, bci.info, &mtime, &bci.attrs); | |
2230 | if (ret < 0) | |
2231 | return ret; | |
2232 | ||
2233 | RGWBucketInstanceMetadataObject *mdo = new RGWBucketInstanceMetadataObject(bci, bci.info.objv_tracker.read_version, mtime); | |
2234 | ||
2235 | *obj = mdo; | |
2236 | ||
2237 | return 0; | |
2238 | } | |
2239 | ||
2240 | int put(RGWRados *store, string& entry, RGWObjVersionTracker& objv_tracker, | |
2241 | real_time mtime, JSONObj *obj, sync_type_t sync_type) override { | |
2242 | RGWBucketCompleteInfo bci, old_bci; | |
2243 | try { | |
2244 | decode_json_obj(bci, obj); | |
2245 | } catch (JSONDecoder::err& e) { | |
2246 | return -EINVAL; | |
2247 | } | |
2248 | ||
2249 | real_time orig_mtime; | |
2250 | RGWObjectCtx obj_ctx(store); | |
2251 | ||
2252 | int ret = store->get_bucket_instance_info(obj_ctx, entry, old_bci.info, | |
2253 | &orig_mtime, &old_bci.attrs); | |
2254 | bool exists = (ret != -ENOENT); | |
2255 | if (ret < 0 && exists) | |
2256 | return ret; | |
2257 | ||
2258 | if (!exists || old_bci.info.bucket.bucket_id != bci.info.bucket.bucket_id) { | |
2259 | /* a new bucket, we need to select a new bucket placement for it */ | |
2260 | auto key(entry); | |
2261 | rgw_bucket_instance_oid_to_key(key); | |
2262 | string tenant_name; | |
2263 | string bucket_name; | |
2264 | string bucket_instance; | |
2265 | parse_bucket(key, &tenant_name, &bucket_name, &bucket_instance); | |
2266 | ||
2267 | RGWZonePlacementInfo rule_info; | |
2268 | bci.info.bucket.name = bucket_name; | |
2269 | bci.info.bucket.bucket_id = bucket_instance; | |
2270 | bci.info.bucket.tenant = tenant_name; | |
2271 | ret = store->select_bucket_location_by_rule(bci.info.placement_rule, &rule_info); | |
2272 | if (ret < 0) { | |
2273 | ldout(store->ctx(), 0) << "ERROR: select_bucket_placement() returned " << ret << dendl; | |
2274 | return ret; | |
2275 | } | |
2276 | bci.info.index_type = rule_info.index_type; | |
2277 | } else { | |
2278 | /* existing bucket, keep its placement */ | |
2279 | bci.info.bucket.explicit_placement = old_bci.info.bucket.explicit_placement; | |
2280 | bci.info.placement_rule = old_bci.info.placement_rule; | |
2281 | } | |
2282 | ||
2283 | if (exists && old_bci.info.datasync_flag_enabled() != bci.info.datasync_flag_enabled()) { | |
2284 | int shards_num = bci.info.num_shards? bci.info.num_shards : 1; | |
2285 | int shard_id = bci.info.num_shards? 0 : -1; | |
2286 | ||
2287 | if (!bci.info.datasync_flag_enabled()) { | |
2288 | ret = store->stop_bi_log_entries(bci.info, -1); | |
2289 | if (ret < 0) { | |
2290 | lderr(store->ctx()) << "ERROR: failed writing bilog" << dendl; | |
2291 | return ret; | |
2292 | } | |
2293 | } else { | |
2294 | ret = store->resync_bi_log_entries(bci.info, -1); | |
2295 | if (ret < 0) { | |
2296 | lderr(store->ctx()) << "ERROR: failed writing bilog" << dendl; | |
2297 | return ret; | |
2298 | } | |
2299 | } | |
2300 | ||
2301 | for (int i = 0; i < shards_num; ++i, ++shard_id) { | |
2302 | ret = store->data_log->add_entry(bci.info.bucket, shard_id); | |
2303 | if (ret < 0) { | |
2304 | lderr(store->ctx()) << "ERROR: failed writing data log" << dendl; | |
2305 | return ret; | |
2306 | } | |
2307 | } | |
2308 | } | |
2309 | ||
2310 | // are we actually going to perform this put, or is it too old? | |
2311 | if (exists && | |
2312 | !check_versions(old_bci.info.objv_tracker.read_version, orig_mtime, | |
2313 | objv_tracker.write_version, mtime, sync_type)) { | |
2314 | objv_tracker.read_version = old_bci.info.objv_tracker.read_version; | |
2315 | return STATUS_NO_APPLY; | |
2316 | } | |
2317 | ||
2318 | /* record the read version (if any), store the new version */ | |
2319 | bci.info.objv_tracker.read_version = old_bci.info.objv_tracker.read_version; | |
2320 | bci.info.objv_tracker.write_version = objv_tracker.write_version; | |
2321 | ||
2322 | ret = store->put_bucket_instance_info(bci.info, false, mtime, &bci.attrs); | |
2323 | if (ret < 0) | |
2324 | return ret; | |
2325 | ||
2326 | objv_tracker = bci.info.objv_tracker; | |
2327 | ||
2328 | ret = store->init_bucket_index(bci.info, bci.info.num_shards); | |
2329 | if (ret < 0) | |
2330 | return ret; | |
2331 | ||
2332 | return STATUS_APPLIED; | |
2333 | } | |
2334 | ||
2335 | struct list_keys_info { | |
2336 | RGWRados *store; | |
2337 | RGWListRawObjsCtx ctx; | |
2338 | }; | |
2339 | ||
2340 | int remove(RGWRados *store, string& entry, RGWObjVersionTracker& objv_tracker) override { | |
2341 | RGWBucketInfo info; | |
2342 | RGWObjectCtx obj_ctx(store); | |
2343 | ||
2344 | int ret = store->get_bucket_instance_info(obj_ctx, entry, info, NULL, NULL); | |
2345 | if (ret < 0 && ret != -ENOENT) | |
2346 | return ret; | |
2347 | ||
2348 | return rgw_bucket_instance_remove_entry(store, entry, &info.objv_tracker); | |
2349 | } | |
2350 | ||
2351 | void get_pool_and_oid(RGWRados *store, const string& key, rgw_pool& pool, string& oid) override { | |
2352 | oid = RGW_BUCKET_INSTANCE_MD_PREFIX + key; | |
2353 | rgw_bucket_instance_key_to_oid(oid); | |
2354 | pool = store->get_zone_params().domain_root; | |
2355 | } | |
2356 | ||
2357 | int list_keys_init(RGWRados *store, void **phandle) override | |
2358 | { | |
2359 | list_keys_info *info = new list_keys_info; | |
2360 | ||
2361 | info->store = store; | |
2362 | ||
2363 | *phandle = (void *)info; | |
2364 | ||
2365 | return 0; | |
2366 | } | |
2367 | ||
2368 | int list_keys_next(void *handle, int max, list<string>& keys, bool *truncated) override { | |
2369 | list_keys_info *info = static_cast<list_keys_info *>(handle); | |
2370 | ||
2371 | string no_filter; | |
2372 | ||
2373 | keys.clear(); | |
2374 | ||
2375 | RGWRados *store = info->store; | |
2376 | ||
2377 | list<string> unfiltered_keys; | |
2378 | ||
2379 | int ret = store->list_raw_objects(store->get_zone_params().domain_root, no_filter, | |
2380 | max, info->ctx, unfiltered_keys, truncated); | |
2381 | if (ret < 0 && ret != -ENOENT) | |
2382 | return ret; | |
2383 | if (ret == -ENOENT) { | |
2384 | if (truncated) | |
2385 | *truncated = false; | |
2386 | return 0; | |
2387 | } | |
2388 | ||
2389 | constexpr int prefix_size = sizeof(RGW_BUCKET_INSTANCE_MD_PREFIX) - 1; | |
2390 | // now filter in the relevant entries | |
2391 | list<string>::iterator iter; | |
2392 | for (iter = unfiltered_keys.begin(); iter != unfiltered_keys.end(); ++iter) { | |
2393 | string& k = *iter; | |
2394 | ||
2395 | if (k.compare(0, prefix_size, RGW_BUCKET_INSTANCE_MD_PREFIX) == 0) { | |
2396 | auto oid = k.substr(prefix_size); | |
2397 | rgw_bucket_instance_oid_to_key(oid); | |
2398 | keys.emplace_back(std::move(oid)); | |
2399 | } | |
2400 | } | |
2401 | ||
2402 | return 0; | |
2403 | } | |
2404 | ||
2405 | void list_keys_complete(void *handle) override { | |
2406 | list_keys_info *info = static_cast<list_keys_info *>(handle); | |
2407 | delete info; | |
2408 | } | |
2409 | ||
2410 | /* | |
2411 | * hash entry for mdlog placement. Use the same hash key we'd have for the bucket entry | |
2412 | * point, so that the log entries end up at the same log shard, so that we process them | |
2413 | * in order | |
2414 | */ | |
2415 | void get_hash_key(const string& section, const string& key, string& hash_key) override { | |
2416 | string k; | |
2417 | int pos = key.find(':'); | |
2418 | if (pos < 0) | |
2419 | k = key; | |
2420 | else | |
2421 | k = key.substr(0, pos); | |
2422 | hash_key = "bucket:" + k; | |
2423 | } | |
2424 | }; | |
2425 | ||
2426 | void rgw_bucket_init(RGWMetadataManager *mm) | |
2427 | { | |
2428 | bucket_meta_handler = new RGWBucketMetadataHandler; | |
2429 | mm->register_handler(bucket_meta_handler); | |
2430 | bucket_instance_meta_handler = new RGWBucketInstanceMetadataHandler; | |
2431 | mm->register_handler(bucket_instance_meta_handler); | |
2432 | } |