1 // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
2 // vim: ts=8 sw=2 smarttab
4 #ifndef CEPH_RGWRADOS_H
5 #define CEPH_RGWRADOS_H
9 #include "include/rados/librados.hpp"
10 #include "include/Context.h"
11 #include "common/RefCountedObj.h"
12 #include "common/RWLock.h"
13 #include "common/ceph_time.h"
14 #include "common/lru_map.h"
15 #include "rgw_common.h"
16 #include "cls/rgw/cls_rgw_types.h"
17 #include "cls/version/cls_version_types.h"
18 #include "cls/log/cls_log_types.h"
19 #include "cls/statelog/cls_statelog_types.h"
20 #include "cls/timeindex/cls_timeindex_types.h"
22 #include "rgw_metadata.h"
23 #include "rgw_meta_sync_status.h"
24 #include "rgw_period_puller.h"
25 #include "rgw_sync_module.h"
31 class RGWMetaNotifier
;
32 class RGWDataNotifier
;
34 class RGWObjectExpirer
;
35 class RGWMetaSyncProcessorThread
;
36 class RGWDataSyncProcessorThread
;
37 class RGWSyncLogTrimThread
;
44 /* flags for put_obj_meta() */
45 #define PUT_OBJ_CREATE 0x01
46 #define PUT_OBJ_EXCL 0x02
47 #define PUT_OBJ_CREATE_EXCL (PUT_OBJ_CREATE | PUT_OBJ_EXCL)
49 #define RGW_OBJ_NS_MULTIPART "multipart"
50 #define RGW_OBJ_NS_SHADOW "shadow"
52 #define RGW_BUCKET_INSTANCE_MD_PREFIX ".bucket.meta."
54 #define RGW_NO_SHARD -1
56 #define RGW_SHARDS_PRIME_0 7877
57 #define RGW_SHARDS_PRIME_1 65521
59 static inline int rgw_shards_mod(unsigned hval
, int max_shards
)
61 if (max_shards
<= RGW_SHARDS_PRIME_0
) {
62 return hval
% RGW_SHARDS_PRIME_0
% max_shards
;
64 return hval
% RGW_SHARDS_PRIME_1
% max_shards
;
67 static inline int rgw_shards_hash(const string
& key
, int max_shards
)
69 return rgw_shards_mod(ceph_str_hash_linux(key
.c_str(), key
.size()), max_shards
);
72 static inline int rgw_shards_max()
74 return RGW_SHARDS_PRIME_1
;
77 static inline void prepend_bucket_marker(const rgw_bucket
& bucket
, const string
& orig_oid
, string
& oid
)
79 if (bucket
.marker
.empty() || orig_oid
.empty()) {
88 static inline void get_obj_bucket_and_oid_loc(const rgw_obj
& obj
, string
& oid
, string
& locator
)
90 const rgw_bucket
& bucket
= obj
.bucket
;
91 prepend_bucket_marker(bucket
, obj
.get_oid(), oid
);
92 const string
& loc
= obj
.key
.get_loc();
94 prepend_bucket_marker(bucket
, loc
, locator
);
100 int rgw_init_ioctx(librados::Rados
*rados
, const rgw_pool
& pool
, librados::IoCtx
& ioctx
, bool create
= false);
102 int rgw_policy_from_attrset(CephContext
*cct
, map
<string
, bufferlist
>& attrset
, RGWAccessControlPolicy
*policy
);
104 static inline bool rgw_raw_obj_to_obj(const rgw_bucket
& bucket
, const rgw_raw_obj
& raw_obj
, rgw_obj
*obj
)
106 ssize_t pos
= raw_obj
.oid
.find('_');
111 if (!rgw_obj_key::parse_raw_oid(raw_obj
.oid
.substr(pos
+ 1), &obj
->key
)) {
114 obj
->bucket
= bucket
;
119 struct rgw_bucket_placement
{
120 string placement_rule
;
123 void dump(Formatter
*f
) const;
126 class rgw_obj_select
{
127 string placement_rule
;
133 rgw_obj_select() : is_raw(false) {}
134 rgw_obj_select(const rgw_obj
& _obj
) : obj(_obj
), is_raw(false) {}
135 rgw_obj_select(const rgw_raw_obj
& _raw_obj
) : raw_obj(_raw_obj
), is_raw(true) {}
136 rgw_obj_select(const rgw_obj_select
& rhs
) {
139 raw_obj
= rhs
.raw_obj
;
145 rgw_raw_obj
get_raw_obj(const RGWZoneGroup
& zonegroup
, const RGWZoneParams
& zone_params
) const;
146 rgw_raw_obj
get_raw_obj(RGWRados
*store
) const;
148 rgw_obj_select
& operator=(const rgw_obj
& rhs
) {
154 rgw_obj_select
& operator=(const rgw_raw_obj
& rhs
) {
160 void set_placement_rule(const string
& rule
) {
161 placement_rule
= rule
;
165 struct compression_block
{
170 void encode(bufferlist
& bl
) const {
171 ENCODE_START(1, 1, bl
);
172 ::encode(old_ofs
, bl
);
173 ::encode(new_ofs
, bl
);
178 void decode(bufferlist::iterator
& bl
) {
180 ::decode(old_ofs
, bl
);
181 ::decode(new_ofs
, bl
);
186 WRITE_CLASS_ENCODER(compression_block
)
188 struct RGWCompressionInfo
{
189 string compression_type
;
191 vector
<compression_block
> blocks
;
193 RGWCompressionInfo() : compression_type("none"), orig_size(0) {}
194 RGWCompressionInfo(const RGWCompressionInfo
& cs_info
) : compression_type(cs_info
.compression_type
),
195 orig_size(cs_info
.orig_size
),
196 blocks(cs_info
.blocks
) {}
198 void encode(bufferlist
& bl
) const {
199 ENCODE_START(1, 1, bl
);
200 ::encode(compression_type
, bl
);
201 ::encode(orig_size
, bl
);
202 ::encode(blocks
, bl
);
206 void decode(bufferlist::iterator
& bl
) {
208 ::decode(compression_type
, bl
);
209 ::decode(orig_size
, bl
);
210 ::decode(blocks
, bl
);
214 WRITE_CLASS_ENCODER(RGWCompressionInfo
)
216 int rgw_compression_info_from_attrset(map
<string
, bufferlist
>& attrs
, bool& need_decompress
, RGWCompressionInfo
& cs_info
);
222 RGWOLHInfo() : removed(false) {}
224 void encode(bufferlist
& bl
) const {
225 ENCODE_START(1, 1, bl
);
226 ::encode(target
, bl
);
227 ::encode(removed
, bl
);
231 void decode(bufferlist::iterator
& bl
) {
233 ::decode(target
, bl
);
234 ::decode(removed
, bl
);
237 static void generate_test_instances(list
<RGWOLHInfo
*>& o
);
238 void dump(Formatter
*f
) const;
240 WRITE_CLASS_ENCODER(RGWOLHInfo
)
242 struct RGWOLHPendingInfo
{
243 ceph::real_time time
;
245 RGWOLHPendingInfo() {}
247 void encode(bufferlist
& bl
) const {
248 ENCODE_START(1, 1, bl
);
253 void decode(bufferlist::iterator
& bl
) {
259 void dump(Formatter
*f
) const;
261 WRITE_CLASS_ENCODER(RGWOLHPendingInfo
)
263 struct RGWUsageBatch
{
264 map
<ceph::real_time
, rgw_usage_log_entry
> m
;
266 void insert(ceph::real_time
& t
, rgw_usage_log_entry
& entry
, bool *account
) {
267 bool exists
= m
.find(t
) != m
.end();
269 m
[t
].aggregate(entry
);
273 struct RGWUsageIter
{
277 RGWUsageIter() : index(0) {}
282 uint64_t extra_data_len
;
284 virtual int handle_data(bufferlist
& bl
, off_t bl_ofs
, off_t bl_len
) = 0;
285 RGWGetDataCB() : extra_data_len(0) {}
286 virtual ~RGWGetDataCB() {}
287 virtual void set_extra_data_len(uint64_t len
) {
288 extra_data_len
= len
;
291 * Flushes any cached data. Used by RGWGetObjFilter.
292 * Return logic same as handle_data.
294 virtual int flush() {
298 * Allows to extend fetch range of RGW object. Used by RGWGetObjFilter.
300 virtual int fixup_range(off_t
& bl_ofs
, off_t
& bl_end
) {
305 class RGWAccessListFilter
{
307 virtual ~RGWAccessListFilter() {}
308 virtual bool filter(string
& name
, string
& key
) = 0;
311 struct RGWCloneRangeInfo
{
318 struct RGWObjManifestPart
{
319 rgw_obj loc
; /* the object where the data is located */
320 uint64_t loc_ofs
; /* the offset at that object where the data is located */
321 uint64_t size
; /* the part size */
323 RGWObjManifestPart() : loc_ofs(0), size(0) {}
325 void encode(bufferlist
& bl
) const {
326 ENCODE_START(2, 2, bl
);
328 ::encode(loc_ofs
, bl
);
333 void decode(bufferlist::iterator
& bl
) {
334 DECODE_START_LEGACY_COMPAT_LEN_32(2, 2, 2, bl
);
336 ::decode(loc_ofs
, bl
);
341 void dump(Formatter
*f
) const;
342 static void generate_test_instances(list
<RGWObjManifestPart
*>& o
);
344 WRITE_CLASS_ENCODER(RGWObjManifestPart
)
347 The manifest defines a set of rules for structuring the object parts.
348 There are a few terms to note:
349 - head: the head part of the object, which is the part that contains
350 the first chunk of data. An object might not have a head (as in the
351 case of multipart-part objects).
352 - stripe: data portion of a single rgw object that resides on a single
354 - part: a collection of stripes that make a contiguous part of an
355 object. A regular object will only have one part (although might have
356 many stripes), a multipart object might have many parts. Each part
357 has a fixed stripe size, although the last stripe of a part might
358 be smaller than that. Consecutive parts may be merged if their stripe
362 struct RGWObjManifestRule
{
363 uint32_t start_part_num
;
365 uint64_t part_size
; /* each part size, 0 if there's no part size, meaning it's unlimited */
366 uint64_t stripe_max_size
; /* underlying obj max size */
367 string override_prefix
;
369 RGWObjManifestRule() : start_part_num(0), start_ofs(0), part_size(0), stripe_max_size(0) {}
370 RGWObjManifestRule(uint32_t _start_part_num
, uint64_t _start_ofs
, uint64_t _part_size
, uint64_t _stripe_max_size
) :
371 start_part_num(_start_part_num
), start_ofs(_start_ofs
), part_size(_part_size
), stripe_max_size(_stripe_max_size
) {}
373 void encode(bufferlist
& bl
) const {
374 ENCODE_START(2, 1, bl
);
375 ::encode(start_part_num
, bl
);
376 ::encode(start_ofs
, bl
);
377 ::encode(part_size
, bl
);
378 ::encode(stripe_max_size
, bl
);
379 ::encode(override_prefix
, bl
);
383 void decode(bufferlist::iterator
& bl
) {
385 ::decode(start_part_num
, bl
);
386 ::decode(start_ofs
, bl
);
387 ::decode(part_size
, bl
);
388 ::decode(stripe_max_size
, bl
);
390 ::decode(override_prefix
, bl
);
393 void dump(Formatter
*f
) const;
395 WRITE_CLASS_ENCODER(RGWObjManifestRule
)
397 class RGWObjManifest
{
399 bool explicit_objs
; /* old manifest? */
400 map
<uint64_t, RGWObjManifestPart
> objs
;
406 string head_placement_rule
;
408 uint64_t max_head_size
;
410 rgw_bucket_placement tail_placement
; /* might be different than the original bucket,
411 as object might have been copied across pools */
412 map
<uint64_t, RGWObjManifestRule
> rules
;
414 string tail_instance
; /* tail object's instance */
416 void convert_to_explicit(const RGWZoneGroup
& zonegroup
, const RGWZoneParams
& zone_params
);
417 int append_explicit(RGWObjManifest
& m
, const RGWZoneGroup
& zonegroup
, const RGWZoneParams
& zone_params
);
418 void append_rules(RGWObjManifest
& m
, map
<uint64_t, RGWObjManifestRule
>::iterator
& iter
, string
*override_prefix
);
420 void update_iterators() {
422 end_iter
.seek(obj_size
);
426 RGWObjManifest() : explicit_objs(false), obj_size(0), head_size(0), max_head_size(0),
427 begin_iter(this), end_iter(this) {}
428 RGWObjManifest(const RGWObjManifest
& rhs
) {
431 RGWObjManifest
& operator=(const RGWObjManifest
& rhs
) {
432 explicit_objs
= rhs
.explicit_objs
;
434 obj_size
= rhs
.obj_size
;
436 head_size
= rhs
.head_size
;
437 max_head_size
= rhs
.max_head_size
;
439 tail_placement
= rhs
.tail_placement
;
441 tail_instance
= rhs
.tail_instance
;
443 begin_iter
.set_manifest(this);
444 end_iter
.set_manifest(this);
446 begin_iter
.seek(rhs
.begin_iter
.get_ofs());
447 end_iter
.seek(rhs
.end_iter
.get_ofs());
452 map
<uint64_t, RGWObjManifestPart
>& get_explicit_objs() {
457 void set_explicit(uint64_t _size
, map
<uint64_t, RGWObjManifestPart
>& _objs
) {
458 explicit_objs
= true;
463 void get_implicit_location(uint64_t cur_part_id
, uint64_t cur_stripe
, uint64_t ofs
, string
*override_prefix
, rgw_obj_select
*location
);
465 void set_trivial_rule(uint64_t tail_ofs
, uint64_t stripe_max_size
) {
466 RGWObjManifestRule
rule(0, tail_ofs
, 0, stripe_max_size
);
468 max_head_size
= tail_ofs
;
471 void set_multipart_part_rule(uint64_t stripe_max_size
, uint64_t part_num
) {
472 RGWObjManifestRule
rule(0, 0, 0, stripe_max_size
);
473 rule
.start_part_num
= part_num
;
478 void encode(bufferlist
& bl
) const {
479 ENCODE_START(7, 6, bl
);
480 ::encode(obj_size
, bl
);
482 ::encode(explicit_objs
, bl
);
484 ::encode(head_size
, bl
);
485 ::encode(max_head_size
, bl
);
486 ::encode(prefix
, bl
);
488 bool encode_tail_bucket
= !(tail_placement
.bucket
== obj
.bucket
);
489 ::encode(encode_tail_bucket
, bl
);
490 if (encode_tail_bucket
) {
491 ::encode(tail_placement
.bucket
, bl
);
493 bool encode_tail_instance
= (tail_instance
!= obj
.key
.instance
);
494 ::encode(encode_tail_instance
, bl
);
495 if (encode_tail_instance
) {
496 ::encode(tail_instance
, bl
);
498 ::encode(head_placement_rule
, bl
);
499 ::encode(tail_placement
.placement_rule
, bl
);
503 void decode(bufferlist::iterator
& bl
) {
504 DECODE_START_LEGACY_COMPAT_LEN_32(7, 2, 2, bl
);
505 ::decode(obj_size
, bl
);
508 ::decode(explicit_objs
, bl
);
510 ::decode(head_size
, bl
);
511 ::decode(max_head_size
, bl
);
512 ::decode(prefix
, bl
);
515 explicit_objs
= true;
517 map
<uint64_t, RGWObjManifestPart
>::iterator iter
= objs
.begin();
518 obj
= iter
->second
.loc
;
519 head_size
= iter
->second
.size
;
520 max_head_size
= head_size
;
524 if (explicit_objs
&& head_size
> 0 && !objs
.empty()) {
525 /* patch up manifest due to issue 16435:
526 * the first object in the explicit objs list might not be the one we need to access, use the
527 * head object instead if set. This would happen if we had an old object that was created
528 * when the explicit objs manifest was around, and it got copied.
530 rgw_obj
& obj_0
= objs
[0].loc
;
531 if (!obj_0
.get_oid().empty() && obj_0
.key
.ns
.empty()) {
533 objs
[0].size
= head_size
;
539 ::decode(tail_placement
.bucket
, bl
);
542 ::decode(need_to_decode
, bl
);
543 if (need_to_decode
) {
544 ::decode(tail_placement
.bucket
, bl
);
546 tail_placement
.bucket
= obj
.bucket
;
553 ::decode(tail_instance
, bl
);
556 ::decode(need_to_decode
, bl
);
557 if (need_to_decode
) {
558 ::decode(tail_instance
, bl
);
560 tail_instance
= obj
.key
.instance
;
563 } else { // old object created before 'tail_instance' field added to manifest
564 tail_instance
= obj
.key
.instance
;
568 ::decode(head_placement_rule
, bl
);
569 ::decode(tail_placement
.placement_rule
, bl
);
576 void dump(Formatter
*f
) const;
577 static void generate_test_instances(list
<RGWObjManifest
*>& o
);
579 int append(RGWObjManifest
& m
, RGWZoneGroup
& zonegroup
, RGWZoneParams
& zone_params
);
580 int append(RGWObjManifest
& m
, RGWRados
*store
);
582 bool get_rule(uint64_t ofs
, RGWObjManifestRule
*rule
);
587 return rules
.empty();
590 bool has_explicit_objs() {
591 return explicit_objs
;
596 if (objs
.size() == 1) {
597 map
<uint64_t, RGWObjManifestPart
>::iterator iter
= objs
.begin();
598 rgw_obj
& o
= iter
->second
.loc
;
601 return (objs
.size() >= 2);
603 return (obj_size
> head_size
);
606 void set_head(const string
& placement_rule
, const rgw_obj
& _o
, uint64_t _s
) {
607 head_placement_rule
= placement_rule
;
611 if (explicit_objs
&& head_size
> 0) {
613 objs
[0].size
= head_size
;
617 const rgw_obj
& get_obj() {
621 void set_tail_placement(const string
& placement_rule
, const rgw_bucket
& _b
) {
622 tail_placement
.placement_rule
= placement_rule
;
623 tail_placement
.bucket
= _b
;
626 const rgw_bucket_placement
& get_tail_placement() {
627 return tail_placement
;
630 const string
& get_head_placement_rule() {
631 return head_placement_rule
;
634 void set_prefix(const string
& _p
) {
638 const string
& get_prefix() {
642 void set_tail_instance(const string
& _ti
) {
646 const string
& get_tail_instance() {
647 return tail_instance
;
650 void set_head_size(uint64_t _s
) {
654 void set_obj_size(uint64_t s
) {
660 uint64_t get_obj_size() {
664 uint64_t get_head_size() {
668 void set_max_head_size(uint64_t s
) {
672 uint64_t get_max_head_size() {
673 return max_head_size
;
677 RGWObjManifest
*manifest
;
678 uint64_t part_ofs
; /* where current part starts */
679 uint64_t stripe_ofs
; /* where current stripe starts */
680 uint64_t ofs
; /* current position within the object */
681 uint64_t stripe_size
; /* current part size */
685 string cur_override_prefix
;
687 rgw_obj_select location
;
689 map
<uint64_t, RGWObjManifestRule
>::iterator rule_iter
;
690 map
<uint64_t, RGWObjManifestRule
>::iterator next_rule_iter
;
692 map
<uint64_t, RGWObjManifestPart
>::iterator explicit_iter
;
703 void update_explicit_pos();
708 void set_manifest(RGWObjManifest
*m
) {
713 obj_iterator() : manifest(NULL
) {
716 explicit obj_iterator(RGWObjManifest
*_m
) : manifest(_m
) {
718 if (!manifest
->empty()) {
722 obj_iterator(RGWObjManifest
*_m
, uint64_t _ofs
) : manifest(_m
) {
724 if (!manifest
->empty()) {
728 void seek(uint64_t ofs
);
731 bool operator==(const obj_iterator
& rhs
) {
732 return (ofs
== rhs
.ofs
);
734 bool operator!=(const obj_iterator
& rhs
) {
735 return (ofs
!= rhs
.ofs
);
737 const rgw_obj_select
& get_location() {
741 /* start of current stripe */
742 uint64_t get_stripe_ofs() {
743 if (manifest
->explicit_objs
) {
744 return explicit_iter
->first
;
749 /* current ofs relative to start of rgw object */
750 uint64_t get_ofs() const {
755 int get_cur_stripe() const {
759 /* current stripe size */
760 uint64_t get_stripe_size() {
761 if (manifest
->explicit_objs
) {
762 return explicit_iter
->second
.size
;
767 /* offset where data starts within current stripe */
768 uint64_t location_ofs() {
769 if (manifest
->explicit_objs
) {
770 return explicit_iter
->second
.loc_ofs
;
772 return 0; /* all stripes start at zero offset */
775 void update_location();
777 friend class RGWObjManifest
;
780 const obj_iterator
& obj_begin();
781 const obj_iterator
& obj_end();
782 obj_iterator
obj_find(uint64_t ofs
);
784 obj_iterator begin_iter
;
785 obj_iterator end_iter
;
788 * simple object generator. Using a simple single rule manifest.
791 RGWObjManifest
*manifest
;
793 uint64_t cur_part_ofs
;
796 uint64_t cur_stripe_size
;
801 rgw_obj_select cur_obj
;
805 RGWObjManifestRule rule
;
808 generator() : manifest(NULL
), last_ofs(0), cur_part_ofs(0), cur_part_id(0),
809 cur_stripe(0), cur_stripe_size(0) {}
810 int create_begin(CephContext
*cct
, RGWObjManifest
*manifest
, const string
& placement_rule
, rgw_bucket
& bucket
, rgw_obj
& obj
);
812 int create_next(uint64_t ofs
);
814 rgw_raw_obj
get_cur_obj(RGWZoneGroup
& zonegroup
, RGWZoneParams
& zone_params
) { return cur_obj
.get_raw_obj(zonegroup
, zone_params
); }
815 rgw_raw_obj
get_cur_obj(RGWRados
*store
) { return cur_obj
.get_raw_obj(store
); }
817 /* total max size of current stripe (including head obj) */
818 uint64_t cur_stripe_max_size() {
819 return cur_stripe_size
;
823 WRITE_CLASS_ENCODER(RGWObjManifest
)
825 struct RGWUploadPartInfo
{
828 uint64_t accounted_size
{0};
830 ceph::real_time modified
;
831 RGWObjManifest manifest
;
832 RGWCompressionInfo cs_info
;
834 RGWUploadPartInfo() : num(0), size(0) {}
836 void encode(bufferlist
& bl
) const {
837 ENCODE_START(4, 2, bl
);
841 ::encode(modified
, bl
);
842 ::encode(manifest
, bl
);
843 ::encode(cs_info
, bl
);
844 ::encode(accounted_size
, bl
);
847 void decode(bufferlist::iterator
& bl
) {
848 DECODE_START_LEGACY_COMPAT_LEN(4, 2, 2, bl
);
852 ::decode(modified
, bl
);
854 ::decode(manifest
, bl
);
856 ::decode(cs_info
, bl
);
857 ::decode(accounted_size
, bl
);
859 accounted_size
= size
;
863 void dump(Formatter
*f
) const;
864 static void generate_test_instances(list
<RGWUploadPartInfo
*>& o
);
866 WRITE_CLASS_ENCODER(RGWUploadPartInfo
)
873 uint64_t size
; //< size of raw object
874 uint64_t accounted_size
{0}; //< size before compression, encryption
875 ceph::real_time mtime
;
880 RGWObjManifest manifest
;
890 uint32_t zone_short_id
;
892 /* important! don't forget to update copy constructor */
894 RGWObjVersionTracker objv_tracker
;
896 map
<string
, bufferlist
> attrset
;
897 RGWObjState() : is_atomic(false), has_attrs(0), exists(false),
898 size(0), epoch(0), fake_tag(false), has_manifest(false),
899 has_data(false), prefetch_data(false), keep_tail(false), is_olh(false),
900 pg_ver(0), zone_short_id(0) {}
901 RGWObjState(const RGWObjState
& rhs
) : obj (rhs
.obj
) {
902 is_atomic
= rhs
.is_atomic
;
903 has_attrs
= rhs
.has_attrs
;
906 accounted_size
= rhs
.accounted_size
;
909 if (rhs
.obj_tag
.length()) {
910 obj_tag
= rhs
.obj_tag
;
912 write_tag
= rhs
.write_tag
;
913 fake_tag
= rhs
.fake_tag
;
914 if (rhs
.has_manifest
) {
915 manifest
= rhs
.manifest
;
917 has_manifest
= rhs
.has_manifest
;
918 shadow_obj
= rhs
.shadow_obj
;
919 has_data
= rhs
.has_data
;
920 if (rhs
.data
.length()) {
923 prefetch_data
= rhs
.prefetch_data
;
924 keep_tail
= rhs
.keep_tail
;
926 objv_tracker
= rhs
.objv_tracker
;
930 bool get_attr(string name
, bufferlist
& dest
) {
931 map
<string
, bufferlist
>::iterator iter
= attrset
.find(name
);
932 if (iter
!= attrset
.end()) {
940 struct RGWRawObjState
{
942 bool has_attrs
{false};
945 ceph::real_time mtime
;
948 bool has_data
{false};
950 bool prefetch_data
{false};
953 /* important! don't forget to update copy constructor */
955 RGWObjVersionTracker objv_tracker
;
957 map
<string
, bufferlist
> attrset
;
959 RGWRawObjState(const RGWRawObjState
& rhs
) : obj (rhs
.obj
) {
960 has_attrs
= rhs
.has_attrs
;
965 if (rhs
.obj_tag
.length()) {
966 obj_tag
= rhs
.obj_tag
;
968 has_data
= rhs
.has_data
;
969 if (rhs
.data
.length()) {
972 prefetch_data
= rhs
.prefetch_data
;
974 objv_tracker
= rhs
.objv_tracker
;
978 struct RGWPoolIterCtx
{
979 librados::IoCtx io_ctx
;
980 librados::NObjectIterator iter
;
983 struct RGWListRawObjsCtx
{
985 RGWPoolIterCtx iter_ctx
;
987 RGWListRawObjsCtx() : initialized(false) {}
990 struct RGWDefaultSystemMetaObjInfo
{
993 void encode(bufferlist
& bl
) const {
994 ENCODE_START(1, 1, bl
);
995 ::encode(default_id
, bl
);
999 void decode(bufferlist::iterator
& bl
) {
1000 DECODE_START(1, bl
);
1001 ::decode(default_id
, bl
);
1005 void dump(Formatter
*f
) const;
1006 void decode_json(JSONObj
*obj
);
1008 WRITE_CLASS_ENCODER(RGWDefaultSystemMetaObjInfo
)
1010 struct RGWNameToId
{
1013 void encode(bufferlist
& bl
) const {
1014 ENCODE_START(1, 1, bl
);
1015 ::encode(obj_id
, bl
);
1019 void decode(bufferlist::iterator
& bl
) {
1020 DECODE_START(1, bl
);
1021 ::decode(obj_id
, bl
);
1025 void dump(Formatter
*f
) const;
1026 void decode_json(JSONObj
*obj
);
1028 WRITE_CLASS_ENCODER(RGWNameToId
)
1030 class RGWSystemMetaObj
{
1038 int store_name(bool exclusive
);
1039 int store_info(bool exclusive
);
1040 int read_info(const string
& obj_id
, bool old_format
= false);
1041 int read_id(const string
& obj_name
, string
& obj_id
);
1042 int read_default(RGWDefaultSystemMetaObjInfo
& default_info
,
1044 /* read and use default id */
1045 int use_default(bool old_format
= false);
1048 RGWSystemMetaObj() : cct(NULL
), store(NULL
) {}
1049 RGWSystemMetaObj(const string
& _name
): name(_name
), cct(NULL
), store(NULL
) {}
1050 RGWSystemMetaObj(const string
& _id
, const string
& _name
) : id(_id
), name(_name
), cct(NULL
), store(NULL
) {}
1051 RGWSystemMetaObj(CephContext
*_cct
, RGWRados
*_store
): cct(_cct
), store(_store
){}
1052 RGWSystemMetaObj(const string
& _name
, CephContext
*_cct
, RGWRados
*_store
): name(_name
), cct(_cct
), store(_store
){}
1053 const string
& get_name() const { return name
; }
1054 const string
& get_id() const { return id
; }
1056 void set_name(const string
& _name
) { name
= _name
;}
1057 void set_id(const string
& _id
) { id
= _id
;}
1058 void clear_id() { id
.clear(); }
1060 virtual ~RGWSystemMetaObj() {}
1062 virtual void encode(bufferlist
& bl
) const {
1063 ENCODE_START(1, 1, bl
);
1069 virtual void decode(bufferlist::iterator
& bl
) {
1070 DECODE_START(1, bl
);
1076 void reinit_instance(CephContext
*_cct
, RGWRados
*_store
) {
1080 int init(CephContext
*_cct
, RGWRados
*_store
, bool setup_obj
= true, bool old_format
= false);
1081 virtual int read_default_id(string
& default_id
, bool old_format
= false);
1082 virtual int set_as_default(bool exclusive
= false);
1083 int delete_default();
1084 virtual int create(bool exclusive
= true);
1085 int delete_obj(bool old_format
= false);
1086 int rename(const string
& new_name
);
1087 int update() { return store_info(false);}
1088 int update_name() { return store_name(false);}
1090 int write(bool exclusive
);
1092 virtual rgw_pool
get_pool(CephContext
*cct
) = 0;
1093 virtual const string
get_default_oid(bool old_format
= false) = 0;
1094 virtual const string
& get_names_oid_prefix() = 0;
1095 virtual const string
& get_info_oid_prefix(bool old_format
= false) = 0;
1096 virtual const string
& get_predefined_name(CephContext
*cct
) = 0;
1098 void dump(Formatter
*f
) const;
1099 void decode_json(JSONObj
*obj
);
1101 WRITE_CLASS_ENCODER(RGWSystemMetaObj
)
1103 struct RGWZonePlacementInfo
{
1104 rgw_pool index_pool
;
1106 rgw_pool data_extra_pool
; /* if not set we should use data_pool */
1107 RGWBucketIndexType index_type
;
1108 std::string compression_type
;
1110 RGWZonePlacementInfo() : index_type(RGWBIType_Normal
) {}
1112 void encode(bufferlist
& bl
) const {
1113 ENCODE_START(6, 1, bl
);
1114 ::encode(index_pool
.to_str(), bl
);
1115 ::encode(data_pool
.to_str(), bl
);
1116 ::encode(data_extra_pool
.to_str(), bl
);
1117 ::encode((uint32_t)index_type
, bl
);
1118 ::encode(compression_type
, bl
);
1122 void decode(bufferlist::iterator
& bl
) {
1123 DECODE_START(6, bl
);
1124 string index_pool_str
;
1125 string data_pool_str
;
1126 ::decode(index_pool_str
, bl
);
1127 index_pool
= rgw_pool(index_pool_str
);
1128 ::decode(data_pool_str
, bl
);
1129 data_pool
= rgw_pool(data_pool_str
);
1130 if (struct_v
>= 4) {
1131 string data_extra_pool_str
;
1132 ::decode(data_extra_pool_str
, bl
);
1133 data_extra_pool
= rgw_pool(data_extra_pool_str
);
1135 if (struct_v
>= 5) {
1138 index_type
= (RGWBucketIndexType
)it
;
1140 if (struct_v
>= 6) {
1141 ::decode(compression_type
, bl
);
1145 const rgw_pool
& get_data_extra_pool() const {
1146 if (data_extra_pool
.empty()) {
1149 return data_extra_pool
;
1151 void dump(Formatter
*f
) const;
1152 void decode_json(JSONObj
*obj
);
1154 WRITE_CLASS_ENCODER(RGWZonePlacementInfo
)
1156 struct RGWZoneParams
: RGWSystemMetaObj
{
1157 rgw_pool domain_root
;
1158 rgw_pool metadata_heap
;
1159 rgw_pool control_pool
;
1163 rgw_pool intent_log_pool
;
1164 rgw_pool usage_log_pool
;
1166 rgw_pool user_keys_pool
;
1167 rgw_pool user_email_pool
;
1168 rgw_pool user_swift_pool
;
1169 rgw_pool user_uid_pool
;
1170 rgw_pool roles_pool
;
1171 rgw_pool reshard_pool
;
1173 RGWAccessKey system_key
;
1175 map
<string
, RGWZonePlacementInfo
> placement_pools
;
1179 map
<string
, string
, ltstr_nocase
> tier_config
;
1181 RGWZoneParams() : RGWSystemMetaObj() {}
1182 RGWZoneParams(const string
& name
) : RGWSystemMetaObj(name
){}
1183 RGWZoneParams(const string
& id
, const string
& name
) : RGWSystemMetaObj(id
, name
) {}
1184 RGWZoneParams(const string
& id
, const string
& name
, const string
& _realm_id
)
1185 : RGWSystemMetaObj(id
, name
), realm_id(_realm_id
) {}
1187 rgw_pool
get_pool(CephContext
*cct
);
1188 const string
get_default_oid(bool old_format
= false) override
;
1189 const string
& get_names_oid_prefix() override
;
1190 const string
& get_info_oid_prefix(bool old_format
= false) override
;
1191 const string
& get_predefined_name(CephContext
*cct
) override
;
1193 int init(CephContext
*_cct
, RGWRados
*_store
, bool setup_obj
= true,
1194 bool old_format
= false);
1195 using RGWSystemMetaObj::init
;
1196 int read_default_id(string
& default_id
, bool old_format
= false) override
;
1197 int set_as_default(bool exclusive
= false) override
;
1198 int create_default(bool old_format
= false);
1199 int create(bool exclusive
= true) override
;
1200 int fix_pool_names();
1202 const string
& get_compression_type(const string
& placement_rule
) const;
1204 void encode(bufferlist
& bl
) const override
{
1205 ENCODE_START(10, 1, bl
);
1206 ::encode(domain_root
, bl
);
1207 ::encode(control_pool
, bl
);
1208 ::encode(gc_pool
, bl
);
1209 ::encode(log_pool
, bl
);
1210 ::encode(intent_log_pool
, bl
);
1211 ::encode(usage_log_pool
, bl
);
1212 ::encode(user_keys_pool
, bl
);
1213 ::encode(user_email_pool
, bl
);
1214 ::encode(user_swift_pool
, bl
);
1215 ::encode(user_uid_pool
, bl
);
1216 RGWSystemMetaObj::encode(bl
);
1217 ::encode(system_key
, bl
);
1218 ::encode(placement_pools
, bl
);
1219 ::encode(metadata_heap
, bl
);
1220 ::encode(realm_id
, bl
);
1221 ::encode(lc_pool
, bl
);
1222 ::encode(tier_config
, bl
);
1223 ::encode(roles_pool
, bl
);
1224 ::encode(reshard_pool
, bl
);
1228 void decode(bufferlist::iterator
& bl
) override
{
1229 DECODE_START(10, bl
);
1230 ::decode(domain_root
, bl
);
1231 ::decode(control_pool
, bl
);
1232 ::decode(gc_pool
, bl
);
1233 ::decode(log_pool
, bl
);
1234 ::decode(intent_log_pool
, bl
);
1235 ::decode(usage_log_pool
, bl
);
1236 ::decode(user_keys_pool
, bl
);
1237 ::decode(user_email_pool
, bl
);
1238 ::decode(user_swift_pool
, bl
);
1239 ::decode(user_uid_pool
, bl
);
1240 if (struct_v
>= 6) {
1241 RGWSystemMetaObj::decode(bl
);
1242 } else if (struct_v
>= 2) {
1247 ::decode(system_key
, bl
);
1249 ::decode(placement_pools
, bl
);
1251 ::decode(metadata_heap
, bl
);
1252 if (struct_v
>= 6) {
1253 ::decode(realm_id
, bl
);
1255 if (struct_v
>= 7) {
1256 ::decode(lc_pool
, bl
);
1258 lc_pool
.init(name
+ ".rgw.lc");
1260 if (struct_v
>= 8) {
1261 ::decode(tier_config
, bl
);
1263 if (struct_v
>= 9) {
1264 ::decode(roles_pool
, bl
);
1266 roles_pool
= name
+ ".rgw.roles";
1268 if (struct_v
>= 10) {
1269 ::decode(reshard_pool
, bl
);
1271 reshard_pool
= name
+ ".rgw.reshard";
1275 void dump(Formatter
*f
) const;
1276 void decode_json(JSONObj
*obj
);
1277 static void generate_test_instances(list
<RGWZoneParams
*>& o
);
1279 bool find_placement(const rgw_data_placement_target
& placement
, string
*placement_id
) {
1280 for (const auto& pp
: placement_pools
) {
1281 const RGWZonePlacementInfo
& info
= pp
.second
;
1282 if (info
.index_pool
== placement
.index_pool
.to_str() &&
1283 info
.data_pool
== placement
.data_pool
.to_str() &&
1284 info
.data_extra_pool
== placement
.data_extra_pool
.to_str()) {
1285 *placement_id
= pp
.first
;
1292 bool get_placement(const string
& placement_id
, RGWZonePlacementInfo
*placement
) const {
1293 auto iter
= placement_pools
.find(placement_id
);
1294 if (iter
== placement_pools
.end()) {
1297 *placement
= iter
->second
;
1302 * return data pool of the head object
1304 bool get_head_data_pool(const string
& placement_id
, const rgw_obj
& obj
, rgw_pool
*pool
) const {
1305 const rgw_data_placement_target
& explicit_placement
= obj
.bucket
.explicit_placement
;
1306 if (!explicit_placement
.data_pool
.empty()) {
1307 if (!obj
.in_extra_data
) {
1308 *pool
= explicit_placement
.data_pool
;
1310 *pool
= explicit_placement
.get_data_extra_pool();
1314 if (placement_id
.empty()) {
1317 auto iter
= placement_pools
.find(placement_id
);
1318 if (iter
== placement_pools
.end()) {
1321 if (!obj
.in_extra_data
) {
1322 *pool
= iter
->second
.data_pool
;
1324 *pool
= iter
->second
.get_data_extra_pool();
1329 WRITE_CLASS_ENCODER(RGWZoneParams
)
1334 list
<string
> endpoints
;
1341 * Represents the number of shards for the bucket index object, a value of zero
1342 * indicates there is no sharding. By default (no sharding, the name of the object
1343 * is '.dir.{marker}', with sharding, the name is '.dir.{marker}.{sharding_id}',
1344 * sharding_id is zero-based value. It is not recommended to set a too large value
1345 * (e.g. thousand) as it increases the cost for bucket listing.
1347 uint32_t bucket_index_max_shards
;
1350 set
<string
> sync_from
; /* list of zones to sync from */
1352 RGWZone() : log_meta(false), log_data(false), read_only(false), bucket_index_max_shards(0),
1353 sync_from_all(true) {}
1355 void encode(bufferlist
& bl
) const {
1356 ENCODE_START(6, 1, bl
);
1358 ::encode(endpoints
, bl
);
1359 ::encode(log_meta
, bl
);
1360 ::encode(log_data
, bl
);
1361 ::encode(bucket_index_max_shards
, bl
);
1363 ::encode(read_only
, bl
);
1364 ::encode(tier_type
, bl
);
1365 ::encode(sync_from_all
, bl
);
1366 ::encode(sync_from
, bl
);
1370 void decode(bufferlist::iterator
& bl
) {
1371 DECODE_START(6, bl
);
1376 ::decode(endpoints
, bl
);
1377 if (struct_v
>= 2) {
1378 ::decode(log_meta
, bl
);
1379 ::decode(log_data
, bl
);
1381 if (struct_v
>= 3) {
1382 ::decode(bucket_index_max_shards
, bl
);
1384 if (struct_v
>= 4) {
1386 ::decode(read_only
, bl
);
1388 if (struct_v
>= 5) {
1389 ::decode(tier_type
, bl
);
1391 if (struct_v
>= 6) {
1392 ::decode(sync_from_all
, bl
);
1393 ::decode(sync_from
, bl
);
1397 void dump(Formatter
*f
) const;
1398 void decode_json(JSONObj
*obj
);
1399 static void generate_test_instances(list
<RGWZone
*>& o
);
1401 bool is_read_only() { return read_only
; }
1403 bool syncs_from(const string
& zone_id
) {
1404 return (sync_from_all
|| sync_from
.find(zone_id
) != sync_from
.end());
1407 WRITE_CLASS_ENCODER(RGWZone
)
1409 struct RGWDefaultZoneGroupInfo
{
1410 string default_zonegroup
;
1412 void encode(bufferlist
& bl
) const {
1413 ENCODE_START(1, 1, bl
);
1414 ::encode(default_zonegroup
, bl
);
1418 void decode(bufferlist::iterator
& bl
) {
1419 DECODE_START(1, bl
);
1420 ::decode(default_zonegroup
, bl
);
1423 void dump(Formatter
*f
) const;
1424 void decode_json(JSONObj
*obj
);
1425 //todo: implement ceph-dencoder
1427 WRITE_CLASS_ENCODER(RGWDefaultZoneGroupInfo
)
1429 struct RGWZoneGroupPlacementTarget
{
1433 bool user_permitted(list
<string
>& user_tags
) {
1437 for (auto& rule
: user_tags
) {
1438 if (tags
.find(rule
) != tags
.end()) {
1445 void encode(bufferlist
& bl
) const {
1446 ENCODE_START(1, 1, bl
);
1452 void decode(bufferlist::iterator
& bl
) {
1453 DECODE_START(1, bl
);
1458 void dump(Formatter
*f
) const;
1459 void decode_json(JSONObj
*obj
);
1461 WRITE_CLASS_ENCODER(RGWZoneGroupPlacementTarget
)
1464 struct RGWZoneGroup
: public RGWSystemMetaObj
{
1466 list
<string
> endpoints
;
1470 map
<string
, RGWZone
> zones
;
1472 map
<string
, RGWZoneGroupPlacementTarget
> placement_targets
;
1473 string default_placement
;
1475 list
<string
> hostnames
;
1476 list
<string
> hostnames_s3website
;
1477 // TODO: Maybe convert hostnames to a map<string,list<string>> for
1478 // endpoint_type->hostnames
1480 20:05 < _robbat21irssi> maybe I do someting like: if (hostname_map.empty()) { populate all map keys from hostnames; };
1481 20:05 < _robbat21irssi> but that's a later compatability migration planning bit
1482 20:06 < yehudasa> more like if (!hostnames.empty()) {
1483 20:06 < yehudasa> for (list<string>::iterator iter = hostnames.begin(); iter != hostnames.end(); ++iter) {
1484 20:06 < yehudasa> hostname_map["s3"].append(iter->second);
1485 20:07 < yehudasa> hostname_map["s3website"].append(iter->second);
1486 20:07 < yehudasa> s/append/push_back/g
1487 20:08 < _robbat21irssi> inner loop over APIs
1488 20:08 < yehudasa> yeah, probably
1489 20:08 < _robbat21irssi> s3, s3website, swift, swith_auth, swift_website
1491 map
<string
, list
<string
> > api_hostname_map
;
1492 map
<string
, list
<string
> > api_endpoints_map
;
1496 RGWZoneGroup(): is_master(false){}
1497 RGWZoneGroup(const std::string
&id
, const std::string
&name
):RGWSystemMetaObj(id
, name
) {}
1498 RGWZoneGroup(const std::string
&_name
):RGWSystemMetaObj(_name
) {}
1499 RGWZoneGroup(const std::string
&_name
, bool _is_master
, CephContext
*cct
, RGWRados
* store
,
1500 const string
& _realm_id
, const list
<string
>& _endpoints
)
1501 : RGWSystemMetaObj(_name
, cct
, store
), endpoints(_endpoints
), is_master(_is_master
),
1502 realm_id(_realm_id
) {}
1504 bool is_master_zonegroup() const { return is_master
;}
1505 void update_master(bool _is_master
) {
1506 is_master
= _is_master
;
1507 post_process_params();
1509 void post_process_params();
1511 void encode(bufferlist
& bl
) const override
{
1512 ENCODE_START(4, 1, bl
);
1514 ::encode(api_name
, bl
);
1515 ::encode(is_master
, bl
);
1516 ::encode(endpoints
, bl
);
1517 ::encode(master_zone
, bl
);
1518 ::encode(zones
, bl
);
1519 ::encode(placement_targets
, bl
);
1520 ::encode(default_placement
, bl
);
1521 ::encode(hostnames
, bl
);
1522 ::encode(hostnames_s3website
, bl
);
1523 RGWSystemMetaObj::encode(bl
);
1524 ::encode(realm_id
, bl
);
1528 void decode(bufferlist::iterator
& bl
) override
{
1529 DECODE_START(4, bl
);
1531 ::decode(api_name
, bl
);
1532 ::decode(is_master
, bl
);
1533 ::decode(endpoints
, bl
);
1534 ::decode(master_zone
, bl
);
1535 ::decode(zones
, bl
);
1536 ::decode(placement_targets
, bl
);
1537 ::decode(default_placement
, bl
);
1538 if (struct_v
>= 2) {
1539 ::decode(hostnames
, bl
);
1541 if (struct_v
>= 3) {
1542 ::decode(hostnames_s3website
, bl
);
1544 if (struct_v
>= 4) {
1545 RGWSystemMetaObj::decode(bl
);
1546 ::decode(realm_id
, bl
);
1553 int read_default_id(string
& default_id
, bool old_format
= false) override
;
1554 int set_as_default(bool exclusive
= false) override
;
1555 int create_default(bool old_format
= false);
1556 int equals(const string
& other_zonegroup
) const;
1557 int add_zone(const RGWZoneParams
& zone_params
, bool *is_master
, bool *read_only
,
1558 const list
<string
>& endpoints
, const string
*ptier_type
,
1559 bool *psync_from_all
, list
<string
>& sync_from
, list
<string
>& sync_from_rm
);
1560 int remove_zone(const std::string
& zone_id
);
1561 int rename_zone(const RGWZoneParams
& zone_params
);
1562 rgw_pool
get_pool(CephContext
*cct
);
1563 const string
get_default_oid(bool old_region_format
= false) override
;
1564 const string
& get_info_oid_prefix(bool old_region_format
= false) override
;
1565 const string
& get_names_oid_prefix() override
;
1566 const string
& get_predefined_name(CephContext
*cct
) override
;
1568 void dump(Formatter
*f
) const;
1569 void decode_json(JSONObj
*obj
);
1570 static void generate_test_instances(list
<RGWZoneGroup
*>& o
);
1572 WRITE_CLASS_ENCODER(RGWZoneGroup
)
1577 map
<string
, RGWZoneGroup
> zonegroups
;
1578 map
<string
, RGWZoneGroup
> zonegroups_by_api
;
1579 map
<string
, uint32_t> short_zone_ids
;
1581 string master_zonegroup
;
1583 void encode(bufferlist
& bl
) const;
1584 void decode(bufferlist::iterator
& bl
);
1586 int update(const RGWZoneGroup
& zonegroup
, CephContext
*cct
);
1588 void dump(Formatter
*f
) const;
1589 void decode_json(JSONObj
*obj
);
1593 zonegroups_by_api
.clear();
1594 master_zonegroup
.clear();
1597 uint32_t get_zone_short_id(const string
& zone_id
) const;
1599 WRITE_CLASS_ENCODER(RGWPeriodMap
)
1601 struct RGWPeriodConfig
1603 RGWQuotaInfo bucket_quota
;
1604 RGWQuotaInfo user_quota
;
1606 void encode(bufferlist
& bl
) const {
1607 ENCODE_START(1, 1, bl
);
1608 ::encode(bucket_quota
, bl
);
1609 ::encode(user_quota
, bl
);
1613 void decode(bufferlist::iterator
& bl
) {
1614 DECODE_START(1, bl
);
1615 ::decode(bucket_quota
, bl
);
1616 ::decode(user_quota
, bl
);
1620 void dump(Formatter
*f
) const;
1621 void decode_json(JSONObj
*obj
);
1623 // the period config must be stored in a local object outside of the period,
1624 // so that it can be used in a default configuration where no realm/period
1626 int read(RGWRados
*store
, const std::string
& realm_id
);
1627 int write(RGWRados
*store
, const std::string
& realm_id
);
1629 static std::string
get_oid(const std::string
& realm_id
);
1630 static rgw_pool
get_pool(CephContext
*cct
);
1632 WRITE_CLASS_ENCODER(RGWPeriodConfig
)
1634 /* for backward comaptability */
1635 struct RGWRegionMap
{
1637 map
<string
, RGWZoneGroup
> regions
;
1639 string master_region
;
1641 RGWQuotaInfo bucket_quota
;
1642 RGWQuotaInfo user_quota
;
1644 void encode(bufferlist
& bl
) const;
1645 void decode(bufferlist::iterator
& bl
);
1647 void dump(Formatter
*f
) const;
1648 void decode_json(JSONObj
*obj
);
1650 WRITE_CLASS_ENCODER(RGWRegionMap
)
1652 struct RGWZoneGroupMap
{
1654 map
<string
, RGWZoneGroup
> zonegroups
;
1655 map
<string
, RGWZoneGroup
> zonegroups_by_api
;
1657 string master_zonegroup
;
1659 RGWQuotaInfo bucket_quota
;
1660 RGWQuotaInfo user_quota
;
1662 /* constract the map */
1663 int read(CephContext
*cct
, RGWRados
*store
);
1665 void encode(bufferlist
& bl
) const;
1666 void decode(bufferlist::iterator
& bl
);
1668 void dump(Formatter
*f
) const;
1669 void decode_json(JSONObj
*obj
);
1671 WRITE_CLASS_ENCODER(RGWZoneGroupMap
)
1675 struct objexp_hint_entry
{
1679 rgw_obj_key obj_key
;
1680 ceph::real_time exp_time
;
1682 void encode(bufferlist
& bl
) const {
1683 ENCODE_START(2, 1, bl
);
1684 ::encode(bucket_name
, bl
);
1685 ::encode(bucket_id
, bl
);
1686 ::encode(obj_key
, bl
);
1687 ::encode(exp_time
, bl
);
1688 ::encode(tenant
, bl
);
1692 void decode(bufferlist::iterator
& bl
) {
1693 // XXX Do we want DECODE_START_LEGACY_COMPAT_LEN(2, 1, 1, bl); ?
1694 DECODE_START(2, bl
);
1695 ::decode(bucket_name
, bl
);
1696 ::decode(bucket_id
, bl
);
1697 ::decode(obj_key
, bl
);
1698 ::decode(exp_time
, bl
);
1699 if (struct_v
>= 2) {
1700 ::decode(tenant
, bl
);
1707 WRITE_CLASS_ENCODER(objexp_hint_entry
)
1711 class RGWRealm
: public RGWSystemMetaObj
1713 string current_period
;
1714 epoch_t epoch
{0}; //< realm epoch, incremented for each new period
1716 int create_control(bool exclusive
);
1717 int delete_control();
1720 RGWRealm(const string
& _id
, const string
& _name
= "") : RGWSystemMetaObj(_id
, _name
) {}
1721 RGWRealm(CephContext
*_cct
, RGWRados
*_store
): RGWSystemMetaObj(_cct
, _store
) {}
1722 RGWRealm(const string
& _name
, CephContext
*_cct
, RGWRados
*_store
): RGWSystemMetaObj(_name
, _cct
, _store
){}
1724 void encode(bufferlist
& bl
) const override
{
1725 ENCODE_START(1, 1, bl
);
1726 RGWSystemMetaObj::encode(bl
);
1727 ::encode(current_period
, bl
);
1728 ::encode(epoch
, bl
);
1732 void decode(bufferlist::iterator
& bl
) override
{
1733 DECODE_START(1, bl
);
1734 RGWSystemMetaObj::decode(bl
);
1735 ::decode(current_period
, bl
);
1736 ::decode(epoch
, bl
);
1740 int create(bool exclusive
= true) override
;
1742 rgw_pool
get_pool(CephContext
*cct
);
1743 const string
get_default_oid(bool old_format
= false) override
;
1744 const string
& get_names_oid_prefix() override
;
1745 const string
& get_info_oid_prefix(bool old_format
= false) override
;
1746 const string
& get_predefined_name(CephContext
*cct
) override
;
1748 using RGWSystemMetaObj::read_id
; // expose as public for radosgw-admin
1750 void dump(Formatter
*f
) const;
1751 void decode_json(JSONObj
*obj
);
1753 const string
& get_current_period() const {
1754 return current_period
;
1756 int set_current_period(RGWPeriod
& period
);
1757 void clear_current_period_and_epoch() {
1758 current_period
.clear();
1761 epoch_t
get_epoch() const { return epoch
; }
1763 string
get_control_oid();
1764 /// send a notify on the realm control object
1765 int notify_zone(bufferlist
& bl
);
1766 /// notify the zone of a new period
1767 int notify_new_period(const RGWPeriod
& period
);
1769 WRITE_CLASS_ENCODER(RGWRealm
)
1771 struct RGWPeriodLatestEpochInfo
{
1774 void encode(bufferlist
& bl
) const {
1775 ENCODE_START(1, 1, bl
);
1776 ::encode(epoch
, bl
);
1780 void decode(bufferlist::iterator
& bl
) {
1781 DECODE_START(1, bl
);
1782 ::decode(epoch
, bl
);
1786 void dump(Formatter
*f
) const;
1787 void decode_json(JSONObj
*obj
);
1789 WRITE_CLASS_ENCODER(RGWPeriodLatestEpochInfo
)
1795 string predecessor_uuid
;
1796 std::vector
<std::string
> sync_status
;
1797 RGWPeriodMap period_map
;
1798 RGWPeriodConfig period_config
;
1799 string master_zonegroup
;
1804 epoch_t realm_epoch
{1}; //< realm epoch when period was made current
1810 int read_latest_epoch(RGWPeriodLatestEpochInfo
& epoch_info
);
1811 int use_latest_epoch();
1812 int use_current_period();
1814 const string
get_period_oid();
1815 const string
get_period_oid_prefix();
1817 // gather the metadata sync status for each shard; only for use on master zone
1818 int update_sync_status(const RGWPeriod
¤t_period
,
1819 std::ostream
& error_stream
, bool force_if_stale
);
1822 RGWPeriod() : epoch(0), cct(NULL
), store(NULL
) {}
1824 RGWPeriod(const string
& period_id
, epoch_t _epoch
= 0)
1825 : id(period_id
), epoch(_epoch
),
1826 cct(NULL
), store(NULL
) {}
1828 const string
& get_id() const { return id
; }
1829 epoch_t
get_epoch() const { return epoch
; }
1830 epoch_t
get_realm_epoch() const { return realm_epoch
; }
1831 const string
& get_predecessor() const { return predecessor_uuid
; }
1832 const string
& get_master_zone() const { return master_zone
; }
1833 const string
& get_master_zonegroup() const { return master_zonegroup
; }
1834 const string
& get_realm() const { return realm_id
; }
1835 const RGWPeriodMap
& get_map() const { return period_map
; }
1836 RGWPeriodConfig
& get_config() { return period_config
; }
1837 const RGWPeriodConfig
& get_config() const { return period_config
; }
1838 const std::vector
<std::string
>& get_sync_status() const { return sync_status
; }
1839 rgw_pool
get_pool(CephContext
*cct
);
1840 const string
& get_latest_epoch_oid();
1841 const string
& get_info_oid_prefix();
1843 void set_user_quota(RGWQuotaInfo
& user_quota
) {
1844 period_config
.user_quota
= user_quota
;
1847 void set_bucket_quota(RGWQuotaInfo
& bucket_quota
) {
1848 period_config
.bucket_quota
= bucket_quota
;
1851 void set_id(const string
& id
) {
1855 void set_epoch(epoch_t epoch
) { this->epoch
= epoch
; }
1856 void set_realm_epoch(epoch_t epoch
) { realm_epoch
= epoch
; }
1858 void set_predecessor(const string
& predecessor
)
1860 predecessor_uuid
= predecessor
;
1863 void set_realm_id(const string
& _realm_id
) {
1864 realm_id
= _realm_id
;
1869 int get_zonegroup(RGWZoneGroup
& zonegroup
,
1870 const string
& zonegroup_id
);
1872 bool is_single_zonegroup(CephContext
*cct
, RGWRados
*store
);
1874 int get_latest_epoch(epoch_t
& epoch
);
1875 int set_latest_epoch(epoch_t epoch
, bool exclusive
= false);
1877 int init(CephContext
*_cct
, RGWRados
*_store
, const string
&period_realm_id
, const string
&period_realm_name
= "",
1878 bool setup_obj
= true);
1879 int init(CephContext
*_cct
, RGWRados
*_store
, bool setup_obj
= true);
1880 int use_next_epoch();
1882 int create(bool exclusive
= true);
1884 int store_info(bool exclusive
);
1885 int add_zonegroup(const RGWZoneGroup
& zonegroup
);
1890 // commit a staging period; only for use on master zone
1891 int commit(RGWRealm
& realm
, const RGWPeriod
¤t_period
,
1892 std::ostream
& error_stream
, bool force_if_stale
= false);
1894 void encode(bufferlist
& bl
) const {
1895 ENCODE_START(1, 1, bl
);
1897 ::encode(epoch
, bl
);
1898 ::encode(realm_epoch
, bl
);
1899 ::encode(predecessor_uuid
, bl
);
1900 ::encode(sync_status
, bl
);
1901 ::encode(period_map
, bl
);
1902 ::encode(master_zone
, bl
);
1903 ::encode(master_zonegroup
, bl
);
1904 ::encode(period_config
, bl
);
1905 ::encode(realm_id
, bl
);
1906 ::encode(realm_name
, bl
);
1910 void decode(bufferlist::iterator
& bl
) {
1911 DECODE_START(1, bl
);
1913 ::decode(epoch
, bl
);
1914 ::decode(realm_epoch
, bl
);
1915 ::decode(predecessor_uuid
, bl
);
1916 ::decode(sync_status
, bl
);
1917 ::decode(period_map
, bl
);
1918 ::decode(master_zone
, bl
);
1919 ::decode(master_zonegroup
, bl
);
1920 ::decode(period_config
, bl
);
1921 ::decode(realm_id
, bl
);
1922 ::decode(realm_name
, bl
);
1925 void dump(Formatter
*f
) const;
1926 void decode_json(JSONObj
*obj
);
1928 static string
get_staging_id(const string
& realm_id
) {
1929 return realm_id
+ ":staging";
1932 WRITE_CLASS_ENCODER(RGWPeriod
)
1934 class RGWDataChangesLog
;
1935 class RGWMetaSyncStatusManager
;
1936 class RGWDataSyncStatusManager
;
1937 class RGWReplicaLogger
;
1938 class RGWCoroutinesManagerRegistry
;
1945 void oid_str(int shard
, string
& oid
);
1946 int get_shard_num(const string
& object
);
1947 string
get_oid(const string
& object
);
1948 int open_ioctx(librados::IoCtx
& ioctx
);
1958 list_state() : cur_shard(0), max_shard(0) {}
1962 virtual bool dump_entry_internal(const cls_statelog_entry
& entry
, Formatter
*f
) {
1967 RGWStateLog(RGWRados
*_store
, int _num_shards
, const string
& _module_name
) :
1968 store(_store
), num_shards(_num_shards
), module_name(_module_name
) {}
1969 virtual ~RGWStateLog() {}
1971 int store_entry(const string
& client_id
, const string
& op_id
, const string
& object
,
1972 uint32_t state
, bufferlist
*bl
, uint32_t *check_state
);
1974 int remove_entry(const string
& client_id
, const string
& op_id
, const string
& object
);
1976 void init_list_entries(const string
& client_id
, const string
& op_id
, const string
& object
,
1979 int list_entries(void *handle
, int max_entries
, list
<cls_statelog_entry
>& entries
, bool *done
);
1981 void finish_list_entries(void *handle
);
1983 virtual void dump_entry(const cls_statelog_entry
& entry
, Formatter
*f
);
1987 * state transitions:
1989 * unknown -> in-progress -> complete
1992 * user can try setting the 'abort' state, and it can only succeed if state is
1995 * state renewal cannot switch state (stays in the same state)
1997 * rgw can switch from in-progress to complete
1998 * rgw can switch from in-progress to error
2000 * rgw can switch from abort to cancelled
2004 class RGWOpState
: public RGWStateLog
{
2006 bool dump_entry_internal(const cls_statelog_entry
& entry
, Formatter
*f
) override
;
2010 OPSTATE_UNKNOWN
= 0,
2011 OPSTATE_IN_PROGRESS
= 1,
2012 OPSTATE_COMPLETE
= 2,
2015 OPSTATE_CANCELLED
= 5,
2018 explicit RGWOpState(RGWRados
*_store
);
2020 int state_from_str(const string
& s
, OpState
*state
);
2021 int set_state(const string
& client_id
, const string
& op_id
, const string
& object
, OpState state
);
2022 int renew_state(const string
& client_id
, const string
& op_id
, const string
& object
, OpState state
);
2025 class RGWOpStateSingleOp
2034 RGWOpState::OpState cur_state
;
2035 ceph::real_time last_update
;
2038 RGWOpStateSingleOp(RGWRados
*store
, const string
& cid
, const string
& oid
, const string
& obj
);
2040 int set_state(RGWOpState::OpState state
);
2044 class RGWGetBucketStats_CB
: public RefCountedObject
{
2047 map
<RGWObjCategory
, RGWStorageStats
> *stats
;
2049 explicit RGWGetBucketStats_CB(rgw_bucket
& _bucket
) : bucket(_bucket
), stats(NULL
) {}
2050 ~RGWGetBucketStats_CB() override
{}
2051 virtual void handle_response(int r
) = 0;
2052 virtual void set_response(map
<RGWObjCategory
, RGWStorageStats
> *_stats
) {
2057 class RGWGetUserStats_CB
: public RefCountedObject
{
2060 RGWStorageStats stats
;
2062 explicit RGWGetUserStats_CB(const rgw_user
& _user
) : user(_user
) {}
2063 ~RGWGetUserStats_CB() override
{}
2064 virtual void handle_response(int r
) = 0;
2065 virtual void set_response(RGWStorageStats
& _stats
) {
2070 class RGWGetDirHeader_CB
;
2071 class RGWGetUserHeader_CB
;
2073 struct rgw_rados_ref
{
2077 librados::IoCtx ioctx
;
2080 class RGWChainedCache
{
2082 virtual ~RGWChainedCache() {}
2083 virtual void chain_cb(const string
& key
, void *data
) = 0;
2084 virtual void invalidate(const string
& key
) = 0;
2085 virtual void invalidate_all() = 0;
2088 RGWChainedCache
*cache
;
2092 Entry(RGWChainedCache
*_c
, const string
& _k
, void *_d
) : cache(_c
), key(_k
), data(_d
) {}
2096 template <class T
, class S
>
2097 class RGWObjectCtxImpl
{
2099 std::map
<T
, S
> objs_state
;
2103 RGWObjectCtxImpl(RGWRados
*_store
) : store(_store
), lock("RGWObjectCtxImpl") {}
2105 S
*get_state(const T
& obj
) {
2107 typename
std::map
<T
, S
>::iterator iter
;
2109 assert (!obj
.empty());
2110 iter
= objs_state
.find(obj
);
2111 if (iter
!= objs_state
.end()) {
2112 result
= &iter
->second
;
2117 result
= &objs_state
[obj
];
2123 void set_atomic(T
& obj
) {
2124 RWLock::WLocker
wl(lock
);
2125 assert (!obj
.empty());
2126 objs_state
[obj
].is_atomic
= true;
2128 void set_prefetch_data(T
& obj
) {
2129 RWLock::WLocker
wl(lock
);
2130 assert (!obj
.empty());
2131 objs_state
[obj
].prefetch_data
= true;
2133 void invalidate(T
& obj
) {
2134 RWLock::WLocker
wl(lock
);
2135 auto iter
= objs_state
.find(obj
);
2136 if (iter
== objs_state
.end()) {
2139 bool is_atomic
= iter
->second
.is_atomic
;
2140 bool prefetch_data
= iter
->second
.prefetch_data
;
2142 objs_state
.erase(iter
);
2144 if (is_atomic
|| prefetch_data
) {
2145 auto& s
= objs_state
[obj
];
2146 s
.is_atomic
= is_atomic
;
2147 s
.prefetch_data
= prefetch_data
;
2153 void RGWObjectCtxImpl
<rgw_obj
, RGWObjState
>::invalidate(rgw_obj
& obj
);
2156 void RGWObjectCtxImpl
<rgw_raw_obj
, RGWRawObjState
>::invalidate(rgw_raw_obj
& obj
);
2158 struct RGWObjectCtx
{
2162 RGWObjectCtxImpl
<rgw_obj
, RGWObjState
> obj
;
2163 RGWObjectCtxImpl
<rgw_raw_obj
, RGWRawObjState
> raw
;
2165 explicit RGWObjectCtx(RGWRados
*_store
) : store(_store
), user_ctx(NULL
), obj(store
), raw(store
) { }
2166 RGWObjectCtx(RGWRados
*_store
, void *_user_ctx
) : store(_store
), user_ctx(_user_ctx
), obj(store
), raw(store
) { }
2170 class RGWAsyncRadosProcessor
;
2173 class RGWChainedCacheImpl
;
2175 struct bucket_info_entry
{
2178 map
<string
, bufferlist
> attrs
;
2181 struct tombstone_entry
{
2182 ceph::real_time mtime
;
2183 uint32_t zone_short_id
;
2186 tombstone_entry() = default;
2187 tombstone_entry(const RGWObjState
& state
)
2188 : mtime(state
.mtime
), zone_short_id(state
.zone_short_id
),
2189 pg_ver(state
.pg_ver
) {}
2192 class RGWIndexCompletionManager
;
2197 friend class RGWMetaNotifier
;
2198 friend class RGWDataNotifier
;
2200 friend class RGWObjectExpirer
;
2201 friend class RGWMetaSyncProcessorThread
;
2202 friend class RGWDataSyncProcessorThread
;
2203 friend class RGWStateLog
;
2204 friend class RGWReplicaLogger
;
2205 friend class RGWReshard
;
2206 friend class RGWBucketReshard
;
2207 friend class BucketIndexLockGuard
;
2209 /** Open the pool used as root for this gateway */
2210 int open_root_pool_ctx();
2211 int open_gc_pool_ctx();
2212 int open_lc_pool_ctx();
2213 int open_objexp_pool_ctx();
2214 int open_reshard_pool_ctx();
2216 int open_pool_ctx(const rgw_pool
& pool
, librados::IoCtx
& io_ctx
);
2217 int open_bucket_index_ctx(const RGWBucketInfo
& bucket_info
, librados::IoCtx
& index_ctx
);
2218 int open_bucket_index(const RGWBucketInfo
& bucket_info
, librados::IoCtx
& index_ctx
, string
& bucket_oid
);
2219 int open_bucket_index_base(const RGWBucketInfo
& bucket_info
, librados::IoCtx
& index_ctx
,
2220 string
& bucket_oid_base
);
2221 int open_bucket_index_shard(const RGWBucketInfo
& bucket_info
, librados::IoCtx
& index_ctx
,
2222 const string
& obj_key
, string
*bucket_obj
, int *shard_id
);
2223 int open_bucket_index_shard(const RGWBucketInfo
& bucket_info
, librados::IoCtx
& index_ctx
,
2224 int shard_id
, string
*bucket_obj
);
2225 int open_bucket_index(const RGWBucketInfo
& bucket_info
, librados::IoCtx
& index_ctx
,
2226 map
<int, string
>& bucket_objs
, int shard_id
= -1, map
<int, string
> *bucket_instance_ids
= NULL
);
2227 template<typename T
>
2228 int open_bucket_index(const RGWBucketInfo
& bucket_info
, librados::IoCtx
& index_ctx
,
2229 map
<int, string
>& oids
, map
<int, T
>& bucket_objs
,
2230 int shard_id
= -1, map
<int, string
> *bucket_instance_ids
= NULL
);
2231 void build_bucket_index_marker(const string
& shard_id_str
, const string
& shard_marker
,
2234 void get_bucket_instance_ids(const RGWBucketInfo
& bucket_info
, int shard_id
, map
<int, string
> *result
);
2236 std::atomic
<int64_t> max_req_id
= { 0 };
2238 Mutex watchers_lock
;
2243 RGWObjectExpirer
*obj_expirer
;
2247 bool run_sync_thread
;
2248 bool run_reshard_thread
;
2250 RGWAsyncRadosProcessor
* async_rados
;
2252 RGWMetaNotifier
*meta_notifier
;
2253 RGWDataNotifier
*data_notifier
;
2254 RGWMetaSyncProcessorThread
*meta_sync_processor_thread
;
2255 map
<string
, RGWDataSyncProcessorThread
*> data_sync_processor_threads
;
2257 RGWSyncLogTrimThread
*sync_log_trimmer
{nullptr};
2259 Mutex meta_sync_thread_lock
;
2260 Mutex data_sync_thread_lock
;
2263 RGWWatcher
**watchers
;
2264 std::set
<int> watchers_set
;
2265 librados::IoCtx root_pool_ctx
; // .rgw
2266 librados::IoCtx control_pool_ctx
; // .rgw.control
2267 bool watch_initialized
;
2269 friend class RGWWatcher
;
2271 Mutex bucket_id_lock
;
2273 // This field represents the number of bucket index object shards
2274 uint32_t bucket_index_max_shards
;
2276 int get_obj_head_ioctx(const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, librados::IoCtx
*ioctx
);
2277 int get_obj_head_ref(const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, rgw_rados_ref
*ref
);
2278 int get_system_obj_ref(const rgw_raw_obj
& obj
, rgw_rados_ref
*ref
, rgw_pool
*pool
= NULL
);
2279 uint64_t max_bucket_id
;
2281 int get_olh_target_state(RGWObjectCtx
& rctx
, const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
,
2282 RGWObjState
*olh_state
, RGWObjState
**target_state
);
2283 int get_system_obj_state_impl(RGWObjectCtx
*rctx
, rgw_raw_obj
& obj
, RGWRawObjState
**state
, RGWObjVersionTracker
*objv_tracker
);
2284 int get_obj_state_impl(RGWObjectCtx
*rctx
, const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, RGWObjState
**state
,
2285 bool follow_olh
, bool assume_noent
= false);
2286 int append_atomic_test(RGWObjectCtx
*rctx
, const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
,
2287 librados::ObjectOperation
& op
, RGWObjState
**state
);
2289 int update_placement_map();
2290 int store_bucket_info(RGWBucketInfo
& info
, map
<string
, bufferlist
> *pattrs
, RGWObjVersionTracker
*objv_tracker
, bool exclusive
);
2292 void remove_rgw_head_obj(librados::ObjectWriteOperation
& op
);
2293 void cls_obj_check_prefix_exist(librados::ObjectOperation
& op
, const string
& prefix
, bool fail_if_exist
);
2294 void cls_obj_check_mtime(librados::ObjectOperation
& op
, const real_time
& mtime
, bool high_precision_time
, RGWCheckMTimeType type
);
2298 std::vector
<librados::Rados
> rados
;
2299 uint32_t next_rados_handle
;
2301 std::map
<pthread_t
, int> rados_map
;
2303 using RGWChainedCacheImpl_bucket_info_entry
= RGWChainedCacheImpl
<bucket_info_entry
>;
2304 RGWChainedCacheImpl_bucket_info_entry
*binfo_cache
;
2306 using tombstone_cache_t
= lru_map
<rgw_obj
, tombstone_entry
>;
2307 tombstone_cache_t
*obj_tombstone_cache
;
2309 librados::IoCtx gc_pool_ctx
; // .rgw.gc
2310 librados::IoCtx lc_pool_ctx
; // .rgw.lc
2311 librados::IoCtx objexp_pool_ctx
;
2312 librados::IoCtx reshard_pool_ctx
;
2314 bool pools_initialized
;
2316 string zonegroup_id
;
2318 string trans_id_suffix
;
2320 RGWQuotaHandler
*quota_handler
;
2324 RGWCoroutinesManagerRegistry
*cr_registry
;
2326 RGWSyncModulesManager
*sync_modules_manager
{nullptr};
2327 RGWSyncModuleInstanceRef sync_module
;
2328 bool writeable_zone
{false};
2330 RGWZoneGroup zonegroup
;
2331 RGWZone zone_public_config
; /* external zone params, e.g., entrypoints, log flags, etc. */
2332 RGWZoneParams zone_params
; /* internal zone params, e.g., rados pools */
2333 uint32_t zone_short_id
;
2335 RGWPeriod current_period
;
2337 RGWIndexCompletionManager
*index_completion_manager
{nullptr};
2339 RGWRados() : lock("rados_timer_lock"), watchers_lock("watchers_lock"), timer(NULL
),
2340 gc(NULL
), lc(NULL
), obj_expirer(NULL
), use_gc_thread(false), use_lc_thread(false), quota_threads(false),
2341 run_sync_thread(false), run_reshard_thread(false), async_rados(nullptr), meta_notifier(NULL
),
2342 data_notifier(NULL
), meta_sync_processor_thread(NULL
),
2343 meta_sync_thread_lock("meta_sync_thread_lock"), data_sync_thread_lock("data_sync_thread_lock"),
2344 num_watchers(0), watchers(NULL
),
2345 watch_initialized(false),
2346 bucket_id_lock("rados_bucket_id"),
2347 bucket_index_max_shards(0),
2348 max_bucket_id(0), cct(NULL
),
2349 next_rados_handle(0),
2350 handle_lock("rados_handle_lock"),
2351 binfo_cache(NULL
), obj_tombstone_cache(nullptr),
2352 pools_initialized(false),
2353 quota_handler(NULL
),
2357 rest_master_conn(NULL
),
2358 meta_mgr(NULL
), data_log(NULL
), reshard(NULL
) {}
2360 uint64_t get_new_req_id() {
2361 return ++max_req_id
;
2364 librados::IoCtx
* get_lc_pool_ctx() {
2365 return &lc_pool_ctx
;
2367 void set_context(CephContext
*_cct
) {
2372 * AmazonS3 errors contain a HostId string, but is an opaque base64 blob; we
2373 * try to be more transparent. This has a wrapper so we can update it when zonegroup/zone are changed.
2375 void init_host_id() {
2376 /* uint64_t needs 16, two '-' separators and a trailing null */
2377 const string
& zone_name
= get_zone().name
;
2378 const string
& zonegroup_name
= zonegroup
.get_name();
2379 char charbuf
[16 + zone_name
.size() + zonegroup_name
.size() + 2 + 1];
2380 snprintf(charbuf
, sizeof(charbuf
), "%llx-%s-%s", (unsigned long long)instance_id(), zone_name
.c_str(), zonegroup_name
.c_str());
2389 RGWRESTConn
*rest_master_conn
;
2390 map
<string
, RGWRESTConn
*> zone_conn_map
;
2391 map
<string
, RGWRESTConn
*> zone_data_sync_from_map
;
2392 map
<string
, RGWRESTConn
*> zone_data_notify_to_map
;
2393 map
<string
, RGWRESTConn
*> zonegroup_conn_map
;
2395 map
<string
, string
> zone_id_by_name
;
2396 map
<string
, RGWZone
> zone_by_id
;
2398 RGWRESTConn
*get_zone_conn_by_id(const string
& id
) {
2399 auto citer
= zone_conn_map
.find(id
);
2400 if (citer
== zone_conn_map
.end()) {
2404 return citer
->second
;
2407 RGWRESTConn
*get_zone_conn_by_name(const string
& name
) {
2408 auto i
= zone_id_by_name
.find(name
);
2409 if (i
== zone_id_by_name
.end()) {
2413 return get_zone_conn_by_id(i
->second
);
2416 bool find_zone_id_by_name(const string
& name
, string
*id
) {
2417 auto i
= zone_id_by_name
.find(name
);
2418 if (i
== zone_id_by_name
.end()) {
2425 int get_zonegroup(const string
& id
, RGWZoneGroup
& zonegroup
) {
2427 if (id
== get_zonegroup().get_id()) {
2428 zonegroup
= get_zonegroup();
2429 } else if (!current_period
.get_id().empty()) {
2430 ret
= current_period
.get_zonegroup(zonegroup
, id
);
2435 RGWRealm
& get_realm() {
2439 RGWZoneParams
& get_zone_params() { return zone_params
; }
2440 RGWZoneGroup
& get_zonegroup() {
2443 RGWZone
& get_zone() {
2444 return zone_public_config
;
2447 bool zone_is_writeable() {
2448 return writeable_zone
&& !get_zone().is_read_only();
2451 uint32_t get_zone_short_id() const {
2452 return zone_short_id
;
2455 bool zone_syncs_from(RGWZone
& target_zone
, RGWZone
& source_zone
);
2457 const RGWQuotaInfo
& get_bucket_quota() {
2458 return current_period
.get_config().bucket_quota
;
2461 const RGWQuotaInfo
& get_user_quota() {
2462 return current_period
.get_config().user_quota
;
2465 const string
& get_current_period_id() {
2466 return current_period
.get_id();
2469 bool has_zonegroup_api(const std::string
& api
) const {
2470 if (!current_period
.get_id().empty()) {
2471 const auto& zonegroups_by_api
= current_period
.get_map().zonegroups_by_api
;
2472 if (zonegroups_by_api
.find(api
) != zonegroups_by_api
.end())
2478 // pulls missing periods for period_history
2479 std::unique_ptr
<RGWPeriodPuller
> period_puller
;
2480 // maintains a connected history of periods
2481 std::unique_ptr
<RGWPeriodHistory
> period_history
;
2483 RGWAsyncRadosProcessor
* get_async_rados() const { return async_rados
; };
2485 RGWMetadataManager
*meta_mgr
;
2487 RGWDataChangesLog
*data_log
;
2489 RGWReshard
*reshard
;
2490 std::shared_ptr
<RGWReshardWait
> reshard_wait
;
2492 virtual ~RGWRados() = default;
2494 tombstone_cache_t
*get_tombstone_cache() {
2495 return obj_tombstone_cache
;
2498 RGWSyncModulesManager
*get_sync_modules_manager() {
2499 return sync_modules_manager
;
2501 const RGWSyncModuleInstanceRef
& get_sync_module() {
2505 int get_required_alignment(const rgw_pool
& pool
, uint64_t *alignment
);
2506 int get_max_chunk_size(const rgw_pool
& pool
, uint64_t *max_chunk_size
);
2507 int get_max_chunk_size(const string
& placement_rule
, const rgw_obj
& obj
, uint64_t *max_chunk_size
);
2509 uint32_t get_max_bucket_shards() {
2510 return rgw_shards_max();
2513 int get_raw_obj_ref(const rgw_raw_obj
& obj
, rgw_rados_ref
*ref
, rgw_pool
*pool
= NULL
);
2515 int list_raw_objects(const rgw_pool
& pool
, const string
& prefix_filter
, int max
,
2516 RGWListRawObjsCtx
& ctx
, list
<string
>& oids
,
2517 bool *is_truncated
);
2519 int list_raw_prefixed_objs(const rgw_pool
& pool
, const string
& prefix
, list
<string
>& result
);
2520 int list_zonegroups(list
<string
>& zonegroups
);
2521 int list_regions(list
<string
>& regions
);
2522 int list_zones(list
<string
>& zones
);
2523 int list_realms(list
<string
>& realms
);
2524 int list_periods(list
<string
>& periods
);
2525 int list_periods(const string
& current_period
, list
<string
>& periods
);
2528 CephContext
*ctx() { return cct
; }
2529 /** do all necessary setup of the storage device */
2530 int initialize(CephContext
*_cct
, bool _use_gc_thread
, bool _use_lc_thread
, bool _quota_threads
, bool _run_sync_thread
, bool _run_reshard_thread
) {
2532 use_gc_thread
= _use_gc_thread
;
2533 use_lc_thread
= _use_lc_thread
;
2534 quota_threads
= _quota_threads
;
2535 run_sync_thread
= _run_sync_thread
;
2536 run_reshard_thread
= _run_reshard_thread
;
2537 return initialize();
2539 /** Initialize the RADOS instance and prepare to do other ops */
2540 virtual int init_rados();
2541 int init_zg_from_period(bool *initialized
);
2542 int init_zg_from_local(bool *creating_defaults
);
2543 int init_complete();
2544 int replace_region_with_zonegroup();
2545 int convert_regionmap();
2549 void schedule_context(Context
*c
);
2551 /** set up a bucket listing. handle is filled in. */
2552 int list_buckets_init(RGWAccessHandle
*handle
);
2554 * get the next bucket in the listing. obj is filled in,
2555 * handle is updated.
2557 int list_buckets_next(rgw_bucket_dir_entry
& obj
, RGWAccessHandle
*handle
);
2560 int log_list_init(const string
& prefix
, RGWAccessHandle
*handle
);
2561 int log_list_next(RGWAccessHandle handle
, string
*name
);
2564 int log_remove(const string
& name
);
2567 int log_show_init(const string
& name
, RGWAccessHandle
*handle
);
2568 int log_show_next(RGWAccessHandle handle
, rgw_log_entry
*entry
);
2570 // log bandwidth info
2571 int log_usage(map
<rgw_user_bucket
, RGWUsageBatch
>& usage_info
);
2572 int read_usage(const rgw_user
& user
, uint64_t start_epoch
, uint64_t end_epoch
, uint32_t max_entries
,
2573 bool *is_truncated
, RGWUsageIter
& read_iter
, map
<rgw_user_bucket
, rgw_usage_log_entry
>& usage
);
2574 int trim_usage(rgw_user
& user
, uint64_t start_epoch
, uint64_t end_epoch
);
2576 int create_pool(const rgw_pool
& pool
);
2579 * create a bucket with name bucket and the given list of attrs
2580 * returns 0 on success, -ERR# otherwise.
2582 int init_bucket_index(RGWBucketInfo
& bucket_info
, int num_shards
);
2583 int select_bucket_placement(RGWUserInfo
& user_info
, const string
& zonegroup_id
, const string
& rule
,
2584 string
*pselected_rule_name
, RGWZonePlacementInfo
*rule_info
);
2585 int select_legacy_bucket_placement(RGWZonePlacementInfo
*rule_info
);
2586 int select_new_bucket_location(RGWUserInfo
& user_info
, const string
& zonegroup_id
, const string
& rule
,
2587 string
*pselected_rule_name
, RGWZonePlacementInfo
*rule_info
);
2588 int select_bucket_location_by_rule(const string
& location_rule
, RGWZonePlacementInfo
*rule_info
);
2589 void create_bucket_id(string
*bucket_id
);
2591 bool get_obj_data_pool(const string
& placement_rule
, const rgw_obj
& obj
, rgw_pool
*pool
);
2592 bool obj_to_raw(const string
& placement_rule
, const rgw_obj
& obj
, rgw_raw_obj
*raw_obj
);
2594 int create_bucket(RGWUserInfo
& owner
, rgw_bucket
& bucket
,
2595 const string
& zonegroup_id
,
2596 const string
& placement_rule
,
2597 const string
& swift_ver_location
,
2598 const RGWQuotaInfo
* pquota_info
,
2599 map
<std::string
,bufferlist
>& attrs
,
2600 RGWBucketInfo
& bucket_info
,
2602 obj_version
*pep_objv
,
2603 ceph::real_time creation_time
,
2604 rgw_bucket
*master_bucket
,
2605 uint32_t *master_num_shards
,
2606 bool exclusive
= true);
2607 int add_bucket_placement(const rgw_pool
& new_pool
);
2608 int remove_bucket_placement(const rgw_pool
& new_pool
);
2609 int list_placement_set(set
<rgw_pool
>& names
);
2610 int create_pools(vector
<rgw_pool
>& pools
, vector
<int>& retcodes
);
2612 RGWCoroutinesManagerRegistry
*get_cr_registry() { return cr_registry
; }
2614 class SystemObject
{
2622 int get_state(RGWRawObjState
**pstate
, RGWObjVersionTracker
*objv_tracker
);
2625 SystemObject(RGWRados
*_store
, RGWObjectCtx
& _ctx
, rgw_raw_obj
& _obj
) : store(_store
), ctx(_ctx
), obj(_obj
), state(NULL
) {}
2627 void invalidate_state();
2629 RGWRados
*get_store() { return store
; }
2630 rgw_raw_obj
& get_obj() { return obj
; }
2631 RGWObjectCtx
& get_ctx() { return ctx
; }
2634 RGWRados::SystemObject
*source
;
2636 struct GetObjState
{
2638 bool has_ref
{false};
2639 uint64_t last_ver
{0};
2643 int get_ref(RGWRados
*store
, rgw_raw_obj
& obj
, rgw_rados_ref
**pref
);
2647 ceph::real_time
*lastmod
;
2649 map
<string
, bufferlist
> *attrs
;
2651 StatParams() : lastmod(NULL
), obj_size(NULL
), attrs(NULL
) {}
2655 rgw_cache_entry_info
*cache_info
;
2656 map
<string
, bufferlist
> *attrs
;
2658 ReadParams() : attrs(NULL
) {}
2661 explicit Read(RGWRados::SystemObject
*_source
) : source(_source
) {}
2663 int stat(RGWObjVersionTracker
*objv_tracker
);
2664 int read(int64_t ofs
, int64_t end
, bufferlist
& bl
, RGWObjVersionTracker
*objv_tracker
);
2665 int get_attr(const char *name
, bufferlist
& dest
);
2669 struct BucketShard
{
2673 librados::IoCtx index_ctx
;
2676 explicit BucketShard(RGWRados
*_store
) : store(_store
), shard_id(-1) {}
2677 int init(const rgw_bucket
& _bucket
, const rgw_obj
& obj
);
2678 int init(const rgw_bucket
& _bucket
, int sid
);
2683 RGWBucketInfo bucket_info
;
2691 bool versioning_disabled
;
2693 bool bs_initialized
;
2696 int get_state(RGWObjState
**pstate
, bool follow_olh
, bool assume_noent
= false);
2697 void invalidate_state();
2699 int prepare_atomic_modification(librados::ObjectWriteOperation
& op
, bool reset_obj
, const string
*ptag
,
2700 const char *ifmatch
, const char *ifnomatch
, bool removal_op
);
2701 int complete_atomic_modification();
2704 Object(RGWRados
*_store
, const RGWBucketInfo
& _bucket_info
, RGWObjectCtx
& _ctx
, const rgw_obj
& _obj
) : store(_store
), bucket_info(_bucket_info
),
2705 ctx(_ctx
), obj(_obj
), bs(store
),
2706 state(NULL
), versioning_disabled(false),
2707 bs_initialized(false) {}
2709 RGWRados
*get_store() { return store
; }
2710 rgw_obj
& get_obj() { return obj
; }
2711 RGWObjectCtx
& get_ctx() { return ctx
; }
2712 RGWBucketInfo
& get_bucket_info() { return bucket_info
; }
2713 int get_manifest(RGWObjManifest
**pmanifest
);
2715 int get_bucket_shard(BucketShard
**pbs
) {
2716 if (!bs_initialized
) {
2717 int r
= bs
.init(bucket_info
.bucket
, obj
);
2721 bs_initialized
= true;
2727 void set_versioning_disabled(bool status
) {
2728 versioning_disabled
= status
;
2731 bool versioning_enabled() {
2732 return (!versioning_disabled
&& bucket_info
.versioning_enabled());
2736 RGWRados::Object
*source
;
2738 struct GetObjState
{
2739 librados::IoCtx io_ctx
;
2741 rgw_raw_obj head_obj
;
2744 struct ConditionParams
{
2745 const ceph::real_time
*mod_ptr
;
2746 const ceph::real_time
*unmod_ptr
;
2747 bool high_precision_time
;
2748 uint32_t mod_zone_id
;
2749 uint64_t mod_pg_ver
;
2750 const char *if_match
;
2751 const char *if_nomatch
;
2754 mod_ptr(NULL
), unmod_ptr(NULL
), high_precision_time(false), mod_zone_id(0), mod_pg_ver(0),
2755 if_match(NULL
), if_nomatch(NULL
) {}
2759 ceph::real_time
*lastmod
;
2761 map
<string
, bufferlist
> *attrs
;
2763 Params() : lastmod(NULL
), obj_size(NULL
), attrs(NULL
) {}
2766 explicit Read(RGWRados::Object
*_source
) : source(_source
) {}
2769 static int range_to_ofs(uint64_t obj_size
, int64_t &ofs
, int64_t &end
);
2770 int read(int64_t ofs
, int64_t end
, bufferlist
& bl
);
2771 int iterate(int64_t ofs
, int64_t end
, RGWGetDataCB
*cb
);
2772 int get_attr(const char *name
, bufferlist
& dest
);
2776 RGWRados::Object
*target
;
2779 ceph::real_time
*mtime
;
2780 map
<std::string
, bufferlist
>* rmattrs
;
2781 const bufferlist
*data
;
2782 RGWObjManifest
*manifest
;
2784 list
<rgw_obj_index_key
> *remove_objs
;
2785 ceph::real_time set_mtime
;
2787 RGWObjCategory category
;
2789 const char *if_match
;
2790 const char *if_nomatch
;
2792 ceph::real_time delete_at
;
2794 const string
*user_data
;
2795 rgw_zone_set
*zones_trace
;
2797 MetaParams() : mtime(NULL
), rmattrs(NULL
), data(NULL
), manifest(NULL
), ptag(NULL
),
2798 remove_objs(NULL
), category(RGW_OBJ_CATEGORY_MAIN
), flags(0),
2799 if_match(NULL
), if_nomatch(NULL
), olh_epoch(0), canceled(false), user_data(nullptr), zones_trace(nullptr) {}
2802 explicit Write(RGWRados::Object
*_target
) : target(_target
) {}
2804 int _do_write_meta(uint64_t size
, uint64_t accounted_size
,
2805 map
<std::string
, bufferlist
>& attrs
,
2808 int write_meta(uint64_t size
, uint64_t accounted_size
,
2809 map
<std::string
, bufferlist
>& attrs
);
2810 int write_data(const char *data
, uint64_t ofs
, uint64_t len
, bool exclusive
);
2814 RGWRados::Object
*target
;
2816 struct DeleteParams
{
2817 rgw_user bucket_owner
;
2818 int versioning_status
;
2819 ACLOwner obj_owner
; /* needed for creation of deletion marker */
2821 string marker_version_id
;
2822 uint32_t bilog_flags
;
2823 list
<rgw_obj_index_key
> *remove_objs
;
2824 ceph::real_time expiration_time
;
2825 ceph::real_time unmod_since
;
2826 ceph::real_time mtime
; /* for setting delete marker mtime */
2827 bool high_precision_time
;
2828 rgw_zone_set
*zones_trace
;
2830 DeleteParams() : versioning_status(0), olh_epoch(0), bilog_flags(0), remove_objs(NULL
), high_precision_time(false), zones_trace(nullptr) {}
2833 struct DeleteResult
{
2837 DeleteResult() : delete_marker(false) {}
2840 explicit Delete(RGWRados::Object
*_target
) : target(_target
) {}
2846 RGWRados::Object
*source
;
2850 RGWObjManifest manifest
;
2853 struct timespec mtime
;
2854 map
<string
, bufferlist
> attrs
;
2856 Result() : has_manifest(false), size(0) {}
2860 librados::IoCtx io_ctx
;
2861 librados::AioCompletion
*completion
;
2864 State() : completion(NULL
), ret(0) {}
2868 explicit Stat(RGWRados::Object
*_source
) : source(_source
) {}
2880 RGWBucketInfo bucket_info
;
2885 Bucket(RGWRados
*_store
, const RGWBucketInfo
& _bucket_info
) : store(_store
), bucket_info(_bucket_info
), bucket(bucket_info
.bucket
),
2886 shard_id(RGW_NO_SHARD
) {}
2887 RGWRados
*get_store() { return store
; }
2888 rgw_bucket
& get_bucket() { return bucket
; }
2889 RGWBucketInfo
& get_bucket_info() { return bucket_info
; }
2891 int update_bucket_id(const string
& new_bucket_id
);
2893 int get_shard_id() { return shard_id
; }
2894 void set_shard_id(int id
) {
2899 RGWRados::Bucket
*target
;
2902 uint16_t bilog_flags
{0};
2904 bool bs_initialized
{false};
2906 bool prepared
{false};
2907 rgw_zone_set
*zones_trace
{nullptr};
2910 int r
= bs
.init(target
->get_bucket(), obj
);
2914 bs_initialized
= true;
2918 void invalidate_bs() {
2919 bs_initialized
= false;
2922 int guard_reshard(BucketShard
**pbs
, std::function
<int(BucketShard
*)> call
);
2925 UpdateIndex(RGWRados::Bucket
*_target
, const rgw_obj
& _obj
) : target(_target
), obj(_obj
),
2926 bs(target
->get_store()) {
2927 blind
= (target
->get_bucket_info().index_type
== RGWBIType_Indexless
);
2930 int get_bucket_shard(BucketShard
**pbs
) {
2931 if (!bs_initialized
) {
2941 void set_bilog_flags(uint16_t flags
) {
2942 bilog_flags
= flags
;
2945 void set_zones_trace(rgw_zone_set
*_zones_trace
) {
2946 zones_trace
= _zones_trace
;
2949 int prepare(RGWModifyOp
, const string
*write_tag
);
2950 int complete(int64_t poolid
, uint64_t epoch
, uint64_t size
,
2951 uint64_t accounted_size
, ceph::real_time
& ut
,
2952 const string
& etag
, const string
& content_type
,
2953 bufferlist
*acl_bl
, RGWObjCategory category
,
2954 list
<rgw_obj_index_key
> *remove_objs
, const string
*user_data
= nullptr);
2955 int complete_del(int64_t poolid
, uint64_t epoch
,
2956 ceph::real_time
& removed_mtime
, /* mtime of removed object */
2957 list
<rgw_obj_index_key
> *remove_objs
);
2960 const string
*get_optag() { return &optag
; }
2962 bool is_prepared() { return prepared
; }
2966 RGWRados::Bucket
*target
;
2967 rgw_obj_key next_marker
;
2973 rgw_obj_key end_marker
;
2976 RGWAccessListFilter
*filter
;
2979 Params() : enforce_ns(true), filter(NULL
), list_versions(false) {}
2983 explicit List(RGWRados::Bucket
*_target
) : target(_target
) {}
2985 int list_objects(int max
, vector
<rgw_bucket_dir_entry
> *result
, map
<string
, bool> *common_prefixes
, bool *is_truncated
);
2986 rgw_obj_key
& get_next_marker() {
2992 /** Write/overwrite an object to the bucket storage. */
2993 virtual int put_system_obj_impl(rgw_raw_obj
& obj
, uint64_t size
, ceph::real_time
*mtime
,
2994 map
<std::string
, bufferlist
>& attrs
, int flags
,
2996 RGWObjVersionTracker
*objv_tracker
,
2997 ceph::real_time set_mtime
/* 0 for don't set */);
2999 virtual int put_system_obj_data(void *ctx
, rgw_raw_obj
& obj
, bufferlist
& bl
,
3000 off_t ofs
, bool exclusive
,
3001 RGWObjVersionTracker
*objv_tracker
= nullptr);
3002 int aio_put_obj_data(void *ctx
, rgw_raw_obj
& obj
, bufferlist
& bl
,
3003 off_t ofs
, bool exclusive
, void **handle
);
3005 int put_system_obj(void *ctx
, rgw_raw_obj
& obj
, const char *data
, size_t len
, bool exclusive
,
3006 ceph::real_time
*mtime
, map
<std::string
, bufferlist
>& attrs
, RGWObjVersionTracker
*objv_tracker
,
3007 ceph::real_time set_mtime
) {
3009 bl
.append(data
, len
);
3010 int flags
= PUT_OBJ_CREATE
;
3012 flags
|= PUT_OBJ_EXCL
;
3014 return put_system_obj_impl(obj
, len
, mtime
, attrs
, flags
, bl
, objv_tracker
, set_mtime
);
3016 int aio_wait(void *handle
);
3017 bool aio_completed(void *handle
);
3019 int on_last_entry_in_listing(RGWBucketInfo
& bucket_info
,
3020 const std::string
& obj_prefix
,
3021 const std::string
& obj_delim
,
3022 std::function
<int(const rgw_bucket_dir_entry
&)> handler
);
3024 bool swift_versioning_enabled(const RGWBucketInfo
& bucket_info
) const {
3025 return bucket_info
.has_swift_versioning() &&
3026 bucket_info
.swift_ver_location
.size();
3029 int swift_versioning_copy(RGWObjectCtx
& obj_ctx
, /* in/out */
3030 const rgw_user
& user
, /* in */
3031 RGWBucketInfo
& bucket_info
, /* in */
3032 rgw_obj
& obj
); /* in */
3033 int swift_versioning_restore(RGWObjectCtx
& obj_ctx
, /* in/out */
3034 const rgw_user
& user
, /* in */
3035 RGWBucketInfo
& bucket_info
, /* in */
3036 rgw_obj
& obj
, /* in */
3037 bool& restored
); /* out */
3038 int copy_obj_to_remote_dest(RGWObjState
*astate
,
3039 map
<string
, bufferlist
>& src_attrs
,
3040 RGWRados::Object::Read
& read_op
,
3041 const rgw_user
& user_id
,
3043 ceph::real_time
*mtime
);
3047 ATTRSMOD_REPLACE
= 1,
3051 int rewrite_obj(RGWBucketInfo
& dest_bucket_info
, rgw_obj
& obj
);
3053 int stat_remote_obj(RGWObjectCtx
& obj_ctx
,
3054 const rgw_user
& user_id
,
3055 const string
& client_id
,
3057 const string
& source_zone
,
3059 RGWBucketInfo
& src_bucket_info
,
3060 real_time
*src_mtime
,
3062 const real_time
*mod_ptr
,
3063 const real_time
*unmod_ptr
,
3064 bool high_precision_time
,
3065 const char *if_match
,
3066 const char *if_nomatch
,
3067 map
<string
, bufferlist
> *pattrs
,
3072 int fetch_remote_obj(RGWObjectCtx
& obj_ctx
,
3073 const rgw_user
& user_id
,
3074 const string
& client_id
,
3075 const string
& op_id
,
3076 bool record_op_state
,
3078 const string
& source_zone
,
3081 RGWBucketInfo
& dest_bucket_info
,
3082 RGWBucketInfo
& src_bucket_info
,
3083 ceph::real_time
*src_mtime
,
3084 ceph::real_time
*mtime
,
3085 const ceph::real_time
*mod_ptr
,
3086 const ceph::real_time
*unmod_ptr
,
3087 bool high_precision_time
,
3088 const char *if_match
,
3089 const char *if_nomatch
,
3092 map
<string
, bufferlist
>& attrs
,
3093 RGWObjCategory category
,
3095 ceph::real_time delete_at
,
3098 ceph::buffer::list
*petag
,
3099 void (*progress_cb
)(off_t
, void *),
3100 void *progress_data
,
3101 rgw_zone_set
*zones_trace
= nullptr);
3104 * dest_obj: the object to copy into
3105 * src_obj: the object to copy from
3106 * attrs: usage depends on attrs_mod parameter
3107 * attrs_mod: the modification mode of the attrs, may have the following values:
3108 * ATTRSMOD_NONE - the attributes of the source object will be
3109 * copied without modifications, attrs parameter is ignored;
3110 * ATTRSMOD_REPLACE - new object will have the attributes provided by attrs
3111 * parameter, source object attributes are not copied;
3112 * ATTRSMOD_MERGE - any conflicting meta keys on the source object's attributes
3113 * are overwritten by values contained in attrs parameter.
3114 * Returns: 0 on success, -ERR# otherwise.
3116 int copy_obj(RGWObjectCtx
& obj_ctx
,
3117 const rgw_user
& user_id
,
3118 const string
& client_id
,
3119 const string
& op_id
,
3121 const string
& source_zone
,
3124 RGWBucketInfo
& dest_bucket_info
,
3125 RGWBucketInfo
& src_bucket_info
,
3126 ceph::real_time
*src_mtime
,
3127 ceph::real_time
*mtime
,
3128 const ceph::real_time
*mod_ptr
,
3129 const ceph::real_time
*unmod_ptr
,
3130 bool high_precision_time
,
3131 const char *if_match
,
3132 const char *if_nomatch
,
3135 map
<std::string
, bufferlist
>& attrs
,
3136 RGWObjCategory category
,
3138 ceph::real_time delete_at
,
3141 ceph::buffer::list
*petag
,
3142 void (*progress_cb
)(off_t
, void *),
3143 void *progress_data
);
3145 int copy_obj_data(RGWObjectCtx
& obj_ctx
,
3146 RGWBucketInfo
& dest_bucket_info
,
3147 RGWRados::Object::Read
& read_op
, off_t end
,
3150 uint64_t max_chunk_size
,
3151 ceph::real_time
*mtime
,
3152 ceph::real_time set_mtime
,
3153 map
<string
, bufferlist
>& attrs
,
3154 RGWObjCategory category
,
3156 ceph::real_time delete_at
,
3159 ceph::buffer::list
*petag
);
3161 int check_bucket_empty(RGWBucketInfo
& bucket_info
);
3165 * bucket: the name of the bucket to delete
3166 * Returns 0 on success, -ERR# otherwise.
3168 int delete_bucket(RGWBucketInfo
& bucket_info
, RGWObjVersionTracker
& objv_tracker
, bool check_empty
= true);
3170 bool is_meta_master();
3173 * Check to see if the bucket metadata is synced
3175 bool is_syncing_bucket_meta(const rgw_bucket
& bucket
);
3176 void wakeup_meta_sync_shards(set
<int>& shard_ids
);
3177 void wakeup_data_sync_shards(const string
& source_zone
, map
<int, set
<string
> >& shard_ids
);
3179 RGWMetaSyncStatusManager
* get_meta_sync_manager();
3180 RGWDataSyncStatusManager
* get_data_sync_manager(const std::string
& source_zone
);
3182 int set_bucket_owner(rgw_bucket
& bucket
, ACLOwner
& owner
);
3183 int set_buckets_enabled(std::vector
<rgw_bucket
>& buckets
, bool enabled
);
3184 int bucket_suspended(rgw_bucket
& bucket
, bool *suspended
);
3186 /** Delete an object.*/
3187 int delete_obj(RGWObjectCtx
& obj_ctx
,
3188 const RGWBucketInfo
& bucket_owner
,
3189 const rgw_obj
& src_obj
,
3190 int versioning_status
,
3191 uint16_t bilog_flags
= 0,
3192 const ceph::real_time
& expiration_time
= ceph::real_time(),
3193 rgw_zone_set
*zones_trace
= nullptr);
3195 /** Delete a raw object.*/
3196 int delete_raw_obj(const rgw_raw_obj
& obj
);
3198 /* Delete a system object */
3199 virtual int delete_system_obj(rgw_raw_obj
& src_obj
, RGWObjVersionTracker
*objv_tracker
= NULL
);
3201 /** Remove an object from the bucket index */
3202 int delete_obj_index(const rgw_obj
& obj
);
3205 * Get an attribute for a system object.
3206 * obj: the object to get attr
3207 * name: name of the attr to retrieve
3208 * dest: bufferlist to store the result in
3209 * Returns: 0 on success, -ERR# otherwise.
3211 virtual int system_obj_get_attr(rgw_raw_obj
& obj
, const char *name
, bufferlist
& dest
);
3213 int system_obj_set_attr(void *ctx
, rgw_raw_obj
& obj
, const char *name
, bufferlist
& bl
,
3214 RGWObjVersionTracker
*objv_tracker
);
3215 virtual int system_obj_set_attrs(void *ctx
, rgw_raw_obj
& obj
,
3216 map
<string
, bufferlist
>& attrs
,
3217 map
<string
, bufferlist
>* rmattrs
,
3218 RGWObjVersionTracker
*objv_tracker
);
3221 * Set an attr on an object.
3222 * bucket: name of the bucket holding the object
3223 * obj: name of the object to set the attr on
3224 * name: the attr to set
3225 * bl: the contents of the attr
3226 * Returns: 0 on success, -ERR# otherwise.
3228 int set_attr(void *ctx
, const RGWBucketInfo
& bucket_info
, rgw_obj
& obj
, const char *name
, bufferlist
& bl
);
3230 int set_attrs(void *ctx
, const RGWBucketInfo
& bucket_info
, rgw_obj
& obj
,
3231 map
<string
, bufferlist
>& attrs
,
3232 map
<string
, bufferlist
>* rmattrs
);
3234 int get_system_obj_state(RGWObjectCtx
*rctx
, rgw_raw_obj
& obj
, RGWRawObjState
**state
, RGWObjVersionTracker
*objv_tracker
);
3235 int get_obj_state(RGWObjectCtx
*rctx
, const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, RGWObjState
**state
,
3236 bool follow_olh
, bool assume_noent
= false);
3237 int get_obj_state(RGWObjectCtx
*rctx
, const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, RGWObjState
**state
) {
3238 return get_obj_state(rctx
, bucket_info
, obj
, state
, true);
3241 virtual int stat_system_obj(RGWObjectCtx
& obj_ctx
,
3242 RGWRados::SystemObject::Read::GetObjState
& state
,
3244 map
<string
, bufferlist
> *attrs
,
3245 ceph::real_time
*lastmod
,
3247 RGWObjVersionTracker
*objv_tracker
);
3249 virtual int get_system_obj(RGWObjectCtx
& obj_ctx
, RGWRados::SystemObject::Read::GetObjState
& read_state
,
3250 RGWObjVersionTracker
*objv_tracker
, rgw_raw_obj
& obj
,
3251 bufferlist
& bl
, off_t ofs
, off_t end
,
3252 map
<string
, bufferlist
> *attrs
,
3253 rgw_cache_entry_info
*cache_info
);
3255 virtual void register_chained_cache(RGWChainedCache
*cache
) {}
3256 virtual bool chain_cache_entry(list
<rgw_cache_entry_info
*>& cache_info_entries
, RGWChainedCache::Entry
*chained_entry
) { return false; }
3258 int iterate_obj(RGWObjectCtx
& ctx
,
3259 const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
,
3260 off_t ofs
, off_t end
,
3261 uint64_t max_chunk_size
,
3262 int (*iterate_obj_cb
)(const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, const rgw_raw_obj
&, off_t
, off_t
, off_t
, bool, RGWObjState
*, void *),
3265 int flush_read_list(struct get_obj_data
*d
);
3267 int get_obj_iterate_cb(RGWObjectCtx
*ctx
, RGWObjState
*astate
,
3268 const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
,
3269 const rgw_raw_obj
& read_obj
,
3270 off_t obj_ofs
, off_t read_ofs
, off_t len
,
3271 bool is_head_obj
, void *arg
);
3273 void get_obj_aio_completion_cb(librados::completion_t cb
, void *arg
);
3276 * a simple object read without keeping state
3279 virtual int raw_obj_stat(rgw_raw_obj
& obj
, uint64_t *psize
, ceph::real_time
*pmtime
, uint64_t *epoch
,
3280 map
<string
, bufferlist
> *attrs
, bufferlist
*first_chunk
,
3281 RGWObjVersionTracker
*objv_tracker
);
3283 int obj_operate(const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, librados::ObjectWriteOperation
*op
);
3284 int obj_operate(const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, librados::ObjectReadOperation
*op
);
3286 int guard_reshard(BucketShard
*bs
, const rgw_obj
& obj_instance
, std::function
<int(BucketShard
*)> call
);
3287 int block_while_resharding(RGWRados::BucketShard
*bs
, string
*new_bucket_id
);
3289 void bucket_index_guard_olh_op(RGWObjState
& olh_state
, librados::ObjectOperation
& op
);
3290 int olh_init_modification(const RGWBucketInfo
& bucket_info
, RGWObjState
& state
, const rgw_obj
& olh_obj
, string
*op_tag
);
3291 int olh_init_modification_impl(const RGWBucketInfo
& bucket_info
, RGWObjState
& state
, const rgw_obj
& olh_obj
, string
*op_tag
);
3292 int bucket_index_link_olh(const RGWBucketInfo
& bucket_info
, RGWObjState
& olh_state
,
3293 const rgw_obj
& obj_instance
, bool delete_marker
,
3294 const string
& op_tag
, struct rgw_bucket_dir_entry_meta
*meta
,
3296 ceph::real_time unmod_since
, bool high_precision_time
, rgw_zone_set
*zones_trace
= nullptr);
3297 int bucket_index_unlink_instance(const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj_instance
, const string
& op_tag
, const string
& olh_tag
, uint64_t olh_epoch
, rgw_zone_set
*zones_trace
= nullptr);
3298 int bucket_index_read_olh_log(const RGWBucketInfo
& bucket_info
, RGWObjState
& state
, const rgw_obj
& obj_instance
, uint64_t ver_marker
,
3299 map
<uint64_t, vector
<rgw_bucket_olh_log_entry
> > *log
, bool *is_truncated
);
3300 int bucket_index_trim_olh_log(const RGWBucketInfo
& bucket_info
, RGWObjState
& obj_state
, const rgw_obj
& obj_instance
, uint64_t ver
);
3301 int bucket_index_clear_olh(const RGWBucketInfo
& bucket_info
, RGWObjState
& state
, const rgw_obj
& obj_instance
);
3302 int apply_olh_log(RGWObjectCtx
& ctx
, RGWObjState
& obj_state
, const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
,
3303 bufferlist
& obj_tag
, map
<uint64_t, vector
<rgw_bucket_olh_log_entry
> >& log
,
3304 uint64_t *plast_ver
, rgw_zone_set
*zones_trace
= nullptr);
3305 int update_olh(RGWObjectCtx
& obj_ctx
, RGWObjState
*state
, const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, rgw_zone_set
*zones_trace
= nullptr);
3306 int set_olh(RGWObjectCtx
& obj_ctx
, RGWBucketInfo
& bucket_info
, const rgw_obj
& target_obj
, bool delete_marker
, rgw_bucket_dir_entry_meta
*meta
,
3307 uint64_t olh_epoch
, ceph::real_time unmod_since
, bool high_precision_time
, rgw_zone_set
*zones_trace
= nullptr);
3308 int unlink_obj_instance(RGWObjectCtx
& obj_ctx
, RGWBucketInfo
& bucket_info
, const rgw_obj
& target_obj
,
3309 uint64_t olh_epoch
, rgw_zone_set
*zones_trace
= nullptr);
3311 void check_pending_olh_entries(map
<string
, bufferlist
>& pending_entries
, map
<string
, bufferlist
> *rm_pending_entries
);
3312 int remove_olh_pending_entries(const RGWBucketInfo
& bucket_info
, RGWObjState
& state
, const rgw_obj
& olh_obj
, map
<string
, bufferlist
>& pending_attrs
);
3313 int follow_olh(const RGWBucketInfo
& bucket_info
, RGWObjectCtx
& ctx
, RGWObjState
*state
, const rgw_obj
& olh_obj
, rgw_obj
*target
);
3314 int get_olh(const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
, RGWOLHInfo
*olh
);
3316 void gen_rand_obj_instance_name(rgw_obj
*target
);
3318 int omap_get_vals(rgw_raw_obj
& obj
, bufferlist
& header
, const std::string
& marker
, uint64_t count
, std::map
<string
, bufferlist
>& m
);
3319 int omap_get_all(rgw_raw_obj
& obj
, bufferlist
& header
, std::map
<string
, bufferlist
>& m
);
3320 int omap_set(rgw_raw_obj
& obj
, const std::string
& key
, bufferlist
& bl
);
3321 int omap_set(rgw_raw_obj
& obj
, map
<std::string
, bufferlist
>& m
);
3322 int omap_del(rgw_raw_obj
& obj
, const std::string
& key
);
3323 int update_containers_stats(map
<string
, RGWBucketEnt
>& m
);
3324 int append_async(rgw_raw_obj
& obj
, size_t size
, bufferlist
& bl
);
3326 int watch(const string
& oid
, uint64_t *watch_handle
, librados::WatchCtx2
*ctx
);
3327 int unwatch(uint64_t watch_handle
);
3328 void add_watcher(int i
);
3329 void remove_watcher(int i
);
3330 virtual bool need_watch_notify() { return false; }
3332 void finalize_watch();
3333 int distribute(const string
& key
, bufferlist
& bl
);
3334 virtual int watch_cb(uint64_t notify_id
,
3336 uint64_t notifier_id
,
3337 bufferlist
& bl
) { return 0; }
3338 void pick_control_oid(const string
& key
, string
& notify_oid
);
3340 virtual void set_cache_enabled(bool state
) {}
3342 void set_atomic(void *ctx
, rgw_obj
& obj
) {
3343 RGWObjectCtx
*rctx
= static_cast<RGWObjectCtx
*>(ctx
);
3344 rctx
->obj
.set_atomic(obj
);
3346 void set_prefetch_data(void *ctx
, rgw_obj
& obj
) {
3347 RGWObjectCtx
*rctx
= static_cast<RGWObjectCtx
*>(ctx
);
3348 rctx
->obj
.set_prefetch_data(obj
);
3350 void set_prefetch_data(void *ctx
, rgw_raw_obj
& obj
) {
3351 RGWObjectCtx
*rctx
= static_cast<RGWObjectCtx
*>(ctx
);
3352 rctx
->raw
.set_prefetch_data(obj
);
3355 int decode_policy(bufferlist
& bl
, ACLOwner
*owner
);
3356 int get_bucket_stats(RGWBucketInfo
& bucket_info
, int shard_id
, string
*bucket_ver
, string
*master_ver
,
3357 map
<RGWObjCategory
, RGWStorageStats
>& stats
, string
*max_marker
);
3358 int get_bucket_stats_async(RGWBucketInfo
& bucket_info
, int shard_id
, RGWGetBucketStats_CB
*cb
);
3359 int get_user_stats(const rgw_user
& user
, RGWStorageStats
& stats
);
3360 int get_user_stats_async(const rgw_user
& user
, RGWGetUserStats_CB
*cb
);
3361 void get_bucket_instance_obj(const rgw_bucket
& bucket
, rgw_raw_obj
& obj
);
3362 void get_bucket_meta_oid(const rgw_bucket
& bucket
, string
& oid
);
3364 int put_bucket_entrypoint_info(const string
& tenant_name
, const string
& bucket_name
, RGWBucketEntryPoint
& entry_point
,
3365 bool exclusive
, RGWObjVersionTracker
& objv_tracker
, ceph::real_time mtime
,
3366 map
<string
, bufferlist
> *pattrs
);
3367 int put_bucket_instance_info(RGWBucketInfo
& info
, bool exclusive
, ceph::real_time mtime
, map
<string
, bufferlist
> *pattrs
);
3368 int get_bucket_entrypoint_info(RGWObjectCtx
& obj_ctx
, const string
& tenant_name
, const string
& bucket_name
,
3369 RGWBucketEntryPoint
& entry_point
, RGWObjVersionTracker
*objv_tracker
,
3370 ceph::real_time
*pmtime
, map
<string
, bufferlist
> *pattrs
, rgw_cache_entry_info
*cache_info
= NULL
);
3371 int get_bucket_instance_info(RGWObjectCtx
& obj_ctx
, const string
& meta_key
, RGWBucketInfo
& info
, ceph::real_time
*pmtime
, map
<string
, bufferlist
> *pattrs
);
3372 int get_bucket_instance_info(RGWObjectCtx
& obj_ctx
, const rgw_bucket
& bucket
, RGWBucketInfo
& info
, ceph::real_time
*pmtime
, map
<string
, bufferlist
> *pattrs
);
3373 int get_bucket_instance_from_oid(RGWObjectCtx
& obj_ctx
, const string
& oid
, RGWBucketInfo
& info
, ceph::real_time
*pmtime
, map
<string
, bufferlist
> *pattrs
,
3374 rgw_cache_entry_info
*cache_info
= NULL
);
3376 int convert_old_bucket_info(RGWObjectCtx
& obj_ctx
, const string
& tenant_name
, const string
& bucket_name
);
3377 static void make_bucket_entry_name(const string
& tenant_name
, const string
& bucket_name
, string
& bucket_entry
);
3378 int get_bucket_info(RGWObjectCtx
& obj_ctx
,
3379 const string
& tenant_name
, const string
& bucket_name
,
3380 RGWBucketInfo
& info
,
3381 ceph::real_time
*pmtime
, map
<string
, bufferlist
> *pattrs
= NULL
);
3382 int put_linked_bucket_info(RGWBucketInfo
& info
, bool exclusive
, ceph::real_time mtime
, obj_version
*pep_objv
,
3383 map
<string
, bufferlist
> *pattrs
, bool create_entry_point
);
3385 int cls_rgw_init_index(librados::IoCtx
& io_ctx
, librados::ObjectWriteOperation
& op
, string
& oid
);
3386 int cls_obj_prepare_op(BucketShard
& bs
, RGWModifyOp op
, string
& tag
, rgw_obj
& obj
, uint16_t bilog_flags
, rgw_zone_set
*zones_trace
= nullptr);
3387 int cls_obj_complete_op(BucketShard
& bs
, const rgw_obj
& obj
, RGWModifyOp op
, string
& tag
, int64_t pool
, uint64_t epoch
,
3388 rgw_bucket_dir_entry
& ent
, RGWObjCategory category
, list
<rgw_obj_index_key
> *remove_objs
, uint16_t bilog_flags
, rgw_zone_set
*zones_trace
= nullptr);
3389 int cls_obj_complete_add(BucketShard
& bs
, const rgw_obj
& obj
, string
& tag
, int64_t pool
, uint64_t epoch
, rgw_bucket_dir_entry
& ent
,
3390 RGWObjCategory category
, list
<rgw_obj_index_key
> *remove_objs
, uint16_t bilog_flags
, rgw_zone_set
*zones_trace
= nullptr);
3391 int cls_obj_complete_del(BucketShard
& bs
, string
& tag
, int64_t pool
, uint64_t epoch
, rgw_obj
& obj
,
3392 ceph::real_time
& removed_mtime
, list
<rgw_obj_index_key
> *remove_objs
, uint16_t bilog_flags
, rgw_zone_set
*zones_trace
= nullptr);
3393 int cls_obj_complete_cancel(BucketShard
& bs
, string
& tag
, rgw_obj
& obj
, uint16_t bilog_flags
, rgw_zone_set
*zones_trace
= nullptr);
3394 int cls_obj_set_bucket_tag_timeout(RGWBucketInfo
& bucket_info
, uint64_t timeout
);
3395 int cls_bucket_list(RGWBucketInfo
& bucket_info
, int shard_id
, rgw_obj_index_key
& start
, const string
& prefix
,
3396 uint32_t num_entries
, bool list_versions
, map
<string
, rgw_bucket_dir_entry
>& m
,
3397 bool *is_truncated
, rgw_obj_index_key
*last_entry
,
3398 bool (*force_check_filter
)(const string
& name
) = NULL
);
3399 int cls_bucket_head(const RGWBucketInfo
& bucket_info
, int shard_id
, map
<string
, struct rgw_bucket_dir_header
>& headers
, map
<int, string
> *bucket_instance_ids
= NULL
);
3400 int cls_bucket_head_async(const RGWBucketInfo
& bucket_info
, int shard_id
, RGWGetDirHeader_CB
*ctx
, int *num_aio
);
3401 int list_bi_log_entries(RGWBucketInfo
& bucket_info
, int shard_id
, string
& marker
, uint32_t max
, std::list
<rgw_bi_log_entry
>& result
, bool *truncated
);
3402 int trim_bi_log_entries(RGWBucketInfo
& bucket_info
, int shard_id
, string
& marker
, string
& end_marker
);
3403 int get_bi_log_status(RGWBucketInfo
& bucket_info
, int shard_id
, map
<int, string
>& max_marker
);
3405 int bi_get_instance(const RGWBucketInfo
& bucket_info
, rgw_obj
& obj
, rgw_bucket_dir_entry
*dirent
);
3406 int bi_get(rgw_bucket
& bucket
, rgw_obj
& obj
, BIIndexType index_type
, rgw_cls_bi_entry
*entry
);
3407 void bi_put(librados::ObjectWriteOperation
& op
, BucketShard
& bs
, rgw_cls_bi_entry
& entry
);
3408 int bi_put(BucketShard
& bs
, rgw_cls_bi_entry
& entry
);
3409 int bi_put(rgw_bucket
& bucket
, rgw_obj
& obj
, rgw_cls_bi_entry
& entry
);
3410 int bi_list(rgw_bucket
& bucket
, int shard_id
, const string
& filter_obj
, const string
& marker
, uint32_t max
, list
<rgw_cls_bi_entry
> *entries
, bool *is_truncated
);
3411 int bi_list(BucketShard
& bs
, const string
& filter_obj
, const string
& marker
, uint32_t max
, list
<rgw_cls_bi_entry
> *entries
, bool *is_truncated
);
3412 int bi_list(rgw_bucket
& bucket
, const string
& obj_name
, const string
& marker
, uint32_t max
,
3413 list
<rgw_cls_bi_entry
> *entries
, bool *is_truncated
);
3414 int bi_remove(BucketShard
& bs
);
3416 int cls_obj_usage_log_add(const string
& oid
, rgw_usage_log_info
& info
);
3417 int cls_obj_usage_log_read(string
& oid
, string
& user
, uint64_t start_epoch
, uint64_t end_epoch
, uint32_t max_entries
,
3418 string
& read_iter
, map
<rgw_user_bucket
, rgw_usage_log_entry
>& usage
, bool *is_truncated
);
3419 int cls_obj_usage_log_trim(string
& oid
, string
& user
, uint64_t start_epoch
, uint64_t end_epoch
);
3421 int key_to_shard_id(const string
& key
, int max_shards
);
3422 void shard_name(const string
& prefix
, unsigned max_shards
, const string
& key
, string
& name
, int *shard_id
);
3423 void shard_name(const string
& prefix
, unsigned max_shards
, const string
& section
, const string
& key
, string
& name
);
3424 void shard_name(const string
& prefix
, unsigned shard_id
, string
& name
);
3425 int get_target_shard_id(const RGWBucketInfo
& bucket_info
, const string
& obj_key
, int *shard_id
);
3426 void time_log_prepare_entry(cls_log_entry
& entry
, const ceph::real_time
& ut
, const string
& section
, const string
& key
, bufferlist
& bl
);
3427 int time_log_add_init(librados::IoCtx
& io_ctx
);
3428 int time_log_add(const string
& oid
, list
<cls_log_entry
>& entries
,
3429 librados::AioCompletion
*completion
, bool monotonic_inc
= true);
3430 int time_log_add(const string
& oid
, const ceph::real_time
& ut
, const string
& section
, const string
& key
, bufferlist
& bl
);
3431 int time_log_list(const string
& oid
, const ceph::real_time
& start_time
, const ceph::real_time
& end_time
,
3432 int max_entries
, list
<cls_log_entry
>& entries
,
3433 const string
& marker
, string
*out_marker
, bool *truncated
);
3434 int time_log_info(const string
& oid
, cls_log_header
*header
);
3435 int time_log_info_async(librados::IoCtx
& io_ctx
, const string
& oid
, cls_log_header
*header
, librados::AioCompletion
*completion
);
3436 int time_log_trim(const string
& oid
, const ceph::real_time
& start_time
, const ceph::real_time
& end_time
,
3437 const string
& from_marker
, const string
& to_marker
,
3438 librados::AioCompletion
*completion
= nullptr);
3440 string
objexp_hint_get_shardname(int shard_num
);
3441 int objexp_key_shard(const rgw_obj_index_key
& key
);
3442 void objexp_get_shard(int shard_num
,
3443 string
& shard
); /* out */
3444 int objexp_hint_add(const ceph::real_time
& delete_at
,
3445 const string
& tenant_name
,
3446 const string
& bucket_name
,
3447 const string
& bucket_id
,
3448 const rgw_obj_index_key
& obj_key
);
3449 int objexp_hint_list(const string
& oid
,
3450 const ceph::real_time
& start_time
,
3451 const ceph::real_time
& end_time
,
3452 const int max_entries
,
3453 const string
& marker
,
3454 list
<cls_timeindex_entry
>& entries
, /* out */
3455 string
*out_marker
, /* out */
3456 bool *truncated
); /* out */
3457 int objexp_hint_parse(cls_timeindex_entry
&ti_entry
,
3458 objexp_hint_entry
& hint_entry
); /* out */
3459 int objexp_hint_trim(const string
& oid
,
3460 const ceph::real_time
& start_time
,
3461 const ceph::real_time
& end_time
,
3462 const string
& from_marker
= std::string(),
3463 const string
& to_marker
= std::string());
3465 int lock_exclusive(rgw_pool
& pool
, const string
& oid
, ceph::timespan
& duration
, string
& zone_id
, string
& owner_id
);
3466 int unlock(rgw_pool
& pool
, const string
& oid
, string
& zone_id
, string
& owner_id
);
3468 void update_gc_chain(rgw_obj
& head_obj
, RGWObjManifest
& manifest
, cls_rgw_obj_chain
*chain
);
3469 int send_chain_to_gc(cls_rgw_obj_chain
& chain
, const string
& tag
, bool sync
);
3470 int gc_operate(string
& oid
, librados::ObjectWriteOperation
*op
);
3471 int gc_aio_operate(string
& oid
, librados::ObjectWriteOperation
*op
);
3472 int gc_operate(string
& oid
, librados::ObjectReadOperation
*op
, bufferlist
*pbl
);
3474 int list_gc_objs(int *index
, string
& marker
, uint32_t max
, bool expired_only
, std::list
<cls_rgw_gc_obj_info
>& result
, bool *truncated
);
3476 int process_expire_objects();
3477 int defer_gc(void *ctx
, const RGWBucketInfo
& bucket_info
, const rgw_obj
& obj
);
3480 int list_lc_progress(const string
& marker
, uint32_t max_entries
, map
<string
, int> *progress_map
);
3482 int bucket_check_index(RGWBucketInfo
& bucket_info
,
3483 map
<RGWObjCategory
, RGWStorageStats
> *existing_stats
,
3484 map
<RGWObjCategory
, RGWStorageStats
> *calculated_stats
);
3485 int bucket_rebuild_index(RGWBucketInfo
& bucket_info
);
3486 int bucket_set_reshard(RGWBucketInfo
& bucket_info
, const cls_rgw_bucket_instance_entry
& entry
);
3487 int remove_objs_from_index(RGWBucketInfo
& bucket_info
, list
<rgw_obj_index_key
>& oid_list
);
3488 int move_rados_obj(librados::IoCtx
& src_ioctx
,
3489 const string
& src_oid
, const string
& src_locator
,
3490 librados::IoCtx
& dst_ioctx
,
3491 const string
& dst_oid
, const string
& dst_locator
);
3492 int fix_head_obj_locator(const RGWBucketInfo
& bucket_info
, bool copy_obj
, bool remove_bad
, rgw_obj_key
& key
);
3493 int fix_tail_obj_locator(const RGWBucketInfo
& bucket_info
, rgw_obj_key
& key
, bool fix
, bool *need_fix
);
3495 int cls_user_get_header(const string
& user_id
, cls_user_header
*header
);
3496 int cls_user_get_header_async(const string
& user_id
, RGWGetUserHeader_CB
*ctx
);
3497 int cls_user_sync_bucket_stats(rgw_raw_obj
& user_obj
, const RGWBucketInfo
& bucket_info
);
3498 int cls_user_list_buckets(rgw_raw_obj
& obj
,
3499 const string
& in_marker
,
3500 const string
& end_marker
,
3502 list
<cls_user_bucket_entry
>& entries
,
3505 int cls_user_add_bucket(rgw_raw_obj
& obj
, const cls_user_bucket_entry
& entry
);
3506 int cls_user_update_buckets(rgw_raw_obj
& obj
, list
<cls_user_bucket_entry
>& entries
, bool add
);
3507 int cls_user_complete_stats_sync(rgw_raw_obj
& obj
);
3508 int complete_sync_user_stats(const rgw_user
& user_id
);
3509 int cls_user_add_bucket(rgw_raw_obj
& obj
, list
<cls_user_bucket_entry
>& entries
);
3510 int cls_user_remove_bucket(rgw_raw_obj
& obj
, const cls_user_bucket
& bucket
);
3512 int check_quota(const rgw_user
& bucket_owner
, rgw_bucket
& bucket
,
3513 RGWQuotaInfo
& user_quota
, RGWQuotaInfo
& bucket_quota
, uint64_t obj_size
);
3515 int check_bucket_shards(const RGWBucketInfo
& bucket_info
, rgw_bucket
& bucket
,
3516 RGWQuotaInfo
& bucket_quota
);
3518 int add_bucket_to_reshard(const RGWBucketInfo
& bucket_info
, uint32_t new_num_shards
);
3520 uint64_t instance_id();
3521 const string
& zone_id() {
3522 return get_zone_params().get_id();
3524 string
unique_id(uint64_t unique_num
) {
3526 snprintf(buf
, sizeof(buf
), ".%llu.%llu", (unsigned long long)instance_id(), (unsigned long long)unique_num
);
3527 string s
= get_zone_params().get_id() + buf
;
3531 void init_unique_trans_id_deps() {
3532 char buf
[16 + 2 + 1]; /* uint64_t needs 16, 2 hyphens add further 2 */
3534 snprintf(buf
, sizeof(buf
), "-%llx-", (unsigned long long)instance_id());
3535 url_encode(string(buf
) + get_zone_params().get_name(), trans_id_suffix
);
3538 /* In order to preserve compability with Swift API, transaction ID
3539 * should contain at least 32 characters satisfying following spec:
3540 * - first 21 chars must be in range [0-9a-f]. Swift uses this
3541 * space for storing fragment of UUID obtained through a call to
3542 * uuid4() function of Python's uuid module;
3543 * - char no. 22 must be a hyphen;
3544 * - at least 10 next characters constitute hex-formatted timestamp
3545 * padded with zeroes if necessary. All bytes must be in [0-9a-f]
3547 * - last, optional part of transaction ID is any url-encoded string
3548 * without restriction on length. */
3549 string
unique_trans_id(const uint64_t unique_num
) {
3550 char buf
[41]; /* 2 + 21 + 1 + 16 (timestamp can consume up to 16) + 1 */
3551 time_t timestamp
= time(NULL
);
3553 snprintf(buf
, sizeof(buf
), "tx%021llx-%010llx",
3554 (unsigned long long)unique_num
,
3555 (unsigned long long)timestamp
);
3557 return string(buf
) + trans_id_suffix
;
3560 void get_log_pool(rgw_pool
& pool
) {
3561 pool
= get_zone_params().log_pool
;
3564 bool need_to_log_data() {
3565 return get_zone().log_data
;
3568 bool need_to_log_metadata() {
3569 return is_meta_master() && get_zone().log_meta
;
3572 librados::Rados
* get_rados_handle();
3574 int delete_raw_obj_aio(const rgw_raw_obj
& obj
, list
<librados::AioCompletion
*>& handles
);
3575 int delete_obj_aio(const rgw_obj
& obj
, RGWBucketInfo
& info
, RGWObjState
*astate
,
3576 list
<librados::AioCompletion
*>& handles
, bool keep_index_consistent
);
3579 * This is a helper method, it generates a list of bucket index objects with the given
3580 * bucket base oid and number of shards.
3582 * bucket_oid_base [in] - base name of the bucket index object;
3583 * num_shards [in] - number of bucket index object shards.
3584 * bucket_objs [out] - filled by this method, a list of bucket index objects.
3586 void get_bucket_index_objects(const string
& bucket_oid_base
, uint32_t num_shards
,
3587 map
<int, string
>& bucket_objs
, int shard_id
= -1);
3590 * Get the bucket index object with the given base bucket index object and object key,
3591 * and the number of bucket index shards.
3593 * bucket_oid_base [in] - bucket object base name.
3594 * obj_key [in] - object key.
3595 * num_shards [in] - number of bucket index shards.
3596 * hash_type [in] - type of hash to find the shard ID.
3597 * bucket_obj [out] - the bucket index object for the given object.
3599 * Return 0 on success, a failure code otherwise.
3601 int get_bucket_index_object(const string
& bucket_oid_base
, const string
& obj_key
,
3602 uint32_t num_shards
, RGWBucketInfo::BIShardsHashType hash_type
, string
*bucket_obj
, int *shard
);
3604 void get_bucket_index_object(const string
& bucket_oid_base
, uint32_t num_shards
,
3605 int shard_id
, string
*bucket_obj
);
3608 * Check the actual on-disk state of the object specified
3609 * by list_state, and fill in the time and size of object.
3610 * Then append any changes to suggested_updates for
3611 * the rgw class' dir_suggest_changes function.
3613 * Note that this can maul list_state; don't use it afterwards. Also
3614 * it expects object to already be filled in from list_state; it only
3615 * sets the size and mtime.
3617 * Returns 0 on success, -ENOENT if the object doesn't exist on disk,
3618 * and -errno on other failures. (-ENOENT is not a failure, and it
3619 * will encode that info as a suggested update.)
3621 int check_disk_state(librados::IoCtx io_ctx
,
3622 const RGWBucketInfo
& bucket_info
,
3623 rgw_bucket_dir_entry
& list_state
,
3624 rgw_bucket_dir_entry
& object
,
3625 bufferlist
& suggested_updates
);
3628 * Init pool iteration
3629 * pool: pool to use for the ctx initialization
3630 * ctx: context object to use for the iteration
3631 * Returns: 0 on success, -ERR# otherwise.
3633 int pool_iterate_begin(const rgw_pool
& pool
, RGWPoolIterCtx
& ctx
);
3636 * Iterate over pool return object names, use optional filter
3637 * ctx: iteration context, initialized with pool_iterate_begin()
3638 * num: max number of objects to return
3639 * objs: a vector that the results will append into
3640 * is_truncated: if not NULL, will hold true iff iteration is complete
3641 * filter: if not NULL, will be used to filter returned objects
3642 * Returns: 0 on success, -ERR# otherwise.
3644 int pool_iterate(RGWPoolIterCtx
& ctx
, uint32_t num
, vector
<rgw_bucket_dir_entry
>& objs
,
3645 bool *is_truncated
, RGWAccessListFilter
*filter
);
3647 uint64_t next_bucket_id();
3650 class RGWStoreManager
{
3652 RGWStoreManager() {}
3653 static RGWRados
*get_storage(CephContext
*cct
, bool use_gc_thread
, bool use_lc_thread
, bool quota_threads
, bool run_sync_thread
, bool run_reshard_thread
) {
3654 RGWRados
*store
= init_storage_provider(cct
, use_gc_thread
, use_lc_thread
, quota_threads
, run_sync_thread
,
3655 run_reshard_thread
);
3658 static RGWRados
*get_raw_storage(CephContext
*cct
) {
3659 RGWRados
*store
= init_raw_storage_provider(cct
);
3662 static RGWRados
*init_storage_provider(CephContext
*cct
, bool use_gc_thread
, bool use_lc_thread
, bool quota_threads
, bool run_sync_thread
, bool run_reshard_thread
);
3663 static RGWRados
*init_raw_storage_provider(CephContext
*cct
);
3664 static void close_storage(RGWRados
*store
);
3669 class RGWChainedCacheImpl
: public RGWChainedCache
{
3672 map
<string
, T
> entries
;
3675 RGWChainedCacheImpl() : lock("RGWChainedCacheImpl::lock") {}
3677 void init(RGWRados
*store
) {
3678 store
->register_chained_cache(this);
3681 bool find(const string
& key
, T
*entry
) {
3682 RWLock::RLocker
rl(lock
);
3683 typename map
<string
, T
>::iterator iter
= entries
.find(key
);
3684 if (iter
== entries
.end()) {
3688 *entry
= iter
->second
;
3692 bool put(RGWRados
*store
, const string
& key
, T
*entry
, list
<rgw_cache_entry_info
*>& cache_info_entries
) {
3693 Entry
chain_entry(this, key
, entry
);
3695 /* we need the store cache to call us under its lock to maintain lock ordering */
3696 return store
->chain_cache_entry(cache_info_entries
, &chain_entry
);
3699 void chain_cb(const string
& key
, void *data
) override
{
3700 T
*entry
= static_cast<T
*>(data
);
3701 RWLock::WLocker
wl(lock
);
3702 entries
[key
] = *entry
;
3705 void invalidate(const string
& key
) override
{
3706 RWLock::WLocker
wl(lock
);
3710 void invalidate_all() override
{
3711 RWLock::WLocker
wl(lock
);
3714 }; /* RGWChainedCacheImpl */
3717 * Base of PUT operation.
3718 * Allow to create chained data transformers like compresors and encryptors.
3720 class RGWPutObjDataProcessor
3723 RGWPutObjDataProcessor(){}
3724 virtual ~RGWPutObjDataProcessor(){}
3725 virtual int handle_data(bufferlist
& bl
, off_t ofs
, void **phandle
, rgw_raw_obj
*pobj
, bool *again
) = 0;
3726 virtual int throttle_data(void *handle
, const rgw_raw_obj
& obj
, uint64_t size
, bool need_to_wait
) = 0;
3727 }; /* RGWPutObjDataProcessor */
3730 class RGWPutObjProcessor
: public RGWPutObjDataProcessor
3734 RGWObjectCtx
& obj_ctx
;
3736 RGWBucketInfo bucket_info
;
3739 virtual int do_complete(size_t accounted_size
, const string
& etag
,
3740 ceph::real_time
*mtime
, ceph::real_time set_mtime
,
3741 map
<string
, bufferlist
>& attrs
, ceph::real_time delete_at
,
3742 const char *if_match
, const char *if_nomatch
, const string
*user_data
,
3743 rgw_zone_set
* zones_trace
= nullptr) = 0;
3746 RGWPutObjProcessor(RGWObjectCtx
& _obj_ctx
, RGWBucketInfo
& _bi
) : store(NULL
),
3751 ~RGWPutObjProcessor() override
{}
3752 virtual int prepare(RGWRados
*_store
, string
*oid_rand
) {
3757 int complete(size_t accounted_size
, const string
& etag
,
3758 ceph::real_time
*mtime
, ceph::real_time set_mtime
,
3759 map
<string
, bufferlist
>& attrs
, ceph::real_time delete_at
,
3760 const char *if_match
= NULL
, const char *if_nomatch
= NULL
, const string
*user_data
= nullptr,
3761 rgw_zone_set
*zones_trace
= nullptr);
3765 bool is_canceled() { return canceled
; }
3766 }; /* RGWPutObjProcessor */
3768 struct put_obj_aio_info
{
3774 #define RGW_PUT_OBJ_MIN_WINDOW_SIZE_DEFAULT (16 * 1024 * 1024)
3776 class RGWPutObjProcessor_Aio
: public RGWPutObjProcessor
3778 list
<struct put_obj_aio_info
> pending
;
3779 uint64_t window_size
{RGW_PUT_OBJ_MIN_WINDOW_SIZE_DEFAULT
};
3780 uint64_t pending_size
{0};
3782 struct put_obj_aio_info
pop_pending();
3783 int wait_pending_front();
3784 bool pending_has_completed();
3786 rgw_raw_obj last_written_obj
;
3789 uint64_t obj_len
{0};
3791 set
<rgw_raw_obj
> written_objs
;
3794 void add_written_obj(const rgw_raw_obj
& obj
) {
3795 written_objs
.insert(obj
);
3798 int drain_pending();
3799 int handle_obj_data(rgw_raw_obj
& obj
, bufferlist
& bl
, off_t ofs
, off_t abs_ofs
, void **phandle
, bool exclusive
);
3802 int prepare(RGWRados
*store
, string
*oid_rand
) override
;
3803 int throttle_data(void *handle
, const rgw_raw_obj
& obj
, uint64_t size
, bool need_to_wait
) override
;
3805 RGWPutObjProcessor_Aio(RGWObjectCtx
& obj_ctx
, RGWBucketInfo
& bucket_info
) : RGWPutObjProcessor(obj_ctx
, bucket_info
) {}
3806 ~RGWPutObjProcessor_Aio() override
;
3807 }; /* RGWPutObjProcessor_Aio */
3809 class RGWPutObjProcessor_Atomic
: public RGWPutObjProcessor_Aio
3811 bufferlist first_chunk
;
3814 off_t next_part_ofs
;
3818 bufferlist pending_data_bl
;
3819 uint64_t max_chunk_size
;
3821 bool versioned_object
;
3831 rgw_raw_obj cur_obj
;
3832 RGWObjManifest manifest
;
3833 RGWObjManifest::generator manifest_gen
;
3835 int write_data(bufferlist
& bl
, off_t ofs
, void **phandle
, rgw_raw_obj
*pobj
, bool exclusive
);
3836 int do_complete(size_t accounted_size
, const string
& etag
,
3837 ceph::real_time
*mtime
, ceph::real_time set_mtime
,
3838 map
<string
, bufferlist
>& attrs
, ceph::real_time delete_at
,
3839 const char *if_match
, const char *if_nomatch
, const string
*user_data
, rgw_zone_set
*zones_trace
) override
;
3841 int prepare_next_part(off_t ofs
);
3842 int complete_parts();
3843 int complete_writing_data();
3845 int prepare_init(RGWRados
*store
, string
*oid_rand
);
3848 ~RGWPutObjProcessor_Atomic() override
{}
3849 RGWPutObjProcessor_Atomic(RGWObjectCtx
& obj_ctx
, RGWBucketInfo
& bucket_info
,
3850 rgw_bucket
& _b
, const string
& _o
, uint64_t _p
, const string
& _t
, bool versioned
) :
3851 RGWPutObjProcessor_Aio(obj_ctx
, bucket_info
),
3858 versioned_object(versioned
),
3863 int prepare(RGWRados
*store
, string
*oid_rand
) override
;
3864 virtual bool immutable_head() { return false; }
3865 int handle_data(bufferlist
& bl
, off_t ofs
, void **phandle
, rgw_raw_obj
*pobj
, bool *again
) override
;
3867 void set_olh_epoch(uint64_t epoch
) {
3871 void set_version_id(const string
& vid
) {
3874 }; /* RGWPutObjProcessor_Atomic */
3876 #define MP_META_SUFFIX ".meta"
3885 RGWMPObj(const string
& _oid
, const string
& _upload_id
) {
3886 init(_oid
, _upload_id
, _upload_id
);
3888 void init(const string
& _oid
, const string
& _upload_id
) {
3889 init(_oid
, _upload_id
, _upload_id
);
3891 void init(const string
& _oid
, const string
& _upload_id
, const string
& part_unique_str
) {
3897 upload_id
= _upload_id
;
3899 meta
= prefix
+ upload_id
+ MP_META_SUFFIX
;
3900 prefix
.append(part_unique_str
);
3902 string
& get_meta() { return meta
; }
3903 string
get_part(int num
) {
3905 snprintf(buf
, 16, ".%d", num
);
3910 string
get_part(string
& part
) {
3916 string
& get_upload_id() {
3922 bool from_meta(string
& meta
) {
3923 int end_pos
= meta
.rfind('.'); // search for ".meta"
3926 int mid_pos
= meta
.rfind('.', end_pos
- 1); // <key>.<upload_id>
3929 oid
= meta
.substr(0, mid_pos
);
3930 upload_id
= meta
.substr(mid_pos
+ 1, end_pos
- mid_pos
- 1);
3931 init(oid
, upload_id
, upload_id
);
3942 class RGWPutObjProcessor_Multipart
: public RGWPutObjProcessor_Atomic
3950 int prepare(RGWRados
*store
, string
*oid_rand
);
3951 int do_complete(size_t accounted_size
, const string
& etag
,
3952 ceph::real_time
*mtime
, ceph::real_time set_mtime
,
3953 map
<string
, bufferlist
>& attrs
, ceph::real_time delete_at
,
3954 const char *if_match
, const char *if_nomatch
, const string
*user_data
,
3955 rgw_zone_set
*zones_trace
) override
;
3957 bool immutable_head() { return true; }
3958 RGWPutObjProcessor_Multipart(RGWObjectCtx
& obj_ctx
, RGWBucketInfo
& bucket_info
, uint64_t _p
, req_state
*_s
) :
3959 RGWPutObjProcessor_Atomic(obj_ctx
, bucket_info
, _s
->bucket
, _s
->object
.name
, _p
, _s
->req_id
, false), s(_s
) {}
3960 void get_mp(RGWMPObj
** _mp
);
3961 }; /* RGWPutObjProcessor_Multipart */