]> git.proxmox.com Git - ceph.git/blame - ceph/src/mds/CDir.h
update sources to 12.2.7
[ceph.git] / ceph / src / mds / CDir.h
CommitLineData
7c673cae
FG
1// -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
2// vim: ts=8 sw=2 smarttab
3/*
4 * Ceph - scalable distributed file system
5 *
6 * Copyright (C) 2004-2006 Sage Weil <sage@newdream.net>
7 *
8 * This is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License version 2.1, as published by the Free Software
11 * Foundation. See file COPYING.
12 *
13 */
14
15
16
17#ifndef CEPH_CDIR_H
18#define CEPH_CDIR_H
19
7c673cae 20#include <iosfwd>
7c673cae 21#include <list>
7c673cae 22#include <map>
94b18763 23#include <set>
7c673cae 24#include <string>
94b18763 25#include <boost/utility/string_view.hpp>
7c673cae 26
94b18763
FG
27#include "common/DecayCounter.h"
28#include "common/bloom_filter.hpp"
29#include "common/config.h"
30#include "include/buffer_fwd.h"
31#include "include/counter.h"
32#include "include/types.h"
7c673cae
FG
33
34#include "CInode.h"
94b18763 35#include "MDSCacheObject.h"
7c673cae
FG
36
37class CDentry;
38class MDCache;
39
40struct ObjectOperation;
41
42ostream& operator<<(ostream& out, const class CDir& dir);
43class CDir : public MDSCacheObject, public Counter<CDir> {
44 friend ostream& operator<<(ostream& out, const class CDir& dir);
45
46public:
181888fb 47 MEMPOOL_CLASS_HELPERS();
7c673cae
FG
48 // -- pins --
49 static const int PIN_DNWAITER = 1;
50 static const int PIN_INOWAITER = 2;
51 static const int PIN_CHILD = 3;
52 static const int PIN_FROZEN = 4;
53 static const int PIN_SUBTREE = 5;
54 static const int PIN_IMPORTING = 7;
55 static const int PIN_IMPORTBOUND = 9;
56 static const int PIN_EXPORTBOUND = 10;
57 static const int PIN_STICKY = 11;
58 static const int PIN_SUBTREETEMP = 12; // used by MDCache::trim_non_auth()
59 const char *pin_name(int p) const override {
60 switch (p) {
61 case PIN_DNWAITER: return "dnwaiter";
62 case PIN_INOWAITER: return "inowaiter";
63 case PIN_CHILD: return "child";
64 case PIN_FROZEN: return "frozen";
65 case PIN_SUBTREE: return "subtree";
66 case PIN_IMPORTING: return "importing";
67 case PIN_IMPORTBOUND: return "importbound";
68 case PIN_EXPORTBOUND: return "exportbound";
69 case PIN_STICKY: return "sticky";
70 case PIN_SUBTREETEMP: return "subtreetemp";
71 default: return generic_pin_name(p);
72 }
73 }
74
75 // -- state --
76 static const unsigned STATE_COMPLETE = (1<< 1); // the complete contents are in cache
77 static const unsigned STATE_FROZENTREE = (1<< 2); // root of tree (bounded by exports)
78 static const unsigned STATE_FREEZINGTREE = (1<< 3); // in process of freezing
79 static const unsigned STATE_FROZENDIR = (1<< 4);
80 static const unsigned STATE_FREEZINGDIR = (1<< 5);
81 static const unsigned STATE_COMMITTING = (1<< 6); // mid-commit
82 static const unsigned STATE_FETCHING = (1<< 7); // currenting fetching
83 static const unsigned STATE_CREATING = (1<< 8);
84 static const unsigned STATE_IMPORTBOUND = (1<<10);
85 static const unsigned STATE_EXPORTBOUND = (1<<11);
86 static const unsigned STATE_EXPORTING = (1<<12);
87 static const unsigned STATE_IMPORTING = (1<<13);
88 static const unsigned STATE_FRAGMENTING = (1<<14);
89 static const unsigned STATE_STICKY = (1<<15); // sticky pin due to inode stickydirs
90 static const unsigned STATE_DNPINNEDFRAG = (1<<16); // dir is refragmenting
91 static const unsigned STATE_ASSIMRSTAT = (1<<17); // assimilating inode->frag rstats
92 static const unsigned STATE_DIRTYDFT = (1<<18); // dirty dirfragtree
93 static const unsigned STATE_BADFRAG = (1<<19); // bad dirfrag
94 static const unsigned STATE_AUXSUBTREE = (1<<20); // no subtree merge
95
96 // common states
97 static const unsigned STATE_CLEAN = 0;
7c673cae
FG
98
99 // these state bits are preserved by an import/export
100 // ...except if the directory is hashed, in which case none of them are!
101 static const unsigned MASK_STATE_EXPORTED =
102 (STATE_COMPLETE|STATE_DIRTY|STATE_DIRTYDFT|STATE_BADFRAG);
103 static const unsigned MASK_STATE_IMPORT_KEPT =
104 (
105 STATE_IMPORTING
106 |STATE_IMPORTBOUND|STATE_EXPORTBOUND
107 |STATE_FROZENTREE
108 |STATE_STICKY);
109 static const unsigned MASK_STATE_EXPORT_KEPT =
110 (STATE_EXPORTING
111 |STATE_IMPORTBOUND|STATE_EXPORTBOUND
112 |STATE_FROZENTREE
113 |STATE_FROZENDIR
114 |STATE_STICKY);
115 static const unsigned MASK_STATE_FRAGMENT_KEPT =
116 (STATE_DIRTY|
117 STATE_EXPORTBOUND |
118 STATE_IMPORTBOUND |
119 STATE_AUXSUBTREE |
120 STATE_REJOINUNDEF);
121
122 // -- rep spec --
123 static const int REP_NONE = 0;
124 static const int REP_ALL = 1;
125 static const int REP_LIST = 2;
126
127
128 static const unsigned EXPORT_NONCE = 1;
129
130
131 // -- wait masks --
132 static const uint64_t WAIT_DENTRY = (1<<0); // wait for item to be in cache
133 static const uint64_t WAIT_COMPLETE = (1<<1); // wait for complete dir contents
134 static const uint64_t WAIT_FROZEN = (1<<2); // auth pins removed
135 static const uint64_t WAIT_CREATED = (1<<3); // new dirfrag is logged
136
137 static const int WAIT_DNLOCK_OFFSET = 4;
138
139 static const uint64_t WAIT_ANY_MASK = (uint64_t)(-1);
140 static const uint64_t WAIT_ATFREEZEROOT = (WAIT_UNFREEZE);
141 static const uint64_t WAIT_ATSUBTREEROOT = (WAIT_SINGLEAUTH);
142
143
144
145
146 public:
147 // context
148 MDCache *cache;
149
150 CInode *inode; // my inode
151 frag_t frag; // my frag
152
153 bool is_lt(const MDSCacheObject *r) const override {
154 return dirfrag() < (static_cast<const CDir*>(r))->dirfrag();
155 }
156
157 fnode_t fnode;
158 snapid_t first;
94b18763 159 mempool::mds_co::compact_map<snapid_t,old_rstat_t> dirty_old_rstat; // [value.first,key]
7c673cae
FG
160
161 // my inodes with dirty rstat data
162 elist<CInode*> dirty_rstat_inodes;
163
164 void resync_accounted_fragstat();
165 void resync_accounted_rstat();
166 void assimilate_dirty_rstat_inodes();
167 void assimilate_dirty_rstat_inodes_finish(MutationRef& mut, EMetaBlob *blob);
168
169protected:
170 version_t projected_version;
94b18763 171 mempool::mds_co::list<fnode_t> projected_fnode;
7c673cae
FG
172
173public:
b32b8144 174 elist<CDentry*> dirty_dentries;
7c673cae
FG
175 elist<CDir*>::item item_dirty, item_new;
176
7c673cae
FG
177public:
178 version_t get_version() const { return fnode.version; }
179 void set_version(version_t v) {
180 assert(projected_fnode.empty());
181 projected_version = fnode.version = v;
182 }
183 version_t get_projected_version() const { return projected_version; }
184
185 const fnode_t *get_projected_fnode() const {
186 if (projected_fnode.empty())
187 return &fnode;
188 else
94b18763 189 return &projected_fnode.back();
7c673cae
FG
190 }
191
192 fnode_t *get_projected_fnode() {
193 if (projected_fnode.empty())
194 return &fnode;
195 else
94b18763 196 return &projected_fnode.back();
7c673cae
FG
197 }
198 fnode_t *project_fnode();
199
200 void pop_and_dirty_projected_fnode(LogSegment *ls);
201 bool is_projected() const { return !projected_fnode.empty(); }
202 version_t pre_dirty(version_t min=0);
203 void _mark_dirty(LogSegment *ls);
204 void _set_dirty_flag() {
205 if (!state_test(STATE_DIRTY)) {
206 state_set(STATE_DIRTY);
207 get(PIN_DIRTY);
208 }
209 }
210 void mark_dirty(version_t pv, LogSegment *ls);
211 void mark_clean();
212
213 bool is_new() { return item_new.is_on_list(); }
214 void mark_new(LogSegment *ls);
215
216 bool is_bad() { return state_test(STATE_BADFRAG); }
217private:
218 void log_mark_dirty();
219
220public:
94b18763
FG
221 typedef mempool::mds_co::map<dentry_key_t, CDentry*> dentry_key_map;
222 typedef mempool::mds_co::set<dentry_key_t> dentry_key_set;
7c673cae
FG
223
224 class scrub_info_t {
225 public:
226 /// inodes we contain with dirty scrub stamps
94b18763 227 dentry_key_map dirty_scrub_stamps; // TODO: make use of this!
7c673cae
FG
228 struct scrub_stamps {
229 version_t version;
230 utime_t time;
231 scrub_stamps() : version(0) {}
232 void operator=(const scrub_stamps &o) {
233 version = o.version;
234 time = o.time;
235 }
236 };
237
238 scrub_stamps recursive_start; // when we last started a recursive scrub
239 scrub_stamps last_recursive; // when we last finished a recursive scrub
240 scrub_stamps last_local; // when we last did a local scrub
241
242 bool directory_scrubbing; /// safety check
243 bool need_scrub_local;
244 bool last_scrub_dirty; /// is scrub info dirty or is it flushed to fnode?
245 bool pending_scrub_error;
246
247 /// these are lists of children in each stage of scrubbing
94b18763
FG
248 dentry_key_set directories_to_scrub;
249 dentry_key_set directories_scrubbing;
250 dentry_key_set directories_scrubbed;
251 dentry_key_set others_to_scrub;
252 dentry_key_set others_scrubbing;
253 dentry_key_set others_scrubbed;
7c673cae
FG
254
255 ScrubHeaderRefConst header;
256
257 scrub_info_t() :
258 directory_scrubbing(false),
259 need_scrub_local(false),
260 last_scrub_dirty(false),
261 pending_scrub_error(false) {}
262 };
263 /**
264 * Call to start this CDir on a new scrub.
265 * @pre It is not currently scrubbing
266 * @pre The CDir is marked complete.
267 * @post It has set up its internal scrubbing state.
268 */
269 void scrub_initialize(const ScrubHeaderRefConst& header);
270 /**
271 * Get the next dentry to scrub. Gives you a CDentry* and its meaning. This
272 * function will give you all directory-representing dentries before any
273 * others.
274 * 0: success, you should scrub this CDentry right now
275 * EAGAIN: is currently fetching the next CDentry into memory for you.
276 * It will activate your callback when done; try again when it does!
277 * ENOENT: there are no remaining dentries to scrub
278 * <0: There was an unexpected error
279 *
280 * @param cb An MDSInternalContext which will be activated only if
281 * we return EAGAIN via rcode, or else ignored
282 * @param dnout CDentry * which you should next scrub, or NULL
283 * @returns a value as described above
284 */
285 int scrub_dentry_next(MDSInternalContext *cb, CDentry **dnout);
286 /**
287 * Get the currently scrubbing dentries. When returned, the passed-in
288 * list will be filled with all CDentry * which have been returned
289 * from scrub_dentry_next() but not sent back via scrub_dentry_finished().
290 */
94b18763 291 void scrub_dentries_scrubbing(std::list<CDentry*> *out_dentries);
7c673cae
FG
292 /**
293 * Report to the CDir that a CDentry has been scrubbed. Call this
294 * for every CDentry returned from scrub_dentry_next().
295 * @param dn The CDentry which has been scrubbed.
296 */
297 void scrub_dentry_finished(CDentry *dn);
298 /**
299 * Call this once all CDentries have been scrubbed, according to
300 * scrub_dentry_next's listing. It finalizes the scrub statistics.
301 */
302 void scrub_finished();
303 /**
304 * Tell the CDir to do a local scrub of itself.
305 * @pre The CDir is_complete().
306 * @returns true if the rstats and directory contents match, false otherwise.
307 */
308 bool scrub_local();
309private:
310 /**
311 * Create a scrub_info_t struct for the scrub_infop pointer.
312 */
313 void scrub_info_create() const;
314 /**
315 * Delete the scrub_infop if it's not got any useful data.
316 */
317 void scrub_maybe_delete_info();
318 /**
319 * Check the given set (presumably one of those in scrub_info_t) for the
320 * next key to scrub and look it up (or fail!).
321 */
94b18763 322 int _next_dentry_on_set(dentry_key_set &dns, bool missing_okay,
7c673cae
FG
323 MDSInternalContext *cb, CDentry **dnout);
324
325
326protected:
94b18763 327 std::unique_ptr<scrub_info_t> scrub_infop; // FIXME not in mempool
7c673cae
FG
328
329 // contents of this directory
94b18763 330 dentry_key_map items; // non-null AND null
7c673cae
FG
331 unsigned num_head_items;
332 unsigned num_head_null;
333 unsigned num_snap_items;
334 unsigned num_snap_null;
335
336 int num_dirty;
337
338 // state
339 version_t committing_version;
340 version_t committed_version;
341
94b18763 342 mempool::mds_co::compact_set<mempool::mds_co::string> stale_items;
7c673cae
FG
343
344 // lock nesting, freeze
345 static int num_frozen_trees;
346 static int num_freezing_trees;
347
348 int dir_auth_pins;
349 int request_pins;
350
351 // cache control (defined for authority; hints for replicas)
352 __s32 dir_rep;
94b18763 353 mempool::mds_co::compact_set<__s32> dir_rep_by; // if dir_rep == REP_LIST
7c673cae
FG
354
355 // popularity
356 dirfrag_load_vec_t pop_me;
357 dirfrag_load_vec_t pop_nested;
358 dirfrag_load_vec_t pop_auth_subtree;
359 dirfrag_load_vec_t pop_auth_subtree_nested;
360
361 utime_t last_popularity_sample;
362
363 load_spread_t pop_spread;
364
28e407b8
AA
365 elist<CInode*> pop_lru_subdirs;
366
7c673cae
FG
367 // and to provide density
368 int num_dentries_nested;
369 int num_dentries_auth_subtree;
370 int num_dentries_auth_subtree_nested;
371
372
373 // friends
374 friend class Migrator;
375 friend class CInode;
376 friend class MDCache;
377 friend class MDiscover;
378 friend class MDBalancer;
379
380 friend class CDirDiscover;
381 friend class CDirExport;
382 friend class C_IO_Dir_TMAP_Fetched;
383 friend class C_IO_Dir_OMAP_Fetched;
384 friend class C_IO_Dir_OMAP_FetchedMore;
385 friend class C_IO_Dir_Committed;
386
94b18763 387 std::unique_ptr<bloom_filter> bloom; // XXX not part of mempool::mds_co
7c673cae
FG
388 /* If you set up the bloom filter, you must keep it accurate!
389 * It's deleted when you mark_complete() and is deliberately not serialized.*/
390
391 public:
392 CDir(CInode *in, frag_t fg, MDCache *mdcache, bool auth);
393
394 const scrub_info_t *scrub_info() const {
395 if (!scrub_infop) {
396 scrub_info_create();
397 }
398 return scrub_infop.get();
399 }
400
401
402 // -- accessors --
403 inodeno_t ino() const { return inode->ino(); } // deprecate me?
404 frag_t get_frag() const { return frag; }
405 dirfrag_t dirfrag() const { return dirfrag_t(inode->ino(), frag); }
406
407 CInode *get_inode() { return inode; }
408 const CInode *get_inode() const { return inode; }
409 CDir *get_parent_dir() { return inode->get_parent_dir(); }
410
94b18763
FG
411 dentry_key_map::iterator begin() { return items.begin(); }
412 dentry_key_map::iterator end() { return items.end(); }
413 dentry_key_map::iterator lower_bound(dentry_key_t key) { return items.lower_bound(key); }
7c673cae
FG
414
415 unsigned get_num_head_items() const { return num_head_items; }
416 unsigned get_num_head_null() const { return num_head_null; }
417 unsigned get_num_snap_items() const { return num_snap_items; }
418 unsigned get_num_snap_null() const { return num_snap_null; }
419 unsigned get_num_any() const { return num_head_items + num_head_null + num_snap_items + num_snap_null; }
420
421 bool check_rstats(bool scrub=false);
422
423 void inc_num_dirty() { num_dirty++; }
424 void dec_num_dirty() {
425 assert(num_dirty > 0);
426 num_dirty--;
427 }
428 int get_num_dirty() const {
429 return num_dirty;
430 }
431
432 int64_t get_frag_size() const {
433 return get_projected_fnode()->fragstat.size();
434 }
435
436 // -- dentries and inodes --
437 public:
94b18763
FG
438 CDentry* lookup_exact_snap(boost::string_view dname, snapid_t last);
439 CDentry* lookup(boost::string_view n, snapid_t snap=CEPH_NOSNAP);
7c673cae 440 CDentry* lookup(const char *n, snapid_t snap=CEPH_NOSNAP) {
94b18763 441 return lookup(boost::string_view(n), snap);
7c673cae
FG
442 }
443
94b18763 444 CDentry* add_null_dentry(boost::string_view dname,
7c673cae 445 snapid_t first=2, snapid_t last=CEPH_NOSNAP);
94b18763 446 CDentry* add_primary_dentry(boost::string_view dname, CInode *in,
7c673cae 447 snapid_t first=2, snapid_t last=CEPH_NOSNAP);
94b18763 448 CDentry* add_remote_dentry(boost::string_view dname, inodeno_t ino, unsigned char d_type,
7c673cae
FG
449 snapid_t first=2, snapid_t last=CEPH_NOSNAP);
450 void remove_dentry( CDentry *dn ); // delete dentry
451 void link_remote_inode( CDentry *dn, inodeno_t ino, unsigned char d_type);
452 void link_remote_inode( CDentry *dn, CInode *in );
453 void link_primary_inode( CDentry *dn, CInode *in );
31f18b77 454 void unlink_inode(CDentry *dn, bool adjust_lru=true);
7c673cae
FG
455 void try_remove_unlinked_dn(CDentry *dn);
456
457 void add_to_bloom(CDentry *dn);
94b18763 458 bool is_in_bloom(boost::string_view name);
7c673cae
FG
459 bool has_bloom() { return (bloom ? true : false); }
460 void remove_bloom() {
461 bloom.reset();
462 }
463private:
464 void link_inode_work( CDentry *dn, CInode *in );
465 void unlink_inode_work( CDentry *dn );
466 void remove_null_dentries();
467 void purge_stale_snap_data(const std::set<snapid_t>& snaps);
468public:
7c673cae
FG
469 void try_remove_dentries_for_stray();
470 bool try_trim_snap_dentry(CDentry *dn, const std::set<snapid_t>& snaps);
471
472
473public:
94b18763
FG
474 void split(int bits, std::list<CDir*>& subs, list<MDSInternalContextBase*>& waiters, bool replay);
475 void merge(std::list<CDir*>& subs, std::list<MDSInternalContextBase*>& waiters, bool replay);
7c673cae
FG
476
477 bool should_split() const {
478 return (int)get_frag_size() > g_conf->mds_bal_split_size;
479 }
480 bool should_split_fast() const;
481 bool should_merge() const {
482 return (int)get_frag_size() < g_conf->mds_bal_merge_size;
483 }
484
485private:
486 void prepare_new_fragment(bool replay);
31f18b77 487 void prepare_old_fragment(map<string_snap_t, std::list<MDSInternalContextBase*> >& dentry_waiters, bool replay);
7c673cae 488 void steal_dentry(CDentry *dn); // from another dir. used by merge/split.
94b18763 489 void finish_old_fragment(std::list<MDSInternalContextBase*>& waiters, bool replay);
7c673cae
FG
490 void init_fragment_pins();
491
492
493 // -- authority --
494 /*
495 * normal: <parent,unknown> !subtree_root
496 * delegation: <mds,unknown> subtree_root
497 * ambiguous: <mds1,mds2> subtree_root
498 * <parent,mds2> subtree_root
499 */
500 mds_authority_t dir_auth;
501
502 std::string get_path() const;
503
504 public:
505 mds_authority_t authority() const override;
506 mds_authority_t get_dir_auth() const { return dir_auth; }
507 void set_dir_auth(mds_authority_t a);
508 void set_dir_auth(mds_rank_t a) { set_dir_auth(mds_authority_t(a, CDIR_AUTH_UNKNOWN)); }
509 bool is_ambiguous_dir_auth() const {
510 return dir_auth.second != CDIR_AUTH_UNKNOWN;
511 }
512 bool is_full_dir_auth() const {
513 return is_auth() && !is_ambiguous_dir_auth();
514 }
515 bool is_full_dir_nonauth() const {
516 return !is_auth() && !is_ambiguous_dir_auth();
517 }
518
519 bool is_subtree_root() const {
520 return dir_auth != CDIR_AUTH_DEFAULT;
521 }
522
523 bool contains(CDir *x); // true if we are x or an ancestor of x
524
525
526 // for giving to clients
527 void get_dist_spec(std::set<mds_rank_t>& ls, mds_rank_t auth) {
528 if (is_rep()) {
529 list_replicas(ls);
530 if (!ls.empty())
531 ls.insert(auth);
532 }
533 }
534 void encode_dirstat(bufferlist& bl, mds_rank_t whoami) {
535 /*
536 * note: encoding matches struct ceph_client_reply_dirfrag
537 */
538 frag_t frag = get_frag();
539 mds_rank_t auth;
540 std::set<mds_rank_t> dist;
541
542 auth = dir_auth.first;
543 if (is_auth())
544 get_dist_spec(dist, whoami);
545
546 ::encode(frag, bl);
547 ::encode(auth, bl);
548 ::encode(dist, bl);
549 }
550
551 void _encode_base(bufferlist& bl) {
552 ::encode(first, bl);
553 ::encode(fnode, bl);
554 ::encode(dir_rep, bl);
555 ::encode(dir_rep_by, bl);
556 }
557 void _decode_base(bufferlist::iterator& p) {
558 ::decode(first, p);
559 ::decode(fnode, p);
560 ::decode(dir_rep, p);
561 ::decode(dir_rep_by, p);
562 }
563 void encode_replica(mds_rank_t who, bufferlist& bl) {
564 __u32 nonce = add_replica(who);
565 ::encode(nonce, bl);
566 _encode_base(bl);
567 }
568 void decode_replica(bufferlist::iterator& p) {
569 __u32 nonce;
570 ::decode(nonce, p);
571 replica_nonce = nonce;
572 _decode_base(p);
573 }
574
575
576
577 // -- state --
578 bool is_complete() { return state & STATE_COMPLETE; }
579 bool is_exporting() { return state & STATE_EXPORTING; }
580 bool is_importing() { return state & STATE_IMPORTING; }
581 bool is_dirty_dft() { return state & STATE_DIRTYDFT; }
582
583 int get_dir_rep() const { return dir_rep; }
584 bool is_rep() const {
585 if (dir_rep == REP_NONE) return false;
586 return true;
587 }
588
589 // -- fetch --
590 object_t get_ondisk_object() {
591 return file_object_t(ino(), frag);
592 }
593 void fetch(MDSInternalContextBase *c, bool ignore_authpinnability=false);
94b18763 594 void fetch(MDSInternalContextBase *c, boost::string_view want_dn, bool ignore_authpinnability=false);
7c673cae
FG
595 void fetch(MDSInternalContextBase *c, const std::set<dentry_key_t>& keys);
596protected:
94b18763 597 mempool::mds_co::compact_set<mempool::mds_co::string> wanted_items;
7c673cae
FG
598
599 void _omap_fetch(MDSInternalContextBase *fin, const std::set<dentry_key_t>& keys);
600 void _omap_fetch_more(
601 bufferlist& hdrbl, std::map<std::string, bufferlist>& omap,
602 MDSInternalContextBase *fin);
603 CDentry *_load_dentry(
94b18763
FG
604 boost::string_view key,
605 boost::string_view dname,
7c673cae
FG
606 snapid_t last,
607 bufferlist &bl,
608 int pos,
609 const std::set<snapid_t> *snaps,
28e407b8 610 bool *force_dirty);
7c673cae
FG
611
612 /**
613 * Mark this fragment as BADFRAG (common part of go_bad and go_bad_dentry)
614 */
615 void _go_bad();
616
617 /**
618 * Go bad due to a damaged dentry (register with damagetable and go BADFRAG)
619 */
94b18763 620 void go_bad_dentry(snapid_t last, boost::string_view dname);
7c673cae
FG
621
622 /**
623 * Go bad due to a damaged header (register with damagetable and go BADFRAG)
624 */
625 void go_bad(bool complete);
626
627 void _omap_fetched(bufferlist& hdrbl, std::map<std::string, bufferlist>& omap,
628 bool complete, int r);
629
630 // -- commit --
94b18763 631 mempool::mds_co::compact_map<version_t, mempool::mds_co::list<MDSInternalContextBase*> > waiting_for_commit;
7c673cae
FG
632 void _commit(version_t want, int op_prio);
633 void _omap_commit(int op_prio);
634 void _encode_dentry(CDentry *dn, bufferlist& bl, const std::set<snapid_t> *snaps);
635 void _committed(int r, version_t v);
636public:
637#if 0 // unused?
638 void wait_for_commit(Context *c, version_t v=0);
639#endif
640 void commit_to(version_t want);
641 void commit(version_t want, MDSInternalContextBase *c,
642 bool ignore_authpinnability=false, int op_prio=-1);
643
644 // -- dirtyness --
645 version_t get_committing_version() const { return committing_version; }
646 version_t get_committed_version() const { return committed_version; }
647 void set_committed_version(version_t v) { committed_version = v; }
648
649 void mark_complete();
650
651
652 // -- reference counting --
653 void first_get() override;
654 void last_put() override;
655
656 void request_pin_get() {
657 if (request_pins == 0) get(PIN_REQUEST);
658 request_pins++;
659 }
660 void request_pin_put() {
661 request_pins--;
662 if (request_pins == 0) put(PIN_REQUEST);
663 }
664
665 // -- waiters --
666protected:
94b18763 667 mempool::mds_co::compact_map< string_snap_t, mempool::mds_co::list<MDSInternalContextBase*> > waiting_on_dentry; // FIXME string_snap_t not in mempool
7c673cae
FG
668
669public:
94b18763 670 bool is_waiting_for_dentry(boost::string_view dname, snapid_t snap) {
7c673cae
FG
671 return waiting_on_dentry.count(string_snap_t(dname, snap));
672 }
94b18763
FG
673 void add_dentry_waiter(boost::string_view dentry, snapid_t snap, MDSInternalContextBase *c);
674 void take_dentry_waiting(boost::string_view dentry, snapid_t first, snapid_t last, std::list<MDSInternalContextBase*>& ls);
7c673cae
FG
675 void take_sub_waiting(std::list<MDSInternalContextBase*>& ls); // dentry or ino
676
677 void add_waiter(uint64_t mask, MDSInternalContextBase *c) override;
678 void take_waiting(uint64_t mask, std::list<MDSInternalContextBase*>& ls) override; // may include dentry waiters
679 void finish_waiting(uint64_t mask, int result = 0); // ditto
680
681
682 // -- import/export --
683 void encode_export(bufferlist& bl);
684 void finish_export(utime_t now);
685 void abort_export() {
686 put(PIN_TEMPEXPORTING);
687 }
688 void decode_import(bufferlist::iterator& blp, utime_t now, LogSegment *ls);
689
690 // -- auth pins --
691 bool can_auth_pin() const override { return is_auth() && !(is_frozen() || is_freezing()); }
692 int get_cum_auth_pins() const { return auth_pins + nested_auth_pins; }
693 int get_auth_pins() const { return auth_pins; }
694 int get_nested_auth_pins() const { return nested_auth_pins; }
695 int get_dir_auth_pins() const { return dir_auth_pins; }
696 void auth_pin(void *who) override;
697 void auth_unpin(void *who) override;
698
699 void adjust_nested_auth_pins(int inc, int dirinc, void *by);
700 void verify_fragstat();
701
702 // -- freezing --
703 bool freeze_tree();
704 void _freeze_tree();
705 void unfreeze_tree();
706
707 bool freeze_dir();
708 void _freeze_dir();
709 void unfreeze_dir();
710
711 void maybe_finish_freeze();
712
713 bool is_freezing() const override { return is_freezing_tree() || is_freezing_dir(); }
714 bool is_freezing_tree() const;
715 bool is_freezing_tree_root() const { return state & STATE_FREEZINGTREE; }
716 bool is_freezing_dir() const { return state & STATE_FREEZINGDIR; }
717
718 bool is_frozen() const override { return is_frozen_dir() || is_frozen_tree(); }
719 bool is_frozen_tree() const;
720 bool is_frozen_tree_root() const { return state & STATE_FROZENTREE; }
721 bool is_frozen_dir() const { return state & STATE_FROZENDIR; }
722
723 bool is_freezeable(bool freezing=false) const {
724 // no nested auth pins.
725 if ((auth_pins-freezing) > 0 || nested_auth_pins > 0)
726 return false;
727
728 // inode must not be frozen.
729 if (!is_subtree_root() && inode->is_frozen())
730 return false;
731
732 return true;
733 }
734 bool is_freezeable_dir(bool freezing=false) const {
735 if ((auth_pins-freezing) > 0 || dir_auth_pins > 0)
736 return false;
737
738 // if not subtree root, inode must not be frozen (tree--frozen_dir is okay).
739 if (!is_subtree_root() && inode->is_frozen() && !inode->is_frozen_dir())
740 return false;
741
742 return true;
743 }
744
745 CDir *get_frozen_tree_root();
746
747
748 ostream& print_db_line_prefix(ostream& out) override;
749 void print(ostream& out) override;
750 void dump(Formatter *f) const;
28e407b8 751 void dump_load(Formatter *f, utime_t now, const DecayRate& rate);
7c673cae
FG
752};
753
754#endif