]>
Commit | Line | Data |
---|---|---|
1 | // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*- | |
2 | // vim: ts=8 sw=2 smarttab | |
3 | /* | |
4 | * 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 | #include "include/types.h" | |
17 | ||
18 | #include "CDir.h" | |
19 | #include "CDentry.h" | |
20 | #include "CInode.h" | |
21 | #include "Mutation.h" | |
22 | ||
23 | #include "MDSMap.h" | |
24 | #include "MDSRank.h" | |
25 | #include "MDCache.h" | |
26 | #include "Locker.h" | |
27 | #include "MDLog.h" | |
28 | #include "LogSegment.h" | |
29 | ||
30 | #include "common/bloom_filter.hpp" | |
31 | #include "include/Context.h" | |
32 | #include "common/Clock.h" | |
33 | ||
34 | #include "osdc/Objecter.h" | |
35 | ||
36 | #include "common/config.h" | |
37 | #include "include/assert.h" | |
38 | #include "include/compat.h" | |
39 | ||
40 | #define dout_context g_ceph_context | |
41 | #define dout_subsys ceph_subsys_mds | |
42 | #undef dout_prefix | |
43 | #define dout_prefix *_dout << "mds." << cache->mds->get_nodeid() << ".cache.dir(" << this->dirfrag() << ") " | |
44 | ||
45 | int CDir::num_frozen_trees = 0; | |
46 | int CDir::num_freezing_trees = 0; | |
47 | ||
48 | class CDirContext : public MDSInternalContextBase | |
49 | { | |
50 | protected: | |
51 | CDir *dir; | |
52 | MDSRank* get_mds() override {return dir->cache->mds;} | |
53 | ||
54 | public: | |
55 | explicit CDirContext(CDir *d) : dir(d) { | |
56 | assert(dir != NULL); | |
57 | } | |
58 | }; | |
59 | ||
60 | ||
61 | class CDirIOContext : public MDSIOContextBase | |
62 | { | |
63 | protected: | |
64 | CDir *dir; | |
65 | MDSRank* get_mds() override {return dir->cache->mds;} | |
66 | ||
67 | public: | |
68 | explicit CDirIOContext(CDir *d) : dir(d) { | |
69 | assert(dir != NULL); | |
70 | } | |
71 | }; | |
72 | ||
73 | ||
74 | // PINS | |
75 | //int cdir_pins[CDIR_NUM_PINS] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0 }; | |
76 | ||
77 | ||
78 | ostream& operator<<(ostream& out, const CDir& dir) | |
79 | { | |
80 | out << "[dir " << dir.dirfrag() << " " << dir.get_path() << "/" | |
81 | << " [" << dir.first << ",head]"; | |
82 | if (dir.is_auth()) { | |
83 | out << " auth"; | |
84 | if (dir.is_replicated()) | |
85 | out << dir.get_replicas(); | |
86 | ||
87 | if (dir.is_projected()) | |
88 | out << " pv=" << dir.get_projected_version(); | |
89 | out << " v=" << dir.get_version(); | |
90 | out << " cv=" << dir.get_committing_version(); | |
91 | out << "/" << dir.get_committed_version(); | |
92 | } else { | |
93 | mds_authority_t a = dir.authority(); | |
94 | out << " rep@" << a.first; | |
95 | if (a.second != CDIR_AUTH_UNKNOWN) | |
96 | out << "," << a.second; | |
97 | out << "." << dir.get_replica_nonce(); | |
98 | } | |
99 | ||
100 | if (dir.is_rep()) out << " REP"; | |
101 | ||
102 | if (dir.get_dir_auth() != CDIR_AUTH_DEFAULT) { | |
103 | if (dir.get_dir_auth().second == CDIR_AUTH_UNKNOWN) | |
104 | out << " dir_auth=" << dir.get_dir_auth().first; | |
105 | else | |
106 | out << " dir_auth=" << dir.get_dir_auth(); | |
107 | } | |
108 | ||
109 | if (dir.get_cum_auth_pins()) | |
110 | out << " ap=" << dir.get_auth_pins() | |
111 | << "+" << dir.get_dir_auth_pins() | |
112 | << "+" << dir.get_nested_auth_pins(); | |
113 | ||
114 | out << " state=" << dir.get_state(); | |
115 | if (dir.state_test(CDir::STATE_COMPLETE)) out << "|complete"; | |
116 | if (dir.state_test(CDir::STATE_FREEZINGTREE)) out << "|freezingtree"; | |
117 | if (dir.state_test(CDir::STATE_FROZENTREE)) out << "|frozentree"; | |
118 | if (dir.state_test(CDir::STATE_AUXSUBTREE)) out << "|auxsubtree"; | |
119 | //if (dir.state_test(CDir::STATE_FROZENTREELEAF)) out << "|frozentreeleaf"; | |
120 | if (dir.state_test(CDir::STATE_FROZENDIR)) out << "|frozendir"; | |
121 | if (dir.state_test(CDir::STATE_FREEZINGDIR)) out << "|freezingdir"; | |
122 | if (dir.state_test(CDir::STATE_EXPORTBOUND)) out << "|exportbound"; | |
123 | if (dir.state_test(CDir::STATE_IMPORTBOUND)) out << "|importbound"; | |
124 | if (dir.state_test(CDir::STATE_BADFRAG)) out << "|badfrag"; | |
125 | if (dir.state_test(CDir::STATE_FRAGMENTING)) out << "|fragmenting"; | |
126 | ||
127 | // fragstat | |
128 | out << " " << dir.fnode.fragstat; | |
129 | if (!(dir.fnode.fragstat == dir.fnode.accounted_fragstat)) | |
130 | out << "/" << dir.fnode.accounted_fragstat; | |
131 | if (g_conf->mds_debug_scatterstat && dir.is_projected()) { | |
132 | const fnode_t *pf = dir.get_projected_fnode(); | |
133 | out << "->" << pf->fragstat; | |
134 | if (!(pf->fragstat == pf->accounted_fragstat)) | |
135 | out << "/" << pf->accounted_fragstat; | |
136 | } | |
137 | ||
138 | // rstat | |
139 | out << " " << dir.fnode.rstat; | |
140 | if (!(dir.fnode.rstat == dir.fnode.accounted_rstat)) | |
141 | out << "/" << dir.fnode.accounted_rstat; | |
142 | if (g_conf->mds_debug_scatterstat && dir.is_projected()) { | |
143 | const fnode_t *pf = dir.get_projected_fnode(); | |
144 | out << "->" << pf->rstat; | |
145 | if (!(pf->rstat == pf->accounted_rstat)) | |
146 | out << "/" << pf->accounted_rstat; | |
147 | } | |
148 | ||
149 | out << " hs=" << dir.get_num_head_items() << "+" << dir.get_num_head_null(); | |
150 | out << ",ss=" << dir.get_num_snap_items() << "+" << dir.get_num_snap_null(); | |
151 | if (dir.get_num_dirty()) | |
152 | out << " dirty=" << dir.get_num_dirty(); | |
153 | ||
154 | if (dir.get_num_ref()) { | |
155 | out << " |"; | |
156 | dir.print_pin_set(out); | |
157 | } | |
158 | ||
159 | out << " " << &dir; | |
160 | return out << "]"; | |
161 | } | |
162 | ||
163 | ||
164 | void CDir::print(ostream& out) | |
165 | { | |
166 | out << *this; | |
167 | } | |
168 | ||
169 | ||
170 | ||
171 | ||
172 | ostream& CDir::print_db_line_prefix(ostream& out) | |
173 | { | |
174 | return out << ceph_clock_now() << " mds." << cache->mds->get_nodeid() << ".cache.dir(" << this->dirfrag() << ") "; | |
175 | } | |
176 | ||
177 | ||
178 | ||
179 | // ------------------------------------------------------------------- | |
180 | // CDir | |
181 | ||
182 | CDir::CDir(CInode *in, frag_t fg, MDCache *mdcache, bool auth) : | |
183 | cache(mdcache), inode(in), frag(fg), | |
184 | first(2), | |
185 | dirty_rstat_inodes(member_offset(CInode, dirty_rstat_item)), | |
186 | projected_version(0), item_dirty(this), item_new(this), | |
187 | num_head_items(0), num_head_null(0), | |
188 | num_snap_items(0), num_snap_null(0), | |
189 | num_dirty(0), committing_version(0), committed_version(0), | |
190 | dir_auth_pins(0), request_pins(0), | |
191 | dir_rep(REP_NONE), | |
192 | pop_me(ceph_clock_now()), | |
193 | pop_nested(ceph_clock_now()), | |
194 | pop_auth_subtree(ceph_clock_now()), | |
195 | pop_auth_subtree_nested(ceph_clock_now()), | |
196 | num_dentries_nested(0), num_dentries_auth_subtree(0), | |
197 | num_dentries_auth_subtree_nested(0), | |
198 | dir_auth(CDIR_AUTH_DEFAULT) | |
199 | { | |
200 | state = STATE_INITIAL; | |
201 | ||
202 | memset(&fnode, 0, sizeof(fnode)); | |
203 | ||
204 | // auth | |
205 | assert(in->is_dir()); | |
206 | if (auth) | |
207 | state |= STATE_AUTH; | |
208 | } | |
209 | ||
210 | /** | |
211 | * Check the recursive statistics on size for consistency. | |
212 | * If mds_debug_scatterstat is enabled, assert for correctness, | |
213 | * otherwise just print out the mismatch and continue. | |
214 | */ | |
215 | bool CDir::check_rstats(bool scrub) | |
216 | { | |
217 | if (!g_conf->mds_debug_scatterstat && !scrub) | |
218 | return true; | |
219 | ||
220 | dout(25) << "check_rstats on " << this << dendl; | |
221 | if (!is_complete() || !is_auth() || is_frozen()) { | |
222 | assert(!scrub); | |
223 | dout(10) << "check_rstats bailing out -- incomplete or non-auth or frozen dir!" << dendl; | |
224 | return true; | |
225 | } | |
226 | ||
227 | frag_info_t frag_info; | |
228 | nest_info_t nest_info; | |
229 | for (map_t::iterator i = items.begin(); i != items.end(); ++i) { | |
230 | if (i->second->last != CEPH_NOSNAP) | |
231 | continue; | |
232 | CDentry::linkage_t *dnl = i->second->get_linkage(); | |
233 | if (dnl->is_primary()) { | |
234 | CInode *in = dnl->get_inode(); | |
235 | nest_info.add(in->inode.accounted_rstat); | |
236 | if (in->is_dir()) | |
237 | frag_info.nsubdirs++; | |
238 | else | |
239 | frag_info.nfiles++; | |
240 | } else if (dnl->is_remote()) | |
241 | frag_info.nfiles++; | |
242 | } | |
243 | ||
244 | bool good = true; | |
245 | // fragstat | |
246 | if(!frag_info.same_sums(fnode.fragstat)) { | |
247 | dout(1) << "mismatch between head items and fnode.fragstat! printing dentries" << dendl; | |
248 | dout(1) << "get_num_head_items() = " << get_num_head_items() | |
249 | << "; fnode.fragstat.nfiles=" << fnode.fragstat.nfiles | |
250 | << " fnode.fragstat.nsubdirs=" << fnode.fragstat.nsubdirs << dendl; | |
251 | good = false; | |
252 | } else { | |
253 | dout(20) << "get_num_head_items() = " << get_num_head_items() | |
254 | << "; fnode.fragstat.nfiles=" << fnode.fragstat.nfiles | |
255 | << " fnode.fragstat.nsubdirs=" << fnode.fragstat.nsubdirs << dendl; | |
256 | } | |
257 | ||
258 | // rstat | |
259 | if (!nest_info.same_sums(fnode.rstat)) { | |
260 | dout(1) << "mismatch between child accounted_rstats and my rstats!" << dendl; | |
261 | dout(1) << "total of child dentrys: " << nest_info << dendl; | |
262 | dout(1) << "my rstats: " << fnode.rstat << dendl; | |
263 | good = false; | |
264 | } else { | |
265 | dout(20) << "total of child dentrys: " << nest_info << dendl; | |
266 | dout(20) << "my rstats: " << fnode.rstat << dendl; | |
267 | } | |
268 | ||
269 | if (!good) { | |
270 | if (!scrub) { | |
271 | for (map_t::iterator i = items.begin(); i != items.end(); ++i) { | |
272 | CDentry *dn = i->second; | |
273 | if (dn->get_linkage()->is_primary()) { | |
274 | CInode *in = dn->get_linkage()->inode; | |
275 | dout(1) << *dn << " rstat " << in->inode.accounted_rstat << dendl; | |
276 | } else { | |
277 | dout(1) << *dn << dendl; | |
278 | } | |
279 | } | |
280 | ||
281 | assert(frag_info.nfiles == fnode.fragstat.nfiles); | |
282 | assert(frag_info.nsubdirs == fnode.fragstat.nsubdirs); | |
283 | assert(nest_info.rbytes == fnode.rstat.rbytes); | |
284 | assert(nest_info.rfiles == fnode.rstat.rfiles); | |
285 | assert(nest_info.rsubdirs == fnode.rstat.rsubdirs); | |
286 | } | |
287 | } | |
288 | dout(10) << "check_rstats complete on " << this << dendl; | |
289 | return good; | |
290 | } | |
291 | ||
292 | CDentry *CDir::lookup(const string& name, snapid_t snap) | |
293 | { | |
294 | dout(20) << "lookup (" << snap << ", '" << name << "')" << dendl; | |
295 | map_t::iterator iter = items.lower_bound(dentry_key_t(snap, name.c_str(), | |
296 | inode->hash_dentry_name(name))); | |
297 | if (iter == items.end()) | |
298 | return 0; | |
299 | if (iter->second->name == name && | |
300 | iter->second->first <= snap && | |
301 | iter->second->last >= snap) { | |
302 | dout(20) << " hit -> " << iter->first << dendl; | |
303 | return iter->second; | |
304 | } | |
305 | dout(20) << " miss -> " << iter->first << dendl; | |
306 | return 0; | |
307 | } | |
308 | ||
309 | CDentry *CDir::lookup_exact_snap(const string& name, snapid_t last) { | |
310 | map_t::iterator p = items.find(dentry_key_t(last, name.c_str(), | |
311 | inode->hash_dentry_name(name))); | |
312 | if (p == items.end()) | |
313 | return NULL; | |
314 | return p->second; | |
315 | } | |
316 | ||
317 | /*** | |
318 | * linking fun | |
319 | */ | |
320 | ||
321 | CDentry* CDir::add_null_dentry(const string& dname, | |
322 | snapid_t first, snapid_t last) | |
323 | { | |
324 | // foreign | |
325 | assert(lookup_exact_snap(dname, last) == 0); | |
326 | ||
327 | // create dentry | |
328 | CDentry* dn = new CDentry(dname, inode->hash_dentry_name(dname), first, last); | |
329 | if (is_auth()) | |
330 | dn->state_set(CDentry::STATE_AUTH); | |
331 | ||
332 | cache->bottom_lru.lru_insert_mid(dn); | |
333 | dn->state_set(CDentry::STATE_BOTTOMLRU); | |
334 | ||
335 | dn->dir = this; | |
336 | dn->version = get_projected_version(); | |
337 | ||
338 | // add to dir | |
339 | assert(items.count(dn->key()) == 0); | |
340 | //assert(null_items.count(dn->name) == 0); | |
341 | ||
342 | items[dn->key()] = dn; | |
343 | if (last == CEPH_NOSNAP) | |
344 | num_head_null++; | |
345 | else | |
346 | num_snap_null++; | |
347 | ||
348 | if (state_test(CDir::STATE_DNPINNEDFRAG)) { | |
349 | dn->get(CDentry::PIN_FRAGMENTING); | |
350 | dn->state_set(CDentry::STATE_FRAGMENTING); | |
351 | } | |
352 | ||
353 | dout(12) << "add_null_dentry " << *dn << dendl; | |
354 | ||
355 | // pin? | |
356 | if (get_num_any() == 1) | |
357 | get(PIN_CHILD); | |
358 | ||
359 | assert(get_num_any() == items.size()); | |
360 | return dn; | |
361 | } | |
362 | ||
363 | ||
364 | CDentry* CDir::add_primary_dentry(const string& dname, CInode *in, | |
365 | snapid_t first, snapid_t last) | |
366 | { | |
367 | // primary | |
368 | assert(lookup_exact_snap(dname, last) == 0); | |
369 | ||
370 | // create dentry | |
371 | CDentry* dn = new CDentry(dname, inode->hash_dentry_name(dname), first, last); | |
372 | if (is_auth()) | |
373 | dn->state_set(CDentry::STATE_AUTH); | |
374 | if (is_auth() || !inode->is_stray()) { | |
375 | cache->lru.lru_insert_mid(dn); | |
376 | } else { | |
377 | cache->bottom_lru.lru_insert_mid(dn); | |
378 | dn->state_set(CDentry::STATE_BOTTOMLRU); | |
379 | } | |
380 | ||
381 | dn->dir = this; | |
382 | dn->version = get_projected_version(); | |
383 | ||
384 | // add to dir | |
385 | assert(items.count(dn->key()) == 0); | |
386 | //assert(null_items.count(dn->name) == 0); | |
387 | ||
388 | items[dn->key()] = dn; | |
389 | ||
390 | dn->get_linkage()->inode = in; | |
391 | in->set_primary_parent(dn); | |
392 | ||
393 | link_inode_work(dn, in); | |
394 | ||
395 | if (dn->last == CEPH_NOSNAP) | |
396 | num_head_items++; | |
397 | else | |
398 | num_snap_items++; | |
399 | ||
400 | if (state_test(CDir::STATE_DNPINNEDFRAG)) { | |
401 | dn->get(CDentry::PIN_FRAGMENTING); | |
402 | dn->state_set(CDentry::STATE_FRAGMENTING); | |
403 | } | |
404 | ||
405 | dout(12) << "add_primary_dentry " << *dn << dendl; | |
406 | ||
407 | // pin? | |
408 | if (get_num_any() == 1) | |
409 | get(PIN_CHILD); | |
410 | assert(get_num_any() == items.size()); | |
411 | return dn; | |
412 | } | |
413 | ||
414 | CDentry* CDir::add_remote_dentry(const string& dname, inodeno_t ino, unsigned char d_type, | |
415 | snapid_t first, snapid_t last) | |
416 | { | |
417 | // foreign | |
418 | assert(lookup_exact_snap(dname, last) == 0); | |
419 | ||
420 | // create dentry | |
421 | CDentry* dn = new CDentry(dname, inode->hash_dentry_name(dname), ino, d_type, first, last); | |
422 | if (is_auth()) | |
423 | dn->state_set(CDentry::STATE_AUTH); | |
424 | cache->lru.lru_insert_mid(dn); | |
425 | ||
426 | dn->dir = this; | |
427 | dn->version = get_projected_version(); | |
428 | ||
429 | // add to dir | |
430 | assert(items.count(dn->key()) == 0); | |
431 | //assert(null_items.count(dn->name) == 0); | |
432 | ||
433 | items[dn->key()] = dn; | |
434 | if (last == CEPH_NOSNAP) | |
435 | num_head_items++; | |
436 | else | |
437 | num_snap_items++; | |
438 | ||
439 | if (state_test(CDir::STATE_DNPINNEDFRAG)) { | |
440 | dn->get(CDentry::PIN_FRAGMENTING); | |
441 | dn->state_set(CDentry::STATE_FRAGMENTING); | |
442 | } | |
443 | ||
444 | dout(12) << "add_remote_dentry " << *dn << dendl; | |
445 | ||
446 | // pin? | |
447 | if (get_num_any() == 1) | |
448 | get(PIN_CHILD); | |
449 | ||
450 | assert(get_num_any() == items.size()); | |
451 | return dn; | |
452 | } | |
453 | ||
454 | ||
455 | ||
456 | void CDir::remove_dentry(CDentry *dn) | |
457 | { | |
458 | dout(12) << "remove_dentry " << *dn << dendl; | |
459 | ||
460 | // there should be no client leases at this point! | |
461 | assert(dn->client_lease_map.empty()); | |
462 | ||
463 | if (state_test(CDir::STATE_DNPINNEDFRAG)) { | |
464 | dn->put(CDentry::PIN_FRAGMENTING); | |
465 | dn->state_clear(CDentry::STATE_FRAGMENTING); | |
466 | } | |
467 | ||
468 | if (dn->get_linkage()->is_null()) { | |
469 | if (dn->last == CEPH_NOSNAP) | |
470 | num_head_null--; | |
471 | else | |
472 | num_snap_null--; | |
473 | } else { | |
474 | if (dn->last == CEPH_NOSNAP) | |
475 | num_head_items--; | |
476 | else | |
477 | num_snap_items--; | |
478 | } | |
479 | ||
480 | if (!dn->get_linkage()->is_null()) | |
481 | // detach inode and dentry | |
482 | unlink_inode_work(dn); | |
483 | ||
484 | // remove from list | |
485 | assert(items.count(dn->key()) == 1); | |
486 | items.erase(dn->key()); | |
487 | ||
488 | // clean? | |
489 | if (dn->is_dirty()) | |
490 | dn->mark_clean(); | |
491 | ||
492 | if (dn->state_test(CDentry::STATE_BOTTOMLRU)) | |
493 | cache->bottom_lru.lru_remove(dn); | |
494 | else | |
495 | cache->lru.lru_remove(dn); | |
496 | delete dn; | |
497 | ||
498 | // unpin? | |
499 | if (get_num_any() == 0) | |
500 | put(PIN_CHILD); | |
501 | assert(get_num_any() == items.size()); | |
502 | } | |
503 | ||
504 | void CDir::link_remote_inode(CDentry *dn, CInode *in) | |
505 | { | |
506 | link_remote_inode(dn, in->ino(), IFTODT(in->get_projected_inode()->mode)); | |
507 | } | |
508 | ||
509 | void CDir::link_remote_inode(CDentry *dn, inodeno_t ino, unsigned char d_type) | |
510 | { | |
511 | dout(12) << "link_remote_inode " << *dn << " remote " << ino << dendl; | |
512 | assert(dn->get_linkage()->is_null()); | |
513 | ||
514 | dn->get_linkage()->set_remote(ino, d_type); | |
515 | ||
516 | if (dn->state_test(CDentry::STATE_BOTTOMLRU)) { | |
517 | cache->bottom_lru.lru_remove(dn); | |
518 | cache->lru.lru_insert_mid(dn); | |
519 | dn->state_clear(CDentry::STATE_BOTTOMLRU); | |
520 | } | |
521 | ||
522 | if (dn->last == CEPH_NOSNAP) { | |
523 | num_head_items++; | |
524 | num_head_null--; | |
525 | } else { | |
526 | num_snap_items++; | |
527 | num_snap_null--; | |
528 | } | |
529 | assert(get_num_any() == items.size()); | |
530 | } | |
531 | ||
532 | void CDir::link_primary_inode(CDentry *dn, CInode *in) | |
533 | { | |
534 | dout(12) << "link_primary_inode " << *dn << " " << *in << dendl; | |
535 | assert(dn->get_linkage()->is_null()); | |
536 | ||
537 | dn->get_linkage()->inode = in; | |
538 | in->set_primary_parent(dn); | |
539 | ||
540 | link_inode_work(dn, in); | |
541 | ||
542 | if (dn->state_test(CDentry::STATE_BOTTOMLRU) && | |
543 | (is_auth() || !inode->is_stray())) { | |
544 | cache->bottom_lru.lru_remove(dn); | |
545 | cache->lru.lru_insert_mid(dn); | |
546 | dn->state_clear(CDentry::STATE_BOTTOMLRU); | |
547 | } | |
548 | ||
549 | if (dn->last == CEPH_NOSNAP) { | |
550 | num_head_items++; | |
551 | num_head_null--; | |
552 | } else { | |
553 | num_snap_items++; | |
554 | num_snap_null--; | |
555 | } | |
556 | ||
557 | assert(get_num_any() == items.size()); | |
558 | } | |
559 | ||
560 | void CDir::link_inode_work( CDentry *dn, CInode *in) | |
561 | { | |
562 | assert(dn->get_linkage()->get_inode() == in); | |
563 | assert(in->get_parent_dn() == dn); | |
564 | ||
565 | // set inode version | |
566 | //in->inode.version = dn->get_version(); | |
567 | ||
568 | // pin dentry? | |
569 | if (in->get_num_ref()) | |
570 | dn->get(CDentry::PIN_INODEPIN); | |
571 | ||
572 | // adjust auth pin count | |
573 | if (in->auth_pins + in->nested_auth_pins) | |
574 | dn->adjust_nested_auth_pins(in->auth_pins + in->nested_auth_pins, in->auth_pins, NULL); | |
575 | ||
576 | // verify open snaprealm parent | |
577 | if (in->snaprealm) | |
578 | in->snaprealm->adjust_parent(); | |
579 | else if (in->is_any_caps()) | |
580 | in->move_to_realm(inode->find_snaprealm()); | |
581 | } | |
582 | ||
583 | void CDir::unlink_inode(CDentry *dn, bool adjust_lru) | |
584 | { | |
585 | if (dn->get_linkage()->is_primary()) { | |
586 | dout(12) << "unlink_inode " << *dn << " " << *dn->get_linkage()->get_inode() << dendl; | |
587 | } else { | |
588 | dout(12) << "unlink_inode " << *dn << dendl; | |
589 | } | |
590 | ||
591 | unlink_inode_work(dn); | |
592 | ||
593 | if (adjust_lru && !dn->state_test(CDentry::STATE_BOTTOMLRU)) { | |
594 | cache->lru.lru_remove(dn); | |
595 | cache->bottom_lru.lru_insert_mid(dn); | |
596 | dn->state_set(CDentry::STATE_BOTTOMLRU); | |
597 | } | |
598 | ||
599 | if (dn->last == CEPH_NOSNAP) { | |
600 | num_head_items--; | |
601 | num_head_null++; | |
602 | } else { | |
603 | num_snap_items--; | |
604 | num_snap_null++; | |
605 | } | |
606 | assert(get_num_any() == items.size()); | |
607 | } | |
608 | ||
609 | ||
610 | void CDir::try_remove_unlinked_dn(CDentry *dn) | |
611 | { | |
612 | assert(dn->dir == this); | |
613 | assert(dn->get_linkage()->is_null()); | |
614 | ||
615 | // no pins (besides dirty)? | |
616 | if (dn->get_num_ref() != dn->is_dirty()) | |
617 | return; | |
618 | ||
619 | // was the dn new? | |
620 | if (dn->is_new()) { | |
621 | dout(10) << "try_remove_unlinked_dn " << *dn << " in " << *this << dendl; | |
622 | if (dn->is_dirty()) | |
623 | dn->mark_clean(); | |
624 | remove_dentry(dn); | |
625 | ||
626 | // NOTE: we may not have any more dirty dentries, but the fnode | |
627 | // still changed, so the directory must remain dirty. | |
628 | } | |
629 | } | |
630 | ||
631 | ||
632 | void CDir::unlink_inode_work( CDentry *dn ) | |
633 | { | |
634 | CInode *in = dn->get_linkage()->get_inode(); | |
635 | ||
636 | if (dn->get_linkage()->is_remote()) { | |
637 | // remote | |
638 | if (in) | |
639 | dn->unlink_remote(dn->get_linkage()); | |
640 | ||
641 | dn->get_linkage()->set_remote(0, 0); | |
642 | } else if (dn->get_linkage()->is_primary()) { | |
643 | // primary | |
644 | // unpin dentry? | |
645 | if (in->get_num_ref()) | |
646 | dn->put(CDentry::PIN_INODEPIN); | |
647 | ||
648 | // unlink auth_pin count | |
649 | if (in->auth_pins + in->nested_auth_pins) | |
650 | dn->adjust_nested_auth_pins(0 - (in->auth_pins + in->nested_auth_pins), 0 - in->auth_pins, NULL); | |
651 | ||
652 | // detach inode | |
653 | in->remove_primary_parent(dn); | |
654 | dn->get_linkage()->inode = 0; | |
655 | } else { | |
656 | assert(!dn->get_linkage()->is_null()); | |
657 | } | |
658 | } | |
659 | ||
660 | void CDir::add_to_bloom(CDentry *dn) | |
661 | { | |
662 | assert(dn->last == CEPH_NOSNAP); | |
663 | if (!bloom) { | |
664 | /* not create bloom filter for incomplete dir that was added by log replay */ | |
665 | if (!is_complete()) | |
666 | return; | |
667 | ||
668 | /* don't maintain bloom filters in standby replay (saves cycles, and also | |
669 | * avoids need to implement clearing it in EExport for #16924) */ | |
670 | if (cache->mds->is_standby_replay()) { | |
671 | return; | |
672 | } | |
673 | ||
674 | unsigned size = get_num_head_items() + get_num_snap_items(); | |
675 | if (size < 100) size = 100; | |
676 | bloom.reset(new bloom_filter(size, 1.0 / size, 0)); | |
677 | } | |
678 | /* This size and false positive probability is completely random.*/ | |
679 | bloom->insert(dn->name.c_str(), dn->name.size()); | |
680 | } | |
681 | ||
682 | bool CDir::is_in_bloom(const string& name) | |
683 | { | |
684 | if (!bloom) | |
685 | return false; | |
686 | return bloom->contains(name.c_str(), name.size()); | |
687 | } | |
688 | ||
689 | void CDir::remove_null_dentries() { | |
690 | dout(12) << "remove_null_dentries " << *this << dendl; | |
691 | ||
692 | CDir::map_t::iterator p = items.begin(); | |
693 | while (p != items.end()) { | |
694 | CDentry *dn = p->second; | |
695 | ++p; | |
696 | if (dn->get_linkage()->is_null() && !dn->is_projected()) | |
697 | remove_dentry(dn); | |
698 | } | |
699 | ||
700 | assert(num_snap_null == 0); | |
701 | assert(num_head_null == 0); | |
702 | assert(get_num_any() == items.size()); | |
703 | } | |
704 | ||
705 | /** remove dirty null dentries for deleted directory. the dirfrag will be | |
706 | * deleted soon, so it's safe to not commit dirty dentries. | |
707 | * | |
708 | * This is called when a directory is being deleted, a prerequisite | |
709 | * of which is that its children have been unlinked: we expect to only see | |
710 | * null, unprojected dentries here. | |
711 | */ | |
712 | void CDir::try_remove_dentries_for_stray() | |
713 | { | |
714 | dout(10) << __func__ << dendl; | |
715 | assert(get_parent_dir()->inode->is_stray()); | |
716 | ||
717 | // clear dirty only when the directory was not snapshotted | |
718 | bool clear_dirty = !inode->snaprealm; | |
719 | ||
720 | CDir::map_t::iterator p = items.begin(); | |
721 | while (p != items.end()) { | |
722 | CDentry *dn = p->second; | |
723 | ++p; | |
724 | if (dn->last == CEPH_NOSNAP) { | |
725 | assert(!dn->is_projected()); | |
726 | assert(dn->get_linkage()->is_null()); | |
727 | if (clear_dirty && dn->is_dirty()) | |
728 | dn->mark_clean(); | |
729 | // It's OK to remove lease prematurely because we will never link | |
730 | // the dentry to inode again. | |
731 | if (dn->is_any_leases()) | |
732 | dn->remove_client_leases(cache->mds->locker); | |
733 | if (dn->get_num_ref() == 0) | |
734 | remove_dentry(dn); | |
735 | } else { | |
736 | assert(!dn->is_projected()); | |
737 | CDentry::linkage_t *dnl= dn->get_linkage(); | |
738 | CInode *in = NULL; | |
739 | if (dnl->is_primary()) { | |
740 | in = dnl->get_inode(); | |
741 | if (clear_dirty && in->is_dirty()) | |
742 | in->mark_clean(); | |
743 | } | |
744 | if (clear_dirty && dn->is_dirty()) | |
745 | dn->mark_clean(); | |
746 | if (dn->get_num_ref() == 0) { | |
747 | remove_dentry(dn); | |
748 | if (in) | |
749 | cache->remove_inode(in); | |
750 | } | |
751 | } | |
752 | } | |
753 | ||
754 | if (clear_dirty && is_dirty()) | |
755 | mark_clean(); | |
756 | } | |
757 | ||
758 | bool CDir::try_trim_snap_dentry(CDentry *dn, const set<snapid_t>& snaps) | |
759 | { | |
760 | assert(dn->last != CEPH_NOSNAP); | |
761 | set<snapid_t>::const_iterator p = snaps.lower_bound(dn->first); | |
762 | CDentry::linkage_t *dnl= dn->get_linkage(); | |
763 | CInode *in = 0; | |
764 | if (dnl->is_primary()) | |
765 | in = dnl->get_inode(); | |
766 | if ((p == snaps.end() || *p > dn->last) && | |
767 | (dn->get_num_ref() == dn->is_dirty()) && | |
768 | (!in || in->get_num_ref() == in->is_dirty())) { | |
769 | dout(10) << " purging snapped " << *dn << dendl; | |
770 | if (in && in->is_dirty()) | |
771 | in->mark_clean(); | |
772 | remove_dentry(dn); | |
773 | if (in) { | |
774 | dout(10) << " purging snapped " << *in << dendl; | |
775 | cache->remove_inode(in); | |
776 | } | |
777 | return true; | |
778 | } | |
779 | return false; | |
780 | } | |
781 | ||
782 | ||
783 | void CDir::purge_stale_snap_data(const set<snapid_t>& snaps) | |
784 | { | |
785 | dout(10) << "purge_stale_snap_data " << snaps << dendl; | |
786 | ||
787 | CDir::map_t::iterator p = items.begin(); | |
788 | while (p != items.end()) { | |
789 | CDentry *dn = p->second; | |
790 | ++p; | |
791 | ||
792 | if (dn->last == CEPH_NOSNAP) | |
793 | continue; | |
794 | ||
795 | try_trim_snap_dentry(dn, snaps); | |
796 | } | |
797 | } | |
798 | ||
799 | ||
800 | /** | |
801 | * steal_dentry -- semi-violently move a dentry from one CDir to another | |
802 | * (*) violently, in that nitems, most pins, etc. are not correctly maintained | |
803 | * on the old CDir corpse; must call finish_old_fragment() when finished. | |
804 | */ | |
805 | void CDir::steal_dentry(CDentry *dn) | |
806 | { | |
807 | dout(15) << "steal_dentry " << *dn << dendl; | |
808 | ||
809 | items[dn->key()] = dn; | |
810 | ||
811 | dn->dir->items.erase(dn->key()); | |
812 | if (dn->dir->items.empty()) | |
813 | dn->dir->put(PIN_CHILD); | |
814 | ||
815 | if (get_num_any() == 0) | |
816 | get(PIN_CHILD); | |
817 | if (dn->get_linkage()->is_null()) { | |
818 | if (dn->last == CEPH_NOSNAP) | |
819 | num_head_null++; | |
820 | else | |
821 | num_snap_null++; | |
822 | } else if (dn->last == CEPH_NOSNAP) { | |
823 | num_head_items++; | |
824 | ||
825 | if (dn->get_linkage()->is_primary()) { | |
826 | CInode *in = dn->get_linkage()->get_inode(); | |
827 | inode_t *pi = in->get_projected_inode(); | |
828 | if (dn->get_linkage()->get_inode()->is_dir()) | |
829 | fnode.fragstat.nsubdirs++; | |
830 | else | |
831 | fnode.fragstat.nfiles++; | |
832 | fnode.rstat.rbytes += pi->accounted_rstat.rbytes; | |
833 | fnode.rstat.rfiles += pi->accounted_rstat.rfiles; | |
834 | fnode.rstat.rsubdirs += pi->accounted_rstat.rsubdirs; | |
835 | fnode.rstat.rsnaprealms += pi->accounted_rstat.rsnaprealms; | |
836 | if (pi->accounted_rstat.rctime > fnode.rstat.rctime) | |
837 | fnode.rstat.rctime = pi->accounted_rstat.rctime; | |
838 | ||
839 | // move dirty inode rstat to new dirfrag | |
840 | if (in->is_dirty_rstat()) | |
841 | dirty_rstat_inodes.push_back(&in->dirty_rstat_item); | |
842 | } else if (dn->get_linkage()->is_remote()) { | |
843 | if (dn->get_linkage()->get_remote_d_type() == DT_DIR) | |
844 | fnode.fragstat.nsubdirs++; | |
845 | else | |
846 | fnode.fragstat.nfiles++; | |
847 | } | |
848 | } else { | |
849 | num_snap_items++; | |
850 | if (dn->get_linkage()->is_primary()) { | |
851 | CInode *in = dn->get_linkage()->get_inode(); | |
852 | if (in->is_dirty_rstat()) | |
853 | dirty_rstat_inodes.push_back(&in->dirty_rstat_item); | |
854 | } | |
855 | } | |
856 | ||
857 | if (dn->auth_pins || dn->nested_auth_pins) { | |
858 | // use the helpers here to maintain the auth_pin invariants on the dir inode | |
859 | int ap = dn->get_num_auth_pins() + dn->get_num_nested_auth_pins(); | |
860 | int dap = dn->get_num_dir_auth_pins(); | |
861 | assert(dap <= ap); | |
862 | adjust_nested_auth_pins(ap, dap, NULL); | |
863 | dn->dir->adjust_nested_auth_pins(-ap, -dap, NULL); | |
864 | } | |
865 | ||
866 | if (dn->is_dirty()) | |
867 | num_dirty++; | |
868 | ||
869 | dn->dir = this; | |
870 | } | |
871 | ||
872 | void CDir::prepare_old_fragment(map<string_snap_t, std::list<MDSInternalContextBase*> >& dentry_waiters, bool replay) | |
873 | { | |
874 | // auth_pin old fragment for duration so that any auth_pinning | |
875 | // during the dentry migration doesn't trigger side effects | |
876 | if (!replay && is_auth()) | |
877 | auth_pin(this); | |
878 | ||
879 | if (!waiting_on_dentry.empty()) { | |
880 | for (auto p = waiting_on_dentry.begin(); p != waiting_on_dentry.end(); ++p) | |
881 | dentry_waiters[p->first].swap(p->second); | |
882 | waiting_on_dentry.clear(); | |
883 | put(PIN_DNWAITER); | |
884 | } | |
885 | } | |
886 | ||
887 | void CDir::prepare_new_fragment(bool replay) | |
888 | { | |
889 | if (!replay && is_auth()) { | |
890 | _freeze_dir(); | |
891 | mark_complete(); | |
892 | } | |
893 | inode->add_dirfrag(this); | |
894 | } | |
895 | ||
896 | void CDir::finish_old_fragment(list<MDSInternalContextBase*>& waiters, bool replay) | |
897 | { | |
898 | // take waiters _before_ unfreeze... | |
899 | if (!replay) { | |
900 | take_waiting(WAIT_ANY_MASK, waiters); | |
901 | if (is_auth()) { | |
902 | auth_unpin(this); // pinned in prepare_old_fragment | |
903 | assert(is_frozen_dir()); | |
904 | unfreeze_dir(); | |
905 | } | |
906 | } | |
907 | ||
908 | assert(nested_auth_pins == 0); | |
909 | assert(dir_auth_pins == 0); | |
910 | assert(auth_pins == 0); | |
911 | ||
912 | num_head_items = num_head_null = 0; | |
913 | num_snap_items = num_snap_null = 0; | |
914 | ||
915 | // this mirrors init_fragment_pins() | |
916 | if (is_auth()) | |
917 | clear_replica_map(); | |
918 | if (is_dirty()) | |
919 | mark_clean(); | |
920 | if (state_test(STATE_IMPORTBOUND)) | |
921 | put(PIN_IMPORTBOUND); | |
922 | if (state_test(STATE_EXPORTBOUND)) | |
923 | put(PIN_EXPORTBOUND); | |
924 | if (is_subtree_root()) | |
925 | put(PIN_SUBTREE); | |
926 | ||
927 | if (auth_pins > 0) | |
928 | put(PIN_AUTHPIN); | |
929 | ||
930 | assert(get_num_ref() == (state_test(STATE_STICKY) ? 1:0)); | |
931 | } | |
932 | ||
933 | void CDir::init_fragment_pins() | |
934 | { | |
935 | if (!replica_map.empty()) | |
936 | get(PIN_REPLICATED); | |
937 | if (state_test(STATE_DIRTY)) | |
938 | get(PIN_DIRTY); | |
939 | if (state_test(STATE_EXPORTBOUND)) | |
940 | get(PIN_EXPORTBOUND); | |
941 | if (state_test(STATE_IMPORTBOUND)) | |
942 | get(PIN_IMPORTBOUND); | |
943 | if (is_subtree_root()) | |
944 | get(PIN_SUBTREE); | |
945 | } | |
946 | ||
947 | void CDir::split(int bits, list<CDir*>& subs, list<MDSInternalContextBase*>& waiters, bool replay) | |
948 | { | |
949 | dout(10) << "split by " << bits << " bits on " << *this << dendl; | |
950 | ||
951 | assert(replay || is_complete() || !is_auth()); | |
952 | ||
953 | list<frag_t> frags; | |
954 | frag.split(bits, frags); | |
955 | ||
956 | vector<CDir*> subfrags(1 << bits); | |
957 | ||
958 | double fac = 1.0 / (double)(1 << bits); // for scaling load vecs | |
959 | ||
960 | version_t rstat_version = inode->get_projected_inode()->rstat.version; | |
961 | version_t dirstat_version = inode->get_projected_inode()->dirstat.version; | |
962 | ||
963 | nest_info_t rstatdiff; | |
964 | frag_info_t fragstatdiff; | |
965 | if (fnode.accounted_rstat.version == rstat_version) | |
966 | rstatdiff.add_delta(fnode.accounted_rstat, fnode.rstat); | |
967 | if (fnode.accounted_fragstat.version == dirstat_version) | |
968 | fragstatdiff.add_delta(fnode.accounted_fragstat, fnode.fragstat); | |
969 | dout(10) << " rstatdiff " << rstatdiff << " fragstatdiff " << fragstatdiff << dendl; | |
970 | ||
971 | map<string_snap_t, std::list<MDSInternalContextBase*> > dentry_waiters; | |
972 | prepare_old_fragment(dentry_waiters, replay); | |
973 | ||
974 | // create subfrag dirs | |
975 | int n = 0; | |
976 | for (list<frag_t>::iterator p = frags.begin(); p != frags.end(); ++p) { | |
977 | CDir *f = new CDir(inode, *p, cache, is_auth()); | |
978 | f->state_set(state & (MASK_STATE_FRAGMENT_KEPT | STATE_COMPLETE)); | |
979 | f->replica_map = replica_map; | |
980 | f->dir_auth = dir_auth; | |
981 | f->init_fragment_pins(); | |
982 | f->set_version(get_version()); | |
983 | ||
984 | f->pop_me = pop_me; | |
985 | f->pop_me.scale(fac); | |
986 | ||
987 | // FIXME; this is an approximation | |
988 | f->pop_nested = pop_nested; | |
989 | f->pop_nested.scale(fac); | |
990 | f->pop_auth_subtree = pop_auth_subtree; | |
991 | f->pop_auth_subtree.scale(fac); | |
992 | f->pop_auth_subtree_nested = pop_auth_subtree_nested; | |
993 | f->pop_auth_subtree_nested.scale(fac); | |
994 | ||
995 | dout(10) << " subfrag " << *p << " " << *f << dendl; | |
996 | subfrags[n++] = f; | |
997 | subs.push_back(f); | |
998 | ||
999 | f->set_dir_auth(get_dir_auth()); | |
1000 | f->prepare_new_fragment(replay); | |
1001 | } | |
1002 | ||
1003 | // repartition dentries | |
1004 | while (!items.empty()) { | |
1005 | CDir::map_t::iterator p = items.begin(); | |
1006 | ||
1007 | CDentry *dn = p->second; | |
1008 | frag_t subfrag = inode->pick_dirfrag(dn->name); | |
1009 | int n = (subfrag.value() & (subfrag.mask() ^ frag.mask())) >> subfrag.mask_shift(); | |
1010 | dout(15) << " subfrag " << subfrag << " n=" << n << " for " << p->first << dendl; | |
1011 | CDir *f = subfrags[n]; | |
1012 | f->steal_dentry(dn); | |
1013 | } | |
1014 | ||
1015 | for (auto& p : dentry_waiters) { | |
1016 | frag_t subfrag = inode->pick_dirfrag(p.first.name); | |
1017 | int n = (subfrag.value() & (subfrag.mask() ^ frag.mask())) >> subfrag.mask_shift(); | |
1018 | CDir *f = subfrags[n]; | |
1019 | ||
1020 | if (f->waiting_on_dentry.empty()) | |
1021 | f->get(PIN_DNWAITER); | |
1022 | f->waiting_on_dentry[p.first].swap(p.second); | |
1023 | } | |
1024 | ||
1025 | // FIXME: handle dirty old rstat | |
1026 | ||
1027 | // fix up new frag fragstats | |
1028 | for (int i=0; i<n; i++) { | |
1029 | CDir *f = subfrags[i]; | |
1030 | f->fnode.rstat.version = rstat_version; | |
1031 | f->fnode.accounted_rstat = f->fnode.rstat; | |
1032 | f->fnode.fragstat.version = dirstat_version; | |
1033 | f->fnode.accounted_fragstat = f->fnode.fragstat; | |
1034 | dout(10) << " rstat " << f->fnode.rstat << " fragstat " << f->fnode.fragstat | |
1035 | << " on " << *f << dendl; | |
1036 | } | |
1037 | ||
1038 | // give any outstanding frag stat differential to first frag | |
1039 | dout(10) << " giving rstatdiff " << rstatdiff << " fragstatdiff" << fragstatdiff | |
1040 | << " to " << *subfrags[0] << dendl; | |
1041 | subfrags[0]->fnode.accounted_rstat.add(rstatdiff); | |
1042 | subfrags[0]->fnode.accounted_fragstat.add(fragstatdiff); | |
1043 | ||
1044 | finish_old_fragment(waiters, replay); | |
1045 | } | |
1046 | ||
1047 | void CDir::merge(list<CDir*>& subs, list<MDSInternalContextBase*>& waiters, bool replay) | |
1048 | { | |
1049 | dout(10) << "merge " << subs << dendl; | |
1050 | ||
1051 | mds_authority_t new_auth = CDIR_AUTH_DEFAULT; | |
1052 | for (auto dir : subs) { | |
1053 | if (dir->get_dir_auth() != CDIR_AUTH_DEFAULT && | |
1054 | dir->get_dir_auth() != new_auth) { | |
1055 | assert(new_auth == CDIR_AUTH_DEFAULT); | |
1056 | new_auth = dir->get_dir_auth(); | |
1057 | } | |
1058 | } | |
1059 | ||
1060 | set_dir_auth(new_auth); | |
1061 | prepare_new_fragment(replay); | |
1062 | ||
1063 | nest_info_t rstatdiff; | |
1064 | frag_info_t fragstatdiff; | |
1065 | bool touched_mtime, touched_chattr; | |
1066 | version_t rstat_version = inode->get_projected_inode()->rstat.version; | |
1067 | version_t dirstat_version = inode->get_projected_inode()->dirstat.version; | |
1068 | ||
1069 | map<string_snap_t, std::list<MDSInternalContextBase*> > dentry_waiters; | |
1070 | ||
1071 | for (auto dir : subs) { | |
1072 | dout(10) << " subfrag " << dir->get_frag() << " " << *dir << dendl; | |
1073 | assert(!dir->is_auth() || dir->is_complete() || replay); | |
1074 | ||
1075 | if (dir->fnode.accounted_rstat.version == rstat_version) | |
1076 | rstatdiff.add_delta(dir->fnode.accounted_rstat, dir->fnode.rstat); | |
1077 | if (dir->fnode.accounted_fragstat.version == dirstat_version) | |
1078 | fragstatdiff.add_delta(dir->fnode.accounted_fragstat, dir->fnode.fragstat, | |
1079 | &touched_mtime, &touched_chattr); | |
1080 | ||
1081 | dir->prepare_old_fragment(dentry_waiters, replay); | |
1082 | ||
1083 | // steal dentries | |
1084 | while (!dir->items.empty()) | |
1085 | steal_dentry(dir->items.begin()->second); | |
1086 | ||
1087 | // merge replica map | |
1088 | for (compact_map<mds_rank_t,unsigned>::iterator p = dir->replicas_begin(); | |
1089 | p != dir->replicas_end(); | |
1090 | ++p) { | |
1091 | unsigned cur = replica_map[p->first]; | |
1092 | if (p->second > cur) | |
1093 | replica_map[p->first] = p->second; | |
1094 | } | |
1095 | ||
1096 | // merge version | |
1097 | if (dir->get_version() > get_version()) | |
1098 | set_version(dir->get_version()); | |
1099 | ||
1100 | // merge state | |
1101 | state_set(dir->get_state() & MASK_STATE_FRAGMENT_KEPT); | |
1102 | ||
1103 | dir->finish_old_fragment(waiters, replay); | |
1104 | inode->close_dirfrag(dir->get_frag()); | |
1105 | } | |
1106 | ||
1107 | if (!dentry_waiters.empty()) { | |
1108 | get(PIN_DNWAITER); | |
1109 | for (auto& p : dentry_waiters) { | |
1110 | waiting_on_dentry[p.first].swap(p.second); | |
1111 | } | |
1112 | } | |
1113 | ||
1114 | if (is_auth() && !replay) | |
1115 | mark_complete(); | |
1116 | ||
1117 | // FIXME: merge dirty old rstat | |
1118 | fnode.rstat.version = rstat_version; | |
1119 | fnode.accounted_rstat = fnode.rstat; | |
1120 | fnode.accounted_rstat.add(rstatdiff); | |
1121 | ||
1122 | fnode.fragstat.version = dirstat_version; | |
1123 | fnode.accounted_fragstat = fnode.fragstat; | |
1124 | fnode.accounted_fragstat.add(fragstatdiff); | |
1125 | ||
1126 | init_fragment_pins(); | |
1127 | } | |
1128 | ||
1129 | ||
1130 | ||
1131 | ||
1132 | void CDir::resync_accounted_fragstat() | |
1133 | { | |
1134 | fnode_t *pf = get_projected_fnode(); | |
1135 | inode_t *pi = inode->get_projected_inode(); | |
1136 | ||
1137 | if (pf->accounted_fragstat.version != pi->dirstat.version) { | |
1138 | pf->fragstat.version = pi->dirstat.version; | |
1139 | dout(10) << "resync_accounted_fragstat " << pf->accounted_fragstat << " -> " << pf->fragstat << dendl; | |
1140 | pf->accounted_fragstat = pf->fragstat; | |
1141 | } | |
1142 | } | |
1143 | ||
1144 | /* | |
1145 | * resync rstat and accounted_rstat with inode | |
1146 | */ | |
1147 | void CDir::resync_accounted_rstat() | |
1148 | { | |
1149 | fnode_t *pf = get_projected_fnode(); | |
1150 | inode_t *pi = inode->get_projected_inode(); | |
1151 | ||
1152 | if (pf->accounted_rstat.version != pi->rstat.version) { | |
1153 | pf->rstat.version = pi->rstat.version; | |
1154 | dout(10) << "resync_accounted_rstat " << pf->accounted_rstat << " -> " << pf->rstat << dendl; | |
1155 | pf->accounted_rstat = pf->rstat; | |
1156 | dirty_old_rstat.clear(); | |
1157 | } | |
1158 | } | |
1159 | ||
1160 | void CDir::assimilate_dirty_rstat_inodes() | |
1161 | { | |
1162 | dout(10) << "assimilate_dirty_rstat_inodes" << dendl; | |
1163 | for (elist<CInode*>::iterator p = dirty_rstat_inodes.begin_use_current(); | |
1164 | !p.end(); ++p) { | |
1165 | CInode *in = *p; | |
1166 | assert(in->is_auth()); | |
1167 | if (in->is_frozen()) | |
1168 | continue; | |
1169 | ||
1170 | inode_t *pi = in->project_inode(); | |
1171 | pi->version = in->pre_dirty(); | |
1172 | ||
1173 | inode->mdcache->project_rstat_inode_to_frag(in, this, 0, 0, NULL); | |
1174 | } | |
1175 | state_set(STATE_ASSIMRSTAT); | |
1176 | dout(10) << "assimilate_dirty_rstat_inodes done" << dendl; | |
1177 | } | |
1178 | ||
1179 | void CDir::assimilate_dirty_rstat_inodes_finish(MutationRef& mut, EMetaBlob *blob) | |
1180 | { | |
1181 | if (!state_test(STATE_ASSIMRSTAT)) | |
1182 | return; | |
1183 | state_clear(STATE_ASSIMRSTAT); | |
1184 | dout(10) << "assimilate_dirty_rstat_inodes_finish" << dendl; | |
1185 | elist<CInode*>::iterator p = dirty_rstat_inodes.begin_use_current(); | |
1186 | while (!p.end()) { | |
1187 | CInode *in = *p; | |
1188 | ++p; | |
1189 | ||
1190 | if (in->is_frozen()) | |
1191 | continue; | |
1192 | ||
1193 | CDentry *dn = in->get_projected_parent_dn(); | |
1194 | ||
1195 | mut->auth_pin(in); | |
1196 | mut->add_projected_inode(in); | |
1197 | ||
1198 | in->clear_dirty_rstat(); | |
1199 | blob->add_primary_dentry(dn, in, true); | |
1200 | } | |
1201 | ||
1202 | if (!dirty_rstat_inodes.empty()) | |
1203 | inode->mdcache->mds->locker->mark_updated_scatterlock(&inode->nestlock); | |
1204 | } | |
1205 | ||
1206 | ||
1207 | ||
1208 | ||
1209 | /**************************************** | |
1210 | * WAITING | |
1211 | */ | |
1212 | ||
1213 | void CDir::add_dentry_waiter(const string& dname, snapid_t snapid, MDSInternalContextBase *c) | |
1214 | { | |
1215 | if (waiting_on_dentry.empty()) | |
1216 | get(PIN_DNWAITER); | |
1217 | waiting_on_dentry[string_snap_t(dname, snapid)].push_back(c); | |
1218 | dout(10) << "add_dentry_waiter dentry " << dname | |
1219 | << " snap " << snapid | |
1220 | << " " << c << " on " << *this << dendl; | |
1221 | } | |
1222 | ||
1223 | void CDir::take_dentry_waiting(const string& dname, snapid_t first, snapid_t last, | |
1224 | list<MDSInternalContextBase*>& ls) | |
1225 | { | |
1226 | if (waiting_on_dentry.empty()) | |
1227 | return; | |
1228 | ||
1229 | string_snap_t lb(dname, first); | |
1230 | string_snap_t ub(dname, last); | |
1231 | compact_map<string_snap_t, list<MDSInternalContextBase*> >::iterator p = waiting_on_dentry.lower_bound(lb); | |
1232 | while (p != waiting_on_dentry.end() && | |
1233 | !(ub < p->first)) { | |
1234 | dout(10) << "take_dentry_waiting dentry " << dname | |
1235 | << " [" << first << "," << last << "] found waiter on snap " | |
1236 | << p->first.snapid | |
1237 | << " on " << *this << dendl; | |
1238 | ls.splice(ls.end(), p->second); | |
1239 | waiting_on_dentry.erase(p++); | |
1240 | } | |
1241 | ||
1242 | if (waiting_on_dentry.empty()) | |
1243 | put(PIN_DNWAITER); | |
1244 | } | |
1245 | ||
1246 | void CDir::take_sub_waiting(list<MDSInternalContextBase*>& ls) | |
1247 | { | |
1248 | dout(10) << "take_sub_waiting" << dendl; | |
1249 | if (!waiting_on_dentry.empty()) { | |
1250 | for (compact_map<string_snap_t, list<MDSInternalContextBase*> >::iterator p = waiting_on_dentry.begin(); | |
1251 | p != waiting_on_dentry.end(); | |
1252 | ++p) | |
1253 | ls.splice(ls.end(), p->second); | |
1254 | waiting_on_dentry.clear(); | |
1255 | put(PIN_DNWAITER); | |
1256 | } | |
1257 | } | |
1258 | ||
1259 | ||
1260 | ||
1261 | void CDir::add_waiter(uint64_t tag, MDSInternalContextBase *c) | |
1262 | { | |
1263 | // hierarchical? | |
1264 | ||
1265 | // at free root? | |
1266 | if (tag & WAIT_ATFREEZEROOT) { | |
1267 | if (!(is_freezing_tree_root() || is_frozen_tree_root() || | |
1268 | is_freezing_dir() || is_frozen_dir())) { | |
1269 | // try parent | |
1270 | dout(10) << "add_waiter " << std::hex << tag << std::dec << " " << c << " should be ATFREEZEROOT, " << *this << " is not root, trying parent" << dendl; | |
1271 | inode->parent->dir->add_waiter(tag, c); | |
1272 | return; | |
1273 | } | |
1274 | } | |
1275 | ||
1276 | // at subtree root? | |
1277 | if (tag & WAIT_ATSUBTREEROOT) { | |
1278 | if (!is_subtree_root()) { | |
1279 | // try parent | |
1280 | dout(10) << "add_waiter " << std::hex << tag << std::dec << " " << c << " should be ATSUBTREEROOT, " << *this << " is not root, trying parent" << dendl; | |
1281 | inode->parent->dir->add_waiter(tag, c); | |
1282 | return; | |
1283 | } | |
1284 | } | |
1285 | ||
1286 | assert(!(tag & WAIT_CREATED) || state_test(STATE_CREATING)); | |
1287 | ||
1288 | MDSCacheObject::add_waiter(tag, c); | |
1289 | } | |
1290 | ||
1291 | ||
1292 | ||
1293 | /* NOTE: this checks dentry waiters too */ | |
1294 | void CDir::take_waiting(uint64_t mask, list<MDSInternalContextBase*>& ls) | |
1295 | { | |
1296 | if ((mask & WAIT_DENTRY) && !waiting_on_dentry.empty()) { | |
1297 | // take all dentry waiters | |
1298 | while (!waiting_on_dentry.empty()) { | |
1299 | compact_map<string_snap_t, list<MDSInternalContextBase*> >::iterator p = waiting_on_dentry.begin(); | |
1300 | dout(10) << "take_waiting dentry " << p->first.name | |
1301 | << " snap " << p->first.snapid << " on " << *this << dendl; | |
1302 | ls.splice(ls.end(), p->second); | |
1303 | waiting_on_dentry.erase(p); | |
1304 | } | |
1305 | put(PIN_DNWAITER); | |
1306 | } | |
1307 | ||
1308 | // waiting | |
1309 | MDSCacheObject::take_waiting(mask, ls); | |
1310 | } | |
1311 | ||
1312 | ||
1313 | void CDir::finish_waiting(uint64_t mask, int result) | |
1314 | { | |
1315 | dout(11) << "finish_waiting mask " << hex << mask << dec << " result " << result << " on " << *this << dendl; | |
1316 | ||
1317 | list<MDSInternalContextBase*> finished; | |
1318 | take_waiting(mask, finished); | |
1319 | if (result < 0) | |
1320 | finish_contexts(g_ceph_context, finished, result); | |
1321 | else | |
1322 | cache->mds->queue_waiters(finished); | |
1323 | } | |
1324 | ||
1325 | ||
1326 | ||
1327 | // dirty/clean | |
1328 | ||
1329 | fnode_t *CDir::project_fnode() | |
1330 | { | |
1331 | assert(get_version() != 0); | |
1332 | fnode_t *p = new fnode_t; | |
1333 | *p = *get_projected_fnode(); | |
1334 | projected_fnode.push_back(p); | |
1335 | ||
1336 | if (scrub_infop && scrub_infop->last_scrub_dirty) { | |
1337 | p->localized_scrub_stamp = scrub_infop->last_local.time; | |
1338 | p->localized_scrub_version = scrub_infop->last_local.version; | |
1339 | p->recursive_scrub_stamp = scrub_infop->last_recursive.time; | |
1340 | p->recursive_scrub_version = scrub_infop->last_recursive.version; | |
1341 | scrub_infop->last_scrub_dirty = false; | |
1342 | scrub_maybe_delete_info(); | |
1343 | } | |
1344 | ||
1345 | dout(10) << "project_fnode " << p << dendl; | |
1346 | return p; | |
1347 | } | |
1348 | ||
1349 | void CDir::pop_and_dirty_projected_fnode(LogSegment *ls) | |
1350 | { | |
1351 | assert(!projected_fnode.empty()); | |
1352 | dout(15) << "pop_and_dirty_projected_fnode " << projected_fnode.front() | |
1353 | << " v" << projected_fnode.front()->version << dendl; | |
1354 | fnode = *projected_fnode.front(); | |
1355 | _mark_dirty(ls); | |
1356 | delete projected_fnode.front(); | |
1357 | projected_fnode.pop_front(); | |
1358 | } | |
1359 | ||
1360 | ||
1361 | version_t CDir::pre_dirty(version_t min) | |
1362 | { | |
1363 | if (min > projected_version) | |
1364 | projected_version = min; | |
1365 | ++projected_version; | |
1366 | dout(10) << "pre_dirty " << projected_version << dendl; | |
1367 | return projected_version; | |
1368 | } | |
1369 | ||
1370 | void CDir::mark_dirty(version_t pv, LogSegment *ls) | |
1371 | { | |
1372 | assert(get_version() < pv); | |
1373 | assert(pv <= projected_version); | |
1374 | fnode.version = pv; | |
1375 | _mark_dirty(ls); | |
1376 | } | |
1377 | ||
1378 | void CDir::_mark_dirty(LogSegment *ls) | |
1379 | { | |
1380 | if (!state_test(STATE_DIRTY)) { | |
1381 | dout(10) << "mark_dirty (was clean) " << *this << " version " << get_version() << dendl; | |
1382 | _set_dirty_flag(); | |
1383 | assert(ls); | |
1384 | } else { | |
1385 | dout(10) << "mark_dirty (already dirty) " << *this << " version " << get_version() << dendl; | |
1386 | } | |
1387 | if (ls) { | |
1388 | ls->dirty_dirfrags.push_back(&item_dirty); | |
1389 | ||
1390 | // if i've never committed, i need to be before _any_ mention of me is trimmed from the journal. | |
1391 | if (committed_version == 0 && !item_new.is_on_list()) | |
1392 | ls->new_dirfrags.push_back(&item_new); | |
1393 | } | |
1394 | } | |
1395 | ||
1396 | void CDir::mark_new(LogSegment *ls) | |
1397 | { | |
1398 | ls->new_dirfrags.push_back(&item_new); | |
1399 | state_clear(STATE_CREATING); | |
1400 | ||
1401 | list<MDSInternalContextBase*> waiters; | |
1402 | take_waiting(CDir::WAIT_CREATED, waiters); | |
1403 | cache->mds->queue_waiters(waiters); | |
1404 | } | |
1405 | ||
1406 | void CDir::mark_clean() | |
1407 | { | |
1408 | dout(10) << "mark_clean " << *this << " version " << get_version() << dendl; | |
1409 | if (state_test(STATE_DIRTY)) { | |
1410 | item_dirty.remove_myself(); | |
1411 | item_new.remove_myself(); | |
1412 | ||
1413 | state_clear(STATE_DIRTY); | |
1414 | put(PIN_DIRTY); | |
1415 | } | |
1416 | } | |
1417 | ||
1418 | // caller should hold auth pin of this | |
1419 | void CDir::log_mark_dirty() | |
1420 | { | |
1421 | if (is_dirty() || is_projected()) | |
1422 | return; // noop if it is already dirty or will be dirty | |
1423 | ||
1424 | version_t pv = pre_dirty(); | |
1425 | mark_dirty(pv, cache->mds->mdlog->get_current_segment()); | |
1426 | } | |
1427 | ||
1428 | void CDir::mark_complete() { | |
1429 | state_set(STATE_COMPLETE); | |
1430 | bloom.reset(); | |
1431 | } | |
1432 | ||
1433 | void CDir::first_get() | |
1434 | { | |
1435 | inode->get(CInode::PIN_DIRFRAG); | |
1436 | } | |
1437 | ||
1438 | void CDir::last_put() | |
1439 | { | |
1440 | inode->put(CInode::PIN_DIRFRAG); | |
1441 | } | |
1442 | ||
1443 | ||
1444 | ||
1445 | /****************************************************************************** | |
1446 | * FETCH and COMMIT | |
1447 | */ | |
1448 | ||
1449 | // ----------------------- | |
1450 | // FETCH | |
1451 | void CDir::fetch(MDSInternalContextBase *c, bool ignore_authpinnability) | |
1452 | { | |
1453 | string want; | |
1454 | return fetch(c, want, ignore_authpinnability); | |
1455 | } | |
1456 | ||
1457 | void CDir::fetch(MDSInternalContextBase *c, const string& want_dn, bool ignore_authpinnability) | |
1458 | { | |
1459 | dout(10) << "fetch on " << *this << dendl; | |
1460 | ||
1461 | assert(is_auth()); | |
1462 | assert(!is_complete()); | |
1463 | ||
1464 | if (!can_auth_pin() && !ignore_authpinnability) { | |
1465 | if (c) { | |
1466 | dout(7) << "fetch waiting for authpinnable" << dendl; | |
1467 | add_waiter(WAIT_UNFREEZE, c); | |
1468 | } else | |
1469 | dout(7) << "fetch not authpinnable and no context" << dendl; | |
1470 | return; | |
1471 | } | |
1472 | ||
1473 | // unlinked directory inode shouldn't have any entry | |
1474 | if (!inode->is_base() && get_parent_dir()->inode->is_stray() && | |
1475 | !inode->snaprealm) { | |
1476 | dout(7) << "fetch dirfrag for unlinked directory, mark complete" << dendl; | |
1477 | if (get_version() == 0) { | |
1478 | assert(inode->is_auth()); | |
1479 | set_version(1); | |
1480 | ||
1481 | if (state_test(STATE_REJOINUNDEF)) { | |
1482 | assert(cache->mds->is_rejoin()); | |
1483 | state_clear(STATE_REJOINUNDEF); | |
1484 | cache->opened_undef_dirfrag(this); | |
1485 | } | |
1486 | } | |
1487 | mark_complete(); | |
1488 | ||
1489 | if (c) | |
1490 | cache->mds->queue_waiter(c); | |
1491 | return; | |
1492 | } | |
1493 | ||
1494 | if (c) add_waiter(WAIT_COMPLETE, c); | |
1495 | if (!want_dn.empty()) wanted_items.insert(want_dn); | |
1496 | ||
1497 | // already fetching? | |
1498 | if (state_test(CDir::STATE_FETCHING)) { | |
1499 | dout(7) << "already fetching; waiting" << dendl; | |
1500 | return; | |
1501 | } | |
1502 | ||
1503 | auth_pin(this); | |
1504 | state_set(CDir::STATE_FETCHING); | |
1505 | ||
1506 | if (cache->mds->logger) cache->mds->logger->inc(l_mds_dir_fetch); | |
1507 | ||
1508 | std::set<dentry_key_t> empty; | |
1509 | _omap_fetch(NULL, empty); | |
1510 | } | |
1511 | ||
1512 | void CDir::fetch(MDSInternalContextBase *c, const std::set<dentry_key_t>& keys) | |
1513 | { | |
1514 | dout(10) << "fetch " << keys.size() << " keys on " << *this << dendl; | |
1515 | ||
1516 | assert(is_auth()); | |
1517 | assert(!is_complete()); | |
1518 | ||
1519 | if (!can_auth_pin()) { | |
1520 | dout(7) << "fetch keys waiting for authpinnable" << dendl; | |
1521 | add_waiter(WAIT_UNFREEZE, c); | |
1522 | return; | |
1523 | } | |
1524 | if (state_test(CDir::STATE_FETCHING)) { | |
1525 | dout(7) << "fetch keys waiting for full fetch" << dendl; | |
1526 | add_waiter(WAIT_COMPLETE, c); | |
1527 | return; | |
1528 | } | |
1529 | ||
1530 | auth_pin(this); | |
1531 | if (cache->mds->logger) cache->mds->logger->inc(l_mds_dir_fetch); | |
1532 | ||
1533 | _omap_fetch(c, keys); | |
1534 | } | |
1535 | ||
1536 | class C_IO_Dir_OMAP_FetchedMore : public CDirIOContext { | |
1537 | MDSInternalContextBase *fin; | |
1538 | public: | |
1539 | bufferlist hdrbl; | |
1540 | bool more = false; | |
1541 | map<string, bufferlist> omap; ///< carry-over from before | |
1542 | map<string, bufferlist> omap_more; ///< new batch | |
1543 | int ret; | |
1544 | C_IO_Dir_OMAP_FetchedMore(CDir *d, MDSInternalContextBase *f) : | |
1545 | CDirIOContext(d), fin(f), ret(0) { } | |
1546 | void finish(int r) { | |
1547 | // merge results | |
1548 | if (omap.empty()) { | |
1549 | omap.swap(omap_more); | |
1550 | } else { | |
1551 | omap.insert(omap_more.begin(), omap_more.end()); | |
1552 | } | |
1553 | if (more) { | |
1554 | dir->_omap_fetch_more(hdrbl, omap, fin); | |
1555 | } else { | |
1556 | dir->_omap_fetched(hdrbl, omap, !fin, r); | |
1557 | if (fin) | |
1558 | fin->complete(r); | |
1559 | } | |
1560 | } | |
1561 | }; | |
1562 | ||
1563 | class C_IO_Dir_OMAP_Fetched : public CDirIOContext { | |
1564 | MDSInternalContextBase *fin; | |
1565 | public: | |
1566 | bufferlist hdrbl; | |
1567 | bool more = false; | |
1568 | map<string, bufferlist> omap; | |
1569 | bufferlist btbl; | |
1570 | int ret1, ret2, ret3; | |
1571 | ||
1572 | C_IO_Dir_OMAP_Fetched(CDir *d, MDSInternalContextBase *f) : | |
1573 | CDirIOContext(d), fin(f), ret1(0), ret2(0), ret3(0) { } | |
1574 | void finish(int r) override { | |
1575 | // check the correctness of backtrace | |
1576 | if (r >= 0 && ret3 != -ECANCELED) | |
1577 | dir->inode->verify_diri_backtrace(btbl, ret3); | |
1578 | if (r >= 0) r = ret1; | |
1579 | if (r >= 0) r = ret2; | |
1580 | if (more) { | |
1581 | dir->_omap_fetch_more(hdrbl, omap, fin); | |
1582 | } else { | |
1583 | dir->_omap_fetched(hdrbl, omap, !fin, r); | |
1584 | if (fin) | |
1585 | fin->complete(r); | |
1586 | } | |
1587 | } | |
1588 | }; | |
1589 | ||
1590 | void CDir::_omap_fetch(MDSInternalContextBase *c, const std::set<dentry_key_t>& keys) | |
1591 | { | |
1592 | C_IO_Dir_OMAP_Fetched *fin = new C_IO_Dir_OMAP_Fetched(this, c); | |
1593 | object_t oid = get_ondisk_object(); | |
1594 | object_locator_t oloc(cache->mds->mdsmap->get_metadata_pool()); | |
1595 | ObjectOperation rd; | |
1596 | rd.omap_get_header(&fin->hdrbl, &fin->ret1); | |
1597 | if (keys.empty()) { | |
1598 | assert(!c); | |
1599 | rd.omap_get_vals("", "", g_conf->mds_dir_keys_per_op, | |
1600 | &fin->omap, &fin->more, &fin->ret2); | |
1601 | } else { | |
1602 | assert(c); | |
1603 | std::set<std::string> str_keys; | |
1604 | for (auto p = keys.begin(); p != keys.end(); ++p) { | |
1605 | string str; | |
1606 | p->encode(str); | |
1607 | str_keys.insert(str); | |
1608 | } | |
1609 | rd.omap_get_vals_by_keys(str_keys, &fin->omap, &fin->ret2); | |
1610 | } | |
1611 | // check the correctness of backtrace | |
1612 | if (g_conf->mds_verify_backtrace > 0 && frag == frag_t()) { | |
1613 | rd.getxattr("parent", &fin->btbl, &fin->ret3); | |
1614 | rd.set_last_op_flags(CEPH_OSD_OP_FLAG_FAILOK); | |
1615 | } else { | |
1616 | fin->ret3 = -ECANCELED; | |
1617 | } | |
1618 | ||
1619 | cache->mds->objecter->read(oid, oloc, rd, CEPH_NOSNAP, NULL, 0, | |
1620 | new C_OnFinisher(fin, cache->mds->finisher)); | |
1621 | } | |
1622 | ||
1623 | void CDir::_omap_fetch_more( | |
1624 | bufferlist& hdrbl, | |
1625 | map<string, bufferlist>& omap, | |
1626 | MDSInternalContextBase *c) | |
1627 | { | |
1628 | // we have more omap keys to fetch! | |
1629 | object_t oid = get_ondisk_object(); | |
1630 | object_locator_t oloc(cache->mds->mdsmap->get_metadata_pool()); | |
1631 | C_IO_Dir_OMAP_FetchedMore *fin = new C_IO_Dir_OMAP_FetchedMore(this, c); | |
1632 | fin->hdrbl.claim(hdrbl); | |
1633 | fin->omap.swap(omap); | |
1634 | ObjectOperation rd; | |
1635 | rd.omap_get_vals(fin->omap.rbegin()->first, | |
1636 | "", /* filter prefix */ | |
1637 | g_conf->mds_dir_keys_per_op, | |
1638 | &fin->omap_more, | |
1639 | &fin->more, | |
1640 | &fin->ret); | |
1641 | cache->mds->objecter->read(oid, oloc, rd, CEPH_NOSNAP, NULL, 0, | |
1642 | new C_OnFinisher(fin, cache->mds->finisher)); | |
1643 | } | |
1644 | ||
1645 | CDentry *CDir::_load_dentry( | |
1646 | const std::string &key, | |
1647 | const std::string &dname, | |
1648 | const snapid_t last, | |
1649 | bufferlist &bl, | |
1650 | const int pos, | |
1651 | const std::set<snapid_t> *snaps, | |
1652 | bool *force_dirty, | |
1653 | list<CInode*> *undef_inodes) | |
1654 | { | |
1655 | bufferlist::iterator q = bl.begin(); | |
1656 | ||
1657 | snapid_t first; | |
1658 | ::decode(first, q); | |
1659 | ||
1660 | // marker | |
1661 | char type; | |
1662 | ::decode(type, q); | |
1663 | ||
1664 | dout(20) << "_fetched pos " << pos << " marker '" << type << "' dname '" << dname | |
1665 | << " [" << first << "," << last << "]" | |
1666 | << dendl; | |
1667 | ||
1668 | bool stale = false; | |
1669 | if (snaps && last != CEPH_NOSNAP) { | |
1670 | set<snapid_t>::const_iterator p = snaps->lower_bound(first); | |
1671 | if (p == snaps->end() || *p > last) { | |
1672 | dout(10) << " skipping stale dentry on [" << first << "," << last << "]" << dendl; | |
1673 | stale = true; | |
1674 | } | |
1675 | } | |
1676 | ||
1677 | /* | |
1678 | * look for existing dentry for _last_ snap, because unlink + | |
1679 | * create may leave a "hole" (epochs during which the dentry | |
1680 | * doesn't exist) but for which no explicit negative dentry is in | |
1681 | * the cache. | |
1682 | */ | |
1683 | CDentry *dn; | |
1684 | if (stale) | |
1685 | dn = lookup_exact_snap(dname, last); | |
1686 | else | |
1687 | dn = lookup(dname, last); | |
1688 | ||
1689 | if (type == 'L') { | |
1690 | // hard link | |
1691 | inodeno_t ino; | |
1692 | unsigned char d_type; | |
1693 | ::decode(ino, q); | |
1694 | ::decode(d_type, q); | |
1695 | ||
1696 | if (stale) { | |
1697 | if (!dn) { | |
1698 | stale_items.insert(key); | |
1699 | *force_dirty = true; | |
1700 | } | |
1701 | return dn; | |
1702 | } | |
1703 | ||
1704 | if (dn) { | |
1705 | if (dn->get_linkage()->get_inode() == 0) { | |
1706 | dout(12) << "_fetched had NEG dentry " << *dn << dendl; | |
1707 | } else { | |
1708 | dout(12) << "_fetched had dentry " << *dn << dendl; | |
1709 | } | |
1710 | } else { | |
1711 | // (remote) link | |
1712 | dn = add_remote_dentry(dname, ino, d_type, first, last); | |
1713 | ||
1714 | // link to inode? | |
1715 | CInode *in = cache->get_inode(ino); // we may or may not have it. | |
1716 | if (in) { | |
1717 | dn->link_remote(dn->get_linkage(), in); | |
1718 | dout(12) << "_fetched got remote link " << ino << " which we have " << *in << dendl; | |
1719 | } else { | |
1720 | dout(12) << "_fetched got remote link " << ino << " (dont' have it)" << dendl; | |
1721 | } | |
1722 | } | |
1723 | } | |
1724 | else if (type == 'I') { | |
1725 | // inode | |
1726 | ||
1727 | // Load inode data before looking up or constructing CInode | |
1728 | InodeStore inode_data; | |
1729 | inode_data.decode_bare(q); | |
1730 | ||
1731 | if (stale) { | |
1732 | if (!dn) { | |
1733 | stale_items.insert(key); | |
1734 | *force_dirty = true; | |
1735 | } | |
1736 | return dn; | |
1737 | } | |
1738 | ||
1739 | bool undef_inode = false; | |
1740 | if (dn) { | |
1741 | CInode *in = dn->get_linkage()->get_inode(); | |
1742 | if (in) { | |
1743 | dout(12) << "_fetched had dentry " << *dn << dendl; | |
1744 | if (in->state_test(CInode::STATE_REJOINUNDEF)) { | |
1745 | undef_inodes->push_back(in); | |
1746 | undef_inode = true; | |
1747 | } | |
1748 | } else | |
1749 | dout(12) << "_fetched had NEG dentry " << *dn << dendl; | |
1750 | } | |
1751 | ||
1752 | if (!dn || undef_inode) { | |
1753 | // add inode | |
1754 | CInode *in = cache->get_inode(inode_data.inode.ino, last); | |
1755 | if (!in || undef_inode) { | |
1756 | if (undef_inode && in) | |
1757 | in->first = first; | |
1758 | else | |
1759 | in = new CInode(cache, true, first, last); | |
1760 | ||
1761 | in->inode = inode_data.inode; | |
1762 | // symlink? | |
1763 | if (in->is_symlink()) | |
1764 | in->symlink = inode_data.symlink; | |
1765 | ||
1766 | in->dirfragtree.swap(inode_data.dirfragtree); | |
1767 | in->xattrs.swap(inode_data.xattrs); | |
1768 | in->old_inodes.swap(inode_data.old_inodes); | |
1769 | if (!in->old_inodes.empty()) { | |
1770 | snapid_t min_first = in->old_inodes.rbegin()->first + 1; | |
1771 | if (min_first > in->first) | |
1772 | in->first = min_first; | |
1773 | } | |
1774 | ||
1775 | in->oldest_snap = inode_data.oldest_snap; | |
1776 | in->decode_snap_blob(inode_data.snap_blob); | |
1777 | if (snaps && !in->snaprealm) | |
1778 | in->purge_stale_snap_data(*snaps); | |
1779 | ||
1780 | if (!undef_inode) { | |
1781 | cache->add_inode(in); // add | |
1782 | dn = add_primary_dentry(dname, in, first, last); // link | |
1783 | } | |
1784 | dout(12) << "_fetched got " << *dn << " " << *in << dendl; | |
1785 | ||
1786 | if (in->inode.is_dirty_rstat()) | |
1787 | in->mark_dirty_rstat(); | |
1788 | ||
1789 | //in->hack_accessed = false; | |
1790 | //in->hack_load_stamp = ceph_clock_now(); | |
1791 | //num_new_inodes_loaded++; | |
1792 | } else { | |
1793 | dout(0) << "_fetched badness: got (but i already had) " << *in | |
1794 | << " mode " << in->inode.mode | |
1795 | << " mtime " << in->inode.mtime << dendl; | |
1796 | string dirpath, inopath; | |
1797 | this->inode->make_path_string(dirpath); | |
1798 | in->make_path_string(inopath); | |
1799 | cache->mds->clog->error() << "loaded dup inode " << inode_data.inode.ino | |
1800 | << " [" << first << "," << last << "] v" << inode_data.inode.version | |
1801 | << " at " << dirpath << "/" << dname | |
1802 | << ", but inode " << in->vino() << " v" << in->inode.version | |
1803 | << " already exists at " << inopath; | |
1804 | return dn; | |
1805 | } | |
1806 | } | |
1807 | } else { | |
1808 | std::ostringstream oss; | |
1809 | oss << "Invalid tag char '" << type << "' pos " << pos; | |
1810 | throw buffer::malformed_input(oss.str()); | |
1811 | } | |
1812 | ||
1813 | return dn; | |
1814 | } | |
1815 | ||
1816 | void CDir::_omap_fetched(bufferlist& hdrbl, map<string, bufferlist>& omap, | |
1817 | bool complete, int r) | |
1818 | { | |
1819 | LogChannelRef clog = cache->mds->clog; | |
1820 | dout(10) << "_fetched header " << hdrbl.length() << " bytes " | |
1821 | << omap.size() << " keys for " << *this << dendl; | |
1822 | ||
1823 | assert(r == 0 || r == -ENOENT || r == -ENODATA); | |
1824 | assert(is_auth()); | |
1825 | assert(!is_frozen()); | |
1826 | ||
1827 | if (hdrbl.length() == 0) { | |
1828 | dout(0) << "_fetched missing object for " << *this << dendl; | |
1829 | ||
1830 | clog->error() << "dir " << dirfrag() << " object missing on disk; some " | |
1831 | "files may be lost (" << get_path() << ")"; | |
1832 | ||
1833 | go_bad(complete); | |
1834 | return; | |
1835 | } | |
1836 | ||
1837 | fnode_t got_fnode; | |
1838 | { | |
1839 | bufferlist::iterator p = hdrbl.begin(); | |
1840 | try { | |
1841 | ::decode(got_fnode, p); | |
1842 | } catch (const buffer::error &err) { | |
1843 | derr << "Corrupt fnode in dirfrag " << dirfrag() | |
1844 | << ": " << err << dendl; | |
1845 | clog->warn() << "Corrupt fnode header in " << dirfrag() << ": " | |
1846 | << err << " (" << get_path() << ")"; | |
1847 | go_bad(complete); | |
1848 | return; | |
1849 | } | |
1850 | if (!p.end()) { | |
1851 | clog->warn() << "header buffer of dir " << dirfrag() << " has " | |
1852 | << hdrbl.length() - p.get_off() << " extra bytes (" | |
1853 | << get_path() << ")"; | |
1854 | go_bad(complete); | |
1855 | return; | |
1856 | } | |
1857 | } | |
1858 | ||
1859 | dout(10) << "_fetched version " << got_fnode.version << dendl; | |
1860 | ||
1861 | // take the loaded fnode? | |
1862 | // only if we are a fresh CDir* with no prior state. | |
1863 | if (get_version() == 0) { | |
1864 | assert(!is_projected()); | |
1865 | assert(!state_test(STATE_COMMITTING)); | |
1866 | fnode = got_fnode; | |
1867 | projected_version = committing_version = committed_version = got_fnode.version; | |
1868 | ||
1869 | if (state_test(STATE_REJOINUNDEF)) { | |
1870 | assert(cache->mds->is_rejoin()); | |
1871 | state_clear(STATE_REJOINUNDEF); | |
1872 | cache->opened_undef_dirfrag(this); | |
1873 | } | |
1874 | } | |
1875 | ||
1876 | list<CInode*> undef_inodes; | |
1877 | ||
1878 | // purge stale snaps? | |
1879 | // only if we have past_parents open! | |
1880 | bool force_dirty = false; | |
1881 | const set<snapid_t> *snaps = NULL; | |
1882 | SnapRealm *realm = inode->find_snaprealm(); | |
1883 | if (!realm->have_past_parents_open()) { | |
1884 | dout(10) << " no snap purge, one or more past parents NOT open" << dendl; | |
1885 | } else if (fnode.snap_purged_thru < realm->get_last_destroyed()) { | |
1886 | snaps = &realm->get_snaps(); | |
1887 | dout(10) << " snap_purged_thru " << fnode.snap_purged_thru | |
1888 | << " < " << realm->get_last_destroyed() | |
1889 | << ", snap purge based on " << *snaps << dendl; | |
1890 | if (get_num_snap_items() == 0) { | |
1891 | fnode.snap_purged_thru = realm->get_last_destroyed(); | |
1892 | force_dirty = true; | |
1893 | } | |
1894 | } | |
1895 | ||
1896 | unsigned pos = omap.size() - 1; | |
1897 | for (map<string, bufferlist>::reverse_iterator p = omap.rbegin(); | |
1898 | p != omap.rend(); | |
1899 | ++p, --pos) { | |
1900 | string dname; | |
1901 | snapid_t last; | |
1902 | dentry_key_t::decode_helper(p->first, dname, last); | |
1903 | ||
1904 | CDentry *dn = NULL; | |
1905 | try { | |
1906 | dn = _load_dentry( | |
1907 | p->first, dname, last, p->second, pos, snaps, | |
1908 | &force_dirty, &undef_inodes); | |
1909 | } catch (const buffer::error &err) { | |
1910 | cache->mds->clog->warn() << "Corrupt dentry '" << dname << "' in " | |
1911 | "dir frag " << dirfrag() << ": " | |
1912 | << err << "(" << get_path() << ")"; | |
1913 | ||
1914 | // Remember that this dentry is damaged. Subsequent operations | |
1915 | // that try to act directly on it will get their EIOs, but this | |
1916 | // dirfrag as a whole will continue to look okay (minus the | |
1917 | // mysteriously-missing dentry) | |
1918 | go_bad_dentry(last, dname); | |
1919 | ||
1920 | // Anyone who was WAIT_DENTRY for this guy will get kicked | |
1921 | // to RetryRequest, and hit the DamageTable-interrogating path. | |
1922 | // Stats will now be bogus because we will think we're complete, | |
1923 | // but have 1 or more missing dentries. | |
1924 | continue; | |
1925 | } | |
1926 | ||
1927 | if (dn && (wanted_items.count(dname) > 0 || !complete)) { | |
1928 | dout(10) << " touching wanted dn " << *dn << dendl; | |
1929 | inode->mdcache->touch_dentry(dn); | |
1930 | } | |
1931 | ||
1932 | /** clean underwater item? | |
1933 | * Underwater item is something that is dirty in our cache from | |
1934 | * journal replay, but was previously flushed to disk before the | |
1935 | * mds failed. | |
1936 | * | |
1937 | * We only do this is committed_version == 0. that implies either | |
1938 | * - this is a fetch after from a clean/empty CDir is created | |
1939 | * (and has no effect, since the dn won't exist); or | |
1940 | * - this is a fetch after _recovery_, which is what we're worried | |
1941 | * about. Items that are marked dirty from the journal should be | |
1942 | * marked clean if they appear on disk. | |
1943 | */ | |
1944 | if (committed_version == 0 && | |
1945 | dn && | |
1946 | dn->get_version() <= got_fnode.version && | |
1947 | dn->is_dirty()) { | |
1948 | dout(10) << "_fetched had underwater dentry " << *dn << ", marking clean" << dendl; | |
1949 | dn->mark_clean(); | |
1950 | ||
1951 | if (dn->get_linkage()->is_primary()) { | |
1952 | assert(dn->get_linkage()->get_inode()->get_version() <= got_fnode.version); | |
1953 | dout(10) << "_fetched had underwater inode " << *dn->get_linkage()->get_inode() << ", marking clean" << dendl; | |
1954 | dn->get_linkage()->get_inode()->mark_clean(); | |
1955 | } | |
1956 | } | |
1957 | } | |
1958 | ||
1959 | //cache->mds->logger->inc("newin", num_new_inodes_loaded); | |
1960 | ||
1961 | // mark complete, !fetching | |
1962 | if (complete) { | |
1963 | wanted_items.clear(); | |
1964 | mark_complete(); | |
1965 | state_clear(STATE_FETCHING); | |
1966 | ||
1967 | if (scrub_infop && scrub_infop->need_scrub_local) { | |
1968 | scrub_infop->need_scrub_local = false; | |
1969 | scrub_local(); | |
1970 | } | |
1971 | } | |
1972 | ||
1973 | // open & force frags | |
1974 | while (!undef_inodes.empty()) { | |
1975 | CInode *in = undef_inodes.front(); | |
1976 | undef_inodes.pop_front(); | |
1977 | in->state_clear(CInode::STATE_REJOINUNDEF); | |
1978 | cache->opened_undef_inode(in); | |
1979 | } | |
1980 | ||
1981 | // dirty myself to remove stale snap dentries | |
1982 | if (force_dirty && !inode->mdcache->is_readonly()) | |
1983 | log_mark_dirty(); | |
1984 | ||
1985 | auth_unpin(this); | |
1986 | ||
1987 | if (complete) { | |
1988 | // kick waiters | |
1989 | finish_waiting(WAIT_COMPLETE, 0); | |
1990 | } | |
1991 | } | |
1992 | ||
1993 | void CDir::_go_bad() | |
1994 | { | |
1995 | if (get_version() == 0) | |
1996 | set_version(1); | |
1997 | state_set(STATE_BADFRAG); | |
1998 | // mark complete, !fetching | |
1999 | mark_complete(); | |
2000 | state_clear(STATE_FETCHING); | |
2001 | auth_unpin(this); | |
2002 | ||
2003 | // kick waiters | |
2004 | finish_waiting(WAIT_COMPLETE, -EIO); | |
2005 | } | |
2006 | ||
2007 | void CDir::go_bad_dentry(snapid_t last, const std::string &dname) | |
2008 | { | |
2009 | dout(10) << "go_bad_dentry " << dname << dendl; | |
2010 | const bool fatal = cache->mds->damage_table.notify_dentry( | |
2011 | inode->ino(), frag, last, dname, get_path() + "/" + dname); | |
2012 | if (fatal) { | |
2013 | cache->mds->damaged(); | |
2014 | ceph_abort(); // unreachable, damaged() respawns us | |
2015 | } | |
2016 | } | |
2017 | ||
2018 | void CDir::go_bad(bool complete) | |
2019 | { | |
2020 | dout(10) << "go_bad " << frag << dendl; | |
2021 | const bool fatal = cache->mds->damage_table.notify_dirfrag( | |
2022 | inode->ino(), frag, get_path()); | |
2023 | if (fatal) { | |
2024 | cache->mds->damaged(); | |
2025 | ceph_abort(); // unreachable, damaged() respawns us | |
2026 | } | |
2027 | ||
2028 | if (complete) | |
2029 | _go_bad(); | |
2030 | else | |
2031 | auth_unpin(this); | |
2032 | } | |
2033 | ||
2034 | // ----------------------- | |
2035 | // COMMIT | |
2036 | ||
2037 | /** | |
2038 | * commit | |
2039 | * | |
2040 | * @param want - min version i want committed | |
2041 | * @param c - callback for completion | |
2042 | */ | |
2043 | void CDir::commit(version_t want, MDSInternalContextBase *c, bool ignore_authpinnability, int op_prio) | |
2044 | { | |
2045 | dout(10) << "commit want " << want << " on " << *this << dendl; | |
2046 | if (want == 0) want = get_version(); | |
2047 | ||
2048 | // preconditions | |
2049 | assert(want <= get_version() || get_version() == 0); // can't commit the future | |
2050 | assert(want > committed_version); // the caller is stupid | |
2051 | assert(is_auth()); | |
2052 | assert(ignore_authpinnability || can_auth_pin()); | |
2053 | ||
2054 | // note: queue up a noop if necessary, so that we always | |
2055 | // get an auth_pin. | |
2056 | if (!c) | |
2057 | c = new C_MDSInternalNoop; | |
2058 | ||
2059 | // auth_pin on first waiter | |
2060 | if (waiting_for_commit.empty()) | |
2061 | auth_pin(this); | |
2062 | waiting_for_commit[want].push_back(c); | |
2063 | ||
2064 | // ok. | |
2065 | _commit(want, op_prio); | |
2066 | } | |
2067 | ||
2068 | class C_IO_Dir_Committed : public CDirIOContext { | |
2069 | version_t version; | |
2070 | public: | |
2071 | C_IO_Dir_Committed(CDir *d, version_t v) : CDirIOContext(d), version(v) { } | |
2072 | void finish(int r) override { | |
2073 | dir->_committed(r, version); | |
2074 | } | |
2075 | }; | |
2076 | ||
2077 | /** | |
2078 | * Flush out the modified dentries in this dir. Keep the bufferlist | |
2079 | * below max_write_size; | |
2080 | */ | |
2081 | void CDir::_omap_commit(int op_prio) | |
2082 | { | |
2083 | dout(10) << "_omap_commit" << dendl; | |
2084 | ||
2085 | unsigned max_write_size = cache->max_dir_commit_size; | |
2086 | unsigned write_size = 0; | |
2087 | ||
2088 | if (op_prio < 0) | |
2089 | op_prio = CEPH_MSG_PRIO_DEFAULT; | |
2090 | ||
2091 | // snap purge? | |
2092 | const set<snapid_t> *snaps = NULL; | |
2093 | SnapRealm *realm = inode->find_snaprealm(); | |
2094 | if (!realm->have_past_parents_open()) { | |
2095 | dout(10) << " no snap purge, one or more past parents NOT open" << dendl; | |
2096 | } else if (fnode.snap_purged_thru < realm->get_last_destroyed()) { | |
2097 | snaps = &realm->get_snaps(); | |
2098 | dout(10) << " snap_purged_thru " << fnode.snap_purged_thru | |
2099 | << " < " << realm->get_last_destroyed() | |
2100 | << ", snap purge based on " << *snaps << dendl; | |
2101 | // fnode.snap_purged_thru = realm->get_last_destroyed(); | |
2102 | } | |
2103 | ||
2104 | set<string> to_remove; | |
2105 | map<string, bufferlist> to_set; | |
2106 | ||
2107 | C_GatherBuilder gather(g_ceph_context, | |
2108 | new C_OnFinisher(new C_IO_Dir_Committed(this, | |
2109 | get_version()), | |
2110 | cache->mds->finisher)); | |
2111 | ||
2112 | SnapContext snapc; | |
2113 | object_t oid = get_ondisk_object(); | |
2114 | object_locator_t oloc(cache->mds->mdsmap->get_metadata_pool()); | |
2115 | ||
2116 | if (!stale_items.empty()) { | |
2117 | for (compact_set<string>::iterator p = stale_items.begin(); | |
2118 | p != stale_items.end(); | |
2119 | ++p) { | |
2120 | to_remove.insert(*p); | |
2121 | write_size += (*p).length(); | |
2122 | } | |
2123 | stale_items.clear(); | |
2124 | } | |
2125 | ||
2126 | for (map_t::iterator p = items.begin(); | |
2127 | p != items.end(); ) { | |
2128 | CDentry *dn = p->second; | |
2129 | ++p; | |
2130 | ||
2131 | string key; | |
2132 | dn->key().encode(key); | |
2133 | ||
2134 | if (dn->last != CEPH_NOSNAP && | |
2135 | snaps && try_trim_snap_dentry(dn, *snaps)) { | |
2136 | dout(10) << " rm " << key << dendl; | |
2137 | write_size += key.length(); | |
2138 | to_remove.insert(key); | |
2139 | continue; | |
2140 | } | |
2141 | ||
2142 | if (!dn->is_dirty() && | |
2143 | (!dn->state_test(CDentry::STATE_FRAGMENTING) || dn->get_linkage()->is_null())) | |
2144 | continue; // skip clean dentries | |
2145 | ||
2146 | if (dn->get_linkage()->is_null()) { | |
2147 | dout(10) << " rm " << dn->name << " " << *dn << dendl; | |
2148 | write_size += key.length(); | |
2149 | to_remove.insert(key); | |
2150 | } else { | |
2151 | dout(10) << " set " << dn->name << " " << *dn << dendl; | |
2152 | bufferlist dnbl; | |
2153 | _encode_dentry(dn, dnbl, snaps); | |
2154 | write_size += key.length() + dnbl.length(); | |
2155 | to_set[key].swap(dnbl); | |
2156 | } | |
2157 | ||
2158 | if (write_size >= max_write_size) { | |
2159 | ObjectOperation op; | |
2160 | op.priority = op_prio; | |
2161 | ||
2162 | // don't create new dirfrag blindly | |
2163 | if (!is_new() && !state_test(CDir::STATE_FRAGMENTING)) | |
2164 | op.stat(NULL, (ceph::real_time*) NULL, NULL); | |
2165 | ||
2166 | if (!to_set.empty()) | |
2167 | op.omap_set(to_set); | |
2168 | if (!to_remove.empty()) | |
2169 | op.omap_rm_keys(to_remove); | |
2170 | ||
2171 | cache->mds->objecter->mutate(oid, oloc, op, snapc, | |
2172 | ceph::real_clock::now(), | |
2173 | 0, gather.new_sub()); | |
2174 | ||
2175 | write_size = 0; | |
2176 | to_set.clear(); | |
2177 | to_remove.clear(); | |
2178 | } | |
2179 | } | |
2180 | ||
2181 | ObjectOperation op; | |
2182 | op.priority = op_prio; | |
2183 | ||
2184 | // don't create new dirfrag blindly | |
2185 | if (!is_new() && !state_test(CDir::STATE_FRAGMENTING)) | |
2186 | op.stat(NULL, (ceph::real_time*)NULL, NULL); | |
2187 | ||
2188 | /* | |
2189 | * save the header at the last moment.. If we were to send it off before other | |
2190 | * updates, but die before sending them all, we'd think that the on-disk state | |
2191 | * was fully committed even though it wasn't! However, since the messages are | |
2192 | * strictly ordered between the MDS and the OSD, and since messages to a given | |
2193 | * PG are strictly ordered, if we simply send the message containing the header | |
2194 | * off last, we cannot get our header into an incorrect state. | |
2195 | */ | |
2196 | bufferlist header; | |
2197 | ::encode(fnode, header); | |
2198 | op.omap_set_header(header); | |
2199 | ||
2200 | if (!to_set.empty()) | |
2201 | op.omap_set(to_set); | |
2202 | if (!to_remove.empty()) | |
2203 | op.omap_rm_keys(to_remove); | |
2204 | ||
2205 | cache->mds->objecter->mutate(oid, oloc, op, snapc, | |
2206 | ceph::real_clock::now(), | |
2207 | 0, gather.new_sub()); | |
2208 | ||
2209 | gather.activate(); | |
2210 | } | |
2211 | ||
2212 | void CDir::_encode_dentry(CDentry *dn, bufferlist& bl, | |
2213 | const set<snapid_t> *snaps) | |
2214 | { | |
2215 | // clear dentry NEW flag, if any. we can no longer silently drop it. | |
2216 | dn->clear_new(); | |
2217 | ||
2218 | ::encode(dn->first, bl); | |
2219 | ||
2220 | // primary or remote? | |
2221 | if (dn->linkage.is_remote()) { | |
2222 | inodeno_t ino = dn->linkage.get_remote_ino(); | |
2223 | unsigned char d_type = dn->linkage.get_remote_d_type(); | |
2224 | dout(14) << " pos " << bl.length() << " dn '" << dn->name << "' remote ino " << ino << dendl; | |
2225 | ||
2226 | // marker, name, ino | |
2227 | bl.append('L'); // remote link | |
2228 | ::encode(ino, bl); | |
2229 | ::encode(d_type, bl); | |
2230 | } else if (dn->linkage.is_primary()) { | |
2231 | // primary link | |
2232 | CInode *in = dn->linkage.get_inode(); | |
2233 | assert(in); | |
2234 | ||
2235 | dout(14) << " pos " << bl.length() << " dn '" << dn->name << "' inode " << *in << dendl; | |
2236 | ||
2237 | // marker, name, inode, [symlink string] | |
2238 | bl.append('I'); // inode | |
2239 | ||
2240 | if (in->is_multiversion()) { | |
2241 | if (!in->snaprealm) { | |
2242 | if (snaps) | |
2243 | in->purge_stale_snap_data(*snaps); | |
2244 | } else if (in->snaprealm->have_past_parents_open()) { | |
2245 | in->purge_stale_snap_data(in->snaprealm->get_snaps()); | |
2246 | } | |
2247 | } | |
2248 | ||
2249 | bufferlist snap_blob; | |
2250 | in->encode_snap_blob(snap_blob); | |
2251 | in->encode_bare(bl, cache->mds->mdsmap->get_up_features(), &snap_blob); | |
2252 | } else { | |
2253 | assert(!dn->linkage.is_null()); | |
2254 | } | |
2255 | } | |
2256 | ||
2257 | void CDir::_commit(version_t want, int op_prio) | |
2258 | { | |
2259 | dout(10) << "_commit want " << want << " on " << *this << dendl; | |
2260 | ||
2261 | // we can't commit things in the future. | |
2262 | // (even the projected future.) | |
2263 | assert(want <= get_version() || get_version() == 0); | |
2264 | ||
2265 | // check pre+postconditions. | |
2266 | assert(is_auth()); | |
2267 | ||
2268 | // already committed? | |
2269 | if (committed_version >= want) { | |
2270 | dout(10) << "already committed " << committed_version << " >= " << want << dendl; | |
2271 | return; | |
2272 | } | |
2273 | // already committing >= want? | |
2274 | if (committing_version >= want) { | |
2275 | dout(10) << "already committing " << committing_version << " >= " << want << dendl; | |
2276 | assert(state_test(STATE_COMMITTING)); | |
2277 | return; | |
2278 | } | |
2279 | ||
2280 | // alrady committed an older version? | |
2281 | if (committing_version > committed_version) { | |
2282 | dout(10) << "already committing older " << committing_version << ", waiting for that to finish" << dendl; | |
2283 | return; | |
2284 | } | |
2285 | ||
2286 | // commit. | |
2287 | committing_version = get_version(); | |
2288 | ||
2289 | // mark committing (if not already) | |
2290 | if (!state_test(STATE_COMMITTING)) { | |
2291 | dout(10) << "marking committing" << dendl; | |
2292 | state_set(STATE_COMMITTING); | |
2293 | } | |
2294 | ||
2295 | if (cache->mds->logger) cache->mds->logger->inc(l_mds_dir_commit); | |
2296 | ||
2297 | _omap_commit(op_prio); | |
2298 | } | |
2299 | ||
2300 | ||
2301 | /** | |
2302 | * _committed | |
2303 | * | |
2304 | * @param v version i just committed | |
2305 | */ | |
2306 | void CDir::_committed(int r, version_t v) | |
2307 | { | |
2308 | if (r < 0) { | |
2309 | // the directory could be partly purged during MDS failover | |
2310 | if (r == -ENOENT && committed_version == 0 && | |
2311 | !inode->is_base() && get_parent_dir()->inode->is_stray()) { | |
2312 | r = 0; | |
2313 | if (inode->snaprealm) | |
2314 | inode->state_set(CInode::STATE_MISSINGOBJS); | |
2315 | } | |
2316 | if (r < 0) { | |
2317 | dout(1) << "commit error " << r << " v " << v << dendl; | |
2318 | cache->mds->clog->error() << "failed to commit dir " << dirfrag() << " object," | |
2319 | << " errno " << r; | |
2320 | cache->mds->handle_write_error(r); | |
2321 | return; | |
2322 | } | |
2323 | } | |
2324 | ||
2325 | dout(10) << "_committed v " << v << " on " << *this << dendl; | |
2326 | assert(is_auth()); | |
2327 | ||
2328 | bool stray = inode->is_stray(); | |
2329 | ||
2330 | // take note. | |
2331 | assert(v > committed_version); | |
2332 | assert(v <= committing_version); | |
2333 | committed_version = v; | |
2334 | ||
2335 | // _all_ commits done? | |
2336 | if (committing_version == committed_version) | |
2337 | state_clear(CDir::STATE_COMMITTING); | |
2338 | ||
2339 | // _any_ commit, even if we've been redirtied, means we're no longer new. | |
2340 | item_new.remove_myself(); | |
2341 | ||
2342 | // dir clean? | |
2343 | if (committed_version == get_version()) | |
2344 | mark_clean(); | |
2345 | ||
2346 | // dentries clean? | |
2347 | for (map_t::iterator it = items.begin(); | |
2348 | it != items.end(); ) { | |
2349 | CDentry *dn = it->second; | |
2350 | ++it; | |
2351 | ||
2352 | // inode? | |
2353 | if (dn->linkage.is_primary()) { | |
2354 | CInode *in = dn->linkage.get_inode(); | |
2355 | assert(in); | |
2356 | assert(in->is_auth()); | |
2357 | ||
2358 | if (committed_version >= in->get_version()) { | |
2359 | if (in->is_dirty()) { | |
2360 | dout(15) << " dir " << committed_version << " >= inode " << in->get_version() << " now clean " << *in << dendl; | |
2361 | in->mark_clean(); | |
2362 | } | |
2363 | } else { | |
2364 | dout(15) << " dir " << committed_version << " < inode " << in->get_version() << " still dirty " << *in << dendl; | |
2365 | assert(in->is_dirty() || in->last < CEPH_NOSNAP); // special case for cow snap items (not predirtied) | |
2366 | } | |
2367 | } | |
2368 | ||
2369 | // dentry | |
2370 | if (committed_version >= dn->get_version()) { | |
2371 | if (dn->is_dirty()) { | |
2372 | dout(15) << " dir " << committed_version << " >= dn " << dn->get_version() << " now clean " << *dn << dendl; | |
2373 | dn->mark_clean(); | |
2374 | ||
2375 | // drop clean null stray dentries immediately | |
2376 | if (stray && | |
2377 | dn->get_num_ref() == 0 && | |
2378 | !dn->is_projected() && | |
2379 | dn->get_linkage()->is_null()) | |
2380 | remove_dentry(dn); | |
2381 | } | |
2382 | } else { | |
2383 | dout(15) << " dir " << committed_version << " < dn " << dn->get_version() << " still dirty " << *dn << dendl; | |
2384 | } | |
2385 | } | |
2386 | ||
2387 | // finishers? | |
2388 | bool were_waiters = !waiting_for_commit.empty(); | |
2389 | ||
2390 | compact_map<version_t, list<MDSInternalContextBase*> >::iterator p = waiting_for_commit.begin(); | |
2391 | while (p != waiting_for_commit.end()) { | |
2392 | compact_map<version_t, list<MDSInternalContextBase*> >::iterator n = p; | |
2393 | ++n; | |
2394 | if (p->first > committed_version) { | |
2395 | dout(10) << " there are waiters for " << p->first << ", committing again" << dendl; | |
2396 | _commit(p->first, -1); | |
2397 | break; | |
2398 | } | |
2399 | cache->mds->queue_waiters(p->second); | |
2400 | waiting_for_commit.erase(p); | |
2401 | p = n; | |
2402 | } | |
2403 | ||
2404 | // try drop dentries in this dirfrag if it's about to be purged | |
2405 | if (!inode->is_base() && get_parent_dir()->inode->is_stray() && | |
2406 | inode->snaprealm) | |
2407 | cache->maybe_eval_stray(inode, true); | |
2408 | ||
2409 | // unpin if we kicked the last waiter. | |
2410 | if (were_waiters && | |
2411 | waiting_for_commit.empty()) | |
2412 | auth_unpin(this); | |
2413 | } | |
2414 | ||
2415 | ||
2416 | ||
2417 | ||
2418 | // IMPORT/EXPORT | |
2419 | ||
2420 | void CDir::encode_export(bufferlist& bl) | |
2421 | { | |
2422 | assert(!is_projected()); | |
2423 | ::encode(first, bl); | |
2424 | ::encode(fnode, bl); | |
2425 | ::encode(dirty_old_rstat, bl); | |
2426 | ::encode(committed_version, bl); | |
2427 | ||
2428 | ::encode(state, bl); | |
2429 | ::encode(dir_rep, bl); | |
2430 | ||
2431 | ::encode(pop_me, bl); | |
2432 | ::encode(pop_auth_subtree, bl); | |
2433 | ||
2434 | ::encode(dir_rep_by, bl); | |
2435 | ::encode(replica_map, bl); | |
2436 | ||
2437 | get(PIN_TEMPEXPORTING); | |
2438 | } | |
2439 | ||
2440 | void CDir::finish_export(utime_t now) | |
2441 | { | |
2442 | state &= MASK_STATE_EXPORT_KEPT; | |
2443 | pop_auth_subtree_nested.sub(now, cache->decayrate, pop_auth_subtree); | |
2444 | pop_me.zero(now); | |
2445 | pop_auth_subtree.zero(now); | |
2446 | put(PIN_TEMPEXPORTING); | |
2447 | dirty_old_rstat.clear(); | |
2448 | } | |
2449 | ||
2450 | void CDir::decode_import(bufferlist::iterator& blp, utime_t now, LogSegment *ls) | |
2451 | { | |
2452 | ::decode(first, blp); | |
2453 | ::decode(fnode, blp); | |
2454 | ::decode(dirty_old_rstat, blp); | |
2455 | projected_version = fnode.version; | |
2456 | ::decode(committed_version, blp); | |
2457 | committing_version = committed_version; | |
2458 | ||
2459 | unsigned s; | |
2460 | ::decode(s, blp); | |
2461 | state &= MASK_STATE_IMPORT_KEPT; | |
2462 | state_set(STATE_AUTH | (s & MASK_STATE_EXPORTED)); | |
2463 | ||
2464 | if (is_dirty()) { | |
2465 | get(PIN_DIRTY); | |
2466 | _mark_dirty(ls); | |
2467 | } | |
2468 | ||
2469 | ::decode(dir_rep, blp); | |
2470 | ||
2471 | ::decode(pop_me, now, blp); | |
2472 | ::decode(pop_auth_subtree, now, blp); | |
2473 | pop_auth_subtree_nested.add(now, cache->decayrate, pop_auth_subtree); | |
2474 | ||
2475 | ::decode(dir_rep_by, blp); | |
2476 | ::decode(replica_map, blp); | |
2477 | if (!replica_map.empty()) get(PIN_REPLICATED); | |
2478 | ||
2479 | replica_nonce = 0; // no longer defined | |
2480 | ||
2481 | // did we import some dirty scatterlock data? | |
2482 | if (dirty_old_rstat.size() || | |
2483 | !(fnode.rstat == fnode.accounted_rstat)) { | |
2484 | cache->mds->locker->mark_updated_scatterlock(&inode->nestlock); | |
2485 | ls->dirty_dirfrag_nest.push_back(&inode->item_dirty_dirfrag_nest); | |
2486 | } | |
2487 | if (!(fnode.fragstat == fnode.accounted_fragstat)) { | |
2488 | cache->mds->locker->mark_updated_scatterlock(&inode->filelock); | |
2489 | ls->dirty_dirfrag_dir.push_back(&inode->item_dirty_dirfrag_dir); | |
2490 | } | |
2491 | if (is_dirty_dft()) { | |
2492 | if (inode->dirfragtreelock.get_state() != LOCK_MIX && | |
2493 | inode->dirfragtreelock.is_stable()) { | |
2494 | // clear stale dirtydft | |
2495 | state_clear(STATE_DIRTYDFT); | |
2496 | } else { | |
2497 | cache->mds->locker->mark_updated_scatterlock(&inode->dirfragtreelock); | |
2498 | ls->dirty_dirfrag_dirfragtree.push_back(&inode->item_dirty_dirfrag_dirfragtree); | |
2499 | } | |
2500 | } | |
2501 | } | |
2502 | ||
2503 | ||
2504 | ||
2505 | ||
2506 | /******************************** | |
2507 | * AUTHORITY | |
2508 | */ | |
2509 | ||
2510 | /* | |
2511 | * if dir_auth.first == parent, auth is same as inode. | |
2512 | * unless .second != unknown, in which case that sticks. | |
2513 | */ | |
2514 | mds_authority_t CDir::authority() const | |
2515 | { | |
2516 | if (is_subtree_root()) | |
2517 | return dir_auth; | |
2518 | else | |
2519 | return inode->authority(); | |
2520 | } | |
2521 | ||
2522 | /** is_subtree_root() | |
2523 | * true if this is an auth delegation point. | |
2524 | * that is, dir_auth != default (parent,unknown) | |
2525 | * | |
2526 | * some key observations: | |
2527 | * if i am auth: | |
2528 | * - any region bound will be an export, or frozen. | |
2529 | * | |
2530 | * note that this DOES heed dir_auth.pending | |
2531 | */ | |
2532 | /* | |
2533 | bool CDir::is_subtree_root() | |
2534 | { | |
2535 | if (dir_auth == CDIR_AUTH_DEFAULT) { | |
2536 | //dout(10) << "is_subtree_root false " << dir_auth << " != " << CDIR_AUTH_DEFAULT | |
2537 | //<< " on " << ino() << dendl; | |
2538 | return false; | |
2539 | } else { | |
2540 | //dout(10) << "is_subtree_root true " << dir_auth << " != " << CDIR_AUTH_DEFAULT | |
2541 | //<< " on " << ino() << dendl; | |
2542 | return true; | |
2543 | } | |
2544 | } | |
2545 | */ | |
2546 | ||
2547 | /** contains(x) | |
2548 | * true if we are x, or an ancestor of x | |
2549 | */ | |
2550 | bool CDir::contains(CDir *x) | |
2551 | { | |
2552 | while (1) { | |
2553 | if (x == this) | |
2554 | return true; | |
2555 | x = x->get_inode()->get_projected_parent_dir(); | |
2556 | if (x == 0) | |
2557 | return false; | |
2558 | } | |
2559 | } | |
2560 | ||
2561 | ||
2562 | ||
2563 | /** set_dir_auth | |
2564 | */ | |
2565 | void CDir::set_dir_auth(mds_authority_t a) | |
2566 | { | |
2567 | dout(10) << "setting dir_auth=" << a | |
2568 | << " from " << dir_auth | |
2569 | << " on " << *this << dendl; | |
2570 | ||
2571 | bool was_subtree = is_subtree_root(); | |
2572 | bool was_ambiguous = dir_auth.second >= 0; | |
2573 | ||
2574 | // set it. | |
2575 | dir_auth = a; | |
2576 | ||
2577 | // new subtree root? | |
2578 | if (!was_subtree && is_subtree_root()) { | |
2579 | dout(10) << " new subtree root, adjusting auth_pins" << dendl; | |
2580 | ||
2581 | // adjust nested auth pins | |
2582 | if (get_cum_auth_pins()) | |
2583 | inode->adjust_nested_auth_pins(-1, NULL); | |
2584 | ||
2585 | // unpin parent of frozen dir/tree? | |
2586 | if (inode->is_auth() && (is_frozen_tree_root() || is_frozen_dir())) | |
2587 | inode->auth_unpin(this); | |
2588 | } | |
2589 | if (was_subtree && !is_subtree_root()) { | |
2590 | dout(10) << " old subtree root, adjusting auth_pins" << dendl; | |
2591 | ||
2592 | // adjust nested auth pins | |
2593 | if (get_cum_auth_pins()) | |
2594 | inode->adjust_nested_auth_pins(1, NULL); | |
2595 | ||
2596 | // pin parent of frozen dir/tree? | |
2597 | if (inode->is_auth() && (is_frozen_tree_root() || is_frozen_dir())) | |
2598 | inode->auth_pin(this); | |
2599 | } | |
2600 | ||
2601 | // newly single auth? | |
2602 | if (was_ambiguous && dir_auth.second == CDIR_AUTH_UNKNOWN) { | |
2603 | list<MDSInternalContextBase*> ls; | |
2604 | take_waiting(WAIT_SINGLEAUTH, ls); | |
2605 | cache->mds->queue_waiters(ls); | |
2606 | } | |
2607 | } | |
2608 | ||
2609 | ||
2610 | /***************************************** | |
2611 | * AUTH PINS and FREEZING | |
2612 | * | |
2613 | * the basic plan is that auth_pins only exist in auth regions, and they | |
2614 | * prevent a freeze (and subsequent auth change). | |
2615 | * | |
2616 | * however, we also need to prevent a parent from freezing if a child is frozen. | |
2617 | * for that reason, the parent inode of a frozen directory is auth_pinned. | |
2618 | * | |
2619 | * the oddity is when the frozen directory is a subtree root. if that's the case, | |
2620 | * the parent inode isn't frozen. which means that when subtree authority is adjusted | |
2621 | * at the bounds, inodes for any frozen bound directories need to get auth_pins at that | |
2622 | * time. | |
2623 | * | |
2624 | */ | |
2625 | ||
2626 | void CDir::auth_pin(void *by) | |
2627 | { | |
2628 | if (auth_pins == 0) | |
2629 | get(PIN_AUTHPIN); | |
2630 | auth_pins++; | |
2631 | ||
2632 | #ifdef MDS_AUTHPIN_SET | |
2633 | auth_pin_set.insert(by); | |
2634 | #endif | |
2635 | ||
2636 | dout(10) << "auth_pin by " << by | |
2637 | << " on " << *this | |
2638 | << " count now " << auth_pins << " + " << nested_auth_pins << dendl; | |
2639 | ||
2640 | // nest pins? | |
2641 | if (!is_subtree_root() && | |
2642 | get_cum_auth_pins() == 1) | |
2643 | inode->adjust_nested_auth_pins(1, by); | |
2644 | } | |
2645 | ||
2646 | void CDir::auth_unpin(void *by) | |
2647 | { | |
2648 | auth_pins--; | |
2649 | ||
2650 | #ifdef MDS_AUTHPIN_SET | |
2651 | assert(auth_pin_set.count(by)); | |
2652 | auth_pin_set.erase(auth_pin_set.find(by)); | |
2653 | #endif | |
2654 | if (auth_pins == 0) | |
2655 | put(PIN_AUTHPIN); | |
2656 | ||
2657 | dout(10) << "auth_unpin by " << by | |
2658 | << " on " << *this | |
2659 | << " count now " << auth_pins << " + " << nested_auth_pins << dendl; | |
2660 | assert(auth_pins >= 0); | |
2661 | ||
2662 | int newcum = get_cum_auth_pins(); | |
2663 | ||
2664 | maybe_finish_freeze(); // pending freeze? | |
2665 | ||
2666 | // nest? | |
2667 | if (!is_subtree_root() && | |
2668 | newcum == 0) | |
2669 | inode->adjust_nested_auth_pins(-1, by); | |
2670 | } | |
2671 | ||
2672 | void CDir::adjust_nested_auth_pins(int inc, int dirinc, void *by) | |
2673 | { | |
2674 | assert(inc); | |
2675 | nested_auth_pins += inc; | |
2676 | dir_auth_pins += dirinc; | |
2677 | ||
2678 | dout(15) << "adjust_nested_auth_pins " << inc << "/" << dirinc << " on " << *this | |
2679 | << " by " << by << " count now " | |
2680 | << auth_pins << " + " << nested_auth_pins << dendl; | |
2681 | assert(nested_auth_pins >= 0); | |
2682 | assert(dir_auth_pins >= 0); | |
2683 | ||
2684 | int newcum = get_cum_auth_pins(); | |
2685 | ||
2686 | maybe_finish_freeze(); // pending freeze? | |
2687 | ||
2688 | // nest? | |
2689 | if (!is_subtree_root()) { | |
2690 | if (newcum == 0) | |
2691 | inode->adjust_nested_auth_pins(-1, by); | |
2692 | else if (newcum == inc) | |
2693 | inode->adjust_nested_auth_pins(1, by); | |
2694 | } | |
2695 | } | |
2696 | ||
2697 | #ifdef MDS_VERIFY_FRAGSTAT | |
2698 | void CDir::verify_fragstat() | |
2699 | { | |
2700 | assert(is_complete()); | |
2701 | if (inode->is_stray()) | |
2702 | return; | |
2703 | ||
2704 | frag_info_t c; | |
2705 | memset(&c, 0, sizeof(c)); | |
2706 | ||
2707 | for (map_t::iterator it = items.begin(); | |
2708 | it != items.end(); | |
2709 | ++it) { | |
2710 | CDentry *dn = it->second; | |
2711 | if (dn->is_null()) | |
2712 | continue; | |
2713 | ||
2714 | dout(10) << " " << *dn << dendl; | |
2715 | if (dn->is_primary()) | |
2716 | dout(10) << " " << *dn->inode << dendl; | |
2717 | ||
2718 | if (dn->is_primary()) { | |
2719 | if (dn->inode->is_dir()) | |
2720 | c.nsubdirs++; | |
2721 | else | |
2722 | c.nfiles++; | |
2723 | } | |
2724 | if (dn->is_remote()) { | |
2725 | if (dn->get_remote_d_type() == DT_DIR) | |
2726 | c.nsubdirs++; | |
2727 | else | |
2728 | c.nfiles++; | |
2729 | } | |
2730 | } | |
2731 | ||
2732 | if (c.nsubdirs != fnode.fragstat.nsubdirs || | |
2733 | c.nfiles != fnode.fragstat.nfiles) { | |
2734 | dout(0) << "verify_fragstat failed " << fnode.fragstat << " on " << *this << dendl; | |
2735 | dout(0) << " i count " << c << dendl; | |
2736 | ceph_abort(); | |
2737 | } else { | |
2738 | dout(0) << "verify_fragstat ok " << fnode.fragstat << " on " << *this << dendl; | |
2739 | } | |
2740 | } | |
2741 | #endif | |
2742 | ||
2743 | /***************************************************************************** | |
2744 | * FREEZING | |
2745 | */ | |
2746 | ||
2747 | // FREEZE TREE | |
2748 | ||
2749 | bool CDir::freeze_tree() | |
2750 | { | |
2751 | assert(!is_frozen()); | |
2752 | assert(!is_freezing()); | |
2753 | ||
2754 | auth_pin(this); | |
2755 | if (is_freezeable(true)) { | |
2756 | _freeze_tree(); | |
2757 | auth_unpin(this); | |
2758 | return true; | |
2759 | } else { | |
2760 | state_set(STATE_FREEZINGTREE); | |
2761 | ++num_freezing_trees; | |
2762 | dout(10) << "freeze_tree waiting " << *this << dendl; | |
2763 | return false; | |
2764 | } | |
2765 | } | |
2766 | ||
2767 | void CDir::_freeze_tree() | |
2768 | { | |
2769 | dout(10) << "_freeze_tree " << *this << dendl; | |
2770 | assert(is_freezeable(true)); | |
2771 | ||
2772 | // twiddle state | |
2773 | if (state_test(STATE_FREEZINGTREE)) { | |
2774 | state_clear(STATE_FREEZINGTREE); // actually, this may get set again by next context? | |
2775 | --num_freezing_trees; | |
2776 | } | |
2777 | state_set(STATE_FROZENTREE); | |
2778 | ++num_frozen_trees; | |
2779 | get(PIN_FROZEN); | |
2780 | ||
2781 | // auth_pin inode for duration of freeze, if we are not a subtree root. | |
2782 | if (is_auth() && !is_subtree_root()) | |
2783 | inode->auth_pin(this); | |
2784 | } | |
2785 | ||
2786 | void CDir::unfreeze_tree() | |
2787 | { | |
2788 | dout(10) << "unfreeze_tree " << *this << dendl; | |
2789 | ||
2790 | if (state_test(STATE_FROZENTREE)) { | |
2791 | // frozen. unfreeze. | |
2792 | state_clear(STATE_FROZENTREE); | |
2793 | --num_frozen_trees; | |
2794 | ||
2795 | put(PIN_FROZEN); | |
2796 | ||
2797 | // unpin (may => FREEZEABLE) FIXME: is this order good? | |
2798 | if (is_auth() && !is_subtree_root()) | |
2799 | inode->auth_unpin(this); | |
2800 | ||
2801 | // waiters? | |
2802 | finish_waiting(WAIT_UNFREEZE); | |
2803 | } else { | |
2804 | finish_waiting(WAIT_FROZEN, -1); | |
2805 | ||
2806 | // freezing. stop it. | |
2807 | assert(state_test(STATE_FREEZINGTREE)); | |
2808 | state_clear(STATE_FREEZINGTREE); | |
2809 | --num_freezing_trees; | |
2810 | auth_unpin(this); | |
2811 | ||
2812 | finish_waiting(WAIT_UNFREEZE); | |
2813 | } | |
2814 | } | |
2815 | ||
2816 | bool CDir::is_freezing_tree() const | |
2817 | { | |
2818 | if (num_freezing_trees == 0) | |
2819 | return false; | |
2820 | const CDir *dir = this; | |
2821 | while (1) { | |
2822 | if (dir->is_freezing_tree_root()) return true; | |
2823 | if (dir->is_subtree_root()) return false; | |
2824 | if (dir->inode->parent) | |
2825 | dir = dir->inode->parent->dir; | |
2826 | else | |
2827 | return false; // root on replica | |
2828 | } | |
2829 | } | |
2830 | ||
2831 | bool CDir::is_frozen_tree() const | |
2832 | { | |
2833 | if (num_frozen_trees == 0) | |
2834 | return false; | |
2835 | const CDir *dir = this; | |
2836 | while (1) { | |
2837 | if (dir->is_frozen_tree_root()) return true; | |
2838 | if (dir->is_subtree_root()) return false; | |
2839 | if (dir->inode->parent) | |
2840 | dir = dir->inode->parent->dir; | |
2841 | else | |
2842 | return false; // root on replica | |
2843 | } | |
2844 | } | |
2845 | ||
2846 | CDir *CDir::get_frozen_tree_root() | |
2847 | { | |
2848 | assert(is_frozen()); | |
2849 | CDir *dir = this; | |
2850 | while (1) { | |
2851 | if (dir->is_frozen_tree_root()) | |
2852 | return dir; | |
2853 | if (dir->inode->parent) | |
2854 | dir = dir->inode->parent->dir; | |
2855 | else | |
2856 | ceph_abort(); | |
2857 | } | |
2858 | } | |
2859 | ||
2860 | class C_Dir_AuthUnpin : public CDirContext { | |
2861 | public: | |
2862 | explicit C_Dir_AuthUnpin(CDir *d) : CDirContext(d) {} | |
2863 | void finish(int r) override { | |
2864 | dir->auth_unpin(dir->get_inode()); | |
2865 | } | |
2866 | }; | |
2867 | ||
2868 | void CDir::maybe_finish_freeze() | |
2869 | { | |
2870 | if (auth_pins != 1 || dir_auth_pins != 0) | |
2871 | return; | |
2872 | ||
2873 | // we can freeze the _dir_ even with nested pins... | |
2874 | if (state_test(STATE_FREEZINGDIR)) { | |
2875 | _freeze_dir(); | |
2876 | auth_unpin(this); | |
2877 | finish_waiting(WAIT_FROZEN); | |
2878 | } | |
2879 | ||
2880 | if (nested_auth_pins != 0) | |
2881 | return; | |
2882 | ||
2883 | if (state_test(STATE_FREEZINGTREE)) { | |
2884 | if (!is_subtree_root() && inode->is_frozen()) { | |
2885 | dout(10) << "maybe_finish_freeze !subtree root and frozen inode, waiting for unfreeze on " << inode << dendl; | |
2886 | // retake an auth_pin... | |
2887 | auth_pin(inode); | |
2888 | // and release it when the parent inode unfreezes | |
2889 | inode->add_waiter(WAIT_UNFREEZE, new C_Dir_AuthUnpin(this)); | |
2890 | return; | |
2891 | } | |
2892 | ||
2893 | _freeze_tree(); | |
2894 | auth_unpin(this); | |
2895 | finish_waiting(WAIT_FROZEN); | |
2896 | } | |
2897 | } | |
2898 | ||
2899 | ||
2900 | ||
2901 | // FREEZE DIR | |
2902 | ||
2903 | bool CDir::freeze_dir() | |
2904 | { | |
2905 | assert(!is_frozen()); | |
2906 | assert(!is_freezing()); | |
2907 | ||
2908 | auth_pin(this); | |
2909 | if (is_freezeable_dir(true)) { | |
2910 | _freeze_dir(); | |
2911 | auth_unpin(this); | |
2912 | return true; | |
2913 | } else { | |
2914 | state_set(STATE_FREEZINGDIR); | |
2915 | dout(10) << "freeze_dir + wait " << *this << dendl; | |
2916 | return false; | |
2917 | } | |
2918 | } | |
2919 | ||
2920 | void CDir::_freeze_dir() | |
2921 | { | |
2922 | dout(10) << "_freeze_dir " << *this << dendl; | |
2923 | //assert(is_freezeable_dir(true)); | |
2924 | // not always true during split because the original fragment may have frozen a while | |
2925 | // ago and we're just now getting around to breaking it up. | |
2926 | ||
2927 | state_clear(STATE_FREEZINGDIR); | |
2928 | state_set(STATE_FROZENDIR); | |
2929 | get(PIN_FROZEN); | |
2930 | ||
2931 | if (is_auth() && !is_subtree_root()) | |
2932 | inode->auth_pin(this); // auth_pin for duration of freeze | |
2933 | } | |
2934 | ||
2935 | ||
2936 | void CDir::unfreeze_dir() | |
2937 | { | |
2938 | dout(10) << "unfreeze_dir " << *this << dendl; | |
2939 | ||
2940 | if (state_test(STATE_FROZENDIR)) { | |
2941 | state_clear(STATE_FROZENDIR); | |
2942 | put(PIN_FROZEN); | |
2943 | ||
2944 | // unpin (may => FREEZEABLE) FIXME: is this order good? | |
2945 | if (is_auth() && !is_subtree_root()) | |
2946 | inode->auth_unpin(this); | |
2947 | ||
2948 | finish_waiting(WAIT_UNFREEZE); | |
2949 | } else { | |
2950 | finish_waiting(WAIT_FROZEN, -1); | |
2951 | ||
2952 | // still freezing. stop. | |
2953 | assert(state_test(STATE_FREEZINGDIR)); | |
2954 | state_clear(STATE_FREEZINGDIR); | |
2955 | auth_unpin(this); | |
2956 | ||
2957 | finish_waiting(WAIT_UNFREEZE); | |
2958 | } | |
2959 | } | |
2960 | ||
2961 | /** | |
2962 | * Slightly less complete than operator<<, because this is intended | |
2963 | * for identifying a directory and its state rather than for dumping | |
2964 | * debug output. | |
2965 | */ | |
2966 | void CDir::dump(Formatter *f) const | |
2967 | { | |
2968 | assert(f != NULL); | |
2969 | ||
2970 | f->dump_stream("path") << get_path(); | |
2971 | ||
2972 | f->dump_stream("dirfrag") << dirfrag(); | |
2973 | f->dump_int("snapid_first", first); | |
2974 | ||
2975 | f->dump_stream("projected_version") << get_projected_version(); | |
2976 | f->dump_stream("version") << get_version(); | |
2977 | f->dump_stream("committing_version") << get_committing_version(); | |
2978 | f->dump_stream("committed_version") << get_committed_version(); | |
2979 | ||
2980 | f->dump_bool("is_rep", is_rep()); | |
2981 | ||
2982 | if (get_dir_auth() != CDIR_AUTH_DEFAULT) { | |
2983 | if (get_dir_auth().second == CDIR_AUTH_UNKNOWN) { | |
2984 | f->dump_stream("dir_auth") << get_dir_auth().first; | |
2985 | } else { | |
2986 | f->dump_stream("dir_auth") << get_dir_auth(); | |
2987 | } | |
2988 | } else { | |
2989 | f->dump_string("dir_auth", ""); | |
2990 | } | |
2991 | ||
2992 | f->open_array_section("states"); | |
2993 | MDSCacheObject::dump_states(f); | |
2994 | if (state_test(CDir::STATE_COMPLETE)) f->dump_string("state", "complete"); | |
2995 | if (state_test(CDir::STATE_FREEZINGTREE)) f->dump_string("state", "freezingtree"); | |
2996 | if (state_test(CDir::STATE_FROZENTREE)) f->dump_string("state", "frozentree"); | |
2997 | if (state_test(CDir::STATE_FROZENDIR)) f->dump_string("state", "frozendir"); | |
2998 | if (state_test(CDir::STATE_FREEZINGDIR)) f->dump_string("state", "freezingdir"); | |
2999 | if (state_test(CDir::STATE_EXPORTBOUND)) f->dump_string("state", "exportbound"); | |
3000 | if (state_test(CDir::STATE_IMPORTBOUND)) f->dump_string("state", "importbound"); | |
3001 | if (state_test(CDir::STATE_BADFRAG)) f->dump_string("state", "badfrag"); | |
3002 | f->close_section(); | |
3003 | ||
3004 | MDSCacheObject::dump(f); | |
3005 | } | |
3006 | ||
3007 | /****** Scrub Stuff *******/ | |
3008 | ||
3009 | void CDir::scrub_info_create() const | |
3010 | { | |
3011 | assert(!scrub_infop); | |
3012 | ||
3013 | // break out of const-land to set up implicit initial state | |
3014 | CDir *me = const_cast<CDir*>(this); | |
3015 | fnode_t *fn = me->get_projected_fnode(); | |
3016 | ||
3017 | std::unique_ptr<scrub_info_t> si(new scrub_info_t()); | |
3018 | ||
3019 | si->last_recursive.version = si->recursive_start.version = | |
3020 | fn->recursive_scrub_version; | |
3021 | si->last_recursive.time = si->recursive_start.time = | |
3022 | fn->recursive_scrub_stamp; | |
3023 | ||
3024 | si->last_local.version = fn->localized_scrub_version; | |
3025 | si->last_local.time = fn->localized_scrub_stamp; | |
3026 | ||
3027 | me->scrub_infop.swap(si); | |
3028 | } | |
3029 | ||
3030 | void CDir::scrub_initialize(const ScrubHeaderRefConst& header) | |
3031 | { | |
3032 | dout(20) << __func__ << dendl; | |
3033 | assert(is_complete()); | |
3034 | assert(header != nullptr); | |
3035 | ||
3036 | // FIXME: weird implicit construction, is someone else meant | |
3037 | // to be calling scrub_info_create first? | |
3038 | scrub_info(); | |
3039 | assert(scrub_infop && !scrub_infop->directory_scrubbing); | |
3040 | ||
3041 | scrub_infop->recursive_start.version = get_projected_version(); | |
3042 | scrub_infop->recursive_start.time = ceph_clock_now(); | |
3043 | ||
3044 | scrub_infop->directories_to_scrub.clear(); | |
3045 | scrub_infop->directories_scrubbing.clear(); | |
3046 | scrub_infop->directories_scrubbed.clear(); | |
3047 | scrub_infop->others_to_scrub.clear(); | |
3048 | scrub_infop->others_scrubbing.clear(); | |
3049 | scrub_infop->others_scrubbed.clear(); | |
3050 | ||
3051 | for (map_t::iterator i = items.begin(); | |
3052 | i != items.end(); | |
3053 | ++i) { | |
3054 | // TODO: handle snapshot scrubbing | |
3055 | if (i->first.snapid != CEPH_NOSNAP) | |
3056 | continue; | |
3057 | ||
3058 | CDentry::linkage_t *dnl = i->second->get_projected_linkage(); | |
3059 | if (dnl->is_primary()) { | |
3060 | if (dnl->get_inode()->is_dir()) | |
3061 | scrub_infop->directories_to_scrub.insert(i->first); | |
3062 | else | |
3063 | scrub_infop->others_to_scrub.insert(i->first); | |
3064 | } else if (dnl->is_remote()) { | |
3065 | // TODO: check remote linkage | |
3066 | } | |
3067 | } | |
3068 | scrub_infop->directory_scrubbing = true; | |
3069 | scrub_infop->header = header; | |
3070 | } | |
3071 | ||
3072 | void CDir::scrub_finished() | |
3073 | { | |
3074 | dout(20) << __func__ << dendl; | |
3075 | assert(scrub_infop && scrub_infop->directory_scrubbing); | |
3076 | ||
3077 | assert(scrub_infop->directories_to_scrub.empty()); | |
3078 | assert(scrub_infop->directories_scrubbing.empty()); | |
3079 | scrub_infop->directories_scrubbed.clear(); | |
3080 | assert(scrub_infop->others_to_scrub.empty()); | |
3081 | assert(scrub_infop->others_scrubbing.empty()); | |
3082 | scrub_infop->others_scrubbed.clear(); | |
3083 | scrub_infop->directory_scrubbing = false; | |
3084 | ||
3085 | scrub_infop->last_recursive = scrub_infop->recursive_start; | |
3086 | scrub_infop->last_scrub_dirty = true; | |
3087 | } | |
3088 | ||
3089 | int CDir::_next_dentry_on_set(set<dentry_key_t>& dns, bool missing_okay, | |
3090 | MDSInternalContext *cb, CDentry **dnout) | |
3091 | { | |
3092 | dentry_key_t dnkey; | |
3093 | CDentry *dn; | |
3094 | ||
3095 | while (!dns.empty()) { | |
3096 | set<dentry_key_t>::iterator front = dns.begin(); | |
3097 | dnkey = *front; | |
3098 | dn = lookup(dnkey.name); | |
3099 | if (!dn) { | |
3100 | if (!is_complete() && | |
3101 | (!has_bloom() || is_in_bloom(dnkey.name))) { | |
3102 | // need to re-read this dirfrag | |
3103 | fetch(cb); | |
3104 | return EAGAIN; | |
3105 | } | |
3106 | // okay, we lost it | |
3107 | if (missing_okay) { | |
3108 | dout(15) << " we no longer have directory dentry " | |
3109 | << dnkey.name << ", assuming it got renamed" << dendl; | |
3110 | dns.erase(dnkey); | |
3111 | continue; | |
3112 | } else { | |
3113 | dout(5) << " we lost dentry " << dnkey.name | |
3114 | << ", bailing out because that's impossible!" << dendl; | |
3115 | ceph_abort(); | |
3116 | } | |
3117 | } | |
3118 | // okay, we got a dentry | |
3119 | dns.erase(dnkey); | |
3120 | ||
3121 | if (dn->get_projected_version() < scrub_infop->last_recursive.version && | |
3122 | !(scrub_infop->header->get_force())) { | |
3123 | dout(15) << " skip dentry " << dnkey.name | |
3124 | << ", no change since last scrub" << dendl; | |
3125 | continue; | |
3126 | } | |
3127 | ||
3128 | *dnout = dn; | |
3129 | return 0; | |
3130 | } | |
3131 | *dnout = NULL; | |
3132 | return ENOENT; | |
3133 | } | |
3134 | ||
3135 | int CDir::scrub_dentry_next(MDSInternalContext *cb, CDentry **dnout) | |
3136 | { | |
3137 | dout(20) << __func__ << dendl; | |
3138 | assert(scrub_infop && scrub_infop->directory_scrubbing); | |
3139 | ||
3140 | dout(20) << "trying to scrub directories underneath us" << dendl; | |
3141 | int rval = _next_dentry_on_set(scrub_infop->directories_to_scrub, true, | |
3142 | cb, dnout); | |
3143 | if (rval == 0) { | |
3144 | dout(20) << __func__ << " inserted to directories scrubbing: " | |
3145 | << *dnout << dendl; | |
3146 | scrub_infop->directories_scrubbing.insert((*dnout)->key()); | |
3147 | } else if (rval == EAGAIN) { | |
3148 | // we don't need to do anything else | |
3149 | } else { // we emptied out the directory scrub set | |
3150 | assert(rval == ENOENT); | |
3151 | dout(20) << "no directories left, moving on to other kinds of dentries" | |
3152 | << dendl; | |
3153 | ||
3154 | rval = _next_dentry_on_set(scrub_infop->others_to_scrub, false, cb, dnout); | |
3155 | if (rval == 0) { | |
3156 | dout(20) << __func__ << " inserted to others scrubbing: " | |
3157 | << *dnout << dendl; | |
3158 | scrub_infop->others_scrubbing.insert((*dnout)->key()); | |
3159 | } | |
3160 | } | |
3161 | dout(20) << " returning " << rval << " with dn=" << *dnout << dendl; | |
3162 | return rval; | |
3163 | } | |
3164 | ||
3165 | void CDir::scrub_dentries_scrubbing(list<CDentry*> *out_dentries) | |
3166 | { | |
3167 | dout(20) << __func__ << dendl; | |
3168 | assert(scrub_infop && scrub_infop->directory_scrubbing); | |
3169 | ||
3170 | for (set<dentry_key_t>::iterator i = | |
3171 | scrub_infop->directories_scrubbing.begin(); | |
3172 | i != scrub_infop->directories_scrubbing.end(); | |
3173 | ++i) { | |
3174 | CDentry *d = lookup(i->name, i->snapid); | |
3175 | assert(d); | |
3176 | out_dentries->push_back(d); | |
3177 | } | |
3178 | for (set<dentry_key_t>::iterator i = scrub_infop->others_scrubbing.begin(); | |
3179 | i != scrub_infop->others_scrubbing.end(); | |
3180 | ++i) { | |
3181 | CDentry *d = lookup(i->name, i->snapid); | |
3182 | assert(d); | |
3183 | out_dentries->push_back(d); | |
3184 | } | |
3185 | } | |
3186 | ||
3187 | void CDir::scrub_dentry_finished(CDentry *dn) | |
3188 | { | |
3189 | dout(20) << __func__ << " on dn " << *dn << dendl; | |
3190 | assert(scrub_infop && scrub_infop->directory_scrubbing); | |
3191 | dentry_key_t dn_key = dn->key(); | |
3192 | if (scrub_infop->directories_scrubbing.erase(dn_key)) { | |
3193 | scrub_infop->directories_scrubbed.insert(dn_key); | |
3194 | } else { | |
3195 | assert(scrub_infop->others_scrubbing.count(dn_key)); | |
3196 | scrub_infop->others_scrubbing.erase(dn_key); | |
3197 | scrub_infop->others_scrubbed.insert(dn_key); | |
3198 | } | |
3199 | } | |
3200 | ||
3201 | void CDir::scrub_maybe_delete_info() | |
3202 | { | |
3203 | if (scrub_infop && | |
3204 | !scrub_infop->directory_scrubbing && | |
3205 | !scrub_infop->need_scrub_local && | |
3206 | !scrub_infop->last_scrub_dirty && | |
3207 | !scrub_infop->pending_scrub_error && | |
3208 | scrub_infop->dirty_scrub_stamps.empty()) { | |
3209 | scrub_infop.reset(); | |
3210 | } | |
3211 | } | |
3212 | ||
3213 | bool CDir::scrub_local() | |
3214 | { | |
3215 | assert(is_complete()); | |
3216 | bool rval = check_rstats(true); | |
3217 | ||
3218 | scrub_info(); | |
3219 | if (rval) { | |
3220 | scrub_infop->last_local.time = ceph_clock_now(); | |
3221 | scrub_infop->last_local.version = get_projected_version(); | |
3222 | scrub_infop->pending_scrub_error = false; | |
3223 | scrub_infop->last_scrub_dirty = true; | |
3224 | } else { | |
3225 | scrub_infop->pending_scrub_error = true; | |
3226 | if (scrub_infop->header->get_repair()) | |
3227 | cache->repair_dirfrag_stats(this); | |
3228 | } | |
3229 | return rval; | |
3230 | } | |
3231 | ||
3232 | std::string CDir::get_path() const | |
3233 | { | |
3234 | std::string path; | |
3235 | get_inode()->make_path_string(path, true); | |
3236 | return path; | |
3237 | } | |
3238 | ||
3239 | bool CDir::should_split_fast() const | |
3240 | { | |
3241 | // Max size a fragment can be before trigger fast splitting | |
3242 | int fast_limit = g_conf->mds_bal_split_size * g_conf->mds_bal_fragment_fast_factor; | |
3243 | ||
3244 | // Fast path: the sum of accounted size and null dentries does not | |
3245 | // exceed threshold: we definitely are not over it. | |
3246 | if (get_frag_size() + get_num_head_null() <= fast_limit) { | |
3247 | return false; | |
3248 | } | |
3249 | ||
3250 | // Fast path: the accounted size of the frag exceeds threshold: we | |
3251 | // definitely are over it | |
3252 | if (get_frag_size() > fast_limit) { | |
3253 | return true; | |
3254 | } | |
3255 | ||
3256 | int64_t effective_size = 0; | |
3257 | ||
3258 | for (const auto &p : items) { | |
3259 | const CDentry *dn = p.second; | |
3260 | if (!dn->get_projected_linkage()->is_null()) { | |
3261 | effective_size++; | |
3262 | } | |
3263 | } | |
3264 | ||
3265 | return effective_size > fast_limit; | |
3266 | } | |
3267 |