| Author | Tokens | Token Proportion | Commits | Commit Proportion |
|---|---|---|---|---|
| Mark Fasheh | 5366 | 54.42% | 23 | 17.29% |
| Joel Becker | 1116 | 11.32% | 18 | 13.53% |
| Joseph Qi | 746 | 7.57% | 12 | 9.02% |
| Sunil Mushran | 680 | 6.90% | 11 | 8.27% |
| Srinivas Eeda | 642 | 6.51% | 3 | 2.26% |
| Jan Kara | 421 | 4.27% | 11 | 8.27% |
| Tao Ma | 232 | 2.35% | 5 | 3.76% |
| Valentin Vidic | 105 | 1.06% | 1 | 0.75% |
| Junxiao Bi | 75 | 0.76% | 3 | 2.26% |
| Younger Liu | 68 | 0.69% | 1 | 0.75% |
| Zhonghua Guo | 44 | 0.45% | 1 | 0.75% |
| Aneesh Kumar K.V | 41 | 0.42% | 1 | 0.75% |
| Heming Zhao via Ocfs2-devel | 37 | 0.38% | 1 | 0.75% |
| Mauricio Faria de Oliveira | 33 | 0.33% | 2 | 1.50% |
| Al Viro | 28 | 0.28% | 3 | 2.26% |
| Kai Li | 25 | 0.25% | 1 | 0.75% |
| Jeff Mahoney | 20 | 0.20% | 1 | 0.75% |
| David Howells | 18 | 0.18% | 1 | 0.75% |
| Jun Piao | 15 | 0.15% | 1 | 0.75% |
| zhangyi (F) | 14 | 0.14% | 1 | 0.75% |
| Eric Sesterhenn / Snakebyte | 14 | 0.14% | 1 | 0.75% |
| Miklos Szeredi | 12 | 0.12% | 1 | 0.75% |
| Christoph Hellwig | 11 | 0.11% | 4 | 3.01% |
| Christophe Jaillet | 11 | 0.11% | 2 | 1.50% |
| Yiwen Jiang | 10 | 0.10% | 1 | 0.75% |
| Jchao Sun | 10 | 0.10% | 1 | 0.75% |
| Leah Rumancik | 8 | 0.08% | 1 | 0.75% |
| Kees Cook | 7 | 0.07% | 2 | 1.50% |
| Zhang Mingyu | 6 | 0.06% | 1 | 0.75% |
| Yi Li | 6 | 0.06% | 1 | 0.75% |
| Andrew Morton | 5 | 0.05% | 1 | 0.75% |
| Arjan van de Ven | 5 | 0.05% | 1 | 0.75% |
| Ingo Molnar | 4 | 0.04% | 1 | 0.75% |
| Deepa Dinamani | 4 | 0.04% | 1 | 0.75% |
| Tiger Yang | 3 | 0.03% | 1 | 0.75% |
| Julia Lawall | 3 | 0.03% | 1 | 0.75% |
| Dave Kleikamp | 3 | 0.03% | 1 | 0.75% |
| Eric W. Biedermann | 2 | 0.02% | 1 | 0.75% |
| Theodore Y. Ts'o | 2 | 0.02% | 1 | 0.75% |
| Robert P. J. Day | 1 | 0.01% | 1 | 0.75% |
| Thomas Gleixner | 1 | 0.01% | 1 | 0.75% |
| Joe Perches | 1 | 0.01% | 1 | 0.75% |
| André Goddard Rosa | 1 | 0.01% | 1 | 0.75% |
| Gustavo A. R. Silva | 1 | 0.01% | 1 | 0.75% |
| Masahiro Yamada | 1 | 0.01% | 1 | 0.75% |
| Lucas De Marchi | 1 | 0.01% | 1 | 0.75% |
| Harshad Shirwadkar | 1 | 0.01% | 1 | 0.75% |
| Total | 9860 | 133 |
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448
// SPDX-License-Identifier: GPL-2.0-or-later /* * journal.c * * Defines functions of journalling api * * Copyright (C) 2003, 2004 Oracle. All rights reserved. */ #include <linux/fs.h> #include <linux/types.h> #include <linux/slab.h> #include <linux/highmem.h> #include <linux/kthread.h> #include <linux/time.h> #include <linux/random.h> #include <linux/delay.h> #include <linux/writeback.h> #include <cluster/masklog.h> #include "ocfs2.h" #include "alloc.h" #include "blockcheck.h" #include "dir.h" #include "dlmglue.h" #include "extent_map.h" #include "heartbeat.h" #include "inode.h" #include "journal.h" #include "localalloc.h" #include "slot_map.h" #include "super.h" #include "sysfile.h" #include "uptodate.h" #include "quota.h" #include "file.h" #include "namei.h" #include "buffer_head_io.h" #include "ocfs2_trace.h" DEFINE_SPINLOCK(trans_inc_lock); #define ORPHAN_SCAN_SCHEDULE_TIMEOUT 300000 static int ocfs2_force_read_journal(struct inode *inode); static int ocfs2_recover_node(struct ocfs2_super *osb, int node_num, int slot_num); static int __ocfs2_recovery_thread(void *arg); static int ocfs2_commit_cache(struct ocfs2_super *osb); static int __ocfs2_wait_on_mount(struct ocfs2_super *osb, int quota); static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb, int dirty, int replayed); static int ocfs2_trylock_journal(struct ocfs2_super *osb, int slot_num); static int ocfs2_recover_orphans(struct ocfs2_super *osb, int slot, enum ocfs2_orphan_reco_type orphan_reco_type); static int ocfs2_commit_thread(void *arg); static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal, int slot_num, struct ocfs2_dinode *la_dinode, struct ocfs2_dinode *tl_dinode, struct ocfs2_quota_recovery *qrec, enum ocfs2_orphan_reco_type orphan_reco_type); static inline int ocfs2_wait_on_mount(struct ocfs2_super *osb) { return __ocfs2_wait_on_mount(osb, 0); } static inline int ocfs2_wait_on_quotas(struct ocfs2_super *osb) { return __ocfs2_wait_on_mount(osb, 1); } /* * This replay_map is to track online/offline slots, so we could recover * offline slots during recovery and mount */ enum ocfs2_replay_state { REPLAY_UNNEEDED = 0, /* Replay is not needed, so ignore this map */ REPLAY_NEEDED, /* Replay slots marked in rm_replay_slots */ REPLAY_DONE /* Replay was already queued */ }; struct ocfs2_replay_map { unsigned int rm_slots; enum ocfs2_replay_state rm_state; unsigned char rm_replay_slots[] __counted_by(rm_slots); }; static void ocfs2_replay_map_set_state(struct ocfs2_super *osb, int state) { if (!osb->replay_map) return; /* If we've already queued the replay, we don't have any more to do */ if (osb->replay_map->rm_state == REPLAY_DONE) return; osb->replay_map->rm_state = state; } int ocfs2_compute_replay_slots(struct ocfs2_super *osb) { struct ocfs2_replay_map *replay_map; int i, node_num; /* If replay map is already set, we don't do it again */ if (osb->replay_map) return 0; replay_map = kzalloc(struct_size(replay_map, rm_replay_slots, osb->max_slots), GFP_KERNEL); if (!replay_map) { mlog_errno(-ENOMEM); return -ENOMEM; } spin_lock(&osb->osb_lock); replay_map->rm_slots = osb->max_slots; replay_map->rm_state = REPLAY_UNNEEDED; /* set rm_replay_slots for offline slot(s) */ for (i = 0; i < replay_map->rm_slots; i++) { if (ocfs2_slot_to_node_num_locked(osb, i, &node_num) == -ENOENT) replay_map->rm_replay_slots[i] = 1; } osb->replay_map = replay_map; spin_unlock(&osb->osb_lock); return 0; } static void ocfs2_queue_replay_slots(struct ocfs2_super *osb, enum ocfs2_orphan_reco_type orphan_reco_type) { struct ocfs2_replay_map *replay_map = osb->replay_map; int i; if (!replay_map) return; if (replay_map->rm_state != REPLAY_NEEDED) return; for (i = 0; i < replay_map->rm_slots; i++) if (replay_map->rm_replay_slots[i]) ocfs2_queue_recovery_completion(osb->journal, i, NULL, NULL, NULL, orphan_reco_type); replay_map->rm_state = REPLAY_DONE; } void ocfs2_free_replay_slots(struct ocfs2_super *osb) { struct ocfs2_replay_map *replay_map = osb->replay_map; if (!osb->replay_map) return; kfree(replay_map); osb->replay_map = NULL; } int ocfs2_recovery_init(struct ocfs2_super *osb) { struct ocfs2_recovery_map *rm; mutex_init(&osb->recovery_lock); osb->disable_recovery = 0; osb->recovery_thread_task = NULL; init_waitqueue_head(&osb->recovery_event); rm = kzalloc(struct_size(rm, rm_entries, osb->max_slots), GFP_KERNEL); if (!rm) { mlog_errno(-ENOMEM); return -ENOMEM; } osb->recovery_map = rm; return 0; } /* we can't grab the goofy sem lock from inside wait_event, so we use * memory barriers to make sure that we'll see the null task before * being woken up */ static int ocfs2_recovery_thread_running(struct ocfs2_super *osb) { mb(); return osb->recovery_thread_task != NULL; } void ocfs2_recovery_exit(struct ocfs2_super *osb) { struct ocfs2_recovery_map *rm; /* disable any new recovery threads and wait for any currently * running ones to exit. Do this before setting the vol_state. */ mutex_lock(&osb->recovery_lock); osb->disable_recovery = 1; mutex_unlock(&osb->recovery_lock); wait_event(osb->recovery_event, !ocfs2_recovery_thread_running(osb)); /* At this point, we know that no more recovery threads can be * launched, so wait for any recovery completion work to * complete. */ if (osb->ocfs2_wq) flush_workqueue(osb->ocfs2_wq); /* * Now that recovery is shut down, and the osb is about to be * freed, the osb_lock is not taken here. */ rm = osb->recovery_map; /* XXX: Should we bug if there are dirty entries? */ kfree(rm); } static int __ocfs2_recovery_map_test(struct ocfs2_super *osb, unsigned int node_num) { int i; struct ocfs2_recovery_map *rm = osb->recovery_map; assert_spin_locked(&osb->osb_lock); for (i = 0; i < rm->rm_used; i++) { if (rm->rm_entries[i] == node_num) return 1; } return 0; } /* Behaves like test-and-set. Returns the previous value */ static int ocfs2_recovery_map_set(struct ocfs2_super *osb, unsigned int node_num) { struct ocfs2_recovery_map *rm = osb->recovery_map; spin_lock(&osb->osb_lock); if (__ocfs2_recovery_map_test(osb, node_num)) { spin_unlock(&osb->osb_lock); return 1; } /* XXX: Can this be exploited? Not from o2dlm... */ BUG_ON(rm->rm_used >= osb->max_slots); rm->rm_entries[rm->rm_used] = node_num; rm->rm_used++; spin_unlock(&osb->osb_lock); return 0; } static void ocfs2_recovery_map_clear(struct ocfs2_super *osb, unsigned int node_num) { int i; struct ocfs2_recovery_map *rm = osb->recovery_map; spin_lock(&osb->osb_lock); for (i = 0; i < rm->rm_used; i++) { if (rm->rm_entries[i] == node_num) break; } if (i < rm->rm_used) { /* XXX: be careful with the pointer math */ memmove(&(rm->rm_entries[i]), &(rm->rm_entries[i + 1]), (rm->rm_used - i - 1) * sizeof(unsigned int)); rm->rm_used--; } spin_unlock(&osb->osb_lock); } static int ocfs2_commit_cache(struct ocfs2_super *osb) { int status = 0; unsigned int flushed; struct ocfs2_journal *journal = NULL; journal = osb->journal; /* Flush all pending commits and checkpoint the journal. */ down_write(&journal->j_trans_barrier); flushed = atomic_read(&journal->j_num_trans); trace_ocfs2_commit_cache_begin(flushed); if (flushed == 0) { up_write(&journal->j_trans_barrier); goto finally; } jbd2_journal_lock_updates(journal->j_journal); status = jbd2_journal_flush(journal->j_journal, 0); jbd2_journal_unlock_updates(journal->j_journal); if (status < 0) { up_write(&journal->j_trans_barrier); mlog_errno(status); goto finally; } ocfs2_inc_trans_id(journal); flushed = atomic_read(&journal->j_num_trans); atomic_set(&journal->j_num_trans, 0); up_write(&journal->j_trans_barrier); trace_ocfs2_commit_cache_end(journal->j_trans_id, flushed); ocfs2_wake_downconvert_thread(osb); wake_up(&journal->j_checkpointed); finally: return status; } handle_t *ocfs2_start_trans(struct ocfs2_super *osb, int max_buffs) { journal_t *journal = osb->journal->j_journal; handle_t *handle; BUG_ON(!osb || !osb->journal->j_journal); if (ocfs2_is_hard_readonly(osb)) return ERR_PTR(-EROFS); BUG_ON(osb->journal->j_state == OCFS2_JOURNAL_FREE); BUG_ON(max_buffs <= 0); /* Nested transaction? Just return the handle... */ if (journal_current_handle()) return jbd2_journal_start(journal, max_buffs); sb_start_intwrite(osb->sb); down_read(&osb->journal->j_trans_barrier); handle = jbd2_journal_start(journal, max_buffs); if (IS_ERR(handle)) { up_read(&osb->journal->j_trans_barrier); sb_end_intwrite(osb->sb); mlog_errno(PTR_ERR(handle)); if (is_journal_aborted(journal)) { ocfs2_abort(osb->sb, "Detected aborted journal\n"); handle = ERR_PTR(-EROFS); } } else { if (!ocfs2_mount_local(osb)) atomic_inc(&(osb->journal->j_num_trans)); } return handle; } int ocfs2_commit_trans(struct ocfs2_super *osb, handle_t *handle) { int ret, nested; struct ocfs2_journal *journal = osb->journal; BUG_ON(!handle); nested = handle->h_ref > 1; ret = jbd2_journal_stop(handle); if (ret < 0) mlog_errno(ret); if (!nested) { up_read(&journal->j_trans_barrier); sb_end_intwrite(osb->sb); } return ret; } /* * 'nblocks' is what you want to add to the current transaction. * * This might call jbd2_journal_restart() which will commit dirty buffers * and then restart the transaction. Before calling * ocfs2_extend_trans(), any changed blocks should have been * dirtied. After calling it, all blocks which need to be changed must * go through another set of journal_access/journal_dirty calls. * * WARNING: This will not release any semaphores or disk locks taken * during the transaction, so make sure they were taken *before* * start_trans or we'll have ordering deadlocks. * * WARNING2: Note that we do *not* drop j_trans_barrier here. This is * good because transaction ids haven't yet been recorded on the * cluster locks associated with this handle. */ int ocfs2_extend_trans(handle_t *handle, int nblocks) { int status, old_nblocks; BUG_ON(!handle); BUG_ON(nblocks < 0); if (!nblocks) return 0; old_nblocks = jbd2_handle_buffer_credits(handle); trace_ocfs2_extend_trans(old_nblocks, nblocks); #ifdef CONFIG_OCFS2_DEBUG_FS status = 1; #else status = jbd2_journal_extend(handle, nblocks, 0); if (status < 0) { mlog_errno(status); goto bail; } #endif if (status > 0) { trace_ocfs2_extend_trans_restart(old_nblocks + nblocks); status = jbd2_journal_restart(handle, old_nblocks + nblocks); if (status < 0) { mlog_errno(status); goto bail; } } status = 0; bail: return status; } /* * Make sure handle has at least 'nblocks' credits available. If it does not * have that many credits available, we will try to extend the handle to have * enough credits. If that fails, we will restart transaction to have enough * credits. Similar notes regarding data consistency and locking implications * as for ocfs2_extend_trans() apply here. */ int ocfs2_assure_trans_credits(handle_t *handle, int nblocks) { int old_nblks = jbd2_handle_buffer_credits(handle); trace_ocfs2_assure_trans_credits(old_nblks); if (old_nblks >= nblocks) return 0; return ocfs2_extend_trans(handle, nblocks - old_nblks); } /* * If we have fewer than thresh credits, extend by OCFS2_MAX_TRANS_DATA. * If that fails, restart the transaction & regain write access for the * buffer head which is used for metadata modifications. * Taken from Ext4: extend_or_restart_transaction() */ int ocfs2_allocate_extend_trans(handle_t *handle, int thresh) { int status, old_nblks; BUG_ON(!handle); old_nblks = jbd2_handle_buffer_credits(handle); trace_ocfs2_allocate_extend_trans(old_nblks, thresh); if (old_nblks < thresh) return 0; status = jbd2_journal_extend(handle, OCFS2_MAX_TRANS_DATA, 0); if (status < 0) { mlog_errno(status); goto bail; } if (status > 0) { status = jbd2_journal_restart(handle, OCFS2_MAX_TRANS_DATA); if (status < 0) mlog_errno(status); } bail: return status; } static inline struct ocfs2_triggers *to_ocfs2_trigger(struct jbd2_buffer_trigger_type *triggers) { return container_of(triggers, struct ocfs2_triggers, ot_triggers); } static void ocfs2_frozen_trigger(struct jbd2_buffer_trigger_type *triggers, struct buffer_head *bh, void *data, size_t size) { struct ocfs2_triggers *ot = to_ocfs2_trigger(triggers); /* * We aren't guaranteed to have the superblock here, so we * must unconditionally compute the ecc data. * __ocfs2_journal_access() will only set the triggers if * metaecc is enabled. */ ocfs2_block_check_compute(data, size, data + ot->ot_offset); } /* * Quota blocks have their own trigger because the struct ocfs2_block_check * offset depends on the blocksize. */ static void ocfs2_dq_frozen_trigger(struct jbd2_buffer_trigger_type *triggers, struct buffer_head *bh, void *data, size_t size) { struct ocfs2_disk_dqtrailer *dqt = ocfs2_block_dqtrailer(size, data); /* * We aren't guaranteed to have the superblock here, so we * must unconditionally compute the ecc data. * __ocfs2_journal_access() will only set the triggers if * metaecc is enabled. */ ocfs2_block_check_compute(data, size, &dqt->dq_check); } /* * Directory blocks also have their own trigger because the * struct ocfs2_block_check offset depends on the blocksize. */ static void ocfs2_db_frozen_trigger(struct jbd2_buffer_trigger_type *triggers, struct buffer_head *bh, void *data, size_t size) { struct ocfs2_dir_block_trailer *trailer = ocfs2_dir_trailer_from_size(size, data); /* * We aren't guaranteed to have the superblock here, so we * must unconditionally compute the ecc data. * __ocfs2_journal_access() will only set the triggers if * metaecc is enabled. */ ocfs2_block_check_compute(data, size, &trailer->db_check); } static void ocfs2_abort_trigger(struct jbd2_buffer_trigger_type *triggers, struct buffer_head *bh) { struct ocfs2_triggers *ot = to_ocfs2_trigger(triggers); mlog(ML_ERROR, "ocfs2_abort_trigger called by JBD2. bh = 0x%lx, " "bh->b_blocknr = %llu\n", (unsigned long)bh, (unsigned long long)bh->b_blocknr); ocfs2_error(ot->sb, "JBD2 has aborted our journal, ocfs2 cannot continue\n"); } static void ocfs2_setup_csum_triggers(struct super_block *sb, enum ocfs2_journal_trigger_type type, struct ocfs2_triggers *ot) { BUG_ON(type >= OCFS2_JOURNAL_TRIGGER_COUNT); switch (type) { case OCFS2_JTR_DI: ot->ot_triggers.t_frozen = ocfs2_frozen_trigger; ot->ot_offset = offsetof(struct ocfs2_dinode, i_check); break; case OCFS2_JTR_EB: ot->ot_triggers.t_frozen = ocfs2_frozen_trigger; ot->ot_offset = offsetof(struct ocfs2_extent_block, h_check); break; case OCFS2_JTR_RB: ot->ot_triggers.t_frozen = ocfs2_frozen_trigger; ot->ot_offset = offsetof(struct ocfs2_refcount_block, rf_check); break; case OCFS2_JTR_GD: ot->ot_triggers.t_frozen = ocfs2_frozen_trigger; ot->ot_offset = offsetof(struct ocfs2_group_desc, bg_check); break; case OCFS2_JTR_DB: ot->ot_triggers.t_frozen = ocfs2_db_frozen_trigger; break; case OCFS2_JTR_XB: ot->ot_triggers.t_frozen = ocfs2_frozen_trigger; ot->ot_offset = offsetof(struct ocfs2_xattr_block, xb_check); break; case OCFS2_JTR_DQ: ot->ot_triggers.t_frozen = ocfs2_dq_frozen_trigger; break; case OCFS2_JTR_DR: ot->ot_triggers.t_frozen = ocfs2_frozen_trigger; ot->ot_offset = offsetof(struct ocfs2_dx_root_block, dr_check); break; case OCFS2_JTR_DL: ot->ot_triggers.t_frozen = ocfs2_frozen_trigger; ot->ot_offset = offsetof(struct ocfs2_dx_leaf, dl_check); break; case OCFS2_JTR_NONE: /* To make compiler happy... */ return; } ot->ot_triggers.t_abort = ocfs2_abort_trigger; ot->sb = sb; } void ocfs2_initialize_journal_triggers(struct super_block *sb, struct ocfs2_triggers triggers[]) { enum ocfs2_journal_trigger_type type; for (type = OCFS2_JTR_DI; type < OCFS2_JOURNAL_TRIGGER_COUNT; type++) ocfs2_setup_csum_triggers(sb, type, &triggers[type]); } static int __ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, struct ocfs2_triggers *triggers, int type) { int status; struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); BUG_ON(!ci || !ci->ci_ops); BUG_ON(!handle); BUG_ON(!bh); trace_ocfs2_journal_access( (unsigned long long)ocfs2_metadata_cache_owner(ci), (unsigned long long)bh->b_blocknr, type, bh->b_size); /* we can safely remove this assertion after testing. */ if (!buffer_uptodate(bh)) { mlog(ML_ERROR, "giving me a buffer that's not uptodate!\n"); mlog(ML_ERROR, "b_blocknr=%llu, b_state=0x%lx\n", (unsigned long long)bh->b_blocknr, bh->b_state); lock_buffer(bh); /* * A previous transaction with a couple of buffer heads fail * to checkpoint, so all the bhs are marked as BH_Write_EIO. * For current transaction, the bh is just among those error * bhs which previous transaction handle. We can't just clear * its BH_Write_EIO and reuse directly, since other bhs are * not written to disk yet and that will cause metadata * inconsistency. So we should set fs read-only to avoid * further damage. */ if (buffer_write_io_error(bh) && !buffer_uptodate(bh)) { unlock_buffer(bh); return ocfs2_error(osb->sb, "A previous attempt to " "write this buffer head failed\n"); } unlock_buffer(bh); } /* Set the current transaction information on the ci so * that the locking code knows whether it can drop it's locks * on this ci or not. We're protected from the commit * thread updating the current transaction id until * ocfs2_commit_trans() because ocfs2_start_trans() took * j_trans_barrier for us. */ ocfs2_set_ci_lock_trans(osb->journal, ci); ocfs2_metadata_cache_io_lock(ci); switch (type) { case OCFS2_JOURNAL_ACCESS_CREATE: case OCFS2_JOURNAL_ACCESS_WRITE: status = jbd2_journal_get_write_access(handle, bh); break; case OCFS2_JOURNAL_ACCESS_UNDO: status = jbd2_journal_get_undo_access(handle, bh); break; default: status = -EINVAL; mlog(ML_ERROR, "Unknown access type!\n"); } if (!status && ocfs2_meta_ecc(osb) && triggers) jbd2_journal_set_triggers(bh, &triggers->ot_triggers); ocfs2_metadata_cache_io_unlock(ci); if (status < 0) mlog(ML_ERROR, "Error %d getting %d access to buffer!\n", status, type); return status; } int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_DI], type); } int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_EB], type); } int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_RB], type); } int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_GD], type); } int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_DB], type); } int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_XB], type); } int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_DQ], type); } int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_DR], type); } int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { struct ocfs2_super *osb = OCFS2_SB(ocfs2_metadata_cache_get_super(ci)); return __ocfs2_journal_access(handle, ci, bh, &osb->s_journal_triggers[OCFS2_JTR_DL], type); } int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *bh, int type) { return __ocfs2_journal_access(handle, ci, bh, NULL, type); } void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh) { int status; trace_ocfs2_journal_dirty((unsigned long long)bh->b_blocknr); status = jbd2_journal_dirty_metadata(handle, bh); if (status) { mlog_errno(status); if (!is_handle_aborted(handle)) { journal_t *journal = handle->h_transaction->t_journal; mlog(ML_ERROR, "jbd2_journal_dirty_metadata failed: " "handle type %u started at line %u, credits %u/%u " "errcode %d. Aborting transaction and journal.\n", handle->h_type, handle->h_line_no, handle->h_requested_credits, jbd2_handle_buffer_credits(handle), status); handle->h_err = status; jbd2_journal_abort_handle(handle); jbd2_journal_abort(journal, status); } } } #define OCFS2_DEFAULT_COMMIT_INTERVAL (HZ * JBD2_DEFAULT_MAX_COMMIT_AGE) void ocfs2_set_journal_params(struct ocfs2_super *osb) { journal_t *journal = osb->journal->j_journal; unsigned long commit_interval = OCFS2_DEFAULT_COMMIT_INTERVAL; if (osb->osb_commit_interval) commit_interval = osb->osb_commit_interval; write_lock(&journal->j_state_lock); journal->j_commit_interval = commit_interval; if (osb->s_mount_opt & OCFS2_MOUNT_BARRIER) journal->j_flags |= JBD2_BARRIER; else journal->j_flags &= ~JBD2_BARRIER; write_unlock(&journal->j_state_lock); } /* * alloc & initialize skeleton for journal structure. * ocfs2_journal_init() will make fs have journal ability. */ int ocfs2_journal_alloc(struct ocfs2_super *osb) { int status = 0; struct ocfs2_journal *journal; journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL); if (!journal) { mlog(ML_ERROR, "unable to alloc journal\n"); status = -ENOMEM; goto bail; } osb->journal = journal; journal->j_osb = osb; atomic_set(&journal->j_num_trans, 0); init_rwsem(&journal->j_trans_barrier); init_waitqueue_head(&journal->j_checkpointed); spin_lock_init(&journal->j_lock); journal->j_trans_id = 1UL; INIT_LIST_HEAD(&journal->j_la_cleanups); INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); journal->j_state = OCFS2_JOURNAL_FREE; bail: return status; } static int ocfs2_journal_submit_inode_data_buffers(struct jbd2_inode *jinode) { struct address_space *mapping = jinode->i_vfs_inode->i_mapping; struct writeback_control wbc = { .sync_mode = WB_SYNC_ALL, .nr_to_write = mapping->nrpages * 2, .range_start = jinode->i_dirty_start, .range_end = jinode->i_dirty_end, }; return filemap_fdatawrite_wbc(mapping, &wbc); } int ocfs2_journal_init(struct ocfs2_super *osb, int *dirty) { int status = -1; struct inode *inode = NULL; /* the journal inode */ journal_t *j_journal = NULL; struct ocfs2_journal *journal = osb->journal; struct ocfs2_dinode *di = NULL; struct buffer_head *bh = NULL; int inode_lock = 0; BUG_ON(!journal); /* already have the inode for our journal */ inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE, osb->slot_num); if (inode == NULL) { status = -EACCES; mlog_errno(status); goto done; } if (is_bad_inode(inode)) { mlog(ML_ERROR, "access error (bad inode)\n"); iput(inode); inode = NULL; status = -EACCES; goto done; } SET_INODE_JOURNAL(inode); OCFS2_I(inode)->ip_open_count++; /* Skip recovery waits here - journal inode metadata never * changes in a live cluster so it can be considered an * exception to the rule. */ status = ocfs2_inode_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY); if (status < 0) { if (status != -ERESTARTSYS) mlog(ML_ERROR, "Could not get lock on journal!\n"); goto done; } inode_lock = 1; di = (struct ocfs2_dinode *)bh->b_data; if (i_size_read(inode) < OCFS2_MIN_JOURNAL_SIZE) { mlog(ML_ERROR, "Journal file size (%lld) is too small!\n", i_size_read(inode)); status = -EINVAL; goto done; } trace_ocfs2_journal_init(i_size_read(inode), (unsigned long long)inode->i_blocks, OCFS2_I(inode)->ip_clusters); /* call the kernels journal init function now */ j_journal = jbd2_journal_init_inode(inode); if (IS_ERR(j_journal)) { mlog(ML_ERROR, "Linux journal layer error\n"); status = PTR_ERR(j_journal); goto done; } trace_ocfs2_journal_init_maxlen(j_journal->j_total_len); *dirty = (le32_to_cpu(di->id1.journal1.ij_flags) & OCFS2_JOURNAL_DIRTY_FL); journal->j_journal = j_journal; journal->j_journal->j_submit_inode_data_buffers = ocfs2_journal_submit_inode_data_buffers; journal->j_journal->j_finish_inode_data_buffers = jbd2_journal_finish_inode_data_buffers; journal->j_inode = inode; journal->j_bh = bh; ocfs2_set_journal_params(osb); journal->j_state = OCFS2_JOURNAL_LOADED; status = 0; done: if (status < 0) { if (inode_lock) ocfs2_inode_unlock(inode, 1); brelse(bh); if (inode) { OCFS2_I(inode)->ip_open_count--; iput(inode); } } return status; } static void ocfs2_bump_recovery_generation(struct ocfs2_dinode *di) { le32_add_cpu(&(di->id1.journal1.ij_recovery_generation), 1); } static u32 ocfs2_get_recovery_generation(struct ocfs2_dinode *di) { return le32_to_cpu(di->id1.journal1.ij_recovery_generation); } static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb, int dirty, int replayed) { int status; unsigned int flags; struct ocfs2_journal *journal = osb->journal; struct buffer_head *bh = journal->j_bh; struct ocfs2_dinode *fe; fe = (struct ocfs2_dinode *)bh->b_data; /* The journal bh on the osb always comes from ocfs2_journal_init() * and was validated there inside ocfs2_inode_lock_full(). It's a * code bug if we mess it up. */ BUG_ON(!OCFS2_IS_VALID_DINODE(fe)); flags = le32_to_cpu(fe->id1.journal1.ij_flags); if (dirty) flags |= OCFS2_JOURNAL_DIRTY_FL; else flags &= ~OCFS2_JOURNAL_DIRTY_FL; fe->id1.journal1.ij_flags = cpu_to_le32(flags); if (replayed) ocfs2_bump_recovery_generation(fe); ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &fe->i_check); status = ocfs2_write_block(osb, bh, INODE_CACHE(journal->j_inode)); if (status < 0) mlog_errno(status); return status; } /* * If the journal has been kmalloc'd it needs to be freed after this * call. */ void ocfs2_journal_shutdown(struct ocfs2_super *osb) { struct ocfs2_journal *journal = NULL; int status = 0; struct inode *inode = NULL; int num_running_trans = 0; BUG_ON(!osb); journal = osb->journal; if (!journal) goto done; inode = journal->j_inode; if (journal->j_state != OCFS2_JOURNAL_LOADED) goto done; /* need to inc inode use count - jbd2_journal_destroy will iput. */ if (!igrab(inode)) BUG(); num_running_trans = atomic_read(&(journal->j_num_trans)); trace_ocfs2_journal_shutdown(num_running_trans); /* Do a commit_cache here. It will flush our journal, *and* * release any locks that are still held. * set the SHUTDOWN flag and release the trans lock. * the commit thread will take the trans lock for us below. */ journal->j_state = OCFS2_JOURNAL_IN_SHUTDOWN; /* The OCFS2_JOURNAL_IN_SHUTDOWN will signal to commit_cache to not * drop the trans_lock (which we want to hold until we * completely destroy the journal. */ if (osb->commit_task) { /* Wait for the commit thread */ trace_ocfs2_journal_shutdown_wait(osb->commit_task); kthread_stop(osb->commit_task); osb->commit_task = NULL; } BUG_ON(atomic_read(&(journal->j_num_trans)) != 0); if (ocfs2_mount_local(osb) && (journal->j_journal->j_flags & JBD2_LOADED)) { jbd2_journal_lock_updates(journal->j_journal); status = jbd2_journal_flush(journal->j_journal, 0); jbd2_journal_unlock_updates(journal->j_journal); if (status < 0) mlog_errno(status); } /* Shutdown the kernel journal system */ if (!jbd2_journal_destroy(journal->j_journal) && !status) { /* * Do not toggle if flush was unsuccessful otherwise * will leave dirty metadata in a "clean" journal */ status = ocfs2_journal_toggle_dirty(osb, 0, 0); if (status < 0) mlog_errno(status); } journal->j_journal = NULL; OCFS2_I(inode)->ip_open_count--; /* unlock our journal */ ocfs2_inode_unlock(inode, 1); brelse(journal->j_bh); journal->j_bh = NULL; journal->j_state = OCFS2_JOURNAL_FREE; done: iput(inode); kfree(journal); osb->journal = NULL; } static void ocfs2_clear_journal_error(struct super_block *sb, journal_t *journal, int slot) { int olderr; olderr = jbd2_journal_errno(journal); if (olderr) { mlog(ML_ERROR, "File system error %d recorded in " "journal %u.\n", olderr, slot); mlog(ML_ERROR, "File system on device %s needs checking.\n", sb->s_id); jbd2_journal_ack_err(journal); jbd2_journal_clear_err(journal); } } int ocfs2_journal_load(struct ocfs2_journal *journal, int local, int replayed) { int status = 0; struct ocfs2_super *osb; BUG_ON(!journal); osb = journal->j_osb; status = jbd2_journal_load(journal->j_journal); if (status < 0) { mlog(ML_ERROR, "Failed to load journal!\n"); goto done; } ocfs2_clear_journal_error(osb->sb, journal->j_journal, osb->slot_num); if (replayed) { jbd2_journal_lock_updates(journal->j_journal); status = jbd2_journal_flush(journal->j_journal, 0); jbd2_journal_unlock_updates(journal->j_journal); if (status < 0) mlog_errno(status); } status = ocfs2_journal_toggle_dirty(osb, 1, replayed); if (status < 0) { mlog_errno(status); goto done; } /* Launch the commit thread */ if (!local) { osb->commit_task = kthread_run(ocfs2_commit_thread, osb, "ocfs2cmt-%s", osb->uuid_str); if (IS_ERR(osb->commit_task)) { status = PTR_ERR(osb->commit_task); osb->commit_task = NULL; mlog(ML_ERROR, "unable to launch ocfs2commit thread, " "error=%d", status); goto done; } } else osb->commit_task = NULL; done: return status; } /* 'full' flag tells us whether we clear out all blocks or if we just * mark the journal clean */ int ocfs2_journal_wipe(struct ocfs2_journal *journal, int full) { int status; BUG_ON(!journal); status = jbd2_journal_wipe(journal->j_journal, full); if (status < 0) { mlog_errno(status); goto bail; } status = ocfs2_journal_toggle_dirty(journal->j_osb, 0, 0); if (status < 0) mlog_errno(status); bail: return status; } static int ocfs2_recovery_completed(struct ocfs2_super *osb) { int empty; struct ocfs2_recovery_map *rm = osb->recovery_map; spin_lock(&osb->osb_lock); empty = (rm->rm_used == 0); spin_unlock(&osb->osb_lock); return empty; } void ocfs2_wait_for_recovery(struct ocfs2_super *osb) { wait_event(osb->recovery_event, ocfs2_recovery_completed(osb)); } /* * JBD Might read a cached version of another nodes journal file. We * don't want this as this file changes often and we get no * notification on those changes. The only way to be sure that we've * got the most up to date version of those blocks then is to force * read them off disk. Just searching through the buffer cache won't * work as there may be pages backing this file which are still marked * up to date. We know things can't change on this file underneath us * as we have the lock by now :) */ static int ocfs2_force_read_journal(struct inode *inode) { int status = 0; int i; u64 v_blkno, p_blkno, p_blocks, num_blocks; struct buffer_head *bh = NULL; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); num_blocks = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode)); v_blkno = 0; while (v_blkno < num_blocks) { status = ocfs2_extent_map_get_blocks(inode, v_blkno, &p_blkno, &p_blocks, NULL); if (status < 0) { mlog_errno(status); goto bail; } for (i = 0; i < p_blocks; i++, p_blkno++) { bh = __find_get_block(osb->sb->s_bdev, p_blkno, osb->sb->s_blocksize); /* block not cached. */ if (!bh) continue; brelse(bh); bh = NULL; /* We are reading journal data which should not * be put in the uptodate cache. */ status = ocfs2_read_blocks_sync(osb, p_blkno, 1, &bh); if (status < 0) { mlog_errno(status); goto bail; } brelse(bh); bh = NULL; } v_blkno += p_blocks; } bail: return status; } struct ocfs2_la_recovery_item { struct list_head lri_list; int lri_slot; struct ocfs2_dinode *lri_la_dinode; struct ocfs2_dinode *lri_tl_dinode; struct ocfs2_quota_recovery *lri_qrec; enum ocfs2_orphan_reco_type lri_orphan_reco_type; }; /* Does the second half of the recovery process. By this point, the * node is marked clean and can actually be considered recovered, * hence it's no longer in the recovery map, but there's still some * cleanup we can do which shouldn't happen within the recovery thread * as locking in that context becomes very difficult if we are to take * recovering nodes into account. * * NOTE: This function can and will sleep on recovery of other nodes * during cluster locking, just like any other ocfs2 process. */ void ocfs2_complete_recovery(struct work_struct *work) { int ret = 0; struct ocfs2_journal *journal = container_of(work, struct ocfs2_journal, j_recovery_work); struct ocfs2_super *osb = journal->j_osb; struct ocfs2_dinode *la_dinode, *tl_dinode; struct ocfs2_la_recovery_item *item, *n; struct ocfs2_quota_recovery *qrec; enum ocfs2_orphan_reco_type orphan_reco_type; LIST_HEAD(tmp_la_list); trace_ocfs2_complete_recovery( (unsigned long long)OCFS2_I(journal->j_inode)->ip_blkno); spin_lock(&journal->j_lock); list_splice_init(&journal->j_la_cleanups, &tmp_la_list); spin_unlock(&journal->j_lock); list_for_each_entry_safe(item, n, &tmp_la_list, lri_list) { list_del_init(&item->lri_list); ocfs2_wait_on_quotas(osb); la_dinode = item->lri_la_dinode; tl_dinode = item->lri_tl_dinode; qrec = item->lri_qrec; orphan_reco_type = item->lri_orphan_reco_type; trace_ocfs2_complete_recovery_slot(item->lri_slot, la_dinode ? le64_to_cpu(la_dinode->i_blkno) : 0, tl_dinode ? le64_to_cpu(tl_dinode->i_blkno) : 0, qrec); if (la_dinode) { ret = ocfs2_complete_local_alloc_recovery(osb, la_dinode); if (ret < 0) mlog_errno(ret); kfree(la_dinode); } if (tl_dinode) { ret = ocfs2_complete_truncate_log_recovery(osb, tl_dinode); if (ret < 0) mlog_errno(ret); kfree(tl_dinode); } ret = ocfs2_recover_orphans(osb, item->lri_slot, orphan_reco_type); if (ret < 0) mlog_errno(ret); if (qrec) { ret = ocfs2_finish_quota_recovery(osb, qrec, item->lri_slot); if (ret < 0) mlog_errno(ret); /* Recovery info is already freed now */ } kfree(item); } trace_ocfs2_complete_recovery_end(ret); } /* NOTE: This function always eats your references to la_dinode and * tl_dinode, either manually on error, or by passing them to * ocfs2_complete_recovery */ static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal, int slot_num, struct ocfs2_dinode *la_dinode, struct ocfs2_dinode *tl_dinode, struct ocfs2_quota_recovery *qrec, enum ocfs2_orphan_reco_type orphan_reco_type) { struct ocfs2_la_recovery_item *item; item = kmalloc(sizeof(struct ocfs2_la_recovery_item), GFP_NOFS); if (!item) { /* Though we wish to avoid it, we are in fact safe in * skipping local alloc cleanup as fsck.ocfs2 is more * than capable of reclaiming unused space. */ kfree(la_dinode); kfree(tl_dinode); if (qrec) ocfs2_free_quota_recovery(qrec); mlog_errno(-ENOMEM); return; } INIT_LIST_HEAD(&item->lri_list); item->lri_la_dinode = la_dinode; item->lri_slot = slot_num; item->lri_tl_dinode = tl_dinode; item->lri_qrec = qrec; item->lri_orphan_reco_type = orphan_reco_type; spin_lock(&journal->j_lock); list_add_tail(&item->lri_list, &journal->j_la_cleanups); queue_work(journal->j_osb->ocfs2_wq, &journal->j_recovery_work); spin_unlock(&journal->j_lock); } /* Called by the mount code to queue recovery the last part of * recovery for it's own and offline slot(s). */ void ocfs2_complete_mount_recovery(struct ocfs2_super *osb) { struct ocfs2_journal *journal = osb->journal; if (ocfs2_is_hard_readonly(osb)) return; /* No need to queue up our truncate_log as regular cleanup will catch * that */ ocfs2_queue_recovery_completion(journal, osb->slot_num, osb->local_alloc_copy, NULL, NULL, ORPHAN_NEED_TRUNCATE); ocfs2_schedule_truncate_log_flush(osb, 0); osb->local_alloc_copy = NULL; /* queue to recover orphan slots for all offline slots */ ocfs2_replay_map_set_state(osb, REPLAY_NEEDED); ocfs2_queue_replay_slots(osb, ORPHAN_NEED_TRUNCATE); ocfs2_free_replay_slots(osb); } void ocfs2_complete_quota_recovery(struct ocfs2_super *osb) { if (osb->quota_rec) { ocfs2_queue_recovery_completion(osb->journal, osb->slot_num, NULL, NULL, osb->quota_rec, ORPHAN_NEED_TRUNCATE); osb->quota_rec = NULL; } } static int __ocfs2_recovery_thread(void *arg) { int status, node_num, slot_num; struct ocfs2_super *osb = arg; struct ocfs2_recovery_map *rm = osb->recovery_map; int *rm_quota = NULL; int rm_quota_used = 0, i; struct ocfs2_quota_recovery *qrec; /* Whether the quota supported. */ int quota_enabled = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA) || OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA); status = ocfs2_wait_on_mount(osb); if (status < 0) { goto bail; } if (quota_enabled) { rm_quota = kcalloc(osb->max_slots, sizeof(int), GFP_NOFS); if (!rm_quota) { status = -ENOMEM; goto bail; } } restart: status = ocfs2_super_lock(osb, 1); if (status < 0) { mlog_errno(status); goto bail; } status = ocfs2_compute_replay_slots(osb); if (status < 0) mlog_errno(status); /* queue recovery for our own slot */ ocfs2_queue_recovery_completion(osb->journal, osb->slot_num, NULL, NULL, NULL, ORPHAN_NO_NEED_TRUNCATE); spin_lock(&osb->osb_lock); while (rm->rm_used) { /* It's always safe to remove entry zero, as we won't * clear it until ocfs2_recover_node() has succeeded. */ node_num = rm->rm_entries[0]; spin_unlock(&osb->osb_lock); slot_num = ocfs2_node_num_to_slot(osb, node_num); trace_ocfs2_recovery_thread_node(node_num, slot_num); if (slot_num == -ENOENT) { status = 0; goto skip_recovery; } /* It is a bit subtle with quota recovery. We cannot do it * immediately because we have to obtain cluster locks from * quota files and we also don't want to just skip it because * then quota usage would be out of sync until some node takes * the slot. So we remember which nodes need quota recovery * and when everything else is done, we recover quotas. */ if (quota_enabled) { for (i = 0; i < rm_quota_used && rm_quota[i] != slot_num; i++) ; if (i == rm_quota_used) rm_quota[rm_quota_used++] = slot_num; } status = ocfs2_recover_node(osb, node_num, slot_num); skip_recovery: if (!status) { ocfs2_recovery_map_clear(osb, node_num); } else { mlog(ML_ERROR, "Error %d recovering node %d on device (%u,%u)!\n", status, node_num, MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); mlog(ML_ERROR, "Volume requires unmount.\n"); } spin_lock(&osb->osb_lock); } spin_unlock(&osb->osb_lock); trace_ocfs2_recovery_thread_end(status); /* Refresh all journal recovery generations from disk */ status = ocfs2_check_journals_nolocks(osb); status = (status == -EROFS) ? 0 : status; if (status < 0) mlog_errno(status); /* Now it is right time to recover quotas... We have to do this under * superblock lock so that no one can start using the slot (and crash) * before we recover it */ if (quota_enabled) { for (i = 0; i < rm_quota_used; i++) { qrec = ocfs2_begin_quota_recovery(osb, rm_quota[i]); if (IS_ERR(qrec)) { status = PTR_ERR(qrec); mlog_errno(status); continue; } ocfs2_queue_recovery_completion(osb->journal, rm_quota[i], NULL, NULL, qrec, ORPHAN_NEED_TRUNCATE); } } ocfs2_super_unlock(osb, 1); /* queue recovery for offline slots */ ocfs2_queue_replay_slots(osb, ORPHAN_NEED_TRUNCATE); bail: mutex_lock(&osb->recovery_lock); if (!status && !ocfs2_recovery_completed(osb)) { mutex_unlock(&osb->recovery_lock); goto restart; } ocfs2_free_replay_slots(osb); osb->recovery_thread_task = NULL; mb(); /* sync with ocfs2_recovery_thread_running */ wake_up(&osb->recovery_event); mutex_unlock(&osb->recovery_lock); if (quota_enabled) kfree(rm_quota); return status; } void ocfs2_recovery_thread(struct ocfs2_super *osb, int node_num) { mutex_lock(&osb->recovery_lock); trace_ocfs2_recovery_thread(node_num, osb->node_num, osb->disable_recovery, osb->recovery_thread_task, osb->disable_recovery ? -1 : ocfs2_recovery_map_set(osb, node_num)); if (osb->disable_recovery) goto out; if (osb->recovery_thread_task) goto out; osb->recovery_thread_task = kthread_run(__ocfs2_recovery_thread, osb, "ocfs2rec-%s", osb->uuid_str); if (IS_ERR(osb->recovery_thread_task)) { mlog_errno((int)PTR_ERR(osb->recovery_thread_task)); osb->recovery_thread_task = NULL; } out: mutex_unlock(&osb->recovery_lock); wake_up(&osb->recovery_event); } static int ocfs2_read_journal_inode(struct ocfs2_super *osb, int slot_num, struct buffer_head **bh, struct inode **ret_inode) { int status = -EACCES; struct inode *inode = NULL; BUG_ON(slot_num >= osb->max_slots); inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE, slot_num); if (!inode || is_bad_inode(inode)) { mlog_errno(status); goto bail; } SET_INODE_JOURNAL(inode); status = ocfs2_read_inode_block_full(inode, bh, OCFS2_BH_IGNORE_CACHE); if (status < 0) { mlog_errno(status); goto bail; } status = 0; bail: if (inode) { if (status || !ret_inode) iput(inode); else *ret_inode = inode; } return status; } /* Does the actual journal replay and marks the journal inode as * clean. Will only replay if the journal inode is marked dirty. */ static int ocfs2_replay_journal(struct ocfs2_super *osb, int node_num, int slot_num) { int status; int got_lock = 0; unsigned int flags; struct inode *inode = NULL; struct ocfs2_dinode *fe; journal_t *journal = NULL; struct buffer_head *bh = NULL; u32 slot_reco_gen; status = ocfs2_read_journal_inode(osb, slot_num, &bh, &inode); if (status) { mlog_errno(status); goto done; } fe = (struct ocfs2_dinode *)bh->b_data; slot_reco_gen = ocfs2_get_recovery_generation(fe); brelse(bh); bh = NULL; /* * As the fs recovery is asynchronous, there is a small chance that * another node mounted (and recovered) the slot before the recovery * thread could get the lock. To handle that, we dirty read the journal * inode for that slot to get the recovery generation. If it is * different than what we expected, the slot has been recovered. * If not, it needs recovery. */ if (osb->slot_recovery_generations[slot_num] != slot_reco_gen) { trace_ocfs2_replay_journal_recovered(slot_num, osb->slot_recovery_generations[slot_num], slot_reco_gen); osb->slot_recovery_generations[slot_num] = slot_reco_gen; status = -EBUSY; goto done; } /* Continue with recovery as the journal has not yet been recovered */ status = ocfs2_inode_lock_full(inode, &bh, 1, OCFS2_META_LOCK_RECOVERY); if (status < 0) { trace_ocfs2_replay_journal_lock_err(status); if (status != -ERESTARTSYS) mlog(ML_ERROR, "Could not lock journal!\n"); goto done; } got_lock = 1; fe = (struct ocfs2_dinode *) bh->b_data; flags = le32_to_cpu(fe->id1.journal1.ij_flags); slot_reco_gen = ocfs2_get_recovery_generation(fe); if (!(flags & OCFS2_JOURNAL_DIRTY_FL)) { trace_ocfs2_replay_journal_skip(node_num); /* Refresh recovery generation for the slot */ osb->slot_recovery_generations[slot_num] = slot_reco_gen; goto done; } /* we need to run complete recovery for offline orphan slots */ ocfs2_replay_map_set_state(osb, REPLAY_NEEDED); printk(KERN_NOTICE "ocfs2: Begin replay journal (node %d, slot %d) on "\ "device (%u,%u)\n", node_num, slot_num, MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters); status = ocfs2_force_read_journal(inode); if (status < 0) { mlog_errno(status); goto done; } journal = jbd2_journal_init_inode(inode); if (IS_ERR(journal)) { mlog(ML_ERROR, "Linux journal layer error\n"); status = PTR_ERR(journal); goto done; } status = jbd2_journal_load(journal); if (status < 0) { mlog_errno(status); BUG_ON(!igrab(inode)); jbd2_journal_destroy(journal); goto done; } ocfs2_clear_journal_error(osb->sb, journal, slot_num); /* wipe the journal */ jbd2_journal_lock_updates(journal); status = jbd2_journal_flush(journal, 0); jbd2_journal_unlock_updates(journal); if (status < 0) mlog_errno(status); /* This will mark the node clean */ flags = le32_to_cpu(fe->id1.journal1.ij_flags); flags &= ~OCFS2_JOURNAL_DIRTY_FL; fe->id1.journal1.ij_flags = cpu_to_le32(flags); /* Increment recovery generation to indicate successful recovery */ ocfs2_bump_recovery_generation(fe); osb->slot_recovery_generations[slot_num] = ocfs2_get_recovery_generation(fe); ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &fe->i_check); status = ocfs2_write_block(osb, bh, INODE_CACHE(inode)); if (status < 0) mlog_errno(status); BUG_ON(!igrab(inode)); jbd2_journal_destroy(journal); printk(KERN_NOTICE "ocfs2: End replay journal (node %d, slot %d) on "\ "device (%u,%u)\n", node_num, slot_num, MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); done: /* drop the lock on this nodes journal */ if (got_lock) ocfs2_inode_unlock(inode, 1); iput(inode); brelse(bh); return status; } /* * Do the most important parts of node recovery: * - Replay it's journal * - Stamp a clean local allocator file * - Stamp a clean truncate log * - Mark the node clean * * If this function completes without error, a node in OCFS2 can be * said to have been safely recovered. As a result, failure during the * second part of a nodes recovery process (local alloc recovery) is * far less concerning. */ static int ocfs2_recover_node(struct ocfs2_super *osb, int node_num, int slot_num) { int status = 0; struct ocfs2_dinode *la_copy = NULL; struct ocfs2_dinode *tl_copy = NULL; trace_ocfs2_recover_node(node_num, slot_num, osb->node_num); /* Should not ever be called to recover ourselves -- in that * case we should've called ocfs2_journal_load instead. */ BUG_ON(osb->node_num == node_num); status = ocfs2_replay_journal(osb, node_num, slot_num); if (status < 0) { if (status == -EBUSY) { trace_ocfs2_recover_node_skip(slot_num, node_num); status = 0; goto done; } mlog_errno(status); goto done; } /* Stamp a clean local alloc file AFTER recovering the journal... */ status = ocfs2_begin_local_alloc_recovery(osb, slot_num, &la_copy); if (status < 0) { mlog_errno(status); goto done; } /* An error from begin_truncate_log_recovery is not * serious enough to warrant halting the rest of * recovery. */ status = ocfs2_begin_truncate_log_recovery(osb, slot_num, &tl_copy); if (status < 0) mlog_errno(status); /* Likewise, this would be a strange but ultimately not so * harmful place to get an error... */ status = ocfs2_clear_slot(osb, slot_num); if (status < 0) mlog_errno(status); /* This will kfree the memory pointed to by la_copy and tl_copy */ ocfs2_queue_recovery_completion(osb->journal, slot_num, la_copy, tl_copy, NULL, ORPHAN_NEED_TRUNCATE); status = 0; done: return status; } /* Test node liveness by trylocking his journal. If we get the lock, * we drop it here. Return 0 if we got the lock, -EAGAIN if node is * still alive (we couldn't get the lock) and < 0 on error. */ static int ocfs2_trylock_journal(struct ocfs2_super *osb, int slot_num) { int status, flags; struct inode *inode = NULL; inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE, slot_num); if (inode == NULL) { mlog(ML_ERROR, "access error\n"); status = -EACCES; goto bail; } if (is_bad_inode(inode)) { mlog(ML_ERROR, "access error (bad inode)\n"); iput(inode); inode = NULL; status = -EACCES; goto bail; } SET_INODE_JOURNAL(inode); flags = OCFS2_META_LOCK_RECOVERY | OCFS2_META_LOCK_NOQUEUE; status = ocfs2_inode_lock_full(inode, NULL, 1, flags); if (status < 0) { if (status != -EAGAIN) mlog_errno(status); goto bail; } ocfs2_inode_unlock(inode, 1); bail: iput(inode); return status; } /* Call this underneath ocfs2_super_lock. It also assumes that the * slot info struct has been updated from disk. */ int ocfs2_mark_dead_nodes(struct ocfs2_super *osb) { unsigned int node_num; int status, i; u32 gen; struct buffer_head *bh = NULL; struct ocfs2_dinode *di; /* This is called with the super block cluster lock, so we * know that the slot map can't change underneath us. */ for (i = 0; i < osb->max_slots; i++) { /* Read journal inode to get the recovery generation */ status = ocfs2_read_journal_inode(osb, i, &bh, NULL); if (status) { mlog_errno(status); goto bail; } di = (struct ocfs2_dinode *)bh->b_data; gen = ocfs2_get_recovery_generation(di); brelse(bh); bh = NULL; spin_lock(&osb->osb_lock); osb->slot_recovery_generations[i] = gen; trace_ocfs2_mark_dead_nodes(i, osb->slot_recovery_generations[i]); if (i == osb->slot_num) { spin_unlock(&osb->osb_lock); continue; } status = ocfs2_slot_to_node_num_locked(osb, i, &node_num); if (status == -ENOENT) { spin_unlock(&osb->osb_lock); continue; } if (__ocfs2_recovery_map_test(osb, node_num)) { spin_unlock(&osb->osb_lock); continue; } spin_unlock(&osb->osb_lock); /* Ok, we have a slot occupied by another node which * is not in the recovery map. We trylock his journal * file here to test if he's alive. */ status = ocfs2_trylock_journal(osb, i); if (!status) { /* Since we're called from mount, we know that * the recovery thread can't race us on * setting / checking the recovery bits. */ ocfs2_recovery_thread(osb, node_num); } else if ((status < 0) && (status != -EAGAIN)) { mlog_errno(status); goto bail; } } status = 0; bail: return status; } /* * Scan timer should get fired every ORPHAN_SCAN_SCHEDULE_TIMEOUT. Add some * randomness to the timeout to minimize multple nodes firing the timer at the * same time. */ static inline unsigned long ocfs2_orphan_scan_timeout(void) { unsigned long time; get_random_bytes(&time, sizeof(time)); time = ORPHAN_SCAN_SCHEDULE_TIMEOUT + (time % 5000); return msecs_to_jiffies(time); } /* * ocfs2_queue_orphan_scan calls ocfs2_queue_recovery_completion for * every slot, queuing a recovery of the slot on the ocfs2_wq thread. This * is done to catch any orphans that are left over in orphan directories. * * It scans all slots, even ones that are in use. It does so to handle the * case described below: * * Node 1 has an inode it was using. The dentry went away due to memory * pressure. Node 1 closes the inode, but it's on the free list. The node * has the open lock. * Node 2 unlinks the inode. It grabs the dentry lock to notify others, * but node 1 has no dentry and doesn't get the message. It trylocks the * open lock, sees that another node has a PR, and does nothing. * Later node 2 runs its orphan dir. It igets the inode, trylocks the * open lock, sees the PR still, and does nothing. * Basically, we have to trigger an orphan iput on node 1. The only way * for this to happen is if node 1 runs node 2's orphan dir. * * ocfs2_queue_orphan_scan gets called every ORPHAN_SCAN_SCHEDULE_TIMEOUT * seconds. It gets an EX lock on os_lockres and checks sequence number * stored in LVB. If the sequence number has changed, it means some other * node has done the scan. This node skips the scan and tracks the * sequence number. If the sequence number didn't change, it means a scan * hasn't happened. The node queues a scan and increments the * sequence number in the LVB. */ static void ocfs2_queue_orphan_scan(struct ocfs2_super *osb) { struct ocfs2_orphan_scan *os; int status, i; u32 seqno = 0; os = &osb->osb_orphan_scan; if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) goto out; trace_ocfs2_queue_orphan_scan_begin(os->os_count, os->os_seqno, atomic_read(&os->os_state)); status = ocfs2_orphan_scan_lock(osb, &seqno); if (status < 0) { if (status != -EAGAIN) mlog_errno(status); goto out; } /* Do no queue the tasks if the volume is being umounted */ if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) goto unlock; if (os->os_seqno != seqno) { os->os_seqno = seqno; goto unlock; } for (i = 0; i < osb->max_slots; i++) ocfs2_queue_recovery_completion(osb->journal, i, NULL, NULL, NULL, ORPHAN_NO_NEED_TRUNCATE); /* * We queued a recovery on orphan slots, increment the sequence * number and update LVB so other node will skip the scan for a while */ seqno++; os->os_count++; os->os_scantime = ktime_get_seconds(); unlock: ocfs2_orphan_scan_unlock(osb, seqno); out: trace_ocfs2_queue_orphan_scan_end(os->os_count, os->os_seqno, atomic_read(&os->os_state)); return; } /* Worker task that gets fired every ORPHAN_SCAN_SCHEDULE_TIMEOUT millsec */ static void ocfs2_orphan_scan_work(struct work_struct *work) { struct ocfs2_orphan_scan *os; struct ocfs2_super *osb; os = container_of(work, struct ocfs2_orphan_scan, os_orphan_scan_work.work); osb = os->os_osb; mutex_lock(&os->os_lock); ocfs2_queue_orphan_scan(osb); if (atomic_read(&os->os_state) == ORPHAN_SCAN_ACTIVE) queue_delayed_work(osb->ocfs2_wq, &os->os_orphan_scan_work, ocfs2_orphan_scan_timeout()); mutex_unlock(&os->os_lock); } void ocfs2_orphan_scan_stop(struct ocfs2_super *osb) { struct ocfs2_orphan_scan *os; os = &osb->osb_orphan_scan; if (atomic_read(&os->os_state) == ORPHAN_SCAN_ACTIVE) { atomic_set(&os->os_state, ORPHAN_SCAN_INACTIVE); mutex_lock(&os->os_lock); cancel_delayed_work(&os->os_orphan_scan_work); mutex_unlock(&os->os_lock); } } void ocfs2_orphan_scan_init(struct ocfs2_super *osb) { struct ocfs2_orphan_scan *os; os = &osb->osb_orphan_scan; os->os_osb = osb; os->os_count = 0; os->os_seqno = 0; mutex_init(&os->os_lock); INIT_DELAYED_WORK(&os->os_orphan_scan_work, ocfs2_orphan_scan_work); } void ocfs2_orphan_scan_start(struct ocfs2_super *osb) { struct ocfs2_orphan_scan *os; os = &osb->osb_orphan_scan; os->os_scantime = ktime_get_seconds(); if (ocfs2_is_hard_readonly(osb) || ocfs2_mount_local(osb)) atomic_set(&os->os_state, ORPHAN_SCAN_INACTIVE); else { atomic_set(&os->os_state, ORPHAN_SCAN_ACTIVE); queue_delayed_work(osb->ocfs2_wq, &os->os_orphan_scan_work, ocfs2_orphan_scan_timeout()); } } struct ocfs2_orphan_filldir_priv { struct dir_context ctx; struct inode *head; struct ocfs2_super *osb; enum ocfs2_orphan_reco_type orphan_reco_type; }; static bool ocfs2_orphan_filldir(struct dir_context *ctx, const char *name, int name_len, loff_t pos, u64 ino, unsigned type) { struct ocfs2_orphan_filldir_priv *p = container_of(ctx, struct ocfs2_orphan_filldir_priv, ctx); struct inode *iter; if (name_len == 1 && !strncmp(".", name, 1)) return true; if (name_len == 2 && !strncmp("..", name, 2)) return true; /* do not include dio entry in case of orphan scan */ if ((p->orphan_reco_type == ORPHAN_NO_NEED_TRUNCATE) && (!strncmp(name, OCFS2_DIO_ORPHAN_PREFIX, OCFS2_DIO_ORPHAN_PREFIX_LEN))) return true; /* Skip bad inodes so that recovery can continue */ iter = ocfs2_iget(p->osb, ino, OCFS2_FI_FLAG_ORPHAN_RECOVERY, 0); if (IS_ERR(iter)) return true; if (!strncmp(name, OCFS2_DIO_ORPHAN_PREFIX, OCFS2_DIO_ORPHAN_PREFIX_LEN)) OCFS2_I(iter)->ip_flags |= OCFS2_INODE_DIO_ORPHAN_ENTRY; /* Skip inodes which are already added to recover list, since dio may * happen concurrently with unlink/rename */ if (OCFS2_I(iter)->ip_next_orphan) { iput(iter); return true; } trace_ocfs2_orphan_filldir((unsigned long long)OCFS2_I(iter)->ip_blkno); /* No locking is required for the next_orphan queue as there * is only ever a single process doing orphan recovery. */ OCFS2_I(iter)->ip_next_orphan = p->head; p->head = iter; return true; } static int ocfs2_queue_orphans(struct ocfs2_super *osb, int slot, struct inode **head, enum ocfs2_orphan_reco_type orphan_reco_type) { int status; struct inode *orphan_dir_inode = NULL; struct ocfs2_orphan_filldir_priv priv = { .ctx.actor = ocfs2_orphan_filldir, .osb = osb, .head = *head, .orphan_reco_type = orphan_reco_type }; orphan_dir_inode = ocfs2_get_system_file_inode(osb, ORPHAN_DIR_SYSTEM_INODE, slot); if (!orphan_dir_inode) { status = -ENOENT; mlog_errno(status); return status; } inode_lock(orphan_dir_inode); status = ocfs2_inode_lock(orphan_dir_inode, NULL, 0); if (status < 0) { mlog_errno(status); goto out; } status = ocfs2_dir_foreach(orphan_dir_inode, &priv.ctx); if (status) { mlog_errno(status); goto out_cluster; } *head = priv.head; out_cluster: ocfs2_inode_unlock(orphan_dir_inode, 0); out: inode_unlock(orphan_dir_inode); iput(orphan_dir_inode); return status; } static int ocfs2_orphan_recovery_can_continue(struct ocfs2_super *osb, int slot) { int ret; spin_lock(&osb->osb_lock); ret = !osb->osb_orphan_wipes[slot]; spin_unlock(&osb->osb_lock); return ret; } static void ocfs2_mark_recovering_orphan_dir(struct ocfs2_super *osb, int slot) { spin_lock(&osb->osb_lock); /* Mark ourselves such that new processes in delete_inode() * know to quit early. */ ocfs2_node_map_set_bit(osb, &osb->osb_recovering_orphan_dirs, slot); while (osb->osb_orphan_wipes[slot]) { /* If any processes are already in the middle of an * orphan wipe on this dir, then we need to wait for * them. */ spin_unlock(&osb->osb_lock); wait_event_interruptible(osb->osb_wipe_event, ocfs2_orphan_recovery_can_continue(osb, slot)); spin_lock(&osb->osb_lock); } spin_unlock(&osb->osb_lock); } static void ocfs2_clear_recovering_orphan_dir(struct ocfs2_super *osb, int slot) { ocfs2_node_map_clear_bit(osb, &osb->osb_recovering_orphan_dirs, slot); } /* * Orphan recovery. Each mounted node has it's own orphan dir which we * must run during recovery. Our strategy here is to build a list of * the inodes in the orphan dir and iget/iput them. The VFS does * (most) of the rest of the work. * * Orphan recovery can happen at any time, not just mount so we have a * couple of extra considerations. * * - We grab as many inodes as we can under the orphan dir lock - * doing iget() outside the orphan dir risks getting a reference on * an invalid inode. * - We must be sure not to deadlock with other processes on the * system wanting to run delete_inode(). This can happen when they go * to lock the orphan dir and the orphan recovery process attempts to * iget() inside the orphan dir lock. This can be avoided by * advertising our state to ocfs2_delete_inode(). */ static int ocfs2_recover_orphans(struct ocfs2_super *osb, int slot, enum ocfs2_orphan_reco_type orphan_reco_type) { int ret = 0; struct inode *inode = NULL; struct inode *iter; struct ocfs2_inode_info *oi; struct buffer_head *di_bh = NULL; struct ocfs2_dinode *di = NULL; trace_ocfs2_recover_orphans(slot); ocfs2_mark_recovering_orphan_dir(osb, slot); ret = ocfs2_queue_orphans(osb, slot, &inode, orphan_reco_type); ocfs2_clear_recovering_orphan_dir(osb, slot); /* Error here should be noted, but we want to continue with as * many queued inodes as we've got. */ if (ret) mlog_errno(ret); while (inode) { oi = OCFS2_I(inode); trace_ocfs2_recover_orphans_iput( (unsigned long long)oi->ip_blkno); iter = oi->ip_next_orphan; oi->ip_next_orphan = NULL; if (oi->ip_flags & OCFS2_INODE_DIO_ORPHAN_ENTRY) { inode_lock(inode); ret = ocfs2_rw_lock(inode, 1); if (ret < 0) { mlog_errno(ret); goto unlock_mutex; } /* * We need to take and drop the inode lock to * force read inode from disk. */ ret = ocfs2_inode_lock(inode, &di_bh, 1); if (ret) { mlog_errno(ret); goto unlock_rw; } di = (struct ocfs2_dinode *)di_bh->b_data; if (di->i_flags & cpu_to_le32(OCFS2_DIO_ORPHANED_FL)) { ret = ocfs2_truncate_file(inode, di_bh, i_size_read(inode)); if (ret < 0) { if (ret != -ENOSPC) mlog_errno(ret); goto unlock_inode; } ret = ocfs2_del_inode_from_orphan(osb, inode, di_bh, 0, 0); if (ret) mlog_errno(ret); } unlock_inode: ocfs2_inode_unlock(inode, 1); brelse(di_bh); di_bh = NULL; unlock_rw: ocfs2_rw_unlock(inode, 1); unlock_mutex: inode_unlock(inode); /* clear dio flag in ocfs2_inode_info */ oi->ip_flags &= ~OCFS2_INODE_DIO_ORPHAN_ENTRY; } else { spin_lock(&oi->ip_lock); /* Set the proper information to get us going into * ocfs2_delete_inode. */ oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED; spin_unlock(&oi->ip_lock); } iput(inode); inode = iter; } return ret; } static int __ocfs2_wait_on_mount(struct ocfs2_super *osb, int quota) { /* This check is good because ocfs2 will wait on our recovery * thread before changing it to something other than MOUNTED * or DISABLED. */ wait_event(osb->osb_mount_event, (!quota && atomic_read(&osb->vol_state) == VOLUME_MOUNTED) || atomic_read(&osb->vol_state) == VOLUME_MOUNTED_QUOTAS || atomic_read(&osb->vol_state) == VOLUME_DISABLED); /* If there's an error on mount, then we may never get to the * MOUNTED flag, but this is set right before * dismount_volume() so we can trust it. */ if (atomic_read(&osb->vol_state) == VOLUME_DISABLED) { trace_ocfs2_wait_on_mount(VOLUME_DISABLED); mlog(0, "mount error, exiting!\n"); return -EBUSY; } return 0; } static int ocfs2_commit_thread(void *arg) { int status; struct ocfs2_super *osb = arg; struct ocfs2_journal *journal = osb->journal; /* we can trust j_num_trans here because _should_stop() is only set in * shutdown and nobody other than ourselves should be able to start * transactions. committing on shutdown might take a few iterations * as final transactions put deleted inodes on the list */ while (!(kthread_should_stop() && atomic_read(&journal->j_num_trans) == 0)) { wait_event_interruptible(osb->checkpoint_event, atomic_read(&journal->j_num_trans) || kthread_should_stop()); status = ocfs2_commit_cache(osb); if (status < 0) { static unsigned long abort_warn_time; /* Warn about this once per minute */ if (printk_timed_ratelimit(&abort_warn_time, 60*HZ)) mlog(ML_ERROR, "status = %d, journal is " "already aborted.\n", status); /* * After ocfs2_commit_cache() fails, j_num_trans has a * non-zero value. Sleep here to avoid a busy-wait * loop. */ msleep_interruptible(1000); } if (kthread_should_stop() && atomic_read(&journal->j_num_trans)){ mlog(ML_KTHREAD, "commit_thread: %u transactions pending on " "shutdown\n", atomic_read(&journal->j_num_trans)); } } return 0; } /* Reads all the journal inodes without taking any cluster locks. Used * for hard readonly access to determine whether any journal requires * recovery. Also used to refresh the recovery generation numbers after * a journal has been recovered by another node. */ int ocfs2_check_journals_nolocks(struct ocfs2_super *osb) { int ret = 0; unsigned int slot; struct buffer_head *di_bh = NULL; struct ocfs2_dinode *di; int journal_dirty = 0; for(slot = 0; slot < osb->max_slots; slot++) { ret = ocfs2_read_journal_inode(osb, slot, &di_bh, NULL); if (ret) { mlog_errno(ret); goto out; } di = (struct ocfs2_dinode *) di_bh->b_data; osb->slot_recovery_generations[slot] = ocfs2_get_recovery_generation(di); if (le32_to_cpu(di->id1.journal1.ij_flags) & OCFS2_JOURNAL_DIRTY_FL) journal_dirty = 1; brelse(di_bh); di_bh = NULL; } out: if (journal_dirty) ret = -EROFS; return ret; }
Information contained on this website is for historical information purposes only and does not indicate or represent copyright ownership.
Created with Cregit http://github.com/cregit/cregit
Version 2.0-RC1