super.c 150 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827282828292830283128322833283428352836283728382839284028412842284328442845284628472848284928502851285228532854285528562857285828592860286128622863286428652866286728682869287028712872287328742875287628772878287928802881288228832884288528862887288828892890289128922893289428952896289728982899290029012902290329042905290629072908290929102911291229132914291529162917291829192920292129222923292429252926292729282929293029312932293329342935293629372938293929402941294229432944294529462947294829492950295129522953295429552956295729582959296029612962296329642965296629672968296929702971297229732974297529762977297829792980298129822983298429852986298729882989299029912992299329942995299629972998299930003001300230033004300530063007300830093010301130123013301430153016301730183019302030213022302330243025302630273028302930303031303230333034303530363037303830393040304130423043304430453046304730483049305030513052305330543055305630573058305930603061306230633064306530663067306830693070307130723073307430753076307730783079308030813082308330843085308630873088308930903091309230933094309530963097309830993100310131023103310431053106310731083109311031113112311331143115311631173118311931203121312231233124312531263127312831293130313131323133313431353136313731383139314031413142314331443145314631473148314931503151315231533154315531563157315831593160316131623163316431653166316731683169317031713172317331743175317631773178317931803181318231833184318531863187318831893190319131923193319431953196319731983199320032013202320332043205320632073208320932103211321232133214321532163217321832193220322132223223322432253226322732283229323032313232323332343235323632373238323932403241324232433244324532463247324832493250325132523253325432553256325732583259326032613262326332643265326632673268326932703271327232733274327532763277327832793280328132823283328432853286328732883289329032913292329332943295329632973298329933003301330233033304330533063307330833093310331133123313331433153316331733183319332033213322332333243325332633273328332933303331333233333334333533363337333833393340334133423343334433453346334733483349335033513352335333543355335633573358335933603361336233633364336533663367336833693370337133723373337433753376337733783379338033813382338333843385338633873388338933903391339233933394339533963397339833993400340134023403340434053406340734083409341034113412341334143415341634173418341934203421342234233424342534263427342834293430343134323433343434353436343734383439344034413442344334443445344634473448344934503451345234533454345534563457345834593460346134623463346434653466346734683469347034713472347334743475347634773478347934803481348234833484348534863487348834893490349134923493349434953496349734983499350035013502350335043505350635073508350935103511351235133514351535163517351835193520352135223523352435253526352735283529353035313532353335343535353635373538353935403541354235433544354535463547354835493550355135523553355435553556355735583559356035613562356335643565356635673568356935703571357235733574357535763577357835793580358135823583358435853586358735883589359035913592359335943595359635973598359936003601360236033604360536063607360836093610361136123613361436153616361736183619362036213622362336243625362636273628362936303631363236333634363536363637363836393640364136423643364436453646364736483649365036513652365336543655365636573658365936603661366236633664366536663667366836693670367136723673367436753676367736783679368036813682368336843685368636873688368936903691369236933694369536963697369836993700370137023703370437053706370737083709371037113712371337143715371637173718371937203721372237233724372537263727372837293730373137323733373437353736373737383739374037413742374337443745374637473748374937503751375237533754375537563757375837593760376137623763376437653766376737683769377037713772377337743775377637773778377937803781378237833784378537863787378837893790379137923793379437953796379737983799380038013802380338043805380638073808380938103811381238133814381538163817381838193820382138223823382438253826382738283829383038313832383338343835383638373838383938403841384238433844384538463847384838493850385138523853385438553856385738583859386038613862386338643865386638673868386938703871387238733874387538763877387838793880388138823883388438853886388738883889389038913892389338943895389638973898389939003901390239033904390539063907390839093910391139123913391439153916391739183919392039213922392339243925392639273928392939303931393239333934393539363937393839393940394139423943394439453946394739483949395039513952395339543955395639573958395939603961396239633964396539663967396839693970397139723973397439753976397739783979398039813982398339843985398639873988398939903991399239933994399539963997399839994000400140024003400440054006400740084009401040114012401340144015401640174018401940204021402240234024402540264027402840294030403140324033403440354036403740384039404040414042404340444045404640474048404940504051405240534054405540564057405840594060406140624063406440654066406740684069407040714072407340744075407640774078407940804081408240834084408540864087408840894090409140924093409440954096409740984099410041014102410341044105410641074108410941104111411241134114411541164117411841194120412141224123412441254126412741284129413041314132413341344135413641374138413941404141414241434144414541464147414841494150415141524153415441554156415741584159416041614162416341644165416641674168416941704171417241734174417541764177417841794180418141824183418441854186418741884189419041914192419341944195419641974198419942004201420242034204420542064207420842094210421142124213421442154216421742184219422042214222422342244225422642274228422942304231423242334234423542364237423842394240424142424243424442454246424742484249425042514252425342544255425642574258425942604261426242634264426542664267426842694270427142724273427442754276427742784279428042814282428342844285428642874288428942904291429242934294429542964297429842994300430143024303430443054306430743084309431043114312431343144315431643174318431943204321432243234324432543264327432843294330433143324333433443354336433743384339434043414342434343444345434643474348434943504351435243534354435543564357435843594360436143624363436443654366436743684369437043714372437343744375437643774378437943804381438243834384438543864387438843894390439143924393439443954396439743984399440044014402440344044405440644074408440944104411441244134414441544164417441844194420442144224423442444254426442744284429443044314432443344344435443644374438443944404441444244434444444544464447444844494450445144524453445444554456445744584459446044614462446344644465446644674468446944704471447244734474447544764477447844794480448144824483448444854486448744884489449044914492449344944495449644974498449945004501450245034504450545064507450845094510451145124513451445154516451745184519452045214522452345244525452645274528452945304531453245334534453545364537453845394540454145424543454445454546454745484549455045514552455345544555455645574558455945604561456245634564456545664567456845694570457145724573457445754576457745784579458045814582458345844585458645874588458945904591459245934594459545964597459845994600460146024603460446054606460746084609461046114612461346144615461646174618461946204621462246234624462546264627462846294630463146324633463446354636463746384639464046414642464346444645464646474648464946504651465246534654465546564657465846594660466146624663466446654666466746684669467046714672467346744675467646774678467946804681468246834684468546864687468846894690469146924693469446954696469746984699470047014702470347044705470647074708470947104711471247134714471547164717471847194720472147224723472447254726472747284729473047314732473347344735473647374738473947404741474247434744474547464747474847494750475147524753475447554756475747584759476047614762476347644765476647674768476947704771477247734774477547764777477847794780478147824783478447854786478747884789479047914792479347944795479647974798479948004801480248034804480548064807480848094810481148124813481448154816481748184819482048214822482348244825482648274828482948304831483248334834483548364837483848394840484148424843484448454846484748484849485048514852485348544855485648574858485948604861486248634864486548664867486848694870487148724873487448754876487748784879488048814882488348844885488648874888488948904891489248934894489548964897489848994900490149024903490449054906490749084909491049114912491349144915491649174918491949204921492249234924492549264927492849294930493149324933493449354936493749384939494049414942494349444945494649474948494949504951495249534954495549564957495849594960496149624963496449654966496749684969497049714972497349744975497649774978497949804981498249834984498549864987498849894990499149924993499449954996499749984999500050015002500350045005500650075008500950105011501250135014501550165017501850195020502150225023502450255026502750285029503050315032503350345035503650375038503950405041504250435044504550465047504850495050505150525053505450555056505750585059506050615062506350645065506650675068506950705071507250735074507550765077507850795080508150825083508450855086508750885089509050915092509350945095509650975098509951005101510251035104510551065107510851095110511151125113511451155116511751185119512051215122512351245125512651275128512951305131513251335134513551365137513851395140514151425143514451455146514751485149515051515152515351545155515651575158515951605161516251635164516551665167516851695170517151725173517451755176517751785179518051815182518351845185518651875188518951905191519251935194519551965197519851995200520152025203520452055206520752085209521052115212521352145215521652175218521952205221522252235224522552265227522852295230523152325233523452355236523752385239524052415242524352445245524652475248524952505251525252535254525552565257525852595260526152625263526452655266526752685269527052715272527352745275527652775278527952805281528252835284528552865287528852895290529152925293529452955296529752985299530053015302530353045305530653075308530953105311531253135314531553165317531853195320532153225323532453255326532753285329533053315332533353345335533653375338533953405341534253435344534553465347534853495350535153525353535453555356535753585359536053615362536353645365536653675368536953705371
  1. /*
  2. * linux/fs/ext4/super.c
  3. *
  4. * Copyright (C) 1992, 1993, 1994, 1995
  5. * Remy Card (card@masi.ibp.fr)
  6. * Laboratoire MASI - Institut Blaise Pascal
  7. * Universite Pierre et Marie Curie (Paris VI)
  8. *
  9. * from
  10. *
  11. * linux/fs/minix/inode.c
  12. *
  13. * Copyright (C) 1991, 1992 Linus Torvalds
  14. *
  15. * Big-endian to little-endian byte-swapping/bitmaps by
  16. * David S. Miller (davem@caip.rutgers.edu), 1995
  17. */
  18. #include <linux/module.h>
  19. #include <linux/string.h>
  20. #include <linux/fs.h>
  21. #include <linux/time.h>
  22. #include <linux/vmalloc.h>
  23. #include <linux/jbd2.h>
  24. #include <linux/slab.h>
  25. #include <linux/init.h>
  26. #include <linux/blkdev.h>
  27. #include <linux/parser.h>
  28. #include <linux/buffer_head.h>
  29. #include <linux/exportfs.h>
  30. #include <linux/vfs.h>
  31. #include <linux/random.h>
  32. #include <linux/mount.h>
  33. #include <linux/namei.h>
  34. #include <linux/quotaops.h>
  35. #include <linux/seq_file.h>
  36. #include <linux/proc_fs.h>
  37. #include <linux/ctype.h>
  38. #include <linux/log2.h>
  39. #include <linux/crc16.h>
  40. #include <linux/cleancache.h>
  41. #include <asm/uaccess.h>
  42. #include <linux/kthread.h>
  43. #include <linux/freezer.h>
  44. #include "ext4.h"
  45. #include "ext4_extents.h"
  46. #include "ext4_jbd2.h"
  47. #include "xattr.h"
  48. #include "acl.h"
  49. #include "mballoc.h"
  50. #define CREATE_TRACE_POINTS
  51. #include <trace/events/ext4.h>
  52. static struct proc_dir_entry *ext4_proc_root;
  53. static struct kset *ext4_kset;
  54. static struct ext4_lazy_init *ext4_li_info;
  55. static struct mutex ext4_li_mtx;
  56. static struct ext4_features *ext4_feat;
  57. static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
  58. unsigned long journal_devnum);
  59. static int ext4_show_options(struct seq_file *seq, struct dentry *root);
  60. static int ext4_commit_super(struct super_block *sb, int sync);
  61. static void ext4_mark_recovery_complete(struct super_block *sb,
  62. struct ext4_super_block *es);
  63. static void ext4_clear_journal_err(struct super_block *sb,
  64. struct ext4_super_block *es);
  65. static int ext4_sync_fs(struct super_block *sb, int wait);
  66. static const char *ext4_decode_error(struct super_block *sb, int errno,
  67. char nbuf[16]);
  68. static int ext4_remount(struct super_block *sb, int *flags, char *data);
  69. static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
  70. static int ext4_unfreeze(struct super_block *sb);
  71. static int ext4_freeze(struct super_block *sb);
  72. static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
  73. const char *dev_name, void *data);
  74. static inline int ext2_feature_set_ok(struct super_block *sb);
  75. static inline int ext3_feature_set_ok(struct super_block *sb);
  76. static int ext4_feature_set_ok(struct super_block *sb, int readonly);
  77. static void ext4_destroy_lazyinit_thread(void);
  78. static void ext4_unregister_li_request(struct super_block *sb);
  79. static void ext4_clear_request_list(void);
  80. #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
  81. static struct file_system_type ext2_fs_type = {
  82. .owner = THIS_MODULE,
  83. .name = "ext2",
  84. .mount = ext4_mount,
  85. .kill_sb = kill_block_super,
  86. .fs_flags = FS_REQUIRES_DEV,
  87. };
  88. #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
  89. #else
  90. #define IS_EXT2_SB(sb) (0)
  91. #endif
  92. #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
  93. static struct file_system_type ext3_fs_type = {
  94. .owner = THIS_MODULE,
  95. .name = "ext3",
  96. .mount = ext4_mount,
  97. .kill_sb = kill_block_super,
  98. .fs_flags = FS_REQUIRES_DEV,
  99. };
  100. #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
  101. #else
  102. #define IS_EXT3_SB(sb) (0)
  103. #endif
  104. static int ext4_verify_csum_type(struct super_block *sb,
  105. struct ext4_super_block *es)
  106. {
  107. if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
  108. EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
  109. return 1;
  110. return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
  111. }
  112. static __le32 ext4_superblock_csum(struct super_block *sb,
  113. struct ext4_super_block *es)
  114. {
  115. struct ext4_sb_info *sbi = EXT4_SB(sb);
  116. int offset = offsetof(struct ext4_super_block, s_checksum);
  117. __u32 csum;
  118. csum = ext4_chksum(sbi, ~0, (char *)es, offset);
  119. return cpu_to_le32(csum);
  120. }
  121. int ext4_superblock_csum_verify(struct super_block *sb,
  122. struct ext4_super_block *es)
  123. {
  124. if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
  125. EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
  126. return 1;
  127. return es->s_checksum == ext4_superblock_csum(sb, es);
  128. }
  129. void ext4_superblock_csum_set(struct super_block *sb,
  130. struct ext4_super_block *es)
  131. {
  132. if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
  133. EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
  134. return;
  135. es->s_checksum = ext4_superblock_csum(sb, es);
  136. }
  137. void *ext4_kvmalloc(size_t size, gfp_t flags)
  138. {
  139. void *ret;
  140. ret = kmalloc(size, flags);
  141. if (!ret)
  142. ret = __vmalloc(size, flags, PAGE_KERNEL);
  143. return ret;
  144. }
  145. void *ext4_kvzalloc(size_t size, gfp_t flags)
  146. {
  147. void *ret;
  148. ret = kzalloc(size, flags);
  149. if (!ret)
  150. ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
  151. return ret;
  152. }
  153. void ext4_kvfree(void *ptr)
  154. {
  155. if (is_vmalloc_addr(ptr))
  156. vfree(ptr);
  157. else
  158. kfree(ptr);
  159. }
  160. ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
  161. struct ext4_group_desc *bg)
  162. {
  163. return le32_to_cpu(bg->bg_block_bitmap_lo) |
  164. (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  165. (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
  166. }
  167. ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
  168. struct ext4_group_desc *bg)
  169. {
  170. return le32_to_cpu(bg->bg_inode_bitmap_lo) |
  171. (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  172. (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
  173. }
  174. ext4_fsblk_t ext4_inode_table(struct super_block *sb,
  175. struct ext4_group_desc *bg)
  176. {
  177. return le32_to_cpu(bg->bg_inode_table_lo) |
  178. (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  179. (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
  180. }
  181. __u32 ext4_free_group_clusters(struct super_block *sb,
  182. struct ext4_group_desc *bg)
  183. {
  184. return le16_to_cpu(bg->bg_free_blocks_count_lo) |
  185. (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  186. (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
  187. }
  188. __u32 ext4_free_inodes_count(struct super_block *sb,
  189. struct ext4_group_desc *bg)
  190. {
  191. return le16_to_cpu(bg->bg_free_inodes_count_lo) |
  192. (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  193. (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
  194. }
  195. __u32 ext4_used_dirs_count(struct super_block *sb,
  196. struct ext4_group_desc *bg)
  197. {
  198. return le16_to_cpu(bg->bg_used_dirs_count_lo) |
  199. (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  200. (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
  201. }
  202. __u32 ext4_itable_unused_count(struct super_block *sb,
  203. struct ext4_group_desc *bg)
  204. {
  205. return le16_to_cpu(bg->bg_itable_unused_lo) |
  206. (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
  207. (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
  208. }
  209. void ext4_block_bitmap_set(struct super_block *sb,
  210. struct ext4_group_desc *bg, ext4_fsblk_t blk)
  211. {
  212. bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
  213. if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
  214. bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
  215. }
  216. void ext4_inode_bitmap_set(struct super_block *sb,
  217. struct ext4_group_desc *bg, ext4_fsblk_t blk)
  218. {
  219. bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
  220. if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
  221. bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
  222. }
  223. void ext4_inode_table_set(struct super_block *sb,
  224. struct ext4_group_desc *bg, ext4_fsblk_t blk)
  225. {
  226. bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
  227. if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
  228. bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
  229. }
  230. void ext4_free_group_clusters_set(struct super_block *sb,
  231. struct ext4_group_desc *bg, __u32 count)
  232. {
  233. bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
  234. if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
  235. bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
  236. }
  237. void ext4_free_inodes_set(struct super_block *sb,
  238. struct ext4_group_desc *bg, __u32 count)
  239. {
  240. bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
  241. if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
  242. bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
  243. }
  244. void ext4_used_dirs_set(struct super_block *sb,
  245. struct ext4_group_desc *bg, __u32 count)
  246. {
  247. bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
  248. if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
  249. bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
  250. }
  251. void ext4_itable_unused_set(struct super_block *sb,
  252. struct ext4_group_desc *bg, __u32 count)
  253. {
  254. bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
  255. if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
  256. bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
  257. }
  258. /* Just increment the non-pointer handle value */
  259. static handle_t *ext4_get_nojournal(void)
  260. {
  261. handle_t *handle = current->journal_info;
  262. unsigned long ref_cnt = (unsigned long)handle;
  263. BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
  264. ref_cnt++;
  265. handle = (handle_t *)ref_cnt;
  266. current->journal_info = handle;
  267. return handle;
  268. }
  269. /* Decrement the non-pointer handle value */
  270. static void ext4_put_nojournal(handle_t *handle)
  271. {
  272. unsigned long ref_cnt = (unsigned long)handle;
  273. BUG_ON(ref_cnt == 0);
  274. ref_cnt--;
  275. handle = (handle_t *)ref_cnt;
  276. current->journal_info = handle;
  277. }
  278. /*
  279. * Wrappers for jbd2_journal_start/end.
  280. *
  281. * The only special thing we need to do here is to make sure that all
  282. * journal_end calls result in the superblock being marked dirty, so
  283. * that sync() will call the filesystem's write_super callback if
  284. * appropriate.
  285. */
  286. handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
  287. {
  288. journal_t *journal;
  289. trace_ext4_journal_start(sb, nblocks, _RET_IP_);
  290. if (sb->s_flags & MS_RDONLY)
  291. return ERR_PTR(-EROFS);
  292. WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
  293. journal = EXT4_SB(sb)->s_journal;
  294. if (!journal)
  295. return ext4_get_nojournal();
  296. /*
  297. * Special case here: if the journal has aborted behind our
  298. * backs (eg. EIO in the commit thread), then we still need to
  299. * take the FS itself readonly cleanly.
  300. */
  301. if (is_journal_aborted(journal)) {
  302. ext4_abort(sb, "Detected aborted journal");
  303. return ERR_PTR(-EROFS);
  304. }
  305. return jbd2_journal_start(journal, nblocks);
  306. }
  307. /*
  308. * The only special thing we need to do here is to make sure that all
  309. * jbd2_journal_stop calls result in the superblock being marked dirty, so
  310. * that sync() will call the filesystem's write_super callback if
  311. * appropriate.
  312. */
  313. int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
  314. {
  315. struct super_block *sb;
  316. int err;
  317. int rc;
  318. if (!ext4_handle_valid(handle)) {
  319. ext4_put_nojournal(handle);
  320. return 0;
  321. }
  322. sb = handle->h_transaction->t_journal->j_private;
  323. err = handle->h_err;
  324. rc = jbd2_journal_stop(handle);
  325. if (!err)
  326. err = rc;
  327. if (err)
  328. __ext4_std_error(sb, where, line, err);
  329. return err;
  330. }
  331. void ext4_journal_abort_handle(const char *caller, unsigned int line,
  332. const char *err_fn, struct buffer_head *bh,
  333. handle_t *handle, int err)
  334. {
  335. char nbuf[16];
  336. const char *errstr = ext4_decode_error(NULL, err, nbuf);
  337. BUG_ON(!ext4_handle_valid(handle));
  338. if (bh)
  339. BUFFER_TRACE(bh, "abort");
  340. if (!handle->h_err)
  341. handle->h_err = err;
  342. if (is_handle_aborted(handle))
  343. return;
  344. printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
  345. caller, line, errstr, err_fn);
  346. jbd2_journal_abort_handle(handle);
  347. }
  348. static void __save_error_info(struct super_block *sb, const char *func,
  349. unsigned int line)
  350. {
  351. struct ext4_super_block *es = EXT4_SB(sb)->s_es;
  352. EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
  353. es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
  354. es->s_last_error_time = cpu_to_le32(get_seconds());
  355. strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
  356. es->s_last_error_line = cpu_to_le32(line);
  357. if (!es->s_first_error_time) {
  358. es->s_first_error_time = es->s_last_error_time;
  359. strncpy(es->s_first_error_func, func,
  360. sizeof(es->s_first_error_func));
  361. es->s_first_error_line = cpu_to_le32(line);
  362. es->s_first_error_ino = es->s_last_error_ino;
  363. es->s_first_error_block = es->s_last_error_block;
  364. }
  365. /*
  366. * Start the daily error reporting function if it hasn't been
  367. * started already
  368. */
  369. if (!es->s_error_count)
  370. mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
  371. es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
  372. }
  373. static void save_error_info(struct super_block *sb, const char *func,
  374. unsigned int line)
  375. {
  376. __save_error_info(sb, func, line);
  377. ext4_commit_super(sb, 1);
  378. }
  379. /*
  380. * The del_gendisk() function uninitializes the disk-specific data
  381. * structures, including the bdi structure, without telling anyone
  382. * else. Once this happens, any attempt to call mark_buffer_dirty()
  383. * (for example, by ext4_commit_super), will cause a kernel OOPS.
  384. * This is a kludge to prevent these oops until we can put in a proper
  385. * hook in del_gendisk() to inform the VFS and file system layers.
  386. */
  387. static int block_device_ejected(struct super_block *sb)
  388. {
  389. struct inode *bd_inode = sb->s_bdev->bd_inode;
  390. struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
  391. return bdi->dev == NULL;
  392. }
  393. static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
  394. {
  395. struct super_block *sb = journal->j_private;
  396. struct ext4_sb_info *sbi = EXT4_SB(sb);
  397. int error = is_journal_aborted(journal);
  398. struct ext4_journal_cb_entry *jce, *tmp;
  399. spin_lock(&sbi->s_md_lock);
  400. list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
  401. list_del_init(&jce->jce_list);
  402. spin_unlock(&sbi->s_md_lock);
  403. jce->jce_func(sb, jce, error);
  404. spin_lock(&sbi->s_md_lock);
  405. }
  406. spin_unlock(&sbi->s_md_lock);
  407. }
  408. /* Deal with the reporting of failure conditions on a filesystem such as
  409. * inconsistencies detected or read IO failures.
  410. *
  411. * On ext2, we can store the error state of the filesystem in the
  412. * superblock. That is not possible on ext4, because we may have other
  413. * write ordering constraints on the superblock which prevent us from
  414. * writing it out straight away; and given that the journal is about to
  415. * be aborted, we can't rely on the current, or future, transactions to
  416. * write out the superblock safely.
  417. *
  418. * We'll just use the jbd2_journal_abort() error code to record an error in
  419. * the journal instead. On recovery, the journal will complain about
  420. * that error until we've noted it down and cleared it.
  421. */
  422. static void ext4_handle_error(struct super_block *sb)
  423. {
  424. if (sb->s_flags & MS_RDONLY)
  425. return;
  426. if (!test_opt(sb, ERRORS_CONT)) {
  427. journal_t *journal = EXT4_SB(sb)->s_journal;
  428. EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
  429. if (journal)
  430. jbd2_journal_abort(journal, -EIO);
  431. }
  432. if (test_opt(sb, ERRORS_RO)) {
  433. ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
  434. sb->s_flags |= MS_RDONLY;
  435. }
  436. if (test_opt(sb, ERRORS_PANIC))
  437. panic("EXT4-fs (device %s): panic forced after error\n",
  438. sb->s_id);
  439. }
  440. void __ext4_error(struct super_block *sb, const char *function,
  441. unsigned int line, const char *fmt, ...)
  442. {
  443. struct va_format vaf;
  444. va_list args;
  445. va_start(args, fmt);
  446. vaf.fmt = fmt;
  447. vaf.va = &args;
  448. printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
  449. sb->s_id, function, line, current->comm, &vaf);
  450. va_end(args);
  451. save_error_info(sb, function, line);
  452. ext4_handle_error(sb);
  453. }
  454. void ext4_error_inode(struct inode *inode, const char *function,
  455. unsigned int line, ext4_fsblk_t block,
  456. const char *fmt, ...)
  457. {
  458. va_list args;
  459. struct va_format vaf;
  460. struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
  461. es->s_last_error_ino = cpu_to_le32(inode->i_ino);
  462. es->s_last_error_block = cpu_to_le64(block);
  463. save_error_info(inode->i_sb, function, line);
  464. va_start(args, fmt);
  465. vaf.fmt = fmt;
  466. vaf.va = &args;
  467. if (block)
  468. printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
  469. "inode #%lu: block %llu: comm %s: %pV\n",
  470. inode->i_sb->s_id, function, line, inode->i_ino,
  471. block, current->comm, &vaf);
  472. else
  473. printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
  474. "inode #%lu: comm %s: %pV\n",
  475. inode->i_sb->s_id, function, line, inode->i_ino,
  476. current->comm, &vaf);
  477. va_end(args);
  478. ext4_handle_error(inode->i_sb);
  479. }
  480. void ext4_error_file(struct file *file, const char *function,
  481. unsigned int line, ext4_fsblk_t block,
  482. const char *fmt, ...)
  483. {
  484. va_list args;
  485. struct va_format vaf;
  486. struct ext4_super_block *es;
  487. struct inode *inode = file->f_dentry->d_inode;
  488. char pathname[80], *path;
  489. es = EXT4_SB(inode->i_sb)->s_es;
  490. es->s_last_error_ino = cpu_to_le32(inode->i_ino);
  491. save_error_info(inode->i_sb, function, line);
  492. path = d_path(&(file->f_path), pathname, sizeof(pathname));
  493. if (IS_ERR(path))
  494. path = "(unknown)";
  495. va_start(args, fmt);
  496. vaf.fmt = fmt;
  497. vaf.va = &args;
  498. if (block)
  499. printk(KERN_CRIT
  500. "EXT4-fs error (device %s): %s:%d: inode #%lu: "
  501. "block %llu: comm %s: path %s: %pV\n",
  502. inode->i_sb->s_id, function, line, inode->i_ino,
  503. block, current->comm, path, &vaf);
  504. else
  505. printk(KERN_CRIT
  506. "EXT4-fs error (device %s): %s:%d: inode #%lu: "
  507. "comm %s: path %s: %pV\n",
  508. inode->i_sb->s_id, function, line, inode->i_ino,
  509. current->comm, path, &vaf);
  510. va_end(args);
  511. ext4_handle_error(inode->i_sb);
  512. }
  513. static const char *ext4_decode_error(struct super_block *sb, int errno,
  514. char nbuf[16])
  515. {
  516. char *errstr = NULL;
  517. switch (errno) {
  518. case -EIO:
  519. errstr = "IO failure";
  520. break;
  521. case -ENOMEM:
  522. errstr = "Out of memory";
  523. break;
  524. case -EROFS:
  525. if (!sb || (EXT4_SB(sb)->s_journal &&
  526. EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
  527. errstr = "Journal has aborted";
  528. else
  529. errstr = "Readonly filesystem";
  530. break;
  531. default:
  532. /* If the caller passed in an extra buffer for unknown
  533. * errors, textualise them now. Else we just return
  534. * NULL. */
  535. if (nbuf) {
  536. /* Check for truncated error codes... */
  537. if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
  538. errstr = nbuf;
  539. }
  540. break;
  541. }
  542. return errstr;
  543. }
  544. /* __ext4_std_error decodes expected errors from journaling functions
  545. * automatically and invokes the appropriate error response. */
  546. void __ext4_std_error(struct super_block *sb, const char *function,
  547. unsigned int line, int errno)
  548. {
  549. char nbuf[16];
  550. const char *errstr;
  551. /* Special case: if the error is EROFS, and we're not already
  552. * inside a transaction, then there's really no point in logging
  553. * an error. */
  554. if (errno == -EROFS && journal_current_handle() == NULL &&
  555. (sb->s_flags & MS_RDONLY))
  556. return;
  557. errstr = ext4_decode_error(sb, errno, nbuf);
  558. printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
  559. sb->s_id, function, line, errstr);
  560. save_error_info(sb, function, line);
  561. ext4_handle_error(sb);
  562. }
  563. /*
  564. * ext4_abort is a much stronger failure handler than ext4_error. The
  565. * abort function may be used to deal with unrecoverable failures such
  566. * as journal IO errors or ENOMEM at a critical moment in log management.
  567. *
  568. * We unconditionally force the filesystem into an ABORT|READONLY state,
  569. * unless the error response on the fs has been set to panic in which
  570. * case we take the easy way out and panic immediately.
  571. */
  572. void __ext4_abort(struct super_block *sb, const char *function,
  573. unsigned int line, const char *fmt, ...)
  574. {
  575. va_list args;
  576. save_error_info(sb, function, line);
  577. va_start(args, fmt);
  578. printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
  579. function, line);
  580. vprintk(fmt, args);
  581. printk("\n");
  582. va_end(args);
  583. if ((sb->s_flags & MS_RDONLY) == 0) {
  584. ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
  585. sb->s_flags |= MS_RDONLY;
  586. EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
  587. if (EXT4_SB(sb)->s_journal)
  588. jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
  589. save_error_info(sb, function, line);
  590. }
  591. if (test_opt(sb, ERRORS_PANIC))
  592. panic("EXT4-fs panic from previous error\n");
  593. }
  594. void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
  595. {
  596. struct va_format vaf;
  597. va_list args;
  598. va_start(args, fmt);
  599. vaf.fmt = fmt;
  600. vaf.va = &args;
  601. printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  602. va_end(args);
  603. }
  604. void __ext4_warning(struct super_block *sb, const char *function,
  605. unsigned int line, const char *fmt, ...)
  606. {
  607. struct va_format vaf;
  608. va_list args;
  609. va_start(args, fmt);
  610. vaf.fmt = fmt;
  611. vaf.va = &args;
  612. printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
  613. sb->s_id, function, line, &vaf);
  614. va_end(args);
  615. }
  616. void __ext4_grp_locked_error(const char *function, unsigned int line,
  617. struct super_block *sb, ext4_group_t grp,
  618. unsigned long ino, ext4_fsblk_t block,
  619. const char *fmt, ...)
  620. __releases(bitlock)
  621. __acquires(bitlock)
  622. {
  623. struct va_format vaf;
  624. va_list args;
  625. struct ext4_super_block *es = EXT4_SB(sb)->s_es;
  626. es->s_last_error_ino = cpu_to_le32(ino);
  627. es->s_last_error_block = cpu_to_le64(block);
  628. __save_error_info(sb, function, line);
  629. va_start(args, fmt);
  630. vaf.fmt = fmt;
  631. vaf.va = &args;
  632. printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
  633. sb->s_id, function, line, grp);
  634. if (ino)
  635. printk(KERN_CONT "inode %lu: ", ino);
  636. if (block)
  637. printk(KERN_CONT "block %llu:", (unsigned long long) block);
  638. printk(KERN_CONT "%pV\n", &vaf);
  639. va_end(args);
  640. if (test_opt(sb, ERRORS_CONT)) {
  641. ext4_commit_super(sb, 0);
  642. return;
  643. }
  644. ext4_unlock_group(sb, grp);
  645. ext4_handle_error(sb);
  646. /*
  647. * We only get here in the ERRORS_RO case; relocking the group
  648. * may be dangerous, but nothing bad will happen since the
  649. * filesystem will have already been marked read/only and the
  650. * journal has been aborted. We return 1 as a hint to callers
  651. * who might what to use the return value from
  652. * ext4_grp_locked_error() to distinguish between the
  653. * ERRORS_CONT and ERRORS_RO case, and perhaps return more
  654. * aggressively from the ext4 function in question, with a
  655. * more appropriate error code.
  656. */
  657. ext4_lock_group(sb, grp);
  658. return;
  659. }
  660. void ext4_update_dynamic_rev(struct super_block *sb)
  661. {
  662. struct ext4_super_block *es = EXT4_SB(sb)->s_es;
  663. if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
  664. return;
  665. ext4_warning(sb,
  666. "updating to rev %d because of new feature flag, "
  667. "running e2fsck is recommended",
  668. EXT4_DYNAMIC_REV);
  669. es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
  670. es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
  671. es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
  672. /* leave es->s_feature_*compat flags alone */
  673. /* es->s_uuid will be set by e2fsck if empty */
  674. /*
  675. * The rest of the superblock fields should be zero, and if not it
  676. * means they are likely already in use, so leave them alone. We
  677. * can leave it up to e2fsck to clean up any inconsistencies there.
  678. */
  679. }
  680. /*
  681. * Open the external journal device
  682. */
  683. static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
  684. {
  685. struct block_device *bdev;
  686. char b[BDEVNAME_SIZE];
  687. bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
  688. if (IS_ERR(bdev))
  689. goto fail;
  690. return bdev;
  691. fail:
  692. ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
  693. __bdevname(dev, b), PTR_ERR(bdev));
  694. return NULL;
  695. }
  696. /*
  697. * Release the journal device
  698. */
  699. static int ext4_blkdev_put(struct block_device *bdev)
  700. {
  701. return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
  702. }
  703. static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
  704. {
  705. struct block_device *bdev;
  706. int ret = -ENODEV;
  707. bdev = sbi->journal_bdev;
  708. if (bdev) {
  709. ret = ext4_blkdev_put(bdev);
  710. sbi->journal_bdev = NULL;
  711. }
  712. return ret;
  713. }
  714. static inline struct inode *orphan_list_entry(struct list_head *l)
  715. {
  716. return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
  717. }
  718. static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
  719. {
  720. struct list_head *l;
  721. ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
  722. le32_to_cpu(sbi->s_es->s_last_orphan));
  723. printk(KERN_ERR "sb_info orphan list:\n");
  724. list_for_each(l, &sbi->s_orphan) {
  725. struct inode *inode = orphan_list_entry(l);
  726. printk(KERN_ERR " "
  727. "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
  728. inode->i_sb->s_id, inode->i_ino, inode,
  729. inode->i_mode, inode->i_nlink,
  730. NEXT_ORPHAN(inode));
  731. }
  732. }
  733. static void ext4_put_super(struct super_block *sb)
  734. {
  735. struct ext4_sb_info *sbi = EXT4_SB(sb);
  736. struct ext4_super_block *es = sbi->s_es;
  737. int i, err;
  738. ext4_unregister_li_request(sb);
  739. dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
  740. flush_workqueue(sbi->dio_unwritten_wq);
  741. destroy_workqueue(sbi->dio_unwritten_wq);
  742. if (sbi->s_journal) {
  743. err = jbd2_journal_destroy(sbi->s_journal);
  744. sbi->s_journal = NULL;
  745. if (err < 0)
  746. ext4_abort(sb, "Couldn't clean up the journal");
  747. }
  748. del_timer(&sbi->s_err_report);
  749. ext4_release_system_zone(sb);
  750. ext4_mb_release(sb);
  751. ext4_ext_release(sb);
  752. ext4_xattr_put_super(sb);
  753. if (!(sb->s_flags & MS_RDONLY)) {
  754. EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
  755. es->s_state = cpu_to_le16(sbi->s_mount_state);
  756. }
  757. if (!(sb->s_flags & MS_RDONLY))
  758. ext4_commit_super(sb, 1);
  759. if (sbi->s_proc) {
  760. remove_proc_entry("options", sbi->s_proc);
  761. remove_proc_entry(sb->s_id, ext4_proc_root);
  762. }
  763. kobject_del(&sbi->s_kobj);
  764. for (i = 0; i < sbi->s_gdb_count; i++)
  765. brelse(sbi->s_group_desc[i]);
  766. ext4_kvfree(sbi->s_group_desc);
  767. ext4_kvfree(sbi->s_flex_groups);
  768. percpu_counter_destroy(&sbi->s_freeclusters_counter);
  769. percpu_counter_destroy(&sbi->s_freeinodes_counter);
  770. percpu_counter_destroy(&sbi->s_dirs_counter);
  771. percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
  772. brelse(sbi->s_sbh);
  773. #ifdef CONFIG_QUOTA
  774. for (i = 0; i < MAXQUOTAS; i++)
  775. kfree(sbi->s_qf_names[i]);
  776. #endif
  777. /* Debugging code just in case the in-memory inode orphan list
  778. * isn't empty. The on-disk one can be non-empty if we've
  779. * detected an error and taken the fs readonly, but the
  780. * in-memory list had better be clean by this point. */
  781. if (!list_empty(&sbi->s_orphan))
  782. dump_orphan_list(sb, sbi);
  783. J_ASSERT(list_empty(&sbi->s_orphan));
  784. invalidate_bdev(sb->s_bdev);
  785. if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
  786. /*
  787. * Invalidate the journal device's buffers. We don't want them
  788. * floating about in memory - the physical journal device may
  789. * hotswapped, and it breaks the `ro-after' testing code.
  790. */
  791. sync_blockdev(sbi->journal_bdev);
  792. invalidate_bdev(sbi->journal_bdev);
  793. ext4_blkdev_remove(sbi);
  794. }
  795. if (sbi->s_mmp_tsk)
  796. kthread_stop(sbi->s_mmp_tsk);
  797. sb->s_fs_info = NULL;
  798. /*
  799. * Now that we are completely done shutting down the
  800. * superblock, we need to actually destroy the kobject.
  801. */
  802. kobject_put(&sbi->s_kobj);
  803. wait_for_completion(&sbi->s_kobj_unregister);
  804. if (sbi->s_chksum_driver)
  805. crypto_free_shash(sbi->s_chksum_driver);
  806. kfree(sbi->s_blockgroup_lock);
  807. kfree(sbi);
  808. }
  809. static struct kmem_cache *ext4_inode_cachep;
  810. /*
  811. * Called inside transaction, so use GFP_NOFS
  812. */
  813. static struct inode *ext4_alloc_inode(struct super_block *sb)
  814. {
  815. struct ext4_inode_info *ei;
  816. ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
  817. if (!ei)
  818. return NULL;
  819. ei->vfs_inode.i_version = 1;
  820. ei->vfs_inode.i_data.writeback_index = 0;
  821. memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
  822. INIT_LIST_HEAD(&ei->i_prealloc_list);
  823. spin_lock_init(&ei->i_prealloc_lock);
  824. ei->i_reserved_data_blocks = 0;
  825. ei->i_reserved_meta_blocks = 0;
  826. ei->i_allocated_meta_blocks = 0;
  827. ei->i_da_metadata_calc_len = 0;
  828. ei->i_da_metadata_calc_last_lblock = 0;
  829. spin_lock_init(&(ei->i_block_reservation_lock));
  830. #ifdef CONFIG_QUOTA
  831. ei->i_reserved_quota = 0;
  832. #endif
  833. ei->jinode = NULL;
  834. INIT_LIST_HEAD(&ei->i_completed_io_list);
  835. spin_lock_init(&ei->i_completed_io_lock);
  836. ei->cur_aio_dio = NULL;
  837. ei->i_sync_tid = 0;
  838. ei->i_datasync_tid = 0;
  839. atomic_set(&ei->i_ioend_count, 0);
  840. atomic_set(&ei->i_aiodio_unwritten, 0);
  841. return &ei->vfs_inode;
  842. }
  843. static int ext4_drop_inode(struct inode *inode)
  844. {
  845. int drop = generic_drop_inode(inode);
  846. trace_ext4_drop_inode(inode, drop);
  847. return drop;
  848. }
  849. static void ext4_i_callback(struct rcu_head *head)
  850. {
  851. struct inode *inode = container_of(head, struct inode, i_rcu);
  852. kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
  853. }
  854. static void ext4_destroy_inode(struct inode *inode)
  855. {
  856. if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
  857. ext4_msg(inode->i_sb, KERN_ERR,
  858. "Inode %lu (%p): orphan list check failed!",
  859. inode->i_ino, EXT4_I(inode));
  860. print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
  861. EXT4_I(inode), sizeof(struct ext4_inode_info),
  862. true);
  863. dump_stack();
  864. }
  865. call_rcu(&inode->i_rcu, ext4_i_callback);
  866. }
  867. static void init_once(void *foo)
  868. {
  869. struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
  870. INIT_LIST_HEAD(&ei->i_orphan);
  871. #ifdef CONFIG_EXT4_FS_XATTR
  872. init_rwsem(&ei->xattr_sem);
  873. #endif
  874. init_rwsem(&ei->i_data_sem);
  875. inode_init_once(&ei->vfs_inode);
  876. }
  877. static int init_inodecache(void)
  878. {
  879. ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
  880. sizeof(struct ext4_inode_info),
  881. 0, (SLAB_RECLAIM_ACCOUNT|
  882. SLAB_MEM_SPREAD),
  883. init_once);
  884. if (ext4_inode_cachep == NULL)
  885. return -ENOMEM;
  886. return 0;
  887. }
  888. static void destroy_inodecache(void)
  889. {
  890. kmem_cache_destroy(ext4_inode_cachep);
  891. }
  892. void ext4_clear_inode(struct inode *inode)
  893. {
  894. invalidate_inode_buffers(inode);
  895. clear_inode(inode);
  896. dquot_drop(inode);
  897. ext4_discard_preallocations(inode);
  898. if (EXT4_I(inode)->jinode) {
  899. jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
  900. EXT4_I(inode)->jinode);
  901. jbd2_free_inode(EXT4_I(inode)->jinode);
  902. EXT4_I(inode)->jinode = NULL;
  903. }
  904. }
  905. static struct inode *ext4_nfs_get_inode(struct super_block *sb,
  906. u64 ino, u32 generation)
  907. {
  908. struct inode *inode;
  909. if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
  910. return ERR_PTR(-ESTALE);
  911. if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
  912. return ERR_PTR(-ESTALE);
  913. /* iget isn't really right if the inode is currently unallocated!!
  914. *
  915. * ext4_read_inode will return a bad_inode if the inode had been
  916. * deleted, so we should be safe.
  917. *
  918. * Currently we don't know the generation for parent directory, so
  919. * a generation of 0 means "accept any"
  920. */
  921. inode = ext4_iget(sb, ino);
  922. if (IS_ERR(inode))
  923. return ERR_CAST(inode);
  924. if (generation && inode->i_generation != generation) {
  925. iput(inode);
  926. return ERR_PTR(-ESTALE);
  927. }
  928. return inode;
  929. }
  930. static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
  931. int fh_len, int fh_type)
  932. {
  933. return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
  934. ext4_nfs_get_inode);
  935. }
  936. static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
  937. int fh_len, int fh_type)
  938. {
  939. return generic_fh_to_parent(sb, fid, fh_len, fh_type,
  940. ext4_nfs_get_inode);
  941. }
  942. /*
  943. * Try to release metadata pages (indirect blocks, directories) which are
  944. * mapped via the block device. Since these pages could have journal heads
  945. * which would prevent try_to_free_buffers() from freeing them, we must use
  946. * jbd2 layer's try_to_free_buffers() function to release them.
  947. */
  948. static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
  949. gfp_t wait)
  950. {
  951. journal_t *journal = EXT4_SB(sb)->s_journal;
  952. WARN_ON(PageChecked(page));
  953. if (!page_has_buffers(page))
  954. return 0;
  955. if (journal)
  956. return jbd2_journal_try_to_free_buffers(journal, page,
  957. wait & ~__GFP_WAIT);
  958. return try_to_free_buffers(page);
  959. }
  960. #ifdef CONFIG_QUOTA
  961. #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
  962. #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
  963. static int ext4_write_dquot(struct dquot *dquot);
  964. static int ext4_acquire_dquot(struct dquot *dquot);
  965. static int ext4_release_dquot(struct dquot *dquot);
  966. static int ext4_mark_dquot_dirty(struct dquot *dquot);
  967. static int ext4_write_info(struct super_block *sb, int type);
  968. static int ext4_quota_on(struct super_block *sb, int type, int format_id,
  969. struct path *path);
  970. static int ext4_quota_on_sysfile(struct super_block *sb, int type,
  971. int format_id);
  972. static int ext4_quota_off(struct super_block *sb, int type);
  973. static int ext4_quota_off_sysfile(struct super_block *sb, int type);
  974. static int ext4_quota_on_mount(struct super_block *sb, int type);
  975. static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
  976. size_t len, loff_t off);
  977. static ssize_t ext4_quota_write(struct super_block *sb, int type,
  978. const char *data, size_t len, loff_t off);
  979. static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
  980. unsigned int flags);
  981. static int ext4_enable_quotas(struct super_block *sb);
  982. static const struct dquot_operations ext4_quota_operations = {
  983. .get_reserved_space = ext4_get_reserved_space,
  984. .write_dquot = ext4_write_dquot,
  985. .acquire_dquot = ext4_acquire_dquot,
  986. .release_dquot = ext4_release_dquot,
  987. .mark_dirty = ext4_mark_dquot_dirty,
  988. .write_info = ext4_write_info,
  989. .alloc_dquot = dquot_alloc,
  990. .destroy_dquot = dquot_destroy,
  991. };
  992. static const struct quotactl_ops ext4_qctl_operations = {
  993. .quota_on = ext4_quota_on,
  994. .quota_off = ext4_quota_off,
  995. .quota_sync = dquot_quota_sync,
  996. .get_info = dquot_get_dqinfo,
  997. .set_info = dquot_set_dqinfo,
  998. .get_dqblk = dquot_get_dqblk,
  999. .set_dqblk = dquot_set_dqblk
  1000. };
  1001. static const struct quotactl_ops ext4_qctl_sysfile_operations = {
  1002. .quota_on_meta = ext4_quota_on_sysfile,
  1003. .quota_off = ext4_quota_off_sysfile,
  1004. .quota_sync = dquot_quota_sync,
  1005. .get_info = dquot_get_dqinfo,
  1006. .set_info = dquot_set_dqinfo,
  1007. .get_dqblk = dquot_get_dqblk,
  1008. .set_dqblk = dquot_set_dqblk
  1009. };
  1010. #endif
  1011. static const struct super_operations ext4_sops = {
  1012. .alloc_inode = ext4_alloc_inode,
  1013. .destroy_inode = ext4_destroy_inode,
  1014. .write_inode = ext4_write_inode,
  1015. .dirty_inode = ext4_dirty_inode,
  1016. .drop_inode = ext4_drop_inode,
  1017. .evict_inode = ext4_evict_inode,
  1018. .put_super = ext4_put_super,
  1019. .sync_fs = ext4_sync_fs,
  1020. .freeze_fs = ext4_freeze,
  1021. .unfreeze_fs = ext4_unfreeze,
  1022. .statfs = ext4_statfs,
  1023. .remount_fs = ext4_remount,
  1024. .show_options = ext4_show_options,
  1025. #ifdef CONFIG_QUOTA
  1026. .quota_read = ext4_quota_read,
  1027. .quota_write = ext4_quota_write,
  1028. #endif
  1029. .bdev_try_to_free_page = bdev_try_to_free_page,
  1030. };
  1031. static const struct super_operations ext4_nojournal_sops = {
  1032. .alloc_inode = ext4_alloc_inode,
  1033. .destroy_inode = ext4_destroy_inode,
  1034. .write_inode = ext4_write_inode,
  1035. .dirty_inode = ext4_dirty_inode,
  1036. .drop_inode = ext4_drop_inode,
  1037. .evict_inode = ext4_evict_inode,
  1038. .put_super = ext4_put_super,
  1039. .statfs = ext4_statfs,
  1040. .remount_fs = ext4_remount,
  1041. .show_options = ext4_show_options,
  1042. #ifdef CONFIG_QUOTA
  1043. .quota_read = ext4_quota_read,
  1044. .quota_write = ext4_quota_write,
  1045. #endif
  1046. .bdev_try_to_free_page = bdev_try_to_free_page,
  1047. };
  1048. static const struct export_operations ext4_export_ops = {
  1049. .fh_to_dentry = ext4_fh_to_dentry,
  1050. .fh_to_parent = ext4_fh_to_parent,
  1051. .get_parent = ext4_get_parent,
  1052. };
  1053. enum {
  1054. Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
  1055. Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
  1056. Opt_nouid32, Opt_debug, Opt_removed,
  1057. Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
  1058. Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
  1059. Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
  1060. Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
  1061. Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
  1062. Opt_data_err_abort, Opt_data_err_ignore,
  1063. Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
  1064. Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
  1065. Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
  1066. Opt_usrquota, Opt_grpquota, Opt_i_version,
  1067. Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
  1068. Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
  1069. Opt_inode_readahead_blks, Opt_journal_ioprio,
  1070. Opt_dioread_nolock, Opt_dioread_lock,
  1071. Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
  1072. Opt_max_dir_size_kb,
  1073. };
  1074. static const match_table_t tokens = {
  1075. {Opt_bsd_df, "bsddf"},
  1076. {Opt_minix_df, "minixdf"},
  1077. {Opt_grpid, "grpid"},
  1078. {Opt_grpid, "bsdgroups"},
  1079. {Opt_nogrpid, "nogrpid"},
  1080. {Opt_nogrpid, "sysvgroups"},
  1081. {Opt_resgid, "resgid=%u"},
  1082. {Opt_resuid, "resuid=%u"},
  1083. {Opt_sb, "sb=%u"},
  1084. {Opt_err_cont, "errors=continue"},
  1085. {Opt_err_panic, "errors=panic"},
  1086. {Opt_err_ro, "errors=remount-ro"},
  1087. {Opt_nouid32, "nouid32"},
  1088. {Opt_debug, "debug"},
  1089. {Opt_removed, "oldalloc"},
  1090. {Opt_removed, "orlov"},
  1091. {Opt_user_xattr, "user_xattr"},
  1092. {Opt_nouser_xattr, "nouser_xattr"},
  1093. {Opt_acl, "acl"},
  1094. {Opt_noacl, "noacl"},
  1095. {Opt_noload, "norecovery"},
  1096. {Opt_noload, "noload"},
  1097. {Opt_removed, "nobh"},
  1098. {Opt_removed, "bh"},
  1099. {Opt_commit, "commit=%u"},
  1100. {Opt_min_batch_time, "min_batch_time=%u"},
  1101. {Opt_max_batch_time, "max_batch_time=%u"},
  1102. {Opt_journal_dev, "journal_dev=%u"},
  1103. {Opt_journal_checksum, "journal_checksum"},
  1104. {Opt_journal_async_commit, "journal_async_commit"},
  1105. {Opt_abort, "abort"},
  1106. {Opt_data_journal, "data=journal"},
  1107. {Opt_data_ordered, "data=ordered"},
  1108. {Opt_data_writeback, "data=writeback"},
  1109. {Opt_data_err_abort, "data_err=abort"},
  1110. {Opt_data_err_ignore, "data_err=ignore"},
  1111. {Opt_offusrjquota, "usrjquota="},
  1112. {Opt_usrjquota, "usrjquota=%s"},
  1113. {Opt_offgrpjquota, "grpjquota="},
  1114. {Opt_grpjquota, "grpjquota=%s"},
  1115. {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
  1116. {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
  1117. {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
  1118. {Opt_grpquota, "grpquota"},
  1119. {Opt_noquota, "noquota"},
  1120. {Opt_quota, "quota"},
  1121. {Opt_usrquota, "usrquota"},
  1122. {Opt_barrier, "barrier=%u"},
  1123. {Opt_barrier, "barrier"},
  1124. {Opt_nobarrier, "nobarrier"},
  1125. {Opt_i_version, "i_version"},
  1126. {Opt_stripe, "stripe=%u"},
  1127. {Opt_delalloc, "delalloc"},
  1128. {Opt_nodelalloc, "nodelalloc"},
  1129. {Opt_mblk_io_submit, "mblk_io_submit"},
  1130. {Opt_nomblk_io_submit, "nomblk_io_submit"},
  1131. {Opt_block_validity, "block_validity"},
  1132. {Opt_noblock_validity, "noblock_validity"},
  1133. {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
  1134. {Opt_journal_ioprio, "journal_ioprio=%u"},
  1135. {Opt_auto_da_alloc, "auto_da_alloc=%u"},
  1136. {Opt_auto_da_alloc, "auto_da_alloc"},
  1137. {Opt_noauto_da_alloc, "noauto_da_alloc"},
  1138. {Opt_dioread_nolock, "dioread_nolock"},
  1139. {Opt_dioread_lock, "dioread_lock"},
  1140. {Opt_discard, "discard"},
  1141. {Opt_nodiscard, "nodiscard"},
  1142. {Opt_init_itable, "init_itable=%u"},
  1143. {Opt_init_itable, "init_itable"},
  1144. {Opt_noinit_itable, "noinit_itable"},
  1145. {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
  1146. {Opt_removed, "check=none"}, /* mount option from ext2/3 */
  1147. {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
  1148. {Opt_removed, "reservation"}, /* mount option from ext2/3 */
  1149. {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
  1150. {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
  1151. {Opt_err, NULL},
  1152. };
  1153. static ext4_fsblk_t get_sb_block(void **data)
  1154. {
  1155. ext4_fsblk_t sb_block;
  1156. char *options = (char *) *data;
  1157. if (!options || strncmp(options, "sb=", 3) != 0)
  1158. return 1; /* Default location */
  1159. options += 3;
  1160. /* TODO: use simple_strtoll with >32bit ext4 */
  1161. sb_block = simple_strtoul(options, &options, 0);
  1162. if (*options && *options != ',') {
  1163. printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
  1164. (char *) *data);
  1165. return 1;
  1166. }
  1167. if (*options == ',')
  1168. options++;
  1169. *data = (void *) options;
  1170. return sb_block;
  1171. }
  1172. #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
  1173. static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
  1174. "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
  1175. #ifdef CONFIG_QUOTA
  1176. static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
  1177. {
  1178. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1179. char *qname;
  1180. if (sb_any_quota_loaded(sb) &&
  1181. !sbi->s_qf_names[qtype]) {
  1182. ext4_msg(sb, KERN_ERR,
  1183. "Cannot change journaled "
  1184. "quota options when quota turned on");
  1185. return -1;
  1186. }
  1187. qname = match_strdup(args);
  1188. if (!qname) {
  1189. ext4_msg(sb, KERN_ERR,
  1190. "Not enough memory for storing quotafile name");
  1191. return -1;
  1192. }
  1193. if (sbi->s_qf_names[qtype] &&
  1194. strcmp(sbi->s_qf_names[qtype], qname)) {
  1195. ext4_msg(sb, KERN_ERR,
  1196. "%s quota file already specified", QTYPE2NAME(qtype));
  1197. kfree(qname);
  1198. return -1;
  1199. }
  1200. sbi->s_qf_names[qtype] = qname;
  1201. if (strchr(sbi->s_qf_names[qtype], '/')) {
  1202. ext4_msg(sb, KERN_ERR,
  1203. "quotafile must be on filesystem root");
  1204. kfree(sbi->s_qf_names[qtype]);
  1205. sbi->s_qf_names[qtype] = NULL;
  1206. return -1;
  1207. }
  1208. set_opt(sb, QUOTA);
  1209. return 1;
  1210. }
  1211. static int clear_qf_name(struct super_block *sb, int qtype)
  1212. {
  1213. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1214. if (sb_any_quota_loaded(sb) &&
  1215. sbi->s_qf_names[qtype]) {
  1216. ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
  1217. " when quota turned on");
  1218. return -1;
  1219. }
  1220. /*
  1221. * The space will be released later when all options are confirmed
  1222. * to be correct
  1223. */
  1224. sbi->s_qf_names[qtype] = NULL;
  1225. return 1;
  1226. }
  1227. #endif
  1228. #define MOPT_SET 0x0001
  1229. #define MOPT_CLEAR 0x0002
  1230. #define MOPT_NOSUPPORT 0x0004
  1231. #define MOPT_EXPLICIT 0x0008
  1232. #define MOPT_CLEAR_ERR 0x0010
  1233. #define MOPT_GTE0 0x0020
  1234. #ifdef CONFIG_QUOTA
  1235. #define MOPT_Q 0
  1236. #define MOPT_QFMT 0x0040
  1237. #else
  1238. #define MOPT_Q MOPT_NOSUPPORT
  1239. #define MOPT_QFMT MOPT_NOSUPPORT
  1240. #endif
  1241. #define MOPT_DATAJ 0x0080
  1242. static const struct mount_opts {
  1243. int token;
  1244. int mount_opt;
  1245. int flags;
  1246. } ext4_mount_opts[] = {
  1247. {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
  1248. {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
  1249. {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
  1250. {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
  1251. {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
  1252. {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
  1253. {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
  1254. {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
  1255. {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
  1256. {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
  1257. {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
  1258. {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
  1259. {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
  1260. {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
  1261. {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
  1262. {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
  1263. EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
  1264. {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
  1265. {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
  1266. {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
  1267. {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
  1268. {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
  1269. {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
  1270. {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
  1271. {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
  1272. {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
  1273. {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
  1274. {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
  1275. {Opt_commit, 0, MOPT_GTE0},
  1276. {Opt_max_batch_time, 0, MOPT_GTE0},
  1277. {Opt_min_batch_time, 0, MOPT_GTE0},
  1278. {Opt_inode_readahead_blks, 0, MOPT_GTE0},
  1279. {Opt_init_itable, 0, MOPT_GTE0},
  1280. {Opt_stripe, 0, MOPT_GTE0},
  1281. {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
  1282. {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
  1283. {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
  1284. #ifdef CONFIG_EXT4_FS_XATTR
  1285. {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
  1286. {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
  1287. #else
  1288. {Opt_user_xattr, 0, MOPT_NOSUPPORT},
  1289. {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
  1290. #endif
  1291. #ifdef CONFIG_EXT4_FS_POSIX_ACL
  1292. {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
  1293. {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
  1294. #else
  1295. {Opt_acl, 0, MOPT_NOSUPPORT},
  1296. {Opt_noacl, 0, MOPT_NOSUPPORT},
  1297. #endif
  1298. {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
  1299. {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
  1300. {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
  1301. {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
  1302. MOPT_SET | MOPT_Q},
  1303. {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
  1304. MOPT_SET | MOPT_Q},
  1305. {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
  1306. EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
  1307. {Opt_usrjquota, 0, MOPT_Q},
  1308. {Opt_grpjquota, 0, MOPT_Q},
  1309. {Opt_offusrjquota, 0, MOPT_Q},
  1310. {Opt_offgrpjquota, 0, MOPT_Q},
  1311. {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
  1312. {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
  1313. {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
  1314. {Opt_max_dir_size_kb, 0, MOPT_GTE0},
  1315. {Opt_err, 0, 0}
  1316. };
  1317. static int handle_mount_opt(struct super_block *sb, char *opt, int token,
  1318. substring_t *args, unsigned long *journal_devnum,
  1319. unsigned int *journal_ioprio, int is_remount)
  1320. {
  1321. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1322. const struct mount_opts *m;
  1323. kuid_t uid;
  1324. kgid_t gid;
  1325. int arg = 0;
  1326. #ifdef CONFIG_QUOTA
  1327. if (token == Opt_usrjquota)
  1328. return set_qf_name(sb, USRQUOTA, &args[0]);
  1329. else if (token == Opt_grpjquota)
  1330. return set_qf_name(sb, GRPQUOTA, &args[0]);
  1331. else if (token == Opt_offusrjquota)
  1332. return clear_qf_name(sb, USRQUOTA);
  1333. else if (token == Opt_offgrpjquota)
  1334. return clear_qf_name(sb, GRPQUOTA);
  1335. #endif
  1336. if (args->from && match_int(args, &arg))
  1337. return -1;
  1338. switch (token) {
  1339. case Opt_noacl:
  1340. case Opt_nouser_xattr:
  1341. ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
  1342. break;
  1343. case Opt_sb:
  1344. return 1; /* handled by get_sb_block() */
  1345. case Opt_removed:
  1346. ext4_msg(sb, KERN_WARNING,
  1347. "Ignoring removed %s option", opt);
  1348. return 1;
  1349. case Opt_resuid:
  1350. uid = make_kuid(current_user_ns(), arg);
  1351. if (!uid_valid(uid)) {
  1352. ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
  1353. return -1;
  1354. }
  1355. sbi->s_resuid = uid;
  1356. return 1;
  1357. case Opt_resgid:
  1358. gid = make_kgid(current_user_ns(), arg);
  1359. if (!gid_valid(gid)) {
  1360. ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
  1361. return -1;
  1362. }
  1363. sbi->s_resgid = gid;
  1364. return 1;
  1365. case Opt_abort:
  1366. sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
  1367. return 1;
  1368. case Opt_i_version:
  1369. sb->s_flags |= MS_I_VERSION;
  1370. return 1;
  1371. case Opt_journal_dev:
  1372. if (is_remount) {
  1373. ext4_msg(sb, KERN_ERR,
  1374. "Cannot specify journal on remount");
  1375. return -1;
  1376. }
  1377. *journal_devnum = arg;
  1378. return 1;
  1379. case Opt_journal_ioprio:
  1380. if (arg < 0 || arg > 7)
  1381. return -1;
  1382. *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
  1383. return 1;
  1384. }
  1385. for (m = ext4_mount_opts; m->token != Opt_err; m++) {
  1386. if (token != m->token)
  1387. continue;
  1388. if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
  1389. return -1;
  1390. if (m->flags & MOPT_EXPLICIT)
  1391. set_opt2(sb, EXPLICIT_DELALLOC);
  1392. if (m->flags & MOPT_CLEAR_ERR)
  1393. clear_opt(sb, ERRORS_MASK);
  1394. if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
  1395. ext4_msg(sb, KERN_ERR, "Cannot change quota "
  1396. "options when quota turned on");
  1397. return -1;
  1398. }
  1399. if (m->flags & MOPT_NOSUPPORT) {
  1400. ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
  1401. } else if (token == Opt_commit) {
  1402. if (arg == 0)
  1403. arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
  1404. sbi->s_commit_interval = HZ * arg;
  1405. } else if (token == Opt_max_batch_time) {
  1406. if (arg == 0)
  1407. arg = EXT4_DEF_MAX_BATCH_TIME;
  1408. sbi->s_max_batch_time = arg;
  1409. } else if (token == Opt_min_batch_time) {
  1410. sbi->s_min_batch_time = arg;
  1411. } else if (token == Opt_inode_readahead_blks) {
  1412. if (arg > (1 << 30))
  1413. return -1;
  1414. if (arg && !is_power_of_2(arg)) {
  1415. ext4_msg(sb, KERN_ERR,
  1416. "EXT4-fs: inode_readahead_blks"
  1417. " must be a power of 2");
  1418. return -1;
  1419. }
  1420. sbi->s_inode_readahead_blks = arg;
  1421. } else if (token == Opt_init_itable) {
  1422. set_opt(sb, INIT_INODE_TABLE);
  1423. if (!args->from)
  1424. arg = EXT4_DEF_LI_WAIT_MULT;
  1425. sbi->s_li_wait_mult = arg;
  1426. } else if (token == Opt_max_dir_size_kb) {
  1427. sbi->s_max_dir_size_kb = arg;
  1428. } else if (token == Opt_stripe) {
  1429. sbi->s_stripe = arg;
  1430. } else if (m->flags & MOPT_DATAJ) {
  1431. if (is_remount) {
  1432. if (!sbi->s_journal)
  1433. ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
  1434. else if (test_opt(sb, DATA_FLAGS) !=
  1435. m->mount_opt) {
  1436. ext4_msg(sb, KERN_ERR,
  1437. "Cannot change data mode on remount");
  1438. return -1;
  1439. }
  1440. } else {
  1441. clear_opt(sb, DATA_FLAGS);
  1442. sbi->s_mount_opt |= m->mount_opt;
  1443. }
  1444. #ifdef CONFIG_QUOTA
  1445. } else if (m->flags & MOPT_QFMT) {
  1446. if (sb_any_quota_loaded(sb) &&
  1447. sbi->s_jquota_fmt != m->mount_opt) {
  1448. ext4_msg(sb, KERN_ERR, "Cannot "
  1449. "change journaled quota options "
  1450. "when quota turned on");
  1451. return -1;
  1452. }
  1453. sbi->s_jquota_fmt = m->mount_opt;
  1454. #endif
  1455. } else {
  1456. if (!args->from)
  1457. arg = 1;
  1458. if (m->flags & MOPT_CLEAR)
  1459. arg = !arg;
  1460. else if (unlikely(!(m->flags & MOPT_SET))) {
  1461. ext4_msg(sb, KERN_WARNING,
  1462. "buggy handling of option %s", opt);
  1463. WARN_ON(1);
  1464. return -1;
  1465. }
  1466. if (arg != 0)
  1467. sbi->s_mount_opt |= m->mount_opt;
  1468. else
  1469. sbi->s_mount_opt &= ~m->mount_opt;
  1470. }
  1471. return 1;
  1472. }
  1473. ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
  1474. "or missing value", opt);
  1475. return -1;
  1476. }
  1477. static int parse_options(char *options, struct super_block *sb,
  1478. unsigned long *journal_devnum,
  1479. unsigned int *journal_ioprio,
  1480. int is_remount)
  1481. {
  1482. #ifdef CONFIG_QUOTA
  1483. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1484. #endif
  1485. char *p;
  1486. substring_t args[MAX_OPT_ARGS];
  1487. int token;
  1488. if (!options)
  1489. return 1;
  1490. while ((p = strsep(&options, ",")) != NULL) {
  1491. if (!*p)
  1492. continue;
  1493. /*
  1494. * Initialize args struct so we know whether arg was
  1495. * found; some options take optional arguments.
  1496. */
  1497. args[0].to = args[0].from = NULL;
  1498. token = match_token(p, tokens, args);
  1499. if (handle_mount_opt(sb, p, token, args, journal_devnum,
  1500. journal_ioprio, is_remount) < 0)
  1501. return 0;
  1502. }
  1503. #ifdef CONFIG_QUOTA
  1504. if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
  1505. if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
  1506. clear_opt(sb, USRQUOTA);
  1507. if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
  1508. clear_opt(sb, GRPQUOTA);
  1509. if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
  1510. ext4_msg(sb, KERN_ERR, "old and new quota "
  1511. "format mixing");
  1512. return 0;
  1513. }
  1514. if (!sbi->s_jquota_fmt) {
  1515. ext4_msg(sb, KERN_ERR, "journaled quota format "
  1516. "not specified");
  1517. return 0;
  1518. }
  1519. } else {
  1520. if (sbi->s_jquota_fmt) {
  1521. ext4_msg(sb, KERN_ERR, "journaled quota format "
  1522. "specified with no journaling "
  1523. "enabled");
  1524. return 0;
  1525. }
  1526. }
  1527. #endif
  1528. return 1;
  1529. }
  1530. static inline void ext4_show_quota_options(struct seq_file *seq,
  1531. struct super_block *sb)
  1532. {
  1533. #if defined(CONFIG_QUOTA)
  1534. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1535. if (sbi->s_jquota_fmt) {
  1536. char *fmtname = "";
  1537. switch (sbi->s_jquota_fmt) {
  1538. case QFMT_VFS_OLD:
  1539. fmtname = "vfsold";
  1540. break;
  1541. case QFMT_VFS_V0:
  1542. fmtname = "vfsv0";
  1543. break;
  1544. case QFMT_VFS_V1:
  1545. fmtname = "vfsv1";
  1546. break;
  1547. }
  1548. seq_printf(seq, ",jqfmt=%s", fmtname);
  1549. }
  1550. if (sbi->s_qf_names[USRQUOTA])
  1551. seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
  1552. if (sbi->s_qf_names[GRPQUOTA])
  1553. seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
  1554. if (test_opt(sb, USRQUOTA))
  1555. seq_puts(seq, ",usrquota");
  1556. if (test_opt(sb, GRPQUOTA))
  1557. seq_puts(seq, ",grpquota");
  1558. #endif
  1559. }
  1560. static const char *token2str(int token)
  1561. {
  1562. const struct match_token *t;
  1563. for (t = tokens; t->token != Opt_err; t++)
  1564. if (t->token == token && !strchr(t->pattern, '='))
  1565. break;
  1566. return t->pattern;
  1567. }
  1568. /*
  1569. * Show an option if
  1570. * - it's set to a non-default value OR
  1571. * - if the per-sb default is different from the global default
  1572. */
  1573. static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
  1574. int nodefs)
  1575. {
  1576. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1577. struct ext4_super_block *es = sbi->s_es;
  1578. int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
  1579. const struct mount_opts *m;
  1580. char sep = nodefs ? '\n' : ',';
  1581. #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
  1582. #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
  1583. if (sbi->s_sb_block != 1)
  1584. SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
  1585. for (m = ext4_mount_opts; m->token != Opt_err; m++) {
  1586. int want_set = m->flags & MOPT_SET;
  1587. if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
  1588. (m->flags & MOPT_CLEAR_ERR))
  1589. continue;
  1590. if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
  1591. continue; /* skip if same as the default */
  1592. if ((want_set &&
  1593. (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
  1594. (!want_set && (sbi->s_mount_opt & m->mount_opt)))
  1595. continue; /* select Opt_noFoo vs Opt_Foo */
  1596. SEQ_OPTS_PRINT("%s", token2str(m->token));
  1597. }
  1598. if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
  1599. le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
  1600. SEQ_OPTS_PRINT("resuid=%u",
  1601. from_kuid_munged(&init_user_ns, sbi->s_resuid));
  1602. if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
  1603. le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
  1604. SEQ_OPTS_PRINT("resgid=%u",
  1605. from_kgid_munged(&init_user_ns, sbi->s_resgid));
  1606. def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
  1607. if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
  1608. SEQ_OPTS_PUTS("errors=remount-ro");
  1609. if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
  1610. SEQ_OPTS_PUTS("errors=continue");
  1611. if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
  1612. SEQ_OPTS_PUTS("errors=panic");
  1613. if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
  1614. SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
  1615. if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
  1616. SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
  1617. if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
  1618. SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
  1619. if (sb->s_flags & MS_I_VERSION)
  1620. SEQ_OPTS_PUTS("i_version");
  1621. if (nodefs || sbi->s_stripe)
  1622. SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
  1623. if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
  1624. if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
  1625. SEQ_OPTS_PUTS("data=journal");
  1626. else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
  1627. SEQ_OPTS_PUTS("data=ordered");
  1628. else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
  1629. SEQ_OPTS_PUTS("data=writeback");
  1630. }
  1631. if (nodefs ||
  1632. sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
  1633. SEQ_OPTS_PRINT("inode_readahead_blks=%u",
  1634. sbi->s_inode_readahead_blks);
  1635. if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
  1636. (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
  1637. SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
  1638. if (nodefs || sbi->s_max_dir_size_kb)
  1639. SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
  1640. ext4_show_quota_options(seq, sb);
  1641. return 0;
  1642. }
  1643. static int ext4_show_options(struct seq_file *seq, struct dentry *root)
  1644. {
  1645. return _ext4_show_options(seq, root->d_sb, 0);
  1646. }
  1647. static int options_seq_show(struct seq_file *seq, void *offset)
  1648. {
  1649. struct super_block *sb = seq->private;
  1650. int rc;
  1651. seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
  1652. rc = _ext4_show_options(seq, sb, 1);
  1653. seq_puts(seq, "\n");
  1654. return rc;
  1655. }
  1656. static int options_open_fs(struct inode *inode, struct file *file)
  1657. {
  1658. return single_open(file, options_seq_show, PDE(inode)->data);
  1659. }
  1660. static const struct file_operations ext4_seq_options_fops = {
  1661. .owner = THIS_MODULE,
  1662. .open = options_open_fs,
  1663. .read = seq_read,
  1664. .llseek = seq_lseek,
  1665. .release = single_release,
  1666. };
  1667. static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
  1668. int read_only)
  1669. {
  1670. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1671. int res = 0;
  1672. if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
  1673. ext4_msg(sb, KERN_ERR, "revision level too high, "
  1674. "forcing read-only mode");
  1675. res = MS_RDONLY;
  1676. }
  1677. if (read_only)
  1678. goto done;
  1679. if (!(sbi->s_mount_state & EXT4_VALID_FS))
  1680. ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
  1681. "running e2fsck is recommended");
  1682. else if ((sbi->s_mount_state & EXT4_ERROR_FS))
  1683. ext4_msg(sb, KERN_WARNING,
  1684. "warning: mounting fs with errors, "
  1685. "running e2fsck is recommended");
  1686. else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
  1687. le16_to_cpu(es->s_mnt_count) >=
  1688. (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
  1689. ext4_msg(sb, KERN_WARNING,
  1690. "warning: maximal mount count reached, "
  1691. "running e2fsck is recommended");
  1692. else if (le32_to_cpu(es->s_checkinterval) &&
  1693. (le32_to_cpu(es->s_lastcheck) +
  1694. le32_to_cpu(es->s_checkinterval) <= get_seconds()))
  1695. ext4_msg(sb, KERN_WARNING,
  1696. "warning: checktime reached, "
  1697. "running e2fsck is recommended");
  1698. if (!sbi->s_journal)
  1699. es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
  1700. if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
  1701. es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
  1702. le16_add_cpu(&es->s_mnt_count, 1);
  1703. es->s_mtime = cpu_to_le32(get_seconds());
  1704. ext4_update_dynamic_rev(sb);
  1705. if (sbi->s_journal)
  1706. EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
  1707. ext4_commit_super(sb, 1);
  1708. done:
  1709. if (test_opt(sb, DEBUG))
  1710. printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
  1711. "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
  1712. sb->s_blocksize,
  1713. sbi->s_groups_count,
  1714. EXT4_BLOCKS_PER_GROUP(sb),
  1715. EXT4_INODES_PER_GROUP(sb),
  1716. sbi->s_mount_opt, sbi->s_mount_opt2);
  1717. cleancache_init_fs(sb);
  1718. return res;
  1719. }
  1720. int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
  1721. {
  1722. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1723. struct flex_groups *new_groups;
  1724. int size;
  1725. if (!sbi->s_log_groups_per_flex)
  1726. return 0;
  1727. size = ext4_flex_group(sbi, ngroup - 1) + 1;
  1728. if (size <= sbi->s_flex_groups_allocated)
  1729. return 0;
  1730. size = roundup_pow_of_two(size * sizeof(struct flex_groups));
  1731. new_groups = ext4_kvzalloc(size, GFP_KERNEL);
  1732. if (!new_groups) {
  1733. ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
  1734. size / (int) sizeof(struct flex_groups));
  1735. return -ENOMEM;
  1736. }
  1737. if (sbi->s_flex_groups) {
  1738. memcpy(new_groups, sbi->s_flex_groups,
  1739. (sbi->s_flex_groups_allocated *
  1740. sizeof(struct flex_groups)));
  1741. ext4_kvfree(sbi->s_flex_groups);
  1742. }
  1743. sbi->s_flex_groups = new_groups;
  1744. sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
  1745. return 0;
  1746. }
  1747. static int ext4_fill_flex_info(struct super_block *sb)
  1748. {
  1749. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1750. struct ext4_group_desc *gdp = NULL;
  1751. ext4_group_t flex_group;
  1752. unsigned int groups_per_flex = 0;
  1753. int i, err;
  1754. sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
  1755. if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
  1756. sbi->s_log_groups_per_flex = 0;
  1757. return 1;
  1758. }
  1759. groups_per_flex = 1 << sbi->s_log_groups_per_flex;
  1760. err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
  1761. if (err)
  1762. goto failed;
  1763. for (i = 0; i < sbi->s_groups_count; i++) {
  1764. gdp = ext4_get_group_desc(sb, i, NULL);
  1765. flex_group = ext4_flex_group(sbi, i);
  1766. atomic_add(ext4_free_inodes_count(sb, gdp),
  1767. &sbi->s_flex_groups[flex_group].free_inodes);
  1768. atomic_add(ext4_free_group_clusters(sb, gdp),
  1769. &sbi->s_flex_groups[flex_group].free_clusters);
  1770. atomic_add(ext4_used_dirs_count(sb, gdp),
  1771. &sbi->s_flex_groups[flex_group].used_dirs);
  1772. }
  1773. return 1;
  1774. failed:
  1775. return 0;
  1776. }
  1777. static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
  1778. struct ext4_group_desc *gdp)
  1779. {
  1780. int offset;
  1781. __u16 crc = 0;
  1782. __le32 le_group = cpu_to_le32(block_group);
  1783. if ((sbi->s_es->s_feature_ro_compat &
  1784. cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
  1785. /* Use new metadata_csum algorithm */
  1786. __u16 old_csum;
  1787. __u32 csum32;
  1788. old_csum = gdp->bg_checksum;
  1789. gdp->bg_checksum = 0;
  1790. csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
  1791. sizeof(le_group));
  1792. csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
  1793. sbi->s_desc_size);
  1794. gdp->bg_checksum = old_csum;
  1795. crc = csum32 & 0xFFFF;
  1796. goto out;
  1797. }
  1798. /* old crc16 code */
  1799. offset = offsetof(struct ext4_group_desc, bg_checksum);
  1800. crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
  1801. crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
  1802. crc = crc16(crc, (__u8 *)gdp, offset);
  1803. offset += sizeof(gdp->bg_checksum); /* skip checksum */
  1804. /* for checksum of struct ext4_group_desc do the rest...*/
  1805. if ((sbi->s_es->s_feature_incompat &
  1806. cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
  1807. offset < le16_to_cpu(sbi->s_es->s_desc_size))
  1808. crc = crc16(crc, (__u8 *)gdp + offset,
  1809. le16_to_cpu(sbi->s_es->s_desc_size) -
  1810. offset);
  1811. out:
  1812. return cpu_to_le16(crc);
  1813. }
  1814. int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
  1815. struct ext4_group_desc *gdp)
  1816. {
  1817. if (ext4_has_group_desc_csum(sb) &&
  1818. (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
  1819. block_group, gdp)))
  1820. return 0;
  1821. return 1;
  1822. }
  1823. void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
  1824. struct ext4_group_desc *gdp)
  1825. {
  1826. if (!ext4_has_group_desc_csum(sb))
  1827. return;
  1828. gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
  1829. }
  1830. /* Called at mount-time, super-block is locked */
  1831. static int ext4_check_descriptors(struct super_block *sb,
  1832. ext4_group_t *first_not_zeroed)
  1833. {
  1834. struct ext4_sb_info *sbi = EXT4_SB(sb);
  1835. ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
  1836. ext4_fsblk_t last_block;
  1837. ext4_fsblk_t block_bitmap;
  1838. ext4_fsblk_t inode_bitmap;
  1839. ext4_fsblk_t inode_table;
  1840. int flexbg_flag = 0;
  1841. ext4_group_t i, grp = sbi->s_groups_count;
  1842. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
  1843. flexbg_flag = 1;
  1844. ext4_debug("Checking group descriptors");
  1845. for (i = 0; i < sbi->s_groups_count; i++) {
  1846. struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
  1847. if (i == sbi->s_groups_count - 1 || flexbg_flag)
  1848. last_block = ext4_blocks_count(sbi->s_es) - 1;
  1849. else
  1850. last_block = first_block +
  1851. (EXT4_BLOCKS_PER_GROUP(sb) - 1);
  1852. if ((grp == sbi->s_groups_count) &&
  1853. !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
  1854. grp = i;
  1855. block_bitmap = ext4_block_bitmap(sb, gdp);
  1856. if (block_bitmap < first_block || block_bitmap > last_block) {
  1857. ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
  1858. "Block bitmap for group %u not in group "
  1859. "(block %llu)!", i, block_bitmap);
  1860. return 0;
  1861. }
  1862. inode_bitmap = ext4_inode_bitmap(sb, gdp);
  1863. if (inode_bitmap < first_block || inode_bitmap > last_block) {
  1864. ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
  1865. "Inode bitmap for group %u not in group "
  1866. "(block %llu)!", i, inode_bitmap);
  1867. return 0;
  1868. }
  1869. inode_table = ext4_inode_table(sb, gdp);
  1870. if (inode_table < first_block ||
  1871. inode_table + sbi->s_itb_per_group - 1 > last_block) {
  1872. ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
  1873. "Inode table for group %u not in group "
  1874. "(block %llu)!", i, inode_table);
  1875. return 0;
  1876. }
  1877. ext4_lock_group(sb, i);
  1878. if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
  1879. ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
  1880. "Checksum for group %u failed (%u!=%u)",
  1881. i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
  1882. gdp)), le16_to_cpu(gdp->bg_checksum));
  1883. if (!(sb->s_flags & MS_RDONLY)) {
  1884. ext4_unlock_group(sb, i);
  1885. return 0;
  1886. }
  1887. }
  1888. ext4_unlock_group(sb, i);
  1889. if (!flexbg_flag)
  1890. first_block += EXT4_BLOCKS_PER_GROUP(sb);
  1891. }
  1892. if (NULL != first_not_zeroed)
  1893. *first_not_zeroed = grp;
  1894. ext4_free_blocks_count_set(sbi->s_es,
  1895. EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
  1896. sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
  1897. return 1;
  1898. }
  1899. /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
  1900. * the superblock) which were deleted from all directories, but held open by
  1901. * a process at the time of a crash. We walk the list and try to delete these
  1902. * inodes at recovery time (only with a read-write filesystem).
  1903. *
  1904. * In order to keep the orphan inode chain consistent during traversal (in
  1905. * case of crash during recovery), we link each inode into the superblock
  1906. * orphan list_head and handle it the same way as an inode deletion during
  1907. * normal operation (which journals the operations for us).
  1908. *
  1909. * We only do an iget() and an iput() on each inode, which is very safe if we
  1910. * accidentally point at an in-use or already deleted inode. The worst that
  1911. * can happen in this case is that we get a "bit already cleared" message from
  1912. * ext4_free_inode(). The only reason we would point at a wrong inode is if
  1913. * e2fsck was run on this filesystem, and it must have already done the orphan
  1914. * inode cleanup for us, so we can safely abort without any further action.
  1915. */
  1916. static void ext4_orphan_cleanup(struct super_block *sb,
  1917. struct ext4_super_block *es)
  1918. {
  1919. unsigned int s_flags = sb->s_flags;
  1920. int nr_orphans = 0, nr_truncates = 0;
  1921. #ifdef CONFIG_QUOTA
  1922. int i;
  1923. #endif
  1924. if (!es->s_last_orphan) {
  1925. jbd_debug(4, "no orphan inodes to clean up\n");
  1926. return;
  1927. }
  1928. if (bdev_read_only(sb->s_bdev)) {
  1929. ext4_msg(sb, KERN_ERR, "write access "
  1930. "unavailable, skipping orphan cleanup");
  1931. return;
  1932. }
  1933. /* Check if feature set would not allow a r/w mount */
  1934. if (!ext4_feature_set_ok(sb, 0)) {
  1935. ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
  1936. "unknown ROCOMPAT features");
  1937. return;
  1938. }
  1939. if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
  1940. if (es->s_last_orphan)
  1941. jbd_debug(1, "Errors on filesystem, "
  1942. "clearing orphan list.\n");
  1943. es->s_last_orphan = 0;
  1944. jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
  1945. return;
  1946. }
  1947. if (s_flags & MS_RDONLY) {
  1948. ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
  1949. sb->s_flags &= ~MS_RDONLY;
  1950. }
  1951. #ifdef CONFIG_QUOTA
  1952. /* Needed for iput() to work correctly and not trash data */
  1953. sb->s_flags |= MS_ACTIVE;
  1954. /* Turn on quotas so that they are updated correctly */
  1955. for (i = 0; i < MAXQUOTAS; i++) {
  1956. if (EXT4_SB(sb)->s_qf_names[i]) {
  1957. int ret = ext4_quota_on_mount(sb, i);
  1958. if (ret < 0)
  1959. ext4_msg(sb, KERN_ERR,
  1960. "Cannot turn on journaled "
  1961. "quota: error %d", ret);
  1962. }
  1963. }
  1964. #endif
  1965. while (es->s_last_orphan) {
  1966. struct inode *inode;
  1967. inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
  1968. if (IS_ERR(inode)) {
  1969. es->s_last_orphan = 0;
  1970. break;
  1971. }
  1972. list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
  1973. dquot_initialize(inode);
  1974. if (inode->i_nlink) {
  1975. ext4_msg(sb, KERN_DEBUG,
  1976. "%s: truncating inode %lu to %lld bytes",
  1977. __func__, inode->i_ino, inode->i_size);
  1978. jbd_debug(2, "truncating inode %lu to %lld bytes\n",
  1979. inode->i_ino, inode->i_size);
  1980. ext4_truncate(inode);
  1981. nr_truncates++;
  1982. } else {
  1983. ext4_msg(sb, KERN_DEBUG,
  1984. "%s: deleting unreferenced inode %lu",
  1985. __func__, inode->i_ino);
  1986. jbd_debug(2, "deleting unreferenced inode %lu\n",
  1987. inode->i_ino);
  1988. nr_orphans++;
  1989. }
  1990. iput(inode); /* The delete magic happens here! */
  1991. }
  1992. #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
  1993. if (nr_orphans)
  1994. ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
  1995. PLURAL(nr_orphans));
  1996. if (nr_truncates)
  1997. ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
  1998. PLURAL(nr_truncates));
  1999. #ifdef CONFIG_QUOTA
  2000. /* Turn quotas off */
  2001. for (i = 0; i < MAXQUOTAS; i++) {
  2002. if (sb_dqopt(sb)->files[i])
  2003. dquot_quota_off(sb, i);
  2004. }
  2005. #endif
  2006. sb->s_flags = s_flags; /* Restore MS_RDONLY status */
  2007. }
  2008. /*
  2009. * Maximal extent format file size.
  2010. * Resulting logical blkno at s_maxbytes must fit in our on-disk
  2011. * extent format containers, within a sector_t, and within i_blocks
  2012. * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
  2013. * so that won't be a limiting factor.
  2014. *
  2015. * However there is other limiting factor. We do store extents in the form
  2016. * of starting block and length, hence the resulting length of the extent
  2017. * covering maximum file size must fit into on-disk format containers as
  2018. * well. Given that length is always by 1 unit bigger than max unit (because
  2019. * we count 0 as well) we have to lower the s_maxbytes by one fs block.
  2020. *
  2021. * Note, this does *not* consider any metadata overhead for vfs i_blocks.
  2022. */
  2023. static loff_t ext4_max_size(int blkbits, int has_huge_files)
  2024. {
  2025. loff_t res;
  2026. loff_t upper_limit = MAX_LFS_FILESIZE;
  2027. /* small i_blocks in vfs inode? */
  2028. if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
  2029. /*
  2030. * CONFIG_LBDAF is not enabled implies the inode
  2031. * i_block represent total blocks in 512 bytes
  2032. * 32 == size of vfs inode i_blocks * 8
  2033. */
  2034. upper_limit = (1LL << 32) - 1;
  2035. /* total blocks in file system block size */
  2036. upper_limit >>= (blkbits - 9);
  2037. upper_limit <<= blkbits;
  2038. }
  2039. /*
  2040. * 32-bit extent-start container, ee_block. We lower the maxbytes
  2041. * by one fs block, so ee_len can cover the extent of maximum file
  2042. * size
  2043. */
  2044. res = (1LL << 32) - 1;
  2045. res <<= blkbits;
  2046. /* Sanity check against vm- & vfs- imposed limits */
  2047. if (res > upper_limit)
  2048. res = upper_limit;
  2049. return res;
  2050. }
  2051. /*
  2052. * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
  2053. * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
  2054. * We need to be 1 filesystem block less than the 2^48 sector limit.
  2055. */
  2056. static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
  2057. {
  2058. loff_t res = EXT4_NDIR_BLOCKS;
  2059. int meta_blocks;
  2060. loff_t upper_limit;
  2061. /* This is calculated to be the largest file size for a dense, block
  2062. * mapped file such that the file's total number of 512-byte sectors,
  2063. * including data and all indirect blocks, does not exceed (2^48 - 1).
  2064. *
  2065. * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
  2066. * number of 512-byte sectors of the file.
  2067. */
  2068. if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
  2069. /*
  2070. * !has_huge_files or CONFIG_LBDAF not enabled implies that
  2071. * the inode i_block field represents total file blocks in
  2072. * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
  2073. */
  2074. upper_limit = (1LL << 32) - 1;
  2075. /* total blocks in file system block size */
  2076. upper_limit >>= (bits - 9);
  2077. } else {
  2078. /*
  2079. * We use 48 bit ext4_inode i_blocks
  2080. * With EXT4_HUGE_FILE_FL set the i_blocks
  2081. * represent total number of blocks in
  2082. * file system block size
  2083. */
  2084. upper_limit = (1LL << 48) - 1;
  2085. }
  2086. /* indirect blocks */
  2087. meta_blocks = 1;
  2088. /* double indirect blocks */
  2089. meta_blocks += 1 + (1LL << (bits-2));
  2090. /* tripple indirect blocks */
  2091. meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
  2092. upper_limit -= meta_blocks;
  2093. upper_limit <<= bits;
  2094. res += 1LL << (bits-2);
  2095. res += 1LL << (2*(bits-2));
  2096. res += 1LL << (3*(bits-2));
  2097. res <<= bits;
  2098. if (res > upper_limit)
  2099. res = upper_limit;
  2100. if (res > MAX_LFS_FILESIZE)
  2101. res = MAX_LFS_FILESIZE;
  2102. return res;
  2103. }
  2104. static ext4_fsblk_t descriptor_loc(struct super_block *sb,
  2105. ext4_fsblk_t logical_sb_block, int nr)
  2106. {
  2107. struct ext4_sb_info *sbi = EXT4_SB(sb);
  2108. ext4_group_t bg, first_meta_bg;
  2109. int has_super = 0;
  2110. first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
  2111. if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
  2112. nr < first_meta_bg)
  2113. return logical_sb_block + nr + 1;
  2114. bg = sbi->s_desc_per_block * nr;
  2115. if (ext4_bg_has_super(sb, bg))
  2116. has_super = 1;
  2117. return (has_super + ext4_group_first_block_no(sb, bg));
  2118. }
  2119. /**
  2120. * ext4_get_stripe_size: Get the stripe size.
  2121. * @sbi: In memory super block info
  2122. *
  2123. * If we have specified it via mount option, then
  2124. * use the mount option value. If the value specified at mount time is
  2125. * greater than the blocks per group use the super block value.
  2126. * If the super block value is greater than blocks per group return 0.
  2127. * Allocator needs it be less than blocks per group.
  2128. *
  2129. */
  2130. static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
  2131. {
  2132. unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
  2133. unsigned long stripe_width =
  2134. le32_to_cpu(sbi->s_es->s_raid_stripe_width);
  2135. int ret;
  2136. if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
  2137. ret = sbi->s_stripe;
  2138. else if (stripe_width <= sbi->s_blocks_per_group)
  2139. ret = stripe_width;
  2140. else if (stride <= sbi->s_blocks_per_group)
  2141. ret = stride;
  2142. else
  2143. ret = 0;
  2144. /*
  2145. * If the stripe width is 1, this makes no sense and
  2146. * we set it to 0 to turn off stripe handling code.
  2147. */
  2148. if (ret <= 1)
  2149. ret = 0;
  2150. return ret;
  2151. }
  2152. /* sysfs supprt */
  2153. struct ext4_attr {
  2154. struct attribute attr;
  2155. ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
  2156. ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
  2157. const char *, size_t);
  2158. int offset;
  2159. };
  2160. static int parse_strtoul(const char *buf,
  2161. unsigned long max, unsigned long *value)
  2162. {
  2163. char *endp;
  2164. *value = simple_strtoul(skip_spaces(buf), &endp, 0);
  2165. endp = skip_spaces(endp);
  2166. if (*endp || *value > max)
  2167. return -EINVAL;
  2168. return 0;
  2169. }
  2170. static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
  2171. struct ext4_sb_info *sbi,
  2172. char *buf)
  2173. {
  2174. return snprintf(buf, PAGE_SIZE, "%llu\n",
  2175. (s64) EXT4_C2B(sbi,
  2176. percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
  2177. }
  2178. static ssize_t session_write_kbytes_show(struct ext4_attr *a,
  2179. struct ext4_sb_info *sbi, char *buf)
  2180. {
  2181. struct super_block *sb = sbi->s_buddy_cache->i_sb;
  2182. if (!sb->s_bdev->bd_part)
  2183. return snprintf(buf, PAGE_SIZE, "0\n");
  2184. return snprintf(buf, PAGE_SIZE, "%lu\n",
  2185. (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
  2186. sbi->s_sectors_written_start) >> 1);
  2187. }
  2188. static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
  2189. struct ext4_sb_info *sbi, char *buf)
  2190. {
  2191. struct super_block *sb = sbi->s_buddy_cache->i_sb;
  2192. if (!sb->s_bdev->bd_part)
  2193. return snprintf(buf, PAGE_SIZE, "0\n");
  2194. return snprintf(buf, PAGE_SIZE, "%llu\n",
  2195. (unsigned long long)(sbi->s_kbytes_written +
  2196. ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
  2197. EXT4_SB(sb)->s_sectors_written_start) >> 1)));
  2198. }
  2199. static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
  2200. struct ext4_sb_info *sbi,
  2201. const char *buf, size_t count)
  2202. {
  2203. unsigned long t;
  2204. if (parse_strtoul(buf, 0x40000000, &t))
  2205. return -EINVAL;
  2206. if (t && !is_power_of_2(t))
  2207. return -EINVAL;
  2208. sbi->s_inode_readahead_blks = t;
  2209. return count;
  2210. }
  2211. static ssize_t sbi_ui_show(struct ext4_attr *a,
  2212. struct ext4_sb_info *sbi, char *buf)
  2213. {
  2214. unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
  2215. return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
  2216. }
  2217. static ssize_t sbi_ui_store(struct ext4_attr *a,
  2218. struct ext4_sb_info *sbi,
  2219. const char *buf, size_t count)
  2220. {
  2221. unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
  2222. unsigned long t;
  2223. if (parse_strtoul(buf, 0xffffffff, &t))
  2224. return -EINVAL;
  2225. *ui = t;
  2226. return count;
  2227. }
  2228. static ssize_t trigger_test_error(struct ext4_attr *a,
  2229. struct ext4_sb_info *sbi,
  2230. const char *buf, size_t count)
  2231. {
  2232. int len = count;
  2233. if (!capable(CAP_SYS_ADMIN))
  2234. return -EPERM;
  2235. if (len && buf[len-1] == '\n')
  2236. len--;
  2237. if (len)
  2238. ext4_error(sbi->s_sb, "%.*s", len, buf);
  2239. return count;
  2240. }
  2241. #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
  2242. static struct ext4_attr ext4_attr_##_name = { \
  2243. .attr = {.name = __stringify(_name), .mode = _mode }, \
  2244. .show = _show, \
  2245. .store = _store, \
  2246. .offset = offsetof(struct ext4_sb_info, _elname), \
  2247. }
  2248. #define EXT4_ATTR(name, mode, show, store) \
  2249. static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
  2250. #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
  2251. #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
  2252. #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
  2253. #define EXT4_RW_ATTR_SBI_UI(name, elname) \
  2254. EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
  2255. #define ATTR_LIST(name) &ext4_attr_##name.attr
  2256. EXT4_RO_ATTR(delayed_allocation_blocks);
  2257. EXT4_RO_ATTR(session_write_kbytes);
  2258. EXT4_RO_ATTR(lifetime_write_kbytes);
  2259. EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
  2260. inode_readahead_blks_store, s_inode_readahead_blks);
  2261. EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
  2262. EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
  2263. EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
  2264. EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
  2265. EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
  2266. EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
  2267. EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
  2268. EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
  2269. EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
  2270. EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
  2271. static struct attribute *ext4_attrs[] = {
  2272. ATTR_LIST(delayed_allocation_blocks),
  2273. ATTR_LIST(session_write_kbytes),
  2274. ATTR_LIST(lifetime_write_kbytes),
  2275. ATTR_LIST(inode_readahead_blks),
  2276. ATTR_LIST(inode_goal),
  2277. ATTR_LIST(mb_stats),
  2278. ATTR_LIST(mb_max_to_scan),
  2279. ATTR_LIST(mb_min_to_scan),
  2280. ATTR_LIST(mb_order2_req),
  2281. ATTR_LIST(mb_stream_req),
  2282. ATTR_LIST(mb_group_prealloc),
  2283. ATTR_LIST(max_writeback_mb_bump),
  2284. ATTR_LIST(extent_max_zeroout_kb),
  2285. ATTR_LIST(trigger_fs_error),
  2286. NULL,
  2287. };
  2288. /* Features this copy of ext4 supports */
  2289. EXT4_INFO_ATTR(lazy_itable_init);
  2290. EXT4_INFO_ATTR(batched_discard);
  2291. EXT4_INFO_ATTR(meta_bg_resize);
  2292. static struct attribute *ext4_feat_attrs[] = {
  2293. ATTR_LIST(lazy_itable_init),
  2294. ATTR_LIST(batched_discard),
  2295. ATTR_LIST(meta_bg_resize),
  2296. NULL,
  2297. };
  2298. static ssize_t ext4_attr_show(struct kobject *kobj,
  2299. struct attribute *attr, char *buf)
  2300. {
  2301. struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
  2302. s_kobj);
  2303. struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
  2304. return a->show ? a->show(a, sbi, buf) : 0;
  2305. }
  2306. static ssize_t ext4_attr_store(struct kobject *kobj,
  2307. struct attribute *attr,
  2308. const char *buf, size_t len)
  2309. {
  2310. struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
  2311. s_kobj);
  2312. struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
  2313. return a->store ? a->store(a, sbi, buf, len) : 0;
  2314. }
  2315. static void ext4_sb_release(struct kobject *kobj)
  2316. {
  2317. struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
  2318. s_kobj);
  2319. complete(&sbi->s_kobj_unregister);
  2320. }
  2321. static const struct sysfs_ops ext4_attr_ops = {
  2322. .show = ext4_attr_show,
  2323. .store = ext4_attr_store,
  2324. };
  2325. static struct kobj_type ext4_ktype = {
  2326. .default_attrs = ext4_attrs,
  2327. .sysfs_ops = &ext4_attr_ops,
  2328. .release = ext4_sb_release,
  2329. };
  2330. static void ext4_feat_release(struct kobject *kobj)
  2331. {
  2332. complete(&ext4_feat->f_kobj_unregister);
  2333. }
  2334. static struct kobj_type ext4_feat_ktype = {
  2335. .default_attrs = ext4_feat_attrs,
  2336. .sysfs_ops = &ext4_attr_ops,
  2337. .release = ext4_feat_release,
  2338. };
  2339. /*
  2340. * Check whether this filesystem can be mounted based on
  2341. * the features present and the RDONLY/RDWR mount requested.
  2342. * Returns 1 if this filesystem can be mounted as requested,
  2343. * 0 if it cannot be.
  2344. */
  2345. static int ext4_feature_set_ok(struct super_block *sb, int readonly)
  2346. {
  2347. if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
  2348. ext4_msg(sb, KERN_ERR,
  2349. "Couldn't mount because of "
  2350. "unsupported optional features (%x)",
  2351. (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
  2352. ~EXT4_FEATURE_INCOMPAT_SUPP));
  2353. return 0;
  2354. }
  2355. if (readonly)
  2356. return 1;
  2357. /* Check that feature set is OK for a read-write mount */
  2358. if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
  2359. ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
  2360. "unsupported optional features (%x)",
  2361. (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
  2362. ~EXT4_FEATURE_RO_COMPAT_SUPP));
  2363. return 0;
  2364. }
  2365. /*
  2366. * Large file size enabled file system can only be mounted
  2367. * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
  2368. */
  2369. if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
  2370. if (sizeof(blkcnt_t) < sizeof(u64)) {
  2371. ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
  2372. "cannot be mounted RDWR without "
  2373. "CONFIG_LBDAF");
  2374. return 0;
  2375. }
  2376. }
  2377. if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
  2378. !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
  2379. ext4_msg(sb, KERN_ERR,
  2380. "Can't support bigalloc feature without "
  2381. "extents feature\n");
  2382. return 0;
  2383. }
  2384. #ifndef CONFIG_QUOTA
  2385. if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
  2386. !readonly) {
  2387. ext4_msg(sb, KERN_ERR,
  2388. "Filesystem with quota feature cannot be mounted RDWR "
  2389. "without CONFIG_QUOTA");
  2390. return 0;
  2391. }
  2392. #endif /* CONFIG_QUOTA */
  2393. return 1;
  2394. }
  2395. /*
  2396. * This function is called once a day if we have errors logged
  2397. * on the file system
  2398. */
  2399. static void print_daily_error_info(unsigned long arg)
  2400. {
  2401. struct super_block *sb = (struct super_block *) arg;
  2402. struct ext4_sb_info *sbi;
  2403. struct ext4_super_block *es;
  2404. sbi = EXT4_SB(sb);
  2405. es = sbi->s_es;
  2406. if (es->s_error_count)
  2407. ext4_msg(sb, KERN_NOTICE, "error count: %u",
  2408. le32_to_cpu(es->s_error_count));
  2409. if (es->s_first_error_time) {
  2410. printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
  2411. sb->s_id, le32_to_cpu(es->s_first_error_time),
  2412. (int) sizeof(es->s_first_error_func),
  2413. es->s_first_error_func,
  2414. le32_to_cpu(es->s_first_error_line));
  2415. if (es->s_first_error_ino)
  2416. printk(": inode %u",
  2417. le32_to_cpu(es->s_first_error_ino));
  2418. if (es->s_first_error_block)
  2419. printk(": block %llu", (unsigned long long)
  2420. le64_to_cpu(es->s_first_error_block));
  2421. printk("\n");
  2422. }
  2423. if (es->s_last_error_time) {
  2424. printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
  2425. sb->s_id, le32_to_cpu(es->s_last_error_time),
  2426. (int) sizeof(es->s_last_error_func),
  2427. es->s_last_error_func,
  2428. le32_to_cpu(es->s_last_error_line));
  2429. if (es->s_last_error_ino)
  2430. printk(": inode %u",
  2431. le32_to_cpu(es->s_last_error_ino));
  2432. if (es->s_last_error_block)
  2433. printk(": block %llu", (unsigned long long)
  2434. le64_to_cpu(es->s_last_error_block));
  2435. printk("\n");
  2436. }
  2437. mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
  2438. }
  2439. /* Find next suitable group and run ext4_init_inode_table */
  2440. static int ext4_run_li_request(struct ext4_li_request *elr)
  2441. {
  2442. struct ext4_group_desc *gdp = NULL;
  2443. ext4_group_t group, ngroups;
  2444. struct super_block *sb;
  2445. unsigned long timeout = 0;
  2446. int ret = 0;
  2447. sb = elr->lr_super;
  2448. ngroups = EXT4_SB(sb)->s_groups_count;
  2449. sb_start_write(sb);
  2450. for (group = elr->lr_next_group; group < ngroups; group++) {
  2451. gdp = ext4_get_group_desc(sb, group, NULL);
  2452. if (!gdp) {
  2453. ret = 1;
  2454. break;
  2455. }
  2456. if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
  2457. break;
  2458. }
  2459. if (group == ngroups)
  2460. ret = 1;
  2461. if (!ret) {
  2462. timeout = jiffies;
  2463. ret = ext4_init_inode_table(sb, group,
  2464. elr->lr_timeout ? 0 : 1);
  2465. if (elr->lr_timeout == 0) {
  2466. timeout = (jiffies - timeout) *
  2467. elr->lr_sbi->s_li_wait_mult;
  2468. elr->lr_timeout = timeout;
  2469. }
  2470. elr->lr_next_sched = jiffies + elr->lr_timeout;
  2471. elr->lr_next_group = group + 1;
  2472. }
  2473. sb_end_write(sb);
  2474. return ret;
  2475. }
  2476. /*
  2477. * Remove lr_request from the list_request and free the
  2478. * request structure. Should be called with li_list_mtx held
  2479. */
  2480. static void ext4_remove_li_request(struct ext4_li_request *elr)
  2481. {
  2482. struct ext4_sb_info *sbi;
  2483. if (!elr)
  2484. return;
  2485. sbi = elr->lr_sbi;
  2486. list_del(&elr->lr_request);
  2487. sbi->s_li_request = NULL;
  2488. kfree(elr);
  2489. }
  2490. static void ext4_unregister_li_request(struct super_block *sb)
  2491. {
  2492. mutex_lock(&ext4_li_mtx);
  2493. if (!ext4_li_info) {
  2494. mutex_unlock(&ext4_li_mtx);
  2495. return;
  2496. }
  2497. mutex_lock(&ext4_li_info->li_list_mtx);
  2498. ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
  2499. mutex_unlock(&ext4_li_info->li_list_mtx);
  2500. mutex_unlock(&ext4_li_mtx);
  2501. }
  2502. static struct task_struct *ext4_lazyinit_task;
  2503. /*
  2504. * This is the function where ext4lazyinit thread lives. It walks
  2505. * through the request list searching for next scheduled filesystem.
  2506. * When such a fs is found, run the lazy initialization request
  2507. * (ext4_rn_li_request) and keep track of the time spend in this
  2508. * function. Based on that time we compute next schedule time of
  2509. * the request. When walking through the list is complete, compute
  2510. * next waking time and put itself into sleep.
  2511. */
  2512. static int ext4_lazyinit_thread(void *arg)
  2513. {
  2514. struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
  2515. struct list_head *pos, *n;
  2516. struct ext4_li_request *elr;
  2517. unsigned long next_wakeup, cur;
  2518. BUG_ON(NULL == eli);
  2519. cont_thread:
  2520. while (true) {
  2521. next_wakeup = MAX_JIFFY_OFFSET;
  2522. mutex_lock(&eli->li_list_mtx);
  2523. if (list_empty(&eli->li_request_list)) {
  2524. mutex_unlock(&eli->li_list_mtx);
  2525. goto exit_thread;
  2526. }
  2527. list_for_each_safe(pos, n, &eli->li_request_list) {
  2528. elr = list_entry(pos, struct ext4_li_request,
  2529. lr_request);
  2530. if (time_after_eq(jiffies, elr->lr_next_sched)) {
  2531. if (ext4_run_li_request(elr) != 0) {
  2532. /* error, remove the lazy_init job */
  2533. ext4_remove_li_request(elr);
  2534. continue;
  2535. }
  2536. }
  2537. if (time_before(elr->lr_next_sched, next_wakeup))
  2538. next_wakeup = elr->lr_next_sched;
  2539. }
  2540. mutex_unlock(&eli->li_list_mtx);
  2541. try_to_freeze();
  2542. cur = jiffies;
  2543. if ((time_after_eq(cur, next_wakeup)) ||
  2544. (MAX_JIFFY_OFFSET == next_wakeup)) {
  2545. cond_resched();
  2546. continue;
  2547. }
  2548. schedule_timeout_interruptible(next_wakeup - cur);
  2549. if (kthread_should_stop()) {
  2550. ext4_clear_request_list();
  2551. goto exit_thread;
  2552. }
  2553. }
  2554. exit_thread:
  2555. /*
  2556. * It looks like the request list is empty, but we need
  2557. * to check it under the li_list_mtx lock, to prevent any
  2558. * additions into it, and of course we should lock ext4_li_mtx
  2559. * to atomically free the list and ext4_li_info, because at
  2560. * this point another ext4 filesystem could be registering
  2561. * new one.
  2562. */
  2563. mutex_lock(&ext4_li_mtx);
  2564. mutex_lock(&eli->li_list_mtx);
  2565. if (!list_empty(&eli->li_request_list)) {
  2566. mutex_unlock(&eli->li_list_mtx);
  2567. mutex_unlock(&ext4_li_mtx);
  2568. goto cont_thread;
  2569. }
  2570. mutex_unlock(&eli->li_list_mtx);
  2571. kfree(ext4_li_info);
  2572. ext4_li_info = NULL;
  2573. mutex_unlock(&ext4_li_mtx);
  2574. return 0;
  2575. }
  2576. static void ext4_clear_request_list(void)
  2577. {
  2578. struct list_head *pos, *n;
  2579. struct ext4_li_request *elr;
  2580. mutex_lock(&ext4_li_info->li_list_mtx);
  2581. list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
  2582. elr = list_entry(pos, struct ext4_li_request,
  2583. lr_request);
  2584. ext4_remove_li_request(elr);
  2585. }
  2586. mutex_unlock(&ext4_li_info->li_list_mtx);
  2587. }
  2588. static int ext4_run_lazyinit_thread(void)
  2589. {
  2590. ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
  2591. ext4_li_info, "ext4lazyinit");
  2592. if (IS_ERR(ext4_lazyinit_task)) {
  2593. int err = PTR_ERR(ext4_lazyinit_task);
  2594. ext4_clear_request_list();
  2595. kfree(ext4_li_info);
  2596. ext4_li_info = NULL;
  2597. printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
  2598. "initialization thread\n",
  2599. err);
  2600. return err;
  2601. }
  2602. ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
  2603. return 0;
  2604. }
  2605. /*
  2606. * Check whether it make sense to run itable init. thread or not.
  2607. * If there is at least one uninitialized inode table, return
  2608. * corresponding group number, else the loop goes through all
  2609. * groups and return total number of groups.
  2610. */
  2611. static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
  2612. {
  2613. ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
  2614. struct ext4_group_desc *gdp = NULL;
  2615. for (group = 0; group < ngroups; group++) {
  2616. gdp = ext4_get_group_desc(sb, group, NULL);
  2617. if (!gdp)
  2618. continue;
  2619. if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
  2620. break;
  2621. }
  2622. return group;
  2623. }
  2624. static int ext4_li_info_new(void)
  2625. {
  2626. struct ext4_lazy_init *eli = NULL;
  2627. eli = kzalloc(sizeof(*eli), GFP_KERNEL);
  2628. if (!eli)
  2629. return -ENOMEM;
  2630. INIT_LIST_HEAD(&eli->li_request_list);
  2631. mutex_init(&eli->li_list_mtx);
  2632. eli->li_state |= EXT4_LAZYINIT_QUIT;
  2633. ext4_li_info = eli;
  2634. return 0;
  2635. }
  2636. static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
  2637. ext4_group_t start)
  2638. {
  2639. struct ext4_sb_info *sbi = EXT4_SB(sb);
  2640. struct ext4_li_request *elr;
  2641. unsigned long rnd;
  2642. elr = kzalloc(sizeof(*elr), GFP_KERNEL);
  2643. if (!elr)
  2644. return NULL;
  2645. elr->lr_super = sb;
  2646. elr->lr_sbi = sbi;
  2647. elr->lr_next_group = start;
  2648. /*
  2649. * Randomize first schedule time of the request to
  2650. * spread the inode table initialization requests
  2651. * better.
  2652. */
  2653. get_random_bytes(&rnd, sizeof(rnd));
  2654. elr->lr_next_sched = jiffies + (unsigned long)rnd %
  2655. (EXT4_DEF_LI_MAX_START_DELAY * HZ);
  2656. return elr;
  2657. }
  2658. static int ext4_register_li_request(struct super_block *sb,
  2659. ext4_group_t first_not_zeroed)
  2660. {
  2661. struct ext4_sb_info *sbi = EXT4_SB(sb);
  2662. struct ext4_li_request *elr;
  2663. ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
  2664. int ret = 0;
  2665. if (sbi->s_li_request != NULL) {
  2666. /*
  2667. * Reset timeout so it can be computed again, because
  2668. * s_li_wait_mult might have changed.
  2669. */
  2670. sbi->s_li_request->lr_timeout = 0;
  2671. return 0;
  2672. }
  2673. if (first_not_zeroed == ngroups ||
  2674. (sb->s_flags & MS_RDONLY) ||
  2675. !test_opt(sb, INIT_INODE_TABLE))
  2676. return 0;
  2677. elr = ext4_li_request_new(sb, first_not_zeroed);
  2678. if (!elr)
  2679. return -ENOMEM;
  2680. mutex_lock(&ext4_li_mtx);
  2681. if (NULL == ext4_li_info) {
  2682. ret = ext4_li_info_new();
  2683. if (ret)
  2684. goto out;
  2685. }
  2686. mutex_lock(&ext4_li_info->li_list_mtx);
  2687. list_add(&elr->lr_request, &ext4_li_info->li_request_list);
  2688. mutex_unlock(&ext4_li_info->li_list_mtx);
  2689. sbi->s_li_request = elr;
  2690. /*
  2691. * set elr to NULL here since it has been inserted to
  2692. * the request_list and the removal and free of it is
  2693. * handled by ext4_clear_request_list from now on.
  2694. */
  2695. elr = NULL;
  2696. if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
  2697. ret = ext4_run_lazyinit_thread();
  2698. if (ret)
  2699. goto out;
  2700. }
  2701. out:
  2702. mutex_unlock(&ext4_li_mtx);
  2703. if (ret)
  2704. kfree(elr);
  2705. return ret;
  2706. }
  2707. /*
  2708. * We do not need to lock anything since this is called on
  2709. * module unload.
  2710. */
  2711. static void ext4_destroy_lazyinit_thread(void)
  2712. {
  2713. /*
  2714. * If thread exited earlier
  2715. * there's nothing to be done.
  2716. */
  2717. if (!ext4_li_info || !ext4_lazyinit_task)
  2718. return;
  2719. kthread_stop(ext4_lazyinit_task);
  2720. }
  2721. static int set_journal_csum_feature_set(struct super_block *sb)
  2722. {
  2723. int ret = 1;
  2724. int compat, incompat;
  2725. struct ext4_sb_info *sbi = EXT4_SB(sb);
  2726. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  2727. EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
  2728. /* journal checksum v2 */
  2729. compat = 0;
  2730. incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
  2731. } else {
  2732. /* journal checksum v1 */
  2733. compat = JBD2_FEATURE_COMPAT_CHECKSUM;
  2734. incompat = 0;
  2735. }
  2736. if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
  2737. ret = jbd2_journal_set_features(sbi->s_journal,
  2738. compat, 0,
  2739. JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
  2740. incompat);
  2741. } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
  2742. ret = jbd2_journal_set_features(sbi->s_journal,
  2743. compat, 0,
  2744. incompat);
  2745. jbd2_journal_clear_features(sbi->s_journal, 0, 0,
  2746. JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
  2747. } else {
  2748. jbd2_journal_clear_features(sbi->s_journal,
  2749. JBD2_FEATURE_COMPAT_CHECKSUM, 0,
  2750. JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
  2751. JBD2_FEATURE_INCOMPAT_CSUM_V2);
  2752. }
  2753. return ret;
  2754. }
  2755. /*
  2756. * Note: calculating the overhead so we can be compatible with
  2757. * historical BSD practice is quite difficult in the face of
  2758. * clusters/bigalloc. This is because multiple metadata blocks from
  2759. * different block group can end up in the same allocation cluster.
  2760. * Calculating the exact overhead in the face of clustered allocation
  2761. * requires either O(all block bitmaps) in memory or O(number of block
  2762. * groups**2) in time. We will still calculate the superblock for
  2763. * older file systems --- and if we come across with a bigalloc file
  2764. * system with zero in s_overhead_clusters the estimate will be close to
  2765. * correct especially for very large cluster sizes --- but for newer
  2766. * file systems, it's better to calculate this figure once at mkfs
  2767. * time, and store it in the superblock. If the superblock value is
  2768. * present (even for non-bigalloc file systems), we will use it.
  2769. */
  2770. static int count_overhead(struct super_block *sb, ext4_group_t grp,
  2771. char *buf)
  2772. {
  2773. struct ext4_sb_info *sbi = EXT4_SB(sb);
  2774. struct ext4_group_desc *gdp;
  2775. ext4_fsblk_t first_block, last_block, b;
  2776. ext4_group_t i, ngroups = ext4_get_groups_count(sb);
  2777. int s, j, count = 0;
  2778. if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
  2779. return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
  2780. sbi->s_itb_per_group + 2);
  2781. first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
  2782. (grp * EXT4_BLOCKS_PER_GROUP(sb));
  2783. last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
  2784. for (i = 0; i < ngroups; i++) {
  2785. gdp = ext4_get_group_desc(sb, i, NULL);
  2786. b = ext4_block_bitmap(sb, gdp);
  2787. if (b >= first_block && b <= last_block) {
  2788. ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
  2789. count++;
  2790. }
  2791. b = ext4_inode_bitmap(sb, gdp);
  2792. if (b >= first_block && b <= last_block) {
  2793. ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
  2794. count++;
  2795. }
  2796. b = ext4_inode_table(sb, gdp);
  2797. if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
  2798. for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
  2799. int c = EXT4_B2C(sbi, b - first_block);
  2800. ext4_set_bit(c, buf);
  2801. count++;
  2802. }
  2803. if (i != grp)
  2804. continue;
  2805. s = 0;
  2806. if (ext4_bg_has_super(sb, grp)) {
  2807. ext4_set_bit(s++, buf);
  2808. count++;
  2809. }
  2810. for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
  2811. ext4_set_bit(EXT4_B2C(sbi, s++), buf);
  2812. count++;
  2813. }
  2814. }
  2815. if (!count)
  2816. return 0;
  2817. return EXT4_CLUSTERS_PER_GROUP(sb) -
  2818. ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
  2819. }
  2820. /*
  2821. * Compute the overhead and stash it in sbi->s_overhead
  2822. */
  2823. int ext4_calculate_overhead(struct super_block *sb)
  2824. {
  2825. struct ext4_sb_info *sbi = EXT4_SB(sb);
  2826. struct ext4_super_block *es = sbi->s_es;
  2827. ext4_group_t i, ngroups = ext4_get_groups_count(sb);
  2828. ext4_fsblk_t overhead = 0;
  2829. char *buf = (char *) get_zeroed_page(GFP_KERNEL);
  2830. memset(buf, 0, PAGE_SIZE);
  2831. if (!buf)
  2832. return -ENOMEM;
  2833. /*
  2834. * Compute the overhead (FS structures). This is constant
  2835. * for a given filesystem unless the number of block groups
  2836. * changes so we cache the previous value until it does.
  2837. */
  2838. /*
  2839. * All of the blocks before first_data_block are overhead
  2840. */
  2841. overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
  2842. /*
  2843. * Add the overhead found in each block group
  2844. */
  2845. for (i = 0; i < ngroups; i++) {
  2846. int blks;
  2847. blks = count_overhead(sb, i, buf);
  2848. overhead += blks;
  2849. if (blks)
  2850. memset(buf, 0, PAGE_SIZE);
  2851. cond_resched();
  2852. }
  2853. sbi->s_overhead = overhead;
  2854. smp_wmb();
  2855. free_page((unsigned long) buf);
  2856. return 0;
  2857. }
  2858. static int ext4_fill_super(struct super_block *sb, void *data, int silent)
  2859. {
  2860. char *orig_data = kstrdup(data, GFP_KERNEL);
  2861. struct buffer_head *bh;
  2862. struct ext4_super_block *es = NULL;
  2863. struct ext4_sb_info *sbi;
  2864. ext4_fsblk_t block;
  2865. ext4_fsblk_t sb_block = get_sb_block(&data);
  2866. ext4_fsblk_t logical_sb_block;
  2867. unsigned long offset = 0;
  2868. unsigned long journal_devnum = 0;
  2869. unsigned long def_mount_opts;
  2870. struct inode *root;
  2871. char *cp;
  2872. const char *descr;
  2873. int ret = -ENOMEM;
  2874. int blocksize, clustersize;
  2875. unsigned int db_count;
  2876. unsigned int i;
  2877. int needs_recovery, has_huge_files, has_bigalloc;
  2878. __u64 blocks_count;
  2879. int err;
  2880. unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
  2881. ext4_group_t first_not_zeroed;
  2882. sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
  2883. if (!sbi)
  2884. goto out_free_orig;
  2885. sbi->s_blockgroup_lock =
  2886. kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
  2887. if (!sbi->s_blockgroup_lock) {
  2888. kfree(sbi);
  2889. goto out_free_orig;
  2890. }
  2891. sb->s_fs_info = sbi;
  2892. sbi->s_sb = sb;
  2893. sbi->s_mount_opt = 0;
  2894. sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
  2895. sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
  2896. sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
  2897. sbi->s_sb_block = sb_block;
  2898. if (sb->s_bdev->bd_part)
  2899. sbi->s_sectors_written_start =
  2900. part_stat_read(sb->s_bdev->bd_part, sectors[1]);
  2901. /* Cleanup superblock name */
  2902. for (cp = sb->s_id; (cp = strchr(cp, '/'));)
  2903. *cp = '!';
  2904. ret = -EINVAL;
  2905. blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
  2906. if (!blocksize) {
  2907. ext4_msg(sb, KERN_ERR, "unable to set blocksize");
  2908. goto out_fail;
  2909. }
  2910. /*
  2911. * The ext4 superblock will not be buffer aligned for other than 1kB
  2912. * block sizes. We need to calculate the offset from buffer start.
  2913. */
  2914. if (blocksize != EXT4_MIN_BLOCK_SIZE) {
  2915. logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
  2916. offset = do_div(logical_sb_block, blocksize);
  2917. } else {
  2918. logical_sb_block = sb_block;
  2919. }
  2920. if (!(bh = sb_bread(sb, logical_sb_block))) {
  2921. ext4_msg(sb, KERN_ERR, "unable to read superblock");
  2922. goto out_fail;
  2923. }
  2924. /*
  2925. * Note: s_es must be initialized as soon as possible because
  2926. * some ext4 macro-instructions depend on its value
  2927. */
  2928. es = (struct ext4_super_block *) (bh->b_data + offset);
  2929. sbi->s_es = es;
  2930. sb->s_magic = le16_to_cpu(es->s_magic);
  2931. if (sb->s_magic != EXT4_SUPER_MAGIC)
  2932. goto cantfind_ext4;
  2933. sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
  2934. /* Warn if metadata_csum and gdt_csum are both set. */
  2935. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  2936. EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
  2937. EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
  2938. ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
  2939. "redundant flags; please run fsck.");
  2940. /* Check for a known checksum algorithm */
  2941. if (!ext4_verify_csum_type(sb, es)) {
  2942. ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
  2943. "unknown checksum algorithm.");
  2944. silent = 1;
  2945. goto cantfind_ext4;
  2946. }
  2947. /* Load the checksum driver */
  2948. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  2949. EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
  2950. sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
  2951. if (IS_ERR(sbi->s_chksum_driver)) {
  2952. ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
  2953. ret = PTR_ERR(sbi->s_chksum_driver);
  2954. sbi->s_chksum_driver = NULL;
  2955. goto failed_mount;
  2956. }
  2957. }
  2958. /* Check superblock checksum */
  2959. if (!ext4_superblock_csum_verify(sb, es)) {
  2960. ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
  2961. "invalid superblock checksum. Run e2fsck?");
  2962. silent = 1;
  2963. goto cantfind_ext4;
  2964. }
  2965. /* Precompute checksum seed for all metadata */
  2966. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  2967. EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
  2968. sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
  2969. sizeof(es->s_uuid));
  2970. /* Set defaults before we parse the mount options */
  2971. def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
  2972. set_opt(sb, INIT_INODE_TABLE);
  2973. if (def_mount_opts & EXT4_DEFM_DEBUG)
  2974. set_opt(sb, DEBUG);
  2975. if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
  2976. set_opt(sb, GRPID);
  2977. if (def_mount_opts & EXT4_DEFM_UID16)
  2978. set_opt(sb, NO_UID32);
  2979. /* xattr user namespace & acls are now defaulted on */
  2980. #ifdef CONFIG_EXT4_FS_XATTR
  2981. set_opt(sb, XATTR_USER);
  2982. #endif
  2983. #ifdef CONFIG_EXT4_FS_POSIX_ACL
  2984. set_opt(sb, POSIX_ACL);
  2985. #endif
  2986. set_opt(sb, MBLK_IO_SUBMIT);
  2987. if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
  2988. set_opt(sb, JOURNAL_DATA);
  2989. else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
  2990. set_opt(sb, ORDERED_DATA);
  2991. else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
  2992. set_opt(sb, WRITEBACK_DATA);
  2993. if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
  2994. set_opt(sb, ERRORS_PANIC);
  2995. else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
  2996. set_opt(sb, ERRORS_CONT);
  2997. else
  2998. set_opt(sb, ERRORS_RO);
  2999. if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
  3000. set_opt(sb, BLOCK_VALIDITY);
  3001. if (def_mount_opts & EXT4_DEFM_DISCARD)
  3002. set_opt(sb, DISCARD);
  3003. sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
  3004. sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
  3005. sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
  3006. sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
  3007. sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
  3008. if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
  3009. set_opt(sb, BARRIER);
  3010. /*
  3011. * enable delayed allocation by default
  3012. * Use -o nodelalloc to turn it off
  3013. */
  3014. if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
  3015. ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
  3016. set_opt(sb, DELALLOC);
  3017. /*
  3018. * set default s_li_wait_mult for lazyinit, for the case there is
  3019. * no mount option specified.
  3020. */
  3021. sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
  3022. if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
  3023. &journal_devnum, &journal_ioprio, 0)) {
  3024. ext4_msg(sb, KERN_WARNING,
  3025. "failed to parse options in superblock: %s",
  3026. sbi->s_es->s_mount_opts);
  3027. }
  3028. sbi->s_def_mount_opt = sbi->s_mount_opt;
  3029. if (!parse_options((char *) data, sb, &journal_devnum,
  3030. &journal_ioprio, 0))
  3031. goto failed_mount;
  3032. if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
  3033. printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
  3034. "with data=journal disables delayed "
  3035. "allocation and O_DIRECT support!\n");
  3036. if (test_opt2(sb, EXPLICIT_DELALLOC)) {
  3037. ext4_msg(sb, KERN_ERR, "can't mount with "
  3038. "both data=journal and delalloc");
  3039. goto failed_mount;
  3040. }
  3041. if (test_opt(sb, DIOREAD_NOLOCK)) {
  3042. ext4_msg(sb, KERN_ERR, "can't mount with "
  3043. "both data=journal and delalloc");
  3044. goto failed_mount;
  3045. }
  3046. if (test_opt(sb, DELALLOC))
  3047. clear_opt(sb, DELALLOC);
  3048. }
  3049. blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
  3050. if (test_opt(sb, DIOREAD_NOLOCK)) {
  3051. if (blocksize < PAGE_SIZE) {
  3052. ext4_msg(sb, KERN_ERR, "can't mount with "
  3053. "dioread_nolock if block size != PAGE_SIZE");
  3054. goto failed_mount;
  3055. }
  3056. }
  3057. sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
  3058. (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
  3059. if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
  3060. (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
  3061. EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
  3062. EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
  3063. ext4_msg(sb, KERN_WARNING,
  3064. "feature flags set on rev 0 fs, "
  3065. "running e2fsck is recommended");
  3066. if (IS_EXT2_SB(sb)) {
  3067. if (ext2_feature_set_ok(sb))
  3068. ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
  3069. "using the ext4 subsystem");
  3070. else {
  3071. ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
  3072. "to feature incompatibilities");
  3073. goto failed_mount;
  3074. }
  3075. }
  3076. if (IS_EXT3_SB(sb)) {
  3077. if (ext3_feature_set_ok(sb))
  3078. ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
  3079. "using the ext4 subsystem");
  3080. else {
  3081. ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
  3082. "to feature incompatibilities");
  3083. goto failed_mount;
  3084. }
  3085. }
  3086. /*
  3087. * Check feature flags regardless of the revision level, since we
  3088. * previously didn't change the revision level when setting the flags,
  3089. * so there is a chance incompat flags are set on a rev 0 filesystem.
  3090. */
  3091. if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
  3092. goto failed_mount;
  3093. if (blocksize < EXT4_MIN_BLOCK_SIZE ||
  3094. blocksize > EXT4_MAX_BLOCK_SIZE) {
  3095. ext4_msg(sb, KERN_ERR,
  3096. "Unsupported filesystem blocksize %d", blocksize);
  3097. goto failed_mount;
  3098. }
  3099. if (sb->s_blocksize != blocksize) {
  3100. /* Validate the filesystem blocksize */
  3101. if (!sb_set_blocksize(sb, blocksize)) {
  3102. ext4_msg(sb, KERN_ERR, "bad block size %d",
  3103. blocksize);
  3104. goto failed_mount;
  3105. }
  3106. brelse(bh);
  3107. logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
  3108. offset = do_div(logical_sb_block, blocksize);
  3109. bh = sb_bread(sb, logical_sb_block);
  3110. if (!bh) {
  3111. ext4_msg(sb, KERN_ERR,
  3112. "Can't read superblock on 2nd try");
  3113. goto failed_mount;
  3114. }
  3115. es = (struct ext4_super_block *)(bh->b_data + offset);
  3116. sbi->s_es = es;
  3117. if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
  3118. ext4_msg(sb, KERN_ERR,
  3119. "Magic mismatch, very weird!");
  3120. goto failed_mount;
  3121. }
  3122. }
  3123. has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
  3124. EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
  3125. sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
  3126. has_huge_files);
  3127. sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
  3128. if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
  3129. sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
  3130. sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
  3131. } else {
  3132. sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
  3133. sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
  3134. if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
  3135. (!is_power_of_2(sbi->s_inode_size)) ||
  3136. (sbi->s_inode_size > blocksize)) {
  3137. ext4_msg(sb, KERN_ERR,
  3138. "unsupported inode size: %d",
  3139. sbi->s_inode_size);
  3140. goto failed_mount;
  3141. }
  3142. if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
  3143. sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
  3144. }
  3145. sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
  3146. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
  3147. if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
  3148. sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
  3149. !is_power_of_2(sbi->s_desc_size)) {
  3150. ext4_msg(sb, KERN_ERR,
  3151. "unsupported descriptor size %lu",
  3152. sbi->s_desc_size);
  3153. goto failed_mount;
  3154. }
  3155. } else
  3156. sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
  3157. sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
  3158. sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
  3159. if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
  3160. goto cantfind_ext4;
  3161. sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
  3162. if (sbi->s_inodes_per_block == 0)
  3163. goto cantfind_ext4;
  3164. sbi->s_itb_per_group = sbi->s_inodes_per_group /
  3165. sbi->s_inodes_per_block;
  3166. sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
  3167. sbi->s_sbh = bh;
  3168. sbi->s_mount_state = le16_to_cpu(es->s_state);
  3169. sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
  3170. sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
  3171. for (i = 0; i < 4; i++)
  3172. sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
  3173. sbi->s_def_hash_version = es->s_def_hash_version;
  3174. i = le32_to_cpu(es->s_flags);
  3175. if (i & EXT2_FLAGS_UNSIGNED_HASH)
  3176. sbi->s_hash_unsigned = 3;
  3177. else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
  3178. #ifdef __CHAR_UNSIGNED__
  3179. es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
  3180. sbi->s_hash_unsigned = 3;
  3181. #else
  3182. es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
  3183. #endif
  3184. }
  3185. /* Handle clustersize */
  3186. clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
  3187. has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
  3188. EXT4_FEATURE_RO_COMPAT_BIGALLOC);
  3189. if (has_bigalloc) {
  3190. if (clustersize < blocksize) {
  3191. ext4_msg(sb, KERN_ERR,
  3192. "cluster size (%d) smaller than "
  3193. "block size (%d)", clustersize, blocksize);
  3194. goto failed_mount;
  3195. }
  3196. sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
  3197. le32_to_cpu(es->s_log_block_size);
  3198. sbi->s_clusters_per_group =
  3199. le32_to_cpu(es->s_clusters_per_group);
  3200. if (sbi->s_clusters_per_group > blocksize * 8) {
  3201. ext4_msg(sb, KERN_ERR,
  3202. "#clusters per group too big: %lu",
  3203. sbi->s_clusters_per_group);
  3204. goto failed_mount;
  3205. }
  3206. if (sbi->s_blocks_per_group !=
  3207. (sbi->s_clusters_per_group * (clustersize / blocksize))) {
  3208. ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
  3209. "clusters per group (%lu) inconsistent",
  3210. sbi->s_blocks_per_group,
  3211. sbi->s_clusters_per_group);
  3212. goto failed_mount;
  3213. }
  3214. } else {
  3215. if (clustersize != blocksize) {
  3216. ext4_warning(sb, "fragment/cluster size (%d) != "
  3217. "block size (%d)", clustersize,
  3218. blocksize);
  3219. clustersize = blocksize;
  3220. }
  3221. if (sbi->s_blocks_per_group > blocksize * 8) {
  3222. ext4_msg(sb, KERN_ERR,
  3223. "#blocks per group too big: %lu",
  3224. sbi->s_blocks_per_group);
  3225. goto failed_mount;
  3226. }
  3227. sbi->s_clusters_per_group = sbi->s_blocks_per_group;
  3228. sbi->s_cluster_bits = 0;
  3229. }
  3230. sbi->s_cluster_ratio = clustersize / blocksize;
  3231. if (sbi->s_inodes_per_group > blocksize * 8) {
  3232. ext4_msg(sb, KERN_ERR,
  3233. "#inodes per group too big: %lu",
  3234. sbi->s_inodes_per_group);
  3235. goto failed_mount;
  3236. }
  3237. /*
  3238. * Test whether we have more sectors than will fit in sector_t,
  3239. * and whether the max offset is addressable by the page cache.
  3240. */
  3241. err = generic_check_addressable(sb->s_blocksize_bits,
  3242. ext4_blocks_count(es));
  3243. if (err) {
  3244. ext4_msg(sb, KERN_ERR, "filesystem"
  3245. " too large to mount safely on this system");
  3246. if (sizeof(sector_t) < 8)
  3247. ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
  3248. ret = err;
  3249. goto failed_mount;
  3250. }
  3251. if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
  3252. goto cantfind_ext4;
  3253. /* check blocks count against device size */
  3254. blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
  3255. if (blocks_count && ext4_blocks_count(es) > blocks_count) {
  3256. ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
  3257. "exceeds size of device (%llu blocks)",
  3258. ext4_blocks_count(es), blocks_count);
  3259. goto failed_mount;
  3260. }
  3261. /*
  3262. * It makes no sense for the first data block to be beyond the end
  3263. * of the filesystem.
  3264. */
  3265. if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
  3266. ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
  3267. "block %u is beyond end of filesystem (%llu)",
  3268. le32_to_cpu(es->s_first_data_block),
  3269. ext4_blocks_count(es));
  3270. goto failed_mount;
  3271. }
  3272. blocks_count = (ext4_blocks_count(es) -
  3273. le32_to_cpu(es->s_first_data_block) +
  3274. EXT4_BLOCKS_PER_GROUP(sb) - 1);
  3275. do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
  3276. if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
  3277. ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
  3278. "(block count %llu, first data block %u, "
  3279. "blocks per group %lu)", sbi->s_groups_count,
  3280. ext4_blocks_count(es),
  3281. le32_to_cpu(es->s_first_data_block),
  3282. EXT4_BLOCKS_PER_GROUP(sb));
  3283. goto failed_mount;
  3284. }
  3285. sbi->s_groups_count = blocks_count;
  3286. sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
  3287. (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
  3288. db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
  3289. EXT4_DESC_PER_BLOCK(sb);
  3290. sbi->s_group_desc = ext4_kvmalloc(db_count *
  3291. sizeof(struct buffer_head *),
  3292. GFP_KERNEL);
  3293. if (sbi->s_group_desc == NULL) {
  3294. ext4_msg(sb, KERN_ERR, "not enough memory");
  3295. ret = -ENOMEM;
  3296. goto failed_mount;
  3297. }
  3298. if (ext4_proc_root)
  3299. sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
  3300. if (sbi->s_proc)
  3301. proc_create_data("options", S_IRUGO, sbi->s_proc,
  3302. &ext4_seq_options_fops, sb);
  3303. bgl_lock_init(sbi->s_blockgroup_lock);
  3304. for (i = 0; i < db_count; i++) {
  3305. block = descriptor_loc(sb, logical_sb_block, i);
  3306. sbi->s_group_desc[i] = sb_bread(sb, block);
  3307. if (!sbi->s_group_desc[i]) {
  3308. ext4_msg(sb, KERN_ERR,
  3309. "can't read group descriptor %d", i);
  3310. db_count = i;
  3311. goto failed_mount2;
  3312. }
  3313. }
  3314. if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
  3315. ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
  3316. goto failed_mount2;
  3317. }
  3318. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
  3319. if (!ext4_fill_flex_info(sb)) {
  3320. ext4_msg(sb, KERN_ERR,
  3321. "unable to initialize "
  3322. "flex_bg meta info!");
  3323. goto failed_mount2;
  3324. }
  3325. sbi->s_gdb_count = db_count;
  3326. get_random_bytes(&sbi->s_next_generation, sizeof(u32));
  3327. spin_lock_init(&sbi->s_next_gen_lock);
  3328. init_timer(&sbi->s_err_report);
  3329. sbi->s_err_report.function = print_daily_error_info;
  3330. sbi->s_err_report.data = (unsigned long) sb;
  3331. err = percpu_counter_init(&sbi->s_freeclusters_counter,
  3332. ext4_count_free_clusters(sb));
  3333. if (!err) {
  3334. err = percpu_counter_init(&sbi->s_freeinodes_counter,
  3335. ext4_count_free_inodes(sb));
  3336. }
  3337. if (!err) {
  3338. err = percpu_counter_init(&sbi->s_dirs_counter,
  3339. ext4_count_dirs(sb));
  3340. }
  3341. if (!err) {
  3342. err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
  3343. }
  3344. if (err) {
  3345. ext4_msg(sb, KERN_ERR, "insufficient memory");
  3346. ret = err;
  3347. goto failed_mount3;
  3348. }
  3349. sbi->s_stripe = ext4_get_stripe_size(sbi);
  3350. sbi->s_max_writeback_mb_bump = 128;
  3351. sbi->s_extent_max_zeroout_kb = 32;
  3352. /*
  3353. * set up enough so that it can read an inode
  3354. */
  3355. if (!test_opt(sb, NOLOAD) &&
  3356. EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
  3357. sb->s_op = &ext4_sops;
  3358. else
  3359. sb->s_op = &ext4_nojournal_sops;
  3360. sb->s_export_op = &ext4_export_ops;
  3361. sb->s_xattr = ext4_xattr_handlers;
  3362. #ifdef CONFIG_QUOTA
  3363. sb->s_qcop = &ext4_qctl_operations;
  3364. sb->dq_op = &ext4_quota_operations;
  3365. if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
  3366. /* Use qctl operations for hidden quota files. */
  3367. sb->s_qcop = &ext4_qctl_sysfile_operations;
  3368. }
  3369. #endif
  3370. memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
  3371. INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
  3372. mutex_init(&sbi->s_orphan_lock);
  3373. sbi->s_resize_flags = 0;
  3374. sb->s_root = NULL;
  3375. needs_recovery = (es->s_last_orphan != 0 ||
  3376. EXT4_HAS_INCOMPAT_FEATURE(sb,
  3377. EXT4_FEATURE_INCOMPAT_RECOVER));
  3378. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
  3379. !(sb->s_flags & MS_RDONLY))
  3380. if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
  3381. goto failed_mount3;
  3382. /*
  3383. * The first inode we look at is the journal inode. Don't try
  3384. * root first: it may be modified in the journal!
  3385. */
  3386. if (!test_opt(sb, NOLOAD) &&
  3387. EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
  3388. if (ext4_load_journal(sb, es, journal_devnum))
  3389. goto failed_mount3;
  3390. } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
  3391. EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
  3392. ext4_msg(sb, KERN_ERR, "required journal recovery "
  3393. "suppressed and not mounted read-only");
  3394. goto failed_mount_wq;
  3395. } else {
  3396. clear_opt(sb, DATA_FLAGS);
  3397. sbi->s_journal = NULL;
  3398. needs_recovery = 0;
  3399. goto no_journal;
  3400. }
  3401. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
  3402. !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
  3403. JBD2_FEATURE_INCOMPAT_64BIT)) {
  3404. ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
  3405. goto failed_mount_wq;
  3406. }
  3407. if (!set_journal_csum_feature_set(sb)) {
  3408. ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
  3409. "feature set");
  3410. goto failed_mount_wq;
  3411. }
  3412. /* We have now updated the journal if required, so we can
  3413. * validate the data journaling mode. */
  3414. switch (test_opt(sb, DATA_FLAGS)) {
  3415. case 0:
  3416. /* No mode set, assume a default based on the journal
  3417. * capabilities: ORDERED_DATA if the journal can
  3418. * cope, else JOURNAL_DATA
  3419. */
  3420. if (jbd2_journal_check_available_features
  3421. (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
  3422. set_opt(sb, ORDERED_DATA);
  3423. else
  3424. set_opt(sb, JOURNAL_DATA);
  3425. break;
  3426. case EXT4_MOUNT_ORDERED_DATA:
  3427. case EXT4_MOUNT_WRITEBACK_DATA:
  3428. if (!jbd2_journal_check_available_features
  3429. (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
  3430. ext4_msg(sb, KERN_ERR, "Journal does not support "
  3431. "requested data journaling mode");
  3432. goto failed_mount_wq;
  3433. }
  3434. default:
  3435. break;
  3436. }
  3437. set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
  3438. sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
  3439. /*
  3440. * The journal may have updated the bg summary counts, so we
  3441. * need to update the global counters.
  3442. */
  3443. percpu_counter_set(&sbi->s_freeclusters_counter,
  3444. ext4_count_free_clusters(sb));
  3445. percpu_counter_set(&sbi->s_freeinodes_counter,
  3446. ext4_count_free_inodes(sb));
  3447. percpu_counter_set(&sbi->s_dirs_counter,
  3448. ext4_count_dirs(sb));
  3449. percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
  3450. no_journal:
  3451. /*
  3452. * Get the # of file system overhead blocks from the
  3453. * superblock if present.
  3454. */
  3455. if (es->s_overhead_clusters)
  3456. sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
  3457. else {
  3458. ret = ext4_calculate_overhead(sb);
  3459. if (ret)
  3460. goto failed_mount_wq;
  3461. }
  3462. /*
  3463. * The maximum number of concurrent works can be high and
  3464. * concurrency isn't really necessary. Limit it to 1.
  3465. */
  3466. EXT4_SB(sb)->dio_unwritten_wq =
  3467. alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
  3468. if (!EXT4_SB(sb)->dio_unwritten_wq) {
  3469. printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
  3470. goto failed_mount_wq;
  3471. }
  3472. /*
  3473. * The jbd2_journal_load will have done any necessary log recovery,
  3474. * so we can safely mount the rest of the filesystem now.
  3475. */
  3476. root = ext4_iget(sb, EXT4_ROOT_INO);
  3477. if (IS_ERR(root)) {
  3478. ext4_msg(sb, KERN_ERR, "get root inode failed");
  3479. ret = PTR_ERR(root);
  3480. root = NULL;
  3481. goto failed_mount4;
  3482. }
  3483. if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
  3484. ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
  3485. iput(root);
  3486. goto failed_mount4;
  3487. }
  3488. sb->s_root = d_make_root(root);
  3489. if (!sb->s_root) {
  3490. ext4_msg(sb, KERN_ERR, "get root dentry failed");
  3491. ret = -ENOMEM;
  3492. goto failed_mount4;
  3493. }
  3494. if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
  3495. sb->s_flags |= MS_RDONLY;
  3496. /* determine the minimum size of new large inodes, if present */
  3497. if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
  3498. sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
  3499. EXT4_GOOD_OLD_INODE_SIZE;
  3500. if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  3501. EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
  3502. if (sbi->s_want_extra_isize <
  3503. le16_to_cpu(es->s_want_extra_isize))
  3504. sbi->s_want_extra_isize =
  3505. le16_to_cpu(es->s_want_extra_isize);
  3506. if (sbi->s_want_extra_isize <
  3507. le16_to_cpu(es->s_min_extra_isize))
  3508. sbi->s_want_extra_isize =
  3509. le16_to_cpu(es->s_min_extra_isize);
  3510. }
  3511. }
  3512. /* Check if enough inode space is available */
  3513. if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
  3514. sbi->s_inode_size) {
  3515. sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
  3516. EXT4_GOOD_OLD_INODE_SIZE;
  3517. ext4_msg(sb, KERN_INFO, "required extra inode space not"
  3518. "available");
  3519. }
  3520. err = ext4_setup_system_zone(sb);
  3521. if (err) {
  3522. ext4_msg(sb, KERN_ERR, "failed to initialize system "
  3523. "zone (%d)", err);
  3524. goto failed_mount4a;
  3525. }
  3526. ext4_ext_init(sb);
  3527. err = ext4_mb_init(sb);
  3528. if (err) {
  3529. ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
  3530. err);
  3531. goto failed_mount5;
  3532. }
  3533. err = ext4_register_li_request(sb, first_not_zeroed);
  3534. if (err)
  3535. goto failed_mount6;
  3536. sbi->s_kobj.kset = ext4_kset;
  3537. init_completion(&sbi->s_kobj_unregister);
  3538. err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
  3539. "%s", sb->s_id);
  3540. if (err)
  3541. goto failed_mount7;
  3542. EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
  3543. ext4_orphan_cleanup(sb, es);
  3544. EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
  3545. if (needs_recovery) {
  3546. ext4_msg(sb, KERN_INFO, "recovery complete");
  3547. ext4_mark_recovery_complete(sb, es);
  3548. }
  3549. if (EXT4_SB(sb)->s_journal) {
  3550. if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
  3551. descr = " journalled data mode";
  3552. else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
  3553. descr = " ordered data mode";
  3554. else
  3555. descr = " writeback data mode";
  3556. } else
  3557. descr = "out journal";
  3558. #ifdef CONFIG_QUOTA
  3559. /* Enable quota usage during mount. */
  3560. if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
  3561. !(sb->s_flags & MS_RDONLY)) {
  3562. ret = ext4_enable_quotas(sb);
  3563. if (ret)
  3564. goto failed_mount7;
  3565. }
  3566. #endif /* CONFIG_QUOTA */
  3567. ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
  3568. "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
  3569. *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
  3570. if (es->s_error_count)
  3571. mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
  3572. kfree(orig_data);
  3573. return 0;
  3574. cantfind_ext4:
  3575. if (!silent)
  3576. ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
  3577. goto failed_mount;
  3578. failed_mount7:
  3579. ext4_unregister_li_request(sb);
  3580. failed_mount6:
  3581. ext4_mb_release(sb);
  3582. failed_mount5:
  3583. ext4_ext_release(sb);
  3584. ext4_release_system_zone(sb);
  3585. failed_mount4a:
  3586. dput(sb->s_root);
  3587. sb->s_root = NULL;
  3588. failed_mount4:
  3589. ext4_msg(sb, KERN_ERR, "mount failed");
  3590. destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
  3591. failed_mount_wq:
  3592. if (sbi->s_journal) {
  3593. jbd2_journal_destroy(sbi->s_journal);
  3594. sbi->s_journal = NULL;
  3595. }
  3596. failed_mount3:
  3597. del_timer(&sbi->s_err_report);
  3598. if (sbi->s_flex_groups)
  3599. ext4_kvfree(sbi->s_flex_groups);
  3600. percpu_counter_destroy(&sbi->s_freeclusters_counter);
  3601. percpu_counter_destroy(&sbi->s_freeinodes_counter);
  3602. percpu_counter_destroy(&sbi->s_dirs_counter);
  3603. percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
  3604. if (sbi->s_mmp_tsk)
  3605. kthread_stop(sbi->s_mmp_tsk);
  3606. failed_mount2:
  3607. for (i = 0; i < db_count; i++)
  3608. brelse(sbi->s_group_desc[i]);
  3609. ext4_kvfree(sbi->s_group_desc);
  3610. failed_mount:
  3611. if (sbi->s_chksum_driver)
  3612. crypto_free_shash(sbi->s_chksum_driver);
  3613. if (sbi->s_proc) {
  3614. remove_proc_entry("options", sbi->s_proc);
  3615. remove_proc_entry(sb->s_id, ext4_proc_root);
  3616. }
  3617. #ifdef CONFIG_QUOTA
  3618. for (i = 0; i < MAXQUOTAS; i++)
  3619. kfree(sbi->s_qf_names[i]);
  3620. #endif
  3621. ext4_blkdev_remove(sbi);
  3622. brelse(bh);
  3623. out_fail:
  3624. sb->s_fs_info = NULL;
  3625. kfree(sbi->s_blockgroup_lock);
  3626. kfree(sbi);
  3627. out_free_orig:
  3628. kfree(orig_data);
  3629. return ret;
  3630. }
  3631. /*
  3632. * Setup any per-fs journal parameters now. We'll do this both on
  3633. * initial mount, once the journal has been initialised but before we've
  3634. * done any recovery; and again on any subsequent remount.
  3635. */
  3636. static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
  3637. {
  3638. struct ext4_sb_info *sbi = EXT4_SB(sb);
  3639. journal->j_commit_interval = sbi->s_commit_interval;
  3640. journal->j_min_batch_time = sbi->s_min_batch_time;
  3641. journal->j_max_batch_time = sbi->s_max_batch_time;
  3642. write_lock(&journal->j_state_lock);
  3643. if (test_opt(sb, BARRIER))
  3644. journal->j_flags |= JBD2_BARRIER;
  3645. else
  3646. journal->j_flags &= ~JBD2_BARRIER;
  3647. if (test_opt(sb, DATA_ERR_ABORT))
  3648. journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
  3649. else
  3650. journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
  3651. write_unlock(&journal->j_state_lock);
  3652. }
  3653. static journal_t *ext4_get_journal(struct super_block *sb,
  3654. unsigned int journal_inum)
  3655. {
  3656. struct inode *journal_inode;
  3657. journal_t *journal;
  3658. BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
  3659. /* First, test for the existence of a valid inode on disk. Bad
  3660. * things happen if we iget() an unused inode, as the subsequent
  3661. * iput() will try to delete it. */
  3662. journal_inode = ext4_iget(sb, journal_inum);
  3663. if (IS_ERR(journal_inode)) {
  3664. ext4_msg(sb, KERN_ERR, "no journal found");
  3665. return NULL;
  3666. }
  3667. if (!journal_inode->i_nlink) {
  3668. make_bad_inode(journal_inode);
  3669. iput(journal_inode);
  3670. ext4_msg(sb, KERN_ERR, "journal inode is deleted");
  3671. return NULL;
  3672. }
  3673. jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
  3674. journal_inode, journal_inode->i_size);
  3675. if (!S_ISREG(journal_inode->i_mode)) {
  3676. ext4_msg(sb, KERN_ERR, "invalid journal inode");
  3677. iput(journal_inode);
  3678. return NULL;
  3679. }
  3680. journal = jbd2_journal_init_inode(journal_inode);
  3681. if (!journal) {
  3682. ext4_msg(sb, KERN_ERR, "Could not load journal inode");
  3683. iput(journal_inode);
  3684. return NULL;
  3685. }
  3686. journal->j_private = sb;
  3687. ext4_init_journal_params(sb, journal);
  3688. return journal;
  3689. }
  3690. static journal_t *ext4_get_dev_journal(struct super_block *sb,
  3691. dev_t j_dev)
  3692. {
  3693. struct buffer_head *bh;
  3694. journal_t *journal;
  3695. ext4_fsblk_t start;
  3696. ext4_fsblk_t len;
  3697. int hblock, blocksize;
  3698. ext4_fsblk_t sb_block;
  3699. unsigned long offset;
  3700. struct ext4_super_block *es;
  3701. struct block_device *bdev;
  3702. BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
  3703. bdev = ext4_blkdev_get(j_dev, sb);
  3704. if (bdev == NULL)
  3705. return NULL;
  3706. blocksize = sb->s_blocksize;
  3707. hblock = bdev_logical_block_size(bdev);
  3708. if (blocksize < hblock) {
  3709. ext4_msg(sb, KERN_ERR,
  3710. "blocksize too small for journal device");
  3711. goto out_bdev;
  3712. }
  3713. sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
  3714. offset = EXT4_MIN_BLOCK_SIZE % blocksize;
  3715. set_blocksize(bdev, blocksize);
  3716. if (!(bh = __bread(bdev, sb_block, blocksize))) {
  3717. ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
  3718. "external journal");
  3719. goto out_bdev;
  3720. }
  3721. es = (struct ext4_super_block *) (bh->b_data + offset);
  3722. if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
  3723. !(le32_to_cpu(es->s_feature_incompat) &
  3724. EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
  3725. ext4_msg(sb, KERN_ERR, "external journal has "
  3726. "bad superblock");
  3727. brelse(bh);
  3728. goto out_bdev;
  3729. }
  3730. if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
  3731. ext4_msg(sb, KERN_ERR, "journal UUID does not match");
  3732. brelse(bh);
  3733. goto out_bdev;
  3734. }
  3735. len = ext4_blocks_count(es);
  3736. start = sb_block + 1;
  3737. brelse(bh); /* we're done with the superblock */
  3738. journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
  3739. start, len, blocksize);
  3740. if (!journal) {
  3741. ext4_msg(sb, KERN_ERR, "failed to create device journal");
  3742. goto out_bdev;
  3743. }
  3744. journal->j_private = sb;
  3745. ll_rw_block(READ, 1, &journal->j_sb_buffer);
  3746. wait_on_buffer(journal->j_sb_buffer);
  3747. if (!buffer_uptodate(journal->j_sb_buffer)) {
  3748. ext4_msg(sb, KERN_ERR, "I/O error on journal device");
  3749. goto out_journal;
  3750. }
  3751. if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
  3752. ext4_msg(sb, KERN_ERR, "External journal has more than one "
  3753. "user (unsupported) - %d",
  3754. be32_to_cpu(journal->j_superblock->s_nr_users));
  3755. goto out_journal;
  3756. }
  3757. EXT4_SB(sb)->journal_bdev = bdev;
  3758. ext4_init_journal_params(sb, journal);
  3759. return journal;
  3760. out_journal:
  3761. jbd2_journal_destroy(journal);
  3762. out_bdev:
  3763. ext4_blkdev_put(bdev);
  3764. return NULL;
  3765. }
  3766. static int ext4_load_journal(struct super_block *sb,
  3767. struct ext4_super_block *es,
  3768. unsigned long journal_devnum)
  3769. {
  3770. journal_t *journal;
  3771. unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
  3772. dev_t journal_dev;
  3773. int err = 0;
  3774. int really_read_only;
  3775. BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
  3776. if (journal_devnum &&
  3777. journal_devnum != le32_to_cpu(es->s_journal_dev)) {
  3778. ext4_msg(sb, KERN_INFO, "external journal device major/minor "
  3779. "numbers have changed");
  3780. journal_dev = new_decode_dev(journal_devnum);
  3781. } else
  3782. journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
  3783. really_read_only = bdev_read_only(sb->s_bdev);
  3784. /*
  3785. * Are we loading a blank journal or performing recovery after a
  3786. * crash? For recovery, we need to check in advance whether we
  3787. * can get read-write access to the device.
  3788. */
  3789. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
  3790. if (sb->s_flags & MS_RDONLY) {
  3791. ext4_msg(sb, KERN_INFO, "INFO: recovery "
  3792. "required on readonly filesystem");
  3793. if (really_read_only) {
  3794. ext4_msg(sb, KERN_ERR, "write access "
  3795. "unavailable, cannot proceed");
  3796. return -EROFS;
  3797. }
  3798. ext4_msg(sb, KERN_INFO, "write access will "
  3799. "be enabled during recovery");
  3800. }
  3801. }
  3802. if (journal_inum && journal_dev) {
  3803. ext4_msg(sb, KERN_ERR, "filesystem has both journal "
  3804. "and inode journals!");
  3805. return -EINVAL;
  3806. }
  3807. if (journal_inum) {
  3808. if (!(journal = ext4_get_journal(sb, journal_inum)))
  3809. return -EINVAL;
  3810. } else {
  3811. if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
  3812. return -EINVAL;
  3813. }
  3814. if (!(journal->j_flags & JBD2_BARRIER))
  3815. ext4_msg(sb, KERN_INFO, "barriers disabled");
  3816. if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
  3817. err = jbd2_journal_wipe(journal, !really_read_only);
  3818. if (!err) {
  3819. char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
  3820. if (save)
  3821. memcpy(save, ((char *) es) +
  3822. EXT4_S_ERR_START, EXT4_S_ERR_LEN);
  3823. err = jbd2_journal_load(journal);
  3824. if (save)
  3825. memcpy(((char *) es) + EXT4_S_ERR_START,
  3826. save, EXT4_S_ERR_LEN);
  3827. kfree(save);
  3828. }
  3829. if (err) {
  3830. ext4_msg(sb, KERN_ERR, "error loading journal");
  3831. jbd2_journal_destroy(journal);
  3832. return err;
  3833. }
  3834. EXT4_SB(sb)->s_journal = journal;
  3835. ext4_clear_journal_err(sb, es);
  3836. if (!really_read_only && journal_devnum &&
  3837. journal_devnum != le32_to_cpu(es->s_journal_dev)) {
  3838. es->s_journal_dev = cpu_to_le32(journal_devnum);
  3839. /* Make sure we flush the recovery flag to disk. */
  3840. ext4_commit_super(sb, 1);
  3841. }
  3842. return 0;
  3843. }
  3844. static int ext4_commit_super(struct super_block *sb, int sync)
  3845. {
  3846. struct ext4_super_block *es = EXT4_SB(sb)->s_es;
  3847. struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
  3848. int error = 0;
  3849. if (!sbh || block_device_ejected(sb))
  3850. return error;
  3851. if (buffer_write_io_error(sbh)) {
  3852. /*
  3853. * Oh, dear. A previous attempt to write the
  3854. * superblock failed. This could happen because the
  3855. * USB device was yanked out. Or it could happen to
  3856. * be a transient write error and maybe the block will
  3857. * be remapped. Nothing we can do but to retry the
  3858. * write and hope for the best.
  3859. */
  3860. ext4_msg(sb, KERN_ERR, "previous I/O error to "
  3861. "superblock detected");
  3862. clear_buffer_write_io_error(sbh);
  3863. set_buffer_uptodate(sbh);
  3864. }
  3865. /*
  3866. * If the file system is mounted read-only, don't update the
  3867. * superblock write time. This avoids updating the superblock
  3868. * write time when we are mounting the root file system
  3869. * read/only but we need to replay the journal; at that point,
  3870. * for people who are east of GMT and who make their clock
  3871. * tick in localtime for Windows bug-for-bug compatibility,
  3872. * the clock is set in the future, and this will cause e2fsck
  3873. * to complain and force a full file system check.
  3874. */
  3875. if (!(sb->s_flags & MS_RDONLY))
  3876. es->s_wtime = cpu_to_le32(get_seconds());
  3877. if (sb->s_bdev->bd_part)
  3878. es->s_kbytes_written =
  3879. cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
  3880. ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
  3881. EXT4_SB(sb)->s_sectors_written_start) >> 1));
  3882. else
  3883. es->s_kbytes_written =
  3884. cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
  3885. ext4_free_blocks_count_set(es,
  3886. EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
  3887. &EXT4_SB(sb)->s_freeclusters_counter)));
  3888. es->s_free_inodes_count =
  3889. cpu_to_le32(percpu_counter_sum_positive(
  3890. &EXT4_SB(sb)->s_freeinodes_counter));
  3891. BUFFER_TRACE(sbh, "marking dirty");
  3892. ext4_superblock_csum_set(sb, es);
  3893. mark_buffer_dirty(sbh);
  3894. if (sync) {
  3895. error = sync_dirty_buffer(sbh);
  3896. if (error)
  3897. return error;
  3898. error = buffer_write_io_error(sbh);
  3899. if (error) {
  3900. ext4_msg(sb, KERN_ERR, "I/O error while writing "
  3901. "superblock");
  3902. clear_buffer_write_io_error(sbh);
  3903. set_buffer_uptodate(sbh);
  3904. }
  3905. }
  3906. return error;
  3907. }
  3908. /*
  3909. * Have we just finished recovery? If so, and if we are mounting (or
  3910. * remounting) the filesystem readonly, then we will end up with a
  3911. * consistent fs on disk. Record that fact.
  3912. */
  3913. static void ext4_mark_recovery_complete(struct super_block *sb,
  3914. struct ext4_super_block *es)
  3915. {
  3916. journal_t *journal = EXT4_SB(sb)->s_journal;
  3917. if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
  3918. BUG_ON(journal != NULL);
  3919. return;
  3920. }
  3921. jbd2_journal_lock_updates(journal);
  3922. if (jbd2_journal_flush(journal) < 0)
  3923. goto out;
  3924. if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
  3925. sb->s_flags & MS_RDONLY) {
  3926. EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
  3927. ext4_commit_super(sb, 1);
  3928. }
  3929. out:
  3930. jbd2_journal_unlock_updates(journal);
  3931. }
  3932. /*
  3933. * If we are mounting (or read-write remounting) a filesystem whose journal
  3934. * has recorded an error from a previous lifetime, move that error to the
  3935. * main filesystem now.
  3936. */
  3937. static void ext4_clear_journal_err(struct super_block *sb,
  3938. struct ext4_super_block *es)
  3939. {
  3940. journal_t *journal;
  3941. int j_errno;
  3942. const char *errstr;
  3943. BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
  3944. journal = EXT4_SB(sb)->s_journal;
  3945. /*
  3946. * Now check for any error status which may have been recorded in the
  3947. * journal by a prior ext4_error() or ext4_abort()
  3948. */
  3949. j_errno = jbd2_journal_errno(journal);
  3950. if (j_errno) {
  3951. char nbuf[16];
  3952. errstr = ext4_decode_error(sb, j_errno, nbuf);
  3953. ext4_warning(sb, "Filesystem error recorded "
  3954. "from previous mount: %s", errstr);
  3955. ext4_warning(sb, "Marking fs in need of filesystem check.");
  3956. EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
  3957. es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
  3958. ext4_commit_super(sb, 1);
  3959. jbd2_journal_clear_err(journal);
  3960. jbd2_journal_update_sb_errno(journal);
  3961. }
  3962. }
  3963. /*
  3964. * Force the running and committing transactions to commit,
  3965. * and wait on the commit.
  3966. */
  3967. int ext4_force_commit(struct super_block *sb)
  3968. {
  3969. journal_t *journal;
  3970. int ret = 0;
  3971. if (sb->s_flags & MS_RDONLY)
  3972. return 0;
  3973. journal = EXT4_SB(sb)->s_journal;
  3974. if (journal)
  3975. ret = ext4_journal_force_commit(journal);
  3976. return ret;
  3977. }
  3978. static int ext4_sync_fs(struct super_block *sb, int wait)
  3979. {
  3980. int ret = 0;
  3981. tid_t target;
  3982. struct ext4_sb_info *sbi = EXT4_SB(sb);
  3983. trace_ext4_sync_fs(sb, wait);
  3984. flush_workqueue(sbi->dio_unwritten_wq);
  3985. /*
  3986. * Writeback quota in non-journalled quota case - journalled quota has
  3987. * no dirty dquots
  3988. */
  3989. dquot_writeback_dquots(sb, -1);
  3990. if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
  3991. if (wait)
  3992. jbd2_log_wait_commit(sbi->s_journal, target);
  3993. }
  3994. return ret;
  3995. }
  3996. /*
  3997. * LVM calls this function before a (read-only) snapshot is created. This
  3998. * gives us a chance to flush the journal completely and mark the fs clean.
  3999. *
  4000. * Note that only this function cannot bring a filesystem to be in a clean
  4001. * state independently. It relies on upper layer to stop all data & metadata
  4002. * modifications.
  4003. */
  4004. static int ext4_freeze(struct super_block *sb)
  4005. {
  4006. int error = 0;
  4007. journal_t *journal;
  4008. if (sb->s_flags & MS_RDONLY)
  4009. return 0;
  4010. journal = EXT4_SB(sb)->s_journal;
  4011. /* Now we set up the journal barrier. */
  4012. jbd2_journal_lock_updates(journal);
  4013. /*
  4014. * Don't clear the needs_recovery flag if we failed to flush
  4015. * the journal.
  4016. */
  4017. error = jbd2_journal_flush(journal);
  4018. if (error < 0)
  4019. goto out;
  4020. /* Journal blocked and flushed, clear needs_recovery flag. */
  4021. EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
  4022. error = ext4_commit_super(sb, 1);
  4023. out:
  4024. /* we rely on upper layer to stop further updates */
  4025. jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
  4026. return error;
  4027. }
  4028. /*
  4029. * Called by LVM after the snapshot is done. We need to reset the RECOVER
  4030. * flag here, even though the filesystem is not technically dirty yet.
  4031. */
  4032. static int ext4_unfreeze(struct super_block *sb)
  4033. {
  4034. if (sb->s_flags & MS_RDONLY)
  4035. return 0;
  4036. /* Reset the needs_recovery flag before the fs is unlocked. */
  4037. EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
  4038. ext4_commit_super(sb, 1);
  4039. return 0;
  4040. }
  4041. /*
  4042. * Structure to save mount options for ext4_remount's benefit
  4043. */
  4044. struct ext4_mount_options {
  4045. unsigned long s_mount_opt;
  4046. unsigned long s_mount_opt2;
  4047. kuid_t s_resuid;
  4048. kgid_t s_resgid;
  4049. unsigned long s_commit_interval;
  4050. u32 s_min_batch_time, s_max_batch_time;
  4051. #ifdef CONFIG_QUOTA
  4052. int s_jquota_fmt;
  4053. char *s_qf_names[MAXQUOTAS];
  4054. #endif
  4055. };
  4056. static int ext4_remount(struct super_block *sb, int *flags, char *data)
  4057. {
  4058. struct ext4_super_block *es;
  4059. struct ext4_sb_info *sbi = EXT4_SB(sb);
  4060. unsigned long old_sb_flags;
  4061. struct ext4_mount_options old_opts;
  4062. int enable_quota = 0;
  4063. ext4_group_t g;
  4064. unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
  4065. int err = 0;
  4066. #ifdef CONFIG_QUOTA
  4067. int i;
  4068. #endif
  4069. char *orig_data = kstrdup(data, GFP_KERNEL);
  4070. /* Store the original options */
  4071. old_sb_flags = sb->s_flags;
  4072. old_opts.s_mount_opt = sbi->s_mount_opt;
  4073. old_opts.s_mount_opt2 = sbi->s_mount_opt2;
  4074. old_opts.s_resuid = sbi->s_resuid;
  4075. old_opts.s_resgid = sbi->s_resgid;
  4076. old_opts.s_commit_interval = sbi->s_commit_interval;
  4077. old_opts.s_min_batch_time = sbi->s_min_batch_time;
  4078. old_opts.s_max_batch_time = sbi->s_max_batch_time;
  4079. #ifdef CONFIG_QUOTA
  4080. old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
  4081. for (i = 0; i < MAXQUOTAS; i++)
  4082. old_opts.s_qf_names[i] = sbi->s_qf_names[i];
  4083. #endif
  4084. if (sbi->s_journal && sbi->s_journal->j_task->io_context)
  4085. journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
  4086. /*
  4087. * Allow the "check" option to be passed as a remount option.
  4088. */
  4089. if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
  4090. err = -EINVAL;
  4091. goto restore_opts;
  4092. }
  4093. if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
  4094. ext4_abort(sb, "Abort forced by user");
  4095. sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
  4096. (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
  4097. es = sbi->s_es;
  4098. if (sbi->s_journal) {
  4099. ext4_init_journal_params(sb, sbi->s_journal);
  4100. set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
  4101. }
  4102. if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
  4103. if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
  4104. err = -EROFS;
  4105. goto restore_opts;
  4106. }
  4107. if (*flags & MS_RDONLY) {
  4108. err = dquot_suspend(sb, -1);
  4109. if (err < 0)
  4110. goto restore_opts;
  4111. /*
  4112. * First of all, the unconditional stuff we have to do
  4113. * to disable replay of the journal when we next remount
  4114. */
  4115. sb->s_flags |= MS_RDONLY;
  4116. /*
  4117. * OK, test if we are remounting a valid rw partition
  4118. * readonly, and if so set the rdonly flag and then
  4119. * mark the partition as valid again.
  4120. */
  4121. if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
  4122. (sbi->s_mount_state & EXT4_VALID_FS))
  4123. es->s_state = cpu_to_le16(sbi->s_mount_state);
  4124. if (sbi->s_journal)
  4125. ext4_mark_recovery_complete(sb, es);
  4126. } else {
  4127. /* Make sure we can mount this feature set readwrite */
  4128. if (!ext4_feature_set_ok(sb, 0)) {
  4129. err = -EROFS;
  4130. goto restore_opts;
  4131. }
  4132. /*
  4133. * Make sure the group descriptor checksums
  4134. * are sane. If they aren't, refuse to remount r/w.
  4135. */
  4136. for (g = 0; g < sbi->s_groups_count; g++) {
  4137. struct ext4_group_desc *gdp =
  4138. ext4_get_group_desc(sb, g, NULL);
  4139. if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
  4140. ext4_msg(sb, KERN_ERR,
  4141. "ext4_remount: Checksum for group %u failed (%u!=%u)",
  4142. g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
  4143. le16_to_cpu(gdp->bg_checksum));
  4144. err = -EINVAL;
  4145. goto restore_opts;
  4146. }
  4147. }
  4148. /*
  4149. * If we have an unprocessed orphan list hanging
  4150. * around from a previously readonly bdev mount,
  4151. * require a full umount/remount for now.
  4152. */
  4153. if (es->s_last_orphan) {
  4154. ext4_msg(sb, KERN_WARNING, "Couldn't "
  4155. "remount RDWR because of unprocessed "
  4156. "orphan inode list. Please "
  4157. "umount/remount instead");
  4158. err = -EINVAL;
  4159. goto restore_opts;
  4160. }
  4161. /*
  4162. * Mounting a RDONLY partition read-write, so reread
  4163. * and store the current valid flag. (It may have
  4164. * been changed by e2fsck since we originally mounted
  4165. * the partition.)
  4166. */
  4167. if (sbi->s_journal)
  4168. ext4_clear_journal_err(sb, es);
  4169. sbi->s_mount_state = le16_to_cpu(es->s_state);
  4170. if (!ext4_setup_super(sb, es, 0))
  4171. sb->s_flags &= ~MS_RDONLY;
  4172. if (EXT4_HAS_INCOMPAT_FEATURE(sb,
  4173. EXT4_FEATURE_INCOMPAT_MMP))
  4174. if (ext4_multi_mount_protect(sb,
  4175. le64_to_cpu(es->s_mmp_block))) {
  4176. err = -EROFS;
  4177. goto restore_opts;
  4178. }
  4179. enable_quota = 1;
  4180. }
  4181. }
  4182. /*
  4183. * Reinitialize lazy itable initialization thread based on
  4184. * current settings
  4185. */
  4186. if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
  4187. ext4_unregister_li_request(sb);
  4188. else {
  4189. ext4_group_t first_not_zeroed;
  4190. first_not_zeroed = ext4_has_uninit_itable(sb);
  4191. ext4_register_li_request(sb, first_not_zeroed);
  4192. }
  4193. ext4_setup_system_zone(sb);
  4194. if (sbi->s_journal == NULL)
  4195. ext4_commit_super(sb, 1);
  4196. #ifdef CONFIG_QUOTA
  4197. /* Release old quota file names */
  4198. for (i = 0; i < MAXQUOTAS; i++)
  4199. if (old_opts.s_qf_names[i] &&
  4200. old_opts.s_qf_names[i] != sbi->s_qf_names[i])
  4201. kfree(old_opts.s_qf_names[i]);
  4202. if (enable_quota) {
  4203. if (sb_any_quota_suspended(sb))
  4204. dquot_resume(sb, -1);
  4205. else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
  4206. EXT4_FEATURE_RO_COMPAT_QUOTA)) {
  4207. err = ext4_enable_quotas(sb);
  4208. if (err)
  4209. goto restore_opts;
  4210. }
  4211. }
  4212. #endif
  4213. ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
  4214. kfree(orig_data);
  4215. return 0;
  4216. restore_opts:
  4217. sb->s_flags = old_sb_flags;
  4218. sbi->s_mount_opt = old_opts.s_mount_opt;
  4219. sbi->s_mount_opt2 = old_opts.s_mount_opt2;
  4220. sbi->s_resuid = old_opts.s_resuid;
  4221. sbi->s_resgid = old_opts.s_resgid;
  4222. sbi->s_commit_interval = old_opts.s_commit_interval;
  4223. sbi->s_min_batch_time = old_opts.s_min_batch_time;
  4224. sbi->s_max_batch_time = old_opts.s_max_batch_time;
  4225. #ifdef CONFIG_QUOTA
  4226. sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
  4227. for (i = 0; i < MAXQUOTAS; i++) {
  4228. if (sbi->s_qf_names[i] &&
  4229. old_opts.s_qf_names[i] != sbi->s_qf_names[i])
  4230. kfree(sbi->s_qf_names[i]);
  4231. sbi->s_qf_names[i] = old_opts.s_qf_names[i];
  4232. }
  4233. #endif
  4234. kfree(orig_data);
  4235. return err;
  4236. }
  4237. static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
  4238. {
  4239. struct super_block *sb = dentry->d_sb;
  4240. struct ext4_sb_info *sbi = EXT4_SB(sb);
  4241. struct ext4_super_block *es = sbi->s_es;
  4242. ext4_fsblk_t overhead = 0;
  4243. u64 fsid;
  4244. s64 bfree;
  4245. if (!test_opt(sb, MINIX_DF))
  4246. overhead = sbi->s_overhead;
  4247. buf->f_type = EXT4_SUPER_MAGIC;
  4248. buf->f_bsize = sb->s_blocksize;
  4249. buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
  4250. bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
  4251. percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
  4252. /* prevent underflow in case that few free space is available */
  4253. buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
  4254. buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
  4255. if (buf->f_bfree < ext4_r_blocks_count(es))
  4256. buf->f_bavail = 0;
  4257. buf->f_files = le32_to_cpu(es->s_inodes_count);
  4258. buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
  4259. buf->f_namelen = EXT4_NAME_LEN;
  4260. fsid = le64_to_cpup((void *)es->s_uuid) ^
  4261. le64_to_cpup((void *)es->s_uuid + sizeof(u64));
  4262. buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
  4263. buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
  4264. return 0;
  4265. }
  4266. /* Helper function for writing quotas on sync - we need to start transaction
  4267. * before quota file is locked for write. Otherwise the are possible deadlocks:
  4268. * Process 1 Process 2
  4269. * ext4_create() quota_sync()
  4270. * jbd2_journal_start() write_dquot()
  4271. * dquot_initialize() down(dqio_mutex)
  4272. * down(dqio_mutex) jbd2_journal_start()
  4273. *
  4274. */
  4275. #ifdef CONFIG_QUOTA
  4276. static inline struct inode *dquot_to_inode(struct dquot *dquot)
  4277. {
  4278. return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
  4279. }
  4280. static int ext4_write_dquot(struct dquot *dquot)
  4281. {
  4282. int ret, err;
  4283. handle_t *handle;
  4284. struct inode *inode;
  4285. inode = dquot_to_inode(dquot);
  4286. handle = ext4_journal_start(inode,
  4287. EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
  4288. if (IS_ERR(handle))
  4289. return PTR_ERR(handle);
  4290. ret = dquot_commit(dquot);
  4291. err = ext4_journal_stop(handle);
  4292. if (!ret)
  4293. ret = err;
  4294. return ret;
  4295. }
  4296. static int ext4_acquire_dquot(struct dquot *dquot)
  4297. {
  4298. int ret, err;
  4299. handle_t *handle;
  4300. handle = ext4_journal_start(dquot_to_inode(dquot),
  4301. EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
  4302. if (IS_ERR(handle))
  4303. return PTR_ERR(handle);
  4304. ret = dquot_acquire(dquot);
  4305. err = ext4_journal_stop(handle);
  4306. if (!ret)
  4307. ret = err;
  4308. return ret;
  4309. }
  4310. static int ext4_release_dquot(struct dquot *dquot)
  4311. {
  4312. int ret, err;
  4313. handle_t *handle;
  4314. handle = ext4_journal_start(dquot_to_inode(dquot),
  4315. EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
  4316. if (IS_ERR(handle)) {
  4317. /* Release dquot anyway to avoid endless cycle in dqput() */
  4318. dquot_release(dquot);
  4319. return PTR_ERR(handle);
  4320. }
  4321. ret = dquot_release(dquot);
  4322. err = ext4_journal_stop(handle);
  4323. if (!ret)
  4324. ret = err;
  4325. return ret;
  4326. }
  4327. static int ext4_mark_dquot_dirty(struct dquot *dquot)
  4328. {
  4329. /* Are we journaling quotas? */
  4330. if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
  4331. EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
  4332. dquot_mark_dquot_dirty(dquot);
  4333. return ext4_write_dquot(dquot);
  4334. } else {
  4335. return dquot_mark_dquot_dirty(dquot);
  4336. }
  4337. }
  4338. static int ext4_write_info(struct super_block *sb, int type)
  4339. {
  4340. int ret, err;
  4341. handle_t *handle;
  4342. /* Data block + inode block */
  4343. handle = ext4_journal_start(sb->s_root->d_inode, 2);
  4344. if (IS_ERR(handle))
  4345. return PTR_ERR(handle);
  4346. ret = dquot_commit_info(sb, type);
  4347. err = ext4_journal_stop(handle);
  4348. if (!ret)
  4349. ret = err;
  4350. return ret;
  4351. }
  4352. /*
  4353. * Turn on quotas during mount time - we need to find
  4354. * the quota file and such...
  4355. */
  4356. static int ext4_quota_on_mount(struct super_block *sb, int type)
  4357. {
  4358. return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
  4359. EXT4_SB(sb)->s_jquota_fmt, type);
  4360. }
  4361. /*
  4362. * Standard function to be called on quota_on
  4363. */
  4364. static int ext4_quota_on(struct super_block *sb, int type, int format_id,
  4365. struct path *path)
  4366. {
  4367. int err;
  4368. if (!test_opt(sb, QUOTA))
  4369. return -EINVAL;
  4370. /* Quotafile not on the same filesystem? */
  4371. if (path->dentry->d_sb != sb)
  4372. return -EXDEV;
  4373. /* Journaling quota? */
  4374. if (EXT4_SB(sb)->s_qf_names[type]) {
  4375. /* Quotafile not in fs root? */
  4376. if (path->dentry->d_parent != sb->s_root)
  4377. ext4_msg(sb, KERN_WARNING,
  4378. "Quota file not on filesystem root. "
  4379. "Journaled quota will not work");
  4380. }
  4381. /*
  4382. * When we journal data on quota file, we have to flush journal to see
  4383. * all updates to the file when we bypass pagecache...
  4384. */
  4385. if (EXT4_SB(sb)->s_journal &&
  4386. ext4_should_journal_data(path->dentry->d_inode)) {
  4387. /*
  4388. * We don't need to lock updates but journal_flush() could
  4389. * otherwise be livelocked...
  4390. */
  4391. jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
  4392. err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
  4393. jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
  4394. if (err)
  4395. return err;
  4396. }
  4397. return dquot_quota_on(sb, type, format_id, path);
  4398. }
  4399. static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
  4400. unsigned int flags)
  4401. {
  4402. int err;
  4403. struct inode *qf_inode;
  4404. unsigned long qf_inums[MAXQUOTAS] = {
  4405. le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
  4406. le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
  4407. };
  4408. BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
  4409. if (!qf_inums[type])
  4410. return -EPERM;
  4411. qf_inode = ext4_iget(sb, qf_inums[type]);
  4412. if (IS_ERR(qf_inode)) {
  4413. ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
  4414. return PTR_ERR(qf_inode);
  4415. }
  4416. err = dquot_enable(qf_inode, type, format_id, flags);
  4417. iput(qf_inode);
  4418. return err;
  4419. }
  4420. /* Enable usage tracking for all quota types. */
  4421. static int ext4_enable_quotas(struct super_block *sb)
  4422. {
  4423. int type, err = 0;
  4424. unsigned long qf_inums[MAXQUOTAS] = {
  4425. le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
  4426. le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
  4427. };
  4428. sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
  4429. for (type = 0; type < MAXQUOTAS; type++) {
  4430. if (qf_inums[type]) {
  4431. err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
  4432. DQUOT_USAGE_ENABLED);
  4433. if (err) {
  4434. ext4_warning(sb,
  4435. "Failed to enable quota (type=%d) "
  4436. "tracking. Please run e2fsck to fix.",
  4437. type);
  4438. return err;
  4439. }
  4440. }
  4441. }
  4442. return 0;
  4443. }
  4444. /*
  4445. * quota_on function that is used when QUOTA feature is set.
  4446. */
  4447. static int ext4_quota_on_sysfile(struct super_block *sb, int type,
  4448. int format_id)
  4449. {
  4450. if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
  4451. return -EINVAL;
  4452. /*
  4453. * USAGE was enabled at mount time. Only need to enable LIMITS now.
  4454. */
  4455. return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
  4456. }
  4457. static int ext4_quota_off(struct super_block *sb, int type)
  4458. {
  4459. struct inode *inode = sb_dqopt(sb)->files[type];
  4460. handle_t *handle;
  4461. /* Force all delayed allocation blocks to be allocated.
  4462. * Caller already holds s_umount sem */
  4463. if (test_opt(sb, DELALLOC))
  4464. sync_filesystem(sb);
  4465. if (!inode)
  4466. goto out;
  4467. /* Update modification times of quota files when userspace can
  4468. * start looking at them */
  4469. handle = ext4_journal_start(inode, 1);
  4470. if (IS_ERR(handle))
  4471. goto out;
  4472. inode->i_mtime = inode->i_ctime = CURRENT_TIME;
  4473. ext4_mark_inode_dirty(handle, inode);
  4474. ext4_journal_stop(handle);
  4475. out:
  4476. return dquot_quota_off(sb, type);
  4477. }
  4478. /*
  4479. * quota_off function that is used when QUOTA feature is set.
  4480. */
  4481. static int ext4_quota_off_sysfile(struct super_block *sb, int type)
  4482. {
  4483. if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
  4484. return -EINVAL;
  4485. /* Disable only the limits. */
  4486. return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
  4487. }
  4488. /* Read data from quotafile - avoid pagecache and such because we cannot afford
  4489. * acquiring the locks... As quota files are never truncated and quota code
  4490. * itself serializes the operations (and no one else should touch the files)
  4491. * we don't have to be afraid of races */
  4492. static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
  4493. size_t len, loff_t off)
  4494. {
  4495. struct inode *inode = sb_dqopt(sb)->files[type];
  4496. ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
  4497. int err = 0;
  4498. int offset = off & (sb->s_blocksize - 1);
  4499. int tocopy;
  4500. size_t toread;
  4501. struct buffer_head *bh;
  4502. loff_t i_size = i_size_read(inode);
  4503. if (off > i_size)
  4504. return 0;
  4505. if (off+len > i_size)
  4506. len = i_size-off;
  4507. toread = len;
  4508. while (toread > 0) {
  4509. tocopy = sb->s_blocksize - offset < toread ?
  4510. sb->s_blocksize - offset : toread;
  4511. bh = ext4_bread(NULL, inode, blk, 0, &err);
  4512. if (err)
  4513. return err;
  4514. if (!bh) /* A hole? */
  4515. memset(data, 0, tocopy);
  4516. else
  4517. memcpy(data, bh->b_data+offset, tocopy);
  4518. brelse(bh);
  4519. offset = 0;
  4520. toread -= tocopy;
  4521. data += tocopy;
  4522. blk++;
  4523. }
  4524. return len;
  4525. }
  4526. /* Write to quotafile (we know the transaction is already started and has
  4527. * enough credits) */
  4528. static ssize_t ext4_quota_write(struct super_block *sb, int type,
  4529. const char *data, size_t len, loff_t off)
  4530. {
  4531. struct inode *inode = sb_dqopt(sb)->files[type];
  4532. ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
  4533. int err = 0;
  4534. int offset = off & (sb->s_blocksize - 1);
  4535. struct buffer_head *bh;
  4536. handle_t *handle = journal_current_handle();
  4537. if (EXT4_SB(sb)->s_journal && !handle) {
  4538. ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
  4539. " cancelled because transaction is not started",
  4540. (unsigned long long)off, (unsigned long long)len);
  4541. return -EIO;
  4542. }
  4543. /*
  4544. * Since we account only one data block in transaction credits,
  4545. * then it is impossible to cross a block boundary.
  4546. */
  4547. if (sb->s_blocksize - offset < len) {
  4548. ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
  4549. " cancelled because not block aligned",
  4550. (unsigned long long)off, (unsigned long long)len);
  4551. return -EIO;
  4552. }
  4553. bh = ext4_bread(handle, inode, blk, 1, &err);
  4554. if (!bh)
  4555. goto out;
  4556. err = ext4_journal_get_write_access(handle, bh);
  4557. if (err) {
  4558. brelse(bh);
  4559. goto out;
  4560. }
  4561. lock_buffer(bh);
  4562. memcpy(bh->b_data+offset, data, len);
  4563. flush_dcache_page(bh->b_page);
  4564. unlock_buffer(bh);
  4565. err = ext4_handle_dirty_metadata(handle, NULL, bh);
  4566. brelse(bh);
  4567. out:
  4568. if (err)
  4569. return err;
  4570. if (inode->i_size < off + len) {
  4571. i_size_write(inode, off + len);
  4572. EXT4_I(inode)->i_disksize = inode->i_size;
  4573. ext4_mark_inode_dirty(handle, inode);
  4574. }
  4575. return len;
  4576. }
  4577. #endif
  4578. static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
  4579. const char *dev_name, void *data)
  4580. {
  4581. return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
  4582. }
  4583. #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
  4584. static inline void register_as_ext2(void)
  4585. {
  4586. int err = register_filesystem(&ext2_fs_type);
  4587. if (err)
  4588. printk(KERN_WARNING
  4589. "EXT4-fs: Unable to register as ext2 (%d)\n", err);
  4590. }
  4591. static inline void unregister_as_ext2(void)
  4592. {
  4593. unregister_filesystem(&ext2_fs_type);
  4594. }
  4595. static inline int ext2_feature_set_ok(struct super_block *sb)
  4596. {
  4597. if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
  4598. return 0;
  4599. if (sb->s_flags & MS_RDONLY)
  4600. return 1;
  4601. if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
  4602. return 0;
  4603. return 1;
  4604. }
  4605. MODULE_ALIAS("ext2");
  4606. #else
  4607. static inline void register_as_ext2(void) { }
  4608. static inline void unregister_as_ext2(void) { }
  4609. static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
  4610. #endif
  4611. #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
  4612. static inline void register_as_ext3(void)
  4613. {
  4614. int err = register_filesystem(&ext3_fs_type);
  4615. if (err)
  4616. printk(KERN_WARNING
  4617. "EXT4-fs: Unable to register as ext3 (%d)\n", err);
  4618. }
  4619. static inline void unregister_as_ext3(void)
  4620. {
  4621. unregister_filesystem(&ext3_fs_type);
  4622. }
  4623. static inline int ext3_feature_set_ok(struct super_block *sb)
  4624. {
  4625. if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
  4626. return 0;
  4627. if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
  4628. return 0;
  4629. if (sb->s_flags & MS_RDONLY)
  4630. return 1;
  4631. if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
  4632. return 0;
  4633. return 1;
  4634. }
  4635. MODULE_ALIAS("ext3");
  4636. #else
  4637. static inline void register_as_ext3(void) { }
  4638. static inline void unregister_as_ext3(void) { }
  4639. static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
  4640. #endif
  4641. static struct file_system_type ext4_fs_type = {
  4642. .owner = THIS_MODULE,
  4643. .name = "ext4",
  4644. .mount = ext4_mount,
  4645. .kill_sb = kill_block_super,
  4646. .fs_flags = FS_REQUIRES_DEV,
  4647. };
  4648. static int __init ext4_init_feat_adverts(void)
  4649. {
  4650. struct ext4_features *ef;
  4651. int ret = -ENOMEM;
  4652. ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
  4653. if (!ef)
  4654. goto out;
  4655. ef->f_kobj.kset = ext4_kset;
  4656. init_completion(&ef->f_kobj_unregister);
  4657. ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
  4658. "features");
  4659. if (ret) {
  4660. kfree(ef);
  4661. goto out;
  4662. }
  4663. ext4_feat = ef;
  4664. ret = 0;
  4665. out:
  4666. return ret;
  4667. }
  4668. static void ext4_exit_feat_adverts(void)
  4669. {
  4670. kobject_put(&ext4_feat->f_kobj);
  4671. wait_for_completion(&ext4_feat->f_kobj_unregister);
  4672. kfree(ext4_feat);
  4673. }
  4674. /* Shared across all ext4 file systems */
  4675. wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
  4676. struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
  4677. static int __init ext4_init_fs(void)
  4678. {
  4679. int i, err;
  4680. ext4_li_info = NULL;
  4681. mutex_init(&ext4_li_mtx);
  4682. ext4_check_flag_values();
  4683. for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
  4684. mutex_init(&ext4__aio_mutex[i]);
  4685. init_waitqueue_head(&ext4__ioend_wq[i]);
  4686. }
  4687. err = ext4_init_pageio();
  4688. if (err)
  4689. return err;
  4690. err = ext4_init_system_zone();
  4691. if (err)
  4692. goto out6;
  4693. ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
  4694. if (!ext4_kset) {
  4695. err = -ENOMEM;
  4696. goto out5;
  4697. }
  4698. ext4_proc_root = proc_mkdir("fs/ext4", NULL);
  4699. err = ext4_init_feat_adverts();
  4700. if (err)
  4701. goto out4;
  4702. err = ext4_init_mballoc();
  4703. if (err)
  4704. goto out3;
  4705. err = ext4_init_xattr();
  4706. if (err)
  4707. goto out2;
  4708. err = init_inodecache();
  4709. if (err)
  4710. goto out1;
  4711. register_as_ext3();
  4712. register_as_ext2();
  4713. err = register_filesystem(&ext4_fs_type);
  4714. if (err)
  4715. goto out;
  4716. return 0;
  4717. out:
  4718. unregister_as_ext2();
  4719. unregister_as_ext3();
  4720. destroy_inodecache();
  4721. out1:
  4722. ext4_exit_xattr();
  4723. out2:
  4724. ext4_exit_mballoc();
  4725. out3:
  4726. ext4_exit_feat_adverts();
  4727. out4:
  4728. if (ext4_proc_root)
  4729. remove_proc_entry("fs/ext4", NULL);
  4730. kset_unregister(ext4_kset);
  4731. out5:
  4732. ext4_exit_system_zone();
  4733. out6:
  4734. ext4_exit_pageio();
  4735. return err;
  4736. }
  4737. static void __exit ext4_exit_fs(void)
  4738. {
  4739. ext4_destroy_lazyinit_thread();
  4740. unregister_as_ext2();
  4741. unregister_as_ext3();
  4742. unregister_filesystem(&ext4_fs_type);
  4743. destroy_inodecache();
  4744. ext4_exit_xattr();
  4745. ext4_exit_mballoc();
  4746. ext4_exit_feat_adverts();
  4747. remove_proc_entry("fs/ext4", NULL);
  4748. kset_unregister(ext4_kset);
  4749. ext4_exit_system_zone();
  4750. ext4_exit_pageio();
  4751. }
  4752. MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
  4753. MODULE_DESCRIPTION("Fourth Extended Filesystem");
  4754. MODULE_LICENSE("GPL");
  4755. module_init(ext4_init_fs)
  4756. module_exit(ext4_exit_fs)