super.c 75 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699
  1. /*
  2. * linux/fs/ext3/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/config.h>
  19. #include <linux/module.h>
  20. #include <linux/string.h>
  21. #include <linux/fs.h>
  22. #include <linux/time.h>
  23. #include <linux/jbd.h>
  24. #include <linux/ext3_fs.h>
  25. #include <linux/ext3_jbd.h>
  26. #include <linux/slab.h>
  27. #include <linux/init.h>
  28. #include <linux/blkdev.h>
  29. #include <linux/parser.h>
  30. #include <linux/smp_lock.h>
  31. #include <linux/buffer_head.h>
  32. #include <linux/vfs.h>
  33. #include <linux/random.h>
  34. #include <linux/mount.h>
  35. #include <linux/namei.h>
  36. #include <linux/quotaops.h>
  37. #include <linux/seq_file.h>
  38. #include <asm/uaccess.h>
  39. #include "xattr.h"
  40. #include "acl.h"
  41. #include "namei.h"
  42. static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
  43. unsigned long journal_devnum);
  44. static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
  45. int);
  46. static void ext3_commit_super (struct super_block * sb,
  47. struct ext3_super_block * es,
  48. int sync);
  49. static void ext3_mark_recovery_complete(struct super_block * sb,
  50. struct ext3_super_block * es);
  51. static void ext3_clear_journal_err(struct super_block * sb,
  52. struct ext3_super_block * es);
  53. static int ext3_sync_fs(struct super_block *sb, int wait);
  54. static const char *ext3_decode_error(struct super_block * sb, int errno,
  55. char nbuf[16]);
  56. static int ext3_remount (struct super_block * sb, int * flags, char * data);
  57. static int ext3_statfs (struct super_block * sb, struct kstatfs * buf);
  58. static void ext3_unlockfs(struct super_block *sb);
  59. static void ext3_write_super (struct super_block * sb);
  60. static void ext3_write_super_lockfs(struct super_block *sb);
  61. /*
  62. * Wrappers for journal_start/end.
  63. *
  64. * The only special thing we need to do here is to make sure that all
  65. * journal_end calls result in the superblock being marked dirty, so
  66. * that sync() will call the filesystem's write_super callback if
  67. * appropriate.
  68. */
  69. handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
  70. {
  71. journal_t *journal;
  72. if (sb->s_flags & MS_RDONLY)
  73. return ERR_PTR(-EROFS);
  74. /* Special case here: if the journal has aborted behind our
  75. * backs (eg. EIO in the commit thread), then we still need to
  76. * take the FS itself readonly cleanly. */
  77. journal = EXT3_SB(sb)->s_journal;
  78. if (is_journal_aborted(journal)) {
  79. ext3_abort(sb, __FUNCTION__,
  80. "Detected aborted journal");
  81. return ERR_PTR(-EROFS);
  82. }
  83. return journal_start(journal, nblocks);
  84. }
  85. /*
  86. * The only special thing we need to do here is to make sure that all
  87. * journal_stop calls result in the superblock being marked dirty, so
  88. * that sync() will call the filesystem's write_super callback if
  89. * appropriate.
  90. */
  91. int __ext3_journal_stop(const char *where, handle_t *handle)
  92. {
  93. struct super_block *sb;
  94. int err;
  95. int rc;
  96. sb = handle->h_transaction->t_journal->j_private;
  97. err = handle->h_err;
  98. rc = journal_stop(handle);
  99. if (!err)
  100. err = rc;
  101. if (err)
  102. __ext3_std_error(sb, where, err);
  103. return err;
  104. }
  105. void ext3_journal_abort_handle(const char *caller, const char *err_fn,
  106. struct buffer_head *bh, handle_t *handle, int err)
  107. {
  108. char nbuf[16];
  109. const char *errstr = ext3_decode_error(NULL, err, nbuf);
  110. if (bh)
  111. BUFFER_TRACE(bh, "abort");
  112. if (!handle->h_err)
  113. handle->h_err = err;
  114. if (is_handle_aborted(handle))
  115. return;
  116. printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
  117. caller, errstr, err_fn);
  118. journal_abort_handle(handle);
  119. }
  120. /* Deal with the reporting of failure conditions on a filesystem such as
  121. * inconsistencies detected or read IO failures.
  122. *
  123. * On ext2, we can store the error state of the filesystem in the
  124. * superblock. That is not possible on ext3, because we may have other
  125. * write ordering constraints on the superblock which prevent us from
  126. * writing it out straight away; and given that the journal is about to
  127. * be aborted, we can't rely on the current, or future, transactions to
  128. * write out the superblock safely.
  129. *
  130. * We'll just use the journal_abort() error code to record an error in
  131. * the journal instead. On recovery, the journal will compain about
  132. * that error until we've noted it down and cleared it.
  133. */
  134. static void ext3_handle_error(struct super_block *sb)
  135. {
  136. struct ext3_super_block *es = EXT3_SB(sb)->s_es;
  137. EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
  138. es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
  139. if (sb->s_flags & MS_RDONLY)
  140. return;
  141. if (test_opt (sb, ERRORS_RO)) {
  142. printk (KERN_CRIT "Remounting filesystem read-only\n");
  143. sb->s_flags |= MS_RDONLY;
  144. } else {
  145. journal_t *journal = EXT3_SB(sb)->s_journal;
  146. EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
  147. if (journal)
  148. journal_abort(journal, -EIO);
  149. }
  150. if (test_opt(sb, ERRORS_PANIC))
  151. panic("EXT3-fs (device %s): panic forced after error\n",
  152. sb->s_id);
  153. ext3_commit_super(sb, es, 1);
  154. }
  155. void ext3_error (struct super_block * sb, const char * function,
  156. const char * fmt, ...)
  157. {
  158. va_list args;
  159. va_start(args, fmt);
  160. printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
  161. vprintk(fmt, args);
  162. printk("\n");
  163. va_end(args);
  164. ext3_handle_error(sb);
  165. }
  166. static const char *ext3_decode_error(struct super_block * sb, int errno,
  167. char nbuf[16])
  168. {
  169. char *errstr = NULL;
  170. switch (errno) {
  171. case -EIO:
  172. errstr = "IO failure";
  173. break;
  174. case -ENOMEM:
  175. errstr = "Out of memory";
  176. break;
  177. case -EROFS:
  178. if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
  179. errstr = "Journal has aborted";
  180. else
  181. errstr = "Readonly filesystem";
  182. break;
  183. default:
  184. /* If the caller passed in an extra buffer for unknown
  185. * errors, textualise them now. Else we just return
  186. * NULL. */
  187. if (nbuf) {
  188. /* Check for truncated error codes... */
  189. if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
  190. errstr = nbuf;
  191. }
  192. break;
  193. }
  194. return errstr;
  195. }
  196. /* __ext3_std_error decodes expected errors from journaling functions
  197. * automatically and invokes the appropriate error response. */
  198. void __ext3_std_error (struct super_block * sb, const char * function,
  199. int errno)
  200. {
  201. char nbuf[16];
  202. const char *errstr;
  203. /* Special case: if the error is EROFS, and we're not already
  204. * inside a transaction, then there's really no point in logging
  205. * an error. */
  206. if (errno == -EROFS && journal_current_handle() == NULL &&
  207. (sb->s_flags & MS_RDONLY))
  208. return;
  209. errstr = ext3_decode_error(sb, errno, nbuf);
  210. printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
  211. sb->s_id, function, errstr);
  212. ext3_handle_error(sb);
  213. }
  214. /*
  215. * ext3_abort is a much stronger failure handler than ext3_error. The
  216. * abort function may be used to deal with unrecoverable failures such
  217. * as journal IO errors or ENOMEM at a critical moment in log management.
  218. *
  219. * We unconditionally force the filesystem into an ABORT|READONLY state,
  220. * unless the error response on the fs has been set to panic in which
  221. * case we take the easy way out and panic immediately.
  222. */
  223. void ext3_abort (struct super_block * sb, const char * function,
  224. const char * fmt, ...)
  225. {
  226. va_list args;
  227. printk (KERN_CRIT "ext3_abort called.\n");
  228. va_start(args, fmt);
  229. printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
  230. vprintk(fmt, args);
  231. printk("\n");
  232. va_end(args);
  233. if (test_opt(sb, ERRORS_PANIC))
  234. panic("EXT3-fs panic from previous error\n");
  235. if (sb->s_flags & MS_RDONLY)
  236. return;
  237. printk(KERN_CRIT "Remounting filesystem read-only\n");
  238. EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
  239. sb->s_flags |= MS_RDONLY;
  240. EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
  241. journal_abort(EXT3_SB(sb)->s_journal, -EIO);
  242. }
  243. void ext3_warning (struct super_block * sb, const char * function,
  244. const char * fmt, ...)
  245. {
  246. va_list args;
  247. va_start(args, fmt);
  248. printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
  249. sb->s_id, function);
  250. vprintk(fmt, args);
  251. printk("\n");
  252. va_end(args);
  253. }
  254. void ext3_update_dynamic_rev(struct super_block *sb)
  255. {
  256. struct ext3_super_block *es = EXT3_SB(sb)->s_es;
  257. if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
  258. return;
  259. ext3_warning(sb, __FUNCTION__,
  260. "updating to rev %d because of new feature flag, "
  261. "running e2fsck is recommended",
  262. EXT3_DYNAMIC_REV);
  263. es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
  264. es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
  265. es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
  266. /* leave es->s_feature_*compat flags alone */
  267. /* es->s_uuid will be set by e2fsck if empty */
  268. /*
  269. * The rest of the superblock fields should be zero, and if not it
  270. * means they are likely already in use, so leave them alone. We
  271. * can leave it up to e2fsck to clean up any inconsistencies there.
  272. */
  273. }
  274. /*
  275. * Open the external journal device
  276. */
  277. static struct block_device *ext3_blkdev_get(dev_t dev)
  278. {
  279. struct block_device *bdev;
  280. char b[BDEVNAME_SIZE];
  281. bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
  282. if (IS_ERR(bdev))
  283. goto fail;
  284. return bdev;
  285. fail:
  286. printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
  287. __bdevname(dev, b), PTR_ERR(bdev));
  288. return NULL;
  289. }
  290. /*
  291. * Release the journal device
  292. */
  293. static int ext3_blkdev_put(struct block_device *bdev)
  294. {
  295. bd_release(bdev);
  296. return blkdev_put(bdev);
  297. }
  298. static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
  299. {
  300. struct block_device *bdev;
  301. int ret = -ENODEV;
  302. bdev = sbi->journal_bdev;
  303. if (bdev) {
  304. ret = ext3_blkdev_put(bdev);
  305. sbi->journal_bdev = NULL;
  306. }
  307. return ret;
  308. }
  309. static inline struct inode *orphan_list_entry(struct list_head *l)
  310. {
  311. return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
  312. }
  313. static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
  314. {
  315. struct list_head *l;
  316. printk(KERN_ERR "sb orphan head is %d\n",
  317. le32_to_cpu(sbi->s_es->s_last_orphan));
  318. printk(KERN_ERR "sb_info orphan list:\n");
  319. list_for_each(l, &sbi->s_orphan) {
  320. struct inode *inode = orphan_list_entry(l);
  321. printk(KERN_ERR " "
  322. "inode %s:%ld at %p: mode %o, nlink %d, next %d\n",
  323. inode->i_sb->s_id, inode->i_ino, inode,
  324. inode->i_mode, inode->i_nlink,
  325. NEXT_ORPHAN(inode));
  326. }
  327. }
  328. static void ext3_put_super (struct super_block * sb)
  329. {
  330. struct ext3_sb_info *sbi = EXT3_SB(sb);
  331. struct ext3_super_block *es = sbi->s_es;
  332. int i;
  333. ext3_xattr_put_super(sb);
  334. journal_destroy(sbi->s_journal);
  335. if (!(sb->s_flags & MS_RDONLY)) {
  336. EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
  337. es->s_state = cpu_to_le16(sbi->s_mount_state);
  338. BUFFER_TRACE(sbi->s_sbh, "marking dirty");
  339. mark_buffer_dirty(sbi->s_sbh);
  340. ext3_commit_super(sb, es, 1);
  341. }
  342. for (i = 0; i < sbi->s_gdb_count; i++)
  343. brelse(sbi->s_group_desc[i]);
  344. kfree(sbi->s_group_desc);
  345. percpu_counter_destroy(&sbi->s_freeblocks_counter);
  346. percpu_counter_destroy(&sbi->s_freeinodes_counter);
  347. percpu_counter_destroy(&sbi->s_dirs_counter);
  348. brelse(sbi->s_sbh);
  349. #ifdef CONFIG_QUOTA
  350. for (i = 0; i < MAXQUOTAS; i++)
  351. kfree(sbi->s_qf_names[i]);
  352. #endif
  353. /* Debugging code just in case the in-memory inode orphan list
  354. * isn't empty. The on-disk one can be non-empty if we've
  355. * detected an error and taken the fs readonly, but the
  356. * in-memory list had better be clean by this point. */
  357. if (!list_empty(&sbi->s_orphan))
  358. dump_orphan_list(sb, sbi);
  359. J_ASSERT(list_empty(&sbi->s_orphan));
  360. invalidate_bdev(sb->s_bdev, 0);
  361. if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
  362. /*
  363. * Invalidate the journal device's buffers. We don't want them
  364. * floating about in memory - the physical journal device may
  365. * hotswapped, and it breaks the `ro-after' testing code.
  366. */
  367. sync_blockdev(sbi->journal_bdev);
  368. invalidate_bdev(sbi->journal_bdev, 0);
  369. ext3_blkdev_remove(sbi);
  370. }
  371. sb->s_fs_info = NULL;
  372. kfree(sbi);
  373. return;
  374. }
  375. static kmem_cache_t *ext3_inode_cachep;
  376. /*
  377. * Called inside transaction, so use GFP_NOFS
  378. */
  379. static struct inode *ext3_alloc_inode(struct super_block *sb)
  380. {
  381. struct ext3_inode_info *ei;
  382. ei = kmem_cache_alloc(ext3_inode_cachep, SLAB_NOFS);
  383. if (!ei)
  384. return NULL;
  385. #ifdef CONFIG_EXT3_FS_POSIX_ACL
  386. ei->i_acl = EXT3_ACL_NOT_CACHED;
  387. ei->i_default_acl = EXT3_ACL_NOT_CACHED;
  388. #endif
  389. ei->i_block_alloc_info = NULL;
  390. ei->vfs_inode.i_version = 1;
  391. return &ei->vfs_inode;
  392. }
  393. static void ext3_destroy_inode(struct inode *inode)
  394. {
  395. kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
  396. }
  397. static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
  398. {
  399. struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
  400. if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
  401. SLAB_CTOR_CONSTRUCTOR) {
  402. INIT_LIST_HEAD(&ei->i_orphan);
  403. #ifdef CONFIG_EXT3_FS_XATTR
  404. init_rwsem(&ei->xattr_sem);
  405. #endif
  406. mutex_init(&ei->truncate_mutex);
  407. inode_init_once(&ei->vfs_inode);
  408. }
  409. }
  410. static int init_inodecache(void)
  411. {
  412. ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
  413. sizeof(struct ext3_inode_info),
  414. 0, (SLAB_RECLAIM_ACCOUNT|
  415. SLAB_MEM_SPREAD),
  416. init_once, NULL);
  417. if (ext3_inode_cachep == NULL)
  418. return -ENOMEM;
  419. return 0;
  420. }
  421. static void destroy_inodecache(void)
  422. {
  423. if (kmem_cache_destroy(ext3_inode_cachep))
  424. printk(KERN_INFO "ext3_inode_cache: not all structures were freed\n");
  425. }
  426. static void ext3_clear_inode(struct inode *inode)
  427. {
  428. struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
  429. #ifdef CONFIG_EXT3_FS_POSIX_ACL
  430. if (EXT3_I(inode)->i_acl &&
  431. EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
  432. posix_acl_release(EXT3_I(inode)->i_acl);
  433. EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
  434. }
  435. if (EXT3_I(inode)->i_default_acl &&
  436. EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
  437. posix_acl_release(EXT3_I(inode)->i_default_acl);
  438. EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
  439. }
  440. #endif
  441. ext3_discard_reservation(inode);
  442. EXT3_I(inode)->i_block_alloc_info = NULL;
  443. kfree(rsv);
  444. }
  445. static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
  446. {
  447. #if defined(CONFIG_QUOTA)
  448. struct ext3_sb_info *sbi = EXT3_SB(sb);
  449. if (sbi->s_jquota_fmt)
  450. seq_printf(seq, ",jqfmt=%s",
  451. (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
  452. if (sbi->s_qf_names[USRQUOTA])
  453. seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
  454. if (sbi->s_qf_names[GRPQUOTA])
  455. seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
  456. if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
  457. seq_puts(seq, ",usrquota");
  458. if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
  459. seq_puts(seq, ",grpquota");
  460. #endif
  461. }
  462. static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
  463. {
  464. struct super_block *sb = vfs->mnt_sb;
  465. if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
  466. seq_puts(seq, ",data=journal");
  467. else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
  468. seq_puts(seq, ",data=ordered");
  469. else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
  470. seq_puts(seq, ",data=writeback");
  471. ext3_show_quota_options(seq, sb);
  472. return 0;
  473. }
  474. #ifdef CONFIG_QUOTA
  475. #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
  476. #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
  477. static int ext3_dquot_initialize(struct inode *inode, int type);
  478. static int ext3_dquot_drop(struct inode *inode);
  479. static int ext3_write_dquot(struct dquot *dquot);
  480. static int ext3_acquire_dquot(struct dquot *dquot);
  481. static int ext3_release_dquot(struct dquot *dquot);
  482. static int ext3_mark_dquot_dirty(struct dquot *dquot);
  483. static int ext3_write_info(struct super_block *sb, int type);
  484. static int ext3_quota_on(struct super_block *sb, int type, int format_id, char *path);
  485. static int ext3_quota_on_mount(struct super_block *sb, int type);
  486. static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
  487. size_t len, loff_t off);
  488. static ssize_t ext3_quota_write(struct super_block *sb, int type,
  489. const char *data, size_t len, loff_t off);
  490. static struct dquot_operations ext3_quota_operations = {
  491. .initialize = ext3_dquot_initialize,
  492. .drop = ext3_dquot_drop,
  493. .alloc_space = dquot_alloc_space,
  494. .alloc_inode = dquot_alloc_inode,
  495. .free_space = dquot_free_space,
  496. .free_inode = dquot_free_inode,
  497. .transfer = dquot_transfer,
  498. .write_dquot = ext3_write_dquot,
  499. .acquire_dquot = ext3_acquire_dquot,
  500. .release_dquot = ext3_release_dquot,
  501. .mark_dirty = ext3_mark_dquot_dirty,
  502. .write_info = ext3_write_info
  503. };
  504. static struct quotactl_ops ext3_qctl_operations = {
  505. .quota_on = ext3_quota_on,
  506. .quota_off = vfs_quota_off,
  507. .quota_sync = vfs_quota_sync,
  508. .get_info = vfs_get_dqinfo,
  509. .set_info = vfs_set_dqinfo,
  510. .get_dqblk = vfs_get_dqblk,
  511. .set_dqblk = vfs_set_dqblk
  512. };
  513. #endif
  514. static struct super_operations ext3_sops = {
  515. .alloc_inode = ext3_alloc_inode,
  516. .destroy_inode = ext3_destroy_inode,
  517. .read_inode = ext3_read_inode,
  518. .write_inode = ext3_write_inode,
  519. .dirty_inode = ext3_dirty_inode,
  520. .delete_inode = ext3_delete_inode,
  521. .put_super = ext3_put_super,
  522. .write_super = ext3_write_super,
  523. .sync_fs = ext3_sync_fs,
  524. .write_super_lockfs = ext3_write_super_lockfs,
  525. .unlockfs = ext3_unlockfs,
  526. .statfs = ext3_statfs,
  527. .remount_fs = ext3_remount,
  528. .clear_inode = ext3_clear_inode,
  529. .show_options = ext3_show_options,
  530. #ifdef CONFIG_QUOTA
  531. .quota_read = ext3_quota_read,
  532. .quota_write = ext3_quota_write,
  533. #endif
  534. };
  535. static struct export_operations ext3_export_ops = {
  536. .get_parent = ext3_get_parent,
  537. };
  538. enum {
  539. Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
  540. Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
  541. Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
  542. Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
  543. Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh,
  544. Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
  545. Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
  546. Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
  547. Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
  548. Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
  549. Opt_grpquota
  550. };
  551. static match_table_t tokens = {
  552. {Opt_bsd_df, "bsddf"},
  553. {Opt_minix_df, "minixdf"},
  554. {Opt_grpid, "grpid"},
  555. {Opt_grpid, "bsdgroups"},
  556. {Opt_nogrpid, "nogrpid"},
  557. {Opt_nogrpid, "sysvgroups"},
  558. {Opt_resgid, "resgid=%u"},
  559. {Opt_resuid, "resuid=%u"},
  560. {Opt_sb, "sb=%u"},
  561. {Opt_err_cont, "errors=continue"},
  562. {Opt_err_panic, "errors=panic"},
  563. {Opt_err_ro, "errors=remount-ro"},
  564. {Opt_nouid32, "nouid32"},
  565. {Opt_nocheck, "nocheck"},
  566. {Opt_nocheck, "check=none"},
  567. {Opt_debug, "debug"},
  568. {Opt_oldalloc, "oldalloc"},
  569. {Opt_orlov, "orlov"},
  570. {Opt_user_xattr, "user_xattr"},
  571. {Opt_nouser_xattr, "nouser_xattr"},
  572. {Opt_acl, "acl"},
  573. {Opt_noacl, "noacl"},
  574. {Opt_reservation, "reservation"},
  575. {Opt_noreservation, "noreservation"},
  576. {Opt_noload, "noload"},
  577. {Opt_nobh, "nobh"},
  578. {Opt_commit, "commit=%u"},
  579. {Opt_journal_update, "journal=update"},
  580. {Opt_journal_inum, "journal=%u"},
  581. {Opt_journal_dev, "journal_dev=%u"},
  582. {Opt_abort, "abort"},
  583. {Opt_data_journal, "data=journal"},
  584. {Opt_data_ordered, "data=ordered"},
  585. {Opt_data_writeback, "data=writeback"},
  586. {Opt_offusrjquota, "usrjquota="},
  587. {Opt_usrjquota, "usrjquota=%s"},
  588. {Opt_offgrpjquota, "grpjquota="},
  589. {Opt_grpjquota, "grpjquota=%s"},
  590. {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
  591. {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
  592. {Opt_grpquota, "grpquota"},
  593. {Opt_noquota, "noquota"},
  594. {Opt_quota, "quota"},
  595. {Opt_usrquota, "usrquota"},
  596. {Opt_barrier, "barrier=%u"},
  597. {Opt_err, NULL},
  598. {Opt_resize, "resize"},
  599. };
  600. static unsigned long get_sb_block(void **data)
  601. {
  602. unsigned long sb_block;
  603. char *options = (char *) *data;
  604. if (!options || strncmp(options, "sb=", 3) != 0)
  605. return 1; /* Default location */
  606. options += 3;
  607. sb_block = simple_strtoul(options, &options, 0);
  608. if (*options && *options != ',') {
  609. printk("EXT3-fs: Invalid sb specification: %s\n",
  610. (char *) *data);
  611. return 1;
  612. }
  613. if (*options == ',')
  614. options++;
  615. *data = (void *) options;
  616. return sb_block;
  617. }
  618. static int parse_options (char *options, struct super_block *sb,
  619. unsigned long *inum, unsigned long *journal_devnum,
  620. unsigned long *n_blocks_count, int is_remount)
  621. {
  622. struct ext3_sb_info *sbi = EXT3_SB(sb);
  623. char * p;
  624. substring_t args[MAX_OPT_ARGS];
  625. int data_opt = 0;
  626. int option;
  627. #ifdef CONFIG_QUOTA
  628. int qtype;
  629. char *qname;
  630. #endif
  631. if (!options)
  632. return 1;
  633. while ((p = strsep (&options, ",")) != NULL) {
  634. int token;
  635. if (!*p)
  636. continue;
  637. token = match_token(p, tokens, args);
  638. switch (token) {
  639. case Opt_bsd_df:
  640. clear_opt (sbi->s_mount_opt, MINIX_DF);
  641. break;
  642. case Opt_minix_df:
  643. set_opt (sbi->s_mount_opt, MINIX_DF);
  644. break;
  645. case Opt_grpid:
  646. set_opt (sbi->s_mount_opt, GRPID);
  647. break;
  648. case Opt_nogrpid:
  649. clear_opt (sbi->s_mount_opt, GRPID);
  650. break;
  651. case Opt_resuid:
  652. if (match_int(&args[0], &option))
  653. return 0;
  654. sbi->s_resuid = option;
  655. break;
  656. case Opt_resgid:
  657. if (match_int(&args[0], &option))
  658. return 0;
  659. sbi->s_resgid = option;
  660. break;
  661. case Opt_sb:
  662. /* handled by get_sb_block() instead of here */
  663. /* *sb_block = match_int(&args[0]); */
  664. break;
  665. case Opt_err_panic:
  666. clear_opt (sbi->s_mount_opt, ERRORS_CONT);
  667. clear_opt (sbi->s_mount_opt, ERRORS_RO);
  668. set_opt (sbi->s_mount_opt, ERRORS_PANIC);
  669. break;
  670. case Opt_err_ro:
  671. clear_opt (sbi->s_mount_opt, ERRORS_CONT);
  672. clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
  673. set_opt (sbi->s_mount_opt, ERRORS_RO);
  674. break;
  675. case Opt_err_cont:
  676. clear_opt (sbi->s_mount_opt, ERRORS_RO);
  677. clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
  678. set_opt (sbi->s_mount_opt, ERRORS_CONT);
  679. break;
  680. case Opt_nouid32:
  681. set_opt (sbi->s_mount_opt, NO_UID32);
  682. break;
  683. case Opt_nocheck:
  684. clear_opt (sbi->s_mount_opt, CHECK);
  685. break;
  686. case Opt_debug:
  687. set_opt (sbi->s_mount_opt, DEBUG);
  688. break;
  689. case Opt_oldalloc:
  690. set_opt (sbi->s_mount_opt, OLDALLOC);
  691. break;
  692. case Opt_orlov:
  693. clear_opt (sbi->s_mount_opt, OLDALLOC);
  694. break;
  695. #ifdef CONFIG_EXT3_FS_XATTR
  696. case Opt_user_xattr:
  697. set_opt (sbi->s_mount_opt, XATTR_USER);
  698. break;
  699. case Opt_nouser_xattr:
  700. clear_opt (sbi->s_mount_opt, XATTR_USER);
  701. break;
  702. #else
  703. case Opt_user_xattr:
  704. case Opt_nouser_xattr:
  705. printk("EXT3 (no)user_xattr options not supported\n");
  706. break;
  707. #endif
  708. #ifdef CONFIG_EXT3_FS_POSIX_ACL
  709. case Opt_acl:
  710. set_opt(sbi->s_mount_opt, POSIX_ACL);
  711. break;
  712. case Opt_noacl:
  713. clear_opt(sbi->s_mount_opt, POSIX_ACL);
  714. break;
  715. #else
  716. case Opt_acl:
  717. case Opt_noacl:
  718. printk("EXT3 (no)acl options not supported\n");
  719. break;
  720. #endif
  721. case Opt_reservation:
  722. set_opt(sbi->s_mount_opt, RESERVATION);
  723. break;
  724. case Opt_noreservation:
  725. clear_opt(sbi->s_mount_opt, RESERVATION);
  726. break;
  727. case Opt_journal_update:
  728. /* @@@ FIXME */
  729. /* Eventually we will want to be able to create
  730. a journal file here. For now, only allow the
  731. user to specify an existing inode to be the
  732. journal file. */
  733. if (is_remount) {
  734. printk(KERN_ERR "EXT3-fs: cannot specify "
  735. "journal on remount\n");
  736. return 0;
  737. }
  738. set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
  739. break;
  740. case Opt_journal_inum:
  741. if (is_remount) {
  742. printk(KERN_ERR "EXT3-fs: cannot specify "
  743. "journal on remount\n");
  744. return 0;
  745. }
  746. if (match_int(&args[0], &option))
  747. return 0;
  748. *inum = option;
  749. break;
  750. case Opt_journal_dev:
  751. if (is_remount) {
  752. printk(KERN_ERR "EXT3-fs: cannot specify "
  753. "journal on remount\n");
  754. return 0;
  755. }
  756. if (match_int(&args[0], &option))
  757. return 0;
  758. *journal_devnum = option;
  759. break;
  760. case Opt_noload:
  761. set_opt (sbi->s_mount_opt, NOLOAD);
  762. break;
  763. case Opt_commit:
  764. if (match_int(&args[0], &option))
  765. return 0;
  766. if (option < 0)
  767. return 0;
  768. if (option == 0)
  769. option = JBD_DEFAULT_MAX_COMMIT_AGE;
  770. sbi->s_commit_interval = HZ * option;
  771. break;
  772. case Opt_data_journal:
  773. data_opt = EXT3_MOUNT_JOURNAL_DATA;
  774. goto datacheck;
  775. case Opt_data_ordered:
  776. data_opt = EXT3_MOUNT_ORDERED_DATA;
  777. goto datacheck;
  778. case Opt_data_writeback:
  779. data_opt = EXT3_MOUNT_WRITEBACK_DATA;
  780. datacheck:
  781. if (is_remount) {
  782. if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
  783. != data_opt) {
  784. printk(KERN_ERR
  785. "EXT3-fs: cannot change data "
  786. "mode on remount\n");
  787. return 0;
  788. }
  789. } else {
  790. sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
  791. sbi->s_mount_opt |= data_opt;
  792. }
  793. break;
  794. #ifdef CONFIG_QUOTA
  795. case Opt_usrjquota:
  796. qtype = USRQUOTA;
  797. goto set_qf_name;
  798. case Opt_grpjquota:
  799. qtype = GRPQUOTA;
  800. set_qf_name:
  801. if (sb_any_quota_enabled(sb)) {
  802. printk(KERN_ERR
  803. "EXT3-fs: Cannot change journalled "
  804. "quota options when quota turned on.\n");
  805. return 0;
  806. }
  807. qname = match_strdup(&args[0]);
  808. if (!qname) {
  809. printk(KERN_ERR
  810. "EXT3-fs: not enough memory for "
  811. "storing quotafile name.\n");
  812. return 0;
  813. }
  814. if (sbi->s_qf_names[qtype] &&
  815. strcmp(sbi->s_qf_names[qtype], qname)) {
  816. printk(KERN_ERR
  817. "EXT3-fs: %s quota file already "
  818. "specified.\n", QTYPE2NAME(qtype));
  819. kfree(qname);
  820. return 0;
  821. }
  822. sbi->s_qf_names[qtype] = qname;
  823. if (strchr(sbi->s_qf_names[qtype], '/')) {
  824. printk(KERN_ERR
  825. "EXT3-fs: quotafile must be on "
  826. "filesystem root.\n");
  827. kfree(sbi->s_qf_names[qtype]);
  828. sbi->s_qf_names[qtype] = NULL;
  829. return 0;
  830. }
  831. set_opt(sbi->s_mount_opt, QUOTA);
  832. break;
  833. case Opt_offusrjquota:
  834. qtype = USRQUOTA;
  835. goto clear_qf_name;
  836. case Opt_offgrpjquota:
  837. qtype = GRPQUOTA;
  838. clear_qf_name:
  839. if (sb_any_quota_enabled(sb)) {
  840. printk(KERN_ERR "EXT3-fs: Cannot change "
  841. "journalled quota options when "
  842. "quota turned on.\n");
  843. return 0;
  844. }
  845. /*
  846. * The space will be released later when all options
  847. * are confirmed to be correct
  848. */
  849. sbi->s_qf_names[qtype] = NULL;
  850. break;
  851. case Opt_jqfmt_vfsold:
  852. sbi->s_jquota_fmt = QFMT_VFS_OLD;
  853. break;
  854. case Opt_jqfmt_vfsv0:
  855. sbi->s_jquota_fmt = QFMT_VFS_V0;
  856. break;
  857. case Opt_quota:
  858. case Opt_usrquota:
  859. set_opt(sbi->s_mount_opt, QUOTA);
  860. set_opt(sbi->s_mount_opt, USRQUOTA);
  861. break;
  862. case Opt_grpquota:
  863. set_opt(sbi->s_mount_opt, QUOTA);
  864. set_opt(sbi->s_mount_opt, GRPQUOTA);
  865. break;
  866. case Opt_noquota:
  867. if (sb_any_quota_enabled(sb)) {
  868. printk(KERN_ERR "EXT3-fs: Cannot change quota "
  869. "options when quota turned on.\n");
  870. return 0;
  871. }
  872. clear_opt(sbi->s_mount_opt, QUOTA);
  873. clear_opt(sbi->s_mount_opt, USRQUOTA);
  874. clear_opt(sbi->s_mount_opt, GRPQUOTA);
  875. break;
  876. #else
  877. case Opt_quota:
  878. case Opt_usrquota:
  879. case Opt_grpquota:
  880. case Opt_usrjquota:
  881. case Opt_grpjquota:
  882. case Opt_offusrjquota:
  883. case Opt_offgrpjquota:
  884. case Opt_jqfmt_vfsold:
  885. case Opt_jqfmt_vfsv0:
  886. printk(KERN_ERR
  887. "EXT3-fs: journalled quota options not "
  888. "supported.\n");
  889. break;
  890. case Opt_noquota:
  891. break;
  892. #endif
  893. case Opt_abort:
  894. set_opt(sbi->s_mount_opt, ABORT);
  895. break;
  896. case Opt_barrier:
  897. if (match_int(&args[0], &option))
  898. return 0;
  899. if (option)
  900. set_opt(sbi->s_mount_opt, BARRIER);
  901. else
  902. clear_opt(sbi->s_mount_opt, BARRIER);
  903. break;
  904. case Opt_ignore:
  905. break;
  906. case Opt_resize:
  907. if (!is_remount) {
  908. printk("EXT3-fs: resize option only available "
  909. "for remount\n");
  910. return 0;
  911. }
  912. if (match_int(&args[0], &option) != 0)
  913. return 0;
  914. *n_blocks_count = option;
  915. break;
  916. case Opt_nobh:
  917. set_opt(sbi->s_mount_opt, NOBH);
  918. break;
  919. default:
  920. printk (KERN_ERR
  921. "EXT3-fs: Unrecognized mount option \"%s\" "
  922. "or missing value\n", p);
  923. return 0;
  924. }
  925. }
  926. #ifdef CONFIG_QUOTA
  927. if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
  928. if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
  929. sbi->s_qf_names[USRQUOTA])
  930. clear_opt(sbi->s_mount_opt, USRQUOTA);
  931. if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
  932. sbi->s_qf_names[GRPQUOTA])
  933. clear_opt(sbi->s_mount_opt, GRPQUOTA);
  934. if ((sbi->s_qf_names[USRQUOTA] &&
  935. (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
  936. (sbi->s_qf_names[GRPQUOTA] &&
  937. (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
  938. printk(KERN_ERR "EXT3-fs: old and new quota "
  939. "format mixing.\n");
  940. return 0;
  941. }
  942. if (!sbi->s_jquota_fmt) {
  943. printk(KERN_ERR "EXT3-fs: journalled quota format "
  944. "not specified.\n");
  945. return 0;
  946. }
  947. } else {
  948. if (sbi->s_jquota_fmt) {
  949. printk(KERN_ERR "EXT3-fs: journalled quota format "
  950. "specified with no journalling "
  951. "enabled.\n");
  952. return 0;
  953. }
  954. }
  955. #endif
  956. return 1;
  957. }
  958. static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
  959. int read_only)
  960. {
  961. struct ext3_sb_info *sbi = EXT3_SB(sb);
  962. int res = 0;
  963. if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
  964. printk (KERN_ERR "EXT3-fs warning: revision level too high, "
  965. "forcing read-only mode\n");
  966. res = MS_RDONLY;
  967. }
  968. if (read_only)
  969. return res;
  970. if (!(sbi->s_mount_state & EXT3_VALID_FS))
  971. printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
  972. "running e2fsck is recommended\n");
  973. else if ((sbi->s_mount_state & EXT3_ERROR_FS))
  974. printk (KERN_WARNING
  975. "EXT3-fs warning: mounting fs with errors, "
  976. "running e2fsck is recommended\n");
  977. else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
  978. le16_to_cpu(es->s_mnt_count) >=
  979. (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
  980. printk (KERN_WARNING
  981. "EXT3-fs warning: maximal mount count reached, "
  982. "running e2fsck is recommended\n");
  983. else if (le32_to_cpu(es->s_checkinterval) &&
  984. (le32_to_cpu(es->s_lastcheck) +
  985. le32_to_cpu(es->s_checkinterval) <= get_seconds()))
  986. printk (KERN_WARNING
  987. "EXT3-fs warning: checktime reached, "
  988. "running e2fsck is recommended\n");
  989. #if 0
  990. /* @@@ We _will_ want to clear the valid bit if we find
  991. inconsistencies, to force a fsck at reboot. But for
  992. a plain journaled filesystem we can keep it set as
  993. valid forever! :) */
  994. es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT3_VALID_FS);
  995. #endif
  996. if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
  997. es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
  998. es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
  999. es->s_mtime = cpu_to_le32(get_seconds());
  1000. ext3_update_dynamic_rev(sb);
  1001. EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
  1002. ext3_commit_super(sb, es, 1);
  1003. if (test_opt(sb, DEBUG))
  1004. printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
  1005. "bpg=%lu, ipg=%lu, mo=%04lx]\n",
  1006. sb->s_blocksize,
  1007. sbi->s_groups_count,
  1008. EXT3_BLOCKS_PER_GROUP(sb),
  1009. EXT3_INODES_PER_GROUP(sb),
  1010. sbi->s_mount_opt);
  1011. printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
  1012. if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
  1013. char b[BDEVNAME_SIZE];
  1014. printk("external journal on %s\n",
  1015. bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
  1016. } else {
  1017. printk("internal journal\n");
  1018. }
  1019. return res;
  1020. }
  1021. /* Called at mount-time, super-block is locked */
  1022. static int ext3_check_descriptors (struct super_block * sb)
  1023. {
  1024. struct ext3_sb_info *sbi = EXT3_SB(sb);
  1025. unsigned long block = le32_to_cpu(sbi->s_es->s_first_data_block);
  1026. struct ext3_group_desc * gdp = NULL;
  1027. int desc_block = 0;
  1028. int i;
  1029. ext3_debug ("Checking group descriptors");
  1030. for (i = 0; i < sbi->s_groups_count; i++)
  1031. {
  1032. if ((i % EXT3_DESC_PER_BLOCK(sb)) == 0)
  1033. gdp = (struct ext3_group_desc *)
  1034. sbi->s_group_desc[desc_block++]->b_data;
  1035. if (le32_to_cpu(gdp->bg_block_bitmap) < block ||
  1036. le32_to_cpu(gdp->bg_block_bitmap) >=
  1037. block + EXT3_BLOCKS_PER_GROUP(sb))
  1038. {
  1039. ext3_error (sb, "ext3_check_descriptors",
  1040. "Block bitmap for group %d"
  1041. " not in group (block %lu)!",
  1042. i, (unsigned long)
  1043. le32_to_cpu(gdp->bg_block_bitmap));
  1044. return 0;
  1045. }
  1046. if (le32_to_cpu(gdp->bg_inode_bitmap) < block ||
  1047. le32_to_cpu(gdp->bg_inode_bitmap) >=
  1048. block + EXT3_BLOCKS_PER_GROUP(sb))
  1049. {
  1050. ext3_error (sb, "ext3_check_descriptors",
  1051. "Inode bitmap for group %d"
  1052. " not in group (block %lu)!",
  1053. i, (unsigned long)
  1054. le32_to_cpu(gdp->bg_inode_bitmap));
  1055. return 0;
  1056. }
  1057. if (le32_to_cpu(gdp->bg_inode_table) < block ||
  1058. le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group >=
  1059. block + EXT3_BLOCKS_PER_GROUP(sb))
  1060. {
  1061. ext3_error (sb, "ext3_check_descriptors",
  1062. "Inode table for group %d"
  1063. " not in group (block %lu)!",
  1064. i, (unsigned long)
  1065. le32_to_cpu(gdp->bg_inode_table));
  1066. return 0;
  1067. }
  1068. block += EXT3_BLOCKS_PER_GROUP(sb);
  1069. gdp++;
  1070. }
  1071. sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
  1072. sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
  1073. return 1;
  1074. }
  1075. /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
  1076. * the superblock) which were deleted from all directories, but held open by
  1077. * a process at the time of a crash. We walk the list and try to delete these
  1078. * inodes at recovery time (only with a read-write filesystem).
  1079. *
  1080. * In order to keep the orphan inode chain consistent during traversal (in
  1081. * case of crash during recovery), we link each inode into the superblock
  1082. * orphan list_head and handle it the same way as an inode deletion during
  1083. * normal operation (which journals the operations for us).
  1084. *
  1085. * We only do an iget() and an iput() on each inode, which is very safe if we
  1086. * accidentally point at an in-use or already deleted inode. The worst that
  1087. * can happen in this case is that we get a "bit already cleared" message from
  1088. * ext3_free_inode(). The only reason we would point at a wrong inode is if
  1089. * e2fsck was run on this filesystem, and it must have already done the orphan
  1090. * inode cleanup for us, so we can safely abort without any further action.
  1091. */
  1092. static void ext3_orphan_cleanup (struct super_block * sb,
  1093. struct ext3_super_block * es)
  1094. {
  1095. unsigned int s_flags = sb->s_flags;
  1096. int nr_orphans = 0, nr_truncates = 0;
  1097. #ifdef CONFIG_QUOTA
  1098. int i;
  1099. #endif
  1100. if (!es->s_last_orphan) {
  1101. jbd_debug(4, "no orphan inodes to clean up\n");
  1102. return;
  1103. }
  1104. if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
  1105. if (es->s_last_orphan)
  1106. jbd_debug(1, "Errors on filesystem, "
  1107. "clearing orphan list.\n");
  1108. es->s_last_orphan = 0;
  1109. jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
  1110. return;
  1111. }
  1112. if (s_flags & MS_RDONLY) {
  1113. printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
  1114. sb->s_id);
  1115. sb->s_flags &= ~MS_RDONLY;
  1116. }
  1117. #ifdef CONFIG_QUOTA
  1118. /* Needed for iput() to work correctly and not trash data */
  1119. sb->s_flags |= MS_ACTIVE;
  1120. /* Turn on quotas so that they are updated correctly */
  1121. for (i = 0; i < MAXQUOTAS; i++) {
  1122. if (EXT3_SB(sb)->s_qf_names[i]) {
  1123. int ret = ext3_quota_on_mount(sb, i);
  1124. if (ret < 0)
  1125. printk(KERN_ERR
  1126. "EXT3-fs: Cannot turn on journalled "
  1127. "quota: error %d\n", ret);
  1128. }
  1129. }
  1130. #endif
  1131. while (es->s_last_orphan) {
  1132. struct inode *inode;
  1133. if (!(inode =
  1134. ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
  1135. es->s_last_orphan = 0;
  1136. break;
  1137. }
  1138. list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
  1139. DQUOT_INIT(inode);
  1140. if (inode->i_nlink) {
  1141. printk(KERN_DEBUG
  1142. "%s: truncating inode %ld to %Ld bytes\n",
  1143. __FUNCTION__, inode->i_ino, inode->i_size);
  1144. jbd_debug(2, "truncating inode %ld to %Ld bytes\n",
  1145. inode->i_ino, inode->i_size);
  1146. ext3_truncate(inode);
  1147. nr_truncates++;
  1148. } else {
  1149. printk(KERN_DEBUG
  1150. "%s: deleting unreferenced inode %ld\n",
  1151. __FUNCTION__, inode->i_ino);
  1152. jbd_debug(2, "deleting unreferenced inode %ld\n",
  1153. inode->i_ino);
  1154. nr_orphans++;
  1155. }
  1156. iput(inode); /* The delete magic happens here! */
  1157. }
  1158. #define PLURAL(x) (x), ((x)==1) ? "" : "s"
  1159. if (nr_orphans)
  1160. printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
  1161. sb->s_id, PLURAL(nr_orphans));
  1162. if (nr_truncates)
  1163. printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
  1164. sb->s_id, PLURAL(nr_truncates));
  1165. #ifdef CONFIG_QUOTA
  1166. /* Turn quotas off */
  1167. for (i = 0; i < MAXQUOTAS; i++) {
  1168. if (sb_dqopt(sb)->files[i])
  1169. vfs_quota_off(sb, i);
  1170. }
  1171. #endif
  1172. sb->s_flags = s_flags; /* Restore MS_RDONLY status */
  1173. }
  1174. #define log2(n) ffz(~(n))
  1175. /*
  1176. * Maximal file size. There is a direct, and {,double-,triple-}indirect
  1177. * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
  1178. * We need to be 1 filesystem block less than the 2^32 sector limit.
  1179. */
  1180. static loff_t ext3_max_size(int bits)
  1181. {
  1182. loff_t res = EXT3_NDIR_BLOCKS;
  1183. /* This constant is calculated to be the largest file size for a
  1184. * dense, 4k-blocksize file such that the total number of
  1185. * sectors in the file, including data and all indirect blocks,
  1186. * does not exceed 2^32. */
  1187. const loff_t upper_limit = 0x1ff7fffd000LL;
  1188. res += 1LL << (bits-2);
  1189. res += 1LL << (2*(bits-2));
  1190. res += 1LL << (3*(bits-2));
  1191. res <<= bits;
  1192. if (res > upper_limit)
  1193. res = upper_limit;
  1194. return res;
  1195. }
  1196. static unsigned long descriptor_loc(struct super_block *sb,
  1197. unsigned long logic_sb_block,
  1198. int nr)
  1199. {
  1200. struct ext3_sb_info *sbi = EXT3_SB(sb);
  1201. unsigned long bg, first_data_block, first_meta_bg;
  1202. int has_super = 0;
  1203. first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block);
  1204. first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
  1205. if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
  1206. nr < first_meta_bg)
  1207. return (logic_sb_block + nr + 1);
  1208. bg = sbi->s_desc_per_block * nr;
  1209. if (ext3_bg_has_super(sb, bg))
  1210. has_super = 1;
  1211. return (first_data_block + has_super + (bg * sbi->s_blocks_per_group));
  1212. }
  1213. static int ext3_fill_super (struct super_block *sb, void *data, int silent)
  1214. {
  1215. struct buffer_head * bh;
  1216. struct ext3_super_block *es = NULL;
  1217. struct ext3_sb_info *sbi;
  1218. unsigned long block;
  1219. unsigned long sb_block = get_sb_block(&data);
  1220. unsigned long logic_sb_block;
  1221. unsigned long offset = 0;
  1222. unsigned long journal_inum = 0;
  1223. unsigned long journal_devnum = 0;
  1224. unsigned long def_mount_opts;
  1225. struct inode *root;
  1226. int blocksize;
  1227. int hblock;
  1228. int db_count;
  1229. int i;
  1230. int needs_recovery;
  1231. __le32 features;
  1232. sbi = kmalloc(sizeof(*sbi), GFP_KERNEL);
  1233. if (!sbi)
  1234. return -ENOMEM;
  1235. sb->s_fs_info = sbi;
  1236. memset(sbi, 0, sizeof(*sbi));
  1237. sbi->s_mount_opt = 0;
  1238. sbi->s_resuid = EXT3_DEF_RESUID;
  1239. sbi->s_resgid = EXT3_DEF_RESGID;
  1240. unlock_kernel();
  1241. blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
  1242. if (!blocksize) {
  1243. printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
  1244. goto out_fail;
  1245. }
  1246. /*
  1247. * The ext3 superblock will not be buffer aligned for other than 1kB
  1248. * block sizes. We need to calculate the offset from buffer start.
  1249. */
  1250. if (blocksize != EXT3_MIN_BLOCK_SIZE) {
  1251. logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
  1252. offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
  1253. } else {
  1254. logic_sb_block = sb_block;
  1255. }
  1256. if (!(bh = sb_bread(sb, logic_sb_block))) {
  1257. printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
  1258. goto out_fail;
  1259. }
  1260. /*
  1261. * Note: s_es must be initialized as soon as possible because
  1262. * some ext3 macro-instructions depend on its value
  1263. */
  1264. es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
  1265. sbi->s_es = es;
  1266. sb->s_magic = le16_to_cpu(es->s_magic);
  1267. if (sb->s_magic != EXT3_SUPER_MAGIC)
  1268. goto cantfind_ext3;
  1269. /* Set defaults before we parse the mount options */
  1270. def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
  1271. if (def_mount_opts & EXT3_DEFM_DEBUG)
  1272. set_opt(sbi->s_mount_opt, DEBUG);
  1273. if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
  1274. set_opt(sbi->s_mount_opt, GRPID);
  1275. if (def_mount_opts & EXT3_DEFM_UID16)
  1276. set_opt(sbi->s_mount_opt, NO_UID32);
  1277. if (def_mount_opts & EXT3_DEFM_XATTR_USER)
  1278. set_opt(sbi->s_mount_opt, XATTR_USER);
  1279. if (def_mount_opts & EXT3_DEFM_ACL)
  1280. set_opt(sbi->s_mount_opt, POSIX_ACL);
  1281. if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
  1282. sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
  1283. else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
  1284. sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
  1285. else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
  1286. sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
  1287. if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
  1288. set_opt(sbi->s_mount_opt, ERRORS_PANIC);
  1289. else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO)
  1290. set_opt(sbi->s_mount_opt, ERRORS_RO);
  1291. sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
  1292. sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
  1293. set_opt(sbi->s_mount_opt, RESERVATION);
  1294. if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
  1295. NULL, 0))
  1296. goto failed_mount;
  1297. sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
  1298. ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
  1299. if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
  1300. (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
  1301. EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
  1302. EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
  1303. printk(KERN_WARNING
  1304. "EXT3-fs warning: feature flags set on rev 0 fs, "
  1305. "running e2fsck is recommended\n");
  1306. /*
  1307. * Check feature flags regardless of the revision level, since we
  1308. * previously didn't change the revision level when setting the flags,
  1309. * so there is a chance incompat flags are set on a rev 0 filesystem.
  1310. */
  1311. features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
  1312. if (features) {
  1313. printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
  1314. "unsupported optional features (%x).\n",
  1315. sb->s_id, le32_to_cpu(features));
  1316. goto failed_mount;
  1317. }
  1318. features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
  1319. if (!(sb->s_flags & MS_RDONLY) && features) {
  1320. printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
  1321. "unsupported optional features (%x).\n",
  1322. sb->s_id, le32_to_cpu(features));
  1323. goto failed_mount;
  1324. }
  1325. blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
  1326. if (blocksize < EXT3_MIN_BLOCK_SIZE ||
  1327. blocksize > EXT3_MAX_BLOCK_SIZE) {
  1328. printk(KERN_ERR
  1329. "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
  1330. blocksize, sb->s_id);
  1331. goto failed_mount;
  1332. }
  1333. hblock = bdev_hardsect_size(sb->s_bdev);
  1334. if (sb->s_blocksize != blocksize) {
  1335. /*
  1336. * Make sure the blocksize for the filesystem is larger
  1337. * than the hardware sectorsize for the machine.
  1338. */
  1339. if (blocksize < hblock) {
  1340. printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
  1341. "device blocksize %d.\n", blocksize, hblock);
  1342. goto failed_mount;
  1343. }
  1344. brelse (bh);
  1345. sb_set_blocksize(sb, blocksize);
  1346. logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
  1347. offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
  1348. bh = sb_bread(sb, logic_sb_block);
  1349. if (!bh) {
  1350. printk(KERN_ERR
  1351. "EXT3-fs: Can't read superblock on 2nd try.\n");
  1352. goto failed_mount;
  1353. }
  1354. es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
  1355. sbi->s_es = es;
  1356. if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
  1357. printk (KERN_ERR
  1358. "EXT3-fs: Magic mismatch, very weird !\n");
  1359. goto failed_mount;
  1360. }
  1361. }
  1362. sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
  1363. if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
  1364. sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
  1365. sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
  1366. } else {
  1367. sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
  1368. sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
  1369. if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
  1370. (sbi->s_inode_size & (sbi->s_inode_size - 1)) ||
  1371. (sbi->s_inode_size > blocksize)) {
  1372. printk (KERN_ERR
  1373. "EXT3-fs: unsupported inode size: %d\n",
  1374. sbi->s_inode_size);
  1375. goto failed_mount;
  1376. }
  1377. }
  1378. sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
  1379. le32_to_cpu(es->s_log_frag_size);
  1380. if (blocksize != sbi->s_frag_size) {
  1381. printk(KERN_ERR
  1382. "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
  1383. sbi->s_frag_size, blocksize);
  1384. goto failed_mount;
  1385. }
  1386. sbi->s_frags_per_block = 1;
  1387. sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
  1388. sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
  1389. sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
  1390. if (EXT3_INODE_SIZE(sb) == 0)
  1391. goto cantfind_ext3;
  1392. sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
  1393. if (sbi->s_inodes_per_block == 0)
  1394. goto cantfind_ext3;
  1395. sbi->s_itb_per_group = sbi->s_inodes_per_group /
  1396. sbi->s_inodes_per_block;
  1397. sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
  1398. sbi->s_sbh = bh;
  1399. sbi->s_mount_state = le16_to_cpu(es->s_state);
  1400. sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb));
  1401. sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb));
  1402. for (i=0; i < 4; i++)
  1403. sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
  1404. sbi->s_def_hash_version = es->s_def_hash_version;
  1405. if (sbi->s_blocks_per_group > blocksize * 8) {
  1406. printk (KERN_ERR
  1407. "EXT3-fs: #blocks per group too big: %lu\n",
  1408. sbi->s_blocks_per_group);
  1409. goto failed_mount;
  1410. }
  1411. if (sbi->s_frags_per_group > blocksize * 8) {
  1412. printk (KERN_ERR
  1413. "EXT3-fs: #fragments per group too big: %lu\n",
  1414. sbi->s_frags_per_group);
  1415. goto failed_mount;
  1416. }
  1417. if (sbi->s_inodes_per_group > blocksize * 8) {
  1418. printk (KERN_ERR
  1419. "EXT3-fs: #inodes per group too big: %lu\n",
  1420. sbi->s_inodes_per_group);
  1421. goto failed_mount;
  1422. }
  1423. if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
  1424. goto cantfind_ext3;
  1425. sbi->s_groups_count = (le32_to_cpu(es->s_blocks_count) -
  1426. le32_to_cpu(es->s_first_data_block) +
  1427. EXT3_BLOCKS_PER_GROUP(sb) - 1) /
  1428. EXT3_BLOCKS_PER_GROUP(sb);
  1429. db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
  1430. EXT3_DESC_PER_BLOCK(sb);
  1431. sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
  1432. GFP_KERNEL);
  1433. if (sbi->s_group_desc == NULL) {
  1434. printk (KERN_ERR "EXT3-fs: not enough memory\n");
  1435. goto failed_mount;
  1436. }
  1437. percpu_counter_init(&sbi->s_freeblocks_counter);
  1438. percpu_counter_init(&sbi->s_freeinodes_counter);
  1439. percpu_counter_init(&sbi->s_dirs_counter);
  1440. bgl_lock_init(&sbi->s_blockgroup_lock);
  1441. for (i = 0; i < db_count; i++) {
  1442. block = descriptor_loc(sb, logic_sb_block, i);
  1443. sbi->s_group_desc[i] = sb_bread(sb, block);
  1444. if (!sbi->s_group_desc[i]) {
  1445. printk (KERN_ERR "EXT3-fs: "
  1446. "can't read group descriptor %d\n", i);
  1447. db_count = i;
  1448. goto failed_mount2;
  1449. }
  1450. }
  1451. if (!ext3_check_descriptors (sb)) {
  1452. printk (KERN_ERR "EXT3-fs: group descriptors corrupted !\n");
  1453. goto failed_mount2;
  1454. }
  1455. sbi->s_gdb_count = db_count;
  1456. get_random_bytes(&sbi->s_next_generation, sizeof(u32));
  1457. spin_lock_init(&sbi->s_next_gen_lock);
  1458. /* per fileystem reservation list head & lock */
  1459. spin_lock_init(&sbi->s_rsv_window_lock);
  1460. sbi->s_rsv_window_root = RB_ROOT;
  1461. /* Add a single, static dummy reservation to the start of the
  1462. * reservation window list --- it gives us a placeholder for
  1463. * append-at-start-of-list which makes the allocation logic
  1464. * _much_ simpler. */
  1465. sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
  1466. sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
  1467. sbi->s_rsv_window_head.rsv_alloc_hit = 0;
  1468. sbi->s_rsv_window_head.rsv_goal_size = 0;
  1469. ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
  1470. /*
  1471. * set up enough so that it can read an inode
  1472. */
  1473. sb->s_op = &ext3_sops;
  1474. sb->s_export_op = &ext3_export_ops;
  1475. sb->s_xattr = ext3_xattr_handlers;
  1476. #ifdef CONFIG_QUOTA
  1477. sb->s_qcop = &ext3_qctl_operations;
  1478. sb->dq_op = &ext3_quota_operations;
  1479. #endif
  1480. INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
  1481. sb->s_root = NULL;
  1482. needs_recovery = (es->s_last_orphan != 0 ||
  1483. EXT3_HAS_INCOMPAT_FEATURE(sb,
  1484. EXT3_FEATURE_INCOMPAT_RECOVER));
  1485. /*
  1486. * The first inode we look at is the journal inode. Don't try
  1487. * root first: it may be modified in the journal!
  1488. */
  1489. if (!test_opt(sb, NOLOAD) &&
  1490. EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
  1491. if (ext3_load_journal(sb, es, journal_devnum))
  1492. goto failed_mount2;
  1493. } else if (journal_inum) {
  1494. if (ext3_create_journal(sb, es, journal_inum))
  1495. goto failed_mount2;
  1496. } else {
  1497. if (!silent)
  1498. printk (KERN_ERR
  1499. "ext3: No journal on filesystem on %s\n",
  1500. sb->s_id);
  1501. goto failed_mount2;
  1502. }
  1503. /* We have now updated the journal if required, so we can
  1504. * validate the data journaling mode. */
  1505. switch (test_opt(sb, DATA_FLAGS)) {
  1506. case 0:
  1507. /* No mode set, assume a default based on the journal
  1508. capabilities: ORDERED_DATA if the journal can
  1509. cope, else JOURNAL_DATA */
  1510. if (journal_check_available_features
  1511. (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
  1512. set_opt(sbi->s_mount_opt, ORDERED_DATA);
  1513. else
  1514. set_opt(sbi->s_mount_opt, JOURNAL_DATA);
  1515. break;
  1516. case EXT3_MOUNT_ORDERED_DATA:
  1517. case EXT3_MOUNT_WRITEBACK_DATA:
  1518. if (!journal_check_available_features
  1519. (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
  1520. printk(KERN_ERR "EXT3-fs: Journal does not support "
  1521. "requested data journaling mode\n");
  1522. goto failed_mount3;
  1523. }
  1524. default:
  1525. break;
  1526. }
  1527. if (test_opt(sb, NOBH)) {
  1528. if (sb->s_blocksize_bits != PAGE_CACHE_SHIFT) {
  1529. printk(KERN_WARNING "EXT3-fs: Ignoring nobh option "
  1530. "since filesystem blocksize doesn't match "
  1531. "pagesize\n");
  1532. clear_opt(sbi->s_mount_opt, NOBH);
  1533. }
  1534. if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
  1535. printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
  1536. "its supported only with writeback mode\n");
  1537. clear_opt(sbi->s_mount_opt, NOBH);
  1538. }
  1539. }
  1540. /*
  1541. * The journal_load will have done any necessary log recovery,
  1542. * so we can safely mount the rest of the filesystem now.
  1543. */
  1544. root = iget(sb, EXT3_ROOT_INO);
  1545. sb->s_root = d_alloc_root(root);
  1546. if (!sb->s_root) {
  1547. printk(KERN_ERR "EXT3-fs: get root inode failed\n");
  1548. iput(root);
  1549. goto failed_mount3;
  1550. }
  1551. if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
  1552. dput(sb->s_root);
  1553. sb->s_root = NULL;
  1554. printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
  1555. goto failed_mount3;
  1556. }
  1557. ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
  1558. /*
  1559. * akpm: core read_super() calls in here with the superblock locked.
  1560. * That deadlocks, because orphan cleanup needs to lock the superblock
  1561. * in numerous places. Here we just pop the lock - it's relatively
  1562. * harmless, because we are now ready to accept write_super() requests,
  1563. * and aviro says that's the only reason for hanging onto the
  1564. * superblock lock.
  1565. */
  1566. EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
  1567. ext3_orphan_cleanup(sb, es);
  1568. EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
  1569. if (needs_recovery)
  1570. printk (KERN_INFO "EXT3-fs: recovery complete.\n");
  1571. ext3_mark_recovery_complete(sb, es);
  1572. printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
  1573. test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
  1574. test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
  1575. "writeback");
  1576. percpu_counter_mod(&sbi->s_freeblocks_counter,
  1577. ext3_count_free_blocks(sb));
  1578. percpu_counter_mod(&sbi->s_freeinodes_counter,
  1579. ext3_count_free_inodes(sb));
  1580. percpu_counter_mod(&sbi->s_dirs_counter,
  1581. ext3_count_dirs(sb));
  1582. lock_kernel();
  1583. return 0;
  1584. cantfind_ext3:
  1585. if (!silent)
  1586. printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
  1587. sb->s_id);
  1588. goto failed_mount;
  1589. failed_mount3:
  1590. journal_destroy(sbi->s_journal);
  1591. failed_mount2:
  1592. for (i = 0; i < db_count; i++)
  1593. brelse(sbi->s_group_desc[i]);
  1594. kfree(sbi->s_group_desc);
  1595. failed_mount:
  1596. #ifdef CONFIG_QUOTA
  1597. for (i = 0; i < MAXQUOTAS; i++)
  1598. kfree(sbi->s_qf_names[i]);
  1599. #endif
  1600. ext3_blkdev_remove(sbi);
  1601. brelse(bh);
  1602. out_fail:
  1603. sb->s_fs_info = NULL;
  1604. kfree(sbi);
  1605. lock_kernel();
  1606. return -EINVAL;
  1607. }
  1608. /*
  1609. * Setup any per-fs journal parameters now. We'll do this both on
  1610. * initial mount, once the journal has been initialised but before we've
  1611. * done any recovery; and again on any subsequent remount.
  1612. */
  1613. static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
  1614. {
  1615. struct ext3_sb_info *sbi = EXT3_SB(sb);
  1616. if (sbi->s_commit_interval)
  1617. journal->j_commit_interval = sbi->s_commit_interval;
  1618. /* We could also set up an ext3-specific default for the commit
  1619. * interval here, but for now we'll just fall back to the jbd
  1620. * default. */
  1621. spin_lock(&journal->j_state_lock);
  1622. if (test_opt(sb, BARRIER))
  1623. journal->j_flags |= JFS_BARRIER;
  1624. else
  1625. journal->j_flags &= ~JFS_BARRIER;
  1626. spin_unlock(&journal->j_state_lock);
  1627. }
  1628. static journal_t *ext3_get_journal(struct super_block *sb, int journal_inum)
  1629. {
  1630. struct inode *journal_inode;
  1631. journal_t *journal;
  1632. /* First, test for the existence of a valid inode on disk. Bad
  1633. * things happen if we iget() an unused inode, as the subsequent
  1634. * iput() will try to delete it. */
  1635. journal_inode = iget(sb, journal_inum);
  1636. if (!journal_inode) {
  1637. printk(KERN_ERR "EXT3-fs: no journal found.\n");
  1638. return NULL;
  1639. }
  1640. if (!journal_inode->i_nlink) {
  1641. make_bad_inode(journal_inode);
  1642. iput(journal_inode);
  1643. printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
  1644. return NULL;
  1645. }
  1646. jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
  1647. journal_inode, journal_inode->i_size);
  1648. if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
  1649. printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
  1650. iput(journal_inode);
  1651. return NULL;
  1652. }
  1653. journal = journal_init_inode(journal_inode);
  1654. if (!journal) {
  1655. printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
  1656. iput(journal_inode);
  1657. return NULL;
  1658. }
  1659. journal->j_private = sb;
  1660. ext3_init_journal_params(sb, journal);
  1661. return journal;
  1662. }
  1663. static journal_t *ext3_get_dev_journal(struct super_block *sb,
  1664. dev_t j_dev)
  1665. {
  1666. struct buffer_head * bh;
  1667. journal_t *journal;
  1668. int start;
  1669. int len;
  1670. int hblock, blocksize;
  1671. unsigned long sb_block;
  1672. unsigned long offset;
  1673. struct ext3_super_block * es;
  1674. struct block_device *bdev;
  1675. bdev = ext3_blkdev_get(j_dev);
  1676. if (bdev == NULL)
  1677. return NULL;
  1678. if (bd_claim(bdev, sb)) {
  1679. printk(KERN_ERR
  1680. "EXT3: failed to claim external journal device.\n");
  1681. blkdev_put(bdev);
  1682. return NULL;
  1683. }
  1684. blocksize = sb->s_blocksize;
  1685. hblock = bdev_hardsect_size(bdev);
  1686. if (blocksize < hblock) {
  1687. printk(KERN_ERR
  1688. "EXT3-fs: blocksize too small for journal device.\n");
  1689. goto out_bdev;
  1690. }
  1691. sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
  1692. offset = EXT3_MIN_BLOCK_SIZE % blocksize;
  1693. set_blocksize(bdev, blocksize);
  1694. if (!(bh = __bread(bdev, sb_block, blocksize))) {
  1695. printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
  1696. "external journal\n");
  1697. goto out_bdev;
  1698. }
  1699. es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
  1700. if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
  1701. !(le32_to_cpu(es->s_feature_incompat) &
  1702. EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
  1703. printk(KERN_ERR "EXT3-fs: external journal has "
  1704. "bad superblock\n");
  1705. brelse(bh);
  1706. goto out_bdev;
  1707. }
  1708. if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
  1709. printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
  1710. brelse(bh);
  1711. goto out_bdev;
  1712. }
  1713. len = le32_to_cpu(es->s_blocks_count);
  1714. start = sb_block + 1;
  1715. brelse(bh); /* we're done with the superblock */
  1716. journal = journal_init_dev(bdev, sb->s_bdev,
  1717. start, len, blocksize);
  1718. if (!journal) {
  1719. printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
  1720. goto out_bdev;
  1721. }
  1722. journal->j_private = sb;
  1723. ll_rw_block(READ, 1, &journal->j_sb_buffer);
  1724. wait_on_buffer(journal->j_sb_buffer);
  1725. if (!buffer_uptodate(journal->j_sb_buffer)) {
  1726. printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
  1727. goto out_journal;
  1728. }
  1729. if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
  1730. printk(KERN_ERR "EXT3-fs: External journal has more than one "
  1731. "user (unsupported) - %d\n",
  1732. be32_to_cpu(journal->j_superblock->s_nr_users));
  1733. goto out_journal;
  1734. }
  1735. EXT3_SB(sb)->journal_bdev = bdev;
  1736. ext3_init_journal_params(sb, journal);
  1737. return journal;
  1738. out_journal:
  1739. journal_destroy(journal);
  1740. out_bdev:
  1741. ext3_blkdev_put(bdev);
  1742. return NULL;
  1743. }
  1744. static int ext3_load_journal(struct super_block *sb,
  1745. struct ext3_super_block *es,
  1746. unsigned long journal_devnum)
  1747. {
  1748. journal_t *journal;
  1749. int journal_inum = le32_to_cpu(es->s_journal_inum);
  1750. dev_t journal_dev;
  1751. int err = 0;
  1752. int really_read_only;
  1753. if (journal_devnum &&
  1754. journal_devnum != le32_to_cpu(es->s_journal_dev)) {
  1755. printk(KERN_INFO "EXT3-fs: external journal device major/minor "
  1756. "numbers have changed\n");
  1757. journal_dev = new_decode_dev(journal_devnum);
  1758. } else
  1759. journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
  1760. really_read_only = bdev_read_only(sb->s_bdev);
  1761. /*
  1762. * Are we loading a blank journal or performing recovery after a
  1763. * crash? For recovery, we need to check in advance whether we
  1764. * can get read-write access to the device.
  1765. */
  1766. if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
  1767. if (sb->s_flags & MS_RDONLY) {
  1768. printk(KERN_INFO "EXT3-fs: INFO: recovery "
  1769. "required on readonly filesystem.\n");
  1770. if (really_read_only) {
  1771. printk(KERN_ERR "EXT3-fs: write access "
  1772. "unavailable, cannot proceed.\n");
  1773. return -EROFS;
  1774. }
  1775. printk (KERN_INFO "EXT3-fs: write access will "
  1776. "be enabled during recovery.\n");
  1777. }
  1778. }
  1779. if (journal_inum && journal_dev) {
  1780. printk(KERN_ERR "EXT3-fs: filesystem has both journal "
  1781. "and inode journals!\n");
  1782. return -EINVAL;
  1783. }
  1784. if (journal_inum) {
  1785. if (!(journal = ext3_get_journal(sb, journal_inum)))
  1786. return -EINVAL;
  1787. } else {
  1788. if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
  1789. return -EINVAL;
  1790. }
  1791. if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
  1792. err = journal_update_format(journal);
  1793. if (err) {
  1794. printk(KERN_ERR "EXT3-fs: error updating journal.\n");
  1795. journal_destroy(journal);
  1796. return err;
  1797. }
  1798. }
  1799. if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
  1800. err = journal_wipe(journal, !really_read_only);
  1801. if (!err)
  1802. err = journal_load(journal);
  1803. if (err) {
  1804. printk(KERN_ERR "EXT3-fs: error loading journal.\n");
  1805. journal_destroy(journal);
  1806. return err;
  1807. }
  1808. EXT3_SB(sb)->s_journal = journal;
  1809. ext3_clear_journal_err(sb, es);
  1810. if (journal_devnum &&
  1811. journal_devnum != le32_to_cpu(es->s_journal_dev)) {
  1812. es->s_journal_dev = cpu_to_le32(journal_devnum);
  1813. sb->s_dirt = 1;
  1814. /* Make sure we flush the recovery flag to disk. */
  1815. ext3_commit_super(sb, es, 1);
  1816. }
  1817. return 0;
  1818. }
  1819. static int ext3_create_journal(struct super_block * sb,
  1820. struct ext3_super_block * es,
  1821. int journal_inum)
  1822. {
  1823. journal_t *journal;
  1824. if (sb->s_flags & MS_RDONLY) {
  1825. printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
  1826. "create journal.\n");
  1827. return -EROFS;
  1828. }
  1829. if (!(journal = ext3_get_journal(sb, journal_inum)))
  1830. return -EINVAL;
  1831. printk(KERN_INFO "EXT3-fs: creating new journal on inode %d\n",
  1832. journal_inum);
  1833. if (journal_create(journal)) {
  1834. printk(KERN_ERR "EXT3-fs: error creating journal.\n");
  1835. journal_destroy(journal);
  1836. return -EIO;
  1837. }
  1838. EXT3_SB(sb)->s_journal = journal;
  1839. ext3_update_dynamic_rev(sb);
  1840. EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
  1841. EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
  1842. es->s_journal_inum = cpu_to_le32(journal_inum);
  1843. sb->s_dirt = 1;
  1844. /* Make sure we flush the recovery flag to disk. */
  1845. ext3_commit_super(sb, es, 1);
  1846. return 0;
  1847. }
  1848. static void ext3_commit_super (struct super_block * sb,
  1849. struct ext3_super_block * es,
  1850. int sync)
  1851. {
  1852. struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
  1853. if (!sbh)
  1854. return;
  1855. es->s_wtime = cpu_to_le32(get_seconds());
  1856. es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
  1857. es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
  1858. BUFFER_TRACE(sbh, "marking dirty");
  1859. mark_buffer_dirty(sbh);
  1860. if (sync)
  1861. sync_dirty_buffer(sbh);
  1862. }
  1863. /*
  1864. * Have we just finished recovery? If so, and if we are mounting (or
  1865. * remounting) the filesystem readonly, then we will end up with a
  1866. * consistent fs on disk. Record that fact.
  1867. */
  1868. static void ext3_mark_recovery_complete(struct super_block * sb,
  1869. struct ext3_super_block * es)
  1870. {
  1871. journal_t *journal = EXT3_SB(sb)->s_journal;
  1872. journal_lock_updates(journal);
  1873. journal_flush(journal);
  1874. if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
  1875. sb->s_flags & MS_RDONLY) {
  1876. EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
  1877. sb->s_dirt = 0;
  1878. ext3_commit_super(sb, es, 1);
  1879. }
  1880. journal_unlock_updates(journal);
  1881. }
  1882. /*
  1883. * If we are mounting (or read-write remounting) a filesystem whose journal
  1884. * has recorded an error from a previous lifetime, move that error to the
  1885. * main filesystem now.
  1886. */
  1887. static void ext3_clear_journal_err(struct super_block * sb,
  1888. struct ext3_super_block * es)
  1889. {
  1890. journal_t *journal;
  1891. int j_errno;
  1892. const char *errstr;
  1893. journal = EXT3_SB(sb)->s_journal;
  1894. /*
  1895. * Now check for any error status which may have been recorded in the
  1896. * journal by a prior ext3_error() or ext3_abort()
  1897. */
  1898. j_errno = journal_errno(journal);
  1899. if (j_errno) {
  1900. char nbuf[16];
  1901. errstr = ext3_decode_error(sb, j_errno, nbuf);
  1902. ext3_warning(sb, __FUNCTION__, "Filesystem error recorded "
  1903. "from previous mount: %s", errstr);
  1904. ext3_warning(sb, __FUNCTION__, "Marking fs in need of "
  1905. "filesystem check.");
  1906. EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
  1907. es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
  1908. ext3_commit_super (sb, es, 1);
  1909. journal_clear_err(journal);
  1910. }
  1911. }
  1912. /*
  1913. * Force the running and committing transactions to commit,
  1914. * and wait on the commit.
  1915. */
  1916. int ext3_force_commit(struct super_block *sb)
  1917. {
  1918. journal_t *journal;
  1919. int ret;
  1920. if (sb->s_flags & MS_RDONLY)
  1921. return 0;
  1922. journal = EXT3_SB(sb)->s_journal;
  1923. sb->s_dirt = 0;
  1924. ret = ext3_journal_force_commit(journal);
  1925. return ret;
  1926. }
  1927. /*
  1928. * Ext3 always journals updates to the superblock itself, so we don't
  1929. * have to propagate any other updates to the superblock on disk at this
  1930. * point. Just start an async writeback to get the buffers on their way
  1931. * to the disk.
  1932. *
  1933. * This implicitly triggers the writebehind on sync().
  1934. */
  1935. static void ext3_write_super (struct super_block * sb)
  1936. {
  1937. if (mutex_trylock(&sb->s_lock) != 0)
  1938. BUG();
  1939. sb->s_dirt = 0;
  1940. }
  1941. static int ext3_sync_fs(struct super_block *sb, int wait)
  1942. {
  1943. tid_t target;
  1944. sb->s_dirt = 0;
  1945. if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
  1946. if (wait)
  1947. log_wait_commit(EXT3_SB(sb)->s_journal, target);
  1948. }
  1949. return 0;
  1950. }
  1951. /*
  1952. * LVM calls this function before a (read-only) snapshot is created. This
  1953. * gives us a chance to flush the journal completely and mark the fs clean.
  1954. */
  1955. static void ext3_write_super_lockfs(struct super_block *sb)
  1956. {
  1957. sb->s_dirt = 0;
  1958. if (!(sb->s_flags & MS_RDONLY)) {
  1959. journal_t *journal = EXT3_SB(sb)->s_journal;
  1960. /* Now we set up the journal barrier. */
  1961. journal_lock_updates(journal);
  1962. journal_flush(journal);
  1963. /* Journal blocked and flushed, clear needs_recovery flag. */
  1964. EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
  1965. ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
  1966. }
  1967. }
  1968. /*
  1969. * Called by LVM after the snapshot is done. We need to reset the RECOVER
  1970. * flag here, even though the filesystem is not technically dirty yet.
  1971. */
  1972. static void ext3_unlockfs(struct super_block *sb)
  1973. {
  1974. if (!(sb->s_flags & MS_RDONLY)) {
  1975. lock_super(sb);
  1976. /* Reser the needs_recovery flag before the fs is unlocked. */
  1977. EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
  1978. ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
  1979. unlock_super(sb);
  1980. journal_unlock_updates(EXT3_SB(sb)->s_journal);
  1981. }
  1982. }
  1983. static int ext3_remount (struct super_block * sb, int * flags, char * data)
  1984. {
  1985. struct ext3_super_block * es;
  1986. struct ext3_sb_info *sbi = EXT3_SB(sb);
  1987. unsigned long n_blocks_count = 0;
  1988. unsigned long old_sb_flags;
  1989. struct ext3_mount_options old_opts;
  1990. int err;
  1991. #ifdef CONFIG_QUOTA
  1992. int i;
  1993. #endif
  1994. /* Store the original options */
  1995. old_sb_flags = sb->s_flags;
  1996. old_opts.s_mount_opt = sbi->s_mount_opt;
  1997. old_opts.s_resuid = sbi->s_resuid;
  1998. old_opts.s_resgid = sbi->s_resgid;
  1999. old_opts.s_commit_interval = sbi->s_commit_interval;
  2000. #ifdef CONFIG_QUOTA
  2001. old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
  2002. for (i = 0; i < MAXQUOTAS; i++)
  2003. old_opts.s_qf_names[i] = sbi->s_qf_names[i];
  2004. #endif
  2005. /*
  2006. * Allow the "check" option to be passed as a remount option.
  2007. */
  2008. if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
  2009. err = -EINVAL;
  2010. goto restore_opts;
  2011. }
  2012. if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
  2013. ext3_abort(sb, __FUNCTION__, "Abort forced by user");
  2014. sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
  2015. ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
  2016. es = sbi->s_es;
  2017. ext3_init_journal_params(sb, sbi->s_journal);
  2018. if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
  2019. n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
  2020. if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
  2021. err = -EROFS;
  2022. goto restore_opts;
  2023. }
  2024. if (*flags & MS_RDONLY) {
  2025. /*
  2026. * First of all, the unconditional stuff we have to do
  2027. * to disable replay of the journal when we next remount
  2028. */
  2029. sb->s_flags |= MS_RDONLY;
  2030. /*
  2031. * OK, test if we are remounting a valid rw partition
  2032. * readonly, and if so set the rdonly flag and then
  2033. * mark the partition as valid again.
  2034. */
  2035. if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
  2036. (sbi->s_mount_state & EXT3_VALID_FS))
  2037. es->s_state = cpu_to_le16(sbi->s_mount_state);
  2038. ext3_mark_recovery_complete(sb, es);
  2039. } else {
  2040. __le32 ret;
  2041. if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
  2042. ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
  2043. printk(KERN_WARNING "EXT3-fs: %s: couldn't "
  2044. "remount RDWR because of unsupported "
  2045. "optional features (%x).\n",
  2046. sb->s_id, le32_to_cpu(ret));
  2047. err = -EROFS;
  2048. goto restore_opts;
  2049. }
  2050. /*
  2051. * Mounting a RDONLY partition read-write, so reread
  2052. * and store the current valid flag. (It may have
  2053. * been changed by e2fsck since we originally mounted
  2054. * the partition.)
  2055. */
  2056. ext3_clear_journal_err(sb, es);
  2057. sbi->s_mount_state = le16_to_cpu(es->s_state);
  2058. if ((ret = ext3_group_extend(sb, es, n_blocks_count))) {
  2059. err = ret;
  2060. goto restore_opts;
  2061. }
  2062. if (!ext3_setup_super (sb, es, 0))
  2063. sb->s_flags &= ~MS_RDONLY;
  2064. }
  2065. }
  2066. #ifdef CONFIG_QUOTA
  2067. /* Release old quota file names */
  2068. for (i = 0; i < MAXQUOTAS; i++)
  2069. if (old_opts.s_qf_names[i] &&
  2070. old_opts.s_qf_names[i] != sbi->s_qf_names[i])
  2071. kfree(old_opts.s_qf_names[i]);
  2072. #endif
  2073. return 0;
  2074. restore_opts:
  2075. sb->s_flags = old_sb_flags;
  2076. sbi->s_mount_opt = old_opts.s_mount_opt;
  2077. sbi->s_resuid = old_opts.s_resuid;
  2078. sbi->s_resgid = old_opts.s_resgid;
  2079. sbi->s_commit_interval = old_opts.s_commit_interval;
  2080. #ifdef CONFIG_QUOTA
  2081. sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
  2082. for (i = 0; i < MAXQUOTAS; i++) {
  2083. if (sbi->s_qf_names[i] &&
  2084. old_opts.s_qf_names[i] != sbi->s_qf_names[i])
  2085. kfree(sbi->s_qf_names[i]);
  2086. sbi->s_qf_names[i] = old_opts.s_qf_names[i];
  2087. }
  2088. #endif
  2089. return err;
  2090. }
  2091. static int ext3_statfs (struct super_block * sb, struct kstatfs * buf)
  2092. {
  2093. struct ext3_sb_info *sbi = EXT3_SB(sb);
  2094. struct ext3_super_block *es = sbi->s_es;
  2095. unsigned long overhead;
  2096. int i;
  2097. if (test_opt (sb, MINIX_DF))
  2098. overhead = 0;
  2099. else {
  2100. unsigned long ngroups;
  2101. ngroups = EXT3_SB(sb)->s_groups_count;
  2102. smp_rmb();
  2103. /*
  2104. * Compute the overhead (FS structures)
  2105. */
  2106. /*
  2107. * All of the blocks before first_data_block are
  2108. * overhead
  2109. */
  2110. overhead = le32_to_cpu(es->s_first_data_block);
  2111. /*
  2112. * Add the overhead attributed to the superblock and
  2113. * block group descriptors. If the sparse superblocks
  2114. * feature is turned on, then not all groups have this.
  2115. */
  2116. for (i = 0; i < ngroups; i++) {
  2117. overhead += ext3_bg_has_super(sb, i) +
  2118. ext3_bg_num_gdb(sb, i);
  2119. cond_resched();
  2120. }
  2121. /*
  2122. * Every block group has an inode bitmap, a block
  2123. * bitmap, and an inode table.
  2124. */
  2125. overhead += (ngroups * (2 + EXT3_SB(sb)->s_itb_per_group));
  2126. }
  2127. buf->f_type = EXT3_SUPER_MAGIC;
  2128. buf->f_bsize = sb->s_blocksize;
  2129. buf->f_blocks = le32_to_cpu(es->s_blocks_count) - overhead;
  2130. buf->f_bfree = percpu_counter_sum(&sbi->s_freeblocks_counter);
  2131. buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
  2132. if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
  2133. buf->f_bavail = 0;
  2134. buf->f_files = le32_to_cpu(es->s_inodes_count);
  2135. buf->f_ffree = percpu_counter_sum(&sbi->s_freeinodes_counter);
  2136. buf->f_namelen = EXT3_NAME_LEN;
  2137. return 0;
  2138. }
  2139. /* Helper function for writing quotas on sync - we need to start transaction before quota file
  2140. * is locked for write. Otherwise the are possible deadlocks:
  2141. * Process 1 Process 2
  2142. * ext3_create() quota_sync()
  2143. * journal_start() write_dquot()
  2144. * DQUOT_INIT() down(dqio_mutex)
  2145. * down(dqio_mutex) journal_start()
  2146. *
  2147. */
  2148. #ifdef CONFIG_QUOTA
  2149. static inline struct inode *dquot_to_inode(struct dquot *dquot)
  2150. {
  2151. return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
  2152. }
  2153. static int ext3_dquot_initialize(struct inode *inode, int type)
  2154. {
  2155. handle_t *handle;
  2156. int ret, err;
  2157. /* We may create quota structure so we need to reserve enough blocks */
  2158. handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
  2159. if (IS_ERR(handle))
  2160. return PTR_ERR(handle);
  2161. ret = dquot_initialize(inode, type);
  2162. err = ext3_journal_stop(handle);
  2163. if (!ret)
  2164. ret = err;
  2165. return ret;
  2166. }
  2167. static int ext3_dquot_drop(struct inode *inode)
  2168. {
  2169. handle_t *handle;
  2170. int ret, err;
  2171. /* We may delete quota structure so we need to reserve enough blocks */
  2172. handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
  2173. if (IS_ERR(handle))
  2174. return PTR_ERR(handle);
  2175. ret = dquot_drop(inode);
  2176. err = ext3_journal_stop(handle);
  2177. if (!ret)
  2178. ret = err;
  2179. return ret;
  2180. }
  2181. static int ext3_write_dquot(struct dquot *dquot)
  2182. {
  2183. int ret, err;
  2184. handle_t *handle;
  2185. struct inode *inode;
  2186. inode = dquot_to_inode(dquot);
  2187. handle = ext3_journal_start(inode,
  2188. EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
  2189. if (IS_ERR(handle))
  2190. return PTR_ERR(handle);
  2191. ret = dquot_commit(dquot);
  2192. err = ext3_journal_stop(handle);
  2193. if (!ret)
  2194. ret = err;
  2195. return ret;
  2196. }
  2197. static int ext3_acquire_dquot(struct dquot *dquot)
  2198. {
  2199. int ret, err;
  2200. handle_t *handle;
  2201. handle = ext3_journal_start(dquot_to_inode(dquot),
  2202. EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
  2203. if (IS_ERR(handle))
  2204. return PTR_ERR(handle);
  2205. ret = dquot_acquire(dquot);
  2206. err = ext3_journal_stop(handle);
  2207. if (!ret)
  2208. ret = err;
  2209. return ret;
  2210. }
  2211. static int ext3_release_dquot(struct dquot *dquot)
  2212. {
  2213. int ret, err;
  2214. handle_t *handle;
  2215. handle = ext3_journal_start(dquot_to_inode(dquot),
  2216. EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
  2217. if (IS_ERR(handle))
  2218. return PTR_ERR(handle);
  2219. ret = dquot_release(dquot);
  2220. err = ext3_journal_stop(handle);
  2221. if (!ret)
  2222. ret = err;
  2223. return ret;
  2224. }
  2225. static int ext3_mark_dquot_dirty(struct dquot *dquot)
  2226. {
  2227. /* Are we journalling quotas? */
  2228. if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
  2229. EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
  2230. dquot_mark_dquot_dirty(dquot);
  2231. return ext3_write_dquot(dquot);
  2232. } else {
  2233. return dquot_mark_dquot_dirty(dquot);
  2234. }
  2235. }
  2236. static int ext3_write_info(struct super_block *sb, int type)
  2237. {
  2238. int ret, err;
  2239. handle_t *handle;
  2240. /* Data block + inode block */
  2241. handle = ext3_journal_start(sb->s_root->d_inode, 2);
  2242. if (IS_ERR(handle))
  2243. return PTR_ERR(handle);
  2244. ret = dquot_commit_info(sb, type);
  2245. err = ext3_journal_stop(handle);
  2246. if (!ret)
  2247. ret = err;
  2248. return ret;
  2249. }
  2250. /*
  2251. * Turn on quotas during mount time - we need to find
  2252. * the quota file and such...
  2253. */
  2254. static int ext3_quota_on_mount(struct super_block *sb, int type)
  2255. {
  2256. return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
  2257. EXT3_SB(sb)->s_jquota_fmt, type);
  2258. }
  2259. /*
  2260. * Standard function to be called on quota_on
  2261. */
  2262. static int ext3_quota_on(struct super_block *sb, int type, int format_id,
  2263. char *path)
  2264. {
  2265. int err;
  2266. struct nameidata nd;
  2267. if (!test_opt(sb, QUOTA))
  2268. return -EINVAL;
  2269. /* Not journalling quota? */
  2270. if (!EXT3_SB(sb)->s_qf_names[USRQUOTA] &&
  2271. !EXT3_SB(sb)->s_qf_names[GRPQUOTA])
  2272. return vfs_quota_on(sb, type, format_id, path);
  2273. err = path_lookup(path, LOOKUP_FOLLOW, &nd);
  2274. if (err)
  2275. return err;
  2276. /* Quotafile not on the same filesystem? */
  2277. if (nd.mnt->mnt_sb != sb) {
  2278. path_release(&nd);
  2279. return -EXDEV;
  2280. }
  2281. /* Quotafile not of fs root? */
  2282. if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
  2283. printk(KERN_WARNING
  2284. "EXT3-fs: Quota file not on filesystem root. "
  2285. "Journalled quota will not work.\n");
  2286. path_release(&nd);
  2287. return vfs_quota_on(sb, type, format_id, path);
  2288. }
  2289. /* Read data from quotafile - avoid pagecache and such because we cannot afford
  2290. * acquiring the locks... As quota files are never truncated and quota code
  2291. * itself serializes the operations (and noone else should touch the files)
  2292. * we don't have to be afraid of races */
  2293. static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
  2294. size_t len, loff_t off)
  2295. {
  2296. struct inode *inode = sb_dqopt(sb)->files[type];
  2297. sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
  2298. int err = 0;
  2299. int offset = off & (sb->s_blocksize - 1);
  2300. int tocopy;
  2301. size_t toread;
  2302. struct buffer_head *bh;
  2303. loff_t i_size = i_size_read(inode);
  2304. if (off > i_size)
  2305. return 0;
  2306. if (off+len > i_size)
  2307. len = i_size-off;
  2308. toread = len;
  2309. while (toread > 0) {
  2310. tocopy = sb->s_blocksize - offset < toread ?
  2311. sb->s_blocksize - offset : toread;
  2312. bh = ext3_bread(NULL, inode, blk, 0, &err);
  2313. if (err)
  2314. return err;
  2315. if (!bh) /* A hole? */
  2316. memset(data, 0, tocopy);
  2317. else
  2318. memcpy(data, bh->b_data+offset, tocopy);
  2319. brelse(bh);
  2320. offset = 0;
  2321. toread -= tocopy;
  2322. data += tocopy;
  2323. blk++;
  2324. }
  2325. return len;
  2326. }
  2327. /* Write to quotafile (we know the transaction is already started and has
  2328. * enough credits) */
  2329. static ssize_t ext3_quota_write(struct super_block *sb, int type,
  2330. const char *data, size_t len, loff_t off)
  2331. {
  2332. struct inode *inode = sb_dqopt(sb)->files[type];
  2333. sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
  2334. int err = 0;
  2335. int offset = off & (sb->s_blocksize - 1);
  2336. int tocopy;
  2337. int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
  2338. size_t towrite = len;
  2339. struct buffer_head *bh;
  2340. handle_t *handle = journal_current_handle();
  2341. mutex_lock(&inode->i_mutex);
  2342. while (towrite > 0) {
  2343. tocopy = sb->s_blocksize - offset < towrite ?
  2344. sb->s_blocksize - offset : towrite;
  2345. bh = ext3_bread(handle, inode, blk, 1, &err);
  2346. if (!bh)
  2347. goto out;
  2348. if (journal_quota) {
  2349. err = ext3_journal_get_write_access(handle, bh);
  2350. if (err) {
  2351. brelse(bh);
  2352. goto out;
  2353. }
  2354. }
  2355. lock_buffer(bh);
  2356. memcpy(bh->b_data+offset, data, tocopy);
  2357. flush_dcache_page(bh->b_page);
  2358. unlock_buffer(bh);
  2359. if (journal_quota)
  2360. err = ext3_journal_dirty_metadata(handle, bh);
  2361. else {
  2362. /* Always do at least ordered writes for quotas */
  2363. err = ext3_journal_dirty_data(handle, bh);
  2364. mark_buffer_dirty(bh);
  2365. }
  2366. brelse(bh);
  2367. if (err)
  2368. goto out;
  2369. offset = 0;
  2370. towrite -= tocopy;
  2371. data += tocopy;
  2372. blk++;
  2373. }
  2374. out:
  2375. if (len == towrite)
  2376. return err;
  2377. if (inode->i_size < off+len-towrite) {
  2378. i_size_write(inode, off+len-towrite);
  2379. EXT3_I(inode)->i_disksize = inode->i_size;
  2380. }
  2381. inode->i_version++;
  2382. inode->i_mtime = inode->i_ctime = CURRENT_TIME;
  2383. ext3_mark_inode_dirty(handle, inode);
  2384. mutex_unlock(&inode->i_mutex);
  2385. return len - towrite;
  2386. }
  2387. #endif
  2388. static struct super_block *ext3_get_sb(struct file_system_type *fs_type,
  2389. int flags, const char *dev_name, void *data)
  2390. {
  2391. return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
  2392. }
  2393. static struct file_system_type ext3_fs_type = {
  2394. .owner = THIS_MODULE,
  2395. .name = "ext3",
  2396. .get_sb = ext3_get_sb,
  2397. .kill_sb = kill_block_super,
  2398. .fs_flags = FS_REQUIRES_DEV,
  2399. };
  2400. static int __init init_ext3_fs(void)
  2401. {
  2402. int err = init_ext3_xattr();
  2403. if (err)
  2404. return err;
  2405. err = init_inodecache();
  2406. if (err)
  2407. goto out1;
  2408. err = register_filesystem(&ext3_fs_type);
  2409. if (err)
  2410. goto out;
  2411. return 0;
  2412. out:
  2413. destroy_inodecache();
  2414. out1:
  2415. exit_ext3_xattr();
  2416. return err;
  2417. }
  2418. static void __exit exit_ext3_fs(void)
  2419. {
  2420. unregister_filesystem(&ext3_fs_type);
  2421. destroy_inodecache();
  2422. exit_ext3_xattr();
  2423. }
  2424. MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
  2425. MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
  2426. MODULE_LICENSE("GPL");
  2427. module_init(init_ext3_fs)
  2428. module_exit(exit_ext3_fs)