super.c 44 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661
  1. /*
  2. * linux/fs/ext2/super.c
  3. *
  4. * Copyright (C) 1992, 1993, 1994, 1995
  5. * Remy Card (card@masi.ibp.fr)
  6. * Laboratoire MASI - Institut Blaise Pascal
  7. * Universite Pierre et Marie Curie (Paris VI)
  8. *
  9. * from
  10. *
  11. * linux/fs/minix/inode.c
  12. *
  13. * Copyright (C) 1991, 1992 Linus Torvalds
  14. *
  15. * Big-endian to little-endian byte-swapping/bitmaps by
  16. * David S. Miller (davem@caip.rutgers.edu), 1995
  17. */
  18. #include <linux/module.h>
  19. #include <linux/string.h>
  20. #include <linux/fs.h>
  21. #include <linux/slab.h>
  22. #include <linux/init.h>
  23. #include <linux/blkdev.h>
  24. #include <linux/parser.h>
  25. #include <linux/random.h>
  26. #include <linux/buffer_head.h>
  27. #include <linux/exportfs.h>
  28. #include <linux/vfs.h>
  29. #include <linux/seq_file.h>
  30. #include <linux/mount.h>
  31. #include <linux/log2.h>
  32. #include <linux/quotaops.h>
  33. #include <linux/uaccess.h>
  34. #include <linux/dax.h>
  35. #include "ext2.h"
  36. #include "xattr.h"
  37. #include "acl.h"
  38. static void ext2_write_super(struct super_block *sb);
  39. static int ext2_remount (struct super_block * sb, int * flags, char * data);
  40. static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf);
  41. static int ext2_sync_fs(struct super_block *sb, int wait);
  42. static int ext2_freeze(struct super_block *sb);
  43. static int ext2_unfreeze(struct super_block *sb);
  44. void ext2_error(struct super_block *sb, const char *function,
  45. const char *fmt, ...)
  46. {
  47. struct va_format vaf;
  48. va_list args;
  49. struct ext2_sb_info *sbi = EXT2_SB(sb);
  50. struct ext2_super_block *es = sbi->s_es;
  51. if (!sb_rdonly(sb)) {
  52. spin_lock(&sbi->s_lock);
  53. sbi->s_mount_state |= EXT2_ERROR_FS;
  54. es->s_state |= cpu_to_le16(EXT2_ERROR_FS);
  55. spin_unlock(&sbi->s_lock);
  56. ext2_sync_super(sb, es, 1);
  57. }
  58. va_start(args, fmt);
  59. vaf.fmt = fmt;
  60. vaf.va = &args;
  61. printk(KERN_CRIT "EXT2-fs (%s): error: %s: %pV\n",
  62. sb->s_id, function, &vaf);
  63. va_end(args);
  64. if (test_opt(sb, ERRORS_PANIC))
  65. panic("EXT2-fs: panic from previous error\n");
  66. if (test_opt(sb, ERRORS_RO)) {
  67. ext2_msg(sb, KERN_CRIT,
  68. "error: remounting filesystem read-only");
  69. sb->s_flags |= MS_RDONLY;
  70. }
  71. }
  72. void ext2_msg(struct super_block *sb, const char *prefix,
  73. const char *fmt, ...)
  74. {
  75. struct va_format vaf;
  76. va_list args;
  77. va_start(args, fmt);
  78. vaf.fmt = fmt;
  79. vaf.va = &args;
  80. printk("%sEXT2-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
  81. va_end(args);
  82. }
  83. /*
  84. * This must be called with sbi->s_lock held.
  85. */
  86. void ext2_update_dynamic_rev(struct super_block *sb)
  87. {
  88. struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  89. if (le32_to_cpu(es->s_rev_level) > EXT2_GOOD_OLD_REV)
  90. return;
  91. ext2_msg(sb, KERN_WARNING,
  92. "warning: updating to rev %d because of "
  93. "new feature flag, running e2fsck is recommended",
  94. EXT2_DYNAMIC_REV);
  95. es->s_first_ino = cpu_to_le32(EXT2_GOOD_OLD_FIRST_INO);
  96. es->s_inode_size = cpu_to_le16(EXT2_GOOD_OLD_INODE_SIZE);
  97. es->s_rev_level = cpu_to_le32(EXT2_DYNAMIC_REV);
  98. /* leave es->s_feature_*compat flags alone */
  99. /* es->s_uuid will be set by e2fsck if empty */
  100. /*
  101. * The rest of the superblock fields should be zero, and if not it
  102. * means they are likely already in use, so leave them alone. We
  103. * can leave it up to e2fsck to clean up any inconsistencies there.
  104. */
  105. }
  106. #ifdef CONFIG_QUOTA
  107. static int ext2_quota_off(struct super_block *sb, int type);
  108. static void ext2_quota_off_umount(struct super_block *sb)
  109. {
  110. int type;
  111. for (type = 0; type < MAXQUOTAS; type++)
  112. ext2_quota_off(sb, type);
  113. }
  114. #else
  115. static inline void ext2_quota_off_umount(struct super_block *sb)
  116. {
  117. }
  118. #endif
  119. static void ext2_put_super (struct super_block * sb)
  120. {
  121. int db_count;
  122. int i;
  123. struct ext2_sb_info *sbi = EXT2_SB(sb);
  124. ext2_quota_off_umount(sb);
  125. if (sbi->s_ea_block_cache) {
  126. ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
  127. sbi->s_ea_block_cache = NULL;
  128. }
  129. if (!sb_rdonly(sb)) {
  130. struct ext2_super_block *es = sbi->s_es;
  131. spin_lock(&sbi->s_lock);
  132. es->s_state = cpu_to_le16(sbi->s_mount_state);
  133. spin_unlock(&sbi->s_lock);
  134. ext2_sync_super(sb, es, 1);
  135. }
  136. db_count = sbi->s_gdb_count;
  137. for (i = 0; i < db_count; i++)
  138. if (sbi->s_group_desc[i])
  139. brelse (sbi->s_group_desc[i]);
  140. kfree(sbi->s_group_desc);
  141. kfree(sbi->s_debts);
  142. percpu_counter_destroy(&sbi->s_freeblocks_counter);
  143. percpu_counter_destroy(&sbi->s_freeinodes_counter);
  144. percpu_counter_destroy(&sbi->s_dirs_counter);
  145. brelse (sbi->s_sbh);
  146. sb->s_fs_info = NULL;
  147. kfree(sbi->s_blockgroup_lock);
  148. kfree(sbi);
  149. }
  150. static struct kmem_cache * ext2_inode_cachep;
  151. static struct inode *ext2_alloc_inode(struct super_block *sb)
  152. {
  153. struct ext2_inode_info *ei;
  154. ei = kmem_cache_alloc(ext2_inode_cachep, GFP_KERNEL);
  155. if (!ei)
  156. return NULL;
  157. ei->i_block_alloc_info = NULL;
  158. ei->vfs_inode.i_version = 1;
  159. #ifdef CONFIG_QUOTA
  160. memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
  161. #endif
  162. return &ei->vfs_inode;
  163. }
  164. static void ext2_i_callback(struct rcu_head *head)
  165. {
  166. struct inode *inode = container_of(head, struct inode, i_rcu);
  167. kmem_cache_free(ext2_inode_cachep, EXT2_I(inode));
  168. }
  169. static void ext2_destroy_inode(struct inode *inode)
  170. {
  171. call_rcu(&inode->i_rcu, ext2_i_callback);
  172. }
  173. static void init_once(void *foo)
  174. {
  175. struct ext2_inode_info *ei = (struct ext2_inode_info *) foo;
  176. rwlock_init(&ei->i_meta_lock);
  177. #ifdef CONFIG_EXT2_FS_XATTR
  178. init_rwsem(&ei->xattr_sem);
  179. #endif
  180. mutex_init(&ei->truncate_mutex);
  181. #ifdef CONFIG_FS_DAX
  182. init_rwsem(&ei->dax_sem);
  183. #endif
  184. inode_init_once(&ei->vfs_inode);
  185. }
  186. static int __init init_inodecache(void)
  187. {
  188. ext2_inode_cachep = kmem_cache_create("ext2_inode_cache",
  189. sizeof(struct ext2_inode_info),
  190. 0, (SLAB_RECLAIM_ACCOUNT|
  191. SLAB_MEM_SPREAD|SLAB_ACCOUNT),
  192. init_once);
  193. if (ext2_inode_cachep == NULL)
  194. return -ENOMEM;
  195. return 0;
  196. }
  197. static void destroy_inodecache(void)
  198. {
  199. /*
  200. * Make sure all delayed rcu free inodes are flushed before we
  201. * destroy cache.
  202. */
  203. rcu_barrier();
  204. kmem_cache_destroy(ext2_inode_cachep);
  205. }
  206. static int ext2_show_options(struct seq_file *seq, struct dentry *root)
  207. {
  208. struct super_block *sb = root->d_sb;
  209. struct ext2_sb_info *sbi = EXT2_SB(sb);
  210. struct ext2_super_block *es = sbi->s_es;
  211. unsigned long def_mount_opts;
  212. spin_lock(&sbi->s_lock);
  213. def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
  214. if (sbi->s_sb_block != 1)
  215. seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
  216. if (test_opt(sb, MINIX_DF))
  217. seq_puts(seq, ",minixdf");
  218. if (test_opt(sb, GRPID))
  219. seq_puts(seq, ",grpid");
  220. if (!test_opt(sb, GRPID) && (def_mount_opts & EXT2_DEFM_BSDGROUPS))
  221. seq_puts(seq, ",nogrpid");
  222. if (!uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT2_DEF_RESUID)) ||
  223. le16_to_cpu(es->s_def_resuid) != EXT2_DEF_RESUID) {
  224. seq_printf(seq, ",resuid=%u",
  225. from_kuid_munged(&init_user_ns, sbi->s_resuid));
  226. }
  227. if (!gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT2_DEF_RESGID)) ||
  228. le16_to_cpu(es->s_def_resgid) != EXT2_DEF_RESGID) {
  229. seq_printf(seq, ",resgid=%u",
  230. from_kgid_munged(&init_user_ns, sbi->s_resgid));
  231. }
  232. if (test_opt(sb, ERRORS_RO)) {
  233. int def_errors = le16_to_cpu(es->s_errors);
  234. if (def_errors == EXT2_ERRORS_PANIC ||
  235. def_errors == EXT2_ERRORS_CONTINUE) {
  236. seq_puts(seq, ",errors=remount-ro");
  237. }
  238. }
  239. if (test_opt(sb, ERRORS_CONT))
  240. seq_puts(seq, ",errors=continue");
  241. if (test_opt(sb, ERRORS_PANIC))
  242. seq_puts(seq, ",errors=panic");
  243. if (test_opt(sb, NO_UID32))
  244. seq_puts(seq, ",nouid32");
  245. if (test_opt(sb, DEBUG))
  246. seq_puts(seq, ",debug");
  247. if (test_opt(sb, OLDALLOC))
  248. seq_puts(seq, ",oldalloc");
  249. #ifdef CONFIG_EXT2_FS_XATTR
  250. if (test_opt(sb, XATTR_USER))
  251. seq_puts(seq, ",user_xattr");
  252. if (!test_opt(sb, XATTR_USER) &&
  253. (def_mount_opts & EXT2_DEFM_XATTR_USER)) {
  254. seq_puts(seq, ",nouser_xattr");
  255. }
  256. #endif
  257. #ifdef CONFIG_EXT2_FS_POSIX_ACL
  258. if (test_opt(sb, POSIX_ACL))
  259. seq_puts(seq, ",acl");
  260. if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT2_DEFM_ACL))
  261. seq_puts(seq, ",noacl");
  262. #endif
  263. if (test_opt(sb, NOBH))
  264. seq_puts(seq, ",nobh");
  265. #if defined(CONFIG_QUOTA)
  266. if (sbi->s_mount_opt & EXT2_MOUNT_USRQUOTA)
  267. seq_puts(seq, ",usrquota");
  268. if (sbi->s_mount_opt & EXT2_MOUNT_GRPQUOTA)
  269. seq_puts(seq, ",grpquota");
  270. #endif
  271. #ifdef CONFIG_FS_DAX
  272. if (sbi->s_mount_opt & EXT2_MOUNT_XIP)
  273. seq_puts(seq, ",xip");
  274. if (sbi->s_mount_opt & EXT2_MOUNT_DAX)
  275. seq_puts(seq, ",dax");
  276. #endif
  277. if (!test_opt(sb, RESERVATION))
  278. seq_puts(seq, ",noreservation");
  279. spin_unlock(&sbi->s_lock);
  280. return 0;
  281. }
  282. #ifdef CONFIG_QUOTA
  283. static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data, size_t len, loff_t off);
  284. static ssize_t ext2_quota_write(struct super_block *sb, int type, const char *data, size_t len, loff_t off);
  285. static int ext2_quota_on(struct super_block *sb, int type, int format_id,
  286. const struct path *path);
  287. static struct dquot **ext2_get_dquots(struct inode *inode)
  288. {
  289. return EXT2_I(inode)->i_dquot;
  290. }
  291. static const struct quotactl_ops ext2_quotactl_ops = {
  292. .quota_on = ext2_quota_on,
  293. .quota_off = ext2_quota_off,
  294. .quota_sync = dquot_quota_sync,
  295. .get_state = dquot_get_state,
  296. .set_info = dquot_set_dqinfo,
  297. .get_dqblk = dquot_get_dqblk,
  298. .set_dqblk = dquot_set_dqblk,
  299. .get_nextdqblk = dquot_get_next_dqblk,
  300. };
  301. #endif
  302. static const struct super_operations ext2_sops = {
  303. .alloc_inode = ext2_alloc_inode,
  304. .destroy_inode = ext2_destroy_inode,
  305. .write_inode = ext2_write_inode,
  306. .evict_inode = ext2_evict_inode,
  307. .put_super = ext2_put_super,
  308. .sync_fs = ext2_sync_fs,
  309. .freeze_fs = ext2_freeze,
  310. .unfreeze_fs = ext2_unfreeze,
  311. .statfs = ext2_statfs,
  312. .remount_fs = ext2_remount,
  313. .show_options = ext2_show_options,
  314. #ifdef CONFIG_QUOTA
  315. .quota_read = ext2_quota_read,
  316. .quota_write = ext2_quota_write,
  317. .get_dquots = ext2_get_dquots,
  318. #endif
  319. };
  320. static struct inode *ext2_nfs_get_inode(struct super_block *sb,
  321. u64 ino, u32 generation)
  322. {
  323. struct inode *inode;
  324. if (ino < EXT2_FIRST_INO(sb) && ino != EXT2_ROOT_INO)
  325. return ERR_PTR(-ESTALE);
  326. if (ino > le32_to_cpu(EXT2_SB(sb)->s_es->s_inodes_count))
  327. return ERR_PTR(-ESTALE);
  328. /*
  329. * ext2_iget isn't quite right if the inode is currently unallocated!
  330. * However ext2_iget currently does appropriate checks to handle stale
  331. * inodes so everything is OK.
  332. */
  333. inode = ext2_iget(sb, ino);
  334. if (IS_ERR(inode))
  335. return ERR_CAST(inode);
  336. if (generation && inode->i_generation != generation) {
  337. /* we didn't find the right inode.. */
  338. iput(inode);
  339. return ERR_PTR(-ESTALE);
  340. }
  341. return inode;
  342. }
  343. static struct dentry *ext2_fh_to_dentry(struct super_block *sb, struct fid *fid,
  344. int fh_len, int fh_type)
  345. {
  346. return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
  347. ext2_nfs_get_inode);
  348. }
  349. static struct dentry *ext2_fh_to_parent(struct super_block *sb, struct fid *fid,
  350. int fh_len, int fh_type)
  351. {
  352. return generic_fh_to_parent(sb, fid, fh_len, fh_type,
  353. ext2_nfs_get_inode);
  354. }
  355. static const struct export_operations ext2_export_ops = {
  356. .fh_to_dentry = ext2_fh_to_dentry,
  357. .fh_to_parent = ext2_fh_to_parent,
  358. .get_parent = ext2_get_parent,
  359. };
  360. static unsigned long get_sb_block(void **data)
  361. {
  362. unsigned long sb_block;
  363. char *options = (char *) *data;
  364. if (!options || strncmp(options, "sb=", 3) != 0)
  365. return 1; /* Default location */
  366. options += 3;
  367. sb_block = simple_strtoul(options, &options, 0);
  368. if (*options && *options != ',') {
  369. printk("EXT2-fs: Invalid sb specification: %s\n",
  370. (char *) *data);
  371. return 1;
  372. }
  373. if (*options == ',')
  374. options++;
  375. *data = (void *) options;
  376. return sb_block;
  377. }
  378. enum {
  379. Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
  380. Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic,
  381. Opt_err_ro, Opt_nouid32, Opt_nocheck, Opt_debug,
  382. Opt_oldalloc, Opt_orlov, Opt_nobh, Opt_user_xattr, Opt_nouser_xattr,
  383. Opt_acl, Opt_noacl, Opt_xip, Opt_dax, Opt_ignore, Opt_err, Opt_quota,
  384. Opt_usrquota, Opt_grpquota, Opt_reservation, Opt_noreservation
  385. };
  386. static const match_table_t tokens = {
  387. {Opt_bsd_df, "bsddf"},
  388. {Opt_minix_df, "minixdf"},
  389. {Opt_grpid, "grpid"},
  390. {Opt_grpid, "bsdgroups"},
  391. {Opt_nogrpid, "nogrpid"},
  392. {Opt_nogrpid, "sysvgroups"},
  393. {Opt_resgid, "resgid=%u"},
  394. {Opt_resuid, "resuid=%u"},
  395. {Opt_sb, "sb=%u"},
  396. {Opt_err_cont, "errors=continue"},
  397. {Opt_err_panic, "errors=panic"},
  398. {Opt_err_ro, "errors=remount-ro"},
  399. {Opt_nouid32, "nouid32"},
  400. {Opt_nocheck, "check=none"},
  401. {Opt_nocheck, "nocheck"},
  402. {Opt_debug, "debug"},
  403. {Opt_oldalloc, "oldalloc"},
  404. {Opt_orlov, "orlov"},
  405. {Opt_nobh, "nobh"},
  406. {Opt_user_xattr, "user_xattr"},
  407. {Opt_nouser_xattr, "nouser_xattr"},
  408. {Opt_acl, "acl"},
  409. {Opt_noacl, "noacl"},
  410. {Opt_xip, "xip"},
  411. {Opt_dax, "dax"},
  412. {Opt_grpquota, "grpquota"},
  413. {Opt_ignore, "noquota"},
  414. {Opt_quota, "quota"},
  415. {Opt_usrquota, "usrquota"},
  416. {Opt_reservation, "reservation"},
  417. {Opt_noreservation, "noreservation"},
  418. {Opt_err, NULL}
  419. };
  420. static int parse_options(char *options, struct super_block *sb)
  421. {
  422. char *p;
  423. struct ext2_sb_info *sbi = EXT2_SB(sb);
  424. substring_t args[MAX_OPT_ARGS];
  425. int option;
  426. kuid_t uid;
  427. kgid_t gid;
  428. if (!options)
  429. return 1;
  430. while ((p = strsep (&options, ",")) != NULL) {
  431. int token;
  432. if (!*p)
  433. continue;
  434. token = match_token(p, tokens, args);
  435. switch (token) {
  436. case Opt_bsd_df:
  437. clear_opt (sbi->s_mount_opt, MINIX_DF);
  438. break;
  439. case Opt_minix_df:
  440. set_opt (sbi->s_mount_opt, MINIX_DF);
  441. break;
  442. case Opt_grpid:
  443. set_opt (sbi->s_mount_opt, GRPID);
  444. break;
  445. case Opt_nogrpid:
  446. clear_opt (sbi->s_mount_opt, GRPID);
  447. break;
  448. case Opt_resuid:
  449. if (match_int(&args[0], &option))
  450. return 0;
  451. uid = make_kuid(current_user_ns(), option);
  452. if (!uid_valid(uid)) {
  453. ext2_msg(sb, KERN_ERR, "Invalid uid value %d", option);
  454. return 0;
  455. }
  456. sbi->s_resuid = uid;
  457. break;
  458. case Opt_resgid:
  459. if (match_int(&args[0], &option))
  460. return 0;
  461. gid = make_kgid(current_user_ns(), option);
  462. if (!gid_valid(gid)) {
  463. ext2_msg(sb, KERN_ERR, "Invalid gid value %d", option);
  464. return 0;
  465. }
  466. sbi->s_resgid = gid;
  467. break;
  468. case Opt_sb:
  469. /* handled by get_sb_block() instead of here */
  470. /* *sb_block = match_int(&args[0]); */
  471. break;
  472. case Opt_err_panic:
  473. clear_opt (sbi->s_mount_opt, ERRORS_CONT);
  474. clear_opt (sbi->s_mount_opt, ERRORS_RO);
  475. set_opt (sbi->s_mount_opt, ERRORS_PANIC);
  476. break;
  477. case Opt_err_ro:
  478. clear_opt (sbi->s_mount_opt, ERRORS_CONT);
  479. clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
  480. set_opt (sbi->s_mount_opt, ERRORS_RO);
  481. break;
  482. case Opt_err_cont:
  483. clear_opt (sbi->s_mount_opt, ERRORS_RO);
  484. clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
  485. set_opt (sbi->s_mount_opt, ERRORS_CONT);
  486. break;
  487. case Opt_nouid32:
  488. set_opt (sbi->s_mount_opt, NO_UID32);
  489. break;
  490. case Opt_nocheck:
  491. clear_opt (sbi->s_mount_opt, CHECK);
  492. break;
  493. case Opt_debug:
  494. set_opt (sbi->s_mount_opt, DEBUG);
  495. break;
  496. case Opt_oldalloc:
  497. set_opt (sbi->s_mount_opt, OLDALLOC);
  498. break;
  499. case Opt_orlov:
  500. clear_opt (sbi->s_mount_opt, OLDALLOC);
  501. break;
  502. case Opt_nobh:
  503. set_opt (sbi->s_mount_opt, NOBH);
  504. break;
  505. #ifdef CONFIG_EXT2_FS_XATTR
  506. case Opt_user_xattr:
  507. set_opt (sbi->s_mount_opt, XATTR_USER);
  508. break;
  509. case Opt_nouser_xattr:
  510. clear_opt (sbi->s_mount_opt, XATTR_USER);
  511. break;
  512. #else
  513. case Opt_user_xattr:
  514. case Opt_nouser_xattr:
  515. ext2_msg(sb, KERN_INFO, "(no)user_xattr options"
  516. "not supported");
  517. break;
  518. #endif
  519. #ifdef CONFIG_EXT2_FS_POSIX_ACL
  520. case Opt_acl:
  521. set_opt(sbi->s_mount_opt, POSIX_ACL);
  522. break;
  523. case Opt_noacl:
  524. clear_opt(sbi->s_mount_opt, POSIX_ACL);
  525. break;
  526. #else
  527. case Opt_acl:
  528. case Opt_noacl:
  529. ext2_msg(sb, KERN_INFO,
  530. "(no)acl options not supported");
  531. break;
  532. #endif
  533. case Opt_xip:
  534. ext2_msg(sb, KERN_INFO, "use dax instead of xip");
  535. set_opt(sbi->s_mount_opt, XIP);
  536. /* Fall through */
  537. case Opt_dax:
  538. #ifdef CONFIG_FS_DAX
  539. ext2_msg(sb, KERN_WARNING,
  540. "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
  541. set_opt(sbi->s_mount_opt, DAX);
  542. #else
  543. ext2_msg(sb, KERN_INFO, "dax option not supported");
  544. #endif
  545. break;
  546. #if defined(CONFIG_QUOTA)
  547. case Opt_quota:
  548. case Opt_usrquota:
  549. set_opt(sbi->s_mount_opt, USRQUOTA);
  550. break;
  551. case Opt_grpquota:
  552. set_opt(sbi->s_mount_opt, GRPQUOTA);
  553. break;
  554. #else
  555. case Opt_quota:
  556. case Opt_usrquota:
  557. case Opt_grpquota:
  558. ext2_msg(sb, KERN_INFO,
  559. "quota operations not supported");
  560. break;
  561. #endif
  562. case Opt_reservation:
  563. set_opt(sbi->s_mount_opt, RESERVATION);
  564. ext2_msg(sb, KERN_INFO, "reservations ON");
  565. break;
  566. case Opt_noreservation:
  567. clear_opt(sbi->s_mount_opt, RESERVATION);
  568. ext2_msg(sb, KERN_INFO, "reservations OFF");
  569. break;
  570. case Opt_ignore:
  571. break;
  572. default:
  573. return 0;
  574. }
  575. }
  576. return 1;
  577. }
  578. static int ext2_setup_super (struct super_block * sb,
  579. struct ext2_super_block * es,
  580. int read_only)
  581. {
  582. int res = 0;
  583. struct ext2_sb_info *sbi = EXT2_SB(sb);
  584. if (le32_to_cpu(es->s_rev_level) > EXT2_MAX_SUPP_REV) {
  585. ext2_msg(sb, KERN_ERR,
  586. "error: revision level too high, "
  587. "forcing read-only mode");
  588. res = MS_RDONLY;
  589. }
  590. if (read_only)
  591. return res;
  592. if (!(sbi->s_mount_state & EXT2_VALID_FS))
  593. ext2_msg(sb, KERN_WARNING,
  594. "warning: mounting unchecked fs, "
  595. "running e2fsck is recommended");
  596. else if ((sbi->s_mount_state & EXT2_ERROR_FS))
  597. ext2_msg(sb, KERN_WARNING,
  598. "warning: mounting fs with errors, "
  599. "running e2fsck is recommended");
  600. else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
  601. le16_to_cpu(es->s_mnt_count) >=
  602. (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
  603. ext2_msg(sb, KERN_WARNING,
  604. "warning: maximal mount count reached, "
  605. "running e2fsck is recommended");
  606. else if (le32_to_cpu(es->s_checkinterval) &&
  607. (le32_to_cpu(es->s_lastcheck) +
  608. le32_to_cpu(es->s_checkinterval) <= get_seconds()))
  609. ext2_msg(sb, KERN_WARNING,
  610. "warning: checktime reached, "
  611. "running e2fsck is recommended");
  612. if (!le16_to_cpu(es->s_max_mnt_count))
  613. es->s_max_mnt_count = cpu_to_le16(EXT2_DFL_MAX_MNT_COUNT);
  614. le16_add_cpu(&es->s_mnt_count, 1);
  615. if (test_opt (sb, DEBUG))
  616. ext2_msg(sb, KERN_INFO, "%s, %s, bs=%lu, fs=%lu, gc=%lu, "
  617. "bpg=%lu, ipg=%lu, mo=%04lx]",
  618. EXT2FS_VERSION, EXT2FS_DATE, sb->s_blocksize,
  619. sbi->s_frag_size,
  620. sbi->s_groups_count,
  621. EXT2_BLOCKS_PER_GROUP(sb),
  622. EXT2_INODES_PER_GROUP(sb),
  623. sbi->s_mount_opt);
  624. return res;
  625. }
  626. static int ext2_check_descriptors(struct super_block *sb)
  627. {
  628. int i;
  629. struct ext2_sb_info *sbi = EXT2_SB(sb);
  630. ext2_debug ("Checking group descriptors");
  631. for (i = 0; i < sbi->s_groups_count; i++) {
  632. struct ext2_group_desc *gdp = ext2_get_group_desc(sb, i, NULL);
  633. ext2_fsblk_t first_block = ext2_group_first_block_no(sb, i);
  634. ext2_fsblk_t last_block;
  635. if (i == sbi->s_groups_count - 1)
  636. last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
  637. else
  638. last_block = first_block +
  639. (EXT2_BLOCKS_PER_GROUP(sb) - 1);
  640. if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
  641. le32_to_cpu(gdp->bg_block_bitmap) > last_block)
  642. {
  643. ext2_error (sb, "ext2_check_descriptors",
  644. "Block bitmap for group %d"
  645. " not in group (block %lu)!",
  646. i, (unsigned long) le32_to_cpu(gdp->bg_block_bitmap));
  647. return 0;
  648. }
  649. if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
  650. le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
  651. {
  652. ext2_error (sb, "ext2_check_descriptors",
  653. "Inode bitmap for group %d"
  654. " not in group (block %lu)!",
  655. i, (unsigned long) le32_to_cpu(gdp->bg_inode_bitmap));
  656. return 0;
  657. }
  658. if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
  659. le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
  660. last_block)
  661. {
  662. ext2_error (sb, "ext2_check_descriptors",
  663. "Inode table for group %d"
  664. " not in group (block %lu)!",
  665. i, (unsigned long) le32_to_cpu(gdp->bg_inode_table));
  666. return 0;
  667. }
  668. }
  669. return 1;
  670. }
  671. /*
  672. * Maximal file size. There is a direct, and {,double-,triple-}indirect
  673. * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
  674. * We need to be 1 filesystem block less than the 2^32 sector limit.
  675. */
  676. static loff_t ext2_max_size(int bits)
  677. {
  678. loff_t res = EXT2_NDIR_BLOCKS;
  679. int meta_blocks;
  680. loff_t upper_limit;
  681. /* This is calculated to be the largest file size for a
  682. * dense, file such that the total number of
  683. * sectors in the file, including data and all indirect blocks,
  684. * does not exceed 2^32 -1
  685. * __u32 i_blocks representing the total number of
  686. * 512 bytes blocks of the file
  687. */
  688. upper_limit = (1LL << 32) - 1;
  689. /* total blocks in file system block size */
  690. upper_limit >>= (bits - 9);
  691. /* indirect blocks */
  692. meta_blocks = 1;
  693. /* double indirect blocks */
  694. meta_blocks += 1 + (1LL << (bits-2));
  695. /* tripple indirect blocks */
  696. meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
  697. upper_limit -= meta_blocks;
  698. upper_limit <<= bits;
  699. res += 1LL << (bits-2);
  700. res += 1LL << (2*(bits-2));
  701. res += 1LL << (3*(bits-2));
  702. res <<= bits;
  703. if (res > upper_limit)
  704. res = upper_limit;
  705. if (res > MAX_LFS_FILESIZE)
  706. res = MAX_LFS_FILESIZE;
  707. return res;
  708. }
  709. static unsigned long descriptor_loc(struct super_block *sb,
  710. unsigned long logic_sb_block,
  711. int nr)
  712. {
  713. struct ext2_sb_info *sbi = EXT2_SB(sb);
  714. unsigned long bg, first_meta_bg;
  715. int has_super = 0;
  716. first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
  717. if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) ||
  718. nr < first_meta_bg)
  719. return (logic_sb_block + nr + 1);
  720. bg = sbi->s_desc_per_block * nr;
  721. if (ext2_bg_has_super(sb, bg))
  722. has_super = 1;
  723. return ext2_group_first_block_no(sb, bg) + has_super;
  724. }
  725. static int ext2_fill_super(struct super_block *sb, void *data, int silent)
  726. {
  727. struct buffer_head * bh;
  728. struct ext2_sb_info * sbi;
  729. struct ext2_super_block * es;
  730. struct inode *root;
  731. unsigned long block;
  732. unsigned long sb_block = get_sb_block(&data);
  733. unsigned long logic_sb_block;
  734. unsigned long offset = 0;
  735. unsigned long def_mount_opts;
  736. long ret = -EINVAL;
  737. int blocksize = BLOCK_SIZE;
  738. int db_count;
  739. int i, j;
  740. __le32 features;
  741. int err;
  742. err = -ENOMEM;
  743. sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
  744. if (!sbi)
  745. goto failed;
  746. sbi->s_blockgroup_lock =
  747. kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
  748. if (!sbi->s_blockgroup_lock) {
  749. kfree(sbi);
  750. goto failed;
  751. }
  752. sb->s_fs_info = sbi;
  753. sbi->s_sb_block = sb_block;
  754. spin_lock_init(&sbi->s_lock);
  755. /*
  756. * See what the current blocksize for the device is, and
  757. * use that as the blocksize. Otherwise (or if the blocksize
  758. * is smaller than the default) use the default.
  759. * This is important for devices that have a hardware
  760. * sectorsize that is larger than the default.
  761. */
  762. blocksize = sb_min_blocksize(sb, BLOCK_SIZE);
  763. if (!blocksize) {
  764. ext2_msg(sb, KERN_ERR, "error: unable to set blocksize");
  765. goto failed_sbi;
  766. }
  767. /*
  768. * If the superblock doesn't start on a hardware sector boundary,
  769. * calculate the offset.
  770. */
  771. if (blocksize != BLOCK_SIZE) {
  772. logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
  773. offset = (sb_block*BLOCK_SIZE) % blocksize;
  774. } else {
  775. logic_sb_block = sb_block;
  776. }
  777. if (!(bh = sb_bread(sb, logic_sb_block))) {
  778. ext2_msg(sb, KERN_ERR, "error: unable to read superblock");
  779. goto failed_sbi;
  780. }
  781. /*
  782. * Note: s_es must be initialized as soon as possible because
  783. * some ext2 macro-instructions depend on its value
  784. */
  785. es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
  786. sbi->s_es = es;
  787. sb->s_magic = le16_to_cpu(es->s_magic);
  788. if (sb->s_magic != EXT2_SUPER_MAGIC)
  789. goto cantfind_ext2;
  790. /* Set defaults before we parse the mount options */
  791. def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
  792. if (def_mount_opts & EXT2_DEFM_DEBUG)
  793. set_opt(sbi->s_mount_opt, DEBUG);
  794. if (def_mount_opts & EXT2_DEFM_BSDGROUPS)
  795. set_opt(sbi->s_mount_opt, GRPID);
  796. if (def_mount_opts & EXT2_DEFM_UID16)
  797. set_opt(sbi->s_mount_opt, NO_UID32);
  798. #ifdef CONFIG_EXT2_FS_XATTR
  799. if (def_mount_opts & EXT2_DEFM_XATTR_USER)
  800. set_opt(sbi->s_mount_opt, XATTR_USER);
  801. #endif
  802. #ifdef CONFIG_EXT2_FS_POSIX_ACL
  803. if (def_mount_opts & EXT2_DEFM_ACL)
  804. set_opt(sbi->s_mount_opt, POSIX_ACL);
  805. #endif
  806. if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC)
  807. set_opt(sbi->s_mount_opt, ERRORS_PANIC);
  808. else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_CONTINUE)
  809. set_opt(sbi->s_mount_opt, ERRORS_CONT);
  810. else
  811. set_opt(sbi->s_mount_opt, ERRORS_RO);
  812. sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
  813. sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
  814. set_opt(sbi->s_mount_opt, RESERVATION);
  815. if (!parse_options((char *) data, sb))
  816. goto failed_mount;
  817. sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
  818. ((EXT2_SB(sb)->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ?
  819. MS_POSIXACL : 0);
  820. sb->s_iflags |= SB_I_CGROUPWB;
  821. if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV &&
  822. (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) ||
  823. EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
  824. EXT2_HAS_INCOMPAT_FEATURE(sb, ~0U)))
  825. ext2_msg(sb, KERN_WARNING,
  826. "warning: feature flags set on rev 0 fs, "
  827. "running e2fsck is recommended");
  828. /*
  829. * Check feature flags regardless of the revision level, since we
  830. * previously didn't change the revision level when setting the flags,
  831. * so there is a chance incompat flags are set on a rev 0 filesystem.
  832. */
  833. features = EXT2_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP);
  834. if (features) {
  835. ext2_msg(sb, KERN_ERR, "error: couldn't mount because of "
  836. "unsupported optional features (%x)",
  837. le32_to_cpu(features));
  838. goto failed_mount;
  839. }
  840. if (!sb_rdonly(sb) && (features = EXT2_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))){
  841. ext2_msg(sb, KERN_ERR, "error: couldn't mount RDWR because of "
  842. "unsupported optional features (%x)",
  843. le32_to_cpu(features));
  844. goto failed_mount;
  845. }
  846. blocksize = BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
  847. if (sbi->s_mount_opt & EXT2_MOUNT_DAX) {
  848. err = bdev_dax_supported(sb, blocksize);
  849. if (err)
  850. goto failed_mount;
  851. }
  852. /* If the blocksize doesn't match, re-read the thing.. */
  853. if (sb->s_blocksize != blocksize) {
  854. brelse(bh);
  855. if (!sb_set_blocksize(sb, blocksize)) {
  856. ext2_msg(sb, KERN_ERR,
  857. "error: bad blocksize %d", blocksize);
  858. goto failed_sbi;
  859. }
  860. logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize;
  861. offset = (sb_block*BLOCK_SIZE) % blocksize;
  862. bh = sb_bread(sb, logic_sb_block);
  863. if(!bh) {
  864. ext2_msg(sb, KERN_ERR, "error: couldn't read"
  865. "superblock on 2nd try");
  866. goto failed_sbi;
  867. }
  868. es = (struct ext2_super_block *) (((char *)bh->b_data) + offset);
  869. sbi->s_es = es;
  870. if (es->s_magic != cpu_to_le16(EXT2_SUPER_MAGIC)) {
  871. ext2_msg(sb, KERN_ERR, "error: magic mismatch");
  872. goto failed_mount;
  873. }
  874. }
  875. sb->s_maxbytes = ext2_max_size(sb->s_blocksize_bits);
  876. sb->s_max_links = EXT2_LINK_MAX;
  877. if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) {
  878. sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE;
  879. sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO;
  880. } else {
  881. sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
  882. sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
  883. if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) ||
  884. !is_power_of_2(sbi->s_inode_size) ||
  885. (sbi->s_inode_size > blocksize)) {
  886. ext2_msg(sb, KERN_ERR,
  887. "error: unsupported inode size: %d",
  888. sbi->s_inode_size);
  889. goto failed_mount;
  890. }
  891. }
  892. sbi->s_frag_size = EXT2_MIN_FRAG_SIZE <<
  893. le32_to_cpu(es->s_log_frag_size);
  894. if (sbi->s_frag_size == 0)
  895. goto cantfind_ext2;
  896. sbi->s_frags_per_block = sb->s_blocksize / sbi->s_frag_size;
  897. sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
  898. sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
  899. sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
  900. if (EXT2_INODE_SIZE(sb) == 0)
  901. goto cantfind_ext2;
  902. sbi->s_inodes_per_block = sb->s_blocksize / EXT2_INODE_SIZE(sb);
  903. if (sbi->s_inodes_per_block == 0 || sbi->s_inodes_per_group == 0)
  904. goto cantfind_ext2;
  905. sbi->s_itb_per_group = sbi->s_inodes_per_group /
  906. sbi->s_inodes_per_block;
  907. sbi->s_desc_per_block = sb->s_blocksize /
  908. sizeof (struct ext2_group_desc);
  909. sbi->s_sbh = bh;
  910. sbi->s_mount_state = le16_to_cpu(es->s_state);
  911. sbi->s_addr_per_block_bits =
  912. ilog2 (EXT2_ADDR_PER_BLOCK(sb));
  913. sbi->s_desc_per_block_bits =
  914. ilog2 (EXT2_DESC_PER_BLOCK(sb));
  915. if (sb->s_magic != EXT2_SUPER_MAGIC)
  916. goto cantfind_ext2;
  917. if (sb->s_blocksize != bh->b_size) {
  918. if (!silent)
  919. ext2_msg(sb, KERN_ERR, "error: unsupported blocksize");
  920. goto failed_mount;
  921. }
  922. if (sb->s_blocksize != sbi->s_frag_size) {
  923. ext2_msg(sb, KERN_ERR,
  924. "error: fragsize %lu != blocksize %lu"
  925. "(not supported yet)",
  926. sbi->s_frag_size, sb->s_blocksize);
  927. goto failed_mount;
  928. }
  929. if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
  930. ext2_msg(sb, KERN_ERR,
  931. "error: #blocks per group too big: %lu",
  932. sbi->s_blocks_per_group);
  933. goto failed_mount;
  934. }
  935. if (sbi->s_frags_per_group > sb->s_blocksize * 8) {
  936. ext2_msg(sb, KERN_ERR,
  937. "error: #fragments per group too big: %lu",
  938. sbi->s_frags_per_group);
  939. goto failed_mount;
  940. }
  941. if (sbi->s_inodes_per_group > sb->s_blocksize * 8) {
  942. ext2_msg(sb, KERN_ERR,
  943. "error: #inodes per group too big: %lu",
  944. sbi->s_inodes_per_group);
  945. goto failed_mount;
  946. }
  947. if (EXT2_BLOCKS_PER_GROUP(sb) == 0)
  948. goto cantfind_ext2;
  949. sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
  950. le32_to_cpu(es->s_first_data_block) - 1)
  951. / EXT2_BLOCKS_PER_GROUP(sb)) + 1;
  952. db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) /
  953. EXT2_DESC_PER_BLOCK(sb);
  954. sbi->s_group_desc = kmalloc (db_count * sizeof (struct buffer_head *), GFP_KERNEL);
  955. if (sbi->s_group_desc == NULL) {
  956. ext2_msg(sb, KERN_ERR, "error: not enough memory");
  957. goto failed_mount;
  958. }
  959. bgl_lock_init(sbi->s_blockgroup_lock);
  960. sbi->s_debts = kcalloc(sbi->s_groups_count, sizeof(*sbi->s_debts), GFP_KERNEL);
  961. if (!sbi->s_debts) {
  962. ext2_msg(sb, KERN_ERR, "error: not enough memory");
  963. goto failed_mount_group_desc;
  964. }
  965. for (i = 0; i < db_count; i++) {
  966. block = descriptor_loc(sb, logic_sb_block, i);
  967. sbi->s_group_desc[i] = sb_bread(sb, block);
  968. if (!sbi->s_group_desc[i]) {
  969. for (j = 0; j < i; j++)
  970. brelse (sbi->s_group_desc[j]);
  971. ext2_msg(sb, KERN_ERR,
  972. "error: unable to read group descriptors");
  973. goto failed_mount_group_desc;
  974. }
  975. }
  976. if (!ext2_check_descriptors (sb)) {
  977. ext2_msg(sb, KERN_ERR, "group descriptors corrupted");
  978. goto failed_mount2;
  979. }
  980. sbi->s_gdb_count = db_count;
  981. get_random_bytes(&sbi->s_next_generation, sizeof(u32));
  982. spin_lock_init(&sbi->s_next_gen_lock);
  983. /* per fileystem reservation list head & lock */
  984. spin_lock_init(&sbi->s_rsv_window_lock);
  985. sbi->s_rsv_window_root = RB_ROOT;
  986. /*
  987. * Add a single, static dummy reservation to the start of the
  988. * reservation window list --- it gives us a placeholder for
  989. * append-at-start-of-list which makes the allocation logic
  990. * _much_ simpler.
  991. */
  992. sbi->s_rsv_window_head.rsv_start = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
  993. sbi->s_rsv_window_head.rsv_end = EXT2_RESERVE_WINDOW_NOT_ALLOCATED;
  994. sbi->s_rsv_window_head.rsv_alloc_hit = 0;
  995. sbi->s_rsv_window_head.rsv_goal_size = 0;
  996. ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
  997. err = percpu_counter_init(&sbi->s_freeblocks_counter,
  998. ext2_count_free_blocks(sb), GFP_KERNEL);
  999. if (!err) {
  1000. err = percpu_counter_init(&sbi->s_freeinodes_counter,
  1001. ext2_count_free_inodes(sb), GFP_KERNEL);
  1002. }
  1003. if (!err) {
  1004. err = percpu_counter_init(&sbi->s_dirs_counter,
  1005. ext2_count_dirs(sb), GFP_KERNEL);
  1006. }
  1007. if (err) {
  1008. ext2_msg(sb, KERN_ERR, "error: insufficient memory");
  1009. goto failed_mount3;
  1010. }
  1011. #ifdef CONFIG_EXT2_FS_XATTR
  1012. sbi->s_ea_block_cache = ext2_xattr_create_cache();
  1013. if (!sbi->s_ea_block_cache) {
  1014. ext2_msg(sb, KERN_ERR, "Failed to create ea_block_cache");
  1015. goto failed_mount3;
  1016. }
  1017. #endif
  1018. /*
  1019. * set up enough so that it can read an inode
  1020. */
  1021. sb->s_op = &ext2_sops;
  1022. sb->s_export_op = &ext2_export_ops;
  1023. sb->s_xattr = ext2_xattr_handlers;
  1024. #ifdef CONFIG_QUOTA
  1025. sb->dq_op = &dquot_operations;
  1026. sb->s_qcop = &ext2_quotactl_ops;
  1027. sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
  1028. #endif
  1029. root = ext2_iget(sb, EXT2_ROOT_INO);
  1030. if (IS_ERR(root)) {
  1031. ret = PTR_ERR(root);
  1032. goto failed_mount3;
  1033. }
  1034. if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
  1035. iput(root);
  1036. ext2_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
  1037. goto failed_mount3;
  1038. }
  1039. sb->s_root = d_make_root(root);
  1040. if (!sb->s_root) {
  1041. ext2_msg(sb, KERN_ERR, "error: get root inode failed");
  1042. ret = -ENOMEM;
  1043. goto failed_mount3;
  1044. }
  1045. if (EXT2_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL))
  1046. ext2_msg(sb, KERN_WARNING,
  1047. "warning: mounting ext3 filesystem as ext2");
  1048. if (ext2_setup_super (sb, es, sb_rdonly(sb)))
  1049. sb->s_flags |= MS_RDONLY;
  1050. ext2_write_super(sb);
  1051. return 0;
  1052. cantfind_ext2:
  1053. if (!silent)
  1054. ext2_msg(sb, KERN_ERR,
  1055. "error: can't find an ext2 filesystem on dev %s.",
  1056. sb->s_id);
  1057. goto failed_mount;
  1058. failed_mount3:
  1059. if (sbi->s_ea_block_cache)
  1060. ext2_xattr_destroy_cache(sbi->s_ea_block_cache);
  1061. percpu_counter_destroy(&sbi->s_freeblocks_counter);
  1062. percpu_counter_destroy(&sbi->s_freeinodes_counter);
  1063. percpu_counter_destroy(&sbi->s_dirs_counter);
  1064. failed_mount2:
  1065. for (i = 0; i < db_count; i++)
  1066. brelse(sbi->s_group_desc[i]);
  1067. failed_mount_group_desc:
  1068. kfree(sbi->s_group_desc);
  1069. kfree(sbi->s_debts);
  1070. failed_mount:
  1071. brelse(bh);
  1072. failed_sbi:
  1073. sb->s_fs_info = NULL;
  1074. kfree(sbi->s_blockgroup_lock);
  1075. kfree(sbi);
  1076. failed:
  1077. return ret;
  1078. }
  1079. static void ext2_clear_super_error(struct super_block *sb)
  1080. {
  1081. struct buffer_head *sbh = EXT2_SB(sb)->s_sbh;
  1082. if (buffer_write_io_error(sbh)) {
  1083. /*
  1084. * Oh, dear. A previous attempt to write the
  1085. * superblock failed. This could happen because the
  1086. * USB device was yanked out. Or it could happen to
  1087. * be a transient write error and maybe the block will
  1088. * be remapped. Nothing we can do but to retry the
  1089. * write and hope for the best.
  1090. */
  1091. ext2_msg(sb, KERN_ERR,
  1092. "previous I/O error to superblock detected\n");
  1093. clear_buffer_write_io_error(sbh);
  1094. set_buffer_uptodate(sbh);
  1095. }
  1096. }
  1097. void ext2_sync_super(struct super_block *sb, struct ext2_super_block *es,
  1098. int wait)
  1099. {
  1100. ext2_clear_super_error(sb);
  1101. spin_lock(&EXT2_SB(sb)->s_lock);
  1102. es->s_free_blocks_count = cpu_to_le32(ext2_count_free_blocks(sb));
  1103. es->s_free_inodes_count = cpu_to_le32(ext2_count_free_inodes(sb));
  1104. es->s_wtime = cpu_to_le32(get_seconds());
  1105. /* unlock before we do IO */
  1106. spin_unlock(&EXT2_SB(sb)->s_lock);
  1107. mark_buffer_dirty(EXT2_SB(sb)->s_sbh);
  1108. if (wait)
  1109. sync_dirty_buffer(EXT2_SB(sb)->s_sbh);
  1110. }
  1111. /*
  1112. * In the second extended file system, it is not necessary to
  1113. * write the super block since we use a mapping of the
  1114. * disk super block in a buffer.
  1115. *
  1116. * However, this function is still used to set the fs valid
  1117. * flags to 0. We need to set this flag to 0 since the fs
  1118. * may have been checked while mounted and e2fsck may have
  1119. * set s_state to EXT2_VALID_FS after some corrections.
  1120. */
  1121. static int ext2_sync_fs(struct super_block *sb, int wait)
  1122. {
  1123. struct ext2_sb_info *sbi = EXT2_SB(sb);
  1124. struct ext2_super_block *es = EXT2_SB(sb)->s_es;
  1125. /*
  1126. * Write quota structures to quota file, sync_blockdev() will write
  1127. * them to disk later
  1128. */
  1129. dquot_writeback_dquots(sb, -1);
  1130. spin_lock(&sbi->s_lock);
  1131. if (es->s_state & cpu_to_le16(EXT2_VALID_FS)) {
  1132. ext2_debug("setting valid to 0\n");
  1133. es->s_state &= cpu_to_le16(~EXT2_VALID_FS);
  1134. }
  1135. spin_unlock(&sbi->s_lock);
  1136. ext2_sync_super(sb, es, wait);
  1137. return 0;
  1138. }
  1139. static int ext2_freeze(struct super_block *sb)
  1140. {
  1141. struct ext2_sb_info *sbi = EXT2_SB(sb);
  1142. /*
  1143. * Open but unlinked files present? Keep EXT2_VALID_FS flag cleared
  1144. * because we have unattached inodes and thus filesystem is not fully
  1145. * consistent.
  1146. */
  1147. if (atomic_long_read(&sb->s_remove_count)) {
  1148. ext2_sync_fs(sb, 1);
  1149. return 0;
  1150. }
  1151. /* Set EXT2_FS_VALID flag */
  1152. spin_lock(&sbi->s_lock);
  1153. sbi->s_es->s_state = cpu_to_le16(sbi->s_mount_state);
  1154. spin_unlock(&sbi->s_lock);
  1155. ext2_sync_super(sb, sbi->s_es, 1);
  1156. return 0;
  1157. }
  1158. static int ext2_unfreeze(struct super_block *sb)
  1159. {
  1160. /* Just write sb to clear EXT2_VALID_FS flag */
  1161. ext2_write_super(sb);
  1162. return 0;
  1163. }
  1164. static void ext2_write_super(struct super_block *sb)
  1165. {
  1166. if (!sb_rdonly(sb))
  1167. ext2_sync_fs(sb, 1);
  1168. }
  1169. static int ext2_remount (struct super_block * sb, int * flags, char * data)
  1170. {
  1171. struct ext2_sb_info * sbi = EXT2_SB(sb);
  1172. struct ext2_super_block * es;
  1173. struct ext2_mount_options old_opts;
  1174. unsigned long old_sb_flags;
  1175. int err;
  1176. sync_filesystem(sb);
  1177. spin_lock(&sbi->s_lock);
  1178. /* Store the old options */
  1179. old_sb_flags = sb->s_flags;
  1180. old_opts.s_mount_opt = sbi->s_mount_opt;
  1181. old_opts.s_resuid = sbi->s_resuid;
  1182. old_opts.s_resgid = sbi->s_resgid;
  1183. /*
  1184. * Allow the "check" option to be passed as a remount option.
  1185. */
  1186. if (!parse_options(data, sb)) {
  1187. err = -EINVAL;
  1188. goto restore_opts;
  1189. }
  1190. sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
  1191. ((sbi->s_mount_opt & EXT2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
  1192. es = sbi->s_es;
  1193. if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT2_MOUNT_DAX) {
  1194. ext2_msg(sb, KERN_WARNING, "warning: refusing change of "
  1195. "dax flag with busy inodes while remounting");
  1196. sbi->s_mount_opt ^= EXT2_MOUNT_DAX;
  1197. }
  1198. if ((bool)(*flags & MS_RDONLY) == sb_rdonly(sb)) {
  1199. spin_unlock(&sbi->s_lock);
  1200. return 0;
  1201. }
  1202. if (*flags & MS_RDONLY) {
  1203. if (le16_to_cpu(es->s_state) & EXT2_VALID_FS ||
  1204. !(sbi->s_mount_state & EXT2_VALID_FS)) {
  1205. spin_unlock(&sbi->s_lock);
  1206. return 0;
  1207. }
  1208. /*
  1209. * OK, we are remounting a valid rw partition rdonly, so set
  1210. * the rdonly flag and then mark the partition as valid again.
  1211. */
  1212. es->s_state = cpu_to_le16(sbi->s_mount_state);
  1213. es->s_mtime = cpu_to_le32(get_seconds());
  1214. spin_unlock(&sbi->s_lock);
  1215. err = dquot_suspend(sb, -1);
  1216. if (err < 0) {
  1217. spin_lock(&sbi->s_lock);
  1218. goto restore_opts;
  1219. }
  1220. ext2_sync_super(sb, es, 1);
  1221. } else {
  1222. __le32 ret = EXT2_HAS_RO_COMPAT_FEATURE(sb,
  1223. ~EXT2_FEATURE_RO_COMPAT_SUPP);
  1224. if (ret) {
  1225. ext2_msg(sb, KERN_WARNING,
  1226. "warning: couldn't remount RDWR because of "
  1227. "unsupported optional features (%x).",
  1228. le32_to_cpu(ret));
  1229. err = -EROFS;
  1230. goto restore_opts;
  1231. }
  1232. /*
  1233. * Mounting a RDONLY partition read-write, so reread and
  1234. * store the current valid flag. (It may have been changed
  1235. * by e2fsck since we originally mounted the partition.)
  1236. */
  1237. sbi->s_mount_state = le16_to_cpu(es->s_state);
  1238. if (!ext2_setup_super (sb, es, 0))
  1239. sb->s_flags &= ~MS_RDONLY;
  1240. spin_unlock(&sbi->s_lock);
  1241. ext2_write_super(sb);
  1242. dquot_resume(sb, -1);
  1243. }
  1244. return 0;
  1245. restore_opts:
  1246. sbi->s_mount_opt = old_opts.s_mount_opt;
  1247. sbi->s_resuid = old_opts.s_resuid;
  1248. sbi->s_resgid = old_opts.s_resgid;
  1249. sb->s_flags = old_sb_flags;
  1250. spin_unlock(&sbi->s_lock);
  1251. return err;
  1252. }
  1253. static int ext2_statfs (struct dentry * dentry, struct kstatfs * buf)
  1254. {
  1255. struct super_block *sb = dentry->d_sb;
  1256. struct ext2_sb_info *sbi = EXT2_SB(sb);
  1257. struct ext2_super_block *es = sbi->s_es;
  1258. u64 fsid;
  1259. spin_lock(&sbi->s_lock);
  1260. if (test_opt (sb, MINIX_DF))
  1261. sbi->s_overhead_last = 0;
  1262. else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
  1263. unsigned long i, overhead = 0;
  1264. smp_rmb();
  1265. /*
  1266. * Compute the overhead (FS structures). This is constant
  1267. * for a given filesystem unless the number of block groups
  1268. * changes so we cache the previous value until it does.
  1269. */
  1270. /*
  1271. * All of the blocks before first_data_block are
  1272. * overhead
  1273. */
  1274. overhead = le32_to_cpu(es->s_first_data_block);
  1275. /*
  1276. * Add the overhead attributed to the superblock and
  1277. * block group descriptors. If the sparse superblocks
  1278. * feature is turned on, then not all groups have this.
  1279. */
  1280. for (i = 0; i < sbi->s_groups_count; i++)
  1281. overhead += ext2_bg_has_super(sb, i) +
  1282. ext2_bg_num_gdb(sb, i);
  1283. /*
  1284. * Every block group has an inode bitmap, a block
  1285. * bitmap, and an inode table.
  1286. */
  1287. overhead += (sbi->s_groups_count *
  1288. (2 + sbi->s_itb_per_group));
  1289. sbi->s_overhead_last = overhead;
  1290. smp_wmb();
  1291. sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
  1292. }
  1293. buf->f_type = EXT2_SUPER_MAGIC;
  1294. buf->f_bsize = sb->s_blocksize;
  1295. buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
  1296. buf->f_bfree = ext2_count_free_blocks(sb);
  1297. es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
  1298. buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
  1299. if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
  1300. buf->f_bavail = 0;
  1301. buf->f_files = le32_to_cpu(es->s_inodes_count);
  1302. buf->f_ffree = ext2_count_free_inodes(sb);
  1303. es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
  1304. buf->f_namelen = EXT2_NAME_LEN;
  1305. fsid = le64_to_cpup((void *)es->s_uuid) ^
  1306. le64_to_cpup((void *)es->s_uuid + sizeof(u64));
  1307. buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
  1308. buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
  1309. spin_unlock(&sbi->s_lock);
  1310. return 0;
  1311. }
  1312. static struct dentry *ext2_mount(struct file_system_type *fs_type,
  1313. int flags, const char *dev_name, void *data)
  1314. {
  1315. return mount_bdev(fs_type, flags, dev_name, data, ext2_fill_super);
  1316. }
  1317. #ifdef CONFIG_QUOTA
  1318. /* Read data from quotafile - avoid pagecache and such because we cannot afford
  1319. * acquiring the locks... As quota files are never truncated and quota code
  1320. * itself serializes the operations (and no one else should touch the files)
  1321. * we don't have to be afraid of races */
  1322. static ssize_t ext2_quota_read(struct super_block *sb, int type, char *data,
  1323. size_t len, loff_t off)
  1324. {
  1325. struct inode *inode = sb_dqopt(sb)->files[type];
  1326. sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
  1327. int err = 0;
  1328. int offset = off & (sb->s_blocksize - 1);
  1329. int tocopy;
  1330. size_t toread;
  1331. struct buffer_head tmp_bh;
  1332. struct buffer_head *bh;
  1333. loff_t i_size = i_size_read(inode);
  1334. if (off > i_size)
  1335. return 0;
  1336. if (off+len > i_size)
  1337. len = i_size-off;
  1338. toread = len;
  1339. while (toread > 0) {
  1340. tocopy = sb->s_blocksize - offset < toread ?
  1341. sb->s_blocksize - offset : toread;
  1342. tmp_bh.b_state = 0;
  1343. tmp_bh.b_size = sb->s_blocksize;
  1344. err = ext2_get_block(inode, blk, &tmp_bh, 0);
  1345. if (err < 0)
  1346. return err;
  1347. if (!buffer_mapped(&tmp_bh)) /* A hole? */
  1348. memset(data, 0, tocopy);
  1349. else {
  1350. bh = sb_bread(sb, tmp_bh.b_blocknr);
  1351. if (!bh)
  1352. return -EIO;
  1353. memcpy(data, bh->b_data+offset, tocopy);
  1354. brelse(bh);
  1355. }
  1356. offset = 0;
  1357. toread -= tocopy;
  1358. data += tocopy;
  1359. blk++;
  1360. }
  1361. return len;
  1362. }
  1363. /* Write to quotafile */
  1364. static ssize_t ext2_quota_write(struct super_block *sb, int type,
  1365. const char *data, size_t len, loff_t off)
  1366. {
  1367. struct inode *inode = sb_dqopt(sb)->files[type];
  1368. sector_t blk = off >> EXT2_BLOCK_SIZE_BITS(sb);
  1369. int err = 0;
  1370. int offset = off & (sb->s_blocksize - 1);
  1371. int tocopy;
  1372. size_t towrite = len;
  1373. struct buffer_head tmp_bh;
  1374. struct buffer_head *bh;
  1375. while (towrite > 0) {
  1376. tocopy = sb->s_blocksize - offset < towrite ?
  1377. sb->s_blocksize - offset : towrite;
  1378. tmp_bh.b_state = 0;
  1379. tmp_bh.b_size = sb->s_blocksize;
  1380. err = ext2_get_block(inode, blk, &tmp_bh, 1);
  1381. if (err < 0)
  1382. goto out;
  1383. if (offset || tocopy != EXT2_BLOCK_SIZE(sb))
  1384. bh = sb_bread(sb, tmp_bh.b_blocknr);
  1385. else
  1386. bh = sb_getblk(sb, tmp_bh.b_blocknr);
  1387. if (unlikely(!bh)) {
  1388. err = -EIO;
  1389. goto out;
  1390. }
  1391. lock_buffer(bh);
  1392. memcpy(bh->b_data+offset, data, tocopy);
  1393. flush_dcache_page(bh->b_page);
  1394. set_buffer_uptodate(bh);
  1395. mark_buffer_dirty(bh);
  1396. unlock_buffer(bh);
  1397. brelse(bh);
  1398. offset = 0;
  1399. towrite -= tocopy;
  1400. data += tocopy;
  1401. blk++;
  1402. }
  1403. out:
  1404. if (len == towrite)
  1405. return err;
  1406. if (inode->i_size < off+len-towrite)
  1407. i_size_write(inode, off+len-towrite);
  1408. inode->i_version++;
  1409. inode->i_mtime = inode->i_ctime = current_time(inode);
  1410. mark_inode_dirty(inode);
  1411. return len - towrite;
  1412. }
  1413. static int ext2_quota_on(struct super_block *sb, int type, int format_id,
  1414. const struct path *path)
  1415. {
  1416. int err;
  1417. struct inode *inode;
  1418. err = dquot_quota_on(sb, type, format_id, path);
  1419. if (err)
  1420. return err;
  1421. inode = d_inode(path->dentry);
  1422. inode_lock(inode);
  1423. EXT2_I(inode)->i_flags |= EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL;
  1424. inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
  1425. S_NOATIME | S_IMMUTABLE);
  1426. inode_unlock(inode);
  1427. mark_inode_dirty(inode);
  1428. return 0;
  1429. }
  1430. static int ext2_quota_off(struct super_block *sb, int type)
  1431. {
  1432. struct inode *inode = sb_dqopt(sb)->files[type];
  1433. int err;
  1434. if (!inode || !igrab(inode))
  1435. goto out;
  1436. err = dquot_quota_off(sb, type);
  1437. if (err)
  1438. goto out_put;
  1439. inode_lock(inode);
  1440. EXT2_I(inode)->i_flags &= ~(EXT2_NOATIME_FL | EXT2_IMMUTABLE_FL);
  1441. inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
  1442. inode_unlock(inode);
  1443. mark_inode_dirty(inode);
  1444. out_put:
  1445. iput(inode);
  1446. return err;
  1447. out:
  1448. return dquot_quota_off(sb, type);
  1449. }
  1450. #endif
  1451. static struct file_system_type ext2_fs_type = {
  1452. .owner = THIS_MODULE,
  1453. .name = "ext2",
  1454. .mount = ext2_mount,
  1455. .kill_sb = kill_block_super,
  1456. .fs_flags = FS_REQUIRES_DEV,
  1457. };
  1458. MODULE_ALIAS_FS("ext2");
  1459. static int __init init_ext2_fs(void)
  1460. {
  1461. int err;
  1462. err = init_inodecache();
  1463. if (err)
  1464. return err;
  1465. err = register_filesystem(&ext2_fs_type);
  1466. if (err)
  1467. goto out;
  1468. return 0;
  1469. out:
  1470. destroy_inodecache();
  1471. return err;
  1472. }
  1473. static void __exit exit_ext2_fs(void)
  1474. {
  1475. unregister_filesystem(&ext2_fs_type);
  1476. destroy_inodecache();
  1477. }
  1478. MODULE_AUTHOR("Remy Card and others");
  1479. MODULE_DESCRIPTION("Second Extended Filesystem");
  1480. MODULE_LICENSE("GPL");
  1481. module_init(init_ext2_fs)
  1482. module_exit(exit_ext2_fs)