1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* -*- mode: c; c-basic-offset: 8; -*- 3 * vim: noexpandtab sw=8 ts=8 sts=0: 4 * 5 * super.c 6 * 7 * load/unload driver, mount/dismount volumes 8 * 9 * Copyright (C) 2002, 2004 Oracle. All rights reserved. 10 */ 11 12 #include <linux/module.h> 13 #include <linux/fs.h> 14 #include <linux/types.h> 15 #include <linux/slab.h> 16 #include <linux/highmem.h> 17 #include <linux/init.h> 18 #include <linux/random.h> 19 #include <linux/statfs.h> 20 #include <linux/moduleparam.h> 21 #include <linux/blkdev.h> 22 #include <linux/socket.h> 23 #include <linux/inet.h> 24 #include <linux/parser.h> 25 #include <linux/crc32.h> 26 #include <linux/debugfs.h> 27 #include <linux/mount.h> 28 #include <linux/seq_file.h> 29 #include <linux/quotaops.h> 30 #include <linux/cleancache.h> 31 #include <linux/signal.h> 32 33 #define CREATE_TRACE_POINTS 34 #include "ocfs2_trace.h" 35 36 #include <cluster/masklog.h> 37 38 #include "ocfs2.h" 39 40 /* this should be the only file to include a version 1 header */ 41 #include "ocfs1_fs_compat.h" 42 43 #include "alloc.h" 44 #include "aops.h" 45 #include "blockcheck.h" 46 #include "dlmglue.h" 47 #include "export.h" 48 #include "extent_map.h" 49 #include "heartbeat.h" 50 #include "inode.h" 51 #include "journal.h" 52 #include "localalloc.h" 53 #include "namei.h" 54 #include "slot_map.h" 55 #include "super.h" 56 #include "sysfile.h" 57 #include "uptodate.h" 58 #include "xattr.h" 59 #include "quota.h" 60 #include "refcounttree.h" 61 #include "suballoc.h" 62 63 #include "buffer_head_io.h" 64 #include "filecheck.h" 65 66 static struct kmem_cache *ocfs2_inode_cachep; 67 struct kmem_cache *ocfs2_dquot_cachep; 68 struct kmem_cache *ocfs2_qf_chunk_cachep; 69 70 static struct dentry *ocfs2_debugfs_root; 71 72 MODULE_AUTHOR("Oracle"); 73 MODULE_LICENSE("GPL"); 74 MODULE_DESCRIPTION("OCFS2 cluster file system"); 75 76 struct mount_options 77 { 78 unsigned long commit_interval; 79 unsigned long mount_opt; 80 unsigned int atime_quantum; 81 signed short slot; 82 int localalloc_opt; 83 unsigned int resv_level; 84 int dir_resv_level; 85 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; 86 }; 87 88 static int ocfs2_parse_options(struct super_block *sb, char *options, 89 struct mount_options *mopt, 90 int is_remount); 91 static int ocfs2_check_set_options(struct super_block *sb, 92 struct mount_options *options); 93 static int ocfs2_show_options(struct seq_file *s, struct dentry *root); 94 static void ocfs2_put_super(struct super_block *sb); 95 static int ocfs2_mount_volume(struct super_block *sb); 96 static int ocfs2_remount(struct super_block *sb, int *flags, char *data); 97 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err); 98 static int ocfs2_initialize_mem_caches(void); 99 static void ocfs2_free_mem_caches(void); 100 static void ocfs2_delete_osb(struct ocfs2_super *osb); 101 102 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf); 103 104 static int ocfs2_sync_fs(struct super_block *sb, int wait); 105 106 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb); 107 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb); 108 static void ocfs2_release_system_inodes(struct ocfs2_super *osb); 109 static int ocfs2_check_volume(struct ocfs2_super *osb); 110 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 111 struct buffer_head *bh, 112 u32 sectsize, 113 struct ocfs2_blockcheck_stats *stats); 114 static int ocfs2_initialize_super(struct super_block *sb, 115 struct buffer_head *bh, 116 int sector_size, 117 struct ocfs2_blockcheck_stats *stats); 118 static int ocfs2_get_sector(struct super_block *sb, 119 struct buffer_head **bh, 120 int block, 121 int sect_size); 122 static struct inode *ocfs2_alloc_inode(struct super_block *sb); 123 static void ocfs2_free_inode(struct inode *inode); 124 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend); 125 static int ocfs2_enable_quotas(struct ocfs2_super *osb); 126 static void ocfs2_disable_quotas(struct ocfs2_super *osb); 127 128 static struct dquot **ocfs2_get_dquots(struct inode *inode) 129 { 130 return OCFS2_I(inode)->i_dquot; 131 } 132 133 static const struct super_operations ocfs2_sops = { 134 .statfs = ocfs2_statfs, 135 .alloc_inode = ocfs2_alloc_inode, 136 .free_inode = ocfs2_free_inode, 137 .drop_inode = ocfs2_drop_inode, 138 .evict_inode = ocfs2_evict_inode, 139 .sync_fs = ocfs2_sync_fs, 140 .put_super = ocfs2_put_super, 141 .remount_fs = ocfs2_remount, 142 .show_options = ocfs2_show_options, 143 .quota_read = ocfs2_quota_read, 144 .quota_write = ocfs2_quota_write, 145 .get_dquots = ocfs2_get_dquots, 146 }; 147 148 enum { 149 Opt_barrier, 150 Opt_err_panic, 151 Opt_err_ro, 152 Opt_intr, 153 Opt_nointr, 154 Opt_hb_none, 155 Opt_hb_local, 156 Opt_hb_global, 157 Opt_data_ordered, 158 Opt_data_writeback, 159 Opt_atime_quantum, 160 Opt_slot, 161 Opt_commit, 162 Opt_localalloc, 163 Opt_localflocks, 164 Opt_stack, 165 Opt_user_xattr, 166 Opt_nouser_xattr, 167 Opt_inode64, 168 Opt_acl, 169 Opt_noacl, 170 Opt_usrquota, 171 Opt_grpquota, 172 Opt_coherency_buffered, 173 Opt_coherency_full, 174 Opt_resv_level, 175 Opt_dir_resv_level, 176 Opt_journal_async_commit, 177 Opt_err_cont, 178 Opt_err, 179 }; 180 181 static const match_table_t tokens = { 182 {Opt_barrier, "barrier=%u"}, 183 {Opt_err_panic, "errors=panic"}, 184 {Opt_err_ro, "errors=remount-ro"}, 185 {Opt_intr, "intr"}, 186 {Opt_nointr, "nointr"}, 187 {Opt_hb_none, OCFS2_HB_NONE}, 188 {Opt_hb_local, OCFS2_HB_LOCAL}, 189 {Opt_hb_global, OCFS2_HB_GLOBAL}, 190 {Opt_data_ordered, "data=ordered"}, 191 {Opt_data_writeback, "data=writeback"}, 192 {Opt_atime_quantum, "atime_quantum=%u"}, 193 {Opt_slot, "preferred_slot=%u"}, 194 {Opt_commit, "commit=%u"}, 195 {Opt_localalloc, "localalloc=%d"}, 196 {Opt_localflocks, "localflocks"}, 197 {Opt_stack, "cluster_stack=%s"}, 198 {Opt_user_xattr, "user_xattr"}, 199 {Opt_nouser_xattr, "nouser_xattr"}, 200 {Opt_inode64, "inode64"}, 201 {Opt_acl, "acl"}, 202 {Opt_noacl, "noacl"}, 203 {Opt_usrquota, "usrquota"}, 204 {Opt_grpquota, "grpquota"}, 205 {Opt_coherency_buffered, "coherency=buffered"}, 206 {Opt_coherency_full, "coherency=full"}, 207 {Opt_resv_level, "resv_level=%u"}, 208 {Opt_dir_resv_level, "dir_resv_level=%u"}, 209 {Opt_journal_async_commit, "journal_async_commit"}, 210 {Opt_err_cont, "errors=continue"}, 211 {Opt_err, NULL} 212 }; 213 214 #ifdef CONFIG_DEBUG_FS 215 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len) 216 { 217 struct ocfs2_cluster_connection *cconn = osb->cconn; 218 struct ocfs2_recovery_map *rm = osb->recovery_map; 219 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan; 220 int i, out = 0; 221 unsigned long flags; 222 223 out += snprintf(buf + out, len - out, 224 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n", 225 "Device", osb->dev_str, osb->uuid_str, 226 osb->fs_generation, osb->vol_label); 227 228 out += snprintf(buf + out, len - out, 229 "%10s => State: %d Flags: 0x%lX\n", "Volume", 230 atomic_read(&osb->vol_state), osb->osb_flags); 231 232 out += snprintf(buf + out, len - out, 233 "%10s => Block: %lu Cluster: %d\n", "Sizes", 234 osb->sb->s_blocksize, osb->s_clustersize); 235 236 out += snprintf(buf + out, len - out, 237 "%10s => Compat: 0x%X Incompat: 0x%X " 238 "ROcompat: 0x%X\n", 239 "Features", osb->s_feature_compat, 240 osb->s_feature_incompat, osb->s_feature_ro_compat); 241 242 out += snprintf(buf + out, len - out, 243 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount", 244 osb->s_mount_opt, osb->s_atime_quantum); 245 246 if (cconn) { 247 out += snprintf(buf + out, len - out, 248 "%10s => Stack: %s Name: %*s " 249 "Version: %d.%d\n", "Cluster", 250 (*osb->osb_cluster_stack == '\0' ? 251 "o2cb" : osb->osb_cluster_stack), 252 cconn->cc_namelen, cconn->cc_name, 253 cconn->cc_version.pv_major, 254 cconn->cc_version.pv_minor); 255 } 256 257 spin_lock_irqsave(&osb->dc_task_lock, flags); 258 out += snprintf(buf + out, len - out, 259 "%10s => Pid: %d Count: %lu WakeSeq: %lu " 260 "WorkSeq: %lu\n", "DownCnvt", 261 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1), 262 osb->blocked_lock_count, osb->dc_wake_sequence, 263 osb->dc_work_sequence); 264 spin_unlock_irqrestore(&osb->dc_task_lock, flags); 265 266 spin_lock(&osb->osb_lock); 267 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:", 268 "Recovery", 269 (osb->recovery_thread_task ? 270 task_pid_nr(osb->recovery_thread_task) : -1)); 271 if (rm->rm_used == 0) 272 out += snprintf(buf + out, len - out, " None\n"); 273 else { 274 for (i = 0; i < rm->rm_used; i++) 275 out += snprintf(buf + out, len - out, " %d", 276 rm->rm_entries[i]); 277 out += snprintf(buf + out, len - out, "\n"); 278 } 279 spin_unlock(&osb->osb_lock); 280 281 out += snprintf(buf + out, len - out, 282 "%10s => Pid: %d Interval: %lu\n", "Commit", 283 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1), 284 osb->osb_commit_interval); 285 286 out += snprintf(buf + out, len - out, 287 "%10s => State: %d TxnId: %lu NumTxns: %d\n", 288 "Journal", osb->journal->j_state, 289 osb->journal->j_trans_id, 290 atomic_read(&osb->journal->j_num_trans)); 291 292 out += snprintf(buf + out, len - out, 293 "%10s => GlobalAllocs: %d LocalAllocs: %d " 294 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n", 295 "Stats", 296 atomic_read(&osb->alloc_stats.bitmap_data), 297 atomic_read(&osb->alloc_stats.local_data), 298 atomic_read(&osb->alloc_stats.bg_allocs), 299 atomic_read(&osb->alloc_stats.moves), 300 atomic_read(&osb->alloc_stats.bg_extends)); 301 302 out += snprintf(buf + out, len - out, 303 "%10s => State: %u Descriptor: %llu Size: %u bits " 304 "Default: %u bits\n", 305 "LocalAlloc", osb->local_alloc_state, 306 (unsigned long long)osb->la_last_gd, 307 osb->local_alloc_bits, osb->local_alloc_default_bits); 308 309 spin_lock(&osb->osb_lock); 310 out += snprintf(buf + out, len - out, 311 "%10s => InodeSlot: %d StolenInodes: %d, " 312 "MetaSlot: %d StolenMeta: %d\n", "Steal", 313 osb->s_inode_steal_slot, 314 atomic_read(&osb->s_num_inodes_stolen), 315 osb->s_meta_steal_slot, 316 atomic_read(&osb->s_num_meta_stolen)); 317 spin_unlock(&osb->osb_lock); 318 319 out += snprintf(buf + out, len - out, "OrphanScan => "); 320 out += snprintf(buf + out, len - out, "Local: %u Global: %u ", 321 os->os_count, os->os_seqno); 322 out += snprintf(buf + out, len - out, " Last Scan: "); 323 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) 324 out += snprintf(buf + out, len - out, "Disabled\n"); 325 else 326 out += snprintf(buf + out, len - out, "%lu seconds ago\n", 327 (unsigned long)(ktime_get_seconds() - os->os_scantime)); 328 329 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n", 330 "Slots", "Num", "RecoGen"); 331 for (i = 0; i < osb->max_slots; ++i) { 332 out += snprintf(buf + out, len - out, 333 "%10s %c %3d %10d\n", 334 " ", 335 (i == osb->slot_num ? '*' : ' '), 336 i, osb->slot_recovery_generations[i]); 337 } 338 339 return out; 340 } 341 342 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file) 343 { 344 struct ocfs2_super *osb = inode->i_private; 345 char *buf = NULL; 346 347 buf = kmalloc(PAGE_SIZE, GFP_KERNEL); 348 if (!buf) 349 goto bail; 350 351 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE)); 352 353 file->private_data = buf; 354 355 return 0; 356 bail: 357 return -ENOMEM; 358 } 359 360 static int ocfs2_debug_release(struct inode *inode, struct file *file) 361 { 362 kfree(file->private_data); 363 return 0; 364 } 365 366 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, 367 size_t nbytes, loff_t *ppos) 368 { 369 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, 370 i_size_read(file->f_mapping->host)); 371 } 372 #else 373 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file) 374 { 375 return 0; 376 } 377 static int ocfs2_debug_release(struct inode *inode, struct file *file) 378 { 379 return 0; 380 } 381 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, 382 size_t nbytes, loff_t *ppos) 383 { 384 return 0; 385 } 386 #endif /* CONFIG_DEBUG_FS */ 387 388 static const struct file_operations ocfs2_osb_debug_fops = { 389 .open = ocfs2_osb_debug_open, 390 .release = ocfs2_debug_release, 391 .read = ocfs2_debug_read, 392 .llseek = generic_file_llseek, 393 }; 394 395 static int ocfs2_sync_fs(struct super_block *sb, int wait) 396 { 397 int status; 398 tid_t target; 399 struct ocfs2_super *osb = OCFS2_SB(sb); 400 401 if (ocfs2_is_hard_readonly(osb)) 402 return -EROFS; 403 404 if (wait) { 405 status = ocfs2_flush_truncate_log(osb); 406 if (status < 0) 407 mlog_errno(status); 408 } else { 409 ocfs2_schedule_truncate_log_flush(osb, 0); 410 } 411 412 if (jbd2_journal_start_commit(osb->journal->j_journal, 413 &target)) { 414 if (wait) 415 jbd2_log_wait_commit(osb->journal->j_journal, 416 target); 417 } 418 return 0; 419 } 420 421 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino) 422 { 423 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA) 424 && (ino == USER_QUOTA_SYSTEM_INODE 425 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE)) 426 return 0; 427 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA) 428 && (ino == GROUP_QUOTA_SYSTEM_INODE 429 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE)) 430 return 0; 431 return 1; 432 } 433 434 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) 435 { 436 struct inode *new = NULL; 437 int status = 0; 438 int i; 439 440 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0); 441 if (IS_ERR(new)) { 442 status = PTR_ERR(new); 443 mlog_errno(status); 444 goto bail; 445 } 446 osb->root_inode = new; 447 448 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0); 449 if (IS_ERR(new)) { 450 status = PTR_ERR(new); 451 mlog_errno(status); 452 goto bail; 453 } 454 osb->sys_root_inode = new; 455 456 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE; 457 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) { 458 if (!ocfs2_need_system_inode(osb, i)) 459 continue; 460 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 461 if (!new) { 462 ocfs2_release_system_inodes(osb); 463 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL; 464 mlog_errno(status); 465 mlog(ML_ERROR, "Unable to load system inode %d, " 466 "possibly corrupt fs?", i); 467 goto bail; 468 } 469 // the array now has one ref, so drop this one 470 iput(new); 471 } 472 473 bail: 474 if (status) 475 mlog_errno(status); 476 return status; 477 } 478 479 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) 480 { 481 struct inode *new = NULL; 482 int status = 0; 483 int i; 484 485 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1; 486 i < NUM_SYSTEM_INODES; 487 i++) { 488 if (!ocfs2_need_system_inode(osb, i)) 489 continue; 490 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 491 if (!new) { 492 ocfs2_release_system_inodes(osb); 493 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL; 494 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n", 495 status, i, osb->slot_num); 496 goto bail; 497 } 498 /* the array now has one ref, so drop this one */ 499 iput(new); 500 } 501 502 bail: 503 if (status) 504 mlog_errno(status); 505 return status; 506 } 507 508 static void ocfs2_release_system_inodes(struct ocfs2_super *osb) 509 { 510 int i; 511 struct inode *inode; 512 513 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) { 514 inode = osb->global_system_inodes[i]; 515 if (inode) { 516 iput(inode); 517 osb->global_system_inodes[i] = NULL; 518 } 519 } 520 521 inode = osb->sys_root_inode; 522 if (inode) { 523 iput(inode); 524 osb->sys_root_inode = NULL; 525 } 526 527 inode = osb->root_inode; 528 if (inode) { 529 iput(inode); 530 osb->root_inode = NULL; 531 } 532 533 if (!osb->local_system_inodes) 534 return; 535 536 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) { 537 if (osb->local_system_inodes[i]) { 538 iput(osb->local_system_inodes[i]); 539 osb->local_system_inodes[i] = NULL; 540 } 541 } 542 543 kfree(osb->local_system_inodes); 544 osb->local_system_inodes = NULL; 545 } 546 547 /* We're allocating fs objects, use GFP_NOFS */ 548 static struct inode *ocfs2_alloc_inode(struct super_block *sb) 549 { 550 struct ocfs2_inode_info *oi; 551 552 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS); 553 if (!oi) 554 return NULL; 555 556 oi->i_sync_tid = 0; 557 oi->i_datasync_tid = 0; 558 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot)); 559 560 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode); 561 return &oi->vfs_inode; 562 } 563 564 static void ocfs2_free_inode(struct inode *inode) 565 { 566 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); 567 } 568 569 static unsigned long long ocfs2_max_file_offset(unsigned int bbits, 570 unsigned int cbits) 571 { 572 unsigned int bytes = 1 << cbits; 573 unsigned int trim = bytes; 574 unsigned int bitshift = 32; 575 576 /* 577 * i_size and all block offsets in ocfs2 are always 64 bits 578 * wide. i_clusters is 32 bits, in cluster-sized units. So on 579 * 64 bit platforms, cluster size will be the limiting factor. 580 */ 581 582 #if BITS_PER_LONG == 32 583 BUILD_BUG_ON(sizeof(sector_t) != 8); 584 /* 585 * We might be limited by page cache size. 586 */ 587 if (bytes > PAGE_SIZE) { 588 bytes = PAGE_SIZE; 589 trim = 1; 590 /* 591 * Shift by 31 here so that we don't get larger than 592 * MAX_LFS_FILESIZE 593 */ 594 bitshift = 31; 595 } 596 #endif 597 598 /* 599 * Trim by a whole cluster when we can actually approach the 600 * on-disk limits. Otherwise we can overflow i_clusters when 601 * an extent start is at the max offset. 602 */ 603 return (((unsigned long long)bytes) << bitshift) - trim; 604 } 605 606 static int ocfs2_remount(struct super_block *sb, int *flags, char *data) 607 { 608 int incompat_features; 609 int ret = 0; 610 struct mount_options parsed_options; 611 struct ocfs2_super *osb = OCFS2_SB(sb); 612 u32 tmp; 613 614 sync_filesystem(sb); 615 616 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) || 617 !ocfs2_check_set_options(sb, &parsed_options)) { 618 ret = -EINVAL; 619 goto out; 620 } 621 622 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL | 623 OCFS2_MOUNT_HB_NONE; 624 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) { 625 ret = -EINVAL; 626 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); 627 goto out; 628 } 629 630 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) != 631 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) { 632 ret = -EINVAL; 633 mlog(ML_ERROR, "Cannot change data mode on remount\n"); 634 goto out; 635 } 636 637 /* Probably don't want this on remount; it might 638 * mess with other nodes */ 639 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) && 640 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) { 641 ret = -EINVAL; 642 mlog(ML_ERROR, "Cannot enable inode64 on remount\n"); 643 goto out; 644 } 645 646 /* We're going to/from readonly mode. */ 647 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 648 /* Disable quota accounting before remounting RO */ 649 if (*flags & SB_RDONLY) { 650 ret = ocfs2_susp_quotas(osb, 0); 651 if (ret < 0) 652 goto out; 653 } 654 /* Lock here so the check of HARD_RO and the potential 655 * setting of SOFT_RO is atomic. */ 656 spin_lock(&osb->osb_lock); 657 if (osb->osb_flags & OCFS2_OSB_HARD_RO) { 658 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n"); 659 ret = -EROFS; 660 goto unlock_osb; 661 } 662 663 if (*flags & SB_RDONLY) { 664 sb->s_flags |= SB_RDONLY; 665 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 666 } else { 667 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) { 668 mlog(ML_ERROR, "Cannot remount RDWR " 669 "filesystem due to previous errors.\n"); 670 ret = -EROFS; 671 goto unlock_osb; 672 } 673 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP); 674 if (incompat_features) { 675 mlog(ML_ERROR, "Cannot remount RDWR because " 676 "of unsupported optional features " 677 "(%x).\n", incompat_features); 678 ret = -EINVAL; 679 goto unlock_osb; 680 } 681 sb->s_flags &= ~SB_RDONLY; 682 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO; 683 } 684 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags); 685 unlock_osb: 686 spin_unlock(&osb->osb_lock); 687 /* Enable quota accounting after remounting RW */ 688 if (!ret && !(*flags & SB_RDONLY)) { 689 if (sb_any_quota_suspended(sb)) 690 ret = ocfs2_susp_quotas(osb, 1); 691 else 692 ret = ocfs2_enable_quotas(osb); 693 if (ret < 0) { 694 /* Return back changes... */ 695 spin_lock(&osb->osb_lock); 696 sb->s_flags |= SB_RDONLY; 697 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 698 spin_unlock(&osb->osb_lock); 699 goto out; 700 } 701 } 702 } 703 704 if (!ret) { 705 /* Only save off the new mount options in case of a successful 706 * remount. */ 707 osb->s_mount_opt = parsed_options.mount_opt; 708 osb->s_atime_quantum = parsed_options.atime_quantum; 709 osb->preferred_slot = parsed_options.slot; 710 if (parsed_options.commit_interval) 711 osb->osb_commit_interval = parsed_options.commit_interval; 712 713 if (!ocfs2_is_hard_readonly(osb)) 714 ocfs2_set_journal_params(osb); 715 716 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 717 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? 718 SB_POSIXACL : 0); 719 } 720 out: 721 return ret; 722 } 723 724 static int ocfs2_sb_probe(struct super_block *sb, 725 struct buffer_head **bh, 726 int *sector_size, 727 struct ocfs2_blockcheck_stats *stats) 728 { 729 int status, tmpstat; 730 struct ocfs1_vol_disk_hdr *hdr; 731 struct ocfs2_dinode *di; 732 int blksize; 733 734 *bh = NULL; 735 736 /* may be > 512 */ 737 *sector_size = bdev_logical_block_size(sb->s_bdev); 738 if (*sector_size > OCFS2_MAX_BLOCKSIZE) { 739 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n", 740 *sector_size, OCFS2_MAX_BLOCKSIZE); 741 status = -EINVAL; 742 goto bail; 743 } 744 745 /* Can this really happen? */ 746 if (*sector_size < OCFS2_MIN_BLOCKSIZE) 747 *sector_size = OCFS2_MIN_BLOCKSIZE; 748 749 /* check block zero for old format */ 750 status = ocfs2_get_sector(sb, bh, 0, *sector_size); 751 if (status < 0) { 752 mlog_errno(status); 753 goto bail; 754 } 755 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data; 756 if (hdr->major_version == OCFS1_MAJOR_VERSION) { 757 mlog(ML_ERROR, "incompatible version: %u.%u\n", 758 hdr->major_version, hdr->minor_version); 759 status = -EINVAL; 760 } 761 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE, 762 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) { 763 mlog(ML_ERROR, "incompatible volume signature: %8s\n", 764 hdr->signature); 765 status = -EINVAL; 766 } 767 brelse(*bh); 768 *bh = NULL; 769 if (status < 0) { 770 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be " 771 "upgraded before mounting with ocfs v2\n"); 772 goto bail; 773 } 774 775 /* 776 * Now check at magic offset for 512, 1024, 2048, 4096 777 * blocksizes. 4096 is the maximum blocksize because it is 778 * the minimum clustersize. 779 */ 780 status = -EINVAL; 781 for (blksize = *sector_size; 782 blksize <= OCFS2_MAX_BLOCKSIZE; 783 blksize <<= 1) { 784 tmpstat = ocfs2_get_sector(sb, bh, 785 OCFS2_SUPER_BLOCK_BLKNO, 786 blksize); 787 if (tmpstat < 0) { 788 status = tmpstat; 789 mlog_errno(status); 790 break; 791 } 792 di = (struct ocfs2_dinode *) (*bh)->b_data; 793 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats)); 794 spin_lock_init(&stats->b_lock); 795 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats); 796 if (tmpstat < 0) { 797 brelse(*bh); 798 *bh = NULL; 799 } 800 if (tmpstat != -EAGAIN) { 801 status = tmpstat; 802 break; 803 } 804 } 805 806 bail: 807 return status; 808 } 809 810 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) 811 { 812 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL; 813 814 if (osb->s_mount_opt & hb_enabled) { 815 if (ocfs2_mount_local(osb)) { 816 mlog(ML_ERROR, "Cannot heartbeat on a locally " 817 "mounted device.\n"); 818 return -EINVAL; 819 } 820 if (ocfs2_userspace_stack(osb)) { 821 mlog(ML_ERROR, "Userspace stack expected, but " 822 "o2cb heartbeat arguments passed to mount\n"); 823 return -EINVAL; 824 } 825 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) && 826 !ocfs2_cluster_o2cb_global_heartbeat(osb)) || 827 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) && 828 ocfs2_cluster_o2cb_global_heartbeat(osb))) { 829 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n"); 830 return -EINVAL; 831 } 832 } 833 834 if (!(osb->s_mount_opt & hb_enabled)) { 835 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) && 836 !ocfs2_userspace_stack(osb)) { 837 mlog(ML_ERROR, "Heartbeat has to be started to mount " 838 "a read-write clustered device.\n"); 839 return -EINVAL; 840 } 841 } 842 843 return 0; 844 } 845 846 /* 847 * If we're using a userspace stack, mount should have passed 848 * a name that matches the disk. If not, mount should not 849 * have passed a stack. 850 */ 851 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb, 852 struct mount_options *mopt) 853 { 854 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) { 855 mlog(ML_ERROR, 856 "cluster stack passed to mount, but this filesystem " 857 "does not support it\n"); 858 return -EINVAL; 859 } 860 861 if (ocfs2_userspace_stack(osb) && 862 strncmp(osb->osb_cluster_stack, mopt->cluster_stack, 863 OCFS2_STACK_LABEL_LEN)) { 864 mlog(ML_ERROR, 865 "cluster stack passed to mount (\"%s\") does not " 866 "match the filesystem (\"%s\")\n", 867 mopt->cluster_stack, 868 osb->osb_cluster_stack); 869 return -EINVAL; 870 } 871 872 return 0; 873 } 874 875 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend) 876 { 877 int type; 878 struct super_block *sb = osb->sb; 879 unsigned int feature[OCFS2_MAXQUOTAS] = { 880 OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 881 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 882 int status = 0; 883 884 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 885 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 886 continue; 887 if (unsuspend) 888 status = dquot_resume(sb, type); 889 else { 890 struct ocfs2_mem_dqinfo *oinfo; 891 892 /* Cancel periodic syncing before suspending */ 893 oinfo = sb_dqinfo(sb, type)->dqi_priv; 894 cancel_delayed_work_sync(&oinfo->dqi_sync_work); 895 status = dquot_suspend(sb, type); 896 } 897 if (status < 0) 898 break; 899 } 900 if (status < 0) 901 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on " 902 "remount (error = %d).\n", status); 903 return status; 904 } 905 906 static int ocfs2_enable_quotas(struct ocfs2_super *osb) 907 { 908 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL }; 909 struct super_block *sb = osb->sb; 910 unsigned int feature[OCFS2_MAXQUOTAS] = { 911 OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 912 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 913 unsigned int ino[OCFS2_MAXQUOTAS] = { 914 LOCAL_USER_QUOTA_SYSTEM_INODE, 915 LOCAL_GROUP_QUOTA_SYSTEM_INODE }; 916 int status; 917 int type; 918 919 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE; 920 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 921 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 922 continue; 923 inode[type] = ocfs2_get_system_file_inode(osb, ino[type], 924 osb->slot_num); 925 if (!inode[type]) { 926 status = -ENOENT; 927 goto out_quota_off; 928 } 929 status = dquot_enable(inode[type], type, QFMT_OCFS2, 930 DQUOT_USAGE_ENABLED); 931 if (status < 0) 932 goto out_quota_off; 933 } 934 935 for (type = 0; type < OCFS2_MAXQUOTAS; type++) 936 iput(inode[type]); 937 return 0; 938 out_quota_off: 939 ocfs2_disable_quotas(osb); 940 for (type = 0; type < OCFS2_MAXQUOTAS; type++) 941 iput(inode[type]); 942 mlog_errno(status); 943 return status; 944 } 945 946 static void ocfs2_disable_quotas(struct ocfs2_super *osb) 947 { 948 int type; 949 struct inode *inode; 950 struct super_block *sb = osb->sb; 951 struct ocfs2_mem_dqinfo *oinfo; 952 953 /* We mostly ignore errors in this function because there's not much 954 * we can do when we see them */ 955 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 956 if (!sb_has_quota_loaded(sb, type)) 957 continue; 958 oinfo = sb_dqinfo(sb, type)->dqi_priv; 959 cancel_delayed_work_sync(&oinfo->dqi_sync_work); 960 inode = igrab(sb->s_dquot.files[type]); 961 /* Turn off quotas. This will remove all dquot structures from 962 * memory and so they will be automatically synced to global 963 * quota files */ 964 dquot_disable(sb, type, DQUOT_USAGE_ENABLED | 965 DQUOT_LIMITS_ENABLED); 966 if (!inode) 967 continue; 968 iput(inode); 969 } 970 } 971 972 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) 973 { 974 struct dentry *root; 975 int status, sector_size; 976 struct mount_options parsed_options; 977 struct inode *inode = NULL; 978 struct ocfs2_super *osb = NULL; 979 struct buffer_head *bh = NULL; 980 char nodestr[12]; 981 struct ocfs2_blockcheck_stats stats; 982 983 trace_ocfs2_fill_super(sb, data, silent); 984 985 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { 986 status = -EINVAL; 987 goto read_super_error; 988 } 989 990 /* probe for superblock */ 991 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats); 992 if (status < 0) { 993 mlog(ML_ERROR, "superblock probe failed!\n"); 994 goto read_super_error; 995 } 996 997 status = ocfs2_initialize_super(sb, bh, sector_size, &stats); 998 osb = OCFS2_SB(sb); 999 if (status < 0) { 1000 mlog_errno(status); 1001 goto read_super_error; 1002 } 1003 brelse(bh); 1004 bh = NULL; 1005 1006 if (!ocfs2_check_set_options(sb, &parsed_options)) { 1007 status = -EINVAL; 1008 goto read_super_error; 1009 } 1010 osb->s_mount_opt = parsed_options.mount_opt; 1011 osb->s_atime_quantum = parsed_options.atime_quantum; 1012 osb->preferred_slot = parsed_options.slot; 1013 osb->osb_commit_interval = parsed_options.commit_interval; 1014 1015 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt); 1016 osb->osb_resv_level = parsed_options.resv_level; 1017 osb->osb_dir_resv_level = parsed_options.resv_level; 1018 if (parsed_options.dir_resv_level == -1) 1019 osb->osb_dir_resv_level = parsed_options.resv_level; 1020 else 1021 osb->osb_dir_resv_level = parsed_options.dir_resv_level; 1022 1023 status = ocfs2_verify_userspace_stack(osb, &parsed_options); 1024 if (status) 1025 goto read_super_error; 1026 1027 sb->s_magic = OCFS2_SUPER_MAGIC; 1028 1029 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) | 1030 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0); 1031 1032 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY, 1033 * heartbeat=none */ 1034 if (bdev_read_only(sb->s_bdev)) { 1035 if (!sb_rdonly(sb)) { 1036 status = -EACCES; 1037 mlog(ML_ERROR, "Readonly device detected but readonly " 1038 "mount was not specified.\n"); 1039 goto read_super_error; 1040 } 1041 1042 /* You should not be able to start a local heartbeat 1043 * on a readonly device. */ 1044 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 1045 status = -EROFS; 1046 mlog(ML_ERROR, "Local heartbeat specified on readonly " 1047 "device.\n"); 1048 goto read_super_error; 1049 } 1050 1051 status = ocfs2_check_journals_nolocks(osb); 1052 if (status < 0) { 1053 if (status == -EROFS) 1054 mlog(ML_ERROR, "Recovery required on readonly " 1055 "file system, but write access is " 1056 "unavailable.\n"); 1057 else 1058 mlog_errno(status); 1059 goto read_super_error; 1060 } 1061 1062 ocfs2_set_ro_flag(osb, 1); 1063 1064 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. " 1065 "Cluster services will not be used for this mount. " 1066 "Recovery will be skipped.\n", osb->dev_str); 1067 } 1068 1069 if (!ocfs2_is_hard_readonly(osb)) { 1070 if (sb_rdonly(sb)) 1071 ocfs2_set_ro_flag(osb, 0); 1072 } 1073 1074 status = ocfs2_verify_heartbeat(osb); 1075 if (status < 0) { 1076 mlog_errno(status); 1077 goto read_super_error; 1078 } 1079 1080 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, 1081 ocfs2_debugfs_root); 1082 if (!osb->osb_debug_root) { 1083 status = -EINVAL; 1084 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n"); 1085 goto read_super_error; 1086 } 1087 1088 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR, 1089 osb->osb_debug_root, 1090 osb, 1091 &ocfs2_osb_debug_fops); 1092 if (!osb->osb_ctxt) { 1093 status = -EINVAL; 1094 mlog_errno(status); 1095 goto read_super_error; 1096 } 1097 1098 if (ocfs2_meta_ecc(osb)) { 1099 status = ocfs2_blockcheck_stats_debugfs_install( 1100 &osb->osb_ecc_stats, 1101 osb->osb_debug_root); 1102 if (status) { 1103 mlog(ML_ERROR, 1104 "Unable to create blockcheck statistics " 1105 "files\n"); 1106 goto read_super_error; 1107 } 1108 } 1109 1110 status = ocfs2_mount_volume(sb); 1111 if (status < 0) 1112 goto read_super_error; 1113 1114 if (osb->root_inode) 1115 inode = igrab(osb->root_inode); 1116 1117 if (!inode) { 1118 status = -EIO; 1119 mlog_errno(status); 1120 goto read_super_error; 1121 } 1122 1123 root = d_make_root(inode); 1124 if (!root) { 1125 status = -ENOMEM; 1126 mlog_errno(status); 1127 goto read_super_error; 1128 } 1129 1130 sb->s_root = root; 1131 1132 ocfs2_complete_mount_recovery(osb); 1133 1134 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL, 1135 &ocfs2_kset->kobj); 1136 if (!osb->osb_dev_kset) { 1137 status = -ENOMEM; 1138 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id); 1139 goto read_super_error; 1140 } 1141 1142 /* Create filecheck sysfs related directories/files at 1143 * /sys/fs/ocfs2/<devname>/filecheck */ 1144 if (ocfs2_filecheck_create_sysfs(osb)) { 1145 status = -ENOMEM; 1146 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at " 1147 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id); 1148 goto read_super_error; 1149 } 1150 1151 if (ocfs2_mount_local(osb)) 1152 snprintf(nodestr, sizeof(nodestr), "local"); 1153 else 1154 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1155 1156 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " 1157 "with %s data mode.\n", 1158 osb->dev_str, nodestr, osb->slot_num, 1159 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : 1160 "ordered"); 1161 1162 atomic_set(&osb->vol_state, VOLUME_MOUNTED); 1163 wake_up(&osb->osb_mount_event); 1164 1165 /* Now we can initialize quotas because we can afford to wait 1166 * for cluster locks recovery now. That also means that truncation 1167 * log recovery can happen but that waits for proper quota setup */ 1168 if (!sb_rdonly(sb)) { 1169 status = ocfs2_enable_quotas(osb); 1170 if (status < 0) { 1171 /* We have to err-out specially here because 1172 * s_root is already set */ 1173 mlog_errno(status); 1174 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1175 wake_up(&osb->osb_mount_event); 1176 return status; 1177 } 1178 } 1179 1180 ocfs2_complete_quota_recovery(osb); 1181 1182 /* Now we wake up again for processes waiting for quotas */ 1183 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS); 1184 wake_up(&osb->osb_mount_event); 1185 1186 /* Start this when the mount is almost sure of being successful */ 1187 ocfs2_orphan_scan_start(osb); 1188 1189 return status; 1190 1191 read_super_error: 1192 brelse(bh); 1193 1194 if (status) 1195 mlog_errno(status); 1196 1197 if (osb) { 1198 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1199 wake_up(&osb->osb_mount_event); 1200 ocfs2_dismount_volume(sb, 1); 1201 } 1202 1203 return status; 1204 } 1205 1206 static struct dentry *ocfs2_mount(struct file_system_type *fs_type, 1207 int flags, 1208 const char *dev_name, 1209 void *data) 1210 { 1211 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super); 1212 } 1213 1214 static struct file_system_type ocfs2_fs_type = { 1215 .owner = THIS_MODULE, 1216 .name = "ocfs2", 1217 .mount = ocfs2_mount, 1218 .kill_sb = kill_block_super, 1219 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, 1220 .next = NULL 1221 }; 1222 MODULE_ALIAS_FS("ocfs2"); 1223 1224 static int ocfs2_check_set_options(struct super_block *sb, 1225 struct mount_options *options) 1226 { 1227 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA && 1228 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1229 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { 1230 mlog(ML_ERROR, "User quotas were requested, but this " 1231 "filesystem does not have the feature enabled.\n"); 1232 return 0; 1233 } 1234 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA && 1235 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1236 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { 1237 mlog(ML_ERROR, "Group quotas were requested, but this " 1238 "filesystem does not have the feature enabled.\n"); 1239 return 0; 1240 } 1241 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL && 1242 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) { 1243 mlog(ML_ERROR, "ACL support requested but extended attributes " 1244 "feature is not enabled\n"); 1245 return 0; 1246 } 1247 /* No ACL setting specified? Use XATTR feature... */ 1248 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL | 1249 OCFS2_MOUNT_NO_POSIX_ACL))) { 1250 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) 1251 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1252 else 1253 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; 1254 } 1255 return 1; 1256 } 1257 1258 static int ocfs2_parse_options(struct super_block *sb, 1259 char *options, 1260 struct mount_options *mopt, 1261 int is_remount) 1262 { 1263 int status, user_stack = 0; 1264 char *p; 1265 u32 tmp; 1266 int token, option; 1267 substring_t args[MAX_OPT_ARGS]; 1268 1269 trace_ocfs2_parse_options(is_remount, options ? options : "(none)"); 1270 1271 mopt->commit_interval = 0; 1272 mopt->mount_opt = OCFS2_MOUNT_NOINTR; 1273 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 1274 mopt->slot = OCFS2_INVALID_SLOT; 1275 mopt->localalloc_opt = -1; 1276 mopt->cluster_stack[0] = '\0'; 1277 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL; 1278 mopt->dir_resv_level = -1; 1279 1280 if (!options) { 1281 status = 1; 1282 goto bail; 1283 } 1284 1285 while ((p = strsep(&options, ",")) != NULL) { 1286 if (!*p) 1287 continue; 1288 1289 token = match_token(p, tokens, args); 1290 switch (token) { 1291 case Opt_hb_local: 1292 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; 1293 break; 1294 case Opt_hb_none: 1295 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE; 1296 break; 1297 case Opt_hb_global: 1298 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL; 1299 break; 1300 case Opt_barrier: 1301 if (match_int(&args[0], &option)) { 1302 status = 0; 1303 goto bail; 1304 } 1305 if (option) 1306 mopt->mount_opt |= OCFS2_MOUNT_BARRIER; 1307 else 1308 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; 1309 break; 1310 case Opt_intr: 1311 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; 1312 break; 1313 case Opt_nointr: 1314 mopt->mount_opt |= OCFS2_MOUNT_NOINTR; 1315 break; 1316 case Opt_err_panic: 1317 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT; 1318 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS; 1319 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 1320 break; 1321 case Opt_err_ro: 1322 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT; 1323 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1324 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS; 1325 break; 1326 case Opt_err_cont: 1327 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS; 1328 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1329 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT; 1330 break; 1331 case Opt_data_ordered: 1332 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; 1333 break; 1334 case Opt_data_writeback: 1335 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; 1336 break; 1337 case Opt_user_xattr: 1338 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR; 1339 break; 1340 case Opt_nouser_xattr: 1341 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR; 1342 break; 1343 case Opt_atime_quantum: 1344 if (match_int(&args[0], &option)) { 1345 status = 0; 1346 goto bail; 1347 } 1348 if (option >= 0) 1349 mopt->atime_quantum = option; 1350 break; 1351 case Opt_slot: 1352 if (match_int(&args[0], &option)) { 1353 status = 0; 1354 goto bail; 1355 } 1356 if (option) 1357 mopt->slot = (s16)option; 1358 break; 1359 case Opt_commit: 1360 if (match_int(&args[0], &option)) { 1361 status = 0; 1362 goto bail; 1363 } 1364 if (option < 0) 1365 return 0; 1366 if (option == 0) 1367 option = JBD2_DEFAULT_MAX_COMMIT_AGE; 1368 mopt->commit_interval = HZ * option; 1369 break; 1370 case Opt_localalloc: 1371 if (match_int(&args[0], &option)) { 1372 status = 0; 1373 goto bail; 1374 } 1375 if (option >= 0) 1376 mopt->localalloc_opt = option; 1377 break; 1378 case Opt_localflocks: 1379 /* 1380 * Changing this during remount could race 1381 * flock() requests, or "unbalance" existing 1382 * ones (e.g., a lock is taken in one mode but 1383 * dropped in the other). If users care enough 1384 * to flip locking modes during remount, we 1385 * could add a "local" flag to individual 1386 * flock structures for proper tracking of 1387 * state. 1388 */ 1389 if (!is_remount) 1390 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS; 1391 break; 1392 case Opt_stack: 1393 /* Check both that the option we were passed 1394 * is of the right length and that it is a proper 1395 * string of the right length. 1396 */ 1397 if (((args[0].to - args[0].from) != 1398 OCFS2_STACK_LABEL_LEN) || 1399 (strnlen(args[0].from, 1400 OCFS2_STACK_LABEL_LEN) != 1401 OCFS2_STACK_LABEL_LEN)) { 1402 mlog(ML_ERROR, 1403 "Invalid cluster_stack option\n"); 1404 status = 0; 1405 goto bail; 1406 } 1407 memcpy(mopt->cluster_stack, args[0].from, 1408 OCFS2_STACK_LABEL_LEN); 1409 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; 1410 /* 1411 * Open code the memcmp here as we don't have 1412 * an osb to pass to 1413 * ocfs2_userspace_stack(). 1414 */ 1415 if (memcmp(mopt->cluster_stack, 1416 OCFS2_CLASSIC_CLUSTER_STACK, 1417 OCFS2_STACK_LABEL_LEN)) 1418 user_stack = 1; 1419 break; 1420 case Opt_inode64: 1421 mopt->mount_opt |= OCFS2_MOUNT_INODE64; 1422 break; 1423 case Opt_usrquota: 1424 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA; 1425 break; 1426 case Opt_grpquota: 1427 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA; 1428 break; 1429 case Opt_coherency_buffered: 1430 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED; 1431 break; 1432 case Opt_coherency_full: 1433 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED; 1434 break; 1435 case Opt_acl: 1436 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1437 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL; 1438 break; 1439 case Opt_noacl: 1440 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; 1441 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 1442 break; 1443 case Opt_resv_level: 1444 if (is_remount) 1445 break; 1446 if (match_int(&args[0], &option)) { 1447 status = 0; 1448 goto bail; 1449 } 1450 if (option >= OCFS2_MIN_RESV_LEVEL && 1451 option < OCFS2_MAX_RESV_LEVEL) 1452 mopt->resv_level = option; 1453 break; 1454 case Opt_dir_resv_level: 1455 if (is_remount) 1456 break; 1457 if (match_int(&args[0], &option)) { 1458 status = 0; 1459 goto bail; 1460 } 1461 if (option >= OCFS2_MIN_RESV_LEVEL && 1462 option < OCFS2_MAX_RESV_LEVEL) 1463 mopt->dir_resv_level = option; 1464 break; 1465 case Opt_journal_async_commit: 1466 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT; 1467 break; 1468 default: 1469 mlog(ML_ERROR, 1470 "Unrecognized mount option \"%s\" " 1471 "or missing value\n", p); 1472 status = 0; 1473 goto bail; 1474 } 1475 } 1476 1477 if (user_stack == 0) { 1478 /* Ensure only one heartbeat mode */ 1479 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | 1480 OCFS2_MOUNT_HB_GLOBAL | 1481 OCFS2_MOUNT_HB_NONE); 1482 if (hweight32(tmp) != 1) { 1483 mlog(ML_ERROR, "Invalid heartbeat mount options\n"); 1484 status = 0; 1485 goto bail; 1486 } 1487 } 1488 1489 status = 1; 1490 1491 bail: 1492 return status; 1493 } 1494 1495 static int ocfs2_show_options(struct seq_file *s, struct dentry *root) 1496 { 1497 struct ocfs2_super *osb = OCFS2_SB(root->d_sb); 1498 unsigned long opts = osb->s_mount_opt; 1499 unsigned int local_alloc_megs; 1500 1501 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) { 1502 seq_printf(s, ",_netdev"); 1503 if (opts & OCFS2_MOUNT_HB_LOCAL) 1504 seq_printf(s, ",%s", OCFS2_HB_LOCAL); 1505 else 1506 seq_printf(s, ",%s", OCFS2_HB_GLOBAL); 1507 } else 1508 seq_printf(s, ",%s", OCFS2_HB_NONE); 1509 1510 if (opts & OCFS2_MOUNT_NOINTR) 1511 seq_printf(s, ",nointr"); 1512 1513 if (opts & OCFS2_MOUNT_DATA_WRITEBACK) 1514 seq_printf(s, ",data=writeback"); 1515 else 1516 seq_printf(s, ",data=ordered"); 1517 1518 if (opts & OCFS2_MOUNT_BARRIER) 1519 seq_printf(s, ",barrier=1"); 1520 1521 if (opts & OCFS2_MOUNT_ERRORS_PANIC) 1522 seq_printf(s, ",errors=panic"); 1523 else if (opts & OCFS2_MOUNT_ERRORS_CONT) 1524 seq_printf(s, ",errors=continue"); 1525 else 1526 seq_printf(s, ",errors=remount-ro"); 1527 1528 if (osb->preferred_slot != OCFS2_INVALID_SLOT) 1529 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); 1530 1531 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); 1532 1533 if (osb->osb_commit_interval) 1534 seq_printf(s, ",commit=%u", 1535 (unsigned) (osb->osb_commit_interval / HZ)); 1536 1537 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits); 1538 if (local_alloc_megs != ocfs2_la_default_mb(osb)) 1539 seq_printf(s, ",localalloc=%d", local_alloc_megs); 1540 1541 if (opts & OCFS2_MOUNT_LOCALFLOCKS) 1542 seq_printf(s, ",localflocks,"); 1543 1544 if (osb->osb_cluster_stack[0]) 1545 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack, 1546 OCFS2_STACK_LABEL_LEN); 1547 if (opts & OCFS2_MOUNT_USRQUOTA) 1548 seq_printf(s, ",usrquota"); 1549 if (opts & OCFS2_MOUNT_GRPQUOTA) 1550 seq_printf(s, ",grpquota"); 1551 1552 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED) 1553 seq_printf(s, ",coherency=buffered"); 1554 else 1555 seq_printf(s, ",coherency=full"); 1556 1557 if (opts & OCFS2_MOUNT_NOUSERXATTR) 1558 seq_printf(s, ",nouser_xattr"); 1559 else 1560 seq_printf(s, ",user_xattr"); 1561 1562 if (opts & OCFS2_MOUNT_INODE64) 1563 seq_printf(s, ",inode64"); 1564 1565 if (opts & OCFS2_MOUNT_POSIX_ACL) 1566 seq_printf(s, ",acl"); 1567 else 1568 seq_printf(s, ",noacl"); 1569 1570 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL) 1571 seq_printf(s, ",resv_level=%d", osb->osb_resv_level); 1572 1573 if (osb->osb_dir_resv_level != osb->osb_resv_level) 1574 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level); 1575 1576 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT) 1577 seq_printf(s, ",journal_async_commit"); 1578 1579 return 0; 1580 } 1581 1582 static int __init ocfs2_init(void) 1583 { 1584 int status; 1585 1586 status = init_ocfs2_uptodate_cache(); 1587 if (status < 0) 1588 goto out1; 1589 1590 status = ocfs2_initialize_mem_caches(); 1591 if (status < 0) 1592 goto out2; 1593 1594 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); 1595 if (!ocfs2_debugfs_root) { 1596 status = -ENOMEM; 1597 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n"); 1598 goto out3; 1599 } 1600 1601 ocfs2_set_locking_protocol(); 1602 1603 status = register_quota_format(&ocfs2_quota_format); 1604 if (status < 0) 1605 goto out3; 1606 status = register_filesystem(&ocfs2_fs_type); 1607 if (!status) 1608 return 0; 1609 1610 unregister_quota_format(&ocfs2_quota_format); 1611 out3: 1612 debugfs_remove(ocfs2_debugfs_root); 1613 ocfs2_free_mem_caches(); 1614 out2: 1615 exit_ocfs2_uptodate_cache(); 1616 out1: 1617 mlog_errno(status); 1618 return status; 1619 } 1620 1621 static void __exit ocfs2_exit(void) 1622 { 1623 unregister_quota_format(&ocfs2_quota_format); 1624 1625 debugfs_remove(ocfs2_debugfs_root); 1626 1627 ocfs2_free_mem_caches(); 1628 1629 unregister_filesystem(&ocfs2_fs_type); 1630 1631 exit_ocfs2_uptodate_cache(); 1632 } 1633 1634 static void ocfs2_put_super(struct super_block *sb) 1635 { 1636 trace_ocfs2_put_super(sb); 1637 1638 ocfs2_sync_blockdev(sb); 1639 ocfs2_dismount_volume(sb, 0); 1640 } 1641 1642 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) 1643 { 1644 struct ocfs2_super *osb; 1645 u32 numbits, freebits; 1646 int status; 1647 struct ocfs2_dinode *bm_lock; 1648 struct buffer_head *bh = NULL; 1649 struct inode *inode = NULL; 1650 1651 trace_ocfs2_statfs(dentry->d_sb, buf); 1652 1653 osb = OCFS2_SB(dentry->d_sb); 1654 1655 inode = ocfs2_get_system_file_inode(osb, 1656 GLOBAL_BITMAP_SYSTEM_INODE, 1657 OCFS2_INVALID_SLOT); 1658 if (!inode) { 1659 mlog(ML_ERROR, "failed to get bitmap inode\n"); 1660 status = -EIO; 1661 goto bail; 1662 } 1663 1664 status = ocfs2_inode_lock(inode, &bh, 0); 1665 if (status < 0) { 1666 mlog_errno(status); 1667 goto bail; 1668 } 1669 1670 bm_lock = (struct ocfs2_dinode *) bh->b_data; 1671 1672 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); 1673 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); 1674 1675 buf->f_type = OCFS2_SUPER_MAGIC; 1676 buf->f_bsize = dentry->d_sb->s_blocksize; 1677 buf->f_namelen = OCFS2_MAX_FILENAME_LEN; 1678 buf->f_blocks = ((sector_t) numbits) * 1679 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1680 buf->f_bfree = ((sector_t) freebits) * 1681 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1682 buf->f_bavail = buf->f_bfree; 1683 buf->f_files = numbits; 1684 buf->f_ffree = freebits; 1685 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN) 1686 & 0xFFFFFFFFUL; 1687 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN, 1688 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL; 1689 1690 brelse(bh); 1691 1692 ocfs2_inode_unlock(inode, 0); 1693 status = 0; 1694 bail: 1695 iput(inode); 1696 1697 if (status) 1698 mlog_errno(status); 1699 1700 return status; 1701 } 1702 1703 static void ocfs2_inode_init_once(void *data) 1704 { 1705 struct ocfs2_inode_info *oi = data; 1706 1707 oi->ip_flags = 0; 1708 oi->ip_open_count = 0; 1709 spin_lock_init(&oi->ip_lock); 1710 ocfs2_extent_map_init(&oi->vfs_inode); 1711 INIT_LIST_HEAD(&oi->ip_io_markers); 1712 INIT_LIST_HEAD(&oi->ip_unwritten_list); 1713 oi->ip_dir_start_lookup = 0; 1714 init_rwsem(&oi->ip_alloc_sem); 1715 init_rwsem(&oi->ip_xattr_sem); 1716 mutex_init(&oi->ip_io_mutex); 1717 1718 oi->ip_blkno = 0ULL; 1719 oi->ip_clusters = 0; 1720 1721 ocfs2_resv_init_once(&oi->ip_la_data_resv); 1722 1723 ocfs2_lock_res_init_once(&oi->ip_rw_lockres); 1724 ocfs2_lock_res_init_once(&oi->ip_inode_lockres); 1725 ocfs2_lock_res_init_once(&oi->ip_open_lockres); 1726 1727 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode), 1728 &ocfs2_inode_caching_ops); 1729 1730 inode_init_once(&oi->vfs_inode); 1731 } 1732 1733 static int ocfs2_initialize_mem_caches(void) 1734 { 1735 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", 1736 sizeof(struct ocfs2_inode_info), 1737 0, 1738 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1739 SLAB_MEM_SPREAD|SLAB_ACCOUNT), 1740 ocfs2_inode_init_once); 1741 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache", 1742 sizeof(struct ocfs2_dquot), 1743 0, 1744 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1745 SLAB_MEM_SPREAD), 1746 NULL); 1747 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache", 1748 sizeof(struct ocfs2_quota_chunk), 1749 0, 1750 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), 1751 NULL); 1752 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep || 1753 !ocfs2_qf_chunk_cachep) { 1754 kmem_cache_destroy(ocfs2_inode_cachep); 1755 kmem_cache_destroy(ocfs2_dquot_cachep); 1756 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1757 return -ENOMEM; 1758 } 1759 1760 return 0; 1761 } 1762 1763 static void ocfs2_free_mem_caches(void) 1764 { 1765 /* 1766 * Make sure all delayed rcu free inodes are flushed before we 1767 * destroy cache. 1768 */ 1769 rcu_barrier(); 1770 kmem_cache_destroy(ocfs2_inode_cachep); 1771 ocfs2_inode_cachep = NULL; 1772 1773 kmem_cache_destroy(ocfs2_dquot_cachep); 1774 ocfs2_dquot_cachep = NULL; 1775 1776 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1777 ocfs2_qf_chunk_cachep = NULL; 1778 } 1779 1780 static int ocfs2_get_sector(struct super_block *sb, 1781 struct buffer_head **bh, 1782 int block, 1783 int sect_size) 1784 { 1785 if (!sb_set_blocksize(sb, sect_size)) { 1786 mlog(ML_ERROR, "unable to set blocksize\n"); 1787 return -EIO; 1788 } 1789 1790 *bh = sb_getblk(sb, block); 1791 if (!*bh) { 1792 mlog_errno(-ENOMEM); 1793 return -ENOMEM; 1794 } 1795 lock_buffer(*bh); 1796 if (!buffer_dirty(*bh)) 1797 clear_buffer_uptodate(*bh); 1798 unlock_buffer(*bh); 1799 ll_rw_block(REQ_OP_READ, 0, 1, bh); 1800 wait_on_buffer(*bh); 1801 if (!buffer_uptodate(*bh)) { 1802 mlog_errno(-EIO); 1803 brelse(*bh); 1804 *bh = NULL; 1805 return -EIO; 1806 } 1807 1808 return 0; 1809 } 1810 1811 static int ocfs2_mount_volume(struct super_block *sb) 1812 { 1813 int status = 0; 1814 int unlock_super = 0; 1815 struct ocfs2_super *osb = OCFS2_SB(sb); 1816 1817 if (ocfs2_is_hard_readonly(osb)) 1818 goto leave; 1819 1820 mutex_init(&osb->obs_trim_fs_mutex); 1821 1822 status = ocfs2_dlm_init(osb); 1823 if (status < 0) { 1824 mlog_errno(status); 1825 if (status == -EBADR && ocfs2_userspace_stack(osb)) 1826 mlog(ML_ERROR, "couldn't mount because cluster name on" 1827 " disk does not match the running cluster name.\n"); 1828 goto leave; 1829 } 1830 1831 status = ocfs2_super_lock(osb, 1); 1832 if (status < 0) { 1833 mlog_errno(status); 1834 goto leave; 1835 } 1836 unlock_super = 1; 1837 1838 /* This will load up the node map and add ourselves to it. */ 1839 status = ocfs2_find_slot(osb); 1840 if (status < 0) { 1841 mlog_errno(status); 1842 goto leave; 1843 } 1844 1845 /* load all node-local system inodes */ 1846 status = ocfs2_init_local_system_inodes(osb); 1847 if (status < 0) { 1848 mlog_errno(status); 1849 goto leave; 1850 } 1851 1852 status = ocfs2_check_volume(osb); 1853 if (status < 0) { 1854 mlog_errno(status); 1855 goto leave; 1856 } 1857 1858 status = ocfs2_truncate_log_init(osb); 1859 if (status < 0) 1860 mlog_errno(status); 1861 1862 leave: 1863 if (unlock_super) 1864 ocfs2_super_unlock(osb, 1); 1865 1866 return status; 1867 } 1868 1869 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) 1870 { 1871 int tmp, hangup_needed = 0; 1872 struct ocfs2_super *osb = NULL; 1873 char nodestr[12]; 1874 1875 trace_ocfs2_dismount_volume(sb); 1876 1877 BUG_ON(!sb); 1878 osb = OCFS2_SB(sb); 1879 BUG_ON(!osb); 1880 1881 /* Remove file check sysfs related directores/files, 1882 * and wait for the pending file check operations */ 1883 ocfs2_filecheck_remove_sysfs(osb); 1884 1885 kset_unregister(osb->osb_dev_kset); 1886 1887 debugfs_remove(osb->osb_ctxt); 1888 1889 /* Orphan scan should be stopped as early as possible */ 1890 ocfs2_orphan_scan_stop(osb); 1891 1892 ocfs2_disable_quotas(osb); 1893 1894 /* All dquots should be freed by now */ 1895 WARN_ON(!llist_empty(&osb->dquot_drop_list)); 1896 /* Wait for worker to be done with the work structure in osb */ 1897 cancel_work_sync(&osb->dquot_drop_work); 1898 1899 ocfs2_shutdown_local_alloc(osb); 1900 1901 ocfs2_truncate_log_shutdown(osb); 1902 1903 /* This will disable recovery and flush any recovery work. */ 1904 ocfs2_recovery_exit(osb); 1905 1906 ocfs2_journal_shutdown(osb); 1907 1908 ocfs2_sync_blockdev(sb); 1909 1910 ocfs2_purge_refcount_trees(osb); 1911 1912 /* No cluster connection means we've failed during mount, so skip 1913 * all the steps which depended on that to complete. */ 1914 if (osb->cconn) { 1915 tmp = ocfs2_super_lock(osb, 1); 1916 if (tmp < 0) { 1917 mlog_errno(tmp); 1918 return; 1919 } 1920 } 1921 1922 if (osb->slot_num != OCFS2_INVALID_SLOT) 1923 ocfs2_put_slot(osb); 1924 1925 if (osb->cconn) 1926 ocfs2_super_unlock(osb, 1); 1927 1928 ocfs2_release_system_inodes(osb); 1929 1930 /* 1931 * If we're dismounting due to mount error, mount.ocfs2 will clean 1932 * up heartbeat. If we're a local mount, there is no heartbeat. 1933 * If we failed before we got a uuid_str yet, we can't stop 1934 * heartbeat. Otherwise, do it. 1935 */ 1936 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str && 1937 !ocfs2_is_hard_readonly(osb)) 1938 hangup_needed = 1; 1939 1940 if (osb->cconn) 1941 ocfs2_dlm_shutdown(osb, hangup_needed); 1942 1943 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats); 1944 debugfs_remove(osb->osb_debug_root); 1945 1946 if (hangup_needed) 1947 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str)); 1948 1949 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); 1950 1951 if (ocfs2_mount_local(osb)) 1952 snprintf(nodestr, sizeof(nodestr), "local"); 1953 else 1954 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1955 1956 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", 1957 osb->dev_str, nodestr); 1958 1959 ocfs2_delete_osb(osb); 1960 kfree(osb); 1961 sb->s_dev = 0; 1962 sb->s_fs_info = NULL; 1963 } 1964 1965 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, 1966 unsigned uuid_bytes) 1967 { 1968 int i, ret; 1969 char *ptr; 1970 1971 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); 1972 1973 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); 1974 if (osb->uuid_str == NULL) 1975 return -ENOMEM; 1976 1977 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { 1978 /* print with null */ 1979 ret = snprintf(ptr, 3, "%02X", uuid[i]); 1980 if (ret != 2) /* drop super cleans up */ 1981 return -EINVAL; 1982 /* then only advance past the last char */ 1983 ptr += 2; 1984 } 1985 1986 return 0; 1987 } 1988 1989 /* Make sure entire volume is addressable by our journal. Requires 1990 osb_clusters_at_boot to be valid and for the journal to have been 1991 initialized by ocfs2_journal_init(). */ 1992 static int ocfs2_journal_addressable(struct ocfs2_super *osb) 1993 { 1994 int status = 0; 1995 u64 max_block = 1996 ocfs2_clusters_to_blocks(osb->sb, 1997 osb->osb_clusters_at_boot) - 1; 1998 1999 /* 32-bit block number is always OK. */ 2000 if (max_block <= (u32)~0ULL) 2001 goto out; 2002 2003 /* Volume is "huge", so see if our journal is new enough to 2004 support it. */ 2005 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb, 2006 OCFS2_FEATURE_COMPAT_JBD2_SB) && 2007 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0, 2008 JBD2_FEATURE_INCOMPAT_64BIT))) { 2009 mlog(ML_ERROR, "The journal cannot address the entire volume. " 2010 "Enable the 'block64' journal option with tunefs.ocfs2"); 2011 status = -EFBIG; 2012 goto out; 2013 } 2014 2015 out: 2016 return status; 2017 } 2018 2019 static int ocfs2_initialize_super(struct super_block *sb, 2020 struct buffer_head *bh, 2021 int sector_size, 2022 struct ocfs2_blockcheck_stats *stats) 2023 { 2024 int status; 2025 int i, cbits, bbits; 2026 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; 2027 struct inode *inode = NULL; 2028 struct ocfs2_journal *journal; 2029 struct ocfs2_super *osb; 2030 u64 total_blocks; 2031 2032 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); 2033 if (!osb) { 2034 status = -ENOMEM; 2035 mlog_errno(status); 2036 goto bail; 2037 } 2038 2039 sb->s_fs_info = osb; 2040 sb->s_op = &ocfs2_sops; 2041 sb->s_d_op = &ocfs2_dentry_ops; 2042 sb->s_export_op = &ocfs2_export_ops; 2043 sb->s_qcop = &dquot_quotactl_sysfile_ops; 2044 sb->dq_op = &ocfs2_quota_operations; 2045 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP; 2046 sb->s_xattr = ocfs2_xattr_handlers; 2047 sb->s_time_gran = 1; 2048 sb->s_flags |= SB_NOATIME; 2049 /* this is needed to support O_LARGEFILE */ 2050 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2051 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); 2052 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); 2053 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid, 2054 sizeof(di->id2.i_super.s_uuid)); 2055 2056 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1; 2057 2058 for (i = 0; i < 3; i++) 2059 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]); 2060 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2061 2062 osb->sb = sb; 2063 osb->s_sectsize_bits = blksize_bits(sector_size); 2064 BUG_ON(!osb->s_sectsize_bits); 2065 2066 spin_lock_init(&osb->dc_task_lock); 2067 init_waitqueue_head(&osb->dc_event); 2068 osb->dc_work_sequence = 0; 2069 osb->dc_wake_sequence = 0; 2070 INIT_LIST_HEAD(&osb->blocked_lock_list); 2071 osb->blocked_lock_count = 0; 2072 spin_lock_init(&osb->osb_lock); 2073 spin_lock_init(&osb->osb_xattr_lock); 2074 ocfs2_init_steal_slots(osb); 2075 2076 mutex_init(&osb->system_file_mutex); 2077 2078 atomic_set(&osb->alloc_stats.moves, 0); 2079 atomic_set(&osb->alloc_stats.local_data, 0); 2080 atomic_set(&osb->alloc_stats.bitmap_data, 0); 2081 atomic_set(&osb->alloc_stats.bg_allocs, 0); 2082 atomic_set(&osb->alloc_stats.bg_extends, 0); 2083 2084 /* Copy the blockcheck stats from the superblock probe */ 2085 osb->osb_ecc_stats = *stats; 2086 2087 ocfs2_init_node_maps(osb); 2088 2089 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", 2090 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); 2091 2092 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); 2093 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { 2094 mlog(ML_ERROR, "Invalid number of node slots (%u)\n", 2095 osb->max_slots); 2096 status = -EINVAL; 2097 goto bail; 2098 } 2099 2100 ocfs2_orphan_scan_init(osb); 2101 2102 status = ocfs2_recovery_init(osb); 2103 if (status) { 2104 mlog(ML_ERROR, "Unable to initialize recovery state\n"); 2105 mlog_errno(status); 2106 goto bail; 2107 } 2108 2109 init_waitqueue_head(&osb->checkpoint_event); 2110 2111 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 2112 2113 osb->slot_num = OCFS2_INVALID_SLOT; 2114 2115 osb->s_xattr_inline_size = le16_to_cpu( 2116 di->id2.i_super.s_xattr_inline_size); 2117 2118 osb->local_alloc_state = OCFS2_LA_UNUSED; 2119 osb->local_alloc_bh = NULL; 2120 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker); 2121 2122 init_waitqueue_head(&osb->osb_mount_event); 2123 2124 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap); 2125 if (status) { 2126 mlog_errno(status); 2127 goto bail; 2128 } 2129 2130 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); 2131 if (!osb->vol_label) { 2132 mlog(ML_ERROR, "unable to alloc vol label\n"); 2133 status = -ENOMEM; 2134 goto bail; 2135 } 2136 2137 osb->slot_recovery_generations = 2138 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), 2139 GFP_KERNEL); 2140 if (!osb->slot_recovery_generations) { 2141 status = -ENOMEM; 2142 mlog_errno(status); 2143 goto bail; 2144 } 2145 2146 init_waitqueue_head(&osb->osb_wipe_event); 2147 osb->osb_orphan_wipes = kcalloc(osb->max_slots, 2148 sizeof(*osb->osb_orphan_wipes), 2149 GFP_KERNEL); 2150 if (!osb->osb_orphan_wipes) { 2151 status = -ENOMEM; 2152 mlog_errno(status); 2153 goto bail; 2154 } 2155 2156 osb->osb_rf_lock_tree = RB_ROOT; 2157 2158 osb->s_feature_compat = 2159 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); 2160 osb->s_feature_ro_compat = 2161 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); 2162 osb->s_feature_incompat = 2163 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); 2164 2165 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { 2166 mlog(ML_ERROR, "couldn't mount because of unsupported " 2167 "optional features (%x).\n", i); 2168 status = -EINVAL; 2169 goto bail; 2170 } 2171 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { 2172 mlog(ML_ERROR, "couldn't mount RDWR because of " 2173 "unsupported optional features (%x).\n", i); 2174 status = -EINVAL; 2175 goto bail; 2176 } 2177 2178 if (ocfs2_clusterinfo_valid(osb)) { 2179 osb->osb_stackflags = 2180 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags; 2181 strlcpy(osb->osb_cluster_stack, 2182 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, 2183 OCFS2_STACK_LABEL_LEN + 1); 2184 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) { 2185 mlog(ML_ERROR, 2186 "couldn't mount because of an invalid " 2187 "cluster stack label (%s) \n", 2188 osb->osb_cluster_stack); 2189 status = -EINVAL; 2190 goto bail; 2191 } 2192 strlcpy(osb->osb_cluster_name, 2193 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster, 2194 OCFS2_CLUSTER_NAME_LEN + 1); 2195 } else { 2196 /* The empty string is identical with classic tools that 2197 * don't know about s_cluster_info. */ 2198 osb->osb_cluster_stack[0] = '\0'; 2199 } 2200 2201 get_random_bytes(&osb->s_next_generation, sizeof(u32)); 2202 2203 /* FIXME 2204 * This should be done in ocfs2_journal_init(), but unknown 2205 * ordering issues will cause the filesystem to crash. 2206 * If anyone wants to figure out what part of the code 2207 * refers to osb->journal before ocfs2_journal_init() is run, 2208 * be my guest. 2209 */ 2210 /* initialize our journal structure */ 2211 2212 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL); 2213 if (!journal) { 2214 mlog(ML_ERROR, "unable to alloc journal\n"); 2215 status = -ENOMEM; 2216 goto bail; 2217 } 2218 osb->journal = journal; 2219 journal->j_osb = osb; 2220 2221 atomic_set(&journal->j_num_trans, 0); 2222 init_rwsem(&journal->j_trans_barrier); 2223 init_waitqueue_head(&journal->j_checkpointed); 2224 spin_lock_init(&journal->j_lock); 2225 journal->j_trans_id = (unsigned long) 1; 2226 INIT_LIST_HEAD(&journal->j_la_cleanups); 2227 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); 2228 journal->j_state = OCFS2_JOURNAL_FREE; 2229 2230 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs); 2231 init_llist_head(&osb->dquot_drop_list); 2232 2233 /* get some pseudo constants for clustersize bits */ 2234 osb->s_clustersize_bits = 2235 le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2236 osb->s_clustersize = 1 << osb->s_clustersize_bits; 2237 2238 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || 2239 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { 2240 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", 2241 osb->s_clustersize); 2242 status = -EINVAL; 2243 goto bail; 2244 } 2245 2246 total_blocks = ocfs2_clusters_to_blocks(osb->sb, 2247 le32_to_cpu(di->i_clusters)); 2248 2249 status = generic_check_addressable(osb->sb->s_blocksize_bits, 2250 total_blocks); 2251 if (status) { 2252 mlog(ML_ERROR, "Volume too large " 2253 "to mount safely on this system"); 2254 status = -EFBIG; 2255 goto bail; 2256 } 2257 2258 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, 2259 sizeof(di->id2.i_super.s_uuid))) { 2260 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); 2261 status = -ENOMEM; 2262 goto bail; 2263 } 2264 2265 strlcpy(osb->vol_label, di->id2.i_super.s_label, 2266 OCFS2_MAX_VOL_LABEL_LEN); 2267 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); 2268 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); 2269 osb->first_cluster_group_blkno = 2270 le64_to_cpu(di->id2.i_super.s_first_cluster_group); 2271 osb->fs_generation = le32_to_cpu(di->i_fs_generation); 2272 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2273 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str, 2274 (unsigned long long)osb->root_blkno, 2275 (unsigned long long)osb->system_dir_blkno, 2276 osb->s_clustersize_bits); 2277 2278 osb->osb_dlm_debug = ocfs2_new_dlm_debug(); 2279 if (!osb->osb_dlm_debug) { 2280 status = -ENOMEM; 2281 mlog_errno(status); 2282 goto bail; 2283 } 2284 2285 atomic_set(&osb->vol_state, VOLUME_INIT); 2286 2287 /* load root, system_dir, and all global system inodes */ 2288 status = ocfs2_init_global_system_inodes(osb); 2289 if (status < 0) { 2290 mlog_errno(status); 2291 goto bail; 2292 } 2293 2294 /* 2295 * global bitmap 2296 */ 2297 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, 2298 OCFS2_INVALID_SLOT); 2299 if (!inode) { 2300 status = -EINVAL; 2301 mlog_errno(status); 2302 goto bail; 2303 } 2304 2305 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; 2306 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters; 2307 iput(inode); 2308 2309 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0, 2310 osb->s_feature_incompat) * 8; 2311 2312 status = ocfs2_init_slot_info(osb); 2313 if (status < 0) { 2314 mlog_errno(status); 2315 goto bail; 2316 } 2317 cleancache_init_shared_fs(sb); 2318 2319 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM); 2320 if (!osb->ocfs2_wq) { 2321 status = -ENOMEM; 2322 mlog_errno(status); 2323 } 2324 2325 bail: 2326 return status; 2327 } 2328 2329 /* 2330 * will return: -EAGAIN if it is ok to keep searching for superblocks 2331 * -EINVAL if there is a bad superblock 2332 * 0 on success 2333 */ 2334 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 2335 struct buffer_head *bh, 2336 u32 blksz, 2337 struct ocfs2_blockcheck_stats *stats) 2338 { 2339 int status = -EAGAIN; 2340 2341 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, 2342 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { 2343 /* We have to do a raw check of the feature here */ 2344 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) & 2345 OCFS2_FEATURE_INCOMPAT_META_ECC) { 2346 status = ocfs2_block_check_validate(bh->b_data, 2347 bh->b_size, 2348 &di->i_check, 2349 stats); 2350 if (status) 2351 goto out; 2352 } 2353 status = -EINVAL; 2354 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) { 2355 mlog(ML_ERROR, "found superblock with incorrect block " 2356 "size: found %u, should be %u\n", 2357 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits), 2358 blksz); 2359 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != 2360 OCFS2_MAJOR_REV_LEVEL || 2361 le16_to_cpu(di->id2.i_super.s_minor_rev_level) != 2362 OCFS2_MINOR_REV_LEVEL) { 2363 mlog(ML_ERROR, "found superblock with bad version: " 2364 "found %u.%u, should be %u.%u\n", 2365 le16_to_cpu(di->id2.i_super.s_major_rev_level), 2366 le16_to_cpu(di->id2.i_super.s_minor_rev_level), 2367 OCFS2_MAJOR_REV_LEVEL, 2368 OCFS2_MINOR_REV_LEVEL); 2369 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { 2370 mlog(ML_ERROR, "bad block number on superblock: " 2371 "found %llu, should be %llu\n", 2372 (unsigned long long)le64_to_cpu(di->i_blkno), 2373 (unsigned long long)bh->b_blocknr); 2374 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || 2375 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { 2376 mlog(ML_ERROR, "bad cluster size found: %u\n", 2377 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); 2378 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { 2379 mlog(ML_ERROR, "bad root_blkno: 0\n"); 2380 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { 2381 mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); 2382 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { 2383 mlog(ML_ERROR, 2384 "Superblock slots found greater than file system " 2385 "maximum: found %u, max %u\n", 2386 le16_to_cpu(di->id2.i_super.s_max_slots), 2387 OCFS2_MAX_SLOTS); 2388 } else { 2389 /* found it! */ 2390 status = 0; 2391 } 2392 } 2393 2394 out: 2395 if (status && status != -EAGAIN) 2396 mlog_errno(status); 2397 return status; 2398 } 2399 2400 static int ocfs2_check_volume(struct ocfs2_super *osb) 2401 { 2402 int status; 2403 int dirty; 2404 int local; 2405 struct ocfs2_dinode *local_alloc = NULL; /* only used if we 2406 * recover 2407 * ourselves. */ 2408 2409 /* Init our journal object. */ 2410 status = ocfs2_journal_init(osb->journal, &dirty); 2411 if (status < 0) { 2412 mlog(ML_ERROR, "Could not initialize journal!\n"); 2413 goto finally; 2414 } 2415 2416 /* Now that journal has been initialized, check to make sure 2417 entire volume is addressable. */ 2418 status = ocfs2_journal_addressable(osb); 2419 if (status) 2420 goto finally; 2421 2422 /* If the journal was unmounted cleanly then we don't want to 2423 * recover anything. Otherwise, journal_load will do that 2424 * dirty work for us :) */ 2425 if (!dirty) { 2426 status = ocfs2_journal_wipe(osb->journal, 0); 2427 if (status < 0) { 2428 mlog_errno(status); 2429 goto finally; 2430 } 2431 } else { 2432 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not " 2433 "unmounted cleanly, recovering it.\n", osb->dev_str); 2434 } 2435 2436 local = ocfs2_mount_local(osb); 2437 2438 /* will play back anything left in the journal. */ 2439 status = ocfs2_journal_load(osb->journal, local, dirty); 2440 if (status < 0) { 2441 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status); 2442 goto finally; 2443 } 2444 2445 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT) 2446 jbd2_journal_set_features(osb->journal->j_journal, 2447 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2448 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 2449 else 2450 jbd2_journal_clear_features(osb->journal->j_journal, 2451 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2452 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 2453 2454 if (dirty) { 2455 /* recover my local alloc if we didn't unmount cleanly. */ 2456 status = ocfs2_begin_local_alloc_recovery(osb, 2457 osb->slot_num, 2458 &local_alloc); 2459 if (status < 0) { 2460 mlog_errno(status); 2461 goto finally; 2462 } 2463 /* we complete the recovery process after we've marked 2464 * ourselves as mounted. */ 2465 } 2466 2467 status = ocfs2_load_local_alloc(osb); 2468 if (status < 0) { 2469 mlog_errno(status); 2470 goto finally; 2471 } 2472 2473 if (dirty) { 2474 /* Recovery will be completed after we've mounted the 2475 * rest of the volume. */ 2476 osb->local_alloc_copy = local_alloc; 2477 local_alloc = NULL; 2478 } 2479 2480 /* go through each journal, trylock it and if you get the 2481 * lock, and it's marked as dirty, set the bit in the recover 2482 * map and launch a recovery thread for it. */ 2483 status = ocfs2_mark_dead_nodes(osb); 2484 if (status < 0) { 2485 mlog_errno(status); 2486 goto finally; 2487 } 2488 2489 status = ocfs2_compute_replay_slots(osb); 2490 if (status < 0) 2491 mlog_errno(status); 2492 2493 finally: 2494 kfree(local_alloc); 2495 2496 if (status) 2497 mlog_errno(status); 2498 return status; 2499 } 2500 2501 /* 2502 * The routine gets called from dismount or close whenever a dismount on 2503 * volume is requested and the osb open count becomes 1. 2504 * It will remove the osb from the global list and also free up all the 2505 * initialized resources and fileobject. 2506 */ 2507 static void ocfs2_delete_osb(struct ocfs2_super *osb) 2508 { 2509 /* This function assumes that the caller has the main osb resource */ 2510 2511 /* ocfs2_initializer_super have already created this workqueue */ 2512 if (osb->ocfs2_wq) 2513 destroy_workqueue(osb->ocfs2_wq); 2514 2515 ocfs2_free_slot_info(osb); 2516 2517 kfree(osb->osb_orphan_wipes); 2518 kfree(osb->slot_recovery_generations); 2519 /* FIXME 2520 * This belongs in journal shutdown, but because we have to 2521 * allocate osb->journal at the start of ocfs2_initialize_osb(), 2522 * we free it here. 2523 */ 2524 kfree(osb->journal); 2525 kfree(osb->local_alloc_copy); 2526 kfree(osb->uuid_str); 2527 kfree(osb->vol_label); 2528 ocfs2_put_dlm_debug(osb->osb_dlm_debug); 2529 memset(osb, 0, sizeof(struct ocfs2_super)); 2530 } 2531 2532 /* Depending on the mount option passed, perform one of the following: 2533 * Put OCFS2 into a readonly state (default) 2534 * Return EIO so that only the process errs 2535 * Fix the error as if fsck.ocfs2 -y 2536 * panic 2537 */ 2538 static int ocfs2_handle_error(struct super_block *sb) 2539 { 2540 struct ocfs2_super *osb = OCFS2_SB(sb); 2541 int rv = 0; 2542 2543 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); 2544 pr_crit("On-disk corruption discovered. " 2545 "Please run fsck.ocfs2 once the filesystem is unmounted.\n"); 2546 2547 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) { 2548 panic("OCFS2: (device %s): panic forced after error\n", 2549 sb->s_id); 2550 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) { 2551 pr_crit("OCFS2: Returning error to the calling process.\n"); 2552 rv = -EIO; 2553 } else { /* default option */ 2554 rv = -EROFS; 2555 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb))) 2556 return rv; 2557 2558 pr_crit("OCFS2: File system is now read-only.\n"); 2559 sb->s_flags |= SB_RDONLY; 2560 ocfs2_set_ro_flag(osb, 0); 2561 } 2562 2563 return rv; 2564 } 2565 2566 int __ocfs2_error(struct super_block *sb, const char *function, 2567 const char *fmt, ...) 2568 { 2569 struct va_format vaf; 2570 va_list args; 2571 2572 va_start(args, fmt); 2573 vaf.fmt = fmt; 2574 vaf.va = &args; 2575 2576 /* Not using mlog here because we want to show the actual 2577 * function the error came from. */ 2578 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV", 2579 sb->s_id, function, &vaf); 2580 2581 va_end(args); 2582 2583 return ocfs2_handle_error(sb); 2584 } 2585 2586 /* Handle critical errors. This is intentionally more drastic than 2587 * ocfs2_handle_error, so we only use for things like journal errors, 2588 * etc. */ 2589 void __ocfs2_abort(struct super_block *sb, const char *function, 2590 const char *fmt, ...) 2591 { 2592 struct va_format vaf; 2593 va_list args; 2594 2595 va_start(args, fmt); 2596 2597 vaf.fmt = fmt; 2598 vaf.va = &args; 2599 2600 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV", 2601 sb->s_id, function, &vaf); 2602 2603 va_end(args); 2604 2605 /* We don't have the cluster support yet to go straight to 2606 * hard readonly in here. Until then, we want to keep 2607 * ocfs2_abort() so that we can at least mark critical 2608 * errors. 2609 * 2610 * TODO: This should abort the journal and alert other nodes 2611 * that our slot needs recovery. */ 2612 2613 /* Force a panic(). This stinks, but it's better than letting 2614 * things continue without having a proper hard readonly 2615 * here. */ 2616 if (!ocfs2_mount_local(OCFS2_SB(sb))) 2617 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 2618 ocfs2_handle_error(sb); 2619 } 2620 2621 /* 2622 * Void signal blockers, because in-kernel sigprocmask() only fails 2623 * when SIG_* is wrong. 2624 */ 2625 void ocfs2_block_signals(sigset_t *oldset) 2626 { 2627 int rc; 2628 sigset_t blocked; 2629 2630 sigfillset(&blocked); 2631 rc = sigprocmask(SIG_BLOCK, &blocked, oldset); 2632 BUG_ON(rc); 2633 } 2634 2635 void ocfs2_unblock_signals(sigset_t *oldset) 2636 { 2637 int rc = sigprocmask(SIG_SETMASK, oldset, NULL); 2638 BUG_ON(rc); 2639 } 2640 2641 module_init(ocfs2_init); 2642 module_exit(ocfs2_exit); 2643