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 1083 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR, 1084 osb->osb_debug_root, 1085 osb, 1086 &ocfs2_osb_debug_fops); 1087 1088 if (ocfs2_meta_ecc(osb)) 1089 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats, 1090 osb->osb_debug_root); 1091 1092 status = ocfs2_mount_volume(sb); 1093 if (status < 0) 1094 goto read_super_error; 1095 1096 if (osb->root_inode) 1097 inode = igrab(osb->root_inode); 1098 1099 if (!inode) { 1100 status = -EIO; 1101 mlog_errno(status); 1102 goto read_super_error; 1103 } 1104 1105 root = d_make_root(inode); 1106 if (!root) { 1107 status = -ENOMEM; 1108 mlog_errno(status); 1109 goto read_super_error; 1110 } 1111 1112 sb->s_root = root; 1113 1114 ocfs2_complete_mount_recovery(osb); 1115 1116 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL, 1117 &ocfs2_kset->kobj); 1118 if (!osb->osb_dev_kset) { 1119 status = -ENOMEM; 1120 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id); 1121 goto read_super_error; 1122 } 1123 1124 /* Create filecheck sysfs related directories/files at 1125 * /sys/fs/ocfs2/<devname>/filecheck */ 1126 if (ocfs2_filecheck_create_sysfs(osb)) { 1127 status = -ENOMEM; 1128 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at " 1129 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id); 1130 goto read_super_error; 1131 } 1132 1133 if (ocfs2_mount_local(osb)) 1134 snprintf(nodestr, sizeof(nodestr), "local"); 1135 else 1136 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1137 1138 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " 1139 "with %s data mode.\n", 1140 osb->dev_str, nodestr, osb->slot_num, 1141 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : 1142 "ordered"); 1143 1144 atomic_set(&osb->vol_state, VOLUME_MOUNTED); 1145 wake_up(&osb->osb_mount_event); 1146 1147 /* Now we can initialize quotas because we can afford to wait 1148 * for cluster locks recovery now. That also means that truncation 1149 * log recovery can happen but that waits for proper quota setup */ 1150 if (!sb_rdonly(sb)) { 1151 status = ocfs2_enable_quotas(osb); 1152 if (status < 0) { 1153 /* We have to err-out specially here because 1154 * s_root is already set */ 1155 mlog_errno(status); 1156 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1157 wake_up(&osb->osb_mount_event); 1158 return status; 1159 } 1160 } 1161 1162 ocfs2_complete_quota_recovery(osb); 1163 1164 /* Now we wake up again for processes waiting for quotas */ 1165 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS); 1166 wake_up(&osb->osb_mount_event); 1167 1168 /* Start this when the mount is almost sure of being successful */ 1169 ocfs2_orphan_scan_start(osb); 1170 1171 return status; 1172 1173 read_super_error: 1174 brelse(bh); 1175 1176 if (status) 1177 mlog_errno(status); 1178 1179 if (osb) { 1180 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1181 wake_up(&osb->osb_mount_event); 1182 ocfs2_dismount_volume(sb, 1); 1183 } 1184 1185 return status; 1186 } 1187 1188 static struct dentry *ocfs2_mount(struct file_system_type *fs_type, 1189 int flags, 1190 const char *dev_name, 1191 void *data) 1192 { 1193 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super); 1194 } 1195 1196 static struct file_system_type ocfs2_fs_type = { 1197 .owner = THIS_MODULE, 1198 .name = "ocfs2", 1199 .mount = ocfs2_mount, 1200 .kill_sb = kill_block_super, 1201 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, 1202 .next = NULL 1203 }; 1204 MODULE_ALIAS_FS("ocfs2"); 1205 1206 static int ocfs2_check_set_options(struct super_block *sb, 1207 struct mount_options *options) 1208 { 1209 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA && 1210 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1211 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { 1212 mlog(ML_ERROR, "User quotas were requested, but this " 1213 "filesystem does not have the feature enabled.\n"); 1214 return 0; 1215 } 1216 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA && 1217 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1218 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { 1219 mlog(ML_ERROR, "Group quotas were requested, but this " 1220 "filesystem does not have the feature enabled.\n"); 1221 return 0; 1222 } 1223 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL && 1224 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) { 1225 mlog(ML_ERROR, "ACL support requested but extended attributes " 1226 "feature is not enabled\n"); 1227 return 0; 1228 } 1229 /* No ACL setting specified? Use XATTR feature... */ 1230 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL | 1231 OCFS2_MOUNT_NO_POSIX_ACL))) { 1232 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) 1233 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1234 else 1235 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; 1236 } 1237 return 1; 1238 } 1239 1240 static int ocfs2_parse_options(struct super_block *sb, 1241 char *options, 1242 struct mount_options *mopt, 1243 int is_remount) 1244 { 1245 int status, user_stack = 0; 1246 char *p; 1247 u32 tmp; 1248 int token, option; 1249 substring_t args[MAX_OPT_ARGS]; 1250 1251 trace_ocfs2_parse_options(is_remount, options ? options : "(none)"); 1252 1253 mopt->commit_interval = 0; 1254 mopt->mount_opt = OCFS2_MOUNT_NOINTR; 1255 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 1256 mopt->slot = OCFS2_INVALID_SLOT; 1257 mopt->localalloc_opt = -1; 1258 mopt->cluster_stack[0] = '\0'; 1259 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL; 1260 mopt->dir_resv_level = -1; 1261 1262 if (!options) { 1263 status = 1; 1264 goto bail; 1265 } 1266 1267 while ((p = strsep(&options, ",")) != NULL) { 1268 if (!*p) 1269 continue; 1270 1271 token = match_token(p, tokens, args); 1272 switch (token) { 1273 case Opt_hb_local: 1274 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; 1275 break; 1276 case Opt_hb_none: 1277 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE; 1278 break; 1279 case Opt_hb_global: 1280 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL; 1281 break; 1282 case Opt_barrier: 1283 if (match_int(&args[0], &option)) { 1284 status = 0; 1285 goto bail; 1286 } 1287 if (option) 1288 mopt->mount_opt |= OCFS2_MOUNT_BARRIER; 1289 else 1290 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; 1291 break; 1292 case Opt_intr: 1293 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; 1294 break; 1295 case Opt_nointr: 1296 mopt->mount_opt |= OCFS2_MOUNT_NOINTR; 1297 break; 1298 case Opt_err_panic: 1299 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT; 1300 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS; 1301 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 1302 break; 1303 case Opt_err_ro: 1304 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT; 1305 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1306 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS; 1307 break; 1308 case Opt_err_cont: 1309 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS; 1310 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1311 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT; 1312 break; 1313 case Opt_data_ordered: 1314 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; 1315 break; 1316 case Opt_data_writeback: 1317 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; 1318 break; 1319 case Opt_user_xattr: 1320 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR; 1321 break; 1322 case Opt_nouser_xattr: 1323 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR; 1324 break; 1325 case Opt_atime_quantum: 1326 if (match_int(&args[0], &option)) { 1327 status = 0; 1328 goto bail; 1329 } 1330 if (option >= 0) 1331 mopt->atime_quantum = option; 1332 break; 1333 case Opt_slot: 1334 if (match_int(&args[0], &option)) { 1335 status = 0; 1336 goto bail; 1337 } 1338 if (option) 1339 mopt->slot = (s16)option; 1340 break; 1341 case Opt_commit: 1342 if (match_int(&args[0], &option)) { 1343 status = 0; 1344 goto bail; 1345 } 1346 if (option < 0) 1347 return 0; 1348 if (option == 0) 1349 option = JBD2_DEFAULT_MAX_COMMIT_AGE; 1350 mopt->commit_interval = HZ * option; 1351 break; 1352 case Opt_localalloc: 1353 if (match_int(&args[0], &option)) { 1354 status = 0; 1355 goto bail; 1356 } 1357 if (option >= 0) 1358 mopt->localalloc_opt = option; 1359 break; 1360 case Opt_localflocks: 1361 /* 1362 * Changing this during remount could race 1363 * flock() requests, or "unbalance" existing 1364 * ones (e.g., a lock is taken in one mode but 1365 * dropped in the other). If users care enough 1366 * to flip locking modes during remount, we 1367 * could add a "local" flag to individual 1368 * flock structures for proper tracking of 1369 * state. 1370 */ 1371 if (!is_remount) 1372 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS; 1373 break; 1374 case Opt_stack: 1375 /* Check both that the option we were passed 1376 * is of the right length and that it is a proper 1377 * string of the right length. 1378 */ 1379 if (((args[0].to - args[0].from) != 1380 OCFS2_STACK_LABEL_LEN) || 1381 (strnlen(args[0].from, 1382 OCFS2_STACK_LABEL_LEN) != 1383 OCFS2_STACK_LABEL_LEN)) { 1384 mlog(ML_ERROR, 1385 "Invalid cluster_stack option\n"); 1386 status = 0; 1387 goto bail; 1388 } 1389 memcpy(mopt->cluster_stack, args[0].from, 1390 OCFS2_STACK_LABEL_LEN); 1391 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; 1392 /* 1393 * Open code the memcmp here as we don't have 1394 * an osb to pass to 1395 * ocfs2_userspace_stack(). 1396 */ 1397 if (memcmp(mopt->cluster_stack, 1398 OCFS2_CLASSIC_CLUSTER_STACK, 1399 OCFS2_STACK_LABEL_LEN)) 1400 user_stack = 1; 1401 break; 1402 case Opt_inode64: 1403 mopt->mount_opt |= OCFS2_MOUNT_INODE64; 1404 break; 1405 case Opt_usrquota: 1406 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA; 1407 break; 1408 case Opt_grpquota: 1409 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA; 1410 break; 1411 case Opt_coherency_buffered: 1412 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED; 1413 break; 1414 case Opt_coherency_full: 1415 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED; 1416 break; 1417 case Opt_acl: 1418 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1419 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL; 1420 break; 1421 case Opt_noacl: 1422 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; 1423 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 1424 break; 1425 case Opt_resv_level: 1426 if (is_remount) 1427 break; 1428 if (match_int(&args[0], &option)) { 1429 status = 0; 1430 goto bail; 1431 } 1432 if (option >= OCFS2_MIN_RESV_LEVEL && 1433 option < OCFS2_MAX_RESV_LEVEL) 1434 mopt->resv_level = option; 1435 break; 1436 case Opt_dir_resv_level: 1437 if (is_remount) 1438 break; 1439 if (match_int(&args[0], &option)) { 1440 status = 0; 1441 goto bail; 1442 } 1443 if (option >= OCFS2_MIN_RESV_LEVEL && 1444 option < OCFS2_MAX_RESV_LEVEL) 1445 mopt->dir_resv_level = option; 1446 break; 1447 case Opt_journal_async_commit: 1448 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT; 1449 break; 1450 default: 1451 mlog(ML_ERROR, 1452 "Unrecognized mount option \"%s\" " 1453 "or missing value\n", p); 1454 status = 0; 1455 goto bail; 1456 } 1457 } 1458 1459 if (user_stack == 0) { 1460 /* Ensure only one heartbeat mode */ 1461 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | 1462 OCFS2_MOUNT_HB_GLOBAL | 1463 OCFS2_MOUNT_HB_NONE); 1464 if (hweight32(tmp) != 1) { 1465 mlog(ML_ERROR, "Invalid heartbeat mount options\n"); 1466 status = 0; 1467 goto bail; 1468 } 1469 } 1470 1471 status = 1; 1472 1473 bail: 1474 return status; 1475 } 1476 1477 static int ocfs2_show_options(struct seq_file *s, struct dentry *root) 1478 { 1479 struct ocfs2_super *osb = OCFS2_SB(root->d_sb); 1480 unsigned long opts = osb->s_mount_opt; 1481 unsigned int local_alloc_megs; 1482 1483 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) { 1484 seq_printf(s, ",_netdev"); 1485 if (opts & OCFS2_MOUNT_HB_LOCAL) 1486 seq_printf(s, ",%s", OCFS2_HB_LOCAL); 1487 else 1488 seq_printf(s, ",%s", OCFS2_HB_GLOBAL); 1489 } else 1490 seq_printf(s, ",%s", OCFS2_HB_NONE); 1491 1492 if (opts & OCFS2_MOUNT_NOINTR) 1493 seq_printf(s, ",nointr"); 1494 1495 if (opts & OCFS2_MOUNT_DATA_WRITEBACK) 1496 seq_printf(s, ",data=writeback"); 1497 else 1498 seq_printf(s, ",data=ordered"); 1499 1500 if (opts & OCFS2_MOUNT_BARRIER) 1501 seq_printf(s, ",barrier=1"); 1502 1503 if (opts & OCFS2_MOUNT_ERRORS_PANIC) 1504 seq_printf(s, ",errors=panic"); 1505 else if (opts & OCFS2_MOUNT_ERRORS_CONT) 1506 seq_printf(s, ",errors=continue"); 1507 else 1508 seq_printf(s, ",errors=remount-ro"); 1509 1510 if (osb->preferred_slot != OCFS2_INVALID_SLOT) 1511 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); 1512 1513 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); 1514 1515 if (osb->osb_commit_interval) 1516 seq_printf(s, ",commit=%u", 1517 (unsigned) (osb->osb_commit_interval / HZ)); 1518 1519 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits); 1520 if (local_alloc_megs != ocfs2_la_default_mb(osb)) 1521 seq_printf(s, ",localalloc=%d", local_alloc_megs); 1522 1523 if (opts & OCFS2_MOUNT_LOCALFLOCKS) 1524 seq_printf(s, ",localflocks,"); 1525 1526 if (osb->osb_cluster_stack[0]) 1527 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack, 1528 OCFS2_STACK_LABEL_LEN); 1529 if (opts & OCFS2_MOUNT_USRQUOTA) 1530 seq_printf(s, ",usrquota"); 1531 if (opts & OCFS2_MOUNT_GRPQUOTA) 1532 seq_printf(s, ",grpquota"); 1533 1534 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED) 1535 seq_printf(s, ",coherency=buffered"); 1536 else 1537 seq_printf(s, ",coherency=full"); 1538 1539 if (opts & OCFS2_MOUNT_NOUSERXATTR) 1540 seq_printf(s, ",nouser_xattr"); 1541 else 1542 seq_printf(s, ",user_xattr"); 1543 1544 if (opts & OCFS2_MOUNT_INODE64) 1545 seq_printf(s, ",inode64"); 1546 1547 if (opts & OCFS2_MOUNT_POSIX_ACL) 1548 seq_printf(s, ",acl"); 1549 else 1550 seq_printf(s, ",noacl"); 1551 1552 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL) 1553 seq_printf(s, ",resv_level=%d", osb->osb_resv_level); 1554 1555 if (osb->osb_dir_resv_level != osb->osb_resv_level) 1556 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level); 1557 1558 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT) 1559 seq_printf(s, ",journal_async_commit"); 1560 1561 return 0; 1562 } 1563 1564 static int __init ocfs2_init(void) 1565 { 1566 int status; 1567 1568 status = init_ocfs2_uptodate_cache(); 1569 if (status < 0) 1570 goto out1; 1571 1572 status = ocfs2_initialize_mem_caches(); 1573 if (status < 0) 1574 goto out2; 1575 1576 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); 1577 1578 ocfs2_set_locking_protocol(); 1579 1580 status = register_quota_format(&ocfs2_quota_format); 1581 if (status < 0) 1582 goto out3; 1583 status = register_filesystem(&ocfs2_fs_type); 1584 if (!status) 1585 return 0; 1586 1587 unregister_quota_format(&ocfs2_quota_format); 1588 out3: 1589 debugfs_remove(ocfs2_debugfs_root); 1590 ocfs2_free_mem_caches(); 1591 out2: 1592 exit_ocfs2_uptodate_cache(); 1593 out1: 1594 mlog_errno(status); 1595 return status; 1596 } 1597 1598 static void __exit ocfs2_exit(void) 1599 { 1600 unregister_quota_format(&ocfs2_quota_format); 1601 1602 debugfs_remove(ocfs2_debugfs_root); 1603 1604 ocfs2_free_mem_caches(); 1605 1606 unregister_filesystem(&ocfs2_fs_type); 1607 1608 exit_ocfs2_uptodate_cache(); 1609 } 1610 1611 static void ocfs2_put_super(struct super_block *sb) 1612 { 1613 trace_ocfs2_put_super(sb); 1614 1615 ocfs2_sync_blockdev(sb); 1616 ocfs2_dismount_volume(sb, 0); 1617 } 1618 1619 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) 1620 { 1621 struct ocfs2_super *osb; 1622 u32 numbits, freebits; 1623 int status; 1624 struct ocfs2_dinode *bm_lock; 1625 struct buffer_head *bh = NULL; 1626 struct inode *inode = NULL; 1627 1628 trace_ocfs2_statfs(dentry->d_sb, buf); 1629 1630 osb = OCFS2_SB(dentry->d_sb); 1631 1632 inode = ocfs2_get_system_file_inode(osb, 1633 GLOBAL_BITMAP_SYSTEM_INODE, 1634 OCFS2_INVALID_SLOT); 1635 if (!inode) { 1636 mlog(ML_ERROR, "failed to get bitmap inode\n"); 1637 status = -EIO; 1638 goto bail; 1639 } 1640 1641 status = ocfs2_inode_lock(inode, &bh, 0); 1642 if (status < 0) { 1643 mlog_errno(status); 1644 goto bail; 1645 } 1646 1647 bm_lock = (struct ocfs2_dinode *) bh->b_data; 1648 1649 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); 1650 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); 1651 1652 buf->f_type = OCFS2_SUPER_MAGIC; 1653 buf->f_bsize = dentry->d_sb->s_blocksize; 1654 buf->f_namelen = OCFS2_MAX_FILENAME_LEN; 1655 buf->f_blocks = ((sector_t) numbits) * 1656 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1657 buf->f_bfree = ((sector_t) freebits) * 1658 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1659 buf->f_bavail = buf->f_bfree; 1660 buf->f_files = numbits; 1661 buf->f_ffree = freebits; 1662 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN) 1663 & 0xFFFFFFFFUL; 1664 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN, 1665 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL; 1666 1667 brelse(bh); 1668 1669 ocfs2_inode_unlock(inode, 0); 1670 status = 0; 1671 bail: 1672 iput(inode); 1673 1674 if (status) 1675 mlog_errno(status); 1676 1677 return status; 1678 } 1679 1680 static void ocfs2_inode_init_once(void *data) 1681 { 1682 struct ocfs2_inode_info *oi = data; 1683 1684 oi->ip_flags = 0; 1685 oi->ip_open_count = 0; 1686 spin_lock_init(&oi->ip_lock); 1687 ocfs2_extent_map_init(&oi->vfs_inode); 1688 INIT_LIST_HEAD(&oi->ip_io_markers); 1689 INIT_LIST_HEAD(&oi->ip_unwritten_list); 1690 oi->ip_dir_start_lookup = 0; 1691 init_rwsem(&oi->ip_alloc_sem); 1692 init_rwsem(&oi->ip_xattr_sem); 1693 mutex_init(&oi->ip_io_mutex); 1694 1695 oi->ip_blkno = 0ULL; 1696 oi->ip_clusters = 0; 1697 1698 ocfs2_resv_init_once(&oi->ip_la_data_resv); 1699 1700 ocfs2_lock_res_init_once(&oi->ip_rw_lockres); 1701 ocfs2_lock_res_init_once(&oi->ip_inode_lockres); 1702 ocfs2_lock_res_init_once(&oi->ip_open_lockres); 1703 1704 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode), 1705 &ocfs2_inode_caching_ops); 1706 1707 inode_init_once(&oi->vfs_inode); 1708 } 1709 1710 static int ocfs2_initialize_mem_caches(void) 1711 { 1712 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", 1713 sizeof(struct ocfs2_inode_info), 1714 0, 1715 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1716 SLAB_MEM_SPREAD|SLAB_ACCOUNT), 1717 ocfs2_inode_init_once); 1718 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache", 1719 sizeof(struct ocfs2_dquot), 1720 0, 1721 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1722 SLAB_MEM_SPREAD), 1723 NULL); 1724 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache", 1725 sizeof(struct ocfs2_quota_chunk), 1726 0, 1727 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), 1728 NULL); 1729 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep || 1730 !ocfs2_qf_chunk_cachep) { 1731 kmem_cache_destroy(ocfs2_inode_cachep); 1732 kmem_cache_destroy(ocfs2_dquot_cachep); 1733 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1734 return -ENOMEM; 1735 } 1736 1737 return 0; 1738 } 1739 1740 static void ocfs2_free_mem_caches(void) 1741 { 1742 /* 1743 * Make sure all delayed rcu free inodes are flushed before we 1744 * destroy cache. 1745 */ 1746 rcu_barrier(); 1747 kmem_cache_destroy(ocfs2_inode_cachep); 1748 ocfs2_inode_cachep = NULL; 1749 1750 kmem_cache_destroy(ocfs2_dquot_cachep); 1751 ocfs2_dquot_cachep = NULL; 1752 1753 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1754 ocfs2_qf_chunk_cachep = NULL; 1755 } 1756 1757 static int ocfs2_get_sector(struct super_block *sb, 1758 struct buffer_head **bh, 1759 int block, 1760 int sect_size) 1761 { 1762 if (!sb_set_blocksize(sb, sect_size)) { 1763 mlog(ML_ERROR, "unable to set blocksize\n"); 1764 return -EIO; 1765 } 1766 1767 *bh = sb_getblk(sb, block); 1768 if (!*bh) { 1769 mlog_errno(-ENOMEM); 1770 return -ENOMEM; 1771 } 1772 lock_buffer(*bh); 1773 if (!buffer_dirty(*bh)) 1774 clear_buffer_uptodate(*bh); 1775 unlock_buffer(*bh); 1776 ll_rw_block(REQ_OP_READ, 0, 1, bh); 1777 wait_on_buffer(*bh); 1778 if (!buffer_uptodate(*bh)) { 1779 mlog_errno(-EIO); 1780 brelse(*bh); 1781 *bh = NULL; 1782 return -EIO; 1783 } 1784 1785 return 0; 1786 } 1787 1788 static int ocfs2_mount_volume(struct super_block *sb) 1789 { 1790 int status = 0; 1791 int unlock_super = 0; 1792 struct ocfs2_super *osb = OCFS2_SB(sb); 1793 1794 if (ocfs2_is_hard_readonly(osb)) 1795 goto leave; 1796 1797 mutex_init(&osb->obs_trim_fs_mutex); 1798 1799 status = ocfs2_dlm_init(osb); 1800 if (status < 0) { 1801 mlog_errno(status); 1802 if (status == -EBADR && ocfs2_userspace_stack(osb)) 1803 mlog(ML_ERROR, "couldn't mount because cluster name on" 1804 " disk does not match the running cluster name.\n"); 1805 goto leave; 1806 } 1807 1808 status = ocfs2_super_lock(osb, 1); 1809 if (status < 0) { 1810 mlog_errno(status); 1811 goto leave; 1812 } 1813 unlock_super = 1; 1814 1815 /* This will load up the node map and add ourselves to it. */ 1816 status = ocfs2_find_slot(osb); 1817 if (status < 0) { 1818 mlog_errno(status); 1819 goto leave; 1820 } 1821 1822 /* load all node-local system inodes */ 1823 status = ocfs2_init_local_system_inodes(osb); 1824 if (status < 0) { 1825 mlog_errno(status); 1826 goto leave; 1827 } 1828 1829 status = ocfs2_check_volume(osb); 1830 if (status < 0) { 1831 mlog_errno(status); 1832 goto leave; 1833 } 1834 1835 status = ocfs2_truncate_log_init(osb); 1836 if (status < 0) 1837 mlog_errno(status); 1838 1839 leave: 1840 if (unlock_super) 1841 ocfs2_super_unlock(osb, 1); 1842 1843 return status; 1844 } 1845 1846 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) 1847 { 1848 int tmp, hangup_needed = 0; 1849 struct ocfs2_super *osb = NULL; 1850 char nodestr[12]; 1851 1852 trace_ocfs2_dismount_volume(sb); 1853 1854 BUG_ON(!sb); 1855 osb = OCFS2_SB(sb); 1856 BUG_ON(!osb); 1857 1858 /* Remove file check sysfs related directores/files, 1859 * and wait for the pending file check operations */ 1860 ocfs2_filecheck_remove_sysfs(osb); 1861 1862 kset_unregister(osb->osb_dev_kset); 1863 1864 debugfs_remove(osb->osb_ctxt); 1865 1866 /* Orphan scan should be stopped as early as possible */ 1867 ocfs2_orphan_scan_stop(osb); 1868 1869 ocfs2_disable_quotas(osb); 1870 1871 /* All dquots should be freed by now */ 1872 WARN_ON(!llist_empty(&osb->dquot_drop_list)); 1873 /* Wait for worker to be done with the work structure in osb */ 1874 cancel_work_sync(&osb->dquot_drop_work); 1875 1876 ocfs2_shutdown_local_alloc(osb); 1877 1878 ocfs2_truncate_log_shutdown(osb); 1879 1880 /* This will disable recovery and flush any recovery work. */ 1881 ocfs2_recovery_exit(osb); 1882 1883 ocfs2_journal_shutdown(osb); 1884 1885 ocfs2_sync_blockdev(sb); 1886 1887 ocfs2_purge_refcount_trees(osb); 1888 1889 /* No cluster connection means we've failed during mount, so skip 1890 * all the steps which depended on that to complete. */ 1891 if (osb->cconn) { 1892 tmp = ocfs2_super_lock(osb, 1); 1893 if (tmp < 0) { 1894 mlog_errno(tmp); 1895 return; 1896 } 1897 } 1898 1899 if (osb->slot_num != OCFS2_INVALID_SLOT) 1900 ocfs2_put_slot(osb); 1901 1902 if (osb->cconn) 1903 ocfs2_super_unlock(osb, 1); 1904 1905 ocfs2_release_system_inodes(osb); 1906 1907 /* 1908 * If we're dismounting due to mount error, mount.ocfs2 will clean 1909 * up heartbeat. If we're a local mount, there is no heartbeat. 1910 * If we failed before we got a uuid_str yet, we can't stop 1911 * heartbeat. Otherwise, do it. 1912 */ 1913 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str && 1914 !ocfs2_is_hard_readonly(osb)) 1915 hangup_needed = 1; 1916 1917 if (osb->cconn) 1918 ocfs2_dlm_shutdown(osb, hangup_needed); 1919 1920 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats); 1921 debugfs_remove(osb->osb_debug_root); 1922 1923 if (hangup_needed) 1924 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str)); 1925 1926 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); 1927 1928 if (ocfs2_mount_local(osb)) 1929 snprintf(nodestr, sizeof(nodestr), "local"); 1930 else 1931 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1932 1933 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", 1934 osb->dev_str, nodestr); 1935 1936 ocfs2_delete_osb(osb); 1937 kfree(osb); 1938 sb->s_dev = 0; 1939 sb->s_fs_info = NULL; 1940 } 1941 1942 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, 1943 unsigned uuid_bytes) 1944 { 1945 int i, ret; 1946 char *ptr; 1947 1948 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); 1949 1950 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); 1951 if (osb->uuid_str == NULL) 1952 return -ENOMEM; 1953 1954 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { 1955 /* print with null */ 1956 ret = snprintf(ptr, 3, "%02X", uuid[i]); 1957 if (ret != 2) /* drop super cleans up */ 1958 return -EINVAL; 1959 /* then only advance past the last char */ 1960 ptr += 2; 1961 } 1962 1963 return 0; 1964 } 1965 1966 /* Make sure entire volume is addressable by our journal. Requires 1967 osb_clusters_at_boot to be valid and for the journal to have been 1968 initialized by ocfs2_journal_init(). */ 1969 static int ocfs2_journal_addressable(struct ocfs2_super *osb) 1970 { 1971 int status = 0; 1972 u64 max_block = 1973 ocfs2_clusters_to_blocks(osb->sb, 1974 osb->osb_clusters_at_boot) - 1; 1975 1976 /* 32-bit block number is always OK. */ 1977 if (max_block <= (u32)~0ULL) 1978 goto out; 1979 1980 /* Volume is "huge", so see if our journal is new enough to 1981 support it. */ 1982 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb, 1983 OCFS2_FEATURE_COMPAT_JBD2_SB) && 1984 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0, 1985 JBD2_FEATURE_INCOMPAT_64BIT))) { 1986 mlog(ML_ERROR, "The journal cannot address the entire volume. " 1987 "Enable the 'block64' journal option with tunefs.ocfs2"); 1988 status = -EFBIG; 1989 goto out; 1990 } 1991 1992 out: 1993 return status; 1994 } 1995 1996 static int ocfs2_initialize_super(struct super_block *sb, 1997 struct buffer_head *bh, 1998 int sector_size, 1999 struct ocfs2_blockcheck_stats *stats) 2000 { 2001 int status; 2002 int i, cbits, bbits; 2003 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; 2004 struct inode *inode = NULL; 2005 struct ocfs2_journal *journal; 2006 struct ocfs2_super *osb; 2007 u64 total_blocks; 2008 2009 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); 2010 if (!osb) { 2011 status = -ENOMEM; 2012 mlog_errno(status); 2013 goto bail; 2014 } 2015 2016 sb->s_fs_info = osb; 2017 sb->s_op = &ocfs2_sops; 2018 sb->s_d_op = &ocfs2_dentry_ops; 2019 sb->s_export_op = &ocfs2_export_ops; 2020 sb->s_qcop = &dquot_quotactl_sysfile_ops; 2021 sb->dq_op = &ocfs2_quota_operations; 2022 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP; 2023 sb->s_xattr = ocfs2_xattr_handlers; 2024 sb->s_time_gran = 1; 2025 sb->s_flags |= SB_NOATIME; 2026 /* this is needed to support O_LARGEFILE */ 2027 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2028 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); 2029 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); 2030 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid, 2031 sizeof(di->id2.i_super.s_uuid)); 2032 2033 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1; 2034 2035 for (i = 0; i < 3; i++) 2036 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]); 2037 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2038 2039 osb->sb = sb; 2040 osb->s_sectsize_bits = blksize_bits(sector_size); 2041 BUG_ON(!osb->s_sectsize_bits); 2042 2043 spin_lock_init(&osb->dc_task_lock); 2044 init_waitqueue_head(&osb->dc_event); 2045 osb->dc_work_sequence = 0; 2046 osb->dc_wake_sequence = 0; 2047 INIT_LIST_HEAD(&osb->blocked_lock_list); 2048 osb->blocked_lock_count = 0; 2049 spin_lock_init(&osb->osb_lock); 2050 spin_lock_init(&osb->osb_xattr_lock); 2051 ocfs2_init_steal_slots(osb); 2052 2053 mutex_init(&osb->system_file_mutex); 2054 2055 atomic_set(&osb->alloc_stats.moves, 0); 2056 atomic_set(&osb->alloc_stats.local_data, 0); 2057 atomic_set(&osb->alloc_stats.bitmap_data, 0); 2058 atomic_set(&osb->alloc_stats.bg_allocs, 0); 2059 atomic_set(&osb->alloc_stats.bg_extends, 0); 2060 2061 /* Copy the blockcheck stats from the superblock probe */ 2062 osb->osb_ecc_stats = *stats; 2063 2064 ocfs2_init_node_maps(osb); 2065 2066 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", 2067 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); 2068 2069 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); 2070 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { 2071 mlog(ML_ERROR, "Invalid number of node slots (%u)\n", 2072 osb->max_slots); 2073 status = -EINVAL; 2074 goto bail; 2075 } 2076 2077 ocfs2_orphan_scan_init(osb); 2078 2079 status = ocfs2_recovery_init(osb); 2080 if (status) { 2081 mlog(ML_ERROR, "Unable to initialize recovery state\n"); 2082 mlog_errno(status); 2083 goto bail; 2084 } 2085 2086 init_waitqueue_head(&osb->checkpoint_event); 2087 2088 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 2089 2090 osb->slot_num = OCFS2_INVALID_SLOT; 2091 2092 osb->s_xattr_inline_size = le16_to_cpu( 2093 di->id2.i_super.s_xattr_inline_size); 2094 2095 osb->local_alloc_state = OCFS2_LA_UNUSED; 2096 osb->local_alloc_bh = NULL; 2097 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker); 2098 2099 init_waitqueue_head(&osb->osb_mount_event); 2100 2101 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap); 2102 if (status) { 2103 mlog_errno(status); 2104 goto bail; 2105 } 2106 2107 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); 2108 if (!osb->vol_label) { 2109 mlog(ML_ERROR, "unable to alloc vol label\n"); 2110 status = -ENOMEM; 2111 goto bail; 2112 } 2113 2114 osb->slot_recovery_generations = 2115 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), 2116 GFP_KERNEL); 2117 if (!osb->slot_recovery_generations) { 2118 status = -ENOMEM; 2119 mlog_errno(status); 2120 goto bail; 2121 } 2122 2123 init_waitqueue_head(&osb->osb_wipe_event); 2124 osb->osb_orphan_wipes = kcalloc(osb->max_slots, 2125 sizeof(*osb->osb_orphan_wipes), 2126 GFP_KERNEL); 2127 if (!osb->osb_orphan_wipes) { 2128 status = -ENOMEM; 2129 mlog_errno(status); 2130 goto bail; 2131 } 2132 2133 osb->osb_rf_lock_tree = RB_ROOT; 2134 2135 osb->s_feature_compat = 2136 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); 2137 osb->s_feature_ro_compat = 2138 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); 2139 osb->s_feature_incompat = 2140 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); 2141 2142 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { 2143 mlog(ML_ERROR, "couldn't mount because of unsupported " 2144 "optional features (%x).\n", i); 2145 status = -EINVAL; 2146 goto bail; 2147 } 2148 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { 2149 mlog(ML_ERROR, "couldn't mount RDWR because of " 2150 "unsupported optional features (%x).\n", i); 2151 status = -EINVAL; 2152 goto bail; 2153 } 2154 2155 if (ocfs2_clusterinfo_valid(osb)) { 2156 osb->osb_stackflags = 2157 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags; 2158 strlcpy(osb->osb_cluster_stack, 2159 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, 2160 OCFS2_STACK_LABEL_LEN + 1); 2161 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) { 2162 mlog(ML_ERROR, 2163 "couldn't mount because of an invalid " 2164 "cluster stack label (%s) \n", 2165 osb->osb_cluster_stack); 2166 status = -EINVAL; 2167 goto bail; 2168 } 2169 strlcpy(osb->osb_cluster_name, 2170 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster, 2171 OCFS2_CLUSTER_NAME_LEN + 1); 2172 } else { 2173 /* The empty string is identical with classic tools that 2174 * don't know about s_cluster_info. */ 2175 osb->osb_cluster_stack[0] = '\0'; 2176 } 2177 2178 get_random_bytes(&osb->s_next_generation, sizeof(u32)); 2179 2180 /* FIXME 2181 * This should be done in ocfs2_journal_init(), but unknown 2182 * ordering issues will cause the filesystem to crash. 2183 * If anyone wants to figure out what part of the code 2184 * refers to osb->journal before ocfs2_journal_init() is run, 2185 * be my guest. 2186 */ 2187 /* initialize our journal structure */ 2188 2189 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL); 2190 if (!journal) { 2191 mlog(ML_ERROR, "unable to alloc journal\n"); 2192 status = -ENOMEM; 2193 goto bail; 2194 } 2195 osb->journal = journal; 2196 journal->j_osb = osb; 2197 2198 atomic_set(&journal->j_num_trans, 0); 2199 init_rwsem(&journal->j_trans_barrier); 2200 init_waitqueue_head(&journal->j_checkpointed); 2201 spin_lock_init(&journal->j_lock); 2202 journal->j_trans_id = (unsigned long) 1; 2203 INIT_LIST_HEAD(&journal->j_la_cleanups); 2204 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); 2205 journal->j_state = OCFS2_JOURNAL_FREE; 2206 2207 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs); 2208 init_llist_head(&osb->dquot_drop_list); 2209 2210 /* get some pseudo constants for clustersize bits */ 2211 osb->s_clustersize_bits = 2212 le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2213 osb->s_clustersize = 1 << osb->s_clustersize_bits; 2214 2215 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || 2216 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { 2217 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", 2218 osb->s_clustersize); 2219 status = -EINVAL; 2220 goto bail; 2221 } 2222 2223 total_blocks = ocfs2_clusters_to_blocks(osb->sb, 2224 le32_to_cpu(di->i_clusters)); 2225 2226 status = generic_check_addressable(osb->sb->s_blocksize_bits, 2227 total_blocks); 2228 if (status) { 2229 mlog(ML_ERROR, "Volume too large " 2230 "to mount safely on this system"); 2231 status = -EFBIG; 2232 goto bail; 2233 } 2234 2235 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, 2236 sizeof(di->id2.i_super.s_uuid))) { 2237 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); 2238 status = -ENOMEM; 2239 goto bail; 2240 } 2241 2242 strlcpy(osb->vol_label, di->id2.i_super.s_label, 2243 OCFS2_MAX_VOL_LABEL_LEN); 2244 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); 2245 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); 2246 osb->first_cluster_group_blkno = 2247 le64_to_cpu(di->id2.i_super.s_first_cluster_group); 2248 osb->fs_generation = le32_to_cpu(di->i_fs_generation); 2249 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2250 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str, 2251 (unsigned long long)osb->root_blkno, 2252 (unsigned long long)osb->system_dir_blkno, 2253 osb->s_clustersize_bits); 2254 2255 osb->osb_dlm_debug = ocfs2_new_dlm_debug(); 2256 if (!osb->osb_dlm_debug) { 2257 status = -ENOMEM; 2258 mlog_errno(status); 2259 goto bail; 2260 } 2261 2262 atomic_set(&osb->vol_state, VOLUME_INIT); 2263 2264 /* load root, system_dir, and all global system inodes */ 2265 status = ocfs2_init_global_system_inodes(osb); 2266 if (status < 0) { 2267 mlog_errno(status); 2268 goto bail; 2269 } 2270 2271 /* 2272 * global bitmap 2273 */ 2274 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, 2275 OCFS2_INVALID_SLOT); 2276 if (!inode) { 2277 status = -EINVAL; 2278 mlog_errno(status); 2279 goto bail; 2280 } 2281 2282 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; 2283 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters; 2284 iput(inode); 2285 2286 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0, 2287 osb->s_feature_incompat) * 8; 2288 2289 status = ocfs2_init_slot_info(osb); 2290 if (status < 0) { 2291 mlog_errno(status); 2292 goto bail; 2293 } 2294 cleancache_init_shared_fs(sb); 2295 2296 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM); 2297 if (!osb->ocfs2_wq) { 2298 status = -ENOMEM; 2299 mlog_errno(status); 2300 } 2301 2302 bail: 2303 return status; 2304 } 2305 2306 /* 2307 * will return: -EAGAIN if it is ok to keep searching for superblocks 2308 * -EINVAL if there is a bad superblock 2309 * 0 on success 2310 */ 2311 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 2312 struct buffer_head *bh, 2313 u32 blksz, 2314 struct ocfs2_blockcheck_stats *stats) 2315 { 2316 int status = -EAGAIN; 2317 2318 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, 2319 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { 2320 /* We have to do a raw check of the feature here */ 2321 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) & 2322 OCFS2_FEATURE_INCOMPAT_META_ECC) { 2323 status = ocfs2_block_check_validate(bh->b_data, 2324 bh->b_size, 2325 &di->i_check, 2326 stats); 2327 if (status) 2328 goto out; 2329 } 2330 status = -EINVAL; 2331 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) { 2332 mlog(ML_ERROR, "found superblock with incorrect block " 2333 "size: found %u, should be %u\n", 2334 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits), 2335 blksz); 2336 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != 2337 OCFS2_MAJOR_REV_LEVEL || 2338 le16_to_cpu(di->id2.i_super.s_minor_rev_level) != 2339 OCFS2_MINOR_REV_LEVEL) { 2340 mlog(ML_ERROR, "found superblock with bad version: " 2341 "found %u.%u, should be %u.%u\n", 2342 le16_to_cpu(di->id2.i_super.s_major_rev_level), 2343 le16_to_cpu(di->id2.i_super.s_minor_rev_level), 2344 OCFS2_MAJOR_REV_LEVEL, 2345 OCFS2_MINOR_REV_LEVEL); 2346 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { 2347 mlog(ML_ERROR, "bad block number on superblock: " 2348 "found %llu, should be %llu\n", 2349 (unsigned long long)le64_to_cpu(di->i_blkno), 2350 (unsigned long long)bh->b_blocknr); 2351 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || 2352 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { 2353 mlog(ML_ERROR, "bad cluster size found: %u\n", 2354 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); 2355 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { 2356 mlog(ML_ERROR, "bad root_blkno: 0\n"); 2357 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { 2358 mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); 2359 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { 2360 mlog(ML_ERROR, 2361 "Superblock slots found greater than file system " 2362 "maximum: found %u, max %u\n", 2363 le16_to_cpu(di->id2.i_super.s_max_slots), 2364 OCFS2_MAX_SLOTS); 2365 } else { 2366 /* found it! */ 2367 status = 0; 2368 } 2369 } 2370 2371 out: 2372 if (status && status != -EAGAIN) 2373 mlog_errno(status); 2374 return status; 2375 } 2376 2377 static int ocfs2_check_volume(struct ocfs2_super *osb) 2378 { 2379 int status; 2380 int dirty; 2381 int local; 2382 struct ocfs2_dinode *local_alloc = NULL; /* only used if we 2383 * recover 2384 * ourselves. */ 2385 2386 /* Init our journal object. */ 2387 status = ocfs2_journal_init(osb->journal, &dirty); 2388 if (status < 0) { 2389 mlog(ML_ERROR, "Could not initialize journal!\n"); 2390 goto finally; 2391 } 2392 2393 /* Now that journal has been initialized, check to make sure 2394 entire volume is addressable. */ 2395 status = ocfs2_journal_addressable(osb); 2396 if (status) 2397 goto finally; 2398 2399 /* If the journal was unmounted cleanly then we don't want to 2400 * recover anything. Otherwise, journal_load will do that 2401 * dirty work for us :) */ 2402 if (!dirty) { 2403 status = ocfs2_journal_wipe(osb->journal, 0); 2404 if (status < 0) { 2405 mlog_errno(status); 2406 goto finally; 2407 } 2408 } else { 2409 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not " 2410 "unmounted cleanly, recovering it.\n", osb->dev_str); 2411 } 2412 2413 local = ocfs2_mount_local(osb); 2414 2415 /* will play back anything left in the journal. */ 2416 status = ocfs2_journal_load(osb->journal, local, dirty); 2417 if (status < 0) { 2418 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status); 2419 goto finally; 2420 } 2421 2422 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT) 2423 jbd2_journal_set_features(osb->journal->j_journal, 2424 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2425 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 2426 else 2427 jbd2_journal_clear_features(osb->journal->j_journal, 2428 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2429 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 2430 2431 if (dirty) { 2432 /* recover my local alloc if we didn't unmount cleanly. */ 2433 status = ocfs2_begin_local_alloc_recovery(osb, 2434 osb->slot_num, 2435 &local_alloc); 2436 if (status < 0) { 2437 mlog_errno(status); 2438 goto finally; 2439 } 2440 /* we complete the recovery process after we've marked 2441 * ourselves as mounted. */ 2442 } 2443 2444 status = ocfs2_load_local_alloc(osb); 2445 if (status < 0) { 2446 mlog_errno(status); 2447 goto finally; 2448 } 2449 2450 if (dirty) { 2451 /* Recovery will be completed after we've mounted the 2452 * rest of the volume. */ 2453 osb->local_alloc_copy = local_alloc; 2454 local_alloc = NULL; 2455 } 2456 2457 /* go through each journal, trylock it and if you get the 2458 * lock, and it's marked as dirty, set the bit in the recover 2459 * map and launch a recovery thread for it. */ 2460 status = ocfs2_mark_dead_nodes(osb); 2461 if (status < 0) { 2462 mlog_errno(status); 2463 goto finally; 2464 } 2465 2466 status = ocfs2_compute_replay_slots(osb); 2467 if (status < 0) 2468 mlog_errno(status); 2469 2470 finally: 2471 kfree(local_alloc); 2472 2473 if (status) 2474 mlog_errno(status); 2475 return status; 2476 } 2477 2478 /* 2479 * The routine gets called from dismount or close whenever a dismount on 2480 * volume is requested and the osb open count becomes 1. 2481 * It will remove the osb from the global list and also free up all the 2482 * initialized resources and fileobject. 2483 */ 2484 static void ocfs2_delete_osb(struct ocfs2_super *osb) 2485 { 2486 /* This function assumes that the caller has the main osb resource */ 2487 2488 /* ocfs2_initializer_super have already created this workqueue */ 2489 if (osb->ocfs2_wq) 2490 destroy_workqueue(osb->ocfs2_wq); 2491 2492 ocfs2_free_slot_info(osb); 2493 2494 kfree(osb->osb_orphan_wipes); 2495 kfree(osb->slot_recovery_generations); 2496 /* FIXME 2497 * This belongs in journal shutdown, but because we have to 2498 * allocate osb->journal at the start of ocfs2_initialize_osb(), 2499 * we free it here. 2500 */ 2501 kfree(osb->journal); 2502 kfree(osb->local_alloc_copy); 2503 kfree(osb->uuid_str); 2504 kfree(osb->vol_label); 2505 ocfs2_put_dlm_debug(osb->osb_dlm_debug); 2506 memset(osb, 0, sizeof(struct ocfs2_super)); 2507 } 2508 2509 /* Depending on the mount option passed, perform one of the following: 2510 * Put OCFS2 into a readonly state (default) 2511 * Return EIO so that only the process errs 2512 * Fix the error as if fsck.ocfs2 -y 2513 * panic 2514 */ 2515 static int ocfs2_handle_error(struct super_block *sb) 2516 { 2517 struct ocfs2_super *osb = OCFS2_SB(sb); 2518 int rv = 0; 2519 2520 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); 2521 pr_crit("On-disk corruption discovered. " 2522 "Please run fsck.ocfs2 once the filesystem is unmounted.\n"); 2523 2524 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) { 2525 panic("OCFS2: (device %s): panic forced after error\n", 2526 sb->s_id); 2527 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) { 2528 pr_crit("OCFS2: Returning error to the calling process.\n"); 2529 rv = -EIO; 2530 } else { /* default option */ 2531 rv = -EROFS; 2532 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb))) 2533 return rv; 2534 2535 pr_crit("OCFS2: File system is now read-only.\n"); 2536 sb->s_flags |= SB_RDONLY; 2537 ocfs2_set_ro_flag(osb, 0); 2538 } 2539 2540 return rv; 2541 } 2542 2543 int __ocfs2_error(struct super_block *sb, const char *function, 2544 const char *fmt, ...) 2545 { 2546 struct va_format vaf; 2547 va_list args; 2548 2549 va_start(args, fmt); 2550 vaf.fmt = fmt; 2551 vaf.va = &args; 2552 2553 /* Not using mlog here because we want to show the actual 2554 * function the error came from. */ 2555 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV", 2556 sb->s_id, function, &vaf); 2557 2558 va_end(args); 2559 2560 return ocfs2_handle_error(sb); 2561 } 2562 2563 /* Handle critical errors. This is intentionally more drastic than 2564 * ocfs2_handle_error, so we only use for things like journal errors, 2565 * etc. */ 2566 void __ocfs2_abort(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 2574 vaf.fmt = fmt; 2575 vaf.va = &args; 2576 2577 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV", 2578 sb->s_id, function, &vaf); 2579 2580 va_end(args); 2581 2582 /* We don't have the cluster support yet to go straight to 2583 * hard readonly in here. Until then, we want to keep 2584 * ocfs2_abort() so that we can at least mark critical 2585 * errors. 2586 * 2587 * TODO: This should abort the journal and alert other nodes 2588 * that our slot needs recovery. */ 2589 2590 /* Force a panic(). This stinks, but it's better than letting 2591 * things continue without having a proper hard readonly 2592 * here. */ 2593 if (!ocfs2_mount_local(OCFS2_SB(sb))) 2594 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 2595 ocfs2_handle_error(sb); 2596 } 2597 2598 /* 2599 * Void signal blockers, because in-kernel sigprocmask() only fails 2600 * when SIG_* is wrong. 2601 */ 2602 void ocfs2_block_signals(sigset_t *oldset) 2603 { 2604 int rc; 2605 sigset_t blocked; 2606 2607 sigfillset(&blocked); 2608 rc = sigprocmask(SIG_BLOCK, &blocked, oldset); 2609 BUG_ON(rc); 2610 } 2611 2612 void ocfs2_unblock_signals(sigset_t *oldset) 2613 { 2614 int rc = sigprocmask(SIG_SETMASK, oldset, NULL); 2615 BUG_ON(rc); 2616 } 2617 2618 module_init(ocfs2_init); 2619 module_exit(ocfs2_exit); 2620