1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * super.c 4 * 5 * load/unload driver, mount/dismount volumes 6 * 7 * Copyright (C) 2002, 2004 Oracle. All rights reserved. 8 */ 9 10 #include <linux/module.h> 11 #include <linux/fs.h> 12 #include <linux/types.h> 13 #include <linux/slab.h> 14 #include <linux/highmem.h> 15 #include <linux/init.h> 16 #include <linux/random.h> 17 #include <linux/statfs.h> 18 #include <linux/moduleparam.h> 19 #include <linux/blkdev.h> 20 #include <linux/socket.h> 21 #include <linux/inet.h> 22 #include <linux/parser.h> 23 #include <linux/crc32.h> 24 #include <linux/debugfs.h> 25 #include <linux/mount.h> 26 #include <linux/seq_file.h> 27 #include <linux/quotaops.h> 28 #include <linux/signal.h> 29 30 #define CREATE_TRACE_POINTS 31 #include "ocfs2_trace.h" 32 33 #include <cluster/masklog.h> 34 35 #include "ocfs2.h" 36 37 /* this should be the only file to include a version 1 header */ 38 #include "ocfs1_fs_compat.h" 39 40 #include "alloc.h" 41 #include "aops.h" 42 #include "blockcheck.h" 43 #include "dlmglue.h" 44 #include "export.h" 45 #include "extent_map.h" 46 #include "heartbeat.h" 47 #include "inode.h" 48 #include "journal.h" 49 #include "localalloc.h" 50 #include "namei.h" 51 #include "slot_map.h" 52 #include "super.h" 53 #include "sysfile.h" 54 #include "uptodate.h" 55 #include "xattr.h" 56 #include "quota.h" 57 #include "refcounttree.h" 58 #include "suballoc.h" 59 60 #include "buffer_head_io.h" 61 #include "filecheck.h" 62 63 static struct kmem_cache *ocfs2_inode_cachep; 64 struct kmem_cache *ocfs2_dquot_cachep; 65 struct kmem_cache *ocfs2_qf_chunk_cachep; 66 67 static struct dentry *ocfs2_debugfs_root; 68 69 MODULE_AUTHOR("Oracle"); 70 MODULE_LICENSE("GPL"); 71 MODULE_DESCRIPTION("OCFS2 cluster file system"); 72 73 struct mount_options 74 { 75 unsigned long commit_interval; 76 unsigned long mount_opt; 77 unsigned int atime_quantum; 78 unsigned short slot; 79 int localalloc_opt; 80 unsigned int resv_level; 81 int dir_resv_level; 82 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; 83 }; 84 85 static int ocfs2_parse_options(struct super_block *sb, char *options, 86 struct mount_options *mopt, 87 int is_remount); 88 static int ocfs2_check_set_options(struct super_block *sb, 89 struct mount_options *options); 90 static int ocfs2_show_options(struct seq_file *s, struct dentry *root); 91 static void ocfs2_put_super(struct super_block *sb); 92 static int ocfs2_mount_volume(struct super_block *sb); 93 static int ocfs2_remount(struct super_block *sb, int *flags, char *data); 94 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err); 95 static int ocfs2_initialize_mem_caches(void); 96 static void ocfs2_free_mem_caches(void); 97 static void ocfs2_delete_osb(struct ocfs2_super *osb); 98 99 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf); 100 101 static int ocfs2_sync_fs(struct super_block *sb, int wait); 102 103 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb); 104 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb); 105 static void ocfs2_release_system_inodes(struct ocfs2_super *osb); 106 static int ocfs2_check_volume(struct ocfs2_super *osb); 107 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 108 struct buffer_head *bh, 109 u32 sectsize, 110 struct ocfs2_blockcheck_stats *stats); 111 static int ocfs2_initialize_super(struct super_block *sb, 112 struct buffer_head *bh, 113 int sector_size, 114 struct ocfs2_blockcheck_stats *stats); 115 static int ocfs2_get_sector(struct super_block *sb, 116 struct buffer_head **bh, 117 int block, 118 int sect_size); 119 static struct inode *ocfs2_alloc_inode(struct super_block *sb); 120 static void ocfs2_free_inode(struct inode *inode); 121 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend); 122 static int ocfs2_enable_quotas(struct ocfs2_super *osb); 123 static void ocfs2_disable_quotas(struct ocfs2_super *osb); 124 125 static struct dquot __rcu **ocfs2_get_dquots(struct inode *inode) 126 { 127 return OCFS2_I(inode)->i_dquot; 128 } 129 130 static const struct super_operations ocfs2_sops = { 131 .statfs = ocfs2_statfs, 132 .alloc_inode = ocfs2_alloc_inode, 133 .free_inode = ocfs2_free_inode, 134 .drop_inode = ocfs2_drop_inode, 135 .evict_inode = ocfs2_evict_inode, 136 .sync_fs = ocfs2_sync_fs, 137 .put_super = ocfs2_put_super, 138 .remount_fs = ocfs2_remount, 139 .show_options = ocfs2_show_options, 140 .quota_read = ocfs2_quota_read, 141 .quota_write = ocfs2_quota_write, 142 .get_dquots = ocfs2_get_dquots, 143 }; 144 145 enum { 146 Opt_barrier, 147 Opt_err_panic, 148 Opt_err_ro, 149 Opt_intr, 150 Opt_nointr, 151 Opt_hb_none, 152 Opt_hb_local, 153 Opt_hb_global, 154 Opt_data_ordered, 155 Opt_data_writeback, 156 Opt_atime_quantum, 157 Opt_slot, 158 Opt_commit, 159 Opt_localalloc, 160 Opt_localflocks, 161 Opt_stack, 162 Opt_user_xattr, 163 Opt_nouser_xattr, 164 Opt_inode64, 165 Opt_acl, 166 Opt_noacl, 167 Opt_usrquota, 168 Opt_grpquota, 169 Opt_coherency_buffered, 170 Opt_coherency_full, 171 Opt_resv_level, 172 Opt_dir_resv_level, 173 Opt_journal_async_commit, 174 Opt_err_cont, 175 Opt_err, 176 }; 177 178 static const match_table_t tokens = { 179 {Opt_barrier, "barrier=%u"}, 180 {Opt_err_panic, "errors=panic"}, 181 {Opt_err_ro, "errors=remount-ro"}, 182 {Opt_intr, "intr"}, 183 {Opt_nointr, "nointr"}, 184 {Opt_hb_none, OCFS2_HB_NONE}, 185 {Opt_hb_local, OCFS2_HB_LOCAL}, 186 {Opt_hb_global, OCFS2_HB_GLOBAL}, 187 {Opt_data_ordered, "data=ordered"}, 188 {Opt_data_writeback, "data=writeback"}, 189 {Opt_atime_quantum, "atime_quantum=%u"}, 190 {Opt_slot, "preferred_slot=%u"}, 191 {Opt_commit, "commit=%u"}, 192 {Opt_localalloc, "localalloc=%d"}, 193 {Opt_localflocks, "localflocks"}, 194 {Opt_stack, "cluster_stack=%s"}, 195 {Opt_user_xattr, "user_xattr"}, 196 {Opt_nouser_xattr, "nouser_xattr"}, 197 {Opt_inode64, "inode64"}, 198 {Opt_acl, "acl"}, 199 {Opt_noacl, "noacl"}, 200 {Opt_usrquota, "usrquota"}, 201 {Opt_grpquota, "grpquota"}, 202 {Opt_coherency_buffered, "coherency=buffered"}, 203 {Opt_coherency_full, "coherency=full"}, 204 {Opt_resv_level, "resv_level=%u"}, 205 {Opt_dir_resv_level, "dir_resv_level=%u"}, 206 {Opt_journal_async_commit, "journal_async_commit"}, 207 {Opt_err_cont, "errors=continue"}, 208 {Opt_err, NULL} 209 }; 210 211 #ifdef CONFIG_DEBUG_FS 212 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len) 213 { 214 struct ocfs2_cluster_connection *cconn = osb->cconn; 215 struct ocfs2_recovery_map *rm = osb->recovery_map; 216 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan; 217 int i, out = 0; 218 unsigned long flags; 219 220 out += scnprintf(buf + out, len - out, 221 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n", 222 "Device", osb->dev_str, osb->uuid_str, 223 osb->fs_generation, osb->vol_label); 224 225 out += scnprintf(buf + out, len - out, 226 "%10s => State: %d Flags: 0x%lX\n", "Volume", 227 atomic_read(&osb->vol_state), osb->osb_flags); 228 229 out += scnprintf(buf + out, len - out, 230 "%10s => Block: %lu Cluster: %d\n", "Sizes", 231 osb->sb->s_blocksize, osb->s_clustersize); 232 233 out += scnprintf(buf + out, len - out, 234 "%10s => Compat: 0x%X Incompat: 0x%X " 235 "ROcompat: 0x%X\n", 236 "Features", osb->s_feature_compat, 237 osb->s_feature_incompat, osb->s_feature_ro_compat); 238 239 out += scnprintf(buf + out, len - out, 240 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount", 241 osb->s_mount_opt, osb->s_atime_quantum); 242 243 if (cconn) { 244 out += scnprintf(buf + out, len - out, 245 "%10s => Stack: %s Name: %*s " 246 "Version: %d.%d\n", "Cluster", 247 (*osb->osb_cluster_stack == '\0' ? 248 "o2cb" : osb->osb_cluster_stack), 249 cconn->cc_namelen, cconn->cc_name, 250 cconn->cc_version.pv_major, 251 cconn->cc_version.pv_minor); 252 } 253 254 spin_lock_irqsave(&osb->dc_task_lock, flags); 255 out += scnprintf(buf + out, len - out, 256 "%10s => Pid: %d Count: %lu WakeSeq: %lu " 257 "WorkSeq: %lu\n", "DownCnvt", 258 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1), 259 osb->blocked_lock_count, osb->dc_wake_sequence, 260 osb->dc_work_sequence); 261 spin_unlock_irqrestore(&osb->dc_task_lock, flags); 262 263 spin_lock(&osb->osb_lock); 264 out += scnprintf(buf + out, len - out, "%10s => Pid: %d Nodes:", 265 "Recovery", 266 (osb->recovery_thread_task ? 267 task_pid_nr(osb->recovery_thread_task) : -1)); 268 if (rm->rm_used == 0) 269 out += scnprintf(buf + out, len - out, " None\n"); 270 else { 271 for (i = 0; i < rm->rm_used; i++) 272 out += scnprintf(buf + out, len - out, " %d", 273 rm->rm_entries[i]); 274 out += scnprintf(buf + out, len - out, "\n"); 275 } 276 spin_unlock(&osb->osb_lock); 277 278 out += scnprintf(buf + out, len - out, 279 "%10s => Pid: %d Interval: %lu\n", "Commit", 280 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1), 281 osb->osb_commit_interval); 282 283 out += scnprintf(buf + out, len - out, 284 "%10s => State: %d TxnId: %lu NumTxns: %d\n", 285 "Journal", osb->journal->j_state, 286 osb->journal->j_trans_id, 287 atomic_read(&osb->journal->j_num_trans)); 288 289 out += scnprintf(buf + out, len - out, 290 "%10s => GlobalAllocs: %d LocalAllocs: %d " 291 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n", 292 "Stats", 293 atomic_read(&osb->alloc_stats.bitmap_data), 294 atomic_read(&osb->alloc_stats.local_data), 295 atomic_read(&osb->alloc_stats.bg_allocs), 296 atomic_read(&osb->alloc_stats.moves), 297 atomic_read(&osb->alloc_stats.bg_extends)); 298 299 out += scnprintf(buf + out, len - out, 300 "%10s => State: %u Descriptor: %llu Size: %u bits " 301 "Default: %u bits\n", 302 "LocalAlloc", osb->local_alloc_state, 303 (unsigned long long)osb->la_last_gd, 304 osb->local_alloc_bits, osb->local_alloc_default_bits); 305 306 spin_lock(&osb->osb_lock); 307 out += scnprintf(buf + out, len - out, 308 "%10s => InodeSlot: %d StolenInodes: %d, " 309 "MetaSlot: %d StolenMeta: %d\n", "Steal", 310 osb->s_inode_steal_slot, 311 atomic_read(&osb->s_num_inodes_stolen), 312 osb->s_meta_steal_slot, 313 atomic_read(&osb->s_num_meta_stolen)); 314 spin_unlock(&osb->osb_lock); 315 316 out += scnprintf(buf + out, len - out, "OrphanScan => "); 317 out += scnprintf(buf + out, len - out, "Local: %u Global: %u ", 318 os->os_count, os->os_seqno); 319 out += scnprintf(buf + out, len - out, " Last Scan: "); 320 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) 321 out += scnprintf(buf + out, len - out, "Disabled\n"); 322 else 323 out += scnprintf(buf + out, len - out, "%lu seconds ago\n", 324 (unsigned long)(ktime_get_seconds() - os->os_scantime)); 325 326 out += scnprintf(buf + out, len - out, "%10s => %3s %10s\n", 327 "Slots", "Num", "RecoGen"); 328 for (i = 0; i < osb->max_slots; ++i) { 329 out += scnprintf(buf + out, len - out, 330 "%10s %c %3d %10d\n", 331 " ", 332 (i == osb->slot_num ? '*' : ' '), 333 i, osb->slot_recovery_generations[i]); 334 } 335 336 return out; 337 } 338 339 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file) 340 { 341 struct ocfs2_super *osb = inode->i_private; 342 char *buf = NULL; 343 344 buf = kmalloc(PAGE_SIZE, GFP_KERNEL); 345 if (!buf) 346 goto bail; 347 348 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE)); 349 350 file->private_data = buf; 351 352 return 0; 353 bail: 354 return -ENOMEM; 355 } 356 357 static int ocfs2_debug_release(struct inode *inode, struct file *file) 358 { 359 kfree(file->private_data); 360 return 0; 361 } 362 363 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, 364 size_t nbytes, loff_t *ppos) 365 { 366 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, 367 i_size_read(file->f_mapping->host)); 368 } 369 #else 370 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file) 371 { 372 return 0; 373 } 374 static int ocfs2_debug_release(struct inode *inode, struct file *file) 375 { 376 return 0; 377 } 378 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, 379 size_t nbytes, loff_t *ppos) 380 { 381 return 0; 382 } 383 #endif /* CONFIG_DEBUG_FS */ 384 385 static const struct file_operations ocfs2_osb_debug_fops = { 386 .open = ocfs2_osb_debug_open, 387 .release = ocfs2_debug_release, 388 .read = ocfs2_debug_read, 389 .llseek = generic_file_llseek, 390 }; 391 392 static int ocfs2_sync_fs(struct super_block *sb, int wait) 393 { 394 int status; 395 tid_t target; 396 struct ocfs2_super *osb = OCFS2_SB(sb); 397 398 if (ocfs2_is_hard_readonly(osb)) 399 return -EROFS; 400 401 if (wait) { 402 status = ocfs2_flush_truncate_log(osb); 403 if (status < 0) 404 mlog_errno(status); 405 } else { 406 ocfs2_schedule_truncate_log_flush(osb, 0); 407 } 408 409 if (jbd2_journal_start_commit(osb->journal->j_journal, 410 &target)) { 411 if (wait) 412 jbd2_log_wait_commit(osb->journal->j_journal, 413 target); 414 } 415 return 0; 416 } 417 418 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino) 419 { 420 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA) 421 && (ino == USER_QUOTA_SYSTEM_INODE 422 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE)) 423 return 0; 424 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA) 425 && (ino == GROUP_QUOTA_SYSTEM_INODE 426 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE)) 427 return 0; 428 return 1; 429 } 430 431 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) 432 { 433 struct inode *new = NULL; 434 int status = 0; 435 int i; 436 437 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0); 438 if (IS_ERR(new)) { 439 status = PTR_ERR(new); 440 mlog_errno(status); 441 goto bail; 442 } 443 osb->root_inode = new; 444 445 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0); 446 if (IS_ERR(new)) { 447 status = PTR_ERR(new); 448 mlog_errno(status); 449 goto bail; 450 } 451 osb->sys_root_inode = new; 452 453 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE; 454 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) { 455 if (!ocfs2_need_system_inode(osb, i)) 456 continue; 457 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 458 if (!new) { 459 ocfs2_release_system_inodes(osb); 460 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL; 461 mlog_errno(status); 462 mlog(ML_ERROR, "Unable to load system inode %d, " 463 "possibly corrupt fs?", i); 464 goto bail; 465 } 466 // the array now has one ref, so drop this one 467 iput(new); 468 } 469 470 bail: 471 if (status) 472 mlog_errno(status); 473 return status; 474 } 475 476 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) 477 { 478 struct inode *new = NULL; 479 int status = 0; 480 int i; 481 482 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1; 483 i < NUM_SYSTEM_INODES; 484 i++) { 485 if (!ocfs2_need_system_inode(osb, i)) 486 continue; 487 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 488 if (!new) { 489 ocfs2_release_system_inodes(osb); 490 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL; 491 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n", 492 status, i, osb->slot_num); 493 goto bail; 494 } 495 /* the array now has one ref, so drop this one */ 496 iput(new); 497 } 498 499 bail: 500 if (status) 501 mlog_errno(status); 502 return status; 503 } 504 505 static void ocfs2_release_system_inodes(struct ocfs2_super *osb) 506 { 507 int i; 508 struct inode *inode; 509 510 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) { 511 inode = osb->global_system_inodes[i]; 512 if (inode) { 513 iput(inode); 514 osb->global_system_inodes[i] = NULL; 515 } 516 } 517 518 inode = osb->sys_root_inode; 519 if (inode) { 520 iput(inode); 521 osb->sys_root_inode = NULL; 522 } 523 524 inode = osb->root_inode; 525 if (inode) { 526 iput(inode); 527 osb->root_inode = NULL; 528 } 529 530 if (!osb->local_system_inodes) 531 return; 532 533 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) { 534 if (osb->local_system_inodes[i]) { 535 iput(osb->local_system_inodes[i]); 536 osb->local_system_inodes[i] = NULL; 537 } 538 } 539 540 kfree(osb->local_system_inodes); 541 osb->local_system_inodes = NULL; 542 } 543 544 /* We're allocating fs objects, use GFP_NOFS */ 545 static struct inode *ocfs2_alloc_inode(struct super_block *sb) 546 { 547 struct ocfs2_inode_info *oi; 548 549 oi = alloc_inode_sb(sb, ocfs2_inode_cachep, GFP_NOFS); 550 if (!oi) 551 return NULL; 552 553 oi->i_sync_tid = 0; 554 oi->i_datasync_tid = 0; 555 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot)); 556 557 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode); 558 return &oi->vfs_inode; 559 } 560 561 static void ocfs2_free_inode(struct inode *inode) 562 { 563 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); 564 } 565 566 static unsigned long long ocfs2_max_file_offset(unsigned int bbits, 567 unsigned int cbits) 568 { 569 unsigned int bytes = 1 << cbits; 570 unsigned int trim = bytes; 571 unsigned int bitshift = 32; 572 573 /* 574 * i_size and all block offsets in ocfs2 are always 64 bits 575 * wide. i_clusters is 32 bits, in cluster-sized units. So on 576 * 64 bit platforms, cluster size will be the limiting factor. 577 */ 578 579 #if BITS_PER_LONG == 32 580 BUILD_BUG_ON(sizeof(sector_t) != 8); 581 /* 582 * We might be limited by page cache size. 583 */ 584 if (bytes > PAGE_SIZE) { 585 bytes = PAGE_SIZE; 586 trim = 1; 587 /* 588 * Shift by 31 here so that we don't get larger than 589 * MAX_LFS_FILESIZE 590 */ 591 bitshift = 31; 592 } 593 #endif 594 595 /* 596 * Trim by a whole cluster when we can actually approach the 597 * on-disk limits. Otherwise we can overflow i_clusters when 598 * an extent start is at the max offset. 599 */ 600 return (((unsigned long long)bytes) << bitshift) - trim; 601 } 602 603 static int ocfs2_remount(struct super_block *sb, int *flags, char *data) 604 { 605 int incompat_features; 606 int ret = 0; 607 struct mount_options parsed_options; 608 struct ocfs2_super *osb = OCFS2_SB(sb); 609 u32 tmp; 610 611 sync_filesystem(sb); 612 613 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) || 614 !ocfs2_check_set_options(sb, &parsed_options)) { 615 ret = -EINVAL; 616 goto out; 617 } 618 619 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL | 620 OCFS2_MOUNT_HB_NONE; 621 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) { 622 ret = -EINVAL; 623 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); 624 goto out; 625 } 626 627 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) != 628 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) { 629 ret = -EINVAL; 630 mlog(ML_ERROR, "Cannot change data mode on remount\n"); 631 goto out; 632 } 633 634 /* Probably don't want this on remount; it might 635 * mess with other nodes */ 636 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) && 637 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) { 638 ret = -EINVAL; 639 mlog(ML_ERROR, "Cannot enable inode64 on remount\n"); 640 goto out; 641 } 642 643 /* We're going to/from readonly mode. */ 644 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 645 /* Disable quota accounting before remounting RO */ 646 if (*flags & SB_RDONLY) { 647 ret = ocfs2_susp_quotas(osb, 0); 648 if (ret < 0) 649 goto out; 650 } 651 /* Lock here so the check of HARD_RO and the potential 652 * setting of SOFT_RO is atomic. */ 653 spin_lock(&osb->osb_lock); 654 if (osb->osb_flags & OCFS2_OSB_HARD_RO) { 655 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n"); 656 ret = -EROFS; 657 goto unlock_osb; 658 } 659 660 if (*flags & SB_RDONLY) { 661 sb->s_flags |= SB_RDONLY; 662 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 663 } else { 664 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) { 665 mlog(ML_ERROR, "Cannot remount RDWR " 666 "filesystem due to previous errors.\n"); 667 ret = -EROFS; 668 goto unlock_osb; 669 } 670 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP); 671 if (incompat_features) { 672 mlog(ML_ERROR, "Cannot remount RDWR because " 673 "of unsupported optional features " 674 "(%x).\n", incompat_features); 675 ret = -EINVAL; 676 goto unlock_osb; 677 } 678 sb->s_flags &= ~SB_RDONLY; 679 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO; 680 } 681 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags); 682 unlock_osb: 683 spin_unlock(&osb->osb_lock); 684 /* Enable quota accounting after remounting RW */ 685 if (!ret && !(*flags & SB_RDONLY)) { 686 if (sb_any_quota_suspended(sb)) 687 ret = ocfs2_susp_quotas(osb, 1); 688 else 689 ret = ocfs2_enable_quotas(osb); 690 if (ret < 0) { 691 /* Return back changes... */ 692 spin_lock(&osb->osb_lock); 693 sb->s_flags |= SB_RDONLY; 694 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 695 spin_unlock(&osb->osb_lock); 696 goto out; 697 } 698 } 699 } 700 701 if (!ret) { 702 /* Only save off the new mount options in case of a successful 703 * remount. */ 704 osb->s_mount_opt = parsed_options.mount_opt; 705 osb->s_atime_quantum = parsed_options.atime_quantum; 706 osb->preferred_slot = parsed_options.slot; 707 if (parsed_options.commit_interval) 708 osb->osb_commit_interval = parsed_options.commit_interval; 709 710 if (!ocfs2_is_hard_readonly(osb)) 711 ocfs2_set_journal_params(osb); 712 713 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 714 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? 715 SB_POSIXACL : 0); 716 } 717 out: 718 return ret; 719 } 720 721 static int ocfs2_sb_probe(struct super_block *sb, 722 struct buffer_head **bh, 723 int *sector_size, 724 struct ocfs2_blockcheck_stats *stats) 725 { 726 int status, tmpstat; 727 struct ocfs1_vol_disk_hdr *hdr; 728 struct ocfs2_dinode *di; 729 int blksize; 730 731 *bh = NULL; 732 733 /* may be > 512 */ 734 *sector_size = bdev_logical_block_size(sb->s_bdev); 735 if (*sector_size > OCFS2_MAX_BLOCKSIZE) { 736 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n", 737 *sector_size, OCFS2_MAX_BLOCKSIZE); 738 status = -EINVAL; 739 goto bail; 740 } 741 742 /* Can this really happen? */ 743 if (*sector_size < OCFS2_MIN_BLOCKSIZE) 744 *sector_size = OCFS2_MIN_BLOCKSIZE; 745 746 /* check block zero for old format */ 747 status = ocfs2_get_sector(sb, bh, 0, *sector_size); 748 if (status < 0) { 749 mlog_errno(status); 750 goto bail; 751 } 752 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data; 753 if (hdr->major_version == OCFS1_MAJOR_VERSION) { 754 mlog(ML_ERROR, "incompatible version: %u.%u\n", 755 hdr->major_version, hdr->minor_version); 756 status = -EINVAL; 757 } 758 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE, 759 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) { 760 mlog(ML_ERROR, "incompatible volume signature: %8s\n", 761 hdr->signature); 762 status = -EINVAL; 763 } 764 brelse(*bh); 765 *bh = NULL; 766 if (status < 0) { 767 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be " 768 "upgraded before mounting with ocfs v2\n"); 769 goto bail; 770 } 771 772 /* 773 * Now check at magic offset for 512, 1024, 2048, 4096 774 * blocksizes. 4096 is the maximum blocksize because it is 775 * the minimum clustersize. 776 */ 777 status = -EINVAL; 778 for (blksize = *sector_size; 779 blksize <= OCFS2_MAX_BLOCKSIZE; 780 blksize <<= 1) { 781 tmpstat = ocfs2_get_sector(sb, bh, 782 OCFS2_SUPER_BLOCK_BLKNO, 783 blksize); 784 if (tmpstat < 0) { 785 status = tmpstat; 786 mlog_errno(status); 787 break; 788 } 789 di = (struct ocfs2_dinode *) (*bh)->b_data; 790 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats)); 791 spin_lock_init(&stats->b_lock); 792 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats); 793 if (tmpstat < 0) { 794 brelse(*bh); 795 *bh = NULL; 796 } 797 if (tmpstat != -EAGAIN) { 798 status = tmpstat; 799 break; 800 } 801 } 802 803 bail: 804 return status; 805 } 806 807 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) 808 { 809 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL; 810 811 if (osb->s_mount_opt & hb_enabled) { 812 if (ocfs2_mount_local(osb)) { 813 mlog(ML_ERROR, "Cannot heartbeat on a locally " 814 "mounted device.\n"); 815 return -EINVAL; 816 } 817 if (ocfs2_userspace_stack(osb)) { 818 mlog(ML_ERROR, "Userspace stack expected, but " 819 "o2cb heartbeat arguments passed to mount\n"); 820 return -EINVAL; 821 } 822 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) && 823 !ocfs2_cluster_o2cb_global_heartbeat(osb)) || 824 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) && 825 ocfs2_cluster_o2cb_global_heartbeat(osb))) { 826 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n"); 827 return -EINVAL; 828 } 829 } 830 831 if (!(osb->s_mount_opt & hb_enabled)) { 832 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) && 833 !ocfs2_userspace_stack(osb)) { 834 mlog(ML_ERROR, "Heartbeat has to be started to mount " 835 "a read-write clustered device.\n"); 836 return -EINVAL; 837 } 838 } 839 840 return 0; 841 } 842 843 /* 844 * If we're using a userspace stack, mount should have passed 845 * a name that matches the disk. If not, mount should not 846 * have passed a stack. 847 */ 848 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb, 849 struct mount_options *mopt) 850 { 851 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) { 852 mlog(ML_ERROR, 853 "cluster stack passed to mount, but this filesystem " 854 "does not support it\n"); 855 return -EINVAL; 856 } 857 858 if (ocfs2_userspace_stack(osb) && 859 strncmp(osb->osb_cluster_stack, mopt->cluster_stack, 860 OCFS2_STACK_LABEL_LEN)) { 861 mlog(ML_ERROR, 862 "cluster stack passed to mount (\"%s\") does not " 863 "match the filesystem (\"%s\")\n", 864 mopt->cluster_stack, 865 osb->osb_cluster_stack); 866 return -EINVAL; 867 } 868 869 return 0; 870 } 871 872 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend) 873 { 874 int type; 875 struct super_block *sb = osb->sb; 876 unsigned int feature[OCFS2_MAXQUOTAS] = { 877 OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 878 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 879 int status = 0; 880 881 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 882 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 883 continue; 884 if (unsuspend) 885 status = dquot_resume(sb, type); 886 else { 887 struct ocfs2_mem_dqinfo *oinfo; 888 889 /* Cancel periodic syncing before suspending */ 890 oinfo = sb_dqinfo(sb, type)->dqi_priv; 891 cancel_delayed_work_sync(&oinfo->dqi_sync_work); 892 status = dquot_suspend(sb, type); 893 } 894 if (status < 0) 895 break; 896 } 897 if (status < 0) 898 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on " 899 "remount (error = %d).\n", status); 900 return status; 901 } 902 903 static int ocfs2_enable_quotas(struct ocfs2_super *osb) 904 { 905 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL }; 906 struct super_block *sb = osb->sb; 907 unsigned int feature[OCFS2_MAXQUOTAS] = { 908 OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 909 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 910 unsigned int ino[OCFS2_MAXQUOTAS] = { 911 LOCAL_USER_QUOTA_SYSTEM_INODE, 912 LOCAL_GROUP_QUOTA_SYSTEM_INODE }; 913 int status; 914 int type; 915 916 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE; 917 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 918 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 919 continue; 920 inode[type] = ocfs2_get_system_file_inode(osb, ino[type], 921 osb->slot_num); 922 if (!inode[type]) { 923 status = -ENOENT; 924 goto out_quota_off; 925 } 926 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2, 927 DQUOT_USAGE_ENABLED); 928 if (status < 0) 929 goto out_quota_off; 930 } 931 932 for (type = 0; type < OCFS2_MAXQUOTAS; type++) 933 iput(inode[type]); 934 return 0; 935 out_quota_off: 936 ocfs2_disable_quotas(osb); 937 for (type = 0; type < OCFS2_MAXQUOTAS; type++) 938 iput(inode[type]); 939 mlog_errno(status); 940 return status; 941 } 942 943 static void ocfs2_disable_quotas(struct ocfs2_super *osb) 944 { 945 int type; 946 struct inode *inode; 947 struct super_block *sb = osb->sb; 948 struct ocfs2_mem_dqinfo *oinfo; 949 950 /* We mostly ignore errors in this function because there's not much 951 * we can do when we see them */ 952 for (type = 0; type < OCFS2_MAXQUOTAS; type++) { 953 if (!sb_has_quota_loaded(sb, type)) 954 continue; 955 if (!sb_has_quota_suspended(sb, type)) { 956 oinfo = sb_dqinfo(sb, type)->dqi_priv; 957 cancel_delayed_work_sync(&oinfo->dqi_sync_work); 958 } 959 inode = igrab(sb->s_dquot.files[type]); 960 /* Turn off quotas. This will remove all dquot structures from 961 * memory and so they will be automatically synced to global 962 * quota files */ 963 dquot_disable(sb, type, DQUOT_USAGE_ENABLED | 964 DQUOT_LIMITS_ENABLED); 965 iput(inode); 966 } 967 } 968 969 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) 970 { 971 struct dentry *root; 972 int status, sector_size; 973 struct mount_options parsed_options; 974 struct inode *inode = NULL; 975 struct ocfs2_super *osb = NULL; 976 struct buffer_head *bh = NULL; 977 char nodestr[12]; 978 struct ocfs2_blockcheck_stats stats; 979 980 trace_ocfs2_fill_super(sb, data, silent); 981 982 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { 983 status = -EINVAL; 984 goto out; 985 } 986 987 /* probe for superblock */ 988 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats); 989 if (status < 0) { 990 mlog(ML_ERROR, "superblock probe failed!\n"); 991 goto out; 992 } 993 994 status = ocfs2_initialize_super(sb, bh, sector_size, &stats); 995 brelse(bh); 996 bh = NULL; 997 if (status < 0) 998 goto out; 999 1000 osb = OCFS2_SB(sb); 1001 1002 if (!ocfs2_check_set_options(sb, &parsed_options)) { 1003 status = -EINVAL; 1004 goto out_super; 1005 } 1006 osb->s_mount_opt = parsed_options.mount_opt; 1007 osb->s_atime_quantum = parsed_options.atime_quantum; 1008 osb->preferred_slot = parsed_options.slot; 1009 osb->osb_commit_interval = parsed_options.commit_interval; 1010 1011 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt); 1012 osb->osb_resv_level = parsed_options.resv_level; 1013 osb->osb_dir_resv_level = parsed_options.resv_level; 1014 if (parsed_options.dir_resv_level == -1) 1015 osb->osb_dir_resv_level = parsed_options.resv_level; 1016 else 1017 osb->osb_dir_resv_level = parsed_options.dir_resv_level; 1018 1019 status = ocfs2_verify_userspace_stack(osb, &parsed_options); 1020 if (status) 1021 goto out_super; 1022 1023 sb->s_magic = OCFS2_SUPER_MAGIC; 1024 1025 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) | 1026 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0); 1027 1028 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY, 1029 * heartbeat=none */ 1030 if (bdev_read_only(sb->s_bdev)) { 1031 if (!sb_rdonly(sb)) { 1032 status = -EACCES; 1033 mlog(ML_ERROR, "Readonly device detected but readonly " 1034 "mount was not specified.\n"); 1035 goto out_super; 1036 } 1037 1038 /* You should not be able to start a local heartbeat 1039 * on a readonly device. */ 1040 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 1041 status = -EROFS; 1042 mlog(ML_ERROR, "Local heartbeat specified on readonly " 1043 "device.\n"); 1044 goto out_super; 1045 } 1046 1047 status = ocfs2_check_journals_nolocks(osb); 1048 if (status < 0) { 1049 if (status == -EROFS) 1050 mlog(ML_ERROR, "Recovery required on readonly " 1051 "file system, but write access is " 1052 "unavailable.\n"); 1053 goto out_super; 1054 } 1055 1056 ocfs2_set_ro_flag(osb, 1); 1057 1058 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. " 1059 "Cluster services will not be used for this mount. " 1060 "Recovery will be skipped.\n", osb->dev_str); 1061 } 1062 1063 if (!ocfs2_is_hard_readonly(osb)) { 1064 if (sb_rdonly(sb)) 1065 ocfs2_set_ro_flag(osb, 0); 1066 } 1067 1068 status = ocfs2_verify_heartbeat(osb); 1069 if (status < 0) 1070 goto out_super; 1071 1072 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, 1073 ocfs2_debugfs_root); 1074 1075 debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root, 1076 osb, &ocfs2_osb_debug_fops); 1077 1078 if (ocfs2_meta_ecc(osb)) { 1079 ocfs2_initialize_journal_triggers(sb, osb->s_journal_triggers); 1080 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats, 1081 osb->osb_debug_root); 1082 } 1083 1084 status = ocfs2_mount_volume(sb); 1085 if (status < 0) 1086 goto out_debugfs; 1087 1088 if (osb->root_inode) 1089 inode = igrab(osb->root_inode); 1090 1091 if (!inode) { 1092 status = -EIO; 1093 goto out_dismount; 1094 } 1095 1096 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL, 1097 &ocfs2_kset->kobj); 1098 if (!osb->osb_dev_kset) { 1099 status = -ENOMEM; 1100 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id); 1101 goto out_dismount; 1102 } 1103 1104 /* Create filecheck sysfs related directories/files at 1105 * /sys/fs/ocfs2/<devname>/filecheck */ 1106 if (ocfs2_filecheck_create_sysfs(osb)) { 1107 status = -ENOMEM; 1108 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at " 1109 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id); 1110 goto out_dismount; 1111 } 1112 1113 root = d_make_root(inode); 1114 if (!root) { 1115 status = -ENOMEM; 1116 goto out_dismount; 1117 } 1118 1119 sb->s_root = root; 1120 1121 ocfs2_complete_mount_recovery(osb); 1122 1123 if (ocfs2_mount_local(osb)) 1124 snprintf(nodestr, sizeof(nodestr), "local"); 1125 else 1126 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1127 1128 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " 1129 "with %s data mode.\n", 1130 osb->dev_str, nodestr, osb->slot_num, 1131 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : 1132 "ordered"); 1133 1134 atomic_set(&osb->vol_state, VOLUME_MOUNTED); 1135 wake_up(&osb->osb_mount_event); 1136 1137 /* Now we can initialize quotas because we can afford to wait 1138 * for cluster locks recovery now. That also means that truncation 1139 * log recovery can happen but that waits for proper quota setup */ 1140 if (!sb_rdonly(sb)) { 1141 status = ocfs2_enable_quotas(osb); 1142 if (status < 0) { 1143 /* We have to err-out specially here because 1144 * s_root is already set */ 1145 mlog_errno(status); 1146 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1147 wake_up(&osb->osb_mount_event); 1148 return status; 1149 } 1150 } 1151 1152 ocfs2_complete_quota_recovery(osb); 1153 1154 /* Now we wake up again for processes waiting for quotas */ 1155 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS); 1156 wake_up(&osb->osb_mount_event); 1157 1158 /* Start this when the mount is almost sure of being successful */ 1159 ocfs2_orphan_scan_start(osb); 1160 1161 return status; 1162 1163 out_dismount: 1164 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1165 wake_up(&osb->osb_mount_event); 1166 ocfs2_free_replay_slots(osb); 1167 ocfs2_dismount_volume(sb, 1); 1168 goto out; 1169 1170 out_debugfs: 1171 debugfs_remove_recursive(osb->osb_debug_root); 1172 out_super: 1173 ocfs2_release_system_inodes(osb); 1174 kfree(osb->recovery_map); 1175 ocfs2_delete_osb(osb); 1176 kfree(osb); 1177 out: 1178 mlog_errno(status); 1179 1180 return status; 1181 } 1182 1183 static struct dentry *ocfs2_mount(struct file_system_type *fs_type, 1184 int flags, 1185 const char *dev_name, 1186 void *data) 1187 { 1188 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super); 1189 } 1190 1191 static struct file_system_type ocfs2_fs_type = { 1192 .owner = THIS_MODULE, 1193 .name = "ocfs2", 1194 .mount = ocfs2_mount, 1195 .kill_sb = kill_block_super, 1196 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, 1197 .next = NULL 1198 }; 1199 MODULE_ALIAS_FS("ocfs2"); 1200 1201 static int ocfs2_check_set_options(struct super_block *sb, 1202 struct mount_options *options) 1203 { 1204 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA && 1205 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1206 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { 1207 mlog(ML_ERROR, "User quotas were requested, but this " 1208 "filesystem does not have the feature enabled.\n"); 1209 return 0; 1210 } 1211 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA && 1212 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1213 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { 1214 mlog(ML_ERROR, "Group quotas were requested, but this " 1215 "filesystem does not have the feature enabled.\n"); 1216 return 0; 1217 } 1218 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL && 1219 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) { 1220 mlog(ML_ERROR, "ACL support requested but extended attributes " 1221 "feature is not enabled\n"); 1222 return 0; 1223 } 1224 /* No ACL setting specified? Use XATTR feature... */ 1225 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL | 1226 OCFS2_MOUNT_NO_POSIX_ACL))) { 1227 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) 1228 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1229 else 1230 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; 1231 } 1232 return 1; 1233 } 1234 1235 static int ocfs2_parse_options(struct super_block *sb, 1236 char *options, 1237 struct mount_options *mopt, 1238 int is_remount) 1239 { 1240 int status, user_stack = 0; 1241 char *p; 1242 u32 tmp; 1243 int token, option; 1244 substring_t args[MAX_OPT_ARGS]; 1245 1246 trace_ocfs2_parse_options(is_remount, options ? options : "(none)"); 1247 1248 mopt->commit_interval = 0; 1249 mopt->mount_opt = OCFS2_MOUNT_NOINTR; 1250 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 1251 mopt->slot = OCFS2_INVALID_SLOT; 1252 mopt->localalloc_opt = -1; 1253 mopt->cluster_stack[0] = '\0'; 1254 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL; 1255 mopt->dir_resv_level = -1; 1256 1257 if (!options) { 1258 status = 1; 1259 goto bail; 1260 } 1261 1262 while ((p = strsep(&options, ",")) != NULL) { 1263 if (!*p) 1264 continue; 1265 1266 token = match_token(p, tokens, args); 1267 switch (token) { 1268 case Opt_hb_local: 1269 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; 1270 break; 1271 case Opt_hb_none: 1272 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE; 1273 break; 1274 case Opt_hb_global: 1275 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL; 1276 break; 1277 case Opt_barrier: 1278 if (match_int(&args[0], &option)) { 1279 status = 0; 1280 goto bail; 1281 } 1282 if (option) 1283 mopt->mount_opt |= OCFS2_MOUNT_BARRIER; 1284 else 1285 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; 1286 break; 1287 case Opt_intr: 1288 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; 1289 break; 1290 case Opt_nointr: 1291 mopt->mount_opt |= OCFS2_MOUNT_NOINTR; 1292 break; 1293 case Opt_err_panic: 1294 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT; 1295 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS; 1296 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 1297 break; 1298 case Opt_err_ro: 1299 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT; 1300 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1301 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS; 1302 break; 1303 case Opt_err_cont: 1304 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS; 1305 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1306 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT; 1307 break; 1308 case Opt_data_ordered: 1309 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; 1310 break; 1311 case Opt_data_writeback: 1312 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; 1313 break; 1314 case Opt_user_xattr: 1315 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR; 1316 break; 1317 case Opt_nouser_xattr: 1318 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR; 1319 break; 1320 case Opt_atime_quantum: 1321 if (match_int(&args[0], &option)) { 1322 status = 0; 1323 goto bail; 1324 } 1325 if (option >= 0) 1326 mopt->atime_quantum = option; 1327 break; 1328 case Opt_slot: 1329 if (match_int(&args[0], &option)) { 1330 status = 0; 1331 goto bail; 1332 } 1333 if (option) 1334 mopt->slot = (u16)option; 1335 break; 1336 case Opt_commit: 1337 if (match_int(&args[0], &option)) { 1338 status = 0; 1339 goto bail; 1340 } 1341 if (option < 0) 1342 return 0; 1343 if (option == 0) 1344 option = JBD2_DEFAULT_MAX_COMMIT_AGE; 1345 mopt->commit_interval = HZ * option; 1346 break; 1347 case Opt_localalloc: 1348 if (match_int(&args[0], &option)) { 1349 status = 0; 1350 goto bail; 1351 } 1352 if (option >= 0) 1353 mopt->localalloc_opt = option; 1354 break; 1355 case Opt_localflocks: 1356 /* 1357 * Changing this during remount could race 1358 * flock() requests, or "unbalance" existing 1359 * ones (e.g., a lock is taken in one mode but 1360 * dropped in the other). If users care enough 1361 * to flip locking modes during remount, we 1362 * could add a "local" flag to individual 1363 * flock structures for proper tracking of 1364 * state. 1365 */ 1366 if (!is_remount) 1367 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS; 1368 break; 1369 case Opt_stack: 1370 /* Check both that the option we were passed 1371 * is of the right length and that it is a proper 1372 * string of the right length. 1373 */ 1374 if (((args[0].to - args[0].from) != 1375 OCFS2_STACK_LABEL_LEN) || 1376 (strnlen(args[0].from, 1377 OCFS2_STACK_LABEL_LEN) != 1378 OCFS2_STACK_LABEL_LEN)) { 1379 mlog(ML_ERROR, 1380 "Invalid cluster_stack option\n"); 1381 status = 0; 1382 goto bail; 1383 } 1384 memcpy(mopt->cluster_stack, args[0].from, 1385 OCFS2_STACK_LABEL_LEN); 1386 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; 1387 /* 1388 * Open code the memcmp here as we don't have 1389 * an osb to pass to 1390 * ocfs2_userspace_stack(). 1391 */ 1392 if (memcmp(mopt->cluster_stack, 1393 OCFS2_CLASSIC_CLUSTER_STACK, 1394 OCFS2_STACK_LABEL_LEN)) 1395 user_stack = 1; 1396 break; 1397 case Opt_inode64: 1398 mopt->mount_opt |= OCFS2_MOUNT_INODE64; 1399 break; 1400 case Opt_usrquota: 1401 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA; 1402 break; 1403 case Opt_grpquota: 1404 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA; 1405 break; 1406 case Opt_coherency_buffered: 1407 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED; 1408 break; 1409 case Opt_coherency_full: 1410 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED; 1411 break; 1412 case Opt_acl: 1413 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1414 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL; 1415 break; 1416 case Opt_noacl: 1417 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL; 1418 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 1419 break; 1420 case Opt_resv_level: 1421 if (is_remount) 1422 break; 1423 if (match_int(&args[0], &option)) { 1424 status = 0; 1425 goto bail; 1426 } 1427 if (option >= OCFS2_MIN_RESV_LEVEL && 1428 option < OCFS2_MAX_RESV_LEVEL) 1429 mopt->resv_level = option; 1430 break; 1431 case Opt_dir_resv_level: 1432 if (is_remount) 1433 break; 1434 if (match_int(&args[0], &option)) { 1435 status = 0; 1436 goto bail; 1437 } 1438 if (option >= OCFS2_MIN_RESV_LEVEL && 1439 option < OCFS2_MAX_RESV_LEVEL) 1440 mopt->dir_resv_level = option; 1441 break; 1442 case Opt_journal_async_commit: 1443 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT; 1444 break; 1445 default: 1446 mlog(ML_ERROR, 1447 "Unrecognized mount option \"%s\" " 1448 "or missing value\n", p); 1449 status = 0; 1450 goto bail; 1451 } 1452 } 1453 1454 if (user_stack == 0) { 1455 /* Ensure only one heartbeat mode */ 1456 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | 1457 OCFS2_MOUNT_HB_GLOBAL | 1458 OCFS2_MOUNT_HB_NONE); 1459 if (hweight32(tmp) != 1) { 1460 mlog(ML_ERROR, "Invalid heartbeat mount options\n"); 1461 status = 0; 1462 goto bail; 1463 } 1464 } 1465 1466 status = 1; 1467 1468 bail: 1469 return status; 1470 } 1471 1472 static int ocfs2_show_options(struct seq_file *s, struct dentry *root) 1473 { 1474 struct ocfs2_super *osb = OCFS2_SB(root->d_sb); 1475 unsigned long opts = osb->s_mount_opt; 1476 unsigned int local_alloc_megs; 1477 1478 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) { 1479 seq_printf(s, ",_netdev"); 1480 if (opts & OCFS2_MOUNT_HB_LOCAL) 1481 seq_printf(s, ",%s", OCFS2_HB_LOCAL); 1482 else 1483 seq_printf(s, ",%s", OCFS2_HB_GLOBAL); 1484 } else 1485 seq_printf(s, ",%s", OCFS2_HB_NONE); 1486 1487 if (opts & OCFS2_MOUNT_NOINTR) 1488 seq_printf(s, ",nointr"); 1489 1490 if (opts & OCFS2_MOUNT_DATA_WRITEBACK) 1491 seq_printf(s, ",data=writeback"); 1492 else 1493 seq_printf(s, ",data=ordered"); 1494 1495 if (opts & OCFS2_MOUNT_BARRIER) 1496 seq_printf(s, ",barrier=1"); 1497 1498 if (opts & OCFS2_MOUNT_ERRORS_PANIC) 1499 seq_printf(s, ",errors=panic"); 1500 else if (opts & OCFS2_MOUNT_ERRORS_CONT) 1501 seq_printf(s, ",errors=continue"); 1502 else 1503 seq_printf(s, ",errors=remount-ro"); 1504 1505 if (osb->preferred_slot != OCFS2_INVALID_SLOT) 1506 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); 1507 1508 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); 1509 1510 if (osb->osb_commit_interval) 1511 seq_printf(s, ",commit=%u", 1512 (unsigned) (osb->osb_commit_interval / HZ)); 1513 1514 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits); 1515 if (local_alloc_megs != ocfs2_la_default_mb(osb)) 1516 seq_printf(s, ",localalloc=%d", local_alloc_megs); 1517 1518 if (opts & OCFS2_MOUNT_LOCALFLOCKS) 1519 seq_printf(s, ",localflocks,"); 1520 1521 if (osb->osb_cluster_stack[0]) 1522 seq_show_option(s, "cluster_stack", osb->osb_cluster_stack); 1523 if (opts & OCFS2_MOUNT_USRQUOTA) 1524 seq_printf(s, ",usrquota"); 1525 if (opts & OCFS2_MOUNT_GRPQUOTA) 1526 seq_printf(s, ",grpquota"); 1527 1528 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED) 1529 seq_printf(s, ",coherency=buffered"); 1530 else 1531 seq_printf(s, ",coherency=full"); 1532 1533 if (opts & OCFS2_MOUNT_NOUSERXATTR) 1534 seq_printf(s, ",nouser_xattr"); 1535 else 1536 seq_printf(s, ",user_xattr"); 1537 1538 if (opts & OCFS2_MOUNT_INODE64) 1539 seq_printf(s, ",inode64"); 1540 1541 if (opts & OCFS2_MOUNT_POSIX_ACL) 1542 seq_printf(s, ",acl"); 1543 else 1544 seq_printf(s, ",noacl"); 1545 1546 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL) 1547 seq_printf(s, ",resv_level=%d", osb->osb_resv_level); 1548 1549 if (osb->osb_dir_resv_level != osb->osb_resv_level) 1550 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level); 1551 1552 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT) 1553 seq_printf(s, ",journal_async_commit"); 1554 1555 return 0; 1556 } 1557 1558 static int __init ocfs2_init(void) 1559 { 1560 int status; 1561 1562 status = init_ocfs2_uptodate_cache(); 1563 if (status < 0) 1564 goto out1; 1565 1566 status = ocfs2_initialize_mem_caches(); 1567 if (status < 0) 1568 goto out2; 1569 1570 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); 1571 1572 ocfs2_set_locking_protocol(); 1573 1574 status = register_quota_format(&ocfs2_quota_format); 1575 if (status < 0) 1576 goto out3; 1577 status = register_filesystem(&ocfs2_fs_type); 1578 if (!status) 1579 return 0; 1580 1581 unregister_quota_format(&ocfs2_quota_format); 1582 out3: 1583 debugfs_remove(ocfs2_debugfs_root); 1584 ocfs2_free_mem_caches(); 1585 out2: 1586 exit_ocfs2_uptodate_cache(); 1587 out1: 1588 mlog_errno(status); 1589 return status; 1590 } 1591 1592 static void __exit ocfs2_exit(void) 1593 { 1594 unregister_quota_format(&ocfs2_quota_format); 1595 1596 debugfs_remove(ocfs2_debugfs_root); 1597 1598 ocfs2_free_mem_caches(); 1599 1600 unregister_filesystem(&ocfs2_fs_type); 1601 1602 exit_ocfs2_uptodate_cache(); 1603 } 1604 1605 static void ocfs2_put_super(struct super_block *sb) 1606 { 1607 trace_ocfs2_put_super(sb); 1608 1609 ocfs2_sync_blockdev(sb); 1610 ocfs2_dismount_volume(sb, 0); 1611 } 1612 1613 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) 1614 { 1615 struct ocfs2_super *osb; 1616 u32 numbits, freebits; 1617 int status; 1618 struct ocfs2_dinode *bm_lock; 1619 struct buffer_head *bh = NULL; 1620 struct inode *inode = NULL; 1621 1622 trace_ocfs2_statfs(dentry->d_sb, buf); 1623 1624 osb = OCFS2_SB(dentry->d_sb); 1625 1626 inode = ocfs2_get_system_file_inode(osb, 1627 GLOBAL_BITMAP_SYSTEM_INODE, 1628 OCFS2_INVALID_SLOT); 1629 if (!inode) { 1630 mlog(ML_ERROR, "failed to get bitmap inode\n"); 1631 status = -EIO; 1632 goto bail; 1633 } 1634 1635 status = ocfs2_inode_lock(inode, &bh, 0); 1636 if (status < 0) { 1637 mlog_errno(status); 1638 goto bail; 1639 } 1640 1641 bm_lock = (struct ocfs2_dinode *) bh->b_data; 1642 1643 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); 1644 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); 1645 1646 buf->f_type = OCFS2_SUPER_MAGIC; 1647 buf->f_bsize = dentry->d_sb->s_blocksize; 1648 buf->f_namelen = OCFS2_MAX_FILENAME_LEN; 1649 buf->f_blocks = ((sector_t) numbits) * 1650 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1651 buf->f_bfree = ((sector_t) freebits) * 1652 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1653 buf->f_bavail = buf->f_bfree; 1654 buf->f_files = numbits; 1655 buf->f_ffree = freebits; 1656 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN) 1657 & 0xFFFFFFFFUL; 1658 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN, 1659 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL; 1660 1661 brelse(bh); 1662 1663 ocfs2_inode_unlock(inode, 0); 1664 status = 0; 1665 bail: 1666 iput(inode); 1667 1668 if (status) 1669 mlog_errno(status); 1670 1671 return status; 1672 } 1673 1674 static void ocfs2_inode_init_once(void *data) 1675 { 1676 struct ocfs2_inode_info *oi = data; 1677 1678 oi->ip_flags = 0; 1679 oi->ip_open_count = 0; 1680 spin_lock_init(&oi->ip_lock); 1681 ocfs2_extent_map_init(&oi->vfs_inode); 1682 INIT_LIST_HEAD(&oi->ip_io_markers); 1683 INIT_LIST_HEAD(&oi->ip_unwritten_list); 1684 oi->ip_dir_start_lookup = 0; 1685 init_rwsem(&oi->ip_alloc_sem); 1686 init_rwsem(&oi->ip_xattr_sem); 1687 mutex_init(&oi->ip_io_mutex); 1688 1689 oi->ip_blkno = 0ULL; 1690 oi->ip_clusters = 0; 1691 oi->ip_next_orphan = NULL; 1692 1693 ocfs2_resv_init_once(&oi->ip_la_data_resv); 1694 1695 ocfs2_lock_res_init_once(&oi->ip_rw_lockres); 1696 ocfs2_lock_res_init_once(&oi->ip_inode_lockres); 1697 ocfs2_lock_res_init_once(&oi->ip_open_lockres); 1698 1699 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode), 1700 &ocfs2_inode_caching_ops); 1701 1702 inode_init_once(&oi->vfs_inode); 1703 } 1704 1705 static int ocfs2_initialize_mem_caches(void) 1706 { 1707 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", 1708 sizeof(struct ocfs2_inode_info), 1709 0, 1710 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1711 SLAB_MEM_SPREAD|SLAB_ACCOUNT), 1712 ocfs2_inode_init_once); 1713 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache", 1714 sizeof(struct ocfs2_dquot), 1715 0, 1716 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1717 SLAB_MEM_SPREAD), 1718 NULL); 1719 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache", 1720 sizeof(struct ocfs2_quota_chunk), 1721 0, 1722 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), 1723 NULL); 1724 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep || 1725 !ocfs2_qf_chunk_cachep) { 1726 kmem_cache_destroy(ocfs2_inode_cachep); 1727 kmem_cache_destroy(ocfs2_dquot_cachep); 1728 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1729 return -ENOMEM; 1730 } 1731 1732 return 0; 1733 } 1734 1735 static void ocfs2_free_mem_caches(void) 1736 { 1737 /* 1738 * Make sure all delayed rcu free inodes are flushed before we 1739 * destroy cache. 1740 */ 1741 rcu_barrier(); 1742 kmem_cache_destroy(ocfs2_inode_cachep); 1743 ocfs2_inode_cachep = NULL; 1744 1745 kmem_cache_destroy(ocfs2_dquot_cachep); 1746 ocfs2_dquot_cachep = NULL; 1747 1748 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1749 ocfs2_qf_chunk_cachep = NULL; 1750 } 1751 1752 static int ocfs2_get_sector(struct super_block *sb, 1753 struct buffer_head **bh, 1754 int block, 1755 int sect_size) 1756 { 1757 if (!sb_set_blocksize(sb, sect_size)) { 1758 mlog(ML_ERROR, "unable to set blocksize\n"); 1759 return -EIO; 1760 } 1761 1762 *bh = sb_getblk(sb, block); 1763 if (!*bh) { 1764 mlog_errno(-ENOMEM); 1765 return -ENOMEM; 1766 } 1767 lock_buffer(*bh); 1768 if (!buffer_dirty(*bh)) 1769 clear_buffer_uptodate(*bh); 1770 unlock_buffer(*bh); 1771 if (bh_read(*bh, 0) < 0) { 1772 mlog_errno(-EIO); 1773 brelse(*bh); 1774 *bh = NULL; 1775 return -EIO; 1776 } 1777 1778 return 0; 1779 } 1780 1781 static int ocfs2_mount_volume(struct super_block *sb) 1782 { 1783 int status = 0; 1784 struct ocfs2_super *osb = OCFS2_SB(sb); 1785 1786 if (ocfs2_is_hard_readonly(osb)) 1787 goto out; 1788 1789 mutex_init(&osb->obs_trim_fs_mutex); 1790 1791 status = ocfs2_dlm_init(osb); 1792 if (status < 0) { 1793 mlog_errno(status); 1794 if (status == -EBADR && ocfs2_userspace_stack(osb)) 1795 mlog(ML_ERROR, "couldn't mount because cluster name on" 1796 " disk does not match the running cluster name.\n"); 1797 goto out; 1798 } 1799 1800 status = ocfs2_super_lock(osb, 1); 1801 if (status < 0) { 1802 mlog_errno(status); 1803 goto out_dlm; 1804 } 1805 1806 /* This will load up the node map and add ourselves to it. */ 1807 status = ocfs2_find_slot(osb); 1808 if (status < 0) { 1809 mlog_errno(status); 1810 goto out_super_lock; 1811 } 1812 1813 /* load all node-local system inodes */ 1814 status = ocfs2_init_local_system_inodes(osb); 1815 if (status < 0) { 1816 mlog_errno(status); 1817 goto out_super_lock; 1818 } 1819 1820 status = ocfs2_check_volume(osb); 1821 if (status < 0) { 1822 mlog_errno(status); 1823 goto out_system_inodes; 1824 } 1825 1826 status = ocfs2_truncate_log_init(osb); 1827 if (status < 0) { 1828 mlog_errno(status); 1829 goto out_check_volume; 1830 } 1831 1832 ocfs2_super_unlock(osb, 1); 1833 return 0; 1834 1835 out_check_volume: 1836 ocfs2_free_replay_slots(osb); 1837 out_system_inodes: 1838 if (osb->local_alloc_state == OCFS2_LA_ENABLED) 1839 ocfs2_shutdown_local_alloc(osb); 1840 ocfs2_release_system_inodes(osb); 1841 /* before journal shutdown, we should release slot_info */ 1842 ocfs2_free_slot_info(osb); 1843 ocfs2_journal_shutdown(osb); 1844 out_super_lock: 1845 ocfs2_super_unlock(osb, 1); 1846 out_dlm: 1847 ocfs2_dlm_shutdown(osb, 0); 1848 out: 1849 return status; 1850 } 1851 1852 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) 1853 { 1854 int tmp, hangup_needed = 0; 1855 struct ocfs2_super *osb = NULL; 1856 char nodestr[12]; 1857 1858 trace_ocfs2_dismount_volume(sb); 1859 1860 BUG_ON(!sb); 1861 osb = OCFS2_SB(sb); 1862 BUG_ON(!osb); 1863 1864 /* Remove file check sysfs related directores/files, 1865 * and wait for the pending file check operations */ 1866 ocfs2_filecheck_remove_sysfs(osb); 1867 1868 kset_unregister(osb->osb_dev_kset); 1869 1870 /* Orphan scan should be stopped as early as possible */ 1871 ocfs2_orphan_scan_stop(osb); 1872 1873 ocfs2_disable_quotas(osb); 1874 1875 /* All dquots should be freed by now */ 1876 WARN_ON(!llist_empty(&osb->dquot_drop_list)); 1877 /* Wait for worker to be done with the work structure in osb */ 1878 cancel_work_sync(&osb->dquot_drop_work); 1879 1880 ocfs2_shutdown_local_alloc(osb); 1881 1882 ocfs2_truncate_log_shutdown(osb); 1883 1884 /* This will disable recovery and flush any recovery work. */ 1885 ocfs2_recovery_exit(osb); 1886 1887 ocfs2_sync_blockdev(sb); 1888 1889 ocfs2_purge_refcount_trees(osb); 1890 1891 /* No cluster connection means we've failed during mount, so skip 1892 * all the steps which depended on that to complete. */ 1893 if (osb->cconn) { 1894 tmp = ocfs2_super_lock(osb, 1); 1895 if (tmp < 0) { 1896 mlog_errno(tmp); 1897 return; 1898 } 1899 } 1900 1901 if (osb->slot_num != OCFS2_INVALID_SLOT) 1902 ocfs2_put_slot(osb); 1903 1904 if (osb->cconn) 1905 ocfs2_super_unlock(osb, 1); 1906 1907 ocfs2_release_system_inodes(osb); 1908 1909 ocfs2_journal_shutdown(osb); 1910 1911 /* 1912 * If we're dismounting due to mount error, mount.ocfs2 will clean 1913 * up heartbeat. If we're a local mount, there is no heartbeat. 1914 * If we failed before we got a uuid_str yet, we can't stop 1915 * heartbeat. Otherwise, do it. 1916 */ 1917 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str && 1918 !ocfs2_is_hard_readonly(osb)) 1919 hangup_needed = 1; 1920 1921 ocfs2_dlm_shutdown(osb, hangup_needed); 1922 1923 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats); 1924 debugfs_remove_recursive(osb->osb_debug_root); 1925 1926 if (hangup_needed) 1927 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str)); 1928 1929 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); 1930 1931 if (ocfs2_mount_local(osb)) 1932 snprintf(nodestr, sizeof(nodestr), "local"); 1933 else 1934 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1935 1936 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", 1937 osb->dev_str, nodestr); 1938 1939 ocfs2_delete_osb(osb); 1940 kfree(osb); 1941 sb->s_dev = 0; 1942 sb->s_fs_info = NULL; 1943 } 1944 1945 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, 1946 unsigned uuid_bytes) 1947 { 1948 int i, ret; 1949 char *ptr; 1950 1951 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); 1952 1953 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); 1954 if (osb->uuid_str == NULL) 1955 return -ENOMEM; 1956 1957 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { 1958 /* print with null */ 1959 ret = snprintf(ptr, 3, "%02X", uuid[i]); 1960 if (ret != 2) /* drop super cleans up */ 1961 return -EINVAL; 1962 /* then only advance past the last char */ 1963 ptr += 2; 1964 } 1965 1966 return 0; 1967 } 1968 1969 /* Make sure entire volume is addressable by our journal. Requires 1970 osb_clusters_at_boot to be valid and for the journal to have been 1971 initialized by ocfs2_journal_init(). */ 1972 static int ocfs2_journal_addressable(struct ocfs2_super *osb) 1973 { 1974 int status = 0; 1975 u64 max_block = 1976 ocfs2_clusters_to_blocks(osb->sb, 1977 osb->osb_clusters_at_boot) - 1; 1978 1979 /* 32-bit block number is always OK. */ 1980 if (max_block <= (u32)~0ULL) 1981 goto out; 1982 1983 /* Volume is "huge", so see if our journal is new enough to 1984 support it. */ 1985 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb, 1986 OCFS2_FEATURE_COMPAT_JBD2_SB) && 1987 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0, 1988 JBD2_FEATURE_INCOMPAT_64BIT))) { 1989 mlog(ML_ERROR, "The journal cannot address the entire volume. " 1990 "Enable the 'block64' journal option with tunefs.ocfs2"); 1991 status = -EFBIG; 1992 goto out; 1993 } 1994 1995 out: 1996 return status; 1997 } 1998 1999 static int ocfs2_initialize_super(struct super_block *sb, 2000 struct buffer_head *bh, 2001 int sector_size, 2002 struct ocfs2_blockcheck_stats *stats) 2003 { 2004 int status; 2005 int i, cbits, bbits; 2006 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; 2007 struct inode *inode = NULL; 2008 struct ocfs2_super *osb; 2009 u64 total_blocks; 2010 2011 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); 2012 if (!osb) { 2013 status = -ENOMEM; 2014 mlog_errno(status); 2015 goto out; 2016 } 2017 2018 sb->s_fs_info = osb; 2019 sb->s_op = &ocfs2_sops; 2020 sb->s_d_op = &ocfs2_dentry_ops; 2021 sb->s_export_op = &ocfs2_export_ops; 2022 sb->s_qcop = &dquot_quotactl_sysfile_ops; 2023 sb->dq_op = &ocfs2_quota_operations; 2024 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP; 2025 sb->s_xattr = ocfs2_xattr_handlers; 2026 sb->s_time_gran = 1; 2027 sb->s_flags |= SB_NOATIME; 2028 /* this is needed to support O_LARGEFILE */ 2029 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2030 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); 2031 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); 2032 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid, 2033 sizeof(di->id2.i_super.s_uuid)); 2034 2035 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1; 2036 2037 for (i = 0; i < 3; i++) 2038 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]); 2039 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2040 2041 osb->sb = sb; 2042 osb->s_sectsize_bits = blksize_bits(sector_size); 2043 BUG_ON(!osb->s_sectsize_bits); 2044 2045 spin_lock_init(&osb->dc_task_lock); 2046 init_waitqueue_head(&osb->dc_event); 2047 osb->dc_work_sequence = 0; 2048 osb->dc_wake_sequence = 0; 2049 INIT_LIST_HEAD(&osb->blocked_lock_list); 2050 osb->blocked_lock_count = 0; 2051 spin_lock_init(&osb->osb_lock); 2052 spin_lock_init(&osb->osb_xattr_lock); 2053 ocfs2_init_steal_slots(osb); 2054 2055 mutex_init(&osb->system_file_mutex); 2056 2057 atomic_set(&osb->alloc_stats.moves, 0); 2058 atomic_set(&osb->alloc_stats.local_data, 0); 2059 atomic_set(&osb->alloc_stats.bitmap_data, 0); 2060 atomic_set(&osb->alloc_stats.bg_allocs, 0); 2061 atomic_set(&osb->alloc_stats.bg_extends, 0); 2062 2063 /* Copy the blockcheck stats from the superblock probe */ 2064 osb->osb_ecc_stats = *stats; 2065 2066 ocfs2_init_node_maps(osb); 2067 2068 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", 2069 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); 2070 2071 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); 2072 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { 2073 mlog(ML_ERROR, "Invalid number of node slots (%u)\n", 2074 osb->max_slots); 2075 status = -EINVAL; 2076 goto out; 2077 } 2078 2079 ocfs2_orphan_scan_init(osb); 2080 2081 status = ocfs2_recovery_init(osb); 2082 if (status) { 2083 mlog(ML_ERROR, "Unable to initialize recovery state\n"); 2084 mlog_errno(status); 2085 goto out; 2086 } 2087 2088 init_waitqueue_head(&osb->checkpoint_event); 2089 2090 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 2091 2092 osb->slot_num = OCFS2_INVALID_SLOT; 2093 2094 osb->s_xattr_inline_size = le16_to_cpu( 2095 di->id2.i_super.s_xattr_inline_size); 2096 2097 osb->local_alloc_state = OCFS2_LA_UNUSED; 2098 osb->local_alloc_bh = NULL; 2099 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker); 2100 2101 init_waitqueue_head(&osb->osb_mount_event); 2102 2103 ocfs2_resmap_init(osb, &osb->osb_la_resmap); 2104 2105 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); 2106 if (!osb->vol_label) { 2107 mlog(ML_ERROR, "unable to alloc vol label\n"); 2108 status = -ENOMEM; 2109 goto out_recovery_map; 2110 } 2111 2112 osb->slot_recovery_generations = 2113 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), 2114 GFP_KERNEL); 2115 if (!osb->slot_recovery_generations) { 2116 status = -ENOMEM; 2117 mlog_errno(status); 2118 goto out_vol_label; 2119 } 2120 2121 init_waitqueue_head(&osb->osb_wipe_event); 2122 osb->osb_orphan_wipes = kcalloc(osb->max_slots, 2123 sizeof(*osb->osb_orphan_wipes), 2124 GFP_KERNEL); 2125 if (!osb->osb_orphan_wipes) { 2126 status = -ENOMEM; 2127 mlog_errno(status); 2128 goto out_slot_recovery_gen; 2129 } 2130 2131 osb->osb_rf_lock_tree = RB_ROOT; 2132 2133 osb->s_feature_compat = 2134 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); 2135 osb->s_feature_ro_compat = 2136 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); 2137 osb->s_feature_incompat = 2138 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); 2139 2140 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { 2141 mlog(ML_ERROR, "couldn't mount because of unsupported " 2142 "optional features (%x).\n", i); 2143 status = -EINVAL; 2144 goto out_orphan_wipes; 2145 } 2146 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { 2147 mlog(ML_ERROR, "couldn't mount RDWR because of " 2148 "unsupported optional features (%x).\n", i); 2149 status = -EINVAL; 2150 goto out_orphan_wipes; 2151 } 2152 2153 if (ocfs2_clusterinfo_valid(osb)) { 2154 /* 2155 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null 2156 * terminated, so make sure no overflow happens here by using 2157 * memcpy. Destination strings will always be null terminated 2158 * because osb is allocated using kzalloc. 2159 */ 2160 osb->osb_stackflags = 2161 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags; 2162 memcpy(osb->osb_cluster_stack, 2163 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, 2164 OCFS2_STACK_LABEL_LEN); 2165 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) { 2166 mlog(ML_ERROR, 2167 "couldn't mount because of an invalid " 2168 "cluster stack label (%s) \n", 2169 osb->osb_cluster_stack); 2170 status = -EINVAL; 2171 goto out_orphan_wipes; 2172 } 2173 memcpy(osb->osb_cluster_name, 2174 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster, 2175 OCFS2_CLUSTER_NAME_LEN); 2176 } else { 2177 /* The empty string is identical with classic tools that 2178 * don't know about s_cluster_info. */ 2179 osb->osb_cluster_stack[0] = '\0'; 2180 } 2181 2182 get_random_bytes(&osb->s_next_generation, sizeof(u32)); 2183 2184 /* 2185 * FIXME 2186 * This should be done in ocfs2_journal_init(), but any inode 2187 * writes back operation will cause the filesystem to crash. 2188 */ 2189 status = ocfs2_journal_alloc(osb); 2190 if (status < 0) 2191 goto out_orphan_wipes; 2192 2193 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs); 2194 init_llist_head(&osb->dquot_drop_list); 2195 2196 /* get some pseudo constants for clustersize bits */ 2197 osb->s_clustersize_bits = 2198 le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2199 osb->s_clustersize = 1 << osb->s_clustersize_bits; 2200 2201 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || 2202 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { 2203 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", 2204 osb->s_clustersize); 2205 status = -EINVAL; 2206 goto out_journal; 2207 } 2208 2209 total_blocks = ocfs2_clusters_to_blocks(osb->sb, 2210 le32_to_cpu(di->i_clusters)); 2211 2212 status = generic_check_addressable(osb->sb->s_blocksize_bits, 2213 total_blocks); 2214 if (status) { 2215 mlog(ML_ERROR, "Volume too large " 2216 "to mount safely on this system"); 2217 status = -EFBIG; 2218 goto out_journal; 2219 } 2220 2221 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, 2222 sizeof(di->id2.i_super.s_uuid))) { 2223 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); 2224 status = -ENOMEM; 2225 goto out_journal; 2226 } 2227 2228 strscpy(osb->vol_label, di->id2.i_super.s_label, 2229 OCFS2_MAX_VOL_LABEL_LEN); 2230 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); 2231 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); 2232 osb->first_cluster_group_blkno = 2233 le64_to_cpu(di->id2.i_super.s_first_cluster_group); 2234 osb->fs_generation = le32_to_cpu(di->i_fs_generation); 2235 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2236 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str, 2237 (unsigned long long)osb->root_blkno, 2238 (unsigned long long)osb->system_dir_blkno, 2239 osb->s_clustersize_bits); 2240 2241 osb->osb_dlm_debug = ocfs2_new_dlm_debug(); 2242 if (!osb->osb_dlm_debug) { 2243 status = -ENOMEM; 2244 mlog_errno(status); 2245 goto out_uuid_str; 2246 } 2247 2248 atomic_set(&osb->vol_state, VOLUME_INIT); 2249 2250 /* load root, system_dir, and all global system inodes */ 2251 status = ocfs2_init_global_system_inodes(osb); 2252 if (status < 0) { 2253 mlog_errno(status); 2254 goto out_dlm_out; 2255 } 2256 2257 /* 2258 * global bitmap 2259 */ 2260 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, 2261 OCFS2_INVALID_SLOT); 2262 if (!inode) { 2263 status = -EINVAL; 2264 mlog_errno(status); 2265 goto out_system_inodes; 2266 } 2267 2268 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; 2269 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters; 2270 iput(inode); 2271 2272 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0, 2273 osb->s_feature_incompat) * 8; 2274 2275 status = ocfs2_init_slot_info(osb); 2276 if (status < 0) { 2277 mlog_errno(status); 2278 goto out_system_inodes; 2279 } 2280 2281 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM); 2282 if (!osb->ocfs2_wq) { 2283 status = -ENOMEM; 2284 mlog_errno(status); 2285 goto out_slot_info; 2286 } 2287 2288 return status; 2289 2290 out_slot_info: 2291 ocfs2_free_slot_info(osb); 2292 out_system_inodes: 2293 ocfs2_release_system_inodes(osb); 2294 out_dlm_out: 2295 ocfs2_put_dlm_debug(osb->osb_dlm_debug); 2296 out_uuid_str: 2297 kfree(osb->uuid_str); 2298 out_journal: 2299 kfree(osb->journal); 2300 out_orphan_wipes: 2301 kfree(osb->osb_orphan_wipes); 2302 out_slot_recovery_gen: 2303 kfree(osb->slot_recovery_generations); 2304 out_vol_label: 2305 kfree(osb->vol_label); 2306 out_recovery_map: 2307 kfree(osb->recovery_map); 2308 out: 2309 kfree(osb); 2310 sb->s_fs_info = NULL; 2311 return status; 2312 } 2313 2314 /* 2315 * will return: -EAGAIN if it is ok to keep searching for superblocks 2316 * -EINVAL if there is a bad superblock 2317 * 0 on success 2318 */ 2319 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 2320 struct buffer_head *bh, 2321 u32 blksz, 2322 struct ocfs2_blockcheck_stats *stats) 2323 { 2324 int status = -EAGAIN; 2325 u32 blksz_bits; 2326 2327 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, 2328 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { 2329 /* We have to do a raw check of the feature here */ 2330 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) & 2331 OCFS2_FEATURE_INCOMPAT_META_ECC) { 2332 status = ocfs2_block_check_validate(bh->b_data, 2333 bh->b_size, 2334 &di->i_check, 2335 stats); 2336 if (status) 2337 goto out; 2338 } 2339 status = -EINVAL; 2340 /* Acceptable block sizes are 512 bytes, 1K, 2K and 4K. */ 2341 blksz_bits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); 2342 if (blksz_bits < 9 || blksz_bits > 12) { 2343 mlog(ML_ERROR, "found superblock with incorrect block " 2344 "size bits: found %u, should be 9, 10, 11, or 12\n", 2345 blksz_bits); 2346 } else if ((1 << le32_to_cpu(blksz_bits)) != blksz) { 2347 mlog(ML_ERROR, "found superblock with incorrect block " 2348 "size: found %u, should be %u\n", 1 << blksz_bits, blksz); 2349 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != 2350 OCFS2_MAJOR_REV_LEVEL || 2351 le16_to_cpu(di->id2.i_super.s_minor_rev_level) != 2352 OCFS2_MINOR_REV_LEVEL) { 2353 mlog(ML_ERROR, "found superblock with bad version: " 2354 "found %u.%u, should be %u.%u\n", 2355 le16_to_cpu(di->id2.i_super.s_major_rev_level), 2356 le16_to_cpu(di->id2.i_super.s_minor_rev_level), 2357 OCFS2_MAJOR_REV_LEVEL, 2358 OCFS2_MINOR_REV_LEVEL); 2359 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { 2360 mlog(ML_ERROR, "bad block number on superblock: " 2361 "found %llu, should be %llu\n", 2362 (unsigned long long)le64_to_cpu(di->i_blkno), 2363 (unsigned long long)bh->b_blocknr); 2364 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || 2365 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { 2366 mlog(ML_ERROR, "bad cluster size found: %u\n", 2367 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); 2368 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { 2369 mlog(ML_ERROR, "bad root_blkno: 0\n"); 2370 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { 2371 mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); 2372 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { 2373 mlog(ML_ERROR, 2374 "Superblock slots found greater than file system " 2375 "maximum: found %u, max %u\n", 2376 le16_to_cpu(di->id2.i_super.s_max_slots), 2377 OCFS2_MAX_SLOTS); 2378 } else { 2379 /* found it! */ 2380 status = 0; 2381 } 2382 } 2383 2384 out: 2385 if (status && status != -EAGAIN) 2386 mlog_errno(status); 2387 return status; 2388 } 2389 2390 static int ocfs2_check_volume(struct ocfs2_super *osb) 2391 { 2392 int status; 2393 int dirty; 2394 int local; 2395 struct ocfs2_dinode *local_alloc = NULL; /* only used if we 2396 * recover 2397 * ourselves. */ 2398 2399 /* Init our journal object. */ 2400 status = ocfs2_journal_init(osb, &dirty); 2401 if (status < 0) { 2402 mlog(ML_ERROR, "Could not initialize journal!\n"); 2403 goto finally; 2404 } 2405 2406 /* Now that journal has been initialized, check to make sure 2407 entire volume is addressable. */ 2408 status = ocfs2_journal_addressable(osb); 2409 if (status) 2410 goto finally; 2411 2412 /* If the journal was unmounted cleanly then we don't want to 2413 * recover anything. Otherwise, journal_load will do that 2414 * dirty work for us :) */ 2415 if (!dirty) { 2416 status = ocfs2_journal_wipe(osb->journal, 0); 2417 if (status < 0) { 2418 mlog_errno(status); 2419 goto finally; 2420 } 2421 } else { 2422 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not " 2423 "unmounted cleanly, recovering it.\n", osb->dev_str); 2424 } 2425 2426 local = ocfs2_mount_local(osb); 2427 2428 /* will play back anything left in the journal. */ 2429 status = ocfs2_journal_load(osb->journal, local, dirty); 2430 if (status < 0) { 2431 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status); 2432 goto finally; 2433 } 2434 2435 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT) 2436 jbd2_journal_set_features(osb->journal->j_journal, 2437 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2438 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 2439 else 2440 jbd2_journal_clear_features(osb->journal->j_journal, 2441 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 2442 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 2443 2444 if (dirty) { 2445 /* recover my local alloc if we didn't unmount cleanly. */ 2446 status = ocfs2_begin_local_alloc_recovery(osb, 2447 osb->slot_num, 2448 &local_alloc); 2449 if (status < 0) { 2450 mlog_errno(status); 2451 goto finally; 2452 } 2453 /* we complete the recovery process after we've marked 2454 * ourselves as mounted. */ 2455 } 2456 2457 status = ocfs2_load_local_alloc(osb); 2458 if (status < 0) { 2459 mlog_errno(status); 2460 goto finally; 2461 } 2462 2463 if (dirty) { 2464 /* Recovery will be completed after we've mounted the 2465 * rest of the volume. */ 2466 osb->local_alloc_copy = local_alloc; 2467 local_alloc = NULL; 2468 } 2469 2470 /* go through each journal, trylock it and if you get the 2471 * lock, and it's marked as dirty, set the bit in the recover 2472 * map and launch a recovery thread for it. */ 2473 status = ocfs2_mark_dead_nodes(osb); 2474 if (status < 0) { 2475 mlog_errno(status); 2476 goto finally; 2477 } 2478 2479 status = ocfs2_compute_replay_slots(osb); 2480 if (status < 0) 2481 mlog_errno(status); 2482 2483 finally: 2484 kfree(local_alloc); 2485 2486 if (status) 2487 mlog_errno(status); 2488 return status; 2489 } 2490 2491 /* 2492 * The routine gets called from dismount or close whenever a dismount on 2493 * volume is requested and the osb open count becomes 1. 2494 * It will remove the osb from the global list and also free up all the 2495 * initialized resources and fileobject. 2496 */ 2497 static void ocfs2_delete_osb(struct ocfs2_super *osb) 2498 { 2499 /* This function assumes that the caller has the main osb resource */ 2500 2501 /* ocfs2_initializer_super have already created this workqueue */ 2502 if (osb->ocfs2_wq) 2503 destroy_workqueue(osb->ocfs2_wq); 2504 2505 ocfs2_free_slot_info(osb); 2506 2507 kfree(osb->osb_orphan_wipes); 2508 kfree(osb->slot_recovery_generations); 2509 /* FIXME 2510 * This belongs in journal shutdown, but because we have to 2511 * allocate osb->journal at the middle of ocfs2_initialize_super(), 2512 * we free it here. 2513 */ 2514 kfree(osb->journal); 2515 kfree(osb->local_alloc_copy); 2516 kfree(osb->uuid_str); 2517 kfree(osb->vol_label); 2518 ocfs2_put_dlm_debug(osb->osb_dlm_debug); 2519 memset(osb, 0, sizeof(struct ocfs2_super)); 2520 } 2521 2522 /* Depending on the mount option passed, perform one of the following: 2523 * Put OCFS2 into a readonly state (default) 2524 * Return EIO so that only the process errs 2525 * Fix the error as if fsck.ocfs2 -y 2526 * panic 2527 */ 2528 static int ocfs2_handle_error(struct super_block *sb) 2529 { 2530 struct ocfs2_super *osb = OCFS2_SB(sb); 2531 int rv = 0; 2532 2533 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); 2534 pr_crit("On-disk corruption discovered. " 2535 "Please run fsck.ocfs2 once the filesystem is unmounted.\n"); 2536 2537 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) { 2538 panic("OCFS2: (device %s): panic forced after error\n", 2539 sb->s_id); 2540 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) { 2541 pr_crit("OCFS2: Returning error to the calling process.\n"); 2542 rv = -EIO; 2543 } else { /* default option */ 2544 rv = -EROFS; 2545 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb))) 2546 return rv; 2547 2548 pr_crit("OCFS2: File system is now read-only.\n"); 2549 sb->s_flags |= SB_RDONLY; 2550 ocfs2_set_ro_flag(osb, 0); 2551 } 2552 2553 return rv; 2554 } 2555 2556 int __ocfs2_error(struct super_block *sb, const char *function, 2557 const char *fmt, ...) 2558 { 2559 struct va_format vaf; 2560 va_list args; 2561 2562 va_start(args, fmt); 2563 vaf.fmt = fmt; 2564 vaf.va = &args; 2565 2566 /* Not using mlog here because we want to show the actual 2567 * function the error came from. */ 2568 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV", 2569 sb->s_id, function, &vaf); 2570 2571 va_end(args); 2572 2573 return ocfs2_handle_error(sb); 2574 } 2575 2576 /* Handle critical errors. This is intentionally more drastic than 2577 * ocfs2_handle_error, so we only use for things like journal errors, 2578 * etc. */ 2579 void __ocfs2_abort(struct super_block *sb, const char *function, 2580 const char *fmt, ...) 2581 { 2582 struct va_format vaf; 2583 va_list args; 2584 2585 va_start(args, fmt); 2586 2587 vaf.fmt = fmt; 2588 vaf.va = &args; 2589 2590 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV", 2591 sb->s_id, function, &vaf); 2592 2593 va_end(args); 2594 2595 /* We don't have the cluster support yet to go straight to 2596 * hard readonly in here. Until then, we want to keep 2597 * ocfs2_abort() so that we can at least mark critical 2598 * errors. 2599 * 2600 * TODO: This should abort the journal and alert other nodes 2601 * that our slot needs recovery. */ 2602 2603 /* Force a panic(). This stinks, but it's better than letting 2604 * things continue without having a proper hard readonly 2605 * here. */ 2606 if (!ocfs2_mount_local(OCFS2_SB(sb))) 2607 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 2608 ocfs2_handle_error(sb); 2609 } 2610 2611 /* 2612 * Void signal blockers, because in-kernel sigprocmask() only fails 2613 * when SIG_* is wrong. 2614 */ 2615 void ocfs2_block_signals(sigset_t *oldset) 2616 { 2617 int rc; 2618 sigset_t blocked; 2619 2620 sigfillset(&blocked); 2621 rc = sigprocmask(SIG_BLOCK, &blocked, oldset); 2622 BUG_ON(rc); 2623 } 2624 2625 void ocfs2_unblock_signals(sigset_t *oldset) 2626 { 2627 int rc = sigprocmask(SIG_SETMASK, oldset, NULL); 2628 BUG_ON(rc); 2629 } 2630 2631 module_init(ocfs2_init); 2632 module_exit(ocfs2_exit); 2633