1 /* -*- mode: c; c-basic-offset: 8; -*- 2 * vim: noexpandtab sw=8 ts=8 sts=0: 3 * 4 * super.c 5 * 6 * load/unload driver, mount/dismount volumes 7 * 8 * Copyright (C) 2002, 2004 Oracle. All rights reserved. 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public 12 * License as published by the Free Software Foundation; either 13 * version 2 of the License, or (at your option) any later version. 14 * 15 * This program is distributed in the hope that it will be useful, 16 * but WITHOUT ANY WARRANTY; without even the implied warranty of 17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 18 * General Public License for more details. 19 * 20 * You should have received a copy of the GNU General Public 21 * License along with this program; if not, write to the 22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 23 * Boston, MA 021110-1307, USA. 24 */ 25 26 #include <linux/module.h> 27 #include <linux/fs.h> 28 #include <linux/types.h> 29 #include <linux/slab.h> 30 #include <linux/highmem.h> 31 #include <linux/init.h> 32 #include <linux/random.h> 33 #include <linux/statfs.h> 34 #include <linux/moduleparam.h> 35 #include <linux/blkdev.h> 36 #include <linux/socket.h> 37 #include <linux/inet.h> 38 #include <linux/parser.h> 39 #include <linux/crc32.h> 40 #include <linux/debugfs.h> 41 #include <linux/mount.h> 42 #include <linux/seq_file.h> 43 #include <linux/quotaops.h> 44 #include <linux/smp_lock.h> 45 46 #define MLOG_MASK_PREFIX ML_SUPER 47 #include <cluster/masklog.h> 48 49 #include "ocfs2.h" 50 51 /* this should be the only file to include a version 1 header */ 52 #include "ocfs1_fs_compat.h" 53 54 #include "alloc.h" 55 #include "blockcheck.h" 56 #include "dlmglue.h" 57 #include "export.h" 58 #include "extent_map.h" 59 #include "heartbeat.h" 60 #include "inode.h" 61 #include "journal.h" 62 #include "localalloc.h" 63 #include "namei.h" 64 #include "slot_map.h" 65 #include "super.h" 66 #include "sysfile.h" 67 #include "uptodate.h" 68 #include "ver.h" 69 #include "xattr.h" 70 #include "quota.h" 71 #include "refcounttree.h" 72 73 #include "buffer_head_io.h" 74 75 static struct kmem_cache *ocfs2_inode_cachep = NULL; 76 struct kmem_cache *ocfs2_dquot_cachep; 77 struct kmem_cache *ocfs2_qf_chunk_cachep; 78 79 /* OCFS2 needs to schedule several differnt types of work which 80 * require cluster locking, disk I/O, recovery waits, etc. Since these 81 * types of work tend to be heavy we avoid using the kernel events 82 * workqueue and schedule on our own. */ 83 struct workqueue_struct *ocfs2_wq = NULL; 84 85 static struct dentry *ocfs2_debugfs_root = NULL; 86 87 MODULE_AUTHOR("Oracle"); 88 MODULE_LICENSE("GPL"); 89 90 struct mount_options 91 { 92 unsigned long commit_interval; 93 unsigned long mount_opt; 94 unsigned int atime_quantum; 95 signed short slot; 96 unsigned int localalloc_opt; 97 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; 98 }; 99 100 static int ocfs2_parse_options(struct super_block *sb, char *options, 101 struct mount_options *mopt, 102 int is_remount); 103 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt); 104 static void ocfs2_put_super(struct super_block *sb); 105 static int ocfs2_mount_volume(struct super_block *sb); 106 static int ocfs2_remount(struct super_block *sb, int *flags, char *data); 107 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err); 108 static int ocfs2_initialize_mem_caches(void); 109 static void ocfs2_free_mem_caches(void); 110 static void ocfs2_delete_osb(struct ocfs2_super *osb); 111 112 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf); 113 114 static int ocfs2_sync_fs(struct super_block *sb, int wait); 115 116 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb); 117 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb); 118 static void ocfs2_release_system_inodes(struct ocfs2_super *osb); 119 static int ocfs2_check_volume(struct ocfs2_super *osb); 120 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 121 struct buffer_head *bh, 122 u32 sectsize, 123 struct ocfs2_blockcheck_stats *stats); 124 static int ocfs2_initialize_super(struct super_block *sb, 125 struct buffer_head *bh, 126 int sector_size, 127 struct ocfs2_blockcheck_stats *stats); 128 static int ocfs2_get_sector(struct super_block *sb, 129 struct buffer_head **bh, 130 int block, 131 int sect_size); 132 static struct inode *ocfs2_alloc_inode(struct super_block *sb); 133 static void ocfs2_destroy_inode(struct inode *inode); 134 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend); 135 static int ocfs2_enable_quotas(struct ocfs2_super *osb); 136 static void ocfs2_disable_quotas(struct ocfs2_super *osb); 137 138 static const struct super_operations ocfs2_sops = { 139 .statfs = ocfs2_statfs, 140 .alloc_inode = ocfs2_alloc_inode, 141 .destroy_inode = ocfs2_destroy_inode, 142 .drop_inode = ocfs2_drop_inode, 143 .clear_inode = ocfs2_clear_inode, 144 .delete_inode = ocfs2_delete_inode, 145 .sync_fs = ocfs2_sync_fs, 146 .put_super = ocfs2_put_super, 147 .remount_fs = ocfs2_remount, 148 .show_options = ocfs2_show_options, 149 .quota_read = ocfs2_quota_read, 150 .quota_write = ocfs2_quota_write, 151 }; 152 153 enum { 154 Opt_barrier, 155 Opt_err_panic, 156 Opt_err_ro, 157 Opt_intr, 158 Opt_nointr, 159 Opt_hb_none, 160 Opt_hb_local, 161 Opt_data_ordered, 162 Opt_data_writeback, 163 Opt_atime_quantum, 164 Opt_slot, 165 Opt_commit, 166 Opt_localalloc, 167 Opt_localflocks, 168 Opt_stack, 169 Opt_user_xattr, 170 Opt_nouser_xattr, 171 Opt_inode64, 172 Opt_acl, 173 Opt_noacl, 174 Opt_usrquota, 175 Opt_grpquota, 176 Opt_err, 177 }; 178 179 static const match_table_t tokens = { 180 {Opt_barrier, "barrier=%u"}, 181 {Opt_err_panic, "errors=panic"}, 182 {Opt_err_ro, "errors=remount-ro"}, 183 {Opt_intr, "intr"}, 184 {Opt_nointr, "nointr"}, 185 {Opt_hb_none, OCFS2_HB_NONE}, 186 {Opt_hb_local, OCFS2_HB_LOCAL}, 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_err, NULL} 203 }; 204 205 #ifdef CONFIG_DEBUG_FS 206 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len) 207 { 208 struct ocfs2_cluster_connection *cconn = osb->cconn; 209 struct ocfs2_recovery_map *rm = osb->recovery_map; 210 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan; 211 int i, out = 0; 212 213 out += snprintf(buf + out, len - out, 214 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n", 215 "Device", osb->dev_str, osb->uuid_str, 216 osb->fs_generation, osb->vol_label); 217 218 out += snprintf(buf + out, len - out, 219 "%10s => State: %d Flags: 0x%lX\n", "Volume", 220 atomic_read(&osb->vol_state), osb->osb_flags); 221 222 out += snprintf(buf + out, len - out, 223 "%10s => Block: %lu Cluster: %d\n", "Sizes", 224 osb->sb->s_blocksize, osb->s_clustersize); 225 226 out += snprintf(buf + out, len - out, 227 "%10s => Compat: 0x%X Incompat: 0x%X " 228 "ROcompat: 0x%X\n", 229 "Features", osb->s_feature_compat, 230 osb->s_feature_incompat, osb->s_feature_ro_compat); 231 232 out += snprintf(buf + out, len - out, 233 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount", 234 osb->s_mount_opt, osb->s_atime_quantum); 235 236 if (cconn) { 237 out += snprintf(buf + out, len - out, 238 "%10s => Stack: %s Name: %*s " 239 "Version: %d.%d\n", "Cluster", 240 (*osb->osb_cluster_stack == '\0' ? 241 "o2cb" : osb->osb_cluster_stack), 242 cconn->cc_namelen, cconn->cc_name, 243 cconn->cc_version.pv_major, 244 cconn->cc_version.pv_minor); 245 } 246 247 spin_lock(&osb->dc_task_lock); 248 out += snprintf(buf + out, len - out, 249 "%10s => Pid: %d Count: %lu WakeSeq: %lu " 250 "WorkSeq: %lu\n", "DownCnvt", 251 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1), 252 osb->blocked_lock_count, osb->dc_wake_sequence, 253 osb->dc_work_sequence); 254 spin_unlock(&osb->dc_task_lock); 255 256 spin_lock(&osb->osb_lock); 257 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:", 258 "Recovery", 259 (osb->recovery_thread_task ? 260 task_pid_nr(osb->recovery_thread_task) : -1)); 261 if (rm->rm_used == 0) 262 out += snprintf(buf + out, len - out, " None\n"); 263 else { 264 for (i = 0; i < rm->rm_used; i++) 265 out += snprintf(buf + out, len - out, " %d", 266 rm->rm_entries[i]); 267 out += snprintf(buf + out, len - out, "\n"); 268 } 269 spin_unlock(&osb->osb_lock); 270 271 out += snprintf(buf + out, len - out, 272 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit", 273 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1), 274 osb->osb_commit_interval, 275 atomic_read(&osb->needs_checkpoint)); 276 277 out += snprintf(buf + out, len - out, 278 "%10s => State: %d TxnId: %lu NumTxns: %d\n", 279 "Journal", osb->journal->j_state, 280 osb->journal->j_trans_id, 281 atomic_read(&osb->journal->j_num_trans)); 282 283 out += snprintf(buf + out, len - out, 284 "%10s => GlobalAllocs: %d LocalAllocs: %d " 285 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n", 286 "Stats", 287 atomic_read(&osb->alloc_stats.bitmap_data), 288 atomic_read(&osb->alloc_stats.local_data), 289 atomic_read(&osb->alloc_stats.bg_allocs), 290 atomic_read(&osb->alloc_stats.moves), 291 atomic_read(&osb->alloc_stats.bg_extends)); 292 293 out += snprintf(buf + out, len - out, 294 "%10s => State: %u Descriptor: %llu Size: %u bits " 295 "Default: %u bits\n", 296 "LocalAlloc", osb->local_alloc_state, 297 (unsigned long long)osb->la_last_gd, 298 osb->local_alloc_bits, osb->local_alloc_default_bits); 299 300 spin_lock(&osb->osb_lock); 301 out += snprintf(buf + out, len - out, 302 "%10s => Slot: %d NumStolen: %d\n", "Steal", 303 osb->s_inode_steal_slot, 304 atomic_read(&osb->s_num_inodes_stolen)); 305 spin_unlock(&osb->osb_lock); 306 307 out += snprintf(buf + out, len - out, "OrphanScan => "); 308 out += snprintf(buf + out, len - out, "Local: %u Global: %u ", 309 os->os_count, os->os_seqno); 310 out += snprintf(buf + out, len - out, " Last Scan: "); 311 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE) 312 out += snprintf(buf + out, len - out, "Disabled\n"); 313 else 314 out += snprintf(buf + out, len - out, "%lu seconds ago\n", 315 (get_seconds() - os->os_scantime.tv_sec)); 316 317 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n", 318 "Slots", "Num", "RecoGen"); 319 for (i = 0; i < osb->max_slots; ++i) { 320 out += snprintf(buf + out, len - out, 321 "%10s %c %3d %10d\n", 322 " ", 323 (i == osb->slot_num ? '*' : ' '), 324 i, osb->slot_recovery_generations[i]); 325 } 326 327 return out; 328 } 329 330 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file) 331 { 332 struct ocfs2_super *osb = inode->i_private; 333 char *buf = NULL; 334 335 buf = kmalloc(PAGE_SIZE, GFP_KERNEL); 336 if (!buf) 337 goto bail; 338 339 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE)); 340 341 file->private_data = buf; 342 343 return 0; 344 bail: 345 return -ENOMEM; 346 } 347 348 static int ocfs2_debug_release(struct inode *inode, struct file *file) 349 { 350 kfree(file->private_data); 351 return 0; 352 } 353 354 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, 355 size_t nbytes, loff_t *ppos) 356 { 357 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, 358 i_size_read(file->f_mapping->host)); 359 } 360 #else 361 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file) 362 { 363 return 0; 364 } 365 static int ocfs2_debug_release(struct inode *inode, struct file *file) 366 { 367 return 0; 368 } 369 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf, 370 size_t nbytes, loff_t *ppos) 371 { 372 return 0; 373 } 374 #endif /* CONFIG_DEBUG_FS */ 375 376 static const struct file_operations ocfs2_osb_debug_fops = { 377 .open = ocfs2_osb_debug_open, 378 .release = ocfs2_debug_release, 379 .read = ocfs2_debug_read, 380 .llseek = generic_file_llseek, 381 }; 382 383 static int ocfs2_sync_fs(struct super_block *sb, int wait) 384 { 385 int status; 386 tid_t target; 387 struct ocfs2_super *osb = OCFS2_SB(sb); 388 389 if (ocfs2_is_hard_readonly(osb)) 390 return -EROFS; 391 392 if (wait) { 393 status = ocfs2_flush_truncate_log(osb); 394 if (status < 0) 395 mlog_errno(status); 396 } else { 397 ocfs2_schedule_truncate_log_flush(osb, 0); 398 } 399 400 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal, 401 &target)) { 402 if (wait) 403 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal, 404 target); 405 } 406 return 0; 407 } 408 409 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino) 410 { 411 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA) 412 && (ino == USER_QUOTA_SYSTEM_INODE 413 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE)) 414 return 0; 415 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA) 416 && (ino == GROUP_QUOTA_SYSTEM_INODE 417 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE)) 418 return 0; 419 return 1; 420 } 421 422 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) 423 { 424 struct inode *new = NULL; 425 int status = 0; 426 int i; 427 428 mlog_entry_void(); 429 430 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0); 431 if (IS_ERR(new)) { 432 status = PTR_ERR(new); 433 mlog_errno(status); 434 goto bail; 435 } 436 osb->root_inode = new; 437 438 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0); 439 if (IS_ERR(new)) { 440 status = PTR_ERR(new); 441 mlog_errno(status); 442 goto bail; 443 } 444 osb->sys_root_inode = new; 445 446 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE; 447 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) { 448 if (!ocfs2_need_system_inode(osb, i)) 449 continue; 450 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 451 if (!new) { 452 ocfs2_release_system_inodes(osb); 453 status = -EINVAL; 454 mlog_errno(status); 455 /* FIXME: Should ERROR_RO_FS */ 456 mlog(ML_ERROR, "Unable to load system inode %d, " 457 "possibly corrupt fs?", i); 458 goto bail; 459 } 460 // the array now has one ref, so drop this one 461 iput(new); 462 } 463 464 bail: 465 mlog_exit(status); 466 return status; 467 } 468 469 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) 470 { 471 struct inode *new = NULL; 472 int status = 0; 473 int i; 474 475 mlog_entry_void(); 476 477 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1; 478 i < NUM_SYSTEM_INODES; 479 i++) { 480 if (!ocfs2_need_system_inode(osb, i)) 481 continue; 482 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 483 if (!new) { 484 ocfs2_release_system_inodes(osb); 485 status = -EINVAL; 486 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n", 487 status, i, osb->slot_num); 488 goto bail; 489 } 490 /* the array now has one ref, so drop this one */ 491 iput(new); 492 } 493 494 bail: 495 mlog_exit(status); 496 return status; 497 } 498 499 static void ocfs2_release_system_inodes(struct ocfs2_super *osb) 500 { 501 int i; 502 struct inode *inode; 503 504 mlog_entry_void(); 505 506 for (i = 0; i < NUM_SYSTEM_INODES; i++) { 507 inode = osb->system_inodes[i]; 508 if (inode) { 509 iput(inode); 510 osb->system_inodes[i] = NULL; 511 } 512 } 513 514 inode = osb->sys_root_inode; 515 if (inode) { 516 iput(inode); 517 osb->sys_root_inode = NULL; 518 } 519 520 inode = osb->root_inode; 521 if (inode) { 522 iput(inode); 523 osb->root_inode = NULL; 524 } 525 526 mlog_exit(0); 527 } 528 529 /* We're allocating fs objects, use GFP_NOFS */ 530 static struct inode *ocfs2_alloc_inode(struct super_block *sb) 531 { 532 struct ocfs2_inode_info *oi; 533 534 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS); 535 if (!oi) 536 return NULL; 537 538 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode); 539 return &oi->vfs_inode; 540 } 541 542 static void ocfs2_destroy_inode(struct inode *inode) 543 { 544 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); 545 } 546 547 static unsigned long long ocfs2_max_file_offset(unsigned int bbits, 548 unsigned int cbits) 549 { 550 unsigned int bytes = 1 << cbits; 551 unsigned int trim = bytes; 552 unsigned int bitshift = 32; 553 554 /* 555 * i_size and all block offsets in ocfs2 are always 64 bits 556 * wide. i_clusters is 32 bits, in cluster-sized units. So on 557 * 64 bit platforms, cluster size will be the limiting factor. 558 */ 559 560 #if BITS_PER_LONG == 32 561 # if defined(CONFIG_LBDAF) 562 BUILD_BUG_ON(sizeof(sector_t) != 8); 563 /* 564 * We might be limited by page cache size. 565 */ 566 if (bytes > PAGE_CACHE_SIZE) { 567 bytes = PAGE_CACHE_SIZE; 568 trim = 1; 569 /* 570 * Shift by 31 here so that we don't get larger than 571 * MAX_LFS_FILESIZE 572 */ 573 bitshift = 31; 574 } 575 # else 576 /* 577 * We are limited by the size of sector_t. Use block size, as 578 * that's what we expose to the VFS. 579 */ 580 bytes = 1 << bbits; 581 trim = 1; 582 bitshift = 31; 583 # endif 584 #endif 585 586 /* 587 * Trim by a whole cluster when we can actually approach the 588 * on-disk limits. Otherwise we can overflow i_clusters when 589 * an extent start is at the max offset. 590 */ 591 return (((unsigned long long)bytes) << bitshift) - trim; 592 } 593 594 static int ocfs2_remount(struct super_block *sb, int *flags, char *data) 595 { 596 int incompat_features; 597 int ret = 0; 598 struct mount_options parsed_options; 599 struct ocfs2_super *osb = OCFS2_SB(sb); 600 601 lock_kernel(); 602 603 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) { 604 ret = -EINVAL; 605 goto out; 606 } 607 608 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) != 609 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) { 610 ret = -EINVAL; 611 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); 612 goto out; 613 } 614 615 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) != 616 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) { 617 ret = -EINVAL; 618 mlog(ML_ERROR, "Cannot change data mode on remount\n"); 619 goto out; 620 } 621 622 /* Probably don't want this on remount; it might 623 * mess with other nodes */ 624 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) && 625 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) { 626 ret = -EINVAL; 627 mlog(ML_ERROR, "Cannot enable inode64 on remount\n"); 628 goto out; 629 } 630 631 /* We're going to/from readonly mode. */ 632 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { 633 /* Disable quota accounting before remounting RO */ 634 if (*flags & MS_RDONLY) { 635 ret = ocfs2_susp_quotas(osb, 0); 636 if (ret < 0) 637 goto out; 638 } 639 /* Lock here so the check of HARD_RO and the potential 640 * setting of SOFT_RO is atomic. */ 641 spin_lock(&osb->osb_lock); 642 if (osb->osb_flags & OCFS2_OSB_HARD_RO) { 643 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n"); 644 ret = -EROFS; 645 goto unlock_osb; 646 } 647 648 if (*flags & MS_RDONLY) { 649 mlog(0, "Going to ro mode.\n"); 650 sb->s_flags |= MS_RDONLY; 651 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 652 } else { 653 mlog(0, "Making ro filesystem writeable.\n"); 654 655 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) { 656 mlog(ML_ERROR, "Cannot remount RDWR " 657 "filesystem due to previous errors.\n"); 658 ret = -EROFS; 659 goto unlock_osb; 660 } 661 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP); 662 if (incompat_features) { 663 mlog(ML_ERROR, "Cannot remount RDWR because " 664 "of unsupported optional features " 665 "(%x).\n", incompat_features); 666 ret = -EINVAL; 667 goto unlock_osb; 668 } 669 sb->s_flags &= ~MS_RDONLY; 670 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO; 671 } 672 unlock_osb: 673 spin_unlock(&osb->osb_lock); 674 /* Enable quota accounting after remounting RW */ 675 if (!ret && !(*flags & MS_RDONLY)) { 676 if (sb_any_quota_suspended(sb)) 677 ret = ocfs2_susp_quotas(osb, 1); 678 else 679 ret = ocfs2_enable_quotas(osb); 680 if (ret < 0) { 681 /* Return back changes... */ 682 spin_lock(&osb->osb_lock); 683 sb->s_flags |= MS_RDONLY; 684 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 685 spin_unlock(&osb->osb_lock); 686 goto out; 687 } 688 } 689 } 690 691 if (!ret) { 692 /* Only save off the new mount options in case of a successful 693 * remount. */ 694 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)) 695 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 696 osb->s_mount_opt = parsed_options.mount_opt; 697 osb->s_atime_quantum = parsed_options.atime_quantum; 698 osb->preferred_slot = parsed_options.slot; 699 if (parsed_options.commit_interval) 700 osb->osb_commit_interval = parsed_options.commit_interval; 701 702 if (!ocfs2_is_hard_readonly(osb)) 703 ocfs2_set_journal_params(osb); 704 } 705 out: 706 unlock_kernel(); 707 return ret; 708 } 709 710 static int ocfs2_sb_probe(struct super_block *sb, 711 struct buffer_head **bh, 712 int *sector_size, 713 struct ocfs2_blockcheck_stats *stats) 714 { 715 int status, tmpstat; 716 struct ocfs1_vol_disk_hdr *hdr; 717 struct ocfs2_dinode *di; 718 int blksize; 719 720 *bh = NULL; 721 722 /* may be > 512 */ 723 *sector_size = bdev_logical_block_size(sb->s_bdev); 724 if (*sector_size > OCFS2_MAX_BLOCKSIZE) { 725 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n", 726 *sector_size, OCFS2_MAX_BLOCKSIZE); 727 status = -EINVAL; 728 goto bail; 729 } 730 731 /* Can this really happen? */ 732 if (*sector_size < OCFS2_MIN_BLOCKSIZE) 733 *sector_size = OCFS2_MIN_BLOCKSIZE; 734 735 /* check block zero for old format */ 736 status = ocfs2_get_sector(sb, bh, 0, *sector_size); 737 if (status < 0) { 738 mlog_errno(status); 739 goto bail; 740 } 741 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data; 742 if (hdr->major_version == OCFS1_MAJOR_VERSION) { 743 mlog(ML_ERROR, "incompatible version: %u.%u\n", 744 hdr->major_version, hdr->minor_version); 745 status = -EINVAL; 746 } 747 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE, 748 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) { 749 mlog(ML_ERROR, "incompatible volume signature: %8s\n", 750 hdr->signature); 751 status = -EINVAL; 752 } 753 brelse(*bh); 754 *bh = NULL; 755 if (status < 0) { 756 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be " 757 "upgraded before mounting with ocfs v2\n"); 758 goto bail; 759 } 760 761 /* 762 * Now check at magic offset for 512, 1024, 2048, 4096 763 * blocksizes. 4096 is the maximum blocksize because it is 764 * the minimum clustersize. 765 */ 766 status = -EINVAL; 767 for (blksize = *sector_size; 768 blksize <= OCFS2_MAX_BLOCKSIZE; 769 blksize <<= 1) { 770 tmpstat = ocfs2_get_sector(sb, bh, 771 OCFS2_SUPER_BLOCK_BLKNO, 772 blksize); 773 if (tmpstat < 0) { 774 status = tmpstat; 775 mlog_errno(status); 776 break; 777 } 778 di = (struct ocfs2_dinode *) (*bh)->b_data; 779 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats)); 780 spin_lock_init(&stats->b_lock); 781 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats); 782 if (tmpstat < 0) { 783 brelse(*bh); 784 *bh = NULL; 785 } 786 if (tmpstat != -EAGAIN) { 787 status = tmpstat; 788 break; 789 } 790 } 791 792 bail: 793 return status; 794 } 795 796 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) 797 { 798 if (ocfs2_mount_local(osb)) { 799 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 800 mlog(ML_ERROR, "Cannot heartbeat on a locally " 801 "mounted device.\n"); 802 return -EINVAL; 803 } 804 } 805 806 if (ocfs2_userspace_stack(osb)) { 807 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 808 mlog(ML_ERROR, "Userspace stack expected, but " 809 "o2cb heartbeat arguments passed to mount\n"); 810 return -EINVAL; 811 } 812 } 813 814 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) { 815 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) && 816 !ocfs2_userspace_stack(osb)) { 817 mlog(ML_ERROR, "Heartbeat has to be started to mount " 818 "a read-write clustered device.\n"); 819 return -EINVAL; 820 } 821 } 822 823 return 0; 824 } 825 826 /* 827 * If we're using a userspace stack, mount should have passed 828 * a name that matches the disk. If not, mount should not 829 * have passed a stack. 830 */ 831 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb, 832 struct mount_options *mopt) 833 { 834 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) { 835 mlog(ML_ERROR, 836 "cluster stack passed to mount, but this filesystem " 837 "does not support it\n"); 838 return -EINVAL; 839 } 840 841 if (ocfs2_userspace_stack(osb) && 842 strncmp(osb->osb_cluster_stack, mopt->cluster_stack, 843 OCFS2_STACK_LABEL_LEN)) { 844 mlog(ML_ERROR, 845 "cluster stack passed to mount (\"%s\") does not " 846 "match the filesystem (\"%s\")\n", 847 mopt->cluster_stack, 848 osb->osb_cluster_stack); 849 return -EINVAL; 850 } 851 852 return 0; 853 } 854 855 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend) 856 { 857 int type; 858 struct super_block *sb = osb->sb; 859 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 860 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 861 int status = 0; 862 863 for (type = 0; type < MAXQUOTAS; type++) { 864 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 865 continue; 866 if (unsuspend) 867 status = vfs_quota_enable( 868 sb_dqopt(sb)->files[type], 869 type, QFMT_OCFS2, 870 DQUOT_SUSPENDED); 871 else 872 status = vfs_quota_disable(sb, type, 873 DQUOT_SUSPENDED); 874 if (status < 0) 875 break; 876 } 877 if (status < 0) 878 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on " 879 "remount (error = %d).\n", status); 880 return status; 881 } 882 883 static int ocfs2_enable_quotas(struct ocfs2_super *osb) 884 { 885 struct inode *inode[MAXQUOTAS] = { NULL, NULL }; 886 struct super_block *sb = osb->sb; 887 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 888 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 889 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE, 890 LOCAL_GROUP_QUOTA_SYSTEM_INODE }; 891 int status; 892 int type; 893 894 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE; 895 for (type = 0; type < MAXQUOTAS; type++) { 896 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 897 continue; 898 inode[type] = ocfs2_get_system_file_inode(osb, ino[type], 899 osb->slot_num); 900 if (!inode[type]) { 901 status = -ENOENT; 902 goto out_quota_off; 903 } 904 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2, 905 DQUOT_USAGE_ENABLED); 906 if (status < 0) 907 goto out_quota_off; 908 } 909 910 for (type = 0; type < MAXQUOTAS; type++) 911 iput(inode[type]); 912 return 0; 913 out_quota_off: 914 ocfs2_disable_quotas(osb); 915 for (type = 0; type < MAXQUOTAS; type++) 916 iput(inode[type]); 917 mlog_errno(status); 918 return status; 919 } 920 921 static void ocfs2_disable_quotas(struct ocfs2_super *osb) 922 { 923 int type; 924 struct inode *inode; 925 struct super_block *sb = osb->sb; 926 927 /* We mostly ignore errors in this function because there's not much 928 * we can do when we see them */ 929 for (type = 0; type < MAXQUOTAS; type++) { 930 if (!sb_has_quota_loaded(sb, type)) 931 continue; 932 inode = igrab(sb->s_dquot.files[type]); 933 /* Turn off quotas. This will remove all dquot structures from 934 * memory and so they will be automatically synced to global 935 * quota files */ 936 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED | 937 DQUOT_LIMITS_ENABLED); 938 if (!inode) 939 continue; 940 iput(inode); 941 } 942 } 943 944 /* Handle quota on quotactl */ 945 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id, 946 char *path, int remount) 947 { 948 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 949 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 950 951 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 952 return -EINVAL; 953 954 if (remount) 955 return 0; /* Just ignore it has been handled in 956 * ocfs2_remount() */ 957 return vfs_quota_enable(sb_dqopt(sb)->files[type], type, 958 format_id, DQUOT_LIMITS_ENABLED); 959 } 960 961 /* Handle quota off quotactl */ 962 static int ocfs2_quota_off(struct super_block *sb, int type, int remount) 963 { 964 if (remount) 965 return 0; /* Ignore now and handle later in 966 * ocfs2_remount() */ 967 return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED); 968 } 969 970 static const struct quotactl_ops ocfs2_quotactl_ops = { 971 .quota_on = ocfs2_quota_on, 972 .quota_off = ocfs2_quota_off, 973 .quota_sync = vfs_quota_sync, 974 .get_info = vfs_get_dqinfo, 975 .set_info = vfs_set_dqinfo, 976 .get_dqblk = vfs_get_dqblk, 977 .set_dqblk = vfs_set_dqblk, 978 }; 979 980 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) 981 { 982 struct dentry *root; 983 int status, sector_size; 984 struct mount_options parsed_options; 985 struct inode *inode = NULL; 986 struct ocfs2_super *osb = NULL; 987 struct buffer_head *bh = NULL; 988 char nodestr[8]; 989 struct ocfs2_blockcheck_stats stats; 990 991 mlog_entry("%p, %p, %i", sb, data, silent); 992 993 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { 994 status = -EINVAL; 995 goto read_super_error; 996 } 997 998 /* probe for superblock */ 999 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats); 1000 if (status < 0) { 1001 mlog(ML_ERROR, "superblock probe failed!\n"); 1002 goto read_super_error; 1003 } 1004 1005 status = ocfs2_initialize_super(sb, bh, sector_size, &stats); 1006 osb = OCFS2_SB(sb); 1007 if (status < 0) { 1008 mlog_errno(status); 1009 goto read_super_error; 1010 } 1011 brelse(bh); 1012 bh = NULL; 1013 1014 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)) 1015 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 1016 1017 osb->s_mount_opt = parsed_options.mount_opt; 1018 osb->s_atime_quantum = parsed_options.atime_quantum; 1019 osb->preferred_slot = parsed_options.slot; 1020 osb->osb_commit_interval = parsed_options.commit_interval; 1021 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt); 1022 osb->local_alloc_bits = osb->local_alloc_default_bits; 1023 if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA && 1024 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1025 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { 1026 status = -EINVAL; 1027 mlog(ML_ERROR, "User quotas were requested, but this " 1028 "filesystem does not have the feature enabled.\n"); 1029 goto read_super_error; 1030 } 1031 if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA && 1032 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1033 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { 1034 status = -EINVAL; 1035 mlog(ML_ERROR, "Group quotas were requested, but this " 1036 "filesystem does not have the feature enabled.\n"); 1037 goto read_super_error; 1038 } 1039 1040 status = ocfs2_verify_userspace_stack(osb, &parsed_options); 1041 if (status) 1042 goto read_super_error; 1043 1044 sb->s_magic = OCFS2_SUPER_MAGIC; 1045 1046 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 1047 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); 1048 1049 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY, 1050 * heartbeat=none */ 1051 if (bdev_read_only(sb->s_bdev)) { 1052 if (!(sb->s_flags & MS_RDONLY)) { 1053 status = -EACCES; 1054 mlog(ML_ERROR, "Readonly device detected but readonly " 1055 "mount was not specified.\n"); 1056 goto read_super_error; 1057 } 1058 1059 /* You should not be able to start a local heartbeat 1060 * on a readonly device. */ 1061 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 1062 status = -EROFS; 1063 mlog(ML_ERROR, "Local heartbeat specified on readonly " 1064 "device.\n"); 1065 goto read_super_error; 1066 } 1067 1068 status = ocfs2_check_journals_nolocks(osb); 1069 if (status < 0) { 1070 if (status == -EROFS) 1071 mlog(ML_ERROR, "Recovery required on readonly " 1072 "file system, but write access is " 1073 "unavailable.\n"); 1074 else 1075 mlog_errno(status); 1076 goto read_super_error; 1077 } 1078 1079 ocfs2_set_ro_flag(osb, 1); 1080 1081 printk(KERN_NOTICE "Readonly device detected. No cluster " 1082 "services will be utilized for this mount. Recovery " 1083 "will be skipped.\n"); 1084 } 1085 1086 if (!ocfs2_is_hard_readonly(osb)) { 1087 if (sb->s_flags & MS_RDONLY) 1088 ocfs2_set_ro_flag(osb, 0); 1089 } 1090 1091 status = ocfs2_verify_heartbeat(osb); 1092 if (status < 0) { 1093 mlog_errno(status); 1094 goto read_super_error; 1095 } 1096 1097 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, 1098 ocfs2_debugfs_root); 1099 if (!osb->osb_debug_root) { 1100 status = -EINVAL; 1101 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n"); 1102 goto read_super_error; 1103 } 1104 1105 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR, 1106 osb->osb_debug_root, 1107 osb, 1108 &ocfs2_osb_debug_fops); 1109 if (!osb->osb_ctxt) { 1110 status = -EINVAL; 1111 mlog_errno(status); 1112 goto read_super_error; 1113 } 1114 1115 if (ocfs2_meta_ecc(osb)) { 1116 status = ocfs2_blockcheck_stats_debugfs_install( 1117 &osb->osb_ecc_stats, 1118 osb->osb_debug_root); 1119 if (status) { 1120 mlog(ML_ERROR, 1121 "Unable to create blockcheck statistics " 1122 "files\n"); 1123 goto read_super_error; 1124 } 1125 } 1126 1127 status = ocfs2_mount_volume(sb); 1128 if (osb->root_inode) 1129 inode = igrab(osb->root_inode); 1130 1131 if (status < 0) 1132 goto read_super_error; 1133 1134 if (!inode) { 1135 status = -EIO; 1136 mlog_errno(status); 1137 goto read_super_error; 1138 } 1139 1140 root = d_alloc_root(inode); 1141 if (!root) { 1142 status = -ENOMEM; 1143 mlog_errno(status); 1144 goto read_super_error; 1145 } 1146 1147 sb->s_root = root; 1148 1149 ocfs2_complete_mount_recovery(osb); 1150 1151 if (ocfs2_mount_local(osb)) 1152 snprintf(nodestr, sizeof(nodestr), "local"); 1153 else 1154 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1155 1156 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " 1157 "with %s data mode.\n", 1158 osb->dev_str, nodestr, osb->slot_num, 1159 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : 1160 "ordered"); 1161 1162 atomic_set(&osb->vol_state, VOLUME_MOUNTED); 1163 wake_up(&osb->osb_mount_event); 1164 1165 /* Now we can initialize quotas because we can afford to wait 1166 * for cluster locks recovery now. That also means that truncation 1167 * log recovery can happen but that waits for proper quota setup */ 1168 if (!(sb->s_flags & MS_RDONLY)) { 1169 status = ocfs2_enable_quotas(osb); 1170 if (status < 0) { 1171 /* We have to err-out specially here because 1172 * s_root is already set */ 1173 mlog_errno(status); 1174 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1175 wake_up(&osb->osb_mount_event); 1176 mlog_exit(status); 1177 return status; 1178 } 1179 } 1180 1181 ocfs2_complete_quota_recovery(osb); 1182 1183 /* Now we wake up again for processes waiting for quotas */ 1184 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS); 1185 wake_up(&osb->osb_mount_event); 1186 1187 /* Start this when the mount is almost sure of being successful */ 1188 ocfs2_orphan_scan_start(osb); 1189 1190 mlog_exit(status); 1191 return status; 1192 1193 read_super_error: 1194 brelse(bh); 1195 1196 if (inode) 1197 iput(inode); 1198 1199 if (osb) { 1200 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1201 wake_up(&osb->osb_mount_event); 1202 ocfs2_dismount_volume(sb, 1); 1203 } 1204 1205 mlog_exit(status); 1206 return status; 1207 } 1208 1209 static int ocfs2_get_sb(struct file_system_type *fs_type, 1210 int flags, 1211 const char *dev_name, 1212 void *data, 1213 struct vfsmount *mnt) 1214 { 1215 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super, 1216 mnt); 1217 } 1218 1219 static void ocfs2_kill_sb(struct super_block *sb) 1220 { 1221 struct ocfs2_super *osb = OCFS2_SB(sb); 1222 1223 /* Failed mount? */ 1224 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED) 1225 goto out; 1226 1227 /* Prevent further queueing of inode drop events */ 1228 spin_lock(&dentry_list_lock); 1229 ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED); 1230 spin_unlock(&dentry_list_lock); 1231 /* Wait for work to finish and/or remove it */ 1232 cancel_work_sync(&osb->dentry_lock_work); 1233 out: 1234 kill_block_super(sb); 1235 } 1236 1237 static struct file_system_type ocfs2_fs_type = { 1238 .owner = THIS_MODULE, 1239 .name = "ocfs2", 1240 .get_sb = ocfs2_get_sb, /* is this called when we mount 1241 * the fs? */ 1242 .kill_sb = ocfs2_kill_sb, 1243 1244 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, 1245 .next = NULL 1246 }; 1247 1248 static int ocfs2_parse_options(struct super_block *sb, 1249 char *options, 1250 struct mount_options *mopt, 1251 int is_remount) 1252 { 1253 int status; 1254 char *p; 1255 1256 mlog_entry("remount: %d, options: \"%s\"\n", is_remount, 1257 options ? options : "(none)"); 1258 1259 mopt->commit_interval = 0; 1260 mopt->mount_opt = 0; 1261 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 1262 mopt->slot = OCFS2_INVALID_SLOT; 1263 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE; 1264 mopt->cluster_stack[0] = '\0'; 1265 1266 if (!options) { 1267 status = 1; 1268 goto bail; 1269 } 1270 1271 while ((p = strsep(&options, ",")) != NULL) { 1272 int token, option; 1273 substring_t args[MAX_OPT_ARGS]; 1274 1275 if (!*p) 1276 continue; 1277 1278 token = match_token(p, tokens, args); 1279 switch (token) { 1280 case Opt_hb_local: 1281 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; 1282 break; 1283 case Opt_hb_none: 1284 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; 1285 break; 1286 case Opt_barrier: 1287 if (match_int(&args[0], &option)) { 1288 status = 0; 1289 goto bail; 1290 } 1291 if (option) 1292 mopt->mount_opt |= OCFS2_MOUNT_BARRIER; 1293 else 1294 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; 1295 break; 1296 case Opt_intr: 1297 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; 1298 break; 1299 case Opt_nointr: 1300 mopt->mount_opt |= OCFS2_MOUNT_NOINTR; 1301 break; 1302 case Opt_err_panic: 1303 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 1304 break; 1305 case Opt_err_ro: 1306 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 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 option = 0; 1330 if (match_int(&args[0], &option)) { 1331 status = 0; 1332 goto bail; 1333 } 1334 if (option) 1335 mopt->slot = (s16)option; 1336 break; 1337 case Opt_commit: 1338 option = 0; 1339 if (match_int(&args[0], &option)) { 1340 status = 0; 1341 goto bail; 1342 } 1343 if (option < 0) 1344 return 0; 1345 if (option == 0) 1346 option = JBD2_DEFAULT_MAX_COMMIT_AGE; 1347 mopt->commit_interval = HZ * option; 1348 break; 1349 case Opt_localalloc: 1350 option = 0; 1351 if (match_int(&args[0], &option)) { 1352 status = 0; 1353 goto bail; 1354 } 1355 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8)) 1356 mopt->localalloc_opt = option; 1357 break; 1358 case Opt_localflocks: 1359 /* 1360 * Changing this during remount could race 1361 * flock() requests, or "unbalance" existing 1362 * ones (e.g., a lock is taken in one mode but 1363 * dropped in the other). If users care enough 1364 * to flip locking modes during remount, we 1365 * could add a "local" flag to individual 1366 * flock structures for proper tracking of 1367 * state. 1368 */ 1369 if (!is_remount) 1370 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS; 1371 break; 1372 case Opt_stack: 1373 /* Check both that the option we were passed 1374 * is of the right length and that it is a proper 1375 * string of the right length. 1376 */ 1377 if (((args[0].to - args[0].from) != 1378 OCFS2_STACK_LABEL_LEN) || 1379 (strnlen(args[0].from, 1380 OCFS2_STACK_LABEL_LEN) != 1381 OCFS2_STACK_LABEL_LEN)) { 1382 mlog(ML_ERROR, 1383 "Invalid cluster_stack option\n"); 1384 status = 0; 1385 goto bail; 1386 } 1387 memcpy(mopt->cluster_stack, args[0].from, 1388 OCFS2_STACK_LABEL_LEN); 1389 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; 1390 break; 1391 case Opt_inode64: 1392 mopt->mount_opt |= OCFS2_MOUNT_INODE64; 1393 break; 1394 case Opt_usrquota: 1395 /* We check only on remount, otherwise features 1396 * aren't yet initialized. */ 1397 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1398 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { 1399 mlog(ML_ERROR, "User quota requested but " 1400 "filesystem feature is not set\n"); 1401 status = 0; 1402 goto bail; 1403 } 1404 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA; 1405 break; 1406 case Opt_grpquota: 1407 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1408 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { 1409 mlog(ML_ERROR, "Group quota requested but " 1410 "filesystem feature is not set\n"); 1411 status = 0; 1412 goto bail; 1413 } 1414 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA; 1415 break; 1416 #ifdef CONFIG_OCFS2_FS_POSIX_ACL 1417 case Opt_acl: 1418 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1419 break; 1420 case Opt_noacl: 1421 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 1422 break; 1423 #else 1424 case Opt_acl: 1425 case Opt_noacl: 1426 printk(KERN_INFO "ocfs2 (no)acl options not supported\n"); 1427 break; 1428 #endif 1429 default: 1430 mlog(ML_ERROR, 1431 "Unrecognized mount option \"%s\" " 1432 "or missing value\n", p); 1433 status = 0; 1434 goto bail; 1435 } 1436 } 1437 1438 status = 1; 1439 1440 bail: 1441 mlog_exit(status); 1442 return status; 1443 } 1444 1445 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) 1446 { 1447 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb); 1448 unsigned long opts = osb->s_mount_opt; 1449 unsigned int local_alloc_megs; 1450 1451 if (opts & OCFS2_MOUNT_HB_LOCAL) 1452 seq_printf(s, ",_netdev,heartbeat=local"); 1453 else 1454 seq_printf(s, ",heartbeat=none"); 1455 1456 if (opts & OCFS2_MOUNT_NOINTR) 1457 seq_printf(s, ",nointr"); 1458 1459 if (opts & OCFS2_MOUNT_DATA_WRITEBACK) 1460 seq_printf(s, ",data=writeback"); 1461 else 1462 seq_printf(s, ",data=ordered"); 1463 1464 if (opts & OCFS2_MOUNT_BARRIER) 1465 seq_printf(s, ",barrier=1"); 1466 1467 if (opts & OCFS2_MOUNT_ERRORS_PANIC) 1468 seq_printf(s, ",errors=panic"); 1469 else 1470 seq_printf(s, ",errors=remount-ro"); 1471 1472 if (osb->preferred_slot != OCFS2_INVALID_SLOT) 1473 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); 1474 1475 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM) 1476 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); 1477 1478 if (osb->osb_commit_interval) 1479 seq_printf(s, ",commit=%u", 1480 (unsigned) (osb->osb_commit_interval / HZ)); 1481 1482 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits); 1483 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE) 1484 seq_printf(s, ",localalloc=%d", local_alloc_megs); 1485 1486 if (opts & OCFS2_MOUNT_LOCALFLOCKS) 1487 seq_printf(s, ",localflocks,"); 1488 1489 if (osb->osb_cluster_stack[0]) 1490 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN, 1491 osb->osb_cluster_stack); 1492 if (opts & OCFS2_MOUNT_USRQUOTA) 1493 seq_printf(s, ",usrquota"); 1494 if (opts & OCFS2_MOUNT_GRPQUOTA) 1495 seq_printf(s, ",grpquota"); 1496 1497 if (opts & OCFS2_MOUNT_NOUSERXATTR) 1498 seq_printf(s, ",nouser_xattr"); 1499 else 1500 seq_printf(s, ",user_xattr"); 1501 1502 if (opts & OCFS2_MOUNT_INODE64) 1503 seq_printf(s, ",inode64"); 1504 1505 #ifdef CONFIG_OCFS2_FS_POSIX_ACL 1506 if (opts & OCFS2_MOUNT_POSIX_ACL) 1507 seq_printf(s, ",acl"); 1508 else 1509 seq_printf(s, ",noacl"); 1510 #endif 1511 1512 return 0; 1513 } 1514 1515 static int __init ocfs2_init(void) 1516 { 1517 int status; 1518 1519 mlog_entry_void(); 1520 1521 ocfs2_print_version(); 1522 1523 status = init_ocfs2_uptodate_cache(); 1524 if (status < 0) { 1525 mlog_errno(status); 1526 goto leave; 1527 } 1528 1529 status = ocfs2_initialize_mem_caches(); 1530 if (status < 0) { 1531 mlog_errno(status); 1532 goto leave; 1533 } 1534 1535 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq"); 1536 if (!ocfs2_wq) { 1537 status = -ENOMEM; 1538 goto leave; 1539 } 1540 1541 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); 1542 if (!ocfs2_debugfs_root) { 1543 status = -EFAULT; 1544 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n"); 1545 } 1546 1547 status = ocfs2_quota_setup(); 1548 if (status) 1549 goto leave; 1550 1551 ocfs2_set_locking_protocol(); 1552 1553 status = register_quota_format(&ocfs2_quota_format); 1554 leave: 1555 if (status < 0) { 1556 ocfs2_quota_shutdown(); 1557 ocfs2_free_mem_caches(); 1558 exit_ocfs2_uptodate_cache(); 1559 } 1560 1561 mlog_exit(status); 1562 1563 if (status >= 0) { 1564 return register_filesystem(&ocfs2_fs_type); 1565 } else 1566 return -1; 1567 } 1568 1569 static void __exit ocfs2_exit(void) 1570 { 1571 mlog_entry_void(); 1572 1573 ocfs2_quota_shutdown(); 1574 1575 if (ocfs2_wq) { 1576 flush_workqueue(ocfs2_wq); 1577 destroy_workqueue(ocfs2_wq); 1578 } 1579 1580 unregister_quota_format(&ocfs2_quota_format); 1581 1582 debugfs_remove(ocfs2_debugfs_root); 1583 1584 ocfs2_free_mem_caches(); 1585 1586 unregister_filesystem(&ocfs2_fs_type); 1587 1588 exit_ocfs2_uptodate_cache(); 1589 1590 mlog_exit_void(); 1591 } 1592 1593 static void ocfs2_put_super(struct super_block *sb) 1594 { 1595 mlog_entry("(0x%p)\n", sb); 1596 1597 lock_kernel(); 1598 1599 ocfs2_sync_blockdev(sb); 1600 ocfs2_dismount_volume(sb, 0); 1601 1602 unlock_kernel(); 1603 1604 mlog_exit_void(); 1605 } 1606 1607 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) 1608 { 1609 struct ocfs2_super *osb; 1610 u32 numbits, freebits; 1611 int status; 1612 struct ocfs2_dinode *bm_lock; 1613 struct buffer_head *bh = NULL; 1614 struct inode *inode = NULL; 1615 1616 mlog_entry("(%p, %p)\n", dentry->d_sb, buf); 1617 1618 osb = OCFS2_SB(dentry->d_sb); 1619 1620 inode = ocfs2_get_system_file_inode(osb, 1621 GLOBAL_BITMAP_SYSTEM_INODE, 1622 OCFS2_INVALID_SLOT); 1623 if (!inode) { 1624 mlog(ML_ERROR, "failed to get bitmap inode\n"); 1625 status = -EIO; 1626 goto bail; 1627 } 1628 1629 status = ocfs2_inode_lock(inode, &bh, 0); 1630 if (status < 0) { 1631 mlog_errno(status); 1632 goto bail; 1633 } 1634 1635 bm_lock = (struct ocfs2_dinode *) bh->b_data; 1636 1637 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); 1638 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); 1639 1640 buf->f_type = OCFS2_SUPER_MAGIC; 1641 buf->f_bsize = dentry->d_sb->s_blocksize; 1642 buf->f_namelen = OCFS2_MAX_FILENAME_LEN; 1643 buf->f_blocks = ((sector_t) numbits) * 1644 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1645 buf->f_bfree = ((sector_t) freebits) * 1646 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1647 buf->f_bavail = buf->f_bfree; 1648 buf->f_files = numbits; 1649 buf->f_ffree = freebits; 1650 1651 brelse(bh); 1652 1653 ocfs2_inode_unlock(inode, 0); 1654 status = 0; 1655 bail: 1656 if (inode) 1657 iput(inode); 1658 1659 mlog_exit(status); 1660 1661 return status; 1662 } 1663 1664 static void ocfs2_inode_init_once(void *data) 1665 { 1666 struct ocfs2_inode_info *oi = data; 1667 1668 oi->ip_flags = 0; 1669 oi->ip_open_count = 0; 1670 spin_lock_init(&oi->ip_lock); 1671 ocfs2_extent_map_init(&oi->vfs_inode); 1672 INIT_LIST_HEAD(&oi->ip_io_markers); 1673 oi->ip_dir_start_lookup = 0; 1674 1675 init_rwsem(&oi->ip_alloc_sem); 1676 init_rwsem(&oi->ip_xattr_sem); 1677 mutex_init(&oi->ip_io_mutex); 1678 1679 oi->ip_blkno = 0ULL; 1680 oi->ip_clusters = 0; 1681 1682 ocfs2_lock_res_init_once(&oi->ip_rw_lockres); 1683 ocfs2_lock_res_init_once(&oi->ip_inode_lockres); 1684 ocfs2_lock_res_init_once(&oi->ip_open_lockres); 1685 1686 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode), 1687 &ocfs2_inode_caching_ops); 1688 1689 inode_init_once(&oi->vfs_inode); 1690 } 1691 1692 static int ocfs2_initialize_mem_caches(void) 1693 { 1694 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", 1695 sizeof(struct ocfs2_inode_info), 1696 0, 1697 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1698 SLAB_MEM_SPREAD), 1699 ocfs2_inode_init_once); 1700 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache", 1701 sizeof(struct ocfs2_dquot), 1702 0, 1703 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1704 SLAB_MEM_SPREAD), 1705 NULL); 1706 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache", 1707 sizeof(struct ocfs2_quota_chunk), 1708 0, 1709 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), 1710 NULL); 1711 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep || 1712 !ocfs2_qf_chunk_cachep) { 1713 if (ocfs2_inode_cachep) 1714 kmem_cache_destroy(ocfs2_inode_cachep); 1715 if (ocfs2_dquot_cachep) 1716 kmem_cache_destroy(ocfs2_dquot_cachep); 1717 if (ocfs2_qf_chunk_cachep) 1718 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1719 return -ENOMEM; 1720 } 1721 1722 return 0; 1723 } 1724 1725 static void ocfs2_free_mem_caches(void) 1726 { 1727 if (ocfs2_inode_cachep) 1728 kmem_cache_destroy(ocfs2_inode_cachep); 1729 ocfs2_inode_cachep = NULL; 1730 1731 if (ocfs2_dquot_cachep) 1732 kmem_cache_destroy(ocfs2_dquot_cachep); 1733 ocfs2_dquot_cachep = NULL; 1734 1735 if (ocfs2_qf_chunk_cachep) 1736 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1737 ocfs2_qf_chunk_cachep = NULL; 1738 } 1739 1740 static int ocfs2_get_sector(struct super_block *sb, 1741 struct buffer_head **bh, 1742 int block, 1743 int sect_size) 1744 { 1745 if (!sb_set_blocksize(sb, sect_size)) { 1746 mlog(ML_ERROR, "unable to set blocksize\n"); 1747 return -EIO; 1748 } 1749 1750 *bh = sb_getblk(sb, block); 1751 if (!*bh) { 1752 mlog_errno(-EIO); 1753 return -EIO; 1754 } 1755 lock_buffer(*bh); 1756 if (!buffer_dirty(*bh)) 1757 clear_buffer_uptodate(*bh); 1758 unlock_buffer(*bh); 1759 ll_rw_block(READ, 1, bh); 1760 wait_on_buffer(*bh); 1761 if (!buffer_uptodate(*bh)) { 1762 mlog_errno(-EIO); 1763 brelse(*bh); 1764 *bh = NULL; 1765 return -EIO; 1766 } 1767 1768 return 0; 1769 } 1770 1771 static int ocfs2_mount_volume(struct super_block *sb) 1772 { 1773 int status = 0; 1774 int unlock_super = 0; 1775 struct ocfs2_super *osb = OCFS2_SB(sb); 1776 1777 mlog_entry_void(); 1778 1779 if (ocfs2_is_hard_readonly(osb)) 1780 goto leave; 1781 1782 status = ocfs2_dlm_init(osb); 1783 if (status < 0) { 1784 mlog_errno(status); 1785 goto leave; 1786 } 1787 1788 status = ocfs2_super_lock(osb, 1); 1789 if (status < 0) { 1790 mlog_errno(status); 1791 goto leave; 1792 } 1793 unlock_super = 1; 1794 1795 /* This will load up the node map and add ourselves to it. */ 1796 status = ocfs2_find_slot(osb); 1797 if (status < 0) { 1798 mlog_errno(status); 1799 goto leave; 1800 } 1801 1802 /* load all node-local system inodes */ 1803 status = ocfs2_init_local_system_inodes(osb); 1804 if (status < 0) { 1805 mlog_errno(status); 1806 goto leave; 1807 } 1808 1809 status = ocfs2_check_volume(osb); 1810 if (status < 0) { 1811 mlog_errno(status); 1812 goto leave; 1813 } 1814 1815 status = ocfs2_truncate_log_init(osb); 1816 if (status < 0) 1817 mlog_errno(status); 1818 1819 leave: 1820 if (unlock_super) 1821 ocfs2_super_unlock(osb, 1); 1822 1823 mlog_exit(status); 1824 return status; 1825 } 1826 1827 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) 1828 { 1829 int tmp, hangup_needed = 0; 1830 struct ocfs2_super *osb = NULL; 1831 char nodestr[8]; 1832 1833 mlog_entry("(0x%p)\n", sb); 1834 1835 BUG_ON(!sb); 1836 osb = OCFS2_SB(sb); 1837 BUG_ON(!osb); 1838 1839 debugfs_remove(osb->osb_ctxt); 1840 1841 /* 1842 * Flush inode dropping work queue so that deletes are 1843 * performed while the filesystem is still working 1844 */ 1845 ocfs2_drop_all_dl_inodes(osb); 1846 1847 /* Orphan scan should be stopped as early as possible */ 1848 ocfs2_orphan_scan_stop(osb); 1849 1850 ocfs2_disable_quotas(osb); 1851 1852 ocfs2_shutdown_local_alloc(osb); 1853 1854 ocfs2_truncate_log_shutdown(osb); 1855 1856 /* This will disable recovery and flush any recovery work. */ 1857 ocfs2_recovery_exit(osb); 1858 1859 ocfs2_journal_shutdown(osb); 1860 1861 ocfs2_sync_blockdev(sb); 1862 1863 ocfs2_purge_refcount_trees(osb); 1864 1865 /* No cluster connection means we've failed during mount, so skip 1866 * all the steps which depended on that to complete. */ 1867 if (osb->cconn) { 1868 tmp = ocfs2_super_lock(osb, 1); 1869 if (tmp < 0) { 1870 mlog_errno(tmp); 1871 return; 1872 } 1873 } 1874 1875 if (osb->slot_num != OCFS2_INVALID_SLOT) 1876 ocfs2_put_slot(osb); 1877 1878 if (osb->cconn) 1879 ocfs2_super_unlock(osb, 1); 1880 1881 ocfs2_release_system_inodes(osb); 1882 1883 /* 1884 * If we're dismounting due to mount error, mount.ocfs2 will clean 1885 * up heartbeat. If we're a local mount, there is no heartbeat. 1886 * If we failed before we got a uuid_str yet, we can't stop 1887 * heartbeat. Otherwise, do it. 1888 */ 1889 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str) 1890 hangup_needed = 1; 1891 1892 if (osb->cconn) 1893 ocfs2_dlm_shutdown(osb, hangup_needed); 1894 1895 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats); 1896 debugfs_remove(osb->osb_debug_root); 1897 1898 if (hangup_needed) 1899 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str)); 1900 1901 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); 1902 1903 if (ocfs2_mount_local(osb)) 1904 snprintf(nodestr, sizeof(nodestr), "local"); 1905 else 1906 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1907 1908 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", 1909 osb->dev_str, nodestr); 1910 1911 ocfs2_delete_osb(osb); 1912 kfree(osb); 1913 sb->s_dev = 0; 1914 sb->s_fs_info = NULL; 1915 } 1916 1917 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, 1918 unsigned uuid_bytes) 1919 { 1920 int i, ret; 1921 char *ptr; 1922 1923 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); 1924 1925 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); 1926 if (osb->uuid_str == NULL) 1927 return -ENOMEM; 1928 1929 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { 1930 /* print with null */ 1931 ret = snprintf(ptr, 3, "%02X", uuid[i]); 1932 if (ret != 2) /* drop super cleans up */ 1933 return -EINVAL; 1934 /* then only advance past the last char */ 1935 ptr += 2; 1936 } 1937 1938 return 0; 1939 } 1940 1941 static int ocfs2_initialize_super(struct super_block *sb, 1942 struct buffer_head *bh, 1943 int sector_size, 1944 struct ocfs2_blockcheck_stats *stats) 1945 { 1946 int status; 1947 int i, cbits, bbits; 1948 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; 1949 struct inode *inode = NULL; 1950 struct ocfs2_journal *journal; 1951 __le32 uuid_net_key; 1952 struct ocfs2_super *osb; 1953 1954 mlog_entry_void(); 1955 1956 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); 1957 if (!osb) { 1958 status = -ENOMEM; 1959 mlog_errno(status); 1960 goto bail; 1961 } 1962 1963 sb->s_fs_info = osb; 1964 sb->s_op = &ocfs2_sops; 1965 sb->s_export_op = &ocfs2_export_ops; 1966 sb->s_qcop = &ocfs2_quotactl_ops; 1967 sb->dq_op = &ocfs2_quota_operations; 1968 sb->s_xattr = ocfs2_xattr_handlers; 1969 sb->s_time_gran = 1; 1970 sb->s_flags |= MS_NOATIME; 1971 /* this is needed to support O_LARGEFILE */ 1972 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); 1973 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); 1974 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); 1975 1976 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1; 1977 1978 for (i = 0; i < 3; i++) 1979 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]); 1980 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash); 1981 1982 osb->sb = sb; 1983 /* Save off for ocfs2_rw_direct */ 1984 osb->s_sectsize_bits = blksize_bits(sector_size); 1985 BUG_ON(!osb->s_sectsize_bits); 1986 1987 spin_lock_init(&osb->dc_task_lock); 1988 init_waitqueue_head(&osb->dc_event); 1989 osb->dc_work_sequence = 0; 1990 osb->dc_wake_sequence = 0; 1991 INIT_LIST_HEAD(&osb->blocked_lock_list); 1992 osb->blocked_lock_count = 0; 1993 spin_lock_init(&osb->osb_lock); 1994 spin_lock_init(&osb->osb_xattr_lock); 1995 ocfs2_init_inode_steal_slot(osb); 1996 1997 atomic_set(&osb->alloc_stats.moves, 0); 1998 atomic_set(&osb->alloc_stats.local_data, 0); 1999 atomic_set(&osb->alloc_stats.bitmap_data, 0); 2000 atomic_set(&osb->alloc_stats.bg_allocs, 0); 2001 atomic_set(&osb->alloc_stats.bg_extends, 0); 2002 2003 /* Copy the blockcheck stats from the superblock probe */ 2004 osb->osb_ecc_stats = *stats; 2005 2006 ocfs2_init_node_maps(osb); 2007 2008 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", 2009 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); 2010 2011 ocfs2_orphan_scan_init(osb); 2012 2013 status = ocfs2_recovery_init(osb); 2014 if (status) { 2015 mlog(ML_ERROR, "Unable to initialize recovery state\n"); 2016 mlog_errno(status); 2017 goto bail; 2018 } 2019 2020 init_waitqueue_head(&osb->checkpoint_event); 2021 atomic_set(&osb->needs_checkpoint, 0); 2022 2023 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 2024 2025 osb->slot_num = OCFS2_INVALID_SLOT; 2026 2027 osb->s_xattr_inline_size = le16_to_cpu( 2028 di->id2.i_super.s_xattr_inline_size); 2029 2030 osb->local_alloc_state = OCFS2_LA_UNUSED; 2031 osb->local_alloc_bh = NULL; 2032 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker); 2033 2034 init_waitqueue_head(&osb->osb_mount_event); 2035 2036 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); 2037 if (!osb->vol_label) { 2038 mlog(ML_ERROR, "unable to alloc vol label\n"); 2039 status = -ENOMEM; 2040 goto bail; 2041 } 2042 2043 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); 2044 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { 2045 mlog(ML_ERROR, "Invalid number of node slots (%u)\n", 2046 osb->max_slots); 2047 status = -EINVAL; 2048 goto bail; 2049 } 2050 mlog(0, "max_slots for this device: %u\n", osb->max_slots); 2051 2052 osb->slot_recovery_generations = 2053 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), 2054 GFP_KERNEL); 2055 if (!osb->slot_recovery_generations) { 2056 status = -ENOMEM; 2057 mlog_errno(status); 2058 goto bail; 2059 } 2060 2061 init_waitqueue_head(&osb->osb_wipe_event); 2062 osb->osb_orphan_wipes = kcalloc(osb->max_slots, 2063 sizeof(*osb->osb_orphan_wipes), 2064 GFP_KERNEL); 2065 if (!osb->osb_orphan_wipes) { 2066 status = -ENOMEM; 2067 mlog_errno(status); 2068 goto bail; 2069 } 2070 2071 osb->osb_rf_lock_tree = RB_ROOT; 2072 2073 osb->s_feature_compat = 2074 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); 2075 osb->s_feature_ro_compat = 2076 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); 2077 osb->s_feature_incompat = 2078 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); 2079 2080 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { 2081 mlog(ML_ERROR, "couldn't mount because of unsupported " 2082 "optional features (%x).\n", i); 2083 status = -EINVAL; 2084 goto bail; 2085 } 2086 if (!(osb->sb->s_flags & MS_RDONLY) && 2087 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { 2088 mlog(ML_ERROR, "couldn't mount RDWR because of " 2089 "unsupported optional features (%x).\n", i); 2090 status = -EINVAL; 2091 goto bail; 2092 } 2093 2094 if (ocfs2_userspace_stack(osb)) { 2095 memcpy(osb->osb_cluster_stack, 2096 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, 2097 OCFS2_STACK_LABEL_LEN); 2098 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; 2099 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) { 2100 mlog(ML_ERROR, 2101 "couldn't mount because of an invalid " 2102 "cluster stack label (%s) \n", 2103 osb->osb_cluster_stack); 2104 status = -EINVAL; 2105 goto bail; 2106 } 2107 } else { 2108 /* The empty string is identical with classic tools that 2109 * don't know about s_cluster_info. */ 2110 osb->osb_cluster_stack[0] = '\0'; 2111 } 2112 2113 get_random_bytes(&osb->s_next_generation, sizeof(u32)); 2114 2115 /* FIXME 2116 * This should be done in ocfs2_journal_init(), but unknown 2117 * ordering issues will cause the filesystem to crash. 2118 * If anyone wants to figure out what part of the code 2119 * refers to osb->journal before ocfs2_journal_init() is run, 2120 * be my guest. 2121 */ 2122 /* initialize our journal structure */ 2123 2124 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL); 2125 if (!journal) { 2126 mlog(ML_ERROR, "unable to alloc journal\n"); 2127 status = -ENOMEM; 2128 goto bail; 2129 } 2130 osb->journal = journal; 2131 journal->j_osb = osb; 2132 2133 atomic_set(&journal->j_num_trans, 0); 2134 init_rwsem(&journal->j_trans_barrier); 2135 init_waitqueue_head(&journal->j_checkpointed); 2136 spin_lock_init(&journal->j_lock); 2137 journal->j_trans_id = (unsigned long) 1; 2138 INIT_LIST_HEAD(&journal->j_la_cleanups); 2139 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); 2140 journal->j_state = OCFS2_JOURNAL_FREE; 2141 2142 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes); 2143 osb->dentry_lock_list = NULL; 2144 2145 /* get some pseudo constants for clustersize bits */ 2146 osb->s_clustersize_bits = 2147 le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2148 osb->s_clustersize = 1 << osb->s_clustersize_bits; 2149 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits); 2150 2151 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || 2152 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { 2153 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", 2154 osb->s_clustersize); 2155 status = -EINVAL; 2156 goto bail; 2157 } 2158 2159 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1) 2160 > (u32)~0UL) { 2161 mlog(ML_ERROR, "Volume might try to write to blocks beyond " 2162 "what jbd can address in 32 bits.\n"); 2163 status = -EINVAL; 2164 goto bail; 2165 } 2166 2167 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, 2168 sizeof(di->id2.i_super.s_uuid))) { 2169 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); 2170 status = -ENOMEM; 2171 goto bail; 2172 } 2173 2174 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key)); 2175 2176 strncpy(osb->vol_label, di->id2.i_super.s_label, 63); 2177 osb->vol_label[63] = '\0'; 2178 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); 2179 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); 2180 osb->first_cluster_group_blkno = 2181 le64_to_cpu(di->id2.i_super.s_first_cluster_group); 2182 osb->fs_generation = le32_to_cpu(di->i_fs_generation); 2183 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2184 mlog(0, "vol_label: %s\n", osb->vol_label); 2185 mlog(0, "uuid: %s\n", osb->uuid_str); 2186 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n", 2187 (unsigned long long)osb->root_blkno, 2188 (unsigned long long)osb->system_dir_blkno); 2189 2190 osb->osb_dlm_debug = ocfs2_new_dlm_debug(); 2191 if (!osb->osb_dlm_debug) { 2192 status = -ENOMEM; 2193 mlog_errno(status); 2194 goto bail; 2195 } 2196 2197 atomic_set(&osb->vol_state, VOLUME_INIT); 2198 2199 /* load root, system_dir, and all global system inodes */ 2200 status = ocfs2_init_global_system_inodes(osb); 2201 if (status < 0) { 2202 mlog_errno(status); 2203 goto bail; 2204 } 2205 2206 /* 2207 * global bitmap 2208 */ 2209 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, 2210 OCFS2_INVALID_SLOT); 2211 if (!inode) { 2212 status = -EINVAL; 2213 mlog_errno(status); 2214 goto bail; 2215 } 2216 2217 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; 2218 iput(inode); 2219 2220 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8; 2221 2222 status = ocfs2_init_slot_info(osb); 2223 if (status < 0) { 2224 mlog_errno(status); 2225 goto bail; 2226 } 2227 2228 bail: 2229 mlog_exit(status); 2230 return status; 2231 } 2232 2233 /* 2234 * will return: -EAGAIN if it is ok to keep searching for superblocks 2235 * -EINVAL if there is a bad superblock 2236 * 0 on success 2237 */ 2238 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 2239 struct buffer_head *bh, 2240 u32 blksz, 2241 struct ocfs2_blockcheck_stats *stats) 2242 { 2243 int status = -EAGAIN; 2244 2245 mlog_entry_void(); 2246 2247 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, 2248 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { 2249 /* We have to do a raw check of the feature here */ 2250 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) & 2251 OCFS2_FEATURE_INCOMPAT_META_ECC) { 2252 status = ocfs2_block_check_validate(bh->b_data, 2253 bh->b_size, 2254 &di->i_check, 2255 stats); 2256 if (status) 2257 goto out; 2258 } 2259 status = -EINVAL; 2260 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) { 2261 mlog(ML_ERROR, "found superblock with incorrect block " 2262 "size: found %u, should be %u\n", 2263 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits), 2264 blksz); 2265 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != 2266 OCFS2_MAJOR_REV_LEVEL || 2267 le16_to_cpu(di->id2.i_super.s_minor_rev_level) != 2268 OCFS2_MINOR_REV_LEVEL) { 2269 mlog(ML_ERROR, "found superblock with bad version: " 2270 "found %u.%u, should be %u.%u\n", 2271 le16_to_cpu(di->id2.i_super.s_major_rev_level), 2272 le16_to_cpu(di->id2.i_super.s_minor_rev_level), 2273 OCFS2_MAJOR_REV_LEVEL, 2274 OCFS2_MINOR_REV_LEVEL); 2275 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { 2276 mlog(ML_ERROR, "bad block number on superblock: " 2277 "found %llu, should be %llu\n", 2278 (unsigned long long)le64_to_cpu(di->i_blkno), 2279 (unsigned long long)bh->b_blocknr); 2280 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || 2281 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { 2282 mlog(ML_ERROR, "bad cluster size found: %u\n", 2283 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); 2284 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { 2285 mlog(ML_ERROR, "bad root_blkno: 0\n"); 2286 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { 2287 mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); 2288 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { 2289 mlog(ML_ERROR, 2290 "Superblock slots found greater than file system " 2291 "maximum: found %u, max %u\n", 2292 le16_to_cpu(di->id2.i_super.s_max_slots), 2293 OCFS2_MAX_SLOTS); 2294 } else { 2295 /* found it! */ 2296 status = 0; 2297 } 2298 } 2299 2300 out: 2301 mlog_exit(status); 2302 return status; 2303 } 2304 2305 static int ocfs2_check_volume(struct ocfs2_super *osb) 2306 { 2307 int status; 2308 int dirty; 2309 int local; 2310 struct ocfs2_dinode *local_alloc = NULL; /* only used if we 2311 * recover 2312 * ourselves. */ 2313 2314 mlog_entry_void(); 2315 2316 /* Init our journal object. */ 2317 status = ocfs2_journal_init(osb->journal, &dirty); 2318 if (status < 0) { 2319 mlog(ML_ERROR, "Could not initialize journal!\n"); 2320 goto finally; 2321 } 2322 2323 /* If the journal was unmounted cleanly then we don't want to 2324 * recover anything. Otherwise, journal_load will do that 2325 * dirty work for us :) */ 2326 if (!dirty) { 2327 status = ocfs2_journal_wipe(osb->journal, 0); 2328 if (status < 0) { 2329 mlog_errno(status); 2330 goto finally; 2331 } 2332 } else { 2333 mlog(ML_NOTICE, "File system was not unmounted cleanly, " 2334 "recovering volume.\n"); 2335 } 2336 2337 local = ocfs2_mount_local(osb); 2338 2339 /* will play back anything left in the journal. */ 2340 status = ocfs2_journal_load(osb->journal, local, dirty); 2341 if (status < 0) { 2342 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status); 2343 goto finally; 2344 } 2345 2346 if (dirty) { 2347 /* recover my local alloc if we didn't unmount cleanly. */ 2348 status = ocfs2_begin_local_alloc_recovery(osb, 2349 osb->slot_num, 2350 &local_alloc); 2351 if (status < 0) { 2352 mlog_errno(status); 2353 goto finally; 2354 } 2355 /* we complete the recovery process after we've marked 2356 * ourselves as mounted. */ 2357 } 2358 2359 mlog(0, "Journal loaded.\n"); 2360 2361 status = ocfs2_load_local_alloc(osb); 2362 if (status < 0) { 2363 mlog_errno(status); 2364 goto finally; 2365 } 2366 2367 if (dirty) { 2368 /* Recovery will be completed after we've mounted the 2369 * rest of the volume. */ 2370 osb->dirty = 1; 2371 osb->local_alloc_copy = local_alloc; 2372 local_alloc = NULL; 2373 } 2374 2375 /* go through each journal, trylock it and if you get the 2376 * lock, and it's marked as dirty, set the bit in the recover 2377 * map and launch a recovery thread for it. */ 2378 status = ocfs2_mark_dead_nodes(osb); 2379 if (status < 0) { 2380 mlog_errno(status); 2381 goto finally; 2382 } 2383 2384 status = ocfs2_compute_replay_slots(osb); 2385 if (status < 0) 2386 mlog_errno(status); 2387 2388 finally: 2389 if (local_alloc) 2390 kfree(local_alloc); 2391 2392 mlog_exit(status); 2393 return status; 2394 } 2395 2396 /* 2397 * The routine gets called from dismount or close whenever a dismount on 2398 * volume is requested and the osb open count becomes 1. 2399 * It will remove the osb from the global list and also free up all the 2400 * initialized resources and fileobject. 2401 */ 2402 static void ocfs2_delete_osb(struct ocfs2_super *osb) 2403 { 2404 mlog_entry_void(); 2405 2406 /* This function assumes that the caller has the main osb resource */ 2407 2408 ocfs2_free_slot_info(osb); 2409 2410 kfree(osb->osb_orphan_wipes); 2411 kfree(osb->slot_recovery_generations); 2412 /* FIXME 2413 * This belongs in journal shutdown, but because we have to 2414 * allocate osb->journal at the start of ocfs2_initalize_osb(), 2415 * we free it here. 2416 */ 2417 kfree(osb->journal); 2418 if (osb->local_alloc_copy) 2419 kfree(osb->local_alloc_copy); 2420 kfree(osb->uuid_str); 2421 ocfs2_put_dlm_debug(osb->osb_dlm_debug); 2422 memset(osb, 0, sizeof(struct ocfs2_super)); 2423 2424 mlog_exit_void(); 2425 } 2426 2427 /* Put OCFS2 into a readonly state, or (if the user specifies it), 2428 * panic(). We do not support continue-on-error operation. */ 2429 static void ocfs2_handle_error(struct super_block *sb) 2430 { 2431 struct ocfs2_super *osb = OCFS2_SB(sb); 2432 2433 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) 2434 panic("OCFS2: (device %s): panic forced after error\n", 2435 sb->s_id); 2436 2437 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); 2438 2439 if (sb->s_flags & MS_RDONLY && 2440 (ocfs2_is_soft_readonly(osb) || 2441 ocfs2_is_hard_readonly(osb))) 2442 return; 2443 2444 printk(KERN_CRIT "File system is now read-only due to the potential " 2445 "of on-disk corruption. Please run fsck.ocfs2 once the file " 2446 "system is unmounted.\n"); 2447 sb->s_flags |= MS_RDONLY; 2448 ocfs2_set_ro_flag(osb, 0); 2449 } 2450 2451 static char error_buf[1024]; 2452 2453 void __ocfs2_error(struct super_block *sb, 2454 const char *function, 2455 const char *fmt, ...) 2456 { 2457 va_list args; 2458 2459 va_start(args, fmt); 2460 vsnprintf(error_buf, sizeof(error_buf), fmt, args); 2461 va_end(args); 2462 2463 /* Not using mlog here because we want to show the actual 2464 * function the error came from. */ 2465 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n", 2466 sb->s_id, function, error_buf); 2467 2468 ocfs2_handle_error(sb); 2469 } 2470 2471 /* Handle critical errors. This is intentionally more drastic than 2472 * ocfs2_handle_error, so we only use for things like journal errors, 2473 * etc. */ 2474 void __ocfs2_abort(struct super_block* sb, 2475 const char *function, 2476 const char *fmt, ...) 2477 { 2478 va_list args; 2479 2480 va_start(args, fmt); 2481 vsnprintf(error_buf, sizeof(error_buf), fmt, args); 2482 va_end(args); 2483 2484 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n", 2485 sb->s_id, function, error_buf); 2486 2487 /* We don't have the cluster support yet to go straight to 2488 * hard readonly in here. Until then, we want to keep 2489 * ocfs2_abort() so that we can at least mark critical 2490 * errors. 2491 * 2492 * TODO: This should abort the journal and alert other nodes 2493 * that our slot needs recovery. */ 2494 2495 /* Force a panic(). This stinks, but it's better than letting 2496 * things continue without having a proper hard readonly 2497 * here. */ 2498 if (!ocfs2_mount_local(OCFS2_SB(sb))) 2499 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 2500 ocfs2_handle_error(sb); 2501 } 2502 2503 module_init(ocfs2_init); 2504 module_exit(ocfs2_exit); 2505