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/utsname.h> 32 #include <linux/init.h> 33 #include <linux/random.h> 34 #include <linux/statfs.h> 35 #include <linux/moduleparam.h> 36 #include <linux/blkdev.h> 37 #include <linux/socket.h> 38 #include <linux/inet.h> 39 #include <linux/parser.h> 40 #include <linux/crc32.h> 41 #include <linux/debugfs.h> 42 #include <linux/mount.h> 43 #include <linux/seq_file.h> 44 #include <linux/quotaops.h> 45 #include <linux/smp_lock.h> 46 47 #define MLOG_MASK_PREFIX ML_SUPER 48 #include <cluster/masklog.h> 49 50 #include "ocfs2.h" 51 52 /* this should be the only file to include a version 1 header */ 53 #include "ocfs1_fs_compat.h" 54 55 #include "alloc.h" 56 #include "blockcheck.h" 57 #include "dlmglue.h" 58 #include "export.h" 59 #include "extent_map.h" 60 #include "heartbeat.h" 61 #include "inode.h" 62 #include "journal.h" 63 #include "localalloc.h" 64 #include "namei.h" 65 #include "slot_map.h" 66 #include "super.h" 67 #include "sysfile.h" 68 #include "uptodate.h" 69 #include "ver.h" 70 #include "xattr.h" 71 #include "quota.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 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 goto bail; 777 } 778 di = (struct ocfs2_dinode *) (*bh)->b_data; 779 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats)); 780 status = ocfs2_verify_volume(di, *bh, blksize, stats); 781 if (status >= 0) 782 goto bail; 783 brelse(*bh); 784 *bh = NULL; 785 if (status != -EAGAIN) 786 break; 787 } 788 789 bail: 790 return status; 791 } 792 793 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) 794 { 795 if (ocfs2_mount_local(osb)) { 796 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 797 mlog(ML_ERROR, "Cannot heartbeat on a locally " 798 "mounted device.\n"); 799 return -EINVAL; 800 } 801 } 802 803 if (ocfs2_userspace_stack(osb)) { 804 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 805 mlog(ML_ERROR, "Userspace stack expected, but " 806 "o2cb heartbeat arguments passed to mount\n"); 807 return -EINVAL; 808 } 809 } 810 811 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) { 812 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) && 813 !ocfs2_userspace_stack(osb)) { 814 mlog(ML_ERROR, "Heartbeat has to be started to mount " 815 "a read-write clustered device.\n"); 816 return -EINVAL; 817 } 818 } 819 820 return 0; 821 } 822 823 /* 824 * If we're using a userspace stack, mount should have passed 825 * a name that matches the disk. If not, mount should not 826 * have passed a stack. 827 */ 828 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb, 829 struct mount_options *mopt) 830 { 831 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) { 832 mlog(ML_ERROR, 833 "cluster stack passed to mount, but this filesystem " 834 "does not support it\n"); 835 return -EINVAL; 836 } 837 838 if (ocfs2_userspace_stack(osb) && 839 strncmp(osb->osb_cluster_stack, mopt->cluster_stack, 840 OCFS2_STACK_LABEL_LEN)) { 841 mlog(ML_ERROR, 842 "cluster stack passed to mount (\"%s\") does not " 843 "match the filesystem (\"%s\")\n", 844 mopt->cluster_stack, 845 osb->osb_cluster_stack); 846 return -EINVAL; 847 } 848 849 return 0; 850 } 851 852 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend) 853 { 854 int type; 855 struct super_block *sb = osb->sb; 856 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 857 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 858 int status = 0; 859 860 for (type = 0; type < MAXQUOTAS; type++) { 861 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 862 continue; 863 if (unsuspend) 864 status = vfs_quota_enable( 865 sb_dqopt(sb)->files[type], 866 type, QFMT_OCFS2, 867 DQUOT_SUSPENDED); 868 else 869 status = vfs_quota_disable(sb, type, 870 DQUOT_SUSPENDED); 871 if (status < 0) 872 break; 873 } 874 if (status < 0) 875 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on " 876 "remount (error = %d).\n", status); 877 return status; 878 } 879 880 static int ocfs2_enable_quotas(struct ocfs2_super *osb) 881 { 882 struct inode *inode[MAXQUOTAS] = { NULL, NULL }; 883 struct super_block *sb = osb->sb; 884 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 885 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 886 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE, 887 LOCAL_GROUP_QUOTA_SYSTEM_INODE }; 888 int status; 889 int type; 890 891 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE; 892 for (type = 0; type < MAXQUOTAS; type++) { 893 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 894 continue; 895 inode[type] = ocfs2_get_system_file_inode(osb, ino[type], 896 osb->slot_num); 897 if (!inode[type]) { 898 status = -ENOENT; 899 goto out_quota_off; 900 } 901 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2, 902 DQUOT_USAGE_ENABLED); 903 if (status < 0) 904 goto out_quota_off; 905 } 906 907 for (type = 0; type < MAXQUOTAS; type++) 908 iput(inode[type]); 909 return 0; 910 out_quota_off: 911 ocfs2_disable_quotas(osb); 912 for (type = 0; type < MAXQUOTAS; type++) 913 iput(inode[type]); 914 mlog_errno(status); 915 return status; 916 } 917 918 static void ocfs2_disable_quotas(struct ocfs2_super *osb) 919 { 920 int type; 921 struct inode *inode; 922 struct super_block *sb = osb->sb; 923 924 /* We mostly ignore errors in this function because there's not much 925 * we can do when we see them */ 926 for (type = 0; type < MAXQUOTAS; type++) { 927 if (!sb_has_quota_loaded(sb, type)) 928 continue; 929 inode = igrab(sb->s_dquot.files[type]); 930 /* Turn off quotas. This will remove all dquot structures from 931 * memory and so they will be automatically synced to global 932 * quota files */ 933 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED | 934 DQUOT_LIMITS_ENABLED); 935 if (!inode) 936 continue; 937 iput(inode); 938 } 939 } 940 941 /* Handle quota on quotactl */ 942 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id, 943 char *path, int remount) 944 { 945 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA, 946 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA}; 947 948 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type])) 949 return -EINVAL; 950 951 if (remount) 952 return 0; /* Just ignore it has been handled in 953 * ocfs2_remount() */ 954 return vfs_quota_enable(sb_dqopt(sb)->files[type], type, 955 format_id, DQUOT_LIMITS_ENABLED); 956 } 957 958 /* Handle quota off quotactl */ 959 static int ocfs2_quota_off(struct super_block *sb, int type, int remount) 960 { 961 if (remount) 962 return 0; /* Ignore now and handle later in 963 * ocfs2_remount() */ 964 return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED); 965 } 966 967 static struct quotactl_ops ocfs2_quotactl_ops = { 968 .quota_on = ocfs2_quota_on, 969 .quota_off = ocfs2_quota_off, 970 .quota_sync = vfs_quota_sync, 971 .get_info = vfs_get_dqinfo, 972 .set_info = vfs_set_dqinfo, 973 .get_dqblk = vfs_get_dqblk, 974 .set_dqblk = vfs_set_dqblk, 975 }; 976 977 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) 978 { 979 struct dentry *root; 980 int status, sector_size; 981 struct mount_options parsed_options; 982 struct inode *inode = NULL; 983 struct ocfs2_super *osb = NULL; 984 struct buffer_head *bh = NULL; 985 char nodestr[8]; 986 struct ocfs2_blockcheck_stats stats; 987 988 mlog_entry("%p, %p, %i", sb, data, silent); 989 990 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { 991 status = -EINVAL; 992 goto read_super_error; 993 } 994 995 /* probe for superblock */ 996 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats); 997 if (status < 0) { 998 mlog(ML_ERROR, "superblock probe failed!\n"); 999 goto read_super_error; 1000 } 1001 1002 status = ocfs2_initialize_super(sb, bh, sector_size, &stats); 1003 osb = OCFS2_SB(sb); 1004 if (status < 0) { 1005 mlog_errno(status); 1006 goto read_super_error; 1007 } 1008 brelse(bh); 1009 bh = NULL; 1010 1011 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)) 1012 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 1013 1014 osb->s_mount_opt = parsed_options.mount_opt; 1015 osb->s_atime_quantum = parsed_options.atime_quantum; 1016 osb->preferred_slot = parsed_options.slot; 1017 osb->osb_commit_interval = parsed_options.commit_interval; 1018 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt); 1019 osb->local_alloc_bits = osb->local_alloc_default_bits; 1020 if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA && 1021 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1022 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { 1023 status = -EINVAL; 1024 mlog(ML_ERROR, "User quotas were requested, but this " 1025 "filesystem does not have the feature enabled.\n"); 1026 goto read_super_error; 1027 } 1028 if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA && 1029 !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1030 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { 1031 status = -EINVAL; 1032 mlog(ML_ERROR, "Group quotas were requested, but this " 1033 "filesystem does not have the feature enabled.\n"); 1034 goto read_super_error; 1035 } 1036 1037 status = ocfs2_verify_userspace_stack(osb, &parsed_options); 1038 if (status) 1039 goto read_super_error; 1040 1041 sb->s_magic = OCFS2_SUPER_MAGIC; 1042 1043 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 1044 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); 1045 1046 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY, 1047 * heartbeat=none */ 1048 if (bdev_read_only(sb->s_bdev)) { 1049 if (!(sb->s_flags & MS_RDONLY)) { 1050 status = -EACCES; 1051 mlog(ML_ERROR, "Readonly device detected but readonly " 1052 "mount was not specified.\n"); 1053 goto read_super_error; 1054 } 1055 1056 /* You should not be able to start a local heartbeat 1057 * on a readonly device. */ 1058 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 1059 status = -EROFS; 1060 mlog(ML_ERROR, "Local heartbeat specified on readonly " 1061 "device.\n"); 1062 goto read_super_error; 1063 } 1064 1065 status = ocfs2_check_journals_nolocks(osb); 1066 if (status < 0) { 1067 if (status == -EROFS) 1068 mlog(ML_ERROR, "Recovery required on readonly " 1069 "file system, but write access is " 1070 "unavailable.\n"); 1071 else 1072 mlog_errno(status); 1073 goto read_super_error; 1074 } 1075 1076 ocfs2_set_ro_flag(osb, 1); 1077 1078 printk(KERN_NOTICE "Readonly device detected. No cluster " 1079 "services will be utilized for this mount. Recovery " 1080 "will be skipped.\n"); 1081 } 1082 1083 if (!ocfs2_is_hard_readonly(osb)) { 1084 if (sb->s_flags & MS_RDONLY) 1085 ocfs2_set_ro_flag(osb, 0); 1086 } 1087 1088 status = ocfs2_verify_heartbeat(osb); 1089 if (status < 0) { 1090 mlog_errno(status); 1091 goto read_super_error; 1092 } 1093 1094 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, 1095 ocfs2_debugfs_root); 1096 if (!osb->osb_debug_root) { 1097 status = -EINVAL; 1098 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n"); 1099 goto read_super_error; 1100 } 1101 1102 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR, 1103 osb->osb_debug_root, 1104 osb, 1105 &ocfs2_osb_debug_fops); 1106 if (!osb->osb_ctxt) { 1107 status = -EINVAL; 1108 mlog_errno(status); 1109 goto read_super_error; 1110 } 1111 1112 if (ocfs2_meta_ecc(osb)) { 1113 status = ocfs2_blockcheck_stats_debugfs_install( 1114 &osb->osb_ecc_stats, 1115 osb->osb_debug_root); 1116 if (status) { 1117 mlog(ML_ERROR, 1118 "Unable to create blockcheck statistics " 1119 "files\n"); 1120 goto read_super_error; 1121 } 1122 } 1123 1124 status = ocfs2_mount_volume(sb); 1125 if (osb->root_inode) 1126 inode = igrab(osb->root_inode); 1127 1128 if (status < 0) 1129 goto read_super_error; 1130 1131 if (!inode) { 1132 status = -EIO; 1133 mlog_errno(status); 1134 goto read_super_error; 1135 } 1136 1137 root = d_alloc_root(inode); 1138 if (!root) { 1139 status = -ENOMEM; 1140 mlog_errno(status); 1141 goto read_super_error; 1142 } 1143 1144 sb->s_root = root; 1145 1146 ocfs2_complete_mount_recovery(osb); 1147 1148 if (ocfs2_mount_local(osb)) 1149 snprintf(nodestr, sizeof(nodestr), "local"); 1150 else 1151 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1152 1153 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " 1154 "with %s data mode.\n", 1155 osb->dev_str, nodestr, osb->slot_num, 1156 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : 1157 "ordered"); 1158 1159 atomic_set(&osb->vol_state, VOLUME_MOUNTED); 1160 wake_up(&osb->osb_mount_event); 1161 1162 /* Now we can initialize quotas because we can afford to wait 1163 * for cluster locks recovery now. That also means that truncation 1164 * log recovery can happen but that waits for proper quota setup */ 1165 if (!(sb->s_flags & MS_RDONLY)) { 1166 status = ocfs2_enable_quotas(osb); 1167 if (status < 0) { 1168 /* We have to err-out specially here because 1169 * s_root is already set */ 1170 mlog_errno(status); 1171 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1172 wake_up(&osb->osb_mount_event); 1173 mlog_exit(status); 1174 return status; 1175 } 1176 } 1177 1178 ocfs2_complete_quota_recovery(osb); 1179 1180 /* Now we wake up again for processes waiting for quotas */ 1181 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS); 1182 wake_up(&osb->osb_mount_event); 1183 1184 /* Start this when the mount is almost sure of being successful */ 1185 ocfs2_orphan_scan_init(osb); 1186 1187 mlog_exit(status); 1188 return status; 1189 1190 read_super_error: 1191 brelse(bh); 1192 1193 if (inode) 1194 iput(inode); 1195 1196 if (osb) { 1197 atomic_set(&osb->vol_state, VOLUME_DISABLED); 1198 wake_up(&osb->osb_mount_event); 1199 ocfs2_dismount_volume(sb, 1); 1200 } 1201 1202 mlog_exit(status); 1203 return status; 1204 } 1205 1206 static int ocfs2_get_sb(struct file_system_type *fs_type, 1207 int flags, 1208 const char *dev_name, 1209 void *data, 1210 struct vfsmount *mnt) 1211 { 1212 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super, 1213 mnt); 1214 } 1215 1216 static struct file_system_type ocfs2_fs_type = { 1217 .owner = THIS_MODULE, 1218 .name = "ocfs2", 1219 .get_sb = ocfs2_get_sb, /* is this called when we mount 1220 * the fs? */ 1221 .kill_sb = kill_block_super, /* set to the generic one 1222 * right now, but do we 1223 * need to change that? */ 1224 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, 1225 .next = NULL 1226 }; 1227 1228 static int ocfs2_parse_options(struct super_block *sb, 1229 char *options, 1230 struct mount_options *mopt, 1231 int is_remount) 1232 { 1233 int status; 1234 char *p; 1235 1236 mlog_entry("remount: %d, options: \"%s\"\n", is_remount, 1237 options ? options : "(none)"); 1238 1239 mopt->commit_interval = 0; 1240 mopt->mount_opt = 0; 1241 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 1242 mopt->slot = OCFS2_INVALID_SLOT; 1243 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE; 1244 mopt->cluster_stack[0] = '\0'; 1245 1246 if (!options) { 1247 status = 1; 1248 goto bail; 1249 } 1250 1251 while ((p = strsep(&options, ",")) != NULL) { 1252 int token, option; 1253 substring_t args[MAX_OPT_ARGS]; 1254 1255 if (!*p) 1256 continue; 1257 1258 token = match_token(p, tokens, args); 1259 switch (token) { 1260 case Opt_hb_local: 1261 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; 1262 break; 1263 case Opt_hb_none: 1264 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; 1265 break; 1266 case Opt_barrier: 1267 if (match_int(&args[0], &option)) { 1268 status = 0; 1269 goto bail; 1270 } 1271 if (option) 1272 mopt->mount_opt |= OCFS2_MOUNT_BARRIER; 1273 else 1274 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; 1275 break; 1276 case Opt_intr: 1277 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; 1278 break; 1279 case Opt_nointr: 1280 mopt->mount_opt |= OCFS2_MOUNT_NOINTR; 1281 break; 1282 case Opt_err_panic: 1283 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 1284 break; 1285 case Opt_err_ro: 1286 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 1287 break; 1288 case Opt_data_ordered: 1289 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; 1290 break; 1291 case Opt_data_writeback: 1292 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; 1293 break; 1294 case Opt_user_xattr: 1295 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR; 1296 break; 1297 case Opt_nouser_xattr: 1298 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR; 1299 break; 1300 case Opt_atime_quantum: 1301 if (match_int(&args[0], &option)) { 1302 status = 0; 1303 goto bail; 1304 } 1305 if (option >= 0) 1306 mopt->atime_quantum = option; 1307 break; 1308 case Opt_slot: 1309 option = 0; 1310 if (match_int(&args[0], &option)) { 1311 status = 0; 1312 goto bail; 1313 } 1314 if (option) 1315 mopt->slot = (s16)option; 1316 break; 1317 case Opt_commit: 1318 option = 0; 1319 if (match_int(&args[0], &option)) { 1320 status = 0; 1321 goto bail; 1322 } 1323 if (option < 0) 1324 return 0; 1325 if (option == 0) 1326 option = JBD2_DEFAULT_MAX_COMMIT_AGE; 1327 mopt->commit_interval = HZ * option; 1328 break; 1329 case Opt_localalloc: 1330 option = 0; 1331 if (match_int(&args[0], &option)) { 1332 status = 0; 1333 goto bail; 1334 } 1335 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8)) 1336 mopt->localalloc_opt = option; 1337 break; 1338 case Opt_localflocks: 1339 /* 1340 * Changing this during remount could race 1341 * flock() requests, or "unbalance" existing 1342 * ones (e.g., a lock is taken in one mode but 1343 * dropped in the other). If users care enough 1344 * to flip locking modes during remount, we 1345 * could add a "local" flag to individual 1346 * flock structures for proper tracking of 1347 * state. 1348 */ 1349 if (!is_remount) 1350 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS; 1351 break; 1352 case Opt_stack: 1353 /* Check both that the option we were passed 1354 * is of the right length and that it is a proper 1355 * string of the right length. 1356 */ 1357 if (((args[0].to - args[0].from) != 1358 OCFS2_STACK_LABEL_LEN) || 1359 (strnlen(args[0].from, 1360 OCFS2_STACK_LABEL_LEN) != 1361 OCFS2_STACK_LABEL_LEN)) { 1362 mlog(ML_ERROR, 1363 "Invalid cluster_stack option\n"); 1364 status = 0; 1365 goto bail; 1366 } 1367 memcpy(mopt->cluster_stack, args[0].from, 1368 OCFS2_STACK_LABEL_LEN); 1369 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; 1370 break; 1371 case Opt_inode64: 1372 mopt->mount_opt |= OCFS2_MOUNT_INODE64; 1373 break; 1374 case Opt_usrquota: 1375 /* We check only on remount, otherwise features 1376 * aren't yet initialized. */ 1377 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1378 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) { 1379 mlog(ML_ERROR, "User quota requested but " 1380 "filesystem feature is not set\n"); 1381 status = 0; 1382 goto bail; 1383 } 1384 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA; 1385 break; 1386 case Opt_grpquota: 1387 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb, 1388 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) { 1389 mlog(ML_ERROR, "Group quota requested but " 1390 "filesystem feature is not set\n"); 1391 status = 0; 1392 goto bail; 1393 } 1394 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA; 1395 break; 1396 #ifdef CONFIG_OCFS2_FS_POSIX_ACL 1397 case Opt_acl: 1398 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; 1399 break; 1400 case Opt_noacl: 1401 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; 1402 break; 1403 #else 1404 case Opt_acl: 1405 case Opt_noacl: 1406 printk(KERN_INFO "ocfs2 (no)acl options not supported\n"); 1407 break; 1408 #endif 1409 default: 1410 mlog(ML_ERROR, 1411 "Unrecognized mount option \"%s\" " 1412 "or missing value\n", p); 1413 status = 0; 1414 goto bail; 1415 } 1416 } 1417 1418 status = 1; 1419 1420 bail: 1421 mlog_exit(status); 1422 return status; 1423 } 1424 1425 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) 1426 { 1427 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb); 1428 unsigned long opts = osb->s_mount_opt; 1429 unsigned int local_alloc_megs; 1430 1431 if (opts & OCFS2_MOUNT_HB_LOCAL) 1432 seq_printf(s, ",_netdev,heartbeat=local"); 1433 else 1434 seq_printf(s, ",heartbeat=none"); 1435 1436 if (opts & OCFS2_MOUNT_NOINTR) 1437 seq_printf(s, ",nointr"); 1438 1439 if (opts & OCFS2_MOUNT_DATA_WRITEBACK) 1440 seq_printf(s, ",data=writeback"); 1441 else 1442 seq_printf(s, ",data=ordered"); 1443 1444 if (opts & OCFS2_MOUNT_BARRIER) 1445 seq_printf(s, ",barrier=1"); 1446 1447 if (opts & OCFS2_MOUNT_ERRORS_PANIC) 1448 seq_printf(s, ",errors=panic"); 1449 else 1450 seq_printf(s, ",errors=remount-ro"); 1451 1452 if (osb->preferred_slot != OCFS2_INVALID_SLOT) 1453 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); 1454 1455 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM) 1456 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); 1457 1458 if (osb->osb_commit_interval) 1459 seq_printf(s, ",commit=%u", 1460 (unsigned) (osb->osb_commit_interval / HZ)); 1461 1462 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits); 1463 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE) 1464 seq_printf(s, ",localalloc=%d", local_alloc_megs); 1465 1466 if (opts & OCFS2_MOUNT_LOCALFLOCKS) 1467 seq_printf(s, ",localflocks,"); 1468 1469 if (osb->osb_cluster_stack[0]) 1470 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN, 1471 osb->osb_cluster_stack); 1472 if (opts & OCFS2_MOUNT_USRQUOTA) 1473 seq_printf(s, ",usrquota"); 1474 if (opts & OCFS2_MOUNT_GRPQUOTA) 1475 seq_printf(s, ",grpquota"); 1476 1477 if (opts & OCFS2_MOUNT_NOUSERXATTR) 1478 seq_printf(s, ",nouser_xattr"); 1479 else 1480 seq_printf(s, ",user_xattr"); 1481 1482 if (opts & OCFS2_MOUNT_INODE64) 1483 seq_printf(s, ",inode64"); 1484 1485 #ifdef CONFIG_OCFS2_FS_POSIX_ACL 1486 if (opts & OCFS2_MOUNT_POSIX_ACL) 1487 seq_printf(s, ",acl"); 1488 else 1489 seq_printf(s, ",noacl"); 1490 #endif 1491 1492 return 0; 1493 } 1494 1495 static int __init ocfs2_init(void) 1496 { 1497 int status; 1498 1499 mlog_entry_void(); 1500 1501 ocfs2_print_version(); 1502 1503 status = init_ocfs2_uptodate_cache(); 1504 if (status < 0) { 1505 mlog_errno(status); 1506 goto leave; 1507 } 1508 1509 status = ocfs2_initialize_mem_caches(); 1510 if (status < 0) { 1511 mlog_errno(status); 1512 goto leave; 1513 } 1514 1515 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq"); 1516 if (!ocfs2_wq) { 1517 status = -ENOMEM; 1518 goto leave; 1519 } 1520 1521 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); 1522 if (!ocfs2_debugfs_root) { 1523 status = -EFAULT; 1524 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n"); 1525 } 1526 1527 status = ocfs2_quota_setup(); 1528 if (status) 1529 goto leave; 1530 1531 ocfs2_set_locking_protocol(); 1532 1533 status = register_quota_format(&ocfs2_quota_format); 1534 leave: 1535 if (status < 0) { 1536 ocfs2_quota_shutdown(); 1537 ocfs2_free_mem_caches(); 1538 exit_ocfs2_uptodate_cache(); 1539 } 1540 1541 mlog_exit(status); 1542 1543 if (status >= 0) { 1544 return register_filesystem(&ocfs2_fs_type); 1545 } else 1546 return -1; 1547 } 1548 1549 static void __exit ocfs2_exit(void) 1550 { 1551 mlog_entry_void(); 1552 1553 ocfs2_quota_shutdown(); 1554 1555 if (ocfs2_wq) { 1556 flush_workqueue(ocfs2_wq); 1557 destroy_workqueue(ocfs2_wq); 1558 } 1559 1560 unregister_quota_format(&ocfs2_quota_format); 1561 1562 debugfs_remove(ocfs2_debugfs_root); 1563 1564 ocfs2_free_mem_caches(); 1565 1566 unregister_filesystem(&ocfs2_fs_type); 1567 1568 exit_ocfs2_uptodate_cache(); 1569 1570 mlog_exit_void(); 1571 } 1572 1573 static void ocfs2_put_super(struct super_block *sb) 1574 { 1575 mlog_entry("(0x%p)\n", sb); 1576 1577 lock_kernel(); 1578 1579 ocfs2_sync_blockdev(sb); 1580 ocfs2_dismount_volume(sb, 0); 1581 1582 unlock_kernel(); 1583 1584 mlog_exit_void(); 1585 } 1586 1587 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) 1588 { 1589 struct ocfs2_super *osb; 1590 u32 numbits, freebits; 1591 int status; 1592 struct ocfs2_dinode *bm_lock; 1593 struct buffer_head *bh = NULL; 1594 struct inode *inode = NULL; 1595 1596 mlog_entry("(%p, %p)\n", dentry->d_sb, buf); 1597 1598 osb = OCFS2_SB(dentry->d_sb); 1599 1600 inode = ocfs2_get_system_file_inode(osb, 1601 GLOBAL_BITMAP_SYSTEM_INODE, 1602 OCFS2_INVALID_SLOT); 1603 if (!inode) { 1604 mlog(ML_ERROR, "failed to get bitmap inode\n"); 1605 status = -EIO; 1606 goto bail; 1607 } 1608 1609 status = ocfs2_inode_lock(inode, &bh, 0); 1610 if (status < 0) { 1611 mlog_errno(status); 1612 goto bail; 1613 } 1614 1615 bm_lock = (struct ocfs2_dinode *) bh->b_data; 1616 1617 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); 1618 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); 1619 1620 buf->f_type = OCFS2_SUPER_MAGIC; 1621 buf->f_bsize = dentry->d_sb->s_blocksize; 1622 buf->f_namelen = OCFS2_MAX_FILENAME_LEN; 1623 buf->f_blocks = ((sector_t) numbits) * 1624 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1625 buf->f_bfree = ((sector_t) freebits) * 1626 (osb->s_clustersize >> osb->sb->s_blocksize_bits); 1627 buf->f_bavail = buf->f_bfree; 1628 buf->f_files = numbits; 1629 buf->f_ffree = freebits; 1630 1631 brelse(bh); 1632 1633 ocfs2_inode_unlock(inode, 0); 1634 status = 0; 1635 bail: 1636 if (inode) 1637 iput(inode); 1638 1639 mlog_exit(status); 1640 1641 return status; 1642 } 1643 1644 static void ocfs2_inode_init_once(void *data) 1645 { 1646 struct ocfs2_inode_info *oi = data; 1647 1648 oi->ip_flags = 0; 1649 oi->ip_open_count = 0; 1650 spin_lock_init(&oi->ip_lock); 1651 ocfs2_extent_map_init(&oi->vfs_inode); 1652 INIT_LIST_HEAD(&oi->ip_io_markers); 1653 oi->ip_created_trans = 0; 1654 oi->ip_last_trans = 0; 1655 oi->ip_dir_start_lookup = 0; 1656 1657 init_rwsem(&oi->ip_alloc_sem); 1658 init_rwsem(&oi->ip_xattr_sem); 1659 mutex_init(&oi->ip_io_mutex); 1660 1661 oi->ip_blkno = 0ULL; 1662 oi->ip_clusters = 0; 1663 1664 ocfs2_lock_res_init_once(&oi->ip_rw_lockres); 1665 ocfs2_lock_res_init_once(&oi->ip_inode_lockres); 1666 ocfs2_lock_res_init_once(&oi->ip_open_lockres); 1667 1668 ocfs2_metadata_cache_init(&oi->vfs_inode); 1669 1670 inode_init_once(&oi->vfs_inode); 1671 } 1672 1673 static int ocfs2_initialize_mem_caches(void) 1674 { 1675 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", 1676 sizeof(struct ocfs2_inode_info), 1677 0, 1678 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1679 SLAB_MEM_SPREAD), 1680 ocfs2_inode_init_once); 1681 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache", 1682 sizeof(struct ocfs2_dquot), 1683 0, 1684 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1685 SLAB_MEM_SPREAD), 1686 NULL); 1687 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache", 1688 sizeof(struct ocfs2_quota_chunk), 1689 0, 1690 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), 1691 NULL); 1692 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep || 1693 !ocfs2_qf_chunk_cachep) { 1694 if (ocfs2_inode_cachep) 1695 kmem_cache_destroy(ocfs2_inode_cachep); 1696 if (ocfs2_dquot_cachep) 1697 kmem_cache_destroy(ocfs2_dquot_cachep); 1698 if (ocfs2_qf_chunk_cachep) 1699 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1700 return -ENOMEM; 1701 } 1702 1703 return 0; 1704 } 1705 1706 static void ocfs2_free_mem_caches(void) 1707 { 1708 if (ocfs2_inode_cachep) 1709 kmem_cache_destroy(ocfs2_inode_cachep); 1710 ocfs2_inode_cachep = NULL; 1711 1712 if (ocfs2_dquot_cachep) 1713 kmem_cache_destroy(ocfs2_dquot_cachep); 1714 ocfs2_dquot_cachep = NULL; 1715 1716 if (ocfs2_qf_chunk_cachep) 1717 kmem_cache_destroy(ocfs2_qf_chunk_cachep); 1718 ocfs2_qf_chunk_cachep = NULL; 1719 } 1720 1721 static int ocfs2_get_sector(struct super_block *sb, 1722 struct buffer_head **bh, 1723 int block, 1724 int sect_size) 1725 { 1726 if (!sb_set_blocksize(sb, sect_size)) { 1727 mlog(ML_ERROR, "unable to set blocksize\n"); 1728 return -EIO; 1729 } 1730 1731 *bh = sb_getblk(sb, block); 1732 if (!*bh) { 1733 mlog_errno(-EIO); 1734 return -EIO; 1735 } 1736 lock_buffer(*bh); 1737 if (!buffer_dirty(*bh)) 1738 clear_buffer_uptodate(*bh); 1739 unlock_buffer(*bh); 1740 ll_rw_block(READ, 1, bh); 1741 wait_on_buffer(*bh); 1742 if (!buffer_uptodate(*bh)) { 1743 mlog_errno(-EIO); 1744 brelse(*bh); 1745 *bh = NULL; 1746 return -EIO; 1747 } 1748 1749 return 0; 1750 } 1751 1752 static int ocfs2_mount_volume(struct super_block *sb) 1753 { 1754 int status = 0; 1755 int unlock_super = 0; 1756 struct ocfs2_super *osb = OCFS2_SB(sb); 1757 1758 mlog_entry_void(); 1759 1760 if (ocfs2_is_hard_readonly(osb)) 1761 goto leave; 1762 1763 status = ocfs2_dlm_init(osb); 1764 if (status < 0) { 1765 mlog_errno(status); 1766 goto leave; 1767 } 1768 1769 status = ocfs2_super_lock(osb, 1); 1770 if (status < 0) { 1771 mlog_errno(status); 1772 goto leave; 1773 } 1774 unlock_super = 1; 1775 1776 /* This will load up the node map and add ourselves to it. */ 1777 status = ocfs2_find_slot(osb); 1778 if (status < 0) { 1779 mlog_errno(status); 1780 goto leave; 1781 } 1782 1783 /* load all node-local system inodes */ 1784 status = ocfs2_init_local_system_inodes(osb); 1785 if (status < 0) { 1786 mlog_errno(status); 1787 goto leave; 1788 } 1789 1790 status = ocfs2_check_volume(osb); 1791 if (status < 0) { 1792 mlog_errno(status); 1793 goto leave; 1794 } 1795 1796 status = ocfs2_truncate_log_init(osb); 1797 if (status < 0) 1798 mlog_errno(status); 1799 1800 leave: 1801 if (unlock_super) 1802 ocfs2_super_unlock(osb, 1); 1803 1804 mlog_exit(status); 1805 return status; 1806 } 1807 1808 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) 1809 { 1810 int tmp, hangup_needed = 0; 1811 struct ocfs2_super *osb = NULL; 1812 char nodestr[8]; 1813 1814 mlog_entry("(0x%p)\n", sb); 1815 1816 BUG_ON(!sb); 1817 osb = OCFS2_SB(sb); 1818 BUG_ON(!osb); 1819 1820 debugfs_remove(osb->osb_ctxt); 1821 1822 /* Orphan scan should be stopped as early as possible */ 1823 ocfs2_orphan_scan_stop(osb); 1824 1825 ocfs2_disable_quotas(osb); 1826 1827 ocfs2_shutdown_local_alloc(osb); 1828 1829 ocfs2_truncate_log_shutdown(osb); 1830 1831 /* This will disable recovery and flush any recovery work. */ 1832 ocfs2_recovery_exit(osb); 1833 1834 ocfs2_journal_shutdown(osb); 1835 1836 ocfs2_sync_blockdev(sb); 1837 1838 /* No cluster connection means we've failed during mount, so skip 1839 * all the steps which depended on that to complete. */ 1840 if (osb->cconn) { 1841 tmp = ocfs2_super_lock(osb, 1); 1842 if (tmp < 0) { 1843 mlog_errno(tmp); 1844 return; 1845 } 1846 } 1847 1848 if (osb->slot_num != OCFS2_INVALID_SLOT) 1849 ocfs2_put_slot(osb); 1850 1851 if (osb->cconn) 1852 ocfs2_super_unlock(osb, 1); 1853 1854 ocfs2_release_system_inodes(osb); 1855 1856 /* 1857 * If we're dismounting due to mount error, mount.ocfs2 will clean 1858 * up heartbeat. If we're a local mount, there is no heartbeat. 1859 * If we failed before we got a uuid_str yet, we can't stop 1860 * heartbeat. Otherwise, do it. 1861 */ 1862 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str) 1863 hangup_needed = 1; 1864 1865 if (osb->cconn) 1866 ocfs2_dlm_shutdown(osb, hangup_needed); 1867 1868 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats); 1869 debugfs_remove(osb->osb_debug_root); 1870 1871 if (hangup_needed) 1872 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str)); 1873 1874 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); 1875 1876 if (ocfs2_mount_local(osb)) 1877 snprintf(nodestr, sizeof(nodestr), "local"); 1878 else 1879 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); 1880 1881 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", 1882 osb->dev_str, nodestr); 1883 1884 ocfs2_delete_osb(osb); 1885 kfree(osb); 1886 sb->s_dev = 0; 1887 sb->s_fs_info = NULL; 1888 } 1889 1890 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, 1891 unsigned uuid_bytes) 1892 { 1893 int i, ret; 1894 char *ptr; 1895 1896 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); 1897 1898 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); 1899 if (osb->uuid_str == NULL) 1900 return -ENOMEM; 1901 1902 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { 1903 /* print with null */ 1904 ret = snprintf(ptr, 3, "%02X", uuid[i]); 1905 if (ret != 2) /* drop super cleans up */ 1906 return -EINVAL; 1907 /* then only advance past the last char */ 1908 ptr += 2; 1909 } 1910 1911 return 0; 1912 } 1913 1914 static int ocfs2_initialize_super(struct super_block *sb, 1915 struct buffer_head *bh, 1916 int sector_size, 1917 struct ocfs2_blockcheck_stats *stats) 1918 { 1919 int status; 1920 int i, cbits, bbits; 1921 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; 1922 struct inode *inode = NULL; 1923 struct ocfs2_journal *journal; 1924 __le32 uuid_net_key; 1925 struct ocfs2_super *osb; 1926 1927 mlog_entry_void(); 1928 1929 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); 1930 if (!osb) { 1931 status = -ENOMEM; 1932 mlog_errno(status); 1933 goto bail; 1934 } 1935 1936 sb->s_fs_info = osb; 1937 sb->s_op = &ocfs2_sops; 1938 sb->s_export_op = &ocfs2_export_ops; 1939 sb->s_qcop = &ocfs2_quotactl_ops; 1940 sb->dq_op = &ocfs2_quota_operations; 1941 sb->s_xattr = ocfs2_xattr_handlers; 1942 sb->s_time_gran = 1; 1943 sb->s_flags |= MS_NOATIME; 1944 /* this is needed to support O_LARGEFILE */ 1945 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); 1946 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); 1947 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); 1948 1949 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1; 1950 1951 for (i = 0; i < 3; i++) 1952 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]); 1953 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash); 1954 1955 osb->sb = sb; 1956 /* Save off for ocfs2_rw_direct */ 1957 osb->s_sectsize_bits = blksize_bits(sector_size); 1958 BUG_ON(!osb->s_sectsize_bits); 1959 1960 spin_lock_init(&osb->dc_task_lock); 1961 init_waitqueue_head(&osb->dc_event); 1962 osb->dc_work_sequence = 0; 1963 osb->dc_wake_sequence = 0; 1964 INIT_LIST_HEAD(&osb->blocked_lock_list); 1965 osb->blocked_lock_count = 0; 1966 spin_lock_init(&osb->osb_lock); 1967 spin_lock_init(&osb->osb_xattr_lock); 1968 ocfs2_init_inode_steal_slot(osb); 1969 1970 atomic_set(&osb->alloc_stats.moves, 0); 1971 atomic_set(&osb->alloc_stats.local_data, 0); 1972 atomic_set(&osb->alloc_stats.bitmap_data, 0); 1973 atomic_set(&osb->alloc_stats.bg_allocs, 0); 1974 atomic_set(&osb->alloc_stats.bg_extends, 0); 1975 1976 /* Copy the blockcheck stats from the superblock probe */ 1977 osb->osb_ecc_stats = *stats; 1978 1979 ocfs2_init_node_maps(osb); 1980 1981 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", 1982 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); 1983 1984 status = ocfs2_recovery_init(osb); 1985 if (status) { 1986 mlog(ML_ERROR, "Unable to initialize recovery state\n"); 1987 mlog_errno(status); 1988 goto bail; 1989 } 1990 1991 init_waitqueue_head(&osb->checkpoint_event); 1992 atomic_set(&osb->needs_checkpoint, 0); 1993 1994 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 1995 1996 osb->slot_num = OCFS2_INVALID_SLOT; 1997 1998 osb->s_xattr_inline_size = le16_to_cpu( 1999 di->id2.i_super.s_xattr_inline_size); 2000 2001 osb->local_alloc_state = OCFS2_LA_UNUSED; 2002 osb->local_alloc_bh = NULL; 2003 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker); 2004 2005 init_waitqueue_head(&osb->osb_mount_event); 2006 2007 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); 2008 if (!osb->vol_label) { 2009 mlog(ML_ERROR, "unable to alloc vol label\n"); 2010 status = -ENOMEM; 2011 goto bail; 2012 } 2013 2014 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); 2015 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { 2016 mlog(ML_ERROR, "Invalid number of node slots (%u)\n", 2017 osb->max_slots); 2018 status = -EINVAL; 2019 goto bail; 2020 } 2021 mlog(0, "max_slots for this device: %u\n", osb->max_slots); 2022 2023 osb->slot_recovery_generations = 2024 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), 2025 GFP_KERNEL); 2026 if (!osb->slot_recovery_generations) { 2027 status = -ENOMEM; 2028 mlog_errno(status); 2029 goto bail; 2030 } 2031 2032 init_waitqueue_head(&osb->osb_wipe_event); 2033 osb->osb_orphan_wipes = kcalloc(osb->max_slots, 2034 sizeof(*osb->osb_orphan_wipes), 2035 GFP_KERNEL); 2036 if (!osb->osb_orphan_wipes) { 2037 status = -ENOMEM; 2038 mlog_errno(status); 2039 goto bail; 2040 } 2041 2042 osb->s_feature_compat = 2043 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); 2044 osb->s_feature_ro_compat = 2045 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); 2046 osb->s_feature_incompat = 2047 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); 2048 2049 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { 2050 mlog(ML_ERROR, "couldn't mount because of unsupported " 2051 "optional features (%x).\n", i); 2052 status = -EINVAL; 2053 goto bail; 2054 } 2055 if (!(osb->sb->s_flags & MS_RDONLY) && 2056 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { 2057 mlog(ML_ERROR, "couldn't mount RDWR because of " 2058 "unsupported optional features (%x).\n", i); 2059 status = -EINVAL; 2060 goto bail; 2061 } 2062 2063 if (ocfs2_userspace_stack(osb)) { 2064 memcpy(osb->osb_cluster_stack, 2065 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, 2066 OCFS2_STACK_LABEL_LEN); 2067 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; 2068 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) { 2069 mlog(ML_ERROR, 2070 "couldn't mount because of an invalid " 2071 "cluster stack label (%s) \n", 2072 osb->osb_cluster_stack); 2073 status = -EINVAL; 2074 goto bail; 2075 } 2076 } else { 2077 /* The empty string is identical with classic tools that 2078 * don't know about s_cluster_info. */ 2079 osb->osb_cluster_stack[0] = '\0'; 2080 } 2081 2082 get_random_bytes(&osb->s_next_generation, sizeof(u32)); 2083 2084 /* FIXME 2085 * This should be done in ocfs2_journal_init(), but unknown 2086 * ordering issues will cause the filesystem to crash. 2087 * If anyone wants to figure out what part of the code 2088 * refers to osb->journal before ocfs2_journal_init() is run, 2089 * be my guest. 2090 */ 2091 /* initialize our journal structure */ 2092 2093 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL); 2094 if (!journal) { 2095 mlog(ML_ERROR, "unable to alloc journal\n"); 2096 status = -ENOMEM; 2097 goto bail; 2098 } 2099 osb->journal = journal; 2100 journal->j_osb = osb; 2101 2102 atomic_set(&journal->j_num_trans, 0); 2103 init_rwsem(&journal->j_trans_barrier); 2104 init_waitqueue_head(&journal->j_checkpointed); 2105 spin_lock_init(&journal->j_lock); 2106 journal->j_trans_id = (unsigned long) 1; 2107 INIT_LIST_HEAD(&journal->j_la_cleanups); 2108 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); 2109 journal->j_state = OCFS2_JOURNAL_FREE; 2110 2111 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes); 2112 osb->dentry_lock_list = NULL; 2113 2114 /* get some pseudo constants for clustersize bits */ 2115 osb->s_clustersize_bits = 2116 le32_to_cpu(di->id2.i_super.s_clustersize_bits); 2117 osb->s_clustersize = 1 << osb->s_clustersize_bits; 2118 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits); 2119 2120 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || 2121 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { 2122 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", 2123 osb->s_clustersize); 2124 status = -EINVAL; 2125 goto bail; 2126 } 2127 2128 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1) 2129 > (u32)~0UL) { 2130 mlog(ML_ERROR, "Volume might try to write to blocks beyond " 2131 "what jbd can address in 32 bits.\n"); 2132 status = -EINVAL; 2133 goto bail; 2134 } 2135 2136 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, 2137 sizeof(di->id2.i_super.s_uuid))) { 2138 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); 2139 status = -ENOMEM; 2140 goto bail; 2141 } 2142 2143 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key)); 2144 2145 strncpy(osb->vol_label, di->id2.i_super.s_label, 63); 2146 osb->vol_label[63] = '\0'; 2147 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); 2148 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); 2149 osb->first_cluster_group_blkno = 2150 le64_to_cpu(di->id2.i_super.s_first_cluster_group); 2151 osb->fs_generation = le32_to_cpu(di->i_fs_generation); 2152 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash); 2153 mlog(0, "vol_label: %s\n", osb->vol_label); 2154 mlog(0, "uuid: %s\n", osb->uuid_str); 2155 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n", 2156 (unsigned long long)osb->root_blkno, 2157 (unsigned long long)osb->system_dir_blkno); 2158 2159 osb->osb_dlm_debug = ocfs2_new_dlm_debug(); 2160 if (!osb->osb_dlm_debug) { 2161 status = -ENOMEM; 2162 mlog_errno(status); 2163 goto bail; 2164 } 2165 2166 atomic_set(&osb->vol_state, VOLUME_INIT); 2167 2168 /* load root, system_dir, and all global system inodes */ 2169 status = ocfs2_init_global_system_inodes(osb); 2170 if (status < 0) { 2171 mlog_errno(status); 2172 goto bail; 2173 } 2174 2175 /* 2176 * global bitmap 2177 */ 2178 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, 2179 OCFS2_INVALID_SLOT); 2180 if (!inode) { 2181 status = -EINVAL; 2182 mlog_errno(status); 2183 goto bail; 2184 } 2185 2186 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; 2187 iput(inode); 2188 2189 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8; 2190 2191 status = ocfs2_init_slot_info(osb); 2192 if (status < 0) { 2193 mlog_errno(status); 2194 goto bail; 2195 } 2196 2197 bail: 2198 mlog_exit(status); 2199 return status; 2200 } 2201 2202 /* 2203 * will return: -EAGAIN if it is ok to keep searching for superblocks 2204 * -EINVAL if there is a bad superblock 2205 * 0 on success 2206 */ 2207 static int ocfs2_verify_volume(struct ocfs2_dinode *di, 2208 struct buffer_head *bh, 2209 u32 blksz, 2210 struct ocfs2_blockcheck_stats *stats) 2211 { 2212 int status = -EAGAIN; 2213 2214 mlog_entry_void(); 2215 2216 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, 2217 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { 2218 /* We have to do a raw check of the feature here */ 2219 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) & 2220 OCFS2_FEATURE_INCOMPAT_META_ECC) { 2221 status = ocfs2_block_check_validate(bh->b_data, 2222 bh->b_size, 2223 &di->i_check, 2224 stats); 2225 if (status) 2226 goto out; 2227 } 2228 status = -EINVAL; 2229 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) { 2230 mlog(ML_ERROR, "found superblock with incorrect block " 2231 "size: found %u, should be %u\n", 2232 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits), 2233 blksz); 2234 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != 2235 OCFS2_MAJOR_REV_LEVEL || 2236 le16_to_cpu(di->id2.i_super.s_minor_rev_level) != 2237 OCFS2_MINOR_REV_LEVEL) { 2238 mlog(ML_ERROR, "found superblock with bad version: " 2239 "found %u.%u, should be %u.%u\n", 2240 le16_to_cpu(di->id2.i_super.s_major_rev_level), 2241 le16_to_cpu(di->id2.i_super.s_minor_rev_level), 2242 OCFS2_MAJOR_REV_LEVEL, 2243 OCFS2_MINOR_REV_LEVEL); 2244 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { 2245 mlog(ML_ERROR, "bad block number on superblock: " 2246 "found %llu, should be %llu\n", 2247 (unsigned long long)le64_to_cpu(di->i_blkno), 2248 (unsigned long long)bh->b_blocknr); 2249 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || 2250 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { 2251 mlog(ML_ERROR, "bad cluster size found: %u\n", 2252 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); 2253 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { 2254 mlog(ML_ERROR, "bad root_blkno: 0\n"); 2255 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { 2256 mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); 2257 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { 2258 mlog(ML_ERROR, 2259 "Superblock slots found greater than file system " 2260 "maximum: found %u, max %u\n", 2261 le16_to_cpu(di->id2.i_super.s_max_slots), 2262 OCFS2_MAX_SLOTS); 2263 } else { 2264 /* found it! */ 2265 status = 0; 2266 } 2267 } 2268 2269 out: 2270 mlog_exit(status); 2271 return status; 2272 } 2273 2274 static int ocfs2_check_volume(struct ocfs2_super *osb) 2275 { 2276 int status; 2277 int dirty; 2278 int local; 2279 struct ocfs2_dinode *local_alloc = NULL; /* only used if we 2280 * recover 2281 * ourselves. */ 2282 2283 mlog_entry_void(); 2284 2285 /* Init our journal object. */ 2286 status = ocfs2_journal_init(osb->journal, &dirty); 2287 if (status < 0) { 2288 mlog(ML_ERROR, "Could not initialize journal!\n"); 2289 goto finally; 2290 } 2291 2292 /* If the journal was unmounted cleanly then we don't want to 2293 * recover anything. Otherwise, journal_load will do that 2294 * dirty work for us :) */ 2295 if (!dirty) { 2296 status = ocfs2_journal_wipe(osb->journal, 0); 2297 if (status < 0) { 2298 mlog_errno(status); 2299 goto finally; 2300 } 2301 } else { 2302 mlog(ML_NOTICE, "File system was not unmounted cleanly, " 2303 "recovering volume.\n"); 2304 } 2305 2306 local = ocfs2_mount_local(osb); 2307 2308 /* will play back anything left in the journal. */ 2309 status = ocfs2_journal_load(osb->journal, local, dirty); 2310 if (status < 0) { 2311 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status); 2312 goto finally; 2313 } 2314 2315 if (dirty) { 2316 /* recover my local alloc if we didn't unmount cleanly. */ 2317 status = ocfs2_begin_local_alloc_recovery(osb, 2318 osb->slot_num, 2319 &local_alloc); 2320 if (status < 0) { 2321 mlog_errno(status); 2322 goto finally; 2323 } 2324 /* we complete the recovery process after we've marked 2325 * ourselves as mounted. */ 2326 } 2327 2328 mlog(0, "Journal loaded.\n"); 2329 2330 status = ocfs2_load_local_alloc(osb); 2331 if (status < 0) { 2332 mlog_errno(status); 2333 goto finally; 2334 } 2335 2336 if (dirty) { 2337 /* Recovery will be completed after we've mounted the 2338 * rest of the volume. */ 2339 osb->dirty = 1; 2340 osb->local_alloc_copy = local_alloc; 2341 local_alloc = NULL; 2342 } 2343 2344 /* go through each journal, trylock it and if you get the 2345 * lock, and it's marked as dirty, set the bit in the recover 2346 * map and launch a recovery thread for it. */ 2347 status = ocfs2_mark_dead_nodes(osb); 2348 if (status < 0) { 2349 mlog_errno(status); 2350 goto finally; 2351 } 2352 2353 status = ocfs2_compute_replay_slots(osb); 2354 if (status < 0) 2355 mlog_errno(status); 2356 2357 finally: 2358 if (local_alloc) 2359 kfree(local_alloc); 2360 2361 mlog_exit(status); 2362 return status; 2363 } 2364 2365 /* 2366 * The routine gets called from dismount or close whenever a dismount on 2367 * volume is requested and the osb open count becomes 1. 2368 * It will remove the osb from the global list and also free up all the 2369 * initialized resources and fileobject. 2370 */ 2371 static void ocfs2_delete_osb(struct ocfs2_super *osb) 2372 { 2373 mlog_entry_void(); 2374 2375 /* This function assumes that the caller has the main osb resource */ 2376 2377 ocfs2_free_slot_info(osb); 2378 2379 kfree(osb->osb_orphan_wipes); 2380 kfree(osb->slot_recovery_generations); 2381 /* FIXME 2382 * This belongs in journal shutdown, but because we have to 2383 * allocate osb->journal at the start of ocfs2_initalize_osb(), 2384 * we free it here. 2385 */ 2386 kfree(osb->journal); 2387 if (osb->local_alloc_copy) 2388 kfree(osb->local_alloc_copy); 2389 kfree(osb->uuid_str); 2390 ocfs2_put_dlm_debug(osb->osb_dlm_debug); 2391 memset(osb, 0, sizeof(struct ocfs2_super)); 2392 2393 mlog_exit_void(); 2394 } 2395 2396 /* Put OCFS2 into a readonly state, or (if the user specifies it), 2397 * panic(). We do not support continue-on-error operation. */ 2398 static void ocfs2_handle_error(struct super_block *sb) 2399 { 2400 struct ocfs2_super *osb = OCFS2_SB(sb); 2401 2402 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) 2403 panic("OCFS2: (device %s): panic forced after error\n", 2404 sb->s_id); 2405 2406 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); 2407 2408 if (sb->s_flags & MS_RDONLY && 2409 (ocfs2_is_soft_readonly(osb) || 2410 ocfs2_is_hard_readonly(osb))) 2411 return; 2412 2413 printk(KERN_CRIT "File system is now read-only due to the potential " 2414 "of on-disk corruption. Please run fsck.ocfs2 once the file " 2415 "system is unmounted.\n"); 2416 sb->s_flags |= MS_RDONLY; 2417 ocfs2_set_ro_flag(osb, 0); 2418 } 2419 2420 static char error_buf[1024]; 2421 2422 void __ocfs2_error(struct super_block *sb, 2423 const char *function, 2424 const char *fmt, ...) 2425 { 2426 va_list args; 2427 2428 va_start(args, fmt); 2429 vsnprintf(error_buf, sizeof(error_buf), fmt, args); 2430 va_end(args); 2431 2432 /* Not using mlog here because we want to show the actual 2433 * function the error came from. */ 2434 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n", 2435 sb->s_id, function, error_buf); 2436 2437 ocfs2_handle_error(sb); 2438 } 2439 2440 /* Handle critical errors. This is intentionally more drastic than 2441 * ocfs2_handle_error, so we only use for things like journal errors, 2442 * etc. */ 2443 void __ocfs2_abort(struct super_block* sb, 2444 const char *function, 2445 const char *fmt, ...) 2446 { 2447 va_list args; 2448 2449 va_start(args, fmt); 2450 vsnprintf(error_buf, sizeof(error_buf), fmt, args); 2451 va_end(args); 2452 2453 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n", 2454 sb->s_id, function, error_buf); 2455 2456 /* We don't have the cluster support yet to go straight to 2457 * hard readonly in here. Until then, we want to keep 2458 * ocfs2_abort() so that we can at least mark critical 2459 * errors. 2460 * 2461 * TODO: This should abort the journal and alert other nodes 2462 * that our slot needs recovery. */ 2463 2464 /* Force a panic(). This stinks, but it's better than letting 2465 * things continue without having a proper hard readonly 2466 * here. */ 2467 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 2468 ocfs2_handle_error(sb); 2469 } 2470 2471 module_init(ocfs2_init); 2472 module_exit(ocfs2_exit); 2473