Linux Audio

Check our new training course

Loading...
v6.2
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
 
   3 * super.c
   4 *
   5 * load/unload driver, mount/dismount volumes
   6 *
   7 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
   8 */
   9
  10#include <linux/module.h>
  11#include <linux/fs.h>
  12#include <linux/types.h>
  13#include <linux/slab.h>
  14#include <linux/highmem.h>
  15#include <linux/init.h>
  16#include <linux/random.h>
  17#include <linux/statfs.h>
  18#include <linux/moduleparam.h>
  19#include <linux/blkdev.h>
  20#include <linux/socket.h>
  21#include <linux/inet.h>
  22#include <linux/parser.h>
  23#include <linux/crc32.h>
  24#include <linux/debugfs.h>
  25#include <linux/mount.h>
  26#include <linux/seq_file.h>
  27#include <linux/quotaops.h>
  28#include <linux/signal.h>
  29
  30#define CREATE_TRACE_POINTS
  31#include "ocfs2_trace.h"
  32
  33#include <cluster/masklog.h>
  34
  35#include "ocfs2.h"
  36
  37/* this should be the only file to include a version 1 header */
  38#include "ocfs1_fs_compat.h"
  39
  40#include "alloc.h"
  41#include "aops.h"
  42#include "blockcheck.h"
  43#include "dlmglue.h"
  44#include "export.h"
  45#include "extent_map.h"
  46#include "heartbeat.h"
  47#include "inode.h"
  48#include "journal.h"
  49#include "localalloc.h"
  50#include "namei.h"
  51#include "slot_map.h"
  52#include "super.h"
  53#include "sysfile.h"
  54#include "uptodate.h"
  55#include "xattr.h"
  56#include "quota.h"
  57#include "refcounttree.h"
  58#include "suballoc.h"
  59
  60#include "buffer_head_io.h"
  61#include "filecheck.h"
  62
  63static struct kmem_cache *ocfs2_inode_cachep;
  64struct kmem_cache *ocfs2_dquot_cachep;
  65struct kmem_cache *ocfs2_qf_chunk_cachep;
  66
  67static struct dentry *ocfs2_debugfs_root;
 
 
 
 
 
 
  68
  69MODULE_AUTHOR("Oracle");
  70MODULE_LICENSE("GPL");
  71MODULE_DESCRIPTION("OCFS2 cluster file system");
  72
  73struct mount_options
  74{
  75	unsigned long	commit_interval;
  76	unsigned long	mount_opt;
  77	unsigned int	atime_quantum;
  78	unsigned short	slot;
  79	int		localalloc_opt;
  80	unsigned int	resv_level;
  81	int		dir_resv_level;
  82	char		cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
  83};
  84
  85static int ocfs2_parse_options(struct super_block *sb, char *options,
  86			       struct mount_options *mopt,
  87			       int is_remount);
  88static int ocfs2_check_set_options(struct super_block *sb,
  89				   struct mount_options *options);
  90static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
  91static void ocfs2_put_super(struct super_block *sb);
  92static int ocfs2_mount_volume(struct super_block *sb);
  93static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
  94static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
  95static int ocfs2_initialize_mem_caches(void);
  96static void ocfs2_free_mem_caches(void);
  97static void ocfs2_delete_osb(struct ocfs2_super *osb);
  98
  99static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 100
 101static int ocfs2_sync_fs(struct super_block *sb, int wait);
 102
 103static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 104static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 105static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 106static int ocfs2_check_volume(struct ocfs2_super *osb);
 107static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 108			       struct buffer_head *bh,
 109			       u32 sectsize,
 110			       struct ocfs2_blockcheck_stats *stats);
 111static int ocfs2_initialize_super(struct super_block *sb,
 112				  struct buffer_head *bh,
 113				  int sector_size,
 114				  struct ocfs2_blockcheck_stats *stats);
 115static int ocfs2_get_sector(struct super_block *sb,
 116			    struct buffer_head **bh,
 117			    int block,
 118			    int sect_size);
 119static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 120static void ocfs2_free_inode(struct inode *inode);
 121static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 122static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 123static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 124
 125static struct dquot **ocfs2_get_dquots(struct inode *inode)
 126{
 127	return OCFS2_I(inode)->i_dquot;
 128}
 129
 130static const struct super_operations ocfs2_sops = {
 131	.statfs		= ocfs2_statfs,
 132	.alloc_inode	= ocfs2_alloc_inode,
 133	.free_inode	= ocfs2_free_inode,
 134	.drop_inode	= ocfs2_drop_inode,
 135	.evict_inode	= ocfs2_evict_inode,
 136	.sync_fs	= ocfs2_sync_fs,
 137	.put_super	= ocfs2_put_super,
 138	.remount_fs	= ocfs2_remount,
 139	.show_options   = ocfs2_show_options,
 140	.quota_read	= ocfs2_quota_read,
 141	.quota_write	= ocfs2_quota_write,
 142	.get_dquots	= ocfs2_get_dquots,
 143};
 144
 145enum {
 146	Opt_barrier,
 147	Opt_err_panic,
 148	Opt_err_ro,
 149	Opt_intr,
 150	Opt_nointr,
 151	Opt_hb_none,
 152	Opt_hb_local,
 153	Opt_hb_global,
 154	Opt_data_ordered,
 155	Opt_data_writeback,
 156	Opt_atime_quantum,
 157	Opt_slot,
 158	Opt_commit,
 159	Opt_localalloc,
 160	Opt_localflocks,
 161	Opt_stack,
 162	Opt_user_xattr,
 163	Opt_nouser_xattr,
 164	Opt_inode64,
 165	Opt_acl,
 166	Opt_noacl,
 167	Opt_usrquota,
 168	Opt_grpquota,
 169	Opt_coherency_buffered,
 170	Opt_coherency_full,
 171	Opt_resv_level,
 172	Opt_dir_resv_level,
 173	Opt_journal_async_commit,
 174	Opt_err_cont,
 175	Opt_err,
 176};
 177
 178static const match_table_t tokens = {
 179	{Opt_barrier, "barrier=%u"},
 180	{Opt_err_panic, "errors=panic"},
 181	{Opt_err_ro, "errors=remount-ro"},
 182	{Opt_intr, "intr"},
 183	{Opt_nointr, "nointr"},
 184	{Opt_hb_none, OCFS2_HB_NONE},
 185	{Opt_hb_local, OCFS2_HB_LOCAL},
 186	{Opt_hb_global, OCFS2_HB_GLOBAL},
 187	{Opt_data_ordered, "data=ordered"},
 188	{Opt_data_writeback, "data=writeback"},
 189	{Opt_atime_quantum, "atime_quantum=%u"},
 190	{Opt_slot, "preferred_slot=%u"},
 191	{Opt_commit, "commit=%u"},
 192	{Opt_localalloc, "localalloc=%d"},
 193	{Opt_localflocks, "localflocks"},
 194	{Opt_stack, "cluster_stack=%s"},
 195	{Opt_user_xattr, "user_xattr"},
 196	{Opt_nouser_xattr, "nouser_xattr"},
 197	{Opt_inode64, "inode64"},
 198	{Opt_acl, "acl"},
 199	{Opt_noacl, "noacl"},
 200	{Opt_usrquota, "usrquota"},
 201	{Opt_grpquota, "grpquota"},
 202	{Opt_coherency_buffered, "coherency=buffered"},
 203	{Opt_coherency_full, "coherency=full"},
 204	{Opt_resv_level, "resv_level=%u"},
 205	{Opt_dir_resv_level, "dir_resv_level=%u"},
 206	{Opt_journal_async_commit, "journal_async_commit"},
 207	{Opt_err_cont, "errors=continue"},
 208	{Opt_err, NULL}
 209};
 210
 211#ifdef CONFIG_DEBUG_FS
 212static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 213{
 214	struct ocfs2_cluster_connection *cconn = osb->cconn;
 215	struct ocfs2_recovery_map *rm = osb->recovery_map;
 216	struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 217	int i, out = 0;
 218	unsigned long flags;
 219
 220	out += scnprintf(buf + out, len - out,
 221			"%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 222			"Device", osb->dev_str, osb->uuid_str,
 223			osb->fs_generation, osb->vol_label);
 224
 225	out += scnprintf(buf + out, len - out,
 226			"%10s => State: %d  Flags: 0x%lX\n", "Volume",
 227			atomic_read(&osb->vol_state), osb->osb_flags);
 228
 229	out += scnprintf(buf + out, len - out,
 230			"%10s => Block: %lu  Cluster: %d\n", "Sizes",
 231			osb->sb->s_blocksize, osb->s_clustersize);
 232
 233	out += scnprintf(buf + out, len - out,
 234			"%10s => Compat: 0x%X  Incompat: 0x%X  "
 235			"ROcompat: 0x%X\n",
 236			"Features", osb->s_feature_compat,
 237			osb->s_feature_incompat, osb->s_feature_ro_compat);
 238
 239	out += scnprintf(buf + out, len - out,
 240			"%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 241			osb->s_mount_opt, osb->s_atime_quantum);
 242
 243	if (cconn) {
 244		out += scnprintf(buf + out, len - out,
 245				"%10s => Stack: %s  Name: %*s  "
 246				"Version: %d.%d\n", "Cluster",
 247				(*osb->osb_cluster_stack == '\0' ?
 248				 "o2cb" : osb->osb_cluster_stack),
 249				cconn->cc_namelen, cconn->cc_name,
 250				cconn->cc_version.pv_major,
 251				cconn->cc_version.pv_minor);
 252	}
 253
 254	spin_lock_irqsave(&osb->dc_task_lock, flags);
 255	out += scnprintf(buf + out, len - out,
 256			"%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 257			"WorkSeq: %lu\n", "DownCnvt",
 258			(osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 259			osb->blocked_lock_count, osb->dc_wake_sequence,
 260			osb->dc_work_sequence);
 261	spin_unlock_irqrestore(&osb->dc_task_lock, flags);
 262
 263	spin_lock(&osb->osb_lock);
 264	out += scnprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 265			"Recovery",
 266			(osb->recovery_thread_task ?
 267			 task_pid_nr(osb->recovery_thread_task) : -1));
 268	if (rm->rm_used == 0)
 269		out += scnprintf(buf + out, len - out, " None\n");
 270	else {
 271		for (i = 0; i < rm->rm_used; i++)
 272			out += scnprintf(buf + out, len - out, " %d",
 273					rm->rm_entries[i]);
 274		out += scnprintf(buf + out, len - out, "\n");
 275	}
 276	spin_unlock(&osb->osb_lock);
 277
 278	out += scnprintf(buf + out, len - out,
 279			"%10s => Pid: %d  Interval: %lu\n", "Commit",
 280			(osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 281			osb->osb_commit_interval);
 282
 283	out += scnprintf(buf + out, len - out,
 284			"%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 285			"Journal", osb->journal->j_state,
 286			osb->journal->j_trans_id,
 287			atomic_read(&osb->journal->j_num_trans));
 288
 289	out += scnprintf(buf + out, len - out,
 290			"%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 291			"SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 292			"Stats",
 293			atomic_read(&osb->alloc_stats.bitmap_data),
 294			atomic_read(&osb->alloc_stats.local_data),
 295			atomic_read(&osb->alloc_stats.bg_allocs),
 296			atomic_read(&osb->alloc_stats.moves),
 297			atomic_read(&osb->alloc_stats.bg_extends));
 298
 299	out += scnprintf(buf + out, len - out,
 300			"%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 301			"Default: %u bits\n",
 302			"LocalAlloc", osb->local_alloc_state,
 303			(unsigned long long)osb->la_last_gd,
 304			osb->local_alloc_bits, osb->local_alloc_default_bits);
 305
 306	spin_lock(&osb->osb_lock);
 307	out += scnprintf(buf + out, len - out,
 308			"%10s => InodeSlot: %d  StolenInodes: %d, "
 309			"MetaSlot: %d  StolenMeta: %d\n", "Steal",
 310			osb->s_inode_steal_slot,
 311			atomic_read(&osb->s_num_inodes_stolen),
 312			osb->s_meta_steal_slot,
 313			atomic_read(&osb->s_num_meta_stolen));
 314	spin_unlock(&osb->osb_lock);
 315
 316	out += scnprintf(buf + out, len - out, "OrphanScan => ");
 317	out += scnprintf(buf + out, len - out, "Local: %u  Global: %u ",
 318			os->os_count, os->os_seqno);
 319	out += scnprintf(buf + out, len - out, " Last Scan: ");
 320	if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 321		out += scnprintf(buf + out, len - out, "Disabled\n");
 322	else
 323		out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
 324				(unsigned long)(ktime_get_seconds() - os->os_scantime));
 325
 326	out += scnprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 327			"Slots", "Num", "RecoGen");
 328	for (i = 0; i < osb->max_slots; ++i) {
 329		out += scnprintf(buf + out, len - out,
 330				"%10s  %c %3d  %10d\n",
 331				" ",
 332				(i == osb->slot_num ? '*' : ' '),
 333				i, osb->slot_recovery_generations[i]);
 334	}
 335
 336	return out;
 337}
 338
 339static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 340{
 341	struct ocfs2_super *osb = inode->i_private;
 342	char *buf = NULL;
 343
 344	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 345	if (!buf)
 346		goto bail;
 347
 348	i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 349
 350	file->private_data = buf;
 351
 352	return 0;
 353bail:
 354	return -ENOMEM;
 355}
 356
 357static int ocfs2_debug_release(struct inode *inode, struct file *file)
 358{
 359	kfree(file->private_data);
 360	return 0;
 361}
 362
 363static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 364				size_t nbytes, loff_t *ppos)
 365{
 366	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 367				       i_size_read(file->f_mapping->host));
 368}
 369#else
 370static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 371{
 372	return 0;
 373}
 374static int ocfs2_debug_release(struct inode *inode, struct file *file)
 375{
 376	return 0;
 377}
 378static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 379				size_t nbytes, loff_t *ppos)
 380{
 381	return 0;
 382}
 383#endif	/* CONFIG_DEBUG_FS */
 384
 385static const struct file_operations ocfs2_osb_debug_fops = {
 386	.open =		ocfs2_osb_debug_open,
 387	.release =	ocfs2_debug_release,
 388	.read =		ocfs2_debug_read,
 389	.llseek =	generic_file_llseek,
 390};
 391
 392static int ocfs2_sync_fs(struct super_block *sb, int wait)
 393{
 394	int status;
 395	tid_t target;
 396	struct ocfs2_super *osb = OCFS2_SB(sb);
 397
 398	if (ocfs2_is_hard_readonly(osb))
 399		return -EROFS;
 400
 401	if (wait) {
 402		status = ocfs2_flush_truncate_log(osb);
 403		if (status < 0)
 404			mlog_errno(status);
 405	} else {
 406		ocfs2_schedule_truncate_log_flush(osb, 0);
 407	}
 408
 409	if (jbd2_journal_start_commit(osb->journal->j_journal,
 410				      &target)) {
 411		if (wait)
 412			jbd2_log_wait_commit(osb->journal->j_journal,
 413					     target);
 414	}
 415	return 0;
 416}
 417
 418static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 419{
 420	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 421	    && (ino == USER_QUOTA_SYSTEM_INODE
 422		|| ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 423		return 0;
 424	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 425	    && (ino == GROUP_QUOTA_SYSTEM_INODE
 426		|| ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 427		return 0;
 428	return 1;
 429}
 430
 431static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 432{
 433	struct inode *new = NULL;
 434	int status = 0;
 435	int i;
 436
 437	new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 438	if (IS_ERR(new)) {
 439		status = PTR_ERR(new);
 440		mlog_errno(status);
 441		goto bail;
 442	}
 443	osb->root_inode = new;
 444
 445	new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 446	if (IS_ERR(new)) {
 447		status = PTR_ERR(new);
 448		mlog_errno(status);
 449		goto bail;
 450	}
 451	osb->sys_root_inode = new;
 452
 453	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 454	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 455		if (!ocfs2_need_system_inode(osb, i))
 456			continue;
 457		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 458		if (!new) {
 459			ocfs2_release_system_inodes(osb);
 460			status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 461			mlog_errno(status);
 
 462			mlog(ML_ERROR, "Unable to load system inode %d, "
 463			     "possibly corrupt fs?", i);
 464			goto bail;
 465		}
 466		// the array now has one ref, so drop this one
 467		iput(new);
 468	}
 469
 470bail:
 471	if (status)
 472		mlog_errno(status);
 473	return status;
 474}
 475
 476static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 477{
 478	struct inode *new = NULL;
 479	int status = 0;
 480	int i;
 481
 482	for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 483	     i < NUM_SYSTEM_INODES;
 484	     i++) {
 485		if (!ocfs2_need_system_inode(osb, i))
 486			continue;
 487		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 488		if (!new) {
 489			ocfs2_release_system_inodes(osb);
 490			status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 491			mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 492			     status, i, osb->slot_num);
 493			goto bail;
 494		}
 495		/* the array now has one ref, so drop this one */
 496		iput(new);
 497	}
 498
 499bail:
 500	if (status)
 501		mlog_errno(status);
 502	return status;
 503}
 504
 505static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 506{
 507	int i;
 508	struct inode *inode;
 509
 510	for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 511		inode = osb->global_system_inodes[i];
 512		if (inode) {
 513			iput(inode);
 514			osb->global_system_inodes[i] = NULL;
 515		}
 516	}
 517
 518	inode = osb->sys_root_inode;
 519	if (inode) {
 520		iput(inode);
 521		osb->sys_root_inode = NULL;
 522	}
 523
 524	inode = osb->root_inode;
 525	if (inode) {
 526		iput(inode);
 527		osb->root_inode = NULL;
 528	}
 529
 530	if (!osb->local_system_inodes)
 531		return;
 532
 533	for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 534		if (osb->local_system_inodes[i]) {
 535			iput(osb->local_system_inodes[i]);
 536			osb->local_system_inodes[i] = NULL;
 537		}
 538	}
 539
 540	kfree(osb->local_system_inodes);
 541	osb->local_system_inodes = NULL;
 542}
 543
 544/* We're allocating fs objects, use GFP_NOFS */
 545static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 546{
 547	struct ocfs2_inode_info *oi;
 548
 549	oi = alloc_inode_sb(sb, ocfs2_inode_cachep, GFP_NOFS);
 550	if (!oi)
 551		return NULL;
 552
 553	oi->i_sync_tid = 0;
 554	oi->i_datasync_tid = 0;
 555	memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
 556
 557	jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 558	return &oi->vfs_inode;
 559}
 560
 561static void ocfs2_free_inode(struct inode *inode)
 562{
 
 563	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 564}
 565
 
 
 
 
 
 566static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 567						unsigned int cbits)
 568{
 569	unsigned int bytes = 1 << cbits;
 570	unsigned int trim = bytes;
 571	unsigned int bitshift = 32;
 572
 573	/*
 574	 * i_size and all block offsets in ocfs2 are always 64 bits
 575	 * wide. i_clusters is 32 bits, in cluster-sized units. So on
 576	 * 64 bit platforms, cluster size will be the limiting factor.
 577	 */
 578
 579#if BITS_PER_LONG == 32
 
 580	BUILD_BUG_ON(sizeof(sector_t) != 8);
 581	/*
 582	 * We might be limited by page cache size.
 583	 */
 584	if (bytes > PAGE_SIZE) {
 585		bytes = PAGE_SIZE;
 586		trim = 1;
 587		/*
 588		 * Shift by 31 here so that we don't get larger than
 589		 * MAX_LFS_FILESIZE
 590		 */
 591		bitshift = 31;
 592	}
 
 
 
 
 
 
 
 
 
 593#endif
 594
 595	/*
 596	 * Trim by a whole cluster when we can actually approach the
 597	 * on-disk limits. Otherwise we can overflow i_clusters when
 598	 * an extent start is at the max offset.
 599	 */
 600	return (((unsigned long long)bytes) << bitshift) - trim;
 601}
 602
 603static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 604{
 605	int incompat_features;
 606	int ret = 0;
 607	struct mount_options parsed_options;
 608	struct ocfs2_super *osb = OCFS2_SB(sb);
 609	u32 tmp;
 610
 611	sync_filesystem(sb);
 612
 613	if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 614	    !ocfs2_check_set_options(sb, &parsed_options)) {
 615		ret = -EINVAL;
 616		goto out;
 617	}
 618
 619	tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 620		OCFS2_MOUNT_HB_NONE;
 621	if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 622		ret = -EINVAL;
 623		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 624		goto out;
 625	}
 626
 627	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 628	    (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 629		ret = -EINVAL;
 630		mlog(ML_ERROR, "Cannot change data mode on remount\n");
 631		goto out;
 632	}
 633
 634	/* Probably don't want this on remount; it might
 635	 * mess with other nodes */
 636	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 637	    (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 638		ret = -EINVAL;
 639		mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 640		goto out;
 641	}
 642
 643	/* We're going to/from readonly mode. */
 644	if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
 645		/* Disable quota accounting before remounting RO */
 646		if (*flags & SB_RDONLY) {
 647			ret = ocfs2_susp_quotas(osb, 0);
 648			if (ret < 0)
 649				goto out;
 650		}
 651		/* Lock here so the check of HARD_RO and the potential
 652		 * setting of SOFT_RO is atomic. */
 653		spin_lock(&osb->osb_lock);
 654		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 655			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 656			ret = -EROFS;
 657			goto unlock_osb;
 658		}
 659
 660		if (*flags & SB_RDONLY) {
 661			sb->s_flags |= SB_RDONLY;
 662			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 663		} else {
 664			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 665				mlog(ML_ERROR, "Cannot remount RDWR "
 666				     "filesystem due to previous errors.\n");
 667				ret = -EROFS;
 668				goto unlock_osb;
 669			}
 670			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 671			if (incompat_features) {
 672				mlog(ML_ERROR, "Cannot remount RDWR because "
 673				     "of unsupported optional features "
 674				     "(%x).\n", incompat_features);
 675				ret = -EINVAL;
 676				goto unlock_osb;
 677			}
 678			sb->s_flags &= ~SB_RDONLY;
 679			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 680		}
 681		trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
 682unlock_osb:
 683		spin_unlock(&osb->osb_lock);
 684		/* Enable quota accounting after remounting RW */
 685		if (!ret && !(*flags & SB_RDONLY)) {
 686			if (sb_any_quota_suspended(sb))
 687				ret = ocfs2_susp_quotas(osb, 1);
 688			else
 689				ret = ocfs2_enable_quotas(osb);
 690			if (ret < 0) {
 691				/* Return back changes... */
 692				spin_lock(&osb->osb_lock);
 693				sb->s_flags |= SB_RDONLY;
 694				osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 695				spin_unlock(&osb->osb_lock);
 696				goto out;
 697			}
 698		}
 699	}
 700
 701	if (!ret) {
 702		/* Only save off the new mount options in case of a successful
 703		 * remount. */
 704		osb->s_mount_opt = parsed_options.mount_opt;
 705		osb->s_atime_quantum = parsed_options.atime_quantum;
 706		osb->preferred_slot = parsed_options.slot;
 707		if (parsed_options.commit_interval)
 708			osb->osb_commit_interval = parsed_options.commit_interval;
 709
 710		if (!ocfs2_is_hard_readonly(osb))
 711			ocfs2_set_journal_params(osb);
 712
 713		sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 714			((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 715							SB_POSIXACL : 0);
 716	}
 717out:
 718	return ret;
 719}
 720
 721static int ocfs2_sb_probe(struct super_block *sb,
 722			  struct buffer_head **bh,
 723			  int *sector_size,
 724			  struct ocfs2_blockcheck_stats *stats)
 725{
 726	int status, tmpstat;
 727	struct ocfs1_vol_disk_hdr *hdr;
 728	struct ocfs2_dinode *di;
 729	int blksize;
 730
 731	*bh = NULL;
 732
 733	/* may be > 512 */
 734	*sector_size = bdev_logical_block_size(sb->s_bdev);
 735	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 736		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 737		     *sector_size, OCFS2_MAX_BLOCKSIZE);
 738		status = -EINVAL;
 739		goto bail;
 740	}
 741
 742	/* Can this really happen? */
 743	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 744		*sector_size = OCFS2_MIN_BLOCKSIZE;
 745
 746	/* check block zero for old format */
 747	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 748	if (status < 0) {
 749		mlog_errno(status);
 750		goto bail;
 751	}
 752	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 753	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 754		mlog(ML_ERROR, "incompatible version: %u.%u\n",
 755		     hdr->major_version, hdr->minor_version);
 756		status = -EINVAL;
 757	}
 758	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 759		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 760		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 761		     hdr->signature);
 762		status = -EINVAL;
 763	}
 764	brelse(*bh);
 765	*bh = NULL;
 766	if (status < 0) {
 767		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 768		     "upgraded before mounting with ocfs v2\n");
 769		goto bail;
 770	}
 771
 772	/*
 773	 * Now check at magic offset for 512, 1024, 2048, 4096
 774	 * blocksizes.  4096 is the maximum blocksize because it is
 775	 * the minimum clustersize.
 776	 */
 777	status = -EINVAL;
 778	for (blksize = *sector_size;
 779	     blksize <= OCFS2_MAX_BLOCKSIZE;
 780	     blksize <<= 1) {
 781		tmpstat = ocfs2_get_sector(sb, bh,
 782					   OCFS2_SUPER_BLOCK_BLKNO,
 783					   blksize);
 784		if (tmpstat < 0) {
 785			status = tmpstat;
 786			mlog_errno(status);
 787			break;
 788		}
 789		di = (struct ocfs2_dinode *) (*bh)->b_data;
 790		memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 791		spin_lock_init(&stats->b_lock);
 792		tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 793		if (tmpstat < 0) {
 794			brelse(*bh);
 795			*bh = NULL;
 796		}
 797		if (tmpstat != -EAGAIN) {
 798			status = tmpstat;
 799			break;
 800		}
 801	}
 802
 803bail:
 804	return status;
 805}
 806
 807static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 808{
 809	u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 810
 811	if (osb->s_mount_opt & hb_enabled) {
 812		if (ocfs2_mount_local(osb)) {
 813			mlog(ML_ERROR, "Cannot heartbeat on a locally "
 814			     "mounted device.\n");
 815			return -EINVAL;
 816		}
 817		if (ocfs2_userspace_stack(osb)) {
 818			mlog(ML_ERROR, "Userspace stack expected, but "
 819			     "o2cb heartbeat arguments passed to mount\n");
 820			return -EINVAL;
 821		}
 822		if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 823		     !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 824		    ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 825		     ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 826			mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 827			return -EINVAL;
 828		}
 829	}
 830
 831	if (!(osb->s_mount_opt & hb_enabled)) {
 832		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 833		    !ocfs2_userspace_stack(osb)) {
 834			mlog(ML_ERROR, "Heartbeat has to be started to mount "
 835			     "a read-write clustered device.\n");
 836			return -EINVAL;
 837		}
 838	}
 839
 840	return 0;
 841}
 842
 843/*
 844 * If we're using a userspace stack, mount should have passed
 845 * a name that matches the disk.  If not, mount should not
 846 * have passed a stack.
 847 */
 848static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 849					struct mount_options *mopt)
 850{
 851	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 852		mlog(ML_ERROR,
 853		     "cluster stack passed to mount, but this filesystem "
 854		     "does not support it\n");
 855		return -EINVAL;
 856	}
 857
 858	if (ocfs2_userspace_stack(osb) &&
 859	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 860		    OCFS2_STACK_LABEL_LEN)) {
 861		mlog(ML_ERROR,
 862		     "cluster stack passed to mount (\"%s\") does not "
 863		     "match the filesystem (\"%s\")\n",
 864		     mopt->cluster_stack,
 865		     osb->osb_cluster_stack);
 866		return -EINVAL;
 867	}
 868
 869	return 0;
 870}
 871
 872static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 873{
 874	int type;
 875	struct super_block *sb = osb->sb;
 876	unsigned int feature[OCFS2_MAXQUOTAS] = {
 877					OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 878					OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 879	int status = 0;
 880
 881	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 882		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 883			continue;
 884		if (unsuspend)
 885			status = dquot_resume(sb, type);
 886		else {
 887			struct ocfs2_mem_dqinfo *oinfo;
 888
 889			/* Cancel periodic syncing before suspending */
 890			oinfo = sb_dqinfo(sb, type)->dqi_priv;
 891			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 892			status = dquot_suspend(sb, type);
 893		}
 894		if (status < 0)
 895			break;
 896	}
 897	if (status < 0)
 898		mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 899		     "remount (error = %d).\n", status);
 900	return status;
 901}
 902
 903static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 904{
 905	struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
 906	struct super_block *sb = osb->sb;
 907	unsigned int feature[OCFS2_MAXQUOTAS] = {
 908					OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 909					OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 910	unsigned int ino[OCFS2_MAXQUOTAS] = {
 911					LOCAL_USER_QUOTA_SYSTEM_INODE,
 912					LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 913	int status;
 914	int type;
 915
 916	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 917	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 918		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 919			continue;
 920		inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 921							osb->slot_num);
 922		if (!inode[type]) {
 923			status = -ENOENT;
 924			goto out_quota_off;
 925		}
 926		status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
 927						DQUOT_USAGE_ENABLED);
 928		if (status < 0)
 929			goto out_quota_off;
 930	}
 931
 932	for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 933		iput(inode[type]);
 934	return 0;
 935out_quota_off:
 936	ocfs2_disable_quotas(osb);
 937	for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 938		iput(inode[type]);
 939	mlog_errno(status);
 940	return status;
 941}
 942
 943static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 944{
 945	int type;
 946	struct inode *inode;
 947	struct super_block *sb = osb->sb;
 948	struct ocfs2_mem_dqinfo *oinfo;
 949
 950	/* We mostly ignore errors in this function because there's not much
 951	 * we can do when we see them */
 952	for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 953		if (!sb_has_quota_loaded(sb, type))
 954			continue;
 
 955		oinfo = sb_dqinfo(sb, type)->dqi_priv;
 956		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 957		inode = igrab(sb->s_dquot.files[type]);
 958		/* Turn off quotas. This will remove all dquot structures from
 959		 * memory and so they will be automatically synced to global
 960		 * quota files */
 961		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 962					DQUOT_LIMITS_ENABLED);
 
 
 963		iput(inode);
 964	}
 965}
 966
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 967static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 968{
 969	struct dentry *root;
 970	int status, sector_size;
 971	struct mount_options parsed_options;
 972	struct inode *inode = NULL;
 973	struct ocfs2_super *osb = NULL;
 974	struct buffer_head *bh = NULL;
 975	char nodestr[12];
 976	struct ocfs2_blockcheck_stats stats;
 977
 978	trace_ocfs2_fill_super(sb, data, silent);
 979
 980	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
 981		status = -EINVAL;
 982		goto out;
 983	}
 984
 985	/* probe for superblock */
 986	status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
 987	if (status < 0) {
 988		mlog(ML_ERROR, "superblock probe failed!\n");
 989		goto out;
 990	}
 991
 992	status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
 
 
 
 
 
 993	brelse(bh);
 994	bh = NULL;
 995	if (status < 0)
 996		goto out;
 997
 998	osb = OCFS2_SB(sb);
 999
1000	if (!ocfs2_check_set_options(sb, &parsed_options)) {
1001		status = -EINVAL;
1002		goto out_super;
1003	}
1004	osb->s_mount_opt = parsed_options.mount_opt;
1005	osb->s_atime_quantum = parsed_options.atime_quantum;
1006	osb->preferred_slot = parsed_options.slot;
1007	osb->osb_commit_interval = parsed_options.commit_interval;
1008
1009	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1010	osb->osb_resv_level = parsed_options.resv_level;
1011	osb->osb_dir_resv_level = parsed_options.resv_level;
1012	if (parsed_options.dir_resv_level == -1)
1013		osb->osb_dir_resv_level = parsed_options.resv_level;
1014	else
1015		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1016
1017	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1018	if (status)
1019		goto out_super;
1020
1021	sb->s_magic = OCFS2_SUPER_MAGIC;
1022
1023	sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1024		((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1025
1026	/* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1027	 * heartbeat=none */
1028	if (bdev_read_only(sb->s_bdev)) {
1029		if (!sb_rdonly(sb)) {
1030			status = -EACCES;
1031			mlog(ML_ERROR, "Readonly device detected but readonly "
1032			     "mount was not specified.\n");
1033			goto out_super;
1034		}
1035
1036		/* You should not be able to start a local heartbeat
1037		 * on a readonly device. */
1038		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1039			status = -EROFS;
1040			mlog(ML_ERROR, "Local heartbeat specified on readonly "
1041			     "device.\n");
1042			goto out_super;
1043		}
1044
1045		status = ocfs2_check_journals_nolocks(osb);
1046		if (status < 0) {
1047			if (status == -EROFS)
1048				mlog(ML_ERROR, "Recovery required on readonly "
1049				     "file system, but write access is "
1050				     "unavailable.\n");
1051			goto out_super;
 
 
1052		}
1053
1054		ocfs2_set_ro_flag(osb, 1);
1055
1056		printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1057		       "Cluster services will not be used for this mount. "
1058		       "Recovery will be skipped.\n", osb->dev_str);
1059	}
1060
1061	if (!ocfs2_is_hard_readonly(osb)) {
1062		if (sb_rdonly(sb))
1063			ocfs2_set_ro_flag(osb, 0);
1064	}
1065
1066	status = ocfs2_verify_heartbeat(osb);
1067	if (status < 0)
1068		goto out_super;
 
 
1069
1070	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1071						 ocfs2_debugfs_root);
 
 
 
 
 
1072
1073	debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1074			    osb, &ocfs2_osb_debug_fops);
 
 
 
 
 
 
 
1075
1076	if (ocfs2_meta_ecc(osb))
1077		ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1078							osb->osb_debug_root);
 
 
 
 
 
 
 
 
1079
1080	status = ocfs2_mount_volume(sb);
1081	if (status < 0)
1082		goto out_debugfs;
1083
1084	if (osb->root_inode)
1085		inode = igrab(osb->root_inode);
1086
1087	if (!inode) {
1088		status = -EIO;
1089		goto out_dismount;
1090	}
1091
1092	osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1093						&ocfs2_kset->kobj);
1094	if (!osb->osb_dev_kset) {
1095		status = -ENOMEM;
1096		mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1097		goto out_dismount;
1098	}
1099
1100	/* Create filecheck sysfs related directories/files at
1101	 * /sys/fs/ocfs2/<devname>/filecheck */
1102	if (ocfs2_filecheck_create_sysfs(osb)) {
1103		status = -ENOMEM;
1104		mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1105			"/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1106		goto out_dismount;
1107	}
1108
1109	root = d_make_root(inode);
1110	if (!root) {
1111		status = -ENOMEM;
1112		goto out_dismount;
 
1113	}
1114
1115	sb->s_root = root;
1116
1117	ocfs2_complete_mount_recovery(osb);
1118
1119	if (ocfs2_mount_local(osb))
1120		snprintf(nodestr, sizeof(nodestr), "local");
1121	else
1122		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1123
1124	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1125	       "with %s data mode.\n",
1126	       osb->dev_str, nodestr, osb->slot_num,
1127	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1128	       "ordered");
1129
1130	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1131	wake_up(&osb->osb_mount_event);
1132
1133	/* Now we can initialize quotas because we can afford to wait
1134	 * for cluster locks recovery now. That also means that truncation
1135	 * log recovery can happen but that waits for proper quota setup */
1136	if (!sb_rdonly(sb)) {
1137		status = ocfs2_enable_quotas(osb);
1138		if (status < 0) {
1139			/* We have to err-out specially here because
1140			 * s_root is already set */
1141			mlog_errno(status);
1142			atomic_set(&osb->vol_state, VOLUME_DISABLED);
1143			wake_up(&osb->osb_mount_event);
1144			return status;
1145		}
1146	}
1147
1148	ocfs2_complete_quota_recovery(osb);
1149
1150	/* Now we wake up again for processes waiting for quotas */
1151	atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1152	wake_up(&osb->osb_mount_event);
1153
1154	/* Start this when the mount is almost sure of being successful */
1155	ocfs2_orphan_scan_start(osb);
1156
1157	return status;
1158
1159out_dismount:
1160	atomic_set(&osb->vol_state, VOLUME_DISABLED);
1161	wake_up(&osb->osb_mount_event);
1162	ocfs2_free_replay_slots(osb);
1163	ocfs2_dismount_volume(sb, 1);
1164	goto out;
1165
1166out_debugfs:
1167	debugfs_remove_recursive(osb->osb_debug_root);
1168out_super:
1169	ocfs2_release_system_inodes(osb);
1170	kfree(osb->recovery_map);
1171	ocfs2_delete_osb(osb);
1172	kfree(osb);
1173out:
1174	mlog_errno(status);
1175
 
 
1176	return status;
1177}
1178
1179static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1180			int flags,
1181			const char *dev_name,
1182			void *data)
1183{
1184	return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1185}
1186
1187static struct file_system_type ocfs2_fs_type = {
1188	.owner          = THIS_MODULE,
1189	.name           = "ocfs2",
1190	.mount          = ocfs2_mount,
1191	.kill_sb        = kill_block_super,
1192	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1193	.next           = NULL
1194};
1195MODULE_ALIAS_FS("ocfs2");
1196
1197static int ocfs2_check_set_options(struct super_block *sb,
1198				   struct mount_options *options)
1199{
1200	if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1201	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1202					 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1203		mlog(ML_ERROR, "User quotas were requested, but this "
1204		     "filesystem does not have the feature enabled.\n");
1205		return 0;
1206	}
1207	if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1208	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1209					 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1210		mlog(ML_ERROR, "Group quotas were requested, but this "
1211		     "filesystem does not have the feature enabled.\n");
1212		return 0;
1213	}
1214	if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1215	    !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1216		mlog(ML_ERROR, "ACL support requested but extended attributes "
1217		     "feature is not enabled\n");
1218		return 0;
1219	}
1220	/* No ACL setting specified? Use XATTR feature... */
1221	if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1222				    OCFS2_MOUNT_NO_POSIX_ACL))) {
1223		if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1224			options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1225		else
1226			options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1227	}
1228	return 1;
1229}
1230
1231static int ocfs2_parse_options(struct super_block *sb,
1232			       char *options,
1233			       struct mount_options *mopt,
1234			       int is_remount)
1235{
1236	int status, user_stack = 0;
1237	char *p;
1238	u32 tmp;
1239	int token, option;
1240	substring_t args[MAX_OPT_ARGS];
1241
1242	trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1243
1244	mopt->commit_interval = 0;
1245	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1246	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1247	mopt->slot = OCFS2_INVALID_SLOT;
1248	mopt->localalloc_opt = -1;
1249	mopt->cluster_stack[0] = '\0';
1250	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1251	mopt->dir_resv_level = -1;
1252
1253	if (!options) {
1254		status = 1;
1255		goto bail;
1256	}
1257
1258	while ((p = strsep(&options, ",")) != NULL) {
 
 
 
1259		if (!*p)
1260			continue;
1261
1262		token = match_token(p, tokens, args);
1263		switch (token) {
1264		case Opt_hb_local:
1265			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1266			break;
1267		case Opt_hb_none:
1268			mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1269			break;
1270		case Opt_hb_global:
1271			mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1272			break;
1273		case Opt_barrier:
1274			if (match_int(&args[0], &option)) {
1275				status = 0;
1276				goto bail;
1277			}
1278			if (option)
1279				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1280			else
1281				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1282			break;
1283		case Opt_intr:
1284			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1285			break;
1286		case Opt_nointr:
1287			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1288			break;
1289		case Opt_err_panic:
1290			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1291			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1292			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1293			break;
1294		case Opt_err_ro:
1295			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1296			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1297			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1298			break;
1299		case Opt_err_cont:
1300			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1301			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1302			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1303			break;
1304		case Opt_data_ordered:
1305			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1306			break;
1307		case Opt_data_writeback:
1308			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1309			break;
1310		case Opt_user_xattr:
1311			mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1312			break;
1313		case Opt_nouser_xattr:
1314			mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1315			break;
1316		case Opt_atime_quantum:
1317			if (match_int(&args[0], &option)) {
1318				status = 0;
1319				goto bail;
1320			}
1321			if (option >= 0)
1322				mopt->atime_quantum = option;
1323			break;
1324		case Opt_slot:
 
1325			if (match_int(&args[0], &option)) {
1326				status = 0;
1327				goto bail;
1328			}
1329			if (option)
1330				mopt->slot = (u16)option;
1331			break;
1332		case Opt_commit:
 
1333			if (match_int(&args[0], &option)) {
1334				status = 0;
1335				goto bail;
1336			}
1337			if (option < 0)
1338				return 0;
1339			if (option == 0)
1340				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1341			mopt->commit_interval = HZ * option;
1342			break;
1343		case Opt_localalloc:
 
1344			if (match_int(&args[0], &option)) {
1345				status = 0;
1346				goto bail;
1347			}
1348			if (option >= 0)
1349				mopt->localalloc_opt = option;
1350			break;
1351		case Opt_localflocks:
1352			/*
1353			 * Changing this during remount could race
1354			 * flock() requests, or "unbalance" existing
1355			 * ones (e.g., a lock is taken in one mode but
1356			 * dropped in the other). If users care enough
1357			 * to flip locking modes during remount, we
1358			 * could add a "local" flag to individual
1359			 * flock structures for proper tracking of
1360			 * state.
1361			 */
1362			if (!is_remount)
1363				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1364			break;
1365		case Opt_stack:
1366			/* Check both that the option we were passed
1367			 * is of the right length and that it is a proper
1368			 * string of the right length.
1369			 */
1370			if (((args[0].to - args[0].from) !=
1371			     OCFS2_STACK_LABEL_LEN) ||
1372			    (strnlen(args[0].from,
1373				     OCFS2_STACK_LABEL_LEN) !=
1374			     OCFS2_STACK_LABEL_LEN)) {
1375				mlog(ML_ERROR,
1376				     "Invalid cluster_stack option\n");
1377				status = 0;
1378				goto bail;
1379			}
1380			memcpy(mopt->cluster_stack, args[0].from,
1381			       OCFS2_STACK_LABEL_LEN);
1382			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1383			/*
1384			 * Open code the memcmp here as we don't have
1385			 * an osb to pass to
1386			 * ocfs2_userspace_stack().
1387			 */
1388			if (memcmp(mopt->cluster_stack,
1389				   OCFS2_CLASSIC_CLUSTER_STACK,
1390				   OCFS2_STACK_LABEL_LEN))
1391				user_stack = 1;
1392			break;
1393		case Opt_inode64:
1394			mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1395			break;
1396		case Opt_usrquota:
1397			mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1398			break;
1399		case Opt_grpquota:
1400			mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1401			break;
1402		case Opt_coherency_buffered:
1403			mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1404			break;
1405		case Opt_coherency_full:
1406			mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1407			break;
1408		case Opt_acl:
1409			mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1410			mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1411			break;
1412		case Opt_noacl:
1413			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1414			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1415			break;
1416		case Opt_resv_level:
1417			if (is_remount)
1418				break;
1419			if (match_int(&args[0], &option)) {
1420				status = 0;
1421				goto bail;
1422			}
1423			if (option >= OCFS2_MIN_RESV_LEVEL &&
1424			    option < OCFS2_MAX_RESV_LEVEL)
1425				mopt->resv_level = option;
1426			break;
1427		case Opt_dir_resv_level:
1428			if (is_remount)
1429				break;
1430			if (match_int(&args[0], &option)) {
1431				status = 0;
1432				goto bail;
1433			}
1434			if (option >= OCFS2_MIN_RESV_LEVEL &&
1435			    option < OCFS2_MAX_RESV_LEVEL)
1436				mopt->dir_resv_level = option;
1437			break;
1438		case Opt_journal_async_commit:
1439			mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1440			break;
1441		default:
1442			mlog(ML_ERROR,
1443			     "Unrecognized mount option \"%s\" "
1444			     "or missing value\n", p);
1445			status = 0;
1446			goto bail;
1447		}
1448	}
1449
1450	if (user_stack == 0) {
1451		/* Ensure only one heartbeat mode */
1452		tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1453					 OCFS2_MOUNT_HB_GLOBAL |
1454					 OCFS2_MOUNT_HB_NONE);
1455		if (hweight32(tmp) != 1) {
1456			mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1457			status = 0;
1458			goto bail;
1459		}
1460	}
1461
1462	status = 1;
1463
1464bail:
1465	return status;
1466}
1467
1468static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1469{
1470	struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1471	unsigned long opts = osb->s_mount_opt;
1472	unsigned int local_alloc_megs;
1473
1474	if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1475		seq_printf(s, ",_netdev");
1476		if (opts & OCFS2_MOUNT_HB_LOCAL)
1477			seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1478		else
1479			seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1480	} else
1481		seq_printf(s, ",%s", OCFS2_HB_NONE);
1482
1483	if (opts & OCFS2_MOUNT_NOINTR)
1484		seq_printf(s, ",nointr");
1485
1486	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1487		seq_printf(s, ",data=writeback");
1488	else
1489		seq_printf(s, ",data=ordered");
1490
1491	if (opts & OCFS2_MOUNT_BARRIER)
1492		seq_printf(s, ",barrier=1");
1493
1494	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1495		seq_printf(s, ",errors=panic");
1496	else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1497		seq_printf(s, ",errors=continue");
1498	else
1499		seq_printf(s, ",errors=remount-ro");
1500
1501	if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1502		seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1503
1504	seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1505
1506	if (osb->osb_commit_interval)
1507		seq_printf(s, ",commit=%u",
1508			   (unsigned) (osb->osb_commit_interval / HZ));
1509
1510	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1511	if (local_alloc_megs != ocfs2_la_default_mb(osb))
1512		seq_printf(s, ",localalloc=%d", local_alloc_megs);
1513
1514	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1515		seq_printf(s, ",localflocks,");
1516
1517	if (osb->osb_cluster_stack[0])
1518		seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1519				  OCFS2_STACK_LABEL_LEN);
1520	if (opts & OCFS2_MOUNT_USRQUOTA)
1521		seq_printf(s, ",usrquota");
1522	if (opts & OCFS2_MOUNT_GRPQUOTA)
1523		seq_printf(s, ",grpquota");
1524
1525	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1526		seq_printf(s, ",coherency=buffered");
1527	else
1528		seq_printf(s, ",coherency=full");
1529
1530	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1531		seq_printf(s, ",nouser_xattr");
1532	else
1533		seq_printf(s, ",user_xattr");
1534
1535	if (opts & OCFS2_MOUNT_INODE64)
1536		seq_printf(s, ",inode64");
1537
1538	if (opts & OCFS2_MOUNT_POSIX_ACL)
1539		seq_printf(s, ",acl");
1540	else
1541		seq_printf(s, ",noacl");
1542
1543	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1544		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1545
1546	if (osb->osb_dir_resv_level != osb->osb_resv_level)
1547		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1548
1549	if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1550		seq_printf(s, ",journal_async_commit");
1551
1552	return 0;
1553}
1554
1555static int __init ocfs2_init(void)
1556{
1557	int status;
1558
1559	status = init_ocfs2_uptodate_cache();
1560	if (status < 0)
1561		goto out1;
1562
1563	status = ocfs2_initialize_mem_caches();
1564	if (status < 0)
1565		goto out2;
1566
 
 
 
 
 
 
1567	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
 
 
 
 
1568
1569	ocfs2_set_locking_protocol();
1570
1571	status = register_quota_format(&ocfs2_quota_format);
1572	if (status < 0)
1573		goto out3;
1574	status = register_filesystem(&ocfs2_fs_type);
1575	if (!status)
1576		return 0;
1577
1578	unregister_quota_format(&ocfs2_quota_format);
1579out3:
 
1580	debugfs_remove(ocfs2_debugfs_root);
 
1581	ocfs2_free_mem_caches();
1582out2:
1583	exit_ocfs2_uptodate_cache();
1584out1:
1585	mlog_errno(status);
1586	return status;
1587}
1588
1589static void __exit ocfs2_exit(void)
1590{
 
 
 
 
 
1591	unregister_quota_format(&ocfs2_quota_format);
1592
1593	debugfs_remove(ocfs2_debugfs_root);
1594
1595	ocfs2_free_mem_caches();
1596
1597	unregister_filesystem(&ocfs2_fs_type);
1598
1599	exit_ocfs2_uptodate_cache();
1600}
1601
1602static void ocfs2_put_super(struct super_block *sb)
1603{
1604	trace_ocfs2_put_super(sb);
1605
1606	ocfs2_sync_blockdev(sb);
1607	ocfs2_dismount_volume(sb, 0);
1608}
1609
1610static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1611{
1612	struct ocfs2_super *osb;
1613	u32 numbits, freebits;
1614	int status;
1615	struct ocfs2_dinode *bm_lock;
1616	struct buffer_head *bh = NULL;
1617	struct inode *inode = NULL;
1618
1619	trace_ocfs2_statfs(dentry->d_sb, buf);
1620
1621	osb = OCFS2_SB(dentry->d_sb);
1622
1623	inode = ocfs2_get_system_file_inode(osb,
1624					    GLOBAL_BITMAP_SYSTEM_INODE,
1625					    OCFS2_INVALID_SLOT);
1626	if (!inode) {
1627		mlog(ML_ERROR, "failed to get bitmap inode\n");
1628		status = -EIO;
1629		goto bail;
1630	}
1631
1632	status = ocfs2_inode_lock(inode, &bh, 0);
1633	if (status < 0) {
1634		mlog_errno(status);
1635		goto bail;
1636	}
1637
1638	bm_lock = (struct ocfs2_dinode *) bh->b_data;
1639
1640	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1641	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1642
1643	buf->f_type = OCFS2_SUPER_MAGIC;
1644	buf->f_bsize = dentry->d_sb->s_blocksize;
1645	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1646	buf->f_blocks = ((sector_t) numbits) *
1647			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
1648	buf->f_bfree = ((sector_t) freebits) *
1649		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1650	buf->f_bavail = buf->f_bfree;
1651	buf->f_files = numbits;
1652	buf->f_ffree = freebits;
1653	buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1654				& 0xFFFFFFFFUL;
1655	buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1656				OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1657
1658	brelse(bh);
1659
1660	ocfs2_inode_unlock(inode, 0);
1661	status = 0;
1662bail:
1663	iput(inode);
 
1664
1665	if (status)
1666		mlog_errno(status);
1667
1668	return status;
1669}
1670
1671static void ocfs2_inode_init_once(void *data)
1672{
1673	struct ocfs2_inode_info *oi = data;
1674
1675	oi->ip_flags = 0;
1676	oi->ip_open_count = 0;
1677	spin_lock_init(&oi->ip_lock);
1678	ocfs2_extent_map_init(&oi->vfs_inode);
1679	INIT_LIST_HEAD(&oi->ip_io_markers);
1680	INIT_LIST_HEAD(&oi->ip_unwritten_list);
1681	oi->ip_dir_start_lookup = 0;
 
1682	init_rwsem(&oi->ip_alloc_sem);
1683	init_rwsem(&oi->ip_xattr_sem);
1684	mutex_init(&oi->ip_io_mutex);
1685
1686	oi->ip_blkno = 0ULL;
1687	oi->ip_clusters = 0;
1688	oi->ip_next_orphan = NULL;
1689
1690	ocfs2_resv_init_once(&oi->ip_la_data_resv);
1691
1692	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1693	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1694	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1695
1696	ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1697				  &ocfs2_inode_caching_ops);
1698
1699	inode_init_once(&oi->vfs_inode);
1700}
1701
1702static int ocfs2_initialize_mem_caches(void)
1703{
1704	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1705				       sizeof(struct ocfs2_inode_info),
1706				       0,
1707				       (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1708						SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1709				       ocfs2_inode_init_once);
1710	ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1711					sizeof(struct ocfs2_dquot),
1712					0,
1713					(SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1714						SLAB_MEM_SPREAD),
1715					NULL);
1716	ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1717					sizeof(struct ocfs2_quota_chunk),
1718					0,
1719					(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1720					NULL);
1721	if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1722	    !ocfs2_qf_chunk_cachep) {
1723		kmem_cache_destroy(ocfs2_inode_cachep);
1724		kmem_cache_destroy(ocfs2_dquot_cachep);
1725		kmem_cache_destroy(ocfs2_qf_chunk_cachep);
 
 
 
1726		return -ENOMEM;
1727	}
1728
1729	return 0;
1730}
1731
1732static void ocfs2_free_mem_caches(void)
1733{
1734	/*
1735	 * Make sure all delayed rcu free inodes are flushed before we
1736	 * destroy cache.
1737	 */
1738	rcu_barrier();
1739	kmem_cache_destroy(ocfs2_inode_cachep);
 
1740	ocfs2_inode_cachep = NULL;
1741
1742	kmem_cache_destroy(ocfs2_dquot_cachep);
 
1743	ocfs2_dquot_cachep = NULL;
1744
1745	kmem_cache_destroy(ocfs2_qf_chunk_cachep);
 
1746	ocfs2_qf_chunk_cachep = NULL;
1747}
1748
1749static int ocfs2_get_sector(struct super_block *sb,
1750			    struct buffer_head **bh,
1751			    int block,
1752			    int sect_size)
1753{
1754	if (!sb_set_blocksize(sb, sect_size)) {
1755		mlog(ML_ERROR, "unable to set blocksize\n");
1756		return -EIO;
1757	}
1758
1759	*bh = sb_getblk(sb, block);
1760	if (!*bh) {
1761		mlog_errno(-ENOMEM);
1762		return -ENOMEM;
1763	}
1764	lock_buffer(*bh);
1765	if (!buffer_dirty(*bh))
1766		clear_buffer_uptodate(*bh);
1767	unlock_buffer(*bh);
1768	if (bh_read(*bh, 0) < 0) {
 
 
1769		mlog_errno(-EIO);
1770		brelse(*bh);
1771		*bh = NULL;
1772		return -EIO;
1773	}
1774
1775	return 0;
1776}
1777
1778static int ocfs2_mount_volume(struct super_block *sb)
1779{
1780	int status = 0;
 
1781	struct ocfs2_super *osb = OCFS2_SB(sb);
1782
1783	if (ocfs2_is_hard_readonly(osb))
1784		goto out;
1785
1786	mutex_init(&osb->obs_trim_fs_mutex);
1787
1788	status = ocfs2_dlm_init(osb);
1789	if (status < 0) {
1790		mlog_errno(status);
1791		if (status == -EBADR && ocfs2_userspace_stack(osb))
1792			mlog(ML_ERROR, "couldn't mount because cluster name on"
1793			" disk does not match the running cluster name.\n");
1794		goto out;
1795	}
1796
1797	status = ocfs2_super_lock(osb, 1);
1798	if (status < 0) {
1799		mlog_errno(status);
1800		goto out_dlm;
1801	}
 
1802
1803	/* This will load up the node map and add ourselves to it. */
1804	status = ocfs2_find_slot(osb);
1805	if (status < 0) {
1806		mlog_errno(status);
1807		goto out_super_lock;
1808	}
1809
1810	/* load all node-local system inodes */
1811	status = ocfs2_init_local_system_inodes(osb);
1812	if (status < 0) {
1813		mlog_errno(status);
1814		goto out_super_lock;
1815	}
1816
1817	status = ocfs2_check_volume(osb);
1818	if (status < 0) {
1819		mlog_errno(status);
1820		goto out_system_inodes;
1821	}
1822
1823	status = ocfs2_truncate_log_init(osb);
1824	if (status < 0) {
1825		mlog_errno(status);
1826		goto out_check_volume;
1827	}
1828
1829	ocfs2_super_unlock(osb, 1);
1830	return 0;
 
1831
1832out_check_volume:
1833	ocfs2_free_replay_slots(osb);
1834out_system_inodes:
1835	if (osb->local_alloc_state == OCFS2_LA_ENABLED)
1836		ocfs2_shutdown_local_alloc(osb);
1837	ocfs2_release_system_inodes(osb);
1838	/* before journal shutdown, we should release slot_info */
1839	ocfs2_free_slot_info(osb);
1840	ocfs2_journal_shutdown(osb);
1841out_super_lock:
1842	ocfs2_super_unlock(osb, 1);
1843out_dlm:
1844	ocfs2_dlm_shutdown(osb, 0);
1845out:
1846	return status;
1847}
1848
1849static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1850{
1851	int tmp, hangup_needed = 0;
1852	struct ocfs2_super *osb = NULL;
1853	char nodestr[12];
1854
1855	trace_ocfs2_dismount_volume(sb);
1856
1857	BUG_ON(!sb);
1858	osb = OCFS2_SB(sb);
1859	BUG_ON(!osb);
1860
1861	/* Remove file check sysfs related directores/files,
1862	 * and wait for the pending file check operations */
1863	ocfs2_filecheck_remove_sysfs(osb);
1864
1865	kset_unregister(osb->osb_dev_kset);
1866
1867	/* Orphan scan should be stopped as early as possible */
1868	ocfs2_orphan_scan_stop(osb);
1869
1870	ocfs2_disable_quotas(osb);
1871
1872	/* All dquots should be freed by now */
1873	WARN_ON(!llist_empty(&osb->dquot_drop_list));
1874	/* Wait for worker to be done with the work structure in osb */
1875	cancel_work_sync(&osb->dquot_drop_work);
1876
1877	ocfs2_shutdown_local_alloc(osb);
1878
1879	ocfs2_truncate_log_shutdown(osb);
1880
1881	/* This will disable recovery and flush any recovery work. */
1882	ocfs2_recovery_exit(osb);
1883
 
 
 
 
 
 
 
 
1884	ocfs2_sync_blockdev(sb);
1885
1886	ocfs2_purge_refcount_trees(osb);
1887
1888	/* No cluster connection means we've failed during mount, so skip
1889	 * all the steps which depended on that to complete. */
1890	if (osb->cconn) {
1891		tmp = ocfs2_super_lock(osb, 1);
1892		if (tmp < 0) {
1893			mlog_errno(tmp);
1894			return;
1895		}
1896	}
1897
1898	if (osb->slot_num != OCFS2_INVALID_SLOT)
1899		ocfs2_put_slot(osb);
1900
1901	if (osb->cconn)
1902		ocfs2_super_unlock(osb, 1);
1903
1904	ocfs2_release_system_inodes(osb);
1905
1906	ocfs2_journal_shutdown(osb);
1907
1908	/*
1909	 * If we're dismounting due to mount error, mount.ocfs2 will clean
1910	 * up heartbeat.  If we're a local mount, there is no heartbeat.
1911	 * If we failed before we got a uuid_str yet, we can't stop
1912	 * heartbeat.  Otherwise, do it.
1913	 */
1914	if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1915	    !ocfs2_is_hard_readonly(osb))
1916		hangup_needed = 1;
1917
1918	ocfs2_dlm_shutdown(osb, hangup_needed);
 
1919
1920	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1921	debugfs_remove_recursive(osb->osb_debug_root);
1922
1923	if (hangup_needed)
1924		ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1925
1926	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1927
1928	if (ocfs2_mount_local(osb))
1929		snprintf(nodestr, sizeof(nodestr), "local");
1930	else
1931		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1932
1933	printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1934	       osb->dev_str, nodestr);
1935
1936	ocfs2_delete_osb(osb);
1937	kfree(osb);
1938	sb->s_dev = 0;
1939	sb->s_fs_info = NULL;
1940}
1941
1942static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1943				unsigned uuid_bytes)
1944{
1945	int i, ret;
1946	char *ptr;
1947
1948	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1949
1950	osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1951	if (osb->uuid_str == NULL)
1952		return -ENOMEM;
1953
1954	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1955		/* print with null */
1956		ret = snprintf(ptr, 3, "%02X", uuid[i]);
1957		if (ret != 2) /* drop super cleans up */
1958			return -EINVAL;
1959		/* then only advance past the last char */
1960		ptr += 2;
1961	}
1962
1963	return 0;
1964}
1965
1966/* Make sure entire volume is addressable by our journal.  Requires
1967   osb_clusters_at_boot to be valid and for the journal to have been
1968   initialized by ocfs2_journal_init(). */
1969static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1970{
1971	int status = 0;
1972	u64 max_block =
1973		ocfs2_clusters_to_blocks(osb->sb,
1974					 osb->osb_clusters_at_boot) - 1;
1975
1976	/* 32-bit block number is always OK. */
1977	if (max_block <= (u32)~0ULL)
1978		goto out;
1979
1980	/* Volume is "huge", so see if our journal is new enough to
1981	   support it. */
1982	if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
1983				       OCFS2_FEATURE_COMPAT_JBD2_SB) &&
1984	      jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
1985					       JBD2_FEATURE_INCOMPAT_64BIT))) {
1986		mlog(ML_ERROR, "The journal cannot address the entire volume. "
1987		     "Enable the 'block64' journal option with tunefs.ocfs2");
1988		status = -EFBIG;
1989		goto out;
1990	}
1991
1992 out:
1993	return status;
1994}
1995
1996static int ocfs2_initialize_super(struct super_block *sb,
1997				  struct buffer_head *bh,
1998				  int sector_size,
1999				  struct ocfs2_blockcheck_stats *stats)
2000{
2001	int status;
2002	int i, cbits, bbits;
2003	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2004	struct inode *inode = NULL;
 
2005	struct ocfs2_super *osb;
2006	u64 total_blocks;
2007
2008	osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2009	if (!osb) {
2010		status = -ENOMEM;
2011		mlog_errno(status);
2012		goto out;
2013	}
2014
2015	sb->s_fs_info = osb;
2016	sb->s_op = &ocfs2_sops;
2017	sb->s_d_op = &ocfs2_dentry_ops;
2018	sb->s_export_op = &ocfs2_export_ops;
2019	sb->s_qcop = &dquot_quotactl_sysfile_ops;
2020	sb->dq_op = &ocfs2_quota_operations;
2021	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2022	sb->s_xattr = ocfs2_xattr_handlers;
2023	sb->s_time_gran = 1;
2024	sb->s_flags |= SB_NOATIME;
2025	/* this is needed to support O_LARGEFILE */
2026	cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2027	bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2028	sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2029	memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2030	       sizeof(di->id2.i_super.s_uuid));
2031
2032	osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2033
2034	for (i = 0; i < 3; i++)
2035		osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2036	osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2037
2038	osb->sb = sb;
 
2039	osb->s_sectsize_bits = blksize_bits(sector_size);
2040	BUG_ON(!osb->s_sectsize_bits);
2041
2042	spin_lock_init(&osb->dc_task_lock);
2043	init_waitqueue_head(&osb->dc_event);
2044	osb->dc_work_sequence = 0;
2045	osb->dc_wake_sequence = 0;
2046	INIT_LIST_HEAD(&osb->blocked_lock_list);
2047	osb->blocked_lock_count = 0;
2048	spin_lock_init(&osb->osb_lock);
2049	spin_lock_init(&osb->osb_xattr_lock);
2050	ocfs2_init_steal_slots(osb);
2051
2052	mutex_init(&osb->system_file_mutex);
2053
2054	atomic_set(&osb->alloc_stats.moves, 0);
2055	atomic_set(&osb->alloc_stats.local_data, 0);
2056	atomic_set(&osb->alloc_stats.bitmap_data, 0);
2057	atomic_set(&osb->alloc_stats.bg_allocs, 0);
2058	atomic_set(&osb->alloc_stats.bg_extends, 0);
2059
2060	/* Copy the blockcheck stats from the superblock probe */
2061	osb->osb_ecc_stats = *stats;
2062
2063	ocfs2_init_node_maps(osb);
2064
2065	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2066		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2067
2068	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2069	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2070		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2071		     osb->max_slots);
2072		status = -EINVAL;
2073		goto out;
2074	}
2075
2076	ocfs2_orphan_scan_init(osb);
2077
2078	status = ocfs2_recovery_init(osb);
2079	if (status) {
2080		mlog(ML_ERROR, "Unable to initialize recovery state\n");
2081		mlog_errno(status);
2082		goto out;
2083	}
2084
2085	init_waitqueue_head(&osb->checkpoint_event);
2086
2087	osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2088
2089	osb->slot_num = OCFS2_INVALID_SLOT;
2090
2091	osb->s_xattr_inline_size = le16_to_cpu(
2092					di->id2.i_super.s_xattr_inline_size);
2093
2094	osb->local_alloc_state = OCFS2_LA_UNUSED;
2095	osb->local_alloc_bh = NULL;
2096	INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2097
2098	init_waitqueue_head(&osb->osb_mount_event);
2099
2100	ocfs2_resmap_init(osb, &osb->osb_la_resmap);
 
 
 
 
2101
2102	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2103	if (!osb->vol_label) {
2104		mlog(ML_ERROR, "unable to alloc vol label\n");
2105		status = -ENOMEM;
2106		goto out_recovery_map;
2107	}
2108
2109	osb->slot_recovery_generations =
2110		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2111			GFP_KERNEL);
2112	if (!osb->slot_recovery_generations) {
2113		status = -ENOMEM;
2114		mlog_errno(status);
2115		goto out_vol_label;
2116	}
2117
2118	init_waitqueue_head(&osb->osb_wipe_event);
2119	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2120					sizeof(*osb->osb_orphan_wipes),
2121					GFP_KERNEL);
2122	if (!osb->osb_orphan_wipes) {
2123		status = -ENOMEM;
2124		mlog_errno(status);
2125		goto out_slot_recovery_gen;
2126	}
2127
2128	osb->osb_rf_lock_tree = RB_ROOT;
2129
2130	osb->s_feature_compat =
2131		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2132	osb->s_feature_ro_compat =
2133		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2134	osb->s_feature_incompat =
2135		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2136
2137	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2138		mlog(ML_ERROR, "couldn't mount because of unsupported "
2139		     "optional features (%x).\n", i);
2140		status = -EINVAL;
2141		goto out_orphan_wipes;
2142	}
2143	if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
 
2144		mlog(ML_ERROR, "couldn't mount RDWR because of "
2145		     "unsupported optional features (%x).\n", i);
2146		status = -EINVAL;
2147		goto out_orphan_wipes;
2148	}
2149
2150	if (ocfs2_clusterinfo_valid(osb)) {
2151		/*
2152		 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2153		 * terminated, so make sure no overflow happens here by using
2154		 * memcpy. Destination strings will always be null terminated
2155		 * because osb is allocated using kzalloc.
2156		 */
2157		osb->osb_stackflags =
2158			OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2159		memcpy(osb->osb_cluster_stack,
2160		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2161		       OCFS2_STACK_LABEL_LEN);
2162		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2163			mlog(ML_ERROR,
2164			     "couldn't mount because of an invalid "
2165			     "cluster stack label (%s) \n",
2166			     osb->osb_cluster_stack);
2167			status = -EINVAL;
2168			goto out_orphan_wipes;
2169		}
2170		memcpy(osb->osb_cluster_name,
2171			OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2172			OCFS2_CLUSTER_NAME_LEN);
2173	} else {
2174		/* The empty string is identical with classic tools that
2175		 * don't know about s_cluster_info. */
2176		osb->osb_cluster_stack[0] = '\0';
2177	}
2178
2179	get_random_bytes(&osb->s_next_generation, sizeof(u32));
2180
2181	/*
2182	 * FIXME
2183	 * This should be done in ocfs2_journal_init(), but any inode
2184	 * writes back operation will cause the filesystem to crash.
 
 
2185	 */
2186	status = ocfs2_journal_alloc(osb);
2187	if (status < 0)
2188		goto out_orphan_wipes;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2189
2190	INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2191	init_llist_head(&osb->dquot_drop_list);
2192
2193	/* get some pseudo constants for clustersize bits */
2194	osb->s_clustersize_bits =
2195		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2196	osb->s_clustersize = 1 << osb->s_clustersize_bits;
2197
2198	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2199	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2200		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2201		     osb->s_clustersize);
2202		status = -EINVAL;
2203		goto out_journal;
2204	}
2205
2206	total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2207						le32_to_cpu(di->i_clusters));
2208
2209	status = generic_check_addressable(osb->sb->s_blocksize_bits,
2210					   total_blocks);
2211	if (status) {
2212		mlog(ML_ERROR, "Volume too large "
2213		     "to mount safely on this system");
2214		status = -EFBIG;
2215		goto out_journal;
2216	}
2217
2218	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2219				 sizeof(di->id2.i_super.s_uuid))) {
2220		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2221		status = -ENOMEM;
2222		goto out_journal;
2223	}
2224
2225	strscpy(osb->vol_label, di->id2.i_super.s_label,
2226		OCFS2_MAX_VOL_LABEL_LEN);
2227	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2228	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2229	osb->first_cluster_group_blkno =
2230		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2231	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2232	osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2233	trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2234				     (unsigned long long)osb->root_blkno,
2235				     (unsigned long long)osb->system_dir_blkno,
2236				     osb->s_clustersize_bits);
2237
2238	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2239	if (!osb->osb_dlm_debug) {
2240		status = -ENOMEM;
2241		mlog_errno(status);
2242		goto out_uuid_str;
2243	}
2244
2245	atomic_set(&osb->vol_state, VOLUME_INIT);
2246
2247	/* load root, system_dir, and all global system inodes */
2248	status = ocfs2_init_global_system_inodes(osb);
2249	if (status < 0) {
2250		mlog_errno(status);
2251		goto out_dlm_out;
2252	}
2253
2254	/*
2255	 * global bitmap
2256	 */
2257	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2258					    OCFS2_INVALID_SLOT);
2259	if (!inode) {
2260		status = -EINVAL;
2261		mlog_errno(status);
2262		goto out_system_inodes;
2263	}
2264
2265	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2266	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2267	iput(inode);
2268
2269	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2270				 osb->s_feature_incompat) * 8;
2271
2272	status = ocfs2_init_slot_info(osb);
2273	if (status < 0) {
2274		mlog_errno(status);
2275		goto out_system_inodes;
2276	}
2277
2278	osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2279	if (!osb->ocfs2_wq) {
2280		status = -ENOMEM;
2281		mlog_errno(status);
2282		goto out_slot_info;
2283	}
 
2284
2285	return status;
2286
2287out_slot_info:
2288	ocfs2_free_slot_info(osb);
2289out_system_inodes:
2290	ocfs2_release_system_inodes(osb);
2291out_dlm_out:
2292	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2293out_uuid_str:
2294	kfree(osb->uuid_str);
2295out_journal:
2296	kfree(osb->journal);
2297out_orphan_wipes:
2298	kfree(osb->osb_orphan_wipes);
2299out_slot_recovery_gen:
2300	kfree(osb->slot_recovery_generations);
2301out_vol_label:
2302	kfree(osb->vol_label);
2303out_recovery_map:
2304	kfree(osb->recovery_map);
2305out:
2306	kfree(osb);
2307	sb->s_fs_info = NULL;
2308	return status;
2309}
2310
2311/*
2312 * will return: -EAGAIN if it is ok to keep searching for superblocks
2313 *              -EINVAL if there is a bad superblock
2314 *              0 on success
2315 */
2316static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2317			       struct buffer_head *bh,
2318			       u32 blksz,
2319			       struct ocfs2_blockcheck_stats *stats)
2320{
2321	int status = -EAGAIN;
2322
2323	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2324		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2325		/* We have to do a raw check of the feature here */
2326		if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2327		    OCFS2_FEATURE_INCOMPAT_META_ECC) {
2328			status = ocfs2_block_check_validate(bh->b_data,
2329							    bh->b_size,
2330							    &di->i_check,
2331							    stats);
2332			if (status)
2333				goto out;
2334		}
2335		status = -EINVAL;
2336		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2337			mlog(ML_ERROR, "found superblock with incorrect block "
2338			     "size: found %u, should be %u\n",
2339			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2340			       blksz);
2341		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2342			   OCFS2_MAJOR_REV_LEVEL ||
2343			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2344			   OCFS2_MINOR_REV_LEVEL) {
2345			mlog(ML_ERROR, "found superblock with bad version: "
2346			     "found %u.%u, should be %u.%u\n",
2347			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
2348			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2349			     OCFS2_MAJOR_REV_LEVEL,
2350			     OCFS2_MINOR_REV_LEVEL);
2351		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2352			mlog(ML_ERROR, "bad block number on superblock: "
2353			     "found %llu, should be %llu\n",
2354			     (unsigned long long)le64_to_cpu(di->i_blkno),
2355			     (unsigned long long)bh->b_blocknr);
2356		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2357			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2358			mlog(ML_ERROR, "bad cluster size found: %u\n",
2359			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2360		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2361			mlog(ML_ERROR, "bad root_blkno: 0\n");
2362		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2363			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2364		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2365			mlog(ML_ERROR,
2366			     "Superblock slots found greater than file system "
2367			     "maximum: found %u, max %u\n",
2368			     le16_to_cpu(di->id2.i_super.s_max_slots),
2369			     OCFS2_MAX_SLOTS);
2370		} else {
2371			/* found it! */
2372			status = 0;
2373		}
2374	}
2375
2376out:
2377	if (status && status != -EAGAIN)
2378		mlog_errno(status);
2379	return status;
2380}
2381
2382static int ocfs2_check_volume(struct ocfs2_super *osb)
2383{
2384	int status;
2385	int dirty;
2386	int local;
2387	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2388						  * recover
2389						  * ourselves. */
2390
2391	/* Init our journal object. */
2392	status = ocfs2_journal_init(osb, &dirty);
2393	if (status < 0) {
2394		mlog(ML_ERROR, "Could not initialize journal!\n");
2395		goto finally;
2396	}
2397
2398	/* Now that journal has been initialized, check to make sure
2399	   entire volume is addressable. */
2400	status = ocfs2_journal_addressable(osb);
2401	if (status)
2402		goto finally;
2403
2404	/* If the journal was unmounted cleanly then we don't want to
2405	 * recover anything. Otherwise, journal_load will do that
2406	 * dirty work for us :) */
2407	if (!dirty) {
2408		status = ocfs2_journal_wipe(osb->journal, 0);
2409		if (status < 0) {
2410			mlog_errno(status);
2411			goto finally;
2412		}
2413	} else {
2414		printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2415		       "unmounted cleanly, recovering it.\n", osb->dev_str);
2416	}
2417
2418	local = ocfs2_mount_local(osb);
2419
2420	/* will play back anything left in the journal. */
2421	status = ocfs2_journal_load(osb->journal, local, dirty);
2422	if (status < 0) {
2423		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2424		goto finally;
2425	}
2426
2427	if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2428		jbd2_journal_set_features(osb->journal->j_journal,
2429				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2430				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2431	else
2432		jbd2_journal_clear_features(osb->journal->j_journal,
2433				JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2434				JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2435
2436	if (dirty) {
2437		/* recover my local alloc if we didn't unmount cleanly. */
2438		status = ocfs2_begin_local_alloc_recovery(osb,
2439							  osb->slot_num,
2440							  &local_alloc);
2441		if (status < 0) {
2442			mlog_errno(status);
2443			goto finally;
2444		}
2445		/* we complete the recovery process after we've marked
2446		 * ourselves as mounted. */
2447	}
2448
2449	status = ocfs2_load_local_alloc(osb);
2450	if (status < 0) {
2451		mlog_errno(status);
2452		goto finally;
2453	}
2454
2455	if (dirty) {
2456		/* Recovery will be completed after we've mounted the
2457		 * rest of the volume. */
 
2458		osb->local_alloc_copy = local_alloc;
2459		local_alloc = NULL;
2460	}
2461
2462	/* go through each journal, trylock it and if you get the
2463	 * lock, and it's marked as dirty, set the bit in the recover
2464	 * map and launch a recovery thread for it. */
2465	status = ocfs2_mark_dead_nodes(osb);
2466	if (status < 0) {
2467		mlog_errno(status);
2468		goto finally;
2469	}
2470
2471	status = ocfs2_compute_replay_slots(osb);
2472	if (status < 0)
2473		mlog_errno(status);
2474
2475finally:
2476	kfree(local_alloc);
2477
2478	if (status)
2479		mlog_errno(status);
2480	return status;
2481}
2482
2483/*
2484 * The routine gets called from dismount or close whenever a dismount on
2485 * volume is requested and the osb open count becomes 1.
2486 * It will remove the osb from the global list and also free up all the
2487 * initialized resources and fileobject.
2488 */
2489static void ocfs2_delete_osb(struct ocfs2_super *osb)
2490{
2491	/* This function assumes that the caller has the main osb resource */
2492
2493	/* ocfs2_initializer_super have already created this workqueue */
2494	if (osb->ocfs2_wq)
2495		destroy_workqueue(osb->ocfs2_wq);
2496
2497	ocfs2_free_slot_info(osb);
2498
2499	kfree(osb->osb_orphan_wipes);
2500	kfree(osb->slot_recovery_generations);
2501	/* FIXME
2502	 * This belongs in journal shutdown, but because we have to
2503	 * allocate osb->journal at the middle of ocfs2_initialize_super(),
2504	 * we free it here.
2505	 */
2506	kfree(osb->journal);
2507	kfree(osb->local_alloc_copy);
2508	kfree(osb->uuid_str);
2509	kfree(osb->vol_label);
2510	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2511	memset(osb, 0, sizeof(struct ocfs2_super));
2512}
2513
2514/* Depending on the mount option passed, perform one of the following:
2515 * Put OCFS2 into a readonly state (default)
2516 * Return EIO so that only the process errs
2517 * Fix the error as if fsck.ocfs2 -y
2518 * panic
2519 */
2520static int ocfs2_handle_error(struct super_block *sb)
2521{
2522	struct ocfs2_super *osb = OCFS2_SB(sb);
2523	int rv = 0;
2524
2525	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2526	pr_crit("On-disk corruption discovered. "
2527		"Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2528
2529	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2530		panic("OCFS2: (device %s): panic forced after error\n",
2531		      sb->s_id);
2532	} else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2533		pr_crit("OCFS2: Returning error to the calling process.\n");
2534		rv = -EIO;
2535	} else { /* default option */
2536		rv = -EROFS;
2537		if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2538			return rv;
2539
2540		pr_crit("OCFS2: File system is now read-only.\n");
2541		sb->s_flags |= SB_RDONLY;
2542		ocfs2_set_ro_flag(osb, 0);
2543	}
2544
2545	return rv;
 
 
 
 
 
 
 
 
 
2546}
2547
2548int __ocfs2_error(struct super_block *sb, const char *function,
2549		  const char *fmt, ...)
 
 
 
2550{
2551	struct va_format vaf;
2552	va_list args;
2553
2554	va_start(args, fmt);
2555	vaf.fmt = fmt;
2556	vaf.va = &args;
2557
2558	/* Not using mlog here because we want to show the actual
2559	 * function the error came from. */
2560	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2561	       sb->s_id, function, &vaf);
2562
2563	va_end(args);
2564
2565	return ocfs2_handle_error(sb);
2566}
2567
2568/* Handle critical errors. This is intentionally more drastic than
2569 * ocfs2_handle_error, so we only use for things like journal errors,
2570 * etc. */
2571void __ocfs2_abort(struct super_block *sb, const char *function,
 
2572		   const char *fmt, ...)
2573{
2574	struct va_format vaf;
2575	va_list args;
2576
2577	va_start(args, fmt);
2578
2579	vaf.fmt = fmt;
2580	vaf.va = &args;
2581
2582	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2583	       sb->s_id, function, &vaf);
2584
2585	va_end(args);
 
 
 
2586
2587	/* We don't have the cluster support yet to go straight to
2588	 * hard readonly in here. Until then, we want to keep
2589	 * ocfs2_abort() so that we can at least mark critical
2590	 * errors.
2591	 *
2592	 * TODO: This should abort the journal and alert other nodes
2593	 * that our slot needs recovery. */
2594
2595	/* Force a panic(). This stinks, but it's better than letting
2596	 * things continue without having a proper hard readonly
2597	 * here. */
2598	if (!ocfs2_mount_local(OCFS2_SB(sb)))
2599		OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2600	ocfs2_handle_error(sb);
2601}
2602
2603/*
2604 * Void signal blockers, because in-kernel sigprocmask() only fails
2605 * when SIG_* is wrong.
2606 */
2607void ocfs2_block_signals(sigset_t *oldset)
2608{
2609	int rc;
2610	sigset_t blocked;
2611
2612	sigfillset(&blocked);
2613	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2614	BUG_ON(rc);
2615}
2616
2617void ocfs2_unblock_signals(sigset_t *oldset)
2618{
2619	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2620	BUG_ON(rc);
2621}
2622
2623module_init(ocfs2_init);
2624module_exit(ocfs2_exit);
v3.15
   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/cleancache.h>
  45
  46#define CREATE_TRACE_POINTS
  47#include "ocfs2_trace.h"
  48
  49#include <cluster/masklog.h>
  50
  51#include "ocfs2.h"
  52
  53/* this should be the only file to include a version 1 header */
  54#include "ocfs1_fs_compat.h"
  55
  56#include "alloc.h"
  57#include "aops.h"
  58#include "blockcheck.h"
  59#include "dlmglue.h"
  60#include "export.h"
  61#include "extent_map.h"
  62#include "heartbeat.h"
  63#include "inode.h"
  64#include "journal.h"
  65#include "localalloc.h"
  66#include "namei.h"
  67#include "slot_map.h"
  68#include "super.h"
  69#include "sysfile.h"
  70#include "uptodate.h"
  71#include "xattr.h"
  72#include "quota.h"
  73#include "refcounttree.h"
  74#include "suballoc.h"
  75
  76#include "buffer_head_io.h"
 
  77
  78static struct kmem_cache *ocfs2_inode_cachep = NULL;
  79struct kmem_cache *ocfs2_dquot_cachep;
  80struct kmem_cache *ocfs2_qf_chunk_cachep;
  81
  82/* OCFS2 needs to schedule several different types of work which
  83 * require cluster locking, disk I/O, recovery waits, etc. Since these
  84 * types of work tend to be heavy we avoid using the kernel events
  85 * workqueue and schedule on our own. */
  86struct workqueue_struct *ocfs2_wq = NULL;
  87
  88static struct dentry *ocfs2_debugfs_root = NULL;
  89
  90MODULE_AUTHOR("Oracle");
  91MODULE_LICENSE("GPL");
  92MODULE_DESCRIPTION("OCFS2 cluster file system");
  93
  94struct mount_options
  95{
  96	unsigned long	commit_interval;
  97	unsigned long	mount_opt;
  98	unsigned int	atime_quantum;
  99	signed short	slot;
 100	int		localalloc_opt;
 101	unsigned int	resv_level;
 102	int		dir_resv_level;
 103	char		cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 104};
 105
 106static int ocfs2_parse_options(struct super_block *sb, char *options,
 107			       struct mount_options *mopt,
 108			       int is_remount);
 109static int ocfs2_check_set_options(struct super_block *sb,
 110				   struct mount_options *options);
 111static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
 112static void ocfs2_put_super(struct super_block *sb);
 113static int ocfs2_mount_volume(struct super_block *sb);
 114static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 115static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 116static int ocfs2_initialize_mem_caches(void);
 117static void ocfs2_free_mem_caches(void);
 118static void ocfs2_delete_osb(struct ocfs2_super *osb);
 119
 120static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 121
 122static int ocfs2_sync_fs(struct super_block *sb, int wait);
 123
 124static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 125static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 126static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 127static int ocfs2_check_volume(struct ocfs2_super *osb);
 128static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 129			       struct buffer_head *bh,
 130			       u32 sectsize,
 131			       struct ocfs2_blockcheck_stats *stats);
 132static int ocfs2_initialize_super(struct super_block *sb,
 133				  struct buffer_head *bh,
 134				  int sector_size,
 135				  struct ocfs2_blockcheck_stats *stats);
 136static int ocfs2_get_sector(struct super_block *sb,
 137			    struct buffer_head **bh,
 138			    int block,
 139			    int sect_size);
 140static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 141static void ocfs2_destroy_inode(struct inode *inode);
 142static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 143static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 144static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 145
 
 
 
 
 
 146static const struct super_operations ocfs2_sops = {
 147	.statfs		= ocfs2_statfs,
 148	.alloc_inode	= ocfs2_alloc_inode,
 149	.destroy_inode	= ocfs2_destroy_inode,
 150	.drop_inode	= ocfs2_drop_inode,
 151	.evict_inode	= ocfs2_evict_inode,
 152	.sync_fs	= ocfs2_sync_fs,
 153	.put_super	= ocfs2_put_super,
 154	.remount_fs	= ocfs2_remount,
 155	.show_options   = ocfs2_show_options,
 156	.quota_read	= ocfs2_quota_read,
 157	.quota_write	= ocfs2_quota_write,
 
 158};
 159
 160enum {
 161	Opt_barrier,
 162	Opt_err_panic,
 163	Opt_err_ro,
 164	Opt_intr,
 165	Opt_nointr,
 166	Opt_hb_none,
 167	Opt_hb_local,
 168	Opt_hb_global,
 169	Opt_data_ordered,
 170	Opt_data_writeback,
 171	Opt_atime_quantum,
 172	Opt_slot,
 173	Opt_commit,
 174	Opt_localalloc,
 175	Opt_localflocks,
 176	Opt_stack,
 177	Opt_user_xattr,
 178	Opt_nouser_xattr,
 179	Opt_inode64,
 180	Opt_acl,
 181	Opt_noacl,
 182	Opt_usrquota,
 183	Opt_grpquota,
 184	Opt_coherency_buffered,
 185	Opt_coherency_full,
 186	Opt_resv_level,
 187	Opt_dir_resv_level,
 
 
 188	Opt_err,
 189};
 190
 191static const match_table_t tokens = {
 192	{Opt_barrier, "barrier=%u"},
 193	{Opt_err_panic, "errors=panic"},
 194	{Opt_err_ro, "errors=remount-ro"},
 195	{Opt_intr, "intr"},
 196	{Opt_nointr, "nointr"},
 197	{Opt_hb_none, OCFS2_HB_NONE},
 198	{Opt_hb_local, OCFS2_HB_LOCAL},
 199	{Opt_hb_global, OCFS2_HB_GLOBAL},
 200	{Opt_data_ordered, "data=ordered"},
 201	{Opt_data_writeback, "data=writeback"},
 202	{Opt_atime_quantum, "atime_quantum=%u"},
 203	{Opt_slot, "preferred_slot=%u"},
 204	{Opt_commit, "commit=%u"},
 205	{Opt_localalloc, "localalloc=%d"},
 206	{Opt_localflocks, "localflocks"},
 207	{Opt_stack, "cluster_stack=%s"},
 208	{Opt_user_xattr, "user_xattr"},
 209	{Opt_nouser_xattr, "nouser_xattr"},
 210	{Opt_inode64, "inode64"},
 211	{Opt_acl, "acl"},
 212	{Opt_noacl, "noacl"},
 213	{Opt_usrquota, "usrquota"},
 214	{Opt_grpquota, "grpquota"},
 215	{Opt_coherency_buffered, "coherency=buffered"},
 216	{Opt_coherency_full, "coherency=full"},
 217	{Opt_resv_level, "resv_level=%u"},
 218	{Opt_dir_resv_level, "dir_resv_level=%u"},
 
 
 219	{Opt_err, NULL}
 220};
 221
 222#ifdef CONFIG_DEBUG_FS
 223static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 224{
 225	struct ocfs2_cluster_connection *cconn = osb->cconn;
 226	struct ocfs2_recovery_map *rm = osb->recovery_map;
 227	struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 228	int i, out = 0;
 
 229
 230	out += snprintf(buf + out, len - out,
 231			"%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 232			"Device", osb->dev_str, osb->uuid_str,
 233			osb->fs_generation, osb->vol_label);
 234
 235	out += snprintf(buf + out, len - out,
 236			"%10s => State: %d  Flags: 0x%lX\n", "Volume",
 237			atomic_read(&osb->vol_state), osb->osb_flags);
 238
 239	out += snprintf(buf + out, len - out,
 240			"%10s => Block: %lu  Cluster: %d\n", "Sizes",
 241			osb->sb->s_blocksize, osb->s_clustersize);
 242
 243	out += snprintf(buf + out, len - out,
 244			"%10s => Compat: 0x%X  Incompat: 0x%X  "
 245			"ROcompat: 0x%X\n",
 246			"Features", osb->s_feature_compat,
 247			osb->s_feature_incompat, osb->s_feature_ro_compat);
 248
 249	out += snprintf(buf + out, len - out,
 250			"%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 251			osb->s_mount_opt, osb->s_atime_quantum);
 252
 253	if (cconn) {
 254		out += snprintf(buf + out, len - out,
 255				"%10s => Stack: %s  Name: %*s  "
 256				"Version: %d.%d\n", "Cluster",
 257				(*osb->osb_cluster_stack == '\0' ?
 258				 "o2cb" : osb->osb_cluster_stack),
 259				cconn->cc_namelen, cconn->cc_name,
 260				cconn->cc_version.pv_major,
 261				cconn->cc_version.pv_minor);
 262	}
 263
 264	spin_lock(&osb->dc_task_lock);
 265	out += snprintf(buf + out, len - out,
 266			"%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 267			"WorkSeq: %lu\n", "DownCnvt",
 268			(osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 269			osb->blocked_lock_count, osb->dc_wake_sequence,
 270			osb->dc_work_sequence);
 271	spin_unlock(&osb->dc_task_lock);
 272
 273	spin_lock(&osb->osb_lock);
 274	out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 275			"Recovery",
 276			(osb->recovery_thread_task ?
 277			 task_pid_nr(osb->recovery_thread_task) : -1));
 278	if (rm->rm_used == 0)
 279		out += snprintf(buf + out, len - out, " None\n");
 280	else {
 281		for (i = 0; i < rm->rm_used; i++)
 282			out += snprintf(buf + out, len - out, " %d",
 283					rm->rm_entries[i]);
 284		out += snprintf(buf + out, len - out, "\n");
 285	}
 286	spin_unlock(&osb->osb_lock);
 287
 288	out += snprintf(buf + out, len - out,
 289			"%10s => Pid: %d  Interval: %lu\n", "Commit",
 290			(osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 291			osb->osb_commit_interval);
 292
 293	out += snprintf(buf + out, len - out,
 294			"%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 295			"Journal", osb->journal->j_state,
 296			osb->journal->j_trans_id,
 297			atomic_read(&osb->journal->j_num_trans));
 298
 299	out += snprintf(buf + out, len - out,
 300			"%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 301			"SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 302			"Stats",
 303			atomic_read(&osb->alloc_stats.bitmap_data),
 304			atomic_read(&osb->alloc_stats.local_data),
 305			atomic_read(&osb->alloc_stats.bg_allocs),
 306			atomic_read(&osb->alloc_stats.moves),
 307			atomic_read(&osb->alloc_stats.bg_extends));
 308
 309	out += snprintf(buf + out, len - out,
 310			"%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 311			"Default: %u bits\n",
 312			"LocalAlloc", osb->local_alloc_state,
 313			(unsigned long long)osb->la_last_gd,
 314			osb->local_alloc_bits, osb->local_alloc_default_bits);
 315
 316	spin_lock(&osb->osb_lock);
 317	out += snprintf(buf + out, len - out,
 318			"%10s => InodeSlot: %d  StolenInodes: %d, "
 319			"MetaSlot: %d  StolenMeta: %d\n", "Steal",
 320			osb->s_inode_steal_slot,
 321			atomic_read(&osb->s_num_inodes_stolen),
 322			osb->s_meta_steal_slot,
 323			atomic_read(&osb->s_num_meta_stolen));
 324	spin_unlock(&osb->osb_lock);
 325
 326	out += snprintf(buf + out, len - out, "OrphanScan => ");
 327	out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 328			os->os_count, os->os_seqno);
 329	out += snprintf(buf + out, len - out, " Last Scan: ");
 330	if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 331		out += snprintf(buf + out, len - out, "Disabled\n");
 332	else
 333		out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 334				(get_seconds() - os->os_scantime.tv_sec));
 335
 336	out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 337			"Slots", "Num", "RecoGen");
 338	for (i = 0; i < osb->max_slots; ++i) {
 339		out += snprintf(buf + out, len - out,
 340				"%10s  %c %3d  %10d\n",
 341				" ",
 342				(i == osb->slot_num ? '*' : ' '),
 343				i, osb->slot_recovery_generations[i]);
 344	}
 345
 346	return out;
 347}
 348
 349static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 350{
 351	struct ocfs2_super *osb = inode->i_private;
 352	char *buf = NULL;
 353
 354	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 355	if (!buf)
 356		goto bail;
 357
 358	i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 359
 360	file->private_data = buf;
 361
 362	return 0;
 363bail:
 364	return -ENOMEM;
 365}
 366
 367static int ocfs2_debug_release(struct inode *inode, struct file *file)
 368{
 369	kfree(file->private_data);
 370	return 0;
 371}
 372
 373static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 374				size_t nbytes, loff_t *ppos)
 375{
 376	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 377				       i_size_read(file->f_mapping->host));
 378}
 379#else
 380static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 381{
 382	return 0;
 383}
 384static int ocfs2_debug_release(struct inode *inode, struct file *file)
 385{
 386	return 0;
 387}
 388static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 389				size_t nbytes, loff_t *ppos)
 390{
 391	return 0;
 392}
 393#endif	/* CONFIG_DEBUG_FS */
 394
 395static const struct file_operations ocfs2_osb_debug_fops = {
 396	.open =		ocfs2_osb_debug_open,
 397	.release =	ocfs2_debug_release,
 398	.read =		ocfs2_debug_read,
 399	.llseek =	generic_file_llseek,
 400};
 401
 402static int ocfs2_sync_fs(struct super_block *sb, int wait)
 403{
 404	int status;
 405	tid_t target;
 406	struct ocfs2_super *osb = OCFS2_SB(sb);
 407
 408	if (ocfs2_is_hard_readonly(osb))
 409		return -EROFS;
 410
 411	if (wait) {
 412		status = ocfs2_flush_truncate_log(osb);
 413		if (status < 0)
 414			mlog_errno(status);
 415	} else {
 416		ocfs2_schedule_truncate_log_flush(osb, 0);
 417	}
 418
 419	if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 420				      &target)) {
 421		if (wait)
 422			jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 423					     target);
 424	}
 425	return 0;
 426}
 427
 428static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 429{
 430	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 431	    && (ino == USER_QUOTA_SYSTEM_INODE
 432		|| ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 433		return 0;
 434	if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 435	    && (ino == GROUP_QUOTA_SYSTEM_INODE
 436		|| ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 437		return 0;
 438	return 1;
 439}
 440
 441static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 442{
 443	struct inode *new = NULL;
 444	int status = 0;
 445	int i;
 446
 447	new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 448	if (IS_ERR(new)) {
 449		status = PTR_ERR(new);
 450		mlog_errno(status);
 451		goto bail;
 452	}
 453	osb->root_inode = new;
 454
 455	new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 456	if (IS_ERR(new)) {
 457		status = PTR_ERR(new);
 458		mlog_errno(status);
 459		goto bail;
 460	}
 461	osb->sys_root_inode = new;
 462
 463	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 464	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 465		if (!ocfs2_need_system_inode(osb, i))
 466			continue;
 467		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 468		if (!new) {
 469			ocfs2_release_system_inodes(osb);
 470			status = -EINVAL;
 471			mlog_errno(status);
 472			/* FIXME: Should ERROR_RO_FS */
 473			mlog(ML_ERROR, "Unable to load system inode %d, "
 474			     "possibly corrupt fs?", i);
 475			goto bail;
 476		}
 477		// the array now has one ref, so drop this one
 478		iput(new);
 479	}
 480
 481bail:
 482	if (status)
 483		mlog_errno(status);
 484	return status;
 485}
 486
 487static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 488{
 489	struct inode *new = NULL;
 490	int status = 0;
 491	int i;
 492
 493	for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 494	     i < NUM_SYSTEM_INODES;
 495	     i++) {
 496		if (!ocfs2_need_system_inode(osb, i))
 497			continue;
 498		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 499		if (!new) {
 500			ocfs2_release_system_inodes(osb);
 501			status = -EINVAL;
 502			mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 503			     status, i, osb->slot_num);
 504			goto bail;
 505		}
 506		/* the array now has one ref, so drop this one */
 507		iput(new);
 508	}
 509
 510bail:
 511	if (status)
 512		mlog_errno(status);
 513	return status;
 514}
 515
 516static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 517{
 518	int i;
 519	struct inode *inode;
 520
 521	for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 522		inode = osb->global_system_inodes[i];
 523		if (inode) {
 524			iput(inode);
 525			osb->global_system_inodes[i] = NULL;
 526		}
 527	}
 528
 529	inode = osb->sys_root_inode;
 530	if (inode) {
 531		iput(inode);
 532		osb->sys_root_inode = NULL;
 533	}
 534
 535	inode = osb->root_inode;
 536	if (inode) {
 537		iput(inode);
 538		osb->root_inode = NULL;
 539	}
 540
 541	if (!osb->local_system_inodes)
 542		return;
 543
 544	for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 545		if (osb->local_system_inodes[i]) {
 546			iput(osb->local_system_inodes[i]);
 547			osb->local_system_inodes[i] = NULL;
 548		}
 549	}
 550
 551	kfree(osb->local_system_inodes);
 552	osb->local_system_inodes = NULL;
 553}
 554
 555/* We're allocating fs objects, use GFP_NOFS */
 556static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 557{
 558	struct ocfs2_inode_info *oi;
 559
 560	oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 561	if (!oi)
 562		return NULL;
 563
 564	oi->i_sync_tid = 0;
 565	oi->i_datasync_tid = 0;
 
 566
 567	jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 568	return &oi->vfs_inode;
 569}
 570
 571static void ocfs2_i_callback(struct rcu_head *head)
 572{
 573	struct inode *inode = container_of(head, struct inode, i_rcu);
 574	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 575}
 576
 577static void ocfs2_destroy_inode(struct inode *inode)
 578{
 579	call_rcu(&inode->i_rcu, ocfs2_i_callback);
 580}
 581
 582static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 583						unsigned int cbits)
 584{
 585	unsigned int bytes = 1 << cbits;
 586	unsigned int trim = bytes;
 587	unsigned int bitshift = 32;
 588
 589	/*
 590	 * i_size and all block offsets in ocfs2 are always 64 bits
 591	 * wide. i_clusters is 32 bits, in cluster-sized units. So on
 592	 * 64 bit platforms, cluster size will be the limiting factor.
 593	 */
 594
 595#if BITS_PER_LONG == 32
 596# if defined(CONFIG_LBDAF)
 597	BUILD_BUG_ON(sizeof(sector_t) != 8);
 598	/*
 599	 * We might be limited by page cache size.
 600	 */
 601	if (bytes > PAGE_CACHE_SIZE) {
 602		bytes = PAGE_CACHE_SIZE;
 603		trim = 1;
 604		/*
 605		 * Shift by 31 here so that we don't get larger than
 606		 * MAX_LFS_FILESIZE
 607		 */
 608		bitshift = 31;
 609	}
 610# else
 611	/*
 612	 * We are limited by the size of sector_t. Use block size, as
 613	 * that's what we expose to the VFS.
 614	 */
 615	bytes = 1 << bbits;
 616	trim = 1;
 617	bitshift = 31;
 618# endif
 619#endif
 620
 621	/*
 622	 * Trim by a whole cluster when we can actually approach the
 623	 * on-disk limits. Otherwise we can overflow i_clusters when
 624	 * an extent start is at the max offset.
 625	 */
 626	return (((unsigned long long)bytes) << bitshift) - trim;
 627}
 628
 629static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 630{
 631	int incompat_features;
 632	int ret = 0;
 633	struct mount_options parsed_options;
 634	struct ocfs2_super *osb = OCFS2_SB(sb);
 635	u32 tmp;
 636
 637	sync_filesystem(sb);
 638
 639	if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 640	    !ocfs2_check_set_options(sb, &parsed_options)) {
 641		ret = -EINVAL;
 642		goto out;
 643	}
 644
 645	tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 646		OCFS2_MOUNT_HB_NONE;
 647	if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 648		ret = -EINVAL;
 649		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 650		goto out;
 651	}
 652
 653	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 654	    (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 655		ret = -EINVAL;
 656		mlog(ML_ERROR, "Cannot change data mode on remount\n");
 657		goto out;
 658	}
 659
 660	/* Probably don't want this on remount; it might
 661	 * mess with other nodes */
 662	if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 663	    (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 664		ret = -EINVAL;
 665		mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 666		goto out;
 667	}
 668
 669	/* We're going to/from readonly mode. */
 670	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 671		/* Disable quota accounting before remounting RO */
 672		if (*flags & MS_RDONLY) {
 673			ret = ocfs2_susp_quotas(osb, 0);
 674			if (ret < 0)
 675				goto out;
 676		}
 677		/* Lock here so the check of HARD_RO and the potential
 678		 * setting of SOFT_RO is atomic. */
 679		spin_lock(&osb->osb_lock);
 680		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 681			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 682			ret = -EROFS;
 683			goto unlock_osb;
 684		}
 685
 686		if (*flags & MS_RDONLY) {
 687			sb->s_flags |= MS_RDONLY;
 688			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 689		} else {
 690			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 691				mlog(ML_ERROR, "Cannot remount RDWR "
 692				     "filesystem due to previous errors.\n");
 693				ret = -EROFS;
 694				goto unlock_osb;
 695			}
 696			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 697			if (incompat_features) {
 698				mlog(ML_ERROR, "Cannot remount RDWR because "
 699				     "of unsupported optional features "
 700				     "(%x).\n", incompat_features);
 701				ret = -EINVAL;
 702				goto unlock_osb;
 703			}
 704			sb->s_flags &= ~MS_RDONLY;
 705			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 706		}
 707		trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
 708unlock_osb:
 709		spin_unlock(&osb->osb_lock);
 710		/* Enable quota accounting after remounting RW */
 711		if (!ret && !(*flags & MS_RDONLY)) {
 712			if (sb_any_quota_suspended(sb))
 713				ret = ocfs2_susp_quotas(osb, 1);
 714			else
 715				ret = ocfs2_enable_quotas(osb);
 716			if (ret < 0) {
 717				/* Return back changes... */
 718				spin_lock(&osb->osb_lock);
 719				sb->s_flags |= MS_RDONLY;
 720				osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 721				spin_unlock(&osb->osb_lock);
 722				goto out;
 723			}
 724		}
 725	}
 726
 727	if (!ret) {
 728		/* Only save off the new mount options in case of a successful
 729		 * remount. */
 730		osb->s_mount_opt = parsed_options.mount_opt;
 731		osb->s_atime_quantum = parsed_options.atime_quantum;
 732		osb->preferred_slot = parsed_options.slot;
 733		if (parsed_options.commit_interval)
 734			osb->osb_commit_interval = parsed_options.commit_interval;
 735
 736		if (!ocfs2_is_hard_readonly(osb))
 737			ocfs2_set_journal_params(osb);
 738
 739		sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 740			((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 741							MS_POSIXACL : 0);
 742	}
 743out:
 744	return ret;
 745}
 746
 747static int ocfs2_sb_probe(struct super_block *sb,
 748			  struct buffer_head **bh,
 749			  int *sector_size,
 750			  struct ocfs2_blockcheck_stats *stats)
 751{
 752	int status, tmpstat;
 753	struct ocfs1_vol_disk_hdr *hdr;
 754	struct ocfs2_dinode *di;
 755	int blksize;
 756
 757	*bh = NULL;
 758
 759	/* may be > 512 */
 760	*sector_size = bdev_logical_block_size(sb->s_bdev);
 761	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 762		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 763		     *sector_size, OCFS2_MAX_BLOCKSIZE);
 764		status = -EINVAL;
 765		goto bail;
 766	}
 767
 768	/* Can this really happen? */
 769	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 770		*sector_size = OCFS2_MIN_BLOCKSIZE;
 771
 772	/* check block zero for old format */
 773	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 774	if (status < 0) {
 775		mlog_errno(status);
 776		goto bail;
 777	}
 778	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 779	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 780		mlog(ML_ERROR, "incompatible version: %u.%u\n",
 781		     hdr->major_version, hdr->minor_version);
 782		status = -EINVAL;
 783	}
 784	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 785		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 786		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 787		     hdr->signature);
 788		status = -EINVAL;
 789	}
 790	brelse(*bh);
 791	*bh = NULL;
 792	if (status < 0) {
 793		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 794		     "upgraded before mounting with ocfs v2\n");
 795		goto bail;
 796	}
 797
 798	/*
 799	 * Now check at magic offset for 512, 1024, 2048, 4096
 800	 * blocksizes.  4096 is the maximum blocksize because it is
 801	 * the minimum clustersize.
 802	 */
 803	status = -EINVAL;
 804	for (blksize = *sector_size;
 805	     blksize <= OCFS2_MAX_BLOCKSIZE;
 806	     blksize <<= 1) {
 807		tmpstat = ocfs2_get_sector(sb, bh,
 808					   OCFS2_SUPER_BLOCK_BLKNO,
 809					   blksize);
 810		if (tmpstat < 0) {
 811			status = tmpstat;
 812			mlog_errno(status);
 813			break;
 814		}
 815		di = (struct ocfs2_dinode *) (*bh)->b_data;
 816		memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 817		spin_lock_init(&stats->b_lock);
 818		tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 819		if (tmpstat < 0) {
 820			brelse(*bh);
 821			*bh = NULL;
 822		}
 823		if (tmpstat != -EAGAIN) {
 824			status = tmpstat;
 825			break;
 826		}
 827	}
 828
 829bail:
 830	return status;
 831}
 832
 833static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 834{
 835	u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 836
 837	if (osb->s_mount_opt & hb_enabled) {
 838		if (ocfs2_mount_local(osb)) {
 839			mlog(ML_ERROR, "Cannot heartbeat on a locally "
 840			     "mounted device.\n");
 841			return -EINVAL;
 842		}
 843		if (ocfs2_userspace_stack(osb)) {
 844			mlog(ML_ERROR, "Userspace stack expected, but "
 845			     "o2cb heartbeat arguments passed to mount\n");
 846			return -EINVAL;
 847		}
 848		if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 849		     !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 850		    ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 851		     ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 852			mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 853			return -EINVAL;
 854		}
 855	}
 856
 857	if (!(osb->s_mount_opt & hb_enabled)) {
 858		if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 859		    !ocfs2_userspace_stack(osb)) {
 860			mlog(ML_ERROR, "Heartbeat has to be started to mount "
 861			     "a read-write clustered device.\n");
 862			return -EINVAL;
 863		}
 864	}
 865
 866	return 0;
 867}
 868
 869/*
 870 * If we're using a userspace stack, mount should have passed
 871 * a name that matches the disk.  If not, mount should not
 872 * have passed a stack.
 873 */
 874static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 875					struct mount_options *mopt)
 876{
 877	if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 878		mlog(ML_ERROR,
 879		     "cluster stack passed to mount, but this filesystem "
 880		     "does not support it\n");
 881		return -EINVAL;
 882	}
 883
 884	if (ocfs2_userspace_stack(osb) &&
 885	    strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 886		    OCFS2_STACK_LABEL_LEN)) {
 887		mlog(ML_ERROR,
 888		     "cluster stack passed to mount (\"%s\") does not "
 889		     "match the filesystem (\"%s\")\n",
 890		     mopt->cluster_stack,
 891		     osb->osb_cluster_stack);
 892		return -EINVAL;
 893	}
 894
 895	return 0;
 896}
 897
 898static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 899{
 900	int type;
 901	struct super_block *sb = osb->sb;
 902	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 903					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 
 904	int status = 0;
 905
 906	for (type = 0; type < MAXQUOTAS; type++) {
 907		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 908			continue;
 909		if (unsuspend)
 910			status = dquot_resume(sb, type);
 911		else {
 912			struct ocfs2_mem_dqinfo *oinfo;
 913
 914			/* Cancel periodic syncing before suspending */
 915			oinfo = sb_dqinfo(sb, type)->dqi_priv;
 916			cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 917			status = dquot_suspend(sb, type);
 918		}
 919		if (status < 0)
 920			break;
 921	}
 922	if (status < 0)
 923		mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 924		     "remount (error = %d).\n", status);
 925	return status;
 926}
 927
 928static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 929{
 930	struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 931	struct super_block *sb = osb->sb;
 932	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 933					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 934	unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 
 
 935					LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 936	int status;
 937	int type;
 938
 939	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 940	for (type = 0; type < MAXQUOTAS; type++) {
 941		if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 942			continue;
 943		inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 944							osb->slot_num);
 945		if (!inode[type]) {
 946			status = -ENOENT;
 947			goto out_quota_off;
 948		}
 949		status = dquot_enable(inode[type], type, QFMT_OCFS2,
 950				      DQUOT_USAGE_ENABLED);
 951		if (status < 0)
 952			goto out_quota_off;
 953	}
 954
 955	for (type = 0; type < MAXQUOTAS; type++)
 956		iput(inode[type]);
 957	return 0;
 958out_quota_off:
 959	ocfs2_disable_quotas(osb);
 960	for (type = 0; type < MAXQUOTAS; type++)
 961		iput(inode[type]);
 962	mlog_errno(status);
 963	return status;
 964}
 965
 966static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 967{
 968	int type;
 969	struct inode *inode;
 970	struct super_block *sb = osb->sb;
 971	struct ocfs2_mem_dqinfo *oinfo;
 972
 973	/* We mostly ignore errors in this function because there's not much
 974	 * we can do when we see them */
 975	for (type = 0; type < MAXQUOTAS; type++) {
 976		if (!sb_has_quota_loaded(sb, type))
 977			continue;
 978		/* Cancel periodic syncing before we grab dqonoff_mutex */
 979		oinfo = sb_dqinfo(sb, type)->dqi_priv;
 980		cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 981		inode = igrab(sb->s_dquot.files[type]);
 982		/* Turn off quotas. This will remove all dquot structures from
 983		 * memory and so they will be automatically synced to global
 984		 * quota files */
 985		dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 986					DQUOT_LIMITS_ENABLED);
 987		if (!inode)
 988			continue;
 989		iput(inode);
 990	}
 991}
 992
 993/* Handle quota on quotactl */
 994static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
 995{
 996	unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 997					     OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 998
 999	if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
1000		return -EINVAL;
1001
1002	return dquot_enable(sb_dqopt(sb)->files[type], type,
1003			    format_id, DQUOT_LIMITS_ENABLED);
1004}
1005
1006/* Handle quota off quotactl */
1007static int ocfs2_quota_off(struct super_block *sb, int type)
1008{
1009	return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1010}
1011
1012static const struct quotactl_ops ocfs2_quotactl_ops = {
1013	.quota_on_meta	= ocfs2_quota_on,
1014	.quota_off	= ocfs2_quota_off,
1015	.quota_sync	= dquot_quota_sync,
1016	.get_info	= dquot_get_dqinfo,
1017	.set_info	= dquot_set_dqinfo,
1018	.get_dqblk	= dquot_get_dqblk,
1019	.set_dqblk	= dquot_set_dqblk,
1020};
1021
1022static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1023{
1024	struct dentry *root;
1025	int status, sector_size;
1026	struct mount_options parsed_options;
1027	struct inode *inode = NULL;
1028	struct ocfs2_super *osb = NULL;
1029	struct buffer_head *bh = NULL;
1030	char nodestr[12];
1031	struct ocfs2_blockcheck_stats stats;
1032
1033	trace_ocfs2_fill_super(sb, data, silent);
1034
1035	if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1036		status = -EINVAL;
1037		goto read_super_error;
1038	}
1039
1040	/* probe for superblock */
1041	status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1042	if (status < 0) {
1043		mlog(ML_ERROR, "superblock probe failed!\n");
1044		goto read_super_error;
1045	}
1046
1047	status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1048	osb = OCFS2_SB(sb);
1049	if (status < 0) {
1050		mlog_errno(status);
1051		goto read_super_error;
1052	}
1053	brelse(bh);
1054	bh = NULL;
 
 
 
 
1055
1056	if (!ocfs2_check_set_options(sb, &parsed_options)) {
1057		status = -EINVAL;
1058		goto read_super_error;
1059	}
1060	osb->s_mount_opt = parsed_options.mount_opt;
1061	osb->s_atime_quantum = parsed_options.atime_quantum;
1062	osb->preferred_slot = parsed_options.slot;
1063	osb->osb_commit_interval = parsed_options.commit_interval;
1064
1065	ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1066	osb->osb_resv_level = parsed_options.resv_level;
1067	osb->osb_dir_resv_level = parsed_options.resv_level;
1068	if (parsed_options.dir_resv_level == -1)
1069		osb->osb_dir_resv_level = parsed_options.resv_level;
1070	else
1071		osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1072
1073	status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1074	if (status)
1075		goto read_super_error;
1076
1077	sb->s_magic = OCFS2_SUPER_MAGIC;
1078
1079	sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1080		((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1081
1082	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1083	 * heartbeat=none */
1084	if (bdev_read_only(sb->s_bdev)) {
1085		if (!(sb->s_flags & MS_RDONLY)) {
1086			status = -EACCES;
1087			mlog(ML_ERROR, "Readonly device detected but readonly "
1088			     "mount was not specified.\n");
1089			goto read_super_error;
1090		}
1091
1092		/* You should not be able to start a local heartbeat
1093		 * on a readonly device. */
1094		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1095			status = -EROFS;
1096			mlog(ML_ERROR, "Local heartbeat specified on readonly "
1097			     "device.\n");
1098			goto read_super_error;
1099		}
1100
1101		status = ocfs2_check_journals_nolocks(osb);
1102		if (status < 0) {
1103			if (status == -EROFS)
1104				mlog(ML_ERROR, "Recovery required on readonly "
1105				     "file system, but write access is "
1106				     "unavailable.\n");
1107			else
1108				mlog_errno(status);
1109			goto read_super_error;
1110		}
1111
1112		ocfs2_set_ro_flag(osb, 1);
1113
1114		printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1115		       "Cluster services will not be used for this mount. "
1116		       "Recovery will be skipped.\n", osb->dev_str);
1117	}
1118
1119	if (!ocfs2_is_hard_readonly(osb)) {
1120		if (sb->s_flags & MS_RDONLY)
1121			ocfs2_set_ro_flag(osb, 0);
1122	}
1123
1124	status = ocfs2_verify_heartbeat(osb);
1125	if (status < 0) {
1126		mlog_errno(status);
1127		goto read_super_error;
1128	}
1129
1130	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1131						 ocfs2_debugfs_root);
1132	if (!osb->osb_debug_root) {
1133		status = -EINVAL;
1134		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1135		goto read_super_error;
1136	}
1137
1138	osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1139					    osb->osb_debug_root,
1140					    osb,
1141					    &ocfs2_osb_debug_fops);
1142	if (!osb->osb_ctxt) {
1143		status = -EINVAL;
1144		mlog_errno(status);
1145		goto read_super_error;
1146	}
1147
1148	if (ocfs2_meta_ecc(osb)) {
1149		status = ocfs2_blockcheck_stats_debugfs_install(
1150						&osb->osb_ecc_stats,
1151						osb->osb_debug_root);
1152		if (status) {
1153			mlog(ML_ERROR,
1154			     "Unable to create blockcheck statistics "
1155			     "files\n");
1156			goto read_super_error;
1157		}
1158	}
1159
1160	status = ocfs2_mount_volume(sb);
1161	if (status < 0)
1162		goto read_super_error;
1163
1164	if (osb->root_inode)
1165		inode = igrab(osb->root_inode);
1166
1167	if (!inode) {
1168		status = -EIO;
1169		mlog_errno(status);
1170		goto read_super_error;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1171	}
1172
1173	root = d_make_root(inode);
1174	if (!root) {
1175		status = -ENOMEM;
1176		mlog_errno(status);
1177		goto read_super_error;
1178	}
1179
1180	sb->s_root = root;
1181
1182	ocfs2_complete_mount_recovery(osb);
1183
1184	if (ocfs2_mount_local(osb))
1185		snprintf(nodestr, sizeof(nodestr), "local");
1186	else
1187		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1188
1189	printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1190	       "with %s data mode.\n",
1191	       osb->dev_str, nodestr, osb->slot_num,
1192	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1193	       "ordered");
1194
1195	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1196	wake_up(&osb->osb_mount_event);
1197
1198	/* Now we can initialize quotas because we can afford to wait
1199	 * for cluster locks recovery now. That also means that truncation
1200	 * log recovery can happen but that waits for proper quota setup */
1201	if (!(sb->s_flags & MS_RDONLY)) {
1202		status = ocfs2_enable_quotas(osb);
1203		if (status < 0) {
1204			/* We have to err-out specially here because
1205			 * s_root is already set */
1206			mlog_errno(status);
1207			atomic_set(&osb->vol_state, VOLUME_DISABLED);
1208			wake_up(&osb->osb_mount_event);
1209			return status;
1210		}
1211	}
1212
1213	ocfs2_complete_quota_recovery(osb);
1214
1215	/* Now we wake up again for processes waiting for quotas */
1216	atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1217	wake_up(&osb->osb_mount_event);
1218
1219	/* Start this when the mount is almost sure of being successful */
1220	ocfs2_orphan_scan_start(osb);
1221
1222	return status;
1223
1224read_super_error:
1225	brelse(bh);
1226
1227	if (osb) {
1228		atomic_set(&osb->vol_state, VOLUME_DISABLED);
1229		wake_up(&osb->osb_mount_event);
1230		ocfs2_dismount_volume(sb, 1);
1231	}
 
 
 
 
 
 
 
 
1232
1233	if (status)
1234		mlog_errno(status);
1235	return status;
1236}
1237
1238static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1239			int flags,
1240			const char *dev_name,
1241			void *data)
1242{
1243	return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1244}
1245
1246static struct file_system_type ocfs2_fs_type = {
1247	.owner          = THIS_MODULE,
1248	.name           = "ocfs2",
1249	.mount          = ocfs2_mount,
1250	.kill_sb        = kill_block_super,
1251	.fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1252	.next           = NULL
1253};
1254MODULE_ALIAS_FS("ocfs2");
1255
1256static int ocfs2_check_set_options(struct super_block *sb,
1257				   struct mount_options *options)
1258{
1259	if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1260	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1261					 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1262		mlog(ML_ERROR, "User quotas were requested, but this "
1263		     "filesystem does not have the feature enabled.\n");
1264		return 0;
1265	}
1266	if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1267	    !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1268					 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1269		mlog(ML_ERROR, "Group quotas were requested, but this "
1270		     "filesystem does not have the feature enabled.\n");
1271		return 0;
1272	}
1273	if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1274	    !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1275		mlog(ML_ERROR, "ACL support requested but extended attributes "
1276		     "feature is not enabled\n");
1277		return 0;
1278	}
1279	/* No ACL setting specified? Use XATTR feature... */
1280	if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1281				    OCFS2_MOUNT_NO_POSIX_ACL))) {
1282		if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1283			options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1284		else
1285			options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1286	}
1287	return 1;
1288}
1289
1290static int ocfs2_parse_options(struct super_block *sb,
1291			       char *options,
1292			       struct mount_options *mopt,
1293			       int is_remount)
1294{
1295	int status, user_stack = 0;
1296	char *p;
1297	u32 tmp;
 
 
1298
1299	trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1300
1301	mopt->commit_interval = 0;
1302	mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1303	mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1304	mopt->slot = OCFS2_INVALID_SLOT;
1305	mopt->localalloc_opt = -1;
1306	mopt->cluster_stack[0] = '\0';
1307	mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1308	mopt->dir_resv_level = -1;
1309
1310	if (!options) {
1311		status = 1;
1312		goto bail;
1313	}
1314
1315	while ((p = strsep(&options, ",")) != NULL) {
1316		int token, option;
1317		substring_t args[MAX_OPT_ARGS];
1318
1319		if (!*p)
1320			continue;
1321
1322		token = match_token(p, tokens, args);
1323		switch (token) {
1324		case Opt_hb_local:
1325			mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1326			break;
1327		case Opt_hb_none:
1328			mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1329			break;
1330		case Opt_hb_global:
1331			mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1332			break;
1333		case Opt_barrier:
1334			if (match_int(&args[0], &option)) {
1335				status = 0;
1336				goto bail;
1337			}
1338			if (option)
1339				mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1340			else
1341				mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1342			break;
1343		case Opt_intr:
1344			mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1345			break;
1346		case Opt_nointr:
1347			mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1348			break;
1349		case Opt_err_panic:
 
 
1350			mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1351			break;
1352		case Opt_err_ro:
 
1353			mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
 
 
 
 
 
 
1354			break;
1355		case Opt_data_ordered:
1356			mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1357			break;
1358		case Opt_data_writeback:
1359			mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1360			break;
1361		case Opt_user_xattr:
1362			mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1363			break;
1364		case Opt_nouser_xattr:
1365			mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1366			break;
1367		case Opt_atime_quantum:
1368			if (match_int(&args[0], &option)) {
1369				status = 0;
1370				goto bail;
1371			}
1372			if (option >= 0)
1373				mopt->atime_quantum = option;
1374			break;
1375		case Opt_slot:
1376			option = 0;
1377			if (match_int(&args[0], &option)) {
1378				status = 0;
1379				goto bail;
1380			}
1381			if (option)
1382				mopt->slot = (s16)option;
1383			break;
1384		case Opt_commit:
1385			option = 0;
1386			if (match_int(&args[0], &option)) {
1387				status = 0;
1388				goto bail;
1389			}
1390			if (option < 0)
1391				return 0;
1392			if (option == 0)
1393				option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1394			mopt->commit_interval = HZ * option;
1395			break;
1396		case Opt_localalloc:
1397			option = 0;
1398			if (match_int(&args[0], &option)) {
1399				status = 0;
1400				goto bail;
1401			}
1402			if (option >= 0)
1403				mopt->localalloc_opt = option;
1404			break;
1405		case Opt_localflocks:
1406			/*
1407			 * Changing this during remount could race
1408			 * flock() requests, or "unbalance" existing
1409			 * ones (e.g., a lock is taken in one mode but
1410			 * dropped in the other). If users care enough
1411			 * to flip locking modes during remount, we
1412			 * could add a "local" flag to individual
1413			 * flock structures for proper tracking of
1414			 * state.
1415			 */
1416			if (!is_remount)
1417				mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1418			break;
1419		case Opt_stack:
1420			/* Check both that the option we were passed
1421			 * is of the right length and that it is a proper
1422			 * string of the right length.
1423			 */
1424			if (((args[0].to - args[0].from) !=
1425			     OCFS2_STACK_LABEL_LEN) ||
1426			    (strnlen(args[0].from,
1427				     OCFS2_STACK_LABEL_LEN) !=
1428			     OCFS2_STACK_LABEL_LEN)) {
1429				mlog(ML_ERROR,
1430				     "Invalid cluster_stack option\n");
1431				status = 0;
1432				goto bail;
1433			}
1434			memcpy(mopt->cluster_stack, args[0].from,
1435			       OCFS2_STACK_LABEL_LEN);
1436			mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1437			/*
1438			 * Open code the memcmp here as we don't have
1439			 * an osb to pass to
1440			 * ocfs2_userspace_stack().
1441			 */
1442			if (memcmp(mopt->cluster_stack,
1443				   OCFS2_CLASSIC_CLUSTER_STACK,
1444				   OCFS2_STACK_LABEL_LEN))
1445				user_stack = 1;
1446			break;
1447		case Opt_inode64:
1448			mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1449			break;
1450		case Opt_usrquota:
1451			mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1452			break;
1453		case Opt_grpquota:
1454			mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1455			break;
1456		case Opt_coherency_buffered:
1457			mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1458			break;
1459		case Opt_coherency_full:
1460			mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1461			break;
1462		case Opt_acl:
1463			mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1464			mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1465			break;
1466		case Opt_noacl:
1467			mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1468			mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1469			break;
1470		case Opt_resv_level:
1471			if (is_remount)
1472				break;
1473			if (match_int(&args[0], &option)) {
1474				status = 0;
1475				goto bail;
1476			}
1477			if (option >= OCFS2_MIN_RESV_LEVEL &&
1478			    option < OCFS2_MAX_RESV_LEVEL)
1479				mopt->resv_level = option;
1480			break;
1481		case Opt_dir_resv_level:
1482			if (is_remount)
1483				break;
1484			if (match_int(&args[0], &option)) {
1485				status = 0;
1486				goto bail;
1487			}
1488			if (option >= OCFS2_MIN_RESV_LEVEL &&
1489			    option < OCFS2_MAX_RESV_LEVEL)
1490				mopt->dir_resv_level = option;
1491			break;
 
 
 
1492		default:
1493			mlog(ML_ERROR,
1494			     "Unrecognized mount option \"%s\" "
1495			     "or missing value\n", p);
1496			status = 0;
1497			goto bail;
1498		}
1499	}
1500
1501	if (user_stack == 0) {
1502		/* Ensure only one heartbeat mode */
1503		tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1504					 OCFS2_MOUNT_HB_GLOBAL |
1505					 OCFS2_MOUNT_HB_NONE);
1506		if (hweight32(tmp) != 1) {
1507			mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1508			status = 0;
1509			goto bail;
1510		}
1511	}
1512
1513	status = 1;
1514
1515bail:
1516	return status;
1517}
1518
1519static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1520{
1521	struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1522	unsigned long opts = osb->s_mount_opt;
1523	unsigned int local_alloc_megs;
1524
1525	if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1526		seq_printf(s, ",_netdev");
1527		if (opts & OCFS2_MOUNT_HB_LOCAL)
1528			seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1529		else
1530			seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1531	} else
1532		seq_printf(s, ",%s", OCFS2_HB_NONE);
1533
1534	if (opts & OCFS2_MOUNT_NOINTR)
1535		seq_printf(s, ",nointr");
1536
1537	if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1538		seq_printf(s, ",data=writeback");
1539	else
1540		seq_printf(s, ",data=ordered");
1541
1542	if (opts & OCFS2_MOUNT_BARRIER)
1543		seq_printf(s, ",barrier=1");
1544
1545	if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1546		seq_printf(s, ",errors=panic");
 
 
1547	else
1548		seq_printf(s, ",errors=remount-ro");
1549
1550	if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1551		seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1552
1553	seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1554
1555	if (osb->osb_commit_interval)
1556		seq_printf(s, ",commit=%u",
1557			   (unsigned) (osb->osb_commit_interval / HZ));
1558
1559	local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1560	if (local_alloc_megs != ocfs2_la_default_mb(osb))
1561		seq_printf(s, ",localalloc=%d", local_alloc_megs);
1562
1563	if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1564		seq_printf(s, ",localflocks,");
1565
1566	if (osb->osb_cluster_stack[0])
1567		seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1568			   osb->osb_cluster_stack);
1569	if (opts & OCFS2_MOUNT_USRQUOTA)
1570		seq_printf(s, ",usrquota");
1571	if (opts & OCFS2_MOUNT_GRPQUOTA)
1572		seq_printf(s, ",grpquota");
1573
1574	if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1575		seq_printf(s, ",coherency=buffered");
1576	else
1577		seq_printf(s, ",coherency=full");
1578
1579	if (opts & OCFS2_MOUNT_NOUSERXATTR)
1580		seq_printf(s, ",nouser_xattr");
1581	else
1582		seq_printf(s, ",user_xattr");
1583
1584	if (opts & OCFS2_MOUNT_INODE64)
1585		seq_printf(s, ",inode64");
1586
1587	if (opts & OCFS2_MOUNT_POSIX_ACL)
1588		seq_printf(s, ",acl");
1589	else
1590		seq_printf(s, ",noacl");
1591
1592	if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1593		seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1594
1595	if (osb->osb_dir_resv_level != osb->osb_resv_level)
1596		seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1597
 
 
 
1598	return 0;
1599}
1600
1601static int __init ocfs2_init(void)
1602{
1603	int status;
1604
1605	status = init_ocfs2_uptodate_cache();
1606	if (status < 0)
1607		goto out1;
1608
1609	status = ocfs2_initialize_mem_caches();
1610	if (status < 0)
1611		goto out2;
1612
1613	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1614	if (!ocfs2_wq) {
1615		status = -ENOMEM;
1616		goto out3;
1617	}
1618
1619	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1620	if (!ocfs2_debugfs_root) {
1621		status = -EFAULT;
1622		mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1623	}
1624
1625	ocfs2_set_locking_protocol();
1626
1627	status = register_quota_format(&ocfs2_quota_format);
1628	if (status < 0)
1629		goto out4;
1630	status = register_filesystem(&ocfs2_fs_type);
1631	if (!status)
1632		return 0;
1633
1634	unregister_quota_format(&ocfs2_quota_format);
1635out4:
1636	destroy_workqueue(ocfs2_wq);
1637	debugfs_remove(ocfs2_debugfs_root);
1638out3:
1639	ocfs2_free_mem_caches();
1640out2:
1641	exit_ocfs2_uptodate_cache();
1642out1:
1643	mlog_errno(status);
1644	return status;
1645}
1646
1647static void __exit ocfs2_exit(void)
1648{
1649	if (ocfs2_wq) {
1650		flush_workqueue(ocfs2_wq);
1651		destroy_workqueue(ocfs2_wq);
1652	}
1653
1654	unregister_quota_format(&ocfs2_quota_format);
1655
1656	debugfs_remove(ocfs2_debugfs_root);
1657
1658	ocfs2_free_mem_caches();
1659
1660	unregister_filesystem(&ocfs2_fs_type);
1661
1662	exit_ocfs2_uptodate_cache();
1663}
1664
1665static void ocfs2_put_super(struct super_block *sb)
1666{
1667	trace_ocfs2_put_super(sb);
1668
1669	ocfs2_sync_blockdev(sb);
1670	ocfs2_dismount_volume(sb, 0);
1671}
1672
1673static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1674{
1675	struct ocfs2_super *osb;
1676	u32 numbits, freebits;
1677	int status;
1678	struct ocfs2_dinode *bm_lock;
1679	struct buffer_head *bh = NULL;
1680	struct inode *inode = NULL;
1681
1682	trace_ocfs2_statfs(dentry->d_sb, buf);
1683
1684	osb = OCFS2_SB(dentry->d_sb);
1685
1686	inode = ocfs2_get_system_file_inode(osb,
1687					    GLOBAL_BITMAP_SYSTEM_INODE,
1688					    OCFS2_INVALID_SLOT);
1689	if (!inode) {
1690		mlog(ML_ERROR, "failed to get bitmap inode\n");
1691		status = -EIO;
1692		goto bail;
1693	}
1694
1695	status = ocfs2_inode_lock(inode, &bh, 0);
1696	if (status < 0) {
1697		mlog_errno(status);
1698		goto bail;
1699	}
1700
1701	bm_lock = (struct ocfs2_dinode *) bh->b_data;
1702
1703	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1704	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1705
1706	buf->f_type = OCFS2_SUPER_MAGIC;
1707	buf->f_bsize = dentry->d_sb->s_blocksize;
1708	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1709	buf->f_blocks = ((sector_t) numbits) *
1710			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
1711	buf->f_bfree = ((sector_t) freebits) *
1712		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1713	buf->f_bavail = buf->f_bfree;
1714	buf->f_files = numbits;
1715	buf->f_ffree = freebits;
1716	buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1717				& 0xFFFFFFFFUL;
1718	buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1719				OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1720
1721	brelse(bh);
1722
1723	ocfs2_inode_unlock(inode, 0);
1724	status = 0;
1725bail:
1726	if (inode)
1727		iput(inode);
1728
1729	if (status)
1730		mlog_errno(status);
1731
1732	return status;
1733}
1734
1735static void ocfs2_inode_init_once(void *data)
1736{
1737	struct ocfs2_inode_info *oi = data;
1738
1739	oi->ip_flags = 0;
1740	oi->ip_open_count = 0;
1741	spin_lock_init(&oi->ip_lock);
1742	ocfs2_extent_map_init(&oi->vfs_inode);
1743	INIT_LIST_HEAD(&oi->ip_io_markers);
 
1744	oi->ip_dir_start_lookup = 0;
1745	mutex_init(&oi->ip_unaligned_aio);
1746	init_rwsem(&oi->ip_alloc_sem);
1747	init_rwsem(&oi->ip_xattr_sem);
1748	mutex_init(&oi->ip_io_mutex);
1749
1750	oi->ip_blkno = 0ULL;
1751	oi->ip_clusters = 0;
 
1752
1753	ocfs2_resv_init_once(&oi->ip_la_data_resv);
1754
1755	ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1756	ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1757	ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1758
1759	ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1760				  &ocfs2_inode_caching_ops);
1761
1762	inode_init_once(&oi->vfs_inode);
1763}
1764
1765static int ocfs2_initialize_mem_caches(void)
1766{
1767	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1768				       sizeof(struct ocfs2_inode_info),
1769				       0,
1770				       (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1771						SLAB_MEM_SPREAD),
1772				       ocfs2_inode_init_once);
1773	ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1774					sizeof(struct ocfs2_dquot),
1775					0,
1776					(SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1777						SLAB_MEM_SPREAD),
1778					NULL);
1779	ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1780					sizeof(struct ocfs2_quota_chunk),
1781					0,
1782					(SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1783					NULL);
1784	if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1785	    !ocfs2_qf_chunk_cachep) {
1786		if (ocfs2_inode_cachep)
1787			kmem_cache_destroy(ocfs2_inode_cachep);
1788		if (ocfs2_dquot_cachep)
1789			kmem_cache_destroy(ocfs2_dquot_cachep);
1790		if (ocfs2_qf_chunk_cachep)
1791			kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1792		return -ENOMEM;
1793	}
1794
1795	return 0;
1796}
1797
1798static void ocfs2_free_mem_caches(void)
1799{
1800	/*
1801	 * Make sure all delayed rcu free inodes are flushed before we
1802	 * destroy cache.
1803	 */
1804	rcu_barrier();
1805	if (ocfs2_inode_cachep)
1806		kmem_cache_destroy(ocfs2_inode_cachep);
1807	ocfs2_inode_cachep = NULL;
1808
1809	if (ocfs2_dquot_cachep)
1810		kmem_cache_destroy(ocfs2_dquot_cachep);
1811	ocfs2_dquot_cachep = NULL;
1812
1813	if (ocfs2_qf_chunk_cachep)
1814		kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1815	ocfs2_qf_chunk_cachep = NULL;
1816}
1817
1818static int ocfs2_get_sector(struct super_block *sb,
1819			    struct buffer_head **bh,
1820			    int block,
1821			    int sect_size)
1822{
1823	if (!sb_set_blocksize(sb, sect_size)) {
1824		mlog(ML_ERROR, "unable to set blocksize\n");
1825		return -EIO;
1826	}
1827
1828	*bh = sb_getblk(sb, block);
1829	if (!*bh) {
1830		mlog_errno(-ENOMEM);
1831		return -ENOMEM;
1832	}
1833	lock_buffer(*bh);
1834	if (!buffer_dirty(*bh))
1835		clear_buffer_uptodate(*bh);
1836	unlock_buffer(*bh);
1837	ll_rw_block(READ, 1, bh);
1838	wait_on_buffer(*bh);
1839	if (!buffer_uptodate(*bh)) {
1840		mlog_errno(-EIO);
1841		brelse(*bh);
1842		*bh = NULL;
1843		return -EIO;
1844	}
1845
1846	return 0;
1847}
1848
1849static int ocfs2_mount_volume(struct super_block *sb)
1850{
1851	int status = 0;
1852	int unlock_super = 0;
1853	struct ocfs2_super *osb = OCFS2_SB(sb);
1854
1855	if (ocfs2_is_hard_readonly(osb))
1856		goto leave;
 
 
1857
1858	status = ocfs2_dlm_init(osb);
1859	if (status < 0) {
1860		mlog_errno(status);
1861		goto leave;
 
 
 
1862	}
1863
1864	status = ocfs2_super_lock(osb, 1);
1865	if (status < 0) {
1866		mlog_errno(status);
1867		goto leave;
1868	}
1869	unlock_super = 1;
1870
1871	/* This will load up the node map and add ourselves to it. */
1872	status = ocfs2_find_slot(osb);
1873	if (status < 0) {
1874		mlog_errno(status);
1875		goto leave;
1876	}
1877
1878	/* load all node-local system inodes */
1879	status = ocfs2_init_local_system_inodes(osb);
1880	if (status < 0) {
1881		mlog_errno(status);
1882		goto leave;
1883	}
1884
1885	status = ocfs2_check_volume(osb);
1886	if (status < 0) {
1887		mlog_errno(status);
1888		goto leave;
1889	}
1890
1891	status = ocfs2_truncate_log_init(osb);
1892	if (status < 0)
1893		mlog_errno(status);
 
 
1894
1895leave:
1896	if (unlock_super)
1897		ocfs2_super_unlock(osb, 1);
1898
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1899	return status;
1900}
1901
1902static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1903{
1904	int tmp, hangup_needed = 0;
1905	struct ocfs2_super *osb = NULL;
1906	char nodestr[12];
1907
1908	trace_ocfs2_dismount_volume(sb);
1909
1910	BUG_ON(!sb);
1911	osb = OCFS2_SB(sb);
1912	BUG_ON(!osb);
1913
1914	debugfs_remove(osb->osb_ctxt);
 
 
 
 
1915
1916	/* Orphan scan should be stopped as early as possible */
1917	ocfs2_orphan_scan_stop(osb);
1918
1919	ocfs2_disable_quotas(osb);
1920
1921	/* All dquots should be freed by now */
1922	WARN_ON(!llist_empty(&osb->dquot_drop_list));
1923	/* Wait for worker to be done with the work structure in osb */
1924	cancel_work_sync(&osb->dquot_drop_work);
1925
1926	ocfs2_shutdown_local_alloc(osb);
1927
 
 
1928	/* This will disable recovery and flush any recovery work. */
1929	ocfs2_recovery_exit(osb);
1930
1931	/*
1932	 * During dismount, when it recovers another node it will call
1933	 * ocfs2_recover_orphans and queue delayed work osb_truncate_log_wq.
1934	 */
1935	ocfs2_truncate_log_shutdown(osb);
1936
1937	ocfs2_journal_shutdown(osb);
1938
1939	ocfs2_sync_blockdev(sb);
1940
1941	ocfs2_purge_refcount_trees(osb);
1942
1943	/* No cluster connection means we've failed during mount, so skip
1944	 * all the steps which depended on that to complete. */
1945	if (osb->cconn) {
1946		tmp = ocfs2_super_lock(osb, 1);
1947		if (tmp < 0) {
1948			mlog_errno(tmp);
1949			return;
1950		}
1951	}
1952
1953	if (osb->slot_num != OCFS2_INVALID_SLOT)
1954		ocfs2_put_slot(osb);
1955
1956	if (osb->cconn)
1957		ocfs2_super_unlock(osb, 1);
1958
1959	ocfs2_release_system_inodes(osb);
1960
 
 
1961	/*
1962	 * If we're dismounting due to mount error, mount.ocfs2 will clean
1963	 * up heartbeat.  If we're a local mount, there is no heartbeat.
1964	 * If we failed before we got a uuid_str yet, we can't stop
1965	 * heartbeat.  Otherwise, do it.
1966	 */
1967	if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1968	    !ocfs2_is_hard_readonly(osb))
1969		hangup_needed = 1;
1970
1971	if (osb->cconn)
1972		ocfs2_dlm_shutdown(osb, hangup_needed);
1973
1974	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1975	debugfs_remove(osb->osb_debug_root);
1976
1977	if (hangup_needed)
1978		ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1979
1980	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1981
1982	if (ocfs2_mount_local(osb))
1983		snprintf(nodestr, sizeof(nodestr), "local");
1984	else
1985		snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1986
1987	printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1988	       osb->dev_str, nodestr);
1989
1990	ocfs2_delete_osb(osb);
1991	kfree(osb);
1992	sb->s_dev = 0;
1993	sb->s_fs_info = NULL;
1994}
1995
1996static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1997				unsigned uuid_bytes)
1998{
1999	int i, ret;
2000	char *ptr;
2001
2002	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2003
2004	osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2005	if (osb->uuid_str == NULL)
2006		return -ENOMEM;
2007
2008	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2009		/* print with null */
2010		ret = snprintf(ptr, 3, "%02X", uuid[i]);
2011		if (ret != 2) /* drop super cleans up */
2012			return -EINVAL;
2013		/* then only advance past the last char */
2014		ptr += 2;
2015	}
2016
2017	return 0;
2018}
2019
2020/* Make sure entire volume is addressable by our journal.  Requires
2021   osb_clusters_at_boot to be valid and for the journal to have been
2022   initialized by ocfs2_journal_init(). */
2023static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2024{
2025	int status = 0;
2026	u64 max_block =
2027		ocfs2_clusters_to_blocks(osb->sb,
2028					 osb->osb_clusters_at_boot) - 1;
2029
2030	/* 32-bit block number is always OK. */
2031	if (max_block <= (u32)~0ULL)
2032		goto out;
2033
2034	/* Volume is "huge", so see if our journal is new enough to
2035	   support it. */
2036	if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2037				       OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2038	      jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2039					       JBD2_FEATURE_INCOMPAT_64BIT))) {
2040		mlog(ML_ERROR, "The journal cannot address the entire volume. "
2041		     "Enable the 'block64' journal option with tunefs.ocfs2");
2042		status = -EFBIG;
2043		goto out;
2044	}
2045
2046 out:
2047	return status;
2048}
2049
2050static int ocfs2_initialize_super(struct super_block *sb,
2051				  struct buffer_head *bh,
2052				  int sector_size,
2053				  struct ocfs2_blockcheck_stats *stats)
2054{
2055	int status;
2056	int i, cbits, bbits;
2057	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2058	struct inode *inode = NULL;
2059	struct ocfs2_journal *journal;
2060	struct ocfs2_super *osb;
2061	u64 total_blocks;
2062
2063	osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2064	if (!osb) {
2065		status = -ENOMEM;
2066		mlog_errno(status);
2067		goto bail;
2068	}
2069
2070	sb->s_fs_info = osb;
2071	sb->s_op = &ocfs2_sops;
2072	sb->s_d_op = &ocfs2_dentry_ops;
2073	sb->s_export_op = &ocfs2_export_ops;
2074	sb->s_qcop = &ocfs2_quotactl_ops;
2075	sb->dq_op = &ocfs2_quota_operations;
 
2076	sb->s_xattr = ocfs2_xattr_handlers;
2077	sb->s_time_gran = 1;
2078	sb->s_flags |= MS_NOATIME;
2079	/* this is needed to support O_LARGEFILE */
2080	cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2081	bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2082	sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
 
 
2083
2084	osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2085
2086	for (i = 0; i < 3; i++)
2087		osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2088	osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2089
2090	osb->sb = sb;
2091	/* Save off for ocfs2_rw_direct */
2092	osb->s_sectsize_bits = blksize_bits(sector_size);
2093	BUG_ON(!osb->s_sectsize_bits);
2094
2095	spin_lock_init(&osb->dc_task_lock);
2096	init_waitqueue_head(&osb->dc_event);
2097	osb->dc_work_sequence = 0;
2098	osb->dc_wake_sequence = 0;
2099	INIT_LIST_HEAD(&osb->blocked_lock_list);
2100	osb->blocked_lock_count = 0;
2101	spin_lock_init(&osb->osb_lock);
2102	spin_lock_init(&osb->osb_xattr_lock);
2103	ocfs2_init_steal_slots(osb);
2104
2105	mutex_init(&osb->system_file_mutex);
2106
2107	atomic_set(&osb->alloc_stats.moves, 0);
2108	atomic_set(&osb->alloc_stats.local_data, 0);
2109	atomic_set(&osb->alloc_stats.bitmap_data, 0);
2110	atomic_set(&osb->alloc_stats.bg_allocs, 0);
2111	atomic_set(&osb->alloc_stats.bg_extends, 0);
2112
2113	/* Copy the blockcheck stats from the superblock probe */
2114	osb->osb_ecc_stats = *stats;
2115
2116	ocfs2_init_node_maps(osb);
2117
2118	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2119		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2120
2121	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2122	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2123		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2124		     osb->max_slots);
2125		status = -EINVAL;
2126		goto bail;
2127	}
2128
2129	ocfs2_orphan_scan_init(osb);
2130
2131	status = ocfs2_recovery_init(osb);
2132	if (status) {
2133		mlog(ML_ERROR, "Unable to initialize recovery state\n");
2134		mlog_errno(status);
2135		goto bail;
2136	}
2137
2138	init_waitqueue_head(&osb->checkpoint_event);
2139
2140	osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2141
2142	osb->slot_num = OCFS2_INVALID_SLOT;
2143
2144	osb->s_xattr_inline_size = le16_to_cpu(
2145					di->id2.i_super.s_xattr_inline_size);
2146
2147	osb->local_alloc_state = OCFS2_LA_UNUSED;
2148	osb->local_alloc_bh = NULL;
2149	INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2150
2151	init_waitqueue_head(&osb->osb_mount_event);
2152
2153	status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2154	if (status) {
2155		mlog_errno(status);
2156		goto bail;
2157	}
2158
2159	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2160	if (!osb->vol_label) {
2161		mlog(ML_ERROR, "unable to alloc vol label\n");
2162		status = -ENOMEM;
2163		goto bail;
2164	}
2165
2166	osb->slot_recovery_generations =
2167		kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2168			GFP_KERNEL);
2169	if (!osb->slot_recovery_generations) {
2170		status = -ENOMEM;
2171		mlog_errno(status);
2172		goto bail;
2173	}
2174
2175	init_waitqueue_head(&osb->osb_wipe_event);
2176	osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2177					sizeof(*osb->osb_orphan_wipes),
2178					GFP_KERNEL);
2179	if (!osb->osb_orphan_wipes) {
2180		status = -ENOMEM;
2181		mlog_errno(status);
2182		goto bail;
2183	}
2184
2185	osb->osb_rf_lock_tree = RB_ROOT;
2186
2187	osb->s_feature_compat =
2188		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2189	osb->s_feature_ro_compat =
2190		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2191	osb->s_feature_incompat =
2192		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2193
2194	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2195		mlog(ML_ERROR, "couldn't mount because of unsupported "
2196		     "optional features (%x).\n", i);
2197		status = -EINVAL;
2198		goto bail;
2199	}
2200	if (!(osb->sb->s_flags & MS_RDONLY) &&
2201	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2202		mlog(ML_ERROR, "couldn't mount RDWR because of "
2203		     "unsupported optional features (%x).\n", i);
2204		status = -EINVAL;
2205		goto bail;
2206	}
2207
2208	if (ocfs2_clusterinfo_valid(osb)) {
 
 
 
 
 
 
2209		osb->osb_stackflags =
2210			OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2211		strlcpy(osb->osb_cluster_stack,
2212		       OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2213		       OCFS2_STACK_LABEL_LEN + 1);
2214		if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2215			mlog(ML_ERROR,
2216			     "couldn't mount because of an invalid "
2217			     "cluster stack label (%s) \n",
2218			     osb->osb_cluster_stack);
2219			status = -EINVAL;
2220			goto bail;
2221		}
2222		strlcpy(osb->osb_cluster_name,
2223			OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2224			OCFS2_CLUSTER_NAME_LEN + 1);
2225	} else {
2226		/* The empty string is identical with classic tools that
2227		 * don't know about s_cluster_info. */
2228		osb->osb_cluster_stack[0] = '\0';
2229	}
2230
2231	get_random_bytes(&osb->s_next_generation, sizeof(u32));
2232
2233	/* FIXME
2234	 * This should be done in ocfs2_journal_init(), but unknown
2235	 * ordering issues will cause the filesystem to crash.
2236	 * If anyone wants to figure out what part of the code
2237	 * refers to osb->journal before ocfs2_journal_init() is run,
2238	 * be my guest.
2239	 */
2240	/* initialize our journal structure */
2241
2242	journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2243	if (!journal) {
2244		mlog(ML_ERROR, "unable to alloc journal\n");
2245		status = -ENOMEM;
2246		goto bail;
2247	}
2248	osb->journal = journal;
2249	journal->j_osb = osb;
2250
2251	atomic_set(&journal->j_num_trans, 0);
2252	init_rwsem(&journal->j_trans_barrier);
2253	init_waitqueue_head(&journal->j_checkpointed);
2254	spin_lock_init(&journal->j_lock);
2255	journal->j_trans_id = (unsigned long) 1;
2256	INIT_LIST_HEAD(&journal->j_la_cleanups);
2257	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2258	journal->j_state = OCFS2_JOURNAL_FREE;
2259
2260	INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2261	init_llist_head(&osb->dquot_drop_list);
2262
2263	/* get some pseudo constants for clustersize bits */
2264	osb->s_clustersize_bits =
2265		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2266	osb->s_clustersize = 1 << osb->s_clustersize_bits;
2267
2268	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2269	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2270		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2271		     osb->s_clustersize);
2272		status = -EINVAL;
2273		goto bail;
2274	}
2275
2276	total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2277						le32_to_cpu(di->i_clusters));
2278
2279	status = generic_check_addressable(osb->sb->s_blocksize_bits,
2280					   total_blocks);
2281	if (status) {
2282		mlog(ML_ERROR, "Volume too large "
2283		     "to mount safely on this system");
2284		status = -EFBIG;
2285		goto bail;
2286	}
2287
2288	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2289				 sizeof(di->id2.i_super.s_uuid))) {
2290		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2291		status = -ENOMEM;
2292		goto bail;
2293	}
2294
2295	strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2296	osb->vol_label[63] = '\0';
2297	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2298	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2299	osb->first_cluster_group_blkno =
2300		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2301	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2302	osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2303	trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2304				     (unsigned long long)osb->root_blkno,
2305				     (unsigned long long)osb->system_dir_blkno,
2306				     osb->s_clustersize_bits);
2307
2308	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2309	if (!osb->osb_dlm_debug) {
2310		status = -ENOMEM;
2311		mlog_errno(status);
2312		goto bail;
2313	}
2314
2315	atomic_set(&osb->vol_state, VOLUME_INIT);
2316
2317	/* load root, system_dir, and all global system inodes */
2318	status = ocfs2_init_global_system_inodes(osb);
2319	if (status < 0) {
2320		mlog_errno(status);
2321		goto bail;
2322	}
2323
2324	/*
2325	 * global bitmap
2326	 */
2327	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2328					    OCFS2_INVALID_SLOT);
2329	if (!inode) {
2330		status = -EINVAL;
2331		mlog_errno(status);
2332		goto bail;
2333	}
2334
2335	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2336	osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2337	iput(inode);
2338
2339	osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2340				 osb->s_feature_incompat) * 8;
2341
2342	status = ocfs2_init_slot_info(osb);
2343	if (status < 0) {
2344		mlog_errno(status);
2345		goto bail;
 
 
 
 
 
 
 
2346	}
2347	cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
2348
2349bail:
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2350	return status;
2351}
2352
2353/*
2354 * will return: -EAGAIN if it is ok to keep searching for superblocks
2355 *              -EINVAL if there is a bad superblock
2356 *              0 on success
2357 */
2358static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2359			       struct buffer_head *bh,
2360			       u32 blksz,
2361			       struct ocfs2_blockcheck_stats *stats)
2362{
2363	int status = -EAGAIN;
2364
2365	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2366		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2367		/* We have to do a raw check of the feature here */
2368		if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2369		    OCFS2_FEATURE_INCOMPAT_META_ECC) {
2370			status = ocfs2_block_check_validate(bh->b_data,
2371							    bh->b_size,
2372							    &di->i_check,
2373							    stats);
2374			if (status)
2375				goto out;
2376		}
2377		status = -EINVAL;
2378		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2379			mlog(ML_ERROR, "found superblock with incorrect block "
2380			     "size: found %u, should be %u\n",
2381			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2382			       blksz);
2383		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2384			   OCFS2_MAJOR_REV_LEVEL ||
2385			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2386			   OCFS2_MINOR_REV_LEVEL) {
2387			mlog(ML_ERROR, "found superblock with bad version: "
2388			     "found %u.%u, should be %u.%u\n",
2389			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
2390			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2391			     OCFS2_MAJOR_REV_LEVEL,
2392			     OCFS2_MINOR_REV_LEVEL);
2393		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2394			mlog(ML_ERROR, "bad block number on superblock: "
2395			     "found %llu, should be %llu\n",
2396			     (unsigned long long)le64_to_cpu(di->i_blkno),
2397			     (unsigned long long)bh->b_blocknr);
2398		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2399			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2400			mlog(ML_ERROR, "bad cluster size found: %u\n",
2401			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2402		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2403			mlog(ML_ERROR, "bad root_blkno: 0\n");
2404		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2405			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2406		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2407			mlog(ML_ERROR,
2408			     "Superblock slots found greater than file system "
2409			     "maximum: found %u, max %u\n",
2410			     le16_to_cpu(di->id2.i_super.s_max_slots),
2411			     OCFS2_MAX_SLOTS);
2412		} else {
2413			/* found it! */
2414			status = 0;
2415		}
2416	}
2417
2418out:
2419	if (status && status != -EAGAIN)
2420		mlog_errno(status);
2421	return status;
2422}
2423
2424static int ocfs2_check_volume(struct ocfs2_super *osb)
2425{
2426	int status;
2427	int dirty;
2428	int local;
2429	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2430						  * recover
2431						  * ourselves. */
2432
2433	/* Init our journal object. */
2434	status = ocfs2_journal_init(osb->journal, &dirty);
2435	if (status < 0) {
2436		mlog(ML_ERROR, "Could not initialize journal!\n");
2437		goto finally;
2438	}
2439
2440	/* Now that journal has been initialized, check to make sure
2441	   entire volume is addressable. */
2442	status = ocfs2_journal_addressable(osb);
2443	if (status)
2444		goto finally;
2445
2446	/* If the journal was unmounted cleanly then we don't want to
2447	 * recover anything. Otherwise, journal_load will do that
2448	 * dirty work for us :) */
2449	if (!dirty) {
2450		status = ocfs2_journal_wipe(osb->journal, 0);
2451		if (status < 0) {
2452			mlog_errno(status);
2453			goto finally;
2454		}
2455	} else {
2456		printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2457		       "unmounted cleanly, recovering it.\n", osb->dev_str);
2458	}
2459
2460	local = ocfs2_mount_local(osb);
2461
2462	/* will play back anything left in the journal. */
2463	status = ocfs2_journal_load(osb->journal, local, dirty);
2464	if (status < 0) {
2465		mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2466		goto finally;
2467	}
2468
 
 
 
 
 
 
 
 
 
2469	if (dirty) {
2470		/* recover my local alloc if we didn't unmount cleanly. */
2471		status = ocfs2_begin_local_alloc_recovery(osb,
2472							  osb->slot_num,
2473							  &local_alloc);
2474		if (status < 0) {
2475			mlog_errno(status);
2476			goto finally;
2477		}
2478		/* we complete the recovery process after we've marked
2479		 * ourselves as mounted. */
2480	}
2481
2482	status = ocfs2_load_local_alloc(osb);
2483	if (status < 0) {
2484		mlog_errno(status);
2485		goto finally;
2486	}
2487
2488	if (dirty) {
2489		/* Recovery will be completed after we've mounted the
2490		 * rest of the volume. */
2491		osb->dirty = 1;
2492		osb->local_alloc_copy = local_alloc;
2493		local_alloc = NULL;
2494	}
2495
2496	/* go through each journal, trylock it and if you get the
2497	 * lock, and it's marked as dirty, set the bit in the recover
2498	 * map and launch a recovery thread for it. */
2499	status = ocfs2_mark_dead_nodes(osb);
2500	if (status < 0) {
2501		mlog_errno(status);
2502		goto finally;
2503	}
2504
2505	status = ocfs2_compute_replay_slots(osb);
2506	if (status < 0)
2507		mlog_errno(status);
2508
2509finally:
2510	kfree(local_alloc);
2511
2512	if (status)
2513		mlog_errno(status);
2514	return status;
2515}
2516
2517/*
2518 * The routine gets called from dismount or close whenever a dismount on
2519 * volume is requested and the osb open count becomes 1.
2520 * It will remove the osb from the global list and also free up all the
2521 * initialized resources and fileobject.
2522 */
2523static void ocfs2_delete_osb(struct ocfs2_super *osb)
2524{
2525	/* This function assumes that the caller has the main osb resource */
2526
 
 
 
 
2527	ocfs2_free_slot_info(osb);
2528
2529	kfree(osb->osb_orphan_wipes);
2530	kfree(osb->slot_recovery_generations);
2531	/* FIXME
2532	 * This belongs in journal shutdown, but because we have to
2533	 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2534	 * we free it here.
2535	 */
2536	kfree(osb->journal);
2537	kfree(osb->local_alloc_copy);
2538	kfree(osb->uuid_str);
 
2539	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2540	memset(osb, 0, sizeof(struct ocfs2_super));
2541}
2542
2543/* Put OCFS2 into a readonly state, or (if the user specifies it),
2544 * panic(). We do not support continue-on-error operation. */
2545static void ocfs2_handle_error(struct super_block *sb)
 
 
 
 
2546{
2547	struct ocfs2_super *osb = OCFS2_SB(sb);
 
 
 
 
 
2548
2549	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2550		panic("OCFS2: (device %s): panic forced after error\n",
2551		      sb->s_id);
 
 
 
 
 
 
 
2552
2553	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
 
 
 
2554
2555	if (sb->s_flags & MS_RDONLY &&
2556	    (ocfs2_is_soft_readonly(osb) ||
2557	     ocfs2_is_hard_readonly(osb)))
2558		return;
2559
2560	printk(KERN_CRIT "File system is now read-only due to the potential "
2561	       "of on-disk corruption. Please run fsck.ocfs2 once the file "
2562	       "system is unmounted.\n");
2563	sb->s_flags |= MS_RDONLY;
2564	ocfs2_set_ro_flag(osb, 0);
2565}
2566
2567static char error_buf[1024];
2568
2569void __ocfs2_error(struct super_block *sb,
2570		   const char *function,
2571		   const char *fmt, ...)
2572{
 
2573	va_list args;
2574
2575	va_start(args, fmt);
2576	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2577	va_end(args);
2578
2579	/* Not using mlog here because we want to show the actual
2580	 * function the error came from. */
2581	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2582	       sb->s_id, function, error_buf);
2583
2584	ocfs2_handle_error(sb);
 
 
2585}
2586
2587/* Handle critical errors. This is intentionally more drastic than
2588 * ocfs2_handle_error, so we only use for things like journal errors,
2589 * etc. */
2590void __ocfs2_abort(struct super_block* sb,
2591		   const char *function,
2592		   const char *fmt, ...)
2593{
 
2594	va_list args;
2595
2596	va_start(args, fmt);
2597	vsnprintf(error_buf, sizeof(error_buf), fmt, args);
 
 
 
 
 
 
2598	va_end(args);
2599
2600	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2601	       sb->s_id, function, error_buf);
2602
2603	/* We don't have the cluster support yet to go straight to
2604	 * hard readonly in here. Until then, we want to keep
2605	 * ocfs2_abort() so that we can at least mark critical
2606	 * errors.
2607	 *
2608	 * TODO: This should abort the journal and alert other nodes
2609	 * that our slot needs recovery. */
2610
2611	/* Force a panic(). This stinks, but it's better than letting
2612	 * things continue without having a proper hard readonly
2613	 * here. */
2614	if (!ocfs2_mount_local(OCFS2_SB(sb)))
2615		OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2616	ocfs2_handle_error(sb);
2617}
2618
2619/*
2620 * Void signal blockers, because in-kernel sigprocmask() only fails
2621 * when SIG_* is wrong.
2622 */
2623void ocfs2_block_signals(sigset_t *oldset)
2624{
2625	int rc;
2626	sigset_t blocked;
2627
2628	sigfillset(&blocked);
2629	rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2630	BUG_ON(rc);
2631}
2632
2633void ocfs2_unblock_signals(sigset_t *oldset)
2634{
2635	int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2636	BUG_ON(rc);
2637}
2638
2639module_init(ocfs2_init);
2640module_exit(ocfs2_exit);