1ccd979bdSMark Fasheh /* -*- mode: c; c-basic-offset: 8; -*- 2ccd979bdSMark Fasheh * vim: noexpandtab sw=8 ts=8 sts=0: 3ccd979bdSMark Fasheh * 4ccd979bdSMark Fasheh * super.c 5ccd979bdSMark Fasheh * 6ccd979bdSMark Fasheh * load/unload driver, mount/dismount volumes 7ccd979bdSMark Fasheh * 8ccd979bdSMark Fasheh * Copyright (C) 2002, 2004 Oracle. All rights reserved. 9ccd979bdSMark Fasheh * 10ccd979bdSMark Fasheh * This program is free software; you can redistribute it and/or 11ccd979bdSMark Fasheh * modify it under the terms of the GNU General Public 12ccd979bdSMark Fasheh * License as published by the Free Software Foundation; either 13ccd979bdSMark Fasheh * version 2 of the License, or (at your option) any later version. 14ccd979bdSMark Fasheh * 15ccd979bdSMark Fasheh * This program is distributed in the hope that it will be useful, 16ccd979bdSMark Fasheh * but WITHOUT ANY WARRANTY; without even the implied warranty of 17ccd979bdSMark Fasheh * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 18ccd979bdSMark Fasheh * General Public License for more details. 19ccd979bdSMark Fasheh * 20ccd979bdSMark Fasheh * You should have received a copy of the GNU General Public 21ccd979bdSMark Fasheh * License along with this program; if not, write to the 22ccd979bdSMark Fasheh * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 23ccd979bdSMark Fasheh * Boston, MA 021110-1307, USA. 24ccd979bdSMark Fasheh */ 25ccd979bdSMark Fasheh 26ccd979bdSMark Fasheh #include <linux/module.h> 27ccd979bdSMark Fasheh #include <linux/fs.h> 28ccd979bdSMark Fasheh #include <linux/types.h> 29ccd979bdSMark Fasheh #include <linux/slab.h> 30ccd979bdSMark Fasheh #include <linux/highmem.h> 31ccd979bdSMark Fasheh #include <linux/utsname.h> 32ccd979bdSMark Fasheh #include <linux/init.h> 33ccd979bdSMark Fasheh #include <linux/random.h> 34ccd979bdSMark Fasheh #include <linux/statfs.h> 35ccd979bdSMark Fasheh #include <linux/moduleparam.h> 36ccd979bdSMark Fasheh #include <linux/blkdev.h> 37ccd979bdSMark Fasheh #include <linux/socket.h> 38ccd979bdSMark Fasheh #include <linux/inet.h> 39ccd979bdSMark Fasheh #include <linux/parser.h> 40ccd979bdSMark Fasheh #include <linux/crc32.h> 41ccd979bdSMark Fasheh #include <linux/debugfs.h> 42d550071cSSunil Mushran #include <linux/mount.h> 43ccd979bdSMark Fasheh 44ccd979bdSMark Fasheh #include <cluster/nodemanager.h> 45ccd979bdSMark Fasheh 46ccd979bdSMark Fasheh #define MLOG_MASK_PREFIX ML_SUPER 47ccd979bdSMark Fasheh #include <cluster/masklog.h> 48ccd979bdSMark Fasheh 49ccd979bdSMark Fasheh #include "ocfs2.h" 50ccd979bdSMark Fasheh 51ccd979bdSMark Fasheh /* this should be the only file to include a version 1 header */ 52ccd979bdSMark Fasheh #include "ocfs1_fs_compat.h" 53ccd979bdSMark Fasheh 54ccd979bdSMark Fasheh #include "alloc.h" 55ccd979bdSMark Fasheh #include "dlmglue.h" 56ccd979bdSMark Fasheh #include "export.h" 57ccd979bdSMark Fasheh #include "extent_map.h" 58ccd979bdSMark Fasheh #include "heartbeat.h" 59ccd979bdSMark Fasheh #include "inode.h" 60ccd979bdSMark Fasheh #include "journal.h" 61ccd979bdSMark Fasheh #include "localalloc.h" 62ccd979bdSMark Fasheh #include "namei.h" 63ccd979bdSMark Fasheh #include "slot_map.h" 64ccd979bdSMark Fasheh #include "super.h" 65ccd979bdSMark Fasheh #include "sysfile.h" 66ccd979bdSMark Fasheh #include "uptodate.h" 67ccd979bdSMark Fasheh #include "ver.h" 68ccd979bdSMark Fasheh 69ccd979bdSMark Fasheh #include "buffer_head_io.h" 70ccd979bdSMark Fasheh 71e18b890bSChristoph Lameter static struct kmem_cache *ocfs2_inode_cachep = NULL; 72ccd979bdSMark Fasheh 73ccd979bdSMark Fasheh /* OCFS2 needs to schedule several differnt types of work which 74ccd979bdSMark Fasheh * require cluster locking, disk I/O, recovery waits, etc. Since these 75ccd979bdSMark Fasheh * types of work tend to be heavy we avoid using the kernel events 76ccd979bdSMark Fasheh * workqueue and schedule on our own. */ 77ccd979bdSMark Fasheh struct workqueue_struct *ocfs2_wq = NULL; 78ccd979bdSMark Fasheh 79ccd979bdSMark Fasheh static struct dentry *ocfs2_debugfs_root = NULL; 80ccd979bdSMark Fasheh 81ccd979bdSMark Fasheh MODULE_AUTHOR("Oracle"); 82ccd979bdSMark Fasheh MODULE_LICENSE("GPL"); 83ccd979bdSMark Fasheh 84c0123adeSTiger Yang struct mount_options 85c0123adeSTiger Yang { 86c0123adeSTiger Yang unsigned long mount_opt; 87c0123adeSTiger Yang unsigned int atime_quantum; 88c0123adeSTiger Yang signed short slot; 89c0123adeSTiger Yang }; 90c0123adeSTiger Yang 91ccd979bdSMark Fasheh static int ocfs2_parse_options(struct super_block *sb, char *options, 92c0123adeSTiger Yang struct mount_options *mopt, 93baf4661aSSunil Mushran int is_remount); 94d550071cSSunil Mushran static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt); 95ccd979bdSMark Fasheh static void ocfs2_put_super(struct super_block *sb); 96ccd979bdSMark Fasheh static int ocfs2_mount_volume(struct super_block *sb); 97ccd979bdSMark Fasheh static int ocfs2_remount(struct super_block *sb, int *flags, char *data); 98ccd979bdSMark Fasheh static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err); 99ccd979bdSMark Fasheh static int ocfs2_initialize_mem_caches(void); 100ccd979bdSMark Fasheh static void ocfs2_free_mem_caches(void); 101ccd979bdSMark Fasheh static void ocfs2_delete_osb(struct ocfs2_super *osb); 102ccd979bdSMark Fasheh 103726c3342SDavid Howells static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf); 104ccd979bdSMark Fasheh 105ccd979bdSMark Fasheh static int ocfs2_sync_fs(struct super_block *sb, int wait); 106ccd979bdSMark Fasheh 107ccd979bdSMark Fasheh static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb); 108ccd979bdSMark Fasheh static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb); 109bddb8eb3SDenis Cheng static void ocfs2_release_system_inodes(struct ocfs2_super *osb); 110ccd979bdSMark Fasheh static int ocfs2_fill_local_node_info(struct ocfs2_super *osb); 111ccd979bdSMark Fasheh static int ocfs2_check_volume(struct ocfs2_super *osb); 112ccd979bdSMark Fasheh static int ocfs2_verify_volume(struct ocfs2_dinode *di, 113ccd979bdSMark Fasheh struct buffer_head *bh, 114ccd979bdSMark Fasheh u32 sectsize); 115ccd979bdSMark Fasheh static int ocfs2_initialize_super(struct super_block *sb, 116ccd979bdSMark Fasheh struct buffer_head *bh, 117ccd979bdSMark Fasheh int sector_size); 118ccd979bdSMark Fasheh static int ocfs2_get_sector(struct super_block *sb, 119ccd979bdSMark Fasheh struct buffer_head **bh, 120ccd979bdSMark Fasheh int block, 121ccd979bdSMark Fasheh int sect_size); 122ccd979bdSMark Fasheh static void ocfs2_write_super(struct super_block *sb); 123ccd979bdSMark Fasheh static struct inode *ocfs2_alloc_inode(struct super_block *sb); 124ccd979bdSMark Fasheh static void ocfs2_destroy_inode(struct inode *inode); 125ccd979bdSMark Fasheh 126ee9b6d61SJosef 'Jeff' Sipek static const struct super_operations ocfs2_sops = { 127ccd979bdSMark Fasheh .statfs = ocfs2_statfs, 128ccd979bdSMark Fasheh .alloc_inode = ocfs2_alloc_inode, 129ccd979bdSMark Fasheh .destroy_inode = ocfs2_destroy_inode, 130ccd979bdSMark Fasheh .drop_inode = ocfs2_drop_inode, 131ccd979bdSMark Fasheh .clear_inode = ocfs2_clear_inode, 132ccd979bdSMark Fasheh .delete_inode = ocfs2_delete_inode, 133ccd979bdSMark Fasheh .sync_fs = ocfs2_sync_fs, 134ccd979bdSMark Fasheh .write_super = ocfs2_write_super, 135ccd979bdSMark Fasheh .put_super = ocfs2_put_super, 136ccd979bdSMark Fasheh .remount_fs = ocfs2_remount, 137d550071cSSunil Mushran .show_options = ocfs2_show_options, 138ccd979bdSMark Fasheh }; 139ccd979bdSMark Fasheh 140ccd979bdSMark Fasheh enum { 141ccd979bdSMark Fasheh Opt_barrier, 142ccd979bdSMark Fasheh Opt_err_panic, 143ccd979bdSMark Fasheh Opt_err_ro, 144ccd979bdSMark Fasheh Opt_intr, 145ccd979bdSMark Fasheh Opt_nointr, 146ccd979bdSMark Fasheh Opt_hb_none, 147ccd979bdSMark Fasheh Opt_hb_local, 148ccd979bdSMark Fasheh Opt_data_ordered, 149ccd979bdSMark Fasheh Opt_data_writeback, 1507f1a37e3STiger Yang Opt_atime_quantum, 151baf4661aSSunil Mushran Opt_slot, 152ccd979bdSMark Fasheh Opt_err, 153ccd979bdSMark Fasheh }; 154ccd979bdSMark Fasheh 155ccd979bdSMark Fasheh static match_table_t tokens = { 156ccd979bdSMark Fasheh {Opt_barrier, "barrier=%u"}, 157ccd979bdSMark Fasheh {Opt_err_panic, "errors=panic"}, 158ccd979bdSMark Fasheh {Opt_err_ro, "errors=remount-ro"}, 159ccd979bdSMark Fasheh {Opt_intr, "intr"}, 160ccd979bdSMark Fasheh {Opt_nointr, "nointr"}, 161ccd979bdSMark Fasheh {Opt_hb_none, OCFS2_HB_NONE}, 162ccd979bdSMark Fasheh {Opt_hb_local, OCFS2_HB_LOCAL}, 163ccd979bdSMark Fasheh {Opt_data_ordered, "data=ordered"}, 164ccd979bdSMark Fasheh {Opt_data_writeback, "data=writeback"}, 1657f1a37e3STiger Yang {Opt_atime_quantum, "atime_quantum=%u"}, 166baf4661aSSunil Mushran {Opt_slot, "preferred_slot=%u"}, 167ccd979bdSMark Fasheh {Opt_err, NULL} 168ccd979bdSMark Fasheh }; 169ccd979bdSMark Fasheh 170ccd979bdSMark Fasheh /* 171ccd979bdSMark Fasheh * write_super and sync_fs ripped right out of ext3. 172ccd979bdSMark Fasheh */ 173ccd979bdSMark Fasheh static void ocfs2_write_super(struct super_block *sb) 174ccd979bdSMark Fasheh { 1757892f2f4SIngo Molnar if (mutex_trylock(&sb->s_lock) != 0) 176ccd979bdSMark Fasheh BUG(); 177ccd979bdSMark Fasheh sb->s_dirt = 0; 178ccd979bdSMark Fasheh } 179ccd979bdSMark Fasheh 180ccd979bdSMark Fasheh static int ocfs2_sync_fs(struct super_block *sb, int wait) 181ccd979bdSMark Fasheh { 182bddb8eb3SDenis Cheng int status; 183ccd979bdSMark Fasheh tid_t target; 184ccd979bdSMark Fasheh struct ocfs2_super *osb = OCFS2_SB(sb); 185ccd979bdSMark Fasheh 186ccd979bdSMark Fasheh sb->s_dirt = 0; 187ccd979bdSMark Fasheh 188ccd979bdSMark Fasheh if (ocfs2_is_hard_readonly(osb)) 189ccd979bdSMark Fasheh return -EROFS; 190ccd979bdSMark Fasheh 191ccd979bdSMark Fasheh if (wait) { 192ccd979bdSMark Fasheh status = ocfs2_flush_truncate_log(osb); 193ccd979bdSMark Fasheh if (status < 0) 194ccd979bdSMark Fasheh mlog_errno(status); 195ccd979bdSMark Fasheh } else { 196ccd979bdSMark Fasheh ocfs2_schedule_truncate_log_flush(osb, 0); 197ccd979bdSMark Fasheh } 198ccd979bdSMark Fasheh 199ccd979bdSMark Fasheh if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) { 200ccd979bdSMark Fasheh if (wait) 201ccd979bdSMark Fasheh log_wait_commit(OCFS2_SB(sb)->journal->j_journal, 202ccd979bdSMark Fasheh target); 203ccd979bdSMark Fasheh } 204ccd979bdSMark Fasheh return 0; 205ccd979bdSMark Fasheh } 206ccd979bdSMark Fasheh 207ccd979bdSMark Fasheh static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) 208ccd979bdSMark Fasheh { 209ccd979bdSMark Fasheh struct inode *new = NULL; 210ccd979bdSMark Fasheh int status = 0; 211ccd979bdSMark Fasheh int i; 212ccd979bdSMark Fasheh 213ccd979bdSMark Fasheh mlog_entry_void(); 214ccd979bdSMark Fasheh 21524c19ef4SMark Fasheh new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE); 216ccd979bdSMark Fasheh if (IS_ERR(new)) { 217ccd979bdSMark Fasheh status = PTR_ERR(new); 218ccd979bdSMark Fasheh mlog_errno(status); 219ccd979bdSMark Fasheh goto bail; 220ccd979bdSMark Fasheh } 221ccd979bdSMark Fasheh osb->root_inode = new; 222ccd979bdSMark Fasheh 22324c19ef4SMark Fasheh new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE); 224ccd979bdSMark Fasheh if (IS_ERR(new)) { 225ccd979bdSMark Fasheh status = PTR_ERR(new); 226ccd979bdSMark Fasheh mlog_errno(status); 227ccd979bdSMark Fasheh goto bail; 228ccd979bdSMark Fasheh } 229ccd979bdSMark Fasheh osb->sys_root_inode = new; 230ccd979bdSMark Fasheh 231ccd979bdSMark Fasheh for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE; 232ccd979bdSMark Fasheh i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) { 233ccd979bdSMark Fasheh new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 234ccd979bdSMark Fasheh if (!new) { 235ccd979bdSMark Fasheh ocfs2_release_system_inodes(osb); 236ccd979bdSMark Fasheh status = -EINVAL; 237ccd979bdSMark Fasheh mlog_errno(status); 238ccd979bdSMark Fasheh /* FIXME: Should ERROR_RO_FS */ 239ccd979bdSMark Fasheh mlog(ML_ERROR, "Unable to load system inode %d, " 240ccd979bdSMark Fasheh "possibly corrupt fs?", i); 241ccd979bdSMark Fasheh goto bail; 242ccd979bdSMark Fasheh } 243ccd979bdSMark Fasheh // the array now has one ref, so drop this one 244ccd979bdSMark Fasheh iput(new); 245ccd979bdSMark Fasheh } 246ccd979bdSMark Fasheh 247ccd979bdSMark Fasheh bail: 248ccd979bdSMark Fasheh mlog_exit(status); 249ccd979bdSMark Fasheh return status; 250ccd979bdSMark Fasheh } 251ccd979bdSMark Fasheh 252ccd979bdSMark Fasheh static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) 253ccd979bdSMark Fasheh { 254ccd979bdSMark Fasheh struct inode *new = NULL; 255ccd979bdSMark Fasheh int status = 0; 256ccd979bdSMark Fasheh int i; 257ccd979bdSMark Fasheh 258ccd979bdSMark Fasheh mlog_entry_void(); 259ccd979bdSMark Fasheh 260ccd979bdSMark Fasheh for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1; 261ccd979bdSMark Fasheh i < NUM_SYSTEM_INODES; 262ccd979bdSMark Fasheh i++) { 263ccd979bdSMark Fasheh new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); 264ccd979bdSMark Fasheh if (!new) { 265ccd979bdSMark Fasheh ocfs2_release_system_inodes(osb); 266ccd979bdSMark Fasheh status = -EINVAL; 267ccd979bdSMark Fasheh mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n", 268ccd979bdSMark Fasheh status, i, osb->slot_num); 269ccd979bdSMark Fasheh goto bail; 270ccd979bdSMark Fasheh } 271ccd979bdSMark Fasheh /* the array now has one ref, so drop this one */ 272ccd979bdSMark Fasheh iput(new); 273ccd979bdSMark Fasheh } 274ccd979bdSMark Fasheh 275ccd979bdSMark Fasheh bail: 276ccd979bdSMark Fasheh mlog_exit(status); 277ccd979bdSMark Fasheh return status; 278ccd979bdSMark Fasheh } 279ccd979bdSMark Fasheh 280bddb8eb3SDenis Cheng static void ocfs2_release_system_inodes(struct ocfs2_super *osb) 281ccd979bdSMark Fasheh { 282bddb8eb3SDenis Cheng int i; 283ccd979bdSMark Fasheh struct inode *inode; 284ccd979bdSMark Fasheh 285ccd979bdSMark Fasheh mlog_entry_void(); 286ccd979bdSMark Fasheh 287ccd979bdSMark Fasheh for (i = 0; i < NUM_SYSTEM_INODES; i++) { 288ccd979bdSMark Fasheh inode = osb->system_inodes[i]; 289ccd979bdSMark Fasheh if (inode) { 290ccd979bdSMark Fasheh iput(inode); 291ccd979bdSMark Fasheh osb->system_inodes[i] = NULL; 292ccd979bdSMark Fasheh } 293ccd979bdSMark Fasheh } 294ccd979bdSMark Fasheh 295ccd979bdSMark Fasheh inode = osb->sys_root_inode; 296ccd979bdSMark Fasheh if (inode) { 297ccd979bdSMark Fasheh iput(inode); 298ccd979bdSMark Fasheh osb->sys_root_inode = NULL; 299ccd979bdSMark Fasheh } 300ccd979bdSMark Fasheh 301ccd979bdSMark Fasheh inode = osb->root_inode; 302ccd979bdSMark Fasheh if (inode) { 303ccd979bdSMark Fasheh iput(inode); 304ccd979bdSMark Fasheh osb->root_inode = NULL; 305ccd979bdSMark Fasheh } 306ccd979bdSMark Fasheh 307bddb8eb3SDenis Cheng mlog_exit(0); 308ccd979bdSMark Fasheh } 309ccd979bdSMark Fasheh 310ccd979bdSMark Fasheh /* We're allocating fs objects, use GFP_NOFS */ 311ccd979bdSMark Fasheh static struct inode *ocfs2_alloc_inode(struct super_block *sb) 312ccd979bdSMark Fasheh { 313ccd979bdSMark Fasheh struct ocfs2_inode_info *oi; 314ccd979bdSMark Fasheh 315e6b4f8daSChristoph Lameter oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS); 316ccd979bdSMark Fasheh if (!oi) 317ccd979bdSMark Fasheh return NULL; 318ccd979bdSMark Fasheh 319ccd979bdSMark Fasheh return &oi->vfs_inode; 320ccd979bdSMark Fasheh } 321ccd979bdSMark Fasheh 322ccd979bdSMark Fasheh static void ocfs2_destroy_inode(struct inode *inode) 323ccd979bdSMark Fasheh { 324ccd979bdSMark Fasheh kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); 325ccd979bdSMark Fasheh } 326ccd979bdSMark Fasheh 3275a254031SMark Fasheh static unsigned long long ocfs2_max_file_offset(unsigned int bbits, 3285a254031SMark Fasheh unsigned int cbits) 329ccd979bdSMark Fasheh { 3305a254031SMark Fasheh unsigned int bytes = 1 << cbits; 3315a254031SMark Fasheh unsigned int trim = bytes; 3325a254031SMark Fasheh unsigned int bitshift = 32; 333ccd979bdSMark Fasheh 3345a254031SMark Fasheh /* 3355a254031SMark Fasheh * i_size and all block offsets in ocfs2 are always 64 bits 3365a254031SMark Fasheh * wide. i_clusters is 32 bits, in cluster-sized units. So on 3375a254031SMark Fasheh * 64 bit platforms, cluster size will be the limiting factor. 338ccd979bdSMark Fasheh */ 339ccd979bdSMark Fasheh 340ccd979bdSMark Fasheh #if BITS_PER_LONG == 32 341ccd979bdSMark Fasheh # if defined(CONFIG_LBD) 3422ecd05aeSAlexey Dobriyan BUILD_BUG_ON(sizeof(sector_t) != 8); 3435a254031SMark Fasheh /* 3445a254031SMark Fasheh * We might be limited by page cache size. 3455a254031SMark Fasheh */ 3465a254031SMark Fasheh if (bytes > PAGE_CACHE_SIZE) { 3475a254031SMark Fasheh bytes = PAGE_CACHE_SIZE; 3485a254031SMark Fasheh trim = 1; 3495a254031SMark Fasheh /* 3505a254031SMark Fasheh * Shift by 31 here so that we don't get larger than 3515a254031SMark Fasheh * MAX_LFS_FILESIZE 3525a254031SMark Fasheh */ 3535a254031SMark Fasheh bitshift = 31; 3545a254031SMark Fasheh } 355ccd979bdSMark Fasheh # else 3565a254031SMark Fasheh /* 3575a254031SMark Fasheh * We are limited by the size of sector_t. Use block size, as 3585a254031SMark Fasheh * that's what we expose to the VFS. 3595a254031SMark Fasheh */ 3605a254031SMark Fasheh bytes = 1 << bbits; 3615a254031SMark Fasheh trim = 1; 3625a254031SMark Fasheh bitshift = 31; 363ccd979bdSMark Fasheh # endif 364ccd979bdSMark Fasheh #endif 365ccd979bdSMark Fasheh 3665a254031SMark Fasheh /* 3675a254031SMark Fasheh * Trim by a whole cluster when we can actually approach the 3685a254031SMark Fasheh * on-disk limits. Otherwise we can overflow i_clusters when 3695a254031SMark Fasheh * an extent start is at the max offset. 3705a254031SMark Fasheh */ 3715a254031SMark Fasheh return (((unsigned long long)bytes) << bitshift) - trim; 372ccd979bdSMark Fasheh } 373ccd979bdSMark Fasheh 374ccd979bdSMark Fasheh static int ocfs2_remount(struct super_block *sb, int *flags, char *data) 375ccd979bdSMark Fasheh { 376ccd979bdSMark Fasheh int incompat_features; 377ccd979bdSMark Fasheh int ret = 0; 378c0123adeSTiger Yang struct mount_options parsed_options; 379ccd979bdSMark Fasheh struct ocfs2_super *osb = OCFS2_SB(sb); 380ccd979bdSMark Fasheh 381c0123adeSTiger Yang if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) { 382ccd979bdSMark Fasheh ret = -EINVAL; 383ccd979bdSMark Fasheh goto out; 384ccd979bdSMark Fasheh } 385ccd979bdSMark Fasheh 386ccd979bdSMark Fasheh if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) != 387c0123adeSTiger Yang (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) { 388ccd979bdSMark Fasheh ret = -EINVAL; 389ccd979bdSMark Fasheh mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); 390ccd979bdSMark Fasheh goto out; 391ccd979bdSMark Fasheh } 392ccd979bdSMark Fasheh 393ccd979bdSMark Fasheh if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) != 394c0123adeSTiger Yang (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) { 395ccd979bdSMark Fasheh ret = -EINVAL; 396ccd979bdSMark Fasheh mlog(ML_ERROR, "Cannot change data mode on remount\n"); 397ccd979bdSMark Fasheh goto out; 398ccd979bdSMark Fasheh } 399ccd979bdSMark Fasheh 400ccd979bdSMark Fasheh /* We're going to/from readonly mode. */ 401ccd979bdSMark Fasheh if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { 402ccd979bdSMark Fasheh /* Lock here so the check of HARD_RO and the potential 403ccd979bdSMark Fasheh * setting of SOFT_RO is atomic. */ 404ccd979bdSMark Fasheh spin_lock(&osb->osb_lock); 405ccd979bdSMark Fasheh if (osb->osb_flags & OCFS2_OSB_HARD_RO) { 406ccd979bdSMark Fasheh mlog(ML_ERROR, "Remount on readonly device is forbidden.\n"); 407ccd979bdSMark Fasheh ret = -EROFS; 408ccd979bdSMark Fasheh goto unlock_osb; 409ccd979bdSMark Fasheh } 410ccd979bdSMark Fasheh 411ccd979bdSMark Fasheh if (*flags & MS_RDONLY) { 412ccd979bdSMark Fasheh mlog(0, "Going to ro mode.\n"); 413ccd979bdSMark Fasheh sb->s_flags |= MS_RDONLY; 414ccd979bdSMark Fasheh osb->osb_flags |= OCFS2_OSB_SOFT_RO; 415ccd979bdSMark Fasheh } else { 416ccd979bdSMark Fasheh mlog(0, "Making ro filesystem writeable.\n"); 417ccd979bdSMark Fasheh 418ccd979bdSMark Fasheh if (osb->osb_flags & OCFS2_OSB_ERROR_FS) { 419ccd979bdSMark Fasheh mlog(ML_ERROR, "Cannot remount RDWR " 420ccd979bdSMark Fasheh "filesystem due to previous errors.\n"); 421ccd979bdSMark Fasheh ret = -EROFS; 422ccd979bdSMark Fasheh goto unlock_osb; 423ccd979bdSMark Fasheh } 424ccd979bdSMark Fasheh incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP); 425ccd979bdSMark Fasheh if (incompat_features) { 426ccd979bdSMark Fasheh mlog(ML_ERROR, "Cannot remount RDWR because " 427ccd979bdSMark Fasheh "of unsupported optional features " 428ccd979bdSMark Fasheh "(%x).\n", incompat_features); 429ccd979bdSMark Fasheh ret = -EINVAL; 430ccd979bdSMark Fasheh goto unlock_osb; 431ccd979bdSMark Fasheh } 432ccd979bdSMark Fasheh sb->s_flags &= ~MS_RDONLY; 433ccd979bdSMark Fasheh osb->osb_flags &= ~OCFS2_OSB_SOFT_RO; 434ccd979bdSMark Fasheh } 435ccd979bdSMark Fasheh unlock_osb: 436ccd979bdSMark Fasheh spin_unlock(&osb->osb_lock); 437ccd979bdSMark Fasheh } 438ccd979bdSMark Fasheh 439ccd979bdSMark Fasheh if (!ret) { 440ccd979bdSMark Fasheh /* Only save off the new mount options in case of a successful 441ccd979bdSMark Fasheh * remount. */ 442c0123adeSTiger Yang osb->s_mount_opt = parsed_options.mount_opt; 443c0123adeSTiger Yang osb->s_atime_quantum = parsed_options.atime_quantum; 444c0123adeSTiger Yang osb->preferred_slot = parsed_options.slot; 445e001e796SMark Fasheh 446e001e796SMark Fasheh if (!ocfs2_is_hard_readonly(osb)) 447e001e796SMark Fasheh ocfs2_set_journal_params(osb); 448ccd979bdSMark Fasheh } 449ccd979bdSMark Fasheh out: 450ccd979bdSMark Fasheh return ret; 451ccd979bdSMark Fasheh } 452ccd979bdSMark Fasheh 453ccd979bdSMark Fasheh static int ocfs2_sb_probe(struct super_block *sb, 454ccd979bdSMark Fasheh struct buffer_head **bh, 455ccd979bdSMark Fasheh int *sector_size) 456ccd979bdSMark Fasheh { 457bddb8eb3SDenis Cheng int status, tmpstat; 458ccd979bdSMark Fasheh struct ocfs1_vol_disk_hdr *hdr; 459ccd979bdSMark Fasheh struct ocfs2_dinode *di; 460ccd979bdSMark Fasheh int blksize; 461ccd979bdSMark Fasheh 462ccd979bdSMark Fasheh *bh = NULL; 463ccd979bdSMark Fasheh 464ccd979bdSMark Fasheh /* may be > 512 */ 465ccd979bdSMark Fasheh *sector_size = bdev_hardsect_size(sb->s_bdev); 466ccd979bdSMark Fasheh if (*sector_size > OCFS2_MAX_BLOCKSIZE) { 467ccd979bdSMark Fasheh mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n", 468ccd979bdSMark Fasheh *sector_size, OCFS2_MAX_BLOCKSIZE); 469ccd979bdSMark Fasheh status = -EINVAL; 470ccd979bdSMark Fasheh goto bail; 471ccd979bdSMark Fasheh } 472ccd979bdSMark Fasheh 473ccd979bdSMark Fasheh /* Can this really happen? */ 474ccd979bdSMark Fasheh if (*sector_size < OCFS2_MIN_BLOCKSIZE) 475ccd979bdSMark Fasheh *sector_size = OCFS2_MIN_BLOCKSIZE; 476ccd979bdSMark Fasheh 477ccd979bdSMark Fasheh /* check block zero for old format */ 478ccd979bdSMark Fasheh status = ocfs2_get_sector(sb, bh, 0, *sector_size); 479ccd979bdSMark Fasheh if (status < 0) { 480ccd979bdSMark Fasheh mlog_errno(status); 481ccd979bdSMark Fasheh goto bail; 482ccd979bdSMark Fasheh } 483ccd979bdSMark Fasheh hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data; 484ccd979bdSMark Fasheh if (hdr->major_version == OCFS1_MAJOR_VERSION) { 485ccd979bdSMark Fasheh mlog(ML_ERROR, "incompatible version: %u.%u\n", 486ccd979bdSMark Fasheh hdr->major_version, hdr->minor_version); 487ccd979bdSMark Fasheh status = -EINVAL; 488ccd979bdSMark Fasheh } 489ccd979bdSMark Fasheh if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE, 490ccd979bdSMark Fasheh strlen(OCFS1_VOLUME_SIGNATURE)) == 0) { 491ccd979bdSMark Fasheh mlog(ML_ERROR, "incompatible volume signature: %8s\n", 492ccd979bdSMark Fasheh hdr->signature); 493ccd979bdSMark Fasheh status = -EINVAL; 494ccd979bdSMark Fasheh } 495ccd979bdSMark Fasheh brelse(*bh); 496ccd979bdSMark Fasheh *bh = NULL; 497ccd979bdSMark Fasheh if (status < 0) { 498ccd979bdSMark Fasheh mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be " 499ccd979bdSMark Fasheh "upgraded before mounting with ocfs v2\n"); 500ccd979bdSMark Fasheh goto bail; 501ccd979bdSMark Fasheh } 502ccd979bdSMark Fasheh 503ccd979bdSMark Fasheh /* 504ccd979bdSMark Fasheh * Now check at magic offset for 512, 1024, 2048, 4096 505ccd979bdSMark Fasheh * blocksizes. 4096 is the maximum blocksize because it is 506ccd979bdSMark Fasheh * the minimum clustersize. 507ccd979bdSMark Fasheh */ 508ccd979bdSMark Fasheh status = -EINVAL; 509ccd979bdSMark Fasheh for (blksize = *sector_size; 510ccd979bdSMark Fasheh blksize <= OCFS2_MAX_BLOCKSIZE; 511ccd979bdSMark Fasheh blksize <<= 1) { 512ccd979bdSMark Fasheh tmpstat = ocfs2_get_sector(sb, bh, 513ccd979bdSMark Fasheh OCFS2_SUPER_BLOCK_BLKNO, 514ccd979bdSMark Fasheh blksize); 515ccd979bdSMark Fasheh if (tmpstat < 0) { 516ccd979bdSMark Fasheh status = tmpstat; 517ccd979bdSMark Fasheh mlog_errno(status); 518ccd979bdSMark Fasheh goto bail; 519ccd979bdSMark Fasheh } 520ccd979bdSMark Fasheh di = (struct ocfs2_dinode *) (*bh)->b_data; 521ccd979bdSMark Fasheh status = ocfs2_verify_volume(di, *bh, blksize); 522ccd979bdSMark Fasheh if (status >= 0) 523ccd979bdSMark Fasheh goto bail; 524ccd979bdSMark Fasheh brelse(*bh); 525ccd979bdSMark Fasheh *bh = NULL; 526ccd979bdSMark Fasheh if (status != -EAGAIN) 527ccd979bdSMark Fasheh break; 528ccd979bdSMark Fasheh } 529ccd979bdSMark Fasheh 530ccd979bdSMark Fasheh bail: 531ccd979bdSMark Fasheh return status; 532ccd979bdSMark Fasheh } 533ccd979bdSMark Fasheh 534c271c5c2SSunil Mushran static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) 535c271c5c2SSunil Mushran { 536c271c5c2SSunil Mushran if (ocfs2_mount_local(osb)) { 537c271c5c2SSunil Mushran if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 538c271c5c2SSunil Mushran mlog(ML_ERROR, "Cannot heartbeat on a locally " 539c271c5c2SSunil Mushran "mounted device.\n"); 540c271c5c2SSunil Mushran return -EINVAL; 541c271c5c2SSunil Mushran } 542c271c5c2SSunil Mushran } 543c271c5c2SSunil Mushran 544c271c5c2SSunil Mushran if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) { 545c271c5c2SSunil Mushran if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) { 546c271c5c2SSunil Mushran mlog(ML_ERROR, "Heartbeat has to be started to mount " 547c271c5c2SSunil Mushran "a read-write clustered device.\n"); 548c271c5c2SSunil Mushran return -EINVAL; 549c271c5c2SSunil Mushran } 550c271c5c2SSunil Mushran } 551c271c5c2SSunil Mushran 552c271c5c2SSunil Mushran return 0; 553c271c5c2SSunil Mushran } 554c271c5c2SSunil Mushran 555ccd979bdSMark Fasheh static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) 556ccd979bdSMark Fasheh { 557ccd979bdSMark Fasheh struct dentry *root; 558ccd979bdSMark Fasheh int status, sector_size; 559c0123adeSTiger Yang struct mount_options parsed_options; 560ccd979bdSMark Fasheh struct inode *inode = NULL; 561ccd979bdSMark Fasheh struct ocfs2_super *osb = NULL; 562ccd979bdSMark Fasheh struct buffer_head *bh = NULL; 563c271c5c2SSunil Mushran char nodestr[8]; 564ccd979bdSMark Fasheh 565ccd979bdSMark Fasheh mlog_entry("%p, %p, %i", sb, data, silent); 566ccd979bdSMark Fasheh 567c0123adeSTiger Yang if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { 568c271c5c2SSunil Mushran status = -EINVAL; 569c271c5c2SSunil Mushran goto read_super_error; 570c271c5c2SSunil Mushran } 571c271c5c2SSunil Mushran 572ccd979bdSMark Fasheh /* for now we only have one cluster/node, make sure we see it 573ccd979bdSMark Fasheh * in the heartbeat universe */ 574c0123adeSTiger Yang if (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL) { 575ccd979bdSMark Fasheh if (!o2hb_check_local_node_heartbeating()) { 576ccd979bdSMark Fasheh status = -EINVAL; 577ccd979bdSMark Fasheh goto read_super_error; 578ccd979bdSMark Fasheh } 579c271c5c2SSunil Mushran } 580ccd979bdSMark Fasheh 581ccd979bdSMark Fasheh /* probe for superblock */ 582ccd979bdSMark Fasheh status = ocfs2_sb_probe(sb, &bh, §or_size); 583ccd979bdSMark Fasheh if (status < 0) { 584ccd979bdSMark Fasheh mlog(ML_ERROR, "superblock probe failed!\n"); 585ccd979bdSMark Fasheh goto read_super_error; 586ccd979bdSMark Fasheh } 587ccd979bdSMark Fasheh 588ccd979bdSMark Fasheh status = ocfs2_initialize_super(sb, bh, sector_size); 589ccd979bdSMark Fasheh osb = OCFS2_SB(sb); 590ccd979bdSMark Fasheh if (status < 0) { 591ccd979bdSMark Fasheh mlog_errno(status); 592ccd979bdSMark Fasheh goto read_super_error; 593ccd979bdSMark Fasheh } 594ccd979bdSMark Fasheh brelse(bh); 595ccd979bdSMark Fasheh bh = NULL; 596c0123adeSTiger Yang osb->s_mount_opt = parsed_options.mount_opt; 597c0123adeSTiger Yang osb->s_atime_quantum = parsed_options.atime_quantum; 598c0123adeSTiger Yang osb->preferred_slot = parsed_options.slot; 599ccd979bdSMark Fasheh 600ccd979bdSMark Fasheh sb->s_magic = OCFS2_SUPER_MAGIC; 601ccd979bdSMark Fasheh 602ccd979bdSMark Fasheh /* Hard readonly mode only if: bdev_read_only, MS_RDONLY, 603ccd979bdSMark Fasheh * heartbeat=none */ 604ccd979bdSMark Fasheh if (bdev_read_only(sb->s_bdev)) { 605ccd979bdSMark Fasheh if (!(sb->s_flags & MS_RDONLY)) { 606ccd979bdSMark Fasheh status = -EACCES; 607ccd979bdSMark Fasheh mlog(ML_ERROR, "Readonly device detected but readonly " 608ccd979bdSMark Fasheh "mount was not specified.\n"); 609ccd979bdSMark Fasheh goto read_super_error; 610ccd979bdSMark Fasheh } 611ccd979bdSMark Fasheh 612ccd979bdSMark Fasheh /* You should not be able to start a local heartbeat 613ccd979bdSMark Fasheh * on a readonly device. */ 614ccd979bdSMark Fasheh if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { 615ccd979bdSMark Fasheh status = -EROFS; 616ccd979bdSMark Fasheh mlog(ML_ERROR, "Local heartbeat specified on readonly " 617ccd979bdSMark Fasheh "device.\n"); 618ccd979bdSMark Fasheh goto read_super_error; 619ccd979bdSMark Fasheh } 620ccd979bdSMark Fasheh 621ccd979bdSMark Fasheh status = ocfs2_check_journals_nolocks(osb); 622ccd979bdSMark Fasheh if (status < 0) { 623ccd979bdSMark Fasheh if (status == -EROFS) 624ccd979bdSMark Fasheh mlog(ML_ERROR, "Recovery required on readonly " 625ccd979bdSMark Fasheh "file system, but write access is " 626ccd979bdSMark Fasheh "unavailable.\n"); 627ccd979bdSMark Fasheh else 628ccd979bdSMark Fasheh mlog_errno(status); 629ccd979bdSMark Fasheh goto read_super_error; 630ccd979bdSMark Fasheh } 631ccd979bdSMark Fasheh 632ccd979bdSMark Fasheh ocfs2_set_ro_flag(osb, 1); 633ccd979bdSMark Fasheh 634ccd979bdSMark Fasheh printk(KERN_NOTICE "Readonly device detected. No cluster " 635ccd979bdSMark Fasheh "services will be utilized for this mount. Recovery " 636ccd979bdSMark Fasheh "will be skipped.\n"); 637ccd979bdSMark Fasheh } 638ccd979bdSMark Fasheh 639ccd979bdSMark Fasheh if (!ocfs2_is_hard_readonly(osb)) { 640ccd979bdSMark Fasheh if (sb->s_flags & MS_RDONLY) 641ccd979bdSMark Fasheh ocfs2_set_ro_flag(osb, 0); 642ccd979bdSMark Fasheh } 643ccd979bdSMark Fasheh 644c271c5c2SSunil Mushran status = ocfs2_verify_heartbeat(osb); 645c271c5c2SSunil Mushran if (status < 0) { 646c271c5c2SSunil Mushran mlog_errno(status); 647c271c5c2SSunil Mushran goto read_super_error; 648c271c5c2SSunil Mushran } 649c271c5c2SSunil Mushran 650ccd979bdSMark Fasheh osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, 651ccd979bdSMark Fasheh ocfs2_debugfs_root); 652ccd979bdSMark Fasheh if (!osb->osb_debug_root) { 653ccd979bdSMark Fasheh status = -EINVAL; 654ccd979bdSMark Fasheh mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n"); 655ccd979bdSMark Fasheh goto read_super_error; 656ccd979bdSMark Fasheh } 657ccd979bdSMark Fasheh 658ccd979bdSMark Fasheh status = ocfs2_mount_volume(sb); 659ccd979bdSMark Fasheh if (osb->root_inode) 660ccd979bdSMark Fasheh inode = igrab(osb->root_inode); 661ccd979bdSMark Fasheh 662ccd979bdSMark Fasheh if (status < 0) 663ccd979bdSMark Fasheh goto read_super_error; 664ccd979bdSMark Fasheh 665ccd979bdSMark Fasheh if (!inode) { 666ccd979bdSMark Fasheh status = -EIO; 667ccd979bdSMark Fasheh mlog_errno(status); 668ccd979bdSMark Fasheh goto read_super_error; 669ccd979bdSMark Fasheh } 670ccd979bdSMark Fasheh 671ccd979bdSMark Fasheh root = d_alloc_root(inode); 672ccd979bdSMark Fasheh if (!root) { 673ccd979bdSMark Fasheh status = -ENOMEM; 674ccd979bdSMark Fasheh mlog_errno(status); 675ccd979bdSMark Fasheh goto read_super_error; 676ccd979bdSMark Fasheh } 677ccd979bdSMark Fasheh 678ccd979bdSMark Fasheh sb->s_root = root; 679ccd979bdSMark Fasheh 680ccd979bdSMark Fasheh ocfs2_complete_mount_recovery(osb); 681ccd979bdSMark Fasheh 682c271c5c2SSunil Mushran if (ocfs2_mount_local(osb)) 683c271c5c2SSunil Mushran snprintf(nodestr, sizeof(nodestr), "local"); 684c271c5c2SSunil Mushran else 685c271c5c2SSunil Mushran snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num); 686c271c5c2SSunil Mushran 687c271c5c2SSunil Mushran printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " 688781ee3e2SSunil Mushran "with %s data mode.\n", 689c271c5c2SSunil Mushran osb->dev_str, nodestr, osb->slot_num, 690ccd979bdSMark Fasheh osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : 691ccd979bdSMark Fasheh "ordered"); 692ccd979bdSMark Fasheh 693ccd979bdSMark Fasheh atomic_set(&osb->vol_state, VOLUME_MOUNTED); 694ccd979bdSMark Fasheh wake_up(&osb->osb_mount_event); 695ccd979bdSMark Fasheh 696ccd979bdSMark Fasheh mlog_exit(status); 697ccd979bdSMark Fasheh return status; 698ccd979bdSMark Fasheh 699ccd979bdSMark Fasheh read_super_error: 700ccd979bdSMark Fasheh if (bh != NULL) 701ccd979bdSMark Fasheh brelse(bh); 702ccd979bdSMark Fasheh 703ccd979bdSMark Fasheh if (inode) 704ccd979bdSMark Fasheh iput(inode); 705ccd979bdSMark Fasheh 706ccd979bdSMark Fasheh if (osb) { 707ccd979bdSMark Fasheh atomic_set(&osb->vol_state, VOLUME_DISABLED); 708ccd979bdSMark Fasheh wake_up(&osb->osb_mount_event); 709ccd979bdSMark Fasheh ocfs2_dismount_volume(sb, 1); 710ccd979bdSMark Fasheh } 711ccd979bdSMark Fasheh 712ccd979bdSMark Fasheh mlog_exit(status); 713ccd979bdSMark Fasheh return status; 714ccd979bdSMark Fasheh } 715ccd979bdSMark Fasheh 716454e2398SDavid Howells static int ocfs2_get_sb(struct file_system_type *fs_type, 717ccd979bdSMark Fasheh int flags, 718ccd979bdSMark Fasheh const char *dev_name, 719454e2398SDavid Howells void *data, 720454e2398SDavid Howells struct vfsmount *mnt) 721ccd979bdSMark Fasheh { 722454e2398SDavid Howells return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super, 723454e2398SDavid Howells mnt); 724ccd979bdSMark Fasheh } 725ccd979bdSMark Fasheh 726ccd979bdSMark Fasheh static struct file_system_type ocfs2_fs_type = { 727ccd979bdSMark Fasheh .owner = THIS_MODULE, 728ccd979bdSMark Fasheh .name = "ocfs2", 729ccd979bdSMark Fasheh .get_sb = ocfs2_get_sb, /* is this called when we mount 730ccd979bdSMark Fasheh * the fs? */ 731ccd979bdSMark Fasheh .kill_sb = kill_block_super, /* set to the generic one 732ccd979bdSMark Fasheh * right now, but do we 733ccd979bdSMark Fasheh * need to change that? */ 7341ba9da2fSMark Fasheh .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, 735ccd979bdSMark Fasheh .next = NULL 736ccd979bdSMark Fasheh }; 737ccd979bdSMark Fasheh 738ccd979bdSMark Fasheh static int ocfs2_parse_options(struct super_block *sb, 739ccd979bdSMark Fasheh char *options, 740c0123adeSTiger Yang struct mount_options *mopt, 741ccd979bdSMark Fasheh int is_remount) 742ccd979bdSMark Fasheh { 743ccd979bdSMark Fasheh int status; 744ccd979bdSMark Fasheh char *p; 745ccd979bdSMark Fasheh 746ccd979bdSMark Fasheh mlog_entry("remount: %d, options: \"%s\"\n", is_remount, 747ccd979bdSMark Fasheh options ? options : "(none)"); 748ccd979bdSMark Fasheh 749c0123adeSTiger Yang mopt->mount_opt = 0; 750c0123adeSTiger Yang mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 751c0123adeSTiger Yang mopt->slot = OCFS2_INVALID_SLOT; 752ccd979bdSMark Fasheh 753ccd979bdSMark Fasheh if (!options) { 754ccd979bdSMark Fasheh status = 1; 755ccd979bdSMark Fasheh goto bail; 756ccd979bdSMark Fasheh } 757ccd979bdSMark Fasheh 758ccd979bdSMark Fasheh while ((p = strsep(&options, ",")) != NULL) { 759ccd979bdSMark Fasheh int token, option; 760ccd979bdSMark Fasheh substring_t args[MAX_OPT_ARGS]; 761ccd979bdSMark Fasheh 762ccd979bdSMark Fasheh if (!*p) 763ccd979bdSMark Fasheh continue; 764ccd979bdSMark Fasheh 765ccd979bdSMark Fasheh token = match_token(p, tokens, args); 766ccd979bdSMark Fasheh switch (token) { 767ccd979bdSMark Fasheh case Opt_hb_local: 768c0123adeSTiger Yang mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; 769ccd979bdSMark Fasheh break; 770ccd979bdSMark Fasheh case Opt_hb_none: 771c0123adeSTiger Yang mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; 772ccd979bdSMark Fasheh break; 773ccd979bdSMark Fasheh case Opt_barrier: 774ccd979bdSMark Fasheh if (match_int(&args[0], &option)) { 775ccd979bdSMark Fasheh status = 0; 776ccd979bdSMark Fasheh goto bail; 777ccd979bdSMark Fasheh } 778ccd979bdSMark Fasheh if (option) 779c0123adeSTiger Yang mopt->mount_opt |= OCFS2_MOUNT_BARRIER; 780ccd979bdSMark Fasheh else 781c0123adeSTiger Yang mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; 782ccd979bdSMark Fasheh break; 783ccd979bdSMark Fasheh case Opt_intr: 784c0123adeSTiger Yang mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; 785ccd979bdSMark Fasheh break; 786ccd979bdSMark Fasheh case Opt_nointr: 787c0123adeSTiger Yang mopt->mount_opt |= OCFS2_MOUNT_NOINTR; 788ccd979bdSMark Fasheh break; 789ccd979bdSMark Fasheh case Opt_err_panic: 790c0123adeSTiger Yang mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 791ccd979bdSMark Fasheh break; 792ccd979bdSMark Fasheh case Opt_err_ro: 793c0123adeSTiger Yang mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; 794ccd979bdSMark Fasheh break; 795ccd979bdSMark Fasheh case Opt_data_ordered: 796c0123adeSTiger Yang mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; 797ccd979bdSMark Fasheh break; 798ccd979bdSMark Fasheh case Opt_data_writeback: 799c0123adeSTiger Yang mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; 800ccd979bdSMark Fasheh break; 8017f1a37e3STiger Yang case Opt_atime_quantum: 8027f1a37e3STiger Yang if (match_int(&args[0], &option)) { 8037f1a37e3STiger Yang status = 0; 8047f1a37e3STiger Yang goto bail; 8057f1a37e3STiger Yang } 8067f1a37e3STiger Yang if (option >= 0) 807c0123adeSTiger Yang mopt->atime_quantum = option; 8087f1a37e3STiger Yang break; 809baf4661aSSunil Mushran case Opt_slot: 810baf4661aSSunil Mushran option = 0; 811baf4661aSSunil Mushran if (match_int(&args[0], &option)) { 812baf4661aSSunil Mushran status = 0; 813baf4661aSSunil Mushran goto bail; 814baf4661aSSunil Mushran } 815baf4661aSSunil Mushran if (option) 816c0123adeSTiger Yang mopt->slot = (s16)option; 817baf4661aSSunil Mushran break; 818ccd979bdSMark Fasheh default: 819ccd979bdSMark Fasheh mlog(ML_ERROR, 820ccd979bdSMark Fasheh "Unrecognized mount option \"%s\" " 821ccd979bdSMark Fasheh "or missing value\n", p); 822ccd979bdSMark Fasheh status = 0; 823ccd979bdSMark Fasheh goto bail; 824ccd979bdSMark Fasheh } 825ccd979bdSMark Fasheh } 826ccd979bdSMark Fasheh 827ccd979bdSMark Fasheh status = 1; 828ccd979bdSMark Fasheh 829ccd979bdSMark Fasheh bail: 830ccd979bdSMark Fasheh mlog_exit(status); 831ccd979bdSMark Fasheh return status; 832ccd979bdSMark Fasheh } 833ccd979bdSMark Fasheh 834d550071cSSunil Mushran static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) 835d550071cSSunil Mushran { 836d550071cSSunil Mushran struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb); 837d550071cSSunil Mushran unsigned long opts = osb->s_mount_opt; 838d550071cSSunil Mushran 839d550071cSSunil Mushran if (opts & OCFS2_MOUNT_HB_LOCAL) 840d550071cSSunil Mushran seq_printf(s, ",_netdev,heartbeat=local"); 841d550071cSSunil Mushran else 842d550071cSSunil Mushran seq_printf(s, ",heartbeat=none"); 843d550071cSSunil Mushran 844d550071cSSunil Mushran if (opts & OCFS2_MOUNT_NOINTR) 845d550071cSSunil Mushran seq_printf(s, ",nointr"); 846d550071cSSunil Mushran 847d550071cSSunil Mushran if (opts & OCFS2_MOUNT_DATA_WRITEBACK) 848d550071cSSunil Mushran seq_printf(s, ",data=writeback"); 849d550071cSSunil Mushran else 850d550071cSSunil Mushran seq_printf(s, ",data=ordered"); 851d550071cSSunil Mushran 852d550071cSSunil Mushran if (opts & OCFS2_MOUNT_BARRIER) 853d550071cSSunil Mushran seq_printf(s, ",barrier=1"); 854d550071cSSunil Mushran 855d550071cSSunil Mushran if (opts & OCFS2_MOUNT_ERRORS_PANIC) 856d550071cSSunil Mushran seq_printf(s, ",errors=panic"); 857d550071cSSunil Mushran else 858d550071cSSunil Mushran seq_printf(s, ",errors=remount-ro"); 859d550071cSSunil Mushran 860d550071cSSunil Mushran if (osb->preferred_slot != OCFS2_INVALID_SLOT) 861d550071cSSunil Mushran seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); 862d550071cSSunil Mushran 863d550071cSSunil Mushran if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM) 864d550071cSSunil Mushran seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); 865d550071cSSunil Mushran 866d550071cSSunil Mushran return 0; 867d550071cSSunil Mushran } 868d550071cSSunil Mushran 869ccd979bdSMark Fasheh static int __init ocfs2_init(void) 870ccd979bdSMark Fasheh { 871ccd979bdSMark Fasheh int status; 872ccd979bdSMark Fasheh 873ccd979bdSMark Fasheh mlog_entry_void(); 874ccd979bdSMark Fasheh 875ccd979bdSMark Fasheh ocfs2_print_version(); 876ccd979bdSMark Fasheh 877ccd979bdSMark Fasheh status = init_ocfs2_uptodate_cache(); 878ccd979bdSMark Fasheh if (status < 0) { 879ccd979bdSMark Fasheh mlog_errno(status); 880ccd979bdSMark Fasheh goto leave; 881ccd979bdSMark Fasheh } 882ccd979bdSMark Fasheh 883ccd979bdSMark Fasheh status = ocfs2_initialize_mem_caches(); 884ccd979bdSMark Fasheh if (status < 0) { 885ccd979bdSMark Fasheh mlog_errno(status); 886ccd979bdSMark Fasheh goto leave; 887ccd979bdSMark Fasheh } 888ccd979bdSMark Fasheh 889ccd979bdSMark Fasheh ocfs2_wq = create_singlethread_workqueue("ocfs2_wq"); 890ccd979bdSMark Fasheh if (!ocfs2_wq) { 891ccd979bdSMark Fasheh status = -ENOMEM; 892ccd979bdSMark Fasheh goto leave; 893ccd979bdSMark Fasheh } 894ccd979bdSMark Fasheh 895ccd979bdSMark Fasheh ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); 896ccd979bdSMark Fasheh if (!ocfs2_debugfs_root) { 897ccd979bdSMark Fasheh status = -EFAULT; 898ccd979bdSMark Fasheh mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n"); 899ccd979bdSMark Fasheh } 900ccd979bdSMark Fasheh 901ccd979bdSMark Fasheh leave: 902ccd979bdSMark Fasheh if (status < 0) { 903ccd979bdSMark Fasheh ocfs2_free_mem_caches(); 904ccd979bdSMark Fasheh exit_ocfs2_uptodate_cache(); 905ccd979bdSMark Fasheh } 906ccd979bdSMark Fasheh 907ccd979bdSMark Fasheh mlog_exit(status); 908ccd979bdSMark Fasheh 909ccd979bdSMark Fasheh if (status >= 0) { 910ccd979bdSMark Fasheh return register_filesystem(&ocfs2_fs_type); 911ccd979bdSMark Fasheh } else 912ccd979bdSMark Fasheh return -1; 913ccd979bdSMark Fasheh } 914ccd979bdSMark Fasheh 915ccd979bdSMark Fasheh static void __exit ocfs2_exit(void) 916ccd979bdSMark Fasheh { 917ccd979bdSMark Fasheh mlog_entry_void(); 918ccd979bdSMark Fasheh 919ccd979bdSMark Fasheh if (ocfs2_wq) { 920ccd979bdSMark Fasheh flush_workqueue(ocfs2_wq); 921ccd979bdSMark Fasheh destroy_workqueue(ocfs2_wq); 922ccd979bdSMark Fasheh } 923ccd979bdSMark Fasheh 924ccd979bdSMark Fasheh debugfs_remove(ocfs2_debugfs_root); 925ccd979bdSMark Fasheh 926ccd979bdSMark Fasheh ocfs2_free_mem_caches(); 927ccd979bdSMark Fasheh 928ccd979bdSMark Fasheh unregister_filesystem(&ocfs2_fs_type); 929ccd979bdSMark Fasheh 930ccd979bdSMark Fasheh exit_ocfs2_uptodate_cache(); 931ccd979bdSMark Fasheh 932ccd979bdSMark Fasheh mlog_exit_void(); 933ccd979bdSMark Fasheh } 934ccd979bdSMark Fasheh 935ccd979bdSMark Fasheh static void ocfs2_put_super(struct super_block *sb) 936ccd979bdSMark Fasheh { 937ccd979bdSMark Fasheh mlog_entry("(0x%p)\n", sb); 938ccd979bdSMark Fasheh 939ccd979bdSMark Fasheh ocfs2_sync_blockdev(sb); 940ccd979bdSMark Fasheh ocfs2_dismount_volume(sb, 0); 941ccd979bdSMark Fasheh 942ccd979bdSMark Fasheh mlog_exit_void(); 943ccd979bdSMark Fasheh } 944ccd979bdSMark Fasheh 945726c3342SDavid Howells static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) 946ccd979bdSMark Fasheh { 947ccd979bdSMark Fasheh struct ocfs2_super *osb; 948ccd979bdSMark Fasheh u32 numbits, freebits; 949ccd979bdSMark Fasheh int status; 950ccd979bdSMark Fasheh struct ocfs2_dinode *bm_lock; 951ccd979bdSMark Fasheh struct buffer_head *bh = NULL; 952ccd979bdSMark Fasheh struct inode *inode = NULL; 953ccd979bdSMark Fasheh 954726c3342SDavid Howells mlog_entry("(%p, %p)\n", dentry->d_sb, buf); 955ccd979bdSMark Fasheh 956726c3342SDavid Howells osb = OCFS2_SB(dentry->d_sb); 957ccd979bdSMark Fasheh 958ccd979bdSMark Fasheh inode = ocfs2_get_system_file_inode(osb, 959ccd979bdSMark Fasheh GLOBAL_BITMAP_SYSTEM_INODE, 960ccd979bdSMark Fasheh OCFS2_INVALID_SLOT); 961ccd979bdSMark Fasheh if (!inode) { 962ccd979bdSMark Fasheh mlog(ML_ERROR, "failed to get bitmap inode\n"); 963ccd979bdSMark Fasheh status = -EIO; 964ccd979bdSMark Fasheh goto bail; 965ccd979bdSMark Fasheh } 966ccd979bdSMark Fasheh 967e63aecb6SMark Fasheh status = ocfs2_inode_lock(inode, &bh, 0); 968ccd979bdSMark Fasheh if (status < 0) { 969ccd979bdSMark Fasheh mlog_errno(status); 970ccd979bdSMark Fasheh goto bail; 971ccd979bdSMark Fasheh } 972ccd979bdSMark Fasheh 973ccd979bdSMark Fasheh bm_lock = (struct ocfs2_dinode *) bh->b_data; 974ccd979bdSMark Fasheh 975ccd979bdSMark Fasheh numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); 976ccd979bdSMark Fasheh freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); 977ccd979bdSMark Fasheh 978ccd979bdSMark Fasheh buf->f_type = OCFS2_SUPER_MAGIC; 979726c3342SDavid Howells buf->f_bsize = dentry->d_sb->s_blocksize; 980ccd979bdSMark Fasheh buf->f_namelen = OCFS2_MAX_FILENAME_LEN; 981ccd979bdSMark Fasheh buf->f_blocks = ((sector_t) numbits) * 982ccd979bdSMark Fasheh (osb->s_clustersize >> osb->sb->s_blocksize_bits); 983ccd979bdSMark Fasheh buf->f_bfree = ((sector_t) freebits) * 984ccd979bdSMark Fasheh (osb->s_clustersize >> osb->sb->s_blocksize_bits); 985ccd979bdSMark Fasheh buf->f_bavail = buf->f_bfree; 986ccd979bdSMark Fasheh buf->f_files = numbits; 987ccd979bdSMark Fasheh buf->f_ffree = freebits; 988ccd979bdSMark Fasheh 989ccd979bdSMark Fasheh brelse(bh); 990ccd979bdSMark Fasheh 991e63aecb6SMark Fasheh ocfs2_inode_unlock(inode, 0); 992ccd979bdSMark Fasheh status = 0; 993ccd979bdSMark Fasheh bail: 994ccd979bdSMark Fasheh if (inode) 995ccd979bdSMark Fasheh iput(inode); 996ccd979bdSMark Fasheh 997ccd979bdSMark Fasheh mlog_exit(status); 998ccd979bdSMark Fasheh 999ccd979bdSMark Fasheh return status; 1000ccd979bdSMark Fasheh } 1001ccd979bdSMark Fasheh 10024ba9b9d0SChristoph Lameter static void ocfs2_inode_init_once(struct kmem_cache *cachep, void *data) 1003ccd979bdSMark Fasheh { 1004ccd979bdSMark Fasheh struct ocfs2_inode_info *oi = data; 1005ccd979bdSMark Fasheh 1006ccd979bdSMark Fasheh oi->ip_flags = 0; 1007ccd979bdSMark Fasheh oi->ip_open_count = 0; 1008ccd979bdSMark Fasheh spin_lock_init(&oi->ip_lock); 100983418978SMark Fasheh ocfs2_extent_map_init(&oi->vfs_inode); 1010ccd979bdSMark Fasheh INIT_LIST_HEAD(&oi->ip_io_markers); 1011ccd979bdSMark Fasheh oi->ip_created_trans = 0; 1012ccd979bdSMark Fasheh oi->ip_last_trans = 0; 1013ccd979bdSMark Fasheh oi->ip_dir_start_lookup = 0; 1014ccd979bdSMark Fasheh 1015ccd979bdSMark Fasheh init_rwsem(&oi->ip_alloc_sem); 1016251b6eccSMark Fasheh mutex_init(&oi->ip_io_mutex); 1017ccd979bdSMark Fasheh 1018ccd979bdSMark Fasheh oi->ip_blkno = 0ULL; 1019ccd979bdSMark Fasheh oi->ip_clusters = 0; 1020ccd979bdSMark Fasheh 1021ccd979bdSMark Fasheh ocfs2_lock_res_init_once(&oi->ip_rw_lockres); 1022e63aecb6SMark Fasheh ocfs2_lock_res_init_once(&oi->ip_inode_lockres); 102350008630STiger Yang ocfs2_lock_res_init_once(&oi->ip_open_lockres); 1024ccd979bdSMark Fasheh 1025ccd979bdSMark Fasheh ocfs2_metadata_cache_init(&oi->vfs_inode); 1026ccd979bdSMark Fasheh 1027ccd979bdSMark Fasheh inode_init_once(&oi->vfs_inode); 1028ccd979bdSMark Fasheh } 1029ccd979bdSMark Fasheh 1030ccd979bdSMark Fasheh static int ocfs2_initialize_mem_caches(void) 1031ccd979bdSMark Fasheh { 1032ccd979bdSMark Fasheh ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", 1033ccd979bdSMark Fasheh sizeof(struct ocfs2_inode_info), 1034fffb60f9SPaul Jackson 0, 1035fffb60f9SPaul Jackson (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| 1036fffb60f9SPaul Jackson SLAB_MEM_SPREAD), 103720c2df83SPaul Mundt ocfs2_inode_init_once); 1038ccd979bdSMark Fasheh if (!ocfs2_inode_cachep) 1039ccd979bdSMark Fasheh return -ENOMEM; 1040ccd979bdSMark Fasheh 1041ccd979bdSMark Fasheh return 0; 1042ccd979bdSMark Fasheh } 1043ccd979bdSMark Fasheh 1044ccd979bdSMark Fasheh static void ocfs2_free_mem_caches(void) 1045ccd979bdSMark Fasheh { 1046ccd979bdSMark Fasheh if (ocfs2_inode_cachep) 1047ccd979bdSMark Fasheh kmem_cache_destroy(ocfs2_inode_cachep); 1048ccd979bdSMark Fasheh 1049ccd979bdSMark Fasheh ocfs2_inode_cachep = NULL; 1050ccd979bdSMark Fasheh } 1051ccd979bdSMark Fasheh 1052ccd979bdSMark Fasheh static int ocfs2_get_sector(struct super_block *sb, 1053ccd979bdSMark Fasheh struct buffer_head **bh, 1054ccd979bdSMark Fasheh int block, 1055ccd979bdSMark Fasheh int sect_size) 1056ccd979bdSMark Fasheh { 1057ccd979bdSMark Fasheh if (!sb_set_blocksize(sb, sect_size)) { 1058ccd979bdSMark Fasheh mlog(ML_ERROR, "unable to set blocksize\n"); 1059ccd979bdSMark Fasheh return -EIO; 1060ccd979bdSMark Fasheh } 1061ccd979bdSMark Fasheh 1062ccd979bdSMark Fasheh *bh = sb_getblk(sb, block); 1063ccd979bdSMark Fasheh if (!*bh) { 1064ccd979bdSMark Fasheh mlog_errno(-EIO); 1065ccd979bdSMark Fasheh return -EIO; 1066ccd979bdSMark Fasheh } 1067ccd979bdSMark Fasheh lock_buffer(*bh); 1068ccd979bdSMark Fasheh if (!buffer_dirty(*bh)) 1069ccd979bdSMark Fasheh clear_buffer_uptodate(*bh); 1070ccd979bdSMark Fasheh unlock_buffer(*bh); 1071ccd979bdSMark Fasheh ll_rw_block(READ, 1, bh); 1072ccd979bdSMark Fasheh wait_on_buffer(*bh); 1073ccd979bdSMark Fasheh return 0; 1074ccd979bdSMark Fasheh } 1075ccd979bdSMark Fasheh 1076ccd979bdSMark Fasheh /* ocfs2 1.0 only allows one cluster and node identity per kernel image. */ 1077ccd979bdSMark Fasheh static int ocfs2_fill_local_node_info(struct ocfs2_super *osb) 1078ccd979bdSMark Fasheh { 1079ccd979bdSMark Fasheh int status; 1080ccd979bdSMark Fasheh 1081ccd979bdSMark Fasheh /* XXX hold a ref on the node while mounte? easy enough, if 1082ccd979bdSMark Fasheh * desirable. */ 1083c271c5c2SSunil Mushran if (ocfs2_mount_local(osb)) 1084c271c5c2SSunil Mushran osb->node_num = 0; 1085c271c5c2SSunil Mushran else 1086ccd979bdSMark Fasheh osb->node_num = o2nm_this_node(); 1087c271c5c2SSunil Mushran 1088ccd979bdSMark Fasheh if (osb->node_num == O2NM_MAX_NODES) { 1089ccd979bdSMark Fasheh mlog(ML_ERROR, "could not find this host's node number\n"); 1090ccd979bdSMark Fasheh status = -ENOENT; 1091ccd979bdSMark Fasheh goto bail; 1092ccd979bdSMark Fasheh } 1093ccd979bdSMark Fasheh 1094781ee3e2SSunil Mushran mlog(0, "I am node %d\n", osb->node_num); 1095ccd979bdSMark Fasheh 1096ccd979bdSMark Fasheh status = 0; 1097ccd979bdSMark Fasheh bail: 1098ccd979bdSMark Fasheh return status; 1099ccd979bdSMark Fasheh } 1100ccd979bdSMark Fasheh 1101ccd979bdSMark Fasheh static int ocfs2_mount_volume(struct super_block *sb) 1102ccd979bdSMark Fasheh { 1103ccd979bdSMark Fasheh int status = 0; 1104ccd979bdSMark Fasheh int unlock_super = 0; 1105ccd979bdSMark Fasheh struct ocfs2_super *osb = OCFS2_SB(sb); 1106ccd979bdSMark Fasheh 1107ccd979bdSMark Fasheh mlog_entry_void(); 1108ccd979bdSMark Fasheh 1109ccd979bdSMark Fasheh if (ocfs2_is_hard_readonly(osb)) 1110ccd979bdSMark Fasheh goto leave; 1111ccd979bdSMark Fasheh 1112ccd979bdSMark Fasheh status = ocfs2_fill_local_node_info(osb); 1113ccd979bdSMark Fasheh if (status < 0) { 1114ccd979bdSMark Fasheh mlog_errno(status); 1115ccd979bdSMark Fasheh goto leave; 1116ccd979bdSMark Fasheh } 1117ccd979bdSMark Fasheh 1118ccd979bdSMark Fasheh status = ocfs2_dlm_init(osb); 1119ccd979bdSMark Fasheh if (status < 0) { 1120ccd979bdSMark Fasheh mlog_errno(status); 1121ccd979bdSMark Fasheh goto leave; 1122ccd979bdSMark Fasheh } 1123ccd979bdSMark Fasheh 1124ccd979bdSMark Fasheh status = ocfs2_super_lock(osb, 1); 1125ccd979bdSMark Fasheh if (status < 0) { 1126ccd979bdSMark Fasheh mlog_errno(status); 1127ccd979bdSMark Fasheh goto leave; 1128ccd979bdSMark Fasheh } 1129ccd979bdSMark Fasheh unlock_super = 1; 1130ccd979bdSMark Fasheh 1131ccd979bdSMark Fasheh /* This will load up the node map and add ourselves to it. */ 1132ccd979bdSMark Fasheh status = ocfs2_find_slot(osb); 1133ccd979bdSMark Fasheh if (status < 0) { 1134ccd979bdSMark Fasheh mlog_errno(status); 1135ccd979bdSMark Fasheh goto leave; 1136ccd979bdSMark Fasheh } 1137ccd979bdSMark Fasheh 1138ccd979bdSMark Fasheh /* load all node-local system inodes */ 1139ccd979bdSMark Fasheh status = ocfs2_init_local_system_inodes(osb); 1140ccd979bdSMark Fasheh if (status < 0) { 1141ccd979bdSMark Fasheh mlog_errno(status); 1142ccd979bdSMark Fasheh goto leave; 1143ccd979bdSMark Fasheh } 1144ccd979bdSMark Fasheh 1145ccd979bdSMark Fasheh status = ocfs2_check_volume(osb); 1146ccd979bdSMark Fasheh if (status < 0) { 1147ccd979bdSMark Fasheh mlog_errno(status); 1148ccd979bdSMark Fasheh goto leave; 1149ccd979bdSMark Fasheh } 1150ccd979bdSMark Fasheh 1151ccd979bdSMark Fasheh status = ocfs2_truncate_log_init(osb); 1152ccd979bdSMark Fasheh if (status < 0) { 1153ccd979bdSMark Fasheh mlog_errno(status); 1154ccd979bdSMark Fasheh goto leave; 1155ccd979bdSMark Fasheh } 1156ccd979bdSMark Fasheh 1157c271c5c2SSunil Mushran if (ocfs2_mount_local(osb)) 1158c271c5c2SSunil Mushran goto leave; 1159c271c5c2SSunil Mushran 1160ccd979bdSMark Fasheh leave: 1161ccd979bdSMark Fasheh if (unlock_super) 1162ccd979bdSMark Fasheh ocfs2_super_unlock(osb, 1); 1163ccd979bdSMark Fasheh 1164ccd979bdSMark Fasheh mlog_exit(status); 1165ccd979bdSMark Fasheh return status; 1166ccd979bdSMark Fasheh } 1167ccd979bdSMark Fasheh 1168ccd979bdSMark Fasheh /* we can't grab the goofy sem lock from inside wait_event, so we use 1169ccd979bdSMark Fasheh * memory barriers to make sure that we'll see the null task before 1170ccd979bdSMark Fasheh * being woken up */ 1171ccd979bdSMark Fasheh static int ocfs2_recovery_thread_running(struct ocfs2_super *osb) 1172ccd979bdSMark Fasheh { 1173ccd979bdSMark Fasheh mb(); 1174ccd979bdSMark Fasheh return osb->recovery_thread_task != NULL; 1175ccd979bdSMark Fasheh } 1176ccd979bdSMark Fasheh 1177ccd979bdSMark Fasheh static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) 1178ccd979bdSMark Fasheh { 1179ccd979bdSMark Fasheh int tmp; 1180ccd979bdSMark Fasheh struct ocfs2_super *osb = NULL; 1181c271c5c2SSunil Mushran char nodestr[8]; 1182ccd979bdSMark Fasheh 1183ccd979bdSMark Fasheh mlog_entry("(0x%p)\n", sb); 1184ccd979bdSMark Fasheh 1185ccd979bdSMark Fasheh BUG_ON(!sb); 1186ccd979bdSMark Fasheh osb = OCFS2_SB(sb); 1187ccd979bdSMark Fasheh BUG_ON(!osb); 1188ccd979bdSMark Fasheh 1189ccd979bdSMark Fasheh ocfs2_shutdown_local_alloc(osb); 1190ccd979bdSMark Fasheh 1191ccd979bdSMark Fasheh ocfs2_truncate_log_shutdown(osb); 1192ccd979bdSMark Fasheh 1193ccd979bdSMark Fasheh /* disable any new recovery threads and wait for any currently 1194ccd979bdSMark Fasheh * running ones to exit. Do this before setting the vol_state. */ 1195c74ec2f7SArjan van de Ven mutex_lock(&osb->recovery_lock); 1196ccd979bdSMark Fasheh osb->disable_recovery = 1; 1197c74ec2f7SArjan van de Ven mutex_unlock(&osb->recovery_lock); 1198ccd979bdSMark Fasheh wait_event(osb->recovery_event, !ocfs2_recovery_thread_running(osb)); 1199ccd979bdSMark Fasheh 1200ccd979bdSMark Fasheh /* At this point, we know that no more recovery threads can be 1201ccd979bdSMark Fasheh * launched, so wait for any recovery completion work to 1202ccd979bdSMark Fasheh * complete. */ 1203ccd979bdSMark Fasheh flush_workqueue(ocfs2_wq); 1204ccd979bdSMark Fasheh 1205ccd979bdSMark Fasheh ocfs2_journal_shutdown(osb); 1206ccd979bdSMark Fasheh 1207ccd979bdSMark Fasheh ocfs2_sync_blockdev(sb); 1208ccd979bdSMark Fasheh 1209ccd979bdSMark Fasheh /* No dlm means we've failed during mount, so skip all the 1210ccd979bdSMark Fasheh * steps which depended on that to complete. */ 1211ccd979bdSMark Fasheh if (osb->dlm) { 1212ccd979bdSMark Fasheh tmp = ocfs2_super_lock(osb, 1); 1213ccd979bdSMark Fasheh if (tmp < 0) { 1214ccd979bdSMark Fasheh mlog_errno(tmp); 1215ccd979bdSMark Fasheh return; 1216ccd979bdSMark Fasheh } 121719b613d4SMark Fasheh } 1218ccd979bdSMark Fasheh 1219ccd979bdSMark Fasheh if (osb->slot_num != OCFS2_INVALID_SLOT) 1220ccd979bdSMark Fasheh ocfs2_put_slot(osb); 1221ccd979bdSMark Fasheh 122219b613d4SMark Fasheh if (osb->dlm) 1223ccd979bdSMark Fasheh ocfs2_super_unlock(osb, 1); 1224ccd979bdSMark Fasheh 1225ccd979bdSMark Fasheh ocfs2_release_system_inodes(osb); 1226ccd979bdSMark Fasheh 122734d024f8SMark Fasheh if (osb->dlm) 1228ccd979bdSMark Fasheh ocfs2_dlm_shutdown(osb); 1229ccd979bdSMark Fasheh 1230ccd979bdSMark Fasheh debugfs_remove(osb->osb_debug_root); 1231ccd979bdSMark Fasheh 1232ccd979bdSMark Fasheh if (!mnt_err) 1233ccd979bdSMark Fasheh ocfs2_stop_heartbeat(osb); 1234ccd979bdSMark Fasheh 1235ccd979bdSMark Fasheh atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); 1236ccd979bdSMark Fasheh 1237c271c5c2SSunil Mushran if (ocfs2_mount_local(osb)) 1238c271c5c2SSunil Mushran snprintf(nodestr, sizeof(nodestr), "local"); 1239c271c5c2SSunil Mushran else 1240c271c5c2SSunil Mushran snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num); 1241c271c5c2SSunil Mushran 1242c271c5c2SSunil Mushran printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", 1243c271c5c2SSunil Mushran osb->dev_str, nodestr); 1244ccd979bdSMark Fasheh 1245ccd979bdSMark Fasheh ocfs2_delete_osb(osb); 1246ccd979bdSMark Fasheh kfree(osb); 1247ccd979bdSMark Fasheh sb->s_dev = 0; 1248ccd979bdSMark Fasheh sb->s_fs_info = NULL; 1249ccd979bdSMark Fasheh } 1250ccd979bdSMark Fasheh 1251ccd979bdSMark Fasheh static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, 1252ccd979bdSMark Fasheh unsigned uuid_bytes) 1253ccd979bdSMark Fasheh { 1254ccd979bdSMark Fasheh int i, ret; 1255ccd979bdSMark Fasheh char *ptr; 1256ccd979bdSMark Fasheh 1257ccd979bdSMark Fasheh BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); 1258ccd979bdSMark Fasheh 1259cd861280SRobert P. J. Day osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); 1260ccd979bdSMark Fasheh if (osb->uuid_str == NULL) 1261ccd979bdSMark Fasheh return -ENOMEM; 1262ccd979bdSMark Fasheh 1263ccd979bdSMark Fasheh for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { 1264ccd979bdSMark Fasheh /* print with null */ 1265ccd979bdSMark Fasheh ret = snprintf(ptr, 3, "%02X", uuid[i]); 1266ccd979bdSMark Fasheh if (ret != 2) /* drop super cleans up */ 1267ccd979bdSMark Fasheh return -EINVAL; 1268ccd979bdSMark Fasheh /* then only advance past the last char */ 1269ccd979bdSMark Fasheh ptr += 2; 1270ccd979bdSMark Fasheh } 1271ccd979bdSMark Fasheh 1272ccd979bdSMark Fasheh return 0; 1273ccd979bdSMark Fasheh } 1274ccd979bdSMark Fasheh 1275ccd979bdSMark Fasheh static int ocfs2_initialize_super(struct super_block *sb, 1276ccd979bdSMark Fasheh struct buffer_head *bh, 1277ccd979bdSMark Fasheh int sector_size) 1278ccd979bdSMark Fasheh { 1279bddb8eb3SDenis Cheng int status; 12805a254031SMark Fasheh int i, cbits, bbits; 12815a254031SMark Fasheh struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; 1282ccd979bdSMark Fasheh struct inode *inode = NULL; 1283ccd979bdSMark Fasheh struct ocfs2_journal *journal; 1284ccd979bdSMark Fasheh __le32 uuid_net_key; 1285ccd979bdSMark Fasheh struct ocfs2_super *osb; 1286ccd979bdSMark Fasheh 1287ccd979bdSMark Fasheh mlog_entry_void(); 1288ccd979bdSMark Fasheh 1289cd861280SRobert P. J. Day osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); 1290ccd979bdSMark Fasheh if (!osb) { 1291ccd979bdSMark Fasheh status = -ENOMEM; 1292ccd979bdSMark Fasheh mlog_errno(status); 1293ccd979bdSMark Fasheh goto bail; 1294ccd979bdSMark Fasheh } 1295ccd979bdSMark Fasheh 1296ccd979bdSMark Fasheh sb->s_fs_info = osb; 1297ccd979bdSMark Fasheh sb->s_op = &ocfs2_sops; 1298ccd979bdSMark Fasheh sb->s_export_op = &ocfs2_export_ops; 1299e0dceaf0SMark Fasheh sb->s_time_gran = 1; 1300ccd979bdSMark Fasheh sb->s_flags |= MS_NOATIME; 1301ccd979bdSMark Fasheh /* this is needed to support O_LARGEFILE */ 13025a254031SMark Fasheh cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); 13035a254031SMark Fasheh bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); 13045a254031SMark Fasheh sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); 1305ccd979bdSMark Fasheh 1306ccd979bdSMark Fasheh osb->sb = sb; 1307ccd979bdSMark Fasheh /* Save off for ocfs2_rw_direct */ 1308ccd979bdSMark Fasheh osb->s_sectsize_bits = blksize_bits(sector_size); 1309ebdec83bSEric Sesterhenn / snakebyte BUG_ON(!osb->s_sectsize_bits); 1310ccd979bdSMark Fasheh 1311ccd979bdSMark Fasheh init_waitqueue_head(&osb->recovery_event); 131234d024f8SMark Fasheh spin_lock_init(&osb->dc_task_lock); 131334d024f8SMark Fasheh init_waitqueue_head(&osb->dc_event); 131434d024f8SMark Fasheh osb->dc_work_sequence = 0; 131534d024f8SMark Fasheh osb->dc_wake_sequence = 0; 1316ccd979bdSMark Fasheh INIT_LIST_HEAD(&osb->blocked_lock_list); 1317ccd979bdSMark Fasheh osb->blocked_lock_count = 0; 1318ccd979bdSMark Fasheh spin_lock_init(&osb->osb_lock); 1319ccd979bdSMark Fasheh 1320ccd979bdSMark Fasheh atomic_set(&osb->alloc_stats.moves, 0); 1321ccd979bdSMark Fasheh atomic_set(&osb->alloc_stats.local_data, 0); 1322ccd979bdSMark Fasheh atomic_set(&osb->alloc_stats.bitmap_data, 0); 1323ccd979bdSMark Fasheh atomic_set(&osb->alloc_stats.bg_allocs, 0); 1324ccd979bdSMark Fasheh atomic_set(&osb->alloc_stats.bg_extends, 0); 1325ccd979bdSMark Fasheh 1326ccd979bdSMark Fasheh ocfs2_init_node_maps(osb); 1327ccd979bdSMark Fasheh 1328ccd979bdSMark Fasheh snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", 1329ccd979bdSMark Fasheh MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); 1330ccd979bdSMark Fasheh 1331c74ec2f7SArjan van de Ven mutex_init(&osb->recovery_lock); 1332ccd979bdSMark Fasheh 1333ccd979bdSMark Fasheh osb->disable_recovery = 0; 1334ccd979bdSMark Fasheh osb->recovery_thread_task = NULL; 1335ccd979bdSMark Fasheh 1336ccd979bdSMark Fasheh init_waitqueue_head(&osb->checkpoint_event); 1337ccd979bdSMark Fasheh atomic_set(&osb->needs_checkpoint, 0); 1338ccd979bdSMark Fasheh 13397f1a37e3STiger Yang osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; 13407f1a37e3STiger Yang 1341ccd979bdSMark Fasheh osb->node_num = O2NM_INVALID_NODE_NUM; 1342ccd979bdSMark Fasheh osb->slot_num = OCFS2_INVALID_SLOT; 1343ccd979bdSMark Fasheh 1344ccd979bdSMark Fasheh osb->local_alloc_state = OCFS2_LA_UNUSED; 1345ccd979bdSMark Fasheh osb->local_alloc_bh = NULL; 1346ccd979bdSMark Fasheh 1347ccd979bdSMark Fasheh ocfs2_setup_hb_callbacks(osb); 1348ccd979bdSMark Fasheh 1349ccd979bdSMark Fasheh init_waitqueue_head(&osb->osb_mount_event); 1350ccd979bdSMark Fasheh 1351ccd979bdSMark Fasheh osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); 1352ccd979bdSMark Fasheh if (!osb->vol_label) { 1353ccd979bdSMark Fasheh mlog(ML_ERROR, "unable to alloc vol label\n"); 1354ccd979bdSMark Fasheh status = -ENOMEM; 1355ccd979bdSMark Fasheh goto bail; 1356ccd979bdSMark Fasheh } 1357ccd979bdSMark Fasheh 1358ccd979bdSMark Fasheh osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); 1359ccd979bdSMark Fasheh if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { 1360ccd979bdSMark Fasheh mlog(ML_ERROR, "Invalid number of node slots (%u)\n", 1361ccd979bdSMark Fasheh osb->max_slots); 1362ccd979bdSMark Fasheh status = -EINVAL; 1363ccd979bdSMark Fasheh goto bail; 1364ccd979bdSMark Fasheh } 1365781ee3e2SSunil Mushran mlog(0, "max_slots for this device: %u\n", osb->max_slots); 1366ccd979bdSMark Fasheh 1367b4df6ed8SMark Fasheh init_waitqueue_head(&osb->osb_wipe_event); 1368b4df6ed8SMark Fasheh osb->osb_orphan_wipes = kcalloc(osb->max_slots, 1369b4df6ed8SMark Fasheh sizeof(*osb->osb_orphan_wipes), 1370b4df6ed8SMark Fasheh GFP_KERNEL); 1371b4df6ed8SMark Fasheh if (!osb->osb_orphan_wipes) { 1372b4df6ed8SMark Fasheh status = -ENOMEM; 1373b4df6ed8SMark Fasheh mlog_errno(status); 1374b4df6ed8SMark Fasheh goto bail; 1375b4df6ed8SMark Fasheh } 1376b4df6ed8SMark Fasheh 1377ccd979bdSMark Fasheh osb->s_feature_compat = 1378ccd979bdSMark Fasheh le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); 1379ccd979bdSMark Fasheh osb->s_feature_ro_compat = 1380ccd979bdSMark Fasheh le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); 1381ccd979bdSMark Fasheh osb->s_feature_incompat = 1382ccd979bdSMark Fasheh le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); 1383ccd979bdSMark Fasheh 1384ccd979bdSMark Fasheh if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { 1385ccd979bdSMark Fasheh mlog(ML_ERROR, "couldn't mount because of unsupported " 1386ccd979bdSMark Fasheh "optional features (%x).\n", i); 1387ccd979bdSMark Fasheh status = -EINVAL; 1388ccd979bdSMark Fasheh goto bail; 1389ccd979bdSMark Fasheh } 1390ccd979bdSMark Fasheh if (!(osb->sb->s_flags & MS_RDONLY) && 1391ccd979bdSMark Fasheh (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { 1392ccd979bdSMark Fasheh mlog(ML_ERROR, "couldn't mount RDWR because of " 1393ccd979bdSMark Fasheh "unsupported optional features (%x).\n", i); 1394ccd979bdSMark Fasheh status = -EINVAL; 1395ccd979bdSMark Fasheh goto bail; 1396ccd979bdSMark Fasheh } 1397ccd979bdSMark Fasheh 1398ccd979bdSMark Fasheh get_random_bytes(&osb->s_next_generation, sizeof(u32)); 1399ccd979bdSMark Fasheh 1400ccd979bdSMark Fasheh /* FIXME 1401ccd979bdSMark Fasheh * This should be done in ocfs2_journal_init(), but unknown 1402ccd979bdSMark Fasheh * ordering issues will cause the filesystem to crash. 1403ccd979bdSMark Fasheh * If anyone wants to figure out what part of the code 1404ccd979bdSMark Fasheh * refers to osb->journal before ocfs2_journal_init() is run, 1405ccd979bdSMark Fasheh * be my guest. 1406ccd979bdSMark Fasheh */ 1407ccd979bdSMark Fasheh /* initialize our journal structure */ 1408ccd979bdSMark Fasheh 1409cd861280SRobert P. J. Day journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL); 1410ccd979bdSMark Fasheh if (!journal) { 1411ccd979bdSMark Fasheh mlog(ML_ERROR, "unable to alloc journal\n"); 1412ccd979bdSMark Fasheh status = -ENOMEM; 1413ccd979bdSMark Fasheh goto bail; 1414ccd979bdSMark Fasheh } 1415ccd979bdSMark Fasheh osb->journal = journal; 1416ccd979bdSMark Fasheh journal->j_osb = osb; 1417ccd979bdSMark Fasheh 1418ccd979bdSMark Fasheh atomic_set(&journal->j_num_trans, 0); 1419ccd979bdSMark Fasheh init_rwsem(&journal->j_trans_barrier); 1420ccd979bdSMark Fasheh init_waitqueue_head(&journal->j_checkpointed); 1421ccd979bdSMark Fasheh spin_lock_init(&journal->j_lock); 1422ccd979bdSMark Fasheh journal->j_trans_id = (unsigned long) 1; 1423ccd979bdSMark Fasheh INIT_LIST_HEAD(&journal->j_la_cleanups); 1424c4028958SDavid Howells INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); 1425ccd979bdSMark Fasheh journal->j_state = OCFS2_JOURNAL_FREE; 1426ccd979bdSMark Fasheh 1427ccd979bdSMark Fasheh /* get some pseudo constants for clustersize bits */ 1428ccd979bdSMark Fasheh osb->s_clustersize_bits = 1429ccd979bdSMark Fasheh le32_to_cpu(di->id2.i_super.s_clustersize_bits); 1430ccd979bdSMark Fasheh osb->s_clustersize = 1 << osb->s_clustersize_bits; 1431ccd979bdSMark Fasheh mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits); 1432ccd979bdSMark Fasheh 1433ccd979bdSMark Fasheh if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || 1434ccd979bdSMark Fasheh osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { 1435ccd979bdSMark Fasheh mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", 1436ccd979bdSMark Fasheh osb->s_clustersize); 1437ccd979bdSMark Fasheh status = -EINVAL; 1438ccd979bdSMark Fasheh goto bail; 1439ccd979bdSMark Fasheh } 1440ccd979bdSMark Fasheh 1441ccd979bdSMark Fasheh if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1) 1442ccd979bdSMark Fasheh > (u32)~0UL) { 1443ccd979bdSMark Fasheh mlog(ML_ERROR, "Volume might try to write to blocks beyond " 1444ccd979bdSMark Fasheh "what jbd can address in 32 bits.\n"); 1445ccd979bdSMark Fasheh status = -EINVAL; 1446ccd979bdSMark Fasheh goto bail; 1447ccd979bdSMark Fasheh } 1448ccd979bdSMark Fasheh 1449ccd979bdSMark Fasheh if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, 1450ccd979bdSMark Fasheh sizeof(di->id2.i_super.s_uuid))) { 1451ccd979bdSMark Fasheh mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); 1452ccd979bdSMark Fasheh status = -ENOMEM; 1453ccd979bdSMark Fasheh goto bail; 1454ccd979bdSMark Fasheh } 1455ccd979bdSMark Fasheh 145678427043SMark Fasheh memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key)); 1457ccd979bdSMark Fasheh 1458ccd979bdSMark Fasheh strncpy(osb->vol_label, di->id2.i_super.s_label, 63); 1459ccd979bdSMark Fasheh osb->vol_label[63] = '\0'; 1460ccd979bdSMark Fasheh osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); 1461ccd979bdSMark Fasheh osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); 1462ccd979bdSMark Fasheh osb->first_cluster_group_blkno = 1463ccd979bdSMark Fasheh le64_to_cpu(di->id2.i_super.s_first_cluster_group); 1464ccd979bdSMark Fasheh osb->fs_generation = le32_to_cpu(di->i_fs_generation); 1465ccd979bdSMark Fasheh mlog(0, "vol_label: %s\n", osb->vol_label); 1466ccd979bdSMark Fasheh mlog(0, "uuid: %s\n", osb->uuid_str); 1467b0697053SMark Fasheh mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n", 1468b0697053SMark Fasheh (unsigned long long)osb->root_blkno, 1469b0697053SMark Fasheh (unsigned long long)osb->system_dir_blkno); 1470ccd979bdSMark Fasheh 1471ccd979bdSMark Fasheh osb->osb_dlm_debug = ocfs2_new_dlm_debug(); 1472ccd979bdSMark Fasheh if (!osb->osb_dlm_debug) { 1473ccd979bdSMark Fasheh status = -ENOMEM; 1474ccd979bdSMark Fasheh mlog_errno(status); 1475ccd979bdSMark Fasheh goto bail; 1476ccd979bdSMark Fasheh } 1477ccd979bdSMark Fasheh 1478ccd979bdSMark Fasheh atomic_set(&osb->vol_state, VOLUME_INIT); 1479ccd979bdSMark Fasheh 1480ccd979bdSMark Fasheh /* load root, system_dir, and all global system inodes */ 1481ccd979bdSMark Fasheh status = ocfs2_init_global_system_inodes(osb); 1482ccd979bdSMark Fasheh if (status < 0) { 1483ccd979bdSMark Fasheh mlog_errno(status); 1484ccd979bdSMark Fasheh goto bail; 1485ccd979bdSMark Fasheh } 1486ccd979bdSMark Fasheh 1487ccd979bdSMark Fasheh /* 1488ccd979bdSMark Fasheh * global bitmap 1489ccd979bdSMark Fasheh */ 1490ccd979bdSMark Fasheh inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, 1491ccd979bdSMark Fasheh OCFS2_INVALID_SLOT); 1492ccd979bdSMark Fasheh if (!inode) { 1493ccd979bdSMark Fasheh status = -EINVAL; 1494ccd979bdSMark Fasheh mlog_errno(status); 1495ccd979bdSMark Fasheh goto bail; 1496ccd979bdSMark Fasheh } 1497ccd979bdSMark Fasheh 1498ccd979bdSMark Fasheh osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; 1499ccd979bdSMark Fasheh iput(inode); 1500ccd979bdSMark Fasheh 1501e9d578a8STao Ma osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8; 1502ccd979bdSMark Fasheh 1503ccd979bdSMark Fasheh status = ocfs2_init_slot_info(osb); 1504ccd979bdSMark Fasheh if (status < 0) { 1505ccd979bdSMark Fasheh mlog_errno(status); 1506ccd979bdSMark Fasheh goto bail; 1507ccd979bdSMark Fasheh } 1508ccd979bdSMark Fasheh 1509ccd979bdSMark Fasheh bail: 1510ccd979bdSMark Fasheh mlog_exit(status); 1511ccd979bdSMark Fasheh return status; 1512ccd979bdSMark Fasheh } 1513ccd979bdSMark Fasheh 1514ccd979bdSMark Fasheh /* 1515ccd979bdSMark Fasheh * will return: -EAGAIN if it is ok to keep searching for superblocks 1516ccd979bdSMark Fasheh * -EINVAL if there is a bad superblock 1517ccd979bdSMark Fasheh * 0 on success 1518ccd979bdSMark Fasheh */ 1519ccd979bdSMark Fasheh static int ocfs2_verify_volume(struct ocfs2_dinode *di, 1520ccd979bdSMark Fasheh struct buffer_head *bh, 1521ccd979bdSMark Fasheh u32 blksz) 1522ccd979bdSMark Fasheh { 1523ccd979bdSMark Fasheh int status = -EAGAIN; 1524ccd979bdSMark Fasheh 1525ccd979bdSMark Fasheh mlog_entry_void(); 1526ccd979bdSMark Fasheh 1527ccd979bdSMark Fasheh if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, 1528ccd979bdSMark Fasheh strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { 1529ccd979bdSMark Fasheh status = -EINVAL; 1530ccd979bdSMark Fasheh if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) { 1531ccd979bdSMark Fasheh mlog(ML_ERROR, "found superblock with incorrect block " 1532ccd979bdSMark Fasheh "size: found %u, should be %u\n", 1533ccd979bdSMark Fasheh 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits), 1534ccd979bdSMark Fasheh blksz); 1535ccd979bdSMark Fasheh } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != 1536ccd979bdSMark Fasheh OCFS2_MAJOR_REV_LEVEL || 1537ccd979bdSMark Fasheh le16_to_cpu(di->id2.i_super.s_minor_rev_level) != 1538ccd979bdSMark Fasheh OCFS2_MINOR_REV_LEVEL) { 1539ccd979bdSMark Fasheh mlog(ML_ERROR, "found superblock with bad version: " 1540ccd979bdSMark Fasheh "found %u.%u, should be %u.%u\n", 1541ccd979bdSMark Fasheh le16_to_cpu(di->id2.i_super.s_major_rev_level), 1542ccd979bdSMark Fasheh le16_to_cpu(di->id2.i_super.s_minor_rev_level), 1543ccd979bdSMark Fasheh OCFS2_MAJOR_REV_LEVEL, 1544ccd979bdSMark Fasheh OCFS2_MINOR_REV_LEVEL); 1545ccd979bdSMark Fasheh } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { 1546ccd979bdSMark Fasheh mlog(ML_ERROR, "bad block number on superblock: " 1547b0697053SMark Fasheh "found %llu, should be %llu\n", 15481ca1a111SMark Fasheh (unsigned long long)le64_to_cpu(di->i_blkno), 1549b0697053SMark Fasheh (unsigned long long)bh->b_blocknr); 1550ccd979bdSMark Fasheh } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || 1551ccd979bdSMark Fasheh le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { 1552ccd979bdSMark Fasheh mlog(ML_ERROR, "bad cluster size found: %u\n", 1553ccd979bdSMark Fasheh 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); 1554ccd979bdSMark Fasheh } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { 1555ccd979bdSMark Fasheh mlog(ML_ERROR, "bad root_blkno: 0\n"); 1556ccd979bdSMark Fasheh } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { 1557ccd979bdSMark Fasheh mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); 1558ccd979bdSMark Fasheh } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { 1559ccd979bdSMark Fasheh mlog(ML_ERROR, 1560ccd979bdSMark Fasheh "Superblock slots found greater than file system " 1561ccd979bdSMark Fasheh "maximum: found %u, max %u\n", 1562ccd979bdSMark Fasheh le16_to_cpu(di->id2.i_super.s_max_slots), 1563ccd979bdSMark Fasheh OCFS2_MAX_SLOTS); 1564ccd979bdSMark Fasheh } else { 1565ccd979bdSMark Fasheh /* found it! */ 1566ccd979bdSMark Fasheh status = 0; 1567ccd979bdSMark Fasheh } 1568ccd979bdSMark Fasheh } 1569ccd979bdSMark Fasheh 1570ccd979bdSMark Fasheh mlog_exit(status); 1571ccd979bdSMark Fasheh return status; 1572ccd979bdSMark Fasheh } 1573ccd979bdSMark Fasheh 1574ccd979bdSMark Fasheh static int ocfs2_check_volume(struct ocfs2_super *osb) 1575ccd979bdSMark Fasheh { 1576bddb8eb3SDenis Cheng int status; 1577ccd979bdSMark Fasheh int dirty; 1578c271c5c2SSunil Mushran int local; 1579ccd979bdSMark Fasheh struct ocfs2_dinode *local_alloc = NULL; /* only used if we 1580ccd979bdSMark Fasheh * recover 1581ccd979bdSMark Fasheh * ourselves. */ 1582ccd979bdSMark Fasheh 1583ccd979bdSMark Fasheh mlog_entry_void(); 1584ccd979bdSMark Fasheh 1585ccd979bdSMark Fasheh /* Init our journal object. */ 1586ccd979bdSMark Fasheh status = ocfs2_journal_init(osb->journal, &dirty); 1587ccd979bdSMark Fasheh if (status < 0) { 1588ccd979bdSMark Fasheh mlog(ML_ERROR, "Could not initialize journal!\n"); 1589ccd979bdSMark Fasheh goto finally; 1590ccd979bdSMark Fasheh } 1591ccd979bdSMark Fasheh 1592ccd979bdSMark Fasheh /* If the journal was unmounted cleanly then we don't want to 1593ccd979bdSMark Fasheh * recover anything. Otherwise, journal_load will do that 1594ccd979bdSMark Fasheh * dirty work for us :) */ 1595ccd979bdSMark Fasheh if (!dirty) { 1596ccd979bdSMark Fasheh status = ocfs2_journal_wipe(osb->journal, 0); 1597ccd979bdSMark Fasheh if (status < 0) { 1598ccd979bdSMark Fasheh mlog_errno(status); 1599ccd979bdSMark Fasheh goto finally; 1600ccd979bdSMark Fasheh } 1601ccd979bdSMark Fasheh } else { 1602ccd979bdSMark Fasheh mlog(ML_NOTICE, "File system was not unmounted cleanly, " 1603ccd979bdSMark Fasheh "recovering volume.\n"); 1604ccd979bdSMark Fasheh } 1605ccd979bdSMark Fasheh 1606c271c5c2SSunil Mushran local = ocfs2_mount_local(osb); 1607c271c5c2SSunil Mushran 1608ccd979bdSMark Fasheh /* will play back anything left in the journal. */ 1609c271c5c2SSunil Mushran ocfs2_journal_load(osb->journal, local); 1610ccd979bdSMark Fasheh 1611ccd979bdSMark Fasheh if (dirty) { 1612ccd979bdSMark Fasheh /* recover my local alloc if we didn't unmount cleanly. */ 1613ccd979bdSMark Fasheh status = ocfs2_begin_local_alloc_recovery(osb, 1614ccd979bdSMark Fasheh osb->slot_num, 1615ccd979bdSMark Fasheh &local_alloc); 1616ccd979bdSMark Fasheh if (status < 0) { 1617ccd979bdSMark Fasheh mlog_errno(status); 1618ccd979bdSMark Fasheh goto finally; 1619ccd979bdSMark Fasheh } 1620ccd979bdSMark Fasheh /* we complete the recovery process after we've marked 1621ccd979bdSMark Fasheh * ourselves as mounted. */ 1622ccd979bdSMark Fasheh } 1623ccd979bdSMark Fasheh 1624ccd979bdSMark Fasheh mlog(0, "Journal loaded.\n"); 1625ccd979bdSMark Fasheh 1626ccd979bdSMark Fasheh status = ocfs2_load_local_alloc(osb); 1627ccd979bdSMark Fasheh if (status < 0) { 1628ccd979bdSMark Fasheh mlog_errno(status); 1629ccd979bdSMark Fasheh goto finally; 1630ccd979bdSMark Fasheh } 1631ccd979bdSMark Fasheh 1632ccd979bdSMark Fasheh if (dirty) { 1633ccd979bdSMark Fasheh /* Recovery will be completed after we've mounted the 1634ccd979bdSMark Fasheh * rest of the volume. */ 1635ccd979bdSMark Fasheh osb->dirty = 1; 1636ccd979bdSMark Fasheh osb->local_alloc_copy = local_alloc; 1637ccd979bdSMark Fasheh local_alloc = NULL; 1638ccd979bdSMark Fasheh } 1639ccd979bdSMark Fasheh 1640ccd979bdSMark Fasheh /* go through each journal, trylock it and if you get the 1641ccd979bdSMark Fasheh * lock, and it's marked as dirty, set the bit in the recover 1642ccd979bdSMark Fasheh * map and launch a recovery thread for it. */ 1643ccd979bdSMark Fasheh status = ocfs2_mark_dead_nodes(osb); 1644ccd979bdSMark Fasheh if (status < 0) 1645ccd979bdSMark Fasheh mlog_errno(status); 1646ccd979bdSMark Fasheh 1647ccd979bdSMark Fasheh finally: 1648ccd979bdSMark Fasheh if (local_alloc) 1649ccd979bdSMark Fasheh kfree(local_alloc); 1650ccd979bdSMark Fasheh 1651ccd979bdSMark Fasheh mlog_exit(status); 1652ccd979bdSMark Fasheh return status; 1653ccd979bdSMark Fasheh } 1654ccd979bdSMark Fasheh 1655ccd979bdSMark Fasheh /* 1656ccd979bdSMark Fasheh * The routine gets called from dismount or close whenever a dismount on 1657ccd979bdSMark Fasheh * volume is requested and the osb open count becomes 1. 1658ccd979bdSMark Fasheh * It will remove the osb from the global list and also free up all the 1659ccd979bdSMark Fasheh * initialized resources and fileobject. 1660ccd979bdSMark Fasheh */ 1661ccd979bdSMark Fasheh static void ocfs2_delete_osb(struct ocfs2_super *osb) 1662ccd979bdSMark Fasheh { 1663ccd979bdSMark Fasheh mlog_entry_void(); 1664ccd979bdSMark Fasheh 1665ccd979bdSMark Fasheh /* This function assumes that the caller has the main osb resource */ 1666ccd979bdSMark Fasheh 1667ccd979bdSMark Fasheh if (osb->slot_info) 1668ccd979bdSMark Fasheh ocfs2_free_slot_info(osb->slot_info); 1669ccd979bdSMark Fasheh 1670b4df6ed8SMark Fasheh kfree(osb->osb_orphan_wipes); 1671ccd979bdSMark Fasheh /* FIXME 1672ccd979bdSMark Fasheh * This belongs in journal shutdown, but because we have to 1673ccd979bdSMark Fasheh * allocate osb->journal at the start of ocfs2_initalize_osb(), 1674ccd979bdSMark Fasheh * we free it here. 1675ccd979bdSMark Fasheh */ 1676ccd979bdSMark Fasheh kfree(osb->journal); 1677ccd979bdSMark Fasheh if (osb->local_alloc_copy) 1678ccd979bdSMark Fasheh kfree(osb->local_alloc_copy); 1679ccd979bdSMark Fasheh kfree(osb->uuid_str); 1680ccd979bdSMark Fasheh ocfs2_put_dlm_debug(osb->osb_dlm_debug); 1681ccd979bdSMark Fasheh memset(osb, 0, sizeof(struct ocfs2_super)); 1682ccd979bdSMark Fasheh 1683ccd979bdSMark Fasheh mlog_exit_void(); 1684ccd979bdSMark Fasheh } 1685ccd979bdSMark Fasheh 1686ccd979bdSMark Fasheh /* Put OCFS2 into a readonly state, or (if the user specifies it), 1687ccd979bdSMark Fasheh * panic(). We do not support continue-on-error operation. */ 1688ccd979bdSMark Fasheh static void ocfs2_handle_error(struct super_block *sb) 1689ccd979bdSMark Fasheh { 1690ccd979bdSMark Fasheh struct ocfs2_super *osb = OCFS2_SB(sb); 1691ccd979bdSMark Fasheh 1692ccd979bdSMark Fasheh if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) 1693ccd979bdSMark Fasheh panic("OCFS2: (device %s): panic forced after error\n", 1694ccd979bdSMark Fasheh sb->s_id); 1695ccd979bdSMark Fasheh 1696ccd979bdSMark Fasheh ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); 1697ccd979bdSMark Fasheh 1698ccd979bdSMark Fasheh if (sb->s_flags & MS_RDONLY && 1699ccd979bdSMark Fasheh (ocfs2_is_soft_readonly(osb) || 1700ccd979bdSMark Fasheh ocfs2_is_hard_readonly(osb))) 1701ccd979bdSMark Fasheh return; 1702ccd979bdSMark Fasheh 1703ccd979bdSMark Fasheh printk(KERN_CRIT "File system is now read-only due to the potential " 1704ccd979bdSMark Fasheh "of on-disk corruption. Please run fsck.ocfs2 once the file " 1705ccd979bdSMark Fasheh "system is unmounted.\n"); 1706ccd979bdSMark Fasheh sb->s_flags |= MS_RDONLY; 1707ccd979bdSMark Fasheh ocfs2_set_ro_flag(osb, 0); 1708ccd979bdSMark Fasheh } 1709ccd979bdSMark Fasheh 1710ccd979bdSMark Fasheh static char error_buf[1024]; 1711ccd979bdSMark Fasheh 1712ccd979bdSMark Fasheh void __ocfs2_error(struct super_block *sb, 1713ccd979bdSMark Fasheh const char *function, 1714ccd979bdSMark Fasheh const char *fmt, ...) 1715ccd979bdSMark Fasheh { 1716ccd979bdSMark Fasheh va_list args; 1717ccd979bdSMark Fasheh 1718ccd979bdSMark Fasheh va_start(args, fmt); 17194a6e617aSAlexey Dobriyan vsnprintf(error_buf, sizeof(error_buf), fmt, args); 1720ccd979bdSMark Fasheh va_end(args); 1721ccd979bdSMark Fasheh 1722ccd979bdSMark Fasheh /* Not using mlog here because we want to show the actual 1723ccd979bdSMark Fasheh * function the error came from. */ 1724ccd979bdSMark Fasheh printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n", 1725ccd979bdSMark Fasheh sb->s_id, function, error_buf); 1726ccd979bdSMark Fasheh 1727ccd979bdSMark Fasheh ocfs2_handle_error(sb); 1728ccd979bdSMark Fasheh } 1729ccd979bdSMark Fasheh 1730ccd979bdSMark Fasheh /* Handle critical errors. This is intentionally more drastic than 1731ccd979bdSMark Fasheh * ocfs2_handle_error, so we only use for things like journal errors, 1732ccd979bdSMark Fasheh * etc. */ 1733ccd979bdSMark Fasheh void __ocfs2_abort(struct super_block* sb, 1734ccd979bdSMark Fasheh const char *function, 1735ccd979bdSMark Fasheh const char *fmt, ...) 1736ccd979bdSMark Fasheh { 1737ccd979bdSMark Fasheh va_list args; 1738ccd979bdSMark Fasheh 1739ccd979bdSMark Fasheh va_start(args, fmt); 17404a6e617aSAlexey Dobriyan vsnprintf(error_buf, sizeof(error_buf), fmt, args); 1741ccd979bdSMark Fasheh va_end(args); 1742ccd979bdSMark Fasheh 1743ccd979bdSMark Fasheh printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n", 1744ccd979bdSMark Fasheh sb->s_id, function, error_buf); 1745ccd979bdSMark Fasheh 1746ccd979bdSMark Fasheh /* We don't have the cluster support yet to go straight to 1747ccd979bdSMark Fasheh * hard readonly in here. Until then, we want to keep 1748ccd979bdSMark Fasheh * ocfs2_abort() so that we can at least mark critical 1749ccd979bdSMark Fasheh * errors. 1750ccd979bdSMark Fasheh * 1751ccd979bdSMark Fasheh * TODO: This should abort the journal and alert other nodes 1752ccd979bdSMark Fasheh * that our slot needs recovery. */ 1753ccd979bdSMark Fasheh 1754ccd979bdSMark Fasheh /* Force a panic(). This stinks, but it's better than letting 1755ccd979bdSMark Fasheh * things continue without having a proper hard readonly 1756ccd979bdSMark Fasheh * here. */ 1757ccd979bdSMark Fasheh OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; 1758ccd979bdSMark Fasheh ocfs2_handle_error(sb); 1759ccd979bdSMark Fasheh } 1760ccd979bdSMark Fasheh 1761ccd979bdSMark Fasheh module_init(ocfs2_init); 1762ccd979bdSMark Fasheh module_exit(ocfs2_exit); 1763