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