1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * super.c
4 *
5 * load/unload driver, mount/dismount volumes
6 *
7 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
8 */
9
10 #include <linux/module.h>
11 #include <linux/fs.h>
12 #include <linux/types.h>
13 #include <linux/slab.h>
14 #include <linux/highmem.h>
15 #include <linux/init.h>
16 #include <linux/random.h>
17 #include <linux/statfs.h>
18 #include <linux/moduleparam.h>
19 #include <linux/blkdev.h>
20 #include <linux/socket.h>
21 #include <linux/inet.h>
22 #include <linux/parser.h>
23 #include <linux/crc32.h>
24 #include <linux/debugfs.h>
25 #include <linux/mount.h>
26 #include <linux/seq_file.h>
27 #include <linux/quotaops.h>
28 #include <linux/cleancache.h>
29 #include <linux/signal.h>
30
31 #define CREATE_TRACE_POINTS
32 #include "ocfs2_trace.h"
33
34 #include <cluster/masklog.h>
35
36 #include "ocfs2.h"
37
38 /* this should be the only file to include a version 1 header */
39 #include "ocfs1_fs_compat.h"
40
41 #include "alloc.h"
42 #include "aops.h"
43 #include "blockcheck.h"
44 #include "dlmglue.h"
45 #include "export.h"
46 #include "extent_map.h"
47 #include "heartbeat.h"
48 #include "inode.h"
49 #include "journal.h"
50 #include "localalloc.h"
51 #include "namei.h"
52 #include "slot_map.h"
53 #include "super.h"
54 #include "sysfile.h"
55 #include "uptodate.h"
56 #include "xattr.h"
57 #include "quota.h"
58 #include "refcounttree.h"
59 #include "suballoc.h"
60
61 #include "buffer_head_io.h"
62 #include "filecheck.h"
63
64 static struct kmem_cache *ocfs2_inode_cachep;
65 struct kmem_cache *ocfs2_dquot_cachep;
66 struct kmem_cache *ocfs2_qf_chunk_cachep;
67
68 static struct dentry *ocfs2_debugfs_root;
69
70 MODULE_AUTHOR("Oracle");
71 MODULE_LICENSE("GPL");
72 MODULE_DESCRIPTION("OCFS2 cluster file system");
73
74 struct mount_options
75 {
76 unsigned long commit_interval;
77 unsigned long mount_opt;
78 unsigned int atime_quantum;
79 unsigned short slot;
80 int localalloc_opt;
81 unsigned int resv_level;
82 int dir_resv_level;
83 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
84 };
85
86 static int ocfs2_parse_options(struct super_block *sb, char *options,
87 struct mount_options *mopt,
88 int is_remount);
89 static int ocfs2_check_set_options(struct super_block *sb,
90 struct mount_options *options);
91 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
92 static void ocfs2_put_super(struct super_block *sb);
93 static int ocfs2_mount_volume(struct super_block *sb);
94 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
95 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
96 static int ocfs2_initialize_mem_caches(void);
97 static void ocfs2_free_mem_caches(void);
98 static void ocfs2_delete_osb(struct ocfs2_super *osb);
99
100 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
101
102 static int ocfs2_sync_fs(struct super_block *sb, int wait);
103
104 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
105 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
106 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
107 static int ocfs2_check_volume(struct ocfs2_super *osb);
108 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
109 struct buffer_head *bh,
110 u32 sectsize,
111 struct ocfs2_blockcheck_stats *stats);
112 static int ocfs2_initialize_super(struct super_block *sb,
113 struct buffer_head *bh,
114 int sector_size,
115 struct ocfs2_blockcheck_stats *stats);
116 static int ocfs2_get_sector(struct super_block *sb,
117 struct buffer_head **bh,
118 int block,
119 int sect_size);
120 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
121 static void ocfs2_free_inode(struct inode *inode);
122 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
123 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
124 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
125
ocfs2_get_dquots(struct inode * inode)126 static struct dquot __rcu **ocfs2_get_dquots(struct inode *inode)
127 {
128 return OCFS2_I(inode)->i_dquot;
129 }
130
131 static const struct super_operations ocfs2_sops = {
132 .statfs = ocfs2_statfs,
133 .alloc_inode = ocfs2_alloc_inode,
134 .free_inode = ocfs2_free_inode,
135 .drop_inode = ocfs2_drop_inode,
136 .evict_inode = ocfs2_evict_inode,
137 .sync_fs = ocfs2_sync_fs,
138 .put_super = ocfs2_put_super,
139 .remount_fs = ocfs2_remount,
140 .show_options = ocfs2_show_options,
141 .quota_read = ocfs2_quota_read,
142 .quota_write = ocfs2_quota_write,
143 .get_dquots = ocfs2_get_dquots,
144 };
145
146 enum {
147 Opt_barrier,
148 Opt_err_panic,
149 Opt_err_ro,
150 Opt_intr,
151 Opt_nointr,
152 Opt_hb_none,
153 Opt_hb_local,
154 Opt_hb_global,
155 Opt_data_ordered,
156 Opt_data_writeback,
157 Opt_atime_quantum,
158 Opt_slot,
159 Opt_commit,
160 Opt_localalloc,
161 Opt_localflocks,
162 Opt_stack,
163 Opt_user_xattr,
164 Opt_nouser_xattr,
165 Opt_inode64,
166 Opt_acl,
167 Opt_noacl,
168 Opt_usrquota,
169 Opt_grpquota,
170 Opt_coherency_buffered,
171 Opt_coherency_full,
172 Opt_resv_level,
173 Opt_dir_resv_level,
174 Opt_journal_async_commit,
175 Opt_err_cont,
176 Opt_err,
177 };
178
179 static const match_table_t tokens = {
180 {Opt_barrier, "barrier=%u"},
181 {Opt_err_panic, "errors=panic"},
182 {Opt_err_ro, "errors=remount-ro"},
183 {Opt_intr, "intr"},
184 {Opt_nointr, "nointr"},
185 {Opt_hb_none, OCFS2_HB_NONE},
186 {Opt_hb_local, OCFS2_HB_LOCAL},
187 {Opt_hb_global, OCFS2_HB_GLOBAL},
188 {Opt_data_ordered, "data=ordered"},
189 {Opt_data_writeback, "data=writeback"},
190 {Opt_atime_quantum, "atime_quantum=%u"},
191 {Opt_slot, "preferred_slot=%u"},
192 {Opt_commit, "commit=%u"},
193 {Opt_localalloc, "localalloc=%d"},
194 {Opt_localflocks, "localflocks"},
195 {Opt_stack, "cluster_stack=%s"},
196 {Opt_user_xattr, "user_xattr"},
197 {Opt_nouser_xattr, "nouser_xattr"},
198 {Opt_inode64, "inode64"},
199 {Opt_acl, "acl"},
200 {Opt_noacl, "noacl"},
201 {Opt_usrquota, "usrquota"},
202 {Opt_grpquota, "grpquota"},
203 {Opt_coherency_buffered, "coherency=buffered"},
204 {Opt_coherency_full, "coherency=full"},
205 {Opt_resv_level, "resv_level=%u"},
206 {Opt_dir_resv_level, "dir_resv_level=%u"},
207 {Opt_journal_async_commit, "journal_async_commit"},
208 {Opt_err_cont, "errors=continue"},
209 {Opt_err, NULL}
210 };
211
212 #ifdef CONFIG_DEBUG_FS
ocfs2_osb_dump(struct ocfs2_super * osb,char * buf,int len)213 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
214 {
215 struct ocfs2_cluster_connection *cconn = osb->cconn;
216 struct ocfs2_recovery_map *rm = osb->recovery_map;
217 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
218 int i, out = 0;
219 unsigned long flags;
220
221 out += scnprintf(buf + out, len - out,
222 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
223 "Device", osb->dev_str, osb->uuid_str,
224 osb->fs_generation, osb->vol_label);
225
226 out += scnprintf(buf + out, len - out,
227 "%10s => State: %d Flags: 0x%lX\n", "Volume",
228 atomic_read(&osb->vol_state), osb->osb_flags);
229
230 out += scnprintf(buf + out, len - out,
231 "%10s => Block: %lu Cluster: %d\n", "Sizes",
232 osb->sb->s_blocksize, osb->s_clustersize);
233
234 out += scnprintf(buf + out, len - out,
235 "%10s => Compat: 0x%X Incompat: 0x%X "
236 "ROcompat: 0x%X\n",
237 "Features", osb->s_feature_compat,
238 osb->s_feature_incompat, osb->s_feature_ro_compat);
239
240 out += scnprintf(buf + out, len - out,
241 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
242 osb->s_mount_opt, osb->s_atime_quantum);
243
244 if (cconn) {
245 out += scnprintf(buf + out, len - out,
246 "%10s => Stack: %s Name: %*s "
247 "Version: %d.%d\n", "Cluster",
248 (*osb->osb_cluster_stack == '\0' ?
249 "o2cb" : osb->osb_cluster_stack),
250 cconn->cc_namelen, cconn->cc_name,
251 cconn->cc_version.pv_major,
252 cconn->cc_version.pv_minor);
253 }
254
255 spin_lock_irqsave(&osb->dc_task_lock, flags);
256 out += scnprintf(buf + out, len - out,
257 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
258 "WorkSeq: %lu\n", "DownCnvt",
259 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
260 osb->blocked_lock_count, osb->dc_wake_sequence,
261 osb->dc_work_sequence);
262 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
263
264 spin_lock(&osb->osb_lock);
265 out += scnprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
266 "Recovery",
267 (osb->recovery_thread_task ?
268 task_pid_nr(osb->recovery_thread_task) : -1));
269 if (rm->rm_used == 0)
270 out += scnprintf(buf + out, len - out, " None\n");
271 else {
272 for (i = 0; i < rm->rm_used; i++)
273 out += scnprintf(buf + out, len - out, " %d",
274 rm->rm_entries[i]);
275 out += scnprintf(buf + out, len - out, "\n");
276 }
277 spin_unlock(&osb->osb_lock);
278
279 out += scnprintf(buf + out, len - out,
280 "%10s => Pid: %d Interval: %lu\n", "Commit",
281 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
282 osb->osb_commit_interval);
283
284 out += scnprintf(buf + out, len - out,
285 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
286 "Journal", osb->journal->j_state,
287 osb->journal->j_trans_id,
288 atomic_read(&osb->journal->j_num_trans));
289
290 out += scnprintf(buf + out, len - out,
291 "%10s => GlobalAllocs: %d LocalAllocs: %d "
292 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
293 "Stats",
294 atomic_read(&osb->alloc_stats.bitmap_data),
295 atomic_read(&osb->alloc_stats.local_data),
296 atomic_read(&osb->alloc_stats.bg_allocs),
297 atomic_read(&osb->alloc_stats.moves),
298 atomic_read(&osb->alloc_stats.bg_extends));
299
300 out += scnprintf(buf + out, len - out,
301 "%10s => State: %u Descriptor: %llu Size: %u bits "
302 "Default: %u bits\n",
303 "LocalAlloc", osb->local_alloc_state,
304 (unsigned long long)osb->la_last_gd,
305 osb->local_alloc_bits, osb->local_alloc_default_bits);
306
307 spin_lock(&osb->osb_lock);
308 out += scnprintf(buf + out, len - out,
309 "%10s => InodeSlot: %d StolenInodes: %d, "
310 "MetaSlot: %d StolenMeta: %d\n", "Steal",
311 osb->s_inode_steal_slot,
312 atomic_read(&osb->s_num_inodes_stolen),
313 osb->s_meta_steal_slot,
314 atomic_read(&osb->s_num_meta_stolen));
315 spin_unlock(&osb->osb_lock);
316
317 out += scnprintf(buf + out, len - out, "OrphanScan => ");
318 out += scnprintf(buf + out, len - out, "Local: %u Global: %u ",
319 os->os_count, os->os_seqno);
320 out += scnprintf(buf + out, len - out, " Last Scan: ");
321 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
322 out += scnprintf(buf + out, len - out, "Disabled\n");
323 else
324 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
325 (unsigned long)(ktime_get_seconds() - os->os_scantime));
326
327 out += scnprintf(buf + out, len - out, "%10s => %3s %10s\n",
328 "Slots", "Num", "RecoGen");
329 for (i = 0; i < osb->max_slots; ++i) {
330 out += scnprintf(buf + out, len - out,
331 "%10s %c %3d %10d\n",
332 " ",
333 (i == osb->slot_num ? '*' : ' '),
334 i, osb->slot_recovery_generations[i]);
335 }
336
337 return out;
338 }
339
ocfs2_osb_debug_open(struct inode * inode,struct file * file)340 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
341 {
342 struct ocfs2_super *osb = inode->i_private;
343 char *buf = NULL;
344
345 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
346 if (!buf)
347 goto bail;
348
349 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
350
351 file->private_data = buf;
352
353 return 0;
354 bail:
355 return -ENOMEM;
356 }
357
ocfs2_debug_release(struct inode * inode,struct file * file)358 static int ocfs2_debug_release(struct inode *inode, struct file *file)
359 {
360 kfree(file->private_data);
361 return 0;
362 }
363
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)364 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
365 size_t nbytes, loff_t *ppos)
366 {
367 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
368 i_size_read(file->f_mapping->host));
369 }
370 #else
ocfs2_osb_debug_open(struct inode * inode,struct file * file)371 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
372 {
373 return 0;
374 }
ocfs2_debug_release(struct inode * inode,struct file * file)375 static int ocfs2_debug_release(struct inode *inode, struct file *file)
376 {
377 return 0;
378 }
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)379 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
380 size_t nbytes, loff_t *ppos)
381 {
382 return 0;
383 }
384 #endif /* CONFIG_DEBUG_FS */
385
386 static const struct file_operations ocfs2_osb_debug_fops = {
387 .open = ocfs2_osb_debug_open,
388 .release = ocfs2_debug_release,
389 .read = ocfs2_debug_read,
390 .llseek = generic_file_llseek,
391 };
392
ocfs2_sync_fs(struct super_block * sb,int wait)393 static int ocfs2_sync_fs(struct super_block *sb, int wait)
394 {
395 int status;
396 tid_t target;
397 struct ocfs2_super *osb = OCFS2_SB(sb);
398
399 if (ocfs2_is_hard_readonly(osb))
400 return -EROFS;
401
402 if (wait) {
403 status = ocfs2_flush_truncate_log(osb);
404 if (status < 0)
405 mlog_errno(status);
406 } else {
407 ocfs2_schedule_truncate_log_flush(osb, 0);
408 }
409
410 if (jbd2_journal_start_commit(osb->journal->j_journal,
411 &target)) {
412 if (wait)
413 jbd2_log_wait_commit(osb->journal->j_journal,
414 target);
415 }
416 return 0;
417 }
418
ocfs2_need_system_inode(struct ocfs2_super * osb,int ino)419 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
420 {
421 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
422 && (ino == USER_QUOTA_SYSTEM_INODE
423 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
424 return 0;
425 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
426 && (ino == GROUP_QUOTA_SYSTEM_INODE
427 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
428 return 0;
429 return 1;
430 }
431
ocfs2_init_global_system_inodes(struct ocfs2_super * osb)432 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
433 {
434 struct inode *new = NULL;
435 int status = 0;
436 int i;
437
438 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
439 if (IS_ERR(new)) {
440 status = PTR_ERR(new);
441 mlog_errno(status);
442 goto bail;
443 }
444 osb->root_inode = new;
445
446 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
447 if (IS_ERR(new)) {
448 status = PTR_ERR(new);
449 mlog_errno(status);
450 goto bail;
451 }
452 osb->sys_root_inode = new;
453
454 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
455 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
456 if (!ocfs2_need_system_inode(osb, i))
457 continue;
458 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
459 if (!new) {
460 ocfs2_release_system_inodes(osb);
461 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
462 mlog_errno(status);
463 mlog(ML_ERROR, "Unable to load system inode %d, "
464 "possibly corrupt fs?", i);
465 goto bail;
466 }
467 // the array now has one ref, so drop this one
468 iput(new);
469 }
470
471 bail:
472 if (status)
473 mlog_errno(status);
474 return status;
475 }
476
ocfs2_init_local_system_inodes(struct ocfs2_super * osb)477 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
478 {
479 struct inode *new = NULL;
480 int status = 0;
481 int i;
482
483 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
484 i < NUM_SYSTEM_INODES;
485 i++) {
486 if (!ocfs2_need_system_inode(osb, i))
487 continue;
488 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
489 if (!new) {
490 ocfs2_release_system_inodes(osb);
491 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
492 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
493 status, i, osb->slot_num);
494 goto bail;
495 }
496 /* the array now has one ref, so drop this one */
497 iput(new);
498 }
499
500 bail:
501 if (status)
502 mlog_errno(status);
503 return status;
504 }
505
ocfs2_release_system_inodes(struct ocfs2_super * osb)506 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
507 {
508 int i;
509 struct inode *inode;
510
511 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
512 inode = osb->global_system_inodes[i];
513 if (inode) {
514 iput(inode);
515 osb->global_system_inodes[i] = NULL;
516 }
517 }
518
519 inode = osb->sys_root_inode;
520 if (inode) {
521 iput(inode);
522 osb->sys_root_inode = NULL;
523 }
524
525 inode = osb->root_inode;
526 if (inode) {
527 iput(inode);
528 osb->root_inode = NULL;
529 }
530
531 if (!osb->local_system_inodes)
532 return;
533
534 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
535 if (osb->local_system_inodes[i]) {
536 iput(osb->local_system_inodes[i]);
537 osb->local_system_inodes[i] = NULL;
538 }
539 }
540
541 kfree(osb->local_system_inodes);
542 osb->local_system_inodes = NULL;
543 }
544
545 /* We're allocating fs objects, use GFP_NOFS */
ocfs2_alloc_inode(struct super_block * sb)546 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
547 {
548 struct ocfs2_inode_info *oi;
549
550 oi = alloc_inode_sb(sb, ocfs2_inode_cachep, GFP_NOFS);
551 if (!oi)
552 return NULL;
553
554 oi->i_sync_tid = 0;
555 oi->i_datasync_tid = 0;
556 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
557
558 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
559 return &oi->vfs_inode;
560 }
561
ocfs2_free_inode(struct inode * inode)562 static void ocfs2_free_inode(struct inode *inode)
563 {
564 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
565 }
566
ocfs2_max_file_offset(unsigned int bbits,unsigned int cbits)567 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
568 unsigned int cbits)
569 {
570 unsigned int bytes = 1 << cbits;
571 unsigned int trim = bytes;
572 unsigned int bitshift = 32;
573
574 /*
575 * i_size and all block offsets in ocfs2 are always 64 bits
576 * wide. i_clusters is 32 bits, in cluster-sized units. So on
577 * 64 bit platforms, cluster size will be the limiting factor.
578 */
579
580 #if BITS_PER_LONG == 32
581 BUILD_BUG_ON(sizeof(sector_t) != 8);
582 /*
583 * We might be limited by page cache size.
584 */
585 if (bytes > PAGE_SIZE) {
586 bytes = PAGE_SIZE;
587 trim = 1;
588 /*
589 * Shift by 31 here so that we don't get larger than
590 * MAX_LFS_FILESIZE
591 */
592 bitshift = 31;
593 }
594 #endif
595
596 /*
597 * Trim by a whole cluster when we can actually approach the
598 * on-disk limits. Otherwise we can overflow i_clusters when
599 * an extent start is at the max offset.
600 */
601 return (((unsigned long long)bytes) << bitshift) - trim;
602 }
603
ocfs2_remount(struct super_block * sb,int * flags,char * data)604 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
605 {
606 int incompat_features;
607 int ret = 0;
608 struct mount_options parsed_options;
609 struct ocfs2_super *osb = OCFS2_SB(sb);
610 u32 tmp;
611
612 sync_filesystem(sb);
613
614 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
615 !ocfs2_check_set_options(sb, &parsed_options)) {
616 ret = -EINVAL;
617 goto out;
618 }
619
620 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
621 OCFS2_MOUNT_HB_NONE;
622 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
623 ret = -EINVAL;
624 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
625 goto out;
626 }
627
628 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
629 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
630 ret = -EINVAL;
631 mlog(ML_ERROR, "Cannot change data mode on remount\n");
632 goto out;
633 }
634
635 /* Probably don't want this on remount; it might
636 * mess with other nodes */
637 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
638 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
639 ret = -EINVAL;
640 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
641 goto out;
642 }
643
644 /* We're going to/from readonly mode. */
645 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
646 /* Disable quota accounting before remounting RO */
647 if (*flags & SB_RDONLY) {
648 ret = ocfs2_susp_quotas(osb, 0);
649 if (ret < 0)
650 goto out;
651 }
652 /* Lock here so the check of HARD_RO and the potential
653 * setting of SOFT_RO is atomic. */
654 spin_lock(&osb->osb_lock);
655 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
656 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
657 ret = -EROFS;
658 goto unlock_osb;
659 }
660
661 if (*flags & SB_RDONLY) {
662 sb->s_flags |= SB_RDONLY;
663 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
664 } else {
665 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
666 mlog(ML_ERROR, "Cannot remount RDWR "
667 "filesystem due to previous errors.\n");
668 ret = -EROFS;
669 goto unlock_osb;
670 }
671 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
672 if (incompat_features) {
673 mlog(ML_ERROR, "Cannot remount RDWR because "
674 "of unsupported optional features "
675 "(%x).\n", incompat_features);
676 ret = -EINVAL;
677 goto unlock_osb;
678 }
679 sb->s_flags &= ~SB_RDONLY;
680 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
681 }
682 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
683 unlock_osb:
684 spin_unlock(&osb->osb_lock);
685 /* Enable quota accounting after remounting RW */
686 if (!ret && !(*flags & SB_RDONLY)) {
687 if (sb_any_quota_suspended(sb))
688 ret = ocfs2_susp_quotas(osb, 1);
689 else
690 ret = ocfs2_enable_quotas(osb);
691 if (ret < 0) {
692 /* Return back changes... */
693 spin_lock(&osb->osb_lock);
694 sb->s_flags |= SB_RDONLY;
695 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
696 spin_unlock(&osb->osb_lock);
697 goto out;
698 }
699 }
700 }
701
702 if (!ret) {
703 /* Only save off the new mount options in case of a successful
704 * remount. */
705 osb->s_mount_opt = parsed_options.mount_opt;
706 osb->s_atime_quantum = parsed_options.atime_quantum;
707 osb->preferred_slot = parsed_options.slot;
708 if (parsed_options.commit_interval)
709 osb->osb_commit_interval = parsed_options.commit_interval;
710
711 if (!ocfs2_is_hard_readonly(osb))
712 ocfs2_set_journal_params(osb);
713
714 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
715 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
716 SB_POSIXACL : 0);
717 }
718 out:
719 return ret;
720 }
721
ocfs2_sb_probe(struct super_block * sb,struct buffer_head ** bh,int * sector_size,struct ocfs2_blockcheck_stats * stats)722 static int ocfs2_sb_probe(struct super_block *sb,
723 struct buffer_head **bh,
724 int *sector_size,
725 struct ocfs2_blockcheck_stats *stats)
726 {
727 int status, tmpstat;
728 struct ocfs1_vol_disk_hdr *hdr;
729 struct ocfs2_dinode *di;
730 int blksize;
731
732 *bh = NULL;
733
734 /* may be > 512 */
735 *sector_size = bdev_logical_block_size(sb->s_bdev);
736 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
737 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
738 *sector_size, OCFS2_MAX_BLOCKSIZE);
739 status = -EINVAL;
740 goto bail;
741 }
742
743 /* Can this really happen? */
744 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
745 *sector_size = OCFS2_MIN_BLOCKSIZE;
746
747 /* check block zero for old format */
748 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
749 if (status < 0) {
750 mlog_errno(status);
751 goto bail;
752 }
753 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
754 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
755 mlog(ML_ERROR, "incompatible version: %u.%u\n",
756 hdr->major_version, hdr->minor_version);
757 status = -EINVAL;
758 }
759 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
760 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
761 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
762 hdr->signature);
763 status = -EINVAL;
764 }
765 brelse(*bh);
766 *bh = NULL;
767 if (status < 0) {
768 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
769 "upgraded before mounting with ocfs v2\n");
770 goto bail;
771 }
772
773 /*
774 * Now check at magic offset for 512, 1024, 2048, 4096
775 * blocksizes. 4096 is the maximum blocksize because it is
776 * the minimum clustersize.
777 */
778 status = -EINVAL;
779 for (blksize = *sector_size;
780 blksize <= OCFS2_MAX_BLOCKSIZE;
781 blksize <<= 1) {
782 tmpstat = ocfs2_get_sector(sb, bh,
783 OCFS2_SUPER_BLOCK_BLKNO,
784 blksize);
785 if (tmpstat < 0) {
786 status = tmpstat;
787 mlog_errno(status);
788 break;
789 }
790 di = (struct ocfs2_dinode *) (*bh)->b_data;
791 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
792 spin_lock_init(&stats->b_lock);
793 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
794 if (tmpstat < 0) {
795 brelse(*bh);
796 *bh = NULL;
797 }
798 if (tmpstat != -EAGAIN) {
799 status = tmpstat;
800 break;
801 }
802 }
803
804 bail:
805 return status;
806 }
807
ocfs2_verify_heartbeat(struct ocfs2_super * osb)808 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
809 {
810 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
811
812 if (osb->s_mount_opt & hb_enabled) {
813 if (ocfs2_mount_local(osb)) {
814 mlog(ML_ERROR, "Cannot heartbeat on a locally "
815 "mounted device.\n");
816 return -EINVAL;
817 }
818 if (ocfs2_userspace_stack(osb)) {
819 mlog(ML_ERROR, "Userspace stack expected, but "
820 "o2cb heartbeat arguments passed to mount\n");
821 return -EINVAL;
822 }
823 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
824 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
825 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
826 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
827 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
828 return -EINVAL;
829 }
830 }
831
832 if (!(osb->s_mount_opt & hb_enabled)) {
833 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
834 !ocfs2_userspace_stack(osb)) {
835 mlog(ML_ERROR, "Heartbeat has to be started to mount "
836 "a read-write clustered device.\n");
837 return -EINVAL;
838 }
839 }
840
841 return 0;
842 }
843
844 /*
845 * If we're using a userspace stack, mount should have passed
846 * a name that matches the disk. If not, mount should not
847 * have passed a stack.
848 */
ocfs2_verify_userspace_stack(struct ocfs2_super * osb,struct mount_options * mopt)849 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
850 struct mount_options *mopt)
851 {
852 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
853 mlog(ML_ERROR,
854 "cluster stack passed to mount, but this filesystem "
855 "does not support it\n");
856 return -EINVAL;
857 }
858
859 if (ocfs2_userspace_stack(osb) &&
860 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
861 OCFS2_STACK_LABEL_LEN)) {
862 mlog(ML_ERROR,
863 "cluster stack passed to mount (\"%s\") does not "
864 "match the filesystem (\"%s\")\n",
865 mopt->cluster_stack,
866 osb->osb_cluster_stack);
867 return -EINVAL;
868 }
869
870 return 0;
871 }
872
ocfs2_susp_quotas(struct ocfs2_super * osb,int unsuspend)873 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
874 {
875 int type;
876 struct super_block *sb = osb->sb;
877 unsigned int feature[OCFS2_MAXQUOTAS] = {
878 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
879 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
880 int status = 0;
881
882 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
883 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
884 continue;
885 if (unsuspend)
886 status = dquot_resume(sb, type);
887 else {
888 struct ocfs2_mem_dqinfo *oinfo;
889
890 /* Cancel periodic syncing before suspending */
891 oinfo = sb_dqinfo(sb, type)->dqi_priv;
892 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
893 status = dquot_suspend(sb, type);
894 }
895 if (status < 0)
896 break;
897 }
898 if (status < 0)
899 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
900 "remount (error = %d).\n", status);
901 return status;
902 }
903
ocfs2_enable_quotas(struct ocfs2_super * osb)904 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
905 {
906 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
907 struct super_block *sb = osb->sb;
908 unsigned int feature[OCFS2_MAXQUOTAS] = {
909 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
910 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
911 unsigned int ino[OCFS2_MAXQUOTAS] = {
912 LOCAL_USER_QUOTA_SYSTEM_INODE,
913 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
914 int status;
915 int type;
916
917 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
918 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
919 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
920 continue;
921 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
922 osb->slot_num);
923 if (!inode[type]) {
924 status = -ENOENT;
925 goto out_quota_off;
926 }
927 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
928 DQUOT_USAGE_ENABLED);
929 if (status < 0)
930 goto out_quota_off;
931 }
932
933 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
934 iput(inode[type]);
935 return 0;
936 out_quota_off:
937 ocfs2_disable_quotas(osb);
938 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
939 iput(inode[type]);
940 mlog_errno(status);
941 return status;
942 }
943
ocfs2_disable_quotas(struct ocfs2_super * osb)944 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
945 {
946 int type;
947 struct inode *inode;
948 struct super_block *sb = osb->sb;
949 struct ocfs2_mem_dqinfo *oinfo;
950
951 /* We mostly ignore errors in this function because there's not much
952 * we can do when we see them */
953 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
954 if (!sb_has_quota_loaded(sb, type))
955 continue;
956 if (!sb_has_quota_suspended(sb, type)) {
957 oinfo = sb_dqinfo(sb, type)->dqi_priv;
958 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
959 }
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 iput(inode);
967 }
968 }
969
ocfs2_fill_super(struct super_block * sb,void * data,int silent)970 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
971 {
972 struct dentry *root;
973 int status, sector_size;
974 struct mount_options parsed_options;
975 struct inode *inode = NULL;
976 struct ocfs2_super *osb = NULL;
977 struct buffer_head *bh = NULL;
978 char nodestr[12];
979 struct ocfs2_blockcheck_stats stats;
980
981 trace_ocfs2_fill_super(sb, data, silent);
982
983 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
984 status = -EINVAL;
985 goto out;
986 }
987
988 /* probe for superblock */
989 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
990 if (status < 0) {
991 mlog(ML_ERROR, "superblock probe failed!\n");
992 goto out;
993 }
994
995 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
996 brelse(bh);
997 bh = NULL;
998 if (status < 0)
999 goto out;
1000
1001 osb = OCFS2_SB(sb);
1002
1003 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1004 status = -EINVAL;
1005 goto out_super;
1006 }
1007 osb->s_mount_opt = parsed_options.mount_opt;
1008 osb->s_atime_quantum = parsed_options.atime_quantum;
1009 osb->preferred_slot = parsed_options.slot;
1010 osb->osb_commit_interval = parsed_options.commit_interval;
1011
1012 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1013 osb->osb_resv_level = parsed_options.resv_level;
1014 osb->osb_dir_resv_level = parsed_options.resv_level;
1015 if (parsed_options.dir_resv_level == -1)
1016 osb->osb_dir_resv_level = parsed_options.resv_level;
1017 else
1018 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1019
1020 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1021 if (status)
1022 goto out_super;
1023
1024 sb->s_magic = OCFS2_SUPER_MAGIC;
1025
1026 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1027 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1028
1029 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1030 * heartbeat=none */
1031 if (bdev_read_only(sb->s_bdev)) {
1032 if (!sb_rdonly(sb)) {
1033 status = -EACCES;
1034 mlog(ML_ERROR, "Readonly device detected but readonly "
1035 "mount was not specified.\n");
1036 goto out_super;
1037 }
1038
1039 /* You should not be able to start a local heartbeat
1040 * on a readonly device. */
1041 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1042 status = -EROFS;
1043 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1044 "device.\n");
1045 goto out_super;
1046 }
1047
1048 status = ocfs2_check_journals_nolocks(osb);
1049 if (status < 0) {
1050 if (status == -EROFS)
1051 mlog(ML_ERROR, "Recovery required on readonly "
1052 "file system, but write access is "
1053 "unavailable.\n");
1054 goto out_super;
1055 }
1056
1057 ocfs2_set_ro_flag(osb, 1);
1058
1059 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1060 "Cluster services will not be used for this mount. "
1061 "Recovery will be skipped.\n", osb->dev_str);
1062 }
1063
1064 if (!ocfs2_is_hard_readonly(osb)) {
1065 if (sb_rdonly(sb))
1066 ocfs2_set_ro_flag(osb, 0);
1067 }
1068
1069 status = ocfs2_verify_heartbeat(osb);
1070 if (status < 0)
1071 goto out_super;
1072
1073 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1074 ocfs2_debugfs_root);
1075
1076 debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1077 osb, &ocfs2_osb_debug_fops);
1078
1079 if (ocfs2_meta_ecc(osb)) {
1080 ocfs2_initialize_journal_triggers(sb, osb->s_journal_triggers);
1081 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1082 osb->osb_debug_root);
1083 }
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(s, "cluster_stack", osb->osb_cluster_stack);
1524 if (opts & OCFS2_MOUNT_USRQUOTA)
1525 seq_printf(s, ",usrquota");
1526 if (opts & OCFS2_MOUNT_GRPQUOTA)
1527 seq_printf(s, ",grpquota");
1528
1529 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1530 seq_printf(s, ",coherency=buffered");
1531 else
1532 seq_printf(s, ",coherency=full");
1533
1534 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1535 seq_printf(s, ",nouser_xattr");
1536 else
1537 seq_printf(s, ",user_xattr");
1538
1539 if (opts & OCFS2_MOUNT_INODE64)
1540 seq_printf(s, ",inode64");
1541
1542 if (opts & OCFS2_MOUNT_POSIX_ACL)
1543 seq_printf(s, ",acl");
1544 else
1545 seq_printf(s, ",noacl");
1546
1547 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1548 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1549
1550 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1551 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1552
1553 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1554 seq_printf(s, ",journal_async_commit");
1555
1556 return 0;
1557 }
1558
ocfs2_init(void)1559 static int __init ocfs2_init(void)
1560 {
1561 int status;
1562
1563 status = init_ocfs2_uptodate_cache();
1564 if (status < 0)
1565 goto out1;
1566
1567 status = ocfs2_initialize_mem_caches();
1568 if (status < 0)
1569 goto out2;
1570
1571 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1572
1573 ocfs2_set_locking_protocol();
1574
1575 register_quota_format(&ocfs2_quota_format);
1576
1577 status = register_filesystem(&ocfs2_fs_type);
1578 if (!status)
1579 return 0;
1580
1581 unregister_quota_format(&ocfs2_quota_format);
1582 debugfs_remove(ocfs2_debugfs_root);
1583 ocfs2_free_mem_caches();
1584 out2:
1585 exit_ocfs2_uptodate_cache();
1586 out1:
1587 mlog_errno(status);
1588 return status;
1589 }
1590
ocfs2_exit(void)1591 static void __exit ocfs2_exit(void)
1592 {
1593 unregister_quota_format(&ocfs2_quota_format);
1594
1595 debugfs_remove(ocfs2_debugfs_root);
1596
1597 ocfs2_free_mem_caches();
1598
1599 unregister_filesystem(&ocfs2_fs_type);
1600
1601 exit_ocfs2_uptodate_cache();
1602 }
1603
ocfs2_put_super(struct super_block * sb)1604 static void ocfs2_put_super(struct super_block *sb)
1605 {
1606 trace_ocfs2_put_super(sb);
1607
1608 ocfs2_sync_blockdev(sb);
1609 ocfs2_dismount_volume(sb, 0);
1610 }
1611
ocfs2_statfs(struct dentry * dentry,struct kstatfs * buf)1612 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1613 {
1614 struct ocfs2_super *osb;
1615 u32 numbits, freebits;
1616 int status;
1617 struct ocfs2_dinode *bm_lock;
1618 struct buffer_head *bh = NULL;
1619 struct inode *inode = NULL;
1620
1621 trace_ocfs2_statfs(dentry->d_sb, buf);
1622
1623 osb = OCFS2_SB(dentry->d_sb);
1624
1625 inode = ocfs2_get_system_file_inode(osb,
1626 GLOBAL_BITMAP_SYSTEM_INODE,
1627 OCFS2_INVALID_SLOT);
1628 if (!inode) {
1629 mlog(ML_ERROR, "failed to get bitmap inode\n");
1630 status = -EIO;
1631 goto bail;
1632 }
1633
1634 status = ocfs2_inode_lock(inode, &bh, 0);
1635 if (status < 0) {
1636 mlog_errno(status);
1637 goto bail;
1638 }
1639
1640 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1641
1642 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1643 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1644
1645 buf->f_type = OCFS2_SUPER_MAGIC;
1646 buf->f_bsize = dentry->d_sb->s_blocksize;
1647 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1648 buf->f_blocks = ((sector_t) numbits) *
1649 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1650 buf->f_bfree = ((sector_t) freebits) *
1651 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1652 buf->f_bavail = buf->f_bfree;
1653 buf->f_files = numbits;
1654 buf->f_ffree = freebits;
1655 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1656 & 0xFFFFFFFFUL;
1657 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1658 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1659
1660 brelse(bh);
1661
1662 ocfs2_inode_unlock(inode, 0);
1663 status = 0;
1664 bail:
1665 iput(inode);
1666
1667 if (status)
1668 mlog_errno(status);
1669
1670 return status;
1671 }
1672
ocfs2_inode_init_once(void * data)1673 static void ocfs2_inode_init_once(void *data)
1674 {
1675 struct ocfs2_inode_info *oi = data;
1676
1677 oi->ip_flags = 0;
1678 oi->ip_open_count = 0;
1679 spin_lock_init(&oi->ip_lock);
1680 ocfs2_extent_map_init(&oi->vfs_inode);
1681 INIT_LIST_HEAD(&oi->ip_io_markers);
1682 INIT_LIST_HEAD(&oi->ip_unwritten_list);
1683 oi->ip_dir_start_lookup = 0;
1684 init_rwsem(&oi->ip_alloc_sem);
1685 init_rwsem(&oi->ip_xattr_sem);
1686 mutex_init(&oi->ip_io_mutex);
1687
1688 oi->ip_blkno = 0ULL;
1689 oi->ip_clusters = 0;
1690 oi->ip_next_orphan = NULL;
1691
1692 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1693
1694 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1695 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1696 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1697
1698 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1699 &ocfs2_inode_caching_ops);
1700
1701 inode_init_once(&oi->vfs_inode);
1702 }
1703
ocfs2_initialize_mem_caches(void)1704 static int ocfs2_initialize_mem_caches(void)
1705 {
1706 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1707 sizeof(struct ocfs2_inode_info),
1708 0,
1709 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1710 SLAB_ACCOUNT),
1711 ocfs2_inode_init_once);
1712 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1713 sizeof(struct ocfs2_dquot),
1714 0,
1715 SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
1716 NULL);
1717 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1718 sizeof(struct ocfs2_quota_chunk),
1719 0,
1720 SLAB_RECLAIM_ACCOUNT,
1721 NULL);
1722 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1723 !ocfs2_qf_chunk_cachep) {
1724 kmem_cache_destroy(ocfs2_inode_cachep);
1725 kmem_cache_destroy(ocfs2_dquot_cachep);
1726 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1727 return -ENOMEM;
1728 }
1729
1730 return 0;
1731 }
1732
ocfs2_free_mem_caches(void)1733 static void ocfs2_free_mem_caches(void)
1734 {
1735 /*
1736 * Make sure all delayed rcu free inodes are flushed before we
1737 * destroy cache.
1738 */
1739 rcu_barrier();
1740 kmem_cache_destroy(ocfs2_inode_cachep);
1741 ocfs2_inode_cachep = NULL;
1742
1743 kmem_cache_destroy(ocfs2_dquot_cachep);
1744 ocfs2_dquot_cachep = NULL;
1745
1746 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1747 ocfs2_qf_chunk_cachep = NULL;
1748 }
1749
ocfs2_get_sector(struct super_block * sb,struct buffer_head ** bh,int block,int sect_size)1750 static int ocfs2_get_sector(struct super_block *sb,
1751 struct buffer_head **bh,
1752 int block,
1753 int sect_size)
1754 {
1755 if (!sb_set_blocksize(sb, sect_size)) {
1756 mlog(ML_ERROR, "unable to set blocksize\n");
1757 return -EIO;
1758 }
1759
1760 *bh = sb_getblk(sb, block);
1761 if (!*bh) {
1762 mlog_errno(-ENOMEM);
1763 return -ENOMEM;
1764 }
1765 lock_buffer(*bh);
1766 if (!buffer_dirty(*bh))
1767 clear_buffer_uptodate(*bh);
1768 unlock_buffer(*bh);
1769 if (bh_read(*bh, 0) < 0) {
1770 mlog_errno(-EIO);
1771 brelse(*bh);
1772 *bh = NULL;
1773 return -EIO;
1774 }
1775
1776 return 0;
1777 }
1778
ocfs2_mount_volume(struct super_block * sb)1779 static int ocfs2_mount_volume(struct super_block *sb)
1780 {
1781 int status = 0;
1782 struct ocfs2_super *osb = OCFS2_SB(sb);
1783
1784 if (ocfs2_is_hard_readonly(osb))
1785 goto out;
1786
1787 mutex_init(&osb->obs_trim_fs_mutex);
1788
1789 status = ocfs2_dlm_init(osb);
1790 if (status < 0) {
1791 mlog_errno(status);
1792 if (status == -EBADR && ocfs2_userspace_stack(osb))
1793 mlog(ML_ERROR, "couldn't mount because cluster name on"
1794 " disk does not match the running cluster name.\n");
1795 goto out;
1796 }
1797
1798 status = ocfs2_super_lock(osb, 1);
1799 if (status < 0) {
1800 mlog_errno(status);
1801 goto out_dlm;
1802 }
1803
1804 /* This will load up the node map and add ourselves to it. */
1805 status = ocfs2_find_slot(osb);
1806 if (status < 0) {
1807 mlog_errno(status);
1808 goto out_super_lock;
1809 }
1810
1811 /* load all node-local system inodes */
1812 status = ocfs2_init_local_system_inodes(osb);
1813 if (status < 0) {
1814 mlog_errno(status);
1815 goto out_super_lock;
1816 }
1817
1818 status = ocfs2_check_volume(osb);
1819 if (status < 0) {
1820 mlog_errno(status);
1821 goto out_system_inodes;
1822 }
1823
1824 status = ocfs2_truncate_log_init(osb);
1825 if (status < 0) {
1826 mlog_errno(status);
1827 goto out_check_volume;
1828 }
1829
1830 ocfs2_super_unlock(osb, 1);
1831 return 0;
1832
1833 out_check_volume:
1834 ocfs2_free_replay_slots(osb);
1835 out_system_inodes:
1836 if (osb->local_alloc_state == OCFS2_LA_ENABLED)
1837 ocfs2_shutdown_local_alloc(osb);
1838 ocfs2_release_system_inodes(osb);
1839 /* before journal shutdown, we should release slot_info */
1840 ocfs2_free_slot_info(osb);
1841 ocfs2_journal_shutdown(osb);
1842 out_super_lock:
1843 ocfs2_super_unlock(osb, 1);
1844 out_dlm:
1845 ocfs2_dlm_shutdown(osb, 0);
1846 out:
1847 return status;
1848 }
1849
ocfs2_dismount_volume(struct super_block * sb,int mnt_err)1850 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1851 {
1852 int tmp, hangup_needed = 0;
1853 struct ocfs2_super *osb = NULL;
1854 char nodestr[12];
1855
1856 trace_ocfs2_dismount_volume(sb);
1857
1858 BUG_ON(!sb);
1859 osb = OCFS2_SB(sb);
1860 BUG_ON(!osb);
1861
1862 /* Remove file check sysfs related directores/files,
1863 * and wait for the pending file check operations */
1864 ocfs2_filecheck_remove_sysfs(osb);
1865
1866 kset_unregister(osb->osb_dev_kset);
1867
1868 /* Orphan scan should be stopped as early as possible */
1869 ocfs2_orphan_scan_stop(osb);
1870
1871 /* Stop quota recovery so that we can disable quotas */
1872 ocfs2_recovery_disable_quota(osb);
1873
1874 ocfs2_disable_quotas(osb);
1875
1876 /* All dquots should be freed by now */
1877 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1878 /* Wait for worker to be done with the work structure in osb */
1879 cancel_work_sync(&osb->dquot_drop_work);
1880
1881 ocfs2_shutdown_local_alloc(osb);
1882
1883 ocfs2_truncate_log_shutdown(osb);
1884
1885 /* This will disable recovery and flush any recovery work. */
1886 ocfs2_recovery_exit(osb);
1887
1888 ocfs2_sync_blockdev(sb);
1889
1890 ocfs2_purge_refcount_trees(osb);
1891
1892 /* No cluster connection means we've failed during mount, so skip
1893 * all the steps which depended on that to complete. */
1894 if (osb->cconn) {
1895 tmp = ocfs2_super_lock(osb, 1);
1896 if (tmp < 0) {
1897 mlog_errno(tmp);
1898 return;
1899 }
1900 }
1901
1902 if (osb->slot_num != OCFS2_INVALID_SLOT)
1903 ocfs2_put_slot(osb);
1904
1905 if (osb->cconn)
1906 ocfs2_super_unlock(osb, 1);
1907
1908 ocfs2_release_system_inodes(osb);
1909
1910 ocfs2_journal_shutdown(osb);
1911
1912 /*
1913 * If we're dismounting due to mount error, mount.ocfs2 will clean
1914 * up heartbeat. If we're a local mount, there is no heartbeat.
1915 * If we failed before we got a uuid_str yet, we can't stop
1916 * heartbeat. Otherwise, do it.
1917 */
1918 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1919 !ocfs2_is_hard_readonly(osb))
1920 hangup_needed = 1;
1921
1922 ocfs2_dlm_shutdown(osb, hangup_needed);
1923
1924 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1925 debugfs_remove_recursive(osb->osb_debug_root);
1926
1927 if (hangup_needed)
1928 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1929
1930 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1931
1932 if (ocfs2_mount_local(osb))
1933 snprintf(nodestr, sizeof(nodestr), "local");
1934 else
1935 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1936
1937 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1938 osb->dev_str, nodestr);
1939
1940 ocfs2_delete_osb(osb);
1941 kfree(osb);
1942 sb->s_dev = 0;
1943 sb->s_fs_info = NULL;
1944 }
1945
ocfs2_setup_osb_uuid(struct ocfs2_super * osb,const unsigned char * uuid,unsigned uuid_bytes)1946 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1947 unsigned uuid_bytes)
1948 {
1949 int i, ret;
1950 char *ptr;
1951
1952 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1953
1954 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1955 if (osb->uuid_str == NULL)
1956 return -ENOMEM;
1957
1958 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1959 /* print with null */
1960 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1961 if (ret != 2) /* drop super cleans up */
1962 return -EINVAL;
1963 /* then only advance past the last char */
1964 ptr += 2;
1965 }
1966
1967 return 0;
1968 }
1969
1970 /* Make sure entire volume is addressable by our journal. Requires
1971 osb_clusters_at_boot to be valid and for the journal to have been
1972 initialized by ocfs2_journal_init(). */
ocfs2_journal_addressable(struct ocfs2_super * osb)1973 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1974 {
1975 int status = 0;
1976 u64 max_block =
1977 ocfs2_clusters_to_blocks(osb->sb,
1978 osb->osb_clusters_at_boot) - 1;
1979
1980 /* 32-bit block number is always OK. */
1981 if (max_block <= (u32)~0ULL)
1982 goto out;
1983
1984 /* Volume is "huge", so see if our journal is new enough to
1985 support it. */
1986 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
1987 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
1988 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
1989 JBD2_FEATURE_INCOMPAT_64BIT))) {
1990 mlog(ML_ERROR, "The journal cannot address the entire volume. "
1991 "Enable the 'block64' journal option with tunefs.ocfs2");
1992 status = -EFBIG;
1993 goto out;
1994 }
1995
1996 out:
1997 return status;
1998 }
1999
ocfs2_initialize_super(struct super_block * sb,struct buffer_head * bh,int sector_size,struct ocfs2_blockcheck_stats * stats)2000 static int ocfs2_initialize_super(struct super_block *sb,
2001 struct buffer_head *bh,
2002 int sector_size,
2003 struct ocfs2_blockcheck_stats *stats)
2004 {
2005 int status;
2006 int i, cbits, bbits;
2007 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2008 struct inode *inode = NULL;
2009 struct ocfs2_super *osb;
2010 u64 total_blocks;
2011
2012 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2013 if (!osb) {
2014 status = -ENOMEM;
2015 mlog_errno(status);
2016 goto out;
2017 }
2018
2019 sb->s_fs_info = osb;
2020 sb->s_op = &ocfs2_sops;
2021 sb->s_d_op = &ocfs2_dentry_ops;
2022 sb->s_export_op = &ocfs2_export_ops;
2023 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2024 sb->dq_op = &ocfs2_quota_operations;
2025 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2026 sb->s_xattr = ocfs2_xattr_handlers;
2027 sb->s_time_gran = 1;
2028 sb->s_flags |= SB_NOATIME;
2029 /* this is needed to support O_LARGEFILE */
2030 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2031 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2032 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2033 super_set_uuid(sb, di->id2.i_super.s_uuid,
2034 sizeof(di->id2.i_super.s_uuid));
2035
2036 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2037
2038 for (i = 0; i < 3; i++)
2039 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2040 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2041
2042 osb->sb = sb;
2043 osb->s_sectsize_bits = blksize_bits(sector_size);
2044 BUG_ON(!osb->s_sectsize_bits);
2045
2046 spin_lock_init(&osb->dc_task_lock);
2047 init_waitqueue_head(&osb->dc_event);
2048 osb->dc_work_sequence = 0;
2049 osb->dc_wake_sequence = 0;
2050 INIT_LIST_HEAD(&osb->blocked_lock_list);
2051 osb->blocked_lock_count = 0;
2052 spin_lock_init(&osb->osb_lock);
2053 spin_lock_init(&osb->osb_xattr_lock);
2054 ocfs2_init_steal_slots(osb);
2055
2056 mutex_init(&osb->system_file_mutex);
2057
2058 atomic_set(&osb->alloc_stats.moves, 0);
2059 atomic_set(&osb->alloc_stats.local_data, 0);
2060 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2061 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2062 atomic_set(&osb->alloc_stats.bg_extends, 0);
2063
2064 /* Copy the blockcheck stats from the superblock probe */
2065 osb->osb_ecc_stats = *stats;
2066
2067 ocfs2_init_node_maps(osb);
2068
2069 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2070 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2071
2072 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2073 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2074 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2075 osb->max_slots);
2076 status = -EINVAL;
2077 goto out;
2078 }
2079
2080 ocfs2_orphan_scan_init(osb);
2081
2082 status = ocfs2_recovery_init(osb);
2083 if (status) {
2084 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2085 mlog_errno(status);
2086 goto out;
2087 }
2088
2089 init_waitqueue_head(&osb->checkpoint_event);
2090
2091 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2092
2093 osb->slot_num = OCFS2_INVALID_SLOT;
2094
2095 osb->s_xattr_inline_size = le16_to_cpu(
2096 di->id2.i_super.s_xattr_inline_size);
2097
2098 osb->local_alloc_state = OCFS2_LA_UNUSED;
2099 osb->local_alloc_bh = NULL;
2100 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2101
2102 init_waitqueue_head(&osb->osb_mount_event);
2103
2104 ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2105
2106 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2107 if (!osb->vol_label) {
2108 mlog(ML_ERROR, "unable to alloc vol label\n");
2109 status = -ENOMEM;
2110 goto out_recovery_map;
2111 }
2112
2113 osb->slot_recovery_generations =
2114 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2115 GFP_KERNEL);
2116 if (!osb->slot_recovery_generations) {
2117 status = -ENOMEM;
2118 mlog_errno(status);
2119 goto out_vol_label;
2120 }
2121
2122 init_waitqueue_head(&osb->osb_wipe_event);
2123 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2124 sizeof(*osb->osb_orphan_wipes),
2125 GFP_KERNEL);
2126 if (!osb->osb_orphan_wipes) {
2127 status = -ENOMEM;
2128 mlog_errno(status);
2129 goto out_slot_recovery_gen;
2130 }
2131
2132 osb->osb_rf_lock_tree = RB_ROOT;
2133
2134 osb->s_feature_compat =
2135 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2136 osb->s_feature_ro_compat =
2137 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2138 osb->s_feature_incompat =
2139 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2140
2141 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2142 mlog(ML_ERROR, "couldn't mount because of unsupported "
2143 "optional features (%x).\n", i);
2144 status = -EINVAL;
2145 goto out_orphan_wipes;
2146 }
2147 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2148 mlog(ML_ERROR, "couldn't mount RDWR because of "
2149 "unsupported optional features (%x).\n", i);
2150 status = -EINVAL;
2151 goto out_orphan_wipes;
2152 }
2153
2154 if (ocfs2_clusterinfo_valid(osb)) {
2155 /*
2156 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2157 * terminated, so make sure no overflow happens here by using
2158 * memcpy. Destination strings will always be null terminated
2159 * because osb is allocated using kzalloc.
2160 */
2161 osb->osb_stackflags =
2162 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2163 memcpy(osb->osb_cluster_stack,
2164 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2165 OCFS2_STACK_LABEL_LEN);
2166 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2167 mlog(ML_ERROR,
2168 "couldn't mount because of an invalid "
2169 "cluster stack label (%s) \n",
2170 osb->osb_cluster_stack);
2171 status = -EINVAL;
2172 goto out_orphan_wipes;
2173 }
2174 memcpy(osb->osb_cluster_name,
2175 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2176 OCFS2_CLUSTER_NAME_LEN);
2177 } else {
2178 /* The empty string is identical with classic tools that
2179 * don't know about s_cluster_info. */
2180 osb->osb_cluster_stack[0] = '\0';
2181 }
2182
2183 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2184
2185 /*
2186 * FIXME
2187 * This should be done in ocfs2_journal_init(), but any inode
2188 * writes back operation will cause the filesystem to crash.
2189 */
2190 status = ocfs2_journal_alloc(osb);
2191 if (status < 0)
2192 goto out_orphan_wipes;
2193
2194 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2195 init_llist_head(&osb->dquot_drop_list);
2196
2197 /* get some pseudo constants for clustersize bits */
2198 osb->s_clustersize_bits =
2199 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2200 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2201
2202 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2203 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2204 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2205 osb->s_clustersize);
2206 status = -EINVAL;
2207 goto out_journal;
2208 }
2209
2210 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2211 le32_to_cpu(di->i_clusters));
2212
2213 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2214 total_blocks);
2215 if (status) {
2216 mlog(ML_ERROR, "Volume too large "
2217 "to mount safely on this system");
2218 status = -EFBIG;
2219 goto out_journal;
2220 }
2221
2222 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2223 sizeof(di->id2.i_super.s_uuid))) {
2224 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2225 status = -ENOMEM;
2226 goto out_journal;
2227 }
2228
2229 strscpy(osb->vol_label, di->id2.i_super.s_label,
2230 OCFS2_MAX_VOL_LABEL_LEN);
2231 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2232 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2233 osb->first_cluster_group_blkno =
2234 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2235 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2236 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2237 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2238 (unsigned long long)osb->root_blkno,
2239 (unsigned long long)osb->system_dir_blkno,
2240 osb->s_clustersize_bits);
2241
2242 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2243 if (!osb->osb_dlm_debug) {
2244 status = -ENOMEM;
2245 mlog_errno(status);
2246 goto out_uuid_str;
2247 }
2248
2249 atomic_set(&osb->vol_state, VOLUME_INIT);
2250
2251 /* load root, system_dir, and all global system inodes */
2252 status = ocfs2_init_global_system_inodes(osb);
2253 if (status < 0) {
2254 mlog_errno(status);
2255 goto out_dlm_out;
2256 }
2257
2258 /*
2259 * global bitmap
2260 */
2261 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2262 OCFS2_INVALID_SLOT);
2263 if (!inode) {
2264 status = -EINVAL;
2265 mlog_errno(status);
2266 goto out_system_inodes;
2267 }
2268
2269 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2270 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2271 iput(inode);
2272
2273 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2274 osb->s_feature_incompat) * 8;
2275
2276 status = ocfs2_init_slot_info(osb);
2277 if (status < 0) {
2278 mlog_errno(status);
2279 goto out_system_inodes;
2280 }
2281 cleancache_init_shared_fs(sb);
2282
2283 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2284 if (!osb->ocfs2_wq) {
2285 status = -ENOMEM;
2286 mlog_errno(status);
2287 goto out_slot_info;
2288 }
2289
2290 return status;
2291
2292 out_slot_info:
2293 ocfs2_free_slot_info(osb);
2294 out_system_inodes:
2295 ocfs2_release_system_inodes(osb);
2296 out_dlm_out:
2297 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2298 out_uuid_str:
2299 kfree(osb->uuid_str);
2300 out_journal:
2301 kfree(osb->journal);
2302 out_orphan_wipes:
2303 kfree(osb->osb_orphan_wipes);
2304 out_slot_recovery_gen:
2305 kfree(osb->slot_recovery_generations);
2306 out_vol_label:
2307 kfree(osb->vol_label);
2308 out_recovery_map:
2309 kfree(osb->recovery_map);
2310 out:
2311 kfree(osb);
2312 sb->s_fs_info = NULL;
2313 return status;
2314 }
2315
2316 /*
2317 * will return: -EAGAIN if it is ok to keep searching for superblocks
2318 * -EINVAL if there is a bad superblock
2319 * 0 on success
2320 */
ocfs2_verify_volume(struct ocfs2_dinode * di,struct buffer_head * bh,u32 blksz,struct ocfs2_blockcheck_stats * stats)2321 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2322 struct buffer_head *bh,
2323 u32 blksz,
2324 struct ocfs2_blockcheck_stats *stats)
2325 {
2326 int status = -EAGAIN;
2327 u32 blksz_bits;
2328
2329 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2330 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2331 /* We have to do a raw check of the feature here */
2332 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2333 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2334 status = ocfs2_block_check_validate(bh->b_data,
2335 bh->b_size,
2336 &di->i_check,
2337 stats);
2338 if (status)
2339 goto out;
2340 }
2341 status = -EINVAL;
2342 /* Acceptable block sizes are 512 bytes, 1K, 2K and 4K. */
2343 blksz_bits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2344 if (blksz_bits < 9 || blksz_bits > 12) {
2345 mlog(ML_ERROR, "found superblock with incorrect block "
2346 "size bits: found %u, should be 9, 10, 11, or 12\n",
2347 blksz_bits);
2348 } else if ((1 << blksz_bits) != blksz) {
2349 mlog(ML_ERROR, "found superblock with incorrect block "
2350 "size: found %u, should be %u\n", 1 << blksz_bits, blksz);
2351 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2352 OCFS2_MAJOR_REV_LEVEL ||
2353 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2354 OCFS2_MINOR_REV_LEVEL) {
2355 mlog(ML_ERROR, "found superblock with bad version: "
2356 "found %u.%u, should be %u.%u\n",
2357 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2358 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2359 OCFS2_MAJOR_REV_LEVEL,
2360 OCFS2_MINOR_REV_LEVEL);
2361 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2362 mlog(ML_ERROR, "bad block number on superblock: "
2363 "found %llu, should be %llu\n",
2364 (unsigned long long)le64_to_cpu(di->i_blkno),
2365 (unsigned long long)bh->b_blocknr);
2366 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2367 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2368 mlog(ML_ERROR, "bad cluster size bit found: %u\n",
2369 le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2370 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2371 mlog(ML_ERROR, "bad root_blkno: 0\n");
2372 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2373 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2374 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2375 mlog(ML_ERROR,
2376 "Superblock slots found greater than file system "
2377 "maximum: found %u, max %u\n",
2378 le16_to_cpu(di->id2.i_super.s_max_slots),
2379 OCFS2_MAX_SLOTS);
2380 } else {
2381 /* found it! */
2382 status = 0;
2383 }
2384 }
2385
2386 out:
2387 if (status && status != -EAGAIN)
2388 mlog_errno(status);
2389 return status;
2390 }
2391
ocfs2_check_volume(struct ocfs2_super * osb)2392 static int ocfs2_check_volume(struct ocfs2_super *osb)
2393 {
2394 int status;
2395 int dirty;
2396 int local;
2397 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2398 * recover
2399 * ourselves. */
2400
2401 /* Init our journal object. */
2402 status = ocfs2_journal_init(osb, &dirty);
2403 if (status < 0) {
2404 mlog(ML_ERROR, "Could not initialize journal!\n");
2405 goto finally;
2406 }
2407
2408 /* Now that journal has been initialized, check to make sure
2409 entire volume is addressable. */
2410 status = ocfs2_journal_addressable(osb);
2411 if (status)
2412 goto finally;
2413
2414 /* If the journal was unmounted cleanly then we don't want to
2415 * recover anything. Otherwise, journal_load will do that
2416 * dirty work for us :) */
2417 if (!dirty) {
2418 status = ocfs2_journal_wipe(osb->journal, 0);
2419 if (status < 0) {
2420 mlog_errno(status);
2421 goto finally;
2422 }
2423 } else {
2424 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2425 "unmounted cleanly, recovering it.\n", osb->dev_str);
2426 }
2427
2428 local = ocfs2_mount_local(osb);
2429
2430 /* will play back anything left in the journal. */
2431 status = ocfs2_journal_load(osb->journal, local, dirty);
2432 if (status < 0) {
2433 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2434 goto finally;
2435 }
2436
2437 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2438 jbd2_journal_set_features(osb->journal->j_journal,
2439 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2440 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2441 else
2442 jbd2_journal_clear_features(osb->journal->j_journal,
2443 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2444 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2445
2446 if (dirty) {
2447 /* recover my local alloc if we didn't unmount cleanly. */
2448 status = ocfs2_begin_local_alloc_recovery(osb,
2449 osb->slot_num,
2450 &local_alloc);
2451 if (status < 0) {
2452 mlog_errno(status);
2453 goto finally;
2454 }
2455 /* we complete the recovery process after we've marked
2456 * ourselves as mounted. */
2457 }
2458
2459 status = ocfs2_load_local_alloc(osb);
2460 if (status < 0) {
2461 mlog_errno(status);
2462 goto finally;
2463 }
2464
2465 if (dirty) {
2466 /* Recovery will be completed after we've mounted the
2467 * rest of the volume. */
2468 osb->local_alloc_copy = local_alloc;
2469 local_alloc = NULL;
2470 }
2471
2472 /* go through each journal, trylock it and if you get the
2473 * lock, and it's marked as dirty, set the bit in the recover
2474 * map and launch a recovery thread for it. */
2475 status = ocfs2_mark_dead_nodes(osb);
2476 if (status < 0) {
2477 mlog_errno(status);
2478 goto finally;
2479 }
2480
2481 status = ocfs2_compute_replay_slots(osb);
2482 if (status < 0)
2483 mlog_errno(status);
2484
2485 finally:
2486 kfree(local_alloc);
2487
2488 if (status)
2489 mlog_errno(status);
2490 return status;
2491 }
2492
2493 /*
2494 * The routine gets called from dismount or close whenever a dismount on
2495 * volume is requested and the osb open count becomes 1.
2496 * It will remove the osb from the global list and also free up all the
2497 * initialized resources and fileobject.
2498 */
ocfs2_delete_osb(struct ocfs2_super * osb)2499 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2500 {
2501 /* This function assumes that the caller has the main osb resource */
2502
2503 /* ocfs2_initializer_super have already created this workqueue */
2504 if (osb->ocfs2_wq)
2505 destroy_workqueue(osb->ocfs2_wq);
2506
2507 ocfs2_free_slot_info(osb);
2508
2509 kfree(osb->osb_orphan_wipes);
2510 kfree(osb->slot_recovery_generations);
2511 /* FIXME
2512 * This belongs in journal shutdown, but because we have to
2513 * allocate osb->journal at the middle of ocfs2_initialize_super(),
2514 * we free it here.
2515 */
2516 kfree(osb->journal);
2517 kfree(osb->local_alloc_copy);
2518 kfree(osb->uuid_str);
2519 kfree(osb->vol_label);
2520 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2521 memset(osb, 0, sizeof(struct ocfs2_super));
2522 }
2523
2524 /* Depending on the mount option passed, perform one of the following:
2525 * Put OCFS2 into a readonly state (default)
2526 * Return EIO so that only the process errs
2527 * Fix the error as if fsck.ocfs2 -y
2528 * panic
2529 */
ocfs2_handle_error(struct super_block * sb)2530 static int ocfs2_handle_error(struct super_block *sb)
2531 {
2532 struct ocfs2_super *osb = OCFS2_SB(sb);
2533 int rv = 0;
2534
2535 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2536 pr_crit("On-disk corruption discovered. "
2537 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2538
2539 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2540 panic("OCFS2: (device %s): panic forced after error\n",
2541 sb->s_id);
2542 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2543 pr_crit("OCFS2: Returning error to the calling process.\n");
2544 rv = -EIO;
2545 } else { /* default option */
2546 rv = -EROFS;
2547 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2548 return rv;
2549
2550 pr_crit("OCFS2: File system is now read-only.\n");
2551 sb->s_flags |= SB_RDONLY;
2552 ocfs2_set_ro_flag(osb, 0);
2553 }
2554
2555 return rv;
2556 }
2557
__ocfs2_error(struct super_block * sb,const char * function,const char * fmt,...)2558 int __ocfs2_error(struct super_block *sb, const char *function,
2559 const char *fmt, ...)
2560 {
2561 struct va_format vaf;
2562 va_list args;
2563
2564 va_start(args, fmt);
2565 vaf.fmt = fmt;
2566 vaf.va = &args;
2567
2568 /* Not using mlog here because we want to show the actual
2569 * function the error came from. */
2570 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2571 sb->s_id, function, &vaf);
2572
2573 va_end(args);
2574
2575 return ocfs2_handle_error(sb);
2576 }
2577
2578 /* Handle critical errors. This is intentionally more drastic than
2579 * ocfs2_handle_error, so we only use for things like journal errors,
2580 * etc. */
__ocfs2_abort(struct super_block * sb,const char * function,const char * fmt,...)2581 void __ocfs2_abort(struct super_block *sb, const char *function,
2582 const char *fmt, ...)
2583 {
2584 struct va_format vaf;
2585 va_list args;
2586
2587 va_start(args, fmt);
2588
2589 vaf.fmt = fmt;
2590 vaf.va = &args;
2591
2592 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2593 sb->s_id, function, &vaf);
2594
2595 va_end(args);
2596
2597 /* We don't have the cluster support yet to go straight to
2598 * hard readonly in here. Until then, we want to keep
2599 * ocfs2_abort() so that we can at least mark critical
2600 * errors.
2601 *
2602 * TODO: This should abort the journal and alert other nodes
2603 * that our slot needs recovery. */
2604
2605 /* Force a panic(). This stinks, but it's better than letting
2606 * things continue without having a proper hard readonly
2607 * here. */
2608 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2609 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2610 ocfs2_handle_error(sb);
2611 }
2612
2613 /*
2614 * Void signal blockers, because in-kernel sigprocmask() only fails
2615 * when SIG_* is wrong.
2616 */
ocfs2_block_signals(sigset_t * oldset)2617 void ocfs2_block_signals(sigset_t *oldset)
2618 {
2619 int rc;
2620 sigset_t blocked;
2621
2622 sigfillset(&blocked);
2623 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2624 BUG_ON(rc);
2625 }
2626
ocfs2_unblock_signals(sigset_t * oldset)2627 void ocfs2_unblock_signals(sigset_t *oldset)
2628 {
2629 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2630 BUG_ON(rc);
2631 }
2632
2633 module_init(ocfs2_init);
2634 module_exit(ocfs2_exit);
2635