ext3 can fail badly when device stops accepting BIO_RW_BARRIER requests
[linux-2.6/pdupreez.git] / fs / ocfs2 / resize.c
blob37835ffcb03985f7bd61a1322538d96ca8f8830d
1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
4 * resize.c
6 * volume resize.
7 * Inspired by ext3/resize.c.
9 * Copyright (C) 2007 Oracle. All rights reserved.
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public
13 * License as published by the Free Software Foundation; either
14 * version 2 of the License, or (at your option) any later version.
16 * This program is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * General Public License for more details.
21 * You should have received a copy of the GNU General Public
22 * License along with this program; if not, write to the
23 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
24 * Boston, MA 021110-1307, USA.
27 #include <linux/fs.h>
28 #include <linux/types.h>
30 #define MLOG_MASK_PREFIX ML_DISK_ALLOC
31 #include <cluster/masklog.h>
33 #include "ocfs2.h"
35 #include "alloc.h"
36 #include "dlmglue.h"
37 #include "inode.h"
38 #include "journal.h"
39 #include "super.h"
40 #include "sysfile.h"
41 #include "uptodate.h"
43 #include "buffer_head_io.h"
44 #include "suballoc.h"
45 #include "resize.h"
48 * Check whether there are new backup superblocks exist
49 * in the last group. If there are some, mark them or clear
50 * them in the bitmap.
52 * Return how many backups we find in the last group.
54 static u16 ocfs2_calc_new_backup_super(struct inode *inode,
55 struct ocfs2_group_desc *gd,
56 int new_clusters,
57 u32 first_new_cluster,
58 u16 cl_cpg,
59 int set)
61 int i;
62 u16 backups = 0;
63 u32 cluster;
64 u64 blkno, gd_blkno, lgd_blkno = le64_to_cpu(gd->bg_blkno);
66 for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
67 blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
68 cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
70 gd_blkno = ocfs2_which_cluster_group(inode, cluster);
71 if (gd_blkno < lgd_blkno)
72 continue;
73 else if (gd_blkno > lgd_blkno)
74 break;
76 if (set)
77 ocfs2_set_bit(cluster % cl_cpg,
78 (unsigned long *)gd->bg_bitmap);
79 else
80 ocfs2_clear_bit(cluster % cl_cpg,
81 (unsigned long *)gd->bg_bitmap);
82 backups++;
85 mlog_exit_void();
86 return backups;
89 static int ocfs2_update_last_group_and_inode(handle_t *handle,
90 struct inode *bm_inode,
91 struct buffer_head *bm_bh,
92 struct buffer_head *group_bh,
93 u32 first_new_cluster,
94 int new_clusters)
96 int ret = 0;
97 struct ocfs2_super *osb = OCFS2_SB(bm_inode->i_sb);
98 struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bm_bh->b_data;
99 struct ocfs2_chain_list *cl = &fe->id2.i_chain;
100 struct ocfs2_chain_rec *cr;
101 struct ocfs2_group_desc *group;
102 u16 chain, num_bits, backups = 0;
103 u16 cl_bpc = le16_to_cpu(cl->cl_bpc);
104 u16 cl_cpg = le16_to_cpu(cl->cl_cpg);
106 mlog_entry("(new_clusters=%d, first_new_cluster = %u)\n",
107 new_clusters, first_new_cluster);
109 ret = ocfs2_journal_access(handle, bm_inode, group_bh,
110 OCFS2_JOURNAL_ACCESS_WRITE);
111 if (ret < 0) {
112 mlog_errno(ret);
113 goto out;
116 group = (struct ocfs2_group_desc *)group_bh->b_data;
118 /* update the group first. */
119 num_bits = new_clusters * cl_bpc;
120 le16_add_cpu(&group->bg_bits, num_bits);
121 le16_add_cpu(&group->bg_free_bits_count, num_bits);
124 * check whether there are some new backup superblocks exist in
125 * this group and update the group bitmap accordingly.
127 if (OCFS2_HAS_COMPAT_FEATURE(osb->sb,
128 OCFS2_FEATURE_COMPAT_BACKUP_SB)) {
129 backups = ocfs2_calc_new_backup_super(bm_inode,
130 group,
131 new_clusters,
132 first_new_cluster,
133 cl_cpg, 1);
134 le16_add_cpu(&group->bg_free_bits_count, -1 * backups);
137 ret = ocfs2_journal_dirty(handle, group_bh);
138 if (ret < 0) {
139 mlog_errno(ret);
140 goto out_rollback;
143 /* update the inode accordingly. */
144 ret = ocfs2_journal_access(handle, bm_inode, bm_bh,
145 OCFS2_JOURNAL_ACCESS_WRITE);
146 if (ret < 0) {
147 mlog_errno(ret);
148 goto out_rollback;
151 chain = le16_to_cpu(group->bg_chain);
152 cr = (&cl->cl_recs[chain]);
153 le32_add_cpu(&cr->c_total, num_bits);
154 le32_add_cpu(&cr->c_free, num_bits);
155 le32_add_cpu(&fe->id1.bitmap1.i_total, num_bits);
156 le32_add_cpu(&fe->i_clusters, new_clusters);
158 if (backups) {
159 le32_add_cpu(&cr->c_free, -1 * backups);
160 le32_add_cpu(&fe->id1.bitmap1.i_used, backups);
163 spin_lock(&OCFS2_I(bm_inode)->ip_lock);
164 OCFS2_I(bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
165 le64_add_cpu(&fe->i_size, new_clusters << osb->s_clustersize_bits);
166 spin_unlock(&OCFS2_I(bm_inode)->ip_lock);
167 i_size_write(bm_inode, le64_to_cpu(fe->i_size));
169 ocfs2_journal_dirty(handle, bm_bh);
171 out_rollback:
172 if (ret < 0) {
173 ocfs2_calc_new_backup_super(bm_inode,
174 group,
175 new_clusters,
176 first_new_cluster,
177 cl_cpg, 0);
178 le16_add_cpu(&group->bg_free_bits_count, backups);
179 le16_add_cpu(&group->bg_bits, -1 * num_bits);
180 le16_add_cpu(&group->bg_free_bits_count, -1 * num_bits);
182 out:
183 mlog_exit(ret);
184 return ret;
187 static int update_backups(struct inode * inode, u32 clusters, char *data)
189 int i, ret = 0;
190 u32 cluster;
191 u64 blkno;
192 struct buffer_head *backup = NULL;
193 struct ocfs2_dinode *backup_di = NULL;
194 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
196 /* calculate the real backups we need to update. */
197 for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
198 blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
199 cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
200 if (cluster > clusters)
201 break;
203 ret = ocfs2_read_block(osb, blkno, &backup, 0, NULL);
204 if (ret < 0) {
205 mlog_errno(ret);
206 break;
209 memcpy(backup->b_data, data, inode->i_sb->s_blocksize);
211 backup_di = (struct ocfs2_dinode *)backup->b_data;
212 backup_di->i_blkno = cpu_to_le64(blkno);
214 ret = ocfs2_write_super_or_backup(osb, backup);
215 brelse(backup);
216 backup = NULL;
217 if (ret < 0) {
218 mlog_errno(ret);
219 break;
223 return ret;
226 static void ocfs2_update_super_and_backups(struct inode *inode,
227 int new_clusters)
229 int ret;
230 u32 clusters = 0;
231 struct buffer_head *super_bh = NULL;
232 struct ocfs2_dinode *super_di = NULL;
233 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
236 * update the superblock last.
237 * It doesn't matter if the write failed.
239 ret = ocfs2_read_block(osb, OCFS2_SUPER_BLOCK_BLKNO,
240 &super_bh, 0, NULL);
241 if (ret < 0) {
242 mlog_errno(ret);
243 goto out;
246 super_di = (struct ocfs2_dinode *)super_bh->b_data;
247 le32_add_cpu(&super_di->i_clusters, new_clusters);
248 clusters = le32_to_cpu(super_di->i_clusters);
250 ret = ocfs2_write_super_or_backup(osb, super_bh);
251 if (ret < 0) {
252 mlog_errno(ret);
253 goto out;
256 if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_COMPAT_BACKUP_SB))
257 ret = update_backups(inode, clusters, super_bh->b_data);
259 out:
260 brelse(super_bh);
261 if (ret)
262 printk(KERN_WARNING "ocfs2: Failed to update super blocks on %s"
263 " during fs resize. This condition is not fatal,"
264 " but fsck.ocfs2 should be run to fix it\n",
265 osb->dev_str);
266 return;
270 * Extend the filesystem to the new number of clusters specified. This entry
271 * point is only used to extend the current filesystem to the end of the last
272 * existing group.
274 int ocfs2_group_extend(struct inode * inode, int new_clusters)
276 int ret;
277 handle_t *handle;
278 struct buffer_head *main_bm_bh = NULL;
279 struct buffer_head *group_bh = NULL;
280 struct inode *main_bm_inode = NULL;
281 struct ocfs2_dinode *fe = NULL;
282 struct ocfs2_group_desc *group = NULL;
283 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
284 u16 cl_bpc;
285 u32 first_new_cluster;
286 u64 lgd_blkno;
288 mlog_entry_void();
290 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
291 return -EROFS;
293 if (new_clusters < 0)
294 return -EINVAL;
295 else if (new_clusters == 0)
296 return 0;
298 main_bm_inode = ocfs2_get_system_file_inode(osb,
299 GLOBAL_BITMAP_SYSTEM_INODE,
300 OCFS2_INVALID_SLOT);
301 if (!main_bm_inode) {
302 ret = -EINVAL;
303 mlog_errno(ret);
304 goto out;
307 mutex_lock(&main_bm_inode->i_mutex);
309 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
310 if (ret < 0) {
311 mlog_errno(ret);
312 goto out_mutex;
315 fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
317 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
318 ocfs2_group_bitmap_size(osb->sb) * 8) {
319 mlog(ML_ERROR, "The disk is too old and small. "
320 "Force to do offline resize.");
321 ret = -EINVAL;
322 goto out_unlock;
325 if (!OCFS2_IS_VALID_DINODE(fe)) {
326 OCFS2_RO_ON_INVALID_DINODE(main_bm_inode->i_sb, fe);
327 ret = -EIO;
328 goto out_unlock;
331 first_new_cluster = le32_to_cpu(fe->i_clusters);
332 lgd_blkno = ocfs2_which_cluster_group(main_bm_inode,
333 first_new_cluster - 1);
335 ret = ocfs2_read_block(osb, lgd_blkno, &group_bh, OCFS2_BH_CACHED,
336 main_bm_inode);
337 if (ret < 0) {
338 mlog_errno(ret);
339 goto out_unlock;
342 group = (struct ocfs2_group_desc *)group_bh->b_data;
344 ret = ocfs2_check_group_descriptor(inode->i_sb, fe, group);
345 if (ret) {
346 mlog_errno(ret);
347 goto out_unlock;
350 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
351 if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters >
352 le16_to_cpu(fe->id2.i_chain.cl_cpg)) {
353 ret = -EINVAL;
354 goto out_unlock;
357 mlog(0, "extend the last group at %llu, new clusters = %d\n",
358 (unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters);
360 handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
361 if (IS_ERR(handle)) {
362 mlog_errno(PTR_ERR(handle));
363 ret = -EINVAL;
364 goto out_unlock;
367 /* update the last group descriptor and inode. */
368 ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode,
369 main_bm_bh, group_bh,
370 first_new_cluster,
371 new_clusters);
372 if (ret) {
373 mlog_errno(ret);
374 goto out_commit;
377 ocfs2_update_super_and_backups(main_bm_inode, new_clusters);
379 out_commit:
380 ocfs2_commit_trans(osb, handle);
381 out_unlock:
382 brelse(group_bh);
383 brelse(main_bm_bh);
385 ocfs2_inode_unlock(main_bm_inode, 1);
387 out_mutex:
388 mutex_unlock(&main_bm_inode->i_mutex);
389 iput(main_bm_inode);
391 out:
392 mlog_exit_void();
393 return ret;
396 static int ocfs2_check_new_group(struct inode *inode,
397 struct ocfs2_dinode *di,
398 struct ocfs2_new_group_input *input,
399 struct buffer_head *group_bh)
401 int ret;
402 struct ocfs2_group_desc *gd;
403 u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc);
404 unsigned int max_bits = le16_to_cpu(di->id2.i_chain.cl_cpg) *
405 le16_to_cpu(di->id2.i_chain.cl_bpc);
408 gd = (struct ocfs2_group_desc *)group_bh->b_data;
410 ret = -EIO;
411 if (!OCFS2_IS_VALID_GROUP_DESC(gd))
412 mlog(ML_ERROR, "Group descriptor # %llu isn't valid.\n",
413 (unsigned long long)le64_to_cpu(gd->bg_blkno));
414 else if (di->i_blkno != gd->bg_parent_dinode)
415 mlog(ML_ERROR, "Group descriptor # %llu has bad parent "
416 "pointer (%llu, expected %llu)\n",
417 (unsigned long long)le64_to_cpu(gd->bg_blkno),
418 (unsigned long long)le64_to_cpu(gd->bg_parent_dinode),
419 (unsigned long long)le64_to_cpu(di->i_blkno));
420 else if (le16_to_cpu(gd->bg_bits) > max_bits)
421 mlog(ML_ERROR, "Group descriptor # %llu has bit count of %u\n",
422 (unsigned long long)le64_to_cpu(gd->bg_blkno),
423 le16_to_cpu(gd->bg_bits));
424 else if (le16_to_cpu(gd->bg_free_bits_count) > le16_to_cpu(gd->bg_bits))
425 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
426 "claims that %u are free\n",
427 (unsigned long long)le64_to_cpu(gd->bg_blkno),
428 le16_to_cpu(gd->bg_bits),
429 le16_to_cpu(gd->bg_free_bits_count));
430 else if (le16_to_cpu(gd->bg_bits) > (8 * le16_to_cpu(gd->bg_size)))
431 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
432 "max bitmap bits of %u\n",
433 (unsigned long long)le64_to_cpu(gd->bg_blkno),
434 le16_to_cpu(gd->bg_bits),
435 8 * le16_to_cpu(gd->bg_size));
436 else if (le16_to_cpu(gd->bg_chain) != input->chain)
437 mlog(ML_ERROR, "Group descriptor # %llu has bad chain %u "
438 "while input has %u set.\n",
439 (unsigned long long)le64_to_cpu(gd->bg_blkno),
440 le16_to_cpu(gd->bg_chain), input->chain);
441 else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc)
442 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
443 "input has %u clusters set\n",
444 (unsigned long long)le64_to_cpu(gd->bg_blkno),
445 le16_to_cpu(gd->bg_bits), input->clusters);
446 else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc)
447 mlog(ML_ERROR, "Group descriptor # %llu has free bit count %u "
448 "but it should have %u set\n",
449 (unsigned long long)le64_to_cpu(gd->bg_blkno),
450 le16_to_cpu(gd->bg_bits),
451 input->frees * cl_bpc);
452 else
453 ret = 0;
455 return ret;
458 static int ocfs2_verify_group_and_input(struct inode *inode,
459 struct ocfs2_dinode *di,
460 struct ocfs2_new_group_input *input,
461 struct buffer_head *group_bh)
463 u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count);
464 u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
465 u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec);
466 u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group);
467 u32 total_clusters = le32_to_cpu(di->i_clusters);
468 int ret = -EINVAL;
470 if (cluster < total_clusters)
471 mlog(ML_ERROR, "add a group which is in the current volume.\n");
472 else if (input->chain >= cl_count)
473 mlog(ML_ERROR, "input chain exceeds the limit.\n");
474 else if (next_free != cl_count && next_free != input->chain)
475 mlog(ML_ERROR,
476 "the add group should be in chain %u\n", next_free);
477 else if (total_clusters + input->clusters < total_clusters)
478 mlog(ML_ERROR, "add group's clusters overflow.\n");
479 else if (input->clusters > cl_cpg)
480 mlog(ML_ERROR, "the cluster exceeds the maximum of a group\n");
481 else if (input->frees > input->clusters)
482 mlog(ML_ERROR, "the free cluster exceeds the total clusters\n");
483 else if (total_clusters % cl_cpg != 0)
484 mlog(ML_ERROR,
485 "the last group isn't full. Use group extend first.\n");
486 else if (input->group != ocfs2_which_cluster_group(inode, cluster))
487 mlog(ML_ERROR, "group blkno is invalid\n");
488 else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh)))
489 mlog(ML_ERROR, "group descriptor check failed.\n");
490 else
491 ret = 0;
493 return ret;
496 /* Add a new group descriptor to global_bitmap. */
497 int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
499 int ret;
500 handle_t *handle;
501 struct buffer_head *main_bm_bh = NULL;
502 struct inode *main_bm_inode = NULL;
503 struct ocfs2_dinode *fe = NULL;
504 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
505 struct buffer_head *group_bh = NULL;
506 struct ocfs2_group_desc *group = NULL;
507 struct ocfs2_chain_list *cl;
508 struct ocfs2_chain_rec *cr;
509 u16 cl_bpc;
511 mlog_entry_void();
513 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
514 return -EROFS;
516 main_bm_inode = ocfs2_get_system_file_inode(osb,
517 GLOBAL_BITMAP_SYSTEM_INODE,
518 OCFS2_INVALID_SLOT);
519 if (!main_bm_inode) {
520 ret = -EINVAL;
521 mlog_errno(ret);
522 goto out;
525 mutex_lock(&main_bm_inode->i_mutex);
527 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
528 if (ret < 0) {
529 mlog_errno(ret);
530 goto out_mutex;
533 fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
535 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
536 ocfs2_group_bitmap_size(osb->sb) * 8) {
537 mlog(ML_ERROR, "The disk is too old and small."
538 " Force to do offline resize.");
539 ret = -EINVAL;
540 goto out_unlock;
543 ret = ocfs2_read_block(osb, input->group, &group_bh, 0, NULL);
544 if (ret < 0) {
545 mlog(ML_ERROR, "Can't read the group descriptor # %llu "
546 "from the device.", (unsigned long long)input->group);
547 goto out_unlock;
550 ocfs2_set_new_buffer_uptodate(inode, group_bh);
552 ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh);
553 if (ret) {
554 mlog_errno(ret);
555 goto out_unlock;
558 mlog(0, "Add a new group %llu in chain = %u, length = %u\n",
559 (unsigned long long)input->group, input->chain, input->clusters);
561 handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS);
562 if (IS_ERR(handle)) {
563 mlog_errno(PTR_ERR(handle));
564 ret = -EINVAL;
565 goto out_unlock;
568 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
569 cl = &fe->id2.i_chain;
570 cr = &cl->cl_recs[input->chain];
572 ret = ocfs2_journal_access(handle, main_bm_inode, group_bh,
573 OCFS2_JOURNAL_ACCESS_WRITE);
574 if (ret < 0) {
575 mlog_errno(ret);
576 goto out_commit;
579 group = (struct ocfs2_group_desc *)group_bh->b_data;
580 group->bg_next_group = cr->c_blkno;
582 ret = ocfs2_journal_dirty(handle, group_bh);
583 if (ret < 0) {
584 mlog_errno(ret);
585 goto out_commit;
588 ret = ocfs2_journal_access(handle, main_bm_inode, main_bm_bh,
589 OCFS2_JOURNAL_ACCESS_WRITE);
590 if (ret < 0) {
591 mlog_errno(ret);
592 goto out_commit;
595 if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) {
596 le16_add_cpu(&cl->cl_next_free_rec, 1);
597 memset(cr, 0, sizeof(struct ocfs2_chain_rec));
600 cr->c_blkno = le64_to_cpu(input->group);
601 le32_add_cpu(&cr->c_total, input->clusters * cl_bpc);
602 le32_add_cpu(&cr->c_free, input->frees * cl_bpc);
604 le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc);
605 le32_add_cpu(&fe->id1.bitmap1.i_used,
606 (input->clusters - input->frees) * cl_bpc);
607 le32_add_cpu(&fe->i_clusters, input->clusters);
609 ocfs2_journal_dirty(handle, main_bm_bh);
611 spin_lock(&OCFS2_I(main_bm_inode)->ip_lock);
612 OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
613 le64_add_cpu(&fe->i_size, input->clusters << osb->s_clustersize_bits);
614 spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock);
615 i_size_write(main_bm_inode, le64_to_cpu(fe->i_size));
617 ocfs2_update_super_and_backups(main_bm_inode, input->clusters);
619 out_commit:
620 ocfs2_commit_trans(osb, handle);
621 out_unlock:
622 brelse(group_bh);
623 brelse(main_bm_bh);
625 ocfs2_inode_unlock(main_bm_inode, 1);
627 out_mutex:
628 mutex_unlock(&main_bm_inode->i_mutex);
629 iput(main_bm_inode);
631 out:
632 mlog_exit_void();
633 return ret;