1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
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.
28 #include <linux/types.h>
30 #define MLOG_MASK_PREFIX ML_DISK_ALLOC
31 #include <cluster/masklog.h>
43 #include "buffer_head_io.h"
48 * Check whether there are new backup superblocks exist
49 * in the last group. If there are some, mark them or clear
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,
57 u32 first_new_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)
73 else if (gd_blkno > lgd_blkno)
77 ocfs2_set_bit(cluster % cl_cpg,
78 (unsigned long *)gd->bg_bitmap);
80 ocfs2_clear_bit(cluster % cl_cpg,
81 (unsigned long *)gd->bg_bitmap);
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,
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);
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,
134 le16_add_cpu(&group->bg_free_bits_count, -1 * backups);
137 ret = ocfs2_journal_dirty(handle, group_bh);
143 /* update the inode accordingly. */
144 ret = ocfs2_journal_access(handle, bm_inode, bm_bh,
145 OCFS2_JOURNAL_ACCESS_WRITE);
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);
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);
173 ocfs2_calc_new_backup_super(bm_inode,
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);
187 static int update_backups(struct inode * inode, u32 clusters, char *data)
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)
203 ret = ocfs2_read_block(osb, blkno, &backup, 0, NULL);
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);
226 static void ocfs2_update_super_and_backups(struct inode *inode,
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,
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);
256 if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_COMPAT_BACKUP_SB))
257 ret = update_backups(inode, clusters, super_bh->b_data);
263 printk(KERN_WARNING "ocfs2: Failed to update super blocks on %s"
264 " during fs resize. This condition is not fatal,"
265 " but fsck.ocfs2 should be run to fix it\n",
271 * Extend the filesystem to the new number of clusters specified. This entry
272 * point is only used to extend the current filesystem to the end of the last
275 int ocfs2_group_extend(struct inode * inode, int new_clusters)
279 struct buffer_head *main_bm_bh = NULL;
280 struct buffer_head *group_bh = NULL;
281 struct inode *main_bm_inode = NULL;
282 struct ocfs2_dinode *fe = NULL;
283 struct ocfs2_group_desc *group = NULL;
284 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
286 u32 first_new_cluster;
291 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
294 if (new_clusters < 0)
296 else if (new_clusters == 0)
299 main_bm_inode = ocfs2_get_system_file_inode(osb,
300 GLOBAL_BITMAP_SYSTEM_INODE,
302 if (!main_bm_inode) {
308 mutex_lock(&main_bm_inode->i_mutex);
310 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
316 fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
318 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
319 ocfs2_group_bitmap_size(osb->sb) * 8) {
320 mlog(ML_ERROR, "The disk is too old and small. "
321 "Force to do offline resize.");
326 if (!OCFS2_IS_VALID_DINODE(fe)) {
327 OCFS2_RO_ON_INVALID_DINODE(main_bm_inode->i_sb, fe);
332 first_new_cluster = le32_to_cpu(fe->i_clusters);
333 lgd_blkno = ocfs2_which_cluster_group(main_bm_inode,
334 first_new_cluster - 1);
336 ret = ocfs2_read_block(osb, lgd_blkno, &group_bh, OCFS2_BH_CACHED,
343 group = (struct ocfs2_group_desc *)group_bh->b_data;
345 ret = ocfs2_check_group_descriptor(inode->i_sb, fe, group);
351 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
352 if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters >
353 le16_to_cpu(fe->id2.i_chain.cl_cpg)) {
358 mlog(0, "extend the last group at %llu, new clusters = %d\n",
359 (unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters);
361 handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
362 if (IS_ERR(handle)) {
363 mlog_errno(PTR_ERR(handle));
368 /* update the last group descriptor and inode. */
369 ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode,
370 main_bm_bh, group_bh,
378 ocfs2_update_super_and_backups(main_bm_inode, new_clusters);
381 ocfs2_commit_trans(osb, handle);
389 ocfs2_inode_unlock(main_bm_inode, 1);
392 mutex_unlock(&main_bm_inode->i_mutex);
400 static int ocfs2_check_new_group(struct inode *inode,
401 struct ocfs2_dinode *di,
402 struct ocfs2_new_group_input *input,
403 struct buffer_head *group_bh)
406 struct ocfs2_group_desc *gd;
407 u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc);
408 unsigned int max_bits = le16_to_cpu(di->id2.i_chain.cl_cpg) *
409 le16_to_cpu(di->id2.i_chain.cl_bpc);
412 gd = (struct ocfs2_group_desc *)group_bh->b_data;
415 if (!OCFS2_IS_VALID_GROUP_DESC(gd))
416 mlog(ML_ERROR, "Group descriptor # %llu isn't valid.\n",
417 (unsigned long long)le64_to_cpu(gd->bg_blkno));
418 else if (di->i_blkno != gd->bg_parent_dinode)
419 mlog(ML_ERROR, "Group descriptor # %llu has bad parent "
420 "pointer (%llu, expected %llu)\n",
421 (unsigned long long)le64_to_cpu(gd->bg_blkno),
422 (unsigned long long)le64_to_cpu(gd->bg_parent_dinode),
423 (unsigned long long)le64_to_cpu(di->i_blkno));
424 else if (le16_to_cpu(gd->bg_bits) > max_bits)
425 mlog(ML_ERROR, "Group descriptor # %llu has bit count of %u\n",
426 (unsigned long long)le64_to_cpu(gd->bg_blkno),
427 le16_to_cpu(gd->bg_bits));
428 else if (le16_to_cpu(gd->bg_free_bits_count) > le16_to_cpu(gd->bg_bits))
429 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
430 "claims that %u are free\n",
431 (unsigned long long)le64_to_cpu(gd->bg_blkno),
432 le16_to_cpu(gd->bg_bits),
433 le16_to_cpu(gd->bg_free_bits_count));
434 else if (le16_to_cpu(gd->bg_bits) > (8 * le16_to_cpu(gd->bg_size)))
435 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
436 "max bitmap bits of %u\n",
437 (unsigned long long)le64_to_cpu(gd->bg_blkno),
438 le16_to_cpu(gd->bg_bits),
439 8 * le16_to_cpu(gd->bg_size));
440 else if (le16_to_cpu(gd->bg_chain) != input->chain)
441 mlog(ML_ERROR, "Group descriptor # %llu has bad chain %u "
442 "while input has %u set.\n",
443 (unsigned long long)le64_to_cpu(gd->bg_blkno),
444 le16_to_cpu(gd->bg_chain), input->chain);
445 else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc)
446 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
447 "input has %u clusters set\n",
448 (unsigned long long)le64_to_cpu(gd->bg_blkno),
449 le16_to_cpu(gd->bg_bits), input->clusters);
450 else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc)
451 mlog(ML_ERROR, "Group descriptor # %llu has free bit count %u "
452 "but it should have %u set\n",
453 (unsigned long long)le64_to_cpu(gd->bg_blkno),
454 le16_to_cpu(gd->bg_bits),
455 input->frees * cl_bpc);
462 static int ocfs2_verify_group_and_input(struct inode *inode,
463 struct ocfs2_dinode *di,
464 struct ocfs2_new_group_input *input,
465 struct buffer_head *group_bh)
467 u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count);
468 u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
469 u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec);
470 u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group);
471 u32 total_clusters = le32_to_cpu(di->i_clusters);
474 if (cluster < total_clusters)
475 mlog(ML_ERROR, "add a group which is in the current volume.\n");
476 else if (input->chain >= cl_count)
477 mlog(ML_ERROR, "input chain exceeds the limit.\n");
478 else if (next_free != cl_count && next_free != input->chain)
480 "the add group should be in chain %u\n", next_free);
481 else if (total_clusters + input->clusters < total_clusters)
482 mlog(ML_ERROR, "add group's clusters overflow.\n");
483 else if (input->clusters > cl_cpg)
484 mlog(ML_ERROR, "the cluster exceeds the maximum of a group\n");
485 else if (input->frees > input->clusters)
486 mlog(ML_ERROR, "the free cluster exceeds the total clusters\n");
487 else if (total_clusters % cl_cpg != 0)
489 "the last group isn't full. Use group extend first.\n");
490 else if (input->group != ocfs2_which_cluster_group(inode, cluster))
491 mlog(ML_ERROR, "group blkno is invalid\n");
492 else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh)))
493 mlog(ML_ERROR, "group descriptor check failed.\n");
500 /* Add a new group descriptor to global_bitmap. */
501 int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
505 struct buffer_head *main_bm_bh = NULL;
506 struct inode *main_bm_inode = NULL;
507 struct ocfs2_dinode *fe = NULL;
508 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
509 struct buffer_head *group_bh = NULL;
510 struct ocfs2_group_desc *group = NULL;
511 struct ocfs2_chain_list *cl;
512 struct ocfs2_chain_rec *cr;
517 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
520 main_bm_inode = ocfs2_get_system_file_inode(osb,
521 GLOBAL_BITMAP_SYSTEM_INODE,
523 if (!main_bm_inode) {
529 mutex_lock(&main_bm_inode->i_mutex);
531 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
537 fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
539 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
540 ocfs2_group_bitmap_size(osb->sb) * 8) {
541 mlog(ML_ERROR, "The disk is too old and small."
542 " Force to do offline resize.");
547 ret = ocfs2_read_block(osb, input->group, &group_bh, 0, NULL);
549 mlog(ML_ERROR, "Can't read the group descriptor # %llu "
550 "from the device.", (unsigned long long)input->group);
554 ocfs2_set_new_buffer_uptodate(inode, group_bh);
556 ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh);
562 mlog(0, "Add a new group %llu in chain = %u, length = %u\n",
563 (unsigned long long)input->group, input->chain, input->clusters);
565 handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS);
566 if (IS_ERR(handle)) {
567 mlog_errno(PTR_ERR(handle));
572 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
573 cl = &fe->id2.i_chain;
574 cr = &cl->cl_recs[input->chain];
576 ret = ocfs2_journal_access(handle, main_bm_inode, group_bh,
577 OCFS2_JOURNAL_ACCESS_WRITE);
583 group = (struct ocfs2_group_desc *)group_bh->b_data;
584 group->bg_next_group = cr->c_blkno;
586 ret = ocfs2_journal_dirty(handle, group_bh);
592 ret = ocfs2_journal_access(handle, main_bm_inode, main_bm_bh,
593 OCFS2_JOURNAL_ACCESS_WRITE);
599 if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) {
600 le16_add_cpu(&cl->cl_next_free_rec, 1);
601 memset(cr, 0, sizeof(struct ocfs2_chain_rec));
604 cr->c_blkno = le64_to_cpu(input->group);
605 le32_add_cpu(&cr->c_total, input->clusters * cl_bpc);
606 le32_add_cpu(&cr->c_free, input->frees * cl_bpc);
608 le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc);
609 le32_add_cpu(&fe->id1.bitmap1.i_used,
610 (input->clusters - input->frees) * cl_bpc);
611 le32_add_cpu(&fe->i_clusters, input->clusters);
613 ocfs2_journal_dirty(handle, main_bm_bh);
615 spin_lock(&OCFS2_I(main_bm_inode)->ip_lock);
616 OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
617 le64_add_cpu(&fe->i_size, input->clusters << osb->s_clustersize_bits);
618 spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock);
619 i_size_write(main_bm_inode, le64_to_cpu(fe->i_size));
621 ocfs2_update_super_and_backups(main_bm_inode, input->clusters);
624 ocfs2_commit_trans(osb, handle);
632 ocfs2_inode_unlock(main_bm_inode, 1);
635 mutex_unlock(&main_bm_inode->i_mutex);