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 #include <cluster/masklog.h>
41 #include "ocfs2_trace.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 u32 cluster, lgd_cluster;
65 u64 blkno, gd_blkno, lgd_blkno = le64_to_cpu(gd->bg_blkno);
67 for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
68 blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
69 cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
71 gd_blkno = ocfs2_which_cluster_group(inode, cluster);
72 if (gd_blkno < lgd_blkno)
74 else if (gd_blkno > lgd_blkno)
77 /* check if already done backup super */
78 lgd_cluster = ocfs2_blocks_to_clusters(inode->i_sb, lgd_blkno);
79 lgd_cluster += old_bg_clusters;
80 if (lgd_cluster >= cluster)
84 ocfs2_set_bit(cluster % cl_cpg,
85 (unsigned long *)gd->bg_bitmap);
87 ocfs2_clear_bit(cluster % cl_cpg,
88 (unsigned long *)gd->bg_bitmap);
95 static int ocfs2_update_last_group_and_inode(handle_t *handle,
96 struct inode *bm_inode,
97 struct buffer_head *bm_bh,
98 struct buffer_head *group_bh,
99 u32 first_new_cluster,
103 struct ocfs2_super *osb = OCFS2_SB(bm_inode->i_sb);
104 struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bm_bh->b_data;
105 struct ocfs2_chain_list *cl = &fe->id2.i_chain;
106 struct ocfs2_chain_rec *cr;
107 struct ocfs2_group_desc *group;
108 u16 chain, num_bits, backups = 0;
109 u16 cl_bpc = le16_to_cpu(cl->cl_bpc);
110 u16 cl_cpg = le16_to_cpu(cl->cl_cpg);
113 trace_ocfs2_update_last_group_and_inode(new_clusters,
116 ret = ocfs2_journal_access_gd(handle, INODE_CACHE(bm_inode),
117 group_bh, OCFS2_JOURNAL_ACCESS_WRITE);
123 group = (struct ocfs2_group_desc *)group_bh->b_data;
125 old_bg_clusters = le16_to_cpu(group->bg_bits) / cl_bpc;
126 /* update the group first. */
127 num_bits = new_clusters * cl_bpc;
128 le16_add_cpu(&group->bg_bits, num_bits);
129 le16_add_cpu(&group->bg_free_bits_count, num_bits);
132 * check whether there are some new backup superblocks exist in
133 * this group and update the group bitmap accordingly.
135 if (OCFS2_HAS_COMPAT_FEATURE(osb->sb,
136 OCFS2_FEATURE_COMPAT_BACKUP_SB)) {
137 backups = ocfs2_calc_new_backup_super(bm_inode,
141 cl_cpg, old_bg_clusters, 1);
142 le16_add_cpu(&group->bg_free_bits_count, -1 * backups);
145 ocfs2_journal_dirty(handle, group_bh);
147 /* update the inode accordingly. */
148 ret = ocfs2_journal_access_di(handle, INODE_CACHE(bm_inode), bm_bh,
149 OCFS2_JOURNAL_ACCESS_WRITE);
155 chain = le16_to_cpu(group->bg_chain);
156 cr = (&cl->cl_recs[chain]);
157 le32_add_cpu(&cr->c_total, num_bits);
158 le32_add_cpu(&cr->c_free, num_bits);
159 le32_add_cpu(&fe->id1.bitmap1.i_total, num_bits);
160 le32_add_cpu(&fe->i_clusters, new_clusters);
163 le32_add_cpu(&cr->c_free, -1 * backups);
164 le32_add_cpu(&fe->id1.bitmap1.i_used, backups);
167 spin_lock(&OCFS2_I(bm_inode)->ip_lock);
168 OCFS2_I(bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
169 le64_add_cpu(&fe->i_size, new_clusters << osb->s_clustersize_bits);
170 spin_unlock(&OCFS2_I(bm_inode)->ip_lock);
171 i_size_write(bm_inode, le64_to_cpu(fe->i_size));
173 ocfs2_journal_dirty(handle, bm_bh);
177 ocfs2_calc_new_backup_super(bm_inode,
181 cl_cpg, old_bg_clusters, 0);
182 le16_add_cpu(&group->bg_free_bits_count, backups);
183 le16_add_cpu(&group->bg_bits, -1 * num_bits);
184 le16_add_cpu(&group->bg_free_bits_count, -1 * num_bits);
192 static int update_backups(struct inode * inode, u32 clusters, char *data)
197 struct buffer_head *backup = NULL;
198 struct ocfs2_dinode *backup_di = NULL;
199 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
201 /* calculate the real backups we need to update. */
202 for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
203 blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
204 cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
205 if (cluster > clusters)
208 ret = ocfs2_read_blocks_sync(osb, blkno, 1, &backup);
214 memcpy(backup->b_data, data, inode->i_sb->s_blocksize);
216 backup_di = (struct ocfs2_dinode *)backup->b_data;
217 backup_di->i_blkno = cpu_to_le64(blkno);
219 ret = ocfs2_write_super_or_backup(osb, backup);
231 static void ocfs2_update_super_and_backups(struct inode *inode,
236 struct buffer_head *super_bh = NULL;
237 struct ocfs2_dinode *super_di = NULL;
238 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
241 * update the superblock last.
242 * It doesn't matter if the write failed.
244 ret = ocfs2_read_blocks_sync(osb, OCFS2_SUPER_BLOCK_BLKNO, 1,
251 super_di = (struct ocfs2_dinode *)super_bh->b_data;
252 le32_add_cpu(&super_di->i_clusters, new_clusters);
253 clusters = le32_to_cpu(super_di->i_clusters);
255 ret = ocfs2_write_super_or_backup(osb, super_bh);
261 if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_COMPAT_BACKUP_SB))
262 ret = update_backups(inode, clusters, super_bh->b_data);
267 printk(KERN_WARNING "ocfs2: Failed to update super blocks on %s"
268 " during fs resize. This condition is not fatal,"
269 " but fsck.ocfs2 should be run to fix it\n",
275 * Extend the filesystem to the new number of clusters specified. This entry
276 * point is only used to extend the current filesystem to the end of the last
279 int ocfs2_group_extend(struct inode * inode, int new_clusters)
283 struct buffer_head *main_bm_bh = NULL;
284 struct buffer_head *group_bh = NULL;
285 struct inode *main_bm_inode = NULL;
286 struct ocfs2_dinode *fe = NULL;
287 struct ocfs2_group_desc *group = NULL;
288 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
290 u32 first_new_cluster;
293 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
296 if (new_clusters < 0)
298 else if (new_clusters == 0)
301 main_bm_inode = ocfs2_get_system_file_inode(osb,
302 GLOBAL_BITMAP_SYSTEM_INODE,
304 if (!main_bm_inode) {
310 mutex_lock(&main_bm_inode->i_mutex);
312 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
318 fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
320 /* main_bm_bh is validated by inode read inside ocfs2_inode_lock(),
321 * so any corruption is a code bug. */
322 BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
324 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
325 ocfs2_group_bitmap_size(osb->sb, 0,
326 osb->s_feature_incompat) * 8) {
327 mlog(ML_ERROR, "The disk is too old and small. "
328 "Force to do offline resize.");
333 first_new_cluster = le32_to_cpu(fe->i_clusters);
334 lgd_blkno = ocfs2_which_cluster_group(main_bm_inode,
335 first_new_cluster - 1);
337 ret = ocfs2_read_group_descriptor(main_bm_inode, fe, lgd_blkno,
343 group = (struct ocfs2_group_desc *)group_bh->b_data;
345 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
346 if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters >
347 le16_to_cpu(fe->id2.i_chain.cl_cpg)) {
353 trace_ocfs2_group_extend(
354 (unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters);
356 handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
357 if (IS_ERR(handle)) {
358 mlog_errno(PTR_ERR(handle));
363 /* update the last group descriptor and inode. */
364 ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode,
365 main_bm_bh, group_bh,
373 ocfs2_update_super_and_backups(main_bm_inode, new_clusters);
376 ocfs2_commit_trans(osb, handle);
381 ocfs2_inode_unlock(main_bm_inode, 1);
384 mutex_unlock(&main_bm_inode->i_mutex);
391 static int ocfs2_check_new_group(struct inode *inode,
392 struct ocfs2_dinode *di,
393 struct ocfs2_new_group_input *input,
394 struct buffer_head *group_bh)
397 struct ocfs2_group_desc *gd =
398 (struct ocfs2_group_desc *)group_bh->b_data;
399 u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc);
401 ret = ocfs2_check_group_descriptor(inode->i_sb, di, group_bh);
406 if (le16_to_cpu(gd->bg_chain) != input->chain)
407 mlog(ML_ERROR, "Group descriptor # %llu has bad chain %u "
408 "while input has %u set.\n",
409 (unsigned long long)le64_to_cpu(gd->bg_blkno),
410 le16_to_cpu(gd->bg_chain), input->chain);
411 else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc)
412 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
413 "input has %u clusters set\n",
414 (unsigned long long)le64_to_cpu(gd->bg_blkno),
415 le16_to_cpu(gd->bg_bits), input->clusters);
416 else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc)
417 mlog(ML_ERROR, "Group descriptor # %llu has free bit count %u "
418 "but it should have %u set\n",
419 (unsigned long long)le64_to_cpu(gd->bg_blkno),
420 le16_to_cpu(gd->bg_bits),
421 input->frees * cl_bpc);
429 static int ocfs2_verify_group_and_input(struct inode *inode,
430 struct ocfs2_dinode *di,
431 struct ocfs2_new_group_input *input,
432 struct buffer_head *group_bh)
434 u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count);
435 u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
436 u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec);
437 u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group);
438 u32 total_clusters = le32_to_cpu(di->i_clusters);
441 if (cluster < total_clusters)
442 mlog(ML_ERROR, "add a group which is in the current volume.\n");
443 else if (input->chain >= cl_count)
444 mlog(ML_ERROR, "input chain exceeds the limit.\n");
445 else if (next_free != cl_count && next_free != input->chain)
447 "the add group should be in chain %u\n", next_free);
448 else if (total_clusters + input->clusters < total_clusters)
449 mlog(ML_ERROR, "add group's clusters overflow.\n");
450 else if (input->clusters > cl_cpg)
451 mlog(ML_ERROR, "the cluster exceeds the maximum of a group\n");
452 else if (input->frees > input->clusters)
453 mlog(ML_ERROR, "the free cluster exceeds the total clusters\n");
454 else if (total_clusters % cl_cpg != 0)
456 "the last group isn't full. Use group extend first.\n");
457 else if (input->group != ocfs2_which_cluster_group(inode, cluster))
458 mlog(ML_ERROR, "group blkno is invalid\n");
459 else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh)))
460 mlog(ML_ERROR, "group descriptor check failed.\n");
467 /* Add a new group descriptor to global_bitmap. */
468 int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
472 struct buffer_head *main_bm_bh = NULL;
473 struct inode *main_bm_inode = NULL;
474 struct ocfs2_dinode *fe = NULL;
475 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
476 struct buffer_head *group_bh = NULL;
477 struct ocfs2_group_desc *group = NULL;
478 struct ocfs2_chain_list *cl;
479 struct ocfs2_chain_rec *cr;
482 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
485 main_bm_inode = ocfs2_get_system_file_inode(osb,
486 GLOBAL_BITMAP_SYSTEM_INODE,
488 if (!main_bm_inode) {
494 mutex_lock(&main_bm_inode->i_mutex);
496 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
502 fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
504 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
505 ocfs2_group_bitmap_size(osb->sb, 0,
506 osb->s_feature_incompat) * 8) {
507 mlog(ML_ERROR, "The disk is too old and small."
508 " Force to do offline resize.");
513 ret = ocfs2_read_blocks_sync(osb, input->group, 1, &group_bh);
515 mlog(ML_ERROR, "Can't read the group descriptor # %llu "
516 "from the device.", (unsigned long long)input->group);
520 ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), group_bh);
522 ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh);
528 trace_ocfs2_group_add((unsigned long long)input->group,
529 input->chain, input->clusters, input->frees);
531 handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS);
532 if (IS_ERR(handle)) {
533 mlog_errno(PTR_ERR(handle));
538 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
539 cl = &fe->id2.i_chain;
540 cr = &cl->cl_recs[input->chain];
542 ret = ocfs2_journal_access_gd(handle, INODE_CACHE(main_bm_inode),
543 group_bh, OCFS2_JOURNAL_ACCESS_WRITE);
549 group = (struct ocfs2_group_desc *)group_bh->b_data;
550 group->bg_next_group = cr->c_blkno;
551 ocfs2_journal_dirty(handle, group_bh);
553 ret = ocfs2_journal_access_di(handle, INODE_CACHE(main_bm_inode),
554 main_bm_bh, OCFS2_JOURNAL_ACCESS_WRITE);
560 if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) {
561 le16_add_cpu(&cl->cl_next_free_rec, 1);
562 memset(cr, 0, sizeof(struct ocfs2_chain_rec));
565 cr->c_blkno = cpu_to_le64(input->group);
566 le32_add_cpu(&cr->c_total, input->clusters * cl_bpc);
567 le32_add_cpu(&cr->c_free, input->frees * cl_bpc);
569 le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc);
570 le32_add_cpu(&fe->id1.bitmap1.i_used,
571 (input->clusters - input->frees) * cl_bpc);
572 le32_add_cpu(&fe->i_clusters, input->clusters);
574 ocfs2_journal_dirty(handle, main_bm_bh);
576 spin_lock(&OCFS2_I(main_bm_inode)->ip_lock);
577 OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
578 le64_add_cpu(&fe->i_size, input->clusters << osb->s_clustersize_bits);
579 spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock);
580 i_size_write(main_bm_inode, le64_to_cpu(fe->i_size));
582 ocfs2_update_super_and_backups(main_bm_inode, input->clusters);
585 ocfs2_commit_trans(osb, handle);
590 ocfs2_inode_unlock(main_bm_inode, 1);
593 mutex_unlock(&main_bm_inode->i_mutex);