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_gd(handle, INODE_CACHE(bm_inode),
110 group_bh, 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 ocfs2_journal_dirty(handle, group_bh);
139 /* update the inode accordingly. */
140 ret = ocfs2_journal_access_di(handle, INODE_CACHE(bm_inode), bm_bh,
141 OCFS2_JOURNAL_ACCESS_WRITE);
147 chain = le16_to_cpu(group->bg_chain);
148 cr = (&cl->cl_recs[chain]);
149 le32_add_cpu(&cr->c_total, num_bits);
150 le32_add_cpu(&cr->c_free, num_bits);
151 le32_add_cpu(&fe->id1.bitmap1.i_total, num_bits);
152 le32_add_cpu(&fe->i_clusters, new_clusters);
155 le32_add_cpu(&cr->c_free, -1 * backups);
156 le32_add_cpu(&fe->id1.bitmap1.i_used, backups);
159 spin_lock(&OCFS2_I(bm_inode)->ip_lock);
160 OCFS2_I(bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
161 le64_add_cpu(&fe->i_size, new_clusters << osb->s_clustersize_bits);
162 spin_unlock(&OCFS2_I(bm_inode)->ip_lock);
163 i_size_write(bm_inode, le64_to_cpu(fe->i_size));
165 ocfs2_journal_dirty(handle, bm_bh);
169 ocfs2_calc_new_backup_super(bm_inode,
174 le16_add_cpu(&group->bg_free_bits_count, backups);
175 le16_add_cpu(&group->bg_bits, -1 * num_bits);
176 le16_add_cpu(&group->bg_free_bits_count, -1 * num_bits);
183 static int update_backups(struct inode * inode, u32 clusters, char *data)
188 struct buffer_head *backup = NULL;
189 struct ocfs2_dinode *backup_di = NULL;
190 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
192 /* calculate the real backups we need to update. */
193 for (i = 0; i < OCFS2_MAX_BACKUP_SUPERBLOCKS; i++) {
194 blkno = ocfs2_backup_super_blkno(inode->i_sb, i);
195 cluster = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
196 if (cluster > clusters)
199 ret = ocfs2_read_blocks_sync(osb, blkno, 1, &backup);
205 memcpy(backup->b_data, data, inode->i_sb->s_blocksize);
207 backup_di = (struct ocfs2_dinode *)backup->b_data;
208 backup_di->i_blkno = cpu_to_le64(blkno);
210 ret = ocfs2_write_super_or_backup(osb, backup);
222 static void ocfs2_update_super_and_backups(struct inode *inode,
227 struct buffer_head *super_bh = NULL;
228 struct ocfs2_dinode *super_di = NULL;
229 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
232 * update the superblock last.
233 * It doesn't matter if the write failed.
235 ret = ocfs2_read_blocks_sync(osb, OCFS2_SUPER_BLOCK_BLKNO, 1,
242 super_di = (struct ocfs2_dinode *)super_bh->b_data;
243 le32_add_cpu(&super_di->i_clusters, new_clusters);
244 clusters = le32_to_cpu(super_di->i_clusters);
246 ret = ocfs2_write_super_or_backup(osb, super_bh);
252 if (OCFS2_HAS_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_COMPAT_BACKUP_SB))
253 ret = update_backups(inode, clusters, super_bh->b_data);
258 printk(KERN_WARNING "ocfs2: Failed to update super blocks on %s"
259 " during fs resize. This condition is not fatal,"
260 " but fsck.ocfs2 should be run to fix it\n",
266 * Extend the filesystem to the new number of clusters specified. This entry
267 * point is only used to extend the current filesystem to the end of the last
270 int ocfs2_group_extend(struct inode * inode, int new_clusters)
274 struct buffer_head *main_bm_bh = NULL;
275 struct buffer_head *group_bh = NULL;
276 struct inode *main_bm_inode = NULL;
277 struct ocfs2_dinode *fe = NULL;
278 struct ocfs2_group_desc *group = NULL;
279 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
281 u32 first_new_cluster;
286 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
289 if (new_clusters < 0)
291 else if (new_clusters == 0)
294 main_bm_inode = ocfs2_get_system_file_inode(osb,
295 GLOBAL_BITMAP_SYSTEM_INODE,
297 if (!main_bm_inode) {
303 mutex_lock(&main_bm_inode->i_mutex);
305 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
311 fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
313 /* main_bm_bh is validated by inode read inside ocfs2_inode_lock(),
314 * so any corruption is a code bug. */
315 BUG_ON(!OCFS2_IS_VALID_DINODE(fe));
317 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
318 ocfs2_group_bitmap_size(osb->sb, 0,
319 osb->s_feature_incompat) * 8) {
320 mlog(ML_ERROR, "The disk is too old and small. "
321 "Force to do offline resize.");
326 first_new_cluster = le32_to_cpu(fe->i_clusters);
327 lgd_blkno = ocfs2_which_cluster_group(main_bm_inode,
328 first_new_cluster - 1);
330 ret = ocfs2_read_group_descriptor(main_bm_inode, fe, lgd_blkno,
336 group = (struct ocfs2_group_desc *)group_bh->b_data;
338 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
339 if (le16_to_cpu(group->bg_bits) / cl_bpc + new_clusters >
340 le16_to_cpu(fe->id2.i_chain.cl_cpg)) {
345 mlog(0, "extend the last group at %llu, new clusters = %d\n",
346 (unsigned long long)le64_to_cpu(group->bg_blkno), new_clusters);
348 handle = ocfs2_start_trans(osb, OCFS2_GROUP_EXTEND_CREDITS);
349 if (IS_ERR(handle)) {
350 mlog_errno(PTR_ERR(handle));
355 /* update the last group descriptor and inode. */
356 ret = ocfs2_update_last_group_and_inode(handle, main_bm_inode,
357 main_bm_bh, group_bh,
365 ocfs2_update_super_and_backups(main_bm_inode, new_clusters);
368 ocfs2_commit_trans(osb, handle);
373 ocfs2_inode_unlock(main_bm_inode, 1);
376 mutex_unlock(&main_bm_inode->i_mutex);
384 static int ocfs2_check_new_group(struct inode *inode,
385 struct ocfs2_dinode *di,
386 struct ocfs2_new_group_input *input,
387 struct buffer_head *group_bh)
390 struct ocfs2_group_desc *gd =
391 (struct ocfs2_group_desc *)group_bh->b_data;
392 u16 cl_bpc = le16_to_cpu(di->id2.i_chain.cl_bpc);
394 ret = ocfs2_check_group_descriptor(inode->i_sb, di, group_bh);
399 if (le16_to_cpu(gd->bg_chain) != input->chain)
400 mlog(ML_ERROR, "Group descriptor # %llu has bad chain %u "
401 "while input has %u set.\n",
402 (unsigned long long)le64_to_cpu(gd->bg_blkno),
403 le16_to_cpu(gd->bg_chain), input->chain);
404 else if (le16_to_cpu(gd->bg_bits) != input->clusters * cl_bpc)
405 mlog(ML_ERROR, "Group descriptor # %llu has bit count %u but "
406 "input has %u clusters set\n",
407 (unsigned long long)le64_to_cpu(gd->bg_blkno),
408 le16_to_cpu(gd->bg_bits), input->clusters);
409 else if (le16_to_cpu(gd->bg_free_bits_count) != input->frees * cl_bpc)
410 mlog(ML_ERROR, "Group descriptor # %llu has free bit count %u "
411 "but it should have %u set\n",
412 (unsigned long long)le64_to_cpu(gd->bg_blkno),
413 le16_to_cpu(gd->bg_bits),
414 input->frees * cl_bpc);
422 static int ocfs2_verify_group_and_input(struct inode *inode,
423 struct ocfs2_dinode *di,
424 struct ocfs2_new_group_input *input,
425 struct buffer_head *group_bh)
427 u16 cl_count = le16_to_cpu(di->id2.i_chain.cl_count);
428 u16 cl_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
429 u16 next_free = le16_to_cpu(di->id2.i_chain.cl_next_free_rec);
430 u32 cluster = ocfs2_blocks_to_clusters(inode->i_sb, input->group);
431 u32 total_clusters = le32_to_cpu(di->i_clusters);
434 if (cluster < total_clusters)
435 mlog(ML_ERROR, "add a group which is in the current volume.\n");
436 else if (input->chain >= cl_count)
437 mlog(ML_ERROR, "input chain exceeds the limit.\n");
438 else if (next_free != cl_count && next_free != input->chain)
440 "the add group should be in chain %u\n", next_free);
441 else if (total_clusters + input->clusters < total_clusters)
442 mlog(ML_ERROR, "add group's clusters overflow.\n");
443 else if (input->clusters > cl_cpg)
444 mlog(ML_ERROR, "the cluster exceeds the maximum of a group\n");
445 else if (input->frees > input->clusters)
446 mlog(ML_ERROR, "the free cluster exceeds the total clusters\n");
447 else if (total_clusters % cl_cpg != 0)
449 "the last group isn't full. Use group extend first.\n");
450 else if (input->group != ocfs2_which_cluster_group(inode, cluster))
451 mlog(ML_ERROR, "group blkno is invalid\n");
452 else if ((ret = ocfs2_check_new_group(inode, di, input, group_bh)))
453 mlog(ML_ERROR, "group descriptor check failed.\n");
460 /* Add a new group descriptor to global_bitmap. */
461 int ocfs2_group_add(struct inode *inode, struct ocfs2_new_group_input *input)
465 struct buffer_head *main_bm_bh = NULL;
466 struct inode *main_bm_inode = NULL;
467 struct ocfs2_dinode *fe = NULL;
468 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
469 struct buffer_head *group_bh = NULL;
470 struct ocfs2_group_desc *group = NULL;
471 struct ocfs2_chain_list *cl;
472 struct ocfs2_chain_rec *cr;
477 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
480 main_bm_inode = ocfs2_get_system_file_inode(osb,
481 GLOBAL_BITMAP_SYSTEM_INODE,
483 if (!main_bm_inode) {
489 mutex_lock(&main_bm_inode->i_mutex);
491 ret = ocfs2_inode_lock(main_bm_inode, &main_bm_bh, 1);
497 fe = (struct ocfs2_dinode *)main_bm_bh->b_data;
499 if (le16_to_cpu(fe->id2.i_chain.cl_cpg) !=
500 ocfs2_group_bitmap_size(osb->sb, 0,
501 osb->s_feature_incompat) * 8) {
502 mlog(ML_ERROR, "The disk is too old and small."
503 " Force to do offline resize.");
508 ret = ocfs2_read_blocks_sync(osb, input->group, 1, &group_bh);
510 mlog(ML_ERROR, "Can't read the group descriptor # %llu "
511 "from the device.", (unsigned long long)input->group);
515 ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), group_bh);
517 ret = ocfs2_verify_group_and_input(main_bm_inode, fe, input, group_bh);
523 mlog(0, "Add a new group %llu in chain = %u, length = %u\n",
524 (unsigned long long)input->group, input->chain, input->clusters);
526 handle = ocfs2_start_trans(osb, OCFS2_GROUP_ADD_CREDITS);
527 if (IS_ERR(handle)) {
528 mlog_errno(PTR_ERR(handle));
533 cl_bpc = le16_to_cpu(fe->id2.i_chain.cl_bpc);
534 cl = &fe->id2.i_chain;
535 cr = &cl->cl_recs[input->chain];
537 ret = ocfs2_journal_access_gd(handle, INODE_CACHE(main_bm_inode),
538 group_bh, OCFS2_JOURNAL_ACCESS_WRITE);
544 group = (struct ocfs2_group_desc *)group_bh->b_data;
545 group->bg_next_group = cr->c_blkno;
546 ocfs2_journal_dirty(handle, group_bh);
548 ret = ocfs2_journal_access_di(handle, INODE_CACHE(main_bm_inode),
549 main_bm_bh, OCFS2_JOURNAL_ACCESS_WRITE);
555 if (input->chain == le16_to_cpu(cl->cl_next_free_rec)) {
556 le16_add_cpu(&cl->cl_next_free_rec, 1);
557 memset(cr, 0, sizeof(struct ocfs2_chain_rec));
560 cr->c_blkno = cpu_to_le64(input->group);
561 le32_add_cpu(&cr->c_total, input->clusters * cl_bpc);
562 le32_add_cpu(&cr->c_free, input->frees * cl_bpc);
564 le32_add_cpu(&fe->id1.bitmap1.i_total, input->clusters *cl_bpc);
565 le32_add_cpu(&fe->id1.bitmap1.i_used,
566 (input->clusters - input->frees) * cl_bpc);
567 le32_add_cpu(&fe->i_clusters, input->clusters);
569 ocfs2_journal_dirty(handle, main_bm_bh);
571 spin_lock(&OCFS2_I(main_bm_inode)->ip_lock);
572 OCFS2_I(main_bm_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
573 le64_add_cpu(&fe->i_size, input->clusters << osb->s_clustersize_bits);
574 spin_unlock(&OCFS2_I(main_bm_inode)->ip_lock);
575 i_size_write(main_bm_inode, le64_to_cpu(fe->i_size));
577 ocfs2_update_super_and_backups(main_bm_inode, input->clusters);
580 ocfs2_commit_trans(osb, handle);
585 ocfs2_inode_unlock(main_bm_inode, 1);
588 mutex_unlock(&main_bm_inode->i_mutex);