ext4: improve ext4_fill_flex_info() a bit
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / resize.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/resize.c
ac27a0ec 3 *
617ba13b 4 * Support for resizing an ext4 filesystem while it is mounted.
ac27a0ec
DK
5 *
6 * Copyright (C) 2001, 2002 Andreas Dilger <adilger@clusterfs.com>
7 *
8 * This could probably be made into a module, because it is not often in use.
9 */
10
11
617ba13b 12#define EXT4FS_DEBUG
ac27a0ec 13
ac27a0ec
DK
14#include <linux/errno.h>
15#include <linux/slab.h>
16
3dcf5451 17#include "ext4_jbd2.h"
717d50e4 18#include "group.h"
ac27a0ec
DK
19
20#define outside(b, first, last) ((b) < (first) || (b) >= (last))
21#define inside(b, first, last) ((b) >= (first) && (b) < (last))
22
23static int verify_group_input(struct super_block *sb,
617ba13b 24 struct ext4_new_group_data *input)
ac27a0ec 25{
617ba13b
MC
26 struct ext4_sb_info *sbi = EXT4_SB(sb);
27 struct ext4_super_block *es = sbi->s_es;
bd81d8ee 28 ext4_fsblk_t start = ext4_blocks_count(es);
617ba13b 29 ext4_fsblk_t end = start + input->blocks_count;
fd2d4291 30 ext4_group_t group = input->group;
617ba13b
MC
31 ext4_fsblk_t itend = input->inode_table + sbi->s_itb_per_group;
32 unsigned overhead = ext4_bg_has_super(sb, group) ?
33 (1 + ext4_bg_num_gdb(sb, group) +
ac27a0ec 34 le16_to_cpu(es->s_reserved_gdt_blocks)) : 0;
617ba13b 35 ext4_fsblk_t metaend = start + overhead;
ac27a0ec 36 struct buffer_head *bh = NULL;
3a5b2ecd 37 ext4_grpblk_t free_blocks_count, offset;
ac27a0ec
DK
38 int err = -EINVAL;
39
40 input->free_blocks_count = free_blocks_count =
41 input->blocks_count - 2 - overhead - sbi->s_itb_per_group;
42
43 if (test_opt(sb, DEBUG))
617ba13b 44 printk(KERN_DEBUG "EXT4-fs: adding %s group %u: %u blocks "
ac27a0ec 45 "(%d free, %u reserved)\n",
617ba13b 46 ext4_bg_has_super(sb, input->group) ? "normal" :
ac27a0ec
DK
47 "no-super", input->group, input->blocks_count,
48 free_blocks_count, input->reserved_blocks);
49
3a5b2ecd 50 ext4_get_group_no_and_offset(sb, start, NULL, &offset);
ac27a0ec 51 if (group != sbi->s_groups_count)
46e665e9 52 ext4_warning(sb, __func__,
ac27a0ec
DK
53 "Cannot add at group %u (only %lu groups)",
54 input->group, sbi->s_groups_count);
3a5b2ecd 55 else if (offset != 0)
46e665e9 56 ext4_warning(sb, __func__, "Last group not full");
ac27a0ec 57 else if (input->reserved_blocks > input->blocks_count / 5)
46e665e9 58 ext4_warning(sb, __func__, "Reserved blocks too high (%u)",
ac27a0ec
DK
59 input->reserved_blocks);
60 else if (free_blocks_count < 0)
46e665e9 61 ext4_warning(sb, __func__, "Bad blocks count %u",
ac27a0ec
DK
62 input->blocks_count);
63 else if (!(bh = sb_bread(sb, end - 1)))
46e665e9 64 ext4_warning(sb, __func__,
2ae02107 65 "Cannot read last block (%llu)",
ac27a0ec
DK
66 end - 1);
67 else if (outside(input->block_bitmap, start, end))
46e665e9 68 ext4_warning(sb, __func__,
bd81d8ee 69 "Block bitmap not in group (block %llu)",
1939e49a 70 (unsigned long long)input->block_bitmap);
ac27a0ec 71 else if (outside(input->inode_bitmap, start, end))
46e665e9 72 ext4_warning(sb, __func__,
bd81d8ee 73 "Inode bitmap not in group (block %llu)",
1939e49a 74 (unsigned long long)input->inode_bitmap);
ac27a0ec
DK
75 else if (outside(input->inode_table, start, end) ||
76 outside(itend - 1, start, end))
46e665e9 77 ext4_warning(sb, __func__,
bd81d8ee 78 "Inode table not in group (blocks %llu-%llu)",
1939e49a 79 (unsigned long long)input->inode_table, itend - 1);
ac27a0ec 80 else if (input->inode_bitmap == input->block_bitmap)
46e665e9 81 ext4_warning(sb, __func__,
bd81d8ee 82 "Block bitmap same as inode bitmap (%llu)",
1939e49a 83 (unsigned long long)input->block_bitmap);
ac27a0ec 84 else if (inside(input->block_bitmap, input->inode_table, itend))
46e665e9 85 ext4_warning(sb, __func__,
bd81d8ee 86 "Block bitmap (%llu) in inode table (%llu-%llu)",
1939e49a
RD
87 (unsigned long long)input->block_bitmap,
88 (unsigned long long)input->inode_table, itend - 1);
ac27a0ec 89 else if (inside(input->inode_bitmap, input->inode_table, itend))
46e665e9 90 ext4_warning(sb, __func__,
bd81d8ee 91 "Inode bitmap (%llu) in inode table (%llu-%llu)",
1939e49a
RD
92 (unsigned long long)input->inode_bitmap,
93 (unsigned long long)input->inode_table, itend - 1);
ac27a0ec 94 else if (inside(input->block_bitmap, start, metaend))
46e665e9 95 ext4_warning(sb, __func__,
bd81d8ee 96 "Block bitmap (%llu) in GDT table"
2ae02107 97 " (%llu-%llu)",
1939e49a
RD
98 (unsigned long long)input->block_bitmap,
99 start, metaend - 1);
ac27a0ec 100 else if (inside(input->inode_bitmap, start, metaend))
46e665e9 101 ext4_warning(sb, __func__,
bd81d8ee 102 "Inode bitmap (%llu) in GDT table"
2ae02107 103 " (%llu-%llu)",
1939e49a
RD
104 (unsigned long long)input->inode_bitmap,
105 start, metaend - 1);
ac27a0ec
DK
106 else if (inside(input->inode_table, start, metaend) ||
107 inside(itend - 1, start, metaend))
46e665e9 108 ext4_warning(sb, __func__,
2ae02107
MC
109 "Inode table (%llu-%llu) overlaps"
110 "GDT table (%llu-%llu)",
1939e49a
RD
111 (unsigned long long)input->inode_table,
112 itend - 1, start, metaend - 1);
ac27a0ec
DK
113 else
114 err = 0;
115 brelse(bh);
116
117 return err;
118}
119
120static struct buffer_head *bclean(handle_t *handle, struct super_block *sb,
617ba13b 121 ext4_fsblk_t blk)
ac27a0ec
DK
122{
123 struct buffer_head *bh;
124 int err;
125
126 bh = sb_getblk(sb, blk);
127 if (!bh)
128 return ERR_PTR(-EIO);
617ba13b 129 if ((err = ext4_journal_get_write_access(handle, bh))) {
ac27a0ec
DK
130 brelse(bh);
131 bh = ERR_PTR(err);
132 } else {
133 lock_buffer(bh);
134 memset(bh->b_data, 0, sb->s_blocksize);
135 set_buffer_uptodate(bh);
136 unlock_buffer(bh);
137 }
138
139 return bh;
140}
141
14904107
ES
142/*
143 * If we have fewer than thresh credits, extend by EXT4_MAX_TRANS_DATA.
144 * If that fails, restart the transaction & regain write access for the
145 * buffer head which is used for block_bitmap modifications.
146 */
147static int extend_or_restart_transaction(handle_t *handle, int thresh,
148 struct buffer_head *bh)
149{
150 int err;
151
152 if (handle->h_buffer_credits >= thresh)
153 return 0;
154
155 err = ext4_journal_extend(handle, EXT4_MAX_TRANS_DATA);
156 if (err < 0)
157 return err;
158 if (err) {
159 if ((err = ext4_journal_restart(handle, EXT4_MAX_TRANS_DATA)))
160 return err;
161 if ((err = ext4_journal_get_write_access(handle, bh)))
162 return err;
163 }
164
165 return 0;
166}
167
ac27a0ec
DK
168/*
169 * Set up the block and inode bitmaps, and the inode table for the new group.
170 * This doesn't need to be part of the main transaction, since we are only
171 * changing blocks outside the actual filesystem. We still do journaling to
172 * ensure the recovery is correct in case of a failure just after resize.
173 * If any part of this fails, we simply abort the resize.
174 */
175static int setup_new_group_blocks(struct super_block *sb,
617ba13b 176 struct ext4_new_group_data *input)
ac27a0ec 177{
617ba13b
MC
178 struct ext4_sb_info *sbi = EXT4_SB(sb);
179 ext4_fsblk_t start = ext4_group_first_block_no(sb, input->group);
180 int reserved_gdb = ext4_bg_has_super(sb, input->group) ?
ac27a0ec 181 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0;
617ba13b 182 unsigned long gdblocks = ext4_bg_num_gdb(sb, input->group);
ac27a0ec
DK
183 struct buffer_head *bh;
184 handle_t *handle;
617ba13b
MC
185 ext4_fsblk_t block;
186 ext4_grpblk_t bit;
ac27a0ec
DK
187 int i;
188 int err = 0, err2;
189
14904107
ES
190 /* This transaction may be extended/restarted along the way */
191 handle = ext4_journal_start_sb(sb, EXT4_MAX_TRANS_DATA);
192
ac27a0ec
DK
193 if (IS_ERR(handle))
194 return PTR_ERR(handle);
195
196 lock_super(sb);
197 if (input->group != sbi->s_groups_count) {
198 err = -EBUSY;
199 goto exit_journal;
200 }
201
202 if (IS_ERR(bh = bclean(handle, sb, input->block_bitmap))) {
203 err = PTR_ERR(bh);
204 goto exit_journal;
205 }
206
617ba13b 207 if (ext4_bg_has_super(sb, input->group)) {
c549a95d 208 ext4_debug("mark backup superblock %#04llx (+0)\n", start);
617ba13b 209 ext4_set_bit(0, bh->b_data);
ac27a0ec
DK
210 }
211
212 /* Copy all of the GDT blocks into the backup in this group */
213 for (i = 0, bit = 1, block = start + 1;
214 i < gdblocks; i++, block++, bit++) {
215 struct buffer_head *gdb;
216
c549a95d 217 ext4_debug("update backup group %#04llx (+%d)\n", block, bit);
ac27a0ec 218
14904107
ES
219 if ((err = extend_or_restart_transaction(handle, 1, bh)))
220 goto exit_bh;
221
ac27a0ec
DK
222 gdb = sb_getblk(sb, block);
223 if (!gdb) {
224 err = -EIO;
225 goto exit_bh;
226 }
617ba13b 227 if ((err = ext4_journal_get_write_access(handle, gdb))) {
ac27a0ec
DK
228 brelse(gdb);
229 goto exit_bh;
230 }
5b615287
ES
231 lock_buffer(gdb);
232 memcpy(gdb->b_data, sbi->s_group_desc[i]->b_data, gdb->b_size);
ac27a0ec 233 set_buffer_uptodate(gdb);
5b615287 234 unlock_buffer(gdb);
617ba13b
MC
235 ext4_journal_dirty_metadata(handle, gdb);
236 ext4_set_bit(bit, bh->b_data);
ac27a0ec
DK
237 brelse(gdb);
238 }
239
240 /* Zero out all of the reserved backup group descriptor table blocks */
241 for (i = 0, bit = gdblocks + 1, block = start + bit;
242 i < reserved_gdb; i++, block++, bit++) {
243 struct buffer_head *gdb;
244
c549a95d 245 ext4_debug("clear reserved block %#04llx (+%d)\n", block, bit);
ac27a0ec 246
14904107
ES
247 if ((err = extend_or_restart_transaction(handle, 1, bh)))
248 goto exit_bh;
249
ac27a0ec
DK
250 if (IS_ERR(gdb = bclean(handle, sb, block))) {
251 err = PTR_ERR(bh);
252 goto exit_bh;
253 }
617ba13b
MC
254 ext4_journal_dirty_metadata(handle, gdb);
255 ext4_set_bit(bit, bh->b_data);
ac27a0ec
DK
256 brelse(gdb);
257 }
c549a95d 258 ext4_debug("mark block bitmap %#04llx (+%llu)\n", input->block_bitmap,
ac27a0ec 259 input->block_bitmap - start);
617ba13b 260 ext4_set_bit(input->block_bitmap - start, bh->b_data);
c549a95d 261 ext4_debug("mark inode bitmap %#04llx (+%llu)\n", input->inode_bitmap,
ac27a0ec 262 input->inode_bitmap - start);
617ba13b 263 ext4_set_bit(input->inode_bitmap - start, bh->b_data);
ac27a0ec
DK
264
265 /* Zero out all of the inode table blocks */
266 for (i = 0, block = input->inode_table, bit = block - start;
267 i < sbi->s_itb_per_group; i++, bit++, block++) {
268 struct buffer_head *it;
269
c549a95d 270 ext4_debug("clear inode block %#04llx (+%d)\n", block, bit);
14904107
ES
271
272 if ((err = extend_or_restart_transaction(handle, 1, bh)))
273 goto exit_bh;
274
ac27a0ec
DK
275 if (IS_ERR(it = bclean(handle, sb, block))) {
276 err = PTR_ERR(it);
277 goto exit_bh;
278 }
617ba13b 279 ext4_journal_dirty_metadata(handle, it);
ac27a0ec 280 brelse(it);
617ba13b 281 ext4_set_bit(bit, bh->b_data);
ac27a0ec 282 }
14904107
ES
283
284 if ((err = extend_or_restart_transaction(handle, 2, bh)))
285 goto exit_bh;
286
617ba13b 287 mark_bitmap_end(input->blocks_count, EXT4_BLOCKS_PER_GROUP(sb),
ac27a0ec 288 bh->b_data);
617ba13b 289 ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK
290 brelse(bh);
291
292 /* Mark unused entries in inode bitmap used */
c549a95d 293 ext4_debug("clear inode bitmap %#04llx (+%llu)\n",
ac27a0ec
DK
294 input->inode_bitmap, input->inode_bitmap - start);
295 if (IS_ERR(bh = bclean(handle, sb, input->inode_bitmap))) {
296 err = PTR_ERR(bh);
297 goto exit_journal;
298 }
299
617ba13b 300 mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), EXT4_BLOCKS_PER_GROUP(sb),
ac27a0ec 301 bh->b_data);
617ba13b 302 ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK
303exit_bh:
304 brelse(bh);
305
306exit_journal:
307 unlock_super(sb);
617ba13b 308 if ((err2 = ext4_journal_stop(handle)) && !err)
ac27a0ec
DK
309 err = err2;
310
311 return err;
312}
313
314/*
315 * Iterate through the groups which hold BACKUP superblock/GDT copies in an
617ba13b 316 * ext4 filesystem. The counters should be initialized to 1, 5, and 7 before
ac27a0ec
DK
317 * calling this for the first time. In a sparse filesystem it will be the
318 * sequence of powers of 3, 5, and 7: 1, 3, 5, 7, 9, 25, 27, 49, 81, ...
319 * For a non-sparse filesystem it will be every group: 1, 2, 3, 4, ...
320 */
617ba13b 321static unsigned ext4_list_backups(struct super_block *sb, unsigned *three,
ac27a0ec
DK
322 unsigned *five, unsigned *seven)
323{
324 unsigned *min = three;
325 int mult = 3;
326 unsigned ret;
327
617ba13b
MC
328 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
329 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER)) {
ac27a0ec
DK
330 ret = *min;
331 *min += 1;
332 return ret;
333 }
334
335 if (*five < *min) {
336 min = five;
337 mult = 5;
338 }
339 if (*seven < *min) {
340 min = seven;
341 mult = 7;
342 }
343
344 ret = *min;
345 *min *= mult;
346
347 return ret;
348}
349
350/*
351 * Check that all of the backup GDT blocks are held in the primary GDT block.
352 * It is assumed that they are stored in group order. Returns the number of
353 * groups in current filesystem that have BACKUPS, or -ve error code.
354 */
355static int verify_reserved_gdb(struct super_block *sb,
356 struct buffer_head *primary)
357{
617ba13b 358 const ext4_fsblk_t blk = primary->b_blocknr;
fd2d4291 359 const ext4_group_t end = EXT4_SB(sb)->s_groups_count;
ac27a0ec
DK
360 unsigned three = 1;
361 unsigned five = 5;
362 unsigned seven = 7;
363 unsigned grp;
364 __le32 *p = (__le32 *)primary->b_data;
365 int gdbackups = 0;
366
617ba13b 367 while ((grp = ext4_list_backups(sb, &three, &five, &seven)) < end) {
bd81d8ee
LV
368 if (le32_to_cpu(*p++) !=
369 grp * EXT4_BLOCKS_PER_GROUP(sb) + blk){
46e665e9 370 ext4_warning(sb, __func__,
2ae02107
MC
371 "reserved GDT %llu"
372 " missing grp %d (%llu)",
ac27a0ec 373 blk, grp,
bd81d8ee
LV
374 grp *
375 (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
376 blk);
ac27a0ec
DK
377 return -EINVAL;
378 }
617ba13b 379 if (++gdbackups > EXT4_ADDR_PER_BLOCK(sb))
ac27a0ec
DK
380 return -EFBIG;
381 }
382
383 return gdbackups;
384}
385
386/*
387 * Called when we need to bring a reserved group descriptor table block into
388 * use from the resize inode. The primary copy of the new GDT block currently
389 * is an indirect block (under the double indirect block in the resize inode).
390 * The new backup GDT blocks will be stored as leaf blocks in this indirect
391 * block, in group order. Even though we know all the block numbers we need,
392 * we check to ensure that the resize inode has actually reserved these blocks.
393 *
394 * Don't need to update the block bitmaps because the blocks are still in use.
395 *
396 * We get all of the error cases out of the way, so that we are sure to not
397 * fail once we start modifying the data on disk, because JBD has no rollback.
398 */
399static int add_new_gdb(handle_t *handle, struct inode *inode,
617ba13b 400 struct ext4_new_group_data *input,
ac27a0ec
DK
401 struct buffer_head **primary)
402{
403 struct super_block *sb = inode->i_sb;
617ba13b
MC
404 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
405 unsigned long gdb_num = input->group / EXT4_DESC_PER_BLOCK(sb);
406 ext4_fsblk_t gdblock = EXT4_SB(sb)->s_sbh->b_blocknr + 1 + gdb_num;
ac27a0ec
DK
407 struct buffer_head **o_group_desc, **n_group_desc;
408 struct buffer_head *dind;
409 int gdbackups;
617ba13b 410 struct ext4_iloc iloc;
ac27a0ec
DK
411 __le32 *data;
412 int err;
413
414 if (test_opt(sb, DEBUG))
415 printk(KERN_DEBUG
617ba13b 416 "EXT4-fs: ext4_add_new_gdb: adding group block %lu\n",
ac27a0ec
DK
417 gdb_num);
418
419 /*
420 * If we are not using the primary superblock/GDT copy don't resize,
421 * because the user tools have no way of handling this. Probably a
422 * bad time to do it anyways.
423 */
617ba13b
MC
424 if (EXT4_SB(sb)->s_sbh->b_blocknr !=
425 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) {
46e665e9 426 ext4_warning(sb, __func__,
ac27a0ec 427 "won't resize using backup superblock at %llu",
617ba13b 428 (unsigned long long)EXT4_SB(sb)->s_sbh->b_blocknr);
ac27a0ec
DK
429 return -EPERM;
430 }
431
432 *primary = sb_bread(sb, gdblock);
433 if (!*primary)
434 return -EIO;
435
436 if ((gdbackups = verify_reserved_gdb(sb, *primary)) < 0) {
437 err = gdbackups;
438 goto exit_bh;
439 }
440
617ba13b 441 data = EXT4_I(inode)->i_data + EXT4_DIND_BLOCK;
ac27a0ec
DK
442 dind = sb_bread(sb, le32_to_cpu(*data));
443 if (!dind) {
444 err = -EIO;
445 goto exit_bh;
446 }
447
448 data = (__le32 *)dind->b_data;
617ba13b 449 if (le32_to_cpu(data[gdb_num % EXT4_ADDR_PER_BLOCK(sb)]) != gdblock) {
46e665e9 450 ext4_warning(sb, __func__,
2ae02107 451 "new group %u GDT block %llu not reserved",
ac27a0ec
DK
452 input->group, gdblock);
453 err = -EINVAL;
454 goto exit_dind;
455 }
456
617ba13b 457 if ((err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh)))
ac27a0ec
DK
458 goto exit_dind;
459
617ba13b 460 if ((err = ext4_journal_get_write_access(handle, *primary)))
ac27a0ec
DK
461 goto exit_sbh;
462
617ba13b 463 if ((err = ext4_journal_get_write_access(handle, dind)))
ac27a0ec
DK
464 goto exit_primary;
465
617ba13b
MC
466 /* ext4_reserve_inode_write() gets a reference on the iloc */
467 if ((err = ext4_reserve_inode_write(handle, inode, &iloc)))
ac27a0ec
DK
468 goto exit_dindj;
469
470 n_group_desc = kmalloc((gdb_num + 1) * sizeof(struct buffer_head *),
216553c4 471 GFP_NOFS);
ac27a0ec
DK
472 if (!n_group_desc) {
473 err = -ENOMEM;
46e665e9 474 ext4_warning(sb, __func__,
ac27a0ec
DK
475 "not enough memory for %lu groups", gdb_num + 1);
476 goto exit_inode;
477 }
478
479 /*
480 * Finally, we have all of the possible failures behind us...
481 *
482 * Remove new GDT block from inode double-indirect block and clear out
483 * the new GDT block for use (which also "frees" the backup GDT blocks
484 * from the reserved inode). We don't need to change the bitmaps for
485 * these blocks, because they are marked as in-use from being in the
486 * reserved inode, and will become GDT blocks (primary and backup).
487 */
617ba13b
MC
488 data[gdb_num % EXT4_ADDR_PER_BLOCK(sb)] = 0;
489 ext4_journal_dirty_metadata(handle, dind);
ac27a0ec
DK
490 brelse(dind);
491 inode->i_blocks -= (gdbackups + 1) * sb->s_blocksize >> 9;
617ba13b 492 ext4_mark_iloc_dirty(handle, inode, &iloc);
ac27a0ec 493 memset((*primary)->b_data, 0, sb->s_blocksize);
617ba13b 494 ext4_journal_dirty_metadata(handle, *primary);
ac27a0ec 495
617ba13b 496 o_group_desc = EXT4_SB(sb)->s_group_desc;
ac27a0ec 497 memcpy(n_group_desc, o_group_desc,
617ba13b 498 EXT4_SB(sb)->s_gdb_count * sizeof(struct buffer_head *));
ac27a0ec 499 n_group_desc[gdb_num] = *primary;
617ba13b
MC
500 EXT4_SB(sb)->s_group_desc = n_group_desc;
501 EXT4_SB(sb)->s_gdb_count++;
ac27a0ec
DK
502 kfree(o_group_desc);
503
e8546d06 504 le16_add_cpu(&es->s_reserved_gdt_blocks, -1);
617ba13b 505 ext4_journal_dirty_metadata(handle, EXT4_SB(sb)->s_sbh);
ac27a0ec
DK
506
507 return 0;
508
509exit_inode:
617ba13b 510 //ext4_journal_release_buffer(handle, iloc.bh);
ac27a0ec
DK
511 brelse(iloc.bh);
512exit_dindj:
617ba13b 513 //ext4_journal_release_buffer(handle, dind);
ac27a0ec 514exit_primary:
617ba13b 515 //ext4_journal_release_buffer(handle, *primary);
ac27a0ec 516exit_sbh:
617ba13b 517 //ext4_journal_release_buffer(handle, *primary);
ac27a0ec
DK
518exit_dind:
519 brelse(dind);
520exit_bh:
521 brelse(*primary);
522
617ba13b 523 ext4_debug("leaving with error %d\n", err);
ac27a0ec
DK
524 return err;
525}
526
527/*
528 * Called when we are adding a new group which has a backup copy of each of
529 * the GDT blocks (i.e. sparse group) and there are reserved GDT blocks.
530 * We need to add these reserved backup GDT blocks to the resize inode, so
531 * that they are kept for future resizing and not allocated to files.
532 *
533 * Each reserved backup GDT block will go into a different indirect block.
534 * The indirect blocks are actually the primary reserved GDT blocks,
535 * so we know in advance what their block numbers are. We only get the
536 * double-indirect block to verify it is pointing to the primary reserved
537 * GDT blocks so we don't overwrite a data block by accident. The reserved
538 * backup GDT blocks are stored in their reserved primary GDT block.
539 */
540static int reserve_backup_gdb(handle_t *handle, struct inode *inode,
617ba13b 541 struct ext4_new_group_data *input)
ac27a0ec
DK
542{
543 struct super_block *sb = inode->i_sb;
617ba13b 544 int reserved_gdb =le16_to_cpu(EXT4_SB(sb)->s_es->s_reserved_gdt_blocks);
ac27a0ec
DK
545 struct buffer_head **primary;
546 struct buffer_head *dind;
617ba13b
MC
547 struct ext4_iloc iloc;
548 ext4_fsblk_t blk;
ac27a0ec
DK
549 __le32 *data, *end;
550 int gdbackups = 0;
551 int res, i;
552 int err;
553
216553c4 554 primary = kmalloc(reserved_gdb * sizeof(*primary), GFP_NOFS);
ac27a0ec
DK
555 if (!primary)
556 return -ENOMEM;
557
617ba13b 558 data = EXT4_I(inode)->i_data + EXT4_DIND_BLOCK;
ac27a0ec
DK
559 dind = sb_bread(sb, le32_to_cpu(*data));
560 if (!dind) {
561 err = -EIO;
562 goto exit_free;
563 }
564
617ba13b 565 blk = EXT4_SB(sb)->s_sbh->b_blocknr + 1 + EXT4_SB(sb)->s_gdb_count;
94460093
JB
566 data = (__le32 *)dind->b_data + (EXT4_SB(sb)->s_gdb_count %
567 EXT4_ADDR_PER_BLOCK(sb));
617ba13b 568 end = (__le32 *)dind->b_data + EXT4_ADDR_PER_BLOCK(sb);
ac27a0ec
DK
569
570 /* Get each reserved primary GDT block and verify it holds backups */
571 for (res = 0; res < reserved_gdb; res++, blk++) {
572 if (le32_to_cpu(*data) != blk) {
46e665e9 573 ext4_warning(sb, __func__,
2ae02107 574 "reserved block %llu"
ac27a0ec
DK
575 " not at offset %ld",
576 blk,
577 (long)(data - (__le32 *)dind->b_data));
578 err = -EINVAL;
579 goto exit_bh;
580 }
581 primary[res] = sb_bread(sb, blk);
582 if (!primary[res]) {
583 err = -EIO;
584 goto exit_bh;
585 }
586 if ((gdbackups = verify_reserved_gdb(sb, primary[res])) < 0) {
587 brelse(primary[res]);
588 err = gdbackups;
589 goto exit_bh;
590 }
591 if (++data >= end)
592 data = (__le32 *)dind->b_data;
593 }
594
595 for (i = 0; i < reserved_gdb; i++) {
617ba13b 596 if ((err = ext4_journal_get_write_access(handle, primary[i]))) {
ac27a0ec
DK
597 /*
598 int j;
599 for (j = 0; j < i; j++)
617ba13b 600 ext4_journal_release_buffer(handle, primary[j]);
ac27a0ec
DK
601 */
602 goto exit_bh;
603 }
604 }
605
617ba13b 606 if ((err = ext4_reserve_inode_write(handle, inode, &iloc)))
ac27a0ec
DK
607 goto exit_bh;
608
609 /*
610 * Finally we can add each of the reserved backup GDT blocks from
611 * the new group to its reserved primary GDT block.
612 */
617ba13b 613 blk = input->group * EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK
614 for (i = 0; i < reserved_gdb; i++) {
615 int err2;
616 data = (__le32 *)primary[i]->b_data;
617 /* printk("reserving backup %lu[%u] = %lu\n",
618 primary[i]->b_blocknr, gdbackups,
619 blk + primary[i]->b_blocknr); */
620 data[gdbackups] = cpu_to_le32(blk + primary[i]->b_blocknr);
617ba13b 621 err2 = ext4_journal_dirty_metadata(handle, primary[i]);
ac27a0ec
DK
622 if (!err)
623 err = err2;
624 }
625 inode->i_blocks += reserved_gdb * sb->s_blocksize >> 9;
617ba13b 626 ext4_mark_iloc_dirty(handle, inode, &iloc);
ac27a0ec
DK
627
628exit_bh:
629 while (--res >= 0)
630 brelse(primary[res]);
631 brelse(dind);
632
633exit_free:
634 kfree(primary);
635
636 return err;
637}
638
639/*
617ba13b 640 * Update the backup copies of the ext4 metadata. These don't need to be part
ac27a0ec
DK
641 * of the main resize transaction, because e2fsck will re-write them if there
642 * is a problem (basically only OOM will cause a problem). However, we
643 * _should_ update the backups if possible, in case the primary gets trashed
644 * for some reason and we need to run e2fsck from a backup superblock. The
645 * important part is that the new block and inode counts are in the backup
646 * superblocks, and the location of the new group metadata in the GDT backups.
647 *
648 * We do not need lock_super() for this, because these blocks are not
649 * otherwise touched by the filesystem code when it is mounted. We don't
650 * need to worry about last changing from sbi->s_groups_count, because the
651 * worst that can happen is that we do not copy the full number of backups
652 * at this time. The resize which changed s_groups_count will backup again.
653 */
654static void update_backups(struct super_block *sb,
655 int blk_off, char *data, int size)
656{
617ba13b 657 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 658 const ext4_group_t last = sbi->s_groups_count;
617ba13b 659 const int bpg = EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK
660 unsigned three = 1;
661 unsigned five = 5;
662 unsigned seven = 7;
fd2d4291 663 ext4_group_t group;
ac27a0ec
DK
664 int rest = sb->s_blocksize - size;
665 handle_t *handle;
666 int err = 0, err2;
667
617ba13b 668 handle = ext4_journal_start_sb(sb, EXT4_MAX_TRANS_DATA);
ac27a0ec
DK
669 if (IS_ERR(handle)) {
670 group = 1;
671 err = PTR_ERR(handle);
672 goto exit_err;
673 }
674
617ba13b 675 while ((group = ext4_list_backups(sb, &three, &five, &seven)) < last) {
ac27a0ec
DK
676 struct buffer_head *bh;
677
678 /* Out of journal space, and can't get more - abort - so sad */
679 if (handle->h_buffer_credits == 0 &&
617ba13b
MC
680 ext4_journal_extend(handle, EXT4_MAX_TRANS_DATA) &&
681 (err = ext4_journal_restart(handle, EXT4_MAX_TRANS_DATA)))
ac27a0ec
DK
682 break;
683
684 bh = sb_getblk(sb, group * bpg + blk_off);
685 if (!bh) {
686 err = -EIO;
687 break;
688 }
617ba13b 689 ext4_debug("update metadata backup %#04lx\n",
ac27a0ec 690 (unsigned long)bh->b_blocknr);
617ba13b 691 if ((err = ext4_journal_get_write_access(handle, bh)))
ac27a0ec
DK
692 break;
693 lock_buffer(bh);
694 memcpy(bh->b_data, data, size);
695 if (rest)
696 memset(bh->b_data + size, 0, rest);
697 set_buffer_uptodate(bh);
698 unlock_buffer(bh);
617ba13b 699 ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK
700 brelse(bh);
701 }
617ba13b 702 if ((err2 = ext4_journal_stop(handle)) && !err)
ac27a0ec
DK
703 err = err2;
704
705 /*
706 * Ugh! Need to have e2fsck write the backup copies. It is too
707 * late to revert the resize, we shouldn't fail just because of
708 * the backup copies (they are only needed in case of corruption).
709 *
710 * However, if we got here we have a journal problem too, so we
711 * can't really start a transaction to mark the superblock.
712 * Chicken out and just set the flag on the hope it will be written
713 * to disk, and if not - we will simply wait until next fsck.
714 */
715exit_err:
716 if (err) {
46e665e9 717 ext4_warning(sb, __func__,
fd2d4291 718 "can't update backup for group %lu (err %d), "
ac27a0ec 719 "forcing fsck on next reboot", group, err);
617ba13b
MC
720 sbi->s_mount_state &= ~EXT4_VALID_FS;
721 sbi->s_es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK
722 mark_buffer_dirty(sbi->s_sbh);
723 }
724}
725
726/* Add group descriptor data to an existing or new group descriptor block.
727 * Ensure we handle all possible error conditions _before_ we start modifying
728 * the filesystem, because we cannot abort the transaction and not have it
729 * write the data to disk.
730 *
731 * If we are on a GDT block boundary, we need to get the reserved GDT block.
732 * Otherwise, we may need to add backup GDT blocks for a sparse group.
733 *
734 * We only need to hold the superblock lock while we are actually adding
735 * in the new group's counts to the superblock. Prior to that we have
736 * not really "added" the group at all. We re-check that we are still
737 * adding in the last group in case things have changed since verifying.
738 */
617ba13b 739int ext4_group_add(struct super_block *sb, struct ext4_new_group_data *input)
ac27a0ec 740{
617ba13b
MC
741 struct ext4_sb_info *sbi = EXT4_SB(sb);
742 struct ext4_super_block *es = sbi->s_es;
743 int reserved_gdb = ext4_bg_has_super(sb, input->group) ?
ac27a0ec
DK
744 le16_to_cpu(es->s_reserved_gdt_blocks) : 0;
745 struct buffer_head *primary = NULL;
617ba13b 746 struct ext4_group_desc *gdp;
ac27a0ec
DK
747 struct inode *inode = NULL;
748 handle_t *handle;
749 int gdb_off, gdb_num;
750 int err, err2;
751
617ba13b
MC
752 gdb_num = input->group / EXT4_DESC_PER_BLOCK(sb);
753 gdb_off = input->group % EXT4_DESC_PER_BLOCK(sb);
ac27a0ec 754
617ba13b
MC
755 if (gdb_off == 0 && !EXT4_HAS_RO_COMPAT_FEATURE(sb,
756 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER)) {
46e665e9 757 ext4_warning(sb, __func__,
ac27a0ec
DK
758 "Can't resize non-sparse filesystem further");
759 return -EPERM;
760 }
761
bd81d8ee
LV
762 if (ext4_blocks_count(es) + input->blocks_count <
763 ext4_blocks_count(es)) {
46e665e9 764 ext4_warning(sb, __func__, "blocks_count overflow\n");
ac27a0ec
DK
765 return -EINVAL;
766 }
767
617ba13b 768 if (le32_to_cpu(es->s_inodes_count) + EXT4_INODES_PER_GROUP(sb) <
ac27a0ec 769 le32_to_cpu(es->s_inodes_count)) {
46e665e9 770 ext4_warning(sb, __func__, "inodes_count overflow\n");
ac27a0ec
DK
771 return -EINVAL;
772 }
773
774 if (reserved_gdb || gdb_off == 0) {
617ba13b
MC
775 if (!EXT4_HAS_COMPAT_FEATURE(sb,
776 EXT4_FEATURE_COMPAT_RESIZE_INODE)){
46e665e9 777 ext4_warning(sb, __func__,
ac27a0ec
DK
778 "No reserved GDT blocks, can't resize");
779 return -EPERM;
780 }
1d1fe1ee
DH
781 inode = ext4_iget(sb, EXT4_RESIZE_INO);
782 if (IS_ERR(inode)) {
46e665e9 783 ext4_warning(sb, __func__,
ac27a0ec 784 "Error opening resize inode");
1d1fe1ee 785 return PTR_ERR(inode);
ac27a0ec
DK
786 }
787 }
788
789 if ((err = verify_group_input(sb, input)))
790 goto exit_put;
791
792 if ((err = setup_new_group_blocks(sb, input)))
793 goto exit_put;
794
795 /*
796 * We will always be modifying at least the superblock and a GDT
797 * block. If we are adding a group past the last current GDT block,
798 * we will also modify the inode and the dindirect block. If we
799 * are adding a group with superblock/GDT backups we will also
800 * modify each of the reserved GDT dindirect blocks.
801 */
617ba13b
MC
802 handle = ext4_journal_start_sb(sb,
803 ext4_bg_has_super(sb, input->group) ?
ac27a0ec
DK
804 3 + reserved_gdb : 4);
805 if (IS_ERR(handle)) {
806 err = PTR_ERR(handle);
807 goto exit_put;
808 }
809
810 lock_super(sb);
811 if (input->group != sbi->s_groups_count) {
46e665e9 812 ext4_warning(sb, __func__,
ac27a0ec
DK
813 "multiple resizers run on filesystem!");
814 err = -EBUSY;
815 goto exit_journal;
816 }
817
617ba13b 818 if ((err = ext4_journal_get_write_access(handle, sbi->s_sbh)))
ac27a0ec
DK
819 goto exit_journal;
820
821 /*
822 * We will only either add reserved group blocks to a backup group
823 * or remove reserved blocks for the first group in a new group block.
824 * Doing both would be mean more complex code, and sane people don't
825 * use non-sparse filesystems anymore. This is already checked above.
826 */
827 if (gdb_off) {
828 primary = sbi->s_group_desc[gdb_num];
617ba13b 829 if ((err = ext4_journal_get_write_access(handle, primary)))
ac27a0ec
DK
830 goto exit_journal;
831
617ba13b 832 if (reserved_gdb && ext4_bg_num_gdb(sb, input->group) &&
ac27a0ec
DK
833 (err = reserve_backup_gdb(handle, inode, input)))
834 goto exit_journal;
835 } else if ((err = add_new_gdb(handle, inode, input, &primary)))
836 goto exit_journal;
837
838 /*
839 * OK, now we've set up the new group. Time to make it active.
840 *
841 * Current kernels don't lock all allocations via lock_super(),
842 * so we have to be safe wrt. concurrent accesses the group
843 * data. So we need to be careful to set all of the relevant
844 * group descriptor data etc. *before* we enable the group.
845 *
846 * The key field here is sbi->s_groups_count: as long as
847 * that retains its old value, nobody is going to access the new
848 * group.
849 *
850 * So first we update all the descriptor metadata for the new
851 * group; then we update the total disk blocks count; then we
852 * update the groups count to enable the group; then finally we
853 * update the free space counts so that the system can start
854 * using the new disk blocks.
855 */
856
857 /* Update group descriptor block for new group */
2856922c
FB
858 gdp = (struct ext4_group_desc *)((char *)primary->b_data +
859 gdb_off * EXT4_DESC_SIZE(sb));
ac27a0ec 860
8fadc143
AR
861 ext4_block_bitmap_set(sb, gdp, input->block_bitmap); /* LV FIXME */
862 ext4_inode_bitmap_set(sb, gdp, input->inode_bitmap); /* LV FIXME */
863 ext4_inode_table_set(sb, gdp, input->inode_table); /* LV FIXME */
ac27a0ec 864 gdp->bg_free_blocks_count = cpu_to_le16(input->free_blocks_count);
617ba13b 865 gdp->bg_free_inodes_count = cpu_to_le16(EXT4_INODES_PER_GROUP(sb));
717d50e4 866 gdp->bg_checksum = ext4_group_desc_csum(sbi, input->group, gdp);
ac27a0ec 867
5f21b0e6
FB
868 /*
869 * We can allocate memory for mb_alloc based on the new group
870 * descriptor
871 */
872 if (test_opt(sb, MBALLOC)) {
873 err = ext4_mb_add_more_groupinfo(sb, input->group, gdp);
874 if (err)
875 goto exit_journal;
876 }
ac27a0ec
DK
877 /*
878 * Make the new blocks and inodes valid next. We do this before
879 * increasing the group count so that once the group is enabled,
880 * all of its blocks and inodes are already valid.
881 *
882 * We always allocate group-by-group, then block-by-block or
883 * inode-by-inode within a group, so enabling these
884 * blocks/inodes before the group is live won't actually let us
885 * allocate the new space yet.
886 */
bd81d8ee 887 ext4_blocks_count_set(es, ext4_blocks_count(es) +
ac27a0ec 888 input->blocks_count);
e8546d06 889 le32_add_cpu(&es->s_inodes_count, EXT4_INODES_PER_GROUP(sb));
ac27a0ec
DK
890
891 /*
892 * We need to protect s_groups_count against other CPUs seeing
893 * inconsistent state in the superblock.
894 *
895 * The precise rules we use are:
896 *
897 * * Writers of s_groups_count *must* hold lock_super
898 * AND
899 * * Writers must perform a smp_wmb() after updating all dependent
900 * data and before modifying the groups count
901 *
902 * * Readers must hold lock_super() over the access
903 * OR
904 * * Readers must perform an smp_rmb() after reading the groups count
905 * and before reading any dependent data.
906 *
907 * NB. These rules can be relaxed when checking the group count
908 * while freeing data, as we can only allocate from a block
909 * group after serialising against the group count, and we can
910 * only then free after serialising in turn against that
911 * allocation.
912 */
913 smp_wmb();
914
915 /* Update the global fs size fields */
916 sbi->s_groups_count++;
917
617ba13b 918 ext4_journal_dirty_metadata(handle, primary);
ac27a0ec
DK
919
920 /* Update the reserved block counts only once the new group is
921 * active. */
bd81d8ee 922 ext4_r_blocks_count_set(es, ext4_r_blocks_count(es) +
ac27a0ec
DK
923 input->reserved_blocks);
924
925 /* Update the free space counts */
aa0dff2d 926 percpu_counter_add(&sbi->s_freeblocks_counter,
ac27a0ec 927 input->free_blocks_count);
aa0dff2d 928 percpu_counter_add(&sbi->s_freeinodes_counter,
617ba13b 929 EXT4_INODES_PER_GROUP(sb));
ac27a0ec 930
617ba13b 931 ext4_journal_dirty_metadata(handle, sbi->s_sbh);
ac27a0ec
DK
932 sb->s_dirt = 1;
933
934exit_journal:
935 unlock_super(sb);
617ba13b 936 if ((err2 = ext4_journal_stop(handle)) && !err)
ac27a0ec
DK
937 err = err2;
938 if (!err) {
939 update_backups(sb, sbi->s_sbh->b_blocknr, (char *)es,
617ba13b 940 sizeof(struct ext4_super_block));
ac27a0ec
DK
941 update_backups(sb, primary->b_blocknr, primary->b_data,
942 primary->b_size);
943 }
944exit_put:
945 iput(inode);
946 return err;
617ba13b 947} /* ext4_group_add */
ac27a0ec
DK
948
949/* Extend the filesystem to the new number of blocks specified. This entry
950 * point is only used to extend the current filesystem to the end of the last
951 * existing group. It can be accessed via ioctl, or by "remount,resize=<size>"
952 * for emergencies (because it has no dependencies on reserved blocks).
953 *
617ba13b 954 * If we _really_ wanted, we could use default values to call ext4_group_add()
ac27a0ec
DK
955 * allow the "remount" trick to work for arbitrary resizing, assuming enough
956 * GDT blocks are reserved to grow to the desired size.
957 */
617ba13b
MC
958int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
959 ext4_fsblk_t n_blocks_count)
ac27a0ec 960{
617ba13b 961 ext4_fsblk_t o_blocks_count;
fd2d4291 962 ext4_group_t o_groups_count;
617ba13b
MC
963 ext4_grpblk_t last;
964 ext4_grpblk_t add;
ac27a0ec
DK
965 struct buffer_head * bh;
966 handle_t *handle;
967 int err;
968 unsigned long freed_blocks;
5f21b0e6
FB
969 ext4_group_t group;
970 struct ext4_group_info *grp;
ac27a0ec
DK
971
972 /* We don't need to worry about locking wrt other resizers just
973 * yet: we're going to revalidate es->s_blocks_count after
974 * taking lock_super() below. */
bd81d8ee 975 o_blocks_count = ext4_blocks_count(es);
617ba13b 976 o_groups_count = EXT4_SB(sb)->s_groups_count;
ac27a0ec
DK
977
978 if (test_opt(sb, DEBUG))
2ae02107 979 printk(KERN_DEBUG "EXT4-fs: extending last group from %llu uto %llu blocks\n",
ac27a0ec
DK
980 o_blocks_count, n_blocks_count);
981
982 if (n_blocks_count == 0 || n_blocks_count == o_blocks_count)
983 return 0;
984
985 if (n_blocks_count > (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 986 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2ae02107 987 " too large to resize to %llu blocks safely\n",
ac27a0ec
DK
988 sb->s_id, n_blocks_count);
989 if (sizeof(sector_t) < 8)
46e665e9 990 ext4_warning(sb, __func__,
ac27a0ec
DK
991 "CONFIG_LBD not enabled\n");
992 return -EINVAL;
993 }
994
995 if (n_blocks_count < o_blocks_count) {
46e665e9 996 ext4_warning(sb, __func__,
ac27a0ec
DK
997 "can't shrink FS - resize aborted");
998 return -EBUSY;
999 }
1000
1001 /* Handle the remaining blocks in the last group only. */
5f21b0e6 1002 ext4_get_group_no_and_offset(sb, o_blocks_count, &group, &last);
ac27a0ec
DK
1003
1004 if (last == 0) {
46e665e9 1005 ext4_warning(sb, __func__,
ac27a0ec
DK
1006 "need to use ext2online to resize further");
1007 return -EPERM;
1008 }
1009
617ba13b 1010 add = EXT4_BLOCKS_PER_GROUP(sb) - last;
ac27a0ec
DK
1011
1012 if (o_blocks_count + add < o_blocks_count) {
46e665e9 1013 ext4_warning(sb, __func__, "blocks_count overflow");
ac27a0ec
DK
1014 return -EINVAL;
1015 }
1016
1017 if (o_blocks_count + add > n_blocks_count)
1018 add = n_blocks_count - o_blocks_count;
1019
1020 if (o_blocks_count + add < n_blocks_count)
46e665e9 1021 ext4_warning(sb, __func__,
2ae02107 1022 "will only finish group (%llu"
ac27a0ec
DK
1023 " blocks, %u new)",
1024 o_blocks_count + add, add);
1025
1026 /* See if the device is actually as big as what was requested */
1027 bh = sb_bread(sb, o_blocks_count + add -1);
1028 if (!bh) {
46e665e9 1029 ext4_warning(sb, __func__,
ac27a0ec
DK
1030 "can't read last block, resize aborted");
1031 return -ENOSPC;
1032 }
1033 brelse(bh);
1034
1035 /* We will update the superblock, one block bitmap, and
617ba13b 1036 * one group descriptor via ext4_free_blocks().
ac27a0ec 1037 */
617ba13b 1038 handle = ext4_journal_start_sb(sb, 3);
ac27a0ec
DK
1039 if (IS_ERR(handle)) {
1040 err = PTR_ERR(handle);
46e665e9 1041 ext4_warning(sb, __func__, "error %d on journal start", err);
ac27a0ec
DK
1042 goto exit_put;
1043 }
1044
1045 lock_super(sb);
bd81d8ee 1046 if (o_blocks_count != ext4_blocks_count(es)) {
46e665e9 1047 ext4_warning(sb, __func__,
ac27a0ec
DK
1048 "multiple resizers run on filesystem!");
1049 unlock_super(sb);
5606bf5d 1050 ext4_journal_stop(handle);
ac27a0ec
DK
1051 err = -EBUSY;
1052 goto exit_put;
1053 }
1054
617ba13b
MC
1055 if ((err = ext4_journal_get_write_access(handle,
1056 EXT4_SB(sb)->s_sbh))) {
46e665e9 1057 ext4_warning(sb, __func__,
ac27a0ec
DK
1058 "error %d on journal write access", err);
1059 unlock_super(sb);
617ba13b 1060 ext4_journal_stop(handle);
ac27a0ec
DK
1061 goto exit_put;
1062 }
bd81d8ee 1063 ext4_blocks_count_set(es, o_blocks_count + add);
617ba13b 1064 ext4_journal_dirty_metadata(handle, EXT4_SB(sb)->s_sbh);
ac27a0ec
DK
1065 sb->s_dirt = 1;
1066 unlock_super(sb);
c549a95d 1067 ext4_debug("freeing blocks %llu through %llu\n", o_blocks_count,
ac27a0ec 1068 o_blocks_count + add);
617ba13b 1069 ext4_free_blocks_sb(handle, sb, o_blocks_count, add, &freed_blocks);
2ae02107 1070 ext4_debug("freed blocks %llu through %llu\n", o_blocks_count,
ac27a0ec 1071 o_blocks_count + add);
617ba13b 1072 if ((err = ext4_journal_stop(handle)))
ac27a0ec 1073 goto exit_put;
5f21b0e6
FB
1074
1075 /*
1076 * Mark mballoc pages as not up to date so that they will be updated
1077 * next time they are loaded by ext4_mb_load_buddy.
1078 */
1079 if (test_opt(sb, MBALLOC)) {
1080 struct ext4_sb_info *sbi = EXT4_SB(sb);
1081 struct inode *inode = sbi->s_buddy_cache;
1082 int blocks_per_page;
1083 int block;
1084 int pnum;
1085 struct page *page;
1086
1087 /* Set buddy page as not up to date */
1088 blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize;
1089 block = group * 2;
1090 pnum = block / blocks_per_page;
1091 page = find_get_page(inode->i_mapping, pnum);
1092 if (page != NULL) {
1093 ClearPageUptodate(page);
1094 page_cache_release(page);
1095 }
1096
1097 /* Set bitmap page as not up to date */
1098 block++;
1099 pnum = block / blocks_per_page;
1100 page = find_get_page(inode->i_mapping, pnum);
1101 if (page != NULL) {
1102 ClearPageUptodate(page);
1103 page_cache_release(page);
1104 }
1105
1106 /* Get the info on the last group */
1107 grp = ext4_get_group_info(sb, group);
1108
1109 /* Update free blocks in group info */
1110 ext4_mb_update_group_info(grp, add);
1111 }
1112
ac27a0ec 1113 if (test_opt(sb, DEBUG))
bd81d8ee
LV
1114 printk(KERN_DEBUG "EXT4-fs: extended group to %llu blocks\n",
1115 ext4_blocks_count(es));
617ba13b
MC
1116 update_backups(sb, EXT4_SB(sb)->s_sbh->b_blocknr, (char *)es,
1117 sizeof(struct ext4_super_block));
ac27a0ec
DK
1118exit_put:
1119 return err;
617ba13b 1120} /* ext4_group_extend */