2 * Copyright (C) International Business Machines Corp., 2000-2004
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
12 * the GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 * note: file system in transition to aggregate/fileset:
24 * file system mount is interpreted as the mount of aggregate,
25 * if not already mounted, and mount of the single/only fileset in
28 * a file system/aggregate is represented by an internal inode
29 * (aka mount inode) initialized with aggregate superblock;
30 * each vfs represents a fileset, and points to its "fileset inode
31 * allocation map inode" (aka fileset inode):
32 * (an aggregate itself is structured recursively as a filset:
33 * an internal vfs is constructed and points to its "fileset inode
34 * allocation map inode" (aka aggregate inode) where each inode
35 * represents a fileset inode) so that inode number is mapped to
36 * on-disk inode in uniform way at both aggregate and fileset level;
38 * each vnode/inode of a fileset is linked to its vfs (to facilitate
39 * per fileset inode operations, e.g., unmount of a fileset, etc.);
40 * each inode points to the mount inode (to facilitate access to
41 * per aggregate information, e.g., block size, etc.) as well as
46 * mntvfs -> fileset ipimap+ -> aggregate ipbmap -> aggregate ipaimap;
47 * fileset vfs -> vp(1) <-> ... <-> vp(n) <->vproot;
51 #include <linux/buffer_head.h>
52 #include <linux/log2.h>
54 #include "jfs_incore.h"
55 #include "jfs_filsys.h"
56 #include "jfs_superblock.h"
59 #include "jfs_metapage.h"
60 #include "jfs_debug.h"
66 static int chkSuper(struct super_block *);
67 static int logMOUNT(struct super_block *sb);
72 * FUNCTION: vfs_mount()
74 * PARAMETER: sb - super block
76 * RETURN: -EBUSY - device already mounted or open for write
77 * -EBUSY - cvrdvp already mounted;
78 * -EBUSY - mount table full
79 * -ENOTDIR- cvrdvp not directory on a device mount
80 * -ENXIO - device open failure
82 int jfs_mount(struct super_block *sb)
84 int rc = 0; /* Return code */
85 struct jfs_sb_info *sbi = JFS_SBI(sb);
86 struct inode *ipaimap = NULL;
87 struct inode *ipaimap2 = NULL;
88 struct inode *ipimap = NULL;
89 struct inode *ipbmap = NULL;
92 * read/validate superblock
93 * (initialize mount inode from the superblock)
95 if ((rc = chkSuper(sb))) {
99 ipaimap = diReadSpecial(sb, AGGREGATE_I, 0);
100 if (ipaimap == NULL) {
101 jfs_err("jfs_mount: Failed to read AGGREGATE_I");
105 sbi->ipaimap = ipaimap;
107 jfs_info("jfs_mount: ipaimap:0x%p", ipaimap);
110 * initialize aggregate inode allocation map
112 if ((rc = diMount(ipaimap))) {
113 jfs_err("jfs_mount: diMount(ipaimap) failed w/rc = %d", rc);
118 * open aggregate block allocation map
120 ipbmap = diReadSpecial(sb, BMAP_I, 0);
121 if (ipbmap == NULL) {
126 jfs_info("jfs_mount: ipbmap:0x%p", ipbmap);
128 sbi->ipbmap = ipbmap;
131 * initialize aggregate block allocation map
133 if ((rc = dbMount(ipbmap))) {
134 jfs_err("jfs_mount: dbMount failed w/rc = %d", rc);
139 * open the secondary aggregate inode allocation map
141 * This is a duplicate of the aggregate inode allocation map.
143 * hand craft a vfs in the same fashion as we did to read ipaimap.
144 * By adding INOSPEREXT (32) to the inode number, we are telling
145 * diReadSpecial that we are reading from the secondary aggregate
146 * inode table. This also creates a unique entry in the inode hash
149 if ((sbi->mntflag & JFS_BAD_SAIT) == 0) {
150 ipaimap2 = diReadSpecial(sb, AGGREGATE_I, 1);
152 jfs_err("jfs_mount: Failed to read AGGREGATE_I");
156 sbi->ipaimap2 = ipaimap2;
158 jfs_info("jfs_mount: ipaimap2:0x%p", ipaimap2);
161 * initialize secondary aggregate inode allocation map
163 if ((rc = diMount(ipaimap2))) {
164 jfs_err("jfs_mount: diMount(ipaimap2) failed, rc = %d",
169 /* Secondary aggregate inode table is not valid */
170 sbi->ipaimap2 = NULL;
173 * mount (the only/single) fileset
176 * open fileset inode allocation map (aka fileset inode)
178 ipimap = diReadSpecial(sb, FILESYSTEM_I, 0);
179 if (ipimap == NULL) {
180 jfs_err("jfs_mount: Failed to read FILESYSTEM_I");
181 /* open fileset secondary inode allocation map */
185 jfs_info("jfs_mount: ipimap:0x%p", ipimap);
187 /* map further access of per fileset inodes by the fileset inode */
188 sbi->ipimap = ipimap;
190 /* initialize fileset inode allocation map */
191 if ((rc = diMount(ipimap))) {
192 jfs_err("jfs_mount: diMount failed w/rc = %d", rc);
201 errout41: /* close fileset inode allocation map inode */
202 diFreeSpecial(ipimap);
204 errout40: /* fileset closed */
206 /* close secondary aggregate inode allocation map */
208 diUnmount(ipaimap2, 1);
209 diFreeSpecial(ipaimap2);
214 /* close aggregate block allocation map */
215 dbUnmount(ipbmap, 1);
216 diFreeSpecial(ipbmap);
218 errout22: /* close aggregate inode allocation map */
220 diUnmount(ipaimap, 1);
222 errout21: /* close aggregate inodes */
223 diFreeSpecial(ipaimap);
224 errout20: /* aggregate closed */
229 jfs_err("Mount JFS Failure: %d", rc);
235 * NAME: jfs_mount_rw(sb, remount)
237 * FUNCTION: Completes read-write mount, or remounts read-only volume
240 int jfs_mount_rw(struct super_block *sb, int remount)
242 struct jfs_sb_info *sbi = JFS_SBI(sb);
246 * If we are re-mounting a previously read-only volume, we want to
247 * re-read the inode and block maps, since fsck.jfs may have updated
251 if (chkSuper(sb) || (sbi->state != FM_CLEAN))
254 truncate_inode_pages(sbi->ipimap->i_mapping, 0);
255 truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
256 diUnmount(sbi->ipimap, 1);
257 if ((rc = diMount(sbi->ipimap))) {
258 jfs_err("jfs_mount_rw: diMount failed!");
262 dbUnmount(sbi->ipbmap, 1);
263 if ((rc = dbMount(sbi->ipbmap))) {
264 jfs_err("jfs_mount_rw: dbMount failed!");
270 * open/initialize log
272 if ((rc = lmLogOpen(sb)))
276 * update file system superblock;
278 if ((rc = updateSuper(sb, FM_MOUNT))) {
279 jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
285 * write MOUNT log record of the file system
295 * validate the superblock of the file system to be mounted and
296 * get the file system parameters.
299 * 0 with fragsize set if check successful
300 * error code if not successful
302 static int chkSuper(struct super_block *sb)
305 struct jfs_sb_info *sbi = JFS_SBI(sb);
306 struct jfs_superblock *j_sb;
307 struct buffer_head *bh;
308 int AIM_bytesize, AIT_bytesize;
309 int expected_AIM_bytesize, expected_AIT_bytesize;
310 s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
311 s64 byte_addr_diff0, byte_addr_diff1;
314 if ((rc = readSuper(sb, &bh)))
316 j_sb = (struct jfs_superblock *)bh->b_data;
319 * validate superblock
321 /* validate fs signature */
322 if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
323 le32_to_cpu(j_sb->s_version) > JFS_VERSION) {
328 bsize = le32_to_cpu(j_sb->s_bsize);
330 if (bsize != PSIZE) {
331 jfs_err("Currently only 4K block size supported!");
337 jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
338 le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
339 (unsigned long long) le64_to_cpu(j_sb->s_size));
341 /* validate the descriptors for Secondary AIM and AIT */
342 if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
343 cpu_to_le32(JFS_BAD_SAIT)) {
344 expected_AIM_bytesize = 2 * PSIZE;
345 AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
346 expected_AIT_bytesize = 4 * PSIZE;
347 AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
348 AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
349 AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
350 byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
351 fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
352 byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
353 if ((AIM_bytesize != expected_AIM_bytesize) ||
354 (AIT_bytesize != expected_AIT_bytesize) ||
355 (byte_addr_diff0 != AIM_bytesize) ||
356 (byte_addr_diff1 <= AIT_bytesize))
357 j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
360 if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
361 cpu_to_le32(JFS_GROUPCOMMIT))
362 j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
364 /* validate fs state */
365 if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
367 jfs_err("jfs_mount: Mount Failure: File System Dirty.");
372 sbi->state = le32_to_cpu(j_sb->s_state);
373 sbi->mntflag = le32_to_cpu(j_sb->s_flag);
376 * JFS always does I/O by 4K pages. Don't tell the buffer cache
377 * that we use anything else (leave s_blocksize alone).
380 sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
382 /* check some fields for possible corruption */
383 if (sbi->l2bsize != ilog2((u32)bsize) ||
385 le32_to_cpu(j_sb->s_state) > FM_STATE_MAX) {
387 jfs_err("jfs_mount: Mount Failure: superblock is corrupt!");
392 * For now, ignore s_pbsize, l2bfactor. All I/O going through buffer
395 sbi->nbperpage = PSIZE >> sbi->l2bsize;
396 sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
397 sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
398 if (sbi->mntflag & JFS_INLINELOG)
399 sbi->logpxd = j_sb->s_logpxd;
401 sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev));
402 memcpy(sbi->uuid, j_sb->s_uuid, sizeof(sbi->uuid));
403 memcpy(sbi->loguuid, j_sb->s_loguuid, sizeof(sbi->uuid));
405 sbi->fsckpxd = j_sb->s_fsckpxd;
406 sbi->ait2 = j_sb->s_ait2;
417 * update synchronously superblock if it is mounted read-write.
419 int updateSuper(struct super_block *sb, uint state)
421 struct jfs_superblock *j_sb;
422 struct jfs_sb_info *sbi = JFS_SBI(sb);
423 struct buffer_head *bh;
426 if (sbi->flag & JFS_NOINTEGRITY) {
427 if (state == FM_DIRTY) {
428 sbi->p_state = state;
430 } else if (state == FM_MOUNT) {
431 sbi->p_state = sbi->state;
433 } else if (state == FM_CLEAN) {
434 state = sbi->p_state;
436 jfs_err("updateSuper: bad state");
437 } else if (sbi->state == FM_DIRTY)
440 if ((rc = readSuper(sb, &bh)))
443 j_sb = (struct jfs_superblock *)bh->b_data;
445 j_sb->s_state = cpu_to_le32(state);
448 if (state == FM_MOUNT) {
449 /* record log's dev_t and mount serial number */
450 j_sb->s_logdev = cpu_to_le32(new_encode_dev(sbi->log->bdev->bd_dev));
451 j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
452 } else if (state == FM_CLEAN) {
454 * If this volume is shared with OS/2, OS/2 will need to
455 * recalculate DASD usage, since we don't deal with it.
457 if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
458 j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
461 mark_buffer_dirty(bh);
462 sync_dirty_buffer(bh);
472 * read superblock by raw sector address
474 int readSuper(struct super_block *sb, struct buffer_head **bpp)
476 /* read in primary superblock */
477 *bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits);
481 /* read in secondary/replicated superblock */
482 *bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits);
493 * function: write a MOUNT log record for file system.
495 * MOUNT record keeps logredo() from processing log records
496 * for this file system past this point in log.
497 * it is harmless if mount fails.
499 * note: MOUNT record is at aggregate level, not at fileset level,
500 * since log records of previous mounts of a fileset
501 * (e.g., AFTER record of extent allocation) have to be processed
502 * to update block allocation map at aggregate level.
504 static int logMOUNT(struct super_block *sb)
506 struct jfs_log *log = JFS_SBI(sb)->log;
511 lrd.type = cpu_to_le16(LOG_MOUNT);
513 lrd.aggregate = cpu_to_le32(new_encode_dev(sb->s_bdev->bd_dev));
514 lmLog(log, NULL, &lrd, NULL);