blob: 9895595fd2f24c5e3128c0138c554430a73e48d8 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Copyright (C) International Business Machines Corp., 2000-2004
3 *
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
Dave Kleikamp63f83c92006-10-02 09:55:27 -05006 * the Free Software Foundation; either version 2 of the License, or
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 * (at your option) any later version.
Dave Kleikamp63f83c92006-10-02 09:55:27 -05008 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * 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.
13 *
14 * You should have received a copy of the GNU General Public License
Dave Kleikamp63f83c92006-10-02 09:55:27 -050015 * along with this program; if not, write to the Free Software
Linus Torvalds1da177e2005-04-16 15:20:36 -070016 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17 */
18
19/*
20 * Module: jfs_mount.c
21 *
22 * note: file system in transition to aggregate/fileset:
23 *
Dave Kleikamp63f83c92006-10-02 09:55:27 -050024 * file system mount is interpreted as the mount of aggregate,
25 * if not already mounted, and mount of the single/only fileset in
Linus Torvalds1da177e2005-04-16 15:20:36 -070026 * the aggregate;
27 *
28 * a file system/aggregate is represented by an internal inode
29 * (aka mount inode) initialized with aggregate superblock;
Dave Kleikamp63f83c92006-10-02 09:55:27 -050030 * each vfs represents a fileset, and points to its "fileset inode
Linus Torvalds1da177e2005-04-16 15:20:36 -070031 * allocation map inode" (aka fileset inode):
Dave Kleikamp63f83c92006-10-02 09:55:27 -050032 * (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
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 * on-disk inode in uniform way at both aggregate and fileset level;
37 *
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
42 * its file set inode.
43 *
Dave Kleikamp63f83c92006-10-02 09:55:27 -050044 * aggregate
Linus Torvalds1da177e2005-04-16 15:20:36 -070045 * ipmnt
46 * mntvfs -> fileset ipimap+ -> aggregate ipbmap -> aggregate ipaimap;
47 * fileset vfs -> vp(1) <-> ... <-> vp(n) <->vproot;
48 */
49
50#include <linux/fs.h>
51#include <linux/buffer_head.h>
52
53#include "jfs_incore.h"
54#include "jfs_filsys.h"
55#include "jfs_superblock.h"
56#include "jfs_dmap.h"
57#include "jfs_imap.h"
58#include "jfs_metapage.h"
59#include "jfs_debug.h"
60
61
62/*
63 * forward references
64 */
65static int chkSuper(struct super_block *);
66static int logMOUNT(struct super_block *sb);
67
68/*
69 * NAME: jfs_mount(sb)
70 *
71 * FUNCTION: vfs_mount()
72 *
73 * PARAMETER: sb - super block
74 *
75 * RETURN: -EBUSY - device already mounted or open for write
76 * -EBUSY - cvrdvp already mounted;
77 * -EBUSY - mount table full
78 * -ENOTDIR- cvrdvp not directory on a device mount
79 * -ENXIO - device open failure
80 */
81int jfs_mount(struct super_block *sb)
82{
Dave Kleikampf720e3b2007-06-06 15:28:35 -050083 int rc = 0; /* Return code */
Linus Torvalds1da177e2005-04-16 15:20:36 -070084 struct jfs_sb_info *sbi = JFS_SBI(sb);
85 struct inode *ipaimap = NULL;
86 struct inode *ipaimap2 = NULL;
87 struct inode *ipimap = NULL;
88 struct inode *ipbmap = NULL;
89
90 /*
Dave Kleikamp63f83c92006-10-02 09:55:27 -050091 * read/validate superblock
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 * (initialize mount inode from the superblock)
93 */
94 if ((rc = chkSuper(sb))) {
95 goto errout20;
96 }
97
98 ipaimap = diReadSpecial(sb, AGGREGATE_I, 0);
99 if (ipaimap == NULL) {
Paul Bolle426d3102010-08-07 12:30:03 +0200100 jfs_err("jfs_mount: Failed to read AGGREGATE_I");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 rc = -EIO;
102 goto errout20;
103 }
104 sbi->ipaimap = ipaimap;
105
106 jfs_info("jfs_mount: ipaimap:0x%p", ipaimap);
107
108 /*
109 * initialize aggregate inode allocation map
110 */
111 if ((rc = diMount(ipaimap))) {
112 jfs_err("jfs_mount: diMount(ipaimap) failed w/rc = %d", rc);
113 goto errout21;
114 }
115
116 /*
117 * open aggregate block allocation map
118 */
119 ipbmap = diReadSpecial(sb, BMAP_I, 0);
120 if (ipbmap == NULL) {
121 rc = -EIO;
122 goto errout22;
123 }
124
125 jfs_info("jfs_mount: ipbmap:0x%p", ipbmap);
126
127 sbi->ipbmap = ipbmap;
128
129 /*
130 * initialize aggregate block allocation map
131 */
132 if ((rc = dbMount(ipbmap))) {
133 jfs_err("jfs_mount: dbMount failed w/rc = %d", rc);
134 goto errout22;
135 }
136
137 /*
138 * open the secondary aggregate inode allocation map
139 *
140 * This is a duplicate of the aggregate inode allocation map.
141 *
142 * hand craft a vfs in the same fashion as we did to read ipaimap.
143 * By adding INOSPEREXT (32) to the inode number, we are telling
144 * diReadSpecial that we are reading from the secondary aggregate
145 * inode table. This also creates a unique entry in the inode hash
146 * table.
147 */
148 if ((sbi->mntflag & JFS_BAD_SAIT) == 0) {
149 ipaimap2 = diReadSpecial(sb, AGGREGATE_I, 1);
Joe Perches09aaa742007-11-13 22:16:08 -0600150 if (!ipaimap2) {
Paul Bolle426d3102010-08-07 12:30:03 +0200151 jfs_err("jfs_mount: Failed to read AGGREGATE_I");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 rc = -EIO;
153 goto errout35;
154 }
155 sbi->ipaimap2 = ipaimap2;
156
157 jfs_info("jfs_mount: ipaimap2:0x%p", ipaimap2);
158
159 /*
160 * initialize secondary aggregate inode allocation map
161 */
162 if ((rc = diMount(ipaimap2))) {
163 jfs_err("jfs_mount: diMount(ipaimap2) failed, rc = %d",
164 rc);
165 goto errout35;
166 }
167 } else
168 /* Secondary aggregate inode table is not valid */
169 sbi->ipaimap2 = NULL;
170
171 /*
Dave Kleikampf720e3b2007-06-06 15:28:35 -0500172 * mount (the only/single) fileset
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 */
174 /*
175 * open fileset inode allocation map (aka fileset inode)
176 */
177 ipimap = diReadSpecial(sb, FILESYSTEM_I, 0);
178 if (ipimap == NULL) {
179 jfs_err("jfs_mount: Failed to read FILESYSTEM_I");
180 /* open fileset secondary inode allocation map */
181 rc = -EIO;
182 goto errout40;
183 }
184 jfs_info("jfs_mount: ipimap:0x%p", ipimap);
185
186 /* map further access of per fileset inodes by the fileset inode */
187 sbi->ipimap = ipimap;
188
189 /* initialize fileset inode allocation map */
190 if ((rc = diMount(ipimap))) {
191 jfs_err("jfs_mount: diMount failed w/rc = %d", rc);
192 goto errout41;
193 }
194
195 goto out;
196
197 /*
Dave Kleikampf720e3b2007-06-06 15:28:35 -0500198 * unwind on error
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 */
200 errout41: /* close fileset inode allocation map inode */
201 diFreeSpecial(ipimap);
202
203 errout40: /* fileset closed */
204
205 /* close secondary aggregate inode allocation map */
206 if (ipaimap2) {
207 diUnmount(ipaimap2, 1);
208 diFreeSpecial(ipaimap2);
209 }
210
211 errout35:
212
213 /* close aggregate block allocation map */
214 dbUnmount(ipbmap, 1);
215 diFreeSpecial(ipbmap);
216
217 errout22: /* close aggregate inode allocation map */
218
219 diUnmount(ipaimap, 1);
220
221 errout21: /* close aggregate inodes */
222 diFreeSpecial(ipaimap);
223 errout20: /* aggregate closed */
224
225 out:
226
227 if (rc)
228 jfs_err("Mount JFS Failure: %d", rc);
229
230 return rc;
231}
232
233/*
234 * NAME: jfs_mount_rw(sb, remount)
235 *
236 * FUNCTION: Completes read-write mount, or remounts read-only volume
237 * as read-write
238 */
239int jfs_mount_rw(struct super_block *sb, int remount)
240{
Dave Kleikamp63f83c92006-10-02 09:55:27 -0500241 struct jfs_sb_info *sbi = JFS_SBI(sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 int rc;
243
244 /*
245 * If we are re-mounting a previously read-only volume, we want to
246 * re-read the inode and block maps, since fsck.jfs may have updated
247 * them.
248 */
249 if (remount) {
250 if (chkSuper(sb) || (sbi->state != FM_CLEAN))
251 return -EINVAL;
252
253 truncate_inode_pages(sbi->ipimap->i_mapping, 0);
254 truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
255 diUnmount(sbi->ipimap, 1);
256 if ((rc = diMount(sbi->ipimap))) {
257 jfs_err("jfs_mount_rw: diMount failed!");
258 return rc;
259 }
260
261 dbUnmount(sbi->ipbmap, 1);
262 if ((rc = dbMount(sbi->ipbmap))) {
263 jfs_err("jfs_mount_rw: dbMount failed!");
264 return rc;
265 }
266 }
267
268 /*
269 * open/initialize log
270 */
271 if ((rc = lmLogOpen(sb)))
272 return rc;
273
274 /*
275 * update file system superblock;
276 */
277 if ((rc = updateSuper(sb, FM_MOUNT))) {
278 jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
279 lmLogClose(sb);
280 return rc;
281 }
282
283 /*
284 * write MOUNT log record of the file system
285 */
286 logMOUNT(sb);
287
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 return rc;
289}
290
291/*
292 * chkSuper()
293 *
Dave Kleikamp63f83c92006-10-02 09:55:27 -0500294 * validate the superblock of the file system to be mounted and
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 * get the file system parameters.
296 *
297 * returns
298 * 0 with fragsize set if check successful
299 * error code if not successful
300 */
301static int chkSuper(struct super_block *sb)
302{
303 int rc = 0;
304 struct jfs_sb_info *sbi = JFS_SBI(sb);
305 struct jfs_superblock *j_sb;
306 struct buffer_head *bh;
307 int AIM_bytesize, AIT_bytesize;
308 int expected_AIM_bytesize, expected_AIT_bytesize;
309 s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
310 s64 byte_addr_diff0, byte_addr_diff1;
311 s32 bsize;
312
313 if ((rc = readSuper(sb, &bh)))
314 return rc;
315 j_sb = (struct jfs_superblock *)bh->b_data;
316
317 /*
318 * validate superblock
319 */
320 /* validate fs signature */
321 if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
322 le32_to_cpu(j_sb->s_version) > JFS_VERSION) {
323 rc = -EINVAL;
324 goto out;
325 }
326
327 bsize = le32_to_cpu(j_sb->s_bsize);
328#ifdef _JFS_4K
329 if (bsize != PSIZE) {
330 jfs_err("Currently only 4K block size supported!");
331 rc = -EINVAL;
332 goto out;
333 }
334#endif /* _JFS_4K */
335
336 jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
337 le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
338 (unsigned long long) le64_to_cpu(j_sb->s_size));
339
340 /* validate the descriptors for Secondary AIM and AIT */
341 if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
342 cpu_to_le32(JFS_BAD_SAIT)) {
343 expected_AIM_bytesize = 2 * PSIZE;
344 AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
345 expected_AIT_bytesize = 4 * PSIZE;
346 AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
347 AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
348 AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
349 byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
350 fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
351 byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
352 if ((AIM_bytesize != expected_AIM_bytesize) ||
353 (AIT_bytesize != expected_AIT_bytesize) ||
354 (byte_addr_diff0 != AIM_bytesize) ||
355 (byte_addr_diff1 <= AIT_bytesize))
356 j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
357 }
358
359 if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
360 cpu_to_le32(JFS_GROUPCOMMIT))
361 j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
362
363 /* validate fs state */
364 if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
365 !(sb->s_flags & MS_RDONLY)) {
366 jfs_err("jfs_mount: Mount Failure: File System Dirty.");
367 rc = -EINVAL;
368 goto out;
369 }
370
371 sbi->state = le32_to_cpu(j_sb->s_state);
372 sbi->mntflag = le32_to_cpu(j_sb->s_flag);
373
374 /*
375 * JFS always does I/O by 4K pages. Don't tell the buffer cache
376 * that we use anything else (leave s_blocksize alone).
377 */
378 sbi->bsize = bsize;
379 sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
380
381 /*
382 * For now, ignore s_pbsize, l2bfactor. All I/O going through buffer
383 * cache.
384 */
385 sbi->nbperpage = PSIZE >> sbi->l2bsize;
386 sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
387 sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
388 if (sbi->mntflag & JFS_INLINELOG)
389 sbi->logpxd = j_sb->s_logpxd;
390 else {
391 sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev));
392 memcpy(sbi->uuid, j_sb->s_uuid, sizeof(sbi->uuid));
393 memcpy(sbi->loguuid, j_sb->s_loguuid, sizeof(sbi->uuid));
394 }
395 sbi->fsckpxd = j_sb->s_fsckpxd;
396 sbi->ait2 = j_sb->s_ait2;
397
398 out:
399 brelse(bh);
400 return rc;
401}
402
403
404/*
405 * updateSuper()
406 *
407 * update synchronously superblock if it is mounted read-write.
408 */
409int updateSuper(struct super_block *sb, uint state)
410{
411 struct jfs_superblock *j_sb;
412 struct jfs_sb_info *sbi = JFS_SBI(sb);
413 struct buffer_head *bh;
414 int rc;
415
416 if (sbi->flag & JFS_NOINTEGRITY) {
417 if (state == FM_DIRTY) {
418 sbi->p_state = state;
419 return 0;
420 } else if (state == FM_MOUNT) {
421 sbi->p_state = sbi->state;
422 state = FM_DIRTY;
423 } else if (state == FM_CLEAN) {
424 state = sbi->p_state;
425 } else
426 jfs_err("updateSuper: bad state");
427 } else if (sbi->state == FM_DIRTY)
428 return 0;
Dave Kleikamp63f83c92006-10-02 09:55:27 -0500429
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 if ((rc = readSuper(sb, &bh)))
431 return rc;
432
433 j_sb = (struct jfs_superblock *)bh->b_data;
434
435 j_sb->s_state = cpu_to_le32(state);
436 sbi->state = state;
437
438 if (state == FM_MOUNT) {
439 /* record log's dev_t and mount serial number */
440 j_sb->s_logdev = cpu_to_le32(new_encode_dev(sbi->log->bdev->bd_dev));
441 j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
442 } else if (state == FM_CLEAN) {
443 /*
444 * If this volume is shared with OS/2, OS/2 will need to
445 * recalculate DASD usage, since we don't deal with it.
446 */
447 if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
448 j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
449 }
450
451 mark_buffer_dirty(bh);
452 sync_dirty_buffer(bh);
453 brelse(bh);
454
455 return 0;
456}
457
458
459/*
460 * readSuper()
461 *
462 * read superblock by raw sector address
463 */
464int readSuper(struct super_block *sb, struct buffer_head **bpp)
465{
466 /* read in primary superblock */
467 *bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits);
468 if (*bpp)
469 return 0;
470
471 /* read in secondary/replicated superblock */
472 *bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits);
473 if (*bpp)
474 return 0;
475
476 return -EIO;
477}
478
479
480/*
481 * logMOUNT()
482 *
483 * function: write a MOUNT log record for file system.
484 *
485 * MOUNT record keeps logredo() from processing log records
486 * for this file system past this point in log.
487 * it is harmless if mount fails.
488 *
Dave Kleikamp63f83c92006-10-02 09:55:27 -0500489 * note: MOUNT record is at aggregate level, not at fileset level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 * since log records of previous mounts of a fileset
Dave Kleikamp63f83c92006-10-02 09:55:27 -0500491 * (e.g., AFTER record of extent allocation) have to be processed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 * to update block allocation map at aggregate level.
493 */
494static int logMOUNT(struct super_block *sb)
495{
496 struct jfs_log *log = JFS_SBI(sb)->log;
497 struct lrd lrd;
498
499 lrd.logtid = 0;
500 lrd.backchain = 0;
501 lrd.type = cpu_to_le16(LOG_MOUNT);
502 lrd.length = 0;
503 lrd.aggregate = cpu_to_le32(new_encode_dev(sb->s_bdev->bd_dev));
504 lmLog(log, NULL, &lrd, NULL);
505
506 return 0;
507}