1 /*
2 * Copyright (c) International Business Machines Corp., 2000-2003
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
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
8 *
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.
13 *
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
17 */
18
19 /*
20 * Module: jfs_mount.c
21 *
22 * note: file system in transition to aggregate/fileset:
23 *
24 * file system mount is interpreted as the mount of aggregate,
25 * if not already mounted, and mount of the single/only fileset in
26 * the aggregate;
27 *
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;
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 *
44 * aggregate
45 * 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/locks.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 */
65 static int chkSuper(struct super_block *);
66 static 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 */
jfs_mount(struct super_block * sb)81 int jfs_mount(struct super_block *sb)
82 {
83 int rc = 0; /* Return code */
84 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 /*
91 * read/validate superblock
92 * (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) {
100 jfs_err("jfs_mount: Faild to read AGGREGATE_I");
101 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);
150 if (ipaimap2 == 0) {
151 jfs_err("jfs_mount: Faild to read AGGREGATE_I");
152 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 = 0;
170
171 /*
172 * mount (the only/single) fileset
173 */
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 /*
198 * unwind on error
199 */
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 */
jfs_mount_rw(struct super_block * sb,int remount)239 int jfs_mount_rw(struct super_block *sb, int remount)
240 {
241 struct jfs_sb_info *sbi = JFS_SBI(sb);
242 struct jfs_log *log;
243 int rc;
244
245 /*
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
248 * them.
249 */
250 if (remount) {
251 if (chkSuper(sb) || (sbi->state != FM_CLEAN))
252 return -EINVAL;
253
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!");
259 return rc;
260 }
261
262 dbUnmount(sbi->ipbmap, 1);
263 if ((rc = dbMount(sbi->ipbmap))) {
264 jfs_err("jfs_mount_rw: dbMount failed!");
265 return rc;
266 }
267 }
268
269 /*
270 * open/initialize log
271 */
272 if ((rc = lmLogOpen(sb, &log)))
273 return rc;
274
275 JFS_SBI(sb)->log = log;
276
277 /*
278 * update file system superblock;
279 */
280 if ((rc = updateSuper(sb, FM_MOUNT))) {
281 jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
282 lmLogClose(sb, log);
283 JFS_SBI(sb)->log = 0;
284 return rc;
285 }
286
287 /*
288 * write MOUNT log record of the file system
289 */
290 logMOUNT(sb);
291
292 return rc;
293 }
294
295 /*
296 * chkSuper()
297 *
298 * validate the superblock of the file system to be mounted and
299 * get the file system parameters.
300 *
301 * returns
302 * 0 with fragsize set if check successful
303 * error code if not successful
304 */
chkSuper(struct super_block * sb)305 static int chkSuper(struct super_block *sb)
306 {
307 int rc = 0;
308 struct jfs_sb_info *sbi = JFS_SBI(sb);
309 struct jfs_superblock *j_sb;
310 struct buffer_head *bh;
311 int AIM_bytesize, AIT_bytesize;
312 int expected_AIM_bytesize, expected_AIT_bytesize;
313 s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
314 s64 byte_addr_diff0, byte_addr_diff1;
315 s32 bsize;
316
317 if ((rc = readSuper(sb, &bh)))
318 return rc;
319 j_sb = (struct jfs_superblock *)bh->b_data;
320
321 /*
322 * validate superblock
323 */
324 /* validate fs signature */
325 if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
326 le32_to_cpu(j_sb->s_version) > JFS_VERSION) {
327 rc = -EINVAL;
328 goto out;
329 }
330
331 bsize = le32_to_cpu(j_sb->s_bsize);
332 #ifdef _JFS_4K
333 if (bsize != PSIZE) {
334 jfs_err("Currently only 4K block size supported!");
335 rc = -EINVAL;
336 goto out;
337 }
338 #endif /* _JFS_4K */
339
340 jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
341 le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
342 (unsigned long long) le64_to_cpu(j_sb->s_size));
343
344 /* validate the descriptors for Secondary AIM and AIT */
345 if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
346 cpu_to_le32(JFS_BAD_SAIT)) {
347 expected_AIM_bytesize = 2 * PSIZE;
348 AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
349 expected_AIT_bytesize = 4 * PSIZE;
350 AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
351 AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
352 AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
353 byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
354 fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
355 byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
356 if ((AIM_bytesize != expected_AIM_bytesize) ||
357 (AIT_bytesize != expected_AIT_bytesize) ||
358 (byte_addr_diff0 != AIM_bytesize) ||
359 (byte_addr_diff1 <= AIT_bytesize))
360 j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
361 }
362
363 if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
364 cpu_to_le32(JFS_GROUPCOMMIT))
365 j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
366
367 /* validate fs state */
368 if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
369 !(sb->s_flags & MS_RDONLY)) {
370 jfs_err("jfs_mount: Mount Failure: File System Dirty.");
371 rc = -EINVAL;
372 goto out;
373 }
374
375 sbi->state = le32_to_cpu(j_sb->s_state);
376 sbi->mntflag = le32_to_cpu(j_sb->s_flag);
377
378 /*
379 * JFS always does I/O by 4K pages. Don't tell the buffer cache
380 * that we use anything else (leave s_blocksize alone).
381 */
382 sbi->bsize = bsize;
383 sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
384
385 /*
386 * For now, ignore s_pbsize, l2bfactor. All I/O going through buffer
387 * cache.
388 */
389 sbi->nbperpage = PSIZE >> sbi->l2bsize;
390 sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
391 sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
392 if (sbi->mntflag & JFS_INLINELOG)
393 sbi->logpxd = j_sb->s_logpxd;
394 else {
395 sbi->logdev = to_kdev_t(le32_to_cpu(j_sb->s_logdev));
396 memcpy(sbi->uuid, j_sb->s_uuid, sizeof(sbi->uuid));
397 memcpy(sbi->loguuid, j_sb->s_loguuid, sizeof(sbi->uuid));
398 }
399 sbi->fsckpxd = j_sb->s_fsckpxd;
400 sbi->ait2 = j_sb->s_ait2;
401
402 out:
403 brelse(bh);
404 return rc;
405 }
406
407
408 /*
409 * updateSuper()
410 *
411 * update synchronously superblock if it is mounted read-write.
412 */
updateSuper(struct super_block * sb,uint state)413 int updateSuper(struct super_block *sb, uint state)
414 {
415 struct jfs_superblock *j_sb;
416 struct jfs_sb_info *sbi = JFS_SBI(sb);
417 struct buffer_head *bh;
418 int rc;
419
420 if (sbi->flag & JFS_NOINTEGRITY) {
421 if (state == FM_DIRTY) {
422 sbi->p_state = state;
423 return 0;
424 } else if (state == FM_MOUNT) {
425 sbi->p_state = sbi->state;
426 state = FM_DIRTY;
427 } else if (state == FM_CLEAN) {
428 state = sbi->p_state;
429 } else
430 jfs_err("updateSuper: bad state");
431 } else if (sbi->state == FM_DIRTY)
432 return 0;
433
434 if ((rc = readSuper(sb, &bh)))
435 return rc;
436
437 j_sb = (struct jfs_superblock *)bh->b_data;
438
439 j_sb->s_state = cpu_to_le32(state);
440 sbi->state = state;
441
442 if (state == FM_MOUNT) {
443 /* record log's dev_t and mount serial number */
444 j_sb->s_logdev = cpu_to_le32(sbi->log->bdev->bd_dev);
445 j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
446 } else if (state == FM_CLEAN) {
447 /*
448 * If this volume is shared with OS/2, OS/2 will need to
449 * recalculate DASD usage, since we don't deal with it.
450 */
451 if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
452 j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
453 }
454
455 mark_buffer_dirty(bh);
456 ll_rw_block(WRITE, 1, &bh);
457 wait_on_buffer(bh);
458 brelse(bh);
459
460 return 0;
461 }
462
463
464 /*
465 * readSuper()
466 *
467 * read superblock by raw sector address
468 */
readSuper(struct super_block * sb,struct buffer_head ** bpp)469 int readSuper(struct super_block *sb, struct buffer_head **bpp)
470 {
471 /* read in primary superblock */
472 *bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits);
473 if (*bpp)
474 return 0;
475
476 /* read in secondary/replicated superblock */
477 *bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits);
478 if (*bpp)
479 return 0;
480
481 return -EIO;
482 }
483
484
485 /*
486 * logMOUNT()
487 *
488 * function: write a MOUNT log record for file system.
489 *
490 * MOUNT record keeps logredo() from processing log records
491 * for this file system past this point in log.
492 * it is harmless if mount fails.
493 *
494 * note: MOUNT record is at aggregate level, not at fileset level,
495 * since log records of previous mounts of a fileset
496 * (e.g., AFTER record of extent allocation) have to be processed
497 * to update block allocation map at aggregate level.
498 */
logMOUNT(struct super_block * sb)499 static int logMOUNT(struct super_block *sb)
500 {
501 struct jfs_log *log = JFS_SBI(sb)->log;
502 struct lrd lrd;
503
504 lrd.logtid = 0;
505 lrd.backchain = 0;
506 lrd.type = cpu_to_le16(LOG_MOUNT);
507 lrd.length = 0;
508 lrd.aggregate = cpu_to_le32(kdev_t_to_nr(sb->s_dev));
509 lmLog(log, NULL, &lrd, NULL);
510
511 return 0;
512 }
513