1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Copyright (C) International Business Machines Corp., 2000-2004
4 */
5
6 /*
7 * Module: jfs_mount.c
8 *
9 * note: file system in transition to aggregate/fileset:
10 *
11 * file system mount is interpreted as the mount of aggregate,
12 * if not already mounted, and mount of the single/only fileset in
13 * the aggregate;
14 *
15 * a file system/aggregate is represented by an internal inode
16 * (aka mount inode) initialized with aggregate superblock;
17 * each vfs represents a fileset, and points to its "fileset inode
18 * allocation map inode" (aka fileset inode):
19 * (an aggregate itself is structured recursively as a filset:
20 * an internal vfs is constructed and points to its "fileset inode
21 * allocation map inode" (aka aggregate inode) where each inode
22 * represents a fileset inode) so that inode number is mapped to
23 * on-disk inode in uniform way at both aggregate and fileset level;
24 *
25 * each vnode/inode of a fileset is linked to its vfs (to facilitate
26 * per fileset inode operations, e.g., unmount of a fileset, etc.);
27 * each inode points to the mount inode (to facilitate access to
28 * per aggregate information, e.g., block size, etc.) as well as
29 * its file set inode.
30 *
31 * aggregate
32 * ipmnt
33 * mntvfs -> fileset ipimap+ -> aggregate ipbmap -> aggregate ipaimap;
34 * fileset vfs -> vp(1) <-> ... <-> vp(n) <->vproot;
35 */
36
37 #include <linux/fs.h>
38 #include <linux/buffer_head.h>
39
40 #include "jfs_incore.h"
41 #include "jfs_filsys.h"
42 #include "jfs_superblock.h"
43 #include "jfs_dmap.h"
44 #include "jfs_imap.h"
45 #include "jfs_metapage.h"
46 #include "jfs_debug.h"
47
48
49 /*
50 * forward references
51 */
52 static int chkSuper(struct super_block *);
53 static int logMOUNT(struct super_block *sb);
54
55 /*
56 * NAME: jfs_mount(sb)
57 *
58 * FUNCTION: vfs_mount()
59 *
60 * PARAMETER: sb - super block
61 *
62 * RETURN: -EBUSY - device already mounted or open for write
63 * -EBUSY - cvrdvp already mounted;
64 * -EBUSY - mount table full
65 * -ENOTDIR- cvrdvp not directory on a device mount
66 * -ENXIO - device open failure
67 */
68 int jfs_mount(struct super_block *sb)
69 {
70 int rc = 0; /* Return code */
71 struct jfs_sb_info *sbi = JFS_SBI(sb);
72 struct inode *ipaimap = NULL;
73 struct inode *ipaimap2 = NULL;
74 struct inode *ipimap = NULL;
75 struct inode *ipbmap = NULL;
76
77 /*
78 * read/validate superblock
79 * (initialize mount inode from the superblock)
80 */
81 if ((rc = chkSuper(sb))) {
82 goto errout20;
83 }
84
85 ipaimap = diReadSpecial(sb, AGGREGATE_I, 0);
86 if (ipaimap == NULL) {
87 jfs_err("jfs_mount: Failed to read AGGREGATE_I");
88 rc = -EIO;
89 goto errout20;
90 }
91 sbi->ipaimap = ipaimap;
92
93 jfs_info("jfs_mount: ipaimap:0x%p", ipaimap);
94
95 /*
96 * initialize aggregate inode allocation map
97 */
98 if ((rc = diMount(ipaimap))) {
99 jfs_err("jfs_mount: diMount(ipaimap) failed w/rc = %d", rc);
100 goto errout21;
101 }
102
103 /*
104 * open aggregate block allocation map
105 */
106 ipbmap = diReadSpecial(sb, BMAP_I, 0);
107 if (ipbmap == NULL) {
108 rc = -EIO;
109 goto errout22;
110 }
111
112 jfs_info("jfs_mount: ipbmap:0x%p", ipbmap);
113
114 sbi->ipbmap = ipbmap;
115
116 /*
117 * initialize aggregate block allocation map
118 */
119 if ((rc = dbMount(ipbmap))) {
120 jfs_err("jfs_mount: dbMount failed w/rc = %d", rc);
121 goto errout22;
122 }
123
124 /*
125 * open the secondary aggregate inode allocation map
126 *
127 * This is a duplicate of the aggregate inode allocation map.
128 *
129 * hand craft a vfs in the same fashion as we did to read ipaimap.
130 * By adding INOSPEREXT (32) to the inode number, we are telling
131 * diReadSpecial that we are reading from the secondary aggregate
132 * inode table. This also creates a unique entry in the inode hash
133 * table.
134 */
135 if ((sbi->mntflag & JFS_BAD_SAIT) == 0) {
136 ipaimap2 = diReadSpecial(sb, AGGREGATE_I, 1);
137 if (!ipaimap2) {
138 jfs_err("jfs_mount: Failed to read AGGREGATE_I");
139 rc = -EIO;
140 goto errout35;
141 }
142 sbi->ipaimap2 = ipaimap2;
143
144 jfs_info("jfs_mount: ipaimap2:0x%p", ipaimap2);
145
146 /*
147 * initialize secondary aggregate inode allocation map
148 */
149 if ((rc = diMount(ipaimap2))) {
150 jfs_err("jfs_mount: diMount(ipaimap2) failed, rc = %d",
151 rc);
152 goto errout35;
153 }
154 } else
155 /* Secondary aggregate inode table is not valid */
156 sbi->ipaimap2 = NULL;
157
158 /*
159 * mount (the only/single) fileset
160 */
161 /*
162 * open fileset inode allocation map (aka fileset inode)
163 */
164 ipimap = diReadSpecial(sb, FILESYSTEM_I, 0);
165 if (ipimap == NULL) {
166 jfs_err("jfs_mount: Failed to read FILESYSTEM_I");
167 /* open fileset secondary inode allocation map */
168 rc = -EIO;
169 goto errout40;
170 }
171 jfs_info("jfs_mount: ipimap:0x%p", ipimap);
172
173 /* map further access of per fileset inodes by the fileset inode */
174 sbi->ipimap = ipimap;
175
176 /* initialize fileset inode allocation map */
177 if ((rc = diMount(ipimap))) {
178 jfs_err("jfs_mount: diMount failed w/rc = %d", rc);
179 goto errout41;
180 }
181
182 goto out;
183
184 /*
185 * unwind on error
186 */
187 errout41: /* close fileset inode allocation map inode */
188 diFreeSpecial(ipimap);
189
190 errout40: /* fileset closed */
191
192 /* close secondary aggregate inode allocation map */
193 if (ipaimap2) {
194 diUnmount(ipaimap2, 1);
195 diFreeSpecial(ipaimap2);
196 }
197
198 errout35:
199
200 /* close aggregate block allocation map */
201 dbUnmount(ipbmap, 1);
202 diFreeSpecial(ipbmap);
203
204 errout22: /* close aggregate inode allocation map */
205
206 diUnmount(ipaimap, 1);
207
208 errout21: /* close aggregate inodes */
209 diFreeSpecial(ipaimap);
210 errout20: /* aggregate closed */
211
212 out:
213
214 if (rc)
215 jfs_err("Mount JFS Failure: %d", rc);
216
217 return rc;
218 }
219
220 /*
221 * NAME: jfs_mount_rw(sb, remount)
222 *
223 * FUNCTION: Completes read-write mount, or remounts read-only volume
224 * as read-write
225 */
226 int jfs_mount_rw(struct super_block *sb, int remount)
227 {
228 struct jfs_sb_info *sbi = JFS_SBI(sb);
229 int rc;
230
231 /*
232 * If we are re-mounting a previously read-only volume, we want to
233 * re-read the inode and block maps, since fsck.jfs may have updated
234 * them.
235 */
236 if (remount) {
237 if (chkSuper(sb) || (sbi->state != FM_CLEAN))
238 return -EINVAL;
239
240 truncate_inode_pages(sbi->ipimap->i_mapping, 0);
241 truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
242 diUnmount(sbi->ipimap, 1);
243 if ((rc = diMount(sbi->ipimap))) {
244 jfs_err("jfs_mount_rw: diMount failed!");
245 return rc;
246 }
247
248 dbUnmount(sbi->ipbmap, 1);
249 if ((rc = dbMount(sbi->ipbmap))) {
250 jfs_err("jfs_mount_rw: dbMount failed!");
251 return rc;
252 }
253 }
254
255 /*
256 * open/initialize log
257 */
258 if ((rc = lmLogOpen(sb)))
259 return rc;
260
261 /*
262 * update file system superblock;
263 */
264 if ((rc = updateSuper(sb, FM_MOUNT))) {
265 jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
266 lmLogClose(sb);
267 return rc;
268 }
269
270 /*
271 * write MOUNT log record of the file system
272 */
273 logMOUNT(sb);
274
275 return rc;
276 }
277
278 /*
279 * chkSuper()
280 *
281 * validate the superblock of the file system to be mounted and
282 * get the file system parameters.
283 *
284 * returns
285 * 0 with fragsize set if check successful
286 * error code if not successful
287 */
288 static int chkSuper(struct super_block *sb)
289 {
290 int rc = 0;
291 struct jfs_sb_info *sbi = JFS_SBI(sb);
292 struct jfs_superblock *j_sb;
293 struct buffer_head *bh;
294 int AIM_bytesize, AIT_bytesize;
295 int expected_AIM_bytesize, expected_AIT_bytesize;
296 s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
297 s64 byte_addr_diff0, byte_addr_diff1;
298 s32 bsize;
299
300 if ((rc = readSuper(sb, &bh)))
301 return rc;
302 j_sb = (struct jfs_superblock *)bh->b_data;
303
304 /*
305 * validate superblock
306 */
307 /* validate fs signature */
308 if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
309 le32_to_cpu(j_sb->s_version) > JFS_VERSION) {
310 rc = -EINVAL;
311 goto out;
312 }
313
314 bsize = le32_to_cpu(j_sb->s_bsize);
315 #ifdef _JFS_4K
316 if (bsize != PSIZE) {
317 jfs_err("Currently only 4K block size supported!");
318 rc = -EINVAL;
319 goto out;
320 }
321 #endif /* _JFS_4K */
322
323 jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
324 le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
325 (unsigned long long) le64_to_cpu(j_sb->s_size));
326
327 /* validate the descriptors for Secondary AIM and AIT */
328 if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
329 cpu_to_le32(JFS_BAD_SAIT)) {
330 expected_AIM_bytesize = 2 * PSIZE;
331 AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
332 expected_AIT_bytesize = 4 * PSIZE;
333 AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
334 AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
335 AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
336 byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
337 fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
338 byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
339 if ((AIM_bytesize != expected_AIM_bytesize) ||
340 (AIT_bytesize != expected_AIT_bytesize) ||
341 (byte_addr_diff0 != AIM_bytesize) ||
342 (byte_addr_diff1 <= AIT_bytesize))
343 j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
344 }
345
346 if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
347 cpu_to_le32(JFS_GROUPCOMMIT))
348 j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
349
350 /* validate fs state */
351 if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
352 !sb_rdonly(sb)) {
353 jfs_err("jfs_mount: Mount Failure: File System Dirty.");
354 rc = -EINVAL;
355 goto out;
356 }
357
358 sbi->state = le32_to_cpu(j_sb->s_state);
359 sbi->mntflag = le32_to_cpu(j_sb->s_flag);
360
361 /*
362 * JFS always does I/O by 4K pages. Don't tell the buffer cache
363 * that we use anything else (leave s_blocksize alone).
364 */
365 sbi->bsize = bsize;
366 sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
367
368 /*
369 * For now, ignore s_pbsize, l2bfactor. All I/O going through buffer
370 * cache.
371 */
372 sbi->nbperpage = PSIZE >> sbi->l2bsize;
373 sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
374 sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
375 if (sbi->mntflag & JFS_INLINELOG)
376 sbi->logpxd = j_sb->s_logpxd;
377 else {
378 sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev));
379 uuid_copy(&sbi->uuid, &j_sb->s_uuid);
380 uuid_copy(&sbi->loguuid, &j_sb->s_loguuid);
381 }
382 sbi->fsckpxd = j_sb->s_fsckpxd;
383 sbi->ait2 = j_sb->s_ait2;
384
385 out:
386 brelse(bh);
387 return rc;
388 }
389
390
391 /*
392 * updateSuper()
393 *
394 * update synchronously superblock if it is mounted read-write.
395 */
396 int updateSuper(struct super_block *sb, uint state)
397 {
398 struct jfs_superblock *j_sb;
399 struct jfs_sb_info *sbi = JFS_SBI(sb);
400 struct buffer_head *bh;
401 int rc;
402
403 if (sbi->flag & JFS_NOINTEGRITY) {
404 if (state == FM_DIRTY) {
405 sbi->p_state = state;
406 return 0;
407 } else if (state == FM_MOUNT) {
408 sbi->p_state = sbi->state;
409 state = FM_DIRTY;
410 } else if (state == FM_CLEAN) {
411 state = sbi->p_state;
412 } else
413 jfs_err("updateSuper: bad state");
414 } else if (sbi->state == FM_DIRTY)
415 return 0;
416
417 if ((rc = readSuper(sb, &bh)))
418 return rc;
419
420 j_sb = (struct jfs_superblock *)bh->b_data;
421
422 j_sb->s_state = cpu_to_le32(state);
423 sbi->state = state;
424
425 if (state == FM_MOUNT) {
426 /* record log's dev_t and mount serial number */
427 j_sb->s_logdev = cpu_to_le32(new_encode_dev(sbi->log->bdev->bd_dev));
428 j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
429 } else if (state == FM_CLEAN) {
430 /*
431 * If this volume is shared with OS/2, OS/2 will need to
432 * recalculate DASD usage, since we don't deal with it.
433 */
434 if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
435 j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
436 }
437
438 mark_buffer_dirty(bh);
439 sync_dirty_buffer(bh);
440 brelse(bh);
441
442 return 0;
443 }
444
445
446 /*
447 * readSuper()
448 *
449 * read superblock by raw sector address
450 */
451 int readSuper(struct super_block *sb, struct buffer_head **bpp)
452 {
453 /* read in primary superblock */
454 *bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits);
455 if (*bpp)
456 return 0;
457
458 /* read in secondary/replicated superblock */
459 *bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits);
460 if (*bpp)
461 return 0;
462
463 return -EIO;
464 }
465
466
467 /*
468 * logMOUNT()
469 *
470 * function: write a MOUNT log record for file system.
471 *
472 * MOUNT record keeps logredo() from processing log records
473 * for this file system past this point in log.
474 * it is harmless if mount fails.
475 *
476 * note: MOUNT record is at aggregate level, not at fileset level,
477 * since log records of previous mounts of a fileset
478 * (e.g., AFTER record of extent allocation) have to be processed
479 * to update block allocation map at aggregate level.
480 */
481 static int logMOUNT(struct super_block *sb)
482 {
483 struct jfs_log *log = JFS_SBI(sb)->log;
484 struct lrd lrd;
485
486 lrd.logtid = 0;
487 lrd.backchain = 0;
488 lrd.type = cpu_to_le16(LOG_MOUNT);
489 lrd.length = 0;
490 lrd.aggregate = cpu_to_le32(new_encode_dev(sb->s_bdev->bd_dev));
491 lmLog(log, NULL, &lrd, NULL);
492
493 return 0;
494 }