1 /*
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
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/buffer_head.h>
52 #include <linux/log2.h>
53 
54 #include "jfs_incore.h"
55 #include "jfs_filsys.h"
56 #include "jfs_superblock.h"
57 #include "jfs_dmap.h"
58 #include "jfs_imap.h"
59 #include "jfs_metapage.h"
60 #include "jfs_debug.h"
61 
62 
63 /*
64  * forward references
65  */
66 static int chkSuper(struct super_block *);
67 static int logMOUNT(struct super_block *sb);
68 
69 /*
70  * NAME:	jfs_mount(sb)
71  *
72  * FUNCTION:	vfs_mount()
73  *
74  * PARAMETER:	sb	- super block
75  *
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
81  */
jfs_mount(struct super_block * sb)82 int jfs_mount(struct super_block *sb)
83 {
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;
90 
91 	/*
92 	 * read/validate superblock
93 	 * (initialize mount inode from the superblock)
94 	 */
95 	if ((rc = chkSuper(sb))) {
96 		goto out;
97 	}
98 
99 	ipaimap = diReadSpecial(sb, AGGREGATE_I, 0);
100 	if (ipaimap == NULL) {
101 		jfs_err("jfs_mount: Failed to read AGGREGATE_I");
102 		rc = -EIO;
103 		goto out;
104 	}
105 	sbi->ipaimap = ipaimap;
106 
107 	jfs_info("jfs_mount: ipaimap:0x%p", ipaimap);
108 
109 	/*
110 	 * initialize aggregate inode allocation map
111 	 */
112 	if ((rc = diMount(ipaimap))) {
113 		jfs_err("jfs_mount: diMount(ipaimap) failed w/rc = %d", rc);
114 		goto err_ipaimap;
115 	}
116 
117 	/*
118 	 * open aggregate block allocation map
119 	 */
120 	ipbmap = diReadSpecial(sb, BMAP_I, 0);
121 	if (ipbmap == NULL) {
122 		rc = -EIO;
123 		goto err_umount_ipaimap;
124 	}
125 
126 	jfs_info("jfs_mount: ipbmap:0x%p", ipbmap);
127 
128 	sbi->ipbmap = ipbmap;
129 
130 	/*
131 	 * initialize aggregate block allocation map
132 	 */
133 	if ((rc = dbMount(ipbmap))) {
134 		jfs_err("jfs_mount: dbMount failed w/rc = %d", rc);
135 		goto err_ipbmap;
136 	}
137 
138 	/*
139 	 * open the secondary aggregate inode allocation map
140 	 *
141 	 * This is a duplicate of the aggregate inode allocation map.
142 	 *
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
147 	 * table.
148 	 */
149 	if ((sbi->mntflag & JFS_BAD_SAIT) == 0) {
150 		ipaimap2 = diReadSpecial(sb, AGGREGATE_I, 1);
151 		if (!ipaimap2) {
152 			jfs_err("jfs_mount: Failed to read AGGREGATE_I");
153 			rc = -EIO;
154 			goto err_umount_ipbmap;
155 		}
156 		sbi->ipaimap2 = ipaimap2;
157 
158 		jfs_info("jfs_mount: ipaimap2:0x%p", ipaimap2);
159 
160 		/*
161 		 * initialize secondary aggregate inode allocation map
162 		 */
163 		if ((rc = diMount(ipaimap2))) {
164 			jfs_err("jfs_mount: diMount(ipaimap2) failed, rc = %d",
165 				rc);
166 			goto err_ipaimap2;
167 		}
168 	} else
169 		/* Secondary aggregate inode table is not valid */
170 		sbi->ipaimap2 = NULL;
171 
172 	/*
173 	 *	mount (the only/single) fileset
174 	 */
175 	/*
176 	 * open fileset inode allocation map (aka fileset inode)
177 	 */
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 */
182 		rc = -EIO;
183 		goto err_umount_ipaimap2;
184 	}
185 	jfs_info("jfs_mount: ipimap:0x%p", ipimap);
186 
187 	/* map further access of per fileset inodes by the fileset inode */
188 	sbi->ipimap = ipimap;
189 
190 	/* initialize fileset inode allocation map */
191 	if ((rc = diMount(ipimap))) {
192 		jfs_err("jfs_mount: diMount failed w/rc = %d", rc);
193 		goto err_ipimap;
194 	}
195 
196 	return rc;
197 
198 	/*
199 	 *	unwind on error
200 	 */
201 err_ipimap:
202 	/* close fileset inode allocation map inode */
203 	diFreeSpecial(ipimap);
204 err_umount_ipaimap2:
205 	/* close secondary aggregate inode allocation map */
206 	if (ipaimap2)
207 		diUnmount(ipaimap2, 1);
208 err_ipaimap2:
209 	/* close aggregate inodes */
210 	if (ipaimap2)
211 		diFreeSpecial(ipaimap2);
212 err_umount_ipbmap:	/* close aggregate block allocation map */
213 	dbUnmount(ipbmap, 1);
214 err_ipbmap:		/* close aggregate inodes */
215 	diFreeSpecial(ipbmap);
216 err_umount_ipaimap:	/* close aggregate inode allocation map */
217 	diUnmount(ipaimap, 1);
218 err_ipaimap:		/* close aggregate inodes */
219 	diFreeSpecial(ipaimap);
220 out:
221 	if (rc)
222 		jfs_err("Mount JFS Failure: %d", rc);
223 
224 	return rc;
225 }
226 
227 /*
228  * NAME:	jfs_mount_rw(sb, remount)
229  *
230  * FUNCTION:	Completes read-write mount, or remounts read-only volume
231  *		as read-write
232  */
jfs_mount_rw(struct super_block * sb,int remount)233 int jfs_mount_rw(struct super_block *sb, int remount)
234 {
235 	struct jfs_sb_info *sbi = JFS_SBI(sb);
236 	int rc;
237 
238 	/*
239 	 * If we are re-mounting a previously read-only volume, we want to
240 	 * re-read the inode and block maps, since fsck.jfs may have updated
241 	 * them.
242 	 */
243 	if (remount) {
244 		if (chkSuper(sb) || (sbi->state != FM_CLEAN))
245 			return -EINVAL;
246 
247 		truncate_inode_pages(sbi->ipimap->i_mapping, 0);
248 		truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
249 		diUnmount(sbi->ipimap, 1);
250 		if ((rc = diMount(sbi->ipimap))) {
251 			jfs_err("jfs_mount_rw: diMount failed!");
252 			return rc;
253 		}
254 
255 		dbUnmount(sbi->ipbmap, 1);
256 		if ((rc = dbMount(sbi->ipbmap))) {
257 			jfs_err("jfs_mount_rw: dbMount failed!");
258 			return rc;
259 		}
260 	}
261 
262 	/*
263 	 * open/initialize log
264 	 */
265 	if ((rc = lmLogOpen(sb)))
266 		return rc;
267 
268 	/*
269 	 * update file system superblock;
270 	 */
271 	if ((rc = updateSuper(sb, FM_MOUNT))) {
272 		jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
273 		lmLogClose(sb);
274 		return rc;
275 	}
276 
277 	/*
278 	 * write MOUNT log record of the file system
279 	 */
280 	logMOUNT(sb);
281 
282 	return rc;
283 }
284 
285 /*
286  *	chkSuper()
287  *
288  * validate the superblock of the file system to be mounted and
289  * get the file system parameters.
290  *
291  * returns
292  *	0 with fragsize set if check successful
293  *	error code if not successful
294  */
chkSuper(struct super_block * sb)295 static int chkSuper(struct super_block *sb)
296 {
297 	int rc = 0;
298 	struct jfs_sb_info *sbi = JFS_SBI(sb);
299 	struct jfs_superblock *j_sb;
300 	struct buffer_head *bh;
301 	int AIM_bytesize, AIT_bytesize;
302 	int expected_AIM_bytesize, expected_AIT_bytesize;
303 	s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
304 	s64 byte_addr_diff0, byte_addr_diff1;
305 	s32 bsize;
306 
307 	if ((rc = readSuper(sb, &bh)))
308 		return rc;
309 	j_sb = (struct jfs_superblock *)bh->b_data;
310 
311 	/*
312 	 * validate superblock
313 	 */
314 	/* validate fs signature */
315 	if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
316 	    le32_to_cpu(j_sb->s_version) > JFS_VERSION) {
317 		rc = -EINVAL;
318 		goto out;
319 	}
320 
321 	bsize = le32_to_cpu(j_sb->s_bsize);
322 #ifdef _JFS_4K
323 	if (bsize != PSIZE) {
324 		jfs_err("Currently only 4K block size supported!");
325 		rc = -EINVAL;
326 		goto out;
327 	}
328 #endif				/* _JFS_4K */
329 
330 	jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
331 		 le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
332 		 (unsigned long long) le64_to_cpu(j_sb->s_size));
333 
334 	/* validate the descriptors for Secondary AIM and AIT */
335 	if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
336 	    cpu_to_le32(JFS_BAD_SAIT)) {
337 		expected_AIM_bytesize = 2 * PSIZE;
338 		AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
339 		expected_AIT_bytesize = 4 * PSIZE;
340 		AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
341 		AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
342 		AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
343 		byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
344 		fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
345 		byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
346 		if ((AIM_bytesize != expected_AIM_bytesize) ||
347 		    (AIT_bytesize != expected_AIT_bytesize) ||
348 		    (byte_addr_diff0 != AIM_bytesize) ||
349 		    (byte_addr_diff1 <= AIT_bytesize))
350 			j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
351 	}
352 
353 	if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
354 	    cpu_to_le32(JFS_GROUPCOMMIT))
355 		j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
356 
357 	/* validate fs state */
358 	if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
359 	    !sb_rdonly(sb)) {
360 		jfs_err("jfs_mount: Mount Failure: File System Dirty.");
361 		rc = -EINVAL;
362 		goto out;
363 	}
364 
365 	sbi->state = le32_to_cpu(j_sb->s_state);
366 	sbi->mntflag = le32_to_cpu(j_sb->s_flag);
367 
368 	/*
369 	 * JFS always does I/O by 4K pages.  Don't tell the buffer cache
370 	 * that we use anything else (leave s_blocksize alone).
371 	 */
372 	sbi->bsize = bsize;
373 	sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
374 
375 	/* check some fields for possible corruption */
376 	if (sbi->l2bsize != ilog2((u32)bsize) ||
377 	    j_sb->pad != 0 ||
378 	    le32_to_cpu(j_sb->s_state) > FM_STATE_MAX) {
379 		rc = -EINVAL;
380 		jfs_err("jfs_mount: Mount Failure: superblock is corrupt!");
381 		goto out;
382 	}
383 
384 	/*
385 	 * For now, ignore s_pbsize, l2bfactor.  All I/O going through buffer
386 	 * cache.
387 	 */
388 	sbi->nbperpage = PSIZE >> sbi->l2bsize;
389 	sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
390 	sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
391 	if (sbi->mntflag & JFS_INLINELOG)
392 		sbi->logpxd = j_sb->s_logpxd;
393 	else {
394 		sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev));
395 		memcpy(sbi->uuid, j_sb->s_uuid, sizeof(sbi->uuid));
396 		memcpy(sbi->loguuid, j_sb->s_loguuid, sizeof(sbi->uuid));
397 	}
398 	sbi->fsckpxd = j_sb->s_fsckpxd;
399 	sbi->ait2 = j_sb->s_ait2;
400 
401       out:
402 	brelse(bh);
403 	return rc;
404 }
405 
406 
407 /*
408  *	updateSuper()
409  *
410  * update synchronously superblock if it is mounted read-write.
411  */
updateSuper(struct super_block * sb,uint state)412 int updateSuper(struct super_block *sb, uint state)
413 {
414 	struct jfs_superblock *j_sb;
415 	struct jfs_sb_info *sbi = JFS_SBI(sb);
416 	struct buffer_head *bh;
417 	int rc;
418 
419 	if (sbi->flag & JFS_NOINTEGRITY) {
420 		if (state == FM_DIRTY) {
421 			sbi->p_state = state;
422 			return 0;
423 		} else if (state == FM_MOUNT) {
424 			sbi->p_state = sbi->state;
425 			state = FM_DIRTY;
426 		} else if (state == FM_CLEAN) {
427 			state = sbi->p_state;
428 		} else
429 			jfs_err("updateSuper: bad state");
430 	} else if (sbi->state == FM_DIRTY)
431 		return 0;
432 
433 	if ((rc = readSuper(sb, &bh)))
434 		return rc;
435 
436 	j_sb = (struct jfs_superblock *)bh->b_data;
437 
438 	j_sb->s_state = cpu_to_le32(state);
439 	sbi->state = state;
440 
441 	if (state == FM_MOUNT) {
442 		/* record log's dev_t and mount serial number */
443 		j_sb->s_logdev = cpu_to_le32(new_encode_dev(sbi->log->bdev->bd_dev));
444 		j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
445 	} else if (state == FM_CLEAN) {
446 		/*
447 		 * If this volume is shared with OS/2, OS/2 will need to
448 		 * recalculate DASD usage, since we don't deal with it.
449 		 */
450 		if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
451 			j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
452 	}
453 
454 	mark_buffer_dirty(bh);
455 	sync_dirty_buffer(bh);
456 	brelse(bh);
457 
458 	return 0;
459 }
460 
461 
462 /*
463  *	readSuper()
464  *
465  * read superblock by raw sector address
466  */
readSuper(struct super_block * sb,struct buffer_head ** bpp)467 int readSuper(struct super_block *sb, struct buffer_head **bpp)
468 {
469 	/* read in primary superblock */
470 	*bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits);
471 	if (*bpp)
472 		return 0;
473 
474 	/* read in secondary/replicated superblock */
475 	*bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits);
476 	if (*bpp)
477 		return 0;
478 
479 	return -EIO;
480 }
481 
482 
483 /*
484  *	logMOUNT()
485  *
486  * function: write a MOUNT log record for file system.
487  *
488  * MOUNT record keeps logredo() from processing log records
489  * for this file system past this point in log.
490  * it is harmless if mount fails.
491  *
492  * note: MOUNT record is at aggregate level, not at fileset level,
493  * since log records of previous mounts of a fileset
494  * (e.g., AFTER record of extent allocation) have to be processed
495  * to update block allocation map at aggregate level.
496  */
logMOUNT(struct super_block * sb)497 static int logMOUNT(struct super_block *sb)
498 {
499 	struct jfs_log *log = JFS_SBI(sb)->log;
500 	struct lrd lrd;
501 
502 	lrd.logtid = 0;
503 	lrd.backchain = 0;
504 	lrd.type = cpu_to_le16(LOG_MOUNT);
505 	lrd.length = 0;
506 	lrd.aggregate = cpu_to_le32(new_encode_dev(sb->s_bdev->bd_dev));
507 	lmLog(log, NULL, &lrd, NULL);
508 
509 	return 0;
510 }
511