xref: /openbmc/linux/fs/jfs/jfs_mount.c (revision 4ce94eab)
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 #include <linux/blkdev.h>
40 #include <linux/log2.h>
41 
42 #include "jfs_incore.h"
43 #include "jfs_filsys.h"
44 #include "jfs_superblock.h"
45 #include "jfs_dmap.h"
46 #include "jfs_imap.h"
47 #include "jfs_metapage.h"
48 #include "jfs_debug.h"
49 
50 
51 /*
52  * forward references
53  */
54 static int chkSuper(struct super_block *);
55 static int logMOUNT(struct super_block *sb);
56 
57 /*
58  * NAME:	jfs_mount(sb)
59  *
60  * FUNCTION:	vfs_mount()
61  *
62  * PARAMETER:	sb	- super block
63  *
64  * RETURN:	-EBUSY	- device already mounted or open for write
65  *		-EBUSY	- cvrdvp already mounted;
66  *		-EBUSY	- mount table full
67  *		-ENOTDIR- cvrdvp not directory on a device mount
68  *		-ENXIO	- device open failure
69  */
70 int jfs_mount(struct super_block *sb)
71 {
72 	int rc = 0;		/* Return code */
73 	struct jfs_sb_info *sbi = JFS_SBI(sb);
74 	struct inode *ipaimap = NULL;
75 	struct inode *ipaimap2 = NULL;
76 	struct inode *ipimap = NULL;
77 	struct inode *ipbmap = NULL;
78 
79 	/*
80 	 * read/validate superblock
81 	 * (initialize mount inode from the superblock)
82 	 */
83 	if ((rc = chkSuper(sb))) {
84 		goto errout20;
85 	}
86 
87 	ipaimap = diReadSpecial(sb, AGGREGATE_I, 0);
88 	if (ipaimap == NULL) {
89 		jfs_err("jfs_mount: Failed to read AGGREGATE_I");
90 		rc = -EIO;
91 		goto errout20;
92 	}
93 	sbi->ipaimap = ipaimap;
94 
95 	jfs_info("jfs_mount: ipaimap:0x%p", ipaimap);
96 
97 	/*
98 	 * initialize aggregate inode allocation map
99 	 */
100 	if ((rc = diMount(ipaimap))) {
101 		jfs_err("jfs_mount: diMount(ipaimap) failed w/rc = %d", rc);
102 		goto errout21;
103 	}
104 
105 	/*
106 	 * open aggregate block allocation map
107 	 */
108 	ipbmap = diReadSpecial(sb, BMAP_I, 0);
109 	if (ipbmap == NULL) {
110 		rc = -EIO;
111 		goto errout22;
112 	}
113 
114 	jfs_info("jfs_mount: ipbmap:0x%p", ipbmap);
115 
116 	sbi->ipbmap = ipbmap;
117 
118 	/*
119 	 * initialize aggregate block allocation map
120 	 */
121 	if ((rc = dbMount(ipbmap))) {
122 		jfs_err("jfs_mount: dbMount failed w/rc = %d", rc);
123 		goto errout22;
124 	}
125 
126 	/*
127 	 * open the secondary aggregate inode allocation map
128 	 *
129 	 * This is a duplicate of the aggregate inode allocation map.
130 	 *
131 	 * hand craft a vfs in the same fashion as we did to read ipaimap.
132 	 * By adding INOSPEREXT (32) to the inode number, we are telling
133 	 * diReadSpecial that we are reading from the secondary aggregate
134 	 * inode table.  This also creates a unique entry in the inode hash
135 	 * table.
136 	 */
137 	if ((sbi->mntflag & JFS_BAD_SAIT) == 0) {
138 		ipaimap2 = diReadSpecial(sb, AGGREGATE_I, 1);
139 		if (!ipaimap2) {
140 			jfs_err("jfs_mount: Failed to read AGGREGATE_I");
141 			rc = -EIO;
142 			goto errout35;
143 		}
144 		sbi->ipaimap2 = ipaimap2;
145 
146 		jfs_info("jfs_mount: ipaimap2:0x%p", ipaimap2);
147 
148 		/*
149 		 * initialize secondary aggregate inode allocation map
150 		 */
151 		if ((rc = diMount(ipaimap2))) {
152 			jfs_err("jfs_mount: diMount(ipaimap2) failed, rc = %d",
153 				rc);
154 			goto errout35;
155 		}
156 	} else
157 		/* Secondary aggregate inode table is not valid */
158 		sbi->ipaimap2 = NULL;
159 
160 	/*
161 	 *	mount (the only/single) fileset
162 	 */
163 	/*
164 	 * open fileset inode allocation map (aka fileset inode)
165 	 */
166 	ipimap = diReadSpecial(sb, FILESYSTEM_I, 0);
167 	if (ipimap == NULL) {
168 		jfs_err("jfs_mount: Failed to read FILESYSTEM_I");
169 		/* open fileset secondary inode allocation map */
170 		rc = -EIO;
171 		goto errout40;
172 	}
173 	jfs_info("jfs_mount: ipimap:0x%p", ipimap);
174 
175 	/* map further access of per fileset inodes by the fileset inode */
176 	sbi->ipimap = ipimap;
177 
178 	/* initialize fileset inode allocation map */
179 	if ((rc = diMount(ipimap))) {
180 		jfs_err("jfs_mount: diMount failed w/rc = %d", rc);
181 		goto errout41;
182 	}
183 
184 	goto out;
185 
186 	/*
187 	 *	unwind on error
188 	 */
189       errout41:		/* close fileset inode allocation map inode */
190 	diFreeSpecial(ipimap);
191 
192       errout40:		/* fileset closed */
193 
194 	/* close secondary aggregate inode allocation map */
195 	if (ipaimap2) {
196 		diUnmount(ipaimap2, 1);
197 		diFreeSpecial(ipaimap2);
198 	}
199 
200       errout35:
201 
202 	/* close aggregate block allocation map */
203 	dbUnmount(ipbmap, 1);
204 	diFreeSpecial(ipbmap);
205 
206       errout22:		/* close aggregate inode allocation map */
207 
208 	diUnmount(ipaimap, 1);
209 
210       errout21:		/* close aggregate inodes */
211 	diFreeSpecial(ipaimap);
212       errout20:		/* aggregate closed */
213 
214       out:
215 
216 	if (rc)
217 		jfs_err("Mount JFS Failure: %d", rc);
218 
219 	return rc;
220 }
221 
222 /*
223  * NAME:	jfs_mount_rw(sb, remount)
224  *
225  * FUNCTION:	Completes read-write mount, or remounts read-only volume
226  *		as read-write
227  */
228 int jfs_mount_rw(struct super_block *sb, int remount)
229 {
230 	struct jfs_sb_info *sbi = JFS_SBI(sb);
231 	int rc;
232 
233 	/*
234 	 * If we are re-mounting a previously read-only volume, we want to
235 	 * re-read the inode and block maps, since fsck.jfs may have updated
236 	 * them.
237 	 */
238 	if (remount) {
239 		if (chkSuper(sb) || (sbi->state != FM_CLEAN))
240 			return -EINVAL;
241 
242 		truncate_inode_pages(sbi->ipimap->i_mapping, 0);
243 		truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
244 		diUnmount(sbi->ipimap, 1);
245 		if ((rc = diMount(sbi->ipimap))) {
246 			jfs_err("jfs_mount_rw: diMount failed!");
247 			return rc;
248 		}
249 
250 		dbUnmount(sbi->ipbmap, 1);
251 		if ((rc = dbMount(sbi->ipbmap))) {
252 			jfs_err("jfs_mount_rw: dbMount failed!");
253 			return rc;
254 		}
255 	}
256 
257 	/*
258 	 * open/initialize log
259 	 */
260 	if ((rc = lmLogOpen(sb)))
261 		return rc;
262 
263 	/*
264 	 * update file system superblock;
265 	 */
266 	if ((rc = updateSuper(sb, FM_MOUNT))) {
267 		jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
268 		lmLogClose(sb);
269 		return rc;
270 	}
271 
272 	/*
273 	 * write MOUNT log record of the file system
274 	 */
275 	logMOUNT(sb);
276 
277 	return rc;
278 }
279 
280 /*
281  *	chkSuper()
282  *
283  * validate the superblock of the file system to be mounted and
284  * get the file system parameters.
285  *
286  * returns
287  *	0 with fragsize set if check successful
288  *	error code if not successful
289  */
290 static int chkSuper(struct super_block *sb)
291 {
292 	int rc = 0;
293 	struct jfs_sb_info *sbi = JFS_SBI(sb);
294 	struct jfs_superblock *j_sb;
295 	struct buffer_head *bh;
296 	int AIM_bytesize, AIT_bytesize;
297 	int expected_AIM_bytesize, expected_AIT_bytesize;
298 	s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
299 	s64 byte_addr_diff0, byte_addr_diff1;
300 	s32 bsize;
301 
302 	if ((rc = readSuper(sb, &bh)))
303 		return rc;
304 	j_sb = (struct jfs_superblock *)bh->b_data;
305 
306 	/*
307 	 * validate superblock
308 	 */
309 	/* validate fs signature */
310 	if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
311 	    le32_to_cpu(j_sb->s_version) > JFS_VERSION) {
312 		rc = -EINVAL;
313 		goto out;
314 	}
315 
316 	bsize = le32_to_cpu(j_sb->s_bsize);
317 #ifdef _JFS_4K
318 	if (bsize != PSIZE) {
319 		jfs_err("Currently only 4K block size supported!");
320 		rc = -EINVAL;
321 		goto out;
322 	}
323 #endif				/* _JFS_4K */
324 
325 	jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
326 		 le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
327 		 (unsigned long long) le64_to_cpu(j_sb->s_size));
328 
329 	/* validate the descriptors for Secondary AIM and AIT */
330 	if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
331 	    cpu_to_le32(JFS_BAD_SAIT)) {
332 		expected_AIM_bytesize = 2 * PSIZE;
333 		AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
334 		expected_AIT_bytesize = 4 * PSIZE;
335 		AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
336 		AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
337 		AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
338 		byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
339 		fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
340 		byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
341 		if ((AIM_bytesize != expected_AIM_bytesize) ||
342 		    (AIT_bytesize != expected_AIT_bytesize) ||
343 		    (byte_addr_diff0 != AIM_bytesize) ||
344 		    (byte_addr_diff1 <= AIT_bytesize))
345 			j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
346 	}
347 
348 	if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
349 	    cpu_to_le32(JFS_GROUPCOMMIT))
350 		j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
351 
352 	/* validate fs state */
353 	if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
354 	    !sb_rdonly(sb)) {
355 		jfs_err("jfs_mount: Mount Failure: File System Dirty.");
356 		rc = -EINVAL;
357 		goto out;
358 	}
359 
360 	sbi->state = le32_to_cpu(j_sb->s_state);
361 	sbi->mntflag = le32_to_cpu(j_sb->s_flag);
362 
363 	/*
364 	 * JFS always does I/O by 4K pages.  Don't tell the buffer cache
365 	 * that we use anything else (leave s_blocksize alone).
366 	 */
367 	sbi->bsize = bsize;
368 	sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
369 
370 	/* check some fields for possible corruption */
371 	if (sbi->l2bsize != ilog2((u32)bsize) ||
372 	    j_sb->pad != 0 ||
373 	    le32_to_cpu(j_sb->s_state) > FM_STATE_MAX) {
374 		rc = -EINVAL;
375 		jfs_err("jfs_mount: Mount Failure: superblock is corrupt!");
376 		goto out;
377 	}
378 
379 	/*
380 	 * For now, ignore s_pbsize, l2bfactor.  All I/O going through buffer
381 	 * cache.
382 	 */
383 	sbi->nbperpage = PSIZE >> sbi->l2bsize;
384 	sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
385 	sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
386 	if (sbi->mntflag & JFS_INLINELOG)
387 		sbi->logpxd = j_sb->s_logpxd;
388 	else {
389 		sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev));
390 		uuid_copy(&sbi->uuid, &j_sb->s_uuid);
391 		uuid_copy(&sbi->loguuid, &j_sb->s_loguuid);
392 	}
393 	sbi->fsckpxd = j_sb->s_fsckpxd;
394 	sbi->ait2 = j_sb->s_ait2;
395 
396       out:
397 	brelse(bh);
398 	return rc;
399 }
400 
401 
402 /*
403  *	updateSuper()
404  *
405  * update synchronously superblock if it is mounted read-write.
406  */
407 int updateSuper(struct super_block *sb, uint state)
408 {
409 	struct jfs_superblock *j_sb;
410 	struct jfs_sb_info *sbi = JFS_SBI(sb);
411 	struct buffer_head *bh;
412 	int rc;
413 
414 	if (sbi->flag & JFS_NOINTEGRITY) {
415 		if (state == FM_DIRTY) {
416 			sbi->p_state = state;
417 			return 0;
418 		} else if (state == FM_MOUNT) {
419 			sbi->p_state = sbi->state;
420 			state = FM_DIRTY;
421 		} else if (state == FM_CLEAN) {
422 			state = sbi->p_state;
423 		} else
424 			jfs_err("updateSuper: bad state");
425 	} else if (sbi->state == FM_DIRTY)
426 		return 0;
427 
428 	if ((rc = readSuper(sb, &bh)))
429 		return rc;
430 
431 	j_sb = (struct jfs_superblock *)bh->b_data;
432 
433 	j_sb->s_state = cpu_to_le32(state);
434 	sbi->state = state;
435 
436 	if (state == FM_MOUNT) {
437 		/* record log's dev_t and mount serial number */
438 		j_sb->s_logdev = cpu_to_le32(new_encode_dev(sbi->log->bdev->bd_dev));
439 		j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
440 	} else if (state == FM_CLEAN) {
441 		/*
442 		 * If this volume is shared with OS/2, OS/2 will need to
443 		 * recalculate DASD usage, since we don't deal with it.
444 		 */
445 		if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
446 			j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
447 	}
448 
449 	mark_buffer_dirty(bh);
450 	sync_dirty_buffer(bh);
451 	brelse(bh);
452 
453 	return 0;
454 }
455 
456 
457 /*
458  *	readSuper()
459  *
460  * read superblock by raw sector address
461  */
462 int readSuper(struct super_block *sb, struct buffer_head **bpp)
463 {
464 	/* read in primary superblock */
465 	*bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits);
466 	if (*bpp)
467 		return 0;
468 
469 	/* read in secondary/replicated superblock */
470 	*bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits);
471 	if (*bpp)
472 		return 0;
473 
474 	return -EIO;
475 }
476 
477 
478 /*
479  *	logMOUNT()
480  *
481  * function: write a MOUNT log record for file system.
482  *
483  * MOUNT record keeps logredo() from processing log records
484  * for this file system past this point in log.
485  * it is harmless if mount fails.
486  *
487  * note: MOUNT record is at aggregate level, not at fileset level,
488  * since log records of previous mounts of a fileset
489  * (e.g., AFTER record of extent allocation) have to be processed
490  * to update block allocation map at aggregate level.
491  */
492 static int logMOUNT(struct super_block *sb)
493 {
494 	struct jfs_log *log = JFS_SBI(sb)->log;
495 	struct lrd lrd;
496 
497 	lrd.logtid = 0;
498 	lrd.backchain = 0;
499 	lrd.type = cpu_to_le16(LOG_MOUNT);
500 	lrd.length = 0;
501 	lrd.aggregate = cpu_to_le32(new_encode_dev(sb->s_bdev->bd_dev));
502 	lmLog(log, NULL, &lrd, NULL);
503 
504 	return 0;
505 }
506