xref: /openbmc/linux/fs/jfs/jfs_mount.c (revision bc33f5e5)
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 out;
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 out;
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 err_ipaimap;
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 err_umount_ipaimap;
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 err_ipbmap;
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 err_umount_ipbmap;
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 err_ipaimap2;
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 err_umount_ipaimap2;
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 err_ipimap;
182 	}
183 
184 	return rc;
185 
186 	/*
187 	 *	unwind on error
188 	 */
189 err_ipimap:
190 	/* close fileset inode allocation map inode */
191 	diFreeSpecial(ipimap);
192 err_umount_ipaimap2:
193 	/* close secondary aggregate inode allocation map */
194 	if (ipaimap2)
195 		diUnmount(ipaimap2, 1);
196 err_ipaimap2:
197 	/* close aggregate inodes */
198 	if (ipaimap2)
199 		diFreeSpecial(ipaimap2);
200 err_umount_ipbmap:	/* close aggregate block allocation map */
201 	dbUnmount(ipbmap, 1);
202 err_ipbmap:		/* close aggregate inodes */
203 	diFreeSpecial(ipbmap);
204 err_umount_ipaimap:	/* close aggregate inode allocation map */
205 	diUnmount(ipaimap, 1);
206 err_ipaimap:		/* close aggregate inodes */
207 	diFreeSpecial(ipaimap);
208 out:
209 	if (rc)
210 		jfs_err("Mount JFS Failure: %d", rc);
211 
212 	return rc;
213 }
214 
215 /*
216  * NAME:	jfs_mount_rw(sb, remount)
217  *
218  * FUNCTION:	Completes read-write mount, or remounts read-only volume
219  *		as read-write
220  */
221 int jfs_mount_rw(struct super_block *sb, int remount)
222 {
223 	struct jfs_sb_info *sbi = JFS_SBI(sb);
224 	int rc;
225 
226 	/*
227 	 * If we are re-mounting a previously read-only volume, we want to
228 	 * re-read the inode and block maps, since fsck.jfs may have updated
229 	 * them.
230 	 */
231 	if (remount) {
232 		if (chkSuper(sb) || (sbi->state != FM_CLEAN))
233 			return -EINVAL;
234 
235 		truncate_inode_pages(sbi->ipimap->i_mapping, 0);
236 		truncate_inode_pages(sbi->ipbmap->i_mapping, 0);
237 		diUnmount(sbi->ipimap, 1);
238 		if ((rc = diMount(sbi->ipimap))) {
239 			jfs_err("jfs_mount_rw: diMount failed!");
240 			return rc;
241 		}
242 
243 		dbUnmount(sbi->ipbmap, 1);
244 		if ((rc = dbMount(sbi->ipbmap))) {
245 			jfs_err("jfs_mount_rw: dbMount failed!");
246 			return rc;
247 		}
248 	}
249 
250 	/*
251 	 * open/initialize log
252 	 */
253 	if ((rc = lmLogOpen(sb)))
254 		return rc;
255 
256 	/*
257 	 * update file system superblock;
258 	 */
259 	if ((rc = updateSuper(sb, FM_MOUNT))) {
260 		jfs_err("jfs_mount: updateSuper failed w/rc = %d", rc);
261 		lmLogClose(sb);
262 		return rc;
263 	}
264 
265 	/*
266 	 * write MOUNT log record of the file system
267 	 */
268 	logMOUNT(sb);
269 
270 	return rc;
271 }
272 
273 /*
274  *	chkSuper()
275  *
276  * validate the superblock of the file system to be mounted and
277  * get the file system parameters.
278  *
279  * returns
280  *	0 with fragsize set if check successful
281  *	error code if not successful
282  */
283 static int chkSuper(struct super_block *sb)
284 {
285 	int rc = 0;
286 	struct jfs_sb_info *sbi = JFS_SBI(sb);
287 	struct jfs_superblock *j_sb;
288 	struct buffer_head *bh;
289 	int AIM_bytesize, AIT_bytesize;
290 	int expected_AIM_bytesize, expected_AIT_bytesize;
291 	s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr;
292 	s64 byte_addr_diff0, byte_addr_diff1;
293 	s32 bsize;
294 
295 	if ((rc = readSuper(sb, &bh)))
296 		return rc;
297 	j_sb = (struct jfs_superblock *)bh->b_data;
298 
299 	/*
300 	 * validate superblock
301 	 */
302 	/* validate fs signature */
303 	if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) ||
304 	    le32_to_cpu(j_sb->s_version) > JFS_VERSION) {
305 		rc = -EINVAL;
306 		goto out;
307 	}
308 
309 	bsize = le32_to_cpu(j_sb->s_bsize);
310 	if (bsize != PSIZE) {
311 		jfs_err("Only 4K block size supported!");
312 		rc = -EINVAL;
313 		goto out;
314 	}
315 
316 	jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx",
317 		 le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state),
318 		 (unsigned long long) le64_to_cpu(j_sb->s_size));
319 
320 	/* validate the descriptors for Secondary AIM and AIT */
321 	if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) !=
322 	    cpu_to_le32(JFS_BAD_SAIT)) {
323 		expected_AIM_bytesize = 2 * PSIZE;
324 		AIM_bytesize = lengthPXD(&(j_sb->s_aim2)) * bsize;
325 		expected_AIT_bytesize = 4 * PSIZE;
326 		AIT_bytesize = lengthPXD(&(j_sb->s_ait2)) * bsize;
327 		AIM_byte_addr = addressPXD(&(j_sb->s_aim2)) * bsize;
328 		AIT_byte_addr = addressPXD(&(j_sb->s_ait2)) * bsize;
329 		byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr;
330 		fsckwsp_addr = addressPXD(&(j_sb->s_fsckpxd)) * bsize;
331 		byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr;
332 		if ((AIM_bytesize != expected_AIM_bytesize) ||
333 		    (AIT_bytesize != expected_AIT_bytesize) ||
334 		    (byte_addr_diff0 != AIM_bytesize) ||
335 		    (byte_addr_diff1 <= AIT_bytesize))
336 			j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT);
337 	}
338 
339 	if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) !=
340 	    cpu_to_le32(JFS_GROUPCOMMIT))
341 		j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT);
342 
343 	/* validate fs state */
344 	if (j_sb->s_state != cpu_to_le32(FM_CLEAN) &&
345 	    !sb_rdonly(sb)) {
346 		jfs_err("jfs_mount: Mount Failure: File System Dirty.");
347 		rc = -EINVAL;
348 		goto out;
349 	}
350 
351 	sbi->state = le32_to_cpu(j_sb->s_state);
352 	sbi->mntflag = le32_to_cpu(j_sb->s_flag);
353 
354 	/*
355 	 * JFS always does I/O by 4K pages.  Don't tell the buffer cache
356 	 * that we use anything else (leave s_blocksize alone).
357 	 */
358 	sbi->bsize = bsize;
359 	sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize);
360 
361 	/* check some fields for possible corruption */
362 	if (sbi->l2bsize != ilog2((u32)bsize) ||
363 	    j_sb->pad != 0 ||
364 	    le32_to_cpu(j_sb->s_state) > FM_STATE_MAX) {
365 		rc = -EINVAL;
366 		jfs_err("jfs_mount: Mount Failure: superblock is corrupt!");
367 		goto out;
368 	}
369 
370 	/*
371 	 * For now, ignore s_pbsize, l2bfactor.  All I/O going through buffer
372 	 * cache.
373 	 */
374 	sbi->nbperpage = PSIZE >> sbi->l2bsize;
375 	sbi->l2nbperpage = L2PSIZE - sbi->l2bsize;
376 	sbi->l2niperblk = sbi->l2bsize - L2DISIZE;
377 	if (sbi->mntflag & JFS_INLINELOG)
378 		sbi->logpxd = j_sb->s_logpxd;
379 	else {
380 		sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev));
381 		uuid_copy(&sbi->uuid, &j_sb->s_uuid);
382 		uuid_copy(&sbi->loguuid, &j_sb->s_loguuid);
383 	}
384 	sbi->fsckpxd = j_sb->s_fsckpxd;
385 	sbi->ait2 = j_sb->s_ait2;
386 
387       out:
388 	brelse(bh);
389 	return rc;
390 }
391 
392 
393 /*
394  *	updateSuper()
395  *
396  * update synchronously superblock if it is mounted read-write.
397  */
398 int updateSuper(struct super_block *sb, uint state)
399 {
400 	struct jfs_superblock *j_sb;
401 	struct jfs_sb_info *sbi = JFS_SBI(sb);
402 	struct buffer_head *bh;
403 	int rc;
404 
405 	if (sbi->flag & JFS_NOINTEGRITY) {
406 		if (state == FM_DIRTY) {
407 			sbi->p_state = state;
408 			return 0;
409 		} else if (state == FM_MOUNT) {
410 			sbi->p_state = sbi->state;
411 			state = FM_DIRTY;
412 		} else if (state == FM_CLEAN) {
413 			state = sbi->p_state;
414 		} else
415 			jfs_err("updateSuper: bad state");
416 	} else if (sbi->state == FM_DIRTY)
417 		return 0;
418 
419 	if ((rc = readSuper(sb, &bh)))
420 		return rc;
421 
422 	j_sb = (struct jfs_superblock *)bh->b_data;
423 
424 	j_sb->s_state = cpu_to_le32(state);
425 	sbi->state = state;
426 
427 	if (state == FM_MOUNT) {
428 		/* record log's dev_t and mount serial number */
429 		j_sb->s_logdev = cpu_to_le32(new_encode_dev(sbi->log->bdev->bd_dev));
430 		j_sb->s_logserial = cpu_to_le32(sbi->log->serial);
431 	} else if (state == FM_CLEAN) {
432 		/*
433 		 * If this volume is shared with OS/2, OS/2 will need to
434 		 * recalculate DASD usage, since we don't deal with it.
435 		 */
436 		if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED))
437 			j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME);
438 	}
439 
440 	mark_buffer_dirty(bh);
441 	sync_dirty_buffer(bh);
442 	brelse(bh);
443 
444 	return 0;
445 }
446 
447 
448 /*
449  *	readSuper()
450  *
451  * read superblock by raw sector address
452  */
453 int readSuper(struct super_block *sb, struct buffer_head **bpp)
454 {
455 	/* read in primary superblock */
456 	*bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits);
457 	if (*bpp)
458 		return 0;
459 
460 	/* read in secondary/replicated superblock */
461 	*bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits);
462 	if (*bpp)
463 		return 0;
464 
465 	return -EIO;
466 }
467 
468 
469 /*
470  *	logMOUNT()
471  *
472  * function: write a MOUNT log record for file system.
473  *
474  * MOUNT record keeps logredo() from processing log records
475  * for this file system past this point in log.
476  * it is harmless if mount fails.
477  *
478  * note: MOUNT record is at aggregate level, not at fileset level,
479  * since log records of previous mounts of a fileset
480  * (e.g., AFTER record of extent allocation) have to be processed
481  * to update block allocation map at aggregate level.
482  */
483 static int logMOUNT(struct super_block *sb)
484 {
485 	struct jfs_log *log = JFS_SBI(sb)->log;
486 	struct lrd lrd;
487 
488 	lrd.logtid = 0;
489 	lrd.backchain = 0;
490 	lrd.type = cpu_to_le16(LOG_MOUNT);
491 	lrd.length = 0;
492 	lrd.aggregate = cpu_to_le32(new_encode_dev(sb->s_bdev->bd_dev));
493 	lmLog(log, NULL, &lrd, NULL);
494 
495 	return 0;
496 }
497