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, mounterror: 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 | |
238 | IWRITE_LOCK(sbi->ipimap, RDWRLOCK_IMAP); |
239 | diUnmount(sbi->ipimap, 1); |
240 | if ((rc = diMount(sbi->ipimap))) { |
241 | IWRITE_UNLOCK(sbi->ipimap); |
242 | jfs_err("jfs_mount_rw: diMount failed!" ); |
243 | return rc; |
244 | } |
245 | IWRITE_UNLOCK(sbi->ipimap); |
246 | |
247 | dbUnmount(ipbmap: sbi->ipbmap, mounterror: 1); |
248 | if ((rc = dbMount(ipbmap: sbi->ipbmap))) { |
249 | jfs_err("jfs_mount_rw: dbMount failed!" ); |
250 | return rc; |
251 | } |
252 | } |
253 | |
254 | /* |
255 | * open/initialize log |
256 | */ |
257 | if ((rc = lmLogOpen(sb))) |
258 | return rc; |
259 | |
260 | /* |
261 | * update file system superblock; |
262 | */ |
263 | if ((rc = updateSuper(sb, FM_MOUNT))) { |
264 | jfs_err("jfs_mount: updateSuper failed w/rc = %d" , rc); |
265 | lmLogClose(sb); |
266 | return rc; |
267 | } |
268 | |
269 | /* |
270 | * write MOUNT log record of the file system |
271 | */ |
272 | logMOUNT(sb); |
273 | |
274 | return rc; |
275 | } |
276 | |
277 | /* |
278 | * chkSuper() |
279 | * |
280 | * validate the superblock of the file system to be mounted and |
281 | * get the file system parameters. |
282 | * |
283 | * returns |
284 | * 0 with fragsize set if check successful |
285 | * error code if not successful |
286 | */ |
287 | static int chkSuper(struct super_block *sb) |
288 | { |
289 | int rc = 0; |
290 | struct jfs_sb_info *sbi = JFS_SBI(sb); |
291 | struct jfs_superblock *j_sb; |
292 | struct buffer_head *bh; |
293 | int AIM_bytesize, AIT_bytesize; |
294 | int expected_AIM_bytesize, expected_AIT_bytesize; |
295 | s64 AIM_byte_addr, AIT_byte_addr, fsckwsp_addr; |
296 | s64 byte_addr_diff0, byte_addr_diff1; |
297 | s32 bsize; |
298 | |
299 | if ((rc = readSuper(sb, &bh))) |
300 | return rc; |
301 | j_sb = (struct jfs_superblock *)bh->b_data; |
302 | |
303 | /* |
304 | * validate superblock |
305 | */ |
306 | /* validate fs signature */ |
307 | if (strncmp(j_sb->s_magic, JFS_MAGIC, 4) || |
308 | le32_to_cpu(j_sb->s_version) > JFS_VERSION) { |
309 | rc = -EINVAL; |
310 | goto out; |
311 | } |
312 | |
313 | bsize = le32_to_cpu(j_sb->s_bsize); |
314 | if (bsize != PSIZE) { |
315 | jfs_err("Only 4K block size supported!" ); |
316 | rc = -EINVAL; |
317 | goto out; |
318 | } |
319 | |
320 | jfs_info("superblock: flag:0x%08x state:0x%08x size:0x%Lx" , |
321 | le32_to_cpu(j_sb->s_flag), le32_to_cpu(j_sb->s_state), |
322 | (unsigned long long) le64_to_cpu(j_sb->s_size)); |
323 | |
324 | /* validate the descriptors for Secondary AIM and AIT */ |
325 | if ((j_sb->s_flag & cpu_to_le32(JFS_BAD_SAIT)) != |
326 | cpu_to_le32(JFS_BAD_SAIT)) { |
327 | expected_AIM_bytesize = 2 * PSIZE; |
328 | AIM_bytesize = lengthPXD(pxd: &(j_sb->s_aim2)) * bsize; |
329 | expected_AIT_bytesize = 4 * PSIZE; |
330 | AIT_bytesize = lengthPXD(pxd: &(j_sb->s_ait2)) * bsize; |
331 | AIM_byte_addr = addressPXD(pxd: &(j_sb->s_aim2)) * bsize; |
332 | AIT_byte_addr = addressPXD(pxd: &(j_sb->s_ait2)) * bsize; |
333 | byte_addr_diff0 = AIT_byte_addr - AIM_byte_addr; |
334 | fsckwsp_addr = addressPXD(pxd: &(j_sb->s_fsckpxd)) * bsize; |
335 | byte_addr_diff1 = fsckwsp_addr - AIT_byte_addr; |
336 | if ((AIM_bytesize != expected_AIM_bytesize) || |
337 | (AIT_bytesize != expected_AIT_bytesize) || |
338 | (byte_addr_diff0 != AIM_bytesize) || |
339 | (byte_addr_diff1 <= AIT_bytesize)) |
340 | j_sb->s_flag |= cpu_to_le32(JFS_BAD_SAIT); |
341 | } |
342 | |
343 | if ((j_sb->s_flag & cpu_to_le32(JFS_GROUPCOMMIT)) != |
344 | cpu_to_le32(JFS_GROUPCOMMIT)) |
345 | j_sb->s_flag |= cpu_to_le32(JFS_GROUPCOMMIT); |
346 | |
347 | /* validate fs state */ |
348 | if (j_sb->s_state != cpu_to_le32(FM_CLEAN) && |
349 | !sb_rdonly(sb)) { |
350 | jfs_err("jfs_mount: Mount Failure: File System Dirty." ); |
351 | rc = -EINVAL; |
352 | goto out; |
353 | } |
354 | |
355 | sbi->state = le32_to_cpu(j_sb->s_state); |
356 | sbi->mntflag = le32_to_cpu(j_sb->s_flag); |
357 | |
358 | /* |
359 | * JFS always does I/O by 4K pages. Don't tell the buffer cache |
360 | * that we use anything else (leave s_blocksize alone). |
361 | */ |
362 | sbi->bsize = bsize; |
363 | sbi->l2bsize = le16_to_cpu(j_sb->s_l2bsize); |
364 | |
365 | /* check some fields for possible corruption */ |
366 | if (sbi->l2bsize != ilog2((u32)bsize) || |
367 | j_sb->pad != 0 || |
368 | le32_to_cpu(j_sb->s_state) > FM_STATE_MAX) { |
369 | rc = -EINVAL; |
370 | jfs_err("jfs_mount: Mount Failure: superblock is corrupt!" ); |
371 | goto out; |
372 | } |
373 | |
374 | /* |
375 | * For now, ignore s_pbsize, l2bfactor. All I/O going through buffer |
376 | * cache. |
377 | */ |
378 | sbi->nbperpage = PSIZE >> sbi->l2bsize; |
379 | sbi->l2nbperpage = L2PSIZE - sbi->l2bsize; |
380 | sbi->l2niperblk = sbi->l2bsize - L2DISIZE; |
381 | if (sbi->mntflag & JFS_INLINELOG) |
382 | sbi->logpxd = j_sb->s_logpxd; |
383 | else { |
384 | sbi->logdev = new_decode_dev(le32_to_cpu(j_sb->s_logdev)); |
385 | uuid_copy(dst: &sbi->uuid, src: &j_sb->s_uuid); |
386 | uuid_copy(dst: &sbi->loguuid, src: &j_sb->s_loguuid); |
387 | } |
388 | sbi->fsckpxd = j_sb->s_fsckpxd; |
389 | sbi->ait2 = j_sb->s_ait2; |
390 | |
391 | out: |
392 | brelse(bh); |
393 | return rc; |
394 | } |
395 | |
396 | |
397 | /* |
398 | * updateSuper() |
399 | * |
400 | * update synchronously superblock if it is mounted read-write. |
401 | */ |
402 | int updateSuper(struct super_block *sb, uint state) |
403 | { |
404 | struct jfs_superblock *j_sb; |
405 | struct jfs_sb_info *sbi = JFS_SBI(sb); |
406 | struct buffer_head *bh; |
407 | int rc; |
408 | |
409 | if (sbi->flag & JFS_NOINTEGRITY) { |
410 | if (state == FM_DIRTY) { |
411 | sbi->p_state = state; |
412 | return 0; |
413 | } else if (state == FM_MOUNT) { |
414 | sbi->p_state = sbi->state; |
415 | state = FM_DIRTY; |
416 | } else if (state == FM_CLEAN) { |
417 | state = sbi->p_state; |
418 | } else |
419 | jfs_err("updateSuper: bad state" ); |
420 | } else if (sbi->state == FM_DIRTY) |
421 | return 0; |
422 | |
423 | if ((rc = readSuper(sb, &bh))) |
424 | return rc; |
425 | |
426 | j_sb = (struct jfs_superblock *)bh->b_data; |
427 | |
428 | j_sb->s_state = cpu_to_le32(state); |
429 | sbi->state = state; |
430 | |
431 | if (state == FM_MOUNT) { |
432 | /* record log's dev_t and mount serial number */ |
433 | j_sb->s_logdev = cpu_to_le32( |
434 | new_encode_dev(sbi->log->bdev_handle->bdev->bd_dev)); |
435 | j_sb->s_logserial = cpu_to_le32(sbi->log->serial); |
436 | } else if (state == FM_CLEAN) { |
437 | /* |
438 | * If this volume is shared with OS/2, OS/2 will need to |
439 | * recalculate DASD usage, since we don't deal with it. |
440 | */ |
441 | if (j_sb->s_flag & cpu_to_le32(JFS_DASD_ENABLED)) |
442 | j_sb->s_flag |= cpu_to_le32(JFS_DASD_PRIME); |
443 | } |
444 | |
445 | mark_buffer_dirty(bh); |
446 | sync_dirty_buffer(bh); |
447 | brelse(bh); |
448 | |
449 | return 0; |
450 | } |
451 | |
452 | |
453 | /* |
454 | * readSuper() |
455 | * |
456 | * read superblock by raw sector address |
457 | */ |
458 | int readSuper(struct super_block *sb, struct buffer_head **bpp) |
459 | { |
460 | /* read in primary superblock */ |
461 | *bpp = sb_bread(sb, SUPER1_OFF >> sb->s_blocksize_bits); |
462 | if (*bpp) |
463 | return 0; |
464 | |
465 | /* read in secondary/replicated superblock */ |
466 | *bpp = sb_bread(sb, SUPER2_OFF >> sb->s_blocksize_bits); |
467 | if (*bpp) |
468 | return 0; |
469 | |
470 | return -EIO; |
471 | } |
472 | |
473 | |
474 | /* |
475 | * logMOUNT() |
476 | * |
477 | * function: write a MOUNT log record for file system. |
478 | * |
479 | * MOUNT record keeps logredo() from processing log records |
480 | * for this file system past this point in log. |
481 | * it is harmless if mount fails. |
482 | * |
483 | * note: MOUNT record is at aggregate level, not at fileset level, |
484 | * since log records of previous mounts of a fileset |
485 | * (e.g., AFTER record of extent allocation) have to be processed |
486 | * to update block allocation map at aggregate level. |
487 | */ |
488 | static int logMOUNT(struct super_block *sb) |
489 | { |
490 | struct jfs_log *log = JFS_SBI(sb)->log; |
491 | struct lrd lrd; |
492 | |
493 | lrd.logtid = 0; |
494 | lrd.backchain = 0; |
495 | lrd.type = cpu_to_le16(LOG_MOUNT); |
496 | lrd.length = 0; |
497 | lrd.aggregate = cpu_to_le32(new_encode_dev(sb->s_bdev->bd_dev)); |
498 | lmLog(log, NULL, &lrd, NULL); |
499 | |
500 | return 0; |
501 | } |
502 | |