xref: /freebsd/sbin/fsck_ffs/gjournal.c (revision a7623790fb345e6dc986dfd31df0ace115e6f2e4)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause AND BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  *
28  * Copyright (c) 1982, 1986, 1989, 1993
29  *	The Regents of the University of California.  All rights reserved.
30  *
31  * Redistribution and use in source and binary forms, with or without
32  * modification, are permitted provided that the following conditions
33  * are met:
34  * 1. Redistributions of source code must retain the above copyright
35  *    notice, this list of conditions and the following disclaimer.
36  * 2. Redistributions in binary form must reproduce the above copyright
37  *    notice, this list of conditions and the following disclaimer in the
38  *    documentation and/or other materials provided with the distribution.
39  * 3. Neither the name of the University nor the names of its contributors
40  *    may be used to endorse or promote products derived from this software
41  *    without specific prior written permission.
42  *
43  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
44  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
45  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
46  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
47  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
48  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
49  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
50  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
51  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
52  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
53  * SUCH DAMAGE.
54  */
55 
56 #include <sys/cdefs.h>
57 __FBSDID("$FreeBSD$");
58 
59 #include <sys/param.h>
60 #include <sys/disklabel.h>
61 #include <sys/mount.h>
62 #include <sys/stat.h>
63 
64 #include <ufs/ufs/ufsmount.h>
65 #include <ufs/ufs/dinode.h>
66 #include <ufs/ffs/fs.h>
67 
68 #include <stdio.h>
69 #include <stdlib.h>
70 #include <stdint.h>
71 #include <libufs.h>
72 #include <strings.h>
73 #include <err.h>
74 #include <assert.h>
75 
76 #include "fsck.h"
77 
78 struct cgchain {
79 	union {
80 		struct cg cgcu_cg;
81 		char cgcu_buf[MAXBSIZE];
82 	} cgc_union;
83 	int	cgc_busy;
84 	int	cgc_dirty;
85 	LIST_ENTRY(cgchain) cgc_next;
86 };
87 #define cgc_cg	cgc_union.cgcu_cg
88 
89 #define	MAX_CACHED_CGS	1024
90 static unsigned ncgs = 0;
91 static LIST_HEAD(, cgchain) cglist = LIST_HEAD_INITIALIZER(cglist);
92 
93 static const char *devnam;
94 static struct uufsd *diskp = NULL;
95 static struct fs *fs = NULL;
96 
97 static void putcgs(void);
98 
99 /*
100  * Return cylinder group from the cache or load it if it is not in the
101  * cache yet.
102  * Don't cache more than MAX_CACHED_CGS cylinder groups.
103  */
104 static struct cgchain *
105 getcg(int cg)
106 {
107 	struct cgchain *cgc;
108 
109 	assert(diskp != NULL && fs != NULL);
110 	LIST_FOREACH(cgc, &cglist, cgc_next) {
111 		if (cgc->cgc_cg.cg_cgx == cg) {
112 			//printf("%s: Found cg=%d\n", __func__, cg);
113 			return (cgc);
114 		}
115 	}
116 	/*
117 	 * Our cache is full? Let's clean it up.
118 	 */
119 	if (ncgs >= MAX_CACHED_CGS) {
120 		//printf("%s: Flushing CGs.\n", __func__);
121 		putcgs();
122 	}
123 	cgc = malloc(sizeof(*cgc));
124 	if (cgc == NULL) {
125 		/*
126 		 * Cannot allocate memory?
127 		 * Let's put all currently loaded and not busy cylinder groups
128 		 * on disk and try again.
129 		 */
130 		//printf("%s: No memory, flushing CGs.\n", __func__);
131 		putcgs();
132 		cgc = malloc(sizeof(*cgc));
133 		if (cgc == NULL)
134 			err(1, "malloc(%zu)", sizeof(*cgc));
135 	}
136 	if (cgget(fsreadfd, fs, cg, &cgc->cgc_cg) == -1)
137 		err(1, "cgget(%d)", cg);
138 	cgc->cgc_busy = 0;
139 	cgc->cgc_dirty = 0;
140 	LIST_INSERT_HEAD(&cglist, cgc, cgc_next);
141 	ncgs++;
142 	//printf("%s: Read cg=%d\n", __func__, cg);
143 	return (cgc);
144 }
145 
146 /*
147  * Mark cylinder group as dirty - it will be written back on putcgs().
148  */
149 static void
150 dirtycg(struct cgchain *cgc)
151 {
152 
153 	cgc->cgc_dirty = 1;
154 }
155 
156 /*
157  * Mark cylinder group as busy - it will not be freed on putcgs().
158  */
159 static void
160 busycg(struct cgchain *cgc)
161 {
162 
163 	cgc->cgc_busy = 1;
164 }
165 
166 /*
167  * Unmark the given cylinder group as busy.
168  */
169 static void
170 unbusycg(struct cgchain *cgc)
171 {
172 
173 	cgc->cgc_busy = 0;
174 }
175 
176 /*
177  * Write back all dirty cylinder groups.
178  * Free all non-busy cylinder groups.
179  */
180 static void
181 putcgs(void)
182 {
183 	struct cgchain *cgc, *cgc2;
184 
185 	assert(diskp != NULL && fs != NULL);
186 	LIST_FOREACH_SAFE(cgc, &cglist, cgc_next, cgc2) {
187 		if (cgc->cgc_busy)
188 			continue;
189 		LIST_REMOVE(cgc, cgc_next);
190 		ncgs--;
191 		if (cgc->cgc_dirty) {
192 			if (cgput(fswritefd, fs, &cgc->cgc_cg) == -1)
193 				err(1, "cgput(%d)", cgc->cgc_cg.cg_cgx);
194 			//printf("%s: Wrote cg=%d\n", __func__,
195 			//    cgc->cgc_cg.cg_cgx);
196 		}
197 		free(cgc);
198 	}
199 }
200 
201 #if 0
202 /*
203  * Free all non-busy cylinder groups without storing the dirty ones.
204  */
205 static void
206 cancelcgs(void)
207 {
208 	struct cgchain *cgc;
209 
210 	assert(diskp != NULL && fs != NULL);
211 	while ((cgc = LIST_FIRST(&cglist)) != NULL) {
212 		if (cgc->cgc_busy)
213 			continue;
214 		LIST_REMOVE(cgc, cgc_next);
215 		//printf("%s: Canceled cg=%d\n", __func__, cgc->cgc_cg.cg_cgx);
216 		free(cgc);
217 	}
218 }
219 #endif
220 
221 /*
222  * Open the given provider, load superblock.
223  */
224 static void
225 opendisk(void)
226 {
227 	if (diskp != NULL)
228 		return;
229 	diskp = &disk;
230 	if (ufs_disk_fillout(diskp, devnam) == -1) {
231 		err(1, "ufs_disk_fillout(%s) failed: %s", devnam,
232 		    diskp->d_error);
233 	}
234 	fs = &diskp->d_fs;
235 }
236 
237 /*
238  * Mark file system as clean, write the super-block back, close the disk.
239  */
240 static void
241 closedisk(void)
242 {
243 
244 	fs->fs_clean = 1;
245 	if (sbwrite(diskp, 0) == -1)
246 		err(1, "sbwrite(%s)", devnam);
247 	if (ufs_disk_close(diskp) == -1)
248 		err(1, "ufs_disk_close(%s)", devnam);
249 	free(diskp);
250 	diskp = NULL;
251 	fs = NULL;
252 }
253 
254 static void
255 blkfree(ufs2_daddr_t bno, long size)
256 {
257 	struct cgchain *cgc;
258 	struct cg *cgp;
259 	ufs1_daddr_t fragno, cgbno;
260 	int i, cg, blk, frags, bbase;
261 	u_int8_t *blksfree;
262 
263 	cg = dtog(fs, bno);
264 	cgc = getcg(cg);
265 	dirtycg(cgc);
266 	cgp = &cgc->cgc_cg;
267 	cgbno = dtogd(fs, bno);
268 	blksfree = cg_blksfree(cgp);
269 	if (size == fs->fs_bsize) {
270 		fragno = fragstoblks(fs, cgbno);
271 		if (!ffs_isfreeblock(fs, blksfree, fragno))
272 			assert(!"blkfree: freeing free block");
273 		ffs_setblock(fs, blksfree, fragno);
274 		ffs_clusteracct(fs, cgp, fragno, 1);
275 		cgp->cg_cs.cs_nbfree++;
276 		fs->fs_cstotal.cs_nbfree++;
277 		fs->fs_cs(fs, cg).cs_nbfree++;
278 	} else {
279 		bbase = cgbno - fragnum(fs, cgbno);
280 		/*
281 		 * decrement the counts associated with the old frags
282 		 */
283 		blk = blkmap(fs, blksfree, bbase);
284 		ffs_fragacct(fs, blk, cgp->cg_frsum, -1);
285 		/*
286 		 * deallocate the fragment
287 		 */
288 		frags = numfrags(fs, size);
289 		for (i = 0; i < frags; i++) {
290 			if (isset(blksfree, cgbno + i))
291 				assert(!"blkfree: freeing free frag");
292 			setbit(blksfree, cgbno + i);
293 		}
294 		cgp->cg_cs.cs_nffree += i;
295 		fs->fs_cstotal.cs_nffree += i;
296 		fs->fs_cs(fs, cg).cs_nffree += i;
297 		/*
298 		 * add back in counts associated with the new frags
299 		 */
300 		blk = blkmap(fs, blksfree, bbase);
301 		ffs_fragacct(fs, blk, cgp->cg_frsum, 1);
302 		/*
303 		 * if a complete block has been reassembled, account for it
304 		 */
305 		fragno = fragstoblks(fs, bbase);
306 		if (ffs_isblock(fs, blksfree, fragno)) {
307 			cgp->cg_cs.cs_nffree -= fs->fs_frag;
308 			fs->fs_cstotal.cs_nffree -= fs->fs_frag;
309 			fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
310 			ffs_clusteracct(fs, cgp, fragno, 1);
311 			cgp->cg_cs.cs_nbfree++;
312 			fs->fs_cstotal.cs_nbfree++;
313 			fs->fs_cs(fs, cg).cs_nbfree++;
314 		}
315 	}
316 }
317 
318 /*
319  * Recursively free all indirect blocks.
320  */
321 static void
322 freeindir(ufs2_daddr_t blk, int level)
323 {
324 	char sblks[MAXBSIZE];
325 	ufs2_daddr_t *blks;
326 	int i;
327 
328 	if (bread(diskp, fsbtodb(fs, blk), (void *)&sblks, (size_t)fs->fs_bsize) == -1)
329 		err(1, "bread: %s", diskp->d_error);
330 	blks = (ufs2_daddr_t *)&sblks;
331 	for (i = 0; i < NINDIR(fs); i++) {
332 		if (blks[i] == 0)
333 			break;
334 		if (level == 0)
335 			blkfree(blks[i], fs->fs_bsize);
336 		else
337 			freeindir(blks[i], level - 1);
338 	}
339 	blkfree(blk, fs->fs_bsize);
340 }
341 
342 #define	dblksize(fs, dino, lbn) \
343 	((dino)->di_size >= smalllblktosize(fs, (lbn) + 1) \
344 	    ? (fs)->fs_bsize \
345 	    : fragroundup(fs, blkoff(fs, (dino)->di_size)))
346 
347 /*
348  * Free all blocks associated with the given inode.
349  */
350 static void
351 clear_inode(struct ufs2_dinode *dino)
352 {
353 	ufs2_daddr_t bn;
354 	int extblocks, i, level;
355 	off_t osize;
356 	long bsize;
357 
358 	extblocks = 0;
359 	if (fs->fs_magic == FS_UFS2_MAGIC && dino->di_extsize > 0)
360 		extblocks = btodb(fragroundup(fs, dino->di_extsize));
361 	/* deallocate external attributes blocks */
362 	if (extblocks > 0) {
363 		osize = dino->di_extsize;
364 		dino->di_blocks -= extblocks;
365 		dino->di_extsize = 0;
366 		for (i = 0; i < UFS_NXADDR; i++) {
367 			if (dino->di_extb[i] == 0)
368 				continue;
369 			blkfree(dino->di_extb[i], sblksize(fs, osize, i));
370 		}
371 	}
372 #define	SINGLE	0	/* index of single indirect block */
373 #define	DOUBLE	1	/* index of double indirect block */
374 #define	TRIPLE	2	/* index of triple indirect block */
375 	/* deallocate indirect blocks */
376 	for (level = SINGLE; level <= TRIPLE; level++) {
377 		if (dino->di_ib[level] == 0)
378 			break;
379 		freeindir(dino->di_ib[level], level);
380 	}
381 	/* deallocate direct blocks and fragments */
382 	for (i = 0; i < UFS_NDADDR; i++) {
383 		bn = dino->di_db[i];
384 		if (bn == 0)
385 			continue;
386 		bsize = dblksize(fs, dino, i);
387 		blkfree(bn, bsize);
388 	}
389 }
390 
391 void
392 gjournal_check(const char *filesys)
393 {
394 	union dinodep dp;
395 	struct cgchain *cgc;
396 	struct cg *cgp;
397 	uint8_t *inosused;
398 	ino_t cino, ino;
399 	int cg;
400 
401 	devnam = filesys;
402 	opendisk();
403 	/* Are there any unreferenced inodes in this file system? */
404 	if (fs->fs_unrefs == 0) {
405 		//printf("No unreferenced inodes.\n");
406 		closedisk();
407 		return;
408 	}
409 
410 	for (cg = 0; cg < fs->fs_ncg; cg++) {
411 		/* Show progress if requested. */
412 		if (got_siginfo) {
413 			printf("%s: phase j: cyl group %d of %d (%d%%)\n",
414 			    cdevname, cg, fs->fs_ncg, cg * 100 / fs->fs_ncg);
415 			got_siginfo = 0;
416 		}
417 		if (got_sigalarm) {
418 			setproctitle("%s pj %d%%", cdevname,
419 			     cg * 100 / fs->fs_ncg);
420 			got_sigalarm = 0;
421 		}
422 		cgc = getcg(cg);
423 		cgp = &cgc->cgc_cg;
424 		/* Are there any unreferenced inodes in this cylinder group? */
425 		if (cgp->cg_unrefs == 0)
426 			continue;
427 		//printf("Analizing cylinder group %d (count=%d)\n", cg, cgp->cg_unrefs);
428 		/*
429 		 * We are going to modify this cylinder group, so we want it to
430 		 * be written back.
431 		 */
432 		dirtycg(cgc);
433 		/* We don't want it to be freed in the meantime. */
434 		busycg(cgc);
435 		inosused = cg_inosused(cgp);
436 		/*
437 		 * Now go through the list of all inodes in this cylinder group
438 		 * to find unreferenced ones.
439 		 */
440 		for (cino = 0; cino < fs->fs_ipg; cino++) {
441 			ino = fs->fs_ipg * cg + cino;
442 			/* Unallocated? Skip it. */
443 			if (isclr(inosused, cino))
444 				continue;
445 			if (getinode(diskp, &dp, ino) == -1)
446 				err(1, "getinode (cg=%d ino=%ju) %s",
447 				    cg, (uintmax_t)ino, diskp->d_error);
448 			/* Not a regular file nor directory? Skip it. */
449 			if (!S_ISREG(dp.dp2->di_mode) &&
450 			    !S_ISDIR(dp.dp2->di_mode))
451 				continue;
452 			/* Has reference(s)? Skip it. */
453 			if (dp.dp2->di_nlink > 0)
454 				continue;
455 			/* printf("Clearing inode=%d (size=%jd)\n", ino,
456 			    (intmax_t)dp.dp2->di_size); */
457 			/* Free inode's blocks. */
458 			clear_inode(dp.dp2);
459 			/* Deallocate it. */
460 			clrbit(inosused, cino);
461 			/* Update position of last used inode. */
462 			if (ino < cgp->cg_irotor)
463 				cgp->cg_irotor = ino;
464 			/* Update statistics. */
465 			cgp->cg_cs.cs_nifree++;
466 			fs->fs_cs(fs, cg).cs_nifree++;
467 			fs->fs_cstotal.cs_nifree++;
468 			cgp->cg_unrefs--;
469 			fs->fs_unrefs--;
470 			/* If this is directory, update related statistics. */
471 			if (S_ISDIR(dp.dp2->di_mode)) {
472 				cgp->cg_cs.cs_ndir--;
473 				fs->fs_cs(fs, cg).cs_ndir--;
474 				fs->fs_cstotal.cs_ndir--;
475 			}
476 			/* Zero-fill the inode. */
477 			*dp.dp2 = ufs2_zino;
478 			/* Write the inode back. */
479 			if (putinode(diskp) == -1)
480 				err(1, "putinode (cg=%d ino=%ju) %s",
481 				    cg, (uintmax_t)ino, diskp->d_error);
482 			if (cgp->cg_unrefs == 0) {
483 				//printf("No more unreferenced inodes in cg=%d.\n", cg);
484 				break;
485 			}
486 		}
487 		/*
488 		 * We don't need this cylinder group anymore, so feel free to
489 		 * free it if needed.
490 		 */
491 		unbusycg(cgc);
492 		/*
493 		 * If there are no more unreferenced inodes, there is no need to
494 		 * check other cylinder groups.
495 		 */
496 		if (fs->fs_unrefs == 0) {
497 			//printf("No more unreferenced inodes (cg=%d/%d).\n", cg,
498 			//    fs->fs_ncg);
499 			break;
500 		}
501 	}
502 	/* Write back modified cylinder groups. */
503 	putcgs();
504 	/* Write back updated statistics and super-block. */
505 	closedisk();
506 }
507