xref: /freebsd/sbin/fsck_ffs/gjournal.c (revision 8a0a413e)
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 *disk = NULL;
95 static struct fs *fs = NULL;
96 struct ufs2_dinode ufs2_zino;
97 
98 static void putcgs(void);
99 
100 /*
101  * Return cylinder group from the cache or load it if it is not in the
102  * cache yet.
103  * Don't cache more than MAX_CACHED_CGS cylinder groups.
104  */
105 static struct cgchain *
106 getcg(int cg)
107 {
108 	struct cgchain *cgc;
109 
110 	assert(disk != NULL && fs != NULL);
111 	LIST_FOREACH(cgc, &cglist, cgc_next) {
112 		if (cgc->cgc_cg.cg_cgx == cg) {
113 			//printf("%s: Found cg=%d\n", __func__, cg);
114 			return (cgc);
115 		}
116 	}
117 	/*
118 	 * Our cache is full? Let's clean it up.
119 	 */
120 	if (ncgs >= MAX_CACHED_CGS) {
121 		//printf("%s: Flushing CGs.\n", __func__);
122 		putcgs();
123 	}
124 	cgc = malloc(sizeof(*cgc));
125 	if (cgc == NULL) {
126 		/*
127 		 * Cannot allocate memory?
128 		 * Let's put all currently loaded and not busy cylinder groups
129 		 * on disk and try again.
130 		 */
131 		//printf("%s: No memory, flushing CGs.\n", __func__);
132 		putcgs();
133 		cgc = malloc(sizeof(*cgc));
134 		if (cgc == NULL)
135 			err(1, "malloc(%zu)", sizeof(*cgc));
136 	}
137 	if (cgread1(disk, cg) == -1)
138 		err(1, "cgread1(%d)", cg);
139 	bcopy(&disk->d_cg, &cgc->cgc_cg, sizeof(cgc->cgc_union));
140 	cgc->cgc_busy = 0;
141 	cgc->cgc_dirty = 0;
142 	LIST_INSERT_HEAD(&cglist, cgc, cgc_next);
143 	ncgs++;
144 	//printf("%s: Read cg=%d\n", __func__, cg);
145 	return (cgc);
146 }
147 
148 /*
149  * Mark cylinder group as dirty - it will be written back on putcgs().
150  */
151 static void
152 dirtycg(struct cgchain *cgc)
153 {
154 
155 	cgc->cgc_dirty = 1;
156 }
157 
158 /*
159  * Mark cylinder group as busy - it will not be freed on putcgs().
160  */
161 static void
162 busycg(struct cgchain *cgc)
163 {
164 
165 	cgc->cgc_busy = 1;
166 }
167 
168 /*
169  * Unmark the given cylinder group as busy.
170  */
171 static void
172 unbusycg(struct cgchain *cgc)
173 {
174 
175 	cgc->cgc_busy = 0;
176 }
177 
178 /*
179  * Write back all dirty cylinder groups.
180  * Free all non-busy cylinder groups.
181  */
182 static void
183 putcgs(void)
184 {
185 	struct cgchain *cgc, *cgc2;
186 
187 	assert(disk != NULL && fs != NULL);
188 	LIST_FOREACH_SAFE(cgc, &cglist, cgc_next, cgc2) {
189 		if (cgc->cgc_busy)
190 			continue;
191 		LIST_REMOVE(cgc, cgc_next);
192 		ncgs--;
193 		if (cgc->cgc_dirty) {
194 			bcopy(&cgc->cgc_cg, &disk->d_cg,
195 			    sizeof(cgc->cgc_union));
196 			if (cgwrite1(disk, cgc->cgc_cg.cg_cgx) == -1)
197 				err(1, "cgwrite1(%d)", cgc->cgc_cg.cg_cgx);
198 			//printf("%s: Wrote cg=%d\n", __func__,
199 			//    cgc->cgc_cg.cg_cgx);
200 		}
201 		free(cgc);
202 	}
203 }
204 
205 #if 0
206 /*
207  * Free all non-busy cylinder groups without storing the dirty ones.
208  */
209 static void
210 cancelcgs(void)
211 {
212 	struct cgchain *cgc;
213 
214 	assert(disk != NULL && fs != NULL);
215 	while ((cgc = LIST_FIRST(&cglist)) != NULL) {
216 		if (cgc->cgc_busy)
217 			continue;
218 		LIST_REMOVE(cgc, cgc_next);
219 		//printf("%s: Canceled cg=%d\n", __func__, cgc->cgc_cg.cg_cgx);
220 		free(cgc);
221 	}
222 }
223 #endif
224 
225 /*
226  * Open the given provider, load superblock.
227  */
228 static void
229 opendisk(void)
230 {
231 	if (disk != NULL)
232 		return;
233 	disk = malloc(sizeof(*disk));
234 	if (disk == NULL)
235 		err(1, "malloc(%zu)", sizeof(*disk));
236 	if (ufs_disk_fillout(disk, devnam) == -1) {
237 		err(1, "ufs_disk_fillout(%s) failed: %s", devnam,
238 		    disk->d_error);
239 	}
240 	fs = &disk->d_fs;
241 }
242 
243 /*
244  * Mark file system as clean, write the super-block back, close the disk.
245  */
246 static void
247 closedisk(void)
248 {
249 
250 	fs->fs_clean = 1;
251 	if (sbwrite(disk, 0) == -1)
252 		err(1, "sbwrite(%s)", devnam);
253 	if (ufs_disk_close(disk) == -1)
254 		err(1, "ufs_disk_close(%s)", devnam);
255 	free(disk);
256 	disk = NULL;
257 	fs = NULL;
258 }
259 
260 static void
261 blkfree(ufs2_daddr_t bno, long size)
262 {
263 	struct cgchain *cgc;
264 	struct cg *cgp;
265 	ufs1_daddr_t fragno, cgbno;
266 	int i, cg, blk, frags, bbase;
267 	u_int8_t *blksfree;
268 
269 	cg = dtog(fs, bno);
270 	cgc = getcg(cg);
271 	dirtycg(cgc);
272 	cgp = &cgc->cgc_cg;
273 	cgbno = dtogd(fs, bno);
274 	blksfree = cg_blksfree(cgp);
275 	if (size == fs->fs_bsize) {
276 		fragno = fragstoblks(fs, cgbno);
277 		if (!ffs_isfreeblock(fs, blksfree, fragno))
278 			assert(!"blkfree: freeing free block");
279 		ffs_setblock(fs, blksfree, fragno);
280 		ffs_clusteracct(fs, cgp, fragno, 1);
281 		cgp->cg_cs.cs_nbfree++;
282 		fs->fs_cstotal.cs_nbfree++;
283 		fs->fs_cs(fs, cg).cs_nbfree++;
284 	} else {
285 		bbase = cgbno - fragnum(fs, cgbno);
286 		/*
287 		 * decrement the counts associated with the old frags
288 		 */
289 		blk = blkmap(fs, blksfree, bbase);
290 		ffs_fragacct(fs, blk, cgp->cg_frsum, -1);
291 		/*
292 		 * deallocate the fragment
293 		 */
294 		frags = numfrags(fs, size);
295 		for (i = 0; i < frags; i++) {
296 			if (isset(blksfree, cgbno + i))
297 				assert(!"blkfree: freeing free frag");
298 			setbit(blksfree, cgbno + i);
299 		}
300 		cgp->cg_cs.cs_nffree += i;
301 		fs->fs_cstotal.cs_nffree += i;
302 		fs->fs_cs(fs, cg).cs_nffree += i;
303 		/*
304 		 * add back in counts associated with the new frags
305 		 */
306 		blk = blkmap(fs, blksfree, bbase);
307 		ffs_fragacct(fs, blk, cgp->cg_frsum, 1);
308 		/*
309 		 * if a complete block has been reassembled, account for it
310 		 */
311 		fragno = fragstoblks(fs, bbase);
312 		if (ffs_isblock(fs, blksfree, fragno)) {
313 			cgp->cg_cs.cs_nffree -= fs->fs_frag;
314 			fs->fs_cstotal.cs_nffree -= fs->fs_frag;
315 			fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
316 			ffs_clusteracct(fs, cgp, fragno, 1);
317 			cgp->cg_cs.cs_nbfree++;
318 			fs->fs_cstotal.cs_nbfree++;
319 			fs->fs_cs(fs, cg).cs_nbfree++;
320 		}
321 	}
322 }
323 
324 /*
325  * Recursively free all indirect blocks.
326  */
327 static void
328 freeindir(ufs2_daddr_t blk, int level)
329 {
330 	char sblks[MAXBSIZE];
331 	ufs2_daddr_t *blks;
332 	int i;
333 
334 	if (bread(disk, fsbtodb(fs, blk), (void *)&sblks, (size_t)fs->fs_bsize) == -1)
335 		err(1, "bread: %s", disk->d_error);
336 	blks = (ufs2_daddr_t *)&sblks;
337 	for (i = 0; i < NINDIR(fs); i++) {
338 		if (blks[i] == 0)
339 			break;
340 		if (level == 0)
341 			blkfree(blks[i], fs->fs_bsize);
342 		else
343 			freeindir(blks[i], level - 1);
344 	}
345 	blkfree(blk, fs->fs_bsize);
346 }
347 
348 #define	dblksize(fs, dino, lbn) \
349 	((dino)->di_size >= smalllblktosize(fs, (lbn) + 1) \
350 	    ? (fs)->fs_bsize \
351 	    : fragroundup(fs, blkoff(fs, (dino)->di_size)))
352 
353 /*
354  * Free all blocks associated with the given inode.
355  */
356 static void
357 clear_inode(struct ufs2_dinode *dino)
358 {
359 	ufs2_daddr_t bn;
360 	int extblocks, i, level;
361 	off_t osize;
362 	long bsize;
363 
364 	extblocks = 0;
365 	if (fs->fs_magic == FS_UFS2_MAGIC && dino->di_extsize > 0)
366 		extblocks = btodb(fragroundup(fs, dino->di_extsize));
367 	/* deallocate external attributes blocks */
368 	if (extblocks > 0) {
369 		osize = dino->di_extsize;
370 		dino->di_blocks -= extblocks;
371 		dino->di_extsize = 0;
372 		for (i = 0; i < UFS_NXADDR; i++) {
373 			if (dino->di_extb[i] == 0)
374 				continue;
375 			blkfree(dino->di_extb[i], sblksize(fs, osize, i));
376 		}
377 	}
378 #define	SINGLE	0	/* index of single indirect block */
379 #define	DOUBLE	1	/* index of double indirect block */
380 #define	TRIPLE	2	/* index of triple indirect block */
381 	/* deallocate indirect blocks */
382 	for (level = SINGLE; level <= TRIPLE; level++) {
383 		if (dino->di_ib[level] == 0)
384 			break;
385 		freeindir(dino->di_ib[level], level);
386 	}
387 	/* deallocate direct blocks and fragments */
388 	for (i = 0; i < UFS_NDADDR; i++) {
389 		bn = dino->di_db[i];
390 		if (bn == 0)
391 			continue;
392 		bsize = dblksize(fs, dino, i);
393 		blkfree(bn, bsize);
394 	}
395 }
396 
397 void
398 gjournal_check(const char *filesys)
399 {
400 	struct ufs2_dinode *dino;
401 	void *p;
402 	struct cgchain *cgc;
403 	struct cg *cgp;
404 	uint8_t *inosused;
405 	ino_t cino, ino;
406 	int cg, mode;
407 
408 	devnam = filesys;
409 	opendisk();
410 	/* Are there any unreferenced inodes in this file system? */
411 	if (fs->fs_unrefs == 0) {
412 		//printf("No unreferenced inodes.\n");
413 		closedisk();
414 		return;
415 	}
416 
417 	for (cg = 0; cg < fs->fs_ncg; cg++) {
418 		/* Show progress if requested. */
419 		if (got_siginfo) {
420 			printf("%s: phase j: cyl group %d of %d (%d%%)\n",
421 			    cdevname, cg, fs->fs_ncg, cg * 100 / fs->fs_ncg);
422 			got_siginfo = 0;
423 		}
424 		if (got_sigalarm) {
425 			setproctitle("%s pj %d%%", cdevname,
426 			     cg * 100 / fs->fs_ncg);
427 			got_sigalarm = 0;
428 		}
429 		cgc = getcg(cg);
430 		cgp = &cgc->cgc_cg;
431 		/* Are there any unreferenced inodes in this cylinder group? */
432 		if (cgp->cg_unrefs == 0)
433 			continue;
434 		//printf("Analizing cylinder group %d (count=%d)\n", cg, cgp->cg_unrefs);
435 		/*
436 		 * We are going to modify this cylinder group, so we want it to
437 		 * be written back.
438 		 */
439 		dirtycg(cgc);
440 		/* We don't want it to be freed in the meantime. */
441 		busycg(cgc);
442 		inosused = cg_inosused(cgp);
443 		/*
444 		 * Now go through the list of all inodes in this cylinder group
445 		 * to find unreferenced ones.
446 		 */
447 		for (cino = 0; cino < fs->fs_ipg; cino++) {
448 			ino = fs->fs_ipg * cg + cino;
449 			/* Unallocated? Skip it. */
450 			if (isclr(inosused, cino))
451 				continue;
452 			if (getino(disk, &p, ino, &mode) == -1)
453 				err(1, "getino(cg=%d ino=%ju)",
454 				    cg, (uintmax_t)ino);
455 			dino = p;
456 			/* Not a regular file nor directory? Skip it. */
457 			if (!S_ISREG(dino->di_mode) && !S_ISDIR(dino->di_mode))
458 				continue;
459 			/* Has reference(s)? Skip it. */
460 			if (dino->di_nlink > 0)
461 				continue;
462 			//printf("Clearing inode=%d (size=%jd)\n", ino, (intmax_t)dino->di_size);
463 			/* Free inode's blocks. */
464 			clear_inode(dino);
465 			/* Deallocate it. */
466 			clrbit(inosused, cino);
467 			/* Update position of last used inode. */
468 			if (ino < cgp->cg_irotor)
469 				cgp->cg_irotor = ino;
470 			/* Update statistics. */
471 			cgp->cg_cs.cs_nifree++;
472 			fs->fs_cs(fs, cg).cs_nifree++;
473 			fs->fs_cstotal.cs_nifree++;
474 			cgp->cg_unrefs--;
475 			fs->fs_unrefs--;
476 			/* If this is directory, update related statistics. */
477 			if (S_ISDIR(dino->di_mode)) {
478 				cgp->cg_cs.cs_ndir--;
479 				fs->fs_cs(fs, cg).cs_ndir--;
480 				fs->fs_cstotal.cs_ndir--;
481 			}
482 			/* Zero-fill the inode. */
483 			*dino = ufs2_zino;
484 			/* Write the inode back. */
485 			if (putino(disk) == -1)
486 				err(1, "putino(cg=%d ino=%ju)",
487 				    cg, (uintmax_t)ino);
488 			if (cgp->cg_unrefs == 0) {
489 				//printf("No more unreferenced inodes in cg=%d.\n", cg);
490 				break;
491 			}
492 		}
493 		/*
494 		 * We don't need this cylinder group anymore, so feel free to
495 		 * free it if needed.
496 		 */
497 		unbusycg(cgc);
498 		/*
499 		 * If there are no more unreferenced inodes, there is no need to
500 		 * check other cylinder groups.
501 		 */
502 		if (fs->fs_unrefs == 0) {
503 			//printf("No more unreferenced inodes (cg=%d/%d).\n", cg,
504 			//    fs->fs_ncg);
505 			break;
506 		}
507 	}
508 	/* Write back modified cylinder groups. */
509 	putcgs();
510 	/* Write back updated statistics and super-block. */
511 	closedisk();
512 }
513