xref: /freebsd/sbin/fsck_ffs/gjournal.c (revision 0957b409)
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 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(diskp != 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 (cgget(diskp, cg, &cgc->cgc_cg) == -1)
138 		err(1, "cgget(%d)", cg);
139 	cgc->cgc_busy = 0;
140 	cgc->cgc_dirty = 0;
141 	LIST_INSERT_HEAD(&cglist, cgc, cgc_next);
142 	ncgs++;
143 	//printf("%s: Read cg=%d\n", __func__, cg);
144 	return (cgc);
145 }
146 
147 /*
148  * Mark cylinder group as dirty - it will be written back on putcgs().
149  */
150 static void
151 dirtycg(struct cgchain *cgc)
152 {
153 
154 	cgc->cgc_dirty = 1;
155 }
156 
157 /*
158  * Mark cylinder group as busy - it will not be freed on putcgs().
159  */
160 static void
161 busycg(struct cgchain *cgc)
162 {
163 
164 	cgc->cgc_busy = 1;
165 }
166 
167 /*
168  * Unmark the given cylinder group as busy.
169  */
170 static void
171 unbusycg(struct cgchain *cgc)
172 {
173 
174 	cgc->cgc_busy = 0;
175 }
176 
177 /*
178  * Write back all dirty cylinder groups.
179  * Free all non-busy cylinder groups.
180  */
181 static void
182 putcgs(void)
183 {
184 	struct cgchain *cgc, *cgc2;
185 
186 	assert(diskp != NULL && fs != NULL);
187 	LIST_FOREACH_SAFE(cgc, &cglist, cgc_next, cgc2) {
188 		if (cgc->cgc_busy)
189 			continue;
190 		LIST_REMOVE(cgc, cgc_next);
191 		ncgs--;
192 		if (cgc->cgc_dirty) {
193 			if (cgput(diskp, &cgc->cgc_cg) == -1)
194 				err(1, "cgput(%d)", cgc->cgc_cg.cg_cgx);
195 			//printf("%s: Wrote cg=%d\n", __func__,
196 			//    cgc->cgc_cg.cg_cgx);
197 		}
198 		free(cgc);
199 	}
200 }
201 
202 #if 0
203 /*
204  * Free all non-busy cylinder groups without storing the dirty ones.
205  */
206 static void
207 cancelcgs(void)
208 {
209 	struct cgchain *cgc;
210 
211 	assert(diskp != NULL && fs != NULL);
212 	while ((cgc = LIST_FIRST(&cglist)) != NULL) {
213 		if (cgc->cgc_busy)
214 			continue;
215 		LIST_REMOVE(cgc, cgc_next);
216 		//printf("%s: Canceled cg=%d\n", __func__, cgc->cgc_cg.cg_cgx);
217 		free(cgc);
218 	}
219 }
220 #endif
221 
222 /*
223  * Open the given provider, load superblock.
224  */
225 static void
226 opendisk(void)
227 {
228 	if (diskp != NULL)
229 		return;
230 	diskp = &disk;
231 	if (ufs_disk_fillout(diskp, devnam) == -1) {
232 		err(1, "ufs_disk_fillout(%s) failed: %s", devnam,
233 		    diskp->d_error);
234 	}
235 	fs = &diskp->d_fs;
236 }
237 
238 /*
239  * Mark file system as clean, write the super-block back, close the disk.
240  */
241 static void
242 closedisk(void)
243 {
244 
245 	fs->fs_clean = 1;
246 	if (sbwrite(diskp, 0) == -1)
247 		err(1, "sbwrite(%s)", devnam);
248 	if (ufs_disk_close(diskp) == -1)
249 		err(1, "ufs_disk_close(%s)", devnam);
250 	free(diskp);
251 	diskp = NULL;
252 	fs = NULL;
253 }
254 
255 static void
256 blkfree(ufs2_daddr_t bno, long size)
257 {
258 	struct cgchain *cgc;
259 	struct cg *cgp;
260 	ufs1_daddr_t fragno, cgbno;
261 	int i, cg, blk, frags, bbase;
262 	u_int8_t *blksfree;
263 
264 	cg = dtog(fs, bno);
265 	cgc = getcg(cg);
266 	dirtycg(cgc);
267 	cgp = &cgc->cgc_cg;
268 	cgbno = dtogd(fs, bno);
269 	blksfree = cg_blksfree(cgp);
270 	if (size == fs->fs_bsize) {
271 		fragno = fragstoblks(fs, cgbno);
272 		if (!ffs_isfreeblock(fs, blksfree, fragno))
273 			assert(!"blkfree: freeing free block");
274 		ffs_setblock(fs, blksfree, fragno);
275 		ffs_clusteracct(fs, cgp, fragno, 1);
276 		cgp->cg_cs.cs_nbfree++;
277 		fs->fs_cstotal.cs_nbfree++;
278 		fs->fs_cs(fs, cg).cs_nbfree++;
279 	} else {
280 		bbase = cgbno - fragnum(fs, cgbno);
281 		/*
282 		 * decrement the counts associated with the old frags
283 		 */
284 		blk = blkmap(fs, blksfree, bbase);
285 		ffs_fragacct(fs, blk, cgp->cg_frsum, -1);
286 		/*
287 		 * deallocate the fragment
288 		 */
289 		frags = numfrags(fs, size);
290 		for (i = 0; i < frags; i++) {
291 			if (isset(blksfree, cgbno + i))
292 				assert(!"blkfree: freeing free frag");
293 			setbit(blksfree, cgbno + i);
294 		}
295 		cgp->cg_cs.cs_nffree += i;
296 		fs->fs_cstotal.cs_nffree += i;
297 		fs->fs_cs(fs, cg).cs_nffree += i;
298 		/*
299 		 * add back in counts associated with the new frags
300 		 */
301 		blk = blkmap(fs, blksfree, bbase);
302 		ffs_fragacct(fs, blk, cgp->cg_frsum, 1);
303 		/*
304 		 * if a complete block has been reassembled, account for it
305 		 */
306 		fragno = fragstoblks(fs, bbase);
307 		if (ffs_isblock(fs, blksfree, fragno)) {
308 			cgp->cg_cs.cs_nffree -= fs->fs_frag;
309 			fs->fs_cstotal.cs_nffree -= fs->fs_frag;
310 			fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
311 			ffs_clusteracct(fs, cgp, fragno, 1);
312 			cgp->cg_cs.cs_nbfree++;
313 			fs->fs_cstotal.cs_nbfree++;
314 			fs->fs_cs(fs, cg).cs_nbfree++;
315 		}
316 	}
317 }
318 
319 /*
320  * Recursively free all indirect blocks.
321  */
322 static void
323 freeindir(ufs2_daddr_t blk, int level)
324 {
325 	char sblks[MAXBSIZE];
326 	ufs2_daddr_t *blks;
327 	int i;
328 
329 	if (bread(diskp, fsbtodb(fs, blk), (void *)&sblks, (size_t)fs->fs_bsize) == -1)
330 		err(1, "bread: %s", diskp->d_error);
331 	blks = (ufs2_daddr_t *)&sblks;
332 	for (i = 0; i < NINDIR(fs); i++) {
333 		if (blks[i] == 0)
334 			break;
335 		if (level == 0)
336 			blkfree(blks[i], fs->fs_bsize);
337 		else
338 			freeindir(blks[i], level - 1);
339 	}
340 	blkfree(blk, fs->fs_bsize);
341 }
342 
343 #define	dblksize(fs, dino, lbn) \
344 	((dino)->di_size >= smalllblktosize(fs, (lbn) + 1) \
345 	    ? (fs)->fs_bsize \
346 	    : fragroundup(fs, blkoff(fs, (dino)->di_size)))
347 
348 /*
349  * Free all blocks associated with the given inode.
350  */
351 static void
352 clear_inode(struct ufs2_dinode *dino)
353 {
354 	ufs2_daddr_t bn;
355 	int extblocks, i, level;
356 	off_t osize;
357 	long bsize;
358 
359 	extblocks = 0;
360 	if (fs->fs_magic == FS_UFS2_MAGIC && dino->di_extsize > 0)
361 		extblocks = btodb(fragroundup(fs, dino->di_extsize));
362 	/* deallocate external attributes blocks */
363 	if (extblocks > 0) {
364 		osize = dino->di_extsize;
365 		dino->di_blocks -= extblocks;
366 		dino->di_extsize = 0;
367 		for (i = 0; i < UFS_NXADDR; i++) {
368 			if (dino->di_extb[i] == 0)
369 				continue;
370 			blkfree(dino->di_extb[i], sblksize(fs, osize, i));
371 		}
372 	}
373 #define	SINGLE	0	/* index of single indirect block */
374 #define	DOUBLE	1	/* index of double indirect block */
375 #define	TRIPLE	2	/* index of triple indirect block */
376 	/* deallocate indirect blocks */
377 	for (level = SINGLE; level <= TRIPLE; level++) {
378 		if (dino->di_ib[level] == 0)
379 			break;
380 		freeindir(dino->di_ib[level], level);
381 	}
382 	/* deallocate direct blocks and fragments */
383 	for (i = 0; i < UFS_NDADDR; i++) {
384 		bn = dino->di_db[i];
385 		if (bn == 0)
386 			continue;
387 		bsize = dblksize(fs, dino, i);
388 		blkfree(bn, bsize);
389 	}
390 }
391 
392 void
393 gjournal_check(const char *filesys)
394 {
395 	union dinodep dp;
396 	struct cgchain *cgc;
397 	struct cg *cgp;
398 	uint8_t *inosused;
399 	ino_t cino, ino;
400 	int cg;
401 
402 	devnam = filesys;
403 	opendisk();
404 	/* Are there any unreferenced inodes in this file system? */
405 	if (fs->fs_unrefs == 0) {
406 		//printf("No unreferenced inodes.\n");
407 		closedisk();
408 		return;
409 	}
410 
411 	for (cg = 0; cg < fs->fs_ncg; cg++) {
412 		/* Show progress if requested. */
413 		if (got_siginfo) {
414 			printf("%s: phase j: cyl group %d of %d (%d%%)\n",
415 			    cdevname, cg, fs->fs_ncg, cg * 100 / fs->fs_ncg);
416 			got_siginfo = 0;
417 		}
418 		if (got_sigalarm) {
419 			setproctitle("%s pj %d%%", cdevname,
420 			     cg * 100 / fs->fs_ncg);
421 			got_sigalarm = 0;
422 		}
423 		cgc = getcg(cg);
424 		cgp = &cgc->cgc_cg;
425 		/* Are there any unreferenced inodes in this cylinder group? */
426 		if (cgp->cg_unrefs == 0)
427 			continue;
428 		//printf("Analizing cylinder group %d (count=%d)\n", cg, cgp->cg_unrefs);
429 		/*
430 		 * We are going to modify this cylinder group, so we want it to
431 		 * be written back.
432 		 */
433 		dirtycg(cgc);
434 		/* We don't want it to be freed in the meantime. */
435 		busycg(cgc);
436 		inosused = cg_inosused(cgp);
437 		/*
438 		 * Now go through the list of all inodes in this cylinder group
439 		 * to find unreferenced ones.
440 		 */
441 		for (cino = 0; cino < fs->fs_ipg; cino++) {
442 			ino = fs->fs_ipg * cg + cino;
443 			/* Unallocated? Skip it. */
444 			if (isclr(inosused, cino))
445 				continue;
446 			if (getinode(diskp, &dp, ino) == -1)
447 				err(1, "getinode (cg=%d ino=%ju) %s",
448 				    cg, (uintmax_t)ino, diskp->d_error);
449 			/* Not a regular file nor directory? Skip it. */
450 			if (!S_ISREG(dp.dp2->di_mode) &&
451 			    !S_ISDIR(dp.dp2->di_mode))
452 				continue;
453 			/* Has reference(s)? Skip it. */
454 			if (dp.dp2->di_nlink > 0)
455 				continue;
456 			/* printf("Clearing inode=%d (size=%jd)\n", ino,
457 			    (intmax_t)dp.dp2->di_size); */
458 			/* Free inode's blocks. */
459 			clear_inode(dp.dp2);
460 			/* Deallocate it. */
461 			clrbit(inosused, cino);
462 			/* Update position of last used inode. */
463 			if (ino < cgp->cg_irotor)
464 				cgp->cg_irotor = ino;
465 			/* Update statistics. */
466 			cgp->cg_cs.cs_nifree++;
467 			fs->fs_cs(fs, cg).cs_nifree++;
468 			fs->fs_cstotal.cs_nifree++;
469 			cgp->cg_unrefs--;
470 			fs->fs_unrefs--;
471 			/* If this is directory, update related statistics. */
472 			if (S_ISDIR(dp.dp2->di_mode)) {
473 				cgp->cg_cs.cs_ndir--;
474 				fs->fs_cs(fs, cg).cs_ndir--;
475 				fs->fs_cstotal.cs_ndir--;
476 			}
477 			/* Zero-fill the inode. */
478 			*dp.dp2 = ufs2_zino;
479 			/* Write the inode back. */
480 			if (putinode(diskp) == -1)
481 				err(1, "putinode (cg=%d ino=%ju) %s",
482 				    cg, (uintmax_t)ino, diskp->d_error);
483 			if (cgp->cg_unrefs == 0) {
484 				//printf("No more unreferenced inodes in cg=%d.\n", cg);
485 				break;
486 			}
487 		}
488 		/*
489 		 * We don't need this cylinder group anymore, so feel free to
490 		 * free it if needed.
491 		 */
492 		unbusycg(cgc);
493 		/*
494 		 * If there are no more unreferenced inodes, there is no need to
495 		 * check other cylinder groups.
496 		 */
497 		if (fs->fs_unrefs == 0) {
498 			//printf("No more unreferenced inodes (cg=%d/%d).\n", cg,
499 			//    fs->fs_ncg);
500 			break;
501 		}
502 	}
503 	/* Write back modified cylinder groups. */
504 	putcgs();
505 	/* Write back updated statistics and super-block. */
506 	closedisk();
507 }
508