xref: /freebsd/sbin/fsck_ffs/gjournal.c (revision d6b92ffa)
1 /*-
2  * Copyright (c) 2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * Copyright (c) 1982, 1986, 1989, 1993
27  *	The Regents of the University of California.  All rights reserved.
28  *
29  * Redistribution and use in source and binary forms, with or without
30  * modification, are permitted provided that the following conditions
31  * are met:
32  * 1. Redistributions of source code must retain the above copyright
33  *    notice, this list of conditions and the following disclaimer.
34  * 2. Redistributions in binary form must reproduce the above copyright
35  *    notice, this list of conditions and the following disclaimer in the
36  *    documentation and/or other materials provided with the distribution.
37  * 3. Neither the name of the University nor the names of its contributors
38  *    may be used to endorse or promote products derived from this software
39  *    without specific prior written permission.
40  *
41  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
42  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
43  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
44  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
45  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
46  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
47  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
48  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
49  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
50  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
51  * SUCH DAMAGE.
52  */
53 
54 #include <sys/cdefs.h>
55 __FBSDID("$FreeBSD$");
56 
57 #include <sys/param.h>
58 #include <sys/disklabel.h>
59 #include <sys/mount.h>
60 #include <sys/stat.h>
61 
62 #include <ufs/ufs/ufsmount.h>
63 #include <ufs/ufs/dinode.h>
64 #include <ufs/ffs/fs.h>
65 
66 #include <stdio.h>
67 #include <stdlib.h>
68 #include <stdint.h>
69 #include <libufs.h>
70 #include <strings.h>
71 #include <err.h>
72 #include <assert.h>
73 
74 #include "fsck.h"
75 
76 struct cgchain {
77 	union {
78 		struct cg cgcu_cg;
79 		char cgcu_buf[MAXBSIZE];
80 	} cgc_union;
81 	int	cgc_busy;
82 	int	cgc_dirty;
83 	LIST_ENTRY(cgchain) cgc_next;
84 };
85 #define cgc_cg	cgc_union.cgcu_cg
86 
87 #define	MAX_CACHED_CGS	1024
88 static unsigned ncgs = 0;
89 static LIST_HEAD(, cgchain) cglist = LIST_HEAD_INITIALIZER(cglist);
90 
91 static const char *devnam;
92 static struct uufsd *disk = NULL;
93 static struct fs *fs = NULL;
94 struct ufs2_dinode ufs2_zino;
95 
96 static void putcgs(void);
97 
98 /*
99  * Return cylinder group from the cache or load it if it is not in the
100  * cache yet.
101  * Don't cache more than MAX_CACHED_CGS cylinder groups.
102  */
103 static struct cgchain *
104 getcg(int cg)
105 {
106 	struct cgchain *cgc;
107 
108 	assert(disk != NULL && fs != NULL);
109 	LIST_FOREACH(cgc, &cglist, cgc_next) {
110 		if (cgc->cgc_cg.cg_cgx == cg) {
111 			//printf("%s: Found cg=%d\n", __func__, cg);
112 			return (cgc);
113 		}
114 	}
115 	/*
116 	 * Our cache is full? Let's clean it up.
117 	 */
118 	if (ncgs >= MAX_CACHED_CGS) {
119 		//printf("%s: Flushing CGs.\n", __func__);
120 		putcgs();
121 	}
122 	cgc = malloc(sizeof(*cgc));
123 	if (cgc == NULL) {
124 		/*
125 		 * Cannot allocate memory?
126 		 * Let's put all currently loaded and not busy cylinder groups
127 		 * on disk and try again.
128 		 */
129 		//printf("%s: No memory, flushing CGs.\n", __func__);
130 		putcgs();
131 		cgc = malloc(sizeof(*cgc));
132 		if (cgc == NULL)
133 			err(1, "malloc(%zu)", sizeof(*cgc));
134 	}
135 	if (cgread1(disk, cg) == -1)
136 		err(1, "cgread1(%d)", cg);
137 	bcopy(&disk->d_cg, &cgc->cgc_cg, sizeof(cgc->cgc_union));
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(disk != 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 			bcopy(&cgc->cgc_cg, &disk->d_cg,
193 			    sizeof(cgc->cgc_union));
194 			if (cgwrite1(disk, cgc->cgc_cg.cg_cgx) == -1)
195 				err(1, "cgwrite1(%d)", cgc->cgc_cg.cg_cgx);
196 			//printf("%s: Wrote cg=%d\n", __func__,
197 			//    cgc->cgc_cg.cg_cgx);
198 		}
199 		free(cgc);
200 	}
201 }
202 
203 #if 0
204 /*
205  * Free all non-busy cylinder groups without storing the dirty ones.
206  */
207 static void
208 cancelcgs(void)
209 {
210 	struct cgchain *cgc;
211 
212 	assert(disk != NULL && fs != NULL);
213 	while ((cgc = LIST_FIRST(&cglist)) != NULL) {
214 		if (cgc->cgc_busy)
215 			continue;
216 		LIST_REMOVE(cgc, cgc_next);
217 		//printf("%s: Canceled cg=%d\n", __func__, cgc->cgc_cg.cg_cgx);
218 		free(cgc);
219 	}
220 }
221 #endif
222 
223 /*
224  * Open the given provider, load superblock.
225  */
226 static void
227 opendisk(void)
228 {
229 	if (disk != NULL)
230 		return;
231 	disk = malloc(sizeof(*disk));
232 	if (disk == NULL)
233 		err(1, "malloc(%zu)", sizeof(*disk));
234 	if (ufs_disk_fillout(disk, devnam) == -1) {
235 		err(1, "ufs_disk_fillout(%s) failed: %s", devnam,
236 		    disk->d_error);
237 	}
238 	fs = &disk->d_fs;
239 }
240 
241 /*
242  * Mark file system as clean, write the super-block back, close the disk.
243  */
244 static void
245 closedisk(void)
246 {
247 
248 	fs->fs_clean = 1;
249 	if (sbwrite(disk, 0) == -1)
250 		err(1, "sbwrite(%s)", devnam);
251 	if (ufs_disk_close(disk) == -1)
252 		err(1, "ufs_disk_close(%s)", devnam);
253 	free(disk);
254 	disk = NULL;
255 	fs = NULL;
256 }
257 
258 static void
259 blkfree(ufs2_daddr_t bno, long size)
260 {
261 	struct cgchain *cgc;
262 	struct cg *cgp;
263 	ufs1_daddr_t fragno, cgbno;
264 	int i, cg, blk, frags, bbase;
265 	u_int8_t *blksfree;
266 
267 	cg = dtog(fs, bno);
268 	cgc = getcg(cg);
269 	dirtycg(cgc);
270 	cgp = &cgc->cgc_cg;
271 	cgbno = dtogd(fs, bno);
272 	blksfree = cg_blksfree(cgp);
273 	if (size == fs->fs_bsize) {
274 		fragno = fragstoblks(fs, cgbno);
275 		if (!ffs_isfreeblock(fs, blksfree, fragno))
276 			assert(!"blkfree: freeing free block");
277 		ffs_setblock(fs, blksfree, fragno);
278 		ffs_clusteracct(fs, cgp, fragno, 1);
279 		cgp->cg_cs.cs_nbfree++;
280 		fs->fs_cstotal.cs_nbfree++;
281 		fs->fs_cs(fs, cg).cs_nbfree++;
282 	} else {
283 		bbase = cgbno - fragnum(fs, cgbno);
284 		/*
285 		 * decrement the counts associated with the old frags
286 		 */
287 		blk = blkmap(fs, blksfree, bbase);
288 		ffs_fragacct(fs, blk, cgp->cg_frsum, -1);
289 		/*
290 		 * deallocate the fragment
291 		 */
292 		frags = numfrags(fs, size);
293 		for (i = 0; i < frags; i++) {
294 			if (isset(blksfree, cgbno + i))
295 				assert(!"blkfree: freeing free frag");
296 			setbit(blksfree, cgbno + i);
297 		}
298 		cgp->cg_cs.cs_nffree += i;
299 		fs->fs_cstotal.cs_nffree += i;
300 		fs->fs_cs(fs, cg).cs_nffree += i;
301 		/*
302 		 * add back in counts associated with the new frags
303 		 */
304 		blk = blkmap(fs, blksfree, bbase);
305 		ffs_fragacct(fs, blk, cgp->cg_frsum, 1);
306 		/*
307 		 * if a complete block has been reassembled, account for it
308 		 */
309 		fragno = fragstoblks(fs, bbase);
310 		if (ffs_isblock(fs, blksfree, fragno)) {
311 			cgp->cg_cs.cs_nffree -= fs->fs_frag;
312 			fs->fs_cstotal.cs_nffree -= fs->fs_frag;
313 			fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
314 			ffs_clusteracct(fs, cgp, fragno, 1);
315 			cgp->cg_cs.cs_nbfree++;
316 			fs->fs_cstotal.cs_nbfree++;
317 			fs->fs_cs(fs, cg).cs_nbfree++;
318 		}
319 	}
320 }
321 
322 /*
323  * Recursively free all indirect blocks.
324  */
325 static void
326 freeindir(ufs2_daddr_t blk, int level)
327 {
328 	char sblks[MAXBSIZE];
329 	ufs2_daddr_t *blks;
330 	int i;
331 
332 	if (bread(disk, fsbtodb(fs, blk), (void *)&sblks, (size_t)fs->fs_bsize) == -1)
333 		err(1, "bread: %s", disk->d_error);
334 	blks = (ufs2_daddr_t *)&sblks;
335 	for (i = 0; i < NINDIR(fs); i++) {
336 		if (blks[i] == 0)
337 			break;
338 		if (level == 0)
339 			blkfree(blks[i], fs->fs_bsize);
340 		else
341 			freeindir(blks[i], level - 1);
342 	}
343 	blkfree(blk, fs->fs_bsize);
344 }
345 
346 #define	dblksize(fs, dino, lbn) \
347 	((dino)->di_size >= smalllblktosize(fs, (lbn) + 1) \
348 	    ? (fs)->fs_bsize \
349 	    : fragroundup(fs, blkoff(fs, (dino)->di_size)))
350 
351 /*
352  * Free all blocks associated with the given inode.
353  */
354 static void
355 clear_inode(struct ufs2_dinode *dino)
356 {
357 	ufs2_daddr_t bn;
358 	int extblocks, i, level;
359 	off_t osize;
360 	long bsize;
361 
362 	extblocks = 0;
363 	if (fs->fs_magic == FS_UFS2_MAGIC && dino->di_extsize > 0)
364 		extblocks = btodb(fragroundup(fs, dino->di_extsize));
365 	/* deallocate external attributes blocks */
366 	if (extblocks > 0) {
367 		osize = dino->di_extsize;
368 		dino->di_blocks -= extblocks;
369 		dino->di_extsize = 0;
370 		for (i = 0; i < UFS_NXADDR; i++) {
371 			if (dino->di_extb[i] == 0)
372 				continue;
373 			blkfree(dino->di_extb[i], sblksize(fs, osize, i));
374 		}
375 	}
376 #define	SINGLE	0	/* index of single indirect block */
377 #define	DOUBLE	1	/* index of double indirect block */
378 #define	TRIPLE	2	/* index of triple indirect block */
379 	/* deallocate indirect blocks */
380 	for (level = SINGLE; level <= TRIPLE; level++) {
381 		if (dino->di_ib[level] == 0)
382 			break;
383 		freeindir(dino->di_ib[level], level);
384 	}
385 	/* deallocate direct blocks and fragments */
386 	for (i = 0; i < UFS_NDADDR; i++) {
387 		bn = dino->di_db[i];
388 		if (bn == 0)
389 			continue;
390 		bsize = dblksize(fs, dino, i);
391 		blkfree(bn, bsize);
392 	}
393 }
394 
395 void
396 gjournal_check(const char *filesys)
397 {
398 	struct ufs2_dinode *dino;
399 	void *p;
400 	struct cgchain *cgc;
401 	struct cg *cgp;
402 	uint8_t *inosused;
403 	ino_t cino, ino;
404 	int cg, mode;
405 
406 	devnam = filesys;
407 	opendisk();
408 	/* Are there any unreferenced inodes in this file system? */
409 	if (fs->fs_unrefs == 0) {
410 		//printf("No unreferenced inodes.\n");
411 		closedisk();
412 		return;
413 	}
414 
415 	for (cg = 0; cg < fs->fs_ncg; cg++) {
416 		/* Show progress if requested. */
417 		if (got_siginfo) {
418 			printf("%s: phase j: cyl group %d of %d (%d%%)\n",
419 			    cdevname, cg, fs->fs_ncg, cg * 100 / fs->fs_ncg);
420 			got_siginfo = 0;
421 		}
422 		if (got_sigalarm) {
423 			setproctitle("%s pj %d%%", cdevname,
424 			     cg * 100 / fs->fs_ncg);
425 			got_sigalarm = 0;
426 		}
427 		cgc = getcg(cg);
428 		cgp = &cgc->cgc_cg;
429 		/* Are there any unreferenced inodes in this cylinder group? */
430 		if (cgp->cg_unrefs == 0)
431 			continue;
432 		//printf("Analizing cylinder group %d (count=%d)\n", cg, cgp->cg_unrefs);
433 		/*
434 		 * We are going to modify this cylinder group, so we want it to
435 		 * be written back.
436 		 */
437 		dirtycg(cgc);
438 		/* We don't want it to be freed in the meantime. */
439 		busycg(cgc);
440 		inosused = cg_inosused(cgp);
441 		/*
442 		 * Now go through the list of all inodes in this cylinder group
443 		 * to find unreferenced ones.
444 		 */
445 		for (cino = 0; cino < fs->fs_ipg; cino++) {
446 			ino = fs->fs_ipg * cg + cino;
447 			/* Unallocated? Skip it. */
448 			if (isclr(inosused, cino))
449 				continue;
450 			if (getino(disk, &p, ino, &mode) == -1)
451 				err(1, "getino(cg=%d ino=%ju)",
452 				    cg, (uintmax_t)ino);
453 			dino = p;
454 			/* Not a regular file nor directory? Skip it. */
455 			if (!S_ISREG(dino->di_mode) && !S_ISDIR(dino->di_mode))
456 				continue;
457 			/* Has reference(s)? Skip it. */
458 			if (dino->di_nlink > 0)
459 				continue;
460 			//printf("Clearing inode=%d (size=%jd)\n", ino, (intmax_t)dino->di_size);
461 			/* Free inode's blocks. */
462 			clear_inode(dino);
463 			/* Deallocate it. */
464 			clrbit(inosused, cino);
465 			/* Update position of last used inode. */
466 			if (ino < cgp->cg_irotor)
467 				cgp->cg_irotor = ino;
468 			/* Update statistics. */
469 			cgp->cg_cs.cs_nifree++;
470 			fs->fs_cs(fs, cg).cs_nifree++;
471 			fs->fs_cstotal.cs_nifree++;
472 			cgp->cg_unrefs--;
473 			fs->fs_unrefs--;
474 			/* If this is directory, update related statistics. */
475 			if (S_ISDIR(dino->di_mode)) {
476 				cgp->cg_cs.cs_ndir--;
477 				fs->fs_cs(fs, cg).cs_ndir--;
478 				fs->fs_cstotal.cs_ndir--;
479 			}
480 			/* Zero-fill the inode. */
481 			*dino = ufs2_zino;
482 			/* Write the inode back. */
483 			if (putino(disk) == -1)
484 				err(1, "putino(cg=%d ino=%ju)",
485 				    cg, (uintmax_t)ino);
486 			if (cgp->cg_unrefs == 0) {
487 				//printf("No more unreferenced inodes in cg=%d.\n", cg);
488 				break;
489 			}
490 		}
491 		/*
492 		 * We don't need this cylinder group anymore, so feel free to
493 		 * free it if needed.
494 		 */
495 		unbusycg(cgc);
496 		/*
497 		 * If there are no more unreferenced inodes, there is no need to
498 		 * check other cylinder groups.
499 		 */
500 		if (fs->fs_unrefs == 0) {
501 			//printf("No more unreferenced inodes (cg=%d/%d).\n", cg,
502 			//    fs->fs_ncg);
503 			break;
504 		}
505 	}
506 	/* Write back modified cylinder groups. */
507 	putcgs();
508 	/* Write back updated statistics and super-block. */
509 	closedisk();
510 }
511