xref: /freebsd/sbin/tunefs/tunefs.c (revision 97549c34ecaf74580941fdc9c5bd1050e4b1f6ce)
1 /*
2  * Copyright (c) 1983, 1993
3  *	The Regents of the University of California.  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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #if 0
31 #ifndef lint
32 static const char copyright[] =
33 "@(#) Copyright (c) 1983, 1993\n\
34 	The Regents of the University of California.  All rights reserved.\n";
35 #endif /* not lint */
36 
37 #ifndef lint
38 static char sccsid[] = "@(#)tunefs.c	8.2 (Berkeley) 4/19/94";
39 #endif /* not lint */
40 #endif
41 #include <sys/cdefs.h>
42 __FBSDID("$FreeBSD$");
43 
44 /*
45  * tunefs: change layout parameters to an existing file system.
46  */
47 #include <sys/param.h>
48 #include <sys/mount.h>
49 #include <sys/disklabel.h>
50 #include <sys/stat.h>
51 
52 #include <ufs/ufs/ufsmount.h>
53 #include <ufs/ufs/dinode.h>
54 #include <ufs/ffs/fs.h>
55 #include <ufs/ufs/dir.h>
56 
57 #include <ctype.h>
58 #include <err.h>
59 #include <fcntl.h>
60 #include <fstab.h>
61 #include <libufs.h>
62 #include <paths.h>
63 #include <stdio.h>
64 #include <stdlib.h>
65 #include <stdint.h>
66 #include <string.h>
67 #include <time.h>
68 #include <unistd.h>
69 
70 /* the optimization warning string template */
71 #define	OPTWARN	"should optimize for %s with minfree %s %d%%"
72 
73 static int blocks;
74 static char clrbuf[MAXBSIZE];
75 static struct uufsd disk;
76 #define	sblock disk.d_fs
77 
78 static void usage(void);
79 static void printfs(void);
80 static int journal_alloc(int64_t size);
81 static void journal_clear(void);
82 static void sbdirty(void);
83 
84 int
85 main(int argc, char *argv[])
86 {
87 	const char *avalue, *jvalue, *Jvalue, *Lvalue, *lvalue, *Nvalue, *nvalue;
88 	const char *tvalue;
89 	const char *special, *on;
90 	const char *name;
91 	int active;
92 	int Aflag, aflag, eflag, evalue, fflag, fvalue, jflag, Jflag, kflag;
93 	int kvalue, Lflag, lflag, mflag, mvalue, Nflag, nflag, oflag, ovalue;
94 	int pflag, sflag, svalue, Svalue, tflag;
95 	int ch, found_arg, i;
96 	const char *chg[2];
97 	struct ufs_args args;
98 	struct statfs stfs;
99 
100 	if (argc < 3)
101 		usage();
102 	Aflag = aflag = eflag = fflag = jflag = Jflag = kflag = Lflag = 0;
103 	lflag = mflag = Nflag = nflag = oflag = pflag = sflag = tflag = 0;
104 	avalue = jvalue = Jvalue = Lvalue = lvalue = Nvalue = nvalue = NULL;
105 	evalue = fvalue = mvalue = ovalue = svalue = Svalue = 0;
106 	active = 0;
107 	found_arg = 0;		/* At least one arg is required. */
108 	while ((ch = getopt(argc, argv, "Aa:e:f:j:J:k:L:l:m:N:n:o:ps:S:t:"))
109 	    != -1)
110 		switch (ch) {
111 
112 		case 'A':
113 			found_arg = 1;
114 			Aflag++;
115 			break;
116 
117 		case 'a':
118 			found_arg = 1;
119 			name = "POSIX.1e ACLs";
120 			avalue = optarg;
121 			if (strcmp(avalue, "enable") &&
122 			    strcmp(avalue, "disable")) {
123 				errx(10, "bad %s (options are %s)",
124 				    name, "`enable' or `disable'");
125 			}
126 			aflag = 1;
127 			break;
128 
129 		case 'e':
130 			found_arg = 1;
131 			name = "maximum blocks per file in a cylinder group";
132 			evalue = atoi(optarg);
133 			if (evalue < 1)
134 				errx(10, "%s must be >= 1 (was %s)",
135 				    name, optarg);
136 			eflag = 1;
137 			break;
138 
139 		case 'f':
140 			found_arg = 1;
141 			name = "average file size";
142 			fvalue = atoi(optarg);
143 			if (fvalue < 1)
144 				errx(10, "%s must be >= 1 (was %s)",
145 				    name, optarg);
146 			fflag = 1;
147 			break;
148 
149 		case 'j':
150 			found_arg = 1;
151 			name = "softdep journaled file system";
152 			jvalue = optarg;
153 			if (strcmp(jvalue, "enable") &&
154 			    strcmp(jvalue, "disable")) {
155 				errx(10, "bad %s (options are %s)",
156 				    name, "`enable' or `disable'");
157 			}
158 			jflag = 1;
159 			break;
160 
161 		case 'J':
162 			found_arg = 1;
163 			name = "gjournaled file system";
164 			Jvalue = optarg;
165 			if (strcmp(Jvalue, "enable") &&
166 			    strcmp(Jvalue, "disable")) {
167 				errx(10, "bad %s (options are %s)",
168 				    name, "`enable' or `disable'");
169 			}
170 			Jflag = 1;
171 			break;
172 
173 		case 'k':
174 			found_arg = 1;
175 			name = "space to hold for metadata blocks";
176 			kvalue = atoi(optarg);
177 			if (kvalue < 0)
178 				errx(10, "bad %s (%s)", name, optarg);
179 			kflag = 1;
180 			break;
181 
182 		case 'L':
183 			found_arg = 1;
184 			name = "volume label";
185 			Lvalue = optarg;
186 			i = -1;
187 			while (isalnum(Lvalue[++i]));
188 			if (Lvalue[i] != '\0') {
189 				errx(10,
190 				"bad %s. Valid characters are alphanumerics.",
191 				    name);
192 			}
193 			if (strlen(Lvalue) >= MAXVOLLEN) {
194 				errx(10, "bad %s. Length is longer than %d.",
195 				    name, MAXVOLLEN - 1);
196 			}
197 			Lflag = 1;
198 			break;
199 
200 		case 'l':
201 			found_arg = 1;
202 			name = "multilabel MAC file system";
203 			lvalue = optarg;
204 			if (strcmp(lvalue, "enable") &&
205 			    strcmp(lvalue, "disable")) {
206 				errx(10, "bad %s (options are %s)",
207 				    name, "`enable' or `disable'");
208 			}
209 			lflag = 1;
210 			break;
211 
212 		case 'm':
213 			found_arg = 1;
214 			name = "minimum percentage of free space";
215 			mvalue = atoi(optarg);
216 			if (mvalue < 0 || mvalue > 99)
217 				errx(10, "bad %s (%s)", name, optarg);
218 			mflag = 1;
219 			break;
220 
221 		case 'N':
222 			found_arg = 1;
223 			name = "NFSv4 ACLs";
224 			Nvalue = optarg;
225 			if (strcmp(Nvalue, "enable") &&
226 			    strcmp(Nvalue, "disable")) {
227 				errx(10, "bad %s (options are %s)",
228 				    name, "`enable' or `disable'");
229 			}
230 			Nflag = 1;
231 			break;
232 
233 		case 'n':
234 			found_arg = 1;
235 			name = "soft updates";
236 			nvalue = optarg;
237 			if (strcmp(nvalue, "enable") != 0 &&
238 			    strcmp(nvalue, "disable") != 0) {
239 				errx(10, "bad %s (options are %s)",
240 				    name, "`enable' or `disable'");
241 			}
242 			nflag = 1;
243 			break;
244 
245 		case 'o':
246 			found_arg = 1;
247 			name = "optimization preference";
248 			if (strcmp(optarg, "space") == 0)
249 				ovalue = FS_OPTSPACE;
250 			else if (strcmp(optarg, "time") == 0)
251 				ovalue = FS_OPTTIME;
252 			else
253 				errx(10,
254 				    "bad %s (options are `space' or `time')",
255 				    name);
256 			oflag = 1;
257 			break;
258 
259 		case 'p':
260 			found_arg = 1;
261 			pflag = 1;
262 			break;
263 
264 		case 's':
265 			found_arg = 1;
266 			name = "expected number of files per directory";
267 			svalue = atoi(optarg);
268 			if (svalue < 1)
269 				errx(10, "%s must be >= 1 (was %s)",
270 				    name, optarg);
271 			sflag = 1;
272 			break;
273 
274 		case 'S':
275 			found_arg = 1;
276 			name = "Softdep Journal Size";
277 			Svalue = atoi(optarg);
278 			if (Svalue < SUJ_MIN)
279 				errx(10, "%s must be >= %d (was %s)",
280 				    name, SUJ_MIN, optarg);
281 			break;
282 
283 		case 't':
284 			found_arg = 1;
285 			name = "trim";
286 			tvalue = optarg;
287 			if (strcmp(tvalue, "enable") != 0 &&
288 			    strcmp(tvalue, "disable") != 0) {
289 				errx(10, "bad %s (options are %s)",
290 				    name, "`enable' or `disable'");
291 			}
292 			tflag = 1;
293 			break;
294 
295 		default:
296 			usage();
297 		}
298 	argc -= optind;
299 	argv += optind;
300 	if (found_arg == 0 || argc != 1)
301 		usage();
302 
303 	on = special = argv[0];
304 	if (ufs_disk_fillout(&disk, special) == -1)
305 		goto err;
306 	if (disk.d_name != special) {
307 		if (statfs(special, &stfs) != 0)
308 			warn("Can't stat %s", special);
309 		if (strcmp(special, stfs.f_mntonname) == 0)
310 			active = 1;
311 	}
312 
313 	if (pflag) {
314 		printfs();
315 		exit(0);
316 	}
317 	if (Lflag) {
318 		name = "volume label";
319 		strncpy(sblock.fs_volname, Lvalue, MAXVOLLEN);
320 	}
321 	if (aflag) {
322 		name = "POSIX.1e ACLs";
323 		if (strcmp(avalue, "enable") == 0) {
324 			if (sblock.fs_flags & FS_ACLS) {
325 				warnx("%s remains unchanged as enabled", name);
326 			} else if (sblock.fs_flags & FS_NFS4ACLS) {
327 				warnx("%s and NFSv4 ACLs are mutually "
328 				    "exclusive", name);
329 			} else {
330 				sblock.fs_flags |= FS_ACLS;
331 				warnx("%s set", name);
332 			}
333 		} else if (strcmp(avalue, "disable") == 0) {
334 			if ((~sblock.fs_flags & FS_ACLS) ==
335 			    FS_ACLS) {
336 				warnx("%s remains unchanged as disabled",
337 				    name);
338 			} else {
339 				sblock.fs_flags &= ~FS_ACLS;
340 				warnx("%s cleared", name);
341 			}
342 		}
343 	}
344 	if (eflag) {
345 		name = "maximum blocks per file in a cylinder group";
346 		if (sblock.fs_maxbpg == evalue)
347 			warnx("%s remains unchanged as %d", name, evalue);
348 		else {
349 			warnx("%s changes from %d to %d",
350 			    name, sblock.fs_maxbpg, evalue);
351 			sblock.fs_maxbpg = evalue;
352 		}
353 	}
354 	if (fflag) {
355 		name = "average file size";
356 		if (sblock.fs_avgfilesize == (unsigned)fvalue) {
357 			warnx("%s remains unchanged as %d", name, fvalue);
358 		}
359 		else {
360 			warnx("%s changes from %d to %d",
361 					name, sblock.fs_avgfilesize, fvalue);
362 			sblock.fs_avgfilesize = fvalue;
363 		}
364 	}
365 	if (jflag) {
366  		name = "soft updates journaling";
367  		if (strcmp(jvalue, "enable") == 0) {
368 			if ((sblock.fs_flags & (FS_DOSOFTDEP | FS_SUJ)) ==
369 			    (FS_DOSOFTDEP | FS_SUJ)) {
370 				warnx("%s remains unchanged as enabled", name);
371 			} else if (sblock.fs_clean == 0) {
372 				warnx("%s cannot be enabled until fsck is run",
373 				    name);
374 			} else if (journal_alloc(Svalue) != 0) {
375 				warnx("%s can not be enabled", name);
376 			} else {
377  				sblock.fs_flags |= FS_DOSOFTDEP | FS_SUJ;
378  				warnx("%s set", name);
379 			}
380  		} else if (strcmp(jvalue, "disable") == 0) {
381 			if ((~sblock.fs_flags & FS_SUJ) == FS_SUJ) {
382 				warnx("%s remains unchanged as disabled", name);
383 			} else {
384 				journal_clear();
385  				sblock.fs_flags &= ~FS_SUJ;
386 				sblock.fs_sujfree = 0;
387  				warnx("%s cleared but soft updates still set.",
388 				    name);
389 
390 				warnx("remove .sujournal to reclaim space");
391 			}
392  		}
393 	}
394 	if (Jflag) {
395 		name = "gjournal";
396 		if (strcmp(Jvalue, "enable") == 0) {
397 			if (sblock.fs_flags & FS_GJOURNAL) {
398 				warnx("%s remains unchanged as enabled", name);
399 			} else {
400 				sblock.fs_flags |= FS_GJOURNAL;
401 				warnx("%s set", name);
402 			}
403 		} else if (strcmp(Jvalue, "disable") == 0) {
404 			if ((~sblock.fs_flags & FS_GJOURNAL) ==
405 			    FS_GJOURNAL) {
406 				warnx("%s remains unchanged as disabled",
407 				    name);
408 			} else {
409 				sblock.fs_flags &= ~FS_GJOURNAL;
410 				warnx("%s cleared", name);
411 			}
412 		}
413 	}
414 	if (kflag) {
415 		name = "space to hold for metadata blocks";
416 		if (sblock.fs_metaspace == kvalue)
417 			warnx("%s remains unchanged as %d", name, kvalue);
418 		else {
419 			kvalue = blknum(&sblock, kvalue);
420 			if (kvalue > sblock.fs_fpg / 2) {
421 				kvalue = blknum(&sblock, sblock.fs_fpg / 2);
422 				warnx("%s cannot exceed half the file system "
423 				    "space", name);
424 			}
425 			warnx("%s changes from %jd to %d",
426 				    name, sblock.fs_metaspace, kvalue);
427 			sblock.fs_metaspace = kvalue;
428 		}
429 	}
430 	if (lflag) {
431 		name = "multilabel";
432 		if (strcmp(lvalue, "enable") == 0) {
433 			if (sblock.fs_flags & FS_MULTILABEL) {
434 				warnx("%s remains unchanged as enabled", name);
435 			} else {
436 				sblock.fs_flags |= FS_MULTILABEL;
437 				warnx("%s set", name);
438 			}
439 		} else if (strcmp(lvalue, "disable") == 0) {
440 			if ((~sblock.fs_flags & FS_MULTILABEL) ==
441 			    FS_MULTILABEL) {
442 				warnx("%s remains unchanged as disabled",
443 				    name);
444 			} else {
445 				sblock.fs_flags &= ~FS_MULTILABEL;
446 				warnx("%s cleared", name);
447 			}
448 		}
449 	}
450 	if (mflag) {
451 		name = "minimum percentage of free space";
452 		if (sblock.fs_minfree == mvalue)
453 			warnx("%s remains unchanged as %d%%", name, mvalue);
454 		else {
455 			warnx("%s changes from %d%% to %d%%",
456 				    name, sblock.fs_minfree, mvalue);
457 			sblock.fs_minfree = mvalue;
458 			if (mvalue >= MINFREE && sblock.fs_optim == FS_OPTSPACE)
459 				warnx(OPTWARN, "time", ">=", MINFREE);
460 			if (mvalue < MINFREE && sblock.fs_optim == FS_OPTTIME)
461 				warnx(OPTWARN, "space", "<", MINFREE);
462 		}
463 	}
464 	if (Nflag) {
465 		name = "NFSv4 ACLs";
466 		if (strcmp(Nvalue, "enable") == 0) {
467 			if (sblock.fs_flags & FS_NFS4ACLS) {
468 				warnx("%s remains unchanged as enabled", name);
469 			} else if (sblock.fs_flags & FS_ACLS) {
470 				warnx("%s and POSIX.1e ACLs are mutually "
471 				    "exclusive", name);
472 			} else {
473 				sblock.fs_flags |= FS_NFS4ACLS;
474 				warnx("%s set", name);
475 			}
476 		} else if (strcmp(Nvalue, "disable") == 0) {
477 			if ((~sblock.fs_flags & FS_NFS4ACLS) ==
478 			    FS_NFS4ACLS) {
479 				warnx("%s remains unchanged as disabled",
480 				    name);
481 			} else {
482 				sblock.fs_flags &= ~FS_NFS4ACLS;
483 				warnx("%s cleared", name);
484 			}
485 		}
486 	}
487 	if (nflag) {
488  		name = "soft updates";
489  		if (strcmp(nvalue, "enable") == 0) {
490 			if (sblock.fs_flags & FS_DOSOFTDEP)
491 				warnx("%s remains unchanged as enabled", name);
492 			else if (sblock.fs_clean == 0) {
493 				warnx("%s cannot be enabled until fsck is run",
494 				    name);
495 			} else {
496  				sblock.fs_flags |= FS_DOSOFTDEP;
497  				warnx("%s set", name);
498 			}
499  		} else if (strcmp(nvalue, "disable") == 0) {
500 			if ((~sblock.fs_flags & FS_DOSOFTDEP) == FS_DOSOFTDEP)
501 				warnx("%s remains unchanged as disabled", name);
502 			else {
503  				sblock.fs_flags &= ~FS_DOSOFTDEP;
504  				warnx("%s cleared", name);
505 			}
506  		}
507 	}
508 	if (oflag) {
509 		name = "optimization preference";
510 		chg[FS_OPTSPACE] = "space";
511 		chg[FS_OPTTIME] = "time";
512 		if (sblock.fs_optim == ovalue)
513 			warnx("%s remains unchanged as %s", name, chg[ovalue]);
514 		else {
515 			warnx("%s changes from %s to %s",
516 				    name, chg[sblock.fs_optim], chg[ovalue]);
517 			sblock.fs_optim = ovalue;
518 			if (sblock.fs_minfree >= MINFREE &&
519 			    ovalue == FS_OPTSPACE)
520 				warnx(OPTWARN, "time", ">=", MINFREE);
521 			if (sblock.fs_minfree < MINFREE && ovalue == FS_OPTTIME)
522 				warnx(OPTWARN, "space", "<", MINFREE);
523 		}
524 	}
525 	if (sflag) {
526 		name = "expected number of files per directory";
527 		if (sblock.fs_avgfpdir == (unsigned)svalue) {
528 			warnx("%s remains unchanged as %d", name, svalue);
529 		}
530 		else {
531 			warnx("%s changes from %d to %d",
532 					name, sblock.fs_avgfpdir, svalue);
533 			sblock.fs_avgfpdir = svalue;
534 		}
535 	}
536 	if (tflag) {
537 		name = "issue TRIM to the disk";
538  		if (strcmp(tvalue, "enable") == 0) {
539 			if (sblock.fs_flags & FS_TRIM)
540 				warnx("%s remains unchanged as enabled", name);
541 			else {
542  				sblock.fs_flags |= FS_TRIM;
543  				warnx("%s set", name);
544 			}
545  		} else if (strcmp(tvalue, "disable") == 0) {
546 			if ((~sblock.fs_flags & FS_TRIM) == FS_TRIM)
547 				warnx("%s remains unchanged as disabled", name);
548 			else {
549  				sblock.fs_flags &= ~FS_TRIM;
550  				warnx("%s cleared", name);
551 			}
552  		}
553 	}
554 
555 	if (sbwrite(&disk, Aflag) == -1)
556 		goto err;
557 	ufs_disk_close(&disk);
558 	if (active) {
559 		bzero(&args, sizeof(args));
560 		if (mount("ufs", on,
561 		    stfs.f_flags | MNT_UPDATE | MNT_RELOAD, &args) < 0)
562 			err(9, "%s: reload", special);
563 		warnx("file system reloaded");
564 	}
565 	exit(0);
566 err:
567 	if (disk.d_error != NULL)
568 		errx(11, "%s: %s", special, disk.d_error);
569 	else
570 		err(12, "%s", special);
571 }
572 
573 static void
574 sbdirty(void)
575 {
576 	disk.d_fs.fs_flags |= FS_UNCLEAN | FS_NEEDSFSCK;
577 	disk.d_fs.fs_clean = 0;
578 }
579 
580 static ufs2_daddr_t
581 journal_balloc(void)
582 {
583 	ufs2_daddr_t blk;
584 	struct cg *cgp;
585 	int valid;
586 	static int contig = 1;
587 
588 	cgp = &disk.d_cg;
589 	for (;;) {
590 		blk = cgballoc(&disk);
591 		if (blk > 0)
592 			break;
593 		/*
594 		 * If we failed to allocate a block from this cg, move to
595 		 * the next.
596 		 */
597 		if (cgwrite(&disk) < 0) {
598 			warn("Failed to write updated cg");
599 			return (-1);
600 		}
601 		while ((valid = cgread(&disk)) == 1) {
602 			/*
603 			 * Try to minimize fragmentation by requiring a minimum
604 			 * number of blocks present.
605 			 */
606 			if (cgp->cg_cs.cs_nbfree > 256 * 1024)
607 				break;
608 			if (contig == 0 && cgp->cg_cs.cs_nbfree)
609 				break;
610 		}
611 		if (valid)
612 			continue;
613 		/*
614 		 * Try once through looking only for large contiguous regions
615 		 * and again taking any space we can find.
616 		 */
617 		if (contig) {
618 			contig = 0;
619 			disk.d_ccg = 0;
620 			warnx("Journal file fragmented.");
621 			continue;
622 		}
623 		warnx("Failed to find sufficient free blocks for the journal");
624 		return -1;
625 	}
626 	if (bwrite(&disk, fsbtodb(&sblock, blk), clrbuf,
627 	    sblock.fs_bsize) <= 0) {
628 		warn("Failed to initialize new block");
629 		return -1;
630 	}
631 	return (blk);
632 }
633 
634 /*
635  * Search a directory block for the SUJ_FILE.
636  */
637 static ino_t
638 dir_search(ufs2_daddr_t blk, int bytes)
639 {
640 	char block[MAXBSIZE];
641 	struct direct *dp;
642 	int off;
643 
644 	if (bread(&disk, fsbtodb(&sblock, blk), block, bytes) <= 0) {
645 		warn("Failed to read dir block");
646 		return (-1);
647 	}
648 	for (off = 0; off < bytes; off += dp->d_reclen) {
649 		dp = (struct direct *)&block[off];
650 		if (dp->d_reclen == 0)
651 			break;
652 		if (dp->d_ino == 0)
653 			continue;
654 		if (dp->d_namlen != strlen(SUJ_FILE))
655 			continue;
656 		if (bcmp(dp->d_name, SUJ_FILE, dp->d_namlen) != 0)
657 			continue;
658 		return (dp->d_ino);
659 	}
660 
661 	return (0);
662 }
663 
664 /*
665  * Search in the ROOTINO for the SUJ_FILE.  If it exists we can not enable
666  * journaling.
667  */
668 static ino_t
669 journal_findfile(void)
670 {
671 	struct ufs1_dinode *dp1;
672 	struct ufs2_dinode *dp2;
673 	ino_t ino;
674 	int mode;
675 	void *ip;
676 	int i;
677 
678 	if (getino(&disk, &ip, ROOTINO, &mode) != 0) {
679 		warn("Failed to get root inode");
680 		return (-1);
681 	}
682 	dp2 = ip;
683 	dp1 = ip;
684 	if (sblock.fs_magic == FS_UFS1_MAGIC) {
685 		if ((off_t)dp1->di_size >= lblktosize(&sblock, NDADDR)) {
686 			warnx("ROOTINO extends beyond direct blocks.");
687 			return (-1);
688 		}
689 		for (i = 0; i < NDADDR; i++) {
690 			if (dp1->di_db[i] == 0)
691 				break;
692 			if ((ino = dir_search(dp1->di_db[i],
693 			    sblksize(&sblock, (off_t)dp1->di_size, i))) != 0)
694 				return (ino);
695 		}
696 	} else {
697 		if ((off_t)dp2->di_size >= lblktosize(&sblock, NDADDR)) {
698 			warnx("ROOTINO extends beyond direct blocks.");
699 			return (-1);
700 		}
701 		for (i = 0; i < NDADDR; i++) {
702 			if (dp2->di_db[i] == 0)
703 				break;
704 			if ((ino = dir_search(dp2->di_db[i],
705 			    sblksize(&sblock, (off_t)dp2->di_size, i))) != 0)
706 				return (ino);
707 		}
708 	}
709 
710 	return (0);
711 }
712 
713 static void
714 dir_clear_block(const char *block, off_t off)
715 {
716 	struct direct *dp;
717 
718 	for (; off < sblock.fs_bsize; off += DIRBLKSIZ) {
719 		dp = (struct direct *)&block[off];
720 		dp->d_ino = 0;
721 		dp->d_reclen = DIRBLKSIZ;
722 		dp->d_type = DT_UNKNOWN;
723 	}
724 }
725 
726 /*
727  * Insert the journal at inode 'ino' into directory blk 'blk' at the first
728  * free offset of 'off'.  DIRBLKSIZ blocks after off are initialized as
729  * empty.
730  */
731 static int
732 dir_insert(ufs2_daddr_t blk, off_t off, ino_t ino)
733 {
734 	struct direct *dp;
735 	char block[MAXBSIZE];
736 
737 	if (bread(&disk, fsbtodb(&sblock, blk), block, sblock.fs_bsize) <= 0) {
738 		warn("Failed to read dir block");
739 		return (-1);
740 	}
741 	bzero(&block[off], sblock.fs_bsize - off);
742 	dp = (struct direct *)&block[off];
743 	dp->d_ino = ino;
744 	dp->d_reclen = DIRBLKSIZ;
745 	dp->d_type = DT_REG;
746 	dp->d_namlen = strlen(SUJ_FILE);
747 	bcopy(SUJ_FILE, &dp->d_name, strlen(SUJ_FILE));
748 	dir_clear_block(block, off + DIRBLKSIZ);
749 	if (bwrite(&disk, fsbtodb(&sblock, blk), block, sblock.fs_bsize) <= 0) {
750 		warn("Failed to write dir block");
751 		return (-1);
752 	}
753 	return (0);
754 }
755 
756 /*
757  * Extend a directory block in 'blk' by copying it to a full size block
758  * and inserting the new journal inode into .sujournal.
759  */
760 static int
761 dir_extend(ufs2_daddr_t blk, ufs2_daddr_t nblk, off_t size, ino_t ino)
762 {
763 	char block[MAXBSIZE];
764 
765 	if (bread(&disk, fsbtodb(&sblock, blk), block,
766 	    roundup(size, sblock.fs_fsize)) <= 0) {
767 		warn("Failed to read dir block");
768 		return (-1);
769 	}
770 	dir_clear_block(block, size);
771 	if (bwrite(&disk, fsbtodb(&sblock, nblk), block, sblock.fs_bsize)
772 	    <= 0) {
773 		warn("Failed to write dir block");
774 		return (-1);
775 	}
776 
777 	return (dir_insert(nblk, size, ino));
778 }
779 
780 /*
781  * Insert the journal file into the ROOTINO directory.  We always extend the
782  * last frag
783  */
784 static int
785 journal_insertfile(ino_t ino)
786 {
787 	struct ufs1_dinode *dp1;
788 	struct ufs2_dinode *dp2;
789 	void *ip;
790 	ufs2_daddr_t nblk;
791 	ufs2_daddr_t blk;
792 	ufs_lbn_t lbn;
793 	int size;
794 	int mode;
795 	int off;
796 
797 	if (getino(&disk, &ip, ROOTINO, &mode) != 0) {
798 		warn("Failed to get root inode");
799 		sbdirty();
800 		return (-1);
801 	}
802 	dp2 = ip;
803 	dp1 = ip;
804 	blk = 0;
805 	size = 0;
806 	nblk = journal_balloc();
807 	if (nblk <= 0)
808 		return (-1);
809 	/*
810 	 * For simplicity sake we aways extend the ROOTINO into a new
811 	 * directory block rather than searching for space and inserting
812 	 * into an existing block.  However, if the rootino has frags
813 	 * have to free them and extend the block.
814 	 */
815 	if (sblock.fs_magic == FS_UFS1_MAGIC) {
816 		lbn = lblkno(&sblock, dp1->di_size);
817 		off = blkoff(&sblock, dp1->di_size);
818 		blk = dp1->di_db[lbn];
819 		size = sblksize(&sblock, (off_t)dp1->di_size, lbn);
820 	} else {
821 		lbn = lblkno(&sblock, dp2->di_size);
822 		off = blkoff(&sblock, dp2->di_size);
823 		blk = dp2->di_db[lbn];
824 		size = sblksize(&sblock, (off_t)dp2->di_size, lbn);
825 	}
826 	if (off != 0) {
827 		if (dir_extend(blk, nblk, off, ino) == -1)
828 			return (-1);
829 	} else {
830 		blk = 0;
831 		if (dir_insert(nblk, 0, ino) == -1)
832 			return (-1);
833 	}
834 	if (sblock.fs_magic == FS_UFS1_MAGIC) {
835 		dp1->di_blocks += (sblock.fs_bsize - size) / DEV_BSIZE;
836 		dp1->di_db[lbn] = nblk;
837 		dp1->di_size = lblktosize(&sblock, lbn+1);
838 	} else {
839 		dp2->di_blocks += (sblock.fs_bsize - size) / DEV_BSIZE;
840 		dp2->di_db[lbn] = nblk;
841 		dp2->di_size = lblktosize(&sblock, lbn+1);
842 	}
843 	if (putino(&disk) < 0) {
844 		warn("Failed to write root inode");
845 		return (-1);
846 	}
847 	if (cgwrite(&disk) < 0) {
848 		warn("Failed to write updated cg");
849 		sbdirty();
850 		return (-1);
851 	}
852 	if (blk) {
853 		if (cgbfree(&disk, blk, size) < 0) {
854 			warn("Failed to write cg");
855 			return (-1);
856 		}
857 	}
858 
859 	return (0);
860 }
861 
862 static int
863 indir_fill(ufs2_daddr_t blk, int level, int *resid)
864 {
865 	char indirbuf[MAXBSIZE];
866 	ufs1_daddr_t *bap1;
867 	ufs2_daddr_t *bap2;
868 	ufs2_daddr_t nblk;
869 	int ncnt;
870 	int cnt;
871 	int i;
872 
873 	bzero(indirbuf, sizeof(indirbuf));
874 	bap1 = (ufs1_daddr_t *)indirbuf;
875 	bap2 = (void *)bap1;
876 	cnt = 0;
877 	for (i = 0; i < NINDIR(&sblock) && *resid != 0; i++) {
878 		nblk = journal_balloc();
879 		if (nblk <= 0)
880 			return (-1);
881 		cnt++;
882 		if (sblock.fs_magic == FS_UFS1_MAGIC)
883 			*bap1++ = nblk;
884 		else
885 			*bap2++ = nblk;
886 		if (level != 0) {
887 			ncnt = indir_fill(nblk, level - 1, resid);
888 			if (ncnt <= 0)
889 				return (-1);
890 			cnt += ncnt;
891 		} else
892 			(*resid)--;
893 	}
894 	if (bwrite(&disk, fsbtodb(&sblock, blk), indirbuf,
895 	    sblock.fs_bsize) <= 0) {
896 		warn("Failed to write indirect");
897 		return (-1);
898 	}
899 	return (cnt);
900 }
901 
902 /*
903  * Clear the flag bits so the journal can be removed.
904  */
905 static void
906 journal_clear(void)
907 {
908 	struct ufs1_dinode *dp1;
909 	struct ufs2_dinode *dp2;
910 	ino_t ino;
911 	int mode;
912 	void *ip;
913 
914 	ino = journal_findfile();
915 	if (ino == (ino_t)-1 || ino == 0) {
916 		warnx("Journal file does not exist");
917 		return;
918 	}
919 	printf("Clearing journal flags from inode %ju\n", (uintmax_t)ino);
920 	if (getino(&disk, &ip, ino, &mode) != 0) {
921 		warn("Failed to get journal inode");
922 		return;
923 	}
924 	dp2 = ip;
925 	dp1 = ip;
926 	if (sblock.fs_magic == FS_UFS1_MAGIC)
927 		dp1->di_flags = 0;
928 	else
929 		dp2->di_flags = 0;
930 	if (putino(&disk) < 0) {
931 		warn("Failed to write journal inode");
932 		return;
933 	}
934 }
935 
936 static int
937 journal_alloc(int64_t size)
938 {
939 	struct ufs1_dinode *dp1;
940 	struct ufs2_dinode *dp2;
941 	ufs2_daddr_t blk;
942 	void *ip;
943 	struct cg *cgp;
944 	int resid;
945 	ino_t ino;
946 	int blks;
947 	int mode;
948 	time_t utime;
949 	int i;
950 
951 	cgp = &disk.d_cg;
952 	ino = 0;
953 
954 	/*
955 	 * If the journal file exists we can't allocate it.
956 	 */
957 	ino = journal_findfile();
958 	if (ino == (ino_t)-1)
959 		return (-1);
960 	if (ino > 0) {
961 		warnx("Journal file %s already exists, please remove.",
962 		    SUJ_FILE);
963 		return (-1);
964 	}
965 	/*
966 	 * If the user didn't supply a size pick one based on the filesystem
967 	 * size constrained with hardcoded MIN and MAX values.  We opt for
968 	 * 1/1024th of the filesystem up to MAX but not exceeding one CG and
969 	 * not less than the MIN.
970 	 */
971 	if (size == 0) {
972 		size = (sblock.fs_size * sblock.fs_bsize) / 1024;
973 		size = MIN(SUJ_MAX, size);
974 		if (size / sblock.fs_fsize > sblock.fs_fpg)
975 			size = sblock.fs_fpg * sblock.fs_fsize;
976 		size = MAX(SUJ_MIN, size);
977 		/* fsck does not support fragments in journal files. */
978 		size = roundup(size, sblock.fs_bsize);
979 	}
980 	resid = blocks = size / sblock.fs_bsize;
981 	if (sblock.fs_cstotal.cs_nbfree < blocks) {
982 		warn("Insufficient free space for %jd byte journal", size);
983 		return (-1);
984 	}
985 	/*
986 	 * Find a cg with enough blocks to satisfy the journal
987 	 * size.  Presently the journal does not span cgs.
988 	 */
989 	while (cgread(&disk) == 1) {
990 		if (cgp->cg_cs.cs_nifree == 0)
991 			continue;
992 		ino = cgialloc(&disk);
993 		if (ino <= 0)
994 			break;
995 		printf("Using inode %ju in cg %d for %jd byte journal\n",
996 		    (uintmax_t)ino, cgp->cg_cgx, size);
997 		if (getino(&disk, &ip, ino, &mode) != 0) {
998 			warn("Failed to get allocated inode");
999 			sbdirty();
1000 			goto out;
1001 		}
1002 		/*
1003 		 * We leave fields unrelated to the number of allocated
1004 		 * blocks and size uninitialized.  This causes legacy
1005 		 * fsck implementations to clear the inode.
1006 		 */
1007 		dp2 = ip;
1008 		dp1 = ip;
1009 		time(&utime);
1010 		if (sblock.fs_magic == FS_UFS1_MAGIC) {
1011 			bzero(dp1, sizeof(*dp1));
1012 			dp1->di_size = size;
1013 			dp1->di_mode = IFREG | IREAD;
1014 			dp1->di_nlink = 1;
1015 			dp1->di_flags = SF_IMMUTABLE | SF_NOUNLINK | UF_NODUMP;
1016 			dp1->di_atime = utime;
1017 			dp1->di_mtime = utime;
1018 			dp1->di_ctime = utime;
1019 		} else {
1020 			bzero(dp2, sizeof(*dp2));
1021 			dp2->di_size = size;
1022 			dp2->di_mode = IFREG | IREAD;
1023 			dp2->di_nlink = 1;
1024 			dp2->di_flags = SF_IMMUTABLE | SF_NOUNLINK | UF_NODUMP;
1025 			dp2->di_atime = utime;
1026 			dp2->di_mtime = utime;
1027 			dp2->di_ctime = utime;
1028 			dp2->di_birthtime = utime;
1029 		}
1030 		for (i = 0; i < NDADDR && resid; i++, resid--) {
1031 			blk = journal_balloc();
1032 			if (blk <= 0)
1033 				goto out;
1034 			if (sblock.fs_magic == FS_UFS1_MAGIC) {
1035 				dp1->di_db[i] = blk;
1036 				dp1->di_blocks++;
1037 			} else {
1038 				dp2->di_db[i] = blk;
1039 				dp2->di_blocks++;
1040 			}
1041 		}
1042 		for (i = 0; i < NIADDR && resid; i++) {
1043 			blk = journal_balloc();
1044 			if (blk <= 0)
1045 				goto out;
1046 			blks = indir_fill(blk, i, &resid) + 1;
1047 			if (blks <= 0) {
1048 				sbdirty();
1049 				goto out;
1050 			}
1051 			if (sblock.fs_magic == FS_UFS1_MAGIC) {
1052 				dp1->di_ib[i] = blk;
1053 				dp1->di_blocks += blks;
1054 			} else {
1055 				dp2->di_ib[i] = blk;
1056 				dp2->di_blocks += blks;
1057 			}
1058 		}
1059 		if (sblock.fs_magic == FS_UFS1_MAGIC)
1060 			dp1->di_blocks *= sblock.fs_bsize / disk.d_bsize;
1061 		else
1062 			dp2->di_blocks *= sblock.fs_bsize / disk.d_bsize;
1063 		if (putino(&disk) < 0) {
1064 			warn("Failed to write inode");
1065 			sbdirty();
1066 			return (-1);
1067 		}
1068 		if (cgwrite(&disk) < 0) {
1069 			warn("Failed to write updated cg");
1070 			sbdirty();
1071 			return (-1);
1072 		}
1073 		if (journal_insertfile(ino) < 0) {
1074 			sbdirty();
1075 			return (-1);
1076 		}
1077 		sblock.fs_sujfree = 0;
1078 		return (0);
1079 	}
1080 	warnx("Insufficient free space for the journal.");
1081 out:
1082 	return (-1);
1083 }
1084 
1085 static void
1086 usage(void)
1087 {
1088 	fprintf(stderr, "%s\n%s\n%s\n%s\n%s\n%s\n",
1089 "usage: tunefs [-A] [-a enable | disable] [-e maxbpg] [-f avgfilesize]",
1090 "              [-J enable | disable] [-j enable | disable] [-k metaspace]",
1091 "              [-L volname] [-l enable | disable] [-m minfree]",
1092 "              [-N enable | disable] [-n enable | disable]",
1093 "              [-o space | time] [-p] [-s avgfpdir] [-t enable | disable]",
1094 "              special | filesystem");
1095 	exit(2);
1096 }
1097 
1098 static void
1099 printfs(void)
1100 {
1101 	warnx("POSIX.1e ACLs: (-a)                                %s",
1102 		(sblock.fs_flags & FS_ACLS)? "enabled" : "disabled");
1103 	warnx("NFSv4 ACLs: (-N)                                   %s",
1104 		(sblock.fs_flags & FS_NFS4ACLS)? "enabled" : "disabled");
1105 	warnx("MAC multilabel: (-l)                               %s",
1106 		(sblock.fs_flags & FS_MULTILABEL)? "enabled" : "disabled");
1107 	warnx("soft updates: (-n)                                 %s",
1108 		(sblock.fs_flags & FS_DOSOFTDEP)? "enabled" : "disabled");
1109 	warnx("soft update journaling: (-j)                       %s",
1110 		(sblock.fs_flags & FS_SUJ)? "enabled" : "disabled");
1111 	warnx("gjournal: (-J)                                     %s",
1112 		(sblock.fs_flags & FS_GJOURNAL)? "enabled" : "disabled");
1113 	warnx("trim: (-t)                                         %s",
1114 		(sblock.fs_flags & FS_TRIM)? "enabled" : "disabled");
1115 	warnx("maximum blocks per file in a cylinder group: (-e)  %d",
1116 	      sblock.fs_maxbpg);
1117 	warnx("average file size: (-f)                            %d",
1118 	      sblock.fs_avgfilesize);
1119 	warnx("average number of files in a directory: (-s)       %d",
1120 	      sblock.fs_avgfpdir);
1121 	warnx("minimum percentage of free space: (-m)             %d%%",
1122 	      sblock.fs_minfree);
1123 	warnx("space to hold for metadata blocks: (-k)            %jd",
1124 	      sblock.fs_metaspace);
1125 	warnx("optimization preference: (-o)                      %s",
1126 	      sblock.fs_optim == FS_OPTSPACE ? "space" : "time");
1127 	if (sblock.fs_minfree >= MINFREE &&
1128 	    sblock.fs_optim == FS_OPTSPACE)
1129 		warnx(OPTWARN, "time", ">=", MINFREE);
1130 	if (sblock.fs_minfree < MINFREE &&
1131 	    sblock.fs_optim == FS_OPTTIME)
1132 		warnx(OPTWARN, "space", "<", MINFREE);
1133 	warnx("volume label: (-L)                                 %s",
1134 		sblock.fs_volname);
1135 }
1136