xref: /freebsd/sbin/tunefs/tunefs.c (revision fd253945ac76a54ff9c11cf02f5458561f711866)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1983, 1993
5  *	The Regents of the University of California.  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  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #if 0
33 #ifndef lint
34 static const char copyright[] =
35 "@(#) Copyright (c) 1983, 1993\n\
36 	The Regents of the University of California.  All rights reserved.\n";
37 #endif /* not lint */
38 
39 #ifndef lint
40 static char sccsid[] = "@(#)tunefs.c	8.2 (Berkeley) 4/19/94";
41 #endif /* not lint */
42 #endif
43 #include <sys/cdefs.h>
44 __FBSDID("$FreeBSD$");
45 
46 /*
47  * tunefs: change layout parameters to an existing file system.
48  */
49 #include <sys/param.h>
50 #include <sys/mount.h>
51 #include <sys/disklabel.h>
52 #include <sys/stat.h>
53 
54 #include <ufs/ufs/ufsmount.h>
55 #include <ufs/ufs/dinode.h>
56 #include <ufs/ffs/fs.h>
57 #include <ufs/ufs/dir.h>
58 
59 #include <ctype.h>
60 #include <err.h>
61 #include <fcntl.h>
62 #include <fstab.h>
63 #include <libufs.h>
64 #include <mntopts.h>
65 #include <paths.h>
66 #include <stdio.h>
67 #include <stdlib.h>
68 #include <stdint.h>
69 #include <string.h>
70 #include <time.h>
71 #include <unistd.h>
72 
73 /* the optimization warning string template */
74 #define	OPTWARN	"should optimize for %s with minfree %s %d%%"
75 
76 static int blocks;
77 static char clrbuf[MAXBSIZE];
78 static struct uufsd disk;
79 #define	sblock disk.d_fs
80 
81 static void usage(void);
82 static void printfs(void);
83 static int journal_alloc(int64_t size);
84 static void journal_clear(void);
85 static void sbdirty(void);
86 
87 int
88 main(int argc, char *argv[])
89 {
90 	const char *avalue, *jvalue, *Jvalue, *Lvalue, *lvalue, *Nvalue, *nvalue;
91 	const char *tvalue;
92 	const char *special, *on;
93 	const char *name;
94 	int active;
95 	int Aflag, aflag, eflag, evalue, fflag, fvalue, jflag, Jflag, kflag;
96 	int kvalue, Lflag, lflag, mflag, mvalue, Nflag, nflag, oflag, ovalue;
97 	int pflag, sflag, svalue, Svalue, tflag;
98 	int ch, found_arg, i;
99 	int iovlen = 0;
100 	const char *chg[2];
101 	struct statfs stfs;
102 	struct iovec *iov = NULL;
103 	char errmsg[255] = {0};
104 
105 	if (argc < 3)
106 		usage();
107 	Aflag = aflag = eflag = fflag = jflag = Jflag = kflag = Lflag = 0;
108 	lflag = mflag = Nflag = nflag = oflag = pflag = sflag = tflag = 0;
109 	avalue = jvalue = Jvalue = Lvalue = lvalue = Nvalue = nvalue = NULL;
110 	evalue = fvalue = mvalue = ovalue = svalue = Svalue = 0;
111 	active = 0;
112 	found_arg = 0;		/* At least one arg is required. */
113 	while ((ch = getopt(argc, argv, "Aa:e:f:j:J:k:L:l:m:N:n:o:ps:S:t:"))
114 	    != -1)
115 		switch (ch) {
116 
117 		case 'A':
118 			found_arg = 1;
119 			Aflag++;
120 			break;
121 
122 		case 'a':
123 			found_arg = 1;
124 			name = "POSIX.1e ACLs";
125 			avalue = optarg;
126 			if (strcmp(avalue, "enable") &&
127 			    strcmp(avalue, "disable")) {
128 				errx(10, "bad %s (options are %s)",
129 				    name, "`enable' or `disable'");
130 			}
131 			aflag = 1;
132 			break;
133 
134 		case 'e':
135 			found_arg = 1;
136 			name = "maximum blocks per file in a cylinder group";
137 			evalue = atoi(optarg);
138 			if (evalue < 1)
139 				errx(10, "%s must be >= 1 (was %s)",
140 				    name, optarg);
141 			eflag = 1;
142 			break;
143 
144 		case 'f':
145 			found_arg = 1;
146 			name = "average file size";
147 			fvalue = atoi(optarg);
148 			if (fvalue < 1)
149 				errx(10, "%s must be >= 1 (was %s)",
150 				    name, optarg);
151 			fflag = 1;
152 			break;
153 
154 		case 'j':
155 			found_arg = 1;
156 			name = "softdep journaled file system";
157 			jvalue = optarg;
158 			if (strcmp(jvalue, "enable") &&
159 			    strcmp(jvalue, "disable")) {
160 				errx(10, "bad %s (options are %s)",
161 				    name, "`enable' or `disable'");
162 			}
163 			jflag = 1;
164 			break;
165 
166 		case 'J':
167 			found_arg = 1;
168 			name = "gjournaled file system";
169 			Jvalue = optarg;
170 			if (strcmp(Jvalue, "enable") &&
171 			    strcmp(Jvalue, "disable")) {
172 				errx(10, "bad %s (options are %s)",
173 				    name, "`enable' or `disable'");
174 			}
175 			Jflag = 1;
176 			break;
177 
178 		case 'k':
179 			found_arg = 1;
180 			name = "space to hold for metadata blocks";
181 			kvalue = atoi(optarg);
182 			if (kvalue < 0)
183 				errx(10, "bad %s (%s)", name, optarg);
184 			kflag = 1;
185 			break;
186 
187 		case 'L':
188 			found_arg = 1;
189 			name = "volume label";
190 			Lvalue = optarg;
191 			i = -1;
192 			while (isalnum(Lvalue[++i]));
193 			if (Lvalue[i] != '\0') {
194 				errx(10,
195 				"bad %s. Valid characters are alphanumerics.",
196 				    name);
197 			}
198 			if (strlen(Lvalue) >= MAXVOLLEN) {
199 				errx(10, "bad %s. Length is longer than %d.",
200 				    name, MAXVOLLEN - 1);
201 			}
202 			Lflag = 1;
203 			break;
204 
205 		case 'l':
206 			found_arg = 1;
207 			name = "multilabel MAC file system";
208 			lvalue = optarg;
209 			if (strcmp(lvalue, "enable") &&
210 			    strcmp(lvalue, "disable")) {
211 				errx(10, "bad %s (options are %s)",
212 				    name, "`enable' or `disable'");
213 			}
214 			lflag = 1;
215 			break;
216 
217 		case 'm':
218 			found_arg = 1;
219 			name = "minimum percentage of free space";
220 			mvalue = atoi(optarg);
221 			if (mvalue < 0 || mvalue > 99)
222 				errx(10, "bad %s (%s)", name, optarg);
223 			mflag = 1;
224 			break;
225 
226 		case 'N':
227 			found_arg = 1;
228 			name = "NFSv4 ACLs";
229 			Nvalue = optarg;
230 			if (strcmp(Nvalue, "enable") &&
231 			    strcmp(Nvalue, "disable")) {
232 				errx(10, "bad %s (options are %s)",
233 				    name, "`enable' or `disable'");
234 			}
235 			Nflag = 1;
236 			break;
237 
238 		case 'n':
239 			found_arg = 1;
240 			name = "soft updates";
241 			nvalue = optarg;
242 			if (strcmp(nvalue, "enable") != 0 &&
243 			    strcmp(nvalue, "disable") != 0) {
244 				errx(10, "bad %s (options are %s)",
245 				    name, "`enable' or `disable'");
246 			}
247 			nflag = 1;
248 			break;
249 
250 		case 'o':
251 			found_arg = 1;
252 			name = "optimization preference";
253 			if (strcmp(optarg, "space") == 0)
254 				ovalue = FS_OPTSPACE;
255 			else if (strcmp(optarg, "time") == 0)
256 				ovalue = FS_OPTTIME;
257 			else
258 				errx(10,
259 				    "bad %s (options are `space' or `time')",
260 				    name);
261 			oflag = 1;
262 			break;
263 
264 		case 'p':
265 			found_arg = 1;
266 			pflag = 1;
267 			break;
268 
269 		case 's':
270 			found_arg = 1;
271 			name = "expected number of files per directory";
272 			svalue = atoi(optarg);
273 			if (svalue < 1)
274 				errx(10, "%s must be >= 1 (was %s)",
275 				    name, optarg);
276 			sflag = 1;
277 			break;
278 
279 		case 'S':
280 			found_arg = 1;
281 			name = "Softdep Journal Size";
282 			Svalue = atoi(optarg);
283 			if (Svalue < SUJ_MIN)
284 				errx(10, "%s must be >= %d (was %s)",
285 				    name, SUJ_MIN, optarg);
286 			break;
287 
288 		case 't':
289 			found_arg = 1;
290 			name = "trim";
291 			tvalue = optarg;
292 			if (strcmp(tvalue, "enable") != 0 &&
293 			    strcmp(tvalue, "disable") != 0) {
294 				errx(10, "bad %s (options are %s)",
295 				    name, "`enable' or `disable'");
296 			}
297 			tflag = 1;
298 			break;
299 
300 		default:
301 			usage();
302 		}
303 	argc -= optind;
304 	argv += optind;
305 	if (found_arg == 0 || argc != 1)
306 		usage();
307 
308 	on = special = argv[0];
309 	if (ufs_disk_fillout(&disk, special) == -1)
310 		goto err;
311 	if (disk.d_name != special) {
312 		if (statfs(special, &stfs) != 0)
313 			warn("Can't stat %s", special);
314 		if (strcmp(special, stfs.f_mntonname) == 0)
315 			active = 1;
316 	}
317 
318 	if (pflag) {
319 		printfs();
320 		exit(0);
321 	}
322 	if (Lflag) {
323 		name = "volume label";
324 		strncpy(sblock.fs_volname, Lvalue, MAXVOLLEN);
325 	}
326 	if (aflag) {
327 		name = "POSIX.1e ACLs";
328 		if (strcmp(avalue, "enable") == 0) {
329 			if (sblock.fs_flags & FS_ACLS) {
330 				warnx("%s remains unchanged as enabled", name);
331 			} else if (sblock.fs_flags & FS_NFS4ACLS) {
332 				warnx("%s and NFSv4 ACLs are mutually "
333 				    "exclusive", name);
334 			} else {
335 				sblock.fs_flags |= FS_ACLS;
336 				warnx("%s set", name);
337 			}
338 		} else if (strcmp(avalue, "disable") == 0) {
339 			if ((~sblock.fs_flags & FS_ACLS) ==
340 			    FS_ACLS) {
341 				warnx("%s remains unchanged as disabled",
342 				    name);
343 			} else {
344 				sblock.fs_flags &= ~FS_ACLS;
345 				warnx("%s cleared", name);
346 			}
347 		}
348 	}
349 	if (eflag) {
350 		name = "maximum blocks per file in a cylinder group";
351 		if (sblock.fs_maxbpg == evalue)
352 			warnx("%s remains unchanged as %d", name, evalue);
353 		else {
354 			warnx("%s changes from %d to %d",
355 			    name, sblock.fs_maxbpg, evalue);
356 			sblock.fs_maxbpg = evalue;
357 		}
358 	}
359 	if (fflag) {
360 		name = "average file size";
361 		if (sblock.fs_avgfilesize == (unsigned)fvalue) {
362 			warnx("%s remains unchanged as %d", name, fvalue);
363 		}
364 		else {
365 			warnx("%s changes from %d to %d",
366 					name, sblock.fs_avgfilesize, fvalue);
367 			sblock.fs_avgfilesize = fvalue;
368 		}
369 	}
370 	if (jflag) {
371  		name = "soft updates journaling";
372  		if (strcmp(jvalue, "enable") == 0) {
373 			if ((sblock.fs_flags & (FS_DOSOFTDEP | FS_SUJ)) ==
374 			    (FS_DOSOFTDEP | FS_SUJ)) {
375 				warnx("%s remains unchanged as enabled", name);
376 			} else if (sblock.fs_clean == 0) {
377 				warnx("%s cannot be enabled until fsck is run",
378 				    name);
379 			} else if (journal_alloc(Svalue) != 0) {
380 				warnx("%s cannot be enabled", name);
381 			} else {
382  				sblock.fs_flags |= FS_DOSOFTDEP | FS_SUJ;
383  				warnx("%s set", name);
384 			}
385  		} else if (strcmp(jvalue, "disable") == 0) {
386 			if ((~sblock.fs_flags & FS_SUJ) == FS_SUJ) {
387 				warnx("%s remains unchanged as disabled", name);
388 			} else {
389 				journal_clear();
390  				sblock.fs_flags &= ~FS_SUJ;
391 				sblock.fs_sujfree = 0;
392  				warnx("%s cleared but soft updates still set.",
393 				    name);
394 
395 				warnx("remove .sujournal to reclaim space");
396 			}
397  		}
398 	}
399 	if (Jflag) {
400 		name = "gjournal";
401 		if (strcmp(Jvalue, "enable") == 0) {
402 			if (sblock.fs_flags & FS_GJOURNAL) {
403 				warnx("%s remains unchanged as enabled", name);
404 			} else {
405 				sblock.fs_flags |= FS_GJOURNAL;
406 				warnx("%s set", name);
407 			}
408 		} else if (strcmp(Jvalue, "disable") == 0) {
409 			if ((~sblock.fs_flags & FS_GJOURNAL) ==
410 			    FS_GJOURNAL) {
411 				warnx("%s remains unchanged as disabled",
412 				    name);
413 			} else {
414 				sblock.fs_flags &= ~FS_GJOURNAL;
415 				warnx("%s cleared", name);
416 			}
417 		}
418 	}
419 	if (kflag) {
420 		name = "space to hold for metadata blocks";
421 		if (sblock.fs_metaspace == kvalue)
422 			warnx("%s remains unchanged as %d", name, kvalue);
423 		else {
424 			kvalue = blknum(&sblock, kvalue);
425 			if (kvalue > sblock.fs_fpg / 2) {
426 				kvalue = blknum(&sblock, sblock.fs_fpg / 2);
427 				warnx("%s cannot exceed half the file system "
428 				    "space", name);
429 			}
430 			warnx("%s changes from %jd to %d",
431 				    name, sblock.fs_metaspace, kvalue);
432 			sblock.fs_metaspace = kvalue;
433 		}
434 	}
435 	if (lflag) {
436 		name = "multilabel";
437 		if (strcmp(lvalue, "enable") == 0) {
438 			if (sblock.fs_flags & FS_MULTILABEL) {
439 				warnx("%s remains unchanged as enabled", name);
440 			} else {
441 				sblock.fs_flags |= FS_MULTILABEL;
442 				warnx("%s set", name);
443 			}
444 		} else if (strcmp(lvalue, "disable") == 0) {
445 			if ((~sblock.fs_flags & FS_MULTILABEL) ==
446 			    FS_MULTILABEL) {
447 				warnx("%s remains unchanged as disabled",
448 				    name);
449 			} else {
450 				sblock.fs_flags &= ~FS_MULTILABEL;
451 				warnx("%s cleared", name);
452 			}
453 		}
454 	}
455 	if (mflag) {
456 		name = "minimum percentage of free space";
457 		if (sblock.fs_minfree == mvalue)
458 			warnx("%s remains unchanged as %d%%", name, mvalue);
459 		else {
460 			warnx("%s changes from %d%% to %d%%",
461 				    name, sblock.fs_minfree, mvalue);
462 			sblock.fs_minfree = mvalue;
463 			if (mvalue >= MINFREE && sblock.fs_optim == FS_OPTSPACE)
464 				warnx(OPTWARN, "time", ">=", MINFREE);
465 			if (mvalue < MINFREE && sblock.fs_optim == FS_OPTTIME)
466 				warnx(OPTWARN, "space", "<", MINFREE);
467 		}
468 	}
469 	if (Nflag) {
470 		name = "NFSv4 ACLs";
471 		if (strcmp(Nvalue, "enable") == 0) {
472 			if (sblock.fs_flags & FS_NFS4ACLS) {
473 				warnx("%s remains unchanged as enabled", name);
474 			} else if (sblock.fs_flags & FS_ACLS) {
475 				warnx("%s and POSIX.1e ACLs are mutually "
476 				    "exclusive", name);
477 			} else {
478 				sblock.fs_flags |= FS_NFS4ACLS;
479 				warnx("%s set", name);
480 			}
481 		} else if (strcmp(Nvalue, "disable") == 0) {
482 			if ((~sblock.fs_flags & FS_NFS4ACLS) ==
483 			    FS_NFS4ACLS) {
484 				warnx("%s remains unchanged as disabled",
485 				    name);
486 			} else {
487 				sblock.fs_flags &= ~FS_NFS4ACLS;
488 				warnx("%s cleared", name);
489 			}
490 		}
491 	}
492 	if (nflag) {
493  		name = "soft updates";
494  		if (strcmp(nvalue, "enable") == 0) {
495 			if (sblock.fs_flags & FS_DOSOFTDEP)
496 				warnx("%s remains unchanged as enabled", name);
497 			else if (sblock.fs_clean == 0) {
498 				warnx("%s cannot be enabled until fsck is run",
499 				    name);
500 			} else {
501  				sblock.fs_flags |= FS_DOSOFTDEP;
502  				warnx("%s set", name);
503 			}
504  		} else if (strcmp(nvalue, "disable") == 0) {
505 			if ((~sblock.fs_flags & FS_DOSOFTDEP) == FS_DOSOFTDEP)
506 				warnx("%s remains unchanged as disabled", name);
507 			else {
508  				sblock.fs_flags &= ~FS_DOSOFTDEP;
509  				warnx("%s cleared", name);
510 			}
511  		}
512 	}
513 	if (oflag) {
514 		name = "optimization preference";
515 		chg[FS_OPTSPACE] = "space";
516 		chg[FS_OPTTIME] = "time";
517 		if (sblock.fs_optim == ovalue)
518 			warnx("%s remains unchanged as %s", name, chg[ovalue]);
519 		else {
520 			warnx("%s changes from %s to %s",
521 				    name, chg[sblock.fs_optim], chg[ovalue]);
522 			sblock.fs_optim = ovalue;
523 			if (sblock.fs_minfree >= MINFREE &&
524 			    ovalue == FS_OPTSPACE)
525 				warnx(OPTWARN, "time", ">=", MINFREE);
526 			if (sblock.fs_minfree < MINFREE && ovalue == FS_OPTTIME)
527 				warnx(OPTWARN, "space", "<", MINFREE);
528 		}
529 	}
530 	if (sflag) {
531 		name = "expected number of files per directory";
532 		if (sblock.fs_avgfpdir == (unsigned)svalue) {
533 			warnx("%s remains unchanged as %d", name, svalue);
534 		}
535 		else {
536 			warnx("%s changes from %d to %d",
537 					name, sblock.fs_avgfpdir, svalue);
538 			sblock.fs_avgfpdir = svalue;
539 		}
540 	}
541 	if (tflag) {
542 		name = "issue TRIM to the disk";
543  		if (strcmp(tvalue, "enable") == 0) {
544 			if (sblock.fs_flags & FS_TRIM)
545 				warnx("%s remains unchanged as enabled", name);
546 			else {
547  				sblock.fs_flags |= FS_TRIM;
548  				warnx("%s set", name);
549 			}
550  		} else if (strcmp(tvalue, "disable") == 0) {
551 			if ((~sblock.fs_flags & FS_TRIM) == FS_TRIM)
552 				warnx("%s remains unchanged as disabled", name);
553 			else {
554  				sblock.fs_flags &= ~FS_TRIM;
555  				warnx("%s cleared", name);
556 			}
557  		}
558 	}
559 
560 	if (sbwrite(&disk, Aflag) == -1)
561 		goto err;
562 	ufs_disk_close(&disk);
563 	if (active) {
564 		build_iovec_argf(&iov, &iovlen, "fstype", "ufs");
565 		build_iovec_argf(&iov, &iovlen, "fspath", "%s", on);
566 		build_iovec(&iov, &iovlen, "errmsg", errmsg, sizeof(errmsg));
567 		if (nmount(iov, iovlen,
568 		    stfs.f_flags | MNT_UPDATE | MNT_RELOAD) < 0) {
569 			if (errmsg[0])
570 				err(9, "%s: reload: %s", special, errmsg);
571 			else
572 				err(9, "%s: reload", special);
573 		}
574 		warnx("file system reloaded");
575 	}
576 	exit(0);
577 err:
578 	if (disk.d_error != NULL)
579 		errx(11, "%s: %s", special, disk.d_error);
580 	else
581 		err(12, "%s", special);
582 }
583 
584 static void
585 sbdirty(void)
586 {
587 	disk.d_fs.fs_flags |= FS_UNCLEAN | FS_NEEDSFSCK;
588 	disk.d_fs.fs_clean = 0;
589 }
590 
591 static ufs2_daddr_t
592 journal_balloc(void)
593 {
594 	ufs2_daddr_t blk;
595 	struct cg *cgp;
596 	int valid;
597 	static int contig = 1;
598 
599 	cgp = &disk.d_cg;
600 	for (;;) {
601 		blk = cgballoc(&disk);
602 		if (blk > 0)
603 			break;
604 		/*
605 		 * If we failed to allocate a block from this cg, move to
606 		 * the next.
607 		 */
608 		if (cgwrite(&disk) < 0) {
609 			warn("Failed to write updated cg");
610 			return (-1);
611 		}
612 		while ((valid = cgread(&disk)) == 1) {
613 			/*
614 			 * Try to minimize fragmentation by requiring a minimum
615 			 * number of blocks present.
616 			 */
617 			if (cgp->cg_cs.cs_nbfree > 256 * 1024)
618 				break;
619 			if (contig == 0 && cgp->cg_cs.cs_nbfree)
620 				break;
621 		}
622 		if (valid)
623 			continue;
624 		/*
625 		 * Try once through looking only for large contiguous regions
626 		 * and again taking any space we can find.
627 		 */
628 		if (contig) {
629 			contig = 0;
630 			disk.d_ccg = 0;
631 			warnx("Journal file fragmented.");
632 			continue;
633 		}
634 		warnx("Failed to find sufficient free blocks for the journal");
635 		return -1;
636 	}
637 	if (bwrite(&disk, fsbtodb(&sblock, blk), clrbuf,
638 	    sblock.fs_bsize) <= 0) {
639 		warn("Failed to initialize new block");
640 		return -1;
641 	}
642 	return (blk);
643 }
644 
645 /*
646  * Search a directory block for the SUJ_FILE.
647  */
648 static ino_t
649 dir_search(ufs2_daddr_t blk, int bytes)
650 {
651 	char block[MAXBSIZE];
652 	struct direct *dp;
653 	int off;
654 
655 	if (bread(&disk, fsbtodb(&sblock, blk), block, bytes) <= 0) {
656 		warn("Failed to read dir block");
657 		return (-1);
658 	}
659 	for (off = 0; off < bytes; off += dp->d_reclen) {
660 		dp = (struct direct *)&block[off];
661 		if (dp->d_reclen == 0)
662 			break;
663 		if (dp->d_ino == 0)
664 			continue;
665 		if (dp->d_namlen != strlen(SUJ_FILE))
666 			continue;
667 		if (bcmp(dp->d_name, SUJ_FILE, dp->d_namlen) != 0)
668 			continue;
669 		return (dp->d_ino);
670 	}
671 
672 	return (0);
673 }
674 
675 /*
676  * Search in the UFS_ROOTINO for the SUJ_FILE.  If it exists we can not enable
677  * journaling.
678  */
679 static ino_t
680 journal_findfile(void)
681 {
682 	union dinodep dp;
683 	ino_t ino;
684 	int i;
685 
686 	if (getinode(&disk, &dp, UFS_ROOTINO) != 0) {
687 		warn("Failed to get root inode: %s", disk.d_error);
688 		return (-1);
689 	}
690 	if (sblock.fs_magic == FS_UFS1_MAGIC) {
691 		if ((off_t)dp.dp1->di_size >= lblktosize(&sblock, UFS_NDADDR)) {
692 			warnx("UFS_ROOTINO extends beyond direct blocks.");
693 			return (-1);
694 		}
695 		for (i = 0; i < UFS_NDADDR; i++) {
696 			if (dp.dp1->di_db[i] == 0)
697 				break;
698 			if ((ino = dir_search(dp.dp1->di_db[i],
699 			    sblksize(&sblock, (off_t)dp.dp1->di_size, i))) != 0)
700 				return (ino);
701 		}
702 	} else {
703 		if ((off_t)dp.dp2->di_size >= lblktosize(&sblock, UFS_NDADDR)) {
704 			warnx("UFS_ROOTINO extends beyond direct blocks.");
705 			return (-1);
706 		}
707 		for (i = 0; i < UFS_NDADDR; i++) {
708 			if (dp.dp2->di_db[i] == 0)
709 				break;
710 			if ((ino = dir_search(dp.dp2->di_db[i],
711 			    sblksize(&sblock, (off_t)dp.dp2->di_size, i))) != 0)
712 				return (ino);
713 		}
714 	}
715 
716 	return (0);
717 }
718 
719 static void
720 dir_clear_block(const char *block, off_t off)
721 {
722 	struct direct *dp;
723 
724 	for (; off < sblock.fs_bsize; off += DIRBLKSIZ) {
725 		dp = (struct direct *)&block[off];
726 		dp->d_ino = 0;
727 		dp->d_reclen = DIRBLKSIZ;
728 		dp->d_type = DT_UNKNOWN;
729 	}
730 }
731 
732 /*
733  * Insert the journal at inode 'ino' into directory blk 'blk' at the first
734  * free offset of 'off'.  DIRBLKSIZ blocks after off are initialized as
735  * empty.
736  */
737 static int
738 dir_insert(ufs2_daddr_t blk, off_t off, ino_t ino)
739 {
740 	struct direct *dp;
741 	char block[MAXBSIZE];
742 
743 	if (bread(&disk, fsbtodb(&sblock, blk), block, sblock.fs_bsize) <= 0) {
744 		warn("Failed to read dir block");
745 		return (-1);
746 	}
747 	bzero(&block[off], sblock.fs_bsize - off);
748 	dp = (struct direct *)&block[off];
749 	dp->d_ino = ino;
750 	dp->d_reclen = DIRBLKSIZ;
751 	dp->d_type = DT_REG;
752 	dp->d_namlen = strlen(SUJ_FILE);
753 	bcopy(SUJ_FILE, &dp->d_name, strlen(SUJ_FILE));
754 	dir_clear_block(block, off + DIRBLKSIZ);
755 	if (bwrite(&disk, fsbtodb(&sblock, blk), block, sblock.fs_bsize) <= 0) {
756 		warn("Failed to write dir block");
757 		return (-1);
758 	}
759 	return (0);
760 }
761 
762 /*
763  * Extend a directory block in 'blk' by copying it to a full size block
764  * and inserting the new journal inode into .sujournal.
765  */
766 static int
767 dir_extend(ufs2_daddr_t blk, ufs2_daddr_t nblk, off_t size, ino_t ino)
768 {
769 	char block[MAXBSIZE];
770 
771 	if (bread(&disk, fsbtodb(&sblock, blk), block,
772 	    roundup(size, sblock.fs_fsize)) <= 0) {
773 		warn("Failed to read dir block");
774 		return (-1);
775 	}
776 	dir_clear_block(block, size);
777 	if (bwrite(&disk, fsbtodb(&sblock, nblk), block, sblock.fs_bsize)
778 	    <= 0) {
779 		warn("Failed to write dir block");
780 		return (-1);
781 	}
782 
783 	return (dir_insert(nblk, size, ino));
784 }
785 
786 /*
787  * Insert the journal file into the UFS_ROOTINO directory.  We always extend the
788  * last frag
789  */
790 static int
791 journal_insertfile(ino_t ino)
792 {
793 	union dinodep dp;
794 	ufs2_daddr_t nblk;
795 	ufs2_daddr_t blk;
796 	ufs_lbn_t lbn;
797 	int size;
798 	int off;
799 
800 	if (getinode(&disk, &dp, UFS_ROOTINO) != 0) {
801 		warn("Failed to get root inode: %s", disk.d_error);
802 		sbdirty();
803 		return (-1);
804 	}
805 	blk = 0;
806 	size = 0;
807 	nblk = journal_balloc();
808 	if (nblk <= 0)
809 		return (-1);
810 	/*
811 	 * For simplicity sake we aways extend the UFS_ROOTINO into a new
812 	 * directory block rather than searching for space and inserting
813 	 * into an existing block.  However, if the rootino has frags
814 	 * have to free them and extend the block.
815 	 */
816 	if (sblock.fs_magic == FS_UFS1_MAGIC) {
817 		lbn = lblkno(&sblock, dp.dp1->di_size);
818 		off = blkoff(&sblock, dp.dp1->di_size);
819 		blk = dp.dp1->di_db[lbn];
820 		size = sblksize(&sblock, (off_t)dp.dp1->di_size, lbn);
821 	} else {
822 		lbn = lblkno(&sblock, dp.dp2->di_size);
823 		off = blkoff(&sblock, dp.dp2->di_size);
824 		blk = dp.dp2->di_db[lbn];
825 		size = sblksize(&sblock, (off_t)dp.dp2->di_size, lbn);
826 	}
827 	if (off != 0) {
828 		if (dir_extend(blk, nblk, off, ino) == -1)
829 			return (-1);
830 	} else {
831 		blk = 0;
832 		if (dir_insert(nblk, 0, ino) == -1)
833 			return (-1);
834 	}
835 	if (sblock.fs_magic == FS_UFS1_MAGIC) {
836 		dp.dp1->di_blocks += (sblock.fs_bsize - size) / DEV_BSIZE;
837 		dp.dp1->di_db[lbn] = nblk;
838 		dp.dp1->di_size = lblktosize(&sblock, lbn+1);
839 	} else {
840 		dp.dp2->di_blocks += (sblock.fs_bsize - size) / DEV_BSIZE;
841 		dp.dp2->di_db[lbn] = nblk;
842 		dp.dp2->di_size = lblktosize(&sblock, lbn+1);
843 	}
844 	if (putinode(&disk) < 0) {
845 		warn("Failed to write root inode: %s", disk.d_error);
846 		return (-1);
847 	}
848 	if (cgwrite(&disk) < 0) {
849 		warn("Failed to write updated cg");
850 		sbdirty();
851 		return (-1);
852 	}
853 	if (blk) {
854 		if (cgbfree(&disk, blk, size) < 0) {
855 			warn("Failed to write cg");
856 			return (-1);
857 		}
858 	}
859 
860 	return (0);
861 }
862 
863 static int
864 indir_fill(ufs2_daddr_t blk, int level, int *resid)
865 {
866 	char indirbuf[MAXBSIZE];
867 	ufs1_daddr_t *bap1;
868 	ufs2_daddr_t *bap2;
869 	ufs2_daddr_t nblk;
870 	int ncnt;
871 	int cnt;
872 	int i;
873 
874 	bzero(indirbuf, sizeof(indirbuf));
875 	bap1 = (ufs1_daddr_t *)indirbuf;
876 	bap2 = (void *)bap1;
877 	cnt = 0;
878 	for (i = 0; i < NINDIR(&sblock) && *resid != 0; i++) {
879 		nblk = journal_balloc();
880 		if (nblk <= 0)
881 			return (-1);
882 		cnt++;
883 		if (sblock.fs_magic == FS_UFS1_MAGIC)
884 			*bap1++ = nblk;
885 		else
886 			*bap2++ = nblk;
887 		if (level != 0) {
888 			ncnt = indir_fill(nblk, level - 1, resid);
889 			if (ncnt <= 0)
890 				return (-1);
891 			cnt += ncnt;
892 		} else
893 			(*resid)--;
894 	}
895 	if (bwrite(&disk, fsbtodb(&sblock, blk), indirbuf,
896 	    sblock.fs_bsize) <= 0) {
897 		warn("Failed to write indirect");
898 		return (-1);
899 	}
900 	return (cnt);
901 }
902 
903 /*
904  * Clear the flag bits so the journal can be removed.
905  */
906 static void
907 journal_clear(void)
908 {
909 	union dinodep dp;
910 	ino_t ino;
911 
912 	ino = journal_findfile();
913 	if (ino == (ino_t)-1 || ino == 0) {
914 		warnx("Journal file does not exist");
915 		return;
916 	}
917 	printf("Clearing journal flags from inode %ju\n", (uintmax_t)ino);
918 	if (getinode(&disk, &dp, ino) != 0) {
919 		warn("Failed to get journal inode: %s", disk.d_error);
920 		return;
921 	}
922 	if (sblock.fs_magic == FS_UFS1_MAGIC)
923 		dp.dp1->di_flags = 0;
924 	else
925 		dp.dp2->di_flags = 0;
926 	if (putinode(&disk) < 0) {
927 		warn("Failed to write journal inode: %s", disk.d_error);
928 		return;
929 	}
930 }
931 
932 static int
933 journal_alloc(int64_t size)
934 {
935 	union dinodep dp;
936 	ufs2_daddr_t blk;
937 	struct cg *cgp;
938 	int resid;
939 	ino_t ino;
940 	int blks;
941 	time_t utime;
942 	int i;
943 
944 	cgp = &disk.d_cg;
945 	ino = 0;
946 
947 	/*
948 	 * If the journal file exists we can't allocate it.
949 	 */
950 	ino = journal_findfile();
951 	if (ino == (ino_t)-1) {
952 		warnx("journal_findfile() failed.");
953 		return (-1);
954 	}
955 	if (ino > 0) {
956 		warnx("Journal file %s already exists, please remove.",
957 		    SUJ_FILE);
958 		return (-1);
959 	}
960 	/*
961 	 * If the user didn't supply a size pick one based on the filesystem
962 	 * size constrained with hardcoded MIN and MAX values.  We opt for
963 	 * 1/1024th of the filesystem up to MAX but not exceeding one CG and
964 	 * not less than the MIN.
965 	 */
966 	if (size == 0) {
967 		size = (sblock.fs_size * sblock.fs_bsize) / 1024;
968 		size = MIN(SUJ_MAX, size);
969 		if (size / sblock.fs_fsize > sblock.fs_fpg)
970 			size = sblock.fs_fpg * sblock.fs_fsize;
971 		size = MAX(SUJ_MIN, size);
972 		/* fsck does not support fragments in journal files. */
973 		size = roundup(size, sblock.fs_bsize);
974 	}
975 	resid = blocks = size / sblock.fs_bsize;
976 	if (sblock.fs_cstotal.cs_nbfree < blocks) {
977 		warn("Insufficient free space for %jd byte journal", size);
978 		return (-1);
979 	}
980 	/*
981 	 * Find a cg with enough blocks to satisfy the journal
982 	 * size.  Presently the journal does not span cgs.
983 	 */
984 	while (cgread(&disk) == 1) {
985 		if (cgp->cg_cs.cs_nifree == 0)
986 			continue;
987 		ino = cgialloc(&disk);
988 		if (ino <= 0)
989 			break;
990 		printf("Using inode %ju in cg %d for %jd byte journal\n",
991 		    (uintmax_t)ino, cgp->cg_cgx, size);
992 		if (getinode(&disk, &dp, ino) != 0) {
993 			warn("Failed to get allocated inode: %s", disk.d_error);
994 			sbdirty();
995 			goto out;
996 		}
997 		/*
998 		 * We leave fields unrelated to the number of allocated
999 		 * blocks and size uninitialized.  This causes legacy
1000 		 * fsck implementations to clear the inode.
1001 		 */
1002 		time(&utime);
1003 		if (sblock.fs_magic == FS_UFS1_MAGIC) {
1004 			bzero(dp.dp1, sizeof(*dp.dp1));
1005 			dp.dp1->di_size = size;
1006 			dp.dp1->di_mode = IFREG | IREAD;
1007 			dp.dp1->di_nlink = 1;
1008 			dp.dp1->di_flags =
1009 			    SF_IMMUTABLE | SF_NOUNLINK | UF_NODUMP;
1010 			dp.dp1->di_atime = utime;
1011 			dp.dp1->di_mtime = utime;
1012 			dp.dp1->di_ctime = utime;
1013 		} else {
1014 			bzero(dp.dp2, sizeof(*dp.dp2));
1015 			dp.dp2->di_size = size;
1016 			dp.dp2->di_mode = IFREG | IREAD;
1017 			dp.dp2->di_nlink = 1;
1018 			dp.dp2->di_flags =
1019 			    SF_IMMUTABLE | SF_NOUNLINK | UF_NODUMP;
1020 			dp.dp2->di_atime = utime;
1021 			dp.dp2->di_mtime = utime;
1022 			dp.dp2->di_ctime = utime;
1023 			dp.dp2->di_birthtime = utime;
1024 		}
1025 		for (i = 0; i < UFS_NDADDR && resid; i++, resid--) {
1026 			blk = journal_balloc();
1027 			if (blk <= 0)
1028 				goto out;
1029 			if (sblock.fs_magic == FS_UFS1_MAGIC) {
1030 				dp.dp1->di_db[i] = blk;
1031 				dp.dp1->di_blocks++;
1032 			} else {
1033 				dp.dp2->di_db[i] = blk;
1034 				dp.dp2->di_blocks++;
1035 			}
1036 		}
1037 		for (i = 0; i < UFS_NIADDR && resid; i++) {
1038 			blk = journal_balloc();
1039 			if (blk <= 0)
1040 				goto out;
1041 			blks = indir_fill(blk, i, &resid) + 1;
1042 			if (blks <= 0) {
1043 				sbdirty();
1044 				goto out;
1045 			}
1046 			if (sblock.fs_magic == FS_UFS1_MAGIC) {
1047 				dp.dp1->di_ib[i] = blk;
1048 				dp.dp1->di_blocks += blks;
1049 			} else {
1050 				dp.dp2->di_ib[i] = blk;
1051 				dp.dp2->di_blocks += blks;
1052 			}
1053 		}
1054 		if (sblock.fs_magic == FS_UFS1_MAGIC)
1055 			dp.dp1->di_blocks *= sblock.fs_bsize / disk.d_bsize;
1056 		else
1057 			dp.dp2->di_blocks *= sblock.fs_bsize / disk.d_bsize;
1058 		if (putinode(&disk) < 0) {
1059 			warn("Failed to write allocated inode: %s",
1060 			    disk.d_error);
1061 			sbdirty();
1062 			return (-1);
1063 		}
1064 		if (cgwrite(&disk) < 0) {
1065 			warn("Failed to write updated cg");
1066 			sbdirty();
1067 			return (-1);
1068 		}
1069 		if (journal_insertfile(ino) < 0) {
1070 			sbdirty();
1071 			return (-1);
1072 		}
1073 		sblock.fs_sujfree = 0;
1074 		return (0);
1075 	}
1076 	warnx("Insufficient free space for the journal.");
1077 out:
1078 	return (-1);
1079 }
1080 
1081 static void
1082 usage(void)
1083 {
1084 	fprintf(stderr, "%s\n%s\n%s\n%s\n%s\n%s\n",
1085 "usage: tunefs [-A] [-a enable | disable] [-e maxbpg] [-f avgfilesize]",
1086 "              [-J enable | disable] [-j enable | disable] [-k metaspace]",
1087 "              [-L volname] [-l enable | disable] [-m minfree]",
1088 "              [-N enable | disable] [-n enable | disable]",
1089 "              [-o space | time] [-p] [-s avgfpdir] [-t enable | disable]",
1090 "              special | filesystem");
1091 	exit(2);
1092 }
1093 
1094 static void
1095 printfs(void)
1096 {
1097 	warnx("POSIX.1e ACLs: (-a)                                %s",
1098 		(sblock.fs_flags & FS_ACLS)? "enabled" : "disabled");
1099 	warnx("NFSv4 ACLs: (-N)                                   %s",
1100 		(sblock.fs_flags & FS_NFS4ACLS)? "enabled" : "disabled");
1101 	warnx("MAC multilabel: (-l)                               %s",
1102 		(sblock.fs_flags & FS_MULTILABEL)? "enabled" : "disabled");
1103 	warnx("soft updates: (-n)                                 %s",
1104 		(sblock.fs_flags & FS_DOSOFTDEP)? "enabled" : "disabled");
1105 	warnx("soft update journaling: (-j)                       %s",
1106 		(sblock.fs_flags & FS_SUJ)? "enabled" : "disabled");
1107 	warnx("gjournal: (-J)                                     %s",
1108 		(sblock.fs_flags & FS_GJOURNAL)? "enabled" : "disabled");
1109 	warnx("trim: (-t)                                         %s",
1110 		(sblock.fs_flags & FS_TRIM)? "enabled" : "disabled");
1111 	warnx("maximum blocks per file in a cylinder group: (-e)  %d",
1112 	      sblock.fs_maxbpg);
1113 	warnx("average file size: (-f)                            %d",
1114 	      sblock.fs_avgfilesize);
1115 	warnx("average number of files in a directory: (-s)       %d",
1116 	      sblock.fs_avgfpdir);
1117 	warnx("minimum percentage of free space: (-m)             %d%%",
1118 	      sblock.fs_minfree);
1119 	warnx("space to hold for metadata blocks: (-k)            %jd",
1120 	      sblock.fs_metaspace);
1121 	warnx("optimization preference: (-o)                      %s",
1122 	      sblock.fs_optim == FS_OPTSPACE ? "space" : "time");
1123 	if (sblock.fs_minfree >= MINFREE &&
1124 	    sblock.fs_optim == FS_OPTSPACE)
1125 		warnx(OPTWARN, "time", ">=", MINFREE);
1126 	if (sblock.fs_minfree < MINFREE &&
1127 	    sblock.fs_optim == FS_OPTTIME)
1128 		warnx(OPTWARN, "space", "<", MINFREE);
1129 	warnx("volume label: (-L)                                 %s",
1130 		sblock.fs_volname);
1131 }
1132