xref: /freebsd/sys/ufs/ffs/ffs_subr.c (revision faf25f48d601ae39f5752602f3020e2e92605625)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 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  *	@(#)ffs_subr.c	8.5 (Berkeley) 3/21/95
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include <sys/param.h>
38 #include <sys/limits.h>
39 
40 #ifndef _KERNEL
41 #include <stdio.h>
42 #include <string.h>
43 #include <stdlib.h>
44 #include <time.h>
45 #include <sys/errno.h>
46 #include <ufs/ufs/dinode.h>
47 #include <ufs/ffs/fs.h>
48 
49 uint32_t calculate_crc32c(uint32_t, const void *, size_t);
50 uint32_t ffs_calc_sbhash(struct fs *);
51 struct malloc_type;
52 #define UFS_MALLOC(size, type, flags) malloc(size)
53 #define UFS_FREE(ptr, type) free(ptr)
54 #define maxphys MAXPHYS
55 
56 #else /* _KERNEL */
57 #include <sys/systm.h>
58 #include <sys/gsb_crc32.h>
59 #include <sys/lock.h>
60 #include <sys/malloc.h>
61 #include <sys/mount.h>
62 #include <sys/vnode.h>
63 #include <sys/bio.h>
64 #include <sys/buf.h>
65 #include <sys/ucred.h>
66 
67 #include <ufs/ufs/quota.h>
68 #include <ufs/ufs/inode.h>
69 #include <ufs/ufs/extattr.h>
70 #include <ufs/ufs/ufsmount.h>
71 #include <ufs/ufs/ufs_extern.h>
72 #include <ufs/ffs/ffs_extern.h>
73 #include <ufs/ffs/fs.h>
74 
75 #define UFS_MALLOC(size, type, flags) malloc(size, type, flags)
76 #define UFS_FREE(ptr, type) free(ptr, type)
77 
78 #endif /* _KERNEL */
79 
80 /*
81  * Verify an inode check-hash.
82  */
83 int
84 ffs_verify_dinode_ckhash(struct fs *fs, struct ufs2_dinode *dip)
85 {
86 	uint32_t ckhash, save_ckhash;
87 
88 	/*
89 	 * Return success if unallocated or we are not doing inode check-hash.
90 	 */
91 	if (dip->di_mode == 0 || (fs->fs_metackhash & CK_INODE) == 0)
92 		return (0);
93 	/*
94 	 * Exclude di_ckhash from the crc32 calculation, e.g., always use
95 	 * a check-hash value of zero when calculating the check-hash.
96 	 */
97 	save_ckhash = dip->di_ckhash;
98 	dip->di_ckhash = 0;
99 	ckhash = calculate_crc32c(~0L, (void *)dip, sizeof(*dip));
100 	dip->di_ckhash = save_ckhash;
101 	if (save_ckhash == ckhash)
102 		return (0);
103 	return (EINVAL);
104 }
105 
106 /*
107  * Update an inode check-hash.
108  */
109 void
110 ffs_update_dinode_ckhash(struct fs *fs, struct ufs2_dinode *dip)
111 {
112 
113 	if (dip->di_mode == 0 || (fs->fs_metackhash & CK_INODE) == 0)
114 		return;
115 	/*
116 	 * Exclude old di_ckhash from the crc32 calculation, e.g., always use
117 	 * a check-hash value of zero when calculating the new check-hash.
118 	 */
119 	dip->di_ckhash = 0;
120 	dip->di_ckhash = calculate_crc32c(~0L, (void *)dip, sizeof(*dip));
121 }
122 
123 /*
124  * These are the low-level functions that actually read and write
125  * the superblock and its associated data.
126  */
127 static off_t sblock_try[] = SBLOCKSEARCH;
128 static int readsuper(void *, struct fs **, off_t, int, int,
129 	int (*)(void *, off_t, void **, int));
130 static int validate_sblock(struct fs *, int);
131 
132 /*
133  * Read a superblock from the devfd device.
134  *
135  * If an alternate superblock is specified, it is read. Otherwise the
136  * set of locations given in the SBLOCKSEARCH list is searched for a
137  * superblock. Memory is allocated for the superblock by the readfunc and
138  * is returned. If filltype is non-NULL, additional memory is allocated
139  * of type filltype and filled in with the superblock summary information.
140  * All memory is freed when any error is returned.
141  *
142  * If a superblock is found, zero is returned. Otherwise one of the
143  * following error values is returned:
144  *     EIO: non-existent or truncated superblock.
145  *     EIO: error reading summary information.
146  *     ENOENT: no usable known superblock found.
147  *     ENOMEM: failed to allocate space for the superblock.
148  *     EINVAL: The previous newfs operation on this volume did not complete.
149  *         The administrator must complete newfs before using this volume.
150  */
151 int
152 ffs_sbget(void *devfd, struct fs **fsp, off_t altsblock,
153     struct malloc_type *filltype,
154     int (*readfunc)(void *devfd, off_t loc, void **bufp, int size))
155 {
156 	struct fs *fs;
157 	struct fs_summary_info *fs_si;
158 	int i, error;
159 	uint64_t size, blks;
160 	uint8_t *space;
161 	int32_t *lp;
162 	char *buf;
163 
164 	fs = NULL;
165 	*fsp = NULL;
166 	if (altsblock >= 0) {
167 		if ((error = readsuper(devfd, &fs, altsblock, 1, 0,
168 		     readfunc)) != 0) {
169 			if (fs != NULL)
170 				UFS_FREE(fs, filltype);
171 			return (error);
172 		}
173 	} else {
174 		for (i = 0; sblock_try[i] != -1; i++) {
175 			if ((error = readsuper(devfd, &fs, sblock_try[i], 0,
176 			     altsblock, readfunc)) == 0)
177 				break;
178 			if (fs != NULL) {
179 				UFS_FREE(fs, filltype);
180 				fs = NULL;
181 			}
182 			if (error == ENOENT)
183 				continue;
184 			return (error);
185 		}
186 		if (sblock_try[i] == -1)
187 			return (ENOENT);
188 	}
189 	/*
190 	 * Read in the superblock summary information.
191 	 */
192 	size = fs->fs_cssize;
193 	blks = howmany(size, fs->fs_fsize);
194 	if (fs->fs_contigsumsize > 0)
195 		size += fs->fs_ncg * sizeof(int32_t);
196 	size += fs->fs_ncg * sizeof(u_int8_t);
197 	if ((fs_si = UFS_MALLOC(sizeof(*fs_si), filltype, M_NOWAIT)) == NULL) {
198 		UFS_FREE(fs, filltype);
199 		return (ENOMEM);
200 	}
201 	bzero(fs_si, sizeof(*fs_si));
202 	fs->fs_si = fs_si;
203 	if ((space = UFS_MALLOC(size, filltype, M_NOWAIT)) == NULL) {
204 		UFS_FREE(fs->fs_si, filltype);
205 		UFS_FREE(fs, filltype);
206 		return (ENOMEM);
207 	}
208 	fs->fs_csp = (struct csum *)space;
209 	for (i = 0; i < blks; i += fs->fs_frag) {
210 		size = fs->fs_bsize;
211 		if (i + fs->fs_frag > blks)
212 			size = (blks - i) * fs->fs_fsize;
213 		buf = NULL;
214 		error = (*readfunc)(devfd,
215 		    dbtob(fsbtodb(fs, fs->fs_csaddr + i)), (void **)&buf, size);
216 		if (error) {
217 			if (buf != NULL)
218 				UFS_FREE(buf, filltype);
219 			UFS_FREE(fs->fs_csp, filltype);
220 			UFS_FREE(fs->fs_si, filltype);
221 			UFS_FREE(fs, filltype);
222 			return (error);
223 		}
224 		memcpy(space, buf, size);
225 		UFS_FREE(buf, filltype);
226 		space += size;
227 	}
228 	if (fs->fs_contigsumsize > 0) {
229 		fs->fs_maxcluster = lp = (int32_t *)space;
230 		for (i = 0; i < fs->fs_ncg; i++)
231 			*lp++ = fs->fs_contigsumsize;
232 		space = (uint8_t *)lp;
233 	}
234 	size = fs->fs_ncg * sizeof(u_int8_t);
235 	fs->fs_contigdirs = (u_int8_t *)space;
236 	bzero(fs->fs_contigdirs, size);
237 	*fsp = fs;
238 	return (0);
239 }
240 
241 /*
242  * Try to read a superblock from the location specified by sblockloc.
243  * Return zero on success or an errno on failure.
244  */
245 static int
246 readsuper(void *devfd, struct fs **fsp, off_t sblockloc, int isaltsblk,
247     int chkhash, int (*readfunc)(void *devfd, off_t loc, void **bufp, int size))
248 {
249 	struct fs *fs;
250 	int error, res;
251 	uint32_t ckhash;
252 
253 	error = (*readfunc)(devfd, sblockloc, (void **)fsp, SBLOCKSIZE);
254 	if (error != 0)
255 		return (error);
256 	fs = *fsp;
257 	if (fs->fs_magic == FS_BAD_MAGIC)
258 		return (EINVAL);
259 	if ((error = validate_sblock(fs, isaltsblk)) != 0)
260 		return (error);
261 	/*
262 	 * If the filesystem has been run on a kernel without
263 	 * metadata check hashes, disable them.
264 	 */
265 	if ((fs->fs_flags & FS_METACKHASH) == 0)
266 		fs->fs_metackhash = 0;
267 	/*
268 	 * Clear any check-hashes that are not maintained
269 	 * by this kernel. Also clear any unsupported flags.
270 	 */
271 	fs->fs_metackhash &= CK_SUPPORTED;
272 	fs->fs_flags &= FS_SUPPORTED;
273 	if (fs->fs_ckhash != (ckhash = ffs_calc_sbhash(fs))) {
274 		if (chkhash == STDSB_NOMSG)
275 			return (EINTEGRITY);
276 		if (chkhash == STDSB_NOHASHFAIL_NOMSG)
277 			return (0);
278 #ifdef _KERNEL
279 		res = uprintf("Superblock check-hash failed: recorded "
280 		    "check-hash 0x%x != computed check-hash 0x%x%s\n",
281 		    fs->fs_ckhash, ckhash,
282 		    chkhash == STDSB_NOHASHFAIL ? " (Ignored)" : "");
283 #else
284 		res = 0;
285 #endif
286 		/*
287 		 * Print check-hash failure if no controlling terminal
288 		 * in kernel or always if in user-mode (libufs).
289 		 */
290 		if (res == 0)
291 			printf("Superblock check-hash failed: recorded "
292 			    "check-hash 0x%x != computed check-hash "
293 			    "0x%x%s\n", fs->fs_ckhash, ckhash,
294 			    chkhash == STDSB_NOHASHFAIL ?
295 			    " (Ignored)" : "");
296 		if (chkhash == STDSB)
297 			return (EINTEGRITY);
298 		/* chkhash == STDSB_NOHASHFAIL */
299 		return (0);
300 	}
301 	/* Have to set for old filesystems that predate this field */
302 	fs->fs_sblockactualloc = sblockloc;
303 	/* Not yet any summary information */
304 	fs->fs_si = NULL;
305 	return (0);
306 }
307 
308 /*
309  * Verify the filesystem values.
310  */
311 #define ILOG2(num) (fls(num) - 1)
312 #undef CHK
313 #define CHK(lhs, op, rhs, fmt)						\
314 	if (lhs op rhs) {						\
315 		printf("UFS%d superblock failed: %s (" #fmt ") %s %s ("	\
316 		    #fmt ")\n", fs->fs_magic == FS_UFS1_MAGIC ? 1 : 2,	\
317 		    #lhs, (intmax_t)lhs, #op, #rhs, (intmax_t)rhs);	\
318 		return (ENOENT);					\
319 	}
320 #define CHK2(lhs1, op1, rhs1, lhs2, op2, rhs2, fmt)			\
321 	if (lhs1 op1 rhs1 && lhs2 op2 rhs2) {				\
322 		printf("UFS%d superblock failed: %s (" #fmt ") %s %s ("	\
323 		    #fmt ") && %s (" #fmt ") %s %s (" #fmt ")\n",	\
324 		    fs->fs_magic == FS_UFS1_MAGIC ? 1 : 2, #lhs1, 	\
325 		    (intmax_t)lhs1, #op1, #rhs1, (intmax_t)rhs1, #lhs2,	\
326 		    (intmax_t)lhs2, #op2, #rhs2, (intmax_t)rhs2);	\
327 		return (ENOENT);					\
328 	}
329 
330 static int
331 validate_sblock(struct fs *fs, int isaltsblk)
332 {
333 	u_long i, sectorsize, cgnum;
334 	u_int64_t maxfilesize, sizepb;
335 
336 	sectorsize = dbtob(1);
337 	if (fs->fs_magic == FS_UFS2_MAGIC) {
338 		if (!isaltsblk) {
339 			CHK(fs->fs_sblockloc, !=, SBLOCK_UFS2, %#jx);
340 			CHK2(fs->fs_sblockactualloc, !=, SBLOCK_UFS2,
341 			    fs->fs_sblockactualloc, !=, 0, %jd);
342 		}
343 		CHK(fs->fs_maxsymlinklen, !=, ((UFS_NDADDR + UFS_NIADDR) *
344 			sizeof(ufs2_daddr_t)), %jd);
345 		CHK(fs->fs_nindir, !=, fs->fs_bsize / sizeof(ufs2_daddr_t),
346 		    %jd);
347 		CHK(fs->fs_inopb, !=, fs->fs_bsize / sizeof(struct ufs2_dinode),
348 		    %jd);
349 	} else if (fs->fs_magic == FS_UFS1_MAGIC) {
350 		if (!isaltsblk) {
351 			CHK(fs->fs_sblockloc, >, SBLOCK_UFS1, %jd);
352 			/*
353 			 * For UFS1 the with a 65536 block size, the first
354 			 * backup superblock is at the same location as the
355 			 * UFS2 superblock. Since SBLOCK_UFS2 is the first
356 			 * location checked, the first backup is the
357 			 * superblock that will be accessed.
358 			 */
359 			if (fs->fs_bsize == SBLOCK_UFS2) {
360 				CHK(fs->fs_sblockactualloc, >, SBLOCK_UFS2,
361 				    %jd);
362 			} else {
363 				CHK2(fs->fs_sblockactualloc, !=, SBLOCK_UFS1,
364 				    fs->fs_sblockactualloc, !=, 0, %jd);
365 			}
366 		}
367 		CHK(fs->fs_nindir, !=, fs->fs_bsize / sizeof(ufs1_daddr_t),
368 		    %jd);
369 		CHK(fs->fs_inopb, !=, fs->fs_bsize / sizeof(struct ufs1_dinode),
370 		    %jd);
371 		CHK(fs->fs_maxsymlinklen, !=, ((UFS_NDADDR + UFS_NIADDR) *
372 			sizeof(ufs1_daddr_t)), %jd);
373 		CHK(fs->fs_old_inodefmt, !=, FS_44INODEFMT, %jd);
374 		CHK(fs->fs_old_cgoffset, !=, 0, %jd);
375 		CHK(fs->fs_old_cgmask, !=, 0xffffffff, %#jx);
376 		CHK(fs->fs_old_rotdelay, !=, 0, %jd);
377 		CHK(fs->fs_old_rps, !=, 60, %jd);
378 		CHK(fs->fs_old_nspf, !=, fs->fs_fsize / sectorsize, %jd);
379 		CHK(fs->fs_old_cpg, !=, 1, %jd);
380 		CHK(fs->fs_old_interleave, !=, 1, %jd);
381 		CHK(fs->fs_old_trackskew, !=, 0, %jd);
382 		CHK(fs->fs_old_cpc, !=, 0, %jd);
383 		CHK(fs->fs_old_postblformat, !=, 1, %jd);
384 		CHK(fs->fs_old_nrpos, !=, 1, %jd);
385 		CHK(fs->fs_old_spc, !=, fs->fs_fpg * fs->fs_old_nspf, %jd);
386 		CHK(fs->fs_old_nsect, !=, fs->fs_old_spc, %jd);
387 		CHK(fs->fs_old_npsect, !=, fs->fs_old_spc, %jd);
388 		CHK(fs->fs_old_ncyl, !=, fs->fs_ncg, %jd);
389 	} else {
390 		/* Bad magic number, so assume not a superblock */
391 		return (ENOENT);
392 	}
393 	CHK(fs->fs_bsize, <, MINBSIZE, %jd);
394 	CHK(fs->fs_bsize, >, MAXBSIZE, %jd);
395 	CHK(fs->fs_bsize, <, roundup(sizeof(struct fs), DEV_BSIZE), %jd);
396 	CHK(fs->fs_sbsize, >, SBLOCKSIZE, %jd);
397 	CHK(powerof2(fs->fs_bsize), ==, 0, %jd);
398 	CHK(fs->fs_fsize, <, sectorsize, %jd);
399 	CHK(fs->fs_fsize, >, fs->fs_bsize, %jd);
400 	CHK(fs->fs_fsize * MAXFRAG, <, fs->fs_bsize, %jd);
401 	CHK(powerof2(fs->fs_fsize), ==, 0, %jd);
402 	CHK(fs->fs_maxbsize, <, fs->fs_bsize, %jd);
403 	CHK(powerof2(fs->fs_maxbsize), ==, 0, %jd);
404 	CHK(fs->fs_maxbsize, >, FS_MAXCONTIG * fs->fs_bsize, %jd);
405 	CHK(fs->fs_bmask, !=, ~(fs->fs_bsize - 1), %#jx);
406 	CHK(fs->fs_fmask, !=, ~(fs->fs_fsize - 1), %#jx);
407 	CHK(fs->fs_qbmask, !=, ~fs->fs_bmask, %#jx);
408 	CHK(fs->fs_qfmask, !=, ~fs->fs_fmask, %#jx);
409 	CHK(fs->fs_bshift, !=, ILOG2(fs->fs_bsize), %jd);
410 	CHK(fs->fs_fshift, !=, ILOG2(fs->fs_fsize), %jd);
411 	CHK(fs->fs_frag, !=, numfrags(fs, fs->fs_bsize), %jd);
412 	CHK(fs->fs_fragshift, !=, ILOG2(fs->fs_frag), %jd);
413 	CHK(fs->fs_frag, >, MAXFRAG, %jd);
414 	CHK(fs->fs_fsbtodb, !=, ILOG2(fs->fs_fsize / sectorsize), %jd);
415 	CHK(fs->fs_sblkno, !=, roundup(
416 	    howmany(fs->fs_sblockloc + SBLOCKSIZE, fs->fs_fsize),
417 	    fs->fs_frag), %jd);
418 	CHK(fs->fs_cblkno, !=, fs->fs_sblkno +
419 	    roundup(howmany(SBLOCKSIZE, fs->fs_fsize), fs->fs_frag), %jd);
420 	CHK(fs->fs_iblkno, !=, fs->fs_cblkno + fs->fs_frag, %jd);
421 	CHK(fs->fs_dblkno, !=, fs->fs_iblkno + fs->fs_ipg / INOPF(fs), %jd);
422 	CHK(fs->fs_cgsize, >, fs->fs_bsize, %jd);
423 	CHK(fs->fs_cssize, !=,
424 		fragroundup(fs, fs->fs_ncg * sizeof(struct csum)), %jd);
425 	/*
426 	 * This test is valid, however older versions of growfs failed
427 	 * to correctly update fs_dsize so will fail this test. Thus we
428 	 * exclude it from the requirements.
429 	 */
430 #ifdef notdef
431 	CHK(fs->fs_dsize, !=, fs->fs_size - fs->fs_sblkno -
432 		fs->fs_ncg * (fs->fs_dblkno - fs->fs_sblkno) -
433 		howmany(fs->fs_cssize, fs->fs_fsize), %jd);
434 #endif
435 	CHK(fs->fs_metaspace, <, 0, %jd);
436 	CHK(fs->fs_metaspace, >, fs->fs_fpg / 2, %jd);
437 	CHK(fs->fs_minfree, >, 99, %jd%%);
438 	maxfilesize = fs->fs_bsize * UFS_NDADDR - 1;
439 	for (sizepb = fs->fs_bsize, i = 0; i < UFS_NIADDR; i++) {
440 		sizepb *= NINDIR(fs);
441 		maxfilesize += sizepb;
442 	}
443 	CHK(fs->fs_maxfilesize, !=, maxfilesize, %jd);
444 	/*
445 	 * These values have a tight interaction with each other that
446 	 * makes it hard to tightly bound them. So we can only check
447 	 * that they are within a broader possible range.
448 	 *
449 	 * The size cannot always be accurately determined, but ensure
450 	 * that it is consistent with the number of cylinder groups (fs_ncg)
451 	 * and the number of fragments per cylinder group (fs_fpg). Ensure
452 	 * that the summary information size is correct and that it starts
453 	 * and ends in the data area of the same cylinder group.
454 	 */
455 	CHK(fs->fs_ncg, <, 1, %jd);
456 	CHK(fs->fs_size, <, 8 * fs->fs_frag, %jd);
457 	CHK(fs->fs_size, <=, (fs->fs_ncg - 1) * fs->fs_fpg, %jd);
458 	CHK(fs->fs_size, >, fs->fs_ncg * fs->fs_fpg, %jd);
459 	CHK(fs->fs_cssize, !=,
460 	    fragroundup(fs, fs->fs_ncg * sizeof(struct csum)), %jd);
461 	cgnum = dtog(fs, fs->fs_csaddr);
462 	CHK(fs->fs_csaddr, <, cgdmin(fs, cgnum), %jd);
463 	CHK(dtog(fs, fs->fs_csaddr + howmany(fs->fs_cssize, fs->fs_fsize)), >,
464 	    cgnum, %jd);
465 	CHK(fs->fs_ipg * fs->fs_ncg, >, (((int64_t)(1)) << 32) - INOPB(fs),
466 	    %jd);
467 	/*
468 	 * With file system clustering it is possible to allocate
469 	 * many contiguous blocks. The kernel variable maxphys defines
470 	 * the maximum transfer size permitted by the controller and/or
471 	 * buffering. The fs_maxcontig parameter controls the maximum
472 	 * number of blocks that the filesystem will read or write
473 	 * in a single transfer. It is calculated when the filesystem
474 	 * is created as maxphys / fs_bsize. The loader uses a maxphys
475 	 * of 128K even when running on a system that supports larger
476 	 * values. If the filesystem was built on a system that supports
477 	 * a larger maxphys (1M is typical) it will have configured
478 	 * fs_maxcontig for that larger system. So we bound the upper
479 	 * allowable limit for fs_maxconfig to be able to at least
480 	 * work with a 1M maxphys on the smallest block size filesystem:
481 	 * 1M / 4096 == 256. There is no harm in allowing the mounting of
482 	 * filesystems that make larger than maxphys I/O requests because
483 	 * those (mostly 32-bit machines) can (very slowly) handle I/O
484 	 * requests that exceed maxphys.
485 	 */
486 	CHK(fs->fs_maxcontig, <, 0, %jd);
487 	CHK(fs->fs_maxcontig, >, MAX(256, maxphys / fs->fs_bsize), %jd);
488 	CHK2(fs->fs_maxcontig, ==, 0, fs->fs_contigsumsize, !=, 0, %jd);
489 	CHK2(fs->fs_maxcontig, >, 1, fs->fs_contigsumsize, !=,
490 	    MIN(fs->fs_maxcontig, FS_MAXCONTIG), %jd);
491 	return (0);
492 }
493 
494 /*
495  * Write a superblock to the devfd device from the memory pointed to by fs.
496  * Write out the superblock summary information if it is present.
497  *
498  * If the write is successful, zero is returned. Otherwise one of the
499  * following error values is returned:
500  *     EIO: failed to write superblock.
501  *     EIO: failed to write superblock summary information.
502  */
503 int
504 ffs_sbput(void *devfd, struct fs *fs, off_t loc,
505     int (*writefunc)(void *devfd, off_t loc, void *buf, int size))
506 {
507 	int i, error, blks, size;
508 	uint8_t *space;
509 
510 	/*
511 	 * If there is summary information, write it first, so if there
512 	 * is an error, the superblock will not be marked as clean.
513 	 */
514 	if (fs->fs_si != NULL && fs->fs_csp != NULL) {
515 		blks = howmany(fs->fs_cssize, fs->fs_fsize);
516 		space = (uint8_t *)fs->fs_csp;
517 		for (i = 0; i < blks; i += fs->fs_frag) {
518 			size = fs->fs_bsize;
519 			if (i + fs->fs_frag > blks)
520 				size = (blks - i) * fs->fs_fsize;
521 			if ((error = (*writefunc)(devfd,
522 			     dbtob(fsbtodb(fs, fs->fs_csaddr + i)),
523 			     space, size)) != 0)
524 				return (error);
525 			space += size;
526 		}
527 	}
528 	fs->fs_fmod = 0;
529 #ifndef _KERNEL
530 	{
531 		struct fs_summary_info *fs_si;
532 
533 		fs->fs_time = time(NULL);
534 		/* Clear the pointers for the duration of writing. */
535 		fs_si = fs->fs_si;
536 		fs->fs_si = NULL;
537 		fs->fs_ckhash = ffs_calc_sbhash(fs);
538 		error = (*writefunc)(devfd, loc, fs, fs->fs_sbsize);
539 		fs->fs_si = fs_si;
540 	}
541 #else /* _KERNEL */
542 	fs->fs_time = time_second;
543 	fs->fs_ckhash = ffs_calc_sbhash(fs);
544 	error = (*writefunc)(devfd, loc, fs, fs->fs_sbsize);
545 #endif /* _KERNEL */
546 	return (error);
547 }
548 
549 /*
550  * Calculate the check-hash for a superblock.
551  */
552 uint32_t
553 ffs_calc_sbhash(struct fs *fs)
554 {
555 	uint32_t ckhash, save_ckhash;
556 
557 	/*
558 	 * A filesystem that was using a superblock ckhash may be moved
559 	 * to an older kernel that does not support ckhashes. The
560 	 * older kernel will clear the FS_METACKHASH flag indicating
561 	 * that it does not update hashes. When the disk is moved back
562 	 * to a kernel capable of ckhashes it disables them on mount:
563 	 *
564 	 *	if ((fs->fs_flags & FS_METACKHASH) == 0)
565 	 *		fs->fs_metackhash = 0;
566 	 *
567 	 * This leaves (fs->fs_metackhash & CK_SUPERBLOCK) == 0) with an
568 	 * old stale value in the fs->fs_ckhash field. Thus the need to
569 	 * just accept what is there.
570 	 */
571 	if ((fs->fs_metackhash & CK_SUPERBLOCK) == 0)
572 		return (fs->fs_ckhash);
573 
574 	save_ckhash = fs->fs_ckhash;
575 	fs->fs_ckhash = 0;
576 	/*
577 	 * If newly read from disk, the caller is responsible for
578 	 * verifying that fs->fs_sbsize <= SBLOCKSIZE.
579 	 */
580 	ckhash = calculate_crc32c(~0L, (void *)fs, fs->fs_sbsize);
581 	fs->fs_ckhash = save_ckhash;
582 	return (ckhash);
583 }
584 
585 /*
586  * Update the frsum fields to reflect addition or deletion
587  * of some frags.
588  */
589 void
590 ffs_fragacct(struct fs *fs, int fragmap, int32_t fraglist[], int cnt)
591 {
592 	int inblk;
593 	int field, subfield;
594 	int siz, pos;
595 
596 	inblk = (int)(fragtbl[fs->fs_frag][fragmap]) << 1;
597 	fragmap <<= 1;
598 	for (siz = 1; siz < fs->fs_frag; siz++) {
599 		if ((inblk & (1 << (siz + (fs->fs_frag % NBBY)))) == 0)
600 			continue;
601 		field = around[siz];
602 		subfield = inside[siz];
603 		for (pos = siz; pos <= fs->fs_frag; pos++) {
604 			if ((fragmap & field) == subfield) {
605 				fraglist[siz] += cnt;
606 				pos += siz;
607 				field <<= siz;
608 				subfield <<= siz;
609 			}
610 			field <<= 1;
611 			subfield <<= 1;
612 		}
613 	}
614 }
615 
616 /*
617  * block operations
618  *
619  * check if a block is available
620  */
621 int
622 ffs_isblock(struct fs *fs, unsigned char *cp, ufs1_daddr_t h)
623 {
624 	unsigned char mask;
625 
626 	switch ((int)fs->fs_frag) {
627 	case 8:
628 		return (cp[h] == 0xff);
629 	case 4:
630 		mask = 0x0f << ((h & 0x1) << 2);
631 		return ((cp[h >> 1] & mask) == mask);
632 	case 2:
633 		mask = 0x03 << ((h & 0x3) << 1);
634 		return ((cp[h >> 2] & mask) == mask);
635 	case 1:
636 		mask = 0x01 << (h & 0x7);
637 		return ((cp[h >> 3] & mask) == mask);
638 	default:
639 #ifdef _KERNEL
640 		panic("ffs_isblock");
641 #endif
642 		break;
643 	}
644 	return (0);
645 }
646 
647 /*
648  * check if a block is free
649  */
650 int
651 ffs_isfreeblock(struct fs *fs, u_char *cp, ufs1_daddr_t h)
652 {
653 
654 	switch ((int)fs->fs_frag) {
655 	case 8:
656 		return (cp[h] == 0);
657 	case 4:
658 		return ((cp[h >> 1] & (0x0f << ((h & 0x1) << 2))) == 0);
659 	case 2:
660 		return ((cp[h >> 2] & (0x03 << ((h & 0x3) << 1))) == 0);
661 	case 1:
662 		return ((cp[h >> 3] & (0x01 << (h & 0x7))) == 0);
663 	default:
664 #ifdef _KERNEL
665 		panic("ffs_isfreeblock");
666 #endif
667 		break;
668 	}
669 	return (0);
670 }
671 
672 /*
673  * take a block out of the map
674  */
675 void
676 ffs_clrblock(struct fs *fs, u_char *cp, ufs1_daddr_t h)
677 {
678 
679 	switch ((int)fs->fs_frag) {
680 	case 8:
681 		cp[h] = 0;
682 		return;
683 	case 4:
684 		cp[h >> 1] &= ~(0x0f << ((h & 0x1) << 2));
685 		return;
686 	case 2:
687 		cp[h >> 2] &= ~(0x03 << ((h & 0x3) << 1));
688 		return;
689 	case 1:
690 		cp[h >> 3] &= ~(0x01 << (h & 0x7));
691 		return;
692 	default:
693 #ifdef _KERNEL
694 		panic("ffs_clrblock");
695 #endif
696 		break;
697 	}
698 }
699 
700 /*
701  * put a block into the map
702  */
703 void
704 ffs_setblock(struct fs *fs, unsigned char *cp, ufs1_daddr_t h)
705 {
706 
707 	switch ((int)fs->fs_frag) {
708 	case 8:
709 		cp[h] = 0xff;
710 		return;
711 	case 4:
712 		cp[h >> 1] |= (0x0f << ((h & 0x1) << 2));
713 		return;
714 	case 2:
715 		cp[h >> 2] |= (0x03 << ((h & 0x3) << 1));
716 		return;
717 	case 1:
718 		cp[h >> 3] |= (0x01 << (h & 0x7));
719 		return;
720 	default:
721 #ifdef _KERNEL
722 		panic("ffs_setblock");
723 #endif
724 		break;
725 	}
726 }
727 
728 /*
729  * Update the cluster map because of an allocation or free.
730  *
731  * Cnt == 1 means free; cnt == -1 means allocating.
732  */
733 void
734 ffs_clusteracct(struct fs *fs, struct cg *cgp, ufs1_daddr_t blkno, int cnt)
735 {
736 	int32_t *sump;
737 	int32_t *lp;
738 	u_char *freemapp, *mapp;
739 	int i, start, end, forw, back, map;
740 	u_int bit;
741 
742 	if (fs->fs_contigsumsize <= 0)
743 		return;
744 	freemapp = cg_clustersfree(cgp);
745 	sump = cg_clustersum(cgp);
746 	/*
747 	 * Allocate or clear the actual block.
748 	 */
749 	if (cnt > 0)
750 		setbit(freemapp, blkno);
751 	else
752 		clrbit(freemapp, blkno);
753 	/*
754 	 * Find the size of the cluster going forward.
755 	 */
756 	start = blkno + 1;
757 	end = start + fs->fs_contigsumsize;
758 	if (end >= cgp->cg_nclusterblks)
759 		end = cgp->cg_nclusterblks;
760 	mapp = &freemapp[start / NBBY];
761 	map = *mapp++;
762 	bit = 1U << (start % NBBY);
763 	for (i = start; i < end; i++) {
764 		if ((map & bit) == 0)
765 			break;
766 		if ((i & (NBBY - 1)) != (NBBY - 1)) {
767 			bit <<= 1;
768 		} else {
769 			map = *mapp++;
770 			bit = 1;
771 		}
772 	}
773 	forw = i - start;
774 	/*
775 	 * Find the size of the cluster going backward.
776 	 */
777 	start = blkno - 1;
778 	end = start - fs->fs_contigsumsize;
779 	if (end < 0)
780 		end = -1;
781 	mapp = &freemapp[start / NBBY];
782 	map = *mapp--;
783 	bit = 1U << (start % NBBY);
784 	for (i = start; i > end; i--) {
785 		if ((map & bit) == 0)
786 			break;
787 		if ((i & (NBBY - 1)) != 0) {
788 			bit >>= 1;
789 		} else {
790 			map = *mapp--;
791 			bit = 1U << (NBBY - 1);
792 		}
793 	}
794 	back = start - i;
795 	/*
796 	 * Account for old cluster and the possibly new forward and
797 	 * back clusters.
798 	 */
799 	i = back + forw + 1;
800 	if (i > fs->fs_contigsumsize)
801 		i = fs->fs_contigsumsize;
802 	sump[i] += cnt;
803 	if (back > 0)
804 		sump[back] -= cnt;
805 	if (forw > 0)
806 		sump[forw] -= cnt;
807 	/*
808 	 * Update cluster summary information.
809 	 */
810 	lp = &sump[fs->fs_contigsumsize];
811 	for (i = fs->fs_contigsumsize; i > 0; i--)
812 		if (*lp-- > 0)
813 			break;
814 	fs->fs_maxcluster[cgp->cg_cgx] = i;
815 }
816