xref: /original-bsd/sys/ufs/ffs/ffs_alloc.c (revision 5998a314)
1 /*	ffs_alloc.c	2.18	82/10/21	*/
2 
3 #include "../h/param.h"
4 #include "../h/systm.h"
5 #include "../h/mount.h"
6 #include "../h/fs.h"
7 #include "../h/conf.h"
8 #include "../h/buf.h"
9 #include "../h/inode.h"
10 #include "../h/dir.h"
11 #include "../h/user.h"
12 #include "../h/quota.h"
13 #include "../h/kernel.h"
14 
15 extern u_long		hashalloc();
16 extern u_long		ialloccg();
17 extern u_long		alloccg();
18 extern daddr_t		alloccgblk();
19 extern daddr_t		fragextend();
20 extern daddr_t		blkpref();
21 extern daddr_t		mapsearch();
22 extern int		inside[], around[];
23 extern unsigned char	*fragtbl[];
24 
25 /*
26  * Allocate a block in the file system.
27  *
28  * The size of the requested block is given, which must be some
29  * multiple of fs_fsize and <= fs_bsize.
30  * A preference may be optionally specified. If a preference is given
31  * the following hierarchy is used to allocate a block:
32  *   1) allocate the requested block.
33  *   2) allocate a rotationally optimal block in the same cylinder.
34  *   3) allocate a block in the same cylinder group.
35  *   4) quadradically rehash into other cylinder groups, until an
36  *      available block is located.
37  * If no block preference is given the following heirarchy is used
38  * to allocate a block:
39  *   1) allocate a block in the cylinder group that contains the
40  *      inode for the file.
41  *   2) quadradically rehash into other cylinder groups, until an
42  *      available block is located.
43  */
44 struct buf *
45 alloc(ip, bpref, size)
46 	register struct inode *ip;
47 	daddr_t bpref;
48 	int size;
49 {
50 	daddr_t bno;
51 	register struct fs *fs;
52 	register struct buf *bp;
53 	int cg;
54 
55 	fs = ip->i_fs;
56 	if ((unsigned)size > fs->fs_bsize || fragoff(fs, size) != 0) {
57 		printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n",
58 		    ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt);
59 		panic("alloc: bad size");
60 	}
61 	if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
62 		goto nospace;
63 	if (u.u_uid != 0 &&
64 	    fs->fs_cstotal.cs_nbfree * fs->fs_frag + fs->fs_cstotal.cs_nffree <
65 	      fs->fs_dsize * fs->fs_minfree / 100)
66 		goto nospace;
67 #ifdef QUOTA
68 	if (chkdq(ip, (long)((unsigned)size/DEV_BSIZE), 0))
69 		return(NULL);
70 #endif
71 	if (bpref >= fs->fs_size)
72 		bpref = 0;
73 	if (bpref == 0)
74 		cg = itog(fs, ip->i_number);
75 	else
76 		cg = dtog(fs, bpref);
77 	bno = (daddr_t)hashalloc(ip, cg, (long)bpref, size, alloccg);
78 	if (bno <= 0)
79 		goto nospace;
80 	bp = getblk(ip->i_dev, fsbtodb(fs, bno), size);
81 	clrbuf(bp);
82 	return (bp);
83 nospace:
84 	fserr(fs, "file system full");
85 	uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
86 	u.u_error = ENOSPC;
87 	return (NULL);
88 }
89 
90 /*
91  * Reallocate a fragment to a bigger size
92  *
93  * The number and size of the old block is given, and a preference
94  * and new size is also specified. The allocator attempts to extend
95  * the original block. Failing that, the regular block allocator is
96  * invoked to get an appropriate block.
97  */
98 struct buf *
99 realloccg(ip, bprev, bpref, osize, nsize)
100 	register struct inode *ip;
101 	daddr_t bprev, bpref;
102 	int osize, nsize;
103 {
104 	daddr_t bno;
105 	register struct fs *fs;
106 	register struct buf *bp, *obp;
107 	int cg;
108 
109 	fs = ip->i_fs;
110 	if ((unsigned)osize > fs->fs_bsize || fragoff(fs, osize) != 0 ||
111 	    (unsigned)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) {
112 		printf("dev = 0x%x, bsize = %d, osize = %d, nsize = %d, fs = %s\n",
113 		    ip->i_dev, fs->fs_bsize, osize, nsize, fs->fs_fsmnt);
114 		panic("realloccg: bad size");
115 	}
116 	if (u.u_uid != 0 &&
117 	    fs->fs_cstotal.cs_nbfree * fs->fs_frag + fs->fs_cstotal.cs_nffree <
118 	      fs->fs_dsize * fs->fs_minfree / 100)
119 		goto nospace;
120 	if (bprev == 0) {
121 		printf("dev = 0x%x, bsize = %d, bprev = %d, fs = %s\n",
122 		    ip->i_dev, fs->fs_bsize, bprev, fs->fs_fsmnt);
123 		panic("realloccg: bad bprev");
124 	}
125 #ifdef QUOTA
126 	if (chkdq(ip, (long)((unsigned)(nsize-osize)/DEV_BSIZE), 0))
127 		return(NULL);
128 #endif
129 	cg = dtog(fs, bprev);
130 	bno = fragextend(ip, cg, (long)bprev, osize, nsize);
131 	if (bno != 0) {
132 		do {
133 			bp = bread(ip->i_dev, fsbtodb(fs, bno), osize);
134 			if (bp->b_flags & B_ERROR) {
135 				brelse(bp);
136 				return (NULL);
137 			}
138 		} while (brealloc(bp, nsize) == 0);
139 		bp->b_flags |= B_DONE;
140 		bzero(bp->b_un.b_addr + osize, nsize - osize);
141 		return (bp);
142 	}
143 	if (bpref >= fs->fs_size)
144 		bpref = 0;
145 	bno = (daddr_t)hashalloc(ip, cg, (long)bpref, nsize, alloccg);
146 	if (bno > 0) {
147 		obp = bread(ip->i_dev, fsbtodb(fs, bprev), osize);
148 		if (obp->b_flags & B_ERROR) {
149 			brelse(obp);
150 			return (NULL);
151 		}
152 		bp = getblk(ip->i_dev, fsbtodb(fs, bno), nsize);
153 		bcopy(obp->b_un.b_addr, bp->b_un.b_addr, (u_int)osize);
154 		bzero(bp->b_un.b_addr + osize, nsize - osize);
155 		brelse(obp);
156 		fre(ip, bprev, (off_t)osize);
157 		return (bp);
158 	}
159 nospace:
160 	/*
161 	 * no space available
162 	 */
163 	fserr(fs, "file system full");
164 	uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
165 	u.u_error = ENOSPC;
166 	return (NULL);
167 }
168 
169 /*
170  * Allocate an inode in the file system.
171  *
172  * A preference may be optionally specified. If a preference is given
173  * the following hierarchy is used to allocate an inode:
174  *   1) allocate the requested inode.
175  *   2) allocate an inode in the same cylinder group.
176  *   3) quadradically rehash into other cylinder groups, until an
177  *      available inode is located.
178  * If no inode preference is given the following heirarchy is used
179  * to allocate an inode:
180  *   1) allocate an inode in cylinder group 0.
181  *   2) quadradically rehash into other cylinder groups, until an
182  *      available inode is located.
183  */
184 struct inode *
185 ialloc(pip, ipref, mode)
186 	register struct inode *pip;
187 	ino_t ipref;
188 	int mode;
189 {
190 	ino_t ino;
191 	register struct fs *fs;
192 	register struct inode *ip;
193 	int cg;
194 
195 	fs = pip->i_fs;
196 	if (fs->fs_cstotal.cs_nifree == 0)
197 		goto noinodes;
198 #ifdef QUOTA
199 	if (chkiq(pip->i_dev, NULL, u.u_uid, 0))
200 		return(NULL);
201 #endif
202 	if (ipref >= fs->fs_ncg * fs->fs_ipg)
203 		ipref = 0;
204 	cg = itog(fs, ipref);
205 	ino = (ino_t)hashalloc(pip, cg, (long)ipref, mode, ialloccg);
206 	if (ino == 0)
207 		goto noinodes;
208 	ip = iget(pip->i_dev, pip->i_fs, ino);
209 	if (ip == NULL) {
210 		ifree(ip, ino, 0);
211 		return (NULL);
212 	}
213 	if (ip->i_mode) {
214 		printf("mode = 0%o, inum = %d, fs = %s\n",
215 		    ip->i_mode, ip->i_number, fs->fs_fsmnt);
216 		panic("ialloc: dup alloc");
217 	}
218 	return (ip);
219 noinodes:
220 	fserr(fs, "out of inodes");
221 	uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
222 	u.u_error = ENOSPC;
223 	return (NULL);
224 }
225 
226 /*
227  * Find a cylinder to place a directory.
228  *
229  * The policy implemented by this algorithm is to select from
230  * among those cylinder groups with above the average number of
231  * free inodes, the one with the smallest number of directories.
232  */
233 dirpref(fs)
234 	register struct fs *fs;
235 {
236 	int cg, minndir, mincg, avgifree;
237 
238 	avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;
239 	minndir = fs->fs_ipg;
240 	mincg = 0;
241 	for (cg = 0; cg < fs->fs_ncg; cg++)
242 		if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
243 		    fs->fs_cs(fs, cg).cs_nifree >= avgifree) {
244 			mincg = cg;
245 			minndir = fs->fs_cs(fs, cg).cs_ndir;
246 		}
247 	return (fs->fs_ipg * mincg);
248 }
249 
250 /*
251  * Select a cylinder to place a large block of data.
252  *
253  * The policy implemented by this algorithm is to maintain a
254  * rotor that sweeps the cylinder groups. When a block is
255  * needed, the rotor is advanced until a cylinder group with
256  * greater than the average number of free blocks is found.
257  */
258 daddr_t
259 blkpref(fs)
260 	register struct fs *fs;
261 {
262 	int cg, avgbfree;
263 
264 	avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
265 	for (cg = fs->fs_cgrotor + 1; cg < fs->fs_ncg; cg++)
266 		if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
267 			fs->fs_cgrotor = cg;
268 			return (fs->fs_fpg * cg + fs->fs_frag);
269 		}
270 	for (cg = 0; cg <= fs->fs_cgrotor; cg++)
271 		if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
272 			fs->fs_cgrotor = cg;
273 			return (fs->fs_fpg * cg + fs->fs_frag);
274 		}
275 	return (NULL);
276 }
277 
278 /*
279  * Implement the cylinder overflow algorithm.
280  *
281  * The policy implemented by this algorithm is:
282  *   1) allocate the block in its requested cylinder group.
283  *   2) quadradically rehash on the cylinder group number.
284  *   3) brute force search for a free block.
285  */
286 /*VARARGS5*/
287 u_long
288 hashalloc(ip, cg, pref, size, allocator)
289 	struct inode *ip;
290 	int cg;
291 	long pref;
292 	int size;	/* size for data blocks, mode for inodes */
293 	u_long (*allocator)();
294 {
295 	register struct fs *fs;
296 	long result;
297 	int i, icg = cg;
298 
299 	fs = ip->i_fs;
300 	/*
301 	 * 1: preferred cylinder group
302 	 */
303 	result = (*allocator)(ip, cg, pref, size);
304 	if (result)
305 		return (result);
306 	/*
307 	 * 2: quadratic rehash
308 	 */
309 	for (i = 1; i < fs->fs_ncg; i *= 2) {
310 		cg += i;
311 		if (cg >= fs->fs_ncg)
312 			cg -= fs->fs_ncg;
313 		result = (*allocator)(ip, cg, 0, size);
314 		if (result)
315 			return (result);
316 	}
317 	/*
318 	 * 3: brute force search
319 	 */
320 	cg = icg;
321 	for (i = 0; i < fs->fs_ncg; i++) {
322 		result = (*allocator)(ip, cg, 0, size);
323 		if (result)
324 			return (result);
325 		cg++;
326 		if (cg == fs->fs_ncg)
327 			cg = 0;
328 	}
329 	return (NULL);
330 }
331 
332 /*
333  * Determine whether a fragment can be extended.
334  *
335  * Check to see if the necessary fragments are available, and
336  * if they are, allocate them.
337  */
338 daddr_t
339 fragextend(ip, cg, bprev, osize, nsize)
340 	struct inode *ip;
341 	int cg;
342 	long bprev;
343 	int osize, nsize;
344 {
345 	register struct fs *fs;
346 	register struct buf *bp;
347 	register struct cg *cgp;
348 	long bno;
349 	int frags, bbase;
350 	int i;
351 
352 	fs = ip->i_fs;
353 	if (fs->fs_cs(fs, cg).cs_nffree < nsize - osize)
354 		return (NULL);
355 	frags = numfrags(fs, nsize);
356 	bbase = fragoff(fs, bprev);
357 	if (bbase > (bprev + frags - 1) % fs->fs_frag) {
358 		/* cannot extend across a block boundry */
359 		return (NULL);
360 	}
361 	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize);
362 	cgp = bp->b_un.b_cg;
363 	if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) {
364 		brelse(bp);
365 		return (NULL);
366 	}
367 	cgp->cg_time = time.tv_sec;
368 	bno = dtogd(fs, bprev);
369 	for (i = numfrags(fs, osize); i < frags; i++)
370 		if (isclr(cgp->cg_free, bno + i)) {
371 			brelse(bp);
372 			return (NULL);
373 		}
374 	/*
375 	 * the current fragment can be extended
376 	 * deduct the count on fragment being extended into
377 	 * increase the count on the remaining fragment (if any)
378 	 * allocate the extended piece
379 	 */
380 	for (i = frags; i < fs->fs_frag - bbase; i++)
381 		if (isclr(cgp->cg_free, bno + i))
382 			break;
383 	cgp->cg_frsum[i - numfrags(fs, osize)]--;
384 	if (i != frags)
385 		cgp->cg_frsum[i - frags]++;
386 	for (i = numfrags(fs, osize); i < frags; i++) {
387 		clrbit(cgp->cg_free, bno + i);
388 		cgp->cg_cs.cs_nffree--;
389 		fs->fs_cstotal.cs_nffree--;
390 		fs->fs_cs(fs, cg).cs_nffree--;
391 	}
392 	fs->fs_fmod++;
393 	bdwrite(bp);
394 	return (bprev);
395 }
396 
397 /*
398  * Determine whether a block can be allocated.
399  *
400  * Check to see if a block of the apprpriate size is available,
401  * and if it is, allocate it.
402  */
403 u_long
404 alloccg(ip, cg, bpref, size)
405 	struct inode *ip;
406 	int cg;
407 	daddr_t bpref;
408 	int size;
409 {
410 	register struct fs *fs;
411 	register struct buf *bp;
412 	register struct cg *cgp;
413 	int bno, frags;
414 	int allocsiz;
415 	register int i;
416 
417 	fs = ip->i_fs;
418 	if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
419 		return (NULL);
420 	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize);
421 	cgp = bp->b_un.b_cg;
422 	if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) {
423 		brelse(bp);
424 		return (NULL);
425 	}
426 	cgp->cg_time = time.tv_sec;
427 	if (size == fs->fs_bsize) {
428 		bno = alloccgblk(fs, cgp, bpref);
429 		bdwrite(bp);
430 		return (bno);
431 	}
432 	/*
433 	 * check to see if any fragments are already available
434 	 * allocsiz is the size which will be allocated, hacking
435 	 * it down to a smaller size if necessary
436 	 */
437 	frags = numfrags(fs, size);
438 	for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)
439 		if (cgp->cg_frsum[allocsiz] != 0)
440 			break;
441 	if (allocsiz == fs->fs_frag) {
442 		/*
443 		 * no fragments were available, so a block will be
444 		 * allocated, and hacked up
445 		 */
446 		if (cgp->cg_cs.cs_nbfree == 0) {
447 			brelse(bp);
448 			return (NULL);
449 		}
450 		bno = alloccgblk(fs, cgp, bpref);
451 		bpref = dtogd(fs, bno);
452 		for (i = frags; i < fs->fs_frag; i++)
453 			setbit(cgp->cg_free, bpref + i);
454 		i = fs->fs_frag - frags;
455 		cgp->cg_cs.cs_nffree += i;
456 		fs->fs_cstotal.cs_nffree += i;
457 		fs->fs_cs(fs, cg).cs_nffree += i;
458 		cgp->cg_frsum[i]++;
459 		bdwrite(bp);
460 		return (bno);
461 	}
462 	bno = mapsearch(fs, cgp, bpref, allocsiz);
463 	if (bno < 0)
464 		return (NULL);
465 	for (i = 0; i < frags; i++)
466 		clrbit(cgp->cg_free, bno + i);
467 	cgp->cg_cs.cs_nffree -= frags;
468 	fs->fs_cstotal.cs_nffree -= frags;
469 	fs->fs_cs(fs, cg).cs_nffree -= frags;
470 	cgp->cg_frsum[allocsiz]--;
471 	if (frags != allocsiz)
472 		cgp->cg_frsum[allocsiz - frags]++;
473 	bdwrite(bp);
474 	return (cg * fs->fs_fpg + bno);
475 }
476 
477 /*
478  * Allocate a block in a cylinder group.
479  *
480  * This algorithm implements the following policy:
481  *   1) allocate the requested block.
482  *   2) allocate a rotationally optimal block in the same cylinder.
483  *   3) allocate the next available block on the block rotor for the
484  *      specified cylinder group.
485  * Note that this routine only allocates fs_bsize blocks; these
486  * blocks may be fragmented by the routine that allocates them.
487  */
488 daddr_t
489 alloccgblk(fs, cgp, bpref)
490 	register struct fs *fs;
491 	register struct cg *cgp;
492 	daddr_t bpref;
493 {
494 	daddr_t bno;
495 	int cylno, pos, delta;
496 	short *cylbp;
497 	register int i;
498 
499 	if (bpref == 0) {
500 		bpref = cgp->cg_rotor;
501 		goto norot;
502 	}
503 	bpref &= ~(fs->fs_frag - 1);
504 	bpref = dtogd(fs, bpref);
505 	/*
506 	 * if the requested block is available, use it
507 	 */
508 /*
509  * disallow sequential layout.
510  *
511 	if (isblock(fs, cgp->cg_free, bpref/fs->fs_frag)) {
512 		bno = bpref;
513 		goto gotit;
514 	}
515  */
516 	/*
517 	 * check for a block available on the same cylinder
518 	 */
519 	cylno = cbtocylno(fs, bpref);
520 	if (cgp->cg_btot[cylno] == 0)
521 		goto norot;
522 	if (fs->fs_cpc == 0) {
523 		/*
524 		 * block layout info is not available, so just have
525 		 * to take any block in this cylinder.
526 		 */
527 		bpref = howmany(fs->fs_spc * cylno, NSPF(fs));
528 		goto norot;
529 	}
530 	/*
531 	 * find a block that is rotationally optimal
532 	 */
533 	cylbp = cgp->cg_b[cylno];
534 	if (fs->fs_rotdelay == 0) {
535 		pos = cbtorpos(fs, bpref);
536 	} else {
537 		/*
538 		 * here we convert ms of delay to frags as:
539 		 * (frags) = (ms) * (rev/sec) * (sect/rev) /
540 		 *	((sect/frag) * (ms/sec))
541 		 * then round up to the next rotational position
542 		 */
543 		bpref += fs->fs_rotdelay * fs->fs_rps * fs->fs_nsect /
544 		    (NSPF(fs) * 1000);
545 		pos = cbtorpos(fs, bpref);
546 		pos = (pos + 1) % NRPOS;
547 	}
548 	/*
549 	 * check the summary information to see if a block is
550 	 * available in the requested cylinder starting at the
551 	 * optimal rotational position and proceeding around.
552 	 */
553 	for (i = pos; i < NRPOS; i++)
554 		if (cylbp[i] > 0)
555 			break;
556 	if (i == NRPOS)
557 		for (i = 0; i < pos; i++)
558 			if (cylbp[i] > 0)
559 				break;
560 	if (cylbp[i] > 0) {
561 		/*
562 		 * found a rotational position, now find the actual
563 		 * block. A panic if none is actually there.
564 		 */
565 		pos = cylno % fs->fs_cpc;
566 		bno = (cylno - pos) * fs->fs_spc / NSPB(fs);
567 		if (fs->fs_postbl[pos][i] == -1) {
568 			printf("pos = %d, i = %d, fs = %s\n",
569 			    pos, i, fs->fs_fsmnt);
570 			panic("alloccgblk: cyl groups corrupted");
571 		}
572 		for (i = fs->fs_postbl[pos][i];; ) {
573 			if (isblock(fs, cgp->cg_free, bno + i)) {
574 				bno = (bno + i) * fs->fs_frag;
575 				goto gotit;
576 			}
577 			delta = fs->fs_rotbl[i];
578 			if (delta <= 0 || delta > MAXBPC - i)
579 				break;
580 			i += delta;
581 		}
582 		printf("pos = %d, i = %d, fs = %s\n", pos, i, fs->fs_fsmnt);
583 		panic("alloccgblk: can't find blk in cyl");
584 	}
585 norot:
586 	/*
587 	 * no blocks in the requested cylinder, so take next
588 	 * available one in this cylinder group.
589 	 */
590 	bno = mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
591 	if (bno < 0)
592 		return (NULL);
593 	cgp->cg_rotor = bno;
594 gotit:
595 	clrblock(fs, cgp->cg_free, (long)(bno/fs->fs_frag));
596 	cgp->cg_cs.cs_nbfree--;
597 	fs->fs_cstotal.cs_nbfree--;
598 	fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--;
599 	cylno = cbtocylno(fs, bno);
600 	cgp->cg_b[cylno][cbtorpos(fs, bno)]--;
601 	cgp->cg_btot[cylno]--;
602 	fs->fs_fmod++;
603 	return (cgp->cg_cgx * fs->fs_fpg + bno);
604 }
605 
606 /*
607  * Determine whether an inode can be allocated.
608  *
609  * Check to see if an inode is available, and if it is,
610  * allocate it using the following policy:
611  *   1) allocate the requested inode.
612  *   2) allocate the next available inode after the requested
613  *      inode in the specified cylinder group.
614  */
615 u_long
616 ialloccg(ip, cg, ipref, mode)
617 	struct inode *ip;
618 	int cg;
619 	daddr_t ipref;
620 	int mode;
621 {
622 	register struct fs *fs;
623 	register struct buf *bp;
624 	register struct cg *cgp;
625 	int i;
626 
627 	fs = ip->i_fs;
628 	if (fs->fs_cs(fs, cg).cs_nifree == 0)
629 		return (NULL);
630 	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize);
631 	cgp = bp->b_un.b_cg;
632 	if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) {
633 		brelse(bp);
634 		return (NULL);
635 	}
636 	cgp->cg_time = time.tv_sec;
637 	if (ipref) {
638 		ipref %= fs->fs_ipg;
639 		if (isclr(cgp->cg_iused, ipref))
640 			goto gotit;
641 	} else
642 		ipref = cgp->cg_irotor;
643 	for (i = 0; i < fs->fs_ipg; i++) {
644 		ipref++;
645 		if (ipref >= fs->fs_ipg)
646 			ipref = 0;
647 		if (isclr(cgp->cg_iused, ipref)) {
648 			cgp->cg_irotor = ipref;
649 			goto gotit;
650 		}
651 	}
652 	brelse(bp);
653 	return (NULL);
654 gotit:
655 	setbit(cgp->cg_iused, ipref);
656 	cgp->cg_cs.cs_nifree--;
657 	fs->fs_cstotal.cs_nifree--;
658 	fs->fs_cs(fs, cg).cs_nifree--;
659 	fs->fs_fmod++;
660 	if ((mode & IFMT) == IFDIR) {
661 		cgp->cg_cs.cs_ndir++;
662 		fs->fs_cstotal.cs_ndir++;
663 		fs->fs_cs(fs, cg).cs_ndir++;
664 	}
665 	bdwrite(bp);
666 	return (cg * fs->fs_ipg + ipref);
667 }
668 
669 /*
670  * Free a block or fragment.
671  *
672  * The specified block or fragment is placed back in the
673  * free map. If a fragment is deallocated, a possible
674  * block reassembly is checked.
675  */
676 fre(ip, bno, size)
677 	register struct inode *ip;
678 	daddr_t bno;
679 	off_t size;
680 {
681 	register struct fs *fs;
682 	register struct cg *cgp;
683 	register struct buf *bp;
684 	int cg, blk, frags, bbase;
685 	register int i;
686 
687 	fs = ip->i_fs;
688 	if ((unsigned)size > fs->fs_bsize || fragoff(fs, size) != 0) {
689 		printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n",
690 		    ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt);
691 		panic("free: bad size");
692 	}
693 	cg = dtog(fs, bno);
694 	if (badblock(fs, bno)) {
695 		printf("bad block %d, ino %d\n", bno, ip->i_number);
696 		return;
697 	}
698 	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize);
699 	cgp = bp->b_un.b_cg;
700 	if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) {
701 		brelse(bp);
702 		return;
703 	}
704 	cgp->cg_time = time.tv_sec;
705 	bno = dtogd(fs, bno);
706 	if (size == fs->fs_bsize) {
707 		if (isblock(fs, cgp->cg_free, bno/fs->fs_frag)) {
708 			printf("dev = 0x%x, block = %d, fs = %s\n",
709 			    ip->i_dev, bno, fs->fs_fsmnt);
710 			panic("free: freeing free block");
711 		}
712 		setblock(fs, cgp->cg_free, bno/fs->fs_frag);
713 		cgp->cg_cs.cs_nbfree++;
714 		fs->fs_cstotal.cs_nbfree++;
715 		fs->fs_cs(fs, cg).cs_nbfree++;
716 		i = cbtocylno(fs, bno);
717 		cgp->cg_b[i][cbtorpos(fs, bno)]++;
718 		cgp->cg_btot[i]++;
719 	} else {
720 		bbase = bno - (bno % fs->fs_frag);
721 		/*
722 		 * decrement the counts associated with the old frags
723 		 */
724 		blk = blkmap(fs, cgp->cg_free, bbase);
725 		fragacct(fs, blk, cgp->cg_frsum, -1);
726 		/*
727 		 * deallocate the fragment
728 		 */
729 		frags = numfrags(fs, size);
730 		for (i = 0; i < frags; i++) {
731 			if (isset(cgp->cg_free, bno + i)) {
732 				printf("dev = 0x%x, block = %d, fs = %s\n",
733 				    ip->i_dev, bno + i, fs->fs_fsmnt);
734 				panic("free: freeing free frag");
735 			}
736 			setbit(cgp->cg_free, bno + i);
737 		}
738 		cgp->cg_cs.cs_nffree += i;
739 		fs->fs_cstotal.cs_nffree += i;
740 		fs->fs_cs(fs, cg).cs_nffree += i;
741 		/*
742 		 * add back in counts associated with the new frags
743 		 */
744 		blk = blkmap(fs, cgp->cg_free, bbase);
745 		fragacct(fs, blk, cgp->cg_frsum, 1);
746 		/*
747 		 * if a complete block has been reassembled, account for it
748 		 */
749 		if (isblock(fs, cgp->cg_free, bbase / fs->fs_frag)) {
750 			cgp->cg_cs.cs_nffree -= fs->fs_frag;
751 			fs->fs_cstotal.cs_nffree -= fs->fs_frag;
752 			fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
753 			cgp->cg_cs.cs_nbfree++;
754 			fs->fs_cstotal.cs_nbfree++;
755 			fs->fs_cs(fs, cg).cs_nbfree++;
756 			i = cbtocylno(fs, bbase);
757 			cgp->cg_b[i][cbtorpos(fs, bbase)]++;
758 			cgp->cg_btot[i]++;
759 		}
760 	}
761 	fs->fs_fmod++;
762 	bdwrite(bp);
763 }
764 
765 /*
766  * Free an inode.
767  *
768  * The specified inode is placed back in the free map.
769  */
770 ifree(ip, ino, mode)
771 	struct inode *ip;
772 	ino_t ino;
773 	int mode;
774 {
775 	register struct fs *fs;
776 	register struct cg *cgp;
777 	register struct buf *bp;
778 	int cg;
779 
780 	fs = ip->i_fs;
781 	if ((unsigned)ino >= fs->fs_ipg*fs->fs_ncg) {
782 		printf("dev = 0x%x, ino = %d, fs = %s\n",
783 		    ip->i_dev, ino, fs->fs_fsmnt);
784 		panic("ifree: range");
785 	}
786 	cg = itog(fs, ino);
787 	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize);
788 	cgp = bp->b_un.b_cg;
789 	if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) {
790 		brelse(bp);
791 		return;
792 	}
793 	cgp->cg_time = time.tv_sec;
794 	ino %= fs->fs_ipg;
795 	if (isclr(cgp->cg_iused, ino)) {
796 		printf("dev = 0x%x, ino = %d, fs = %s\n",
797 		    ip->i_dev, ino, fs->fs_fsmnt);
798 		panic("ifree: freeing free inode");
799 	}
800 	clrbit(cgp->cg_iused, ino);
801 	cgp->cg_cs.cs_nifree++;
802 	fs->fs_cstotal.cs_nifree++;
803 	fs->fs_cs(fs, cg).cs_nifree++;
804 	if ((mode & IFMT) == IFDIR) {
805 		cgp->cg_cs.cs_ndir--;
806 		fs->fs_cstotal.cs_ndir--;
807 		fs->fs_cs(fs, cg).cs_ndir--;
808 	}
809 	fs->fs_fmod++;
810 	bdwrite(bp);
811 }
812 
813 /*
814  * Find a block of the specified size in the specified cylinder group.
815  *
816  * It is a panic if a request is made to find a block if none are
817  * available.
818  */
819 daddr_t
820 mapsearch(fs, cgp, bpref, allocsiz)
821 	register struct fs *fs;
822 	register struct cg *cgp;
823 	daddr_t bpref;
824 	int allocsiz;
825 {
826 	daddr_t bno;
827 	int start, len, loc, i;
828 	int blk, field, subfield, pos;
829 
830 	/*
831 	 * find the fragment by searching through the free block
832 	 * map for an appropriate bit pattern
833 	 */
834 	if (bpref)
835 		start = dtogd(fs, bpref) / NBBY;
836 	else
837 		start = cgp->cg_frotor / NBBY;
838 	len = howmany(fs->fs_fpg, NBBY) - start;
839 	loc = scanc(len, &cgp->cg_free[start], fragtbl[fs->fs_frag],
840 		1 << (allocsiz - 1 + (fs->fs_frag % NBBY)));
841 	if (loc == 0) {
842 		len = start + 1;
843 		start = 0;
844 		loc = scanc(len, &cgp->cg_free[start], fragtbl[fs->fs_frag],
845 			1 << (allocsiz - 1 + (fs->fs_frag % NBBY)));
846 		if (loc == 0) {
847 			printf("start = %d, len = %d, fs = %s\n",
848 			    start, len, fs->fs_fsmnt);
849 			panic("alloccg: map corrupted");
850 			return (-1);
851 		}
852 	}
853 	bno = (start + len - loc) * NBBY;
854 	cgp->cg_frotor = bno;
855 	/*
856 	 * found the byte in the map
857 	 * sift through the bits to find the selected frag
858 	 */
859 	for (i = bno + NBBY; bno < i; bno += fs->fs_frag) {
860 		blk = blkmap(fs, cgp->cg_free, bno);
861 		blk <<= 1;
862 		field = around[allocsiz];
863 		subfield = inside[allocsiz];
864 		for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) {
865 			if ((blk & field) == subfield)
866 				return (bno + pos);
867 			field <<= 1;
868 			subfield <<= 1;
869 		}
870 	}
871 	printf("bno = %d, fs = %s\n", bno, fs->fs_fsmnt);
872 	panic("alloccg: block not in map");
873 	return (-1);
874 }
875 
876 /*
877  * Getfs maps a device number into a pointer to the incore super block.
878  *
879  * The algorithm is a linear search through the mount table. A
880  * consistency check of the super block magic number is performed.
881  *
882  * panic: no fs -- the device is not mounted.
883  *	this "cannot happen"
884  */
885 struct fs *
886 getfs(dev)
887 	dev_t dev;
888 {
889 	register struct mount *mp;
890 	register struct fs *fs;
891 
892 	for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++) {
893 		if (mp->m_bufp == NULL || mp->m_dev != dev)
894 			continue;
895 		fs = mp->m_bufp->b_un.b_fs;
896 		if (fs->fs_magic != FS_MAGIC) {
897 			printf("dev = 0x%x, fs = %s\n", dev, fs->fs_fsmnt);
898 			panic("getfs: bad magic");
899 		}
900 		return (fs);
901 	}
902 	printf("dev = 0x%x\n", dev);
903 	panic("getfs: no fs");
904 	return (NULL);
905 }
906 
907 /*
908  * Fserr prints the name of a file system with an error diagnostic.
909  *
910  * The form of the error message is:
911  *	fs: error message
912  */
913 fserr(fs, cp)
914 	struct fs *fs;
915 	char *cp;
916 {
917 
918 	printf("%s: %s\n", fs->fs_fsmnt, cp);
919 }
920 
921 /*
922  * Getfsx returns the index in the file system
923  * table of the specified device.  The swap device
924  * is also assigned a pseudo-index.  The index may
925  * be used as a compressed indication of the location
926  * of a block, recording
927  *	<getfsx(dev),blkno>
928  * rather than
929  *	<dev, blkno>
930  * provided the information need remain valid only
931  * as long as the file system is mounted.
932  */
933 getfsx(dev)
934 	dev_t dev;
935 {
936 	register struct mount *mp;
937 
938 	if (dev == swapdev)
939 		return (MSWAPX);
940 	for(mp = &mount[0]; mp < &mount[NMOUNT]; mp++)
941 		if (mp->m_dev == dev)
942 			return (mp - &mount[0]);
943 	return (-1);
944 }
945 
946 /*
947  * Update is the internal name of 'sync'.  It goes through the disk
948  * queues to initiate sandbagged IO; goes through the inodes to write
949  * modified nodes; and it goes through the mount table to initiate
950  * the writing of the modified super blocks.
951  */
952 update()
953 {
954 	register struct inode *ip;
955 	register struct mount *mp;
956 	struct fs *fs;
957 
958 	if (updlock)
959 		return;
960 	updlock++;
961 	/*
962 	 * Write back modified superblocks.
963 	 * Consistency check that the superblock
964 	 * of each file system is still in the buffer cache.
965 	 */
966 	for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++) {
967 		if (mp->m_bufp == NULL)
968 			continue;
969 		fs = mp->m_bufp->b_un.b_fs;
970 		if (fs->fs_fmod == 0)
971 			continue;
972 		if (fs->fs_ronly != 0) {		/* XXX */
973 			printf("fs = %s\n", fs->fs_fsmnt);
974 			panic("update: rofs mod");
975 		}
976 		fs->fs_fmod = 0;
977 		fs->fs_time = time.tv_sec;
978 		sbupdate(mp);
979 	}
980 	/*
981 	 * Write back each (modified) inode.
982 	 */
983 	for (ip = inode; ip < inodeNINODE; ip++) {
984 		if ((ip->i_flag & ILOCKED) != 0 || ip->i_count == 0)
985 			continue;
986 		ip->i_flag |= ILOCKED;
987 		ip->i_count++;
988 		iupdat(ip, &time, &time, 0);
989 		iput(ip);
990 	}
991 	updlock = 0;
992 	/*
993 	 * Force stale buffer cache information to be flushed,
994 	 * for all devices.
995 	 */
996 	bflush(NODEV);
997 }
998 
999