Home | History | Annotate | Line # | Download | only in ffs
ffs_alloc.c revision 1.96
      1 /*	$NetBSD: ffs_alloc.c,v 1.96 2006/11/16 01:33:53 christos Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 2002 Networks Associates Technology, Inc.
      5  * All rights reserved.
      6  *
      7  * This software was developed for the FreeBSD Project by Marshall
      8  * Kirk McKusick and Network Associates Laboratories, the Security
      9  * Research Division of Network Associates, Inc. under DARPA/SPAWAR
     10  * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS
     11  * research program
     12  *
     13  * Copyright (c) 1982, 1986, 1989, 1993
     14  *	The Regents of the University of California.  All rights reserved.
     15  *
     16  * Redistribution and use in source and binary forms, with or without
     17  * modification, are permitted provided that the following conditions
     18  * are met:
     19  * 1. Redistributions of source code must retain the above copyright
     20  *    notice, this list of conditions and the following disclaimer.
     21  * 2. Redistributions in binary form must reproduce the above copyright
     22  *    notice, this list of conditions and the following disclaimer in the
     23  *    documentation and/or other materials provided with the distribution.
     24  * 3. Neither the name of the University nor the names of its contributors
     25  *    may be used to endorse or promote products derived from this software
     26  *    without specific prior written permission.
     27  *
     28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     38  * SUCH DAMAGE.
     39  *
     40  *	@(#)ffs_alloc.c	8.19 (Berkeley) 7/13/95
     41  */
     42 
     43 #include <sys/cdefs.h>
     44 __KERNEL_RCSID(0, "$NetBSD: ffs_alloc.c,v 1.96 2006/11/16 01:33:53 christos Exp $");
     45 
     46 #if defined(_KERNEL_OPT)
     47 #include "opt_ffs.h"
     48 #include "opt_quota.h"
     49 #endif
     50 
     51 #include <sys/param.h>
     52 #include <sys/systm.h>
     53 #include <sys/buf.h>
     54 #include <sys/proc.h>
     55 #include <sys/vnode.h>
     56 #include <sys/mount.h>
     57 #include <sys/kernel.h>
     58 #include <sys/syslog.h>
     59 #include <sys/kauth.h>
     60 
     61 #include <miscfs/specfs/specdev.h>
     62 #include <ufs/ufs/quota.h>
     63 #include <ufs/ufs/ufsmount.h>
     64 #include <ufs/ufs/inode.h>
     65 #include <ufs/ufs/ufs_extern.h>
     66 #include <ufs/ufs/ufs_bswap.h>
     67 
     68 #include <ufs/ffs/fs.h>
     69 #include <ufs/ffs/ffs_extern.h>
     70 
     71 static daddr_t ffs_alloccg(struct inode *, int, daddr_t, int);
     72 static daddr_t ffs_alloccgblk(struct inode *, struct buf *, daddr_t);
     73 #ifdef XXXUBC
     74 static daddr_t ffs_clusteralloc(struct inode *, int, daddr_t, int);
     75 #endif
     76 static ino_t ffs_dirpref(struct inode *);
     77 static daddr_t ffs_fragextend(struct inode *, int, daddr_t, int, int);
     78 static void ffs_fserr(struct fs *, u_int, const char *);
     79 static daddr_t ffs_hashalloc(struct inode *, int, daddr_t, int,
     80     daddr_t (*)(struct inode *, int, daddr_t, int));
     81 static daddr_t ffs_nodealloccg(struct inode *, int, daddr_t, int);
     82 static int32_t ffs_mapsearch(struct fs *, struct cg *,
     83 				      daddr_t, int);
     84 #if defined(DIAGNOSTIC) || defined(DEBUG)
     85 #ifdef XXXUBC
     86 static int ffs_checkblk(struct inode *, daddr_t, long size);
     87 #endif
     88 #endif
     89 
     90 /* if 1, changes in optimalization strategy are logged */
     91 int ffs_log_changeopt = 0;
     92 
     93 /* in ffs_tables.c */
     94 extern const int inside[], around[];
     95 extern const u_char * const fragtbl[];
     96 
     97 /*
     98  * Allocate a block in the file system.
     99  *
    100  * The size of the requested block is given, which must be some
    101  * multiple of fs_fsize and <= fs_bsize.
    102  * A preference may be optionally specified. If a preference is given
    103  * the following hierarchy is used to allocate a block:
    104  *   1) allocate the requested block.
    105  *   2) allocate a rotationally optimal block in the same cylinder.
    106  *   3) allocate a block in the same cylinder group.
    107  *   4) quadradically rehash into other cylinder groups, until an
    108  *      available block is located.
    109  * If no block preference is given the following hierarchy is used
    110  * to allocate a block:
    111  *   1) allocate a block in the cylinder group that contains the
    112  *      inode for the file.
    113  *   2) quadradically rehash into other cylinder groups, until an
    114  *      available block is located.
    115  */
    116 int
    117 ffs_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref, int size,
    118     kauth_cred_t cred, daddr_t *bnp)
    119 {
    120 	struct fs *fs;
    121 	daddr_t bno;
    122 	int cg;
    123 #ifdef QUOTA
    124 	int error;
    125 #endif
    126 
    127 	fs = ip->i_fs;
    128 
    129 #ifdef UVM_PAGE_TRKOWN
    130 	if (ITOV(ip)->v_type == VREG &&
    131 	    lblktosize(fs, (voff_t)lbn) < round_page(ITOV(ip)->v_size)) {
    132 		struct vm_page *pg;
    133 		struct uvm_object *uobj = &ITOV(ip)->v_uobj;
    134 		voff_t off = trunc_page(lblktosize(fs, lbn));
    135 		voff_t endoff = round_page(lblktosize(fs, lbn) + size);
    136 
    137 		simple_lock(&uobj->vmobjlock);
    138 		while (off < endoff) {
    139 			pg = uvm_pagelookup(uobj, off);
    140 			KASSERT(pg != NULL);
    141 			KASSERT(pg->owner == curproc->p_pid);
    142 			off += PAGE_SIZE;
    143 		}
    144 		simple_unlock(&uobj->vmobjlock);
    145 	}
    146 #endif
    147 
    148 	*bnp = 0;
    149 #ifdef DIAGNOSTIC
    150 	if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
    151 		printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n",
    152 		    ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt);
    153 		panic("ffs_alloc: bad size");
    154 	}
    155 	if (cred == NOCRED)
    156 		panic("ffs_alloc: missing credential");
    157 #endif /* DIAGNOSTIC */
    158 	if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
    159 		goto nospace;
    160 	if (kauth_cred_geteuid(cred) != 0 && freespace(fs, fs->fs_minfree) <= 0)
    161 		goto nospace;
    162 #ifdef QUOTA
    163 	if ((error = chkdq(ip, btodb(size), cred, 0)) != 0)
    164 		return (error);
    165 #endif
    166 	if (bpref >= fs->fs_size)
    167 		bpref = 0;
    168 	if (bpref == 0)
    169 		cg = ino_to_cg(fs, ip->i_number);
    170 	else
    171 		cg = dtog(fs, bpref);
    172 	bno = ffs_hashalloc(ip, cg, bpref, size, ffs_alloccg);
    173 	if (bno > 0) {
    174 		DIP_ADD(ip, blocks, btodb(size));
    175 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
    176 		*bnp = bno;
    177 		return (0);
    178 	}
    179 #ifdef QUOTA
    180 	/*
    181 	 * Restore user's disk quota because allocation failed.
    182 	 */
    183 	(void) chkdq(ip, -btodb(size), cred, FORCE);
    184 #endif
    185 nospace:
    186 	ffs_fserr(fs, kauth_cred_geteuid(cred), "file system full");
    187 	uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
    188 	return (ENOSPC);
    189 }
    190 
    191 /*
    192  * Reallocate a fragment to a bigger size
    193  *
    194  * The number and size of the old block is given, and a preference
    195  * and new size is also specified. The allocator attempts to extend
    196  * the original block. Failing that, the regular block allocator is
    197  * invoked to get an appropriate block.
    198  */
    199 int
    200 ffs_realloccg(struct inode *ip, daddr_t lbprev, daddr_t bpref, int osize,
    201     int nsize, kauth_cred_t cred, struct buf **bpp, daddr_t *blknop)
    202 {
    203 	struct fs *fs;
    204 	struct buf *bp;
    205 	int cg, request, error;
    206 	daddr_t bprev, bno;
    207 
    208 	fs = ip->i_fs;
    209 #ifdef UVM_PAGE_TRKOWN
    210 	if (ITOV(ip)->v_type == VREG) {
    211 		struct vm_page *pg;
    212 		struct uvm_object *uobj = &ITOV(ip)->v_uobj;
    213 		voff_t off = trunc_page(lblktosize(fs, lbprev));
    214 		voff_t endoff = round_page(lblktosize(fs, lbprev) + osize);
    215 
    216 		simple_lock(&uobj->vmobjlock);
    217 		while (off < endoff) {
    218 			pg = uvm_pagelookup(uobj, off);
    219 			KASSERT(pg != NULL);
    220 			KASSERT(pg->owner == curproc->p_pid);
    221 			KASSERT((pg->flags & PG_CLEAN) == 0);
    222 			off += PAGE_SIZE;
    223 		}
    224 		simple_unlock(&uobj->vmobjlock);
    225 	}
    226 #endif
    227 
    228 #ifdef DIAGNOSTIC
    229 	if ((u_int)osize > fs->fs_bsize || fragoff(fs, osize) != 0 ||
    230 	    (u_int)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) {
    231 		printf(
    232 		    "dev = 0x%x, bsize = %d, osize = %d, nsize = %d, fs = %s\n",
    233 		    ip->i_dev, fs->fs_bsize, osize, nsize, fs->fs_fsmnt);
    234 		panic("ffs_realloccg: bad size");
    235 	}
    236 	if (cred == NOCRED)
    237 		panic("ffs_realloccg: missing credential");
    238 #endif /* DIAGNOSTIC */
    239 	if (kauth_cred_geteuid(cred) != 0 && freespace(fs, fs->fs_minfree) <= 0)
    240 		goto nospace;
    241 	if (fs->fs_magic == FS_UFS2_MAGIC)
    242 		bprev = ufs_rw64(ip->i_ffs2_db[lbprev], UFS_FSNEEDSWAP(fs));
    243 	else
    244 		bprev = ufs_rw32(ip->i_ffs1_db[lbprev], UFS_FSNEEDSWAP(fs));
    245 
    246 	if (bprev == 0) {
    247 		printf("dev = 0x%x, bsize = %d, bprev = %" PRId64 ", fs = %s\n",
    248 		    ip->i_dev, fs->fs_bsize, bprev, fs->fs_fsmnt);
    249 		panic("ffs_realloccg: bad bprev");
    250 	}
    251 	/*
    252 	 * Allocate the extra space in the buffer.
    253 	 */
    254 	if (bpp != NULL &&
    255 	    (error = bread(ITOV(ip), lbprev, osize, NOCRED, &bp)) != 0) {
    256 		brelse(bp);
    257 		return (error);
    258 	}
    259 #ifdef QUOTA
    260 	if ((error = chkdq(ip, btodb(nsize - osize), cred, 0)) != 0) {
    261 		if (bpp != NULL) {
    262 			brelse(bp);
    263 		}
    264 		return (error);
    265 	}
    266 #endif
    267 	/*
    268 	 * Check for extension in the existing location.
    269 	 */
    270 	cg = dtog(fs, bprev);
    271 	if ((bno = ffs_fragextend(ip, cg, bprev, osize, nsize)) != 0) {
    272 		DIP_ADD(ip, blocks, btodb(nsize - osize));
    273 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
    274 
    275 		if (bpp != NULL) {
    276 			if (bp->b_blkno != fsbtodb(fs, bno))
    277 				panic("bad blockno");
    278 			allocbuf(bp, nsize, 1);
    279 			bp->b_flags |= B_DONE;
    280 			memset(bp->b_data + osize, 0, nsize - osize);
    281 			*bpp = bp;
    282 		}
    283 		if (blknop != NULL) {
    284 			*blknop = bno;
    285 		}
    286 		return (0);
    287 	}
    288 	/*
    289 	 * Allocate a new disk location.
    290 	 */
    291 	if (bpref >= fs->fs_size)
    292 		bpref = 0;
    293 	switch ((int)fs->fs_optim) {
    294 	case FS_OPTSPACE:
    295 		/*
    296 		 * Allocate an exact sized fragment. Although this makes
    297 		 * best use of space, we will waste time relocating it if
    298 		 * the file continues to grow. If the fragmentation is
    299 		 * less than half of the minimum free reserve, we choose
    300 		 * to begin optimizing for time.
    301 		 */
    302 		request = nsize;
    303 		if (fs->fs_minfree < 5 ||
    304 		    fs->fs_cstotal.cs_nffree >
    305 		    fs->fs_dsize * fs->fs_minfree / (2 * 100))
    306 			break;
    307 
    308 		if (ffs_log_changeopt) {
    309 			log(LOG_NOTICE,
    310 				"%s: optimization changed from SPACE to TIME\n",
    311 				fs->fs_fsmnt);
    312 		}
    313 
    314 		fs->fs_optim = FS_OPTTIME;
    315 		break;
    316 	case FS_OPTTIME:
    317 		/*
    318 		 * At this point we have discovered a file that is trying to
    319 		 * grow a small fragment to a larger fragment. To save time,
    320 		 * we allocate a full sized block, then free the unused portion.
    321 		 * If the file continues to grow, the `ffs_fragextend' call
    322 		 * above will be able to grow it in place without further
    323 		 * copying. If aberrant programs cause disk fragmentation to
    324 		 * grow within 2% of the free reserve, we choose to begin
    325 		 * optimizing for space.
    326 		 */
    327 		request = fs->fs_bsize;
    328 		if (fs->fs_cstotal.cs_nffree <
    329 		    fs->fs_dsize * (fs->fs_minfree - 2) / 100)
    330 			break;
    331 
    332 		if (ffs_log_changeopt) {
    333 			log(LOG_NOTICE,
    334 				"%s: optimization changed from TIME to SPACE\n",
    335 				fs->fs_fsmnt);
    336 		}
    337 
    338 		fs->fs_optim = FS_OPTSPACE;
    339 		break;
    340 	default:
    341 		printf("dev = 0x%x, optim = %d, fs = %s\n",
    342 		    ip->i_dev, fs->fs_optim, fs->fs_fsmnt);
    343 		panic("ffs_realloccg: bad optim");
    344 		/* NOTREACHED */
    345 	}
    346 	bno = ffs_hashalloc(ip, cg, bpref, request, ffs_alloccg);
    347 	if (bno > 0) {
    348 		if (!DOINGSOFTDEP(ITOV(ip)))
    349 			ffs_blkfree(fs, ip->i_devvp, bprev, (long)osize,
    350 			    ip->i_number);
    351 		if (nsize < request)
    352 			ffs_blkfree(fs, ip->i_devvp, bno + numfrags(fs, nsize),
    353 			    (long)(request - nsize), ip->i_number);
    354 		DIP_ADD(ip, blocks, btodb(nsize - osize));
    355 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
    356 		if (bpp != NULL) {
    357 			bp->b_blkno = fsbtodb(fs, bno);
    358 			allocbuf(bp, nsize, 1);
    359 			bp->b_flags |= B_DONE;
    360 			memset(bp->b_data + osize, 0, (u_int)nsize - osize);
    361 			*bpp = bp;
    362 		}
    363 		if (blknop != NULL) {
    364 			*blknop = bno;
    365 		}
    366 		return (0);
    367 	}
    368 #ifdef QUOTA
    369 	/*
    370 	 * Restore user's disk quota because allocation failed.
    371 	 */
    372 	(void) chkdq(ip, -btodb(nsize - osize), cred, FORCE);
    373 #endif
    374 	if (bpp != NULL) {
    375 		brelse(bp);
    376 	}
    377 
    378 nospace:
    379 	/*
    380 	 * no space available
    381 	 */
    382 	ffs_fserr(fs, kauth_cred_geteuid(cred), "file system full");
    383 	uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
    384 	return (ENOSPC);
    385 }
    386 
    387 #if 0
    388 /*
    389  * Reallocate a sequence of blocks into a contiguous sequence of blocks.
    390  *
    391  * The vnode and an array of buffer pointers for a range of sequential
    392  * logical blocks to be made contiguous is given. The allocator attempts
    393  * to find a range of sequential blocks starting as close as possible
    394  * from the end of the allocation for the logical block immediately
    395  * preceding the current range. If successful, the physical block numbers
    396  * in the buffer pointers and in the inode are changed to reflect the new
    397  * allocation. If unsuccessful, the allocation is left unchanged. The
    398  * success in doing the reallocation is returned. Note that the error
    399  * return is not reflected back to the user. Rather the previous block
    400  * allocation will be used.
    401 
    402  */
    403 #ifdef XXXUBC
    404 #ifdef DEBUG
    405 #include <sys/sysctl.h>
    406 int prtrealloc = 0;
    407 struct ctldebug debug15 = { "prtrealloc", &prtrealloc };
    408 #endif
    409 #endif
    410 
    411 /*
    412  * NOTE: when re-enabling this, it must be updated for UFS2.
    413  */
    414 
    415 int doasyncfree = 1;
    416 
    417 int
    418 ffs_reallocblks(void *v)
    419 {
    420 #ifdef XXXUBC
    421 	struct vop_reallocblks_args /* {
    422 		struct vnode *a_vp;
    423 		struct cluster_save *a_buflist;
    424 	} */ *ap = v;
    425 	struct fs *fs;
    426 	struct inode *ip;
    427 	struct vnode *vp;
    428 	struct buf *sbp, *ebp;
    429 	int32_t *bap, *ebap = NULL, *sbap;	/* XXX ondisk32 */
    430 	struct cluster_save *buflist;
    431 	daddr_t start_lbn, end_lbn, soff, newblk, blkno;
    432 	struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp;
    433 	int i, len, start_lvl, end_lvl, pref, ssize;
    434 #endif /* XXXUBC */
    435 
    436 	/* XXXUBC don't reallocblks for now */
    437 	return ENOSPC;
    438 
    439 #ifdef XXXUBC
    440 	vp = ap->a_vp;
    441 	ip = VTOI(vp);
    442 	fs = ip->i_fs;
    443 	if (fs->fs_contigsumsize <= 0)
    444 		return (ENOSPC);
    445 	buflist = ap->a_buflist;
    446 	len = buflist->bs_nchildren;
    447 	start_lbn = buflist->bs_children[0]->b_lblkno;
    448 	end_lbn = start_lbn + len - 1;
    449 #ifdef DIAGNOSTIC
    450 	for (i = 0; i < len; i++)
    451 		if (!ffs_checkblk(ip,
    452 		   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
    453 			panic("ffs_reallocblks: unallocated block 1");
    454 	for (i = 1; i < len; i++)
    455 		if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
    456 			panic("ffs_reallocblks: non-logical cluster");
    457 	blkno = buflist->bs_children[0]->b_blkno;
    458 	ssize = fsbtodb(fs, fs->fs_frag);
    459 	for (i = 1; i < len - 1; i++)
    460 		if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize))
    461 			panic("ffs_reallocblks: non-physical cluster %d", i);
    462 #endif
    463 	/*
    464 	 * If the latest allocation is in a new cylinder group, assume that
    465 	 * the filesystem has decided to move and do not force it back to
    466 	 * the previous cylinder group.
    467 	 */
    468 	if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
    469 	    dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
    470 		return (ENOSPC);
    471 	if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
    472 	    ufs_getlbns(vp, end_lbn, end_ap, &end_lvl))
    473 		return (ENOSPC);
    474 	/*
    475 	 * Get the starting offset and block map for the first block.
    476 	 */
    477 	if (start_lvl == 0) {
    478 		sbap = &ip->i_ffs1_db[0];
    479 		soff = start_lbn;
    480 	} else {
    481 		idp = &start_ap[start_lvl - 1];
    482 		if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) {
    483 			brelse(sbp);
    484 			return (ENOSPC);
    485 		}
    486 		sbap = (int32_t *)sbp->b_data;
    487 		soff = idp->in_off;
    488 	}
    489 	/*
    490 	 * Find the preferred location for the cluster.
    491 	 */
    492 	pref = ffs_blkpref(ip, start_lbn, soff, sbap);
    493 	/*
    494 	 * If the block range spans two block maps, get the second map.
    495 	 */
    496 	if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
    497 		ssize = len;
    498 	} else {
    499 #ifdef DIAGNOSTIC
    500 		if (start_ap[start_lvl-1].in_lbn == idp->in_lbn)
    501 			panic("ffs_reallocblk: start == end");
    502 #endif
    503 		ssize = len - (idp->in_off + 1);
    504 		if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp))
    505 			goto fail;
    506 		ebap = (int32_t *)ebp->b_data;	/* XXX ondisk32 */
    507 	}
    508 	/*
    509 	 * Search the block map looking for an allocation of the desired size.
    510 	 */
    511 	if ((newblk = (daddr_t)ffs_hashalloc(ip, dtog(fs, pref), (long)pref,
    512 	    len, ffs_clusteralloc)) == 0)
    513 		goto fail;
    514 	/*
    515 	 * We have found a new contiguous block.
    516 	 *
    517 	 * First we have to replace the old block pointers with the new
    518 	 * block pointers in the inode and indirect blocks associated
    519 	 * with the file.
    520 	 */
    521 #ifdef DEBUG
    522 	if (prtrealloc)
    523 		printf("realloc: ino %d, lbns %d-%d\n\told:", ip->i_number,
    524 		    start_lbn, end_lbn);
    525 #endif
    526 	blkno = newblk;
    527 	for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) {
    528 		daddr_t ba;
    529 
    530 		if (i == ssize) {
    531 			bap = ebap;
    532 			soff = -i;
    533 		}
    534 		/* XXX ondisk32 */
    535 		ba = ufs_rw32(*bap, UFS_FSNEEDSWAP(fs));
    536 #ifdef DIAGNOSTIC
    537 		if (!ffs_checkblk(ip,
    538 		   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
    539 			panic("ffs_reallocblks: unallocated block 2");
    540 		if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != ba)
    541 			panic("ffs_reallocblks: alloc mismatch");
    542 #endif
    543 #ifdef DEBUG
    544 		if (prtrealloc)
    545 			printf(" %d,", ba);
    546 #endif
    547  		if (DOINGSOFTDEP(vp)) {
    548  			if (sbap == &ip->i_ffs1_db[0] && i < ssize)
    549  				softdep_setup_allocdirect(ip, start_lbn + i,
    550  				    blkno, ba, fs->fs_bsize, fs->fs_bsize,
    551  				    buflist->bs_children[i]);
    552  			else
    553  				softdep_setup_allocindir_page(ip, start_lbn + i,
    554  				    i < ssize ? sbp : ebp, soff + i, blkno,
    555  				    ba, buflist->bs_children[i]);
    556  		}
    557 		/* XXX ondisk32 */
    558 		*bap++ = ufs_rw32((u_int32_t)blkno, UFS_FSNEEDSWAP(fs));
    559 	}
    560 	/*
    561 	 * Next we must write out the modified inode and indirect blocks.
    562 	 * For strict correctness, the writes should be synchronous since
    563 	 * the old block values may have been written to disk. In practise
    564 	 * they are almost never written, but if we are concerned about
    565 	 * strict correctness, the `doasyncfree' flag should be set to zero.
    566 	 *
    567 	 * The test on `doasyncfree' should be changed to test a flag
    568 	 * that shows whether the associated buffers and inodes have
    569 	 * been written. The flag should be set when the cluster is
    570 	 * started and cleared whenever the buffer or inode is flushed.
    571 	 * We can then check below to see if it is set, and do the
    572 	 * synchronous write only when it has been cleared.
    573 	 */
    574 	if (sbap != &ip->i_ffs1_db[0]) {
    575 		if (doasyncfree)
    576 			bdwrite(sbp);
    577 		else
    578 			bwrite(sbp);
    579 	} else {
    580 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
    581 		if (!doasyncfree)
    582 			ffs_update(vp, NULL, NULL, 1);
    583 	}
    584 	if (ssize < len) {
    585 		if (doasyncfree)
    586 			bdwrite(ebp);
    587 		else
    588 			bwrite(ebp);
    589 	}
    590 	/*
    591 	 * Last, free the old blocks and assign the new blocks to the buffers.
    592 	 */
    593 #ifdef DEBUG
    594 	if (prtrealloc)
    595 		printf("\n\tnew:");
    596 #endif
    597 	for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) {
    598 		if (!DOINGSOFTDEP(vp))
    599 			ffs_blkfree(fs, ip->i_devvp,
    600 			    dbtofsb(fs, buflist->bs_children[i]->b_blkno),
    601 			    fs->fs_bsize, ip->i_number);
    602 		buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
    603 #ifdef DEBUG
    604 		if (!ffs_checkblk(ip,
    605 		   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
    606 			panic("ffs_reallocblks: unallocated block 3");
    607 		if (prtrealloc)
    608 			printf(" %d,", blkno);
    609 #endif
    610 	}
    611 #ifdef DEBUG
    612 	if (prtrealloc) {
    613 		prtrealloc--;
    614 		printf("\n");
    615 	}
    616 #endif
    617 	return (0);
    618 
    619 fail:
    620 	if (ssize < len)
    621 		brelse(ebp);
    622 	if (sbap != &ip->i_ffs1_db[0])
    623 		brelse(sbp);
    624 	return (ENOSPC);
    625 #endif /* XXXUBC */
    626 }
    627 #endif /* 0 */
    628 
    629 /*
    630  * Allocate an inode in the file system.
    631  *
    632  * If allocating a directory, use ffs_dirpref to select the inode.
    633  * If allocating in a directory, the following hierarchy is followed:
    634  *   1) allocate the preferred inode.
    635  *   2) allocate an inode in the same cylinder group.
    636  *   3) quadradically rehash into other cylinder groups, until an
    637  *      available inode is located.
    638  * If no inode preference is given the following hierarchy is used
    639  * to allocate an inode:
    640  *   1) allocate an inode in cylinder group 0.
    641  *   2) quadradically rehash into other cylinder groups, until an
    642  *      available inode is located.
    643  */
    644 int
    645 ffs_valloc(struct vnode *pvp, int mode, kauth_cred_t cred,
    646     struct vnode **vpp)
    647 {
    648 	struct inode *pip;
    649 	struct fs *fs;
    650 	struct inode *ip;
    651 	struct timespec ts;
    652 	ino_t ino, ipref;
    653 	int cg, error;
    654 
    655 	*vpp = NULL;
    656 	pip = VTOI(pvp);
    657 	fs = pip->i_fs;
    658 	if (fs->fs_cstotal.cs_nifree == 0)
    659 		goto noinodes;
    660 
    661 	if ((mode & IFMT) == IFDIR)
    662 		ipref = ffs_dirpref(pip);
    663 	else
    664 		ipref = pip->i_number;
    665 	if (ipref >= fs->fs_ncg * fs->fs_ipg)
    666 		ipref = 0;
    667 	cg = ino_to_cg(fs, ipref);
    668 	/*
    669 	 * Track number of dirs created one after another
    670 	 * in a same cg without intervening by files.
    671 	 */
    672 	if ((mode & IFMT) == IFDIR) {
    673 		if (fs->fs_contigdirs[cg] < 255)
    674 			fs->fs_contigdirs[cg]++;
    675 	} else {
    676 		if (fs->fs_contigdirs[cg] > 0)
    677 			fs->fs_contigdirs[cg]--;
    678 	}
    679 	ino = (ino_t)ffs_hashalloc(pip, cg, ipref, mode, ffs_nodealloccg);
    680 	if (ino == 0)
    681 		goto noinodes;
    682 	error = VFS_VGET(pvp->v_mount, ino, vpp);
    683 	if (error) {
    684 		ffs_vfree(pvp, ino, mode);
    685 		return (error);
    686 	}
    687 	KASSERT((*vpp)->v_type == VNON);
    688 	ip = VTOI(*vpp);
    689 	if (ip->i_mode) {
    690 #if 0
    691 		printf("mode = 0%o, inum = %d, fs = %s\n",
    692 		    ip->i_mode, ip->i_number, fs->fs_fsmnt);
    693 #else
    694 		printf("dmode %x mode %x dgen %x gen %x\n",
    695 		    DIP(ip, mode), ip->i_mode,
    696 		    DIP(ip, gen), ip->i_gen);
    697 		printf("size %llx blocks %llx\n",
    698 		    (long long)DIP(ip, size), (long long)DIP(ip, blocks));
    699 		printf("ino %llu ipref %llu\n", (unsigned long long)ino,
    700 		    (unsigned long long)ipref);
    701 #if 0
    702 		error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
    703 		    (int)fs->fs_bsize, NOCRED, &bp);
    704 #endif
    705 
    706 #endif
    707 		panic("ffs_valloc: dup alloc");
    708 	}
    709 	if (DIP(ip, blocks)) {				/* XXX */
    710 		printf("free inode %s/%llu had %" PRId64 " blocks\n",
    711 		    fs->fs_fsmnt, (unsigned long long)ino, DIP(ip, blocks));
    712 		DIP_ASSIGN(ip, blocks, 0);
    713 	}
    714 	ip->i_flag &= ~IN_SPACECOUNTED;
    715 	ip->i_flags = 0;
    716 	DIP_ASSIGN(ip, flags, 0);
    717 	/*
    718 	 * Set up a new generation number for this inode.
    719 	 */
    720 	ip->i_gen++;
    721 	DIP_ASSIGN(ip, gen, ip->i_gen);
    722 	if (fs->fs_magic == FS_UFS2_MAGIC) {
    723 		vfs_timestamp(&ts);
    724 		ip->i_ffs2_birthtime = ts.tv_sec;
    725 		ip->i_ffs2_birthnsec = ts.tv_nsec;
    726 	}
    727 	return (0);
    728 noinodes:
    729 	ffs_fserr(fs, kauth_cred_geteuid(cred), "out of inodes");
    730 	uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
    731 	return (ENOSPC);
    732 }
    733 
    734 /*
    735  * Find a cylinder group in which to place a directory.
    736  *
    737  * The policy implemented by this algorithm is to allocate a
    738  * directory inode in the same cylinder group as its parent
    739  * directory, but also to reserve space for its files inodes
    740  * and data. Restrict the number of directories which may be
    741  * allocated one after another in the same cylinder group
    742  * without intervening allocation of files.
    743  *
    744  * If we allocate a first level directory then force allocation
    745  * in another cylinder group.
    746  */
    747 static ino_t
    748 ffs_dirpref(struct inode *pip)
    749 {
    750 	register struct fs *fs;
    751 	int cg, prefcg;
    752 	int64_t dirsize, cgsize, curdsz;
    753 	int avgifree, avgbfree, avgndir;
    754 	int minifree, minbfree, maxndir;
    755 	int mincg, minndir;
    756 	int maxcontigdirs;
    757 
    758 	fs = pip->i_fs;
    759 
    760 	avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;
    761 	avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
    762 	avgndir = fs->fs_cstotal.cs_ndir / fs->fs_ncg;
    763 
    764 	/*
    765 	 * Force allocation in another cg if creating a first level dir.
    766 	 */
    767 	if (ITOV(pip)->v_flag & VROOT) {
    768 		prefcg = random() % fs->fs_ncg;
    769 		mincg = prefcg;
    770 		minndir = fs->fs_ipg;
    771 		for (cg = prefcg; cg < fs->fs_ncg; cg++)
    772 			if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
    773 			    fs->fs_cs(fs, cg).cs_nifree >= avgifree &&
    774 			    fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
    775 				mincg = cg;
    776 				minndir = fs->fs_cs(fs, cg).cs_ndir;
    777 			}
    778 		for (cg = 0; cg < prefcg; cg++)
    779 			if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
    780 			    fs->fs_cs(fs, cg).cs_nifree >= avgifree &&
    781 			    fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
    782 				mincg = cg;
    783 				minndir = fs->fs_cs(fs, cg).cs_ndir;
    784 			}
    785 		return ((ino_t)(fs->fs_ipg * mincg));
    786 	}
    787 
    788 	/*
    789 	 * Count various limits which used for
    790 	 * optimal allocation of a directory inode.
    791 	 */
    792 	maxndir = min(avgndir + fs->fs_ipg / 16, fs->fs_ipg);
    793 	minifree = avgifree - fs->fs_ipg / 4;
    794 	if (minifree < 0)
    795 		minifree = 0;
    796 	minbfree = avgbfree - fragstoblks(fs, fs->fs_fpg) / 4;
    797 	if (minbfree < 0)
    798 		minbfree = 0;
    799 	cgsize = (int64_t)fs->fs_fsize * fs->fs_fpg;
    800 	dirsize = (int64_t)fs->fs_avgfilesize * fs->fs_avgfpdir;
    801 	if (avgndir != 0) {
    802 		curdsz = (cgsize - (int64_t)avgbfree * fs->fs_bsize) / avgndir;
    803 		if (dirsize < curdsz)
    804 			dirsize = curdsz;
    805 	}
    806 	if (cgsize < dirsize * 255)
    807 		maxcontigdirs = cgsize / dirsize;
    808 	else
    809 		maxcontigdirs = 255;
    810 	if (fs->fs_avgfpdir > 0)
    811 		maxcontigdirs = min(maxcontigdirs,
    812 				    fs->fs_ipg / fs->fs_avgfpdir);
    813 	if (maxcontigdirs == 0)
    814 		maxcontigdirs = 1;
    815 
    816 	/*
    817 	 * Limit number of dirs in one cg and reserve space for
    818 	 * regular files, but only if we have no deficit in
    819 	 * inodes or space.
    820 	 */
    821 	prefcg = ino_to_cg(fs, pip->i_number);
    822 	for (cg = prefcg; cg < fs->fs_ncg; cg++)
    823 		if (fs->fs_cs(fs, cg).cs_ndir < maxndir &&
    824 		    fs->fs_cs(fs, cg).cs_nifree >= minifree &&
    825 	    	    fs->fs_cs(fs, cg).cs_nbfree >= minbfree) {
    826 			if (fs->fs_contigdirs[cg] < maxcontigdirs)
    827 				return ((ino_t)(fs->fs_ipg * cg));
    828 		}
    829 	for (cg = 0; cg < prefcg; cg++)
    830 		if (fs->fs_cs(fs, cg).cs_ndir < maxndir &&
    831 		    fs->fs_cs(fs, cg).cs_nifree >= minifree &&
    832 	    	    fs->fs_cs(fs, cg).cs_nbfree >= minbfree) {
    833 			if (fs->fs_contigdirs[cg] < maxcontigdirs)
    834 				return ((ino_t)(fs->fs_ipg * cg));
    835 		}
    836 	/*
    837 	 * This is a backstop when we are deficient in space.
    838 	 */
    839 	for (cg = prefcg; cg < fs->fs_ncg; cg++)
    840 		if (fs->fs_cs(fs, cg).cs_nifree >= avgifree)
    841 			return ((ino_t)(fs->fs_ipg * cg));
    842 	for (cg = 0; cg < prefcg; cg++)
    843 		if (fs->fs_cs(fs, cg).cs_nifree >= avgifree)
    844 			break;
    845 	return ((ino_t)(fs->fs_ipg * cg));
    846 }
    847 
    848 /*
    849  * Select the desired position for the next block in a file.  The file is
    850  * logically divided into sections. The first section is composed of the
    851  * direct blocks. Each additional section contains fs_maxbpg blocks.
    852  *
    853  * If no blocks have been allocated in the first section, the policy is to
    854  * request a block in the same cylinder group as the inode that describes
    855  * the file. If no blocks have been allocated in any other section, the
    856  * policy is to place the section in a cylinder group with a greater than
    857  * average number of free blocks.  An appropriate cylinder group is found
    858  * by using a rotor that sweeps the cylinder groups. When a new group of
    859  * blocks is needed, the sweep begins in the cylinder group following the
    860  * cylinder group from which the previous allocation was made. The sweep
    861  * continues until a cylinder group with greater than the average number
    862  * of free blocks is found. If the allocation is for the first block in an
    863  * indirect block, the information on the previous allocation is unavailable;
    864  * here a best guess is made based upon the logical block number being
    865  * allocated.
    866  *
    867  * If a section is already partially allocated, the policy is to
    868  * contiguously allocate fs_maxcontig blocks.  The end of one of these
    869  * contiguous blocks and the beginning of the next is laid out
    870  * contigously if possible.
    871  */
    872 daddr_t
    873 ffs_blkpref_ufs1(struct inode *ip, daddr_t lbn, int indx,
    874     int32_t *bap /* XXX ondisk32 */)
    875 {
    876 	struct fs *fs;
    877 	int cg;
    878 	int avgbfree, startcg;
    879 
    880 	fs = ip->i_fs;
    881 	if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
    882 		if (lbn < NDADDR + NINDIR(fs)) {
    883 			cg = ino_to_cg(fs, ip->i_number);
    884 			return (fs->fs_fpg * cg + fs->fs_frag);
    885 		}
    886 		/*
    887 		 * Find a cylinder with greater than average number of
    888 		 * unused data blocks.
    889 		 */
    890 		if (indx == 0 || bap[indx - 1] == 0)
    891 			startcg =
    892 			    ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
    893 		else
    894 			startcg = dtog(fs,
    895 				ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1);
    896 		startcg %= fs->fs_ncg;
    897 		avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
    898 		for (cg = startcg; cg < fs->fs_ncg; cg++)
    899 			if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
    900 				return (fs->fs_fpg * cg + fs->fs_frag);
    901 			}
    902 		for (cg = 0; cg < startcg; cg++)
    903 			if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
    904 				return (fs->fs_fpg * cg + fs->fs_frag);
    905 			}
    906 		return (0);
    907 	}
    908 	/*
    909 	 * We just always try to lay things out contiguously.
    910 	 */
    911 	return ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
    912 }
    913 
    914 daddr_t
    915 ffs_blkpref_ufs2(struct inode *ip, daddr_t lbn, int indx, int64_t *bap)
    916 {
    917 	struct fs *fs;
    918 	int cg;
    919 	int avgbfree, startcg;
    920 
    921 	fs = ip->i_fs;
    922 	if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
    923 		if (lbn < NDADDR + NINDIR(fs)) {
    924 			cg = ino_to_cg(fs, ip->i_number);
    925 			return (fs->fs_fpg * cg + fs->fs_frag);
    926 		}
    927 		/*
    928 		 * Find a cylinder with greater than average number of
    929 		 * unused data blocks.
    930 		 */
    931 		if (indx == 0 || bap[indx - 1] == 0)
    932 			startcg =
    933 			    ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
    934 		else
    935 			startcg = dtog(fs,
    936 				ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1);
    937 		startcg %= fs->fs_ncg;
    938 		avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
    939 		for (cg = startcg; cg < fs->fs_ncg; cg++)
    940 			if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
    941 				return (fs->fs_fpg * cg + fs->fs_frag);
    942 			}
    943 		for (cg = 0; cg < startcg; cg++)
    944 			if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
    945 				return (fs->fs_fpg * cg + fs->fs_frag);
    946 			}
    947 		return (0);
    948 	}
    949 	/*
    950 	 * We just always try to lay things out contiguously.
    951 	 */
    952 	return ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
    953 }
    954 
    955 
    956 /*
    957  * Implement the cylinder overflow algorithm.
    958  *
    959  * The policy implemented by this algorithm is:
    960  *   1) allocate the block in its requested cylinder group.
    961  *   2) quadradically rehash on the cylinder group number.
    962  *   3) brute force search for a free block.
    963  */
    964 /*VARARGS5*/
    965 static daddr_t
    966 ffs_hashalloc(struct inode *ip, int cg, daddr_t pref,
    967     int size /* size for data blocks, mode for inodes */,
    968     daddr_t (*allocator)(struct inode *, int, daddr_t, int))
    969 {
    970 	struct fs *fs;
    971 	daddr_t result;
    972 	int i, icg = cg;
    973 
    974 	fs = ip->i_fs;
    975 	/*
    976 	 * 1: preferred cylinder group
    977 	 */
    978 	result = (*allocator)(ip, cg, pref, size);
    979 	if (result)
    980 		return (result);
    981 	/*
    982 	 * 2: quadratic rehash
    983 	 */
    984 	for (i = 1; i < fs->fs_ncg; i *= 2) {
    985 		cg += i;
    986 		if (cg >= fs->fs_ncg)
    987 			cg -= fs->fs_ncg;
    988 		result = (*allocator)(ip, cg, 0, size);
    989 		if (result)
    990 			return (result);
    991 	}
    992 	/*
    993 	 * 3: brute force search
    994 	 * Note that we start at i == 2, since 0 was checked initially,
    995 	 * and 1 is always checked in the quadratic rehash.
    996 	 */
    997 	cg = (icg + 2) % fs->fs_ncg;
    998 	for (i = 2; i < fs->fs_ncg; i++) {
    999 		result = (*allocator)(ip, cg, 0, size);
   1000 		if (result)
   1001 			return (result);
   1002 		cg++;
   1003 		if (cg == fs->fs_ncg)
   1004 			cg = 0;
   1005 	}
   1006 	return (0);
   1007 }
   1008 
   1009 /*
   1010  * Determine whether a fragment can be extended.
   1011  *
   1012  * Check to see if the necessary fragments are available, and
   1013  * if they are, allocate them.
   1014  */
   1015 static daddr_t
   1016 ffs_fragextend(struct inode *ip, int cg, daddr_t bprev, int osize, int nsize)
   1017 {
   1018 	struct fs *fs;
   1019 	struct cg *cgp;
   1020 	struct buf *bp;
   1021 	daddr_t bno;
   1022 	int frags, bbase;
   1023 	int i, error;
   1024 	u_int8_t *blksfree;
   1025 
   1026 	fs = ip->i_fs;
   1027 	if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize))
   1028 		return (0);
   1029 	frags = numfrags(fs, nsize);
   1030 	bbase = fragnum(fs, bprev);
   1031 	if (bbase > fragnum(fs, (bprev + frags - 1))) {
   1032 		/* cannot extend across a block boundary */
   1033 		return (0);
   1034 	}
   1035 	error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
   1036 		(int)fs->fs_cgsize, NOCRED, &bp);
   1037 	if (error) {
   1038 		brelse(bp);
   1039 		return (0);
   1040 	}
   1041 	cgp = (struct cg *)bp->b_data;
   1042 	if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs))) {
   1043 		brelse(bp);
   1044 		return (0);
   1045 	}
   1046 	cgp->cg_old_time = ufs_rw32(time_second, UFS_FSNEEDSWAP(fs));
   1047 	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
   1048 	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
   1049 		cgp->cg_time = ufs_rw64(time_second, UFS_FSNEEDSWAP(fs));
   1050 	bno = dtogd(fs, bprev);
   1051 	blksfree = cg_blksfree(cgp, UFS_FSNEEDSWAP(fs));
   1052 	for (i = numfrags(fs, osize); i < frags; i++)
   1053 		if (isclr(blksfree, bno + i)) {
   1054 			brelse(bp);
   1055 			return (0);
   1056 		}
   1057 	/*
   1058 	 * the current fragment can be extended
   1059 	 * deduct the count on fragment being extended into
   1060 	 * increase the count on the remaining fragment (if any)
   1061 	 * allocate the extended piece
   1062 	 */
   1063 	for (i = frags; i < fs->fs_frag - bbase; i++)
   1064 		if (isclr(blksfree, bno + i))
   1065 			break;
   1066 	ufs_add32(cgp->cg_frsum[i - numfrags(fs, osize)], -1, UFS_FSNEEDSWAP(fs));
   1067 	if (i != frags)
   1068 		ufs_add32(cgp->cg_frsum[i - frags], 1, UFS_FSNEEDSWAP(fs));
   1069 	for (i = numfrags(fs, osize); i < frags; i++) {
   1070 		clrbit(blksfree, bno + i);
   1071 		ufs_add32(cgp->cg_cs.cs_nffree, -1, UFS_FSNEEDSWAP(fs));
   1072 		fs->fs_cstotal.cs_nffree--;
   1073 		fs->fs_cs(fs, cg).cs_nffree--;
   1074 	}
   1075 	fs->fs_fmod = 1;
   1076 	if (DOINGSOFTDEP(ITOV(ip)))
   1077 		softdep_setup_blkmapdep(bp, fs, bprev);
   1078 	ACTIVECG_CLR(fs, cg);
   1079 	bdwrite(bp);
   1080 	return (bprev);
   1081 }
   1082 
   1083 /*
   1084  * Determine whether a block can be allocated.
   1085  *
   1086  * Check to see if a block of the appropriate size is available,
   1087  * and if it is, allocate it.
   1088  */
   1089 static daddr_t
   1090 ffs_alloccg(struct inode *ip, int cg, daddr_t bpref, int size)
   1091 {
   1092 	struct fs *fs = ip->i_fs;
   1093 	struct cg *cgp;
   1094 	struct buf *bp;
   1095 	int32_t bno;
   1096 	daddr_t blkno;
   1097 	int error, frags, allocsiz, i;
   1098 	u_int8_t *blksfree;
   1099 #ifdef FFS_EI
   1100 	const int needswap = UFS_FSNEEDSWAP(fs);
   1101 #endif
   1102 
   1103 	if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
   1104 		return (0);
   1105 	error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
   1106 		(int)fs->fs_cgsize, NOCRED, &bp);
   1107 	if (error) {
   1108 		brelse(bp);
   1109 		return (0);
   1110 	}
   1111 	cgp = (struct cg *)bp->b_data;
   1112 	if (!cg_chkmagic(cgp, needswap) ||
   1113 	    (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) {
   1114 		brelse(bp);
   1115 		return (0);
   1116 	}
   1117 	cgp->cg_old_time = ufs_rw32(time_second, needswap);
   1118 	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
   1119 	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
   1120 		cgp->cg_time = ufs_rw64(time_second, needswap);
   1121 	if (size == fs->fs_bsize) {
   1122 		blkno = ffs_alloccgblk(ip, bp, bpref);
   1123 		ACTIVECG_CLR(fs, cg);
   1124 		bdwrite(bp);
   1125 		return (blkno);
   1126 	}
   1127 	/*
   1128 	 * check to see if any fragments are already available
   1129 	 * allocsiz is the size which will be allocated, hacking
   1130 	 * it down to a smaller size if necessary
   1131 	 */
   1132 	blksfree = cg_blksfree(cgp, needswap);
   1133 	frags = numfrags(fs, size);
   1134 	for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)
   1135 		if (cgp->cg_frsum[allocsiz] != 0)
   1136 			break;
   1137 	if (allocsiz == fs->fs_frag) {
   1138 		/*
   1139 		 * no fragments were available, so a block will be
   1140 		 * allocated, and hacked up
   1141 		 */
   1142 		if (cgp->cg_cs.cs_nbfree == 0) {
   1143 			brelse(bp);
   1144 			return (0);
   1145 		}
   1146 		blkno = ffs_alloccgblk(ip, bp, bpref);
   1147 		bno = dtogd(fs, blkno);
   1148 		for (i = frags; i < fs->fs_frag; i++)
   1149 			setbit(blksfree, bno + i);
   1150 		i = fs->fs_frag - frags;
   1151 		ufs_add32(cgp->cg_cs.cs_nffree, i, needswap);
   1152 		fs->fs_cstotal.cs_nffree += i;
   1153 		fs->fs_cs(fs, cg).cs_nffree += i;
   1154 		fs->fs_fmod = 1;
   1155 		ufs_add32(cgp->cg_frsum[i], 1, needswap);
   1156 		ACTIVECG_CLR(fs, cg);
   1157 		bdwrite(bp);
   1158 		return (blkno);
   1159 	}
   1160 	bno = ffs_mapsearch(fs, cgp, bpref, allocsiz);
   1161 #if 0
   1162 	/*
   1163 	 * XXX fvdl mapsearch will panic, and never return -1
   1164 	 *          also: returning NULL as daddr_t ?
   1165 	 */
   1166 	if (bno < 0) {
   1167 		brelse(bp);
   1168 		return (0);
   1169 	}
   1170 #endif
   1171 	for (i = 0; i < frags; i++)
   1172 		clrbit(blksfree, bno + i);
   1173 	ufs_add32(cgp->cg_cs.cs_nffree, -frags, needswap);
   1174 	fs->fs_cstotal.cs_nffree -= frags;
   1175 	fs->fs_cs(fs, cg).cs_nffree -= frags;
   1176 	fs->fs_fmod = 1;
   1177 	ufs_add32(cgp->cg_frsum[allocsiz], -1, needswap);
   1178 	if (frags != allocsiz)
   1179 		ufs_add32(cgp->cg_frsum[allocsiz - frags], 1, needswap);
   1180 	blkno = cg * fs->fs_fpg + bno;
   1181 	if (DOINGSOFTDEP(ITOV(ip)))
   1182 		softdep_setup_blkmapdep(bp, fs, blkno);
   1183 	ACTIVECG_CLR(fs, cg);
   1184 	bdwrite(bp);
   1185 	return blkno;
   1186 }
   1187 
   1188 /*
   1189  * Allocate a block in a cylinder group.
   1190  *
   1191  * This algorithm implements the following policy:
   1192  *   1) allocate the requested block.
   1193  *   2) allocate a rotationally optimal block in the same cylinder.
   1194  *   3) allocate the next available block on the block rotor for the
   1195  *      specified cylinder group.
   1196  * Note that this routine only allocates fs_bsize blocks; these
   1197  * blocks may be fragmented by the routine that allocates them.
   1198  */
   1199 static daddr_t
   1200 ffs_alloccgblk(struct inode *ip, struct buf *bp, daddr_t bpref)
   1201 {
   1202 	struct fs *fs = ip->i_fs;
   1203 	struct cg *cgp;
   1204 	daddr_t blkno;
   1205 	int32_t bno;
   1206 	u_int8_t *blksfree;
   1207 #ifdef FFS_EI
   1208 	const int needswap = UFS_FSNEEDSWAP(fs);
   1209 #endif
   1210 
   1211 	cgp = (struct cg *)bp->b_data;
   1212 	blksfree = cg_blksfree(cgp, needswap);
   1213 	if (bpref == 0 || dtog(fs, bpref) != ufs_rw32(cgp->cg_cgx, needswap)) {
   1214 		bpref = ufs_rw32(cgp->cg_rotor, needswap);
   1215 	} else {
   1216 		bpref = blknum(fs, bpref);
   1217 		bno = dtogd(fs, bpref);
   1218 		/*
   1219 		 * if the requested block is available, use it
   1220 		 */
   1221 		if (ffs_isblock(fs, blksfree, fragstoblks(fs, bno)))
   1222 			goto gotit;
   1223 	}
   1224 	/*
   1225 	 * Take the next available block in this cylinder group.
   1226 	 */
   1227 	bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
   1228 	if (bno < 0)
   1229 		return (0);
   1230 	cgp->cg_rotor = ufs_rw32(bno, needswap);
   1231 gotit:
   1232 	blkno = fragstoblks(fs, bno);
   1233 	ffs_clrblock(fs, blksfree, blkno);
   1234 	ffs_clusteracct(fs, cgp, blkno, -1);
   1235 	ufs_add32(cgp->cg_cs.cs_nbfree, -1, needswap);
   1236 	fs->fs_cstotal.cs_nbfree--;
   1237 	fs->fs_cs(fs, ufs_rw32(cgp->cg_cgx, needswap)).cs_nbfree--;
   1238 	if ((fs->fs_magic == FS_UFS1_MAGIC) &&
   1239 	    ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0)) {
   1240 		int cylno;
   1241 		cylno = old_cbtocylno(fs, bno);
   1242 		KASSERT(cylno >= 0);
   1243 		KASSERT(cylno < fs->fs_old_ncyl);
   1244 		KASSERT(old_cbtorpos(fs, bno) >= 0);
   1245 		KASSERT(fs->fs_old_nrpos == 0 || old_cbtorpos(fs, bno) < fs->fs_old_nrpos);
   1246 		ufs_add16(old_cg_blks(fs, cgp, cylno, needswap)[old_cbtorpos(fs, bno)], -1,
   1247 		    needswap);
   1248 		ufs_add32(old_cg_blktot(cgp, needswap)[cylno], -1, needswap);
   1249 	}
   1250 	fs->fs_fmod = 1;
   1251 	blkno = ufs_rw32(cgp->cg_cgx, needswap) * fs->fs_fpg + bno;
   1252 	if (DOINGSOFTDEP(ITOV(ip)))
   1253 		softdep_setup_blkmapdep(bp, fs, blkno);
   1254 	return (blkno);
   1255 }
   1256 
   1257 #ifdef XXXUBC
   1258 /*
   1259  * Determine whether a cluster can be allocated.
   1260  *
   1261  * We do not currently check for optimal rotational layout if there
   1262  * are multiple choices in the same cylinder group. Instead we just
   1263  * take the first one that we find following bpref.
   1264  */
   1265 
   1266 /*
   1267  * This function must be fixed for UFS2 if re-enabled.
   1268  */
   1269 static daddr_t
   1270 ffs_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len)
   1271 {
   1272 	struct fs *fs;
   1273 	struct cg *cgp;
   1274 	struct buf *bp;
   1275 	int i, got, run, bno, bit, map;
   1276 	u_char *mapp;
   1277 	int32_t *lp;
   1278 
   1279 	fs = ip->i_fs;
   1280 	if (fs->fs_maxcluster[cg] < len)
   1281 		return (0);
   1282 	if (bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
   1283 	    NOCRED, &bp))
   1284 		goto fail;
   1285 	cgp = (struct cg *)bp->b_data;
   1286 	if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs)))
   1287 		goto fail;
   1288 	/*
   1289 	 * Check to see if a cluster of the needed size (or bigger) is
   1290 	 * available in this cylinder group.
   1291 	 */
   1292 	lp = &cg_clustersum(cgp, UFS_FSNEEDSWAP(fs))[len];
   1293 	for (i = len; i <= fs->fs_contigsumsize; i++)
   1294 		if (ufs_rw32(*lp++, UFS_FSNEEDSWAP(fs)) > 0)
   1295 			break;
   1296 	if (i > fs->fs_contigsumsize) {
   1297 		/*
   1298 		 * This is the first time looking for a cluster in this
   1299 		 * cylinder group. Update the cluster summary information
   1300 		 * to reflect the true maximum sized cluster so that
   1301 		 * future cluster allocation requests can avoid reading
   1302 		 * the cylinder group map only to find no clusters.
   1303 		 */
   1304 		lp = &cg_clustersum(cgp, UFS_FSNEEDSWAP(fs))[len - 1];
   1305 		for (i = len - 1; i > 0; i--)
   1306 			if (ufs_rw32(*lp--, UFS_FSNEEDSWAP(fs)) > 0)
   1307 				break;
   1308 		fs->fs_maxcluster[cg] = i;
   1309 		goto fail;
   1310 	}
   1311 	/*
   1312 	 * Search the cluster map to find a big enough cluster.
   1313 	 * We take the first one that we find, even if it is larger
   1314 	 * than we need as we prefer to get one close to the previous
   1315 	 * block allocation. We do not search before the current
   1316 	 * preference point as we do not want to allocate a block
   1317 	 * that is allocated before the previous one (as we will
   1318 	 * then have to wait for another pass of the elevator
   1319 	 * algorithm before it will be read). We prefer to fail and
   1320 	 * be recalled to try an allocation in the next cylinder group.
   1321 	 */
   1322 	if (dtog(fs, bpref) != cg)
   1323 		bpref = 0;
   1324 	else
   1325 		bpref = fragstoblks(fs, dtogd(fs, blknum(fs, bpref)));
   1326 	mapp = &cg_clustersfree(cgp, UFS_FSNEEDSWAP(fs))[bpref / NBBY];
   1327 	map = *mapp++;
   1328 	bit = 1 << (bpref % NBBY);
   1329 	for (run = 0, got = bpref;
   1330 		got < ufs_rw32(cgp->cg_nclusterblks, UFS_FSNEEDSWAP(fs)); got++) {
   1331 		if ((map & bit) == 0) {
   1332 			run = 0;
   1333 		} else {
   1334 			run++;
   1335 			if (run == len)
   1336 				break;
   1337 		}
   1338 		if ((got & (NBBY - 1)) != (NBBY - 1)) {
   1339 			bit <<= 1;
   1340 		} else {
   1341 			map = *mapp++;
   1342 			bit = 1;
   1343 		}
   1344 	}
   1345 	if (got == ufs_rw32(cgp->cg_nclusterblks, UFS_FSNEEDSWAP(fs)))
   1346 		goto fail;
   1347 	/*
   1348 	 * Allocate the cluster that we have found.
   1349 	 */
   1350 #ifdef DIAGNOSTIC
   1351 	for (i = 1; i <= len; i++)
   1352 		if (!ffs_isblock(fs, cg_blksfree(cgp, UFS_FSNEEDSWAP(fs)),
   1353 		    got - run + i))
   1354 			panic("ffs_clusteralloc: map mismatch");
   1355 #endif
   1356 	bno = cg * fs->fs_fpg + blkstofrags(fs, got - run + 1);
   1357 	if (dtog(fs, bno) != cg)
   1358 		panic("ffs_clusteralloc: allocated out of group");
   1359 	len = blkstofrags(fs, len);
   1360 	for (i = 0; i < len; i += fs->fs_frag)
   1361 		if ((got = ffs_alloccgblk(ip, bp, bno + i)) != bno + i)
   1362 			panic("ffs_clusteralloc: lost block");
   1363 	ACTIVECG_CLR(fs, cg);
   1364 	bdwrite(bp);
   1365 	return (bno);
   1366 
   1367 fail:
   1368 	brelse(bp);
   1369 	return (0);
   1370 }
   1371 #endif /* XXXUBC */
   1372 
   1373 /*
   1374  * Determine whether an inode can be allocated.
   1375  *
   1376  * Check to see if an inode is available, and if it is,
   1377  * allocate it using the following policy:
   1378  *   1) allocate the requested inode.
   1379  *   2) allocate the next available inode after the requested
   1380  *      inode in the specified cylinder group.
   1381  */
   1382 static daddr_t
   1383 ffs_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode)
   1384 {
   1385 	struct fs *fs = ip->i_fs;
   1386 	struct cg *cgp;
   1387 	struct buf *bp, *ibp;
   1388 	u_int8_t *inosused;
   1389 	int error, start, len, loc, map, i;
   1390 	int32_t initediblk;
   1391 	struct ufs2_dinode *dp2;
   1392 #ifdef FFS_EI
   1393 	const int needswap = UFS_FSNEEDSWAP(fs);
   1394 #endif
   1395 
   1396 	if (fs->fs_cs(fs, cg).cs_nifree == 0)
   1397 		return (0);
   1398 	error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
   1399 		(int)fs->fs_cgsize, NOCRED, &bp);
   1400 	if (error) {
   1401 		brelse(bp);
   1402 		return (0);
   1403 	}
   1404 	cgp = (struct cg *)bp->b_data;
   1405 	if (!cg_chkmagic(cgp, needswap) || cgp->cg_cs.cs_nifree == 0) {
   1406 		brelse(bp);
   1407 		return (0);
   1408 	}
   1409 	cgp->cg_old_time = ufs_rw32(time_second, needswap);
   1410 	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
   1411 	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
   1412 		cgp->cg_time = ufs_rw64(time_second, needswap);
   1413 	inosused = cg_inosused(cgp, needswap);
   1414 	if (ipref) {
   1415 		ipref %= fs->fs_ipg;
   1416 		if (isclr(inosused, ipref))
   1417 			goto gotit;
   1418 	}
   1419 	start = ufs_rw32(cgp->cg_irotor, needswap) / NBBY;
   1420 	len = howmany(fs->fs_ipg - ufs_rw32(cgp->cg_irotor, needswap),
   1421 		NBBY);
   1422 	loc = skpc(0xff, len, &inosused[start]);
   1423 	if (loc == 0) {
   1424 		len = start + 1;
   1425 		start = 0;
   1426 		loc = skpc(0xff, len, &inosused[0]);
   1427 		if (loc == 0) {
   1428 			printf("cg = %d, irotor = %d, fs = %s\n",
   1429 			    cg, ufs_rw32(cgp->cg_irotor, needswap),
   1430 				fs->fs_fsmnt);
   1431 			panic("ffs_nodealloccg: map corrupted");
   1432 			/* NOTREACHED */
   1433 		}
   1434 	}
   1435 	i = start + len - loc;
   1436 	map = inosused[i];
   1437 	ipref = i * NBBY;
   1438 	for (i = 1; i < (1 << NBBY); i <<= 1, ipref++) {
   1439 		if ((map & i) == 0) {
   1440 			cgp->cg_irotor = ufs_rw32(ipref, needswap);
   1441 			goto gotit;
   1442 		}
   1443 	}
   1444 	printf("fs = %s\n", fs->fs_fsmnt);
   1445 	panic("ffs_nodealloccg: block not in map");
   1446 	/* NOTREACHED */
   1447 gotit:
   1448 	if (DOINGSOFTDEP(ITOV(ip)))
   1449 		softdep_setup_inomapdep(bp, ip, cg * fs->fs_ipg + ipref);
   1450 	setbit(inosused, ipref);
   1451 	ufs_add32(cgp->cg_cs.cs_nifree, -1, needswap);
   1452 	fs->fs_cstotal.cs_nifree--;
   1453 	fs->fs_cs(fs, cg).cs_nifree--;
   1454 	fs->fs_fmod = 1;
   1455 	if ((mode & IFMT) == IFDIR) {
   1456 		ufs_add32(cgp->cg_cs.cs_ndir, 1, needswap);
   1457 		fs->fs_cstotal.cs_ndir++;
   1458 		fs->fs_cs(fs, cg).cs_ndir++;
   1459 	}
   1460 	/*
   1461 	 * Check to see if we need to initialize more inodes.
   1462 	 */
   1463 	initediblk = ufs_rw32(cgp->cg_initediblk, needswap);
   1464 	if (fs->fs_magic == FS_UFS2_MAGIC &&
   1465 	    ipref + INOPB(fs) > initediblk &&
   1466 	    initediblk < ufs_rw32(cgp->cg_niblk, needswap)) {
   1467 		ibp = getblk(ip->i_devvp, fsbtodb(fs,
   1468 		    ino_to_fsba(fs, cg * fs->fs_ipg + initediblk)),
   1469 		    (int)fs->fs_bsize, 0, 0);
   1470 		    memset(ibp->b_data, 0, fs->fs_bsize);
   1471 		    dp2 = (struct ufs2_dinode *)(ibp->b_data);
   1472 		    for (i = 0; i < INOPB(fs); i++) {
   1473 			/*
   1474 			 * Don't bother to swap, it's supposed to be
   1475 			 * random, after all.
   1476 			 */
   1477 			dp2->di_gen = (arc4random() & INT32_MAX) / 2 + 1;
   1478 			dp2++;
   1479 		}
   1480 		bawrite(ibp);
   1481 		initediblk += INOPB(fs);
   1482 		cgp->cg_initediblk = ufs_rw32(initediblk, needswap);
   1483 	}
   1484 
   1485 	ACTIVECG_CLR(fs, cg);
   1486 	bdwrite(bp);
   1487 	return (cg * fs->fs_ipg + ipref);
   1488 }
   1489 
   1490 /*
   1491  * Free a block or fragment.
   1492  *
   1493  * The specified block or fragment is placed back in the
   1494  * free map. If a fragment is deallocated, a possible
   1495  * block reassembly is checked.
   1496  */
   1497 void
   1498 ffs_blkfree(struct fs *fs, struct vnode *devvp, daddr_t bno, long size,
   1499     ino_t inum)
   1500 {
   1501 	struct cg *cgp;
   1502 	struct buf *bp;
   1503 	struct ufsmount *ump;
   1504 	int32_t fragno, cgbno;
   1505 	daddr_t cgblkno;
   1506 	int i, error, cg, blk, frags, bbase;
   1507 	u_int8_t *blksfree;
   1508 	dev_t dev;
   1509 	const int needswap = UFS_FSNEEDSWAP(fs);
   1510 
   1511 	cg = dtog(fs, bno);
   1512 	if (devvp->v_type != VBLK) {
   1513 		/* devvp is a snapshot */
   1514 		dev = VTOI(devvp)->i_devvp->v_rdev;
   1515 		cgblkno = fragstoblks(fs, cgtod(fs, cg));
   1516 	} else {
   1517 		dev = devvp->v_rdev;
   1518 		ump = VFSTOUFS(devvp->v_specmountpoint);
   1519 		cgblkno = fsbtodb(fs, cgtod(fs, cg));
   1520 		if (TAILQ_FIRST(&ump->um_snapshots) != NULL &&
   1521 		    ffs_snapblkfree(fs, devvp, bno, size, inum))
   1522 			return;
   1523 	}
   1524 	if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0 ||
   1525 	    fragnum(fs, bno) + numfrags(fs, size) > fs->fs_frag) {
   1526 		printf("dev = 0x%x, bno = %" PRId64 " bsize = %d, "
   1527 		       "size = %ld, fs = %s\n",
   1528 		    dev, bno, fs->fs_bsize, size, fs->fs_fsmnt);
   1529 		panic("blkfree: bad size");
   1530 	}
   1531 
   1532 	if (bno >= fs->fs_size) {
   1533 		printf("bad block %" PRId64 ", ino %llu\n", bno,
   1534 		    (unsigned long long)inum);
   1535 		ffs_fserr(fs, inum, "bad block");
   1536 		return;
   1537 	}
   1538 	error = bread(devvp, cgblkno, (int)fs->fs_cgsize, NOCRED, &bp);
   1539 	if (error) {
   1540 		brelse(bp);
   1541 		return;
   1542 	}
   1543 	cgp = (struct cg *)bp->b_data;
   1544 	if (!cg_chkmagic(cgp, needswap)) {
   1545 		brelse(bp);
   1546 		return;
   1547 	}
   1548 	cgp->cg_old_time = ufs_rw32(time_second, needswap);
   1549 	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
   1550 	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
   1551 		cgp->cg_time = ufs_rw64(time_second, needswap);
   1552 	cgbno = dtogd(fs, bno);
   1553 	blksfree = cg_blksfree(cgp, needswap);
   1554 	if (size == fs->fs_bsize) {
   1555 		fragno = fragstoblks(fs, cgbno);
   1556 		if (!ffs_isfreeblock(fs, blksfree, fragno)) {
   1557 			if (devvp->v_type != VBLK) {
   1558 				/* devvp is a snapshot */
   1559 				brelse(bp);
   1560 				return;
   1561 			}
   1562 			printf("dev = 0x%x, block = %" PRId64 ", fs = %s\n",
   1563 			    dev, bno, fs->fs_fsmnt);
   1564 			panic("blkfree: freeing free block");
   1565 		}
   1566 		ffs_setblock(fs, blksfree, fragno);
   1567 		ffs_clusteracct(fs, cgp, fragno, 1);
   1568 		ufs_add32(cgp->cg_cs.cs_nbfree, 1, needswap);
   1569 		fs->fs_cstotal.cs_nbfree++;
   1570 		fs->fs_cs(fs, cg).cs_nbfree++;
   1571 		if ((fs->fs_magic == FS_UFS1_MAGIC) &&
   1572 		    ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0)) {
   1573 			i = old_cbtocylno(fs, cgbno);
   1574 			KASSERT(i >= 0);
   1575 			KASSERT(i < fs->fs_old_ncyl);
   1576 			KASSERT(old_cbtorpos(fs, cgbno) >= 0);
   1577 			KASSERT(fs->fs_old_nrpos == 0 || old_cbtorpos(fs, cgbno) < fs->fs_old_nrpos);
   1578 			ufs_add16(old_cg_blks(fs, cgp, i, needswap)[old_cbtorpos(fs, cgbno)], 1,
   1579 			    needswap);
   1580 			ufs_add32(old_cg_blktot(cgp, needswap)[i], 1, needswap);
   1581 		}
   1582 	} else {
   1583 		bbase = cgbno - fragnum(fs, cgbno);
   1584 		/*
   1585 		 * decrement the counts associated with the old frags
   1586 		 */
   1587 		blk = blkmap(fs, blksfree, bbase);
   1588 		ffs_fragacct(fs, blk, cgp->cg_frsum, -1, needswap);
   1589 		/*
   1590 		 * deallocate the fragment
   1591 		 */
   1592 		frags = numfrags(fs, size);
   1593 		for (i = 0; i < frags; i++) {
   1594 			if (isset(blksfree, cgbno + i)) {
   1595 				printf("dev = 0x%x, block = %" PRId64
   1596 				       ", fs = %s\n",
   1597 				    dev, bno + i, fs->fs_fsmnt);
   1598 				panic("blkfree: freeing free frag");
   1599 			}
   1600 			setbit(blksfree, cgbno + i);
   1601 		}
   1602 		ufs_add32(cgp->cg_cs.cs_nffree, i, needswap);
   1603 		fs->fs_cstotal.cs_nffree += i;
   1604 		fs->fs_cs(fs, cg).cs_nffree += i;
   1605 		/*
   1606 		 * add back in counts associated with the new frags
   1607 		 */
   1608 		blk = blkmap(fs, blksfree, bbase);
   1609 		ffs_fragacct(fs, blk, cgp->cg_frsum, 1, needswap);
   1610 		/*
   1611 		 * if a complete block has been reassembled, account for it
   1612 		 */
   1613 		fragno = fragstoblks(fs, bbase);
   1614 		if (ffs_isblock(fs, blksfree, fragno)) {
   1615 			ufs_add32(cgp->cg_cs.cs_nffree, -fs->fs_frag, needswap);
   1616 			fs->fs_cstotal.cs_nffree -= fs->fs_frag;
   1617 			fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
   1618 			ffs_clusteracct(fs, cgp, fragno, 1);
   1619 			ufs_add32(cgp->cg_cs.cs_nbfree, 1, needswap);
   1620 			fs->fs_cstotal.cs_nbfree++;
   1621 			fs->fs_cs(fs, cg).cs_nbfree++;
   1622 			if ((fs->fs_magic == FS_UFS1_MAGIC) &&
   1623 			    ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0)) {
   1624 				i = old_cbtocylno(fs, bbase);
   1625 				KASSERT(i >= 0);
   1626 				KASSERT(i < fs->fs_old_ncyl);
   1627 				KASSERT(old_cbtorpos(fs, bbase) >= 0);
   1628 				KASSERT(fs->fs_old_nrpos == 0 || old_cbtorpos(fs, bbase) < fs->fs_old_nrpos);
   1629 				ufs_add16(old_cg_blks(fs, cgp, i, needswap)[old_cbtorpos(fs,
   1630 				    bbase)], 1, needswap);
   1631 				ufs_add32(old_cg_blktot(cgp, needswap)[i], 1, needswap);
   1632 			}
   1633 		}
   1634 	}
   1635 	fs->fs_fmod = 1;
   1636 	ACTIVECG_CLR(fs, cg);
   1637 	bdwrite(bp);
   1638 }
   1639 
   1640 #if defined(DIAGNOSTIC) || defined(DEBUG)
   1641 #ifdef XXXUBC
   1642 /*
   1643  * Verify allocation of a block or fragment. Returns true if block or
   1644  * fragment is allocated, false if it is free.
   1645  */
   1646 static int
   1647 ffs_checkblk(struct inode *ip, daddr_t bno, long size)
   1648 {
   1649 	struct fs *fs;
   1650 	struct cg *cgp;
   1651 	struct buf *bp;
   1652 	int i, error, frags, free;
   1653 
   1654 	fs = ip->i_fs;
   1655 	if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
   1656 		printf("bsize = %d, size = %ld, fs = %s\n",
   1657 		    fs->fs_bsize, size, fs->fs_fsmnt);
   1658 		panic("checkblk: bad size");
   1659 	}
   1660 	if (bno >= fs->fs_size)
   1661 		panic("checkblk: bad block %d", bno);
   1662 	error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, dtog(fs, bno))),
   1663 		(int)fs->fs_cgsize, NOCRED, &bp);
   1664 	if (error) {
   1665 		brelse(bp);
   1666 		return 0;
   1667 	}
   1668 	cgp = (struct cg *)bp->b_data;
   1669 	if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs))) {
   1670 		brelse(bp);
   1671 		return 0;
   1672 	}
   1673 	bno = dtogd(fs, bno);
   1674 	if (size == fs->fs_bsize) {
   1675 		free = ffs_isblock(fs, cg_blksfree(cgp, UFS_FSNEEDSWAP(fs)),
   1676 			fragstoblks(fs, bno));
   1677 	} else {
   1678 		frags = numfrags(fs, size);
   1679 		for (free = 0, i = 0; i < frags; i++)
   1680 			if (isset(cg_blksfree(cgp, UFS_FSNEEDSWAP(fs)), bno + i))
   1681 				free++;
   1682 		if (free != 0 && free != frags)
   1683 			panic("checkblk: partially free fragment");
   1684 	}
   1685 	brelse(bp);
   1686 	return (!free);
   1687 }
   1688 #endif /* XXXUBC */
   1689 #endif /* DIAGNOSTIC */
   1690 
   1691 /*
   1692  * Free an inode.
   1693  */
   1694 int
   1695 ffs_vfree(struct vnode *vp, ino_t ino, int mode)
   1696 {
   1697 
   1698 	if (DOINGSOFTDEP(vp)) {
   1699 		softdep_freefile(vp, ino, mode);
   1700 		return (0);
   1701 	}
   1702 	return ffs_freefile(VTOI(vp)->i_fs, VTOI(vp)->i_devvp, ino, mode);
   1703 }
   1704 
   1705 /*
   1706  * Do the actual free operation.
   1707  * The specified inode is placed back in the free map.
   1708  */
   1709 int
   1710 ffs_freefile(struct fs *fs, struct vnode *devvp, ino_t ino, int mode)
   1711 {
   1712 	struct cg *cgp;
   1713 	struct buf *bp;
   1714 	int error, cg;
   1715 	daddr_t cgbno;
   1716 	u_int8_t *inosused;
   1717 	dev_t dev;
   1718 #ifdef FFS_EI
   1719 	const int needswap = UFS_FSNEEDSWAP(fs);
   1720 #endif
   1721 
   1722 	cg = ino_to_cg(fs, ino);
   1723 	if (devvp->v_type != VBLK) {
   1724 		/* devvp is a snapshot */
   1725 		dev = VTOI(devvp)->i_devvp->v_rdev;
   1726 		cgbno = fragstoblks(fs, cgtod(fs, cg));
   1727 	} else {
   1728 		dev = devvp->v_rdev;
   1729 		cgbno = fsbtodb(fs, cgtod(fs, cg));
   1730 	}
   1731 	if ((u_int)ino >= fs->fs_ipg * fs->fs_ncg)
   1732 		panic("ifree: range: dev = 0x%x, ino = %llu, fs = %s",
   1733 		    dev, (unsigned long long)ino, fs->fs_fsmnt);
   1734 	error = bread(devvp, cgbno, (int)fs->fs_cgsize, NOCRED, &bp);
   1735 	if (error) {
   1736 		brelse(bp);
   1737 		return (error);
   1738 	}
   1739 	cgp = (struct cg *)bp->b_data;
   1740 	if (!cg_chkmagic(cgp, needswap)) {
   1741 		brelse(bp);
   1742 		return (0);
   1743 	}
   1744 	cgp->cg_old_time = ufs_rw32(time_second, needswap);
   1745 	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
   1746 	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
   1747 		cgp->cg_time = ufs_rw64(time_second, needswap);
   1748 	inosused = cg_inosused(cgp, needswap);
   1749 	ino %= fs->fs_ipg;
   1750 	if (isclr(inosused, ino)) {
   1751 		printf("ifree: dev = 0x%x, ino = %llu, fs = %s\n",
   1752 		    dev, (unsigned long long)ino + cg * fs->fs_ipg,
   1753 		    fs->fs_fsmnt);
   1754 		if (fs->fs_ronly == 0)
   1755 			panic("ifree: freeing free inode");
   1756 	}
   1757 	clrbit(inosused, ino);
   1758 	if (ino < ufs_rw32(cgp->cg_irotor, needswap))
   1759 		cgp->cg_irotor = ufs_rw32(ino, needswap);
   1760 	ufs_add32(cgp->cg_cs.cs_nifree, 1, needswap);
   1761 	fs->fs_cstotal.cs_nifree++;
   1762 	fs->fs_cs(fs, cg).cs_nifree++;
   1763 	if ((mode & IFMT) == IFDIR) {
   1764 		ufs_add32(cgp->cg_cs.cs_ndir, -1, needswap);
   1765 		fs->fs_cstotal.cs_ndir--;
   1766 		fs->fs_cs(fs, cg).cs_ndir--;
   1767 	}
   1768 	fs->fs_fmod = 1;
   1769 	ACTIVECG_CLR(fs, cg);
   1770 	bdwrite(bp);
   1771 	return (0);
   1772 }
   1773 
   1774 /*
   1775  * Check to see if a file is free.
   1776  */
   1777 int
   1778 ffs_checkfreefile(struct fs *fs, struct vnode *devvp, ino_t ino)
   1779 {
   1780 	struct cg *cgp;
   1781 	struct buf *bp;
   1782 	daddr_t cgbno;
   1783 	int ret, cg;
   1784 	u_int8_t *inosused;
   1785 
   1786 	cg = ino_to_cg(fs, ino);
   1787 	if (devvp->v_type != VBLK) {
   1788 		/* devvp is a snapshot */
   1789 		cgbno = fragstoblks(fs, cgtod(fs, cg));
   1790 	} else
   1791 		cgbno = fsbtodb(fs, cgtod(fs, cg));
   1792 	if ((u_int)ino >= fs->fs_ipg * fs->fs_ncg)
   1793 		return 1;
   1794 	if (bread(devvp, cgbno, (int)fs->fs_cgsize, NOCRED, &bp)) {
   1795 		brelse(bp);
   1796 		return 1;
   1797 	}
   1798 	cgp = (struct cg *)bp->b_data;
   1799 	if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs))) {
   1800 		brelse(bp);
   1801 		return 1;
   1802 	}
   1803 	inosused = cg_inosused(cgp, UFS_FSNEEDSWAP(fs));
   1804 	ino %= fs->fs_ipg;
   1805 	ret = isclr(inosused, ino);
   1806 	brelse(bp);
   1807 	return ret;
   1808 }
   1809 
   1810 /*
   1811  * Find a block of the specified size in the specified cylinder group.
   1812  *
   1813  * It is a panic if a request is made to find a block if none are
   1814  * available.
   1815  */
   1816 static int32_t
   1817 ffs_mapsearch(struct fs *fs, struct cg *cgp, daddr_t bpref, int allocsiz)
   1818 {
   1819 	int32_t bno;
   1820 	int start, len, loc, i;
   1821 	int blk, field, subfield, pos;
   1822 	int ostart, olen;
   1823 	u_int8_t *blksfree;
   1824 #ifdef FFS_EI
   1825 	const int needswap = UFS_FSNEEDSWAP(fs);
   1826 #endif
   1827 
   1828 	/*
   1829 	 * find the fragment by searching through the free block
   1830 	 * map for an appropriate bit pattern
   1831 	 */
   1832 	if (bpref)
   1833 		start = dtogd(fs, bpref) / NBBY;
   1834 	else
   1835 		start = ufs_rw32(cgp->cg_frotor, needswap) / NBBY;
   1836 	blksfree = cg_blksfree(cgp, needswap);
   1837 	len = howmany(fs->fs_fpg, NBBY) - start;
   1838 	ostart = start;
   1839 	olen = len;
   1840 	loc = scanc((u_int)len,
   1841 		(const u_char *)&blksfree[start],
   1842 		(const u_char *)fragtbl[fs->fs_frag],
   1843 		(1 << (allocsiz - 1 + (fs->fs_frag & (NBBY - 1)))));
   1844 	if (loc == 0) {
   1845 		len = start + 1;
   1846 		start = 0;
   1847 		loc = scanc((u_int)len,
   1848 			(const u_char *)&blksfree[0],
   1849 			(const u_char *)fragtbl[fs->fs_frag],
   1850 			(1 << (allocsiz - 1 + (fs->fs_frag & (NBBY - 1)))));
   1851 		if (loc == 0) {
   1852 			printf("start = %d, len = %d, fs = %s\n",
   1853 			    ostart, olen, fs->fs_fsmnt);
   1854 			printf("offset=%d %ld\n",
   1855 				ufs_rw32(cgp->cg_freeoff, needswap),
   1856 				(long)blksfree - (long)cgp);
   1857 			printf("cg %d\n", cgp->cg_cgx);
   1858 			panic("ffs_alloccg: map corrupted");
   1859 			/* NOTREACHED */
   1860 		}
   1861 	}
   1862 	bno = (start + len - loc) * NBBY;
   1863 	cgp->cg_frotor = ufs_rw32(bno, needswap);
   1864 	/*
   1865 	 * found the byte in the map
   1866 	 * sift through the bits to find the selected frag
   1867 	 */
   1868 	for (i = bno + NBBY; bno < i; bno += fs->fs_frag) {
   1869 		blk = blkmap(fs, blksfree, bno);
   1870 		blk <<= 1;
   1871 		field = around[allocsiz];
   1872 		subfield = inside[allocsiz];
   1873 		for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) {
   1874 			if ((blk & field) == subfield)
   1875 				return (bno + pos);
   1876 			field <<= 1;
   1877 			subfield <<= 1;
   1878 		}
   1879 	}
   1880 	printf("bno = %d, fs = %s\n", bno, fs->fs_fsmnt);
   1881 	panic("ffs_alloccg: block not in map");
   1882 	/* return (-1); */
   1883 }
   1884 
   1885 /*
   1886  * Update the cluster map because of an allocation or free.
   1887  *
   1888  * Cnt == 1 means free; cnt == -1 means allocating.
   1889  */
   1890 void
   1891 ffs_clusteracct(struct fs *fs, struct cg *cgp, int32_t blkno, int cnt)
   1892 {
   1893 	int32_t *sump;
   1894 	int32_t *lp;
   1895 	u_char *freemapp, *mapp;
   1896 	int i, start, end, forw, back, map, bit;
   1897 #ifdef FFS_EI
   1898 	const int needswap = UFS_FSNEEDSWAP(fs);
   1899 #endif
   1900 
   1901 	if (fs->fs_contigsumsize <= 0)
   1902 		return;
   1903 	freemapp = cg_clustersfree(cgp, needswap);
   1904 	sump = cg_clustersum(cgp, needswap);
   1905 	/*
   1906 	 * Allocate or clear the actual block.
   1907 	 */
   1908 	if (cnt > 0)
   1909 		setbit(freemapp, blkno);
   1910 	else
   1911 		clrbit(freemapp, blkno);
   1912 	/*
   1913 	 * Find the size of the cluster going forward.
   1914 	 */
   1915 	start = blkno + 1;
   1916 	end = start + fs->fs_contigsumsize;
   1917 	if (end >= ufs_rw32(cgp->cg_nclusterblks, needswap))
   1918 		end = ufs_rw32(cgp->cg_nclusterblks, needswap);
   1919 	mapp = &freemapp[start / NBBY];
   1920 	map = *mapp++;
   1921 	bit = 1 << (start % NBBY);
   1922 	for (i = start; i < end; i++) {
   1923 		if ((map & bit) == 0)
   1924 			break;
   1925 		if ((i & (NBBY - 1)) != (NBBY - 1)) {
   1926 			bit <<= 1;
   1927 		} else {
   1928 			map = *mapp++;
   1929 			bit = 1;
   1930 		}
   1931 	}
   1932 	forw = i - start;
   1933 	/*
   1934 	 * Find the size of the cluster going backward.
   1935 	 */
   1936 	start = blkno - 1;
   1937 	end = start - fs->fs_contigsumsize;
   1938 	if (end < 0)
   1939 		end = -1;
   1940 	mapp = &freemapp[start / NBBY];
   1941 	map = *mapp--;
   1942 	bit = 1 << (start % NBBY);
   1943 	for (i = start; i > end; i--) {
   1944 		if ((map & bit) == 0)
   1945 			break;
   1946 		if ((i & (NBBY - 1)) != 0) {
   1947 			bit >>= 1;
   1948 		} else {
   1949 			map = *mapp--;
   1950 			bit = 1 << (NBBY - 1);
   1951 		}
   1952 	}
   1953 	back = start - i;
   1954 	/*
   1955 	 * Account for old cluster and the possibly new forward and
   1956 	 * back clusters.
   1957 	 */
   1958 	i = back + forw + 1;
   1959 	if (i > fs->fs_contigsumsize)
   1960 		i = fs->fs_contigsumsize;
   1961 	ufs_add32(sump[i], cnt, needswap);
   1962 	if (back > 0)
   1963 		ufs_add32(sump[back], -cnt, needswap);
   1964 	if (forw > 0)
   1965 		ufs_add32(sump[forw], -cnt, needswap);
   1966 
   1967 	/*
   1968 	 * Update cluster summary information.
   1969 	 */
   1970 	lp = &sump[fs->fs_contigsumsize];
   1971 	for (i = fs->fs_contigsumsize; i > 0; i--)
   1972 		if (ufs_rw32(*lp--, needswap) > 0)
   1973 			break;
   1974 	fs->fs_maxcluster[ufs_rw32(cgp->cg_cgx, needswap)] = i;
   1975 }
   1976 
   1977 /*
   1978  * Fserr prints the name of a file system with an error diagnostic.
   1979  *
   1980  * The form of the error message is:
   1981  *	fs: error message
   1982  */
   1983 static void
   1984 ffs_fserr(struct fs *fs, u_int uid, const char *cp)
   1985 {
   1986 
   1987 	log(LOG_ERR, "uid %d, pid %d, command %s, on %s: %s\n",
   1988 	    uid, curproc->p_pid, curproc->p_comm, fs->fs_fsmnt, cp);
   1989 }
   1990