[BACK]Return to ffs_alloc.c CVS log [TXT][DIR] Up to [cvs.NetBSD.org] / src / sys / ufs / ffs

Annotation of src/sys/ufs/ffs/ffs_alloc.c, Revision 1.68.2.5

1.68.2.5! skrll       1: /*     $NetBSD: ffs_alloc.c,v 1.78 2004/08/29 10:13:48 hannken Exp $   */
1.2       cgd         2:
1.1       mycroft     3: /*
1.60      fvdl        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:  *
1.1       mycroft    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.
1.68.2.2  skrll      24:  * 3. Neither the name of the University nor the names of its contributors
1.1       mycroft    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:  *
1.18      fvdl       40:  *     @(#)ffs_alloc.c 8.19 (Berkeley) 7/13/95
1.1       mycroft    41:  */
1.53      lukem      42:
                     43: #include <sys/cdefs.h>
1.68.2.5! skrll      44: __KERNEL_RCSID(0, "$NetBSD: ffs_alloc.c,v 1.78 2004/08/29 10:13:48 hannken Exp $");
1.17      mrg        45:
1.43      mrg        46: #if defined(_KERNEL_OPT)
1.27      thorpej    47: #include "opt_ffs.h"
1.21      scottr     48: #include "opt_quota.h"
1.22      scottr     49: #endif
1.1       mycroft    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>
1.29      mrg        59:
1.68.2.2  skrll      60: #include <miscfs/specfs/specdev.h>
1.1       mycroft    61: #include <ufs/ufs/quota.h>
1.19      bouyer     62: #include <ufs/ufs/ufsmount.h>
1.1       mycroft    63: #include <ufs/ufs/inode.h>
1.9       christos   64: #include <ufs/ufs/ufs_extern.h>
1.19      bouyer     65: #include <ufs/ufs/ufs_bswap.h>
1.1       mycroft    66:
                     67: #include <ufs/ffs/fs.h>
                     68: #include <ufs/ffs/ffs_extern.h>
                     69:
1.58      fvdl       70: static daddr_t ffs_alloccg __P((struct inode *, int, daddr_t, int));
                     71: static daddr_t ffs_alloccgblk __P((struct inode *, struct buf *, daddr_t));
1.55      matt       72: #ifdef XXXUBC
1.58      fvdl       73: static daddr_t ffs_clusteralloc __P((struct inode *, int, daddr_t, int));
1.55      matt       74: #endif
1.50      lukem      75: static ino_t ffs_dirpref __P((struct inode *));
1.58      fvdl       76: static daddr_t ffs_fragextend __P((struct inode *, int, daddr_t, int, int));
1.30      fvdl       77: static void ffs_fserr __P((struct fs *, u_int, char *));
1.58      fvdl       78: static daddr_t ffs_hashalloc __P((struct inode *, int, daddr_t, int,
                     79:     daddr_t (*)(struct inode *, int, daddr_t, int)));
                     80: static daddr_t ffs_nodealloccg __P((struct inode *, int, daddr_t, int));
1.60      fvdl       81: static int32_t ffs_mapsearch __P((struct fs *, struct cg *,
1.58      fvdl       82:                                      daddr_t, int));
1.18      fvdl       83: #if defined(DIAGNOSTIC) || defined(DEBUG)
1.55      matt       84: #ifdef XXXUBC
1.58      fvdl       85: static int ffs_checkblk __P((struct inode *, daddr_t, long size));
1.18      fvdl       86: #endif
1.55      matt       87: #endif
1.23      drochner   88:
1.34      jdolecek   89: /* if 1, changes in optimalization strategy are logged */
                     90: int ffs_log_changeopt = 0;
                     91:
1.23      drochner   92: /* in ffs_tables.c */
1.40      jdolecek   93: extern const int inside[], around[];
                     94: extern const u_char * const fragtbl[];
1.1       mycroft    95:
                     96: /*
                     97:  * Allocate a block in the file system.
                     98:  *
                     99:  * The size of the requested block is given, which must be some
                    100:  * multiple of fs_fsize and <= fs_bsize.
                    101:  * A preference may be optionally specified. If a preference is given
                    102:  * the following hierarchy is used to allocate a block:
                    103:  *   1) allocate the requested block.
                    104:  *   2) allocate a rotationally optimal block in the same cylinder.
                    105:  *   3) allocate a block in the same cylinder group.
                    106:  *   4) quadradically rehash into other cylinder groups, until an
                    107:  *      available block is located.
1.47      wiz       108:  * If no block preference is given the following hierarchy is used
1.1       mycroft   109:  * to allocate a block:
                    110:  *   1) allocate a block in the cylinder group that contains the
                    111:  *      inode for the file.
                    112:  *   2) quadradically rehash into other cylinder groups, until an
                    113:  *      available block is located.
                    114:  */
1.9       christos  115: int
1.1       mycroft   116: ffs_alloc(ip, lbn, bpref, size, cred, bnp)
1.33      augustss  117:        struct inode *ip;
1.58      fvdl      118:        daddr_t lbn, bpref;
1.1       mycroft   119:        int size;
                    120:        struct ucred *cred;
1.58      fvdl      121:        daddr_t *bnp;
1.1       mycroft   122: {
1.62      fvdl      123:        struct fs *fs;
1.58      fvdl      124:        daddr_t bno;
1.9       christos  125:        int cg;
                    126: #ifdef QUOTA
                    127:        int error;
                    128: #endif
1.1       mycroft   129:
1.62      fvdl      130:        fs = ip->i_fs;
                    131:
1.37      chs       132: #ifdef UVM_PAGE_TRKOWN
1.51      chs       133:        if (ITOV(ip)->v_type == VREG &&
                    134:            lblktosize(fs, (voff_t)lbn) < round_page(ITOV(ip)->v_size)) {
1.37      chs       135:                struct vm_page *pg;
1.51      chs       136:                struct uvm_object *uobj = &ITOV(ip)->v_uobj;
1.49      lukem     137:                voff_t off = trunc_page(lblktosize(fs, lbn));
                    138:                voff_t endoff = round_page(lblktosize(fs, lbn) + size);
1.37      chs       139:
                    140:                simple_lock(&uobj->vmobjlock);
                    141:                while (off < endoff) {
                    142:                        pg = uvm_pagelookup(uobj, off);
                    143:                        KASSERT(pg != NULL);
                    144:                        KASSERT(pg->owner == curproc->p_pid);
                    145:                        KASSERT((pg->flags & PG_CLEAN) == 0);
                    146:                        off += PAGE_SIZE;
                    147:                }
                    148:                simple_unlock(&uobj->vmobjlock);
                    149:        }
                    150: #endif
                    151:
1.1       mycroft   152:        *bnp = 0;
                    153: #ifdef DIAGNOSTIC
                    154:        if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
1.13      christos  155:                printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n",
1.1       mycroft   156:                    ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt);
                    157:                panic("ffs_alloc: bad size");
                    158:        }
                    159:        if (cred == NOCRED)
1.56      provos    160:                panic("ffs_alloc: missing credential");
1.1       mycroft   161: #endif /* DIAGNOSTIC */
                    162:        if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
                    163:                goto nospace;
                    164:        if (cred->cr_uid != 0 && freespace(fs, fs->fs_minfree) <= 0)
                    165:                goto nospace;
                    166: #ifdef QUOTA
1.60      fvdl      167:        if ((error = chkdq(ip, btodb(size), cred, 0)) != 0)
1.1       mycroft   168:                return (error);
                    169: #endif
                    170:        if (bpref >= fs->fs_size)
                    171:                bpref = 0;
                    172:        if (bpref == 0)
                    173:                cg = ino_to_cg(fs, ip->i_number);
                    174:        else
                    175:                cg = dtog(fs, bpref);
1.58      fvdl      176:        bno = ffs_hashalloc(ip, cg, (long)bpref, size,
1.9       christos  177:                                     ffs_alloccg);
1.1       mycroft   178:        if (bno > 0) {
1.65      kristerw  179:                DIP_ADD(ip, blocks, btodb(size));
1.1       mycroft   180:                ip->i_flag |= IN_CHANGE | IN_UPDATE;
                    181:                *bnp = bno;
                    182:                return (0);
                    183:        }
                    184: #ifdef QUOTA
                    185:        /*
                    186:         * Restore user's disk quota because allocation failed.
                    187:         */
1.60      fvdl      188:        (void) chkdq(ip, -btodb(size), cred, FORCE);
1.1       mycroft   189: #endif
                    190: nospace:
                    191:        ffs_fserr(fs, cred->cr_uid, "file system full");
                    192:        uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
                    193:        return (ENOSPC);
                    194: }
                    195:
                    196: /*
                    197:  * Reallocate a fragment to a bigger size
                    198:  *
                    199:  * The number and size of the old block is given, and a preference
                    200:  * and new size is also specified. The allocator attempts to extend
                    201:  * the original block. Failing that, the regular block allocator is
                    202:  * invoked to get an appropriate block.
                    203:  */
1.9       christos  204: int
1.37      chs       205: ffs_realloccg(ip, lbprev, bpref, osize, nsize, cred, bpp, blknop)
1.33      augustss  206:        struct inode *ip;
1.58      fvdl      207:        daddr_t lbprev;
                    208:        daddr_t bpref;
1.1       mycroft   209:        int osize, nsize;
                    210:        struct ucred *cred;
                    211:        struct buf **bpp;
1.58      fvdl      212:        daddr_t *blknop;
1.1       mycroft   213: {
1.62      fvdl      214:        struct fs *fs;
1.1       mycroft   215:        struct buf *bp;
                    216:        int cg, request, error;
1.58      fvdl      217:        daddr_t bprev, bno;
1.25      thorpej   218:
1.62      fvdl      219:        fs = ip->i_fs;
1.37      chs       220: #ifdef UVM_PAGE_TRKOWN
                    221:        if (ITOV(ip)->v_type == VREG) {
                    222:                struct vm_page *pg;
1.51      chs       223:                struct uvm_object *uobj = &ITOV(ip)->v_uobj;
1.49      lukem     224:                voff_t off = trunc_page(lblktosize(fs, lbprev));
                    225:                voff_t endoff = round_page(lblktosize(fs, lbprev) + osize);
1.37      chs       226:
                    227:                simple_lock(&uobj->vmobjlock);
                    228:                while (off < endoff) {
                    229:                        pg = uvm_pagelookup(uobj, off);
                    230:                        KASSERT(pg != NULL);
                    231:                        KASSERT(pg->owner == curproc->p_pid);
                    232:                        KASSERT((pg->flags & PG_CLEAN) == 0);
                    233:                        off += PAGE_SIZE;
                    234:                }
                    235:                simple_unlock(&uobj->vmobjlock);
                    236:        }
                    237: #endif
                    238:
1.1       mycroft   239: #ifdef DIAGNOSTIC
                    240:        if ((u_int)osize > fs->fs_bsize || fragoff(fs, osize) != 0 ||
                    241:            (u_int)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) {
1.13      christos  242:                printf(
1.1       mycroft   243:                    "dev = 0x%x, bsize = %d, osize = %d, nsize = %d, fs = %s\n",
                    244:                    ip->i_dev, fs->fs_bsize, osize, nsize, fs->fs_fsmnt);
                    245:                panic("ffs_realloccg: bad size");
                    246:        }
                    247:        if (cred == NOCRED)
1.56      provos    248:                panic("ffs_realloccg: missing credential");
1.1       mycroft   249: #endif /* DIAGNOSTIC */
                    250:        if (cred->cr_uid != 0 && freespace(fs, fs->fs_minfree) <= 0)
                    251:                goto nospace;
1.60      fvdl      252:        if (fs->fs_magic == FS_UFS2_MAGIC)
                    253:                bprev = ufs_rw64(ip->i_ffs2_db[lbprev], UFS_FSNEEDSWAP(fs));
                    254:        else
                    255:                bprev = ufs_rw32(ip->i_ffs1_db[lbprev], UFS_FSNEEDSWAP(fs));
                    256:
                    257:        if (bprev == 0) {
1.59      tsutsui   258:                printf("dev = 0x%x, bsize = %d, bprev = %" PRId64 ", fs = %s\n",
                    259:                    ip->i_dev, fs->fs_bsize, bprev, fs->fs_fsmnt);
1.1       mycroft   260:                panic("ffs_realloccg: bad bprev");
                    261:        }
                    262:        /*
                    263:         * Allocate the extra space in the buffer.
                    264:         */
1.37      chs       265:        if (bpp != NULL &&
                    266:            (error = bread(ITOV(ip), lbprev, osize, NOCRED, &bp)) != 0) {
1.1       mycroft   267:                brelse(bp);
                    268:                return (error);
                    269:        }
                    270: #ifdef QUOTA
1.60      fvdl      271:        if ((error = chkdq(ip, btodb(nsize - osize), cred, 0)) != 0) {
1.44      chs       272:                if (bpp != NULL) {
                    273:                        brelse(bp);
                    274:                }
1.1       mycroft   275:                return (error);
                    276:        }
                    277: #endif
                    278:        /*
                    279:         * Check for extension in the existing location.
                    280:         */
                    281:        cg = dtog(fs, bprev);
1.60      fvdl      282:        if ((bno = ffs_fragextend(ip, cg, bprev, osize, nsize)) != 0) {
1.65      kristerw  283:                DIP_ADD(ip, blocks, btodb(nsize - osize));
1.1       mycroft   284:                ip->i_flag |= IN_CHANGE | IN_UPDATE;
1.37      chs       285:
                    286:                if (bpp != NULL) {
                    287:                        if (bp->b_blkno != fsbtodb(fs, bno))
                    288:                                panic("bad blockno");
1.68.2.2  skrll     289:                        allocbuf(bp, nsize, 1);
1.37      chs       290:                        bp->b_flags |= B_DONE;
                    291:                        memset(bp->b_data + osize, 0, nsize - osize);
                    292:                        *bpp = bp;
                    293:                }
                    294:                if (blknop != NULL) {
                    295:                        *blknop = bno;
                    296:                }
1.1       mycroft   297:                return (0);
                    298:        }
                    299:        /*
                    300:         * Allocate a new disk location.
                    301:         */
                    302:        if (bpref >= fs->fs_size)
                    303:                bpref = 0;
                    304:        switch ((int)fs->fs_optim) {
                    305:        case FS_OPTSPACE:
                    306:                /*
                    307:                 * Allocate an exact sized fragment. Although this makes
                    308:                 * best use of space, we will waste time relocating it if
                    309:                 * the file continues to grow. If the fragmentation is
                    310:                 * less than half of the minimum free reserve, we choose
                    311:                 * to begin optimizing for time.
                    312:                 */
                    313:                request = nsize;
                    314:                if (fs->fs_minfree < 5 ||
                    315:                    fs->fs_cstotal.cs_nffree >
                    316:                    fs->fs_dsize * fs->fs_minfree / (2 * 100))
                    317:                        break;
1.34      jdolecek  318:
                    319:                if (ffs_log_changeopt) {
                    320:                        log(LOG_NOTICE,
                    321:                                "%s: optimization changed from SPACE to TIME\n",
                    322:                                fs->fs_fsmnt);
                    323:                }
                    324:
1.1       mycroft   325:                fs->fs_optim = FS_OPTTIME;
                    326:                break;
                    327:        case FS_OPTTIME:
                    328:                /*
                    329:                 * At this point we have discovered a file that is trying to
                    330:                 * grow a small fragment to a larger fragment. To save time,
                    331:                 * we allocate a full sized block, then free the unused portion.
                    332:                 * If the file continues to grow, the `ffs_fragextend' call
                    333:                 * above will be able to grow it in place without further
                    334:                 * copying. If aberrant programs cause disk fragmentation to
                    335:                 * grow within 2% of the free reserve, we choose to begin
                    336:                 * optimizing for space.
                    337:                 */
                    338:                request = fs->fs_bsize;
                    339:                if (fs->fs_cstotal.cs_nffree <
                    340:                    fs->fs_dsize * (fs->fs_minfree - 2) / 100)
                    341:                        break;
1.34      jdolecek  342:
                    343:                if (ffs_log_changeopt) {
                    344:                        log(LOG_NOTICE,
                    345:                                "%s: optimization changed from TIME to SPACE\n",
                    346:                                fs->fs_fsmnt);
                    347:                }
                    348:
1.1       mycroft   349:                fs->fs_optim = FS_OPTSPACE;
                    350:                break;
                    351:        default:
1.13      christos  352:                printf("dev = 0x%x, optim = %d, fs = %s\n",
1.1       mycroft   353:                    ip->i_dev, fs->fs_optim, fs->fs_fsmnt);
                    354:                panic("ffs_realloccg: bad optim");
                    355:                /* NOTREACHED */
                    356:        }
1.58      fvdl      357:        bno = ffs_hashalloc(ip, cg, bpref, request, ffs_alloccg);
1.1       mycroft   358:        if (bno > 0) {
1.30      fvdl      359:                if (!DOINGSOFTDEP(ITOV(ip)))
1.68.2.2  skrll     360:                        ffs_blkfree(fs, ip->i_devvp, bprev, (long)osize,
                    361:                            ip->i_number);
1.1       mycroft   362:                if (nsize < request)
1.68.2.2  skrll     363:                        ffs_blkfree(fs, ip->i_devvp, bno + numfrags(fs, nsize),
                    364:                            (long)(request - nsize), ip->i_number);
1.65      kristerw  365:                DIP_ADD(ip, blocks, btodb(nsize - osize));
1.1       mycroft   366:                ip->i_flag |= IN_CHANGE | IN_UPDATE;
1.37      chs       367:                if (bpp != NULL) {
                    368:                        bp->b_blkno = fsbtodb(fs, bno);
1.68.2.2  skrll     369:                        allocbuf(bp, nsize, 1);
1.37      chs       370:                        bp->b_flags |= B_DONE;
                    371:                        memset(bp->b_data + osize, 0, (u_int)nsize - osize);
                    372:                        *bpp = bp;
                    373:                }
                    374:                if (blknop != NULL) {
                    375:                        *blknop = bno;
                    376:                }
1.1       mycroft   377:                return (0);
                    378:        }
                    379: #ifdef QUOTA
                    380:        /*
                    381:         * Restore user's disk quota because allocation failed.
                    382:         */
1.60      fvdl      383:        (void) chkdq(ip, -btodb(nsize - osize), cred, FORCE);
1.1       mycroft   384: #endif
1.37      chs       385:        if (bpp != NULL) {
                    386:                brelse(bp);
                    387:        }
                    388:
1.1       mycroft   389: nospace:
                    390:        /*
                    391:         * no space available
                    392:         */
                    393:        ffs_fserr(fs, cred->cr_uid, "file system full");
                    394:        uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
                    395:        return (ENOSPC);
                    396: }
                    397:
                    398: /*
                    399:  * Reallocate a sequence of blocks into a contiguous sequence of blocks.
                    400:  *
                    401:  * The vnode and an array of buffer pointers for a range of sequential
                    402:  * logical blocks to be made contiguous is given. The allocator attempts
1.60      fvdl      403:  * to find a range of sequential blocks starting as close as possible
                    404:  * from the end of the allocation for the logical block immediately
                    405:  * preceding the current range. If successful, the physical block numbers
                    406:  * in the buffer pointers and in the inode are changed to reflect the new
                    407:  * allocation. If unsuccessful, the allocation is left unchanged. The
                    408:  * success in doing the reallocation is returned. Note that the error
                    409:  * return is not reflected back to the user. Rather the previous block
                    410:  * allocation will be used.
                    411:
1.1       mycroft   412:  */
1.55      matt      413: #ifdef XXXUBC
1.3       mycroft   414: #ifdef DEBUG
1.1       mycroft   415: #include <sys/sysctl.h>
1.5       mycroft   416: int prtrealloc = 0;
                    417: struct ctldebug debug15 = { "prtrealloc", &prtrealloc };
1.1       mycroft   418: #endif
1.55      matt      419: #endif
1.1       mycroft   420:
1.60      fvdl      421: /*
                    422:  * NOTE: when re-enabling this, it must be updated for UFS2.
                    423:  */
                    424:
1.18      fvdl      425: int doasyncfree = 1;
                    426:
1.1       mycroft   427: int
1.9       christos  428: ffs_reallocblks(v)
                    429:        void *v;
                    430: {
1.55      matt      431: #ifdef XXXUBC
1.1       mycroft   432:        struct vop_reallocblks_args /* {
                    433:                struct vnode *a_vp;
                    434:                struct cluster_save *a_buflist;
1.9       christos  435:        } */ *ap = v;
1.1       mycroft   436:        struct fs *fs;
                    437:        struct inode *ip;
                    438:        struct vnode *vp;
                    439:        struct buf *sbp, *ebp;
1.58      fvdl      440:        int32_t *bap, *ebap = NULL, *sbap;      /* XXX ondisk32 */
1.1       mycroft   441:        struct cluster_save *buflist;
1.58      fvdl      442:        daddr_t start_lbn, end_lbn, soff, newblk, blkno;
1.1       mycroft   443:        struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp;
                    444:        int i, len, start_lvl, end_lvl, pref, ssize;
1.55      matt      445: #endif /* XXXUBC */
1.1       mycroft   446:
1.37      chs       447:        /* XXXUBC don't reallocblks for now */
                    448:        return ENOSPC;
                    449:
1.55      matt      450: #ifdef XXXUBC
1.1       mycroft   451:        vp = ap->a_vp;
                    452:        ip = VTOI(vp);
                    453:        fs = ip->i_fs;
                    454:        if (fs->fs_contigsumsize <= 0)
                    455:                return (ENOSPC);
                    456:        buflist = ap->a_buflist;
                    457:        len = buflist->bs_nchildren;
                    458:        start_lbn = buflist->bs_children[0]->b_lblkno;
                    459:        end_lbn = start_lbn + len - 1;
                    460: #ifdef DIAGNOSTIC
1.18      fvdl      461:        for (i = 0; i < len; i++)
                    462:                if (!ffs_checkblk(ip,
                    463:                   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
                    464:                        panic("ffs_reallocblks: unallocated block 1");
1.1       mycroft   465:        for (i = 1; i < len; i++)
                    466:                if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
1.18      fvdl      467:                        panic("ffs_reallocblks: non-logical cluster");
                    468:        blkno = buflist->bs_children[0]->b_blkno;
                    469:        ssize = fsbtodb(fs, fs->fs_frag);
                    470:        for (i = 1; i < len - 1; i++)
                    471:                if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize))
                    472:                        panic("ffs_reallocblks: non-physical cluster %d", i);
1.1       mycroft   473: #endif
                    474:        /*
                    475:         * If the latest allocation is in a new cylinder group, assume that
                    476:         * the filesystem has decided to move and do not force it back to
                    477:         * the previous cylinder group.
                    478:         */
                    479:        if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
                    480:            dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
                    481:                return (ENOSPC);
                    482:        if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
                    483:            ufs_getlbns(vp, end_lbn, end_ap, &end_lvl))
                    484:                return (ENOSPC);
                    485:        /*
                    486:         * Get the starting offset and block map for the first block.
                    487:         */
                    488:        if (start_lvl == 0) {
1.60      fvdl      489:                sbap = &ip->i_ffs1_db[0];
1.1       mycroft   490:                soff = start_lbn;
                    491:        } else {
                    492:                idp = &start_ap[start_lvl - 1];
                    493:                if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) {
                    494:                        brelse(sbp);
                    495:                        return (ENOSPC);
                    496:                }
1.60      fvdl      497:                sbap = (int32_t *)sbp->b_data;
1.1       mycroft   498:                soff = idp->in_off;
                    499:        }
                    500:        /*
                    501:         * Find the preferred location for the cluster.
                    502:         */
                    503:        pref = ffs_blkpref(ip, start_lbn, soff, sbap);
                    504:        /*
                    505:         * If the block range spans two block maps, get the second map.
                    506:         */
                    507:        if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
                    508:                ssize = len;
                    509:        } else {
                    510: #ifdef DIAGNOSTIC
                    511:                if (start_ap[start_lvl-1].in_lbn == idp->in_lbn)
                    512:                        panic("ffs_reallocblk: start == end");
                    513: #endif
                    514:                ssize = len - (idp->in_off + 1);
                    515:                if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp))
                    516:                        goto fail;
1.58      fvdl      517:                ebap = (int32_t *)ebp->b_data;  /* XXX ondisk32 */
1.1       mycroft   518:        }
                    519:        /*
                    520:         * Search the block map looking for an allocation of the desired size.
                    521:         */
1.58      fvdl      522:        if ((newblk = (daddr_t)ffs_hashalloc(ip, dtog(fs, pref), (long)pref,
1.9       christos  523:            len, ffs_clusteralloc)) == 0)
1.1       mycroft   524:                goto fail;
                    525:        /*
                    526:         * We have found a new contiguous block.
                    527:         *
                    528:         * First we have to replace the old block pointers with the new
                    529:         * block pointers in the inode and indirect blocks associated
                    530:         * with the file.
                    531:         */
1.5       mycroft   532: #ifdef DEBUG
                    533:        if (prtrealloc)
1.13      christos  534:                printf("realloc: ino %d, lbns %d-%d\n\told:", ip->i_number,
1.5       mycroft   535:                    start_lbn, end_lbn);
                    536: #endif
1.1       mycroft   537:        blkno = newblk;
                    538:        for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) {
1.58      fvdl      539:                daddr_t ba;
1.30      fvdl      540:
                    541:                if (i == ssize) {
1.1       mycroft   542:                        bap = ebap;
1.30      fvdl      543:                        soff = -i;
                    544:                }
1.58      fvdl      545:                /* XXX ondisk32 */
1.30      fvdl      546:                ba = ufs_rw32(*bap, UFS_FSNEEDSWAP(fs));
1.1       mycroft   547: #ifdef DIAGNOSTIC
1.18      fvdl      548:                if (!ffs_checkblk(ip,
                    549:                   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
                    550:                        panic("ffs_reallocblks: unallocated block 2");
1.30      fvdl      551:                if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != ba)
1.1       mycroft   552:                        panic("ffs_reallocblks: alloc mismatch");
                    553: #endif
1.5       mycroft   554: #ifdef DEBUG
                    555:                if (prtrealloc)
1.30      fvdl      556:                        printf(" %d,", ba);
1.5       mycroft   557: #endif
1.30      fvdl      558:                if (DOINGSOFTDEP(vp)) {
1.60      fvdl      559:                        if (sbap == &ip->i_ffs1_db[0] && i < ssize)
1.30      fvdl      560:                                softdep_setup_allocdirect(ip, start_lbn + i,
                    561:                                    blkno, ba, fs->fs_bsize, fs->fs_bsize,
                    562:                                    buflist->bs_children[i]);
                    563:                        else
                    564:                                softdep_setup_allocindir_page(ip, start_lbn + i,
                    565:                                    i < ssize ? sbp : ebp, soff + i, blkno,
                    566:                                    ba, buflist->bs_children[i]);
                    567:                }
1.58      fvdl      568:                /* XXX ondisk32 */
                    569:                *bap++ = ufs_rw32((int32_t)blkno, UFS_FSNEEDSWAP(fs));
1.1       mycroft   570:        }
                    571:        /*
                    572:         * Next we must write out the modified inode and indirect blocks.
                    573:         * For strict correctness, the writes should be synchronous since
                    574:         * the old block values may have been written to disk. In practise
                    575:         * they are almost never written, but if we are concerned about
                    576:         * strict correctness, the `doasyncfree' flag should be set to zero.
                    577:         *
                    578:         * The test on `doasyncfree' should be changed to test a flag
                    579:         * that shows whether the associated buffers and inodes have
                    580:         * been written. The flag should be set when the cluster is
                    581:         * started and cleared whenever the buffer or inode is flushed.
                    582:         * We can then check below to see if it is set, and do the
                    583:         * synchronous write only when it has been cleared.
                    584:         */
1.60      fvdl      585:        if (sbap != &ip->i_ffs1_db[0]) {
1.1       mycroft   586:                if (doasyncfree)
                    587:                        bdwrite(sbp);
                    588:                else
                    589:                        bwrite(sbp);
                    590:        } else {
                    591:                ip->i_flag |= IN_CHANGE | IN_UPDATE;
1.28      mycroft   592:                if (!doasyncfree)
                    593:                        VOP_UPDATE(vp, NULL, NULL, 1);
1.1       mycroft   594:        }
1.25      thorpej   595:        if (ssize < len) {
1.1       mycroft   596:                if (doasyncfree)
                    597:                        bdwrite(ebp);
                    598:                else
                    599:                        bwrite(ebp);
1.25      thorpej   600:        }
1.1       mycroft   601:        /*
                    602:         * Last, free the old blocks and assign the new blocks to the buffers.
                    603:         */
1.5       mycroft   604: #ifdef DEBUG
                    605:        if (prtrealloc)
1.13      christos  606:                printf("\n\tnew:");
1.5       mycroft   607: #endif
1.1       mycroft   608:        for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) {
1.30      fvdl      609:                if (!DOINGSOFTDEP(vp))
1.68.2.2  skrll     610:                        ffs_blkfree(fs, ip->i_devvp,
1.30      fvdl      611:                            dbtofsb(fs, buflist->bs_children[i]->b_blkno),
1.68.2.2  skrll     612:                            fs->fs_bsize, ip->i_number);
1.1       mycroft   613:                buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
1.5       mycroft   614: #ifdef DEBUG
1.18      fvdl      615:                if (!ffs_checkblk(ip,
                    616:                   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
                    617:                        panic("ffs_reallocblks: unallocated block 3");
1.5       mycroft   618:                if (prtrealloc)
1.13      christos  619:                        printf(" %d,", blkno);
1.5       mycroft   620: #endif
                    621:        }
                    622: #ifdef DEBUG
                    623:        if (prtrealloc) {
                    624:                prtrealloc--;
1.13      christos  625:                printf("\n");
1.1       mycroft   626:        }
1.5       mycroft   627: #endif
1.1       mycroft   628:        return (0);
                    629:
                    630: fail:
                    631:        if (ssize < len)
                    632:                brelse(ebp);
1.60      fvdl      633:        if (sbap != &ip->i_ffs1_db[0])
1.1       mycroft   634:                brelse(sbp);
                    635:        return (ENOSPC);
1.55      matt      636: #endif /* XXXUBC */
1.1       mycroft   637: }
                    638:
                    639: /*
                    640:  * Allocate an inode in the file system.
                    641:  *
                    642:  * If allocating a directory, use ffs_dirpref to select the inode.
                    643:  * If allocating in a directory, the following hierarchy is followed:
                    644:  *   1) allocate the preferred inode.
                    645:  *   2) allocate an inode in the same cylinder group.
                    646:  *   3) quadradically rehash into other cylinder groups, until an
                    647:  *      available inode is located.
1.47      wiz       648:  * If no inode preference is given the following hierarchy is used
1.1       mycroft   649:  * to allocate an inode:
                    650:  *   1) allocate an inode in cylinder group 0.
                    651:  *   2) quadradically rehash into other cylinder groups, until an
                    652:  *      available inode is located.
                    653:  */
1.9       christos  654: int
                    655: ffs_valloc(v)
                    656:        void *v;
                    657: {
1.1       mycroft   658:        struct vop_valloc_args /* {
                    659:                struct vnode *a_pvp;
                    660:                int a_mode;
                    661:                struct ucred *a_cred;
                    662:                struct vnode **a_vpp;
1.9       christos  663:        } */ *ap = v;
1.33      augustss  664:        struct vnode *pvp = ap->a_pvp;
                    665:        struct inode *pip;
                    666:        struct fs *fs;
                    667:        struct inode *ip;
1.60      fvdl      668:        struct timespec ts;
1.1       mycroft   669:        mode_t mode = ap->a_mode;
                    670:        ino_t ino, ipref;
                    671:        int cg, error;
                    672:
                    673:        *ap->a_vpp = NULL;
                    674:        pip = VTOI(pvp);
                    675:        fs = pip->i_fs;
                    676:        if (fs->fs_cstotal.cs_nifree == 0)
                    677:                goto noinodes;
                    678:
                    679:        if ((mode & IFMT) == IFDIR)
1.50      lukem     680:                ipref = ffs_dirpref(pip);
                    681:        else
                    682:                ipref = pip->i_number;
1.1       mycroft   683:        if (ipref >= fs->fs_ncg * fs->fs_ipg)
                    684:                ipref = 0;
                    685:        cg = ino_to_cg(fs, ipref);
1.50      lukem     686:        /*
                    687:         * Track number of dirs created one after another
                    688:         * in a same cg without intervening by files.
                    689:         */
                    690:        if ((mode & IFMT) == IFDIR) {
1.63      fvdl      691:                if (fs->fs_contigdirs[cg] < 255)
1.50      lukem     692:                        fs->fs_contigdirs[cg]++;
                    693:        } else {
                    694:                if (fs->fs_contigdirs[cg] > 0)
                    695:                        fs->fs_contigdirs[cg]--;
                    696:        }
1.60      fvdl      697:        ino = (ino_t)ffs_hashalloc(pip, cg, ipref, mode, ffs_nodealloccg);
1.1       mycroft   698:        if (ino == 0)
                    699:                goto noinodes;
1.68.2.3  skrll     700:        error = VFS_VGET(pvp->v_mount, ino, ap->a_vpp);
1.1       mycroft   701:        if (error) {
                    702:                VOP_VFREE(pvp, ino, mode);
                    703:                return (error);
                    704:        }
                    705:        ip = VTOI(*ap->a_vpp);
1.60      fvdl      706:        if (ip->i_mode) {
                    707: #if 0
1.13      christos  708:                printf("mode = 0%o, inum = %d, fs = %s\n",
1.60      fvdl      709:                    ip->i_mode, ip->i_number, fs->fs_fsmnt);
                    710: #else
                    711:                printf("dmode %x mode %x dgen %x gen %x\n",
                    712:                    DIP(ip, mode), ip->i_mode,
                    713:                    DIP(ip, gen), ip->i_gen);
                    714:                printf("size %llx blocks %llx\n",
                    715:                    (long long)DIP(ip, size), (long long)DIP(ip, blocks));
                    716:                printf("ino %u ipref %u\n", ino, ipref);
                    717: #if 0
                    718:                error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
                    719:                    (int)fs->fs_bsize, NOCRED, &bp);
                    720: #endif
                    721:
                    722: #endif
1.1       mycroft   723:                panic("ffs_valloc: dup alloc");
                    724:        }
1.60      fvdl      725:        if (DIP(ip, blocks)) {                          /* XXX */
                    726:                printf("free inode %s/%d had %" PRId64 " blocks\n",
                    727:                    fs->fs_fsmnt, ino, DIP(ip, blocks));
1.65      kristerw  728:                DIP_ASSIGN(ip, blocks, 0);
1.1       mycroft   729:        }
1.57      hannken   730:        ip->i_flag &= ~IN_SPACECOUNTED;
1.61      fvdl      731:        ip->i_flags = 0;
1.65      kristerw  732:        DIP_ASSIGN(ip, flags, 0);
1.1       mycroft   733:        /*
                    734:         * Set up a new generation number for this inode.
                    735:         */
1.60      fvdl      736:        ip->i_gen++;
1.65      kristerw  737:        DIP_ASSIGN(ip, gen, ip->i_gen);
1.60      fvdl      738:        if (fs->fs_magic == FS_UFS2_MAGIC) {
                    739:                TIMEVAL_TO_TIMESPEC(&time, &ts);
                    740:                ip->i_ffs2_birthtime = ts.tv_sec;
                    741:                ip->i_ffs2_birthnsec = ts.tv_nsec;
                    742:        }
1.1       mycroft   743:        return (0);
                    744: noinodes:
                    745:        ffs_fserr(fs, ap->a_cred->cr_uid, "out of inodes");
                    746:        uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
                    747:        return (ENOSPC);
                    748: }
                    749:
                    750: /*
1.50      lukem     751:  * Find a cylinder group in which to place a directory.
1.42      sommerfe  752:  *
1.50      lukem     753:  * The policy implemented by this algorithm is to allocate a
                    754:  * directory inode in the same cylinder group as its parent
                    755:  * directory, but also to reserve space for its files inodes
                    756:  * and data. Restrict the number of directories which may be
                    757:  * allocated one after another in the same cylinder group
                    758:  * without intervening allocation of files.
1.42      sommerfe  759:  *
1.50      lukem     760:  * If we allocate a first level directory then force allocation
                    761:  * in another cylinder group.
1.1       mycroft   762:  */
                    763: static ino_t
1.50      lukem     764: ffs_dirpref(pip)
                    765:        struct inode *pip;
1.1       mycroft   766: {
1.50      lukem     767:        register struct fs *fs;
1.68.2.2  skrll     768:        int cg, prefcg;
                    769:        int64_t dirsize, cgsize;
1.50      lukem     770:        int avgifree, avgbfree, avgndir, curdirsize;
                    771:        int minifree, minbfree, maxndir;
                    772:        int mincg, minndir;
                    773:        int maxcontigdirs;
                    774:
                    775:        fs = pip->i_fs;
1.1       mycroft   776:
                    777:        avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;
1.50      lukem     778:        avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
                    779:        avgndir = fs->fs_cstotal.cs_ndir / fs->fs_ncg;
                    780:
                    781:        /*
                    782:         * Force allocation in another cg if creating a first level dir.
                    783:         */
                    784:        if (ITOV(pip)->v_flag & VROOT) {
                    785:                prefcg = random() % fs->fs_ncg;
                    786:                mincg = prefcg;
                    787:                minndir = fs->fs_ipg;
                    788:                for (cg = prefcg; cg < fs->fs_ncg; cg++)
                    789:                        if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
                    790:                            fs->fs_cs(fs, cg).cs_nifree >= avgifree &&
                    791:                            fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
1.42      sommerfe  792:                                mincg = cg;
1.50      lukem     793:                                minndir = fs->fs_cs(fs, cg).cs_ndir;
1.42      sommerfe  794:                        }
1.50      lukem     795:                for (cg = 0; cg < prefcg; cg++)
                    796:                        if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
                    797:                            fs->fs_cs(fs, cg).cs_nifree >= avgifree &&
                    798:                            fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
                    799:                                mincg = cg;
                    800:                                minndir = fs->fs_cs(fs, cg).cs_ndir;
1.42      sommerfe  801:                        }
1.50      lukem     802:                return ((ino_t)(fs->fs_ipg * mincg));
1.42      sommerfe  803:        }
1.50      lukem     804:
                    805:        /*
                    806:         * Count various limits which used for
                    807:         * optimal allocation of a directory inode.
                    808:         */
                    809:        maxndir = min(avgndir + fs->fs_ipg / 16, fs->fs_ipg);
                    810:        minifree = avgifree - fs->fs_ipg / 4;
                    811:        if (minifree < 0)
                    812:                minifree = 0;
1.54      mycroft   813:        minbfree = avgbfree - fragstoblks(fs, fs->fs_fpg) / 4;
1.50      lukem     814:        if (minbfree < 0)
                    815:                minbfree = 0;
                    816:        cgsize = fs->fs_fsize * fs->fs_fpg;
                    817:        dirsize = fs->fs_avgfilesize * fs->fs_avgfpdir;
                    818:        curdirsize = avgndir ? (cgsize - avgbfree * fs->fs_bsize) / avgndir : 0;
                    819:        if (dirsize < curdirsize)
                    820:                dirsize = curdirsize;
                    821:        maxcontigdirs = min(cgsize / dirsize, 255);
                    822:        if (fs->fs_avgfpdir > 0)
                    823:                maxcontigdirs = min(maxcontigdirs,
                    824:                                    fs->fs_ipg / fs->fs_avgfpdir);
                    825:        if (maxcontigdirs == 0)
                    826:                maxcontigdirs = 1;
                    827:
                    828:        /*
                    829:         * Limit number of dirs in one cg and reserve space for
                    830:         * regular files, but only if we have no deficit in
                    831:         * inodes or space.
                    832:         */
                    833:        prefcg = ino_to_cg(fs, pip->i_number);
                    834:        for (cg = prefcg; cg < fs->fs_ncg; cg++)
                    835:                if (fs->fs_cs(fs, cg).cs_ndir < maxndir &&
                    836:                    fs->fs_cs(fs, cg).cs_nifree >= minifree &&
                    837:                    fs->fs_cs(fs, cg).cs_nbfree >= minbfree) {
                    838:                        if (fs->fs_contigdirs[cg] < maxcontigdirs)
                    839:                                return ((ino_t)(fs->fs_ipg * cg));
                    840:                }
                    841:        for (cg = 0; cg < prefcg; cg++)
                    842:                if (fs->fs_cs(fs, cg).cs_ndir < maxndir &&
                    843:                    fs->fs_cs(fs, cg).cs_nifree >= minifree &&
                    844:                    fs->fs_cs(fs, cg).cs_nbfree >= minbfree) {
                    845:                        if (fs->fs_contigdirs[cg] < maxcontigdirs)
                    846:                                return ((ino_t)(fs->fs_ipg * cg));
                    847:                }
                    848:        /*
                    849:         * This is a backstop when we are deficient in space.
                    850:         */
                    851:        for (cg = prefcg; cg < fs->fs_ncg; cg++)
                    852:                if (fs->fs_cs(fs, cg).cs_nifree >= avgifree)
                    853:                        return ((ino_t)(fs->fs_ipg * cg));
                    854:        for (cg = 0; cg < prefcg; cg++)
                    855:                if (fs->fs_cs(fs, cg).cs_nifree >= avgifree)
                    856:                        break;
                    857:        return ((ino_t)(fs->fs_ipg * cg));
1.1       mycroft   858: }
                    859:
                    860: /*
                    861:  * Select the desired position for the next block in a file.  The file is
                    862:  * logically divided into sections. The first section is composed of the
                    863:  * direct blocks. Each additional section contains fs_maxbpg blocks.
                    864:  *
                    865:  * If no blocks have been allocated in the first section, the policy is to
                    866:  * request a block in the same cylinder group as the inode that describes
                    867:  * the file. If no blocks have been allocated in any other section, the
                    868:  * policy is to place the section in a cylinder group with a greater than
                    869:  * average number of free blocks.  An appropriate cylinder group is found
                    870:  * by using a rotor that sweeps the cylinder groups. When a new group of
                    871:  * blocks is needed, the sweep begins in the cylinder group following the
                    872:  * cylinder group from which the previous allocation was made. The sweep
                    873:  * continues until a cylinder group with greater than the average number
                    874:  * of free blocks is found. If the allocation is for the first block in an
                    875:  * indirect block, the information on the previous allocation is unavailable;
                    876:  * here a best guess is made based upon the logical block number being
                    877:  * allocated.
                    878:  *
                    879:  * If a section is already partially allocated, the policy is to
                    880:  * contiguously allocate fs_maxcontig blocks.  The end of one of these
1.60      fvdl      881:  * contiguous blocks and the beginning of the next is laid out
                    882:  * contigously if possible.
1.1       mycroft   883:  */
1.58      fvdl      884: daddr_t
1.60      fvdl      885: ffs_blkpref_ufs1(ip, lbn, indx, bap)
1.1       mycroft   886:        struct inode *ip;
1.58      fvdl      887:        daddr_t lbn;
1.1       mycroft   888:        int indx;
1.58      fvdl      889:        int32_t *bap;   /* XXX ondisk32 */
1.1       mycroft   890: {
1.33      augustss  891:        struct fs *fs;
                    892:        int cg;
1.1       mycroft   893:        int avgbfree, startcg;
                    894:
                    895:        fs = ip->i_fs;
                    896:        if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
1.31      fvdl      897:                if (lbn < NDADDR + NINDIR(fs)) {
1.1       mycroft   898:                        cg = ino_to_cg(fs, ip->i_number);
                    899:                        return (fs->fs_fpg * cg + fs->fs_frag);
                    900:                }
                    901:                /*
                    902:                 * Find a cylinder with greater than average number of
                    903:                 * unused data blocks.
                    904:                 */
                    905:                if (indx == 0 || bap[indx - 1] == 0)
                    906:                        startcg =
                    907:                            ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
                    908:                else
1.19      bouyer    909:                        startcg = dtog(fs,
1.30      fvdl      910:                                ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1);
1.1       mycroft   911:                startcg %= fs->fs_ncg;
                    912:                avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
                    913:                for (cg = startcg; cg < fs->fs_ncg; cg++)
                    914:                        if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
                    915:                                return (fs->fs_fpg * cg + fs->fs_frag);
                    916:                        }
1.52      lukem     917:                for (cg = 0; cg < startcg; cg++)
1.1       mycroft   918:                        if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
                    919:                                return (fs->fs_fpg * cg + fs->fs_frag);
                    920:                        }
1.35      thorpej   921:                return (0);
1.1       mycroft   922:        }
                    923:        /*
1.60      fvdl      924:         * We just always try to lay things out contiguously.
                    925:         */
                    926:        return ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
                    927: }
                    928:
                    929: daddr_t
                    930: ffs_blkpref_ufs2(ip, lbn, indx, bap)
                    931:        struct inode *ip;
                    932:        daddr_t lbn;
                    933:        int indx;
                    934:        int64_t *bap;
                    935: {
                    936:        struct fs *fs;
                    937:        int cg;
                    938:        int avgbfree, startcg;
                    939:
                    940:        fs = ip->i_fs;
                    941:        if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
                    942:                if (lbn < NDADDR + NINDIR(fs)) {
                    943:                        cg = ino_to_cg(fs, ip->i_number);
                    944:                        return (fs->fs_fpg * cg + fs->fs_frag);
                    945:                }
1.1       mycroft   946:                /*
1.60      fvdl      947:                 * Find a cylinder with greater than average number of
                    948:                 * unused data blocks.
1.1       mycroft   949:                 */
1.60      fvdl      950:                if (indx == 0 || bap[indx - 1] == 0)
                    951:                        startcg =
                    952:                            ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
                    953:                else
                    954:                        startcg = dtog(fs,
                    955:                                ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1);
                    956:                startcg %= fs->fs_ncg;
                    957:                avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
                    958:                for (cg = startcg; cg < fs->fs_ncg; cg++)
                    959:                        if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
                    960:                                return (fs->fs_fpg * cg + fs->fs_frag);
                    961:                        }
                    962:                for (cg = 0; cg < startcg; cg++)
                    963:                        if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
                    964:                                return (fs->fs_fpg * cg + fs->fs_frag);
                    965:                        }
                    966:                return (0);
                    967:        }
                    968:        /*
                    969:         * We just always try to lay things out contiguously.
                    970:         */
                    971:        return ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
1.1       mycroft   972: }
                    973:
1.60      fvdl      974:
1.1       mycroft   975: /*
                    976:  * Implement the cylinder overflow algorithm.
                    977:  *
                    978:  * The policy implemented by this algorithm is:
                    979:  *   1) allocate the block in its requested cylinder group.
                    980:  *   2) quadradically rehash on the cylinder group number.
                    981:  *   3) brute force search for a free block.
                    982:  */
                    983: /*VARARGS5*/
1.58      fvdl      984: static daddr_t
1.1       mycroft   985: ffs_hashalloc(ip, cg, pref, size, allocator)
                    986:        struct inode *ip;
                    987:        int cg;
1.58      fvdl      988:        daddr_t pref;
1.1       mycroft   989:        int size;       /* size for data blocks, mode for inodes */
1.58      fvdl      990:        daddr_t (*allocator) __P((struct inode *, int, daddr_t, int));
1.1       mycroft   991: {
1.33      augustss  992:        struct fs *fs;
1.58      fvdl      993:        daddr_t result;
1.1       mycroft   994:        int i, icg = cg;
                    995:
                    996:        fs = ip->i_fs;
                    997:        /*
                    998:         * 1: preferred cylinder group
                    999:         */
                   1000:        result = (*allocator)(ip, cg, pref, size);
                   1001:        if (result)
                   1002:                return (result);
                   1003:        /*
                   1004:         * 2: quadratic rehash
                   1005:         */
                   1006:        for (i = 1; i < fs->fs_ncg; i *= 2) {
                   1007:                cg += i;
                   1008:                if (cg >= fs->fs_ncg)
                   1009:                        cg -= fs->fs_ncg;
                   1010:                result = (*allocator)(ip, cg, 0, size);
                   1011:                if (result)
                   1012:                        return (result);
                   1013:        }
                   1014:        /*
                   1015:         * 3: brute force search
                   1016:         * Note that we start at i == 2, since 0 was checked initially,
                   1017:         * and 1 is always checked in the quadratic rehash.
                   1018:         */
                   1019:        cg = (icg + 2) % fs->fs_ncg;
                   1020:        for (i = 2; i < fs->fs_ncg; i++) {
                   1021:                result = (*allocator)(ip, cg, 0, size);
                   1022:                if (result)
                   1023:                        return (result);
                   1024:                cg++;
                   1025:                if (cg == fs->fs_ncg)
                   1026:                        cg = 0;
                   1027:        }
1.35      thorpej  1028:        return (0);
1.1       mycroft  1029: }
                   1030:
                   1031: /*
                   1032:  * Determine whether a fragment can be extended.
                   1033:  *
                   1034:  * Check to see if the necessary fragments are available, and
                   1035:  * if they are, allocate them.
                   1036:  */
1.58      fvdl     1037: static daddr_t
1.1       mycroft  1038: ffs_fragextend(ip, cg, bprev, osize, nsize)
                   1039:        struct inode *ip;
                   1040:        int cg;
1.58      fvdl     1041:        daddr_t bprev;
1.1       mycroft  1042:        int osize, nsize;
                   1043: {
1.33      augustss 1044:        struct fs *fs;
                   1045:        struct cg *cgp;
1.1       mycroft  1046:        struct buf *bp;
1.58      fvdl     1047:        daddr_t bno;
1.1       mycroft  1048:        int frags, bbase;
                   1049:        int i, error;
1.62      fvdl     1050:        u_int8_t *blksfree;
1.1       mycroft  1051:
                   1052:        fs = ip->i_fs;
                   1053:        if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize))
1.35      thorpej  1054:                return (0);
1.1       mycroft  1055:        frags = numfrags(fs, nsize);
                   1056:        bbase = fragnum(fs, bprev);
                   1057:        if (bbase > fragnum(fs, (bprev + frags - 1))) {
                   1058:                /* cannot extend across a block boundary */
1.35      thorpej  1059:                return (0);
1.1       mycroft  1060:        }
                   1061:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
                   1062:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1063:        if (error) {
                   1064:                brelse(bp);
1.35      thorpej  1065:                return (0);
1.1       mycroft  1066:        }
                   1067:        cgp = (struct cg *)bp->b_data;
1.30      fvdl     1068:        if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs))) {
1.1       mycroft  1069:                brelse(bp);
1.35      thorpej  1070:                return (0);
1.1       mycroft  1071:        }
1.62      fvdl     1072:        cgp->cg_old_time = ufs_rw32(time.tv_sec, UFS_FSNEEDSWAP(fs));
1.68.2.2  skrll    1073:        if ((fs->fs_magic != FS_UFS1_MAGIC) ||
                   1074:            (fs->fs_old_flags & FS_FLAGS_UPDATED))
                   1075:                cgp->cg_time = ufs_rw64(time.tv_sec, UFS_FSNEEDSWAP(fs));
1.1       mycroft  1076:        bno = dtogd(fs, bprev);
1.62      fvdl     1077:        blksfree = cg_blksfree(cgp, UFS_FSNEEDSWAP(fs));
1.1       mycroft  1078:        for (i = numfrags(fs, osize); i < frags; i++)
1.62      fvdl     1079:                if (isclr(blksfree, bno + i)) {
1.1       mycroft  1080:                        brelse(bp);
1.35      thorpej  1081:                        return (0);
1.1       mycroft  1082:                }
                   1083:        /*
                   1084:         * the current fragment can be extended
                   1085:         * deduct the count on fragment being extended into
                   1086:         * increase the count on the remaining fragment (if any)
                   1087:         * allocate the extended piece
                   1088:         */
                   1089:        for (i = frags; i < fs->fs_frag - bbase; i++)
1.62      fvdl     1090:                if (isclr(blksfree, bno + i))
1.1       mycroft  1091:                        break;
1.30      fvdl     1092:        ufs_add32(cgp->cg_frsum[i - numfrags(fs, osize)], -1, UFS_FSNEEDSWAP(fs));
1.1       mycroft  1093:        if (i != frags)
1.30      fvdl     1094:                ufs_add32(cgp->cg_frsum[i - frags], 1, UFS_FSNEEDSWAP(fs));
1.1       mycroft  1095:        for (i = numfrags(fs, osize); i < frags; i++) {
1.62      fvdl     1096:                clrbit(blksfree, bno + i);
1.30      fvdl     1097:                ufs_add32(cgp->cg_cs.cs_nffree, -1, UFS_FSNEEDSWAP(fs));
1.1       mycroft  1098:                fs->fs_cstotal.cs_nffree--;
                   1099:                fs->fs_cs(fs, cg).cs_nffree--;
                   1100:        }
                   1101:        fs->fs_fmod = 1;
1.30      fvdl     1102:        if (DOINGSOFTDEP(ITOV(ip)))
                   1103:                softdep_setup_blkmapdep(bp, fs, bprev);
1.68.2.2  skrll    1104:        ACTIVECG_CLR(fs, cg);
1.1       mycroft  1105:        bdwrite(bp);
                   1106:        return (bprev);
                   1107: }
                   1108:
                   1109: /*
                   1110:  * Determine whether a block can be allocated.
                   1111:  *
                   1112:  * Check to see if a block of the appropriate size is available,
                   1113:  * and if it is, allocate it.
                   1114:  */
1.58      fvdl     1115: static daddr_t
1.1       mycroft  1116: ffs_alloccg(ip, cg, bpref, size)
                   1117:        struct inode *ip;
                   1118:        int cg;
1.58      fvdl     1119:        daddr_t bpref;
1.1       mycroft  1120:        int size;
                   1121: {
1.62      fvdl     1122:        struct fs *fs = ip->i_fs;
1.30      fvdl     1123:        struct cg *cgp;
1.1       mycroft  1124:        struct buf *bp;
1.60      fvdl     1125:        int32_t bno;
                   1126:        daddr_t blkno;
1.30      fvdl     1127:        int error, frags, allocsiz, i;
1.62      fvdl     1128:        u_int8_t *blksfree;
1.30      fvdl     1129: #ifdef FFS_EI
                   1130:        const int needswap = UFS_FSNEEDSWAP(fs);
                   1131: #endif
1.1       mycroft  1132:
                   1133:        if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
1.35      thorpej  1134:                return (0);
1.1       mycroft  1135:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
                   1136:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1137:        if (error) {
                   1138:                brelse(bp);
1.35      thorpej  1139:                return (0);
1.1       mycroft  1140:        }
                   1141:        cgp = (struct cg *)bp->b_data;
1.19      bouyer   1142:        if (!cg_chkmagic(cgp, needswap) ||
1.1       mycroft  1143:            (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) {
                   1144:                brelse(bp);
1.35      thorpej  1145:                return (0);
1.1       mycroft  1146:        }
1.62      fvdl     1147:        cgp->cg_old_time = ufs_rw32(time.tv_sec, needswap);
1.68.2.2  skrll    1148:        if ((fs->fs_magic != FS_UFS1_MAGIC) ||
                   1149:            (fs->fs_old_flags & FS_FLAGS_UPDATED))
                   1150:                cgp->cg_time = ufs_rw64(time.tv_sec, needswap);
1.1       mycroft  1151:        if (size == fs->fs_bsize) {
1.60      fvdl     1152:                blkno = ffs_alloccgblk(ip, bp, bpref);
1.68.2.2  skrll    1153:                ACTIVECG_CLR(fs, cg);
1.1       mycroft  1154:                bdwrite(bp);
1.60      fvdl     1155:                return (blkno);
1.1       mycroft  1156:        }
                   1157:        /*
                   1158:         * check to see if any fragments are already available
                   1159:         * allocsiz is the size which will be allocated, hacking
                   1160:         * it down to a smaller size if necessary
                   1161:         */
1.62      fvdl     1162:        blksfree = cg_blksfree(cgp, needswap);
1.1       mycroft  1163:        frags = numfrags(fs, size);
                   1164:        for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)
                   1165:                if (cgp->cg_frsum[allocsiz] != 0)
                   1166:                        break;
                   1167:        if (allocsiz == fs->fs_frag) {
                   1168:                /*
                   1169:                 * no fragments were available, so a block will be
                   1170:                 * allocated, and hacked up
                   1171:                 */
                   1172:                if (cgp->cg_cs.cs_nbfree == 0) {
                   1173:                        brelse(bp);
1.35      thorpej  1174:                        return (0);
1.1       mycroft  1175:                }
1.60      fvdl     1176:                blkno = ffs_alloccgblk(ip, bp, bpref);
                   1177:                bno = dtogd(fs, blkno);
1.1       mycroft  1178:                for (i = frags; i < fs->fs_frag; i++)
1.62      fvdl     1179:                        setbit(blksfree, bno + i);
1.1       mycroft  1180:                i = fs->fs_frag - frags;
1.19      bouyer   1181:                ufs_add32(cgp->cg_cs.cs_nffree, i, needswap);
1.1       mycroft  1182:                fs->fs_cstotal.cs_nffree += i;
1.30      fvdl     1183:                fs->fs_cs(fs, cg).cs_nffree += i;
1.1       mycroft  1184:                fs->fs_fmod = 1;
1.19      bouyer   1185:                ufs_add32(cgp->cg_frsum[i], 1, needswap);
1.68.2.2  skrll    1186:                ACTIVECG_CLR(fs, cg);
1.1       mycroft  1187:                bdwrite(bp);
1.60      fvdl     1188:                return (blkno);
1.1       mycroft  1189:        }
1.30      fvdl     1190:        bno = ffs_mapsearch(fs, cgp, bpref, allocsiz);
                   1191: #if 0
                   1192:        /*
                   1193:         * XXX fvdl mapsearch will panic, and never return -1
1.58      fvdl     1194:         *          also: returning NULL as daddr_t ?
1.30      fvdl     1195:         */
1.1       mycroft  1196:        if (bno < 0) {
                   1197:                brelse(bp);
1.35      thorpej  1198:                return (0);
1.1       mycroft  1199:        }
1.30      fvdl     1200: #endif
1.1       mycroft  1201:        for (i = 0; i < frags; i++)
1.62      fvdl     1202:                clrbit(blksfree, bno + i);
1.19      bouyer   1203:        ufs_add32(cgp->cg_cs.cs_nffree, -frags, needswap);
1.1       mycroft  1204:        fs->fs_cstotal.cs_nffree -= frags;
                   1205:        fs->fs_cs(fs, cg).cs_nffree -= frags;
                   1206:        fs->fs_fmod = 1;
1.19      bouyer   1207:        ufs_add32(cgp->cg_frsum[allocsiz], -1, needswap);
1.1       mycroft  1208:        if (frags != allocsiz)
1.19      bouyer   1209:                ufs_add32(cgp->cg_frsum[allocsiz - frags], 1, needswap);
1.30      fvdl     1210:        blkno = cg * fs->fs_fpg + bno;
                   1211:        if (DOINGSOFTDEP(ITOV(ip)))
                   1212:                softdep_setup_blkmapdep(bp, fs, blkno);
1.68.2.2  skrll    1213:        ACTIVECG_CLR(fs, cg);
1.1       mycroft  1214:        bdwrite(bp);
1.30      fvdl     1215:        return blkno;
1.1       mycroft  1216: }
                   1217:
                   1218: /*
                   1219:  * Allocate a block in a cylinder group.
                   1220:  *
                   1221:  * This algorithm implements the following policy:
                   1222:  *   1) allocate the requested block.
                   1223:  *   2) allocate a rotationally optimal block in the same cylinder.
                   1224:  *   3) allocate the next available block on the block rotor for the
                   1225:  *      specified cylinder group.
                   1226:  * Note that this routine only allocates fs_bsize blocks; these
                   1227:  * blocks may be fragmented by the routine that allocates them.
                   1228:  */
1.58      fvdl     1229: static daddr_t
1.30      fvdl     1230: ffs_alloccgblk(ip, bp, bpref)
                   1231:        struct inode *ip;
                   1232:        struct buf *bp;
1.58      fvdl     1233:        daddr_t bpref;
1.1       mycroft  1234: {
1.62      fvdl     1235:        struct fs *fs = ip->i_fs;
1.30      fvdl     1236:        struct cg *cgp;
1.60      fvdl     1237:        daddr_t blkno;
                   1238:        int32_t bno;
                   1239:        u_int8_t *blksfree;
1.30      fvdl     1240: #ifdef FFS_EI
                   1241:        const int needswap = UFS_FSNEEDSWAP(fs);
                   1242: #endif
1.1       mycroft  1243:
1.30      fvdl     1244:        cgp = (struct cg *)bp->b_data;
1.60      fvdl     1245:        blksfree = cg_blksfree(cgp, needswap);
1.30      fvdl     1246:        if (bpref == 0 || dtog(fs, bpref) != ufs_rw32(cgp->cg_cgx, needswap)) {
1.19      bouyer   1247:                bpref = ufs_rw32(cgp->cg_rotor, needswap);
1.60      fvdl     1248:        } else {
                   1249:                bpref = blknum(fs, bpref);
                   1250:                bno = dtogd(fs, bpref);
1.1       mycroft  1251:                /*
1.60      fvdl     1252:                 * if the requested block is available, use it
1.1       mycroft  1253:                 */
1.60      fvdl     1254:                if (ffs_isblock(fs, blksfree, fragstoblks(fs, bno)))
                   1255:                        goto gotit;
1.1       mycroft  1256:        }
                   1257:        /*
1.60      fvdl     1258:         * Take the next available block in this cylinder group.
1.1       mycroft  1259:         */
1.30      fvdl     1260:        bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
1.1       mycroft  1261:        if (bno < 0)
1.35      thorpej  1262:                return (0);
1.60      fvdl     1263:        cgp->cg_rotor = ufs_rw32(bno, needswap);
1.1       mycroft  1264: gotit:
                   1265:        blkno = fragstoblks(fs, bno);
1.60      fvdl     1266:        ffs_clrblock(fs, blksfree, blkno);
1.30      fvdl     1267:        ffs_clusteracct(fs, cgp, blkno, -1);
1.19      bouyer   1268:        ufs_add32(cgp->cg_cs.cs_nbfree, -1, needswap);
1.1       mycroft  1269:        fs->fs_cstotal.cs_nbfree--;
1.19      bouyer   1270:        fs->fs_cs(fs, ufs_rw32(cgp->cg_cgx, needswap)).cs_nbfree--;
1.68.2.2  skrll    1271:        if ((fs->fs_magic == FS_UFS1_MAGIC) &&
                   1272:            ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0)) {
                   1273:                int cylno;
                   1274:                cylno = old_cbtocylno(fs, bno);
                   1275:                KASSERT(cylno >= 0);
                   1276:                KASSERT(cylno < fs->fs_old_ncyl);
                   1277:                KASSERT(old_cbtorpos(fs, bno) >= 0);
                   1278:                KASSERT(fs->fs_old_nrpos == 0 || old_cbtorpos(fs, bno) < fs->fs_old_nrpos);
                   1279:                ufs_add16(old_cg_blks(fs, cgp, cylno, needswap)[old_cbtorpos(fs, bno)], -1,
                   1280:                    needswap);
                   1281:                ufs_add32(old_cg_blktot(cgp, needswap)[cylno], -1, needswap);
                   1282:        }
1.1       mycroft  1283:        fs->fs_fmod = 1;
1.30      fvdl     1284:        blkno = ufs_rw32(cgp->cg_cgx, needswap) * fs->fs_fpg + bno;
                   1285:        if (DOINGSOFTDEP(ITOV(ip)))
                   1286:                softdep_setup_blkmapdep(bp, fs, blkno);
                   1287:        return (blkno);
1.1       mycroft  1288: }
                   1289:
1.55      matt     1290: #ifdef XXXUBC
1.1       mycroft  1291: /*
                   1292:  * Determine whether a cluster can be allocated.
                   1293:  *
                   1294:  * We do not currently check for optimal rotational layout if there
                   1295:  * are multiple choices in the same cylinder group. Instead we just
                   1296:  * take the first one that we find following bpref.
                   1297:  */
1.60      fvdl     1298:
                   1299: /*
                   1300:  * This function must be fixed for UFS2 if re-enabled.
                   1301:  */
1.58      fvdl     1302: static daddr_t
1.1       mycroft  1303: ffs_clusteralloc(ip, cg, bpref, len)
                   1304:        struct inode *ip;
                   1305:        int cg;
1.58      fvdl     1306:        daddr_t bpref;
1.1       mycroft  1307:        int len;
                   1308: {
1.33      augustss 1309:        struct fs *fs;
                   1310:        struct cg *cgp;
1.1       mycroft  1311:        struct buf *bp;
1.18      fvdl     1312:        int i, got, run, bno, bit, map;
1.1       mycroft  1313:        u_char *mapp;
1.5       mycroft  1314:        int32_t *lp;
1.1       mycroft  1315:
                   1316:        fs = ip->i_fs;
1.5       mycroft  1317:        if (fs->fs_maxcluster[cg] < len)
1.35      thorpej  1318:                return (0);
1.1       mycroft  1319:        if (bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
                   1320:            NOCRED, &bp))
                   1321:                goto fail;
                   1322:        cgp = (struct cg *)bp->b_data;
1.30      fvdl     1323:        if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs)))
1.1       mycroft  1324:                goto fail;
                   1325:        /*
                   1326:         * Check to see if a cluster of the needed size (or bigger) is
                   1327:         * available in this cylinder group.
                   1328:         */
1.30      fvdl     1329:        lp = &cg_clustersum(cgp, UFS_FSNEEDSWAP(fs))[len];
1.1       mycroft  1330:        for (i = len; i <= fs->fs_contigsumsize; i++)
1.30      fvdl     1331:                if (ufs_rw32(*lp++, UFS_FSNEEDSWAP(fs)) > 0)
1.1       mycroft  1332:                        break;
1.5       mycroft  1333:        if (i > fs->fs_contigsumsize) {
                   1334:                /*
                   1335:                 * This is the first time looking for a cluster in this
                   1336:                 * cylinder group. Update the cluster summary information
                   1337:                 * to reflect the true maximum sized cluster so that
                   1338:                 * future cluster allocation requests can avoid reading
                   1339:                 * the cylinder group map only to find no clusters.
                   1340:                 */
1.30      fvdl     1341:                lp = &cg_clustersum(cgp, UFS_FSNEEDSWAP(fs))[len - 1];
1.5       mycroft  1342:                for (i = len - 1; i > 0; i--)
1.30      fvdl     1343:                        if (ufs_rw32(*lp--, UFS_FSNEEDSWAP(fs)) > 0)
1.5       mycroft  1344:                                break;
                   1345:                fs->fs_maxcluster[cg] = i;
1.1       mycroft  1346:                goto fail;
1.5       mycroft  1347:        }
1.1       mycroft  1348:        /*
                   1349:         * Search the cluster map to find a big enough cluster.
                   1350:         * We take the first one that we find, even if it is larger
                   1351:         * than we need as we prefer to get one close to the previous
                   1352:         * block allocation. We do not search before the current
                   1353:         * preference point as we do not want to allocate a block
                   1354:         * that is allocated before the previous one (as we will
                   1355:         * then have to wait for another pass of the elevator
                   1356:         * algorithm before it will be read). We prefer to fail and
                   1357:         * be recalled to try an allocation in the next cylinder group.
                   1358:         */
                   1359:        if (dtog(fs, bpref) != cg)
                   1360:                bpref = 0;
                   1361:        else
                   1362:                bpref = fragstoblks(fs, dtogd(fs, blknum(fs, bpref)));
1.30      fvdl     1363:        mapp = &cg_clustersfree(cgp, UFS_FSNEEDSWAP(fs))[bpref / NBBY];
1.1       mycroft  1364:        map = *mapp++;
                   1365:        bit = 1 << (bpref % NBBY);
1.19      bouyer   1366:        for (run = 0, got = bpref;
1.30      fvdl     1367:                got < ufs_rw32(cgp->cg_nclusterblks, UFS_FSNEEDSWAP(fs)); got++) {
1.1       mycroft  1368:                if ((map & bit) == 0) {
                   1369:                        run = 0;
                   1370:                } else {
                   1371:                        run++;
                   1372:                        if (run == len)
                   1373:                                break;
                   1374:                }
1.18      fvdl     1375:                if ((got & (NBBY - 1)) != (NBBY - 1)) {
1.1       mycroft  1376:                        bit <<= 1;
                   1377:                } else {
                   1378:                        map = *mapp++;
                   1379:                        bit = 1;
                   1380:                }
                   1381:        }
1.30      fvdl     1382:        if (got == ufs_rw32(cgp->cg_nclusterblks, UFS_FSNEEDSWAP(fs)))
1.1       mycroft  1383:                goto fail;
                   1384:        /*
                   1385:         * Allocate the cluster that we have found.
                   1386:         */
1.30      fvdl     1387: #ifdef DIAGNOSTIC
1.18      fvdl     1388:        for (i = 1; i <= len; i++)
1.30      fvdl     1389:                if (!ffs_isblock(fs, cg_blksfree(cgp, UFS_FSNEEDSWAP(fs)),
                   1390:                    got - run + i))
1.18      fvdl     1391:                        panic("ffs_clusteralloc: map mismatch");
1.30      fvdl     1392: #endif
1.18      fvdl     1393:        bno = cg * fs->fs_fpg + blkstofrags(fs, got - run + 1);
                   1394:        if (dtog(fs, bno) != cg)
                   1395:                panic("ffs_clusteralloc: allocated out of group");
1.1       mycroft  1396:        len = blkstofrags(fs, len);
                   1397:        for (i = 0; i < len; i += fs->fs_frag)
1.30      fvdl     1398:                if ((got = ffs_alloccgblk(ip, bp, bno + i)) != bno + i)
1.1       mycroft  1399:                        panic("ffs_clusteralloc: lost block");
1.68.2.2  skrll    1400:        ACTIVECG_CLR(fs, cg);
1.8       cgd      1401:        bdwrite(bp);
1.1       mycroft  1402:        return (bno);
                   1403:
                   1404: fail:
                   1405:        brelse(bp);
                   1406:        return (0);
                   1407: }
1.55      matt     1408: #endif /* XXXUBC */
1.1       mycroft  1409:
                   1410: /*
                   1411:  * Determine whether an inode can be allocated.
                   1412:  *
                   1413:  * Check to see if an inode is available, and if it is,
                   1414:  * allocate it using the following policy:
                   1415:  *   1) allocate the requested inode.
                   1416:  *   2) allocate the next available inode after the requested
                   1417:  *      inode in the specified cylinder group.
                   1418:  */
1.58      fvdl     1419: static daddr_t
1.1       mycroft  1420: ffs_nodealloccg(ip, cg, ipref, mode)
                   1421:        struct inode *ip;
                   1422:        int cg;
1.58      fvdl     1423:        daddr_t ipref;
1.1       mycroft  1424:        int mode;
                   1425: {
1.62      fvdl     1426:        struct fs *fs = ip->i_fs;
1.33      augustss 1427:        struct cg *cgp;
1.60      fvdl     1428:        struct buf *bp, *ibp;
                   1429:        u_int8_t *inosused;
1.1       mycroft  1430:        int error, start, len, loc, map, i;
1.60      fvdl     1431:        int32_t initediblk;
                   1432:        struct ufs2_dinode *dp2;
1.19      bouyer   1433: #ifdef FFS_EI
1.30      fvdl     1434:        const int needswap = UFS_FSNEEDSWAP(fs);
1.19      bouyer   1435: #endif
1.1       mycroft  1436:
                   1437:        if (fs->fs_cs(fs, cg).cs_nifree == 0)
1.35      thorpej  1438:                return (0);
1.1       mycroft  1439:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
                   1440:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1441:        if (error) {
                   1442:                brelse(bp);
1.35      thorpej  1443:                return (0);
1.1       mycroft  1444:        }
                   1445:        cgp = (struct cg *)bp->b_data;
1.19      bouyer   1446:        if (!cg_chkmagic(cgp, needswap) || cgp->cg_cs.cs_nifree == 0) {
1.1       mycroft  1447:                brelse(bp);
1.35      thorpej  1448:                return (0);
1.1       mycroft  1449:        }
1.62      fvdl     1450:        cgp->cg_old_time = ufs_rw32(time.tv_sec, needswap);
1.68.2.2  skrll    1451:        if ((fs->fs_magic != FS_UFS1_MAGIC) ||
                   1452:            (fs->fs_old_flags & FS_FLAGS_UPDATED))
                   1453:                cgp->cg_time = ufs_rw64(time.tv_sec, needswap);
1.60      fvdl     1454:        inosused = cg_inosused(cgp, needswap);
1.1       mycroft  1455:        if (ipref) {
                   1456:                ipref %= fs->fs_ipg;
1.60      fvdl     1457:                if (isclr(inosused, ipref))
1.1       mycroft  1458:                        goto gotit;
                   1459:        }
1.19      bouyer   1460:        start = ufs_rw32(cgp->cg_irotor, needswap) / NBBY;
                   1461:        len = howmany(fs->fs_ipg - ufs_rw32(cgp->cg_irotor, needswap),
                   1462:                NBBY);
1.60      fvdl     1463:        loc = skpc(0xff, len, &inosused[start]);
1.1       mycroft  1464:        if (loc == 0) {
                   1465:                len = start + 1;
                   1466:                start = 0;
1.60      fvdl     1467:                loc = skpc(0xff, len, &inosused[0]);
1.1       mycroft  1468:                if (loc == 0) {
1.13      christos 1469:                        printf("cg = %d, irotor = %d, fs = %s\n",
1.19      bouyer   1470:                            cg, ufs_rw32(cgp->cg_irotor, needswap),
                   1471:                                fs->fs_fsmnt);
1.1       mycroft  1472:                        panic("ffs_nodealloccg: map corrupted");
                   1473:                        /* NOTREACHED */
                   1474:                }
                   1475:        }
                   1476:        i = start + len - loc;
1.60      fvdl     1477:        map = inosused[i];
1.1       mycroft  1478:        ipref = i * NBBY;
                   1479:        for (i = 1; i < (1 << NBBY); i <<= 1, ipref++) {
                   1480:                if ((map & i) == 0) {
1.19      bouyer   1481:                        cgp->cg_irotor = ufs_rw32(ipref, needswap);
1.1       mycroft  1482:                        goto gotit;
                   1483:                }
                   1484:        }
1.13      christos 1485:        printf("fs = %s\n", fs->fs_fsmnt);
1.1       mycroft  1486:        panic("ffs_nodealloccg: block not in map");
                   1487:        /* NOTREACHED */
                   1488: gotit:
1.30      fvdl     1489:        if (DOINGSOFTDEP(ITOV(ip)))
                   1490:                softdep_setup_inomapdep(bp, ip, cg * fs->fs_ipg + ipref);
1.60      fvdl     1491:        setbit(inosused, ipref);
1.19      bouyer   1492:        ufs_add32(cgp->cg_cs.cs_nifree, -1, needswap);
1.1       mycroft  1493:        fs->fs_cstotal.cs_nifree--;
1.30      fvdl     1494:        fs->fs_cs(fs, cg).cs_nifree--;
1.1       mycroft  1495:        fs->fs_fmod = 1;
                   1496:        if ((mode & IFMT) == IFDIR) {
1.19      bouyer   1497:                ufs_add32(cgp->cg_cs.cs_ndir, 1, needswap);
1.1       mycroft  1498:                fs->fs_cstotal.cs_ndir++;
                   1499:                fs->fs_cs(fs, cg).cs_ndir++;
                   1500:        }
1.60      fvdl     1501:        /*
                   1502:         * Check to see if we need to initialize more inodes.
                   1503:         */
                   1504:        initediblk = ufs_rw32(cgp->cg_initediblk, needswap);
                   1505:        if (fs->fs_magic == FS_UFS2_MAGIC &&
                   1506:            ipref + INOPB(fs) > initediblk &&
                   1507:            initediblk < ufs_rw32(cgp->cg_niblk, needswap)) {
                   1508:                ibp = getblk(ip->i_devvp, fsbtodb(fs,
                   1509:                    ino_to_fsba(fs, cg * fs->fs_ipg + initediblk)),
                   1510:                    (int)fs->fs_bsize, 0, 0);
                   1511:                    memset(ibp->b_data, 0, fs->fs_bsize);
                   1512:                    dp2 = (struct ufs2_dinode *)(ibp->b_data);
                   1513:                    for (i = 0; i < INOPB(fs); i++) {
                   1514:                        /*
                   1515:                         * Don't bother to swap, it's supposed to be
                   1516:                         * random, after all.
                   1517:                         */
1.68.2.2  skrll    1518:                        dp2->di_gen = (arc4random() & INT32_MAX) / 2 + 1;
1.60      fvdl     1519:                        dp2++;
                   1520:                }
                   1521:                bawrite(ibp);
                   1522:                initediblk += INOPB(fs);
                   1523:                cgp->cg_initediblk = ufs_rw32(initediblk, needswap);
                   1524:        }
                   1525:
1.68.2.2  skrll    1526:        ACTIVECG_CLR(fs, cg);
1.1       mycroft  1527:        bdwrite(bp);
                   1528:        return (cg * fs->fs_ipg + ipref);
                   1529: }
                   1530:
                   1531: /*
                   1532:  * Free a block or fragment.
                   1533:  *
                   1534:  * The specified block or fragment is placed back in the
                   1535:  * free map. If a fragment is deallocated, a possible
                   1536:  * block reassembly is checked.
                   1537:  */
1.9       christos 1538: void
1.68.2.2  skrll    1539: ffs_blkfree(fs, devvp, bno, size, inum)
                   1540:        struct fs *fs;
                   1541:        struct vnode *devvp;
1.58      fvdl     1542:        daddr_t bno;
1.1       mycroft  1543:        long size;
1.68.2.2  skrll    1544:        ino_t inum;
1.1       mycroft  1545: {
1.33      augustss 1546:        struct cg *cgp;
1.1       mycroft  1547:        struct buf *bp;
1.68.2.2  skrll    1548:        struct ufsmount *ump;
1.60      fvdl     1549:        int32_t fragno, cgbno;
1.68.2.2  skrll    1550:        daddr_t cgblkno;
1.1       mycroft  1551:        int i, error, cg, blk, frags, bbase;
1.62      fvdl     1552:        u_int8_t *blksfree;
1.68.2.2  skrll    1553:        dev_t dev;
1.30      fvdl     1554:        const int needswap = UFS_FSNEEDSWAP(fs);
1.1       mycroft  1555:
1.68.2.2  skrll    1556:        cg = dtog(fs, bno);
                   1557:        if (devvp->v_type != VBLK) {
                   1558:                /* devvp is a snapshot */
                   1559:                dev = VTOI(devvp)->i_devvp->v_rdev;
                   1560:                cgblkno = fragstoblks(fs, cgtod(fs, cg));
                   1561:        } else {
                   1562:                dev = devvp->v_rdev;
                   1563:                ump = VFSTOUFS(devvp->v_specmountpoint);
                   1564:                cgblkno = fsbtodb(fs, cgtod(fs, cg));
                   1565:                if (TAILQ_FIRST(&ump->um_snapshots) != NULL &&
                   1566:                    ffs_snapblkfree(fs, devvp, bno, size, inum))
                   1567:                        return;
                   1568:        }
1.30      fvdl     1569:        if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0 ||
                   1570:            fragnum(fs, bno) + numfrags(fs, size) > fs->fs_frag) {
1.59      tsutsui  1571:                printf("dev = 0x%x, bno = %" PRId64 " bsize = %d, "
1.58      fvdl     1572:                       "size = %ld, fs = %s\n",
1.68.2.2  skrll    1573:                    dev, bno, fs->fs_bsize, size, fs->fs_fsmnt);
1.1       mycroft  1574:                panic("blkfree: bad size");
                   1575:        }
1.68.2.2  skrll    1576:
1.60      fvdl     1577:        if (bno >= fs->fs_size) {
1.68.2.2  skrll    1578:                printf("bad block %" PRId64 ", ino %d\n", bno, inum);
                   1579:                ffs_fserr(fs, inum, "bad block");
1.1       mycroft  1580:                return;
                   1581:        }
1.68.2.2  skrll    1582:        error = bread(devvp, cgblkno, (int)fs->fs_cgsize, NOCRED, &bp);
1.1       mycroft  1583:        if (error) {
                   1584:                brelse(bp);
                   1585:                return;
                   1586:        }
                   1587:        cgp = (struct cg *)bp->b_data;
1.19      bouyer   1588:        if (!cg_chkmagic(cgp, needswap)) {
1.1       mycroft  1589:                brelse(bp);
                   1590:                return;
                   1591:        }
1.62      fvdl     1592:        cgp->cg_old_time = ufs_rw32(time.tv_sec, needswap);
1.68.2.2  skrll    1593:        if ((fs->fs_magic != FS_UFS1_MAGIC) ||
                   1594:            (fs->fs_old_flags & FS_FLAGS_UPDATED))
                   1595:                cgp->cg_time = ufs_rw64(time.tv_sec, needswap);
1.60      fvdl     1596:        cgbno = dtogd(fs, bno);
1.62      fvdl     1597:        blksfree = cg_blksfree(cgp, needswap);
1.1       mycroft  1598:        if (size == fs->fs_bsize) {
1.60      fvdl     1599:                fragno = fragstoblks(fs, cgbno);
1.62      fvdl     1600:                if (!ffs_isfreeblock(fs, blksfree, fragno)) {
1.68.2.2  skrll    1601:                        if (devvp->v_type != VBLK) {
                   1602:                                /* devvp is a snapshot */
                   1603:                                brelse(bp);
                   1604:                                return;
                   1605:                        }
1.59      tsutsui  1606:                        printf("dev = 0x%x, block = %" PRId64 ", fs = %s\n",
1.68.2.2  skrll    1607:                            dev, bno, fs->fs_fsmnt);
1.1       mycroft  1608:                        panic("blkfree: freeing free block");
                   1609:                }
1.62      fvdl     1610:                ffs_setblock(fs, blksfree, fragno);
1.60      fvdl     1611:                ffs_clusteracct(fs, cgp, fragno, 1);
1.19      bouyer   1612:                ufs_add32(cgp->cg_cs.cs_nbfree, 1, needswap);
1.1       mycroft  1613:                fs->fs_cstotal.cs_nbfree++;
                   1614:                fs->fs_cs(fs, cg).cs_nbfree++;
1.68.2.2  skrll    1615:                if ((fs->fs_magic == FS_UFS1_MAGIC) &&
                   1616:                    ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0)) {
                   1617:                        i = old_cbtocylno(fs, cgbno);
                   1618:                        KASSERT(i >= 0);
                   1619:                        KASSERT(i < fs->fs_old_ncyl);
                   1620:                        KASSERT(old_cbtorpos(fs, cgbno) >= 0);
                   1621:                        KASSERT(fs->fs_old_nrpos == 0 || old_cbtorpos(fs, cgbno) < fs->fs_old_nrpos);
                   1622:                        ufs_add16(old_cg_blks(fs, cgp, i, needswap)[old_cbtorpos(fs, cgbno)], 1,
                   1623:                            needswap);
                   1624:                        ufs_add32(old_cg_blktot(cgp, needswap)[i], 1, needswap);
                   1625:                }
1.1       mycroft  1626:        } else {
1.60      fvdl     1627:                bbase = cgbno - fragnum(fs, cgbno);
1.1       mycroft  1628:                /*
                   1629:                 * decrement the counts associated with the old frags
                   1630:                 */
1.62      fvdl     1631:                blk = blkmap(fs, blksfree, bbase);
1.19      bouyer   1632:                ffs_fragacct(fs, blk, cgp->cg_frsum, -1, needswap);
1.1       mycroft  1633:                /*
                   1634:                 * deallocate the fragment
                   1635:                 */
                   1636:                frags = numfrags(fs, size);
                   1637:                for (i = 0; i < frags; i++) {
1.62      fvdl     1638:                        if (isset(blksfree, cgbno + i)) {
1.59      tsutsui  1639:                                printf("dev = 0x%x, block = %" PRId64
                   1640:                                       ", fs = %s\n",
1.68.2.2  skrll    1641:                                    dev, bno + i, fs->fs_fsmnt);
1.1       mycroft  1642:                                panic("blkfree: freeing free frag");
                   1643:                        }
1.62      fvdl     1644:                        setbit(blksfree, cgbno + i);
1.1       mycroft  1645:                }
1.19      bouyer   1646:                ufs_add32(cgp->cg_cs.cs_nffree, i, needswap);
1.1       mycroft  1647:                fs->fs_cstotal.cs_nffree += i;
1.30      fvdl     1648:                fs->fs_cs(fs, cg).cs_nffree += i;
1.1       mycroft  1649:                /*
                   1650:                 * add back in counts associated with the new frags
                   1651:                 */
1.62      fvdl     1652:                blk = blkmap(fs, blksfree, bbase);
1.19      bouyer   1653:                ffs_fragacct(fs, blk, cgp->cg_frsum, 1, needswap);
1.1       mycroft  1654:                /*
                   1655:                 * if a complete block has been reassembled, account for it
                   1656:                 */
1.60      fvdl     1657:                fragno = fragstoblks(fs, bbase);
1.62      fvdl     1658:                if (ffs_isblock(fs, blksfree, fragno)) {
1.19      bouyer   1659:                        ufs_add32(cgp->cg_cs.cs_nffree, -fs->fs_frag, needswap);
1.1       mycroft  1660:                        fs->fs_cstotal.cs_nffree -= fs->fs_frag;
                   1661:                        fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
1.60      fvdl     1662:                        ffs_clusteracct(fs, cgp, fragno, 1);
1.19      bouyer   1663:                        ufs_add32(cgp->cg_cs.cs_nbfree, 1, needswap);
1.1       mycroft  1664:                        fs->fs_cstotal.cs_nbfree++;
                   1665:                        fs->fs_cs(fs, cg).cs_nbfree++;
1.68.2.2  skrll    1666:                        if ((fs->fs_magic == FS_UFS1_MAGIC) &&
                   1667:                            ((fs->fs_old_flags & FS_FLAGS_UPDATED) == 0)) {
                   1668:                                i = old_cbtocylno(fs, bbase);
                   1669:                                KASSERT(i >= 0);
                   1670:                                KASSERT(i < fs->fs_old_ncyl);
                   1671:                                KASSERT(old_cbtorpos(fs, bbase) >= 0);
                   1672:                                KASSERT(fs->fs_old_nrpos == 0 || old_cbtorpos(fs, bbase) < fs->fs_old_nrpos);
                   1673:                                ufs_add16(old_cg_blks(fs, cgp, i, needswap)[old_cbtorpos(fs,
                   1674:                                    bbase)], 1, needswap);
                   1675:                                ufs_add32(old_cg_blktot(cgp, needswap)[i], 1, needswap);
                   1676:                        }
1.1       mycroft  1677:                }
                   1678:        }
                   1679:        fs->fs_fmod = 1;
1.68.2.2  skrll    1680:        ACTIVECG_CLR(fs, cg);
1.1       mycroft  1681:        bdwrite(bp);
                   1682: }
                   1683:
1.18      fvdl     1684: #if defined(DIAGNOSTIC) || defined(DEBUG)
1.55      matt     1685: #ifdef XXXUBC
1.18      fvdl     1686: /*
                   1687:  * Verify allocation of a block or fragment. Returns true if block or
                   1688:  * fragment is allocated, false if it is free.
                   1689:  */
                   1690: static int
                   1691: ffs_checkblk(ip, bno, size)
                   1692:        struct inode *ip;
1.58      fvdl     1693:        daddr_t bno;
1.18      fvdl     1694:        long size;
                   1695: {
                   1696:        struct fs *fs;
                   1697:        struct cg *cgp;
                   1698:        struct buf *bp;
                   1699:        int i, error, frags, free;
                   1700:
                   1701:        fs = ip->i_fs;
                   1702:        if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
                   1703:                printf("bsize = %d, size = %ld, fs = %s\n",
                   1704:                    fs->fs_bsize, size, fs->fs_fsmnt);
                   1705:                panic("checkblk: bad size");
                   1706:        }
1.60      fvdl     1707:        if (bno >= fs->fs_size)
1.18      fvdl     1708:                panic("checkblk: bad block %d", bno);
                   1709:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, dtog(fs, bno))),
                   1710:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1711:        if (error) {
                   1712:                brelse(bp);
                   1713:                return 0;
                   1714:        }
                   1715:        cgp = (struct cg *)bp->b_data;
1.30      fvdl     1716:        if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs))) {
1.18      fvdl     1717:                brelse(bp);
                   1718:                return 0;
                   1719:        }
                   1720:        bno = dtogd(fs, bno);
                   1721:        if (size == fs->fs_bsize) {
1.30      fvdl     1722:                free = ffs_isblock(fs, cg_blksfree(cgp, UFS_FSNEEDSWAP(fs)),
1.19      bouyer   1723:                        fragstoblks(fs, bno));
1.18      fvdl     1724:        } else {
                   1725:                frags = numfrags(fs, size);
                   1726:                for (free = 0, i = 0; i < frags; i++)
1.30      fvdl     1727:                        if (isset(cg_blksfree(cgp, UFS_FSNEEDSWAP(fs)), bno + i))
1.18      fvdl     1728:                                free++;
                   1729:                if (free != 0 && free != frags)
                   1730:                        panic("checkblk: partially free fragment");
                   1731:        }
                   1732:        brelse(bp);
                   1733:        return (!free);
                   1734: }
1.55      matt     1735: #endif /* XXXUBC */
1.18      fvdl     1736: #endif /* DIAGNOSTIC */
                   1737:
1.1       mycroft  1738: /*
                   1739:  * Free an inode.
1.30      fvdl     1740:  */
                   1741: int
                   1742: ffs_vfree(v)
                   1743:        void *v;
                   1744: {
                   1745:        struct vop_vfree_args /* {
                   1746:                struct vnode *a_pvp;
                   1747:                ino_t a_ino;
                   1748:                int a_mode;
                   1749:        } */ *ap = v;
                   1750:
                   1751:        if (DOINGSOFTDEP(ap->a_pvp)) {
                   1752:                softdep_freefile(ap);
                   1753:                return (0);
                   1754:        }
1.68.2.4  skrll    1755:        return (ffs_freefile(VTOI(ap->a_pvp)->i_fs, VTOI(ap->a_pvp)->i_devvp,
                   1756:            ap->a_ino, ap->a_mode));
1.30      fvdl     1757: }
                   1758:
                   1759: /*
                   1760:  * Do the actual free operation.
1.1       mycroft  1761:  * The specified inode is placed back in the free map.
                   1762:  */
                   1763: int
1.68.2.4  skrll    1764: ffs_freefile(fs, devvp, ino, mode)
                   1765:        struct fs *fs;
                   1766:        struct vnode *devvp;
                   1767:        ino_t ino;
                   1768:        int mode;
1.9       christos 1769: {
1.33      augustss 1770:        struct cg *cgp;
1.1       mycroft  1771:        struct buf *bp;
                   1772:        int error, cg;
1.68.2.2  skrll    1773:        daddr_t cgbno;
1.62      fvdl     1774:        u_int8_t *inosused;
1.68.2.4  skrll    1775:        dev_t dev;
1.19      bouyer   1776: #ifdef FFS_EI
1.30      fvdl     1777:        const int needswap = UFS_FSNEEDSWAP(fs);
1.19      bouyer   1778: #endif
1.1       mycroft  1779:
1.68.2.2  skrll    1780:        cg = ino_to_cg(fs, ino);
1.68.2.4  skrll    1781:        if (devvp->v_type != VBLK) {
                   1782:                /* devvp is a snapshot */
                   1783:                dev = VTOI(devvp)->i_devvp->v_rdev;
1.68.2.2  skrll    1784:                cgbno = fragstoblks(fs, cgtod(fs, cg));
                   1785:        } else {
1.68.2.4  skrll    1786:                dev = devvp->v_rdev;
1.68.2.2  skrll    1787:                cgbno = fsbtodb(fs, cgtod(fs, cg));
                   1788:        }
1.1       mycroft  1789:        if ((u_int)ino >= fs->fs_ipg * fs->fs_ncg)
1.56      provos   1790:                panic("ifree: range: dev = 0x%x, ino = %d, fs = %s",
1.68.2.4  skrll    1791:                    dev, ino, fs->fs_fsmnt);
                   1792:        error = bread(devvp, cgbno, (int)fs->fs_cgsize, NOCRED, &bp);
1.1       mycroft  1793:        if (error) {
                   1794:                brelse(bp);
1.30      fvdl     1795:                return (error);
1.1       mycroft  1796:        }
                   1797:        cgp = (struct cg *)bp->b_data;
1.19      bouyer   1798:        if (!cg_chkmagic(cgp, needswap)) {
1.1       mycroft  1799:                brelse(bp);
                   1800:                return (0);
                   1801:        }
1.62      fvdl     1802:        cgp->cg_old_time = ufs_rw32(time.tv_sec, needswap);
1.68.2.2  skrll    1803:        if ((fs->fs_magic != FS_UFS1_MAGIC) ||
                   1804:            (fs->fs_old_flags & FS_FLAGS_UPDATED))
                   1805:                cgp->cg_time = ufs_rw64(time.tv_sec, needswap);
1.62      fvdl     1806:        inosused = cg_inosused(cgp, needswap);
1.1       mycroft  1807:        ino %= fs->fs_ipg;
1.62      fvdl     1808:        if (isclr(inosused, ino)) {
1.13      christos 1809:                printf("dev = 0x%x, ino = %d, fs = %s\n",
1.68.2.4  skrll    1810:                    dev, ino, fs->fs_fsmnt);
1.1       mycroft  1811:                if (fs->fs_ronly == 0)
                   1812:                        panic("ifree: freeing free inode");
                   1813:        }
1.62      fvdl     1814:        clrbit(inosused, ino);
1.19      bouyer   1815:        if (ino < ufs_rw32(cgp->cg_irotor, needswap))
                   1816:                cgp->cg_irotor = ufs_rw32(ino, needswap);
                   1817:        ufs_add32(cgp->cg_cs.cs_nifree, 1, needswap);
1.1       mycroft  1818:        fs->fs_cstotal.cs_nifree++;
                   1819:        fs->fs_cs(fs, cg).cs_nifree++;
1.68.2.4  skrll    1820:        if ((mode & IFMT) == IFDIR) {
1.19      bouyer   1821:                ufs_add32(cgp->cg_cs.cs_ndir, -1, needswap);
1.1       mycroft  1822:                fs->fs_cstotal.cs_ndir--;
                   1823:                fs->fs_cs(fs, cg).cs_ndir--;
                   1824:        }
                   1825:        fs->fs_fmod = 1;
                   1826:        bdwrite(bp);
                   1827:        return (0);
                   1828: }
                   1829:
                   1830: /*
1.68.2.2  skrll    1831:  * Check to see if a file is free.
                   1832:  */
                   1833: int
                   1834: ffs_checkfreefile(fs, devvp, ino)
                   1835:        struct fs *fs;
                   1836:        struct vnode *devvp;
                   1837:        ino_t ino;
                   1838: {
                   1839:        struct cg *cgp;
                   1840:        struct buf *bp;
                   1841:        daddr_t cgbno;
                   1842:        int ret, cg;
                   1843:        u_int8_t *inosused;
                   1844:
                   1845:        cg = ino_to_cg(fs, ino);
                   1846:        if (devvp->v_type != VBLK) {
                   1847:                /* devvp is a snapshot */
                   1848:                cgbno = fragstoblks(fs, cgtod(fs, cg));
                   1849:        } else
                   1850:                cgbno = fsbtodb(fs, cgtod(fs, cg));
                   1851:        if ((u_int)ino >= fs->fs_ipg * fs->fs_ncg)
                   1852:                return 1;
                   1853:        if (bread(devvp, cgbno, (int)fs->fs_cgsize, NOCRED, &bp)) {
                   1854:                brelse(bp);
                   1855:                return 1;
                   1856:        }
                   1857:        cgp = (struct cg *)bp->b_data;
                   1858:        if (!cg_chkmagic(cgp, UFS_FSNEEDSWAP(fs))) {
                   1859:                brelse(bp);
                   1860:                return 1;
                   1861:        }
                   1862:        inosused = cg_inosused(cgp, UFS_FSNEEDSWAP(fs));
                   1863:        ino %= fs->fs_ipg;
                   1864:        ret = isclr(inosused, ino);
                   1865:        brelse(bp);
                   1866:        return ret;
                   1867: }
                   1868:
                   1869: /*
1.1       mycroft  1870:  * Find a block of the specified size in the specified cylinder group.
                   1871:  *
                   1872:  * It is a panic if a request is made to find a block if none are
                   1873:  * available.
                   1874:  */
1.60      fvdl     1875: static int32_t
1.30      fvdl     1876: ffs_mapsearch(fs, cgp, bpref, allocsiz)
1.33      augustss 1877:        struct fs *fs;
                   1878:        struct cg *cgp;
1.58      fvdl     1879:        daddr_t bpref;
1.1       mycroft  1880:        int allocsiz;
                   1881: {
1.60      fvdl     1882:        int32_t bno;
1.1       mycroft  1883:        int start, len, loc, i;
                   1884:        int blk, field, subfield, pos;
1.19      bouyer   1885:        int ostart, olen;
1.62      fvdl     1886:        u_int8_t *blksfree;
1.30      fvdl     1887: #ifdef FFS_EI
                   1888:        const int needswap = UFS_FSNEEDSWAP(fs);
                   1889: #endif
1.1       mycroft  1890:
                   1891:        /*
                   1892:         * find the fragment by searching through the free block
                   1893:         * map for an appropriate bit pattern
                   1894:         */
                   1895:        if (bpref)
                   1896:                start = dtogd(fs, bpref) / NBBY;
                   1897:        else
1.19      bouyer   1898:                start = ufs_rw32(cgp->cg_frotor, needswap) / NBBY;
1.62      fvdl     1899:        blksfree = cg_blksfree(cgp, needswap);
1.1       mycroft  1900:        len = howmany(fs->fs_fpg, NBBY) - start;
1.19      bouyer   1901:        ostart = start;
                   1902:        olen = len;
1.45      lukem    1903:        loc = scanc((u_int)len,
1.62      fvdl     1904:                (const u_char *)&blksfree[start],
1.45      lukem    1905:                (const u_char *)fragtbl[fs->fs_frag],
1.54      mycroft  1906:                (1 << (allocsiz - 1 + (fs->fs_frag & (NBBY - 1)))));
1.1       mycroft  1907:        if (loc == 0) {
                   1908:                len = start + 1;
                   1909:                start = 0;
1.45      lukem    1910:                loc = scanc((u_int)len,
1.62      fvdl     1911:                        (const u_char *)&blksfree[0],
1.45      lukem    1912:                        (const u_char *)fragtbl[fs->fs_frag],
1.54      mycroft  1913:                        (1 << (allocsiz - 1 + (fs->fs_frag & (NBBY - 1)))));
1.1       mycroft  1914:                if (loc == 0) {
1.13      christos 1915:                        printf("start = %d, len = %d, fs = %s\n",
1.19      bouyer   1916:                            ostart, olen, fs->fs_fsmnt);
1.20      ross     1917:                        printf("offset=%d %ld\n",
1.19      bouyer   1918:                                ufs_rw32(cgp->cg_freeoff, needswap),
1.62      fvdl     1919:                                (long)blksfree - (long)cgp);
                   1920:                        printf("cg %d\n", cgp->cg_cgx);
1.1       mycroft  1921:                        panic("ffs_alloccg: map corrupted");
                   1922:                        /* NOTREACHED */
                   1923:                }
                   1924:        }
                   1925:        bno = (start + len - loc) * NBBY;
1.19      bouyer   1926:        cgp->cg_frotor = ufs_rw32(bno, needswap);
1.1       mycroft  1927:        /*
                   1928:         * found the byte in the map
                   1929:         * sift through the bits to find the selected frag
                   1930:         */
                   1931:        for (i = bno + NBBY; bno < i; bno += fs->fs_frag) {
1.62      fvdl     1932:                blk = blkmap(fs, blksfree, bno);
1.1       mycroft  1933:                blk <<= 1;
                   1934:                field = around[allocsiz];
                   1935:                subfield = inside[allocsiz];
                   1936:                for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) {
                   1937:                        if ((blk & field) == subfield)
                   1938:                                return (bno + pos);
                   1939:                        field <<= 1;
                   1940:                        subfield <<= 1;
                   1941:                }
                   1942:        }
1.60      fvdl     1943:        printf("bno = %d, fs = %s\n", bno, fs->fs_fsmnt);
1.1       mycroft  1944:        panic("ffs_alloccg: block not in map");
1.58      fvdl     1945:        /* return (-1); */
1.1       mycroft  1946: }
                   1947:
                   1948: /*
                   1949:  * Update the cluster map because of an allocation or free.
                   1950:  *
                   1951:  * Cnt == 1 means free; cnt == -1 means allocating.
                   1952:  */
1.9       christos 1953: void
1.30      fvdl     1954: ffs_clusteracct(fs, cgp, blkno, cnt)
1.1       mycroft  1955:        struct fs *fs;
                   1956:        struct cg *cgp;
1.60      fvdl     1957:        int32_t blkno;
1.1       mycroft  1958:        int cnt;
                   1959: {
1.4       cgd      1960:        int32_t *sump;
1.5       mycroft  1961:        int32_t *lp;
1.1       mycroft  1962:        u_char *freemapp, *mapp;
                   1963:        int i, start, end, forw, back, map, bit;
1.30      fvdl     1964: #ifdef FFS_EI
                   1965:        const int needswap = UFS_FSNEEDSWAP(fs);
                   1966: #endif
1.1       mycroft  1967:
                   1968:        if (fs->fs_contigsumsize <= 0)
                   1969:                return;
1.19      bouyer   1970:        freemapp = cg_clustersfree(cgp, needswap);
                   1971:        sump = cg_clustersum(cgp, needswap);
1.1       mycroft  1972:        /*
                   1973:         * Allocate or clear the actual block.
                   1974:         */
                   1975:        if (cnt > 0)
                   1976:                setbit(freemapp, blkno);
                   1977:        else
                   1978:                clrbit(freemapp, blkno);
                   1979:        /*
                   1980:         * Find the size of the cluster going forward.
                   1981:         */
                   1982:        start = blkno + 1;
                   1983:        end = start + fs->fs_contigsumsize;
1.19      bouyer   1984:        if (end >= ufs_rw32(cgp->cg_nclusterblks, needswap))
                   1985:                end = ufs_rw32(cgp->cg_nclusterblks, needswap);
1.1       mycroft  1986:        mapp = &freemapp[start / NBBY];
                   1987:        map = *mapp++;
                   1988:        bit = 1 << (start % NBBY);
                   1989:        for (i = start; i < end; i++) {
                   1990:                if ((map & bit) == 0)
                   1991:                        break;
                   1992:                if ((i & (NBBY - 1)) != (NBBY - 1)) {
                   1993:                        bit <<= 1;
                   1994:                } else {
                   1995:                        map = *mapp++;
                   1996:                        bit = 1;
                   1997:                }
                   1998:        }
                   1999:        forw = i - start;
                   2000:        /*
                   2001:         * Find the size of the cluster going backward.
                   2002:         */
                   2003:        start = blkno - 1;
                   2004:        end = start - fs->fs_contigsumsize;
                   2005:        if (end < 0)
                   2006:                end = -1;
                   2007:        mapp = &freemapp[start / NBBY];
                   2008:        map = *mapp--;
                   2009:        bit = 1 << (start % NBBY);
                   2010:        for (i = start; i > end; i--) {
                   2011:                if ((map & bit) == 0)
                   2012:                        break;
                   2013:                if ((i & (NBBY - 1)) != 0) {
                   2014:                        bit >>= 1;
                   2015:                } else {
                   2016:                        map = *mapp--;
                   2017:                        bit = 1 << (NBBY - 1);
                   2018:                }
                   2019:        }
                   2020:        back = start - i;
                   2021:        /*
                   2022:         * Account for old cluster and the possibly new forward and
                   2023:         * back clusters.
                   2024:         */
                   2025:        i = back + forw + 1;
                   2026:        if (i > fs->fs_contigsumsize)
                   2027:                i = fs->fs_contigsumsize;
1.19      bouyer   2028:        ufs_add32(sump[i], cnt, needswap);
1.1       mycroft  2029:        if (back > 0)
1.19      bouyer   2030:                ufs_add32(sump[back], -cnt, needswap);
1.1       mycroft  2031:        if (forw > 0)
1.19      bouyer   2032:                ufs_add32(sump[forw], -cnt, needswap);
                   2033:
1.5       mycroft  2034:        /*
                   2035:         * Update cluster summary information.
                   2036:         */
                   2037:        lp = &sump[fs->fs_contigsumsize];
                   2038:        for (i = fs->fs_contigsumsize; i > 0; i--)
1.19      bouyer   2039:                if (ufs_rw32(*lp--, needswap) > 0)
1.5       mycroft  2040:                        break;
1.19      bouyer   2041:        fs->fs_maxcluster[ufs_rw32(cgp->cg_cgx, needswap)] = i;
1.1       mycroft  2042: }
                   2043:
                   2044: /*
                   2045:  * Fserr prints the name of a file system with an error diagnostic.
                   2046:  *
                   2047:  * The form of the error message is:
                   2048:  *     fs: error message
                   2049:  */
                   2050: static void
                   2051: ffs_fserr(fs, uid, cp)
                   2052:        struct fs *fs;
                   2053:        u_int uid;
                   2054:        char *cp;
                   2055: {
                   2056:
1.64      gmcgarry 2057:        log(LOG_ERR, "uid %d, pid %d, command %s, on %s: %s\n",
                   2058:            uid, curproc->p_pid, curproc->p_comm, fs->fs_fsmnt, cp);
1.1       mycroft  2059: }

CVSweb <webmaster@jp.NetBSD.org>