X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/743f1ef7466d7d6ad19a6b75ae83f34f2df22add..961945a80c6f995f4567dbce8881af0bbdee211c:/usr/src/sys/ufs/ffs/ffs_alloc.c diff --git a/usr/src/sys/ufs/ffs/ffs_alloc.c b/usr/src/sys/ufs/ffs/ffs_alloc.c index 3c11738343..e5a86e1bdc 100644 --- a/usr/src/sys/ufs/ffs/ffs_alloc.c +++ b/usr/src/sys/ufs/ffs/ffs_alloc.c @@ -1,8 +1,4 @@ -/* Copyright (c) 1981 Regents of the University of California */ - -static char vers[] = "@(#)ffs_alloc.c 1.7 %G%"; - -/* alloc.c 4.8 81/03/08 */ +/* ffs_alloc.c 2.20 82/12/17 */ #include "../h/param.h" #include "../h/systm.h" @@ -13,20 +9,40 @@ static char vers[] = "@(#)ffs_alloc.c 1.7 %G%"; #include "../h/inode.h" #include "../h/dir.h" #include "../h/user.h" +#include "../h/quota.h" +#include "../h/kernel.h" -extern long hashalloc(); -extern long ialloccg(); +extern u_long hashalloc(); +extern ino_t ialloccg(); extern daddr_t alloccg(); extern daddr_t alloccgblk(); extern daddr_t fragextend(); extern daddr_t blkpref(); extern daddr_t mapsearch(); extern int inside[], around[]; -extern unsigned char fragtbl[]; +extern unsigned char *fragtbl[]; +/* + * Allocate a block in the file system. + * + * The size of the requested block is given, which must be some + * multiple of fs_fsize and <= fs_bsize. + * A preference may be optionally specified. If a preference is given + * the following hierarchy is used to allocate a block: + * 1) allocate the requested block. + * 2) allocate a rotationally optimal block in the same cylinder. + * 3) allocate a block in the same cylinder group. + * 4) quadradically rehash into other cylinder groups, until an + * available block is located. + * If no block preference is given the following heirarchy is used + * to allocate a block: + * 1) allocate a block in the cylinder group that contains the + * inode for the file. + * 2) quadradically rehash into other cylinder groups, until an + * available block is located. + */ struct buf * -alloc(dev, ip, bpref, size) - dev_t dev; +alloc(ip, bpref, size) register struct inode *ip; daddr_t bpref; int size; @@ -36,21 +52,33 @@ alloc(dev, ip, bpref, size) register struct buf *bp; int cg; - if ((unsigned)size > BSIZE || size % FSIZE != 0) + fs = ip->i_fs; + if ((unsigned)size > fs->fs_bsize || fragoff(fs, size) != 0) { + printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n", + ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt); panic("alloc: bad size"); - fs = getfs(dev); - if (size == BSIZE && - (fs->fs_nbfree == 0 || - (!suser() && fs->fs_nbfree < fs->fs_nbfree * minfree / 100))) + } + if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0) + goto nospace; + if (u.u_uid != 0 && + fs->fs_cstotal.cs_nbfree * fs->fs_frag + fs->fs_cstotal.cs_nffree < + fs->fs_dsize * fs->fs_minfree / 100) goto nospace; +#ifdef QUOTA + if (chkdq(ip, (long)((unsigned)size/DEV_BSIZE), 0)) + return(NULL); +#endif + if (bpref >= fs->fs_size) + bpref = 0; if (bpref == 0) - cg = itog(ip->i_number, fs); + cg = itog(fs, ip->i_number); else - cg = dtog(bpref, fs); - bno = hashalloc(dev, fs, cg, (long)bpref, size, alloccg); - if (bno == 0) + cg = dtog(fs, bpref); + bno = (daddr_t)hashalloc(ip, cg, (long)bpref, size, + (u_long (*)())alloccg); + if (bno <= 0) goto nospace; - bp = getblk(dev, bno, size); + bp = getblk(ip->i_dev, fsbtodb(fs, bno), size); clrbuf(bp); return (bp); nospace: @@ -60,9 +88,16 @@ nospace: return (NULL); } +/* + * Reallocate a fragment to a bigger size + * + * The number and size of the old block is given, and a preference + * and new size is also specified. The allocator attempts to extend + * the original block. Failing that, the regular block allocator is + * invoked to get an appropriate block. + */ struct buf * -realloccg(dev, ip, bprev, bpref, osize, nsize) - dev_t dev; +realloccg(ip, bprev, bpref, osize, nsize) register struct inode *ip; daddr_t bprev, bpref; int osize, nsize; @@ -70,40 +105,60 @@ realloccg(dev, ip, bprev, bpref, osize, nsize) daddr_t bno; register struct fs *fs; register struct buf *bp, *obp; - caddr_t cp; int cg; - if ((unsigned)osize > BSIZE || osize % FSIZE != 0 || - (unsigned)nsize > BSIZE || nsize % FSIZE != 0) + fs = ip->i_fs; + if ((unsigned)osize > fs->fs_bsize || fragoff(fs, osize) != 0 || + (unsigned)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) { + printf("dev = 0x%x, bsize = %d, osize = %d, nsize = %d, fs = %s\n", + ip->i_dev, fs->fs_bsize, osize, nsize, fs->fs_fsmnt); panic("realloccg: bad size"); - fs = getfs(dev); - if (bprev == 0) + } + if (u.u_uid != 0 && + fs->fs_cstotal.cs_nbfree * fs->fs_frag + fs->fs_cstotal.cs_nffree < + fs->fs_dsize * fs->fs_minfree / 100) + goto nospace; + if (bprev == 0) { + printf("dev = 0x%x, bsize = %d, bprev = %d, fs = %s\n", + ip->i_dev, fs->fs_bsize, bprev, fs->fs_fsmnt); panic("realloccg: bad bprev"); - else - cg = dtog(bprev, fs); - bno = fragextend(dev, fs, cg, (long)bprev, osize, nsize); + } +#ifdef QUOTA + if (chkdq(ip, (long)((unsigned)(nsize-osize)/DEV_BSIZE), 0)) + return(NULL); +#endif + cg = dtog(fs, bprev); + bno = fragextend(ip, cg, (long)bprev, osize, nsize); if (bno != 0) { - bp = bread(dev, bno, osize); - bp->b_bcount = nsize; - blkclr(bp->b_un.b_addr + osize, nsize - osize); + do { + bp = bread(ip->i_dev, fsbtodb(fs, bno), osize); + if (bp->b_flags & B_ERROR) { + brelse(bp); + return (NULL); + } + } while (brealloc(bp, nsize) == 0); + bp->b_flags |= B_DONE; + bzero(bp->b_un.b_addr + osize, (unsigned)nsize - osize); return (bp); } - bno = hashalloc(dev, fs, cg, (long)bpref, nsize, alloccg); - if (bno != 0) { - /* - * make a new copy - */ - obp = bread(dev, bprev, osize); - bp = getblk(dev, bno, nsize); - cp = bp->b_un.b_addr; - bp->b_un.b_addr = obp->b_un.b_addr; - obp->b_un.b_addr = cp; - obp->b_flags |= B_INVAL; + if (bpref >= fs->fs_size) + bpref = 0; + bno = (daddr_t)hashalloc(ip, cg, (long)bpref, nsize, + (u_long (*)())alloccg); + if (bno > 0) { + obp = bread(ip->i_dev, fsbtodb(fs, bprev), osize); + if (obp->b_flags & B_ERROR) { + brelse(obp); + return (NULL); + } + bp = getblk(ip->i_dev, fsbtodb(fs, bno), nsize); + bcopy(obp->b_un.b_addr, bp->b_un.b_addr, (u_int)osize); + bzero(bp->b_un.b_addr + osize, (unsigned)nsize - osize); brelse(obp); - fre(dev, bprev, osize); - blkclr(bp->b_un.b_addr + osize, nsize - osize); - return(bp); + free(ip, bprev, (off_t)osize); + return (bp); } +nospace: /* * no space available */ @@ -113,101 +168,192 @@ realloccg(dev, ip, bprev, bpref, osize, nsize) return (NULL); } +/* + * Allocate an inode in the file system. + * + * A preference may be optionally specified. If a preference is given + * the following hierarchy is used to allocate an inode: + * 1) allocate the requested inode. + * 2) allocate an inode in the same cylinder group. + * 3) quadradically rehash into other cylinder groups, until an + * available inode is located. + * If no inode preference is given the following heirarchy is used + * to allocate an inode: + * 1) allocate an inode in cylinder group 0. + * 2) quadradically rehash into other cylinder groups, until an + * available inode is located. + */ struct inode * -ialloc(dev, ipref, mode) - dev_t dev; +ialloc(pip, ipref, mode) + register struct inode *pip; ino_t ipref; int mode; { - daddr_t ino; + ino_t ino; register struct fs *fs; register struct inode *ip; int cg; - fs = getfs(dev); - if (fs->fs_nifree == 0) + fs = pip->i_fs; + if (fs->fs_cstotal.cs_nifree == 0) goto noinodes; - cg = itog(ipref, fs); - ino = hashalloc(dev, fs, cg, (long)ipref, mode, ialloccg); +#ifdef QUOTA + if (chkiq(pip->i_dev, (struct inode *)NULL, u.u_uid, 0)) + return(NULL); +#endif + if (ipref >= fs->fs_ncg * fs->fs_ipg) + ipref = 0; + cg = itog(fs, ipref); + ino = (ino_t)hashalloc(pip, cg, (long)ipref, mode, ialloccg); if (ino == 0) goto noinodes; - ip = iget(dev, ino); + ip = iget(pip->i_dev, pip->i_fs, ino); if (ip == NULL) { - ifree(dev, ino); + ifree(ip, ino, 0); return (NULL); } - if (ip->i_mode) + if (ip->i_mode) { + printf("mode = 0%o, inum = %d, fs = %s\n", + ip->i_mode, ip->i_number, fs->fs_fsmnt); panic("ialloc: dup alloc"); + } return (ip); noinodes: fserr(fs, "out of inodes"); - uprintf("\n%s: create failed, no inodes free\n", fs->fs_fsmnt); + uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt); u.u_error = ENOSPC; return (NULL); } /* - * find a cylinder to place a directory + * Find a cylinder to place a directory. + * + * The policy implemented by this algorithm is to select from + * among those cylinder groups with above the average number of + * free inodes, the one with the smallest number of directories. */ -dirpref(dev) - dev_t dev; -{ +ino_t +dirpref(fs) register struct fs *fs; +{ int cg, minndir, mincg, avgifree; - fs = getfs(dev); - avgifree = fs->fs_nifree / fs->fs_ncg; + avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg; minndir = fs->fs_ipg; mincg = 0; for (cg = 0; cg < fs->fs_ncg; cg++) - if (fs->fs_cs(cg).cs_ndir < minndir && - fs->fs_cs(cg).cs_nifree >= avgifree) { + if (fs->fs_cs(fs, cg).cs_ndir < minndir && + fs->fs_cs(fs, cg).cs_nifree >= avgifree) { mincg = cg; - minndir = fs->fs_cs(cg).cs_ndir; + minndir = fs->fs_cs(fs, cg).cs_ndir; } - return (fs->fs_ipg * mincg); + return ((ino_t)(fs->fs_ipg * mincg)); } /* - * select a cylinder to place a large block of data + * Select the desired position for the next block in a file. The file is + * logically divided into sections. The first section is composed of the + * direct blocks. Each additional section contains fs_maxbpg blocks. + * + * If no blocks have been allocated in the first section, the policy is to + * request a block in the same cylinder group as the inode that describes + * the file. If no blocks have been allocated in any other section, the + * policy is to place the section in a cylinder group with a greater than + * average number of free blocks. An appropriate cylinder group is found + * by maintaining a rotor that sweeps the cylinder groups. When a new + * group of blocks is needed, the rotor is advanced until a cylinder group + * with greater than the average number of free blocks is found. + * + * If a section is already partially allocated, the policy is to + * contiguously allocate fs_maxcontig blocks. The end of one of these + * contiguous blocks and the beginning of the next is physically separated + * so that the disk head will be in transit between them for at least + * fs_rotdelay milliseconds. This is to allow time for the processor to + * schedule another I/O transfer. */ -blkpref(dev) - dev_t dev; +daddr_t +blkpref(ip, lbn, indx, bap) + struct inode *ip; + daddr_t lbn; + int indx; + daddr_t *bap; { register struct fs *fs; int cg, avgbfree; + daddr_t nextblk; - fs = getfs(dev); - avgbfree = fs->fs_nbfree / fs->fs_ncg; - for (cg = fs->fs_cgrotor + 1; cg < fs->fs_ncg; cg++) - if (fs->fs_cs(cg).cs_nbfree >= avgbfree) { - fs->fs_cgrotor = cg; - return (fs->fs_fpg * cg + FRAG); - } - for (cg = 0; cg <= fs->fs_cgrotor; cg++) - if (fs->fs_cs(cg).cs_nbfree >= avgbfree) { - fs->fs_cgrotor = cg; - return (fs->fs_fpg * cg + FRAG); + fs = ip->i_fs; + if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) { + if (lbn < NDADDR) { + cg = itog(fs, ip->i_number); + return (fs->fs_fpg * cg + fs->fs_frag); } - return (0); + /* + * Find a cylinder with greater than average number of + * unused data blocks. + */ + avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; + for (cg = fs->fs_cgrotor + 1; cg < fs->fs_ncg; cg++) + if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { + fs->fs_cgrotor = cg; + return (fs->fs_fpg * cg + fs->fs_frag); + } + for (cg = 0; cg <= fs->fs_cgrotor; cg++) + if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { + fs->fs_cgrotor = cg; + return (fs->fs_fpg * cg + fs->fs_frag); + } + return (NULL); + } + /* + * One or more previous blocks have been laid out. If less + * than fs_maxcontig previous blocks are contiguous, the + * next block is requested contiguously, otherwise it is + * requested rotationally delayed by fs_rotdelay milliseconds. + */ + nextblk = bap[indx - 1] + fs->fs_frag; + if (indx > fs->fs_maxcontig && + bap[indx - fs->fs_maxcontig] + fs->fs_frag * fs->fs_maxcontig + != nextblk) + return (nextblk); + if (fs->fs_rotdelay != 0) + /* + * Here we convert ms of delay to frags as: + * (frags) = (ms) * (rev/sec) * (sect/rev) / + * ((sect/frag) * (ms/sec)) + * then round up to the next block. + */ + nextblk += roundup(fs->fs_rotdelay * fs->fs_rps * fs->fs_nsect / + (NSPF(fs) * 1000), fs->fs_frag); + return (nextblk); } -long -hashalloc(dev, fs, cg, pref, size, allocator) - dev_t dev; - register struct fs *fs; +/* + * Implement the cylinder overflow algorithm. + * + * The policy implemented by this algorithm is: + * 1) allocate the block in its requested cylinder group. + * 2) quadradically rehash on the cylinder group number. + * 3) brute force search for a free block. + */ +/*VARARGS5*/ +u_long +hashalloc(ip, cg, pref, size, allocator) + struct inode *ip; int cg; long pref; int size; /* size for data blocks, mode for inodes */ - long (*allocator)(); + u_long (*allocator)(); { + register struct fs *fs; long result; int i, icg = cg; + fs = ip->i_fs; /* * 1: preferred cylinder group */ - result = (*allocator)(dev, fs, cg, pref, size); + result = (*allocator)(ip, cg, pref, size); if (result) return (result); /* @@ -217,7 +363,7 @@ hashalloc(dev, fs, cg, pref, size, allocator) cg += i; if (cg >= fs->fs_ncg) cg -= fs->fs_ncg; - result = (*allocator)(dev, fs, cg, 0, size); + result = (*allocator)(ip, cg, 0, size); if (result) return (result); } @@ -226,95 +372,113 @@ hashalloc(dev, fs, cg, pref, size, allocator) */ cg = icg; for (i = 0; i < fs->fs_ncg; i++) { - result = (*allocator)(dev, fs, cg, 0, size); + result = (*allocator)(ip, cg, 0, size); if (result) return (result); cg++; if (cg == fs->fs_ncg) cg = 0; } - return (0); + return (NULL); } +/* + * Determine whether a fragment can be extended. + * + * Check to see if the necessary fragments are available, and + * if they are, allocate them. + */ daddr_t -fragextend(dev, fs, cg, bprev, osize, nsize) - dev_t dev; - register struct fs *fs; +fragextend(ip, cg, bprev, osize, nsize) + struct inode *ip; int cg; long bprev; int osize, nsize; { + register struct fs *fs; register struct buf *bp; register struct cg *cgp; long bno; int frags, bbase; int i; - frags = nsize / FSIZE; - bbase = bprev % FRAG; - if (bbase > (bprev + frags - 1) % FRAG) { + fs = ip->i_fs; + if (fs->fs_cs(fs, cg).cs_nffree < nsize - osize) + return (NULL); + frags = numfrags(fs, nsize); + bbase = fragoff(fs, bprev); + if (bbase > (bprev + frags - 1) % fs->fs_frag) { /* cannot extend across a block boundry */ - return (0); + return (NULL); } - bp = bread(dev, cgtod(cg, fs), BSIZE); - if (bp->b_flags & B_ERROR) - return (0); + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize); cgp = bp->b_un.b_cg; - bno = bprev % fs->fs_fpg; - for (i = osize / FSIZE; i < frags; i++) { - if (isclr(cgp->cg_free, bno + i)) - break; + if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { + brelse(bp); + return (NULL); } - if (i == frags) { - /* - * the current fragment can be extended - * deduct the count on fragment being extended into - * increase the count on the remaining fragment (if any) - * allocate the extended piece - */ - for (i = frags; i < FRAG - bbase; i++) - if (isclr(cgp->cg_free, bno + i)) - break; - cgp->cg_frsum[i - osize / FSIZE]--; - if (i != frags) - cgp->cg_frsum[i - frags]++; - for (i = osize / FSIZE; i < frags; i++) { - clrbit(cgp->cg_free, bno + i); - cgp->cg_nffree--; - fs->fs_nffree--; + cgp->cg_time = time.tv_sec; + bno = dtogd(fs, bprev); + for (i = numfrags(fs, osize); i < frags; i++) + if (isclr(cgp->cg_free, bno + i)) { + brelse(bp); + return (NULL); } - fs->fs_fmod++; - bdwrite(bp); - return (bprev); + /* + * the current fragment can be extended + * deduct the count on fragment being extended into + * increase the count on the remaining fragment (if any) + * allocate the extended piece + */ + for (i = frags; i < fs->fs_frag - bbase; i++) + if (isclr(cgp->cg_free, bno + i)) + break; + cgp->cg_frsum[i - numfrags(fs, osize)]--; + if (i != frags) + cgp->cg_frsum[i - frags]++; + for (i = numfrags(fs, osize); i < frags; i++) { + clrbit(cgp->cg_free, bno + i); + cgp->cg_cs.cs_nffree--; + fs->fs_cstotal.cs_nffree--; + fs->fs_cs(fs, cg).cs_nffree--; } - brelse(bp); - return (0); + fs->fs_fmod++; + bdwrite(bp); + return (bprev); } +/* + * Determine whether a block can be allocated. + * + * Check to see if a block of the apprpriate size is available, + * and if it is, allocate it. + */ daddr_t -alloccg(dev, fs, cg, bpref, size) - dev_t dev; - register struct fs *fs; +alloccg(ip, cg, bpref, size) + struct inode *ip; int cg; daddr_t bpref; int size; { + register struct fs *fs; register struct buf *bp; register struct cg *cgp; int bno, frags; int allocsiz; register int i; - bp = bread(dev, cgtod(cg, fs), BSIZE); - if (bp->b_flags & B_ERROR) - return (0); + fs = ip->i_fs; + if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize) + return (NULL); + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize); cgp = bp->b_un.b_cg; - if (size == BSIZE) { - if (cgp->cg_nbfree == 0) { - brelse(bp); - return (0); - } - bno = alloccgblk(dev, fs, cgp, bpref); + if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { + brelse(bp); + return (NULL); + } + cgp->cg_time = time.tv_sec; + if (size == fs->fs_bsize) { + bno = alloccgblk(fs, cgp, bpref); bdwrite(bp); return (bno); } @@ -323,37 +487,41 @@ alloccg(dev, fs, cg, bpref, size) * allocsiz is the size which will be allocated, hacking * it down to a smaller size if necessary */ - frags = size / FSIZE; - for (allocsiz = frags; allocsiz < FRAG; allocsiz++) + frags = numfrags(fs, size); + for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++) if (cgp->cg_frsum[allocsiz] != 0) break; - if (allocsiz == FRAG) { + if (allocsiz == fs->fs_frag) { /* * no fragments were available, so a block will be * allocated, and hacked up */ - if (cgp->cg_nbfree == 0) { + if (cgp->cg_cs.cs_nbfree == 0) { brelse(bp); - return (0); + return (NULL); } - bno = alloccgblk(dev, fs, cgp, bpref); - bpref = bno % fs->fs_fpg; - for (i = frags; i < FRAG; i++) + bno = alloccgblk(fs, cgp, bpref); + bpref = dtogd(fs, bno); + for (i = frags; i < fs->fs_frag; i++) setbit(cgp->cg_free, bpref + i); - i = FRAG - frags; - cgp->cg_nffree += i; - fs->fs_nffree += i; + i = fs->fs_frag - frags; + cgp->cg_cs.cs_nffree += i; + fs->fs_cstotal.cs_nffree += i; + fs->fs_cs(fs, cg).cs_nffree += i; + fs->fs_fmod++; cgp->cg_frsum[i]++; bdwrite(bp); return (bno); } bno = mapsearch(fs, cgp, bpref, allocsiz); - if (bno == 0) - return (0); + if (bno < 0) + return (NULL); for (i = 0; i < frags; i++) clrbit(cgp->cg_free, bno + i); - cgp->cg_nffree -= frags; - fs->fs_nffree -= frags; + cgp->cg_cs.cs_nffree -= frags; + fs->fs_cstotal.cs_nffree -= frags; + fs->fs_cs(fs, cg).cs_nffree -= frags; + fs->fs_fmod++; cgp->cg_frsum[allocsiz]--; if (frags != allocsiz) cgp->cg_frsum[allocsiz - frags]++; @@ -361,93 +529,146 @@ alloccg(dev, fs, cg, bpref, size) return (cg * fs->fs_fpg + bno); } +/* + * Allocate a block in a cylinder group. + * + * This algorithm implements the following policy: + * 1) allocate the requested block. + * 2) allocate a rotationally optimal block in the same cylinder. + * 3) allocate the next available block on the block rotor for the + * specified cylinder group. + * Note that this routine only allocates fs_bsize blocks; these + * blocks may be fragmented by the routine that allocates them. + */ daddr_t -alloccgblk(dev, fs, cgp, bpref) - dev_t dev; - struct fs *fs; +alloccgblk(fs, cgp, bpref) + register struct fs *fs; register struct cg *cgp; daddr_t bpref; { daddr_t bno; - int cylno, pos; + int cylno, pos, delta; short *cylbp; - int i, j; + register int i; if (bpref == 0) { bpref = cgp->cg_rotor; - } else { - bpref &= ~(FRAG - 1); - bpref %= fs->fs_fpg; - /* - * if the requested block is available, use it - */ - if (isblock(cgp->cg_free, bpref/FRAG)) { - bno = bpref; - goto gotit; - } + goto norot; + } + bpref &= ~(fs->fs_frag - 1); + bpref = dtogd(fs, bpref); + /* + * if the requested block is available, use it + */ + if (isblock(fs, cgp->cg_free, bpref/fs->fs_frag)) { + bno = bpref; + goto gotit; + } + /* + * check for a block available on the same cylinder + */ + cylno = cbtocylno(fs, bpref); + if (cgp->cg_btot[cylno] == 0) + goto norot; + if (fs->fs_cpc == 0) { /* - * check for a block available on the same cylinder - * beginning with one which is rotationally optimal + * block layout info is not available, so just have + * to take any block in this cylinder. */ - i = bpref * NSPF; - cylno = i / fs->fs_spc; - cylbp = cgp->cg_b[cylno]; - pos = (i + (ROTDELAY == 0) ? - 0 : 1 + ROTDELAY * HZ * fs->fs_nsect / (NSPF * 1000)) % - fs->fs_nsect * NRPOS / fs->fs_nsect; - for (i = pos; i < NRPOS; i++) + bpref = howmany(fs->fs_spc * cylno, NSPF(fs)); + goto norot; + } + /* + * check the summary information to see if a block is + * available in the requested cylinder starting at the + * requested rotational position and proceeding around. + */ + cylbp = cgp->cg_b[cylno]; + pos = cbtorpos(fs, bpref); + for (i = pos; i < NRPOS; i++) + if (cylbp[i] > 0) + break; + if (i == NRPOS) + for (i = 0; i < pos; i++) if (cylbp[i] > 0) break; - if (i == NRPOS) - for (i = 0; i < pos; i++) - if (cylbp[i] > 0) - break; - if (cylbp[i] > 0) { - bpref = cylno * fs->fs_spc / (NSPF * FRAG); - for (j = fs->fs_postbl[i]; j > -1; j = fs->fs_rotbl[j]) { - if (isblock(cgp->cg_free, bpref + j)) { - bno = (bpref + j) * FRAG; - goto gotit; - } + if (cylbp[i] > 0) { + /* + * found a rotational position, now find the actual + * block. A panic if none is actually there. + */ + pos = cylno % fs->fs_cpc; + bno = (cylno - pos) * fs->fs_spc / NSPB(fs); + if (fs->fs_postbl[pos][i] == -1) { + printf("pos = %d, i = %d, fs = %s\n", + pos, i, fs->fs_fsmnt); + panic("alloccgblk: cyl groups corrupted"); + } + for (i = fs->fs_postbl[pos][i];; ) { + if (isblock(fs, cgp->cg_free, bno + i)) { + bno = (bno + i) * fs->fs_frag; + goto gotit; } - panic("alloccgblk: can't find blk in cyl"); + delta = fs->fs_rotbl[i]; + if (delta <= 0 || delta > MAXBPC - i) + break; + i += delta; } + printf("pos = %d, i = %d, fs = %s\n", pos, i, fs->fs_fsmnt); + panic("alloccgblk: can't find blk in cyl"); } - bno = mapsearch(fs, cgp, bpref, FRAG); - if (bno == 0) - return (0); +norot: + /* + * no blocks in the requested cylinder, so take next + * available one in this cylinder group. + */ + bno = mapsearch(fs, cgp, bpref, (int)fs->fs_frag); + if (bno < 0) + return (NULL); cgp->cg_rotor = bno; gotit: - clrblock(cgp->cg_free, bno/FRAG); - cgp->cg_nbfree--; - fs->fs_nbfree--; - fs->fs_cs(cgp->cg_cgx).cs_nbfree--; - i = bno * NSPF; - cgp->cg_b[i/fs->fs_spc][i%fs->fs_nsect*NRPOS/fs->fs_nsect]--; + clrblock(fs, cgp->cg_free, (long)(bno/fs->fs_frag)); + cgp->cg_cs.cs_nbfree--; + fs->fs_cstotal.cs_nbfree--; + fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--; + cylno = cbtocylno(fs, bno); + cgp->cg_b[cylno][cbtorpos(fs, bno)]--; + cgp->cg_btot[cylno]--; fs->fs_fmod++; return (cgp->cg_cgx * fs->fs_fpg + bno); } -long -ialloccg(dev, fs, cg, ipref, mode) - dev_t dev; - register struct fs *fs; +/* + * Determine whether an inode can be allocated. + * + * Check to see if an inode is available, and if it is, + * allocate it using the following policy: + * 1) allocate the requested inode. + * 2) allocate the next available inode after the requested + * inode in the specified cylinder group. + */ +ino_t +ialloccg(ip, cg, ipref, mode) + struct inode *ip; int cg; daddr_t ipref; int mode; { + register struct fs *fs; register struct buf *bp; register struct cg *cgp; int i; - bp = bread(dev, cgtod(cg, fs), BSIZE); - if (bp->b_flags & B_ERROR) - return (0); + fs = ip->i_fs; + if (fs->fs_cs(fs, cg).cs_nifree == 0) + return (NULL); + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize); cgp = bp->b_un.b_cg; - if (cgp->cg_nifree == 0) { + if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { brelse(bp); - return (0); + return (NULL); } + cgp->cg_time = time.tv_sec; if (ipref) { ipref %= fs->fs_ipg; if (isclr(cgp->cg_iused, ipref)) @@ -464,25 +685,33 @@ ialloccg(dev, fs, cg, ipref, mode) } } brelse(bp); - return (0); + return (NULL); gotit: setbit(cgp->cg_iused, ipref); - cgp->cg_nifree--; - fs->fs_nifree--; - fs->fs_cs(cg).cs_nifree--; + cgp->cg_cs.cs_nifree--; + fs->fs_cstotal.cs_nifree--; + fs->fs_cs(fs, cg).cs_nifree--; fs->fs_fmod++; if ((mode & IFMT) == IFDIR) { - cgp->cg_ndir++; - fs->fs_cs(cg).cs_ndir++; + cgp->cg_cs.cs_ndir++; + fs->fs_cstotal.cs_ndir++; + fs->fs_cs(fs, cg).cs_ndir++; } bdwrite(bp); return (cg * fs->fs_ipg + ipref); } -fre(dev, bno, size) - dev_t dev; +/* + * Free a block or fragment. + * + * The specified block or fragment is placed back in the + * free map. If a fragment is deallocated, a possible + * block reassembly is checked. + */ +free(ip, bno, size) + register struct inode *ip; daddr_t bno; - int size; + off_t size; { register struct fs *fs; register struct cg *cgp; @@ -490,105 +719,135 @@ fre(dev, bno, size) int cg, blk, frags, bbase; register int i; - if ((unsigned)size > BSIZE || size % FSIZE != 0) + fs = ip->i_fs; + if ((unsigned)size > fs->fs_bsize || fragoff(fs, size) != 0) { + printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n", + ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt); panic("free: bad size"); - fs = getfs(dev); - cg = dtog(bno, fs); - if (badblock(fs, bno)) - return; - bp = bread(dev, cgtod(cg, fs), BSIZE); - if (bp->b_flags & B_ERROR) + } + cg = dtog(fs, bno); + if (badblock(fs, bno)) { + printf("bad block %d, ino %d\n", bno, ip->i_number); return; + } + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize); cgp = bp->b_un.b_cg; - bno %= fs->fs_fpg; - if (size == BSIZE) { - if (isblock(cgp->cg_free, bno/FRAG)) + if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { + brelse(bp); + return; + } + cgp->cg_time = time.tv_sec; + bno = dtogd(fs, bno); + if (size == fs->fs_bsize) { + if (isblock(fs, cgp->cg_free, bno/fs->fs_frag)) { + printf("dev = 0x%x, block = %d, fs = %s\n", + ip->i_dev, bno, fs->fs_fsmnt); panic("free: freeing free block"); - setblock(cgp->cg_free, bno/FRAG); - cgp->cg_nbfree++; - fs->fs_nbfree++; - fs->fs_cs(cg).cs_nbfree++; - i = bno * NSPF; - cgp->cg_b[i/fs->fs_spc][i%fs->fs_nsect*NRPOS/fs->fs_nsect]++; + } + setblock(fs, cgp->cg_free, bno/fs->fs_frag); + cgp->cg_cs.cs_nbfree++; + fs->fs_cstotal.cs_nbfree++; + fs->fs_cs(fs, cg).cs_nbfree++; + i = cbtocylno(fs, bno); + cgp->cg_b[i][cbtorpos(fs, bno)]++; + cgp->cg_btot[i]++; } else { - bbase = bno - (bno % FRAG); + bbase = bno - (bno % fs->fs_frag); /* * decrement the counts associated with the old frags */ - blk = ((cgp->cg_free[bbase / NBBY] >> (bbase % NBBY)) & - (0xff >> (NBBY - FRAG))); - fragacct(blk, cgp->cg_frsum, -1); + blk = blkmap(fs, cgp->cg_free, bbase); + fragacct(fs, blk, cgp->cg_frsum, -1); /* * deallocate the fragment */ - frags = size / FSIZE; + frags = numfrags(fs, size); for (i = 0; i < frags; i++) { - if (isset(cgp->cg_free, bno + i)) + if (isset(cgp->cg_free, bno + i)) { + printf("dev = 0x%x, block = %d, fs = %s\n", + ip->i_dev, bno + i, fs->fs_fsmnt); panic("free: freeing free frag"); + } setbit(cgp->cg_free, bno + i); - cgp->cg_nffree++; - fs->fs_nffree++; } + cgp->cg_cs.cs_nffree += i; + fs->fs_cstotal.cs_nffree += i; + fs->fs_cs(fs, cg).cs_nffree += i; /* * add back in counts associated with the new frags */ - blk = ((cgp->cg_free[bbase / NBBY] >> (bbase % NBBY)) & - (0xff >> (NBBY - FRAG))); - fragacct(blk, cgp->cg_frsum, 1); + blk = blkmap(fs, cgp->cg_free, bbase); + fragacct(fs, blk, cgp->cg_frsum, 1); /* * if a complete block has been reassembled, account for it */ - if (isblock(cgp->cg_free, bbase / FRAG)) { - cgp->cg_nffree -= FRAG; - fs->fs_nffree -= FRAG; - cgp->cg_nbfree++; - fs->fs_nbfree++; - fs->fs_cs(cg).cs_nbfree++; - i = bbase * NSPF; - cgp->cg_b[i / fs->fs_spc] - [i % fs->fs_nsect * NRPOS / fs->fs_nsect]++; + if (isblock(fs, cgp->cg_free, bbase / fs->fs_frag)) { + cgp->cg_cs.cs_nffree -= fs->fs_frag; + fs->fs_cstotal.cs_nffree -= fs->fs_frag; + fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag; + cgp->cg_cs.cs_nbfree++; + fs->fs_cstotal.cs_nbfree++; + fs->fs_cs(fs, cg).cs_nbfree++; + i = cbtocylno(fs, bbase); + cgp->cg_b[i][cbtorpos(fs, bbase)]++; + cgp->cg_btot[i]++; } } fs->fs_fmod++; bdwrite(bp); } -ifree(dev, ino, mode) - dev_t dev; +/* + * Free an inode. + * + * The specified inode is placed back in the free map. + */ +ifree(ip, ino, mode) + struct inode *ip; ino_t ino; int mode; { register struct fs *fs; register struct cg *cgp; register struct buf *bp; - int i; int cg; - fs = getfs(dev); - if ((unsigned)ino >= fs->fs_ipg*fs->fs_ncg) + fs = ip->i_fs; + if ((unsigned)ino >= fs->fs_ipg*fs->fs_ncg) { + printf("dev = 0x%x, ino = %d, fs = %s\n", + ip->i_dev, ino, fs->fs_fsmnt); panic("ifree: range"); - cg = itog(ino, fs); - bp = bread(dev, cgtod(cg, fs), BSIZE); - if (bp->b_flags & B_ERROR) - return; + } + cg = itog(fs, ino); + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize); cgp = bp->b_un.b_cg; + if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { + brelse(bp); + return; + } + cgp->cg_time = time.tv_sec; ino %= fs->fs_ipg; - if (isclr(cgp->cg_iused, ino)) + if (isclr(cgp->cg_iused, ino)) { + printf("dev = 0x%x, ino = %d, fs = %s\n", + ip->i_dev, ino, fs->fs_fsmnt); panic("ifree: freeing free inode"); + } clrbit(cgp->cg_iused, ino); - cgp->cg_nifree++; - fs->fs_nifree++; - fs->fs_cs(cg).cs_nifree++; + cgp->cg_cs.cs_nifree++; + fs->fs_cstotal.cs_nifree++; + fs->fs_cs(fs, cg).cs_nifree++; if ((mode & IFMT) == IFDIR) { - cgp->cg_ndir--; - fs->fs_cs(cg).cs_ndir--; + cgp->cg_cs.cs_ndir--; + fs->fs_cstotal.cs_ndir--; + fs->fs_cs(fs, cg).cs_ndir--; } fs->fs_fmod++; bdwrite(bp); } /* - * find a block of the specified size in the specified cylinder group + * Find a block of the specified size in the specified cylinder group. + * * It is a panic if a request is made to find a block if none are * available. */ @@ -608,21 +867,19 @@ mapsearch(fs, cgp, bpref, allocsiz) * map for an appropriate bit pattern */ if (bpref) - start = bpref % fs->fs_fpg / NBBY; + start = dtogd(fs, bpref) / NBBY; else start = cgp->cg_frotor / NBBY; - len = roundup(fs->fs_fpg - 1, NBBY) / NBBY - start; - loc = scanc(len, &cgp->cg_free[start], fragtbl, 1 << (allocsiz - 1)); + len = howmany(fs->fs_fpg, NBBY) - start; + loc = scanc(len, &cgp->cg_free[start], fragtbl[fs->fs_frag], + 1 << (allocsiz - 1 + (fs->fs_frag % NBBY))); if (loc == 0) { - len = start - 1; - start = (cgdmin(cgp->cg_cgx, fs) - - cgbase(cgp->cg_cgx, fs)) / NBBY; - loc = scanc(len, &cgp->cg_free[start], fragtbl, - 1 << (allocsiz - 1)); - if (loc == 0) { - panic("alloccg: map corrupted"); - return (0); - } + len = start + 1; + start = 0; + loc = scanc(len, &cgp->cg_free[start], fragtbl[fs->fs_frag], + 1 << (allocsiz - 1 + (fs->fs_frag % NBBY))); + if (loc == 0) + return (-1); } bno = (start + len - loc) * NBBY; cgp->cg_frotor = bno; @@ -630,101 +887,27 @@ mapsearch(fs, cgp, bpref, allocsiz) * found the byte in the map * sift through the bits to find the selected frag */ - for (i = 0; i < NBBY; i += FRAG) { - blk = (cgp->cg_free[bno / NBBY] >> i) & (0xff >> NBBY - FRAG); + for (i = bno + NBBY; bno < i; bno += fs->fs_frag) { + blk = blkmap(fs, cgp->cg_free, bno); blk <<= 1; field = around[allocsiz]; subfield = inside[allocsiz]; - for (pos = 0; pos <= FRAG - allocsiz; pos++) { - if ((blk & field) == subfield) { - return (bno + i + pos); - } + for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) { + if ((blk & field) == subfield) + return (bno + pos); field <<= 1; subfield <<= 1; } } + printf("bno = %d, fs = %s\n", bno, fs->fs_fsmnt); panic("alloccg: block not in map"); - return (0); -} - -/* - * update the frsum fields to reflect addition or deletion - * of some frags - */ -fragacct(fragmap, fraglist, cnt) - int fragmap; - short fraglist[]; - int cnt; -{ - int inblk; - register int field, subfield; - register int siz, pos; - - inblk = (int)(fragtbl[fragmap]) << 1; - fragmap <<= 1; - for (siz = 1; siz < FRAG; siz++) { - if (((1 << siz) & inblk) == 0) - continue; - field = around[siz]; - subfield = inside[siz]; - for (pos = siz; pos <= FRAG; pos++) { - if ((fragmap & field) == subfield) { - fraglist[siz] += cnt; - pos += siz; - field <<= siz; - subfield <<= siz; - } - field <<= 1; - subfield <<= 1; - } - } -} - -badblock(fs, bn) - register struct fs *fs; - daddr_t bn; -{ - - if ((unsigned)bn >= fs->fs_size || bn < cgdmin(dtog(bn, fs), fs)) { - fserr(fs, "bad block"); - return (1); - } - return (0); -} - -/* - * getfs maps a device number into - * a pointer to the incore super - * block. The algorithm is a linear - * search through the mount table. - * A consistency check of the - * in core free-block and i-node - * counts is performed. - * - * panic: no fs -- the device is not mounted. - * this "cannot happen" - */ -struct fs * -getfs(dev) - dev_t dev; -{ - register struct mount *mp; - register struct fs *fs; - - for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++) - if (mp->m_bufp != NULL && mp->m_dev == dev) { - fs = mp->m_bufp->b_un.b_fs; - if (fs->fs_magic != FS_MAGIC) - panic("getfs: bad magic"); - return (fs); - } - panic("getfs: no fs"); - return (NULL); + return (-1); } /* - * Fserr prints the name of a file system - * with an error diagnostic, in the form + * Fserr prints the name of a file system with an error diagnostic. + * + * The form of the error message is: * fs: error message */ fserr(fs, cp) @@ -734,90 +917,3 @@ fserr(fs, cp) printf("%s: %s\n", fs->fs_fsmnt, cp); } - -/* - * Getfsx returns the index in the file system - * table of the specified device. The swap device - * is also assigned a pseudo-index. The index may - * be used as a compressed indication of the location - * of a block, recording - * - * rather than - * - * provided the information need remain valid only - * as long as the file system is mounted. - */ -getfsx(dev) - dev_t dev; -{ - register struct mount *mp; - - if (dev == swapdev) - return (MSWAPX); - for(mp = &mount[0]; mp < &mount[NMOUNT]; mp++) - if (mp->m_dev == dev) - return (mp - &mount[0]); - return (-1); -} - -/* - * Update is the internal name of 'sync'. It goes through the disk - * queues to initiate sandbagged IO; goes through the inodes to write - * modified nodes; and it goes through the mount table to initiate modified - * super blocks. - */ -update() -{ - register struct inode *ip; - register struct mount *mp; - register struct buf *bp; - struct fs *fs; - time_t tim; - int i, blks; - - if (updlock) - return; - updlock++; - /* - * Write back modified superblocks. - * Consistency check that the superblock - * of each file system is still in the buffer cache. - */ - for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++) - if (mp->m_bufp != NULL) { - fs = mp->m_bufp->b_un.b_fs; - if (fs->fs_fmod == 0) - continue; - if (fs->fs_ronly != 0) - panic("update: rofs mod"); - bp = getblk(mp->m_dev, SBLOCK, BSIZE); - fs->fs_fmod = 0; - fs->fs_time = TIME; - if (bp->b_un.b_fs != fs) - panic("update: bad b_fs"); - bwrite(bp); - blks = howmany(cssize(fs), BSIZE); - for (i = 0; i < blks; i++) { - bp = getblk(mp->m_dev, csaddr(fs) + (i * FRAG), - BSIZE); - bwrite(bp); - } - } - /* - * Write back each (modified) inode. - */ - for (ip = inode; ip < inodeNINODE; ip++) - if((ip->i_flag&ILOCK)==0 && ip->i_count) { - ip->i_flag |= ILOCK; - ip->i_count++; - tim = TIME; - iupdat(ip, &tim, &tim, 0); - iput(ip); - } - updlock = 0; - /* - * Force stale buffer cache information to be flushed, - * for all devices. - */ - bflush(NODEV); -}