X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/d65bd829f61c24841302f3eccd11bb193ebefd48..4c5ed28aa382bb34796c47f745f2abe32ad9bb97:/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 e34c2b3308..3095be9f80 100644 --- a/usr/src/sys/ufs/ffs/ffs_alloc.c +++ b/usr/src/sys/ufs/ffs/ffs_alloc.c @@ -1,26 +1,37 @@ -/* ffs_alloc.c 2.21 83/01/14 */ +/* + * Copyright (c) 1982, 1986, 1989 Regents of the University of California. + * All rights reserved. + * + * %sccs.include.redist.c% + * + * @(#)ffs_alloc.c 7.30 (Berkeley) %G% + */ + +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +#include +#include -#include "../h/param.h" -#include "../h/systm.h" -#include "../h/mount.h" -#include "../h/fs.h" -#include "../h/conf.h" -#include "../h/buf.h" -#include "../h/inode.h" -#include "../h/dir.h" -#include "../h/user.h" -#include "../h/quota.h" -#include "../h/kernel.h" +extern u_long nextgennumber; -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[]; +static daddr_t ffs_alloccg __P((struct inode *, int, daddr_t, int)); +static daddr_t ffs_alloccgblk __P((struct fs *, struct cg *, daddr_t)); +static ino_t ffs_dirpref __P((struct fs *)); +static daddr_t ffs_fragextend __P((struct inode *, int, long, int, int)); +static void ffs_fserr __P((struct fs *, u_int, char *)); +static u_long ffs_hashalloc + __P((struct inode *, int, long, int, u_long (*)())); +static ino_t ffs_ialloccg __P((struct inode *, int, daddr_t, int)); +static daddr_t ffs_mapsearch __P((struct fs *, struct cg *, daddr_t, int)); /* * Allocate a block in the file system. @@ -41,32 +52,32 @@ extern unsigned char *fragtbl[]; * 2) quadradically rehash into other cylinder groups, until an * available block is located. */ -struct buf * -alloc(ip, bpref, size) +ffs_alloc(ip, lbn, bpref, size, bnp) register struct inode *ip; - daddr_t bpref; + daddr_t lbn, bpref; int size; + daddr_t *bnp; { daddr_t bno; register struct fs *fs; register struct buf *bp; - int cg; + int cg, error; + struct ucred *cred = curproc->p_ucred; /* XXX */ + *bnp = 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"); + panic("ffs_alloc: bad size"); } 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) + if (cred->cr_uid != 0 && freespace(fs, fs->fs_minfree) <= 0) goto nospace; #ifdef QUOTA - if (chkdq(ip, (long)((unsigned)size/DEV_BSIZE), 0)) - return(NULL); + if (error = chkdq(ip, (long)btodb(size), cred, 0)) + return (error); #endif if (bpref >= fs->fs_size) bpref = 0; @@ -74,18 +85,24 @@ alloc(ip, bpref, size) cg = itog(fs, ip->i_number); else cg = dtog(fs, bpref); - bno = (daddr_t)hashalloc(ip, cg, (long)bpref, size, - (u_long (*)())alloccg); - if (bno <= 0) - goto nospace; - bp = getblk(ip->i_dev, fsbtodb(fs, bno), size); - clrbuf(bp); - return (bp); + bno = (daddr_t)ffs_hashalloc(ip, cg, (long)bpref, size, + (u_long (*)())ffs_alloccg); + if (bno > 0) { + ip->i_blocks += btodb(size); + ip->i_flag |= IUPD|ICHG; + *bnp = bno; + return (0); + } +#ifdef QUOTA + /* + * Restore user's disk quota because allocation failed. + */ + (void) chkdq(ip, (long)-btodb(size), cred, FORCE); +#endif nospace: - fserr(fs, "file system full"); + ffs_fserr(fs, cred->cr_uid, "file system full"); uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt); - u.u_error = ENOSPC; - return (NULL); + return (ENOSPC); } /* @@ -96,84 +113,159 @@ nospace: * the original block. Failing that, the regular block allocator is * invoked to get an appropriate block. */ -struct buf * -realloccg(ip, bprev, bpref, osize, nsize) +ffs_realloccg(ip, lbprev, bpref, osize, nsize, bpp) register struct inode *ip; - daddr_t bprev, bpref; + off_t lbprev; + daddr_t bpref; int osize, nsize; + struct buf **bpp; { - daddr_t bno; register struct fs *fs; - register struct buf *bp, *obp; - int cg; + struct buf *bp, *obp; + int cg, request, error; + daddr_t bprev, bno; + struct ucred *cred = curproc->p_ucred; /* XXX */ + *bpp = 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", + 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"); + panic("ffs_realloccg: bad size"); } - 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) + if (cred->cr_uid != 0 && freespace(fs, fs->fs_minfree) <= 0) goto nospace; - if (bprev == 0) { + if ((bprev = ip->i_db[lbprev]) == 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"); + panic("ffs_realloccg: bad bprev"); + } + /* + * Allocate the extra space in the buffer. + */ + if (error = bread(ITOV(ip), lbprev, osize, NOCRED, &bp)) { + brelse(bp); + return (error); } #ifdef QUOTA - if (chkdq(ip, (long)((unsigned)(nsize-osize)/DEV_BSIZE), 0)) - return(NULL); + if (error = chkdq(ip, (long)btodb(nsize - osize), cred, 0)) { + brelse(bp); + return (error); + } #endif + /* + * Check for extension in the existing location. + */ cg = dtog(fs, bprev); - bno = fragextend(ip, cg, (long)bprev, osize, nsize); - if (bno != 0) { - 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); + if (bno = ffs_fragextend(ip, cg, (long)bprev, osize, nsize)) { + if (bp->b_blkno != fsbtodb(fs, bno)) + panic("bad blockno"); + ip->i_blocks += btodb(nsize - osize); + ip->i_flag |= IUPD|ICHG; + allocbuf(bp, nsize); bp->b_flags |= B_DONE; bzero(bp->b_un.b_addr + osize, (unsigned)nsize - osize); - return (bp); + *bpp = bp; + return (0); } + /* + * Allocate a new disk location. + */ if (bpref >= fs->fs_size) bpref = 0; - bno = (daddr_t)hashalloc(ip, cg, (long)bpref, nsize, - (u_long (*)())alloccg); + switch ((int)fs->fs_optim) { + case FS_OPTSPACE: + /* + * Allocate an exact sized fragment. Although this makes + * best use of space, we will waste time relocating it if + * the file continues to grow. If the fragmentation is + * less than half of the minimum free reserve, we choose + * to begin optimizing for time. + */ + request = nsize; + if (fs->fs_minfree < 5 || + fs->fs_cstotal.cs_nffree > + fs->fs_dsize * fs->fs_minfree / (2 * 100)) + break; + log(LOG_NOTICE, "%s: optimization changed from SPACE to TIME\n", + fs->fs_fsmnt); + fs->fs_optim = FS_OPTTIME; + break; + case FS_OPTTIME: + /* + * At this point we have discovered a file that is trying to + * grow a small fragment to a larger fragment. To save time, + * we allocate a full sized block, then free the unused portion. + * If the file continues to grow, the `ffs_fragextend' call + * above will be able to grow it in place without further + * copying. If aberrant programs cause disk fragmentation to + * grow within 2% of the free reserve, we choose to begin + * optimizing for space. + */ + request = fs->fs_bsize; + if (fs->fs_cstotal.cs_nffree < + fs->fs_dsize * (fs->fs_minfree - 2) / 100) + break; + log(LOG_NOTICE, "%s: optimization changed from TIME to SPACE\n", + fs->fs_fsmnt); + fs->fs_optim = FS_OPTSPACE; + break; + default: + printf("dev = 0x%x, optim = %d, fs = %s\n", + ip->i_dev, fs->fs_optim, fs->fs_fsmnt); + panic("ffs_realloccg: bad optim"); + /* NOTREACHED */ + } + bno = (daddr_t)ffs_hashalloc(ip, cg, (long)bpref, request, + (u_long (*)())ffs_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); +#ifdef SECSIZE + obp = bread(ip->i_dev, fsbtodb(fs, bprev), osize, + fs->fs_dbsize); +#else SECSIZE + count = howmany(osize, CLBYTES); + for (i = 0; i < count; i++) +#ifdef SECSIZE + munhash(ip->i_dev, bn + i * CLBYTES / fs->fs_dbsize); +#else SECSIZE + munhash(ip->i_dev, bn + i * CLBYTES / DEV_BSIZE); +#endif SECSIZE + ffs_blkfree(ip, bprev, (off_t)osize); + if (nsize < request) + ffs_blkfree(ip, bno + numfrags(fs, nsize), + (off_t)(request - nsize)); + ip->i_blocks += btodb(nsize - osize); + ip->i_flag |= IUPD|ICHG; + allocbuf(bp, nsize); + bp->b_flags |= B_DONE; bzero(bp->b_un.b_addr + osize, (unsigned)nsize - osize); - brelse(obp); - free(ip, bprev, (off_t)osize); - return (bp); + *bpp = bp; + return (0); } +#ifdef QUOTA + /* + * Restore user's disk quota because allocation failed. + */ + (void) chkdq(ip, (long)-btodb(nsize - osize), cred, FORCE); +#endif + brelse(bp); nospace: /* * no space available */ - fserr(fs, "file system full"); + ffs_fserr(fs, cred->cr_uid, "file system full"); uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt); - u.u_error = ENOSPC; - return (NULL); + return (ENOSPC); } /* * 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. + * If allocating a directory, use ffs_dirpref to select the inode. + * If allocating in a directory, the following hierarchy is followed: + * 1) allocate the preferred inode. * 2) allocate an inode in the same cylinder group. * 3) quadradically rehash into other cylinder groups, until an * available inode is located. @@ -183,46 +275,62 @@ nospace: * 2) quadradically rehash into other cylinder groups, until an * available inode is located. */ -struct inode * -ialloc(pip, ipref, mode) - register struct inode *pip; - ino_t ipref; +ffs_valloc(pvp, mode, cred, vpp) + register struct vnode *pvp; int mode; + struct ucred *cred; + struct vnode **vpp; { - ino_t ino; + register struct inode *pip; register struct fs *fs; register struct inode *ip; - int cg; + ino_t ino, ipref; + int cg, error; + *vpp = NULL; + pip = VTOI(pvp); fs = pip->i_fs; if (fs->fs_cstotal.cs_nifree == 0) goto noinodes; -#ifdef QUOTA - if (chkiq(pip->i_dev, (struct inode *)NULL, u.u_uid, 0)) - return(NULL); -#endif + + if ((mode & IFMT) == IFDIR) + ipref = ffs_dirpref(fs); + else + ipref = pip->i_number; if (ipref >= fs->fs_ncg * fs->fs_ipg) ipref = 0; cg = itog(fs, ipref); - ino = (ino_t)hashalloc(pip, cg, (long)ipref, mode, ialloccg); + ino = (ino_t)ffs_hashalloc(pip, cg, (long)ipref, mode, ffs_ialloccg); if (ino == 0) goto noinodes; - ip = iget(pip->i_dev, pip->i_fs, ino); - if (ip == NULL) { - ifree(ip, ino, 0); - return (NULL); + error = ffs_vget(pvp->v_mount, ino, vpp); + if (error) { + ffs_vfree(pvp, ino, mode); + return (error); } + ip = VTOI(*vpp); 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"); + panic("ffs_valloc: dup alloc"); + } + if (ip->i_blocks) { /* XXX */ + printf("free inode %s/%d had %d blocks\n", + fs->fs_fsmnt, ino, ip->i_blocks); + ip->i_blocks = 0; } - return (ip); + ip->i_flags = 0; + /* + * Set up a new generation number for this inode. + */ + if (++nextgennumber < (u_long)time.tv_sec) + nextgennumber = time.tv_sec; + ip->i_gen = nextgennumber; + return (0); noinodes: - fserr(fs, "out of inodes"); + ffs_fserr(fs, cred->cr_uid, "out of inodes"); uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt); - u.u_error = ENOSPC; - return (NULL); + return (ENOSPC); } /* @@ -232,8 +340,8 @@ noinodes: * among those cylinder groups with above the average number of * free inodes, the one with the smallest number of directories. */ -ino_t -dirpref(fs) +static ino_t +ffs_dirpref(fs) register struct fs *fs; { int cg, minndir, mincg, avgifree; @@ -260,9 +368,14 @@ dirpref(fs) * 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. + * by using a rotor that sweeps the cylinder groups. When a new group of + * blocks is needed, the sweep begins in the cylinder group following the + * cylinder group from which the previous allocation was made. The sweep + * continues until a cylinder group with greater than the average number + * of free blocks is found. If the allocation is for the first block in an + * indirect block, the information on the previous allocation is unavailable; + * here a best guess is made based upon the logical block number being + * allocated. * * If a section is already partially allocated, the policy is to * contiguously allocate fs_maxcontig blocks. The end of one of these @@ -272,14 +385,15 @@ dirpref(fs) * schedule another I/O transfer. */ daddr_t -blkpref(ip, lbn, indx, bap) +ffs_blkpref(ip, lbn, indx, bap) struct inode *ip; daddr_t lbn; int indx; daddr_t *bap; { register struct fs *fs; - int cg, avgbfree; + register int cg; + int avgbfree, startcg; daddr_t nextblk; fs = ip->i_fs; @@ -292,13 +406,18 @@ blkpref(ip, lbn, indx, bap) * Find a cylinder with greater than average number of * unused data blocks. */ + if (indx == 0 || bap[indx - 1] == 0) + startcg = itog(fs, ip->i_number) + lbn / fs->fs_maxbpg; + else + startcg = dtog(fs, bap[indx - 1]) + 1; + startcg %= fs->fs_ncg; avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg; - for (cg = fs->fs_cgrotor + 1; cg < fs->fs_ncg; cg++) + for (cg = startcg; 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++) + for (cg = 0; cg <= startcg; cg++) if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) { fs->fs_cgrotor = cg; return (fs->fs_fpg * cg + fs->fs_frag); @@ -313,7 +432,7 @@ blkpref(ip, lbn, indx, bap) */ nextblk = bap[indx - 1] + fs->fs_frag; if (indx > fs->fs_maxcontig && - bap[indx - fs->fs_maxcontig] + fs->fs_frag * fs->fs_maxcontig + bap[indx - fs->fs_maxcontig] + blkstofrags(fs, fs->fs_maxcontig) != nextblk) return (nextblk); if (fs->fs_rotdelay != 0) @@ -337,8 +456,8 @@ blkpref(ip, lbn, indx, bap) * 3) brute force search for a free block. */ /*VARARGS5*/ -u_long -hashalloc(ip, cg, pref, size, allocator) +static u_long +ffs_hashalloc(ip, cg, pref, size, allocator) struct inode *ip; int cg; long pref; @@ -369,11 +488,11 @@ hashalloc(ip, cg, pref, size, allocator) } /* * 3: brute force search - * Note that we start at i == 3, since 0 was checked initially, - * and 1 and 2 are always checked in the quadratic rehash. + * Note that we start at i == 2, since 0 was checked initially, + * and 1 is always checked in the quadratic rehash. */ - cg = icg; - for (i = 3; i < fs->fs_ncg; i++) { + cg = (icg + 2) % fs->fs_ncg; + for (i = 2; i < fs->fs_ncg; i++) { result = (*allocator)(ip, cg, 0, size); if (result) return (result); @@ -390,39 +509,49 @@ hashalloc(ip, cg, pref, size, allocator) * Check to see if the necessary fragments are available, and * if they are, allocate them. */ -daddr_t -fragextend(ip, cg, bprev, osize, nsize) +static daddr_t +ffs_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; + struct buf *bp; long bno; int frags, bbase; - int i; + int i, error; fs = ip->i_fs; - if (fs->fs_cs(fs, cg).cs_nffree < nsize - osize) + if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, 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 */ + bbase = fragnum(fs, bprev); + if (bbase > fragnum(fs, (bprev + frags - 1))) { + /* cannot extend across a block boundary */ return (NULL); } - bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize); +#ifdef SECSIZE + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, + fs->fs_dbsize); +#else SECSIZE + error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), + (int)fs->fs_cgsize, NOCRED, &bp); + if (error) { + brelse(bp); + return (NULL); + } +#endif SECSIZE cgp = bp->b_un.b_cg; - if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { + if (!cg_chkmagic(cgp)) { brelse(bp); return (NULL); } 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)) { + if (isclr(cg_blksfree(cgp), bno + i)) { brelse(bp); return (NULL); } @@ -433,18 +562,18 @@ fragextend(ip, cg, bprev, osize, nsize) * allocate the extended piece */ for (i = frags; i < fs->fs_frag - bbase; i++) - if (isclr(cgp->cg_free, bno + i)) + if (isclr(cg_blksfree(cgp), 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); + clrbit(cg_blksfree(cgp), bno + i); cgp->cg_cs.cs_nffree--; fs->fs_cstotal.cs_nffree--; fs->fs_cs(fs, cg).cs_nffree--; } - fs->fs_fmod++; + fs->fs_fmod = 1; bdwrite(bp); return (bprev); } @@ -455,34 +584,42 @@ fragextend(ip, cg, bprev, osize, nsize) * Check to see if a block of the apprpriate size is available, * and if it is, allocate it. */ -daddr_t -alloccg(ip, cg, bpref, size) +static daddr_t +ffs_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; + struct buf *bp; register int i; + int error, bno, frags, allocsiz; 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_cgsize); - cgp = bp->b_un.b_cg; - if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { +#ifdef SECSIZE + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, + fs->fs_dbsize); +#else SECSIZE + error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), + (int)fs->fs_cgsize, NOCRED, &bp); + if (error) { brelse(bp); return (NULL); } - if (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize) +#endif SECSIZE + cgp = bp->b_un.b_cg; + if (!cg_chkmagic(cgp) || + (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) { + brelse(bp); return (NULL); + } cgp->cg_time = time.tv_sec; if (size == fs->fs_bsize) { - bno = alloccgblk(fs, cgp, bpref); + bno = ffs_alloccgblk(fs, cgp, bpref); bdwrite(bp); return (bno); } @@ -504,28 +641,30 @@ alloccg(ip, cg, bpref, size) brelse(bp); return (NULL); } - bno = alloccgblk(fs, cgp, bpref); + bno = ffs_alloccgblk(fs, cgp, bpref); bpref = dtogd(fs, bno); for (i = frags; i < fs->fs_frag; i++) - setbit(cgp->cg_free, bpref + i); + setbit(cg_blksfree(cgp), bpref + 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++; + fs->fs_fmod = 1; cgp->cg_frsum[i]++; bdwrite(bp); return (bno); } - bno = mapsearch(fs, cgp, bpref, allocsiz); - if (bno < 0) + bno = ffs_mapsearch(fs, cgp, bpref, allocsiz); + if (bno < 0) { + brelse(bp); return (NULL); + } for (i = 0; i < frags; i++) - clrbit(cgp->cg_free, bno + i); + clrbit(cg_blksfree(cgp), bno + i); cgp->cg_cs.cs_nffree -= frags; fs->fs_cstotal.cs_nffree -= frags; fs->fs_cs(fs, cg).cs_nffree -= frags; - fs->fs_fmod++; + fs->fs_fmod = 1; cgp->cg_frsum[allocsiz]--; if (frags != allocsiz) cgp->cg_frsum[allocsiz - frags]++; @@ -544,8 +683,8 @@ alloccg(ip, cg, bpref, size) * Note that this routine only allocates fs_bsize blocks; these * blocks may be fragmented by the routine that allocates them. */ -daddr_t -alloccgblk(fs, cgp, bpref) +static daddr_t +ffs_alloccgblk(fs, cgp, bpref) register struct fs *fs; register struct cg *cgp; daddr_t bpref; @@ -559,12 +698,12 @@ alloccgblk(fs, cgp, bpref) bpref = cgp->cg_rotor; goto norot; } - bpref &= ~(fs->fs_frag - 1); + bpref = blknum(fs, bpref); bpref = dtogd(fs, bpref); /* * if the requested block is available, use it */ - if (isblock(fs, cgp->cg_free, bpref/fs->fs_frag)) { + if (ffs_isblock(fs, cg_blksfree(cgp), fragstoblks(fs, bpref))) { bno = bpref; goto gotit; } @@ -572,7 +711,7 @@ alloccgblk(fs, cgp, bpref) * check for a block available on the same cylinder */ cylno = cbtocylno(fs, bpref); - if (cgp->cg_btot[cylno] == 0) + if (cg_blktot(cgp)[cylno] == 0) goto norot; if (fs->fs_cpc == 0) { /* @@ -587,12 +726,12 @@ alloccgblk(fs, cgp, bpref) * available in the requested cylinder starting at the * requested rotational position and proceeding around. */ - cylbp = cgp->cg_b[cylno]; + cylbp = cg_blks(fs, cgp, cylno); pos = cbtorpos(fs, bpref); - for (i = pos; i < NRPOS; i++) + for (i = pos; i < fs->fs_nrpos; i++) if (cylbp[i] > 0) break; - if (i == NRPOS) + if (i == fs->fs_nrpos) for (i = 0; i < pos; i++) if (cylbp[i] > 0) break; @@ -603,45 +742,46 @@ alloccgblk(fs, cgp, bpref) */ pos = cylno % fs->fs_cpc; bno = (cylno - pos) * fs->fs_spc / NSPB(fs); - if (fs->fs_postbl[pos][i] == -1) { + if (fs_postbl(fs, pos)[i] == -1) { printf("pos = %d, i = %d, fs = %s\n", pos, i, fs->fs_fsmnt); - panic("alloccgblk: cyl groups corrupted"); + panic("ffs_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; + for (i = fs_postbl(fs, pos)[i];; ) { + if (ffs_isblock(fs, cg_blksfree(cgp), bno + i)) { + bno = blkstofrags(fs, (bno + i)); goto gotit; } - delta = fs->fs_rotbl[i]; - if (delta <= 0 || delta > MAXBPC - i) + delta = fs_rotbl(fs)[i]; + if (delta <= 0 || + delta + i > fragstoblks(fs, fs->fs_fpg)) break; i += delta; } printf("pos = %d, i = %d, fs = %s\n", pos, i, fs->fs_fsmnt); - panic("alloccgblk: can't find blk in cyl"); + panic("ffs_alloccgblk: can't find blk in cyl"); } 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); + bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag); if (bno < 0) return (NULL); cgp->cg_rotor = bno; gotit: - clrblock(fs, cgp->cg_free, (long)(bno/fs->fs_frag)); + ffs_clrblock(fs, cg_blksfree(cgp), (long)fragstoblks(fs, bno)); 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++; + cg_blks(fs, cgp, cylno)[cbtorpos(fs, bno)]--; + cg_blktot(cgp)[cylno]--; + fs->fs_fmod = 1; return (cgp->cg_cgx * fs->fs_fpg + bno); } - + /* * Determine whether an inode can be allocated. * @@ -651,53 +791,75 @@ gotit: * 2) allocate the next available inode after the requested * inode in the specified cylinder group. */ -ino_t -ialloccg(ip, cg, ipref, mode) +static ino_t +ffs_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; + struct buf *bp; + int error, start, len, loc, map, i; 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_cgsize); - cgp = bp->b_un.b_cg; - if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { +#ifdef SECSIZE + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, + fs->fs_dbsize); +#else SECSIZE + error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), + (int)fs->fs_cgsize, NOCRED, &bp); + if (error) { brelse(bp); return (NULL); } - if (cgp->cg_cs.cs_nifree == 0) +#endif SECSIZE + cgp = bp->b_un.b_cg; + if (!cg_chkmagic(cgp) || cgp->cg_cs.cs_nifree == 0) { + brelse(bp); return (NULL); + } cgp->cg_time = time.tv_sec; if (ipref) { ipref %= fs->fs_ipg; - if (isclr(cgp->cg_iused, ipref)) + if (isclr(cg_inosused(cgp), ipref)) goto gotit; - } else - ipref = cgp->cg_irotor; - for (i = 0; i < fs->fs_ipg; i++) { - ipref++; - if (ipref >= fs->fs_ipg) - ipref = 0; - if (isclr(cgp->cg_iused, ipref)) { + } + start = cgp->cg_irotor / NBBY; + len = howmany(fs->fs_ipg - cgp->cg_irotor, NBBY); + loc = skpc(0xff, len, &cg_inosused(cgp)[start]); + if (loc == 0) { + len = start + 1; + start = 0; + loc = skpc(0xff, len, &cg_inosused(cgp)[0]); + if (loc == 0) { + printf("cg = %s, irotor = %d, fs = %s\n", + cg, cgp->cg_irotor, fs->fs_fsmnt); + panic("ffs_ialloccg: map corrupted"); + /* NOTREACHED */ + } + } + i = start + len - loc; + map = cg_inosused(cgp)[i]; + ipref = i * NBBY; + for (i = 1; i < (1 << NBBY); i <<= 1, ipref++) { + if ((map & i) == 0) { cgp->cg_irotor = ipref; goto gotit; } } - brelse(bp); - return (NULL); + printf("fs = %s\n", fs->fs_fsmnt); + panic("ffs_ialloccg: block not in map"); + /* NOTREACHED */ gotit: - setbit(cgp->cg_iused, ipref); + setbit(cg_inosused(cgp), ipref); cgp->cg_cs.cs_nifree--; fs->fs_cstotal.cs_nifree--; fs->fs_cs(fs, cg).cs_nifree--; - fs->fs_fmod++; + fs->fs_fmod = 1; if ((mode & IFMT) == IFDIR) { cgp->cg_cs.cs_ndir++; fs->fs_cstotal.cs_ndir++; @@ -714,67 +876,79 @@ gotit: * free map. If a fragment is deallocated, a possible * block reassembly is checked. */ -free(ip, bno, size) +ffs_blkfree(ip, bno, size) register struct inode *ip; daddr_t bno; off_t size; { register struct fs *fs; register struct cg *cgp; - register struct buf *bp; - int cg, blk, frags, bbase; + struct buf *bp; + int error, cg, blk, frags, bbase; register int i; + struct ucred *cred = curproc->p_ucred; /* XXX */ 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"); + panic("blkfree: bad size"); } cg = dtog(fs, bno); - if (badblock(fs, bno)) { + if ((unsigned)bno >= fs->fs_size) { printf("bad block %d, ino %d\n", bno, ip->i_number); + ffs_fserr(fs, cred->cr_uid, "bad block"); + return; + } +#ifdef SECSIZE + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, + fs->fs_dbsize); +#else SECSIZE + error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), + (int)fs->fs_cgsize, NOCRED, &bp); + if (error) { + brelse(bp); return; } - bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize); +#endif SECSIZE cgp = bp->b_un.b_cg; - if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { + if (!cg_chkmagic(cgp)) { 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)) { + if (ffs_isblock(fs, cg_blksfree(cgp), fragstoblks(fs, bno))) { printf("dev = 0x%x, block = %d, fs = %s\n", ip->i_dev, bno, fs->fs_fsmnt); - panic("free: freeing free block"); + panic("blkfree: freeing free block"); } - setblock(fs, cgp->cg_free, bno/fs->fs_frag); + ffs_setblock(fs, cg_blksfree(cgp), fragstoblks(fs, bno)); 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]++; + cg_blks(fs, cgp, i)[cbtorpos(fs, bno)]++; + cg_blktot(cgp)[i]++; } else { - bbase = bno - (bno % fs->fs_frag); + bbase = bno - fragnum(fs, bno); /* * decrement the counts associated with the old frags */ - blk = blkmap(fs, cgp->cg_free, bbase); - fragacct(fs, blk, cgp->cg_frsum, -1); + blk = blkmap(fs, cg_blksfree(cgp), bbase); + ffs_fragacct(fs, blk, cgp->cg_frsum, -1); /* * deallocate the fragment */ frags = numfrags(fs, size); for (i = 0; i < frags; i++) { - if (isset(cgp->cg_free, bno + i)) { + if (isset(cg_blksfree(cgp), bno + i)) { printf("dev = 0x%x, block = %d, fs = %s\n", ip->i_dev, bno + i, fs->fs_fsmnt); - panic("free: freeing free frag"); + panic("blkfree: freeing free frag"); } - setbit(cgp->cg_free, bno + i); + setbit(cg_blksfree(cgp), bno + i); } cgp->cg_cs.cs_nffree += i; fs->fs_cstotal.cs_nffree += i; @@ -782,12 +956,13 @@ free(ip, bno, size) /* * add back in counts associated with the new frags */ - blk = blkmap(fs, cgp->cg_free, bbase); - fragacct(fs, blk, cgp->cg_frsum, 1); + blk = blkmap(fs, cg_blksfree(cgp), bbase); + ffs_fragacct(fs, blk, cgp->cg_frsum, 1); /* * if a complete block has been reassembled, account for it */ - if (isblock(fs, cgp->cg_free, bbase / fs->fs_frag)) { + if (ffs_isblock(fs, cg_blksfree(cgp), + (daddr_t)fragstoblks(fs, bbase))) { 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; @@ -795,11 +970,11 @@ free(ip, bno, size) 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]++; + cg_blks(fs, cgp, i)[cbtorpos(fs, bbase)]++; + cg_blktot(cgp)[i]++; } } - fs->fs_fmod++; + fs->fs_fmod = 1; bdwrite(bp); } @@ -808,37 +983,51 @@ free(ip, bno, size) * * The specified inode is placed back in the free map. */ -ifree(ip, ino, mode) - struct inode *ip; +void +ffs_vfree(pvp, ino, mode) + struct vnode *pvp; ino_t ino; int mode; { register struct fs *fs; register struct cg *cgp; - register struct buf *bp; - int cg; + register struct inode *pip; + struct buf *bp; + int error, cg; - 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"); - } + pip = VTOI(pvp); + fs = pip->i_fs; + if ((u_int)ino >= fs->fs_ipg * fs->fs_ncg) + panic("ifree: range: dev = 0x%x, ino = %d, fs = %s\n", + pip->i_dev, ino, fs->fs_fsmnt); cg = itog(fs, ino); - bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize); +#ifdef SECSIZE + bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize, + fs->fs_dbsize); +#else SECSIZE + error = bread(pip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), + (int)fs->fs_cgsize, NOCRED, &bp); + if (error) { + brelse(bp); + return; + } +#endif SECSIZE cgp = bp->b_un.b_cg; - if (bp->b_flags & B_ERROR || cgp->cg_magic != CG_MAGIC) { + if (!cg_chkmagic(cgp)) { brelse(bp); return; } cgp->cg_time = time.tv_sec; ino %= fs->fs_ipg; - if (isclr(cgp->cg_iused, ino)) { + if (isclr(cg_inosused(cgp), ino)) { printf("dev = 0x%x, ino = %d, fs = %s\n", - ip->i_dev, ino, fs->fs_fsmnt); - panic("ifree: freeing free inode"); + pip->i_dev, ino, fs->fs_fsmnt); + if (fs->fs_ronly == 0) + panic("ifree: freeing free inode"); } - clrbit(cgp->cg_iused, ino); + clrbit(cg_inosused(cgp), ino); + if (ino < cgp->cg_irotor) + cgp->cg_irotor = ino; cgp->cg_cs.cs_nifree++; fs->fs_cstotal.cs_nifree++; fs->fs_cs(fs, cg).cs_nifree++; @@ -847,7 +1036,7 @@ ifree(ip, ino, mode) fs->fs_cstotal.cs_ndir--; fs->fs_cs(fs, cg).cs_ndir--; } - fs->fs_fmod++; + fs->fs_fmod = 1; bdwrite(bp); } @@ -857,8 +1046,8 @@ ifree(ip, ino, mode) * It is a panic if a request is made to find a block if none are * available. */ -daddr_t -mapsearch(fs, cgp, bpref, allocsiz) +static daddr_t +ffs_mapsearch(fs, cgp, bpref, allocsiz) register struct fs *fs; register struct cg *cgp; daddr_t bpref; @@ -877,15 +1066,21 @@ mapsearch(fs, cgp, bpref, allocsiz) else start = cgp->cg_frotor / NBBY; 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))); + loc = scanc((unsigned)len, (u_char *)&cg_blksfree(cgp)[start], + (u_char *)fragtbl[fs->fs_frag], + (u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY)))); if (loc == 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); + loc = scanc((unsigned)len, (u_char *)&cg_blksfree(cgp)[0], + (u_char *)fragtbl[fs->fs_frag], + (u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY)))); + if (loc == 0) { + printf("start = %d, len = %d, fs = %s\n", + start, len, fs->fs_fsmnt); + panic("ffs_alloccg: map corrupted"); + /* NOTREACHED */ + } } bno = (start + len - loc) * NBBY; cgp->cg_frotor = bno; @@ -894,7 +1089,7 @@ mapsearch(fs, cgp, bpref, allocsiz) * sift through the bits to find the selected frag */ for (i = bno + NBBY; bno < i; bno += fs->fs_frag) { - blk = blkmap(fs, cgp->cg_free, bno); + blk = blkmap(fs, cg_blksfree(cgp), bno); blk <<= 1; field = around[allocsiz]; subfield = inside[allocsiz]; @@ -906,7 +1101,7 @@ mapsearch(fs, cgp, bpref, allocsiz) } } printf("bno = %d, fs = %s\n", bno, fs->fs_fsmnt); - panic("alloccg: block not in map"); + panic("ffs_alloccg: block not in map"); return (-1); } @@ -916,10 +1111,12 @@ mapsearch(fs, cgp, bpref, allocsiz) * The form of the error message is: * fs: error message */ -fserr(fs, cp) +static void +ffs_fserr(fs, uid, cp) struct fs *fs; + u_int uid; char *cp; { - printf("%s: %s\n", fs->fs_fsmnt, cp); + log(LOG_ERR, "uid %d on %s: %s\n", uid, fs->fs_fsmnt, cp); }