-/* lfs_alloc.c 2.18 82/10/21 */
+/*
+ * Copyright (c) 1982, 1986, 1989 Regents of the University of California.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms are permitted
+ * provided that the above copyright notice and this paragraph are
+ * duplicated in all such forms and that any documentation,
+ * advertising materials, and other materials related to such
+ * distribution and use acknowledge that the software was developed
+ * by the University of California, Berkeley. The name of the
+ * University may not be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
+ * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
+ *
+ * @(#)lfs_alloc.c 7.12 (Berkeley) %G%
+ */
-#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"
+#include "param.h"
+#include "systm.h"
+#include "mount.h"
+#include "buf.h"
+#include "user.h"
+#include "vnode.h"
+#include "kernel.h"
+#include "syslog.h"
+#include "cmap.h"
+#include "../ufs/quota.h"
+#include "../ufs/inode.h"
+#include "../ufs/fs.h"
extern u_long hashalloc();
-extern u_long ialloccg();
-extern u_long alloccg();
+extern ino_t ialloccg();
+extern daddr_t alloccg();
extern daddr_t alloccgblk();
extern daddr_t fragextend();
extern daddr_t blkpref();
* 2) quadradically rehash into other cylinder groups, until an
* available block is located.
*/
-struct buf *
-alloc(ip, bpref, size)
+alloc(ip, bpref, size, bpp, flags)
register struct inode *ip;
daddr_t bpref;
int size;
+ struct buf **bpp;
+ int flags;
{
daddr_t bno;
register struct fs *fs;
register struct buf *bp;
- int cg;
+ int cg, error;
+ *bpp = 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",
}
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 (u.u_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), 0))
+ return (error);
#endif
if (bpref >= fs->fs_size)
bpref = 0;
cg = itog(fs, ip->i_number);
else
cg = dtog(fs, bpref);
- bno = (daddr_t)hashalloc(ip, cg, (long)bpref, size, alloccg);
+ 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);
+ ip->i_blocks += btodb(size);
+ ip->i_flag |= IUPD|ICHG;
+#ifdef SECSIZE
+ bp = getblk(ip->i_dev, fsbtodb(fs, bno), size, fs->fs_dbsize);
+#else SECSIZE
+ bp = getblk(ip->i_devvp, fsbtodb(fs, bno), size);
+ if (flags & B_CLRBUF)
+ clrbuf(bp);
+ *bpp = bp;
+ return (0);
nospace:
fserr(fs, "file system full");
uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
- u.u_error = ENOSPC;
- return (NULL);
+ return (ENOSPC);
}
/*
* the original block. Failing that, the regular block allocator is
* invoked to get an appropriate block.
*/
-struct buf *
-realloccg(ip, bprev, bpref, osize, nsize)
+realloccg(ip, bprev, bpref, osize, nsize, bpp)
register struct inode *ip;
daddr_t bprev, 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;
+ daddr_t bno, bn;
+ int i, error, count;
+ *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) {
ip->i_dev, fs->fs_bsize, osize, nsize, fs->fs_fsmnt);
panic("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 (u.u_uid != 0 && freespace(fs, fs->fs_minfree) <= 0)
goto nospace;
if (bprev == 0) {
printf("dev = 0x%x, bsize = %d, bprev = %d, fs = %s\n",
panic("realloccg: bad bprev");
}
#ifdef QUOTA
- if (chkdq(ip, (long)((unsigned)(nsize-osize)/DEV_BSIZE), 0))
- return(NULL);
+ if (error = chkdq(ip, (long)btodb(nsize - osize), 0))
+ return (error);
#endif
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) {
+#ifdef SECSIZE
+ bp = bread(ip->i_dev, fsbtodb(fs, bno), osize,
+ fs->fs_dbsize);
+#else SECSIZE
+ error = bread(ip->i_devvp, fsbtodb(fs, bno),
+ osize, NOCRED, &bp);
+ if (error) {
brelse(bp);
- return (NULL);
+ return (error);
}
} while (brealloc(bp, nsize) == 0);
bp->b_flags |= B_DONE;
- bzero(bp->b_un.b_addr + osize, nsize - osize);
- return (bp);
+ bzero(bp->b_un.b_addr + osize, (unsigned)nsize - osize);
+ ip->i_blocks += btodb(nsize - osize);
+ ip->i_flag |= IUPD|ICHG;
+ *bpp = bp;
+ return (0);
}
if (bpref >= fs->fs_size)
bpref = 0;
- bno = (daddr_t)hashalloc(ip, cg, (long)bpref, nsize, 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
+ * `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("realloccg: bad optim");
+ /* NOTREACHED */
+ }
+ bno = (daddr_t)hashalloc(ip, cg, (long)bpref, request,
+ (u_long (*)())alloccg);
if (bno > 0) {
- obp = bread(ip->i_dev, fsbtodb(fs, bprev), osize);
- if (obp->b_flags & B_ERROR) {
+#ifdef SECSIZE
+ obp = bread(ip->i_dev, fsbtodb(fs, bprev), osize,
+ fs->fs_dbsize);
+#else SECSIZE
+ error = bread(ip->i_devvp, fsbtodb(fs, bprev),
+ osize, NOCRED, &obp);
+ if (error) {
brelse(obp);
- return (NULL);
+ return (error);
}
- bp = getblk(ip->i_dev, fsbtodb(fs, bno), nsize);
+ bn = fsbtodb(fs, bno);
+#ifdef SECSIZE
+ bp = getblk(ip->i_dev, bn, nsize, fs->fs_dbsize);
+#else SECSIZE
+ bp = getblk(ip->i_devvp, bn, nsize);
+#endif SECSIZE
bcopy(obp->b_un.b_addr, bp->b_un.b_addr, (u_int)osize);
- bzero(bp->b_un.b_addr + osize, nsize - osize);
+ 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
+ bzero(bp->b_un.b_addr + osize, (unsigned)nsize - osize);
+ if (obp->b_flags & B_DELWRI) {
+ obp->b_flags &= ~B_DELWRI;
+ u.u_ru.ru_oublock--; /* delete charge */
+ }
brelse(obp);
- fre(ip, bprev, (off_t)osize);
- return (bp);
+ blkfree(ip, bprev, (off_t)osize);
+ if (nsize < request)
+ blkfree(ip, bno + numfrags(fs, nsize),
+ (off_t)(request - nsize));
+ ip->i_blocks += btodb(nsize - osize);
+ ip->i_flag |= IUPD|ICHG;
+ *bpp = bp;
+ return (0);
}
nospace:
/*
*/
fserr(fs, "file system full");
uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
- u.u_error = ENOSPC;
- return (NULL);
+ return (ENOSPC);
}
/*
* 2) quadradically rehash into other cylinder groups, until an
* available inode is located.
*/
-struct inode *
-ialloc(pip, ipref, mode)
+ialloc(pip, ipref, mode, ipp)
register struct inode *pip;
ino_t ipref;
int mode;
+ struct inode **ipp;
{
ino_t ino;
register struct fs *fs;
register struct inode *ip;
- int cg;
+ int cg, error;
+ *ipp = 0;
fs = pip->i_fs;
if (fs->fs_cstotal.cs_nifree == 0)
goto noinodes;
#ifdef QUOTA
- if (chkiq(pip->i_dev, NULL, u.u_uid, 0))
- return(NULL);
+ if (error = chkiq(pip->i_dev, (struct inode *)NULL, u.u_uid, 0))
+ return (error);
#endif
if (ipref >= fs->fs_ncg * fs->fs_ipg)
ipref = 0;
ino = (ino_t)hashalloc(pip, cg, (long)ipref, mode, 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 = iget(pip, ino, ipp);
+ if (error) {
+ ifree(pip, ino, 0);
+ return (error);
}
+ ip = *ipp;
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);
+ 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;
+ }
+ 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");
uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
- u.u_error = ENOSPC;
- return (NULL);
+ return (ENOSPC);
}
/*
* among those cylinder groups with above the average number of
* free inodes, the one with the smallest number of directories.
*/
+ino_t
dirpref(fs)
register struct fs *fs;
{
mincg = cg;
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.
- *
- * The policy implemented by this algorithm is to maintain a
- * rotor that sweeps the cylinder groups. When a block is
- * needed, the rotor is advanced until a cylinder group with
- * greater than the average number of free blocks is found.
+ * 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 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
+ * 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.
*/
daddr_t
-blkpref(fs)
- register struct fs *fs;
+blkpref(ip, lbn, indx, bap)
+ struct inode *ip;
+ daddr_t lbn;
+ int indx;
+ daddr_t *bap;
{
- int cg, avgbfree;
+ register struct fs *fs;
+ register int cg;
+ int avgbfree, startcg;
+ daddr_t nextblk;
- 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;
+ 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 (NULL);
+ /*
+ * 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 = 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 <= startcg; 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] + blkstofrags(fs, 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);
}
/*
}
/*
* 3: brute force search
+ * 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 = 0; 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);
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_bsize);
+#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);
}
* 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--;
* Check to see if a block of the apprpriate size is available,
* and if it is, allocate it.
*/
-u_long
+daddr_t
alloccg(ip, cg, bpref, size)
struct inode *ip;
int cg;
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_bsize);
+#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) ||
+ (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) {
brelse(bp);
return (NULL);
}
bno = 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++;
cgp->cg_frsum[i]++;
bdwrite(bp);
return (bno);
}
bno = mapsearch(fs, cgp, bpref, allocsiz);
- if (bno < 0)
+ 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++;
cgp->cg_frsum[allocsiz]--;
if (frags != allocsiz)
cgp->cg_frsum[allocsiz - frags]++;
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
*/
-/*
- * disallow sequential layout.
- *
- if (isblock(fs, cgp->cg_free, bpref/fs->fs_frag)) {
+ if (isblock(fs, cg_blksfree(cgp), fragstoblks(fs, bpref))) {
bno = bpref;
goto gotit;
}
- */
/*
* 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) {
/*
bpref = howmany(fs->fs_spc * cylno, NSPF(fs));
goto norot;
}
- /*
- * find a block that is rotationally optimal
- */
- cylbp = cgp->cg_b[cylno];
- if (fs->fs_rotdelay == 0) {
- pos = cbtorpos(fs, bpref);
- } else {
- /*
- * 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 rotational position
- */
- bpref += fs->fs_rotdelay * fs->fs_rps * fs->fs_nsect /
- (NSPF(fs) * 1000);
- pos = cbtorpos(fs, bpref);
- pos = (pos + 1) % NRPOS;
- }
/*
* check the summary information to see if a block is
* available in the requested cylinder starting at the
- * optimal rotational position and proceeding around.
+ * requested rotational position and proceeding around.
*/
- for (i = pos; i < NRPOS; i++)
+ cylbp = cg_blks(fs, cgp, cylno);
+ pos = cbtorpos(fs, bpref);
+ 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;
*/
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");
}
- 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 (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;
}
return (NULL);
cgp->cg_rotor = bno;
gotit:
- clrblock(fs, cgp->cg_free, (long)(bno/fs->fs_frag));
+ 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]--;
+ cg_blks(fs, cgp, cylno)[cbtorpos(fs, bno)]--;
+ cg_blktot(cgp)[cylno]--;
fs->fs_fmod++;
return (cgp->cg_cgx * fs->fs_fpg + bno);
}
-
+
/*
* Determine whether an inode can be allocated.
*
* 2) allocate the next available inode after the requested
* inode in the specified cylinder group.
*/
-u_long
+ino_t
ialloccg(ip, cg, ipref, mode)
struct inode *ip;
int cg;
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_bsize);
+#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) || 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("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("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--;
* free map. If a fragment is deallocated, a possible
* block reassembly is checked.
*/
-fre(ip, bno, size)
+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;
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)) {
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);
+#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;
+ }
+#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 (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);
+ 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);
+ blk = blkmap(fs, cg_blksfree(cgp), bbase);
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;
/*
* add back in counts associated with the new frags
*/
- blk = blkmap(fs, cgp->cg_free, bbase);
+ blk = blkmap(fs, cg_blksfree(cgp), bbase);
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 (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;
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++;
{
register struct fs *fs;
register struct cg *cgp;
- register struct buf *bp;
- int cg;
+ struct buf *bp;
+ int error, cg;
fs = ip->i_fs;
if ((unsigned)ino >= fs->fs_ipg*fs->fs_ncg) {
panic("ifree: range");
}
cg = itog(fs, ino);
- bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_bsize);
+#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;
+ }
+#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");
}
- 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++;
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)));
+ 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("alloccg: map corrupted");
- return (-1);
+ /* NOTREACHED */
}
}
bno = (start + len - loc) * NBBY;
* 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];
return (-1);
}
-/*
- * 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 super block magic number 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)
- continue;
- fs = mp->m_bufp->b_un.b_fs;
- if (fs->fs_magic != FS_MAGIC) {
- printf("dev = 0x%x, fs = %s\n", dev, fs->fs_fsmnt);
- panic("getfs: bad magic");
- }
- return (fs);
- }
- printf("dev = 0x%x\n", dev);
- panic("getfs: no fs");
- return (NULL);
-}
-
/*
* Fserr prints the name of a file system with an error diagnostic.
*
char *cp;
{
- printf("%s: %s\n", fs->fs_fsmnt, cp);
+ log(LOG_ERR, "%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
- * <getfsx(dev),blkno>
- * rather than
- * <dev, blkno>
- * 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
- * the writing of the modified super blocks.
- */
-update()
-{
- register struct inode *ip;
- register struct mount *mp;
- struct fs *fs;
-
- 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)
- continue;
- fs = mp->m_bufp->b_un.b_fs;
- if (fs->fs_fmod == 0)
- continue;
- if (fs->fs_ronly != 0) { /* XXX */
- printf("fs = %s\n", fs->fs_fsmnt);
- panic("update: rofs mod");
- }
- fs->fs_fmod = 0;
- fs->fs_time = time.tv_sec;
- sbupdate(mp);
- }
- /*
- * Write back each (modified) inode.
- */
- for (ip = inode; ip < inodeNINODE; ip++) {
- if ((ip->i_flag & ILOCKED) != 0 || ip->i_count == 0)
- continue;
- ip->i_flag |= ILOCKED;
- ip->i_count++;
- iupdat(ip, &time, &time, 0);
- iput(ip);
- }
- updlock = 0;
- /*
- * Force stale buffer cache information to be flushed,
- * for all devices.
- */
- bflush(NODEV);
-}
-