2001-12-19 13:05:20 +03:00
|
|
|
/* $NetBSD: setup.c,v 1.52 2001/12/19 10:05:20 fvdl Exp $ */
|
1995-03-18 17:54:19 +03:00
|
|
|
|
1993-03-21 12:45:37 +03:00
|
|
|
/*
|
1994-06-08 22:57:30 +04:00
|
|
|
* Copyright (c) 1980, 1986, 1993
|
|
|
|
* The Regents of the University of California. All rights reserved.
|
1993-03-21 12:45:37 +03:00
|
|
|
*
|
|
|
|
* Redistribution and use in source and binary forms, with or without
|
|
|
|
* modification, are permitted provided that the following conditions
|
|
|
|
* are met:
|
|
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer.
|
|
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
|
|
* documentation and/or other materials provided with the distribution.
|
|
|
|
* 3. All advertising materials mentioning features or use of this software
|
|
|
|
* must display the following acknowledgement:
|
|
|
|
* This product includes software developed by the University of
|
|
|
|
* California, Berkeley and its contributors.
|
|
|
|
* 4. Neither the name of the University nor the names of its contributors
|
|
|
|
* may be used to endorse or promote products derived from this software
|
|
|
|
* without specific prior written permission.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
|
|
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
|
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
|
|
* SUCH DAMAGE.
|
|
|
|
*/
|
|
|
|
|
1997-09-14 18:36:29 +04:00
|
|
|
#include <sys/cdefs.h>
|
1993-03-21 12:45:37 +03:00
|
|
|
#ifndef lint
|
1995-03-18 17:54:19 +03:00
|
|
|
#if 0
|
1997-09-16 20:44:43 +04:00
|
|
|
static char sccsid[] = "@(#)setup.c 8.10 (Berkeley) 5/9/95";
|
1995-03-18 17:54:19 +03:00
|
|
|
#else
|
2001-12-19 13:05:20 +03:00
|
|
|
__RCSID("$NetBSD: setup.c,v 1.52 2001/12/19 10:05:20 fvdl Exp $");
|
1995-03-18 17:54:19 +03:00
|
|
|
#endif
|
1993-03-21 12:45:37 +03:00
|
|
|
#endif /* not lint */
|
|
|
|
|
|
|
|
#include <sys/param.h>
|
1994-04-25 22:28:42 +04:00
|
|
|
#include <sys/time.h>
|
1993-03-21 12:45:37 +03:00
|
|
|
#include <sys/stat.h>
|
|
|
|
#include <sys/ioctl.h>
|
1998-11-12 19:19:47 +03:00
|
|
|
#define FSTYPENAMES
|
1993-03-21 12:45:37 +03:00
|
|
|
#include <sys/disklabel.h>
|
|
|
|
#include <sys/file.h>
|
1994-12-05 23:15:31 +03:00
|
|
|
|
1997-09-16 20:44:43 +04:00
|
|
|
#include <ufs/ufs/dinode.h>
|
1998-03-18 20:01:23 +03:00
|
|
|
#include <ufs/ufs/ufs_bswap.h>
|
1997-09-16 20:44:43 +04:00
|
|
|
#include <ufs/ffs/fs.h>
|
1998-03-18 20:01:23 +03:00
|
|
|
#include <ufs/ffs/ffs_extern.h>
|
1997-09-16 20:44:43 +04:00
|
|
|
|
|
|
|
#include <ctype.h>
|
|
|
|
#include <err.h>
|
1993-03-21 12:45:37 +03:00
|
|
|
#include <errno.h>
|
1997-09-20 10:16:23 +04:00
|
|
|
#include <stdio.h>
|
|
|
|
#include <stdlib.h>
|
1993-03-21 12:45:37 +03:00
|
|
|
#include <string.h>
|
1996-09-23 20:18:31 +04:00
|
|
|
|
1993-03-21 12:45:37 +03:00
|
|
|
#include "fsck.h"
|
1994-12-05 23:15:31 +03:00
|
|
|
#include "extern.h"
|
1996-09-28 02:45:10 +04:00
|
|
|
#include "fsutil.h"
|
1993-03-21 12:45:37 +03:00
|
|
|
|
|
|
|
#define POWEROF2(num) (((num) & ((num) - 1)) == 0)
|
|
|
|
|
1997-09-16 20:44:43 +04:00
|
|
|
static void badsb __P((int, char *));
|
1998-07-27 00:32:42 +04:00
|
|
|
static int calcsb __P((const char *, int, struct fs *));
|
|
|
|
static struct disklabel *getdisklabel __P((const char *, int));
|
1996-09-23 20:18:31 +04:00
|
|
|
static int readsb __P((int));
|
1993-03-21 12:45:37 +03:00
|
|
|
|
1997-09-16 20:44:43 +04:00
|
|
|
/*
|
|
|
|
* Read in a superblock finding an alternate if necessary.
|
|
|
|
* Return 1 if successful, 0 if unsuccessful, -1 if filesystem
|
|
|
|
* is already clean (preen mode only).
|
|
|
|
*/
|
1994-12-05 23:15:31 +03:00
|
|
|
int
|
1993-03-21 12:45:37 +03:00
|
|
|
setup(dev)
|
1998-07-27 00:32:42 +04:00
|
|
|
const char *dev;
|
1993-03-21 12:45:37 +03:00
|
|
|
{
|
|
|
|
long cg, size, asked, i, j;
|
|
|
|
long bmapsize;
|
|
|
|
struct disklabel *lp;
|
1994-06-08 22:57:30 +04:00
|
|
|
off_t sizepb;
|
|
|
|
struct stat statb;
|
1993-03-21 12:45:37 +03:00
|
|
|
struct fs proto;
|
1995-04-13 01:24:07 +04:00
|
|
|
int doskipclean;
|
1996-05-21 21:25:56 +04:00
|
|
|
u_int64_t maxfilesize;
|
2001-09-02 05:58:30 +04:00
|
|
|
struct csum *ccsp;
|
1993-03-21 12:45:37 +03:00
|
|
|
|
|
|
|
havesb = 0;
|
1994-06-08 22:57:30 +04:00
|
|
|
fswritefd = -1;
|
1995-04-13 01:24:07 +04:00
|
|
|
doskipclean = skipclean;
|
1993-03-21 12:45:37 +03:00
|
|
|
if (stat(dev, &statb) < 0) {
|
|
|
|
printf("Can't stat %s: %s\n", dev, strerror(errno));
|
|
|
|
return (0);
|
|
|
|
}
|
2001-11-16 08:35:40 +03:00
|
|
|
if (!forceimage && !S_ISCHR(statb.st_mode)) {
|
1993-03-21 12:45:37 +03:00
|
|
|
pfatal("%s is not a character device", dev);
|
|
|
|
if (reply("CONTINUE") == 0)
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
if ((fsreadfd = open(dev, O_RDONLY)) < 0) {
|
|
|
|
printf("Can't open %s: %s\n", dev, strerror(errno));
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
if (preen == 0)
|
|
|
|
printf("** %s", dev);
|
|
|
|
if (nflag || (fswritefd = open(dev, O_WRONLY)) < 0) {
|
|
|
|
fswritefd = -1;
|
|
|
|
if (preen)
|
|
|
|
pfatal("NO WRITE ACCESS");
|
|
|
|
printf(" (NO WRITE)");
|
|
|
|
}
|
|
|
|
if (preen == 0)
|
|
|
|
printf("\n");
|
|
|
|
fsmodified = 0;
|
|
|
|
lfdir = 0;
|
|
|
|
initbarea(&sblk);
|
|
|
|
initbarea(&asblk);
|
|
|
|
sblk.b_un.b_buf = malloc(SBSIZE);
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock = malloc(SBSIZE);
|
1993-03-21 12:45:37 +03:00
|
|
|
asblk.b_un.b_buf = malloc(SBSIZE);
|
1998-03-18 20:01:23 +03:00
|
|
|
altsblock = malloc(SBSIZE);
|
|
|
|
if (sblk.b_un.b_buf == NULL || asblk.b_un.b_buf == NULL ||
|
|
|
|
sblock == NULL || altsblock == NULL)
|
1997-09-16 20:44:43 +04:00
|
|
|
errx(EEXIT, "cannot allocate space for superblock");
|
2001-11-16 08:35:40 +03:00
|
|
|
if (!forceimage && (lp = getdisklabel(NULL, fsreadfd)) != NULL)
|
1993-03-21 12:45:37 +03:00
|
|
|
dev_bsize = secsize = lp->d_secsize;
|
|
|
|
else
|
|
|
|
dev_bsize = secsize = DEV_BSIZE;
|
|
|
|
/*
|
|
|
|
* Read in the superblock, looking for alternates if necessary
|
|
|
|
*/
|
|
|
|
if (readsb(1) == 0) {
|
2001-11-16 08:35:40 +03:00
|
|
|
if (bflag || preen || forceimage ||
|
|
|
|
calcsb(dev, fsreadfd, &proto) == 0)
|
1993-03-21 12:45:37 +03:00
|
|
|
return(0);
|
|
|
|
if (reply("LOOK FOR ALTERNATE SUPERBLOCKS") == 0)
|
|
|
|
return (0);
|
|
|
|
for (cg = 0; cg < proto.fs_ncg; cg++) {
|
|
|
|
bflag = fsbtodb(&proto, cgsblock(&proto, cg));
|
|
|
|
if (readsb(0) != 0)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (cg >= proto.fs_ncg) {
|
|
|
|
printf("%s %s\n%s %s\n%s %s\n",
|
|
|
|
"SEARCH FOR ALTERNATE SUPER-BLOCK",
|
|
|
|
"FAILED. YOU MUST USE THE",
|
2001-07-05 02:43:35 +04:00
|
|
|
"-b OPTION TO fsck_ffs TO SPECIFY THE",
|
1993-03-21 12:45:37 +03:00
|
|
|
"LOCATION OF AN ALTERNATE",
|
|
|
|
"SUPER-BLOCK TO SUPPLY NEEDED",
|
1996-04-05 05:44:24 +04:00
|
|
|
"INFORMATION; SEE fsck_ffs(8).");
|
1993-03-21 12:45:37 +03:00
|
|
|
return(0);
|
|
|
|
}
|
1995-04-13 01:24:07 +04:00
|
|
|
doskipclean = 0;
|
1993-03-21 12:45:37 +03:00
|
|
|
pwarn("USING ALTERNATE SUPERBLOCK AT %d\n", bflag);
|
|
|
|
}
|
1995-04-13 01:24:07 +04:00
|
|
|
if (debug)
|
1998-03-18 20:01:23 +03:00
|
|
|
printf("clean = %d\n", sblock->fs_clean);
|
|
|
|
if (doswap)
|
|
|
|
doskipclean = 0;
|
|
|
|
if (sblock->fs_clean & FS_ISCLEAN) {
|
implement a 'force check' flag, '-f'. I used the SunOS name, but the Digital
semantics. now:
(1) dirty file systems will always be checked; nothing new there.
(2) if not '-f' clean file systems will _NEVER_ be checked,
i.e. they won't be checked even if -p isn't specified. This
allows one to 'fsck -p ; fsck' to preen, then clean up
anything that 'fsck -p' barfs on, without waiting for the
clean file systems to be checked again.
(3) if '-f' clean file systems will ALWAYS be checked. This
allows people to put 'fsck -fp' into /etc/rc on systems
where they're leery of the FS clean flag state, need
the extra reliability, and can afford time 'wasted'
in checks.
The assumption made here is that if a file system is marked clean, it
_IS CLEAN_, really, and shouldn't be checked unless fsck is explicitly
told to (with -f). This should be a valid assumption, but may not be in
the presence of file system bugs. Documentation updated to note '-f'.
1995-07-12 05:49:16 +04:00
|
|
|
if (doskipclean) {
|
|
|
|
pwarn("%sile system is clean; not checking\n",
|
|
|
|
preen ? "f" : "** F");
|
1995-04-13 01:24:07 +04:00
|
|
|
return (-1);
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (!preen && !doswap)
|
1995-04-13 01:24:07 +04:00
|
|
|
pwarn("** File system is already clean\n");
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
maxfsblock = sblock->fs_size;
|
|
|
|
maxino = sblock->fs_ncg * sblock->fs_ipg;
|
|
|
|
sizepb = sblock->fs_bsize;
|
|
|
|
maxfilesize = sblock->fs_bsize * NDADDR - 1;
|
1996-05-21 21:25:56 +04:00
|
|
|
for (i = 0; i < NIADDR; i++) {
|
1998-03-18 20:01:23 +03:00
|
|
|
sizepb *= NINDIR(sblock);
|
1996-05-21 21:25:56 +04:00
|
|
|
maxfilesize += sizepb;
|
|
|
|
}
|
1993-03-21 12:45:37 +03:00
|
|
|
/*
|
|
|
|
* Check and potentially fix certain fields in the super block.
|
|
|
|
*/
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_optim != FS_OPTTIME && sblock->fs_optim != FS_OPTSPACE) {
|
1993-03-21 12:45:37 +03:00
|
|
|
pfatal("UNDEFINED OPTIMIZATION IN SUPERBLOCK");
|
|
|
|
if (reply("SET TO DEFAULT") == 1) {
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_optim = FS_OPTTIME;
|
1993-03-21 12:45:37 +03:00
|
|
|
sbdirty();
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if ((sblock->fs_minfree < 0 || sblock->fs_minfree > 99)) {
|
1993-03-21 12:45:37 +03:00
|
|
|
pfatal("IMPOSSIBLE MINFREE=%d IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_minfree);
|
1993-03-21 12:45:37 +03:00
|
|
|
if (reply("SET TO DEFAULT") == 1) {
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_minfree = 10;
|
1993-03-21 12:45:37 +03:00
|
|
|
sbdirty();
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_interleave < 1 ||
|
|
|
|
sblock->fs_interleave > sblock->fs_nsect) {
|
1993-03-21 12:45:37 +03:00
|
|
|
pwarn("IMPOSSIBLE INTERLEAVE=%d IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_interleave);
|
|
|
|
sblock->fs_interleave = 1;
|
1993-03-21 12:45:37 +03:00
|
|
|
if (preen)
|
|
|
|
printf(" (FIXED)\n");
|
|
|
|
if (preen || reply("SET TO DEFAULT") == 1) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_npsect < sblock->fs_nsect ||
|
|
|
|
sblock->fs_npsect > sblock->fs_nsect*2) {
|
1993-03-21 12:45:37 +03:00
|
|
|
pwarn("IMPOSSIBLE NPSECT=%d IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_npsect);
|
|
|
|
sblock->fs_npsect = sblock->fs_nsect;
|
1993-03-21 12:45:37 +03:00
|
|
|
if (preen)
|
|
|
|
printf(" (FIXED)\n");
|
|
|
|
if (preen || reply("SET TO DEFAULT") == 1) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_bmask != ~(sblock->fs_bsize - 1)) {
|
1997-09-14 18:36:29 +04:00
|
|
|
pwarn("INCORRECT BMASK=0x%x IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_bmask);
|
|
|
|
sblock->fs_bmask = ~(sblock->fs_bsize - 1);
|
1996-05-21 21:25:56 +04:00
|
|
|
if (preen)
|
|
|
|
printf(" (FIXED)\n");
|
|
|
|
if (preen || reply("FIX") == 1) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_fmask != ~(sblock->fs_fsize - 1)) {
|
1997-09-14 18:36:29 +04:00
|
|
|
pwarn("INCORRECT FMASK=0x%x IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_fmask);
|
|
|
|
sblock->fs_fmask = ~(sblock->fs_fsize - 1);
|
1996-05-21 21:25:56 +04:00
|
|
|
if (preen)
|
|
|
|
printf(" (FIXED)\n");
|
|
|
|
if (preen || reply("FIX") == 1) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_inodefmt >= FS_44INODEFMT) {
|
|
|
|
if (sblock->fs_maxfilesize != maxfilesize) {
|
2001-01-05 05:02:57 +03:00
|
|
|
pwarn("INCORRECT MAXFILESIZE=%lld IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
(unsigned long long)sblock->fs_maxfilesize);
|
|
|
|
sblock->fs_maxfilesize = maxfilesize;
|
1996-05-21 21:25:56 +04:00
|
|
|
if (preen)
|
|
|
|
printf(" (FIXED)\n");
|
|
|
|
if (preen || reply("FIX") == 1) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_maxsymlinklen != MAXSYMLINKLEN) {
|
1996-05-21 21:25:56 +04:00
|
|
|
pwarn("INCORRECT MAXSYMLINKLEN=%d IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_maxsymlinklen);
|
|
|
|
sblock->fs_maxsymlinklen = MAXSYMLINKLEN;
|
1996-05-21 21:25:56 +04:00
|
|
|
if (preen)
|
|
|
|
printf(" (FIXED)\n");
|
|
|
|
if (preen || reply("FIX") == 1) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_qbmask != ~sblock->fs_bmask) {
|
2001-01-05 05:02:57 +03:00
|
|
|
pwarn("INCORRECT QBMASK=%llx IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
(unsigned long long)sblock->fs_qbmask);
|
|
|
|
sblock->fs_qbmask = ~sblock->fs_bmask;
|
1996-05-21 21:25:56 +04:00
|
|
|
if (preen)
|
|
|
|
printf(" (FIXED)\n");
|
|
|
|
if (preen || reply("FIX") == 1) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_qfmask != ~sblock->fs_fmask) {
|
2001-01-05 05:02:57 +03:00
|
|
|
pwarn("INCORRECT QFMASK=%llx IN SUPERBLOCK",
|
1998-03-18 20:01:23 +03:00
|
|
|
(unsigned long long)sblock->fs_qfmask);
|
|
|
|
sblock->fs_qfmask = ~sblock->fs_fmask;
|
1996-05-21 21:25:56 +04:00
|
|
|
if (preen)
|
|
|
|
printf(" (FIXED)\n");
|
|
|
|
if (preen || reply("FIX") == 1) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
}
|
1994-06-08 22:57:30 +04:00
|
|
|
newinofmt = 1;
|
|
|
|
} else {
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_qbmask = ~sblock->fs_bmask;
|
|
|
|
sblock->fs_qfmask = ~sblock->fs_fmask;
|
1994-06-08 22:57:30 +04:00
|
|
|
newinofmt = 0;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Convert to new inode format.
|
|
|
|
*/
|
1998-03-18 20:01:23 +03:00
|
|
|
if (cvtlevel >= 2 && sblock->fs_inodefmt < FS_44INODEFMT) {
|
1994-06-08 22:57:30 +04:00
|
|
|
if (preen)
|
|
|
|
pwarn("CONVERTING TO NEW INODE FORMAT\n");
|
|
|
|
else if (!reply("CONVERT TO NEW INODE FORMAT"))
|
|
|
|
return(0);
|
|
|
|
doinglevel2++;
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_inodefmt = FS_44INODEFMT;
|
|
|
|
sblock->fs_maxfilesize = maxfilesize;
|
|
|
|
sblock->fs_maxsymlinklen = MAXSYMLINKLEN;
|
|
|
|
sblock->fs_qbmask = ~sblock->fs_bmask;
|
|
|
|
sblock->fs_qfmask = ~sblock->fs_fmask;
|
1994-06-08 22:57:30 +04:00
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Convert to new cylinder group format.
|
|
|
|
*/
|
1998-03-18 20:01:23 +03:00
|
|
|
if (cvtlevel >= 1 && sblock->fs_postblformat == FS_42POSTBLFMT) {
|
1994-06-08 22:57:30 +04:00
|
|
|
if (preen)
|
|
|
|
pwarn("CONVERTING TO NEW CYLINDER GROUP FORMAT\n");
|
|
|
|
else if (!reply("CONVERT TO NEW CYLINDER GROUP FORMAT"))
|
|
|
|
return(0);
|
|
|
|
doinglevel1++;
|
1998-03-18 20:01:23 +03:00
|
|
|
sblock->fs_postblformat = FS_DYNAMICPOSTBLFMT;
|
|
|
|
sblock->fs_nrpos = 8;
|
|
|
|
sblock->fs_postbloff =
|
|
|
|
(char *)(&sblock->fs_opostbl[0][0]) -
|
|
|
|
(char *)(&sblock->fs_firstfield);
|
|
|
|
sblock->fs_rotbloff = &sblock->fs_space[0] -
|
|
|
|
(u_char *)(&sblock->fs_firstfield);
|
|
|
|
sblock->fs_cgsize =
|
|
|
|
fragroundup(sblock, CGSIZE(sblock));
|
1994-06-08 22:57:30 +04:00
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
1993-03-21 12:45:37 +03:00
|
|
|
}
|
1994-06-29 15:01:35 +04:00
|
|
|
if (asblk.b_dirty && !bflag) {
|
1998-03-18 20:01:23 +03:00
|
|
|
memmove((struct fs*)sblk.b_un.b_fs, sblock, SBSIZE);
|
|
|
|
if (needswap)
|
2001-08-17 06:18:46 +04:00
|
|
|
ffs_sb_swap(sblock, (struct fs*)sblk.b_un.b_fs);
|
1998-03-18 20:01:23 +03:00
|
|
|
memmove(asblk.b_un.b_fs, sblk.b_un.b_fs, (size_t)sblock->fs_sbsize);
|
1993-03-21 12:45:37 +03:00
|
|
|
flush(fswritefd, &asblk);
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* read in the summary info.
|
|
|
|
*/
|
|
|
|
asked = 0;
|
2001-09-02 05:58:30 +04:00
|
|
|
sblock->fs_csp = (struct csum *)calloc(1, sblock->fs_cssize);
|
1998-03-18 20:01:23 +03:00
|
|
|
for (i = 0, j = 0; i < sblock->fs_cssize; i += sblock->fs_bsize, j++) {
|
|
|
|
size = sblock->fs_cssize - i < sblock->fs_bsize ?
|
|
|
|
sblock->fs_cssize - i : sblock->fs_bsize;
|
2001-09-02 05:58:30 +04:00
|
|
|
ccsp = (struct csum *)((char *)sblock->fs_csp + i);
|
|
|
|
if (bread(fsreadfd, (char *)ccsp,
|
1998-03-18 20:01:23 +03:00
|
|
|
fsbtodb(sblock, sblock->fs_csaddr + j * sblock->fs_frag),
|
1993-03-21 12:45:37 +03:00
|
|
|
size) != 0 && !asked) {
|
|
|
|
pfatal("BAD SUMMARY INFORMATION");
|
1999-11-15 22:18:24 +03:00
|
|
|
if (reply("CONTINUE") == 0) {
|
|
|
|
markclean = 0;
|
1997-09-16 20:44:43 +04:00
|
|
|
exit(EEXIT);
|
1999-11-15 22:18:24 +03:00
|
|
|
}
|
1993-03-21 12:45:37 +03:00
|
|
|
asked++;
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
if (doswap) {
|
2001-09-02 05:58:30 +04:00
|
|
|
ffs_csum_swap(ccsp, ccsp, size);
|
|
|
|
bwrite(fswritefd, (char *)ccsp,
|
1999-11-15 22:18:24 +03:00
|
|
|
fsbtodb(sblock,
|
|
|
|
sblock->fs_csaddr + j * sblock->fs_frag),
|
|
|
|
size);
|
1998-03-18 20:01:23 +03:00
|
|
|
}
|
2001-09-02 05:58:30 +04:00
|
|
|
if (needswap)
|
|
|
|
ffs_csum_swap(ccsp, ccsp, size);
|
1993-03-21 12:45:37 +03:00
|
|
|
}
|
|
|
|
/*
|
|
|
|
* allocate and initialize the necessary maps
|
|
|
|
*/
|
1995-03-21 04:30:09 +03:00
|
|
|
bmapsize = roundup(howmany(maxfsblock, NBBY), sizeof(int16_t));
|
1993-03-21 12:45:37 +03:00
|
|
|
blockmap = calloc((unsigned)bmapsize, sizeof (char));
|
|
|
|
if (blockmap == NULL) {
|
|
|
|
printf("cannot alloc %u bytes for blockmap\n",
|
|
|
|
(unsigned)bmapsize);
|
1994-12-05 23:15:31 +03:00
|
|
|
goto badsblabel;
|
1993-03-21 12:45:37 +03:00
|
|
|
}
|
|
|
|
statemap = calloc((unsigned)(maxino + 1), sizeof(char));
|
|
|
|
if (statemap == NULL) {
|
|
|
|
printf("cannot alloc %u bytes for statemap\n",
|
|
|
|
(unsigned)(maxino + 1));
|
1994-12-05 23:15:31 +03:00
|
|
|
goto badsblabel;
|
1993-03-21 12:45:37 +03:00
|
|
|
}
|
1994-06-08 22:57:30 +04:00
|
|
|
typemap = calloc((unsigned)(maxino + 1), sizeof(char));
|
|
|
|
if (typemap == NULL) {
|
|
|
|
printf("cannot alloc %u bytes for typemap\n",
|
|
|
|
(unsigned)(maxino + 1));
|
1994-12-05 23:15:31 +03:00
|
|
|
goto badsblabel;
|
1994-06-08 22:57:30 +04:00
|
|
|
}
|
1995-03-21 04:30:09 +03:00
|
|
|
lncntp = (int16_t *)calloc((unsigned)(maxino + 1), sizeof(int16_t));
|
1993-03-21 12:45:37 +03:00
|
|
|
if (lncntp == NULL) {
|
|
|
|
printf("cannot alloc %u bytes for lncntp\n",
|
1997-09-16 12:37:01 +04:00
|
|
|
(unsigned)((maxino + 1) * sizeof(int16_t)));
|
1994-12-05 23:15:31 +03:00
|
|
|
goto badsblabel;
|
1993-03-21 12:45:37 +03:00
|
|
|
}
|
2001-01-09 08:39:27 +03:00
|
|
|
/*
|
|
|
|
* cs_ndir may be inaccurate, particularly if we're using the -b
|
|
|
|
* option, so set a minimum to prevent bogus subdirectory reconnects
|
|
|
|
* and really inefficient directory scans.
|
|
|
|
* Also set a maximum in case the value is too large.
|
|
|
|
*/
|
1998-03-18 20:01:23 +03:00
|
|
|
numdirs = sblock->fs_cstotal.cs_ndir;
|
2001-01-09 08:39:27 +03:00
|
|
|
if (numdirs < 1024)
|
|
|
|
numdirs = 1024;
|
|
|
|
if (numdirs > maxino + 1)
|
|
|
|
numdirs = maxino + 1;
|
1993-03-21 12:45:37 +03:00
|
|
|
inplast = 0;
|
|
|
|
listmax = numdirs + 10;
|
|
|
|
inpsort = (struct inoinfo **)calloc((unsigned)listmax,
|
|
|
|
sizeof(struct inoinfo *));
|
|
|
|
inphead = (struct inoinfo **)calloc((unsigned)numdirs,
|
|
|
|
sizeof(struct inoinfo *));
|
|
|
|
if (inpsort == NULL || inphead == NULL) {
|
|
|
|
printf("cannot alloc %u bytes for inphead\n",
|
1997-09-16 12:37:01 +04:00
|
|
|
(unsigned)(numdirs * sizeof(struct inoinfo *)));
|
1994-12-05 23:15:31 +03:00
|
|
|
goto badsblabel;
|
1993-03-21 12:45:37 +03:00
|
|
|
}
|
1999-11-15 22:18:24 +03:00
|
|
|
cgrp = malloc(sblock->fs_cgsize);
|
|
|
|
if (cgrp == NULL) {
|
|
|
|
printf("cannot alloc %u bytes for cylinder group\n",
|
|
|
|
sblock->fs_cgsize);
|
|
|
|
goto badsblabel;
|
|
|
|
}
|
1993-03-21 12:45:37 +03:00
|
|
|
bufinit();
|
1999-11-15 22:18:24 +03:00
|
|
|
if (sblock->fs_flags & FS_DOSOFTDEP)
|
|
|
|
usedsoftdep = 1;
|
|
|
|
else
|
|
|
|
usedsoftdep = 0;
|
1993-03-21 12:45:37 +03:00
|
|
|
return (1);
|
|
|
|
|
1994-12-05 23:15:31 +03:00
|
|
|
badsblabel:
|
1998-03-18 20:01:23 +03:00
|
|
|
markclean=0;
|
|
|
|
ckfini();
|
1993-03-21 12:45:37 +03:00
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Read in the super block and its summary info.
|
|
|
|
*/
|
1996-09-23 20:18:31 +04:00
|
|
|
static int
|
1993-03-21 12:45:37 +03:00
|
|
|
readsb(listerr)
|
|
|
|
int listerr;
|
|
|
|
{
|
1997-09-16 20:44:43 +04:00
|
|
|
ufs_daddr_t super = bflag ? bflag : SBOFF / dev_bsize;
|
1998-03-18 20:01:23 +03:00
|
|
|
struct fs *fs;
|
1993-03-21 12:45:37 +03:00
|
|
|
|
1998-03-18 20:01:23 +03:00
|
|
|
if (bread(fsreadfd, (char *)sblk.b_un.b_fs, super, (long)SBSIZE) != 0)
|
1993-03-21 12:45:37 +03:00
|
|
|
return (0);
|
|
|
|
sblk.b_bno = super;
|
|
|
|
sblk.b_size = SBSIZE;
|
1998-03-18 20:01:23 +03:00
|
|
|
|
|
|
|
fs = sblk.b_un.b_fs;
|
|
|
|
/* auto detect byte order */
|
|
|
|
if( fs->fs_magic == FS_MAGIC) {
|
|
|
|
if (endian == 0 || BYTE_ORDER == endian) {
|
|
|
|
needswap = 0;
|
|
|
|
doswap = do_blkswap = do_dirswap = 0;
|
|
|
|
} else {
|
|
|
|
needswap = 1;
|
|
|
|
doswap = do_blkswap = do_dirswap = 1;
|
|
|
|
}
|
|
|
|
} else if (fs->fs_magic == bswap32(FS_MAGIC)) {
|
|
|
|
if (endian == 0 || BYTE_ORDER != endian) {
|
|
|
|
needswap = 1;
|
|
|
|
doswap = do_blkswap = do_dirswap = 0;
|
|
|
|
} else {
|
|
|
|
needswap = 0;
|
|
|
|
doswap = do_blkswap = do_dirswap = 1;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
badsb(listerr, "MAGIC NUMBER WRONG");
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
if (doswap) {
|
|
|
|
if (preen)
|
|
|
|
errx(EEXIT, "incompatible options -B and -p");
|
|
|
|
if (nflag)
|
|
|
|
errx(EEXIT, "incompatible options -B and -n");
|
|
|
|
if (endian == LITTLE_ENDIAN) {
|
1999-05-02 00:04:14 +04:00
|
|
|
if (!reply("CONVERT TO LITTLE ENDIAN"))
|
1998-03-18 20:01:23 +03:00
|
|
|
return 0;
|
|
|
|
} else if (endian == BIG_ENDIAN) {
|
1999-05-02 00:04:14 +04:00
|
|
|
if (!reply("CONVERT TO BIG ENDIAN"))
|
1998-03-18 20:01:23 +03:00
|
|
|
return 0;
|
|
|
|
} else
|
|
|
|
pfatal("INTERNAL ERROR: unknown endian");
|
|
|
|
}
|
|
|
|
if (needswap)
|
|
|
|
printf("** Swapped byte order\n");
|
|
|
|
/* swap SB byte order if asked */
|
|
|
|
if (doswap)
|
2001-08-17 06:18:46 +04:00
|
|
|
ffs_sb_swap(sblk.b_un.b_fs, sblk.b_un.b_fs);
|
1998-03-18 20:01:23 +03:00
|
|
|
|
|
|
|
memmove(sblock, sblk.b_un.b_fs, SBSIZE);
|
|
|
|
if (needswap)
|
2001-08-17 06:18:46 +04:00
|
|
|
ffs_sb_swap(sblk.b_un.b_fs, sblock);
|
1998-03-18 20:01:23 +03:00
|
|
|
|
1993-03-21 12:45:37 +03:00
|
|
|
/*
|
|
|
|
* run a few consistency checks of the super block
|
|
|
|
*/
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_ncg < 1)
|
1993-03-21 12:45:37 +03:00
|
|
|
{ badsb(listerr, "NCG OUT OF RANGE"); return (0); }
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_cpg < 1)
|
1993-03-21 12:45:37 +03:00
|
|
|
{ badsb(listerr, "CPG OUT OF RANGE"); return (0); }
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_ncg * sblock->fs_cpg < sblock->fs_ncyl ||
|
|
|
|
(sblock->fs_ncg - 1) * sblock->fs_cpg >= sblock->fs_ncyl)
|
1993-03-21 12:45:37 +03:00
|
|
|
{ badsb(listerr, "NCYL LESS THAN NCG*CPG"); return (0); }
|
1998-03-18 20:01:23 +03:00
|
|
|
if (sblock->fs_sbsize > SBSIZE)
|
1993-03-21 12:45:37 +03:00
|
|
|
{ badsb(listerr, "SIZE PREPOSTEROUSLY LARGE"); return (0); }
|
|
|
|
/*
|
|
|
|
* Compute block size that the filesystem is based on,
|
|
|
|
* according to fsbtodb, and adjust superblock block number
|
|
|
|
* so we can tell if this is an alternate later.
|
|
|
|
*/
|
|
|
|
super *= dev_bsize;
|
1998-03-18 20:01:23 +03:00
|
|
|
dev_bsize = sblock->fs_fsize / fsbtodb(sblock, 1);
|
1993-03-21 12:45:37 +03:00
|
|
|
sblk.b_bno = super / dev_bsize;
|
1998-03-18 20:01:23 +03:00
|
|
|
|
1994-06-08 22:57:30 +04:00
|
|
|
if (bflag) {
|
|
|
|
havesb = 1;
|
|
|
|
return (1);
|
|
|
|
}
|
1993-03-21 12:45:37 +03:00
|
|
|
/*
|
|
|
|
* Set all possible fields that could differ, then do check
|
|
|
|
* of whole super block against an alternate super block.
|
|
|
|
* When an alternate super-block is specified this check is skipped.
|
|
|
|
*/
|
1998-03-18 20:01:23 +03:00
|
|
|
getblk(&asblk, cgsblock(sblock, sblock->fs_ncg - 1), sblock->fs_sbsize);
|
1993-03-21 12:45:37 +03:00
|
|
|
if (asblk.b_errs)
|
|
|
|
return (0);
|
1998-03-18 20:01:23 +03:00
|
|
|
/* swap SB byte order if asked */
|
|
|
|
if (doswap)
|
2001-08-17 06:18:46 +04:00
|
|
|
ffs_sb_swap(asblk.b_un.b_fs, asblk.b_un.b_fs);
|
1998-03-18 20:01:23 +03:00
|
|
|
|
|
|
|
memmove(altsblock, asblk.b_un.b_fs, sblock->fs_sbsize);
|
|
|
|
if (needswap)
|
2001-08-17 06:18:46 +04:00
|
|
|
ffs_sb_swap(asblk.b_un.b_fs, altsblock);
|
2001-01-26 20:37:16 +03:00
|
|
|
if (cmpsblks(sblock, altsblock)) {
|
1994-12-28 03:03:49 +03:00
|
|
|
if (debug) {
|
|
|
|
long *nlp, *olp, *endlp;
|
|
|
|
|
|
|
|
printf("superblock mismatches\n");
|
1998-03-18 20:01:23 +03:00
|
|
|
nlp = (long *)altsblock;
|
|
|
|
olp = (long *)sblock;
|
|
|
|
endlp = olp + (sblock->fs_sbsize / sizeof *olp);
|
1994-12-28 03:03:49 +03:00
|
|
|
for ( ; olp < endlp; olp++, nlp++) {
|
|
|
|
if (*olp == *nlp)
|
|
|
|
continue;
|
1998-03-18 20:01:23 +03:00
|
|
|
printf("offset %ld, original %lx, alternate %lx\n",
|
|
|
|
(long)(olp - (long *)sblock), *olp, *nlp);
|
1994-12-28 03:03:49 +03:00
|
|
|
}
|
|
|
|
}
|
1993-03-21 12:45:37 +03:00
|
|
|
badsb(listerr,
|
|
|
|
"VALUES IN SUPER BLOCK DISAGREE WITH THOSE IN FIRST ALTERNATE");
|
|
|
|
return (0);
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
/* Now we know the SB is valid, we can write it back if needed */
|
|
|
|
if (doswap) {
|
|
|
|
sbdirty();
|
|
|
|
dirty(&asblk);
|
|
|
|
}
|
1993-03-21 12:45:37 +03:00
|
|
|
havesb = 1;
|
|
|
|
return (1);
|
|
|
|
}
|
|
|
|
|
2001-01-26 20:37:16 +03:00
|
|
|
int
|
|
|
|
cmpsblks(const struct fs *sb, struct fs *asb)
|
|
|
|
{
|
|
|
|
|
2001-09-18 12:38:28 +04:00
|
|
|
/*
|
|
|
|
* Copy fields which we don't care if they're different in the
|
|
|
|
* alternate superblocks, as they're either likely to be
|
|
|
|
* different because they're per-cylinder-group specific, or
|
|
|
|
* because they're transient details which are only maintained
|
|
|
|
* in the primary superblock.
|
|
|
|
*/
|
2001-01-26 20:37:16 +03:00
|
|
|
asb->fs_firstfield = sb->fs_firstfield;
|
|
|
|
asb->fs_unused_1 = sb->fs_unused_1;
|
|
|
|
asb->fs_time = sb->fs_time;
|
|
|
|
asb->fs_cstotal = sb->fs_cstotal;
|
|
|
|
asb->fs_cgrotor = sb->fs_cgrotor;
|
|
|
|
asb->fs_fmod = sb->fs_fmod;
|
|
|
|
asb->fs_clean = sb->fs_clean;
|
|
|
|
asb->fs_ronly = sb->fs_ronly;
|
|
|
|
asb->fs_flags = sb->fs_flags;
|
|
|
|
asb->fs_maxcontig = sb->fs_maxcontig;
|
|
|
|
asb->fs_minfree = sb->fs_minfree;
|
|
|
|
asb->fs_optim = sb->fs_optim;
|
|
|
|
asb->fs_rotdelay = sb->fs_rotdelay;
|
|
|
|
asb->fs_maxbpg = sb->fs_maxbpg;
|
2001-09-02 05:58:30 +04:00
|
|
|
memmove(asb->fs_ocsp, sb->fs_ocsp, sizeof sb->fs_ocsp);
|
Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory
The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.
The original FreeBSD commit messages are attached:
=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.
------
One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.
First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:
1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35
2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50
You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html
Test Results
tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66
"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.
------
Algorithm description
The old dirpref algorithm is described in comments:
/*
* 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.
*/
A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.
What I mean by a big file system ?
1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.
The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.
My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/
My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.
My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.
The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:
int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */
These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.
I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.
Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====
=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.
Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.
Reviewed by: mckusick
=====
=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
2001-09-06 06:16:00 +04:00
|
|
|
asb->fs_contigdirs = sb->fs_contigdirs;
|
2001-09-02 05:58:30 +04:00
|
|
|
asb->fs_csp = sb->fs_csp;
|
2001-01-26 20:37:16 +03:00
|
|
|
asb->fs_maxcluster = sb->fs_maxcluster;
|
|
|
|
memmove(asb->fs_fsmnt, sb->fs_fsmnt, sizeof sb->fs_fsmnt);
|
Incorporate the enhanced ffs_dirpref() by Grigoriy Orlov, as found in
FreeBSD (three commits; the initial work, man page updates, and a fix
to ffs_reload()), with the following differences:
- Be consistent between newfs(8) and tunefs(8) as to the options which
set and control the tuning parameters for this work (avgfilesize & avgfpdir)
- Use u_int16_t instead of u_int8_t to keep track of the number of
contiguous directories (suggested by Chuck Silvers)
- Work within our FFS_EI framework
- Ensure that fs->fs_maxclusters and fs->fs_contigdirs don't point to
the same area of memory
The new algorithm has a marked performance increase, especially when
performing tasks such as untarring pkgsrc.tar.gz, etc.
The original FreeBSD commit messages are attached:
=====
mckusick 2001/04/10 01:39:00 PDT
Directory layout preference improvements from Grigoriy Orlov <gluk@ptci.ru>.
His description of the problem and solution follow. My own tests show
speedups on typical filesystem intensive workloads of 5% to 12% which
is very impressive considering the small amount of code change involved.
------
One day I noticed that some file operations run much faster on
small file systems then on big ones. I've looked at the ffs
algorithms, thought about them, and redesigned the dirpref algorithm.
First I want to describe the results of my tests. These results are old
and I have improved the algorithm after these tests were done. Nevertheless
they show how big the perfomance speedup may be. I have done two file/directory
intensive tests on a two OpenBSD systems with old and new dirpref algorithm.
The first test is "tar -xzf ports.tar.gz", the second is "rm -rf ports".
The ports.tar.gz file is the ports collection from the OpenBSD 2.8 release.
It contains 6596 directories and 13868 files. The test systems are:
1. Celeron-450, 128Mb, two IDE drives, the system at wd0, file system for
test is at wd1. Size of test file system is 8 Gb, number of cg=991,
size of cg is 8m, block size = 8k, fragment size = 1k OpenBSD-current
from Dec 2000 with BUFCACHEPERCENT=35
2. PIII-600, 128Mb, two IBM DTLA-307045 IDE drives at i815e, the system
at wd0, file system for test is at wd1. Size of test file system is 40 Gb,
number of cg=5324, size of cg is 8m, block size = 8k, fragment size = 1k
OpenBSD-current from Dec 2000 with BUFCACHEPERCENT=50
You can get more info about the test systems and methods at:
http://www.ptci.ru/gluk/dirpref/old/dirpref.html
Test Results
tar -xzf ports.tar.gz rm -rf ports
mode old dirpref new dirpref speedup old dirprefnew dirpref speedup
First system
normal 667 472 1.41 477 331 1.44
async 285 144 1.98 130 14 9.29
sync 768 616 1.25 477 334 1.43
softdep 413 252 1.64 241 38 6.34
Second system
normal 329 81 4.06 263.5 93.5 2.81
async 302 25.7 11.75 112 2.26 49.56
sync 281 57.0 4.93 263 90.5 2.9
softdep 341 40.6 8.4 284 4.76 59.66
"old dirpref" and "new dirpref" columns give a test time in seconds.
speedup - speed increasement in times, ie. old dirpref / new dirpref.
------
Algorithm description
The old dirpref algorithm is described in comments:
/*
* 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.
*/
A new directory is allocated in a different cylinder groups than its
parent directory resulting in a directory tree that is spreaded across
all the cylinder groups. This spreading out results in a non-optimal
access to the directories and files. When we have a small filesystem
it is not a problem but when the filesystem is big then perfomance
degradation becomes very apparent.
What I mean by a big file system ?
1. A big filesystem is a filesystem which occupy 20-30 or more percent
of total drive space, i.e. first and last cylinder are physically
located relatively far from each other.
2. It has a relatively large number of cylinder groups, for example
more cylinder groups than 50% of the buffers in the buffer cache.
The first results in long access times, while the second results in
many buffers being used by metadata operations. Such operations use
cylinder group blocks and on-disk inode blocks. The cylinder group
block (fs->fs_cblkno) contains struct cg, inode and block bit maps.
It is 2k in size for the default filesystem parameters. If new and
parent directories are located in different cylinder groups then the
system performs more input/output operations and uses more buffers.
On filesystems with many cylinder groups, lots of cache buffers are
used for metadata operations.
My solution for this problem is very simple. I allocate many directories
in one cylinder group. I also do some things, so that the new allocation
method does not cause excessive fragmentation and all directory inodes
will not be located at a location far from its file's inodes and data.
The algorithm is:
/*
* Find a cylinder group to place a directory.
*
* The policy implemented by this algorithm is to allocate a
* directory inode in the same cylinder group as its parent
* directory, but also to reserve space for its files inodes
* and data. Restrict the number of directories which may be
* allocated one after another in the same cylinder group
* without intervening allocation of files.
*
* If we allocate a first level directory then force allocation
* in another cylinder group.
*/
My early versions of dirpref give me a good results for a wide range of
file operations and different filesystem capacities except one case:
those applications that create their entire directory structure first
and only later fill this structure with files.
My solution for such and similar cases is to limit a number of
directories which may be created one after another in the same cylinder
group without intervening file creations. For this purpose, I allocate
an array of counters at mount time. This array is linked to the superblock
fs->fs_contigdirs[cg]. Each time a directory is created the counter
increases and each time a file is created the counter decreases. A 60Gb
filesystem with 8mb/cg requires 10kb of memory for the counters array.
The maxcontigdirs is a maximum number of directories which may be created
without an intervening file creation. I found in my tests that the best
performance occurs when I restrict the number of directories in one cylinder
group such that all its files may be located in the same cylinder group.
There may be some deterioration in performance if all the file inodes
are in the same cylinder group as its containing directory, but their
data partially resides in a different cylinder group. The maxcontigdirs
value is calculated to try to prevent this condition. Since there is
no way to know how many files and directories will be allocated later
I added two optimization parameters in superblock/tunefs. They are:
int32_t fs_avgfilesize; /* expected average file size */
int32_t fs_avgfpdir; /* expected # of files per directory */
These parameters have reasonable defaults but may be tweeked for special
uses of a filesystem. They are only necessary in rare cases like better
tuning a filesystem being used to store a squid cache.
I have been using this algorithm for about 3 months. I have done
a lot of testing on filesystems with different capacities, average
filesize, average number of files per directory, and so on. I think
this algorithm has no negative impact on filesystem perfomance. It
works better than the default one in all cases. The new dirpref
will greatly improve untarring/removing/coping of big directories,
decrease load on cvs servers and much more. The new dirpref doesn't
speedup a compilation process, but also doesn't slow it down.
Obtained from: Grigoriy Orlov <gluk@ptci.ru>
=====
=====
iedowse 2001/04/23 17:37:17 PDT
Pre-dirpref versions of fsck may zero out the new superblock fields
fs_contigdirs, fs_avgfilesize and fs_avgfpdir. This could cause
panics if these fields were zeroed while a filesystem was mounted
read-only, and then remounted read-write.
Add code to ffs_reload() which copies the fs_contigdirs pointer
from the previous superblock, and reinitialises fs_avgf* if necessary.
Reviewed by: mckusick
=====
=====
nik 2001/04/10 03:36:44 PDT
Add information about the new options to newfs and tunefs which set the
expected average file size and number of files per directory. Could do
with some fleshing out.
=====
2001-09-06 06:16:00 +04:00
|
|
|
memmove(asb->fs_snapinum,
|
|
|
|
sb->fs_snapinum, sizeof sb->fs_snapinum);
|
|
|
|
asb->fs_avgfilesize = sb->fs_avgfilesize;
|
|
|
|
asb->fs_avgfpdir = sb->fs_avgfpdir;
|
2001-12-19 13:05:20 +03:00
|
|
|
asb->fs_pendingblocks = sb->fs_pendingblocks;
|
|
|
|
asb->fs_pendinginodes = sb->fs_pendinginodes;
|
2001-01-26 20:37:16 +03:00
|
|
|
memmove(asb->fs_sparecon,
|
|
|
|
sb->fs_sparecon, sizeof sb->fs_sparecon);
|
|
|
|
/*
|
2001-09-18 12:38:28 +04:00
|
|
|
* The following should not have to be copied, but need to be.
|
2001-01-26 20:37:16 +03:00
|
|
|
*/
|
|
|
|
asb->fs_fsbtodb = sb->fs_fsbtodb;
|
|
|
|
asb->fs_interleave = sb->fs_interleave;
|
|
|
|
asb->fs_npsect = sb->fs_npsect;
|
|
|
|
asb->fs_nrpos = sb->fs_nrpos;
|
|
|
|
asb->fs_qbmask = sb->fs_qbmask;
|
|
|
|
asb->fs_qfmask = sb->fs_qfmask;
|
|
|
|
asb->fs_state = sb->fs_state;
|
|
|
|
asb->fs_maxfilesize = sb->fs_maxfilesize;
|
|
|
|
|
2001-09-18 12:38:28 +04:00
|
|
|
/*
|
|
|
|
* Compare the superblocks, effectively checking every other
|
|
|
|
* field to see if they differ.
|
|
|
|
*/
|
2001-01-26 20:37:16 +03:00
|
|
|
return (memcmp(sb, asb, (int)sb->fs_sbsize));
|
|
|
|
}
|
|
|
|
|
1997-09-16 20:44:43 +04:00
|
|
|
static void
|
1993-03-21 12:45:37 +03:00
|
|
|
badsb(listerr, s)
|
|
|
|
int listerr;
|
|
|
|
char *s;
|
|
|
|
{
|
|
|
|
|
|
|
|
if (!listerr)
|
|
|
|
return;
|
|
|
|
if (preen)
|
1996-09-23 20:18:31 +04:00
|
|
|
printf("%s: ", cdevname());
|
1993-03-21 12:45:37 +03:00
|
|
|
pfatal("BAD SUPER BLOCK: %s\n", s);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Calculate a prototype superblock based on information in the disk label.
|
|
|
|
* When done the cgsblock macro can be calculated and the fs_ncg field
|
|
|
|
* can be used. Do NOT attempt to use other macros without verifying that
|
|
|
|
* their needed information is available!
|
|
|
|
*/
|
1997-09-16 20:44:43 +04:00
|
|
|
static int
|
1993-03-21 12:45:37 +03:00
|
|
|
calcsb(dev, devfd, fs)
|
1998-07-27 00:32:42 +04:00
|
|
|
const char *dev;
|
1993-03-21 12:45:37 +03:00
|
|
|
int devfd;
|
1997-09-14 18:36:29 +04:00
|
|
|
struct fs *fs;
|
1993-03-21 12:45:37 +03:00
|
|
|
{
|
1997-09-14 18:36:29 +04:00
|
|
|
struct disklabel *lp;
|
|
|
|
struct partition *pp;
|
|
|
|
char *cp;
|
1993-03-21 12:45:37 +03:00
|
|
|
int i;
|
|
|
|
|
1994-09-23 18:26:58 +04:00
|
|
|
cp = strchr(dev, '\0') - 1;
|
2001-11-16 08:35:40 +03:00
|
|
|
if ((cp == (char *)-1 || (*cp < 'a' || *cp > 'p')) && !isdigit(*cp)) {
|
1993-03-21 12:45:37 +03:00
|
|
|
pfatal("%s: CANNOT FIGURE OUT FILE SYSTEM PARTITION\n", dev);
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
lp = getdisklabel(dev, devfd);
|
|
|
|
if (isdigit(*cp))
|
|
|
|
pp = &lp->d_partitions[0];
|
|
|
|
else
|
|
|
|
pp = &lp->d_partitions[*cp - 'a'];
|
|
|
|
if (pp->p_fstype != FS_BSDFFS) {
|
|
|
|
pfatal("%s: NOT LABELED AS A BSD FILE SYSTEM (%s)\n",
|
|
|
|
dev, pp->p_fstype < FSMAXTYPES ?
|
|
|
|
fstypenames[pp->p_fstype] : "unknown");
|
|
|
|
return (0);
|
|
|
|
}
|
1998-03-18 20:01:23 +03:00
|
|
|
/* avoid divide by 0 */
|
|
|
|
if (pp->p_fsize == 0 || pp->p_frag == 0)
|
|
|
|
return (0);
|
1994-09-23 18:26:58 +04:00
|
|
|
memset(fs, 0, sizeof(struct fs));
|
1993-03-21 12:45:37 +03:00
|
|
|
fs->fs_fsize = pp->p_fsize;
|
|
|
|
fs->fs_frag = pp->p_frag;
|
|
|
|
fs->fs_cpg = pp->p_cpg;
|
|
|
|
fs->fs_size = pp->p_size;
|
|
|
|
fs->fs_ntrak = lp->d_ntracks;
|
|
|
|
fs->fs_nsect = lp->d_nsectors;
|
|
|
|
fs->fs_spc = lp->d_secpercyl;
|
|
|
|
fs->fs_nspf = fs->fs_fsize / lp->d_secsize;
|
|
|
|
fs->fs_sblkno = roundup(
|
|
|
|
howmany(lp->d_bbsize + lp->d_sbsize, fs->fs_fsize),
|
|
|
|
fs->fs_frag);
|
|
|
|
fs->fs_cgmask = 0xffffffff;
|
|
|
|
for (i = fs->fs_ntrak; i > 1; i >>= 1)
|
|
|
|
fs->fs_cgmask <<= 1;
|
|
|
|
if (!POWEROF2(fs->fs_ntrak))
|
|
|
|
fs->fs_cgmask <<= 1;
|
|
|
|
fs->fs_cgoffset = roundup(
|
|
|
|
howmany(fs->fs_nsect, NSPF(fs)), fs->fs_frag);
|
|
|
|
fs->fs_fpg = (fs->fs_cpg * fs->fs_spc) / NSPF(fs);
|
|
|
|
fs->fs_ncg = howmany(fs->fs_size / fs->fs_spc, fs->fs_cpg);
|
|
|
|
for (fs->fs_fsbtodb = 0, i = NSPF(fs); i > 1; i >>= 1)
|
|
|
|
fs->fs_fsbtodb++;
|
|
|
|
dev_bsize = lp->d_secsize;
|
|
|
|
return (1);
|
|
|
|
}
|
|
|
|
|
1996-09-23 20:18:31 +04:00
|
|
|
static struct disklabel *
|
1993-03-21 12:45:37 +03:00
|
|
|
getdisklabel(s, fd)
|
1998-07-27 00:32:42 +04:00
|
|
|
const char *s;
|
1993-03-21 12:45:37 +03:00
|
|
|
int fd;
|
|
|
|
{
|
|
|
|
static struct disklabel lab;
|
|
|
|
|
|
|
|
if (ioctl(fd, DIOCGDINFO, (char *)&lab) < 0) {
|
|
|
|
if (s == NULL)
|
|
|
|
return ((struct disklabel *)NULL);
|
|
|
|
pwarn("ioctl (GCINFO): %s\n", strerror(errno));
|
1997-09-16 20:44:43 +04:00
|
|
|
errx(EEXIT, "%s: can't read disk label", s);
|
1993-03-21 12:45:37 +03:00
|
|
|
}
|
|
|
|
return (&lab);
|
|
|
|
}
|