NetBSD/sbin/newfs/newfs.c

907 lines
24 KiB
C
Raw Normal View History

/* $NetBSD: newfs.c,v 1.85 2004/11/15 12:21:29 he Exp $ */
/*
* Copyright (c) 1983, 1989, 1993, 1994
* The Regents of the University of California. All rights reserved.
*
* 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. 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.
*/
1993-03-21 12:45:37 +03:00
/*
* Copyright (c) 2002 Networks Associates Technology, Inc.
* All rights reserved.
*
* This software was developed for the FreeBSD Project by Marshall
* Kirk McKusick and Network Associates Laboratories, the Security
* Research Division of Network Associates, Inc. under DARPA/SPAWAR
* contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS
* research program
*
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.
*/
1997-07-01 02:20:30 +04:00
#include <sys/cdefs.h>
1993-03-21 12:45:37 +03:00
#ifndef lint
1997-07-01 02:20:30 +04:00
__COPYRIGHT("@(#) Copyright (c) 1983, 1989, 1993, 1994\n\
The Regents of the University of California. All rights reserved.\n");
1993-03-21 12:45:37 +03:00
#endif /* not lint */
#ifndef lint
#if 0
static char sccsid[] = "@(#)newfs.c 8.13 (Berkeley) 5/1/95";
#else
__RCSID("$NetBSD: newfs.c,v 1.85 2004/11/15 12:21:29 he Exp $");
#endif
1993-03-21 12:45:37 +03:00
#endif /* not lint */
/*
* newfs: friendly front end to mkfs
*/
#include <sys/param.h>
#include <sys/ioctl.h>
#include <sys/disklabel.h>
#include <sys/file.h>
#include <sys/mount.h>
#include <sys/sysctl.h>
#include <sys/wait.h>
1993-03-21 12:45:37 +03:00
#include <ufs/ufs/dir.h>
#include <ufs/ufs/dinode.h>
1998-03-01 05:20:01 +03:00
#include <ufs/ufs/ufsmount.h>
#include <ufs/ffs/fs.h>
#include <ctype.h>
#include <disktab.h>
#include <err.h>
1993-03-21 12:45:37 +03:00
#include <errno.h>
#include <grp.h>
2003-11-01 21:42:00 +03:00
#include <limits.h>
#include <paths.h>
#include <pwd.h>
#include <signal.h>
2003-11-01 21:42:00 +03:00
#include <stdint.h>
1993-03-21 12:45:37 +03:00
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <syslog.h>
#include <unistd.h>
#include <util.h>
#include "mntopts.h"
1997-07-01 02:20:30 +04:00
#include "dkcksum.h"
#include "extern.h"
struct mntopt mopts[] = {
MOPT_STDOPTS,
MOPT_ASYNC,
MOPT_UPDATE,
2002-09-21 22:43:31 +04:00
MOPT_GETARGS,
1997-01-30 12:56:10 +03:00
MOPT_NOATIME,
{ NULL },
};
2000-12-01 14:52:54 +03:00
static struct disklabel *getdisklabel(char *, int);
static void rewritelabel(char *, int, struct disklabel *);
static gid_t mfs_group(const char *);
static uid_t mfs_user(const char *);
static int64_t strsuftoi64(const char *, const char *, int64_t, int64_t, int *);
2000-12-01 14:52:54 +03:00
static void usage(void);
int main(int, char *[]);
1993-03-21 12:45:37 +03:00
#define COMPAT /* allow non-labeled disks */
/*
* The following two constants set the default block and fragment sizes.
* Both constants must be a power of 2 and meet the following constraints:
* MINBSIZE <= DESBLKSIZE <= MAXBSIZE
* sectorsize <= DESFRAGSIZE <= DESBLKSIZE
* DESBLKSIZE / DESFRAGSIZE <= 8
*/
/*
* For file systems smaller than SMALL_FSSIZE we use the S_DFL_* defaults,
* otherwise if less than MEDIUM_FSSIZE use M_DFL_*, otherwise use
* L_DFL_*.
*/
#define SMALL_FSSIZE (20*1024*2)
#define S_DFL_FRAGSIZE 512
#define MEDIUM_FSSIZE (1000*1024*2)
#define M_DFL_FRAGSIZE 1024
#define L_DFL_FRAGSIZE 2048
#define DFL_FRAG_BLK 8
1993-03-21 12:45:37 +03:00
/* Apple requires the fragment size to be at least APPLEUFS_DIRBLKSIZ
* but the block size cannot be larger than Darwin's PAGE_SIZE. See
* the mount check in Darwin's ffs_mountfs for an explanation.
*/
#define APPLEUFS_DFL_FRAGSIZE APPLEUFS_DIRBLKSIZ /* 1024 */
#define APPLEUFS_DFL_BLKSIZE 4096 /* default Darwin PAGE_SIZE */
/*
* Default sector size.
*/
#define DFL_SECSIZE 512
1993-03-21 12:45:37 +03:00
/*
* MAXBLKPG determines the maximum number of data blocks which are
* placed in a single cylinder group. The default is one indirect
* block worth of data blocks.
*/
#define MAXBLKPG_UFS1(bsize) ((bsize) / sizeof(int32_t))
#define MAXBLKPG_UFS2(bsize) ((bsize) / sizeof(int64_t))
1993-03-21 12:45:37 +03:00
/*
* Each file system has a number of inodes statically allocated.
* We allocate one inode slot per NFPI fragments, expecting this
* to be far more than we will ever need.
*/
#define NFPI 4
int mfs; /* run as the memory based filesystem */
int Nflag; /* run without writing file system */
int Oflag = 1; /* format as an 4.3BSD file system */
int64_t fssize; /* file system size */
1993-03-21 12:45:37 +03:00
int sectorsize; /* bytes/sector */
int fsize = 0; /* fragment size */
int bsize = 0; /* block size */
int maxbsize = 0; /* maximum clustering */
1993-03-21 12:45:37 +03:00
int minfree = MINFREE; /* free space threshold */
int opt = DEFAULTOPT; /* optimization preference (space or time) */
int density; /* number of bytes per inode */
2004-09-07 06:18:06 +04:00
int num_inodes; /* number of inodes (overrides density) */
int maxcontig = 0; /* max contiguous blocks to allocate */
1993-03-21 12:45:37 +03:00
int maxbpg; /* maximum blocks per file in a cyl group */
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
int avgfilesize = AVFILESIZ;/* expected average file size */
int avgfpdir = AFPDIR; /* expected number of files per directory */
int mntflags = MNT_ASYNC; /* flags to be passed to mount */
1993-03-21 12:45:37 +03:00
u_long memleft; /* virtual memory available */
caddr_t membase; /* start address of memory based filesystem */
int needswap; /* Filesystem not in native byte order */
1993-03-21 12:45:37 +03:00
#ifdef COMPAT
char *disktype;
int unlabeled;
#endif
char *appleufs_volname = 0; /* Apple UFS volume name */
int isappleufs = 0;
1993-03-21 12:45:37 +03:00
char device[MAXPATHLEN];
int
2000-12-01 14:52:54 +03:00
main(int argc, char *argv[])
1993-03-21 12:45:37 +03:00
{
struct partition *pp = NULL;
struct disklabel *lp = NULL;
1993-03-21 12:45:37 +03:00
struct partition oldpartition;
struct statvfs *mp;
struct stat sb;
int ch, fsi, fso, len, n, Fflag, Iflag, Zflag;
uint ptn = 0; /* gcc -Wuninitialised */
char *cp, *s1, *s2, *special;
const char *opstring;
int byte_sized = 0;
#ifdef MFS
struct mfs_args args;
char mountfromname[100];
pid_t pid, res;
struct statvfs sf;
2002-01-18 11:37:08 +03:00
int status;
#endif
mode_t mfsmode = 01777; /* default mode for a /tmp-type directory */
uid_t mfsuid = 0; /* user root */
gid_t mfsgid = 0; /* group wheel */
1993-03-21 12:45:37 +03:00
cp = NULL;
fsi = fso = -1;
Fflag = Iflag = Zflag = 0;
2001-02-20 01:48:57 +03:00
if (strstr(getprogname(), "mfs")) {
1993-03-21 12:45:37 +03:00
mfs = 1;
} else {
/* Undocumented, for ease of testing */
if (argv[1] != NULL && !strcmp(argv[1], "-mfs")) {
argv++;
argc--;
mfs = 1;
}
1993-03-21 12:45:37 +03:00
}
opstring = mfs ?
"NT:a:b:d:e:f:g:h:i:m:n:o:p:s:u:" :
"B:FINO:S:T:Za:b:d:e:f:g:h:i:l:m:n:o:p:r:s:u:v:";
while ((ch = getopt(argc, argv, opstring)) != -1)
switch (ch) {
case 'B':
if (strcmp(optarg, "be") == 0) {
#if BYTE_ORDER == LITTLE_ENDIAN
needswap = 1;
#endif
} else if (strcmp(optarg, "le") == 0) {
#if BYTE_ORDER == BIG_ENDIAN
needswap = 1;
#endif
} else
usage();
break;
case 'F':
Fflag = 1;
break;
case 'I':
Iflag = 1;
break;
1993-03-21 12:45:37 +03:00
case 'N':
Nflag = 1;
break;
case 'O':
Oflag = strsuftoi64("format", optarg, 0, 2, NULL);
1993-03-21 12:45:37 +03:00
break;
case 'S':
sectorsize = strsuftoi64("sector size",
optarg, 1, INT_MAX, NULL);
1993-03-21 12:45:37 +03:00
break;
#ifdef COMPAT
case 'T':
disktype = optarg;
break;
#endif
case 'Z':
Zflag = 1;
break;
1993-03-21 12:45:37 +03:00
case 'a':
maxcontig = strsuftoi64("maximum contiguous blocks",
optarg, 1, INT_MAX, NULL);
1993-03-21 12:45:37 +03:00
break;
case 'b':
bsize = strsuftoi64("block size",
optarg, MINBSIZE, MAXBSIZE, NULL);
1993-03-21 12:45:37 +03:00
break;
case 'd':
maxbsize = strsuftoi64("maximum extent size",
optarg, 0, INT_MAX, NULL);
1993-03-21 12:45:37 +03:00
break;
case 'e':
maxbpg = strsuftoi64(
"blocks per file in a cylinder group",
optarg, 1, INT_MAX, NULL);
1993-03-21 12:45:37 +03:00
break;
case 'f':
fsize = strsuftoi64("fragment size",
optarg, 1, MAXBSIZE, NULL);
1993-03-21 12:45:37 +03:00
break;
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
case 'g':
if (mfs)
mfsgid = mfs_group(optarg);
else {
avgfilesize = strsuftoi64("average file size",
optarg, 1, INT_MAX, NULL);
}
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
break;
case 'h':
avgfpdir = strsuftoi64("expected files per directory",
optarg, 1, INT_MAX, NULL);
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
break;
1993-03-21 12:45:37 +03:00
case 'i':
density = strsuftoi64("bytes per inode",
optarg, 1, INT_MAX, NULL);
1993-03-21 12:45:37 +03:00
break;
case 'm':
minfree = strsuftoi64("free space %",
optarg, 0, 99, NULL);
1993-03-21 12:45:37 +03:00
break;
case 'n':
num_inodes = strsuftoi64("number of inodes",
optarg, 1, INT_MAX, NULL);
break;
1993-03-21 12:45:37 +03:00
case 'o':
if (mfs)
getmntopts(optarg, mopts, &mntflags, 0);
else {
if (strcmp(optarg, "space") == 0)
opt = FS_OPTSPACE;
else if (strcmp(optarg, "time") == 0)
opt = FS_OPTTIME;
else
1997-07-01 02:20:30 +04:00
errx(1, "%s %s",
"unknown optimization preference: ",
"use `space' or `time'.");
}
1993-03-21 12:45:37 +03:00
break;
case 'p':
if (mfs) {
if ((mfsmode = strtol(optarg, NULL, 8)) <= 0)
errx(1, "bad mode `%s'", optarg);
} else
errx(1, "unknown option 'p'");
1993-03-21 12:45:37 +03:00
break;
case 's':
fssize = strsuftoi64("file system size",
optarg, INT64_MIN, INT64_MAX, &byte_sized);
1993-03-21 12:45:37 +03:00
break;
case 'u':
if (mfs)
mfsuid = mfs_user(optarg);
else
errx(1, "deprecated option 'u'");
1993-03-21 12:45:37 +03:00
break;
case 'v':
appleufs_volname = optarg;
if (strchr(appleufs_volname, ':') || strchr(appleufs_volname, '/'))
errx(1,"Apple UFS volume name cannot contain ':' or '/'");
if (appleufs_volname[0] == '\0')
errx(1,"Apple UFS volume name cannot be zero length");
isappleufs = 1;
break;
1993-03-21 12:45:37 +03:00
case '?':
default:
usage();
}
argc -= optind;
argv += optind;
#ifdef MFS
/* This is enough to get through the correct kernel code paths */
memset(&args, 0, sizeof args);
args.fspec = mountfromname;
if (mntflags & (MNT_GETARGS | MNT_UPDATE)) {
if (mount(MOUNT_MFS, argv[1], mntflags, &args) < 0)
err(1, "mount `%s' failed", argv[1]);
if (mntflags & MNT_GETARGS)
printf("base=%p, size=%ld\n", args.base, args.size);
exit(0);
}
#endif
1993-03-21 12:45:37 +03:00
if (argc != 2 && (mfs || argc != 1))
usage();
memset(&oldpartition, 0, sizeof oldpartition);
memset(&sb, 0, sizeof sb);
1993-03-21 12:45:37 +03:00
special = argv[0];
if (Fflag || mfs) {
/*
* It's a file system image or an MFS,
* no label, use fixed default for sectorsize.
*/
if (sectorsize == 0)
sectorsize = DFL_SECSIZE;
if (mfs) {
/* Default filesystem size to that of supplied device */
if (fssize == 0)
stat(special, &sb);
} else {
/* creating image in a regular file */
int fl;
if (Nflag)
fl = O_RDONLY;
else {
if (fssize > 0)
fl = O_RDWR | O_CREAT;
else
fl = O_RDWR;
}
fsi = open(special, fl, 0777);
if (fsi == -1)
err(1, "can't open file %s", special);
if (fstat(fsi, &sb) == -1)
err(1, "can't fstat opened %s", special);
if (!Nflag)
fso = fsi;
}
} else { /* !Fflag && !mfs */
fsi = opendisk(special, O_RDONLY, device, sizeof(device), 0);
special = device;
if (fsi < 0 || fstat(fsi, &sb) == -1)
err(1, "%s: open for read", special);
if (!Nflag) {
fso = open(special, O_WRONLY, 0);
if (fso < 0)
err(1, "%s: open for write", special);
/* Bail if target special is mounted */
n = getmntinfo(&mp, MNT_NOWAIT);
if (n == 0)
err(1, "%s: getmntinfo", special);
len = sizeof(_PATH_DEV) - 1;
s1 = special;
if (strncmp(_PATH_DEV, s1, len) == 0)
s1 += len;
while (--n >= 0) {
s2 = mp->f_mntfromname;
if (strncmp(_PATH_DEV, s2, len) == 0) {
s2 += len - 1;
*s2 = 'r';
}
if (strcmp(s1, s2) == 0 ||
strcmp(s1, &s2[1]) == 0)
errx(1, "%s is mounted on %s",
special, mp->f_mntonname);
++mp;
}
}
1993-03-21 12:45:37 +03:00
#ifdef COMPAT
if (disktype == NULL)
disktype = argv[1];
1993-03-21 12:45:37 +03:00
#endif
lp = getdisklabel(special, fsi);
if (sectorsize == 0) {
sectorsize = lp->d_secsize;
if (sectorsize <= 0)
errx(1, "no default sector size");
}
ptn = strchr(special, '\0')[-1] - 'a';
if (ptn < lp->d_npartitions && ptn == DISKPART(sb.st_rdev)) {
/* Assume partition really does match the label */
pp = &lp->d_partitions[ptn];
oldpartition = *pp;
if (pp->p_size == 0)
errx(1, "`%c' partition is unavailable",
'a' + ptn);
if (pp->p_fstype == FS_APPLEUFS)
isappleufs = 1;
if (!Iflag) {
if (isappleufs) {
if (pp->p_fstype != FS_APPLEUFS)
errx(1, "`%c' partition type is not `Apple UFS'", 'a' + ptn);
} else {
if (pp->p_fstype != FS_BSDFFS)
errx(1, "`%c' partition type is not `4.2BSD'", 'a' + ptn);
}
}
}
} /* !Fflag && !mfs */
if (byte_sized)
fssize /= sectorsize;
if (fssize <= 0) {
if (sb.st_size != 0)
fssize += sb.st_size / sectorsize;
else
fssize += oldpartition.p_size;
if (fssize <= 0)
errx(1, "Unable to determine file system size");
1993-03-21 12:45:37 +03:00
}
if (pp != NULL && fssize > pp->p_size)
errx(1, "maximum file system size on `%s' is %d sectors",
special, pp->p_size);
/* XXXLUKEM: only ftruncate() regular files ? (dsl: or at all?) */
if (Fflag && fso != -1
&& ftruncate(fso, (off_t)fssize * sectorsize) == -1)
err(1, "can't ftruncate %s to %" PRId64, special, fssize);
if (Zflag && fso != -1) { /* pre-zero (and de-sparce) the file */
char *buf;
int bufsize, i;
off_t bufrem;
struct statvfs sfs;
if (fstatvfs(fso, &sfs) == -1) {
warn("can't fstatvfs `%s'", special);
bufsize = 8192;
} else
bufsize = sfs.f_iosize;
if ((buf = calloc(1, bufsize)) == NULL)
err(1, "can't malloc buffer of %d",
bufsize);
bufrem = fssize * sectorsize;
printf(
"Creating file system image in `%s', size %lld bytes, in %d byte chunks.\n",
special, (long long)bufrem, bufsize);
while (bufrem > 0) {
i = write(fso, buf, MIN(bufsize, bufrem));
if (i == -1)
err(1, "writing image");
bufrem -= i;
}
free(buf);
}
/* Sort out fragment and block sizes */
1993-03-21 12:45:37 +03:00
if (fsize == 0) {
fsize = bsize / DFL_FRAG_BLK;
if (fsize == 0)
fsize = oldpartition.p_fsize;
if (fsize <= 0) {
if (isappleufs) {
fsize = APPLEUFS_DFL_FRAGSIZE;
} else {
if (fssize < SMALL_FSSIZE)
fsize = S_DFL_FRAGSIZE;
else if (fssize < MEDIUM_FSSIZE)
fsize = M_DFL_FRAGSIZE;
else
fsize = L_DFL_FRAGSIZE;
if (fsize < sectorsize)
fsize = sectorsize;
}
}
1993-03-21 12:45:37 +03:00
}
if (bsize == 0) {
bsize = oldpartition.p_frag * oldpartition.p_fsize;
if (bsize <= 0) {
if (isappleufs)
bsize = APPLEUFS_DFL_BLKSIZE;
else
bsize = DFL_FRAG_BLK * fsize;
}
1993-03-21 12:45:37 +03:00
}
if (isappleufs && (fsize < APPLEUFS_DFL_FRAGSIZE)) {
warnx("Warning: chosen fsize of %d is less than Apple UFS minimum of %d",
fsize, APPLEUFS_DFL_FRAGSIZE);
}
if (isappleufs && (bsize > APPLEUFS_DFL_BLKSIZE)) {
warnx("Warning: chosen bsize of %d is greater than Apple UFS maximum of %d",
bsize, APPLEUFS_DFL_BLKSIZE);
}
/*
* Maxcontig sets the default for the maximum number of blocks
* that may be allocated sequentially. With filesystem clustering
* it is possible to allocate contiguous blocks up to the maximum
* transfer size permitted by the controller or buffering.
*/
if (maxcontig == 0)
maxcontig = MAX(1, MIN(MAXPHYS, MAXBSIZE) / bsize);
1993-03-21 12:45:37 +03:00
if (density == 0)
density = NFPI * fsize;
1994-02-06 11:19:56 +03:00
if (minfree < MINFREE && opt != FS_OPTSPACE) {
1997-07-01 02:20:30 +04:00
warnx("%s %s %d%%", "Warning: changing optimization to space",
"because minfree is less than", MINFREE);
1993-03-21 12:45:37 +03:00
opt = FS_OPTSPACE;
}
if (maxbpg == 0) {
if (Oflag <= 1)
maxbpg = MAXBLKPG_UFS1(bsize);
else
maxbpg = MAXBLKPG_UFS2(bsize);
1993-03-21 12:45:37 +03:00
}
mkfs(pp, special, fsi, fso, mfsmode, mfsuid, mfsgid);
if (!Nflag && pp != NULL
&& memcmp(pp, &oldpartition, sizeof(oldpartition)))
1993-03-21 12:45:37 +03:00
rewritelabel(special, fso, lp);
if (fsi != -1 && fsi != fso)
close(fsi);
if (fso != -1)
1993-03-21 12:45:37 +03:00
close(fso);
#ifdef MFS
if (mfs) {
switch (pid = fork()) {
case -1:
perror("mfs");
exit(10);
case 0:
1998-07-28 23:22:54 +04:00
(void)snprintf(mountfromname, sizeof(mountfromname),
"mfs:%d", getpid());
break;
default:
1998-07-28 23:22:54 +04:00
(void)snprintf(mountfromname, sizeof(mountfromname),
"mfs:%d", pid);
for (;;) {
/*
* spin until the mount succeeds
* or the child exits
*/
usleep(1);
/*
* XXX Here is a race condition: another process
* can mount a filesystem which hides our
* ramdisk before we see the success.
*/
if (statvfs(argv[1], &sf) < 0)
err(88, "statvfs %s", argv[1]);
if (!strcmp(sf.f_mntfromname, mountfromname) &&
!strncmp(sf.f_mntonname, argv[1],
MNAMELEN) &&
!strcmp(sf.f_fstypename, "mfs"))
exit(0);
res = waitpid(pid, &status, WNOHANG);
if (res == -1)
err(11, "waitpid");
if (res != pid)
continue;
if (WIFEXITED(status)) {
if (WEXITSTATUS(status) == 0)
exit(0);
errx(1, "%s: mount: %s", argv[1],
strerror(WEXITSTATUS(status)));
} else
errx(11, "abnormal termination");
}
/* NOTREACHED */
}
(void) setsid();
(void) close(0);
(void) close(1);
(void) close(2);
(void) chdir("/");
args.export.ex_root = -2;
if (mntflags & MNT_RDONLY)
args.export.ex_flags = MNT_EXRDONLY;
else
args.export.ex_flags = 0;
1993-03-21 12:45:37 +03:00
args.base = membase;
args.size = fssize * sectorsize;
if (mount(MOUNT_MFS, argv[1], mntflags, &args) < 0)
exit(errno); /* parent prints message */
1993-03-21 12:45:37 +03:00
}
#endif
exit(0);
}
#ifdef COMPAT
const char lmsg[] = "%s: can't read disk label; disk type must be specified";
1993-03-21 12:45:37 +03:00
#else
const char lmsg[] = "%s: can't read disk label";
1993-03-21 12:45:37 +03:00
#endif
1997-07-01 02:20:30 +04:00
static struct disklabel *
getdisklabel(char *s, int fd)
1993-03-21 12:45:37 +03:00
{
static struct disklabel lab;
#ifdef COMPAT
if (disktype) {
struct disklabel *lp;
unlabeled++;
lp = getdiskbyname(disktype);
if (lp == NULL)
errx(1, "%s: unknown disk type", disktype);
return (lp);
}
1993-03-21 12:45:37 +03:00
#endif
if (ioctl(fd, DIOCGDINFO, &lab) < 0) {
warn("ioctl (GDINFO)");
1997-07-01 02:20:30 +04:00
errx(1, lmsg, s);
1993-03-21 12:45:37 +03:00
}
return (&lab);
}
1997-07-01 02:20:30 +04:00
static void
rewritelabel(char *s, int fd, struct disklabel *lp)
1993-03-21 12:45:37 +03:00
{
#ifdef COMPAT
if (unlabeled)
return;
#endif
lp->d_checksum = 0;
lp->d_checksum = dkcksum(lp);
if (ioctl(fd, DIOCWDINFO, (char *)lp) < 0) {
if (errno == ESRCH)
return;
warn("ioctl (WDINFO)");
1997-07-01 02:20:30 +04:00
errx(1, "%s: can't rewrite disk label", s);
1993-03-21 12:45:37 +03:00
}
1998-10-20 05:48:14 +04:00
#if __vax__
1993-03-21 12:45:37 +03:00
if (lp->d_type == DTYPE_SMD && lp->d_flags & D_BADSECT) {
1997-07-01 02:20:30 +04:00
int i;
1993-03-21 12:45:37 +03:00
int cfd;
daddr_t alt;
off_t loff;
1993-03-21 12:45:37 +03:00
char specname[64];
char blk[1024];
char *cp;
/*
* Make name for 'c' partition.
*/
2003-07-13 12:05:58 +04:00
strlcpy(specname, s, sizeof(specname));
1993-03-21 12:45:37 +03:00
cp = specname + strlen(specname) - 1;
if (!isdigit((unsigned char)*cp))
1993-03-21 12:45:37 +03:00
*cp = 'c';
cfd = open(specname, O_WRONLY);
if (cfd < 0)
err(1, "%s: open", specname);
if ((loff = getlabeloffset()) < 0)
err(1, "getlabeloffset()");
memset(blk, 0, sizeof(blk));
*(struct disklabel *)(blk + loff) = *lp;
1993-03-21 12:45:37 +03:00
alt = lp->d_ncylinders * lp->d_secpercyl - lp->d_nsectors;
for (i = 1; i < 11 && i < lp->d_nsectors; i += 2) {
off_t offset;
offset = alt + i;
offset *= lp->d_secsize;
if (lseek(cfd, offset, SEEK_SET) == -1)
1997-07-01 02:20:30 +04:00
err(1, "lseek to badsector area: ");
1993-03-21 12:45:37 +03:00
if (write(cfd, blk, lp->d_secsize) < lp->d_secsize)
warn("alternate label %d write", i/2);
1993-03-21 12:45:37 +03:00
}
close(cfd);
}
#endif
}
static gid_t
mfs_group(const char *gname)
{
struct group *gp;
if (!(gp = getgrnam(gname)) && !isdigit((unsigned char)*gname))
errx(1, "unknown gname %s", gname);
return gp ? gp->gr_gid : atoi(gname);
}
static uid_t
mfs_user(const char *uname)
{
struct passwd *pp;
if (!(pp = getpwnam(uname)) && !isdigit((unsigned char)*uname))
errx(1, "unknown user %s", uname);
return pp ? pp->pw_uid : atoi(uname);
}
static int64_t
strsuftoi64(const char *desc, const char *arg, int64_t min, int64_t max, int *num_suffix)
{
int64_t result, r1;
int shift = 0;
char *ep;
errno = 0;
r1 = strtoll(arg, &ep, 10);
if (ep[0] != '\0' && ep[1] != '\0')
errx(1, "%s `%s' is not a valid number.", desc, arg);
switch (ep[0]) {
case '\0':
case 's': case 'S':
if (num_suffix != NULL)
*num_suffix = 0;
break;
case 'g': case 'G':
shift += 10;
/* FALLTHROUGH */
case 'm': case 'M':
shift += 10;
/* FALLTHROUGH */
case 'k': case 'K':
shift += 10;
/* FALLTHROUGH */
case 'b': case 'B':
if (num_suffix != NULL)
*num_suffix = 1;
break;
default:
errx(1, "`%s' is not a valid suffix for %s.", ep, desc);
}
result = r1 << shift;
if (errno == ERANGE || result >> shift != r1)
errx(1, "%s `%s' is too large to convert.", desc, arg);
if (result < min)
errx(1, "%s `%s' (%" PRId64 ") is less than the minimum (%" PRId64 ").",
desc, arg, result, min);
if (result > max)
errx(1, "%s `%s' (%" PRId64 ") is greater than the maximum (%" PRId64 ").",
desc, arg, result, max);
return result;
}
#define NEWFS 1
#define MFS_MOUNT 2
#define BOTH NEWFS | MFS_MOUNT
struct help_strings {
int flags;
const char *str;
} const help_strings[] = {
{ NEWFS, "-B byteorder\tbyte order (`be' or `le')" },
{ NEWFS, "-F \t\tcreate file system image in regular file" },
{ NEWFS, "-I \t\tdo not check that the file system type is '4.2BSD'" },
{ BOTH, "-N \t\tdo not create file system, just print out "
"parameters" },
{ NEWFS, "-O N\t\tfilesystem format: 0 ==> 4.3BSD, 1 ==> FFS, 2 ==> UFS2" },
{ NEWFS, "-S secsize\tsector size" },
#ifdef COMPAT
{ NEWFS, "-T disktype\tdisk type" },
#endif
{ NEWFS, "-Z \t\tpre-zero the image file" },
{ BOTH, "-a maxcontig\tmaximum contiguous blocks" },
{ BOTH, "-b bsize\tblock size" },
{ BOTH, "-d maxbsize\tmaximum extent size" },
{ BOTH, "-e maxbpg\tmaximum blocks per file in a cylinder group"
},
{ BOTH, "-f fsize\tfrag size" },
{ NEWFS, "-g avgfilesize\taverage file size" },
{ MFS_MOUNT, "-g groupname\tgroup name of mount point" },
{ BOTH, "-h avgfpdir\taverage files per directory" },
{ BOTH, "-i density\tnumber of bytes per inode" },
{ BOTH, "-m minfree\tminimum free space %%" },
{ BOTH, "-n inodes\tnumber of inodes (overrides -i density)" },
{ BOTH, "-o optim\toptimization preference (`space' or `time')"
},
{ MFS_MOUNT, "-p perm\t\tpermissions (in octal)" },
{ BOTH, "-s fssize\tfile system size (sectors)" },
{ MFS_MOUNT, "-u username\tuser name of mount point" },
{ NEWFS, "-v volname\tApple UFS volume name" },
{ 0, NULL }
};
1997-07-01 02:20:30 +04:00
static void
2000-12-01 14:52:54 +03:00
usage(void)
1993-03-21 12:45:37 +03:00
{
int match;
const struct help_strings *hs;
1993-03-21 12:45:37 +03:00
if (mfs) {
fprintf(stderr,
"usage: %s [ fsoptions ] special-device mount-point\n",
2001-02-20 01:48:57 +03:00
getprogname());
1993-03-21 12:45:37 +03:00
} else
fprintf(stderr,
"usage: %s [ fsoptions ] special-device%s\n",
2001-02-20 01:48:57 +03:00
getprogname(),
1993-03-21 12:45:37 +03:00
#ifdef COMPAT
" [disk-type]");
1993-03-21 12:45:37 +03:00
#else
"");
#endif
fprintf(stderr, "where fsoptions are:\n");
match = mfs ? MFS_MOUNT : NEWFS;
for (hs = help_strings; hs->flags != 0; hs++)
if (hs->flags & match)
fprintf(stderr, "\t%s\n", hs->str);
1993-03-21 12:45:37 +03:00
exit(1);
}