/* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License, Version 1.0 only * (the "License"). You may not use this file except in compliance * with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2005 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ /* Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */ /* All Rights Reserved */ /* * University Copyright- Copyright (c) 1982, 1986, 1988 * The Regents of the University of California * All Rights Reserved * * University Acknowledgment- Portions of this document are derived from * software developed by the University of California, Berkeley, and its * contributors. */ #pragma ident "%Z%%M% %I% %E% SMI" /* * The maximum supported file system size (in sectors) is the * number of frags that can be represented in an int32_t field * (INT_MAX) times the maximum number of sectors per frag. Since * the maximum frag size is MAXBSIZE, the maximum number of sectors * per frag is MAXBSIZE/DEV_BSIZE. */ #define FS_MAX (((diskaddr_t)INT_MAX) * (MAXBSIZE/DEV_BSIZE)) /* * make file system for cylinder-group style file systems * * usage: * * mkfs [-F FSType] [-V] [-G [-P]] [-M dirname] [-m] [options] * [-o specific_options] special size * [nsect ntrack bsize fsize cpg minfree rps nbpi opt apc rotdelay * 2 3 4 5 6 7 8 9 10 11 12 * nrpos maxcontig mtb] * 13 14 15 * * where specific_options are: * N - no create * nsect - The number of sectors per track * ntrack - The number of tracks per cylinder * bsize - block size * fragsize - fragment size * cgsize - The number of disk cylinders per cylinder group. * free - minimum free space * rps - rotational speed (rev/sec). * nbpi - number of data bytes per allocated inode * opt - optimization (space, time) * apc - number of alternates * gap - gap size * nrpos - number of rotational positions * maxcontig - maximum number of logical blocks that will be * allocated contiguously before inserting rotational delay * mtb - if "y", set up file system for eventual growth to over a * a terabyte * -P Do not grow the file system, but print on stdout the maximal * size in sectors to which the file system can be increased. The calculated * size is limited by the value provided by the operand size. * * Note that -P is a project-private interface and together with -G intended * to be used only by the growfs script. It is therefore purposely not * documented in the man page. * The -P option is covered by PSARC case 2003/422. */ /* * The following constants set the defaults used for the number * of sectors/track (fs_nsect), and number of tracks/cyl (fs_ntrak). * * NSECT NTRAK * 72MB CDC 18 9 * 30MB CDC 18 5 * 720KB Diskette 9 2 */ #define DFLNSECT 32 #define DFLNTRAK 16 /* * 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 * DEV_BSIZE <= DESFRAGSIZE <= DESBLKSIZE * DESBLKSIZE / DESFRAGSIZE <= 8 */ #define DESBLKSIZE 8192 #define DESFRAGSIZE 1024 /* * The maximum number of cylinders in a group depends upon how much * information can be stored on a single cylinder. The default is to * use 16 cylinders per group. This is effectively tradition - it was * the largest value acceptable under SunOs 4.1 */ #define DESCPG 16 /* desired fs_cpg */ /* * MINFREE gives the minimum acceptable percentage of file system * blocks which may be free. If the freelist drops below this level * only the superuser may continue to allocate blocks. This may * be set to 0 if no reserve of free blocks is deemed necessary, * however throughput drops by fifty percent if the file system * is run at between 90% and 100% full; thus the default value of * fs_minfree is 10%. With 10% free space, fragmentation is not a * problem, so we choose to optimize for time. */ #define MINFREE 10 #define DEFAULTOPT FS_OPTTIME /* * ROTDELAY gives the minimum number of milliseconds to initiate * another disk transfer on the same cylinder. It is no longer used * and will always default to 0. */ #define ROTDELAY 0 /* * 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(bsize) ((bsize) / sizeof (daddr32_t)) /* * Each file system has a number of inodes statically allocated. * We allocate one inode slot per NBPI bytes, expecting this * to be far more than we will ever need. */ #define NBPI 2048 /* Number Bytes Per Inode */ #define MTB_NBPI (MB) /* Number Bytes Per Inode for multi-terabyte */ /* * Disks are assumed to rotate at 60HZ, unless otherwise specified. */ #define DEFHZ 60 /* * Cylinder group related limits. * * For each cylinder we keep track of the availability of blocks at different * rotational positions, so that we can lay out the data to be picked * up with minimum rotational latency. NRPOS is the number of rotational * positions which we distinguish. With NRPOS 8 the resolution of our * summary information is 2ms for a typical 3600 rpm drive. */ #define NRPOS 8 /* number distinct rotational positions */ /* * range_check "user_supplied" flag values. */ #define RC_DEFAULT 0 #define RC_KEYWORD 1 #define RC_POSITIONAL 2 #ifndef STANDALONE #include #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "roll_log.h" #define bcopy(f, t, n) (void) memcpy(t, f, n) #define bzero(s, n) (void) memset(s, 0, n) #define bcmp(s, d, n) memcmp(s, d, n) #define index(s, r) strchr(s, r) #define rindex(s, r) strrchr(s, r) #include #include #include #include #include /* for ENDIAN defines */ #include #include #include extern offset_t llseek(); extern char *getfullblkname(); extern long lrand48(); extern int optind; extern char *optarg; /* * The size of a cylinder group is calculated by CGSIZE. The maximum size * is limited by the fact that cylinder groups are at most one block. * Its size is derived from the size of the maps maintained in the * cylinder group and the (struct cg) size. */ #define CGSIZE(fs) \ /* base cg */ (sizeof (struct cg) + \ /* blktot size */ (fs)->fs_cpg * sizeof (long) + \ /* blks size */ (fs)->fs_cpg * (fs)->fs_nrpos * sizeof (short) + \ /* inode map */ howmany((fs)->fs_ipg, NBBY) + \ /* block map */ howmany((fs)->fs_cpg * (fs)->fs_spc / NSPF(fs), NBBY)) /* * We limit the size of the inode map to be no more than a * third of the cylinder group space, since we must leave at * least an equal amount of space for the block map. * * N.B.: MAXIpG must be a multiple of INOPB(fs). */ #define MAXIpG(fs) roundup((fs)->fs_bsize * NBBY / 3, INOPB(fs)) /* * Same as MAXIpG, but parameterized by the block size (b) and the * cylinder group divisor (d), which is the reciprocal of the fraction of the * cylinder group overhead block that is used for the inode map. So for * example, if d = 5, the macro's computation assumes that 1/5 of the * cylinder group overhead block can be dedicated to the inode map. */ #define MAXIpG_B(b, d) roundup((b) * NBBY / (d), (b) / sizeof (struct dinode)) #define UMASK 0755 #define MAXINOPB (MAXBSIZE / sizeof (struct dinode)) #define POWEROF2(num) (((num) & ((num) - 1)) == 0) #define MB (1024*1024) #define BETWEEN(x, l, h) ((x) >= (l) && (x) <= (h)) /* * Used to set the inode generation number. Since both inodes and dinodes * are dealt with, we really need a pointer to an icommon here. */ #define IRANDOMIZE(icp) (icp)->ic_gen = lrand48(); /* * Flags for number() */ #define ALLOW_PERCENT 0x01 /* allow trailing `%' on number */ #define ALLOW_MS1 0x02 /* allow trailing `ms', state 1 */ #define ALLOW_MS2 0x04 /* allow trailing `ms', state 2 */ #define ALLOW_END_ONLY 0x08 /* must be at end of number & suffixes */ #define MAXAIO 1000 /* maximum number of outstanding I/O's we'll manage */ #define BLOCK 1 /* block in aiowait */ #define NOBLOCK 0 /* don't block in aiowait */ #define RELEASE 1 /* free an aio buffer after use */ #define SAVE 0 /* don't free the buffer */ typedef struct aio_trans { aio_result_t resultbuf; diskaddr_t bno; char *buffer; int size; int release; struct aio_trans *next; } aio_trans; typedef struct aio_results { int max; int outstanding; int maxpend; aio_trans *trans; } aio_results; int aio_inited = 0; aio_results results; /* * Allow up to MAXBUF aio requests that each have a unique buffer. * More aio's might be done, but not using memory through the getbuf() * interface. This can be raised, but you run into the potential of * using more memory than is physically available on the machine, * and if you start swapping, you can forget about performance. * To prevent this, we also limit the total memory used for a given * type of buffer to MAXBUFMEM. * * Tests indicate a cylinder group's worth of inodes takes: * * NBPI Size of Inode Buffer * 2k 1688k * 8k 424k * * initcg() stores all the inodes for a cylinder group in one buffer, * so allowing 20 buffers could take 32 MB if not limited by MAXBUFMEM. */ #define MAXBUF 20 #define MAXBUFMEM (8 * 1024 * 1024) /* * header information for buffers managed by getbuf() and freebuf() */ typedef struct bufhdr { struct bufhdr *head; struct bufhdr *next; } bufhdr; int bufhdrsize; bufhdr inodebuf = { NULL, NULL }; bufhdr cgsumbuf = { NULL, NULL }; #define SECTORS_PER_TERABYTE (1LL << 31) /* * The following constant specifies an upper limit for file system size * that is actually a lot bigger than we expect to support with UFS. (Since * it's specified in sectors, the file system size would be 2**44 * 512, * which is 2**53, which is 8192 Terabytes.) However, it's useful * for checking the basic sanity of a size value that is input on the * command line. */ #define FS_SIZE_UPPER_LIMIT 0x100000000000LL /* * Forward declarations */ static char *getbuf(bufhdr *bufhead, int size); static void freebuf(char *buf); static void freetrans(aio_trans *transp); static aio_trans *get_aiop(); static aio_trans *wait_for_write(int block); static void initcg(int cylno); static void fsinit(); static int makedir(struct direct *protodir, int entries); static void iput(struct inode *ip); static void rdfs(diskaddr_t bno, int size, char *bf); static void wtfs(diskaddr_t bno, int size, char *bf); static void awtfs(diskaddr_t bno, int size, char *bf, int release); static void wtfs_breakup(diskaddr_t bno, int size, char *bf); static int isblock(struct fs *fs, unsigned char *cp, int h); static void clrblock(struct fs *fs, unsigned char *cp, int h); static void setblock(struct fs *fs, unsigned char *cp, int h); static void usage(); static void dump_fscmd(char *fsys, int fsi); static uint64_t number(uint64_t d_value, char *param, int flags); static int match(char *s); static char checkopt(char *optim); static char checkmtb(char *mtbarg); static void range_check(long *varp, char *name, long minimum, long maximum, long def_val, int user_supplied); static void range_check_64(uint64_t *varp, char *name, uint64_t minimum, uint64_t maximum, uint64_t def_val, int user_supplied); static daddr32_t alloc(int size, int mode); static diskaddr_t get_max_size(int fd); static long get_max_track_size(int fd); static void block_sigint(sigset_t *old_mask); static void unblock_sigint(sigset_t *old_mask); static void recover_from_sigint(int signum); static int confirm_abort(void); static int getline(FILE *fp, char *loc, int maxlen); static void flush_writes(void); static long compute_maxcpg(long, long, long, long, long); static int in_64bit_mode(void); static int validate_size(int fd, diskaddr_t size); static void dump_sblock(void); union { struct fs fs; char pad[SBSIZE]; } fsun; #define sblock fsun.fs struct csum *fscs; union cgun { struct cg cg; char pad[MAXBSIZE]; } cgun; #define acg cgun.cg /* * Size of screen in cols in which to fit output */ #define WIDTH 80 struct dinode zino[MAXBSIZE / sizeof (struct dinode)]; /* * file descriptors used for rdfs(fsi) and wtfs(fso). * Initialized to an illegal file descriptor number. */ int fsi = -1; int fso = -1; /* * The BIG parameter is machine dependent. It should be a longlong integer * constant that can be used by the number parser to check the validity * of numeric parameters. */ #define BIG 0x7fffffffffffffffLL /* Used to indicate to number() that a bogus value should cause us to exit */ #define NO_DEFAULT LONG_MIN /* * The *_flag variables are used to indicate that the user specified * the values, rather than that we made them up ourselves. We can * complain about the user giving us bogus values. */ /* semi-constants */ long sectorsize = DEV_BSIZE; /* bytes/sector from param.h */ long bbsize = BBSIZE; /* boot block size */ long sbsize = SBSIZE; /* superblock size */ /* parameters */ diskaddr_t fssize_db; /* file system size in disk blocks */ diskaddr_t fssize_frag; /* file system size in frags */ long cpg; /* cylinders/cylinder group */ int cpg_flag = RC_DEFAULT; long rotdelay = -1; /* rotational delay between blocks */ int rotdelay_flag = RC_DEFAULT; long maxcontig; /* max contiguous blocks to allocate */ int maxcontig_flag = RC_DEFAULT; long nsect = DFLNSECT; /* sectors per track */ int nsect_flag = RC_DEFAULT; long ntrack = DFLNTRAK; /* tracks per cylinder group */ int ntrack_flag = RC_DEFAULT; long bsize = DESBLKSIZE; /* filesystem block size */ int bsize_flag = RC_DEFAULT; long fragsize = DESFRAGSIZE; /* filesystem fragment size */ int fragsize_flag = RC_DEFAULT; long minfree = MINFREE; /* fs_minfree */ int minfree_flag = RC_DEFAULT; long rps = DEFHZ; /* revolutions/second of drive */ int rps_flag = RC_DEFAULT; long nbpi = NBPI; /* number of bytes per inode */ int nbpi_flag = RC_DEFAULT; long nrpos = NRPOS; /* number of rotational positions */ int nrpos_flag = RC_DEFAULT; long apc = 0; /* alternate sectors per cylinder */ int apc_flag = RC_DEFAULT; char opt = 't'; /* optimization style, `t' or `s' */ char mtb = 'n'; /* multi-terabyte format, 'y' or 'n' */ long debug = 0; /* enable debugging output */ int spc_flag = 0; /* alternate sectors specified or */ /* found */ /* global state */ int Nflag; /* do not write to disk */ int mflag; /* return the command line used to create this FS */ int rflag; /* report the superblock in an easily-parsed form */ int Rflag; /* dump the superblock in binary */ char *fsys; time_t mkfstime; char *string; /* * logging support */ int ismdd; /* true if device is a SVM device */ int islog; /* true if ufs or SVM logging is enabled */ int islogok; /* true if ufs/SVM log state is good */ static int isufslog; /* true if ufs logging is enabled */ static int waslog; /* true when ufs logging disabled during grow */ /* * growfs defines, globals, and forward references */ #define NOTENOUGHSPACE 33 int grow; static int Pflag; /* probe to which size the fs can be grown */ int ismounted; char *directory; diskaddr_t grow_fssize; long grow_fs_size; long grow_fs_ncg; diskaddr_t grow_fs_csaddr; long grow_fs_cssize; int grow_fs_clean; struct csum *grow_fscs; diskaddr_t grow_sifrag; int test; int testforce; diskaddr_t testfrags; int inlockexit; int isbad; void lockexit(int); void randomgeneration(void); void checksummarysize(void); void checksblock(void); void growinit(char *); void checkdev(char *, char *); void checkmount(struct mnttab *, char *); struct dinode *gdinode(ino_t); int csfraginrange(daddr32_t); struct csfrag *findcsfrag(daddr32_t, struct csfrag **); void checkindirect(ino_t, daddr32_t *, daddr32_t, int); void addcsfrag(ino_t, daddr32_t, struct csfrag **); void delcsfrag(daddr32_t, struct csfrag **); void checkdirect(ino_t, daddr32_t *, daddr32_t *, int); void findcsfragino(void); void fixindirect(daddr32_t, int); void fixdirect(caddr_t, daddr32_t, daddr32_t *, int); void fixcsfragino(void); void extendsummaryinfo(void); int notenoughspace(void); void unalloccsfragino(void); void unalloccsfragfree(void); void findcsfragfree(void); void copycsfragino(void); void rdcg(long); void wtcg(void); void flcg(void); void allocfrags(long, daddr32_t *, long *); void alloccsfragino(void); void alloccsfragfree(void); void freefrags(daddr32_t, long, long); int findfreerange(long *, long *); void resetallocinfo(void); void extendcg(long); void ulockfs(void); void wlockfs(void); void clockfs(void); void wtsb(void); static int64_t checkfragallocated(daddr32_t); static struct csum *read_summaryinfo(struct fs *); static diskaddr_t probe_summaryinfo(); void main(int argc, char *argv[]) { long i, mincpc, mincpg, ibpcl; long cylno, rpos, blk, j, warn = 0; long mincpgcnt, maxcpg; uint64_t used, bpcg, inospercg; long mapcramped, inodecramped; long postblsize, rotblsize, totalsbsize; FILE *mnttab; struct mnttab mntp; char *special; struct statvfs64 fs; struct dk_cinfo dkcinfo; char pbuf[sizeof (uint64_t) * 3 + 1]; int width, plen; uint64_t num; int c, saverr; diskaddr_t max_fssize; long tmpmaxcontig = -1; struct sigaction sigact; uint64_t nbytes64; int remaining_cg; int do_dot = 0; (void) setlocale(LC_ALL, ""); #if !defined(TEXT_DOMAIN) #define TEXT_DOMAIN "SYS_TEST" #endif (void) textdomain(TEXT_DOMAIN); while ((c = getopt(argc, argv, "F:bmo:VPGM:T:t:")) != EOF) { switch (c) { case 'F': string = optarg; if (strcmp(string, "ufs") != 0) usage(); break; case 'm': /* return command line used to create this FS */ mflag++; break; case 'o': /* * ufs specific options. */ string = optarg; while (*string != '\0') { if (match("nsect=")) { nsect = number(DFLNSECT, "nsect", 0); nsect_flag = RC_KEYWORD; } else if (match("ntrack=")) { ntrack = number(DFLNTRAK, "ntrack", 0); ntrack_flag = RC_KEYWORD; } else if (match("bsize=")) { bsize = number(DESBLKSIZE, "bsize", 0); bsize_flag = RC_KEYWORD; } else if (match("fragsize=")) { fragsize = number(DESFRAGSIZE, "fragsize", 0); fragsize_flag = RC_KEYWORD; } else if (match("cgsize=")) { cpg = number(DESCPG, "cgsize", 0); cpg_flag = RC_KEYWORD; } else if (match("free=")) { minfree = number(MINFREE, "free", ALLOW_PERCENT); minfree_flag = RC_KEYWORD; } else if (match("maxcontig=")) { tmpmaxcontig = number(-1, "maxcontig", 0); maxcontig_flag = RC_KEYWORD; } else if (match("nrpos=")) { nrpos = number(NRPOS, "nrpos", 0); nrpos_flag = RC_KEYWORD; } else if (match("rps=")) { rps = number(DEFHZ, "rps", 0); rps_flag = RC_KEYWORD; } else if (match("nbpi=")) { nbpi = number(NBPI, "nbpi", 0); nbpi_flag = RC_KEYWORD; } else if (match("opt=")) { opt = checkopt(string); } else if (match("mtb=")) { mtb = checkmtb(string); } else if (match("apc=")) { apc = number(0, "apc", 0); apc_flag = RC_KEYWORD; } else if (match("gap=")) { (void) number(0, "gap", ALLOW_MS1); rotdelay = ROTDELAY; rotdelay_flag = RC_DEFAULT; } else if (match("debug=")) { debug = number(0, "debug", 0); } else if (match("N")) { Nflag++; } else if (match("calcsb")) { rflag++; Nflag++; } else if (match("calcbinsb")) { rflag++; Rflag++; Nflag++; } else if (*string == '\0') { break; } else { (void) fprintf(stderr, gettext( "illegal option: %s\n"), string); usage(); } if (*string == ',') string++; if (*string == ' ') string++; } break; case 'V': { char *opt_text; int opt_count; (void) fprintf(stdout, gettext("mkfs -F ufs ")); for (opt_count = 1; opt_count < argc; opt_count++) { opt_text = argv[opt_count]; if (opt_text) (void) fprintf(stdout, " %s ", opt_text); } (void) fprintf(stdout, "\n"); } break; case 'b': /* do nothing for this */ break; case 'M': /* grow the mounted file system */ directory = optarg; /* FALLTHROUGH */ case 'G': /* grow the file system */ grow = 1; break; case 'P': /* probe the file system growing size */ Pflag = 1; grow = 1; /* probe mode implies fs growing */ break; case 'T': /* For testing */ testforce = 1; /* FALLTHROUGH */ case 't': test = 1; string = optarg; testfrags = number(NO_DEFAULT, "testfrags", 0); break; case '?': usage(); break; } } #ifdef MKFS_DEBUG /* * Turning on MKFS_DEBUG causes mkfs to produce a filesystem * that can be reproduced by setting the time to 0 and seeding * the random number generator to a constant. */ mkfstime = 0; /* reproducible results */ #else (void) time(&mkfstime); #endif if (optind >= (argc - 1)) { if (optind > (argc - 1)) { (void) fprintf(stderr, gettext("special not specified\n")); usage(); } else if (mflag == 0) { (void) fprintf(stderr, gettext("size not specified\n")); usage(); } } argc -= optind; argv = &argv[optind]; fsys = argv[0]; fsi = open64(fsys, O_RDONLY); if (fsi < 0) { (void) fprintf(stderr, gettext("%s: cannot open\n"), fsys); lockexit(32); } if (mflag) { dump_fscmd(fsys, fsi); lockexit(0); } /* * The task of setting all of the configuration parameters for a * UFS file system is basically a matter of solving n equations * in m variables. Typically, m is greater than n, so there is * usually more than one valid solution. Since this is usually * an under-constrained problem, it's not always obvious what the * "best" configuration is. * * In general, the approach is to * 1. Determine the values for the file system parameters * that are externally contrained and therefore not adjustable * by mkfs (such as the device's size and maxtransfer size). * 2. Acquire the user's requested setting for all configuration * values that can be set on the command line. * 3. Determine the final value of all configuration values, by * the following approach: * - set the file system block size (fs_bsize). Although * this could be regarded as an adjustable parameter, in * fact, it's pretty much a constant. At this time, it's * generally set to 8k (with older hardware, it can * sometimes make sense to set it to 4k, but those * situations are pretty rare now). * - re-adjust the maximum file system size based on the * value of the file system block size. Since the * frag size can't be any larger than a file system * block, and the number of frags in the file system * has to fit into 31 bits, the file system block size * affects the maximum file system size. * - now that the real maximum file system is known, set the * actual size of the file system to be created to * MIN(requested size, maximum file system size). * - now validate, and if necessary, adjust the following * values: * rotdelay * nsect * maxcontig * apc * frag_size * rps * minfree * nrpos * nrack * nbpi * - calculate maxcpg (the maximum value of the cylinders-per- * cylinder-group configuration parameters). There are two * algorithms for calculating maxcpg: an old one, which is * used for file systems of less than 1 terabyte, and a * new one, implemented in the function compute_maxcpg(), * which is used for file systems of greater than 1 TB. * The difference between them is that compute_maxcpg() * really tries to maximize the cpg value. The old * algorithm fails to take advantage of smaller frags and * lower inode density when determining the maximum cpg, * and thus comes up with much lower numbers in some * configurations. At some point, we might use the * new algorithm for determining maxcpg for all file * systems, but at this time, the changes implemented for * multi-terabyte UFS are NOT being automatically applied * to UFS file systems of less than a terabyte (in the * interest of not changing existing UFS policy too much * until the ramifications of the changes are well-understood * and have been evaluated for their effects on performance.) * - check the current values of the configuration parameters * against the various constraints imposed by UFS. These * include: * * There must be at least one inode in each * cylinder group. * * The cylinder group overhead block, which * contains the inode and frag bigmaps, must fit * within one file system block. * * The space required for inode maps should * occupy no more than a third of the cylinder * group overhead block. * * The rotational position tables have to fit * within the available space in the super block. * Adjust the configuration values that can be adjusted * so that these constraints are satisfied. The * configuration values that are adjustable are: * * frag size * * cylinders per group * * inode density (can be increased) * * number of rotational positions (the rotational * position tables are eliminated altogether if * there isn't enough room for them.) * 4. Set the values for all the dependent configuration * values (those that aren't settable on the command * line and which are completely dependent on the * adjustable parameters). This include cpc (cycles * per cylinder, spc (sectors-per-cylinder), and many others. */ max_fssize = get_max_size(fsi); /* * Get and check positional arguments, if any. */ switch (argc - 1) { default: usage(); /*NOTREACHED*/ case 15: mtb = checkmtb(argv[15]); /* FALLTHROUGH */ case 14: string = argv[14]; tmpmaxcontig = number(-1, "maxcontig", 0); maxcontig_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 13: string = argv[13]; nrpos = number(NRPOS, "nrpos", 0); nrpos_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 12: string = argv[12]; rotdelay = ROTDELAY; rotdelay_flag = RC_DEFAULT; /* FALLTHROUGH */ case 11: string = argv[11]; apc = number(0, "apc", 0); apc_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 10: opt = checkopt(argv[10]); /* FALLTHROUGH */ case 9: string = argv[9]; nbpi = number(NBPI, "nbpi", 0); nbpi_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 8: string = argv[8]; rps = number(DEFHZ, "rps", 0); rps_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 7: string = argv[7]; minfree = number(MINFREE, "free", ALLOW_PERCENT); minfree_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 6: string = argv[6]; cpg = number(DESCPG, "cgsize", 0); cpg_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 5: string = argv[5]; fragsize = number(DESFRAGSIZE, "fragsize", 0); fragsize_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 4: string = argv[4]; bsize = number(DESBLKSIZE, "bsize", 0); bsize_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 3: string = argv[3]; ntrack = number(DFLNTRAK, "ntrack", 0); ntrack_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 2: string = argv[2]; nsect = number(DFLNSECT, "nsect", 0); nsect_flag = RC_POSITIONAL; /* FALLTHROUGH */ case 1: string = argv[1]; fssize_db = number(max_fssize, "size", 0); } if ((maxcontig_flag == RC_DEFAULT) || (tmpmaxcontig == -1) || (maxcontig == -1)) { long maxtrax = get_max_track_size(fsi); maxcontig = maxtrax / bsize; } else { maxcontig = tmpmaxcontig; } if (rotdelay == -1) { /* default by newfs and mkfs */ rotdelay = ROTDELAY; } if (cpg_flag == RC_DEFAULT) { /* If not explicity set, use default */ cpg = DESCPG; } /* * Now that we have the semi-sane args, either positional, via -o, * or by defaulting, handle inter-dependencies and range checks. */ /* * Settle the file system block size first, since it's a fixed * parameter once set and so many other parameters, including * max_fssize, depend on it. */ range_check(&bsize, "bsize", MINBSIZE, MAXBSIZE, DESBLKSIZE, bsize_flag); if (!POWEROF2(bsize)) { (void) fprintf(stderr, gettext("block size must be a power of 2, not %ld\n"), bsize); bsize = DESBLKSIZE; (void) fprintf(stderr, gettext("mkfs: bsize reset to default %ld\n"), bsize); } if (fssize_db > max_fssize && validate_size(fsi, fssize_db)) { (void) fprintf(stderr, gettext( "Warning: the requested size of this file system\n" "(%lld sectors) is greater than the size of the\n" "device reported by the driver (%lld sectors).\n" "However, a read of the device at the requested size\n" "does succeed, so the requested size will be used.\n"), fssize_db, max_fssize); max_fssize = fssize_db; } /* * Since the maximum allocatable unit (the frag) must be less than * or equal to bsize, and the number of frags must be less than or * equal to INT_MAX, the total size of the file system (in * bytes) must be less than or equal to bsize * INT_MAX. */ if (max_fssize > ((diskaddr_t)bsize/DEV_BSIZE) * INT_MAX) max_fssize = ((diskaddr_t)bsize/DEV_BSIZE) * INT_MAX; range_check_64(&fssize_db, "size", 1024LL, max_fssize, max_fssize, 1); if (fssize_db >= SECTORS_PER_TERABYTE) { mtb = 'y'; if (!in_64bit_mode()) { (void) fprintf(stderr, gettext( "mkfs: Warning: Creating a file system greater than 1 terabyte on a\n" " system running a 32-bit kernel. This file system will not be\n" " accessible until the system is rebooted with a 64-bit kernel.\n")); } } /* * 32K based on max block size of 64K, and rotational layout * test of nsect <= (256 * sectors/block). Current block size * limit is not 64K, but it's growing soon. */ range_check(&nsect, "nsect", 1, 32768, DFLNSECT, nsect_flag); range_check(&apc, "apc", 0, nsect - 1, 0, apc_flag); if (mtb == 'y') fragsize = bsize; range_check(&fragsize, "fragsize", sectorsize, bsize, MAX(bsize / MAXFRAG, MIN(DESFRAGSIZE, bsize)), fragsize_flag); if ((bsize / MAXFRAG) > fragsize) { (void) fprintf(stderr, gettext( "fragment size %ld is too small, minimum with block size %ld is %ld\n"), fragsize, bsize, bsize / MAXFRAG); (void) fprintf(stderr, gettext("mkfs: fragsize reset to minimum %ld\n"), bsize / MAXFRAG); fragsize = bsize / MAXFRAG; } if (!POWEROF2(fragsize)) { (void) fprintf(stderr, gettext("fragment size must be a power of 2, not %ld\n"), fragsize); fragsize = MAX(bsize / MAXFRAG, MIN(DESFRAGSIZE, bsize)); (void) fprintf(stderr, gettext("mkfs: fragsize reset to %ld\n"), fragsize); } /* At this point, bsize must be >= fragsize, so no need to check it */ if (bsize < PAGESIZE) { (void) fprintf(stderr, gettext( "WARNING: filesystem block size (%ld) is smaller than " "memory page size (%ld).\nResulting filesystem can not be " "mounted on this system.\n\n"), bsize, (long)PAGESIZE); } range_check(&rps, "rps", 1, 1000, DEFHZ, rps_flag); range_check(&minfree, "free", 0, 99, MINFREE, minfree_flag); range_check(&nrpos, "nrpos", 1, nsect, MIN(nsect, NRPOS), nrpos_flag); /* * ntrack is the number of tracks per cylinder. * The ntrack value must be between 1 and the total number of * sectors in the file system. */ range_check(&ntrack, "ntrack", 1, fssize_db > INT_MAX ? INT_MAX : (uint32_t)fssize_db, DFLNTRAK, ntrack_flag); /* * nbpi is variable, but 2MB seems a reasonable upper limit, * as 4MB tends to cause problems (using otherwise-default * parameters). The true limit is where we end up with one * inode per cylinder group. If this file system is being * configured for multi-terabyte access, nbpi must be at least 1MB. */ if (mtb == 'y' && nbpi < MTB_NBPI) { (void) fprintf(stderr, gettext("mkfs: bad value for nbpi: " "must be at least 1048576 for multi-terabyte, " "nbpi reset to default 1048576\n")); nbpi = MTB_NBPI; } if (mtb == 'y') range_check(&nbpi, "nbpi", MTB_NBPI, 2 * MB, MTB_NBPI, nbpi_flag); else range_check(&nbpi, "nbpi", DEV_BSIZE, 2 * MB, NBPI, nbpi_flag); /* * maxcpg is another variably-limited parameter. Calculate * the limit based on what we've got for its dependent * variables. Effectively, it's how much space is left in the * superblock after all the other bits are accounted for. We * only fill in sblock fields so we can use MAXIpG. * * If the calculation of maxcpg below (for the mtb == 'n' * case) is changed, update newfs as well. * * For old-style, non-MTB format file systems, use the old * algorithm for calculating the maximum cylinder group size, * even though it limits the cylinder group more than necessary. * Since layout can affect performance, we don't want to change * the default layout for non-MTB file systems at this time. * However, for MTB file systems, use the new maxcpg calculation, * which really maxes out the cylinder group size. */ sblock.fs_bsize = bsize; sblock.fs_inopb = sblock.fs_bsize / sizeof (struct dinode); if (mtb == 'n') { maxcpg = (bsize - sizeof (struct cg) - howmany(MAXIpG(&sblock), NBBY)) / (sizeof (long) + nrpos * sizeof (short) + nsect / (MAXFRAG * NBBY)); } else { maxcpg = compute_maxcpg(bsize, fragsize, nbpi, nrpos, nsect * ntrack); } if (cpg == -1) cpg = maxcpg; /* * mincpg is variable in complex ways, so we really can't * do a sane lower-end limit check at this point. */ range_check(&cpg, "cgsize", 1, maxcpg, MIN(maxcpg, DESCPG), cpg_flag); /* * get the controller info */ ismdd = 0; islog = 0; islogok = 0; waslog = 0; if (ioctl(fsi, DKIOCINFO, &dkcinfo) == 0) /* * if it is an MDD (disksuite) device */ if (dkcinfo.dki_ctype == DKC_MD) { ismdd++; /* * check the logging device */ if (ioctl(fsi, _FIOISLOG, NULL) == 0) { islog++; if (ioctl(fsi, _FIOISLOGOK, NULL) == 0) islogok++; } } /* * Do not grow the file system, but print on stdout the maximum * size in sectors to which the file system can be increased. * The calculated size is limited by fssize_db. * Note that we don't lock the filesystem and therefore under rare * conditions (the filesystem is mounted, the free block count is * almost zero, and the superuser is still changing it) the calculated * size can be imprecise. */ if (Pflag) { (void) printf("%llu\n", probe_summaryinfo()); exit(0); } /* * If we're growing an existing filesystem, then we're about * to start doing things that can require recovery efforts if * we get interrupted, so make sure we get a chance to do so. */ if (grow) { sigact.sa_handler = recover_from_sigint; sigemptyset(&sigact.sa_mask); sigact.sa_flags = SA_RESTART; if (sigaction(SIGINT, &sigact, (struct sigaction *)NULL) < 0) { perror(gettext("Could not register SIGINT handler")); lockexit(3); } } if (!Nflag) { /* * Check if MNTTAB is trustable */ if (statvfs64(MNTTAB, &fs) < 0) { (void) fprintf(stderr, gettext("can't statvfs %s\n"), MNTTAB); exit(32); } if (strcmp(MNTTYPE_MNTFS, fs.f_basetype) != 0) { (void) fprintf(stderr, gettext( "%s file system type is not %s, can't mkfs\n"), MNTTAB, MNTTYPE_MNTFS); exit(32); } special = getfullblkname(fsys); checkdev(fsys, special); /* * If we found the block device name, * then check the mount table. * if mounted, and growing write lock the file system * */ if ((special != NULL) && (*special != '\0')) { if ((mnttab = fopen(MNTTAB, "r")) == NULL) { (void) fprintf(stderr, gettext( "can't open %s\n"), MNTTAB); exit(32); } while ((getmntent(mnttab, &mntp)) == NULL) { if (grow) { checkmount(&mntp, special); continue; } if (strcmp(special, mntp.mnt_special) == 0) { (void) fprintf(stderr, gettext( "%s is mounted, can't mkfs\n"), special); exit(32); } } (void) fclose(mnttab); } if (directory && (ismounted == 0)) { (void) fprintf(stderr, gettext("%s is not mounted\n"), special); lockexit(32); } fso = (grow) ? open64(fsys, O_WRONLY) : creat64(fsys, 0666); if (fso < 0) { saverr = errno; (void) fprintf(stderr, gettext("%s: cannot create: %s\n"), fsys, strerror(saverr)); lockexit(32); } } else { /* * For the -N case, a file descriptor is needed for the llseek() * in wtfs(). See the comment in wtfs() for more information. * * Get a file descriptor that's read-only so that this code * doesn't accidentally write to the file. */ fso = open64(fsys, O_RDONLY); if (fso < 0) { saverr = errno; (void) fprintf(stderr, gettext("%s: cannot open: %s\n"), fsys, strerror(saverr)); lockexit(32); } } /* * seed random # generator (for ic_generation) */ #ifdef MKFS_DEBUG srand48(12962); /* reproducible results */ #else srand48((long)(time((time_t *)NULL) + getpid())); #endif if (grow) { growinit(fsys); goto grow00; } /* * Validate the given file system size. * Verify that its last block can actually be accessed. * * Note: it's ok to use sblock as a buffer because it is immediately * overwritten by the rdfs() of the superblock in the next line. * * ToDo: Because the size checking is done in rdfs()/wtfs(), the * error message for specifying an illegal size is very unfriendly. * In the future, one could replace the rdfs()/wtfs() calls * below with in-line calls to read() or write(). This allows better * error messages to be put in place. */ rdfs(fssize_db - 1, (int)sectorsize, (char *)&sblock); /* * make the fs unmountable */ rdfs((diskaddr_t)(SBOFF / sectorsize), (int)sbsize, (char *)&sblock); sblock.fs_magic = -1; sblock.fs_clean = FSBAD; sblock.fs_state = FSOKAY - sblock.fs_time; wtfs((diskaddr_t)(SBOFF / sectorsize), (int)sbsize, (char *)&sblock); bzero(&sblock, (size_t)sbsize); sblock.fs_nsect = nsect; sblock.fs_ntrak = ntrack; /* * Validate specified/determined spc * and calculate minimum cylinders per group. */ /* * sectors/cyl = tracks/cyl * sectors/track */ sblock.fs_spc = sblock.fs_ntrak * sblock.fs_nsect; grow00: if (apc_flag) { sblock.fs_spc -= apc; } /* * Have to test for this separately from apc_flag, due to * the growfs case.... */ if (sblock.fs_spc != sblock.fs_ntrak * sblock.fs_nsect) { spc_flag = 1; } if (grow) goto grow10; sblock.fs_nrpos = nrpos; sblock.fs_bsize = bsize; sblock.fs_fsize = fragsize; sblock.fs_minfree = minfree; grow10: if (nbpi < sblock.fs_fsize) { (void) fprintf(stderr, gettext( "warning: wasteful data byte allocation / inode (nbpi):\n")); (void) fprintf(stderr, gettext( "%ld smaller than allocatable fragment size of %d\n"), nbpi, sblock.fs_fsize); } if (grow) goto grow20; if (opt == 's') sblock.fs_optim = FS_OPTSPACE; else sblock.fs_optim = FS_OPTTIME; sblock.fs_bmask = ~(sblock.fs_bsize - 1); sblock.fs_fmask = ~(sblock.fs_fsize - 1); /* * Planning now for future expansion. */ #if defined(_BIG_ENDIAN) sblock.fs_qbmask.val[0] = 0; sblock.fs_qbmask.val[1] = ~sblock.fs_bmask; sblock.fs_qfmask.val[0] = 0; sblock.fs_qfmask.val[1] = ~sblock.fs_fmask; #endif #if defined(_LITTLE_ENDIAN) sblock.fs_qbmask.val[0] = ~sblock.fs_bmask; sblock.fs_qbmask.val[1] = 0; sblock.fs_qfmask.val[0] = ~sblock.fs_fmask; sblock.fs_qfmask.val[1] = 0; #endif for (sblock.fs_bshift = 0, i = sblock.fs_bsize; i > 1; i >>= 1) sblock.fs_bshift++; for (sblock.fs_fshift = 0, i = sblock.fs_fsize; i > 1; i >>= 1) sblock.fs_fshift++; sblock.fs_frag = numfrags(&sblock, sblock.fs_bsize); for (sblock.fs_fragshift = 0, i = sblock.fs_frag; i > 1; i >>= 1) sblock.fs_fragshift++; if (sblock.fs_frag > MAXFRAG) { (void) fprintf(stderr, gettext( "fragment size %d is too small, minimum with block size %d is %d\n"), sblock.fs_fsize, sblock.fs_bsize, sblock.fs_bsize / MAXFRAG); lockexit(32); } sblock.fs_nindir = sblock.fs_bsize / sizeof (daddr32_t); sblock.fs_inopb = sblock.fs_bsize / sizeof (struct dinode); sblock.fs_nspf = sblock.fs_fsize / sectorsize; for (sblock.fs_fsbtodb = 0, i = NSPF(&sblock); i > 1; i >>= 1) sblock.fs_fsbtodb++; /* * Compute the super-block, cylinder group, and inode blocks. * Note that these "blkno" are really fragment addresses. * For example, on an 8K/1K (block/fragment) system, fs_sblkno is 16, * fs_cblkno is 24, and fs_iblkno is 32. This is why CGSIZE is so * important: only 1 FS block is allocated for the cg struct (fragment * numbers 24 through 31). */ sblock.fs_sblkno = roundup(howmany(bbsize + sbsize, sblock.fs_fsize), sblock.fs_frag); sblock.fs_cblkno = (daddr32_t)(sblock.fs_sblkno + roundup(howmany(sbsize, sblock.fs_fsize), sblock.fs_frag)); sblock.fs_iblkno = sblock.fs_cblkno + sblock.fs_frag; sblock.fs_cgoffset = roundup( howmany(sblock.fs_nsect, NSPF(&sblock)), sblock.fs_frag); for (sblock.fs_cgmask = -1, i = sblock.fs_ntrak; i > 1; i >>= 1) sblock.fs_cgmask <<= 1; if (!POWEROF2(sblock.fs_ntrak)) sblock.fs_cgmask <<= 1; /* * Validate specified/determined spc * and calculate minimum cylinders per group. */ for (sblock.fs_cpc = NSPB(&sblock), i = sblock.fs_spc; sblock.fs_cpc > 1 && (i & 1) == 0; sblock.fs_cpc >>= 1, i >>= 1) /* void */; mincpc = sblock.fs_cpc; /* if these calculations are changed, check dump_fscmd also */ bpcg = (uint64_t)sblock.fs_spc * sectorsize; inospercg = (uint64_t)roundup(bpcg / sizeof (struct dinode), INOPB(&sblock)); if (inospercg > MAXIpG(&sblock)) inospercg = MAXIpG(&sblock); used = (uint64_t)(sblock.fs_iblkno + inospercg / INOPF(&sblock)) * NSPF(&sblock); mincpgcnt = (long)howmany((uint64_t)sblock.fs_cgoffset * (~sblock.fs_cgmask) + used, sblock.fs_spc); mincpg = roundup(mincpgcnt, mincpc); /* * Insure that cylinder group with mincpg has enough space * for block maps */ sblock.fs_cpg = mincpg; sblock.fs_ipg = (int32_t)inospercg; mapcramped = 0; /* * Make sure the cg struct fits within the file system block. * Use larger block sizes until it fits */ while (CGSIZE(&sblock) > sblock.fs_bsize) { mapcramped = 1; if (sblock.fs_bsize < MAXBSIZE) { sblock.fs_bsize <<= 1; if ((i & 1) == 0) { i >>= 1; } else { sblock.fs_cpc <<= 1; mincpc <<= 1; mincpg = roundup(mincpgcnt, mincpc); sblock.fs_cpg = mincpg; } sblock.fs_frag <<= 1; sblock.fs_fragshift += 1; if (sblock.fs_frag <= MAXFRAG) continue; } /* * Looped far enough. The fragment is now as large as the * filesystem block! */ if (sblock.fs_fsize == sblock.fs_bsize) { (void) fprintf(stderr, gettext( "There is no block size that can support this disk\n")); lockexit(32); } /* * Try a larger fragment. Double the fragment size. */ sblock.fs_frag >>= 1; sblock.fs_fragshift -= 1; sblock.fs_fsize <<= 1; sblock.fs_nspf <<= 1; } /* * Insure that cylinder group with mincpg has enough space for inodes */ inodecramped = 0; used *= sectorsize; nbytes64 = (uint64_t)mincpg * bpcg - used; inospercg = (uint64_t)roundup((nbytes64 / nbpi), INOPB(&sblock)); sblock.fs_ipg = (int32_t)inospercg; while (inospercg > MAXIpG(&sblock)) { inodecramped = 1; if (mincpc == 1 || sblock.fs_frag == 1 || sblock.fs_bsize == MINBSIZE) break; nbytes64 = (uint64_t)mincpg * bpcg - used; (void) fprintf(stderr, gettext("With a block size of %d %s %lu\n"), sblock.fs_bsize, gettext("minimum bytes per inode is"), (uint32_t)(nbytes64 / MAXIpG(&sblock) + 1)); sblock.fs_bsize >>= 1; sblock.fs_frag >>= 1; sblock.fs_fragshift -= 1; mincpc >>= 1; sblock.fs_cpg = roundup(mincpgcnt, mincpc); if (CGSIZE(&sblock) > sblock.fs_bsize) { sblock.fs_bsize <<= 1; break; } mincpg = sblock.fs_cpg; nbytes64 = (uint64_t)mincpg * bpcg - used; inospercg = (uint64_t)roundup((nbytes64 / nbpi), INOPB(&sblock)); sblock.fs_ipg = (int32_t)inospercg; } if (inodecramped) { if (inospercg > MAXIpG(&sblock)) { nbytes64 = (uint64_t)mincpg * bpcg - used; (void) fprintf(stderr, gettext( "Minimum bytes per inode is %d\n"), (uint32_t)(nbytes64 / MAXIpG(&sblock) + 1)); } else if (!mapcramped) { (void) fprintf(stderr, gettext( "With %ld bytes per inode, minimum cylinders per group is %ld\n"), nbpi, mincpg); } } if (mapcramped) { (void) fprintf(stderr, gettext( "With %d sectors per cylinder, minimum cylinders " "per group is %ld\n"), sblock.fs_spc, mincpg); } if (inodecramped || mapcramped) { /* * To make this at least somewhat comprehensible in * the world of i18n, figure out what we're going to * say and then say it all at one time. The days of * needing to scrimp on string space are behind us.... */ if ((sblock.fs_bsize != bsize) && (sblock.fs_fsize != fragsize)) { (void) fprintf(stderr, gettext( "This requires the block size to be changed from %ld to %d\n" "and the fragment size to be changed from %ld to %d\n"), bsize, sblock.fs_bsize, fragsize, sblock.fs_fsize); } else if (sblock.fs_bsize != bsize) { (void) fprintf(stderr, gettext( "This requires the block size to be changed from %ld to %d\n"), bsize, sblock.fs_bsize); } else if (sblock.fs_fsize != fragsize) { (void) fprintf(stderr, gettext( "This requires the fragment size to be changed from %ld to %d\n"), fragsize, sblock.fs_fsize); } else { (void) fprintf(stderr, gettext( "Unable to make filesystem fit with the given constraints\n")); } (void) fprintf(stderr, gettext( "Please re-run mkfs with corrected parameters\n")); lockexit(32); } /* * Calculate the number of cylinders per group */ sblock.fs_cpg = cpg; if (sblock.fs_cpg % mincpc != 0) { (void) fprintf(stderr, gettext( "Warning: cylinder groups must have a multiple " "of %ld cylinders with the given\n parameters\n"), mincpc); sblock.fs_cpg = roundup(sblock.fs_cpg, mincpc); (void) fprintf(stderr, gettext("Rounded cgsize up to %d\n"), sblock.fs_cpg); } /* * Must insure there is enough space for inodes */ /* if these calculations are changed, check dump_fscmd also */ nbytes64 = (uint64_t)sblock.fs_cpg * bpcg - used; sblock.fs_ipg = roundup((uint32_t)(nbytes64 / nbpi), INOPB(&sblock)); /* * Slim down cylinders per group, until the inodes can fit. */ while (sblock.fs_ipg > MAXIpG(&sblock)) { inodecramped = 1; sblock.fs_cpg -= mincpc; nbytes64 = (uint64_t)sblock.fs_cpg * bpcg - used; sblock.fs_ipg = roundup((uint32_t)(nbytes64 / nbpi), INOPB(&sblock)); } /* * Must insure there is enough space to hold block map. * Cut down on cylinders per group, until the cg struct fits in a * filesystem block. */ while (CGSIZE(&sblock) > sblock.fs_bsize) { mapcramped = 1; sblock.fs_cpg -= mincpc; nbytes64 = (uint64_t)sblock.fs_cpg * bpcg - used; sblock.fs_ipg = roundup((uint32_t)(nbytes64 / nbpi), INOPB(&sblock)); } sblock.fs_fpg = (sblock.fs_cpg * sblock.fs_spc) / NSPF(&sblock); if ((sblock.fs_cpg * sblock.fs_spc) % NSPB(&sblock) != 0) { (void) fprintf(stderr, gettext("newfs: panic (fs_cpg * fs_spc) %% NSPF != 0\n")); lockexit(32); } if (sblock.fs_cpg < mincpg) { (void) fprintf(stderr, gettext( "With the given parameters, cgsize must be at least %ld; please re-run mkfs\n"), mincpg); lockexit(32); } sblock.fs_cgsize = fragroundup(&sblock, CGSIZE(&sblock)); grow20: /* * Now have size for file system and nsect and ntrak. * Determine number of cylinders and blocks in the file system. */ fssize_frag = (int64_t)dbtofsb(&sblock, fssize_db); if (fssize_frag > INT_MAX) { (void) fprintf(stderr, gettext( "There are too many fragments in the system, increase fragment size\n"), mincpg); lockexit(32); } sblock.fs_size = (int32_t)fssize_frag; sblock.fs_ncyl = (int32_t)(fssize_frag * NSPF(&sblock) / sblock.fs_spc); if (fssize_frag * NSPF(&sblock) > (uint64_t)sblock.fs_ncyl * sblock.fs_spc) { sblock.fs_ncyl++; warn = 1; } if (sblock.fs_ncyl < 1) { (void) fprintf(stderr, gettext( "file systems must have at least one cylinder\n")); lockexit(32); } if (grow) goto grow30; /* * Determine feasability/values of rotational layout tables. * * The size of the rotational layout tables is limited by the size * of the file system block, fs_bsize. The amount of space * available for tables is calculated as (fs_bsize - sizeof (struct * fs)). The size of these tables is inversely proportional to the * block size of the file system. The size increases if sectors per * track are not powers of two, because more cylinders must be * described by the tables before the rotational pattern repeats * (fs_cpc). */ sblock.fs_postblformat = FS_DYNAMICPOSTBLFMT; sblock.fs_sbsize = fragroundup(&sblock, sizeof (struct fs)); sblock.fs_npsect = sblock.fs_nsect; if (sblock.fs_ntrak == 1) { sblock.fs_cpc = 0; goto next; } postblsize = sblock.fs_nrpos * sblock.fs_cpc * sizeof (short); rotblsize = sblock.fs_cpc * sblock.fs_spc / NSPB(&sblock); totalsbsize = sizeof (struct fs) + rotblsize; /* do static allocation if nrpos == 8 and fs_cpc == 16 */ if (sblock.fs_nrpos == 8 && sblock.fs_cpc <= 16) { /* use old static table space */ sblock.fs_postbloff = (char *)(&sblock.fs_opostbl[0][0]) - (char *)(&sblock.fs_link); sblock.fs_rotbloff = &sblock.fs_space[0] - (uchar_t *)(&sblock.fs_link); } else { /* use 4.3 dynamic table space */ sblock.fs_postbloff = &sblock.fs_space[0] - (uchar_t *)(&sblock.fs_link); sblock.fs_rotbloff = sblock.fs_postbloff + postblsize; totalsbsize += postblsize; } if (totalsbsize > sblock.fs_bsize || sblock.fs_nsect > (1 << NBBY) * NSPB(&sblock)) { (void) fprintf(stderr, gettext( "Warning: insufficient space in super block for\n" "rotational layout tables with nsect %d, ntrack %d, " "and nrpos %d.\nOmitting tables - file system " "performance may be impaired.\n"), sblock.fs_nsect, sblock.fs_ntrak, sblock.fs_nrpos); /* * Setting fs_cpc to 0 tells alloccgblk() in ufs_alloc.c to * ignore the positional layout table and rotational * position table. */ sblock.fs_cpc = 0; goto next; } sblock.fs_sbsize = fragroundup(&sblock, totalsbsize); /* * calculate the available blocks for each rotational position */ for (cylno = 0; cylno < sblock.fs_cpc; cylno++) for (rpos = 0; rpos < sblock.fs_nrpos; rpos++) fs_postbl(&sblock, cylno)[rpos] = -1; for (i = (rotblsize - 1) * sblock.fs_frag; i >= 0; i -= sblock.fs_frag) { cylno = cbtocylno(&sblock, i); rpos = cbtorpos(&sblock, i); blk = fragstoblks(&sblock, i); if (fs_postbl(&sblock, cylno)[rpos] == -1) fs_rotbl(&sblock)[blk] = 0; else fs_rotbl(&sblock)[blk] = fs_postbl(&sblock, cylno)[rpos] - blk; fs_postbl(&sblock, cylno)[rpos] = blk; } next: grow30: /* * Compute/validate number of cylinder groups. * Note that if an excessively large filesystem is specified * (e.g., more than 16384 cylinders for an 8K filesystem block), it * does not get detected until checksummarysize() */ sblock.fs_ncg = sblock.fs_ncyl / sblock.fs_cpg; if (sblock.fs_ncyl % sblock.fs_cpg) sblock.fs_ncg++; sblock.fs_dblkno = sblock.fs_iblkno + sblock.fs_ipg / INOPF(&sblock); i = MIN(~sblock.fs_cgmask, sblock.fs_ncg - 1); ibpcl = cgdmin(&sblock, i) - cgbase(&sblock, i); if (ibpcl >= sblock.fs_fpg) { (void) fprintf(stderr, gettext( "inode blocks/cyl group (%d) >= data blocks (%d)\n"), cgdmin(&sblock, i) - cgbase(&sblock, i) / sblock.fs_frag, sblock.fs_fpg / sblock.fs_frag); if ((ibpcl < 0) || (sblock.fs_fpg < 0)) { (void) fprintf(stderr, gettext( "number of cylinders per cylinder group (%d) must be decreased.\n"), sblock.fs_cpg); } else { (void) fprintf(stderr, gettext( "number of cylinders per cylinder group (%d) must be increased.\n"), sblock.fs_cpg); } (void) fprintf(stderr, gettext( "Note that cgsize may have been adjusted to allow struct cg to fit.\n")); lockexit(32); } j = sblock.fs_ncg - 1; if ((i = fssize_frag - j * sblock.fs_fpg) < sblock.fs_fpg && cgdmin(&sblock, j) - cgbase(&sblock, j) > i) { (void) fprintf(stderr, gettext( "Warning: inode blocks/cyl group (%d) >= data " "blocks (%ld) in last\n cylinder group. This " "implies %ld sector(s) cannot be allocated.\n"), (cgdmin(&sblock, j) - cgbase(&sblock, j)) / sblock.fs_frag, i / sblock.fs_frag, i * NSPF(&sblock)); sblock.fs_ncg--; sblock.fs_ncyl -= sblock.fs_ncyl % sblock.fs_cpg; sblock.fs_size = fssize_frag = (int64_t)sblock.fs_ncyl * (int64_t)sblock.fs_spc / (int64_t)NSPF(&sblock); warn = 0; } if (warn && !spc_flag) { (void) fprintf(stderr, gettext( "Warning: %d sector(s) in last cylinder unallocated\n"), sblock.fs_spc - (uint32_t)(fssize_frag * NSPF(&sblock) - (uint64_t)(sblock.fs_ncyl - 1) * sblock.fs_spc)); } /* * fill in remaining fields of the super block */ /* * The csum records are stored in cylinder group 0, starting at * cgdmin, the first data block. */ sblock.fs_csaddr = cgdmin(&sblock, 0); sblock.fs_cssize = fragroundup(&sblock, sblock.fs_ncg * sizeof (struct csum)); i = sblock.fs_bsize / sizeof (struct csum); sblock.fs_csmask = ~(i - 1); for (sblock.fs_csshift = 0; i > 1; i >>= 1) sblock.fs_csshift++; fscs = (struct csum *)calloc(1, sblock.fs_cssize); checksummarysize(); if (mtb == 'y') { sblock.fs_magic = MTB_UFS_MAGIC; sblock.fs_version = MTB_UFS_VERSION_1; } else { sblock.fs_magic = FS_MAGIC; } if (grow) { bcopy((caddr_t)grow_fscs, (caddr_t)fscs, (int)grow_fs_cssize); extendsummaryinfo(); goto grow40; } sblock.fs_rotdelay = rotdelay; sblock.fs_maxcontig = maxcontig; sblock.fs_maxbpg = MAXBLKPG(sblock.fs_bsize); sblock.fs_rps = rps; sblock.fs_cgrotor = 0; sblock.fs_cstotal.cs_ndir = 0; sblock.fs_cstotal.cs_nbfree = 0; sblock.fs_cstotal.cs_nifree = 0; sblock.fs_cstotal.cs_nffree = 0; sblock.fs_fmod = 0; sblock.fs_ronly = 0; sblock.fs_time = mkfstime; sblock.fs_state = FSOKAY - sblock.fs_time; sblock.fs_clean = FSCLEAN; grow40: /* * If all that's needed is a dump of the superblock we * would use by default, we've got it now. So, splat it * out and leave. */ if (rflag) { dump_sblock(); lockexit(0); } /* * Dump out summary information about file system. */ (void) fprintf(stderr, gettext( "%s:\t%lld sectors in %d cylinders of %d tracks, %d sectors\n"), fsys, (uint64_t)sblock.fs_size * NSPF(&sblock), sblock.fs_ncyl, sblock.fs_ntrak, sblock.fs_nsect); (void) fprintf(stderr, gettext( "\t%.1fMB in %d cyl groups (%d c/g, %.2fMB/g, %d i/g)\n"), (float)sblock.fs_size * sblock.fs_fsize / MB, sblock.fs_ncg, sblock.fs_cpg, (float)sblock.fs_fpg * sblock.fs_fsize / MB, sblock.fs_ipg); /* * Now build the cylinders group blocks and * then print out indices of cylinder groups. */ (void) fprintf(stderr, gettext( "super-block backups (for fsck -F ufs -o b=#) at:\n")); for (width = cylno = 0; cylno < sblock.fs_ncg && cylno < 10; cylno++) { if ((grow == 0) || (cylno >= grow_fs_ncg)) initcg(cylno); num = fsbtodb(&sblock, (uint64_t)cgsblock(&sblock, cylno)); (void) sprintf(pbuf, " %llu,", num); plen = strlen(pbuf); if ((width + plen) > (WIDTH - 1)) { width = plen; (void) fprintf(stderr, "\n"); } else { width += plen; } (void) fprintf(stderr, "%s", pbuf); } (void) fprintf(stderr, "\n"); remaining_cg = sblock.fs_ncg - cylno; /* * If there are more than 300 cylinder groups still to be * initialized, print a "." for every 50 cylinder groups. */ if (remaining_cg > 300) { (void) fprintf(stderr, gettext( "Initializing cylinder groups:\n")); do_dot = 1; } /* * Now initialize all cylinder groups between the first ten * and the last ten. * * If the number of cylinder groups was less than 10, all of the * cylinder group offsets would have printed in the last loop * and cylno will already be equal to sblock.fs_ncg and so this * loop will not be entered. If there are less than 20 cylinder * groups, cylno is already less than fs_ncg - 10, so this loop * won't be entered in that case either. */ i = 0; for (; cylno < sblock.fs_ncg - 10; cylno++) { if ((grow == 0) || (cylno >= grow_fs_ncg)) initcg(cylno); if (do_dot && cylno % 50 == 0) { (void) fprintf(stderr, "."); i++; if (i == WIDTH - 1) { (void) fprintf(stderr, "\n"); i = 0; } } } /* * Now print the cylinder group offsets for the last 10 * cylinder groups, if any are left. */ if (do_dot) { (void) fprintf(stderr, gettext( "\nsuper-block backups for last 10 cylinder groups at:\n")); } for (width = 0; cylno < sblock.fs_ncg; cylno++) { if ((grow == 0) || (cylno >= grow_fs_ncg)) initcg(cylno); num = fsbtodb(&sblock, (uint64_t)cgsblock(&sblock, cylno)); (void) sprintf(pbuf, " %llu,", num); plen = strlen(pbuf); if ((width + plen) > (WIDTH - 1)) { width = plen; (void) fprintf(stderr, "\n"); } else { width += plen; } (void) fprintf(stderr, "%s", pbuf); } (void) fprintf(stderr, "\n"); if (Nflag) lockexit(0); if (grow) goto grow50; /* * Now construct the initial file system, * then write out the super-block. */ fsinit(); grow50: /* * write the superblock and csum information */ wtsb(); /* * extend the last cylinder group in the original file system */ if (grow) { extendcg(grow_fs_ncg-1); wtsb(); } /* * Write out the duplicate super blocks to the first 10 * cylinder groups (or fewer, if there are fewer than 10 * cylinder groups). */ for (cylno = 0; cylno < sblock.fs_ncg && cylno < 10; cylno++) awtfs(fsbtodb(&sblock, (uint64_t)cgsblock(&sblock, cylno)), (int)sbsize, (char *)&sblock, SAVE); /* * Now write out duplicate super blocks to the remaining * cylinder groups. In the case of multi-terabyte file * systems, just write out the super block to the last ten * cylinder groups (or however many are left). */ if (mtb == 'y') { if (sblock.fs_ncg <= 10) cylno = sblock.fs_ncg; else if (sblock.fs_ncg <= 20) cylno = 10; else cylno = sblock.fs_ncg - 10; } for (; cylno < sblock.fs_ncg; cylno++) awtfs(fsbtodb(&sblock, (uint64_t)cgsblock(&sblock, cylno)), (int)sbsize, (char *)&sblock, SAVE); /* * Flush out all the AIO writes we've done. It's not * necessary to do this explicitly, but it's the only * way to report any errors from those writes. */ flush_writes(); /* * set clean flag */ if (grow) sblock.fs_clean = grow_fs_clean; else sblock.fs_clean = FSCLEAN; sblock.fs_time = mkfstime; sblock.fs_state = FSOKAY - sblock.fs_time; wtfs((diskaddr_t)(SBOFF / sectorsize), sbsize, (char *)&sblock); isbad = 0; if (fsync(fso) == -1) { saverr = errno; (void) fprintf(stderr, gettext("mkfs: fsync failed on write disk: %s\n"), strerror(saverr)); /* we're just cleaning up, so keep going */ } if (close(fsi) == -1) { saverr = errno; (void) fprintf(stderr, gettext("mkfs: close failed on read disk: %s\n"), strerror(saverr)); /* we're just cleaning up, so keep going */ } if (close(fso) == -1) { saverr = errno; (void) fprintf(stderr, gettext("mkfs: close failed on write disk: %s\n"), strerror(saverr)); /* we're just cleaning up, so keep going */ } fsi = fso = -1; #ifndef STANDALONE lockexit(0); #endif } /* * Figure out how big the partition we're dealing with is. * The value returned is in disk blocks (sectors); */ static diskaddr_t get_max_size(int fd) { struct vtoc vtoc; dk_gpt_t *efi_vtoc; int is_efi = 0; diskaddr_t slicesize; int index = read_vtoc(fd, &vtoc); if (index < 0) { if (index == VT_ENOTSUP || index == VT_ERROR) { /* it might be an EFI label */ is_efi = 1; index = efi_alloc_and_read(fd, &efi_vtoc); } } if (index < 0) { switch (index) { case VT_ERROR: break; case VT_EIO: errno = EIO; break; case VT_EINVAL: errno = EINVAL; } perror(gettext("Can not determine partition size")); lockexit(32); } if (is_efi) { slicesize = efi_vtoc->efi_parts[index].p_size; efi_free(efi_vtoc); } else { /* * In the vtoc struct, p_size is a 32-bit signed quantity. * In the dk_gpt struct (efi's version of the vtoc), p_size * is an unsigned 64-bit quantity. By casting the vtoc's * psize to an unsigned 32-bit quantity, it will be copied * to 'slicesize' (an unsigned 64-bit diskaddr_t) without * sign extension. */ slicesize = (uint32_t)vtoc.v_part[index].p_size; } if (debug) { (void) fprintf(stderr, "get_max_size: index = %d, p_size = %lld, dolimit = %d\n", index, slicesize, (slicesize > FS_MAX)); } /* * The next line limits a UFS file system to the maximum * supported size. */ if (slicesize > FS_MAX) return (FS_MAX); return (slicesize); } static long get_max_track_size(int fd) { struct dk_cinfo ci; long track_size = -1; if (ioctl(fd, DKIOCINFO, &ci) == 0) { track_size = ci.dki_maxtransfer * DEV_BSIZE; } if ((track_size < 0)) { int error = 0; int maxphys; int gotit = 0; gotit = fsgetmaxphys(&maxphys, &error); if (gotit) { track_size = MIN(MB, maxphys); } else { (void) fprintf(stderr, gettext( "Warning: Could not get system value for maxphys. The value for\n" "maxcontig will default to 1MB.\n")); track_size = MB; } } return (track_size); } /* * Initialize a cylinder group. */ static void initcg(int cylno) { diskaddr_t cbase, d; diskaddr_t dlower; /* last data block before cg metadata */ diskaddr_t dupper; /* first data block after cg metadata */ diskaddr_t dmax; int64_t i; struct csum *cs; struct dinode *inode_buffer; int size; /* * Variables used to store intermediate results as a part of * the internal implementation of the cbtocylno() macros. */ diskaddr_t bno; /* UFS block number (not sector number) */ int cbcylno; /* current cylinder number */ int cbcylno_sect; /* sector offset within cylinder */ int cbsect_incr; /* amount to increment sector offset */ /* * Variables used to store intermediate results as a part of * the internal implementation of the cbtorpos() macros. */ short *cgblks; /* pointer to array of free blocks in cg */ int trackrpos; /* tmp variable for rotation position */ int trackoff; /* offset within a track */ int trackoff_incr; /* amount to increment trackoff */ int rpos; /* rotation position of current block */ int rpos_incr; /* amount to increment rpos per block */ union cgun *icgun; /* local pointer to a cg summary block */ #define icg (icgun->cg) icgun = (union cgun *)getbuf(&cgsumbuf, sizeof (union cgun)); /* * Determine block bounds for cylinder group. * Allow space for super block summary information in first * cylinder group. */ cbase = cgbase(&sblock, cylno); dmax = cbase + sblock.fs_fpg; if (dmax > sblock.fs_size) /* last cg may be smaller than normal */ dmax = sblock.fs_size; dlower = cgsblock(&sblock, cylno) - cbase; dupper = cgdmin(&sblock, cylno) - cbase; if (cylno == 0) dupper += howmany(sblock.fs_cssize, sblock.fs_fsize); cs = fscs + cylno; icg.cg_time = mkfstime; icg.cg_magic = CG_MAGIC; icg.cg_cgx = cylno; /* last one gets whatever's left */ if (cylno == sblock.fs_ncg - 1) icg.cg_ncyl = sblock.fs_ncyl - (sblock.fs_cpg * cylno); else icg.cg_ncyl = sblock.fs_cpg; icg.cg_niblk = sblock.fs_ipg; icg.cg_ndblk = dmax - cbase; icg.cg_cs.cs_ndir = 0; icg.cg_cs.cs_nffree = 0; icg.cg_cs.cs_nbfree = 0; icg.cg_cs.cs_nifree = 0; icg.cg_rotor = 0; icg.cg_frotor = 0; icg.cg_irotor = 0; icg.cg_btotoff = &icg.cg_space[0] - (uchar_t *)(&icg.cg_link); icg.cg_boff = icg.cg_btotoff + sblock.fs_cpg * sizeof (long); icg.cg_iusedoff = icg.cg_boff + sblock.fs_cpg * sblock.fs_nrpos * sizeof (short); icg.cg_freeoff = icg.cg_iusedoff + howmany(sblock.fs_ipg, NBBY); icg.cg_nextfreeoff = icg.cg_freeoff + howmany(sblock.fs_cpg * sblock.fs_spc / NSPF(&sblock), NBBY); for (i = 0; i < sblock.fs_frag; i++) { icg.cg_frsum[i] = 0; } bzero((caddr_t)cg_inosused(&icg), icg.cg_freeoff - icg.cg_iusedoff); icg.cg_cs.cs_nifree += sblock.fs_ipg; if (cylno == 0) for (i = 0; i < UFSROOTINO; i++) { setbit(cg_inosused(&icg), i); icg.cg_cs.cs_nifree--; } /* * Initialize all the inodes in the cylinder group using * random numbers. */ size = sblock.fs_ipg * sizeof (struct dinode); inode_buffer = (struct dinode *)getbuf(&inodebuf, size); for (i = 0; i < sblock.fs_ipg; i++) { IRANDOMIZE(&(inode_buffer[i].di_ic)); } /* * Write all inodes in a single write for performance. */ awtfs(fsbtodb(&sblock, (uint64_t)cgimin(&sblock, cylno)), (int)size, (char *)inode_buffer, RELEASE); bzero((caddr_t)cg_blktot(&icg), icg.cg_boff - icg.cg_btotoff); bzero((caddr_t)cg_blks(&sblock, &icg, 0), icg.cg_iusedoff - icg.cg_boff); bzero((caddr_t)cg_blksfree(&icg), icg.cg_nextfreeoff - icg.cg_freeoff); if (cylno > 0) { for (d = 0; d < dlower; d += sblock.fs_frag) { setblock(&sblock, cg_blksfree(&icg), d/sblock.fs_frag); icg.cg_cs.cs_nbfree++; cg_blktot(&icg)[cbtocylno(&sblock, d)]++; cg_blks(&sblock, &icg, cbtocylno(&sblock, d)) [cbtorpos(&sblock, d)]++; } sblock.fs_dsize += dlower; } sblock.fs_dsize += icg.cg_ndblk - dupper; if ((i = dupper % sblock.fs_frag) != 0) { icg.cg_frsum[sblock.fs_frag - i]++; for (d = dupper + sblock.fs_frag - i; dupper < d; dupper++) { setbit(cg_blksfree(&icg), dupper); icg.cg_cs.cs_nffree++; } } /* * WARNING: The following code is somewhat confusing, but * results in a substantial performance improvement in mkfs. * * Instead of using cbtocylno() and cbtorpos() macros, we * keep track of all the intermediate state of those macros * in some variables. This allows simple addition to be * done to calculate the results as we step through the * blocks in an orderly fashion instead of the slower * multiplication and division the macros are forced to * used so they can support random input. (Multiplication, * division, and remainder operations typically take about * 10x as many processor cycles as other operations.) * * The basic idea is to take code: * * for (x = starting_x; x < max; x++) * y = (x * c) / z * * and rewrite it to take advantage of the fact that * the variable x is incrementing in an orderly way: * * intermediate = starting_x * c * yval = intermediate / z * for (x = starting_x; x < max; x++) { * y = yval; * intermediate += c * if (intermediate > z) { * yval++; * intermediate -= z * } * } * * Performance has improved as much as 4X using this code. */ /* * Initialize the starting points for all the cbtocylno() * macro variables and figure out the increments needed each * time through the loop. */ cbcylno_sect = dupper * NSPF(&sblock); cbsect_incr = sblock.fs_frag * NSPF(&sblock); cbcylno = cbcylno_sect / sblock.fs_spc; cbcylno_sect %= sblock.fs_spc; cgblks = cg_blks(&sblock, &icg, cbcylno); bno = dupper / sblock.fs_frag; /* * Initialize the starting points for all the cbtorpos() * macro variables and figure out the increments needed each * time through the loop. * * It's harder to simplify the cbtorpos() macro if there were * alternate sectors specified (or if they previously existed * in the growfs case). Since this is rare, we just revert to * using the macros in this case and skip the variable setup. */ if (!spc_flag) { trackrpos = (cbcylno_sect % sblock.fs_nsect) * sblock.fs_nrpos; rpos = trackrpos / sblock.fs_nsect; trackoff = trackrpos % sblock.fs_nsect; trackoff_incr = cbsect_incr * sblock.fs_nrpos; rpos_incr = (trackoff_incr / sblock.fs_nsect) % sblock.fs_nrpos; trackoff_incr = trackoff_incr % sblock.fs_nsect; } /* * Loop through all the blocks, marking them free and * updating totals kept in the superblock and cg summary. */ for (d = dupper; d + sblock.fs_frag <= dmax - cbase; ) { setblock(&sblock, cg_blksfree(&icg), bno); icg.cg_cs.cs_nbfree++; cg_blktot(&icg)[cbcylno]++; if (!spc_flag) cgblks[rpos]++; else cg_blks(&sblock, &icg, cbtocylno(&sblock, d)) [cbtorpos(&sblock, d)]++; d += sblock.fs_frag; bno++; /* * Increment the sector offset within the cylinder * for the cbtocylno() macro reimplementation. If * we're beyond the end of the cylinder, update the * cylinder number, calculate the offset in the * new cylinder, and update the cgblks pointer * to the next rotational position. */ cbcylno_sect += cbsect_incr; if (cbcylno_sect >= sblock.fs_spc) { cbcylno++; cbcylno_sect -= sblock.fs_spc; cgblks += sblock.fs_nrpos; } /* * If there aren't alternate sectors, increment the * rotational position variables for the cbtorpos() * reimplementation. Note that we potentially * increment rpos twice. Once by rpos_incr, and one * more time when we wrap to a new track because * trackoff >= fs_nsect. */ if (!spc_flag) { trackoff += trackoff_incr; rpos += rpos_incr; if (trackoff >= sblock.fs_nsect) { trackoff -= sblock.fs_nsect; rpos++; } if (rpos >= sblock.fs_nrpos) rpos -= sblock.fs_nrpos; } } if (d < dmax - cbase) { icg.cg_frsum[dmax - cbase - d]++; for (; d < dmax - cbase; d++) { setbit(cg_blksfree(&icg), d); icg.cg_cs.cs_nffree++; } } sblock.fs_cstotal.cs_ndir += icg.cg_cs.cs_ndir; sblock.fs_cstotal.cs_nffree += icg.cg_cs.cs_nffree; sblock.fs_cstotal.cs_nbfree += icg.cg_cs.cs_nbfree; sblock.fs_cstotal.cs_nifree += icg.cg_cs.cs_nifree; *cs = icg.cg_cs; awtfs(fsbtodb(&sblock, (uint64_t)cgtod(&sblock, cylno)), sblock.fs_bsize, (char *)&icg, RELEASE); } /* * initialize the file system */ struct inode node; #define LOSTDIR #ifdef LOSTDIR #define PREDEFDIR 3 #else #define PREDEFDIR 2 #endif struct direct root_dir[] = { { UFSROOTINO, sizeof (struct direct), 1, "." }, { UFSROOTINO, sizeof (struct direct), 2, ".." }, #ifdef LOSTDIR { LOSTFOUNDINO, sizeof (struct direct), 10, "lost+found" }, #endif }; #ifdef LOSTDIR struct direct lost_found_dir[] = { { LOSTFOUNDINO, sizeof (struct direct), 1, "." }, { UFSROOTINO, sizeof (struct direct), 2, ".." }, { 0, DIRBLKSIZ, 0, 0 }, }; #endif char buf[MAXBSIZE]; static void fsinit() { int i; /* * initialize the node */ node.i_atime = mkfstime; node.i_mtime = mkfstime; node.i_ctime = mkfstime; #ifdef LOSTDIR /* * create the lost+found directory */ (void) makedir(lost_found_dir, 2); for (i = DIRBLKSIZ; i < sblock.fs_bsize; i += DIRBLKSIZ) { bcopy(&lost_found_dir[2], &buf[i], DIRSIZ(&lost_found_dir[2])); } node.i_number = LOSTFOUNDINO; node.i_smode = node.i_mode = IFDIR | 0700; node.i_nlink = 2; node.i_size = sblock.fs_bsize; node.i_db[0] = alloc((int)node.i_size, node.i_mode); node.i_blocks = btodb(fragroundup(&sblock, (int)node.i_size)); IRANDOMIZE(&node.i_ic); wtfs(fsbtodb(&sblock, (uint64_t)node.i_db[0]), (int)node.i_size, buf); iput(&node); #endif /* * create the root directory */ node.i_number = UFSROOTINO; node.i_mode = node.i_smode = IFDIR | UMASK; node.i_nlink = PREDEFDIR; node.i_size = makedir(root_dir, PREDEFDIR); node.i_db[0] = alloc(sblock.fs_fsize, node.i_mode); /* i_size < 2GB because we are initializing the file system */ node.i_blocks = btodb(fragroundup(&sblock, (int)node.i_size)); IRANDOMIZE(&node.i_ic); wtfs(fsbtodb(&sblock, (uint64_t)node.i_db[0]), sblock.fs_fsize, buf); iput(&node); } /* * construct a set of directory entries in "buf". * return size of directory. */ static int makedir(struct direct *protodir, int entries) { char *cp; int i; ushort_t spcleft; spcleft = DIRBLKSIZ; for (cp = buf, i = 0; i < entries - 1; i++) { protodir[i].d_reclen = DIRSIZ(&protodir[i]); bcopy(&protodir[i], cp, protodir[i].d_reclen); cp += protodir[i].d_reclen; spcleft -= protodir[i].d_reclen; } protodir[i].d_reclen = spcleft; bcopy(&protodir[i], cp, DIRSIZ(&protodir[i])); return (DIRBLKSIZ); } /* * allocate a block or frag */ static daddr32_t alloc(int size, int mode) { int i, frag; daddr32_t d; rdfs(fsbtodb(&sblock, (uint64_t)cgtod(&sblock, 0)), sblock.fs_cgsize, (char *)&acg); if (acg.cg_magic != CG_MAGIC) { (void) fprintf(stderr, gettext("cg 0: bad magic number\n")); lockexit(32); } if (acg.cg_cs.cs_nbfree == 0) { (void) fprintf(stderr, gettext("first cylinder group ran out of space\n")); lockexit(32); } for (d = 0; d < acg.cg_ndblk; d += sblock.fs_frag) if (isblock(&sblock, cg_blksfree(&acg), d / sblock.fs_frag)) goto goth; (void) fprintf(stderr, gettext("internal error: can't find block in cyl 0\n")); lockexit(32); goth: clrblock(&sblock, cg_blksfree(&acg), d / sblock.fs_frag); acg.cg_cs.cs_nbfree--; sblock.fs_cstotal.cs_nbfree--; fscs[0].cs_nbfree--; if (mode & IFDIR) { acg.cg_cs.cs_ndir++; sblock.fs_cstotal.cs_ndir++; fscs[0].cs_ndir++; } cg_blktot(&acg)[cbtocylno(&sblock, d)]--; cg_blks(&sblock, &acg, cbtocylno(&sblock, d))[cbtorpos(&sblock, d)]--; if (size != sblock.fs_bsize) { frag = howmany(size, sblock.fs_fsize); fscs[0].cs_nffree += sblock.fs_frag - frag; sblock.fs_cstotal.cs_nffree += sblock.fs_frag - frag; acg.cg_cs.cs_nffree += sblock.fs_frag - frag; acg.cg_frsum[sblock.fs_frag - frag]++; for (i = frag; i < sblock.fs_frag; i++) setbit(cg_blksfree(&acg), d + i); } wtfs(fsbtodb(&sblock, (uint64_t)cgtod(&sblock, 0)), sblock.fs_cgsize, (char *)&acg); return (d); } /* * Allocate an inode on the disk */ static void iput(struct inode *ip) { struct dinode buf[MAXINOPB]; diskaddr_t d; rdfs(fsbtodb(&sblock, (uint64_t)cgtod(&sblock, 0)), sblock.fs_cgsize, (char *)&acg); if (acg.cg_magic != CG_MAGIC) { (void) fprintf(stderr, gettext("cg 0: bad magic number\n")); lockexit(32); } acg.cg_cs.cs_nifree--; setbit(cg_inosused(&acg), ip->i_number); wtfs(fsbtodb(&sblock, (uint64_t)cgtod(&sblock, 0)), sblock.fs_cgsize, (char *)&acg); sblock.fs_cstotal.cs_nifree--; fscs[0].cs_nifree--; if ((int)ip->i_number >= sblock.fs_ipg * sblock.fs_ncg) { (void) fprintf(stderr, gettext("fsinit: inode value out of range (%d).\n"), ip->i_number); lockexit(32); } d = fsbtodb(&sblock, (uint64_t)itod(&sblock, (int)ip->i_number)); rdfs(d, sblock.fs_bsize, (char *)buf); buf[itoo(&sblock, (int)ip->i_number)].di_ic = ip->i_ic; wtfs(d, sblock.fs_bsize, (char *)buf); } /* * getbuf() -- Get a buffer for use in an AIO operation. Buffer * is zero'd the first time returned, left with whatever * was in memory after that. This function actually gets * enough memory the first time it's called to support * MAXBUF buffers like a slab allocator. When all the * buffers are in use, it waits for an aio to complete * and make a buffer available. * * Never returns an error. Either succeeds or exits. */ static char * getbuf(bufhdr *bufhead, int size) { bufhdr *pbuf; bufhdr *prev; int i; int buf_size, max_bufs; /* * Initialize all the buffers */ if (bufhead->head == NULL) { /* * round up the size of our buffer header to a * 16 byte boundary so the address we return to * the caller is "suitably aligned". */ bufhdrsize = (sizeof (bufhdr) + 15) & ~15; /* * Add in our header to the buffer and round it all up to * a 16 byte boundry so each member of the slab is aligned. */ buf_size = (size + bufhdrsize + 15) & ~15; /* * Limit number of buffers to lesser of MAXBUFMEM's worth * or MAXBUF, whichever is less. */ max_bufs = MAXBUFMEM / buf_size; if (max_bufs > MAXBUF) max_bufs = MAXBUF; pbuf = (bufhdr *)calloc(max_bufs, buf_size); if (pbuf == NULL) { perror("calloc"); lockexit(32); } bufhead->head = bufhead; prev = bufhead; for (i = 0; i < max_bufs; i++) { pbuf->head = bufhead; prev->next = pbuf; prev = pbuf; pbuf = (bufhdr *)((char *)pbuf + buf_size); } } /* * Get an available buffer, waiting for I/O if necessary */ wait_for_write(NOBLOCK); while (bufhead->next == NULL) wait_for_write(BLOCK); /* * Take the buffer off the list */ pbuf = bufhead->next; bufhead->next = pbuf->next; pbuf->next = NULL; /* * return the empty buffer space just past the header */ return ((char *)pbuf + bufhdrsize); } /* * freebuf() -- Free a buffer gotten previously through getbuf. * Puts the buffer back on the appropriate list for * later use. Never calls free(). * * Assumes that SIGINT is blocked. */ static void freebuf(char *buf) { bufhdr *pbuf; bufhdr *bufhead; /* * get the header for this buffer */ pbuf = (bufhdr *)(buf - bufhdrsize); /* * Put it back on the list of available buffers */ bufhead = pbuf->head; pbuf->next = bufhead->next; bufhead->next = pbuf; } /* * freetrans() -- Free a transaction gotten previously through getaiop. * Puts the transaction struct back on the appropriate list for * later use. Never calls free(). * * Assumes that SIGINT is blocked. */ static void freetrans(aio_trans *transp) { /* * free the buffer associated with this AIO if needed */ if (transp->release == RELEASE) freebuf(transp->buffer); /* * Put transaction on the free list */ transp->next = results.trans; results.trans = transp; } /* * wait_for_write() -- Wait for an aio write to complete. Return * the transaction structure for that write. * * Blocks SIGINT if necessary. */ aio_trans * wait_for_write(int block) { aio_trans *transp; aio_result_t *resultp; static struct timeval zero_wait = { 0, 0 }; sigset_t old_mask; /* * If we know there aren't any outstanding transactions, just return */ if (results.outstanding == 0) return ((aio_trans *) 0); block_sigint(&old_mask); resultp = aiowait(block ? NULL : &zero_wait); if (resultp == NULL || (resultp == (aio_result_t *)-1 && errno == EINVAL)) { unblock_sigint(&old_mask); return ((aio_trans *) 0); } results.outstanding--; transp = (aio_trans *)resultp; if (resultp->aio_return != transp->size) { if (resultp->aio_return == -1) { /* * The aiowrite() may have failed because the * kernel didn't have enough memory to do the job. * Flush all pending writes and try a normal * write(). wtfs_breakup() will call exit if it * fails, so we don't worry about errors here. */ flush_writes(); wtfs_breakup(transp->bno, transp->size, transp->buffer); } else { (void) fprintf(stderr, gettext( "short write (%d of %d bytes) on sector %lld\n"), resultp->aio_return, transp->size, transp->bno); /* * Don't unblock SIGINT, to avoid potential * looping due to queued interrupts and * error handling. */ lockexit(32); } } resultp->aio_return = 0; freetrans(transp); unblock_sigint(&old_mask); return (transp); } /* * flush_writes() -- flush all the outstanding aio writes. */ static void flush_writes(void) { while (wait_for_write(BLOCK)) ; } /* * get_aiop() -- find and return an aio_trans structure on which a new * aio can be done. Blocks on aiowait() if needed. Reaps * all outstanding completed aio's. * * Assumes that SIGINT is blocked. */ aio_trans * get_aiop() { int i; aio_trans *transp; aio_trans *prev; /* * initialize aio stuff */ if (!aio_inited) { aio_inited = 1; results.maxpend = 0; results.outstanding = 0; results.max = MAXAIO; results.trans = (aio_trans *)calloc(results.max, sizeof (aio_trans)); if (results.trans == NULL) { perror("calloc"); lockexit(32); } /* * Initialize the linked list of aio transaction * structures. Note that the final "next" pointer * will be NULL since we got the buffer from calloc(). */ prev = results.trans; for (i = 1; i < results.max; i++) { prev->next = &(results.trans[i]); prev = prev->next; } } wait_for_write(NOBLOCK); while (results.trans == NULL) wait_for_write(BLOCK); transp = results.trans; results.trans = results.trans->next; transp->next = 0; transp->resultbuf.aio_return = AIO_INPROGRESS; return (transp); } /* * read a block from the file system */ static void rdfs(diskaddr_t bno, int size, char *bf) { int n, saverr; /* * In case we need any data that's pending in an aiowrite(), * we wait for them all to complete before doing a read. */ flush_writes(); /* * Note: the llseek() can succeed, even if the offset is out of range. * It's not until the file i/o operation (the read()) that one knows * for sure if the raw device can handle the offset. */ if (llseek(fsi, (offset_t)bno * sectorsize, 0) < 0) { saverr = errno; (void) fprintf(stderr, gettext("seek error on sector %lld: %s\n"), bno, strerror(saverr)); lockexit(32); } n = read(fsi, bf, size); if (n != size) { saverr = errno; if (n == -1) (void) fprintf(stderr, gettext("read error on sector %lld: %s\n"), bno, strerror(saverr)); else (void) fprintf(stderr, gettext( "short read (%d of %d bytes) on sector %lld\n"), n, size, bno); lockexit(32); } } /* * write a block to the file system */ static void wtfs(diskaddr_t bno, int size, char *bf) { int n, saverr; if (fso == -1) return; /* * Note: the llseek() can succeed, even if the offset is out of range. * It's not until the file i/o operation (the write()) that one knows * for sure if the raw device can handle the offset. */ if (llseek(fso, (offset_t)bno * sectorsize, 0) < 0) { saverr = errno; (void) fprintf(stderr, gettext("seek error on sector %lld: %s\n"), bno, strerror(saverr)); lockexit(32); } if (Nflag) return; n = write(fso, bf, size); if (n != size) { saverr = errno; if (n == -1) (void) fprintf(stderr, gettext("write error on sector %lld: %s\n"), bno, strerror(saverr)); else (void) fprintf(stderr, gettext( "short write (%d of %d bytes) on sector %lld\n"), n, size, bno); lockexit(32); } } /* * write a block to the file system -- buffered with aio */ static void awtfs(diskaddr_t bno, int size, char *bf, int release) { int n; aio_trans *transp; sigset_t old_mask; if (fso == -1) return; /* * We need to keep things consistent if we get interrupted, * so defer any expected interrupts for the time being. */ block_sigint(&old_mask); if (Nflag) { if (release == RELEASE) freebuf(bf); } else { transp = get_aiop(); transp->bno = bno; transp->buffer = bf; transp->size = size; transp->release = release; n = aiowrite(fso, bf, size, (off_t)bno * sectorsize, SEEK_SET, &transp->resultbuf); if (n < 0) { /* * The aiowrite() may have failed because the * kernel didn't have enough memory to do the job. * Flush all pending writes and try a normal * write(). wtfs_breakup() will call exit if it * fails, so we don't worry about errors here. */ flush_writes(); wtfs_breakup(transp->bno, transp->size, transp->buffer); freetrans(transp); } else { /* * Keep track of our pending writes. */ results.outstanding++; if (results.outstanding > results.maxpend) results.maxpend = results.outstanding; } } unblock_sigint(&old_mask); } /* * write a block to the file system, but break it up into sbsize * chunks to avoid forcing a large amount of memory to be locked down. * Only used as a fallback when an aio write has failed. */ static void wtfs_breakup(diskaddr_t bno, int size, char *bf) { int n, saverr; int wsize; int block_incr = sbsize / sectorsize; if (size < sbsize) wsize = size; else wsize = sbsize; n = 0; while (size) { /* * Note: the llseek() can succeed, even if the offset is * out of range. It's not until the file i/o operation * (the write()) that one knows for sure if the raw device * can handle the offset. */ if (llseek(fso, (offset_t)bno * sectorsize, 0) < 0) { saverr = errno; (void) fprintf(stderr, gettext("seek error on sector %lld: %s\n"), bno, strerror(saverr)); lockexit(32); } n = write(fso, bf, wsize); if (n == -1) { saverr = errno; (void) fprintf(stderr, gettext("write error on sector %lld: %s\n"), bno, strerror(saverr)); lockexit(32); } if (n != wsize) { saverr = errno; (void) fprintf(stderr, gettext( "short write (%d of %d bytes) on sector %lld\n"), n, size, bno); lockexit(32); } bno += block_incr; bf += wsize; size -= wsize; if (size < wsize) wsize = size; } } /* * check if a block is available */ static int isblock(struct fs *fs, unsigned char *cp, int h) { unsigned char mask; switch (fs->fs_frag) { case 8: return (cp[h] == 0xff); case 4: mask = 0x0f << ((h & 0x1) << 2); return ((cp[h >> 1] & mask) == mask); case 2: mask = 0x03 << ((h & 0x3) << 1); return ((cp[h >> 2] & mask) == mask); case 1: mask = 0x01 << (h & 0x7); return ((cp[h >> 3] & mask) == mask); default: (void) fprintf(stderr, "isblock bad fs_frag %d\n", fs->fs_frag); return (0); } } /* * take a block out of the map */ static void clrblock(struct fs *fs, unsigned char *cp, int h) { switch ((fs)->fs_frag) { case 8: cp[h] = 0; return; case 4: cp[h >> 1] &= ~(0x0f << ((h & 0x1) << 2)); return; case 2: cp[h >> 2] &= ~(0x03 << ((h & 0x3) << 1)); return; case 1: cp[h >> 3] &= ~(0x01 << (h & 0x7)); return; default: (void) fprintf(stderr, gettext("clrblock: bad fs_frag value %d\n"), fs->fs_frag); return; } } /* * put a block into the map */ static void setblock(struct fs *fs, unsigned char *cp, int h) { switch (fs->fs_frag) { case 8: cp[h] = 0xff; return; case 4: cp[h >> 1] |= (0x0f << ((h & 0x1) << 2)); return; case 2: cp[h >> 2] |= (0x03 << ((h & 0x3) << 1)); return; case 1: cp[h >> 3] |= (0x01 << (h & 0x7)); return; default: (void) fprintf(stderr, gettext("setblock: bad fs_frag value %d\n"), fs->fs_frag); return; } } static void usage() { (void) fprintf(stderr, gettext("ufs usage: mkfs [-F FSType] [-V] [-m] [-o options] " "special " /* param 0 */ "size(sectors) \\ \n")); /* param 1 */ (void) fprintf(stderr, "[nsect " /* param 2 */ "ntrack " /* param 3 */ "bsize " /* param 4 */ "fragsize " /* param 5 */ "cpg " /* param 6 */ "free " /* param 7 */ "rps " /* param 8 */ "nbpi " /* param 9 */ "opt " /* param 10 */ "apc " /* param 11 */ "gap " /* param 12 */ "nrpos " /* param 13 */ "maxcontig " /* param 14 */ "mtb]\n"); /* param 15 */ (void) fprintf(stderr, gettext(" -m : dump fs cmd line used to make this partition\n" " -V :print this command line and return\n" " -o :ufs options: :nsect=%d,ntrack=%d,bsize=%d,fragsize=%d\n" " -o :ufs options: :cgsize=%d,free=%d,rps=%d,nbpi=%d,opt=%c\n" " -o :ufs options: :apc=%d,gap=%d,nrpos=%d,maxcontig=%d\n" " -o :ufs options: :mtb=%c,calcsb,calcbinsb\n" "NOTE that all -o suboptions: must be separated only by commas so as to\n" "be parsed as a single argument\n"), nsect, ntrack, bsize, fragsize, cpg, sblock.fs_minfree, rps, nbpi, opt, apc, (rotdelay == -1) ? 0 : rotdelay, sblock.fs_nrpos, maxcontig, mtb); lockexit(32); } /*ARGSUSED*/ static void dump_fscmd(char *fsys, int fsi) { int64_t used, bpcg, inospercg; int64_t nbpi; uint64_t nbytes64; bzero((char *)&sblock, sizeof (sblock)); rdfs((diskaddr_t)SBLOCK, SBSIZE, (char *)&sblock); /* * ensure a valid file system and if not, exit with error or else * we will end up computing block numbers etc and dividing by zero * which will cause floating point errors in this routine. */ if ((sblock.fs_magic != FS_MAGIC) && (sblock.fs_magic != MTB_UFS_MAGIC)) { (void) fprintf(stderr, gettext( "[not currently a valid file system - bad superblock]\n")); lockexit(32); } if (sblock.fs_magic == MTB_UFS_MAGIC && (sblock.fs_version > MTB_UFS_VERSION_1 || sblock.fs_version < MTB_UFS_VERSION_MIN)) { (void) fprintf(stderr, gettext( "Unknown version of UFS format: %d\n"), sblock.fs_version); lockexit(32); } /* * Compute a reasonable nbpi value. * The algorithm for "used" is copied from code * in main() verbatim. * The nbpi equation is taken from main where the * fs_ipg value is set for the last time. The INOPB(...) - 1 * is used to account for the roundup. * The problem is that a range of nbpi values map to * the same file system layout. So it is not possible * to calculate the exact value specified when the file * system was created. So instead we determine the top * end of the range of values. */ bpcg = sblock.fs_spc * sectorsize; inospercg = (int64_t)roundup(bpcg / sizeof (struct dinode), INOPB(&sblock)); if (inospercg > MAXIpG(&sblock)) inospercg = MAXIpG(&sblock); used = (int64_t) (sblock.fs_iblkno + inospercg / INOPF(&sblock)) * NSPF(&sblock); used *= sectorsize; nbytes64 = (uint64_t)sblock.fs_cpg * bpcg - used; /* * The top end of the range of values for nbpi may not be * a valid command line value for mkfs. Report the bottom * end instead. */ nbpi = (int64_t)(nbytes64 / (sblock.fs_ipg)); (void) fprintf(stdout, gettext("mkfs -F ufs -o "), fsys); (void) fprintf(stdout, "nsect=%d,ntrack=%d,", sblock.fs_nsect, sblock.fs_ntrak); (void) fprintf(stdout, "bsize=%d,fragsize=%d,cgsize=%d,free=%d,", sblock.fs_bsize, sblock.fs_fsize, sblock.fs_cpg, sblock.fs_minfree); (void) fprintf(stdout, "rps=%d,nbpi=%lld,opt=%c,apc=%d,gap=%d,", sblock.fs_rps, nbpi, (sblock.fs_optim == FS_OPTSPACE) ? 's' : 't', (sblock.fs_ntrak * sblock.fs_nsect) - sblock.fs_spc, sblock.fs_rotdelay); (void) fprintf(stdout, "nrpos=%d,maxcontig=%d,mtb=%c ", sblock.fs_nrpos, sblock.fs_maxcontig, ((sblock.fs_magic == MTB_UFS_MAGIC) ? 'y' : 'n')); (void) fprintf(stdout, "%s %lld\n", fsys, fsbtodb(&sblock, sblock.fs_size)); bzero((char *)&sblock, sizeof (sblock)); } /* number ************************************************************* */ /* */ /* Convert a numeric string arg to binary */ /* */ /* Args: d_value - default value, if have parse error */ /* param - the name of the argument, for error messages */ /* flags - parser state and what's allowed in the arg */ /* Global arg: string - pointer to command arg */ /* */ /* Valid forms: 123 | 123k | 123*123 | 123x123 */ /* */ /* Return: converted number */ /* */ /* ******************************************************************** */ static uint64_t number(uint64_t d_value, char *param, int flags) { char *cs; uint64_t n, t; uint64_t cut = BIG / 10; /* limit to avoid overflow */ int minus = 0; cs = string; if (*cs == '-') { minus = 1; cs += 1; } if ((*cs < '0') || (*cs > '9')) { goto bail_out; } n = 0; while ((*cs >= '0') && (*cs <= '9') && (n <= cut)) { n = n*10 + *cs++ - '0'; } if (minus) n = -n; for (;;) { switch (*cs++) { case 'k': if (flags & ALLOW_END_ONLY) goto bail_out; if (n > (BIG / 1024)) goto overflow; n *= 1024; continue; case '*': case 'x': if (flags & ALLOW_END_ONLY) goto bail_out; string = cs; t = number(d_value, param, flags); if (n > (BIG / t)) goto overflow; n *= t; cs = string + 1; /* adjust for -- below */ /* recursion has read rest of expression */ /* FALLTHROUGH */ case ',': case '\0': cs--; string = cs; return (n); case '%': if (flags & ALLOW_END_ONLY) goto bail_out; if (flags & ALLOW_PERCENT) { flags &= ~ALLOW_PERCENT; flags |= ALLOW_END_ONLY; continue; } goto bail_out; case 'm': if (flags & ALLOW_END_ONLY) goto bail_out; if (flags & ALLOW_MS1) { flags &= ~ALLOW_MS1; flags |= ALLOW_MS2; continue; } goto bail_out; case 's': if (flags & ALLOW_END_ONLY) goto bail_out; if (flags & ALLOW_MS2) { flags &= ~ALLOW_MS2; flags |= ALLOW_END_ONLY; continue; } goto bail_out; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': overflow: (void) fprintf(stderr, gettext("mkfs: value for %s overflowed\n"), param); while ((*cs != '\0') && (*cs != ',')) cs++; string = cs; return (BIG); default: bail_out: (void) fprintf(stderr, gettext( "mkfs: bad numeric arg for %s: \"%s\"\n"), param, string); while ((*cs != '\0') && (*cs != ',')) cs++; string = cs; if (d_value != NO_DEFAULT) { (void) fprintf(stderr, gettext("mkfs: %s reset to default %lld\n"), param, d_value); return (d_value); } lockexit(2); } } /* never gets here */ } /* match ************************************************************** */ /* */ /* Compare two text strings for equality */ /* */ /* Arg: s - pointer to string to match with a command arg */ /* Global arg: string - pointer to command arg */ /* */ /* Return: 1 if match, 0 if no match */ /* If match, also reset `string' to point to the text */ /* that follows the matching text. */ /* */ /* ******************************************************************** */ static int match(char *s) { char *cs; cs = string; while (*cs++ == *s) { if (*s++ == '\0') { goto true; } } if (*s != '\0') { return (0); } true: cs--; string = cs; return (1); } /* * GROWFS ROUTINES */ /* ARGSUSED */ void lockexit(int exitstatus) { if (Pflag) { /* the probe mode neither changes nor locks the filesystem */ exit(exitstatus); } /* * flush the dirty cylinder group */ if (inlockexit == 0) { inlockexit = 1; flcg(); } if (aio_inited) { flush_writes(); } /* * make sure the file system is unlocked before exiting */ if ((inlockexit == 1) && (!isbad)) { inlockexit = 2; ulockfs(); /* * if logging was enabled, then re-enable it */ if (waslog) { if (rl_log_control(fsys, _FIOLOGENABLE) != RL_SUCCESS) { (void) fprintf(stderr, gettext( "failed to re-enable logging\n")); } } } else if (grow) { if (isbad) { (void) fprintf(stderr, gettext( "Filesystem is currently inconsistent. It " "must be repaired with fsck(1M)\nbefore being " "used. Use the following command to " "do this:\n\n\tfsck %s\n\n"), fsys); if (ismounted) { (void) fprintf(stderr, gettext( "You will be told that the filesystem " "is already mounted, and asked if you\n" "wish to continue. Answer `yes' to " "this question.\n\n")); } (void) fprintf(stderr, gettext( "One problem should be reported, that " "the summary information is bad.\n" "You will then be asked if it " "should be salvaged. Answer `yes' " "to\nthis question.\n\n")); } if (ismounted) { /* * In theory, there's no way to get here without * isbad also being set, but be robust in the * face of future code changes. */ (void) fprintf(stderr, gettext( "The filesystem is currently mounted " "read-only and write-locked. ")); if (isbad) { (void) fprintf(stderr, gettext( "After\nrunning fsck, unlock the " "filesystem and ")); } else { (void) fprintf(stderr, gettext( "Unlock the filesystem\nand ")); } (void) fprintf(stderr, gettext( "re-enable writing with\nthe following " "command:\n\n\tlockfs -u %s\n\n"), directory); } } exit(exitstatus); } void randomgeneration() { int i; struct dinode *dp; /* * always perform fsirand(1) function... newfs will notice that * the inodes have been randomized and will not call fsirand itself */ for (i = 0, dp = zino; i < sblock.fs_inopb; ++i, ++dp) IRANDOMIZE(&dp->di_ic); } /* * Check the size of the summary information. * Fields in sblock are not changed in this function. * * For an 8K filesystem block, the maximum number of cylinder groups is 16384. * MAXCSBUFS {32} * 8K {FS block size} * divided by (sizeof csum) {16} * * Note that MAXCSBUFS is not used in the kernel; as of Solaris 2.6 build 32, * this is the only place where it's referenced. */ void checksummarysize() { diskaddr_t dmax; diskaddr_t dmin; int64_t cg0frags; int64_t cg0blocks; int64_t maxncg; int64_t maxfrags; uint64_t fs_size; uint64_t maxfs_blocks; /* filesystem blocks for max filesystem size */ /* * compute the maximum summary info size */ dmin = cgdmin(&sblock, 0); dmax = cgbase(&sblock, 0) + sblock.fs_fpg; fs_size = (grow) ? grow_fs_size : sblock.fs_size; if (dmax > fs_size) dmax = fs_size; cg0frags = dmax - dmin; cg0blocks = cg0frags / sblock.fs_frag; cg0frags = cg0blocks * sblock.fs_frag; maxncg = (longlong_t)cg0blocks * (longlong_t)(sblock.fs_bsize / sizeof (struct csum)); maxfs_blocks = FS_MAX; if (maxncg > ((longlong_t)maxfs_blocks / (longlong_t)sblock.fs_fpg) + 1) maxncg = ((longlong_t)maxfs_blocks / (longlong_t)sblock.fs_fpg) + 1; maxfrags = maxncg * (longlong_t)sblock.fs_fpg; if (maxfrags > maxfs_blocks) maxfrags = maxfs_blocks; /* * remember for later processing in extendsummaryinfo() */ if (test) grow_sifrag = dmin + (cg0blocks * sblock.fs_frag); if (testfrags == 0) testfrags = cg0frags; if (testforce) if (testfrags > cg0frags) { (void) fprintf(stderr, gettext("Too many test frags (%lld); " "try %lld\n"), testfrags, cg0frags); lockexit(32); } /* * if summary info is too large (too many cg's) tell the user and exit */ if ((longlong_t)sblock.fs_size > maxfrags) { (void) fprintf(stderr, gettext( "Too many cylinder groups with %llu sectors;\n try " "increasing cgsize, or decreasing fssize to %llu\n"), fsbtodb(&sblock, (uint64_t)sblock.fs_size), fsbtodb(&sblock, (uint64_t)maxfrags)); lockexit(32); } } void checksblock() { /* * make sure this is a file system */ if ((sblock.fs_magic != FS_MAGIC) && (sblock.fs_magic != MTB_UFS_MAGIC)) { (void) fprintf(stderr, gettext("Bad superblock; magic number wrong\n")); lockexit(32); } if (sblock.fs_magic == MTB_UFS_MAGIC && sblock.fs_version > MTB_UFS_VERSION_1) { (void) fprintf(stderr, gettext("Unrecognized version of UFS\n")); lockexit(32); } if (sblock.fs_ncg < 1) { (void) fprintf(stderr, gettext("Bad superblock; ncg out of range\n")); lockexit(32); } if (sblock.fs_cpg < 1) { (void) fprintf(stderr, gettext("Bad superblock; cpg out of range\n")); lockexit(32); } if (sblock.fs_ncg * sblock.fs_cpg < sblock.fs_ncyl || (sblock.fs_ncg - 1) * sblock.fs_cpg >= sblock.fs_ncyl) { (void) fprintf(stderr, gettext("Bad superblock; ncyl out of range\n")); lockexit(32); } if (sblock.fs_sbsize <= 0 || sblock.fs_sbsize > sblock.fs_bsize) { (void) fprintf(stderr, gettext( "Bad superblock; superblock size out of range\n")); lockexit(32); } } /* * Roll the embedded log, if any, and set up the global variables * islog, islogok and isufslog. */ static void logsetup(char *devstr) { void *buf, *ud_buf; extent_block_t *ebp; ml_unit_t *ul; ml_odunit_t *ud; /* * Does the superblock indicate that we are supposed to have a log ? */ if (sblock.fs_logbno == 0) { /* * No log present, nothing to do. */ islogok = 0; islog = 0; isufslog = 0; return; } else { /* * There's a log in a yet unknown state, attempt to roll it. */ islog = 1; islogok = 0; isufslog = 0; /* * We failed to roll the log, bail out. */ if (rl_roll_log(devstr) != RL_SUCCESS) return; isufslog = 1; /* log is not okay; check the fs */ if ((FSOKAY != (sblock.fs_state + sblock.fs_time)) || (sblock.fs_clean != FSLOG)) return; /* get the log allocation block */ buf = (void *)malloc(DEV_BSIZE); if (buf == (void *) NULL) return; ud_buf = (void *)malloc(DEV_BSIZE); if (ud_buf == (void *) NULL) { free(buf); return; } rdfs((diskaddr_t)logbtodb(&sblock, sblock.fs_logbno), DEV_BSIZE, buf); ebp = (extent_block_t *)buf; /* log allocation block is not okay; check the fs */ if (ebp->type != LUFS_EXTENTS) { free(buf); free(ud_buf); return; } /* get the log state block(s) */ rdfs((diskaddr_t)logbtodb(&sblock, ebp->extents[0].pbno), DEV_BSIZE, ud_buf); ud = (ml_odunit_t *)ud_buf; ul = (ml_unit_t *)malloc(sizeof (*ul)); ul->un_ondisk = *ud; /* log state is okay */ if ((ul->un_chksum == ul->un_head_ident + ul->un_tail_ident) && (ul->un_version == LUFS_VERSION_LATEST) && (ul->un_badlog == 0)) islogok = 1; free(ud_buf); free(buf); free(ul); } } void growinit(char *devstr) { int i; char buf[DEV_BSIZE]; /* * Read and verify the superblock */ rdfs((diskaddr_t)(SBOFF / sectorsize), (int)sbsize, (char *)&sblock); checksblock(); if (sblock.fs_postblformat != FS_DYNAMICPOSTBLFMT) { (void) fprintf(stderr, gettext("old file system format; can't growfs\n")); lockexit(32); } /* * can't shrink a file system */ grow_fssize = fsbtodb(&sblock, (uint64_t)sblock.fs_size); if (fssize_db < grow_fssize) { (void) fprintf(stderr, gettext("%lld sectors < current size of %lld sectors\n"), fssize_db, grow_fssize); lockexit(32); } /* * can't grow a system to over a terabyte unless it was set up * as an MTB UFS file system. */ if (mtb == 'y' && sblock.fs_magic != MTB_UFS_MAGIC) { if (fssize_db >= SECTORS_PER_TERABYTE) { (void) fprintf(stderr, gettext( "File system was not set up with the multi-terabyte format.\n")); (void) fprintf(stderr, gettext( "Its size cannot be increased to a terabyte or more.\n")); } else { (void) fprintf(stderr, gettext( "Cannot convert file system to multi-terabyte format.\n")); } lockexit(32); } logsetup(devstr); /* * can't growfs when logging device has errors */ if ((islog && !islogok) || ((FSOKAY == (sblock.fs_state + sblock.fs_time)) && (sblock.fs_clean == FSLOG && !islog))) { (void) fprintf(stderr, gettext("logging device has errors; can't growfs\n")); lockexit(32); } /* * disable ufs logging for growing */ if (isufslog) { if (rl_log_control(devstr, _FIOLOGDISABLE) != RL_SUCCESS) { (void) fprintf(stderr, gettext( "failed to disable logging\n")); lockexit(32); } islog = 0; waslog = 1; } /* * if mounted write lock the file system to be grown */ if (ismounted) wlockfs(); /* * refresh dynamic superblock state - disabling logging will have * changed the amount of free space available in the file system */ rdfs((diskaddr_t)(SBOFF / sectorsize), sbsize, (char *)&sblock); /* * make sure device is big enough */ rdfs((diskaddr_t)fssize_db - 1, DEV_BSIZE, buf); wtfs((diskaddr_t)fssize_db - 1, DEV_BSIZE, buf); /* * read current summary information */ grow_fscs = read_summaryinfo(&sblock); /* * save some current size related fields from the superblock * These are used in extendsummaryinfo() */ grow_fs_size = sblock.fs_size; grow_fs_ncg = sblock.fs_ncg; grow_fs_csaddr = (diskaddr_t)sblock.fs_csaddr; grow_fs_cssize = sblock.fs_cssize; /* * save and reset the clean flag */ if (FSOKAY == (sblock.fs_state + sblock.fs_time)) grow_fs_clean = sblock.fs_clean; else grow_fs_clean = FSBAD; sblock.fs_clean = FSBAD; sblock.fs_state = FSOKAY - sblock.fs_time; isbad = 1; wtfs((diskaddr_t)(SBOFF / sectorsize), sbsize, (char *)&sblock); } void checkdev(char *rdev, char *bdev) { struct stat64 statarea; if (stat64(bdev, &statarea) < 0) { (void) fprintf(stderr, gettext("can't check mount point; ")); (void) fprintf(stderr, gettext("can't stat %s\n"), bdev); lockexit(32); } if ((statarea.st_mode & S_IFMT) != S_IFBLK) { (void) fprintf(stderr, gettext( "can't check mount point; %s is not a block device\n"), bdev); lockexit(32); } if (stat64(rdev, &statarea) < 0) { (void) fprintf(stderr, gettext("can't stat %s\n"), rdev); lockexit(32); } if ((statarea.st_mode & S_IFMT) != S_IFCHR) { (void) fprintf(stderr, gettext("%s is not a character device\n"), rdev); lockexit(32); } } void checkmount(struct mnttab *mntp, char *bdevname) { struct stat64 statdir; struct stat64 statdev; if (strcmp(bdevname, mntp->mnt_special) == 0) { if (stat64(mntp->mnt_mountp, &statdir) == -1) { (void) fprintf(stderr, gettext("can't stat %s\n"), mntp->mnt_mountp); lockexit(32); } if (stat64(mntp->mnt_special, &statdev) == -1) { (void) fprintf(stderr, gettext("can't stat %s\n"), mntp->mnt_special); lockexit(32); } if (statdir.st_dev != statdev.st_rdev) { (void) fprintf(stderr, gettext( "%s is not mounted on %s; mnttab(4) wrong\n"), mntp->mnt_special, mntp->mnt_mountp); lockexit(32); } ismounted = 1; if (directory) { if (strcmp(mntp->mnt_mountp, directory) != 0) { (void) fprintf(stderr, gettext("%s is mounted on %s, not %s\n"), bdevname, mntp->mnt_mountp, directory); lockexit(32); } } else { if (grow) (void) fprintf(stderr, gettext( "%s is mounted on %s; can't growfs\n"), bdevname, mntp->mnt_mountp); else (void) fprintf(stderr, gettext("%s is mounted, can't mkfs\n"), bdevname); lockexit(32); } } } struct dinode *dibuf = 0; diskaddr_t difrag = 0; struct dinode * gdinode(ino_t ino) { /* * read the block of inodes containing inode number ino */ if (dibuf == 0) dibuf = (struct dinode *)malloc((unsigned)sblock.fs_bsize); if (itod(&sblock, ino) != difrag) { difrag = itod(&sblock, ino); rdfs(fsbtodb(&sblock, (uint64_t)difrag), (int)sblock.fs_bsize, (char *)dibuf); } return (dibuf + (ino % INOPB(&sblock))); } /* * structure that manages the frags we need for extended summary info * These frags can be: * free * data block * alloc block */ struct csfrag { struct csfrag *next; /* next entry */ daddr32_t ofrag; /* old frag */ daddr32_t nfrag; /* new frag */ long cylno; /* cylno of nfrag */ long frags; /* number of frags */ long size; /* size in bytes */ ino_t ino; /* inode number */ long fixed; /* Boolean - Already fixed? */ }; struct csfrag *csfrag; /* state unknown */ struct csfrag *csfragino; /* frags belonging to an inode */ struct csfrag *csfragfree; /* frags that are free */ daddr32_t maxcsfrag = 0; /* maximum in range */ daddr32_t mincsfrag = 0x7fffffff; /* minimum in range */ int csfraginrange(daddr32_t frag) { return ((frag >= mincsfrag) && (frag <= maxcsfrag)); } struct csfrag * findcsfrag(daddr32_t frag, struct csfrag **cfap) { struct csfrag *cfp; if (!csfraginrange(frag)) return (NULL); for (cfp = *cfap; cfp; cfp = cfp->next) if (cfp->ofrag == frag) return (cfp); return (NULL); } void checkindirect(ino_t ino, daddr32_t *fragsp, daddr32_t frag, int level) { int i; int ne = sblock.fs_bsize / sizeof (daddr32_t); daddr32_t fsb[MAXBSIZE / sizeof (daddr32_t)]; if (frag == 0) return; rdfs(fsbtodb(&sblock, frag), (int)sblock.fs_bsize, (char *)fsb); checkdirect(ino, fragsp, fsb, sblock.fs_bsize / sizeof (daddr32_t)); if (level) for (i = 0; i < ne && *fragsp; ++i) checkindirect(ino, fragsp, fsb[i], level-1); } void addcsfrag(ino_t ino, daddr32_t frag, struct csfrag **cfap) { struct csfrag *cfp, *curr, *prev; /* * establish a range for faster checking in csfraginrange() */ if (frag > maxcsfrag) maxcsfrag = frag; if (frag < mincsfrag) mincsfrag = frag; /* * if this frag belongs to an inode and is not the start of a block * then see if it is part of a frag range for this inode */ if (ino && (frag % sblock.fs_frag)) for (cfp = *cfap; cfp; cfp = cfp->next) { if (ino != cfp->ino) continue; if (frag != cfp->ofrag + cfp->frags) continue; cfp->frags++; cfp->size += sblock.fs_fsize; return; } /* * allocate a csfrag entry and insert it in an increasing order into the * specified list */ cfp = (struct csfrag *)calloc(1, sizeof (struct csfrag)); cfp->ino = ino; cfp->ofrag = frag; cfp->frags = 1; cfp->size = sblock.fs_fsize; for (prev = NULL, curr = *cfap; curr != NULL; prev = curr, curr = curr->next) { if (frag < curr->ofrag) { cfp->next = curr; if (prev) prev->next = cfp; /* middle element */ else *cfap = cfp; /* first element */ break; } if (curr->next == NULL) { curr->next = cfp; /* last element */ break; } } if (*cfap == NULL) /* will happen only once */ *cfap = cfp; } void delcsfrag(daddr32_t frag, struct csfrag **cfap) { struct csfrag *cfp; struct csfrag **cfpp; /* * free up entry whose beginning frag matches */ for (cfpp = cfap; *cfpp; cfpp = &(*cfpp)->next) { if (frag == (*cfpp)->ofrag) { cfp = *cfpp; *cfpp = (*cfpp)->next; free((char *)cfp); return; } } } /* * See whether any of the direct blocks in the array pointed by "db" and of * length "ne" are within the range of frags needed to extend the cylinder * summary. If so, remove those frags from the "as-yet-unclassified" list * (csfrag) and add them to the "owned-by-inode" list (csfragino). * For each such frag found, decrement the frag count pointed to by fragsp. * "ino" is the inode that contains (either directly or indirectly) the frags * being checked. */ void checkdirect(ino_t ino, daddr32_t *fragsp, daddr32_t *db, int ne) { int i; int j; int found; diskaddr_t frag; /* * scan for allocation within the new summary info range */ for (i = 0; i < ne && *fragsp; ++i) { if ((frag = *db++) != 0) { found = 0; for (j = 0; j < sblock.fs_frag && *fragsp; ++j) { if (found || (found = csfraginrange(frag))) { addcsfrag(ino, frag, &csfragino); delcsfrag(frag, &csfrag); } ++frag; --(*fragsp); } } } } void findcsfragino() { int i; int j; daddr32_t frags; struct dinode *dp; /* * scan all old inodes looking for allocations in the new * summary info range. Move the affected frag from the * generic csfrag list onto the `owned-by-inode' list csfragino. */ for (i = UFSROOTINO; i < grow_fs_ncg*sblock.fs_ipg && csfrag; ++i) { dp = gdinode((ino_t)i); switch (dp->di_mode & IFMT) { case IFSHAD : case IFLNK : case IFDIR : case IFREG : break; default : continue; } frags = dbtofsb(&sblock, dp->di_blocks); checkdirect((ino_t)i, &frags, &dp->di_db[0], NDADDR+NIADDR); for (j = 0; j < NIADDR && frags; ++j) checkindirect((ino_t)i, &frags, dp->di_ib[j], j); } } void fixindirect(daddr32_t frag, int level) { int i; int ne = sblock.fs_bsize / sizeof (daddr32_t); daddr32_t fsb[MAXBSIZE / sizeof (daddr32_t)]; if (frag == 0) return; rdfs(fsbtodb(&sblock, (uint64_t)frag), (int)sblock.fs_bsize, (char *)fsb); fixdirect((caddr_t)fsb, frag, fsb, ne); if (level) for (i = 0; i < ne; ++i) fixindirect(fsb[i], level-1); } void fixdirect(caddr_t bp, daddr32_t frag, daddr32_t *db, int ne) { int i; struct csfrag *cfp; for (i = 0; i < ne; ++i, ++db) { if (*db == 0) continue; if ((cfp = findcsfrag(*db, &csfragino)) == NULL) continue; *db = cfp->nfrag; cfp->fixed = 1; wtfs(fsbtodb(&sblock, (uint64_t)frag), (int)sblock.fs_bsize, bp); } } void fixcsfragino() { int i; struct dinode *dp; struct csfrag *cfp; for (cfp = csfragino; cfp; cfp = cfp->next) { if (cfp->fixed) continue; dp = gdinode((ino_t)cfp->ino); fixdirect((caddr_t)dibuf, difrag, dp->di_db, NDADDR+NIADDR); for (i = 0; i < NIADDR; ++i) fixindirect(dp->di_ib[i], i); } } /* * Read the cylinders summary information specified by settings in the * passed 'fs' structure into a new allocated array of csum structures. * The caller is responsible for freeing the returned array. * Return a pointer to an array of csum structures. */ static struct csum * read_summaryinfo(struct fs *fsp) { struct csum *csp; int i; if ((csp = malloc((size_t)fsp->fs_cssize)) == NULL) { (void) fprintf(stderr, gettext("cannot create csum list," " not enough memory\n")); exit(32); } for (i = 0; i < fsp->fs_cssize; i += fsp->fs_bsize) { rdfs(fsbtodb(fsp, (uint64_t)(fsp->fs_csaddr + numfrags(fsp, i))), (int)(fsp->fs_cssize - i < fsp->fs_bsize ? fsp->fs_cssize - i : fsp->fs_bsize), ((caddr_t)csp) + i); } return (csp); } /* * Check the allocation of fragments that are to be made part of a csum block. * A fragment is allocated if it is either in the csfragfree list or, it is * in the csfragino list and has new frags associated with it. * Return the number of allocated fragments. */ int64_t checkfragallocated(daddr32_t frag) { struct csfrag *cfp; /* * Since the lists are sorted we can break the search if the asked * frag is smaller then the one in the list. */ for (cfp = csfragfree; cfp != NULL && frag >= cfp->ofrag; cfp = cfp->next) { if (frag == cfp->ofrag) return (1); } for (cfp = csfragino; cfp != NULL && frag >= cfp->ofrag; cfp = cfp->next) { if (frag == cfp->ofrag && cfp->nfrag != 0) return (cfp->frags); } return (0); } /* * Figure out how much the filesystem can be grown. The limiting factor is * the available free space needed to extend the cg summary info block. * The free space is determined in three steps: * - Try to extend the cg summary block to the required size. * - Find free blocks in last cg. * - Find free space in the last already allocated fragment of the summary info * block, and use it for additional csum structures. * Return the maximum size of the new filesystem or 0 if it can't be grown. * Please note that this function leaves the global list pointers csfrag, * csfragfree, and csfragino initialized, and the caller is responsible for * freeing the lists. */ diskaddr_t probe_summaryinfo() { /* fragments by which the csum block can be extended. */ int64_t growth_csum_frags = 0; /* fragments by which the filesystem can be extended. */ int64_t growth_fs_frags = 0; int64_t new_fs_cssize; /* size of csum blk in the new FS */ int64_t new_fs_ncg; /* number of cg in the new FS */ int64_t spare_csum; daddr32_t oldfrag_daddr; daddr32_t newfrag_daddr; daddr32_t daddr; int i; /* * read and verify the superblock */ rdfs((diskaddr_t)(SBOFF / sectorsize), (int)sbsize, (char *)&sblock); checksblock(); /* * check how much we can extend the cg summary info block */ /* * read current summary information */ fscs = read_summaryinfo(&sblock); /* * build list of frags needed for cg summary info block extension */ oldfrag_daddr = howmany(sblock.fs_cssize, sblock.fs_fsize) + sblock.fs_csaddr; new_fs_ncg = howmany(dbtofsb(&sblock, fssize_db), sblock.fs_fpg); new_fs_cssize = fragroundup(&sblock, new_fs_ncg * sizeof (struct csum)); newfrag_daddr = howmany(new_fs_cssize, sblock.fs_fsize) + sblock.fs_csaddr; /* * add all of the frags that are required to grow the cyl summary to the * csfrag list, which is the generic/unknown list, since at this point * we don't yet know the state of those frags. */ for (daddr = oldfrag_daddr; daddr < newfrag_daddr; daddr++) addcsfrag((ino_t)0, daddr, &csfrag); /* * filter free fragments and allocate them. Note that the free frags * must be allocated first otherwise they could be grabbed by * alloccsfragino() for data frags. */ findcsfragfree(); alloccsfragfree(); /* * filter fragments owned by inodes and allocate them */ grow_fs_ncg = sblock.fs_ncg; /* findcsfragino() needs this glob. var. */ findcsfragino(); alloccsfragino(); if (notenoughspace()) { /* * check how many consecutive fragments could be allocated * in both lists. */ int64_t tmp_frags; for (daddr = oldfrag_daddr; daddr < newfrag_daddr; daddr += tmp_frags) { if ((tmp_frags = checkfragallocated(daddr)) > 0) growth_csum_frags += tmp_frags; else break; } } else { /* * We have all we need for the new desired size, * so clean up and report back. */ return (fssize_db); } /* * given the number of fragments by which the csum block can be grown * compute by how many new fragments the FS can be increased. * It is the number of csum instances per fragment multiplied by * `growth_csum_frags' and the number of fragments per cylinder group. */ growth_fs_frags = howmany(sblock.fs_fsize, sizeof (struct csum)) * growth_csum_frags * sblock.fs_fpg; /* * compute free fragments in the last cylinder group */ rdcg(sblock.fs_ncg - 1); growth_fs_frags += sblock.fs_fpg - acg.cg_ndblk; /* * compute how many csum instances are unused in the old csum block. * For each unused csum instance the FS can be grown by one cylinder * group without extending the csum block. */ spare_csum = howmany(sblock.fs_cssize, sizeof (struct csum)) - sblock.fs_ncg; if (spare_csum > 0) growth_fs_frags += spare_csum * sblock.fs_fpg; /* * recalculate the new filesystem size in sectors, shorten it by * the requested size `fssize_db' if necessary. */ if (growth_fs_frags > 0) { diskaddr_t sect; sect = (sblock.fs_size + growth_fs_frags) * sblock.fs_nspf; return ((sect > fssize_db) ? fssize_db : sect); } return (0); } void extendsummaryinfo() { int64_t i; int localtest = test; int64_t frags; daddr32_t oldfrag; daddr32_t newfrag; /* * if no-write (-N), don't bother */ if (Nflag) return; again: flcg(); /* * summary info did not change size -- do nothing unless in test mode */ if (grow_fs_cssize == sblock.fs_cssize) if (!localtest) return; /* * build list of frags needed for additional summary information */ oldfrag = howmany(grow_fs_cssize, sblock.fs_fsize) + grow_fs_csaddr; newfrag = howmany(sblock.fs_cssize, sblock.fs_fsize) + grow_fs_csaddr; /* * add all of the frags that are required to grow the cyl summary to the * csfrag list, which is the generic/unknown list, since at this point * we don't yet know the state of those frags. */ for (i = oldfrag, frags = 0; i < newfrag; ++i, ++frags) addcsfrag((ino_t)0, (diskaddr_t)i, &csfrag); /* * reduce the number of data blocks in the file system (fs_dsize) by * the number of frags that need to be added to the cyl summary */ sblock.fs_dsize -= (newfrag - oldfrag); /* * In test mode, we move more data than necessary from * cylinder group 0. The lookup/allocate/move code can be * better stressed without having to create HUGE file systems. */ if (localtest) for (i = newfrag; i < grow_sifrag; ++i) { if (frags >= testfrags) break; frags++; addcsfrag((ino_t)0, (diskaddr_t)i, &csfrag); } /* * move frags to free or inode lists, depending on owner */ findcsfragfree(); findcsfragino(); /* * if not all frags can be located, file system must be inconsistent */ if (csfrag) { isbad = 1; /* should already be set, but make sure */ lockexit(32); } /* * allocate the free frags. Note that the free frags must be allocated * first otherwise they could be grabbed by alloccsfragino() for data * frags. */ alloccsfragfree(); /* * allocate extra space for inode frags */ alloccsfragino(); /* * not enough space */ if (notenoughspace()) { unalloccsfragfree(); unalloccsfragino(); if (localtest && !testforce) { localtest = 0; goto again; } (void) fprintf(stderr, gettext("Not enough free space\n")); lockexit(NOTENOUGHSPACE); } /* * copy the data from old frags to new frags */ copycsfragino(); /* * fix the inodes to point to the new frags */ fixcsfragino(); /* * We may have moved more frags than we needed. Free them. */ rdcg((long)0); for (i = newfrag; i <= maxcsfrag; ++i) setbit(cg_blksfree(&acg), i-cgbase(&sblock, 0)); wtcg(); flcg(); } /* * Check if all fragments in the `csfragino' list were reallocated. */ int notenoughspace() { struct csfrag *cfp; /* * If any element in the csfragino array has a "new frag location" * of 0, the allocfrags() function was unsuccessful in allocating * space for moving the frag represented by this array element. */ for (cfp = csfragino; cfp; cfp = cfp->next) if (cfp->nfrag == 0) return (1); return (0); } void unalloccsfragino() { struct csfrag *cfp; while ((cfp = csfragino) != NULL) { if (cfp->nfrag) freefrags(cfp->nfrag, cfp->frags, cfp->cylno); delcsfrag(cfp->ofrag, &csfragino); } } void unalloccsfragfree() { struct csfrag *cfp; while ((cfp = csfragfree) != NULL) { freefrags(cfp->ofrag, cfp->frags, cfp->cylno); delcsfrag(cfp->ofrag, &csfragfree); } } /* * For each frag in the "as-yet-unclassified" list (csfrag), see if * it's free (i.e., its bit is set in the free frag bit map). If so, * move it from the "as-yet-unclassified" list to the csfragfree list. */ void findcsfragfree() { struct csfrag *cfp; struct csfrag *cfpnext; /* * move free frags onto the free-frag list */ rdcg((long)0); for (cfp = csfrag; cfp; cfp = cfpnext) { cfpnext = cfp->next; if (isset(cg_blksfree(&acg), cfp->ofrag - cgbase(&sblock, 0))) { addcsfrag(cfp->ino, cfp->ofrag, &csfragfree); delcsfrag(cfp->ofrag, &csfrag); } } } void copycsfragino() { struct csfrag *cfp; char buf[MAXBSIZE]; /* * copy data from old frags to newly allocated frags */ for (cfp = csfragino; cfp; cfp = cfp->next) { rdfs(fsbtodb(&sblock, (uint64_t)cfp->ofrag), (int)cfp->size, buf); wtfs(fsbtodb(&sblock, (uint64_t)cfp->nfrag), (int)cfp->size, buf); } } long curcylno = -1; int cylnodirty = 0; void rdcg(long cylno) { if (cylno != curcylno) { flcg(); curcylno = cylno; rdfs(fsbtodb(&sblock, (uint64_t)cgtod(&sblock, curcylno)), (int)sblock.fs_cgsize, (char *)&acg); } } void flcg() { if (cylnodirty) { if (debug && Pflag) { (void) fprintf(stderr, "Assert: cylnodirty set in probe mode\n"); return; } resetallocinfo(); wtfs(fsbtodb(&sblock, (uint64_t)cgtod(&sblock, curcylno)), (int)sblock.fs_cgsize, (char *)&acg); cylnodirty = 0; } curcylno = -1; } void wtcg() { if (!Pflag) { /* probe mode should never write to disk */ cylnodirty = 1; } } void allocfrags(long frags, daddr32_t *fragp, long *cylnop) { int i; int j; long bits; long bit; /* * Allocate a free-frag range in an old cylinder group */ for (i = 0, *fragp = 0; i < grow_fs_ncg; ++i) { if (((fscs+i)->cs_nffree < frags) && ((fscs+i)->cs_nbfree == 0)) continue; rdcg((long)i); bit = bits = 0; while (findfreerange(&bit, &bits)) { if (frags <= bits) { for (j = 0; j < frags; ++j) clrbit(cg_blksfree(&acg), bit+j); wtcg(); *cylnop = i; *fragp = bit + cgbase(&sblock, i); return; } bit += bits; } } } /* * Allocate space for frags that need to be moved in order to free up space for * expanding the cylinder summary info. * For each frag that needs to be moved (each frag or range of frags in * the csfragino list), allocate a new location and store the frag number * of that new location in the nfrag field of the csfrag struct. * If a new frag can't be allocated for any element in the csfragino list, * set the new frag number for that element to 0 and return immediately. * The notenoughspace() function will detect this condition. */ void alloccsfragino() { struct csfrag *cfp; /* * allocate space for inode frag ranges */ for (cfp = csfragino; cfp; cfp = cfp->next) { allocfrags(cfp->frags, &cfp->nfrag, &cfp->cylno); if (cfp->nfrag == 0) break; } } void alloccsfragfree() { struct csfrag *cfp; /* * allocate the free frags needed for extended summary info */ rdcg((long)0); for (cfp = csfragfree; cfp; cfp = cfp->next) clrbit(cg_blksfree(&acg), cfp->ofrag - cgbase(&sblock, 0)); wtcg(); } void freefrags(daddr32_t frag, long frags, long cylno) { int i; /* * free frags */ rdcg(cylno); for (i = 0; i < frags; ++i) { setbit(cg_blksfree(&acg), (frag+i) - cgbase(&sblock, cylno)); } wtcg(); } int findfreerange(long *bitp, long *bitsp) { long bit; /* * find a range of free bits in a cylinder group bit map */ for (bit = *bitp, *bitsp = 0; bit < acg.cg_ndblk; ++bit) if (isset(cg_blksfree(&acg), bit)) break; if (bit >= acg.cg_ndblk) return (0); *bitp = bit; *bitsp = 1; for (++bit; bit < acg.cg_ndblk; ++bit, ++(*bitsp)) { if ((bit % sblock.fs_frag) == 0) break; if (isclr(cg_blksfree(&acg), bit)) break; } return (1); } void resetallocinfo() { long cno; long bit; long bits; /* * Compute the free blocks/frags info and update the appropriate * inmemory superblock, summary info, and cylinder group fields */ sblock.fs_cstotal.cs_nffree -= acg.cg_cs.cs_nffree; sblock.fs_cstotal.cs_nbfree -= acg.cg_cs.cs_nbfree; acg.cg_cs.cs_nffree = 0; acg.cg_cs.cs_nbfree = 0; bzero((caddr_t)acg.cg_frsum, sizeof (acg.cg_frsum)); bzero((caddr_t)cg_blktot(&acg), (int)(acg.cg_iusedoff-acg.cg_btotoff)); bit = bits = 0; while (findfreerange(&bit, &bits)) { if (bits == sblock.fs_frag) { acg.cg_cs.cs_nbfree++; cno = cbtocylno(&sblock, bit); cg_blktot(&acg)[cno]++; cg_blks(&sblock, &acg, cno)[cbtorpos(&sblock, bit)]++; } else { acg.cg_cs.cs_nffree += bits; acg.cg_frsum[bits]++; } bit += bits; } *(fscs + acg.cg_cgx) = acg.cg_cs; sblock.fs_cstotal.cs_nffree += acg.cg_cs.cs_nffree; sblock.fs_cstotal.cs_nbfree += acg.cg_cs.cs_nbfree; } void extendcg(long cylno) { int i; diskaddr_t dupper; diskaddr_t cbase; diskaddr_t dmax; /* * extend the cylinder group at the end of the old file system * if it was partially allocated becase of lack of space */ flcg(); rdcg(cylno); dupper = acg.cg_ndblk; if (cylno == sblock.fs_ncg - 1) acg.cg_ncyl = sblock.fs_ncyl - (sblock.fs_cpg * cylno); else acg.cg_ncyl = sblock.fs_cpg; cbase = cgbase(&sblock, cylno); dmax = cbase + sblock.fs_fpg; if (dmax > sblock.fs_size) dmax = sblock.fs_size; acg.cg_ndblk = dmax - cbase; for (i = dupper; i < acg.cg_ndblk; ++i) setbit(cg_blksfree(&acg), i); sblock.fs_dsize += (acg.cg_ndblk - dupper); wtcg(); flcg(); } struct lockfs lockfs; int lockfd; int islocked; int lockfskey; char lockfscomment[128]; void ulockfs() { /* * if the file system was locked, unlock it before exiting */ if (islocked == 0) return; /* * first, check if the lock held */ lockfs.lf_flags = LOCKFS_MOD; if (ioctl(lockfd, _FIOLFSS, &lockfs) == -1) { perror(directory); lockexit(32); } if (LOCKFS_IS_MOD(&lockfs)) { (void) fprintf(stderr, gettext("FILE SYSTEM CHANGED DURING GROWFS!\n")); (void) fprintf(stderr, gettext(" See lockfs(1), umount(1), and fsck(1)\n")); lockexit(32); } /* * unlock the file system */ lockfs.lf_lock = LOCKFS_ULOCK; lockfs.lf_flags = 0; lockfs.lf_key = lockfskey; clockfs(); if (ioctl(lockfd, _FIOLFS, &lockfs) == -1) { perror(directory); lockexit(32); } } void wlockfs() { /* * if no-write (-N), don't bother */ if (Nflag) return; /* * open the mountpoint, and write lock the file system */ if ((lockfd = open64(directory, O_RDONLY)) == -1) { perror(directory); lockexit(32); } /* * check if it is already locked */ if (ioctl(lockfd, _FIOLFSS, &lockfs) == -1) { perror(directory); lockexit(32); } if (lockfs.lf_lock != LOCKFS_WLOCK) { lockfs.lf_lock = LOCKFS_WLOCK; lockfs.lf_flags = 0; lockfs.lf_key = 0; clockfs(); if (ioctl(lockfd, _FIOLFS, &lockfs) == -1) { perror(directory); lockexit(32); } } islocked = 1; lockfskey = lockfs.lf_key; } void clockfs() { time_t t; char *ct; (void) time(&t); ct = ctime(&t); ct[strlen(ct)-1] = '\0'; (void) sprintf(lockfscomment, "%s -- mkfs pid %d", ct, getpid()); lockfs.lf_comlen = strlen(lockfscomment)+1; lockfs.lf_comment = lockfscomment; } /* * Write the csum records and the superblock */ void wtsb() { long i; /* * write summary information */ for (i = 0; i < sblock.fs_cssize; i += sblock.fs_bsize) wtfs(fsbtodb(&sblock, (uint64_t)(sblock.fs_csaddr + numfrags(&sblock, i))), (int)(sblock.fs_cssize - i < sblock.fs_bsize ? sblock.fs_cssize - i : sblock.fs_bsize), ((char *)fscs) + i); /* * write superblock */ sblock.fs_time = mkfstime; wtfs((diskaddr_t)(SBOFF / sectorsize), sbsize, (char *)&sblock); } /* * Verify that the optimization selection is reasonable, and advance * the global "string" appropriately. */ static char checkopt(char *optim) { char opt; int limit = strcspn(optim, ","); switch (limit) { case 0: /* missing indicator (have comma or nul) */ (void) fprintf(stderr, gettext( "mkfs: missing optimization flag reset to `t' (time)\n")); opt = 't'; break; case 1: /* single-character indicator */ opt = *optim; if ((opt != 's') && (opt != 't')) { (void) fprintf(stderr, gettext( "mkfs: bad optimization value `%c' reset to `t' (time)\n"), opt); opt = 't'; } break; default: /* multi-character indicator */ (void) fprintf(stderr, gettext( "mkfs: bad optimization value `%*.*s' reset to `t' (time)\n"), limit, limit, optim); opt = 't'; break; } string += limit; return (opt); } /* * Verify that the mtb selection is reasonable, and advance * the global "string" appropriately. */ static char checkmtb(char *mtbarg) { char mtbc; int limit = strcspn(mtbarg, ","); switch (limit) { case 0: /* missing indicator (have comma or nul) */ (void) fprintf(stderr, gettext( "mkfs: missing mtb flag reset to `n' (no mtb support)\n")); mtbc = 'n'; break; case 1: /* single-character indicator */ mtbc = tolower(*mtbarg); if ((mtbc != 'y') && (mtbc != 'n')) { (void) fprintf(stderr, gettext( "mkfs: bad mtb value `%c' reset to `n' (no mtb support)\n"), mtbc); mtbc = 'n'; } break; default: /* multi-character indicator */ (void) fprintf(stderr, gettext( "mkfs: bad mtb value `%*.*s' reset to `n' (no mtb support)\n"), limit, limit, mtbarg); opt = 'n'; break; } string += limit; return (mtbc); } /* * Verify that a value is in a range. If it is not, resets it to * its default value if one is supplied, exits otherwise. * * When testing, can compare user_supplied to RC_KEYWORD or RC_POSITIONAL. */ static void range_check(long *varp, char *name, long minimum, long maximum, long def_val, int user_supplied) { if ((*varp < minimum) || (*varp > maximum)) { if (user_supplied != RC_DEFAULT) { (void) fprintf(stderr, gettext( "mkfs: bad value for %s: %ld must be between %ld and %ld\n"), name, *varp, minimum, maximum); } if (def_val != NO_DEFAULT) { if (user_supplied) { (void) fprintf(stderr, gettext("mkfs: %s reset to default %ld\n"), name, def_val); } *varp = def_val; return; } lockexit(2); /*NOTREACHED*/ } } /* * Verify that a value is in a range. If it is not, resets it to * its default value if one is supplied, exits otherwise. * * When testing, can compare user_supplied to RC_KEYWORD or RC_POSITIONAL. */ static void range_check_64(uint64_t *varp, char *name, uint64_t minimum, uint64_t maximum, uint64_t def_val, int user_supplied) { if ((*varp < minimum) || (*varp > maximum)) { if (user_supplied != RC_DEFAULT) { (void) fprintf(stderr, gettext( "mkfs: bad value for %s: %lld must be between %lld and %lld\n"), name, *varp, minimum, maximum); } if (def_val != NO_DEFAULT) { if (user_supplied) { (void) fprintf(stderr, gettext("mkfs: %s reset to default %lld\n"), name, def_val); } *varp = def_val; return; } lockexit(2); /*NOTREACHED*/ } } /* * Blocks SIGINT from delivery. Returns the previous mask in the * buffer provided, so that mask may be later restored. */ static void block_sigint(sigset_t *old_mask) { sigset_t block_mask; if (sigemptyset(&block_mask) < 0) { fprintf(stderr, gettext("Could not clear signal mask\n")); lockexit(3); } if (sigaddset(&block_mask, SIGINT) < 0) { fprintf(stderr, gettext("Could not set signal mask\n")); lockexit(3); } if (sigprocmask(SIG_BLOCK, &block_mask, old_mask) < 0) { fprintf(stderr, gettext("Could not block SIGINT\n")); lockexit(3); } } /* * Restores the signal mask that was in force before a call * to block_sigint(). This may actually still have SIGINT blocked, * if we've been recursively invoked. */ static void unblock_sigint(sigset_t *old_mask) { if (sigprocmask(SIG_UNBLOCK, old_mask, (sigset_t *)NULL) < 0) { fprintf(stderr, gettext("Could not restore signal mask\n")); lockexit(3); } } /* * Attempt to be somewhat graceful about being interrupted, rather than * just silently leaving the filesystem in an unusable state. * * The kernel has blocked SIGINT upon entry, so we don't have to worry * about recursion if the user starts pounding on the keyboard. */ static void recover_from_sigint(int signum) { if (fso > -1) { if ((Nflag != 0) || confirm_abort()) { lockexit(4); } } } static int confirm_abort(void) { char line[80]; printf(gettext("\n\nAborting at this point will leave the filesystem " "in an inconsistent\nstate. If you do choose to stop, " "you will be given instructions on how to\nrecover " "the filesystem. Do you wish to cancel the filesystem " "grow\noperation (y/n)?")); if (getline(stdin, line, sizeof (line)) == EOF) line[0] = 'y'; printf("\n"); if (line[0] == 'y' || line[0] == 'Y') return (1); else { return (0); } } static int getline(FILE *fp, char *loc, int maxlen) { int n; char *p, *lastloc; p = loc; lastloc = &p[maxlen-1]; while ((n = getc(fp)) != '\n') { if (n == EOF) return (EOF); if (!isspace(n) && p < lastloc) *p++ = n; } *p = 0; return (p - loc); } /* * Calculate the maximum value of cylinders-per-group for a file * system with the characteristics: * * bsize - file system block size * fragsize - frag size * nbpi - number of bytes of disk space per inode * nrpos - number of rotational positions * spc - sectors per cylinder * * These five characteristic are not adjustable (by this function). * The only attribute of the file system which IS adjusted by this * function in order to maximize cylinders-per-group is the proportion * of the cylinder group overhead block used for the inode map. The * inode map cannot occupy more than one-third of the cylinder group * overhead block, but it's OK for it to occupy less than one-third * of the overhead block. * * The setting of nbpi determines one possible value for the maximum * size of a cylinder group. It does so because it determines the total * number of inodes in the file system (file system size is fixed, and * nbpi is fixed, so the total number of inodes is fixed too). The * cylinder group has to be small enough so that the number of inodes * in the cylinder group is less than or equal to the number of bits * in one-third (or whatever proportion is assumed) of a file system * block. The details of the calculation are: * * The macro MAXIpG_B(bsize, inode_divisor) determines the maximum * number of inodes that can be in a cylinder group, given the * proportion of the cylinder group overhead block used for the * inode bitmaps (an inode_divisor of 3 means that 1/3 of the * block is used for inode bitmaps; an inode_divisor of 12 means * that 1/12 of the block is used for inode bitmaps.) * * Once the number of inodes per cylinder group is known, the * maximum value of cylinders-per-group (determined by nbpi) * is calculated by the formula * * maxcpg_given_nbpi = (size of a cylinder group)/(size of a cylinder) * * = (inodes-per-cg * nbpi)/(spc * DEV_BSIZE) * * (Interestingly, the size of the file system never enters * into this calculation.) * * Another possible value for the maximum cylinder group size is determined * by frag_size and nrpos. The frags in the cylinder group must be * representable in the frag bitmaps in the cylinder overhead block and the * rotational positions for each cylinder must be represented in the * rotational position tables. The calculation of the maximum cpg * value, given the frag and nrpos vales, is: * * maxcpg_given_fragsize = * (available space in the overhead block) / (size of per-cylinder data) * * The available space in the overhead block = * bsize - sizeof (struct cg) - space_used_for_inode_bitmaps * * The size of the per-cylinder data is: * sizeof(long) # for the "blocks avail per cylinder" field * + nrpos * sizeof(short) # for the rotational position table entry * + frags-per-cylinder/NBBY # number of bytes to represent this * # cylinder in the frag bitmap * * The two calculated maximum values of cylinder-per-group will typically * turn out to be different, since they are derived from two different * constraints. Usually, maxcpg_given_nbpi is much bigger than * maxcpg_given_fragsize. But they can be brought together by * adjusting the proportion of the overhead block dedicated to * the inode bitmaps. Decreasing the proportion of the cylinder * group overhead block used for inode maps will decrease * maxcpg_given_nbpi and increase maxcpg_given_fragsize. * * This function calculates the initial values of maxcpg_given_nbpi * and maxcpg_given_fragsize assuming that 1/3 of the cg overhead * block is used for inode bitmaps. Then it decreases the proportion * of the cg overhead block used for inode bitmaps (by increasing * the value of inode_divisor) until maxcpg_given_nbpi and * maxcpg_given_fragsize are the same, or stop changing, or * maxcpg_given_nbpi is less than maxcpg_given_fragsize. * * The loop terminates when any of the following occur: * * maxcpg_given_fragsize is greater than or equal to * maxcpg_given_nbpi * * neither maxcpg_given_fragsize nor maxcpg_given_nbpi * change in the expected direction * * The loop is guaranteed to terminate because it only continues * while maxcpg_given_fragsize and maxcpg_given_nbpi are approaching * each other. As soon they cross each other, or neither one changes * in the direction of the other, or one of them moves in the wrong * direction, the loop completes. */ static long compute_maxcpg(long bsize, long fragsize, long nbpi, long nrpos, long spc) { int maxcpg_given_nbpi; /* in cylinders */ int maxcpg_given_fragsize; /* in cylinders */ int spf; /* sectors per frag */ int inode_divisor; int old_max_given_frag = 0; int old_max_given_nbpi = INT_MAX; spf = fragsize / DEV_BSIZE; inode_divisor = 3; while (1) { maxcpg_given_nbpi = (((int64_t)(MAXIpG_B(bsize, inode_divisor))) * nbpi) / (DEV_BSIZE * ((int64_t)spc)); maxcpg_given_fragsize = (bsize - (sizeof (struct cg)) - (bsize / inode_divisor)) / (sizeof (long) + nrpos * sizeof (short) + (spc / spf) / NBBY); if (maxcpg_given_fragsize >= maxcpg_given_nbpi) return (maxcpg_given_nbpi); /* * If neither value moves toward the other, return the * least of the old values (we use the old instead of the * new because: if the old is the same as the new, it * doesn't matter which ones we use. If one of the * values changed, but in the wrong direction, the * new values are suspect. Better use the old. This * shouldn't happen, but it's best to check. */ if (!(maxcpg_given_nbpi < old_max_given_nbpi) && !(maxcpg_given_fragsize > old_max_given_frag)) return (MIN(old_max_given_nbpi, old_max_given_frag)); /* * This is probably impossible, but if one of the maxcpg * values moved in the "right" direction and one moved * in the "wrong" direction (that is, the two values moved * in the same direction), the previous conditional won't * recognize that the values aren't converging (since at * least one value moved in the "right" direction, the * last conditional says "keep going"). * * Just to make absolutely certain that the loop terminates, * check for one of the values moving in the "wrong" direction * and terminate the loop if it happens. */ if (maxcpg_given_nbpi > old_max_given_nbpi || maxcpg_given_fragsize < old_max_given_frag) return (MIN(old_max_given_nbpi, old_max_given_frag)); old_max_given_nbpi = maxcpg_given_nbpi; old_max_given_frag = maxcpg_given_fragsize; inode_divisor++; } } static int in_64bit_mode(void) { /* cmd must be an absolute path, for security */ char *cmd = "/usr/bin/isainfo -b"; char buf[BUFSIZ]; FILE *ptr; int retval = 0; putenv("IFS= \t"); if ((ptr = popen(cmd, "r")) != NULL) { if (fgets(buf, BUFSIZ, ptr) != NULL && strncmp(buf, "64", 2) == 0) retval = 1; (void) pclose(ptr); } return (retval); } /* * validate_size * * Return 1 if the device appears to be at least "size" sectors long. * Return 0 if it's shorter or we can't read it. */ static int validate_size(int fd, diskaddr_t size) { char buf[DEV_BSIZE]; int rc; if ((llseek(fd, (offset_t)((size - 1) * DEV_BSIZE), SEEK_SET) == -1) || (read(fd, buf, DEV_BSIZE)) != DEV_BSIZE) rc = 0; else rc = 1; return (rc); } /* * Print every field of the calculated superblock, along with * its value. To make parsing easier on the caller, the value * is printed first, then the name. Additionally, there's only * one name/value pair per line. All values are reported in * hexadecimal (with the traditional 0x prefix), as that's slightly * easier for humans to read. Not that they're expected to, but * debugging happens. */ static void dump_sblock(void) { int row, column, pending, written; caddr_t source; if (Rflag) { pending = sizeof (sblock); source = (caddr_t)&sblock; do { written = write(fileno(stdout), source, pending); pending -= written; source += written; } while ((pending > 0) && (written > 0)); if (written < 0) { perror(gettext("Binary dump of superblock failed")); lockexit(1); } return; } else { printf("0x%x sblock.fs_link\n", sblock.fs_link); printf("0x%x sblock.fs_rolled\n", sblock.fs_rolled); printf("0x%x sblock.fs_sblkno\n", sblock.fs_sblkno); printf("0x%x sblock.fs_cblkno\n", sblock.fs_cblkno); printf("0x%x sblock.fs_iblkno\n", sblock.fs_iblkno); printf("0x%x sblock.fs_dblkno\n", sblock.fs_dblkno); printf("0x%x sblock.fs_cgoffset\n", sblock.fs_cgoffset); printf("0x%x sblock.fs_cgmask\n", sblock.fs_cgmask); printf("0x%x sblock.fs_time\n", sblock.fs_time); printf("0x%x sblock.fs_size\n", sblock.fs_size); printf("0x%x sblock.fs_dsize\n", sblock.fs_dsize); printf("0x%x sblock.fs_ncg\n", sblock.fs_ncg); printf("0x%x sblock.fs_bsize\n", sblock.fs_bsize); printf("0x%x sblock.fs_fsize\n", sblock.fs_fsize); printf("0x%x sblock.fs_frag\n", sblock.fs_frag); printf("0x%x sblock.fs_minfree\n", sblock.fs_minfree); printf("0x%x sblock.fs_rotdelay\n", sblock.fs_rotdelay); printf("0x%x sblock.fs_rps\n", sblock.fs_rps); printf("0x%x sblock.fs_bmask\n", sblock.fs_bmask); printf("0x%x sblock.fs_fmask\n", sblock.fs_fmask); printf("0x%x sblock.fs_bshift\n", sblock.fs_bshift); printf("0x%x sblock.fs_fshift\n", sblock.fs_fshift); printf("0x%x sblock.fs_maxcontig\n", sblock.fs_maxcontig); printf("0x%x sblock.fs_maxbpg\n", sblock.fs_maxbpg); printf("0x%x sblock.fs_fragshift\n", sblock.fs_fragshift); printf("0x%x sblock.fs_fsbtodb\n", sblock.fs_fsbtodb); printf("0x%x sblock.fs_sbsize\n", sblock.fs_sbsize); printf("0x%x sblock.fs_csmask\n", sblock.fs_csmask); printf("0x%x sblock.fs_csshift\n", sblock.fs_csshift); printf("0x%x sblock.fs_nindir\n", sblock.fs_nindir); printf("0x%x sblock.fs_inopb\n", sblock.fs_inopb); printf("0x%x sblock.fs_nspf\n", sblock.fs_nspf); printf("0x%x sblock.fs_optim\n", sblock.fs_optim); #ifdef _LITTLE_ENDIAN printf("0x%x sblock.fs_state\n", sblock.fs_state); #else printf("0x%x sblock.fs_npsect\n", sblock.fs_npsect); #endif printf("0x%x sblock.fs_si\n", sblock.fs_si); printf("0x%x sblock.fs_trackskew\n", sblock.fs_trackskew); printf("0x%x sblock.fs_id[0]\n", sblock.fs_id[0]); printf("0x%x sblock.fs_id[1]\n", sblock.fs_id[1]); printf("0x%x sblock.fs_csaddr\n", sblock.fs_csaddr); printf("0x%x sblock.fs_cssize\n", sblock.fs_cssize); printf("0x%x sblock.fs_cgsize\n", sblock.fs_cgsize); printf("0x%x sblock.fs_ntrak\n", sblock.fs_ntrak); printf("0x%x sblock.fs_nsect\n", sblock.fs_nsect); printf("0x%x sblock.fs_spc\n", sblock.fs_spc); printf("0x%x sblock.fs_ncyl\n", sblock.fs_ncyl); printf("0x%x sblock.fs_cpg\n", sblock.fs_cpg); printf("0x%x sblock.fs_ipg\n", sblock.fs_ipg); printf("0x%x sblock.fs_fpg\n", sblock.fs_fpg); printf("0x%x sblock.fs_cstotal\n", sblock.fs_cstotal); printf("0x%x sblock.fs_fmod\n", sblock.fs_fmod); printf("0x%x sblock.fs_clean\n", sblock.fs_clean); printf("0x%x sblock.fs_ronly\n", sblock.fs_ronly); printf("0x%x sblock.fs_flags\n", sblock.fs_flags); printf("0x%x sblock.fs_fsmnt\n", sblock.fs_fsmnt); printf("0x%x sblock.fs_cgrotor\n", sblock.fs_cgrotor); printf("0x%x sblock.fs_u.fs_csp\n", sblock.fs_u.fs_csp); printf("0x%x sblock.fs_cpc\n", sblock.fs_cpc); /* * No macros are defined for the dimensions of the * opostbl array. */ for (row = 0; row < 16; row++) { for (column = 0; column < 8; column++) { printf("0x%x sblock.fs_opostbl[%d][%d]\n", sblock.fs_opostbl[row][column], row, column); } } /* * Ditto the size of sparecon. */ for (row = 0; row < 51; row++) { printf("0x%x sblock.fs_sparecon[%d]\n", sblock.fs_sparecon[row], row); } printf("0x%x sblock.fs_version\n", sblock.fs_version); printf("0x%x sblock.fs_logbno\n", sblock.fs_logbno); printf("0x%x sblock.fs_reclaim\n", sblock.fs_reclaim); printf("0x%x sblock.fs_sparecon2\n", sblock.fs_sparecon2); #ifdef _LITTLE_ENDIAN printf("0x%x sblock.fs_npsect\n", sblock.fs_npsect); #else printf("0x%x sblock.fs_state\n", sblock.fs_state); #endif printf("0x%llx sblock.fs_qbmask\n", sblock.fs_qbmask); printf("0x%llx sblock.fs_qfmask\n", sblock.fs_qfmask); printf("0x%x sblock.fs_postblformat\n", sblock.fs_postblformat); printf("0x%x sblock.fs_nrpos\n", sblock.fs_nrpos); printf("0x%x sblock.fs_postbloff\n", sblock.fs_postbloff); printf("0x%x sblock.fs_rotbloff\n", sblock.fs_rotbloff); printf("0x%x sblock.fs_magic\n", sblock.fs_magic); /* * fs_space isn't of much use in this context, so we'll * just ignore it for now. */ } }