/* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (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 */ #pragma ident "%Z%%M% %I% %E% SMI" /* from UCB 5.2 9/11/85 */ /* * newfs: friendly front end to mkfs * * Copyright 2008 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static unsigned int number(char *, char *, int, int); static int64_t number64(char *, char *, int, int64_t); static diskaddr_t getdiskbydev(char *); static int yes(void); static int notrand(char *); static void usage(); static diskaddr_t get_device_size(int, char *); static diskaddr_t brute_force_get_device_size(int); static int validate_size(char *disk, diskaddr_t size); static void exenv(void); static struct fs *read_sb(char *); /*PRINTFLIKE1*/ static void fatal(char *fmt, ...); #define EPATH "PATH=/usr/sbin:/sbin:" #define CPATH "/sbin" /* an EPATH element */ #define MB (1024 * 1024) #define GBSEC ((1024 * 1024 * 1024) / DEV_BSIZE) /* sectors in a GB */ #define MINFREESEC ((64 * 1024 * 1024) / DEV_BSIZE) /* sectors in 64 MB */ #define MINCPG (16) /* traditional */ #define MAXDEFDENSITY (8 * 1024) /* arbitrary */ #define MINDENSITY (2 * 1024) /* traditional */ #define MIN_MTB_DENSITY (1024 * 1024) #define POWEROF2(num) (((num) & ((num) - 1)) == 0) #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 /* For use with number() */ #define NR_NONE 0 #define NR_PERCENT 0x01 /* * 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 #ifdef DEBUG #define dprintf(x) printf x #else #define dprintf(x) #endif static int Nflag; /* run mkfs without writing file system */ static int Tflag; /* set up file system for growth to over 1 TB */ static int verbose; /* show mkfs line before exec */ static int fsize = 0; /* fragment size */ static int fsize_flag = 0; /* fragment size was specified on cmd line */ static int bsize; /* block size */ static int ntracks; /* # tracks/cylinder */ static int ntracks_set = 0; /* true if the user specified ntracks */ static int optim = FS_OPTTIME; /* optimization, t(ime) or s(pace) */ static int nsectors; /* # sectors/track */ static int cpg; /* cylinders/cylinder group */ static int cpg_set = 0; /* true if the user specified cpg */ static int minfree = -1; /* free space threshold */ static int rpm; /* revolutions/minute of drive */ static int rpm_set = 0; /* true if the user specified rpm */ static int nrpos = 8; /* # of distinguished rotational positions */ /* 8 is the historical default */ static int nrpos_set = 0; /* true if the user specified nrpos */ static int density = 0; /* number of bytes per inode */ static int apc; /* alternates per cylinder */ static int apc_set = 0; /* true if the user specified apc */ static int rot = -1; /* rotational delay (msecs) */ static int rot_set = 0; /* true if the user specified rot */ static int maxcontig = -1; /* maximum number of contig blocks */ static int text_sb = 0; /* no disk changes; just final sb text dump */ static int binary_sb = 0; /* no disk changes; just final sb binary dump */ static int label_type; /* see types below */ /* * The variable use_efi_dflts is an indicator of whether to use EFI logic * or the geometry logic in laying out the filesystem. This is decided * based on the size/type of the disk and is used only for non-EFI labeled * disks and removable media. */ static int use_efi_dflts = 0; static int isremovable = 0; static char device[MAXPATHLEN]; static char cmd[BUFSIZ]; extern char *getfullrawname(); /* from libadm */ int main(int argc, char *argv[]) { char *special, *name; struct stat64 st; int status; int option; struct fs *sbp; /* Pointer to superblock (if present) */ diskaddr_t actual_fssize; diskaddr_t max_possible_fssize; diskaddr_t req_fssize = 0; diskaddr_t fssize = 0; char *req_fssize_str = NULL; /* requested size argument */ (void) setlocale(LC_ALL, ""); #if !defined(TEXT_DOMAIN) #define TEXT_DOMAIN "SYS_TEST" #endif (void) textdomain(TEXT_DOMAIN); opterr = 0; /* We print our own errors, disable getopt's message */ while ((option = getopt(argc, argv, "vNBSs:C:d:t:o:a:b:f:c:m:n:r:i:T")) != EOF) { switch (option) { case 'S': text_sb++; break; case 'B': binary_sb++; break; case 'v': verbose++; break; case 'N': Nflag++; break; case 's': /* * The maximum file system size is a lot smaller * than FS_SIZE_UPPER_LIMIT, but until we find out * the device size and block size, we don't know * what it is. So save the requested size in a * string so that we can print it out later if we * determine it's too big. */ req_fssize = number64("fssize", optarg, NR_NONE, FS_SIZE_UPPER_LIMIT); if (req_fssize < 1024) fatal(gettext( "%s: fssize must be at least 1024"), optarg); req_fssize_str = strdup(optarg); if (req_fssize_str == NULL) fatal(gettext( "Insufficient memory for string copy.")); break; case 'C': maxcontig = number("maxcontig", optarg, NR_NONE, -1); if (maxcontig < 0) fatal(gettext("%s: bad maxcontig"), optarg); break; case 'd': rot = number("rotdelay", optarg, NR_NONE, 0); rot_set = 1; if (rot < 0 || rot > 1000) fatal(gettext( "%s: bad rotational delay"), optarg); break; case 't': ntracks = number("ntrack", optarg, NR_NONE, 16); ntracks_set = 1; if ((ntracks < 0) || (ntracks > INT_MAX)) fatal(gettext("%s: bad total tracks"), optarg); break; case 'o': if (strcmp(optarg, "space") == 0) optim = FS_OPTSPACE; else if (strcmp(optarg, "time") == 0) optim = FS_OPTTIME; else fatal(gettext( "%s: bad optimization preference (options are `space' or `time')"), optarg); break; case 'a': apc = number("apc", optarg, NR_NONE, 0); apc_set = 1; if (apc < 0 || apc > 32768) /* see mkfs.c */ fatal(gettext( "%s: bad alternates per cyl"), optarg); break; case 'b': bsize = number("bsize", optarg, NR_NONE, DESBLKSIZE); if (bsize < MINBSIZE || bsize > MAXBSIZE) fatal(gettext( "%s: bad block size"), optarg); break; case 'f': fsize = number("fragsize", optarg, NR_NONE, DESFRAGSIZE); fsize_flag++; /* xxx ought to test against bsize for upper limit */ if (fsize < DEV_BSIZE) fatal(gettext("%s: bad frag size"), optarg); break; case 'c': cpg = number("cpg", optarg, NR_NONE, 16); cpg_set = 1; if (cpg < 1) fatal(gettext("%s: bad cylinders/group"), optarg); break; case 'm': minfree = number("minfree", optarg, NR_PERCENT, 10); if (minfree < 0 || minfree > 99) fatal(gettext("%s: bad free space %%"), optarg); break; case 'n': nrpos = number("nrpos", optarg, NR_NONE, 8); nrpos_set = 1; if (nrpos <= 0) fatal(gettext( "%s: bad number of rotational positions"), optarg); break; case 'r': rpm = number("rpm", optarg, NR_NONE, 3600); rpm_set = 1; if (rpm < 0) fatal(gettext("%s: bad revs/minute"), optarg); break; case 'i': /* xxx ought to test against fsize */ density = number("nbpi", optarg, NR_NONE, 2048); if (density < DEV_BSIZE) fatal(gettext("%s: bad bytes per inode"), optarg); break; case 'T': Tflag++; break; default: usage(); fatal(gettext("-%c: unknown flag"), optopt); } } /* At this point, there should only be one argument left: */ /* The raw-special-device itself. If not, print usage message. */ if ((argc - optind) != 1) { usage(); exit(1); } name = argv[optind]; special = getfullrawname(name); if (special == NULL) { (void) fprintf(stderr, gettext("newfs: malloc failed\n")); exit(1); } if (*special == '\0') { if (strchr(name, '/') != NULL) { if (stat64(name, &st) < 0) { (void) fprintf(stderr, gettext("newfs: %s: %s\n"), name, strerror(errno)); exit(2); } fatal(gettext("%s: not a raw disk device"), name); } (void) snprintf(device, sizeof (device), "/dev/rdsk/%s", name); if ((special = getfullrawname(device)) == NULL) { (void) fprintf(stderr, gettext("newfs: malloc failed\n")); exit(1); } if (*special == '\0') { (void) snprintf(device, sizeof (device), "/dev/%s", name); if ((special = getfullrawname(device)) == NULL) { (void) fprintf(stderr, gettext("newfs: malloc failed\n")); exit(1); } if (*special == '\0') fatal(gettext( "%s: not a raw disk device"), name); } } /* * getdiskbydev() determines the characteristics of the special * device on which the file system will be built. In the case * of devices with SMI labels (that is, non-EFI labels), the * following characteristics are set (if they were not already * set on the command line, since the command line settings * take precedence): * * nsectors - sectors per track * ntracks - tracks per cylinder * rpm - disk revolutions per minute * * apc is NOT set * * getdiskbydev() also sets the following quantities for all * devices, if not already set: * * bsize - file system block size * maxcontig * label_type (efi, vtoc, or other) * * getdiskbydev() returns the actual size of the device, in * sectors. */ actual_fssize = getdiskbydev(special); if (req_fssize == 0) { fssize = actual_fssize; } else { /* * If the user specified a size larger than what we've * determined as the actual size of the device, see if the * size specified by the user can be read. If so, use it, * since some devices and volume managers may not support * the vtoc and EFI interfaces we use to determine device * size. */ if (req_fssize > actual_fssize && validate_size(special, req_fssize)) { (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"), req_fssize, actual_fssize); fssize = req_fssize; } else { fssize = MIN(req_fssize, actual_fssize); } } if (label_type == LABEL_TYPE_VTOC) { if (nsectors < 0) fatal(gettext("%s: no default #sectors/track"), special); if (!use_efi_dflts) { if (ntracks < 0) fatal(gettext("%s: no default #tracks"), special); } if (rpm < 0) fatal(gettext( "%s: no default revolutions/minute value"), special); if (rpm < 60) { (void) fprintf(stderr, gettext("Warning: setting rpm to 60\n")); rpm = 60; } } if (label_type == LABEL_TYPE_EFI || label_type == LABEL_TYPE_OTHER) { if (ntracks_set) (void) fprintf(stderr, gettext( "Warning: ntracks is obsolete for this device and will be ignored.\n")); if (cpg_set) (void) fprintf(stderr, gettext( "Warning: cylinders/group is obsolete for this device and will be ignored.\n")); if (rpm_set) (void) fprintf(stderr, gettext( "Warning: rpm is obsolete for this device and will be ignored.\n")); if (rot_set) (void) fprintf(stderr, gettext( "Warning: rotational delay is obsolete for this device and" " will be ignored.\n")); if (nrpos_set) (void) fprintf(stderr, gettext( "Warning: number of rotational positions is obsolete for this device and\n" "will be ignored.\n")); if (apc_set) (void) fprintf(stderr, gettext( "Warning: number of alternate sectors per cylinder is obsolete for this\n" "device and will be ignored.\n")); /* * We need these for the call to mkfs, even though they are * meaningless. */ rpm = 60; nrpos = 1; apc = 0; rot = -1; /* * These values are set to produce a file system with * a cylinder group size of 48MB. For disks with * non-EFI labels, most geometries result in cylinder * groups of around 40 - 50 MB, so we arbitrarily choose * 48MB for disks with EFI labels. mkfs will reduce * cylinders per group even further if necessary. */ cpg = 16; nsectors = 128; ntracks = 48; /* * mkfs produces peculiar results for file systems * that are smaller than one cylinder so don't allow * them to be created (this check is only made for * disks with EFI labels. Eventually, it should probably * be enforced for all disks.) */ if (fssize < nsectors * ntracks) { fatal(gettext( "file system size must be at least %d sectors"), nsectors * ntracks); } } if (fssize > INT_MAX) Tflag = 1; /* * If the user requested that the file system be set up for * eventual growth to over a terabyte, or if it's already greater * than a terabyte, set the inode density (nbpi) to MIN_MTB_DENSITY * (unless the user has specified a larger nbpi), set the frag size * equal to the block size, and set the cylinders-per-group value * passed to mkfs to -1, which tells mkfs to make cylinder groups * as large as possible. */ if (Tflag) { if (density < MIN_MTB_DENSITY) density = MIN_MTB_DENSITY; fsize = bsize; cpg = -1; /* says make cyl groups as big as possible */ } else { if (fsize == 0) fsize = DESFRAGSIZE; } if (!POWEROF2(fsize)) { (void) fprintf(stderr, gettext( "newfs: fragment size must a power of 2, not %d\n"), fsize); fsize = bsize/8; (void) fprintf(stderr, gettext( "newfs: fragsize reset to %ld\n"), fsize); } /* * The file system is limited in size by the fragment size. * The number of fragments in the file system must fit into * a signed 32-bit quantity, so the number of sectors in the * file system is INT_MAX * the number of sectors in a frag. */ max_possible_fssize = ((uint64_t)fsize)/DEV_BSIZE * INT_MAX; if (fssize > max_possible_fssize) fssize = max_possible_fssize; /* * Now fssize is the final size of the file system (in sectors). * If it's less than what the user requested, print a message. */ if (fssize < req_fssize) { (void) fprintf(stderr, gettext( "newfs: requested size of %s disk blocks is too large.\n"), req_fssize_str); (void) fprintf(stderr, gettext( "newfs: Resetting size to %lld\n"), fssize); } /* * fssize now equals the size (in sectors) of the file system * that will be created. */ /* XXX - following defaults are both here and in mkfs */ if (density <= 0) { if (fssize < GBSEC) density = MINDENSITY; else density = (int)((((longlong_t)fssize + (GBSEC - 1)) / GBSEC) * MINDENSITY); if (density <= 0) density = MINDENSITY; if (density > MAXDEFDENSITY) density = MAXDEFDENSITY; } if (cpg == 0) { /* * maxcpg calculation adapted from mkfs * In the case of disks with EFI labels, cpg has * already been set, so we won't enter this code. */ long maxcpg, maxipg; maxipg = roundup(bsize * NBBY / 3, bsize / sizeof (struct inode)); maxcpg = (bsize - sizeof (struct cg) - howmany(maxipg, NBBY)) / (sizeof (long) + nrpos * sizeof (short) + nsectors / (MAXFRAG * NBBY)); cpg = (fssize / GBSEC) * 32; if (cpg > maxcpg) cpg = maxcpg; if (cpg <= 0) cpg = MINCPG; } if (minfree < 0) { minfree = (int)(((float)MINFREESEC / fssize) * 100); if (minfree > 10) minfree = 10; if (minfree <= 0) minfree = 1; } #ifdef i386 /* Bug 1170182 */ if (ntracks > 32 && (ntracks % 16) != 0) { ntracks -= (ntracks % 16); } #endif /* * Confirmation */ if (isatty(fileno(stdin)) && !Nflag) { /* * If we can read a valid superblock, report the mount * point on which this filesystem was last mounted. */ if (((sbp = read_sb(special)) != 0) && (*sbp->fs_fsmnt != '\0')) { (void) printf(gettext( "newfs: %s last mounted as %s\n"), special, sbp->fs_fsmnt); } (void) printf(gettext( "newfs: construct a new file system %s: (y/n)? "), special); (void) fflush(stdout); if (!yes()) exit(0); } dprintf(("DeBuG newfs : nsect=%d ntrak=%d cpg=%d\n", nsectors, ntracks, cpg)); /* * If alternates-per-cylinder is ever implemented: * need to get apc from dp->d_apc if no -a switch??? */ (void) snprintf(cmd, sizeof (cmd), "pfexec mkfs -F ufs " "%s%s%s%s %lld %d %d %d %d %d %d %d %d %s %d %d %d %d %s", Nflag ? "-o N " : "", binary_sb ? "-o calcbinsb " : "", text_sb ? "-o calcsb " : "", special, fssize, nsectors, ntracks, bsize, fsize, cpg, minfree, rpm/60, density, optim == FS_OPTSPACE ? "s" : "t", apc, rot, nrpos, maxcontig, Tflag ? "y" : "n"); if (verbose) { (void) printf("%s\n", cmd); (void) fflush(stdout); } exenv(); if (status = system(cmd)) exit(status >> 8); if (Nflag) exit(0); (void) snprintf(cmd, sizeof (cmd), "/usr/sbin/fsirand %s", special); if (notrand(special) && (status = system(cmd)) != 0) (void) fprintf(stderr, gettext("%s: failed, status = %d\n"), cmd, status); return (0); } static void exenv(void) { char *epath; /* executable file path */ char *cpath; /* current path */ if ((cpath = getenv("PATH")) == NULL) { (void) fprintf(stderr, gettext("newfs: no PATH in env\n")); /* * Background: the Bourne shell interpolates "." into * the path where said path starts with a colon, ends * with a colon, or has two adjacent colons. Thus, * the path ":/sbin::/usr/sbin:" is equivalent to * ".:/sbin:.:/usr/sbin:.". Now, we have no cpath, * and epath ends in a colon (to make for easy * catenation in the normal case). By the above, if * we use "", then "." becomes part of path. That's * bad, so use CPATH (which is just a duplicate of some * element in EPATH). No point in opening ourselves * up to a Trojan horse attack when we don't have to.... */ cpath = CPATH; } if ((epath = malloc(strlen(EPATH) + strlen(cpath) + 1)) == NULL) { (void) fprintf(stderr, gettext("newfs: malloc failed\n")); exit(1); } (void) strcpy(epath, EPATH); (void) strcat(epath, cpath); if (putenv(epath) < 0) { (void) fprintf(stderr, gettext("newfs: putenv failed\n")); exit(1); } } static int yes(void) { int i, b; i = b = getchar(); while (b != '\n' && b != '\0' && b != EOF) b = getchar(); return (i == 'y'); } /* * xxx Caller must run fmt through gettext(3) for us, if we ever * xxx go the i18n route.... */ static void fatal(char *fmt, ...) { va_list pvar; (void) fprintf(stderr, "newfs: "); va_start(pvar, fmt); (void) vfprintf(stderr, fmt, pvar); va_end(pvar); (void) putc('\n', stderr); exit(10); } static diskaddr_t getdiskbydev(char *disk) { struct dk_geom g; struct dk_cinfo ci; diskaddr_t actual_size; int fd; if ((fd = open64(disk, 0)) < 0) { perror(disk); exit(1); } /* * get_device_size() determines the actual size of the * device, and also the disk's attributes, such as geometry. */ actual_size = get_device_size(fd, disk); if (label_type == LABEL_TYPE_VTOC) { /* * Geometry information does not make sense for removable media * anyway, so indicate mkfs to use EFI default parameters. */ if (ioctl(fd, DKIOCREMOVABLE, &isremovable)) { dprintf(("DeBuG newfs : Unable to determine if %s is" " Removable Media. Proceeding with system" " determined parameters.\n", disk)); isremovable = 0; } else if (isremovable && !Tflag) use_efi_dflts = 1; if (ioctl(fd, DKIOCGGEOM, &g)) fatal(gettext( "%s: Unable to read Disk geometry"), disk); if (((g.dkg_ncyl * g.dkg_nhead * g.dkg_nsect) > CHSLIMIT) && !Tflag) { use_efi_dflts = 1; } dprintf(("DeBuG newfs : geom=%ld, CHSLIMIT=%d " "isremovable = %d use_efi_dflts = %d\n", g.dkg_ncyl * g.dkg_nhead * g.dkg_nsect, CHSLIMIT, isremovable, use_efi_dflts)); /* * The ntracks that is passed to mkfs is decided here based * on 'use_efi_dflts' and whether ntracks was specified as a * command line parameter to newfs. * If ntracks of -1 is passed to mkfs, mkfs uses DEF_TRACKS_EFI * and DEF_SECTORS_EFI for ntracks and nsectors respectively. */ if (nsectors == 0) nsectors = g.dkg_nsect; if (ntracks == 0) ntracks = use_efi_dflts ? -1 : g.dkg_nhead; if (rpm == 0) rpm = ((int)g.dkg_rpm <= 0) ? 3600: g.dkg_rpm; } if (bsize == 0) bsize = DESBLKSIZE; /* * Adjust maxcontig by the device's maxtransfer. If maxtransfer * information is not available, default to the min of a MB and * maxphys. */ if (maxcontig == -1 && ioctl(fd, DKIOCINFO, &ci) == 0) { maxcontig = ci.dki_maxtransfer * DEV_BSIZE; if (maxcontig < 0) { int error, gotit, maxphys; gotit = fsgetmaxphys(&maxphys, &error); /* * If we cannot get the maxphys value, default * to ufs_maxmaxphys (MB). */ if (gotit) { maxcontig = MIN(maxphys, MB); } else { (void) fprintf(stderr, gettext( "Warning: Could not get system value for maxphys. The value for maxcontig\n" "will default to 1MB.\n")); maxcontig = MB; } } maxcontig /= bsize; } (void) close(fd); return (actual_size); } /* * Figure out how big the partition we're dealing with is. */ static diskaddr_t get_device_size(int fd, char *name) { struct vtoc vtoc; dk_gpt_t *efi_vtoc; diskaddr_t slicesize; int index = read_vtoc(fd, &vtoc); if (index >= 0) { label_type = LABEL_TYPE_VTOC; } else { if (index == VT_ENOTSUP || index == VT_ERROR) { /* it might be an EFI label */ index = efi_alloc_and_read(fd, &efi_vtoc); if (index >= 0) label_type = LABEL_TYPE_EFI; } } if (index < 0) { /* * Since both attempts to read the label failed, we're * going to fall back to a brute force approach to * determining the device's size: see how far out we can * perform reads on the device. */ slicesize = brute_force_get_device_size(fd); if (slicesize == 0) { switch (index) { case VT_ERROR: (void) fprintf(stderr, gettext( "newfs: %s: %s\n"), name, strerror(errno)); exit(10); /*NOTREACHED*/ case VT_EIO: fatal(gettext( "%s: I/O error accessing VTOC"), name); /*NOTREACHED*/ case VT_EINVAL: fatal(gettext( "%s: Invalid field in VTOC"), name); /*NOTREACHED*/ default: fatal(gettext( "%s: unknown error accessing VTOC"), name); /*NOTREACHED*/ } } else { label_type = LABEL_TYPE_OTHER; } } if (label_type == LABEL_TYPE_EFI) { slicesize = efi_vtoc->efi_parts[index].p_size; efi_free(efi_vtoc); } else if (label_type == LABEL_TYPE_VTOC) { /* * 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; } return (slicesize); } /* * brute_force_get_device_size * * Determine the size of the device by seeing how far we can * read. Doing an llseek( , , SEEK_END) would probably work * in most cases, but we've seen at least one third-party driver * which doesn't correctly support the SEEK_END option when the * the device is greater than a terabyte. */ static diskaddr_t brute_force_get_device_size(int fd) { diskaddr_t min_fail = 0; diskaddr_t max_succeed = 0; diskaddr_t cur_db_off; char buf[DEV_BSIZE]; /* * First, see if we can read the device at all, just to * eliminate errors that have nothing to do with the * device's size. */ if (((llseek(fd, (offset_t)0, SEEK_SET)) == -1) || ((read(fd, buf, DEV_BSIZE)) == -1)) return (0); /* can't determine size */ /* * Now, go sequentially through the multiples of 4TB * to find the first read that fails (this isn't strictly * the most efficient way to find the actual size if the * size really could be anything between 0 and 2**64 bytes. * We expect the sizes to be less than 16 TB for some time, * so why do a bunch of reads that are larger than that? * However, this algorithm *will* work for sizes of greater * than 16 TB. We're just not optimizing for those sizes.) */ for (cur_db_off = SECTORS_PER_TERABYTE * 4; min_fail == 0 && cur_db_off < FS_SIZE_UPPER_LIMIT; cur_db_off += 4 * SECTORS_PER_TERABYTE) { if (((llseek(fd, (offset_t)(cur_db_off * DEV_BSIZE), SEEK_SET)) == -1) || ((read(fd, buf, DEV_BSIZE)) != DEV_BSIZE)) min_fail = cur_db_off; else max_succeed = cur_db_off; } if (min_fail == 0) return (0); /* * We now know that the size of the device is less than * min_fail and greater than or equal to max_succeed. Now * keep splitting the difference until the actual size in * sectors in known. We also know that the difference * between max_succeed and min_fail at this time is * 4 * SECTORS_PER_TERABYTE, which is a power of two, which * simplifies the math below. */ while (min_fail - max_succeed > 1) { cur_db_off = max_succeed + (min_fail - max_succeed)/2; if (((llseek(fd, (offset_t)(cur_db_off * DEV_BSIZE), SEEK_SET)) == -1) || ((read(fd, buf, DEV_BSIZE)) != DEV_BSIZE)) min_fail = cur_db_off; else max_succeed = cur_db_off; } /* the size is the last successfully read sector offset plus one */ return (max_succeed + 1); } /* * 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(char *disk, diskaddr_t size) { char buf[DEV_BSIZE]; int fd, rc; if ((fd = open64(disk, O_RDONLY)) < 0) { perror(disk); exit(1); } if ((llseek(fd, (offset_t)((size - 1) * DEV_BSIZE), SEEK_SET) == -1) || (read(fd, buf, DEV_BSIZE)) != DEV_BSIZE) rc = 0; else rc = 1; (void) close(fd); return (rc); } /* * read_sb(char * rawdev) - Attempt to read the superblock from a raw device * * Returns: * 0 : * Could not read a valid superblock for a variety of reasons. * Since 'newfs' handles any fatal conditions, we're not going * to make any guesses as to why this is failing or what should * be done about it. * * struct fs *: * A pointer to (what we think is) a valid superblock. The * space for the superblock is static (inside the function) * since we will only be reading the values from it. */ struct fs * read_sb(char *fsdev) { static struct fs sblock; struct stat64 statb; int dskfd; char *bufp = NULL; int bufsz = 0; if (stat64(fsdev, &statb) < 0) return (0); if ((dskfd = open64(fsdev, O_RDONLY)) < 0) return (0); /* * We need a buffer whose size is a multiple of DEV_BSIZE in order * to read from a raw device (which we were probably passed). */ bufsz = ((sizeof (sblock) / DEV_BSIZE) + 1) * DEV_BSIZE; if ((bufp = malloc(bufsz)) == NULL) { (void) close(dskfd); return (0); } if (llseek(dskfd, (offset_t)SBOFF, SEEK_SET) < 0 || read(dskfd, bufp, bufsz) < 0) { (void) close(dskfd); free(bufp); return (0); } (void) close(dskfd); /* Done with the file */ (void) memcpy(&sblock, bufp, sizeof (sblock)); free(bufp); /* Don't need this anymore */ if (((sblock.fs_magic != FS_MAGIC) && (sblock.fs_magic != MTB_UFS_MAGIC)) || sblock.fs_ncg < 1 || sblock.fs_cpg < 1) return (0); if (sblock.fs_ncg * sblock.fs_cpg < sblock.fs_ncyl || (sblock.fs_ncg - 1) * sblock.fs_cpg >= sblock.fs_ncyl) return (0); if (sblock.fs_sbsize < 0 || sblock.fs_sbsize > SBSIZE) return (0); return (&sblock); } /* * Read the UFS file system on the raw device SPECIAL. If it does not * appear to be a UFS file system, return non-zero, indicating that * fsirand should be called (and it will spit out an error message). * If it is a UFS file system, take a look at the inodes in the first * cylinder group. If they appear to be randomized (non-zero), return * zero, which will cause fsirand to not be called. If the inode generation * counts are all zero, then we must call fsirand, so return non-zero. */ #define RANDOMIZED 0 #define NOT_RANDOMIZED 1 static int notrand(char *special) { long fsbuf[SBSIZE / sizeof (long)]; struct dinode dibuf[MAXBSIZE/sizeof (struct dinode)]; struct fs *fs; struct dinode *dip; offset_t seekaddr; int bno, inum; int fd; fs = (struct fs *)fsbuf; if ((fd = open64(special, 0)) == -1) return (NOT_RANDOMIZED); if (llseek(fd, (offset_t)SBLOCK * DEV_BSIZE, 0) == -1 || read(fd, (char *)fs, SBSIZE) != SBSIZE || ((fs->fs_magic != FS_MAGIC) && (fs->fs_magic != MTB_UFS_MAGIC))) { (void) close(fd); return (NOT_RANDOMIZED); } /* looks like a UFS file system; read the first cylinder group */ bsize = INOPB(fs) * sizeof (struct dinode); inum = 0; while (inum < fs->fs_ipg) { bno = itod(fs, inum); seekaddr = (offset_t)fsbtodb(fs, bno) * DEV_BSIZE; if (llseek(fd, seekaddr, 0) == -1 || read(fd, (char *)dibuf, bsize) != bsize) { (void) close(fd); return (NOT_RANDOMIZED); } for (dip = dibuf; dip < &dibuf[INOPB(fs)]; dip++) { if (dip->di_gen != 0) { (void) close(fd); return (RANDOMIZED); } inum++; } } (void) close(fd); return (NOT_RANDOMIZED); } static void usage(void) { (void) fprintf(stderr, gettext( "usage: newfs [ -v ] [ mkfs-options ] raw-special-device\n")); (void) fprintf(stderr, gettext("where mkfs-options are:\n")); (void) fprintf(stderr, gettext( "\t-N do not create file system, just print out parameters\n")); (void) fprintf(stderr, gettext( "\t-T configure file system for eventual growth to over a terabyte\n")); (void) fprintf(stderr, gettext("\t-s file system size (sectors)\n")); (void) fprintf(stderr, gettext("\t-b block size\n")); (void) fprintf(stderr, gettext("\t-f frag size\n")); (void) fprintf(stderr, gettext("\t-t tracks/cylinder\n")); (void) fprintf(stderr, gettext("\t-c cylinders/group\n")); (void) fprintf(stderr, gettext("\t-m minimum free space %%\n")); (void) fprintf(stderr, gettext( "\t-o optimization preference (`space' or `time')\n")); (void) fprintf(stderr, gettext("\t-r revolutions/minute\n")); (void) fprintf(stderr, gettext("\t-i number of bytes per inode\n")); (void) fprintf(stderr, gettext( "\t-a number of alternates per cylinder\n")); (void) fprintf(stderr, gettext("\t-C maxcontig\n")); (void) fprintf(stderr, gettext("\t-d rotational delay\n")); (void) fprintf(stderr, gettext( "\t-n number of rotational positions\n")); (void) fprintf(stderr, gettext( "\t-S print a textual version of the calculated superblock to stdout\n")); (void) fprintf(stderr, gettext( "\t-B dump a binary version of the calculated superblock to stdout\n")); } /* * Error-detecting version of atoi(3). Adapted from mkfs' number(). */ static unsigned int number(char *param, char *value, int flags, int def_value) { char *cs; int n; int cut = INT_MAX / 10; /* limit to avoid overflow */ int minus = 0; cs = value; 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 '\0': return (n); case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': (void) fprintf(stderr, gettext( "newfs: value for %s overflowed, using %d\n"), param, def_value); return (def_value); case '%': if (flags & NR_PERCENT) break; /* FALLTHROUGH */ default: bail_out: fatal(gettext("bad numeric arg for %s: \"%s\""), param, value); } } /* NOTREACHED */ } /* * Error-detecting version of atoi(3). Adapted from mkfs' number(). */ static int64_t number64(char *param, char *value, int flags, int64_t def_value) { char *cs; int64_t n; int64_t cut = FS_SIZE_UPPER_LIMIT/ 10; /* limit to avoid overflow */ int minus = 0; cs = value; 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 '\0': return (n); case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': (void) fprintf(stderr, gettext( "newfs: value for %s overflowed, using %d\n"), param, def_value); return (def_value); case '%': if (flags & NR_PERCENT) break; /* FALLTHROUGH */ default: bail_out: fatal(gettext("bad numeric arg for %s: \"%s\""), param, value); } } /* NOTREACHED */ }