newfs_msdos: rework error handling for eventual use in makefs

Return -1 on errors from mkfs_msdos() instead of err()/errx(), to
allow different consumers to handle errors as appropriate.

Obtained from:	NetBSD
Sponsored by:	The FreeBSD Foundation
This commit is contained in:
Ed Maste 2015-11-23 18:58:00 +00:00
parent b3e7589994
commit 99f4158c82
2 changed files with 191 additions and 98 deletions

View File

@ -214,12 +214,11 @@ static const u_int8_t bootcode[] = {
static volatile sig_atomic_t got_siginfo;
static void infohandler(int);
static void check_mounted(const char *, mode_t);
static void getstdfmt(const char *, struct bpb *);
static void getdiskinfo(int, const char *, const char *, int,
struct bpb *);
static int check_mounted(const char *, mode_t);
static int getstdfmt(const char *, struct bpb *);
static int getdiskinfo(int, const char *, const char *, int, struct bpb *);
static void print_bpb(struct bpb *);
static u_int ckgeom(const char *, u_int, const char *);
static int ckgeom(const char *, u_int, const char *);
static void mklabel(u_int8_t *, const char *);
static int oklabel(const char *);
static void setstr(u_int8_t *, const char *, size_t);
@ -247,20 +246,31 @@ int mkfs_msdos(const char *fname, const char *dtype,
struct msdos_options o = *op;
if (o.OEM_string && strlen(o.OEM_string) > 8) {
errx(1, "%s: bad OEM string", o.OEM_string);
warnx("%s: bad OEM string", o.OEM_string);
return -1;
}
if (o.create_size) {
if (o.no_create)
errx(1, "create (-C) is incompatible with -N");
if (o.no_create) {
warnx("create (-C) is incompatible with -N");
return -1;
}
fd = open(fname, O_RDWR | O_CREAT | O_TRUNC, 0644);
if (fd == -1)
errx(1, "failed to create %s", fname);
if (ftruncate(fd, o.create_size))
errx(1, "failed to initialize %jd bytes", (intmax_t)o.create_size);
} else if ((fd = open(fname, o.no_create ? O_RDONLY : O_RDWR)) == -1)
err(1, "%s", fname);
if (fstat(fd, &sb))
err(1, "%s", fname);
if (fd == -1) {
warnx("failed to create %s", fname);
return -1;
}
if (ftruncate(fd, o.create_size)) {
warnx("failed to initialize %jd bytes", (intmax_t)o.create_size);
return -1;
}
} else if ((fd = open(fname, o.no_create ? O_RDONLY : O_RDWR)) == -1) {
warn("%s", fname);
return -1;
}
if (fstat(fd, &sb)) {
warn("%s", fname);
return -1;
}
if (o.create_size) {
if (!S_ISREG(sb.st_mode))
warnx("warning, %s is not a regular file", fname);
@ -269,12 +279,16 @@ int mkfs_msdos(const char *fname, const char *dtype,
warnx("warning, %s is not a character device", fname);
}
if (!o.no_create)
check_mounted(fname, sb.st_mode);
if (o.offset && o.offset != lseek(fd, o.offset, SEEK_SET))
errx(1, "cannot seek to %jd", (intmax_t)o.offset);
if (check_mounted(fname, sb.st_mode) == -1)
return -1;
if (o.offset && o.offset != lseek(fd, o.offset, SEEK_SET)) {
warnx("cannot seek to %jd", (intmax_t)o.offset);
return -1;
}
memset(&bpb, 0, sizeof(bpb));
if (o.floppy) {
getstdfmt(o.floppy, &bpb);
if (getstdfmt(o.floppy, &bpb) == -1)
return -1;
bpb.bpbHugeSectors = bpb.bpbSectors;
bpb.bpbSectors = 0;
bpb.bpbBigFATsecs = bpb.bpbFATsecs;
@ -313,58 +327,81 @@ int mkfs_msdos(const char *fname, const char *dtype,
bpb.bpbSecPerClust = 64; /* otherwise 32k */
}
}
if (!powerof2(bpb.bpbBytesPerSec))
errx(1, "bytes/sector (%u) is not a power of 2", bpb.bpbBytesPerSec);
if (bpb.bpbBytesPerSec < MINBPS)
errx(1, "bytes/sector (%u) is too small; minimum is %u",
if (!powerof2(bpb.bpbBytesPerSec)) {
warnx("bytes/sector (%u) is not a power of 2", bpb.bpbBytesPerSec);
return -1;
}
if (bpb.bpbBytesPerSec < MINBPS) {
warnx("bytes/sector (%u) is too small; minimum is %u",
bpb.bpbBytesPerSec, MINBPS);
if (o.volume_label && !oklabel(o.volume_label))
errx(1, "%s: bad volume label", o.volume_label);
return -1;
}
if (o.volume_label && !oklabel(o.volume_label)) {
warnx("%s: bad volume label", o.volume_label);
return -1;
}
if (!(fat = o.fat_type)) {
if (o.floppy)
fat = 12;
else if (!o.directory_entries && (o.info_sector || o.backup_sector))
fat = 32;
}
if ((fat == 32 && o.directory_entries) || (fat != 32 && (o.info_sector || o.backup_sector)))
errx(1, "-%c is not a legal FAT%s option",
if ((fat == 32 && o.directory_entries) || (fat != 32 && (o.info_sector || o.backup_sector))) {
warnx("-%c is not a legal FAT%s option",
fat == 32 ? 'e' : o.info_sector ? 'i' : 'k',
fat == 32 ? "32" : "12/16");
return -1;
}
if (o.floppy && fat == 32)
bpb.bpbRootDirEnts = 0;
if (fat != 0 && fat != 12 && fat != 16 && fat != 32) {
errx(1, "%d: bad FAT type", fat);
warnx("%d: bad FAT type", fat);
return -1;
}
if (o.block_size) {
if (!powerof2(o.block_size))
errx(1, "block size (%u) is not a power of 2", o.block_size);
if (o.block_size < bpb.bpbBytesPerSec)
errx(1, "block size (%u) is too small; minimum is %u",
if (!powerof2(o.block_size)) {
warnx("block size (%u) is not a power of 2", o.block_size);
return -1;
}
if (o.block_size < bpb.bpbBytesPerSec) {
warnx("block size (%u) is too small; minimum is %u",
o.block_size, bpb.bpbBytesPerSec);
if (o.block_size > bpb.bpbBytesPerSec * MAXSPC)
errx(1, "block size (%u) is too large; maximum is %u",
return -1;
}
if (o.block_size > bpb.bpbBytesPerSec * MAXSPC) {
warnx("block size (%u) is too large; maximum is %u",
o.block_size, bpb.bpbBytesPerSec * MAXSPC);
return -1;
}
bpb.bpbSecPerClust = o.block_size / bpb.bpbBytesPerSec;
}
if (o.sectors_per_cluster) {
if (!powerof2(o.sectors_per_cluster))
errx(1, "sectors/cluster (%u) is not a power of 2", o.sectors_per_cluster);
if (!powerof2(o.sectors_per_cluster)) {
warnx("sectors/cluster (%u) is not a power of 2",
o.sectors_per_cluster);
return -1;
}
bpb.bpbSecPerClust = o.sectors_per_cluster;
}
if (o.reserved_sectors)
bpb.bpbResSectors = o.reserved_sectors;
if (o.num_FAT) {
if (o.num_FAT > MAXNFT)
errx(1, "number of FATs (%u) is too large; maximum is %u",
if (o.num_FAT > MAXNFT) {
warnx("number of FATs (%u) is too large; maximum is %u",
o.num_FAT, MAXNFT);
return -1;
}
bpb.bpbFATs = o.num_FAT;
}
if (o.directory_entries)
bpb.bpbRootDirEnts = o.directory_entries;
if (o.media_descriptor_set) {
if (o.media_descriptor < 0xf0)
errx(1, "illegal media descriptor (%#x)", o.media_descriptor);
if (o.media_descriptor < 0xf0) {
warnx("illegal media descriptor (%#x)", o.media_descriptor);
return -1;
}
bpb.bpbMedia = o.media_descriptor;
}
if (o.sectors_per_fat)
@ -380,15 +417,21 @@ int mkfs_msdos(const char *fname, const char *dtype,
bname = o.bootstrap;
if (!strchr(bname, '/')) {
snprintf(buf, sizeof(buf), "/boot/%s", bname);
if (!(bname = strdup(buf)))
err(1, NULL);
if (!(bname = strdup(buf))) {
warn(NULL);
return -1;
}
}
if ((fd1 = open(bname, O_RDONLY)) == -1 || fstat(fd1, &sb)) {
warn("%s", bname);
return -1;
}
if ((fd1 = open(bname, O_RDONLY)) == -1 || fstat(fd1, &sb))
err(1, "%s", bname);
if (!S_ISREG(sb.st_mode) || sb.st_size % bpb.bpbBytesPerSec ||
sb.st_size < bpb.bpbBytesPerSec ||
sb.st_size > bpb.bpbBytesPerSec * MAXU16)
errx(1, "%s: inappropriate file type or format", bname);
sb.st_size > bpb.bpbBytesPerSec * MAXU16) {
warnx("%s: inappropriate file type or format", bname);
return -1;
}
bss = sb.st_size / bpb.bpbBytesPerSec;
}
if (!bpb.bpbFATs)
@ -420,27 +463,35 @@ int mkfs_msdos(const char *fname, const char *dtype,
x = bss;
if (fat == 32) {
if (!bpb.bpbFSInfo) {
if (x == MAXU16 || x == bpb.bpbBackup)
errx(1, "no room for info sector");
if (x == MAXU16 || x == bpb.bpbBackup) {
warnx("no room for info sector");
return -1;
}
bpb.bpbFSInfo = x;
}
if (bpb.bpbFSInfo != MAXU16 && x <= bpb.bpbFSInfo)
x = bpb.bpbFSInfo + 1;
if (!bpb.bpbBackup) {
if (x == MAXU16)
errx(1, "no room for backup sector");
if (x == MAXU16) {
warnx("no room for backup sector");
return -1;
}
bpb.bpbBackup = x;
} else if (bpb.bpbBackup != MAXU16 && bpb.bpbBackup == bpb.bpbFSInfo)
errx(1, "backup sector would overwrite info sector");
} else if (bpb.bpbBackup != MAXU16 && bpb.bpbBackup == bpb.bpbFSInfo) {
warnx("backup sector would overwrite info sector");
return -1;
}
if (bpb.bpbBackup != MAXU16 && x <= bpb.bpbBackup)
x = bpb.bpbBackup + 1;
}
if (!bpb.bpbResSectors)
bpb.bpbResSectors = fat == 32 ?
MAX(x, MAX(16384 / bpb.bpbBytesPerSec, 4)) : x;
else if (bpb.bpbResSectors < x)
errx(1, "too few reserved sectors (need %d have %d)", x,
else if (bpb.bpbResSectors < x) {
warnx("too few reserved sectors (need %d have %d)", x,
bpb.bpbResSectors);
return -1;
}
if (fat != 32 && !bpb.bpbRootDirEnts)
bpb.bpbRootDirEnts = DEFRDE;
rds = howmany(bpb.bpbRootDirEnts, bpb.bpbBytesPerSec / sizeof(struct de));
@ -457,12 +508,16 @@ int mkfs_msdos(const char *fname, const char *dtype,
bpb.bpbSecPerClust <= bpb.bpbHugeSectors;
bpb.bpbSecPerClust <<= 1)
continue;
if (fat != 32 && bpb.bpbBigFATsecs > MAXU16)
errx(1, "too many sectors/FAT for FAT12/16");
if (fat != 32 && bpb.bpbBigFATsecs > MAXU16) {
warnx("too many sectors/FAT for FAT12/16");
return -1;
}
x1 = bpb.bpbResSectors + rds;
x = bpb.bpbBigFATsecs ? bpb.bpbBigFATsecs : 1;
if (x1 + (u_int64_t)x * bpb.bpbFATs > bpb.bpbHugeSectors)
errx(1, "meta data exceeds file system size");
if (x1 + (u_int64_t)x * bpb.bpbFATs > bpb.bpbHugeSectors) {
warnx("meta data exceeds file system size");
return -1;
}
x1 += x * bpb.bpbFATs;
x = (u_int64_t)(bpb.bpbHugeSectors - x1) * bpb.bpbBytesPerSec * NPB /
(bpb.bpbSecPerClust * bpb.bpbBytesPerSec * NPB + fat /
@ -481,9 +536,11 @@ int mkfs_msdos(const char *fname, const char *dtype,
if (bpb.bpbBigFATsecs < x2)
warnx("warning: sectors/FAT limits file system to %u clusters",
cls);
if (cls < mincls(fat))
errx(1, "%u clusters too few clusters for FAT%u, need %u", cls, fat,
if (cls < mincls(fat)) {
warnx("%u clusters too few clusters for FAT%u, need %u", cls, fat,
mincls(fat));
return -1;
}
if (cls > maxcls(fat)) {
cls = maxcls(fat);
bpb.bpbHugeSectors = x1 + (cls + 1) * bpb.bpbSecPerClust - 1;
@ -511,14 +568,18 @@ int mkfs_msdos(const char *fname, const char *dtype,
gettimeofday(&tv, NULL);
now = tv.tv_sec;
tm = localtime(&now);
if (!(img = malloc(bpb.bpbBytesPerSec)))
err(1, NULL);
if (!(img = malloc(bpb.bpbBytesPerSec))) {
warn(NULL);
return -1;
}
dir = bpb.bpbResSectors + (bpb.bpbFATsecs ? bpb.bpbFATsecs :
bpb.bpbBigFATsecs) * bpb.bpbFATs;
memset(&si_sa, 0, sizeof(si_sa));
si_sa.sa_handler = infohandler;
if (sigaction(SIGINFO, &si_sa, NULL) == -1)
err(1, "sigaction SIGINFO");
if (sigaction(SIGINFO, &si_sa, NULL) == -1) {
warn("sigaction SIGINFO");
return -1;
}
for (lsn = 0; lsn < dir + (fat == 32 ? bpb.bpbSecPerClust : rds); lsn++) {
if (got_siginfo) {
fprintf(stderr,"%s: writing sector %u of %u (%u%%)\n",
@ -533,14 +594,20 @@ int mkfs_msdos(const char *fname, const char *dtype,
fat == 32 && bpb.bpbBackup != MAXU16 &&
bss <= bpb.bpbBackup && x >= bpb.bpbBackup) {
x -= bpb.bpbBackup;
if (!x && lseek(fd1, o.offset, SEEK_SET))
err(1, "%s", bname);
if (!x && lseek(fd1, o.offset, SEEK_SET)) {
warn("%s", bname);
return -1;
}
}
if (o.bootstrap && x < bss) {
if ((n = read(fd1, img, bpb.bpbBytesPerSec)) == -1)
err(1, "%s", bname);
if ((unsigned)n != bpb.bpbBytesPerSec)
errx(1, "%s: can't read sector %u", bname, x);
if ((n = read(fd1, img, bpb.bpbBytesPerSec)) == -1) {
warn("%s", bname);
return -1;
}
if ((unsigned)n != bpb.bpbBytesPerSec) {
warnx("%s: can't read sector %u", bname, x);
return -1;
}
} else
memset(img, 0, bpb.bpbBytesPerSec);
if (!lsn ||
@ -627,19 +694,23 @@ int mkfs_msdos(const char *fname, const char *dtype,
(u_int)tm->tm_mday;
mk2(de->deMDate, x);
}
if ((n = write(fd, img, bpb.bpbBytesPerSec)) == -1)
err(1, "%s", fname);
if ((unsigned)n != bpb.bpbBytesPerSec)
errx(1, "%s: can't write sector %u", fname, lsn);
if ((n = write(fd, img, bpb.bpbBytesPerSec)) == -1) {
warn("%s", fname);
return -1;
}
if ((unsigned)n != bpb.bpbBytesPerSec) {
warnx("%s: can't write sector %u", fname, lsn);
return -1;
}
}
}
return 0;
}
/*
* Exit with error if file system is mounted.
* return -1 with error if file system is mounted.
*/
static void
static int
check_mounted(const char *fname, mode_t mode)
{
struct statfs *mp;
@ -647,8 +718,10 @@ check_mounted(const char *fname, mode_t mode)
size_t len;
int n, r;
if (!(n = getmntinfo(&mp, MNT_NOWAIT)))
err(1, "getmntinfo");
if (!(n = getmntinfo(&mp, MNT_NOWAIT))) {
warn("getmntinfo");
return -1;
}
len = strlen(_PATH_DEV);
s1 = fname;
if (!strncmp(s1, _PATH_DEV, len))
@ -659,30 +732,36 @@ check_mounted(const char *fname, mode_t mode)
if (!strncmp(s2, _PATH_DEV, len))
s2 += len;
if ((r && s2 != mp->f_mntfromname && !strcmp(s1 + 1, s2)) ||
!strcmp(s1, s2))
errx(1, "%s is mounted on %s", fname, mp->f_mntonname);
!strcmp(s1, s2)) {
warnx("%s is mounted on %s", fname, mp->f_mntonname);
return -1;
}
}
return 0;
}
/*
* Get a standard format.
*/
static void
static int
getstdfmt(const char *fmt, struct bpb *bpb)
{
u_int x, i;
x = sizeof(stdfmt) / sizeof(stdfmt[0]);
for (i = 0; i < x && strcmp(fmt, stdfmt[i].name); i++);
if (i == x)
errx(1, "%s: unknown standard format", fmt);
if (i == x) {
warnx("%s: unknown standard format", fmt);
return -1;
}
*bpb = stdfmt[i].bpb;
return 0;
}
/*
* Get disk slice, partition, and geometry information.
*/
static void
static int
getdiskinfo(int fd, const char *fname, const char *dtype, __unused int oflag,
struct bpb *bpb)
{
@ -750,16 +829,26 @@ getdiskinfo(int fd, const char *fname, const char *dtype, __unused int oflag,
lp = &dlp;
}
if (bpb->bpbBytesPerSec == 0)
bpb->bpbBytesPerSec = ckgeom(fname, lp->d_secsize, "bytes/sector");
if (bpb->bpbSecPerTrack == 0)
bpb->bpbSecPerTrack = ckgeom(fname, lp->d_nsectors, "sectors/track");
if (bpb->bpbHeads == 0)
bpb->bpbHeads = ckgeom(fname, lp->d_ntracks, "drive heads");
if (bpb->bpbBytesPerSec == 0) {
if (ckgeom(fname, lp->d_secsize, "bytes/sector") == -1)
return -1;
bpb->bpbBytesPerSec = lp->d_secsize;
}
if (bpb->bpbSecPerTrack == 0) {
if (ckgeom(fname, lp->d_nsectors, "sectors/track") == -1)
return -1;
bpb->bpbSecPerTrack = lp->d_nsectors;
}
if (bpb->bpbHeads == 0) {
if (ckgeom(fname, lp->d_ntracks, "drive heads") == -1)
return -1;
bpb->bpbHeads = lp->d_ntracks;
}
if (bpb->bpbHugeSectors == 0)
bpb->bpbHugeSectors = lp->d_secperunit;
if (bpb->bpbHiddenSecs == 0)
bpb->bpbHiddenSecs = hs;
return 0;
}
/*
@ -796,14 +885,18 @@ print_bpb(struct bpb *bpb)
/*
* Check a disk geometry value.
*/
static u_int
static int
ckgeom(const char *fname, u_int val, const char *msg)
{
if (!val)
errx(1, "%s: no default %s", fname, msg);
if (val > MAXU16)
errx(1, "%s: illegal %s %d", fname, msg, val);
return val;
if (!val) {
warnx("%s: no default %s", fname, msg);
return -1;
}
if (val > MAXU16) {
warnx("%s: illegal %s %d", fname, msg, val);
return -1;
}
return 0;
}
/*

View File

@ -159,7 +159,7 @@ main(int argc, char *argv[])
err(1, NULL);
}
dtype = *argv;
return mkfs_msdos(fname, dtype, &o);
return !!mkfs_msdos(fname, dtype, &o);
}
/*