summaryrefslogtreecommitdiffstats
path: root/disk-utils/mkswap.c
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--disk-utils/mkswap.c746
1 files changed, 746 insertions, 0 deletions
diff --git a/disk-utils/mkswap.c b/disk-utils/mkswap.c
new file mode 100644
index 0000000..30c4ea8
--- /dev/null
+++ b/disk-utils/mkswap.c
@@ -0,0 +1,746 @@
+/*
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * mkswap.c - set up a linux swap device
+ *
+ * Copyright (C) 1991 Linus Torvalds
+ * 20.12.91 - time began. Got VM working yesterday by doing this by hand.
+ *
+ * Copyright (C) 1999 Jakub Jelinek <jj@ultra.linux.cz>
+ * Copyright (C) 2007-2014 Karel Zak <kzak@redhat.com>
+ */
+
+#include <stdio.h>
+#include <unistd.h>
+#include <string.h>
+#include <fcntl.h>
+#include <stdlib.h>
+#include <limits.h>
+#include <sys/utsname.h>
+#include <sys/stat.h>
+#include <sys/ioctl.h>
+#include <errno.h>
+#include <getopt.h>
+#include <assert.h>
+#ifdef HAVE_LIBSELINUX
+# include <selinux/selinux.h>
+# include <selinux/context.h>
+# include "selinux-utils.h"
+#endif
+#ifdef HAVE_LINUX_FIEMAP_H
+# include <linux/fs.h>
+# include <linux/fiemap.h>
+#endif
+
+#include "linux_version.h"
+#include "swapheader.h"
+#include "strutils.h"
+#include "nls.h"
+#include "blkdev.h"
+#include "pathnames.h"
+#include "all-io.h"
+#include "xalloc.h"
+#include "c.h"
+#include "closestream.h"
+#include "ismounted.h"
+#include "optutils.h"
+#include "bitops.h"
+
+#ifdef HAVE_LIBUUID
+# include <uuid.h>
+#endif
+
+#ifdef HAVE_LIBBLKID
+# include <blkid.h>
+#endif
+
+#define MIN_GOODPAGES 10
+
+#define SELINUX_SWAPFILE_TYPE "swapfile_t"
+
+enum ENDIANNESS {
+ ENDIANNESS_NATIVE,
+ ENDIANNESS_LITTLE,
+ ENDIANNESS_BIG,
+};
+
+struct mkswap_control {
+ struct swap_header_v1_2 *hdr; /* swap header */
+ void *signature_page;/* buffer with swap header */
+
+ char *devname; /* device or file name */
+ const char *lockmode; /* as specified by --lock */
+ struct stat devstat; /* stat() result */
+ int fd; /* swap file descriptor */
+
+ unsigned long long npages; /* number of pages */
+ unsigned long nbadpages; /* number of bad pages */
+
+ int user_pagesize; /* --pagesize */
+ int pagesize; /* final pagesize used for the header */
+
+ char *opt_label; /* LABEL as specified on command line */
+ unsigned char *uuid; /* UUID parsed by libbuuid */
+
+ size_t nbad_extents;
+
+ enum ENDIANNESS endianness;
+
+ unsigned int check:1, /* --check */
+ verbose:1, /* --verbose */
+ quiet:1, /* --quiet */
+ force:1; /* --force */
+};
+
+static uint32_t cpu32_to_endianness(uint32_t v, enum ENDIANNESS e)
+{
+ switch (e) {
+ case ENDIANNESS_NATIVE: return v;
+ case ENDIANNESS_LITTLE: return cpu_to_le32(v);
+ case ENDIANNESS_BIG: return cpu_to_be32(v);
+ }
+ abort();
+}
+
+static void init_signature_page(struct mkswap_control *ctl)
+{
+ const int kernel_pagesize = getpagesize();
+
+ if (ctl->user_pagesize) {
+ if (ctl->user_pagesize < 0 || !is_power_of_2(ctl->user_pagesize) ||
+ (size_t) ctl->user_pagesize < sizeof(struct swap_header_v1_2) + 10)
+ errx(EXIT_FAILURE,
+ _("Bad user-specified page size %u"),
+ ctl->user_pagesize);
+ if (!ctl->quiet && ctl->user_pagesize != kernel_pagesize)
+ warnx(_("Using user-specified page size %d, "
+ "instead of the system value %d"),
+ ctl->user_pagesize, kernel_pagesize);
+ ctl->pagesize = ctl->user_pagesize;
+ } else
+ ctl->pagesize = kernel_pagesize;
+
+ ctl->signature_page = xcalloc(1, ctl->pagesize);
+ ctl->hdr = (struct swap_header_v1_2 *) ctl->signature_page;
+}
+
+static void deinit_signature_page(struct mkswap_control *ctl)
+{
+ free(ctl->signature_page);
+
+ ctl->hdr = NULL;
+ ctl->signature_page = NULL;
+}
+
+static void set_signature(const struct mkswap_control *ctl)
+{
+ char *sp = (char *) ctl->signature_page;
+
+ assert(sp);
+ memcpy(sp + ctl->pagesize - SWAP_SIGNATURE_SZ, SWAP_SIGNATURE, SWAP_SIGNATURE_SZ);
+}
+
+static void set_uuid_and_label(const struct mkswap_control *ctl)
+{
+ assert(ctl);
+ assert(ctl->hdr);
+
+ /* set UUID */
+ if (ctl->uuid)
+ memcpy(ctl->hdr->uuid, ctl->uuid, sizeof(ctl->hdr->uuid));
+
+ /* set LABEL */
+ if (ctl->opt_label) {
+ xstrncpy(ctl->hdr->volume_name,
+ ctl->opt_label, sizeof(ctl->hdr->volume_name));
+ if (!ctl->quiet
+ && strlen(ctl->opt_label) > strlen(ctl->hdr->volume_name))
+ warnx(_("Label was truncated."));
+ }
+
+ /* report results */
+ if (!ctl->quiet && (ctl->uuid || ctl->opt_label)) {
+ if (ctl->opt_label)
+ printf("LABEL=%s, ", ctl->hdr->volume_name);
+ else
+ printf(_("no label, "));
+#ifdef HAVE_LIBUUID
+ if (ctl->uuid) {
+ char uuid_string[UUID_STR_LEN];
+ uuid_unparse(ctl->uuid, uuid_string);
+ printf("UUID=%s\n", uuid_string);
+ } else
+#endif
+ printf(_("no uuid\n"));
+ }
+}
+
+static void __attribute__((__noreturn__)) usage(void)
+{
+ FILE *out = stdout;
+
+ fputs(USAGE_HEADER, out);
+ fprintf(out, _(" %s [options] device [size]\n"), program_invocation_short_name);
+
+ fputs(USAGE_SEPARATOR, out);
+ fputs(_("Set up a Linux swap area.\n"), out);
+
+ fputs(USAGE_OPTIONS, out);
+ fputs(_(" -c, --check check bad blocks before creating the swap area\n"), out);
+ fputs(_(" -f, --force allow swap size area be larger than device\n"), out);
+ fputs(_(" -q, --quiet suppress output and warning messages\n"), out);
+ fputs(_(" -p, --pagesize SIZE specify page size in bytes\n"), out);
+ fputs(_(" -L, --label LABEL specify label\n"), out);
+ fputs(_(" -v, --swapversion NUM specify swap-space version number\n"), out);
+ fputs(_(" -U, --uuid UUID specify the uuid to use\n"), out);
+ fprintf(out,
+ _(" -e, --endianness=<value> specify the endianness to use "
+ "(%s, %s or %s)\n"), "native", "little", "big");
+ fputs(_(" --verbose verbose output\n"), out);
+
+ fprintf(out,
+ _(" --lock[=<mode>] use exclusive device lock (%s, %s or %s)\n"), "yes", "no", "nonblock");
+
+ printf(USAGE_HELP_OPTIONS(27));
+
+ printf(USAGE_MAN_TAIL("mkswap(8)"));
+ exit(EXIT_SUCCESS);
+}
+
+static void page_bad(struct mkswap_control *ctl, unsigned int page)
+{
+ const unsigned long max_badpages =
+ (ctl->pagesize - 1024 - 128 * sizeof(int) - 10) / sizeof(int);
+
+ if (ctl->nbadpages == max_badpages)
+ errx(EXIT_FAILURE, _("too many bad pages: %lu"), max_badpages);
+
+ ctl->hdr->badpages[ctl->nbadpages] = page;
+ ctl->nbadpages++;
+}
+
+static void check_blocks(struct mkswap_control *ctl)
+{
+ unsigned int current_page = 0;
+ int do_seek = 1;
+ char *buffer;
+
+ assert(ctl);
+ assert(ctl->fd > -1);
+
+ buffer = xmalloc(ctl->pagesize);
+ while (current_page < ctl->npages) {
+ ssize_t rc;
+ off_t offset = (off_t) current_page * ctl->pagesize;
+
+ if (do_seek && lseek(ctl->fd, offset, SEEK_SET) != offset)
+ errx(EXIT_FAILURE, _("seek failed in check_blocks"));
+
+ rc = read(ctl->fd, buffer, ctl->pagesize);
+ do_seek = (rc < 0 || rc != ctl->pagesize);
+ if (do_seek)
+ page_bad(ctl, current_page);
+ current_page++;
+ }
+
+ if (!ctl->quiet)
+ printf(P_("%lu bad page\n", "%lu bad pages\n", ctl->nbadpages), ctl->nbadpages);
+ free(buffer);
+}
+
+
+#ifdef HAVE_LINUX_FIEMAP_H
+static void warn_extent(struct mkswap_control *ctl, const char *msg, uint64_t off)
+{
+ if (ctl->nbad_extents == 0) {
+ fputc('\n', stderr);
+ fprintf(stderr, _(
+
+ "mkswap: %s contains holes or other unsupported extents.\n"
+ " This swap file can be rejected by kernel on swap activation!\n"),
+ ctl->devname);
+
+ if (ctl->verbose)
+ fputc('\n', stderr);
+ else
+ fprintf(stderr, _(
+ " Use --verbose for more details.\n"));
+
+ }
+ if (ctl->verbose) {
+ fputs(" - ", stderr);
+ fprintf(stderr, msg, off);
+ fputc('\n', stderr);
+ }
+ ctl->nbad_extents++;
+}
+
+static void check_extents(struct mkswap_control *ctl)
+{
+ char buf[BUFSIZ] = { 0 };
+ struct fiemap *fiemap = (struct fiemap *) buf;
+ int last = 0;
+ uint64_t last_logical = 0;
+
+ memset(fiemap, 0, sizeof(struct fiemap));
+
+ do {
+ int rc;
+ size_t n, i;
+
+ fiemap->fm_length = ~0ULL;
+ fiemap->fm_flags = FIEMAP_FLAG_SYNC;
+ fiemap->fm_extent_count =
+ (sizeof(buf) - sizeof(*fiemap)) / sizeof(struct fiemap_extent);
+
+ rc = ioctl(ctl->fd, FS_IOC_FIEMAP, (unsigned long) fiemap);
+ if (rc < 0)
+ return;
+
+ n = fiemap->fm_mapped_extents;
+ if (n == 0)
+ break;
+
+ for (i = 0; i < n; i++) {
+ struct fiemap_extent *e = &fiemap->fm_extents[i];
+
+ if (e->fe_logical > last_logical)
+ warn_extent(ctl, _("hole detected at offset %ju"),
+ (uintmax_t) last_logical);
+
+ last_logical = (e->fe_logical + e->fe_length);
+
+ if (e->fe_flags & FIEMAP_EXTENT_LAST)
+ last = 1;
+ if (e->fe_flags & FIEMAP_EXTENT_DATA_INLINE)
+ warn_extent(ctl, _("data inline extent at offset %ju"),
+ (uintmax_t) e->fe_logical);
+ if (e->fe_flags & FIEMAP_EXTENT_SHARED)
+ warn_extent(ctl, _("shared extent at offset %ju"),
+ (uintmax_t) e->fe_logical);
+ if (e->fe_flags & FIEMAP_EXTENT_DELALLOC)
+ warn_extent(ctl, _("unallocated extent at offset %ju"),
+ (uintmax_t) e->fe_logical);
+
+ if (!ctl->verbose && ctl->nbad_extents)
+ goto done;
+ }
+ fiemap->fm_start = fiemap->fm_extents[n - 1].fe_logical
+ + fiemap->fm_extents[n - 1].fe_length;
+ } while (last == 0);
+
+ if (last_logical < (uint64_t) ctl->devstat.st_size)
+ warn_extent(ctl, _("hole detected at offset %ju"),
+ (uintmax_t) last_logical);
+done:
+ if (ctl->nbad_extents)
+ fputc('\n', stderr);
+}
+#endif /* HAVE_LINUX_FIEMAP_H */
+
+/* return size in pages */
+static unsigned long long get_size(const struct mkswap_control *ctl)
+{
+ int fd;
+ unsigned long long size;
+
+ fd = open(ctl->devname, O_RDONLY);
+ if (fd < 0)
+ err(EXIT_FAILURE, _("cannot open %s"), ctl->devname);
+ if (blkdev_get_size(fd, &size) < 0)
+ err(EXIT_FAILURE, _("cannot determine size of %s"), ctl->devname);
+ size /= ctl->pagesize;
+
+ close(fd);
+ return size;
+}
+
+#ifdef HAVE_LIBBLKID
+static blkid_probe new_prober(const struct mkswap_control *ctl)
+{
+ blkid_probe pr = blkid_new_probe();
+ if (!pr)
+ errx(EXIT_FAILURE, _("unable to alloc new libblkid probe"));
+ if (blkid_probe_set_device(pr, ctl->fd, 0, 0))
+ errx(EXIT_FAILURE, _("unable to assign device to libblkid probe"));
+ return pr;
+}
+#endif
+
+static void open_device(struct mkswap_control *ctl)
+{
+ assert(ctl);
+ assert(ctl->devname);
+
+ if (stat(ctl->devname, &ctl->devstat) < 0)
+ err(EXIT_FAILURE, _("stat of %s failed"), ctl->devname);
+ ctl->fd = open_blkdev_or_file(&ctl->devstat, ctl->devname, O_RDWR);
+ if (ctl->fd < 0)
+ err(EXIT_FAILURE, _("cannot open %s"), ctl->devname);
+
+ if (blkdev_lock(ctl->fd, ctl->devname, ctl->lockmode) != 0)
+ exit(EXIT_FAILURE);
+
+ if (ctl->check && S_ISREG(ctl->devstat.st_mode)) {
+ ctl->check = 0;
+ if (!ctl->quiet)
+ warnx(_("warning: checking bad blocks from swap file is not supported: %s"),
+ ctl->devname);
+ }
+}
+
+static void wipe_device(struct mkswap_control *ctl)
+{
+ char *type = NULL;
+ int zap = 1;
+#ifdef HAVE_LIBBLKID
+ blkid_probe pr = NULL;
+ const char *v = NULL;
+#endif
+ if (!ctl->force) {
+ if (lseek(ctl->fd, 0, SEEK_SET) != 0)
+ errx(EXIT_FAILURE, _("unable to rewind swap-device"));
+
+#ifdef HAVE_LIBBLKID
+ pr = new_prober(ctl);
+ blkid_probe_enable_partitions(pr, 1);
+ blkid_probe_enable_superblocks(pr, 0);
+
+ if (blkid_do_fullprobe(pr) == 0 &&
+ blkid_probe_lookup_value(pr, "PTTYPE", &v, NULL) == 0 && v) {
+ type = xstrdup(v);
+ zap = 0;
+ }
+#else
+ /* don't zap if compiled without libblkid */
+ zap = 0;
+#endif
+ }
+
+ if (zap) {
+ /*
+ * Wipe bootbits
+ */
+ char buf[1024] = { '\0' };
+
+ if (lseek(ctl->fd, 0, SEEK_SET) != 0)
+ errx(EXIT_FAILURE, _("unable to rewind swap-device"));
+
+ if (write_all(ctl->fd, buf, sizeof(buf)))
+ errx(EXIT_FAILURE, _("unable to erase bootbits sectors"));
+#ifdef HAVE_LIBBLKID
+ /*
+ * Wipe rest of the device
+ */
+ if (!pr)
+ pr = new_prober(ctl);
+
+ blkid_probe_enable_superblocks(pr, 1);
+ blkid_probe_enable_partitions(pr, 0);
+ blkid_probe_set_superblocks_flags(pr, BLKID_SUBLKS_MAGIC|BLKID_SUBLKS_TYPE);
+
+ while (blkid_do_probe(pr) == 0) {
+ const char *data = NULL;
+
+ if (!ctl->quiet
+ && blkid_probe_lookup_value(pr, "TYPE", &data, NULL) == 0 && data)
+ warnx(_("%s: warning: wiping old %s signature."), ctl->devname, data);
+ blkid_do_wipe(pr, 0);
+ }
+#endif
+ } else if (!ctl->quiet) {
+ warnx(_("%s: warning: don't erase bootbits sectors"),
+ ctl->devname);
+ if (type)
+ fprintf(stderr, _(" (%s partition table detected). "), type);
+ else
+ fprintf(stderr, _(" (compiled without libblkid). "));
+ fprintf(stderr, _("Use -f to force.\n"));
+ }
+ free(type);
+#ifdef HAVE_LIBBLKID
+ blkid_free_probe(pr);
+#endif
+}
+
+#define SIGNATURE_OFFSET 1024
+
+static void write_header_to_device(struct mkswap_control *ctl)
+{
+ assert(ctl);
+ assert(ctl->fd > -1);
+ assert(ctl->signature_page);
+
+ if (lseek(ctl->fd, SIGNATURE_OFFSET, SEEK_SET) != SIGNATURE_OFFSET)
+ errx(EXIT_FAILURE, _("unable to rewind swap-device"));
+
+ if (write_all(ctl->fd, (char *) ctl->signature_page + SIGNATURE_OFFSET,
+ ctl->pagesize - SIGNATURE_OFFSET) == -1)
+ err(EXIT_FAILURE,
+ _("%s: unable to write signature page"),
+ ctl->devname);
+}
+
+int main(int argc, char **argv)
+{
+ struct mkswap_control ctl = { .fd = -1, .endianness = ENDIANNESS_NATIVE };
+ int c, permMask;
+ uint64_t sz;
+ int version = SWAP_VERSION;
+ char *block_count = NULL, *strsz = NULL;
+#ifdef HAVE_LIBUUID
+ const char *opt_uuid = NULL;
+ uuid_t uuid_dat;
+#endif
+ enum {
+ OPT_LOCK = CHAR_MAX + 1,
+ OPT_VERBOSE
+ };
+ static const struct option longopts[] = {
+ { "check", no_argument, NULL, 'c' },
+ { "force", no_argument, NULL, 'f' },
+ { "quiet", no_argument, NULL, 'q' },
+ { "pagesize", required_argument, NULL, 'p' },
+ { "label", required_argument, NULL, 'L' },
+ { "swapversion", required_argument, NULL, 'v' },
+ { "uuid", required_argument, NULL, 'U' },
+ { "endianness", required_argument, NULL, 'e' },
+ { "version", no_argument, NULL, 'V' },
+ { "help", no_argument, NULL, 'h' },
+ { "lock", optional_argument, NULL, OPT_LOCK },
+ { "verbose", no_argument, NULL, OPT_VERBOSE },
+ { NULL, 0, NULL, 0 }
+ };
+
+ static const ul_excl_t excl[] = { /* rows and cols in ASCII order */
+ { 'c', 'q' },
+ { 0 }
+ };
+ int excl_st[ARRAY_SIZE(excl)] = UL_EXCL_STATUS_INIT;
+
+ setlocale(LC_ALL, "");
+ bindtextdomain(PACKAGE, LOCALEDIR);
+ textdomain(PACKAGE);
+ close_stdout_atexit();
+
+ while((c = getopt_long(argc, argv, "cfp:qL:v:U:e:Vh", longopts, NULL)) != -1) {
+
+ err_exclusive_options(c, longopts, excl, excl_st);
+
+ switch (c) {
+ case 'c':
+ ctl.check = 1;
+ break;
+ case 'f':
+ ctl.force = 1;
+ break;
+ case 'p':
+ ctl.user_pagesize = strtou32_or_err(optarg, _("parsing page size failed"));
+ break;
+ case 'q':
+ ctl.quiet = 1;
+ break;
+ case 'L':
+ ctl.opt_label = optarg;
+ break;
+ case 'v':
+ version = strtos32_or_err(optarg, _("parsing version number failed"));
+ if (version != SWAP_VERSION)
+ errx(EXIT_FAILURE,
+ _("swapspace version %d is not supported"), version);
+ break;
+ case 'U':
+#ifdef HAVE_LIBUUID
+ opt_uuid = optarg;
+#else
+ warnx(_("warning: ignoring -U (UUIDs are unsupported by %s)"),
+ program_invocation_short_name);
+#endif
+ break;
+ case 'e':
+ if (strcmp(optarg, "native") == 0) {
+ ctl.endianness = ENDIANNESS_NATIVE;
+ } else if (strcmp(optarg, "little") == 0) {
+ ctl.endianness = ENDIANNESS_LITTLE;
+ } else if (strcmp(optarg, "big") == 0) {
+ ctl.endianness = ENDIANNESS_BIG;
+ } else {
+ errx(EXIT_FAILURE,
+ _("invalid endianness %s is not supported"), optarg);
+ }
+ break;
+ case 'V':
+ print_version(EXIT_SUCCESS);
+ break;
+ case OPT_LOCK:
+ ctl.lockmode = "1";
+ if (optarg) {
+ if (*optarg == '=')
+ optarg++;
+ ctl.lockmode = optarg;
+ }
+ break;
+ case OPT_VERBOSE:
+ ctl.verbose = 1;
+ break;
+ case 'h':
+ usage();
+ default:
+ errtryhelp(EXIT_FAILURE);
+ }
+ }
+
+ if (optind < argc)
+ ctl.devname = argv[optind++];
+ if (optind < argc)
+ block_count = argv[optind++];
+ if (optind != argc) {
+ warnx(_("only one device argument is currently supported"));
+ errtryhelp(EXIT_FAILURE);
+ }
+
+#ifdef HAVE_LIBUUID
+ if(opt_uuid) {
+ if (strcmp(opt_uuid, "clear") == 0)
+ uuid_clear(uuid_dat);
+ else if (strcmp(opt_uuid, "random") == 0)
+ uuid_generate_random(uuid_dat);
+ else if (strcmp(opt_uuid, "time") == 0)
+ uuid_generate_time(uuid_dat);
+ else if (uuid_parse(opt_uuid, uuid_dat) != 0)
+ errx(EXIT_FAILURE, _("error: parsing UUID failed"));
+ } else
+ uuid_generate(uuid_dat);
+ ctl.uuid = uuid_dat;
+#endif
+
+ init_signature_page(&ctl); /* get pagesize and allocate signature page */
+
+ if (!ctl.devname) {
+ warnx(_("error: Nowhere to set up swap on?"));
+ errtryhelp(EXIT_FAILURE);
+ }
+ if (block_count) {
+ /* this silly user specified the number of blocks explicitly */
+ uint64_t blks = strtou64_or_err(block_count,
+ _("invalid block count argument"));
+ ctl.npages = blks / (ctl.pagesize / 1024);
+ }
+
+ sz = get_size(&ctl);
+ if (!ctl.npages)
+ ctl.npages = sz;
+ else if (ctl.npages > sz && !ctl.force)
+ errx(EXIT_FAILURE,
+ _("error: "
+ "size %llu KiB is larger than device size %"PRIu64" KiB"),
+ ctl.npages * (ctl.pagesize / 1024), sz * (ctl.pagesize / 1024));
+
+ if (ctl.npages < MIN_GOODPAGES)
+ errx(EXIT_FAILURE,
+ _("error: swap area needs to be at least %ld KiB"),
+ (long)(MIN_GOODPAGES * ctl.pagesize / 1024));
+ if (ctl.npages > UINT32_MAX) {
+ /* true when swap is bigger than 17.59 terabytes */
+ ctl.npages = UINT32_MAX;
+ if (!ctl.quiet)
+ warnx(_("warning: truncating swap area to %llu KiB"),
+ ctl.npages * ctl.pagesize / 1024);
+ }
+
+ if (is_mounted(ctl.devname))
+ errx(EXIT_FAILURE, _("error: "
+ "%s is mounted; will not make swapspace"),
+ ctl.devname);
+
+ open_device(&ctl);
+ permMask = S_ISBLK(ctl.devstat.st_mode) ? 07007 : 07077;
+ if (!ctl.quiet && (ctl.devstat.st_mode & permMask) != 0)
+ warnx(_("%s: insecure permissions %04o, fix with: chmod %04o %s"),
+ ctl.devname, ctl.devstat.st_mode & 07777,
+ ~permMask & 0666, ctl.devname);
+ if (!ctl.quiet
+ && getuid() == 0 && S_ISREG(ctl.devstat.st_mode) && ctl.devstat.st_uid != 0)
+ warnx(_("%s: insecure file owner %d, fix with: chown 0:0 %s"),
+ ctl.devname, ctl.devstat.st_uid, ctl.devname);
+
+
+ if (ctl.check)
+ check_blocks(&ctl);
+#ifdef HAVE_LINUX_FIEMAP_H
+ if (!ctl.quiet && S_ISREG(ctl.devstat.st_mode))
+ check_extents(&ctl);
+#endif
+
+ wipe_device(&ctl);
+
+ assert(ctl.hdr);
+ ctl.hdr->version = cpu32_to_endianness(version, ctl.endianness);
+ ctl.hdr->last_page = cpu32_to_endianness(ctl.npages - 1, ctl.endianness);
+ ctl.hdr->nr_badpages = cpu32_to_endianness(ctl.nbadpages, ctl.endianness);
+
+ if ((ctl.npages - MIN_GOODPAGES) < ctl.nbadpages)
+ errx(EXIT_FAILURE, _("Unable to set up swap-space: unreadable"));
+
+ sz = (ctl.npages - ctl.nbadpages - 1) * ctl.pagesize;
+ strsz = size_to_human_string(SIZE_SUFFIX_SPACE | SIZE_SUFFIX_3LETTER, sz);
+
+ if (!ctl.quiet)
+ printf(_("Setting up swapspace version %d, size = %s (%"PRIu64" bytes)\n"),
+ version, strsz, sz);
+ free(strsz);
+
+ set_signature(&ctl);
+ set_uuid_and_label(&ctl);
+
+ write_header_to_device(&ctl);
+
+ deinit_signature_page(&ctl);
+
+#ifdef HAVE_LIBSELINUX
+ if (S_ISREG(ctl.devstat.st_mode) && is_selinux_enabled() > 0) {
+ const char *context_string;
+ char *oldcontext;
+ context_t newcontext;
+
+ if (fgetfilecon(ctl.fd, &oldcontext) < 0) {
+ if (errno != ENODATA)
+ err(EXIT_FAILURE,
+ _("%s: unable to obtain selinux file label"),
+ ctl.devname);
+ if (ul_selinux_get_default_context(ctl.devname,
+ ctl.devstat.st_mode, &oldcontext))
+ errx(EXIT_FAILURE,
+ _("%s: unable to obtain default selinux file label"),
+ ctl.devname);
+ }
+ if (!(newcontext = context_new(oldcontext)))
+ errx(EXIT_FAILURE, _("unable to create new selinux context"));
+ if (context_type_set(newcontext, SELINUX_SWAPFILE_TYPE))
+ errx(EXIT_FAILURE, _("couldn't compute selinux context"));
+
+ context_string = context_str(newcontext);
+
+ if (strcmp(context_string, oldcontext)!=0) {
+ if (fsetfilecon(ctl.fd, context_string) && errno != ENOTSUP)
+ err(EXIT_FAILURE, _("unable to relabel %s to %s"),
+ ctl.devname, context_string);
+ }
+ context_free(newcontext);
+ freecon(oldcontext);
+ }
+#endif
+ /*
+ * A subsequent swapon() will fail if the signature
+ * is not actually on disk. (This is a kernel bug.)
+ * The fsync() in close_fd() will take care of writing.
+ */
+ if (close_fd(ctl.fd) != 0)
+ err(EXIT_FAILURE, _("write failed"));
+ return EXIT_SUCCESS;
+}