wall: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

swapon: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

setpriv: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

fallocate: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

swaplabel: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

mkswap: unify write check to a file descriptor

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

mkfs.minix: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

mkfs.cramfs: unify write check to a file descriptor

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

mkfs.bfs: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

fsck.minix: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

fsck.cramfs: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

wdctl: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

sfdisk: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

cfdisk: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

resizepart: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

partx: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

fdformat: check writing to a file descriptor was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

include: add close_fd() for noticing write errors before close()

Essentially this helper function is similar to close_stream(), but for
file descriptors.

When a file descriptors are close()'d status of write is often
overlooked.  The close_fd() will try to determine what happen to writes
with fsync() before closing the file descriptor.

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

libblkid: check writing to a file was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

pg: check writing to a file was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

agetty: check writing to a file was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

setpriv: check writing to a file was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

cfdisk: check writing to a file was successful

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

setsid: exit when control terminal cannot be set

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

hexdump: add long options to the command

Includes update to bash completion, and manual as well.

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

setpriv: allow login and group name option arguments

For an average user names are easier to use than uid and gid numbers.

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

utmpdump: add option to write to a file

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

blkid, hwclock, ldattach: use program_invocation_short_name

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

bash-completion: add mount and umount

Signed-off-by: Sami Kerola <kerolasa@iki.fi>

Fix various typos

wipefs: clean up code

This is C, use int & bits for boolean stuff rather than army of
variables.

Signed-off-by: Karel Zak <kzak@redhat.com>

wipefs: add --backup

[kzak@redhat.com: - don't wipe if failed to create a backup file
                  - use ts_scsi_debug_init in the regression tests]

Signed-off-by: Ondrej Oprala <ooprala@redhat.com>
Signed-off-by: Karel Zak <kzak@redhat.com>

tests: partx tests cleanup

 * use ts_scsi_debug_init

 * use headers in output

 * use udevadm

Signed-off-by: Karel Zak <kzak@redhat.com>

tests: add partx regression tests

Signed-off-by: Ondrej Oprala <ooprala@redhat.com>

build-sys: release++ (v2.23)

Signed-off-by: Karel Zak <kzak@redhat.com>

docs: update v2.23-ReleaseNotes

Signed-off-by: Karel Zak <kzak@redhat.com>

docs: update AUTHORS file

Signed-off-by: Karel Zak <kzak@redhat.com>

po: merge changes

Signed-off-by: Karel Zak <kzak@redhat.com>

po: update uk.po (from translationproject.org)

po: update pl.po (from translationproject.org)

po: update nl.po (from translationproject.org)

po: update hr.po (from translationproject.org)

po: update fr.po (from translationproject.org)

po: update da.po (from translationproject.org)

po: update cs.po (from translationproject.org)

tests: make mount tests backwardly compatible with deprecated-mount/

Signed-off-by: Karel Zak <kzak@redhat.com>

mount: (deprecated) check loopcxt_ return values

Signed-off-by: Karel Zak <kzak@redhat.com>

mount: (deprecated) remove duplicate xstrndup()

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: use 'check' for all utils with dependencies

Signed-off-by: Karel Zak <kzak@redhat.com>

AUTHORS: fix prlimit typo

Signed-off-by: Davidlohr Bueso <dave@gnu.org>

libfdisk: gpt: remove leftover comments

I must have left those in some commit.

Signed-off-by: Davidlohr Bueso <dave@gnu.org>

rtcwake: pass -h to shutdown

The sysvinit shutdown requires you use -h when using -H or -P.
Other inits just don't seem to care.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

build-sys: fix non-linux build

Make enabling fdformat depend on being a linux build.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>

dmesg: minor cleanup of EXCL_xxx_ERR macros

Delete EXCL_ACT_ERR and EXCL_SYS_ERR defines:
usage is removed in 43d2eeef5dd3a64c372dbf60e4a32aa9e97b4645:
    "dmesg: use err_exclusive_options()"

Signed-off-by: Karel Zak <kzak@redhat.com>

rtcwake: add support for "freeze" mode

Supported by Linux since 3.9 (upstream commit
7e73c5ae6e7991a6c01f6d096ff8afaef4458c36).

Signed-off-by: Karel Zak <kzak@redhat.com>

umount: use devno to identify --all-target devices

 # mount -t tmpfs tmp /mnt/A
 # mount --bind /mnt/A /mnt/B
 # mount -t tmpfs tmp /mnt/C

the command:

  umount --all-targets /mnt/A

has to umount only A and B, but not C. The C is a another filesystem.

Signed-off-by: Karel Zak <kzak@redhat.com>

libfdisk: (gpt) fix EFI UUID usage

Signed-off-by: Karel Zak <kzak@redhat.com>

fdisk: fix list types

Signed-off-by: Karel Zak <kzak@redhat.com>

libfdisk: (gpt) fix valid_pmbr()

Signed-off-by: Karel Zak <kzak@redhat.com>

libfdisk: (gpt) use le64_to_cpu() for alternative LBA checks

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: remove obsolete code

 ... forgot in the previous commit to the swap prober.

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: ubifs is LE

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: make swap detection arch independent

Signed-off-by: Karel Zak <kzak@redhat.com>

tests: fix compiler warning

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: use pkg-config results for libuser

Our code depends on libuser >= 0.58, it's better to completely rely on
pkg-config results rather than also call UL_CHECK_LIB (that overwrites
previous pkg-config have_user= result independently on libuser
version).

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: fix UL_REQUIRES_HAVE

Signed-off-by: Karel Zak <kzak@redhat.com>

fdisks: fix compiler warning

Signed-off-by: Karel Zak <kzak@redhat.com>

sulogin: fix compiler warnings

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: fix nsenter requirements

Signed-off-by: Karel Zak <kzak@redhat.com>

include: fix SYS_{unshare,nsenter} usage

Signed-off-by: Karel Zak <kzak@redhat.com>

lscpu: set 32-bit / 64-bit for ppc

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: use more robust code in silicon_raid checksum prober

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: fix silicon raid checksum code on BE

Signed-off-by: Karel Zak <kzak@redhat.com>

tests: add BE minix test

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: arch independent minix detection

It seems that on-disk MINIX FS superblock is native-endian.

Addresses: https://bugzilla.redhat.com/show_bug.cgi?id=833841
Signed-off-by: Karel Zak <kzak@redhat.com>

tests: consolidate scsi_debug usage

Signed-off-by: Karel Zak <kzak@redhat.com>

libmount: add debug message to guess FS function

Signed-off-by: Karel Zak <kzak@redhat.com>

tests: use udevadm settle rather than sleep

Signed-off-by: Karel Zak <kzak@redhat.com>

libmount: fix mount.nfs segfault, rely on assert() rather than on nonnull

We use
   mnt_optstr_append_option(&o, mnt_fs_get_vfs_options(fs), NULL);

in mount.nfs, unfortunately mnt_optstr_append_option() has been marked
ass nonnull(1, 2). That's wrong because append and prepend should
robust enough to accept NULL as option name.

The patch also removes almost all __attribute__((nonnull). It seems
better to rely on assert() to have usable feedback. In many cases
(nonnull) is premature optimization for the library. This attribute
makes sense for things like strlen() or so...

Addresses: https://bugzilla.redhat.com/show_bug.cgi?id=948274
Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: release++ (v2.23-rc2)

Signed-off-by: Karel Zak <kzak@redhat.com>

docs: update v2.23-ReleaseNotes

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: make wall optional in bash-completion

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: add --disable-fdformat

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: add --disable-hwclock

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: add --disable-bfs

Signed-off-by: Karel Zak <kzak@redhat.com>

build-sys: add --disable-cytune

Signed-off-by: Karel Zak <kzak@redhat.com>

docs: update AUTHORS file

Signed-off-by: Karel Zak <kzak@redhat.com>

po: merge changes

Signed-off-by: Karel Zak <kzak@redhat.com>

blkid: use xalloc everywhere

Signed-off-by: Karel Zak <kzak@redhat.com>

tests: make line(1) optional

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: remove duplicate includes

Signed-off-by: Karel Zak <kzak@redhat.com>

lib/xalloc: add xstrndup()

Signed-off-by: Karel Zak <kzak@redhat.com>

tests: update lscpu tests

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: fix ZFS superblok offset for SBMAGIC

 # wipefs /dev/sdd1

 offset               type
 ----------------------------------------------------------------
 0x65000              zfs_member   [raid]
                      LABEL: zm1
                      UUID:  18067835215330778964

 and "wipefs -o 0x65000 /dev/sdd1" seems to run in an endless loop.

Reported-by: Helmut Hullen <Hullen@t-online.de>
Signed-off-by: Karel Zak <kzak@redhat.com>

bash-completion: add swapoff

[kzak@redhat.com: remove short options]

Signed-off-by: Bernhard Voelker <mail@bernhard-voelker.de>
Signed-off-by: Karel Zak <kzak@redhat.com>

lscpu: add Model name

Signed-off-by: Karel Zak <kzak@redhat.com>

su: remove '+' from getopt_long()

 # /bin/su - -c 'echo test'
 su: user -c does not exist

Addresses: https://bugzilla.redhat.com/show_bug.cgi?id=950495
Signed-off-by: Karel Zak <kzak@redhat.com>

loopdev: sync capacity after setting it

I recently tried to mount an hfsplus file system from an image file with
a partition table by using the loop offset and sizelimit options to specify
the location of the file system.

hfsplus stores some metadata at a set offset from the end of the partition,
so it's sensitive to the device size reported by the kernel.

It worked with this:

But failed with this:

/dev/loop0: [0089]:2 (<imagefile>), offset 32768, sizelimit 102400000
/dev/loop1: [0089]:2 (<imagefile>), offset 32768, sizelimit 102400000

/proc/partitions shows the correct number of blocks to match the sizelimit.

But if I set a breakpoint in mount before the mount syscall, I could see:
102400000
102432768

The kernel loop driver will set the gendisk capacity of the device at
LOOP_SET_STATUS64 but won't sync it to the block device until one of two
conditions are met: All open file descriptors referring to the device are
closed (and it will sync when re-opened) or if the LOOP_SET_CAPACITY ioctl
is called to sync it. Since mount opens the device and passes it directly
to the mount syscall after LOOP_SET_STATUS64 without closing and reopening
it, the sizelimit argument is effectively ignroed. The capacity needs to
be synced immediately for it to work as expected.

This patch adds the LOOP_SET_CAPACITY call to loopctx_setup_device since
the device isn't yet released to the user, so it's safe to sync the capacity
immediately.

[kzak@redhat.com: - port to the current git HEAD,
                  - use uint64_t]

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: export blkid_init_debug()

Signed-off-by: Karel Zak <kzak@redhat.com>

libblkid: support LIBBLKID_DEBUG= only

 ... the BLKID_DEBUG= is not more supported.

Signed-off-by: Karel Zak <kzak@redhat.com>