libmount: cleanup post-umount operations

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

libmount: fix MS_BIND|MS_REC usage, improve some bitwise operations

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

libmount: add post-mount checks to detect ro/rw

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

libmount: avoid duplicate loopdevs

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

mount: (new) print error messages

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

libmount: allow to toggle ro/rw and mount again

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

libmount: allow to use UUID and LABEL for owner/group mount options in fstab

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

libmount: add functions to export syscall/helpers status

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

tests: fix mount/special to be compatible with libmount

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

libmount: fix auto-loopdev

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

libmount: support umount /path/file.img

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

tests: add exceptions for libmount

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

libmount: make options in fstab optional

 .. to be compatible with old mount(8).

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

tests: update tests with fdisk output

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

tests: remove unssuported losetup -s

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

tests: remove unexpected set -x

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

tests: cleanup paths

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

build-sys: suport static build for new [u]mount

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

build-sys: add --enable-new-mount

 Now we have three versions of the mount(8) utility

 * old mount(8)
   --enable-mount [default]

 * old mount(8) linked with libmount
   This is this is necessary for systems without mtab file.
   --enable-libmount-mount

 - new mount(8)
   This is completely new pure-libmount based mount(8).
   --enable-new-mount

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

libmount: add --pass-fd to samples/mount

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

libmount: add support for encrypted loopdevs

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

losetup: add note about non-root users to losetup.8

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

losetup: move xgetpass() to lib/xgetpass.c

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

build-sys: move swapon from mount/ to sys-utils/

 ... to make it independent on mount stuff.

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

swapon: merge swap_constants.h into swapon.c

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

docs: add info about virt-what to TODO

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

mount: add hint that context= has to be quoted

Addresses: http://bugzilla.redhat.com/show_bug.cgi?id=747038
Signed-off-by: Karel Zak <kzak@redhat.com>

mount: update info about check= mount option

Addresses: http://bugzilla.redhat.com/show_bug.cgi?id=770172
Signed-off-by: Karel Zak <kzak@redhat.com>

losetup: add --partscan option

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

loopdev: improve debug messages

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

loopdev: support LO_FLAGS_PARTSCAN flag (kernel 3.2)

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

sfdisk: use is_blkdev

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

blockdev: make functions static

make the functions in blockdev to statics.

Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>

lscpu: fix compiler warnings

The string format is not being passed triggering:

    lscpu.c: In function ‘read_hypervisor’:
    lscpu.c:545:4: warning: format not a string literal and no format arguments
    lscpu.c: In function ‘get_cell_header’:
    lscpu.c:904:2: warning: format not a string literal and no format arguments
    lscpu.c:904:2: warning: format not a string literal and no format arguments

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

blkdev: add is_blkdev function

We should have the most basic of checks in this library to see whether or not a block device is being used.

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

fdisk: remove PACKED macro

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

sfdisk: remove PACKED macro

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

sfdisk: remove do_warn

Replace it with warnx() instead.

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

sfdisk: remove fatal

Since fatal() exists with a standard 1 code, it can easily be replaced by errx(3).
The xalloc lib is also introduced in this patch.

[kzak@redhat,com: - fix coding style]

Signed-off-by: Davidlohr Bueso <dave@gnu.org>
Signed-off-by: Karel Zak <kzak@redhat.com>

column: don't check result from xmalloc()

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

build-sys: add checkxalloc to toplevel Makefile

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

column: use xalloc libs

Signed-off-by: Dave Reisner <dreisner@archlinux.org>

lscpu: use xalloc libs

Signed-off-by: Dave Reisner <dreisner@archlinux.org>

tools: add new code checking script

Finds usage of strdup, malloc, calloc, and realloc when xalloc.h is
included.

http://marc.info/?l=util-linux-ng&m=132438338929925&w=2

Signed-off-by: Dave Reisner <dreisner@archlinux.org>

libmount: fixed a small typo

fdisk: fix incorrect position of is_ide_cdrom_or_tape() call

Reported-by: Guillem Jover <guillem@hadrons.org>
Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

libmount: fix typo in function name

Build broken in c70d9d7.

Signed-off-by: Dave Reisner <dreisner@archlinux.org>

libmount: consolidate MNT_FS_* code

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

mount: don't canonicalize the source of a pseudofs

In line with previous patches, the source of a pseudofs is strictly
semantic; do not attempt to resolve it to a filesystem path.

As a side effect of this, nspec needs to be initialized to NULL since
it may not be assigned null by spec_to_devname when the FS is deemed
a pseudofs.

Signed-off-by: Dave Reisner <dreisner@archlinux.org>

libmount: avoid resolving pseudofs source on update

Similar to the previous commit, when mounting a pseudofs, trust the
source of the fs, as it cannot be matched to backing device on the
filesystem.

Similar situation, but slightly different bug:

  # mkdir /foo /root/foo
  # echo 'foo  /foo  tmpfs  defaults  0  0' >> /etc/fstab
  # cd /root
  # mount -a
  # mount | grep /foo
  /root/foo on /foo type tmpfs (rw,relatime)

This fixes the pure libmount based mount utility, but the legacy mount
tool will still fail to handle this properly.

Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Karel Zak <kzak@redhat.com>

libmount; fix possible mem-leak

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

libmount: always trust the source of a pseudofs

Since nodev FSs have no backing block device, it's unreasonable to try
and resolve the source back to a path on the filesystem. Always trust
the source of the FS given in /etc/fstab to avoid remounting the device
when a directory of the same name exists in the current working
directory.

Example reproduction:

  # mkdir /foo
  # echo 'foo  /foo  tmpfs  defaults  0  0' >> /etc/fstab
  # mount -a
  # cd /root; mkdir foo
  # mount -a
  # mount | grep /foo
  foo on /foo type tmpfs (rw,relatime)
  /root/foo on /foo type tmpfs (rw,relatime)

[kzak@redhat.com: - use fs->flags to detect pseudo FS]

Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Karel Zak <kzak@redhat.com>

fdisk: use CDROM_GET_CAPABILITY ioctl

And replace the current archaic logic of is_ide_cdrom_or_tape().

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

mount: append inverting options for mount.<type>

 fstab:
server://foo   /mnt/foo    nfs    user,exec

The mount(8) does not append the option "exec" to /sbin/mount.<type>
helper's command line. This is no problem when executed by non-root
user as it reads the options from fstab only.

.. but when executed by root (UID=0) then the mount.<type> helper
follows the command where the "exec" option is missing. This is bug.

original version (strace output):

 execve("/sbin/mount.nfs", ["/sbin/mount.nfs",
        "sr.net.home:/mnt/store", "/mnt/store", "-o",
        "rw,nosuid,nodev,noauto,user"], [/* 21 vars */]) = 0

fixed version:

 execve("/sbin/mount.nfs", ["/sbin/mount.nfs",
        "sr.net.home:/mnt/store", "/mnt/store", "-o",
        "rw,nosuid,nodev,noauto,user,exec"], [/* 21 vars */]) = 0
                                    ^^^^^

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

lib,tt: don't trim last line char

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

losetup: cleanup headers, cmdline checks

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

losetup: cleanup man page

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

losetup: cleanup usage()

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

losetup: check  to control debug output

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

losetup: print info about one device

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

losetup: rewrite loop setup

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

losetup: rewrite -f

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

losetup: rewrite -d and -D

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

losetup: rewrite -c

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

losetup: rewrite -j

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

losetup: rewrite -a

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

lib,loopdev: add missing functions

 - improve loop_info usage (don't call ioctl more than once)
 - add functions to get devno and inode of the backing file
 - add function for compare any file with backing file by devno + inode
   or by filename

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

libblkid; fix memory leak in blkid_probe_get_wholedisk_probe()

Reported-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Karel Zak <kzak@redhat.com>

prlimit: use xalloc libs

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

losetup: move to sys-utils directory

Now losetup(8) is independent on mount/ stuff at all.

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

losetup: remove mount-specific headers

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

losetup: remove unnecessary header files

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

losetup: remove 32-bit ioctls

 ... kernel 2.4 ?! Who cares?

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

losetup: remove #ifdef LOOP_SET_FD nonsense

It's build-system task to enable/disable utils according to accessible
system features.

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

losetup: remove obsolete #ifdef MAIN

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

libmount: don't use "new" in header file

Reported-by: ben <benscott@nwlink.com>
Signed-off-by: Karel Zak <kzak@redhat.com>

tests: update fdisk tests

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: improve some messages

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: Fix bad invalid flag 0x00000 warning message

This splits check_dos_label() and dos_init() off from get_boot() and gets
rid of the invalid flag 0x00000 warning message due to a check for MBR
signs in zeroized buffer:

memset(MBRbuffer, 0, 512);

if (what == create_empty_dos)
goto got_dos_table;
[...]
got_dos_table:
if (!valid_part_table_flag(MBRbuffer)) {
[...]
if (!valid_part_table_flag(pe->sectorbuffer))
fprintf(stderr, _("Warning: invalid flag 0x%04x of partition "
[...]

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: avoid an intermediate variable in command_prompt()

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: don't use get_existing_partition() if disk label isn't dos

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: remove unused guess_device_type() function

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: get rid of get_boot(create_empty_sun) call

create_sunlabel() should create a new empty SUN disklabel without checking data
itself writes to memory and initialize internal related fdisk variables.

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: get rid of type_open global variable

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: rename try() and tryprocpt() functions

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: drop one parameter from try() function

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: move code for renaming SGI bootfile to SGI module

This moves the code for renaming SGI bootfile from command_prompt() to
sgi_set_bootfilename() function.

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: remove possibly_osf_label variable that is only assigned

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

fdisk: remove unused 'require' member in enum action

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

agetty: add static and const qualifiers

Remove also a static qualifier in main() function.

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

libblkid: minor docs changes

Signed-off-by: Francesco Cosoleto <cosoleto@gmail.com>

wipefs: use xalloc lib

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

lists: add list sorting routine

We need a list sorting function, just to mention one example user that could
benefit is the lib/tt code to sort columns.

This patch adds list_sort() which uses the Merge Sort algorithm, behaving
nicely in O(nlog(n)), heavily based on the kernel's implementation[1].  The
private data (void *priv) passed to the comparison function as been removed to
adopt a qsort(3)-like syntax, and IMHO we don't really need it anyways.

[1]: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=blob;f=lib/list_sort.c;h=d7325c6b103f0be078ff3672c35c468ed35738f1;hb=HEAD

[kzak@redhat.com: - use size_t in list_sort()]

Signed-off-by: Davidlohr Bueso <dave@gnu.org>
Signed-off-by: Karel Zak <kzak@redhat.com>

findmnt: support alternative location of fstab

Signed-off-by: Dave Reisner <dreisner@archlinux.org>
Signed-off-by: Karel Zak <kzak@redhat.com>

findmnt: use warnx for libmnt error callback

libmount doesn't set errno appropriately here:

$ findmnt -s
findmnt: /etc/fstab: parse error at line 20: Success

Signed-off-by: Dave Reisner <dreisner@archlinux.org>

setterm: order man page items alphabetically

Signed-off-by: Jari Aalto <jari.aalto@cante.net>

prlimit: return proper type

The add_prlim() function's definition currently returns a bogus structure.

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

docs: add hwclock -c to TODO file

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

partx: don't treat empty partition table as error

Phillip wrote:
 If a partition table was detected, but it had no partitions
 in it, a poorly worded error message was printed and further
 action halted.  For partx -a, this gave an unnecessary error.

Reported-by: Phillip Susi <psusi@cfl.rr.com>
Signed-off-by: Karel Zak <kzak@redhat.com>

partx: cleanup usage() and man page, add undocumented --verbose

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