blkdiscard: fix underflow when offset is greater than device size
If offset (range[0]) is greater than device size (blksize), the variable 'end'
will be greater than blksize, and range[1] (length) will be recalculated.
The underflow happens when subtracting range[0] (offset) from blksize, thus
range[1] will be the result of an underflow. The bug leads to unwanted behavior
from the program, where range[1] is likely to be a high number and then will
discard a considerable amount of blocks from the device. The fix consists of
exitting the program with an error message when the condition stated above is
true. Spotted while auditing the code.
Signed-off-by: Raphael S. Carvalho <raphaelsc@cloudius-systems.com>
Helmut Grohne [Tue, 7 Oct 2014 21:07:19 +0000 (23:07 +0200)]
build-sys: check for libtoolize rather than libtool
The system libtool program has architecture dependent behaviour. It is
therefore unavailable in cross build environments. The only place it was
used in util-linux is autogen.sh to determine the availability of
libtool. All other places correctly use libtoolize or
$(top_builddir)/libtool.
Karel Zak [Tue, 7 Oct 2014 10:44:34 +0000 (12:44 +0200)]
libblkid: zeroize errno on blkid_probe_get_buffer() success
Since 37f4060225df0591ab8e1dd676dbc8115d900d4f prober functions are
sensitive to errno, it seems more robust to set errno=0 with in
blkid_probe_get_buffer() on success than set the zero on all places
where we call blkid_probe_get_buffer().
Addresses: https://github.com/karelzak/util-linux/issues/119 Signed-off-by: Karel Zak <kzak@redhat.com>
Sami Kerola [Sun, 28 Sep 2014 19:51:36 +0000 (20:51 +0100)]
hexdump: remove deprecation message
Message 'calling hexdump as od has been deprecated in favor of GNU
coreutils od' has informed the hexdump not to be used like that for three
and half years, and five releases. It is time to get rid of notice.
Karel Zak [Wed, 1 Oct 2014 09:38:49 +0000 (11:38 +0200)]
cfdisk: don't modify ncurses LINES and COLS
On Sun, Sep 28, 2014 at 01:48:38AM -0700, pv4 wrote:
If ncurses is build with --enable-reentrant, building
util-linux-2.25.1 fails with the folllowing error:
disk-utils/cfdisk.c: In function 'resize':
disk-utils/cfdisk.c:202:9: error: lvalue required as left operand of assignment
LINES = ws.ws_row;
^
disk-utils/cfdisk.c:203:8: error: lvalue required as left operand of assignment
COLS = ws.ws_col;
^
man COLS gives the following:
Depending on the configuration, these may be actual variables,
or macros (see curs_threads(3X)) which provide read-only
access to curses's state. In either case, applications should
treat them as read-only to avoid confusing the library.
Addresses: https://github.com/karelzak/util-linux/issues/125 Signed-off-by: Karel Zak <kzak@redhat.com>
An example is, in one terminal "sudo su -; echo $$", and in
another terminal, "kill -9 $PID" (the pid of the su -). It
should not print "(core dumped)", unless the kill signal
specified so, e.g. kill -7 or kill -11.
libmount: don't return root if path contains // or ending /
If the path fx. is /foo/bar/ the initial stripoff will replace the last slash
with \0 and return a pointer to that exact \0 character. The same thing will
happen if the path contains // somewhere.
This enables libmount to build on non-linux, which also
means we can build fsck on non-linux again.
(Since the context part of libmount still needs porting,
building the mount utility has instead been restricted
to only build on Linux.)
This has been build-tested on Debian GNU/kFreeBSD.
Signed-off-by: Andreas Henriksson <andreas@fatal.se>
Robert Schiele [Thu, 18 Sep 2014 06:09:59 +0000 (08:09 +0200)]
umount: read mountinfo if --detach-loop given
Don't use the optimization not to read mountinfo from commit 6a52473ecd877227f6f7da2b95da0b51593ffec1 if --detach-loop was given
since we need the name of the loop device in that case and with the
optimization this is not present and thus the detach operation
obviously fails.
Signed-off-by: Robert Schiele <rschiele@gmail.com>
Sami Kerola [Mon, 11 Aug 2014 22:04:26 +0000 (23:04 +0100)]
last: make is_phantom() when kernel config does not include audit support
When kernel CONFIG_AUDIT is not set the /proc/<pid>/loginuid information
is not present resulting live sessions to be marked 'gone - no logout' in
last(1) print out. To go-around this change makes last(1) to look
/dev/<tty> device ownership as a substitute of loginuid.
The go-around seems to work fairly well, but it has it short comings.
For example after closing a X window session the /dev/ttyN file seems to
be owned by root, not the user who had it before entering to the X
session. While that is suboptimal it is still better than an attmempt to
determine uid_t by looking owner of the /proc/<struct utmp ut_pid>, that
is a login(1) process running as root.
The issue was found using Archlinux installation.
$ pacman -Qi linux
Name : linux
Version : 3.16-2
[...]
Build Date : Mon Aug 4 18:06:51 2014
Karel Zak [Tue, 16 Sep 2014 13:30:03 +0000 (15:30 +0200)]
libmount: hide details about failed search in fstab/mtab
The current code returns -errno when not found "mount /foo" in fstab
and mtab does not exist (or /etc/mtab points to non-mounted /proc).
This is problem because the return value is too low-level and maybe
misinterpreted by top level code. It's better to always return
MNT_ERR_NOFSTAB when not found in fstab/mtab.
Reported-by: Dylan Cali <calid1984@gmail.com> Signed-off-by: Karel Zak <kzak@redhat.com>
build-sys: create parent directory for $(PATHFILES)
When building outside the source tree there is nothing to guarantee the
target directory exists before writing to it. Most of the time this just
happens to work because something else creates the directory but not
always. For example:
$ mkdir build
$ cd build
$ ../configure --disable-dependency-tracking
$ make libuuid/uuid.pc
GEN libuuid/uuid.pc
/bin/sh: line 2: libuuid/uuid.pc.tmp: No such file or directory
make: *** [libuuid/uuid.pc] Error 1
If you delete logical partition and then create a new one than fdisk
(and cfdisk) might write EBR to the first sector on the device. That's
wrong of course; because you will lost MBR (primary partitions).
Sami Kerola [Wed, 20 Aug 2014 19:06:58 +0000 (22:06 +0300)]
tests: apply travis clang PATH workaround
As seen in Travis build log the clang is not found, causing one of the
tests to fail. This change adds a workaround from Travis issue 2607 to
util-linux.
Reference: https://travis-ci.org/karelzak/util-linux/jobs/33056101#L3630
Reference: https://github.com/travis-ci/travis-ci/issues/2607 Signed-off-by: Sami Kerola <kerolasa@iki.fi>
Aurelien Jarno [Thu, 14 Aug 2014 22:26:24 +0000 (00:26 +0200)]
tests: fix fdisk/bsd for the two possible sectors/offsets
BSD disklabels depend on the endianess, but also on the architecture as
one can see in include/pt-bsd.h with the BSD_LABELSECTOR and
BSD_LABELOFFSET #define. This cause the testsuite to fail on the
affected architectures as the disk image is not the same as the expected
result.
Commit 180b3a7e tried to fix the endianess, but the special PowerPC case
has been chosen as a reference, so it still fails for example on MIPS BE
or S/390.
This patch fixes the testsuite by converting the md5sums to the expected
values, still depending on the endianess, but also for the two possible
sectors/offsets. This has been tested on Alpha, MIPS, PowerPC and x86-64,
so this should cover all 4 cases.
Cc: Andreas Henriksson <andreas@fatal.se> Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
When tailf is not using inotify, it seems to sleep for 0.25s between
checks. Just giving a 0.1s time window between update and removal of the
input file thus gives little chance for tailf to succeed.
Similar between startup and append to file, make sure initial
content is read before appending additional data by bumping the time.
This should possibly be bumped much larger to make sure that
the tailf process actually gets a chance to run at all in the
given time window. Otherwise it might fail on really slow/overloaded
machines.
The drawback would then ofcourse be to increase the time it takes
to run the testsuite.
Signed-off-by: Andreas Henriksson <andreas@fatal.se>
various: erase all traces of the long-obsolete xiafs
The xiafs filesystem was removed from the kernel fifteen years ago,
and any kernel that contained it reached end of life ten years ago.
It's time to stop mentioning it in the mount man page and elsewhere.
tests: Fall back on host/getent in ts_resolve_host
Neither dig or nslookup would be my first option for resolving,
so add the host utility of ISC/bind fame and the (slow) getent
utility shipped with libc.
projects / thirdparty / util-linux.git / log
