Karel Zak [Fri, 1 Feb 2013 14:59:58 +0000 (15:59 +0100)]
libblkid: remove optimization from verify( funrtion
Now libblkid (the cache based part) tries to probe for the cached
filesystem firstly. This optimization is broken, because:
* new another superblock could be on the device and the original
is already obsolete
* we still need to probe for partitions and raids
* the code was too fragile
The patch also suggests lsblk --fs in blkid.8 for end users. lsblk
read information from used db.
Reported-by: Andreas Hofmeister <andi@collax.com> Signed-off-by: Karel Zak <kzak@redhat.com>
Karel Zak [Wed, 12 Dec 2012 09:36:32 +0000 (10:36 +0100)]
chfn: clean up exit status
The old version in some cases (but not always) returns -1 (255) on
error. It seems better to cleanup the code and don't return internal
return codes by exit().
Karel Zak [Mon, 26 Nov 2012 13:30:22 +0000 (14:30 +0100)]
mount: sanitize paths from non-root users
$ mount /root/.ssh/../../dev/sda2
mount: only root can mount UUID=17bc65ec-4125-4e7c-8a7d-e2795064c736 on /boot
this is too promiscuous. It seems better to ignore on command line
specified paths which are not resolve-able for non-root users.
Fixed version:
$ mount /root/.ssh/../../dev/sda2
mount: /root/.ssh/../../dev/sda2: Permission denied
$ mount /dev/sda2
mount: only root can mount UUID=17bc65ec-4125-4e7c-8a7d-e2795064c736 on /boot
Note that this bug has no relation to mount(2) permissions evaluation
in suid mode. The way how non-root user specifies paths on command
line is completely irrelevant for comparison with fstab entries.
Commit c550f728f724360f99aae0fdb45b0589d9a347e0 added O_EXCL when
opening the thing to erase. This broke the wipefs utility when used
on anything which isn't an unmounted filesystem. eg. If you use it on
a block device containing partitions, then it won't work because the
kernel recognizes the partitions and so thinks the device is in use.
This change adds the --force option which, when used, undoes the above
flag change. However you still have to use --force most of the time
when erasing something that isn't a plain unmounted filesystem.
Signed-off-by: Richard W.M. Jones <rjones@redhat.com>
Toshi Kani [Tue, 23 Oct 2012 19:49:46 +0000 (13:49 -0600)]
lscpu: Fix issue found on CPU hot-remove
read_basicinfo() relies on sysfs cpu directories
"/sys/devices/system/cpu/cpu%d" with assumption that cpu
logical number %d is always sequentially assigned for all
CPUs. However, this assumption is not correct with CPU
hot-remove operation since it removes a target sysfs cpu
directory after it is ejected. As a result, lscpu may not
recognize all CPUs.
The issue can be easily reproduced on KVM or VirtualBox,
which supports CPU eject operation, as follows.
1) The system has 4 CPUs
$ lscpu -a -e
CPU NODE SOCKET CORE L1d:L1i:L2 ONLINE
0 0 0 0 0:0:0 yes
1 0 1 1 1:1:1 yes
2 0 2 2 2:2:2 yes
3 0 3 3 3:3:3 yes
3) lscpu no longer recognizes cpu3 after cpu2 is ejected
$ lscpu -a -e
CPU NODE SOCKET CORE L1d:L1i:L2 ONLINE
0 0 0 0 0:0:0 yes
1 0 1 1 1:1:1 yes
The following changes are made to address this issue.
- Use maxcpus to allocate and parse bitmaps.
- Set desc->ncpu from cpu/present, which includes both on-line
and off-line CPUs.
- Add is_cpu_present() to check if a CPU is present. Ejected
CPUs are not present.
[kzak@redhat.com: - read also /sys/devices/system/cpu/possible mask to
determine maximal number of CPUs,
- err() if possible mask is not found in /sys]
Signed-off-by: Toshi Kani <toshi.kani@hp.com> Signed-off-by: Karel Zak <kzak@redhat.com>
The reason, as Daniel it reported, was that the script will close stderr
twice, once as timing file and atexit() in function close_stdout(). This
commit fixes the problem.
Reported-by: Daniel Narvaez <dwnarvaez@gmail.com> Signed-off-by: Sami Kerola <kerolasa@iki.fi>
Karel Zak [Wed, 17 Oct 2012 09:43:39 +0000 (11:43 +0200)]
lib/loopdev: check for /sys
The current loopdev code prefers /sys to get information about
loop devices. The old methods like scan /dev are fallback solution
only. Unfortunately, the code does not check if /sys is mounted.
Addresses: http://blog.flameeyes.eu/2012/10/sophistication-can-be-bad Signed-off-by: Karel Zak <kzak@redhat.com>
libblkid/src/superblocks/befs.c: In function 'get_uuid':
libblkid/src/superblocks/befs.c:353:6: warning: dereferencing type-punned pointer will break strict-aliasing rules [-Wstrict-aliasing]
Karel Zak [Mon, 15 Oct 2012 09:10:50 +0000 (11:10 +0200)]
libmount: don't use umount optimization for -l or -f
The options -l (lazy) and -f (force) means that the mountpoint may be
unreadable (for example because NFS server is unreadable). So we
should not try to be smart in this case and we should try to minimize
number of situations when stat() or readlink() is used for the
mountpoint.
Bjørn Mork [Mon, 1 Oct 2012 23:54:49 +0000 (01:54 +0200)]
dmesg: fix console-level off-by-one
commit f06ec64f dmesg; support level names (e.g. --console-level=alert)
introduced an off-by-one error. The kernel will print messages with
a *higher* level than the console-level. The bug made it impossible to
set the level for debugging, like it is documented in e.g
Documentation/networking/netconsole.txt :
nemi:/tmp# dmesg -n 8
dmesg: unknown level '8'
And attempting to set the "emerg" level would result in an invalid 0 value:
Restoring the old behaviour for numeric levels, and mapping the level
names so that "dmesg -n debug" behaves as expected: logging everything
at level "debug" and higher.
[kzak@redhat.com: - add comment to parse_level()]
Signed-off-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: Karel Zak <kzak@redhat.com>
Karel Zak [Tue, 25 Sep 2012 14:47:18 +0000 (16:47 +0200)]
libmount: optimize mtab and utab parsing in umount
create 8000 NFS mountpoints:
#!/bin/bash
mount=/tmp/mount
if [ ! -d $mount ]; then
mkdir -p $mount
fi
for dir in {1..8000}; do
if [ ! -d $mount/$dir ]; then
mkdir -p $mount/$dir
fi
echo mount $dir
mount -t nfs 127.0.0.1:/ $mount/$dir
done
old version:
time ./umount /tmp/mount/2255
real 0m1.254s
user 0m1.002s
sys 0m0.238s
new version:
time ./umount /tmp/mount/2244
real 0m0.332s
user 0m0.111s
sys 0m0.218s
Reported-by: chenditang <chendt.fnst@cn.fujitsu.com> Signed-off-by: Karel Zak <kzak@redhat.com>
Milan Broz [Mon, 10 Sep 2012 10:58:39 +0000 (12:58 +0200)]
mkswap: wipe old signature if there is no part table by default
Historically mkswap avoids wiping any signature on "whole disk",
until force option is given.
While the idea is that it should not wipe possible boot loader,
in reality it leads to many situations where e.g. LUKS device
is overwritten but still can be detected as LUKS (but unusable).
Patch chnges behaviour that only if partition table is detected,
signatures are not wiped.
Also it removes check for block device - loop device can now
map partitions in-kernel, so using mkswap on disk image in file
should behave the same as on disk.
Also it adds warning that know signature was wiped.
Peter Rajnoha [Thu, 20 Sep 2012 11:17:13 +0000 (13:17 +0200)]
lsblk: fix processing of dm partition mappings
As blkid_devno_to_wholedisk returns parent dm device for
a partition mapping, the condition used in lsblk incorrectly
checked the parent-child relationship.
In this particular case, we need to process the dm partition
mapping like any other non-partition device as dm devices always
use proper holders/slaves sysfs hierarchy instead of
/sys/block/<parent>/<name> that is used for real partitions.
Example (test1 is a partition mapping and sdb1 is a real partition):
$ lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 128M 0 disk
`-test (dm-0) 253:0 0 128M 0 dm
`-test1 (dm-1) 253:1 0 127M 0 part
sdb 8:16 0 128M 0 disk
`-sdb1 8:17 0 127M 0 disk
Before this patch (test1 skipped!):
$ lsblk -s /dev/mapper/test1
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
test (dm-0) 253:0 0 128M 0 dm
`-sda 8:0 0 128M 0 disk
$ lsblk -s /dev/sdb1
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sdb1 8:17 0 127M 0 disk
`-sdb 8:16 0 128M 0 disk
With this patch (test1 processed correctly):
$ lsblk -s /dev/mapper/test1
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
test1 (dm-1) 253:1 0 127M 0 part
`-test (dm-0) 253:0 0 128M 0 dm
`-sda 8:0 0 128M 0 disk
$ lsblk -s /dev/sdb1
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sdb1 8:17 0 127M 0 disk
`-sdb 8:16 0 128M 0 disk
Peter Rajnoha [Thu, 20 Sep 2012 07:39:57 +0000 (09:39 +0200)]
lib/sysfs: fix sysfs_devname_to_devno for dm partitions
Partitions mapped by device-mapper are not like real partitions where
there's a /sys/block/<parent>/<name>/dev sysfs path. We need to look
at /sys/block/<name>/dev like we do for any other non-partition devices.
The mapped partition is not found otherwise.
For example, this bug shows up in lsblk while specifying a device
on command line while that device is a dm mapping over a partition:
$lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 128M 0 disk
`-test (dm-0) 253:0 0 128M 0 dm
`-test1 (dm-1) 253:1 0 127M 0 part
Before this patch:
$lsblk /dev/mapper/test1
lsblk: dm-1: unknown device name
With this patch:
$lsblk /dev/mapper/test1
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
test1 (dm-1) 253:1 0 127M 0 part
Petr Uzel [Thu, 20 Sep 2012 13:53:50 +0000 (15:53 +0200)]
setterm: make error message more informative
If `setterm -dump` fails because of lack of permission to read
/dev/vcsa, it should not report that it couldn't read /dev/vcsa0.
This could be misleading if there is only /dev/vcsa, but not /dev/vcsa0.
(Note: /dev/vcsa0 does not exist and the user does not have read
permission on /dev/vcsa in this case).
Addresses: https://bugzilla.novell.com/show_bug.cgi?id=780615 Reported-by: Christopher Yeleighton <giecrilj@stegny.2a.pl> Signed-off-by: Petr Uzel <petr.uzel@suse.cz>
Tom Gundersen [Thu, 20 Sep 2012 22:02:46 +0000 (22:02 +0000)]
hwclock: don't warp the systemtime if the RTC is in UTC
A sideeffect of 839be2ba6b44fa9dc927f081d547ebadec9de19c is that we now
warp the systemtime according to the timezone, on the first call of
--systz. This is not always the correct thing to do, and causes a
regression for us in Arch Linux.
The behavior is correct if the RTC, and hence the systemtime, is
in localtime. However, if the systemtime is already in UTC we don't
want to touch it when we set the kernel timezone (which we still need to
do as some filesystems use this information).
Karel Zak [Fri, 21 Sep 2012 10:05:12 +0000 (12:05 +0200)]
hwclock: wait time in hwclock is not long enough for some ARM devices
On Sun, Sep 16, 2012 at 02:11:48PM +0800, Queen Adam wrote:
> I'm using Archlinux ARM for my Melo A100 box.
> The hwclock always timeout when using select() in rtc.c.
>
> After I change the timeout from 5 to 10, the problem is solved.
>
> In fact the timeout in my ARM box seems only to be a little larger
> than 5s.
Karel Zak [Tue, 18 Sep 2012 09:40:44 +0000 (11:40 +0200)]
libmount: append loop option when automatically create loopdev
The loop option is optional, mount(8) is able to detect that the
source path is regular file (image) with known filesystem -- then a
loop device is automatically created. In this case we have to store
"loop" option to mtab on systems without autoclear loopdev flag.
Disable an alarm use the same logic used to enable it: first try RTC_WKALM_SET
with the "enabled" flag set to false, if it fails fall back to RTC_AIE_OFF.
Signed-off-by: Giacomo <giacomo.perale@gmail.com> Signed-off-by: Karel Zak <kzak@redhat.com>
Karel Zak [Wed, 12 Sep 2012 12:27:12 +0000 (14:27 +0200)]
libmount: don't remove user= when executed by root
The original mount(8) allows to store arbitrary user= option to mtab
file if called by root user. For example:
# mount -f foo /bar -t xxx -o rw,user=kzak
the new mount removes the 'user=' and 'users' options at all for root
user. This is regression. The original functionality is necessary by
'sshfs' where fuse writes to mtab file by mount(8).
Reported-by: Juergen Daubert <jue@jue.li> (and 'horrorStruck' on IRC) Signed-off-by: Karel Zak <kzak@redhat.com>
Karel Zak [Tue, 11 Sep 2012 07:30:16 +0000 (09:30 +0200)]
libblkid: fix copyright noticed to be consistent (so all are LGPL)
The libblkid library was intended to be released under the LGPL.
There is very little significant code in version.c; just a version
parser, and the original copyright holder (Theodore Ts'o,
tytso@mit.edu) has confirmed that he had intended to relicense all of
the files under the LGPL during a time when he was the sole author.
Acked-by: "Theodore Ts'o" <tytso@mit.edu> Reported-by: John Calcote <john.calcote@gmail.com> Signed-off-by: Karel Zak <kzak@redhat.com>
projects / thirdparty / util-linux.git / log
