Martin Pitt [Sat, 6 Nov 2010 17:40:07 +0000 (13:40 -0400)]
Fix hang on setting invalid VFAT labels
VFAT does not support several characters in file names and labels, in
particular: " * / : < > ? \ |
(http://en.wikipedia.org/wiki/File_Allocation_Table)
When using one of those, mlabel interactively asks what to do, and thus causes
a hang. Now check for invalid characters and return a proper error instead.
Also update the test case to verify that we get an Error on invalid labels, and
that valid labels still work.
Martin Pitt [Sat, 6 Nov 2010 17:28:39 +0000 (13:28 -0400)]
simplify validate_and_escape_label()
Cleanup from 3d42082
Martin Pitt [Sat, 6 Nov 2010 17:13:35 +0000 (13:13 -0400)]
Escape NTFS labels properly
Caught by recent test suite update.
Martin Pitt [Sat, 6 Nov 2010 17:09:04 +0000 (13:09 -0400)]
testsuite: use shorter label for renaming
The previous one was too long for some file systems like XFS. This produced a
correct DBusException, but it's not what we primarily want to test for.
Jiro SEKIBA [Fri, 15 Oct 2010 04:48:07 +0000 (13:48 +0900)]
udisks: escape backslash for label string
validate_and_escape_label will only escapes '"'.
However, string '\"' will be escaped to '\\"' which leads unmatched
double quote for shell string in result.
This adds escaping '\' in validate_and_escape_label to avoid the problem.
Signed-off-by: Jiro SEKIBA <jir@unicus.jp>
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Martin Pitt [Sat, 6 Nov 2010 17:00:27 +0000 (13:00 -0400)]
testsuite: Check correct handling of labels with funny characters
Try to change to a label containing backslashes and quotes. This will make
things fall over quite badly in a lot of cases right now.
Martin Pitt [Sat, 6 Nov 2010 16:42:09 +0000 (12:42 -0400)]
Fix building with Linux 2.6.37
<linux/fs.h> uses uint64_t without including it itself. Arguably a kernel bug,
but it costs us nothing to defend against it.
Marti Raudsepp [Wed, 20 Oct 2010 16:39:02 +0000 (19:39 +0300)]
udisks: Avoid deprecated polkit_authority_get() function
Use polkit_authority_get_sync() instead.
This was introduced in polkit-1 commit
322513b0d, thus bump the configure.ac
check to 0.97.
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Martin Pitt [Sun, 24 Oct 2010 13:48:51 +0000 (09:48 -0400)]
testsuite: Test forced removal of standard devices
Corresponding to the previous test for forced removal of LUKS encrypted
devices, also test handling of forced removal for plain unencrypted devices.
Martin Pitt [Sun, 24 Oct 2010 13:31:51 +0000 (15:31 +0200)]
testsuite: Check forced removal of LUKS encrypted device
Add two new methods remove_device() and readd_devices() which simulate device
removal and readdition.
Add new test case for force-removal handling of a mounted LUKS device: verify
that it cleans up properly and device can be remounted again afterwards.
This reproduces the cleanup failure in bug 30332 and verifies that the fix for
tihs (
16529b69f) works.
Martin Pitt [Thu, 21 Oct 2010 07:35:29 +0000 (09:35 +0200)]
Fix uninitialized variable in part-id
With gcc 4.5:
part-id.c: In function ‘main’:
part-id.c:186:9: error: ‘partition_number’ may be used uninitialized in this function
This only affected the error path, where we would return an arbitrary value in
get_part_table_device_file()'s out_partition_number (but return a failure
result). No big deal, but let's initialize it for quiescing gcc.
Martin Pitt [Thu, 21 Oct 2010 07:33:00 +0000 (09:33 +0200)]
partutil: Explicitly use PART_TYPE_UNKNOWN
Previously we were comparing PartitionScheme values against the undefined enum
value -1, which causes a warning with gcc 4.5. Define -1 explicitly as
"PART_TYPE_UNKNOWN" and only use that in the code, which is cleaner.
Thanks to Marti Raudsepp for pointing this out!
Jiro SEKIBA [Fri, 15 Oct 2010 04:44:22 +0000 (13:44 +0900)]
udisks: fix a bug of unquoted/unescaped label string
This patch quotes/escapes label string for nilfs-tune.
Signed-off-by: Jiro SEKIBA <jir@unicus.jp>
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Jiro SEKIBA [Thu, 14 Oct 2010 05:56:19 +0000 (14:56 +0900)]
Add nilfs2 label rename
Support renaming label of nilfs2 partition. Renaming is done by nilfs2-tune,
which is in nilfs2-util > 2.0.20.
This patch enables the label test excluded for lacking utility.
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
David Zeuthen [Fri, 1 Oct 2010 16:02:24 +0000 (12:02 -0400)]
Also export UDISKS_DM_TARGET_PARAMS for 'multipath' target
This is safe - no key material or similar is ever exported
cf. CVE-2010-1149. Additionally, this unbreaks multipath support, see
http://people.freedesktop.org/~david/multipath-restored.png
Signed-off-by: David Zeuthen <davidz@redhat.com>
Martin Pitt [Wed, 29 Sep 2010 11:50:17 +0000 (13:50 +0200)]
testsuite: Move to scsi_debug
Drop the usage of loop devices and move to scsi_debug. This provides much more
realistic behaviour (such as partitioning) and avoids using a kind of device
which udisks special-cases. scsi_debug fake devices can also be forcibly
removed (unlike loop and md), thus we can write test cases to check the cleanup
in this case.
Since scsi_debug only provides one (independent) test device, reorganize the
code a little to either just use one that one device (for most tests such as FS
or Smart), or use two partitions on that (for the LVM PVs).
Drop the "Loop" test case; it is not that interesting.
Martin Pitt [Tue, 28 Sep 2010 07:53:20 +0000 (09:53 +0200)]
Fix creating NTFS on raw devices
Similarly to the other file systems, add a -F option to mkntfs, to avoid
failing on "this is not a partition blabla".
Martin Pitt [Mon, 27 Sep 2010 18:40:22 +0000 (20:40 +0200)]
Fix mkfs.ext* arguments
Pass exactly one -F argument to mkfs.ext*, not one for each D-BUS method option
that we got (which meant none at all if there were no options). This fixes
creating ext* file systems on unpartitioned devices when supplying no option.
Martin Pitt [Mon, 27 Sep 2010 15:05:20 +0000 (17:05 +0200)]
testsuite: LVM: fix naming assumptions
The LVM tests previously assumed that LVM devices had their primary device node
in /dev/mapper/. However, under the current udev rule device names should match
the kernel name, i. e. /dev/dm-*, and /dev/mapper/ should merely contain
symlinks.
Drop the /dev/mapper/* assumptions and just check whether there is a by-id/
symlink with the LV name.
Martin Pitt [Mon, 27 Sep 2010 14:43:55 +0000 (16:43 +0200)]
testsuite: Fix SMART test for SSDs
SSDs do not have a "start-stop-count" (at least mine doesn't). Use
"power-on-hours" in the test suite instead.
Martin Pitt [Mon, 27 Sep 2010 13:20:35 +0000 (15:20 +0200)]
testsuite: fix race with SMART data
On fast machines, the Smart.test_0_hd_status progresses too fast for the SMART
data to be collected at the time when its properties are tested. Now wait for
the data to become available.
Mathieu Trudel-Lapierre [Mon, 27 Sep 2010 12:52:26 +0000 (14:52 +0200)]
Bug 30332 — udisks does not tear down LUKS devices when the slave is forcibly removed
In the event of the forced removal of a crypto device, use the luks_holder
property since it is still available to figure out which underlying cleartext
LUKS device to teardown, instead of scanning through all available devices
(because the cleartext device already has had its properties cleaned up).
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Martin Pitt [Tue, 24 Aug 2010 12:44:31 +0000 (14:44 +0200)]
profile.h: Add missing include
access() and F_OK are defined in unistd.h. Not strictly necessary for udisks
itself, but easier for stealing the file for other projects.
Martin Pitt [Sun, 15 Aug 2010 15:39:42 +0000 (17:39 +0200)]
Fix btrfs properties
Fix the maximum label size, and availability for online resizing. Thanks Chris
Ball!
Martin Pitt [Sun, 15 Aug 2010 15:34:34 +0000 (17:34 +0200)]
Bug 26258 — initial btrfs support
Add basic support for creating and checking btrfs. This does not yet support
spanning multiple devices, RAID, etc.
We disable online fsck for now. It works, but is said to be too unstable still.
Also bump the size of the loop devices in the test suite, since btrfs needs at
least 256 MB.
Jiro SEKIBA [Sun, 15 Aug 2010 15:23:39 +0000 (17:23 +0200)]
Add nilfs2 file system support
This needs util-linux-ng > 2.18.
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Matt Zimmerman [Mon, 28 Jun 2010 09:05:58 +0000 (11:05 +0200)]
Add udisks --eject option
As a companion to --mount, --unmount, and --detach, it's useful for testing and
provides a policykit-authenticated alternative to eject(1).
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Martin Pitt [Thu, 3 Jun 2010 10:09:29 +0000 (12:09 +0200)]
Bug 28075 — Do not have files executable on NTFS
Set fmask on NTFS devices to avoid having files being executable. This provides
a better default for almost all NTFS use cases under Linux, and avoids
confusing "show or execute?" questions in nautilus.
This is similar to what commit 7e7ec1 did for vfat, except that NTFS does not
have this handy "showexec" option.
Martin Pitt [Wed, 2 Jun 2010 13:56:07 +0000 (15:56 +0200)]
Fix long hangs on probing nonexistant floppy drives
A lot of modern machines do not have a floppy drive any more, but still have a
floppy controller somewhere (or at least the BIOS pretends to). Trying to
open(/dev/fd0) on these machines causes long hangs, which lead to long desktop
startup times.
To fix this, avoid probing floppies for media in update_info().
https://launchpad.net/bugs/539515
David Zeuthen [Tue, 1 Jun 2010 21:22:38 +0000 (17:22 -0400)]
Avoid using entities in introspection XML
Unfortunately GMarkup doesn't understand entities so
gdbus-binding-tool will not work with these XML files without this
patch.
David Zeuthen [Tue, 25 May 2010 14:59:01 +0000 (10:59 -0400)]
Don't probe non-data discs for partition tables
Suggested by Kay Sievers. This is related to
https://bugzilla.kernel.org/show_bug.cgi?id=15757
Signed-off-by: David Zeuthen <davidz@redhat.com>
Martin Pitt [Thu, 20 May 2010 20:12:44 +0000 (22:12 +0200)]
testsuite: mkfs() code cleanup
Replace the long if ladder for mkfs special cases with option dictionaries.
This makes it easier to extend in the future and the code easier to read.
Martin Pitt [Wed, 12 May 2010 12:00:18 +0000 (14:00 +0200)]
Bug 28075 — Do not have all files executable on vfat
Enable the "showexec" vfat mount option, to avoid data files being executable
(which causes confusing question dialogs in nautilus which only have one
sensible answer).
See http://www.kernel.org/doc/Documentation/filesystems/vfat.txt (search for
"showexec") for the details of this option.
David Zeuthen [Wed, 19 May 2010 17:01:31 +0000 (13:01 -0400)]
Update to latest LVM2 API
See
http://lists.freedesktop.org/archives/devkit-devel/2010-April/000783.html
Martin Pitt [Wed, 12 May 2010 11:00:16 +0000 (13:00 +0200)]
testsuite: Check file permissions
Add a check to the filesystem test cases for sane permissions of data files and
executables. In particular, this ensures that the executable bit is set where
it makes sense (directories and programs), but not on mere data files.
We currently ignore known-broken file systems like vfat and ntfs.
This is the groundwork to fix this behaviour at least on vfat. Its current
behaviour is quite unnerving, since trying to open a file on them in e. g.
GNOME will always ask for "do you want to open or run this file".
David Zeuthen [Fri, 9 Apr 2010 15:56:27 +0000 (11:56 -0400)]
Post-release version bump to 1.0.2
David Zeuthen [Fri, 9 Apr 2010 15:45:44 +0000 (11:45 -0400)]
Update NEWS for release
Martin Pitt [Fri, 9 Apr 2010 11:19:44 +0000 (13:19 +0200)]
add information about CVE-2010-1149 to NEWS
Martin Pitt [Wed, 7 Apr 2010 07:50:19 +0000 (09:50 +0200)]
Hide Sony E-Book launcher partition
It only contains Windows software and is read-only, and thus useless.
https://launchpad.net/bugs/546924
Martin Pitt [Wed, 7 Apr 2010 07:27:18 +0000 (09:27 +0200)]
job-drive-benchmark.c: Fix data types in error messages
We previously used wrongly sized data types in the error messages, which causes
crashes on platforms where they do not coincidentally match in size.
https://launchpad.net/bugs/527202
Martin Pitt [Wed, 7 Apr 2010 06:56:05 +0000 (08:56 +0200)]
part-id: Fix DM partition table detection
If DM_NAME does not end in 'pN' (i. e. is not a partition), do not treat it as
malformed, but just go ahead probing it. This fixes partition table detection
for kpartx'ed DM devices.
Re-enable the test case which originally uncovered this bug.
Martin Pitt [Wed, 7 Apr 2010 06:36:42 +0000 (08:36 +0200)]
testsuite: Check presence and properties of loop devices
Martin Pitt [Wed, 7 Apr 2010 06:12:42 +0000 (08:12 +0200)]
Bug 27494 — publicly exports dm key information
Change udisks-dm-export to only export UDISKS_DM_TARGETS_PARAMS for "linear"
types. It is the only one we care about for now and know how to interpret.
"crypto" types have information about the encryption key in the target
parameters, which we must not leak.
Also add appropriate comments to the two places which currently evaluate
UDISKS_DM_TARGETS_PARAMS.
Martin Pitt [Tue, 6 Apr 2010 21:01:38 +0000 (23:01 +0200)]
testsuite: Test detection of kpartx LVM partitions
Add a new LVM test case where we create two partitions of the LV and check
their udisks properties.
This is primarily meant to test the rather involved logic of udisks-part-id to
determine which device that partition is a slave of (it involves parsing
UDISKS_DM_TARGETS_PARAMS and various DM_* properties).
This also apparently uncovered a bug: the partition table does not have correct
properties in udisks. Comment out the two checks for now, to be investigated
later.
Martin Pitt [Tue, 6 Apr 2010 18:34:36 +0000 (20:34 +0200)]
testsuite: Check that our udev probers do not leak key information
This reproduces https://bugs.freedesktop.org/show_bug.cgi?id=27494
Martin Pitt [Fri, 19 Mar 2010 09:09:59 +0000 (10:09 +0100)]
Fix exit code of umount.udisks
Exit with return code 0 if the unmount succeeded.
https://launchpad.net/bugs/541740
Martin Pitt [Thu, 18 Mar 2010 13:42:24 +0000 (14:42 +0100)]
Allow other rules to set a more specific presentation icon
If previous rules already set UDISKS_PRESENTATION_ICON_NAME (like in a future
media-player-info, to attach more specific icons to particular devices), do not
overwrite it with our generic icons; just set them if we do not have an icon at
all yet.
Also put the icon rules into their own visual section now, they were previously
in the middle of the "card reader type" rules.
Martin Pitt [Wed, 17 Mar 2010 14:46:03 +0000 (15:46 +0100)]
Set multimedia-player-ipod icon for iPod media players
Part of https://launchpad.net/bugs/540235
David Zeuthen [Mon, 15 Mar 2010 23:42:29 +0000 (19:42 -0400)]
Update NEWS for release
Ooops, should have done this before 1.0.0 - this will have to do.
David Zeuthen [Mon, 15 Mar 2010 23:38:42 +0000 (19:38 -0400)]
Post-release version bump to 1.0.1
Signed-off-by: David Zeuthen <davidz@redhat.com>
David Zeuthen [Mon, 15 Mar 2010 22:37:21 +0000 (18:37 -0400)]
Add TODO item about UTF-8 and D-Bus properties
David Zeuthen [Mon, 15 Mar 2010 22:31:58 +0000 (18:31 -0400)]
Guarantee that DeviceFilePresentation always contain something meaningful
This makes life easier for UI tools - they can always just display
:DeviceFilePresentation instead of having to check whether it's blank
and if so use :DeviceFile.
Signed-off-by: David Zeuthen <davidz@redhat.com>
David Zeuthen [Mon, 15 Mar 2010 22:22:48 +0000 (18:22 -0400)]
Make it possible to disable LVM2 and dm-multipath support
Evidently, the device-mapper and LVM2 side of this needs some more
work so certain enterprise distros may want to disable this for now.
Signed-off-by: David Zeuthen <davidz@redhat.com>
David Zeuthen [Mon, 15 Mar 2010 19:38:53 +0000 (15:38 -0400)]
Update TODO
Signed-off-by: David Zeuthen <davidz@redhat.com>
David Zeuthen [Mon, 15 Mar 2010 18:04:59 +0000 (14:04 -0400)]
Fix partition table probing during coldplug
Commit
8aa58f68ae3b53f6ba6d8c51b4f66bc8135b4123 introduced a subtle
assumption that the device file given on the command line for
udisks-part-id is the same as what udev_device_get_devnode() returns
for the udev_device object. This is actually not true when handling an
event as the node passed is a temporary one.
Signed-off-by: David Zeuthen <davidz@redhat.com>
Martin Pitt [Mon, 15 Mar 2010 07:41:32 +0000 (08:41 +0100)]
udisks-part-id: Simplify code
In 8aa58f we changed part-id to keep one singe udev_device around. Simplify
get_part_table_device_file() to not get the syspath from our device, just to
get a new udev device from that syspath again (which is exactly the same as the
device that we already have).
This also fixes a double udev_device_unref(), although this seems to have been
harmless.
Martin Pitt [Sun, 14 Mar 2010 19:58:28 +0000 (20:58 +0100)]
Bug 27072 — does not force-unmount for drives in fstab
force_removal() previously only checked for entries in our own mtab.
This caused forced unmounts for drives in /etc/fstab not to work.
Now check for drives in /etc/fstab as well.
Martin Pitt [Fri, 12 Mar 2010 19:07:24 +0000 (20:07 +0100)]
TODO update for property update race conditions
Martin Pitt [Fri, 12 Mar 2010 18:56:34 +0000 (19:56 +0100)]
testsuite: Add option to disable workarounds for race conditions
Some D-BUS API methods cause properties to not be up to date yet when a method
call finishes, thus we do an udevadm settle as a workaround in the test suite.
Those bugs should eventually get fixed properly, but it's unnerving to have the
tests fail on them when you are working on something else.
Split the udevadm settles into the existing .sync() for legitimate syncs, and a
new sync_workaround() for race condition workarounds.
Add --no-workarounds CLI option to disable those workaround syncs, so that
these race conditions can be detected and fixed.
Martin Pitt [Fri, 12 Mar 2010 18:14:24 +0000 (19:14 +0100)]
testsuite: Use loop devices instead of ram disks
On many platforms, the default size of RAM disks is too small for some file
systems or LVM devices. Rewrite using temporary loop devices.
Martin Pitt [Fri, 12 Mar 2010 17:30:23 +0000 (18:30 +0100)]
udisks-part-id: Check ID_FS_TYPE from environment first
We need to do that for the initial udev run where blkid and udisks-part-id are
called in the same round; the database is not updated at that time yet.
Martin Pitt [Fri, 12 Mar 2010 16:42:25 +0000 (17:42 +0100)]
Bug 26996 — wrongly detects partition on unpartitioned VFAT device
A standard VFAT header matches the signature of an MBR, so we detected VFAT
file systems on raw unpartitioned devices as both VFAT and MBR partition table.
This confuses gnome-disk-utility/gvfs, and causes them to be ignored in GNOME.
The standard approach in the kernel/blkid is to only detect an MBR if it does
not have a valid VFAT signature. Since blkid already does that for us, do not
output PARTITION_TABLE properties in udisks-part-id if we already know that we
have a valid ID_FS_TYPE.
Martin Pitt [Fri, 12 Mar 2010 16:27:03 +0000 (17:27 +0100)]
udisks-part-id: Shuffle code to keep udev_device in main()
Reorder the code to get and keep an udev_device struct in main() and pass it
down to get_part_table_device_file(), instead of temporarily creating them in
that function and get_syspath(). This avoids duplicate lookups and also
allows us to do further queries on the udev_device later on (which we need to
fix bz#26996)
Martin Pitt [Fri, 12 Mar 2010 15:15:52 +0000 (16:15 +0100)]
testsuite: Check partitions with filesystems
So far we only checked handling of partitions without file systems, which did
not detect the regression from commit cfc6698. Now test both combinations.
Note that we have to jump through some hoops here, since
Partition{Create,Modify} internally writes the new partition type/flags/etc,
and thus does not care about a broken part-id. Thus we trigger a change event
on block devices to actually check what the udev probers do.
Michael Biebl [Tue, 17 Nov 2009 03:18:14 +0000 (21:18 -0600)]
Bug 25424 — Update to stable API of udev/gudev/dbus
Remove G_UDEV_API_IS_SUBJECT_TO_CHANGE and
LIBUDEV_I_KNOW_THE_API_IS_SUBJECT_TO_CHANGE.
Bump the build dependency on libudev to >= 143 and gudev-1.0 to >= 147
when those APIs were declared stable.
We also depend on dbus >= 1.0, so remove DBUS_API_SUBJECT_TO_CHANGE as
well.
Signed-off-by: David Zeuthen <davidz@redhat.com>
tuxce [Fri, 12 Mar 2010 12:12:41 +0000 (13:12 +0100)]
Bug 25012 Add "usefree" to vfat available options
When mounting a vfat volume, system scan the entire partition to know free
space available, using "usefree" option avoid this scan and use the free
clusters info on file system.
Modern kernels should not really need this any more, even on a 500 GB vfat
external disk it only takes two or three seconds to probe the file system.
However, this regressed in the kernel already twice in the past, so merely
allowing the mount option (without making it the default) provides a workaround
for such cases.
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Martin Pitt [Fri, 12 Mar 2010 12:10:06 +0000 (13:10 +0100)]
testsuite: Get along with running gvfs/nautilus
Send a SIGSTOP to all gvfs-gdu-volume-monitor processes while the test suite is
running, to avoid nautilus windows popping up by the dozens, and causing "busy"
failures.
Martin Pitt [Fri, 12 Mar 2010 12:06:51 +0000 (13:06 +0100)]
testsuite: Add TODO item to test new support for loop devices
David Zeuthen [Thu, 11 Mar 2010 21:38:32 +0000 (16:38 -0500)]
Handle loopback devices
http://people.freedesktop.org/~david/udisks-loop-devices.png
it even works with kpartx partitions
http://people.freedesktop.org/~david/palimpsest-loop-kpartx-partition.png
In order for this to work well, you will this patch
http://people.freedesktop.org/~david/0001-Generate-change-uevent-for-loop-device.patch
which is also needed in order to make the /dev/disk hierarchy work
well. In lieu of this patch, you can generate the 'change' uevent
yourself. I've submitted this patch upstream through Kay Sievers.
David Zeuthen [Thu, 11 Mar 2010 19:21:02 +0000 (14:21 -0500)]
Revert "Bug 26996 — wrongly detects partition on unpartitioned VFAT device"
This commit actually broke partition table parsing for the partitions
itself, e.g. this output
UDISKS_PARTITION=1
UDISKS_PARTITION_SCHEME=mbr
UDISKS_PARTITION_NUMBER=1
UDISKS_PARTITION_TYPE=0x83
UDISKS_PARTITION_SIZE=
51202335744
UDISKS_PARTITION_LABEL=
UDISKS_PARTITION_UUID=
UDISKS_PARTITION_FLAGS=boot
UDISKS_PARTITION_SLAVE=/sys/devices/pci0000:00/0000:00:1f.2/host0/target0:0:0/0:0:0:0/block/sda
UDISKS_PARTITION_OFFSET=32256
UDISKS_PARTITION_ALIGNMENT_OFFSET=0
if the partition in question had a recognizable filesystem.
This reverts commit
cfc669846a53815517d32d0b7b96cbc4935cea7a.
David Zeuthen [Thu, 11 Mar 2010 15:55:05 +0000 (10:55 -0500)]
Update TODO
Martin Pitt [Thu, 11 Mar 2010 06:58:34 +0000 (07:58 +0100)]
Bug 26996 — wrongly detects partition on unpartitioned VFAT device
A standard VFAT header matches the signature of an MBR, so we detected
VFAT file systems on raw unpartitioned devices as both VFAT and MBR partition
table. This confuses gnome-disk-utility/gvfs, and causes them to be ignored in
GNOME.
The standard approach in the kernel/blkid is to only detect an MBR if it does
not have a valid VFAT signature. Since blkid already does that for us, just
skip the udisks-part-id invocation completely if we already know that we have a
valid ID_FS_TYPE.
Martin Pitt [Wed, 10 Mar 2010 16:45:38 +0000 (17:45 +0100)]
testsuite: Ensure that volumes on raw partitions don't have a partition table
This reproduces bz#26996.
Michael Biebl [Tue, 9 Mar 2010 13:12:20 +0000 (14:12 +0100)]
Drop obsolete rule for suspended LVM devices
udisks-dm-export does not set UDISKS_DM_{STATE,NAME} any more, and recent LVM
versions check and flag private devices themselves, so just drop our own rules
for that.
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Michael Biebl [Tue, 9 Mar 2010 12:59:40 +0000 (13:59 +0100)]
Drop obsolete rules for private LVM devices
udisks-dm-export does not set UDISKS_DM_{UUID,NAME} any more, and recent LVM
versions check and flag private devices themselves, so just drop our own rules
for that.
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Michael Biebl [Tue, 9 Mar 2010 12:55:54 +0000 (13:55 +0100)]
Respect DM_UDEV_DISABLE_DISK_RULES_FLAG from current LVM
Michael Biebl [Fri, 5 Mar 2010 07:23:30 +0000 (08:23 +0100)]
Include udisks-tcp-bridge.xml in dist tarball
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Michael Biebl [Fri, 5 Mar 2010 07:19:54 +0000 (08:19 +0100)]
Include profile.h in dist tarball
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Martin Pitt [Wed, 3 Mar 2010 21:48:50 +0000 (22:48 +0100)]
Update .gitignore files
Martin Pitt [Wed, 3 Mar 2010 21:23:47 +0000 (22:23 +0100)]
fix compiler warnings
Fix a couple of unchecked function call results, and a strict pointer aliasing
violation, which resulted in compiler warnings.
Martin Pitt [Wed, 3 Mar 2010 21:06:09 +0000 (22:06 +0100)]
in maintainer mode, build with -Werror
With --enable-maintainer-mode, or with --enable-strict, build with -Werror.
David Zeuthen [Wed, 3 Mar 2010 21:12:47 +0000 (16:12 -0500)]
Initialize variable before use
Martin Pitt [Wed, 3 Mar 2010 20:20:00 +0000 (21:20 +0100)]
Bug 26725 — add profiling
Add a PROFILE() macro (src/profile.h) which, when building with
--enable-profiling, adds fake access() calls as trace points for startup speed
profiling.
http://people.gnome.org/~federico/news-2006-03.html#login-time-2 describes how
to use strace and turn its output into a nice graph:
# strace -tttfo /tmp/trace src/udisks-daemon
[...]
$ plot-timeline.py -o /tmp/trace.png /tmp/trace
Instrument the most important/complex methods in daemon.c and device.c, and
main().
Daniele Napolitano [Tue, 2 Mar 2010 07:18:47 +0000 (08:18 +0100)]
Bug 26822 — Fix benchmarking for large files
Use lseek64() instead of lseek() in job-drive-benchmark.c (as in the rest of
the code), to also work with large devices.
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
Martin Pitt [Mon, 1 Mar 2010 12:36:49 +0000 (13:36 +0100)]
fix media detection of CD-ROMs
Use ID_CDROM_MEDIA property for media detection in update_info(), not
ID_CDROM_MEDIA_STATE. The latter is not set by udev's cdrom_id for pure CD-ROM
media (since it's sometimes wrong), and we are already using ID_CDROM_MEDIA in
our udev rules, so let's be consistent here.
This was exposed by commit c8d02e0, which previously shadowed this bug.
David Zeuthen [Sat, 27 Feb 2010 16:59:56 +0000 (11:59 -0500)]
Include stdout in mkfs error output
This is to help expose this bug
mke2fs insists on user interaction even if stdin is not a tty and -F is passed
https://bugzilla.redhat.com/show_bug.cgi?id=569021
David Zeuthen [Thu, 25 Feb 2010 22:19:51 +0000 (17:19 -0500)]
Bug 24424 - Move UDISKS_MEDIA_AVAILABLE out of the part table prober
This change (magically) makes floppy disks work in GNOME. Woot.
David Zeuthen [Thu, 25 Feb 2010 17:28:40 +0000 (12:28 -0500)]
We're using the UDISKS_ prefix, not DKD_
Also pointed out in
http://bugs.freedesktop.org/show_bug.cgi?id=26720
David Zeuthen [Wed, 24 Feb 2010 16:36:21 +0000 (11:36 -0500)]
Add a --disable-remote-access option
Certain Enterprise Linux vendors might not want to support the udisks
1.0.x ABI for seven years.
David Zeuthen [Tue, 23 Feb 2010 21:48:33 +0000 (16:48 -0500)]
Update detach helper for current sysfs interface
Writing 'suspend' to power/level doesn't seem to work these days
(worked in F12 but broken in F13). Also use the new remove/ sysfs
attribute.
David Zeuthen [Tue, 23 Feb 2010 20:56:12 +0000 (15:56 -0500)]
Improve debug output and support for the detach helper
This allows easily invoking the tool from a shell, like this
$ sudo /usr/libexec/udisks-helper-drive-detach /dev/sdb
Detaching device: /dev/sdb
SYNCHRONIZE CACHE: FAILED: No such file or directory
(Continuing despite SYNCHRONIZE CACHE failure.)
STOP UNIT: OK
Unbind USB interface driver: OK
Suspend USB Device: OK
$ echo $?
0
David Zeuthen [Tue, 23 Feb 2010 17:31:07 +0000 (12:31 -0500)]
Use DM_NAME instead of DM_UUID when determining partition number
Gah, looks like DM_UUID is not always set for partitions on multipath
devices. Let's hope it's the case for DM_NAME.
Martin Pitt [Wed, 17 Feb 2010 11:29:37 +0000 (12:29 +0100)]
testsuite: Fix LVM tests
Turns out that current LVM versions/rules do create by-id symlinks for RAID
mirror images, on purpose. So do not fail on them, just check that they do not
have a real IdUsage/IdUuid.
David Zeuthen [Tue, 16 Feb 2010 19:24:27 +0000 (14:24 -0500)]
Update for new liblvm2app library
Several things changed in liblvm2app
- sizes are now reported in bytes instead of sectors
- a couple of new functions were added
You will need lvm2 version 2.02.61 or later (on Fedora Rawhide, this
is available as lvm2-2.02.61-1.fc13.x86_64). Unfortunately the liblvm2
pc version number is stuck at 2.1 so we can't require this version in
configure.in. I've asked the LVM team to fix that.
David Zeuthen [Mon, 15 Feb 2010 23:28:42 +0000 (18:28 -0500)]
Add doc/udisks-overrides.txt so 'make dist' works
Hmm, need to investigate later why this is needed.
David Zeuthen [Mon, 15 Feb 2010 22:51:15 +0000 (17:51 -0500)]
Fix compiler warning
David Zeuthen [Mon, 15 Feb 2010 20:15:42 +0000 (15:15 -0500)]
Export alignment offset for a partition
David Zeuthen [Mon, 15 Feb 2010 19:53:54 +0000 (14:53 -0500)]
Rework partition table probing
In particular, always respect UDISKS_PARTITION_* and
UDISKS_PARTITION_TABLE_* udev properties.
Also, temporarily, use g_getenv() to retrieve properties in
udisks-part-id. It seems like libudev doesn't return properties
correctly in some cases involving dm-multipath. Will need to
investigate.
Martin Pitt [Mon, 15 Feb 2010 16:02:30 +0000 (17:02 +0100)]
Add support for "disable polling" udev property
Do not poll a device which has UDISKS_DISABLE_POLLING set. Some known-broken
CD-ROM drives cause a lot of CPU activity when being polled, so allow admins to
suppress polling with a local udev rule like
SUBSYSTEM=="block", ENV{ID_VENDOR}=="Optiarc*", ENV{ID_MODEL}=="*AD-7640S*", \
ENV{UDISKS_DISABLE_POLLING}="1"
https://bugs.freedesktop.org/show_bug.cgi?id=26508
Michael Biebl [Fri, 12 Feb 2010 11:12:27 +0000 (12:12 +0100)]
Make sure udisks-helper-mdadm-expand is included in the dist tarball
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>