Shut up valgrind complaing about LOOP_GET_STATUS64 ioctl

Signed-off-by: David Zeuthen <davidz@redhat.com>

Use g_dbus_interface_dup_object() and check return value

This fixes potential (multi-threading related) problems where the
GDBusObject is removed just before or in the middle of handling a
call. Also bump required GLib version to (unreleased) 2.31.13 since
that's the version this API was introduced in.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Do coldplug for block devices twice

Otherwise we may miss some associations between LUKS devices and other
block devices. This bug only became evident once we started sorting
block devices before processing them.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Make sure that Drive:SortKey sorts e.g. sdz before sdaa

Also make sure that we e.g. process sdb before sdad to get
Drive:SortKey set using sdb rather than sdad.

With this change we get the expected output that corresponds to kernel
probing order. Which (on good days) may correspond to slot and
enclosure order.

MODEL                     REVISION  SERIAL               DEVICE
--------------------------------------------------------------------------
HP LOGICAL VOLUME         3.00      5001438008F61790     sda
SEAGATE ST3300657SS       0006      3SJ1S7K600009051M0CE sdad sdb
SEAGATE ST3300657SS       0006      3SJ1S4MH00009052RG6Z sdae sdc
SEAGATE ST3300657SS       0006      3SJ1RFP900009051HZ5S sdaf sdd
SEAGATE ST3300657SS       0006      3SJ1S66L00009052QGZA sdag sde
SEAGATE ST3300657SS       0006      3SJ1RPFJ00009052BZ10 sdah sdf
SEAGATE ST3300657SS       0006      3SJ1RWZ100009101T12H sdai sdg
SEAGATE ST3300657SS       0006      3SJ1S7C500009052RKPP sdaj sdh
SEAGATE ST3300657SS       0006      3SJ1S40K00009101T1XU sdak sdi
SEAGATE ST3300657SS       0006      3SJ1RJS700009050VYEA sdal sdj
SEAGATE ST3300657SS       0006      3SJ1RSJC00009052MWTB sdam sdk
SEAGATE ST3300657SS       0006      3SJ1QNMQ00009052NECM sdan sdl
SEAGATE ST3300657SS       0006      3SJ1S3NE00009101TB5Y sdao sdm
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30012560      sdn
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30064749      sdo
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30039741      sdp
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30064055      sdq
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30030732      sdr
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30010853      sds
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30024753      sdt
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30065104      sdu
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30031826      sdv
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30025874      sdw
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30063724      sdx
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30064486      sdy
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30032003      sdz
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30021345      sdaa
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30067287      sdab
WDC WD1002FAEX-00Y9A0     01.01V10  WD-WCAW30014347      sdac

Signed-off-by: David Zeuthen <davidz@redhat.com>

udisksctl: sort the drives using Drive:SortKey for 'status' verb

Signed-off-by: David Zeuthen <davidz@redhat.com>

udisksctl: nuke PORT column in output of 'status' verb

Signed-off-by: David Zeuthen <davidz@redhat.com>

With /media and /run on tmfs, switch to temporary store for mounted-fs

Signed-off-by: David Zeuthen <davidz@redhat.com>

Prefer mounting in /run/user/$USER/media instead of /media

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix .gir make dependencies

Make UDisks-2.0.gir build depend on libudisks2.la. This fixes parallel builds,
as well as ensures that the gir/typelibs are updated on any code change that
affects the library.

For mount options, switch from comment=udisks to x-udisks

Recent util-linux (2.21+) supports any x-* option.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix 'make distcheck'

Signed-off-by: David Zeuthen <davidz@redhat.com>

Update POTFILES.in

Signed-off-by: David Zeuthen <davidz@redhat.com>

Use gobject-introspection Makefile

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add targets for release management

Right now it's hard-coded for me but we can change that later if
needed etc.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add Transifex URL to README file

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add Danish translation

This is just to check that the translations are actually being used
properly.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Move polkit .policy file from policy/ to data/

Signed-off-by: David Zeuthen <davidz@redhat.com>

Also update the error name in D-Bus docs

Signed-off-by: David Zeuthen <davidz@redhat.com>

Use org.freedesktop.UDisks2.* for errors, not org.freedesktop.UDisks.*

This will break apps using the D-Bus library but will not break apps
using the libudisks2 library.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix up doc page

Signed-off-by: David Zeuthen <davidz@redhat.com>

Remove copyright and license from docs and add link to online docs

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add 2012 to copyright years

Signed-off-by: David Zeuthen <davidz@redhat.com>

Rename systemd service to udisks2 and fix a couple of typos

Signed-off-by: David Zeuthen <davidz@redhat.com>

Install a systemd service file if systemd is used

This makes dbus defer starting udisks to systemd instead of doing it
itself and thus puts the daemon under proper systemd supervision.

Can be disabled by passing --without-systemdsystemunitdir to configure

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix use of memset(3)

Pointed out here https://bugs.freedesktop.org/show_bug.cgi?id=45039

Signed-off-by: David Zeuthen <davidz@redhat.com>

Ensure PATH is set

Apparently PATH may be empty if activated by recent D-Bus daemons.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Post-release version bump to version 1.91.0

Signed-off-by: David Zeuthen <davidz@redhat.com>

Update NEWS for release

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix typos in udisks(8) man page and reword a couple of things

Signed-off-by: David Zeuthen <davidz@redhat.com>

Clarify access control rules in udisks(8) man page

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix hints for PC floppy drives

Signed-off-by: David Zeuthen <davidz@redhat.com>

Don't accept a blank options string when adding/updating fstab entries

Signed-off-by: David Zeuthen <davidz@redhat.com>

Change how devices referenced in /etc/fstab and /etc/crypttab are handled

Hitherto we've required extra authorization (typically requiring the
user to authenticate as root) if a device is referenced in fstab or
crypttab and this actually limits the utility of these files -
e.g. why should the user need to authenticate as root just because
there are options in the fstab entry for the device? In fact, this
behavior is so annoying that people use e.g. the 'users' option for a
device which opens up extra security hole insofar that remote users
can now access the device.

Therefore, change the behavior so we treat fstab and crypttab devices
just like any other device, e.g. require the filesystem and
filesystem-system authorizations.

We also introduce options comment=udisks-auth [1] and x-udisks-auth
that can be used in fstab and crypttab respectively to force admin
authentication (the polkit actions in question are filesystem-fstab
encrypted-unlock-crypttab).

This has also been hooked up in GNOME's Disks
application

 http://people.freedesktop.org/~david/udisks-fstab-option-udisks-auth.png
 http://people.freedesktop.org/~david/udisks-crypttab-option-x-udisks-auth.png

Also mention this behavior in the udisks(8) man page.

Note that if a device is mounted at start-up, a user just can't
willy-nilly unmount the device - by default, he will still need to
authenticate as root in order to unmount the device (the polkit
action in question is 'filesystem-unmount-others')

[1] : this will be changed to x-udisks-auth when a future util-linux
release supporting arbitrary x-* options lands

Signed-off-by: David Zeuthen <davidz@redhat.com>

Just call it "fstab device" instead of "non-user fstab device"

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add some safe and useful ntfs-3g allowed mount options

Remove invalid assertion that all managed mounts are in /media

This is not true as we support mounting items referenced in /etc/fstab.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Don't set Drive:Ejectable to TRUE for floppy drives

Signed-off-by: David Zeuthen <davidz@redhat.com>

Adjust sort-key calculation

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix up how Drive:SortKey and Drive:TimeMediaDetected is calculated

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add work-around for kernel devices not going on media removal if mounted

Signed-off-by: David Zeuthen <davidz@redhat.com>

Use reading/writing instead of backup/restore in authentication prompts

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add convenience routines to get block devices by label and uuid

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix escaping issues when handling LABEL= and UUID= entries in fstab and crypttab

Signed-off-by: David Zeuthen <davidz@redhat.com>

Nuke debug spew

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add some more thumb drive matches

Signed-off-by: David Zeuthen <davidz@redhat.com>

Support new ID_DRIVE_THUMB udev property and 'thumb' media type

This will also set Drive:Removable to FALSE to override what the
hardware (wrongly) reports. So also add new Drive:Ejectable property
since we still want to display an Eject icon for thumb drives.

Use the icon 'media-removable' for drives of this type.

http://people.freedesktop.org/~david/gdu2-ID_DRIVE_THUMB-support.png

Signed-off-by: David Zeuthen <davidz@redhat.com>

Force ReadOnly to TRUE for optical drives

At least do this until the kernel sets the 'ro' property properly.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix loop device cleanup if backing device is not known

Signed-off-by: David Zeuthen <davidz@redhat.com>

Don't fail LoopSetup() because fstat(2) fails

With this change, this method now works with fds from a GVfs FUSE
mount (where not even uid 0 may open files on).

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add Encrypted.ChangePassphrase() method

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix up docs

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add Block.OpenForRestore() method

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add Block.OpenForBackup() method used for creating disk images

See

 http://people.freedesktop.org/~david/gdu2-create-disk-image-1.png
 http://people.freedesktop.org/~david/gdu2-create-disk-image-2.png
 http://people.freedesktop.org/~david/gdu2-create-disk-image-3.png
 http://people.freedesktop.org/~david/gdu2-create-disk-image-4.png
 http://people.freedesktop.org/~david/gdu2-create-disk-image-5.png
 http://people.freedesktop.org/~david/gdu2-create-disk-image-6.png
 http://people.freedesktop.org/~david/gdu2-create-disk-image-7.png

Signed-off-by: David Zeuthen <davidz@redhat.com>

Don't cleanup loop devices if backing device has major 0

E.g. NFS mounts.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix compile warning

Signed-off-by: David Zeuthen <davidz@redhat.com>

UDisksCleanup: Nuke error handling from user-level API

Signed-off-by: David Zeuthen <davidz@redhat.com>

Greatly simplify cleanup logic and routines

Signed-off-by: David Zeuthen <davidz@redhat.com>

Use /etc/bash_completion.d, not /etc/profile.d for bash completion script

Also use _DATA, not _SCRIPTS, to ensure that the installed file is not
executable.

Signed-off-by: David Zeuthen <davidz@redhat.com>

integration tests: small simplification

Introduce "no_options" constant for an empty options variant.

Use non-deprecated mutex and thread primitives in GLib 2.31 and later

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add new Job:Objects property and set it

This can be used to render spinners, e.g.

 http://people.freedesktop.org/~david/gdu2-spinners.png

Signed-off-by: David Zeuthen <davidz@redhat.com>

Don't distribute generated sources in tarballs

... instead, insist that these sources are regenerated at build time.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Update misc files

Signed-off-by: David Zeuthen <davidz@redhat.com>

Move udisks man page to section 8

... from section 7. This is to avoid conflicts with the udisks(7) man
page from udisks 1.x.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add shell-snippet showing how to update the ATA smart data

  gdbus call --system --dest org.freedesktop.UDisks2 \
             --object-path /org/freedesktop/UDisks2/drives/WDC_WD1002FAEX_00Y9A0_WD_WCAW30039835 \
             --method org.freedesktop.UDisks2.Drive.Ata.SmartUpdate \
             "{'atasmart_blob': <'$ATA_SMART_BLOB'>}"

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add new Block:HintPartitionable property

This can be used in the UI to determine if a "Format Disk" (to
e.g. create a partition table) button should be offered.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add ReadOnly property to the Block interface

Signed-off-by: David Zeuthen <davidz@redhat.com>

Wipe device before formatting it

It's better to not trust various mkfs tools etc. to get it right.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Actually wait for the right object when formatting a block device

Signed-off-by: David Zeuthen <davidz@redhat.com>

Allow user to manipulate partitions of a loop device if he set it up

Signed-off-by: David Zeuthen <davidz@redhat.com>

Use LO_FLAGS_PARTSCAN when setting up a loop device

Actually, use its value of 8 until 3.2 has been out long enough for us
to require it to be available and installed when compiling the code.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Pass UDisksObject so $(udisks2.device) is properly expanded

Signed-off-by: David Zeuthen <davidz@redhat.com>

Allow creating partition tables with Block.Format()

Signed-off-by: David Zeuthen <davidz@redhat.com>

Support creating dos partitions

Signed-off-by: David Zeuthen <davidz@redhat.com>

UDisksClient: add get_partitions() convenience method

Signed-off-by: David Zeuthen <davidz@redhat.com>

Don't support dos partitions of type 0x00

Signed-off-by: David Zeuthen <davidz@redhat.com>

For dos partitions of type 0, output 0x00, not 0x0

Signed-off-by: David Zeuthen <davidz@redhat.com>

Refuse to change partition type to an extended partition

Signed-off-by: David Zeuthen <davidz@redhat.com>

Introduce UDisksPartitionTypeInfoFlags

Signed-off-by: David Zeuthen <davidz@redhat.com>

When creating a partition, ensure size is a multiple of 512

Also account for parted(8) expecting the @end parameter to be
inclusive...

Signed-off-by: David Zeuthen <davidz@redhat.com>

When creating a GPT partition, don't insist that the size is divisible by 1 MiB

... only insist that the partition start is. Also take the last 33
LBAs into account - this is not space we can use.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Fix ref count bug

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add the concept of partition table subtypes

This makes it possible to clean up the UI like this

 http://people.freedesktop.org/~david/gdu2-partition-table-subtypes.png

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add PartitionTable.CreatePartition() method

Right now it only works for 'gpt' (because parted(8) expects different
formats depending on the partition table type (!)) and it doesn't yet
set the partition type. Apart from that it works great.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Synthesize 'change' uevent when deleting a partition

Otherwise the udev database is not updated...

Signed-off-by: David Zeuthen <davidz@redhat.com>

integration tests: Test loop device creation/teardown

Test in read/write and readonly modes.

integration tests: Test LUKS

Add integration test suite

This is the beginning of porting udisks1's "tests/run" integration test suite.
When running the suite from the source tree, this tests the daemon and library
from source tree; falls back on the installed system ones otherwise.

Currently covered are:

 - Creation and label handling on all supported file systems (including swap
   and empty)
 - cleanup after drive removal without unmount
 - UDisksDrive properties
 - picks up changes from command line tools like mkfs
 - correct file/dir/mountpoint permissions
 - SMART

Call the right completion method

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add a Partition.Delete() method

Signed-off-by: David Zeuthen <davidz@redhat.com>

Export a Drive.Removable property

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add Btrfs to list of known filesystems

Signed-off-by: David Zeuthen <davidz@redhat.com>

Handle the odd case when there is no partition table for a partition

This can happen when a device is ejected and it changes size to
zero. In that situation, old stale partition objects may be lying
around.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Add support for creating encrypted devices in Block.Format()

Signed-off-by: David Zeuthen <davidz@redhat.com>

Use luks-${UUID} instead of LUKS-udisks2-${UUID} to match the rest of the OS

In particular, systemd uses luks-${UUID}

Signed-off-by: David Zeuthen <davidz@redhat.com>

Support option 'take-ownership' in the Block.Format() method

Signed-off-by: David Zeuthen <davidz@redhat.com>

Rename Block.CreateFilesystem to Block.Format

... and other misc changes.

Signed-off-by: David Zeuthen <davidz@redhat.com>

Bug 41931 — Add file system creation support

For now this only supports the "label" option. Further features (LUKS, taking
ownership, etc.) will be delivered in future patches.

We still have a rather weird bug here: When udisksd calls mkfs.ext4, there is
no change uevent on the block device when it finishes, thus udev and udisks are
missing the newly created file system. This does not happen when calling
mkfs.ext4 from a shell. Work around this by triggering a change uevent after
mkfs.ext4 ourselves.

https://bugs.freedesktop.org/show_bug.cgi?id=41931

Signed-off-by: David Zeuthen <davidz@redhat.com>

Return "Unknown" for unknown partition types

Signed-off-by: David Zeuthen <davidz@redhat.com>

Partition: Add Set{Name,Type,Flags} methods and implement for dos and gpt

Using the parted(8), sfdisk(8) and sgdisk(8) commands.

Signed-off-by: David Zeuthen <davidz@redhat.com>