In the old code QByteArray fstabContents was actually empty.
Also, writeData function was opening file in append mode,
thus nothing was actually written.
Split writeData function into two:
* one for block devices
* another for writing fstab file
BUG: 417205
The output of `sfdisk --json /dev/sdb` is not necessarily
valid JSON. Then, no partition information is stored,
no first-valid-lba in particular. This leads to new partitions
being made from sector 0, which is invalid on a GPT table.
The workaround is to manually fix the known-broken JSON
from sfdisk. This is amply documented in a standalone
static function.
FIXES#425097
The GPT type and attributes can be set since the commits 0529ebf (Add
support for setting the specific GPT type) and 0ffec31 (Add new job to
set the GPT partition attributes).
But these two data from existing partitions are not read and are missing
though.
This reads the GPT type and attributes data at scanning from the json
output, after the GPT name/label and uuid are read.
The GPT partition layout supports partition attributes.
The CLI sfdisk sets the partition attributes using the option
--part-attrs. See the examples below:
$ cat <<EOF | sfdisk disk.img
label: gpt
type=C12A7328-F81F-11D2-BA4B-00A0C93EC93B, size=64M
type=0FC63DAF-8483-4772-8E79-3D69D8477DE4
EOF
(...)
$ sfdisk --part-attrs disk.img 1 RequiredPartition,NoBlockIOProtocol,LegacyBIOSBootable,48
(...)
$ sfdisk --part-attrs disk.img 2 60,61,62,63
(...)
$ sfdisk --dump disk.img
(...)
disk.img1 : start= 2048, size= 131072, type=C12A7328-F81F-11D2-BA4B-00A0C93EC93B, uuid=BBE806BB-8567-A843-9FF5-9B6B35D2908E, attrs="RequiredPartition NoBlockIOProtocol LegacyBIOSBootable GUID:48"
disk.img2 : start= 133120, size= 1963999, type=0FC63DAF-8483-4772-8E79-3D69D8477DE4, uuid=7DB39F08-C138-664B-B38E-ED2DEB549AA6, attrs="GUID:60,61,62,63"
This commit introduces the new job set-partition-attributes that is used
in the new-operation to set the attributes of the partition. The job
uses the newly introduced method setPartitionAttributes that is
implemented by the sfdisk and dummy backends.
Note: This is a copypaste of what was done for GPT partition label in
commit f585f6c (Add new job to set the GPT partition label) and GPT
partition UUID in commit 1dde035 (Add new job to set the GPT partition
UUID).
Note: RequiredPartition, NoBlockIOProtocol, LegacyBIOSBootable are
key words for 0, 1 and 2.
The GPT UUID are generated by the backends automatically once a
partition is created.
This reads the UUID that was generated after the creation of a partition
and stores it back to the object Partition.
The GPT partition layout supports unique GUID partitions.
The CLI sfdisk sets the partition UUID using the option --part-uuid.
See the examples below:
$ cat <<EOF | sfdisk disk.img
label: gpt
type=C12A7328-F81F-11D2-BA4B-00A0C93EC93B, size=64M
type=0FC63DAF-8483-4772-8E79-3D69D8477DE4
EOF
(...)
$ sfdisk --dump disk.img
(...)
disk.img1 : start= 2048, size= 131072, type=C12A7328-F81F-11D2-BA4B-00A0C93EC93B, uuid=BBE806BB-8567-A843-9FF5-9B6B35D2908E
disk.img2 : start= 133120, size= 1963999, type=0FC63DAF-8483-4772-8E79-3D69D8477DE4, uuid=7DB39F08-C138-664B-B38E-ED2DEB549AA6
$ sfdisk --part-uuid disk.img 1 00000000^C000-0000-0000-000000000000
(...)
$ sfdisk --part-uuid disk.img 2 11111111-1111-1111-1111-111111111111
(...)
$ sfdisk --dump disk.img
(...)
disk.img1 : start= 2048, size= 131072, type=C12A7328-F81F-11D2-BA4B-00A0C93EC93B, uuid=00000000-0000-0000-0000-000000000000
disk.img2 : start= 133120, size= 1963999, type=0FC63DAF-8483-4772-8E79-3D69D8477DE4, uuid=11111111-1111-1111-1111-111111111111
This commit introduces the new job set-partition-uuid that is used in
the new-operation to set the UUID of the partition. The job uses the
newly introduced method setPartitionUUID that is implemented by the
sfdisk and dummy backends.
Note: This is a copypaste of what was done for GTP partition label in
commit f585f6c (Add new job to set the GPT partition label).
Summary: Add Q_UNUSED() to mark the unused parameters
Test Plan: Builds with fewer warnings on FreeBSD
Differential Revision: https://phabricator.kde.org/D24155
The GPT partition layout supports naming partitions.
The support for the partition label was added in the backend libparted
since commit 28fa6ac (Add support for GTP partition labels).
The libparted was removed later by commit 8fa1814 (Remove libparted
backend) and no backend sets the partition label.
The CLI sfdisk sets the partition label using the option --part-label.
See the examples below:
$ cat <<EOF | sfdisk disk.img
label: gpt
type=C12A7328-F81F-11D2-BA4B-00A0C93EC93B, size=64M
type=0FC63DAF-8483-4772-8E79-3D69D8477DE4
EOF
(...)
$ sfdisk --part-label disk.img 1 efi
(...)
$ sfdisk --part-label disk.img 2 rootfs
(...)
$ sfdisk --dump disk.img
(...)
disk.img1 : start= 2048, size= 131072, type=C12A7328-F81F-11D2-BA4B-00A0C93EC93B, uuid=D08E5B2A-4649-9F4A-AEA3-6C3048888EAA, name="efi"
disk.img2 : start= 133120, size= 1963999, type=0FC63DAF-8483-4772-8E79-3D69D8477DE4, uuid=0BEEFE82-19EA-DC4C-BB6A-27B6DA0C3BD2, name="rootfs"
This commit introduces the new job set-partition-label that is used in
the new-operation to set the label of the partition. The job uses the
newly introduced method setPartitionLabel that is implemented by the
sfdisk and dummy backends.
The object QVariant is a native object in the Qt world that represents
any type of value: boolean, integer, string, floating point, date,
UUID...
This change updates the API to use the object QVariant and provide a
more flexible API to represent the filesystem features.
The GPT partition layout defines the type of the partition using a
UUID[1]. The type of the partition is not related to the underlying
filesystem but to its functionality: rootfs, home, var, swap, esp...
The UUID 4f68bce3-e8cd-4db1-96e7-fbcaf984b709 defines the root partition
(/), the UUID 933ac7e1-2eb4-4f13-b844-0e14e2aef915 defines the home
partition (/home), the UUID 4d21b016-b534-45c2-a9fb-5c16e091fd2d defines
the variable partition (/var)... Also, the UUIDs may vary depending on
the architecture (x86, x86_64, arm or aarch64).
KPMcore guesses the type depending on the underlying filesystem.
This commit enables the setting of a more specific GPT type if it is
specified through the method Partition::setType(). It falls back to the
previous guess type if the type is left unset.
[1]: https://systemd.io/DISCOVERABLE_PARTITIONS/
In some cases, it may be necessary to create a filesystem with specific features
enabled/disabled.
This PR makes it possible to create filesystems this way. It does so
by introducing a new m_Features member variable and the createWithFeatures()
function to the FileSystem base class. The latter function is implemented
for the btrfs, ext2/3/4 and FAT filesystems.
Additionnally, the CreateFileSystemJob has been modified to enable
creating/formatting filesystems with specific features enabled.
Differential Revision: https://phabricator.kde.org/D21903
CCBUG: 342178
Summary: BUG: 411226
Test Plan: I couldn't test it since my device does not support SMART
Reviewers: stikonas
Reviewed By: stikonas
Subscribers: #kde_partition_manager
Tags: #kde_partition_manager
Differential Revision: https://phabricator.kde.org/D23572
Summary: BUG: 411226
Test Plan: I couldn't test it since my device does not support SMART
Reviewers: stikonas
Reviewed By: stikonas
Subscribers: #kde_partition_manager
Tags: #kde_partition_manager
Differential Revision: https://phabricator.kde.org/D23572
Summary:
kauth helpers have a qcoreapplication started which auto-terminates after
10 seconds of idle time. this would be an added time out on top of our
life time management that is not necessary. to prevent this additional
delay simply quit the qapplication directly.
this makes helper termination almost instantaneous even when the dbus
client disappears due to crash or SIGINT, and largely removes the
need to call stopHelper altogether.
BUG: 408245
Test Plan:
both partitionmanager and calamares instantly quit the helper no matter how the client process gets terminated
other than that everything seems to work as before
Reviewers: stikonas
Differential Revision: https://phabricator.kde.org/D21559
Summary:
kauth helpers have a qcoreapplication started which auto-terminates after
10 seconds of idle time. this would be an added time out on top of our
life time management that is not necessary. to prevent this additional
delay simply quit the qapplication directly.
this makes helper termination almost instantaneous even when the dbus
client disappears due to crash or SIGINT, and largely removes the
need to call stopHelper altogether.
BUG: 408245
Test Plan:
both partitionmanager and calamares instantly quit the helper no matter how the client process gets terminated
other than that everything seems to work as before
Reviewers: stikonas
Differential Revision: https://phabricator.kde.org/D21559
Summary:
smartmontools 7.0 reports user capacity as an object that contains bytes and blocks value, instead of being an int that just contains bytes
This should fix incorrect calculation of bad blocks threshold and incorrect display of "Overall" status
Test Plan:
run partitionmanager on a system that has smartmontools 7, open smart status
overall status should be reported correctly
Reviewers: stikonas
Reviewed By: stikonas
Differential Revision: https://phabricator.kde.org/D21311
Summary: QDbus already provides enough security to the calls made by the application to the helper. Hence no need to encrypt, sign the requests of the application and verify their integrity at the helper side.
Reviewers: stikonas, cjlcarvalho
Reviewed By: stikonas
Subscribers: #kde_partition_manager
Tags: #kde_partition_manager
Differential Revision: https://phabricator.kde.org/D21275
Install the policy files to the cmake provided locations instead of hardcoding the path to /etc/dbus-1
Differential Revision: https://phabricator.kde.org/D21162
Summary: Even though constantifying these parameters won't change anything(since they are being passed by value), it is done to assure they conform to rest of the coding style
Reviewers: stikonas, cjlcarvalho, #kde_partition_manager
Reviewed By: cjlcarvalho
Subscribers: #kde_partition_manager
Tags: #kde_partition_manager
Differential Revision: https://phabricator.kde.org/D20106
Summary: unsigned long long is a "C" data type
Reviewers: stikonas, cjlcarvalho
Reviewed By: cjlcarvalho
Subscribers: #kde_partition_manager
Tags: #kde_partition_manager
Differential Revision: https://phabricator.kde.org/D20076
Summary:
lsblk command does not return model name as output in some specific cases, specially when you have a micro SD in your computer (perhaps it should happen with a normal SD card as well, but I haven't tested it).
So I included a verification in SfdiskBackend::scanDevice to check for kname in lsblk and use it as the name in these cases.
Reviewers: stikonas
Reviewed By: stikonas
Tags: #kde_partition_manager
Differential Revision: https://phabricator.kde.org/D16577
Add CopyTargetByteArray. CopySourceByteArray is not implemented yet.
This is only suitable for reading small amount of data such as GPT header
or FAT boot sector location. Not meant for copying whole partition because
data has to be transfered over DBus.
Differential Revision: https://phabricator.kde.org/D16487
Summary:
the motivation here is to terminate the "server" helper when the client
disappears. dbus supports this use case natively via service registration
events which we can easily handle via QDBusServiceWatcher.
instead of repeatedly poking the client we'll simply monitor its dbus
service now. this is cheaper, less code and doesn't risk timing out
randomly.
Test Plan:
- on neon ISO build kpmcore & calamares & pm
- calamares manages to actually partition stuff
- partitionmanager also starts properly
- also the same again on the installed system.
Reviewers: stikonas, bshah
Reviewed By: bshah
Subscribers: bshah
Differential Revision: https://phabricator.kde.org/D14646
Andrius Štikonas
Adriaan de Groot
l10n daemon script
Gaël PORTAY
Yuri Chornoivan
Arnaud Ferraris
Caio Jordão Carvalho
Shubham Jangra
Harald Sitter
Valerii Malov
Antonio Rojas
Albert Astals Cid
Anthony Fieroni
Simon Depiets