Change the VolumeManager Factory to be a context manager.
All code using VolumeManager was updated to the following
with statement:
with VolumeManager(...).new as volume_manager:
volume_manager.some_member()
This is related to Issue #2412
Use proper Union declaration for system variable and add
consistency layer into Filesystem/VolumeManager classes to
meet the type declaration as well as to simplify further
refactoring on these classes
Change the FileSystem class to be a context manager. All code using
FileSystem was updated to the following with statement:
with FileSystem.new(...) as filesystem:
filesystem.some_member()
This is related to Issue #2412
For the btrfs volume management, allow to put a volume into a specific
parent volume. If not specified the volume is below the default volume
This Fixes#2316
In a volume setup the special volume declaration
<volume name="@root=identifier"/> was only evaluated for the
LVM volume manager. In case of btrfs a hardcoded root volume
name '@' was used. This commit allows to specify a custom
name for the root volume for btrfs as well and also allows
to specify that there should be no such root volume.
Example:
<volume name="@root=@"/>
Name the root volume '@'. If not specified this stays as
the default to stay compatible
<volume name="@root=/"/>
Indicate no root volume is wanted. All subvolumes resides
below root (/)
<volume name="@root=foo"/>
Name the root volume 'foo'
This is related to Issue #2316 and a first patch to
address the requested changes
The /@/.snapshots subvolume was not mounted as /.snapshots in the root
filesystem snapshot. This is now necessary for snapper to work.
While at it, create 1/snapshot as plain subvolume, it does not make sense
to snapshot @ itself.
Along with creating a filesystem including device mapper features
like dm_verity (see verity_blocks) or dm_crypt/dm_integrity (see luks)
there is always the question where to store the metadata information
required to setup the device map. This can include information about
blocksizes, offset addresses and more. The create_verification_metadata()
method allows to write a signed custom data block of a documented
format at the end of the given block special which stores this type
of information such that tools at boot time gets the opportunity to
read this information. In this commit only information connected
to the dm_verity feature activated via the verity_blocks attribute
will be part of the verification block. With future changes other
data might be added
So far the verity support was only available with the
overlayroot layout and the read-only squashfs root. This
commit adds a new attribute: verity_blocks="number|all"
which allows to create the verity setup also on the
standard root partition
In addition to the change it was needed to extend the
Filesystem API with an additional optional paramter to
allow setup of the filesystem UUID. Having the opportunity
to set the UUID at filesystem creation is generally useful
and with regards to this particular change it became also
required
veritysetup was called as part of the disk builder. However,
the veritysetup should be a responsibility of the classes which
implements the sync_data method. This allows to use the creation
of a verity hash format right after sync as a feature to these
classes and in a broader scope. In addition to that change the
VeritySetup::format method now returns the metadata from the
format call and stores it as debug information to the log file.
A concept to persistently store the verification metadata as part
of the partition(s) will be a next step.
The modifications in this commit allows the unit tests
to run on both, pytest 6.x (nose test layout) and the new
pytest 7.x (xunit test layout). This Fixes#2072 in a
much nicer way. Thanks much to @smarlowucf
This reverts commit 0dc2e803e0.
The pytest interface from version v6 to v7 has received
changes which requires the tests to be adapted to work for
either the old or the new interface. As there are still many
distributions which uses v6 as the standard we decided to
revert back the adaptions done to support v7 and create
a version requirement to v6 in .virtualenv.dev-requirements.txt
This Fixes#2072
The pytest interface setup() method call has changed
in a way that an additional parameter is passed to
the method which leads to a python error at invocation
time if the setup method does not define it.
Due to the change on the temporary directory an error
in the volume manager classes for LVM and btrfs was
exposed. There was code which uses a sub-part of the
volume mount point directory which is a tempdir based
directory name. The sub-part was choosen by an index
based path split which worked for /tmp but no longer for
/var/tmp. With tempdir now being also a commandline
option this code has to become more robust.
Moving use of mkdtemp, NamedTemporaryFile and TemporaryDirectory
into its own class called Temporary: By default all temporary
data is created below /var/tmp but can be changed via the
global commandline option --temp-dir. This Fixes#1870
The new attribute <volume ... filesystem_check="true|false"/>
allows to change the default value for the fs_passno field in
the generated fstab file. By default kiwi sets "0" in this
field and leaves it up to the user to customize this as
appropriate via script code. Coding changes to the fstab file
via scripts are not very user friendly and with respect that
systemd takes over control and generates checkers depending on
the value of fs_passno it would be good if there is a way to
explicitly specify if checks to the filesystem are wanted or
not. Therefore the new attribute now exists. If set to: true
this results in a value of "2" for the fs_passno field. Please
note the root/boot and efi entries are excluded from this
setup. This Fixes#1728
This commit refactors VolumeManager to turn it into a proper
factory class and to also include type hints to facilitate it's
use from an API POV. Related to #1498
In addition to the custom size of the root volume it's now
also possible to setup the name of the root volume as follows:
<volume name="@root=rootlv"/>
If no name for the root volume is specified the default
name: LVRoot applies as before. This Fixes#1530
A vmx image is the same disk as an oem just without the dracut
repart/resize feature. This difference is better handled with
an oemconfig parameter <oem-resize> which allows to switch resize
on or off. The extra image type vmx will be dropped and an XSLT
stylesheet automatically transforms a vmx description to be a
oem image with the resize feature switched off.
This Fixes#1425
Using the --inplace option in rsync helps to save space on
syncing the rootfs data and prevents e.g OBS workers from
running out of VM space when transfering root filesystem
data. Also using --inplace allows to keep hardlinks intact.
This is related to bsc#1096738
With regards to Issue #1486 a discussion came up that the
way factories are implemented are questionable when thinking
about strong typing for the public kiwi interface. This
commit refactors the FileSystem factory to be a real factory
and to add type hints such that its use from an api perspective
is clear and enforced.
instances of VolumeManager creates temporary directories but
only stores the latest one. The cleanup leaves former directories
behind which is fixed by this commit
In preparation to allow a chroot operation into the loop
mounted disk this commit refactors the process when filesystems
gets umounted and also fixes the canonical order for calling
the destructors. Related to Issue #1464
This commit makes sure to empty RPM macros files are created during
the build. Additionally it also ensures that no needless `rm` and `mkdir`
calls are done.
Fixes#1316
The former implementation builds a new device map which is
a subset of the low level device map. However due to the
additional swap partition the maps provided by the VolumeManager
classes are now incomplete. Instead this commit changes the
VolumeManager interface to receive the current low level
device_map and merged device changes on top of it when
needed.
In an OEM deployment that requested the creation of a swap
partition via <oem-swap> that swap partition was created
at first boot and was always the last partition on the disk.
This was required because it could not be placed before
any other partition without destroying those partition
contents. This process leaves the system in an inflexible
condition if the storage device can change its geometry
dynamicly as it's the case for SAN systems. The typical
deployment target for OEM images are SAN storage clusters
and it's cumbersome to resize the root partition if swap
is last.
This commit Fixes#1231 and changes the handling of swap if
requested via <oem-swap> as follows:
1. The swap space is created as part of the image build process
and no longer on first boot of the image via dracut code.
This increases the size of the non compressed .raw disk image
by the configured swap space size or the default. The
compressed versions are not affected since zero initialized
swap space compresses to almost no space. Deployment of
the image however also deploys the swap partition which
increases deployment time. For big swap configurations
it's advisable to switch off image verification via
oem-skip-verify. For very big swap configurations it's
also recommended to prevent kiwi from adding them as part
of the image and let them be created on first boot via
a systemd service that e.g places a swap file, or creates
a swap volume when possible such that the fexibility to
resize the rootfs is still available.
2. The setup of the swap space is now explicit. It's no longer
calculated by twice times RAM size because on newer machines
this could lead to huge numbers. Either the kiwi encoded
default swap size applies or the user configured value.
3. LVM based oem disks creates the swap space as logical volume.
The volume is created as part of the image build process
and no longer on first boot. The swap volume at build time
of the image is of a minimal size and gets resized on first
boot.
4. The move of the swap creation into the builder code also
handles swap per configured device persistency schema like
any other devices. This means by default swap is mounted
via by-uuid name and thus also Fixes#1259
This commit makes sure the chrooted operations of over a loop device
including btrfs subvolumes and root as an snapshot include the
`/.snapshots` subvolume mount. This is necessary for certain operations
that depend on the layout, e.g. grub configuration using `grub2-mkconfig`.
