This commits adds implementation of reporting device plugin status
into the Module's status.
The following fields are reported:
1) number of targeted nodes ( based on selector field)
2) number of desired podsto be run
3) current number of the pods actually running

Currently webhook verifies that module's name + namespace length
does not exceed 40 chars in case version is defined. This commit removes
the version condition and always verifies that length

Since the name of the volumes and volume mounts that are configured
for service-account pull-secrets are constructed from "pull-secret"
and the actual pull-secret name, then it can result in names being more
then 63 chars which will cause an error. This comment move the volumes
names to be "pull-secret-hash(pull-secret-name)"

Try loading or unloading kernel modules until either:
- the Module is deleted, or;
- the Pod template has changed.

emove the inProgress field of the NMC statuses.

This commits adds implementation of reporting worker pods status
into the Module's status.
The following fields are reported:
1) number of targeted nodes ( based on selector field)
2) number of desired nodes that kernel module should be deployed (based
   on NMCs configured)
3) current number of the nodes that the kernel module is already
   deployed to ( based on NMC spec and status configs)

Because this changes the deployment name in the CSV, it should allow for
OLM upgrades despite the change in label selectors.

Bump versions for some of our generators.

operator-sdk now sets that field for us.
Make new environment variables for the signer and worker image names
mandatory.

* Remove deprecated module ready labels on nodes (#668)

This change removes 'kmm.node.kubernetes.io/<module>.ready' labels on
all nodes, as it was deprecated in v1.1 and removed in v2.0.

* Fix handling of unknown kernels (#670)

Fix scheduling data generation for nodes that run an unconfigured
kernel and do not have the kmod loaded yet.

* Build / sign: handle all owner kinds (#673)

Make the build / sign events reconciler handle any owner for jobs, not
only Module.

* Refactor worker to support different type of image to be mounted (#677)

Currently worker only supports remote image mounting (from a repo).
In the future we might want to support also mounting an image from
a local tar file. This Pr lays a foundation for that. The basic idea
is that all the code should stay the same, and only the image mounting
implementation should change, as well as adding additional cobra
commands. The sequence to add a new image mounter option is:
1. Define a new cobra command and mount it under root command
2. the PreRun command in the new command should initialize the worker
   with the new image mounter
3. define a new load/unload cobra commands that should be exact replicas
   of the existing load/unload cobra command. The copy must be done due
   to the fact that in cobra a command cannot be a child of multiple
   commands
Changes in the PR:
1. kmod cobra command now initializes the worker. it will initiaze it
   with remote image mounter implementation. In case we need to add ad
   different mounter, we just define a different kmod cobra command and
   it will initialize the worker with the different image mounter
   implmentation
2. Change ImagePuller interface to ImageMounter interface
3. Move ImageMounter interface definition to a dedicated file
4. Move the implementation of the RemoteImageMounter into the dedicated
   file
5 update unit-tests

* Fixing DevicePlugin upgrade from v1.x to v2.x (#679)

In v1.x we had 2 type of Daemonsets: DevicePlugin and ModuleLoader,
which could be distingished by the DaemonsetRole label
In v2.x, there only 1 type of daemonset, DevicePlugin, so the
DaemonsetRole labels was removed.
During DevicePlugin reconciliation, reconcile get all the device plugin
Ds based only on module name label, which means that in case there is a
v1.1 ModuleLoader DS running it will also be chosen. Once reconciler
tries to reconcile ModuleLoader DS, it will fail
This PR, updates the function that chooses the DevicePlugin DS, by
removing old ModuleLoader Ds based on Role label

---------

Co-authored-by: Yevgeny Shnaidman <[email protected]>

Create links as is when extracting files to the worker Pod filesystem.

Signed-off-by: Enrique Belarte Luque <[email protected]>
Co-authored-by: Enrique Belarte Luque <[email protected]>

Co-authored-by: Quentin Barrand <[email protected]>

This is required for the BuildSignEvents reconciler to remove the
finalizer.

Co-authored-by: Quentin Barrand <[email protected]>

* Add a webhook for namespace deletion (#719)

This commit makes KMM set the kmm.node.k8s.io/contains-modules label on
all namespaces that contain at least one Module.
It also adds a new webhook to the bundle and the corresponding handler
in the manager.
The new webhook rejects namespace deletions if the
kmm.node.k8s.io/contains-modules label is present on the namespace.
This avoids entering situations where the namespace is being deleted and
KMM cannot create unloading Pods to honor Module deletion.

* Restrict checks for image existence (#734)

Before populating the NodeModulesConfig object, only check if built or
signed image exist on the registry.
Add a log message when the NodeModulesConfig is not populated for that
reason.

* Make slight changes to the CRDs (#736)

Module: make moduleName an optional field.
ManagedClusterModule: make spokeNamespace a required field.