Also addresses a number of issues:
- StopHostNetworking isn't plumbed, win-sshproxy leaks on hyperv
- Wait api and print output doesn't work properly on Windows
- API forwarding doesn't work on WSL
- Terminal corruption with after start/stop on Windows
- Gvproxy is forcefully killed vs gracefully quit
- Switching rootful/rootless does not update /var/run/docker.sock on the guest
- File already closed error on init
- HyperV backend is publishing Unix sockets when it should be named pipes
- User-mode networking doesn't always work
- Stop state outside of lock boundaries
- WSL blocks parallel machined (should be supported)
[NO NEW TESTS NEEDED]
Signed-off-by: Jason T. Greene <jason.greene@redhat.com>
This PR is only a few cleanup items where code is no longer used. More
to come I'm sure
[NO NEW TESTS NEEDED]
Signed-off-by: Brent Baude <bbaude@redhat.com>
Fix a bug where if a vm exists, created by some other process/user, and
you attempt to make a podman machine with the same name.
Signed-off-by: Brent Baude <bbaude@redhat.com>
Align the behaviour of `podman kube play file.yaml` to Kubernetes' by forcing
an image pull when `imagePullPolicy` is omitted and the container image does
not specify a tag.
Signed-off-by: Maurizio Porrato <mporrato@redhat.com>
Added the module github.com/crc-org/crc/ as a dependency.
Updated `decompress.go` and `copy_test.go` in compression
so that `CopySparse` from crc-org/crc/v2/pkg/os is used instead
of the local version in `copy.go`.
Deleted `copy.go` that is not used anymore.
Signed-off-by: Mario Loriedo <mario.loriedo@gmail.com>
SSH uses 22 as default so it is really not necessary to require the
port. The backend code already does this but the parsing in the
frontend always tried to parse the port.
[NO NEW TESTS NEEDED] This would require actual remote host ssh setup in
CI so it is not possible to be check but I verified it locally.
Fixes https://issues.redhat.com/browse/RHEL-17776
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Sets up USB passthrough for machine. Additionally moves `SetOptions` out
from `pkg/machine/config.go` to its own file in
`pkg/machine/define/setopts.go`.
[NO NEW TESTS NEEDED]
Signed-off-by: Jake Correnti <jakecorrenti+github@proton.me>
Changes the signature for `getDefaultDevices` to take
a `vmconfigs.MachineConfig`.
[NO NEW TESTS NEEDED]
Signed-off-by: Jake Correnti <jakecorrenti+github@proton.me>
This is completely untested as I do not have access to a freebsd system
but it compiles and changes look simple enough to assume it works.
[NO NEW TESTS NEEDED]
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Moving from Go module v4 to v5 prepares us for public releases.
Move done using gomove [1] as with the v3 and v4 moves.
[1] https://github.com/KSubedi/gomove
Signed-off-by: Matt Heon <mheon@redhat.com>
Re-enable the commented-out code that passed the path to a logfile to
gvproxy when the user passes `--log-level=debug`
Signed-off-by: Jake Correnti <jakecorrenti+github@proton.me>
Adds the functionality for `podman machine set --rootful` for AppleHV,
QEMU, and HyperV. Abstracts the functionality out to a method of
`MachineConfig`. WSL currently uses a function `SetRootful` that is
provided by the `machine` package, which will eventually get changed
when WSL moves to the refactored structure.
Re-enables the "set rootful with docker sock change" test.
[NO NEW TESTS NEEDED]
Signed-off-by: Jake Correnti <jakecorrenti+github@proton.me>
Signed-off-by: Brent Baude <bbaude@redhat.com>
gz by definition is not able to preserve the sparse nature of files. using some code from the crc project and gluing it together with our decompression code, we can re-create the sparseness of a file. one downside is the operation is a little bit slower, but i think the gains from the sparse file are well worth it in IO alone.
there are a number of todo's in this PR that would be ripe for quick hitting fixes.
[NO NEW TESTS NEEDED]
Signed-off-by: Brent Baude <baude@redhat.com>
in various use cases, the required machine dirs are not created. the
machine dirs are runtimedir, datadir, and configdir. Example in Linux
would be:
configDir /<HOME>/.config/containers/podman/machine/<provider>
dataDir /<HOME>/.local/share/containers/podman/machine/<provider>
runtimeDir /run/user/1000/podman/machine
now we blindly create them without checking for their existence (because
it is faster).
this fixes a bug where runtimedir does not exist on macos after a reboot
[NO NEW TESTS NEEDED]
Signed-off-by: Brent Baude <baude@redhat.com>
this pr represents the podman 5 maching refactoring for HyperV. with
the exception of already skipped tests, all local tests pass.
Signed-off-by: Brent Baude <bbaude@redhat.com>
this is the second provider done (qemu first). all tests pass on arm64 hardware locally ... the hybrid pull from oci registries limit this to arm64 only.
calling gvproxy, waiting for it, and then vfkit seems to still be problematic. this would be an area that should be cleaned up once all providers are implemented.
Signed-off-by: Brent Baude <bbaude@redhat.com>
The intial refactor used specifically qemu for testing and infra bring
up. But the whole point was to have things interfaced. This PR results
in an interface experience like podman 4 using the same term `provider`
to generically represent 'a provider' like qemu/applehv/etc.
This PR is required to move forward with new providers.
Also renamed pkg/machine/p5 to pkg/machine/shim.
[NO NEW TESTS REQUIRED]
Signed-off-by: Brent Baude <bbaude@redhat.com>
The following PR is the leading PR for refactoring podman machine with
the following goals:
* less duplication/more re-use
* common configuration file between providers
* more consistentency in how machines are handled by providers
The goal of this PR is the rough refactor. There are still rough spots
for sure, specifically around the podman socket and pipe. This
implemention is only for Linux. All other providers are still present
but will not compile or work. This is why tests for them have been
temporarily suspended.
The ready socket code is another area that needs to be smoothed over.
Right now, the ready socket code is still in QEMU. Preferably it would
be moved to a generic spot where all three approaches to readiness
socket use can be defined.
It should also be noted:
* all machine related tests pass.
* make validate for Linux passes
* Apple QEMU was largely removed
* More code pruning is possible; will become clearer when other
providers are complete.
the dir pkg/machine/p5 is not permanent. i had to seperate this from
machine initially due to circular import problems. i think when all
providers are done (or nearly done), it can be placed and named
properly.
Signed-off-by: Brent Baude <bbaude@redhat.com>
this is a logical place to get changes upstream before they grow out of
control. this pr is the first in an effort to deduplicate machine code
and streamline code flow.
a lot of code is simply moved to eliminate circular imports. names and
specific paths can ultimately be changed. i dont like some of the
descriptive interface names, etc. ultimately, i think once we have the
"old" code sanitized, we can re-use some of those.
clearly some of what is in here is temporary and will either be deleted,
changed, or moved again as this effort comes to a close.
right now, the machine code does not use any of the "new" code. you
will see in `init` and `rm` some commented out code that hooks it. i'm
afraid things will get worse before they get better (way worse).
[NO NEW TESTS NEEDED]
Signed-off-by: Brent Baude <bbaude@redhat.com>
This mirrors how the Docker API handles things, allowing us to be
more compatible with Docker and more verbose on the Libpod API.
Stats are given as per network interface in the container, but
still aggregated for `podman stats` and `podman pod stats`
display (so the CLI does not change, only the Libpod and Compat
APIs).
Signed-off-by: Matt Heon <mheon@redhat.com>
The pasta network mode has been added in podman v4.4 and this causes a
conflict with named networks that could also be called "pasta". To not
break anything we had special logic to prefer the named network over the
network mode. Now with 5.0 we can break this and remove this awkward
special handling from the code.
Containers created with 4.X that use a named network pasta will also
continue to work fine, this chnage will only effect the creation of new
containers with a named network pasta and instead always used the
network mode pasta. We now also block the creation of networks with the
name "pasta".
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
The current field separator comma of the inspect annotation conflicts with the mount options of --volumes-from as the mount options itself can be comma separated.
Signed-off-by: Vikas Goel <vikas.goel@gmail.com>
Fix the way we set skipTLSVerify on the client side
to ensure that the push stage in farm build takes into
account the configuration in the farm node's registries.conf
when the user hasn't set it on the client side.
Signed-off-by: Urvashi Mohnani <umohnani@redhat.com>
Podman Desktop [1] is looking into improving the user experience which
requires to know the source of an image. Consider the user triggers an
image pull and Podman Desktop wants to figure out whether the image name
refers to a Red Hat registry, for instance, to prompt installing the RH
auth extension.
Since the input values of images may be a short name [2], Podman Desktop
has no means to figure out the (potential) source of the image. Hence,
add a new `/resolve` endpoint to allow external callers to figure out
the (potential) fully-qualified image name of a given value.
With the new endpoint, Podman Desktop can ask Podman directly to resolve
the image name and then make an informed decision whether to prompt the
user to perform certain tasks or not. This for sure can also be used
for any other registry (e.g., Quay, Docker Hub).
[1] https://github.com/containers/podman-desktop/issues/5771
[2] https://www.redhat.com/sysadmin/container-image-short-names
Signed-off-by: Valentin Rothberg <vrothberg@redhat.com>
We now no longer write containers.conf, instead system connections and
farms are written to a new file called podman-connections.conf.
This is a major rework and I had to change a lot of things to get this
to compile again with my c/common changes.
It is a breaking change for users as connections/farms added before this
commit can now no longer be removed or modified directly. However because
the logic keeps reading from containers.conf the old connections can
still be used to connect to a remote host.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
When inspecting a container that does not define any health check, the health field should return nil. This matches docker behavior.
Signed-off-by: Ashley Cui <acui@redhat.com>
SpecGen is our primary container creation abstraction, and is
used to connect our CLI to the Libpod container creation backend.
Because container creation has a million options (I exaggerate
only slightly), the struct is composed of several other structs,
many of which are quite large.
The core problem is that SpecGen is also an API type - it's used
in remote Podman. There, we have a client and a server, and we
want to respect the server's containers.conf. But how do we tell
what parts of SpecGen were set by the client explicitly, and what
parts were not? If we're not using nullable values, an explicit
empty string and a value never being set are identical - and we
can't tell if it's safe to grab a default from the server's
containers.conf.
Fortunately, we only really need to do this for booleans. An
empty string is sufficient to tell us that a string was unset
(even if the user explicitly gave us an empty string for an
option, filling in a default from the config file is acceptable).
This makes things a lot simpler. My initial attempt at this
changed everything, including strings, and it was far larger and
more painful.
Also, begin the first steps of removing all uses of
containers.conf defaults from client-side. Two are gone entirely,
the rest are marked as remove-when-possible.
[NO NEW TESTS NEEDED] This is just a refactor.
Signed-off-by: Matt Heon <mheon@redhat.com>
