- Listen before starting the vm
- Fix a device race caused by lazy hv_vsock init by waiting on network manager
[NO NEW TESTS NEEDED]
Signed-off-by: Jason T. Greene <jason.greene@redhat.com>
... instead of using a multi-variable state machine.
The net effect of this code is exactly the same as the previous implementation,
except:
- the operation after Write() returns an error might differ
- If the file ends with zeroes, we don't Seek(-1), and
we don't create a hole at all if it is too small, preferring
to save a syscall.
But this formulation is hopefully easier to prove correct.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
- Add more test cases
- Test that we create the expected (large) holes;
don't enforce anything for the <zerosThresholt ones.
O)nly changes test code, should not change behavior.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
When we want the original image to be gzip, explicitly ask for that
instead of assuming the containers.conf defaults do that.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
When we want the original image to be gzip, explicitly ask for that
instead of assuming the containers.conf defaults do that.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Fixes the "machine rm --save-ignition --save-image" test so that it no longer
uses the `{{ .Image }}` format string.
Fixes the "init should cleanup on failure" test so that it no longer
uses the `{{ .Image }}` and `{{ .ConfigPath }}` format strings.
Signed-off-by: Jake Correnti <jakecorrenti+github@proton.me>
We don't care about the provenance of the machine image, so this is no
longer applicable to have when displaying info.
Signed-off-by: Jake Correnti <jakecorrenti+github@proton.me>
Changes the order in which the machine-specific files are removed in
`Remove()`. Removes the system connections first, then removes the
`configPath` last. `configPath` is removed last, because in the case of
an error with any of the previous files, the removal can be attempted
again since the machine still "exists".
Made the errors in `Remove` hard errors instead of soft errors.
Added the implementation for the QEMU-specific file removal.
[NO NEW TESTS NEEDED]
Signed-off-by: Jake Correnti <jakecorrenti+github@proton.me>
The ID filed in the Event struct is duplicated for no reason, since the
Details struct is directly embedded in the Event the ID filed is
basically duplicate on the same level multiple times. Removing this one
should be be safe and make no change to the resulting json.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
When gvproxy exits it will delete the pidfile itself so we need to
account for that and juts ignore the case, it just means gvproxy was
able to exit successfully on its own.
Also remove the useless defer and return the error so we can get an
error exit code not just a print on stderr.
Currently it shows this error which is not helpful to any user:
unable to clean up gvproxy: "unable to read gvproxy pid file /run/user/1000/podman/gvproxy.pid: open /run/user/1000/podman/gvproxy.pid: no such file or directory"
[NO NEW TESTS NEEDED] TODO: make machine tests check stderr for such
things.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
ToHumanReadable() exists twice now, there is no reason for this just
call the function on the backend event type is fine as this still has to
be used there.
It also fixes a bug where the wrong event type was passed to the
template which did not match the docs and json output.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Add new event type in cmd/podman to better match the docker format.
Signed-off-by: AhmedGrati <ahmedgrati1999@gmail.com>
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
Copy all proxy envs into the VM on each start, this allows for updates
without having to recrate the VM. This is implemented via shell script
that is passed via ssh to the VM after it is started.
With that we now use the same logic for all providers the old fw_cfg
logic for qemu has been removed and the WSL code as well which keeps the
behavior the same.
There is a small risk now because we only update the env via ssh that
processes started before will have the old incorrect env but it should
really only effect core system processes which likely do not need them
anyway. The podman system service should not be started at this point
so it should be good enough.
It also fixes the broken behavior with SSL_CERT_FILE/SSL_CERT_DIR which
were updated on each start which is not correct as the files are only
copied with ignition so these should not be updated and just set
statically when the VM was created.
e2e test has been added to ensure the behavior works as expected.
Signed-off-by: Paul Holzinger <pholzing@redhat.com>
