Multiple kernels kver_cmd to return an incoherent string

[Peter Mattingly]

Hello,

A customer had contacted with an issue related to this bug. Taking a look at the stacktrace for the issue:

```
ls -la
total 16
drwxr-xr-x. 4 root root 4096 Feb 10 02:53 .
dr-xr-xr-x. 40 root root 4096 Feb 10 02:58 ..
drwxr-xr-x. 6 root root 4096 Feb 10 02:54 4.18.0-372.32.1.el8_6.x86_64
drwxr-xr-x. 6 root root 4096 Feb 10 02:49 4.18.0-372.9.1.el8.x86_64

<stack trace>
Running curthooks.builtin_curthooks...
Traceback (most recent call last):
File "/tmp/tmp1x5k7i8r/target/curtin/curtin-hooks", line 133, in <module>
main()
File "/tmp/tmp1x5k7i8r/target/curtin/curtin-hooks", line 129, in main
curthook(config, target, state)
File "/tmp/tmp1x5k7i8r/target/curtin/curtin-hooks", line 114, in curthook
curthooks.builtin_curthooks(cfg, target, state)
File "/curtin/curtin/commands/curthooks.py", line 1831, in builtin_curthooks
redhat_update_initramfs(target, cfg)
File "/curtin/curtin/commands/curthooks.py", line 1672, in redhat_update_initramfs
in_chroot.subp(dracut_cmd, capture=True)
File "/curtin/curtin/util.py", line 787, in subp
return subp(*args, **kwargs)
File "/curtin/curtin/util.py", line 275, in subp
return _subp(*args, **kwargs)
File "/curtin/curtin/util.py", line 139, in _subp
raise ProcessExecutionError(stdout=out, stderr=err,
curtin.util.ProcessExecutionError: Unexpected error while running command.
Command: ['unshare', '--fork', '--pid', '--', 'chroot', '/tmp/tmp1x5k7i8r/target', 'dracut', '-f', '/boot/initramfs-4.18.0-372.9.1.el8.x86_644.18.0-372.32.1.el8_6.x86_64.img', '4.18.0-372.9.1.el8.x86_644.18.0-372.32.1.el8_6.x86_64']
Exit code: 1
Reason: -
Stdout: ''
Stderr: dracut: Cannot find module directory /lib/modules/4.18.0-372.9.1.el8.x86_644.18.0-372.32.1.el8_6.x86_64/
dracut: and --no-kernel was not specified
```

Then following the stack trace, this line in the source can be found:

```
kver_cmd = ['rpm', '-q', '--queryformat', '%{VERSION}-%{RELEASE}.%{ARCH}', 'kernel']
```

Then, if this command is run on a system with at leas two kernels installed, the returned string is malformed:

```
rpm -q --queryformat='%{VERSION}-%{RELEASE}.%{ARCH}' kernel
4.18.0-477.10.1.el8_8.x86_64
```

Which causes an issue in the customer's use-case.

As this issue is time sensitive, would it be possible to request a workaround while a fix is developed?

Thank you.

[Nicholas Fries (nicfries)]

We have a workaround (we ensure only one kernel exists in our image). This issue can occur if a user builds and image and in the process runs a yum update. In this case, two kernels may exist in /boot causing this issue.

[Dagmawi Biru (dagbiru)]

The issue appears to be sourced here:
https://github.com/canonical/curtin/blob/master/curtin/commands/curthooks.py#L1708

[Peter Neal (peterneal)]

Bringing this back to life,

Yes, on a system with two or more kernels present,

rpm -q --queryformat='%{VERSION}-%{RELEASE}.%{ARCH}' kernel

will output the results for both kernels concatenated together, i.e. if the system has both 'kernel-4.18.0-372.9.1.el8' and 'kernel-4.18.0-372.32.1.el8_6' present, this command returns `4.18.0-372.9.1.el8.x86_644.18.0-372.32.1.el8_6.x86_64`. This gets passed to dracut, fails, and fails the deployment.

I propose simply changing the query to `rpm -q --queryformat='%{VERSION}-%{RELEASE}.%{ARCH}\n' kernel`, capturing the output in some array $kvers, then proceed inside of something like `for(kver in kvers): dracut -f kver`.

This should also be applied to oracle linux but for the kernel-uek package.

It's not convenient to purge extra kernels during image build, and there are cases where we want > 1 kernels available on deployment.