There appears to be an issue with newer QEMU versions (spotted with Arch
Linux and C9S containers) which causes the respective GH Action to hang
when booting a QEMU VM in combination with the `-cpu max` parameter.
During (a particularly painful) debugging session I once managed to get
some output from such "frozen" machine (using `earlycon` and
`earlyprintk` kernel cmdline options), and in that particular case the
VM died with a trap caused by an invalid opcode.
I couldn't reproduce this locally, only in GH Actions environment with
Arch Linux and C9S containers. Also, so far I haven't found out which
specific CPUID flag causes this, but using the `IvyBridge-v2` feature
set seems to mitigate the issue.
Due to parallel probing of the linux kernel `/dev/sd*` can't be used to
reliably address a hard disk. This can be seen by the many spurious
failures of the dracut CI, where `mdadm` failed with error 524 or tests
failed due to the success marker message written to the wrong disk.
* don't rely on `/dev/sd*` but use disk ids and `/dev/disk/by-id/ata-disk_<name>`
* specify the exact qemu machine architecture `-M q35` needed for the
disk ids. A later patch will move this to `run-qemu`, when all tests are converted
* due to `-M q35` the interface names have changed from
`ens2` -> `enp0s1` and `ens3` -> `enp0s2`
If the host supports kvm, use is. If this is not the case but the kqemu
module is loaded, run qemu with kqemu optimization. Otherwise fall-back to
pure usermode qemu.
Also create the root filesystem from within the virtual env, this way there
will be no nasty suprises w.r.t interactions with the host machine. This
is important when we test booting to LVM, LUKS, and mdraid.
Frantisek Sumsal