`cosa run` sometimes fails with "swtpm: Could not open UnixIO socket: No such file or directory"

[jlebon]

Happened just now again:

$ cosa run
warning at $.storage.files.2.mode, line 19 col 14: permissions unset, defaulting to 0644
qemu-system-x86_64: -chardev socket,id=chrtpm,path=/tmp/cosa-tpmstate.tQ8yMB/swtpm-sock: Failed to connect socket /tmp/cosa-tpmstate.tQ8y
swtpm: Could not open UnixIO socket: No such file or directory

At that point, I just rerun and it works. But we should investigate what's going on there.

[jlebon]

FCOS pipeline recently failed with this related error:

+ cosa fetch --strict
info: Missing CAP_SYS_ADMIN; using virt
Formatting 'cache/cache2.qcow2', fmt=qcow2 size=10737418240 cluster_size=65536 lazy_refcounts=off refcount_bits=16
removed 'cache/cache.qcow2'
Config commit: 8db8857f3f480f0909d4d4eb6bcd759d1748034c
Using manifest: /home/jenkins/workspace/fedora-coreos/fedora-coreos-fedora-coreos-pipeline/src/config/manifest.yaml
Running: rpm-ostree compose tree --repo=/home/jenkins/workspace/fedora-coreos/fedora-coreos-fedora-coreos-pipeline/tmp/repo --cachedir=/home/jenkins/workspace/fedora-coreos/fedora-coreos-fedora-coreos-pipeline/cache --touch-if-changed /home/jenkins/workspace/fedora-coreos/fedora-coreos-fedora-coreos-pipeline/tmp/build/tmp/treecompose.changed --unified-core /home/jenkins/workspace/fedora-coreos/fedora-coreos-fedora-coreos-pipeline/src/config/manifest.yaml --download-only --ex-lockfile=/home/jenkins/workspace/fedora-coreos/fedora-coreos-fedora-coreos-pipeline/src/config/manifest-lock.x86_64.json --ex-lockfile=/home/jenkins/workspace/fedora-coreos/fedora-coreos-fedora-coreos-pipeline/src/config/manifest-lock.overrides.x86_64.yaml --ex-lockfile-strict
qemu-system-x86_64: -chardev socket,id=chrtpm,path=/tmp/mantle-qemu938177700/swtpm-sock: Failed to connect socket /tmp/mantle-qemu938177700/swtpm-sock: No such file or directory
Error: exit status 1
2020-05-09T21:24:58Z cli: exit status 1
fatal: Failed to run 'kola qemuexec'

[jlebon]

Hit the same error as the pipeline locally:

$ cosa run --qemu-image /srv/fcos/tmp/fastbuild/fastbuildinitrd-fedora-coreos-qemu.qcow2 --qemu-multipath -c
qemu-system-x86_64: -drive if=none,id=mpath10,file=nbd:unix:/tmp/mantle-qemu099162466/disk738004825.socket,auto-read-only=off,media=disk: Failed to connect socket /tmp/mantle-qemu099162466/disk738004825.socket: No such file or directory
Error: exit status 1
2020-05-13T20:46:30Z cli: exit status 1

Something like swtpm creating the socket racing with qemu starting?

[jlebon]

https://github.com/coreos/coreos-assembler/pull/1499

[c4rt0]

Presumably fixed by https://github.com/coreos/coreos-assembler/pull/1499