Some minor quirks to fix after update from 43 to 44 wrt new directory hash certificate system & other things

in case this may help someone else. I updated to 44 yesterday. I have run into a few minor error, easily fixable:

  1. CA certificate bundle: I use a script to alert myself that a freefilesync backup has been completely successfully or not. Today I was getting ffs errors. I found that msmtp (and possibly other mail clients) the CA certificate bundle has moved as part of Fedora 44’s new directory-hash certificate system. If you use msmtp and have tls_trust_file pointing to /etc/pki/tls/certs/ca-bundle.crt, that file no longer exists. Update it to: /etc/pki/ca-trust/extracted/pem/tls-ca-bundle.pem

Other applications that reference the old bundle path directly may have the same issue.

  1. Also FreeFileSync - I was getting errors that previously weren’t showing as errors. Two new exclusion patterns are needed after the upgrade:
  • .ssh/agent/* — OpenSSH 10.1 (shipped with F44) moved the SSH agent socket from /tmp to ~/.ssh/agent/. FreeFileSync will error trying to copy the socket file.
  • *-wal and *-shm — not strictly F44-related, but if you back up home directories with running applications like Slack, these SQLite transient files can appear and disappear mid-backup causing errors.

3. Nvidia CUDA / Docker containers with GPU passthrough: If you run Docker containers that use Nvidia GPU passthrough (e.g. for AI/ML workloads), your containers may fail to start after the upgrade. The error will reference a missing library from your old driver version, something like open /usr/lib64/libEGL_nvidia.so.580.142: no such file or directory. This is because the CDI spec at /etc/cdi/nvidia.yaml was generated under your old driver and still references the old library paths. Regenerate it to match your new driver:

sudo nvidia-ctk cdi generate --output=/etc/cdi/nvidia.yaml

If your containers still don’t start after that, also check /etc/docker/daemon.json and ensure "default-runtime": "nvidia" is set.

1 Like