autofs: ioctl error handling may cause threads to hang in D state on Oracle Linux 8.8 (similar to Red Hat Bugzilla 2023740) #178 #179
Description
Since the previously reported issue was closed (#178), I’m opening a new one to provide additional context and clarify reproducibility steps.
The problem can be consistently reproduced by performing an autofs RPM upgrade or downgrade. After this change, certain automount operations hang indefinitely unless the system is rebooted.
Notably:
• There is no documentation explicitly stating that a system restart is required after upgrading or downgrading the autofs package.
• The command dnf needs-restarting -r does not flag any services requiring a reboot, which may mislead administrators into assuming a restart is unnecessary.
This can lead to production-impacting hangs, particularly for services (e.g., MongoDB FTDC) that rely on automounted paths.
Please advise if a restart should be formally required after autofs RPM changes, or if this behavior can be mitigated in future releases.