Bugzilla – Bug 1190024
haveged service has become obsolete with recent kernels (>= 5.6)
Last modified: 2024-02-15 15:06:54 UTC
According to https://github.com/jirka-h/haveged/commit/297bdf1fc52fc6f59d0495f911d4e594b4d29190, starting from Linux kernel v5.6, the HAVEGED *service* has become obsolete. So it might make sense: - drop the unit file from the package - not install haveged by default
Secteam, can you please reconsider this bug ?
This is an autogenerated message for OBS integration: This bug (1190024) was mentioned in https://build.opensuse.org/request/show/980343 Factory / patterns-base
we reverted it i think on feedback from darix: r141 | msmeissn | 2021-11-02 08:19:20 | b15b4981794dc7c2d3e6d63e7a4ffd85 | unknown | - revert last change, e.g. for VMs where we are not being fed entropy from the host or similar setups. -------------------------------------------------------------------- - Improvements on the linux kernel random subsystem have made the haveged service/daemon obsolete, remove the service files, initrd modules and udev rules, the other components are still useful.
(In reply to Marcus Meissner from comment #3) > we reverted it i think on feedback from darix: > > r141 | msmeissn | 2021-11-02 08:19:20 | b15b4981794dc7c2d3e6d63e7a4ffd85 | > unknown | > > - revert last change, e.g. for VMs where we are not being fed entropy > from the host or similar setups. From my understanding, kernel >= 5.6 got support for HAVEGED algorithm which should replace haveged completely. Unless those VMs ran an older kernel, I'm not sure to understand why this needed to be reverted. Can you provide more details ?
We can just speak from our production VMs that we still saw stalls without haveged when waiting for entropy.
TBH as long as we ship the haveged binary we should have the service file. if you want to reduce the package to just the shared library, then we could drop the service file too.
I am just curious ... what would be the motivation to still ship the binaries but not the service files anymore?
We could also drop the binary haveged and ship the library only if we were sure that the support added in kernel 5.6 is enough. But you reported some issues with some VMs of yours without haveged in initrd so it would be interesting to figure out why HAVEGE implementation in the kernel is not working as good as haveged.