On Sat, Apr 25, 2020 at 2:43 AM Khem Raj <[email protected]> wrote: > > > > On 4/23/20 10:14 AM, Andre McCurdy wrote: > > On Thu, Apr 23, 2020 at 1:30 AM Alex Kiernan <[email protected]> wrote: > >> On Wed, Apr 22, 2020 at 7:48 PM Andre McCurdy <[email protected]> wrote: > >>> On Wed, Apr 22, 2020 at 6:52 AM Alex Kiernan <[email protected]> > >>> wrote: > >>>> > >>>> sshd.socket launches the templated [email protected], so by moving the > >>>> sshdgenkeys.service to sshd.socket, key generation can start in advance > >>>> of a connection. > >>> > >>> Is deferring key generation until the first connection a bug or by design? > >> > >> If you've installed it, presumably you're expecting connections. But I > >> agree you could argue it either way. > > > > The trade off is between slower first boot (or every boot for > > read-only rootfs if host keys are not persistent) -vs- slower first > > incoming ssh connection. > > > > Since the first incoming ssh connection may never happen (especially > > if ssh is only used for debugging or development), optimising it at > > the expense of boot time doesn't seem to be an obvious improvement. > > > > I tend to agree. Perhaps it could be covered under read-only-rootfs > DISTRO_FEATURE >
I just ran both through systemd-analyze. For qemuarm, this adds 5s to the total boot time, but graphical.target is reached in the same time. Clearly you can cut this one both ways - I don't have a specific problem this fixes, it was more an annoyance (for me). > > > > > > > -- Alex Kiernan
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#137462): https://lists.openembedded.org/g/openembedded-core/message/137462 Mute This Topic: https://lists.openembedded.org/mt/73195919/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
