[systemd-devel] removal of RD_TIMESTAMP support in initrd

Lennart Poettering lennart at poettering.net
Mon Jul 22 15:44:22 PDT 2013


On Fri, 19.07.13 22:02, Dave Reisner (d at falconindy.com) wrote:

> Hi,
> 
> With systemd 206 near release, I'd like to understand if commit
> c72aadd1851096ea is going to stand. This commit removes support for
> reading RD_TIMESTAMP in the initramfs, and thus makes systemd mandatory
> for measuring initramfs runtime.
> 
> If this is the intended future, please help me to understand the
> rationale behind removing 25 lines of code for a useful feature -- one
> which has extremely low overhead and is self contained code-wise.
> 
> I strongly object to the way this was done and would appreciate a more
> "official" explanation.

(I didn't remove the lines in question, but I am fine with it.)

So, we try to keep our code base clean. We delete stuff from time to
time if it it's obsolete and not important to keep for
compatibility. RD_TIMESTAMP is obsolete, we can do this much better now
(and more comprehensively) via serializing/deserializing systemd state
between inird and the host OS. And the stuff is not important for
compatibility, since this always has been one of those features that if
detected are used, but if they aren't it doesn't matter much. It's
nothing that prevents you to boot, causes services to fail, corrupts
anything. It's really just about whether "systemd-analyze" will you show
3 boot time values split out instead of 2...

I do understand that you are currently not running systemd in the initrd
[1], so for you for now this is indeed a loss of functionality. I am
sorry for that, but please understand this as gentle push to maybe use
systemd in the initrd, or even better maybe just adopt dracut?

Lennart

[1] I remember Arch's Tom Gunderson working with Harald on improving the
mounting logic in darcut/systemd quite a bit at the last
hackfest. Because of Arch we now have much nicer mount code in
Dracut/systemd. It would be a pitty if that'd be lost to Arch itself,
where this came from...

-- 
Lennart Poettering - Red Hat, Inc.


More information about the systemd-devel mailing list