Monday, 2025-09-08

clarkbre the fips issue is the command already removed from centos 10 stream?15:27
fungisounds like that's how it came to our attention15:28
fungii.e. trying to directly port the fips jobs from centos 9 to 1015:28
clarkbreading the justification for the change I honestly don't think those things matter too much for our use case. But considering this is a "security" stance feature I can understand why taking a stronger stance makes sense15:30
clarkbspecifically we only care about whether our workload can work and don't care about existing luks keys or ssh keys being generated properly according to fips15:30
fungiyeah, it's possible the job could just adjust the kernel cli arguments15:31
fungithough i'm increasingly unconvinced of the usefulness of that testing upstream, it's a very specific set of requirements for one particular country's government and military systems15:33
clarkbyes I definitely don't think we should build custom images just for that functioanlity. But if people can reboot into something that works well enough I won't stop them15:33
fungirequirements that by even objective standards are not particularly more secure, just ones that have been approved for use15:34
fungi(e.g. quick to forbid certain weak primitives even when used outside a security context, but *very* slow to accept newer algorithms)15:36
opendevreviewMerged openstack/governance master: Mark "Migrate from wsgi scripts" goal as completed  https://review.opendev.org/c/openstack/governance/+/95881719:01
opendevreviewMerged openstack/governance master: Show inactive project status in project.yaml  https://review.opendev.org/c/openstack/governance/+/95822919:01
spotz[m]Because I forgot while Matrix was down last week... Jonathan Wright from Alma will join us tomorrow21:14
fungivery cool!21:16
spotz[m]That's Jonathan:)21:17
JonathanWright[m]hello world21:17
fungiahoy!21:17
JonathanWright[m]Ok cool, the matrix bridge has a constant spinny-wheel for me in this channel but messages seem to work (just no history).21:18
fungiif you're talking about oftc's matrix bridge specifically, yeah i understand it's not super great21:19
spotz[m]Works for me:)21:22
sean-k-mooneyfungi: hum even in out downstream distop we do the fips enabling post install21:56
sean-k-mooneyi thikn we can enable it durign the install if and only if you have the install image your servers21:57
sean-k-mooneyhttps://github.com/openstack-k8s-operators/edpm-ansible/blob/da4023cb996a2804da88bef1d88ccd31db89ff4a/roles/edpm_bootstrap/tasks/fips.yml#L1921:59
sean-k-mooneyi know in image mode i.e. bootc they are going to do that a bit diffently but im not sure how you get form a non bootc install to a bootc one22:00
sean-k-mooneyi.e. for anyoen upgrading22:00
sean-k-mooneyi dont actully know how that is ment ot work with c10s or rocky or alma but given thsoe all branched for fedora 40 and that is targeted to fedroa 42 im not sure if that will impact the 10 based release22:05
clarkbsean-k-mooney: it sounds like c10s is already affected. Probably because it is somewhat forward looking too?22:06
sean-k-mooneyfun well it after fedroa and before rhel but our downstream ci is mostly rhel/centos 9 based. we have some usage of 10 but i dont think we have any 10 + fips yet22:07
sean-k-mooneyclarkb: currently we use dib to build our hardened images for installer provisioned nodes22:10
sean-k-mooneybut there are also experimetn to move that to using bootc in the future https://github.com/openstack-k8s-operators/edpm-image-builder/blob/0753302c26bc3123be82afa517c6ddc74b50a966/bootc/Containerfile#L14522:10
sean-k-mooneyi guess that the plan for 10 but i have not been invoved in any of that22:12
clarkbsean-k-mooney: I suspect that both dib and bootc will require that the builds also happen in a fips environment based on the fedora post list of issues22:15
sean-k-mooneymaybe it will likely depend on how well the isolation between the two envs is.22:16
clarkbI guess openssh-server will generate host keys on first start if not present (this way you don't have the asme host key everywhere) but other things may need similar treatment if they don't already do it22:17
sean-k-mooneyboth are effectivly runing in a chroot of a form so as long as you dont mix any host executable into that env it might be ok22:17
clarkbsean-k-mooney: the problem is its a kernel setting22:17
clarkbcurrently I think that only affects the kernel behavior but the proposed change in the fedora post has everything keying off of that22:18
clarkbso you basically need the kernel to be in fips mode to have anything else in fips mode which creates a fun bootstrapping problem for tools like bootc and dib22:18
sean-k-mooneyyou mena fips=1 22:19
sean-k-mooneybut that not actully a kernel parmater for the kernel to use22:19
sean-k-mooneyits a kernal paramter that is sued by the installer i think22:19
JayFit signals dracut, I'm guessing?22:19
JayFanaconda, sure, makes sense22:19
JayFkernel command line parameters are great way to sneak config in :D 22:19
clarkb"Instead, we will turn the fips=1 kernel command line flag into the single source of truth for whether FIPS mode is enabled. We have already removed or will remove some of the separate knobs and instead automate them to follow the single source of truth"22:19
sean-k-mooneyso the problem thye are tryign to solve i think is makeing sure when the intall isntall is happeing you dont install anyting that is not fips compatiable at any point22:20
JayFdepends on the context that quote comes from; lots of distros control userspace tooling across the entire distro by respecting something in /proc/cmdline 22:20
clarkbI read that as openssl, openssh, etc etc etc need to check the kernel fips mode to know if they are in fips mode22:20
JayFoh I see what you mean22:21
JayFand I suppose they weren't as nice as IPA to provide an override :( 22:21
sean-k-mooneymaybe22:21
clarkb"The system-wide cryptographic policy will automatically follow the fips=1 kernel command line flag to remove the need for the manual switch previously performed by fips-mode-setup"22:21
clarkbanyway my point is mostly that this goes beyond "just build a fips image from scratch" because I suspect to bootstrap that we may already need some amount of fips enablement22:22
JayFwhich IRL might just be someone inserting a fips-y installer cd22:22
clarkbat least to do it properly and avoid the problems listed in that post like generating keys using non fips mode22:22
JayFbut in the cloud when you wanna get an image is a tough bootstrap problem22:22
sean-k-mooneywhat would be ideal woudl be if centos/rocky/alma had a fips cloud image22:23
sean-k-mooneythat dib could just use as a base22:23
clarkboh ok there is some fancy bind mount thing going on too that maybe dib/bootc could also provide to address that problem22:23
JonathanWright[m]sean-k-mooney: That's doable ya know ;)22:23
clarkbfor the record I do not want to have fips images in opendev22:23
JayFProblem successfully delegated, run before JonathanWright[m] changes his mind! ;) 22:24
JayFclarkb: tbh I think I agree with fungi's stance that it's ... weird that we test fips mode22:24
sean-k-mooneyclarkb: ya it is a bit of a weird edgecase22:24
JonathanWright[m]haha. seems like spotz should've invited me here sooner22:24
JayFgiven it's a US-specific requirement in an international community22:24
JonathanWright[m]I can see a FIPS image being quite useful so I bet there's other demand as well22:24
sean-k-mooneysince if we add a fips oen then whatever the eu one is will be next ectra22:24
JayFbingo22:24
clarkbI'm more than happy to have people enable fips inside of the test environment however that makes sense. But its a lot of work to manage the images we already have and doubling their number for one feature is not reasonable imo22:25
sean-k-mooneyJonathanWright[m]: is almalinux joining the list of images in opendev ci?22:26
sean-k-mooneythe x86_64 v2 build target is a very interesting capablity that alma has over both c10s and rocky 1022:27
JonathanWright[m]Honestly that's been on my list for...a long time.  I've been away from openstack for quite some time and I need to get re-familiar with things or ideally, find someone within the alma community that'd be willing to step up to the plate to help with alma/openstack stuff.22:28
JonathanWright[m]v2 has been far more popular than even we expected, it's been very fun to see the fanfare around it when people realize they can upgrade to 10 on their old hardware.22:29
sean-k-mooneyya i belive most of my dev hardware cant actully run v3.22:30
clarkbmy personal fileserver cannot v322:30
JonathanWright[m]Well, you now have a solution to get to 10, including EPEL :)22:30
sean-k-mooneyi mean i normally use debian/ubuntu but ill need to retrie that hardwar or use alma or debian going forward22:30
spotz[m]Jonathan Wright: Hey you never asked:) And I just assume all my friends know each other:)22:55
spotz[m]You should come to Summit, there I invited you some place:)22:56
JonathanWright[m]Summit is so $$$22:56
JonathanWright[m]Wait, what summit?22:57
spotz[m]No the travel to Paris and hotel are $$$, tickets are $:)22:57
JonathanWright[m]My brain immediately when to RH summit, but I'm thinking there must be an openstack summit or something22:57
spotz[m]Welcome to an OpenStack channel, Summit means OpenInfra Summit:)22:57
JonathanWright[m]Oh that one!  I think my wife my murder me if I go to Paris without her22:58
JonathanWright[m]Oct 17-19!  Ooof bit late now but maybe next year!22:58
spotz[m]Not sure where/when we'll be next year but will let you know22:58
JonathanWright[m]If it's in the US I can almost guarantee I could swing it.  Europe just let me know and it will depend on what other travel we have booked up.22:59
JayFWhere in US are you if you don't mind the ask?23:04

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!