15:11:10 <pjgeorg> #startmeeting Kmods SIG
15:11:10 <centbot> Meeting started Mon Jun 28 15:11:10 2021 UTC.  The chair is pjgeorg. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:11:10 <centbot> Useful Commands: #action #agreed #help #info #idea #link #topic.
15:11:56 <pjgeorg> Starting a little late. Anyway, our first meeting this week. I only have a few points to go through. So meeting might be short.
15:12:20 <pjgeorg> In case you have any topics just let me know.
15:12:28 <pjgeorg> #topic Election
15:12:36 <pjgeorg> According to the SIG description there are two co-chairs. Both are to be elected once a year.
15:13:03 <pjgeorg> Currently I'm listed as initial chair, but have never been elected. This I suggest we just do a (fast) election.
15:13:12 <billings> I nominate pjgeorg
15:13:25 <jcpunk> I nominate billings and pjgeorge
15:13:58 <billings> sounds good to me. :)
15:14:04 <jcpunk> :)
15:15:09 <pjgeorg> Ok, so officially I think we have to vote one after the other. So votes for first chair (there's no difference between first and second)
15:15:29 <billings> is it a public vote?
15:15:29 <jcpunk> I vote +1 for pjgeorge
15:15:37 <jcpunk> I vote +1 for billings
15:15:53 <jcpunk> For our initial election I think public is fine
15:15:56 <billings> Ok
15:16:05 <pjgeorg> Yes, public is fine.
15:16:06 <billings> yes, I also vote for pjeorge.
15:16:23 <billings> for first chair
15:16:38 <billings> er, pjgeorg
15:16:46 <jcpunk> (I voted in first/second order)
15:16:59 <pjgeorg> I vote +1 for billings (as second chair)
15:17:20 <pjgeorg> So we are done. Me and billings are now officially elected as co-chairs.
15:17:27 <jcpunk> :)
15:17:35 <pjgeorg> #chair pjgeorg billings
15:17:35 <centbot> Current chairs: billings pjgeorg
15:18:15 <pjgeorg> #agreed pjgeorg and billings have been elected as co-chairs
15:18:33 <pjgeorg> Anything more to say about this topic?
15:18:41 <billings> Not for me.
15:18:45 <jcpunk> I'm thrilled to have you both!
15:19:03 <billings> June 2022 will be the next election.
15:19:20 <jcpunk> SIG chairs are invited to view the monthly CentOS Board meeting, I'll get an invite sent your way
15:19:33 <billings> Thank you
15:19:42 <pjgeorg> Thanks.
15:20:03 <pjgeorg> So we can move on to the next topic.
15:20:37 <pjgeorg> #topic Quartely report
15:21:13 <pjgeorg> We have been asked to send in a quarterly SIG report by July 1st.
15:21:47 <pjgeorg> As there is not yet much to report I suggest that I just write a short report about the bootup progress and our first meeting.
15:21:58 <billings> Well, I suspect having just started, they probably will be fine with the fact that we've had our first meeting and elected chairs?
15:22:09 <billings> ok.
15:22:30 <jcpunk> I'd also add a note on the big issue we're looking at - how to structure the packages
15:22:41 <billings> https://pagure.io/centos-sig-kmods/sig/issue/4
15:22:50 <pjgeorg> Yes, will mention this issue there.
15:23:06 <billings> And the licensing question
15:23:30 <billings> (re: signed kmods)
15:23:53 <pjgeorg> Probably easier: I'll write a draft and post it at pagure by tomorrow. You can then comment and add topics I missed.
15:23:59 <billings> Thanks.
15:24:02 <jcpunk> +1
15:25:21 <pjgeorg> #action pjgeorg to post draft for quarterly report at pagure by tomorrow
15:25:38 <jcpunk> Do you folks have a bugzilla.redhat.com account?
15:25:43 <pjgeorg> Yes
15:25:50 <billings> I use my Fedora account to post there.
15:26:21 <jcpunk> excellent, feel free to add me to the CC list on any bugs you think I should know about
15:27:04 <billings> Will do.
15:27:46 <pjgeorg> Same. Just to avoid confusion, I think you can not find me there via my nickname pjgeorg, but as peter.georg
15:28:19 <pjgeorg> #topic Open issues
15:28:47 <billings> Now that we have build tags, we need the centos-release-kmods package, right?
15:29:00 <jcpunk> yep
15:29:06 <pjgeorg> Yes, there is an open issue for this.
15:29:20 <billings> https://pagure.io/centos-sig-kmods/sig/issue/3
15:29:50 <pjgeorg> Is already assigned to me, I'll take care of it.
15:30:05 <billings> You're right.  Thanks
15:30:45 <pjgeorg> Probably not time sensitive as we'll probably need some discussion in https://pagure.io/centos-sig-kmods/sig/issue/4 before we actually build our first package in CBS.
15:31:04 <jcpunk> Indeed
15:32:07 <jcpunk> I'll see if I can get some elrepo folks to look and provide some thoughts
15:32:32 <pjgeorg> That'd be great. I still need to sign up to their elrepo-devel mailing list.
15:32:47 <billings> in the centos-sig-kmods pagure page, I see a isci-kmod and a kmod-isci repo. Are they the same thing?
15:33:25 <pjgeorg> No. These are two different approaches how to package kmods.
15:33:29 <billings> ok.
15:33:46 <pjgeorg> isci-driver and isci-kmod are inspired by the RPMFusion kmods2 Guidelines.
15:34:04 <pjgeorg> The kmod-isci package is done similar to how Red Hat packaged kmod-kvdo
15:34:45 <pjgeorg> Afaik the last one is also similar to how ELRepo is organizing their packages.
15:34:47 <billings> oh I see
15:35:24 <pjgeorg> I just posted both approaches so we can compare and then decide which is the better approach for us.
15:35:48 * arrfab pops-in, curious
15:36:07 <billings> isci-kmod creates a kmod-isci metapackage and a kmod-isci-%{kernel_version}, and hte metapackage requires the versioned kmod
15:36:54 <pjgeorg> Yes, see here for a detailed description: https://rpmfusion.org/Packaging/KernelModules/Kmods2
15:37:19 <billings> *nod*
15:38:01 <billings> so, I've been building OpenAFS kmod packages for a while, and I jump through some crazy hoops to make it so it can build so it automatically uses the kernel version available in the buildroot
15:38:34 <billings> its mostly broken in Fedora because it relies on kernel-headers version/release == kernel-devel version/release
15:39:25 <billings> I basically have this eval'd at the beginning of the spec file: https://github.com/jsbillings/openafs/blob/master/find-installed-kversion.sh
15:39:38 <pjgeorg> Can we find your spec file somewhere online? Might be interesting to have a look at it.
15:39:54 <billings> https://github.com/jsbillings/openafs/blob/master/openafs-kmod.spec
15:40:13 <billings> see line 43 and 44
15:41:17 <jcpunk> Has that one drifted much from the one in the openafs sources?
15:41:32 <billings> I try to incorporate from the openafs sources when I can
15:41:49 <billings> but my kmod package separates out the kernel module from the rest of the build (which is in the openafs.spec in the same repo)
15:42:05 <billings> I do this because I'm constantly rebuilding the kmod, but I don't need to rebuild the whole source each time
15:42:30 <billings> this will be even more important now that IBM has relicensed the base code used in OpenAFS 1.0.0 to GPL
15:42:44 <jcpunk> pjgeorge fyi this is the other spec we're talking about http://git.openafs.org/?p=openafs.git;a=tree;f=src/packaging/RedHat;h=467af0503e5d6cc07111f2f8221dc95b69b61d18;hb=HEAD
15:42:56 <billings> yup
15:44:00 <jcpunk> the kmod-tool.sh approach is interesting for helping consistency
15:44:09 <billings> although I only build against the openaf-stable-1_8_x head
15:44:25 <billings> the master branch is largely untested in production
15:45:12 <billings> http://git.openafs.org/?p=openafs.git;a=tree;f=src/packaging/RedHat;h=306faa1a3586ad08ec1511eb9a6a5e80508acf1c;hb=refs/heads/openafs-stable-1_8_x is the head that I was talking about
15:49:14 <pjgeorg> RPMFusion is using kmodtool. The example I posted on pagure (isci-*) is not using kmodtool as I wanted to do everything explicitly. Always helps me to better understand what's going on.
15:49:23 <billings> Right.
15:49:34 <billings> Another approach in the OpenAFS code is dkms
15:49:54 <jcpunk> Explicit is good, implicit tends to become magic, tends to become unknown over time
15:50:01 <billings> (which is part of the upstream openafs code, and in my non-kmod spec)
15:50:22 <billings> and also, there's the akmods method
15:50:40 <jcpunk> I'm not too familiar with akmods, I'll have to read up on those some more
15:50:56 <jcpunk> I looked at them a few years back.... but that is long gone
15:50:56 <billings> https://rpmfusion.org/Packaging/KernelModules/Akmods
15:51:37 <jcpunk> I tend to like DKMS in theory, but not so much in practice.  It doesn't work well at scale with Secure Boot
15:51:42 <billings> akmods is a bit akin to dkms except it creates a kmod package
15:52:02 <billings> and yes, it won't be signed
15:52:12 <pjgeorg> akmods and dkms are useful for distributions like Fedora or upstream projects that do not want to rebuild for each kernel release. Not sure this is what we want to do in this SIG?
15:52:41 <billings> I know that the ZFS folks care, but I *really* don't want to build zfs kmods in this SIG if I can
15:52:43 <jcpunk> Personally, I'd rather have "ready to go" RPMs for folks
15:53:07 <pjgeorg> +1
15:53:29 <billings> Yes. I think at least to start, lets focus on kmods we can provide
15:53:38 <pjgeorg> With dkms packages you never know whether they actually work or you'll just end up with a weird compiler error.
15:53:48 <billings> Agreed.
15:54:01 <jcpunk> Indeed, +1 for "just GPL for now" and "sometimes you get weird complier errors with dkms"
15:54:28 <billings> And my experience with OpenAFS is that the EL kernels do change mid-stream enough to break builds
15:54:38 <jcpunk> eventually I'd like to get some other stuff in here, but we've got enough to tackle thus far
15:55:01 <jcpunk> I can confirm some suprises in OpenAFS and EL kmods from SL
15:55:40 <billings> Almost at the end of the hour
15:55:54 <pjgeorg> #agreed The SIG will work on GPL only for now
15:56:14 <pjgeorg> #agreed The SIG will work on kmods only for now (no dkms/akmods)
15:56:51 <billings> I do hope to build OpenAFS kmods too, using a standard approach decided by the SIG
15:58:23 <pjgeorg> This means https://pagure.io/centos-sig-kmods/sig/issue/4 is what we should work on.
16:00:27 <pjgeorg> This will probably still take some time and discussion. But we'll get there.
16:00:49 <pjgeorg> Any other open topics?
16:01:05 <jcpunk> nothing comes to mind
16:01:07 <billings> I can't think of anything
16:01:50 <pjgeorg> Ok, then I think I can end this meeting and we'll meet again in two weeks.
16:02:09 <billings> Thanks for running the meeting!
16:02:13 <jcpunk> +1
16:02:44 <pjgeorg> Thanks for attending!
16:02:50 <pjgeorg> #endmeeting