15:00:09 <pjgeorg> #startmeeting Kmods SIG
15:00:09 <centbot> Meeting started Mon Aug 23 15:00:09 2021 UTC.  The chair is pjgeorg. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:09 <centbot> Useful Commands: #action #agreed #help #info #idea #link #topic.
15:00:21 <pjgeorg> #chair pjgeorg billings
15:00:21 <centbot> Current chairs: billings pjgeorg
15:01:07 <pjgeorg> #topic Open issues
15:05:19 <pjgeorg> billings: Are you present?
15:05:59 <billings> pjgeorg: sorry, was on conf call
15:06:15 <billings> I'm here now
15:06:25 <pjgeorg> No problem.
15:06:37 <pjgeorg> Do you have any open issues to discuss?
15:06:53 <billings> No.
15:07:02 <billings> I've been super busy in my last couple weeks at this job
15:07:53 <pjgeorg> I can understand that.
15:08:15 <pjgeorg> Maybe just a short update from my side.
15:08:53 <billings> Thanks
15:09:33 <pjgeorg> Concerning the talk for the Dojo: I haven't really worked on it due to time constraints. Concerning you are currently busy, shall I take care of this and discuss with you once I have something to show?
15:09:55 <billings> Did we submit a paper?
15:09:59 <billings> or talk, rather
15:10:26 <billings> I don't remember if we actually had an accepted talk to give
15:10:29 <pjgeorg> Not yet, but we have been asked to give a talk.
15:10:46 <billings> Oh, ok.
15:12:14 <pjgeorg> But yes, we haven't submitted anything yet. Deadline is in two weeks.
15:13:03 <billings> Ok.  Do you want to collaborate with some google slides?
15:13:10 <billings> I'd be happy to help with that
15:13:39 <billings> Might be nice to have a template we can use, does the CentOS devel/board have a template we can use?
15:14:31 <pjgeorg> I think last time we agreed to look into different services for collaboration. google docs is fine for me.
15:14:44 <pjgeorg> I do not know about any template.
15:15:45 <pjgeorg> I'll set something up in google docs in the next days and share the link with you, ok?
15:15:52 <billings> ok.  I was looking at a markdown slides projects: https://gist.github.com/johnloy/27dd124ad40e210e91c70dd1c24ac8c8
15:15:57 <billings> Sure.
15:16:12 <billings> but none of them jumped out as something I'm thrilled with
15:16:28 <billings> (other than pandoc if you like that kind of thing)
15:17:23 <pjgeorg> HackMD is the only one on this list I have been working with so far.
15:17:28 <billings> ok
15:17:45 <billings> I'll ask the centos-devel if there are slide templates we can use
15:17:56 <billings> it might be one of those things that will come along with the new logo presentation
15:17:57 <pjgeorg> Ok.
15:18:55 <billings> A LOOONG time ago I used latex to do transparencies
15:19:07 <pjgeorg> I still do :)
15:19:09 <billings> (that tells you how long ago it was)
15:20:27 <pjgeorg> Anything else to discuss about the talk now?
15:20:35 <billings> No, I don't think so.
15:20:53 <billings> will we have another meeting before it's due?
15:21:07 <billings> if not, we should schedule a meeting to make sure it is ready
15:21:57 <pjgeorg> This is the last regular meeting. The deadline is on Monday 00:01 UTC.
15:22:18 <pjgeorg> But the talk has not to be ready before the deadline.
15:22:35 <pjgeorg> We need to agree on a speaker, title and finish an abstract by then.
15:22:57 <billings> ok
15:23:37 <pjgeorg> Title is easy: "Kmods SIG Introduction"
15:24:13 <billings> We can probably lift a lot of the text from the SIG creation
15:24:25 <pjgeorg> Yes
15:26:20 <pjgeorg> I'll prepare an abstract this week and add it to https://pagure.io/centos-sig-kmods/sig/issue/10 for discussion.
15:27:49 <pjgeorg> To other topics I have been working on since the last meeting:
15:29:02 <pjgeorg> I had a look at DUDs and how we might be able to provide these in an automated way, i.e. using CBS. Not really anything to discuss yet. Probably have to ask infra about ideas what one could do.
15:29:11 <billings> excellent
15:29:19 <billings> I'm curious how those are built
15:29:59 <pjgeorg> Building is quite easy. E.g., see https://access.redhat.com/solutions/6078541
15:30:00 <jcpunk> #link https://anaconda-installer.readthedocs.io/en/latest/driverdisc.html#ddv3-structure
15:30:40 <pjgeorg> I'm just not sure one can something like this in CBS/koji. Afaik koji only supports images created via kickstart files.
15:32:42 <billings> too bad pagure doesn't have the equivalent of github actions, we could kick off a build that way
15:33:27 <billings> maybe create an ansible playbook that generates it from a kmod package
15:34:42 <pjgeorg> For me there are currently 3 open questions:
15:34:51 <pjgeorg> 1. How to create DUDs in an automated way
15:35:12 <pjgeorg> 2. Do we want to create one DUD for every kmod or one DUD containing all kmods?
15:35:27 <pjgeorg> 3. How to get the DUD(s) on mirror.centos.org
15:35:54 <billings> Good questions.
15:36:15 <billings> I get the impression that the elrepo guys hand-build their DUDs against the installer kernel
15:36:38 <billings> not all kmods we create make sense to be a DUD, too
15:37:23 <pjgeorg> I do not know elrepo's infrastructure. No idea how they do it.
15:37:35 <billings> for example, if I chose to create a v4l2loopback kmod for 8-stream, we probably don't need it during install.
15:37:38 <pjgeorg> I think the most critical is No 3 and it can only be answered by infra.
15:38:30 <pjgeorg> That's correct. Probably only stuff related to network and disks (scsi, p/sata, etc.).
15:39:28 * billings nods
15:39:35 <pjgeorg> Deciding which kmods to include is the easy part at the end.
15:40:22 <billings> I think that DUDs are typically only built for each installer kernel only, so it's not something that needs as much automation
15:40:55 <pjgeorg> For CentOS Stream the installer iso is afaik re-created for every compose.
15:41:34 <billings> ah
15:41:56 <billings> then we will need to tie into that automation somehow
15:43:04 <pjgeorg> I'll open an issue with centos-infra to get their opinion. Probably the best way to move forward.
15:47:59 <pjgeorg> Apart from that I recently wanted to check how complicated it is to port a kernel module from a newer upstream kernel release back to EL8. For this test I used the new exfat module. Turned out to be easier than I first expected. However this is a rather simple module. For more complex kmods, e.g. btrfs, it is obviously a lot more work.
15:48:28 <pjgeorg> Do you think there is any need for a kmod-exfat package? If yes, I can push it.
15:49:43 <billings> it really depends on what other subsystems it relies on, and if it has changed significantly since
15:50:10 <billings> for example, the upstream kafs module relies on fscache, and it was rewritten since 4.18 so I can't trivially backport it
15:51:31 <pjgeorg> Yep, it really depends on the particular module.
15:51:33 <billings> however, I also have to cherry-pick a change from upstream kernels to get it to build on the CentOS Stream kernel, because yet another part of the kernel changed (in patches) that I need to address.
15:53:33 <pjgeorg> That's true for many inkernel modules. I hope that this process will be easier for el9. For el9 we should get a git history of the kernel sources, for el8 we only have a tarball.
15:55:16 <billings> Ok.  5 til
15:55:32 <billings> Lets just create some action items
15:55:50 <billings> #action billings will ask about slide templates
15:56:31 <pjgeorg> #action pjgeorg Setup slides in google Docs and share link
15:56:52 <pjgeorg> #action pjgeorg Write abstract for Dojo talk
15:57:15 <pjgeorg> #action pjgeorg Open issue with centos-infra concerning DUDs
15:57:38 <pjgeorg> Anything missing?
15:58:07 <billings> looking through history
15:58:29 <billings> looks good
15:59:45 <pjgeorg> Good, then we are done for today. Have a nice day!
16:00:07 <pjgeorg> #endmeeting