13:01:53 <jki> #startmeeting CIP IRC weekly meeting 13:01:53 <collab-meetbot> Meeting started Thu Aug 7 13:01:53 2025 UTC and is due to finish in 60 minutes. The chair is jki. Information about MeetBot at http://wiki.debian.org/MeetBot. 13:01:53 <collab-meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 13:01:53 <collab-meetbot> The meeting name has been set to 'cip_irc_weekly_meeting' 13:02:01 <jki> #topic AI review 13:02:16 <jki> none on my list, and I found none in the past weeks 13:02:22 <jki> 5 13:02:24 <jki> 4 13:02:25 <jki> 3 13:02:27 <jki> 2 13:02:28 <jki> 1 13:02:31 <jki> #topic Kernel maintenance updates 13:02:43 <masami> This week reported 4 new CVEs and 8 updated CVEs. 13:02:44 <pave1> I'm reviewing 6.12.40 and 41. 13:02:45 <uli> i'm preparing 4.19 13:02:57 <iwamatsu__> I reviewed 6.12.40 and 41. 13:04:11 <jki> anything to add? 13:04:24 <jki> 5 13:04:25 <jki> 4 13:04:27 <jki> 3 13:04:29 <jki> 2 13:04:30 <jki> 1 13:04:32 <jki> #topic Kernel release status 13:04:42 <jki> all lights green right now 13:05:25 <jki> any issues upcoming? 13:05:36 <jki> 5 13:05:38 <jki> 4 13:05:40 <jki> 3 13:05:42 <jki> 2 13:05:44 <jki> 1 13:05:46 <jki> #topic Kernel testing 13:06:04 <patersonc> Arisu-san has been continuing to get our boards added to the new KernelCI 13:06:22 <arisut> I could send test to the cip boards on lava, currently going on improving the work 13:06:25 <arisut> https://lava.ciplatform.org/scheduler/alljobs?page=1&length=25&search=kci-staging#table 13:06:34 <arisut> from KernelCI 13:06:34 <patersonc> Thanks :) 13:06:57 <patersonc> It looks like not all boards boot when using the merged config 13:07:29 <patersonc> arisut please ping me if there's anything specific you'd like me to investigate 13:07:38 <jki> some examples at hand? 13:07:52 <jki> and we do have configs for them in our repo that used to boot? 13:07:55 <patersonc> jki: https://lava.ciplatform.org/scheduler/job/1298034 13:08:02 <arisut> sure, currently I just need to finalize the PR and be sure that what we expect is what we get 13:08:14 <patersonc> arisut: Thanks 13:09:12 <patersonc> Does anyone know if the 6.12 merged CIP config is meant to work with de0-nano? 13:09:54 <patersonc> Maybe we don't need to support it, depending on the blank cell in https://wiki.linuxfoundation.org/civilinfrastructureplatform/ciptesting/cipreferencehardware#cip_reference_hardware 13:09:55 <jki> likely not yet 13:10:11 <jki> if there is no nano-soc config for 6.12 uploaded 13:10:23 <iwamatsu__> not yet 13:10:34 <jki> expected error 13:10:35 <patersonc> Maybe Arisu-san we skip nano-soc and iwg20m for 6.12? 13:11:12 <arisut> actually we could support it, one of the problem is that currently KernelCI is mixing up configurations 13:11:15 <jki> we can simply fix the config, I would say 13:11:37 <arisut> patersonc, I don't mind to push as is and improve it later 13:11:42 <arisut> same for riscv 13:11:47 <patersonc> Again it's a question of do we build/test boards not listed as reference h/w ? 13:12:23 <jki> oh, we didn't decide whether to keep the nanosoc in support, right 13:12:25 <arisut> nano is listed 13:12:30 <arisut> for v6.1 13:12:41 <arisut> and 4.19/5.10 13:13:08 <patersonc> Sure. Was it working with those branches? 13:13:26 <arisut> we currently don't know, actually 13:13:36 <arisut> as I said my PR is still in progress 13:13:41 <arisut> is not yet merged 13:14:12 <patersonc> Sure. I'd remove the not-working ones from your PR for now. Then we can add more as we test they work? 13:14:19 <arisut> and configurations are not always used as expected 13:14:47 <patersonc> arisut: Should I talk to Denys about getting me set up so I can push/test with staging? Or would it clash with your work? 13:15:09 <arisut> patersonc, currently my PR is not stable to be merged. we are still not applying the right configurations to the right boards 13:15:54 <arisut> and some configurations are not yet merged in the *-cip sub configurations 13:17:49 <patersonc> Question for the kernel team - are we currently maintaining the in-tree defconfigs for each arch? 13:18:32 <jki> which defconfigs? those inside the kernel tree? they are per-arch, obviously 13:18:55 <pave1> That's iwamatsu-san question, but I don't believe we touch those configs. 13:19:39 <pave1> ...or actually... 13:19:58 <pave1> ...we are getting patches to arch/arm64/configs/defconfig when the driver is merged, etc, 13:20:04 <pave1> and we merge those patches. 13:20:14 <pave1> I guess you could call that "maintaining" :-) 13:20:39 <jki> yeah, would be strange to see normal stable fixes touching those defconfigs 13:21:10 <patersonc> Okay. Then I'll aim to include defconfig builds in the kernelci setup 13:21:42 <patersonc> I think it's worth making sure they at least build - as I assume a lot of users would use them as a first step 13:21:51 <pave1> Actually, that happens, too. 72ce323e17d0f6a6d586cdded4dc38cdcba31b6d . It should not, but when stable 13:22:18 <pave1> team picks up dependencies for a fix, they tend to pick lot of interesting stuff. 13:24:17 <arisut> My PR is currently not finished, at this time working on a unfinished PR would clash with my work as the code could change later on 13:24:53 <patersonc> Sure 13:26:10 <pave1> patersonc: The way I see renesas updates .. I'd say their goal is for defconfig to boot / work on that hardware. 13:26:35 <patersonc> Yea 13:27:05 <pave1> So.. yes, I'd say testing it builds an boots would be useful. 13:27:12 <patersonc> Okay 13:28:39 <arisut> another question is if we want also to test mainline kernel on the cip boards or only cip kernel? 13:29:00 <arisut> as sometime mainline kernel could be useful as reference 13:29:06 <pave1> Older -stable kernels may not work on all cip boards. 13:29:21 <pave1> Where it works, it is useful to make sure it keeps working. 13:29:27 <pave1> So that we catch bugs early. 13:30:09 <patersonc> Agreed. Should that can be part of the standard Maestro though rather than the CIP yaml files Arisu? 13:30:29 <arisut> sure 13:30:57 <patersonc> I'm happy for the boards to be used for any KernelCI testing - part of CIP's contribution etc. 13:31:24 <arisut> actually I'm not sure what is better, I think they are still cip board 13:31:48 <arisut> so having them in one cip file is still better, if we want to do some changes 13:32:04 <jki> can we prioritize board usage? 13:32:23 <arisut> maybe I don't remember sorry 13:32:36 <arisut> but that is a good question 13:32:43 <jki> we should give boards into general testing, but if they become "overused", cip should be first 13:32:52 <arisut> right 13:33:25 <arisut> will check if is possible 13:33:37 <patersonc> Sure. We can investigate/add support once we start having capacity issues :) 13:34:56 <jki> anything else on testing? 13:35:23 <jki> 5 13:35:25 <jki> 4 13:35:26 <jki> 3 13:35:28 <jki> 2 13:35:30 <jki> 1 13:35:32 <jki> #topic AOB 13:35:54 <patersonc> o/ I have an AOB about CVEs 13:36:18 <jki> go ahead 13:36:37 <patersonc> Do we have a process for monitoring the CVEs that get created to see if they should be applied to our self-maintained SLTS kernels? 13:36:51 <patersonc> I assume that stable will sort out the LTS based kernels? 13:37:20 <patersonc> But is someone looking at each CVE "fix" and seeing if it should be backported to 4.4 and 4.19? 13:37:25 <pave1> Not sure that assumption is correct :-) 13:37:34 <patersonc> pave1: Sure :P 13:38:22 <pave1> When patch fails to apply to our -cip kernels, we take a look if it looks serious. 13:38:39 <pave1> CVEs is just another ID for patches. 13:38:48 <pave1> So yes, we kind of do that. 13:38:58 <iwamatsu__> I checked CVEs sometime, and backport. 13:39:14 <patersonc> So we don't have a mechanism as part of cip-kernel-sec? 13:39:43 <patersonc> I ask, because a couple were flagged to me recently 13:39:45 <pave1> But stable team may not backport patch if it looks too complex or does not look like serious-enough problem. 13:40:42 <patersonc> Here's an example: https://gitlab.com/cip-project/cip-kernel/cip-kernel-sec/-/blob/master/issues/CVE-2025-21917.yml. It's been labelled as introduced in v3.0-rc1, and "fixed" in CIP 5.10 onwards. But no "fix" in 4.4 and 4.19 13:40:59 <iwamatsu__> We are still stuck with CVE management, and we don't have a process for how to deal with unfixed CVEs. 13:41:25 <jki> we tried to filter at least irrelevant CVEs based on our configs 13:41:51 <jki> irrelevant = not used in CIP configurations, thus not officially supported 13:42:16 <pave1> I'm sure you can find many more such issue. Greg ~ automatically creates CVE for each stable patch. 13:42:16 <patersonc> We have these yaml files - is there a way to compare the "introduced-by" with the "fixed-version" fields? 13:42:51 <patersonc> iwamatsu__: Sure. But this example hasn't been marked as "irrelevant" 13:43:02 <pave1> For the CVE-2025-21917, look at the description, and note that it does not describe anything. 13:43:27 <pave1> So yes, you can (probably) run some kind of script. 13:43:33 <pave1> You'll get 1000+ results. 13:43:38 <pave1> Then you can laugh... or cry :-). 13:44:36 <jki> tooling and maintenance capacity need to be increased to track CVEs on "paper", with VEX output or whatever 13:45:07 <iwamatsu__> I thought it might be necessary to keep a record of checking for unfixed CVEs as CIP. 13:45:35 <iwamatsu__> I am checking CVEs via web interface 13:45:46 <jki> we have the KNOWN-BUGS file, but that is only for prominent ones 13:46:11 <pave1> Yeah, putting every single CVE would make that file completely useless. 13:46:27 <pave1> Yeah, putting every single CVE there would make that file completely useless. 13:47:02 <patersonc> Maybe we need to first work out which CVEs aren't fixed, then work out how many could be easily. 13:47:09 <jki> we should probably discuss the existing process and possible enhancements/costs during the extended meetup 13:47:24 <patersonc> jki: Good shout 13:47:25 <iwamatsu__> Can we use cip-kernel-sec? 13:47:33 <patersonc> Makes sense to me 13:47:43 <jki> there is no "just do X" or "just spend some extra hour" to address this 13:47:58 <pave1> If you want to see log of various patches not backported 13:48:06 <pave1> ...which is basically CVEs... 13:48:21 <pave1> ...you can take a look at v4.4.org and v4.19.org 13:48:35 <pave1> in git@gitlab.com:cip-project/cip-kernel/lts-commit-list.git repository. 13:48:53 <pave1> That's where the work is recorded. 13:49:05 <arisut> makes sense 13:49:22 <pave1> There's aproximately 9000 patches not applied to 4.4, and 1500 not applied to 4.19. 13:50:42 <pave1> (that was wc, so its less than that, but you get the idea). 13:51:14 <pave1> (7000 and 1000). 13:54:19 <jki> let me try to prepare some discussion about that for Amsterdam 13:54:38 <jki> do we have some material to start from? 13:55:55 <patersonc> uli did you present a bit about the CVE tools a while back? 13:56:03 <pave1> I did take 9 randomly selected CVEs and tried to review them at one point. 13:56:25 <uli> patersonc: not specifically, it was about the maintenance process in general 13:56:36 <patersonc> okay 13:56:37 <uli> actually only mentioned cves when somebody asked about it :) 13:57:05 <jki> yeah, these questions will increase... 13:57:25 <jki> anyway, will share with you upfront for alignment 13:57:34 <jki> anything else for today? 13:57:43 <iwamatsu__> I'm on vacation next week. 13:58:23 <jki> ok 13:58:28 <jki> enjoy :) 13:58:38 <iwamatsu__> :-) 13:58:52 <pave1> I decided 3 were not security issue, one could not be determined in reasonable time, 2 were "ok, maybe that should be fixed", rest was "that's really low severity". 14:00:13 <jki> ok, let's close... 14:00:16 <patersonc> For CVEs we've spotted like this, should we backport the patch and send to cip-dev? How does cip-kernel-sec then get updated? 14:00:53 <pave1> patersonc: If you believe you see a real security issue that's "bad" and want it fixed... 14:01:01 <pave1> yes, backport, cip-dev and uli. 14:02:00 <patersonc> Okay 14:02:04 <pave1> Please have real description of a bug ("uid 123 can echo baz into /sys/foo to crash the system"). 14:02:07 <masami> if patch is merged into git repo, cip-kernel-sec can be updated. 14:02:26 <patersonc> Thanks masami 14:02:42 <jki> great! but I suspect too much manual work still ;) 14:02:52 <jki> ok... anything else? 14:03:10 <jki> 5 14:03:11 <jki> 4 14:03:13 <jki> 3 14:03:15 <jki> 2 14:03:17 <jki> 1 14:03:18 <jki> #endmeeting