20:00:04 <carnil> #startmeeting 20:00:04 <MeetBot> Meeting started Wed Oct 29 20:00:04 2025 UTC. The chair is carnil. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:00:04 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic. 20:00:16 <waldi> hi 20:00:39 <carnil> #chair carnil bwh ukleinek waldi 20:00:39 <MeetBot> Current chairs: bwh carnil ukleinek waldi 20:01:01 <carnil> On request I have added on top two additional agenda items to discuss things before the buglist 20:01:14 <carnil> is there anything else someone wants to bring in? 20:01:23 <waldi> nope 20:01:26 * ukleinek doesn't 20:01:37 <carnil> let'start then, I guess two are enough :) 20:01:42 <carnil> #topic 20:01:53 <carnil> #topic future of iptables-legacy 20:02:00 * ukleinek inserts a "hi" a few lines above 20:02:00 <carnil> waldi: it's your stage 20:02:05 <waldi> yeah 20:02:44 * ukleinek was surprised to learn that yocto still defaults to iptables-legacy (and most people seem to stick to that, because there was a build failure if you explicitly select nft) 20:02:53 <waldi> this is the same as the architecture baseline thing. do we want to support everything until the bitter end or should we do a planned exit 20:03:13 <waldi> and iptables legacy is now hidden behind even another set of settings 20:03:45 <ukleinek> iptables legacy is different from the iptables compatible binaries that do nft behind the curtain, right? 20:04:01 <waldi> iptables-legacy vs iptables-nft 20:04:23 * ukleinek nods and would support to let iptables-legacy die for forky 20:04:28 <waldi> iptables-legacy contains a whole bunch of kernel code 20:05:03 * carnil wonders if https://codesearch.debian.net/search?q=iptables-legacy&literal=0 gives us a wide enough view where work is needed 20:05:07 <bwh> Hi, sorry I'm late 20:05:16 <carnil> hi bwh, no problem! 20:05:27 <carnil> bwh: we are discussing the iptables-legancy item 20:05:55 <ukleinek> huh, I'm surprised that there are explicit calls for iptables-legacy-* 20:05:59 <carnil> there are a lot of things in unstable which refer to iptables-legancy invocations apparently 20:06:23 <bwh> I think it would be reasonable to do an MBF for those 20:07:21 <bwh> But nftables may be so different that upstream will have trouble switching even within this release cycle. I don't know 20:07:26 <waldi> carnil: most look like fallbacks. but direct references to iptables-legacy will break many things already 20:07:32 <ukleinek> I wonder if there are any known incompatibilities that are relevant in practise 20:07:56 <ukleinek> bwh: deprecating iptables-nft would be a bigger challenge 20:08:22 <waldi> bwh: i don't think anyone proposed deprecating iptables(-nft) 20:08:23 <bwh> ukleinek: Right but we don't need to care about that 20:08:41 <ukleinek> there are even several matches in src:linux 20:08:57 <waldi> ukleinek: test code, for the legacy code paths i would assume 20:09:31 <ukleinek> probably right, they are all below tools/testing/selftests/ 20:09:52 <bwh> Someone should maybe take an action to find out what actually depends on this, then? 20:09:59 * ukleinek thinks we agree to deprecate iptables-legacy? 20:09:59 <waldi> i will do 20:10:04 <ukleinek> 👍 20:10:23 <waldi> i would propose we start the MBF, drop the legacy binaries ASAP and the kernel support after forky 20:10:51 <bwh> I think may depend on the scale of what needs fixing... 20:11:12 <ukleinek> ack, but declaring that as goal sounds right 20:11:14 <carnil> #action waldi tries to find out what still depends on iptables-legacy (not only as fallback), propose a plan for MBF and dropping legacy binaries ASAP (and kernel support after forky) 20:11:19 <carnil> ^^ waldi fine enough? 20:11:37 <waldi> yep 20:11:45 <carnil> thanks a lot for bringing it on the table 20:12:04 <carnil> next topic is proposed by santiago, not sure if he can join 20:12:13 <carnil> #topic Proposal about maintaining firmware-nonfree in old releases 20:12:16 <santiago> hi (I'm half here) 20:12:26 <santiago> thanks! I don't want to take too much of time of the meeting, so I'll try to be brief 20:12:32 <carnil> I pinged after our last meeting santiago and asked if he can post the proposal on the list. 20:12:39 <carnil> santiago: please go ahead :) 20:12:45 <santiago> security support for firmware-nonfree is difficult because of different reasons (documented in the email I sent): high risk of regressions, lack of information of updates, old kernels not able to load updated files, ... 20:12:53 <santiago> the idea is to introduce per-kernel firmware-nonfree packages, aiming to minimize regressions and to make sure uploaded firmware files can be loaded by target "target" kernel 20:13:00 <santiago> this of course introduces complexity, since it means maintaining an additional package, and not "simply" backporting from newer releases. but the latter is too challenging 20:13:12 <santiago> what does the kernel team think about this idea? does it make sense? 20:13:34 <santiago> and should the hypothetical firmware-nonfree-6.1 source package maintained under the kernel team umbrella? 20:13:57 <ukleinek> Is it really only about firmware-nonfree? 20:14:08 <bwh> The issue I have with this is that there isn't such a simple relationship between firmware-nonfree and kernel versions 20:14:27 <carnil> santiago: when I first heard the idea I found it might be more complex than the src:linux-X.Y case, because the firmware files all live in /lib/firmware/ and owuld need breaks,replaces relations 20:14:33 <bwh> yes 20:14:48 <santiago> that's true 20:15:10 * ukleinek imagines that tracking the set of firmwares used for a given kernel is difficult, each stable update might result in changes 20:15:25 <bwh> However: just thinking out loud, what we could do is install everything under /lib/firmware/6.1 and then have each kernel package create a symlink from /lib/firmware/<release> to 6.1 20:15:39 <waldi> bwh: can we change the loading path? 20:16:13 * ukleinek removes the line he was about to type that would result in the third one with that thought 20:16:23 <bwh> the kernel already has logic for loading from /lib/firmware/<release> (and it looks there first) so we should just need to create those as symlinks 20:16:37 <waldi> okay 20:17:24 * ukleinek doesn't understand the need for symlinks 20:17:44 <waldi> ukleinek: because we don't want a firmware-nonfree-6.1.2-arm64 20:17:49 <bwh> exactly 20:18:23 <ukleinek> ah, and the firmware loader only checks /lib/firmware/6.1.2-arm64 and not /lib/firmware/6.1 20:18:31 <bwh> yes 20:19:27 <waldi> so it would be possible to create per-version firmware packages. but what would it contain? 20:19:41 <ukleinek> and would firmware-non-free-6.1 contain all firmwares, or only those that got an update since firmware-nonfree/stable? 20:20:36 <bwh> To be decided, but I think it would be simplest to include all the same things that are in firmware-nonfree/bookworm, just with the package name and installation path change 20:21:10 <waldi> and this would then be updated to latest upstream? 20:21:36 <ukleinek> "We will also avoid backporting new upstream versions" 20:21:58 <santiago> sorry people, I need to go now :( (really sorry for just bringing the topic, and without able to contribute further to the meeting) 20:22:14 <ukleinek> santiago: then next week? 20:22:27 <bwh> I can spend some of my LTS time to try this out 20:22:31 <carnil> I think that would not really help, the newest version might contain firmware files which the kernel will not load, right? 20:22:34 <carnil> bwh: ok! 20:23:09 <carnil> bwh: it would be great if you can do that, and then maybe summarize/give an overview in one of the next meetings once you found time to do so? 20:23:17 <santiago> ukleinek, by the time of the kernel team meetings, it's actually easier for me by email 20:23:19 <ukleinek> I think it solves a real problem, so we should consider doing it. My gut feeling would be to only ship updated files to keep the source small 20:23:24 <santiago> gtg now 20:23:25 <santiago> cheers! 20:23:41 <bwh> Longer term, it might be worth doing something like this for non-LTS to reduce the number of different blobs installed, but let's not discuss that now 20:24:39 <carnil> I think we have some ideas, and if bwh can have a look at this on LTS time, then we can now move to the second half of the meeting and go over the bugs (likely not all) 20:24:55 <ukleinek> bwh: I'd expect that if you try it, that you will get a stronger opinion on the actual implementation details 20:25:45 <carnil> do we move to bugs? 20:26:03 * ukleinek nods 20:26:07 <carnil> #topic #1119161: (C, ) sudden memory leak/full wipes out all disk partitions 20:26:54 <bwh> I don't think this is actionable 20:27:01 <carnil> this is a new report; very light on details (and severity might be exagerated). user reports that filesystem got corrupted 20:27:19 <carnil> but there is nothing really to act/look on 20:27:56 <bwh> It doesn't sound like we are going to be able to get any logs from them either 20:28:03 <ukleinek> ..ooOO(win10 is out of support :-) 20:29:05 <bwh> ukleinek: There is extended support in many places! And I'm fairly sure it doesn't self-destruct at end of support (any more often than usual) 20:29:34 * ukleinek is surprised that an OOM problem crashes windows 20:30:04 <carnil> so I think we can write back to the user that there is unfortunately no real actionable information we can help with, and we would close the bug (free to reopen if for instance could provide some previous kernel logs for debugging) 20:30:14 <bwh> OK 20:30:41 <carnil> #action carnil responds to #1119161 that we need more actionable information (logs) to help with and are going to close the bug for the time beeing 20:30:43 <ukleinek> Does a kernel log really help? 20:31:08 <bwh> I suspect not but I don't know what else could help us debug this 20:31:27 <carnil> #topic #1109268: (i, ) ath11k_pci may trigger prompt for nonexistent firmware file during installation 20:31:34 <ukleinek> I think the only thing that would help is getting the HD on one's desk 20:31:45 <iam_tj[m]> We'd need to see the partition table and what signatures each has BEFORE the user tried to recover it 20:31:57 <carnil> this was an action for ukleinek last time, as the hardware works suggested to downgrade to normal/minor 20:32:06 <carnil> ukleinek: still wanting to take care of it? 20:32:23 <bwh> I think I have an action to improve the messages 20:32:24 * ukleinek failed to work on his action items from last week and will catch up 20:32:53 <carnil> #action ukleinek takes care of #1109268 for further investigation (and adjusting bug severity) 20:33:04 <carnil> #topic #1112627: (i, u) linux-image-6.16.3+deb14-amd64: Intel audio no longer works: DMAR: [DMA Write NO_PASID] Request device [00:1b.0] ... non-zero reserved fields in PTE 20:33:41 <carnil> summary: intel_iommu=off helps, audio works in this case, but logs get spammed (this is my understanidng after rereading). Reporter will still try to do a BIOS update. 20:34:14 <carnil> I have no other ideas than suggesting to live with the intel_iommu=off and try after a BIOS update 20:34:32 <ukleinek> sounds reasonable 20:34:37 <carnil> (but this likely means we keep the bug open further?) 20:35:22 <waldi> we can close it. or can we do anything? 20:35:54 <ukleinek> 6.12 is good, is that because the iommu is off there? 20:37:42 <carnil> we have CONFIG_INTEL_IOMMU_DEFAULT_ON_INTGPU_OFF=y as well there already 20:38:15 <bwh> The custom-built 6.12 worked because it had the IOMMU off by default 20:38:32 <bwh> or... no, you are right 20:38:52 <carnil> that is right, but problem is that linux-image-6.12.38+deb13-amd64 was working well for the reporter 20:39:26 <bwh> so, there may be a firmware problem but the driver did avoid it previously 20:39:36 <bwh> Let's send this upstream then 20:40:03 <carnil> bwh: if you have an idea on what to report, would you like to take over the action on this bug? 20:40:04 <ukleinek> let the reporter complete the bisect first? With a consistent .config? 20:40:21 <bwh> #action bwh will forward #1114557 upstream 20:40:31 <bwh> sorry 20:40:47 <bwh> #action bwh will forward #1112627 upstream 20:40:51 <carnil> thank you 20:41:02 <carnil> #topic #1114557: (i, u+) linux-image-6.12.43+deb13-amd64: Jieli touchscreen and stylus no longer supported 20:41:25 <carnil> wait, nothing to do, 4th iteration of patches seem to finalizing. is candiate for stable series as well, so we will pick up the fix later 20:41:32 <ukleinek> no action item apart from keeping it on the radar 20:41:43 <carnil> right 20:41:46 <bwh> OK 20:41:48 <carnil> #topic #1114884: (i, ) linux-image-6.1.0-39-amd64: AMD Ryzen 9 7950X, linux-image-6.1.0-39-amd64 hangs on boot while 6.1.0-37-amd64 works fine 20:42:07 <carnil> this was taken by kathara, ukleinek did you approach her to see if she still wants to take care of it? 20:42:19 * ukleinek blushes 20:42:58 <carnil> oh sorry was not meant this way 20:43:18 <carnil> just asking if we need still to ping kathara. (or she is reading here) 20:43:25 <ukleinek> carnil: you took that wrong, the blame is on my side only 20:43:46 * ukleinek will care for it (discussing the bug with her) 20:44:16 <carnil> #action ukleinek discusses with kathara for #1114884 debugging/triaging 20:44:29 <carnil> #topic #1114912: (i, ) linux-image-amd64: KVM GPU passthrough causes kernel crash and system hang on Debian 13 after VM shutdown 20:44:45 <ukleinek> I think that is another bug that is on my list 20:44:55 <carnil> reporter provided the results for the requested tests. I think it make sense to keep an action here for ukleinek 20:45:03 * ukleinek nods 20:45:15 <carnil> #action ukleinek takes care of further debugging of #1114912 20:45:26 <carnil> #topic #1115581: (i, ) linux-image-6.16.3+deb14-rt-amd64: Boot hang with Creative Sound Blaster AE-7 [1102:0010] 20:45:59 <carnil> so far no reply from the contacts with a sound card with same pci ids. So wait for a reply? 20:46:14 <carnil> or do we have other ideas on what to try next? 20:46:21 <ukleinek> #action ukleinek blushes further, will go through #1115581 and maybe ping again 20:46:53 <carnil> topic #1118437: (i, u) linux-image-6.17.2-amd64: kernel crash in interrupt after receiving an ip packet on veth from xsk from user space 20:47:10 <carnil> there were some iterations of the patch sumission, it has not yet landed in mainline (checked today earlier) 20:47:13 <carnil> -> wait 20:47:21 * ukleinek nods 20:47:27 <carnil> #topic #1118465: (i, u) Regression: linux-image-5.10.0-36-arm64 fails to bring up LAN7800 USB NIC: "NO-CARRIER" and "state DOWN" 20:47:39 <carnil> sad news here, the patch did not work to fix the problem :-/ 20:48:09 <bwh> Right, so I should probably add some more debugging 20:48:29 <bwh> I don't think bisection is needed as there was only one commit that changed the usbnet driver 20:49:12 <bwh> #action bwh will try to debug #1118465 properly instead of guessing 20:49:27 <ukleinek> bwh: you could ask for testing two kernel that just differ by that patch to confirm this is really the culprit, #bisect-light 20:49:36 <bwh> yes 20:49:48 <carnil> thank you bwh, Axel will be happy 20:50:01 <carnil> #topic #1118508: (i, M) linux-image-6.12.38+deb13-amd64: Frequent kernel panic during/just after boot on Lenovo P15v Gen1 with external USB hub in Samsung display 20:50:17 <carnil> still same situation: waiting for problem to trigger again and (hopefully) logs from that. No news yet -> action: wait 20:50:26 <carnil> #topic #1118762: (i, ) linux-image-6.16.12+deb14+1-amd64: external monitors disconnect every several seconds 20:50:29 <carnil> this is new 20:51:27 <carnil> I miss here a kernel log to see if there are warnings, errors, hints 20:51:44 <bwh> Right, we need kernel logs at the least 20:51:46 <ukleinek> + is "every several seconds" a constant amount of time? 20:52:20 * ukleinek doesn't know thunderbold, maybe there is some tracing that could be enabled? 20:53:08 <bwh> I think a fair amount of this stuff is firmware-controlled but probably some of it can be traced in the kernel 20:53:20 <carnil> #action carnil will ask Vincent in #1118762 for kernel (full) kernel logs up to when problem triggers; clarification on "disconnection" times 20:53:23 <bwh> and it is the kernel that regressed... 20:53:25 <carnil> what I additonal wonder is: 20:53:29 <carnil> Kernel 6.7.12 is still fine. 20:53:49 <carnil> is there a typo or did he meant really 6.7.12? 20:54:16 <ukleinek> carnil: why do you doubt? 20:54:22 <carnil> there is a relation to #1072063 which explains that 20:55:09 <bwh> so now both monitors are affected and they also only turn on for a few seconds 20:55:38 <ukleinek> so that might be the same issue that "only" shifted the symptoms 20:55:53 <carnil> so 6.7.12 -> 6.16.9 (only one monitor affected), after updating from 6.16.9 -> 6.16.12 both are affected 20:55:59 <bwh> Yes it's possible that they will get fixed together, but let's not assume that 20:56:19 <carnil> so will keep the above action and ask for complete logs first as a start 20:56:42 <ukleinek> Let them try 6.17.2?! 20:56:58 <carnil> ukleinek: right as well 20:57:30 <carnil> so it is almost 22:00, do you still have energy to go through the remaining bugs? 20:57:48 <waldi> not really 20:57:54 <bwh> We have 1 more important 20:58:20 * ukleinek still has to clean the kitchen, so +1 for not stretching the meeting's end 20:58:35 <carnil> let's do the last one then, and for the rest we have a couple of 'wait" and actions from the previous meeting, so they can remain IMHO assigned 20:58:45 <carnil> so last bug 20:58:58 <carnil> #topic #1119137: (i, ) linux-image-6.12.48+deb13-amd64: CPU stall when Debian Trixie runs as guest in VirtualBox environment 20:59:42 <bwh> Can we reassign this to VirtualBox? 21:00:05 <ukleinek> alternatively let them reproduce with kvm 21:00:37 <carnil> the reporter said: I will attempt to 21:00:38 <carnil> create a Trixie VM with QEmu and see whether that works. 21:00:57 <bwh> Well we know the kernel works fine in KVM 21:00:58 <carnil> so I would say we ping the reporte to ask if that has happened. if it is not reproducible then reassign it to virtualbox 21:01:18 <carnil> correct at least in all cases I ever encountered :) 21:01:31 <carnil> how is the VM sized? (if same or buster works?) 21:01:46 <carnil> bwh: so you would want to straight reassign? 21:02:47 <bwh> Since the reporter said they would also test in QEMU, I suppose it is reasonable to wait for that first 21:03:09 <waldi> yeah 21:03:22 <ukleinek> ack 21:03:23 <bwh> Oh, the kernel log has "[ 0.000000] [Firmware Bug]: TSC doesn't count with P0 frequency!" right at the top, I wonder how important that is 21:03:47 <ukleinek> oh, and does that happen on the host kernel, too? 21:03:51 <waldi> it is quite 21:04:10 <waldi> without counter, all stale checking is unreliable 21:04:55 <bwh> carnil: The VM seems to be ~32 GiB RAM 21:04:58 <ukleinek> is that something that virtual box gets wrong, or is that a hardware issue? 21:05:06 <bwh> I think that would be VB 21:06:16 <ukleinek> so asking for a test with qemu sounds sensible, if that works, reassign to vbox 21:06:35 <bwh> We should probably reply to say we're waiting to hear how that went, yes 21:06:48 <ukleinek> maybe point out that message in the reply and specifically ask if that happens with qemu/kvm, too 21:06:57 <carnil> ok 21:06:58 <bwh> #action bwh will reply to #1119137 to ask about QEMU results, then likely reassign to virtualbox 21:07:21 <carnil> maybe it be worth reducing the amount ov vcpus as well for the vm to test 21:07:25 <carnil> thank you 21:07:33 <carnil> so let's switch to AOB 21:07:41 <carnil> #topic AOB 21:07:56 <bwh> Is there a plan to upload 6.17 to unstable? 21:07:57 <carnil> I guess the only sensible thing to discuss here is who takes care of the next meeting 21:08:07 <waldi> i assembled the list of source packages for iptables/ebtables/arptables legacy, it contains 23 names 21:08:08 <carnil> bwh: I was importing 6.17.6 21:08:18 <bwh> carnil: Grea 21:08:21 <carnil> and if waldi agrees we could upload that to unstable? 21:08:22 <bwh> +t 21:08:45 <waldi> carnil: sure. i can do that or you. and upload 6.18-rc3 or so to experimental 21:09:27 <carnil> bwh: waldi: let me finish the 6.17.6 import, I can do it but I was asking since you did all the work for 6.17 and might want to do the final switch 21:09:59 <waldi> i won't be available next week 21:10:11 <carnil> ok I take care of 6.17.6 -> unstable 21:10:31 <waldi> okay 21:10:32 <bwh> and I will chair next week 21:10:51 <carnil> #action bwh will chair next weeks meeting 21:11:09 <carnil> okay let's stop here, sorry it was more than we could handle this time 21:11:17 <carnil> #endmeeting