13:01:33 <jki> #startmeeting CIP IRC weekly meeting
13:01:33 <collab-meetbot`> Meeting started Thu Jan  4 13:01:33 2024 UTC and is due to finish in 60 minutes.  The chair is jki. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:01:33 <collab-meetbot`> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
13:01:33 <collab-meetbot`> The meeting name has been set to 'cip_irc_weekly_meeting'
13:01:40 <jki> #topic AI review
13:01:47 <jki> - prepare blog entry on SLTS kernel state and challenges [Jan]
13:01:54 <jki> still on my list...
13:02:10 <jki> I don't have more recorded - anything missing?
13:02:20 <jki> 5
13:02:22 <jki> 4
13:02:23 <jki> 3
13:02:25 <pave1> I think that's all.
13:02:25 <jki> 2
13:02:27 <jki> 1
13:02:29 <jki> #topic Kernel maintenance updates
13:02:47 <masami> This week reported 4 new CVEs and 2 updated CVEs.
13:02:52 <pave1> I did reviews, 6.1.70 and 71.
13:02:56 <uli> reviewing 6.1.70
13:03:13 <pave1> masami: I see you now track 4.4-st. Nice, thanks!
13:03:34 <masami> pavel: your welcome :)
13:04:57 <jki> anything else?
13:05:47 <jki> no corrupted filesystems or broken wifi so far in this year? ;)
13:06:12 <jki> 5
13:06:14 <jki> 4
13:06:15 <jki> 3
13:06:16 <pave1> Just broken hw, but saving that for "aob" section :-)
13:06:20 <jki> 2
13:06:22 <jki> 1
13:06:23 <uli> raid1 in 4.4 fixed last year, but only released this year; does that count?
13:06:39 <jki> all good :)
13:06:43 <jki> then let's move on
13:06:48 <jki> #topic Kernel release status
13:07:13 <jki> nobuhiro's script should go here
13:07:28 <jki> 4.4
13:07:36 <pave1> basically we are ok. Continue? :-)
13:07:52 <uli> i guess. :) cip83 is out
13:08:39 <jki> yeah, I think I only saw no update to 5.10-rt yesterday
13:08:57 <jki> so, 4.4, 4.19, 6.1 - all recently updated
13:09:07 <jki> 5.10 vanilla as well
13:09:20 <pave1> I believe 5.10-rt is due next month?
13:10:09 <jki> end of this month
13:10:12 <jki> ok
13:10:26 <jki> no problems ahead as well, I assume
13:10:31 <jki> then moving on...
13:10:39 <jki> 3
13:10:40 <jki> 2
13:10:42 <jki> 1
13:10:44 <jki> #topic Kernel testing
13:10:58 <patersonc> Hello
13:11:34 <patersonc> I've spent a bit of time looking into some of the SW update use cases in LAVA, but other then that I've mainly been on holiday feasting
13:12:21 <jki> there were a lot of discussions on the related issue, indeed
13:13:45 <patersonc> On the positive side, all of the labs stayed online throughout the holiday period
13:14:09 <jki> that's kind of them
13:14:25 <jki> other testing topics?
13:14:29 <pave1> You may want to check 6.6-stable results
13:14:50 <pave1> https://gitlab.com/cip-project/cip-testing/linux-stable-rc-ci/-/pipelines/1125990082
13:15:03 <pave1> de0nanosoc always fails.
13:15:15 <pave1> iirc it is not a fluke.
13:15:16 <patersonc> Thanks pave1, will do
13:15:22 <pave1> (y)
13:16:49 <jki> is that board in our lab?
13:17:38 <patersonc> You have one
13:17:49 <patersonc> Other two are in the denx lab
13:18:06 <jki> so the issue is not lab-specific, rather board-related
13:19:21 <patersonc> More than likely, but we'll have to investigate
13:19:27 <jki> looks like the siemens config passes
13:19:28 <jki> ok
13:19:46 <jki> then move on...?
13:19:52 <jki> 5
13:19:53 <jki> 4
13:19:55 <jki> 3
13:19:57 <jki> 2
13:19:59 <jki> 1
13:20:01 <jki> #topic AOB
13:20:11 <pave1> Ok, so buggy hardware.
13:20:26 <patersonc> jkl: Yes, it's multi_v7_defconfig that's failing
13:20:43 <patersonc> So maybe a kernel configuration issue
13:21:00 <pave1> After the kernel patch discussion about the Renesas R-V board, I don't believe we should make that official "cip board".
13:21:24 <pave1> OTOH the kernel patches seem quite reasonable.
13:21:46 <pave1> I see no problem taking them.
13:21:54 <pave1> But the problems in userland really seem bad.
13:22:24 <jki> we have up to option 3 in Chris' list at least (userland from other sources, just for basic testing)
13:22:27 <patersonc> The emulation approach that jkl suggested could work, but would be a lot of work to get running
13:22:44 <jki> yes, I'm afraid of that as well
13:23:04 <patersonc> I don't know if it's a route RISC-V is planning in general or not
13:23:12 <jki> but you still wanted to check if the rebuild impact could be limited to certain packages
13:23:40 <patersonc> Yes, prabhakarlad is looking into that
13:23:42 <jki> no, this is no general risc-v topic, it remains specific to your IP
13:24:17 <jki> that is the reason why there will be no general userland solution
13:24:50 <pave1> One small chance would be...
13:25:17 <pave1> ...if it was discovered that starting at 0x50000 is a bad idea because it is not on 2MB (or whatever) boundary
13:25:31 <pave1> and it impacts performance in general due to increased TLB pressure.
13:25:56 <jki> ...that should have been discovered earlier, no?
13:26:32 <pave1> Dunno. putting code this low seems strange.
13:26:57 <pave1> Or maybe that putting code this low is bad for security because bigger guard area is needed.
13:27:51 <jki> well, the binutils patch was reject upstream already IIRC
13:28:16 <pave1> Send them a chocolate or something :-).
13:28:23 <patersonc> Ha
13:29:30 <pave1> Dunno. 0x50000 is in the middle of PMD, so it seems like wrong start address. Putting code at start of PUD could be a tiny bit better for TLBs and thus performance.
13:29:44 <pave1> If it is, that could convince someone. And chocolate, I guess :-).
13:30:07 <prabhakarlad> jki: yep binutils patch was rejected. Palmer suggested to handle it in the toolchain.
13:30:34 <pave1> And can it be handled in the toolchain?
13:30:41 <jki> what does "handle it in the toolchain" mean?
13:32:12 <prabhakarlad> https://paste.debian.net/1303035/
13:32:32 <prabhakarlad> let me find the complete link.
13:33:27 <jki> "and we might want that as a tunable for
13:33:27 <jki> distros anyway (for huge page alignment, for example)."
13:33:37 <jki> that's what pavel is talking about
13:33:50 <pave1> Yep :-).
13:33:57 <jki> you need to follow up on that - before the gates are closing for trixie!
13:36:19 <patersonc> So for now, is CIP okay to take the kernel support as-is?
13:36:49 <patersonc> On the assumption that "official" reference platform acceptance is only if a working solution is found for userspace
13:36:50 <pave1> Kernel patches look reasonable.
13:36:54 <jki> because even if we had that tunable, it still takes Debian (or any other distro) to tune that as needed, for whole riscv64
13:39:50 <jki> anything else?
13:40:09 <pave1> So...
13:40:22 <pave1> Is the plan to just apply the kernel patches even when we don't have the userspace?
13:40:57 <jki> provided renesas looks after testing, it should be a way forward
13:41:18 <jki> but they also need to push harder on a userspace solution IMHO
13:41:29 <pave1> Ok, sounds good. So I expect another series, not marked "RFC" when testing is sorted out.
13:42:03 <patersonc> I'll add some support for simple boot testing to start with, using a Poky image
13:42:54 <patersonc> And test the RFC with it
13:43:02 <patersonc> If it works we'll submit a v1
13:44:06 <pave1> Could we also get some qemu risc-v testing?
13:44:19 <pave1> Right now I don't see one.
13:44:31 <pave1> https://gitlab.com/cip-project/cip-testing/linux-stable-rc-ci/-/pipelines/1125990380
13:45:13 <patersonc> Yes this is needed
13:45:27 <pave1> Thank you!
13:45:48 <jki> we should soon have isar-cip-core generating riscv images more reliably again
13:45:54 <jki> if that helps as well
13:46:03 <patersonc> yep
13:47:05 <jki> ok, I can plug something together based on my pending isar patches
13:47:18 <jki> more topics for today?
13:47:41 <jki> 4
13:47:42 <jki> 3
13:47:44 <jki> 2
13:47:46 <jki> 1
13:47:48 <jki> #endmeeting