*** monstr <monstr!~monstr@nat-35.starnet.cz> has joined #cip | 06:53 | |
*** rajm <rajm!~robert@82.27.50.32> has joined #cip | 06:54 | |
*** frieder <frieder!~frieder@i5C75E691.versanet.de> has joined #cip | 07:25 | |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has joined #cip | 08:45 | |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has quit IRC (Ping timeout: 250 seconds) | 10:42 | |
*** prabhakar <prabhakar!~prabhakar@217.163.141.2> has quit IRC (Ping timeout: 252 seconds) | 10:42 | |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has joined #cip | 12:06 | |
*** masami <masami!~masami@FL1-219-107-72-235.tky.mesh.ad.jp> has joined #cip | 12:45 | |
*** monstr <monstr!~monstr@nat-35.starnet.cz> has quit IRC (Read error: Connection reset by peer) | 12:55 | |
*** monstr <monstr!~monstr@nat-35.starnet.cz> has joined #cip | 12:55 | |
iwamatsu | hello | 13:00 |
---|---|---|
masami | hello | 13:01 |
arisut | hello | 13:01 |
*** jki <jki!~jki@46.128.89.130> has joined #cip | 13:02 | |
iwamatsu | Hi Jan, | 13:03 |
jki | hi! | 13:03 |
jki | meeting started already - or do I have to? | 13:03 |
patersonc | Hello | 13:03 |
iwamatsu | not start yet. | 13:04 |
jki | ah, ok | 13:04 |
jki | then let me try to handle that | 13:04 |
jki | #startmeeting CIP IRC weekly meeting | 13:04 |
collab-meetbot | Meeting started Thu Nov 30 13:04:50 2023 UTC and is due to finish in 60 minutes. The chair is jki. Information about MeetBot at http://wiki.debian.org/MeetBot. | 13:04 |
collab-meetbot | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 13:04 |
collab-meetbot | The meeting name has been set to 'cip_irc_weekly_meeting' | 13:04 |
*** collab-meetbot changes topic to " (Meeting topic: CIP IRC weekly meeting)" | 13:04 | |
iwamatsu | I am going to be in charge today | 13:04 |
jki | oh, sorry, I was too fast then :) | 13:05 |
jki | #topic AI review | 13:05 |
*** collab-meetbot changes topic to "AI review (Meeting topic: CIP IRC weekly meeting)" | 13:05 | |
iwamatsu | no problem, Can I take it? | 13:05 |
jki | sure - go ahead! | 13:05 |
pave1 | Hello! | 13:05 |
iwamatsu | ok | 13:05 |
iwamatsu | hi pave1. | 13:05 |
iwamatsu | Nothing on the list | 13:06 |
iwamatsu | moving on | 13:06 |
iwamatsu | 5 | 13:06 |
iwamatsu | 4 | 13:06 |
iwamatsu | 3 | 13:06 |
iwamatsu | 2 | 13:06 |
iwamatsu | 1 | 13:06 |
iwamatsu | #topic Kernel maintenance updates | 13:06 |
masami | This week reported 1 new CVEs and 6 updated CVEs. | 13:07 |
pave1 | I did reviews, 6.1.63 and .64 | 13:07 |
jki | #topic Kernel maintenance updates | 13:07 |
*** collab-meetbot changes topic to "Kernel maintenance updates (Meeting topic: CIP IRC weekly meeting)" | 13:07 | |
jki | (seems I can only change the topic now) | 13:07 |
iwamatsu | jki: thank you | 13:07 |
iwamatsu | I reviewed 6.1.64. | 13:10 |
iwamatsu | Anything else? | 13:10 |
iwamatsu | 3 | 13:10 |
iwamatsu | 2 | 13:10 |
iwamatsu | 1 | 13:10 |
iwamatsu | #topic Kernel release status | 13:10 |
iwamatsu | 4.4 | 13:10 |
pave1 | 4.4, 4.19, 5.10: -rt is on schedule. | 13:10 |
iwamatsu | 4.19 | 13:11 |
iwamatsu | on track | 13:11 |
iwamatsu | 5.10 | 13:12 |
iwamatsu | on trak | 13:12 |
iwamatsu | 6.10 | 13:12 |
iwamatsu | -cip and cip-rt is late. | 13:13 |
iwamatsu | I am going to release -cip after this meeting. | 13:13 |
pave1 | 6.1-rt: we don't have suitable rt release. I tried to ping maintainers for one. | 13:13 |
iwamatsu | ok, thanks. | 13:14 |
iwamatsu | Anything else? | 13:14 |
iwamatsu | 5 | 13:14 |
iwamatsu | 4 | 13:14 |
iwamatsu | 3 | 13:14 |
iwamatsu | 2 | 13:14 |
iwamatsu | 1 | 13:14 |
iwamatsu | #topic Kernel testing | 13:14 |
patersonc | I've not much to share this week | 13:15 |
iwamatsu | Got it. | 13:15 |
arisut | nothing from me | 13:15 |
iwamatsu | Anything else? | 13:15 |
iwamatsu | 5 | 13:16 |
iwamatsu | 4 | 13:16 |
iwamatsu | 3 | 13:16 |
iwamatsu | 2 | 13:16 |
iwamatsu | 1 | 13:16 |
iwamatsu | #topic AOB | 13:16 |
iwamatsu | F2F meeting before OSSJ. | 13:17 |
pave1 | Will there be some kind of remote bridge? | 13:17 |
patersonc | At the E-TSC next week, I'd like to re-visit the RZ/Five reference platform proposal - is that okay? | 13:18 |
jki | pavel: I assume so | 13:18 |
jki | patersonc: are the linker problems resolved by now? | 13:19 |
patersonc | jkl: This is one of the sticking points. It would require the filesystem to be built specifically for RZ/Five, which doesn't really work with Debian packages | 13:20 |
jki | yes, that will be challenging | 13:20 |
iwamatsu | Other topic? | 13:21 |
jki | who will run the meeting next week? | 13:21 |
jki | I'm on a plane, may or may not be online... | 13:22 |
patersonc | I won't be around next week | 13:22 |
iwamatsu | I can take over | 13:22 |
pave1 | thank you! | 13:23 |
iwamatsu | Anything else? | 13:23 |
iwamatsu | 5 | 13:23 |
iwamatsu | 4 | 13:23 |
iwamatsu | 3 | 13:23 |
iwamatsu | 2 | 13:23 |
iwamatsu | 1 | 13:23 |
iwamatsu | #endmeeting | 13:23 |
jki | thanks, folks! | 13:24 |
iwamatsu | jki: please re-type "#endmeeting" | 13:24 |
jki | #endmeeting | 13:24 |
collab-meetbot | Meeting ended Thu Nov 30 13:24:31 2023 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 13:24 |
collab-meetbot | Minutes: http://ircbot.wl.linuxfoundation.org/meetings/cip/2023/11/cip.2023-11-30-13.04.html | 13:24 |
collab-meetbot | Minutes (text): http://ircbot.wl.linuxfoundation.org/meetings/cip/2023/11/cip.2023-11-30-13.04.txt | 13:24 |
collab-meetbot | Log: http://ircbot.wl.linuxfoundation.org/meetings/cip/2023/11/cip.2023-11-30-13.04.log.html | 13:24 |
*** collab-meetbot changes topic to "Civil Infrastructure Platform Project. CIP mailing list at https://lists.cip-project.org/g/cip-dev | CIP kernel meeting every Thursday at 12:00 UTC | Find the meeting logs at https://ircbot.wl.linuxfoundation.org/meetings/cip/ and chat logs at https://ircbot.wl.linuxfoundation.org/logs/%23cip/" | 13:24 | |
jki | :) | 13:24 |
pave1 | Thank you, see you next week. | 13:24 |
patersonc | jkl: On the cip-core topic... | 13:24 |
jki | see (some of you) in Japan! | 13:24 |
masami | thank you | 13:24 |
iwamatsu | see you next week | 13:24 |
arisut | see you | 13:24 |
*** masami <masami!~masami@FL1-219-107-72-235.tky.mesh.ad.jp> has quit IRC (Quit: Leaving) | 13:24 | |
patersonc | Do you think there's an option to not support a reference platform in cip-core Or is the mandatory for a CIP reference platform? | 13:25 |
jki | patersonc: yes? | 13:25 |
jki | I don't think we have such a rule somewhere, it's more a practical concern | 13:25 |
jki | how many packages would you have to patch in practice? | 13:25 |
jki | we only saw issues with some - or do all have this in principle? | 13:26 |
jki | ...with some that did no dynamic linking IIRC | 13:26 |
patersonc | I think in theory everything should be compiled differently, but in practice we've only seen issues with a few | 13:26 |
jki | reminds me of our "fun" with that Quark processor... | 13:28 |
jki | we need to have a picture what it will practically mean for testing the kernel | 13:28 |
patersonc | Sure - I'm putting something together | 13:29 |
jki | also, as there is not much activitiy on the tiny profile anymore, userspace would remaing out of scope for CIP on that board | 13:29 |
patersonc | Yes that's a shame | 13:29 |
jki | if we are expanding our test images via isar-cip-core, coverage for the RZ/Five could suffer, even kernel-wise | 13:30 |
patersonc | Deby could have been an option because it compiles the packages from source, so we could have added our flag | 13:30 |
patersonc | On the kernel side of things the only "issue" is a large number of core riscv patches that would need backporting | 13:30 |
jki | ...or if our test infrastructure would need special care only for that board | 13:31 |
pave1 | Could we get a summary of the bug and the workaround, somewhere? | 13:33 |
pave1 | I believe we really should aim for workaround at hw or kernel level... | 13:33 |
pave1 | Because if userland has to care, that is no longer riscv architecture, that is something slightly different. | 13:34 |
pave1 | ...and will cause problems. For example security will be very different. | 13:35 |
patersonc | Sure, and we've fixed the kernel side as much as can be done without redesigning the hardware | 13:36 |
pave1 | So what is the end result? | 13:37 |
pave1 | Part of virtual space of each process is unusable. | 13:38 |
pave1 | ... is unusable for normal mappings? | 13:38 |
jki | + is always accessable by userspace, no? | 13:39 |
jki | is this security-wise fixable at all in software? | 13:40 |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has quit IRC (Quit: Client closed) | 13:44 | |
patersonc | Access to naughty areas is blocked by the memory protection unit | 13:45 |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has joined #cip | 13:45 | |
*** prabhakar <prabhakar!~prabhakar@217.163.141.2> has joined #cip | 13:46 | |
pave1 | Aha, ok so we can't run normal binaries... | 13:47 |
pave1 | ...because they may have fixed mappings at that areas? | 13:47 |
pave1 | But security is not completely broken... | 13:47 |
patersonc | Correct | 13:47 |
pave1 | ...because we can use MPU to prevent such accesses. | 13:47 |
patersonc | The TEXT_START_ADDR needs to be set from 0x10000 to 0x50000 | 13:48 |
pave1 | I guess using MPU faults to emulate such accesses (ala math fpu emulation for 386) is not feasible? | 13:50 |
prabhakar | pave1: basically on RISC-V we have PMP regions (16 on rz/five) available on memory protection unit (MPU). This allows users to configure regions to give permission for R/W/X in M/S/U modes. | 13:55 |
pave1 | Prabhakar: Thanks. Is there description of MPU unit somewhere? | 13:57 |
prabhakar | so currently for rz/five we only allow M mode to access the ILM/DLM regions any access tried from S/U (kernel/user space) will cause a panic/fault. | 13:57 |
prabhakarlad | let me check. | 13:58 |
patersonc | https://www.renesas.com/eu/en/products/microcontrollers-microprocessors/rz-mpus/rzfive-general-purpose-microprocessors-risc-v-cpu-core-andes-ax45mp-single-10-ghz-2ch-gigabit-ethernet | 13:58 |
patersonc | "Download Manual HW" - blue button at the top | 14:00 |
prabhakarlad | https://www.andestech.com/wp-content/uploads/AX45MP-1C-Rev.-5.0.0-Datasheet.pdf | 14:01 |
pave1 | prabhakarlad: Thank you! | 14:01 |
prabhakarlad | page 263, i termed it as MPU as its used commonly in arm but in risc-v we its called "Physical Memory Protection Unit" | 14:02 |
pave1 | patersonc: I went through those, but... if it is there it is hidden in all the hardware stuff. | 14:02 |
prabhakarlad | Opensbi patch which does this change https://github.com/riscv-software-src/opensbi/commit/dea0922f867f3d681ad3191fb562a082ea4a339f which is already accepted upstream. | 14:06 |
pave1 | Ok, this is rather small and elegant. | 14:08 |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has quit IRC (Quit: Client closed) | 14:08 | |
pave1 | But as a result board can no longer run normal risc-v binaries, which is a problem. | 14:09 |
pave1 | Proper solution would to be new architecture -- rv64-nolowmem. | 14:09 |
pave1 | Or convince everyone that memory below 0x5_0000 should be reserved on all rv64 linux platforms. | 14:10 |
patersonc | That may be a hard sell | 14:11 |
jki | those faults will be taken to M mode, right? any chance to replay them from there as regular page faults to S mode? | 14:11 |
pave1 | jki: normal page fault handler will map the area and retry execution. | 14:12 |
pave1 | jki: That can't be done here; | 14:12 |
pave1 | jki: we'd have to emulate the instruction. | 14:12 |
jki | so the kernel gets at least the information about the access, and we "only" need to emulate? | 14:14 |
pave1 | jki: Ok, that would work. But performance penalty would be high. | 14:16 |
jki | maybe KVM for RISC-V has instruction emulation support which could be reused here (code-wise, not configuration-wise as there is no hypervisor support on that chip) | 14:16 |
jki | question remains how many apps this will affect in practice | 14:16 |
jki | but, yes, all that remains a de-facto show stopper for distros | 14:17 |
pave1 | jki: yes. :-(. | 14:17 |
pave1 | Ok. I guess we should not advertise this as risc-v platform. | 14:17 |
pave1 | It can not run normal risc-v distribution. | 14:17 |
pave1 | Sorry :-(. | 14:17 |
patersonc | :) | 14:19 |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has joined #cip | 14:20 | |
prabhakar | pave1: for user space we plan to propose similar patch https://paste.debian.net/1299716/ for isar-cip-core. | 14:22 |
jki | that alone does not help | 14:24 |
jki | we need a list of Debian packages that require recompilation against such patched binutils | 14:24 |
pave1 | prabhakar: Patch alone looks good, but... | 14:25 |
jki | in theory, all debian packages could be rebuild via isar[-cip-core], but... no... | 14:25 |
patersonc | Indeed | 14:26 |
patersonc | Perhaps we could generate a test image this way for kernel testing - but not a great setup for "production" cip-core | 14:27 |
jki | even generating that for testing only would be a lot of effort - at least regarding CI time | 14:28 |
patersonc | Sure | 14:29 |
patersonc | Either that or use Renesas' Poky based BSP | 14:29 |
patersonc | Or work out what ASUS have done for their Debian bases Tinker V board | 14:30 |
patersonc | s/bases/based | 14:30 |
patersonc | https://tinker-board.asus.com/series/tinker-v.html | 14:31 |
jki | can you summarize what Asus did? | 14:31 |
patersonc | I don't actually know! | 14:31 |
patersonc | I can't see any docs, but they say they have a Debian based OS | 14:32 |
patersonc | I'll see if I can find out | 14:32 |
*** jki <jki!~jki@46.128.89.130> has quit IRC (Ping timeout: 255 seconds) | 14:37 | |
*** jki <jki!~jki@46.128.89.130> has joined #cip | 14:50 | |
*** jki <jki!~jki@46.128.89.130> has quit IRC (Remote host closed the connection) | 14:58 | |
*** monstr <monstr!~monstr@nat-35.starnet.cz> has quit IRC (Remote host closed the connection) | 17:21 | |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has quit IRC (Quit: Client closed) | 20:02 | |
*** frieder <frieder!~frieder@i5C75E691.versanet.de> has quit IRC (Remote host closed the connection) | 20:16 | |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has joined #cip | 20:40 | |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has quit IRC (Quit: Client closed) | 21:14 | |
*** prabhakar <prabhakar!~prabhakar@217.163.141.2> has quit IRC (Quit: Connection closed) | 21:14 | |
*** prabhakar <prabhakar!~prabhakar@217.163.141.2> has joined #cip | 21:14 | |
*** prabhakarlad <prabhakarlad!~prabhakar@217.163.141.2> has joined #cip | 21:15 | |
*** rajm <rajm!~robert@82.27.50.32> has quit IRC (Ping timeout: 256 seconds) | 22:47 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!