| *** igordc has joined #kata-dev | 00:44 | |
| *** changcheng has joined #kata-dev | 02:16 | |
| *** igordc has quit IRC | 03:18 | |
| *** igordc has joined #kata-dev | 03:18 | |
| *** pcaruana has joined #kata-dev | 06:30 | |
| *** igordc has quit IRC | 06:53 | |
| *** tmhoang1 has joined #kata-dev | 06:59 | |
| *** sgarzare has joined #kata-dev | 07:08 | |
| *** sameo has joined #kata-dev | 07:29 | |
| *** jodh has joined #kata-dev | 07:30 | |
| *** davidgiluk has joined #kata-dev | 08:02 | |
| *** lpetrut has joined #kata-dev | 08:49 | |
| *** pcaruana has quit IRC | 10:32 | |
| *** pcaruana has joined #kata-dev | 10:58 | |
| *** pcaruana has quit IRC | 12:05 | |
| *** tmhoang1 has quit IRC | 12:07 | |
| *** devimc has joined #kata-dev | 12:17 | |
| *** pcaruana has joined #kata-dev | 12:19 | |
| *** fuentess has joined #kata-dev | 12:27 | |
| *** tobberydberg_ is now known as tobberydberg | 13:01 | |
| *** tmhoang1 has joined #kata-dev | 13:49 | |
| davidgiluk | does someone know where the kata alpha builds with virtiofs are - I'd like to try debugging the existing staticly built virtiofsd | 15:36 |
|---|---|---|
| kata-irc-bot1 | <salvador.fuentes> hi davidgiluk, you can find the assets here: https://github.com/kata-containers/runtime/releases/tag/1.9.0-alpha2 | 15:38 |
| kata-irc-bot1 | <salvador.fuentes> I just needed to apply runtime PR 1994 and rebuild kata-runtime | 15:39 |
| davidgiluk | salvador.fuentes: Ah excellent, I want to see if I can see what's going on with your static build | 15:39 |
| kata-irc-bot1 | <salvador.fuentes> btw, I've just tested the patch you provided on the PR and still cannot make virtiofsd work | 15:39 |
| kata-irc-bot1 | <salvador.fuentes> thanks davidgiluk | 15:40 |
| davidgiluk | salvador.fuentes: OK, can you pass -d to virtiofsd to see if it'll spit out any more debug? Also, which kernel are you on? | 15:40 |
| kata-irc-bot1 | <salvador.fuentes> sure, I'll try with -d | 15:41 |
| kata-irc-bot1 | <salvador.fuentes> my kernel is 5.0.0-1018 | 15:41 |
| davidgiluk | ok, so that's a nice new one, not the same problem Naresh had, which was an older crusty kernel | 15:46 |
| fuentess | davidgiluk: I've tested virtiofsd with -d and still no more information, updated the PR with this comment as well | 15:57 |
| *** pcaruana has quit IRC | 15:57 | |
| *** eernst has joined #kata-dev | 15:58 | |
| davidgiluk | fuentess: Hmm I wonder where the output goes, with -d it should be quite chatty | 15:58 |
| *** eernst has quit IRC | 15:59 | |
| fuentess | davidgiluk: this is the command being run: /opt/kata/bin/virtiofsd --fd=3 -o source=/run/kata-containers/shared/sandboxes/14b49f4045a185eaaf84b3aaff63ff2f0fa683874c4dfbb61a75a9e3822c54e5 -o cache=always --syslog -d -f | 15:59 |
| davidgiluk | you'd hope it's all piled into syslog | 16:02 |
| kata-irc-bot1 | <claire> Hey y’all! If you’re interested in helping to create a Kata user survey - @allison @jimmy and I are available for a kickoff discussion about it on October 9th at 10:00am PT. If that works for you, please let us know this week and we’ll get it booked. Thanks! | 16:07 |
| *** tmhoang1 has quit IRC | 16:13 | |
| *** igordc has joined #kata-dev | 16:14 | |
| davidgiluk | fuentess: Oh that audit message makes more sense! | 16:19 |
| davidgiluk | fuentess: Whenever seccomp gets invoked it just nukes the process completely before it gets a chance to log anything | 16:34 |
| kata-irc-bot1 | <eric.ernst> seccomp is the worst :) | 16:35 |
| kata-irc-bot1 | <eric.ernst> (or the best) | 16:35 |
| kata-irc-bot1 | <eric.ernst> Seems profile isn't appropriate then? | 16:35 |
| davidgiluk | eric.ernst: So, hmm who is setting up the seccomp at the point virtiofsd is run - before it sets up it's own? | 16:36 |
| kata-irc-bot1 | <eric.ernst> I think it is its own -- we don't apply any. | 16:36 |
| davidgiluk | hmm and my reading is we don't set it up until after we've done the listen() call though | 16:38 |
| kata-irc-bot1 | <eric.ernst> should be a simple enough test to just remove the seccomp handling in virtiofsd to see? | 16:39 |
| davidgiluk | yeh I just suggested in the bug to add the listen() call to the seccomp list | 16:40 |
| *** igordc has quit IRC | 16:43 | |
| davidgiluk | fuentess: Note that the static tar has a /opt/kata/bin/virtiofsd and an /opt/kata/bin/virtiofsd-x86_64 which is ~4 months older | 16:53 |
| *** lpetrut has quit IRC | 16:53 | |
| fuentess | davidgiluk: thanks for the suggestion, I'm trying it now | 16:53 |
| fuentess | davidgiluk: yeah, I think the old one is the one that works with the old implementation, which uses nemu in our case | 16:54 |
| davidgiluk | fuentess: I'm going to be surprised if it fixes it, because in theory the listen we do is before we invoke seccomp; unless there's another listen somewhere, and then ...why? | 16:54 |
| davidgiluk | ok, replicated it here | 17:00 |
| *** jodh has quit IRC | 17:00 | |
| *** dklyle has quit IRC | 17:01 | |
| *** dklyle has joined #kata-dev | 17:01 | |
| davidgiluk | that makes no sense | 17:11 |
| davidgiluk | I see a coredump of virtio_loop->fuse_info->__vsyslog_chk->time and that's where it gets the sigsys of apparently a listen call - wth would you do a listen in time() | 17:12 |
| davidgiluk | fuentess: OK, I think I see it | 17:14 |
| davidgiluk | fuentess: on x86 201 is actually _NR_time - not listen | 17:14 |
| fuentess | davidgiluk: ooh, ok, great. so the listen change would work? | 17:15 |
| fuentess | still building it with the change here... | 17:15 |
| davidgiluk | fuentess: Cancel that, the listen change wont help, it needs to be time() not listen() | 17:15 |
| davidgiluk | fuentess: Now, the next question is why your static build hits it | 17:16 |
| fuentess | davidgiluk: ok, cancelling and tryint with time | 17:16 |
| fuentess | yeah, only happens on static build | 17:16 |
| davidgiluk | fuentess: I think the dynamic version gets clever and uses something from the vdso to do time() calls, the static version is....static | 17:18 |
| *** igordc has joined #kata-dev | 17:28 | |
| fuentess | davidgiluk: yeah, that fixes the issue | 17:31 |
| fuentess | I can launch qemu successfully with virtiofsd | 17:31 |
| fuentess | thanks for the help | 17:31 |
| davidgiluk | wahey! | 17:32 |
| fuentess | davidgiluk: once you add the patch, can you please share it, so that we can grab it and apply it when builiding it? | 17:34 |
| davidgiluk | sure, will do | 17:34 |
| fuentess | davidgiluk: thank you :) | 17:35 |
| davidgiluk | you want a patch against 0.3? | 17:37 |
| *** sgarzare has quit IRC | 17:38 | |
| fuentess | davidgiluk: that would be great | 17:38 |
| davidgiluk | seccomp lists are quite tricky, because you never quite know what another version of a library you link against is going to use | 17:47 |
| *** pcaruana has joined #kata-dev | 18:04 | |
| *** sameo has quit IRC | 18:04 | |
| *** igordc has quit IRC | 18:29 | |
| *** igordc has joined #kata-dev | 18:32 | |
| *** pcaruana has quit IRC | 18:36 | |
| *** eernst has joined #kata-dev | 18:46 | |
| *** eernst has quit IRC | 18:51 | |
| *** eernst_ has joined #kata-dev | 18:51 | |
| *** eernst_ has quit IRC | 18:51 | |
| *** eernst has joined #kata-dev | 18:51 | |
| *** davidgiluk has quit IRC | 19:12 | |
| *** lpetrut has joined #kata-dev | 19:47 | |
| *** lpetrut has quit IRC | 19:58 | |
| kata-irc-bot1 | <kshitij> Hi I have a question regarding ARP replies on virtual endpoints. I have an interface mgmt0 from which I ping another interface. I can see in “ifconfig” Tx and Rx stats going up . On the host in the bridge I can see the MAC addresses being learnt and reply packets reaching mgmt0. However, I don’t see ARP resolved in the container. Am I missing something? I have a custom build initrd (added Priority field to the Route struct) | 20:34 |
| *** devimc has quit IRC | 21:16 | |
| *** fuentess has quit IRC | 21:37 | |
| *** sameo has joined #kata-dev | 22:10 | |
| *** sameo has quit IRC | 22:21 | |
| *** sameo has joined #kata-dev | 22:32 | |
| *** eernst has quit IRC | 23:01 | |
| *** eernst has joined #kata-dev | 23:04 | |
| *** eernst has quit IRC | 23:08 | |
| *** sameo has quit IRC | 23:48 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!