| *** m3thos has quit IRC | 00:01 | |
| *** igordc has joined #kata-dev | 00:42 | |
| *** sameo has quit IRC | 00:49 | |
| *** dklyle_ has joined #kata-dev | 01:27 | |
| *** david-lyle has quit IRC | 01:31 | |
| *** david-lyle has joined #kata-dev | 02:12 | |
| *** igordc has quit IRC | 02:15 | |
| *** dklyle_ has quit IRC | 02:15 | |
| *** jodh has joined #kata-dev | 08:24 | |
| *** sameo has joined #kata-dev | 08:39 | |
| *** davidgiluk has joined #kata-dev | 09:01 | |
| *** gwhaley has joined #kata-dev | 09:28 | |
| kata-irc-bot1 | <graham.whaley> @manohar.r.castelino - I thought we had the ubuntu kernel vsock thing understood - that it is only certain host kernel versions that have the problem. Do we still have work to do there? Are we hoping to nail down the exact broke/fixed patches for instance? /cc @james.o.hunt @julio.montes | 09:35 |
|---|---|---|
| kata-irc-bot1 | <james.o.hunt> @graham.whaley - nothing more for us to do - I identified the failing kernel and Canonical have a new kernel in testing fwics. The only problem is that users might try to use vsock with the latest (in other words broken) kernels. We *could* put some sort of heuristic check into the runtime but that is going to be horrid and fragile. | 09:40 |
| *** fdegir has quit IRC | 10:57 | |
| *** fdegir has joined #kata-dev | 10:58 | |
| *** devimc has joined #kata-dev | 13:35 | |
| *** lpetrut has joined #kata-dev | 13:45 | |
| *** jodh has quit IRC | 14:23 | |
| *** jodh has joined #kata-dev | 14:27 | |
| *** irclogbot_0 has joined #kata-dev | 14:29 | |
| *** devimc has quit IRC | 14:52 | |
| *** devimc has joined #kata-dev | 14:53 | |
| *** david-lyle has quit IRC | 14:57 | |
| *** tmhoang has joined #kata-dev | 15:21 | |
| kata-irc-bot1 | <eric.ernst> @julio.montes around? | 16:39 |
| kata-irc-bot1 | <eric.ernst> @julio.montes - I am pretty concnered about the vsock kernel bug. I think we should look @ masking the CID to only be u32. because I don't think the "real" kernel fix will be available for a few weeks. WDYT? Withtout this, anythinng with vsock=true will fail (all Firecracker, a good amount of QEMU configs of Kata) | 16:43 |
| kata-irc-bot1 | <eric.ernst> It'll be a few weeks before it gets fixed on distros, AFAICT -- I think we should make this work around. | 16:44 |
| *** tmhoang has quit IRC | 16:44 | |
| kata-irc-bot1 | <eric.ernst> And if the workaround has to be 'smart' to handle s390? | 16:44 |
| devimc | sounds good, the problem is s390x | 16:44 |
| kata-irc-bot1 | <eric.ernst> I didn't grock what's unique for s390x - can you help me understand? | 16:47 |
| devimc | https://github.com/kata-containers/runtime/pull/960#pullrequestreview-180294614 | 16:49 |
| kata-irc-bot1 | <eric.ernst> right; can't we keep it u64, but mask out upper 32b? Maybe I need to see what is in master now. | 16:50 |
| kata-irc-bot1 | <eric.ernst> Yes, that's already what we do -- I am not sure if we can really work around on our side, then, after taking a closer look @ https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7fbe078c37aba3088359c9256c1a1d0c3e39ee81 | 16:54 |
| kata-irc-bot1 | <eric.ernst> Unless Stefan has a suggestion. | 16:54 |
| *** sameo has quit IRC | 16:56 | |
| *** dklyle has joined #kata-dev | 17:29 | |
| davidgiluk | eric.ernst: Is the s390x problem purely due to endianness? | 17:42 |
| gwhaley | @devimc: ^^ | 18:05 |
| *** jodh has quit IRC | 18:08 | |
| devimc | davidgiluk, yes | 18:09 |
| devimc | davidgiluk, https://github.com/kata-containers/runtime/issues/947 | 18:09 |
| *** sameo has joined #kata-dev | 18:09 | |
| davidgiluk | devimc: Yeh endianness and int sizes always make a fun combination | 18:10 |
| devimc | yes | 18:11 |
| *** gwhaley has quit IRC | 18:16 | |
| *** igordc has joined #kata-dev | 19:07 | |
| *** lpetrut has quit IRC | 19:40 | |
| *** davidgiluk has quit IRC | 20:08 | |
| *** sameo has quit IRC | 20:40 | |
| *** m3thos has joined #kata-dev | 21:11 | |
| *** igordc has quit IRC | 22:04 | |
| *** devimc has quit IRC | 22:50 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!