*** tmhoang has joined #kata-dev | 07:06 | |
*** tmhoang has quit IRC | 07:12 | |
*** sameo has joined #kata-dev | 07:24 | |
*** jodh has joined #kata-dev | 07:41 | |
*** gwhaley has joined #kata-dev | 07:58 | |
*** davidgiluk has joined #kata-dev | 08:03 | |
*** pcaruana has joined #kata-dev | 08:52 | |
davidgiluk | stefanha: While the virtiofsd code to handle hot-add-ram isn't safe, I don't think that's actually the problem that kata is hitting | 08:57 |
---|---|---|
stefanha | davidgiluk: Interesting. Any idea why the address is out of range? | 09:23 |
stefanha | davidgiluk: Unfortunately I've fallen ill again (!). My daughter got sick again last Wednesday and I caught it over the weekend. | 09:24 |
davidgiluk | stefanha: She's got a great career as a microbiologist | 09:24 |
stefanha | davidgiluk: Ah, I see you've updated https://github.com/kata-containers/runtime/issues/1745#issuecomment-502598048 | 09:25 |
stefanha | Excellent, thank you! | 09:25 |
davidgiluk | stefanha: I added a comment to the issue, it's actually an easier case - it's just Kata is using a plain ram object to hot add rather than a ram-from-file/share=on | 09:25 |
stefanha | So it's just a question for fixing up kata-runtime's QMP memory hotplug command | 09:25 |
stefanha | That's good | 09:25 |
davidgiluk | yeh | 09:26 |
davidgiluk | stefanha: The interesting side bug is that means kata has been hotplugging normal RAM rather than hugepages, which I assume is not what kata was expecting to do if there main ram was hugepage | 09:26 |
stefanha | Yep | 09:26 |
*** gwhaley has quit IRC | 11:01 | |
*** gwhaley has joined #kata-dev | 12:07 | |
*** fuentess has joined #kata-dev | 12:55 | |
*** pcaruana has quit IRC | 13:26 | |
*** pcaruana|afk| has joined #kata-dev | 13:26 | |
*** pcaruana has joined #kata-dev | 13:48 | |
*** pcaruana|afk| has quit IRC | 13:51 | |
*** pcaruana has quit IRC | 14:47 | |
*** dklyle has joined #kata-dev | 14:55 | |
*** sameo has quit IRC | 15:31 | |
*** sameo has joined #kata-dev | 15:35 | |
*** pcaruana has joined #kata-dev | 15:40 | |
*** sameo has quit IRC | 15:54 | |
*** sameo has joined #kata-dev | 15:59 | |
*** sameo has quit IRC | 16:05 | |
*** igordc has joined #kata-dev | 16:44 | |
*** gwhaley has quit IRC | 16:58 | |
*** jodh has quit IRC | 17:00 | |
kata-irc-bot | <sebastien.boeuf> davidgiluk: stefanha: still around? | 17:37 |
davidgiluk | I am | 17:38 |
kata-irc-bot | <sebastien.boeuf> cool I have a question for you regarding https://gitlab.com/virtio-fs/linux/blob/virtio-fs/fs/fuse/virtio_fs.c#L525-527 | 17:39 |
* davidgiluk looks | 17:40 | |
kata-irc-bot | <sebastien.boeuf> davidgiluk: I think we should be able to use virtio-fs without the shared memory region, even when CONFIG_DAX_DRIVER is there | 17:40 |
kata-irc-bot | <sebastien.boeuf> if the device provides a cache_size=0, then the driver should not try to use the shared region at all, wdyt? | 17:40 |
davidgiluk | Right, I have patches in the current dev world that allow cache_size=0, and I have a kernel hack but I hadn't cleaned it up yet | 17:41 |
kata-irc-bot | <sebastien.boeuf> I would `return 0` instead of `return -ENXIO` | 17:41 |
kata-irc-bot | <sebastien.boeuf> oh glad to hear it | 17:41 |
davidgiluk | seb: I'd used it to get ARM working | 17:41 |
kata-irc-bot | <sebastien.boeuf> ok gotcha | 17:42 |
kata-irc-bot | <sebastien.boeuf> I need those to make cloud-hypervisor working with virtio-fs | 17:42 |
kata-irc-bot | <sebastien.boeuf> because right now, I didn't implement the shared memory region | 17:42 |
davidgiluk | ah ok | 17:42 |
kata-irc-bot | <sebastien.boeuf> but it's still valid and working if I rely purely on the virtqueues to get the fs sharing | 17:43 |
kata-irc-bot | <sebastien.boeuf> obviously I want to get the shared memory region working for performance at some point | 17:43 |
davidgiluk | yeh that's perfectly valid | 17:43 |
kata-irc-bot | <sebastien.boeuf> ok cool thanks Dave, I'm glad you agree with this use case! | 17:43 |
davidgiluk | I'm just fighting some other bugs but then I'll clean that up | 17:45 |
kata-irc-bot | <sebastien.boeuf> great :thumbsup: | 17:45 |
davidgiluk | gmmaha: Did you see my reply about hotplug on virtiofs ? | 17:47 |
kata-irc-bot | <eric.ernst> @jose.carlos.venegas.m @salvador.fuentes hey ya'll: see https://github.com/kata-containers/agent/pull/581 | 17:55 |
kata-irc-bot | <eric.ernst> PR is fine, but two instances of flakey CI - curious if you have input. I'm restarting the Ubuntu one, but wondering if we should make test more robust of OBS connectivity failure (add retry?) | 17:56 |
gmmaha | davidgiluk: yes sir. I did see it. Will try and patch up kata to turn sharing on when hotplug + virtio-fs is sent. | 18:04 |
gmmaha | does hotplug work with just qemu on your end? | 18:04 |
gmmaha | i know stefanha was going to give it a spin.. was just curious.. | 18:05 |
kata-irc-bot | <salvador.fuentes> @eric.ernst I saw the issues, one was a network timeout when installing deps on centos, the other is a shimv2 failure which I havent seen before, have restarted both of them and lets see if the shimv2 error appears again... | 18:08 |
gmmaha | davidgiluk: seems like i missed the chat history here where stefanha has commented. Will fix that on the kata side of things. | 18:13 |
davidgiluk | gmmaha: When you do let me know; I know the code in virtiofsd needs fixing, but it should get further; but it might explode a bit more spectacularly :-) | 18:25 |
gmmaha | hahah | 18:25 |
gmmaha | will do! | 18:25 |
gmmaha | i have barely 48 hours before i am off for a few weeks.. will try and sneak it in before. | 18:26 |
davidgiluk | yeh things creep up before holidays | 18:28 |
*** gmmaha has quit IRC | 18:28 | |
*** pcaruana has quit IRC | 18:44 | |
*** gmmaha has joined #kata-dev | 18:45 | |
*** pcaruana has joined #kata-dev | 19:17 | |
*** davidgiluk has quit IRC | 19:22 | |
*** pcaruana has quit IRC | 19:27 | |
*** sameo has joined #kata-dev | 21:38 | |
*** fuentess has quit IRC | 21:54 | |
*** ChanServ has quit IRC | 22:09 | |
*** ChanServ has joined #kata-dev | 22:14 | |
*** niven.freenode.net sets mode: +o ChanServ | 22:14 | |
*** sameo has quit IRC | 22:55 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!