| kata-irc-bot | <liuj97> We are going to contribute a PCI passthrough implementation for dragonball, which has been in use for a while in Alibaba. | 01:34 |
|---|---|---|
| kata-irc-bot | <liuj97> But for pci-ivshmem, we haven't worked on that yet. What's the specific use case for it? | 01:34 |
| kata-irc-bot | <liuj97> For qemu/firecracker/clh, we need help from community to enable them:$ | 01:35 |
| kata-irc-bot | <liuj97> We are going to work on pci once the runtime-rs gets merged into the main branch | 01:38 |
| kata-irc-bot | <zkaiser> @liuj97 Can you keep me in the loop? I may have some requirements based on the current work I am doing on enabling GPUDirect RDMA with Kata. The current implementation has some deficiencies that I would like to prevent happening in the new implementation. Thanks! | 08:29 |
| kata-irc-bot | <zkaiser> @liuj97 ivshmem use-case is for several services that we're trying to enable in Kata are using shmem communication, we do not have any specific performance requirements, so the latency and bandwidth are negligible | 08:30 |
| kata-irc-bot | <zkaiser> QEMU and Acrn have support for ivshmem, I would say start there, test and maybe add to Dragonball. | 08:31 |
| kata-irc-bot | <zkaiser> As discussed: https://github.com/kata-containers/kata-containers/pull/4492 | 15:34 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!